* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure
2016-07-20 14:24 13% [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure Tomasz Kulasek
2016-07-20 15:01 4% ` Thomas Monjalon
2016-07-21 15:24 11% ` [dpdk-dev] [PATCH v2] " Tomasz Kulasek
@ 2016-07-31 7:46 4% ` Vlad Zolotarov
2 siblings, 0 replies; 200+ results
From: Vlad Zolotarov @ 2016-07-31 7:46 UTC (permalink / raw)
To: Tomasz Kulasek, dev
On 07/20/2016 05:24 PM, Tomasz Kulasek wrote:
> This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
> changes in rte_eth_dev and rte_eth_desc_lim structures.
>
> In 16.11, we plan to introduce rte_eth_tx_prep() function to do
> necessary preparations of packet burst to be safely transmitted on
> device for desired HW offloads (set/reset checksum field according to
> the hardware requirements) and check HW constraints (number of segments
> per packet, etc).
>
> While the limitations and requirements may differ for devices, it
> requires to extend rte_eth_dev structure with new function pointer
> "tx_pkt_prep" which can be implemented in the driver to prepare and
> verify packets, in devices specific way, before burst, what should to
> prevent application to send malformed packets.
>
> Also new fields will be introduced in rte_eth_desc_lim: nb_seg_max and
> nb_mtu_seg_max, providing an information about max segments in TSO and
> non TSO packets acceptable by device.
>
> Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Acked-by: Vlad Zolotarov <vladz@scylladb.com>
> ---
> doc/guides/rel_notes/deprecation.rst | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index f502f86..485aacb 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -41,3 +41,10 @@ Deprecation Notices
> * The mempool functions for single/multi producer/consumer are deprecated and
> will be removed in 16.11.
> It is replaced by rte_mempool_generic_get/put functions.
> +
> +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
> + extended with new function pointer ``tx_pkt_prep`` allowing verification
> + and processing of packet burst to meet HW specific requirements before
> + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
> + segments limit to be transmitted by device for TSO/non-TSO packets.
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH v2] log: remove history dump
2016-07-29 13:01 9% [dpdk-dev] [PATCH] log: remove history dump Thomas Monjalon
@ 2016-07-29 13:50 9% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-29 13:50 UTC (permalink / raw)
To: dev
The log history feature was deprecated in 16.07.
The remaining empty functions are removed in 16.11.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
v2: fix LIBABIVER and compilation of test
---
app/test/test.h | 5 +++-
doc/guides/rel_notes/deprecation.rst | 3 ---
doc/guides/rel_notes/release_16_11.rst | 4 +++-
lib/librte_eal/bsdapp/eal/Makefile | 2 +-
lib/librte_eal/common/eal_common_log.c | 19 ---------------
lib/librte_eal/common/include/rte_log.h | 41 ---------------------------------
lib/librte_eal/linuxapp/eal/Makefile | 2 +-
7 files changed, 9 insertions(+), 67 deletions(-)
diff --git a/app/test/test.h b/app/test/test.h
index 467b9c0..74d6021 100644
--- a/app/test/test.h
+++ b/app/test/test.h
@@ -33,9 +33,12 @@
#ifndef _TEST_H_
#define _TEST_H_
+
#include <stddef.h>
#include <sys/queue.h>
-#include "rte_log.h"
+
+#include <rte_common.h>
+#include <rte_log.h>
#define TEST_SUCCESS (0)
#define TEST_FAILED (-1)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 8263d03..96db661 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -8,9 +8,6 @@ API and ABI deprecation notices are to be posted here.
Deprecation Notices
-------------------
-* The log history is deprecated.
- It is voided in 16.07 and will be removed in release 16.11.
-
* The ethdev library file will be renamed from libethdev.* to librte_ethdev.*
in release 16.11 in order to have a more consistent namespace.
diff --git a/doc/guides/rel_notes/release_16_11.rst b/doc/guides/rel_notes/release_16_11.rst
index a6e3307..0b9022d 100644
--- a/doc/guides/rel_notes/release_16_11.rst
+++ b/doc/guides/rel_notes/release_16_11.rst
@@ -94,6 +94,8 @@ API Changes
This section is a comment. Make sure to start the actual text at the margin.
+* The log history is removed.
+
ABI Changes
-----------
@@ -131,7 +133,7 @@ The libraries prepended with a plus sign were incremented in this version.
librte_cmdline.so.2
librte_cryptodev.so.1
librte_distributor.so.1
- librte_eal.so.2
+ + librte_eal.so.3
librte_hash.so.2
librte_ip_frag.so.1
librte_ivshmem.so.1
diff --git a/lib/librte_eal/bsdapp/eal/Makefile b/lib/librte_eal/bsdapp/eal/Makefile
index 988cbbc..7a0fea5 100644
--- a/lib/librte_eal/bsdapp/eal/Makefile
+++ b/lib/librte_eal/bsdapp/eal/Makefile
@@ -48,7 +48,7 @@ LDLIBS += -lgcc_s
EXPORT_MAP := rte_eal_version.map
-LIBABIVER := 2
+LIBABIVER := 3
# specific to linuxapp exec-env
SRCS-$(CONFIG_RTE_EXEC_ENV_BSDAPP) := eal.c
diff --git a/lib/librte_eal/common/eal_common_log.c b/lib/librte_eal/common/eal_common_log.c
index 7916c78..967991a 100644
--- a/lib/librte_eal/common/eal_common_log.c
+++ b/lib/librte_eal/common/eal_common_log.c
@@ -64,19 +64,6 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg, log_cur_msg);
/* default logs */
-int
-rte_log_add_in_history(const char *buf __rte_unused, size_t size __rte_unused)
-{
- return 0;
-}
-
-void
-rte_log_set_history(int enable)
-{
- if (enable)
- RTE_LOG(WARNING, EAL, "The log history is deprecated.\n");
-}
-
/* Change the stream that will be used by logging system */
int
rte_openlog_stream(FILE *f)
@@ -131,12 +118,6 @@ int rte_log_cur_msg_logtype(void)
return RTE_PER_LCORE(log_cur_msg).logtype;
}
-/* Dump log history to file */
-void
-rte_log_dump_history(FILE *out __rte_unused)
-{
-}
-
/*
* Generates a log message The message will be sent in the stream
* defined by the previous call to rte_openlog_stream().
diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
index b1add04..919563c 100644
--- a/lib/librte_eal/common/include/rte_log.h
+++ b/lib/librte_eal/common/include/rte_log.h
@@ -42,8 +42,6 @@
* This file provides a log API to RTE applications.
*/
-#include "rte_common.h" /* for __rte_deprecated macro */
-
#ifdef __cplusplus
extern "C" {
#endif
@@ -181,45 +179,6 @@ int rte_log_cur_msg_loglevel(void);
int rte_log_cur_msg_logtype(void);
/**
- * @deprecated
- * Enable or disable the history (enabled by default)
- *
- * @param enable
- * true to enable, or 0 to disable history.
- */
-__rte_deprecated
-void rte_log_set_history(int enable);
-
-/**
- * @deprecated
- * Dump the log history to a file
- *
- * @param f
- * A pointer to a file for output
- */
-__rte_deprecated
-void rte_log_dump_history(FILE *f);
-
-/**
- * @deprecated
- * Add a log message to the history.
- *
- * This function can be called from a user-defined log stream. It adds
- * the given message in the history that can be dumped using
- * rte_log_dump_history().
- *
- * @param buf
- * A data buffer containing the message to be saved in the history.
- * @param size
- * The length of the data buffer.
- * @return
- * - 0: Success.
- * - (-ENOBUFS) if there is no room to store the message.
- */
-__rte_deprecated
-int rte_log_add_in_history(const char *buf, size_t size);
-
-/**
* Generates a log message.
*
* The message will be sent in the stream defined by the previous call
diff --git a/lib/librte_eal/linuxapp/eal/Makefile b/lib/librte_eal/linuxapp/eal/Makefile
index 182729c..3a7631a 100644
--- a/lib/librte_eal/linuxapp/eal/Makefile
+++ b/lib/librte_eal/linuxapp/eal/Makefile
@@ -37,7 +37,7 @@ ARCH_DIR ?= $(RTE_ARCH)
EXPORT_MAP := rte_eal_version.map
VPATH += $(RTE_SDK)/lib/librte_eal/common/arch/$(ARCH_DIR)
-LIBABIVER := 2
+LIBABIVER := 3
VPATH += $(RTE_SDK)/lib/librte_eal/common
--
2.7.0
^ permalink raw reply [relevance 9%]
* [dpdk-dev] [PATCH] doc: postpone mempool ABI breakage
@ 2016-07-29 13:41 8% Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-29 13:41 UTC (permalink / raw)
To: olivier.matz; +Cc: dev
It was planned to remove some mempool functions which are deprecated
since 16.07.
As no other mempool ABI change is planned in 16.11, it is better
to postpone and group every mempool ABI changes in 17.02.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
doc/guides/rel_notes/deprecation.rst | 15 +++++++--------
1 file changed, 7 insertions(+), 8 deletions(-)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 1b953fe..96db661 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -41,15 +41,14 @@ Deprecation Notices
PKT_RX_QINQ_STRIPPED, that are better described. The old flags and
their behavior will be kept in 16.07 and will be removed in 16.11.
-* The APIs rte_mempool_count and rte_mempool_free_count are being deprecated
- on the basis that they are confusing to use - free_count actually returns
- the number of allocated entries, not the number of free entries as expected.
- They are being replaced by rte_mempool_avail_count and
- rte_mempool_in_use_count respectively.
+* mempool: The functions ``rte_mempool_count`` and ``rte_mempool_free_count``
+ will be removed in 17.02.
+ They are replaced by ``rte_mempool_avail_count`` and
+ ``rte_mempool_in_use_count`` respectively.
-* The mempool functions for single/multi producer/consumer are deprecated and
- will be removed in 16.11.
- It is replaced by rte_mempool_generic_get/put functions.
+* mempool: The functions for single/multi producer/consumer are deprecated
+ and will be removed in 17.02.
+ It is replaced by ``rte_mempool_generic_get/put`` functions.
* The ``rte_ivshmem`` feature (including library and EAL code) will be removed
in 16.11 because it has some design issues which are not planned to be fixed.
--
2.7.0
^ permalink raw reply [relevance 8%]
* [dpdk-dev] [PATCH] log: remove history dump
@ 2016-07-29 13:01 9% Thomas Monjalon
2016-07-29 13:50 9% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-29 13:01 UTC (permalink / raw)
To: david.marchand; +Cc: dev
The log history feature was deprecated in 16.07.
The remaining empty functions are removed in 16.11.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
doc/guides/rel_notes/deprecation.rst | 3 ---
doc/guides/rel_notes/release_16_11.rst | 4 +++-
lib/librte_eal/common/eal_common_log.c | 19 ---------------
lib/librte_eal/common/include/rte_log.h | 41 ---------------------------------
4 files changed, 3 insertions(+), 64 deletions(-)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index d2dc4a9..1b953fe 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -8,9 +8,6 @@ API and ABI deprecation notices are to be posted here.
Deprecation Notices
-------------------
-* The log history is deprecated.
- It is voided in 16.07 and will be removed in release 16.11.
-
* The ethdev library file will be renamed from libethdev.* to librte_ethdev.*
in release 16.11 in order to have a more consistent namespace.
diff --git a/doc/guides/rel_notes/release_16_11.rst b/doc/guides/rel_notes/release_16_11.rst
index a6e3307..0b9022d 100644
--- a/doc/guides/rel_notes/release_16_11.rst
+++ b/doc/guides/rel_notes/release_16_11.rst
@@ -94,6 +94,8 @@ API Changes
This section is a comment. Make sure to start the actual text at the margin.
+* The log history is removed.
+
ABI Changes
-----------
@@ -131,7 +133,7 @@ The libraries prepended with a plus sign were incremented in this version.
librte_cmdline.so.2
librte_cryptodev.so.1
librte_distributor.so.1
- librte_eal.so.2
+ + librte_eal.so.3
librte_hash.so.2
librte_ip_frag.so.1
librte_ivshmem.so.1
diff --git a/lib/librte_eal/common/eal_common_log.c b/lib/librte_eal/common/eal_common_log.c
index 7916c78..967991a 100644
--- a/lib/librte_eal/common/eal_common_log.c
+++ b/lib/librte_eal/common/eal_common_log.c
@@ -64,19 +64,6 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg, log_cur_msg);
/* default logs */
-int
-rte_log_add_in_history(const char *buf __rte_unused, size_t size __rte_unused)
-{
- return 0;
-}
-
-void
-rte_log_set_history(int enable)
-{
- if (enable)
- RTE_LOG(WARNING, EAL, "The log history is deprecated.\n");
-}
-
/* Change the stream that will be used by logging system */
int
rte_openlog_stream(FILE *f)
@@ -131,12 +118,6 @@ int rte_log_cur_msg_logtype(void)
return RTE_PER_LCORE(log_cur_msg).logtype;
}
-/* Dump log history to file */
-void
-rte_log_dump_history(FILE *out __rte_unused)
-{
-}
-
/*
* Generates a log message The message will be sent in the stream
* defined by the previous call to rte_openlog_stream().
diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
index b1add04..919563c 100644
--- a/lib/librte_eal/common/include/rte_log.h
+++ b/lib/librte_eal/common/include/rte_log.h
@@ -42,8 +42,6 @@
* This file provides a log API to RTE applications.
*/
-#include "rte_common.h" /* for __rte_deprecated macro */
-
#ifdef __cplusplus
extern "C" {
#endif
@@ -181,45 +179,6 @@ int rte_log_cur_msg_loglevel(void);
int rte_log_cur_msg_logtype(void);
/**
- * @deprecated
- * Enable or disable the history (enabled by default)
- *
- * @param enable
- * true to enable, or 0 to disable history.
- */
-__rte_deprecated
-void rte_log_set_history(int enable);
-
-/**
- * @deprecated
- * Dump the log history to a file
- *
- * @param f
- * A pointer to a file for output
- */
-__rte_deprecated
-void rte_log_dump_history(FILE *f);
-
-/**
- * @deprecated
- * Add a log message to the history.
- *
- * This function can be called from a user-defined log stream. It adds
- * the given message in the history that can be dumped using
- * rte_log_dump_history().
- *
- * @param buf
- * A data buffer containing the message to be saved in the history.
- * @param size
- * The length of the data buffer.
- * @return
- * - 0: Success.
- * - (-ENOBUFS) if there is no room to store the message.
- */
-__rte_deprecated
-int rte_log_add_in_history(const char *buf, size_t size);
-
-/**
* Generates a log message.
*
* The message will be sent in the stream defined by the previous call
--
2.7.0
^ permalink raw reply [relevance 9%]
* [dpdk-dev] [PATCH] ivshmem: remove integration in dpdk
@ 2016-07-29 12:28 1% David Marchand
0 siblings, 0 replies; 200+ results
From: David Marchand @ 2016-07-29 12:28 UTC (permalink / raw)
To: dev; +Cc: thomas.monjalon, anatoly.burakov
Following discussions on the mailing list [1] and since nobody stood up to
implement the necessary cleanups, here is the ivshmem integration removal.
There is not much to say about this patch, a lot of code is being removed.
The default configuration file for packet_ordering example is replaced with
the "native" x86 file.
The only tricky part is in eal_memory with the memseg index stuff.
More cleanups can be done after this but will come in subsequent patchsets.
[1]: http://dpdk.org/ml/archives/dev/2016-June/040844.html
Signed-off-by: David Marchand <david.marchand@6wind.com>
---
MAINTAINERS | 8 -
app/test/Makefile | 1 -
app/test/autotest_data.py | 6 -
app/test/test.c | 3 -
app/test/test.h | 1 -
app/test/test_ivshmem.c | 433 ------------
config/defconfig_arm64-armv8a-linuxapp-gcc | 1 -
config/defconfig_x86_64-ivshmem-linuxapp-gcc | 49 --
config/defconfig_x86_64-ivshmem-linuxapp-icc | 49 --
doc/api/doxy-api-index.md | 1 -
doc/api/doxy-api.conf | 1 -
doc/api/examples.dox | 2 -
doc/guides/linux_gsg/build_dpdk.rst | 2 +-
doc/guides/linux_gsg/quick_start.rst | 14 +-
doc/guides/prog_guide/img/ivshmem.png | Bin 44920 -> 0 bytes
doc/guides/prog_guide/index.rst | 1 -
doc/guides/prog_guide/ivshmem_lib.rst | 160 -----
doc/guides/prog_guide/source_org.rst | 1 -
doc/guides/rel_notes/deprecation.rst | 3 -
doc/guides/rel_notes/release_16_11.rst | 3 +
examples/Makefile | 1 -
examples/l2fwd-ivshmem/Makefile | 43 --
examples/l2fwd-ivshmem/guest/Makefile | 50 --
examples/l2fwd-ivshmem/guest/guest.c | 452 -------------
examples/l2fwd-ivshmem/host/Makefile | 50 --
examples/l2fwd-ivshmem/host/host.c | 895 -------------------------
examples/l2fwd-ivshmem/include/common.h | 111 ----
examples/packet_ordering/Makefile | 2 +-
lib/Makefile | 1 -
lib/librte_eal/common/eal_common_memzone.c | 12 -
lib/librte_eal/common/eal_private.h | 22 -
lib/librte_eal/common/include/rte_memory.h | 3 -
lib/librte_eal/common/include/rte_memzone.h | 7 +-
lib/librte_eal/common/malloc_heap.c | 8 -
lib/librte_eal/linuxapp/eal/Makefile | 9 -
lib/librte_eal/linuxapp/eal/eal.c | 10 -
lib/librte_eal/linuxapp/eal/eal_ivshmem.c | 954 ---------------------------
lib/librte_eal/linuxapp/eal/eal_memory.c | 30 +-
lib/librte_ivshmem/Makefile | 54 --
lib/librte_ivshmem/rte_ivshmem.c | 919 --------------------------
lib/librte_ivshmem/rte_ivshmem.h | 165 -----
lib/librte_ivshmem/rte_ivshmem_version.map | 12 -
mk/rte.app.mk | 1 -
43 files changed, 13 insertions(+), 4537 deletions(-)
delete mode 100644 app/test/test_ivshmem.c
delete mode 100644 config/defconfig_x86_64-ivshmem-linuxapp-gcc
delete mode 100644 config/defconfig_x86_64-ivshmem-linuxapp-icc
delete mode 100644 doc/guides/prog_guide/img/ivshmem.png
delete mode 100644 doc/guides/prog_guide/ivshmem_lib.rst
delete mode 100644 examples/l2fwd-ivshmem/Makefile
delete mode 100644 examples/l2fwd-ivshmem/guest/Makefile
delete mode 100644 examples/l2fwd-ivshmem/guest/guest.c
delete mode 100644 examples/l2fwd-ivshmem/host/Makefile
delete mode 100644 examples/l2fwd-ivshmem/host/host.c
delete mode 100644 examples/l2fwd-ivshmem/include/common.h
delete mode 100644 lib/librte_eal/linuxapp/eal/eal_ivshmem.c
delete mode 100644 lib/librte_ivshmem/Makefile
delete mode 100644 lib/librte_ivshmem/rte_ivshmem.c
delete mode 100644 lib/librte_ivshmem/rte_ivshmem.h
delete mode 100644 lib/librte_ivshmem/rte_ivshmem_version.map
diff --git a/MAINTAINERS b/MAINTAINERS
index 6536c6b..5e3d825 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -546,14 +546,6 @@ F: app/test/test_cmdline*
F: examples/cmdline/
F: doc/guides/sample_app_ug/cmd_line.rst
-Qemu IVSHMEM
-M: Anatoly Burakov <anatoly.burakov@intel.com>
-F: lib/librte_ivshmem/
-F: lib/librte_eal/linuxapp/eal/eal_ivshmem.c
-F: doc/guides/prog_guide/ivshmem_lib.rst
-F: app/test/test_ivshmem.c
-F: examples/l2fwd-ivshmem/
-
Key/Value parsing
M: Olivier Matz <olivier.matz@6wind.com>
F: lib/librte_kvargs/
diff --git a/app/test/Makefile b/app/test/Makefile
index 49ea195..611d77a 100644
--- a/app/test/Makefile
+++ b/app/test/Makefile
@@ -167,7 +167,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_KNI) += test_kni.c
SRCS-$(CONFIG_RTE_LIBRTE_POWER) += test_power.c test_power_acpi_cpufreq.c
SRCS-$(CONFIG_RTE_LIBRTE_POWER) += test_power_kvm_vm.c
SRCS-y += test_common.c
-SRCS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += test_ivshmem.c
SRCS-$(CONFIG_RTE_LIBRTE_DISTRIBUTOR) += test_distributor.c
SRCS-$(CONFIG_RTE_LIBRTE_DISTRIBUTOR) += test_distributor_perf.c
diff --git a/app/test/autotest_data.py b/app/test/autotest_data.py
index defd46e..9e8fd94 100644
--- a/app/test/autotest_data.py
+++ b/app/test/autotest_data.py
@@ -174,12 +174,6 @@ parallel_test_group_list = [
"Report" : None,
},
{
- "Name" : "IVSHMEM autotest",
- "Command" : "ivshmem_autotest",
- "Func" : default_autotest,
- "Report" : None,
- },
- {
"Name" : "Memcpy autotest",
"Command" : "memcpy_autotest",
"Func" : default_autotest,
diff --git a/app/test/test.c b/app/test/test.c
index ccad0e3..cd0e784 100644
--- a/app/test/test.c
+++ b/app/test/test.c
@@ -95,9 +95,6 @@ do_recursive_call(void)
{ "test_memory_flags", no_action },
{ "test_file_prefix", no_action },
{ "test_no_huge_flag", no_action },
-#ifdef RTE_LIBRTE_IVSHMEM
- { "test_ivshmem", test_ivshmem },
-#endif
};
if (recursive_call == NULL)
diff --git a/app/test/test.h b/app/test/test.h
index 467b9c0..b250c84 100644
--- a/app/test/test.h
+++ b/app/test/test.h
@@ -235,7 +235,6 @@ int test_pci_run;
int test_mp_secondary(void);
-int test_ivshmem(void);
int test_set_rxtx_conf(cmdline_fixed_string_t mode);
int test_set_rxtx_anchor(cmdline_fixed_string_t type);
int test_set_rxtx_sc(cmdline_fixed_string_t type);
diff --git a/app/test/test_ivshmem.c b/app/test/test_ivshmem.c
deleted file mode 100644
index ae9fd6c..0000000
--- a/app/test/test_ivshmem.c
+++ /dev/null
@@ -1,433 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#include <fcntl.h>
-#include <limits.h>
-#include <unistd.h>
-#include <string.h>
-#include <sys/mman.h>
-#include <sys/wait.h>
-#include <stdio.h>
-
-#include <cmdline_parse.h>
-
-#include "test.h"
-
-#include <rte_common.h>
-#include <rte_ivshmem.h>
-#include <rte_string_fns.h>
-#include "process.h"
-
-#define DUPLICATE_METADATA "duplicate"
-#define METADATA_NAME "metadata"
-#define NONEXISTENT_METADATA "nonexistent"
-#define FIRST_TEST 'a'
-
-#define launch_proc(ARGV) process_dup(ARGV, \
- sizeof(ARGV)/(sizeof(ARGV[0])), "test_ivshmem")
-
-#define ASSERT(cond,msg) do { \
- if (!(cond)) { \
- printf("**** TEST %s() failed: %s\n", \
- __func__, msg); \
- return -1; \
- } \
-} while(0)
-
-static char*
-get_current_prefix(char * prefix, int size)
-{
- char path[PATH_MAX] = {0};
- char buf[PATH_MAX] = {0};
-
- /* get file for config (fd is always 3) */
- snprintf(path, sizeof(path), "/proc/self/fd/%d", 3);
-
- /* return NULL on error */
- if (readlink(path, buf, sizeof(buf)) == -1)
- return NULL;
-
- /* get the basename */
- snprintf(buf, sizeof(buf), "%s", basename(buf));
-
- /* copy string all the way from second char up to start of _config */
- snprintf(prefix, size, "%.*s",
- (int)(strnlen(buf, sizeof(buf)) - sizeof("_config")),
- &buf[1]);
-
- return prefix;
-}
-
-static struct rte_ivshmem_metadata*
-mmap_metadata(const char *name)
-{
- int fd;
- char pathname[PATH_MAX];
- struct rte_ivshmem_metadata *metadata;
-
- snprintf(pathname, sizeof(pathname),
- "/var/run/.dpdk_ivshmem_metadata_%s", name);
-
- fd = open(pathname, O_RDWR, 0660);
- if (fd < 0)
- return NULL;
-
- metadata = (struct rte_ivshmem_metadata*) mmap(NULL,
- sizeof(struct rte_ivshmem_metadata), PROT_READ | PROT_WRITE,
- MAP_SHARED, fd, 0);
-
- if (metadata == MAP_FAILED)
- return NULL;
-
- close(fd);
-
- return metadata;
-}
-
-static int
-create_duplicate(void)
-{
- /* create a metadata that another process will then try to overwrite */
- ASSERT (rte_ivshmem_metadata_create(DUPLICATE_METADATA) == 0,
- "Creating metadata failed");
- return 0;
-}
-
-static int
-test_ivshmem_create_lots_of_memzones(void)
-{
- int i;
- char name[IVSHMEM_NAME_LEN];
- const struct rte_memzone *mz;
-
- ASSERT(rte_ivshmem_metadata_create(METADATA_NAME) == 0,
- "Failed to create metadata");
-
- for (i = 0; i < RTE_LIBRTE_IVSHMEM_MAX_ENTRIES; i++) {
- snprintf(name, sizeof(name), "mz_%i", i);
-
- mz = rte_memzone_reserve(name, RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY, 0);
- ASSERT(mz != NULL, "Failed to reserve memzone");
-
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, METADATA_NAME) == 0,
- "Failed to add memzone");
- }
- mz = rte_memzone_reserve("one too many", RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY, 0);
- ASSERT(mz != NULL, "Failed to reserve memzone");
-
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, METADATA_NAME) < 0,
- "Metadata should have been full");
-
- return 0;
-}
-
-static int
-test_ivshmem_create_duplicate_memzone(void)
-{
- const struct rte_memzone *mz;
-
- ASSERT(rte_ivshmem_metadata_create(METADATA_NAME) == 0,
- "Failed to create metadata");
-
- mz = rte_memzone_reserve("mz", RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY, 0);
- ASSERT(mz != NULL, "Failed to reserve memzone");
-
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, METADATA_NAME) == 0,
- "Failed to add memzone");
-
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, METADATA_NAME) < 0,
- "Added the same memzone twice");
-
- return 0;
-}
-
-static int
-test_ivshmem_api_test(void)
-{
- const struct rte_memzone * mz;
- struct rte_mempool * mp;
- struct rte_ring * r;
- char buf[BUFSIZ];
-
- memset(buf, 0, sizeof(buf));
-
- r = rte_ring_create("ring", 1, SOCKET_ID_ANY, 0);
- mp = rte_mempool_create("mempool", 1, 1, 1, 1, NULL, NULL, NULL, NULL,
- SOCKET_ID_ANY, 0);
- mz = rte_memzone_reserve("memzone", 64, SOCKET_ID_ANY, 0);
-
- ASSERT(r != NULL, "Failed to create ring");
- ASSERT(mp != NULL, "Failed to create mempool");
- ASSERT(mz != NULL, "Failed to reserve memzone");
-
- /* try to create NULL metadata */
- ASSERT(rte_ivshmem_metadata_create(NULL) < 0,
- "Created metadata with NULL name");
-
- /* create valid metadata to do tests on */
- ASSERT(rte_ivshmem_metadata_create(METADATA_NAME) == 0,
- "Failed to create metadata");
-
- /* test adding memzone */
- ASSERT(rte_ivshmem_metadata_add_memzone(NULL, NULL) < 0,
- "Added NULL memzone to NULL metadata");
- ASSERT(rte_ivshmem_metadata_add_memzone(NULL, METADATA_NAME) < 0,
- "Added NULL memzone");
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, NULL) < 0,
- "Added memzone to NULL metadata");
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, NONEXISTENT_METADATA) < 0,
- "Added memzone to nonexistent metadata");
-
- /* test adding ring */
- ASSERT(rte_ivshmem_metadata_add_ring(NULL, NULL) < 0,
- "Added NULL ring to NULL metadata");
- ASSERT(rte_ivshmem_metadata_add_ring(NULL, METADATA_NAME) < 0,
- "Added NULL ring");
- ASSERT(rte_ivshmem_metadata_add_ring(r, NULL) < 0,
- "Added ring to NULL metadata");
- ASSERT(rte_ivshmem_metadata_add_ring(r, NONEXISTENT_METADATA) < 0,
- "Added ring to nonexistent metadata");
-
- /* test adding mempool */
- ASSERT(rte_ivshmem_metadata_add_mempool(NULL, NULL) < 0,
- "Added NULL mempool to NULL metadata");
- ASSERT(rte_ivshmem_metadata_add_mempool(NULL, METADATA_NAME) < 0,
- "Added NULL mempool");
- ASSERT(rte_ivshmem_metadata_add_mempool(mp, NULL) < 0,
- "Added mempool to NULL metadata");
- ASSERT(rte_ivshmem_metadata_add_mempool(mp, NONEXISTENT_METADATA) < 0,
- "Added mempool to nonexistent metadata");
-
- /* test creating command line */
- ASSERT(rte_ivshmem_metadata_cmdline_generate(NULL, sizeof(buf), METADATA_NAME) < 0,
- "Written command line into NULL buffer");
- ASSERT(strnlen(buf, sizeof(buf)) == 0, "Buffer is not empty");
-
- ASSERT(rte_ivshmem_metadata_cmdline_generate(buf, 0, METADATA_NAME) < 0,
- "Written command line into small buffer");
- ASSERT(strnlen(buf, sizeof(buf)) == 0, "Buffer is not empty");
-
- ASSERT(rte_ivshmem_metadata_cmdline_generate(buf, sizeof(buf), NULL) < 0,
- "Written command line for NULL metadata");
- ASSERT(strnlen(buf, sizeof(buf)) == 0, "Buffer is not empty");
-
- ASSERT(rte_ivshmem_metadata_cmdline_generate(buf, sizeof(buf),
- NONEXISTENT_METADATA) < 0,
- "Writen command line for nonexistent metadata");
- ASSERT(strnlen(buf, sizeof(buf)) == 0, "Buffer is not empty");
-
- /* add stuff to config */
- ASSERT(rte_ivshmem_metadata_add_memzone(mz, METADATA_NAME) == 0,
- "Failed to add memzone to valid config");
- ASSERT(rte_ivshmem_metadata_add_ring(r, METADATA_NAME) == 0,
- "Failed to add ring to valid config");
- ASSERT(rte_ivshmem_metadata_add_mempool(mp, METADATA_NAME) == 0,
- "Failed to add mempool to valid config");
-
- /* create config */
- ASSERT(rte_ivshmem_metadata_cmdline_generate(buf, sizeof(buf),
- METADATA_NAME) == 0, "Failed to write command-line");
-
- /* check if something was written */
- ASSERT(strnlen(buf, sizeof(buf)) != 0, "Buffer is empty");
-
- /* make sure we don't segfault */
- rte_ivshmem_metadata_dump(stdout, NULL);
-
- /* dump our metadata */
- rte_ivshmem_metadata_dump(stdout, METADATA_NAME);
-
- return 0;
-}
-
-static int
-test_ivshmem_create_duplicate_metadata(void)
-{
- ASSERT(rte_ivshmem_metadata_create(DUPLICATE_METADATA) < 0,
- "Creating duplicate metadata should have failed");
-
- return 0;
-}
-
-static int
-test_ivshmem_create_metadata_config(void)
-{
- struct rte_ivshmem_metadata *metadata;
-
- rte_ivshmem_metadata_create(METADATA_NAME);
-
- metadata = mmap_metadata(METADATA_NAME);
-
- ASSERT(metadata != MAP_FAILED, "Metadata mmaping failed");
-
- ASSERT(metadata->magic_number == IVSHMEM_MAGIC,
- "Magic number is not that magic");
-
- ASSERT(strncmp(metadata->name, METADATA_NAME, sizeof(metadata->name)) == 0,
- "Name has not been set up");
-
- ASSERT(metadata->entry[0].offset == 0, "Offest is not initialized");
- ASSERT(metadata->entry[0].mz.addr == 0, "mz.addr is not initialized");
- ASSERT(metadata->entry[0].mz.len == 0, "mz.len is not initialized");
-
- return 0;
-}
-
-static int
-test_ivshmem_create_multiple_metadata_configs(void)
-{
- int i;
- char name[IVSHMEM_NAME_LEN];
- struct rte_ivshmem_metadata *metadata;
-
- for (i = 0; i < RTE_LIBRTE_IVSHMEM_MAX_METADATA_FILES / 2; i++) {
- snprintf(name, sizeof(name), "test_%d", i);
- rte_ivshmem_metadata_create(name);
- metadata = mmap_metadata(name);
-
- ASSERT(metadata->magic_number == IVSHMEM_MAGIC,
- "Magic number is not that magic");
-
- ASSERT(strncmp(metadata->name, name, sizeof(metadata->name)) == 0,
- "Name has not been set up");
- }
-
- return 0;
-}
-
-static int
-test_ivshmem_create_too_many_metadata_configs(void)
-{
- int i;
- char name[IVSHMEM_NAME_LEN];
-
- for (i = 0; i < RTE_LIBRTE_IVSHMEM_MAX_METADATA_FILES; i++) {
- snprintf(name, sizeof(name), "test_%d", i);
- ASSERT(rte_ivshmem_metadata_create(name) == 0,
- "Create config file failed");
- }
-
- ASSERT(rte_ivshmem_metadata_create(name) < 0,
- "Create config file didn't fail");
-
- return 0;
-}
-
-enum rte_ivshmem_tests {
- _test_ivshmem_api_test = 0,
- _test_ivshmem_create_metadata_config,
- _test_ivshmem_create_multiple_metadata_configs,
- _test_ivshmem_create_too_many_metadata_configs,
- _test_ivshmem_create_duplicate_metadata,
- _test_ivshmem_create_lots_of_memzones,
- _test_ivshmem_create_duplicate_memzone,
- _last_test,
-};
-
-#define RTE_IVSHMEM_TEST_ID "RTE_IVSHMEM_TEST_ID"
-
-static int
-launch_all_tests_on_secondary_processes(void)
-{
- int ret = 0;
- char id;
- char testid;
- char tmp[PATH_MAX] = {0};
- char prefix[PATH_MAX] = {0};
-
- get_current_prefix(tmp, sizeof(tmp));
-
- snprintf(prefix, sizeof(prefix), "--file-prefix=%s", tmp);
-
- const char *argv[] = { prgname, "-c", "1", "-n", "3",
- "--proc-type=secondary", prefix };
-
- for (id = 0; id < _last_test; id++) {
- testid = (char)(FIRST_TEST + id);
- setenv(RTE_IVSHMEM_TEST_ID, &testid, 1);
- if (launch_proc(argv) != 0)
- return -1;
- }
- return ret;
-}
-
-int
-test_ivshmem(void)
-{
- int testid;
-
- /* We want to have a clean execution for every test without exposing
- * private global data structures in rte_ivshmem so we launch each test
- * on a different secondary process. */
- if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
-
- /* first, create metadata */
- ASSERT(create_duplicate() == 0, "Creating metadata failed");
-
- return launch_all_tests_on_secondary_processes();
- }
-
- testid = *(getenv(RTE_IVSHMEM_TEST_ID)) - FIRST_TEST;
-
- printf("Secondary process running test %d \n", testid);
-
- switch (testid) {
- case _test_ivshmem_api_test:
- return test_ivshmem_api_test();
-
- case _test_ivshmem_create_metadata_config:
- return test_ivshmem_create_metadata_config();
-
- case _test_ivshmem_create_multiple_metadata_configs:
- return test_ivshmem_create_multiple_metadata_configs();
-
- case _test_ivshmem_create_too_many_metadata_configs:
- return test_ivshmem_create_too_many_metadata_configs();
-
- case _test_ivshmem_create_duplicate_metadata:
- return test_ivshmem_create_duplicate_metadata();
-
- case _test_ivshmem_create_lots_of_memzones:
- return test_ivshmem_create_lots_of_memzones();
-
- case _test_ivshmem_create_duplicate_memzone:
- return test_ivshmem_create_duplicate_memzone();
-
- default:
- break;
- }
-
- return -1;
-}
-
-REGISTER_TEST_COMMAND(ivshmem_autotest, test_ivshmem);
diff --git a/config/defconfig_arm64-armv8a-linuxapp-gcc b/config/defconfig_arm64-armv8a-linuxapp-gcc
index 1a17126..73f4733 100644
--- a/config/defconfig_arm64-armv8a-linuxapp-gcc
+++ b/config/defconfig_arm64-armv8a-linuxapp-gcc
@@ -44,7 +44,6 @@ CONFIG_RTE_TOOLCHAIN_GCC=y
CONFIG_RTE_EAL_IGB_UIO=n
-CONFIG_RTE_LIBRTE_IVSHMEM=n
CONFIG_RTE_LIBRTE_FM10K_PMD=n
CONFIG_RTE_LIBRTE_I40E_PMD=n
diff --git a/config/defconfig_x86_64-ivshmem-linuxapp-gcc b/config/defconfig_x86_64-ivshmem-linuxapp-gcc
deleted file mode 100644
index 41ac5c3..0000000
--- a/config/defconfig_x86_64-ivshmem-linuxapp-gcc
+++ /dev/null
@@ -1,49 +0,0 @@
-# BSD LICENSE
-#
-# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
-# All rights reserved.
-#
-# Redistribution and use in source and binary forms, with or without
-# modification, are permitted provided that the following conditions
-# are met:
-#
-# * Redistributions of source code must retain the above copyright
-# notice, this list of conditions and the following disclaimer.
-# * Redistributions in binary form must reproduce the above copyright
-# notice, this list of conditions and the following disclaimer in
-# the documentation and/or other materials provided with the
-# distribution.
-# * Neither the name of Intel Corporation nor the names of its
-# contributors may be used to endorse or promote products derived
-# from this software without specific prior written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-#
-
-#
-# use default config
-#
-
-#include "defconfig_x86_64-native-linuxapp-gcc"
-
-#
-# Compile IVSHMEM library
-#
-CONFIG_RTE_LIBRTE_IVSHMEM=y
-CONFIG_RTE_LIBRTE_IVSHMEM_DEBUG=n
-CONFIG_RTE_LIBRTE_IVSHMEM_MAX_PCI_DEVS=4
-CONFIG_RTE_LIBRTE_IVSHMEM_MAX_ENTRIES=128
-CONFIG_RTE_LIBRTE_IVSHMEM_MAX_METADATA_FILES=32
-
-# Set EAL to single file segments
-CONFIG_RTE_EAL_SINGLE_FILE_SEGMENTS=y
\ No newline at end of file
diff --git a/config/defconfig_x86_64-ivshmem-linuxapp-icc b/config/defconfig_x86_64-ivshmem-linuxapp-icc
deleted file mode 100644
index 77fec93..0000000
--- a/config/defconfig_x86_64-ivshmem-linuxapp-icc
+++ /dev/null
@@ -1,49 +0,0 @@
-# BSD LICENSE
-#
-# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
-# All rights reserved.
-#
-# Redistribution and use in source and binary forms, with or without
-# modification, are permitted provided that the following conditions
-# are met:
-#
-# * Redistributions of source code must retain the above copyright
-# notice, this list of conditions and the following disclaimer.
-# * Redistributions in binary form must reproduce the above copyright
-# notice, this list of conditions and the following disclaimer in
-# the documentation and/or other materials provided with the
-# distribution.
-# * Neither the name of Intel Corporation nor the names of its
-# contributors may be used to endorse or promote products derived
-# from this software without specific prior written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-#
-
-#
-# use default config
-#
-
-#include "defconfig_x86_64-native-linuxapp-icc"
-
-#
-# Compile IVSHMEM library
-#
-CONFIG_RTE_LIBRTE_IVSHMEM=y
-CONFIG_RTE_LIBRTE_IVSHMEM_DEBUG=n
-CONFIG_RTE_LIBRTE_IVSHMEM_MAX_PCI_DEVS=4
-CONFIG_RTE_LIBRTE_IVSHMEM_MAX_ENTRIES=128
-CONFIG_RTE_LIBRTE_IVSHMEM_MAX_METADATA_FILES=32
-
-# Set EAL to single file segments
-CONFIG_RTE_EAL_SINGLE_FILE_SEGMENTS=y
diff --git a/doc/api/doxy-api-index.md b/doc/api/doxy-api-index.md
index 2284a53..6675f96 100644
--- a/doc/api/doxy-api-index.md
+++ b/doc/api/doxy-api-index.md
@@ -108,7 +108,6 @@ There are many libraries, so their headers may be grouped by topics:
[reorder] (@ref rte_reorder.h),
[tailq] (@ref rte_tailq.h),
[bitmap] (@ref rte_bitmap.h),
- [ivshmem] (@ref rte_ivshmem.h)
- **packet framework**:
* [port] (@ref rte_port.h):
diff --git a/doc/api/doxy-api.conf b/doc/api/doxy-api.conf
index af5d6dd..9dc7ae5 100644
--- a/doc/api/doxy-api.conf
+++ b/doc/api/doxy-api.conf
@@ -43,7 +43,6 @@ INPUT = doc/api/doxy-api-index.md \
lib/librte_ether \
lib/librte_hash \
lib/librte_ip_frag \
- lib/librte_ivshmem \
lib/librte_jobstats \
lib/librte_kni \
lib/librte_kvargs \
diff --git a/doc/api/examples.dox b/doc/api/examples.dox
index 200af0b..1626852 100644
--- a/doc/api/examples.dox
+++ b/doc/api/examples.dox
@@ -40,8 +40,6 @@
@example ipv4_multicast/main.c
@example kni/main.c
@example l2fwd-crypto/main.c
-@example l2fwd-ivshmem/guest/guest.c
-@example l2fwd-ivshmem/host/host.c
@example l2fwd-jobstats/main.c
@example l2fwd-keepalive/main.c
@example l2fwd/main.c
diff --git a/doc/guides/linux_gsg/build_dpdk.rst b/doc/guides/linux_gsg/build_dpdk.rst
index f8007b3..474598a 100644
--- a/doc/guides/linux_gsg/build_dpdk.rst
+++ b/doc/guides/linux_gsg/build_dpdk.rst
@@ -75,7 +75,7 @@ where:
* ``ARCH`` can be: ``i686``, ``x86_64``, ``ppc_64``
-* ``MACHINE`` can be: ``native``, ``ivshmem``, ``power8``
+* ``MACHINE`` can be: ``native``, ``power8``
* ``EXECENV`` can be: ``linuxapp``, ``bsdapp``
diff --git a/doc/guides/linux_gsg/quick_start.rst b/doc/guides/linux_gsg/quick_start.rst
index 8789b58..6e858c2 100644
--- a/doc/guides/linux_gsg/quick_start.rst
+++ b/doc/guides/linux_gsg/quick_start.rst
@@ -126,19 +126,15 @@ Some options in the script prompt the user for further data before proceeding.
[3] ppc_64-power8-linuxapp-gcc
- [4] x86_64-ivshmem-linuxapp-gcc
+ [4] x86_64-native-bsdapp-clang
- [5] x86_64-ivshmem-linuxapp-icc
+ [5] x86_64-native-bsdapp-gcc
- [6] x86_64-native-bsdapp-clang
+ [6] x86_64-native-linuxapp-clang
- [7] x86_64-native-bsdapp-gcc
+ [7] x86_64-native-linuxapp-gcc
- [8] x86_64-native-linuxapp-clang
-
- [9] x86_64-native-linuxapp-gcc
-
- [10] x86_64-native-linuxapp-icc
+ [8] x86_64-native-linuxapp-icc
------------------------------------------------------------------------
diff --git a/doc/guides/prog_guide/img/ivshmem.png b/doc/guides/prog_guide/img/ivshmem.png
deleted file mode 100644
index 2b34a2cfdf3d65025c1f942f5cc1f2a1ef05892c..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001
literal 44920
zcmdSA1Cu4q^ESL=dj~tVZR3n>+qP})*x0eH9ox2TduGRa&;5J;U*bIxCpx;LJF7DD
zs;e@qDkBx-#1UYzVSoJifgmX%qV(ej2oLbz3k?Z;vK2Xl2fTndD~SVs)J)@^0dK&~
zg=B?({HTwE`!Iw6-orRZXgdG+fjIczA0&ka5eNA2my4)|i?Y3$i@Twd=?_^;V<&rO
z;2oj7nVF@rss9z5<BuOTr;;K<Djs@Qx=;q_Ll}H#*~oyKE=lw{MG<85s-HA&m#Ziu
z;ND>9-XQ2h_o_cdMC%k)Gv~eGkfGy|=e;-lXD;d227iDgbslzcwH|k^`F1pRG@8$J
zIiJL)5e0++e^~<?Z<YD^`FqQ=CZ)g&%Ku#`IT|J=CWL)_Iv=Zz;!{(@ezBMlQc;aj
z&&|%xGAbUMq@|}vI35fe&gKbWvD>6;`F*{bTidP~e7-$ONJ!LmqX&)Tuv(%b67buk
zSO9<Nl*$d#va&;TEt^4jJZ`&@6gj<61U><M!B7XY`NCGda&oa+HYR3f1$lYBZ;*$_
z$M{>zFp25asErK`jAoNqK95W4qobn%jVu|tpP!#7KYxZC9v%{pj89L~-)Fz)zy0mI
z@`PWGys){&k_#O-cF+g+7iK#Hk8|AZ4J3XieBsH#X_1$gS3j{LVP{vvTY>q?**G~l
z(I&&j#{Pyd#rC@YdcQenoZ%nXY;#7z$G1zt2)Y0T1$9_bRWHAh=YK6C9-!B4Hy8**
zc-jBWkWBqMI(o1$j=$n-`VSY{!^w;~xZ7EA8g)V2b^8TSC8$l0gP7K*b!!w{+^Q!`
zG_=Cx!`?tJ^u2gWu^8;oX4{SZ4f`HZDJfXHu1B?|tHDULxk`10f0t`$Kpzb=^n4UR
zhl6vPj3PJDYBqp3noTM7Sg+LJU3Wd}y}Z2azMkZY1Fc=}cBN&?yyLcEd%Dvj@b!8^
z;PY`e64l$53zUv${>Kp#5uZ1rT)DKOhqQgY$=djEG~xUvnM%3N``OO*lINyV*|x`Y
z0z)Q?T{eNL=fUU7d2x1@lP$TtLA%YV*y~i%<^AbuAWh$|T*7Rt-PP)|LbdXz(CW)o
z%T97dzEdIU;eh{wEC&_KV|B~6(mP?G+xa5QijFga)nZv-DxG$LP%z}_W*bYp+ZC+7
zzP`1+6Z7`ARHW@jGeVk<)0kpo>T0_yO)7(a4L{m%Oa`@(R{w9m&e!9ts27Qn&9L{)
zChIifgX^sh8o=jqmVL57{Dp^H6;Iq;kt7=L*UJH?(_sX0tMfFMcmhF-_f^x1if@U0
zZp|XSD#H@P1%TN9{i<cDi74alkHjshbE>od=8R=&HY5WBL*+8i^`%kFWIPoZ0s;b$
z8Klh)dw|Y|X&x)0vh&u}JS{jFn7-i{9Q1nCN;O%&OW7=TRjIf#_1YZb(etHBntAKF
zf>0$@pj8IWN^@%_)o@60u9NtMQBfoZpROD`dV2*QGj`2W6`K?z6{(aA6J@kp>|%Pp
zzqHHQmm^Cov$<VW_$10mE8cH<2r%`1>WM9DPPqQKE^D^#2Eip49D3ei-}*?jf5?u+
z;?0%F;xzepBl~^aiHqJ=2?KGs^rTX{X?C;SC3TeJ`?Plchn1D}3uip}@}(CDgxqr$
zbQH<!ahh7*^OXXC<6DIVY9g-$HK}Bxh>p%5C2`UuAQ1Ob)b$jjV$aV9kOWkE4&3DY
zv(PpvBEtwH*AAtuw=C;7Mo1#(MGhUyK;IuPN=L)UDJIT%3dajPWg0W2VtUA4QAlOz
zd66=2RPsofzQRna<bimpcR5ok{;4_bOyoh=iZG!ubCTmNmrXPCp9hC4YrCU5Zo0Lf
z>8&l2L4e(@>G|?9{hBK^O#&rQ$x0_CCQb$=#-P>6o0LUG_O&8~r*7IC2&YlvdYEKc
zvX_Kp=)4WO&#zYwB4mmugf@qQdKLw(he<IR4ke7V+3D7$qN8TbtRxk619kAG7NZtV
zgqO?!ANqW}+E8B#KW0P!6(_F0ev<27^TcDn&3j~D*fz%e1vH>Jx?vJ!`TA)vN#Js|
zwt9lVrvtd`?(D3B>-468Dp8N})|SHn!ntX^%^LHgHv&S;cNh>nZ?mgl6NsS`^c^x8
zo-SX{XA9!hJ^N8KKLhJOfc&83M~y07saCz%9**xNK}CX!BWsWdJ5kg9W+f`;vWs>Z
zuzPcMK%B8k$duUol18{T4^Fx_M<9N`ufLp7!WbvZl5$~br8JeWdYTIytq+R;C*ofT
zqmqfN{*ujANHYmmF^|L<^DZSp2mF4YkLs*;n;Nz=#&u{xBZd?RJdzd-u>=D6{-2L&
z-1_(@#c6s<kImloo_?5m9{I!y|5L(A5?pXUA}UI0aJkdd*)(<?4@O1F`{_~$NIE}J
zRuB8EW+P`1!QL9n3M|B~H`|mGkX4cWg%zljh}Ox8zfnoi-cHL1J$b3==|AL*jG<wc
zNL6v4WtmMFsKzp~p{t<0CD$T*<)HD+AC9NaeXN$N%BMrAZ|^4<)VsM+tu-{#&&$a?
z5VP2AmAa*u(?)jhB<-4i$v7P_t%3+Ijb8LQn|z*D9>N8b@Lot3kk47>W`_;?hmV-^
zy?#FLq+|e*7&06S^R_5qK^AD01RHBIgcySsI4B-EoTH!F{4pWT51cCt{RDv8-iOQG
zk>+_d4k&s3JI(i!jf1G@68L%%M!`cjw(LdB6#b1+j&Ghh1BA!doZL^~7g2%aJO~H~
zae^6c3ja-IDry!D3$wGg(5;ivrZZ<__?djEdV1^vY5%mnHB`smn}n6dVR^%i16}DZ
z4lFG!{%}+Oxx`xTO8~{Cm=N5b$KNJJVmLf}C8lA~wTSs8gKI3-Rdowb60AQt6OtwX
z3r_(4!sO6ee2zs)NhtXW{XS+7+sX%)^op}A3O&xF!EZLvZ_EiT9czdMar-ze^caCn
z-C11p)>s2iX-0SloO)Cxego%rYsEbiOgy|85HQHHRo2*a&Kz#1An*pGBdyD|dp>Sg
zn^@YL*wg^D#Cr_pvF#s+*HA&;k_kOkNh#Z3w<CtQn;p;>ZXpHgwcBN?mE?;7h^U&w
z_5uFZy|r9$CMZS^_eu8<_Z-O3yjz<Bn0_-_V6{QP5i=GvrXZ~cu^ZA=0ta7MP@@UR
z@ITFZ>w*3GE!0pcAtEaJE1Soi{dhXtIwcwZrbl4O!xVt|`ge}DE}?6h5mCn$tDL18
z6z&x!EP-iU<87fyESLWT2Z)VkHR%0ZwrF7z1QL#6>bm$3fq{s&Va5iNTse~}#^}!h
zc#;S6gB$;4ERG$Y%BBL7D$T}WkHZA<=+#w?N)ch<fs`_bQdD&8zWPRr-&K>O9^v{_
zxwr$RPAecV@?c9vye<){5~M%w3PqdPLe*p5caL;`*b~rPNbWZpwMD^JJ@D>o@6eLW
z_asJ5fm#U=^;ajUQjWgjNO-}NIVP2L!HKqcj!EyoL?s(Lz`Xmgv9jv?t~e4Z{YHq*
zSJni$ILd<(t&B4-QvW<00fQpI@NNi=U+bb|{76>Sj%NVR;+IUfAV7XCusM}r%)Adp
zUWX?HJe^o0ek+Zl7W_tYg75eR-7bQT{!^vws5U=e5wc9nq|^XK_Q`8Eg0jwv0~!T^
zWg?=JbGMF?`<oqL5vLidV(3%>18{&~5~NeK+>V_?qAEC5>H)FhCyA;F26p~UJeEjA
z{S#3!CQUOm=!0}kzTi<*Sh$cn?>qs|TQ}y)s6Y<6phGcYy8IMT`nAxgy&0nb#~K}#
z-{|@}c_(cA31Gn=ix`~z4KwFU2Tm!X5pt>qJvY|8N6M`{mx0VIn@%}G&HWq_7AAHi
zhR$LP7C6}>vIMiH($S4I7kP%wAr?`Sf}H)qhwVpPL2ujqk#e&nKCWMS_SS~H)dFyt
zi9~ftugL(o+%b%nkVK*Cei0FIEB@wpJ>rzD0su#6MF4fl<zUEEIs>fC<T9h1ibh_+
zZ|^~7-4V&ajC&Zd9Lm3A_gSF;Il7P=+&ep|xhYmzb${`2@#CF>(~ysf9dePAQJ+Nj
zS#B})yy4)6TIy2};NUh;sitTq6kS}F1+aoug9`Ecx)dB@F<?@WUn>5Dw&Y{*4=Exh
zNW1y>sQJaeA+9dHjLD=c>GE%Gi%P`39B5e{1AG#N2>O3|RLCh3#7ncYWz1zpW>r0w
zfc3mHFO5m5%o;Wc7sLfNy`S28@W>xOceQ6|AMVl7JW`Rj;1yk1bvLrI&{%aPW2NS}
zWYnqK2aH31@_?ZrWtmc6ii9Tvd5BDfNjn-`xL_}r#7K=fwx*bpgNT=5!)2-|LQbLr
zqg1hq{5+P)9*{h?5VF|a=%xERs|Mh**8)4o<*jQ>C%Pg!sv@I7k%4iF2n#f1NX1E`
zF$KinpBNm|j2d_b$9QB(nw$7vdi%;-v?atv;|K8#w|Anl>Y^?Hzo7wXi}rgg5To|>
zC^@0Ym7;PdO2rglJf#Y$Iukv$K~W@%_7|K$mp3xj(S-rn;{IvF%IG=x!s4TqF*>Y8
zJNc@ki`+sAKX+!0ipx+0y=ZJgN9E}#9MGjwaX^<^METK*EQ}I!SH4}+DwJIK5w)j<
zp5i*Xcs7becA39D`dDd)234g1T;Lh<<EGbs5mnsQU1kjoSw%E7_6>sE$ww7PCmIV7
zg)dt!v9VU5R$#0-E&@h}09wyunBpEV!lGrvR<g-P6E0*}dWo-H$eENJ=x#=jd1lqj
z{)G|KmhG@}b4X=Bvj@aWBd)2EPKiSZeXTM~|DTRBYp<D66Jds-X_N(panoGj(T*-V
z&>6Bnf;g#MYaG<$k(afIYD#56Kh!-Uk?{tkqbRxEJVu`&#!5UV>)XUqKbE@cf!2ai
zoirag*XpXLj=k0-qROmE3OW5F+}mBZe<~8{*_13RDQ5&~PM_?w_T;fgHstFfE;6U6
zn0VS*nK5+vYU(5pv=?<S&|a%JG(|%_Zn)XR+@)ty1zDy;bDF}v>G4S+DZILqyEwly
zYU+*JW9xl@$fBcMc!XJSIbyOYAeleT>;cYZ|Dn@FT$GHO?%9?oD^X^JK~_{<c~Z_C
z7w_Q6=@}H|8N?4$GZeJ=bkIb##G0v~us~TvTolu)8rei^V6qd>*apaV)O`P0*)C%6
zIW)R}uE067hLk%t8hK&qm$-Qq%6_8lXJ*tq8p~4H{apP@ipUA0t}VfqVu3RyH}$|j
z0@L@ofZ9THbs+!D7g{Cg`ZXY{No(fNin&bc0SdhJMk`m5>;8@jD#&vwXSM@OLSTWO
zk1erwEJ*UjTrH^-?jn<AiKGWW$CFU2P=o?0M^$L)VJ%dL&>zpN_}sNP?s~02q!jWI
z?iFGOTBSn@c92!I?G`bstEPgbz{3M3l72`f1_;V}-?Il9=0}uJjd;6iV8UBQ4P@Gp
zz>n$pxWZpyxoG#)sa+v>Sh&CTSb9mB!>Sz)g~!3yQXO5$LemuHVDde~HEF=9jM_`4
zm?km+Q3?kRT%QI1>khiITL`*Lm5&R+jVz;vnjqU5I{gnhOq}IfMn?_woZ~-;Ncujd
z|7hVAjcXW5@GA1YV_7Yq&^sx;-*Fl9e|%RV?xy1YD!}zKd%>kCJb=kNlEffgB&|bN
zW>-iMSzKSkpbB?DI`-}KPd3WGxiLqq%8^NG_WKYSH2owb)muJbn%`HZB~Lq2$Q&Dc
zp!?0ng($mX9(rnrS>SAfB~}%`fI|#z4YS4t8oAcD3sjO-_pDGFtw6cnYkCmGf2>>N
zC=?M5)W!te3P?f_t<jesA0Jin#o@K?^dCMUMW9hAc+0<Gy&=lp0|s2kVk2OEetms`
z+}>LU1pIjWpe>TeAWrgKLP<F$ZNL0=pS2SR35b^~VEF=JE~;8Cgf8=B>+Gs>Y#s@y
zt!4Q42RKSnp!EJwZM^9at#)~H(6=3;Mq{(Vrs*DIjVZu(et+>;IbC&J<C{{m;K)qn
zFqBvbioX(*SNckcwI_=kpN*4vyzDWhGavD}T#nk_n>JSWOUR1ciH^FdNx=5W!PgoX
z9vmy>SfayX)hNYR#Maww?Yw+6$}`z!rnd*}r^S=D7!ARGyd1<~3Hb3^EupHG15K!;
z!nuaW>!Faal|Uk{P^-sxc(P%v8;4|XZ;#9E0?ot(g@QFoWQ;f)R~i<Q@B^>(;0S>M
zoREhHfr~4{VzCSb002x%)^vfHDl3}H6_8v?dPc?h9({S@a*4UQl^|=_>bT5_k}dCy
zj6eYsD&fcTrMpR%nLI7Fa0UfnEJIf2&@eGIHIzsuo!rl($r88G*bV_ff%(*TC7DXP
z5)crO^yTYVhtH}U2esS?%)u?T+MBqW|1T~CuJz>a!Fn_wj>^~Z17q4&9^S0WeTpA2
zwUODC;r^EY?*S()@slB;8CLvLqiPxOgV<Y}h6j5Pw#8Ht_g!qFZtuW9l+M{ek=lKY
zW3dZFiI*PeoQcGHcn06Eg8Lf^Ne<RP(FnYMF0n_y(?2!7R*TXawSR8=8m&joAH*{O
zqw1Cb^Q3HfJ|Zrf-%$vO_-X%khW>V8aC*?hcj-abyY$Tf={5x|N7l^`;5O;3j_P50
zuRQekQv>?OjnF+#vj!Ljk?pOSjDY``hVlMWC%bVgBzTqf<J+u414tuQ{r~lgnYu@p
zDgKmGNIsh*@Qo#4_o1k_ks$?058>oIPowLEV{35e`Y+i|p@3E~;8#%t17q~a&{j5?
zl`=(2jT8#NwiKZ0t?}X+C2dQ9Ek5q4MgEwdpGRqO#&Kev-@USRuS|u{X(GDF0`G~_
z63R6`>8_Eu_eF;2o|>gKfVT>z!(38_>tNOw)={DXe_=jasDCca-{Zy_O?$%C?7+`t
zx2|ct#09exb@Q*)i0#Pz6U}+!MZ6wVbjVpM*1*!~AU5G!hL;d4nk6T$r3M0f5C7L1
zH(cKC0~g`E#Bfgl!Qm#_IzAKS2gA;RImPm<15Br{0y06cP_~uPJ)Ard>0*MX_5tN)
z)Me`AH$<9zW_S^PVL+WivQ@LNJ=dp?K69>*8Etl13#k3(&X;iK+#Sl{p||0oz|8JI
z8I;JGWy3;Q$ZZ;XhbIL6^gE1x7e7co3O9?JqytUEyYrrX>wPsD=c03GysqHc=rq4~
zX()9M-2M3Aq?|BQ-52j*5yw!dwcl__GP_l8XqZ?r=;suUx}n4R%8Sdsg{<31!27S~
zFh*6s8ru6dRydf1-;FZRj4T|`@?w0UtN*wykUOE|IXrQ@kxj)NPd1-4;GQLkU9US{
z94~J^ysNG>k2gzRjO;ITVw<Klr+YTt?~GHQ_p5~OoZf47OA~9tAG}5BN4T4HhkMAD
zI$oadwP05nJqcrsg}j;LAQT<^DRKfip0{h)XLs9o_DbFsrovo#6^NYnS*M*7vQ7P3
zmJIHiPd}!woI{HqO!8Ncdl749#+YtUS5C&vzf83}?c6*6-SgHwT;Df-X1~v*g$v`!
zf8BGULo83|+0^N|X$I;TymKG&OvkMZbR*XE!d%d~(&^d1cQ@j3*7k%+?>;2=zUdsH
zbWy<WRILX873&K8G52J8t79W2fWSA9S|ItN_ZML^N*|4WFyV4E5vOMBw&fe9Dih>V
z1AOoboEQgeEIgq4;mBFJJgd&oiMH+yCr=Tx`~}<K$h`Nfskcv7(8HIg4ZeP&6m$b4
zoky0RvSeIm?F2w*#f99c_k;#HYvSmHe=+n(BLHR93=W{(A(K3;wG>o>N2c!4e4wk7
zKL7QJRQGJ37q-Zx(M-oczvr62uXoQMUq?{3GtG;k5C6^qDlL%8J=zqxgYfZ?q60;>
z@zf!4`7=+*now>0`$M4_3EXPzk!w!6v??z`Mi1IwCZYW*q5W@~mo(7@UgJ<_35wGd
z>XTGgPqLRi<BewOB2QWSX81SAMK>7g(btkKNs;V#-MaHyV38Lm<dqYL@ohlPEz*2_
zKABLDbu3gHH_RH_?c#N5d=5&`@=}Yj%!1_Y^PWr0GP|6DZ&eP&$9+M3H0wX|ts_HO
z)B<iJe)z28(JcM#dz(L006*(c2Q1>oD-MGlt2{XxqB65%Fuog4LSes+y5J94@Q>gl
zp`=H}iNn6M@G3QDLWR+3O{7PI@djN9x0=oY_Aird!!<m&FQRuJxmYgcCkBR~h%f*r
zfH?BKXj%!WfL9yrIXelK)(7F`)08OG7O<=|%$-E)=LNvm@MOAeU$P_`TmfGtUNP<I
zgl1j{JSwR8Ms46BJ(<4HQB-)GPWI|CTz>Ce?)R#F!%k-r(zD<x+TIhiUDm$dVfQS*
zN=cJw-+b=0$<yR0?}^_Z&V^gb5eqHk&1&%$#6J0W`SNhz;fs<8<9Gp8U(B#gj7qZb
zAX`|FVOv(0*#T(*%VvOt<RdPkuy?5dBNzwaC;K1~P}rec<Uu&)MCQEmA*Agm%{TD8
zoxuhD<H|Mrmu1#B0695c{{UZ*W3Ab&(4Pj^6ci5U54!}ujG{lxeeYd-mIyj~vaiAt
zL}bv#I)Md=bXe7mYk0i5F4MrIZZeL=K3hV&dM$;>Q{|b)4$WZ-oyqb)TIJUzv<A)-
z*oRGFYf!!;?)&{5*G8ROX%igJIThCX&^qjFyNaEup7h7<mMs<gscH5bTV2KCD)+K7
zKkD|Jma&O)I)}Y3!5*5UMxFN%i^eGpS;0c1{IQzkblguyH8L0BJb#QFizoPVNXK~_
zYqCBXPZ8)Si#cWdeM*C;q+X5#k;urkV_7bzXtitLB5zZA7l960#jPq4geNgQ77bD=
za{tDCo(<+hDiTbt^V|AM??dCmi`*=@ooh{e8gEhfAni8IcyLI+*IN9R!Cr?F1JNc+
zEtP)($*SkzK~tAPgvMkSc)9YTKEMwl1fDySIwB%1Z8&HpaG@3hP)j03EwPvr596C!
z696ln_`ag1$KIJEUgX!2HZEP>+$q>T+Jx3L{hmY#>+ogbnE;QTLQ%qAOEuB*`!$Q$
z%w|H*49y@zEL1p2*OE~EOhHVSx5FEBzERjg>robK<ucUY8FOkz$MV+aU?m=^^`O|O
z#ulRiEon*~UxOM`tmVPo$D`~Tn*T6c1bvgRWV(&SG}~tONoM!SUS!M_wA)uALI^)~
zKGU8QU4Y5^)xi!~*{v{r<{!-xy6PxY()$|C<N6zq!<U6(x>B4min%6<ICRaUX9>0z
z2`rb|WdDLl{amR;ot<<@2#H9O8`^vw%t2WY-!NQZXP;L~GA49n+#`P-$|>RV%CxgY
zLiZtGUMA(?HDJDqV}^_iB%Q*xxT);)xB&4ran1pTbu9(MXnHyZCuGLBMtx?34xwZ(
zaUDSXBs|)~N@ICW>nGlvmb%^Ek#201m}p?tk{Cuna!6r@LjeEgVsU*NNEuHtz#u3?
z$ixsF%4m6whnz&gD>IqTBWsj*6F0*p*+(8`9C%TF%^exE&OGyP%at&H)K4Nr6EF%V
z66k_PwpJFJJYi2AA{;{LK^N3N6oTRG`!!L(sXX-4#=M*GrD2tt+HD3l@5Z;J4XCw_
z1gl7Bi&o`!;L;%k@uME$d6wTIMyX_1@mJOJlWT#}H|qr+VTvi9G9uWHJR=YJbcgVj
zSS9z05YU}OR~#+9NDP<1!mA~1L(}H(3=UQlfqqj;gur3)HBIBSe)8z%0QXSezAZPp
z7Gp<2jRxnIUfV1kVnuIs9EcOF?T|5dud6D2d%MNhh4a)X!%p=j?ZISzlNzl#Ff|JY
zR}J)!_!vI7J=h}A<i@N3xS74cDu=*%vSTi!SQTdQrdN^ry1k29Uh6@Ad_vqVAN8tm
z$z47HoTwNS#u87<95!H3-bcq;2rN*4MTYQCR`*0iV=QqgFJ5LTwe0A#s9v?i{pk4m
zp(X=`;cIm@^Pzmrsb$YpsA;cn-gOjObWTT|LH*UCsMa+Eug}1c`M3jzLLdLcs4>2s
zs7?5sS8!gf2!72Gv_B=v>GEV;OhE6}fIR0S`&r9aeuedsy$VLc?jKQvt7!jZ9UL;S
zE&O)lG%xeOBm+G>VYMgzl)3kF=>oW~Gj?~*g7_N`LdAvES=c(OM>gI!zIAn?)W>gA
z?P)4}Mt^cY+C+0uhC5SB&rNK8cVP6u5$Ely>>sG90N1j){aTJxLRYkk_a?!GaHvvF
zbzlXv_Vn+Ns{!s&SLpcn!2{*Jgpbu<f=wdsKeEWjNL$0oiZmY99SFLe^e&;xC~+&d
ztK?IY-$%6zio6zF`>ZFK^~|?%-d@MCeI#wQ0{edL>@P$b|AUS)7EIb2>u=Y}9$4O!
z&?9{^XHTxQNX3I_8=*gVs)G7%nCz1lmo@M-i%l%<;_KOX-*tc2xx8w%`%MWMW`3Jb
z{36IEATG1~d-8F!r;EVvSf8FZOcx19XXgGMGR|5*E|I^{?}-bQq8%SSRJoiZZQt~X
zk3rg^fd>6g3!sp-4%%2(>Lf8T)*!fKY-~u9_~U}Mr7yhr<!RcNlfcHWISo=bz*+2v
z`L+Q){EWrsmX7(&&TGc|k-ItH<xq{W=K#MzLD5WAZlj(X*Y~>VW=Vk~^&@?`%z){v
zi97FY_0)7oA_kB}#po<`MQQ(9Sx0!T6p_u=Wwf!{Aud*F-amc2a;>HIzoq=mjneV>
zmV)d4E43~zN|%}jDE!EqT0J1*Ly?e7`5aABNDO)auXDW>g0Dpc4@IZWJP6#wvV;+|
z3)A7i?2<n-KX7PAmMc?<$`Z#EoF^ibKDT>Ur-64;3<eE6Z<;dBy+kXYT2%rLHObwF
z4DBC1LuhmChUCOn9+<z9t|f01%x@HU>k8L=8K+T|Gpdg~=)$b1*--L#1j3r<{7HcL
z>96KcBhrL+zp%t4@!kbjlq1fLb^m?|7WNJxum`hh1dC9r3=-5c9Z?iBHaIO6beJ8>
z>d3s5f{=A13p7bfh#TDzH#l!wn5&Gtq@K_@vb;@&I-9r&m{Xh~a_>Db+q=_(S2Y0O
za?7>BFFe%<lDuloAw;*>r_hqL!w>Xj31Ehf=;<|8nAA=4aDYjBfgwfo(;zogmPyf=
zO@s4SQJJ)Wlqt$4>2m$2WZh`ZQ=<3nR$V#g3Fs^R)f@>C@7<7jFC7g9@pC6r2`U%L
z8X3OWb6m|}91mt;ji%O~2F--&R=DxbBSA{yW%GC(J!np%SdxC9Ng8R)MX`gyEUc{C
z5i};3N(JN#_UK+6K|M3C=a{MdzGTqNudhwjUCQ;OW18df3CB6&Z_8u`)-~WBi3j-q
ze=gZSy7@nH%^CPRAt_)OG+&(614qp8<X|i2?|Y*3d`uF)S;sLmFJr+-7@&TY3%U`<
zryOSd=6+^v!+P~R7NtUf7iOOkYxkK<C94d=#Agvs4f?=+^`|0o6;mURCvB3e=?xu`
zP9O_sTcMQq_yv05oHCLN=Rc5QPv4*4NC4IkC~nfGk;CwTIQ&U?`+Ubq@$S4>eG>EG
z<D633CtfpDW3hYlW~?~-PEQ0u#jE!_zetKdD56%ZU?ry*`XSNUyr!H{Ap)Nr_Ex==
zXcUVV=Jr{@Pun2-`3BSs2}}$2&f_)X2t5=x|3o+@(Ap0iAbdDvl#+(w<hA-2G?=A(
zs1$94khLIHvqljdsAo91=Xl)UnBQUKvkFV_RIrIcB@wdT$%}~iEw>`XusBG@*!@1G
z8Q8<K9_xk)3V@p9C$w?MxUha<q-WJZV5PIDCf+LbIlX$f`rkg16(W&!>n?VuuESM|
zcM&Tf39+u>A^D8bv~Q`aDcJsr-p}LQZ=%_5m!LfqNOh)S=ss}N*qPVb+3XI2&_%h`
zf}Y7i1;zxhZ;Jf*FL|?9g=t9~)i1mAo@zb<PO=@gx}o8IUI++CR}4MYlg4SOr_hPA
zlJ8Vg{l;t3k{8mAf2ijyld&@|S*Z~V{+yHJ_Lt^vM~pS%4-et^Gi&d{R82}DB%1Wn
zbJaT<v47k7t&}AL56vht)j?lnF6rlg=1=v*=&j<dP106XfcJTMk9`yKgIje$hfJm1
z5C^jDsY(I256iG#pb1EEhewBQ%4!>V^aBO7A1mGlce6pk86aT+L%6nd{znNF<5Ej8
z`D1d0MuaBrEBZHurXO-e@8P>w{LjQBv@eIgfhh<={=Yr_BqC~1;@4DRSr=6Q+nl8I
ze*uX>5<Mt6T3A{k;MR*Oww-3edsF?TQk?$)vcpMO*+_f!h^88uZyzJ}8Uzqm*`k}^
z7yJF9D*+$d8-B7GTh-fp564pu9e?_P8qnKxFC@p@*ip8~*d$fx1W3vjd+jH%K<h$S
zsr}IL{XF~oyMFhtpQ`>+LK&Dk2nbJrPG+5X#u0W32GES-Nb~KT53r0rN0{U4Y1qhN
z9Piuq6<rnEOnuJKdPL-xAVk*~B|UN-7u0OLNtZ5#Rag<P>mLR`?eyOQ-`-?bXztt>
z@Kt-10U337g7K6*B+#Rr;>mh-cIqj(Iiha3h(9Ue!UON@HIb0^mln0wzzKYq9<M*F
zVb4o&0RRW^ZT@FHXA5Q3rzI59j)1F@9{L{cc)uJVddv5?;>-eQ9p`!r<O9~^!5##_
z5wB@GT9}@bvS?y@Dg?$%!9DZ;G~1)u>Cvxf-}YP&)HhpVK4W1F3GU||73iP^;xZXF
z@@UYoQnT;|UzU-x(}VKk)BB&m5EX;)V8Y#XAsim1OT`5*z;rjgu5~k$GWngb>)1FQ
zdUZtS4rTW3<bu)g)6hBxz1KYpi=O>AKhuBM-DGX7nX<RW?H^5FwQY+@KDwo*V0zGM
zUcVmbF3QA-+i2j)@Vu*G`6-<uR!obp%SZer$T86p;3`G_&q0MRBxinvAPRH7&t(!Z
zzmH%(uv^beciD-GyOBnP8PW*tIFKJ(Sl4-zu-vPEF11W`a+8L4SqBUefZ+J0*dAJv
zDVmu7u=WT2dclf75A`@?6hv>7`FSwB6ELar@a%J&p<(t+y>^0V@QBnfm3AK=LA_g#
zYHX5y>HnUq^d+FkzXs@i?S0b^fKV>>DYqf5gP}8{sdI>G%dK*G_((T&S+Z%2y>qJB
zjD3vVf}l-kU~?|&&M;f+j%nC}QX1I_{U>Tx?`Mj@RlDF`GhD~EQ&y}04=a|9u3haD
zUjye$(TK_>^&Q*d_2Q5?b;Na%Qn^)N?XGX9Nd~)&b@B@-)Bf0ksKkUQYjjvJ&!)~8
z=C%|WE;0<(c8!&Yod`O!n<f<3-%kj;DSqCb0cNg~K1RFw{Zn<9fT(oR!(aoHa4*0|
z4g(Z*KT?wciGAT6)+01y=PK?wP~JYVeY74Jx|(g;XLiAO2h7;9$Dxpfq+>&U++y9P
zEmCOWS7;)^oplivk)!Vr4L-NobVx@yPQe>2=r9$A+?vIpHl(^jag4s906dgd(q)zS
zwVrw~S9{iZr}5XmFPW6Y%TL_Ooii7+!8KCjKRr({DD`!({tj?r=^-h5ZyV~l(_aT5
zx4Cu8>}`(W5|>2QNcDD&#m2;2`XT<lkFp?*QGvdZpk*}Yy?vBW^aem){wlgRx);Oo
zOeR!Hr5tSfP|g!&*ySZkpcj^9c?^hG=qxKaM+ks3A^1Erv|Cu)tXW&@eFg9@UBFOf
zb|^neU@+TIUDSGtVknx0heVg1wlWLj|A89;;eZXqM=sbT8Fz~C=~f$U%DucGPz<UE
z7)XbP;J8ECrE%*}z&0E}C<zdCkpN6iHq8>bW{jICD@3^T=Kb;pihE}%0?4mIeuyp=
z<AJp6Lg>&Pj&tY8U>D5`ZsTo<3?XViRKLBg{o*jt4rW(kcx*Zk2Kx*?3Oynl=hs!y
zwZpRx6*CR@PQpt8pJAlb!S^A8Gs{OvnK9QvU`1`DwrYDtW@e#d8B9!W?W$ch)KLn-
zS_s9JNc}BIU682tgfbru?zKfHWd(P;ol;PTk<p|4%cJ`9C)9-2Bb)5Xx@?FK4(we-
z6iGch7nsHB&?XeiN7(YuC8~}12yAz>CemiB0%S~zV2kaAO(=1~kcqaz$hK^fBkNEd
zwTSsDnj&&5*y>Oh9vxySxlxa0XD50|L3O5stO>BUv+7<Q#WIx!UNcRhr=x6=itCwY
zMG_K_x%(i`o)uj({kRI~?oLoqK9*wT?!SSwZ5-<%)y$L=C3>E7moiCp%RWdUFp__k
zkSu5_M&)gmK@CL&s0t-d=!*J9qKy2gX8%1Rq3k0ka^JWa#8Vbc=(CQMHC;Z~5%r1z
zlp66-T*)`Q7tD@tOkP*O2q9DqHc53SM@bd+@(W-kVj=ol|K;8p-XpPHBqmvAF3KR>
z&^)<FfV&%^l76~u`eRI8pOe2nCBJ36=a#jO37*iH;H=FUl;5!3fLxgUrvACmR<h{*
zV%>|5MyLEKz7S1U8Lm2gh0D}btNWPp#@%$1y{g-yxmGEzPVAAdv*9b~rcL<ZBJUZr
z33}~cDWNyz9T7<p@^2x@s5=UVM}D#&;(pfI4?GNI@rH2xBt98N(TC!Ai=76>n2t#~
zD;wg2F3NJk$KXK6V1)p{Ow~{w4?WRp$d;JuQ?-V=`btrE2*-9Nv$vFz;s?<9y@QJO
z2BxST2P*_~$C?O5F0PT-&2K6-Q6l3l;v;~O&QNssPD-RkbP`q_=yn1p=}Y^FS&&IX
zehpb}{%m-^Tl~8M&6&rlUeZRTP?6@aMQMbj;|K)30KNjUN+0s$uFbP1@IKo_{#NIk
z=J03NJ*>k`VhalG$=^mbX`i)Et%vsHp)VRRpCTrR+SGSyhK$K+r<?PvlTt;p-7k|7
zsF*0rUa;O2fJ256aFdjP+R9=U+_K6E2K#T?3?M{&FytQ`2HU@G%j<eFBxIKO5OGzB
zQvNBq3ck!#T9P>jNSb{Yq8zz2&^emWR*?)mU*saxKRMB6rkjYb)yO-H<<b)A$X0L*
zJI4%zBpqT~Y6a7JDY>IhS$=H(&N+))@!8pqz_lDBjBDin;<qZYjb+TBu7AH2YmBt2
z6W?wbBzarI%FR}MDMBcVoQDw#Iz`IGzDbJspiZ3`9M*dqly#k?PHZ(p0<%uaYYB2a
z;R>+XIbb&Ndf9*04PAy#RnuLKdQn-b{PLI%;UUG(Ok>?Bd$NFa;)4k@n5mSN`;bQ6
zb=%bA1o$3T241{myrc3fnNfW=MCZ2Y>nZXdZuW5J_;|V_J6ua>4i1{=M|Amrg}obP
z_)=VYt-qge`yJnDV)~i!y`+s8ZneeczG~;7^*9q$1$QC&nRE5^Y}U|TcVBWaa=_`!
z;KUMHn;UvxEuX&k65~eGh9e@Y2d>nIm7ROt?~ifI>R}h;J<y8YSl#Aaz_`mncbs(y
zyeK^EmDtCS67^qD&Y^rvh8Fd@k|MI3usgkT&^E#Z`D+eeUD>{79|6xo#_l$&#L3RE
zFJU~ZesJ*SB)PnJd*j&6_&RFa8Lz}_#7k&Q&+yW>9C&>-jV1eEeLk@GPlUd^<3IEW
z;x~5}oYQN62fvN~*bi>nj5zhk_Bj#*XGlVF>T8^icdjHADOI%cKwRT;tP20|KoVF%
z*Vmta8fqcAo2}}Uk(pBSg&n&Wq3kB0kkWeAAOC_oT>MaWc6iRc>1!cG%DQM85jT&c
zB=vUxQJZEDkNg+F4+UZgHvWWdr2@SfcZv1Yr>B^cp*QZHuSuKsRxl-SurDhxhPfN#
zSBOlYxj-76=%=iTd0g{=)<0oQ9F?&rNM>^|ABsNDiN85+-eU5_(7L~*pWty^Xv!kl
zJbc&pEG0zm_in)WcBK(<<jP~Xd)1hrA}qt}&TQCqXp$)aEiws-s0%g7%W&kEG7{b7
z<Cb<fIh>X>Vx7Ciw5w*)Vzj1@#q`BC_{I^cu29D+VQZaC0SgN1c-Bql;lHLnAAir(
zya18DUBcmmRdq>zJN_)@XMzz6|Ej2$2Lqk$P7h$u4g(e7=BI%XSwbHaJxD|}^e~@q
zgodd!tDbe}uO6l6;`)`vUL~&PMcvue^xUemcNF@usN#Yz0K&#$FYJ{@ukrjZX%rqI
z5NQOo1I7oQz$fdOX8ZmXmF@F<k8Jbza@%9V$y849c(9LW^u@A$Nb}YDz=(L(U$VaH
zQMtk(rU&iaVVrz`b%_j!UDuM<o?ffj#pjFHhzh)&51g5+-wgFqnN^;(vhcxSLOOKk
zBYS)T%pcDZ#(uVC*ysM5XbW1NPBBtFN-yz$by92<#K5%q5BBjS2nI`F@opbf;I6?#
zW0s-eKC?*ajmXq4`Kh+9gl^$DFlHYfBjNg$&b#)pS&GDH5&yb26>oodFf-^MuqomK
zV;whbcvZ!Qt%?E++2!zA%zjv}NL?fxKim%3Rh~bY3>#Q7A6)c18k!JzIFm+;rG$2S
zV^f5OfbU+lOwFjcoaEa3xd~UvtwrFK<DeX3M#2$$YZqgyqW>ojk-#6mO5eVM>!~}e
z^~8k5)2N?bLWjKBhJ70PCYqgSsv6kve{V{~rDN?r%={-13@S5s3mK+O^5trjARrBi
zzIu%SWE;06Pch}cjOgFXG0uWI=1=fVe*qOM500Uov5Pty5?DW@{->x@yA&%G)Bh47
zyn@0&_Z>C?eHWD3+T7PQ**oJ8pL$c}&{Ho9v}M^mpd^=<duf9aL_*fZ=t^U^UmT)R
zde|5gPWXE||0*33E8frRORj&&x?xj5`nuvGioxSR{&TuXK9ojPkAUfbA`(JyRC}yZ
zEC1hSatzu*wtK)<M_?VJ!639H+x3=>AKyKcQzK%Yu?Cb&?Q$Q`z30Gm(a~qtrODeM
z91&ugf8*n_L}FaFFY55osKmQa{n5^-KTO-2wSpVyTwDSIU+~KfsKqjnO?QIb<Fnu|
zDXq}Q@(UDX4J1sF^IwCTFhIk_#>i#`$O`}!^(1q6J3BV#7`{Jb8y)A4I2Mw1kn`ZE
zR2jenYJ0Hya)`FOE2O*>gKLT2&JcS@Zlt*H<Vg+cplCu6^3r~it}h=-&J&YZ?^-G8
zHh9$nhPSz+*u&<F`GvxQ(M^P<(O9cmu}A-3%FC?hjSvwBh0P9L)3aTDaBf2@rx8f2
z=xz(x$a5}}J2|uizIA6`r2{kh$I`F+Nh?mlRr7T%9jynM@t%pYfx+rcT*I1m69T9}
zR6*X4ms{c8QItGMwB1;qD)j4^sk4oGZdbI<&R_?-!3=KqY#hLjVoZTXGyKbw_QQkg
z3J9Gx5G1(H2bd&0C(!F*?ts#|X0`G#9wA_gO5{|n+0U7Q=<DjfWm@fE|FgF@D5O|W
z57tZlCbmS#=^EP(3h36*my=w8{#MsBXbeskP}nzV?LhnA>f5{8(43`?zRG!-$6Ccy
zpdyAr3s5~@-0=w@e=R`Z3GAvJ{AuNM0z;y=EPl@8o-A>WP%k-f071vd<j;RQfOKr0
zS<7n7_aoy~^M%H<X|;iilc$pg8)CU}*`gH(=-MBzug9d^nwo+uHFi9Eu9B;n^>7nh
zx&{8=^ovR7t!)3j;(P)MhnX4t@k{WMs3P`-!jsQv!9CwikndmIB6$^$a8h<D2^mB>
z_tM8k4W7e-ERZC`)r8pn%o(ck><2Z=2o%(PmmTCS8XjC|_2575f4^o>{UIiqn!ykF
z)P$=3HTdBmS?zbvc&ex29!Ab159zI!5t@GIDzh_~Z-^fvkn|;F<VH&zjFYe36brO$
z2arf8g1fo3mzZi|Em!!U1@QEs4@(REP2^eOY)n$DT4`9<JgHY7#wW=|Znk@O`;N_#
z`ns#YsA-rMYM;yfAcK5Bfi!asTqMkajrG^9@4pP}dXUTWa<vXM-)$lB-#1W$w+jC|
zsJW+GrG)UmPKhWba_zAlTyWCulcj0UY$ADz<z$MKavnEgvQyx1Rk{L3X4(I9JJ1|F
z!DpGvwNUd5z$3&eu9KRWge5rM1_N8}kaPo#w_jfj4hzGIFPFRS5~H#JnJ1!;Y~qz6
z4){;v>9C%bo~EY%6S5Yn@As?ql-OXKSCN@Z5CF<(n^}H7&4<~#4{L4w9Fp5Yka(|j
zYq_Z1YWtw^=Xu5ztNqE24aV&;5AqamZeU$$U@L!YI$NkgS{m%@V}X<n@!J1Mlf_)u
zfbg~#%uaxfc#gcaP05kV^Or_$CnQaxTAtwK#^i@gXVZ=HB<DZ0=As*bTVN+V>n4UC
z@dLs|V;Il={RAA;3s2uzjsG_uwA=TuzsExi!$;P3!Zk!D$48{@mT#n-p5W(D%Okz(
zCkLUaae2-DvesR|@j%1lF7DLun2+Ul1!~uU@N&UW+3TxFkw<<Gxs0JS(I{mTasP;@
z0Q(~NM6{T4_4j+)7ufx+KDnqwWK1ZZ*Wxua9v8|w-;SpKPlb(I$n%#K)P{)`DI4gu
z$n^)bpsYnR&>ZVK=!$o>#%0U@;N8X}#|l9_pzw6VQ7Y@Aym%icqMKPVl#-oZw{0}W
zzh?APjH7FTdeK202u_oiH&7%#led4<Oba-+TkYt|B{F@$(y+T_ZTmg<w?|x!ZXYf^
z|4&yFb93Y6D$UmOib_V0TLYiBvr?<icGrvBOg?5|1;X)rU_HRdA1>#`=TJDzs7D{1
zFOyP9rrTB|{+B#%_Pc4Go89({nt>&C{pwWv?pLuCXMgt{-xiI1qbQ#EjEqPo<Kf*k
z^IX-DCKST)YQG;-8sXz~qpJOOE)VaF6=zp!`Tqk*X@RFRv@l3*H|%1$F*>OC^Ysu=
zO!&u#k0?1Qp*g|w_t&$)<z^e&L@M3DZ-#D4wOg{N2np6rJ76Kt%W_V(OmD+lqq#AJ
zxHYuj>k+2Uzd7-j%f^}b;T$8l4)>ekhZ%u5U;%<=g07w(e)rosjY_2&!az8Zy7Wlw
zvB75si#c_FBg`_SKacO5%OlU%HcU!j`WIKu?{nA3ehs2_b1mLz!Bp#3KHz91rtkXs
zawpC*A<4wDGBbnJ*VhLP2d7Lbc2C@fA|8v|S0bBrXuh%!EF2geAMbx$(T}sPX@hj%
z_AI>|huhxXE>kQFThF<@yAztqWJ4KMz!U-&>k*FvyJHMIl~xf=80dPji~u~<uLdq@
zDpt^;a_+TenpR$HtL+>md>v=ybTpn~6!;s4espRIJ@74f-F=}%7G0BP4tRoMrZy8<
zH<i_kA==tcAP4<nX=h_*9q_-wv(;Dhhn-Vt|E!d1CgAC+x<65f_+q{o?!WS&Gf9Q1
z1b9gaKJO=$L9uzwzZMBv^5Ed$lqq7$&r=*bD*h`qx^VvfjcG|H?(3GN<-W>yC{#)#
zL%%UAeG?cZ-8MO3;?8aF569BkZLv|Ykm(GzydKX(fwev#PaF0Xmp47%YVT(&wL5dZ
zoO7gg`Q^LOKG7%z-VH-ww{)ms1$)=CCMVZ*!tx<`DmNEs@`4)uv^<Dq$w5Iuv!(I`
zM=gb#FCzs0dCH}7Q6|`IRw1=LUs|-g+*nLTAr58nWdSA<fZs7Os^>Srn!s@#GLW^r
z9{bbT{3u=)W@bkBYaJ0WvHgDw@&Xm3mse<LXekAUUp5QLu9f;d{6G}*PG)h)<Hfz_
zWk^4YGt<V31-hb~;lw6PUH2&iT(ssq@qr_dMFLkvWgrY>#~%qE*P@z}$;Q-WFBFrQ
zn;Tm?jb2nm<tHROR^B)a>cnEXGLz%J$bXfCRyfoR$5UA}{@ZX2oq5ZTr*pz6i3o}4
zOS(W`r`%kx==%==Ys+dUm9u7nMRK(`cCA9M_xRqIOp7(T+SMedB%Q$W$Q1uLeqiAo
z2X)fQ%1VpGNk~llSCfhbOmNtmVg8;c(00>=hA?{uZ!Yx*nHr%KVGH2Az~ZjCJRvaN
z$3^9+qG|36+NPY63R|^?=t0q;V-<&x&#i5W@AdN-vx_8o{_<p;vDk}0Bh4_ioUI5;
z+Gt#*ZddDjbK(T5{6cRZWyIO<btvaJL3@u3;%qx;m0AeTA>=&nkiybo@Lc6YhNQ%a
zsj^s?26_I`emsdVJoK%x<a9jbVzIc9+IDTCuea^)H?fP0i^OK6&m%X@MhE)Wxvi%K
zzkhNySuMc=@vdqWb0Ogb%-)K7V&Bb6p;ZZmQY7}t%8tvfArY{Px*IpQd&0udylkq(
zP)Nzhh{(wy>9m_wvJt@1H-S}=v%qRTQ!CVe#R<2qbdgcVbiSUKrev?PE#h^aw%eVW
z2_^@_F{(54V^zSmqM$i!Rs(?eq}m}Ei`}DFvDwTip%ONfNL7iT)B3BBO+srP$MET<
zGlFlFxkyArBq1j^oWtk+UlpooIyN>tFqHUjk4<ho+BK5`c%l+`?z~vZEjm1t$ry6~
z;2@>k;q@r3Qm>(b84asI@LNV0DKoO0^p@M<ABZ>J)rQe~O|o+CwBJ)rL(AN~CSEA|
zNXdA1OR^u&HpNpI_r%#t+K$KAli#wh%ac)q`odFV3df%li|~*`C)~<E_66y)tfd13
z{d;6f6HHEVdN}<f6F7c)eifa)Ot{fzwLQrvIIpA5q(2O}S~w2%>BFPBb}x?L^vk6K
z<ErQ=FKPGm>7x-$!Xw0%1|5NGo!?vTJE{ZToH+@O3eg&+Gl%)%_y7-&7e#aCzTxrV
z!%LyUflTl-+5>xy=^Zq9<-FWbkg`kFM(<l*glL|Qi<*wb-yyMXdMs}6jEBKp+0=W-
z+g;>}OBJS{Qi40!Q1>h))35!6b90(Aia=kc@4)EkZ?mk~3yFZE9%*3edgR}ytUZVI
zwO<PETEJ4kq_SbzOgc7fwAhQs)Ikp(;R}iRVPJzV3kmswG~4ckg}LGomWe?>nd7Pq
zba&DP4{Armu4cC(xqevV<2@abeM@92D&g;=sgNeYv~b1aRgNa?QL?Acak2>@yrUuI
zNe3fxw&?4NHt@P{hqGBP=lQ-rRXl4HMg^I;)7oq_Pn|B7ClQTLQ&WYof*t^;4W<&A
z45Omr;_?O{>#V19IhVXmGS_(T213CTi8V180Vkm_R;+^raeO8bI_!Hw9P?b?=D7k<
zxC3B$AoJ7JMrred$Yid-PV3j(1%js~R)Nc+zFyB+uAf!W#Ph$NFXnlM)?SJ4Cv6fw
zzAYUmr|5rt-qHUmYnbQtJ^CclE^5YESYrMinvnSZvmp6=+KBlV3<c+(lx1BcpKVp7
z?--gu`+OUwZO`WSvKL-Bl7gAZ<Z`NTaJx5f?(Ip>du(hxg8k6_J{BRhZLQON>2^?k
zm0RY8A1Pq;$6wF?pisW_mvNyUJHE#IUq@0hq3Q{V8O8?YhG_K`KLdDOzqI<t(dohy
z_4o|?kH7ZaOuMx>(K|+|xO6iUlWgoG_AK<8Zk0@?Cz#?BZMmz5oW_T#6w$v+b<BCc
zdNoMs>1!nL1l})Mj2Fukm$T=~q9~KK$pQv^R^Ny1e5qp~;4sQK_(xckd=(uTowB<u
zz;|Z-HOC0cq06P{TX#W^E^v6<s+fbh0<!o#;C~t1;HeT#sb|`;qFgd1=RoTzXU3_Z
zcG7=HAEA-2kb6Aaf%n|Q2Hmtl^$CYl4;5o58urcbh3{T(p?s_wLt~&<j~{VE_B^Jx
z`OTApuT*QlWZiTN-!k+Nz7(e^RBLmHs8yE4ni8ZH5_;%-9X;RccuW$06J9<?4OLPR
z9P|&L>~E`alzS~lR0XalG~J3_Qb5DP4gby_p?n;173vduA%2DWNNPc;b;FBeXVb#v
z>*;Y)Gq&o7mpOG1n6BM<xA2U1$IXfH#SG3u9K@rm;n9jEw3mr`M}8zeH*y)_Q6!X&
zf9}&1f7h~5(Zx->#1a#vn!%9;6yC&=l2^zKaVMm=ZnN{e2Grfvz2^!?V^Q}S3c?E;
zjDnEdSx3GO5h}&D;c`0U!HVZGgybXQUfPDUdr%)jnD{aqjyBEccVc;bc{(G)lxVgP
zJ&W5x&Q|?mv`SW`%v&4V)AoK!7aE%8GUTF9g*z#*FqR{dP@eMCdH43rtQFl0jZ{>g
zBWvZmu3<Q_5Jt?+&JA`Gs%9)Xh6c~S73DPjPJ7{UgVD3NgU)>U9cgFyl~A1i*BTbz
z!UN4q{lYE<w1c^j$!kLT6wM9Nu9|0(Iy8MU!^34m<Om5@_k=yf1A?f9@LT^x@1xs7
zg%7BE=9utc6{d>Oq+lPfA@NZP%UY~2YNUS1<x56j0jYR}(A#tC=jmJ_`s>|u^mxxf
zEDxjICg)OFYxh2W6*c7Bn%N{{xDQj||3}kRhDFsyYYPyiLmH%U;G?^{Lt2pT&Y@E}
zha3s%p+Qol8)oQ`PU#$wZs8v9eeOU09A=*#>+H4G8#S0drY0lAc^>xS1la@F^#)bV
zG%YXDMl$-hirfftgPc@h@&y&r5Km761LHrO%Q~Bi6|dx3<@?oK->zADuV=QTy&fej
zHuVkqP4ww~A#d#^%!c&S9?T%IhM-z52nhYlA?J2kj?hfecEnFL$)|yOS6$d)rN(;^
zv6l%)>954tjD>%n;YNLxflT*viNKE&Hj^n3+?*z<n|NjL`FwF6IXv1yj+zvTcjP0Z
zCR&-bsZZeIDVCz%Z=rpE0F45|=i^_t!GZ4H*g~}FUU8GhTA@EbRH|cOTsJ=ju5HAA
zfAL(iTys!8Bc8(ElOG3JTlB)>Oo>DNsb~)Xs^s{f?oD)q2x?yTKDNhp<xINwwCVx<
z#|IYS&U*?LH6xfHX;yos7Wggx*UfMp*d$A7;@&%j1$(~|Q>~Pxu)PuK!dKBwBUjPH
zI16vRxZ7qyD8J4)*_{x>uTQYL-2bBzU3@#D%%tDj%m`8P<Q5QnIFXV8k7#ro^`KFD
zi*LWBmEahDg&yDQLm7Bu+CVXw*j~+IUV!-Q+^!wK(40h<(f_r1{3eZ>MJrUxH?GcF
zN8f5_#w9pLSNSjO{p8@V<OjjWhVG*rzcDQ^onT1;?W8)zuR;&J6{?_T&TXbwXD!5_
zY3W4m#CS%}c7|8hYe`mKrGGHksO7=u2@hoihOuis_I2#0Bt|Ru`7|vZJYzo;`@y48
z@K5UCmo$DGR}picii!)!I?uJK`GbU4Ho9Ta&-=XPz_ZV@?*>&|4eDt5rT2!}I<**B
zgh1&-kv>Cv6`P5a#bZ_=#|UM1Ka}Wr6WKR-^ncwh`jh76LzJ%u@?{!-2evLUW_OV5
z-ZrOy(hAZ_Cha#tv~skCy_zeB-p=V*7^qF9QZsNHQVepP_7fHFEB5bl8)gW2pIHo$
z#bSM^nD8)>muEd=H{G>2o84K@*ukb==2I>cU4%X<Jx*RA8hkNSC}JR%)kj79f|60k
zEEQS=hH%lirQ14A>+Bdvx$E+n?%M8AthLM}$}IujUq9EU-ZZoz>C%LIHMad|e(D9>
zi@D3=mCs4I^_m|~7fd2wL?6X*jeT6;$U5XVIt2s|v=;v3pZ~U&K497#kstK*SkCv=
z#F+>!qdronKgE_HJpIYejNDW9<77OUHGB}4P7u_RPHggvj!U8h6C{X7p)xn6TK@(m
z=*cnpcNDJGA|nxYR*9R^pjv+{+N!Q-JyE|j<`tIK^X^BZM{j4~gk_!a86}?;cZUO<
zxNF}FZsGYGs|)v=y(^1j1Wbsbz_?G~>1;B7g#-P0-T1CP1e6T<JG!_3DeWhU?kq{6
z;D6l9*hMr#m_#8I1e23iB`>Fv>$!SR1Av_akV}`r+pl;G{s3f9FkPq6AMO%$I)h52
zfBVCU!?e%<uH&_2g}U6#*M#(x+cP$LHcWidIBEBi{mdrv>f>8^m9JB5brW)OQyv#r
zM9_U-YnY^rTmf0*CZT1E=+s$0jlj`r%p<{Yitc6x19pLjJlEEPeV`tuS1?2C5LS>6
zldThHTUdKFZ_NE(%r}712QbTGoNnEwmfJTW1r(I5vQ<g3Sk&K7y#Kc1nq?%?HLH{@
z0hz)Vcw=6^*LtBiGCI3$S4x;hdXBGW9X1>XHy1Mc=?@0xs>lj2n<~fl)C5=N`YTgQ
zkwV2uT{^-ezz#%}e;n@t)U7IzYw5GhRnLxF6RtY+WT&79BfN&dE|EqU>Eh*2$u+ta
zDFkvGvb9fxD#=Zp4^Mp4h@xTNeAk%${nyB-8J=JwMSU+UTLNi{55LO@(q360b=xcc
z%51MpPT00s#q19m>535=ZCp&Yf4JBC`U5pZKS3r`d`efDr))iE!a_IN1i}373(hy)
zYNIDCw)}!`{IxdT=)39fC)L{p2v@|O2tb0?A2~xIB!0WEF&Gsq<pe@S@_CDopYVbh
z{)P(mVH7Ep34VALi>JHR&`!0U&(H|csITQ?Mm6_r3wz`l>(0iKk(;_a&8}wZ@voYp
zm66xX0BE|%1??#o{k%6@^NSmEU4(PeFb4p^j@M0DphZ4=5-`;WU!9eTd_02D4vuD8
z;^n98%~o<KOzy*I&CFGb5I3^wal$__A?alUk)JUUXFSn1ck6neQtR;8eh|mc=8vkc
zX$OmUKC9&Yxi*@<KLi#TpFC{g`}2$e>4W*w?vnM_VP-hFFxg+M7XdF%9EKycPzlI1
z%{zO3?;oZXxZ|yaAPH|+j)lT*Q7Z%m@*`uk1H479m)#>o2qNlvTy2P~jjU_fo!LVp
zGcKXjbpPaWHqDOQBBJNrW8Ff`Lhj8fDf~<qOq$qGDVDIYdsXL^gv{X>(8<|VqR2Q`
z3PyFK&cMHuZdWI6#&Ny%)+M_f#Jh-vJ(5<$4|xpLp9$SKB#_57q^_&0tK4<}k9A>0
zZs485(KzdFA`cwbjn@ydyB)}1(3>N~BO+or8aPhvW`H=~cqptSdkG^kw)qo#&cPlp
zu*<&k)A1V(E(B+tj$0}MLtXRnKfOBXqfmrN9<N`sK;N+?HD{G|i+c`juLp5OlXjJ}
z*MjrvM*jfO*}>(aJ-?>%<boTf*HFOON;AHX`)&agXOR$}!KBQfq^Lrtn~?+e*3Hol
zkm0GiU3*Es*iI%k5CD4LJ(@wQh#*S?glg&*k+Be+>)_q4XRcPaYupiSyCakixo8Ns
z^JO2bti~Bp=c}G`eOLMS)oh33;na%4&Zw#9fiB>96}mPeWH@j${EWLr_WIAxmvldu
z=kT2CXUAl|jnkKC!h(0|+{pP?9GLi%t=%Nv_f@*JPD{QlHy&R~YcNJ2e=%{XXHIWL
zG{cB7M;>-QTUc6|N^Ecz-H5iWV}k8}6B-BpCEUQKfWt?QJ`s14<D<s)WT&mfESz^B
ze#U=xW#_{Zqi2=e91-Mih41sjN=WVp-FF1eI!=~J-j83Lc=~PjSZ`V^k<sFL#|vN<
z)wiFno|8hg{We9+WNlm9ITpVl)<ip=9{n2r>;7BYekyuKJR<E#6?$*Y!otYw_7g=`
z=Y%iICOWoV1dZ&3SXmG5y8wgT;I#wlDq7!mIruN>5Bojcj+=MOD;2ZE(n%7%F^*#m
zT+5G+fTCL$aY=3~P!x!0Z##vPMc^d5cB2px1NoDd<%8dnV<QPVIy&fOWPO`ju4n}L
zNMhvno@gN+C^Xu@M2a<dvRs>4+p0R~>F@Q7^WD|aXA5ImISBXG3o@@258b4*$G`Vv
z92^{QK$68BOtqYAU}JuJwjII!PKs1N$=v^JkP!}O7R3X)L?r_q=?x*fuIHnCL?6=N
z<zs$8*-7rYxASPJkuPi+liEazbZ^o_@+@MR1=?LCmp}<@=u05|=PINonKSe<Ofr{;
zeazrC@?H<8L73=d0Cg2o_ni^--`iFsKv9N^l-I#VK9moT0MeuV;p+kGaPdFt5{G>y
zzt12t=EfT4h%gFjS5MhqT|FSyU2zG;W!qL8xYMt5bkx(;3|wlrRx|%Leff<lko{oL
zRnJI>WyFT&+G?!L2fnFc?K2T~cX>$WaWKt6LraTk>wQ34w>oy-BmwC@Pm$FsG@ule
zqI|Up6fAWuuWor97pD``Qil?0EnF{_n`_$p;qdbNS3$(jc|zm#dF)AO|B&0#(KC=>
zi`~>vV@Iu%9avA7YWEUPT>G0FUx;p=Cx~&C&&9ES#FX{6Xqv3D{JdE$_Qa{f`q}n}
z)6Akcaa`iSM_6F?9b8URjMMB2K}>Yn`Vo3RlXytu;eFE)In6=YvMbTcxI8ZI;SAF`
z(wB|=Nued=vk}cWeQ6@%(<vn>sg)t@@?_^cRiY-gf5jke(sp-H@ozyKoF#)T7guwF
z9&c$7fM!*_o%4tr?N-@P^NB(D2nMw!3=rUw2?PpN<HRGuv~(I<#5BBdH#goMhR7p5
zR=Vms^5_Y5#|2SYm~4Dpv->V74G07GqYUI<am_=8wc}=k264l76n{O4&X$a#pe9<;
z_<V*GQLkY;@BDHMTl^hkg*)ZP)+}e5a?vq$aS!W1kr-oTXE!<<9i<&N8HfC9#-NHF
z%$khs`wih&28h2W@g`~K|8)g$e0ZLMK7YSH@b;@a*>S8hX%+1!T?W#%2h^X*Yo7jG
zXXLl2TV(UByGxM!4cCmp@Ox7MLeO}E^QVrCTUqKGA<EFxNMSLMn4fW$8Wg6;j~C|Q
z*>U8J0kZ07MjqXwZpaf*hAPJBSrnd@dT8NQcl_mjC{gZ4IrAJQ6DCH<&$ppL*w87O
zaar#qH~vguYVmIl<k`+z8tf>LF_l)!!T9sdN%G5dF3Cs17(1nvI+Ug3R&11Zzn8nY
z*v226=ZZ0Nf{q2W#maGJp_f*M#009BqYgY*KJqY)FFV3$Ik5qhaT!|;X07X&+afeK
z5muD^kJ}9eG#k#yD|j3q_VV?!Sri9EZ`x`ssnL3=cFDWNpSH1tWV&Y4hS;#(;MJzW
zr}ts6BJ=DKte<}Vp^7uQz(AcrMy>)qKI4&PEDBXA{XhoUk$~6n0f(mY*|9iV6jAT~
z`<&<WeX;3!7l9jKA?9T7&trF^C712%F;T2d*4BO@<=^bXdu-$XpMLld&q|BlOfJl4
z-qYI&DAW+m`aIm6Dkec?{5Bi5Ssb@#R=rmQs@e;$WbI=+T`7BW5TU!>>))=Nn}R4K
z&lCBVUgVnpR4->{y}G*E17TrqH`m9l{b>>zPeH8R_&!F=L1A_LC9lW7`S&>93Q^Z+
zIywKqDN{_4<o#7jmxo6Q6qYYX`$`d%T!C#@`V6msD6%~K35syNoQbFnXw^26n6Bag
zmzqYIF~9J!kLhEVR@FlZZ~D`5ANM=(e)P_Y_*ZbMM&7Tg*X*6^37w<09T_(~r`VhO
zrF`*+)#OeO7v}3Nr6U3G6BTTdsg2GCMP&<;1$ggLE+1S*jn^$sxMI5R^W;?gx1R~T
zSR|)9(a}DqvGMuejyZHJsv8wC%rDDl;6Ua(!{ZWC9(f=MJCOAHs6}(4VtxP7f!)z}
z7OE?0)T|R7EK(W=!-wWaKo-_;YKV+qIkhF76X=J(K&5cM2ya8zejSUcKKrTel~F-*
z_Id2d_wmM(AJ`Rz{u0o&w9oU;*`EfCy@Q^fTy*4#wzjt4PnDB#VJEDJ;<i^@cKw$W
zP6Fhr#oyVwmri={F=czO!|yfL>7(sW3}BHT(UNpNyqAu=E)YY6UOK@^9x1;6>haOB
ziRWK$6D8Hg&qS8LVBr$)hQOF}+QVpRi9dTSId)kB+Bq-qR`w==POeok!SDs%=RrDI
zY9YcVq}hDmP#i<GOi&ye=}O33yGvO}Qm9F;c;6PnbmV6=f2Zs?%{u*vQIEJ8ByJ6d
zeRtqB*7MIvKi;u5P{v0<jOL6t($-c2a!r+|geI5^BDLQ*3_^PNYQ@U5j!BP!{<Ze=
z)mg{7$Y8>8BS{$qf6K-?eO$k}^YnhnOKdR*!Z)S*IlhYNG^5H?+h#M8ca}AO>f#ze
zeuI0(jW9KD>>^eH^s$H^j@~<6OHFxJ)p&JjpJ-`gFrWJg-D$EY>HjCqhZ_Il@N-dD
zwcKOZ(!FSZlht*ewX~_*`1k$R6aNdL0is`)S=kZy0Gq@T4Vqrf2DHAzrRS<XUEzII
zpA?9O#6X&l`SmWPuz^X+9HeX6vN-peDbM&a#QKV{^>+Qlzs-HmU3l5z)mTGB@TBWF
zE~c|wP=x4A%O)1uq$Poc=`uujow+mMXTOw8WbhGBWN`bhyILA)KDNLln(kVaKk$W_
zZ4?vJ&u}N@c>$8N8Ds+4f#DZ!YIV%(P#}YWJoAYXyA3<n%RKlV;AbBmTwIae-e=pw
zX_8)SzdN>LyacxNOkvDf13y~jxqeF5J2smLrpH~;{<BE4pmg^3O#5BxH!_Uglarov
zSb_PDDX3*PqrFeymBSy#PhTA%aW#>_xMIvxmv7bMy`l=3l?ht70uo5AJY7f6U9|h&
zZ*dW|MSPY_G1PvB=+Y=6yq;WGP|A{)n}T)-WL7=chl&bXspc1K?!9Gh*RgyyRhQuC
zOXQ0}m^|^blBtb02IAi4`IcGOxtta2pk#0;30I~*O&M{v)gTE5w0Q3JR@I9SSN4uh
zCLNU$G|KW=D?B=Av_fo)xR>!6>4e}9?37~=C^fNC-zTd0H_!gsVvE}B)Yh+GObXPe
zCGN?`Amf+@<q%ok$ke=j#R>nF`jQF8J;`azQpW(Z%;I)$!2^|T#CZhKS!Jpc`Lvi3
zphYBTMeLbb<;*<g!DHm4ISu|R5Ln9@x(@^NUM?nGmyt!1V7?hDGpMBX2T${?8mM~|
zD&w$eS3$h4kq=5rhUXq#H)W%M`;JY4u--&!gX#UAO^aXvn4E1G+0Sj54dl>>=<Kq+
zbn1QvL9<2fHCV#PRvv4FijAEcIS2E-Tvq53nPDgk?PtubaaqX1;lXwVAZ}#EhFgF?
zb!lnEhR&pnQq-f=2c}p;j?qXr8WK=U)lGrhn<Y$ND@jUlJ}N(yMCnhvP-P_|iQW6v
z@jVN9x@}xoRb9X@<vtOh!G4e!s<7*4;qLttq}*g<20<!@dUa;TPVlU+@Pr^$N@+on
zfYJ*`4kQN92?~7R=&FVDEt@#)nz1j_GO%6kZSm6{7#8`;lp!Ho^y{}$f0%#y{nV~K
zNp8<AMbaZZJR;H_NT1MFSFW267Cf$6jz{K05uP0^F;!wtwrlC5ENq11(w*T!8oZ^O
zHceEj@}A^5-Q5l3fu_AV;u&kfW>`N1b~3D|SuC<BlZx3QV}u#88MNQ;R(_(6*#i2G
zeXH$$GhOajH(r@=Qp9uUTm)vgsGzl|pio;M1XeFG1ziH_3&$P}904tgukxZ&c3f+p
zEykdha9Vf|YTp0uyM+w8rtd-+4r5z?Yr4y(rkc(p<A_7y{r+I*qi+yI=Bgnb#OYgl
zRl;$4p+B(PY+B%vFl8|lil@w``gITa(W&;@<0$-O3Wa#{aejIlYa%jH|E3{Mlav)e
znDjfmKKVr6b0oQ284b8oX>Y-QyGT-6g`4Q$L+Z_CUya#H(9G~ZCq8L4c^-iF1cAW{
zVEcB=f@FZ$6QFLy?9`$$$q^U6JvK?i8kUyEXG9bZpBJ?5s%cE!CD>Yzv&%lUIa-Rn
z0X7yV<QU4k855C4ZWQKq6%c{QT#0w5$r5d067DePWG!2$emz<AX42It<{;3N*zn1-
zh}|Qh(r+WoLFybD>YAatUHr{SqdwA=oy|P239@gVUJw=P)*91d;!8h9HO$Rxz@L}J
zz=oFSp6q~Sweyqw!XCSDmLJRCU)e1=O&Pbq+*?5*2C`0%&yZfs>}Qz0P7kQ^#43Ie
zX(ocAAmiH`y4ja;T*I|#gtK$w7HT^cUtYr37KzHyD;+1mhNkqZ;JdwflGD{kHumBL
z3jPvb)A8f0xD{^$<vAt0Yy7Dhqz8TJ9wlMDKo1+a8eh=sjuQz;KI5RLF0j3A>|}vP
zazQMs23M>rwQVhc+;>|9CFG$GyM}4G05A*n4CT@{8E<2Y*%9QN=i#y+Ot0ZaE?>0Y
z*8rMT*Db%U;U{=Ya=Ubb@Pnj1;xxoygQG>%;Lb)G78mYy>FnSg+-MFgTz{`psj|Xi
zRq34Mf+N(n)vtrW_Td0nH8SRyD&0%TDG$FpCmyfLH*Bs<K>Y}QoE}B`ICH8Mi?(*?
zc-3>DP`5$^bR?CpH|UrGcO{d}%Sjmq&FjQGW!Z1(Q%E=iFsG^~er$8c0DuZj+t2cz
z%I%UG(-Z_n&#IG3*I>X{fd~=}DQ1oy!(#iO3f79mp!39|7~NFlRPgLbvlV36E;4x-
z8+?uD?gl;=O0?YSLM(mG_<GU%pHlJu9Wg0d#$1@Rz@GuLw)Q7Y0E+<JV1Yqv-g(CT
zz64d3isVJ@`_i$ghi6ZpR6rQR_x8BxI1`JKF>m%y@EiNk2O&N%8;GC~Q+i44?QWDD
zrH}OOYPt`U>o1oQ6IW~<T3Ux-pw(mfH|GvL$MLu>=Yf~9#<|CNOmC8dT*Vlw{YoiF
zdlc#vEOvA3uR0djJPOqkysuXZL|dvmQNOZ=4>&}P{CtCgT=<LRDk6UQ@K?Je1=CZ0
zvcKzI2X{?iap+*F&=?1o^b1j@NC?hv6K@|Skryv&-!0Yr=Zw9&f5UPhq&(M-f_5Tx
z_Ck2(kvfaQNFFRO&})q@W?}>T(lm)HC+816!6((KD3SaG$bx;{sqX)$SfDSy*EzQ0
z6Tn^0(6;qlsH$z_U;{f}MZ=S1h6+kfg?D%T+Nn23AN73-`4PM)fJ1>H<@Eu2ZEQ%{
zcjJId?U~roL+KtLqmi+bsMi;!cs*)jC>ol81RuA^b_+ARJGwS0)X2iZ^v}WZ3ZMu5
zjBmvP(or)^vU_uPD4sycE*&1kB+YMeVsddfr?$d}npBUKAokd#nByD5aF#1x;AeP{
zoSqovexDQjBMak|_bqkT994#ZUb>)1xgvTXT)>)!nO3uZ&h>Rhygy7_rFB5>!CgQg
zL0cEp;jqA}ML@%NZaQ8Sj|?Gyb@!@E?s4HXAS^AHh}UT`Y^ztn`y?L?8noba4=6J0
zL^JyY0n-4)Vp76LKoB^(|2EnecY;W<KKC>hd8EVk4wb0Bmxe3F8S_y0#hvmkXMWH)
zgLp`1KB$mwoCJ_CvuMWoM%|fTXTFPPEDBaRVKZmQh|Wa(mt$VdB2o7b4ua6!b;w9W
zIZLom(|ujPtFb~j(44fK^|98w&hXR?=$n5Y6&emwt(gj4<H%ya-i~j%JS-n+6ZR6N
ztVBB;uzBJ6%T`MTD4Wu)_4Wa^{;8cpIsdMv#1mv>m9T`z7=#{@o-YQqZ8%b}ny1z_
zLhJ@A)M!1@Q-s<MG&=u=096fUsNU)t599tB>AXohlHf==VJt04b_TdgDZapaOXX0G
zHa5Hls-eu13eHwE4!kWo(EhX3Sl#e*qq~M4i=*&tYXg+%B7i?1PGnSC6)ig>l~1eM
zOP{B20wCeZ{b~(mz;uL}BXk5;HJXkbI9Uwb0o7G6psWGbS;+pfpCD3<&W-ksM7hVA
zQeOZkVuoC5R1@e_;LTbXCCs}W@)1~&FdJ_dcZYD9bJ)I&ylH-Ov0qDhvC>>enI#g(
zz@m`mG5NnwO0p037zh0=dR<h|BODWIgn@Nr?kuRCT~DdSm9^JQ<8O(0`Ep8NCa0nb
zSEBO+l$EfW>t&+s%$KW$=GR&uM;;}Mx+XtdBsIU_9_Ek8puk1G`q<uqRpT=upccNZ
zh!9qhyDt*&aYeC_J=dT3F$?_Y(3s4cRPVGdIU^EOKx+fQC7^Gr_$P5M7?V>MT-O!Z
zW3l`y_I1I8^XAHRhV8x$nKgPaI{>ZvRvan`m6q|E)w9+sB78eb4FW0B+n8X}258WK
z!RFUyYV@%DrH#iC$x_!_D7^Wo`ik<~x&ZlEsd_`+U{y$W<o5ONAl*Ptt3X$%?SSh@
z=i^-k4R&TJ;>BEp^KT;$v&0hv6jc7H>0;6yeh1wzBE4EY$d7K8Yz^>1KqdqwG{ywL
zWW`7uY_nQ^wJFCJZ2~{$zDAOp1PBa$=|)t&=ykL4<mxP7miV=uD5#s<`G1dv<l~Rb
zkKmyCrAo_$7Qg43n4tSt1fKa=iT}ks$M5o>=wJkwqDm6Ljk%5q>^5=lW#;XVZxIWw
zR@q5h!*y`HGX(I0LkFvyHe$^028PIx>RCcwRh;?b{F@jRKforX7=OL0Fl5#}EJXIO
zTr5RwX6sq~aZ{&jVd}TQi@)enQYK7HOw3D#R-{IY4B|$Ha`<mrP-<jWXYqwey8|#Z
zicvGo$@hBWSpu5vEc4x)ihwGtY!anVhC%)>9#yP6sMhnfCPh3H1-P~}A~N|r)CK4p
z9$3u09-R~PI4DhgE4XWjGMlZZNBn9uEwV!VtVGec(JA{)>}2T!&>k2w8K>Ue%9C1x
z(kwzS&8>UOLSC*dly&DA=;pjd94DI);~aq!A4sSaWDsc?vfGgit!}SWSXJev3IE~5
zpqI@B5*lXSgJMq)p}dZZrPC@Uc5bdW!~p?xD!N!hyToP?GZVZ|y6|O=T?%c6`kAIP
zUz=2{Oym$QhKh*+u^eL}oxyWFvQZJML@W63rw!%Zj(OT)BLl%5LzFu*K*YH$>fW0p
zWE-<ws9d?+Qa54?c)TUFfV>+?eDNR0*HJ=jdguRy=!u(@w<Yj@u!v7(J#QG1-5AP|
zzZ=<$-HPqgzFgzkQ3|F|V_e#Ysn<?;6_a<lR(Jlr`gX?L;J7qvc@Ww`2Vu1Q3aO(J
z3>kcDlhew1kLK4X@W!$`1cjA{S^>;$4v3!nrFo{?WDx^Hf1afj5+**g4*D4IpK*tQ
zNSr0bl$<eP^fL5zyEM-xvwoA2fdLNkz~oPj4^-Y^m@aJtrW(HXkpT<+5GczH1?{{^
zw8)5>qNG=!rJ&^|TTrO69;YPq+6B49fnLT6Jd8oteiU3R)ut^%+4$o(fh|rFA}WZ5
zwpqgfQFWP_0F_gWEe%7G&Y})CeI|uVevd&ptms_v*hb}aXG3d)E{5HH(cRhbfEdAO
zG9)3UINZiZIGDP@5H~;s`#>Ib8@iI_b1ctJ&Um0W50f<Dc^E@oLB%*V^-ieZ60%Rl
zQ1P@3!zAM^J$V^n)l@DAf%q0Dm|4F<J5J*(wa8FT`H*Lj(Vrd@W3hQKedNEOZpqm!
zX^Y2Vqs8)r<TGlXdFF!uK9iSa8(}1?x}nze%klH=J)FB^19YRHogMdWcHWS!anmn?
zwbt8V46RSEGu_CY6;qNVTC38jzHDSkm81)1OQoxd{RP{r+=kdf_Hql&SD_|tRK0!%
zD`bl2&`T4xhU3yw{|Sq0s_qdY@O17!0r_q(6nW3MGT4VfU#(OUl@M#R*Q4^DQB7~?
zNUr?r8JD3Hgn*!DX`aIO-T!|!GW(H3u-FjAQ~u*9g)?jP*$ez6H~&)R-4c^O>@yVW
zb~2U;ph*ke+dLsARa|nBF79nQ65VZ2T7XG<alEob-yZBBio`7%a-SiHeC3y~(1gr#
zorxk0nG4<Fp7Lo{+y`d!J5u#I(qnUL7KyVKdJH<5SFTm%TGZwUAkXhEe1<G_Dax$H
z$9I_qMhh^wLOYbN;tp{Mp>H#R=Psru(N_SN43EVip@}aTm3>N1n-WH+ipotP5c5UP
z$UuPYWi<+-tfi$l43u&mu^H)cP%`vV#gh@eoM<Ov^CLYL`sTM~LX*g6RCbw@PxRE*
zdciwf0beovOz~dn<^$4-L)L>txabJ&ydjGMO^}mFLBScI*`g34N`2K-J2#~*3c3s5
z!`>jrpfk|zJ$`SJmBDO&yjNkqhp8#nEO?bL7mEI3@4%&KafS(4O^uN59_|@_<ruw4
z0lE>wJiWdqQLFu;%yO2byR6)hO-oy&Ry=ap0cb7+ADRC%z)Nta%1!(l{uPg#goI+e
z%XExk8#EDSwfnl)aF45!#e{3)?U#d?Aj28c9^aam(dFM_hM<Go`%+t{mlL%^&%YxT
zpGRaQnxxN0m30}Slyuer=ujQo?PmdLbR-ObOsva?*-*s;AgcuLhY7zL{ggS4<nXct
zv0wuAFx6*5BVkz1W3qqB8@WgGq0G9V?LOZoMF9$IElu)VMM9*a`_~u<U2l$_E<~GQ
z4?(xkEZAKd!O)5yvem;aOGj+A=Rru^G&1=HgKK?6ABsz7VkK@WDx%801*-8%pNnyI
zs)}o-MeZ{iQhbHw4<0v}%s)eXuG^h~TG5Vj6#H7EJ@X@D=Hi!yV)MwQHJz<n@nb<5
z&fm7+aa$hMkVh2QGy&lf@*GQK1LH2sAXB8K>V-Dms7+=BcNw#g%?=lNgxx$e`{hNo
zhGmP{>umfsR63)GH`?L%yJHjWwe-O;763GEzRibgOzg?u60doqyQp)p3egW3Hh^pV
zEYYf|gwVZ54S+*EmvEWqb2Q|8;+wd4Q>yORaz?*4IHBAAD$yX_CtWYnR<Nbq^;ekH
zN05q*uF9%A$RvS@@U$oVt>L12w=a7c2r=ApmOXYQrSzp2VXz{F90bDN3;yLuIt04C
z0Xk~2+eS{6VQY)Sr^T|6%c9&NXfz(u9zID>Rf*Q!lDKDZBz0k7Q|BJMPLJd<uVx<A
zE#3#NCui(B@q_#%!UnjZ54Ok98M<-*nlfoyLC(La6<xmIy$K22?27Z@)BgS&`G}n+
zH#b)`fD3}gPmTmc@O46wh9n614SS}rKRZ85TyeU)iO_DC<qk9jMWxx?<EIN&q~CEf
z;_m8qH`W`$ZQGbu&U3U#2#3HkMpQQs)MTIoS0;_##z0Dlt>XEP_+n6ioIht$d+I$N
zhL`CHJJ)+Ii!aa1s@u+D4xkOE$E|-hf}ZYkFTY=`ofI+V`fH#pzmwBLz&S*doSsD-
zozz!|dkj(sYH3$t^)a0OqE{z@AN$hy=iq(H6?sSZ1*m1Ke8_#v+9cjD3tZgjNsc8^
zdVDA1ib>>u&jD5!KHS4^5PNlyzxUU=zne#q3-#J1Z=_fq>eIHltjDwzq1#zjr~gb6
z4sZQ-v#PZk{#7SLdw?>z>IVg(s7wk>q`Cz=7sU~hdZ#N0RZz3&0VsGL%wNRULY5(H
z*yWla%^b908`Z^pVPGuZ-`von<I?$*+c5~N-JTao)r6X_w)vD}DX&O343pd_<SmO;
zqs$~{@5B_2-Yum-r~J2!8JtX33>TP(A#aQ{X9uC1ks<&$?iC%0>=GBBfzL>uvHk?2
zFEZ!UG_H+N9{9E@C~H}08N2U}X*VzZ!-n&`y9A(}>W?DujewJGG>~%m*a?EtPZ^gi
zNX~83HyIs@5~Nm8n|;H5E(*XLW#X0}HxtBktcVnLzo1cmWP5=}jegXPM%DMvt|mX8
zgiDs<TcOl1GIo<-8mJsqfR@MzT^U=)jU+DDaP>bM3w?~5Z4PgLP}c{|Vy2!g=T+}b
zOc_;-TOXh^I~^M(eR_9TU2(LpR@6dYcKT4Pnq59yf1Tw~q&9*jP`istEs<igaJr?K
zoG<p3jp7bRVdCm?uJoBa?Ul{zX5zCol7jb@cyr=Wy*M;nTk+1&^c%o!1;a-ilbU^V
z(Y~6es^tJAMTxhWVV$FX7>Pk*rku&G5U}HnhE3@J$OGTQ0Tis=#olBuz#?hZmpd`&
zujH0K)uRswlFVvMYCiLSt`zt2e5*Eb6hb-y9Vb0+<8_S#v;%g0Ir_B%!d>DHt2>RL
z=iZ^Hn4EZccyWMrM)yD0BS9)^aav*F<(oZNLR@Jn6AZ`-Bhq&c_#`#p%|Iq^VPQco
z;;ilm2N($sucONY00)veoQ8_oQ-fPFY84o;WD)6S2gESB1m3WlnZP~^o>kw&{~Fry
zCBgq1%bFX_x^&(-XCWFM>$kRA<id;AavmJkW8~XNuOj2<Vj@C)g;Nvm&C-jl=Q`yo
z@=_{Qy+<}!8qo_MN=lvP?*~p$A>03&<?p4R#du-K?+ToG(U+l9MizgVA(trW$Sw=L
zZ>Yu|)e0<@G(G0WdJ7u@I-%8C!tb>yp=8?2z*RYwDmBB*#!iEyE^Ss-rC;!uMbR7c
zl3@UW9S_7m^$a+7lI^30?OUG;bC=wQ|Fm&c%raMnSr1q1t<^5SG3IznS_<Rh=_rn&
zMlQI5(W+HW?G2LF8Z2D+XXv3nW&69$2O-$Zk*25BM|^H5H)|Y@_`Rb1Bjt4X!&jlF
zrbP>Ie(}4B8yEL%E@rmqkivdPL_Ar^fckVkC%phTri1_SC%=5~1?|EbuRlBb@$e(&
zp+fSutF7A<UGdvshOL{^O{iMj(l0(9v%c3+f_*@rx>`PwjR;RDNUmlsW9i&qe&Asg
zU=3Bvr`)goyb^GC($B#*#56#-WgY?zf30HHxl?j!Ly#B%H3)E{dUY)^l})Fx;YIJA
zb3yydL%d+Nl-QeY;WWQ^BhNUte8AI5ms=i8{=GY0P;`wR^Ndm8bR8U#bMkyg(yR?x
zH=KF^dx|ZO2&37lI+zRqFagjz>cLY}Mu{ZweCzPJ7l^!^nGyY?j4Sp0w?VA%{i%`)
z^CW|^V|lHIkrs8+<d0>j3mH9|zt9H8P(ap8*n{?}FR}m*%WB}g<Mur)#N4n6yPU%_
zwMLYMvH2|P_1#kw8;I5_D8CwVpUYt$p%Ed>dniXmMMx`TnZ<?;Qoh2{jX16IuT5(5
zQR4=oQZ7hHpmeR7X4*MCRJq}uHzB{%xlR<qG?=lfU^MG(L|35_Jq68n-@=w7+7p(k
zqQg{uGUJhPHHaA__fZ}@Av~-Zs`2dmAKe~w0Ya^B6S^<8+J)cwRW5uHce6x$9vxcD
zcg+R3TFvC20Q?qaZH4oUpoPyvhswXG1(Heu{zQko4brs(K*mn{)Q-^wfY^?6moz=<
zy)y63cVDB&Q}czQVaxvR2$<;*$b#%}9Lm>cRy||Ipfa;YAxy>mOt>V{SA{Ii)3B!a
z<n`E~{X<qja<!Bag`+#JsfqtmcREyC|I{S%b@oGLC=+eP4?EMFllAhcFx4PxwwZ<Z
z422CpnXZmk?1gXELS?UySH36Xn*$esa$3pMZ`4-ENl3!e#ynBKZQJ##9~`Kce8-!U
zm~NnsG4LR_O%)SXUEqidwF{0352Vul0)k_GaGlu_c_P#b#E3VId`3t%s#{)_to3Gi
zUO+r0?VyIcJSOJrP+T#nm#36hE}pb#=;^JEnY*zkd47fXZ`yN$UzLy5yM?Z1-dZt?
zna<?z`Xf`Vk}F1t1(!%t;Aj~P6e<<v7<b|dR*Y_FCyVFHZ0#2sgrEF4*EohY0$s%}
z_9w*#(#3Q;PUZ!=$7ox~ZOW%cW|_pi<6yx_*8tF=L__0ZsySwFPr8IH_78DHqLxPi
zt@nRQo7Wa-7=CU?tMe%mdMj+lk%3C9Xm`T~=yxCz{!!ni7S^0+3F@BQjeBcWbe+s{
z3NBFuqnzbRR|7tH^XpxcnA}TB2XSRgi7pH4k{<g*zBqJ-SG-HTe*i%!N(oO4W0M4;
zQB>V{Kl=G8dCTq`wT_wFxv4I{E))H`(s|l_P@)C#!$^+V6;mtTg*hi(q)sGrb#!a*
z(dZ}CHn&9KUIGIJaGc)ejXJz@Qs6jRTJB>%yG6Ld8-v@Nq8^K7%D+ujs4Aa4b!t=O
zU=M4v7IN<4#mw3Szt^f>g?iuSwSt11>w>SpJcomso7^s*{q?MTD+EGLQUkX`R-%sd
zBXOtrI=PH*waXNLspEnhyKS!O9Ol6hDP2wE*cg9InqI!N6xd1eFpm_s&DwuLL7r|3
zEak<Rl)R#1Ma6E#vz)2$DPL<~oIG8v2`#2^Yd!T>I*a63%X}+0sZ~gowl#ERD(2h9
zKrRSB?6KE??vRx!GP&28S2N;ECu{L%MRnj_h3xF!D8bGpMMt`~dHr!P*l^Q*<m(lc
zy?c$niZ6&_qQ3dvW^_WhF(NUlkNZo0ob~kk2wcjoh$`$9T#42>B{9~>NcfdT_$tgz
zA;nziS_(gf8ePNdHeZ`*&6$|8ixVw_`4csUsK%JTJ(&Oe*BL)~ME{{gB>zk?bJqk;
zrV@G6Z<ri*#gC0{drABp3Z6DLgCR}5q`lIwd+>pw5y=|0_igR|iJ57|uP2repPhmP
zB_Bd@_l^>Nv|-pfhF_j0IZ6eDBqimB8^CJ*1F9TfcS(E(SeVF%<X+vsQdqi_)DN8e
zdHiP^$gO->?Ox>vw%;8H&L>di;`=$U{~ac^OyFb`63q4KmG^c~zQ3mGTKd!Wh57hR
zXtmd?gEdZjC0-)qzI8?mnm3pnlB7{()6$L9gk%;g*fznRah{zY{h^;uPk4olTMWGw
zW9vO<6D}Ew&3%C$f%yP%hI+ESR?8@zL_dP%J0H%5o32_8*QAoo?ap(9P-WvxGO^DQ
zbb{0l{x{Sv`$Z!~j$bt~qd`6Qf24>Bh}pYX7@upHj-}_zyDQCA%`=3Z{d&t*IzyK+
zF7{-QM1Hsdnu@<P*?0-j4FJ&aT6!eE?=-$#N3c=A>#(Ur4KjXgt9ZuQF1NXI-CZ9=
zHp{)E2Dwu?IeRz!;B)4YZM6N%V8c~I``GBm>r^I&!Si=i^@TBYOW3l7^1>56YYZ|T
zV(O^f+?dvAAS{3TH1WQM6@ar;3@sG=Z42NKI_4uEi0S=~8IiH{=Aq5Mmerlvs`81_
zy1xMtCFHEx{KDtBjgMRrk>yvC{6!y@`hEdQ-{+v!JmEmTm}_C+<xE5iZ_@4PPY-`r
z7heR&+VrJGu%CY;+(@h)CN`xb$^8t-vf9V-=^WjSH2wX0Uzwaty^}fA`_pY(u>!!B
z?33`V_}AZVPf+(Z9r5<2^9(;6zdCPfKZ`LXAozsxu2`8P4vmUCu8^*xt0x2{!u9Zb
zM6G#btq7?6g!o<Yz#pW)OI07>Z4me6C7S+kgIHWz+0vgHxapz+EvL6X#^b)jjW&1M
z6G(%I9Dd9a!(rHBcFVUT^5&9bpW(Xu?%#h&Dn|wMdUFKEBv}VEp1KaPY<8qu$9vN*
zfH+RASM5CF8sNMYeSOat&N_O>Nvpp<b$R=bZ2n`-Rq`+Z?r$Y$H@2x%Lq(w>fz+(h
zzB+ghZltliWdHqsv+Kzpmn+lS9!%OK1<uMyh?hsGR>DQh`mQATBR&`5aPgshD-@Rc
zd8OzWO3on#LCZv`44v{C822eyK3h2Caieot^#Pdiv-%PmcrN!xSxwJq-m1*Ve)JK3
z60Y$VV0R*48_7I9YV10!etEt6_J>2Xvbq<_Z1WP6eeQNy$Swdl*9{qh#GbFFGWx&u
zKbgs)I@##(a<QNP2)+g>2d}Zs_9NTH@-=mo3@C&rKqwl80W0-BFV0fjPvzr3_ZQcg
zu0chpOuBQ#PCspBM@jQkklz463}yW%%Hv8;OhH%!fHP58icYeplB23VZsJ7JSnOCm
z4ri;Y&-J9JAGg;=OlajLW>Jj@phWfFpiFwBjdG|CR5fZV7A!G5QOY=Sc#v&aRisL%
zv22@LQt?a>Y!>-M<cTY)k#ZY)A=%Ln%PXF7SSN%`y@tG#-0>bn$G5r8*yg@8@<qp)
zBEV;Kn<c2vdOsPh7gdsr+UHL%jal9a*no(hB@`dFH#kfGiINPm)R%}z)CX047kkwk
z7Un>wv@lPBkGro+q1NCnKC;lmp09i5A9Y#w`BdXjPMs~U=D)-}(d8#XIA14`wKg9Y
z0_WQ=Pm|@bG*`>)SPqsmnyq(RgIwyT$1mgGycg<O=oXi0oTF)Z4Ul9MdwX@Le;~s5
zSP2VS2SGf9E4&M{L^G3kYVy&I$Rh%6JOHCwsB7$)oE_gMZG!v+f0iz7gHIt*Kr5k|
zsl_=da?fH?B%|T-r~}h@|DPBIEvZ5D@$>K)Cwn-naQ*Io7TdddPU7o*Qd;0m&$hUO
zjv4m+{(vDg>rHLy7LT&e>}8kcrGIWlu_J>3JCKf^l+izO2`8#LA8M#Q$l!mnDx;1U
zebdT0dneeWbn`=Ul&8q)CS^7asarT?qF5=Gwtst5x5UPN5Y%{2Xz1~yM^i&j$bXHX
zLZ!fdm7Lhfh7!Q9Ov?55yu<f<@U9OH$!kWaqT;PES3iTa*#J^2fGCTK%v@QGYXw%q
zN^3vwuiiiq*^9Qfe+I8En1pK&dB7O9-+fix-;ShN>l?K&wJ5~w<=CCRfB62J_$;|_
zzeOkNUl%@%{OwZDwQJ`(>@Qas@gJGS6A17M9F9UcspbZHyEF#6<HxD^^HsR<Z{$!C
zDPAhQkk0*2M|ute)=zw7`JF2OymUf*z$D`ptNmuuX;2aj3tzR1D)3vB0BglGm+u~?
zcEPQ#B^QgC%`0<oW`Gq?QO~Rt^2!X#{*Yd}*Kza0`)sfv#%V=xsu~fH^Zro{wxMpa
za-L1%caokCO`8#<;+rO^>`|7t(;MjCqhQOAFy_zEk=FIbHeRvMtitaHf8R8g2aCn*
z->HxAdAHrlXW<(0kNLRYscKHnw10{^cB|0Jlnvw*h1}{~p5?gR&+uEO3N75ee;xga
zq4@Bd4Rztp#bgosXohHu(qvxxev6^9ZpvOxs^Ys$y!+G4N<(@}37tWOkocT1Yk+Lg
zf}mov3rt=s-LvTtRGHJOf82&?<`vN1uCSD+ru=3XRfGKs<|f+14-SA+3xkH@rM&`H
zn5P9hVz76~y9;Tl@*<r1a~xd7{A1Irs6I?x7R{Q+q*u9byju(XYy3fTP64b1PWa=E
zqB{lt4!ohXq!V7w^1UFI6@u{ext$jIP&N`{i@C=>CbFOSzXMo`6}7aN0>>1Oc1Bfk
z0z_K(CSoOCciZcmh4biipc*=|a4icpLB#5x^OK+S^XSCl!>glTL&hj}Ku?uW?s~tZ
zi);MZm;-$lpm*cdi`hk=mi`>^(`~wx;582PC$A-bj@St+fXcuI5JYz0zr4YCWU)eh
z(SPUtk9mI%a&Ru*Zy~9U+1D4#gJ{1Yl;h*sVHTF4GHd|a^-_zx!nG;Ihg~pHL7ESI
zdKn1JzlR%>R$er6BUUR4>ZeRzb<E<tl9QCCd$!PdQNeF}Z&Y;axi-}E8(;qbnYEAt
zTFH{3hZCf13cU`6SAM_pd<KI}GpmqsF8`oS?KcGrhj(YgS#-eHe-Czlrli>}5bboN
zpYl-ENl9!cBHqIH^B47mkoe`-tG7aAjoidm(98j*0r95+QE$4lSr`CN*xMG-)JE$a
z{*Ea4*$W0dsmv`~Rzr`ms3`mZeorr8|5a?S;q*q<IS3u|#etb}uTYrx%hdf!_0!U4
zxfJR*?`S81y;JZHMY6C?Z)Gi=Km9v?FlKM@^`sB1m>a4Yr%(_Q8~vP=k5`B`Yn@(o
zYG#3i!@YQcMB<+z*TrNKi*p+;@s4*e*x8}zWn{3cMW6P}o6Ws{f+jj6CGz9L(j|6B
zMTUf{je}+_nd|iCALnq5ICoE8_%neL*XgYzi0<t9xY_`Zz3gn<f21VLmxrby4-tVa
zSs^_fQ>28kZ|H@%AWq910boz>=qYeK-V_gt{lvSJzKNR9WVOlW;jU3$#}NQcRZWZz
z7<iqY;QixVZLHJ(`)_=M0zeH1fG(2S(kxlk$`(C?rl$NLHsTbu<HYN4D`aD<&SSSq
z$;P$U;4@8TYVif8PKcJ@0a^A=rdNqK=FZQ5x1y=|n7#>a+T!-xyirXuv`eu3Wv_sP
zGLU8A)NKYok|S{?&AkWnf^PRpFJdCoCQ_q%h7~6p2=RHJU25G1p^oYZaTj`fN2dS?
zqB^dSb6kO5C3!HN*xjDP6qMh8{^H5?>vf(cQh1VyZ;PZjfp6vwo2hl*JcE~l-+4$v
z<E@*o&p28+*9LTUGX{qu08DkvM`*|wnc!p&Y|`e#c;J`uWOyJ89P07kDA=SCtGYDU
zx9Pbx1fmC}z%D9KKemu7Q0je$#F6BB8R6V8I}a>`;>qMBI6x+_5%35LZ{EW@fu%a9
zqT<^?v*?gxFCU8ltPwh)p#oq;mQh6~h6fmg(E}cK^Bp4Asa7;1|MW7S{o()XnKAGV
z&v%fVWb)xTG(3#@16_~WuZ~_+4nxKU($CT;k2u#t#nJnB-si3TV2`l0!lS;h>=cel
zCv~c9PQ%9bDB0(iC(^O^kgK+L#3<?H=Sdn-fftx|RTVBksQVy>f(3hr8nf>U(;jLx
zsACcV339cCk$iNR0IT%-{^;(bktT{1i0Wpn`)_nv)8RiYPdY)oEDAZ@<OImUcff{C
zMK>aifX@&S_k8gu4`ND#M<?X8gfu*_E@oIDglKbBH|98c=USnc?^|H0GrQ=@6VGh1
zqYhvvSU}4WfOFoEIQy5scO3v9ZQ|5a92%TepWtOy#Y+QI=TY7m?N@A~rh`H0=$R&9
zF%r0Mr$r2exU=m5eF_oj8G-1re{N^?+9Ps^lVpO*ziIoVu~aqKnozZBQo)M9Iox|n
z!(+UmLHjTOgSIHuHQ%eh3~uQC2k4YRd}{ytpGprcs5>|)zTovKI(+)CG0gz;N8B_G
zop)~W!qY^Ay!lrHG{ww>E0K~Rbns)<DV`Y&w&wD&yrBENKH(~=PiUKAd;Tm`6e}ET
zvhA@)`#|0}r7jtES$K4ime2h0*uuKtgx!ifo5biB>G(kAGi|5)J&n|VHh886_~(m9
z2sxWC2;9P`V*)?(vKC%vRsc{*OPf`ESlgqNRm?0zBk%HEKZC>1ldbYNs}wviGTsxO
z6u{mx>nIl)Upr;yv_S*yJMYmCz}+|GPxwv)_%n_kfA~exZ8V483LL^x8*haifG}r_
zh95?Lq#EVVVQ3uZi!H=?U5b_PI@hk8m9$C%B`N#E!YKd0b(iYODy{&)FH{tp|7e2{
zrWWk*J$N*h@DxZy20yqQ63w??;f_ZkC!4wj``o$8!c1udgz{;oJdeWy>9jQKtDA$<
z1n3(9S=JUEAr<}4>#=*60aWHr4>Zu|d6G5zh&2E28+&6e7HkX29h#RdM&BcX3E%#6
z23)8A@0ht@kKriL7*ox!Yz9{9!A)&r`faZClh*46pE5|KGNI>2tjjRi0}nW%z(lRn
z178;^&D4E;dGPY@4?5qTcPdvLXZc4KA71P~T>(4Dvb-d@2LPLKhJ^<?&PxnFWUE#f
zmUZc3Pn9gK4ZMa=#ybnv4G@0<2}Y#3ScGvwAl?0Ou-5Xb^eQYW!-8#?IXMvLk52G5
z`d9v#w&JWv6Q&y6n5{r-bq~xRr%FEoC#aU|3kAWMQK}l9AOJUYw+ORs`6Z_-FIjuD
z%4wLJXzbCbsE|QsxxjqS<WR_nf|qkdzJtRKYS~y&lK=E`|A0<#Y;g?Z1GgbS*L4BN
z$HViV>S2=iM<5CZv1}SZ+|0Xq#=P`U_pJ}gEhv5}c{%7LDZOe)Zg<2e+0Ke>eIUyI
z@Er&+ZuAD<OK-zqZwe*p<`3_Bj=gHR0;-1+&6owvbb8plY47k;OnBzB`GEr@?+^z)
zmni)g>3AmzuUdX;w^xOWwtRYElI9!b1jxegYEfGI9!4b$lKy59*tjJZ+O$wI`$eyi
zxNbvr^1A@k-XJ*OSm2*F-NH=LxyuMk@w)FvnQgM$^smBGfB|IvuW>r+V(&YSD_!M;
zd#uf{9{?Y9#2*~W-X`y@z)89ASfBx>le6|7RN^+AD%+#fre@{sh|i96#v1R%ysT#D
zs>v!~l--;LA}eB{uP3}FGrDdwOn?^%=)fjz2OLjLtL@(^GXcJhW+|kW>ezGgw+nWi
zjF`#fLO94^({jF&{h_hfe!!}l%ww@pPZIJOF4PZG+t3NB_Na8`0j_4fO7mo`f{rOr
z)25oana_*T2A;x{l^&Y7QywYV?S-uP&hIV(7Vu*-p^=+2?i6&LUT6=NVosnAt`*y?
zfBD<ge=})Mue~+1Xk?#px|A5$45x%}?gJdfdQ+_Du!Li6cFR`G%&Pm7FWWF?ZbQv}
z@Cvo12M0zHG00E2e|k3SvsrIM54Q?p$VfxOTfub-y2DPm^7z+kyegZ&%cxr2Hj%m5
zF5pSfyKXf=<`=i);Bxqu0oQ+ia!)-1w7~KHvY>Q;1rp@ku<?!tOgp_~aoD8WOOiu6
zmI2JYD8Q3e!v+j|{-f%%mG};Ju6(Qh+(NFUzFv5K0XP)D<Nzx8J`4tYI)Impaf&Q$
zYayojVFSR$i+#jCOr*h$<~O9eO84Z!St$jDC&?Sd%-6j~*yQY7h4WMFc!HmP{GzUS
z|2A#*I6;3k{CfZX=J$#I1-Gq2gyM6{&;=!6fLfNsKuAjg3|B+%ULKSK;8bvt&x8N8
zQ289Rcf{_Fh|(OM-+c41b`9JZL-+7zji3V1JlXyN$*MNT4&XLJP9pKok5rB-n?z2R
z!unGYUCXFhizutK*QI&^t!gvh*GHbaX;rZm5Dq|^?xaxBElB`CrL}&Bj{R2Ofcq0i
zSdnkOv24+ojmL}mc}29lDTfE|8!v!w{CRnDjg-gsr-;*6sV>UjwpRZzuu46&OUhIf
z(PJb=oUYdd$e}pi05BHdpL)Ns<{&)t5D5Nmg-6jn{RfDf%_$Vo5a9H834Prctd%DR
zPSfP<KvaQrqXYr)H(iT;rN-Se{KVKkn#azR7o%!IyW}c!=9X;vgNibU&VO4OphoX%
zfwPuxTK@#wg=-$#YEe$`$&IYz)ciHx9NfXq^RtoT0mo*jUCF=Img-U*G6FtF=aT@z
zKgEA3z3Dyx$S)`;v8Yx?_^0Hq9V#Lx`z65kTtAsQDG^(O*D!Ss;1X?qcOm@{ZBZ%N
z8iWaKONcaIzPxwzKFm*pU<GjGozpL(8A&XZ{Pauul;i&N!<YED9VS9zAtvC&LJ0p6
zRZj5k_F@>XEvi{JFH3L!r@VLkj=Ou~h11x!)1<MjCh5e+9ox3kXks-pv28VMY}<`(
zyD@sEzxBL));fQ{nK%3Gp6}ZG+7~`ZLlvWgE^{|-87??aPGgi9KL27Mu86(+hEtia
zWVYMOP_eWiu$W9uJ;dHYlgA6;m4m0S46*N7n~JZ7#vRC+0Pb6micDw1%e(DB{1|t2
zkugk;gooqB-yOBsNo6(Z!BhQTkV>=Wl0aS3^hW=X^m9igZ1Q&xgM6+ggJ;Nr3@fp#
zrf*{*WWimMxWCMCg;p-H@bZKJ-#Z7Ue5^3u3&5ZO8Oh-(#3Py&8bMl&?eEw{Ip-~N
zK|5c#%mRo0wCLC(rnyZ;icz%>N~f3_H@ZOzE_L^Z5kz<*p?-Nn_6{z$8e(Y8`5Kjt
zK<bjSqlf^NJ*W22#s;~A&Ww5@<*V;FOAD=}{bxq`zGeoYu5?0f3FmFsc&GIy5d8O<
zu5gEW;lHS)g{KNp1(=e@3Jter4lKu4wn?ym7~=c{%r<F(T&V;pDM(@SzXYoMxa>*L
zoqZevVJe2%J4ZJk-su%=-c}gKtlfW8=aGHzxR2Paj|<Q&KPfb`f{mII;0%VnPrxLT
z{-$8hQ&I+ztf*UR9E9QT)ZKaemDL|S&%3GZ#LgOfTw3DrTjGcq5}vT8HveO6v-YEx
zcr@pL*g3b+$x*gRQ1#R?N`x`Su{%e9Yj5fXiL^0ItQ6yxL6YZ%DMJ9izq<mQ#=5&j
z2YF&(j^=U;_+i;row(?Ti{&->_?(T66NO~1xm>osA<WMf4tH~@U_~C1gZPdCcgO(l
zSv2jM5GKVm`ovHJ3x$+bM1fv*J(lvtcNo6k6v0#;tZEi72FHTxGl@3jLsN!xYBJ-p
z?pYiYe`e8e<l3hgzs!Cq-_QAEP1#Mvc4g%&e8J$U<ZLe9U@7Sf30@5i5MqabwA2i@
zMGrgV?iVyDYGIqS_YK7+Gfn<r^rSV9m4i8&_&_BL5IjZ8QvHbo_)W7?lH5aOtZ=zP
znWq#Vj|F+HC^r5<^ou9WB>UvM1h}>CA%q1FSIiA!Rrl%h?<h-4o%(1z4fz_`XDW{7
zR*$?sqnWd7;}{H2WE^SXngCrT3WK&8qP^qOO2Z{U!8p#mM?0oLEn)S~Xqrdd#siOC
z*?JVPQ45*q-P&woQxqsvk|J9XlERcPqmyy|ukJJQw|Hnv_?L4nFgE`5(#;TvNryRY
z0ex>lPR>UUH2_Cfft$fTPFlrI1312h!yahWJqwMjs<FLWOBgq|rvBzjG1mC2GK%Ha
zUQ*_iDAQN|2)?eS?5!K#5`up2Mnmtwn!Uh2E(a;R6AyzNi0<l;UZ#GSi3fsCanI%(
zOxpKrrUe}};mzv$brLVBStdhxWB5gAj=vo&yVv)~-wNu~BQM2&^>URRA`(vZSe|3g
zH|EEYnw{9=MxM0&hsX}A_`8;a4t>|7$fqVgnnJVzkp$yx*IoiL2W5afH~h|9HvdX9
z39E=yej)0vkG2_ii<(M^en-NVfmbj8odV&Bq*=K(cq}2kiSu`eL*Y{=gYwhU)5wJ|
zFewR_RK$-^3r;^YsH-zBE9pnlFZH6L63VYaFjmB5F1{nlGIBB!06?&=shOFoP@$ld
zMmLej=+NwKzce~E9S6yyR{N*PhBFElt+QQpGw+!%{9CV8pAYnd=icIyv(eac19M+z
zo61#LuM?!SN`C&+BZO%G!TxhEhJ}Np`f&bS;(voI=;aXWyagKJLLUBgIP%JkPzL}S
zNURf}e_|3Vq3<HI?b4xrWX{4|6ANF$FB*8I+EEd4Y=2+Gjta&2&(e|>f~AOt0ML+f
z8}aH=*xv9XAoJovmqglYlbRC{nb<JMvcj`HUv+__m#$RLyv@@J9>;N734Le_QqGn^
zB_^l?f^aJ&R|sl?6kp_1n6Xd<5g_1({FgK2>!7V)HaG!)-@Vo$(v}rfKb-u_#e(-b
zpQH?lJAEJjF`R!t-UdVHOc01n#a*>tb9F{cu+PQZZ7~`0R&D3Y*t^qOJ=z<xUwlTc
zAq`YIGV=tT1lpacp)j5H6mWkcxZvw`?`Du3k*kqXS^5O0Tl`6KtUOyMZhU<F>H3P4
z=jA$sBE4**a_sH5^hAS<G?%Twznw4R;VGG!A=yF?lwTEzL%n7ZvP&2mQrUkDtVH!<
zVP}<<ij~FX`b!MPqoRo2Ef73Q;eWQ%ly0z`l=CB5tpGXQMMffqhKE;34c5D>8W0f?
z-CE*tg_+RP<nCgSNQ=!_L8valYdC`luZheM3;0jB{$!dziJ@hpJASDV+@l}em<xpj
z{+!mQbKTn-W<a>kxQ27yAHm|@hTUNe_Xm%W2&1R!w+;;&aWg9`DQV`eD2*k$J6xxI
zngiK;*ikF5>e=~3J)CJRGgYZJK?Z(IcDR(pquN1~+&C|GtY1LnvlZ|eQxAPA_!T{F
z(<~XLau7#ieU+0@5RP5V5aw*aO~KV^40;tyU}jX*n$ky8O1DE8ED6Klk?YK_SP<{z
zs{iXmjy|Y`vhE8_8p@;rYf3>+G+qx;7#T0I2F!{SXv}z>v4x?}zrRW?cI=0ovgZJ+
zFnE%%Mx>$s{N)<!d!f%+2&N<u((8EQ*Da)lmq|DKJ`mlv6$s<CY4h^tt)sl3+kOo&
zzSI{a`UxFd5&Sj%mBamgqouvqd3~|g`MgQ$Kr-2TjChwfG$z{2g6MZ|9+ao$e)Q!y
zAS<nv#06^07$rwjM{W6L2iyJ3(o$tdn4tJ1vgJWn(Om<VWNu(rMjMp+pzuw-q|7^3
z4$OE{#6hM`y<RP{#+jrPsF{N@3Px0Uxzw;&ZG)w$9%th_YBYAbyurjS@b}QboiO1O
zH7S4TU~)joSr%xO!KjuxkVuZlQ$WU~rL~n+9OHI?nZ@1<#Ye~ts~(UfNXKy&k~Y9a
zFuO@(PBNjCPf6>nC)R+@3fV2~M=6?x8`E3DFe1;0RYsrqwP>rZsg9JJ9sAW=4|VZX
zaun|;Ck%Ywz{FRWZYo4N6?Hr$+PjX|%f2Z@63?e7hf_}butJWsw7kGTu9;_ViSiva
z=dP8QrFBLp9UC7~oA}%ye28+6{}R|aLvE)?m1yUGM!+Y%*NHUDVuI-8$wft9S$+5m
z!gJ~p`_f5&7aRAu&X@EQNJ@sp2?)m)a`CTVm_0mP^7uLvu)>xoa{Z@LN4F&Z;~poy
zpBN7$3+Z-v+!7}qdxJaP<I+5!B5u{3b$;H({ODgO6E0ITACB_9)p+cm)<hCwuE?i-
zp+uPlgs#N{jkduu3?2$4dvfgeI7gjN7kpeQ>3K$A9A~bf%iS179xY|W3>UAg$Ipz&
zX`&QQI|#->Y4WHBJJ#r|#H@qy>of|>JVZz<E&x|hC<ee=Pp!E?Ni6$wLlPt*RcEND
z4xR;I=?a=IJT4N}STdNYP2)04XOwk$zZGF4vG4RSSaGw;J5Qlx6w?TUo{R+<grcB)
z9n)P7f>D4Fok65HT2_B=(%g2t)Mp{kcrm~{Nwn17908$uzD#8z)+q7>!BXxufgm-R
zxhIrfm>R=QtvSyq)-76LsZ3Qs<zNI<=RwMgALC<MWf;txMt{nx|Kagplo;1wlH#Bt
z4f0K~F;8ETj$Ru7ZwrV5YMS6OP2DgBpC|!B6nM`uoi@jW;m?N4F|iPrj2~&HkZ+T$
zpI&o@PP_o~WV)4dp?)vXsAfIgVuUyX5zaO+&PaF4<a{*hzeDEX9b3r%FGoivSt;0T
zGR4g>%Zf`+rdkHzo7Tc8n^1`0%9#&KQmY0|=TSNG@;USHJ3*vD>0mxAS2*W?y&t4P
z1eYed=GysYn%N+DJR45CBos(uXS#i=S1V#u9G87xuXZ}6?orVn9Efszt-n^fK(h*&
zoM|MSM+(n6bpb4f0n8;PrE&W8Wx{MvG4Z~oa!_Nl_!{^f8Vcwy!xWZEoy7(0{IIdc
zrk_fZ=x*w?9w)dI-LwiP*tu@3^c$wKyy<%A5EZoIGqQuBcUm@ie7Cbk9IbIxC=_AN
z@)WJ@eqAqQ%O@jgq@07o&d(toisBK-m>y|}A`(3S+7N3^5{)xVjDma28{4PO0qa-~
z%E_oJDklANGG0|ljgd(Senx+m@?N{nb#^z@0>%JTx+a1Qusf|E;Ae_{bbl5xu-X`p
zi}mTS26c2i42%x^`o+)xSQ^iBPN0fK7oJBFSR+af4#e<vdoVG5=V<x5i|p<DT1GUE
z90#6wGhJ~RL%T(MeJa$A&A4Vu;nyt0bd>glAOQRA=DVZ}#fp%*(UFzEt;OIMT!m#~
z(v!wzU;^*E{|xRLwNr`8BTj>DiYk`BY|Ef|DzPiZy`f<SJGFVS=}KP!9{YV~x%Nh8
zXGk+{+ex^4K!+@~^_phPC{CdgTTrn37QK9;%1unh)a-nV{H%VGv(G-yvUVK%89vlY
zOc>*APh<?S7<q$RrG=Bx)gw1nZJc*TgFUvLvIy4`$G$T$V}^wiEo3W-K7ak!@9$aj
zUKIiGYH{*BNg+Y+_&eclh7$d8A^e}&)>mSthK~Y;L2rNH=b|MUd^98*sa?U*V1};u
zfggCcl`Ilo-^afFS!uu)U!0Q2<W~*EHyyY(()r1O&nJY^Ac!&4)pnuKID#_hL)c=y
zD})}c^Xd7vbNEiIp8`mx1w$e=dXH<D|2cp6&HUMivx|HIMav*CrV&uczYaPVdASx}
z=MmMKXl|cu8f_@_pu;Zp=)`NeKOc7X><_bD*pi+amX<f_JM`aBFfFQSWi&#|!JQ6~
zr&SR>fD*jW16M9ih;3cO%rcvRvvOrm>*5vbDpdI<&N*>o^s-~#c}}S_MZ9uE5AvGa
z!5D^ix4M%@Kw_MrUv|J2491VPl?%@yw=^Gd7}Qp!#Em+Y1AwwpC{Lx9+0>LhF1O&k
zy|;awAF$YkT-GrXZ<b(O=O|C|=Ylm}J&plMV8p~arUcn3H1%+F=I&L+Zl7CYWH!b(
zTNy&nMTe*>Ap_7~k9Ks3eTsOA5qq#3*X;3#<1fKqaH%mbBweP0tAq9u&bsifm_X`Y
zp2<I)>jbF@po0^X8Xkhb*<!~OuGxJ-Ndc}G;_{na!URoMigzv^sffSM%3w0UaE~<Y
z2CekgjON45HE03&tmC6dLlws{n3<b|incWX91gp7-t3s`nUH0oC9?E%lf$*S=d}t%
z5vYfL$`M55-c?Hj_Pam%Pe1RWsvMT#h(4&wGF=gQ=D~Sx6W>vF_)rM!pYgIK(ZY*M
zsaSvU&Q*@~kK3s()JXgEvT2T75<F89&^rL*3@)NOHHGol&Qc`*0sN<_gh=cO0kG<>
zB&NQFQr@^3nrpF21%GsTpHMg$Sg=x;f4@0M@~+;(k4~$#-tgF(c#O|;lx`HK&F^4W
zRc;Xlz_pq_kgM-|y?iVXwt3>iJLW`8&7SRBb@AhBVh99$6quV+Eb{!ud@_eN*&5wa
zr)o#pcUYB3H24yB6i;Tt6BoMwqTThVM!5NGO~qJav25t>$crEV_i5SV$sKpUZg~DJ
zN-wC0uEen`mu>g+nXyLV<(L;=<viwqx)TuCVIOK$oTs`g;vQodR;fANv+c2!A29dH
zJSIK!fF$(C3})h!J=d5S)kDC2&B631s`eOQRTJKtCB#OOI7ikKb_pr<;1Jg2z6(WS
zmPfi!*KUS+AF7S_nTV(F!DK0ZK0clpG{)X3+1iu;aQpU#snv=8VZDYh@4FsaJn6eD
z=IxK`?Nj(1Pybs<=!Z!AN6*jk(w(Y}aaFb*N<X(-Oq@+jI4V~vu-DHG;y!K@QRExx
zPeh$T6OMSkEdSesyeDr}^L36mSsm+x{%_30CY>i%_p(OxpTuuL;^mVr^mVPWsCdsq
z)z;T<Wpkl#s1#1;Ki6<k1pf>}d@lR@7T+-q3OQ`eY3S$d5(r~I)(DIv)7tz!KY~hJ
zeB%iTgIa0@PvHvil|OgKi2{KvVh=U|Z~uMhu~Dtx&vx}p5HAWW%@A3+@|c6!RH`iq
z1mEx%(nf?5MBp*1yq?K$VwE)B+b?c|tZ8=~Cm$O()BsJ-^xOEyQ}wv<+uQ@d`~`uE
za}V*fXeeLujrGr37pC~q)Bv1bfVi_}IHnPr`4b$qs7y712FK&@;6~nRDtA^7kJz;w
z2M#gJ$<cUgv`yt`^&EsJiRM^2u_{)zA|y7ks%cFhx+J_$DaiMM*I04U3!>hLi`U$+
z*T-8(GeE2O8lF4yTml(yRVblMoH#RxoVif*avz5>cZ<PtBz<wo#WccYUExRItA0bh
zz;Ki@)`XajzF|`6CZ=mC_BDKu*nk%<1sW7w;*X$w2C&Xi3Jh;Y+ffLEAPgOd*olj_
zL=#^%%;IoX#OY4^tp3ERC(lx$L&%p%S%)a?j7AAbp5rm7%rfTsF;vvPaN2D4$~Bd@
z)k8p4%@$I5kV#o0b|pYZkZc*4e`aa2U6MO%^Y^>Dyu6g3GfPo;yR0=B&5&Tro6@i8
z`Kv`_2kCy8w(8OF-al7n$dQ2kP*OKkvn6}$JTl$?PGYV5O7RPI16-{b2mqxu3ZXv&
zO=OtdQX?MBdn2++s^hlePkHLvH2B_@FF7ZK94IyUPT{wL{qTm=_>%LP?byvXyjrZs
z*^cY2e!fj0)a84`T~Oo+axL_ay=zF;Zi)~AJh}FT=fcjzkyaKUY%U}S5yu;|Z+sN3
zZ`B_LGkX~w&GzRvJE{()*udTrajQKsQ>KP|AHlK**8^}(oVGbxm7}X**O@RZ<aLpB
zH#dXbw=C54WF)4ODqQPYXweO;pOBiT_KiOw3uVp~E7Dp}!I$*>(U{4E%r=Y6?D$Xm
z5RkYLWgNfK)qhcUDJJ>^M<ylb3H!0mgNt->&s43d1-ERg{BijFvg7RS@V3yFifr-R
zV8Mo$mUVRzLL23)aU?ZeR@~z8Jbc`0$8btV8mjClg05K!_B%YRlnXt|MhpF5W%Qtk
zTli_(ul(=YU8W;$FQM*d`*yEjj@_Q$78yJrQT-N0Vj$uHW9JVc^_&eD>Zp$GAdT|d
zUAg{QeWqhwVMj19U93=ybpMOlvGnTrqx)RI{52}d?iqoJXSu7x4JcR$j06WQrw&#(
zC#a~ydXq{aPWWMSIXQ+?EomhDSWm`$arsI#0;3mq6qDqx=uQWhfL}U-XU8VMP1^0=
ziHXA+h#zv99dmc}Zh*C(L#()yn;QZ`;4KsO#goGSi&te8w7i7`aE_pbDNp_9!HTZC
z;zCjZs<?I5kYXPnyQ$`?amZs<r+o1A)43K}8oi>PM|Iz5hJXD`Y@+wBYjM@S=OWXh
z4?@rj%V9om`&jaHFR;@g70%K0EDp9xBwIR?Jris}laWihDxDkqN;nZq@R@nBr4Tc;
zEHj~+XOx;6PMeQJX5GqBi&R16ltr1Du4bo*=h;=crRJ|3sWg?)9&Xa)6uj;u&!cZg
z8+qJAT($*X68?I5=U#$m0Q9x4>rm^<1rX|4Z`I&B?w>XGG_8>9+77?Eo-=|Vk1lAN
zcypu%IiNre%ERUk)n3~G$0Z~H&UWcPk;%ZI{4WRvGdmkG&RaWVWvoW>t3S4X?goq#
zXU#Y#GdtaPpx7*RS-EP<KBfzI_#oB(G0Zcp_-(*wgr$sd)^Eds!TX8z3CenIQQf;Z
z2y)Z=8zB4y(?8#St-1=3`<^=ec4E*3=^4b{!sIGP?^-z__f|CN0S{y40>pe`F)oLI
z<yyuMLHD(^lvU2m2^QA5uW`-`nZvX~^}Ablz$cO*I|(HC7~9DjApEy_xlve=at!x3
zwcl1&5Yud6;_QhdGMo47YkzmfuY1b*;2C)af>z7T5BH&LZSaAwPhDW2MPB99wjUZD
zwBjk0K&gsAZ?qu<t^J>L?0><uLZ2F0grJB%m!u|Y?`LHHFxZxx^-~)@98!13>l_Ir
zKKN&NL`+Qn>HpCMnynT2<C_@JZ*g3fIbaUB9E_@?1hs;Tbts)msesRRw->cTEf2Ie
zn2l*>CO&{?6F2tvMDhEx6VY<^{{@0Ju$QMF^dp(X{6gw(b-Px+5*KB2kAAU>U)<iU
zYmnD4yovv#gV?;3>&=9RWRIPnF`bg{5n!EyBFxs(Cd@3G*(uwNWUl75r^*j<22nqz
z^^K>UjaASw>EZBZ$z`g`iF(e5deXUong()RDPMvdvbuAhX)HI_QN^p<b@1%oNOBZC
z?P*4%uK(`A{pMcm=Nlk^uqoC0N7%#wa%kqL-Z*9CNc(zBSt=N=3AK}6Vq;X#mrf~K
z+_zyJ8GoQwA5uN3&<cw@znN}!;%oD&76nBAPKLt^UAHecgW2x55z8*QrL7xNhE?<|
z4!26o4z`a=XN~EQLvvWjCEZ_z8%x9|wZ$H(>(S_1G|~!YtKW-9JWFN!G5HWDe^@<&
zJl^;x)s6`FbHo?lWbCSf|Gk!Ey+dkh?T0X95h1UX7_M~-F&iUKsTN9-)+o>)i<?vM
zkAoNT6$Fn?SJOxzIcsh%i^*H^l~7|S56@YMp4niy*N&_sGb15GJ@Sah{_wl?i!;59
z@7G?#FkqW41Ufmva0AhlmE6|A0865?NlxEPoCR=q<8<S%T5#p)W$27rxH0v9h#A{k
z4vQR{W%F}tEE_cr|F3EN>8o|iM@!>zWR7XIgBXpWp*B0}+IuHeMAHmX5yW%MdFyHX
z24T4rT#k4-K})sIm|l6uf|p<mSySYanBDEG)@}n5Ak3_q=t|jB7o+P=oji<TgBgp&
zEf%F=H6DdwyTXL|zy-ow!hvqdXlqTP9qTx`z$1Sv!8!=zdxrkB_sZ1A#Z3xM&{~rF
z*S#V~BsJ;k=4l~$g*_PGVh8ilw*H=n1y`fR%*XJpR+g)cw!z!4-?|W^ua;V!bBoIh
zhZrEKi0n2@5-ZwkYHDbj>0n3P)eXQeW|Nzu4ArN_QWTxp2SJ+)=V^;<ZsspH**eN*
z!mU>jUM_UzA@uA1r)6%8G#<IdPI&uQg9io-D6on6*P^*Gg3eEVoTQP4W|uL;u~>lj
zKeXF57)46AJ<dj%fhlGk7;p3*BXOKZT@}RMm+JO+ImvF_h9gvCZ&Plfltnk2gFQn6
z0BcuK-Z$&)c73m$_Ei0b$rxa{IM5<Dt((E_8fg@d^VJmgdV}z8^X@ysSiL7c`vG!8
zj`YLG<JJDbBNO{0gYk)UPXNPTkAh$EPC%l^G0(M_ivMDf!)V*hsElw8Tz{k$eG@D5
zd&GH=ZAu!xot~Hu85+N(pQ)(9Ocn;J8~yeB<W2^U>bt`1<j?B7x#3!TAZz~}e&!KB
z&(iIV#WCb{|AtedcP2_1ut$=EF;tRUqyUmsL#aXd(`t~?KCu<L_^D!vzOAML*>D<_
ziW?gyN(5q+erX|AP`0G5D$8@+L4ouqg+M-szGC&~eupG_@o3vB%VhV#Y=%#^o8*EO
z(kR(F4O`<EAU9G`<k)6|+9?ZLwY%zmtk5R{VOV!^W{=bhJK1boUYHkARQ>0yb7{ur
z1HMLiNYfh`q7-4)t2M%)r%+m4P!{m4{h@>dU9TO6VNyC?^bO@XR2rHLezfV9okp(T
zX~QUKc&%VcVm3?~imERuf8pR0QtEWQL)zOxS4<|8x@W*=0>LOX{?Dhh9ER#3;U2%V
zRRj8NEck1zhf3C0K}rlVSr=fLB&X?(zXnh3N4~4WDhBs7i%ulXw<#gbB$Z~n!6X8;
zyX3nnh9$Qp*Ve^yz|-UC{VmLQeFlknaou|EgWEQA;Q$t@rNg6<8_C}J{q<J$%Lvl$
z3#dEE1GdKqww(KhNr(1v6gII-{|#EP+I22?*V_-f0yiS=6*3e*94>t-m)x{k(=pnM
zR4HYI3?R)&xc#e63(IKanh~RI&B=;s69^7pDW3BT$y`A`$bRAie6Mn74mG*ote&U}
zC=>b{w-@?JQ36ZNV8%}lU(Mi2(D`tZcBL((f|~IAY*}UpZbg)hpwYx*W5Ho4ec5)|
zewnM8<U0$_-8HFd^A3QrgG>(ZPAZAZ^<Wbs>D6d%gP@vO9B@|DRXRRi*+gnI0?Jlf
z+lOoZFUhDIVe<Qp$Q_d2ZLuGOwYpzCD40sXU`I*-;hKjD`#kjxD0E!AaEOPtu*MCF
z4%nB`2v(qH=yBUXgj~hKW6JTE;g&%7cMnL$7YeF2rpT#_rBa;tcpghPY8(*<pYs2f
zZos*wJ{6&KSiBqMM(Q3zWqnz@zHE(Aqcl;kWMYfv3W_ytyL&r9-$n_oh+R5V7VsJx
z%Vt;#O$|L#O4u0>vQ2+1?5IW<ZE^IatcSZE7V4pb<mav!PrHSE<G6kA@`jF9U`#mr
zL#8W%)JD@bM<Xcx+h64h8wUsKaFG07dpnb#ONZDeWkt-P8_lIHP3ccD%L{<K=V}5J
zgPxjp>KK&PF5iXl@DutoF-*`cWdh~19`^AKxl~2O1T4qV6koD<ANlh-iVQ(wii^F_
z2QT=531cbRY0~<y!b*+Mx};$)bL%D<EA&GZ6eo62`0cH+0NxF+YTNk1YQods1UIAi
zm?|7MuT1-P#U#pitIUC46L_x7c7276R(jj-XMS+Cr%XO|3j_y5%m-@9o0+1byc;AP
zf%29koeAF^628~nF-pidQk>B}g6Y`+|9Wo!XyO~tcA7OqS&&Xk9^i>Bb}Wu!K4LL1
zv7FQz1RpJ{bH%N*Dwodfo{=6=xUaa0bNr<QuNS_1kL|zMKdsW6hXC)fv4pF8$#~Vy
zT0?c(hqc@oD!9$tAWOX>dmKg#K_UKA?QA0Mli&gOJ($#FTXH@w+0!MRF-aRujrrr_
zuL&3)2W^Y-cYkpHO9`0CxD`+yTgk_vxA41((?`F;Pc*{ScxNqXRrb*6%&#I)t6)1F
zuYt`yE(a_qe+(1ogn-ltR2|$|+=&<&yTUQ+*0~nfM2r*YpS^u#pJmpN;shJaVk&qX
zOSR$WEKkQpKm<Vl`t9nc`n7sQEw3oCc95$gw_GSXqX_<!UN`|_GkrZ3)hHLN&n9lm
z6q2L#Oj3QbR<sI!@Y<a(NZq=oW5))5r9b`{%E?=B20#wm7|YBJK3LIsJ#f*X2=F3L
z-O{1>_<6eRb5q=Nb|CC!s9f;WdMD`8$!VrTxqYRhxyZ7*g5Tn_2!0%?W&+Fi?fBRC
zdQ>$S1O{8IOX~!CnCZ=b8>9;rA{0A-;kB*P*9d8h8ln{p&TXx|xn*lU>YKKId$YqK
zGf8!$?mCfmo(lWg`M@hoE_xTVYjqis+o~inF)LYRsKMsdQ#X|zC@<*sx_SMwX7NUL
z^-ZUL1^+31X+uXVz>+Nc-mKGmD>~Jt1vTURgGugt3+TS5F6(PE%fR`C4~g@l;-Tvl
zqN-=gvrE;Mt+n+V9mLIZX670ZBB_8^ww)Cd9xl}iF~AHD50@7E|66+QAaey({Ve(q
z199>Gi)|nCVF7sLtc~2;bc<5XxKJ(d>=`8gPA1P0_f&TxEx%r)G2XCuv&Kjj*8|o_
zLq~YwqYLe7>?|+rYDEkjPBdY)qeQM~Y9GP|j&mM4A$}YPz7q0+JQN(KZ+DAy{kkP&
z*H&LyK77&8FvM@9$Nj&3&ihW?`sISF+s(w&7^GnVkTFEVHg;siq=4DS!v%TKv|3nX
z9AuI7YCSaWni=odiG7bg&kW%q8yV$&OpQk`qRBAtzw!aJtg}7F5BMNgnUQ`iIgc1F
znAvw~p&85^rnJE#bxLQwSNy@UUeifK%2~?#tJ_Uf8XYQ~=+FzYkFX>2uRB(U5@sNN
z%LIp|L7HS5s`#I3L%51OXHc1`(A{0;C1|U{7?IYr2aQ1nmB8m37$JNl7xw$a^Eg@V
za>Rx?pv(%0D9VTQP(<&UsP@Lnje@Bp(|oZC$JsfeOy2JASpfSHN9>_KoLbX8--=;C
zzu97fSwdcb57E<}0T)A2AWvR=l^)CMeip8JSzUeip;h{xL#hnXU5})^LAo9|9T0%`
zW&F+AHwC6~4Qiopk|T%s@>3IE*MR$CB2Hh!=6x18v2`{j0MXA~l>Oq<_V5j@56AnI
ziK?6%ZJN*V3z$858+`jUSzrv}<+u5=3HRHMZeLUUqlw_NJ=K6uS9-$_ngSV(rag0^
zEam)nsJvKA;!*qVv`!zLp5Rehfoxy0-|Hx%KX9<I3%(kGM!A)<6vDY|f4(7Wzp^+#
zOxEXqI}Gqb;rRGqoT<l{Il|{^UL~gL?6eE&7DW&=KzApBh8=&l#{wLf^H^=tn2V^4
z7|`U?3PQeq1Q9R5kftzFsvvF(iueKQG;+1hs(h;vusHsOh-QEOl9Vb}TA~r@gml#m
zTko7#H5d@7sDwZq=~^+bD^t^o8E;SWIXa4#y4T-4=K1>}nph1=7^QUdTm_wZ^AOF9
zphK8~#g-MKEaw7_tb334U0PP&7sZK1Yo$(l+v+=&=XYbO4wIgf`uCwN$uH3Ylj{2`
z0)T||D?zGe_QYBNX@CTd+hs_tE^Y6XKJ)zr<ADN|<ASPqfB&|Qx%6K`ijT#2;9WHV
zcKpGgN`-_xTT@O2+S7dv6G8F{<_t){8_!cc+-J?Yy^^P5D|s_vW1SD@>Qt7EX5gP=
zE)4_`dHVQalOzHw)HmxvD^apd`qS>;6tGVyY*P|v`p)o6_|*c$%c<!$889u<Qs_f{
z3h7%b!#cu#)C9<%vRFRjs(RW<EM(J~2OQX$3Zmq91*Y3XyHP-jW<-qtJfRbak)n~M
zvEt+4ySaRs8%A6$YswY?^!1=F)+wZK6h`*H%Gfx#*mCI)ch~^j^)T4frFs)zaPVMc
zV?Nrb%S$j=z6B`b$yyBmXs1yj?`}(*$YOP_DU<Vr01t-S*r{D5#EmQITNMN}cd3js
z>G_qq#Nk$?)D1;BLfWHm3;Pr>d1UUr?}QGo7XKQj@Zd^%tqs$<xJ~IY>ajxh2tQ3N
zRKbspg%pB<he!5I(|ee$*l<%R1k{p5a`{*c0;5DpK%s<57~<ePjiVG?4+x{-4q4C2
zk0>atmz!FW<Km>k+JxLmkZ!{H(ZfC$pI^Cih-V@N#cPL{2XcFqFK5KOoO+^NUxf59
zr9xL)Ax3?rfz^~&SkhBWe2D#aD6xX-cH$b8j^UK?cBrW8o{P9shkgT?qcA|}V6hMY
zl>E7KFk<jX-0Nbxw?w`Hxhm0U=QM|hNA%v!)w@9xI8hZq9nRzLLx8!Y@L3!`?&9X`
zztR--T|i*=olUZsqjj|T&K2El1v4@U5^s)_?~&yL{CxCA6gmLj|5y=LjcCTskX0>D
zEK9>{YN_7sg~^YwzwEg9X0*DjS%{jZEZVXT;}$fWO-{wPkO8XIM3vwtpJmHw1lo1l
zstS!jZrrsASx3NZmPXyQ*KL=*WroiWoC`LDM+)p-l{>V?BuDa^r52)8!r=2|c5O@1
zKkMP4{a4IKR;;0&WPfkH)>Dj$2-@?pxMMCZWIX>1+vmqw81zD7kkz@_;%6~B;z05>
z3cYLb^pRb|@iaiZ+-fBYv~nul5vg5EgeW5QJ&$<9KlLQK+b`1?C!#FH)3@-CKd-Ph
z2Tpr<rr$ksn;?8e;z<-eBBwv$Q{o4otP|`ASKkp1Ha9=ov4y$`T}q4hhwS8xIhuUB
zec3SSy4QE{6lN{du`igLp=Yc8Ia_$U5R)d*!5WU*E%~pP02zH=b={#$1=Q**jPw@h
z6@O`2fOtjX!i-MI6F(Q&Jl&q;w;aMNdJYG##c$jmjB}i=ImH>pVNfWk|IFXHjHBbH
zW20!^^b`{GI5)<##f?~t|Mxyh$<EA4L*77Yi0AnKWr2ZuqsS1(0^-G;5kf(JvXV*?
J)d1t*{{tHm(mwzI
diff --git a/doc/guides/prog_guide/index.rst b/doc/guides/prog_guide/index.rst
index 07a4d35..e5a50a8 100644
--- a/doc/guides/prog_guide/index.rst
+++ b/doc/guides/prog_guide/index.rst
@@ -43,7 +43,6 @@ Programmer's Guide
mbuf_lib
poll_mode_drv
cryptodev_lib
- ivshmem_lib
link_bonding_poll_mode_drv_lib
timer_lib
hash_lib
diff --git a/doc/guides/prog_guide/ivshmem_lib.rst b/doc/guides/prog_guide/ivshmem_lib.rst
deleted file mode 100644
index b8a32e4..0000000
--- a/doc/guides/prog_guide/ivshmem_lib.rst
+++ /dev/null
@@ -1,160 +0,0 @@
-.. BSD LICENSE
- Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- All rights reserved.
-
- Redistribution and use in source and binary forms, with or without
- modification, are permitted provided that the following conditions
- are met:
-
- * Redistributions of source code must retain the above copyright
- notice, this list of conditions and the following disclaimer.
- * Redistributions in binary form must reproduce the above copyright
- notice, this list of conditions and the following disclaimer in
- the documentation and/or other materials provided with the
- distribution.
- * Neither the name of Intel Corporation nor the names of its
- contributors may be used to endorse or promote products derived
- from this software without specific prior written permission.
-
- THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-IVSHMEM Library
-===============
-
-The DPDK IVSHMEM library facilitates fast zero-copy data sharing among virtual machines
-(host-to-guest or guest-to-guest) by means of QEMU's IVSHMEM mechanism.
-
-The library works by providing a command line for QEMU to map several hugepages into a single IVSHMEM device.
-For the guest to know what is inside any given IVSHMEM device
-(and to distinguish between DPDK and non-DPDK IVSHMEM devices),
-a metadata file is also mapped into the IVSHMEM segment.
-No work needs to be done by the guest application to map IVSHMEM devices into memory;
-they are automatically recognized by the DPDK Environment Abstraction Layer (EAL).
-
-A typical DPDK IVSHMEM use case looks like the following.
-
-
-.. figure:: img/ivshmem.*
-
- Typical Ivshmem use case
-
-
-The same could work with several virtual machines, providing host-to-VM or VM-to-VM communication.
-The maximum number of metadata files is 32 (by default) and each metadata file can contain different (or even the same) hugepages.
-The only constraint is that each VM has to have access to the memory it is sharing with other entities (be it host or another VM).
-For example, if the user wants to share the same memzone across two VMs, each VM must have that memzone in its metadata file.
-
-IVHSHMEM Library API Overview
------------------------------
-
-The following is a simple guide to using the IVSHMEM Library API:
-
-* Call rte_ivshmem_metadata_create() to create a new metadata file.
- The metadata name is used to distinguish between multiple metadata files.
-
-* Populate each metadata file with DPDK data structures.
- This can be done using the following API calls:
-
- * rte_ivhshmem_metadata_add_memzone() to add rte_memzone to metadata file
-
- * rte_ivshmem_metadata_add_ring() to add rte_ring to metadata file
-
- * rte_ivshmem_metadata_add_mempool() to add rte_mempool to metadata file
-
-* Finally, call rte_ivshmem_metadata_cmdline_generate() to generate the command line for QEMU.
- Multiple metadata files (and thus multiple command lines) can be supplied to a single VM.
-
-.. note::
-
- Only data structures fully residing in DPDK hugepage memory work correctly.
- Supported data structures created by malloc(), mmap()
- or otherwise using non-DPDK memory cause undefined behavior and even a segmentation fault.
- Specifically, because the memzone field in an rte_ring refers to a memzone structure residing in local memory,
- accessing the memzone field in a shared rte_ring will cause an immediate segmentation fault.
-
-IVSHMEM Environment Configuration
----------------------------------
-
-The steps needed to successfully run IVSHMEM applications are the following:
-
-* Compile a special version of QEMU from sources.
-
- The source code can be found on the QEMU website (currently, version 1.4.x is supported, but version 1.5.x is known to work also),
- however, the source code will need to be patched to support using regular files as the IVSHMEM memory backend.
- The patch is not included in the DPDK package,
- but is available on the `Intel®DPDK-vswitch project webpage <https://01.org/packet-processing/intel%C2%AE-ovdk>`_
- (either separately or in a DPDK vSwitch package).
-
-* Enable IVSHMEM library in the DPDK build configuration.
-
- In the default configuration, IVSHMEM library is not compiled. To compile the IVSHMEM library,
- one has to either use one of the provided IVSHMEM targets
- (for example, x86_64-ivshmem-linuxapp-gcc),
- or set CONFIG_RTE_LIBRTE_IVSHMEM to "y" in the build configuration.
-
-* Set up hugepage memory on the virtual machine.
-
- The guest applications run as regular DPDK (primary) processes and thus need their own hugepage memory set up inside the VM.
- The process is identical to the one described in the *DPDK Getting Started Guide*.
-
-Best Practices for Writing IVSHMEM Applications
------------------------------------------------
-
-When considering the use of IVSHMEM for sharing memory, security implications need to be carefully evaluated.
-IVSHMEM is not suitable for untrusted guests, as IVSHMEM is essentially a window into the host process memory.
-This also has implications for the multiple VM scenarios.
-While the IVSHMEM library tries to share as little memory as possible,
-it is quite probable that data designated for one VM might also be present in an IVSMHMEM device designated for another VM.
-Consequently, any shared memory corruption will affect both host and all VMs sharing that particular memory.
-
-IVSHMEM applications essentially behave like multi-process applications,
-so it is important to implement access serialization to data and thread safety.
-DPDK ring structures are already thread-safe, however,
-any custom data structures that the user might need would have to be thread-safe also.
-
-Similar to regular DPDK multi-process applications,
-it is not recommended to use function pointers as functions might have different memory addresses in different processes.
-
-It is best to avoid freeing the rte_mbuf structure on a different machine from where it was allocated,
-that is, if the mbuf was allocated on the host, the host should free it.
-Consequently, any packet transmission and reception should also happen on the same machine (whether virtual or physical).
-Failing to do so may lead to data corruption in the mempool cache.
-
-Despite the IVSHMEM mechanism being zero-copy and having good performance,
-it is still desirable to do processing in batches and follow other procedures described in
-:ref:`Performance Optimization <Performance_Optimization>`.
-
-Best Practices for Running IVSHMEM Applications
------------------------------------------------
-
-For performance reasons,
-it is best to pin host processes and QEMU processes to different cores so that they do not interfere with each other.
-If NUMA support is enabled, it is also desirable to keep host process' hugepage memory and QEMU process on the same NUMA node.
-
-For the best performance across all NUMA nodes, each QEMU core should be pinned to host CPU core on the appropriate NUMA node.
-QEMU's virtual NUMA nodes should also be set up to correspond to physical NUMA nodes.
-More on how to set up DPDK and QEMU NUMA support can be found in *DPDK Getting Started Guide* and
-`QEMU documentation <http://qemu.weilnetz.de/qemu-doc.html>`_ respectively.
-A script called cpu_layout.py is provided with the DPDK package (in the tools directory)
-that can be used to identify which CPU cores correspond to which NUMA node.
-
-The QEMU IVSHMEM command line creation should be considered the last step before starting the virtual machine.
-Currently, there is no hot plug support for QEMU IVSHMEM devices,
-so one cannot add additional memory to an IVSHMEM device once it has been created.
-Therefore, the correct sequence to run an IVSHMEM application is to run host application first,
-obtain the command lines for each IVSHMEM device and then run all QEMU instances with guest applications afterwards.
-
-It is important to note that once QEMU is started, it holds on to the hugepages it uses for IVSHMEM devices.
-As a result, if the user wishes to shut down or restart the IVSHMEM host application,
-it is not enough to simply shut the application down.
-The virtual machine must also be shut down (if not, it will hold onto outdated host data).
diff --git a/doc/guides/prog_guide/source_org.rst b/doc/guides/prog_guide/source_org.rst
index 0c06d47..d9c140f 100644
--- a/doc/guides/prog_guide/source_org.rst
+++ b/doc/guides/prog_guide/source_org.rst
@@ -70,7 +70,6 @@ The lib directory contains::
+-- librte_ether # Generic interface to poll mode driver
+-- librte_hash # Hash library
+-- librte_ip_frag # IP fragmentation library
- +-- librte_ivshmem # QEMU IVSHMEM library
+-- librte_kni # Kernel NIC interface
+-- librte_kvargs # Argument parsing library
+-- librte_lpm # Longest prefix match library
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index d2dc4a9..be03262 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -54,9 +54,6 @@ Deprecation Notices
will be removed in 16.11.
It is replaced by rte_mempool_generic_get/put functions.
-* The ``rte_ivshmem`` feature (including library and EAL code) will be removed
- in 16.11 because it has some design issues which are not planned to be fixed.
-
* The vhost-cuse will be removed in 16.11. Since v2.1, a large majority of
development effort has gone to vhost-user, such as multiple-queue, live
migration, reconnect etc. Therefore, vhost-user should be used instead.
diff --git a/doc/guides/rel_notes/release_16_11.rst b/doc/guides/rel_notes/release_16_11.rst
index a6e3307..f7a2ceb 100644
--- a/doc/guides/rel_notes/release_16_11.rst
+++ b/doc/guides/rel_notes/release_16_11.rst
@@ -94,6 +94,9 @@ API Changes
This section is a comment. Make sure to start the actual text at the margin.
+* The ``rte_ivshmem`` feature (including library and EAL code) has been removed
+ in 16.11 because it had some design issues which were not planned to be fixed.
+
ABI Changes
-----------
diff --git a/examples/Makefile b/examples/Makefile
index 18b41b9..d49c7f2 100644
--- a/examples/Makefile
+++ b/examples/Makefile
@@ -61,7 +61,6 @@ ifneq ($(PQOS_INSTALL_PATH),)
DIRS-y += l2fwd-cat
endif
DIRS-$(CONFIG_RTE_LIBRTE_CRYPTODEV) += l2fwd-crypto
-DIRS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += l2fwd-ivshmem
DIRS-$(CONFIG_RTE_LIBRTE_JOBSTATS) += l2fwd-jobstats
DIRS-y += l2fwd-keepalive
DIRS-y += l2fwd-keepalive/ka-agent
diff --git a/examples/l2fwd-ivshmem/Makefile b/examples/l2fwd-ivshmem/Makefile
deleted file mode 100644
index 5f1d172..0000000
--- a/examples/l2fwd-ivshmem/Makefile
+++ /dev/null
@@ -1,43 +0,0 @@
-# BSD LICENSE
-#
-# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
-# All rights reserved.
-#
-# Redistribution and use in source and binary forms, with or without
-# modification, are permitted provided that the following conditions
-# are met:
-#
-# * Redistributions of source code must retain the above copyright
-# notice, this list of conditions and the following disclaimer.
-# * Redistributions in binary form must reproduce the above copyright
-# notice, this list of conditions and the following disclaimer in
-# the documentation and/or other materials provided with the
-# distribution.
-# * Neither the name of Intel Corporation nor the names of its
-# contributors may be used to endorse or promote products derived
-# from this software without specific prior written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-ifeq ($(RTE_SDK),)
-$(error "Please define RTE_SDK environment variable")
-endif
-
-# Default target, can be overriden by command line or environment
-RTE_TARGET ?= x86_64-ivshmem-linuxapp-gcc
-
-include $(RTE_SDK)/mk/rte.vars.mk
-
-DIRS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += host guest
-
-include $(RTE_SDK)/mk/rte.extsubdir.mk
diff --git a/examples/l2fwd-ivshmem/guest/Makefile b/examples/l2fwd-ivshmem/guest/Makefile
deleted file mode 100644
index 3ca73b4..0000000
--- a/examples/l2fwd-ivshmem/guest/Makefile
+++ /dev/null
@@ -1,50 +0,0 @@
-# BSD LICENSE
-#
-# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
-# All rights reserved.
-#
-# Redistribution and use in source and binary forms, with or without
-# modification, are permitted provided that the following conditions
-# are met:
-#
-# * Redistributions of source code must retain the above copyright
-# notice, this list of conditions and the following disclaimer.
-# * Redistributions in binary form must reproduce the above copyright
-# notice, this list of conditions and the following disclaimer in
-# the documentation and/or other materials provided with the
-# distribution.
-# * Neither the name of Intel Corporation nor the names of its
-# contributors may be used to endorse or promote products derived
-# from this software without specific prior written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-ifeq ($(RTE_SDK),)
-$(error "Please define RTE_SDK environment variable")
-endif
-
-# Default target, can be overriden by command line or environment
-RTE_TARGET ?= x86_64-ivshmem-linuxapp-gcc
-
-include $(RTE_SDK)/mk/rte.vars.mk
-
-# binary name
-APP = guest
-
-# all source are stored in SRCS-y
-SRCS-y := guest.c
-
-CFLAGS += -O3
-CFLAGS += $(WERROR_FLAGS)
-
-include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/l2fwd-ivshmem/guest/guest.c b/examples/l2fwd-ivshmem/guest/guest.c
deleted file mode 100644
index 7c49521..0000000
--- a/examples/l2fwd-ivshmem/guest/guest.c
+++ /dev/null
@@ -1,452 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <stdint.h>
-#include <unistd.h>
-#include <getopt.h>
-#include <signal.h>
-#include <sys/mman.h>
-#include <sys/types.h>
-#include <sys/stat.h>
-#include <sys/queue.h>
-#include <sys/file.h>
-#include <unistd.h>
-#include <limits.h>
-#include <errno.h>
-#include <sys/ioctl.h>
-#include <sys/time.h>
-
-#include <rte_common.h>
-#include <rte_eal_memconfig.h>
-#include <rte_log.h>
-#include <rte_memory.h>
-#include <rte_memcpy.h>
-#include <rte_memzone.h>
-#include <rte_eal.h>
-#include <rte_per_lcore.h>
-#include <rte_launch.h>
-#include <rte_atomic.h>
-#include <rte_cycles.h>
-#include <rte_prefetch.h>
-#include <rte_lcore.h>
-#include <rte_per_lcore.h>
-#include <rte_branch_prediction.h>
-#include <rte_interrupts.h>
-#include <rte_pci.h>
-#include <rte_random.h>
-#include <rte_debug.h>
-#include <rte_ether.h>
-#include <rte_ethdev.h>
-#include <rte_ring.h>
-#include <rte_mempool.h>
-#include <rte_mbuf.h>
-#include <rte_ivshmem.h>
-
-#include "../include/common.h"
-
-#define MAX_RX_QUEUE_PER_LCORE 16
-#define MAX_TX_QUEUE_PER_PORT 16
-struct lcore_queue_conf {
- unsigned n_rx_port;
- unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
- struct mbuf_table rx_mbufs[RTE_MAX_ETHPORTS];
- struct vm_port_param * port_param[MAX_RX_QUEUE_PER_LCORE];
-} __rte_cache_aligned;
-static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
-
-/* Print out statistics on packets dropped */
-static void
-print_stats(void)
-{
- uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
- unsigned portid;
-
- total_packets_dropped = 0;
- total_packets_tx = 0;
- total_packets_rx = 0;
-
- const char clr[] = { 27, '[', '2', 'J', '\0' };
- const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
-
- /* Clear screen and move to top left */
- printf("%s%s", clr, topLeft);
-
- printf("\nPort statistics ====================================");
-
- for (portid = 0; portid < ctrl->nb_ports; portid++) {
- /* skip ports that are not enabled */
- printf("\nStatistics for port %u ------------------------------"
- "\nPackets sent: %24"PRIu64
- "\nPackets received: %20"PRIu64
- "\nPackets dropped: %21"PRIu64,
- portid,
- ctrl->vm_ports[portid].stats.tx,
- ctrl->vm_ports[portid].stats.rx,
- ctrl->vm_ports[portid].stats.dropped);
-
- total_packets_dropped += ctrl->vm_ports[portid].stats.dropped;
- total_packets_tx += ctrl->vm_ports[portid].stats.tx;
- total_packets_rx += ctrl->vm_ports[portid].stats.rx;
- }
- printf("\nAggregate statistics ==============================="
- "\nTotal packets sent: %18"PRIu64
- "\nTotal packets received: %14"PRIu64
- "\nTotal packets dropped: %15"PRIu64,
- total_packets_tx,
- total_packets_rx,
- total_packets_dropped);
- printf("\n====================================================\n");
-}
-
-/* display usage */
-static void
-l2fwd_ivshmem_usage(const char *prgname)
-{
- printf("%s [EAL options] -- [-q NQ -T PERIOD]\n"
- " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
- " -T PERIOD: statistics will be refreshed each PERIOD seconds (0 to disable, 10 default, 86400 maximum)\n",
- prgname);
-}
-
-static unsigned int
-l2fwd_ivshmem_parse_nqueue(const char *q_arg)
-{
- char *end = NULL;
- unsigned long n;
-
- /* parse hexadecimal string */
- n = strtoul(q_arg, &end, 10);
- if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
- return 0;
- if (n == 0)
- return 0;
- if (n >= MAX_RX_QUEUE_PER_LCORE)
- return 0;
-
- return n;
-}
-
-static int
-l2fwd_ivshmem_parse_timer_period(const char *q_arg)
-{
- char *end = NULL;
- int n;
-
- /* parse number string */
- n = strtol(q_arg, &end, 10);
- if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
- return -1;
- if (n >= MAX_TIMER_PERIOD)
- return -1;
-
- return n;
-}
-
-/* Parse the argument given in the command line of the application */
-static int
-l2fwd_ivshmem_parse_args(int argc, char **argv)
-{
- int opt, ret;
- char **argvopt;
- int option_index;
- char *prgname = argv[0];
- static struct option lgopts[] = {
- {NULL, 0, 0, 0}
- };
-
- argvopt = argv;
-
- while ((opt = getopt_long(argc, argvopt, "q:p:T:",
- lgopts, &option_index)) != EOF) {
-
- switch (opt) {
-
- /* nqueue */
- case 'q':
- l2fwd_ivshmem_rx_queue_per_lcore = l2fwd_ivshmem_parse_nqueue(optarg);
- if (l2fwd_ivshmem_rx_queue_per_lcore == 0) {
- printf("invalid queue number\n");
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- break;
-
- /* timer period */
- case 'T':
- timer_period = l2fwd_ivshmem_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
- if (timer_period < 0) {
- printf("invalid timer period\n");
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- break;
-
- /* long options */
- case 0:
- l2fwd_ivshmem_usage(prgname);
- return -1;
-
- default:
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- }
-
- if (optind >= 0)
- argv[optind-1] = prgname;
-
- ret = optind-1;
- optind = 0; /* reset getopt lib */
- return ret;
-}
-
-/*
- * this loop is getting packets from RX rings of each port, and puts them
- * into TX rings of destination ports.
- */
-static void
-fwd_loop(void)
-{
-
- struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
- struct rte_mbuf **m_table;
- struct rte_mbuf *m;
- struct rte_ring *rx, *tx;
- unsigned lcore_id, len;
- uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
- unsigned i, j, portid, nb_rx;
- struct lcore_queue_conf *qconf;
- struct ether_hdr *eth;
- void *tmp;
-
- prev_tsc = 0;
- timer_tsc = 0;
-
- lcore_id = rte_lcore_id();
- qconf = &lcore_queue_conf[lcore_id];
-
- if (qconf->n_rx_port == 0) {
- RTE_LOG(INFO, L2FWD_IVSHMEM, "lcore %u has nothing to do\n", lcore_id);
- return;
- }
-
- RTE_LOG(INFO, L2FWD_IVSHMEM, "entering main loop on lcore %u\n", lcore_id);
-
- for (i = 0; i < qconf->n_rx_port; i++) {
- portid = qconf->rx_port_list[i];
- RTE_LOG(INFO, L2FWD_IVSHMEM, " -- lcoreid=%u portid=%u\n", lcore_id,
- portid);
- }
-
- while (ctrl->state == STATE_FWD) {
- cur_tsc = rte_rdtsc();
-
- diff_tsc = cur_tsc - prev_tsc;
-
- /*
- * Read packet from RX queues and send it to TX queues
- */
- for (i = 0; i < qconf->n_rx_port; i++) {
-
- portid = qconf->rx_port_list[i];
-
- len = qconf->rx_mbufs[portid].len;
-
- rx = ctrl->vm_ports[portid].rx_ring;
- tx = ctrl->vm_ports[portid].dst->tx_ring;
-
- m_table = qconf->rx_mbufs[portid].m_table;
-
- /* if we have something in the queue, try and transmit it down */
- if (len != 0) {
-
- /* if we succeed in sending the packets down, mark queue as free */
- if (rte_ring_enqueue_bulk(tx, (void**) m_table, len) == 0) {
- ctrl->vm_ports[portid].stats.tx += len;
- qconf->rx_mbufs[portid].len = 0;
- len = 0;
- }
- }
-
- nb_rx = rte_ring_count(rx);
-
- nb_rx = RTE_MIN(nb_rx, (unsigned) MAX_PKT_BURST);
-
- if (nb_rx == 0)
- continue;
-
- /* if we can get packets into the m_table */
- if (nb_rx < (RTE_DIM(qconf->rx_mbufs[portid].m_table) - len)) {
-
- /* this situation cannot exist, so if we fail to dequeue, that
- * means something went horribly wrong, hence the failure. */
- if (rte_ring_dequeue_bulk(rx, (void**) pkts_burst, nb_rx) < 0) {
- ctrl->state = STATE_FAIL;
- return;
- }
-
- ctrl->vm_ports[portid].stats.rx += nb_rx;
-
- /* put packets into the queue */
- for (j = 0; j < nb_rx; j++) {
- m = pkts_burst[j];
-
- rte_prefetch0(rte_pktmbuf_mtod(m, void *));
-
- m_table[len + j] = m;
-
- eth = rte_pktmbuf_mtod(m, struct ether_hdr *);
-
- /* 02:00:00:00:00:xx */
- tmp = ð->d_addr.addr_bytes[0];
- *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)portid << 40);
-
- /* src addr */
- ether_addr_copy(&ctrl->vm_ports[portid].dst->ethaddr,
- ð->s_addr);
- }
- qconf->rx_mbufs[portid].len += nb_rx;
-
- }
-
- }
-
- /* if timer is enabled */
- if (timer_period > 0) {
-
- /* advance the timer */
- timer_tsc += diff_tsc;
-
- /* if timer has reached its timeout */
- if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
-
- /* do this only on master core */
- if (lcore_id == rte_get_master_lcore()) {
- print_stats();
- /* reset the timer */
- timer_tsc = 0;
- }
- }
- }
-
- prev_tsc = cur_tsc;
- }
-}
-
-static int
-l2fwd_ivshmem_launch_one_lcore(__attribute__((unused)) void *dummy)
-{
- fwd_loop();
- return 0;
-}
-
-int
-main(int argc, char **argv)
-{
- struct lcore_queue_conf *qconf;
- const struct rte_memzone * mz;
- int ret;
- uint8_t portid;
- unsigned rx_lcore_id, lcore_id;
-
- /* init EAL */
- ret = rte_eal_init(argc, argv);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
- argc -= ret;
- argv += ret;
-
- /* parse application arguments (after the EAL ones) */
- ret = l2fwd_ivshmem_parse_args(argc, argv);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "Invalid l2fwd-ivshmem arguments\n");
-
- /* find control structure */
- mz = rte_memzone_lookup(CTRL_MZ_NAME);
- if (mz == NULL)
- rte_exit(EXIT_FAILURE, "Cannot find control memzone\n");
-
- ctrl = (struct ivshmem_ctrl*) mz->addr;
-
- /* lock the ctrl so that we don't have conflicts with anything else */
- rte_spinlock_lock(&ctrl->lock);
-
- if (ctrl->state == STATE_FWD)
- rte_exit(EXIT_FAILURE, "Forwarding already started!\n");
-
- rx_lcore_id = 0;
- qconf = NULL;
-
- /* Initialize the port/queue configuration of each logical core */
- for (portid = 0; portid < ctrl->nb_ports; portid++) {
-
- /* get the lcore_id for this port */
- while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
- lcore_queue_conf[rx_lcore_id].n_rx_port ==
- l2fwd_ivshmem_rx_queue_per_lcore) {
- rx_lcore_id++;
- if (rx_lcore_id >= RTE_MAX_LCORE)
- rte_exit(EXIT_FAILURE, "Not enough cores\n");
- }
-
- if (qconf != &lcore_queue_conf[rx_lcore_id])
- /* Assigned a new logical core in the loop above. */
- qconf = &lcore_queue_conf[rx_lcore_id];
-
- qconf->rx_port_list[qconf->n_rx_port] = portid;
- qconf->port_param[qconf->n_rx_port] = &ctrl->vm_ports[portid];
- qconf->n_rx_port++;
-
- printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
- }
-
- sigsetup();
-
- /* indicate that we are ready to forward */
- ctrl->state = STATE_FWD;
-
- /* unlock */
- rte_spinlock_unlock(&ctrl->lock);
-
- /* launch per-lcore init on every lcore */
- rte_eal_mp_remote_launch(l2fwd_ivshmem_launch_one_lcore, NULL, CALL_MASTER);
- RTE_LCORE_FOREACH_SLAVE(lcore_id) {
- if (rte_eal_wait_lcore(lcore_id) < 0)
- return -1;
- }
-
- return 0;
-}
diff --git a/examples/l2fwd-ivshmem/host/Makefile b/examples/l2fwd-ivshmem/host/Makefile
deleted file mode 100644
index f91419e..0000000
--- a/examples/l2fwd-ivshmem/host/Makefile
+++ /dev/null
@@ -1,50 +0,0 @@
-# BSD LICENSE
-#
-# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
-# All rights reserved.
-#
-# Redistribution and use in source and binary forms, with or without
-# modification, are permitted provided that the following conditions
-# are met:
-#
-# * Redistributions of source code must retain the above copyright
-# notice, this list of conditions and the following disclaimer.
-# * Redistributions in binary form must reproduce the above copyright
-# notice, this list of conditions and the following disclaimer in
-# the documentation and/or other materials provided with the
-# distribution.
-# * Neither the name of Intel Corporation nor the names of its
-# contributors may be used to endorse or promote products derived
-# from this software without specific prior written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-ifeq ($(RTE_SDK),)
-$(error "Please define RTE_SDK environment variable")
-endif
-
-# Default target, can be overriden by command line or environment
-RTE_TARGET ?= x86_64-ivshmem-linuxapp-gcc
-
-include $(RTE_SDK)/mk/rte.vars.mk
-
-# binary name
-APP = host
-
-# all source are stored in SRCS-y
-SRCS-y := host.c
-
-CFLAGS += -O3
-CFLAGS += $(WERROR_FLAGS)
-
-include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/l2fwd-ivshmem/host/host.c b/examples/l2fwd-ivshmem/host/host.c
deleted file mode 100644
index da7b00d..0000000
--- a/examples/l2fwd-ivshmem/host/host.c
+++ /dev/null
@@ -1,895 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#include <unistd.h>
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include <limits.h>
-#include <inttypes.h>
-#include <getopt.h>
-#include <signal.h>
-
-#include <rte_eal.h>
-#include <rte_cycles.h>
-#include <rte_eal_memconfig.h>
-#include <rte_debug.h>
-#include <rte_ether.h>
-#include <rte_ethdev.h>
-#include <rte_string_fns.h>
-#include <rte_ivshmem.h>
-#include <rte_ring.h>
-#include <rte_mempool.h>
-#include <rte_mbuf.h>
-
-#include "../include/common.h"
-
-/*
- * Configurable number of RX/TX ring descriptors
- */
-#define RTE_TEST_RX_DESC_DEFAULT 128
-#define RTE_TEST_TX_DESC_DEFAULT 512
-static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
-static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
-
-#define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
-
-/* mask of enabled ports */
-static uint32_t l2fwd_ivshmem_enabled_port_mask = 0;
-
-static struct ether_addr l2fwd_ivshmem_ports_eth_addr[RTE_MAX_ETHPORTS];
-
-#define NB_MBUF 8192
-
-#define MAX_RX_QUEUE_PER_LCORE 16
-#define MAX_TX_QUEUE_PER_PORT 16
-struct lcore_queue_conf {
- unsigned n_rx_port;
- unsigned rx_port_list[MAX_RX_QUEUE_PER_LCORE];
- struct vm_port_param * port_param[MAX_RX_QUEUE_PER_LCORE];
- struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
- struct mbuf_table rx_mbufs[RTE_MAX_ETHPORTS];
-} __rte_cache_aligned;
-static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
-
-static const struct rte_eth_conf port_conf = {
- .rxmode = {
- .split_hdr_size = 0,
- .header_split = 0, /**< Header Split disabled */
- .hw_ip_checksum = 0, /**< IP checksum offload disabled */
- .hw_vlan_filter = 0, /**< VLAN filtering disabled */
- .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
- .hw_strip_crc = 0, /**< CRC stripped by hardware */
- },
- .txmode = {
- .mq_mode = ETH_MQ_TX_NONE,
- },
-};
-
-#define METADATA_NAME "l2fwd_ivshmem"
-#define CMDLINE_OPT_FWD_CONF "fwd-conf"
-
-#define QEMU_CMD_FMT "/tmp/ivshmem_qemu_cmdline_%s"
-
-struct port_statistics port_statistics[RTE_MAX_ETHPORTS];
-
-struct rte_mempool * l2fwd_ivshmem_pktmbuf_pool = NULL;
-
-/* Print out statistics on packets dropped */
-static void
-print_stats(void)
-{
- uint64_t total_packets_dropped, total_packets_tx, total_packets_rx;
- uint64_t total_vm_packets_dropped = 0;
- uint64_t total_vm_packets_tx, total_vm_packets_rx;
- unsigned portid;
-
- total_packets_dropped = 0;
- total_packets_tx = 0;
- total_packets_rx = 0;
- total_vm_packets_tx = 0;
- total_vm_packets_rx = 0;
-
- const char clr[] = { 27, '[', '2', 'J', '\0' };
- const char topLeft[] = { 27, '[', '1', ';', '1', 'H','\0' };
-
- /* Clear screen and move to top left */
- printf("%s%s", clr, topLeft);
-
- printf("\nPort statistics ====================================");
-
- for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
- /* skip disabled ports */
- if ((l2fwd_ivshmem_enabled_port_mask & (1 << portid)) == 0)
- continue;
- printf("\nStatistics for port %u ------------------------------"
- "\nPackets sent: %24"PRIu64
- "\nPackets received: %20"PRIu64
- "\nPackets dropped: %21"PRIu64,
- portid,
- port_statistics[portid].tx,
- port_statistics[portid].rx,
- port_statistics[portid].dropped);
-
- total_packets_dropped += port_statistics[portid].dropped;
- total_packets_tx += port_statistics[portid].tx;
- total_packets_rx += port_statistics[portid].rx;
- }
-
- printf("\nVM statistics ======================================");
- for (portid = 0; portid < ctrl->nb_ports; portid++) {
- printf("\nStatistics for port %u ------------------------------"
- "\nPackets sent: %24"PRIu64
- "\nPackets received: %20"PRIu64,
- portid,
- ctrl->vm_ports[portid].stats.tx,
- ctrl->vm_ports[portid].stats.rx);
-
- total_vm_packets_dropped += ctrl->vm_ports[portid].stats.dropped;
- total_vm_packets_tx += ctrl->vm_ports[portid].stats.tx;
- total_vm_packets_rx += ctrl->vm_ports[portid].stats.rx;
- }
- printf("\nAggregate statistics ==============================="
- "\nTotal packets sent: %18"PRIu64
- "\nTotal packets received: %14"PRIu64
- "\nTotal packets dropped: %15"PRIu64
- "\nTotal VM packets sent: %15"PRIu64
- "\nTotal VM packets received: %11"PRIu64,
- total_packets_tx,
- total_packets_rx,
- total_packets_dropped,
- total_vm_packets_tx,
- total_vm_packets_rx);
- printf("\n====================================================\n");
-}
-
-static int
-print_to_file(const char *cmdline, const char *config_name)
-{
- FILE *file;
- char path[PATH_MAX];
-
- snprintf(path, sizeof(path), QEMU_CMD_FMT, config_name);
- file = fopen(path, "w");
- if (file == NULL) {
- RTE_LOG(ERR, L2FWD_IVSHMEM, "Could not open '%s' \n", path);
- return -1;
- }
-
- RTE_LOG(DEBUG, L2FWD_IVSHMEM, "QEMU command line for config '%s': %s \n",
- config_name, cmdline);
-
- fprintf(file, "%s\n", cmdline);
- fclose(file);
- return 0;
-}
-
-static int
-generate_ivshmem_cmdline(const char *config_name)
-{
- char cmdline[PATH_MAX];
- if (rte_ivshmem_metadata_cmdline_generate(cmdline, sizeof(cmdline),
- config_name) < 0)
- return -1;
-
- if (print_to_file(cmdline, config_name) < 0)
- return -1;
-
- rte_ivshmem_metadata_dump(stdout, config_name);
- return 0;
-}
-
-/* display usage */
-static void
-l2fwd_ivshmem_usage(const char *prgname)
-{
- printf("%s [EAL options] -- -p PORTMASK [-q NQ -T PERIOD]\n"
- " -p PORTMASK: hexadecimal bitmask of ports to configure\n"
- " -q NQ: number of queue (=ports) per lcore (default is 1)\n"
- " -T PERIOD: statistics will be refreshed each PERIOD seconds "
- "(0 to disable, 10 default, 86400 maximum)\n",
- prgname);
-}
-
-static unsigned int
-l2fwd_ivshmem_parse_nqueue(const char *q_arg)
-{
- char *end = NULL;
- unsigned long n;
-
- /* parse hexadecimal string */
- n = strtoul(q_arg, &end, 10);
- if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
- return 0;
- if (n == 0)
- return 0;
- if (n >= MAX_RX_QUEUE_PER_LCORE)
- return 0;
-
- return n;
-}
-
-static int
-l2fwd_ivshmem_parse_portmask(const char *portmask)
-{
- char *end = NULL;
- unsigned long pm;
-
- /* parse hexadecimal string */
- pm = strtoul(portmask, &end, 16);
- if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
- return -1;
-
- if (pm == 0)
- return -1;
-
- return pm;
-}
-
-static int
-l2fwd_ivshmem_parse_timer_period(const char *q_arg)
-{
- char *end = NULL;
- int n;
-
- /* parse number string */
- n = strtol(q_arg, &end, 10);
- if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
- return -1;
- if (n >= MAX_TIMER_PERIOD)
- return -1;
-
- return n;
-}
-
-/* Parse the argument given in the command line of the application */
-static int
-l2fwd_ivshmem_parse_args(int argc, char **argv)
-{
- int opt, ret;
- char **argvopt;
- int option_index;
- char *prgname = argv[0];
- static struct option lgopts[] = {
- {CMDLINE_OPT_FWD_CONF, 1, 0, 0},
- {NULL, 0, 0, 0}
- };
-
- argvopt = argv;
-
- while ((opt = getopt_long(argc, argvopt, "q:p:T:",
- lgopts, &option_index)) != EOF) {
-
- switch (opt) {
- /* portmask */
- case 'p':
- l2fwd_ivshmem_enabled_port_mask = l2fwd_ivshmem_parse_portmask(optarg);
- if (l2fwd_ivshmem_enabled_port_mask == 0) {
- printf("invalid portmask\n");
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- break;
-
- /* nqueue */
- case 'q':
- l2fwd_ivshmem_rx_queue_per_lcore = l2fwd_ivshmem_parse_nqueue(optarg);
- if (l2fwd_ivshmem_rx_queue_per_lcore == 0) {
- printf("invalid queue number\n");
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- break;
-
- /* timer period */
- case 'T':
- timer_period = l2fwd_ivshmem_parse_timer_period(optarg) * 1000 * TIMER_MILLISECOND;
- if (timer_period < 0) {
- printf("invalid timer period\n");
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- break;
-
- /* long options */
- case 0:
- l2fwd_ivshmem_usage(prgname);
- return -1;
-
- default:
- l2fwd_ivshmem_usage(prgname);
- return -1;
- }
- }
-
- if (optind >= 0)
- argv[optind-1] = prgname;
-
- ret = optind-1;
- optind = 0; /* reset getopt lib */
- return ret;
-}
-
-/* Check the link status of all ports in up to 9s, and print them finally */
-static void
-check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
-{
-#define CHECK_INTERVAL 100 /* 100ms */
-#define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
- uint8_t portid, count, all_ports_up, print_flag = 0;
- struct rte_eth_link link;
-
- printf("\nChecking link status");
- fflush(stdout);
- for (count = 0; count <= MAX_CHECK_TIME; count++) {
- all_ports_up = 1;
- for (portid = 0; portid < port_num; portid++) {
- if ((port_mask & (1 << portid)) == 0)
- continue;
- memset(&link, 0, sizeof(link));
- rte_eth_link_get_nowait(portid, &link);
- /* print link status if flag set */
- if (print_flag == 1) {
- if (link.link_status)
- printf("Port %d Link Up - speed %u "
- "Mbps - %s\n", (uint8_t)portid,
- (unsigned)link.link_speed,
- (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
- ("full-duplex") : ("half-duplex\n"));
- else
- printf("Port %d Link Down\n",
- (uint8_t)portid);
- continue;
- }
- /* clear all_ports_up flag if any link down */
- if (link.link_status == ETH_LINK_DOWN) {
- all_ports_up = 0;
- break;
- }
- }
- /* after finally printing all link status, get out */
- if (print_flag == 1)
- break;
-
- if (all_ports_up == 0) {
- printf(".");
- fflush(stdout);
- rte_delay_ms(CHECK_INTERVAL);
- }
-
- /* set the print_flag if all ports up or timeout */
- if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
- print_flag = 1;
- printf("done\n");
- }
- }
-}
-
-/* Send the burst of packets on an output interface */
-static int
-l2fwd_ivshmem_send_burst(struct lcore_queue_conf *qconf, unsigned n, uint8_t port)
-{
- struct rte_mbuf **m_table;
- unsigned ret;
- unsigned queueid =0;
-
- m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
-
- ret = rte_eth_tx_burst(port, (uint16_t) queueid, m_table, (uint16_t) n);
- port_statistics[port].tx += ret;
- if (unlikely(ret < n)) {
- port_statistics[port].dropped += (n - ret);
- do {
- rte_pktmbuf_free(m_table[ret]);
- } while (++ret < n);
- }
-
- return 0;
-}
-
-/* Enqueue packets for TX and prepare them to be sent on the network */
-static int
-l2fwd_ivshmem_send_packet(struct rte_mbuf *m, uint8_t port)
-{
- unsigned lcore_id, len;
- struct lcore_queue_conf *qconf;
-
- lcore_id = rte_lcore_id();
-
- qconf = &lcore_queue_conf[lcore_id];
- len = qconf->tx_mbufs[port].len;
- qconf->tx_mbufs[port].m_table[len] = m;
- len++;
-
- /* enough pkts to be sent */
- if (unlikely(len == MAX_PKT_BURST)) {
- l2fwd_ivshmem_send_burst(qconf, MAX_PKT_BURST, port);
- len = 0;
- }
-
- qconf->tx_mbufs[port].len = len;
- return 0;
-}
-
-static int
-l2fwd_ivshmem_receive_burst(struct lcore_queue_conf *qconf, unsigned portid,
- unsigned vm_port)
-{
- struct rte_mbuf ** m;
- struct rte_ring * rx;
- unsigned len, pkt_idx;
-
- m = qconf->rx_mbufs[portid].m_table;
- len = qconf->rx_mbufs[portid].len;
- rx = qconf->port_param[vm_port]->rx_ring;
-
- /* if enqueueing failed, ring is probably full, so drop the packets */
- if (rte_ring_enqueue_bulk(rx, (void**) m, len) < 0) {
- port_statistics[portid].dropped += len;
-
- pkt_idx = 0;
- do {
- rte_pktmbuf_free(m[pkt_idx]);
- } while (++pkt_idx < len);
- }
- else
- /* increment rx stats by however many packets we managed to receive */
- port_statistics[portid].rx += len;
-
- return 0;
-}
-
-/* Enqueue packets for RX and prepare them to be sent to VM */
-static int
-l2fwd_ivshmem_receive_packets(struct rte_mbuf ** m, unsigned n, unsigned portid,
- unsigned vm_port)
-{
- unsigned lcore_id, len, pkt_idx;
- struct lcore_queue_conf *qconf;
-
- lcore_id = rte_lcore_id();
-
- qconf = &lcore_queue_conf[lcore_id];
-
- len = qconf->rx_mbufs[portid].len;
- pkt_idx = 0;
-
- /* enqueue packets */
- while (pkt_idx < n && len < MAX_PKT_BURST * 2) {
- qconf->rx_mbufs[portid].m_table[len++] = m[pkt_idx++];
- }
-
- /* increment queue len by however many packets we managed to receive */
- qconf->rx_mbufs[portid].len += pkt_idx;
-
- /* drop the unreceived packets */
- if (unlikely(pkt_idx < n)) {
- port_statistics[portid].dropped += n - pkt_idx;
- do {
- rte_pktmbuf_free(m[pkt_idx]);
- } while (++pkt_idx < n);
- }
-
- /* drain the queue halfway through the maximum capacity */
- if (unlikely(qconf->rx_mbufs[portid].len >= MAX_PKT_BURST))
- l2fwd_ivshmem_receive_burst(qconf, portid, vm_port);
-
- return 0;
-}
-
-/* loop for host forwarding mode.
- * the data flow is as follows:
- * 1) get packets from TX queue and send it out from a given port
- * 2) RX packets from given port and enqueue them on RX ring
- * 3) dequeue packets from TX ring and put them on TX queue for a given port
- */
-static void
-fwd_loop(void)
-{
- struct rte_mbuf *pkts_burst[MAX_PKT_BURST * 2];
- struct rte_mbuf *m;
- unsigned lcore_id;
- uint64_t prev_tsc, diff_tsc, cur_tsc, timer_tsc;
- unsigned i, j, portid, nb_rx;
- struct lcore_queue_conf *qconf;
- struct rte_ring *tx;
- const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
-
- prev_tsc = 0;
- timer_tsc = 0;
-
- lcore_id = rte_lcore_id();
- qconf = &lcore_queue_conf[lcore_id];
-
- if (qconf->n_rx_port == 0) {
- RTE_LOG(INFO, L2FWD_IVSHMEM, "lcore %u has nothing to do\n", lcore_id);
- return;
- }
-
- RTE_LOG(INFO, L2FWD_IVSHMEM, "entering main loop on lcore %u\n", lcore_id);
-
- for (i = 0; i < qconf->n_rx_port; i++) {
-
- portid = qconf->rx_port_list[i];
- RTE_LOG(INFO, L2FWD_IVSHMEM, " -- lcoreid=%u portid=%u\n", lcore_id,
- portid);
- }
-
- while (ctrl->state == STATE_FWD) {
-
- cur_tsc = rte_rdtsc();
-
- /*
- * Burst queue drain
- */
- diff_tsc = cur_tsc - prev_tsc;
- if (unlikely(diff_tsc > drain_tsc)) {
-
- /*
- * TX
- */
- for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
- if (qconf->tx_mbufs[portid].len == 0)
- continue;
- l2fwd_ivshmem_send_burst(qconf,
- qconf->tx_mbufs[portid].len,
- (uint8_t) portid);
- qconf->tx_mbufs[portid].len = 0;
- }
-
- /*
- * RX
- */
- for (i = 0; i < qconf->n_rx_port; i++) {
- portid = qconf->rx_port_list[i];
- if (qconf->rx_mbufs[portid].len == 0)
- continue;
- l2fwd_ivshmem_receive_burst(qconf, portid, i);
- qconf->rx_mbufs[portid].len = 0;
- }
-
- /* if timer is enabled */
- if (timer_period > 0) {
-
- /* advance the timer */
- timer_tsc += diff_tsc;
-
- /* if timer has reached its timeout */
- if (unlikely(timer_tsc >= (uint64_t) timer_period)) {
-
- /* do this only on master core */
- if (lcore_id == rte_get_master_lcore()) {
- print_stats();
- /* reset the timer */
- timer_tsc = 0;
- }
- }
- }
-
- prev_tsc = cur_tsc;
- }
-
- /*
- * packet RX and forwarding
- */
- for (i = 0; i < qconf->n_rx_port; i++) {
-
- /* RX packets from port and put them on RX ring */
- portid = qconf->rx_port_list[i];
- nb_rx = rte_eth_rx_burst((uint8_t) portid, 0,
- pkts_burst, MAX_PKT_BURST);
-
- if (nb_rx != 0)
- l2fwd_ivshmem_receive_packets(pkts_burst, nb_rx, portid, i);
-
- /* dequeue packets from TX ring and send them to TX queue */
- tx = qconf->port_param[i]->tx_ring;
-
- nb_rx = rte_ring_count(tx);
-
- nb_rx = RTE_MIN(nb_rx, (unsigned) MAX_PKT_BURST);
-
- if (nb_rx == 0)
- continue;
-
- /* should not happen */
- if (unlikely(rte_ring_dequeue_bulk(tx, (void**) pkts_burst, nb_rx) < 0)) {
- ctrl->state = STATE_FAIL;
- return;
- }
-
- for (j = 0; j < nb_rx; j++) {
- m = pkts_burst[j];
- l2fwd_ivshmem_send_packet(m, portid);
- }
- }
- }
-}
-
-static int
-l2fwd_ivshmem_launch_one_lcore(__attribute__((unused)) void *dummy)
-{
- fwd_loop();
- return 0;
-}
-
-int main(int argc, char **argv)
-{
- char name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
- struct lcore_queue_conf *qconf;
- struct rte_eth_dev_info dev_info;
- uint8_t portid, port_nr;
- uint8_t nb_ports, nb_ports_available;
- uint8_t nb_ports_in_mask;
- int ret;
- unsigned lcore_id, rx_lcore_id;
-
- /* init EAL */
- ret = rte_eal_init(argc, argv);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
- argc -= ret;
- argv += ret;
-
- /* parse application arguments (after the EAL ones) */
- ret = l2fwd_ivshmem_parse_args(argc, argv);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "Invalid l2fwd-ivshmem arguments\n");
-
- /* create a shared mbuf pool */
- l2fwd_ivshmem_pktmbuf_pool =
- rte_pktmbuf_pool_create(MBUF_MP_NAME, NB_MBUF, 32,
- 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
- if (l2fwd_ivshmem_pktmbuf_pool == NULL)
- rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
-
- nb_ports = rte_eth_dev_count();
- if (nb_ports == 0)
- rte_exit(EXIT_FAILURE, "No Ethernet ports - bye\n");
-
- /*
- * reserve memzone to communicate with VMs - we cannot use rte_malloc here
- * because while it is technically possible, it is a very bad idea to share
- * the heap between two primary processes.
- */
- ctrl_mz = rte_memzone_reserve(CTRL_MZ_NAME, sizeof(struct ivshmem_ctrl),
- SOCKET_ID_ANY, 0);
- if (ctrl_mz == NULL)
- rte_exit(EXIT_FAILURE, "Cannot reserve control memzone\n");
- ctrl = (struct ivshmem_ctrl*) ctrl_mz->addr;
-
- memset(ctrl, 0, sizeof(struct ivshmem_ctrl));
-
- /*
- * Each port is assigned an output port.
- */
- nb_ports_in_mask = 0;
- for (portid = 0; portid < nb_ports; portid++) {
- /* skip ports that are not enabled */
- if ((l2fwd_ivshmem_enabled_port_mask & (1 << portid)) == 0)
- continue;
- if (portid % 2) {
- ctrl->vm_ports[nb_ports_in_mask].dst = &ctrl->vm_ports[nb_ports_in_mask-1];
- ctrl->vm_ports[nb_ports_in_mask-1].dst = &ctrl->vm_ports[nb_ports_in_mask];
- }
-
- nb_ports_in_mask++;
-
- rte_eth_dev_info_get(portid, &dev_info);
- }
- if (nb_ports_in_mask % 2) {
- printf("Notice: odd number of ports in portmask.\n");
- ctrl->vm_ports[nb_ports_in_mask-1].dst =
- &ctrl->vm_ports[nb_ports_in_mask-1];
- }
-
- rx_lcore_id = 0;
- qconf = NULL;
-
- printf("Initializing ports configuration...\n");
-
- nb_ports_available = nb_ports;
-
- /* Initialise each port */
- for (portid = 0; portid < nb_ports; portid++) {
-
- /* skip ports that are not enabled */
- if ((l2fwd_ivshmem_enabled_port_mask & (1 << portid)) == 0) {
- printf("Skipping disabled port %u\n", (unsigned) portid);
- nb_ports_available--;
- continue;
- }
-
- /* init port */
- printf("Initializing port %u... ", (unsigned) portid);
- fflush(stdout);
- ret = rte_eth_dev_configure(portid, 1, 1, &port_conf);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
- ret, (unsigned) portid);
-
- rte_eth_macaddr_get(portid,&l2fwd_ivshmem_ports_eth_addr[portid]);
-
- /* init one RX queue */
- fflush(stdout);
- ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
- rte_eth_dev_socket_id(portid),
- NULL,
- l2fwd_ivshmem_pktmbuf_pool);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup:err=%d, port=%u\n",
- ret, (unsigned) portid);
-
- /* init one TX queue on each port */
- fflush(stdout);
- ret = rte_eth_tx_queue_setup(portid, 0, nb_txd,
- rte_eth_dev_socket_id(portid),
- NULL);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup:err=%d, port=%u\n",
- ret, (unsigned) portid);
-
- /* Start device */
- ret = rte_eth_dev_start(portid);
- if (ret < 0)
- rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
- ret, (unsigned) portid);
-
- printf("done: \n");
-
- rte_eth_promiscuous_enable(portid);
-
- printf("Port %u, MAC address: %02X:%02X:%02X:%02X:%02X:%02X\n\n",
- (unsigned) portid,
- l2fwd_ivshmem_ports_eth_addr[portid].addr_bytes[0],
- l2fwd_ivshmem_ports_eth_addr[portid].addr_bytes[1],
- l2fwd_ivshmem_ports_eth_addr[portid].addr_bytes[2],
- l2fwd_ivshmem_ports_eth_addr[portid].addr_bytes[3],
- l2fwd_ivshmem_ports_eth_addr[portid].addr_bytes[4],
- l2fwd_ivshmem_ports_eth_addr[portid].addr_bytes[5]);
-
- /* initialize port stats */
- memset(&port_statistics, 0, sizeof(port_statistics));
- }
-
- if (!nb_ports_available) {
- rte_exit(EXIT_FAILURE,
- "All available ports are disabled. Please set portmask.\n");
- }
- port_nr = 0;
-
- /* Initialize the port/queue configuration of each logical core */
- for (portid = 0; portid < nb_ports; portid++) {
- if ((l2fwd_ivshmem_enabled_port_mask & (1 << portid)) == 0)
- continue;
-
- /* get the lcore_id for this port */
- while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
- lcore_queue_conf[rx_lcore_id].n_rx_port ==
- l2fwd_ivshmem_rx_queue_per_lcore) {
- rx_lcore_id++;
- if (rx_lcore_id >= RTE_MAX_LCORE)
- rte_exit(EXIT_FAILURE, "Not enough cores\n");
- }
-
- if (qconf != &lcore_queue_conf[rx_lcore_id])
- /* Assigned a new logical core in the loop above. */
- qconf = &lcore_queue_conf[rx_lcore_id];
-
-
- rte_eth_macaddr_get(portid, &ctrl->vm_ports[port_nr].ethaddr);
-
- qconf->rx_port_list[qconf->n_rx_port] = portid;
- qconf->port_param[qconf->n_rx_port] = &ctrl->vm_ports[port_nr];
- qconf->n_rx_port++;
- port_nr++;
- printf("Lcore %u: RX port %u\n", rx_lcore_id, (unsigned) portid);
- }
-
- check_all_ports_link_status(nb_ports_available, l2fwd_ivshmem_enabled_port_mask);
-
- /* create rings for each VM port (several ports can be on the same VM).
- * note that we store the pointers in ctrl - that way, they are the same
- * and valid across all VMs because ctrl is also in DPDK memory */
- for (portid = 0; portid < nb_ports_available; portid++) {
-
- /* RX ring. SP/SC because it's only used by host and a single VM */
- snprintf(name, sizeof(name), "%s%i", RX_RING_PREFIX, portid);
- r = rte_ring_create(name, NB_MBUF,
- SOCKET_ID_ANY, RING_F_SP_ENQ | RING_F_SC_DEQ);
- if (r == NULL)
- rte_exit(EXIT_FAILURE, "Cannot create ring %s\n", name);
-
- ctrl->vm_ports[portid].rx_ring = r;
-
- /* TX ring. SP/SC because it's only used by host and a single VM */
- snprintf(name, sizeof(name), "%s%i", TX_RING_PREFIX, portid);
- r = rte_ring_create(name, NB_MBUF,
- SOCKET_ID_ANY, RING_F_SP_ENQ | RING_F_SC_DEQ);
- if (r == NULL)
- rte_exit(EXIT_FAILURE, "Cannot create ring %s\n", name);
-
- ctrl->vm_ports[portid].tx_ring = r;
- }
-
- /* create metadata, output cmdline */
- if (rte_ivshmem_metadata_create(METADATA_NAME) < 0)
- rte_exit(EXIT_FAILURE, "Cannot create IVSHMEM metadata\n");
-
- if (rte_ivshmem_metadata_add_memzone(ctrl_mz, METADATA_NAME))
- rte_exit(EXIT_FAILURE, "Cannot add memzone to IVSHMEM metadata\n");
-
- if (rte_ivshmem_metadata_add_mempool(l2fwd_ivshmem_pktmbuf_pool, METADATA_NAME))
- rte_exit(EXIT_FAILURE, "Cannot add mbuf mempool to IVSHMEM metadata\n");
-
- for (portid = 0; portid < nb_ports_available; portid++) {
- if (rte_ivshmem_metadata_add_ring(ctrl->vm_ports[portid].rx_ring,
- METADATA_NAME) < 0)
- rte_exit(EXIT_FAILURE, "Cannot add ring %s to IVSHMEM metadata\n",
- ctrl->vm_ports[portid].rx_ring->name);
- if (rte_ivshmem_metadata_add_ring(ctrl->vm_ports[portid].tx_ring,
- METADATA_NAME) < 0)
- rte_exit(EXIT_FAILURE, "Cannot add ring %s to IVSHMEM metadata\n",
- ctrl->vm_ports[portid].tx_ring->name);
- }
- generate_ivshmem_cmdline(METADATA_NAME);
-
- ctrl->nb_ports = nb_ports_available;
-
- printf("Waiting for VM to initialize...\n");
-
- /* wait for VM to initialize */
- while (ctrl->state != STATE_FWD) {
- if (ctrl->state == STATE_FAIL)
- rte_exit(EXIT_FAILURE, "VM reported failure\n");
-
- sleep(1);
- }
-
- printf("Done!\n");
-
- sigsetup();
-
- /* launch per-lcore init on every lcore */
- rte_eal_mp_remote_launch(l2fwd_ivshmem_launch_one_lcore, NULL, CALL_MASTER);
- RTE_LCORE_FOREACH_SLAVE(lcore_id) {
- if (rte_eal_wait_lcore(lcore_id) < 0)
- return -1;
- }
-
- if (ctrl->state == STATE_FAIL)
- rte_exit(EXIT_FAILURE, "VM reported failure\n");
-
- return 0;
-}
diff --git a/examples/l2fwd-ivshmem/include/common.h b/examples/l2fwd-ivshmem/include/common.h
deleted file mode 100644
index 8564d32..0000000
--- a/examples/l2fwd-ivshmem/include/common.h
+++ /dev/null
@@ -1,111 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#ifndef _IVSHMEM_COMMON_H_
-#define _IVSHMEM_COMMON_H_
-
-#define RTE_LOGTYPE_L2FWD_IVSHMEM RTE_LOGTYPE_USER1
-
-#define CTRL_MZ_NAME "CTRL_MEMZONE"
-#define MBUF_MP_NAME "MBUF_MEMPOOL"
-#define RX_RING_PREFIX "RX_"
-#define TX_RING_PREFIX "TX_"
-
-/* A tsc-based timer responsible for triggering statistics printout */
-#define TIMER_MILLISECOND 2000000ULL /* around 1ms at 2 Ghz */
-#define MAX_TIMER_PERIOD 86400 /* 1 day max */
-static int64_t timer_period = 10 * TIMER_MILLISECOND * 1000; /* default period is 10 seconds */
-
-#define DIM(x)\
- (sizeof(x)/sizeof(x)[0])
-
-#define MAX_PKT_BURST 32
-
-const struct rte_memzone * ctrl_mz;
-
-enum l2fwd_state {
- STATE_NONE = 0,
- STATE_FWD,
- STATE_EXIT,
- STATE_FAIL
-};
-
-/* Per-port statistics struct */
-struct port_statistics {
- uint64_t tx;
- uint64_t rx;
- uint64_t dropped;
-} __rte_cache_aligned;
-
-struct mbuf_table {
- unsigned len;
- struct rte_mbuf *m_table[MAX_PKT_BURST * 2]; /**< allow up to two bursts */
-};
-
-struct vm_port_param {
- struct rte_ring * rx_ring; /**< receiving ring for current port */
- struct rte_ring * tx_ring; /**< transmitting ring for current port */
- struct vm_port_param * dst; /**< current port's destination port */
- volatile struct port_statistics stats; /**< statistics for current port */
- struct ether_addr ethaddr; /**< Ethernet address of the port */
-};
-
-/* control structure, to synchronize host and VM */
-struct ivshmem_ctrl {
- rte_spinlock_t lock;
- uint8_t nb_ports; /**< total nr of ports */
- volatile enum l2fwd_state state; /**< report state */
- struct vm_port_param vm_ports[RTE_MAX_ETHPORTS];
-};
-
-struct ivshmem_ctrl * ctrl;
-
-static unsigned int l2fwd_ivshmem_rx_queue_per_lcore = 1;
-
-static void sighandler(int __rte_unused s)
-{
- ctrl->state = STATE_EXIT;
-}
-
-static void sigsetup(void)
-{
- struct sigaction sigIntHandler;
-
- sigIntHandler.sa_handler = sighandler;
- sigemptyset(&sigIntHandler.sa_mask);
- sigIntHandler.sa_flags = 0;
-
- sigaction(SIGINT, &sigIntHandler, NULL);
-}
-
-#endif /* _IVSHMEM_COMMON_H_ */
diff --git a/examples/packet_ordering/Makefile b/examples/packet_ordering/Makefile
index 9e080a3..de066c4 100644
--- a/examples/packet_ordering/Makefile
+++ b/examples/packet_ordering/Makefile
@@ -34,7 +34,7 @@ $(error "Please define RTE_SDK environment variable")
endif
# Default target, can be overridden by command line or environment
-RTE_TARGET ?= x86_64-ivshmem-linuxapp-gcc
+RTE_TARGET ?= x86_64-native-linuxapp-gcc
include $(RTE_SDK)/mk/rte.vars.mk
diff --git a/lib/Makefile b/lib/Makefile
index ca7c02f..990f23a 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -61,7 +61,6 @@ DIRS-$(CONFIG_RTE_LIBRTE_PDUMP) += librte_pdump
ifeq ($(CONFIG_RTE_EXEC_ENV_LINUXAPP),y)
DIRS-$(CONFIG_RTE_LIBRTE_KNI) += librte_kni
-DIRS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += librte_ivshmem
endif
include $(RTE_SDK)/mk/rte.subdir.mk
diff --git a/lib/librte_eal/common/eal_common_memzone.c b/lib/librte_eal/common/eal_common_memzone.c
index 1bd0a33..64f4e0a 100644
--- a/lib/librte_eal/common/eal_common_memzone.c
+++ b/lib/librte_eal/common/eal_common_memzone.c
@@ -337,19 +337,7 @@ rte_memzone_free(const struct rte_memzone *mz)
idx = ((uintptr_t)mz - (uintptr_t)mcfg->memzone);
idx = idx / sizeof(struct rte_memzone);
-#ifdef RTE_LIBRTE_IVSHMEM
- /*
- * If ioremap_addr is set, it's an IVSHMEM memzone and we cannot
- * free it.
- */
- if (mcfg->memzone[idx].ioremap_addr != 0) {
- rte_rwlock_write_unlock(&mcfg->mlock);
- return -EINVAL;
- }
-#endif
-
addr = mcfg->memzone[idx].addr;
-
if (addr == NULL)
ret = -EINVAL;
else if (mcfg->memzone_cnt == 0) {
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 857dc3e..0bda493 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -126,28 +126,6 @@ int rte_eal_log_init(const char *id, int facility);
*/
int rte_eal_pci_init(void);
-#ifdef RTE_LIBRTE_IVSHMEM
-/**
- * Init the memory from IVSHMEM devices
- *
- * This function is private to EAL.
- *
- * @return
- * 0 on success, negative on error
- */
-int rte_eal_ivshmem_init(void);
-
-/**
- * Init objects in IVSHMEM devices
- *
- * This function is private to EAL.
- *
- * @return
- * 0 on success, negative on error
- */
-int rte_eal_ivshmem_obj_init(void);
-#endif
-
struct rte_pci_driver;
struct rte_pci_device;
diff --git a/lib/librte_eal/common/include/rte_memory.h b/lib/librte_eal/common/include/rte_memory.h
index 0661109..d9e8c21 100644
--- a/lib/librte_eal/common/include/rte_memory.h
+++ b/lib/librte_eal/common/include/rte_memory.h
@@ -107,9 +107,6 @@ struct rte_memseg {
void *addr; /**< Start virtual address. */
uint64_t addr_64; /**< Makes sure addr is always 64 bits */
};
-#ifdef RTE_LIBRTE_IVSHMEM
- phys_addr_t ioremap_addr; /**< Real physical address inside the VM */
-#endif
size_t len; /**< Length of the segment. */
uint64_t hugepage_sz; /**< The pagesize of underlying memory */
int32_t socket_id; /**< NUMA socket ID. */
diff --git a/lib/librte_eal/common/include/rte_memzone.h b/lib/librte_eal/common/include/rte_memzone.h
index f69b5a8..dae98f5 100644
--- a/lib/librte_eal/common/include/rte_memzone.h
+++ b/lib/librte_eal/common/include/rte_memzone.h
@@ -82,9 +82,6 @@ struct rte_memzone {
void *addr; /**< Start virtual address. */
uint64_t addr_64; /**< Makes sure addr is always 64-bits */
};
-#ifdef RTE_LIBRTE_IVSHMEM
- phys_addr_t ioremap_addr; /**< Real physical address inside the VM */
-#endif
size_t len; /**< Length of the memzone. */
uint64_t hugepage_sz; /**< The page size of underlying memory */
@@ -256,12 +253,10 @@ const struct rte_memzone *rte_memzone_reserve_bounded(const char *name,
/**
* Free a memzone.
*
- * Note: an IVSHMEM zone cannot be freed.
- *
* @param mz
* A pointer to the memzone
* @return
- * -EINVAL - invalid parameter, IVSHMEM memzone.
+ * -EINVAL - invalid parameter.
* 0 - success
*/
int rte_memzone_free(const struct rte_memzone *mz);
diff --git a/lib/librte_eal/common/malloc_heap.c b/lib/librte_eal/common/malloc_heap.c
index 763fa32..267a4c6 100644
--- a/lib/librte_eal/common/malloc_heap.c
+++ b/lib/librte_eal/common/malloc_heap.c
@@ -221,14 +221,6 @@ rte_eal_malloc_heap_init(void)
for (ms = &mcfg->memseg[0], ms_cnt = 0;
(ms_cnt < RTE_MAX_MEMSEG) && (ms->len > 0);
ms_cnt++, ms++) {
-#ifdef RTE_LIBRTE_IVSHMEM
- /*
- * if segment has ioremap address set, it's an IVSHMEM segment and
- * it is not memory to allocate from.
- */
- if (ms->ioremap_addr != 0)
- continue;
-#endif
malloc_heap_add_memseg(&mcfg->malloc_heaps[ms->socket_id], ms);
}
diff --git a/lib/librte_eal/linuxapp/eal/Makefile b/lib/librte_eal/linuxapp/eal/Makefile
index 182729c..0baa571 100644
--- a/lib/librte_eal/linuxapp/eal/Makefile
+++ b/lib/librte_eal/linuxapp/eal/Makefile
@@ -44,12 +44,6 @@ VPATH += $(RTE_SDK)/lib/librte_eal/common
CFLAGS += -I$(SRCDIR)/include
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common/include
-ifeq ($(CONFIG_RTE_LIBRTE_IVSHMEM),y)
-# workaround for circular dependency eal -> ivshmem -> ring/mempool -> eal
-CFLAGS += -I$(RTE_SDK)/lib/librte_ring
-CFLAGS += -I$(RTE_SDK)/lib/librte_mempool
-CFLAGS += -I$(RTE_SDK)/lib/librte_ivshmem
-endif
CFLAGS += $(WERROR_FLAGS) -O3
LDLIBS += -ldl
@@ -76,9 +70,6 @@ SRCS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += eal_lcore.c
SRCS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += eal_timer.c
SRCS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += eal_interrupts.c
SRCS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += eal_alarm.c
-ifeq ($(CONFIG_RTE_LIBRTE_IVSHMEM),y)
-SRCS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += eal_ivshmem.c
-endif
# from common dir
SRCS-$(CONFIG_RTE_EXEC_ENV_LINUXAPP) += eal_common_lcore.c
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index 3fb2188..d5b81a3 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -797,11 +797,6 @@ rte_eal_init(int argc, char **argv)
rte_panic("Cannot init VFIO\n");
#endif
-#ifdef RTE_LIBRTE_IVSHMEM
- if (rte_eal_ivshmem_init() < 0)
- rte_panic("Cannot init IVSHMEM\n");
-#endif
-
if (rte_eal_memory_init() < 0)
rte_panic("Cannot init memory\n");
@@ -814,11 +809,6 @@ rte_eal_init(int argc, char **argv)
if (rte_eal_tailqs_init() < 0)
rte_panic("Cannot init tail queues for objects\n");
-#ifdef RTE_LIBRTE_IVSHMEM
- if (rte_eal_ivshmem_obj_init() < 0)
- rte_panic("Cannot init IVSHMEM objects\n");
-#endif
-
if (rte_eal_log_init(logid, internal_config.syslog_facility) < 0)
rte_panic("Cannot init logs\n");
diff --git a/lib/librte_eal/linuxapp/eal/eal_ivshmem.c b/lib/librte_eal/linuxapp/eal/eal_ivshmem.c
deleted file mode 100644
index 67b3caf..0000000
--- a/lib/librte_eal/linuxapp/eal/eal_ivshmem.c
+++ /dev/null
@@ -1,954 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#ifdef RTE_LIBRTE_IVSHMEM /* hide it from coverage */
-
-#include <stdint.h>
-#include <unistd.h>
-#include <inttypes.h>
-#include <sys/mman.h>
-#include <sys/file.h>
-#include <string.h>
-#include <sys/queue.h>
-
-#include <rte_log.h>
-#include <rte_pci.h>
-#include <rte_memory.h>
-#include <rte_eal.h>
-#include <rte_eal_memconfig.h>
-#include <rte_string_fns.h>
-#include <rte_errno.h>
-#include <rte_ring.h>
-#include <rte_malloc.h>
-#include <rte_common.h>
-#include <rte_ivshmem.h>
-
-#include "eal_internal_cfg.h"
-#include "eal_private.h"
-
-#define PCI_VENDOR_ID_IVSHMEM 0x1Af4
-#define PCI_DEVICE_ID_IVSHMEM 0x1110
-
-#define IVSHMEM_MAGIC 0x0BADC0DE
-
-#define IVSHMEM_RESOURCE_PATH "/sys/bus/pci/devices/%04x:%02x:%02x.%x/resource2"
-#define IVSHMEM_CONFIG_PATH "/var/run/.%s_ivshmem_config"
-
-#define PHYS 0x1
-#define VIRT 0x2
-#define IOREMAP 0x4
-#define FULL (PHYS|VIRT|IOREMAP)
-
-#define METADATA_SIZE_ALIGNED \
- (RTE_ALIGN_CEIL(sizeof(struct rte_ivshmem_metadata),pagesz))
-
-#define CONTAINS(x,y)\
- (((y).addr_64 >= (x).addr_64) && ((y).addr_64 < (x).addr_64 + (x).len))
-
-#define DIM(x) (sizeof(x)/sizeof(x[0]))
-
-struct ivshmem_pci_device {
- char path[PATH_MAX];
- phys_addr_t ioremap_addr;
-};
-
-/* data type to store in config */
-struct ivshmem_segment {
- struct rte_ivshmem_metadata_entry entry;
- uint64_t align;
- char path[PATH_MAX];
-};
-struct ivshmem_shared_config {
- struct ivshmem_segment segment[RTE_MAX_MEMSEG];
- uint32_t segment_idx;
- struct ivshmem_pci_device pci_devs[RTE_LIBRTE_IVSHMEM_MAX_PCI_DEVS];
- uint32_t pci_devs_idx;
-};
-static struct ivshmem_shared_config * ivshmem_config;
-static int memseg_idx;
-static int pagesz;
-
-/* Tailq heads to add rings to */
-TAILQ_HEAD(rte_ring_list, rte_tailq_entry);
-
-/*
- * Utility functions
- */
-
-static int
-is_ivshmem_device(struct rte_pci_device * dev)
-{
- return dev->id.vendor_id == PCI_VENDOR_ID_IVSHMEM
- && dev->id.device_id == PCI_DEVICE_ID_IVSHMEM;
-}
-
-static void *
-map_metadata(int fd, uint64_t len)
-{
- size_t metadata_len = sizeof(struct rte_ivshmem_metadata);
- size_t aligned_len = METADATA_SIZE_ALIGNED;
-
- return mmap(NULL, metadata_len, PROT_READ | PROT_WRITE,
- MAP_SHARED, fd, len - aligned_len);
-}
-
-static void
-unmap_metadata(void * ptr)
-{
- munmap(ptr, sizeof(struct rte_ivshmem_metadata));
-}
-
-static int
-has_ivshmem_metadata(int fd, uint64_t len)
-{
- struct rte_ivshmem_metadata metadata;
- void * ptr;
-
- ptr = map_metadata(fd, len);
-
- if (ptr == MAP_FAILED)
- return -1;
-
- metadata = *(struct rte_ivshmem_metadata*) (ptr);
-
- unmap_metadata(ptr);
-
- return metadata.magic_number == IVSHMEM_MAGIC;
-}
-
-static void
-remove_segment(struct ivshmem_segment * ms, int len, int idx)
-{
- int i;
-
- for (i = idx; i < len - 1; i++)
- memcpy(&ms[i], &ms[i+1], sizeof(struct ivshmem_segment));
- memset(&ms[len-1], 0, sizeof(struct ivshmem_segment));
-}
-
-static int
-overlap(const struct rte_memzone * mz1, const struct rte_memzone * mz2)
-{
- uint64_t start1, end1, start2, end2;
- uint64_t p_start1, p_end1, p_start2, p_end2;
- uint64_t i_start1, i_end1, i_start2, i_end2;
- int result = 0;
-
- /* gather virtual addresses */
- start1 = mz1->addr_64;
- end1 = mz1->addr_64 + mz1->len;
- start2 = mz2->addr_64;
- end2 = mz2->addr_64 + mz2->len;
-
- /* gather physical addresses */
- p_start1 = mz1->phys_addr;
- p_end1 = mz1->phys_addr + mz1->len;
- p_start2 = mz2->phys_addr;
- p_end2 = mz2->phys_addr + mz2->len;
-
- /* gather ioremap addresses */
- i_start1 = mz1->ioremap_addr;
- i_end1 = mz1->ioremap_addr + mz1->len;
- i_start2 = mz2->ioremap_addr;
- i_end2 = mz2->ioremap_addr + mz2->len;
-
- /* check for overlap in virtual addresses */
- if (start1 >= start2 && start1 < end2)
- result |= VIRT;
- if (start2 >= start1 && start2 < end1)
- result |= VIRT;
-
- /* check for overlap in physical addresses */
- if (p_start1 >= p_start2 && p_start1 < p_end2)
- result |= PHYS;
- if (p_start2 >= p_start1 && p_start2 < p_end1)
- result |= PHYS;
-
- /* check for overlap in ioremap addresses */
- if (i_start1 >= i_start2 && i_start1 < i_end2)
- result |= IOREMAP;
- if (i_start2 >= i_start1 && i_start2 < i_end1)
- result |= IOREMAP;
-
- return result;
-}
-
-static int
-adjacent(const struct rte_memzone * mz1, const struct rte_memzone * mz2)
-{
- uint64_t start1, end1, start2, end2;
- uint64_t p_start1, p_end1, p_start2, p_end2;
- uint64_t i_start1, i_end1, i_start2, i_end2;
- int result = 0;
-
- /* gather virtual addresses */
- start1 = mz1->addr_64;
- end1 = mz1->addr_64 + mz1->len;
- start2 = mz2->addr_64;
- end2 = mz2->addr_64 + mz2->len;
-
- /* gather physical addresses */
- p_start1 = mz1->phys_addr;
- p_end1 = mz1->phys_addr + mz1->len;
- p_start2 = mz2->phys_addr;
- p_end2 = mz2->phys_addr + mz2->len;
-
- /* gather ioremap addresses */
- i_start1 = mz1->ioremap_addr;
- i_end1 = mz1->ioremap_addr + mz1->len;
- i_start2 = mz2->ioremap_addr;
- i_end2 = mz2->ioremap_addr + mz2->len;
-
- /* check if segments are virtually adjacent */
- if (start1 == end2)
- result |= VIRT;
- if (start2 == end1)
- result |= VIRT;
-
- /* check if segments are physically adjacent */
- if (p_start1 == p_end2)
- result |= PHYS;
- if (p_start2 == p_end1)
- result |= PHYS;
-
- /* check if segments are ioremap-adjacent */
- if (i_start1 == i_end2)
- result |= IOREMAP;
- if (i_start2 == i_end1)
- result |= IOREMAP;
-
- return result;
-}
-
-static int
-has_adjacent_segments(struct ivshmem_segment * ms, int len)
-{
- int i, j;
-
- for (i = 0; i < len; i++)
- for (j = i + 1; j < len; j++) {
- /* we're only interested in fully adjacent segments; partially
- * adjacent segments can coexist.
- */
- if (adjacent(&ms[i].entry.mz, &ms[j].entry.mz) == FULL)
- return 1;
- }
- return 0;
-}
-
-static int
-has_overlapping_segments(struct ivshmem_segment * ms, int len)
-{
- int i, j;
-
- for (i = 0; i < len; i++)
- for (j = i + 1; j < len; j++)
- if (overlap(&ms[i].entry.mz, &ms[j].entry.mz))
- return 1;
- return 0;
-}
-
-static int
-seg_compare(const void * a, const void * b)
-{
- const struct ivshmem_segment * s1 = (const struct ivshmem_segment*) a;
- const struct ivshmem_segment * s2 = (const struct ivshmem_segment*) b;
-
- /* move unallocated zones to the end */
- if (s1->entry.mz.addr == NULL && s2->entry.mz.addr == NULL)
- return 0;
- if (s1->entry.mz.addr == 0)
- return 1;
- if (s2->entry.mz.addr == 0)
- return -1;
-
- return s1->entry.mz.phys_addr > s2->entry.mz.phys_addr;
-}
-
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
-static void
-entry_dump(struct rte_ivshmem_metadata_entry *e)
-{
- RTE_LOG(DEBUG, EAL, "\tvirt: %p-%p\n", e->mz.addr,
- RTE_PTR_ADD(e->mz.addr, e->mz.len));
- RTE_LOG(DEBUG, EAL, "\tphys: 0x%" PRIx64 "-0x%" PRIx64 "\n",
- e->mz.phys_addr,
- e->mz.phys_addr + e->mz.len);
- RTE_LOG(DEBUG, EAL, "\tio: 0x%" PRIx64 "-0x%" PRIx64 "\n",
- e->mz.ioremap_addr,
- e->mz.ioremap_addr + e->mz.len);
- RTE_LOG(DEBUG, EAL, "\tlen: 0x%" PRIx64 "\n", e->mz.len);
- RTE_LOG(DEBUG, EAL, "\toff: 0x%" PRIx64 "\n", e->offset);
-}
-#endif
-
-
-
-/*
- * Actual useful code
- */
-
-/* read through metadata mapped from the IVSHMEM device */
-static int
-read_metadata(char * path, int path_len, int fd, uint64_t flen)
-{
- struct rte_ivshmem_metadata metadata;
- struct rte_ivshmem_metadata_entry * entry;
- int idx, i;
- void * ptr;
-
- ptr = map_metadata(fd, flen);
-
- if (ptr == MAP_FAILED)
- return -1;
-
- metadata = *(struct rte_ivshmem_metadata*) (ptr);
-
- unmap_metadata(ptr);
-
- RTE_LOG(DEBUG, EAL, "Parsing metadata for \"%s\"\n", metadata.name);
-
- idx = ivshmem_config->segment_idx;
-
- for (i = 0; i < RTE_LIBRTE_IVSHMEM_MAX_ENTRIES &&
- idx <= RTE_MAX_MEMSEG; i++) {
-
- if (idx == RTE_MAX_MEMSEG) {
- RTE_LOG(ERR, EAL, "Not enough memory segments!\n");
- return -1;
- }
-
- entry = &metadata.entry[i];
-
- /* stop on uninitialized memzone */
- if (entry->mz.len == 0)
- break;
-
- /* copy metadata entry */
- memcpy(&ivshmem_config->segment[idx].entry, entry,
- sizeof(struct rte_ivshmem_metadata_entry));
-
- /* copy path */
- snprintf(ivshmem_config->segment[idx].path, path_len, "%s", path);
-
- idx++;
- }
- ivshmem_config->segment_idx = idx;
-
- return 0;
-}
-
-/* check through each segment and look for adjacent or overlapping ones. */
-static int
-cleanup_segments(struct ivshmem_segment * ms, int tbl_len)
-{
- struct ivshmem_segment * s, * tmp;
- int i, j, concat, seg_adjacent, seg_overlapping;
- uint64_t start1, start2, end1, end2, p_start1, p_start2, i_start1, i_start2;
-
- qsort(ms, tbl_len, sizeof(struct ivshmem_segment),
- seg_compare);
-
- while (has_overlapping_segments(ms, tbl_len) ||
- has_adjacent_segments(ms, tbl_len)) {
-
- for (i = 0; i < tbl_len; i++) {
- s = &ms[i];
-
- concat = 0;
-
- for (j = i + 1; j < tbl_len; j++) {
- tmp = &ms[j];
-
- /* check if this segment is overlapping with existing segment,
- * or is adjacent to existing segment */
- seg_overlapping = overlap(&s->entry.mz, &tmp->entry.mz);
- seg_adjacent = adjacent(&s->entry.mz, &tmp->entry.mz);
-
- /* check if segments fully overlap or are fully adjacent */
- if ((seg_adjacent == FULL) || (seg_overlapping == FULL)) {
-
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
- RTE_LOG(DEBUG, EAL, "Concatenating segments\n");
- RTE_LOG(DEBUG, EAL, "Segment %i:\n", i);
- entry_dump(&s->entry);
- RTE_LOG(DEBUG, EAL, "Segment %i:\n", j);
- entry_dump(&tmp->entry);
-#endif
-
- start1 = s->entry.mz.addr_64;
- start2 = tmp->entry.mz.addr_64;
- p_start1 = s->entry.mz.phys_addr;
- p_start2 = tmp->entry.mz.phys_addr;
- i_start1 = s->entry.mz.ioremap_addr;
- i_start2 = tmp->entry.mz.ioremap_addr;
- end1 = s->entry.mz.addr_64 + s->entry.mz.len;
- end2 = tmp->entry.mz.addr_64 + tmp->entry.mz.len;
-
- /* settle for minimum start address and maximum length */
- s->entry.mz.addr_64 = RTE_MIN(start1, start2);
- s->entry.mz.phys_addr = RTE_MIN(p_start1, p_start2);
- s->entry.mz.ioremap_addr = RTE_MIN(i_start1, i_start2);
- s->entry.offset = RTE_MIN(s->entry.offset, tmp->entry.offset);
- s->entry.mz.len = RTE_MAX(end1, end2) - s->entry.mz.addr_64;
- concat = 1;
-
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
- RTE_LOG(DEBUG, EAL, "Resulting segment:\n");
- entry_dump(&s->entry);
-
-#endif
- }
- /* if segments not fully overlap, we have an error condition.
- * adjacent segments can coexist.
- */
- else if (seg_overlapping > 0) {
- RTE_LOG(ERR, EAL, "Segments %i and %i overlap!\n", i, j);
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
- RTE_LOG(DEBUG, EAL, "Segment %i:\n", i);
- entry_dump(&s->entry);
- RTE_LOG(DEBUG, EAL, "Segment %i:\n", j);
- entry_dump(&tmp->entry);
-#endif
- return -1;
- }
- if (concat)
- break;
- }
- /* if we concatenated, remove segment at j */
- if (concat) {
- remove_segment(ms, tbl_len, j);
- tbl_len--;
- break;
- }
- }
- }
-
- return tbl_len;
-}
-
-static int
-create_shared_config(void)
-{
- char path[PATH_MAX];
- int fd;
-
- /* build ivshmem config file path */
- snprintf(path, sizeof(path), IVSHMEM_CONFIG_PATH,
- internal_config.hugefile_prefix);
-
- fd = open(path, O_CREAT | O_RDWR, 0600);
-
- if (fd < 0) {
- RTE_LOG(ERR, EAL, "Could not open %s: %s\n", path, strerror(errno));
- return -1;
- }
-
- /* try ex-locking first - if the file is locked, we have a problem */
- if (flock(fd, LOCK_EX | LOCK_NB) == -1) {
- RTE_LOG(ERR, EAL, "Locking %s failed: %s\n", path, strerror(errno));
- close(fd);
- return -1;
- }
-
- if (ftruncate(fd, sizeof(struct ivshmem_shared_config)) < 0) {
- RTE_LOG(ERR, EAL, "ftruncate failed: %s\n", strerror(errno));
- return -1;
- }
-
- ivshmem_config = mmap(NULL, sizeof(struct ivshmem_shared_config),
- PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
-
- if (ivshmem_config == MAP_FAILED)
- return -1;
-
- memset(ivshmem_config, 0, sizeof(struct ivshmem_shared_config));
-
- /* change the exclusive lock we got earlier to a shared lock */
- if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
- RTE_LOG(ERR, EAL, "Locking %s failed: %s \n", path, strerror(errno));
- return -1;
- }
-
- close(fd);
-
- return 0;
-}
-
-/* open shared config file and, if present, map the config.
- * having no config file is not an error condition, as we later check if
- * ivshmem_config is NULL (if it is, that means nothing was mapped). */
-static int
-open_shared_config(void)
-{
- char path[PATH_MAX];
- int fd;
-
- /* build ivshmem config file path */
- snprintf(path, sizeof(path), IVSHMEM_CONFIG_PATH,
- internal_config.hugefile_prefix);
-
- fd = open(path, O_RDONLY);
-
- /* if the file doesn't exist, just return success */
- if (fd < 0 && errno == ENOENT)
- return 0;
- /* else we have an error condition */
- else if (fd < 0) {
- RTE_LOG(ERR, EAL, "Could not open %s: %s\n",
- path, strerror(errno));
- return -1;
- }
-
- /* try ex-locking first - if the lock *does* succeed, this means it's a
- * stray config file, so it should be deleted.
- */
- if (flock(fd, LOCK_EX | LOCK_NB) != -1) {
-
- /* if we can't remove the file, something is wrong */
- if (unlink(path) < 0) {
- RTE_LOG(ERR, EAL, "Could not remove %s: %s\n", path,
- strerror(errno));
- return -1;
- }
-
- /* release the lock */
- flock(fd, LOCK_UN);
- close(fd);
-
- /* return success as having a stray config file is equivalent to not
- * having config file at all.
- */
- return 0;
- }
-
- ivshmem_config = mmap(NULL, sizeof(struct ivshmem_shared_config),
- PROT_READ, MAP_SHARED, fd, 0);
-
- if (ivshmem_config == MAP_FAILED)
- return -1;
-
- /* place a shared lock on config file */
- if (flock(fd, LOCK_SH | LOCK_NB) == -1) {
- RTE_LOG(ERR, EAL, "Locking %s failed: %s \n", path, strerror(errno));
- return -1;
- }
-
- close(fd);
-
- return 0;
-}
-
-/*
- * This function does the following:
- *
- * 1) Builds a table of ivshmem_segments with proper offset alignment
- * 2) Cleans up that table so that we don't have any overlapping or adjacent
- * memory segments
- * 3) Creates memsegs from this table and maps them into memory.
- */
-static inline int
-map_all_segments(void)
-{
- struct ivshmem_segment ms_tbl[RTE_MAX_MEMSEG];
- struct ivshmem_pci_device * pci_dev;
- struct rte_mem_config * mcfg;
- struct ivshmem_segment * seg;
- int fd, fd_zero;
- unsigned i, j;
- struct rte_memzone mz;
- struct rte_memseg ms;
- void * base_addr;
- uint64_t align, len;
- phys_addr_t ioremap_addr;
-
- ioremap_addr = 0;
-
- memset(ms_tbl, 0, sizeof(ms_tbl));
- memset(&mz, 0, sizeof(struct rte_memzone));
- memset(&ms, 0, sizeof(struct rte_memseg));
-
- /* first, build a table of memsegs to map, to avoid failed mmaps due to
- * overlaps
- */
- for (i = 0; i < ivshmem_config->segment_idx && i <= RTE_MAX_MEMSEG; i++) {
- if (i == RTE_MAX_MEMSEG) {
- RTE_LOG(ERR, EAL, "Too many segments requested!\n");
- return -1;
- }
-
- seg = &ivshmem_config->segment[i];
-
- /* copy segment to table */
- memcpy(&ms_tbl[i], seg, sizeof(struct ivshmem_segment));
-
- /* find ioremap addr */
- for (j = 0; j < DIM(ivshmem_config->pci_devs); j++) {
- pci_dev = &ivshmem_config->pci_devs[j];
- if (!strncmp(pci_dev->path, seg->path, sizeof(pci_dev->path))) {
- ioremap_addr = pci_dev->ioremap_addr;
- break;
- }
- }
- if (ioremap_addr == 0) {
- RTE_LOG(ERR, EAL, "Cannot find ioremap addr!\n");
- return -1;
- }
-
- /* work out alignments */
- align = seg->entry.mz.addr_64 -
- RTE_ALIGN_FLOOR(seg->entry.mz.addr_64, 0x1000);
- len = RTE_ALIGN_CEIL(seg->entry.mz.len + align, 0x1000);
-
- /* save original alignments */
- ms_tbl[i].align = align;
-
- /* create a memory zone */
- mz.addr_64 = seg->entry.mz.addr_64 - align;
- mz.len = len;
- mz.hugepage_sz = seg->entry.mz.hugepage_sz;
- mz.phys_addr = seg->entry.mz.phys_addr - align;
-
- /* find true physical address */
- mz.ioremap_addr = ioremap_addr + seg->entry.offset - align;
-
- ms_tbl[i].entry.offset = seg->entry.offset - align;
-
- memcpy(&ms_tbl[i].entry.mz, &mz, sizeof(struct rte_memzone));
- }
-
- /* clean up the segments */
- memseg_idx = cleanup_segments(ms_tbl, ivshmem_config->segment_idx);
-
- if (memseg_idx < 0)
- return -1;
-
- mcfg = rte_eal_get_configuration()->mem_config;
-
- fd_zero = open("/dev/zero", O_RDWR);
-
- if (fd_zero < 0) {
- RTE_LOG(ERR, EAL, "Cannot open /dev/zero: %s\n", strerror(errno));
- return -1;
- }
-
- /* create memsegs and put them into DPDK memory */
- for (i = 0; i < (unsigned) memseg_idx; i++) {
-
- seg = &ms_tbl[i];
-
- ms.addr_64 = seg->entry.mz.addr_64;
- ms.hugepage_sz = seg->entry.mz.hugepage_sz;
- ms.len = seg->entry.mz.len;
- ms.nchannel = rte_memory_get_nchannel();
- ms.nrank = rte_memory_get_nrank();
- ms.phys_addr = seg->entry.mz.phys_addr;
- ms.ioremap_addr = seg->entry.mz.ioremap_addr;
- ms.socket_id = seg->entry.mz.socket_id;
-
- base_addr = mmap(ms.addr, ms.len,
- PROT_READ | PROT_WRITE, MAP_PRIVATE, fd_zero, 0);
-
- if (base_addr == MAP_FAILED || base_addr != ms.addr) {
- RTE_LOG(ERR, EAL, "Cannot map /dev/zero!\n");
- return -1;
- }
-
- fd = open(seg->path, O_RDWR);
-
- if (fd < 0) {
- RTE_LOG(ERR, EAL, "Cannot open %s: %s\n", seg->path,
- strerror(errno));
- return -1;
- }
-
- munmap(ms.addr, ms.len);
-
- base_addr = mmap(ms.addr, ms.len,
- PROT_READ | PROT_WRITE, MAP_SHARED, fd,
- seg->entry.offset);
-
-
- if (base_addr == MAP_FAILED || base_addr != ms.addr) {
- RTE_LOG(ERR, EAL, "Cannot map segment into memory: "
- "expected %p got %p (%s)\n", ms.addr, base_addr,
- strerror(errno));
- return -1;
- }
-
- RTE_LOG(DEBUG, EAL, "Memory segment mapped: %p (len %" PRIx64 ") at "
- "offset 0x%" PRIx64 "\n",
- ms.addr, ms.len, seg->entry.offset);
-
- /* put the pointers back into their real positions using original
- * alignment */
- ms.addr_64 += seg->align;
- ms.phys_addr += seg->align;
- ms.ioremap_addr += seg->align;
- ms.len -= seg->align;
-
- /* at this point, the rest of DPDK memory is not initialized, so we
- * expect memsegs to be empty */
- memcpy(&mcfg->memseg[i], &ms,
- sizeof(struct rte_memseg));
-
- close(fd);
-
- RTE_LOG(DEBUG, EAL, "IVSHMEM segment found, size: 0x%lx\n",
- ms.len);
- }
-
- return 0;
-}
-
-/* this happens at a later stage, after general EAL memory initialization */
-int
-rte_eal_ivshmem_obj_init(void)
-{
- struct rte_ring_list* ring_list = NULL;
- struct rte_mem_config * mcfg;
- struct ivshmem_segment * seg;
- struct rte_memzone * mz;
- struct rte_ring * r;
- struct rte_tailq_entry *te;
- unsigned i, ms, idx;
- uint64_t offset;
-
- /* secondary process would not need any object discovery - it'll all
- * already be in shared config */
- if (rte_eal_process_type() != RTE_PROC_PRIMARY || ivshmem_config == NULL)
- return 0;
-
- /* check that we have an initialised ring tail queue */
- ring_list = RTE_TAILQ_LOOKUP(RTE_TAILQ_RING_NAME, rte_ring_list);
- if (ring_list == NULL) {
- RTE_LOG(ERR, EAL, "No rte_ring tailq found!\n");
- return -1;
- }
-
- mcfg = rte_eal_get_configuration()->mem_config;
-
- /* create memzones */
- for (i = 0; i < ivshmem_config->segment_idx && i <= RTE_MAX_MEMZONE; i++) {
-
- seg = &ivshmem_config->segment[i];
-
- /* add memzone */
- if (mcfg->memzone_cnt == RTE_MAX_MEMZONE) {
- RTE_LOG(ERR, EAL, "No more memory zones available!\n");
- return -1;
- }
-
- idx = mcfg->memzone_cnt;
-
- RTE_LOG(DEBUG, EAL, "Found memzone: '%s' at %p (len 0x%" PRIx64 ")\n",
- seg->entry.mz.name, seg->entry.mz.addr, seg->entry.mz.len);
-
- memcpy(&mcfg->memzone[idx], &seg->entry.mz,
- sizeof(struct rte_memzone));
-
- /* find ioremap address */
- for (ms = 0; ms <= RTE_MAX_MEMSEG; ms++) {
- if (ms == RTE_MAX_MEMSEG) {
- RTE_LOG(ERR, EAL, "Physical address of segment not found!\n");
- return -1;
- }
- if (CONTAINS(mcfg->memseg[ms], mcfg->memzone[idx])) {
- offset = mcfg->memzone[idx].addr_64 -
- mcfg->memseg[ms].addr_64;
- mcfg->memzone[idx].ioremap_addr = mcfg->memseg[ms].ioremap_addr +
- offset;
- break;
- }
- }
-
- mcfg->memzone_cnt++;
- }
-
- rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
-
- /* find rings */
- for (i = 0; i < mcfg->memzone_cnt; i++) {
- mz = &mcfg->memzone[i];
-
- /* check if memzone has a ring prefix */
- if (strncmp(mz->name, RTE_RING_MZ_PREFIX,
- sizeof(RTE_RING_MZ_PREFIX) - 1) != 0)
- continue;
-
- r = (struct rte_ring*) (mz->addr_64);
-
- te = rte_zmalloc("RING_TAILQ_ENTRY", sizeof(*te), 0);
- if (te == NULL) {
- RTE_LOG(ERR, EAL, "Cannot allocate ring tailq entry!\n");
- return -1;
- }
-
- te->data = (void *) r;
-
- TAILQ_INSERT_TAIL(ring_list, te, next);
-
- RTE_LOG(DEBUG, EAL, "Found ring: '%s' at %p\n", r->name, mz->addr);
- }
- rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
-
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
- rte_memzone_dump(stdout);
- rte_ring_list_dump(stdout);
-#endif
-
- return 0;
-}
-
-/* initialize ivshmem structures */
-int rte_eal_ivshmem_init(void)
-{
- struct rte_pci_device * dev;
- struct rte_pci_resource * res;
- int fd, ret;
- char path[PATH_MAX];
-
- /* initialize everything to 0 */
- memset(path, 0, sizeof(path));
- ivshmem_config = NULL;
-
- pagesz = getpagesize();
-
- RTE_LOG(DEBUG, EAL, "Searching for IVSHMEM devices...\n");
-
- if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
-
- if (open_shared_config() < 0) {
- RTE_LOG(ERR, EAL, "Could not open IVSHMEM config!\n");
- return -1;
- }
- }
- else {
-
- TAILQ_FOREACH(dev, &pci_device_list, next) {
-
- if (is_ivshmem_device(dev)) {
-
- /* IVSHMEM memory is always on BAR2 */
- res = &dev->mem_resource[2];
-
- /* if we don't have a BAR2 */
- if (res->len == 0)
- continue;
-
- /* construct pci device path */
- snprintf(path, sizeof(path), IVSHMEM_RESOURCE_PATH,
- dev->addr.domain, dev->addr.bus, dev->addr.devid,
- dev->addr.function);
-
- /* try to find memseg */
- fd = open(path, O_RDWR);
- if (fd < 0) {
- RTE_LOG(ERR, EAL, "Could not open %s\n", path);
- return -1;
- }
-
- /* check if it's a DPDK IVSHMEM device */
- ret = has_ivshmem_metadata(fd, res->len);
-
- /* is DPDK device */
- if (ret == 1) {
-
- /* config file creation is deferred until the first
- * DPDK device is found. then, it has to be created
- * only once. */
- if (ivshmem_config == NULL &&
- create_shared_config() < 0) {
- RTE_LOG(ERR, EAL, "Could not create IVSHMEM config!\n");
- close(fd);
- return -1;
- }
-
- if (read_metadata(path, sizeof(path), fd, res->len) < 0) {
- RTE_LOG(ERR, EAL, "Could not read metadata from"
- " device %02x:%02x.%x!\n", dev->addr.bus,
- dev->addr.devid, dev->addr.function);
- close(fd);
- return -1;
- }
-
- if (ivshmem_config->pci_devs_idx == RTE_LIBRTE_IVSHMEM_MAX_PCI_DEVS) {
- RTE_LOG(WARNING, EAL,
- "IVSHMEM PCI device limit exceeded. Increase "
- "CONFIG_RTE_LIBRTE_IVSHMEM_MAX_PCI_DEVS in "
- "your config file.\n");
- break;
- }
-
- RTE_LOG(INFO, EAL, "Found IVSHMEM device %02x:%02x.%x\n",
- dev->addr.bus, dev->addr.devid, dev->addr.function);
-
- ivshmem_config->pci_devs[ivshmem_config->pci_devs_idx].ioremap_addr = res->phys_addr;
- snprintf(ivshmem_config->pci_devs[ivshmem_config->pci_devs_idx].path,
- sizeof(ivshmem_config->pci_devs[ivshmem_config->pci_devs_idx].path),
- "%s", path);
-
- ivshmem_config->pci_devs_idx++;
- }
- /* failed to read */
- else if (ret < 0) {
- RTE_LOG(ERR, EAL, "Could not read IVSHMEM device: %s\n",
- strerror(errno));
- close(fd);
- return -1;
- }
- /* not a DPDK device */
- else
- RTE_LOG(DEBUG, EAL, "Skipping non-DPDK IVSHMEM device\n");
-
- /* close the BAR fd */
- close(fd);
- }
- }
- }
-
- /* ivshmem_config is not NULL only if config was created and/or mapped */
- if (ivshmem_config) {
- if (map_all_segments() < 0) {
- RTE_LOG(ERR, EAL, "Mapping IVSHMEM segments failed!\n");
- return -1;
- }
- }
- else {
- RTE_LOG(DEBUG, EAL, "No IVSHMEM configuration found! \n");
- }
-
- return 0;
-}
-
-#endif
diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c b/lib/librte_eal/linuxapp/eal/eal_memory.c
index 41e0a92..992a1b1 100644
--- a/lib/librte_eal/linuxapp/eal/eal_memory.c
+++ b/lib/librte_eal/linuxapp/eal/eal_memory.c
@@ -1436,15 +1436,8 @@ rte_eal_hugepage_init(void)
free(tmp_hp);
tmp_hp = NULL;
- /* find earliest free memseg - this is needed because in case of IVSHMEM,
- * segments might have already been initialized */
- for (j = 0; j < RTE_MAX_MEMSEG; j++)
- if (mcfg->memseg[j].addr == NULL) {
- /* move to previous segment and exit loop */
- j--;
- break;
- }
-
+ /* first memseg index shall be 0 after incrementing it below */
+ j = -1;
for (i = 0; i < nr_hugefiles; i++) {
new_memseg = 0;
@@ -1597,15 +1590,6 @@ rte_eal_hugepage_attach(void)
if (mcfg->memseg[s].len == 0)
break;
-#ifdef RTE_LIBRTE_IVSHMEM
- /*
- * if segment has ioremap address set, it's an IVSHMEM segment and
- * doesn't need mapping as it was already mapped earlier
- */
- if (mcfg->memseg[s].ioremap_addr != 0)
- continue;
-#endif
-
/*
* fdzero is mmapped to get a contiguous block of virtual
* addresses of the appropriate memseg size.
@@ -1644,16 +1628,6 @@ rte_eal_hugepage_attach(void)
void *addr, *base_addr;
uintptr_t offset = 0;
size_t mapping_size;
-#ifdef RTE_LIBRTE_IVSHMEM
- /*
- * if segment has ioremap address set, it's an IVSHMEM segment and
- * doesn't need mapping as it was already mapped earlier
- */
- if (mcfg->memseg[s].ioremap_addr != 0) {
- s++;
- continue;
- }
-#endif
/*
* free previously mapped memory so we can map the
* hugepages into the space
diff --git a/lib/librte_ivshmem/Makefile b/lib/librte_ivshmem/Makefile
deleted file mode 100644
index c099438..0000000
--- a/lib/librte_ivshmem/Makefile
+++ /dev/null
@@ -1,54 +0,0 @@
-# BSD LICENSE
-#
-# Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
-# All rights reserved.
-#
-# Redistribution and use in source and binary forms, with or without
-# modification, are permitted provided that the following conditions
-# are met:
-#
-# * Redistributions of source code must retain the above copyright
-# notice, this list of conditions and the following disclaimer.
-# * Redistributions in binary form must reproduce the above copyright
-# notice, this list of conditions and the following disclaimer in
-# the documentation and/or other materials provided with the
-# distribution.
-# * Neither the name of Intel Corporation nor the names of its
-# contributors may be used to endorse or promote products derived
-# from this software without specific prior written permission.
-#
-# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-include $(RTE_SDK)/mk/rte.vars.mk
-
-# library name
-LIB = librte_ivshmem.a
-
-CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR) -O3
-
-EXPORT_MAP := rte_ivshmem_version.map
-
-LIBABIVER := 1
-
-# all source are stored in SRCS-y
-SRCS-$(CONFIG_RTE_LIBRTE_IVSHMEM) := rte_ivshmem.c
-
-# install includes
-SYMLINK-$(CONFIG_RTE_LIBRTE_IVSHMEM)-include := rte_ivshmem.h
-
-# this lib needs EAL, ring and mempool
-DEPDIRS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += lib/librte_eal
-DEPDIRS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += lib/librte_ring
-DEPDIRS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += lib/librte_mempool
-
-include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/lib/librte_ivshmem/rte_ivshmem.c b/lib/librte_ivshmem/rte_ivshmem.c
deleted file mode 100644
index c26edb6..0000000
--- a/lib/librte_ivshmem/rte_ivshmem.c
+++ /dev/null
@@ -1,919 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-#include <fcntl.h>
-#include <limits.h>
-#include <unistd.h>
-#include <sys/mman.h>
-#include <string.h>
-#include <stdio.h>
-
-#include <rte_eal_memconfig.h>
-#include <rte_memory.h>
-#include <rte_ivshmem.h>
-#include <rte_string_fns.h>
-#include <rte_common.h>
-#include <rte_log.h>
-#include <rte_debug.h>
-#include <rte_spinlock.h>
-#include <rte_common.h>
-#include <rte_malloc.h>
-
-#include "rte_ivshmem.h"
-
-#define IVSHMEM_CONFIG_FILE_FMT "/var/run/.dpdk_ivshmem_metadata_%s"
-#define IVSHMEM_QEMU_CMD_LINE_HEADER_FMT "-device ivshmem,size=%" PRIu64 "M,shm=fd%s"
-#define IVSHMEM_QEMU_CMD_FD_FMT ":%s:0x%" PRIx64 ":0x%" PRIx64
-#define IVSHMEM_QEMU_CMDLINE_BUFSIZE 1024
-#define IVSHMEM_MAX_PAGES (1 << 12)
-#define adjacent(x,y) (((x).phys_addr+(x).len)==(y).phys_addr)
-#define METADATA_SIZE_ALIGNED \
- (RTE_ALIGN_CEIL(sizeof(struct rte_ivshmem_metadata),pagesz))
-
-#define GET_PAGEMAP_ADDR(in,addr,dlm,err) \
-{ \
- char *end; \
- errno = 0; \
- addr = strtoull((in), &end, 16); \
- if (errno != 0 || *end != (dlm)) { \
- RTE_LOG(ERR, EAL, err); \
- goto error; \
- } \
- (in) = end + 1; \
-}
-
-static int pagesz;
-
-struct memseg_cache_entry {
- char filepath[PATH_MAX];
- uint64_t offset;
- uint64_t len;
-};
-
-struct ivshmem_config {
- struct rte_ivshmem_metadata * metadata;
- struct memseg_cache_entry memseg_cache[IVSHMEM_MAX_PAGES];
- /**< account for multiple files per segment case */
- struct flock lock;
- rte_spinlock_t sl;
-};
-
-static struct ivshmem_config
-ivshmem_global_config[RTE_LIBRTE_IVSHMEM_MAX_METADATA_FILES];
-
-static rte_spinlock_t global_cfg_sl;
-
-static struct ivshmem_config *
-get_config_by_name(const char * name)
-{
- struct rte_ivshmem_metadata * config;
- unsigned i;
-
- for (i = 0; i < RTE_DIM(ivshmem_global_config); i++) {
- config = ivshmem_global_config[i].metadata;
- if (config == NULL)
- return NULL;
- if (strncmp(name, config->name, IVSHMEM_NAME_LEN) == 0)
- return &ivshmem_global_config[i];
- }
-
- return NULL;
-}
-
-static int
-overlap(const struct rte_memzone * s1, const struct rte_memzone * s2)
-{
- uint64_t start1, end1, start2, end2;
-
- start1 = s1->addr_64;
- end1 = s1->addr_64 + s1->len;
- start2 = s2->addr_64;
- end2 = s2->addr_64 + s2->len;
-
- if (start1 >= start2 && start1 < end2)
- return 1;
- if (start2 >= start1 && start2 < end1)
- return 1;
-
- return 0;
-}
-
-static struct rte_memzone *
-get_memzone_by_addr(const void * addr)
-{
- struct rte_memzone * tmp, * mz;
- struct rte_mem_config * mcfg;
- int i;
-
- mcfg = rte_eal_get_configuration()->mem_config;
- mz = NULL;
-
- /* find memzone for the ring */
- for (i = 0; i < RTE_MAX_MEMZONE; i++) {
- tmp = &mcfg->memzone[i];
-
- if (tmp->addr_64 == (uint64_t) addr) {
- mz = tmp;
- break;
- }
- }
-
- return mz;
-}
-
-static int
-entry_compare(const void * a, const void * b)
-{
- const struct rte_ivshmem_metadata_entry * e1 =
- (const struct rte_ivshmem_metadata_entry*) a;
- const struct rte_ivshmem_metadata_entry * e2 =
- (const struct rte_ivshmem_metadata_entry*) b;
-
- /* move unallocated zones to the end */
- if (e1->mz.addr == NULL && e2->mz.addr == NULL)
- return 0;
- if (e1->mz.addr == 0)
- return 1;
- if (e2->mz.addr == 0)
- return -1;
-
- return e1->mz.phys_addr > e2->mz.phys_addr;
-}
-
-/* fills hugepage cache entry for a given start virt_addr */
-static int
-get_hugefile_by_virt_addr(uint64_t virt_addr, struct memseg_cache_entry * e)
-{
- uint64_t start_addr, end_addr;
- char *start,*path_end;
- char buf[PATH_MAX*2];
- FILE *f;
-
- start = NULL;
- path_end = NULL;
- start_addr = 0;
-
- memset(e->filepath, 0, sizeof(e->filepath));
-
- /* open /proc/self/maps */
- f = fopen("/proc/self/maps", "r");
- if (f == NULL) {
- RTE_LOG(ERR, EAL, "cannot open /proc/self/maps!\n");
- return -1;
- }
-
- /* parse maps */
- while (fgets(buf, sizeof(buf), f) != NULL) {
-
- /* get endptr to end of start addr */
- start = buf;
-
- GET_PAGEMAP_ADDR(start,start_addr,'-',
- "Cannot find start address in maps!\n");
-
- /* if start address is bigger than our address, skip */
- if (start_addr > virt_addr)
- continue;
-
- GET_PAGEMAP_ADDR(start,end_addr,' ',
- "Cannot find end address in maps!\n");
-
- /* if end address is less than our address, skip */
- if (end_addr <= virt_addr)
- continue;
-
- /* find where the path starts */
- start = strstr(start, "/");
-
- if (start == NULL)
- continue;
-
- /* at this point, we know that this is our map.
- * now let's find the file */
- path_end = strstr(start, "\n");
- break;
- }
-
- if (path_end == NULL) {
- RTE_LOG(ERR, EAL, "Hugefile path not found!\n");
- goto error;
- }
-
- /* calculate offset and copy the file path */
- snprintf(e->filepath, RTE_PTR_DIFF(path_end, start) + 1, "%s", start);
-
- e->offset = virt_addr - start_addr;
-
- fclose(f);
-
- return 0;
-error:
- fclose(f);
- return -1;
-}
-
-/*
- * This is a complex function. What it does is the following:
- * 1. Goes through metadata and gets list of hugepages involved
- * 2. Sorts the hugepages by size (1G first)
- * 3. Goes through metadata again and writes correct offsets
- * 4. Goes through pages and finds out their filenames, offsets etc.
- */
-static int
-build_config(struct rte_ivshmem_metadata * metadata)
-{
- struct rte_ivshmem_metadata_entry * e_local;
- struct memseg_cache_entry * ms_local;
- struct rte_memseg pages[IVSHMEM_MAX_PAGES];
- struct rte_ivshmem_metadata_entry *entry;
- struct memseg_cache_entry * c_entry, * prev_entry;
- struct ivshmem_config * config;
- unsigned i, j, mz_iter, ms_iter;
- uint64_t biggest_len;
- int biggest_idx;
-
- /* return error if we try to use an unknown config file */
- config = get_config_by_name(metadata->name);
- if (config == NULL) {
- RTE_LOG(ERR, EAL, "Cannot find IVSHMEM config %s!\n", metadata->name);
- goto fail_e;
- }
-
- memset(pages, 0, sizeof(pages));
-
- e_local = malloc(sizeof(config->metadata->entry));
- if (e_local == NULL)
- goto fail_e;
- ms_local = malloc(sizeof(config->memseg_cache));
- if (ms_local == NULL)
- goto fail_ms;
-
-
- /* make local copies before doing anything */
- memcpy(e_local, config->metadata->entry, sizeof(config->metadata->entry));
- memcpy(ms_local, config->memseg_cache, sizeof(config->memseg_cache));
-
- qsort(e_local, RTE_DIM(config->metadata->entry), sizeof(struct rte_ivshmem_metadata_entry),
- entry_compare);
-
- /* first pass - collect all huge pages */
- for (mz_iter = 0; mz_iter < RTE_DIM(config->metadata->entry); mz_iter++) {
-
- entry = &e_local[mz_iter];
-
- uint64_t start_addr = RTE_ALIGN_FLOOR(entry->mz.addr_64,
- entry->mz.hugepage_sz);
- uint64_t offset = entry->mz.addr_64 - start_addr;
- uint64_t len = RTE_ALIGN_CEIL(entry->mz.len + offset,
- entry->mz.hugepage_sz);
-
- if (entry->mz.addr_64 == 0 || start_addr == 0 || len == 0)
- continue;
-
- int start_page;
-
- /* find first unused page - mz are phys_addr sorted so we don't have to
- * look out for holes */
- for (i = 0; i < RTE_DIM(pages); i++) {
-
- /* skip if we already have this page */
- if (pages[i].addr_64 == start_addr) {
- start_addr += entry->mz.hugepage_sz;
- len -= entry->mz.hugepage_sz;
- continue;
- }
- /* we found a new page */
- else if (pages[i].addr_64 == 0) {
- start_page = i;
- break;
- }
- }
- if (i == RTE_DIM(pages)) {
- RTE_LOG(ERR, EAL, "Cannot find unused page!\n");
- goto fail;
- }
-
- /* populate however many pages the memzone has */
- for (i = start_page; i < RTE_DIM(pages) && len != 0; i++) {
-
- pages[i].addr_64 = start_addr;
- pages[i].len = entry->mz.hugepage_sz;
- start_addr += entry->mz.hugepage_sz;
- len -= entry->mz.hugepage_sz;
- }
- /* if there's still length left */
- if (len != 0) {
- RTE_LOG(ERR, EAL, "Not enough space for pages!\n");
- goto fail;
- }
- }
-
- /* second pass - sort pages by size */
- for (i = 0; i < RTE_DIM(pages); i++) {
-
- if (pages[i].addr == NULL)
- break;
-
- biggest_len = 0;
- biggest_idx = -1;
-
- /*
- * browse all entries starting at 'i', and find the
- * entry with the smallest addr
- */
- for (j=i; j< RTE_DIM(pages); j++) {
- if (pages[j].addr == NULL)
- break;
- if (biggest_len == 0 ||
- pages[j].len > biggest_len) {
- biggest_len = pages[j].len;
- biggest_idx = j;
- }
- }
-
- /* should not happen */
- if (biggest_idx == -1) {
- RTE_LOG(ERR, EAL, "Error sorting by size!\n");
- goto fail;
- }
- if (i != (unsigned) biggest_idx) {
- struct rte_memseg tmp;
-
- memcpy(&tmp, &pages[biggest_idx], sizeof(struct rte_memseg));
-
- /* we don't want to break contiguousness, so instead of just
- * swapping segments, we move all the preceding segments to the
- * right and then put the old segment @ biggest_idx in place of
- * segment @ i */
- for (j = biggest_idx - 1; j >= i; j--) {
- memcpy(&pages[j+1], &pages[j], sizeof(struct rte_memseg));
- memset(&pages[j], 0, sizeof(struct rte_memseg));
- if (j == 0)
- break;
- }
-
- /* put old biggest segment to its new place */
- memcpy(&pages[i], &tmp, sizeof(struct rte_memseg));
- }
- }
-
- /* third pass - write correct offsets */
- for (mz_iter = 0; mz_iter < RTE_DIM(config->metadata->entry); mz_iter++) {
-
- uint64_t offset = 0;
-
- entry = &e_local[mz_iter];
-
- if (entry->mz.addr_64 == 0)
- break;
-
- /* find page for current memzone */
- for (i = 0; i < RTE_DIM(pages); i++) {
- /* we found our page */
- if (entry->mz.addr_64 >= pages[i].addr_64 &&
- entry->mz.addr_64 < pages[i].addr_64 + pages[i].len) {
- entry->offset = (entry->mz.addr_64 - pages[i].addr_64) +
- offset;
- break;
- }
- offset += pages[i].len;
- }
- if (i == RTE_DIM(pages)) {
- RTE_LOG(ERR, EAL, "Page not found!\n");
- goto fail;
- }
- }
-
- ms_iter = 0;
- prev_entry = NULL;
-
- /* fourth pass - create proper memseg cache */
- for (i = 0; i < RTE_DIM(pages) &&
- ms_iter <= RTE_DIM(config->memseg_cache); i++) {
- if (pages[i].addr_64 == 0)
- break;
-
-
- if (ms_iter == RTE_DIM(pages)) {
- RTE_LOG(ERR, EAL, "The universe has collapsed!\n");
- goto fail;
- }
-
- c_entry = &ms_local[ms_iter];
- c_entry->len = pages[i].len;
-
- if (get_hugefile_by_virt_addr(pages[i].addr_64, c_entry) < 0)
- goto fail;
-
- /* if previous entry has the same filename and is contiguous,
- * clear current entry and increase previous entry's length
- */
- if (prev_entry != NULL &&
- strncmp(c_entry->filepath, prev_entry->filepath,
- sizeof(c_entry->filepath)) == 0 &&
- prev_entry->offset + prev_entry->len == c_entry->offset) {
- prev_entry->len += pages[i].len;
- memset(c_entry, 0, sizeof(struct memseg_cache_entry));
- }
- else {
- prev_entry = c_entry;
- ms_iter++;
- }
- }
-
- /* update current configuration with new valid data */
- memcpy(config->metadata->entry, e_local, sizeof(config->metadata->entry));
- memcpy(config->memseg_cache, ms_local, sizeof(config->memseg_cache));
-
- free(ms_local);
- free(e_local);
-
- return 0;
-fail:
- free(ms_local);
-fail_ms:
- free(e_local);
-fail_e:
- return -1;
-}
-
-static int
-add_memzone_to_metadata(const struct rte_memzone * mz,
- struct ivshmem_config * config)
-{
- struct rte_ivshmem_metadata_entry * entry;
- unsigned i, idx;
- struct rte_mem_config *mcfg;
-
- if (mz->len == 0) {
- RTE_LOG(ERR, EAL, "Trying to add an empty memzone\n");
- return -1;
- }
-
- rte_spinlock_lock(&config->sl);
-
- mcfg = rte_eal_get_configuration()->mem_config;
-
- /* it prevents the memzone being freed while we add it to the metadata */
- rte_rwlock_write_lock(&mcfg->mlock);
-
- /* find free slot in this config */
- for (i = 0; i < RTE_DIM(config->metadata->entry); i++) {
- entry = &config->metadata->entry[i];
-
- if (&entry->mz.addr_64 != 0 && overlap(mz, &entry->mz)) {
- RTE_LOG(ERR, EAL, "Overlapping memzones!\n");
- goto fail;
- }
-
- /* if addr is zero, the memzone is probably free */
- if (entry->mz.addr_64 == 0) {
- RTE_LOG(DEBUG, EAL, "Adding memzone '%s' at %p to metadata %s\n",
- mz->name, mz->addr, config->metadata->name);
- memcpy(&entry->mz, mz, sizeof(struct rte_memzone));
-
- /* run config file parser */
- if (build_config(config->metadata) < 0)
- goto fail;
-
- break;
- }
- }
-
- /* if we reached the maximum, that means we have no place in config */
- if (i == RTE_DIM(config->metadata->entry)) {
- RTE_LOG(ERR, EAL, "No space left in IVSHMEM metadata %s!\n",
- config->metadata->name);
- goto fail;
- }
-
- idx = ((uintptr_t)mz - (uintptr_t)mcfg->memzone);
- idx = idx / sizeof(struct rte_memzone);
-
- /* mark the memzone not freeable */
- mcfg->memzone[idx].ioremap_addr = mz->phys_addr;
-
- rte_rwlock_write_unlock(&mcfg->mlock);
- rte_spinlock_unlock(&config->sl);
- return 0;
-fail:
- rte_rwlock_write_unlock(&mcfg->mlock);
- rte_spinlock_unlock(&config->sl);
- return -1;
-}
-
-static int
-add_ring_to_metadata(const struct rte_ring * r,
- struct ivshmem_config * config)
-{
- struct rte_memzone * mz;
-
- mz = get_memzone_by_addr(r);
-
- if (!mz) {
- RTE_LOG(ERR, EAL, "Cannot find memzone for ring!\n");
- return -1;
- }
-
- return add_memzone_to_metadata(mz, config);
-}
-
-static int
-add_mempool_memzone_to_metadata(const void *addr,
- struct ivshmem_config *config)
-{
- struct rte_memzone *mz;
-
- mz = get_memzone_by_addr(addr);
-
- if (!mz) {
- RTE_LOG(ERR, EAL, "Cannot find memzone for mempool!\n");
- return -1;
- }
-
- return add_memzone_to_metadata(mz, config);
-}
-
-static int
-add_mempool_to_metadata(const struct rte_mempool *mp,
- struct ivshmem_config *config)
-{
- struct rte_mempool_memhdr *memhdr;
- int ret;
-
- ret = add_mempool_memzone_to_metadata(mp, config);
- if (ret < 0)
- return -1;
-
- STAILQ_FOREACH(memhdr, &mp->mem_list, next) {
- ret = add_mempool_memzone_to_metadata(memhdr->addr, config);
- if (ret < 0)
- return -1;
- }
-
- /* mempool consists of memzone and ring */
- return add_ring_to_metadata(mp->pool_data, config);
-}
-
-int
-rte_ivshmem_metadata_add_ring(const struct rte_ring * r, const char * name)
-{
- struct ivshmem_config * config;
-
- if (name == NULL || r == NULL)
- return -1;
-
- config = get_config_by_name(name);
-
- if (config == NULL) {
- RTE_LOG(ERR, EAL, "Cannot find IVSHMEM config %s!\n", name);
- return -1;
- }
-
- return add_ring_to_metadata(r, config);
-}
-
-int
-rte_ivshmem_metadata_add_memzone(const struct rte_memzone * mz, const char * name)
-{
- struct ivshmem_config * config;
-
- if (name == NULL || mz == NULL)
- return -1;
-
- config = get_config_by_name(name);
-
- if (config == NULL) {
- RTE_LOG(ERR, EAL, "Cannot find IVSHMEM config %s!\n", name);
- return -1;
- }
-
- return add_memzone_to_metadata(mz, config);
-}
-
-int
-rte_ivshmem_metadata_add_mempool(const struct rte_mempool * mp, const char * name)
-{
- struct ivshmem_config * config;
-
- if (name == NULL || mp == NULL)
- return -1;
-
- config = get_config_by_name(name);
-
- if (config == NULL) {
- RTE_LOG(ERR, EAL, "Cannot find IVSHMEM config %s!\n", name);
- return -1;
- }
-
- return add_mempool_to_metadata(mp, config);
-}
-
-static inline void
-ivshmem_config_path(char *buffer, size_t bufflen, const char *name)
-{
- snprintf(buffer, bufflen, IVSHMEM_CONFIG_FILE_FMT, name);
-}
-
-
-
-static inline
-void *ivshmem_metadata_create(const char *name, size_t size,
- struct flock *lock)
-{
- int retval, fd;
- void *metadata_addr;
- char pathname[PATH_MAX];
-
- ivshmem_config_path(pathname, sizeof(pathname), name);
-
- fd = open(pathname, O_RDWR | O_CREAT, 0660);
- if (fd < 0) {
- RTE_LOG(ERR, EAL, "Cannot open '%s'\n", pathname);
- return NULL;
- }
-
- size = METADATA_SIZE_ALIGNED;
-
- retval = fcntl(fd, F_SETLK, lock);
- if (retval < 0){
- close(fd);
- RTE_LOG(ERR, EAL, "Cannot create lock on '%s'. Is another "
- "process using it?\n", pathname);
- return NULL;
- }
-
- retval = ftruncate(fd, size);
- if (retval < 0){
- close(fd);
- RTE_LOG(ERR, EAL, "Cannot resize '%s'\n", pathname);
- return NULL;
- }
-
- metadata_addr = mmap(NULL, size,
- PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
-
- if (metadata_addr == MAP_FAILED){
- RTE_LOG(ERR, EAL, "Cannot mmap memory for '%s'\n", pathname);
-
- /* we don't care if we can't unlock */
- fcntl(fd, F_UNLCK, lock);
- close(fd);
-
- return NULL;
- }
-
- return metadata_addr;
-}
-
-int rte_ivshmem_metadata_create(const char *name)
-{
- struct ivshmem_config * ivshmem_config;
- unsigned index;
-
- if (pagesz == 0)
- pagesz = getpagesize();
-
- if (name == NULL)
- return -1;
-
- rte_spinlock_lock(&global_cfg_sl);
-
- for (index = 0; index < RTE_DIM(ivshmem_global_config); index++) {
- if (ivshmem_global_config[index].metadata == NULL) {
- ivshmem_config = &ivshmem_global_config[index];
- break;
- }
- }
-
- if (index == RTE_DIM(ivshmem_global_config)) {
- RTE_LOG(ERR, EAL, "Cannot create more ivshmem config files. "
- "Maximum has been reached\n");
- rte_spinlock_unlock(&global_cfg_sl);
- return -1;
- }
-
- ivshmem_config->lock.l_type = F_WRLCK;
- ivshmem_config->lock.l_whence = SEEK_SET;
-
- ivshmem_config->lock.l_start = 0;
- ivshmem_config->lock.l_len = METADATA_SIZE_ALIGNED;
-
- ivshmem_global_config[index].metadata = ((struct rte_ivshmem_metadata *)
- ivshmem_metadata_create(
- name,
- sizeof(struct rte_ivshmem_metadata),
- &ivshmem_config->lock));
-
- if (ivshmem_global_config[index].metadata == NULL) {
- rte_spinlock_unlock(&global_cfg_sl);
- return -1;
- }
-
- /* Metadata setup */
- memset(ivshmem_config->metadata, 0, sizeof(struct rte_ivshmem_metadata));
- ivshmem_config->metadata->magic_number = IVSHMEM_MAGIC;
- snprintf(ivshmem_config->metadata->name,
- sizeof(ivshmem_config->metadata->name), "%s", name);
-
- rte_spinlock_unlock(&global_cfg_sl);
-
- return 0;
-}
-
-int
-rte_ivshmem_metadata_cmdline_generate(char *buffer, unsigned size, const char *name)
-{
- const struct memseg_cache_entry * ms_cache, *entry;
- struct ivshmem_config * config;
- char cmdline[IVSHMEM_QEMU_CMDLINE_BUFSIZE], *cmdline_ptr;
- char cfg_file_path[PATH_MAX];
- unsigned remaining_len, tmplen, iter;
- uint64_t shared_mem_size, zero_size, total_size;
-
- if (buffer == NULL || name == NULL)
- return -1;
-
- config = get_config_by_name(name);
-
- if (config == NULL) {
- RTE_LOG(ERR, EAL, "Config %s not found!\n", name);
- return -1;
- }
-
- rte_spinlock_lock(&config->sl);
-
- /* prepare metadata file path */
- snprintf(cfg_file_path, sizeof(cfg_file_path), IVSHMEM_CONFIG_FILE_FMT,
- config->metadata->name);
-
- ms_cache = config->memseg_cache;
-
- cmdline_ptr = cmdline;
- remaining_len = sizeof(cmdline);
-
- shared_mem_size = 0;
- iter = 0;
-
- while ((ms_cache[iter].len != 0) && (iter < RTE_DIM(config->metadata->entry))) {
-
- entry = &ms_cache[iter];
-
- /* Offset and sizes within the current pathname */
- tmplen = snprintf(cmdline_ptr, remaining_len, IVSHMEM_QEMU_CMD_FD_FMT,
- entry->filepath, entry->offset, entry->len);
-
- shared_mem_size += entry->len;
-
- cmdline_ptr = RTE_PTR_ADD(cmdline_ptr, tmplen);
- remaining_len -= tmplen;
-
- if (remaining_len == 0) {
- RTE_LOG(ERR, EAL, "Command line too long!\n");
- rte_spinlock_unlock(&config->sl);
- return -1;
- }
-
- iter++;
- }
-
- total_size = rte_align64pow2(shared_mem_size + METADATA_SIZE_ALIGNED);
- zero_size = total_size - shared_mem_size - METADATA_SIZE_ALIGNED;
-
- /* add /dev/zero to command-line to fill the space */
- tmplen = snprintf(cmdline_ptr, remaining_len, IVSHMEM_QEMU_CMD_FD_FMT,
- "/dev/zero",
- (uint64_t)0x0,
- zero_size);
-
- cmdline_ptr = RTE_PTR_ADD(cmdline_ptr, tmplen);
- remaining_len -= tmplen;
-
- if (remaining_len == 0) {
- RTE_LOG(ERR, EAL, "Command line too long!\n");
- rte_spinlock_unlock(&config->sl);
- return -1;
- }
-
- /* add metadata file to the end of command-line */
- tmplen = snprintf(cmdline_ptr, remaining_len, IVSHMEM_QEMU_CMD_FD_FMT,
- cfg_file_path,
- (uint64_t)0x0,
- METADATA_SIZE_ALIGNED);
-
- cmdline_ptr = RTE_PTR_ADD(cmdline_ptr, tmplen);
- remaining_len -= tmplen;
-
- if (remaining_len == 0) {
- RTE_LOG(ERR, EAL, "Command line too long!\n");
- rte_spinlock_unlock(&config->sl);
- return -1;
- }
-
- /* if current length of the command line is bigger than the buffer supplied
- * by the user, or if command-line is bigger than what IVSHMEM accepts */
- if ((sizeof(cmdline) - remaining_len) > size) {
- RTE_LOG(ERR, EAL, "Buffer is too short!\n");
- rte_spinlock_unlock(&config->sl);
- return -1;
- }
- /* complete the command-line */
- snprintf(buffer, size,
- IVSHMEM_QEMU_CMD_LINE_HEADER_FMT,
- total_size >> 20,
- cmdline);
-
- rte_spinlock_unlock(&config->sl);
-
- return 0;
-}
-
-void
-rte_ivshmem_metadata_dump(FILE *f, const char *name)
-{
- unsigned i = 0;
- struct ivshmem_config * config;
- struct rte_ivshmem_metadata_entry *entry;
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
- uint64_t addr;
- uint64_t end, hugepage_sz;
- struct memseg_cache_entry e;
-#endif
-
- if (name == NULL)
- return;
-
- /* return error if we try to use an unknown config file */
- config = get_config_by_name(name);
- if (config == NULL) {
- RTE_LOG(ERR, EAL, "Cannot find IVSHMEM config %s!\n", name);
- return;
- }
-
- rte_spinlock_lock(&config->sl);
-
- entry = &config->metadata->entry[0];
-
- while (entry->mz.addr != NULL && i < RTE_DIM(config->metadata->entry)) {
-
- fprintf(f, "Entry %u: name:<%-20s>, phys:0x%-15lx, len:0x%-15lx, "
- "virt:%-15p, off:0x%-15lx\n",
- i,
- entry->mz.name,
- entry->mz.phys_addr,
- entry->mz.len,
- entry->mz.addr,
- entry->offset);
- i++;
-
-#ifdef RTE_LIBRTE_IVSHMEM_DEBUG
- fprintf(f, "\tHugepage files:\n");
-
- hugepage_sz = entry->mz.hugepage_sz;
- addr = RTE_ALIGN_FLOOR(entry->mz.addr_64, hugepage_sz);
- end = addr + RTE_ALIGN_CEIL(entry->mz.len + (entry->mz.addr_64 - addr),
- hugepage_sz);
-
- for (; addr < end; addr += hugepage_sz) {
- memset(&e, 0, sizeof(e));
-
- get_hugefile_by_virt_addr(addr, &e);
-
- fprintf(f, "\t0x%"PRIx64 "-0x%" PRIx64 " offset: 0x%" PRIx64 " %s\n",
- addr, addr + hugepage_sz, e.offset, e.filepath);
- }
-#endif
- entry++;
- }
-
- rte_spinlock_unlock(&config->sl);
-}
diff --git a/lib/librte_ivshmem/rte_ivshmem.h b/lib/librte_ivshmem/rte_ivshmem.h
deleted file mode 100644
index a5d36d6..0000000
--- a/lib/librte_ivshmem/rte_ivshmem.h
+++ /dev/null
@@ -1,165 +0,0 @@
-/*-
- * BSD LICENSE
- *
- * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
- * All rights reserved.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of Intel Corporation nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- */
-
-#ifndef RTE_IVSHMEM_H_
-#define RTE_IVSHMEM_H_
-
-#include <rte_memzone.h>
-#include <rte_mempool.h>
-
-/**
- * @file
- *
- * The RTE IVSHMEM interface provides functions to create metadata files
- * describing memory segments to be shared via QEMU IVSHMEM.
- */
-
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#define IVSHMEM_MAGIC 0x0BADC0DE
-#define IVSHMEM_NAME_LEN 32
-
-/**
- * Structure that holds IVSHMEM shared metadata entry.
- */
-struct rte_ivshmem_metadata_entry {
- struct rte_memzone mz; /**< shared memzone */
- uint64_t offset; /**< offset of memzone within IVSHMEM device */
-};
-
-/**
- * Structure that holds IVSHMEM metadata.
- */
-struct rte_ivshmem_metadata {
- int magic_number; /**< magic number */
- char name[IVSHMEM_NAME_LEN]; /**< name of the metadata file */
- struct rte_ivshmem_metadata_entry entry[RTE_LIBRTE_IVSHMEM_MAX_ENTRIES];
- /**< metadata entries */
-};
-
-/**
- * Creates metadata file with a given name
- *
- * @param name
- * Name of metadata file to be created
- *
- * @return
- * - On success, zero
- * - On failure, a negative value
- */
-int rte_ivshmem_metadata_create(const char * name);
-
-/**
- * Adds memzone to a specific metadata file
- *
- * @param mz
- * Memzone to be added
- * @param md_name
- * Name of metadata file for the memzone to be added to
- *
- * @return
- * - On success, zero
- * - On failure, a negative value
- */
-int rte_ivshmem_metadata_add_memzone(const struct rte_memzone * mz,
- const char * md_name);
-
-/**
- * Adds a ring descriptor to a specific metadata file
- *
- * @param r
- * Ring descriptor to be added
- * @param md_name
- * Name of metadata file for the ring to be added to
- *
- * @return
- * - On success, zero
- * - On failure, a negative value
- */
-int rte_ivshmem_metadata_add_ring(const struct rte_ring * r,
- const char * md_name);
-
-/**
- * Adds a mempool to a specific metadata file
- *
- * @param mp
- * Mempool to be added
- * @param md_name
- * Name of metadata file for the mempool to be added to
- *
- * @return
- * - On success, zero
- * - On failure, a negative value
- */
-int rte_ivshmem_metadata_add_mempool(const struct rte_mempool * mp,
- const char * md_name);
-
-
-/**
- * Generates the QEMU command-line for IVSHMEM device for a given metadata file.
- * This function is to be called after all the objects were added.
- *
- * @param buffer
- * Buffer to be filled with the command line arguments.
- * @param size
- * Size of the buffer.
- * @param name
- * Name of metadata file to generate QEMU command-line parameters for
- *
- * @return
- * - On success, zero
- * - On failure, a negative value
- */
-int rte_ivshmem_metadata_cmdline_generate(char *buffer, unsigned size,
- const char *name);
-
-
-/**
- * Dump all metadata entries from a given metadata file to the console.
- *
- * @param f
- * A pointer to a file for output
- * @name
- * Name of the metadata file to be dumped to console.
- */
-void rte_ivshmem_metadata_dump(FILE *f, const char *name);
-
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* RTE_IVSHMEM_H_ */
diff --git a/lib/librte_ivshmem/rte_ivshmem_version.map b/lib/librte_ivshmem/rte_ivshmem_version.map
deleted file mode 100644
index 5a393dd..0000000
--- a/lib/librte_ivshmem/rte_ivshmem_version.map
+++ /dev/null
@@ -1,12 +0,0 @@
-DPDK_2.0 {
- global:
-
- rte_ivshmem_metadata_add_mempool;
- rte_ivshmem_metadata_add_memzone;
- rte_ivshmem_metadata_add_ring;
- rte_ivshmem_metadata_cmdline_generate;
- rte_ivshmem_metadata_create;
- rte_ivshmem_metadata_dump;
-
- local: *;
-};
diff --git a/mk/rte.app.mk b/mk/rte.app.mk
index eb28e11..1a0095b 100644
--- a/mk/rte.app.mk
+++ b/mk/rte.app.mk
@@ -62,7 +62,6 @@ _LDLIBS-y += -L$(RTE_SDK_BIN)/lib
ifeq ($(CONFIG_RTE_EXEC_ENV_LINUXAPP),y)
_LDLIBS-$(CONFIG_RTE_LIBRTE_KNI) += -lrte_kni
-_LDLIBS-$(CONFIG_RTE_LIBRTE_IVSHMEM) += -lrte_ivshmem
endif
_LDLIBS-$(CONFIG_RTE_LIBRTE_PIPELINE) += -lrte_pipeline
--
1.9.1
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v1] doc: add template release notes for 16.11
@ 2016-07-29 11:23 6% John McNamara
0 siblings, 0 replies; 200+ results
From: John McNamara @ 2016-07-29 11:23 UTC (permalink / raw)
To: dev; +Cc: John McNamara
Add template release notes for DPDK 16.11 with inline
comments and explanations of the various sections.
Signed-off-by: John McNamara <john.mcnamara@intel.com>
---
doc/guides/rel_notes/index.rst | 1 +
doc/guides/rel_notes/release_16_11.rst | 205 +++++++++++++++++++++++++++++++++
2 files changed, 206 insertions(+)
create mode 100644 doc/guides/rel_notes/release_16_11.rst
diff --git a/doc/guides/rel_notes/index.rst b/doc/guides/rel_notes/index.rst
index 52c63b4..7e51b2c 100644
--- a/doc/guides/rel_notes/index.rst
+++ b/doc/guides/rel_notes/index.rst
@@ -36,6 +36,7 @@ Release Notes
:numbered:
rel_description
+ release_16_11
release_16_07
release_16_04
release_2_2
diff --git a/doc/guides/rel_notes/release_16_11.rst b/doc/guides/rel_notes/release_16_11.rst
new file mode 100644
index 0000000..a6e3307
--- /dev/null
+++ b/doc/guides/rel_notes/release_16_11.rst
@@ -0,0 +1,205 @@
+DPDK Release 16.11
+==================
+
+.. **Read this first.**
+
+ The text below explains how to update the release notes.
+
+ Use proper spelling, capitalization and punctuation in all sections.
+
+ Variable and config names should be quoted as fixed width text: ``LIKE_THIS``.
+
+ Build the docs and view the output file to ensure the changes are correct::
+
+ make doc-guides-html
+
+ firefox build/doc/html/guides/rel_notes/release_16_11.html
+
+
+New Features
+------------
+
+.. This section should contain new features added in this release. Sample format:
+
+ * **Add a title in the past tense with a full stop.**
+
+ Add a short 1-2 sentence description in the past tense. The description
+ should be enough to allow someone scanning the release notes to understand
+ the new feature.
+
+ If the feature adds a lot of sub-features you can use a bullet list like this.
+
+ * Added feature foo to do something.
+ * Enhanced feature bar to do something else.
+
+ Refer to the previous release notes for examples.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+Resolved Issues
+---------------
+
+.. This section should contain bug fixes added to the relevant sections. Sample format:
+
+ * **code/section Fixed issue in the past tense with a full stop.**
+
+ Add a short 1-2 sentence description of the resolved issue in the past tense.
+ The title should contain the code/lib section like a commit message.
+ Add the entries in alphabetic order in the relevant sections below.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+EAL
+~~~
+
+
+Drivers
+~~~~~~~
+
+
+Libraries
+~~~~~~~~~
+
+
+Examples
+~~~~~~~~
+
+
+Other
+~~~~~
+
+
+Known Issues
+------------
+
+.. This section should contain new known issues in this release. Sample format:
+
+ * **Add title in present tense with full stop.**
+
+ Add a short 1-2 sentence description of the known issue in the present
+ tense. Add information on any known workarounds.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+API Changes
+-----------
+
+.. This section should contain API changes. Sample format:
+
+ * Add a short 1-2 sentence description of the API change. Use fixed width
+ quotes for ``rte_function_names`` or ``rte_struct_names``. Use the past tense.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+ABI Changes
+-----------
+
+.. This section should contain ABI changes. Sample format:
+
+ * Add a short 1-2 sentence description of the ABI change that was announced in
+ the previous releases and made in this release. Use fixed width quotes for
+ ``rte_function_names`` or ``rte_struct_names``. Use the past tense.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+
+Shared Library Versions
+-----------------------
+
+.. Update any library version updated in this release and prepend with a ``+``
+ sign, like this:
+
+ libethdev.so.4
+ librte_acl.so.2
+ + librte_cfgfile.so.2
+ librte_cmdline.so.2
+
+
+
+The libraries prepended with a plus sign were incremented in this version.
+
+.. code-block:: diff
+
+ libethdev.so.4
+ librte_acl.so.2
+ librte_cfgfile.so.2
+ librte_cmdline.so.2
+ librte_cryptodev.so.1
+ librte_distributor.so.1
+ librte_eal.so.2
+ librte_hash.so.2
+ librte_ip_frag.so.1
+ librte_ivshmem.so.1
+ librte_jobstats.so.1
+ librte_kni.so.2
+ librte_kvargs.so.1
+ librte_lpm.so.2
+ librte_mbuf.so.2
+ librte_mempool.so.2
+ librte_meter.so.1
+ librte_pdump.so.1
+ librte_pipeline.so.3
+ librte_pmd_bond.so.1
+ librte_pmd_ring.so.2
+ librte_port.so.3
+ librte_power.so.1
+ librte_reorder.so.1
+ librte_ring.so.1
+ librte_sched.so.1
+ librte_table.so.2
+ librte_timer.so.1
+ librte_vhost.so.3
+
+
+Tested Platforms
+----------------
+
+.. This section should contain a list of platforms that were tested with this release.
+
+ The format is:
+
+ #. Platform name.
+
+ * Platform details.
+ * Platform details.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+Tested NICs
+-----------
+
+.. This section should contain a list of NICs that were tested with this release.
+
+ The format is:
+
+ #. NIC name.
+
+ * NIC details.
+ * NIC details.
+
+ This section is a comment. Make sure to start the actual text at the margin.
+
+
+Tested OSes
+-----------
+
+.. This section should contain a list of OSes that were tested with this release.
+ The format is as follows, in alphabetical order:
+
+ * CentOS 7.0
+ * Fedora 23
+ * Fedora 24
+ * FreeBSD 10.3
+ * Red Hat Enterprise Linux 7.2
+ * SUSE Enterprise Linux 12
+ * Ubuntu 15.10
+ * Ubuntu 16.04 LTS
+ * Wind River Linux 8
+
+ This section is a comment. Make sure to start the actual text at the margin.
--
2.7.4
^ permalink raw reply [relevance 6%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change of struct rte_port_source_params and rte_port_sink_params
2016-07-27 10:42 7% ` Thomas Monjalon
@ 2016-07-28 18:28 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-28 18:28 UTC (permalink / raw)
To: Zhang, Roy Fan
Cc: Dumitrescu, Cristian, dev, Panu Matilainen, Singh, Jasvinder
2016-07-27 12:42, Thomas Monjalon:
> 2016-07-27 10:08, Dumitrescu, Cristian:
> > As Thomas mentioned, today is probably the last day to discuss ABI changes. This one is pretty small and straightforward, any issues with it?
> >
> > Panu had a concern that the change from "char *" to "const char *" is too small to be regarded as ABI breakage and we should simply go ahead and do it. My conservative proposal was to put a notice anyway.
> >
> > Nonetheless, what I would like to get from Thomas and Panu is a path forward for this now:
> > a) If we agree to consider this an ABI change, please merge the notice for 16.7;
>
> Panu was noticing 3 things (and I agree with them):
> - it is an API change
> - they can be grouped in only one list item
> - it is better to wait having more changes to break an API
>
> About the third point, in this specific case, I think it is acceptable because:
> - it should not break the ABI
> - the impact of the API change is really small
> - I'm not sure the packet framework should be considered as a DPDK API.
>
> > b) If we agree this is too small for an ABI change, please let us agree now
> > to accept our quick patch for 16.11 for this change.
>
> For an API deprecation notice (reworded),
> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Applied, thanks
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-27 8:33 4% ` Thomas Monjalon
@ 2016-07-28 18:04 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-28 18:04 UTC (permalink / raw)
To: Olivier Matz; +Cc: dev, jerin.jacob, bruce.richardson
> > For 16.11, the mbuf structure will be modified implying ABI breakage.
> > Some discussions already took place here:
> > http://www.dpdk.org/dev/patchwork/patch/12878/
> >
> > Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
>
> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: John Daley <johndale@cisco.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Applied, thanks
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 16:25 7% ` Jerin Jacob
@ 2016-07-28 17:07 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-28 17:07 UTC (permalink / raw)
To: Jerin Jacob, Ananyev, Konstantin, Tomasz Kulasek; +Cc: dev
2016-07-28 21:55, Jerin Jacob:
> On Thu, Jul 28, 2016 at 04:52:45PM +0200, Thomas Monjalon wrote:
> > 2016-07-28 19:29, Jerin Jacob:
> > > Above things worries me, I wouldn't have cared if the changes are not comes
> > > in fastpath and I don't think this sort of issues will never get fixed any time
> > > soon in this community.
> > >
> > > So I given up.
> >
> > I feel something goes wrong here but I cannot understand your sentence.
> > Please could you reword/explain Jerin?
>
> I guess you have removed the context from the email. Never mind.
>
> 1) IMHO, Introducing a new fast path API which has "performance impact"
> on existing other PMD should get the consensus from the other PMD maintainers.
> At least, bare minimum, send a patch much in advance with the
> implementation of ethdev API as well as PMD
> driver implementation to get feedback from other developers _before_ ABI
> change announcement rather just debating on hypothetical points.
I totally agree with you and it was my first comment in this thread:
http://dpdk.org/ml/archives/dev/2016-July/044366.html
Unfortunately it is difficult to have a formal process so it is not
so strict currently. You are welcome to suggest how to improve the
process for the next releases.
> 2) What I can understand from the discussion is that it is the
> workaround for an HW limitation.
> At this point, I am not sure tx_prep is the only way to address it and
> do other PMD have similar
> restriction?. If yes, Can we have abstract it in a proper way the usage
> part will be very clear from PMD and application perspective?
I feel the tx_prep can be interesting to solve a current problem.
However, as you say, there are maybe other solutions to consider.
That's why I think we can keep this deprecation notice and follow up
with a patch-based discussion. We will be able to discard this change
if something better is found.
As an example, we have just removed a deprecation notice which has
never been implemented:
http://dpdk.org/browse/dpdk/commit/?id=16695af340
I know this process is not perfect and the ethdev API is far from perfect,
so we must continue improving our process to define a good API.
Konstantin, Tomasz,
I generally prefer waiting for a consensus. For this case, I'll make an
exception and apply the deprecation notice.
Please make an effort to better explain your next patches and meet
a clear consensus. We'll review your patches very carefully and keep
the right to reject them.
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 14:52 4% ` Thomas Monjalon
@ 2016-07-28 16:25 7% ` Jerin Jacob
2016-07-28 17:07 4% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Jerin Jacob @ 2016-07-28 16:25 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: Ananyev, Konstantin, dev
On Thu, Jul 28, 2016 at 04:52:45PM +0200, Thomas Monjalon wrote:
> 2016-07-28 19:29, Jerin Jacob:
> > Above things worries me, I wouldn't have cared if the changes are not comes
> > in fastpath and I don't think this sort of issues will never get fixed any time
> > soon in this community.
> >
> > So I given up.
>
> I feel something goes wrong here but I cannot understand your sentence.
> Please could you reword/explain Jerin?
I guess you have removed the context from the email. Never mind.
1) IMHO, Introducing a new fast path API which has "performance impact"
on existing other PMD should get the consensus from the other PMD maintainers.
At least, bare minimum, send a patch much in advance with the
implementation of ethdev API as well as PMD
driver implementation to get feedback from other developers _before_ ABI
change announcement rather just debating on hypothetical points.
2) What I can understand from the discussion is that it is the
workaround for an HW limitation.
At this point, I am not sure tx_prep is the only way to address it and
do other PMD have similar
restriction?. If yes, Can we have abstract it in a proper way the usage
part will be very clear from PMD and application perspective?
Jerin
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] doc: announce ivshmem support removal
2016-07-28 9:20 3% ` Christian Ehrhardt
@ 2016-07-28 15:23 0% ` Mauricio Vasquez
0 siblings, 0 replies; 200+ results
From: Mauricio Vasquez @ 2016-07-28 15:23 UTC (permalink / raw)
To: Christian Ehrhardt, Jan Viktorin; +Cc: Thomas Monjalon, anatoly.burakov, dev
Hello All,
Here in Politecnico di Torino we use the ivshmem technology from a
research point of view.
Our research efforts focus in optimizing the inter-Virtual Network
Function communication, currently we have implemented two versions of
our prototype, they are described in [1] and [2].
Unfortunately, we do not have the human resources to implement the
improvements that are necessary for ivshmem in DPDK, however we could
provide some feedback and testing for possible patches.
Best Regards,
Mauricio Vasquez.
[1]
https://www.researchgate.net/publication/305699120_Transparent_Optimization_of_Inter-Virtual_Network_Function_Communication_in_Open_vSwitch
[2]
https://www.researchgate.net/publication/305699122_A_Transparent_Highway_for_inter-Virtual_Network_Function_Communication_with_Open_vSwitch
On 07/28/2016 11:20 AM, Christian Ehrhardt wrote:
> Hi Thomas,
> just my two cents as Ubuntu DPDK maintainer (and part of the Debian Team
> that does the same).
> (It seems I can reuse that line for all posts about the deprecation notices
> :-) )
>
> While IVSHMEM was enabled (as it was the default) I never heard of any
> users of what we provided so far.
> But that is expected considering that not all qemu bits are landed either.
> Since it will follow the process of "deprecation notice -> grace period ->
> ABI bump" on removal, I think packaging and consuming applications should
> be fine.
>
> I'd agree to Hiroshi that, if really needed a clean re-implementation more
> properly separated from EAL likely is the best way to go.
>
> I think it is a good change to drop rather complex code in favor of
> stabilizing the main paths:
> Acked-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
>
>
> Christian Ehrhardt
> Software Engineer, Ubuntu Server
> Canonical Ltd
>
> On Wed, Jul 27, 2016 at 9:08 PM, Jan Viktorin <viktorin@rehivetech.com>
> wrote:
>
>> On Wed, 20 Jul 2016 18:35:46 +0200
>> Thomas Monjalon <thomas.monjalon@6wind.com> wrote:
>>
>>> There was a prior call with an explanation of what needs to be done:
>>> http://dpdk.org/ml/archives/dev/2016-June/040844.html
>>> - Qemu patch upstreamed
>>> - IVSHMEM PCI device managed by a PCI driver
>>> - No DPDK objects (ring/mempool) allocated by EAL
>>>
>>> As nobody seems interested, it is time to remove this code which
>>> makes EAL improvements harder.
>>>
>>> Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
>>> Acked-by: David Marchand <david.marchand@6wind.com>
>>> Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
>> Acked-by: Jan Viktorin <viktorin@rehivetech.com
>>
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption
@ 2016-07-28 14:54 3% ` Alex Williamson
0 siblings, 0 replies; 200+ results
From: Alex Williamson @ 2016-07-28 14:54 UTC (permalink / raw)
To: Burakov, Anatoly; +Cc: Thomas Monjalon, dev
On Thu, 28 Jul 2016 09:42:13 +0000
"Burakov, Anatoly" <anatoly.burakov@intel.com> wrote:
> > Hi,
> >
> > 2016-07-27 16:14, Alex Williamson:
> > > I took a quick look at the dpdk vfio code and spotted an invalid
> > > assumption that should probably be corrected ASAP.
> >
> > It can theoretically be a bug but the value may never change in the kernel,
> > right?
> > So when you say ASAP, I feel it can wait the next DPDK release (we plan to
> > release today).
> > Do you agree?
>
> Unless there are imminent plans to change this in the kernel, I think it can wait for next release.
I don't have any plans to change it, but this relationship is not a
guaranteed part of the ABI. I reserve the right to make such changes
in the future. Thanks,
Alex
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 13:59 4% ` Jerin Jacob
@ 2016-07-28 14:52 4% ` Thomas Monjalon
2016-07-28 16:25 7% ` Jerin Jacob
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-28 14:52 UTC (permalink / raw)
To: Jerin Jacob; +Cc: Ananyev, Konstantin, dev
2016-07-28 19:29, Jerin Jacob:
> Above things worries me, I wouldn't have cared if the changes are not comes
> in fastpath and I don't think this sort of issues will never get fixed any time
> soon in this community.
>
> So I given up.
I feel something goes wrong here but I cannot understand your sentence.
Please could you reword/explain Jerin?
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 13:58 4% ` Olivier MATZ
@ 2016-07-28 14:21 4% ` Ananyev, Konstantin
0 siblings, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-07-28 14:21 UTC (permalink / raw)
To: Olivier MATZ, Jerin Jacob; +Cc: Thomas Monjalon, dev
Hi Olivier,
>
> Hi,
>
> Jumping into this thread, it looks it's the last pending patch remaining for the release.
>
> For reference, the idea of tx_prep() was mentionned some time ago in http://dpdk.org/ml/archives/dev/2014-May/002504.html
>
> Few comments below.
>
> On 07/28/2016 03:01 PM, Ananyev, Konstantin wrote:
> > Right now to make HW TX offloads to work user is required to do particular actions:
> > 1. set mbuf.ol_flags properly.
> > 2. setup mbuf.tx_offload fields properly.
> > 3. update L3/L4 header fields in a particular way.
> >
> > We move #3 into tx_prep(), to hide that complexity from the user simplify things for him.
> > Though if he still prefers to do #3 by himself - that's ok too.
>
> I think moving #3 out of the application is a good idea. Today, for TSO, the offload dpdk API requires to set a specific pseudo header
> checksum (which does not include the ip len, as expected by Intel drivers), and set the IP checksum to 0.
>
> In our app, the network stack sets the TCP checksum to the standard pseudo header checksum, and before sending the mbuf:
> - packets are split in sw if the driver does not support tso
> - the tcp csum is patched to match dpdk api if the driver supports tso
>
> In the patchs I've recently sent adding tso support for virtio-pmd, it conforms to the dpdk API (phdr csum without ip len), so the tx function
> need to patch the mbuf data inside the driver, which is something what we want to avoid, for some good reasons explained by Konstantin.
Yep, that would be another good use-case for tx_prep() I suppose.
>
> So, I think having a tx_prep would also be the occasion to standardize a bit the dpdk offload api, and let the driver-specific stuff in tx_prep().
>
> Konstantin, any opinion about this?
Yes, that sounds like a good thing to me.
>
> >>> Another main purpose of tx_prep(): for multi-segment packets is to
> >>> check that number of segments doesn't exceed HW limit.
> >>> Again right now users have to do that on their own.
>
> If calling tx_prep() is optional, does it mean that this check may be done twice? (once in tx_prep() and once in tx_burst())
I meant 'optional' in a way, that if user doesn't want to use tx_prep() and
do step #3 from the above on his own (what happens now) that is still ok.
But I think step #3 (modify packet's data) still needs to be done before tx_burst() is called for the packets.
>
> What would be the advantage of doing this check in tx_prep() instead of keeping it in tx_burst(), as it does not touch the mbuf data?
>
> >>> 3. Having it a s separate function would allow user control when/where
> >>> to call it, let say only for some packets, or probably call tx_prep()
> >>> on one core, and do actual tx_burst() for these packets on the other.
>
> Yes, from what I remember, the pipeline model was the main reason why we do not just modify the packet in tx_burst(). Right?
Yes.
>
> >>>>> If you refer as lost cycles here something like:
> >>>>> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP); then
> >>>>> yes.
> >>>>> Though comparing to actual work need to be done for most HW TX
> >>>>> offloads, I think it is neglectable.
> >>>>
> >>>> Not sure.
>
> I think doing this test on a per-bulk basis should not impact performance.
>
> > To be honest, I don't understand what is your concern here.
> > That proposed change doesn't break any existing functionality, doesn't
> > introduce any new requirements to the existing API, and wouldn't
> > introduce any performance regression for existing apps.
> > It is a an extension, and user is free not to use it, if it doesn't fit his needs.
> > From other side there are users who are interested in that
> > functionality, and they do have use-cases for it.
>
> In my opinion, using tx_prep() will implicitly become mandatory as soon as the application want to do offload. An application that won't use
> it will have to prepare the mbuf, and this preparation will depend on the device, which is not acceptable inside an application.
Yes, I also hope that most apps that do use TX offloads will start to use it,
as I think it will be much more convenient way, then what we have right now.
I just liked to emphasis that user wouldn't be forced to.
Konstantin
>
>
> So, to conclude, the api change notification looks good to me, even if there is still some room to discuss the implementation details.
>
>
> Acked-by: Olivier Matz <olivier.matz@6wind.com>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 13:01 4% ` Ananyev, Konstantin
2016-07-28 13:58 4% ` Olivier MATZ
@ 2016-07-28 13:59 4% ` Jerin Jacob
2016-07-28 14:52 4% ` Thomas Monjalon
1 sibling, 1 reply; 200+ results
From: Jerin Jacob @ 2016-07-28 13:59 UTC (permalink / raw)
To: Ananyev, Konstantin; +Cc: Thomas Monjalon, dev
On Thu, Jul 28, 2016 at 01:01:16PM +0000, Ananyev, Konstantin wrote:
>
>
> > > >
> > > > Not according to proposal. It can't be too as application has no
> > > > idea what PMD driver does with "prep" what is the implication on a
> > > > HW if application does not
> > >
> > > Why application writer wouldn't have an idea?
> > > We would document what tx_prep() supposed to do, and in what cases user don't need it.
> >
> > But how he/she detect that on that run-time ?
>
> By the application logic for example.
> If let say is doing the l2fwd for that group of packets, it would know
> that it doesn't need to do tx_prep().
>
> To be honest, I don't understand what is your concern here.
> That proposed change doesn't break any existing functionality,
> doesn't introduce any new requirements to the existing API,
> and wouldn't introduce any performance regression for existing apps.
Yes for the existing application but no for ANY application that uses tx_prep() in future,
that run on the PMD where callback is NULL(one/two PMDs vs N PMDs)
> It is a an extension, and user is free not to use it, if it doesn't fit his needs.
If it is a workaround for a specific HW then why to change the normative
"fastpath" ethdev specification. You could give your fixup as internal
PMD driver routine and be done with it. It is as simple as that.
> From other side there are users who are interested in that functionality,
> and they do have use-cases for it.
> So what worries you?
Above things worries me, I wouldn't have cared if the changes are not comes
in fastpath and I don't think this sort of issues will never get fixed any time
soon in this community.
So I given up.
Jerin
> Konstantin
>
> >
> > > Then it would be up to the user:
> > > - not to use it at all (one segment per packet, no HW TX offloads)
> >
> > We already have TX flags for that
> >
> > > - not to use tx_prep(), and make necessary preparations himself,
> > > that what people have to do now.
> > > - use tx_prep()
> > >
> > > Konstantin
> > >
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 13:01 4% ` Ananyev, Konstantin
@ 2016-07-28 13:58 4% ` Olivier MATZ
2016-07-28 14:21 4% ` Ananyev, Konstantin
2016-07-28 13:59 4% ` Jerin Jacob
1 sibling, 1 reply; 200+ results
From: Olivier MATZ @ 2016-07-28 13:58 UTC (permalink / raw)
To: Ananyev, Konstantin, Jerin Jacob; +Cc: Thomas Monjalon, dev
Hi,
Jumping into this thread, it looks it's the last pending patch remaining
for the release.
For reference, the idea of tx_prep() was mentionned some time ago in
http://dpdk.org/ml/archives/dev/2014-May/002504.html
Few comments below.
On 07/28/2016 03:01 PM, Ananyev, Konstantin wrote:
> Right now to make HW TX offloads to work user is required to do particular actions:
> 1. set mbuf.ol_flags properly.
> 2. setup mbuf.tx_offload fields properly.
> 3. update L3/L4 header fields in a particular way.
>
> We move #3 into tx_prep(), to hide that complexity from the user simplify things for him.
> Though if he still prefers to do #3 by himself - that's ok too.
I think moving #3 out of the application is a good idea. Today, for TSO,
the offload dpdk API requires to set a specific pseudo header checksum
(which does not include the ip len, as expected by Intel drivers), and
set the IP checksum to 0.
In our app, the network stack sets the TCP checksum to the standard
pseudo header checksum, and before sending the mbuf:
- packets are split in sw if the driver does not support tso
- the tcp csum is patched to match dpdk api if the driver supports tso
In the patchs I've recently sent adding tso support for virtio-pmd, it
conforms to the dpdk API (phdr csum without ip len), so the tx function
need to patch the mbuf data inside the driver, which is something what
we want to avoid, for some good reasons explained by Konstantin.
So, I think having a tx_prep would also be the occasion to standardize a
bit the dpdk offload api, and let the driver-specific stuff in tx_prep().
Konstantin, any opinion about this?
>>> Another main purpose of tx_prep(): for multi-segment packets is to
>>> check that number of segments doesn't exceed HW limit.
>>> Again right now users have to do that on their own.
If calling tx_prep() is optional, does it mean that this check may be
done twice? (once in tx_prep() and once in tx_burst())
What would be the advantage of doing this check in tx_prep() instead of
keeping it in tx_burst(), as it does not touch the mbuf data?
>>> 3. Having it a s separate function would allow user control when/where
>>> to call it, let say only for some packets, or probably call tx_prep()
>>> on one core, and do actual tx_burst() for these packets on the other.
Yes, from what I remember, the pipeline model was the main reason why we
do not just modify the packet in tx_burst(). Right?
>>>>> If you refer as lost cycles here something like:
>>>>> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP); then
>>>>> yes.
>>>>> Though comparing to actual work need to be done for most HW TX
>>>>> offloads, I think it is neglectable.
>>>>
>>>> Not sure.
I think doing this test on a per-bulk basis should not impact performance.
> To be honest, I don't understand what is your concern here.
> That proposed change doesn't break any existing functionality,
> doesn't introduce any new requirements to the existing API,
> and wouldn't introduce any performance regression for existing apps.
> It is a an extension, and user is free not to use it, if it doesn't fit his needs.
> From other side there are users who are interested in that functionality,
> and they do have use-cases for it.
In my opinion, using tx_prep() will implicitly become mandatory as soon
as the application want to do offload. An application that won't use it
will have to prepare the mbuf, and this preparation will depend on the
device, which is not acceptable inside an application.
So, to conclude, the api change notification looks good to me, even if
there is still some room to discuss the implementation details.
Acked-by: Olivier Matz <olivier.matz@6wind.com>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 11:38 4% ` Jerin Jacob
2016-07-28 12:07 4% ` Avi Kivity
@ 2016-07-28 13:01 4% ` Ananyev, Konstantin
2016-07-28 13:58 4% ` Olivier MATZ
2016-07-28 13:59 4% ` Jerin Jacob
1 sibling, 2 replies; 200+ results
From: Ananyev, Konstantin @ 2016-07-28 13:01 UTC (permalink / raw)
To: Jerin Jacob; +Cc: Thomas Monjalon, dev
> -----Original Message-----
> From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> Sent: Thursday, July 28, 2016 12:39 PM
> To: Ananyev, Konstantin <konstantin.ananyev@intel.com>
> Cc: Thomas Monjalon <thomas.monjalon@6wind.com>; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
>
> On Thu, Jul 28, 2016 at 10:36:07AM +0000, Ananyev, Konstantin wrote:
> > > If it does not cope up then it can skip tx'ing in the actual tx
> > > burst itself and move the "skipped" tx packets to end of the list in
> > > the tx burst so that application can take the action on "skipped"
> > > packet after the tx burst
> >
> > Sorry, that's too cryptic for me.
> > Can you reword it somehow?
>
> OK.
> 1) lets say application requests 32 packets to send it using tx_burst.
> 2) packets are from p0 to p31
> 3) in driver due to some reason, it is not able to send the packets due to some constraints in the driver(say expect p2 and p16 everything
> else sent successfully by the driver)
> 4) driver can move p2 and p16 at pkt[0] and pkt[1] on tx_burst and return 30
> 5) application can take action on p2 and p16 based the return value of 30(ie 32-30 = 2 packets needs to handle at pkt[0] and pkt[1]
That would introduce packets reordering and unnecessary complicate the PMD TX functions.
Again it would require changes in *all* existing PMD tx functions.
So we don't plan to do things that way.
>
>
> >
> > >
> > >
> > > > Instead it just setups the ol_flags, fills tx_offload fields and calls tx_prep().
> > > > Please read the original Tomasz's patch, I think he explained
> > > > possible use-cases with lot of details.
> > >
> > > Sorry, it is not very clear in terms of use cases.
> >
> > Ok, what I meant to say:
> > Right now, if user wants to use HW TX cksum/TSO offloads he might have to:
> > - setup ipv4 header cksum field.
> > - calculate the pseudo header checksum
> > - setup tcp/udp cksum field.
> >
> > Rules how these calculations need to be done and which fields need to
> > be updated, may vary depending on HW underneath and requested offloads.
> > tx_prep() - supposed to hide all these nuances from user and allow him
> > to use TX HW offloads in a transparent way.
>
> Not sure I understand it completely. Bit contradicting with below statement
> |We would document what tx_prep() supposed to do, and in what cases user
> |don't need it.
How that contradicts?
Right now to make HW TX offloads to work user is required to do particular actions:
1. set mbuf.ol_flags properly.
2. setup mbuf.tx_offload fields properly.
3. update L3/L4 header fields in a particular way.
We move #3 into tx_prep(), to hide that complexity from the user simplify things for him.
Though if he still prefers to do #3 by himself - that's ok too.
>
> How about introducing a new ethdev generic eal command-line mode OR new ethdev_configure hint that PMD driver is in "tx_prep-
> >tx_burst" mode instead of just tx_burst? That way no fast-path performance degradation for the PMD that does not need it
User might want to send different packets over different devices,
or even over different queues over the same device.
Or even he might want to call tx_prep() for one group of packets,
but skip for different group of packets for the same TX queue.
So I think we should allow user to decide when/where to call it.
>
>
> >
> > Another main purpose of tx_prep(): for multi-segment packets is to
> > check that number of segments doesn't exceed HW limit.
> > Again right now users have to do that on their own.
> >
> > >
> > > In HW perspective, It it tries to avoid the illegal state. But not
> > > sure calling "back to back" tx prepare and then tx burst how does it
> > > improve the situation as the check illegal state check introduce in
> > > actual tx burst it self.
> > >
> > > In SW perspective, its try to avoid sending malformed packets. In my
> > > view the same can achieved with existing tx burst it self as PMD is
> > > the one finally send the packets on the wire.
> >
> > Ok, so your question is: why not to put that functionality into
> > tx_burst() itself, right?
> > For few reasons:
> > 1. putting that functionality into tx_burst() would introduce unnecessary
> > slowdown for cases when that functionality is not needed
> > (one segment per packet, no HW offloads).
>
> These parameters can be configured on init time
No always, see above.
>
> > 2. User might don't want to use tx_prep() - he/she might have its
> > own analog, which he/she belives is faster/smarter,etc.
>
> That's the current mode. Right?
Yes.
> > 3. Having it a s separate function would allow user control when/where
> > to call it, let say only for some packets, or probably call tx_prep()
> > on one core, and do actual tx_burst() for these packets on the other.
> Why to process it under tx_prep() as application can always process the packet in one core
Because not every application wants to do it over the same core.
Some apps would like to do it on the same core, some apps would like to do it on different core.
With proposed API both models are possible.
>
> > >
> > > proposal quote:
> > >
> > > 1. Introduce rte_eth_tx_prep() function to do necessary preparations of
> > > packet burst to be safely transmitted on device for desired HW
> > > offloads (set/reset checksum field according to the hardware
> > > requirements) and check HW constraints (number of segments per
> > > packet, etc).
> > >
> > > While the limitations and requirements may differ for devices, it
> > > requires to extend rte_eth_dev structure with new function pointer
> > > "tx_pkt_prep" which can be implemented in the driver to prepare and
> > > verify packets, in devices specific way, before burst, what should to
> > > prevent application to send malformed packets.
> > >
> > >
> > > >
> > > > > and what if the PMD does not implement that callback then it is of waste cycles. Right?
> > > >
> > > > If you refer as lost cycles here something like:
> > > > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP); then
> > > > yes.
> > > > Though comparing to actual work need to be done for most HW TX
> > > > offloads, I think it is neglectable.
> > >
> > > Not sure.
> > >
> > > > Again, as I said before, it is totally voluntary for the application.
> > >
> > > Not according to proposal. It can't be too as application has no
> > > idea what PMD driver does with "prep" what is the implication on a
> > > HW if application does not
> >
> > Why application writer wouldn't have an idea?
> > We would document what tx_prep() supposed to do, and in what cases user don't need it.
>
> But how he/she detect that on that run-time ?
By the application logic for example.
If let say is doing the l2fwd for that group of packets, it would know
that it doesn't need to do tx_prep().
To be honest, I don't understand what is your concern here.
That proposed change doesn't break any existing functionality,
doesn't introduce any new requirements to the existing API,
and wouldn't introduce any performance regression for existing apps.
It is a an extension, and user is free not to use it, if it doesn't fit his needs.
>From other side there are users who are interested in that functionality,
and they do have use-cases for it.
So what worries you?
Konstantin
>
> > Then it would be up to the user:
> > - not to use it at all (one segment per packet, no HW TX offloads)
>
> We already have TX flags for that
>
> > - not to use tx_prep(), and make necessary preparations himself,
> > that what people have to do now.
> > - use tx_prep()
> >
> > Konstantin
> >
^ permalink raw reply [relevance 4%]
* [dpdk-dev] DPDK Stable Releases and Long Term Support
@ 2016-07-28 12:33 3% Mcnamara, John
0 siblings, 0 replies; 200+ results
From: Mcnamara, John @ 2016-07-28 12:33 UTC (permalink / raw)
To: dev
This document sets out the guidelines for DPDK Stable Releases and Long Term
Support releases (LTS) based on the initial RFC and comments:
http://dpdk.org/ml/archives/dev/2016-June/040256.html.
In particular it incorporates suggestions for a Stable Release structure as
well as a Long Term Support release.
Introduction
------------
The purpose of the DPDK Stable Releases will be to maintain releases of DPDK
with backported fixes over an extended period of time. This will provide
downstream consumers of DPDK with a stable target on which to base
applications or packages.
The Long Term Support release (LTS) will be a designation applied to a Stable
Release to indicate longer support.
Stable Releases
---------------
Any major release of DPDK can be designated as a Stable Release if a
maintainer volunteers to maintain it.
A Stable Release will be used to backport fixes from a N release back to a N-1
release, for example, from 16.11 to 16.07.
The duration of a stable release should be one complete release cycle. It can
be longer, up to 1 year, if a maintainer continues to support the stable
branch, or if users supply backported fixes, however the explicit commitment
should be for one release cycle.
The release cadence can be determined by the maintainer based on the number of
bugfixes and the criticality of the bugs. However, releases should be
coordinated with the validation engineers to ensure that a tagged release has
been tested.
LTS Release
-----------
A stable release can be designated as an LTS release based on community
agreement and a commitment from a maintainer. An LTS release will have a
maintenance duration of 2 years.
It is anticipated that there should be at least 4 releases per year of the LTS
or approximately 1 every 3 months. However, the cadence can be shorter or
longer depending on the number and criticality of the backported
fixes. Releases should be coordinated with the validation engineers to ensure
that a tagged release has been tested.
Initial Stable Release
----------------------
The initial DPDK Stable Release will be 16.07. It will be viewed as a trial of
the Stable Release/LTS policy to determine what are the best working practices
for DPDK.
The maintainer for the initial release will be Yuanhan Liu
<yuanhan.liu@linux.intel.com>. It is hoped that other community members will
volunteer as maintainers for other Stable Releases.
The initial targeted release for LTS is proposed to be 16.11 based on the
results of the work carried out on the 16.07 Stable Release.
A list has been set up for Stable Release/LTS specific discussions:
<stable@dpdk.org>. This address can also be used for CCing maintainers on bug
fix submissions.
What changes should be backported
---------------------------------
The backporting should be limited to bug fixes.
Features should not be backported to stable releases. It may be acceptable, in
limited cases, to back port features for the LTS release where:
* There is a justifiable use case (for example a new PMD).
* The change is non-invasive.
* The work of preparing the backport is done by the proposer.
* There is support within the community.
Testing
-------
Stable and LTS releases should be tested before release/tagging.
Intel will provide validation engineers to test the 16.07 Stable Release and
the initial LTS tree. Other community members should provide testing for other
stable releases.
The validation will consist of compilation testing on the range of OSes
supported by the master release and functional/performance testing on the
current major/LTS release of the following OSes:
* Ubuntu
* RHEL
* SuSE
* FreeBSD
Releasing
---------
A Stable Release will be released by:
* Tagging the release with YY.MM.nn (year, month, number) or similar.
* Uploading a tarball of the release to dpdk.org.
* Sending an announcement to the <announce@dpdk.org> list.
ABI
---
The Stable Release should not be seen as a way of breaking or circumventing
the DPDK ABI policy.
Review of the Stable Release/LTS guidelines
-------------------------------------------
This document serves as a set of guidelines for the planned Stable
Releases/LTS activities. However, the actual process can be reviewed and
amended over time, based on experiences and feedback.
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 11:38 4% ` Jerin Jacob
@ 2016-07-28 12:07 4% ` Avi Kivity
2016-07-28 13:01 4% ` Ananyev, Konstantin
1 sibling, 0 replies; 200+ results
From: Avi Kivity @ 2016-07-28 12:07 UTC (permalink / raw)
To: Jerin Jacob, Ananyev, Konstantin; +Cc: Thomas Monjalon, dev
On 07/28/2016 02:38 PM, Jerin Jacob wrote:
> On Thu, Jul 28, 2016 at 10:36:07AM +0000, Ananyev, Konstantin wrote:
>>> If it does not cope up then it can skip tx'ing in the actual tx burst
>>> itself and move the "skipped" tx packets to end of the list in the tx
>>> burst so that application can take the action on "skipped" packet after
>>> the tx burst
>> Sorry, that's too cryptic for me.
>> Can you reword it somehow?
> OK.
> 1) lets say application requests 32 packets to send it using tx_burst.
> 2) packets are from p0 to p31
> 3) in driver due to some reason, it is not able to send the packets due to some
> constraints in the driver(say expect p2 and p16 everything else sent
> successfully by the driver)
> 4) driver can move p2 and p16 at pkt[0] and pkt[1] on tx_burst and
> return 30
> 5) application can take action on p2 and p16 based the return value of
> 30(ie 32-30 = 2 packets needs to handle at pkt[0] and pkt[1]
That can cause reordering; while it is legal, it reduces tcp performance.
Better to preserve the application-provided order.
>
>>>
>>>> Instead it just setups the ol_flags, fills tx_offload fields and calls tx_prep().
>>>> Please read the original Tomasz's patch, I think he explained possible use-cases
>>>> with lot of details.
>>> Sorry, it is not very clear in terms of use cases.
>> Ok, what I meant to say:
>> Right now, if user wants to use HW TX cksum/TSO offloads he might have to:
>> - setup ipv4 header cksum field.
>> - calculate the pseudo header checksum
>> - setup tcp/udp cksum field.
>>
>> Rules how these calculations need to be done and which fields need to be updated,
>> may vary depending on HW underneath and requested offloads.
>> tx_prep() - supposed to hide all these nuances from user and allow him to use TX HW offloads
>> in a transparent way.
> Not sure I understand it completely. Bit contradicting with below
> statement
> |We would document what tx_prep() supposed to do, and in what cases user
> |don't need it.
>
> How about introducing a new ethdev generic eal command-line mode OR
> new ethdev_configure hint that PMD driver is in "tx_prep->tx_burst" mode
> instead of just tx_burst? That way no fast-path performance degradation
> for the PMD that does not need it
>
>
>> Another main purpose of tx_prep(): for multi-segment packets is to check
>> that number of segments doesn't exceed HW limit.
>> Again right now users have to do that on their own.
>>
>>> In HW perspective, It it tries to avoid the illegal state. But not sure
>>> calling "back to back" tx prepare and then tx burst how does it improve the
>>> situation as the check illegal state check introduce in actual tx burst
>>> it self.
>>>
>>> In SW perspective, its try to avoid sending malformed packets. In my
>>> view the same can achieved with existing tx burst it self as PMD is the
>>> one finally send the packets on the wire.
>> Ok, so your question is: why not to put that functionality into
>> tx_burst() itself, right?
>> For few reasons:
>> 1. putting that functionality into tx_burst() would introduce unnecessary
>> slowdown for cases when that functionality is not needed
>> (one segment per packet, no HW offloads).
> These parameters can be configured on init time
>
>> 2. User might don't want to use tx_prep() - he/she might have its
>> own analog, which he/she belives is faster/smarter,etc.
> That's the current mode. Right?
>> 3. Having it a s separate function would allow user control when/where
>> to call it, let say only for some packets, or probably call tx_prep()
>> on one core, and do actual tx_burst() for these packets on the other.
> Why to process it under tx_prep() as application can always process the
> packet in one core
>
>>> proposal quote:
>>>
>>> 1. Introduce rte_eth_tx_prep() function to do necessary preparations of
>>> packet burst to be safely transmitted on device for desired HW
>>> offloads (set/reset checksum field according to the hardware
>>> requirements) and check HW constraints (number of segments per
>>> packet, etc).
>>>
>>> While the limitations and requirements may differ for devices, it
>>> requires to extend rte_eth_dev structure with new function pointer
>>> "tx_pkt_prep" which can be implemented in the driver to prepare and
>>> verify packets, in devices specific way, before burst, what should to
>>> prevent application to send malformed packets.
>>>
>>>
>>>>> and what if the PMD does not implement that callback then it is of waste cycles. Right?
>>>> If you refer as lost cycles here something like:
>>>> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP);
>>>> then yes.
>>>> Though comparing to actual work need to be done for most HW TX offloads,
>>>> I think it is neglectable.
>>> Not sure.
>>>
>>>> Again, as I said before, it is totally voluntary for the application.
>>> Not according to proposal. It can't be too as application has no idea
>>> what PMD driver does with "prep" what is the implication on a HW if
>>> application does not
>> Why application writer wouldn't have an idea?
>> We would document what tx_prep() supposed to do, and in what cases user don't need it.
> But how he/she detect that on that run-time ?
>
>> Then it would be up to the user:
>> - not to use it at all (one segment per packet, no HW TX offloads)
> We already have TX flags for that
>
>> - not to use tx_prep(), and make necessary preparations himself,
>> that what people have to do now.
>> - use tx_prep()
>>
>> Konstantin
>>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-21 15:24 11% ` [dpdk-dev] [PATCH v2] " Tomasz Kulasek
2016-07-21 22:48 4% ` Ananyev, Konstantin
@ 2016-07-28 12:04 4% ` Avi Kivity
1 sibling, 0 replies; 200+ results
From: Avi Kivity @ 2016-07-28 12:04 UTC (permalink / raw)
To: Tomasz Kulasek, dev; +Cc: Vladislav Zolotarov, Takuya ASADA
On 07/21/2016 06:24 PM, Tomasz Kulasek wrote:
> This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
> changes in rte_eth_dev and rte_eth_desc_lim structures.
>
> As discussed in that thread:
>
> http://dpdk.org/ml/archives/dev/2015-September/023603.html
>
> Different NIC models depending on HW offload requested might impose
> different requirements on packets to be TX-ed in terms of:
>
> - Max number of fragments per packet allowed
> - Max number of fragments per TSO segments
> - The way pseudo-header checksum should be pre-calculated
> - L3/L4 header fields filling
> - etc.
>
>
> MOTIVATION:
> -----------
>
> 1) Some work cannot (and didn't should) be done in rte_eth_tx_burst.
> However, this work is sometimes required, and now, it's an
> application issue.
>
> 2) Different hardware may have different requirements for TX offloads,
> other subset can be supported and so on.
>
> 3) Some parameters (eg. number of segments in ixgbe driver) may hung
> device. These parameters may be vary for different devices.
>
> For example i40e HW allows 8 fragments per packet, but that is after
> TSO segmentation. While ixgbe has a 38-fragment pre-TSO limit.
>
> 4) Fields in packet may require different initialization (like eg. will
> require pseudo-header checksum precalculation, sometimes in a
> different way depending on packet type, and so on). Now application
> needs to care about it.
>
> 5) Using additional API (rte_eth_tx_prep) before rte_eth_tx_burst let to
> prepare packet burst in acceptable form for specific device.
>
> 6) Some additional checks may be done in debug mode keeping tx_burst
> implementation clean.
Thanks a lot for this. Seastar suffered from this issue and had to
apply NIC-specific workarounds.
The proposal will work well for seastar.
>
> PROPOSAL:
> ---------
>
> To help user to deal with all these varieties we propose to:
>
> 1. Introduce rte_eth_tx_prep() function to do necessary preparations of
> packet burst to be safely transmitted on device for desired HW
> offloads (set/reset checksum field according to the hardware
> requirements) and check HW constraints (number of segments per
> packet, etc).
>
> While the limitations and requirements may differ for devices, it
> requires to extend rte_eth_dev structure with new function pointer
> "tx_pkt_prep" which can be implemented in the driver to prepare and
> verify packets, in devices specific way, before burst, what should to
> prevent application to send malformed packets.
>
> 2. Also new fields will be introduced in rte_eth_desc_lim:
> nb_seg_max and nb_mtu_seg_max, providing an information about max
> segments in TSO and non-TSO packets acceptable by device.
>
> This information is useful for application to not create/limit
> malicious packet.
>
>
> APPLICATION (CASE OF USE):
> --------------------------
>
> 1) Application should to initialize burst of packets to send, set
> required tx offload flags and required fields, like l2_len, l3_len,
> l4_len, and tso_segsz
>
> 2) Application passes burst to the rte_eth_tx_prep to check conditions
> required to send packets through the NIC.
>
> 3) The result of rte_eth_tx_prep can be used to send valid packets
> and/or restore invalid if function fails.
>
> eg.
>
> for (i = 0; i < nb_pkts; i++) {
>
> /* initialize or process packet */
>
> bufs[i]->tso_segsz = 800;
> bufs[i]->ol_flags = PKT_TX_TCP_SEG | PKT_TX_IPV4
> | PKT_TX_IP_CKSUM;
> bufs[i]->l2_len = sizeof(struct ether_hdr);
> bufs[i]->l3_len = sizeof(struct ipv4_hdr);
> bufs[i]->l4_len = sizeof(struct tcp_hdr);
> }
>
> /* Prepare burst of TX packets */
> nb_prep = rte_eth_tx_prep(port, 0, bufs, nb_pkts);
>
> if (nb_prep < nb_pkts) {
> printf("tx_prep failed\n");
>
> /* drop or restore invalid packets */
>
> }
>
> /* Send burst of TX packets */
> nb_tx = rte_eth_tx_burst(port, 0, bufs, nb_prep);
>
> /* Free any unsent packets. */
>
>
>
> Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> ---
> doc/guides/rel_notes/deprecation.rst | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index f502f86..485aacb 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -41,3 +41,10 @@ Deprecation Notices
> * The mempool functions for single/multi producer/consumer are deprecated and
> will be removed in 16.11.
> It is replaced by rte_mempool_generic_get/put functions.
> +
> +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
> + extended with new function pointer ``tx_pkt_prep`` allowing verification
> + and processing of packet burst to meet HW specific requirements before
> + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
> + segments limit to be transmitted by device for TSO/non-TSO packets.
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 10:36 4% ` Ananyev, Konstantin
@ 2016-07-28 11:38 4% ` Jerin Jacob
2016-07-28 12:07 4% ` Avi Kivity
2016-07-28 13:01 4% ` Ananyev, Konstantin
0 siblings, 2 replies; 200+ results
From: Jerin Jacob @ 2016-07-28 11:38 UTC (permalink / raw)
To: Ananyev, Konstantin; +Cc: Thomas Monjalon, dev
On Thu, Jul 28, 2016 at 10:36:07AM +0000, Ananyev, Konstantin wrote:
> > If it does not cope up then it can skip tx'ing in the actual tx burst
> > itself and move the "skipped" tx packets to end of the list in the tx
> > burst so that application can take the action on "skipped" packet after
> > the tx burst
>
> Sorry, that's too cryptic for me.
> Can you reword it somehow?
OK.
1) lets say application requests 32 packets to send it using tx_burst.
2) packets are from p0 to p31
3) in driver due to some reason, it is not able to send the packets due to some
constraints in the driver(say expect p2 and p16 everything else sent
successfully by the driver)
4) driver can move p2 and p16 at pkt[0] and pkt[1] on tx_burst and
return 30
5) application can take action on p2 and p16 based the return value of
30(ie 32-30 = 2 packets needs to handle at pkt[0] and pkt[1]
>
> >
> >
> > > Instead it just setups the ol_flags, fills tx_offload fields and calls tx_prep().
> > > Please read the original Tomasz's patch, I think he explained possible use-cases
> > > with lot of details.
> >
> > Sorry, it is not very clear in terms of use cases.
>
> Ok, what I meant to say:
> Right now, if user wants to use HW TX cksum/TSO offloads he might have to:
> - setup ipv4 header cksum field.
> - calculate the pseudo header checksum
> - setup tcp/udp cksum field.
>
> Rules how these calculations need to be done and which fields need to be updated,
> may vary depending on HW underneath and requested offloads.
> tx_prep() - supposed to hide all these nuances from user and allow him to use TX HW offloads
> in a transparent way.
Not sure I understand it completely. Bit contradicting with below
statement
|We would document what tx_prep() supposed to do, and in what cases user
|don't need it.
How about introducing a new ethdev generic eal command-line mode OR
new ethdev_configure hint that PMD driver is in "tx_prep->tx_burst" mode
instead of just tx_burst? That way no fast-path performance degradation
for the PMD that does not need it
>
> Another main purpose of tx_prep(): for multi-segment packets is to check
> that number of segments doesn't exceed HW limit.
> Again right now users have to do that on their own.
>
> >
> > In HW perspective, It it tries to avoid the illegal state. But not sure
> > calling "back to back" tx prepare and then tx burst how does it improve the
> > situation as the check illegal state check introduce in actual tx burst
> > it self.
> >
> > In SW perspective, its try to avoid sending malformed packets. In my
> > view the same can achieved with existing tx burst it self as PMD is the
> > one finally send the packets on the wire.
>
> Ok, so your question is: why not to put that functionality into
> tx_burst() itself, right?
> For few reasons:
> 1. putting that functionality into tx_burst() would introduce unnecessary
> slowdown for cases when that functionality is not needed
> (one segment per packet, no HW offloads).
These parameters can be configured on init time
> 2. User might don't want to use tx_prep() - he/she might have its
> own analog, which he/she belives is faster/smarter,etc.
That's the current mode. Right?
> 3. Having it a s separate function would allow user control when/where
> to call it, let say only for some packets, or probably call tx_prep()
> on one core, and do actual tx_burst() for these packets on the other.
Why to process it under tx_prep() as application can always process the
packet in one core
> >
> > proposal quote:
> >
> > 1. Introduce rte_eth_tx_prep() function to do necessary preparations of
> > packet burst to be safely transmitted on device for desired HW
> > offloads (set/reset checksum field according to the hardware
> > requirements) and check HW constraints (number of segments per
> > packet, etc).
> >
> > While the limitations and requirements may differ for devices, it
> > requires to extend rte_eth_dev structure with new function pointer
> > "tx_pkt_prep" which can be implemented in the driver to prepare and
> > verify packets, in devices specific way, before burst, what should to
> > prevent application to send malformed packets.
> >
> >
> > >
> > > > and what if the PMD does not implement that callback then it is of waste cycles. Right?
> > >
> > > If you refer as lost cycles here something like:
> > > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP);
> > > then yes.
> > > Though comparing to actual work need to be done for most HW TX offloads,
> > > I think it is neglectable.
> >
> > Not sure.
> >
> > > Again, as I said before, it is totally voluntary for the application.
> >
> > Not according to proposal. It can't be too as application has no idea
> > what PMD driver does with "prep" what is the implication on a HW if
> > application does not
>
> Why application writer wouldn't have an idea?
> We would document what tx_prep() supposed to do, and in what cases user don't need it.
But how he/she detect that on that run-time ?
> Then it would be up to the user:
> - not to use it at all (one segment per packet, no HW TX offloads)
We already have TX flags for that
> - not to use tx_prep(), and make necessary preparations himself,
> that what people have to do now.
> - use tx_prep()
>
> Konstantin
>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-28 2:13 4% ` Jerin Jacob
@ 2016-07-28 10:36 4% ` Ananyev, Konstantin
2016-07-28 11:38 4% ` Jerin Jacob
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-07-28 10:36 UTC (permalink / raw)
To: Jerin Jacob; +Cc: Thomas Monjalon, dev
> > > >
> > > > > -----Original Message-----
> > > > > From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> > > > > Sent: Wednesday, July 27, 2016 6:11 PM
> > > > > To: Thomas Monjalon <thomas.monjalon@6wind.com>
> > > > > Cc: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org;
> > > > > Ananyev, Konstantin <konstantin.ananyev@intel.com>
> > > > > Subject: Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for
> > > > > rte_eth_dev structure
> > > > >
> > > > > On Wed, Jul 27, 2016 at 01:59:01AM -0700, Thomas Monjalon wrote:
> > > > > > > > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > > > > > > > ---
> > > > > > > > +* In 16.11 ABI changes are plained: the ``rte_eth_dev``
> > > > > > > > +structure will be
> > > > > > > > + extended with new function pointer ``tx_pkt_prep`` allowing
> > > > > > > > +verification
> > > > > > > > + and processing of packet burst to meet HW specific
> > > > > > > > +requirements before
> > > > > > > > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > > > > > > > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing
> > > > > > > > +information about number of
> > > > > > > > + segments limit to be transmitted by device for TSO/non-TSO packets.
> > > > > > >
> > > > > > > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> > > > > >
> > > > > > I think I understand you want to split the TX processing:
> > > > > > 1/ modify/write in mbufs
> > > > > > 2/ write in HW
> > > > > > and let application decide:
> > > > > > - where the TX prep is done (which core)
> > > > >
> > > > > In what basics applications knows when and where to call tx_pkt_prep in fast path.
> > > > > if all the time it needs to call before tx_burst then the PMD won't
> > > > > have/don't need this callback waste cycles in fast path.Is this the expected behavior ?
> > > > > Anything think it as compile time to make other PMDs wont suffer because of this change.
> > > >
> > > > Not sure what suffering you are talking about...
> > > > Current model - i.e. when application does preparations (or doesn't if
> > > > none is required) on its own and just call tx_burst() would still be there.
> > > > If the app doesn't want to use tx_prep() by some reason - that still
> > > > ok, and decision is up to the particular app.
> > >
> > > So my question is in what basics application decides to call the preparation.
> > > Can you tell me the use case in application perspective?
> >
> > I suppose one most common use-case when application uses HW TX offloads,
> > and don't' to cope on its own which L3/L4 header fields need to be filled
> > for that particular dev_type/hw_offload combination.
>
> If it does not cope up then it can skip tx'ing in the actual tx burst
> itself and move the "skipped" tx packets to end of the list in the tx
> burst so that application can take the action on "skipped" packet after
> the tx burst
Sorry, that's too cryptic for me.
Can you reword it somehow?
>
>
> > Instead it just setups the ol_flags, fills tx_offload fields and calls tx_prep().
> > Please read the original Tomasz's patch, I think he explained possible use-cases
> > with lot of details.
>
> Sorry, it is not very clear in terms of use cases.
Ok, what I meant to say:
Right now, if user wants to use HW TX cksum/TSO offloads he might have to:
- setup ipv4 header cksum field.
- calculate the pseudo header checksum
- setup tcp/udp cksum field.
Rules how these calculations need to be done and which fields need to be updated,
may vary depending on HW underneath and requested offloads.
tx_prep() - supposed to hide all these nuances from user and allow him to use TX HW offloads
in a transparent way.
Another main purpose of tx_prep(): for multi-segment packets is to check
that number of segments doesn't exceed HW limit.
Again right now users have to do that on their own.
>
> In HW perspective, It it tries to avoid the illegal state. But not sure
> calling "back to back" tx prepare and then tx burst how does it improve the
> situation as the check illegal state check introduce in actual tx burst
> it self.
>
> In SW perspective, its try to avoid sending malformed packets. In my
> view the same can achieved with existing tx burst it self as PMD is the
> one finally send the packets on the wire.
Ok, so your question is: why not to put that functionality into
tx_burst() itself, right?
For few reasons:
1. putting that functionality into tx_burst() would introduce unnecessary
slowdown for cases when that functionality is not needed
(one segment per packet, no HW offloads).
2. User might don't want to use tx_prep() - he/she might have its
own analog, which he/she belives is faster/smarter,etc.
3. Having it a s separate function would allow user control when/where
to call it, let say only for some packets, or probably call tx_prep()
on one core, and do actual tx_burst() for these packets on the other.
>
> proposal quote:
>
> 1. Introduce rte_eth_tx_prep() function to do necessary preparations of
> packet burst to be safely transmitted on device for desired HW
> offloads (set/reset checksum field according to the hardware
> requirements) and check HW constraints (number of segments per
> packet, etc).
>
> While the limitations and requirements may differ for devices, it
> requires to extend rte_eth_dev structure with new function pointer
> "tx_pkt_prep" which can be implemented in the driver to prepare and
> verify packets, in devices specific way, before burst, what should to
> prevent application to send malformed packets.
>
>
> >
> > > and what if the PMD does not implement that callback then it is of waste cycles. Right?
> >
> > If you refer as lost cycles here something like:
> > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP);
> > then yes.
> > Though comparing to actual work need to be done for most HW TX offloads,
> > I think it is neglectable.
>
> Not sure.
>
> > Again, as I said before, it is totally voluntary for the application.
>
> Not according to proposal. It can't be too as application has no idea
> what PMD driver does with "prep" what is the implication on a HW if
> application does not
Why application writer wouldn't have an idea?
We would document what tx_prep() supposed to do, and in what cases user don't need it.
Then it would be up to the user:
- not to use it at all (one segment per packet, no HW TX offloads)
- not to use tx_prep(), and make necessary preparations himself,
that what people have to do now.
- use tx_prep()
Konstantin
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH] doc: remove deprecation notice related to new flow types
@ 2016-07-28 10:15 7% Rahul Lakkireddy
0 siblings, 0 replies; 200+ results
From: Rahul Lakkireddy @ 2016-07-28 10:15 UTC (permalink / raw)
To: dev; +Cc: Kumar Sanghvi, Nirranjan Kirubaharan
Remove deprecation notice pertaining to introduction of new flow
types in favor of a more generic filtering infrastructure proposal.
Signed-off-by: Rahul Lakkireddy <rahul.lakkireddy@chelsio.com>
Signed-off-by: Kumar Sanghvi <kumaras@chelsio.com>
---
doc/guides/rel_notes/deprecation.rst | 4 ----
1 file changed, 4 deletions(-)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f502f86..7fc1185 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -23,10 +23,6 @@ Deprecation Notices
do not need to care about the kind of devices that are being used, making it
easier to add new buses later.
-* ABI changes are planned for adding four new flow types. This impacts
- RTE_ETH_FLOW_MAX. The release 2.2 does not contain these ABI changes,
- but release 2.3 will. [postponed]
-
* The mbuf flags PKT_RX_VLAN_PKT and PKT_RX_QINQ_PKT are deprecated and
are respectively replaced by PKT_RX_VLAN_STRIPPED and
PKT_RX_QINQ_STRIPPED, that are better described. The old flags and
--
2.5.3
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] removal of old deprecation notice for Chelsio filtering
2016-07-28 8:29 4% [dpdk-dev] removal of old deprecation notice for Chelsio filtering Thomas Monjalon
@ 2016-07-28 10:12 0% ` Rahul Lakkireddy
0 siblings, 0 replies; 200+ results
From: Rahul Lakkireddy @ 2016-07-28 10:12 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev, Kumar Sanghvi, Nirranjan Kirubaharan
Hi Thomas,
On Thursday, July 07/28/16, 2016 at 01:29:20 -0700, Thomas Monjalon wrote:
> Hi Rahul,
>
> We still have this deprecation notice:
>
> * ABI changes are planned for adding four new flow types. This impacts
> RTE_ETH_FLOW_MAX. The release 2.2 does not contain these ABI changes,
> but release 2.3 will. [postponed]
>
> Do you agree that we can remove it now we have a better generic filtering
> API approach?
>
Yes. We can remove this deprecation notice in favor of the generic
filtering proposal in discussion. I'll send a patch now.
Thanks,
Rahul
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] doc: announce renaming of ethdev library
2016-07-28 9:29 3% ` Christian Ehrhardt
@ 2016-07-28 9:52 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-28 9:52 UTC (permalink / raw)
To: Christian Ehrhardt; +Cc: Jan Viktorin, dev
2016-07-28 11:29, Christian Ehrhardt:
> Just curious, do we already know by looking ahead if ethdev will get an ABI
> bump anyway?
> So will the transition be:
> a) libethdev4 -> librte_ethdev5
> b)libethdev4 -> librte_ethdev4
> If it is b) would/should one provide a compat symlink then in your Opinion?
Good point.
We'll make a symlink if the version stay the same.
Maybe it will be bumped because of a rework of the hotplug API.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] doc: announce renaming of ethdev library
@ 2016-07-28 9:29 3% ` Christian Ehrhardt
2016-07-28 9:52 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Christian Ehrhardt @ 2016-07-28 9:29 UTC (permalink / raw)
To: Jan Viktorin; +Cc: dev, Thomas Monjalon
Hi Thomas,
just my two cents as Ubuntu DPDK maintainer (and part of the Debian Team
that does the same).
(Yeah I really could reuse it three times :-) )
It will be a bit of effort to adapt, but should be no rocket-science.
I like that eventually the namespace will be cleaner.
Just curious, do we already know by looking ahead if ethdev will get an ABI
bump anyway?
So will the transition be:
a) libethdev4 -> librte_ethdev5
b)libethdev4 -> librte_ethdev4
If it is b) would/should one provide a compat symlink then in your Opinion?
Anyway, for now I think it is fair to say:
Acked-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd
On Wed, Jul 27, 2016 at 6:33 PM, Jan Viktorin <viktorin@rehivetech.com>
wrote:
> On Tue, 26 Jul 2016 18:22:21 +0200
> Thomas Monjalon <thomas.monjalon@6wind.com> wrote:
>
> > The right name of ethdev should be dpdk_netdev. However:
> > 1/ We are using rte_ prefix in the code and library names.
> > 2/ The API uses rte_ethdev
> > That's why 16.11 will just have the rte_ prefix prepended to
> > the library filename as every other libraries.
> >
> > Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> >
> Acked-by: Jan Viktorin <viktorin@rehivetech.com>
>
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] doc: announce ivshmem support removal
@ 2016-07-28 9:20 3% ` Christian Ehrhardt
2016-07-28 15:23 0% ` Mauricio Vasquez
0 siblings, 1 reply; 200+ results
From: Christian Ehrhardt @ 2016-07-28 9:20 UTC (permalink / raw)
To: Jan Viktorin; +Cc: Thomas Monjalon, anatoly.burakov, dev
Hi Thomas,
just my two cents as Ubuntu DPDK maintainer (and part of the Debian Team
that does the same).
(It seems I can reuse that line for all posts about the deprecation notices
:-) )
While IVSHMEM was enabled (as it was the default) I never heard of any
users of what we provided so far.
But that is expected considering that not all qemu bits are landed either.
Since it will follow the process of "deprecation notice -> grace period ->
ABI bump" on removal, I think packaging and consuming applications should
be fine.
I'd agree to Hiroshi that, if really needed a clean re-implementation more
properly separated from EAL likely is the best way to go.
I think it is a good change to drop rather complex code in favor of
stabilizing the main paths:
Acked-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd
On Wed, Jul 27, 2016 at 9:08 PM, Jan Viktorin <viktorin@rehivetech.com>
wrote:
> On Wed, 20 Jul 2016 18:35:46 +0200
> Thomas Monjalon <thomas.monjalon@6wind.com> wrote:
>
> > There was a prior call with an explanation of what needs to be done:
> > http://dpdk.org/ml/archives/dev/2016-June/040844.html
> > - Qemu patch upstreamed
> > - IVSHMEM PCI device managed by a PCI driver
> > - No DPDK objects (ring/mempool) allocated by EAL
> >
> > As nobody seems interested, it is time to remove this code which
> > makes EAL improvements harder.
> >
> > Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> > Acked-by: David Marchand <david.marchand@6wind.com>
> > Acked-by: Maxime Coquelin <maxime.coquelin@redhat.com>
>
> Acked-by: Jan Viktorin <viktorin@rehivetech.com
>
^ permalink raw reply [relevance 3%]
* [dpdk-dev] removal of old deprecation notice for Chelsio filtering
@ 2016-07-28 8:29 4% Thomas Monjalon
2016-07-28 10:12 0% ` Rahul Lakkireddy
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-28 8:29 UTC (permalink / raw)
To: Rahul Lakkireddy; +Cc: Kumar Sanghvi, dev
Hi Rahul,
We still have this deprecation notice:
* ABI changes are planned for adding four new flow types. This impacts
RTE_ETH_FLOW_MAX. The release 2.2 does not contain these ABI changes,
but release 2.3 will. [postponed]
Do you agree that we can remove it now we have a better generic filtering
API approach?
Thank you
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption
2016-07-27 22:14 3% [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption Alex Williamson
2016-07-28 6:54 0% ` Thomas Monjalon
@ 2016-07-28 8:06 0% ` Santosh Shukla
1 sibling, 0 replies; 200+ results
From: Santosh Shukla @ 2016-07-28 8:06 UTC (permalink / raw)
To: Alex Williamson; +Cc: anatoly.burakov, dev
On Thu, Jul 28, 2016 at 03:44:57AM +0530, Alex Williamson wrote:
> Hi,
>
> I took a quick look at the dpdk vfio code and spotted an invalid
> assumption that should probably be corrected ASAP. That is:
>
> lib/librte_eal/linuxapp/eal/eal_vfio.h:
> #define VFIO_GET_REGION_ADDR(x) ((uint64_t) x << 40ULL)
> #define VFIO_GET_REGION_IDX(x) (x >> 40)
Yes. I agree. We need some way to carry essential vfio region info in pci_dev,
needed for pread/pwrite. currently, rte_intr_handle only has vfio_dev_fd but
thats not sufficient information. I stumbled while adding ioport support in vfio
and took a short path to define region_idx thatway. To get-rid of this, Possible
approach could be;
- add essential vfio region specific info (ie.. offset, idx, flag) in
rte_intr_handle.
- or pull dev_fd to rte_pci_device{}; and define region specific details.
Thanks.
> Region offset to index is an implementation detail of the kernel, the
> vfio API defines that the offset of a given region (BAR) is found via
> the offset field of struct vfio_region_info returned via the
> VFIO_DEVICE_GET_REGION_INFO ioctl. You're free to cache the offset
> into any sort of local variable you like, but the kernel may change the
> implementation of region index to offset at any point in time. This is
> explicitly not part of the ABI. Is there a place to file a bug, or is
> this sufficient? Thanks,
>
> Alex
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption
2016-07-27 22:14 3% [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption Alex Williamson
@ 2016-07-28 6:54 0% ` Thomas Monjalon
2016-07-28 8:06 0% ` Santosh Shukla
1 sibling, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-28 6:54 UTC (permalink / raw)
To: Alex Williamson; +Cc: dev, anatoly.burakov
Hi,
2016-07-27 16:14, Alex Williamson:
> I took a quick look at the dpdk vfio code and spotted an invalid
> assumption that should probably be corrected ASAP.
It can theoretically be a bug but the value may never change in the kernel,
right?
So when you say ASAP, I feel it can wait the next DPDK release
(we plan to release today).
Do you agree?
> That is:
>
> lib/librte_eal/linuxapp/eal/eal_vfio.h:
> #define VFIO_GET_REGION_ADDR(x) ((uint64_t) x << 40ULL)
> #define VFIO_GET_REGION_IDX(x) (x >> 40)
>
> Region offset to index is an implementation detail of the kernel, the
> vfio API defines that the offset of a given region (BAR) is found via
> the offset field of struct vfio_region_info returned via the
> VFIO_DEVICE_GET_REGION_INFO ioctl. You're free to cache the offset
> into any sort of local variable you like, but the kernel may change the
> implementation of region index to offset at any point in time. This is
> explicitly not part of the ABI. Is there a place to file a bug, or is
> this sufficient? Thanks,
Thanks for the report. This email is sufficient :)
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
` (3 preceding siblings ...)
2016-07-28 2:35 4% ` John Daley (johndale)
@ 2016-07-28 2:39 4% ` Jerin Jacob
4 siblings, 0 replies; 200+ results
From: Jerin Jacob @ 2016-07-28 2:39 UTC (permalink / raw)
To: Olivier Matz; +Cc: dev, thomas.monjalon, bruce.richardson
On Wed, Jul 20, 2016 at 09:16:14AM +0200, Olivier Matz wrote:
> For 16.11, the mbuf structure will be modified implying ABI breakage.
> Some discussions already took place here:
> http://www.dpdk.org/dev/patchwork/patch/12878/
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
` (2 preceding siblings ...)
2016-07-27 9:34 4% ` Ananyev, Konstantin
@ 2016-07-28 2:35 4% ` John Daley (johndale)
2016-07-28 2:39 4% ` Jerin Jacob
4 siblings, 0 replies; 200+ results
From: John Daley (johndale) @ 2016-07-28 2:35 UTC (permalink / raw)
To: Olivier Matz, dev; +Cc: jerin.jacob, thomas.monjalon, bruce.richardson
>
> For 16.11, the mbuf structure will be modified implying ABI breakage.
> Some discussions already took place here:
> http://www.dpdk.org/dev/patchwork/patch/12878/
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> ---
Acked-by: John Daley <johndale@cisco.com>
Also, definitely +1 on trying to get m->next into the first cache line.
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-27 20:51 4% ` Ananyev, Konstantin
@ 2016-07-28 2:13 4% ` Jerin Jacob
2016-07-28 10:36 4% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Jerin Jacob @ 2016-07-28 2:13 UTC (permalink / raw)
To: Ananyev, Konstantin; +Cc: Thomas Monjalon, dev
On Wed, Jul 27, 2016 at 08:51:09PM +0000, Ananyev, Konstantin wrote:
>
> >
> > On Wed, Jul 27, 2016 at 05:33:01PM +0000, Ananyev, Konstantin wrote:
> > >
> > >
> > > > -----Original Message-----
> > > > From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> > > > Sent: Wednesday, July 27, 2016 6:11 PM
> > > > To: Thomas Monjalon <thomas.monjalon@6wind.com>
> > > > Cc: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org;
> > > > Ananyev, Konstantin <konstantin.ananyev@intel.com>
> > > > Subject: Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for
> > > > rte_eth_dev structure
> > > >
> > > > On Wed, Jul 27, 2016 at 01:59:01AM -0700, Thomas Monjalon wrote:
> > > > > > > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > > > > > > ---
> > > > > > > +* In 16.11 ABI changes are plained: the ``rte_eth_dev``
> > > > > > > +structure will be
> > > > > > > + extended with new function pointer ``tx_pkt_prep`` allowing
> > > > > > > +verification
> > > > > > > + and processing of packet burst to meet HW specific
> > > > > > > +requirements before
> > > > > > > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > > > > > > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing
> > > > > > > +information about number of
> > > > > > > + segments limit to be transmitted by device for TSO/non-TSO packets.
> > > > > >
> > > > > > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> > > > >
> > > > > I think I understand you want to split the TX processing:
> > > > > 1/ modify/write in mbufs
> > > > > 2/ write in HW
> > > > > and let application decide:
> > > > > - where the TX prep is done (which core)
> > > >
> > > > In what basics applications knows when and where to call tx_pkt_prep in fast path.
> > > > if all the time it needs to call before tx_burst then the PMD won't
> > > > have/don't need this callback waste cycles in fast path.Is this the expected behavior ?
> > > > Anything think it as compile time to make other PMDs wont suffer because of this change.
> > >
> > > Not sure what suffering you are talking about...
> > > Current model - i.e. when application does preparations (or doesn't if
> > > none is required) on its own and just call tx_burst() would still be there.
> > > If the app doesn't want to use tx_prep() by some reason - that still
> > > ok, and decision is up to the particular app.
> >
> > So my question is in what basics application decides to call the preparation.
> > Can you tell me the use case in application perspective?
>
> I suppose one most common use-case when application uses HW TX offloads,
> and don't' to cope on its own which L3/L4 header fields need to be filled
> for that particular dev_type/hw_offload combination.
If it does not cope up then it can skip tx'ing in the actual tx burst
itself and move the "skipped" tx packets to end of the list in the tx
burst so that application can take the action on "skipped" packet after
the tx burst
> Instead it just setups the ol_flags, fills tx_offload fields and calls tx_prep().
> Please read the original Tomasz's patch, I think he explained possible use-cases
> with lot of details.
Sorry, it is not very clear in terms of use cases.
In HW perspective, It it tries to avoid the illegal state. But not sure
calling "back to back" tx prepare and then tx burst how does it improve the
situation as the check illegal state check introduce in actual tx burst
it self.
In SW perspective, its try to avoid sending malformed packets. In my
view the same can achieved with existing tx burst it self as PMD is the
one finally send the packets on the wire.
proposal quote:
1. Introduce rte_eth_tx_prep() function to do necessary preparations of
packet burst to be safely transmitted on device for desired HW
offloads (set/reset checksum field according to the hardware
requirements) and check HW constraints (number of segments per
packet, etc).
While the limitations and requirements may differ for devices, it
requires to extend rte_eth_dev structure with new function pointer
"tx_pkt_prep" which can be implemented in the driver to prepare and
verify packets, in devices specific way, before burst, what should to
prevent application to send malformed packets.
>
> > and what if the PMD does not implement that callback then it is of waste cycles. Right?
>
> If you refer as lost cycles here something like:
> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP);
> then yes.
> Though comparing to actual work need to be done for most HW TX offloads,
> I think it is neglectable.
Not sure.
> Again, as I said before, it is totally voluntary for the application.
Not according to proposal. It can't be too as application has no idea
what PMD driver does with "prep" what is the implication on a HW if
application does not
Jerin
> Konstantin
>
> >
> > Jerin
> >
> >
> > > Konstantin
> > >
> > > >
> > > >
> > > > > - what to do if the TX prep fail
> > > > > So adding some processing in this first part becomes "not too
> > > > > expensive" or "manageable" from the application point of view.
> > > > >
> > > > > If I well understand the intent,
> > > > >
> > > > > Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com> (except
> > > > > typos ;)
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption
@ 2016-07-27 22:14 3% Alex Williamson
2016-07-28 6:54 0% ` Thomas Monjalon
2016-07-28 8:06 0% ` Santosh Shukla
0 siblings, 2 replies; 200+ results
From: Alex Williamson @ 2016-07-27 22:14 UTC (permalink / raw)
To: anatoly.burakov; +Cc: dev
Hi,
I took a quick look at the dpdk vfio code and spotted an invalid
assumption that should probably be corrected ASAP. That is:
lib/librte_eal/linuxapp/eal/eal_vfio.h:
#define VFIO_GET_REGION_ADDR(x) ((uint64_t) x << 40ULL)
#define VFIO_GET_REGION_IDX(x) (x >> 40)
Region offset to index is an implementation detail of the kernel, the
vfio API defines that the offset of a given region (BAR) is found via
the offset field of struct vfio_region_info returned via the
VFIO_DEVICE_GET_REGION_INFO ioctl. You're free to cache the offset
into any sort of local variable you like, but the kernel may change the
implementation of region index to offset at any point in time. This is
explicitly not part of the ABI. Is there a place to file a bug, or is
this sufficient? Thanks,
Alex
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-27 17:41 4% ` Jerin Jacob
@ 2016-07-27 20:51 4% ` Ananyev, Konstantin
2016-07-28 2:13 4% ` Jerin Jacob
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-07-27 20:51 UTC (permalink / raw)
To: Jerin Jacob; +Cc: Thomas Monjalon, dev
>
> On Wed, Jul 27, 2016 at 05:33:01PM +0000, Ananyev, Konstantin wrote:
> >
> >
> > > -----Original Message-----
> > > From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> > > Sent: Wednesday, July 27, 2016 6:11 PM
> > > To: Thomas Monjalon <thomas.monjalon@6wind.com>
> > > Cc: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org;
> > > Ananyev, Konstantin <konstantin.ananyev@intel.com>
> > > Subject: Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for
> > > rte_eth_dev structure
> > >
> > > On Wed, Jul 27, 2016 at 01:59:01AM -0700, Thomas Monjalon wrote:
> > > > > > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > > > > > ---
> > > > > > +* In 16.11 ABI changes are plained: the ``rte_eth_dev``
> > > > > > +structure will be
> > > > > > + extended with new function pointer ``tx_pkt_prep`` allowing
> > > > > > +verification
> > > > > > + and processing of packet burst to meet HW specific
> > > > > > +requirements before
> > > > > > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > > > > > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing
> > > > > > +information about number of
> > > > > > + segments limit to be transmitted by device for TSO/non-TSO packets.
> > > > >
> > > > > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> > > >
> > > > I think I understand you want to split the TX processing:
> > > > 1/ modify/write in mbufs
> > > > 2/ write in HW
> > > > and let application decide:
> > > > - where the TX prep is done (which core)
> > >
> > > In what basics applications knows when and where to call tx_pkt_prep in fast path.
> > > if all the time it needs to call before tx_burst then the PMD won't
> > > have/don't need this callback waste cycles in fast path.Is this the expected behavior ?
> > > Anything think it as compile time to make other PMDs wont suffer because of this change.
> >
> > Not sure what suffering you are talking about...
> > Current model - i.e. when application does preparations (or doesn't if
> > none is required) on its own and just call tx_burst() would still be there.
> > If the app doesn't want to use tx_prep() by some reason - that still
> > ok, and decision is up to the particular app.
>
> So my question is in what basics application decides to call the preparation.
> Can you tell me the use case in application perspective?
I suppose one most common use-case when application uses HW TX offloads,
and don't' to cope on its own which L3/L4 header fields need to be filled
for that particular dev_type/hw_offload combination.
Instead it just setups the ol_flags, fills tx_offload fields and calls tx_prep().
Please read the original Tomasz's patch, I think he explained possible use-cases
with lot of details.
> and what if the PMD does not implement that callback then it is of waste cycles. Right?
If you refer as lost cycles here something like:
RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->tx_prep, -ENOTSUP);
then yes.
Though comparing to actual work need to be done for most HW TX offloads,
I think it is neglectable.
Again, as I said before, it is totally voluntary for the application.
Konstantin
>
> Jerin
>
>
> > Konstantin
> >
> > >
> > >
> > > > - what to do if the TX prep fail
> > > > So adding some processing in this first part becomes "not too
> > > > expensive" or "manageable" from the application point of view.
> > > >
> > > > If I well understand the intent,
> > > >
> > > > Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com> (except
> > > > typos ;)
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-27 17:33 4% ` Ananyev, Konstantin
@ 2016-07-27 17:41 4% ` Jerin Jacob
2016-07-27 20:51 4% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Jerin Jacob @ 2016-07-27 17:41 UTC (permalink / raw)
To: Ananyev, Konstantin; +Cc: Thomas Monjalon, Kulasek, TomaszX, dev
On Wed, Jul 27, 2016 at 05:33:01PM +0000, Ananyev, Konstantin wrote:
>
>
> > -----Original Message-----
> > From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> > Sent: Wednesday, July 27, 2016 6:11 PM
> > To: Thomas Monjalon <thomas.monjalon@6wind.com>
> > Cc: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org; Ananyev, Konstantin <konstantin.ananyev@intel.com>
> > Subject: Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
> >
> > On Wed, Jul 27, 2016 at 01:59:01AM -0700, Thomas Monjalon wrote:
> > > > > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > > > > ---
> > > > > +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure
> > > > > +will be
> > > > > + extended with new function pointer ``tx_pkt_prep`` allowing
> > > > > +verification
> > > > > + and processing of packet burst to meet HW specific requirements
> > > > > +before
> > > > > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > > > > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information
> > > > > +about number of
> > > > > + segments limit to be transmitted by device for TSO/non-TSO packets.
> > > >
> > > > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> > >
> > > I think I understand you want to split the TX processing:
> > > 1/ modify/write in mbufs
> > > 2/ write in HW
> > > and let application decide:
> > > - where the TX prep is done (which core)
> >
> > In what basics applications knows when and where to call tx_pkt_prep in fast path.
> > if all the time it needs to call before tx_burst then the PMD won't have/don't need this callback waste cycles in fast path.Is this the expected
> > behavior ?
> > Anything think it as compile time to make other PMDs wont suffer because of this change.
>
> Not sure what suffering you are talking about...
> Current model - i.e. when application does preparations (or doesn't if none is required)
> on its own and just call tx_burst() would still be there.
> If the app doesn't want to use tx_prep() by some reason - that still ok,
> and decision is up to the particular app.
So my question is in what basics application decides to call the preparation.
Can you tell me the use case in application perspective?
and what if the PMD does not implement that callback then it is of waste
cycles. Right?
Jerin
> Konstantin
>
> >
> >
> > > - what to do if the TX prep fail
> > > So adding some processing in this first part becomes "not too
> > > expensive" or "manageable" from the application point of view.
> > >
> > > If I well understand the intent,
> > >
> > > Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com> (except typos ;)
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-27 17:10 4% ` Jerin Jacob
@ 2016-07-27 17:33 4% ` Ananyev, Konstantin
2016-07-27 17:41 4% ` Jerin Jacob
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-07-27 17:33 UTC (permalink / raw)
To: Jerin Jacob, Thomas Monjalon; +Cc: Kulasek, TomaszX, dev
> -----Original Message-----
> From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> Sent: Wednesday, July 27, 2016 6:11 PM
> To: Thomas Monjalon <thomas.monjalon@6wind.com>
> Cc: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org; Ananyev, Konstantin <konstantin.ananyev@intel.com>
> Subject: Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
>
> On Wed, Jul 27, 2016 at 01:59:01AM -0700, Thomas Monjalon wrote:
> > > > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > > > ---
> > > > +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure
> > > > +will be
> > > > + extended with new function pointer ``tx_pkt_prep`` allowing
> > > > +verification
> > > > + and processing of packet burst to meet HW specific requirements
> > > > +before
> > > > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > > > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information
> > > > +about number of
> > > > + segments limit to be transmitted by device for TSO/non-TSO packets.
> > >
> > > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> >
> > I think I understand you want to split the TX processing:
> > 1/ modify/write in mbufs
> > 2/ write in HW
> > and let application decide:
> > - where the TX prep is done (which core)
>
> In what basics applications knows when and where to call tx_pkt_prep in fast path.
> if all the time it needs to call before tx_burst then the PMD won't have/don't need this callback waste cycles in fast path.Is this the expected
> behavior ?
> Anything think it as compile time to make other PMDs wont suffer because of this change.
Not sure what suffering you are talking about...
Current model - i.e. when application does preparations (or doesn't if none is required)
on its own and just call tx_burst() would still be there.
If the app doesn't want to use tx_prep() by some reason - that still ok,
and decision is up to the particular app.
Konstantin
>
>
> > - what to do if the TX prep fail
> > So adding some processing in this first part becomes "not too
> > expensive" or "manageable" from the application point of view.
> >
> > If I well understand the intent,
> >
> > Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com> (except typos ;)
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-27 8:59 4% ` Thomas Monjalon
@ 2016-07-27 17:10 4% ` Jerin Jacob
2016-07-27 17:33 4% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Jerin Jacob @ 2016-07-27 17:10 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: Kulasek, TomaszX, dev, Ananyev, Konstantin
On Wed, Jul 27, 2016 at 01:59:01AM -0700, Thomas Monjalon wrote:
> > > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > > ---
> > > +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
> > > + extended with new function pointer ``tx_pkt_prep`` allowing verification
> > > + and processing of packet burst to meet HW specific requirements before
> > > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
> > > + segments limit to be transmitted by device for TSO/non-TSO packets.
> >
> > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
>
> I think I understand you want to split the TX processing:
> 1/ modify/write in mbufs
> 2/ write in HW
> and let application decide:
> - where the TX prep is done (which core)
In what basics applications knows when and where to call tx_pkt_prep in fast path.
if all the time it needs to call before tx_burst then the PMD won't have/don't need this
callback waste cycles in fast path.Is this the expected behavior ?
Anything think it as compile time to make other PMDs wont suffer because
of this change.
> - what to do if the TX prep fail
> So adding some processing in this first part becomes "not too expensive" or
> "manageable" from the application point of view.
>
> If I well understand the intent,
>
> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> (except typos ;)
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change of struct rte_port_source_params and rte_port_sink_params
2016-07-27 10:08 9% ` Dumitrescu, Cristian
@ 2016-07-27 10:42 7% ` Thomas Monjalon
2016-07-28 18:28 4% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-27 10:42 UTC (permalink / raw)
To: Dumitrescu, Cristian, Zhang, Roy Fan
Cc: dev, Panu Matilainen, Singh, Jasvinder
2016-07-27 10:08, Dumitrescu, Cristian:
> As Thomas mentioned, today is probably the last day to discuss ABI changes. This one is pretty small and straightforward, any issues with it?
>
> Panu had a concern that the change from "char *" to "const char *" is too small to be regarded as ABI breakage and we should simply go ahead and do it. My conservative proposal was to put a notice anyway.
>
> Nonetheless, what I would like to get from Thomas and Panu is a path forward for this now:
> a) If we agree to consider this an ABI change, please merge the notice for 16.7;
Panu was noticing 3 things (and I agree with them):
- it is an API change
- they can be grouped in only one list item
- it is better to wait having more changes to break an API
About the third point, in this specific case, I think it is acceptable because:
- it should not break the ABI
- the impact of the API change is really small
- I'm not sure the packet framework should be considered as a DPDK API.
> b) If we agree this is too small for an ABI change, please let us agree now
> to accept our quick patch for 16.11 for this change.
For an API deprecation notice (reworded),
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> > -----Original Message-----
> > The ABI changes are planned for rte_port_source_params and
> > rte_port_sink_params, which will be supported from release 16.11. Here
> > announces that ABI changes in detail.
> >
> > Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> > Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
> > ---
> > +* ABI will change for rte_port_source_params struct. The member
> > file_name
> > + data type will be changed from char * to const char *. This change targets
> > + release 16.11
> > +
> > +* ABI will change for rte_port_sink_params struct. The member file_name
> > + data type will be changed from char * to const char *. This change targets
> > + release 16.11
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change of struct rte_port_source_params and rte_port_sink_params
@ 2016-07-27 10:08 9% ` Dumitrescu, Cristian
2016-07-27 10:42 7% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Dumitrescu, Cristian @ 2016-07-27 10:08 UTC (permalink / raw)
To: Thomas Monjalon, Panu Matilainen; +Cc: Zhang, Roy Fan, Singh, Jasvinder, dev
As Thomas mentioned, today is probably the last day to discuss ABI changes. This one is pretty small and straightforward, any issues with it?
Panu had a concern that the change from "char *" to "const char *" is too small to be regarded as ABI breakage and we should simply go ahead and do it. My conservative proposal was to put a notice anyway.
Nonetheless, what I would like to get from Thomas and Panu is a path forward for this now:
a) If we agree to consider this an ABI change, please merge the notice for 16.7;
b) If we agree this is too small for an ABI change, please let us agree now to accept our quick patch for 16.11 for this change.
Thanks,
Cristian
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Fan Zhang
> Sent: Thursday, May 19, 2016 3:19 PM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH v2] doc: announce ABI change of struct
> rte_port_source_params and rte_port_sink_params
>
> The ABI changes are planned for rte_port_source_params and
> rte_port_sink_params, which will be supported from release 16.11. Here
> announces that ABI changes in detail.
>
> Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
> ---
> doc/guides/rel_notes/deprecation.rst | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst
> b/doc/guides/rel_notes/deprecation.rst
> index fffe9c7..4f3fefe 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -74,3 +74,11 @@ Deprecation Notices
> a handle, like the way kernel exposes an fd to user for locating a
> specific file, and to keep all major structures internally, so that
> we are likely to be free from ABI violations in future.
> +
> +* ABI will change for rte_port_source_params struct. The member
> file_name
> + data type will be changed from char * to const char *. This change targets
> + release 16.11
> +
> +* ABI will change for rte_port_sink_params struct. The member file_name
> + data type will be changed from char * to const char *. This change targets
> + release 16.11
> --
> 2.5.5
^ permalink raw reply [relevance 9%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
2016-07-20 8:54 4% ` Ferruh Yigit
2016-07-27 8:33 4% ` Thomas Monjalon
@ 2016-07-27 9:34 4% ` Ananyev, Konstantin
2016-07-28 2:35 4% ` John Daley (johndale)
2016-07-28 2:39 4% ` Jerin Jacob
4 siblings, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-07-27 9:34 UTC (permalink / raw)
To: Olivier Matz, dev; +Cc: jerin.jacob, thomas.monjalon, Richardson, Bruce
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Olivier Matz
> Sent: Wednesday, July 20, 2016 8:16 AM
> To: dev@dpdk.org
> Cc: jerin.jacob@caviumnetworks.com; thomas.monjalon@6wind.com; Richardson, Bruce <bruce.richardson@intel.com>
> Subject: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
>
> For 16.11, the mbuf structure will be modified implying ABI breakage.
> Some discussions already took place here:
> http://www.dpdk.org/dev/patchwork/patch/12878/
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> ---
>
> v1->v2:
> - reword the sentences to keep things more open, as suggested by Bruce
>
> doc/guides/rel_notes/deprecation.rst | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index f502f86..b9f5a93 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -41,3 +41,9 @@ Deprecation Notices
> * The mempool functions for single/multi producer/consumer are deprecated and
> will be removed in 16.11.
> It is replaced by rte_mempool_generic_get/put functions.
> +
> +* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some fields
> + may be reordered to facilitate the writing of ``data_off``, ``refcnt``, and
> + ``nb_segs`` in one operation, because some platforms have an overhead if the
> + store address is not naturally aligned. Other mbuf fields, such as the
> + ``port`` field, may be moved or removed as part of this mbuf work.
> --
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> 2.8.1
^ permalink raw reply [relevance 4%]
* [dpdk-dev] last days for deprecation notices
@ 2016-07-27 9:15 3% Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-27 9:15 UTC (permalink / raw)
To: dev
Hi everybody,
There are some announces pending to make some changes in 16.11 which
will break the API or ABI:
http://dpdk.org/dev/patchwork/project/dpdk/list/?q=announce
Some of them are really good but will probably not happen because there
is no visible consensus (or often no discussion at all).
Such changes need 3 "meaningful" acks to be accepted:
http://dpdk.org/browse/dpdk/tree/doc/guides/contributing/versioning.rst#n56
Note: 3 acks from the same company are not really "meaningful" ;)
The release is planned on Thursday 28, so it is almost the last day
to discuss these changes.
Thanks
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-21 22:48 4% ` Ananyev, Konstantin
@ 2016-07-27 8:59 4% ` Thomas Monjalon
2016-07-27 17:10 4% ` Jerin Jacob
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-27 8:59 UTC (permalink / raw)
To: Kulasek, TomaszX; +Cc: dev, Ananyev, Konstantin
> > Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> > ---
> > +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
> > + extended with new function pointer ``tx_pkt_prep`` allowing verification
> > + and processing of packet burst to meet HW specific requirements before
> > + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> > + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
> > + segments limit to be transmitted by device for TSO/non-TSO packets.
>
> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
I think I understand you want to split the TX processing:
1/ modify/write in mbufs
2/ write in HW
and let application decide:
- where the TX prep is done (which core)
- what to do if the TX prep fail
So adding some processing in this first part becomes "not too expensive" or
"manageable" from the application point of view.
If I well understand the intent,
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
(except typos ;)
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
2016-07-20 8:54 4% ` Ferruh Yigit
@ 2016-07-27 8:33 4% ` Thomas Monjalon
2016-07-28 18:04 4% ` Thomas Monjalon
2016-07-27 9:34 4% ` Ananyev, Konstantin
` (2 subsequent siblings)
4 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-27 8:33 UTC (permalink / raw)
To: Olivier Matz; +Cc: dev, jerin.jacob, bruce.richardson
> For 16.11, the mbuf structure will be modified implying ABI breakage.
> Some discussions already took place here:
> http://www.dpdk.org/dev/patchwork/patch/12878/
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-21 15:24 11% ` [dpdk-dev] [PATCH v2] " Tomasz Kulasek
@ 2016-07-21 22:48 4% ` Ananyev, Konstantin
2016-07-27 8:59 4% ` Thomas Monjalon
2016-07-28 12:04 4% ` Avi Kivity
1 sibling, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-07-21 22:48 UTC (permalink / raw)
To: Kulasek, TomaszX, dev
>
> This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
> changes in rte_eth_dev and rte_eth_desc_lim structures.
>
> As discussed in that thread:
>
> http://dpdk.org/ml/archives/dev/2015-September/023603.html
>
> Different NIC models depending on HW offload requested might impose
> different requirements on packets to be TX-ed in terms of:
>
> - Max number of fragments per packet allowed
> - Max number of fragments per TSO segments
> - The way pseudo-header checksum should be pre-calculated
> - L3/L4 header fields filling
> - etc.
>
>
> MOTIVATION:
> -----------
>
> 1) Some work cannot (and didn't should) be done in rte_eth_tx_burst.
> However, this work is sometimes required, and now, it's an
> application issue.
>
> 2) Different hardware may have different requirements for TX offloads,
> other subset can be supported and so on.
>
> 3) Some parameters (eg. number of segments in ixgbe driver) may hung
> device. These parameters may be vary for different devices.
>
> For example i40e HW allows 8 fragments per packet, but that is after
> TSO segmentation. While ixgbe has a 38-fragment pre-TSO limit.
>
> 4) Fields in packet may require different initialization (like eg. will
> require pseudo-header checksum precalculation, sometimes in a
> different way depending on packet type, and so on). Now application
> needs to care about it.
>
> 5) Using additional API (rte_eth_tx_prep) before rte_eth_tx_burst let to
> prepare packet burst in acceptable form for specific device.
>
> 6) Some additional checks may be done in debug mode keeping tx_burst
> implementation clean.
>
>
> PROPOSAL:
> ---------
>
> To help user to deal with all these varieties we propose to:
>
> 1. Introduce rte_eth_tx_prep() function to do necessary preparations of
> packet burst to be safely transmitted on device for desired HW
> offloads (set/reset checksum field according to the hardware
> requirements) and check HW constraints (number of segments per
> packet, etc).
>
> While the limitations and requirements may differ for devices, it
> requires to extend rte_eth_dev structure with new function pointer
> "tx_pkt_prep" which can be implemented in the driver to prepare and
> verify packets, in devices specific way, before burst, what should to
> prevent application to send malformed packets.
>
> 2. Also new fields will be introduced in rte_eth_desc_lim:
> nb_seg_max and nb_mtu_seg_max, providing an information about max
> segments in TSO and non-TSO packets acceptable by device.
>
> This information is useful for application to not create/limit
> malicious packet.
>
>
> APPLICATION (CASE OF USE):
> --------------------------
>
> 1) Application should to initialize burst of packets to send, set
> required tx offload flags and required fields, like l2_len, l3_len,
> l4_len, and tso_segsz
>
> 2) Application passes burst to the rte_eth_tx_prep to check conditions
> required to send packets through the NIC.
>
> 3) The result of rte_eth_tx_prep can be used to send valid packets
> and/or restore invalid if function fails.
>
> eg.
>
> for (i = 0; i < nb_pkts; i++) {
>
> /* initialize or process packet */
>
> bufs[i]->tso_segsz = 800;
> bufs[i]->ol_flags = PKT_TX_TCP_SEG | PKT_TX_IPV4
> | PKT_TX_IP_CKSUM;
> bufs[i]->l2_len = sizeof(struct ether_hdr);
> bufs[i]->l3_len = sizeof(struct ipv4_hdr);
> bufs[i]->l4_len = sizeof(struct tcp_hdr);
> }
>
> /* Prepare burst of TX packets */
> nb_prep = rte_eth_tx_prep(port, 0, bufs, nb_pkts);
>
> if (nb_prep < nb_pkts) {
> printf("tx_prep failed\n");
>
> /* drop or restore invalid packets */
>
> }
>
> /* Send burst of TX packets */
> nb_tx = rte_eth_tx_burst(port, 0, bufs, nb_prep);
>
> /* Free any unsent packets. */
>
>
>
> Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
> ---
> doc/guides/rel_notes/deprecation.rst | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index f502f86..485aacb 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -41,3 +41,10 @@ Deprecation Notices
> * The mempool functions for single/multi producer/consumer are deprecated and
> will be removed in 16.11.
> It is replaced by rte_mempool_generic_get/put functions.
> +
> +* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
> + extended with new function pointer ``tx_pkt_prep`` allowing verification
> + and processing of packet burst to meet HW specific requirements before
> + transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
> + ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
> + segments limit to be transmitted by device for TSO/non-TSO packets.
> --
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> 1.7.9.5
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] mempool: adjust name string size in related data types
2016-07-21 14:25 0% ` Olivier Matz
@ 2016-07-21 21:16 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-21 21:16 UTC (permalink / raw)
To: Zoltan Kiss; +Cc: dev, Olivier Matz, Zoltan Kiss, Bruce Richardson
2016-07-21 16:25, Olivier Matz:
> On 07/21/2016 03:47 PM, Zoltan Kiss wrote:
> > On 21/07/16 14:40, Olivier Matz wrote:
> >> On 07/20/2016 07:16 PM, Zoltan Kiss wrote:
> >>> A recent patch brought up an issue about the size of the 'name' fields:
> >>>
> >>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
> >>>
> >>> These relations should be observed:
> >>>
> >>> 1. Each ring creates a memzone with a prefixed name:
> >>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
> >>>
> >>> 2. There are some mempool handlers which create a ring with a prefixed
> >>> name:
> >>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE -
> >>> strlen(RTE_MEMPOOL_MZ_PREFIX)
> >>>
> >>> 3. A mempool can create up to RTE_MAX_MEMZONE pre and postfixed
> >>> memzones:
> >>> sprintf(postfix, "_%d", RTE_MAX_MEMZONE)
> >>> RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
> >>> strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen(postfix)
> >>>
> >>> Setting all of them to 32 hides this restriction from the application.
> >>> This patch decreases the mempool and ring string size to accommodate for
> >>> these prefixes, but it doesn't apply the 3rd constraint. Applications
> >>> relying on these constants need to be recompiled, otherwise they'll run
> >>> into ENAMETOOLONG issues.
> >>> The size of the arrays are kept 32 for ABI compatibility, it can be
> >>> decreased next time the ABI changes.
> >>>
> >>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
> >>
> >> Looks like to be a good compromise for the 16.07 release. One question
> >> however: why not taking in account the 3rd constraint? Because it may
> >> not completly fix the issue if the mempool is fragmented.
> >>
> >> We could define RTE_MEMPOOL_NAMESIZE to 24
> >> = 32 - len('mp_') - len('_0123'))
> >
> > I was trying to figure out a compile time macro for strlen(postfix), but
> > I could not. Your suggestion would work only until someone increases
> > RTE_MAX_MEMZONE above 9999. As the likelihood of fragmenting a pool over
> > 99 memzones seems small, I did not bother to fix this with an ugly hack,
> > but if you think we should include it, let me know!
>
> Ok, looks fair, thanks for the clarification.
>
> Acked-by: Olivier Matz <olivier.matz@6wind.com>
Applied, thanks
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v2] doc: announce ABI change for rte_eth_dev structure
2016-07-20 14:24 13% [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure Tomasz Kulasek
2016-07-20 15:01 4% ` Thomas Monjalon
@ 2016-07-21 15:24 11% ` Tomasz Kulasek
2016-07-21 22:48 4% ` Ananyev, Konstantin
2016-07-28 12:04 4% ` Avi Kivity
2016-07-31 7:46 4% ` [dpdk-dev] [PATCH] " Vlad Zolotarov
2 siblings, 2 replies; 200+ results
From: Tomasz Kulasek @ 2016-07-21 15:24 UTC (permalink / raw)
To: dev
This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
changes in rte_eth_dev and rte_eth_desc_lim structures.
As discussed in that thread:
http://dpdk.org/ml/archives/dev/2015-September/023603.html
Different NIC models depending on HW offload requested might impose
different requirements on packets to be TX-ed in terms of:
- Max number of fragments per packet allowed
- Max number of fragments per TSO segments
- The way pseudo-header checksum should be pre-calculated
- L3/L4 header fields filling
- etc.
MOTIVATION:
-----------
1) Some work cannot (and didn't should) be done in rte_eth_tx_burst.
However, this work is sometimes required, and now, it's an
application issue.
2) Different hardware may have different requirements for TX offloads,
other subset can be supported and so on.
3) Some parameters (eg. number of segments in ixgbe driver) may hung
device. These parameters may be vary for different devices.
For example i40e HW allows 8 fragments per packet, but that is after
TSO segmentation. While ixgbe has a 38-fragment pre-TSO limit.
4) Fields in packet may require different initialization (like eg. will
require pseudo-header checksum precalculation, sometimes in a
different way depending on packet type, and so on). Now application
needs to care about it.
5) Using additional API (rte_eth_tx_prep) before rte_eth_tx_burst let to
prepare packet burst in acceptable form for specific device.
6) Some additional checks may be done in debug mode keeping tx_burst
implementation clean.
PROPOSAL:
---------
To help user to deal with all these varieties we propose to:
1. Introduce rte_eth_tx_prep() function to do necessary preparations of
packet burst to be safely transmitted on device for desired HW
offloads (set/reset checksum field according to the hardware
requirements) and check HW constraints (number of segments per
packet, etc).
While the limitations and requirements may differ for devices, it
requires to extend rte_eth_dev structure with new function pointer
"tx_pkt_prep" which can be implemented in the driver to prepare and
verify packets, in devices specific way, before burst, what should to
prevent application to send malformed packets.
2. Also new fields will be introduced in rte_eth_desc_lim:
nb_seg_max and nb_mtu_seg_max, providing an information about max
segments in TSO and non-TSO packets acceptable by device.
This information is useful for application to not create/limit
malicious packet.
APPLICATION (CASE OF USE):
--------------------------
1) Application should to initialize burst of packets to send, set
required tx offload flags and required fields, like l2_len, l3_len,
l4_len, and tso_segsz
2) Application passes burst to the rte_eth_tx_prep to check conditions
required to send packets through the NIC.
3) The result of rte_eth_tx_prep can be used to send valid packets
and/or restore invalid if function fails.
eg.
for (i = 0; i < nb_pkts; i++) {
/* initialize or process packet */
bufs[i]->tso_segsz = 800;
bufs[i]->ol_flags = PKT_TX_TCP_SEG | PKT_TX_IPV4
| PKT_TX_IP_CKSUM;
bufs[i]->l2_len = sizeof(struct ether_hdr);
bufs[i]->l3_len = sizeof(struct ipv4_hdr);
bufs[i]->l4_len = sizeof(struct tcp_hdr);
}
/* Prepare burst of TX packets */
nb_prep = rte_eth_tx_prep(port, 0, bufs, nb_pkts);
if (nb_prep < nb_pkts) {
printf("tx_prep failed\n");
/* drop or restore invalid packets */
}
/* Send burst of TX packets */
nb_tx = rte_eth_tx_burst(port, 0, bufs, nb_prep);
/* Free any unsent packets. */
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
---
doc/guides/rel_notes/deprecation.rst | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f502f86..485aacb 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -41,3 +41,10 @@ Deprecation Notices
* The mempool functions for single/multi producer/consumer are deprecated and
will be removed in 16.11.
It is replaced by rte_mempool_generic_get/put functions.
+
+* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
+ extended with new function pointer ``tx_pkt_prep`` allowing verification
+ and processing of packet burst to meet HW specific requirements before
+ transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
+ ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
+ segments limit to be transmitted by device for TSO/non-TSO packets.
--
1.7.9.5
^ permalink raw reply [relevance 11%]
* Re: [dpdk-dev] [PATCH v2] mempool: adjust name string size in related data types
2016-07-21 13:47 0% ` Zoltan Kiss
@ 2016-07-21 14:25 0% ` Olivier Matz
2016-07-21 21:16 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-07-21 14:25 UTC (permalink / raw)
To: Zoltan Kiss, Zoltan Kiss, dev; +Cc: Bruce Richardson
On 07/21/2016 03:47 PM, Zoltan Kiss wrote:
>
>
> On 21/07/16 14:40, Olivier Matz wrote:
>> Hi Zoltan,
>>
>>
>> On 07/20/2016 07:16 PM, Zoltan Kiss wrote:
>>> A recent patch brought up an issue about the size of the 'name' fields:
>>>
>>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>>
>>> These relations should be observed:
>>>
>>> 1. Each ring creates a memzone with a prefixed name:
>>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>>>
>>> 2. There are some mempool handlers which create a ring with a prefixed
>>> name:
>>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE -
>>> strlen(RTE_MEMPOOL_MZ_PREFIX)
>>>
>>> 3. A mempool can create up to RTE_MAX_MEMZONE pre and postfixed
>>> memzones:
>>> sprintf(postfix, "_%d", RTE_MAX_MEMZONE)
>>> RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
>>> strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen(postfix)
>>>
>>> Setting all of them to 32 hides this restriction from the application.
>>> This patch decreases the mempool and ring string size to accommodate for
>>> these prefixes, but it doesn't apply the 3rd constraint. Applications
>>> relying on these constants need to be recompiled, otherwise they'll run
>>> into ENAMETOOLONG issues.
>>> The size of the arrays are kept 32 for ABI compatibility, it can be
>>> decreased next time the ABI changes.
>>>
>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>>
>> Looks like to be a good compromise for the 16.07 release. One question
>> however: why not taking in account the 3rd constraint? Because it may
>> not completly fix the issue if the mempool is fragmented.
>>
>> We could define RTE_MEMPOOL_NAMESIZE to 24
>> = 32 - len('mp_') - len('_0123'))
>
> I was trying to figure out a compile time macro for strlen(postfix), but
> I could not. Your suggestion would work only until someone increases
> RTE_MAX_MEMZONE above 9999. As the likelihood of fragmenting a pool over
> 99 memzones seems small, I did not bother to fix this with an ugly hack,
> but if you think we should include it, let me know!
Ok, looks fair, thanks for the clarification.
Acked-by: Olivier Matz <olivier.matz@6wind.com>
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v2] mempool: adjust name string size in related data types
2016-07-21 13:40 0% ` Olivier Matz
@ 2016-07-21 13:47 0% ` Zoltan Kiss
2016-07-21 14:25 0% ` Olivier Matz
0 siblings, 1 reply; 200+ results
From: Zoltan Kiss @ 2016-07-21 13:47 UTC (permalink / raw)
To: Olivier Matz, Zoltan Kiss, dev; +Cc: Bruce Richardson
On 21/07/16 14:40, Olivier Matz wrote:
> Hi Zoltan,
>
>
> On 07/20/2016 07:16 PM, Zoltan Kiss wrote:
>> A recent patch brought up an issue about the size of the 'name' fields:
>>
>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>
>> These relations should be observed:
>>
>> 1. Each ring creates a memzone with a prefixed name:
>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>>
>> 2. There are some mempool handlers which create a ring with a prefixed
>> name:
>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE - strlen(RTE_MEMPOOL_MZ_PREFIX)
>>
>> 3. A mempool can create up to RTE_MAX_MEMZONE pre and postfixed memzones:
>> sprintf(postfix, "_%d", RTE_MAX_MEMZONE)
>> RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
>> strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen(postfix)
>>
>> Setting all of them to 32 hides this restriction from the application.
>> This patch decreases the mempool and ring string size to accommodate for
>> these prefixes, but it doesn't apply the 3rd constraint. Applications
>> relying on these constants need to be recompiled, otherwise they'll run
>> into ENAMETOOLONG issues.
>> The size of the arrays are kept 32 for ABI compatibility, it can be
>> decreased next time the ABI changes.
>>
>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>
> Looks like to be a good compromise for the 16.07 release. One question
> however: why not taking in account the 3rd constraint? Because it may
> not completly fix the issue if the mempool is fragmented.
>
> We could define RTE_MEMPOOL_NAMESIZE to 24
> = 32 - len('mp_') - len('_0123'))
I was trying to figure out a compile time macro for strlen(postfix), but
I could not. Your suggestion would work only until someone increases
RTE_MAX_MEMZONE above 9999. As the likelihood of fragmenting a pool over
99 memzones seems small, I did not bother to fix this with an ugly hack,
but if you think we should include it, let me know!
>
> Thanks,
> Olivier
>
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v2] mempool: adjust name string size in related data types
2016-07-20 17:16 12% ` [dpdk-dev] [PATCH v2] " Zoltan Kiss
@ 2016-07-21 13:40 0% ` Olivier Matz
2016-07-21 13:47 0% ` Zoltan Kiss
0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-07-21 13:40 UTC (permalink / raw)
To: Zoltan Kiss, dev; +Cc: Bruce Richardson
Hi Zoltan,
On 07/20/2016 07:16 PM, Zoltan Kiss wrote:
> A recent patch brought up an issue about the size of the 'name' fields:
>
> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>
> These relations should be observed:
>
> 1. Each ring creates a memzone with a prefixed name:
> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>
> 2. There are some mempool handlers which create a ring with a prefixed
> name:
> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE - strlen(RTE_MEMPOOL_MZ_PREFIX)
>
> 3. A mempool can create up to RTE_MAX_MEMZONE pre and postfixed memzones:
> sprintf(postfix, "_%d", RTE_MAX_MEMZONE)
> RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
> strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen(postfix)
>
> Setting all of them to 32 hides this restriction from the application.
> This patch decreases the mempool and ring string size to accommodate for
> these prefixes, but it doesn't apply the 3rd constraint. Applications
> relying on these constants need to be recompiled, otherwise they'll run
> into ENAMETOOLONG issues.
> The size of the arrays are kept 32 for ABI compatibility, it can be
> decreased next time the ABI changes.
>
> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
Looks like to be a good compromise for the 16.07 release. One question
however: why not taking in account the 3rd constraint? Because it may
not completly fix the issue if the mempool is fragmented.
We could define RTE_MEMPOOL_NAMESIZE to 24
= 32 - len('mp_') - len('_0123'))
Thanks,
Olivier
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-20 10:41 2% ` Adrien Mazarguil
@ 2016-07-21 3:18 0% ` Lu, Wenzhuo
0 siblings, 0 replies; 200+ results
From: Lu, Wenzhuo @ 2016-07-21 3:18 UTC (permalink / raw)
To: Adrien Mazarguil
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
Hi Adrien,
> -----Original Message-----
> From: Adrien Mazarguil [mailto:adrien.mazarguil@6wind.com]
> Sent: Wednesday, July 20, 2016 6:41 PM
> To: Lu, Wenzhuo
> Cc: dev@dpdk.org; Thomas Monjalon; Zhang, Helin; Wu, Jingjing; Rasesh Mody;
> Ajit Khaparde; Rahul Lakkireddy; Jan Medala; John Daley; Chen, Jing D; Ananyev,
> Konstantin; Matej Vido; Alejandro Lucero; Sony Chacko; Jerin Jacob; De Lara
> Guarch, Pablo; Olga Shern
> Subject: Re: [RFC] Generic flow director/filtering/classification API
>
> Hi Wenzhuo,
>
> On Wed, Jul 20, 2016 at 02:16:51AM +0000, Lu, Wenzhuo wrote:
> [...]
> > > So, today an application cannot combine N-tuple and FDIR flow rules
> > > and get a reliable outcome, unless it is designed for specific
> > > devices with a known behavior.
> > >
> > > > What's the right behavior of PMD if APP want to create a flow
> > > > director rule
> > > which has a higher or even equal priority than an existing n-tuple
> > > rule? Should PMD return fail?
> > >
> > > First remember applications only deal with the generic API, PMDs are
> > > responsible for choosing the most appropriate HW implementation to
> > > use according to the requested flow rules (FDIR, N-tuple or anything else).
> > >
> > > For the specific case of FDIR vs N-tuple, if the underlying HW
> > > supports both I do not see why the PMD would create a N-tuple rule.
> > > Doesn't FDIR support everything N-tuple can do and much more?
> > Talking about the filters, fdir can cover n-tuple. I think that's why i40e only
> supports fdir but not n-tuple. But n-tuple has its own highlight. As we know, at
> least on intel NICs, fdir only supports per device mask. But n-tuple can support
> per rule mask.
> > As every pattern has spec and mask both, we cannot guarantee the masks are
> same. I think ixgbe will try to use n-tuple first if can. Because even the masks are
> different, we can support them all.
>
> OK, makes sense. In that case existing rules may indeed prevent subsequent
> ones from getting created if their priority is wrong. I do not think there is a way
> around that if the application needs this exact ordering.
Agree. I don’t see any workaround either. PMD has to return fail sometimes.
>
> > > Assuming such a thing happened anyway, that the PMD had to create a
> > > rule using a high priority filter type and that the application
> > > requests the creation of a rule that can only be done using a lower
> > > priority filter type, but also requested a higher priority for that rule, then yes,
> it should obviously fail.
> > >
> > > That is, unless the PMD can perform some kind of workaround to have both.
> > >
> > > > If so, do we need more fail reasons? According to this RFC, I
> > > > think we need
> > > return " EEXIST: collision with an existing rule. ", but it's not
> > > very clear, APP doesn't know the problem is priority, maybe more detailed
> reason is helpful.
> > >
> > > Possibly, I've defined a basic set of errors, there are quite a
> > > number of errno values to choose from. However I think we should not
> define too many values.
> > > In my opinion the basic set covers every possible failure:
> > >
> > > - EINVAL: invalid format, rule is broken or cannot be understood by the PMD
> > > anyhow.
> > >
> > > - ENOTSUP: pattern/actions look fine but something in the requested rule is
> > > not supported and thus cannot be applied.
> > >
> > > - EEXIST: pattern/actions are fine and could have been applied if only some
> > > other rule did not prevent the PMD to do it (I see it as the closest thing
> > > to "ETOOBAD" which unfortunately does not exist).
> > >
> > > - ENOMEM: like EEXIST, except it is due to the lack of resources not because
> > > of another rule. I wasn't sure which of ENOMEM or ENOSPC was better but
> > > settled on ENOMEM as it is well known. Still open to debate.
> > >
> > > Errno values are only useful to get a rough idea of the reason, and
> > > another mechanism is needed to pinpoint the exact problem for
> > > debugging/reporting purposes, something like:
> > >
> > > enum rte_flow_error_type {
> > > RTE_FLOW_ERROR_TYPE_NONE,
> > > RTE_FLOW_ERROR_TYPE_UNKNOWN,
> > > RTE_FLOW_ERROR_TYPE_PRIORITY,
> > > RTE_FLOW_ERROR_TYPE_PATTERN,
> > > RTE_FLOW_ERROR_TYPE_ACTION,
> > > };
> > >
> > > struct rte_flow_error {
> > > enum rte_flow_error_type type;
> > > void *offset; /* Points to the exact pattern item or action. */
> > > const char *message;
> > > };
> > When we are using a CLI and it fails, normally it will let us know
> > which parameter is not appropriate. So, I think it’s a good idea to
> > have this error structure :)
>
> Agreed.
>
> > > Then either provide an optional struct rte_flow_error pointer to
> > > rte_flow_validate(), or a separate function (rte_flow_analyze()?),
> > > since processing this may be quite expensive and applications may
> > > not care about the exact reason.
> > Agree the processing may be too expensive. Maybe we can say it's optional to
> return error details. And that's a good question that what APP should do if
> creating the rule fails. I believe normally it will choose handle the rule by itself.
> But I think it's not bad to feedback more. Or even the APP want to adjust the
> rules, it cannot be an option for lack of info.
>
> All right then, I'll add it to the specification.
>
> int
> rte_flow_validate(uint8_t port_id,
> const struct rte_flow_pattern *pattern,
> const struct rte_flow_actions *actions,
> struct rte_flow_error *error);
>
> With error possibly NULL if the application does not care. Is it fine for you?
Yes, it looks good to me. Thanks for that :)
>
> [...]
> > > > > > > - PMDs, not applications, are responsible for maintaining flow rules
> > > > > > > configuration when stopping and restarting a port or performing
> other
> > > > > > > actions which may affect them. They can only be destroyed explicitly.
> > > > > > Don’t understand " They can only be destroyed explicitly."
> > > > >
> > > > > This part says that as long as an application has not called
> > > > > rte_flow_destroy() on a flow rule, it never disappears, whatever
> > > > > happens to the port (stopped, restarted). The application is not
> > > > > responsible for re-creating rules after that.
> > > > >
> > > > > Note that according to the specification, this may translate to
> > > > > not being able to stop a port as long as a flow rule is present,
> > > > > depending on how nice the PMD intends to be with applications.
> > > > > Implementation can be done in small steps with minimal amount of
> > > > > code on
> > > the PMD side.
> > > > Does it mean PMD should store and maintain all the rules? Why not
> > > > let rte do
> > > that? I think if PMD maintain all the rules, it means every kind of
> > > NIC should have a copy of code for the rules. But if rte do that,
> > > only one copy of code need to be maintained, right?
> > >
> > > I've considered having rules stored in a common format understood at
> > > the RTE level and not specific to each PMD and decided that the
> > > opaque rte_flow pointer was a better choice for the following reasons:
> > >
> > > - Even though flow rules management is done in the control path, processing
> > > must be as fast as possible. Letting PMDs store flow rules using their own
> > > internal representation gives them the chance to achieve better
> > > performance.
> > Not quite understand. I think we're talking about maintain the rules by SW. I
> don’t think there's something need to be optimized according to specific NICs. If
> we need to optimize the code, I think we need to consider the CPU, OS ... and
> some common means. I'm wrong?
>
> Perhaps we were talking about different things, here I was only explaining why
> rte_flow (the result of creating a flow rule) should be opaque and fully managed
> by the PMD. More on the SW side of things below.
>
> > > - An opaque context managed by PMDs would probably have to be stored
> > > somewhere as well anyway.
> > >
> > > - PMDs may not need to allocate/store anything at all if they exclusively
> > > rely on HW state for everything. In my opinion, the generic API has enough
> > > constraints for this to work and maintain consistency between flow
> > > rules. Note this is currently how most PMDs implement FDIR and other
> > > filter types.
> > Yes, the rules are stored by HW. But considering stop/start the device, the
> rules in HW will lose. we have to store the rules by SW and re-program them
> when restarting the device.
>
> Assume a HW capable of keeping flow rules programmed even during a
> stop/start cycle (e.g. mlx4/mlx5 may be able to do it from DPDK point of view),
> don't you think it is more efficient to standardize on this behavior and let PMDs
> restore flow rules for HW that do not support it regardless of whether it would
> be done by RTE or the application (SW)?
Didn’t know that. As some NICs have already had the ability to keep the rules during a stop/start cycle, maybe it could be a trend :)
>
> > And in existing code, we store the filters by SW at least on Intel NICs. But I
> think we cannot reuse them, because considering the priority and which
> category of filter should be chosen, I think we need a whole new table for
> generic API. I think it’s what's designed now, right?
>
> So I understand you'd want RTE to help your PMD keep track of the flow rules it
> created?
Yes. But as you said before, it’s not a good idea for mlx4/mlx5, because their HW doesn't need SW to re-program the rules after stopping/starting. If we make it a common mechanism, it just wastes time for mlx4/mlx5.
>
> Nothing wrong with that, all I'm saying is that it should be entirely optional. RTE
> should not automatically maintain a list. PMDs have to call RTE helpers if they
> need help to maintain a context. These helpers are not defined in this API yet
> because it is difficult to know what will be useful in advance.
>
> > > - RTE can (and will) provide helpers to avoid most of the code redundancy,
> > > PMDs are free to use them or manage everything by themselves.
> > >
> > > - Given that the opaque rte_flow pointer associated with a flow rule is to
> > > be stored by the application, PMDs do not even have to keep references to
> > > them.
> > Don’t understand. More details?
>
> In an application:
>
> rte_flow *foo = rte_flow_create(...);
>
> In the above example, foo cannot be dereferenced by the application nor RTE,
> only the PMD is aware of its contents. This object can only be used with
> rte_flow*() functions.
>
> PMDs are thus free to make this object grow as needed when adding internal
> features without breaking any kind of public API/ABI.
>
> What I meant is, given that the application is supposed to store foo somewhere
> in order to destroy it later, the PMD does not have to keep track of that pointer
> assuming it does not need to access it later on its own for some reason.
>
> > > - The flow rules format described in this specification (pattern / actions)
> > > will be used by applications directly, and will be free to arrange them in
> > > lists, trees or in any other way if they need to keep flow specifications
> > > around for further processing.
> > Who will create the lists, trees or something else? According to previous
> discussion, I think the APP will program the rules one by one. So if APP organize
> the rules to lists, trees..., PMD doesn’t know that.
> > And you said " Given that the opaque rte_flow pointer associated with a flow
> rule is to be stored by the application ". I'm lost here.
>
> I guess that's because we're discussing two different things, flow rule
> specifications and flow rule objects. Let me sum it up:
>
> - Flow rule specifications are the patterns/actions combinations provided by
> applications to rte_flow_create(). Applications can store those as needed
> and organize them as they wish (hash, tree, list). Neither PMDs nor RTE
> will do it for them.
>
> - Flow rule objects (struct rte_flow *) are generated when a flow rule is
> created. Applications must keep these around if they want to manipulate
> them later (i.e. destroy or query existing rules).
Thanks for this clarification. So the specifications can be different with objects, right? The specifications are what the APP wants, the objects are what the APP really gets. As rte_flow_create can fail. Right?
>
> Then PMDs *may* need to keep and arrange flow rule objects internally for
> management purposes. Could be because HW requires it, detecting conflicting
> rules, managing priorities and so on. Possible reasons are not described in this
> API because these are thought as PMD-specific needs.
Got it.
>
> > > > When the port is stopped and restarted, rte can reconfigure the
> > > > rules. Is the
> > > concern that PMD may adjust the sequence of the rules according to
> > > the priority, so every NIC has a different list of rules? But PMD
> > > can adjust them again when rte reconfiguring the rules.
> > >
> > > What about PMDs able to stop and restart ports without destroying
> > > their own flow rules? If we assume flow rules must be destroyed when
> > > stopping a port, these PMDs are needlessly penalized with slower
> > > stop/start cycles. Think about it assuming thousands of flow rules.
> > I believe the rules maintained by SW should not be destroyed, because they're
> used to be re-programed when the device starts again.
>
> Do we agree that applications should not care? Flow rules configured before
> stopping a port must still be there after restarting it.
Yes, agree.
>
> What we seem to not agree about is that you think RTE should be responsible
> for restoring flow rules of devices that lose them when stopped. I think doing so
> is unfair to devices for which it is not the case and not really nice to applications,
> so my opinion is that the PMD is responsible for restoring flow rules however it
> wants. It is free to use RTE helpers to keep their track, as long as it's all managed
> internally.
What I think is RTE can store the flow rules and recreate them after restarting, in the function like rte_dev_start, so APP knows nothing about it. But according to the discussing above, I think the design doesn't support it, right?
RTE doesn't store the flow rules objects and event it stores them, there's no way designed to re-program the objects. And also considering some HW doesn't need to be re-programed. I think it's OK to let PMD maintain the rules as the re-programing is a NIC specific requirement.
>
> > > Thus from an application point of view, whatever happens when
> > > stopping and restarting a port should not matter. If a flow rule was
> > > present before, it must still be present afterwards. If the PMD had
> > > to destroy flow rules and re-create them, it does not actually matter if they
> differ slightly at the HW level, as long as:
> > >
> > > - Existing opaque flow rule pointers (rte_flow) are still valid to the PMD
> > > and refer to the same rules.
> > >
> > > - The overall behavior of all rules is the same.
> > >
> > > The list of rules you think of (patterns / actions) is maintained by
> > > applications (not RTE), and only if they need them. RTE would needlessly
> duplicate this.
> > As said before, need more details to understand this. Maybe an example
> > is better :)
>
> The generic format both RTE and applications might understand is the one
> described in this API (struct rte_flow_pattern and struct rte_flow_actions).
>
> If we wanted RTE to maintain some sort of per-port state for flow rule
> specifications, it would have to be a copy of these structures arranged somehow
> (list or something else).
>
> If we consider that PMDs need to keep a context object associated to a flow
> rule (the opaque struct rte_flow *), then RTE would most likely have to store it
> along with the flow specification.
>
> Such a list may not be useful to applications (list lookups take time), so they
> would implement their own redundant method. They might also require extra
> room to attach some application context to flow rules. A generic list cannot plan
> for it.
>
> Applications know what they want to do with flow rules and are responsible for
> managing them efficiently with RTE out of the way.
>
> I'm not sure if this answered your question, if not, please describe a scenario
> where a RTE-managed list of flow rules would be mandatory.
Got your point and agree :)
>
> --
> Adrien Mazarguil
> 6WIND
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v2] validate_abi: build faster by augmenting make with job count
@ 2016-07-20 19:02 9% ` Neil Horman
0 siblings, 0 replies; 200+ results
From: Neil Horman @ 2016-07-20 19:02 UTC (permalink / raw)
To: dev; +Cc: Neil Horman, Neil Horman, Thomas Monjalon, Mcnamara, John
John Mcnamara and I were discussing enhacing the validate_abi script to build
the dpdk tree faster with multiple jobs. Theres no reason not to do it, so this
implements that requirement. It uses a MAKE_JOBS variable that can be set by
the user to limit the job count. By default the job count is set to the number
of online cpus.
Signed-off-by: Neil Horman <nhorman@tuxdrier.com>
CC: Thomas Monjalon <thomas.monjalon@6wind.com>
CC: "Mcnamara, John" <john.mcnamara@intel.com>
---
Change notes
v2) switch variable to DPDK_MAKE_JOBS
make provision for no existance of lscpu
---
scripts/validate-abi.sh | 15 +++++++++++++--
1 file changed, 13 insertions(+), 2 deletions(-)
diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index c36ad61..feda6c8 100755
--- a/scripts/validate-abi.sh
+++ b/scripts/validate-abi.sh
@@ -97,6 +97,17 @@ fixup_config() {
#trap on ctrl-c to clean up
trap cleanup_and_exit SIGINT
+if [ -z "$DPDK_MAKE_JOBS" ]
+then
+ # This counts the number of cpus on the system
+ if [ -e /usr/bin/lscpu ]
+ then
+ DPDK_MAKE_JOBS=`lscpu -p=cpu | grep -v "#" | wc -l`
+ else
+ DPDK_MAKE_JOBS=1
+ fi
+fi
+
#Save the current branch
CURRENT_BRANCH=`git branch | grep \* | cut -d' ' -f2`
@@ -183,7 +194,7 @@ log "INFO" "Configuring DPDK $TAG1"
make config T=$TARGET O=$TARGET > $VERBOSE 2>&1
log "INFO" "Building DPDK $TAG1. This might take a moment"
-make O=$TARGET > $VERBOSE 2>&1
+make -j$DPDK_MAKE_JOBS O=$TARGET > $VERBOSE 2>&1
if [ $? -ne 0 ]
then
@@ -214,7 +225,7 @@ log "INFO" "Configuring DPDK $TAG2"
make config T=$TARGET O=$TARGET > $VERBOSE 2>&1
log "INFO" "Building DPDK $TAG2. This might take a moment"
-make O=$TARGET > $VERBOSE 2>&1
+make -j$DPDK_MAKE_JOBS O=$TARGET > $VERBOSE 2>&1
if [ $? -ne 0 ]
then
--
2.5.5
^ permalink raw reply [relevance 9%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-20 13:37 4% ` Olivier Matz
2016-07-20 14:01 0% ` Richardson, Bruce
@ 2016-07-20 17:20 0% ` Zoltan Kiss
1 sibling, 0 replies; 200+ results
From: Zoltan Kiss @ 2016-07-20 17:20 UTC (permalink / raw)
To: Olivier Matz, Zoltan Kiss, dev
On 20/07/16 14:37, Olivier Matz wrote:
> Hi,
>
> On 07/20/2016 02:41 PM, Zoltan Kiss wrote:
>>
>>
>> On 19/07/16 17:17, Olivier Matz wrote:
>>> Hi Zoltan,
>>>
>>> On 07/19/2016 05:59 PM, Zoltan Kiss wrote:
>>>>
>>>>
>>>> On 19/07/16 16:37, Olivier Matz wrote:
>>>>> Hi Zoltan,
>>>>>
>>>>> On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
>>>>>> A recent fix brought up an issue about the size of the 'name' fields:
>>>>>>
>>>>>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>>>>>
>>>>>> These relations should be observed:
>>>>>>
>>>>>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>>>>>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE -
>>>>>> strlen(RTE_MEMPOOL_MZ_PREFIX)
>>>>>>
>>>>>> Setting all of them to 32 hides this restriction from the application.
>>>>>> This patch increases the memzone string size to accomodate for these
>>>>>> prefixes, and the same happens with the ring name string. The ABI
>>>>>> needs to
>>>>>> be broken to fix this API issue, this way doesn't break applications
>>>>>> previously not failing due to the truncating bug now fixed.
>>>>>>
>>>>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>>>>>
>>>>> I agree it is a problem for an application because it cannot know what
>>>>> is the maximum name length. On the other hand, breaking the ABI for
>>>>> this
>>>>> looks a bit overkill. Maybe we could reduce RTE_MEMPOOL_NAMESIZE and
>>>>> RTE_RING_NAMESIZE instead of increasing RTE_MEMZONE_NAMESIZE? That way,
>>>>> we could keep the ABI as is.
>>>>
>>>> But that would break the ABI too, wouldn't it? Unless you keep the array
>>>> the same size (32 bytes) by using RTE_MEMZONE_NAMESIZE.
>>>
>>> Yes, that was the idea.
>>>
>>>> And even then, the API breaks anyway. There are applications - I have at
>>>> least some - which use all 32 bytes to store the name. Decrease that
>>>> would cause headache to change the naming scheme, because it's a 30
>>>> character long id, and chopping the last few chars would cause name
>>>> collisions and annoying bugs.
>>>
>>> Before my patch (85cf0079), long names were silently truncated when
>>> mempool created its ring and/or memzones. Now, it returns an error.
>>
>> With 16.04 an application could operate as expected if the first 26
>> character were unique. Your patch revealed the problem that characters
>> after these were left out of the name. Now applications fail where this
>> never been a bug because their naming scheme guarantees the uniqueness
>> on the first 26 chars (or makes it very unlikely)
>> Where the first 26 is not unique, it failed earlier too, because at
>> memzone creation it checks for duplicate names.
>
> Yes, I understand that there is a behavior change for applications using
> names larger than 26 between 16.04 and 16.07. I also understand that
> there is no way for an application to know what is the maximum usable
> size for a mempool or a ring.
>
>
>>> I'm not getting why changing the struct to something like below would
>>> break the API, since it would already return an error today.
>>>
>>> #define RTE_MEMPOOL_NAMESIZE \
>>
>> Wait, this would mean applications need to recompile to use the smaller
>> value. AFAIK that's an ABI break too, right? At the moment I don't see a
>> way to fix this without breaking the ABI
>
> With this modification, if you don't recompile the application, its
> behavior will still be the same as today -> it will return ENAMETOOLONG.
> If you recompile it, the application will be aware of the maximum
> length. To me, it seems to be a acceptable compromise for this release.
>
> The patch you're proposing also changes the ABI of librte_ring and
> librte_eal, which cannot be accepted for the release.
Ok, I've sent a new version with this approach.
>
>
>>
>>> (RTE_MEMZONE_NAMESIZE - sizeof(pool_prefix) - sizeof(ring prefix))
>>> struct rte_mempool {
>>> union {
>>> char name[RTE_MEMPOOL_NAMESIZE];
>>> char pad[32];
>>> };
>>> ...
>>> }
>>>
>>> Anyway, it may not be the proper solution since it supposes that a
>>> mempool includes a ring based on a memzone, which is not always true now
>>> with mempool handlers.
>>
>> Oh, as we dug deeper it gets better!
>> Indeed, we don't necessarily have this ring + memzone pair underneath,
>> but the user is not aware of that, and I think we should keep it that
>> way. It should only care that the string passed shouldn't be bigger than
>> a certain amount.
>
> Yes. What I'm just saying here is that it's not a good solution to write
> in the #define that "a mempool is based on a ring + a memzone", because
> if some someone adds a new mempool handler replacing the ring, and also
> creating a memzone prefixed by something larger than "rg_", we will have
> to break the ABI again.
If someone adds a new handler, (s)he needs to keep in mind what's the
max size for pool name, and any derived object using that name as a base
should check if it fits.
>
>
>> Also, even though we don't necessarily have the ring, we still reserve
>> memzone's in rte_mempool_populate_default(). And their name has a 3
>> letter prefix, and a "_%d" postfix, where the %d could be as much as
>> RTE_MAX_MEMZONE in worst case (2560 by default) So actually:
>>
>> RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
>> strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen("_2560")
>>
>>
>> As a side note, there is another bug around here: rte_ring_create()
>> doesn't check for name duplications. However the user of the library can
>> lookup based on the name with rte_ring_lookup(), and it will return the
>> first ring with that name
>
> The name uniqueness is checked by rte_memzone_reserve().
>
>
>>>>> It would even be better to get rid of this static char[] for the
>>>>> structure names and replace it by an allocated const char *. I didn't
>>>>> check it's feasible for memzones. What do you think?
>>>>
>>>> It would work too, but I don't think it would help a lot. We would still
>>>> need max sizes for the names. Storing them somewhere else won't help us
>>>> in this problem.
>>>
>>> Why should we have a maximum length for the names?
>>
>> What happens if an application loads DPDK and create a mempool with a
>> name string 2 million characters long? Maybe nothing we should worry
>> about, but in general I think unlimited length function parameters are
>> problematic at the very least. The length should be passed at least
>> (which also creates a max due to the size of the param). But I think it
>> would be just easier to have these maximums set, observing the above
>> constrains.
>
> I think have a maximum name length brings more problems than not having
> it, especially ABI problems.
>
>
> Regards,
> Olivier
>
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v2] mempool: adjust name string size in related data types
2016-07-19 14:37 3% ` [dpdk-dev] [PATCH] mempool: adjust name string size in related data types Zoltan Kiss
2016-07-19 15:37 4% ` Olivier Matz
@ 2016-07-20 17:16 12% ` Zoltan Kiss
2016-07-21 13:40 0% ` Olivier Matz
1 sibling, 1 reply; 200+ results
From: Zoltan Kiss @ 2016-07-20 17:16 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, Bruce Richardson
A recent patch brought up an issue about the size of the 'name' fields:
85cf0079 mem: avoid memzone/mempool/ring name truncation
These relations should be observed:
1. Each ring creates a memzone with a prefixed name:
RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
2. There are some mempool handlers which create a ring with a prefixed
name:
RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE - strlen(RTE_MEMPOOL_MZ_PREFIX)
3. A mempool can create up to RTE_MAX_MEMZONE pre and postfixed memzones:
sprintf(postfix, "_%d", RTE_MAX_MEMZONE)
RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen(postfix)
Setting all of them to 32 hides this restriction from the application.
This patch decreases the mempool and ring string size to accommodate for
these prefixes, but it doesn't apply the 3rd constraint. Applications
relying on these constants need to be recompiled, otherwise they'll run
into ENAMETOOLONG issues.
The size of the arrays are kept 32 for ABI compatibility, it can be
decreased next time the ABI changes.
Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
---
Notes:
v2: keep arrays 32 bytes and decrease the max sizes to maintain ABI
compatibility
lib/librte_mempool/rte_mempool.h | 11 +++++++++--
lib/librte_ring/rte_ring.h | 12 ++++++++++--
2 files changed, 19 insertions(+), 4 deletions(-)
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 4a8fbb1..059ad9e 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -123,7 +123,9 @@ struct rte_mempool_objsz {
/**< Total size of an object (header + elt + trailer). */
};
-#define RTE_MEMPOOL_NAMESIZE 32 /**< Maximum length of a memory pool. */
+/**< Maximum length of a memory pool's name. */
+#define RTE_MEMPOOL_NAMESIZE (RTE_RING_NAMESIZE - \
+ sizeof(RTE_MEMPOOL_MZ_PREFIX) + 1)
#define RTE_MEMPOOL_MZ_PREFIX "MP_"
/* "MP_<name>" */
@@ -208,7 +210,12 @@ struct rte_mempool_memhdr {
* The RTE mempool structure.
*/
struct rte_mempool {
- char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
+ /*
+ * Note: this field kept the RTE_MEMZONE_NAMESIZE size due to ABI
+ * compatibility requirements, it could be changed to
+ * RTE_MEMPOOL_NAMESIZE next time the ABI changes
+ */
+ char name[RTE_MEMZONE_NAMESIZE]; /**< Name of mempool. */
union {
void *pool_data; /**< Ring or pool to store objects. */
uint64_t pool_id; /**< External mempool identifier. */
diff --git a/lib/librte_ring/rte_ring.h b/lib/librte_ring/rte_ring.h
index eb45e41..0e22e69 100644
--- a/lib/librte_ring/rte_ring.h
+++ b/lib/librte_ring/rte_ring.h
@@ -100,6 +100,7 @@ extern "C" {
#include <rte_lcore.h>
#include <rte_atomic.h>
#include <rte_branch_prediction.h>
+#include <rte_memzone.h>
#define RTE_TAILQ_RING_NAME "RTE_RING"
@@ -126,8 +127,10 @@ struct rte_ring_debug_stats {
} __rte_cache_aligned;
#endif
-#define RTE_RING_NAMESIZE 32 /**< The maximum length of a ring name. */
#define RTE_RING_MZ_PREFIX "RG_"
+/**< The maximum length of a ring name. */
+#define RTE_RING_NAMESIZE (RTE_MEMZONE_NAMESIZE - \
+ sizeof(RTE_RING_MZ_PREFIX) + 1)
#ifndef RTE_RING_PAUSE_REP_COUNT
#define RTE_RING_PAUSE_REP_COUNT 0 /**< Yield after pause num of times, no yield
@@ -147,7 +150,12 @@ struct rte_memzone; /* forward declaration, so as not to require memzone.h */
* a problem.
*/
struct rte_ring {
- char name[RTE_RING_NAMESIZE]; /**< Name of the ring. */
+ /*
+ * Note: this field kept the RTE_MEMZONE_NAMESIZE size due to ABI
+ * compatibility requirements, it could be changed to RTE_RING_NAMESIZE
+ * next time the ABI changes
+ */
+ char name[RTE_MEMZONE_NAMESIZE]; /**< Name of the ring. */
int flags; /**< Flags supplied at creation. */
const struct rte_memzone *memzone;
/**< Memzone, if any, containing the rte_ring */
--
1.9.1
^ permalink raw reply [relevance 12%]
* [dpdk-dev] [PATCH] validate_abi: build faster by augmenting make with job count
@ 2016-07-20 17:09 9% Neil Horman
0 siblings, 0 replies; 200+ results
From: Neil Horman @ 2016-07-20 17:09 UTC (permalink / raw)
To: dev; +Cc: Neil Horman, Neil Horman, Thomas Monjalon, Mcnamara, John
From: Neil Horman <nhorman@redhat.com>
John Mcnamara and I were discussing enhacing the validate_abi script to build
the dpdk tree faster with multiple jobs. Theres no reason not to do it, so this
implements that requirement. It uses a MAKE_JOBS variable that can be set by
the user to limit the job count. By default the job count is set to the number
of online cpus.
Signed-off-by: Neil Horman <nhorman@tuxdrier.com>
CC: Thomas Monjalon <thomas.monjalon@6wind.com>
CC: "Mcnamara, John" <john.mcnamara@intel.com>
---
scripts/validate-abi.sh | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index c36ad61..1c9627b 100755
--- a/scripts/validate-abi.sh
+++ b/scripts/validate-abi.sh
@@ -97,6 +97,12 @@ fixup_config() {
#trap on ctrl-c to clean up
trap cleanup_and_exit SIGINT
+if [ -z "$MAKE_JOBS" ]
+then
+ # This counts the number of cpus on the system
+ MAKE_JOBS=`lscpu -p=cpu | grep -v "#" | wc -l`
+fi
+
#Save the current branch
CURRENT_BRANCH=`git branch | grep \* | cut -d' ' -f2`
@@ -183,7 +189,7 @@ log "INFO" "Configuring DPDK $TAG1"
make config T=$TARGET O=$TARGET > $VERBOSE 2>&1
log "INFO" "Building DPDK $TAG1. This might take a moment"
-make O=$TARGET > $VERBOSE 2>&1
+make -j$MAKE_JOBS O=$TARGET > $VERBOSE 2>&1
if [ $? -ne 0 ]
then
@@ -214,7 +220,7 @@ log "INFO" "Configuring DPDK $TAG2"
make config T=$TARGET O=$TARGET > $VERBOSE 2>&1
log "INFO" "Building DPDK $TAG2. This might take a moment"
-make O=$TARGET > $VERBOSE 2>&1
+make -j$MAKE_JOBS O=$TARGET > $VERBOSE 2>&1
if [ $? -ne 0 ]
then
--
2.5.5
^ permalink raw reply [relevance 9%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure
2016-07-20 15:22 7% ` Thomas Monjalon
@ 2016-07-20 15:42 4% ` Kulasek, TomaszX
0 siblings, 0 replies; 200+ results
From: Kulasek, TomaszX @ 2016-07-20 15:42 UTC (permalink / raw)
To: Thomas Monjalon, Ananyev, Konstantin; +Cc: dev
> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Wednesday, July 20, 2016 17:22
> To: Ananyev, Konstantin <konstantin.ananyev@intel.com>
> Cc: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev
> structure
>
> 2016-07-20 15:13, Ananyev, Konstantin:
> > Hi Thomas,
> >
> > > Hi,
> > >
> > > This patch announces an interesting change in the DPDK design.
> > >
> > > 2016-07-20 16:24, Tomasz Kulasek:
> > > > This is an ABI deprecation notice for DPDK 16.11 in librte_ether
> > > > about changes in rte_eth_dev and rte_eth_desc_lim structures.
> > > >
> > > > In 16.11, we plan to introduce rte_eth_tx_prep() function to do
> > > > necessary preparations of packet burst to be safely transmitted on
> > > > device for desired HW offloads (set/reset checksum field according
> > > > to the hardware requirements) and check HW constraints (number of
> > > > segments per packet, etc).
> > > >
> > > > While the limitations and requirements may differ for devices, it
> > > > requires to extend rte_eth_dev structure with new function pointer
> > > > "tx_pkt_prep" which can be implemented in the driver to prepare
> > > > and verify packets, in devices specific way, before burst, what
> > > > should to prevent application to send malformed packets.
> > > >
> > > > Also new fields will be introduced in rte_eth_desc_lim: nb_seg_max
> > > > and nb_mtu_seg_max, providing an information about max segments in
> > > > TSO and non TSO packets acceptable by device.
> > >
> > > We cannot acknowledge such notice without a prior design discussion.
> > > Please explain why you plan to work on this change and give a draft of
> the new structures (a RFC patch would be ideal).
> >
> > I think it is not really a deprecation note, but announce ABI change for
> rte_ethdev.h structures.
>
> An ABI break requires a deprecation notice. So it is :)
>
> > The plan is to implement what was proposed & discussed the following
> thread:
> > http://dpdk.org/ml/archives/dev/2015-September/023603.html
>
> Please could you summarize it here?
Hi Thomas,
The implementation of rte_eth_tx_prep() will be similar to the rte_eth_tx_burst(), passing same arguments to the driver, so packets can be checked and modified before real burst will be done.
The API for new function will be implemented in the fallowed way:
+/**
+ * Process a burst of output packets on a transmit queue of an Ethernet device.
+ *
+ * The rte_eth_tx_prep() function is invoked to prepare output packets to be
+ * transmitted on the output queue *queue_id* of the Ethernet device designated
+ * by its *port_id*.
+ * The *nb_pkts* parameter is the number of packets to be prepared which are
+ * supplied in the *tx_pkts* array of *rte_mbuf* structures, each of them
+ * allocated from a pool created with rte_pktmbuf_pool_create().
+ * For each packet to send, the rte_eth_tx_prep() function performs
+ * the following operations:
+ *
+ * - Check if packet meets devices requirements for tx offloads.
+ *
+ * - Check limitations about number of segments.
+ *
+ * - Check additional requirements when debug is enabled.
+ *
+ * - Update and/or reset required checksums when tx offload is set for packet.
+ *
+ * The rte_eth_tx_prep() function returns the number of packets ready it
+ * actually sent. A return value equal to *nb_pkts* means that all packets
+ * are valid and ready to be sent.
+ *
+ * @param port_id
+ * The port identifier of the Ethernet device.
+ * @param queue_id
+ * The index of the transmit queue through which output packets must be
+ * sent.
+ * The value must be in the range [0, nb_tx_queue - 1] previously supplied
+ * to rte_eth_dev_configure().
+ * @param tx_pkts
+ * The address of an array of *nb_pkts* pointers to *rte_mbuf* structures
+ * which contain the output packets.
+ * @param nb_pkts
+ * The maximum number of packets to process.
+ * @return
+ * The number of packets correct and ready to be sent. The return value can be
+ * less than the value of the *tx_pkts* parameter when some packet doesn't
+ * meet devices requirements with rte_errno set appropriately.
+ */
+static inline uint16_t
+rte_eth_tx_prep(uint8_t port_id, uint16_t queue_id, struct rte_mbuf **tx_pkts,
+ uint16_t nb_pkts)
+{
+ struct rte_eth_dev *dev = &rte_eth_devices[port_id];
+
+ if (!dev->tx_pkt_prep) {
+ rte_errno = -ENOTSUP;
+ return 0;
+ }
+
+#ifdef RTE_LIBRTE_ETHDEV_DEBUG
+ if (queue_id >= dev->data->nb_tx_queues) {
+ RTE_PMD_DEBUG_TRACE("Invalid TX queue_id=%d\n", queue_id);
+ rte_errno = -EINVAL;
+ return 0;
+ }
+#endif
+
+ return (*dev->tx_pkt_prep)(dev->data->tx_queues[queue_id], tx_pkts, nb_pkts);
+}
Tomasz
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure
2016-07-20 15:13 7% ` Ananyev, Konstantin
@ 2016-07-20 15:22 7% ` Thomas Monjalon
2016-07-20 15:42 4% ` Kulasek, TomaszX
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-20 15:22 UTC (permalink / raw)
To: Ananyev, Konstantin; +Cc: Kulasek, TomaszX, dev
2016-07-20 15:13, Ananyev, Konstantin:
> Hi Thomas,
>
> > Hi,
> >
> > This patch announces an interesting change in the DPDK design.
> >
> > 2016-07-20 16:24, Tomasz Kulasek:
> > > This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
> > > changes in rte_eth_dev and rte_eth_desc_lim structures.
> > >
> > > In 16.11, we plan to introduce rte_eth_tx_prep() function to do
> > > necessary preparations of packet burst to be safely transmitted on
> > > device for desired HW offloads (set/reset checksum field according to
> > > the hardware requirements) and check HW constraints (number of
> > > segments per packet, etc).
> > >
> > > While the limitations and requirements may differ for devices, it
> > > requires to extend rte_eth_dev structure with new function pointer
> > > "tx_pkt_prep" which can be implemented in the driver to prepare and
> > > verify packets, in devices specific way, before burst, what should to
> > > prevent application to send malformed packets.
> > >
> > > Also new fields will be introduced in rte_eth_desc_lim: nb_seg_max and
> > > nb_mtu_seg_max, providing an information about max segments in TSO and
> > > non TSO packets acceptable by device.
> >
> > We cannot acknowledge such notice without a prior design discussion.
> > Please explain why you plan to work on this change and give a draft of the new structures (a RFC patch would be ideal).
>
> I think it is not really a deprecation note, but announce ABI change for rte_ethdev.h structures.
An ABI break requires a deprecation notice. So it is :)
> The plan is to implement what was proposed & discussed the following thread:
> http://dpdk.org/ml/archives/dev/2015-September/023603.html
Please could you summarize it here?
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure
2016-07-20 15:01 4% ` Thomas Monjalon
@ 2016-07-20 15:13 7% ` Ananyev, Konstantin
2016-07-20 15:22 7% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-07-20 15:13 UTC (permalink / raw)
To: Thomas Monjalon, Kulasek, TomaszX; +Cc: dev
Hi Thomas,
> Hi,
>
> This patch announces an interesting change in the DPDK design.
>
> 2016-07-20 16:24, Tomasz Kulasek:
> > This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
> > changes in rte_eth_dev and rte_eth_desc_lim structures.
> >
> > In 16.11, we plan to introduce rte_eth_tx_prep() function to do
> > necessary preparations of packet burst to be safely transmitted on
> > device for desired HW offloads (set/reset checksum field according to
> > the hardware requirements) and check HW constraints (number of
> > segments per packet, etc).
> >
> > While the limitations and requirements may differ for devices, it
> > requires to extend rte_eth_dev structure with new function pointer
> > "tx_pkt_prep" which can be implemented in the driver to prepare and
> > verify packets, in devices specific way, before burst, what should to
> > prevent application to send malformed packets.
> >
> > Also new fields will be introduced in rte_eth_desc_lim: nb_seg_max and
> > nb_mtu_seg_max, providing an information about max segments in TSO and
> > non TSO packets acceptable by device.
>
> We cannot acknowledge such notice without a prior design discussion.
> Please explain why you plan to work on this change and give a draft of the new structures (a RFC patch would be ideal).
I think it is not really a deprecation note, but announce ABI change for rte_ethdev.h structures.
The plan is to implement what was proposed & discussed the following thread:
http://dpdk.org/ml/archives/dev/2015-September/023603.html
Konstantin
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure
2016-07-20 14:24 13% [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure Tomasz Kulasek
@ 2016-07-20 15:01 4% ` Thomas Monjalon
2016-07-20 15:13 7% ` Ananyev, Konstantin
2016-07-21 15:24 11% ` [dpdk-dev] [PATCH v2] " Tomasz Kulasek
2016-07-31 7:46 4% ` [dpdk-dev] [PATCH] " Vlad Zolotarov
2 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-20 15:01 UTC (permalink / raw)
To: Tomasz Kulasek; +Cc: dev
Hi,
This patch announces an interesting change in the DPDK design.
2016-07-20 16:24, Tomasz Kulasek:
> This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
> changes in rte_eth_dev and rte_eth_desc_lim structures.
>
> In 16.11, we plan to introduce rte_eth_tx_prep() function to do
> necessary preparations of packet burst to be safely transmitted on
> device for desired HW offloads (set/reset checksum field according to
> the hardware requirements) and check HW constraints (number of segments
> per packet, etc).
>
> While the limitations and requirements may differ for devices, it
> requires to extend rte_eth_dev structure with new function pointer
> "tx_pkt_prep" which can be implemented in the driver to prepare and
> verify packets, in devices specific way, before burst, what should to
> prevent application to send malformed packets.
>
> Also new fields will be introduced in rte_eth_desc_lim: nb_seg_max and
> nb_mtu_seg_max, providing an information about max segments in TSO and
> non TSO packets acceptable by device.
We cannot acknowledge such notice without a prior design discussion.
Please explain why you plan to work on this change and give a draft of
the new structures (a RFC patch would be ideal).
Thanks
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure
@ 2016-07-20 14:24 13% Tomasz Kulasek
2016-07-20 15:01 4% ` Thomas Monjalon
` (2 more replies)
0 siblings, 3 replies; 200+ results
From: Tomasz Kulasek @ 2016-07-20 14:24 UTC (permalink / raw)
To: dev
This is an ABI deprecation notice for DPDK 16.11 in librte_ether about
changes in rte_eth_dev and rte_eth_desc_lim structures.
In 16.11, we plan to introduce rte_eth_tx_prep() function to do
necessary preparations of packet burst to be safely transmitted on
device for desired HW offloads (set/reset checksum field according to
the hardware requirements) and check HW constraints (number of segments
per packet, etc).
While the limitations and requirements may differ for devices, it
requires to extend rte_eth_dev structure with new function pointer
"tx_pkt_prep" which can be implemented in the driver to prepare and
verify packets, in devices specific way, before burst, what should to
prevent application to send malformed packets.
Also new fields will be introduced in rte_eth_desc_lim: nb_seg_max and
nb_mtu_seg_max, providing an information about max segments in TSO and
non TSO packets acceptable by device.
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
---
doc/guides/rel_notes/deprecation.rst | 7 +++++++
1 file changed, 7 insertions(+)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f502f86..485aacb 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -41,3 +41,10 @@ Deprecation Notices
* The mempool functions for single/multi producer/consumer are deprecated and
will be removed in 16.11.
It is replaced by rte_mempool_generic_get/put functions.
+
+* In 16.11 ABI changes are plained: the ``rte_eth_dev`` structure will be
+ extended with new function pointer ``tx_pkt_prep`` allowing verification
+ and processing of packet burst to meet HW specific requirements before
+ transmit. Also new fields will be added to the ``rte_eth_desc_lim`` structure:
+ ``nb_seg_max`` and ``nb_mtu_seg_max`` provideing information about number of
+ segments limit to be transmitted by device for TSO/non-TSO packets.
--
1.7.9.5
^ permalink raw reply [relevance 13%]
* [dpdk-dev] [PATCH] unify tools naming
@ 2016-07-20 14:24 3% Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-20 14:24 UTC (permalink / raw)
To: dev
The following tools may be installed system-wise.
It may be cleaner and more convenient to find them with the same
dpdk- prefix (especially for autocompletion).
Moreover, the script dpdk_nic_bind.py deserves a new name because it is
not restricted to NICs and can be used for e.g. crypto.
These files are renamed:
pmdinfogen -> dpdk-pmdinfogen
pmdinfo.py -> dpdk-pmdinfo.py
dpdk_pdump -> dpdk-pdump
dpdk_proc_info -> dpdk-procinfo
dpdk_nic_bind.py -> dpdk-devbind.py
setup.sh -> dpdk-setup.sh
The tools pmdinfogen, pmdinfo.py and dpdk_pdump are new in 16.07.
The scripts dpdk_nic_bind.py and setup.sh may have been used with
previous releases by end users. That's why a symbolic link still
provide the old name in the installed tools directory.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
It would be good to have this rename in 16.07 before that pmdinfo.py
and dpdk_pdump are widely used.
A possible addition to this patch could be renaming the test apps:
test -> dpdk-test
testacl -> dpdk-testacl
testpipeline -> dpdk-testpipeline
testpmd -> dpdk-testpmd
---
MAINTAINERS | 2 +-
app/pdump/Makefile | 2 +-
app/proc_info/Makefile | 2 +-
buildtools/pmdinfogen/Makefile | 2 +-
doc/guides/faq/faq.rst | 2 +-
doc/guides/linux_gsg/build_dpdk.rst | 14 +++++++-------
doc/guides/linux_gsg/nic_perf_intel_platform.rst | 6 +++---
doc/guides/linux_gsg/quick_start.rst | 12 ++++++------
doc/guides/nics/bnx2x.rst | 4 ++--
doc/guides/nics/cxgbe.rst | 4 ++--
doc/guides/nics/ena.rst | 2 +-
doc/guides/nics/enic.rst | 6 +++---
doc/guides/nics/i40e.rst | 4 ++--
doc/guides/nics/intel_vf.rst | 8 ++++----
doc/guides/nics/nfp.rst | 8 ++++----
doc/guides/nics/qede.rst | 2 +-
doc/guides/nics/thunderx.rst | 16 ++++++++--------
doc/guides/nics/virtio.rst | 2 +-
doc/guides/prog_guide/dev_kit_build_system.rst | 14 +++++++-------
doc/guides/rel_notes/release_16_07.rst | 3 +++
doc/guides/sample_app_ug/pdump.rst | 14 +++++++-------
doc/guides/sample_app_ug/proc_info.rst | 8 ++++----
doc/guides/testpmd_app_ug/testpmd_funcs.rst | 10 +++++-----
doc/guides/xen/pkt_switch.rst | 2 +-
lib/librte_eal/common/eal_common_options.c | 4 ++--
mk/internal/rte.compile-pre.mk | 2 +-
mk/rte.sdkinstall.mk | 16 ++++++++++------
mk/rte.sdktest.mk | 4 ++--
tools/{dpdk_nic_bind.py => dpdk-devbind.py} | 0
tools/{pmdinfo.py => dpdk-pmdinfo.py} | 4 +---
tools/{setup.sh => dpdk-setup.sh} | 24 ++++++++++++------------
31 files changed, 104 insertions(+), 99 deletions(-)
rename tools/{dpdk_nic_bind.py => dpdk-devbind.py} (100%)
rename tools/{pmdinfo.py => dpdk-pmdinfo.py} (99%)
rename tools/{setup.sh => dpdk-setup.sh} (95%)
diff --git a/MAINTAINERS b/MAINTAINERS
index 2996b09..3bfcc9f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -70,7 +70,7 @@ F: scripts/validate-abi.sh
Driver information
F: buildtools/pmdinfogen/
-F: tools/pmdinfo.py
+F: tools/dpdk-pmdinfo.py
Environment Abstraction Layer
diff --git a/app/pdump/Makefile b/app/pdump/Makefile
index d85bb08..536198f 100644
--- a/app/pdump/Makefile
+++ b/app/pdump/Makefile
@@ -33,7 +33,7 @@ include $(RTE_SDK)/mk/rte.vars.mk
ifeq ($(CONFIG_RTE_LIBRTE_PDUMP),y)
-APP = dpdk_pdump
+APP = dpdk-pdump
CFLAGS += $(WERROR_FLAGS)
diff --git a/app/proc_info/Makefile b/app/proc_info/Makefile
index 33e058e..e051e03 100644
--- a/app/proc_info/Makefile
+++ b/app/proc_info/Makefile
@@ -31,7 +31,7 @@
include $(RTE_SDK)/mk/rte.vars.mk
-APP = dpdk_proc_info
+APP = dpdk-procinfo
CFLAGS += $(WERROR_FLAGS)
diff --git a/buildtools/pmdinfogen/Makefile b/buildtools/pmdinfogen/Makefile
index 3885d3b..bd8f900 100644
--- a/buildtools/pmdinfogen/Makefile
+++ b/buildtools/pmdinfogen/Makefile
@@ -34,7 +34,7 @@ include $(RTE_SDK)/mk/rte.vars.mk
#
# library name
#
-HOSTAPP = pmdinfogen
+HOSTAPP = dpdk-pmdinfogen
#
# all sources are stored in SRCS-y
diff --git a/doc/guides/faq/faq.rst b/doc/guides/faq/faq.rst
index 3228b92..8d1ea6c 100644
--- a/doc/guides/faq/faq.rst
+++ b/doc/guides/faq/faq.rst
@@ -50,7 +50,7 @@ When you stop and restart the test application, it looks to see if the pages are
If you look in the directory, you will see ``n`` number of 2M pages files. If you specified 1024, you will see 1024 page files.
These are then placed in memory segments to get contiguous memory.
-If you need to change the number of pages, it is easier to first remove the pages. The tools/setup.sh script provides an option to do this.
+If you need to change the number of pages, it is easier to first remove the pages. The tools/dpdk-setup.sh script provides an option to do this.
See the "Quick Start Setup Script" section in the :ref:`DPDK Getting Started Guide <linux_gsg>` for more information.
diff --git a/doc/guides/linux_gsg/build_dpdk.rst b/doc/guides/linux_gsg/build_dpdk.rst
index fb2c481..f8007b3 100644
--- a/doc/guides/linux_gsg/build_dpdk.rst
+++ b/doc/guides/linux_gsg/build_dpdk.rst
@@ -198,7 +198,7 @@ however please consult your distributions documentation to make sure that is the
Also, to use VFIO, both kernel and BIOS must support and be configured to use IO virtualization (such as Intel® VT-d).
For proper operation of VFIO when running DPDK applications as a non-privileged user, correct permissions should also be set up.
-This can be done by using the DPDK setup script (called setup.sh and located in the tools directory).
+This can be done by using the DPDK setup script (called dpdk-setup.sh and located in the tools directory).
.. _linux_gsg_binding_kernel:
@@ -224,7 +224,7 @@ and to bind and unbind those ports from the different kernel modules, including
The following are some examples of how the script can be used.
A full description of the script and its parameters can be obtained by calling the script with the ``--help`` or ``--usage`` options.
Note that the uio or vfio kernel modules to be used, should be loaded into the kernel before
-running the ``dpdk_nic_bind.py`` script.
+running the ``dpdk-devbind.py`` script.
.. warning::
@@ -238,14 +238,14 @@ running the ``dpdk_nic_bind.py`` script.
.. warning::
- While any user can run the dpdk_nic_bind.py script to view the status of the network ports,
+ While any user can run the dpdk-devbind.py script to view the status of the network ports,
binding or unbinding network ports requires root privileges.
To see the status of all network ports on the system:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --status
+ ./tools/dpdk-devbind.py --status
Network devices using DPDK-compatible driver
============================================
@@ -267,16 +267,16 @@ To bind device ``eth1``,``04:00.1``, to the ``uio_pci_generic`` driver:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind=uio_pci_generic 04:00.1
+ ./tools/dpdk-devbind.py --bind=uio_pci_generic 04:00.1
or, alternatively,
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind=uio_pci_generic eth1
+ ./tools/dpdk-devbind.py --bind=uio_pci_generic eth1
To restore device ``82:00.0`` to its original kernel binding:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind=ixgbe 82:00.0
+ ./tools/dpdk-devbind.py --bind=ixgbe 82:00.0
diff --git a/doc/guides/linux_gsg/nic_perf_intel_platform.rst b/doc/guides/linux_gsg/nic_perf_intel_platform.rst
index b433732..d4a8362 100644
--- a/doc/guides/linux_gsg/nic_perf_intel_platform.rst
+++ b/doc/guides/linux_gsg/nic_perf_intel_platform.rst
@@ -192,12 +192,12 @@ Configurations before running DPDK
# Bind ports 82:00.0 and 85:00.0 to dpdk driver
- ./dpdk_folder/tools/dpdk_nic_bind.py -b igb_uio 82:00.0 85:00.0
+ ./dpdk_folder/tools/dpdk-devbind.py -b igb_uio 82:00.0 85:00.0
# Check the port driver status
- ./dpdk_folder/tools/dpdk_nic_bind.py --status
+ ./dpdk_folder/tools/dpdk-devbind.py --status
- See ``dpdk_nic_bind.py --help`` for more details.
+ See ``dpdk-devbind.py --help`` for more details.
More details about DPDK setup and Linux kernel requirements see :ref:`linux_gsg_compiling_dpdk`.
diff --git a/doc/guides/linux_gsg/quick_start.rst b/doc/guides/linux_gsg/quick_start.rst
index 1e0f8ff..8789b58 100644
--- a/doc/guides/linux_gsg/quick_start.rst
+++ b/doc/guides/linux_gsg/quick_start.rst
@@ -33,7 +33,7 @@
Quick Start Setup Script
========================
-The setup.sh script, found in the tools subdirectory, allows the user to perform the following tasks:
+The dpdk-setup.sh script, found in the tools subdirectory, allows the user to perform the following tasks:
* Build the DPDK libraries
@@ -63,7 +63,7 @@ the user may compile their own application that links in the EAL libraries to cr
Script Organization
-------------------
-The setup.sh script is logically organized into a series of steps that a user performs in sequence.
+The dpdk-setup.sh script is logically organized into a series of steps that a user performs in sequence.
Each step provides a number of options that guide the user to completing the desired task.
The following is a brief synopsis of each step.
@@ -98,17 +98,17 @@ The final step has options for restoring the system to its original state.
Use Cases
---------
-The following are some example of how to use the setup.sh script.
+The following are some example of how to use the dpdk-setup.sh script.
The script should be run using the source command.
Some options in the script prompt the user for further data before proceeding.
.. warning::
- The setup.sh script should be run with root privileges.
+ The dpdk-setup.sh script should be run with root privileges.
.. code-block:: console
- source tools/setup.sh
+ source tools/dpdk-setup.sh
------------------------------------------------------------------------
@@ -269,7 +269,7 @@ The following selection demonstrates the launch of the test application to run o
Applications
------------
-Once the user has run the setup.sh script, built one of the EAL targets and set up hugepages (if using one of the Linux EAL targets),
+Once the user has run the dpdk-setup.sh script, built one of the EAL targets and set up hugepages (if using one of the Linux EAL targets),
the user can then move on to building and running their application or one of the examples provided.
The examples in the /examples directory provide a good starting point to gain an understanding of the operation of the DPDK.
diff --git a/doc/guides/nics/bnx2x.rst b/doc/guides/nics/bnx2x.rst
index df8fb47..6453168 100644
--- a/doc/guides/nics/bnx2x.rst
+++ b/doc/guides/nics/bnx2x.rst
@@ -207,7 +207,7 @@ devices managed by ``librte_pmd_bnx2x`` in Linux operating system.
#. Bind the QLogic adapters to ``igb_uio`` or ``vfio-pci`` loaded in the
previous step::
- ./tools/dpdk_nic_bind.py --bind igb_uio 0000:84:00.0 0000:84:00.1
+ ./tools/dpdk-devbind.py --bind igb_uio 0000:84:00.0 0000:84:00.1
or
@@ -219,7 +219,7 @@ devices managed by ``librte_pmd_bnx2x`` in Linux operating system.
sudo chmod 0666 /dev/vfio/*
- ./tools/dpdk_nic_bind.py --bind vfio-pci 0000:84:00.0 0000:84:00.1
+ ./tools/dpdk-devbind.py --bind vfio-pci 0000:84:00.0 0000:84:00.1
#. Start ``testpmd`` with basic parameters:
diff --git a/doc/guides/nics/cxgbe.rst b/doc/guides/nics/cxgbe.rst
index d718f19..d8236b0 100644
--- a/doc/guides/nics/cxgbe.rst
+++ b/doc/guides/nics/cxgbe.rst
@@ -285,7 +285,7 @@ devices managed by librte_pmd_cxgbe in Linux operating system.
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind igb_uio 0000:02:00.4
+ ./tools/dpdk-devbind.py --bind igb_uio 0000:02:00.4
or
@@ -297,7 +297,7 @@ devices managed by librte_pmd_cxgbe in Linux operating system.
sudo chmod 0666 /dev/vfio/*
- ./tools/dpdk_nic_bind.py --bind vfio-pci 0000:02:00.4
+ ./tools/dpdk-devbind.py --bind vfio-pci 0000:02:00.4
.. note::
diff --git a/doc/guides/nics/ena.rst b/doc/guides/nics/ena.rst
index 9f93848..073b35a 100644
--- a/doc/guides/nics/ena.rst
+++ b/doc/guides/nics/ena.rst
@@ -225,7 +225,7 @@ devices managed by librte_pmd_ena.
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind=igb_uio 0000:02:00.1
+ ./tools/dpdk-devbind.py --bind=igb_uio 0000:02:00.1
#. Start testpmd with basic parameters:
diff --git a/doc/guides/nics/enic.rst b/doc/guides/nics/enic.rst
index e67c3db..2e7d7a4 100644
--- a/doc/guides/nics/enic.rst
+++ b/doc/guides/nics/enic.rst
@@ -177,13 +177,13 @@ Prerequisites
- DPDK suite should be configured based on the user's decision to use VFIO or
UIO framework
- If the vNIC device(s) to be used is bound to the kernel mode Ethernet driver
- (enic), use 'ifconfig' to bring the interface down. The dpdk_nic_bind.py tool
+ (enic), use 'ifconfig' to bring the interface down. The dpdk-devbind.py tool
can then be used to unbind the device's bus id from the enic kernel mode
driver.
- Bind the intended vNIC to vfio-pci in case the user wants ENIC PMD to use
- VFIO framework using dpdk_nic_bind.py.
+ VFIO framework using dpdk-devbind.py.
- Bind the intended vNIC to igb_uio in case the user wants ENIC PMD to use
- UIO framework using dpdk_nic_bind.py.
+ UIO framework using dpdk-devbind.py.
At this point the system should be ready to run DPDK applications. Once the
application runs to completion, the vNIC can be detached from vfio-pci or
diff --git a/doc/guides/nics/i40e.rst b/doc/guides/nics/i40e.rst
index da695af..4d12b10 100644
--- a/doc/guides/nics/i40e.rst
+++ b/doc/guides/nics/i40e.rst
@@ -164,13 +164,13 @@ devices managed by ``librte_pmd_i40e`` in the Linux operating system.
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind igb_uio 0000:83:00.0
+ ./tools/dpdk-devbind.py --bind igb_uio 0000:83:00.0
Or setup VFIO permissions for regular users and then bind to ``vfio-pci``:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind vfio-pci 0000:83:00.0
+ ./tools/dpdk-devbind.py --bind vfio-pci 0000:83:00.0
#. Start ``testpmd`` with basic parameters:
diff --git a/doc/guides/nics/intel_vf.rst b/doc/guides/nics/intel_vf.rst
index a68198f..95a79b5 100644
--- a/doc/guides/nics/intel_vf.rst
+++ b/doc/guides/nics/intel_vf.rst
@@ -151,7 +151,7 @@ For example,
modprobe uio
insmod igb_uio
- ./dpdk_nic_bind.py -b igb_uio bb:ss.f
+ ./dpdk-devbind.py -b igb_uio bb:ss.f
echo 2 > /sys/bus/pci/devices/0000\:bb\:ss.f/max_vfs (To enable two VFs on a specific PCI device)
Launch the DPDK testpmd/example or your own host daemon application using the DPDK PMD library.
@@ -236,7 +236,7 @@ For example,
modprobe uio
insmod igb_uio
- ./dpdk_nic_bind.py -b igb_uio bb:ss.f
+ ./dpdk-devbind.py -b igb_uio bb:ss.f
echo 2 > /sys/bus/pci/devices/0000\:bb\:ss.f/max_vfs (To enable two VFs on a specific PCI device)
Launch the DPDK testpmd/example or your own host daemon application using the DPDK PMD library.
@@ -285,7 +285,7 @@ For example,
.. code-block:: console
insmod igb_uio
- ./dpdk_nic_bind.py -b igb_uio bb:ss.f
+ ./dpdk-devbind.py -b igb_uio bb:ss.f
echo 2 > /sys/bus/pci/devices/0000\:bb\:ss.f/max_vfs (To enable two VFs on a specific pci device)
Launch DPDK testpmd/example or your own host daemon application using the DPDK PMD library.
@@ -406,7 +406,7 @@ The setup procedure is as follows:
modprobe uio
insmod igb_uio
- ./dpdk_nic_bind.py -b igb_uio 02:00.0 02:00.1 0e:00.0 0e:00.1
+ ./dpdk-devbind.py -b igb_uio 02:00.0 02:00.1 0e:00.0 0e:00.1
echo 2 > /sys/bus/pci/devices/0000\:02\:00.0/max_vfs
echo 2 > /sys/bus/pci/devices/0000\:02\:00.1/max_vfs
echo 2 > /sys/bus/pci/devices/0000\:0e\:00.0/max_vfs
diff --git a/doc/guides/nics/nfp.rst b/doc/guides/nics/nfp.rst
index e4ebc71..4ef6e02 100644
--- a/doc/guides/nics/nfp.rst
+++ b/doc/guides/nics/nfp.rst
@@ -242,9 +242,9 @@ Using the NFP PMD is not different to using other PMDs. Usual steps are:
useful for installing the UIO modules and for binding the right device to those
modules avoiding doing so manually:
- * **setup.sh**
- * **dpdk_nic_bind.py**
+ * **dpdk-setup.sh**
+ * **dpdk-devbind.py**
- Configuration may be performed by running setup.sh which invokes
- dpdk_nic_bind.py as needed. Executing setup.sh will display a menu of
+ Configuration may be performed by running dpdk-setup.sh which invokes
+ dpdk-devbind.py as needed. Executing dpdk-setup.sh will display a menu of
configuration options.
diff --git a/doc/guides/nics/qede.rst b/doc/guides/nics/qede.rst
index f7ca8eb..53d749c 100644
--- a/doc/guides/nics/qede.rst
+++ b/doc/guides/nics/qede.rst
@@ -177,7 +177,7 @@ devices managed by ``librte_pmd_qede`` in Linux operating system.
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind igb_uio 0000:84:00.0 0000:84:00.1 \
+ ./tools/dpdk-devbind.py --bind igb_uio 0000:84:00.0 0000:84:00.1 \
0000:84:00.2 0000:84:00.3
#. Start ``testpmd`` with basic parameters:
diff --git a/doc/guides/nics/thunderx.rst b/doc/guides/nics/thunderx.rst
index e38f260..248b1af 100644
--- a/doc/guides/nics/thunderx.rst
+++ b/doc/guides/nics/thunderx.rst
@@ -146,7 +146,7 @@ managed by ``librte_pmd_thunderx_nicvf`` in the Linux operating system.
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind vfio-pci 0002:01:00.2
+ ./tools/dpdk-devbind.py --bind vfio-pci 0002:01:00.2
#. Start ``testpmd`` with basic parameters:
@@ -246,11 +246,11 @@ This section provides instructions to configure SR-IOV with Linux OS.
Unless ``thunder-nicvf`` driver is in use make sure your kernel config includes ``CONFIG_THUNDER_NIC_VF`` setting.
-#. Verify PF/VF bind using ``dpdk_nic_bind.py``:
+#. Verify PF/VF bind using ``dpdk-devbind.py``:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --status
+ ./tools/dpdk-devbind.py --status
Example output:
@@ -268,18 +268,18 @@ This section provides instructions to configure SR-IOV with Linux OS.
modprobe vfio-pci
-#. Bind VF devices to ``vfio-pci`` using ``dpdk_nic_bind.py``:
+#. Bind VF devices to ``vfio-pci`` using ``dpdk-devbind.py``:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --bind vfio-pci 0002:01:00.1
- ./tools/dpdk_nic_bind.py --bind vfio-pci 0002:01:00.2
+ ./tools/dpdk-devbind.py --bind vfio-pci 0002:01:00.1
+ ./tools/dpdk-devbind.py --bind vfio-pci 0002:01:00.2
-#. Verify VF bind using ``dpdk_nic_bind.py``:
+#. Verify VF bind using ``dpdk-devbind.py``:
.. code-block:: console
- ./tools/dpdk_nic_bind.py --status
+ ./tools/dpdk-devbind.py --status
Example output:
diff --git a/doc/guides/nics/virtio.rst b/doc/guides/nics/virtio.rst
index c6335d4..5431015 100644
--- a/doc/guides/nics/virtio.rst
+++ b/doc/guides/nics/virtio.rst
@@ -172,7 +172,7 @@ Host2VM communication example
modprobe uio
echo 512 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
modprobe uio_pci_generic
- python tools/dpdk_nic_bind.py -b uio_pci_generic 00:03.0
+ python tools/dpdk-devbind.py -b uio_pci_generic 00:03.0
We use testpmd as the forwarding application in this example.
diff --git a/doc/guides/prog_guide/dev_kit_build_system.rst b/doc/guides/prog_guide/dev_kit_build_system.rst
index 18a3010..fa2411f 100644
--- a/doc/guides/prog_guide/dev_kit_build_system.rst
+++ b/doc/guides/prog_guide/dev_kit_build_system.rst
@@ -309,11 +309,11 @@ Misc
Internally Generated Build Tools
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-``app/pmdinfogen``
+``app/dpdk-pmdinfogen``
-``pmdinfogen`` scans an object (.o) file for various well known symbol names. These
-well known symbol names are defined by various macros and used to export
+``dpdk-pmdinfogen`` scans an object (.o) file for various well known symbol names.
+These well known symbol names are defined by various macros and used to export
important information about hardware support and usage for pmd files. For
instance the macro:
@@ -328,10 +328,10 @@ Creates the following symbol:
static char this_pmd_name0[] __attribute__((used)) = "<name>";
-Which pmdinfogen scans for. Using this information other relevant bits of data
-can be exported from the object file and used to produce a hardware support
-description, that pmdinfogen then encodes into a json formatted string in the
-following format:
+Which ``dpdk-pmdinfogen`` scans for. Using this information other relevant
+bits of data can be exported from the object file and used to produce a
+hardware support description, that ``dpdk-pmdinfogen`` then encodes into a
+json formatted string in the following format:
.. code-block:: c
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index d3a144f..b79e710 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -289,6 +289,9 @@ API Changes
* The function ``rte_eth_dev_set_mtu`` adds a new return value ``-EBUSY``, which
indicates the operation is forbidden because the port is running.
+* The script ``dpdk_nic_bind.py`` is renamed to ``dpdk-devbind.py``.
+ And the script ``setup.sh`` is renamed to ``dpdk-setup.sh``.
+
ABI Changes
-----------
diff --git a/doc/guides/sample_app_ug/pdump.rst b/doc/guides/sample_app_ug/pdump.rst
index ceb038e..ac0e7c9 100644
--- a/doc/guides/sample_app_ug/pdump.rst
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -30,15 +30,15 @@
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-dpdk_pdump Application
+dpdk-pdump Application
======================
-The ``dpdk_pdump`` tool is a Data Plane Development Kit (DPDK) tool that runs as
+The ``dpdk-pdump`` tool is a Data Plane Development Kit (DPDK) tool that runs as
a DPDK secondary process and is capable of enabling packet capture on dpdk ports.
.. Note::
- * The ``dpdk_pdump`` tool depends on libpcap based PMD which is disabled
+ * The ``dpdk-pdump`` tool depends on libpcap based PMD which is disabled
by default in the build configuration files,
owing to an external dependency on the libpcap development files
which must be installed on the board.
@@ -53,7 +53,7 @@ The tool has a number of command line options:
.. code-block:: console
- ./build/app/dpdk_pdump --
+ ./build/app/dpdk-pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue_id>),
(rx-dev=<iface or pcap file> |
@@ -95,10 +95,10 @@ PCI address (or) name of the eth device on which packets should be captured.
.. Note::
- * As of now the ``dpdk_pdump`` tool cannot capture the packets of virtual devices
+ * As of now the ``dpdk-pdump`` tool cannot capture the packets of virtual devices
in the primary process due to a bug in the ethdev library. Due to this bug, in a multi process context,
when the primary and secondary have different ports set, then the secondary process
- (here the ``dpdk_pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
+ (here the ``dpdk-pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
``queue``:
Queue id of the eth device on which packets should be captured. The user can pass a queue value of ``*`` to enable
@@ -141,4 +141,4 @@ Example
.. code-block:: console
- $ sudo ./build/app/dpdk_pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
+ $ sudo ./build/app/dpdk-pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
diff --git a/doc/guides/sample_app_ug/proc_info.rst b/doc/guides/sample_app_ug/proc_info.rst
index 542950b..73f2195 100644
--- a/doc/guides/sample_app_ug/proc_info.rst
+++ b/doc/guides/sample_app_ug/proc_info.rst
@@ -30,10 +30,10 @@
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-dpdk_proc_info Application
-==========================
+dpdk-procinfo Application
+=========================
-The dpdk_proc_info application is a Data Plane Development Kit (DPDK) application
+The dpdk-procinfo application is a Data Plane Development Kit (DPDK) application
that runs as a DPDK secondary process and is capable of retrieving port
statistics, resetting port statistics and printing DPDK memory information.
This application extends the original functionality that was supported by
@@ -45,7 +45,7 @@ The application has a number of command line options:
.. code-block:: console
- ./$(RTE_TARGET)/app/dpdk_proc_info -- -m | [-p PORTMASK] [--stats | --xstats |
+ ./$(RTE_TARGET)/app/dpdk-procinfo -- -m | [-p PORTMASK] [--stats | --xstats |
--stats-reset | --xstats-reset]
Parameters
diff --git a/doc/guides/testpmd_app_ug/testpmd_funcs.rst b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
index 30e410d..f87e0c2 100644
--- a/doc/guides/testpmd_app_ug/testpmd_funcs.rst
+++ b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
@@ -999,7 +999,7 @@ For example, to move a pci device using ixgbe under DPDK management:
.. code-block:: console
# Check the status of the available devices.
- ./tools/dpdk_nic_bind.py --status
+ ./tools/dpdk-devbind.py --status
Network devices using DPDK-compatible driver
============================================
@@ -1011,11 +1011,11 @@ For example, to move a pci device using ixgbe under DPDK management:
# Bind the device to igb_uio.
- sudo ./tools/dpdk_nic_bind.py -b igb_uio 0000:0a:00.0
+ sudo ./tools/dpdk-devbind.py -b igb_uio 0000:0a:00.0
# Recheck the status of the devices.
- ./tools/dpdk_nic_bind.py --status
+ ./tools/dpdk-devbind.py --status
Network devices using DPDK-compatible driver
============================================
0000:0a:00.0 '82599ES 10-Gigabit' drv=igb_uio unused=
@@ -1118,9 +1118,9 @@ For example, to move a pci device under kernel management:
.. code-block:: console
- sudo ./tools/dpdk_nic_bind.py -b ixgbe 0000:0a:00.0
+ sudo ./tools/dpdk-devbind.py -b ixgbe 0000:0a:00.0
- ./tools/dpdk_nic_bind.py --status
+ ./tools/dpdk-devbind.py --status
Network devices using DPDK-compatible driver
============================================
diff --git a/doc/guides/xen/pkt_switch.rst b/doc/guides/xen/pkt_switch.rst
index 3a6fc47..00a8f0c 100644
--- a/doc/guides/xen/pkt_switch.rst
+++ b/doc/guides/xen/pkt_switch.rst
@@ -323,7 +323,7 @@ Building and Running the Switching Backend
.. code-block:: console
modprobe uio_pci_generic
- python tools/dpdk_nic_bind.py -b uio_pci_generic 0000:09:00:00.0
+ python tools/dpdk-devbind.py -b uio_pci_generic 0000:09:00:00.0
In this case, 0000:09:00.0 is the PCI address for the NIC controller.
diff --git a/lib/librte_eal/common/eal_common_options.c b/lib/librte_eal/common/eal_common_options.c
index 0a594d7..481c732 100644
--- a/lib/librte_eal/common/eal_common_options.c
+++ b/lib/librte_eal/common/eal_common_options.c
@@ -116,9 +116,9 @@ TAILQ_HEAD_INITIALIZER(solib_list);
static const char *default_solib_dir = RTE_EAL_PMD_PATH;
/*
- * Stringified version of solib path used by pmdinfo.py
+ * Stringified version of solib path used by dpdk-pmdinfo.py
* Note: PLEASE DO NOT ALTER THIS without making a corresponding
- * change to tools/pmdinfo.py
+ * change to tools/dpdk-pmdinfo.py
*/
static const char dpdk_solib_path[] __attribute__((used)) =
"DPDK_PLUGIN_PATH=" RTE_EAL_PMD_PATH;
diff --git a/mk/internal/rte.compile-pre.mk b/mk/internal/rte.compile-pre.mk
index 9c25ff6..f740179 100644
--- a/mk/internal/rte.compile-pre.mk
+++ b/mk/internal/rte.compile-pre.mk
@@ -84,7 +84,7 @@ C_TO_O = $(CC) -Wp,-MD,$(call obj2dep,$(@)).tmp $(CFLAGS) \
C_TO_O_STR = $(subst ','\'',$(C_TO_O)) #'# fix syntax highlight
C_TO_O_DISP = $(if $(V),"$(C_TO_O_STR)"," CC $(@)")
endif
-PMDINFO_GEN = $(RTE_SDK_BIN)/app/pmdinfogen $@ $@.pmd.c
+PMDINFO_GEN = $(RTE_SDK_BIN)/app/dpdk-pmdinfogen $@ $@.pmd.c
PMDINFO_CC = $(CC) $(CFLAGS) -c -o $@.pmd.o $@.pmd.c
PMDINFO_LD = $(CROSS)ld $(LDFLAGS) -r -o $@.o $@.pmd.o $@
PMDINFO_TO_O = if grep -q 'PMD_REGISTER_DRIVER(.*)' $<; then \
diff --git a/mk/rte.sdkinstall.mk b/mk/rte.sdkinstall.mk
index 7cd352c..5217063 100644
--- a/mk/rte.sdkinstall.mk
+++ b/mk/rte.sdkinstall.mk
@@ -117,18 +117,22 @@ install-runtime:
$(Q)cp -a $O/lib/* $(DESTDIR)$(libdir)
$(Q)$(call rte_mkdir, $(DESTDIR)$(bindir))
$(Q)tar -cf - -C $O --exclude 'app/*.map' \
- --exclude app/pmdinfogen \
+ --exclude app/dpdk-pmdinfogen \
--exclude 'app/cmdline*' --exclude app/test \
--exclude app/testacl --exclude app/testpipeline app | \
tar -xf - -C $(DESTDIR)$(bindir) --strip-components=1 \
--keep-newer-files --warning=no-ignore-newer
$(Q)$(call rte_mkdir, $(DESTDIR)$(datadir))
$(Q)cp -a $(RTE_SDK)/tools $(DESTDIR)$(datadir)
+ $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk-setup.sh, \
+ $(DESTDIR)$(datadir)/tools/setup.sh)
+ $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk-devbind.py, \
+ $(DESTDIR)$(datadir)/tools/dpdk_nic_bind.py)
$(Q)$(call rte_mkdir, $(DESTDIR)$(sbindir))
- $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk_nic_bind.py, \
- $(DESTDIR)$(sbindir)/dpdk_nic_bind)
- $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/pmdinfo.py, \
- $(DESTDIR)$(bindir)/dpdk_pmdinfo)
+ $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk-devbind.py, \
+ $(DESTDIR)$(sbindir)/dpdk-devbind)
+ $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk-pmdinfo.py, \
+ $(DESTDIR)$(bindir)/dpdk-pmdinfo)
install-kmod:
ifneq ($(wildcard $O/kmod/*),)
@@ -146,7 +150,7 @@ install-sdk:
$(Q)cp -a $(RTE_SDK)/scripts $(DESTDIR)$(sdkdir)
$(Q)$(call rte_mkdir, $(DESTDIR)$(targetdir)/app)
$(Q)cp -a $O/.config $(DESTDIR)$(targetdir)
- $(Q)cp -a $O/app/pmdinfogen $(DESTDIR)$(targetdir)/app
+ $(Q)cp -a $O/app/dpdk-pmdinfogen $(DESTDIR)$(targetdir)/app
$(Q)$(call rte_symlink, $(DESTDIR)$(includedir), $(DESTDIR)$(targetdir)/include)
$(Q)$(call rte_symlink, $(DESTDIR)$(libdir), $(DESTDIR)$(targetdir)/lib)
diff --git a/mk/rte.sdktest.mk b/mk/rte.sdktest.mk
index ff57181..ddbbbf6 100644
--- a/mk/rte.sdktest.mk
+++ b/mk/rte.sdktest.mk
@@ -66,7 +66,7 @@ test fast_test perf_test:
fi
# this is a special target to ease the pain of running coverage tests
-# this runs all the autotests, cmdline_test script and dpdk_proc_info
+# this runs all the autotests, cmdline_test script and dpdk-procinfo
coverage:
@mkdir -p $(AUTOTEST_DIR) ; \
cd $(AUTOTEST_DIR) ; \
@@ -78,7 +78,7 @@ coverage:
$(RTE_OUTPUT)/app/test \
$(RTE_TARGET) \
$(BLACKLIST) $(WHITELIST) ; \
- $(RTE_OUTPUT)/app/dpdk_proc_info --file-prefix=ring_perf -- -m; \
+ $(RTE_OUTPUT)/app/dpdk-procinfo --file-prefix=ring_perf -- -m; \
else \
echo "No test found, please do a 'make build' first, or specify O=" ;\
fi
diff --git a/tools/dpdk_nic_bind.py b/tools/dpdk-devbind.py
similarity index 100%
rename from tools/dpdk_nic_bind.py
rename to tools/dpdk-devbind.py
diff --git a/tools/pmdinfo.py b/tools/dpdk-pmdinfo.py
similarity index 99%
rename from tools/pmdinfo.py
rename to tools/dpdk-pmdinfo.py
index 662034a..dcc8db8 100755
--- a/tools/pmdinfo.py
+++ b/tools/dpdk-pmdinfo.py
@@ -1,6 +1,5 @@
#!/usr/bin/env python
# -------------------------------------------------------------------------
-# scripts/pmdinfo.py
#
# Utility to dump PMD_INFO_STRING support from an object file
#
@@ -569,8 +568,7 @@ def main(stream=None):
optparser = OptionParser(
usage='usage: %prog [-hrtp] [-d <pci id file] <elf-file>',
description="Dump pmd hardware support info",
- add_help_option=True,
- prog='pmdinfo.py')
+ add_help_option=True)
optparser.add_option('-r', '--raw',
action='store_true', dest='raw_output',
help='Dump raw json strings')
diff --git a/tools/setup.sh b/tools/dpdk-setup.sh
similarity index 95%
rename from tools/setup.sh
rename to tools/dpdk-setup.sh
index 6097ab7..ac81b2e 100755
--- a/tools/setup.sh
+++ b/tools/dpdk-setup.sh
@@ -32,7 +32,7 @@
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
-# Run with "source /path/to/setup.sh"
+# Run with "source /path/to/dpdk-setup.sh"
#
#
@@ -422,13 +422,13 @@ grep_meminfo()
}
#
-# Calls dpdk_nic_bind.py --status to show the NIC and what they
+# Calls dpdk-devbind.py --status to show the NIC and what they
# are all bound to, in terms of drivers.
#
show_nics()
{
if [ -d /sys/module/vfio_pci -o -d /sys/module/igb_uio ]; then
- ${RTE_SDK}/tools/dpdk_nic_bind.py --status
+ ${RTE_SDK}/tools/dpdk-devbind.py --status
else
echo "# Please load the 'igb_uio' or 'vfio-pci' kernel module before "
echo "# querying or adjusting NIC device bindings"
@@ -436,16 +436,16 @@ show_nics()
}
#
-# Uses dpdk_nic_bind.py to move devices to work with vfio-pci
+# Uses dpdk-devbind.py to move devices to work with vfio-pci
#
bind_nics_to_vfio()
{
if [ -d /sys/module/vfio_pci ]; then
- ${RTE_SDK}/tools/dpdk_nic_bind.py --status
+ ${RTE_SDK}/tools/dpdk-devbind.py --status
echo ""
echo -n "Enter PCI address of device to bind to VFIO driver: "
read PCI_PATH
- sudo ${RTE_SDK}/tools/dpdk_nic_bind.py -b vfio-pci $PCI_PATH &&
+ sudo ${RTE_SDK}/tools/dpdk-devbind.py -b vfio-pci $PCI_PATH &&
echo "OK"
else
echo "# Please load the 'vfio-pci' kernel module before querying or "
@@ -454,16 +454,16 @@ bind_nics_to_vfio()
}
#
-# Uses dpdk_nic_bind.py to move devices to work with igb_uio
+# Uses dpdk-devbind.py to move devices to work with igb_uio
#
bind_nics_to_igb_uio()
{
if [ -d /sys/module/igb_uio ]; then
- ${RTE_SDK}/tools/dpdk_nic_bind.py --status
+ ${RTE_SDK}/tools/dpdk-devbind.py --status
echo ""
echo -n "Enter PCI address of device to bind to IGB UIO driver: "
read PCI_PATH
- sudo ${RTE_SDK}/tools/dpdk_nic_bind.py -b igb_uio $PCI_PATH && echo "OK"
+ sudo ${RTE_SDK}/tools/dpdk-devbind.py -b igb_uio $PCI_PATH && echo "OK"
else
echo "# Please load the 'igb_uio' kernel module before querying or "
echo "# adjusting NIC device bindings"
@@ -471,18 +471,18 @@ bind_nics_to_igb_uio()
}
#
-# Uses dpdk_nic_bind.py to move devices to work with kernel drivers again
+# Uses dpdk-devbind.py to move devices to work with kernel drivers again
#
unbind_nics()
{
- ${RTE_SDK}/tools/dpdk_nic_bind.py --status
+ ${RTE_SDK}/tools/dpdk-devbind.py --status
echo ""
echo -n "Enter PCI address of device to unbind: "
read PCI_PATH
echo ""
echo -n "Enter name of kernel driver to bind the device to: "
read DRV
- sudo ${RTE_SDK}/tools/dpdk_nic_bind.py -b $DRV $PCI_PATH && echo "OK"
+ sudo ${RTE_SDK}/tools/dpdk-devbind.py -b $DRV $PCI_PATH && echo "OK"
}
#
--
2.7.0
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-20 13:37 4% ` Olivier Matz
@ 2016-07-20 14:01 0% ` Richardson, Bruce
2016-07-20 17:20 0% ` Zoltan Kiss
1 sibling, 0 replies; 200+ results
From: Richardson, Bruce @ 2016-07-20 14:01 UTC (permalink / raw)
To: Olivier Matz, Zoltan Kiss, Zoltan Kiss, dev
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Olivier Matz
> Sent: Wednesday, July 20, 2016 2:37 PM
> To: Zoltan Kiss <zoltan.kiss@linaro.org>; Zoltan Kiss
> <zoltan.kiss@schaman.hu>; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] mempool: adjust name string size in
> related data types
>
> Hi,
>
> On 07/20/2016 02:41 PM, Zoltan Kiss wrote:
> >
> >
> > On 19/07/16 17:17, Olivier Matz wrote:
> >> Hi Zoltan,
> >>
> >> On 07/19/2016 05:59 PM, Zoltan Kiss wrote:
> >>>
> >>>
> >>> On 19/07/16 16:37, Olivier Matz wrote:
> >>>> Hi Zoltan,
> >>>>
> >>>> On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
> >>>>> A recent fix brought up an issue about the size of the 'name'
> fields:
> >>>>>
> >>>>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
> >>>>>
> >>>>> These relations should be observed:
> >>>>>
> >>>>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
> >>>>> strlen(RTE_RING_MZ_PREFIX) RTE_MEMPOOL_NAMESIZE <=
> >>>>> RTE_RING_NAMESIZE -
> >>>>> strlen(RTE_MEMPOOL_MZ_PREFIX)
> >>>>>
> >>>>> Setting all of them to 32 hides this restriction from the
> application.
> >>>>> This patch increases the memzone string size to accomodate for
> >>>>> these prefixes, and the same happens with the ring name string.
> >>>>> The ABI needs to be broken to fix this API issue, this way doesn't
> >>>>> break applications previously not failing due to the truncating
> >>>>> bug now fixed.
> >>>>>
> >>>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
> >>>>
> >>>> I agree it is a problem for an application because it cannot know
> >>>> what is the maximum name length. On the other hand, breaking the
> >>>> ABI for this looks a bit overkill. Maybe we could reduce
> >>>> RTE_MEMPOOL_NAMESIZE and RTE_RING_NAMESIZE instead of increasing
> >>>> RTE_MEMZONE_NAMESIZE? That way, we could keep the ABI as is.
> >>>
> >>> But that would break the ABI too, wouldn't it? Unless you keep the
> >>> array the same size (32 bytes) by using RTE_MEMZONE_NAMESIZE.
> >>
> >> Yes, that was the idea.
> >>
> >>> And even then, the API breaks anyway. There are applications - I
> >>> have at least some - which use all 32 bytes to store the name.
> >>> Decrease that would cause headache to change the naming scheme,
> >>> because it's a 30 character long id, and chopping the last few chars
> >>> would cause name collisions and annoying bugs.
> >>
> >> Before my patch (85cf0079), long names were silently truncated when
> >> mempool created its ring and/or memzones. Now, it returns an error.
> >
> > With 16.04 an application could operate as expected if the first 26
> > character were unique. Your patch revealed the problem that characters
> > after these were left out of the name. Now applications fail where
> > this never been a bug because their naming scheme guarantees the
> > uniqueness on the first 26 chars (or makes it very unlikely) Where the
> > first 26 is not unique, it failed earlier too, because at memzone
> > creation it checks for duplicate names.
>
> Yes, I understand that there is a behavior change for applications using
> names larger than 26 between 16.04 and 16.07. I also understand that there
> is no way for an application to know what is the maximum usable size for a
> mempool or a ring.
>
>
> >> I'm not getting why changing the struct to something like below would
> >> break the API, since it would already return an error today.
> >>
> >> #define RTE_MEMPOOL_NAMESIZE \
> >
> > Wait, this would mean applications need to recompile to use the
> > smaller value. AFAIK that's an ABI break too, right? At the moment I
> > don't see a way to fix this without breaking the ABI
>
> With this modification, if you don't recompile the application, its
> behavior will still be the same as today -> it will return ENAMETOOLONG.
> If you recompile it, the application will be aware of the maximum length.
> To me, it seems to be a acceptable compromise for this release.
>
> The patch you're proposing also changes the ABI of librte_ring and
> librte_eal, which cannot be accepted for the release.
>
>
> >
> >> (RTE_MEMZONE_NAMESIZE - sizeof(pool_prefix) - sizeof(ring
> prefix))
> >> struct rte_mempool {
> >> union {
> >> char name[RTE_MEMPOOL_NAMESIZE];
> >> char pad[32];
> >> };
> >> ...
> >> }
> >>
> >> Anyway, it may not be the proper solution since it supposes that a
> >> mempool includes a ring based on a memzone, which is not always true
> >> now with mempool handlers.
> >
> > Oh, as we dug deeper it gets better!
> > Indeed, we don't necessarily have this ring + memzone pair underneath,
> > but the user is not aware of that, and I think we should keep it that
> > way. It should only care that the string passed shouldn't be bigger
> > than a certain amount.
>
> Yes. What I'm just saying here is that it's not a good solution to write
> in the #define that "a mempool is based on a ring + a memzone", because if
> some someone adds a new mempool handler replacing the ring, and also
> creating a memzone prefixed by something larger than "rg_", we will have
> to break the ABI again.
>
>
> > Also, even though we don't necessarily have the ring, we still reserve
> > memzone's in rte_mempool_populate_default(). And their name has a 3
> > letter prefix, and a "_%d" postfix, where the %d could be as much as
> > RTE_MAX_MEMZONE in worst case (2560 by default) So actually:
> >
> > RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
> > strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen("_2560")
> >
> >
> > As a side note, there is another bug around here: rte_ring_create()
> > doesn't check for name duplications. However the user of the library
> > can lookup based on the name with rte_ring_lookup(), and it will
> > return the first ring with that name
>
> The name uniqueness is checked by rte_memzone_reserve().
>
>
> >>>> It would even be better to get rid of this static char[] for the
> >>>> structure names and replace it by an allocated const char *. I
> >>>> didn't check it's feasible for memzones. What do you think?
> >>>
> >>> It would work too, but I don't think it would help a lot. We would
> >>> still need max sizes for the names. Storing them somewhere else
> >>> won't help us in this problem.
> >>
> >> Why should we have a maximum length for the names?
> >
> > What happens if an application loads DPDK and create a mempool with a
> > name string 2 million characters long? Maybe nothing we should worry
> > about, but in general I think unlimited length function parameters are
> > problematic at the very least. The length should be passed at least
> > (which also creates a max due to the size of the param). But I think
> > it would be just easier to have these maximums set, observing the
> > above constrains.
>
> I think have a maximum name length brings more problems than not having
> it, especially ABI problems.
>
I disagree. I think we should have reasonable max names, and allow functions to return an error in case of a name being too long. However, what I think we also need to do is to guarantee a minimum maximum name length to allow apps to ensure they never hit that name-too-long error. We can guarantee that for ring/mempool etc, that the max allowed name will always be at least 20 characters, for example. That gives plenty of scope for adjusting the max as we need to, while giving others reasonable guarantees too.
/Bruce
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-20 12:41 4% ` Zoltan Kiss
@ 2016-07-20 13:37 4% ` Olivier Matz
2016-07-20 14:01 0% ` Richardson, Bruce
2016-07-20 17:20 0% ` Zoltan Kiss
0 siblings, 2 replies; 200+ results
From: Olivier Matz @ 2016-07-20 13:37 UTC (permalink / raw)
To: Zoltan Kiss, Zoltan Kiss, dev
Hi,
On 07/20/2016 02:41 PM, Zoltan Kiss wrote:
>
>
> On 19/07/16 17:17, Olivier Matz wrote:
>> Hi Zoltan,
>>
>> On 07/19/2016 05:59 PM, Zoltan Kiss wrote:
>>>
>>>
>>> On 19/07/16 16:37, Olivier Matz wrote:
>>>> Hi Zoltan,
>>>>
>>>> On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
>>>>> A recent fix brought up an issue about the size of the 'name' fields:
>>>>>
>>>>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>>>>
>>>>> These relations should be observed:
>>>>>
>>>>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>>>>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE -
>>>>> strlen(RTE_MEMPOOL_MZ_PREFIX)
>>>>>
>>>>> Setting all of them to 32 hides this restriction from the application.
>>>>> This patch increases the memzone string size to accomodate for these
>>>>> prefixes, and the same happens with the ring name string. The ABI
>>>>> needs to
>>>>> be broken to fix this API issue, this way doesn't break applications
>>>>> previously not failing due to the truncating bug now fixed.
>>>>>
>>>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>>>>
>>>> I agree it is a problem for an application because it cannot know what
>>>> is the maximum name length. On the other hand, breaking the ABI for
>>>> this
>>>> looks a bit overkill. Maybe we could reduce RTE_MEMPOOL_NAMESIZE and
>>>> RTE_RING_NAMESIZE instead of increasing RTE_MEMZONE_NAMESIZE? That way,
>>>> we could keep the ABI as is.
>>>
>>> But that would break the ABI too, wouldn't it? Unless you keep the array
>>> the same size (32 bytes) by using RTE_MEMZONE_NAMESIZE.
>>
>> Yes, that was the idea.
>>
>>> And even then, the API breaks anyway. There are applications - I have at
>>> least some - which use all 32 bytes to store the name. Decrease that
>>> would cause headache to change the naming scheme, because it's a 30
>>> character long id, and chopping the last few chars would cause name
>>> collisions and annoying bugs.
>>
>> Before my patch (85cf0079), long names were silently truncated when
>> mempool created its ring and/or memzones. Now, it returns an error.
>
> With 16.04 an application could operate as expected if the first 26
> character were unique. Your patch revealed the problem that characters
> after these were left out of the name. Now applications fail where this
> never been a bug because their naming scheme guarantees the uniqueness
> on the first 26 chars (or makes it very unlikely)
> Where the first 26 is not unique, it failed earlier too, because at
> memzone creation it checks for duplicate names.
Yes, I understand that there is a behavior change for applications using
names larger than 26 between 16.04 and 16.07. I also understand that
there is no way for an application to know what is the maximum usable
size for a mempool or a ring.
>> I'm not getting why changing the struct to something like below would
>> break the API, since it would already return an error today.
>>
>> #define RTE_MEMPOOL_NAMESIZE \
>
> Wait, this would mean applications need to recompile to use the smaller
> value. AFAIK that's an ABI break too, right? At the moment I don't see a
> way to fix this without breaking the ABI
With this modification, if you don't recompile the application, its
behavior will still be the same as today -> it will return ENAMETOOLONG.
If you recompile it, the application will be aware of the maximum
length. To me, it seems to be a acceptable compromise for this release.
The patch you're proposing also changes the ABI of librte_ring and
librte_eal, which cannot be accepted for the release.
>
>> (RTE_MEMZONE_NAMESIZE - sizeof(pool_prefix) - sizeof(ring prefix))
>> struct rte_mempool {
>> union {
>> char name[RTE_MEMPOOL_NAMESIZE];
>> char pad[32];
>> };
>> ...
>> }
>>
>> Anyway, it may not be the proper solution since it supposes that a
>> mempool includes a ring based on a memzone, which is not always true now
>> with mempool handlers.
>
> Oh, as we dug deeper it gets better!
> Indeed, we don't necessarily have this ring + memzone pair underneath,
> but the user is not aware of that, and I think we should keep it that
> way. It should only care that the string passed shouldn't be bigger than
> a certain amount.
Yes. What I'm just saying here is that it's not a good solution to write
in the #define that "a mempool is based on a ring + a memzone", because
if some someone adds a new mempool handler replacing the ring, and also
creating a memzone prefixed by something larger than "rg_", we will have
to break the ABI again.
> Also, even though we don't necessarily have the ring, we still reserve
> memzone's in rte_mempool_populate_default(). And their name has a 3
> letter prefix, and a "_%d" postfix, where the %d could be as much as
> RTE_MAX_MEMZONE in worst case (2560 by default) So actually:
>
> RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
> strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen("_2560")
>
>
> As a side note, there is another bug around here: rte_ring_create()
> doesn't check for name duplications. However the user of the library can
> lookup based on the name with rte_ring_lookup(), and it will return the
> first ring with that name
The name uniqueness is checked by rte_memzone_reserve().
>>>> It would even be better to get rid of this static char[] for the
>>>> structure names and replace it by an allocated const char *. I didn't
>>>> check it's feasible for memzones. What do you think?
>>>
>>> It would work too, but I don't think it would help a lot. We would still
>>> need max sizes for the names. Storing them somewhere else won't help us
>>> in this problem.
>>
>> Why should we have a maximum length for the names?
>
> What happens if an application loads DPDK and create a mempool with a
> name string 2 million characters long? Maybe nothing we should worry
> about, but in general I think unlimited length function parameters are
> problematic at the very least. The length should be passed at least
> (which also creates a max due to the size of the param). But I think it
> would be just easier to have these maximums set, observing the above
> constrains.
I think have a maximum name length brings more problems than not having
it, especially ABI problems.
Regards,
Olivier
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-19 16:17 0% ` Olivier Matz
@ 2016-07-20 12:41 4% ` Zoltan Kiss
2016-07-20 13:37 4% ` Olivier Matz
0 siblings, 1 reply; 200+ results
From: Zoltan Kiss @ 2016-07-20 12:41 UTC (permalink / raw)
To: Olivier Matz, Zoltan Kiss, dev
On 19/07/16 17:17, Olivier Matz wrote:
> Hi Zoltan,
>
> On 07/19/2016 05:59 PM, Zoltan Kiss wrote:
>>
>>
>> On 19/07/16 16:37, Olivier Matz wrote:
>>> Hi Zoltan,
>>>
>>> On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
>>>> A recent fix brought up an issue about the size of the 'name' fields:
>>>>
>>>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>>>
>>>> These relations should be observed:
>>>>
>>>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>>>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE -
>>>> strlen(RTE_MEMPOOL_MZ_PREFIX)
>>>>
>>>> Setting all of them to 32 hides this restriction from the application.
>>>> This patch increases the memzone string size to accomodate for these
>>>> prefixes, and the same happens with the ring name string. The ABI
>>>> needs to
>>>> be broken to fix this API issue, this way doesn't break applications
>>>> previously not failing due to the truncating bug now fixed.
>>>>
>>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>>>
>>> I agree it is a problem for an application because it cannot know what
>>> is the maximum name length. On the other hand, breaking the ABI for this
>>> looks a bit overkill. Maybe we could reduce RTE_MEMPOOL_NAMESIZE and
>>> RTE_RING_NAMESIZE instead of increasing RTE_MEMZONE_NAMESIZE? That way,
>>> we could keep the ABI as is.
>>
>> But that would break the ABI too, wouldn't it? Unless you keep the array
>> the same size (32 bytes) by using RTE_MEMZONE_NAMESIZE.
>
> Yes, that was the idea.
>
>> And even then, the API breaks anyway. There are applications - I have at
>> least some - which use all 32 bytes to store the name. Decrease that
>> would cause headache to change the naming scheme, because it's a 30
>> character long id, and chopping the last few chars would cause name
>> collisions and annoying bugs.
>
> Before my patch (85cf0079), long names were silently truncated when
> mempool created its ring and/or memzones. Now, it returns an error.
With 16.04 an application could operate as expected if the first 26
character were unique. Your patch revealed the problem that characters
after these were left out of the name. Now applications fail where this
never been a bug because their naming scheme guarantees the uniqueness
on the first 26 chars (or makes it very unlikely)
Where the first 26 is not unique, it failed earlier too, because at
memzone creation it checks for duplicate names.
>
> I'm not getting why changing the struct to something like below would
> break the API, since it would already return an error today.
>
> #define RTE_MEMPOOL_NAMESIZE \
Wait, this would mean applications need to recompile to use the smaller
value. AFAIK that's an ABI break too, right? At the moment I don't see a
way to fix this without breaking the ABI
> (RTE_MEMZONE_NAMESIZE - sizeof(pool_prefix) - sizeof(ring prefix))
> struct rte_mempool {
> union {
> char name[RTE_MEMPOOL_NAMESIZE];
> char pad[32];
> };
> ...
> }
>
> Anyway, it may not be the proper solution since it supposes that a
> mempool includes a ring based on a memzone, which is not always true now
> with mempool handlers.
Oh, as we dug deeper it gets better!
Indeed, we don't necessarily have this ring + memzone pair underneath,
but the user is not aware of that, and I think we should keep it that
way. It should only care that the string passed shouldn't be bigger than
a certain amount.
Also, even though we don't necessarily have the ring, we still reserve
memzone's in rte_mempool_populate_default(). And their name has a 3
letter prefix, and a "_%d" postfix, where the %d could be as much as
RTE_MAX_MEMZONE in worst case (2560 by default) So actually:
RTE_MEMPOOL_NAMESIZE <= RTE_MEMZONE_NAMESIZE -
strlen(RTE_MEMPOOL_MZ_PREFIX) - strlen("_2560")
As a side note, there is another bug around here: rte_ring_create()
doesn't check for name duplications. However the user of the library can
lookup based on the name with rte_ring_lookup(), and it will return the
first ring with that name
>
>>> It would even be better to get rid of this static char[] for the
>>> structure names and replace it by an allocated const char *. I didn't
>>> check it's feasible for memzones. What do you think?
>>
>> It would work too, but I don't think it would help a lot. We would still
>> need max sizes for the names. Storing them somewhere else won't help us
>> in this problem.
>
> Why should we have a maximum length for the names?
What happens if an application loads DPDK and create a mempool with a
name string 2 million characters long? Maybe nothing we should worry
about, but in general I think unlimited length function parameters are
problematic at the very least. The length should be passed at least
(which also creates a max due to the size of the param). But I think it
would be just easier to have these maximums set, observing the above
constrains.
>
>
> Thanks,
> Olivier
>
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-20 2:16 0% ` Lu, Wenzhuo
@ 2016-07-20 10:41 2% ` Adrien Mazarguil
2016-07-21 3:18 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Adrien Mazarguil @ 2016-07-20 10:41 UTC (permalink / raw)
To: Lu, Wenzhuo
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
Hi Wenzhuo,
On Wed, Jul 20, 2016 at 02:16:51AM +0000, Lu, Wenzhuo wrote:
[...]
> > So, today an application cannot combine N-tuple and FDIR flow rules and get a
> > reliable outcome, unless it is designed for specific devices with a known
> > behavior.
> >
> > > What's the right behavior of PMD if APP want to create a flow director rule
> > which has a higher or even equal priority than an existing n-tuple rule? Should
> > PMD return fail?
> >
> > First remember applications only deal with the generic API, PMDs are
> > responsible for choosing the most appropriate HW implementation to use
> > according to the requested flow rules (FDIR, N-tuple or anything else).
> >
> > For the specific case of FDIR vs N-tuple, if the underlying HW supports both I do
> > not see why the PMD would create a N-tuple rule. Doesn't FDIR support
> > everything N-tuple can do and much more?
> Talking about the filters, fdir can cover n-tuple. I think that's why i40e only supports fdir but not n-tuple. But n-tuple has its own highlight. As we know, at least on intel NICs, fdir only supports per device mask. But n-tuple can support per rule mask.
> As every pattern has spec and mask both, we cannot guarantee the masks are same. I think ixgbe will try to use n-tuple first if can. Because even the masks are different, we can support them all.
OK, makes sense. In that case existing rules may indeed prevent subsequent
ones from getting created if their priority is wrong. I do not think there
is a way around that if the application needs this exact ordering.
> > Assuming such a thing happened anyway, that the PMD had to create a rule
> > using a high priority filter type and that the application requests the creation of a
> > rule that can only be done using a lower priority filter type, but also requested a
> > higher priority for that rule, then yes, it should obviously fail.
> >
> > That is, unless the PMD can perform some kind of workaround to have both.
> >
> > > If so, do we need more fail reasons? According to this RFC, I think we need
> > return " EEXIST: collision with an existing rule. ", but it's not very clear, APP
> > doesn't know the problem is priority, maybe more detailed reason is helpful.
> >
> > Possibly, I've defined a basic set of errors, there are quite a number of errno
> > values to choose from. However I think we should not define too many values.
> > In my opinion the basic set covers every possible failure:
> >
> > - EINVAL: invalid format, rule is broken or cannot be understood by the PMD
> > anyhow.
> >
> > - ENOTSUP: pattern/actions look fine but something in the requested rule is
> > not supported and thus cannot be applied.
> >
> > - EEXIST: pattern/actions are fine and could have been applied if only some
> > other rule did not prevent the PMD to do it (I see it as the closest thing
> > to "ETOOBAD" which unfortunately does not exist).
> >
> > - ENOMEM: like EEXIST, except it is due to the lack of resources not because
> > of another rule. I wasn't sure which of ENOMEM or ENOSPC was better but
> > settled on ENOMEM as it is well known. Still open to debate.
> >
> > Errno values are only useful to get a rough idea of the reason, and another
> > mechanism is needed to pinpoint the exact problem for debugging/reporting
> > purposes, something like:
> >
> > enum rte_flow_error_type {
> > RTE_FLOW_ERROR_TYPE_NONE,
> > RTE_FLOW_ERROR_TYPE_UNKNOWN,
> > RTE_FLOW_ERROR_TYPE_PRIORITY,
> > RTE_FLOW_ERROR_TYPE_PATTERN,
> > RTE_FLOW_ERROR_TYPE_ACTION,
> > };
> >
> > struct rte_flow_error {
> > enum rte_flow_error_type type;
> > void *offset; /* Points to the exact pattern item or action. */
> > const char *message;
> > };
> When we are using a CLI and it fails, normally it will let us know which parameter is not appropriate. So, I think it’s a good idea to have this error structure :)
Agreed.
> > Then either provide an optional struct rte_flow_error pointer to
> > rte_flow_validate(), or a separate function (rte_flow_analyze()?), since
> > processing this may be quite expensive and applications may not care about the
> > exact reason.
> Agree the processing may be too expensive. Maybe we can say it's optional to return error details. And that's a good question that what APP should do if creating the rule fails. I believe normally it will choose handle the rule by itself. But I think it's not bad to feedback more. Or even the APP want to adjust the rules, it cannot be an option for lack of info.
All right then, I'll add it to the specification.
int
rte_flow_validate(uint8_t port_id,
const struct rte_flow_pattern *pattern,
const struct rte_flow_actions *actions,
struct rte_flow_error *error);
With error possibly NULL if the application does not care. Is it fine for
you?
[...]
> > > > > > - PMDs, not applications, are responsible for maintaining flow rules
> > > > > > configuration when stopping and restarting a port or performing other
> > > > > > actions which may affect them. They can only be destroyed explicitly.
> > > > > Don’t understand " They can only be destroyed explicitly."
> > > >
> > > > This part says that as long as an application has not called
> > > > rte_flow_destroy() on a flow rule, it never disappears, whatever
> > > > happens to the port (stopped, restarted). The application is not
> > > > responsible for re-creating rules after that.
> > > >
> > > > Note that according to the specification, this may translate to not
> > > > being able to stop a port as long as a flow rule is present,
> > > > depending on how nice the PMD intends to be with applications.
> > > > Implementation can be done in small steps with minimal amount of code on
> > the PMD side.
> > > Does it mean PMD should store and maintain all the rules? Why not let rte do
> > that? I think if PMD maintain all the rules, it means every kind of NIC should have
> > a copy of code for the rules. But if rte do that, only one copy of code need to be
> > maintained, right?
> >
> > I've considered having rules stored in a common format understood at the RTE
> > level and not specific to each PMD and decided that the opaque rte_flow pointer
> > was a better choice for the following reasons:
> >
> > - Even though flow rules management is done in the control path, processing
> > must be as fast as possible. Letting PMDs store flow rules using their own
> > internal representation gives them the chance to achieve better
> > performance.
> Not quite understand. I think we're talking about maintain the rules by SW. I don’t think there's something need to be optimized according to specific NICs. If we need to optimize the code, I think we need to consider the CPU, OS ... and some common means. I'm wrong?
Perhaps we were talking about different things, here I was only explaining
why rte_flow (the result of creating a flow rule) should be opaque and fully
managed by the PMD. More on the SW side of things below.
> > - An opaque context managed by PMDs would probably have to be stored
> > somewhere as well anyway.
> >
> > - PMDs may not need to allocate/store anything at all if they exclusively
> > rely on HW state for everything. In my opinion, the generic API has enough
> > constraints for this to work and maintain consistency between flow
> > rules. Note this is currently how most PMDs implement FDIR and other
> > filter types.
> Yes, the rules are stored by HW. But considering stop/start the device, the rules in HW will lose. we have to store the rules by SW and re-program them when restarting the device.
Assume a HW capable of keeping flow rules programmed even during a
stop/start cycle (e.g. mlx4/mlx5 may be able to do it from DPDK point of
view), don't you think it is more efficient to standardize on this behavior
and let PMDs restore flow rules for HW that do not support it regardless of
whether it would be done by RTE or the application (SW)?
> And in existing code, we store the filters by SW at least on Intel NICs. But I think we cannot reuse them, because considering the priority and which category of filter should be chosen, I think we need a whole new table for generic API. I think it’s what's designed now, right?
So I understand you'd want RTE to help your PMD keep track of the flow rules
it created?
Nothing wrong with that, all I'm saying is that it should be entirely
optional. RTE should not automatically maintain a list. PMDs have to call
RTE helpers if they need help to maintain a context. These helpers are not
defined in this API yet because it is difficult to know what will be useful
in advance.
> > - RTE can (and will) provide helpers to avoid most of the code redundancy,
> > PMDs are free to use them or manage everything by themselves.
> >
> > - Given that the opaque rte_flow pointer associated with a flow rule is to
> > be stored by the application, PMDs do not even have to keep references to
> > them.
> Don’t understand. More details?
In an application:
rte_flow *foo = rte_flow_create(...);
In the above example, foo cannot be dereferenced by the application nor RTE,
only the PMD is aware of its contents. This object can only be used with
rte_flow*() functions.
PMDs are thus free to make this object grow as needed when adding internal
features without breaking any kind of public API/ABI.
What I meant is, given that the application is supposed to store foo
somewhere in order to destroy it later, the PMD does not have to keep track
of that pointer assuming it does not need to access it later on its own for
some reason.
> > - The flow rules format described in this specification (pattern / actions)
> > will be used by applications directly, and will be free to arrange them in
> > lists, trees or in any other way if they need to keep flow specifications
> > around for further processing.
> Who will create the lists, trees or something else? According to previous discussion, I think the APP will program the rules one by one. So if APP organize the rules to lists, trees..., PMD doesn’t know that.
> And you said " Given that the opaque rte_flow pointer associated with a flow rule is to be stored by the application ". I'm lost here.
I guess that's because we're discussing two different things, flow rule
specifications and flow rule objects. Let me sum it up:
- Flow rule specifications are the patterns/actions combinations provided by
applications to rte_flow_create(). Applications can store those as needed
and organize them as they wish (hash, tree, list). Neither PMDs nor RTE
will do it for them.
- Flow rule objects (struct rte_flow *) are generated when a flow rule is
created. Applications must keep these around if they want to manipulate
them later (i.e. destroy or query existing rules).
Then PMDs *may* need to keep and arrange flow rule objects internally for
management purposes. Could be because HW requires it, detecting conflicting
rules, managing priorities and so on. Possible reasons are not described in
this API because these are thought as PMD-specific needs.
> > > When the port is stopped and restarted, rte can reconfigure the rules. Is the
> > concern that PMD may adjust the sequence of the rules according to the priority,
> > so every NIC has a different list of rules? But PMD can adjust them again when
> > rte reconfiguring the rules.
> >
> > What about PMDs able to stop and restart ports without destroying their own
> > flow rules? If we assume flow rules must be destroyed when stopping a port,
> > these PMDs are needlessly penalized with slower stop/start cycles. Think about
> > it assuming thousands of flow rules.
> I believe the rules maintained by SW should not be destroyed, because they're used to be re-programed when the device starts again.
Do we agree that applications should not care? Flow rules configured before
stopping a port must still be there after restarting it.
What we seem to not agree about is that you think RTE should be responsible
for restoring flow rules of devices that lose them when stopped. I think
doing so is unfair to devices for which it is not the case and not really
nice to applications, so my opinion is that the PMD is responsible for
restoring flow rules however it wants. It is free to use RTE helpers to keep
their track, as long as it's all managed internally.
> > Thus from an application point of view, whatever happens when stopping and
> > restarting a port should not matter. If a flow rule was present before, it must
> > still be present afterwards. If the PMD had to destroy flow rules and re-create
> > them, it does not actually matter if they differ slightly at the HW level, as long as:
> >
> > - Existing opaque flow rule pointers (rte_flow) are still valid to the PMD
> > and refer to the same rules.
> >
> > - The overall behavior of all rules is the same.
> >
> > The list of rules you think of (patterns / actions) is maintained by applications
> > (not RTE), and only if they need them. RTE would needlessly duplicate this.
> As said before, need more details to understand this. Maybe an example is better :)
The generic format both RTE and applications might understand is the one
described in this API (struct rte_flow_pattern and struct
rte_flow_actions).
If we wanted RTE to maintain some sort of per-port state for flow rule
specifications, it would have to be a copy of these structures arranged
somehow (list or something else).
If we consider that PMDs need to keep a context object associated to a flow
rule (the opaque struct rte_flow *), then RTE would most likely have to
store it along with the flow specification.
Such a list may not be useful to applications (list lookups take time), so
they would implement their own redundant method. They might also require
extra room to attach some application context to flow rules. A generic list
cannot plan for it.
Applications know what they want to do with flow rules and are responsible
for managing them efficiently with RTE out of the way.
I'm not sure if this answered your question, if not, please describe a
scenario where a RTE-managed list of flow rules would be mandatory.
--
Adrien Mazarguil
6WIND
^ permalink raw reply [relevance 2%]
* Re: [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
@ 2016-07-20 8:54 4% ` Ferruh Yigit
2016-07-27 8:33 4% ` Thomas Monjalon
` (3 subsequent siblings)
4 siblings, 0 replies; 200+ results
From: Ferruh Yigit @ 2016-07-20 8:54 UTC (permalink / raw)
To: Olivier Matz, dev; +Cc: jerin.jacob, thomas.monjalon, bruce.richardson
On 7/20/2016 8:16 AM, Olivier Matz wrote:
> For 16.11, the mbuf structure will be modified implying ABI breakage.
> Some discussions already took place here:
> http://www.dpdk.org/dev/patchwork/patch/12878/
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> ---
>
> v1->v2:
> - reword the sentences to keep things more open, as suggested by Bruce
>
> doc/guides/rel_notes/deprecation.rst | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index f502f86..b9f5a93 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -41,3 +41,9 @@ Deprecation Notices
> * The mempool functions for single/multi producer/consumer are deprecated and
> will be removed in 16.11.
> It is replaced by rte_mempool_generic_get/put functions.
> +
> +* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some fields
> + may be reordered to facilitate the writing of ``data_off``, ``refcnt``, and
> + ``nb_segs`` in one operation, because some platforms have an overhead if the
> + store address is not naturally aligned. Other mbuf fields, such as the
> + ``port`` field, may be moved or removed as part of this mbuf work.
>
Not directly related to this patch, but generally for deprecation
notices, does it make sense to tag explicitly which library effected, like:
* librte_mbuf [perhaps with version here]:
Explanation about deprecation ...
For this case it is more clear which library effected, but sometimes
that is not obvious from deprecation notice.
Also when checked for if specific library effected, it is harder to find
with current notes.
Thanks,
ferruh
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH v2] doc: announce ABI change for mbuf structure
2016-07-19 14:01 13% [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure Olivier Matz
2016-07-19 14:40 4% ` Bruce Richardson
@ 2016-07-20 7:16 13% ` Olivier Matz
2016-07-20 8:54 4% ` Ferruh Yigit
` (4 more replies)
1 sibling, 5 replies; 200+ results
From: Olivier Matz @ 2016-07-20 7:16 UTC (permalink / raw)
To: dev; +Cc: jerin.jacob, thomas.monjalon, bruce.richardson
For 16.11, the mbuf structure will be modified implying ABI breakage.
Some discussions already took place here:
http://www.dpdk.org/dev/patchwork/patch/12878/
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
v1->v2:
- reword the sentences to keep things more open, as suggested by Bruce
doc/guides/rel_notes/deprecation.rst | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f502f86..b9f5a93 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -41,3 +41,9 @@ Deprecation Notices
* The mempool functions for single/multi producer/consumer are deprecated and
will be removed in 16.11.
It is replaced by rte_mempool_generic_get/put functions.
+
+* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some fields
+ may be reordered to facilitate the writing of ``data_off``, ``refcnt``, and
+ ``nb_segs`` in one operation, because some platforms have an overhead if the
+ store address is not naturally aligned. Other mbuf fields, such as the
+ ``port`` field, may be moved or removed as part of this mbuf work.
--
2.8.1
^ permalink raw reply [relevance 13%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-19 13:12 0% ` Adrien Mazarguil
@ 2016-07-20 2:16 0% ` Lu, Wenzhuo
2016-07-20 10:41 2% ` Adrien Mazarguil
0 siblings, 1 reply; 200+ results
From: Lu, Wenzhuo @ 2016-07-20 2:16 UTC (permalink / raw)
To: Adrien Mazarguil
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
Hi Adrien,
> -----Original Message-----
> From: Adrien Mazarguil [mailto:adrien.mazarguil@6wind.com]
> Sent: Tuesday, July 19, 2016 9:12 PM
> To: Lu, Wenzhuo
> Cc: dev@dpdk.org; Thomas Monjalon; Zhang, Helin; Wu, Jingjing; Rasesh Mody;
> Ajit Khaparde; Rahul Lakkireddy; Jan Medala; John Daley; Chen, Jing D; Ananyev,
> Konstantin; Matej Vido; Alejandro Lucero; Sony Chacko; Jerin Jacob; De Lara
> Guarch, Pablo; Olga Shern
> Subject: Re: [RFC] Generic flow director/filtering/classification API
>
> On Tue, Jul 19, 2016 at 08:11:48AM +0000, Lu, Wenzhuo wrote:
> > Hi Adrien,
> > Thanks for your clarification. Most of my questions are clear, but still
> something may need to be discussed, comment below.
>
> Hi Wenzhuo,
>
> Please see below.
>
> [...]
> > > > > Requirements for a new API:
> > > > >
> > > > > - Flexible and extensible without causing API/ABI problems for existing
> > > > > applications.
> > > > > - Should be unambiguous and easy to use.
> > > > > - Support existing filtering features and actions listed in `Filter types`_.
> > > > > - Support packet alteration.
> > > > > - In case of overlapping filters, their priority should be well documented.
> > > > Does that mean we don't guarantee the consistent of priority? The
> > > > priority can
> > > be different on different NICs. So the behavior of the actions can be
> different.
> > > Right?
> > >
> > > No, the intent is precisely to define what happens in order to get a
> > > consistent result across different devices, and document cases with
> undefined behavior.
> > > There must be no room left for interpretation.
> > >
> > > For example, the API must describe what happens when two overlapping
> > > filters (e.g. one matching an Ethernet header, another one matching
> > > an IP header) match a given packet at a given priority level.
> > >
> > > It is documented in section 4.1.1 (priorities) as "undefined behavior".
> > > Applications remain free to do it and deal with consequences, at
> > > least they know they cannot expect a consistent outcome, unless they
> > > use different priority levels for both rules, see also 4.4.5 (flow rules priority).
> > >
> > > > Seems the users still need to aware the some details of the HW? Do
> > > > we need
> > > to add the negotiation for the priority?
> > >
> > > Priorities as defined in this document may not be directly mappable
> > > to HW capabilities (e.g. HW does not support enough priorities, or
> > > that some corner case make them not work as described), in which
> > > case the PMD may choose to simulate priorities (again 4.4.5), as
> > > long as the end result follows the specification.
> > >
> > > So users must not be aware of some HW details, the PMD does and must
> > > perform the needed workarounds to suit their expectations. Users may
> > > only be impacted by errors while attempting to create rules that are
> > > either unsupported or would cause them (or existing rules) to diverge from
> the spec.
> > The problem is sometime the priority of the filters is fixed according
> > to
> > > HW's implementation. For example, on ixgbe, n-tuple has a higher
> > > priority than flow director.
>
> As a side note I did not know that N-tuple had a higher priority than flow
> director on ixgbe, priorities among filter types do not seem to be documented at
> all in DPDK. This is one of the reasons I think we need a generic API to handle
> flow configuration.
Totally agree with you. We haven't documented the info well enough. And even we do that, users have to study the details of every NIC, it can still make the filters very hard to use. I believe a generic API is very helpful here :)
>
>
> So, today an application cannot combine N-tuple and FDIR flow rules and get a
> reliable outcome, unless it is designed for specific devices with a known
> behavior.
>
> > What's the right behavior of PMD if APP want to create a flow director rule
> which has a higher or even equal priority than an existing n-tuple rule? Should
> PMD return fail?
>
> First remember applications only deal with the generic API, PMDs are
> responsible for choosing the most appropriate HW implementation to use
> according to the requested flow rules (FDIR, N-tuple or anything else).
>
> For the specific case of FDIR vs N-tuple, if the underlying HW supports both I do
> not see why the PMD would create a N-tuple rule. Doesn't FDIR support
> everything N-tuple can do and much more?
Talking about the filters, fdir can cover n-tuple. I think that's why i40e only supports fdir but not n-tuple. But n-tuple has its own highlight. As we know, at least on intel NICs, fdir only supports per device mask. But n-tuple can support per rule mask.
As every pattern has spec and mask both, we cannot guarantee the masks are same. I think ixgbe will try to use n-tuple first if can. Because even the masks are different, we can support them all.
>
> Assuming such a thing happened anyway, that the PMD had to create a rule
> using a high priority filter type and that the application requests the creation of a
> rule that can only be done using a lower priority filter type, but also requested a
> higher priority for that rule, then yes, it should obviously fail.
>
> That is, unless the PMD can perform some kind of workaround to have both.
>
> > If so, do we need more fail reasons? According to this RFC, I think we need
> return " EEXIST: collision with an existing rule. ", but it's not very clear, APP
> doesn't know the problem is priority, maybe more detailed reason is helpful.
>
> Possibly, I've defined a basic set of errors, there are quite a number of errno
> values to choose from. However I think we should not define too many values.
> In my opinion the basic set covers every possible failure:
>
> - EINVAL: invalid format, rule is broken or cannot be understood by the PMD
> anyhow.
>
> - ENOTSUP: pattern/actions look fine but something in the requested rule is
> not supported and thus cannot be applied.
>
> - EEXIST: pattern/actions are fine and could have been applied if only some
> other rule did not prevent the PMD to do it (I see it as the closest thing
> to "ETOOBAD" which unfortunately does not exist).
>
> - ENOMEM: like EEXIST, except it is due to the lack of resources not because
> of another rule. I wasn't sure which of ENOMEM or ENOSPC was better but
> settled on ENOMEM as it is well known. Still open to debate.
>
> Errno values are only useful to get a rough idea of the reason, and another
> mechanism is needed to pinpoint the exact problem for debugging/reporting
> purposes, something like:
>
> enum rte_flow_error_type {
> RTE_FLOW_ERROR_TYPE_NONE,
> RTE_FLOW_ERROR_TYPE_UNKNOWN,
> RTE_FLOW_ERROR_TYPE_PRIORITY,
> RTE_FLOW_ERROR_TYPE_PATTERN,
> RTE_FLOW_ERROR_TYPE_ACTION,
> };
>
> struct rte_flow_error {
> enum rte_flow_error_type type;
> void *offset; /* Points to the exact pattern item or action. */
> const char *message;
> };
When we are using a CLI and it fails, normally it will let us know which parameter is not appropriate. So, I think it’s a good idea to have this error structure :)
>
> Then either provide an optional struct rte_flow_error pointer to
> rte_flow_validate(), or a separate function (rte_flow_analyze()?), since
> processing this may be quite expensive and applications may not care about the
> exact reason.
Agree the processing may be too expensive. Maybe we can say it's optional to return error details. And that's a good question that what APP should do if creating the rule fails. I believe normally it will choose handle the rule by itself. But I think it's not bad to feedback more. Or even the APP want to adjust the rules, it cannot be an option for lack of info.
>
> What do you suggest?
>
> > > > > Behavior
> > > > > --------
> > > > >
> > > > > - API operations are synchronous and blocking (``EAGAIN`` cannot be
> > > > > returned).
> > > > >
> > > > > - There is no provision for reentrancy/multi-thread safety, although
> nothing
> > > > > should prevent different devices from being configured at the same
> > > > > time. PMDs may protect their control path functions accordingly.
> > > > >
> > > > > - Stopping the data path (TX/RX) should not be necessary when
> > > > > managing
> > > flow
> > > > > rules. If this cannot be achieved naturally or with workarounds (such as
> > > > > temporarily replacing the burst function pointers), an appropriate error
> > > > > code must be returned (``EBUSY``).
> > > > PMD cannot stop the data path without adding lock. So I think if
> > > > some rules
> > > cannot be applied without stopping rx/tx, PMD has to return fail.
> > > > Or let the APP to stop the data path.
> > >
> > > Agreed, that is the intent. If the PMD cannot touch flow rules for
> > > some reason even after trying really hard, then it just returns EBUSY.
> > >
> > > Perhaps we should write down that applications may get a different
> > > outcome after stopping the data path if they get EBUSY?
> > Agree, it's better to describe more about the APP. BTW, I checked the
> > behavior of ixgbe/igb, I think we can add/delete filters during
> > runtime. Hopefully we'll not hit too many EBUSY problems on other NICs
> > :)
>
> OK, I will add it.
>
> > > > > - PMDs, not applications, are responsible for maintaining flow rules
> > > > > configuration when stopping and restarting a port or performing other
> > > > > actions which may affect them. They can only be destroyed explicitly.
> > > > Don’t understand " They can only be destroyed explicitly."
> > >
> > > This part says that as long as an application has not called
> > > rte_flow_destroy() on a flow rule, it never disappears, whatever
> > > happens to the port (stopped, restarted). The application is not
> > > responsible for re-creating rules after that.
> > >
> > > Note that according to the specification, this may translate to not
> > > being able to stop a port as long as a flow rule is present,
> > > depending on how nice the PMD intends to be with applications.
> > > Implementation can be done in small steps with minimal amount of code on
> the PMD side.
> > Does it mean PMD should store and maintain all the rules? Why not let rte do
> that? I think if PMD maintain all the rules, it means every kind of NIC should have
> a copy of code for the rules. But if rte do that, only one copy of code need to be
> maintained, right?
>
> I've considered having rules stored in a common format understood at the RTE
> level and not specific to each PMD and decided that the opaque rte_flow pointer
> was a better choice for the following reasons:
>
> - Even though flow rules management is done in the control path, processing
> must be as fast as possible. Letting PMDs store flow rules using their own
> internal representation gives them the chance to achieve better
> performance.
Not quite understand. I think we're talking about maintain the rules by SW. I don’t think there's something need to be optimized according to specific NICs. If we need to optimize the code, I think we need to consider the CPU, OS ... and some common means. I'm wrong?
>
> - An opaque context managed by PMDs would probably have to be stored
> somewhere as well anyway.
>
> - PMDs may not need to allocate/store anything at all if they exclusively
> rely on HW state for everything. In my opinion, the generic API has enough
> constraints for this to work and maintain consistency between flow
> rules. Note this is currently how most PMDs implement FDIR and other
> filter types.
Yes, the rules are stored by HW. But considering stop/start the device, the rules in HW will lose. we have to store the rules by SW and re-program them when restarting the device.
And in existing code, we store the filters by SW at least on Intel NICs. But I think we cannot reuse them, because considering the priority and which category of filter should be chosen, I think we need a whole new table for generic API. I think it’s what's designed now, right?
>
> - RTE can (and will) provide helpers to avoid most of the code redundancy,
> PMDs are free to use them or manage everything by themselves.
>
> - Given that the opaque rte_flow pointer associated with a flow rule is to
> be stored by the application, PMDs do not even have to keep references to
> them.
Don’t understand. More details?
>
> - The flow rules format described in this specification (pattern / actions)
> will be used by applications directly, and will be free to arrange them in
> lists, trees or in any other way if they need to keep flow specifications
> around for further processing.
Who will create the lists, trees or something else? According to previous discussion, I think the APP will program the rules one by one. So if APP organize the rules to lists, trees..., PMD doesn’t know that.
And you said " Given that the opaque rte_flow pointer associated with a flow rule is to be stored by the application ". I'm lost here.
>
> > When the port is stopped and restarted, rte can reconfigure the rules. Is the
> concern that PMD may adjust the sequence of the rules according to the priority,
> so every NIC has a different list of rules? But PMD can adjust them again when
> rte reconfiguring the rules.
>
> What about PMDs able to stop and restart ports without destroying their own
> flow rules? If we assume flow rules must be destroyed when stopping a port,
> these PMDs are needlessly penalized with slower stop/start cycles. Think about
> it assuming thousands of flow rules.
I believe the rules maintained by SW should not be destroyed, because they're used to be re-programed when the device starts again.
>
> Thus from an application point of view, whatever happens when stopping and
> restarting a port should not matter. If a flow rule was present before, it must
> still be present afterwards. If the PMD had to destroy flow rules and re-create
> them, it does not actually matter if they differ slightly at the HW level, as long as:
>
> - Existing opaque flow rule pointers (rte_flow) are still valid to the PMD
> and refer to the same rules.
>
> - The overall behavior of all rules is the same.
>
> The list of rules you think of (patterns / actions) is maintained by applications
> (not RTE), and only if they need them. RTE would needlessly duplicate this.
As said before, need more details to understand this. Maybe an example is better :)
>
> --
> Adrien Mazarguil
> 6WIND
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-19 15:59 3% ` Zoltan Kiss
@ 2016-07-19 16:17 0% ` Olivier Matz
2016-07-20 12:41 4% ` Zoltan Kiss
0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-07-19 16:17 UTC (permalink / raw)
To: Zoltan Kiss, Zoltan Kiss, dev
Hi Zoltan,
On 07/19/2016 05:59 PM, Zoltan Kiss wrote:
>
>
> On 19/07/16 16:37, Olivier Matz wrote:
>> Hi Zoltan,
>>
>> On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
>>> A recent fix brought up an issue about the size of the 'name' fields:
>>>
>>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>>
>>> These relations should be observed:
>>>
>>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE -
>>> strlen(RTE_MEMPOOL_MZ_PREFIX)
>>>
>>> Setting all of them to 32 hides this restriction from the application.
>>> This patch increases the memzone string size to accomodate for these
>>> prefixes, and the same happens with the ring name string. The ABI
>>> needs to
>>> be broken to fix this API issue, this way doesn't break applications
>>> previously not failing due to the truncating bug now fixed.
>>>
>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>>
>> I agree it is a problem for an application because it cannot know what
>> is the maximum name length. On the other hand, breaking the ABI for this
>> looks a bit overkill. Maybe we could reduce RTE_MEMPOOL_NAMESIZE and
>> RTE_RING_NAMESIZE instead of increasing RTE_MEMZONE_NAMESIZE? That way,
>> we could keep the ABI as is.
>
> But that would break the ABI too, wouldn't it? Unless you keep the array
> the same size (32 bytes) by using RTE_MEMZONE_NAMESIZE.
Yes, that was the idea.
> And even then, the API breaks anyway. There are applications - I have at
> least some - which use all 32 bytes to store the name. Decrease that
> would cause headache to change the naming scheme, because it's a 30
> character long id, and chopping the last few chars would cause name
> collisions and annoying bugs.
Before my patch (85cf0079), long names were silently truncated when
mempool created its ring and/or memzones. Now, it returns an error.
I'm not getting why changing the struct to something like below would
break the API, since it would already return an error today.
#define RTE_MEMPOOL_NAMESIZE \
(RTE_MEMZONE_NAMESIZE - sizeof(pool_prefix) - sizeof(ring prefix))
struct rte_mempool {
union {
char name[RTE_MEMPOOL_NAMESIZE];
char pad[32];
};
...
}
Anyway, it may not be the proper solution since it supposes that a
mempool includes a ring based on a memzone, which is not always true now
with mempool handlers.
>> It would even be better to get rid of this static char[] for the
>> structure names and replace it by an allocated const char *. I didn't
>> check it's feasible for memzones. What do you think?
>
> It would work too, but I don't think it would help a lot. We would still
> need max sizes for the names. Storing them somewhere else won't help us
> in this problem.
Why should we have a maximum length for the names?
Thanks,
Olivier
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-19 15:37 4% ` Olivier Matz
@ 2016-07-19 15:59 3% ` Zoltan Kiss
2016-07-19 16:17 0% ` Olivier Matz
0 siblings, 1 reply; 200+ results
From: Zoltan Kiss @ 2016-07-19 15:59 UTC (permalink / raw)
To: Olivier Matz, Zoltan Kiss, dev
On 19/07/16 16:37, Olivier Matz wrote:
> Hi Zoltan,
>
> On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
>> A recent fix brought up an issue about the size of the 'name' fields:
>>
>> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>>
>> These relations should be observed:
>>
>> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
>> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE - strlen(RTE_MEMPOOL_MZ_PREFIX)
>>
>> Setting all of them to 32 hides this restriction from the application.
>> This patch increases the memzone string size to accomodate for these
>> prefixes, and the same happens with the ring name string. The ABI needs to
>> be broken to fix this API issue, this way doesn't break applications
>> previously not failing due to the truncating bug now fixed.
>>
>> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
>
> I agree it is a problem for an application because it cannot know what
> is the maximum name length. On the other hand, breaking the ABI for this
> looks a bit overkill. Maybe we could reduce RTE_MEMPOOL_NAMESIZE and
> RTE_RING_NAMESIZE instead of increasing RTE_MEMZONE_NAMESIZE? That way,
> we could keep the ABI as is.
But that would break the ABI too, wouldn't it? Unless you keep the array
the same size (32 bytes) by using RTE_MEMZONE_NAMESIZE.
And even then, the API breaks anyway. There are applications - I have at
least some - which use all 32 bytes to store the name. Decrease that
would cause headache to change the naming scheme, because it's a 30
character long id, and chopping the last few chars would cause name
collisions and annoying bugs.
>
> It would even be better to get rid of this static char[] for the
> structure names and replace it by an allocated const char *. I didn't
> check it's feasible for memzones. What do you think?
It would work too, but I don't think it would help a lot. We would still
need max sizes for the names. Storing them somewhere else won't help us
in this problem.
>
> In any case, I think it's a bit late for 16.07 for this kind of fix.
>
> Regards,
> Olivier
>
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
2016-07-19 14:37 3% ` [dpdk-dev] [PATCH] mempool: adjust name string size in related data types Zoltan Kiss
@ 2016-07-19 15:37 4% ` Olivier Matz
2016-07-19 15:59 3% ` Zoltan Kiss
2016-07-20 17:16 12% ` [dpdk-dev] [PATCH v2] " Zoltan Kiss
1 sibling, 1 reply; 200+ results
From: Olivier Matz @ 2016-07-19 15:37 UTC (permalink / raw)
To: Zoltan Kiss, dev
Hi Zoltan,
On 07/19/2016 04:37 PM, Zoltan Kiss wrote:
> A recent fix brought up an issue about the size of the 'name' fields:
>
> 85cf0079 mem: avoid memzone/mempool/ring name truncation
>
> These relations should be observed:
>
> RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
> RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE - strlen(RTE_MEMPOOL_MZ_PREFIX)
>
> Setting all of them to 32 hides this restriction from the application.
> This patch increases the memzone string size to accomodate for these
> prefixes, and the same happens with the ring name string. The ABI needs to
> be broken to fix this API issue, this way doesn't break applications
> previously not failing due to the truncating bug now fixed.
>
> Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
I agree it is a problem for an application because it cannot know what
is the maximum name length. On the other hand, breaking the ABI for this
looks a bit overkill. Maybe we could reduce RTE_MEMPOOL_NAMESIZE and
RTE_RING_NAMESIZE instead of increasing RTE_MEMZONE_NAMESIZE? That way,
we could keep the ABI as is.
It would even be better to get rid of this static char[] for the
structure names and replace it by an allocated const char *. I didn't
check it's feasible for memzones. What do you think?
In any case, I think it's a bit late for 16.07 for this kind of fix.
Regards,
Olivier
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure
2016-07-19 15:07 4% ` Richardson, Bruce
@ 2016-07-19 15:28 4% ` Olivier Matz
0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-07-19 15:28 UTC (permalink / raw)
To: Richardson, Bruce; +Cc: dev, jerin.jacob, thomas.monjalon
On 07/19/2016 05:07 PM, Richardson, Bruce wrote:
>
>
>> -----Original Message-----
>> From: Olivier Matz [mailto:olivier.matz@6wind.com]
>> Sent: Tuesday, July 19, 2016 4:04 PM
>> To: Richardson, Bruce <bruce.richardson@intel.com>
>> Cc: dev@dpdk.org; jerin.jacob@caviumnetworks.com;
>> thomas.monjalon@6wind.com
>> Subject: Re: [dpdk-dev] [PATCH] doc: announce ABI change for mbuf
>> structure
>>
>> Hi Bruce,
>>
>> On 07/19/2016 04:40 PM, Bruce Richardson wrote:
>>> On Tue, Jul 19, 2016 at 04:01:15PM +0200, Olivier Matz wrote:
>>>> For 16.11, the mbuf structure will be modified implying ABI breakage.
>>>> Some discussions already took place here:
>>>> http://www.dpdk.org/dev/patchwork/patch/12878/
>>>>
>>>> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
>>>> ---
>>>> doc/guides/rel_notes/deprecation.rst | 6 ++++++
>>>> 1 file changed, 6 insertions(+)
>>>>
>>>> diff --git a/doc/guides/rel_notes/deprecation.rst
>>>> b/doc/guides/rel_notes/deprecation.rst
>>>> index f502f86..2245bc2 100644
>>>> --- a/doc/guides/rel_notes/deprecation.rst
>>>> +++ b/doc/guides/rel_notes/deprecation.rst
>>>> @@ -41,3 +41,9 @@ Deprecation Notices
>>>> * The mempool functions for single/multi producer/consumer are
>> deprecated and
>>>> will be removed in 16.11.
>>>> It is replaced by rte_mempool_generic_get/put functions.
>>>> +
>>>> +* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure:
>>>> +some
>>>> + fields will be reordered to facilitate the writing of
>>>> +``data_off``,
>>>> + ``refcnt``, and ``nb_segs`` in one operation. Indeed, some
>>>> +platforms
>>>> + have an overhead if the store address is not naturally aligned.
>>>> +The
>>>> + useless ``port`` field will also be removed at the same occasion.
>>>> --
>>>
>>> Have we fully bottomed out on the mbuf changes. I'm not sure that once
>>> patches start getting considered for merge, new opinions may come
>>> forward. For instance, is the "port" field really "useless"?
>>>
>>> Would it not be better to put in a less specific deprecation notice?
>>> What happens if this notice goes in and the final changes are
>>> different from those called out here?
>>
>> Yes, you are right. What about the following text?
>>
>> ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some
>> fields may be reordered to facilitate the writing of ``data_off``,
>> ``refcnt``, and ``nb_segs`` in one operation. Indeed, some platforms have
>> an overhead if the store address is not naturally aligned. The ``port``
>> field may also be removed at the same occasion.
>>
> Better. Two suggestions:
> 1. change "Indeed" to "because" and join the sentences.
> 2. change the last sentence to be even more general: "Other mbuf fields, such as the port field, may be moved or removed as part of this mbuf work".
It's much better indeed ;)
Thanks Bruce, I'll submit a v2.
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure
2016-07-19 15:04 7% ` Olivier Matz
@ 2016-07-19 15:07 4% ` Richardson, Bruce
2016-07-19 15:28 4% ` Olivier Matz
0 siblings, 1 reply; 200+ results
From: Richardson, Bruce @ 2016-07-19 15:07 UTC (permalink / raw)
To: Olivier Matz; +Cc: dev, jerin.jacob, thomas.monjalon
> -----Original Message-----
> From: Olivier Matz [mailto:olivier.matz@6wind.com]
> Sent: Tuesday, July 19, 2016 4:04 PM
> To: Richardson, Bruce <bruce.richardson@intel.com>
> Cc: dev@dpdk.org; jerin.jacob@caviumnetworks.com;
> thomas.monjalon@6wind.com
> Subject: Re: [dpdk-dev] [PATCH] doc: announce ABI change for mbuf
> structure
>
> Hi Bruce,
>
> On 07/19/2016 04:40 PM, Bruce Richardson wrote:
> > On Tue, Jul 19, 2016 at 04:01:15PM +0200, Olivier Matz wrote:
> >> For 16.11, the mbuf structure will be modified implying ABI breakage.
> >> Some discussions already took place here:
> >> http://www.dpdk.org/dev/patchwork/patch/12878/
> >>
> >> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> >> ---
> >> doc/guides/rel_notes/deprecation.rst | 6 ++++++
> >> 1 file changed, 6 insertions(+)
> >>
> >> diff --git a/doc/guides/rel_notes/deprecation.rst
> >> b/doc/guides/rel_notes/deprecation.rst
> >> index f502f86..2245bc2 100644
> >> --- a/doc/guides/rel_notes/deprecation.rst
> >> +++ b/doc/guides/rel_notes/deprecation.rst
> >> @@ -41,3 +41,9 @@ Deprecation Notices
> >> * The mempool functions for single/multi producer/consumer are
> deprecated and
> >> will be removed in 16.11.
> >> It is replaced by rte_mempool_generic_get/put functions.
> >> +
> >> +* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure:
> >> +some
> >> + fields will be reordered to facilitate the writing of
> >> +``data_off``,
> >> + ``refcnt``, and ``nb_segs`` in one operation. Indeed, some
> >> +platforms
> >> + have an overhead if the store address is not naturally aligned.
> >> +The
> >> + useless ``port`` field will also be removed at the same occasion.
> >> --
> >
> > Have we fully bottomed out on the mbuf changes. I'm not sure that once
> > patches start getting considered for merge, new opinions may come
> > forward. For instance, is the "port" field really "useless"?
> >
> > Would it not be better to put in a less specific deprecation notice?
> > What happens if this notice goes in and the final changes are
> > different from those called out here?
>
> Yes, you are right. What about the following text?
>
> ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some
> fields may be reordered to facilitate the writing of ``data_off``,
> ``refcnt``, and ``nb_segs`` in one operation. Indeed, some platforms have
> an overhead if the store address is not naturally aligned. The ``port``
> field may also be removed at the same occasion.
>
Better. Two suggestions:
1. change "Indeed" to "because" and join the sentences.
2. change the last sentence to be even more general: "Other mbuf fields, such as the port field, may be moved or removed as part of this mbuf work".
/Bruce
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure
2016-07-19 14:40 4% ` Bruce Richardson
@ 2016-07-19 15:04 7% ` Olivier Matz
2016-07-19 15:07 4% ` Richardson, Bruce
0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-07-19 15:04 UTC (permalink / raw)
To: Bruce Richardson; +Cc: dev, jerin.jacob, thomas.monjalon
Hi Bruce,
On 07/19/2016 04:40 PM, Bruce Richardson wrote:
> On Tue, Jul 19, 2016 at 04:01:15PM +0200, Olivier Matz wrote:
>> For 16.11, the mbuf structure will be modified implying ABI breakage.
>> Some discussions already took place here:
>> http://www.dpdk.org/dev/patchwork/patch/12878/
>>
>> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
>> ---
>> doc/guides/rel_notes/deprecation.rst | 6 ++++++
>> 1 file changed, 6 insertions(+)
>>
>> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
>> index f502f86..2245bc2 100644
>> --- a/doc/guides/rel_notes/deprecation.rst
>> +++ b/doc/guides/rel_notes/deprecation.rst
>> @@ -41,3 +41,9 @@ Deprecation Notices
>> * The mempool functions for single/multi producer/consumer are deprecated and
>> will be removed in 16.11.
>> It is replaced by rte_mempool_generic_get/put functions.
>> +
>> +* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some
>> + fields will be reordered to facilitate the writing of ``data_off``,
>> + ``refcnt``, and ``nb_segs`` in one operation. Indeed, some platforms
>> + have an overhead if the store address is not naturally aligned. The
>> + useless ``port`` field will also be removed at the same occasion.
>> --
>
> Have we fully bottomed out on the mbuf changes. I'm not sure that once patches
> start getting considered for merge, new opinions may come forward. For instance,
> is the "port" field really "useless"?
>
> Would it not be better to put in a less specific deprecation notice? What happens
> if this notice goes in and the final changes are different from those called out
> here?
Yes, you are right. What about the following text?
ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some
fields may be reordered to facilitate the writing of ``data_off``,
``refcnt``, and ``nb_segs`` in one operation. Indeed, some platforms
have an overhead if the store address is not naturally aligned. The
``port`` field may also be removed at the same occasion.
Thanks,
Olivier
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure
2016-07-19 14:01 13% [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure Olivier Matz
@ 2016-07-19 14:40 4% ` Bruce Richardson
2016-07-19 15:04 7% ` Olivier Matz
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
1 sibling, 1 reply; 200+ results
From: Bruce Richardson @ 2016-07-19 14:40 UTC (permalink / raw)
To: Olivier Matz; +Cc: dev, jerin.jacob, thomas.monjalon
On Tue, Jul 19, 2016 at 04:01:15PM +0200, Olivier Matz wrote:
> For 16.11, the mbuf structure will be modified implying ABI breakage.
> Some discussions already took place here:
> http://www.dpdk.org/dev/patchwork/patch/12878/
>
> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> ---
> doc/guides/rel_notes/deprecation.rst | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index f502f86..2245bc2 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -41,3 +41,9 @@ Deprecation Notices
> * The mempool functions for single/multi producer/consumer are deprecated and
> will be removed in 16.11.
> It is replaced by rte_mempool_generic_get/put functions.
> +
> +* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some
> + fields will be reordered to facilitate the writing of ``data_off``,
> + ``refcnt``, and ``nb_segs`` in one operation. Indeed, some platforms
> + have an overhead if the store address is not naturally aligned. The
> + useless ``port`` field will also be removed at the same occasion.
> --
Have we fully bottomed out on the mbuf changes. I'm not sure that once patches
start getting considered for merge, new opinions may come forward. For instance,
is the "port" field really "useless"?
Would it not be better to put in a less specific deprecation notice? What happens
if this notice goes in and the final changes are different from those called out
here?
/Bruce
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH] mempool: adjust name string size in related data types
@ 2016-07-19 14:37 3% ` Zoltan Kiss
2016-07-19 15:37 4% ` Olivier Matz
2016-07-20 17:16 12% ` [dpdk-dev] [PATCH v2] " Zoltan Kiss
0 siblings, 2 replies; 200+ results
From: Zoltan Kiss @ 2016-07-19 14:37 UTC (permalink / raw)
To: dev; +Cc: olivier.matz
A recent fix brought up an issue about the size of the 'name' fields:
85cf0079 mem: avoid memzone/mempool/ring name truncation
These relations should be observed:
RTE_RING_NAMESIZE <= RTE_MEMZONE_NAMESIZE - strlen(RTE_RING_MZ_PREFIX)
RTE_MEMPOOL_NAMESIZE <= RTE_RING_NAMESIZE - strlen(RTE_MEMPOOL_MZ_PREFIX)
Setting all of them to 32 hides this restriction from the application.
This patch increases the memzone string size to accomodate for these
prefixes, and the same happens with the ring name string. The ABI needs to
be broken to fix this API issue, this way doesn't break applications
previously not failing due to the truncating bug now fixed.
Signed-off-by: Zoltan Kiss <zoltan.kiss@schaman.hu>
---
lib/librte_eal/common/include/rte_memzone.h | 2 +-
lib/librte_mempool/rte_mempool.h | 4 +++-
lib/librte_ring/rte_ring.h | 5 ++++-
3 files changed, 8 insertions(+), 3 deletions(-)
diff --git a/lib/librte_eal/common/include/rte_memzone.h b/lib/librte_eal/common/include/rte_memzone.h
index f69b5a8..ba3a1f0 100644
--- a/lib/librte_eal/common/include/rte_memzone.h
+++ b/lib/librte_eal/common/include/rte_memzone.h
@@ -74,7 +74,7 @@ extern "C" {
*/
struct rte_memzone {
-#define RTE_MEMZONE_NAMESIZE 32 /**< Maximum length of memory zone name.*/
+#define RTE_MEMZONE_NAMESIZE (32 + 6) /**< Maximum length of memory zone name.*/
char name[RTE_MEMZONE_NAMESIZE]; /**< Name of the memory zone. */
phys_addr_t phys_addr; /**< Start physical address. */
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 4a8fbb1..61e8d19 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -123,7 +123,9 @@ struct rte_mempool_objsz {
/**< Total size of an object (header + elt + trailer). */
};
-#define RTE_MEMPOOL_NAMESIZE 32 /**< Maximum length of a memory pool. */
+/**< Maximum length of a memory pool's name. */
+#define RTE_MEMPOOL_NAMESIZE (RTE_RING_NAMESIZE - \
+ sizeof(RTE_MEMPOOL_MZ_PREFIX) + 1)
#define RTE_MEMPOOL_MZ_PREFIX "MP_"
/* "MP_<name>" */
diff --git a/lib/librte_ring/rte_ring.h b/lib/librte_ring/rte_ring.h
index eb45e41..d6185de 100644
--- a/lib/librte_ring/rte_ring.h
+++ b/lib/librte_ring/rte_ring.h
@@ -100,6 +100,7 @@ extern "C" {
#include <rte_lcore.h>
#include <rte_atomic.h>
#include <rte_branch_prediction.h>
+#include <rte_memzone.h>
#define RTE_TAILQ_RING_NAME "RTE_RING"
@@ -126,8 +127,10 @@ struct rte_ring_debug_stats {
} __rte_cache_aligned;
#endif
-#define RTE_RING_NAMESIZE 32 /**< The maximum length of a ring name. */
#define RTE_RING_MZ_PREFIX "RG_"
+/**< The maximum length of a ring name. */
+#define RTE_RING_NAMESIZE (RTE_MEMZONE_NAMESIZE - \
+ sizeof(RTE_RING_MZ_PREFIX) + 1)
#ifndef RTE_RING_PAUSE_REP_COUNT
#define RTE_RING_PAUSE_REP_COUNT 0 /**< Yield after pause num of times, no yield
--
1.9.1
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure
@ 2016-07-19 14:01 13% Olivier Matz
2016-07-19 14:40 4% ` Bruce Richardson
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
0 siblings, 2 replies; 200+ results
From: Olivier Matz @ 2016-07-19 14:01 UTC (permalink / raw)
To: dev; +Cc: jerin.jacob, thomas.monjalon
For 16.11, the mbuf structure will be modified implying ABI breakage.
Some discussions already took place here:
http://www.dpdk.org/dev/patchwork/patch/12878/
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
doc/guides/rel_notes/deprecation.rst | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f502f86..2245bc2 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -41,3 +41,9 @@ Deprecation Notices
* The mempool functions for single/multi producer/consumer are deprecated and
will be removed in 16.11.
It is replaced by rte_mempool_generic_get/put functions.
+
+* ABI changes are planned for 16.11 in the ``rte_mbuf`` structure: some
+ fields will be reordered to facilitate the writing of ``data_off``,
+ ``refcnt``, and ``nb_segs`` in one operation. Indeed, some platforms
+ have an overhead if the store address is not naturally aligned. The
+ useless ``port`` field will also be removed at the same occasion.
--
2.8.1
^ permalink raw reply [relevance 13%]
* Re: [dpdk-dev] [PATCH] rte_delay_us can be replaced with user function
@ 2016-07-19 13:17 3% ` Wiles, Keith
0 siblings, 0 replies; 200+ results
From: Wiles, Keith @ 2016-07-19 13:17 UTC (permalink / raw)
To: jozmarti; +Cc: dev
Hi Jozef,
I have two quick comments inline.
> On Jul 19, 2016, at 7:42 AM, jozmarti@cisco.com wrote:
>
> From: Jozef Martiniak <jozmarti@cisco.com>
>
> when running single-core, some drivers tend to call rte_delay_us for a
> long time, and that is causing packet drops.
> Attached patch introduces 2 new functions:
>
> void rte_delay_us_callback_register(void(*userfunc)(unsigned));
> void rte_delay_us_callback_unregister(void);
>
> First one replaces rte_delay_us with userfunc and second one restores
> original rte_delay_us.
> Test user_delay_us is included.
>
> Signed-off-by: Jozef Martiniak <jozmarti@cisco.com>
> ---
> app/test/test_cycles.c | 39 ++++++++++++++++++++++
> lib/librte_eal/common/eal_common_timer.c | 19 +++++++++++
> lib/librte_eal/common/include/generic/rte_cycles.h | 13 ++++++++
> 3 files changed, 71 insertions(+)
>
> diff --git a/app/test/test_cycles.c b/app/test/test_cycles.c
> index f6c043a..2b44a53 100644
> --- a/app/test/test_cycles.c
> +++ b/app/test/test_cycles.c
> @@ -90,3 +90,42 @@ test_cycles(void)
> }
>
> REGISTER_TEST_COMMAND(cycles_autotest, test_cycles);
> +
> +/*
> + * rte_delay_us_callback test
> + *
> + * - check if callback is correctly registered/unregistered
> + *
> + */
> +
> +static int pattern;
> +static void my_rte_delay_us(unsigned us)
> +{
> + pattern += us;
> +}
> +
> +static int
> +test_user_delay_us(void)
> +{
> + pattern = 0;
> +
> + rte_delay_us_callback_register(my_rte_delay_us);
> +
> + rte_delay_us(2);
> + if (pattern != 2)
> + return -1;
> +
> + rte_delay_us(3);
> + if (pattern != 5)
> + return -1;
> +
> + rte_delay_us_callback_unregister();
> +
> + rte_delay_us(3);
> + if (pattern != 5)
> + return -1;
> +
> + return 0;
> +}
> +
> +REGISTER_TEST_COMMAND(user_delay_us, test_user_delay_us);
> diff --git a/lib/librte_eal/common/eal_common_timer.c b/lib/librte_eal/common/eal_common_timer.c
> index c4227cd..a982562 100644
> --- a/lib/librte_eal/common/eal_common_timer.c
> +++ b/lib/librte_eal/common/eal_common_timer.c
> @@ -47,9 +47,18 @@
> /* The frequency of the RDTSC timer resolution */
> static uint64_t eal_tsc_resolution_hz;
>
> +/* User function which replaces rte_delay_us function */
> +static void (*rte_delay_us_override)(unsigned) = NULL;
> +
> void
> rte_delay_us(unsigned us)
> {
> + if (unlikely(rte_delay_us_override != NULL))
> + {
> + rte_delay_us_override(us);
> + return;
> + }
> +
> const uint64_t start = rte_get_timer_cycles();
> const uint64_t ticks = (uint64_t)us * rte_get_timer_hz() / 1E6;
> while ((rte_get_timer_cycles() - start) < ticks)
> @@ -84,3 +93,13 @@ set_tsc_freq(void)
> RTE_LOG(DEBUG, EAL, "TSC frequency is ~%" PRIu64 " KHz\n", freq / 1000);
> eal_tsc_resolution_hz = freq;
> }
> +
> +void rte_delay_us_callback_register(void (*userfunc)(unsigned))
> +{
> + rte_delay_us_override = userfunc;
> +}
> +
> +void rte_delay_us_callback_unregister(void)
> +{
> + rte_delay_us_override = NULL;
> +}
I guess I would have used the rte_delay_us_callback_register(NULL) to unregister, but this is fine.
> diff --git a/lib/librte_eal/common/include/generic/rte_cycles.h b/lib/librte_eal/common/include/generic/rte_cycles.h
> index 8cc21f2..274f798 100644
> --- a/lib/librte_eal/common/include/generic/rte_cycles.h
> +++ b/lib/librte_eal/common/include/generic/rte_cycles.h
> @@ -202,4 +202,17 @@ rte_delay_ms(unsigned ms)
> rte_delay_us(ms * 1000);
> }
>
> +/**
> + * Replace rte_delay_us with user defined function.
> + *
> + * @param userfunc
> + * User function which replaces rte_delay_us.
> + */
> +void rte_delay_us_callback_register(void(*userfunc)(unsigned));
> +
> +/**
> + * Unregister user callback function. Restores original rte_delay_us.
> + */
> +void rte_delay_us_callback_unregister(void);
Just a note we need to add these two new APIs to the map file for ABI checking.
Other then these two comments I would give this one a +1 unless someone else has some comments.
> +
> #endif /* _RTE_CYCLES_H_ */
> --
> 2.1.4
>
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v1] doc: fix release notes for 16.07
@ 2016-07-19 13:16 13% John McNamara
0 siblings, 0 replies; 200+ results
From: John McNamara @ 2016-07-19 13:16 UTC (permalink / raw)
To: dev; +Cc: John McNamara
Fix grammar, spelling and formatting of DPDK 16.07 release notes.
Signed-off-by: John McNamara <john.mcnamara@intel.com>
---
doc/guides/rel_notes/release_16_07.rst | 134 +++++++++++++++++----------------
1 file changed, 71 insertions(+), 63 deletions(-)
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index d3a144f..38887a5 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -40,16 +40,16 @@ New Features
* **Added mempool external cache for non-EAL thread.**
- Added new functions to create, free or flush a user-owned mempool
- cache for non-EAL threads. Previously, cache was always disabled
- on these threads.
+ Added new functions to create, free or flush a user-owned mempool
+ cache for non-EAL threads. Previously the cache was always disabled
+ on these threads.
* **Changed the memory allocation in mempool library.**
* Added ability to allocate a large mempool in virtually fragmented memory.
* Added new APIs to populate a mempool with memory.
* Added an API to free a mempool.
- * Modified the API of rte_mempool_obj_iter() function.
+ * Modified the API of the ``rte_mempool_obj_iter()`` function.
* Dropped specific Xen Dom0 code.
* Dropped specific anonymous mempool code in testpmd.
@@ -63,10 +63,10 @@ New Features
* **Added mailbox interrupt support for ixgbe and igb VFs.**
- When the physical NIC link comes down or up, the PF driver will send a
+ When the physical NIC link comes up or down, the PF driver will send a
mailbox message to notify each VF. To handle this link up/down event,
- add mailbox interrupts support to receive the message and allow the app to
- register a callback for it.
+ support have been added for a mailbox interrupt to receive the message and
+ allow the application to register a callback for it.
* **Updated the ixgbe base driver.**
@@ -74,51 +74,49 @@ New Features
following:
* Added sgmii link for X550.
- * Added mac link setup for X550a SFP and SFP+.
+ * Added MAC link setup for X550a SFP and SFP+.
* Added KR support for X550em_a.
- * Added new phy definitions for M88E1500.
+ * Added new PHY definitions for M88E1500.
* Added support for the VLVF to be bypassed when adding/removing a VFTA entry.
* Added X550a flow control auto negotiation support.
* **Updated the i40e base driver.**
- Updated the i40e base driver, which includes support for new devices IDs.
+ Updated the i40e base driver including support for new devices IDs.
-* **Supported virtio on IBM POWER8.**
+* **Added support for virtio on IBM POWER8.**
The ioports are mapped in memory when using Linux UIO.
-* **Virtio support for containers.**
+* **Added support for Virtio in containers.**
Add a new virtual device, named virtio-user, to support virtio for containers.
Known limitations:
* Control queue and multi-queue are not supported yet.
- * Cannot work with --huge-unlink.
- * Cannot work with --no-huge.
- * Cannot work when there are more than VHOST_MEMORY_MAX_NREGIONS(8) hugepages.
- * Root privilege is a must for sorting hugepages by physical address.
- * Can only be used with vhost user backend.
+ * Doesn't work with ``--huge-unlink``.
+ * Doesn't work with ``--no-huge``.
+ * Doesn't work when there are more than ``VHOST_MEMORY_MAX_NREGIONS(8)`` hugepages.
+ * Root privilege is required for sorting hugepages by physical address.
+ * Can only be used with the vhost user backend.
* **Added vhost-user client mode.**
- DPDK vhost-user could be the server as well as the client. It supports
- server mode only before, now it also supports client mode. Client mode
- is enabled when ``RTE_VHOST_USER_CLIENT`` flag is set while calling
+ DPDK vhost-user now supports client mode as well as server mode. Client mode
+ is enabled when the ``RTE_VHOST_USER_CLIENT`` flag is set while calling
``rte_vhost_driver_register``.
- When DPDK vhost-user restarts from normal or abnormal quit (say crash),
- the client mode would allow DPDK to establish the connect again. Note
- that a brand new QEMU version (v2.7 or above) is needed, otherwise, the
- reconnect won't work.
+ When DPDK vhost-user restarts from an normal or abnormal exit (such as a
+ crash), the client mode allows DPDK to establish the connection again. Note
+ that QEMU version v2.7 or above is required for this feature.
- DPDK vhost-user will also try to reconnect by default when
+ DPDK vhost-user will also try to reconnect by default when:
- * the first connect fails (when QEMU is not started yet)
- * the connection is broken (when QEMU restarts)
+ * The first connect fails (when QEMU is not started yet).
+ * The connection is broken (when QEMU restarts).
- It can be turned off if flag ``RTE_VHOST_USER_NO_RECONNECT`` is set.
+ It can be turned off by setting the ``RTE_VHOST_USER_NO_RECONNECT`` flag.
* **Added NSH packet recognition in i40e.**
@@ -127,7 +125,7 @@ New Features
Now AESNI MB PMD supports 128/192/256-bit counter mode AES encryption and
decryption.
-* **Added support of AES counter mode for Intel QuickAssist devices.**
+* **Added support for AES counter mode with Intel QuickAssist devices.**
Enabled support for the AES CTR algorithm for Intel QuickAssist devices.
Provided support for algorithm-chaining operations.
@@ -141,33 +139,33 @@ New Features
The following features/modifications have been added to rte_hash library:
- * Enabled application developers to use an extra flag for rte_hash creation
- to specify default behavior (multi-thread safe/unsafe) with rte_hash_add_key
- function.
- * Changed Cuckoo search algorithm to breadth first search for multi-writer
- routine and split Cuckoo Search and Move operations in order to reduce
- transactional code region and improve TSX performance.
- * Added a hash multi-writer test case for test app.
+ * Enabled application developers to use an extra flag for ``rte_hash``
+ creation to specify default behavior (multi-thread safe/unsafe) with the
+ ``rte_hash_add_key`` function.
+ * Changed the Cuckoo Hash Search algorithm to breadth first search for
+ multi-writer routines and split Cuckoo Hash Search and Move operations in
+ order to reduce transactional code region and improve TSX performance.
+ * Added a hash multi-writer test case to the test app.
* **Improved IP Pipeline Application.**
- The following features have been added to ip_pipeline application:
+ The following features have been added to the ip_pipeline application:
- * Configure the MAC address in the routing pipeline and automatic routes
+ * Configure the MAC address in the routing pipeline and automatic route
updates with change in link state.
* Enable RSS per network interface through the configuration file.
* Streamline the CLI code.
* **Added keepalive enhancements.**
- Adds support for reporting of core states other than dead to
+ Added support for reporting of core states other than dead to
monitoring applications, enabling the support of broader liveness
reporting to external processes.
* **Added packet capture framework.**
- * A new library ``librte_pdump`` is added to provide packet capture API.
- * A new ``app/pdump`` tool is added to capture packets in DPDK.
+ * A new library ``librte_pdump`` is added to provide a packet capture API.
+ * A new ``app/pdump`` tool is added to demonstrate capture packets in DPDK.
* **Added floating VEB support for i40e PF driver.**
@@ -197,18 +195,20 @@ EAL
* **igb_uio: Fixed possible mmap failure for Linux >= 4.5.**
- mmaping the iomem range of the PCI device fails for kernels that
- enabled CONFIG_IO_STRICT_DEVMEM option:
+ The mmaping of the iomem range of the PCI device fails for kernels that
+ enabled the ``CONFIG_IO_STRICT_DEVMEM`` option. The error seen by the
+ user is as similar to the following::
- EAL: pci_map_resource():
- cannot mmap(39, 0x7f1c51800000, 0x100000, 0x0):
- Invalid argument (0xffffffffffffffff)
+ EAL: pci_map_resource():
- CONFIG_IO_STRICT_DEVMEM is introduced in Linux v4.5
+ cannot mmap(39, 0x7f1c51800000, 0x100000, 0x0):
+ Invalid argument (0xffffffffffffffff)
- Updated igb_uio to stop reserving PCI memory resources, from
- kernel point of view iomem region looks like idle and mmap worked
- again. This matches uio_pci_generic usage.
+ The ``CONFIG_IO_STRICT_DEVMEM`` kernel option was introduced in Linux v4.5.
+
+ The issues was resolve by updating ``igb_uio`` to stop reserving PCI memory
+ resources. From the kernel point of view the iomem region looks like idle
+ and mmap works again. This matches the ``uio_pci_generic`` usage.
Drivers
@@ -234,9 +234,9 @@ Libraries
* **mbuf: Fixed refcnt update when detaching.**
- Fix the ``rte_pktmbuf_detach()`` function to decrement the direct
- mbuf's reference counter. The previous behavior was not to affect
- the reference counter. It lead a memory leak of the direct mbuf.
+ Fix the ``rte_pktmbuf_detach()`` function to decrement the direct mbuf's
+ reference counter. The previous behavior was not to affect the reference
+ counter. This lead to a memory leak of the direct mbuf.
Examples
@@ -266,9 +266,17 @@ API Changes
* Add a short 1-2 sentence description of the API change. Use fixed width
quotes for ``rte_function_names`` or ``rte_struct_names``. Use the past tense.
-* The following counters are removed from ``rte_eth_stats`` structure:
- ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
- tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* The following counters are removed from the ``rte_eth_stats`` structure:
+
+ * ``ibadcrc``
+ * ``ibadlen``
+ * ``imcasts``
+ * ``fdirmatch``
+ * ``fdirmiss``
+ * ``tx_pause_xon``
+ * ``rx_pause_xon``
+ * ``tx_pause_xoff``
+ * ``rx_pause_xoff``
* The extended statistics are fetched by ids with ``rte_eth_xstats_get``
after a lookup by name ``rte_eth_xstats_get_names``.
@@ -280,8 +288,8 @@ API Changes
``rte_vhost_avail_entries``.
* All existing vhost APIs and callbacks with ``virtio_net`` struct pointer
- as the parameter have been changed due to the ABI refactoring mentioned
- below: it's replaced by ``int vid``.
+ as the parameter have been changed due to the ABI refactoring described
+ below. It is replaced by ``int vid``.
* The function ``rte_vhost_enqueue_burst`` no longer supports concurrent enqueuing
packets to the same queue.
@@ -297,15 +305,15 @@ ABI Changes
the previous releases and made in this release. Use fixed width quotes for
``rte_function_names`` or ``rte_struct_names``. Use the past tense.
-* The ``rte_port_source_params`` structure has new fields to support PCAP file.
+* The ``rte_port_source_params`` structure has new fields to support PCAP files.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
- to support number of queues configured by software.
+ to support the number of queues configured by software.
-* vhost ABI refactoring has been made: ``virtio_net`` structure is never
- exported to application any more. Instead, a handle, ``vid``, has been
- used to represent this structure internally.
+* A Vhost ABI refactoring has been made: the ``virtio_net`` structure is no
+ longer exported directly to the application. Instead, a handle, ``vid``, has
+ been used to represent this structure internally.
Shared Library Versions
--
2.7.4
^ permalink raw reply [relevance 13%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-19 8:11 0% ` Lu, Wenzhuo
@ 2016-07-19 13:12 0% ` Adrien Mazarguil
2016-07-20 2:16 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Adrien Mazarguil @ 2016-07-19 13:12 UTC (permalink / raw)
To: Lu, Wenzhuo
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
On Tue, Jul 19, 2016 at 08:11:48AM +0000, Lu, Wenzhuo wrote:
> Hi Adrien,
> Thanks for your clarification. Most of my questions are clear, but still something may need to be discussed, comment below.
Hi Wenzhuo,
Please see below.
[...]
> > > > Requirements for a new API:
> > > >
> > > > - Flexible and extensible without causing API/ABI problems for existing
> > > > applications.
> > > > - Should be unambiguous and easy to use.
> > > > - Support existing filtering features and actions listed in `Filter types`_.
> > > > - Support packet alteration.
> > > > - In case of overlapping filters, their priority should be well documented.
> > > Does that mean we don't guarantee the consistent of priority? The priority can
> > be different on different NICs. So the behavior of the actions can be different.
> > Right?
> >
> > No, the intent is precisely to define what happens in order to get a consistent
> > result across different devices, and document cases with undefined behavior.
> > There must be no room left for interpretation.
> >
> > For example, the API must describe what happens when two overlapping filters
> > (e.g. one matching an Ethernet header, another one matching an IP header)
> > match a given packet at a given priority level.
> >
> > It is documented in section 4.1.1 (priorities) as "undefined behavior".
> > Applications remain free to do it and deal with consequences, at least they know
> > they cannot expect a consistent outcome, unless they use different priority
> > levels for both rules, see also 4.4.5 (flow rules priority).
> >
> > > Seems the users still need to aware the some details of the HW? Do we need
> > to add the negotiation for the priority?
> >
> > Priorities as defined in this document may not be directly mappable to HW
> > capabilities (e.g. HW does not support enough priorities, or that some corner
> > case make them not work as described), in which case the PMD may choose to
> > simulate priorities (again 4.4.5), as long as the end result follows the
> > specification.
> >
> > So users must not be aware of some HW details, the PMD does and must
> > perform the needed workarounds to suit their expectations. Users may only be
> > impacted by errors while attempting to create rules that are either unsupported
> > or would cause them (or existing rules) to diverge from the spec.
> The problem is sometime the priority of the filters is fixed according to
> > HW's implementation. For example, on ixgbe, n-tuple has a higher
> > priority than flow director.
As a side note I did not know that N-tuple had a higher priority than flow
director on ixgbe, priorities among filter types do not seem to be
documented at all in DPDK. This is one of the reasons I think we need a
generic API to handle flow configuration.
So, today an application cannot combine N-tuple and FDIR flow rules and get
a reliable outcome, unless it is designed for specific devices with a known
behavior.
> What's the right behavior of PMD if APP want to create a flow director rule which has a higher or even equal priority than an existing n-tuple rule? Should PMD return fail?
First remember applications only deal with the generic API, PMDs are
responsible for choosing the most appropriate HW implementation to use
according to the requested flow rules (FDIR, N-tuple or anything else).
For the specific case of FDIR vs N-tuple, if the underlying HW supports both
I do not see why the PMD would create a N-tuple rule. Doesn't FDIR support
everything N-tuple can do and much more?
Assuming such a thing happened anyway, that the PMD had to create a rule
using a high priority filter type and that the application requests the
creation of a rule that can only be done using a lower priority filter type,
but also requested a higher priority for that rule, then yes, it should
obviously fail.
That is, unless the PMD can perform some kind of workaround to have both.
> If so, do we need more fail reasons? According to this RFC, I think we need return " EEXIST: collision with an existing rule. ", but it's not very clear, APP doesn't know the problem is priority, maybe more detailed reason is helpful.
Possibly, I've defined a basic set of errors, there are quite a number of
errno values to choose from. However I think we should not define too many
values. In my opinion the basic set covers every possible failure:
- EINVAL: invalid format, rule is broken or cannot be understood by the PMD
anyhow.
- ENOTSUP: pattern/actions look fine but something in the requested rule is
not supported and thus cannot be applied.
- EEXIST: pattern/actions are fine and could have been applied if only some
other rule did not prevent the PMD to do it (I see it as the closest thing
to "ETOOBAD" which unfortunately does not exist).
- ENOMEM: like EEXIST, except it is due to the lack of resources not because
of another rule. I wasn't sure which of ENOMEM or ENOSPC was better but
settled on ENOMEM as it is well known. Still open to debate.
Errno values are only useful to get a rough idea of the reason, and another
mechanism is needed to pinpoint the exact problem for debugging/reporting
purposes, something like:
enum rte_flow_error_type {
RTE_FLOW_ERROR_TYPE_NONE,
RTE_FLOW_ERROR_TYPE_UNKNOWN,
RTE_FLOW_ERROR_TYPE_PRIORITY,
RTE_FLOW_ERROR_TYPE_PATTERN,
RTE_FLOW_ERROR_TYPE_ACTION,
};
struct rte_flow_error {
enum rte_flow_error_type type;
void *offset; /* Points to the exact pattern item or action. */
const char *message;
};
Then either provide an optional struct rte_flow_error pointer to
rte_flow_validate(), or a separate function (rte_flow_analyze()?), since
processing this may be quite expensive and applications may not care about
the exact reason.
What do you suggest?
> > > > Behavior
> > > > --------
> > > >
> > > > - API operations are synchronous and blocking (``EAGAIN`` cannot be
> > > > returned).
> > > >
> > > > - There is no provision for reentrancy/multi-thread safety, although nothing
> > > > should prevent different devices from being configured at the same
> > > > time. PMDs may protect their control path functions accordingly.
> > > >
> > > > - Stopping the data path (TX/RX) should not be necessary when managing
> > flow
> > > > rules. If this cannot be achieved naturally or with workarounds (such as
> > > > temporarily replacing the burst function pointers), an appropriate error
> > > > code must be returned (``EBUSY``).
> > > PMD cannot stop the data path without adding lock. So I think if some rules
> > cannot be applied without stopping rx/tx, PMD has to return fail.
> > > Or let the APP to stop the data path.
> >
> > Agreed, that is the intent. If the PMD cannot touch flow rules for some reason
> > even after trying really hard, then it just returns EBUSY.
> >
> > Perhaps we should write down that applications may get a different outcome
> > after stopping the data path if they get EBUSY?
> Agree, it's better to describe more about the APP. BTW, I checked the behavior of ixgbe/igb, I think we can add/delete filters during runtime. Hopefully we'll not hit too many EBUSY problems on other NICs :)
OK, I will add it.
> > > > - PMDs, not applications, are responsible for maintaining flow rules
> > > > configuration when stopping and restarting a port or performing other
> > > > actions which may affect them. They can only be destroyed explicitly.
> > > Don’t understand " They can only be destroyed explicitly."
> >
> > This part says that as long as an application has not called
> > rte_flow_destroy() on a flow rule, it never disappears, whatever happens to the
> > port (stopped, restarted). The application is not responsible for re-creating rules
> > after that.
> >
> > Note that according to the specification, this may translate to not being able to
> > stop a port as long as a flow rule is present, depending on how nice the PMD
> > intends to be with applications. Implementation can be done in small steps with
> > minimal amount of code on the PMD side.
> Does it mean PMD should store and maintain all the rules? Why not let rte do that? I think if PMD maintain all the rules, it means every kind of NIC should have a copy of code for the rules. But if rte do that, only one copy of code need to be maintained, right?
I've considered having rules stored in a common format understood at the RTE
level and not specific to each PMD and decided that the opaque rte_flow
pointer was a better choice for the following reasons:
- Even though flow rules management is done in the control path, processing
must be as fast as possible. Letting PMDs store flow rules using their own
internal representation gives them the chance to achieve better
performance.
- An opaque context managed by PMDs would probably have to be stored
somewhere as well anyway.
- PMDs may not need to allocate/store anything at all if they exclusively
rely on HW state for everything. In my opinion, the generic API has enough
constraints for this to work and maintain consistency between flow
rules. Note this is currently how most PMDs implement FDIR and other
filter types.
- RTE can (and will) provide helpers to avoid most of the code redundancy,
PMDs are free to use them or manage everything by themselves.
- Given that the opaque rte_flow pointer associated with a flow rule is to
be stored by the application, PMDs do not even have to keep references to
them.
- The flow rules format described in this specification (pattern / actions)
will be used by applications directly, and will be free to arrange them in
lists, trees or in any other way if they need to keep flow specifications
around for further processing.
> When the port is stopped and restarted, rte can reconfigure the rules. Is the concern that PMD may adjust the sequence of the rules according to the priority, so every NIC has a different list of rules? But PMD can adjust them again when rte reconfiguring the rules.
What about PMDs able to stop and restart ports without destroying their own
flow rules? If we assume flow rules must be destroyed when stopping a port,
these PMDs are needlessly penalized with slower stop/start cycles. Think
about it assuming thousands of flow rules.
Thus from an application point of view, whatever happens when stopping and
restarting a port should not matter. If a flow rule was present before, it
must still be present afterwards. If the PMD had to destroy flow rules and
re-create them, it does not actually matter if they differ slightly at the
HW level, as long as:
- Existing opaque flow rule pointers (rte_flow) are still valid to the PMD
and refer to the same rules.
- The overall behavior of all rules is the same.
The list of rules you think of (patterns / actions) is maintained by
applications (not RTE), and only if they need them. RTE would needlessly
duplicate this.
--
Adrien Mazarguil
6WIND
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-07 10:26 2% ` Adrien Mazarguil
@ 2016-07-19 8:11 0% ` Lu, Wenzhuo
2016-07-19 13:12 0% ` Adrien Mazarguil
0 siblings, 1 reply; 200+ results
From: Lu, Wenzhuo @ 2016-07-19 8:11 UTC (permalink / raw)
To: Adrien Mazarguil
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
Hi Adrien,
Thanks for your clarification. Most of my questions are clear, but still something may need to be discussed, comment below.
> -----Original Message-----
> From: Adrien Mazarguil [mailto:adrien.mazarguil@6wind.com]
> Sent: Thursday, July 7, 2016 6:27 PM
> To: Lu, Wenzhuo
> Cc: dev@dpdk.org; Thomas Monjalon; Zhang, Helin; Wu, Jingjing; Rasesh Mody;
> Ajit Khaparde; Rahul Lakkireddy; Jan Medala; John Daley; Chen, Jing D; Ananyev,
> Konstantin; Matej Vido; Alejandro Lucero; Sony Chacko; Jerin Jacob; De Lara
> Guarch, Pablo; Olga Shern
> Subject: Re: [RFC] Generic flow director/filtering/classification API
>
> Hi Lu Wenzhuo,
>
> Thanks for your feedback, I'm replying below as well.
>
> On Thu, Jul 07, 2016 at 07:14:18AM +0000, Lu, Wenzhuo wrote:
> > Hi Adrien,
> > I have some questions, please see inline, thanks.
> >
> > > -----Original Message-----
> > > From: Adrien Mazarguil [mailto:adrien.mazarguil@6wind.com]
> > > Sent: Wednesday, July 6, 2016 2:17 AM
> > > To: dev@dpdk.org
> > > Cc: Thomas Monjalon; Zhang, Helin; Wu, Jingjing; Rasesh Mody; Ajit
> > > Khaparde; Rahul Lakkireddy; Lu, Wenzhuo; Jan Medala; John Daley;
> > > Chen, Jing D; Ananyev, Konstantin; Matej Vido; Alejandro Lucero;
> > > Sony Chacko; Jerin Jacob; De Lara Guarch, Pablo; Olga Shern
> > > Subject: [RFC] Generic flow director/filtering/classification API
> > >
> > >
> > > Requirements for a new API:
> > >
> > > - Flexible and extensible without causing API/ABI problems for existing
> > > applications.
> > > - Should be unambiguous and easy to use.
> > > - Support existing filtering features and actions listed in `Filter types`_.
> > > - Support packet alteration.
> > > - In case of overlapping filters, their priority should be well documented.
> > Does that mean we don't guarantee the consistent of priority? The priority can
> be different on different NICs. So the behavior of the actions can be different.
> Right?
>
> No, the intent is precisely to define what happens in order to get a consistent
> result across different devices, and document cases with undefined behavior.
> There must be no room left for interpretation.
>
> For example, the API must describe what happens when two overlapping filters
> (e.g. one matching an Ethernet header, another one matching an IP header)
> match a given packet at a given priority level.
>
> It is documented in section 4.1.1 (priorities) as "undefined behavior".
> Applications remain free to do it and deal with consequences, at least they know
> they cannot expect a consistent outcome, unless they use different priority
> levels for both rules, see also 4.4.5 (flow rules priority).
>
> > Seems the users still need to aware the some details of the HW? Do we need
> to add the negotiation for the priority?
>
> Priorities as defined in this document may not be directly mappable to HW
> capabilities (e.g. HW does not support enough priorities, or that some corner
> case make them not work as described), in which case the PMD may choose to
> simulate priorities (again 4.4.5), as long as the end result follows the
> specification.
>
> So users must not be aware of some HW details, the PMD does and must
> perform the needed workarounds to suit their expectations. Users may only be
> impacted by errors while attempting to create rules that are either unsupported
> or would cause them (or existing rules) to diverge from the spec.
The problem is sometime the priority of the filters is fixed according to HW's implementation. For example, on ixgbe, n-tuple has a higher priority than flow director. What's the right behavior of PMD if APP want to create a flow director rule which has a higher or even equal priority than an existing n-tuple rule? Should PMD return fail?
If so, do we need more fail reasons? According to this RFC, I think we need return " EEXIST: collision with an existing rule. ", but it's not very clear, APP doesn't know the problem is priority, maybe more detailed reason is helpful.
> > > Behavior
> > > --------
> > >
> > > - API operations are synchronous and blocking (``EAGAIN`` cannot be
> > > returned).
> > >
> > > - There is no provision for reentrancy/multi-thread safety, although nothing
> > > should prevent different devices from being configured at the same
> > > time. PMDs may protect their control path functions accordingly.
> > >
> > > - Stopping the data path (TX/RX) should not be necessary when managing
> flow
> > > rules. If this cannot be achieved naturally or with workarounds (such as
> > > temporarily replacing the burst function pointers), an appropriate error
> > > code must be returned (``EBUSY``).
> > PMD cannot stop the data path without adding lock. So I think if some rules
> cannot be applied without stopping rx/tx, PMD has to return fail.
> > Or let the APP to stop the data path.
>
> Agreed, that is the intent. If the PMD cannot touch flow rules for some reason
> even after trying really hard, then it just returns EBUSY.
>
> Perhaps we should write down that applications may get a different outcome
> after stopping the data path if they get EBUSY?
Agree, it's better to describe more about the APP. BTW, I checked the behavior of ixgbe/igb, I think we can add/delete filters during runtime. Hopefully we'll not hit too many EBUSY problems on other NICs :)
>
> > > - PMDs, not applications, are responsible for maintaining flow rules
> > > configuration when stopping and restarting a port or performing other
> > > actions which may affect them. They can only be destroyed explicitly.
> > Don’t understand " They can only be destroyed explicitly."
>
> This part says that as long as an application has not called
> rte_flow_destroy() on a flow rule, it never disappears, whatever happens to the
> port (stopped, restarted). The application is not responsible for re-creating rules
> after that.
>
> Note that according to the specification, this may translate to not being able to
> stop a port as long as a flow rule is present, depending on how nice the PMD
> intends to be with applications. Implementation can be done in small steps with
> minimal amount of code on the PMD side.
Does it mean PMD should store and maintain all the rules? Why not let rte do that? I think if PMD maintain all the rules, it means every kind of NIC should have a copy of code for the rules. But if rte do that, only one copy of code need to be maintained, right?
When the port is stopped and restarted, rte can reconfigure the rules. Is the concern that PMD may adjust the sequence of the rules according to the priority, so every NIC has a different list of rules? But PMD can adjust them again when rte reconfiguring the rules.
>
> --
> Adrien Mazarguil
> 6WIND
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v6 04/17] eal: remove duplicate function declaration
2016-07-12 6:01 3% ` [dpdk-dev] [PATCH v6 04/17] eal: remove duplicate function declaration Shreyansh Jain
@ 2016-07-14 17:13 0% ` viktorin
0 siblings, 0 replies; 200+ results
From: viktorin @ 2016-07-14 17:13 UTC (permalink / raw)
To: Shreyansh Jain; +Cc: dev, thomas.monjalon, david.marchand
On Tue, 12 Jul 2016 11:31:09 +0530
Shreyansh Jain <shreyansh.jain@nxp.com> wrote:
> rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
> introduction.
> This function has been exported in ABI, so remove it from eal_private.h
>
> Fixes: e57f20e05177 ("eal: make vdev init path generic for both virtual and pci devices")
> Signed-off-by: David Marchand <david.marchand@6wind.com>
> Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
> ---
Reviewed-by: Jan Viktorin <viktorin@rehivetech.com>
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] rte_ether: Driver-specific stats getting overwritten
@ 2016-07-14 15:50 3% ` Remy Horton
0 siblings, 0 replies; 200+ results
From: Remy Horton @ 2016-07-14 15:50 UTC (permalink / raw)
To: Igor Ryzhov, Thomas Monjalon; +Cc: dev
On 14/07/2016 14:51, Igor Ryzhov wrote:
[..]
> How about deleting rx_nombuf from rte_eth_stats? Do you think this
> counter is necessary? It just shows enormous numbers in case of a
> lack of processing speed. But we already have imissed counter which
> shows real number of packets, dropped for the same reason.
Deleting it has API/ABI breakage issues. There is also lack of
consistency between drivers as to what imissed includes, as some don't
implement it at all whereas others include filtered packets as well.
>> 14 июля 2016 г., в 16:37, Thomas Monjalon
>> <thomas.monjalon@6wind.com> написал(а):
>>
[..]
>> Yes it is strange and has always been like that. Why not moving the
>> assignment before calling the driver callback?
Think I'll do that. Easier than updating all the drivers that don't fill
it in..
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v6 04/17] eal: remove duplicate function declaration
@ 2016-07-12 6:01 3% ` Shreyansh Jain
2016-07-14 17:13 0% ` viktorin
0 siblings, 1 reply; 200+ results
From: Shreyansh Jain @ 2016-07-12 6:01 UTC (permalink / raw)
To: dev; +Cc: viktorin, thomas.monjalon, david.marchand
rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
introduction.
This function has been exported in ABI, so remove it from eal_private.h
Fixes: e57f20e05177 ("eal: make vdev init path generic for both virtual and pci devices")
Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
---
lib/librte_eal/common/eal_private.h | 7 -------
lib/librte_eal/linuxapp/eal/eal.c | 1 +
2 files changed, 1 insertion(+), 7 deletions(-)
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 857dc3e..06a68f6 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -259,13 +259,6 @@ int rte_eal_intr_init(void);
int rte_eal_alarm_init(void);
/**
- * This function initialises any virtual devices
- *
- * This function is private to the EAL.
- */
-int rte_eal_dev_init(void);
-
-/**
* Function is to check if the kernel module(like, vfio, vfio_iommu_type1,
* etc.) loaded.
*
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index 3fb2188..fe9c704 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -70,6 +70,7 @@
#include <rte_cpuflags.h>
#include <rte_interrupts.h>
#include <rte_pci.h>
+#include <rte_dev.h>
#include <rte_devargs.h>
#include <rte_common.h>
#include <rte_version.h>
--
2.7.4
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] librte_pmd_bond: fix exported symbol versioning
2016-07-11 11:27 3% ` [dpdk-dev] [PATCH v2] " Christian Ehrhardt
@ 2016-07-11 12:58 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-11 12:58 UTC (permalink / raw)
To: Christian Ehrhardt; +Cc: Eric Kinzie, dev
2016-07-11 13:27, Christian Ehrhardt:
> *update in v2*
> - add missing changes in rte_eth_bond_8023ad.h
>
> The older versions of rte_eth_bond_8023ad_conf_get and
> rte_eth_bond_8023ad_setup were available in the old way since 2.0 - at
> least according to the map file.
>
> But versioning in the code was set to 16.04.
> That breaks compatibility checks for 2.0 on that library.
>
> For example with the dpdk abi checker:
> http://people.canonical.com/~paelzer/compat_report.html
>
> To fix, version the old symbols on the 2.0 version as they were
> initially added to the map file.
>
> See http://people.canonical.com/~paelzer/compat_report.html
>
> Fixes: dc40f17a ("net/bonding: allow external state machine in mode 4")
>
> Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
Applied, thanks
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v2] librte_pmd_bond: fix exported symbol versioning
2016-07-06 11:39 3% [dpdk-dev] [PATCH] librte_pmd_bond: fix exported symbol versioning Christian Ehrhardt
@ 2016-07-11 11:27 3% ` Christian Ehrhardt
2016-07-11 12:58 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Christian Ehrhardt @ 2016-07-11 11:27 UTC (permalink / raw)
To: Eric Kinzie, christian.ehrhardt, thomas.monjalon, dev
*update in v2*
- add missing changes in rte_eth_bond_8023ad.h
The older versions of rte_eth_bond_8023ad_conf_get and
rte_eth_bond_8023ad_setup were available in the old way since 2.0 - at
least according to the map file.
But versioning in the code was set to 16.04.
That breaks compatibility checks for 2.0 on that library.
For example with the dpdk abi checker:
http://people.canonical.com/~paelzer/compat_report.html
To fix, version the old symbols on the 2.0 version as they were
initially added to the map file.
See http://people.canonical.com/~paelzer/compat_report.html
Fixes: dc40f17a ("net/bonding: allow external state machine in mode 4")
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
---
drivers/net/bonding/rte_eth_bond_8023ad.c | 12 ++++++------
drivers/net/bonding/rte_eth_bond_8023ad.h | 4 ++--
2 files changed, 8 insertions(+), 8 deletions(-)
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index 48a50e4..2f7ae70 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -1068,7 +1068,7 @@ bond_mode_8023ad_conf_assign(struct mode8023ad_private *mode4,
}
static void
-bond_mode_8023ad_setup_v1604(struct rte_eth_dev *dev,
+bond_mode_8023ad_setup_v20(struct rte_eth_dev *dev,
struct rte_eth_bond_8023ad_conf *conf)
{
struct rte_eth_bond_8023ad_conf def_conf;
@@ -1214,7 +1214,7 @@ free_out:
}
int
-rte_eth_bond_8023ad_conf_get_v1604(uint8_t port_id,
+rte_eth_bond_8023ad_conf_get_v20(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf)
{
struct rte_eth_dev *bond_dev;
@@ -1229,7 +1229,7 @@ rte_eth_bond_8023ad_conf_get_v1604(uint8_t port_id,
bond_mode_8023ad_conf_get(bond_dev, conf);
return 0;
}
-VERSION_SYMBOL(rte_eth_bond_8023ad_conf_get, _v1604, 16.04);
+VERSION_SYMBOL(rte_eth_bond_8023ad_conf_get, _v20, 2.0);
int
rte_eth_bond_8023ad_conf_get_v1607(uint8_t port_id,
@@ -1278,7 +1278,7 @@ bond_8023ad_setup_validate(uint8_t port_id,
}
int
-rte_eth_bond_8023ad_setup_v1604(uint8_t port_id,
+rte_eth_bond_8023ad_setup_v20(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf)
{
struct rte_eth_dev *bond_dev;
@@ -1289,11 +1289,11 @@ rte_eth_bond_8023ad_setup_v1604(uint8_t port_id,
return err;
bond_dev = &rte_eth_devices[port_id];
- bond_mode_8023ad_setup_v1604(bond_dev, conf);
+ bond_mode_8023ad_setup_v20(bond_dev, conf);
return 0;
}
-VERSION_SYMBOL(rte_eth_bond_8023ad_setup, _v1604, 16.04);
+VERSION_SYMBOL(rte_eth_bond_8023ad_setup, _v20, 2.0);
int
rte_eth_bond_8023ad_setup_v1607(uint8_t port_id,
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.h b/drivers/net/bonding/rte_eth_bond_8023ad.h
index 1de34bc..6b8ff57 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.h
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.h
@@ -188,7 +188,7 @@ int
rte_eth_bond_8023ad_conf_get(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf);
int
-rte_eth_bond_8023ad_conf_get_v1604(uint8_t port_id,
+rte_eth_bond_8023ad_conf_get_v20(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf);
int
rte_eth_bond_8023ad_conf_get_v1607(uint8_t port_id,
@@ -209,7 +209,7 @@ int
rte_eth_bond_8023ad_setup(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf);
int
-rte_eth_bond_8023ad_setup_v1604(uint8_t port_id,
+rte_eth_bond_8023ad_setup_v20(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf);
int
rte_eth_bond_8023ad_setup_v1607(uint8_t port_id,
--
2.7.4
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH] cryptodev: move new cryptodev type to bottom of enum
2016-07-06 14:05 3% [dpdk-dev] [PATCH] cryptodev: move new cryptodev type to bottom of enum Pablo de Lara
@ 2016-07-08 17:52 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-08 17:52 UTC (permalink / raw)
To: Pablo de Lara; +Cc: dev, declan.doherty
2016-07-06 15:05, Pablo de Lara:
> New cryptodev type for the new KASUMI PMD was added
> in the cryptodev type enum, but not at the end of it,
> causing an ABI breakage.
>
> Fixes: 2773c86d061a ("crypto/kasumi: add driver for KASUMI library")
>
> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> Reported-by: Ferruh Yigit <ferruh.yigit@intel.com>
Applied, thanks
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v3 10/10] maintainers: add section for pmdinfo
@ 2016-07-08 14:42 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-08 14:42 UTC (permalink / raw)
To: Neil Horman; +Cc: dev
The author of this feature is Neil Horman.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
---
MAINTAINERS | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index a59191e..f996c2e 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -68,6 +68,10 @@ F: lib/librte_compat/
F: doc/guides/rel_notes/deprecation.rst
F: scripts/validate-abi.sh
+Driver information
+F: buildtools/pmdinfogen/
+F: tools/pmdinfo.py
+
Environment Abstraction Layer
-----------------------------
--
2.7.0
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-07 23:15 0% ` Chandran, Sugesh
@ 2016-07-08 13:03 0% ` Adrien Mazarguil
0 siblings, 0 replies; 200+ results
From: Adrien Mazarguil @ 2016-07-08 13:03 UTC (permalink / raw)
To: Chandran, Sugesh
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Lu, Wenzhuo, Jan Medala,
John Daley, Chen, Jing D, Ananyev, Konstantin, Matej Vido,
Alejandro Lucero, Sony Chacko, Jerin Jacob, De Lara Guarch,
Pablo, Olga Shern
Hi Sugesh,
On Thu, Jul 07, 2016 at 11:15:07PM +0000, Chandran, Sugesh wrote:
> Hi Adrien,
>
> Thank you for proposing this. It would be really useful for application such as OVS-DPDK.
> Please find my comments and questions inline below prefixed with [Sugesh]. Most of them are from the perspective of enabling these APIs in application such as OVS-DPDK.
Thanks, I'm replying below.
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Adrien Mazarguil
> > Sent: Tuesday, July 5, 2016 7:17 PM
> > To: dev@dpdk.org
> > Cc: Thomas Monjalon <thomas.monjalon@6wind.com>; Zhang, Helin
> > <helin.zhang@intel.com>; Wu, Jingjing <jingjing.wu@intel.com>; Rasesh
> > Mody <rasesh.mody@qlogic.com>; Ajit Khaparde
> > <ajit.khaparde@broadcom.com>; Rahul Lakkireddy
> > <rahul.lakkireddy@chelsio.com>; Lu, Wenzhuo <wenzhuo.lu@intel.com>;
> > Jan Medala <jan@semihalf.com>; John Daley <johndale@cisco.com>; Chen,
> > Jing D <jing.d.chen@intel.com>; Ananyev, Konstantin
> > <konstantin.ananyev@intel.com>; Matej Vido <matejvido@gmail.com>;
> > Alejandro Lucero <alejandro.lucero@netronome.com>; Sony Chacko
> > <sony.chacko@qlogic.com>; Jerin Jacob
> > <jerin.jacob@caviumnetworks.com>; De Lara Guarch, Pablo
> > <pablo.de.lara.guarch@intel.com>; Olga Shern <olgas@mellanox.com>
> > Subject: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
> >
> > Hi All,
> >
> > First, forgive me for this large message, I know our mailboxes already
> > suffer quite a bit from the amount of traffic on this ML.
> >
> > This is not exactly yet another thread about how flow director should be
> > extended, rather about a brand new API to handle filtering and
> > classification for incoming packets in the most PMD-generic and
> > application-friendly fashion we can come up with. Reasons described below.
> >
> > I think this topic is important enough to include both the users of this API
> > as well as PMD maintainers. So far I have CC'ed librte_ether (especially
> > rte_eth_ctrl.h contributors), testpmd and PMD maintainers (with and
> > without
> > a .filter_ctrl implementation), but if you know application maintainers
> > other than testpmd who use FDIR or might be interested in this discussion,
> > feel free to add them.
> >
> > The issues we found with the current approach are already summarized in
> > the
> > following document, but here is a quick summary for TL;DR folks:
> >
> > - PMDs do not expose a common set of filter types and even when they do,
> > their behavior more or less differs.
> >
> > - Applications need to determine and adapt to device-specific limitations
> > and quirks on their own, without help from PMDs.
> >
> > - Writing an application that creates flow rules targeting all devices
> > supported by DPDK is thus difficult, if not impossible.
> >
> > - The current API has too many unspecified areas (particularly regarding
> > side effects of flow rules) that make PMD implementation tricky.
> >
> > This RFC API handles everything currently supported by .filter_ctrl, the
> > idea being to reimplement all of these to make them fully usable by
> > applications in a more generic and well defined fashion. It has a very small
> > set of mandatory features and an easy method to let applications probe for
> > supported capabilities.
> >
> > The only downside is more work for the software control side of PMDs
> > because
> > they have to adapt to the API instead of the reverse. I think helpers can be
> > added to EAL to assist with this.
> >
> > HTML version:
> >
> > https://rawgit.com/6WIND/rte_flow/master/rte_flow.html
> >
> > PDF version:
> >
> > https://rawgit.com/6WIND/rte_flow/master/rte_flow.pdf
> >
> > Related draft header file (for reference while reading the specification):
> >
> > https://raw.githubusercontent.com/6WIND/rte_flow/master/rte_flow.h
> >
> > Git tree for completeness (latest .rst version can be retrieved from here):
> >
> > https://github.com/6WIND/rte_flow
> >
> > What follows is the ReST source of the above, for inline comments and
> > discussion. I intend to update that specification accordingly.
> >
> > ========================
> > Generic filter interface
> > ========================
> >
> > .. footer::
> >
> > v0.6
> >
> > .. contents::
> > .. sectnum::
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Overview
> > ========
> >
> > DPDK provides several competing interfaces added over time to perform
> > packet
> > matching and related actions such as filtering and classification.
> >
> > They must be extended to implement the features supported by newer
> > devices
> > in order to expose them to applications, however the current design has
> > several drawbacks:
> >
> > - Complicated filter combinations which have not been hard-coded cannot be
> > expressed.
> > - Prone to API/ABI breakage when new features must be added to an
> > existing
> > filter type, which frequently happens.
> >
> > From an application point of view:
> >
> > - Having disparate interfaces, all optional and lacking in features does not
> > make this API easy to use.
> > - Seemingly arbitrary built-in limitations of filter types based on the
> > device they were initially designed for.
> > - Undefined relationship between different filter types.
> > - High complexity, considerable undocumented and/or undefined behavior.
> >
> > Considering the growing number of devices supported by DPDK, adding a
> > new
> > filter type each time a new feature must be implemented is not sustainable
> > in the long term. Applications not written to target a specific device
> > cannot really benefit from such an API.
> >
> > For these reasons, this document defines an extensible unified API that
> > encompasses and supersedes these legacy filter types.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Current API
> > ===========
> >
> > Rationale
> > ---------
> >
> > The reason several competing (and mostly overlapping) filtering APIs are
> > present in DPDK is due to its nature as a thin layer between hardware and
> > software.
> >
> > Each subsequent interface has been added to better match the capabilities
> > and limitations of the latest supported device, which usually happened to
> > need an incompatible configuration approach. Because of this, many ended
> > up
> > device-centric and not usable by applications that were not written for that
> > particular device.
> >
> > This document is not the first attempt to address this proliferation issue,
> > in fact a lot of work has already been done both to create a more generic
> > interface while somewhat keeping compatibility with legacy ones through a
> > common call interface (``rte_eth_dev_filter_ctrl()`` with the
> > ``.filter_ctrl`` PMD callback in ``rte_ethdev.h``).
> >
> > Today, these previously incompatible interfaces are known as filter types
> > (``RTE_ETH_FILTER_*`` from ``enum rte_filter_type`` in ``rte_eth_ctrl.h``).
> >
> > However while trivial to extend with new types, it only shifted the
> > underlying problem as applications still need to be written for one kind of
> > filter type, which, as described in the following sections, is not
> > necessarily implemented by all PMDs that support filtering.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Filter types
> > ------------
> >
> > This section summarizes the capabilities of each filter type.
> >
> > Although the following list is exhaustive, the description of individual
> > types may contain inaccuracies due to the lack of documentation or usage
> > examples.
> >
> > Note: names are prefixed with ``RTE_ETH_FILTER_``.
> >
> > ``MACVLAN``
> > ~~~~~~~~~~~
> >
> > Matching:
> >
> > - L2 source/destination addresses.
> > - Optional 802.1Q VLAN ID.
> > - Masking individual fields on a rule basis is not supported.
> >
> > Action:
> >
> > - Packets are redirected either to a given VF device using its ID or to the
> > PF.
> >
> > ``ETHERTYPE``
> > ~~~~~~~~~~~~~
> >
> > Matching:
> >
> > - L2 source/destination addresses (optional).
> > - Ethertype (no VLAN ID?).
> > - Masking individual fields on a rule basis is not supported.
> >
> > Action:
> >
> > - Receive packets on a given queue.
> > - Drop packets.
> >
> > ``FLEXIBLE``
> > ~~~~~~~~~~~~
> >
> > Matching:
> >
> > - At most 128 consecutive bytes anywhere in packets.
> > - Masking is supported with byte granularity.
> > - Priorities are supported (relative to this filter type, undefined
> > otherwise).
> >
> > Action:
> >
> > - Receive packets on a given queue.
> >
> > ``SYN``
> > ~~~~~~~
> >
> > Matching:
> >
> > - TCP SYN packets only.
> > - One high priority bit can be set to give the highest possible priority to
> > this type when other filters with different types are configured.
> >
> > Action:
> >
> > - Receive packets on a given queue.
> >
> > ``NTUPLE``
> > ~~~~~~~~~~
> >
> > Matching:
> >
> > - Source/destination IPv4 addresses (optional in 2-tuple mode).
> > - Source/destination TCP/UDP port (mandatory in 2 and 5-tuple modes).
> > - L4 protocol (2 and 5-tuple modes).
> > - Masking individual fields is supported.
> > - TCP flags.
> > - Up to 7 levels of priority relative to this filter type, undefined
> > otherwise.
> > - No IPv6.
> >
> > Action:
> >
> > - Receive packets on a given queue.
> >
> > ``TUNNEL``
> > ~~~~~~~~~~
> >
> > Matching:
> >
> > - Outer L2 source/destination addresses.
> > - Inner L2 source/destination addresses.
> > - Inner VLAN ID.
> > - IPv4/IPv6 source (destination?) address.
> > - Tunnel type to match (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE,
> > 802.1BR
> > E-Tag).
> > - Tenant ID for tunneling protocols that have one.
> > - Any combination of the above can be specified.
> > - Masking individual fields on a rule basis is not supported.
> >
> > Action:
> >
> > - Receive packets on a given queue.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``FDIR``
> > ~~~~~~~~
> >
> > Queries:
> >
> > - Device capabilities and limitations.
> > - Device statistics about configured filters (resource usage, collisions).
> > - Device configuration (matching input set and masks)
> >
> > Matching:
> >
> > - Device mode of operation: none (to disable filtering), signature
> > (hash-based dispatching from masked fields) or perfect (either MAC VLAN
> > or
> > tunnel).
> > - L2 Ethertype.
> > - Outer L2 destination address (MAC VLAN mode).
> > - Inner L2 destination address, tunnel type (NVGRE, VXLAN) and tunnel ID
> > (tunnel mode).
> > - IPv4 source/destination addresses, ToS, TTL and protocol fields.
> > - IPv6 source/destination addresses, TC, protocol and hop limits fields.
> > - UDP source/destination IPv4/IPv6 and ports.
> > - TCP source/destination IPv4/IPv6 and ports.
> > - SCTP source/destination IPv4/IPv6, ports and verification tag field.
> > - Note, only one protocol type at once (either only L2 Ethertype, basic
> > IPv6, IPv4+UDP, IPv4+TCP and so on).
> > - VLAN TCI (extended API).
> > - At most 16 bytes to match in payload (extended API). A global device
> > look-up table specifies for each possible protocol layer (unknown, raw,
> > L2, L3, L4) the offset to use for each byte (they do not need to be
> > contiguous) and the related bitmask.
> > - Whether packet is addressed to PF or VF, in that case its ID can be
> > matched as well (extended API).
> > - Masking most of the above fields is supported, but simultaneously affects
> > all filters configured on a device.
> > - Input set can be modified in a similar fashion for a given device to
> > ignore individual fields of filters (i.e. do not match the destination
> > address in a IPv4 filter, refer to **RTE_ETH_INPUT_SET_**
> > macros). Configuring this also affects RSS processing on **i40e**.
> > - Filters can also provide 32 bits of arbitrary data to return as part of
> > matched packets.
> >
> > Action:
> >
> > - **RTE_ETH_FDIR_ACCEPT**: receive (accept) packet on a given queue.
> > - **RTE_ETH_FDIR_REJECT**: drop packet immediately.
> > - **RTE_ETH_FDIR_PASSTHRU**: similar to accept for the last filter in list,
> > otherwise process it with subsequent filters.
> > - For accepted packets and if requested by filter, either 32 bits of
> > arbitrary data and four bytes of matched payload (only in case of flex
> > bytes matching), or eight bytes of matched payload (flex also) are added
> > to meta data.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``HASH``
> > ~~~~~~~~
> >
> > Not an actual filter type. Provides and retrieves the global device
> > configuration (per port or entire NIC) for hash functions and their
> > properties.
> >
> > Hash function selection: "default" (keep current), XOR or Toeplitz.
> >
> > This function can be configured per flow type (**RTE_ETH_FLOW_**
> > definitions), supported types are:
> >
> > - Unknown.
> > - Raw.
> > - Fragmented or non-fragmented IPv4.
> > - Non-fragmented IPv4 with L4 (TCP, UDP, SCTP or other).
> > - Fragmented or non-fragmented IPv6.
> > - Non-fragmented IPv6 with L4 (TCP, UDP, SCTP or other).
> > - L2 payload.
> > - IPv6 with extensions.
> > - IPv6 with L4 (TCP, UDP) and extensions.
> >
> > ``L2_TUNNEL``
> > ~~~~~~~~~~~~~
> >
> > Matching:
> >
> > - All packets received on a given port.
> >
> > Action:
> >
> > - Add tunnel encapsulation (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE,
> > 802.1BR E-Tag) using the provided Ethertype and tunnel ID (only E-Tag
> > is implemented at the moment).
> > - VF ID to use for tag insertion (currently unused).
> > - Destination pool for tag based forwarding (pools are IDs that can be
> > affected to ports, duplication occurs if the same ID is shared by several
> > ports of the same NIC).
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Driver support
> > --------------
> >
> > ======== ======= ========= ======== === ====== ====== ==== ====
> > =========
> > Driver MACVLAN ETHERTYPE FLEXIBLE SYN NTUPLE TUNNEL FDIR HASH
> > L2_TUNNEL
> > ======== ======= ========= ======== === ====== ====== ==== ====
> > =========
> > bnx2x
> > cxgbe
> > e1000 yes yes yes yes
> > ena
> > enic yes
> > fm10k
> > i40e yes yes yes yes yes
> > ixgbe yes yes yes yes yes
> > mlx4
> > mlx5 yes
> > szedata2
> > ======== ======= ========= ======== === ====== ====== ==== ====
> > =========
> >
> > Flow director
> > -------------
> >
> > Flow director (FDIR) is the name of the most capable filter type, which
> > covers most features offered by others. As such, it is the most widespread
> > in PMDs that support filtering (i.e. all of them besides **e1000**).
> >
> > It is also the only type that allows an arbitrary 32 bits value provided by
> > applications to be attached to a filter and returned with matching packets
> > instead of relying on the destination queue to recognize flows.
> >
> > Unfortunately, even FDIR requires applications to be aware of low-level
> > capabilities and limitations (most of which come directly from **ixgbe** and
> > **i40e**):
> >
> > - Bitmasks are set globally per device (port?), not per filter.
> [Sugesh] This means application cannot define filters that matches on arbitrary different offsets?
> If that’s the case, I assume the application has to program bitmask in advance. Otherwise how
> the API framework deduce this bitmask information from the rules?? Its not very clear to me
> that how application pass down the bitmask information for multiple filters on same port?
This is my understanding of how flow director currently works, perhaps
someome more familiar with it can answer this question better than I could.
Let me take an example, if particular device can only handle a single IPv4
mask common to all flow rules (say only to match destination addresses),
updating that mask to also match the source address affects all defined and
future flow rules simultaneously.
That is how FDIR currently works and I think it is wrong, as it penalizes
devices that do support individual bit-masks per rule, and is a little
awkward from an application point of view.
What I suggest for the new API instead is the ability to specify one
bit-mask per rule, and let the PMD deal with HW limitations by automatically
configuring global bitmasks from the first added rule, then refusing to add
subsequent rules if they specify a conflicting bit-mask. Existing rules
remain unaffected that way, and applications do not have to be extra
cautious.
> > - Configuration state is not expected to be saved by the driver, and
> > stopping/restarting a port requires the application to perform it again
> > (API documentation is also unclear about this).
> > - Monolithic approach with ABI issues as soon as a new kind of flow or
> > combination needs to be supported.
> > - Cryptic global statistics/counters.
> > - Unclear about how priorities are managed; filters seem to be arranged as a
> > linked list in hardware (possibly related to configuration order).
> >
> > Packet alteration
> > -----------------
> >
> > One interesting feature is that the L2 tunnel filter type implements the
> > ability to alter incoming packets through a filter (in this case to
> > encapsulate them), thus the **mlx5** flow encap/decap features are not a
> > foreign concept.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Proposed API
> > ============
> >
> > Terminology
> > -----------
> >
> > - **Filtering API**: overall framework affecting the fate of selected
> > packets, covers everything described in this document.
> > - **Matching pattern**: properties to look for in received packets, a
> > combination of any number of items.
> > - **Pattern item**: part of a pattern that either matches packet data
> > (protocol header, payload or derived information), or specifies properties
> > of the pattern itself.
> > - **Actions**: what needs to be done when a packet matches a pattern.
> > - **Flow rule**: this is the result of combining a *matching pattern* with
> > *actions*.
> > - **Filter rule**: a less generic term than *flow rule*, can otherwise be
> > used interchangeably.
> > - **Hit**: a flow rule is said to be *hit* when processing a matching
> > packet.
> >
> > Requirements
> > ------------
> >
> > As described in the previous section, there is a growing need for a common
> > method to configure filtering and related actions in a hardware independent
> > fashion.
> >
> > The filtering API should not disallow any filter combination by design and
> > must remain as simple as possible to use. It can simply be defined as a
> > method to perform one or several actions on selected packets.
> >
> > PMDs are aware of the capabilities of the device they manage and should be
> > responsible for preventing unsupported or conflicting combinations.
> >
> > This approach is fundamentally different as it places most of the burden on
> > the software side of the PMD instead of having device capabilities directly
> > mapped to API functions, then expecting applications to work around
> > ensuing
> > compatibility issues.
> >
> > Requirements for a new API:
> >
> > - Flexible and extensible without causing API/ABI problems for existing
> > applications.
> > - Should be unambiguous and easy to use.
> > - Support existing filtering features and actions listed in `Filter types`_.
> > - Support packet alteration.
> > - In case of overlapping filters, their priority should be well documented.
> > - Support filter queries (for example to retrieve counters).
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > High level design
> > -----------------
> >
> > The chosen approach to make filtering as generic as possible is by
> > expressing matching patterns through lists of items instead of the flat
> > structures used in DPDK today, enabling combinations that are not
> > predefined
> > and thus being more versatile.
> >
> > Flow rules can have several distinct actions (such as counting,
> > encapsulating, decapsulating before redirecting packets to a particular
> > queue, etc.), instead of relying on several rules to achieve this and having
> > applications deal with hardware implementation details regarding their
> > order.
> >
> > Support for different priority levels on a rule basis is provided, for
> > example in order to force a more specific rule come before a more generic
> > one for packets matched by both, however hardware support for more than
> > a
> > single priority level cannot be guaranteed. When supported, the number of
> > available priority levels is usually low, which is why they can also be
> > implemented in software by PMDs (e.g. to simulate missing priority levels by
> > reordering rules).
> >
> > In order to remain as hardware agnostic as possible, by default all rules
> > are considered to have the same priority, which means that the order
> > between
> > overlapping rules (when a packet is matched by several filters) is
> > undefined, packet duplication may even occur as a result.
> >
> > PMDs may refuse to create overlapping rules at a given priority level when
> > they can be detected (e.g. if a pattern matches an existing filter).
> >
> > Thus predictable results for a given priority level can only be achieved
> > with non-overlapping rules, using perfect matching on all protocol layers.
> >
> > Support for multiple actions per rule may be implemented internally on top
> > of non-default hardware priorities, as a result both features may not be
> > simultaneously available to applications.
> >
> > Considering that allowed pattern/actions combinations cannot be known in
> > advance and would result in an unpractically large number of capabilities to
> > expose, a method is provided to validate a given rule from the current
> > device configuration state without actually adding it (akin to a "dry run"
> > mode).
> >
> > This enables applications to check if the rule types they need is supported
> > at initialization time, before starting their data path. This method can be
> > used anytime, its only requirement being that the resources needed by a
> > rule
> > must exist (e.g. a target RX queue must be configured first).
> >
> > Each defined rule is associated with an opaque handle managed by the PMD,
> > applications are responsible for keeping it. These can be used for queries
> > and rules management, such as retrieving counters or other data and
> > destroying them.
> >
> > Handles must be destroyed before releasing associated resources such as
> > queues.
> >
> > Integration
> > -----------
> >
> > To avoid ABI breakage, this new interface will be implemented through the
> > existing filtering control framework (``rte_eth_dev_filter_ctrl()``) using
> > **RTE_ETH_FILTER_GENERIC** as a new filter type.
> >
> > However a public front-end API described in `Rules management`_ will
> > be added as the preferred method to use it.
> >
> > Once discussions with the community have converged to a definite API,
> > legacy
> > filter types should be deprecated and a deadline defined to remove their
> > support entirely.
> >
> > PMDs will have to be gradually converted to **RTE_ETH_FILTER_GENERIC**
> > or
> > drop filtering support entirely. Less maintained PMDs for older hardware
> > may
> > lose support at this point.
> >
> > The notion of filter type will then be deprecated and subsequently dropped
> > to avoid confusion between both frameworks.
> >
> > Implementation details
> > ======================
> >
> > Flow rule
> > ---------
> >
> > A flow rule is the combination of a matching pattern with a list of actions,
> > and is the basis of this API.
> >
> > Priorities
> > ~~~~~~~~~~
> >
> > A priority can be assigned to a matching pattern.
> >
> > The default priority level is 0 and is also the highest. Support for more
> > than a single priority level in hardware is not guaranteed.
> >
> > If a packet is matched by several filters at a given priority level, the
> > outcome is undefined. It can take any path and can even be duplicated.
> >
> > Matching pattern
> > ~~~~~~~~~~~~~~~~
> >
> > A matching pattern comprises any number of items of various types.
> >
> > Items are arranged in a list to form a matching pattern for packets. They
> > fall in two categories:
> >
> > - Protocol matching (ANY, RAW, ETH, IPV4, IPV6, ICMP, UDP, TCP, VXLAN and
> > so
> > on), usually associated with a specification structure. These must be
> > stacked in the same order as the protocol layers to match, starting from
> > L2.
> >
> > - Affecting how the pattern is processed (END, VOID, INVERT, PF, VF,
> > SIGNATURE and so on), often without a specification structure. Since they
> > are meta data that does not match packet contents, these can be specified
> > anywhere within item lists without affecting the protocol matching items.
> >
> > Most item specifications can be optionally paired with a mask to narrow the
> > specific fields or bits to be matched.
> >
> > - Items are defined with ``struct rte_flow_item``.
> > - Patterns are defined with ``struct rte_flow_pattern``.
> >
> > Example of an item specification matching an Ethernet header:
> >
> > +-----------------------------------------+
> > | Ethernet |
> > +==========+=========+====================+
> > | ``spec`` | ``src`` | ``00:01:02:03:04`` |
> > | +---------+--------------------+
> > | | ``dst`` | ``00:2a:66:00:01`` |
> > +----------+---------+--------------------+
> > | ``mask`` | ``src`` | ``00:ff:ff:ff:00`` |
> > | +---------+--------------------+
> > | | ``dst`` | ``00:00:00:00:ff`` |
> > +----------+---------+--------------------+
> >
> > Non-masked bits stand for any value, Ethernet headers with the following
> > properties are thus matched:
> >
> > - ``src``: ``??:01:02:03:??``
> > - ``dst``: ``??:??:??:??:01``
> >
> > Except for meta types that do not need one, ``spec`` must be a valid pointer
> > to a structure of the related item type. A ``mask`` of the same type can be
> > provided to tell which bits in ``spec`` are to be matched.
> >
> > A mask is normally only needed for ``spec`` fields matching packet data,
> > ignored otherwise. See individual item types for more information.
> >
> > A ``NULL`` mask pointer is allowed and is similar to matching with a full
> > mask (all ones) ``spec`` fields supported by hardware, the remaining fields
> > are ignored (all zeroes), there is thus no error checking for unsupported
> > fields.
> >
> > Matching pattern items for packet data must be naturally stacked (ordered
> > from lowest to highest protocol layer), as in the following examples:
> >
> > +--------------+
> > | TCPv4 as L4 |
> > +===+==========+
> > | 0 | Ethernet |
> > +---+----------+
> > | 1 | IPv4 |
> > +---+----------+
> > | 2 | TCP |
> > +---+----------+
> >
> > +----------------+
> > | TCPv6 in VXLAN |
> > +===+============+
> > | 0 | Ethernet |
> > +---+------------+
> > | 1 | IPv4 |
> > +---+------------+
> > | 2 | UDP |
> > +---+------------+
> > | 3 | VXLAN |
> > +---+------------+
> > | 4 | Ethernet |
> > +---+------------+
> > | 5 | IPv6 |
> > +---+------------+
> > | 6 | TCP |
> > +---+------------+
> >
> > +-----------------------------+
> > | TCPv4 as L4 with meta items |
> > +===+=========================+
> > | 0 | VOID |
> > +---+-------------------------+
> > | 1 | Ethernet |
> > +---+-------------------------+
> > | 2 | VOID |
> > +---+-------------------------+
> > | 3 | IPv4 |
> > +---+-------------------------+
> > | 4 | TCP |
> > +---+-------------------------+
> > | 5 | VOID |
> > +---+-------------------------+
> > | 6 | VOID |
> > +---+-------------------------+
> >
> > The above example shows how meta items do not affect packet data
> > matching
> > items, as long as those remain stacked properly. The resulting matching
> > pattern is identical to "TCPv4 as L4".
> >
> > +----------------+
> > | UDPv6 anywhere |
> > +===+============+
> > | 0 | IPv6 |
> > +---+------------+
> > | 1 | UDP |
> > +---+------------+
> >
> > If supported by the PMD, omitting one or several protocol layers at the
> > bottom of the stack as in the above example (missing an Ethernet
> > specification) enables hardware to look anywhere in packets.
> >
> > It is unspecified whether the payload of supported encapsulations
> > (e.g. VXLAN inner packet) is matched by such a pattern, which may apply to
> > inner, outer or both packets.
> >
> > +---------------------+
> > | Invalid, missing L3 |
> > +===+=================+
> > | 0 | Ethernet |
> > +---+-----------------+
> > | 1 | UDP |
> > +---+-----------------+
> >
> > The above pattern is invalid due to a missing L3 specification between L2
> > and L4. It is only allowed at the bottom and at the top of the stack.
> >
> > Meta item types
> > ~~~~~~~~~~~~~~~
> >
> > These do not match packet data but affect how the pattern is processed,
> > most
> > of them do not need a specification structure. This particularity allows
> > them to be specified anywhere without affecting other item types.
> >
> > ``END``
> > ^^^^^^^
> >
> > End marker for item lists. Prevents further processing of items, thereby
> > ending the pattern.
> >
> > - Its numeric value is **0** for convenience.
> > - PMD support is mandatory.
> > - Both ``spec`` and ``mask`` are ignored.
> >
> > +--------------------+
> > | END |
> > +==========+=========+
> > | ``spec`` | ignored |
> > +----------+---------+
> > | ``mask`` | ignored |
> > +----------+---------+
> >
> > ``VOID``
> > ^^^^^^^^
> >
> > Used as a placeholder for convenience. It is ignored and simply discarded by
> > PMDs.
> >
> > - PMD support is mandatory.
> > - Both ``spec`` and ``mask`` are ignored.
> >
> > +--------------------+
> > | VOID |
> > +==========+=========+
> > | ``spec`` | ignored |
> > +----------+---------+
> > | ``mask`` | ignored |
> > +----------+---------+
> >
> > One usage example for this type is generating rules that share a common
> > prefix quickly without reallocating memory, only by updating item types:
> >
> > +------------------------+
> > | TCP, UDP or ICMP as L4 |
> > +===+====================+
> > | 0 | Ethernet |
> > +---+--------------------+
> > | 1 | IPv4 |
> > +---+------+------+------+
> > | 2 | UDP | VOID | VOID |
> > +---+------+------+------+
> > | 3 | VOID | TCP | VOID |
> > +---+------+------+------+
> > | 4 | VOID | VOID | ICMP |
> > +---+------+------+------+
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``INVERT``
> > ^^^^^^^^^^
> >
> > Inverted matching, i.e. process packets that do not match the pattern.
> >
> > - Both ``spec`` and ``mask`` are ignored.
> >
> > +--------------------+
> > | INVERT |
> > +==========+=========+
> > | ``spec`` | ignored |
> > +----------+---------+
> > | ``mask`` | ignored |
> > +----------+---------+
> >
> > Usage example in order to match non-TCPv4 packets only:
> >
> > +--------------------+
> > | Anything but TCPv4 |
> > +===+================+
> > | 0 | INVERT |
> > +---+----------------+
> > | 1 | Ethernet |
> > +---+----------------+
> > | 2 | IPv4 |
> > +---+----------------+
> > | 3 | TCP |
> > +---+----------------+
> >
> > ``PF``
> > ^^^^^^
> >
> > Matches packets addressed to the physical function of the device.
> >
> > - Both ``spec`` and ``mask`` are ignored.
> >
> > +--------------------+
> > | PF |
> > +==========+=========+
> > | ``spec`` | ignored |
> > +----------+---------+
> > | ``mask`` | ignored |
> > +----------+---------+
> >
> > ``VF``
> > ^^^^^^
> >
> > Matches packets addressed to the given virtual function ID of the device.
> >
> > - Only ``spec`` needs to be defined, ``mask`` is ignored.
> >
> > +----------------------------------------+
> > | VF |
> > +==========+=========+===================+
> > | ``spec`` | ``vf`` | destination VF ID |
> > +----------+---------+-------------------+
> > | ``mask`` | ignored |
> > +----------+-----------------------------+
> >
> > ``SIGNATURE``
> > ^^^^^^^^^^^^^
> >
> > Requests hash-based signature dispatching for this rule.
> >
> > Considering this is a global setting on devices that support it, all
> > subsequent filter rules may have to be created with it as well.
> >
> > - Only ``spec`` needs to be defined, ``mask`` is ignored.
> >
> > +--------------------+
> > | SIGNATURE |
> > +==========+=========+
> > | ``spec`` | TBD |
> > +----------+---------+
> > | ``mask`` | ignored |
> > +----------+---------+
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Data matching item types
> > ~~~~~~~~~~~~~~~~~~~~~~~~
> >
> > Most of these are basically protocol header definitions with associated
> > bitmasks. They must be specified (stacked) from lowest to highest protocol
> > layer.
> >
> > The following list is not exhaustive as new protocols will be added in the
> > future.
> >
> > ``ANY``
> > ^^^^^^^
> >
> > Matches any protocol in place of the current layer, a single ANY may also
> > stand for several protocol layers.
> >
> > This is usually specified as the first pattern item when looking for a
> > protocol anywhere in a packet.
> >
> > - A maximum value of **0** requests matching any number of protocol
> > layers
> > above or equal to the minimum value, a maximum value lower than the
> > minimum one is otherwise invalid.
> > - Only ``spec`` needs to be defined, ``mask`` is ignored.
> >
> > +-----------------------------------------------------------------------+
> > | ANY |
> > +==========+=========+====================================
> > ==============+
> > | ``spec`` | ``min`` | minimum number of layers covered |
> > | +---------+--------------------------------------------------+
> > | | ``max`` | maximum number of layers covered, 0 for infinity |
> > +----------+---------+--------------------------------------------------+
> > | ``mask`` | ignored |
> > +----------+------------------------------------------------------------+
> >
> > Example for VXLAN TCP payload matching regardless of outer L3 (IPv4 or
> > IPv6)
> > and L4 (UDP) both matched by the first ANY specification, and inner L3 (IPv4
> > or IPv6) matched by the second ANY specification:
> >
> > +----------------------------------+
> > | TCP in VXLAN with wildcards |
> > +===+==============================+
> > | 0 | Ethernet |
> > +---+-----+----------+---------+---+
> > | 1 | ANY | ``spec`` | ``min`` | 2 |
> > | | | +---------+---+
> > | | | | ``max`` | 2 |
> > +---+-----+----------+---------+---+
> > | 2 | VXLAN |
> > +---+------------------------------+
> > | 3 | Ethernet |
> > +---+-----+----------+---------+---+
> > | 4 | ANY | ``spec`` | ``min`` | 1 |
> > | | | +---------+---+
> > | | | | ``max`` | 1 |
> > +---+-----+----------+---------+---+
> > | 5 | TCP |
> > +---+------------------------------+
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``RAW``
> > ^^^^^^^
> >
> > Matches a string of a given length at a given offset (in bytes), or anywhere
> > in the payload of the current protocol layer (including L2 header if used as
> > the first item in the stack).
> >
> > This does not increment the protocol layer count as it is not a protocol
> > definition. Subsequent RAW items modulate the first absolute one with
> > relative offsets.
> >
> > - Using **-1** as the ``offset`` of the first RAW item makes its absolute
> > offset not fixed, i.e. the pattern is searched everywhere.
> > - ``mask`` only affects the pattern.
> >
> > +--------------------------------------------------------------+
> > | RAW |
> > +==========+=============+================================
> > =====+
> > | ``spec`` | ``offset`` | absolute or relative pattern offset |
> > | +-------------+-------------------------------------+
> > | | ``length`` | pattern length |
> > | +-------------+-------------------------------------+
> > | | ``pattern`` | byte string of the above length |
> > +----------+-------------+-------------------------------------+
> > | ``mask`` | ``offset`` | ignored |
> > | +-------------+-------------------------------------+
> > | | ``length`` | ignored |
> > | +-------------+-------------------------------------+
> > | | ``pattern`` | bitmask with the same byte length |
> > +----------+-------------+-------------------------------------+
> >
> > Example pattern looking for several strings at various offsets of a UDP
> > payload, using combined RAW items:
> >
> > +------------------------------------------+
> > | UDP payload matching |
> > +===+======================================+
> > | 0 | Ethernet |
> > +---+--------------------------------------+
> > | 1 | IPv4 |
> > +---+--------------------------------------+
> > | 2 | UDP |
> > +---+-----+----------+-------------+-------+
> > | 3 | RAW | ``spec`` | ``offset`` | -1 |
> > | | | +-------------+-------+
> > | | | | ``length`` | 3 |
> > | | | +-------------+-------+
> > | | | | ``pattern`` | "foo" |
> > +---+-----+----------+-------------+-------+
> > | 4 | RAW | ``spec`` | ``offset`` | 20 |
> > | | | +-------------+-------+
> > | | | | ``length`` | 3 |
> > | | | +-------------+-------+
> > | | | | ``pattern`` | "bar" |
> > +---+-----+----------+-------------+-------+
> > | 5 | RAW | ``spec`` | ``offset`` | -30 |
> > | | | +-------------+-------+
> > | | | | ``length`` | 3 |
> > | | | +-------------+-------+
> > | | | | ``pattern`` | "baz" |
> > +---+-----+----------+-------------+-------+
> >
> > This translates to:
> >
> > - Locate "foo" in UDP payload, remember its offset.
> > - Check "bar" at "foo"'s offset plus 20 bytes.
> > - Check "baz" at "foo"'s offset minus 30 bytes.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``ETH``
> > ^^^^^^^
> >
> > Matches an Ethernet header.
> >
> > - ``dst``: destination MAC.
> > - ``src``: source MAC.
> > - ``type``: EtherType.
> > - ``tags``: number of 802.1Q/ad tags defined.
> > - ``tag[]``: 802.1Q/ad tag definitions, innermost first. For each one:
> >
> > - ``tpid``: Tag protocol identifier.
> > - ``tci``: Tag control information.
> >
> > ``IPV4``
> > ^^^^^^^^
> >
> > Matches an IPv4 header.
> >
> > - ``src``: source IP address.
> > - ``dst``: destination IP address.
> > - ``tos``: ToS/DSCP field.
> > - ``ttl``: TTL field.
> > - ``proto``: protocol number for the next layer.
> >
> > ``IPV6``
> > ^^^^^^^^
> >
> > Matches an IPv6 header.
> >
> > - ``src``: source IP address.
> > - ``dst``: destination IP address.
> > - ``tc``: traffic class field.
> > - ``nh``: Next header field (protocol).
> > - ``hop_limit``: hop limit field (TTL).
> >
> > ``ICMP``
> > ^^^^^^^^
> >
> > Matches an ICMP header.
> >
> > - TBD.
> >
> > ``UDP``
> > ^^^^^^^
> >
> > Matches a UDP header.
> >
> > - ``sport``: source port.
> > - ``dport``: destination port.
> > - ``length``: UDP length.
> > - ``checksum``: UDP checksum.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``TCP``
> > ^^^^^^^
> >
> > Matches a TCP header.
> >
> > - ``sport``: source port.
> > - ``dport``: destination port.
> > - All other TCP fields and bits.
> >
> > ``VXLAN``
> > ^^^^^^^^^
> >
> > Matches a VXLAN header.
> >
> > - TBD.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Actions
> > ~~~~~~~
> >
> > Each possible action is represented by a type. Some have associated
> > configuration structures. Several actions combined in a list can be affected
> > to a flow rule. That list is not ordered.
> >
> > At least one action must be defined in a filter rule in order to do
> > something with matched packets.
> >
> > - Actions are defined with ``struct rte_flow_action``.
> > - A list of actions is defined with ``struct rte_flow_actions``.
> >
> > They fall in three categories:
> >
> > - Terminating actions (such as QUEUE, DROP, RSS, PF, VF) that prevent
> > processing matched packets by subsequent flow rules, unless overridden
> > with PASSTHRU.
> >
> > - Non terminating actions (PASSTHRU, DUP) that leave matched packets up
> > for
> > additional processing by subsequent flow rules.
> >
> > - Other non terminating meta actions that do not affect the fate of packets
> > (END, VOID, ID, COUNT).
> >
> > When several actions are combined in a flow rule, they should all have
> > different types (e.g. dropping a packet twice is not possible). However
> > considering the VOID type is an exception to this rule, the defined behavior
> > is for PMDs to only take into account the last action of a given type found
> > in the list. PMDs still perform error checking on the entire list.
> >
> > *Note that PASSTHRU is the only action able to override a terminating rule.*
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Example of an action that redirects packets to queue index 10:
> >
> > +----------------+
> > | QUEUE |
> > +===========+====+
> > | ``queue`` | 10 |
> > +-----------+----+
> >
> > Action lists examples, their order is not significant, applications must
> > consider all actions to be performed simultaneously:
> >
> > +----------------+
> > | Count and drop |
> > +=======+========+
> > | COUNT | |
> > +-------+--------+
> > | DROP | |
> > +-------+--------+
> >
> > +--------------------------+
> > | Tag, count and redirect |
> > +=======+===========+======+
> > | ID | ``id`` | 0x2a |
> > +-------+-----------+------+
> > | COUNT | |
> > +-------+-----------+------+
> > | QUEUE | ``queue`` | 10 |
> > +-------+-----------+------+
> >
> > +-----------------------+
> > | Redirect to queue 5 |
> > +=======+===============+
> > | DROP | |
> > +-------+-----------+---+
> > | QUEUE | ``queue`` | 5 |
> > +-------+-----------+---+
> >
> > In the above example, considering both actions are performed
> > simultaneously,
> > its end result is that only QUEUE has any effect.
> >
> > +-----------------------+
> > | Redirect to queue 3 |
> > +=======+===========+===+
> > | QUEUE | ``queue`` | 5 |
> > +-------+-----------+---+
> > | VOID | |
> > +-------+-----------+---+
> > | QUEUE | ``queue`` | 3 |
> > +-------+-----------+---+
> >
> > As previously described, only the last action of a given type found in the
> > list is taken into account. The above example also shows that VOID is
> > ignored.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Action types
> > ~~~~~~~~~~~~
> >
> > Common action types are described in this section. Like pattern item types,
> > this list is not exhaustive as new actions will be added in the future.
> >
> > ``END`` (action)
> > ^^^^^^^^^^^^^^^^
> >
> > End marker for action lists. Prevents further processing of actions, thereby
> > ending the list.
> >
> > - Its numeric value is **0** for convenience.
> > - PMD support is mandatory.
> > - No configurable property.
> >
> > +---------------+
> > | END |
> > +===============+
> > | no properties |
> > +---------------+
> >
> > ``VOID`` (action)
> > ^^^^^^^^^^^^^^^^^
> >
> > Used as a placeholder for convenience. It is ignored and simply discarded by
> > PMDs.
> >
> > - PMD support is mandatory.
> > - No configurable property.
> >
> > +---------------+
> > | VOID |
> > +===============+
> > | no properties |
> > +---------------+
> >
> > ``PASSTHRU``
> > ^^^^^^^^^^^^
> >
> > Leaves packets up for additional processing by subsequent flow rules. This
> > is the default when a rule does not contain a terminating action, but can be
> > specified to force a rule to become non-terminating.
> >
> > - No configurable property.
> >
> > +---------------+
> > | PASSTHRU |
> > +===============+
> > | no properties |
> > +---------------+
> >
> > Example to copy a packet to a queue and continue processing by subsequent
> > flow rules:
> [Sugesh] If a packet get copied to a queue, it’s a termination action.
> How can its possible to do subsequent action after the packet already
> moved to the queue. ?How it differs from DUP action?
> Am I missing anything here?
Devices may not support the combination of QUEUE + PASSTHRU (i.e. making
QUEUE non-terminating). However these same devices may expose the ability to
copy a packet to another (sniffer) queue all while keeping the rule
terminating (QUEUE + DUP but no PASSTHRU).
DUP with two rules, assuming priorties and PASSTRHU are supported:
- pattern X, priority 0; actions: QUEUE 5, PASSTHRU (non-terminating)
- pattern X, priority 1; actions: QUEUE 6 (terminating)
DUP with two actions on a single rule and a single priority:
- pattern X, priority 0; actions: DUP 5, QUEUE 6 (terminating)
If supported, from an application point of view the end result is similar in
both cases (note the second case may be implemented by the PMD using two HW
rules internally).
However the second case does not waste a priority level and clearly states
the intent to the PMD which is more likely to be supported. If HW supports
DUP directly it is even faster since there is a single rule. That is why I
thought having DUP as an action would be useful.
> > +--------------------------+
> > | Copy to queue 8 |
> > +==========+===============+
> > | PASSTHRU | |
> > +----------+-----------+---+
> > | QUEUE | ``queue`` | 8 |
> > +----------+-----------+---+
> >
> > ``ID``
> > ^^^^^^
> >
> > Attaches a 32 bit value to packets.
> >
> > +----------------------------------------------+
> > | ID |
> > +========+=====================================+
> > | ``id`` | 32 bit value to return with packets |
> > +--------+-------------------------------------+
> >
> [Sugesh] I assume the application has to program the flow
> with a unique ID and matching packets are stamped with this ID
> when reporting to the software. The uniqueness of ID is NOT
> guaranteed by the API framework. Correct me if I am wrong here.
You are right, if the way I wrote it is not clear enough, I'm open to
suggestions to improve it.
> [Sugesh] Is it a limitation to use only 32 bit ID? Is it possible to have a
> 64 bit ID? So that application can use the control plane flow pointer
> Itself as an ID. Does it make sense?
I've specified a 32 bit ID for now because this is what FDIR supports and
also what existing devices can report today AFAIK (i40e and mlx5).
We could use 64 bit for future-proofness in a separate action like "ID64"
when at least one device supports it.
To PMD maintainers: please comment if you know devices that support tagging
matching packets with more than 32 bits of user-provided data!
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``QUEUE``
> > ^^^^^^^^^
> >
> > Assigns packets to a given queue index.
> >
> > - Terminating by default.
> >
> > +--------------------------------+
> > | QUEUE |
> > +===========+====================+
> > | ``queue`` | queue index to use |
> > +-----------+--------------------+
> >
> > ``DROP``
> > ^^^^^^^^
> >
> > Drop packets.
> >
> > - No configurable property.
> > - Terminating by default.
> > - PASSTHRU overrides this action if both are specified.
> >
> > +---------------+
> > | DROP |
> > +===============+
> > | no properties |
> > +---------------+
> >
> > ``COUNT``
> > ^^^^^^^^^
> >
> [Sugesh] Should we really have to set count action explicitly for every rule?
> IMHO it would be great to be an implicit action. Most of the application would be
> interested in the stats of almost all the filters/flows .
I can see why, but no, it must be explicitly requested because you may want
to know in advance when it is not supported. Also considering it is
something else to be done by HW (a separate action), we can assume enabling
this may slow things down a bit.
HW limitations may also prevent you from having as many flow counters as you
want, in which case you probably want to carefully pick which rules have
them.
I think this target is most useful with DROP, VF and PF actions since
those are currently the only ones where SW may not see the related packets.
> > Enables hits counter for this rule.
> >
> > This counter can be retrieved and reset through ``rte_flow_query()``, see
> > ``struct rte_flow_query_count``.
> >
> > - Counters can be retrieved with ``rte_flow_query()``.
> > - No configurable property.
> >
> > +---------------+
> > | COUNT |
> > +===============+
> > | no properties |
> > +---------------+
> >
> > Query structure to retrieve and reset the flow rule hits counter:
> >
> > +------------------------------------------------+
> > | COUNT query |
> > +===========+=====+==============================+
> > | ``reset`` | in | reset counter after query |
> > +-----------+-----+------------------------------+
> > | ``hits`` | out | number of hits for this flow |
> > +-----------+-----+------------------------------+
> >
> > ``DUP``
> > ^^^^^^^
> >
> > Duplicates packets to a given queue index.
> >
> > This is normally combined with QUEUE, however when used alone, it is
> > actually similar to QUEUE + PASSTHRU.
> >
> > - Non-terminating by default.
> >
> > +------------------------------------------------+
> > | DUP |
> > +===========+====================================+
> > | ``queue`` | queue index to duplicate packet to |
> > +-----------+------------------------------------+
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``RSS``
> > ^^^^^^^
> >
> > Similar to QUEUE, except RSS is additionally performed on packets to spread
> > them among several queues according to the provided parameters.
> >
> > - Terminating by default.
> >
> > +---------------------------------------------+
> > | RSS |
> > +==============+==============================+
> > | ``rss_conf`` | RSS parameters |
> > +--------------+------------------------------+
> > | ``queues`` | number of entries in queue[] |
> > +--------------+------------------------------+
> > | ``queue[]`` | queue indices to use |
> > +--------------+------------------------------+
> >
> > ``PF`` (action)
> > ^^^^^^^^^^^^^^^
> >
> > Redirects packets to the physical function (PF) of the current device.
> >
> > - No configurable property.
> > - Terminating by default.
> >
> > +---------------+
> > | PF |
> > +===============+
> > | no properties |
> > +---------------+
> >
> > ``VF`` (action)
> > ^^^^^^^^^^^^^^^
> >
> > Redirects packets to the virtual function (VF) of the current device with
> > the specified ID.
> >
> > - Terminating by default.
> >
> > +---------------------------------------+
> > | VF |
> > +========+==============================+
> > | ``id`` | VF ID to redirect packets to |
> > +--------+------------------------------+
> >
> > Planned types
> > ~~~~~~~~~~~~~
> >
> > Other action types are planned but not defined yet. These actions will add
> > the ability to alter matching packets in several ways, such as performing
> > encapsulation/decapsulation of tunnel headers on specific flows.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Rules management
> > ----------------
> >
> > A simple API with only four functions is provided to fully manage flows.
> >
> > Each created flow rule is associated with an opaque, PMD-specific handle
> > pointer. The application is responsible for keeping it until the rule is
> > destroyed.
> >
> > Flows rules are defined with ``struct rte_flow``.
> >
> > Validation
> > ~~~~~~~~~~
> >
> > Given that expressing a definite set of device capabilities with this API is
> > not practical, a dedicated function is provided to check if a flow rule is
> > supported and can be created.
> >
> > ::
> >
> > int
> > rte_flow_validate(uint8_t port_id,
> > const struct rte_flow_pattern *pattern,
> > const struct rte_flow_actions *actions);
> >
> > While this function has no effect on the target device, the flow rule is
> > validated against its current configuration state and the returned value
> > should be considered valid by the caller for that state only.
> >
> > The returned value is guaranteed to remain valid only as long as no
> > successful calls to rte_flow_create() or rte_flow_destroy() are made in the
> > meantime and no device parameter affecting flow rules in any way are
> > modified, due to possible collisions or resource limitations (although in
> > such cases ``EINVAL`` should not be returned).
> >
> > Arguments:
> >
> > - ``port_id``: port identifier of Ethernet device.
> > - ``pattern``: pattern specification to check.
> > - ``actions``: actions associated with the flow definition.
> >
> > Return value:
> >
> > - **0** if flow rule is valid and can be created. A negative errno value
> > otherwise (``rte_errno`` is also set), the following errors are defined.
> > - ``-EINVAL``: unknown or invalid rule specification.
> > - ``-ENOTSUP``: valid but unsupported rule specification (e.g. partial masks
> > are unsupported).
> > - ``-EEXIST``: collision with an existing rule.
> > - ``-ENOMEM``: not enough resources.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Creation
> > ~~~~~~~~
> >
> > Creating a flow rule is similar to validating one, except the rule is
> > actually created.
> >
> > ::
> >
> > struct rte_flow *
> > rte_flow_create(uint8_t port_id,
> > const struct rte_flow_pattern *pattern,
> > const struct rte_flow_actions *actions);
> >
> > Arguments:
> >
> > - ``port_id``: port identifier of Ethernet device.
> > - ``pattern``: pattern specification to add.
> > - ``actions``: actions associated with the flow definition.
> >
> > Return value:
> >
> > A valid flow pointer in case of success, NULL otherwise and ``rte_errno`` is
> > set to the positive version of one of the error codes defined for
> > ``rte_flow_validate()``.
> [Sugesh] : Kind of implementation specific query. What if application
> try to add duplicate rules? Does the API create new flow entry for every
> API call?
If an application adds duplicate rules at a given priority level, the second
one may return an error depending on the PMD. Collisions are sometimes
trivial to detect (such as the same pattern twice), others not so much (one
matching an Ethernet header only, the other one matching an IP header only).
Either way if a packet is matched by two rules at a given priority level,
what happens is described in 3.3 (High level design) and 4.4.1 (Priorities).
Applications are responsible for not relying on the PMD to detect these, or
should use a single priority level for each rule to make things clear.
However since the number of HW priority levels is finite and possibly small,
they must also make sure not to waste them. My advice is to only use
priority levels when it cannot be proven that rules do not collide.
If all you have is perfect matching rules without wildcards and all of them
match the same number of layers, a single priority level is fine.
> [Sugesh] Another concern is the cost and time of installing these rules
> in the hardware. Can we make these APIs time bound(or at least an option to
> set the time limit to execute these APIs), so that
> Application doesn’t have to wait so long when installing and deleting flows with
> slow hardware/NIC. What do you think? Most of the datapath flow installations are
> dynamic and triggered only when there is
> an ingress traffic. Delay in flow insertion/deletion have unpredictable consequences.
This API is (currently) aimed at the control path only, and must indeed be
assumed to be slow. Creating million of rules may take quite long as it may
involve syscalls and other time-consuming synchronization things on the PMD
side.
So currently there is no plan to have rules added from the data path with
time constraints. I think it would be implemented through a different set of
functions anyway.
I do not think adding time limits is practical, even specifying in the API
that creating a single flow rule must take less than a maximum number of
seconds in order to be effective is too much of a constraint (applications
that create all flows during init may not care after all).
You should consider in any case that modifying flow rules will always be
slower than receiving packets, there is no way around that. Applications
have to live with it and provide a software fallback for incoming packets
while managing flow rules.
Moreover, think about what happens when you hit the maximum number of flow
rules and cannot create any more. Applications need to implement some kind
of fallback in their data path.
Offloading flows in HW is also only useful if they live much longer than the
time taken to create and delete them. Perhaps applications may choose to do
so after detecting long lived flows such as TCP sessions.
You may have one separate control thread dedicated to manage flows and
keep your normal control thread unaffected by delays. Several threads can
even be dedicated, one per device.
> [Sugesh] Another query is on the synchronization part. What if same rules are
> handled from different threads? Is application responsible for handling the concurrent
> hardware programming?
Like most (if not all) DPDK APIs, applications are responsible for managing
locking issues as decribed in 4.3 (Behavior). Since this is a control path
API and applications usually have a single control thread, locking should
not be necessary in most cases.
Regarding my above comment about using several control threads to manage
different devices, section 4.3 says:
"There is no provision for reentrancy/multi-thread safety, although nothing
should prevent different devices from being configured at the same
time. PMDs may protect their control path functions accordingly."
I'd like to emphasize it is not "per port" but "per device", since in a few
cases a configurable resource is shared by several ports. It may be
difficult for applications to determine which ports are shared by a given
device but this falls outside the scope of this API.
Do you think adding the guarantee that it is always safe to configure two
different ports simultaneously without locking from the application side is
necessary? In which case the PMD would be responsible for locking shared
resources.
> > Destruction
> > ~~~~~~~~~~~
> >
> > Flow rules destruction is not automatic, and a queue should not be released
> > if any are still attached to it. Applications must take care of performing
> > this step before releasing resources.
> >
> > ::
> >
> > int
> > rte_flow_destroy(uint8_t port_id,
> > struct rte_flow *flow);
> >
> >
> [Sugesh] I would suggest having a clean-up API is really useful as the releasing of
> Queue(is it applicable for releasing of port too?) is not guaranteeing the automatic flow
> destruction.
Would something like rte_flow_flush(port_id) do the trick? I wanted to
emphasize in this first draft that applications should really keep the flow
pointers around in order to manage/destroy them. It is their responsibility,
not PMD's.
> This way application can initialize the port,
> clean-up all the existing rules and create new rules on a clean slate.
No resource can be released as long as a flow rule is using it (bad things
may happen otherwise), all flow rules must be destroyed first, thus none can
possibly remain after initializing a port. It is assumed that PMDs do
automatic clean up during init if necessary to ensure this.
> > Failure to destroy a flow rule may occur when other flow rules depend on it,
> > and destroying it would result in an inconsistent state.
> >
> > This function is only guaranteed to succeed if flow rules are destroyed in
> > reverse order of their creation.
> >
> > Arguments:
> >
> > - ``port_id``: port identifier of Ethernet device.
> > - ``flow``: flow rule to destroy.
> >
> > Return value:
> >
> > - **0** on success, a negative errno value otherwise and ``rte_errno`` is
> > set.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Query
> > ~~~~~
> >
> > Query an existing flow rule.
> >
> > This function allows retrieving flow-specific data such as counters. Data
> > is gathered by special actions which must be present in the flow rule
> > definition.
> >
> > ::
> >
> > int
> > rte_flow_query(uint8_t port_id,
> > struct rte_flow *flow,
> > enum rte_flow_action_type action,
> > void *data);
> >
> > Arguments:
> >
> > - ``port_id``: port identifier of Ethernet device.
> > - ``flow``: flow rule to query.
> > - ``action``: action type to query.
> > - ``data``: pointer to storage for the associated query data type.
> >
> > Return value:
> >
> > - **0** on success, a negative errno value otherwise and ``rte_errno`` is
> > set.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Behavior
> > --------
> >
> > - API operations are synchronous and blocking (``EAGAIN`` cannot be
> > returned).
> >
> > - There is no provision for reentrancy/multi-thread safety, although nothing
> > should prevent different devices from being configured at the same
> > time. PMDs may protect their control path functions accordingly.
> >
> > - Stopping the data path (TX/RX) should not be necessary when managing
> > flow
> > rules. If this cannot be achieved naturally or with workarounds (such as
> > temporarily replacing the burst function pointers), an appropriate error
> > code must be returned (``EBUSY``).
> >
> > - PMDs, not applications, are responsible for maintaining flow rules
> > configuration when stopping and restarting a port or performing other
> > actions which may affect them. They can only be destroyed explicitly.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> [Sugesh] Query all the rules for a specific port/queue?? Useful when adding and
> deleting ports and queues dynamically according to the need. I am not sure
> what are the other different usecases for these APIs. But I feel it makes much easier to
> manage flows from the application. What do you think?
Not sure, that seems to fall out of the scope of this API. As described,
applications already store the related rte_flow pointers. Accordingly, they
know how many rules are associated to a given port. They need both a port ID
and a flow rule pointer to destroy them after all.
Now perhaps something to convert back an existing rte_flow to a pattern and
a list of actions, however I cannot see an immediate use case for it.
What you describe seems to be doable through a front-end API, I think
keeping this one as low-level as possible with only basic actions is better
right now. I'll keep your suggestion in mind.
> > Compatibility
> > -------------
> >
> > No known hardware implementation supports all the features described in
> > this
> > document.
> >
> > Unsupported features or combinations are not expected to be fully
> > emulated
> > in software by PMDs for performance reasons. Partially supported features
> > may be completed in software as long as hardware performs most of the
> > work
> > (such as queue redirection and packet recognition).
> >
> > However PMDs are expected to do their best to satisfy application requests
> > by working around hardware limitations as long as doing so does not affect
> > the behavior of existing flow rules.
> >
> > The following sections provide a few examples of such cases, they are based
> > on limitations built into the previous APIs.
> >
> > Global bitmasks
> > ~~~~~~~~~~~~~~~
> >
> > Each flow rule comes with its own, per-layer bitmasks, while hardware may
> > support only a single, device-wide bitmask for a given layer type, so that
> > two IPv4 rules cannot use different bitmasks.
> >
> > The expected behavior in this case is that PMDs automatically configure
> > global bitmasks according to the needs of the first created flow rule.
> >
> > Subsequent rules are allowed only if their bitmasks match those, the
> > ``EEXIST`` error code should be returned otherwise.
> >
> > Unsupported layer types
> > ~~~~~~~~~~~~~~~~~~~~~~~
> >
> > Many protocols can be simulated by crafting patterns with the `RAW`_ type.
> >
> > PMDs can rely on this capability to simulate support for protocols with
> > fixed headers not directly recognized by hardware.
> >
> > ``ANY`` pattern item
> > ~~~~~~~~~~~~~~~~~~~~
> >
> > This pattern item stands for anything, which can be difficult to translate
> > to something hardware would understand, particularly if followed by more
> > specific types.
> >
> > Consider the following pattern:
> >
> > +---+--------------------------------+
> > | 0 | ETHER |
> > +---+--------------------------------+
> > | 1 | ANY (``min`` = 1, ``max`` = 1) |
> > +---+--------------------------------+
> > | 2 | TCP |
> > +---+--------------------------------+
> >
> > Knowing that TCP does not make sense with something other than IPv4 and
> > IPv6
> > as L3, such a pattern may be translated to two flow rules instead:
> >
> > +---+--------------------+
> > | 0 | ETHER |
> > +---+--------------------+
> > | 1 | IPV4 (zeroed mask) |
> > +---+--------------------+
> > | 2 | TCP |
> > +---+--------------------+
> >
> > +---+--------------------+
> > | 0 | ETHER |
> > +---+--------------------+
> > | 1 | IPV6 (zeroed mask) |
> > +---+--------------------+
> > | 2 | TCP |
> > +---+--------------------+
> >
> > Note that as soon as a ANY rule covers several layers, this approach may
> > yield a large number of hidden flow rules. It is thus suggested to only
> > support the most common scenarios (anything as L2 and/or L3).
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > Unsupported actions
> > ~~~~~~~~~~~~~~~~~~~
> >
> > - When combined with a `QUEUE`_ action, packet counting (`COUNT`_) and
> > tagging (`ID`_) may be implemented in software as long as the target queue
> > is used by a single rule.
> >
> > - A rule specifying both `DUP`_ + `QUEUE`_ may be translated to two hidden
> > rules combining `QUEUE`_ and `PASSTHRU`_.
> >
> > - When a single target queue is provided, `RSS`_ can also be implemented
> > through `QUEUE`_.
> >
> > Flow rules priority
> > ~~~~~~~~~~~~~~~~~~~
> >
> > While it would naturally make sense, flow rules cannot be assumed to be
> > processed by hardware in the same order as their creation for several
> > reasons:
> >
> > - They may be managed internally as a tree or a hash table instead of a
> > list.
> > - Removing a flow rule before adding another one can either put the new
> > rule
> > at the end of the list or reuse a freed entry.
> > - Duplication may occur when packets are matched by several rules.
> >
> > For overlapping rules (particularly in order to use the `PASSTHRU`_ action)
> > predictable behavior is only guaranteed by using different priority levels.
> >
> > Priority levels are not necessarily implemented in hardware, or may be
> > severely limited (e.g. a single priority bit).
> >
> > For these reasons, priority levels may be implemented purely in software by
> > PMDs.
> >
> > - For devices expecting flow rules to be added in the correct order, PMDs
> > may destroy and re-create existing rules after adding a new one with
> > a higher priority.
> >
> > - A configurable number of dummy or empty rules can be created at
> > initialization time to save high priority slots for later.
> >
> > - In order to save priority levels, PMDs may evaluate whether rules are
> > likely to collide and adjust their priority accordingly.
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > API migration
> > =============
> >
> > Exhaustive list of deprecated filter types and how to convert them to
> > generic flow rules.
> >
> > ``MACVLAN`` to ``ETH`` → ``VF``, ``PF``
> > ---------------------------------------
> >
> > `MACVLAN`_ can be translated to a basic `ETH`_ flow rule with a `VF
> > (action)`_ or `PF (action)`_ terminating action.
> >
> > +------------------------------------+
> > | MACVLAN |
> > +--------------------------+---------+
> > | Pattern | Actions |
> > +===+=====+==========+=====+=========+
> > | 0 | ETH | ``spec`` | any | VF, |
> > | | +----------+-----+ PF |
> > | | | ``mask`` | any | |
> > +---+-----+----------+-----+---------+
> >
> > ``ETHERTYPE`` to ``ETH`` → ``QUEUE``, ``DROP``
> > ----------------------------------------------
> >
> > `ETHERTYPE`_ is basically an `ETH`_ flow rule with `QUEUE`_ or `DROP`_ as
> > a terminating action.
> >
> > +------------------------------------+
> > | ETHERTYPE |
> > +--------------------------+---------+
> > | Pattern | Actions |
> > +===+=====+==========+=====+=========+
> > | 0 | ETH | ``spec`` | any | QUEUE, |
> > | | +----------+-----+ DROP |
> > | | | ``mask`` | any | |
> > +---+-----+----------+-----+---------+
> >
> > ``FLEXIBLE`` to ``RAW`` → ``QUEUE``
> > -----------------------------------
> >
> > `FLEXIBLE`_ can be translated to one `RAW`_ pattern with `QUEUE`_ as the
> > terminating action and a defined priority level.
> >
> > +------------------------------------+
> > | FLEXIBLE |
> > +--------------------------+---------+
> > | Pattern | Actions |
> > +===+=====+==========+=====+=========+
> > | 0 | RAW | ``spec`` | any | QUEUE |
> > | | +----------+-----+ |
> > | | | ``mask`` | any | |
> > +---+-----+----------+-----+---------+
> >
> > ``SYN`` to ``TCP`` → ``QUEUE``
> > ------------------------------
> >
> > `SYN`_ is a `TCP`_ rule with only the ``syn`` bit enabled and masked, and
> > `QUEUE`_ as the terminating action.
> >
> > Priority level can be set to simulate the high priority bit.
> >
> > +---------------------------------------------+
> > | SYN |
> > +-----------------------------------+---------+
> > | Pattern | Actions |
> > +===+======+==========+=============+=========+
> > | 0 | ETH | ``spec`` | N/A | QUEUE |
> > | | +----------+-------------+ |
> > | | | ``mask`` | empty | |
> > +---+------+----------+-------------+ |
> > | 1 | IPV4 | ``spec`` | N/A | |
> > | | +----------+-------------+ |
> > | | | ``mask`` | empty | |
> > +---+------+----------+-------------+ |
> > | 2 | TCP | ``spec`` | ``syn`` = 1 | |
> > | | +----------+-------------+ |
> > | | | ``mask`` | ``syn`` = 1 | |
> > +---+------+----------+-------------+---------+
> >
> > ``NTUPLE`` to ``IPV4``, ``TCP``, ``UDP`` → ``QUEUE``
> > ----------------------------------------------------
> >
> > `NTUPLE`_ is similar to specifying an empty L2, `IPV4`_ as L3 with `TCP`_ or
> > `UDP`_ as L4 and `QUEUE`_ as the terminating action.
> >
> > A priority level can be specified as well.
> >
> > +---------------------------------------+
> > | NTUPLE |
> > +-----------------------------+---------+
> > | Pattern | Actions |
> > +===+======+==========+=======+=========+
> > | 0 | ETH | ``spec`` | N/A | QUEUE |
> > | | +----------+-------+ |
> > | | | ``mask`` | empty | |
> > +---+------+----------+-------+ |
> > | 1 | IPV4 | ``spec`` | any | |
> > | | +----------+-------+ |
> > | | | ``mask`` | any | |
> > +---+------+----------+-------+ |
> > | 2 | TCP, | ``spec`` | any | |
> > | | UDP +----------+-------+ |
> > | | | ``mask`` | any | |
> > +---+------+----------+-------+---------+
> >
> > ``TUNNEL`` to ``ETH``, ``IPV4``, ``IPV6``, ``VXLAN`` (or other) → ``QUEUE``
> > ---------------------------------------------------------------------------
> >
> > `TUNNEL`_ matches common IPv4 and IPv6 L3/L4-based tunnel types.
> >
> > In the following table, `ANY`_ is used to cover the optional L4.
> >
> > +------------------------------------------------+
> > | TUNNEL |
> > +--------------------------------------+---------+
> > | Pattern | Actions |
> > +===+=========+==========+=============+=========+
> > | 0 | ETH | ``spec`` | any | QUEUE |
> > | | +----------+-------------+ |
> > | | | ``mask`` | any | |
> > +---+---------+----------+-------------+ |
> > | 1 | IPV4, | ``spec`` | any | |
> > | | IPV6 +----------+-------------+ |
> > | | | ``mask`` | any | |
> > +---+---------+----------+-------------+ |
> > | 2 | ANY | ``spec`` | ``min`` = 0 | |
> > | | | +-------------+ |
> > | | | | ``max`` = 0 | |
> > | | +----------+-------------+ |
> > | | | ``mask`` | N/A | |
> > +---+---------+----------+-------------+ |
> > | 3 | VXLAN, | ``spec`` | any | |
> > | | GENEVE, +----------+-------------+ |
> > | | TEREDO, | ``mask`` | any | |
> > | | NVGRE, | | | |
> > | | GRE, | | | |
> > | | ... | | | |
> > +---+---------+----------+-------------+---------+
> >
> > .. raw:: pdf
> >
> > PageBreak
> >
> > ``FDIR`` to most item types → ``QUEUE``, ``DROP``, ``PASSTHRU``
> > ---------------------------------------------------------------
> >
> > `FDIR`_ is more complex than any other type, there are several methods to
> > emulate its functionality. It is summarized for the most part in the table
> > below.
> >
> > A few features are intentionally not supported:
> >
> > - The ability to configure the matching input set and masks for the entire
> > device, PMDs should take care of it automatically according to flow rules.
> >
> > - Returning four or eight bytes of matched data when using flex bytes
> > filtering. Although a specific action could implement it, it conflicts
> > with the much more useful 32 bits tagging on devices that support it.
> >
> > - Side effects on RSS processing of the entire device. Flow rules that
> > conflict with the current device configuration should not be
> > allowed. Similarly, device configuration should not be allowed when it
> > affects existing flow rules.
> >
> > - Device modes of operation. "none" is unsupported since filtering cannot be
> > disabled as long as a flow rule is present.
> >
> > - "MAC VLAN" or "tunnel" perfect matching modes should be automatically
> > set
> > according to the created flow rules.
> >
> > +----------------------------------------------+
> > | FDIR |
> > +---------------------------------+------------+
> > | Pattern | Actions |
> > +===+============+==========+=====+============+
> > | 0 | ETH, | ``spec`` | any | QUEUE, |
> > | | RAW +----------+-----+ DROP, |
> > | | | ``mask`` | any | PASSTHRU |
> > +---+------------+----------+-----+------------+
> > | 1 | IPV4, | ``spec`` | any | ID |
> > | | IPV6 +----------+-----+ (optional) |
> > | | | ``mask`` | any | |
> > +---+------------+----------+-----+ |
> > | 2 | TCP, | ``spec`` | any | |
> > | | UDP, +----------+-----+ |
> > | | SCTP | ``mask`` | any | |
> > +---+------------+----------+-----+ |
> > | 3 | VF, | ``spec`` | any | |
> > | | PF, +----------+-----+ |
> > | | SIGNATURE | ``mask`` | any | |
> > | | (optional) | | | |
> > +---+------------+----------+-----+------------+
> >
> > ``HASH``
> > ~~~~~~~~
> >
> > Hashing configuration is set per rule through the `SIGNATURE`_ item.
> >
> > Since it is usually a global device setting, all flow rules created with
> > this item may have to share the same specification.
> >
> > ``L2_TUNNEL`` to ``VOID`` → ``VXLAN`` (or others)
> > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >
> > All packets are matched. This type alters incoming packets to encapsulate
> > them in a chosen tunnel type, optionally redirect them to a VF as well.
> >
> > The destination pool for tag based forwarding can be emulated with other
> > flow rules using `DUP`_ as the action.
> >
> > +----------------------------------------+
> > | L2_TUNNEL |
> > +---------------------------+------------+
> > | Pattern | Actions |
> > +===+======+==========+=====+============+
> > | 0 | VOID | ``spec`` | N/A | VXLAN, |
> > | | | | | GENEVE, |
> > | | | | | ... |
> > | | +----------+-----+------------+
> > | | | ``mask`` | N/A | VF |
> > | | | | | (optional) |
> > +---+------+----------+-----+------------+
> >
> > --
> > Adrien Mazarguil
> > 6WIND
--
Adrien Mazarguil
6WIND
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-05 18:16 2% [dpdk-dev] [RFC] Generic flow director/filtering/classification API Adrien Mazarguil
2016-07-07 7:14 0% ` Lu, Wenzhuo
2016-07-07 23:15 0% ` Chandran, Sugesh
@ 2016-07-08 11:11 0% ` Liang, Cunming
2 siblings, 0 replies; 200+ results
From: Liang, Cunming @ 2016-07-08 11:11 UTC (permalink / raw)
To: dev, Thomas Monjalon, Helin Zhang, Jingjing Wu, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Wenzhuo Lu, Jan Medala,
John Daley, Jing Chen, Konstantin Ananyev, Matej Vido,
Alejandro Lucero, Sony Chacko, Jerin Jacob, Pablo de Lara,
Olga Shern
Hi Adrien,
On 7/6/2016 2:16 AM, Adrien Mazarguil wrote:
> Hi All,
>
> First, forgive me for this large message, I know our mailboxes already
> suffer quite a bit from the amount of traffic on this ML.
>
> This is not exactly yet another thread about how flow director should be
> extended, rather about a brand new API to handle filtering and
> classification for incoming packets in the most PMD-generic and
> application-friendly fashion we can come up with. Reasons described below.
>
> I think this topic is important enough to include both the users of this API
> as well as PMD maintainers. So far I have CC'ed librte_ether (especially
> rte_eth_ctrl.h contributors), testpmd and PMD maintainers (with and without
> a .filter_ctrl implementation), but if you know application maintainers
> other than testpmd who use FDIR or might be interested in this discussion,
> feel free to add them.
>
> The issues we found with the current approach are already summarized in the
> following document, but here is a quick summary for TL;DR folks:
>
> - PMDs do not expose a common set of filter types and even when they do,
> their behavior more or less differs.
>
> - Applications need to determine and adapt to device-specific limitations
> and quirks on their own, without help from PMDs.
>
> - Writing an application that creates flow rules targeting all devices
> supported by DPDK is thus difficult, if not impossible.
>
> - The current API has too many unspecified areas (particularly regarding
> side effects of flow rules) that make PMD implementation tricky.
>
> This RFC API handles everything currently supported by .filter_ctrl, the
> idea being to reimplement all of these to make them fully usable by
> applications in a more generic and well defined fashion. It has a very small
> set of mandatory features and an easy method to let applications probe for
> supported capabilities.
>
> The only downside is more work for the software control side of PMDs because
> they have to adapt to the API instead of the reverse. I think helpers can be
> added to EAL to assist with this.
>
> HTML version:
>
> https://rawgit.com/6WIND/rte_flow/master/rte_flow.html
>
> PDF version:
>
> https://rawgit.com/6WIND/rte_flow/master/rte_flow.pdf
>
> Related draft header file (for reference while reading the specification):
>
> https://raw.githubusercontent.com/6WIND/rte_flow/master/rte_flow.h
>
> Git tree for completeness (latest .rst version can be retrieved from here):
>
> https://github.com/6WIND/rte_flow
>
> What follows is the ReST source of the above, for inline comments and
> discussion. I intend to update that specification accordingly.
>
> ========================
> Generic filter interface
> ========================
>
> .. footer::
>
> v0.6
>
> .. contents::
> .. sectnum::
> .. raw:: pdf
>
> PageBreak
>
> Overview
> ========
>
> DPDK provides several competing interfaces added over time to perform packet
> matching and related actions such as filtering and classification.
>
> They must be extended to implement the features supported by newer devices
> in order to expose them to applications, however the current design has
> several drawbacks:
>
> - Complicated filter combinations which have not been hard-coded cannot be
> expressed.
> - Prone to API/ABI breakage when new features must be added to an existing
> filter type, which frequently happens.
>
> From an application point of view:
>
> - Having disparate interfaces, all optional and lacking in features does not
> make this API easy to use.
> - Seemingly arbitrary built-in limitations of filter types based on the
> device they were initially designed for.
> - Undefined relationship between different filter types.
> - High complexity, considerable undocumented and/or undefined behavior.
>
> Considering the growing number of devices supported by DPDK, adding a new
> filter type each time a new feature must be implemented is not sustainable
> in the long term. Applications not written to target a specific device
> cannot really benefit from such an API.
>
> For these reasons, this document defines an extensible unified API that
> encompasses and supersedes these legacy filter types.
>
> .. raw:: pdf
>
> PageBreak
>
> Current API
> ===========
>
> Rationale
> ---------
>
> The reason several competing (and mostly overlapping) filtering APIs are
> present in DPDK is due to its nature as a thin layer between hardware and
> software.
>
> Each subsequent interface has been added to better match the capabilities
> and limitations of the latest supported device, which usually happened to
> need an incompatible configuration approach. Because of this, many ended up
> device-centric and not usable by applications that were not written for that
> particular device.
>
> This document is not the first attempt to address this proliferation issue,
> in fact a lot of work has already been done both to create a more generic
> interface while somewhat keeping compatibility with legacy ones through a
> common call interface (``rte_eth_dev_filter_ctrl()`` with the
> ``.filter_ctrl`` PMD callback in ``rte_ethdev.h``).
>
> Today, these previously incompatible interfaces are known as filter types
> (``RTE_ETH_FILTER_*`` from ``enum rte_filter_type`` in ``rte_eth_ctrl.h``).
>
> However while trivial to extend with new types, it only shifted the
> underlying problem as applications still need to be written for one kind of
> filter type, which, as described in the following sections, is not
> necessarily implemented by all PMDs that support filtering.
>
> .. raw:: pdf
>
> PageBreak
>
> Filter types
> ------------
>
> This section summarizes the capabilities of each filter type.
>
> Although the following list is exhaustive, the description of individual
> types may contain inaccuracies due to the lack of documentation or usage
> examples.
>
> Note: names are prefixed with ``RTE_ETH_FILTER_``.
>
> ``MACVLAN``
> ~~~~~~~~~~~
>
> Matching:
>
> - L2 source/destination addresses.
> - Optional 802.1Q VLAN ID.
> - Masking individual fields on a rule basis is not supported.
>
> Action:
>
> - Packets are redirected either to a given VF device using its ID or to the
> PF.
>
> ``ETHERTYPE``
> ~~~~~~~~~~~~~
>
> Matching:
>
> - L2 source/destination addresses (optional).
> - Ethertype (no VLAN ID?).
> - Masking individual fields on a rule basis is not supported.
>
> Action:
>
> - Receive packets on a given queue.
> - Drop packets.
>
> ``FLEXIBLE``
> ~~~~~~~~~~~~
>
> Matching:
>
> - At most 128 consecutive bytes anywhere in packets.
> - Masking is supported with byte granularity.
> - Priorities are supported (relative to this filter type, undefined
> otherwise).
>
> Action:
>
> - Receive packets on a given queue.
>
> ``SYN``
> ~~~~~~~
>
> Matching:
>
> - TCP SYN packets only.
> - One high priority bit can be set to give the highest possible priority to
> this type when other filters with different types are configured.
>
> Action:
>
> - Receive packets on a given queue.
>
> ``NTUPLE``
> ~~~~~~~~~~
>
> Matching:
>
> - Source/destination IPv4 addresses (optional in 2-tuple mode).
> - Source/destination TCP/UDP port (mandatory in 2 and 5-tuple modes).
> - L4 protocol (2 and 5-tuple modes).
> - Masking individual fields is supported.
> - TCP flags.
> - Up to 7 levels of priority relative to this filter type, undefined
> otherwise.
> - No IPv6.
>
> Action:
>
> - Receive packets on a given queue.
>
> ``TUNNEL``
> ~~~~~~~~~~
>
> Matching:
>
> - Outer L2 source/destination addresses.
> - Inner L2 source/destination addresses.
> - Inner VLAN ID.
> - IPv4/IPv6 source (destination?) address.
> - Tunnel type to match (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE, 802.1BR
> E-Tag).
> - Tenant ID for tunneling protocols that have one.
> - Any combination of the above can be specified.
> - Masking individual fields on a rule basis is not supported.
>
> Action:
>
> - Receive packets on a given queue.
>
> .. raw:: pdf
>
> PageBreak
>
> ``FDIR``
> ~~~~~~~~
>
> Queries:
>
> - Device capabilities and limitations.
> - Device statistics about configured filters (resource usage, collisions).
> - Device configuration (matching input set and masks)
>
> Matching:
>
> - Device mode of operation: none (to disable filtering), signature
> (hash-based dispatching from masked fields) or perfect (either MAC VLAN or
> tunnel).
> - L2 Ethertype.
> - Outer L2 destination address (MAC VLAN mode).
> - Inner L2 destination address, tunnel type (NVGRE, VXLAN) and tunnel ID
> (tunnel mode).
> - IPv4 source/destination addresses, ToS, TTL and protocol fields.
> - IPv6 source/destination addresses, TC, protocol and hop limits fields.
> - UDP source/destination IPv4/IPv6 and ports.
> - TCP source/destination IPv4/IPv6 and ports.
> - SCTP source/destination IPv4/IPv6, ports and verification tag field.
> - Note, only one protocol type at once (either only L2 Ethertype, basic
> IPv6, IPv4+UDP, IPv4+TCP and so on).
> - VLAN TCI (extended API).
> - At most 16 bytes to match in payload (extended API). A global device
> look-up table specifies for each possible protocol layer (unknown, raw,
> L2, L3, L4) the offset to use for each byte (they do not need to be
> contiguous) and the related bitmask.
> - Whether packet is addressed to PF or VF, in that case its ID can be
> matched as well (extended API).
> - Masking most of the above fields is supported, but simultaneously affects
> all filters configured on a device.
> - Input set can be modified in a similar fashion for a given device to
> ignore individual fields of filters (i.e. do not match the destination
> address in a IPv4 filter, refer to **RTE_ETH_INPUT_SET_**
> macros). Configuring this also affects RSS processing on **i40e**.
> - Filters can also provide 32 bits of arbitrary data to return as part of
> matched packets.
>
> Action:
>
> - **RTE_ETH_FDIR_ACCEPT**: receive (accept) packet on a given queue.
> - **RTE_ETH_FDIR_REJECT**: drop packet immediately.
> - **RTE_ETH_FDIR_PASSTHRU**: similar to accept for the last filter in list,
> otherwise process it with subsequent filters.
> - For accepted packets and if requested by filter, either 32 bits of
> arbitrary data and four bytes of matched payload (only in case of flex
> bytes matching), or eight bytes of matched payload (flex also) are added
> to meta data.
>
> .. raw:: pdf
>
> PageBreak
>
> ``HASH``
> ~~~~~~~~
>
> Not an actual filter type. Provides and retrieves the global device
> configuration (per port or entire NIC) for hash functions and their
> properties.
>
> Hash function selection: "default" (keep current), XOR or Toeplitz.
>
> This function can be configured per flow type (**RTE_ETH_FLOW_**
> definitions), supported types are:
>
> - Unknown.
> - Raw.
> - Fragmented or non-fragmented IPv4.
> - Non-fragmented IPv4 with L4 (TCP, UDP, SCTP or other).
> - Fragmented or non-fragmented IPv6.
> - Non-fragmented IPv6 with L4 (TCP, UDP, SCTP or other).
> - L2 payload.
> - IPv6 with extensions.
> - IPv6 with L4 (TCP, UDP) and extensions.
>
> ``L2_TUNNEL``
> ~~~~~~~~~~~~~
>
> Matching:
>
> - All packets received on a given port.
>
> Action:
>
> - Add tunnel encapsulation (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE,
> 802.1BR E-Tag) using the provided Ethertype and tunnel ID (only E-Tag
> is implemented at the moment).
> - VF ID to use for tag insertion (currently unused).
> - Destination pool for tag based forwarding (pools are IDs that can be
> affected to ports, duplication occurs if the same ID is shared by several
> ports of the same NIC).
>
> .. raw:: pdf
>
> PageBreak
>
> Driver support
> --------------
>
> ======== ======= ========= ======== === ====== ====== ==== ==== =========
> Driver MACVLAN ETHERTYPE FLEXIBLE SYN NTUPLE TUNNEL FDIR HASH L2_TUNNEL
> ======== ======= ========= ======== === ====== ====== ==== ==== =========
> bnx2x
> cxgbe
> e1000 yes yes yes yes
> ena
> enic yes
> fm10k
> i40e yes yes yes yes yes
> ixgbe yes yes yes yes yes
> mlx4
> mlx5 yes
> szedata2
> ======== ======= ========= ======== === ====== ====== ==== ==== =========
>
> Flow director
> -------------
>
> Flow director (FDIR) is the name of the most capable filter type, which
> covers most features offered by others. As such, it is the most widespread
> in PMDs that support filtering (i.e. all of them besides **e1000**).
>
> It is also the only type that allows an arbitrary 32 bits value provided by
> applications to be attached to a filter and returned with matching packets
> instead of relying on the destination queue to recognize flows.
>
> Unfortunately, even FDIR requires applications to be aware of low-level
> capabilities and limitations (most of which come directly from **ixgbe** and
> **i40e**):
>
> - Bitmasks are set globally per device (port?), not per filter.
> - Configuration state is not expected to be saved by the driver, and
> stopping/restarting a port requires the application to perform it again
> (API documentation is also unclear about this).
> - Monolithic approach with ABI issues as soon as a new kind of flow or
> combination needs to be supported.
> - Cryptic global statistics/counters.
> - Unclear about how priorities are managed; filters seem to be arranged as a
> linked list in hardware (possibly related to configuration order).
>
> Packet alteration
> -----------------
>
> One interesting feature is that the L2 tunnel filter type implements the
> ability to alter incoming packets through a filter (in this case to
> encapsulate them), thus the **mlx5** flow encap/decap features are not a
> foreign concept.
>
> .. raw:: pdf
>
> PageBreak
>
> Proposed API
> ============
>
> Terminology
> -----------
>
> - **Filtering API**: overall framework affecting the fate of selected
> packets, covers everything described in this document.
> - **Matching pattern**: properties to look for in received packets, a
> combination of any number of items.
> - **Pattern item**: part of a pattern that either matches packet data
> (protocol header, payload or derived information), or specifies properties
> of the pattern itself.
> - **Actions**: what needs to be done when a packet matches a pattern.
> - **Flow rule**: this is the result of combining a *matching pattern* with
> *actions*.
> - **Filter rule**: a less generic term than *flow rule*, can otherwise be
> used interchangeably.
> - **Hit**: a flow rule is said to be *hit* when processing a matching
> packet.
>
> Requirements
> ------------
>
> As described in the previous section, there is a growing need for a common
> method to configure filtering and related actions in a hardware independent
> fashion.
>
> The filtering API should not disallow any filter combination by design and
> must remain as simple as possible to use. It can simply be defined as a
> method to perform one or several actions on selected packets.
>
> PMDs are aware of the capabilities of the device they manage and should be
> responsible for preventing unsupported or conflicting combinations.
>
> This approach is fundamentally different as it places most of the burden on
> the software side of the PMD instead of having device capabilities directly
> mapped to API functions, then expecting applications to work around ensuing
> compatibility issues.
>
> Requirements for a new API:
>
> - Flexible and extensible without causing API/ABI problems for existing
> applications.
> - Should be unambiguous and easy to use.
> - Support existing filtering features and actions listed in `Filter types`_.
> - Support packet alteration.
> - In case of overlapping filters, their priority should be well documented.
> - Support filter queries (for example to retrieve counters).
>
> .. raw:: pdf
>
> PageBreak
>
> High level design
> -----------------
>
> The chosen approach to make filtering as generic as possible is by
> expressing matching patterns through lists of items instead of the flat
> structures used in DPDK today, enabling combinations that are not predefined
> and thus being more versatile.
>
> Flow rules can have several distinct actions (such as counting,
> encapsulating, decapsulating before redirecting packets to a particular
> queue, etc.), instead of relying on several rules to achieve this and having
> applications deal with hardware implementation details regarding their
> order.
>
> Support for different priority levels on a rule basis is provided, for
> example in order to force a more specific rule come before a more generic
> one for packets matched by both, however hardware support for more than a
> single priority level cannot be guaranteed. When supported, the number of
> available priority levels is usually low, which is why they can also be
> implemented in software by PMDs (e.g. to simulate missing priority levels by
> reordering rules).
>
> In order to remain as hardware agnostic as possible, by default all rules
> are considered to have the same priority, which means that the order between
> overlapping rules (when a packet is matched by several filters) is
> undefined, packet duplication may even occur as a result.
>
> PMDs may refuse to create overlapping rules at a given priority level when
> they can be detected (e.g. if a pattern matches an existing filter).
>
> Thus predictable results for a given priority level can only be achieved
> with non-overlapping rules, using perfect matching on all protocol layers.
>
> Support for multiple actions per rule may be implemented internally on top
> of non-default hardware priorities, as a result both features may not be
> simultaneously available to applications.
>
> Considering that allowed pattern/actions combinations cannot be known in
> advance and would result in an unpractically large number of capabilities to
> expose, a method is provided to validate a given rule from the current
> device configuration state without actually adding it (akin to a "dry run"
> mode).
>
> This enables applications to check if the rule types they need is supported
> at initialization time, before starting their data path. This method can be
> used anytime, its only requirement being that the resources needed by a rule
> must exist (e.g. a target RX queue must be configured first).
>
> Each defined rule is associated with an opaque handle managed by the PMD,
> applications are responsible for keeping it. These can be used for queries
> and rules management, such as retrieving counters or other data and
> destroying them.
>
> Handles must be destroyed before releasing associated resources such as
> queues.
>
> Integration
> -----------
>
> To avoid ABI breakage, this new interface will be implemented through the
> existing filtering control framework (``rte_eth_dev_filter_ctrl()``) using
> **RTE_ETH_FILTER_GENERIC** as a new filter type.
>
> However a public front-end API described in `Rules management`_ will
> be added as the preferred method to use it.
>
> Once discussions with the community have converged to a definite API, legacy
> filter types should be deprecated and a deadline defined to remove their
> support entirely.
>
> PMDs will have to be gradually converted to **RTE_ETH_FILTER_GENERIC** or
> drop filtering support entirely. Less maintained PMDs for older hardware may
> lose support at this point.
>
> The notion of filter type will then be deprecated and subsequently dropped
> to avoid confusion between both frameworks.
>
> Implementation details
> ======================
>
> Flow rule
> ---------
>
> A flow rule is the combination of a matching pattern with a list of actions,
> and is the basis of this API.
>
> Priorities
> ~~~~~~~~~~
>
> A priority can be assigned to a matching pattern.
>
> The default priority level is 0 and is also the highest. Support for more
> than a single priority level in hardware is not guaranteed.
>
> If a packet is matched by several filters at a given priority level, the
> outcome is undefined. It can take any path and can even be duplicated.
>
> Matching pattern
> ~~~~~~~~~~~~~~~~
>
> A matching pattern comprises any number of items of various types.
>
> Items are arranged in a list to form a matching pattern for packets. They
> fall in two categories:
>
> - Protocol matching (ANY, RAW, ETH, IPV4, IPV6, ICMP, UDP, TCP, VXLAN and so
> on), usually associated with a specification structure. These must be
> stacked in the same order as the protocol layers to match, starting from
> L2.
>
> - Affecting how the pattern is processed (END, VOID, INVERT, PF, VF,
> SIGNATURE and so on), often without a specification structure. Since they
> are meta data that does not match packet contents, these can be specified
> anywhere within item lists without affecting the protocol matching items.
>
> Most item specifications can be optionally paired with a mask to narrow the
> specific fields or bits to be matched.
>
> - Items are defined with ``struct rte_flow_item``.
> - Patterns are defined with ``struct rte_flow_pattern``.
>
> Example of an item specification matching an Ethernet header:
>
> +-----------------------------------------+
> | Ethernet |
> +==========+=========+====================+
> | ``spec`` | ``src`` | ``00:01:02:03:04`` |
> | +---------+--------------------+
> | | ``dst`` | ``00:2a:66:00:01`` |
> +----------+---------+--------------------+
> | ``mask`` | ``src`` | ``00:ff:ff:ff:00`` |
> | +---------+--------------------+
> | | ``dst`` | ``00:00:00:00:ff`` |
> +----------+---------+--------------------+
>
> Non-masked bits stand for any value, Ethernet headers with the following
> properties are thus matched:
>
> - ``src``: ``??:01:02:03:??``
> - ``dst``: ``??:??:??:??:01``
>
> Except for meta types that do not need one, ``spec`` must be a valid pointer
> to a structure of the related item type. A ``mask`` of the same type can be
> provided to tell which bits in ``spec`` are to be matched.
>
> A mask is normally only needed for ``spec`` fields matching packet data,
> ignored otherwise. See individual item types for more information.
>
> A ``NULL`` mask pointer is allowed and is similar to matching with a full
> mask (all ones) ``spec`` fields supported by hardware, the remaining fields
> are ignored (all zeroes), there is thus no error checking for unsupported
> fields.
>
> Matching pattern items for packet data must be naturally stacked (ordered
> from lowest to highest protocol layer), as in the following examples:
>
> +--------------+
> | TCPv4 as L4 |
> +===+==========+
> | 0 | Ethernet |
> +---+----------+
> | 1 | IPv4 |
> +---+----------+
> | 2 | TCP |
> +---+----------+
>
> +----------------+
> | TCPv6 in VXLAN |
> +===+============+
> | 0 | Ethernet |
> +---+------------+
> | 1 | IPv4 |
> +---+------------+
> | 2 | UDP |
> +---+------------+
> | 3 | VXLAN |
> +---+------------+
> | 4 | Ethernet |
> +---+------------+
> | 5 | IPv6 |
> +---+------------+
> | 6 | TCP |
> +---+------------+
>
> +-----------------------------+
> | TCPv4 as L4 with meta items |
> +===+=========================+
> | 0 | VOID |
> +---+-------------------------+
> | 1 | Ethernet |
> +---+-------------------------+
> | 2 | VOID |
> +---+-------------------------+
> | 3 | IPv4 |
> +---+-------------------------+
> | 4 | TCP |
> +---+-------------------------+
> | 5 | VOID |
> +---+-------------------------+
> | 6 | VOID |
> +---+-------------------------+
>
> The above example shows how meta items do not affect packet data matching
> items, as long as those remain stacked properly. The resulting matching
> pattern is identical to "TCPv4 as L4".
>
> +----------------+
> | UDPv6 anywhere |
> +===+============+
> | 0 | IPv6 |
> +---+------------+
> | 1 | UDP |
> +---+------------+
>
> If supported by the PMD, omitting one or several protocol layers at the
> bottom of the stack as in the above example (missing an Ethernet
> specification) enables hardware to look anywhere in packets.
>
> It is unspecified whether the payload of supported encapsulations
> (e.g. VXLAN inner packet) is matched by such a pattern, which may apply to
> inner, outer or both packets.
>
> +---------------------+
> | Invalid, missing L3 |
> +===+=================+
> | 0 | Ethernet |
> +---+-----------------+
> | 1 | UDP |
> +---+-----------------+
>
> The above pattern is invalid due to a missing L3 specification between L2
> and L4. It is only allowed at the bottom and at the top of the stack.
>
> Meta item types
> ~~~~~~~~~~~~~~~
>
> These do not match packet data but affect how the pattern is processed, most
> of them do not need a specification structure. This particularity allows
> them to be specified anywhere without affecting other item types.
[LC] For the meta item(END, VOID, INVERT) and some data matching type
like ANY and RAW,
it's all PMD responsible to understand the key character and to parse
the header graph?
>
> ``END``
> ^^^^^^^
>
> End marker for item lists. Prevents further processing of items, thereby
> ending the pattern.
>
> - Its numeric value is **0** for convenience.
> - PMD support is mandatory.
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | END |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> ``VOID``
> ^^^^^^^^
>
> Used as a placeholder for convenience. It is ignored and simply discarded by
> PMDs.
>
> - PMD support is mandatory.
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | VOID |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> One usage example for this type is generating rules that share a common
> prefix quickly without reallocating memory, only by updating item types:
>
> +------------------------+
> | TCP, UDP or ICMP as L4 |
> +===+====================+
> | 0 | Ethernet |
> +---+--------------------+
> | 1 | IPv4 |
> +---+------+------+------+
> | 2 | UDP | VOID | VOID |
> +---+------+------+------+
> | 3 | VOID | TCP | VOID |
> +---+------+------+------+
> | 4 | VOID | VOID | ICMP |
> +---+------+------+------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``INVERT``
> ^^^^^^^^^^
>
> Inverted matching, i.e. process packets that do not match the pattern.
>
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | INVERT |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> Usage example in order to match non-TCPv4 packets only:
>
> +--------------------+
> | Anything but TCPv4 |
> +===+================+
> | 0 | INVERT |
> +---+----------------+
> | 1 | Ethernet |
> +---+----------------+
> | 2 | IPv4 |
> +---+----------------+
> | 3 | TCP |
> +---+----------------+
>
> ``PF``
> ^^^^^^
>
> Matches packets addressed to the physical function of the device.
>
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | PF |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> ``VF``
> ^^^^^^
>
> Matches packets addressed to the given virtual function ID of the device.
>
> - Only ``spec`` needs to be defined, ``mask`` is ignored.
>
> +----------------------------------------+
> | VF |
> +==========+=========+===================+
> | ``spec`` | ``vf`` | destination VF ID |
> +----------+---------+-------------------+
> | ``mask`` | ignored |
> +----------+-----------------------------+
>
> ``SIGNATURE``
> ^^^^^^^^^^^^^
>
> Requests hash-based signature dispatching for this rule.
>
> Considering this is a global setting on devices that support it, all
> subsequent filter rules may have to be created with it as well.
>
> - Only ``spec`` needs to be defined, ``mask`` is ignored.
>
> +--------------------+
> | SIGNATURE |
> +==========+=========+
> | ``spec`` | TBD |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> .. raw:: pdf
>
> PageBreak
>
> Data matching item types
> ~~~~~~~~~~~~~~~~~~~~~~~~
>
> Most of these are basically protocol header definitions with associated
> bitmasks. They must be specified (stacked) from lowest to highest protocol
> layer.
>
> The following list is not exhaustive as new protocols will be added in the
> future.
>
> ``ANY``
> ^^^^^^^
>
> Matches any protocol in place of the current layer, a single ANY may also
> stand for several protocol layers.
>
> This is usually specified as the first pattern item when looking for a
> protocol anywhere in a packet.
>
> - A maximum value of **0** requests matching any number of protocol layers
> above or equal to the minimum value, a maximum value lower than the
> minimum one is otherwise invalid.
> - Only ``spec`` needs to be defined, ``mask`` is ignored.
>
> +-----------------------------------------------------------------------+
> | ANY |
> +==========+=========+==================================================+
> | ``spec`` | ``min`` | minimum number of layers covered |
> | +---------+--------------------------------------------------+
> | | ``max`` | maximum number of layers covered, 0 for infinity |
> +----------+---------+--------------------------------------------------+
> | ``mask`` | ignored |
> +----------+------------------------------------------------------------+
>
> Example for VXLAN TCP payload matching regardless of outer L3 (IPv4 or IPv6)
> and L4 (UDP) both matched by the first ANY specification, and inner L3 (IPv4
> or IPv6) matched by the second ANY specification:
>
> +----------------------------------+
> | TCP in VXLAN with wildcards |
> +===+==============================+
> | 0 | Ethernet |
> +---+-----+----------+---------+---+
> | 1 | ANY | ``spec`` | ``min`` | 2 |
> | | | +---------+---+
> | | | | ``max`` | 2 |
> +---+-----+----------+---------+---+
> | 2 | VXLAN |
> +---+------------------------------+
> | 3 | Ethernet |
> +---+-----+----------+---------+---+
> | 4 | ANY | ``spec`` | ``min`` | 1 |
> | | | +---------+---+
> | | | | ``max`` | 1 |
> +---+-----+----------+---------+---+
> | 5 | TCP |
> +---+------------------------------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``RAW``
> ^^^^^^^
>
> Matches a string of a given length at a given offset (in bytes), or anywhere
> in the payload of the current protocol layer (including L2 header if used as
> the first item in the stack).
>
> This does not increment the protocol layer count as it is not a protocol
> definition. Subsequent RAW items modulate the first absolute one with
> relative offsets.
>
> - Using **-1** as the ``offset`` of the first RAW item makes its absolute
> offset not fixed, i.e. the pattern is searched everywhere.
> - ``mask`` only affects the pattern.
The RAW matching type allow offset & length which support anchor setting
setting and string match.
It's not defined for a user defined packet layout. Sometimes, comparing
payload raw data after a header require
{offset, length}. One typical case is 5-tuples matching. The 'PORT' of
transport layer is an offset to the IP header.
It can't address by IP/ANY, as it requires to extract key from the field
in ANY.
>
> +--------------------------------------------------------------+
> | RAW |
> +==========+=============+=====================================+
> | ``spec`` | ``offset`` | absolute or relative pattern offset |
> | +-------------+-------------------------------------+
> | | ``length`` | pattern length |
> | +-------------+-------------------------------------+
> | | ``pattern`` | byte string of the above length |
> +----------+-------------+-------------------------------------+
> | ``mask`` | ``offset`` | ignored |
> | +-------------+-------------------------------------+
> | | ``length`` | ignored |
> | +-------------+-------------------------------------+
> | | ``pattern`` | bitmask with the same byte length |
> +----------+-------------+-------------------------------------+
>
> Example pattern looking for several strings at various offsets of a UDP
> payload, using combined RAW items:
>
> +------------------------------------------+
> | UDP payload matching |
> +===+======================================+
> | 0 | Ethernet |
> +---+--------------------------------------+
> | 1 | IPv4 |
> +---+--------------------------------------+
> | 2 | UDP |
> +---+-----+----------+-------------+-------+
> | 3 | RAW | ``spec`` | ``offset`` | -1 |
> | | | +-------------+-------+
> | | | | ``length`` | 3 |
> | | | +-------------+-------+
> | | | | ``pattern`` | "foo" |
> +---+-----+----------+-------------+-------+
> | 4 | RAW | ``spec`` | ``offset`` | 20 |
> | | | +-------------+-------+
> | | | | ``length`` | 3 |
> | | | +-------------+-------+
> | | | | ``pattern`` | "bar" |
> +---+-----+----------+-------------+-------+
> | 5 | RAW | ``spec`` | ``offset`` | -30 |
> | | | +-------------+-------+
> | | | | ``length`` | 3 |
> | | | +-------------+-------+
> | | | | ``pattern`` | "baz" |
> +---+-----+----------+-------------+-------+
>
> This translates to:
>
> - Locate "foo" in UDP payload, remember its offset.
> - Check "bar" at "foo"'s offset plus 20 bytes.
> - Check "baz" at "foo"'s offset minus 30 bytes.
>
> .. raw:: pdf
>
> PageBreak
>
> ``ETH``
> ^^^^^^^
>
> Matches an Ethernet header.
>
> - ``dst``: destination MAC.
> - ``src``: source MAC.
> - ``type``: EtherType.
> - ``tags``: number of 802.1Q/ad tags defined.
> - ``tag[]``: 802.1Q/ad tag definitions, innermost first. For each one:
>
> - ``tpid``: Tag protocol identifier.
> - ``tci``: Tag control information.
>
> ``IPV4``
> ^^^^^^^^
>
> Matches an IPv4 header.
>
> - ``src``: source IP address.
> - ``dst``: destination IP address.
> - ``tos``: ToS/DSCP field.
> - ``ttl``: TTL field.
> - ``proto``: protocol number for the next layer.
>
> ``IPV6``
> ^^^^^^^^
>
> Matches an IPv6 header.
>
> - ``src``: source IP address.
> - ``dst``: destination IP address.
> - ``tc``: traffic class field.
> - ``nh``: Next header field (protocol).
> - ``hop_limit``: hop limit field (TTL).
>
> ``ICMP``
> ^^^^^^^^
>
> Matches an ICMP header.
>
> - TBD.
>
> ``UDP``
> ^^^^^^^
>
> Matches a UDP header.
>
> - ``sport``: source port.
> - ``dport``: destination port.
> - ``length``: UDP length.
> - ``checksum``: UDP checksum.
>
> .. raw:: pdf
>
> PageBreak
>
> ``TCP``
> ^^^^^^^
>
> Matches a TCP header.
>
> - ``sport``: source port.
> - ``dport``: destination port.
> - All other TCP fields and bits.
>
> ``VXLAN``
> ^^^^^^^^^
>
> Matches a VXLAN header.
>
> - TBD.
>
> .. raw:: pdf
>
> PageBreak
>
> Actions
> ~~~~~~~
>
> Each possible action is represented by a type. Some have associated
> configuration structures. Several actions combined in a list can be affected
> to a flow rule. That list is not ordered.
>
> At least one action must be defined in a filter rule in order to do
> something with matched packets.
>
> - Actions are defined with ``struct rte_flow_action``.
> - A list of actions is defined with ``struct rte_flow_actions``.
>
> They fall in three categories:
>
> - Terminating actions (such as QUEUE, DROP, RSS, PF, VF) that prevent
> processing matched packets by subsequent flow rules, unless overridden
> with PASSTHRU.
>
> - Non terminating actions (PASSTHRU, DUP) that leave matched packets up for
> additional processing by subsequent flow rules.
>
> - Other non terminating meta actions that do not affect the fate of packets
> (END, VOID, ID, COUNT).
>
> When several actions are combined in a flow rule, they should all have
> different types (e.g. dropping a packet twice is not possible). However
> considering the VOID type is an exception to this rule, the defined behavior
> is for PMDs to only take into account the last action of a given type found
> in the list. PMDs still perform error checking on the entire list.
>
> *Note that PASSTHRU is the only action able to override a terminating rule.*
[LC] I'm wondering how to address the meta data carried by mbuf, there's
no mentioned here.
For packets hit one specific flow, usually there's something for CPU to
identify the flow.
FDIR and RSS as an example, has id or key in mbuf. In addition, some
meta may pointed by userdata in mbuf.
Any view on it ?
>
> .. raw:: pdf
>
> PageBreak
>
> Example of an action that redirects packets to queue index 10:
>
> +----------------+
> | QUEUE |
> +===========+====+
> | ``queue`` | 10 |
> +-----------+----+
>
> Action lists examples, their order is not significant, applications must
> consider all actions to be performed simultaneously:
>
> +----------------+
> | Count and drop |
> +=======+========+
> | COUNT | |
> +-------+--------+
> | DROP | |
> +-------+--------+
>
> +--------------------------+
> | Tag, count and redirect |
> +=======+===========+======+
> | ID | ``id`` | 0x2a |
> +-------+-----------+------+
> | COUNT | |
> +-------+-----------+------+
> | QUEUE | ``queue`` | 10 |
> +-------+-----------+------+
>
> +-----------------------+
> | Redirect to queue 5 |
> +=======+===============+
> | DROP | |
> +-------+-----------+---+
> | QUEUE | ``queue`` | 5 |
> +-------+-----------+---+
>
> In the above example, considering both actions are performed simultaneously,
> its end result is that only QUEUE has any effect.
>
> +-----------------------+
> | Redirect to queue 3 |
> +=======+===========+===+
> | QUEUE | ``queue`` | 5 |
> +-------+-----------+---+
> | VOID | |
> +-------+-----------+---+
> | QUEUE | ``queue`` | 3 |
> +-------+-----------+---+
>
> As previously described, only the last action of a given type found in the
> list is taken into account. The above example also shows that VOID is
> ignored.
>
> .. raw:: pdf
>
> PageBreak
>
> Action types
> ~~~~~~~~~~~~
>
> Common action types are described in this section. Like pattern item types,
> this list is not exhaustive as new actions will be added in the future.
>
> ``END`` (action)
> ^^^^^^^^^^^^^^^^
>
> End marker for action lists. Prevents further processing of actions, thereby
> ending the list.
>
> - Its numeric value is **0** for convenience.
> - PMD support is mandatory.
> - No configurable property.
>
> +---------------+
> | END |
> +===============+
> | no properties |
> +---------------+
>
> ``VOID`` (action)
> ^^^^^^^^^^^^^^^^^
>
> Used as a placeholder for convenience. It is ignored and simply discarded by
> PMDs.
>
> - PMD support is mandatory.
> - No configurable property.
>
> +---------------+
> | VOID |
> +===============+
> | no properties |
> +---------------+
>
> ``PASSTHRU``
> ^^^^^^^^^^^^
>
> Leaves packets up for additional processing by subsequent flow rules. This
> is the default when a rule does not contain a terminating action, but can be
> specified to force a rule to become non-terminating.
>
> - No configurable property.
>
> +---------------+
> | PASSTHRU |
> +===============+
> | no properties |
> +---------------+
>
> Example to copy a packet to a queue and continue processing by subsequent
> flow rules:
>
> +--------------------------+
> | Copy to queue 8 |
> +==========+===============+
> | PASSTHRU | |
> +----------+-----------+---+
> | QUEUE | ``queue`` | 8 |
> +----------+-----------+---+
>
> ``ID``
> ^^^^^^
>
> Attaches a 32 bit value to packets.
>
> +----------------------------------------------+
> | ID |
> +========+=====================================+
> | ``id`` | 32 bit value to return with packets |
> +--------+-------------------------------------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``QUEUE``
> ^^^^^^^^^
>
> Assigns packets to a given queue index.
>
> - Terminating by default.
>
> +--------------------------------+
> | QUEUE |
> +===========+====================+
> | ``queue`` | queue index to use |
> +-----------+--------------------+
>
> ``DROP``
> ^^^^^^^^
>
> Drop packets.
>
> - No configurable property.
> - Terminating by default.
> - PASSTHRU overrides this action if both are specified.
>
> +---------------+
> | DROP |
> +===============+
> | no properties |
> +---------------+
>
> ``COUNT``
> ^^^^^^^^^
>
> Enables hits counter for this rule.
>
> This counter can be retrieved and reset through ``rte_flow_query()``, see
> ``struct rte_flow_query_count``.
>
> - Counters can be retrieved with ``rte_flow_query()``.
> - No configurable property.
>
> +---------------+
> | COUNT |
> +===============+
> | no properties |
> +---------------+
>
> Query structure to retrieve and reset the flow rule hits counter:
>
> +------------------------------------------------+
> | COUNT query |
> +===========+=====+==============================+
> | ``reset`` | in | reset counter after query |
> +-----------+-----+------------------------------+
> | ``hits`` | out | number of hits for this flow |
> +-----------+-----+------------------------------+
>
> ``DUP``
> ^^^^^^^
>
> Duplicates packets to a given queue index.
>
> This is normally combined with QUEUE, however when used alone, it is
> actually similar to QUEUE + PASSTHRU.
>
> - Non-terminating by default.
>
> +------------------------------------------------+
> | DUP |
> +===========+====================================+
> | ``queue`` | queue index to duplicate packet to |
> +-----------+------------------------------------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``RSS``
> ^^^^^^^
>
> Similar to QUEUE, except RSS is additionally performed on packets to spread
> them among several queues according to the provided parameters.
>
> - Terminating by default.
>
> +---------------------------------------------+
> | RSS |
> +==============+==============================+
> | ``rss_conf`` | RSS parameters |
> +--------------+------------------------------+
> | ``queues`` | number of entries in queue[] |
> +--------------+------------------------------+
> | ``queue[]`` | queue indices to use |
> +--------------+------------------------------+
>
> ``PF`` (action)
> ^^^^^^^^^^^^^^^
>
> Redirects packets to the physical function (PF) of the current device.
>
> - No configurable property.
> - Terminating by default.
>
> +---------------+
> | PF |
> +===============+
> | no properties |
> +---------------+
>
> ``VF`` (action)
> ^^^^^^^^^^^^^^^
>
> Redirects packets to the virtual function (VF) of the current device with
> the specified ID.
>
> - Terminating by default.
>
> +---------------------------------------+
> | VF |
> +========+==============================+
> | ``id`` | VF ID to redirect packets to |
> +--------+------------------------------+
>
> Planned types
> ~~~~~~~~~~~~~
>
> Other action types are planned but not defined yet. These actions will add
> the ability to alter matching packets in several ways, such as performing
> encapsulation/decapsulation of tunnel headers on specific flows.
>
> .. raw:: pdf
>
> PageBreak
>
> Rules management
> ----------------
>
> A simple API with only four functions is provided to fully manage flows.
>
> Each created flow rule is associated with an opaque, PMD-specific handle
> pointer. The application is responsible for keeping it until the rule is
> destroyed.
>
> Flows rules are defined with ``struct rte_flow``.
>
> Validation
> ~~~~~~~~~~
>
> Given that expressing a definite set of device capabilities with this API is
> not practical, a dedicated function is provided to check if a flow rule is
> supported and can be created.
>
> ::
>
> int
> rte_flow_validate(uint8_t port_id,
> const struct rte_flow_pattern *pattern,
> const struct rte_flow_actions *actions);
>
> While this function has no effect on the target device, the flow rule is
> validated against its current configuration state and the returned value
> should be considered valid by the caller for that state only.
>
> The returned value is guaranteed to remain valid only as long as no
> successful calls to rte_flow_create() or rte_flow_destroy() are made in the
> meantime and no device parameter affecting flow rules in any way are
> modified, due to possible collisions or resource limitations (although in
> such cases ``EINVAL`` should not be returned).
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``pattern``: pattern specification to check.
> - ``actions``: actions associated with the flow definition.
>
> Return value:
>
> - **0** if flow rule is valid and can be created. A negative errno value
> otherwise (``rte_errno`` is also set), the following errors are defined.
> - ``-EINVAL``: unknown or invalid rule specification.
> - ``-ENOTSUP``: valid but unsupported rule specification (e.g. partial masks
> are unsupported).
> - ``-EEXIST``: collision with an existing rule.
> - ``-ENOMEM``: not enough resources.
>
> .. raw:: pdf
>
> PageBreak
>
> Creation
> ~~~~~~~~
>
> Creating a flow rule is similar to validating one, except the rule is
> actually created.
>
> ::
>
> struct rte_flow *
> rte_flow_create(uint8_t port_id,
> const struct rte_flow_pattern *pattern,
> const struct rte_flow_actions *actions);
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``pattern``: pattern specification to add.
> - ``actions``: actions associated with the flow definition.
>
> Return value:
>
> A valid flow pointer in case of success, NULL otherwise and ``rte_errno`` is
> set to the positive version of one of the error codes defined for
> ``rte_flow_validate()``.
>
> Destruction
> ~~~~~~~~~~~
>
> Flow rules destruction is not automatic, and a queue should not be released
> if any are still attached to it. Applications must take care of performing
> this step before releasing resources.
>
> ::
>
> int
> rte_flow_destroy(uint8_t port_id,
> struct rte_flow *flow);
>
>
> Failure to destroy a flow rule may occur when other flow rules depend on it,
> and destroying it would result in an inconsistent state.
>
> This function is only guaranteed to succeed if flow rules are destroyed in
> reverse order of their creation.
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``flow``: flow rule to destroy.
>
> Return value:
>
> - **0** on success, a negative errno value otherwise and ``rte_errno`` is
> set.
>
> .. raw:: pdf
>
> PageBreak
>
> Query
> ~~~~~
>
> Query an existing flow rule.
>
> This function allows retrieving flow-specific data such as counters. Data
> is gathered by special actions which must be present in the flow rule
> definition.
>
> ::
>
> int
> rte_flow_query(uint8_t port_id,
> struct rte_flow *flow,
> enum rte_flow_action_type action,
> void *data);
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``flow``: flow rule to query.
> - ``action``: action type to query.
> - ``data``: pointer to storage for the associated query data type.
>
> Return value:
>
> - **0** on success, a negative errno value otherwise and ``rte_errno`` is
> set.
>
> .. raw:: pdf
>
> PageBreak
>
> Behavior
> --------
>
> - API operations are synchronous and blocking (``EAGAIN`` cannot be
> returned).
>
> - There is no provision for reentrancy/multi-thread safety, although nothing
> should prevent different devices from being configured at the same
> time. PMDs may protect their control path functions accordingly.
>
> - Stopping the data path (TX/RX) should not be necessary when managing flow
> rules. If this cannot be achieved naturally or with workarounds (such as
> temporarily replacing the burst function pointers), an appropriate error
> code must be returned (``EBUSY``).
>
> - PMDs, not applications, are responsible for maintaining flow rules
> configuration when stopping and restarting a port or performing other
> actions which may affect them. They can only be destroyed explicitly.
>
> .. raw:: pdf
>
> PageBreak
>
> Compatibility
> -------------
>
> No known hardware implementation supports all the features described in this
> document.
>
> Unsupported features or combinations are not expected to be fully emulated
> in software by PMDs for performance reasons. Partially supported features
> may be completed in software as long as hardware performs most of the work
> (such as queue redirection and packet recognition).
>
> However PMDs are expected to do their best to satisfy application requests
> by working around hardware limitations as long as doing so does not affect
> the behavior of existing flow rules.
>
> The following sections provide a few examples of such cases, they are based
> on limitations built into the previous APIs.
>
> Global bitmasks
> ~~~~~~~~~~~~~~~
>
> Each flow rule comes with its own, per-layer bitmasks, while hardware may
> support only a single, device-wide bitmask for a given layer type, so that
> two IPv4 rules cannot use different bitmasks.
>
> The expected behavior in this case is that PMDs automatically configure
> global bitmasks according to the needs of the first created flow rule.
>
> Subsequent rules are allowed only if their bitmasks match those, the
> ``EEXIST`` error code should be returned otherwise.
>
> Unsupported layer types
> ~~~~~~~~~~~~~~~~~~~~~~~
>
> Many protocols can be simulated by crafting patterns with the `RAW`_ type.
>
> PMDs can rely on this capability to simulate support for protocols with
> fixed headers not directly recognized by hardware.
>
> ``ANY`` pattern item
> ~~~~~~~~~~~~~~~~~~~~
>
> This pattern item stands for anything, which can be difficult to translate
> to something hardware would understand, particularly if followed by more
> specific types.
>
> Consider the following pattern:
>
> +---+--------------------------------+
> | 0 | ETHER |
> +---+--------------------------------+
> | 1 | ANY (``min`` = 1, ``max`` = 1) |
> +---+--------------------------------+
> | 2 | TCP |
> +---+--------------------------------+
>
> Knowing that TCP does not make sense with something other than IPv4 and IPv6
> as L3, such a pattern may be translated to two flow rules instead:
>
> +---+--------------------+
> | 0 | ETHER |
> +---+--------------------+
> | 1 | IPV4 (zeroed mask) |
> +---+--------------------+
> | 2 | TCP |
> +---+--------------------+
>
> +---+--------------------+
> | 0 | ETHER |
> +---+--------------------+
> | 1 | IPV6 (zeroed mask) |
> +---+--------------------+
> | 2 | TCP |
> +---+--------------------+
>
> Note that as soon as a ANY rule covers several layers, this approach may
> yield a large number of hidden flow rules. It is thus suggested to only
> support the most common scenarios (anything as L2 and/or L3).
>
> .. raw:: pdf
>
> PageBreak
>
> Unsupported actions
> ~~~~~~~~~~~~~~~~~~~
>
> - When combined with a `QUEUE`_ action, packet counting (`COUNT`_) and
> tagging (`ID`_) may be implemented in software as long as the target queue
> is used by a single rule.
>
> - A rule specifying both `DUP`_ + `QUEUE`_ may be translated to two hidden
> rules combining `QUEUE`_ and `PASSTHRU`_.
>
> - When a single target queue is provided, `RSS`_ can also be implemented
> through `QUEUE`_.
>
> Flow rules priority
> ~~~~~~~~~~~~~~~~~~~
>
> While it would naturally make sense, flow rules cannot be assumed to be
> processed by hardware in the same order as their creation for several
> reasons:
>
> - They may be managed internally as a tree or a hash table instead of a
> list.
> - Removing a flow rule before adding another one can either put the new rule
> at the end of the list or reuse a freed entry.
> - Duplication may occur when packets are matched by several rules.
>
> For overlapping rules (particularly in order to use the `PASSTHRU`_ action)
> predictable behavior is only guaranteed by using different priority levels.
>
> Priority levels are not necessarily implemented in hardware, or may be
> severely limited (e.g. a single priority bit).
>
> For these reasons, priority levels may be implemented purely in software by
> PMDs.
>
> - For devices expecting flow rules to be added in the correct order, PMDs
> may destroy and re-create existing rules after adding a new one with
> a higher priority.
>
> - A configurable number of dummy or empty rules can be created at
> initialization time to save high priority slots for later.
>
> - In order to save priority levels, PMDs may evaluate whether rules are
> likely to collide and adjust their priority accordingly.
>
> .. raw:: pdf
>
> PageBreak
>
> API migration
> =============
>
> Exhaustive list of deprecated filter types and how to convert them to
> generic flow rules.
>
> ``MACVLAN`` to ``ETH`` → ``VF``, ``PF``
> ---------------------------------------
>
> `MACVLAN`_ can be translated to a basic `ETH`_ flow rule with a `VF
> (action)`_ or `PF (action)`_ terminating action.
>
> +------------------------------------+
> | MACVLAN |
> +--------------------------+---------+
> | Pattern | Actions |
> +===+=====+==========+=====+=========+
> | 0 | ETH | ``spec`` | any | VF, |
> | | +----------+-----+ PF |
> | | | ``mask`` | any | |
> +---+-----+----------+-----+---------+
>
> ``ETHERTYPE`` to ``ETH`` → ``QUEUE``, ``DROP``
> ----------------------------------------------
>
> `ETHERTYPE`_ is basically an `ETH`_ flow rule with `QUEUE`_ or `DROP`_ as
> a terminating action.
>
> +------------------------------------+
> | ETHERTYPE |
> +--------------------------+---------+
> | Pattern | Actions |
> +===+=====+==========+=====+=========+
> | 0 | ETH | ``spec`` | any | QUEUE, |
> | | +----------+-----+ DROP |
> | | | ``mask`` | any | |
> +---+-----+----------+-----+---------+
>
> ``FLEXIBLE`` to ``RAW`` → ``QUEUE``
> -----------------------------------
>
> `FLEXIBLE`_ can be translated to one `RAW`_ pattern with `QUEUE`_ as the
> terminating action and a defined priority level.
>
> +------------------------------------+
> | FLEXIBLE |
> +--------------------------+---------+
> | Pattern | Actions |
> +===+=====+==========+=====+=========+
> | 0 | RAW | ``spec`` | any | QUEUE |
> | | +----------+-----+ |
> | | | ``mask`` | any | |
> +---+-----+----------+-----+---------+
>
> ``SYN`` to ``TCP`` → ``QUEUE``
> ------------------------------
>
> `SYN`_ is a `TCP`_ rule with only the ``syn`` bit enabled and masked, and
> `QUEUE`_ as the terminating action.
>
> Priority level can be set to simulate the high priority bit.
>
> +---------------------------------------------+
> | SYN |
> +-----------------------------------+---------+
> | Pattern | Actions |
> +===+======+==========+=============+=========+
> | 0 | ETH | ``spec`` | N/A | QUEUE |
> | | +----------+-------------+ |
> | | | ``mask`` | empty | |
> +---+------+----------+-------------+ |
> | 1 | IPV4 | ``spec`` | N/A | |
> | | +----------+-------------+ |
> | | | ``mask`` | empty | |
> +---+------+----------+-------------+ |
> | 2 | TCP | ``spec`` | ``syn`` = 1 | |
> | | +----------+-------------+ |
> | | | ``mask`` | ``syn`` = 1 | |
> +---+------+----------+-------------+---------+
>
> ``NTUPLE`` to ``IPV4``, ``TCP``, ``UDP`` → ``QUEUE``
> ----------------------------------------------------
>
> `NTUPLE`_ is similar to specifying an empty L2, `IPV4`_ as L3 with `TCP`_ or
> `UDP`_ as L4 and `QUEUE`_ as the terminating action.
>
> A priority level can be specified as well.
>
> +---------------------------------------+
> | NTUPLE |
> +-----------------------------+---------+
> | Pattern | Actions |
> +===+======+==========+=======+=========+
> | 0 | ETH | ``spec`` | N/A | QUEUE |
> | | +----------+-------+ |
> | | | ``mask`` | empty | |
> +---+------+----------+-------+ |
> | 1 | IPV4 | ``spec`` | any | |
> | | +----------+-------+ |
> | | | ``mask`` | any | |
> +---+------+----------+-------+ |
> | 2 | TCP, | ``spec`` | any | |
> | | UDP +----------+-------+ |
> | | | ``mask`` | any | |
> +---+------+----------+-------+---------+
>
> ``TUNNEL`` to ``ETH``, ``IPV4``, ``IPV6``, ``VXLAN`` (or other) → ``QUEUE``
> ---------------------------------------------------------------------------
>
> `TUNNEL`_ matches common IPv4 and IPv6 L3/L4-based tunnel types.
>
> In the following table, `ANY`_ is used to cover the optional L4.
>
> +------------------------------------------------+
> | TUNNEL |
> +--------------------------------------+---------+
> | Pattern | Actions |
> +===+=========+==========+=============+=========+
> | 0 | ETH | ``spec`` | any | QUEUE |
> | | +----------+-------------+ |
> | | | ``mask`` | any | |
> +---+---------+----------+-------------+ |
> | 1 | IPV4, | ``spec`` | any | |
> | | IPV6 +----------+-------------+ |
> | | | ``mask`` | any | |
> +---+---------+----------+-------------+ |
> | 2 | ANY | ``spec`` | ``min`` = 0 | |
> | | | +-------------+ |
> | | | | ``max`` = 0 | |
> | | +----------+-------------+ |
> | | | ``mask`` | N/A | |
> +---+---------+----------+-------------+ |
> | 3 | VXLAN, | ``spec`` | any | |
> | | GENEVE, +----------+-------------+ |
> | | TEREDO, | ``mask`` | any | |
> | | NVGRE, | | | |
> | | GRE, | | | |
> | | ... | | | |
> +---+---------+----------+-------------+---------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``FDIR`` to most item types → ``QUEUE``, ``DROP``, ``PASSTHRU``
> ---------------------------------------------------------------
>
> `FDIR`_ is more complex than any other type, there are several methods to
> emulate its functionality. It is summarized for the most part in the table
> below.
>
> A few features are intentionally not supported:
>
> - The ability to configure the matching input set and masks for the entire
> device, PMDs should take care of it automatically according to flow rules.
>
> - Returning four or eight bytes of matched data when using flex bytes
> filtering. Although a specific action could implement it, it conflicts
> with the much more useful 32 bits tagging on devices that support it.
>
> - Side effects on RSS processing of the entire device. Flow rules that
> conflict with the current device configuration should not be
> allowed. Similarly, device configuration should not be allowed when it
> affects existing flow rules.
>
> - Device modes of operation. "none" is unsupported since filtering cannot be
> disabled as long as a flow rule is present.
>
> - "MAC VLAN" or "tunnel" perfect matching modes should be automatically set
> according to the created flow rules.
>
> +----------------------------------------------+
> | FDIR |
> +---------------------------------+------------+
> | Pattern | Actions |
> +===+============+==========+=====+============+
> | 0 | ETH, | ``spec`` | any | QUEUE, |
> | | RAW +----------+-----+ DROP, |
> | | | ``mask`` | any | PASSTHRU |
> +---+------------+----------+-----+------------+
> | 1 | IPV4, | ``spec`` | any | ID |
> | | IPV6 +----------+-----+ (optional) |
> | | | ``mask`` | any | |
> +---+------------+----------+-----+ |
> | 2 | TCP, | ``spec`` | any | |
> | | UDP, +----------+-----+ |
> | | SCTP | ``mask`` | any | |
> +---+------------+----------+-----+ |
> | 3 | VF, | ``spec`` | any | |
> | | PF, +----------+-----+ |
> | | SIGNATURE | ``mask`` | any | |
> | | (optional) | | | |
> +---+------------+----------+-----+------------+
>
> ``HASH``
> ~~~~~~~~
>
> Hashing configuration is set per rule through the `SIGNATURE`_ item.
>
> Since it is usually a global device setting, all flow rules created with
> this item may have to share the same specification.
>
> ``L2_TUNNEL`` to ``VOID`` → ``VXLAN`` (or others)
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> All packets are matched. This type alters incoming packets to encapsulate
> them in a chosen tunnel type, optionally redirect them to a VF as well.
>
> The destination pool for tag based forwarding can be emulated with other
> flow rules using `DUP`_ as the action.
>
> +----------------------------------------+
> | L2_TUNNEL |
> +---------------------------+------------+
> | Pattern | Actions |
> +===+======+==========+=====+============+
> | 0 | VOID | ``spec`` | N/A | VXLAN, |
> | | | | | GENEVE, |
> | | | | | ... |
> | | +----------+-----+------------+
> | | | ``mask`` | N/A | VF |
> | | | | | (optional) |
> +---+------+----------+-----+------------+
>
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v2 10/10] maintainers: add section for pmdinfo
@ 2016-07-08 10:14 4% ` Thomas Monjalon
1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-08 10:14 UTC (permalink / raw)
To: Neil Horman; +Cc: dev
The author of this feature is Neil Horman.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
---
MAINTAINERS | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index a59191e..f996c2e 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -68,6 +68,10 @@ F: lib/librte_compat/
F: doc/guides/rel_notes/deprecation.rst
F: scripts/validate-abi.sh
+Driver information
+F: buildtools/pmdinfogen/
+F: tools/pmdinfo.py
+
Environment Abstraction Layer
-----------------------------
--
2.7.0
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-05 18:16 2% [dpdk-dev] [RFC] Generic flow director/filtering/classification API Adrien Mazarguil
2016-07-07 7:14 0% ` Lu, Wenzhuo
@ 2016-07-07 23:15 0% ` Chandran, Sugesh
2016-07-08 13:03 0% ` Adrien Mazarguil
2016-07-08 11:11 0% ` Liang, Cunming
2 siblings, 1 reply; 200+ results
From: Chandran, Sugesh @ 2016-07-07 23:15 UTC (permalink / raw)
To: Adrien Mazarguil, dev
Cc: Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Lu, Wenzhuo, Jan Medala,
John Daley, Chen, Jing D, Ananyev, Konstantin, Matej Vido,
Alejandro Lucero, Sony Chacko, Jerin Jacob, De Lara Guarch,
Pablo, Olga Shern
Hi Adrien,
Thank you for proposing this. It would be really useful for application such as OVS-DPDK.
Please find my comments and questions inline below prefixed with [Sugesh]. Most of them are from the perspective of enabling these APIs in application such as OVS-DPDK.
Regards
_Sugesh
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Adrien Mazarguil
> Sent: Tuesday, July 5, 2016 7:17 PM
> To: dev@dpdk.org
> Cc: Thomas Monjalon <thomas.monjalon@6wind.com>; Zhang, Helin
> <helin.zhang@intel.com>; Wu, Jingjing <jingjing.wu@intel.com>; Rasesh
> Mody <rasesh.mody@qlogic.com>; Ajit Khaparde
> <ajit.khaparde@broadcom.com>; Rahul Lakkireddy
> <rahul.lakkireddy@chelsio.com>; Lu, Wenzhuo <wenzhuo.lu@intel.com>;
> Jan Medala <jan@semihalf.com>; John Daley <johndale@cisco.com>; Chen,
> Jing D <jing.d.chen@intel.com>; Ananyev, Konstantin
> <konstantin.ananyev@intel.com>; Matej Vido <matejvido@gmail.com>;
> Alejandro Lucero <alejandro.lucero@netronome.com>; Sony Chacko
> <sony.chacko@qlogic.com>; Jerin Jacob
> <jerin.jacob@caviumnetworks.com>; De Lara Guarch, Pablo
> <pablo.de.lara.guarch@intel.com>; Olga Shern <olgas@mellanox.com>
> Subject: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
>
> Hi All,
>
> First, forgive me for this large message, I know our mailboxes already
> suffer quite a bit from the amount of traffic on this ML.
>
> This is not exactly yet another thread about how flow director should be
> extended, rather about a brand new API to handle filtering and
> classification for incoming packets in the most PMD-generic and
> application-friendly fashion we can come up with. Reasons described below.
>
> I think this topic is important enough to include both the users of this API
> as well as PMD maintainers. So far I have CC'ed librte_ether (especially
> rte_eth_ctrl.h contributors), testpmd and PMD maintainers (with and
> without
> a .filter_ctrl implementation), but if you know application maintainers
> other than testpmd who use FDIR or might be interested in this discussion,
> feel free to add them.
>
> The issues we found with the current approach are already summarized in
> the
> following document, but here is a quick summary for TL;DR folks:
>
> - PMDs do not expose a common set of filter types and even when they do,
> their behavior more or less differs.
>
> - Applications need to determine and adapt to device-specific limitations
> and quirks on their own, without help from PMDs.
>
> - Writing an application that creates flow rules targeting all devices
> supported by DPDK is thus difficult, if not impossible.
>
> - The current API has too many unspecified areas (particularly regarding
> side effects of flow rules) that make PMD implementation tricky.
>
> This RFC API handles everything currently supported by .filter_ctrl, the
> idea being to reimplement all of these to make them fully usable by
> applications in a more generic and well defined fashion. It has a very small
> set of mandatory features and an easy method to let applications probe for
> supported capabilities.
>
> The only downside is more work for the software control side of PMDs
> because
> they have to adapt to the API instead of the reverse. I think helpers can be
> added to EAL to assist with this.
>
> HTML version:
>
> https://rawgit.com/6WIND/rte_flow/master/rte_flow.html
>
> PDF version:
>
> https://rawgit.com/6WIND/rte_flow/master/rte_flow.pdf
>
> Related draft header file (for reference while reading the specification):
>
> https://raw.githubusercontent.com/6WIND/rte_flow/master/rte_flow.h
>
> Git tree for completeness (latest .rst version can be retrieved from here):
>
> https://github.com/6WIND/rte_flow
>
> What follows is the ReST source of the above, for inline comments and
> discussion. I intend to update that specification accordingly.
>
> ========================
> Generic filter interface
> ========================
>
> .. footer::
>
> v0.6
>
> .. contents::
> .. sectnum::
> .. raw:: pdf
>
> PageBreak
>
> Overview
> ========
>
> DPDK provides several competing interfaces added over time to perform
> packet
> matching and related actions such as filtering and classification.
>
> They must be extended to implement the features supported by newer
> devices
> in order to expose them to applications, however the current design has
> several drawbacks:
>
> - Complicated filter combinations which have not been hard-coded cannot be
> expressed.
> - Prone to API/ABI breakage when new features must be added to an
> existing
> filter type, which frequently happens.
>
> From an application point of view:
>
> - Having disparate interfaces, all optional and lacking in features does not
> make this API easy to use.
> - Seemingly arbitrary built-in limitations of filter types based on the
> device they were initially designed for.
> - Undefined relationship between different filter types.
> - High complexity, considerable undocumented and/or undefined behavior.
>
> Considering the growing number of devices supported by DPDK, adding a
> new
> filter type each time a new feature must be implemented is not sustainable
> in the long term. Applications not written to target a specific device
> cannot really benefit from such an API.
>
> For these reasons, this document defines an extensible unified API that
> encompasses and supersedes these legacy filter types.
>
> .. raw:: pdf
>
> PageBreak
>
> Current API
> ===========
>
> Rationale
> ---------
>
> The reason several competing (and mostly overlapping) filtering APIs are
> present in DPDK is due to its nature as a thin layer between hardware and
> software.
>
> Each subsequent interface has been added to better match the capabilities
> and limitations of the latest supported device, which usually happened to
> need an incompatible configuration approach. Because of this, many ended
> up
> device-centric and not usable by applications that were not written for that
> particular device.
>
> This document is not the first attempt to address this proliferation issue,
> in fact a lot of work has already been done both to create a more generic
> interface while somewhat keeping compatibility with legacy ones through a
> common call interface (``rte_eth_dev_filter_ctrl()`` with the
> ``.filter_ctrl`` PMD callback in ``rte_ethdev.h``).
>
> Today, these previously incompatible interfaces are known as filter types
> (``RTE_ETH_FILTER_*`` from ``enum rte_filter_type`` in ``rte_eth_ctrl.h``).
>
> However while trivial to extend with new types, it only shifted the
> underlying problem as applications still need to be written for one kind of
> filter type, which, as described in the following sections, is not
> necessarily implemented by all PMDs that support filtering.
>
> .. raw:: pdf
>
> PageBreak
>
> Filter types
> ------------
>
> This section summarizes the capabilities of each filter type.
>
> Although the following list is exhaustive, the description of individual
> types may contain inaccuracies due to the lack of documentation or usage
> examples.
>
> Note: names are prefixed with ``RTE_ETH_FILTER_``.
>
> ``MACVLAN``
> ~~~~~~~~~~~
>
> Matching:
>
> - L2 source/destination addresses.
> - Optional 802.1Q VLAN ID.
> - Masking individual fields on a rule basis is not supported.
>
> Action:
>
> - Packets are redirected either to a given VF device using its ID or to the
> PF.
>
> ``ETHERTYPE``
> ~~~~~~~~~~~~~
>
> Matching:
>
> - L2 source/destination addresses (optional).
> - Ethertype (no VLAN ID?).
> - Masking individual fields on a rule basis is not supported.
>
> Action:
>
> - Receive packets on a given queue.
> - Drop packets.
>
> ``FLEXIBLE``
> ~~~~~~~~~~~~
>
> Matching:
>
> - At most 128 consecutive bytes anywhere in packets.
> - Masking is supported with byte granularity.
> - Priorities are supported (relative to this filter type, undefined
> otherwise).
>
> Action:
>
> - Receive packets on a given queue.
>
> ``SYN``
> ~~~~~~~
>
> Matching:
>
> - TCP SYN packets only.
> - One high priority bit can be set to give the highest possible priority to
> this type when other filters with different types are configured.
>
> Action:
>
> - Receive packets on a given queue.
>
> ``NTUPLE``
> ~~~~~~~~~~
>
> Matching:
>
> - Source/destination IPv4 addresses (optional in 2-tuple mode).
> - Source/destination TCP/UDP port (mandatory in 2 and 5-tuple modes).
> - L4 protocol (2 and 5-tuple modes).
> - Masking individual fields is supported.
> - TCP flags.
> - Up to 7 levels of priority relative to this filter type, undefined
> otherwise.
> - No IPv6.
>
> Action:
>
> - Receive packets on a given queue.
>
> ``TUNNEL``
> ~~~~~~~~~~
>
> Matching:
>
> - Outer L2 source/destination addresses.
> - Inner L2 source/destination addresses.
> - Inner VLAN ID.
> - IPv4/IPv6 source (destination?) address.
> - Tunnel type to match (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE,
> 802.1BR
> E-Tag).
> - Tenant ID for tunneling protocols that have one.
> - Any combination of the above can be specified.
> - Masking individual fields on a rule basis is not supported.
>
> Action:
>
> - Receive packets on a given queue.
>
> .. raw:: pdf
>
> PageBreak
>
> ``FDIR``
> ~~~~~~~~
>
> Queries:
>
> - Device capabilities and limitations.
> - Device statistics about configured filters (resource usage, collisions).
> - Device configuration (matching input set and masks)
>
> Matching:
>
> - Device mode of operation: none (to disable filtering), signature
> (hash-based dispatching from masked fields) or perfect (either MAC VLAN
> or
> tunnel).
> - L2 Ethertype.
> - Outer L2 destination address (MAC VLAN mode).
> - Inner L2 destination address, tunnel type (NVGRE, VXLAN) and tunnel ID
> (tunnel mode).
> - IPv4 source/destination addresses, ToS, TTL and protocol fields.
> - IPv6 source/destination addresses, TC, protocol and hop limits fields.
> - UDP source/destination IPv4/IPv6 and ports.
> - TCP source/destination IPv4/IPv6 and ports.
> - SCTP source/destination IPv4/IPv6, ports and verification tag field.
> - Note, only one protocol type at once (either only L2 Ethertype, basic
> IPv6, IPv4+UDP, IPv4+TCP and so on).
> - VLAN TCI (extended API).
> - At most 16 bytes to match in payload (extended API). A global device
> look-up table specifies for each possible protocol layer (unknown, raw,
> L2, L3, L4) the offset to use for each byte (they do not need to be
> contiguous) and the related bitmask.
> - Whether packet is addressed to PF or VF, in that case its ID can be
> matched as well (extended API).
> - Masking most of the above fields is supported, but simultaneously affects
> all filters configured on a device.
> - Input set can be modified in a similar fashion for a given device to
> ignore individual fields of filters (i.e. do not match the destination
> address in a IPv4 filter, refer to **RTE_ETH_INPUT_SET_**
> macros). Configuring this also affects RSS processing on **i40e**.
> - Filters can also provide 32 bits of arbitrary data to return as part of
> matched packets.
>
> Action:
>
> - **RTE_ETH_FDIR_ACCEPT**: receive (accept) packet on a given queue.
> - **RTE_ETH_FDIR_REJECT**: drop packet immediately.
> - **RTE_ETH_FDIR_PASSTHRU**: similar to accept for the last filter in list,
> otherwise process it with subsequent filters.
> - For accepted packets and if requested by filter, either 32 bits of
> arbitrary data and four bytes of matched payload (only in case of flex
> bytes matching), or eight bytes of matched payload (flex also) are added
> to meta data.
>
> .. raw:: pdf
>
> PageBreak
>
> ``HASH``
> ~~~~~~~~
>
> Not an actual filter type. Provides and retrieves the global device
> configuration (per port or entire NIC) for hash functions and their
> properties.
>
> Hash function selection: "default" (keep current), XOR or Toeplitz.
>
> This function can be configured per flow type (**RTE_ETH_FLOW_**
> definitions), supported types are:
>
> - Unknown.
> - Raw.
> - Fragmented or non-fragmented IPv4.
> - Non-fragmented IPv4 with L4 (TCP, UDP, SCTP or other).
> - Fragmented or non-fragmented IPv6.
> - Non-fragmented IPv6 with L4 (TCP, UDP, SCTP or other).
> - L2 payload.
> - IPv6 with extensions.
> - IPv6 with L4 (TCP, UDP) and extensions.
>
> ``L2_TUNNEL``
> ~~~~~~~~~~~~~
>
> Matching:
>
> - All packets received on a given port.
>
> Action:
>
> - Add tunnel encapsulation (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE,
> 802.1BR E-Tag) using the provided Ethertype and tunnel ID (only E-Tag
> is implemented at the moment).
> - VF ID to use for tag insertion (currently unused).
> - Destination pool for tag based forwarding (pools are IDs that can be
> affected to ports, duplication occurs if the same ID is shared by several
> ports of the same NIC).
>
> .. raw:: pdf
>
> PageBreak
>
> Driver support
> --------------
>
> ======== ======= ========= ======== === ====== ====== ==== ====
> =========
> Driver MACVLAN ETHERTYPE FLEXIBLE SYN NTUPLE TUNNEL FDIR HASH
> L2_TUNNEL
> ======== ======= ========= ======== === ====== ====== ==== ====
> =========
> bnx2x
> cxgbe
> e1000 yes yes yes yes
> ena
> enic yes
> fm10k
> i40e yes yes yes yes yes
> ixgbe yes yes yes yes yes
> mlx4
> mlx5 yes
> szedata2
> ======== ======= ========= ======== === ====== ====== ==== ====
> =========
>
> Flow director
> -------------
>
> Flow director (FDIR) is the name of the most capable filter type, which
> covers most features offered by others. As such, it is the most widespread
> in PMDs that support filtering (i.e. all of them besides **e1000**).
>
> It is also the only type that allows an arbitrary 32 bits value provided by
> applications to be attached to a filter and returned with matching packets
> instead of relying on the destination queue to recognize flows.
>
> Unfortunately, even FDIR requires applications to be aware of low-level
> capabilities and limitations (most of which come directly from **ixgbe** and
> **i40e**):
>
> - Bitmasks are set globally per device (port?), not per filter.
[Sugesh] This means application cannot define filters that matches on arbitrary different offsets?
If that’s the case, I assume the application has to program bitmask in advance. Otherwise how
the API framework deduce this bitmask information from the rules?? Its not very clear to me
that how application pass down the bitmask information for multiple filters on same port?
> - Configuration state is not expected to be saved by the driver, and
> stopping/restarting a port requires the application to perform it again
> (API documentation is also unclear about this).
> - Monolithic approach with ABI issues as soon as a new kind of flow or
> combination needs to be supported.
> - Cryptic global statistics/counters.
> - Unclear about how priorities are managed; filters seem to be arranged as a
> linked list in hardware (possibly related to configuration order).
>
> Packet alteration
> -----------------
>
> One interesting feature is that the L2 tunnel filter type implements the
> ability to alter incoming packets through a filter (in this case to
> encapsulate them), thus the **mlx5** flow encap/decap features are not a
> foreign concept.
>
> .. raw:: pdf
>
> PageBreak
>
> Proposed API
> ============
>
> Terminology
> -----------
>
> - **Filtering API**: overall framework affecting the fate of selected
> packets, covers everything described in this document.
> - **Matching pattern**: properties to look for in received packets, a
> combination of any number of items.
> - **Pattern item**: part of a pattern that either matches packet data
> (protocol header, payload or derived information), or specifies properties
> of the pattern itself.
> - **Actions**: what needs to be done when a packet matches a pattern.
> - **Flow rule**: this is the result of combining a *matching pattern* with
> *actions*.
> - **Filter rule**: a less generic term than *flow rule*, can otherwise be
> used interchangeably.
> - **Hit**: a flow rule is said to be *hit* when processing a matching
> packet.
>
> Requirements
> ------------
>
> As described in the previous section, there is a growing need for a common
> method to configure filtering and related actions in a hardware independent
> fashion.
>
> The filtering API should not disallow any filter combination by design and
> must remain as simple as possible to use. It can simply be defined as a
> method to perform one or several actions on selected packets.
>
> PMDs are aware of the capabilities of the device they manage and should be
> responsible for preventing unsupported or conflicting combinations.
>
> This approach is fundamentally different as it places most of the burden on
> the software side of the PMD instead of having device capabilities directly
> mapped to API functions, then expecting applications to work around
> ensuing
> compatibility issues.
>
> Requirements for a new API:
>
> - Flexible and extensible without causing API/ABI problems for existing
> applications.
> - Should be unambiguous and easy to use.
> - Support existing filtering features and actions listed in `Filter types`_.
> - Support packet alteration.
> - In case of overlapping filters, their priority should be well documented.
> - Support filter queries (for example to retrieve counters).
>
> .. raw:: pdf
>
> PageBreak
>
> High level design
> -----------------
>
> The chosen approach to make filtering as generic as possible is by
> expressing matching patterns through lists of items instead of the flat
> structures used in DPDK today, enabling combinations that are not
> predefined
> and thus being more versatile.
>
> Flow rules can have several distinct actions (such as counting,
> encapsulating, decapsulating before redirecting packets to a particular
> queue, etc.), instead of relying on several rules to achieve this and having
> applications deal with hardware implementation details regarding their
> order.
>
> Support for different priority levels on a rule basis is provided, for
> example in order to force a more specific rule come before a more generic
> one for packets matched by both, however hardware support for more than
> a
> single priority level cannot be guaranteed. When supported, the number of
> available priority levels is usually low, which is why they can also be
> implemented in software by PMDs (e.g. to simulate missing priority levels by
> reordering rules).
>
> In order to remain as hardware agnostic as possible, by default all rules
> are considered to have the same priority, which means that the order
> between
> overlapping rules (when a packet is matched by several filters) is
> undefined, packet duplication may even occur as a result.
>
> PMDs may refuse to create overlapping rules at a given priority level when
> they can be detected (e.g. if a pattern matches an existing filter).
>
> Thus predictable results for a given priority level can only be achieved
> with non-overlapping rules, using perfect matching on all protocol layers.
>
> Support for multiple actions per rule may be implemented internally on top
> of non-default hardware priorities, as a result both features may not be
> simultaneously available to applications.
>
> Considering that allowed pattern/actions combinations cannot be known in
> advance and would result in an unpractically large number of capabilities to
> expose, a method is provided to validate a given rule from the current
> device configuration state without actually adding it (akin to a "dry run"
> mode).
>
> This enables applications to check if the rule types they need is supported
> at initialization time, before starting their data path. This method can be
> used anytime, its only requirement being that the resources needed by a
> rule
> must exist (e.g. a target RX queue must be configured first).
>
> Each defined rule is associated with an opaque handle managed by the PMD,
> applications are responsible for keeping it. These can be used for queries
> and rules management, such as retrieving counters or other data and
> destroying them.
>
> Handles must be destroyed before releasing associated resources such as
> queues.
>
> Integration
> -----------
>
> To avoid ABI breakage, this new interface will be implemented through the
> existing filtering control framework (``rte_eth_dev_filter_ctrl()``) using
> **RTE_ETH_FILTER_GENERIC** as a new filter type.
>
> However a public front-end API described in `Rules management`_ will
> be added as the preferred method to use it.
>
> Once discussions with the community have converged to a definite API,
> legacy
> filter types should be deprecated and a deadline defined to remove their
> support entirely.
>
> PMDs will have to be gradually converted to **RTE_ETH_FILTER_GENERIC**
> or
> drop filtering support entirely. Less maintained PMDs for older hardware
> may
> lose support at this point.
>
> The notion of filter type will then be deprecated and subsequently dropped
> to avoid confusion between both frameworks.
>
> Implementation details
> ======================
>
> Flow rule
> ---------
>
> A flow rule is the combination of a matching pattern with a list of actions,
> and is the basis of this API.
>
> Priorities
> ~~~~~~~~~~
>
> A priority can be assigned to a matching pattern.
>
> The default priority level is 0 and is also the highest. Support for more
> than a single priority level in hardware is not guaranteed.
>
> If a packet is matched by several filters at a given priority level, the
> outcome is undefined. It can take any path and can even be duplicated.
>
> Matching pattern
> ~~~~~~~~~~~~~~~~
>
> A matching pattern comprises any number of items of various types.
>
> Items are arranged in a list to form a matching pattern for packets. They
> fall in two categories:
>
> - Protocol matching (ANY, RAW, ETH, IPV4, IPV6, ICMP, UDP, TCP, VXLAN and
> so
> on), usually associated with a specification structure. These must be
> stacked in the same order as the protocol layers to match, starting from
> L2.
>
> - Affecting how the pattern is processed (END, VOID, INVERT, PF, VF,
> SIGNATURE and so on), often without a specification structure. Since they
> are meta data that does not match packet contents, these can be specified
> anywhere within item lists without affecting the protocol matching items.
>
> Most item specifications can be optionally paired with a mask to narrow the
> specific fields or bits to be matched.
>
> - Items are defined with ``struct rte_flow_item``.
> - Patterns are defined with ``struct rte_flow_pattern``.
>
> Example of an item specification matching an Ethernet header:
>
> +-----------------------------------------+
> | Ethernet |
> +==========+=========+====================+
> | ``spec`` | ``src`` | ``00:01:02:03:04`` |
> | +---------+--------------------+
> | | ``dst`` | ``00:2a:66:00:01`` |
> +----------+---------+--------------------+
> | ``mask`` | ``src`` | ``00:ff:ff:ff:00`` |
> | +---------+--------------------+
> | | ``dst`` | ``00:00:00:00:ff`` |
> +----------+---------+--------------------+
>
> Non-masked bits stand for any value, Ethernet headers with the following
> properties are thus matched:
>
> - ``src``: ``??:01:02:03:??``
> - ``dst``: ``??:??:??:??:01``
>
> Except for meta types that do not need one, ``spec`` must be a valid pointer
> to a structure of the related item type. A ``mask`` of the same type can be
> provided to tell which bits in ``spec`` are to be matched.
>
> A mask is normally only needed for ``spec`` fields matching packet data,
> ignored otherwise. See individual item types for more information.
>
> A ``NULL`` mask pointer is allowed and is similar to matching with a full
> mask (all ones) ``spec`` fields supported by hardware, the remaining fields
> are ignored (all zeroes), there is thus no error checking for unsupported
> fields.
>
> Matching pattern items for packet data must be naturally stacked (ordered
> from lowest to highest protocol layer), as in the following examples:
>
> +--------------+
> | TCPv4 as L4 |
> +===+==========+
> | 0 | Ethernet |
> +---+----------+
> | 1 | IPv4 |
> +---+----------+
> | 2 | TCP |
> +---+----------+
>
> +----------------+
> | TCPv6 in VXLAN |
> +===+============+
> | 0 | Ethernet |
> +---+------------+
> | 1 | IPv4 |
> +---+------------+
> | 2 | UDP |
> +---+------------+
> | 3 | VXLAN |
> +---+------------+
> | 4 | Ethernet |
> +---+------------+
> | 5 | IPv6 |
> +---+------------+
> | 6 | TCP |
> +---+------------+
>
> +-----------------------------+
> | TCPv4 as L4 with meta items |
> +===+=========================+
> | 0 | VOID |
> +---+-------------------------+
> | 1 | Ethernet |
> +---+-------------------------+
> | 2 | VOID |
> +---+-------------------------+
> | 3 | IPv4 |
> +---+-------------------------+
> | 4 | TCP |
> +---+-------------------------+
> | 5 | VOID |
> +---+-------------------------+
> | 6 | VOID |
> +---+-------------------------+
>
> The above example shows how meta items do not affect packet data
> matching
> items, as long as those remain stacked properly. The resulting matching
> pattern is identical to "TCPv4 as L4".
>
> +----------------+
> | UDPv6 anywhere |
> +===+============+
> | 0 | IPv6 |
> +---+------------+
> | 1 | UDP |
> +---+------------+
>
> If supported by the PMD, omitting one or several protocol layers at the
> bottom of the stack as in the above example (missing an Ethernet
> specification) enables hardware to look anywhere in packets.
>
> It is unspecified whether the payload of supported encapsulations
> (e.g. VXLAN inner packet) is matched by such a pattern, which may apply to
> inner, outer or both packets.
>
> +---------------------+
> | Invalid, missing L3 |
> +===+=================+
> | 0 | Ethernet |
> +---+-----------------+
> | 1 | UDP |
> +---+-----------------+
>
> The above pattern is invalid due to a missing L3 specification between L2
> and L4. It is only allowed at the bottom and at the top of the stack.
>
> Meta item types
> ~~~~~~~~~~~~~~~
>
> These do not match packet data but affect how the pattern is processed,
> most
> of them do not need a specification structure. This particularity allows
> them to be specified anywhere without affecting other item types.
>
> ``END``
> ^^^^^^^
>
> End marker for item lists. Prevents further processing of items, thereby
> ending the pattern.
>
> - Its numeric value is **0** for convenience.
> - PMD support is mandatory.
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | END |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> ``VOID``
> ^^^^^^^^
>
> Used as a placeholder for convenience. It is ignored and simply discarded by
> PMDs.
>
> - PMD support is mandatory.
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | VOID |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> One usage example for this type is generating rules that share a common
> prefix quickly without reallocating memory, only by updating item types:
>
> +------------------------+
> | TCP, UDP or ICMP as L4 |
> +===+====================+
> | 0 | Ethernet |
> +---+--------------------+
> | 1 | IPv4 |
> +---+------+------+------+
> | 2 | UDP | VOID | VOID |
> +---+------+------+------+
> | 3 | VOID | TCP | VOID |
> +---+------+------+------+
> | 4 | VOID | VOID | ICMP |
> +---+------+------+------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``INVERT``
> ^^^^^^^^^^
>
> Inverted matching, i.e. process packets that do not match the pattern.
>
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | INVERT |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> Usage example in order to match non-TCPv4 packets only:
>
> +--------------------+
> | Anything but TCPv4 |
> +===+================+
> | 0 | INVERT |
> +---+----------------+
> | 1 | Ethernet |
> +---+----------------+
> | 2 | IPv4 |
> +---+----------------+
> | 3 | TCP |
> +---+----------------+
>
> ``PF``
> ^^^^^^
>
> Matches packets addressed to the physical function of the device.
>
> - Both ``spec`` and ``mask`` are ignored.
>
> +--------------------+
> | PF |
> +==========+=========+
> | ``spec`` | ignored |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> ``VF``
> ^^^^^^
>
> Matches packets addressed to the given virtual function ID of the device.
>
> - Only ``spec`` needs to be defined, ``mask`` is ignored.
>
> +----------------------------------------+
> | VF |
> +==========+=========+===================+
> | ``spec`` | ``vf`` | destination VF ID |
> +----------+---------+-------------------+
> | ``mask`` | ignored |
> +----------+-----------------------------+
>
> ``SIGNATURE``
> ^^^^^^^^^^^^^
>
> Requests hash-based signature dispatching for this rule.
>
> Considering this is a global setting on devices that support it, all
> subsequent filter rules may have to be created with it as well.
>
> - Only ``spec`` needs to be defined, ``mask`` is ignored.
>
> +--------------------+
> | SIGNATURE |
> +==========+=========+
> | ``spec`` | TBD |
> +----------+---------+
> | ``mask`` | ignored |
> +----------+---------+
>
> .. raw:: pdf
>
> PageBreak
>
> Data matching item types
> ~~~~~~~~~~~~~~~~~~~~~~~~
>
> Most of these are basically protocol header definitions with associated
> bitmasks. They must be specified (stacked) from lowest to highest protocol
> layer.
>
> The following list is not exhaustive as new protocols will be added in the
> future.
>
> ``ANY``
> ^^^^^^^
>
> Matches any protocol in place of the current layer, a single ANY may also
> stand for several protocol layers.
>
> This is usually specified as the first pattern item when looking for a
> protocol anywhere in a packet.
>
> - A maximum value of **0** requests matching any number of protocol
> layers
> above or equal to the minimum value, a maximum value lower than the
> minimum one is otherwise invalid.
> - Only ``spec`` needs to be defined, ``mask`` is ignored.
>
> +-----------------------------------------------------------------------+
> | ANY |
> +==========+=========+====================================
> ==============+
> | ``spec`` | ``min`` | minimum number of layers covered |
> | +---------+--------------------------------------------------+
> | | ``max`` | maximum number of layers covered, 0 for infinity |
> +----------+---------+--------------------------------------------------+
> | ``mask`` | ignored |
> +----------+------------------------------------------------------------+
>
> Example for VXLAN TCP payload matching regardless of outer L3 (IPv4 or
> IPv6)
> and L4 (UDP) both matched by the first ANY specification, and inner L3 (IPv4
> or IPv6) matched by the second ANY specification:
>
> +----------------------------------+
> | TCP in VXLAN with wildcards |
> +===+==============================+
> | 0 | Ethernet |
> +---+-----+----------+---------+---+
> | 1 | ANY | ``spec`` | ``min`` | 2 |
> | | | +---------+---+
> | | | | ``max`` | 2 |
> +---+-----+----------+---------+---+
> | 2 | VXLAN |
> +---+------------------------------+
> | 3 | Ethernet |
> +---+-----+----------+---------+---+
> | 4 | ANY | ``spec`` | ``min`` | 1 |
> | | | +---------+---+
> | | | | ``max`` | 1 |
> +---+-----+----------+---------+---+
> | 5 | TCP |
> +---+------------------------------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``RAW``
> ^^^^^^^
>
> Matches a string of a given length at a given offset (in bytes), or anywhere
> in the payload of the current protocol layer (including L2 header if used as
> the first item in the stack).
>
> This does not increment the protocol layer count as it is not a protocol
> definition. Subsequent RAW items modulate the first absolute one with
> relative offsets.
>
> - Using **-1** as the ``offset`` of the first RAW item makes its absolute
> offset not fixed, i.e. the pattern is searched everywhere.
> - ``mask`` only affects the pattern.
>
> +--------------------------------------------------------------+
> | RAW |
> +==========+=============+================================
> =====+
> | ``spec`` | ``offset`` | absolute or relative pattern offset |
> | +-------------+-------------------------------------+
> | | ``length`` | pattern length |
> | +-------------+-------------------------------------+
> | | ``pattern`` | byte string of the above length |
> +----------+-------------+-------------------------------------+
> | ``mask`` | ``offset`` | ignored |
> | +-------------+-------------------------------------+
> | | ``length`` | ignored |
> | +-------------+-------------------------------------+
> | | ``pattern`` | bitmask with the same byte length |
> +----------+-------------+-------------------------------------+
>
> Example pattern looking for several strings at various offsets of a UDP
> payload, using combined RAW items:
>
> +------------------------------------------+
> | UDP payload matching |
> +===+======================================+
> | 0 | Ethernet |
> +---+--------------------------------------+
> | 1 | IPv4 |
> +---+--------------------------------------+
> | 2 | UDP |
> +---+-----+----------+-------------+-------+
> | 3 | RAW | ``spec`` | ``offset`` | -1 |
> | | | +-------------+-------+
> | | | | ``length`` | 3 |
> | | | +-------------+-------+
> | | | | ``pattern`` | "foo" |
> +---+-----+----------+-------------+-------+
> | 4 | RAW | ``spec`` | ``offset`` | 20 |
> | | | +-------------+-------+
> | | | | ``length`` | 3 |
> | | | +-------------+-------+
> | | | | ``pattern`` | "bar" |
> +---+-----+----------+-------------+-------+
> | 5 | RAW | ``spec`` | ``offset`` | -30 |
> | | | +-------------+-------+
> | | | | ``length`` | 3 |
> | | | +-------------+-------+
> | | | | ``pattern`` | "baz" |
> +---+-----+----------+-------------+-------+
>
> This translates to:
>
> - Locate "foo" in UDP payload, remember its offset.
> - Check "bar" at "foo"'s offset plus 20 bytes.
> - Check "baz" at "foo"'s offset minus 30 bytes.
>
> .. raw:: pdf
>
> PageBreak
>
> ``ETH``
> ^^^^^^^
>
> Matches an Ethernet header.
>
> - ``dst``: destination MAC.
> - ``src``: source MAC.
> - ``type``: EtherType.
> - ``tags``: number of 802.1Q/ad tags defined.
> - ``tag[]``: 802.1Q/ad tag definitions, innermost first. For each one:
>
> - ``tpid``: Tag protocol identifier.
> - ``tci``: Tag control information.
>
> ``IPV4``
> ^^^^^^^^
>
> Matches an IPv4 header.
>
> - ``src``: source IP address.
> - ``dst``: destination IP address.
> - ``tos``: ToS/DSCP field.
> - ``ttl``: TTL field.
> - ``proto``: protocol number for the next layer.
>
> ``IPV6``
> ^^^^^^^^
>
> Matches an IPv6 header.
>
> - ``src``: source IP address.
> - ``dst``: destination IP address.
> - ``tc``: traffic class field.
> - ``nh``: Next header field (protocol).
> - ``hop_limit``: hop limit field (TTL).
>
> ``ICMP``
> ^^^^^^^^
>
> Matches an ICMP header.
>
> - TBD.
>
> ``UDP``
> ^^^^^^^
>
> Matches a UDP header.
>
> - ``sport``: source port.
> - ``dport``: destination port.
> - ``length``: UDP length.
> - ``checksum``: UDP checksum.
>
> .. raw:: pdf
>
> PageBreak
>
> ``TCP``
> ^^^^^^^
>
> Matches a TCP header.
>
> - ``sport``: source port.
> - ``dport``: destination port.
> - All other TCP fields and bits.
>
> ``VXLAN``
> ^^^^^^^^^
>
> Matches a VXLAN header.
>
> - TBD.
>
> .. raw:: pdf
>
> PageBreak
>
> Actions
> ~~~~~~~
>
> Each possible action is represented by a type. Some have associated
> configuration structures. Several actions combined in a list can be affected
> to a flow rule. That list is not ordered.
>
> At least one action must be defined in a filter rule in order to do
> something with matched packets.
>
> - Actions are defined with ``struct rte_flow_action``.
> - A list of actions is defined with ``struct rte_flow_actions``.
>
> They fall in three categories:
>
> - Terminating actions (such as QUEUE, DROP, RSS, PF, VF) that prevent
> processing matched packets by subsequent flow rules, unless overridden
> with PASSTHRU.
>
> - Non terminating actions (PASSTHRU, DUP) that leave matched packets up
> for
> additional processing by subsequent flow rules.
>
> - Other non terminating meta actions that do not affect the fate of packets
> (END, VOID, ID, COUNT).
>
> When several actions are combined in a flow rule, they should all have
> different types (e.g. dropping a packet twice is not possible). However
> considering the VOID type is an exception to this rule, the defined behavior
> is for PMDs to only take into account the last action of a given type found
> in the list. PMDs still perform error checking on the entire list.
>
> *Note that PASSTHRU is the only action able to override a terminating rule.*
>
> .. raw:: pdf
>
> PageBreak
>
> Example of an action that redirects packets to queue index 10:
>
> +----------------+
> | QUEUE |
> +===========+====+
> | ``queue`` | 10 |
> +-----------+----+
>
> Action lists examples, their order is not significant, applications must
> consider all actions to be performed simultaneously:
>
> +----------------+
> | Count and drop |
> +=======+========+
> | COUNT | |
> +-------+--------+
> | DROP | |
> +-------+--------+
>
> +--------------------------+
> | Tag, count and redirect |
> +=======+===========+======+
> | ID | ``id`` | 0x2a |
> +-------+-----------+------+
> | COUNT | |
> +-------+-----------+------+
> | QUEUE | ``queue`` | 10 |
> +-------+-----------+------+
>
> +-----------------------+
> | Redirect to queue 5 |
> +=======+===============+
> | DROP | |
> +-------+-----------+---+
> | QUEUE | ``queue`` | 5 |
> +-------+-----------+---+
>
> In the above example, considering both actions are performed
> simultaneously,
> its end result is that only QUEUE has any effect.
>
> +-----------------------+
> | Redirect to queue 3 |
> +=======+===========+===+
> | QUEUE | ``queue`` | 5 |
> +-------+-----------+---+
> | VOID | |
> +-------+-----------+---+
> | QUEUE | ``queue`` | 3 |
> +-------+-----------+---+
>
> As previously described, only the last action of a given type found in the
> list is taken into account. The above example also shows that VOID is
> ignored.
>
> .. raw:: pdf
>
> PageBreak
>
> Action types
> ~~~~~~~~~~~~
>
> Common action types are described in this section. Like pattern item types,
> this list is not exhaustive as new actions will be added in the future.
>
> ``END`` (action)
> ^^^^^^^^^^^^^^^^
>
> End marker for action lists. Prevents further processing of actions, thereby
> ending the list.
>
> - Its numeric value is **0** for convenience.
> - PMD support is mandatory.
> - No configurable property.
>
> +---------------+
> | END |
> +===============+
> | no properties |
> +---------------+
>
> ``VOID`` (action)
> ^^^^^^^^^^^^^^^^^
>
> Used as a placeholder for convenience. It is ignored and simply discarded by
> PMDs.
>
> - PMD support is mandatory.
> - No configurable property.
>
> +---------------+
> | VOID |
> +===============+
> | no properties |
> +---------------+
>
> ``PASSTHRU``
> ^^^^^^^^^^^^
>
> Leaves packets up for additional processing by subsequent flow rules. This
> is the default when a rule does not contain a terminating action, but can be
> specified to force a rule to become non-terminating.
>
> - No configurable property.
>
> +---------------+
> | PASSTHRU |
> +===============+
> | no properties |
> +---------------+
>
> Example to copy a packet to a queue and continue processing by subsequent
> flow rules:
[Sugesh] If a packet get copied to a queue, it’s a termination action.
How can its possible to do subsequent action after the packet already
moved to the queue. ?How it differs from DUP action?
Am I missing anything here?
>
> +--------------------------+
> | Copy to queue 8 |
> +==========+===============+
> | PASSTHRU | |
> +----------+-----------+---+
> | QUEUE | ``queue`` | 8 |
> +----------+-----------+---+
>
> ``ID``
> ^^^^^^
>
> Attaches a 32 bit value to packets.
>
> +----------------------------------------------+
> | ID |
> +========+=====================================+
> | ``id`` | 32 bit value to return with packets |
> +--------+-------------------------------------+
>
[Sugesh] I assume the application has to program the flow
with a unique ID and matching packets are stamped with this ID
when reporting to the software. The uniqueness of ID is NOT
guaranteed by the API framework. Correct me if I am wrong here.
[Sugesh] Is it a limitation to use only 32 bit ID? Is it possible to have a
64 bit ID? So that application can use the control plane flow pointer
Itself as an ID. Does it make sense?
> .. raw:: pdf
>
> PageBreak
>
> ``QUEUE``
> ^^^^^^^^^
>
> Assigns packets to a given queue index.
>
> - Terminating by default.
>
> +--------------------------------+
> | QUEUE |
> +===========+====================+
> | ``queue`` | queue index to use |
> +-----------+--------------------+
>
> ``DROP``
> ^^^^^^^^
>
> Drop packets.
>
> - No configurable property.
> - Terminating by default.
> - PASSTHRU overrides this action if both are specified.
>
> +---------------+
> | DROP |
> +===============+
> | no properties |
> +---------------+
>
> ``COUNT``
> ^^^^^^^^^
>
[Sugesh] Should we really have to set count action explicitly for every rule?
IMHO it would be great to be an implicit action. Most of the application would be
interested in the stats of almost all the filters/flows .
> Enables hits counter for this rule.
>
> This counter can be retrieved and reset through ``rte_flow_query()``, see
> ``struct rte_flow_query_count``.
>
> - Counters can be retrieved with ``rte_flow_query()``.
> - No configurable property.
>
> +---------------+
> | COUNT |
> +===============+
> | no properties |
> +---------------+
>
> Query structure to retrieve and reset the flow rule hits counter:
>
> +------------------------------------------------+
> | COUNT query |
> +===========+=====+==============================+
> | ``reset`` | in | reset counter after query |
> +-----------+-----+------------------------------+
> | ``hits`` | out | number of hits for this flow |
> +-----------+-----+------------------------------+
>
> ``DUP``
> ^^^^^^^
>
> Duplicates packets to a given queue index.
>
> This is normally combined with QUEUE, however when used alone, it is
> actually similar to QUEUE + PASSTHRU.
>
> - Non-terminating by default.
>
> +------------------------------------------------+
> | DUP |
> +===========+====================================+
> | ``queue`` | queue index to duplicate packet to |
> +-----------+------------------------------------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``RSS``
> ^^^^^^^
>
> Similar to QUEUE, except RSS is additionally performed on packets to spread
> them among several queues according to the provided parameters.
>
> - Terminating by default.
>
> +---------------------------------------------+
> | RSS |
> +==============+==============================+
> | ``rss_conf`` | RSS parameters |
> +--------------+------------------------------+
> | ``queues`` | number of entries in queue[] |
> +--------------+------------------------------+
> | ``queue[]`` | queue indices to use |
> +--------------+------------------------------+
>
> ``PF`` (action)
> ^^^^^^^^^^^^^^^
>
> Redirects packets to the physical function (PF) of the current device.
>
> - No configurable property.
> - Terminating by default.
>
> +---------------+
> | PF |
> +===============+
> | no properties |
> +---------------+
>
> ``VF`` (action)
> ^^^^^^^^^^^^^^^
>
> Redirects packets to the virtual function (VF) of the current device with
> the specified ID.
>
> - Terminating by default.
>
> +---------------------------------------+
> | VF |
> +========+==============================+
> | ``id`` | VF ID to redirect packets to |
> +--------+------------------------------+
>
> Planned types
> ~~~~~~~~~~~~~
>
> Other action types are planned but not defined yet. These actions will add
> the ability to alter matching packets in several ways, such as performing
> encapsulation/decapsulation of tunnel headers on specific flows.
>
> .. raw:: pdf
>
> PageBreak
>
> Rules management
> ----------------
>
> A simple API with only four functions is provided to fully manage flows.
>
> Each created flow rule is associated with an opaque, PMD-specific handle
> pointer. The application is responsible for keeping it until the rule is
> destroyed.
>
> Flows rules are defined with ``struct rte_flow``.
>
> Validation
> ~~~~~~~~~~
>
> Given that expressing a definite set of device capabilities with this API is
> not practical, a dedicated function is provided to check if a flow rule is
> supported and can be created.
>
> ::
>
> int
> rte_flow_validate(uint8_t port_id,
> const struct rte_flow_pattern *pattern,
> const struct rte_flow_actions *actions);
>
> While this function has no effect on the target device, the flow rule is
> validated against its current configuration state and the returned value
> should be considered valid by the caller for that state only.
>
> The returned value is guaranteed to remain valid only as long as no
> successful calls to rte_flow_create() or rte_flow_destroy() are made in the
> meantime and no device parameter affecting flow rules in any way are
> modified, due to possible collisions or resource limitations (although in
> such cases ``EINVAL`` should not be returned).
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``pattern``: pattern specification to check.
> - ``actions``: actions associated with the flow definition.
>
> Return value:
>
> - **0** if flow rule is valid and can be created. A negative errno value
> otherwise (``rte_errno`` is also set), the following errors are defined.
> - ``-EINVAL``: unknown or invalid rule specification.
> - ``-ENOTSUP``: valid but unsupported rule specification (e.g. partial masks
> are unsupported).
> - ``-EEXIST``: collision with an existing rule.
> - ``-ENOMEM``: not enough resources.
>
> .. raw:: pdf
>
> PageBreak
>
> Creation
> ~~~~~~~~
>
> Creating a flow rule is similar to validating one, except the rule is
> actually created.
>
> ::
>
> struct rte_flow *
> rte_flow_create(uint8_t port_id,
> const struct rte_flow_pattern *pattern,
> const struct rte_flow_actions *actions);
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``pattern``: pattern specification to add.
> - ``actions``: actions associated with the flow definition.
>
> Return value:
>
> A valid flow pointer in case of success, NULL otherwise and ``rte_errno`` is
> set to the positive version of one of the error codes defined for
> ``rte_flow_validate()``.
[Sugesh] : Kind of implementation specific query. What if application
try to add duplicate rules? Does the API create new flow entry for every
API call?
[Sugesh] Another concern is the cost and time of installing these rules
in the hardware. Can we make these APIs time bound(or at least an option to
set the time limit to execute these APIs), so that
Application doesn’t have to wait so long when installing and deleting flows with
slow hardware/NIC. What do you think? Most of the datapath flow installations are
dynamic and triggered only when there is
an ingress traffic. Delay in flow insertion/deletion have unpredictable consequences.
[Sugesh] Another query is on the synchronization part. What if same rules are
handled from different threads? Is application responsible for handling the concurrent
hardware programming?
>
> Destruction
> ~~~~~~~~~~~
>
> Flow rules destruction is not automatic, and a queue should not be released
> if any are still attached to it. Applications must take care of performing
> this step before releasing resources.
>
> ::
>
> int
> rte_flow_destroy(uint8_t port_id,
> struct rte_flow *flow);
>
>
[Sugesh] I would suggest having a clean-up API is really useful as the releasing of
Queue(is it applicable for releasing of port too?) is not guaranteeing the automatic flow
destruction. This way application can initialize the port,
clean-up all the existing rules and create new rules on a clean slate.
> Failure to destroy a flow rule may occur when other flow rules depend on it,
> and destroying it would result in an inconsistent state.
>
> This function is only guaranteed to succeed if flow rules are destroyed in
> reverse order of their creation.
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``flow``: flow rule to destroy.
>
> Return value:
>
> - **0** on success, a negative errno value otherwise and ``rte_errno`` is
> set.
>
> .. raw:: pdf
>
> PageBreak
>
> Query
> ~~~~~
>
> Query an existing flow rule.
>
> This function allows retrieving flow-specific data such as counters. Data
> is gathered by special actions which must be present in the flow rule
> definition.
>
> ::
>
> int
> rte_flow_query(uint8_t port_id,
> struct rte_flow *flow,
> enum rte_flow_action_type action,
> void *data);
>
> Arguments:
>
> - ``port_id``: port identifier of Ethernet device.
> - ``flow``: flow rule to query.
> - ``action``: action type to query.
> - ``data``: pointer to storage for the associated query data type.
>
> Return value:
>
> - **0** on success, a negative errno value otherwise and ``rte_errno`` is
> set.
>
> .. raw:: pdf
>
> PageBreak
>
> Behavior
> --------
>
> - API operations are synchronous and blocking (``EAGAIN`` cannot be
> returned).
>
> - There is no provision for reentrancy/multi-thread safety, although nothing
> should prevent different devices from being configured at the same
> time. PMDs may protect their control path functions accordingly.
>
> - Stopping the data path (TX/RX) should not be necessary when managing
> flow
> rules. If this cannot be achieved naturally or with workarounds (such as
> temporarily replacing the burst function pointers), an appropriate error
> code must be returned (``EBUSY``).
>
> - PMDs, not applications, are responsible for maintaining flow rules
> configuration when stopping and restarting a port or performing other
> actions which may affect them. They can only be destroyed explicitly.
>
> .. raw:: pdf
>
> PageBreak
>
[Sugesh] Query all the rules for a specific port/queue?? Useful when adding and
deleting ports and queues dynamically according to the need. I am not sure
what are the other different usecases for these APIs. But I feel it makes much easier to
manage flows from the application. What do you think?
> Compatibility
> -------------
>
> No known hardware implementation supports all the features described in
> this
> document.
>
> Unsupported features or combinations are not expected to be fully
> emulated
> in software by PMDs for performance reasons. Partially supported features
> may be completed in software as long as hardware performs most of the
> work
> (such as queue redirection and packet recognition).
>
> However PMDs are expected to do their best to satisfy application requests
> by working around hardware limitations as long as doing so does not affect
> the behavior of existing flow rules.
>
> The following sections provide a few examples of such cases, they are based
> on limitations built into the previous APIs.
>
> Global bitmasks
> ~~~~~~~~~~~~~~~
>
> Each flow rule comes with its own, per-layer bitmasks, while hardware may
> support only a single, device-wide bitmask for a given layer type, so that
> two IPv4 rules cannot use different bitmasks.
>
> The expected behavior in this case is that PMDs automatically configure
> global bitmasks according to the needs of the first created flow rule.
>
> Subsequent rules are allowed only if their bitmasks match those, the
> ``EEXIST`` error code should be returned otherwise.
>
> Unsupported layer types
> ~~~~~~~~~~~~~~~~~~~~~~~
>
> Many protocols can be simulated by crafting patterns with the `RAW`_ type.
>
> PMDs can rely on this capability to simulate support for protocols with
> fixed headers not directly recognized by hardware.
>
> ``ANY`` pattern item
> ~~~~~~~~~~~~~~~~~~~~
>
> This pattern item stands for anything, which can be difficult to translate
> to something hardware would understand, particularly if followed by more
> specific types.
>
> Consider the following pattern:
>
> +---+--------------------------------+
> | 0 | ETHER |
> +---+--------------------------------+
> | 1 | ANY (``min`` = 1, ``max`` = 1) |
> +---+--------------------------------+
> | 2 | TCP |
> +---+--------------------------------+
>
> Knowing that TCP does not make sense with something other than IPv4 and
> IPv6
> as L3, such a pattern may be translated to two flow rules instead:
>
> +---+--------------------+
> | 0 | ETHER |
> +---+--------------------+
> | 1 | IPV4 (zeroed mask) |
> +---+--------------------+
> | 2 | TCP |
> +---+--------------------+
>
> +---+--------------------+
> | 0 | ETHER |
> +---+--------------------+
> | 1 | IPV6 (zeroed mask) |
> +---+--------------------+
> | 2 | TCP |
> +---+--------------------+
>
> Note that as soon as a ANY rule covers several layers, this approach may
> yield a large number of hidden flow rules. It is thus suggested to only
> support the most common scenarios (anything as L2 and/or L3).
>
> .. raw:: pdf
>
> PageBreak
>
> Unsupported actions
> ~~~~~~~~~~~~~~~~~~~
>
> - When combined with a `QUEUE`_ action, packet counting (`COUNT`_) and
> tagging (`ID`_) may be implemented in software as long as the target queue
> is used by a single rule.
>
> - A rule specifying both `DUP`_ + `QUEUE`_ may be translated to two hidden
> rules combining `QUEUE`_ and `PASSTHRU`_.
>
> - When a single target queue is provided, `RSS`_ can also be implemented
> through `QUEUE`_.
>
> Flow rules priority
> ~~~~~~~~~~~~~~~~~~~
>
> While it would naturally make sense, flow rules cannot be assumed to be
> processed by hardware in the same order as their creation for several
> reasons:
>
> - They may be managed internally as a tree or a hash table instead of a
> list.
> - Removing a flow rule before adding another one can either put the new
> rule
> at the end of the list or reuse a freed entry.
> - Duplication may occur when packets are matched by several rules.
>
> For overlapping rules (particularly in order to use the `PASSTHRU`_ action)
> predictable behavior is only guaranteed by using different priority levels.
>
> Priority levels are not necessarily implemented in hardware, or may be
> severely limited (e.g. a single priority bit).
>
> For these reasons, priority levels may be implemented purely in software by
> PMDs.
>
> - For devices expecting flow rules to be added in the correct order, PMDs
> may destroy and re-create existing rules after adding a new one with
> a higher priority.
>
> - A configurable number of dummy or empty rules can be created at
> initialization time to save high priority slots for later.
>
> - In order to save priority levels, PMDs may evaluate whether rules are
> likely to collide and adjust their priority accordingly.
>
> .. raw:: pdf
>
> PageBreak
>
> API migration
> =============
>
> Exhaustive list of deprecated filter types and how to convert them to
> generic flow rules.
>
> ``MACVLAN`` to ``ETH`` → ``VF``, ``PF``
> ---------------------------------------
>
> `MACVLAN`_ can be translated to a basic `ETH`_ flow rule with a `VF
> (action)`_ or `PF (action)`_ terminating action.
>
> +------------------------------------+
> | MACVLAN |
> +--------------------------+---------+
> | Pattern | Actions |
> +===+=====+==========+=====+=========+
> | 0 | ETH | ``spec`` | any | VF, |
> | | +----------+-----+ PF |
> | | | ``mask`` | any | |
> +---+-----+----------+-----+---------+
>
> ``ETHERTYPE`` to ``ETH`` → ``QUEUE``, ``DROP``
> ----------------------------------------------
>
> `ETHERTYPE`_ is basically an `ETH`_ flow rule with `QUEUE`_ or `DROP`_ as
> a terminating action.
>
> +------------------------------------+
> | ETHERTYPE |
> +--------------------------+---------+
> | Pattern | Actions |
> +===+=====+==========+=====+=========+
> | 0 | ETH | ``spec`` | any | QUEUE, |
> | | +----------+-----+ DROP |
> | | | ``mask`` | any | |
> +---+-----+----------+-----+---------+
>
> ``FLEXIBLE`` to ``RAW`` → ``QUEUE``
> -----------------------------------
>
> `FLEXIBLE`_ can be translated to one `RAW`_ pattern with `QUEUE`_ as the
> terminating action and a defined priority level.
>
> +------------------------------------+
> | FLEXIBLE |
> +--------------------------+---------+
> | Pattern | Actions |
> +===+=====+==========+=====+=========+
> | 0 | RAW | ``spec`` | any | QUEUE |
> | | +----------+-----+ |
> | | | ``mask`` | any | |
> +---+-----+----------+-----+---------+
>
> ``SYN`` to ``TCP`` → ``QUEUE``
> ------------------------------
>
> `SYN`_ is a `TCP`_ rule with only the ``syn`` bit enabled and masked, and
> `QUEUE`_ as the terminating action.
>
> Priority level can be set to simulate the high priority bit.
>
> +---------------------------------------------+
> | SYN |
> +-----------------------------------+---------+
> | Pattern | Actions |
> +===+======+==========+=============+=========+
> | 0 | ETH | ``spec`` | N/A | QUEUE |
> | | +----------+-------------+ |
> | | | ``mask`` | empty | |
> +---+------+----------+-------------+ |
> | 1 | IPV4 | ``spec`` | N/A | |
> | | +----------+-------------+ |
> | | | ``mask`` | empty | |
> +---+------+----------+-------------+ |
> | 2 | TCP | ``spec`` | ``syn`` = 1 | |
> | | +----------+-------------+ |
> | | | ``mask`` | ``syn`` = 1 | |
> +---+------+----------+-------------+---------+
>
> ``NTUPLE`` to ``IPV4``, ``TCP``, ``UDP`` → ``QUEUE``
> ----------------------------------------------------
>
> `NTUPLE`_ is similar to specifying an empty L2, `IPV4`_ as L3 with `TCP`_ or
> `UDP`_ as L4 and `QUEUE`_ as the terminating action.
>
> A priority level can be specified as well.
>
> +---------------------------------------+
> | NTUPLE |
> +-----------------------------+---------+
> | Pattern | Actions |
> +===+======+==========+=======+=========+
> | 0 | ETH | ``spec`` | N/A | QUEUE |
> | | +----------+-------+ |
> | | | ``mask`` | empty | |
> +---+------+----------+-------+ |
> | 1 | IPV4 | ``spec`` | any | |
> | | +----------+-------+ |
> | | | ``mask`` | any | |
> +---+------+----------+-------+ |
> | 2 | TCP, | ``spec`` | any | |
> | | UDP +----------+-------+ |
> | | | ``mask`` | any | |
> +---+------+----------+-------+---------+
>
> ``TUNNEL`` to ``ETH``, ``IPV4``, ``IPV6``, ``VXLAN`` (or other) → ``QUEUE``
> ---------------------------------------------------------------------------
>
> `TUNNEL`_ matches common IPv4 and IPv6 L3/L4-based tunnel types.
>
> In the following table, `ANY`_ is used to cover the optional L4.
>
> +------------------------------------------------+
> | TUNNEL |
> +--------------------------------------+---------+
> | Pattern | Actions |
> +===+=========+==========+=============+=========+
> | 0 | ETH | ``spec`` | any | QUEUE |
> | | +----------+-------------+ |
> | | | ``mask`` | any | |
> +---+---------+----------+-------------+ |
> | 1 | IPV4, | ``spec`` | any | |
> | | IPV6 +----------+-------------+ |
> | | | ``mask`` | any | |
> +---+---------+----------+-------------+ |
> | 2 | ANY | ``spec`` | ``min`` = 0 | |
> | | | +-------------+ |
> | | | | ``max`` = 0 | |
> | | +----------+-------------+ |
> | | | ``mask`` | N/A | |
> +---+---------+----------+-------------+ |
> | 3 | VXLAN, | ``spec`` | any | |
> | | GENEVE, +----------+-------------+ |
> | | TEREDO, | ``mask`` | any | |
> | | NVGRE, | | | |
> | | GRE, | | | |
> | | ... | | | |
> +---+---------+----------+-------------+---------+
>
> .. raw:: pdf
>
> PageBreak
>
> ``FDIR`` to most item types → ``QUEUE``, ``DROP``, ``PASSTHRU``
> ---------------------------------------------------------------
>
> `FDIR`_ is more complex than any other type, there are several methods to
> emulate its functionality. It is summarized for the most part in the table
> below.
>
> A few features are intentionally not supported:
>
> - The ability to configure the matching input set and masks for the entire
> device, PMDs should take care of it automatically according to flow rules.
>
> - Returning four or eight bytes of matched data when using flex bytes
> filtering. Although a specific action could implement it, it conflicts
> with the much more useful 32 bits tagging on devices that support it.
>
> - Side effects on RSS processing of the entire device. Flow rules that
> conflict with the current device configuration should not be
> allowed. Similarly, device configuration should not be allowed when it
> affects existing flow rules.
>
> - Device modes of operation. "none" is unsupported since filtering cannot be
> disabled as long as a flow rule is present.
>
> - "MAC VLAN" or "tunnel" perfect matching modes should be automatically
> set
> according to the created flow rules.
>
> +----------------------------------------------+
> | FDIR |
> +---------------------------------+------------+
> | Pattern | Actions |
> +===+============+==========+=====+============+
> | 0 | ETH, | ``spec`` | any | QUEUE, |
> | | RAW +----------+-----+ DROP, |
> | | | ``mask`` | any | PASSTHRU |
> +---+------------+----------+-----+------------+
> | 1 | IPV4, | ``spec`` | any | ID |
> | | IPV6 +----------+-----+ (optional) |
> | | | ``mask`` | any | |
> +---+------------+----------+-----+ |
> | 2 | TCP, | ``spec`` | any | |
> | | UDP, +----------+-----+ |
> | | SCTP | ``mask`` | any | |
> +---+------------+----------+-----+ |
> | 3 | VF, | ``spec`` | any | |
> | | PF, +----------+-----+ |
> | | SIGNATURE | ``mask`` | any | |
> | | (optional) | | | |
> +---+------------+----------+-----+------------+
>
> ``HASH``
> ~~~~~~~~
>
> Hashing configuration is set per rule through the `SIGNATURE`_ item.
>
> Since it is usually a global device setting, all flow rules created with
> this item may have to share the same specification.
>
> ``L2_TUNNEL`` to ``VOID`` → ``VXLAN`` (or others)
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> All packets are matched. This type alters incoming packets to encapsulate
> them in a chosen tunnel type, optionally redirect them to a VF as well.
>
> The destination pool for tag based forwarding can be emulated with other
> flow rules using `DUP`_ as the action.
>
> +----------------------------------------+
> | L2_TUNNEL |
> +---------------------------+------------+
> | Pattern | Actions |
> +===+======+==========+=====+============+
> | 0 | VOID | ``spec`` | N/A | VXLAN, |
> | | | | | GENEVE, |
> | | | | | ... |
> | | +----------+-----+------------+
> | | | ``mask`` | N/A | VF |
> | | | | | (optional) |
> +---+------+----------+-----+------------+
>
> --
> Adrien Mazarguil
> 6WIND
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH 11/11] maintainers: add section for pmdinfo
2016-07-07 15:36 4% ` [dpdk-dev] [PATCH 11/11] maintainers: add section for pmdinfo Thomas Monjalon
@ 2016-07-07 16:14 0% ` Neil Horman
0 siblings, 0 replies; 200+ results
From: Neil Horman @ 2016-07-07 16:14 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev
On Thu, Jul 07, 2016 at 05:36:30PM +0200, Thomas Monjalon wrote:
> The author of this feature is Neil Horman.
>
> Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> ---
> MAINTAINERS | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/MAINTAINERS b/MAINTAINERS
> index a59191e..f996c2e 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -68,6 +68,10 @@ F: lib/librte_compat/
> F: doc/guides/rel_notes/deprecation.rst
> F: scripts/validate-abi.sh
>
> +Driver information
> +F: buildtools/pmdinfogen/
> +F: tools/pmdinfo.py
> +
>
> Environment Abstraction Layer
> -----------------------------
> --
> 2.7.0
>
>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH 11/11] maintainers: add section for pmdinfo
@ 2016-07-07 15:36 4% ` Thomas Monjalon
2016-07-07 16:14 0% ` Neil Horman
1 sibling, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-07-07 15:36 UTC (permalink / raw)
To: Neil Horman; +Cc: dev
The author of this feature is Neil Horman.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
MAINTAINERS | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index a59191e..f996c2e 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -68,6 +68,10 @@ F: lib/librte_compat/
F: doc/guides/rel_notes/deprecation.rst
F: scripts/validate-abi.sh
+Driver information
+F: buildtools/pmdinfogen/
+F: tools/pmdinfo.py
+
Environment Abstraction Layer
-----------------------------
--
2.7.0
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-07 7:14 0% ` Lu, Wenzhuo
@ 2016-07-07 10:26 2% ` Adrien Mazarguil
2016-07-19 8:11 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Adrien Mazarguil @ 2016-07-07 10:26 UTC (permalink / raw)
To: Lu, Wenzhuo
Cc: dev, Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
Hi Lu Wenzhuo,
Thanks for your feedback, I'm replying below as well.
On Thu, Jul 07, 2016 at 07:14:18AM +0000, Lu, Wenzhuo wrote:
> Hi Adrien,
> I have some questions, please see inline, thanks.
>
> > -----Original Message-----
> > From: Adrien Mazarguil [mailto:adrien.mazarguil@6wind.com]
> > Sent: Wednesday, July 6, 2016 2:17 AM
> > To: dev@dpdk.org
> > Cc: Thomas Monjalon; Zhang, Helin; Wu, Jingjing; Rasesh Mody; Ajit Khaparde;
> > Rahul Lakkireddy; Lu, Wenzhuo; Jan Medala; John Daley; Chen, Jing D; Ananyev,
> > Konstantin; Matej Vido; Alejandro Lucero; Sony Chacko; Jerin Jacob; De Lara
> > Guarch, Pablo; Olga Shern
> > Subject: [RFC] Generic flow director/filtering/classification API
> >
> >
> > Requirements for a new API:
> >
> > - Flexible and extensible without causing API/ABI problems for existing
> > applications.
> > - Should be unambiguous and easy to use.
> > - Support existing filtering features and actions listed in `Filter types`_.
> > - Support packet alteration.
> > - In case of overlapping filters, their priority should be well documented.
> Does that mean we don't guarantee the consistent of priority? The priority can be different on different NICs. So the behavior of the actions can be different. Right?
No, the intent is precisely to define what happens in order to get a
consistent result across different devices, and document cases with
undefined behavior. There must be no room left for interpretation.
For example, the API must describe what happens when two overlapping filters
(e.g. one matching an Ethernet header, another one matching an IP header)
match a given packet at a given priority level.
It is documented in section 4.1.1 (priorities) as "undefined behavior".
Applications remain free to do it and deal with consequences, at least they
know they cannot expect a consistent outcome, unless they use different
priority levels for both rules, see also 4.4.5 (flow rules priority).
> Seems the users still need to aware the some details of the HW? Do we need to add the negotiation for the priority?
Priorities as defined in this document may not be directly mappable to HW
capabilities (e.g. HW does not support enough priorities, or that some
corner case make them not work as described), in which case the PMD may
choose to simulate priorities (again 4.4.5), as long as the end result
follows the specification.
So users must not be aware of some HW details, the PMD does and must perform
the needed workarounds to suit their expectations. Users may only be
impacted by errors while attempting to create rules that are either
unsupported or would cause them (or existing rules) to diverge from the
spec.
> > Flow rules can have several distinct actions (such as counting,
> > encapsulating, decapsulating before redirecting packets to a particular
> > queue, etc.), instead of relying on several rules to achieve this and having
> > applications deal with hardware implementation details regarding their
> > order.
> I think normally HW doesn't support several actions in one rule. If a rule has several actions, seems HW has to split it to several rules. The order can still be a problem.
Note that, except for a very small subset of pattern items and actions,
supporting multiple actions for a given rule is not mandatory, and can be
emulated as you said by having to split them into several rules each with
its own priority if possible (see 3.3 "high level design").
Also, a rule "action" as defined in this API can be just about anything, for
example combining a queue redirection with 32-bit tagging. FDIR supports
many cases of what can be described as several actions, see 5.7 "FDIR to
most item types → QUEUE, DROP, PASSTHRU".
If you were thinking about having two queue targets for a given rule, then
I agree with you - that is why a rule cannot have more than a single action
of a given type (see 4.1.5 actions), to avoid such abuse from applications.
Applications must use several pass-through rules with different priority
levels if they want to perform a given action several times on a given
packet. Again, PMDs support is not mandatory as pass-through is optional.
> > ``ETH``
> > ^^^^^^^
> >
> > Matches an Ethernet header.
> >
> > - ``dst``: destination MAC.
> > - ``src``: source MAC.
> > - ``type``: EtherType.
> > - ``tags``: number of 802.1Q/ad tags defined.
> > - ``tag[]``: 802.1Q/ad tag definitions, innermost first. For each one:
> >
> > - ``tpid``: Tag protocol identifier.
> > - ``tci``: Tag control information.
> "ETH" means all the parameters, dst, src, type... need to be matched? The same question for IPv4, IPv6 ...
Yes, it's basically the description of all Ethernet header fields including
VLAN tags (same for other protocols). Please see the linked draft header
file which should make all of this easier to understand:
https://raw.githubusercontent.com/6WIND/rte_flow/master/rte_flow.h
> > ``UDP``
> > ^^^^^^^
> >
> > Matches a UDP header.
> >
> > - ``sport``: source port.
> > - ``dport``: destination port.
> > - ``length``: UDP length.
> > - ``checksum``: UDP checksum.
> Why checksum? Do we need to filter the packets by checksum value?
Well, I've decided to include all protocol header fields for completeness
(so the ABI does not need to be broken later then they become necessary, or
require another pattern item), not that all of them make sense in a pattern.
In this specific case, all PMDs I know of must reject a pattern
specification with a nonzero mask for the checksum field, because none of
them support it.
> > ``VOID`` (action)
> > ^^^^^^^^^^^^^^^^^
> >
> > Used as a placeholder for convenience. It is ignored and simply discarded by
> > PMDs.
> Don't understand why we need VOID. If it’s about the format. Why not guarantee it in rte layer?
I'm not sure to understand your question about rte layer, but this type is
fully managed by the PMD and is not supposed to be translated to a hardware
action.
I think it may come handy in some cases (like the VOID pattern item), so it
is defined just in case. Should be relatively trivial to support.
Applications may find a use for it when they want to statically define
templates for flow rules, when they need room for some reason.
> > Behavior
> > --------
> >
> > - API operations are synchronous and blocking (``EAGAIN`` cannot be
> > returned).
> >
> > - There is no provision for reentrancy/multi-thread safety, although nothing
> > should prevent different devices from being configured at the same
> > time. PMDs may protect their control path functions accordingly.
> >
> > - Stopping the data path (TX/RX) should not be necessary when managing flow
> > rules. If this cannot be achieved naturally or with workarounds (such as
> > temporarily replacing the burst function pointers), an appropriate error
> > code must be returned (``EBUSY``).
> PMD cannot stop the data path without adding lock. So I think if some rules cannot be applied without stopping rx/tx, PMD has to return fail.
> Or let the APP to stop the data path.
Agreed, that is the intent. If the PMD cannot touch flow rules for some
reason even after trying really hard, then it just returns EBUSY.
Perhaps we should write down that applications may get a different outcome
after stopping the data path if they get EBUSY?
> > - PMDs, not applications, are responsible for maintaining flow rules
> > configuration when stopping and restarting a port or performing other
> > actions which may affect them. They can only be destroyed explicitly.
> Don’t understand " They can only be destroyed explicitly."
This part says that as long as an application has not called
rte_flow_destroy() on a flow rule, it never disappears, whatever happens to
the port (stopped, restarted). The application is not responsible for
re-creating rules after that.
Note that according to the specification, this may translate to not being
able to stop a port as long as a flow rule is present, depending on how nice
the PMD intends to be with applications. Implementation can be done in small
steps with minimal amount of code on the PMD side.
> If a new rule has conflict with an old one, what should we do? Return fail?
That should not happen. If say 100 rules have been created with various
priorities and the port is happily running with them, stopping the port may
require the PMD to destroy them, it then has to re-create all 100 of them
exactly as they were automatically when restarting the port.
If re-creating them is not possible for some reason, the port cannot be
restarted as long as rules that cannot be added back haven't been destroyed
by the application. Frankly, this should not happen.
To manage this case, I suggest preventing applications from doing things
that conflict with existing flow rules while the port is stopped (just like
when it is not stopped, as described in 5.7 "FDIR to most item types").
> > ``ANY`` pattern item
> > ~~~~~~~~~~~~~~~~~~~~
> >
> > This pattern item stands for anything, which can be difficult to translate
> > to something hardware would understand, particularly if followed by more
> > specific types.
> >
> > Consider the following pattern:
> >
> > +---+--------------------------------+
> > | 0 | ETHER |
> > +---+--------------------------------+
> > | 1 | ANY (``min`` = 1, ``max`` = 1) |
> > +---+--------------------------------+
> > | 2 | TCP |
> > +---+--------------------------------+
> >
> > Knowing that TCP does not make sense with something other than IPv4 and IPv6
> > as L3, such a pattern may be translated to two flow rules instead:
> >
> > +---+--------------------+
> > | 0 | ETHER |
> > +---+--------------------+
> > | 1 | IPV4 (zeroed mask) |
> > +---+--------------------+
> > | 2 | TCP |
> > +---+--------------------+
> >
> > +---+--------------------+
> > | 0 | ETHER |
> > +---+--------------------+
> > | 1 | IPV6 (zeroed mask) |
> > +---+--------------------+
> > | 2 | TCP |
> > +---+--------------------+
> >
> > Note that as soon as a ANY rule covers several layers, this approach may
> > yield a large number of hidden flow rules. It is thus suggested to only
> > support the most common scenarios (anything as L2 and/or L3).
> I think "any" may make things confusing. How about if the NIC doesn't support IPv6? Should we return fail for this rule?
In a sense you are right, ANY relies on HW capabilities so you cannot know
that it won't match unsupported protocols. The above example would be
somewhat useless for a conscious application which should really have
created two separate flow rules (and gotten an error on the IPv6 one).
So an ANY flow rule only able to match v4 packets won't return an error.
ANY can be useful to match outer packets when only a tunnel header and the
inner packet are meaningful to the application. HW that does not recognize
the outer packet is not able to recognize the inner one anyway.
This section only says that PMDs should do their best to make HW match what
they can when faced with ANY.
Also once again, ANY support is not mandatory.
> > Flow rules priority
> > ~~~~~~~~~~~~~~~~~~~
> >
> > While it would naturally make sense, flow rules cannot be assumed to be
> > processed by hardware in the same order as their creation for several
> > reasons:
> >
> > - They may be managed internally as a tree or a hash table instead of a
> > list.
> > - Removing a flow rule before adding another one can either put the new rule
> > at the end of the list or reuse a freed entry.
> > - Duplication may occur when packets are matched by several rules.
> >
> > For overlapping rules (particularly in order to use the `PASSTHRU`_ action)
> > predictable behavior is only guaranteed by using different priority levels.
> >
> > Priority levels are not necessarily implemented in hardware, or may be
> > severely limited (e.g. a single priority bit).
> >
> > For these reasons, priority levels may be implemented purely in software by
> > PMDs.
> >
> > - For devices expecting flow rules to be added in the correct order, PMDs
> > may destroy and re-create existing rules after adding a new one with
> > a higher priority.
> >
> > - A configurable number of dummy or empty rules can be created at
> > initialization time to save high priority slots for later.
> >
> > - In order to save priority levels, PMDs may evaluate whether rules are
> > likely to collide and adjust their priority accordingly.
> If there's 3 rules, r1, r2,r3. The rules say the priority is r1 > r2 > r3. If PMD can only support r1 > r3 > r2, or doesn't support r2. Should PMD apply r1 and r3 or totally not support them all?
Remember that the API lets applications create only one rule at a time. If
all 3 rules are not supported together but individually are, the answer
depends on what the application does:
1. r1 OK, r2 FAIL => application chooses to stop here, thus only r1 works as
expected (may roll back and remove r1 as a result).
2. r1 OK, r2 FAIL, r3 OK => application chooses to ignore the fact r2 failed
and added r3 anyway, so it should end up with r1 > r3.
Applications should do as described in 1, they need to check for errors if
they want consistency.
This document describes only the basic functions, but may be extended later
with methods to add several flow rules at once, so rules that depend on
others can be added together and a single failure is returned without the
need for a rollback at the application level.
> A generic question, is the parsing supposed to be done by rte or PMD?
Actually, a bit of both. EAL will certainly at least provide helpers to
assist PMDs. This specification defines only the public-facing API for now,
but our goal is really to have something that is not too difficult to
implement both for applications and PMDs.
These helpers can be defined later with the first implementation.
--
Adrien Mazarguil
6WIND
^ permalink raw reply [relevance 2%]
* Re: [dpdk-dev] [PATCH v4] Pci: Add the class_id support
2016-07-06 11:08 3% ` Ferruh Yigit
@ 2016-07-07 7:46 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-07 7:46 UTC (permalink / raw)
To: Ferruh Yigit; +Cc: Ziye Yang, dev
2016-07-06 12:08, Ferruh Yigit:
> On 6/14/2016 3:52 PM, Thomas Monjalon wrote:
> > 2016-05-24 20:50, Ziye Yang:
> >> This patch is used to add the class_id (class_code,
> >> subclass_code, programming_interface) support for
> >> pci_device probe. With this patch, it will be
> >> flexible for users to probe a class of devices
> >> by class_id.
> >>
> >>
> >> Signed-off-by: Ziye Yang <ziye.yang@intel.com>
> >
> > Applied, thanks
> >
> Hi Thomas, Ziye,
>
> Is modification in public "struct rte_pci_id" is a ABI break?
> If so, it requires eal LIBABIVER increase and release notes update.
Not really sure. I was thinking that it is used only by drivers
but not by applications.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [RFC] Generic flow director/filtering/classification API
2016-07-05 18:16 2% [dpdk-dev] [RFC] Generic flow director/filtering/classification API Adrien Mazarguil
@ 2016-07-07 7:14 0% ` Lu, Wenzhuo
2016-07-07 10:26 2% ` Adrien Mazarguil
2016-07-07 23:15 0% ` Chandran, Sugesh
2016-07-08 11:11 0% ` Liang, Cunming
2 siblings, 1 reply; 200+ results
From: Lu, Wenzhuo @ 2016-07-07 7:14 UTC (permalink / raw)
To: Adrien Mazarguil, dev
Cc: Thomas Monjalon, Zhang, Helin, Wu, Jingjing, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Jan Medala, John Daley, Chen,
Jing D, Ananyev, Konstantin, Matej Vido, Alejandro Lucero,
Sony Chacko, Jerin Jacob, De Lara Guarch, Pablo, Olga Shern
Hi Adrien,
I have some questions, please see inline, thanks.
> -----Original Message-----
> From: Adrien Mazarguil [mailto:adrien.mazarguil@6wind.com]
> Sent: Wednesday, July 6, 2016 2:17 AM
> To: dev@dpdk.org
> Cc: Thomas Monjalon; Zhang, Helin; Wu, Jingjing; Rasesh Mody; Ajit Khaparde;
> Rahul Lakkireddy; Lu, Wenzhuo; Jan Medala; John Daley; Chen, Jing D; Ananyev,
> Konstantin; Matej Vido; Alejandro Lucero; Sony Chacko; Jerin Jacob; De Lara
> Guarch, Pablo; Olga Shern
> Subject: [RFC] Generic flow director/filtering/classification API
>
>
> Requirements for a new API:
>
> - Flexible and extensible without causing API/ABI problems for existing
> applications.
> - Should be unambiguous and easy to use.
> - Support existing filtering features and actions listed in `Filter types`_.
> - Support packet alteration.
> - In case of overlapping filters, their priority should be well documented.
Does that mean we don't guarantee the consistent of priority? The priority can be different on different NICs. So the behavior of the actions can be different. Right?
Seems the users still need to aware the some details of the HW? Do we need to add the negotiation for the priority?
>
> Flow rules can have several distinct actions (such as counting,
> encapsulating, decapsulating before redirecting packets to a particular
> queue, etc.), instead of relying on several rules to achieve this and having
> applications deal with hardware implementation details regarding their
> order.
I think normally HW doesn't support several actions in one rule. If a rule has several actions, seems HW has to split it to several rules. The order can still be a problem.
>
> ``ETH``
> ^^^^^^^
>
> Matches an Ethernet header.
>
> - ``dst``: destination MAC.
> - ``src``: source MAC.
> - ``type``: EtherType.
> - ``tags``: number of 802.1Q/ad tags defined.
> - ``tag[]``: 802.1Q/ad tag definitions, innermost first. For each one:
>
> - ``tpid``: Tag protocol identifier.
> - ``tci``: Tag control information.
"ETH" means all the parameters, dst, src, type... need to be matched? The same question for IPv4, IPv6 ...
>
> ``UDP``
> ^^^^^^^
>
> Matches a UDP header.
>
> - ``sport``: source port.
> - ``dport``: destination port.
> - ``length``: UDP length.
> - ``checksum``: UDP checksum.
Why checksum? Do we need to filter the packets by checksum value?
>
> ``VOID`` (action)
> ^^^^^^^^^^^^^^^^^
>
> Used as a placeholder for convenience. It is ignored and simply discarded by
> PMDs.
Don't understand why we need VOID. If it’s about the format. Why not guarantee it in rte layer?
>
> Behavior
> --------
>
> - API operations are synchronous and blocking (``EAGAIN`` cannot be
> returned).
>
> - There is no provision for reentrancy/multi-thread safety, although nothing
> should prevent different devices from being configured at the same
> time. PMDs may protect their control path functions accordingly.
>
> - Stopping the data path (TX/RX) should not be necessary when managing flow
> rules. If this cannot be achieved naturally or with workarounds (such as
> temporarily replacing the burst function pointers), an appropriate error
> code must be returned (``EBUSY``).
PMD cannot stop the data path without adding lock. So I think if some rules cannot be applied without stopping rx/tx, PMD has to return fail.
Or let the APP to stop the data path.
>
> - PMDs, not applications, are responsible for maintaining flow rules
> configuration when stopping and restarting a port or performing other
> actions which may affect them. They can only be destroyed explicitly.
Don’t understand " They can only be destroyed explicitly." If a new rule has conflict with an old one, what should we do? Return fail?
>
> ``ANY`` pattern item
> ~~~~~~~~~~~~~~~~~~~~
>
> This pattern item stands for anything, which can be difficult to translate
> to something hardware would understand, particularly if followed by more
> specific types.
>
> Consider the following pattern:
>
> +---+--------------------------------+
> | 0 | ETHER |
> +---+--------------------------------+
> | 1 | ANY (``min`` = 1, ``max`` = 1) |
> +---+--------------------------------+
> | 2 | TCP |
> +---+--------------------------------+
>
> Knowing that TCP does not make sense with something other than IPv4 and IPv6
> as L3, such a pattern may be translated to two flow rules instead:
>
> +---+--------------------+
> | 0 | ETHER |
> +---+--------------------+
> | 1 | IPV4 (zeroed mask) |
> +---+--------------------+
> | 2 | TCP |
> +---+--------------------+
>
> +---+--------------------+
> | 0 | ETHER |
> +---+--------------------+
> | 1 | IPV6 (zeroed mask) |
> +---+--------------------+
> | 2 | TCP |
> +---+--------------------+
>
> Note that as soon as a ANY rule covers several layers, this approach may
> yield a large number of hidden flow rules. It is thus suggested to only
> support the most common scenarios (anything as L2 and/or L3).
I think "any" may make things confusing. How about if the NIC doesn't support IPv6? Should we return fail for this rule?
>
> Flow rules priority
> ~~~~~~~~~~~~~~~~~~~
>
> While it would naturally make sense, flow rules cannot be assumed to be
> processed by hardware in the same order as their creation for several
> reasons:
>
> - They may be managed internally as a tree or a hash table instead of a
> list.
> - Removing a flow rule before adding another one can either put the new rule
> at the end of the list or reuse a freed entry.
> - Duplication may occur when packets are matched by several rules.
>
> For overlapping rules (particularly in order to use the `PASSTHRU`_ action)
> predictable behavior is only guaranteed by using different priority levels.
>
> Priority levels are not necessarily implemented in hardware, or may be
> severely limited (e.g. a single priority bit).
>
> For these reasons, priority levels may be implemented purely in software by
> PMDs.
>
> - For devices expecting flow rules to be added in the correct order, PMDs
> may destroy and re-create existing rules after adding a new one with
> a higher priority.
>
> - A configurable number of dummy or empty rules can be created at
> initialization time to save high priority slots for later.
>
> - In order to save priority levels, PMDs may evaluate whether rules are
> likely to collide and adjust their priority accordingly.
If there's 3 rules, r1, r2,r3. The rules say the priority is r1 > r2 > r3. If PMD can only support r1 > r3 > r2, or doesn't support r2. Should PMD apply r1 and r3 or totally not support them all?
A generic question, is the parsing supposed to be done by rte or PMD?
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH] cryptodev: move new cryptodev type to bottom of enum
@ 2016-07-06 14:05 3% Pablo de Lara
2016-07-08 17:52 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Pablo de Lara @ 2016-07-06 14:05 UTC (permalink / raw)
To: dev; +Cc: declan.doherty, Pablo de Lara
New cryptodev type for the new KASUMI PMD was added
in the cryptodev type enum, but not at the end of it,
causing an ABI breakage.
Fixes: 2773c86d061a ("crypto/kasumi: add driver for KASUMI library")
Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
Reported-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
lib/librte_cryptodev/rte_cryptodev.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/librte_cryptodev/rte_cryptodev.h b/lib/librte_cryptodev/rte_cryptodev.h
index 7768f0a..508c1f7 100644
--- a/lib/librte_cryptodev/rte_cryptodev.h
+++ b/lib/librte_cryptodev/rte_cryptodev.h
@@ -67,9 +67,9 @@ enum rte_cryptodev_type {
RTE_CRYPTODEV_NULL_PMD = 1, /**< Null crypto PMD */
RTE_CRYPTODEV_AESNI_GCM_PMD, /**< AES-NI GCM PMD */
RTE_CRYPTODEV_AESNI_MB_PMD, /**< AES-NI multi buffer PMD */
- RTE_CRYPTODEV_KASUMI_PMD, /**< KASUMI PMD */
RTE_CRYPTODEV_QAT_SYM_PMD, /**< QAT PMD Symmetric Crypto */
RTE_CRYPTODEV_SNOW3G_PMD, /**< SNOW 3G PMD */
+ RTE_CRYPTODEV_KASUMI_PMD, /**< KASUMI PMD */
};
extern const char **rte_cyptodev_names;
--
2.5.5
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v3 1/3] kasumi: add new KASUMI PMD
2016-07-06 11:26 3% ` Ferruh Yigit
2016-07-06 13:07 0% ` Thomas Monjalon
@ 2016-07-06 13:22 0% ` De Lara Guarch, Pablo
1 sibling, 0 replies; 200+ results
From: De Lara Guarch, Pablo @ 2016-07-06 13:22 UTC (permalink / raw)
To: Yigit, Ferruh, dev; +Cc: Doherty, Declan, Jain, Deepak K
> -----Original Message-----
> From: Yigit, Ferruh
> Sent: Wednesday, July 06, 2016 12:26 PM
> To: De Lara Guarch, Pablo; dev@dpdk.org
> Cc: Doherty, Declan; Jain, Deepak K
> Subject: Re: [dpdk-dev] [PATCH v3 1/3] kasumi: add new KASUMI PMD
>
> On 6/20/2016 3:40 PM, Pablo de Lara wrote:
> > Added new SW PMD which makes use of the libsso_kasumi SW library,
> > which provides wireless algorithms KASUMI F8 and F9
> > in software.
> >
> > This PMD supports cipher-only, hash-only and chained operations
> > ("cipher then hash" and "hash then cipher") of the following
> > algorithms:
> > - RTE_CRYPTO_SYM_CIPHER_KASUMI_F8
> > - RTE_CRYPTO_SYM_AUTH_KASUMI_F9
> >
> > Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> > Acked-by: Jain, Deepak K <deepak.k.jain@intel.com>
>
> ...
>
> > --- a/lib/librte_cryptodev/rte_cryptodev.h
> > +++ b/lib/librte_cryptodev/rte_cryptodev.h
> > @@ -59,12 +59,15 @@ extern "C" {
> > /**< Intel QAT Symmetric Crypto PMD device name */
> > #define CRYPTODEV_NAME_SNOW3G_PMD ("cryptodev_snow3g_pmd")
> > /**< SNOW 3G PMD device name */
> > +#define CRYPTODEV_NAME_KASUMI_PMD ("cryptodev_kasumi_pmd")
> > +/**< KASUMI PMD device name */
> >
> > /** Crypto device type */
> > enum rte_cryptodev_type {
> > RTE_CRYPTODEV_NULL_PMD = 1, /**< Null crypto PMD */
> > RTE_CRYPTODEV_AESNI_GCM_PMD, /**< AES-NI GCM PMD */
> > RTE_CRYPTODEV_AESNI_MB_PMD, /**< AES-NI multi buffer PMD
> */
> > + RTE_CRYPTODEV_KASUMI_PMD, /**< KASUMI PMD */
> Does adding new field into middle cause a ABI breakage?
> Since now value of below fields changed.
Right! Thanks for the catch, will send a patch to fix that.
>
> Btw, librte_cryptodev is not listed in release notes, "shared library
> versions" section, not sure if this is intentional.
>
> > RTE_CRYPTODEV_QAT_SYM_PMD, /**< QAT PMD Symmetric
> Crypto */
> > RTE_CRYPTODEV_SNOW3G_PMD, /**< SNOW 3G PMD */
> > };
>
> ...
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v3 1/3] kasumi: add new KASUMI PMD
2016-07-06 11:26 3% ` Ferruh Yigit
@ 2016-07-06 13:07 0% ` Thomas Monjalon
2016-07-06 13:22 0% ` De Lara Guarch, Pablo
1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-06 13:07 UTC (permalink / raw)
To: Ferruh Yigit
Cc: dev, Pablo de Lara, declan.doherty, deepak.k.jain, reshma.pattan
2016-07-06 12:26, Ferruh Yigit:
> On 6/20/2016 3:40 PM, Pablo de Lara wrote:
> > enum rte_cryptodev_type {
> > RTE_CRYPTODEV_NULL_PMD = 1, /**< Null crypto PMD */
> > RTE_CRYPTODEV_AESNI_GCM_PMD, /**< AES-NI GCM PMD */
> > RTE_CRYPTODEV_AESNI_MB_PMD, /**< AES-NI multi buffer PMD */
> > + RTE_CRYPTODEV_KASUMI_PMD, /**< KASUMI PMD */
> Does adding new field into middle cause a ABI breakage?
> Since now value of below fields changed.
>
> Btw, librte_cryptodev is not listed in release notes, "shared library
> versions" section, not sure if this is intentional.
Good catch!
Now that crypto is not experimental anymore, we must add cryptodev in
release notes. librte_pdump is also missing in this list.
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH] librte_pmd_bond: fix exported symbol versioning
@ 2016-07-06 11:39 3% Christian Ehrhardt
2016-07-11 11:27 3% ` [dpdk-dev] [PATCH v2] " Christian Ehrhardt
0 siblings, 1 reply; 200+ results
From: Christian Ehrhardt @ 2016-07-06 11:39 UTC (permalink / raw)
To: Eric Kinzie, christian.ehrhardt, thomas.monjalon, dev
The older versions of rte_eth_bond_8023ad_conf_get and
rte_eth_bond_8023ad_setup were available in the old way since 2.0 - at
least according to the map file.
But versioning in the code was set to 16.04.
That breaks compatibility checks for 2.0 on that library.
For example with the dpdk abi checker:
http://people.canonical.com/~paelzer/compat_report.html
To fix, version the old symbols on the 2.0 version as they were
initially added to the map file.
See http://people.canonical.com/~paelzer/compat_report.html
Fixes: dc40f17a ("net/bonding: allow external state machine in mode 4")
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
---
drivers/net/bonding/rte_eth_bond_8023ad.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index 48a50e4..2f7ae70 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -1068,7 +1068,7 @@ bond_mode_8023ad_conf_assign(struct mode8023ad_private *mode4,
}
static void
-bond_mode_8023ad_setup_v1604(struct rte_eth_dev *dev,
+bond_mode_8023ad_setup_v20(struct rte_eth_dev *dev,
struct rte_eth_bond_8023ad_conf *conf)
{
struct rte_eth_bond_8023ad_conf def_conf;
@@ -1214,7 +1214,7 @@ free_out:
}
int
-rte_eth_bond_8023ad_conf_get_v1604(uint8_t port_id,
+rte_eth_bond_8023ad_conf_get_v20(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf)
{
struct rte_eth_dev *bond_dev;
@@ -1229,7 +1229,7 @@ rte_eth_bond_8023ad_conf_get_v1604(uint8_t port_id,
bond_mode_8023ad_conf_get(bond_dev, conf);
return 0;
}
-VERSION_SYMBOL(rte_eth_bond_8023ad_conf_get, _v1604, 16.04);
+VERSION_SYMBOL(rte_eth_bond_8023ad_conf_get, _v20, 2.0);
int
rte_eth_bond_8023ad_conf_get_v1607(uint8_t port_id,
@@ -1278,7 +1278,7 @@ bond_8023ad_setup_validate(uint8_t port_id,
}
int
-rte_eth_bond_8023ad_setup_v1604(uint8_t port_id,
+rte_eth_bond_8023ad_setup_v20(uint8_t port_id,
struct rte_eth_bond_8023ad_conf *conf)
{
struct rte_eth_dev *bond_dev;
@@ -1289,11 +1289,11 @@ rte_eth_bond_8023ad_setup_v1604(uint8_t port_id,
return err;
bond_dev = &rte_eth_devices[port_id];
- bond_mode_8023ad_setup_v1604(bond_dev, conf);
+ bond_mode_8023ad_setup_v20(bond_dev, conf);
return 0;
}
-VERSION_SYMBOL(rte_eth_bond_8023ad_setup, _v1604, 16.04);
+VERSION_SYMBOL(rte_eth_bond_8023ad_setup, _v20, 2.0);
int
rte_eth_bond_8023ad_setup_v1607(uint8_t port_id,
--
2.7.4
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v3 1/3] kasumi: add new KASUMI PMD
@ 2016-07-06 11:26 3% ` Ferruh Yigit
2016-07-06 13:07 0% ` Thomas Monjalon
2016-07-06 13:22 0% ` De Lara Guarch, Pablo
0 siblings, 2 replies; 200+ results
From: Ferruh Yigit @ 2016-07-06 11:26 UTC (permalink / raw)
To: Pablo de Lara, dev; +Cc: declan.doherty, deepak.k.jain
On 6/20/2016 3:40 PM, Pablo de Lara wrote:
> Added new SW PMD which makes use of the libsso_kasumi SW library,
> which provides wireless algorithms KASUMI F8 and F9
> in software.
>
> This PMD supports cipher-only, hash-only and chained operations
> ("cipher then hash" and "hash then cipher") of the following
> algorithms:
> - RTE_CRYPTO_SYM_CIPHER_KASUMI_F8
> - RTE_CRYPTO_SYM_AUTH_KASUMI_F9
>
> Signed-off-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
> Acked-by: Jain, Deepak K <deepak.k.jain@intel.com>
...
> --- a/lib/librte_cryptodev/rte_cryptodev.h
> +++ b/lib/librte_cryptodev/rte_cryptodev.h
> @@ -59,12 +59,15 @@ extern "C" {
> /**< Intel QAT Symmetric Crypto PMD device name */
> #define CRYPTODEV_NAME_SNOW3G_PMD ("cryptodev_snow3g_pmd")
> /**< SNOW 3G PMD device name */
> +#define CRYPTODEV_NAME_KASUMI_PMD ("cryptodev_kasumi_pmd")
> +/**< KASUMI PMD device name */
>
> /** Crypto device type */
> enum rte_cryptodev_type {
> RTE_CRYPTODEV_NULL_PMD = 1, /**< Null crypto PMD */
> RTE_CRYPTODEV_AESNI_GCM_PMD, /**< AES-NI GCM PMD */
> RTE_CRYPTODEV_AESNI_MB_PMD, /**< AES-NI multi buffer PMD */
> + RTE_CRYPTODEV_KASUMI_PMD, /**< KASUMI PMD */
Does adding new field into middle cause a ABI breakage?
Since now value of below fields changed.
Btw, librte_cryptodev is not listed in release notes, "shared library
versions" section, not sure if this is intentional.
> RTE_CRYPTODEV_QAT_SYM_PMD, /**< QAT PMD Symmetric Crypto */
> RTE_CRYPTODEV_SNOW3G_PMD, /**< SNOW 3G PMD */
> };
...
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v4] Pci: Add the class_id support
@ 2016-07-06 11:08 3% ` Ferruh Yigit
2016-07-07 7:46 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Ferruh Yigit @ 2016-07-06 11:08 UTC (permalink / raw)
To: Thomas Monjalon, Ziye Yang; +Cc: dev
On 6/14/2016 3:52 PM, Thomas Monjalon wrote:
> 2016-05-24 20:50, Ziye Yang:
>> This patch is used to add the class_id (class_code,
>> subclass_code, programming_interface) support for
>> pci_device probe. With this patch, it will be
>> flexible for users to probe a class of devices
>> by class_id.
>>
>>
>> Signed-off-by: Ziye Yang <ziye.yang@intel.com>
>
> Applied, thanks
>
Hi Thomas, Ziye,
Is modification in public "struct rte_pci_id" is a ABI break?
If so, it requires eal LIBABIVER increase and release notes update.
Regards,
ferruh
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] mk: filter duplicate configuration entries
@ 2016-07-06 5:37 3% ` Christian Ehrhardt
0 siblings, 0 replies; 200+ results
From: Christian Ehrhardt @ 2016-07-06 5:37 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: Ferruh Yigit, dev
Hi,
I came up with something very similar when looking for tac replacements
yesterday, but had no time to finish things.
But your suggestion is even shorter - I had found "sed -n '1{h;T;};G;h;$p;'
file" or "sed -n '1!G;h;$p'".
That removes the tac dependency, which I agree is a good thing.
To chain things up without a temp file one would need the "in-place"
features of sed&awk which I'm not sure they are available (awk >=4.1 and
only GNU awk).
sed -i is only used in validate-abi.sh which might not be used on all
platforms to count as "-i is there already so I can use it".
And I really don't want to break anyone due to that change, just naively
clean up the resulting config a bit.
Also we already have a temp file .config_tmp in the same scope and remove
it on our own.
So it is not that much different to create and remove a second one for that
section.
Thanks for both of your feedback, submitting v3 now ...
Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd
On Tue, Jul 5, 2016 at 9:47 PM, Thomas Monjalon <thomas.monjalon@6wind.com>
wrote:
> 2016-07-05 17:47, Ferruh Yigit:
> > On 6/30/2016 1:00 PM, Christian Ehrhardt wrote:
> > > + tac $(RTE_OUTPUT)/.config_tmp >
> $(RTE_OUTPUT)/.config_tmp_reverse ; \
> > Now we are adding new binary dependency (tac) to build system
>
> tac can be replaced by sed '1!G;h;$!d'
>
>
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [RFC] Generic flow director/filtering/classification API
@ 2016-07-05 18:16 2% Adrien Mazarguil
2016-07-07 7:14 0% ` Lu, Wenzhuo
` (2 more replies)
0 siblings, 3 replies; 200+ results
From: Adrien Mazarguil @ 2016-07-05 18:16 UTC (permalink / raw)
To: dev
Cc: Thomas Monjalon, Helin Zhang, Jingjing Wu, Rasesh Mody,
Ajit Khaparde, Rahul Lakkireddy, Wenzhuo Lu, Jan Medala,
John Daley, Jing Chen, Konstantin Ananyev, Matej Vido,
Alejandro Lucero, Sony Chacko, Jerin Jacob, Pablo de Lara,
Olga Shern
Hi All,
First, forgive me for this large message, I know our mailboxes already
suffer quite a bit from the amount of traffic on this ML.
This is not exactly yet another thread about how flow director should be
extended, rather about a brand new API to handle filtering and
classification for incoming packets in the most PMD-generic and
application-friendly fashion we can come up with. Reasons described below.
I think this topic is important enough to include both the users of this API
as well as PMD maintainers. So far I have CC'ed librte_ether (especially
rte_eth_ctrl.h contributors), testpmd and PMD maintainers (with and without
a .filter_ctrl implementation), but if you know application maintainers
other than testpmd who use FDIR or might be interested in this discussion,
feel free to add them.
The issues we found with the current approach are already summarized in the
following document, but here is a quick summary for TL;DR folks:
- PMDs do not expose a common set of filter types and even when they do,
their behavior more or less differs.
- Applications need to determine and adapt to device-specific limitations
and quirks on their own, without help from PMDs.
- Writing an application that creates flow rules targeting all devices
supported by DPDK is thus difficult, if not impossible.
- The current API has too many unspecified areas (particularly regarding
side effects of flow rules) that make PMD implementation tricky.
This RFC API handles everything currently supported by .filter_ctrl, the
idea being to reimplement all of these to make them fully usable by
applications in a more generic and well defined fashion. It has a very small
set of mandatory features and an easy method to let applications probe for
supported capabilities.
The only downside is more work for the software control side of PMDs because
they have to adapt to the API instead of the reverse. I think helpers can be
added to EAL to assist with this.
HTML version:
https://rawgit.com/6WIND/rte_flow/master/rte_flow.html
PDF version:
https://rawgit.com/6WIND/rte_flow/master/rte_flow.pdf
Related draft header file (for reference while reading the specification):
https://raw.githubusercontent.com/6WIND/rte_flow/master/rte_flow.h
Git tree for completeness (latest .rst version can be retrieved from here):
https://github.com/6WIND/rte_flow
What follows is the ReST source of the above, for inline comments and
discussion. I intend to update that specification accordingly.
========================
Generic filter interface
========================
.. footer::
v0.6
.. contents::
.. sectnum::
.. raw:: pdf
PageBreak
Overview
========
DPDK provides several competing interfaces added over time to perform packet
matching and related actions such as filtering and classification.
They must be extended to implement the features supported by newer devices
in order to expose them to applications, however the current design has
several drawbacks:
- Complicated filter combinations which have not been hard-coded cannot be
expressed.
- Prone to API/ABI breakage when new features must be added to an existing
filter type, which frequently happens.
>From an application point of view:
- Having disparate interfaces, all optional and lacking in features does not
make this API easy to use.
- Seemingly arbitrary built-in limitations of filter types based on the
device they were initially designed for.
- Undefined relationship between different filter types.
- High complexity, considerable undocumented and/or undefined behavior.
Considering the growing number of devices supported by DPDK, adding a new
filter type each time a new feature must be implemented is not sustainable
in the long term. Applications not written to target a specific device
cannot really benefit from such an API.
For these reasons, this document defines an extensible unified API that
encompasses and supersedes these legacy filter types.
.. raw:: pdf
PageBreak
Current API
===========
Rationale
---------
The reason several competing (and mostly overlapping) filtering APIs are
present in DPDK is due to its nature as a thin layer between hardware and
software.
Each subsequent interface has been added to better match the capabilities
and limitations of the latest supported device, which usually happened to
need an incompatible configuration approach. Because of this, many ended up
device-centric and not usable by applications that were not written for that
particular device.
This document is not the first attempt to address this proliferation issue,
in fact a lot of work has already been done both to create a more generic
interface while somewhat keeping compatibility with legacy ones through a
common call interface (``rte_eth_dev_filter_ctrl()`` with the
``.filter_ctrl`` PMD callback in ``rte_ethdev.h``).
Today, these previously incompatible interfaces are known as filter types
(``RTE_ETH_FILTER_*`` from ``enum rte_filter_type`` in ``rte_eth_ctrl.h``).
However while trivial to extend with new types, it only shifted the
underlying problem as applications still need to be written for one kind of
filter type, which, as described in the following sections, is not
necessarily implemented by all PMDs that support filtering.
.. raw:: pdf
PageBreak
Filter types
------------
This section summarizes the capabilities of each filter type.
Although the following list is exhaustive, the description of individual
types may contain inaccuracies due to the lack of documentation or usage
examples.
Note: names are prefixed with ``RTE_ETH_FILTER_``.
``MACVLAN``
~~~~~~~~~~~
Matching:
- L2 source/destination addresses.
- Optional 802.1Q VLAN ID.
- Masking individual fields on a rule basis is not supported.
Action:
- Packets are redirected either to a given VF device using its ID or to the
PF.
``ETHERTYPE``
~~~~~~~~~~~~~
Matching:
- L2 source/destination addresses (optional).
- Ethertype (no VLAN ID?).
- Masking individual fields on a rule basis is not supported.
Action:
- Receive packets on a given queue.
- Drop packets.
``FLEXIBLE``
~~~~~~~~~~~~
Matching:
- At most 128 consecutive bytes anywhere in packets.
- Masking is supported with byte granularity.
- Priorities are supported (relative to this filter type, undefined
otherwise).
Action:
- Receive packets on a given queue.
``SYN``
~~~~~~~
Matching:
- TCP SYN packets only.
- One high priority bit can be set to give the highest possible priority to
this type when other filters with different types are configured.
Action:
- Receive packets on a given queue.
``NTUPLE``
~~~~~~~~~~
Matching:
- Source/destination IPv4 addresses (optional in 2-tuple mode).
- Source/destination TCP/UDP port (mandatory in 2 and 5-tuple modes).
- L4 protocol (2 and 5-tuple modes).
- Masking individual fields is supported.
- TCP flags.
- Up to 7 levels of priority relative to this filter type, undefined
otherwise.
- No IPv6.
Action:
- Receive packets on a given queue.
``TUNNEL``
~~~~~~~~~~
Matching:
- Outer L2 source/destination addresses.
- Inner L2 source/destination addresses.
- Inner VLAN ID.
- IPv4/IPv6 source (destination?) address.
- Tunnel type to match (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE, 802.1BR
E-Tag).
- Tenant ID for tunneling protocols that have one.
- Any combination of the above can be specified.
- Masking individual fields on a rule basis is not supported.
Action:
- Receive packets on a given queue.
.. raw:: pdf
PageBreak
``FDIR``
~~~~~~~~
Queries:
- Device capabilities and limitations.
- Device statistics about configured filters (resource usage, collisions).
- Device configuration (matching input set and masks)
Matching:
- Device mode of operation: none (to disable filtering), signature
(hash-based dispatching from masked fields) or perfect (either MAC VLAN or
tunnel).
- L2 Ethertype.
- Outer L2 destination address (MAC VLAN mode).
- Inner L2 destination address, tunnel type (NVGRE, VXLAN) and tunnel ID
(tunnel mode).
- IPv4 source/destination addresses, ToS, TTL and protocol fields.
- IPv6 source/destination addresses, TC, protocol and hop limits fields.
- UDP source/destination IPv4/IPv6 and ports.
- TCP source/destination IPv4/IPv6 and ports.
- SCTP source/destination IPv4/IPv6, ports and verification tag field.
- Note, only one protocol type at once (either only L2 Ethertype, basic
IPv6, IPv4+UDP, IPv4+TCP and so on).
- VLAN TCI (extended API).
- At most 16 bytes to match in payload (extended API). A global device
look-up table specifies for each possible protocol layer (unknown, raw,
L2, L3, L4) the offset to use for each byte (they do not need to be
contiguous) and the related bitmask.
- Whether packet is addressed to PF or VF, in that case its ID can be
matched as well (extended API).
- Masking most of the above fields is supported, but simultaneously affects
all filters configured on a device.
- Input set can be modified in a similar fashion for a given device to
ignore individual fields of filters (i.e. do not match the destination
address in a IPv4 filter, refer to **RTE_ETH_INPUT_SET_**
macros). Configuring this also affects RSS processing on **i40e**.
- Filters can also provide 32 bits of arbitrary data to return as part of
matched packets.
Action:
- **RTE_ETH_FDIR_ACCEPT**: receive (accept) packet on a given queue.
- **RTE_ETH_FDIR_REJECT**: drop packet immediately.
- **RTE_ETH_FDIR_PASSTHRU**: similar to accept for the last filter in list,
otherwise process it with subsequent filters.
- For accepted packets and if requested by filter, either 32 bits of
arbitrary data and four bytes of matched payload (only in case of flex
bytes matching), or eight bytes of matched payload (flex also) are added
to meta data.
.. raw:: pdf
PageBreak
``HASH``
~~~~~~~~
Not an actual filter type. Provides and retrieves the global device
configuration (per port or entire NIC) for hash functions and their
properties.
Hash function selection: "default" (keep current), XOR or Toeplitz.
This function can be configured per flow type (**RTE_ETH_FLOW_**
definitions), supported types are:
- Unknown.
- Raw.
- Fragmented or non-fragmented IPv4.
- Non-fragmented IPv4 with L4 (TCP, UDP, SCTP or other).
- Fragmented or non-fragmented IPv6.
- Non-fragmented IPv6 with L4 (TCP, UDP, SCTP or other).
- L2 payload.
- IPv6 with extensions.
- IPv6 with L4 (TCP, UDP) and extensions.
``L2_TUNNEL``
~~~~~~~~~~~~~
Matching:
- All packets received on a given port.
Action:
- Add tunnel encapsulation (VXLAN, GENEVE, TEREDO, NVGRE, IP over GRE,
802.1BR E-Tag) using the provided Ethertype and tunnel ID (only E-Tag
is implemented at the moment).
- VF ID to use for tag insertion (currently unused).
- Destination pool for tag based forwarding (pools are IDs that can be
affected to ports, duplication occurs if the same ID is shared by several
ports of the same NIC).
.. raw:: pdf
PageBreak
Driver support
--------------
======== ======= ========= ======== === ====== ====== ==== ==== =========
Driver MACVLAN ETHERTYPE FLEXIBLE SYN NTUPLE TUNNEL FDIR HASH L2_TUNNEL
======== ======= ========= ======== === ====== ====== ==== ==== =========
bnx2x
cxgbe
e1000 yes yes yes yes
ena
enic yes
fm10k
i40e yes yes yes yes yes
ixgbe yes yes yes yes yes
mlx4
mlx5 yes
szedata2
======== ======= ========= ======== === ====== ====== ==== ==== =========
Flow director
-------------
Flow director (FDIR) is the name of the most capable filter type, which
covers most features offered by others. As such, it is the most widespread
in PMDs that support filtering (i.e. all of them besides **e1000**).
It is also the only type that allows an arbitrary 32 bits value provided by
applications to be attached to a filter and returned with matching packets
instead of relying on the destination queue to recognize flows.
Unfortunately, even FDIR requires applications to be aware of low-level
capabilities and limitations (most of which come directly from **ixgbe** and
**i40e**):
- Bitmasks are set globally per device (port?), not per filter.
- Configuration state is not expected to be saved by the driver, and
stopping/restarting a port requires the application to perform it again
(API documentation is also unclear about this).
- Monolithic approach with ABI issues as soon as a new kind of flow or
combination needs to be supported.
- Cryptic global statistics/counters.
- Unclear about how priorities are managed; filters seem to be arranged as a
linked list in hardware (possibly related to configuration order).
Packet alteration
-----------------
One interesting feature is that the L2 tunnel filter type implements the
ability to alter incoming packets through a filter (in this case to
encapsulate them), thus the **mlx5** flow encap/decap features are not a
foreign concept.
.. raw:: pdf
PageBreak
Proposed API
============
Terminology
-----------
- **Filtering API**: overall framework affecting the fate of selected
packets, covers everything described in this document.
- **Matching pattern**: properties to look for in received packets, a
combination of any number of items.
- **Pattern item**: part of a pattern that either matches packet data
(protocol header, payload or derived information), or specifies properties
of the pattern itself.
- **Actions**: what needs to be done when a packet matches a pattern.
- **Flow rule**: this is the result of combining a *matching pattern* with
*actions*.
- **Filter rule**: a less generic term than *flow rule*, can otherwise be
used interchangeably.
- **Hit**: a flow rule is said to be *hit* when processing a matching
packet.
Requirements
------------
As described in the previous section, there is a growing need for a common
method to configure filtering and related actions in a hardware independent
fashion.
The filtering API should not disallow any filter combination by design and
must remain as simple as possible to use. It can simply be defined as a
method to perform one or several actions on selected packets.
PMDs are aware of the capabilities of the device they manage and should be
responsible for preventing unsupported or conflicting combinations.
This approach is fundamentally different as it places most of the burden on
the software side of the PMD instead of having device capabilities directly
mapped to API functions, then expecting applications to work around ensuing
compatibility issues.
Requirements for a new API:
- Flexible and extensible without causing API/ABI problems for existing
applications.
- Should be unambiguous and easy to use.
- Support existing filtering features and actions listed in `Filter types`_.
- Support packet alteration.
- In case of overlapping filters, their priority should be well documented.
- Support filter queries (for example to retrieve counters).
.. raw:: pdf
PageBreak
High level design
-----------------
The chosen approach to make filtering as generic as possible is by
expressing matching patterns through lists of items instead of the flat
structures used in DPDK today, enabling combinations that are not predefined
and thus being more versatile.
Flow rules can have several distinct actions (such as counting,
encapsulating, decapsulating before redirecting packets to a particular
queue, etc.), instead of relying on several rules to achieve this and having
applications deal with hardware implementation details regarding their
order.
Support for different priority levels on a rule basis is provided, for
example in order to force a more specific rule come before a more generic
one for packets matched by both, however hardware support for more than a
single priority level cannot be guaranteed. When supported, the number of
available priority levels is usually low, which is why they can also be
implemented in software by PMDs (e.g. to simulate missing priority levels by
reordering rules).
In order to remain as hardware agnostic as possible, by default all rules
are considered to have the same priority, which means that the order between
overlapping rules (when a packet is matched by several filters) is
undefined, packet duplication may even occur as a result.
PMDs may refuse to create overlapping rules at a given priority level when
they can be detected (e.g. if a pattern matches an existing filter).
Thus predictable results for a given priority level can only be achieved
with non-overlapping rules, using perfect matching on all protocol layers.
Support for multiple actions per rule may be implemented internally on top
of non-default hardware priorities, as a result both features may not be
simultaneously available to applications.
Considering that allowed pattern/actions combinations cannot be known in
advance and would result in an unpractically large number of capabilities to
expose, a method is provided to validate a given rule from the current
device configuration state without actually adding it (akin to a "dry run"
mode).
This enables applications to check if the rule types they need is supported
at initialization time, before starting their data path. This method can be
used anytime, its only requirement being that the resources needed by a rule
must exist (e.g. a target RX queue must be configured first).
Each defined rule is associated with an opaque handle managed by the PMD,
applications are responsible for keeping it. These can be used for queries
and rules management, such as retrieving counters or other data and
destroying them.
Handles must be destroyed before releasing associated resources such as
queues.
Integration
-----------
To avoid ABI breakage, this new interface will be implemented through the
existing filtering control framework (``rte_eth_dev_filter_ctrl()``) using
**RTE_ETH_FILTER_GENERIC** as a new filter type.
However a public front-end API described in `Rules management`_ will
be added as the preferred method to use it.
Once discussions with the community have converged to a definite API, legacy
filter types should be deprecated and a deadline defined to remove their
support entirely.
PMDs will have to be gradually converted to **RTE_ETH_FILTER_GENERIC** or
drop filtering support entirely. Less maintained PMDs for older hardware may
lose support at this point.
The notion of filter type will then be deprecated and subsequently dropped
to avoid confusion between both frameworks.
Implementation details
======================
Flow rule
---------
A flow rule is the combination of a matching pattern with a list of actions,
and is the basis of this API.
Priorities
~~~~~~~~~~
A priority can be assigned to a matching pattern.
The default priority level is 0 and is also the highest. Support for more
than a single priority level in hardware is not guaranteed.
If a packet is matched by several filters at a given priority level, the
outcome is undefined. It can take any path and can even be duplicated.
Matching pattern
~~~~~~~~~~~~~~~~
A matching pattern comprises any number of items of various types.
Items are arranged in a list to form a matching pattern for packets. They
fall in two categories:
- Protocol matching (ANY, RAW, ETH, IPV4, IPV6, ICMP, UDP, TCP, VXLAN and so
on), usually associated with a specification structure. These must be
stacked in the same order as the protocol layers to match, starting from
L2.
- Affecting how the pattern is processed (END, VOID, INVERT, PF, VF,
SIGNATURE and so on), often without a specification structure. Since they
are meta data that does not match packet contents, these can be specified
anywhere within item lists without affecting the protocol matching items.
Most item specifications can be optionally paired with a mask to narrow the
specific fields or bits to be matched.
- Items are defined with ``struct rte_flow_item``.
- Patterns are defined with ``struct rte_flow_pattern``.
Example of an item specification matching an Ethernet header:
+-----------------------------------------+
| Ethernet |
+==========+=========+====================+
| ``spec`` | ``src`` | ``00:01:02:03:04`` |
| +---------+--------------------+
| | ``dst`` | ``00:2a:66:00:01`` |
+----------+---------+--------------------+
| ``mask`` | ``src`` | ``00:ff:ff:ff:00`` |
| +---------+--------------------+
| | ``dst`` | ``00:00:00:00:ff`` |
+----------+---------+--------------------+
Non-masked bits stand for any value, Ethernet headers with the following
properties are thus matched:
- ``src``: ``??:01:02:03:??``
- ``dst``: ``??:??:??:??:01``
Except for meta types that do not need one, ``spec`` must be a valid pointer
to a structure of the related item type. A ``mask`` of the same type can be
provided to tell which bits in ``spec`` are to be matched.
A mask is normally only needed for ``spec`` fields matching packet data,
ignored otherwise. See individual item types for more information.
A ``NULL`` mask pointer is allowed and is similar to matching with a full
mask (all ones) ``spec`` fields supported by hardware, the remaining fields
are ignored (all zeroes), there is thus no error checking for unsupported
fields.
Matching pattern items for packet data must be naturally stacked (ordered
from lowest to highest protocol layer), as in the following examples:
+--------------+
| TCPv4 as L4 |
+===+==========+
| 0 | Ethernet |
+---+----------+
| 1 | IPv4 |
+---+----------+
| 2 | TCP |
+---+----------+
+----------------+
| TCPv6 in VXLAN |
+===+============+
| 0 | Ethernet |
+---+------------+
| 1 | IPv4 |
+---+------------+
| 2 | UDP |
+---+------------+
| 3 | VXLAN |
+---+------------+
| 4 | Ethernet |
+---+------------+
| 5 | IPv6 |
+---+------------+
| 6 | TCP |
+---+------------+
+-----------------------------+
| TCPv4 as L4 with meta items |
+===+=========================+
| 0 | VOID |
+---+-------------------------+
| 1 | Ethernet |
+---+-------------------------+
| 2 | VOID |
+---+-------------------------+
| 3 | IPv4 |
+---+-------------------------+
| 4 | TCP |
+---+-------------------------+
| 5 | VOID |
+---+-------------------------+
| 6 | VOID |
+---+-------------------------+
The above example shows how meta items do not affect packet data matching
items, as long as those remain stacked properly. The resulting matching
pattern is identical to "TCPv4 as L4".
+----------------+
| UDPv6 anywhere |
+===+============+
| 0 | IPv6 |
+---+------------+
| 1 | UDP |
+---+------------+
If supported by the PMD, omitting one or several protocol layers at the
bottom of the stack as in the above example (missing an Ethernet
specification) enables hardware to look anywhere in packets.
It is unspecified whether the payload of supported encapsulations
(e.g. VXLAN inner packet) is matched by such a pattern, which may apply to
inner, outer or both packets.
+---------------------+
| Invalid, missing L3 |
+===+=================+
| 0 | Ethernet |
+---+-----------------+
| 1 | UDP |
+---+-----------------+
The above pattern is invalid due to a missing L3 specification between L2
and L4. It is only allowed at the bottom and at the top of the stack.
Meta item types
~~~~~~~~~~~~~~~
These do not match packet data but affect how the pattern is processed, most
of them do not need a specification structure. This particularity allows
them to be specified anywhere without affecting other item types.
``END``
^^^^^^^
End marker for item lists. Prevents further processing of items, thereby
ending the pattern.
- Its numeric value is **0** for convenience.
- PMD support is mandatory.
- Both ``spec`` and ``mask`` are ignored.
+--------------------+
| END |
+==========+=========+
| ``spec`` | ignored |
+----------+---------+
| ``mask`` | ignored |
+----------+---------+
``VOID``
^^^^^^^^
Used as a placeholder for convenience. It is ignored and simply discarded by
PMDs.
- PMD support is mandatory.
- Both ``spec`` and ``mask`` are ignored.
+--------------------+
| VOID |
+==========+=========+
| ``spec`` | ignored |
+----------+---------+
| ``mask`` | ignored |
+----------+---------+
One usage example for this type is generating rules that share a common
prefix quickly without reallocating memory, only by updating item types:
+------------------------+
| TCP, UDP or ICMP as L4 |
+===+====================+
| 0 | Ethernet |
+---+--------------------+
| 1 | IPv4 |
+---+------+------+------+
| 2 | UDP | VOID | VOID |
+---+------+------+------+
| 3 | VOID | TCP | VOID |
+---+------+------+------+
| 4 | VOID | VOID | ICMP |
+---+------+------+------+
.. raw:: pdf
PageBreak
``INVERT``
^^^^^^^^^^
Inverted matching, i.e. process packets that do not match the pattern.
- Both ``spec`` and ``mask`` are ignored.
+--------------------+
| INVERT |
+==========+=========+
| ``spec`` | ignored |
+----------+---------+
| ``mask`` | ignored |
+----------+---------+
Usage example in order to match non-TCPv4 packets only:
+--------------------+
| Anything but TCPv4 |
+===+================+
| 0 | INVERT |
+---+----------------+
| 1 | Ethernet |
+---+----------------+
| 2 | IPv4 |
+---+----------------+
| 3 | TCP |
+---+----------------+
``PF``
^^^^^^
Matches packets addressed to the physical function of the device.
- Both ``spec`` and ``mask`` are ignored.
+--------------------+
| PF |
+==========+=========+
| ``spec`` | ignored |
+----------+---------+
| ``mask`` | ignored |
+----------+---------+
``VF``
^^^^^^
Matches packets addressed to the given virtual function ID of the device.
- Only ``spec`` needs to be defined, ``mask`` is ignored.
+----------------------------------------+
| VF |
+==========+=========+===================+
| ``spec`` | ``vf`` | destination VF ID |
+----------+---------+-------------------+
| ``mask`` | ignored |
+----------+-----------------------------+
``SIGNATURE``
^^^^^^^^^^^^^
Requests hash-based signature dispatching for this rule.
Considering this is a global setting on devices that support it, all
subsequent filter rules may have to be created with it as well.
- Only ``spec`` needs to be defined, ``mask`` is ignored.
+--------------------+
| SIGNATURE |
+==========+=========+
| ``spec`` | TBD |
+----------+---------+
| ``mask`` | ignored |
+----------+---------+
.. raw:: pdf
PageBreak
Data matching item types
~~~~~~~~~~~~~~~~~~~~~~~~
Most of these are basically protocol header definitions with associated
bitmasks. They must be specified (stacked) from lowest to highest protocol
layer.
The following list is not exhaustive as new protocols will be added in the
future.
``ANY``
^^^^^^^
Matches any protocol in place of the current layer, a single ANY may also
stand for several protocol layers.
This is usually specified as the first pattern item when looking for a
protocol anywhere in a packet.
- A maximum value of **0** requests matching any number of protocol layers
above or equal to the minimum value, a maximum value lower than the
minimum one is otherwise invalid.
- Only ``spec`` needs to be defined, ``mask`` is ignored.
+-----------------------------------------------------------------------+
| ANY |
+==========+=========+==================================================+
| ``spec`` | ``min`` | minimum number of layers covered |
| +---------+--------------------------------------------------+
| | ``max`` | maximum number of layers covered, 0 for infinity |
+----------+---------+--------------------------------------------------+
| ``mask`` | ignored |
+----------+------------------------------------------------------------+
Example for VXLAN TCP payload matching regardless of outer L3 (IPv4 or IPv6)
and L4 (UDP) both matched by the first ANY specification, and inner L3 (IPv4
or IPv6) matched by the second ANY specification:
+----------------------------------+
| TCP in VXLAN with wildcards |
+===+==============================+
| 0 | Ethernet |
+---+-----+----------+---------+---+
| 1 | ANY | ``spec`` | ``min`` | 2 |
| | | +---------+---+
| | | | ``max`` | 2 |
+---+-----+----------+---------+---+
| 2 | VXLAN |
+---+------------------------------+
| 3 | Ethernet |
+---+-----+----------+---------+---+
| 4 | ANY | ``spec`` | ``min`` | 1 |
| | | +---------+---+
| | | | ``max`` | 1 |
+---+-----+----------+---------+---+
| 5 | TCP |
+---+------------------------------+
.. raw:: pdf
PageBreak
``RAW``
^^^^^^^
Matches a string of a given length at a given offset (in bytes), or anywhere
in the payload of the current protocol layer (including L2 header if used as
the first item in the stack).
This does not increment the protocol layer count as it is not a protocol
definition. Subsequent RAW items modulate the first absolute one with
relative offsets.
- Using **-1** as the ``offset`` of the first RAW item makes its absolute
offset not fixed, i.e. the pattern is searched everywhere.
- ``mask`` only affects the pattern.
+--------------------------------------------------------------+
| RAW |
+==========+=============+=====================================+
| ``spec`` | ``offset`` | absolute or relative pattern offset |
| +-------------+-------------------------------------+
| | ``length`` | pattern length |
| +-------------+-------------------------------------+
| | ``pattern`` | byte string of the above length |
+----------+-------------+-------------------------------------+
| ``mask`` | ``offset`` | ignored |
| +-------------+-------------------------------------+
| | ``length`` | ignored |
| +-------------+-------------------------------------+
| | ``pattern`` | bitmask with the same byte length |
+----------+-------------+-------------------------------------+
Example pattern looking for several strings at various offsets of a UDP
payload, using combined RAW items:
+------------------------------------------+
| UDP payload matching |
+===+======================================+
| 0 | Ethernet |
+---+--------------------------------------+
| 1 | IPv4 |
+---+--------------------------------------+
| 2 | UDP |
+---+-----+----------+-------------+-------+
| 3 | RAW | ``spec`` | ``offset`` | -1 |
| | | +-------------+-------+
| | | | ``length`` | 3 |
| | | +-------------+-------+
| | | | ``pattern`` | "foo" |
+---+-----+----------+-------------+-------+
| 4 | RAW | ``spec`` | ``offset`` | 20 |
| | | +-------------+-------+
| | | | ``length`` | 3 |
| | | +-------------+-------+
| | | | ``pattern`` | "bar" |
+---+-----+----------+-------------+-------+
| 5 | RAW | ``spec`` | ``offset`` | -30 |
| | | +-------------+-------+
| | | | ``length`` | 3 |
| | | +-------------+-------+
| | | | ``pattern`` | "baz" |
+---+-----+----------+-------------+-------+
This translates to:
- Locate "foo" in UDP payload, remember its offset.
- Check "bar" at "foo"'s offset plus 20 bytes.
- Check "baz" at "foo"'s offset minus 30 bytes.
.. raw:: pdf
PageBreak
``ETH``
^^^^^^^
Matches an Ethernet header.
- ``dst``: destination MAC.
- ``src``: source MAC.
- ``type``: EtherType.
- ``tags``: number of 802.1Q/ad tags defined.
- ``tag[]``: 802.1Q/ad tag definitions, innermost first. For each one:
- ``tpid``: Tag protocol identifier.
- ``tci``: Tag control information.
``IPV4``
^^^^^^^^
Matches an IPv4 header.
- ``src``: source IP address.
- ``dst``: destination IP address.
- ``tos``: ToS/DSCP field.
- ``ttl``: TTL field.
- ``proto``: protocol number for the next layer.
``IPV6``
^^^^^^^^
Matches an IPv6 header.
- ``src``: source IP address.
- ``dst``: destination IP address.
- ``tc``: traffic class field.
- ``nh``: Next header field (protocol).
- ``hop_limit``: hop limit field (TTL).
``ICMP``
^^^^^^^^
Matches an ICMP header.
- TBD.
``UDP``
^^^^^^^
Matches a UDP header.
- ``sport``: source port.
- ``dport``: destination port.
- ``length``: UDP length.
- ``checksum``: UDP checksum.
.. raw:: pdf
PageBreak
``TCP``
^^^^^^^
Matches a TCP header.
- ``sport``: source port.
- ``dport``: destination port.
- All other TCP fields and bits.
``VXLAN``
^^^^^^^^^
Matches a VXLAN header.
- TBD.
.. raw:: pdf
PageBreak
Actions
~~~~~~~
Each possible action is represented by a type. Some have associated
configuration structures. Several actions combined in a list can be affected
to a flow rule. That list is not ordered.
At least one action must be defined in a filter rule in order to do
something with matched packets.
- Actions are defined with ``struct rte_flow_action``.
- A list of actions is defined with ``struct rte_flow_actions``.
They fall in three categories:
- Terminating actions (such as QUEUE, DROP, RSS, PF, VF) that prevent
processing matched packets by subsequent flow rules, unless overridden
with PASSTHRU.
- Non terminating actions (PASSTHRU, DUP) that leave matched packets up for
additional processing by subsequent flow rules.
- Other non terminating meta actions that do not affect the fate of packets
(END, VOID, ID, COUNT).
When several actions are combined in a flow rule, they should all have
different types (e.g. dropping a packet twice is not possible). However
considering the VOID type is an exception to this rule, the defined behavior
is for PMDs to only take into account the last action of a given type found
in the list. PMDs still perform error checking on the entire list.
*Note that PASSTHRU is the only action able to override a terminating rule.*
.. raw:: pdf
PageBreak
Example of an action that redirects packets to queue index 10:
+----------------+
| QUEUE |
+===========+====+
| ``queue`` | 10 |
+-----------+----+
Action lists examples, their order is not significant, applications must
consider all actions to be performed simultaneously:
+----------------+
| Count and drop |
+=======+========+
| COUNT | |
+-------+--------+
| DROP | |
+-------+--------+
+--------------------------+
| Tag, count and redirect |
+=======+===========+======+
| ID | ``id`` | 0x2a |
+-------+-----------+------+
| COUNT | |
+-------+-----------+------+
| QUEUE | ``queue`` | 10 |
+-------+-----------+------+
+-----------------------+
| Redirect to queue 5 |
+=======+===============+
| DROP | |
+-------+-----------+---+
| QUEUE | ``queue`` | 5 |
+-------+-----------+---+
In the above example, considering both actions are performed simultaneously,
its end result is that only QUEUE has any effect.
+-----------------------+
| Redirect to queue 3 |
+=======+===========+===+
| QUEUE | ``queue`` | 5 |
+-------+-----------+---+
| VOID | |
+-------+-----------+---+
| QUEUE | ``queue`` | 3 |
+-------+-----------+---+
As previously described, only the last action of a given type found in the
list is taken into account. The above example also shows that VOID is
ignored.
.. raw:: pdf
PageBreak
Action types
~~~~~~~~~~~~
Common action types are described in this section. Like pattern item types,
this list is not exhaustive as new actions will be added in the future.
``END`` (action)
^^^^^^^^^^^^^^^^
End marker for action lists. Prevents further processing of actions, thereby
ending the list.
- Its numeric value is **0** for convenience.
- PMD support is mandatory.
- No configurable property.
+---------------+
| END |
+===============+
| no properties |
+---------------+
``VOID`` (action)
^^^^^^^^^^^^^^^^^
Used as a placeholder for convenience. It is ignored and simply discarded by
PMDs.
- PMD support is mandatory.
- No configurable property.
+---------------+
| VOID |
+===============+
| no properties |
+---------------+
``PASSTHRU``
^^^^^^^^^^^^
Leaves packets up for additional processing by subsequent flow rules. This
is the default when a rule does not contain a terminating action, but can be
specified to force a rule to become non-terminating.
- No configurable property.
+---------------+
| PASSTHRU |
+===============+
| no properties |
+---------------+
Example to copy a packet to a queue and continue processing by subsequent
flow rules:
+--------------------------+
| Copy to queue 8 |
+==========+===============+
| PASSTHRU | |
+----------+-----------+---+
| QUEUE | ``queue`` | 8 |
+----------+-----------+---+
``ID``
^^^^^^
Attaches a 32 bit value to packets.
+----------------------------------------------+
| ID |
+========+=====================================+
| ``id`` | 32 bit value to return with packets |
+--------+-------------------------------------+
.. raw:: pdf
PageBreak
``QUEUE``
^^^^^^^^^
Assigns packets to a given queue index.
- Terminating by default.
+--------------------------------+
| QUEUE |
+===========+====================+
| ``queue`` | queue index to use |
+-----------+--------------------+
``DROP``
^^^^^^^^
Drop packets.
- No configurable property.
- Terminating by default.
- PASSTHRU overrides this action if both are specified.
+---------------+
| DROP |
+===============+
| no properties |
+---------------+
``COUNT``
^^^^^^^^^
Enables hits counter for this rule.
This counter can be retrieved and reset through ``rte_flow_query()``, see
``struct rte_flow_query_count``.
- Counters can be retrieved with ``rte_flow_query()``.
- No configurable property.
+---------------+
| COUNT |
+===============+
| no properties |
+---------------+
Query structure to retrieve and reset the flow rule hits counter:
+------------------------------------------------+
| COUNT query |
+===========+=====+==============================+
| ``reset`` | in | reset counter after query |
+-----------+-----+------------------------------+
| ``hits`` | out | number of hits for this flow |
+-----------+-----+------------------------------+
``DUP``
^^^^^^^
Duplicates packets to a given queue index.
This is normally combined with QUEUE, however when used alone, it is
actually similar to QUEUE + PASSTHRU.
- Non-terminating by default.
+------------------------------------------------+
| DUP |
+===========+====================================+
| ``queue`` | queue index to duplicate packet to |
+-----------+------------------------------------+
.. raw:: pdf
PageBreak
``RSS``
^^^^^^^
Similar to QUEUE, except RSS is additionally performed on packets to spread
them among several queues according to the provided parameters.
- Terminating by default.
+---------------------------------------------+
| RSS |
+==============+==============================+
| ``rss_conf`` | RSS parameters |
+--------------+------------------------------+
| ``queues`` | number of entries in queue[] |
+--------------+------------------------------+
| ``queue[]`` | queue indices to use |
+--------------+------------------------------+
``PF`` (action)
^^^^^^^^^^^^^^^
Redirects packets to the physical function (PF) of the current device.
- No configurable property.
- Terminating by default.
+---------------+
| PF |
+===============+
| no properties |
+---------------+
``VF`` (action)
^^^^^^^^^^^^^^^
Redirects packets to the virtual function (VF) of the current device with
the specified ID.
- Terminating by default.
+---------------------------------------+
| VF |
+========+==============================+
| ``id`` | VF ID to redirect packets to |
+--------+------------------------------+
Planned types
~~~~~~~~~~~~~
Other action types are planned but not defined yet. These actions will add
the ability to alter matching packets in several ways, such as performing
encapsulation/decapsulation of tunnel headers on specific flows.
.. raw:: pdf
PageBreak
Rules management
----------------
A simple API with only four functions is provided to fully manage flows.
Each created flow rule is associated with an opaque, PMD-specific handle
pointer. The application is responsible for keeping it until the rule is
destroyed.
Flows rules are defined with ``struct rte_flow``.
Validation
~~~~~~~~~~
Given that expressing a definite set of device capabilities with this API is
not practical, a dedicated function is provided to check if a flow rule is
supported and can be created.
::
int
rte_flow_validate(uint8_t port_id,
const struct rte_flow_pattern *pattern,
const struct rte_flow_actions *actions);
While this function has no effect on the target device, the flow rule is
validated against its current configuration state and the returned value
should be considered valid by the caller for that state only.
The returned value is guaranteed to remain valid only as long as no
successful calls to rte_flow_create() or rte_flow_destroy() are made in the
meantime and no device parameter affecting flow rules in any way are
modified, due to possible collisions or resource limitations (although in
such cases ``EINVAL`` should not be returned).
Arguments:
- ``port_id``: port identifier of Ethernet device.
- ``pattern``: pattern specification to check.
- ``actions``: actions associated with the flow definition.
Return value:
- **0** if flow rule is valid and can be created. A negative errno value
otherwise (``rte_errno`` is also set), the following errors are defined.
- ``-EINVAL``: unknown or invalid rule specification.
- ``-ENOTSUP``: valid but unsupported rule specification (e.g. partial masks
are unsupported).
- ``-EEXIST``: collision with an existing rule.
- ``-ENOMEM``: not enough resources.
.. raw:: pdf
PageBreak
Creation
~~~~~~~~
Creating a flow rule is similar to validating one, except the rule is
actually created.
::
struct rte_flow *
rte_flow_create(uint8_t port_id,
const struct rte_flow_pattern *pattern,
const struct rte_flow_actions *actions);
Arguments:
- ``port_id``: port identifier of Ethernet device.
- ``pattern``: pattern specification to add.
- ``actions``: actions associated with the flow definition.
Return value:
A valid flow pointer in case of success, NULL otherwise and ``rte_errno`` is
set to the positive version of one of the error codes defined for
``rte_flow_validate()``.
Destruction
~~~~~~~~~~~
Flow rules destruction is not automatic, and a queue should not be released
if any are still attached to it. Applications must take care of performing
this step before releasing resources.
::
int
rte_flow_destroy(uint8_t port_id,
struct rte_flow *flow);
Failure to destroy a flow rule may occur when other flow rules depend on it,
and destroying it would result in an inconsistent state.
This function is only guaranteed to succeed if flow rules are destroyed in
reverse order of their creation.
Arguments:
- ``port_id``: port identifier of Ethernet device.
- ``flow``: flow rule to destroy.
Return value:
- **0** on success, a negative errno value otherwise and ``rte_errno`` is
set.
.. raw:: pdf
PageBreak
Query
~~~~~
Query an existing flow rule.
This function allows retrieving flow-specific data such as counters. Data
is gathered by special actions which must be present in the flow rule
definition.
::
int
rte_flow_query(uint8_t port_id,
struct rte_flow *flow,
enum rte_flow_action_type action,
void *data);
Arguments:
- ``port_id``: port identifier of Ethernet device.
- ``flow``: flow rule to query.
- ``action``: action type to query.
- ``data``: pointer to storage for the associated query data type.
Return value:
- **0** on success, a negative errno value otherwise and ``rte_errno`` is
set.
.. raw:: pdf
PageBreak
Behavior
--------
- API operations are synchronous and blocking (``EAGAIN`` cannot be
returned).
- There is no provision for reentrancy/multi-thread safety, although nothing
should prevent different devices from being configured at the same
time. PMDs may protect their control path functions accordingly.
- Stopping the data path (TX/RX) should not be necessary when managing flow
rules. If this cannot be achieved naturally or with workarounds (such as
temporarily replacing the burst function pointers), an appropriate error
code must be returned (``EBUSY``).
- PMDs, not applications, are responsible for maintaining flow rules
configuration when stopping and restarting a port or performing other
actions which may affect them. They can only be destroyed explicitly.
.. raw:: pdf
PageBreak
Compatibility
-------------
No known hardware implementation supports all the features described in this
document.
Unsupported features or combinations are not expected to be fully emulated
in software by PMDs for performance reasons. Partially supported features
may be completed in software as long as hardware performs most of the work
(such as queue redirection and packet recognition).
However PMDs are expected to do their best to satisfy application requests
by working around hardware limitations as long as doing so does not affect
the behavior of existing flow rules.
The following sections provide a few examples of such cases, they are based
on limitations built into the previous APIs.
Global bitmasks
~~~~~~~~~~~~~~~
Each flow rule comes with its own, per-layer bitmasks, while hardware may
support only a single, device-wide bitmask for a given layer type, so that
two IPv4 rules cannot use different bitmasks.
The expected behavior in this case is that PMDs automatically configure
global bitmasks according to the needs of the first created flow rule.
Subsequent rules are allowed only if their bitmasks match those, the
``EEXIST`` error code should be returned otherwise.
Unsupported layer types
~~~~~~~~~~~~~~~~~~~~~~~
Many protocols can be simulated by crafting patterns with the `RAW`_ type.
PMDs can rely on this capability to simulate support for protocols with
fixed headers not directly recognized by hardware.
``ANY`` pattern item
~~~~~~~~~~~~~~~~~~~~
This pattern item stands for anything, which can be difficult to translate
to something hardware would understand, particularly if followed by more
specific types.
Consider the following pattern:
+---+--------------------------------+
| 0 | ETHER |
+---+--------------------------------+
| 1 | ANY (``min`` = 1, ``max`` = 1) |
+---+--------------------------------+
| 2 | TCP |
+---+--------------------------------+
Knowing that TCP does not make sense with something other than IPv4 and IPv6
as L3, such a pattern may be translated to two flow rules instead:
+---+--------------------+
| 0 | ETHER |
+---+--------------------+
| 1 | IPV4 (zeroed mask) |
+---+--------------------+
| 2 | TCP |
+---+--------------------+
+---+--------------------+
| 0 | ETHER |
+---+--------------------+
| 1 | IPV6 (zeroed mask) |
+---+--------------------+
| 2 | TCP |
+---+--------------------+
Note that as soon as a ANY rule covers several layers, this approach may
yield a large number of hidden flow rules. It is thus suggested to only
support the most common scenarios (anything as L2 and/or L3).
.. raw:: pdf
PageBreak
Unsupported actions
~~~~~~~~~~~~~~~~~~~
- When combined with a `QUEUE`_ action, packet counting (`COUNT`_) and
tagging (`ID`_) may be implemented in software as long as the target queue
is used by a single rule.
- A rule specifying both `DUP`_ + `QUEUE`_ may be translated to two hidden
rules combining `QUEUE`_ and `PASSTHRU`_.
- When a single target queue is provided, `RSS`_ can also be implemented
through `QUEUE`_.
Flow rules priority
~~~~~~~~~~~~~~~~~~~
While it would naturally make sense, flow rules cannot be assumed to be
processed by hardware in the same order as their creation for several
reasons:
- They may be managed internally as a tree or a hash table instead of a
list.
- Removing a flow rule before adding another one can either put the new rule
at the end of the list or reuse a freed entry.
- Duplication may occur when packets are matched by several rules.
For overlapping rules (particularly in order to use the `PASSTHRU`_ action)
predictable behavior is only guaranteed by using different priority levels.
Priority levels are not necessarily implemented in hardware, or may be
severely limited (e.g. a single priority bit).
For these reasons, priority levels may be implemented purely in software by
PMDs.
- For devices expecting flow rules to be added in the correct order, PMDs
may destroy and re-create existing rules after adding a new one with
a higher priority.
- A configurable number of dummy or empty rules can be created at
initialization time to save high priority slots for later.
- In order to save priority levels, PMDs may evaluate whether rules are
likely to collide and adjust their priority accordingly.
.. raw:: pdf
PageBreak
API migration
=============
Exhaustive list of deprecated filter types and how to convert them to
generic flow rules.
``MACVLAN`` to ``ETH`` → ``VF``, ``PF``
---------------------------------------
`MACVLAN`_ can be translated to a basic `ETH`_ flow rule with a `VF
(action)`_ or `PF (action)`_ terminating action.
+------------------------------------+
| MACVLAN |
+--------------------------+---------+
| Pattern | Actions |
+===+=====+==========+=====+=========+
| 0 | ETH | ``spec`` | any | VF, |
| | +----------+-----+ PF |
| | | ``mask`` | any | |
+---+-----+----------+-----+---------+
``ETHERTYPE`` to ``ETH`` → ``QUEUE``, ``DROP``
----------------------------------------------
`ETHERTYPE`_ is basically an `ETH`_ flow rule with `QUEUE`_ or `DROP`_ as
a terminating action.
+------------------------------------+
| ETHERTYPE |
+--------------------------+---------+
| Pattern | Actions |
+===+=====+==========+=====+=========+
| 0 | ETH | ``spec`` | any | QUEUE, |
| | +----------+-----+ DROP |
| | | ``mask`` | any | |
+---+-----+----------+-----+---------+
``FLEXIBLE`` to ``RAW`` → ``QUEUE``
-----------------------------------
`FLEXIBLE`_ can be translated to one `RAW`_ pattern with `QUEUE`_ as the
terminating action and a defined priority level.
+------------------------------------+
| FLEXIBLE |
+--------------------------+---------+
| Pattern | Actions |
+===+=====+==========+=====+=========+
| 0 | RAW | ``spec`` | any | QUEUE |
| | +----------+-----+ |
| | | ``mask`` | any | |
+---+-----+----------+-----+---------+
``SYN`` to ``TCP`` → ``QUEUE``
------------------------------
`SYN`_ is a `TCP`_ rule with only the ``syn`` bit enabled and masked, and
`QUEUE`_ as the terminating action.
Priority level can be set to simulate the high priority bit.
+---------------------------------------------+
| SYN |
+-----------------------------------+---------+
| Pattern | Actions |
+===+======+==========+=============+=========+
| 0 | ETH | ``spec`` | N/A | QUEUE |
| | +----------+-------------+ |
| | | ``mask`` | empty | |
+---+------+----------+-------------+ |
| 1 | IPV4 | ``spec`` | N/A | |
| | +----------+-------------+ |
| | | ``mask`` | empty | |
+---+------+----------+-------------+ |
| 2 | TCP | ``spec`` | ``syn`` = 1 | |
| | +----------+-------------+ |
| | | ``mask`` | ``syn`` = 1 | |
+---+------+----------+-------------+---------+
``NTUPLE`` to ``IPV4``, ``TCP``, ``UDP`` → ``QUEUE``
----------------------------------------------------
`NTUPLE`_ is similar to specifying an empty L2, `IPV4`_ as L3 with `TCP`_ or
`UDP`_ as L4 and `QUEUE`_ as the terminating action.
A priority level can be specified as well.
+---------------------------------------+
| NTUPLE |
+-----------------------------+---------+
| Pattern | Actions |
+===+======+==========+=======+=========+
| 0 | ETH | ``spec`` | N/A | QUEUE |
| | +----------+-------+ |
| | | ``mask`` | empty | |
+---+------+----------+-------+ |
| 1 | IPV4 | ``spec`` | any | |
| | +----------+-------+ |
| | | ``mask`` | any | |
+---+------+----------+-------+ |
| 2 | TCP, | ``spec`` | any | |
| | UDP +----------+-------+ |
| | | ``mask`` | any | |
+---+------+----------+-------+---------+
``TUNNEL`` to ``ETH``, ``IPV4``, ``IPV6``, ``VXLAN`` (or other) → ``QUEUE``
---------------------------------------------------------------------------
`TUNNEL`_ matches common IPv4 and IPv6 L3/L4-based tunnel types.
In the following table, `ANY`_ is used to cover the optional L4.
+------------------------------------------------+
| TUNNEL |
+--------------------------------------+---------+
| Pattern | Actions |
+===+=========+==========+=============+=========+
| 0 | ETH | ``spec`` | any | QUEUE |
| | +----------+-------------+ |
| | | ``mask`` | any | |
+---+---------+----------+-------------+ |
| 1 | IPV4, | ``spec`` | any | |
| | IPV6 +----------+-------------+ |
| | | ``mask`` | any | |
+---+---------+----------+-------------+ |
| 2 | ANY | ``spec`` | ``min`` = 0 | |
| | | +-------------+ |
| | | | ``max`` = 0 | |
| | +----------+-------------+ |
| | | ``mask`` | N/A | |
+---+---------+----------+-------------+ |
| 3 | VXLAN, | ``spec`` | any | |
| | GENEVE, +----------+-------------+ |
| | TEREDO, | ``mask`` | any | |
| | NVGRE, | | | |
| | GRE, | | | |
| | ... | | | |
+---+---------+----------+-------------+---------+
.. raw:: pdf
PageBreak
``FDIR`` to most item types → ``QUEUE``, ``DROP``, ``PASSTHRU``
---------------------------------------------------------------
`FDIR`_ is more complex than any other type, there are several methods to
emulate its functionality. It is summarized for the most part in the table
below.
A few features are intentionally not supported:
- The ability to configure the matching input set and masks for the entire
device, PMDs should take care of it automatically according to flow rules.
- Returning four or eight bytes of matched data when using flex bytes
filtering. Although a specific action could implement it, it conflicts
with the much more useful 32 bits tagging on devices that support it.
- Side effects on RSS processing of the entire device. Flow rules that
conflict with the current device configuration should not be
allowed. Similarly, device configuration should not be allowed when it
affects existing flow rules.
- Device modes of operation. "none" is unsupported since filtering cannot be
disabled as long as a flow rule is present.
- "MAC VLAN" or "tunnel" perfect matching modes should be automatically set
according to the created flow rules.
+----------------------------------------------+
| FDIR |
+---------------------------------+------------+
| Pattern | Actions |
+===+============+==========+=====+============+
| 0 | ETH, | ``spec`` | any | QUEUE, |
| | RAW +----------+-----+ DROP, |
| | | ``mask`` | any | PASSTHRU |
+---+------------+----------+-----+------------+
| 1 | IPV4, | ``spec`` | any | ID |
| | IPV6 +----------+-----+ (optional) |
| | | ``mask`` | any | |
+---+------------+----------+-----+ |
| 2 | TCP, | ``spec`` | any | |
| | UDP, +----------+-----+ |
| | SCTP | ``mask`` | any | |
+---+------------+----------+-----+ |
| 3 | VF, | ``spec`` | any | |
| | PF, +----------+-----+ |
| | SIGNATURE | ``mask`` | any | |
| | (optional) | | | |
+---+------------+----------+-----+------------+
``HASH``
~~~~~~~~
Hashing configuration is set per rule through the `SIGNATURE`_ item.
Since it is usually a global device setting, all flow rules created with
this item may have to share the same specification.
``L2_TUNNEL`` to ``VOID`` → ``VXLAN`` (or others)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
All packets are matched. This type alters incoming packets to encapsulate
them in a chosen tunnel type, optionally redirect them to a VF as well.
The destination pool for tag based forwarding can be emulated with other
flow rules using `DUP`_ as the action.
+----------------------------------------+
| L2_TUNNEL |
+---------------------------+------------+
| Pattern | Actions |
+===+======+==========+=====+============+
| 0 | VOID | ``spec`` | N/A | VXLAN, |
| | | | | GENEVE, |
| | | | | ... |
| | +----------+-----+------------+
| | | ``mask`` | N/A | VF |
| | | | | (optional) |
+---+------+----------+-----+------------+
--
Adrien Mazarguil
6WIND
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH 01/18] doc: add template for release notes 16.11
@ 2016-07-05 15:41 6% ` Olivier Matz
0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-07-05 15:41 UTC (permalink / raw)
To: dev
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
doc/guides/rel_notes/release_16_11.rst | 160 +++++++++++++++++++++++++++++++++
1 file changed, 160 insertions(+)
create mode 100644 doc/guides/rel_notes/release_16_11.rst
diff --git a/doc/guides/rel_notes/release_16_11.rst b/doc/guides/rel_notes/release_16_11.rst
new file mode 100644
index 0000000..0106bc9
--- /dev/null
+++ b/doc/guides/rel_notes/release_16_11.rst
@@ -0,0 +1,160 @@
+DPDK Release 16.11
+==================
+
+.. **Read this first.**
+
+ The text below explains how to update the release notes.
+
+ Use proper spelling, capitalization and punctuation in all sections.
+
+ Variable and config names should be quoted as fixed width text: ``LIKE_THIS``.
+
+ Build the docs and view the output file to ensure the changes are correct::
+
+ make doc-guides-html
+
+ firefox build/doc/html/guides/rel_notes/release_16_11.html
+
+
+New Features
+------------
+
+.. This section should contain new features added in this release. Sample format:
+
+ * **Add a title in the past tense with a full stop.**
+
+ Add a short 1-2 sentence description in the past tense. The description
+ should be enough to allow someone scanning the release notes to understand
+ the new feature.
+
+ If the feature adds a lot of sub-features you can use a bullet list like this.
+
+ * Added feature foo to do something.
+ * Enhanced feature bar to do something else.
+
+ Refer to the previous release notes for examples.
+
+
+Resolved Issues
+---------------
+
+.. This section should contain bug fixes added to the relevant sections. Sample format:
+
+ * **code/section Fixed issue in the past tense with a full stop.**
+
+ Add a short 1-2 sentence description of the resolved issue in the past tense.
+ The title should contain the code/lib section like a commit message.
+ Add the entries in alphabetic order in the relevant sections below.
+
+
+EAL
+~~~
+
+
+Drivers
+~~~~~~~
+
+
+Libraries
+~~~~~~~~~
+
+
+Examples
+~~~~~~~~
+
+
+Other
+~~~~~
+
+
+Known Issues
+------------
+
+.. This section should contain new known issues in this release. Sample format:
+
+ * **Add title in present tense with full stop.**
+
+ Add a short 1-2 sentence description of the known issue in the present
+ tense. Add information on any known workarounds.
+
+
+API Changes
+-----------
+
+.. This section should contain API changes. Sample format:
+
+ * Add a short 1-2 sentence description of the API change. Use fixed width
+ quotes for ``rte_function_names`` or ``rte_struct_names``. Use the past tense.
+
+
+ABI Changes
+-----------
+
+.. * Add a short 1-2 sentence description of the ABI change that was announced in
+ the previous releases and made in this release. Use fixed width quotes for
+ ``rte_function_names`` or ``rte_struct_names``. Use the past tense.
+
+
+Shared Library Versions
+-----------------------
+
+.. Update any library version updated in this release and prepend with a ``+`` sign.
+
+The libraries prepended with a plus sign were incremented in this version.
+
+.. code-block:: diff
+
+ libethdev.so.3
+ librte_acl.so.2
+ librte_cfgfile.so.2
+ librte_cmdline.so.2
+ librte_distributor.so.1
+ librte_eal.so.2
+ librte_hash.so.2
+ librte_ip_frag.so.1
+ librte_ivshmem.so.1
+ librte_jobstats.so.1
+ librte_kni.so.2
+ librte_kvargs.so.1
+ librte_lpm.so.2
+ librte_mbuf.so.2
+ librte_mempool.so.2
+ librte_meter.so.1
+ librte_pipeline.so.3
+ librte_pmd_bond.so.1
+ librte_pmd_ring.so.2
+ librte_port.so.2
+ librte_power.so.1
+ librte_reorder.so.1
+ librte_ring.so.1
+ librte_sched.so.1
+ librte_table.so.2
+ librte_timer.so.1
+ librte_vhost.so.2
+
+
+Tested Platforms
+----------------
+
+.. This section should contain a list of platforms that were tested with this
+ release.
+
+ The format is:
+
+ #. Platform name.
+
+ - Platform details.
+ - Platform details.
+
+
+Tested NICs
+-----------
+
+.. This section should contain a list of NICs that were tested with this release.
+
+ The format is:
+
+ #. NIC name.
+
+ - NIC details.
+ - NIC details.
--
2.8.1
^ permalink raw reply [relevance 6%]
* [dpdk-dev] [PATCH v9 7/7] tools: query binaries for support information
2016-07-04 1:14 2% ` [dpdk-dev] [PATCH v9 4/7] pmdinfogen: parse driver to generate code to export Thomas Monjalon
@ 2016-07-04 1:14 2% ` Thomas Monjalon
1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-04 1:14 UTC (permalink / raw)
To: Neil Horman; +Cc: dev, Panu Matilainen
From: Neil Horman <nhorman@tuxdriver.com>
This tool searches for the primer string PMD_INFO_STRING= in any ELF binary,
and, if found parses the remainder of the string as a json encoded string,
outputting the results in either a human readable or raw, script parseable
format.
Note that, in the case of dynamically linked applications, pmdinfo.py will
scan for implicitly linked PMDs by searching the specified binaries
.dynamic section for DT_NEEDED entries that contain the substring
librte_pmd. The DT_RUNPATH, LD_LIBRARY_PATH, /usr/lib and /lib are
searched for these libraries, in that order.
If a file is specified with no path, it is assumed to be a PMD DSO, and the
LD_LIBRARY_PATH, /usr/lib[64]/ and /lib[64] is searched for it.
Currently the tool can output data in 3 formats:
a) raw, suitable for scripting, where the raw JSON strings are dumped out
b) table format (default) where hex pci ids are dumped in a table format
c) pretty, where a user supplied pci.ids file is used to print out vendor
and device strings
There is a dependency on pyelftools.
The script is not yet compatible with Python 3.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Remy Horton <remy.horton@intel.com>
---
MAINTAINERS | 1 +
lib/librte_eal/common/eal_common_options.c | 2 +-
mk/rte.sdkinstall.mk | 2 +
tools/dpdk-pmdinfo.py | 628 +++++++++++++++++++++++++++++
4 files changed, 632 insertions(+), 1 deletion(-)
create mode 100755 tools/dpdk-pmdinfo.py
diff --git a/MAINTAINERS b/MAINTAINERS
index 1a8a3b7..1e972f0 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -71,6 +71,7 @@ F: scripts/validate-abi.sh
Driver information
M: Neil Horman <nhorman@tuxdriver.com>
F: buildtools/pmdinfogen/
+F: tools/dpdk-pmdinfo.py
Environment Abstraction Layer
diff --git a/lib/librte_eal/common/eal_common_options.c b/lib/librte_eal/common/eal_common_options.c
index 7e9f7b8..b562c8a 100644
--- a/lib/librte_eal/common/eal_common_options.c
+++ b/lib/librte_eal/common/eal_common_options.c
@@ -115,7 +115,7 @@ TAILQ_HEAD_INITIALIZER(solib_list);
/* Default path of external loadable drivers */
static const char *default_solib_dir = RTE_EAL_PMD_PATH;
-/* Stringified version of default solib path */
+/* Stringified version of default solib path used by dpdk-pmdinfo.py */
static const char dpdk_solib_path[] __attribute__((used)) =
"DPDK_PLUGIN_PATH=" RTE_EAL_PMD_PATH;
diff --git a/mk/rte.sdkinstall.mk b/mk/rte.sdkinstall.mk
index 2b92157..76be308 100644
--- a/mk/rte.sdkinstall.mk
+++ b/mk/rte.sdkinstall.mk
@@ -126,6 +126,8 @@ install-runtime:
$(Q)$(call rte_mkdir, $(DESTDIR)$(sbindir))
$(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk_nic_bind.py, \
$(DESTDIR)$(sbindir)/dpdk_nic_bind)
+ $(Q)$(call rte_symlink, $(DESTDIR)$(datadir)/tools/dpdk-pmdinfo.py, \
+ $(DESTDIR)$(bindir)/dpdk-pmdinfo)
install-kmod:
ifneq ($(wildcard $O/kmod/*),)
diff --git a/tools/dpdk-pmdinfo.py b/tools/dpdk-pmdinfo.py
new file mode 100755
index 0000000..b8a9be2
--- /dev/null
+++ b/tools/dpdk-pmdinfo.py
@@ -0,0 +1,628 @@
+#!/usr/bin/python
+# -------------------------------------------------------------------------
+#
+# Utility to dump PMD_INFO_STRING support from an object file
+#
+# -------------------------------------------------------------------------
+import os
+import sys
+from optparse import OptionParser
+import string
+import json
+
+# For running from development directory. It should take precedence over the
+# installed pyelftools.
+sys.path.insert(0, '.')
+
+
+from elftools import __version__
+from elftools.common.exceptions import ELFError
+from elftools.common.py3compat import (
+ ifilter, byte2int, bytes2str, itervalues, str2bytes)
+from elftools.elf.elffile import ELFFile
+from elftools.elf.dynamic import DynamicSection, DynamicSegment
+from elftools.elf.enums import ENUM_D_TAG
+from elftools.elf.segments import InterpSegment
+from elftools.elf.sections import SymbolTableSection
+from elftools.elf.gnuversions import (
+ GNUVerSymSection, GNUVerDefSection,
+ GNUVerNeedSection,
+)
+from elftools.elf.relocation import RelocationSection
+from elftools.elf.descriptions import (
+ describe_ei_class, describe_ei_data, describe_ei_version,
+ describe_ei_osabi, describe_e_type, describe_e_machine,
+ describe_e_version_numeric, describe_p_type, describe_p_flags,
+ describe_sh_type, describe_sh_flags,
+ describe_symbol_type, describe_symbol_bind, describe_symbol_visibility,
+ describe_symbol_shndx, describe_reloc_type, describe_dyn_tag,
+ describe_ver_flags,
+)
+from elftools.elf.constants import E_FLAGS
+from elftools.dwarf.dwarfinfo import DWARFInfo
+from elftools.dwarf.descriptions import (
+ describe_reg_name, describe_attr_value, set_global_machine_arch,
+ describe_CFI_instructions, describe_CFI_register_rule,
+ describe_CFI_CFA_rule,
+)
+from elftools.dwarf.constants import (
+ DW_LNS_copy, DW_LNS_set_file, DW_LNE_define_file)
+from elftools.dwarf.callframe import CIE, FDE
+
+raw_output = False
+pcidb = None
+
+# ===========================================
+
+
+class Vendor:
+ """
+ Class for vendors. This is the top level class
+ for the devices belong to a specific vendor.
+ self.devices is the device dictionary
+ subdevices are in each device.
+ """
+
+ def __init__(self, vendorStr):
+ """
+ Class initializes with the raw line from pci.ids
+ Parsing takes place inside __init__
+ """
+ self.ID = vendorStr.split()[0]
+ self.name = vendorStr.replace("%s " % self.ID, "").rstrip()
+ self.devices = {}
+
+ def add_device(self, deviceStr):
+ """
+ Adds a device to self.devices
+ takes the raw line from pci.ids
+ """
+ s = deviceStr.strip()
+ devID = s.split()[0]
+ if devID in self.devices:
+ pass
+ else:
+ self.devices[devID] = Device(deviceStr)
+
+ def report(self):
+ print self.ID, self.name
+ for id, dev in self.devices.items():
+ dev.report()
+
+ def find_device(self, devid):
+ # convert to a hex string and remove 0x
+ devid = hex(devid)[2:]
+ try:
+ return self.devices[devid]
+ except:
+ return Device("%s Unknown Device" % devid)
+
+
+class Device:
+
+ def __init__(self, deviceStr):
+ """
+ Class for each device.
+ Each vendor has its own devices dictionary.
+ """
+ s = deviceStr.strip()
+ self.ID = s.split()[0]
+ self.name = s.replace("%s " % self.ID, "")
+ self.subdevices = {}
+
+ def report(self):
+ print "\t%s\t%s" % (self.ID, self.name)
+ for subID, subdev in self.subdevices.items():
+ subdev.report()
+
+ def add_sub_device(self, subDeviceStr):
+ """
+ Adds a subvendor, subdevice to device.
+ Uses raw line from pci.ids
+ """
+ s = subDeviceStr.strip()
+ spl = s.split()
+ subVendorID = spl[0]
+ subDeviceID = spl[1]
+ subDeviceName = s.split(" ")[-1]
+ devID = "%s:%s" % (subVendorID, subDeviceID)
+ self.subdevices[devID] = SubDevice(
+ subVendorID, subDeviceID, subDeviceName)
+
+ def find_subid(self, subven, subdev):
+ subven = hex(subven)[2:]
+ subdev = hex(subdev)[2:]
+ devid = "%s:%s" % (subven, subdev)
+
+ try:
+ return self.subdevices[devid]
+ except:
+ if (subven == "ffff" and subdev == "ffff"):
+ return SubDevice("ffff", "ffff", "(All Subdevices)")
+ else:
+ return SubDevice(subven, subdev, "(Unknown Subdevice)")
+
+
+class SubDevice:
+ """
+ Class for subdevices.
+ """
+
+ def __init__(self, vendor, device, name):
+ """
+ Class initializes with vendorid, deviceid and name
+ """
+ self.vendorID = vendor
+ self.deviceID = device
+ self.name = name
+
+ def report(self):
+ print "\t\t%s\t%s\t%s" % (self.vendorID, self.deviceID, self.name)
+
+
+class PCIIds:
+ """
+ Top class for all pci.ids entries.
+ All queries will be asked to this class.
+ PCIIds.vendors["0e11"].devices["0046"].\
+ subdevices["0e11:4091"].name = "Smart Array 6i"
+ """
+
+ def __init__(self, filename):
+ """
+ Prepares the directories.
+ Checks local data file.
+ Tries to load from local, if not found, downloads from web
+ """
+ self.version = ""
+ self.date = ""
+ self.vendors = {}
+ self.contents = None
+ self.read_local(filename)
+ self.parse()
+
+ def report_vendors(self):
+ """Reports the vendors
+ """
+ for vid, v in self.vendors.items():
+ print v.ID, v.name
+
+ def report(self, vendor=None):
+ """
+ Reports everything for all vendors or a specific vendor
+ PCIIds.report() reports everything
+ PCIIDs.report("0e11") reports only "Compaq Computer Corporation"
+ """
+ if vendor is not None:
+ self.vendors[vendor].report()
+ else:
+ for vID, v in self.vendors.items():
+ v.report()
+
+ def find_vendor(self, vid):
+ # convert vid to a hex string and remove the 0x
+ vid = hex(vid)[2:]
+
+ try:
+ return self.vendors[vid]
+ except:
+ return Vendor("%s Unknown Vendor" % (vid))
+
+ def find_date(self, content):
+ for l in content:
+ if l.find("Date:") > -1:
+ return l.split()[-2].replace("-", "")
+ return None
+
+ def parse(self):
+ if len(self.contents) < 1:
+ print "data/%s-pci.ids not found" % self.date
+ else:
+ vendorID = ""
+ deviceID = ""
+ for l in self.contents:
+ if l[0] == "#":
+ continue
+ elif len(l.strip()) == 0:
+ continue
+ else:
+ if l.find("\t\t") == 0:
+ self.vendors[vendorID].devices[
+ deviceID].add_sub_device(l)
+ elif l.find("\t") == 0:
+ deviceID = l.strip().split()[0]
+ self.vendors[vendorID].add_device(l)
+ else:
+ vendorID = l.split()[0]
+ self.vendors[vendorID] = Vendor(l)
+
+ def read_local(self, filename):
+ """
+ Reads the local file
+ """
+ self.contents = open(filename).readlines()
+ self.date = self.find_date(self.contents)
+
+ def load_local(self):
+ """
+ Loads database from local. If there is no file,
+ it creates a new one from web
+ """
+ self.date = idsfile[0].split("/")[1].split("-")[0]
+ self.read_local()
+
+
+# =======================================
+
+def search_file(filename, search_path):
+ """ Given a search path, find file with requested name """
+ for path in string.split(search_path, ":"):
+ candidate = os.path.join(path, filename)
+ if os.path.exists(candidate):
+ return os.path.abspath(candidate)
+ return None
+
+
+class ReadElf(object):
+ """ display_* methods are used to emit output into the output stream
+ """
+
+ def __init__(self, file, output):
+ """ file:
+ stream object with the ELF file to read
+
+ output:
+ output stream to write to
+ """
+ self.elffile = ELFFile(file)
+ self.output = output
+
+ # Lazily initialized if a debug dump is requested
+ self._dwarfinfo = None
+
+ self._versioninfo = None
+
+ def _section_from_spec(self, spec):
+ """ Retrieve a section given a "spec" (either number or name).
+ Return None if no such section exists in the file.
+ """
+ try:
+ num = int(spec)
+ if num < self.elffile.num_sections():
+ return self.elffile.get_section(num)
+ else:
+ return None
+ except ValueError:
+ # Not a number. Must be a name then
+ return self.elffile.get_section_by_name(str2bytes(spec))
+
+ def pretty_print_pmdinfo(self, pmdinfo):
+ global pcidb
+
+ for i in pmdinfo["pci_ids"]:
+ vendor = pcidb.find_vendor(i[0])
+ device = vendor.find_device(i[1])
+ subdev = device.find_subid(i[2], i[3])
+ print("%s (%s) : %s (%s) %s" %
+ (vendor.name, vendor.ID, device.name,
+ device.ID, subdev.name))
+
+ def parse_pmd_info_string(self, mystring):
+ global raw_output
+ global pcidb
+
+ optional_pmd_info = [{'id': 'params', 'tag': 'PMD PARAMETERS'}]
+
+ i = mystring.index("=")
+ mystring = mystring[i + 2:]
+ pmdinfo = json.loads(mystring)
+
+ if raw_output:
+ print(pmdinfo)
+ return
+
+ print("PMD NAME: " + pmdinfo["name"])
+ for i in optional_pmd_info:
+ try:
+ print("%s: %s" % (i['tag'], pmdinfo[i['id']]))
+ except KeyError as e:
+ continue
+
+ if (len(pmdinfo["pci_ids"]) != 0):
+ print("PMD HW SUPPORT:")
+ if pcidb is not None:
+ self.pretty_print_pmdinfo(pmdinfo)
+ else:
+ print("VENDOR\t DEVICE\t SUBVENDOR\t SUBDEVICE")
+ for i in pmdinfo["pci_ids"]:
+ print("0x%04x\t 0x%04x\t 0x%04x\t\t 0x%04x" %
+ (i[0], i[1], i[2], i[3]))
+
+ print("")
+
+ def display_pmd_info_strings(self, section_spec):
+ """ Display a strings dump of a section. section_spec is either a
+ section number or a name.
+ """
+ section = self._section_from_spec(section_spec)
+ if section is None:
+ return
+
+ data = section.data()
+ dataptr = 0
+
+ while dataptr < len(data):
+ while (dataptr < len(data) and
+ not (32 <= byte2int(data[dataptr]) <= 127)):
+ dataptr += 1
+
+ if dataptr >= len(data):
+ break
+
+ endptr = dataptr
+ while endptr < len(data) and byte2int(data[endptr]) != 0:
+ endptr += 1
+
+ mystring = bytes2str(data[dataptr:endptr])
+ rc = mystring.find("PMD_INFO_STRING")
+ if (rc != -1):
+ self.parse_pmd_info_string(mystring)
+
+ dataptr = endptr
+
+ def find_librte_eal(self, section):
+ for tag in section.iter_tags():
+ if tag.entry.d_tag == 'DT_NEEDED':
+ if "librte_eal" in tag.needed:
+ return tag.needed
+ return None
+
+ def search_for_autoload_path(self):
+ scanelf = self
+ scanfile = None
+ library = None
+
+ section = self._section_from_spec(".dynamic")
+ try:
+ eallib = self.find_librte_eal(section)
+ if eallib is not None:
+ ldlibpath = os.environ.get('LD_LIBRARY_PATH')
+ if ldlibpath is None:
+ ldlibpath = ""
+ dtr = self.get_dt_runpath(section)
+ library = search_file(eallib,
+ dtr + ":" + ldlibpath +
+ ":/usr/lib64:/lib64:/usr/lib:/lib")
+ if library is None:
+ return (None, None)
+ if raw_output is False:
+ print("Scanning for autoload path in %s" % library)
+ scanfile = open(library, 'rb')
+ scanelf = ReadElf(scanfile, sys.stdout)
+ except AttributeError:
+ # Not a dynamic binary
+ pass
+ except ELFError:
+ scanfile.close()
+ return (None, None)
+
+ section = scanelf._section_from_spec(".rodata")
+ if section is None:
+ if scanfile is not None:
+ scanfile.close()
+ return (None, None)
+
+ data = section.data()
+ dataptr = 0
+
+ while dataptr < len(data):
+ while (dataptr < len(data) and
+ not (32 <= byte2int(data[dataptr]) <= 127)):
+ dataptr += 1
+
+ if dataptr >= len(data):
+ break
+
+ endptr = dataptr
+ while endptr < len(data) and byte2int(data[endptr]) != 0:
+ endptr += 1
+
+ mystring = bytes2str(data[dataptr:endptr])
+ rc = mystring.find("DPDK_PLUGIN_PATH")
+ if (rc != -1):
+ rc = mystring.find("=")
+ return (mystring[rc + 1:], library)
+
+ dataptr = endptr
+ if scanfile is not None:
+ scanfile.close()
+ return (None, None)
+
+ def get_dt_runpath(self, dynsec):
+ for tag in dynsec.iter_tags():
+ if tag.entry.d_tag == 'DT_RUNPATH':
+ return tag.runpath
+ return ""
+
+ def process_dt_needed_entries(self):
+ """ Look to see if there are any DT_NEEDED entries in the binary
+ And process those if there are
+ """
+ global raw_output
+ runpath = ""
+ ldlibpath = os.environ.get('LD_LIBRARY_PATH')
+ if ldlibpath is None:
+ ldlibpath = ""
+
+ dynsec = self._section_from_spec(".dynamic")
+ try:
+ runpath = self.get_dt_runpath(dynsec)
+ except AttributeError:
+ # dynsec is None, just return
+ return
+
+ for tag in dynsec.iter_tags():
+ if tag.entry.d_tag == 'DT_NEEDED':
+ rc = tag.needed.find("librte_pmd")
+ if (rc != -1):
+ library = search_file(tag.needed,
+ runpath + ":" + ldlibpath +
+ ":/usr/lib64:/lib64:/usr/lib:/lib")
+ if library is not None:
+ if raw_output is False:
+ print("Scanning %s for pmd information" % library)
+ with open(library, 'rb') as file:
+ try:
+ libelf = ReadElf(file, sys.stdout)
+ except ELFError as e:
+ print("%s is no an ELF file" % library)
+ continue
+ libelf.process_dt_needed_entries()
+ libelf.display_pmd_info_strings(".rodata")
+ file.close()
+
+
+def scan_autoload_path(autoload_path):
+ global raw_output
+
+ if os.path.exists(autoload_path) is False:
+ return
+
+ try:
+ dirs = os.listdir(autoload_path)
+ except OSError as e:
+ # Couldn't read the directory, give up
+ return
+
+ for d in dirs:
+ dpath = os.path.join(autoload_path, d)
+ if os.path.isdir(dpath):
+ scan_autoload_path(dpath)
+ if os.path.isfile(dpath):
+ try:
+ file = open(dpath, 'rb')
+ readelf = ReadElf(file, sys.stdout)
+ except ELFError as e:
+ # this is likely not an elf file, skip it
+ continue
+ except IOError as e:
+ # No permission to read the file, skip it
+ continue
+
+ if raw_output is False:
+ print("Hw Support for library %s" % d)
+ readelf.display_pmd_info_strings(".rodata")
+ file.close()
+
+
+def scan_for_autoload_pmds(dpdk_path):
+ """
+ search the specified application or path for a pmd autoload path
+ then scan said path for pmds and report hw support
+ """
+ global raw_output
+
+ if (os.path.isfile(dpdk_path) is False):
+ if raw_output is False:
+ print("Must specify a file name")
+ return
+
+ file = open(dpdk_path, 'rb')
+ try:
+ readelf = ReadElf(file, sys.stdout)
+ except ElfError as e:
+ if raw_output is False:
+ print("Unable to parse %s" % file)
+ return
+
+ (autoload_path, scannedfile) = readelf.search_for_autoload_path()
+ if (autoload_path is None or autoload_path is ""):
+ if (raw_output is False):
+ print("No autoload path configured in %s" % dpdk_path)
+ return
+ if (raw_output is False):
+ if (scannedfile is None):
+ scannedfile = dpdk_path
+ print("Found autoload path %s in %s" % (autoload_path, scannedfile))
+
+ file.close()
+ if (raw_output is False):
+ print("Discovered Autoload HW Support:")
+ scan_autoload_path(autoload_path)
+ return
+
+
+def main(stream=None):
+ global raw_output
+ global pcidb
+
+ optparser = OptionParser(
+ usage='usage: %prog [-hrtp] [-d <pci id file] <elf-file>',
+ description="Dump pmd hardware support info",
+ add_help_option=True,
+ prog='pmdinfo.py')
+ optparser.add_option('-r', '--raw',
+ action='store_true', dest='raw_output',
+ help='Dump raw json strings')
+ optparser.add_option("-d", "--pcidb", dest="pcifile",
+ help="specify a pci database "
+ "to get vendor names from",
+ default="/usr/share/hwdata/pci.ids", metavar="FILE")
+ optparser.add_option("-t", "--table", dest="tblout",
+ help="output information on hw support as a hex table",
+ action='store_true')
+ optparser.add_option("-p", "--plugindir", dest="pdir",
+ help="scan dpdk for autoload plugins",
+ action='store_true')
+
+ options, args = optparser.parse_args()
+
+ if options.raw_output:
+ raw_output = True
+
+ if options.pcifile:
+ pcidb = PCIIds(options.pcifile)
+ if pcidb is None:
+ print("Pci DB file not found")
+ exit(1)
+
+ if options.tblout:
+ options.pcifile = None
+ pcidb = None
+
+ if (len(args) == 0):
+ optparser.print_usage()
+ exit(1)
+
+ if options.pdir is True:
+ exit(scan_for_autoload_pmds(args[0]))
+
+ ldlibpath = os.environ.get('LD_LIBRARY_PATH')
+ if (ldlibpath is None):
+ ldlibpath = ""
+
+ if (os.path.exists(args[0]) is True):
+ myelffile = args[0]
+ else:
+ myelffile = search_file(
+ args[0], ldlibpath + ":/usr/lib64:/lib64:/usr/lib:/lib")
+
+ if (myelffile is None):
+ print("File not found")
+ sys.exit(1)
+
+ with open(myelffile, 'rb') as file:
+ try:
+ readelf = ReadElf(file, sys.stdout)
+ readelf.process_dt_needed_entries()
+ readelf.display_pmd_info_strings(".rodata")
+ sys.exit(0)
+
+ except ELFError as ex:
+ sys.stderr.write('ELF error: %s\n' % ex)
+ sys.exit(1)
+
+
+# -------------------------------------------------------------------------
+if __name__ == '__main__':
+ main()
--
2.7.0
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH v9 4/7] pmdinfogen: parse driver to generate code to export
@ 2016-07-04 1:14 2% ` Thomas Monjalon
2016-07-04 1:14 2% ` [dpdk-dev] [PATCH v9 7/7] tools: query binaries for support information Thomas Monjalon
1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-07-04 1:14 UTC (permalink / raw)
To: Neil Horman; +Cc: dev, Panu Matilainen
From: Neil Horman <nhorman@tuxdriver.com>
dpdk-pmdinfogen is a tool used to parse object files and build JSON
strings for use in later determining hardware support in a DSO or
application binary.
dpdk-pmdinfogen looks for the non-exported symbol names rte_pmd_name<n>
(where n is a integer counter) and <name>_pci_table_export.
It records the name of each of these tuples, using the later to find
the symbolic name of the PCI table for physical devices that the object
supports. With this information, it outputs a C file with a single line
of the form:
static char *<name>_pmd_info[] __attribute__((used)) = " \
PMD_INFO_STRING=<json_string>";
Where <name> is the arbitrary name of the PMD, and <json_string> is the
JSON encoded string that hold relevant PMD information, including the PMD
name, type and optional array of PCI device/vendor IDs that the driver
supports.
This C file is suitable for compiling to object code, then relocatably
linking into the parent file from which the C was generated. This creates
an entry in the string table of the object that can inform a later tool
about hardware support.
Note 1: When installed as part of a SDK package, dpdk-pmdinfogen should
be built for the SDK target. It is not handled currently.
Note 2: Some generated files are not cleaned by "make clean".
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Panu Matilainen <pmatilai@redhat.com>
Acked-by: Remy Horton <remy.horton@intel.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
GNUmakefile | 2 +-
MAINTAINERS | 4 +
GNUmakefile => buildtools/Makefile | 17 +-
GNUmakefile => buildtools/pmdinfogen/Makefile | 21 +-
buildtools/pmdinfogen/pmdinfogen.c | 451 +++++++++++++++++++++++++
buildtools/pmdinfogen/pmdinfogen.h | 99 ++++++
doc/guides/prog_guide/dev_kit_build_system.rst | 15 +-
mk/rte.sdkbuild.mk | 2 +-
mk/rte.sdkinstall.mk | 3 +
9 files changed, 587 insertions(+), 27 deletions(-)
copy GNUmakefile => buildtools/Makefile (87%)
copy GNUmakefile => buildtools/pmdinfogen/Makefile (84%)
create mode 100644 buildtools/pmdinfogen/pmdinfogen.c
create mode 100644 buildtools/pmdinfogen/pmdinfogen.h
diff --git a/GNUmakefile b/GNUmakefile
index b59e4b6..00fe0db 100644
--- a/GNUmakefile
+++ b/GNUmakefile
@@ -40,6 +40,6 @@ export RTE_SDK
# directory list
#
-ROOTDIRS-y := lib drivers app
+ROOTDIRS-y := buildtools lib drivers app
include $(RTE_SDK)/mk/rte.sdkroot.mk
diff --git a/MAINTAINERS b/MAINTAINERS
index a59191e..1a8a3b7 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -68,6 +68,10 @@ F: lib/librte_compat/
F: doc/guides/rel_notes/deprecation.rst
F: scripts/validate-abi.sh
+Driver information
+M: Neil Horman <nhorman@tuxdriver.com>
+F: buildtools/pmdinfogen/
+
Environment Abstraction Layer
-----------------------------
diff --git a/GNUmakefile b/buildtools/Makefile
similarity index 87%
copy from GNUmakefile
copy to buildtools/Makefile
index b59e4b6..35a42ff 100644
--- a/GNUmakefile
+++ b/buildtools/Makefile
@@ -1,6 +1,6 @@
# BSD LICENSE
#
-# Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+# Copyright(c) 2016 Neil Horman. All rights reserved.
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
@@ -29,17 +29,8 @@
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-#
-# Head Makefile for compiling rte SDK
-#
-
-RTE_SDK := $(CURDIR)
-export RTE_SDK
-
-#
-# directory list
-#
+include $(RTE_SDK)/mk/rte.vars.mk
-ROOTDIRS-y := lib drivers app
+DIRS-y += pmdinfogen
-include $(RTE_SDK)/mk/rte.sdkroot.mk
+include $(RTE_SDK)/mk/rte.subdir.mk
diff --git a/GNUmakefile b/buildtools/pmdinfogen/Makefile
similarity index 84%
copy from GNUmakefile
copy to buildtools/pmdinfogen/Makefile
index b59e4b6..327927e 100644
--- a/GNUmakefile
+++ b/buildtools/pmdinfogen/Makefile
@@ -1,6 +1,6 @@
# BSD LICENSE
#
-# Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+# Copyright(c) 2016 Neil Horman. All rights reserved.
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
@@ -29,17 +29,16 @@
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-#
-# Head Makefile for compiling rte SDK
-#
+include $(RTE_SDK)/mk/rte.vars.mk
-RTE_SDK := $(CURDIR)
-export RTE_SDK
+HOSTAPP_DIR = buildtools
+HOSTAPP = dpdk-pmdinfogen
-#
-# directory list
-#
+SRCS-y += pmdinfogen.c
+
+HOST_CFLAGS += $(WERROR_FLAGS) -g
+HOST_CFLAGS += -I$(RTE_OUTPUT)/include
-ROOTDIRS-y := lib drivers app
+DEPDIRS-y += lib/librte_eal
-include $(RTE_SDK)/mk/rte.sdkroot.mk
+include $(RTE_SDK)/mk/rte.hostapp.mk
diff --git a/buildtools/pmdinfogen/pmdinfogen.c b/buildtools/pmdinfogen/pmdinfogen.c
new file mode 100644
index 0000000..101bce1
--- /dev/null
+++ b/buildtools/pmdinfogen/pmdinfogen.c
@@ -0,0 +1,451 @@
+/* Postprocess pmd object files to export hw support
+ *
+ * Copyright 2016 Neil Horman <nhorman@tuxdriver.com>
+ * Based in part on modpost.c from the linux kernel
+ *
+ * This software may be used and distributed according to the terms
+ * of the GNU General Public License V2, incorporated herein by reference.
+ */
+
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <ctype.h>
+#include <string.h>
+#include <limits.h>
+#include <stdbool.h>
+#include <errno.h>
+#include <rte_common.h>
+#include "pmdinfogen.h"
+
+#ifdef RTE_ARCH_64
+#define ADDR_SIZE 64
+#else
+#define ADDR_SIZE 32
+#endif
+
+
+static void *
+grab_file(const char *filename, unsigned long *size)
+{
+ struct stat st;
+ void *map = MAP_FAILED;
+ int fd;
+
+ fd = open(filename, O_RDONLY);
+ if (fd < 0)
+ return NULL;
+ if (fstat(fd, &st))
+ goto failed;
+
+ *size = st.st_size;
+ map = mmap(NULL, *size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
+
+failed:
+ close(fd);
+ if (map == MAP_FAILED)
+ return NULL;
+ return map;
+}
+
+/*
+ * Return a copy of the next line in a mmap'ed file.
+ * spaces in the beginning of the line is trimmed away.
+ * Return a pointer to a static buffer.
+ */
+static void
+release_file(void *file, unsigned long size)
+{
+ munmap(file, size);
+}
+
+/*
+ * Note, it seems odd that we have both a CONVERT_NATIVE and a TO_NATIVE macro
+ * below. We do this because the values passed to TO_NATIVE may themselves be
+ * macros and need both macros here to get expanded. Specifically its the width
+ * variable we are concerned with, because it needs to get expanded prior to
+ * string concatenation
+ */
+#define CONVERT_NATIVE(fend, width, x) ({ \
+typeof(x) ___x; \
+if ((fend) == ELFDATA2LSB) \
+ ___x = rte_le_to_cpu_##width(x); \
+else \
+ ___x = rte_be_to_cpu_##width(x); \
+ ___x; \
+})
+
+#define TO_NATIVE(fend, width, x) CONVERT_NATIVE(fend, width, x)
+
+static int
+parse_elf(struct elf_info *info, const char *filename)
+{
+ unsigned int i;
+ Elf_Ehdr *hdr;
+ Elf_Shdr *sechdrs;
+ Elf_Sym *sym;
+ int endian;
+ unsigned int symtab_idx = ~0U, symtab_shndx_idx = ~0U;
+
+ hdr = grab_file(filename, &info->size);
+ if (!hdr) {
+ perror(filename);
+ return -ENOENT;
+ }
+ info->hdr = hdr;
+ if (info->size < sizeof(*hdr)) {
+ /* file too small, assume this is an empty .o file */
+ return 0;
+ }
+ /* Is this a valid ELF file? */
+ if ((hdr->e_ident[EI_MAG0] != ELFMAG0) ||
+ (hdr->e_ident[EI_MAG1] != ELFMAG1) ||
+ (hdr->e_ident[EI_MAG2] != ELFMAG2) ||
+ (hdr->e_ident[EI_MAG3] != ELFMAG3)) {
+ /* Not an ELF file - silently ignore it */
+ return 0;
+ }
+
+ if (!hdr->e_ident[EI_DATA]) {
+ /* Unknown endian */
+ return 0;
+ }
+
+ endian = hdr->e_ident[EI_DATA];
+
+ /* Fix endianness in ELF header */
+ hdr->e_type = TO_NATIVE(endian, 16, hdr->e_type);
+ hdr->e_machine = TO_NATIVE(endian, 16, hdr->e_machine);
+ hdr->e_version = TO_NATIVE(endian, 32, hdr->e_version);
+ hdr->e_entry = TO_NATIVE(endian, ADDR_SIZE, hdr->e_entry);
+ hdr->e_phoff = TO_NATIVE(endian, ADDR_SIZE, hdr->e_phoff);
+ hdr->e_shoff = TO_NATIVE(endian, ADDR_SIZE, hdr->e_shoff);
+ hdr->e_flags = TO_NATIVE(endian, 32, hdr->e_flags);
+ hdr->e_ehsize = TO_NATIVE(endian, 16, hdr->e_ehsize);
+ hdr->e_phentsize = TO_NATIVE(endian, 16, hdr->e_phentsize);
+ hdr->e_phnum = TO_NATIVE(endian, 16, hdr->e_phnum);
+ hdr->e_shentsize = TO_NATIVE(endian, 16, hdr->e_shentsize);
+ hdr->e_shnum = TO_NATIVE(endian, 16, hdr->e_shnum);
+ hdr->e_shstrndx = TO_NATIVE(endian, 16, hdr->e_shstrndx);
+
+ sechdrs = RTE_PTR_ADD(hdr, hdr->e_shoff);
+ info->sechdrs = sechdrs;
+
+ /* Check if file offset is correct */
+ if (hdr->e_shoff > info->size) {
+ fprintf(stderr, "section header offset=%lu in file '%s' "
+ "is bigger than filesize=%lu\n",
+ (unsigned long)hdr->e_shoff,
+ filename, info->size);
+ return 0;
+ }
+
+ if (hdr->e_shnum == SHN_UNDEF) {
+ /*
+ * There are more than 64k sections,
+ * read count from .sh_size.
+ */
+ info->num_sections = TO_NATIVE(endian, 32, sechdrs[0].sh_size);
+ } else {
+ info->num_sections = hdr->e_shnum;
+ }
+ if (hdr->e_shstrndx == SHN_XINDEX)
+ info->secindex_strings =
+ TO_NATIVE(endian, 32, sechdrs[0].sh_link);
+ else
+ info->secindex_strings = hdr->e_shstrndx;
+
+ /* Fix endianness in section headers */
+ for (i = 0; i < info->num_sections; i++) {
+ sechdrs[i].sh_name =
+ TO_NATIVE(endian, 32, sechdrs[i].sh_name);
+ sechdrs[i].sh_type =
+ TO_NATIVE(endian, 32, sechdrs[i].sh_type);
+ sechdrs[i].sh_flags =
+ TO_NATIVE(endian, 32, sechdrs[i].sh_flags);
+ sechdrs[i].sh_addr =
+ TO_NATIVE(endian, ADDR_SIZE, sechdrs[i].sh_addr);
+ sechdrs[i].sh_offset =
+ TO_NATIVE(endian, ADDR_SIZE, sechdrs[i].sh_offset);
+ sechdrs[i].sh_size =
+ TO_NATIVE(endian, 32, sechdrs[i].sh_size);
+ sechdrs[i].sh_link =
+ TO_NATIVE(endian, 32, sechdrs[i].sh_link);
+ sechdrs[i].sh_info =
+ TO_NATIVE(endian, 32, sechdrs[i].sh_info);
+ sechdrs[i].sh_addralign =
+ TO_NATIVE(endian, ADDR_SIZE, sechdrs[i].sh_addralign);
+ sechdrs[i].sh_entsize =
+ TO_NATIVE(endian, ADDR_SIZE, sechdrs[i].sh_entsize);
+ }
+ /* Find symbol table. */
+ for (i = 1; i < info->num_sections; i++) {
+ int nobits = sechdrs[i].sh_type == SHT_NOBITS;
+
+ if (!nobits && sechdrs[i].sh_offset > info->size) {
+ fprintf(stderr, "%s is truncated. "
+ "sechdrs[i].sh_offset=%lu > sizeof(*hrd)=%zu\n",
+ filename, (unsigned long)sechdrs[i].sh_offset,
+ sizeof(*hdr));
+ return 0;
+ }
+
+ if (sechdrs[i].sh_type == SHT_SYMTAB) {
+ unsigned int sh_link_idx;
+ symtab_idx = i;
+ info->symtab_start = RTE_PTR_ADD(hdr,
+ sechdrs[i].sh_offset);
+ info->symtab_stop = RTE_PTR_ADD(hdr,
+ sechdrs[i].sh_offset + sechdrs[i].sh_size);
+ sh_link_idx = sechdrs[i].sh_link;
+ info->strtab = RTE_PTR_ADD(hdr,
+ sechdrs[sh_link_idx].sh_offset);
+ }
+
+ /* 32bit section no. table? ("more than 64k sections") */
+ if (sechdrs[i].sh_type == SHT_SYMTAB_SHNDX) {
+ symtab_shndx_idx = i;
+ info->symtab_shndx_start = RTE_PTR_ADD(hdr,
+ sechdrs[i].sh_offset);
+ info->symtab_shndx_stop = RTE_PTR_ADD(hdr,
+ sechdrs[i].sh_offset + sechdrs[i].sh_size);
+ }
+ }
+ if (!info->symtab_start)
+ fprintf(stderr, "%s has no symtab?\n", filename);
+
+ /* Fix endianness in symbols */
+ for (sym = info->symtab_start; sym < info->symtab_stop; sym++) {
+ sym->st_shndx = TO_NATIVE(endian, 16, sym->st_shndx);
+ sym->st_name = TO_NATIVE(endian, 32, sym->st_name);
+ sym->st_value = TO_NATIVE(endian, ADDR_SIZE, sym->st_value);
+ sym->st_size = TO_NATIVE(endian, ADDR_SIZE, sym->st_size);
+ }
+
+ if (symtab_shndx_idx != ~0U) {
+ Elf32_Word *p;
+ if (symtab_idx != sechdrs[symtab_shndx_idx].sh_link)
+ fprintf(stderr,
+ "%s: SYMTAB_SHNDX has bad sh_link: %u!=%u\n",
+ filename, sechdrs[symtab_shndx_idx].sh_link,
+ symtab_idx);
+ /* Fix endianness */
+ for (p = info->symtab_shndx_start; p < info->symtab_shndx_stop; p++)
+ *p = TO_NATIVE(endian, 32, *p);
+ }
+
+ return 1;
+}
+
+static void
+parse_elf_finish(struct elf_info *info)
+{
+ struct pmd_driver *tmp, *idx = info->drivers;
+ release_file(info->hdr, info->size);
+ while (idx) {
+ tmp = idx->next;
+ free(idx);
+ idx = tmp;
+ }
+}
+
+static const char *
+get_sym_name(struct elf_info *elf, Elf_Sym *sym)
+{
+ if (sym)
+ return elf->strtab + sym->st_name;
+ else
+ return "(unknown)";
+}
+
+static void *
+get_sym_value(struct elf_info *info, const Elf_Sym *sym)
+{
+ return RTE_PTR_ADD(info->hdr,
+ info->sechdrs[sym->st_shndx].sh_offset + sym->st_value);
+}
+
+static Elf_Sym *
+find_sym_in_symtab(struct elf_info *info, const char *name, Elf_Sym *last)
+{
+ Elf_Sym *idx;
+ if (last)
+ idx = last+1;
+ else
+ idx = info->symtab_start;
+
+ for (; idx < info->symtab_stop; idx++) {
+ const char *n = get_sym_name(info, idx);
+ if (!strncmp(n, name, strlen(name)))
+ return idx;
+ }
+ return NULL;
+}
+
+struct opt_tag {
+ const char *suffix;
+ const char *json_id;
+};
+
+static const struct opt_tag opt_tags[] = {
+ {"_param_string_export", "params"},
+};
+
+static int
+complete_pmd_entry(struct elf_info *info, struct pmd_driver *drv)
+{
+ const char *tname;
+ int i;
+ char tmpsymname[128];
+ Elf_Sym *tmpsym;
+
+ drv->name = get_sym_value(info, drv->name_sym);
+
+ for (i = 0; i < PMD_OPT_MAX; i++) {
+ memset(tmpsymname, 0, 128);
+ sprintf(tmpsymname, "__%s%s", drv->name, opt_tags[i].suffix);
+ tmpsym = find_sym_in_symtab(info, tmpsymname, NULL);
+ if (!tmpsym)
+ continue;
+ drv->opt_vals[i] = get_sym_value(info, tmpsym);
+ }
+
+ memset(tmpsymname, 0, 128);
+ sprintf(tmpsymname, "__%s_pci_table_export", drv->name);
+
+ tmpsym = find_sym_in_symtab(info, tmpsymname, NULL);
+
+ /*
+ * If this returns NULL, then this is a PMD_VDEV, because
+ * it has no pci table reference
+ */
+ if (!tmpsym) {
+ drv->pci_tbl = NULL;
+ return 0;
+ }
+
+ tname = get_sym_value(info, tmpsym);
+ tmpsym = find_sym_in_symtab(info, tname, NULL);
+ if (!tmpsym) {
+ fprintf(stderr, "No symbol %s\n", tname);
+ return -ENOENT;
+ }
+
+ drv->pci_tbl = (struct rte_pci_id *)get_sym_value(info, tmpsym);
+ if (!drv->pci_tbl) {
+ fprintf(stderr, "Failed to get PCI table %s\n", tname);
+ return -ENOENT;
+ }
+
+ return 0;
+}
+
+static int
+locate_pmd_entries(struct elf_info *info)
+{
+ Elf_Sym *last = NULL;
+ struct pmd_driver *new;
+
+ info->drivers = NULL;
+
+ do {
+ new = calloc(sizeof(struct pmd_driver), 1);
+ new->name_sym = find_sym_in_symtab(info, "rte_pmd_name", last);
+ last = new->name_sym;
+ if (!new->name_sym)
+ free(new);
+ else {
+ if (complete_pmd_entry(info, new)) {
+ fprintf(stderr, "Failed to complete pmd entry\n");
+ free(new);
+ return -ENOENT;
+ } else {
+ new->next = info->drivers;
+ info->drivers = new;
+ }
+ }
+ } while (last);
+
+ return 0;
+}
+
+static void
+output_pmd_info_string(struct elf_info *info, char *outfile)
+{
+ FILE *ofd;
+ struct pmd_driver *drv;
+ struct rte_pci_id *pci_ids;
+ int idx = 0;
+
+ ofd = fopen(outfile, "w+");
+ if (!ofd) {
+ fprintf(stderr, "Unable to open output file\n");
+ return;
+ }
+
+ drv = info->drivers;
+
+ while (drv) {
+ fprintf(ofd, "const char %s_pmd_info[] __attribute__((used)) = "
+ "\"PMD_INFO_STRING= {",
+ drv->name);
+ fprintf(ofd, "\\\"name\\\" : \\\"%s\\\", ", drv->name);
+
+ for (idx = 0; idx < PMD_OPT_MAX; idx++) {
+ if (drv->opt_vals[idx])
+ fprintf(ofd, "\\\"%s\\\" : \\\"%s\\\", ",
+ opt_tags[idx].json_id,
+ drv->opt_vals[idx]);
+ }
+
+ pci_ids = drv->pci_tbl;
+ fprintf(ofd, "\\\"pci_ids\\\" : [");
+
+ while (pci_ids && pci_ids->device_id) {
+ fprintf(ofd, "[%d, %d, %d, %d]",
+ pci_ids->vendor_id, pci_ids->device_id,
+ pci_ids->subsystem_vendor_id,
+ pci_ids->subsystem_device_id);
+ pci_ids++;
+ if (pci_ids->device_id)
+ fprintf(ofd, ",");
+ else
+ fprintf(ofd, " ");
+ }
+ fprintf(ofd, "]}\";");
+ drv = drv->next;
+ }
+
+ fclose(ofd);
+}
+
+int main(int argc, char **argv)
+{
+ struct elf_info info;
+ int rc;
+
+ if (argc < 3) {
+ fprintf(stderr,
+ "usage: dpdk-pmdinfogen <object file> <c output file>\n");
+ exit(127);
+ }
+
+ rc = parse_elf(&info, argv[1]);
+ if (rc < 0)
+ exit(-rc);
+
+ rc = locate_pmd_entries(&info);
+ if (rc < 0)
+ goto error;
+
+ if (info.drivers) {
+ output_pmd_info_string(&info, argv[2]);
+ rc = 0;
+ } else {
+ rc = -1;
+ fprintf(stderr, "No drivers registered\n");
+ }
+
+error:
+ parse_elf_finish(&info);
+ exit(-rc);
+}
diff --git a/buildtools/pmdinfogen/pmdinfogen.h b/buildtools/pmdinfogen/pmdinfogen.h
new file mode 100644
index 0000000..7e57702
--- /dev/null
+++ b/buildtools/pmdinfogen/pmdinfogen.h
@@ -0,0 +1,99 @@
+/* Postprocess pmd object files to export hw support
+ *
+ * Copyright 2016 Neil Horman <nhorman@tuxdriver.com>
+ * Based in part on modpost.c from the linux kernel
+ *
+ * This software may be used and distributed according to the terms
+ * of the GNU General Public License V2, incorporated herein by reference.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <elf.h>
+#include <rte_config.h>
+#include <rte_pci.h>
+#include <rte_byteorder.h>
+
+/* On BSD-alike OSes elf.h defines these according to host's word size */
+#undef ELF_ST_BIND
+#undef ELF_ST_TYPE
+#undef ELF_R_SYM
+#undef ELF_R_TYPE
+
+/*
+ * Define ELF64_* to ELF_*, the latter being defined in both 32 and 64 bit
+ * flavors in elf.h. This makes our code a bit more generic between arches
+ * and allows us to support 32 bit code in the future should we ever want to
+ */
+#ifdef RTE_ARCH_64
+#define Elf_Ehdr Elf64_Ehdr
+#define Elf_Shdr Elf64_Shdr
+#define Elf_Sym Elf64_Sym
+#define Elf_Addr Elf64_Addr
+#define Elf_Sword Elf64_Sxword
+#define Elf_Section Elf64_Half
+#define ELF_ST_BIND ELF64_ST_BIND
+#define ELF_ST_TYPE ELF64_ST_TYPE
+
+#define Elf_Rel Elf64_Rel
+#define Elf_Rela Elf64_Rela
+#define ELF_R_SYM ELF64_R_SYM
+#define ELF_R_TYPE ELF64_R_TYPE
+#else
+#define Elf_Ehdr Elf32_Ehdr
+#define Elf_Shdr Elf32_Shdr
+#define Elf_Sym Elf32_Sym
+#define Elf_Addr Elf32_Addr
+#define Elf_Sword Elf32_Sxword
+#define Elf_Section Elf32_Half
+#define ELF_ST_BIND ELF32_ST_BIND
+#define ELF_ST_TYPE ELF32_ST_TYPE
+
+#define Elf_Rel Elf32_Rel
+#define Elf_Rela Elf32_Rela
+#define ELF_R_SYM ELF32_R_SYM
+#define ELF_R_TYPE ELF32_R_TYPE
+#endif
+
+
+enum opt_params {
+ PMD_PARAM_STRING = 0,
+ PMD_OPT_MAX
+};
+
+struct pmd_driver {
+ Elf_Sym *name_sym;
+ const char *name;
+ struct rte_pci_id *pci_tbl;
+ struct pmd_driver *next;
+
+ const char *opt_vals[PMD_OPT_MAX];
+};
+
+struct elf_info {
+ unsigned long size;
+ Elf_Ehdr *hdr;
+ Elf_Shdr *sechdrs;
+ Elf_Sym *symtab_start;
+ Elf_Sym *symtab_stop;
+ char *strtab;
+
+ /* support for 32bit section numbers */
+
+ unsigned int num_sections; /* max_secindex + 1 */
+ unsigned int secindex_strings;
+ /* if Nth symbol table entry has .st_shndx = SHN_XINDEX,
+ * take shndx from symtab_shndx_start[N] instead
+ */
+ Elf32_Word *symtab_shndx_start;
+ Elf32_Word *symtab_shndx_stop;
+
+ struct pmd_driver *drivers;
+};
diff --git a/doc/guides/prog_guide/dev_kit_build_system.rst b/doc/guides/prog_guide/dev_kit_build_system.rst
index dedd18a..fa34fe0 100644
--- a/doc/guides/prog_guide/dev_kit_build_system.rst
+++ b/doc/guides/prog_guide/dev_kit_build_system.rst
@@ -70,7 +70,7 @@ Each build directory contains include files, libraries, and applications:
...
~/DEV/DPDK$ ls i686-native-linuxapp-gcc
- app build hostapp include kmod lib Makefile
+ app build buildtools include kmod lib Makefile
~/DEV/DPDK$ ls i686-native-linuxapp-gcc/app/
@@ -307,6 +307,7 @@ Misc
Internally Generated Build Tools
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+``dpdk-pmdinfogen`` scans an object (.o) file for various well known symbol names.
These well known symbol names are defined by various macros and used to export
important information about hardware support and usage for PMD files. For
instance the macro:
@@ -321,6 +322,18 @@ Creates the following symbol:
static char rte_pmd_name0[] __attribute__((used)) = "<name>";
+Which dpdk-pmdinfogen scans for. Using this information other relevant bits of
+data can be exported from the object file and used to produce a hardware support
+description, that dpdk-pmdinfogen then encodes into a json formatted string in
+the following format:
+
+.. code-block:: C
+
+ static char <name_pmd_string>="PMD_INFO_STRING=\"{'name' : '<name>', ...}\"";
+
+These strings can then be searched for by external tools to determine the
+hardware support of a given library or application.
+
.. _Useful_Variables_Provided_by_the_Build_System:
Useful Variables Provided by the Build System
diff --git a/mk/rte.sdkbuild.mk b/mk/rte.sdkbuild.mk
index f1a163a..5edbf50 100644
--- a/mk/rte.sdkbuild.mk
+++ b/mk/rte.sdkbuild.mk
@@ -63,7 +63,7 @@ build: $(ROOTDIRS-y)
.PHONY: clean
clean: $(CLEANDIRS)
@rm -rf $(RTE_OUTPUT)/include $(RTE_OUTPUT)/app \
- $(RTE_OUTPUT)/lib $(RTE_OUTPUT)/kmod
+ $(RTE_OUTPUT)/lib $(RTE_OUTPUT)/kmod $(RTE_OUTPUT)/buildtools
@[ -d $(RTE_OUTPUT)/include ] || mkdir -p $(RTE_OUTPUT)/include
@$(RTE_SDK)/scripts/gen-config-h.sh $(RTE_OUTPUT)/.config \
> $(RTE_OUTPUT)/include/rte_config.h
diff --git a/mk/rte.sdkinstall.mk b/mk/rte.sdkinstall.mk
index abdab0f..2b92157 100644
--- a/mk/rte.sdkinstall.mk
+++ b/mk/rte.sdkinstall.mk
@@ -141,10 +141,13 @@ install-sdk:
$(Q)$(call rte_mkdir, $(DESTDIR)$(sdkdir))
$(Q)cp -a $(RTE_SDK)/mk $(DESTDIR)$(sdkdir)
$(Q)cp -a $(RTE_SDK)/scripts $(DESTDIR)$(sdkdir)
+ $(Q)cp -a $O/buildtools $(DESTDIR)$(sdkdir)
$(Q)$(call rte_mkdir, $(DESTDIR)$(targetdir))
$(Q)cp -a $O/.config $(DESTDIR)$(targetdir)
$(Q)$(call rte_symlink, $(DESTDIR)$(includedir), $(DESTDIR)$(targetdir)/include)
$(Q)$(call rte_symlink, $(DESTDIR)$(libdir), $(DESTDIR)$(targetdir)/lib)
+ $(Q)$(call rte_symlink, $(DESTDIR)$(sdkdir)/buildtools, \
+ $(DESTDIR)$(targetdir)/buildtools)
install-doc:
ifneq ($(wildcard $O/doc),)
--
2.7.0
^ permalink raw reply [relevance 2%]
* Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
2016-06-30 11:15 7% ` Mcnamara, John
@ 2016-06-30 11:40 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-30 11:40 UTC (permalink / raw)
To: Mcnamara, John
Cc: Panu Matilainen, Yuanhan Liu, dev, Xie, Huawei, Rich Lane,
Tetsuya Mukawa
2016-06-30 11:15, Mcnamara, John:
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Panu Matilainen
> > On 06/30/2016 10:57 AM, Yuanhan Liu wrote:
> > > On Thu, Jun 30, 2016 at 10:39:45AM +0300, Panu Matilainen wrote:
> > >> On 06/07/2016 06:51 AM, Yuanhan Liu wrote:
> > >>> v3: - adapted the new vhost ABI/API changes to tep_term example, to
> > make
> > >>> sure not break build at least.
> > >>> - bumped the ABI version to 3
> > >>>
> > >>> NOTE: I created a branch at dpdk.org [0] for more conveinient testing:
> > >>>
> > >>> [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
> > >>>
> > >>>
> > >>> Every time we introduce a new feature to vhost, we are likely to
> > >>> break ABI. Moreover, some cleanups (such as the one from Ilya to
> > >>> remove vec_buf
> > >> >from vhost_virtqueue struct) also break ABI.
> > >>>
> > >>> This patch set is meant to resolve above issue ultimately, by hiding
> > >>> virtio_net structure (as well as few others) internaly, and export
> > >>> the virtio_net dev strut to applications by a number, vid, like the
> > >>> way kernel exposes an fd to user space.
> > >>>
> > >>> Back to the patch set, the first part of this set makes some changes
> > >>> to vhost example, vhost-pmd and vhost, bit by bit, to remove the
> > >>> dependence to "virtio_net" struct. And then do the final change to
> > >>> make the current APIs to adapt to using "vid".
> > >>>
> > >>> After that, "vrtio_net_device_ops" is the only left open struct that
> > >>> an application can acces, therefore, it's the only place that might
> > >>> introduce potential ABI breakage in future for extension. Hence, I
> > >>> made few more
> > >>> (5) space reservation, to make sure we will not break ABI for a long
> > >>> time, and hopefuly, forever.
> > >>
> > >> Been intending to say this for a while but seems I never actually got
> > >> around to do so:
> > >>
> > >> This is a really fine example of how to refactor an API against
> > >> constant ABI breakages, thank you Yuanhan!
> > >
> > > Panu, thanks!
> > >
> > >> Exported structs are one of the biggest obstacles in keeping a stable
> > >> ABI while adding new features, and while its not always possible to
> > >> hide everything to this extent, the damage (erm,
> > >> exposure) can usually be considerably limited by careful API design.
> > >
> > > Agreed.
> > >
> > >> Since the first and the foremost objection against doing this in the
> > >> DPDK context is always "but performance!", I'm curious as to what
> > >> sort of numbers you're getting with the new API vs the old one? I'm
> > >> really hoping other libraries would follow suit after seeing that its
> > >> possible to provide a future-proof API/ABI without sacrificing
> > >> performance :)
> > >
> > > From my (limited) test, nope, I see no performance drop at all, not
> > > even a little.
> >
> > Awesome!
> >
> > With that, hopefully others will see the light and follow its example.
> > If nothing else, they ought to get a bit envious when you can add features
> > left and right without ever having to wait for API/ABI break periods etc
> > ;)
>
> Agreed. We should be doing more of this type of refactoring work to make the API/ABI less easier to break.
+1
But we must check the possible performance degradation with care :)
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
2016-06-30 9:05 7% ` Panu Matilainen
@ 2016-06-30 11:15 7% ` Mcnamara, John
2016-06-30 11:40 4% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Mcnamara, John @ 2016-06-30 11:15 UTC (permalink / raw)
To: Panu Matilainen, Yuanhan Liu
Cc: dev, Xie, Huawei, Thomas Monjalon, Rich Lane, Tetsuya Mukawa
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Panu Matilainen
> Sent: Thursday, June 30, 2016 10:05 AM
> To: Yuanhan Liu <yuanhan.liu@linux.intel.com>
> Cc: dev@dpdk.org; Xie, Huawei <huawei.xie@intel.com>; Thomas Monjalon
> <thomas.monjalon@6wind.com>; Rich Lane <rich.lane@bigswitch.com>; Tetsuya
> Mukawa <mukawa@igel.co.jp>
> Subject: Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
>
> On 06/30/2016 10:57 AM, Yuanhan Liu wrote:
> > On Thu, Jun 30, 2016 at 10:39:45AM +0300, Panu Matilainen wrote:
> >> On 06/07/2016 06:51 AM, Yuanhan Liu wrote:
> >>> v3: - adapted the new vhost ABI/API changes to tep_term example, to
> make
> >>> sure not break build at least.
> >>> - bumped the ABI version to 3
> >>>
> >>> NOTE: I created a branch at dpdk.org [0] for more conveinient testing:
> >>>
> >>> [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
> >>>
> >>>
> >>> Every time we introduce a new feature to vhost, we are likely to
> >>> break ABI. Moreover, some cleanups (such as the one from Ilya to
> >>> remove vec_buf
> >> >from vhost_virtqueue struct) also break ABI.
> >>>
> >>> This patch set is meant to resolve above issue ultimately, by hiding
> >>> virtio_net structure (as well as few others) internaly, and export
> >>> the virtio_net dev strut to applications by a number, vid, like the
> >>> way kernel exposes an fd to user space.
> >>>
> >>> Back to the patch set, the first part of this set makes some changes
> >>> to vhost example, vhost-pmd and vhost, bit by bit, to remove the
> >>> dependence to "virtio_net" struct. And then do the final change to
> >>> make the current APIs to adapt to using "vid".
> >>>
> >>> After that, "vrtio_net_device_ops" is the only left open struct that
> >>> an application can acces, therefore, it's the only place that might
> >>> introduce potential ABI breakage in future for extension. Hence, I
> >>> made few more
> >>> (5) space reservation, to make sure we will not break ABI for a long
> >>> time, and hopefuly, forever.
> >>
> >> Been intending to say this for a while but seems I never actually got
> >> around to do so:
> >>
> >> This is a really fine example of how to refactor an API against
> >> constant ABI breakages, thank you Yuanhan!
> >
> > Panu, thanks!
> >
> >> Exported structs are one of the biggest obstacles in keeping a stable
> >> ABI while adding new features, and while its not always possible to
> >> hide everything to this extent, the damage (erm,
> >> exposure) can usually be considerably limited by careful API design.
> >
> > Agreed.
> >
> >> Since the first and the foremost objection against doing this in the
> >> DPDK context is always "but performance!", I'm curious as to what
> >> sort of numbers you're getting with the new API vs the old one? I'm
> >> really hoping other libraries would follow suit after seeing that its
> >> possible to provide a future-proof API/ABI without sacrificing
> >> performance :)
> >
> > From my (limited) test, nope, I see no performance drop at all, not
> > even a little.
>
> Awesome!
>
> With that, hopefully others will see the light and follow its example.
> If nothing else, they ought to get a bit envious when you can add features
> left and right without ever having to wait for API/ABI break periods etc
> ;)
Agreed. We should be doing more of this type of refactoring work to make the API/ABI less easier to break.
John
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
2016-06-30 7:57 4% ` Yuanhan Liu
@ 2016-06-30 9:05 7% ` Panu Matilainen
2016-06-30 11:15 7% ` Mcnamara, John
0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-06-30 9:05 UTC (permalink / raw)
To: Yuanhan Liu; +Cc: dev, huawei.xie, Thomas Monjalon, Rich Lane, Tetsuya Mukawa
On 06/30/2016 10:57 AM, Yuanhan Liu wrote:
> On Thu, Jun 30, 2016 at 10:39:45AM +0300, Panu Matilainen wrote:
>> On 06/07/2016 06:51 AM, Yuanhan Liu wrote:
>>> v3: - adapted the new vhost ABI/API changes to tep_term example, to make
>>> sure not break build at least.
>>> - bumped the ABI version to 3
>>>
>>> NOTE: I created a branch at dpdk.org [0] for more conveinient testing:
>>>
>>> [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
>>>
>>>
>>> Every time we introduce a new feature to vhost, we are likely to break
>>> ABI. Moreover, some cleanups (such as the one from Ilya to remove vec_buf
>> >from vhost_virtqueue struct) also break ABI.
>>>
>>> This patch set is meant to resolve above issue ultimately, by hiding
>>> virtio_net structure (as well as few others) internaly, and export the
>>> virtio_net dev strut to applications by a number, vid, like the way
>>> kernel exposes an fd to user space.
>>>
>>> Back to the patch set, the first part of this set makes some changes to
>>> vhost example, vhost-pmd and vhost, bit by bit, to remove the dependence
>>> to "virtio_net" struct. And then do the final change to make the current
>>> APIs to adapt to using "vid".
>>>
>>> After that, "vrtio_net_device_ops" is the only left open struct that an
>>> application can acces, therefore, it's the only place that might introduce
>>> potential ABI breakage in future for extension. Hence, I made few more
>>> (5) space reservation, to make sure we will not break ABI for a long time,
>>> and hopefuly, forever.
>>
>> Been intending to say this for a while but seems I never actually got around
>> to do so:
>>
>> This is a really fine example of how to refactor an API against constant ABI
>> breakages, thank you Yuanhan!
>
> Panu, thanks!
>
>> Exported structs are one of the biggest
>> obstacles in keeping a stable ABI while adding new features, and while its
>> not always possible to hide everything to this extent, the damage (erm,
>> exposure) can usually be considerably limited by careful API design.
>
> Agreed.
>
>> Since the first and the foremost objection against doing this in the DPDK
>> context is always "but performance!", I'm curious as to what sort of numbers
>> you're getting with the new API vs the old one? I'm really hoping other
>> libraries would follow suit after seeing that its possible to provide a
>> future-proof API/ABI without sacrificing performance :)
>
> From my (limited) test, nope, I see no performance drop at all, not even a
> little.
Awesome!
With that, hopefully others will see the light and follow its example.
If nothing else, they ought to get a bit envious when you can add
features left and right without ever having to wait for API/ABI break
periods etc ;)
- Panu -
>
> --yliu
>
^ permalink raw reply [relevance 7%]
* Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
2016-06-30 7:39 9% ` Panu Matilainen
@ 2016-06-30 7:57 4% ` Yuanhan Liu
2016-06-30 9:05 7% ` Panu Matilainen
0 siblings, 1 reply; 200+ results
From: Yuanhan Liu @ 2016-06-30 7:57 UTC (permalink / raw)
To: Panu Matilainen
Cc: dev, huawei.xie, Thomas Monjalon, Rich Lane, Tetsuya Mukawa
On Thu, Jun 30, 2016 at 10:39:45AM +0300, Panu Matilainen wrote:
> On 06/07/2016 06:51 AM, Yuanhan Liu wrote:
> >v3: - adapted the new vhost ABI/API changes to tep_term example, to make
> > sure not break build at least.
> > - bumped the ABI version to 3
> >
> >NOTE: I created a branch at dpdk.org [0] for more conveinient testing:
> >
> > [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
> >
> >
> >Every time we introduce a new feature to vhost, we are likely to break
> >ABI. Moreover, some cleanups (such as the one from Ilya to remove vec_buf
> >from vhost_virtqueue struct) also break ABI.
> >
> >This patch set is meant to resolve above issue ultimately, by hiding
> >virtio_net structure (as well as few others) internaly, and export the
> >virtio_net dev strut to applications by a number, vid, like the way
> >kernel exposes an fd to user space.
> >
> >Back to the patch set, the first part of this set makes some changes to
> >vhost example, vhost-pmd and vhost, bit by bit, to remove the dependence
> >to "virtio_net" struct. And then do the final change to make the current
> >APIs to adapt to using "vid".
> >
> >After that, "vrtio_net_device_ops" is the only left open struct that an
> >application can acces, therefore, it's the only place that might introduce
> >potential ABI breakage in future for extension. Hence, I made few more
> >(5) space reservation, to make sure we will not break ABI for a long time,
> >and hopefuly, forever.
>
> Been intending to say this for a while but seems I never actually got around
> to do so:
>
> This is a really fine example of how to refactor an API against constant ABI
> breakages, thank you Yuanhan!
Panu, thanks!
> Exported structs are one of the biggest
> obstacles in keeping a stable ABI while adding new features, and while its
> not always possible to hide everything to this extent, the damage (erm,
> exposure) can usually be considerably limited by careful API design.
Agreed.
> Since the first and the foremost objection against doing this in the DPDK
> context is always "but performance!", I'm curious as to what sort of numbers
> you're getting with the new API vs the old one? I'm really hoping other
> libraries would follow suit after seeing that its possible to provide a
> future-proof API/ABI without sacrificing performance :)
>From my (limited) test, nope, I see no performance drop at all, not even a
little.
--yliu
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
2016-06-14 12:00 4% ` Yuanhan Liu
@ 2016-06-30 7:39 9% ` Panu Matilainen
2016-06-30 7:57 4% ` Yuanhan Liu
1 sibling, 1 reply; 200+ results
From: Panu Matilainen @ 2016-06-30 7:39 UTC (permalink / raw)
To: Yuanhan Liu, dev
Cc: huawei.xie, Thomas Monjalon, Traynor Kevin, Rich Lane, Tetsuya Mukawa
On 06/07/2016 06:51 AM, Yuanhan Liu wrote:
> v3: - adapted the new vhost ABI/API changes to tep_term example, to make
> sure not break build at least.
> - bumped the ABI version to 3
>
> NOTE: I created a branch at dpdk.org [0] for more conveinient testing:
>
> [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
>
>
> Every time we introduce a new feature to vhost, we are likely to break
> ABI. Moreover, some cleanups (such as the one from Ilya to remove vec_buf
> from vhost_virtqueue struct) also break ABI.
>
> This patch set is meant to resolve above issue ultimately, by hiding
> virtio_net structure (as well as few others) internaly, and export the
> virtio_net dev strut to applications by a number, vid, like the way
> kernel exposes an fd to user space.
>
> Back to the patch set, the first part of this set makes some changes to
> vhost example, vhost-pmd and vhost, bit by bit, to remove the dependence
> to "virtio_net" struct. And then do the final change to make the current
> APIs to adapt to using "vid".
>
> After that, "vrtio_net_device_ops" is the only left open struct that an
> application can acces, therefore, it's the only place that might introduce
> potential ABI breakage in future for extension. Hence, I made few more
> (5) space reservation, to make sure we will not break ABI for a long time,
> and hopefuly, forever.
Been intending to say this for a while but seems I never actually got
around to do so:
This is a really fine example of how to refactor an API against constant
ABI breakages, thank you Yuanhan! Exported structs are one of the
biggest obstacles in keeping a stable ABI while adding new features, and
while its not always possible to hide everything to this extent, the
damage (erm, exposure) can usually be considerably limited by careful
API design.
Since the first and the foremost objection against doing this in the
DPDK context is always "but performance!", I'm curious as to what sort
of numbers you're getting with the new API vs the old one? I'm really
hoping other libraries would follow suit after seeing that its possible
to provide a future-proof API/ABI without sacrificing performance :)
Thanks again,
- Panu -
^ permalink raw reply [relevance 9%]
* Re: [dpdk-dev] [PATCH 0/3] ethdev: add helper functions to get eth_dev and dev private data
@ 2016-06-22 21:47 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-22 21:47 UTC (permalink / raw)
To: Ferruh Yigit; +Cc: dev, Declan Doherty
2016-02-17 14:20, Ferruh Yigit:
> This is to provide abstraction and reduce global variable access.
>
> Global variable rte_eth_devices kept exported to not break ABI.
>
> Bonding driver not selected on purpose, just it seems it is using
> rte_eth_devices heavily.
The struct rte_eth_dev is marked internal.
It is a good goal to remove access to the global array rte_eth_devices,
but the fix must be in the code accessing it only (bonding).
This patchset is rejected.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v5 09/17] crypto: get rid of crypto driver register callback
@ 2016-06-22 13:44 3% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-22 13:44 UTC (permalink / raw)
To: Neil Horman; +Cc: Shreyansh Jain, dev, viktorin
2016-06-22 09:27, Neil Horman:
> > +++ b/lib/librte_cryptodev/rte_cryptodev_version.map
> > - rte_cryptodev_pmd_driver_register;
> NAK, you can't just remove exported symbols without going through the
> deprecation process. Better still would be to only expose it for DPDK_16.04 and
> hide it in the next release
This function is not called by the application.
Thus there is no ABI break.
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v16 1/3] mempool: support mempool handler operations
2016-06-22 9:27 3% ` [dpdk-dev] [PATCH v16 0/3] mempool: add mempool handler feature David Hunt
@ 2016-06-22 9:27 1% ` David Hunt
0 siblings, 0 replies; 200+ results
From: David Hunt @ 2016-06-22 9:27 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain, David Hunt
Until now, the objects stored in a mempool were internally stored in a
ring. This patch introduces the possibility to register external handlers
replacing the ring.
The default behavior remains unchanged, but calling the new function
rte_mempool_set_ops_byname() right after rte_mempool_create_empty() allows
the user to change the handler that will be used when populating
the mempool.
This patch also adds a set of default ops (function callbacks) based
on rte_ring.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---
app/test/test_mempool_perf.c | 1 -
doc/guides/prog_guide/mempool_lib.rst | 32 +++-
doc/guides/rel_notes/deprecation.rst | 9 -
lib/librte_mempool/Makefile | 2 +
lib/librte_mempool/rte_mempool.c | 67 +++-----
lib/librte_mempool/rte_mempool.h | 255 ++++++++++++++++++++++++++---
lib/librte_mempool/rte_mempool_ops.c | 151 +++++++++++++++++
lib/librte_mempool/rte_mempool_ring.c | 161 ++++++++++++++++++
lib/librte_mempool/rte_mempool_version.map | 13 +-
9 files changed, 610 insertions(+), 81 deletions(-)
create mode 100644 lib/librte_mempool/rte_mempool_ops.c
create mode 100644 lib/librte_mempool/rte_mempool_ring.c
diff --git a/app/test/test_mempool_perf.c b/app/test/test_mempool_perf.c
index c5e3576..c5f8455 100644
--- a/app/test/test_mempool_perf.c
+++ b/app/test/test_mempool_perf.c
@@ -161,7 +161,6 @@ per_lcore_mempool_test(__attribute__((unused)) void *arg)
n_get_bulk);
if (unlikely(ret < 0)) {
rte_mempool_dump(stdout, mp);
- rte_ring_dump(stdout, mp->ring);
/* in this case, objects are lost... */
return -1;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index c3afc2e..1943fc4 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -34,7 +34,8 @@ Mempool Library
===============
A memory pool is an allocator of a fixed-sized object.
-In the DPDK, it is identified by name and uses a ring to store free objects.
+In the DPDK, it is identified by name and uses a mempool handler to store free objects.
+The default mempool handler is ring based.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
@@ -127,6 +128,35 @@ The maximum size of the cache is static and is defined at compilation time (CONF
A mempool in Memory with its Associated Ring
+Mempool Handlers
+------------------------
+
+This allows external memory subsystems, such as external hardware memory
+management systems and software based memory allocators, to be used with DPDK.
+
+There are two aspects to a mempool handler.
+
+* Adding the code for your new mempool operations (ops). This is achieved by
+ adding a new mempool ops code, and using the ``REGISTER_MEMPOOL_OPS`` macro.
+
+* Using the new API to call ``rte_mempool_create_empty()`` and
+ ``rte_mempool_set_ops_byname()`` to create a new mempool and specifying which
+ ops to use.
+
+Several different mempool handlers may be used in the same application. A new
+mempool can be created by using the ``rte_mempool_create_empty()`` function,
+then using ``rte_mempool_set_ops_byname()`` to point the mempool to the
+relevant mempool handler callback (ops) structure.
+
+Legacy applications may continue to use the old ``rte_mempool_create()`` API
+call, which uses a ring based mempool handler by default. These applications
+will need to be modified to use a new mempool handler.
+
+For applications that use ``rte_pktmbuf_create()``, there is a config setting
+(``RTE_MBUF_DEFAULT_MEMPOOL_OPS``) that allows the application to make use of
+an alternative mempool handler.
+
+
Use Cases
---------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f75183f..3cbc19e 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -34,15 +34,6 @@ Deprecation Notices
compact API. The ones that remain are backwards compatible and use the
per-lcore default cache if available. This change targets release 16.07.
-* The rte_mempool struct will be changed in 16.07 to facilitate the new
- external mempool manager functionality.
- The ring element will be replaced with a more generic 'pool' opaque pointer
- to allow new mempool handlers to use their own user-defined mempool
- layout. Also newly added to rte_mempool is a handler index.
- The existing API will be backward compatible, but there will be new API
- functions added to facilitate the creation of mempools using an external
- handler. The 16.07 release will contain these changes.
-
* A librte_vhost public structures refactor is planned for DPDK 16.07
that requires both ABI and API change.
The proposed refactor would expose DPDK vhost dev to applications as
diff --git a/lib/librte_mempool/Makefile b/lib/librte_mempool/Makefile
index 43423e0..a4c089e 100644
--- a/lib/librte_mempool/Makefile
+++ b/lib/librte_mempool/Makefile
@@ -42,6 +42,8 @@ LIBABIVER := 2
# all source are stored in SRCS-y
SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ring.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_MEMPOOL)-include := rte_mempool.h
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index af71edd..e6a83d0 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -148,7 +148,7 @@ mempool_add_elem(struct rte_mempool *mp, void *obj, phys_addr_t physaddr)
#endif
/* enqueue in ring */
- rte_ring_sp_enqueue(mp->ring, obj);
+ rte_mempool_ops_enqueue_bulk(mp, &obj, 1);
}
/* call obj_cb() for each mempool element */
@@ -303,40 +303,6 @@ rte_mempool_xmem_usage(__rte_unused void *vaddr, uint32_t elt_num,
return (size_t)paddr_idx << pg_shift;
}
-/* create the internal ring */
-static int
-rte_mempool_ring_create(struct rte_mempool *mp)
-{
- int rg_flags = 0, ret;
- char rg_name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
-
- ret = snprintf(rg_name, sizeof(rg_name),
- RTE_MEMPOOL_MZ_FORMAT, mp->name);
- if (ret < 0 || ret >= (int)sizeof(rg_name))
- return -ENAMETOOLONG;
-
- /* ring flags */
- if (mp->flags & MEMPOOL_F_SP_PUT)
- rg_flags |= RING_F_SP_ENQ;
- if (mp->flags & MEMPOOL_F_SC_GET)
- rg_flags |= RING_F_SC_DEQ;
-
- /* Allocate the ring that will be used to store objects.
- * Ring functions will return appropriate errors if we are
- * running as a secondary process etc., so no checks made
- * in this function for that condition.
- */
- r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
- mp->socket_id, rg_flags);
- if (r == NULL)
- return -rte_errno;
-
- mp->ring = r;
- mp->flags |= MEMPOOL_F_RING_CREATED;
- return 0;
-}
-
/* free a memchunk allocated with rte_memzone_reserve() */
static void
rte_mempool_memchunk_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
@@ -354,7 +320,7 @@ rte_mempool_free_memchunks(struct rte_mempool *mp)
void *elt;
while (!STAILQ_EMPTY(&mp->elt_list)) {
- rte_ring_sc_dequeue(mp->ring, &elt);
+ rte_mempool_ops_dequeue_bulk(mp, &elt, 1);
(void)elt;
STAILQ_REMOVE_HEAD(&mp->elt_list, next);
mp->populated_size--;
@@ -386,10 +352,11 @@ rte_mempool_populate_phys(struct rte_mempool *mp, char *vaddr,
int ret;
/* create the internal ring if not already done */
- if ((mp->flags & MEMPOOL_F_RING_CREATED) == 0) {
- ret = rte_mempool_ring_create(mp);
- if (ret < 0)
+ if ((mp->flags & MEMPOOL_F_POOL_CREATED) == 0) {
+ ret = rte_mempool_ops_alloc(mp);
+ if (ret != 0)
return ret;
+ mp->flags |= MEMPOOL_F_POOL_CREATED;
}
/* mempool is already populated */
@@ -703,7 +670,7 @@ rte_mempool_free(struct rte_mempool *mp)
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_mempool_free_memchunks(mp);
- rte_ring_free(mp->ring);
+ rte_mempool_ops_free(mp);
rte_memzone_free(mp->mz);
}
@@ -815,6 +782,7 @@ rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, 0));
te->data = mp;
+
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_INSERT_TAIL(mempool_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
@@ -844,6 +812,19 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
if (mp == NULL)
return NULL;
+ /*
+ * Since we have 4 combinations of the SP/SC/MP/MC examine the flags to
+ * set the correct index into the table of ops structs.
+ */
+ if (flags & (MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET))
+ rte_mempool_set_ops_byname(mp, "ring_sp_sc", NULL);
+ else if (flags & MEMPOOL_F_SP_PUT)
+ rte_mempool_set_ops_byname(mp, "ring_sp_mc", NULL);
+ else if (flags & MEMPOOL_F_SC_GET)
+ rte_mempool_set_ops_byname(mp, "ring_mp_sc", NULL);
+ else
+ rte_mempool_set_ops_byname(mp, "ring_mp_mc", NULL);
+
/* call the mempool priv initializer */
if (mp_init)
mp_init(mp, mp_init_arg);
@@ -930,7 +911,7 @@ rte_mempool_count(const struct rte_mempool *mp)
unsigned count;
unsigned lcore_id;
- count = rte_ring_count(mp->ring);
+ count = rte_mempool_ops_get_count(mp);
if (mp->cache_size == 0)
return count;
@@ -1119,7 +1100,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
fprintf(f, "mempool <%s>@%p\n", mp->name, mp);
fprintf(f, " flags=%x\n", mp->flags);
- fprintf(f, " ring=<%s>@%p\n", mp->ring->name, mp->ring);
+ fprintf(f, " pool=%p\n", mp->pool_data);
fprintf(f, " phys_addr=0x%" PRIx64 "\n", mp->mz->phys_addr);
fprintf(f, " nb_mem_chunks=%u\n", mp->nb_mem_chunks);
fprintf(f, " size=%"PRIu32"\n", mp->size);
@@ -1140,7 +1121,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
}
cache_count = rte_mempool_dump_cache(f, mp);
- common_count = rte_ring_count(mp->ring);
+ common_count = rte_mempool_ops_get_count(mp);
if ((cache_count + common_count) > mp->size)
common_count = mp->size - cache_count;
fprintf(f, " common_pool_count=%u\n", common_count);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 60339bd..0a1777c 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -67,6 +67,7 @@
#include <inttypes.h>
#include <sys/queue.h>
+#include <rte_spinlock.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_lcore.h>
@@ -203,10 +204,14 @@ struct rte_mempool_memhdr {
*/
struct rte_mempool {
char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
- struct rte_ring *ring; /**< Ring to store objects. */
- const struct rte_memzone *mz; /**< Memzone where pool is allocated */
+ union {
+ void *pool_data; /**< Ring or pool to store objects. */
+ uint64_t pool_id; /**< External mempool identifier. */
+ };
+ void *pool_config; /**< optional args for ops alloc. */
+ const struct rte_memzone *mz; /**< Memzone where pool is alloc'd. */
int flags; /**< Flags of the mempool. */
- int socket_id; /**< Socket id passed at mempool creation. */
+ int socket_id; /**< Socket id passed at create. */
uint32_t size; /**< Max size of the mempool. */
uint32_t cache_size; /**< Size of per-lcore local cache. */
uint32_t cache_flushthresh;
@@ -217,6 +222,14 @@ struct rte_mempool {
uint32_t trailer_size; /**< Size of trailer (after elt). */
unsigned private_data_size; /**< Size of private data. */
+ /**
+ * Index into rte_mempool_ops_table array of mempool ops
+ * structs, which contain callback function pointers.
+ * We're using an index here rather than pointers to the callbacks
+ * to facilitate any secondary processes that may want to use
+ * this mempool.
+ */
+ int32_t ops_index;
struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
@@ -235,7 +248,7 @@ struct rte_mempool {
#define MEMPOOL_F_NO_CACHE_ALIGN 0x0002 /**< Do not align objs on cache lines.*/
#define MEMPOOL_F_SP_PUT 0x0004 /**< Default put is "single-producer".*/
#define MEMPOOL_F_SC_GET 0x0008 /**< Default get is "single-consumer".*/
-#define MEMPOOL_F_RING_CREATED 0x0010 /**< Internal: ring is created */
+#define MEMPOOL_F_POOL_CREATED 0x0010 /**< Internal: pool is created. */
#define MEMPOOL_F_NO_PHYS_CONTIG 0x0020 /**< Don't need physically contiguous objs. */
/**
@@ -325,6 +338,215 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
#define __mempool_check_cookies(mp, obj_table_const, n, free) do {} while(0)
#endif /* RTE_LIBRTE_MEMPOOL_DEBUG */
+#define RTE_MEMPOOL_OPS_NAMESIZE 32 /**< Max length of ops struct name. */
+
+/**
+ * Prototype for implementation specific data provisioning function.
+ *
+ * The function should provide the implementation specific memory for
+ * for use by the other mempool ops functions in a given mempool ops struct.
+ * E.g. the default ops provides an instance of the rte_ring for this purpose.
+ * it will most likely point to a different type of data structure, and
+ * will be transparent to the application programmer.
+ * This function should set mp->pool_data.
+ */
+typedef int (*rte_mempool_alloc_t)(struct rte_mempool *mp);
+
+/**
+ * Free the opaque private data pointed to by mp->pool_data pointer.
+ */
+typedef void (*rte_mempool_free_t)(struct rte_mempool *mp);
+
+/**
+ * Enqueue an object into the external pool.
+ */
+typedef int (*rte_mempool_enqueue_t)(struct rte_mempool *mp,
+ void * const *obj_table, unsigned int n);
+
+/**
+ * Dequeue an object from the external pool.
+ */
+typedef int (*rte_mempool_dequeue_t)(struct rte_mempool *mp,
+ void **obj_table, unsigned int n);
+
+/**
+ * Return the number of available objects in the external pool.
+ */
+typedef unsigned (*rte_mempool_get_count)(const struct rte_mempool *mp);
+
+/** Structure defining mempool operations structure */
+struct rte_mempool_ops {
+ char name[RTE_MEMPOOL_OPS_NAMESIZE]; /**< Name of mempool ops struct. */
+ rte_mempool_alloc_t alloc; /**< Allocate private data. */
+ rte_mempool_free_t free; /**< Free the external pool. */
+ rte_mempool_enqueue_t enqueue; /**< Enqueue an object. */
+ rte_mempool_dequeue_t dequeue; /**< Dequeue an object. */
+ rte_mempool_get_count get_count; /**< Get qty of available objs. */
+} __rte_cache_aligned;
+
+#define RTE_MEMPOOL_MAX_OPS_IDX 16 /**< Max registered ops structs */
+
+/**
+ * Structure storing the table of registered ops structs, each of which contain
+ * the function pointers for the mempool ops functions.
+ * Each process has its own storage for this ops struct array so that
+ * the mempools can be shared across primary and secondary processes.
+ * The indices used to access the array are valid across processes, whereas
+ * any function pointers stored directly in the mempool struct would not be.
+ * This results in us simply having "ops_index" in the mempool struct.
+ */
+struct rte_mempool_ops_table {
+ rte_spinlock_t sl; /**< Spinlock for add/delete. */
+ uint32_t num_ops; /**< Number of used ops structs in the table. */
+ /**
+ * Storage for all possible ops structs.
+ */
+ struct rte_mempool_ops ops[RTE_MEMPOOL_MAX_OPS_IDX];
+} __rte_cache_aligned;
+
+/** Array of registered ops structs. */
+extern struct rte_mempool_ops_table rte_mempool_ops_table;
+
+/**
+ * @internal Get the mempool ops struct from its index.
+ *
+ * @param ops_index
+ * The index of the ops struct in the ops struct table. It must be a valid
+ * index: (0 <= idx < num_ops).
+ * @return
+ * The pointer to the ops struct in the table.
+ */
+static inline struct rte_mempool_ops *
+rte_mempool_get_ops(int ops_index)
+{
+ RTE_VERIFY((ops_index >= 0) && (ops_index < RTE_MEMPOOL_MAX_OPS_IDX));
+
+ return &rte_mempool_ops_table.ops[ops_index];
+}
+
+/**
+ * @internal Wrapper for mempool_ops alloc callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * - 0: Success; successfully allocated mempool pool_data.
+ * - <0: Error; code of alloc function.
+ */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp);
+
+/**
+ * @internal Wrapper for mempool_ops dequeue callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to get.
+ * @return
+ * - 0: Success; got n objects.
+ * - <0: Error; code of dequeue function.
+ */
+static inline int
+rte_mempool_ops_dequeue_bulk(struct rte_mempool *mp,
+ void **obj_table, unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_get_ops(mp->ops_index);
+ return ops->dequeue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops enqueue callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to put.
+ * @return
+ * - 0: Success; n objects supplied.
+ * - <0: Error; code of enqueue function.
+ */
+static inline int
+rte_mempool_ops_enqueue_bulk(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_get_ops(mp->ops_index);
+ return ops->enqueue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops get_count callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * The number of available objects in the external pool.
+ */
+unsigned
+rte_mempool_ops_get_count(const struct rte_mempool *mp);
+
+/**
+ * @internal wrapper for mempool_ops free callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ */
+void
+rte_mempool_ops_free(struct rte_mempool *mp);
+
+/**
+ * Set the ops of a mempool.
+ *
+ * This can only be done on a mempool that is not populated, i.e. just after
+ * a call to rte_mempool_create_empty().
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param name
+ * Name of the ops structure to use for this mempool.
+ * @param pool_config
+ * Opaque data that can be passed by the application to the ops functions.
+ * @return
+ * - 0: Success; the mempool is now using the requested ops functions.
+ * - -EINVAL - Invalid ops struct name provided.
+ * - -EEXIST - mempool already has an ops struct assigned.
+ */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config);
+
+/**
+ * Register mempool operations.
+ *
+ * @param ops
+ * Pointer to an ops structure to register.
+ * @return
+ * - >=0: Success; return the index of the ops struct in the table.
+ * - -EINVAL - some missing callbacks while registering ops struct.
+ * - -ENOSPC - the maximum number of ops structs has been reached.
+ */
+int rte_mempool_register_ops(const struct rte_mempool_ops *ops);
+
+/**
+ * Macro to statically register the ops of a mempool handler.
+ * Note that the rte_mempool_register_ops fails silently here when
+ * more then RTE_MEMPOOL_MAX_OPS_IDX is registered.
+ */
+#define MEMPOOL_REGISTER_OPS(ops) \
+ void mp_hdlr_init_##ops(void); \
+ void __attribute__((constructor, used)) mp_hdlr_init_##ops(void)\
+ { \
+ rte_mempool_register_ops(&ops); \
+ }
+
/**
* An object callback function for mempool.
*
@@ -774,7 +996,7 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
cache->len += n;
if (cache->len >= flushthresh) {
- rte_ring_mp_enqueue_bulk(mp->ring, &cache->objs[cache_size],
+ rte_mempool_ops_enqueue_bulk(mp, &cache->objs[cache_size],
cache->len - cache_size);
cache->len = cache_size;
}
@@ -785,19 +1007,10 @@ ring_enqueue:
/* push remaining objects in ring */
#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
- if (is_mp) {
- if (rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
- else {
- if (rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
+ if (rte_mempool_ops_enqueue_bulk(mp, obj_table, n) < 0)
+ rte_panic("cannot put objects in mempool\n");
#else
- if (is_mp)
- rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n);
- else
- rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n);
+ rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
#endif
}
@@ -945,7 +1158,8 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
uint32_t req = n + (cache_size - cache->len);
/* How many do we require i.e. number to fill the cache + the request */
- ret = rte_ring_mc_dequeue_bulk(mp->ring, &cache->objs[cache->len], req);
+ ret = rte_mempool_ops_dequeue_bulk(mp,
+ &cache->objs[cache->len], req);
if (unlikely(ret < 0)) {
/*
* In the offchance that we are buffer constrained,
@@ -972,10 +1186,7 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
ring_dequeue:
/* get remaining objects from ring */
- if (is_mc)
- ret = rte_ring_mc_dequeue_bulk(mp->ring, obj_table, n);
- else
- ret = rte_ring_sc_dequeue_bulk(mp->ring, obj_table, n);
+ ret = rte_mempool_ops_dequeue_bulk(mp, obj_table, n);
if (ret < 0)
__MEMPOOL_STAT_ADD(mp, get_fail, n);
diff --git a/lib/librte_mempool/rte_mempool_ops.c b/lib/librte_mempool/rte_mempool_ops.c
new file mode 100644
index 0000000..fd0b64c
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ops.c
@@ -0,0 +1,151 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Copyright(c) 2016 6WIND S.A.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_mempool.h>
+#include <rte_errno.h>
+
+/* indirect jump table to support external memory pools. */
+struct rte_mempool_ops_table rte_mempool_ops_table = {
+ .sl = RTE_SPINLOCK_INITIALIZER,
+ .num_ops = 0
+};
+
+/* add a new ops struct in rte_mempool_ops_table, return its index. */
+int
+rte_mempool_register_ops(const struct rte_mempool_ops *h)
+{
+ struct rte_mempool_ops *ops;
+ int16_t ops_index;
+
+ rte_spinlock_lock(&rte_mempool_ops_table.sl);
+
+ if (rte_mempool_ops_table.num_ops >=
+ RTE_MEMPOOL_MAX_OPS_IDX) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Maximum number of mempool ops structs exceeded\n");
+ return -ENOSPC;
+ }
+
+ if (h->alloc == NULL || h->enqueue == NULL ||
+ h->dequeue == NULL || h->get_count == NULL) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Missing callback while registering mempool ops\n");
+ return -EINVAL;
+ }
+
+ if (strlen(h->name) >= sizeof(ops->name) - 1) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(DEBUG, EAL, "%s(): mempool_ops <%s>: name too long\n",
+ __func__, h->name);
+ rte_errno = EEXIST;
+ return -EEXIST;
+ }
+
+ ops_index = rte_mempool_ops_table.num_ops++;
+ ops = &rte_mempool_ops_table.ops[ops_index];
+ snprintf(ops->name, sizeof(ops->name), "%s", h->name);
+ ops->alloc = h->alloc;
+ ops->enqueue = h->enqueue;
+ ops->dequeue = h->dequeue;
+ ops->get_count = h->get_count;
+
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+
+ return ops_index;
+}
+
+/* wrapper to allocate an external mempool's private (pool) data. */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_get_ops(mp->ops_index);
+ return ops->alloc(mp);
+}
+
+/* wrapper to free an external pool ops. */
+void
+rte_mempool_ops_free(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_get_ops(mp->ops_index);
+ if (ops->free == NULL)
+ return;
+ ops->free(mp);
+}
+
+/* wrapper to get available objects in an external mempool. */
+unsigned int
+rte_mempool_ops_get_count(const struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_get_ops(mp->ops_index);
+ return ops->get_count(mp);
+}
+
+/* sets mempool ops previously registered by rte_mempool_register_ops. */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config)
+{
+ struct rte_mempool_ops *ops = NULL;
+ unsigned i;
+
+ /* too late, the mempool is already populated. */
+ if (mp->flags & MEMPOOL_F_POOL_CREATED)
+ return -EEXIST;
+
+ for (i = 0; i < rte_mempool_ops_table.num_ops; i++) {
+ if (!strcmp(name,
+ rte_mempool_ops_table.ops[i].name)) {
+ ops = &rte_mempool_ops_table.ops[i];
+ break;
+ }
+ }
+
+ if (ops == NULL)
+ return -EINVAL;
+
+ mp->ops_index = i;
+ mp->pool_config = pool_config;
+ return 0;
+}
diff --git a/lib/librte_mempool/rte_mempool_ring.c b/lib/librte_mempool/rte_mempool_ring.c
new file mode 100644
index 0000000..b9aa64d
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ring.c
@@ -0,0 +1,161 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_errno.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+
+static int
+common_ring_mp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_mp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_sp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_mc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_mc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_sc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static unsigned
+common_ring_get_count(const struct rte_mempool *mp)
+{
+ return rte_ring_count(mp->pool_data);
+}
+
+
+static int
+common_ring_alloc(struct rte_mempool *mp)
+{
+ int rg_flags = 0, ret;
+ char rg_name[RTE_RING_NAMESIZE];
+ struct rte_ring *r;
+
+ ret = snprintf(rg_name, sizeof(rg_name),
+ RTE_MEMPOOL_MZ_FORMAT, mp->name);
+ if (ret < 0 || ret >= (int)sizeof(rg_name)) {
+ rte_errno = ENAMETOOLONG;
+ return -rte_errno;
+ }
+
+ /* ring flags */
+ if (mp->flags & MEMPOOL_F_SP_PUT)
+ rg_flags |= RING_F_SP_ENQ;
+ if (mp->flags & MEMPOOL_F_SC_GET)
+ rg_flags |= RING_F_SC_DEQ;
+
+ /*
+ * Allocate the ring that will be used to store objects.
+ * Ring functions will return appropriate errors if we are
+ * running as a secondary process etc., so no checks made
+ * in this function for that condition.
+ */
+ r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
+ mp->socket_id, rg_flags);
+ if (r == NULL)
+ return -rte_errno;
+
+ mp->pool_data = r;
+
+ return 0;
+}
+
+static void
+common_ring_free(struct rte_mempool *mp)
+{
+ rte_ring_free(mp->pool_data);
+}
+
+/*
+ * The following 4 declarations of mempool ops structs address
+ * the need for the backward compatible mempool handlers for
+ * single/multi producers and single/multi consumers as dictated by the
+ * flags provided to the rte_mempool_create function
+ */
+static const struct rte_mempool_ops ops_mp_mc = {
+ .name = "ring_mp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_sc = {
+ .name = "ring_sp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_mp_sc = {
+ .name = "ring_mp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_mc = {
+ .name = "ring_sp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+MEMPOOL_REGISTER_OPS(ops_mp_mc);
+MEMPOOL_REGISTER_OPS(ops_sp_sc);
+MEMPOOL_REGISTER_OPS(ops_mp_sc);
+MEMPOOL_REGISTER_OPS(ops_sp_mc);
diff --git a/lib/librte_mempool/rte_mempool_version.map b/lib/librte_mempool/rte_mempool_version.map
index f63461b..a4a6c1f 100644
--- a/lib/librte_mempool/rte_mempool_version.map
+++ b/lib/librte_mempool/rte_mempool_version.map
@@ -20,15 +20,18 @@ DPDK_16.7 {
global:
rte_mempool_check_cookies;
- rte_mempool_obj_iter;
- rte_mempool_mem_iter;
rte_mempool_create_empty;
+ rte_mempool_free;
+ rte_mempool_mem_iter;
+ rte_mempool_obj_iter;
+ rte_mempool_ops_table;
+ rte_mempool_populate_anon;
+ rte_mempool_populate_default;
rte_mempool_populate_phys;
rte_mempool_populate_phys_tab;
rte_mempool_populate_virt;
- rte_mempool_populate_default;
- rte_mempool_populate_anon;
- rte_mempool_free;
+ rte_mempool_register_ops;
+ rte_mempool_set_ops_byname;
local: *;
} DPDK_2.0;
--
2.5.5
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v16 0/3] mempool: add mempool handler feature
2016-06-19 12:05 3% ` [dpdk-dev] [PATCH v15 0/3] mempool: add mempool handler feature David Hunt
2016-06-19 12:05 1% ` [dpdk-dev] [PATCH v15 1/3] mempool: support mempool handler operations David Hunt
@ 2016-06-22 9:27 3% ` David Hunt
2016-06-22 9:27 1% ` [dpdk-dev] [PATCH v16 1/3] mempool: support mempool handler operations David Hunt
1 sibling, 1 reply; 200+ results
From: David Hunt @ 2016-06-22 9:27 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the Mempool Handler patch set.
It's re-based on top of the latest head as of 20/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v16 changes:
* Changed rte_mempool_ops_get() to rte_mempool_get_ops()
* Changed rte_mempool_ops_register() to rte_mempool_register_ops()
* Applied missing changes that should have been in v15
v15 changes:
* Changed rte_mempool_ops_get() to rte_mempool_get_ops()
* Did some minor tweaks to comments after the previous change of function
names from put/get to enqueue/dequeue
* Added missing spinlock_unlock in rte_mempool_ops_register()
* Added check for null in ops_free
* removed un-needed return statement
v14 changes:
* set MEMPOOL_F_RING_CREATED flag after rte_mempool_ring_create() is called.
* Changed name of feature from "external mempool manager" to "mempool handler"
and updated comments and release notes accordingly.
* Added a comment for newly added pool_config param in
rte_mempool_set_ops_byname.
v13 changes:
* Added in extra opaque data (pool_config) to mempool struct for mempool
configuration by the ops functions. For example, this can be used to pass
device names or device flags to the underlying alloc function.
* Added mempool_config param to rte_mempool_set_ops_byname()
v12 changes:
* Fixed a comment (function pram h -> ops)
* fixed a typo (callbacki)
v11 changes:
* Fixed comments (added '.' where needed for consistency)
* removed ABI breakage notice for mempool manager in deprecation.rst
* Added description of the external mempool manager functionality to
doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
* renamed rte_mempool_default.c to rte_mempool_ring.c
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The Mempool Handler feature is an extension to the mempool API that allows
users to add and use an alternative mempool handler, which allows
external memory subsystems such as external hardware memory management
systems and software based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool handler will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing.
There are two aspects to mempool handlers.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops_byname to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempools ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several mempool handlers may be used in the same application. A new
mempool can then be created by using the new rte_mempool_create_empty function,
then calling rte_mempool_set_ops_byname to point the mempool to the relevant
mempool handler callback (ops) structure.
Legacy applications will continue to use the old rte_mempool_create API call,
which uses a ring based mempool handler by default. These applications
will need to be modified to use a new mempool handler.
A mempool handler needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name
void *pool_config);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool handler using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support mempool handler operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test mempool handler
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v5 04/17] eal: remove duplicate function declaration
@ 2016-06-22 9:06 3% ` Shreyansh Jain
0 siblings, 0 replies; 200+ results
From: Shreyansh Jain @ 2016-06-22 9:06 UTC (permalink / raw)
To: dev; +Cc: viktorin, thomas.monjalon
rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
introduction.
This function has been exported in ABI, so remove it from eal_private.h
Fixes: e57f20e05177 ("eal: make vdev init path generic for both virtual and pci devices")
Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
---
lib/librte_eal/common/eal_private.h | 7 -------
lib/librte_eal/linuxapp/eal/eal.c | 1 +
2 files changed, 1 insertion(+), 7 deletions(-)
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 857dc3e..06a68f6 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -259,13 +259,6 @@ int rte_eal_intr_init(void);
int rte_eal_alarm_init(void);
/**
- * This function initialises any virtual devices
- *
- * This function is private to the EAL.
- */
-int rte_eal_dev_init(void);
-
-/**
* Function is to check if the kernel module(like, vfio, vfio_iommu_type1,
* etc.) loaded.
*
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index 4f22c18..29fba52 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -70,6 +70,7 @@
#include <rte_cpuflags.h>
#include <rte_interrupts.h>
#include <rte_pci.h>
+#include <rte_dev.h>
#include <rte_devargs.h>
#include <rte_common.h>
#include <rte_version.h>
--
2.7.4
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v4 04/17] eal: remove duplicate function declaration
@ 2016-06-21 12:02 3% ` Shreyansh Jain
` (2 subsequent siblings)
3 siblings, 0 replies; 200+ results
From: Shreyansh Jain @ 2016-06-21 12:02 UTC (permalink / raw)
To: dev; +Cc: viktorin, thomas.monjalon
rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
introduction.
This function has been exported in ABI, so remove it from eal_private.h
Fixes: e57f20e05177 ("eal: make vdev init path generic for both virtual and pci devices")
Signed-off-by: David Marchand <david.marchand@6wind.com>
Signed-off-by: Shreyansh Jain <shreyansh.jain@nxp.com>
---
lib/librte_eal/common/eal_private.h | 7 -------
lib/librte_eal/linuxapp/eal/eal.c | 1 +
2 files changed, 1 insertion(+), 7 deletions(-)
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 857dc3e..06a68f6 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -259,13 +259,6 @@ int rte_eal_intr_init(void);
int rte_eal_alarm_init(void);
/**
- * This function initialises any virtual devices
- *
- * This function is private to the EAL.
- */
-int rte_eal_dev_init(void);
-
-/**
* Function is to check if the kernel module(like, vfio, vfio_iommu_type1,
* etc.) loaded.
*
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index 4f22c18..29fba52 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -70,6 +70,7 @@
#include <rte_cpuflags.h>
#include <rte_interrupts.h>
#include <rte_pci.h>
+#include <rte_dev.h>
#include <rte_devargs.h>
#include <rte_common.h>
#include <rte_version.h>
--
2.7.4
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2] i40e: modify the meaning of single VLAN type
2016-06-21 11:06 0% ` Panu Matilainen
@ 2016-06-21 11:28 0% ` Bruce Richardson
0 siblings, 0 replies; 200+ results
From: Bruce Richardson @ 2016-06-21 11:28 UTC (permalink / raw)
To: Panu Matilainen; +Cc: Beilei Xing, jingjing.wu, dev, thomas.monjalon, nhorman
On Tue, Jun 21, 2016 at 02:06:38PM +0300, Panu Matilainen wrote:
> On 06/21/2016 01:29 PM, Bruce Richardson wrote:
> >On Mon, Jun 13, 2016 at 04:03:32PM +0800, Beilei Xing wrote:
> >>In current i40e codebase, if single VLAN header is added in a packet,
> >>it's treated as inner VLAN. Generally, a single VLAN header is
> >>treated as the outer VLAN header. So change corresponding register
> >>for single VLAN.
> >>At the meanwhile, change the meanings of inner VLAN and outer VLAN.
> >>
> >>Signed-off-by: Beilei Xing <beilei.xing@intel.com>
> >
> >This patch changes the ABI, since an app written to the original API as specified
> >e.g. to set a single vlan header, would no longer work with this change.
> >Therefore, even though the original behaviour was inconsistent with other drivers
> >it may still need to be preserved.
> >
> >I'm thinking that we may need to provide appropriately versioned copies of the
> >vlan_offload_set and vlan_tpid_set functions for backward compatibility with
> >the old ABI.
> >
> >Any other comments or thoughts on this?
> >Neil, Thomas, Panu - is this fix something that we need to provide backward
> >version-compatibility for, or given that the functions are being called through
> >a generic ethdev API mean that this can just go in as a straight bug-fix?
>
> Since it's currently inconsistent with everything else, I'd just call it a
> bug-fix and leave it at that.
>
Yep, makes sense.
> Besides, I dont think you could version it via the ordinary means even if
> you wanted to, due to the way its called through eth_dev_ops etc.
>
Good point, never thought of that! :-(
> - Panu -
Thanks for the guidance.
/Bruce
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v2] i40e: modify the meaning of single VLAN type
2016-06-21 10:29 4% ` Bruce Richardson
@ 2016-06-21 11:06 0% ` Panu Matilainen
2016-06-21 11:28 0% ` Bruce Richardson
0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-06-21 11:06 UTC (permalink / raw)
To: Bruce Richardson, Beilei Xing; +Cc: jingjing.wu, dev, thomas.monjalon, nhorman
On 06/21/2016 01:29 PM, Bruce Richardson wrote:
> On Mon, Jun 13, 2016 at 04:03:32PM +0800, Beilei Xing wrote:
>> In current i40e codebase, if single VLAN header is added in a packet,
>> it's treated as inner VLAN. Generally, a single VLAN header is
>> treated as the outer VLAN header. So change corresponding register
>> for single VLAN.
>> At the meanwhile, change the meanings of inner VLAN and outer VLAN.
>>
>> Signed-off-by: Beilei Xing <beilei.xing@intel.com>
>
> This patch changes the ABI, since an app written to the original API as specified
> e.g. to set a single vlan header, would no longer work with this change.
> Therefore, even though the original behaviour was inconsistent with other drivers
> it may still need to be preserved.
>
> I'm thinking that we may need to provide appropriately versioned copies of the
> vlan_offload_set and vlan_tpid_set functions for backward compatibility with
> the old ABI.
>
> Any other comments or thoughts on this?
> Neil, Thomas, Panu - is this fix something that we need to provide backward
> version-compatibility for, or given that the functions are being called through
> a generic ethdev API mean that this can just go in as a straight bug-fix?
Since it's currently inconsistent with everything else, I'd just call it
a bug-fix and leave it at that.
Besides, I dont think you could version it via the ordinary means even
if you wanted to, due to the way its called through eth_dev_ops etc.
- Panu -
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v2] i40e: modify the meaning of single VLAN type
2016-06-13 8:03 4% ` [dpdk-dev] [PATCH v2] i40e: modify the meaning of single " Beilei Xing
@ 2016-06-21 10:29 4% ` Bruce Richardson
2016-06-21 11:06 0% ` Panu Matilainen
0 siblings, 1 reply; 200+ results
From: Bruce Richardson @ 2016-06-21 10:29 UTC (permalink / raw)
To: Beilei Xing; +Cc: jingjing.wu, dev, thomas.monjalon, nhorman, pmatilai
On Mon, Jun 13, 2016 at 04:03:32PM +0800, Beilei Xing wrote:
> In current i40e codebase, if single VLAN header is added in a packet,
> it's treated as inner VLAN. Generally, a single VLAN header is
> treated as the outer VLAN header. So change corresponding register
> for single VLAN.
> At the meanwhile, change the meanings of inner VLAN and outer VLAN.
>
> Signed-off-by: Beilei Xing <beilei.xing@intel.com>
This patch changes the ABI, since an app written to the original API as specified
e.g. to set a single vlan header, would no longer work with this change.
Therefore, even though the original behaviour was inconsistent with other drivers
it may still need to be preserved.
I'm thinking that we may need to provide appropriately versioned copies of the
vlan_offload_set and vlan_tpid_set functions for backward compatibility with
the old ABI.
Any other comments or thoughts on this?
Neil, Thomas, Panu - is this fix something that we need to provide backward
version-compatibility for, or given that the functions are being called through
a generic ethdev API mean that this can just go in as a straight bug-fix?
/Bruce
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v3 1/2] mempool: add stack (lifo) mempool handler
@ 2016-06-21 9:44 3% ` Olivier Matz
0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-06-21 9:44 UTC (permalink / raw)
To: Ananyev, Konstantin, Jerin Jacob
Cc: Thomas Monjalon, dev, Hunt, David, viktorin, shreyansh.jain
Hi,
On 06/21/2016 11:28 AM, Ananyev, Konstantin wrote:
>>>> I was proposing only to move only the new
>>>> handler(lib/librte_mempool/rte_mempool_stack.c). Not any library or any
>>>> other common code.
>>>>
>>>> Just like DPDK crypto device, Even if it is software implementation its
>>>> better to move in driver/crypto instead of lib/librte_cryptodev
>>>>
>>>> "lib/librte_mempool/arch/" is not correct place as it is platform specific
>>>> not architecture specific and HW mempool device may be PCIe or platform
>>>> device.
>>>
>>> Ok, but why rte_mempool_stack.c has to be moved?
>>
>> Just thought of having all the mempool handlers at one place.
>> We can't move all HW mempool handlers at lib/librte_mempool/
>
> Yep, but from your previous mail I thought we might have specific ones
> for specific devices, no?
> If so, why to put them in one place, why just not in:
> Drivers/xxx_dev/xxx_mempool.[h,c]
> ?
> And keep generic ones in lib/librte_mempool
> ?
I think all drivers (generic or not) should be at the same place for
consistency.
I'm not sure having them lib/librte_mempool is really a problem today,
but once hardware-dependent handler are pushed, we may move all of them
in drivers/mempool because I think we should avoid to have hw-specific
code in lib/.
I don't think it will cause ABI/API breakage since the user always
talk to the generic mempool API.
Regards
Olivier
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v15 1/3] mempool: support mempool handler operations
2016-06-19 12:05 3% ` [dpdk-dev] [PATCH v15 0/3] mempool: add mempool handler feature David Hunt
@ 2016-06-19 12:05 1% ` David Hunt
2016-06-22 9:27 3% ` [dpdk-dev] [PATCH v16 0/3] mempool: add mempool handler feature David Hunt
1 sibling, 0 replies; 200+ results
From: David Hunt @ 2016-06-19 12:05 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain, David Hunt
Until now, the objects stored in a mempool were internally stored in a
ring. This patch introduces the possibility to register external handlers
replacing the ring.
The default behavior remains unchanged, but calling the new function
rte_mempool_set_ops_byname() right after rte_mempool_create_empty() allows
the user to change the handler that will be used when populating
the mempool.
This patch also adds a set of default ops (function callbacks) based
on rte_ring.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---
app/test/test_mempool_perf.c | 1 -
doc/guides/prog_guide/mempool_lib.rst | 32 +++-
doc/guides/rel_notes/deprecation.rst | 9 -
lib/librte_mempool/Makefile | 2 +
lib/librte_mempool/rte_mempool.c | 67 +++-----
lib/librte_mempool/rte_mempool.h | 255 ++++++++++++++++++++++++++---
lib/librte_mempool/rte_mempool_ops.c | 150 +++++++++++++++++
lib/librte_mempool/rte_mempool_ring.c | 161 ++++++++++++++++++
lib/librte_mempool/rte_mempool_version.map | 13 +-
9 files changed, 609 insertions(+), 81 deletions(-)
create mode 100644 lib/librte_mempool/rte_mempool_ops.c
create mode 100644 lib/librte_mempool/rte_mempool_ring.c
diff --git a/app/test/test_mempool_perf.c b/app/test/test_mempool_perf.c
index c5e3576..c5f8455 100644
--- a/app/test/test_mempool_perf.c
+++ b/app/test/test_mempool_perf.c
@@ -161,7 +161,6 @@ per_lcore_mempool_test(__attribute__((unused)) void *arg)
n_get_bulk);
if (unlikely(ret < 0)) {
rte_mempool_dump(stdout, mp);
- rte_ring_dump(stdout, mp->ring);
/* in this case, objects are lost... */
return -1;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index c3afc2e..1943fc4 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -34,7 +34,8 @@ Mempool Library
===============
A memory pool is an allocator of a fixed-sized object.
-In the DPDK, it is identified by name and uses a ring to store free objects.
+In the DPDK, it is identified by name and uses a mempool handler to store free objects.
+The default mempool handler is ring based.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
@@ -127,6 +128,35 @@ The maximum size of the cache is static and is defined at compilation time (CONF
A mempool in Memory with its Associated Ring
+Mempool Handlers
+------------------------
+
+This allows external memory subsystems, such as external hardware memory
+management systems and software based memory allocators, to be used with DPDK.
+
+There are two aspects to a mempool handler.
+
+* Adding the code for your new mempool operations (ops). This is achieved by
+ adding a new mempool ops code, and using the ``REGISTER_MEMPOOL_OPS`` macro.
+
+* Using the new API to call ``rte_mempool_create_empty()`` and
+ ``rte_mempool_set_ops_byname()`` to create a new mempool and specifying which
+ ops to use.
+
+Several different mempool handlers may be used in the same application. A new
+mempool can be created by using the ``rte_mempool_create_empty()`` function,
+then using ``rte_mempool_set_ops_byname()`` to point the mempool to the
+relevant mempool handler callback (ops) structure.
+
+Legacy applications may continue to use the old ``rte_mempool_create()`` API
+call, which uses a ring based mempool handler by default. These applications
+will need to be modified to use a new mempool handler.
+
+For applications that use ``rte_pktmbuf_create()``, there is a config setting
+(``RTE_MBUF_DEFAULT_MEMPOOL_OPS``) that allows the application to make use of
+an alternative mempool handler.
+
+
Use Cases
---------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f75183f..3cbc19e 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -34,15 +34,6 @@ Deprecation Notices
compact API. The ones that remain are backwards compatible and use the
per-lcore default cache if available. This change targets release 16.07.
-* The rte_mempool struct will be changed in 16.07 to facilitate the new
- external mempool manager functionality.
- The ring element will be replaced with a more generic 'pool' opaque pointer
- to allow new mempool handlers to use their own user-defined mempool
- layout. Also newly added to rte_mempool is a handler index.
- The existing API will be backward compatible, but there will be new API
- functions added to facilitate the creation of mempools using an external
- handler. The 16.07 release will contain these changes.
-
* A librte_vhost public structures refactor is planned for DPDK 16.07
that requires both ABI and API change.
The proposed refactor would expose DPDK vhost dev to applications as
diff --git a/lib/librte_mempool/Makefile b/lib/librte_mempool/Makefile
index 43423e0..a4c089e 100644
--- a/lib/librte_mempool/Makefile
+++ b/lib/librte_mempool/Makefile
@@ -42,6 +42,8 @@ LIBABIVER := 2
# all source are stored in SRCS-y
SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ring.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_MEMPOOL)-include := rte_mempool.h
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index af71edd..e6a83d0 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -148,7 +148,7 @@ mempool_add_elem(struct rte_mempool *mp, void *obj, phys_addr_t physaddr)
#endif
/* enqueue in ring */
- rte_ring_sp_enqueue(mp->ring, obj);
+ rte_mempool_ops_enqueue_bulk(mp, &obj, 1);
}
/* call obj_cb() for each mempool element */
@@ -303,40 +303,6 @@ rte_mempool_xmem_usage(__rte_unused void *vaddr, uint32_t elt_num,
return (size_t)paddr_idx << pg_shift;
}
-/* create the internal ring */
-static int
-rte_mempool_ring_create(struct rte_mempool *mp)
-{
- int rg_flags = 0, ret;
- char rg_name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
-
- ret = snprintf(rg_name, sizeof(rg_name),
- RTE_MEMPOOL_MZ_FORMAT, mp->name);
- if (ret < 0 || ret >= (int)sizeof(rg_name))
- return -ENAMETOOLONG;
-
- /* ring flags */
- if (mp->flags & MEMPOOL_F_SP_PUT)
- rg_flags |= RING_F_SP_ENQ;
- if (mp->flags & MEMPOOL_F_SC_GET)
- rg_flags |= RING_F_SC_DEQ;
-
- /* Allocate the ring that will be used to store objects.
- * Ring functions will return appropriate errors if we are
- * running as a secondary process etc., so no checks made
- * in this function for that condition.
- */
- r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
- mp->socket_id, rg_flags);
- if (r == NULL)
- return -rte_errno;
-
- mp->ring = r;
- mp->flags |= MEMPOOL_F_RING_CREATED;
- return 0;
-}
-
/* free a memchunk allocated with rte_memzone_reserve() */
static void
rte_mempool_memchunk_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
@@ -354,7 +320,7 @@ rte_mempool_free_memchunks(struct rte_mempool *mp)
void *elt;
while (!STAILQ_EMPTY(&mp->elt_list)) {
- rte_ring_sc_dequeue(mp->ring, &elt);
+ rte_mempool_ops_dequeue_bulk(mp, &elt, 1);
(void)elt;
STAILQ_REMOVE_HEAD(&mp->elt_list, next);
mp->populated_size--;
@@ -386,10 +352,11 @@ rte_mempool_populate_phys(struct rte_mempool *mp, char *vaddr,
int ret;
/* create the internal ring if not already done */
- if ((mp->flags & MEMPOOL_F_RING_CREATED) == 0) {
- ret = rte_mempool_ring_create(mp);
- if (ret < 0)
+ if ((mp->flags & MEMPOOL_F_POOL_CREATED) == 0) {
+ ret = rte_mempool_ops_alloc(mp);
+ if (ret != 0)
return ret;
+ mp->flags |= MEMPOOL_F_POOL_CREATED;
}
/* mempool is already populated */
@@ -703,7 +670,7 @@ rte_mempool_free(struct rte_mempool *mp)
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_mempool_free_memchunks(mp);
- rte_ring_free(mp->ring);
+ rte_mempool_ops_free(mp);
rte_memzone_free(mp->mz);
}
@@ -815,6 +782,7 @@ rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, 0));
te->data = mp;
+
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_INSERT_TAIL(mempool_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
@@ -844,6 +812,19 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
if (mp == NULL)
return NULL;
+ /*
+ * Since we have 4 combinations of the SP/SC/MP/MC examine the flags to
+ * set the correct index into the table of ops structs.
+ */
+ if (flags & (MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET))
+ rte_mempool_set_ops_byname(mp, "ring_sp_sc", NULL);
+ else if (flags & MEMPOOL_F_SP_PUT)
+ rte_mempool_set_ops_byname(mp, "ring_sp_mc", NULL);
+ else if (flags & MEMPOOL_F_SC_GET)
+ rte_mempool_set_ops_byname(mp, "ring_mp_sc", NULL);
+ else
+ rte_mempool_set_ops_byname(mp, "ring_mp_mc", NULL);
+
/* call the mempool priv initializer */
if (mp_init)
mp_init(mp, mp_init_arg);
@@ -930,7 +911,7 @@ rte_mempool_count(const struct rte_mempool *mp)
unsigned count;
unsigned lcore_id;
- count = rte_ring_count(mp->ring);
+ count = rte_mempool_ops_get_count(mp);
if (mp->cache_size == 0)
return count;
@@ -1119,7 +1100,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
fprintf(f, "mempool <%s>@%p\n", mp->name, mp);
fprintf(f, " flags=%x\n", mp->flags);
- fprintf(f, " ring=<%s>@%p\n", mp->ring->name, mp->ring);
+ fprintf(f, " pool=%p\n", mp->pool_data);
fprintf(f, " phys_addr=0x%" PRIx64 "\n", mp->mz->phys_addr);
fprintf(f, " nb_mem_chunks=%u\n", mp->nb_mem_chunks);
fprintf(f, " size=%"PRIu32"\n", mp->size);
@@ -1140,7 +1121,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
}
cache_count = rte_mempool_dump_cache(f, mp);
- common_count = rte_ring_count(mp->ring);
+ common_count = rte_mempool_ops_get_count(mp);
if ((cache_count + common_count) > mp->size)
common_count = mp->size - cache_count;
fprintf(f, " common_pool_count=%u\n", common_count);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 60339bd..2d7c980 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -67,6 +67,7 @@
#include <inttypes.h>
#include <sys/queue.h>
+#include <rte_spinlock.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_lcore.h>
@@ -203,10 +204,14 @@ struct rte_mempool_memhdr {
*/
struct rte_mempool {
char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
- struct rte_ring *ring; /**< Ring to store objects. */
- const struct rte_memzone *mz; /**< Memzone where pool is allocated */
+ union {
+ void *pool_data; /**< Ring or pool to store objects. */
+ uint64_t pool_id; /**< External mempool identifier. */
+ };
+ void *pool_config; /**< optional args for ops alloc. */
+ const struct rte_memzone *mz; /**< Memzone where pool is alloc'd. */
int flags; /**< Flags of the mempool. */
- int socket_id; /**< Socket id passed at mempool creation. */
+ int socket_id; /**< Socket id passed at create. */
uint32_t size; /**< Max size of the mempool. */
uint32_t cache_size; /**< Size of per-lcore local cache. */
uint32_t cache_flushthresh;
@@ -217,6 +222,14 @@ struct rte_mempool {
uint32_t trailer_size; /**< Size of trailer (after elt). */
unsigned private_data_size; /**< Size of private data. */
+ /**
+ * Index into rte_mempool_ops_table array of mempool ops
+ * structs, which contain callback function pointers.
+ * We're using an index here rather than pointers to the callbacks
+ * to facilitate any secondary processes that may want to use
+ * this mempool.
+ */
+ int32_t ops_index;
struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
@@ -235,7 +248,7 @@ struct rte_mempool {
#define MEMPOOL_F_NO_CACHE_ALIGN 0x0002 /**< Do not align objs on cache lines.*/
#define MEMPOOL_F_SP_PUT 0x0004 /**< Default put is "single-producer".*/
#define MEMPOOL_F_SC_GET 0x0008 /**< Default get is "single-consumer".*/
-#define MEMPOOL_F_RING_CREATED 0x0010 /**< Internal: ring is created */
+#define MEMPOOL_F_POOL_CREATED 0x0010 /**< Internal: pool is created. */
#define MEMPOOL_F_NO_PHYS_CONTIG 0x0020 /**< Don't need physically contiguous objs. */
/**
@@ -325,6 +338,215 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
#define __mempool_check_cookies(mp, obj_table_const, n, free) do {} while(0)
#endif /* RTE_LIBRTE_MEMPOOL_DEBUG */
+#define RTE_MEMPOOL_OPS_NAMESIZE 32 /**< Max length of ops struct name. */
+
+/**
+ * Prototype for implementation specific data provisioning function.
+ *
+ * The function should provide the implementation specific memory for
+ * for use by the other mempool ops functions in a given mempool ops struct.
+ * E.g. the default ops provides an instance of the rte_ring for this purpose.
+ * it will most likely point to a different type of data structure, and
+ * will be transparent to the application programmer.
+ * This function should set mp->pool_data.
+ */
+typedef int (*rte_mempool_alloc_t)(struct rte_mempool *mp);
+
+/**
+ * Free the opaque private data pointed to by mp->pool_data pointer.
+ */
+typedef void (*rte_mempool_free_t)(struct rte_mempool *mp);
+
+/**
+ * Enqueue an object into the external pool.
+ */
+typedef int (*rte_mempool_enqueue_t)(struct rte_mempool *mp,
+ void * const *obj_table, unsigned int n);
+
+/**
+ * Dequeue an object from the external pool.
+ */
+typedef int (*rte_mempool_dequeue_t)(struct rte_mempool *mp,
+ void **obj_table, unsigned int n);
+
+/**
+ * Return the number of available objects in the external pool.
+ */
+typedef unsigned (*rte_mempool_get_count)(const struct rte_mempool *mp);
+
+/** Structure defining mempool operations structure */
+struct rte_mempool_ops {
+ char name[RTE_MEMPOOL_OPS_NAMESIZE]; /**< Name of mempool ops struct. */
+ rte_mempool_alloc_t alloc; /**< Allocate private data. */
+ rte_mempool_free_t free; /**< Free the external pool. */
+ rte_mempool_enqueue_t enqueue; /**< Enqueue an object. */
+ rte_mempool_dequeue_t dequeue; /**< Dequeue an object. */
+ rte_mempool_get_count get_count; /**< Get qty of available objs. */
+} __rte_cache_aligned;
+
+#define RTE_MEMPOOL_MAX_OPS_IDX 16 /**< Max registered ops structs */
+
+/**
+ * Structure storing the table of registered ops structs, each of which contain
+ * the function pointers for the mempool ops functions.
+ * Each process has its own storage for this ops struct array so that
+ * the mempools can be shared across primary and secondary processes.
+ * The indices used to access the array are valid across processes, whereas
+ * any function pointers stored directly in the mempool struct would not be.
+ * This results in us simply having "ops_index" in the mempool struct.
+ */
+struct rte_mempool_ops_table {
+ rte_spinlock_t sl; /**< Spinlock for add/delete. */
+ uint32_t num_ops; /**< Number of used ops structs in the table. */
+ /**
+ * Storage for all possible ops structs.
+ */
+ struct rte_mempool_ops ops[RTE_MEMPOOL_MAX_OPS_IDX];
+} __rte_cache_aligned;
+
+/** Array of registered ops structs. */
+extern struct rte_mempool_ops_table rte_mempool_ops_table;
+
+/**
+ * @internal Get the mempool ops struct from its index.
+ *
+ * @param ops_index
+ * The index of the ops struct in the ops struct table. It must be a valid
+ * index: (0 <= idx < num_ops).
+ * @return
+ * The pointer to the ops struct in the table.
+ */
+static inline struct rte_mempool_ops *
+rte_mempool_ops_get(int ops_index)
+{
+ RTE_VERIFY(ops_index < RTE_MEMPOOL_MAX_OPS_IDX);
+
+ return &rte_mempool_ops_table.ops[ops_index];
+}
+
+/**
+ * @internal Wrapper for mempool_ops alloc callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * - 0: Success; successfully allocated mempool pool_data.
+ * - <0: Error; code of alloc function.
+ */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp);
+
+/**
+ * @internal Wrapper for mempool_ops get callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to get.
+ * @return
+ * - 0: Success; got n objects.
+ * - <0: Error; code of get function.
+ */
+static inline int
+rte_mempool_ops_dequeue_bulk(struct rte_mempool *mp,
+ void **obj_table, unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->dequeue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops put callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to put.
+ * @return
+ * - 0: Success; n objects supplied.
+ * - <0: Error; code of put function.
+ */
+static inline int
+rte_mempool_ops_enqueue_bulk(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->enqueue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops get_count callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * The number of available objects in the external pool.
+ */
+unsigned
+rte_mempool_ops_get_count(const struct rte_mempool *mp);
+
+/**
+ * @internal wrapper for mempool_ops free callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ */
+void
+rte_mempool_ops_free(struct rte_mempool *mp);
+
+/**
+ * Set the ops of a mempool.
+ *
+ * This can only be done on a mempool that is not populated, i.e. just after
+ * a call to rte_mempool_create_empty().
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param name
+ * Name of the ops structure to use for this mempool.
+ * @param pool_config
+ * Opaque data that can be passed by the application to the ops functions.
+ * @return
+ * - 0: Success; the mempool is now using the requested ops functions.
+ * - -EINVAL - Invalid ops struct name provided.
+ * - -EEXIST - mempool already has an ops struct assigned.
+ */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config);
+
+/**
+ * Register mempool operations.
+ *
+ * @param ops
+ * Pointer to an ops structure to register.
+ * @return
+ * - >=0: Success; return the index of the ops struct in the table.
+ * - -EINVAL - some missing callbacks while registering ops struct.
+ * - -ENOSPC - the maximum number of ops structs has been reached.
+ */
+int rte_mempool_ops_register(const struct rte_mempool_ops *ops);
+
+/**
+ * Macro to statically register the ops of a mempool handler.
+ * Note that the rte_mempool_ops_register fails silently here when
+ * more then RTE_MEMPOOL_MAX_OPS_IDX is registered.
+ */
+#define MEMPOOL_REGISTER_OPS(ops) \
+ void mp_hdlr_init_##ops(void); \
+ void __attribute__((constructor, used)) mp_hdlr_init_##ops(void)\
+ { \
+ rte_mempool_ops_register(&ops); \
+ }
+
/**
* An object callback function for mempool.
*
@@ -774,7 +996,7 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
cache->len += n;
if (cache->len >= flushthresh) {
- rte_ring_mp_enqueue_bulk(mp->ring, &cache->objs[cache_size],
+ rte_mempool_ops_enqueue_bulk(mp, &cache->objs[cache_size],
cache->len - cache_size);
cache->len = cache_size;
}
@@ -785,19 +1007,10 @@ ring_enqueue:
/* push remaining objects in ring */
#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
- if (is_mp) {
- if (rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
- else {
- if (rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
+ if (rte_mempool_ops_enqueue_bulk(mp, obj_table, n) < 0)
+ rte_panic("cannot put objects in mempool\n");
#else
- if (is_mp)
- rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n);
- else
- rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n);
+ rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
#endif
}
@@ -945,7 +1158,8 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
uint32_t req = n + (cache_size - cache->len);
/* How many do we require i.e. number to fill the cache + the request */
- ret = rte_ring_mc_dequeue_bulk(mp->ring, &cache->objs[cache->len], req);
+ ret = rte_mempool_ops_dequeue_bulk(mp,
+ &cache->objs[cache->len], req);
if (unlikely(ret < 0)) {
/*
* In the offchance that we are buffer constrained,
@@ -972,10 +1186,7 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
ring_dequeue:
/* get remaining objects from ring */
- if (is_mc)
- ret = rte_ring_mc_dequeue_bulk(mp->ring, obj_table, n);
- else
- ret = rte_ring_sc_dequeue_bulk(mp->ring, obj_table, n);
+ ret = rte_mempool_ops_dequeue_bulk(mp, obj_table, n);
if (ret < 0)
__MEMPOOL_STAT_ADD(mp, get_fail, n);
diff --git a/lib/librte_mempool/rte_mempool_ops.c b/lib/librte_mempool/rte_mempool_ops.c
new file mode 100644
index 0000000..7977a14
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ops.c
@@ -0,0 +1,150 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Copyright(c) 2016 6WIND S.A.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_mempool.h>
+#include <rte_errno.h>
+
+/* indirect jump table to support external memory pools. */
+struct rte_mempool_ops_table rte_mempool_ops_table = {
+ .sl = RTE_SPINLOCK_INITIALIZER,
+ .num_ops = 0
+};
+
+/* add a new ops struct in rte_mempool_ops_table, return its index. */
+int
+rte_mempool_ops_register(const struct rte_mempool_ops *h)
+{
+ struct rte_mempool_ops *ops;
+ int16_t ops_index;
+
+ rte_spinlock_lock(&rte_mempool_ops_table.sl);
+
+ if (rte_mempool_ops_table.num_ops >=
+ RTE_MEMPOOL_MAX_OPS_IDX) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Maximum number of mempool ops structs exceeded\n");
+ return -ENOSPC;
+ }
+
+ if (h->alloc == NULL || h->enqueue == NULL ||
+ h->dequeue == NULL || h->get_count == NULL) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Missing callback while registering mempool ops\n");
+ return -EINVAL;
+ }
+
+ if (strlen(h->name) >= sizeof(ops->name) - 1) {
+ RTE_LOG(DEBUG, EAL, "%s(): mempool_ops <%s>: name too long\n",
+ __func__, h->name);
+ rte_errno = EEXIST;
+ return -EEXIST;
+ }
+
+ ops_index = rte_mempool_ops_table.num_ops++;
+ ops = &rte_mempool_ops_table.ops[ops_index];
+ snprintf(ops->name, sizeof(ops->name), "%s", h->name);
+ ops->alloc = h->alloc;
+ ops->enqueue = h->enqueue;
+ ops->dequeue = h->dequeue;
+ ops->get_count = h->get_count;
+
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+
+ return ops_index;
+}
+
+/* wrapper to allocate an external mempool's private (pool) data. */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->alloc(mp);
+}
+
+/* wrapper to free an external pool ops. */
+void
+rte_mempool_ops_free(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ if (ops->free == NULL)
+ return;
+ return ops->free(mp);
+}
+
+/* wrapper to get available objects in an external mempool. */
+unsigned int
+rte_mempool_ops_get_count(const struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->get_count(mp);
+}
+
+/* sets mempool ops previously registered by rte_mempool_ops_register. */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config)
+{
+ struct rte_mempool_ops *ops = NULL;
+ unsigned i;
+
+ /* too late, the mempool is already populated. */
+ if (mp->flags & MEMPOOL_F_POOL_CREATED)
+ return -EEXIST;
+
+ for (i = 0; i < rte_mempool_ops_table.num_ops; i++) {
+ if (!strcmp(name,
+ rte_mempool_ops_table.ops[i].name)) {
+ ops = &rte_mempool_ops_table.ops[i];
+ break;
+ }
+ }
+
+ if (ops == NULL)
+ return -EINVAL;
+
+ mp->ops_index = i;
+ mp->pool_config = pool_config;
+ return 0;
+}
diff --git a/lib/librte_mempool/rte_mempool_ring.c b/lib/librte_mempool/rte_mempool_ring.c
new file mode 100644
index 0000000..b9aa64d
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ring.c
@@ -0,0 +1,161 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_errno.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+
+static int
+common_ring_mp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_mp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_sp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_mc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_mc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_sc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static unsigned
+common_ring_get_count(const struct rte_mempool *mp)
+{
+ return rte_ring_count(mp->pool_data);
+}
+
+
+static int
+common_ring_alloc(struct rte_mempool *mp)
+{
+ int rg_flags = 0, ret;
+ char rg_name[RTE_RING_NAMESIZE];
+ struct rte_ring *r;
+
+ ret = snprintf(rg_name, sizeof(rg_name),
+ RTE_MEMPOOL_MZ_FORMAT, mp->name);
+ if (ret < 0 || ret >= (int)sizeof(rg_name)) {
+ rte_errno = ENAMETOOLONG;
+ return -rte_errno;
+ }
+
+ /* ring flags */
+ if (mp->flags & MEMPOOL_F_SP_PUT)
+ rg_flags |= RING_F_SP_ENQ;
+ if (mp->flags & MEMPOOL_F_SC_GET)
+ rg_flags |= RING_F_SC_DEQ;
+
+ /*
+ * Allocate the ring that will be used to store objects.
+ * Ring functions will return appropriate errors if we are
+ * running as a secondary process etc., so no checks made
+ * in this function for that condition.
+ */
+ r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
+ mp->socket_id, rg_flags);
+ if (r == NULL)
+ return -rte_errno;
+
+ mp->pool_data = r;
+
+ return 0;
+}
+
+static void
+common_ring_free(struct rte_mempool *mp)
+{
+ rte_ring_free(mp->pool_data);
+}
+
+/*
+ * The following 4 declarations of mempool ops structs address
+ * the need for the backward compatible mempool handlers for
+ * single/multi producers and single/multi consumers as dictated by the
+ * flags provided to the rte_mempool_create function
+ */
+static const struct rte_mempool_ops ops_mp_mc = {
+ .name = "ring_mp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_sc = {
+ .name = "ring_sp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_mp_sc = {
+ .name = "ring_mp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_mc = {
+ .name = "ring_sp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+MEMPOOL_REGISTER_OPS(ops_mp_mc);
+MEMPOOL_REGISTER_OPS(ops_sp_sc);
+MEMPOOL_REGISTER_OPS(ops_mp_sc);
+MEMPOOL_REGISTER_OPS(ops_sp_mc);
diff --git a/lib/librte_mempool/rte_mempool_version.map b/lib/librte_mempool/rte_mempool_version.map
index f63461b..6209ec2 100644
--- a/lib/librte_mempool/rte_mempool_version.map
+++ b/lib/librte_mempool/rte_mempool_version.map
@@ -20,15 +20,18 @@ DPDK_16.7 {
global:
rte_mempool_check_cookies;
- rte_mempool_obj_iter;
- rte_mempool_mem_iter;
rte_mempool_create_empty;
+ rte_mempool_free;
+ rte_mempool_mem_iter;
+ rte_mempool_obj_iter;
+ rte_mempool_ops_register;
+ rte_mempool_ops_table;
+ rte_mempool_populate_anon;
+ rte_mempool_populate_default;
rte_mempool_populate_phys;
rte_mempool_populate_phys_tab;
rte_mempool_populate_virt;
- rte_mempool_populate_default;
- rte_mempool_populate_anon;
- rte_mempool_free;
+ rte_mempool_set_ops_byname;
local: *;
} DPDK_2.0;
--
2.5.5
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v15 0/3] mempool: add mempool handler feature
2016-06-17 13:53 3% ` [dpdk-dev] [PATCH v14 0/3] mempool: add mempool handler feature David Hunt
2016-06-17 13:53 1% ` [dpdk-dev] [PATCH v14 1/3] mempool: support mempool handler operations David Hunt
@ 2016-06-19 12:05 3% ` David Hunt
2016-06-19 12:05 1% ` [dpdk-dev] [PATCH v15 1/3] mempool: support mempool handler operations David Hunt
2016-06-22 9:27 3% ` [dpdk-dev] [PATCH v16 0/3] mempool: add mempool handler feature David Hunt
1 sibling, 2 replies; 200+ results
From: David Hunt @ 2016-06-19 12:05 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the Mempool Handler feature patch set.
It's re-based on top of the latest head as of 19/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v15 changes:
* Changed rte_mempool_ops_get() to rte_mempool_get_ops()
* Did some minor tweaks to comments after the previous change of function
names from put/get to enqueue/dequeue
* Added missing spinlock_unlock in rte_mempool_ops_register()
* Added check for null in ops_free
* removed unneeded return statement
v14 changes:
* set MEMPOOL_F_RING_CREATED flag after rte_mempool_ring_create() is called.
* Changed name of feature from "external mempool manager" to "mempool handler"
and updated comments and release notes accordingly.
* Added a comment for newly added pool_config param in
rte_mempool_set_ops_byname.
v13 changes:
* Added in extra opaque data (pool_config) to mempool struct for mempool
configuration by the ops functions. For example, this can be used to pass
device names or device flags to the underlying alloc function.
* Added mempool_config param to rte_mempool_set_ops_byname()
v12 changes:
* Fixed a comment (function pram h -> ops)
* fixed a typo (callbacki)
v11 changes:
* Fixed comments (added '.' where needed for consistency)
* removed ABI breakage notice for mempool manager in deprecation.rst
* Added description of the external mempool manager functionality to
doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
* renamed rte_mempool_default.c to rte_mempool_ring.c
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The Mempool Handler feature is an extension to the mempool API that allows
users to add and use an alternative mempool handler, which allows
external memory subsystems such as external hardware memory management
systems and software based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool handler will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing.
There are two aspects to mempool handlers.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops_byname to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempools ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several mempool handlers may be used in the same application. A new
mempool can then be created by using the new rte_mempool_create_empty function,
then calling rte_mempool_set_ops_byname to point the mempool to the relevant
mempool handler callback (ops) structure.
Legacy applications will continue to use the old rte_mempool_create API call,
which uses a ring based mempool handler by default. These applications
will need to be modified to use a new mempool handler.
A mempool handler needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name
void *pool_config);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool handler using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support mempool handler operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test mempool handler
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [RFC] librte_vhost: Add unix domain socket fd registration
@ 2016-06-17 15:32 3% Aaron Conole
0 siblings, 0 replies; 200+ results
From: Aaron Conole @ 2016-06-17 15:32 UTC (permalink / raw)
To: dev, Huawei Xie, Yuanhan Liu
Prior to this commit, the only way to add a vhost-user socket to the
system is by relying on librte_vhost to open the unix domain socket and
add it to the unix socket list. This is problematic for applications
which would like to set the permissions, or applications which are not
directly allowed to open sockets due to policy restrictions.
This patch provides a new API and ABI to allow application developers to
acquire the unix domain socket via whatever mechanism fits and pass it
to the vhost driver registration process.
Signed-off-by: Aaron Conole <aconole@redhat.com>
---
doc/guides/prog_guide/vhost_lib.rst | 8 +++++
lib/librte_vhost/rte_vhost_version.map | 6 ++++
lib/librte_vhost/rte_virtio_net.h | 6 ++++
lib/librte_vhost/vhost_user/vhost-net-user.c | 47 ++++++++++++++++++----------
4 files changed, 50 insertions(+), 17 deletions(-)
diff --git a/doc/guides/prog_guide/vhost_lib.rst b/doc/guides/prog_guide/vhost_lib.rst
index 48e1fff..22d0c6d 100644
--- a/doc/guides/prog_guide/vhost_lib.rst
+++ b/doc/guides/prog_guide/vhost_lib.rst
@@ -49,6 +49,14 @@ Vhost API Overview
For vhost-user, a Unix domain socket server will be created with the parameter as
the local socket path.
+ Alternately, rte_vhost_driver_register_socket registers a unix domain
+ socket into the system.
+ This socket descriptor should be acquired by the host application through
+ some mechanism (either fd passing or by performing the unix domain socket
+ allocation).
+ The file descriptor passed in this way must still be a Unix domain socket
+ server.
+
* Vhost session start
rte_vhost_driver_session_start starts the vhost session loop.
diff --git a/lib/librte_vhost/rte_vhost_version.map b/lib/librte_vhost/rte_vhost_version.map
index 3d8709e..fe58967 100644
--- a/lib/librte_vhost/rte_vhost_version.map
+++ b/lib/librte_vhost/rte_vhost_version.map
@@ -20,3 +20,9 @@ DPDK_2.1 {
rte_vhost_driver_unregister;
} DPDK_2.0;
+
+DPDK_16.7 {
+ global:
+
+ rte_vhost_driver_register_socket;
+} DPDK_2.1;
diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
index 600b20b..d2959ff 100644
--- a/lib/librte_vhost/rte_virtio_net.h
+++ b/lib/librte_vhost/rte_virtio_net.h
@@ -236,6 +236,12 @@ int rte_vhost_enable_guest_notification(struct virtio_net *dev, uint16_t queue_i
/* Register vhost driver. dev_name could be different for multiple instance support. */
int rte_vhost_driver_register(const char *dev_name);
+/* Register vhost driver using the provided unix domain socket. The socket MUST
+ * already be fully created and in a listening state (by calling listen()).
+ */
+int rte_vhost_driver_register_socket(const char *dev_name,
+ int vhost_unix_socket);
+
/* Unregister vhost driver. This is only meaningful to vhost user. */
int rte_vhost_driver_unregister(const char *dev_name);
diff --git a/lib/librte_vhost/vhost_user/vhost-net-user.c b/lib/librte_vhost/vhost_user/vhost-net-user.c
index df2bd64..0fe72db 100644
--- a/lib/librte_vhost/vhost_user/vhost-net-user.c
+++ b/lib/librte_vhost/vhost_user/vhost-net-user.c
@@ -446,45 +446,58 @@ vserver_message_handler(int connfd, void *dat, int *remove)
}
}
+
/**
- * Creates and initialise the vhost server.
+ * Appends a socket to the vhost server polling list
*/
int
-rte_vhost_driver_register(const char *path)
+rte_vhost_driver_register_socket(const char *dev_name, int vhost_unix_socket)
{
struct vhost_server *vserver;
- pthread_mutex_lock(&g_vhost_server.server_mutex);
-
- if (g_vhost_server.vserver_cnt == MAX_VHOST_SERVER) {
- RTE_LOG(ERR, VHOST_CONFIG,
- "error: the number of servers reaches maximum\n");
- pthread_mutex_unlock(&g_vhost_server.server_mutex);
- return -1;
- }
-
vserver = calloc(sizeof(struct vhost_server), 1);
if (vserver == NULL) {
- pthread_mutex_unlock(&g_vhost_server.server_mutex);
return -1;
}
- vserver->listenfd = uds_socket(path);
- if (vserver->listenfd < 0) {
+ vserver->listenfd = vhost_unix_socket;
+ vserver->path = strdup(dev_name);
+ if (!vserver->path) {
free(vserver);
- pthread_mutex_unlock(&g_vhost_server.server_mutex);
return -1;
}
- vserver->path = strdup(path);
+ pthread_mutex_lock(&g_vhost_server.server_mutex);
+
+ if (g_vhost_server.vserver_cnt == MAX_VHOST_SERVER) {
+ RTE_LOG(ERR, VHOST_CONFIG,
+ "error: the number of servers reaches maximum\n");
+ pthread_mutex_unlock(&g_vhost_server.server_mutex);
+ free(vserver->path);
+ free(vserver);
+ return -1;
+ }
fdset_add(&g_vhost_server.fdset, vserver->listenfd,
vserver_new_vq_conn, NULL, vserver);
g_vhost_server.server[g_vhost_server.vserver_cnt++] = vserver;
pthread_mutex_unlock(&g_vhost_server.server_mutex);
+}
- return 0;
+
+/**
+ * Creates and initialise the vhost server.
+ */
+int
+rte_vhost_driver_register(const char *dev_name)
+{
+
+ int listenfd = uds_socket(dev_name);
+ if (listenfd < 0)
+ return -1;
+
+ return rte_vhost_driver_register_socket(dev_name, listenfd);
}
--
2.5.5
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v14 1/3] mempool: support mempool handler operations
2016-06-17 13:53 3% ` [dpdk-dev] [PATCH v14 0/3] mempool: add mempool handler feature David Hunt
@ 2016-06-17 13:53 1% ` David Hunt
2016-06-19 12:05 3% ` [dpdk-dev] [PATCH v15 0/3] mempool: add mempool handler feature David Hunt
1 sibling, 0 replies; 200+ results
From: David Hunt @ 2016-06-17 13:53 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain, David Hunt
Until now, the objects stored in a mempool were internally stored in a
ring. This patch introduces the possibility to register external handlers
replacing the ring.
The default behavior remains unchanged, but calling the new function
rte_mempool_set_ops_byname() right after rte_mempool_create_empty() allows
the user to change the handler that will be used when populating
the mempool.
This patch also adds a set of default ops (function callbacks) based
on rte_ring.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---
app/test/test_mempool_perf.c | 1 -
doc/guides/prog_guide/mempool_lib.rst | 32 +++-
doc/guides/rel_notes/deprecation.rst | 9 -
lib/librte_mempool/Makefile | 2 +
lib/librte_mempool/rte_mempool.c | 67 +++-----
lib/librte_mempool/rte_mempool.h | 255 ++++++++++++++++++++++++++---
lib/librte_mempool/rte_mempool_ops.c | 150 +++++++++++++++++
lib/librte_mempool/rte_mempool_ring.c | 161 ++++++++++++++++++
lib/librte_mempool/rte_mempool_version.map | 13 +-
9 files changed, 609 insertions(+), 81 deletions(-)
create mode 100644 lib/librte_mempool/rte_mempool_ops.c
create mode 100644 lib/librte_mempool/rte_mempool_ring.c
diff --git a/app/test/test_mempool_perf.c b/app/test/test_mempool_perf.c
index c5e3576..c5f8455 100644
--- a/app/test/test_mempool_perf.c
+++ b/app/test/test_mempool_perf.c
@@ -161,7 +161,6 @@ per_lcore_mempool_test(__attribute__((unused)) void *arg)
n_get_bulk);
if (unlikely(ret < 0)) {
rte_mempool_dump(stdout, mp);
- rte_ring_dump(stdout, mp->ring);
/* in this case, objects are lost... */
return -1;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index c3afc2e..1943fc4 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -34,7 +34,8 @@ Mempool Library
===============
A memory pool is an allocator of a fixed-sized object.
-In the DPDK, it is identified by name and uses a ring to store free objects.
+In the DPDK, it is identified by name and uses a mempool handler to store free objects.
+The default mempool handler is ring based.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
@@ -127,6 +128,35 @@ The maximum size of the cache is static and is defined at compilation time (CONF
A mempool in Memory with its Associated Ring
+Mempool Handlers
+------------------------
+
+This allows external memory subsystems, such as external hardware memory
+management systems and software based memory allocators, to be used with DPDK.
+
+There are two aspects to a mempool handler.
+
+* Adding the code for your new mempool operations (ops). This is achieved by
+ adding a new mempool ops code, and using the ``REGISTER_MEMPOOL_OPS`` macro.
+
+* Using the new API to call ``rte_mempool_create_empty()`` and
+ ``rte_mempool_set_ops_byname()`` to create a new mempool and specifying which
+ ops to use.
+
+Several different mempool handlers may be used in the same application. A new
+mempool can be created by using the ``rte_mempool_create_empty()`` function,
+then using ``rte_mempool_set_ops_byname()`` to point the mempool to the
+relevant mempool handler callback (ops) structure.
+
+Legacy applications may continue to use the old ``rte_mempool_create()`` API
+call, which uses a ring based mempool handler by default. These applications
+will need to be modified to use a new mempool handler.
+
+For applications that use ``rte_pktmbuf_create()``, there is a config setting
+(``RTE_MBUF_DEFAULT_MEMPOOL_OPS``) that allows the application to make use of
+an alternative mempool handler.
+
+
Use Cases
---------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index f75183f..3cbc19e 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -34,15 +34,6 @@ Deprecation Notices
compact API. The ones that remain are backwards compatible and use the
per-lcore default cache if available. This change targets release 16.07.
-* The rte_mempool struct will be changed in 16.07 to facilitate the new
- external mempool manager functionality.
- The ring element will be replaced with a more generic 'pool' opaque pointer
- to allow new mempool handlers to use their own user-defined mempool
- layout. Also newly added to rte_mempool is a handler index.
- The existing API will be backward compatible, but there will be new API
- functions added to facilitate the creation of mempools using an external
- handler. The 16.07 release will contain these changes.
-
* A librte_vhost public structures refactor is planned for DPDK 16.07
that requires both ABI and API change.
The proposed refactor would expose DPDK vhost dev to applications as
diff --git a/lib/librte_mempool/Makefile b/lib/librte_mempool/Makefile
index 43423e0..a4c089e 100644
--- a/lib/librte_mempool/Makefile
+++ b/lib/librte_mempool/Makefile
@@ -42,6 +42,8 @@ LIBABIVER := 2
# all source are stored in SRCS-y
SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ring.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_MEMPOOL)-include := rte_mempool.h
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index af71edd..e6a83d0 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -148,7 +148,7 @@ mempool_add_elem(struct rte_mempool *mp, void *obj, phys_addr_t physaddr)
#endif
/* enqueue in ring */
- rte_ring_sp_enqueue(mp->ring, obj);
+ rte_mempool_ops_enqueue_bulk(mp, &obj, 1);
}
/* call obj_cb() for each mempool element */
@@ -303,40 +303,6 @@ rte_mempool_xmem_usage(__rte_unused void *vaddr, uint32_t elt_num,
return (size_t)paddr_idx << pg_shift;
}
-/* create the internal ring */
-static int
-rte_mempool_ring_create(struct rte_mempool *mp)
-{
- int rg_flags = 0, ret;
- char rg_name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
-
- ret = snprintf(rg_name, sizeof(rg_name),
- RTE_MEMPOOL_MZ_FORMAT, mp->name);
- if (ret < 0 || ret >= (int)sizeof(rg_name))
- return -ENAMETOOLONG;
-
- /* ring flags */
- if (mp->flags & MEMPOOL_F_SP_PUT)
- rg_flags |= RING_F_SP_ENQ;
- if (mp->flags & MEMPOOL_F_SC_GET)
- rg_flags |= RING_F_SC_DEQ;
-
- /* Allocate the ring that will be used to store objects.
- * Ring functions will return appropriate errors if we are
- * running as a secondary process etc., so no checks made
- * in this function for that condition.
- */
- r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
- mp->socket_id, rg_flags);
- if (r == NULL)
- return -rte_errno;
-
- mp->ring = r;
- mp->flags |= MEMPOOL_F_RING_CREATED;
- return 0;
-}
-
/* free a memchunk allocated with rte_memzone_reserve() */
static void
rte_mempool_memchunk_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
@@ -354,7 +320,7 @@ rte_mempool_free_memchunks(struct rte_mempool *mp)
void *elt;
while (!STAILQ_EMPTY(&mp->elt_list)) {
- rte_ring_sc_dequeue(mp->ring, &elt);
+ rte_mempool_ops_dequeue_bulk(mp, &elt, 1);
(void)elt;
STAILQ_REMOVE_HEAD(&mp->elt_list, next);
mp->populated_size--;
@@ -386,10 +352,11 @@ rte_mempool_populate_phys(struct rte_mempool *mp, char *vaddr,
int ret;
/* create the internal ring if not already done */
- if ((mp->flags & MEMPOOL_F_RING_CREATED) == 0) {
- ret = rte_mempool_ring_create(mp);
- if (ret < 0)
+ if ((mp->flags & MEMPOOL_F_POOL_CREATED) == 0) {
+ ret = rte_mempool_ops_alloc(mp);
+ if (ret != 0)
return ret;
+ mp->flags |= MEMPOOL_F_POOL_CREATED;
}
/* mempool is already populated */
@@ -703,7 +670,7 @@ rte_mempool_free(struct rte_mempool *mp)
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_mempool_free_memchunks(mp);
- rte_ring_free(mp->ring);
+ rte_mempool_ops_free(mp);
rte_memzone_free(mp->mz);
}
@@ -815,6 +782,7 @@ rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, 0));
te->data = mp;
+
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_INSERT_TAIL(mempool_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
@@ -844,6 +812,19 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
if (mp == NULL)
return NULL;
+ /*
+ * Since we have 4 combinations of the SP/SC/MP/MC examine the flags to
+ * set the correct index into the table of ops structs.
+ */
+ if (flags & (MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET))
+ rte_mempool_set_ops_byname(mp, "ring_sp_sc", NULL);
+ else if (flags & MEMPOOL_F_SP_PUT)
+ rte_mempool_set_ops_byname(mp, "ring_sp_mc", NULL);
+ else if (flags & MEMPOOL_F_SC_GET)
+ rte_mempool_set_ops_byname(mp, "ring_mp_sc", NULL);
+ else
+ rte_mempool_set_ops_byname(mp, "ring_mp_mc", NULL);
+
/* call the mempool priv initializer */
if (mp_init)
mp_init(mp, mp_init_arg);
@@ -930,7 +911,7 @@ rte_mempool_count(const struct rte_mempool *mp)
unsigned count;
unsigned lcore_id;
- count = rte_ring_count(mp->ring);
+ count = rte_mempool_ops_get_count(mp);
if (mp->cache_size == 0)
return count;
@@ -1119,7 +1100,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
fprintf(f, "mempool <%s>@%p\n", mp->name, mp);
fprintf(f, " flags=%x\n", mp->flags);
- fprintf(f, " ring=<%s>@%p\n", mp->ring->name, mp->ring);
+ fprintf(f, " pool=%p\n", mp->pool_data);
fprintf(f, " phys_addr=0x%" PRIx64 "\n", mp->mz->phys_addr);
fprintf(f, " nb_mem_chunks=%u\n", mp->nb_mem_chunks);
fprintf(f, " size=%"PRIu32"\n", mp->size);
@@ -1140,7 +1121,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
}
cache_count = rte_mempool_dump_cache(f, mp);
- common_count = rte_ring_count(mp->ring);
+ common_count = rte_mempool_ops_get_count(mp);
if ((cache_count + common_count) > mp->size)
common_count = mp->size - cache_count;
fprintf(f, " common_pool_count=%u\n", common_count);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 60339bd..2d7c980 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -67,6 +67,7 @@
#include <inttypes.h>
#include <sys/queue.h>
+#include <rte_spinlock.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_lcore.h>
@@ -203,10 +204,14 @@ struct rte_mempool_memhdr {
*/
struct rte_mempool {
char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
- struct rte_ring *ring; /**< Ring to store objects. */
- const struct rte_memzone *mz; /**< Memzone where pool is allocated */
+ union {
+ void *pool_data; /**< Ring or pool to store objects. */
+ uint64_t pool_id; /**< External mempool identifier. */
+ };
+ void *pool_config; /**< optional args for ops alloc. */
+ const struct rte_memzone *mz; /**< Memzone where pool is alloc'd. */
int flags; /**< Flags of the mempool. */
- int socket_id; /**< Socket id passed at mempool creation. */
+ int socket_id; /**< Socket id passed at create. */
uint32_t size; /**< Max size of the mempool. */
uint32_t cache_size; /**< Size of per-lcore local cache. */
uint32_t cache_flushthresh;
@@ -217,6 +222,14 @@ struct rte_mempool {
uint32_t trailer_size; /**< Size of trailer (after elt). */
unsigned private_data_size; /**< Size of private data. */
+ /**
+ * Index into rte_mempool_ops_table array of mempool ops
+ * structs, which contain callback function pointers.
+ * We're using an index here rather than pointers to the callbacks
+ * to facilitate any secondary processes that may want to use
+ * this mempool.
+ */
+ int32_t ops_index;
struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
@@ -235,7 +248,7 @@ struct rte_mempool {
#define MEMPOOL_F_NO_CACHE_ALIGN 0x0002 /**< Do not align objs on cache lines.*/
#define MEMPOOL_F_SP_PUT 0x0004 /**< Default put is "single-producer".*/
#define MEMPOOL_F_SC_GET 0x0008 /**< Default get is "single-consumer".*/
-#define MEMPOOL_F_RING_CREATED 0x0010 /**< Internal: ring is created */
+#define MEMPOOL_F_POOL_CREATED 0x0010 /**< Internal: pool is created. */
#define MEMPOOL_F_NO_PHYS_CONTIG 0x0020 /**< Don't need physically contiguous objs. */
/**
@@ -325,6 +338,215 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
#define __mempool_check_cookies(mp, obj_table_const, n, free) do {} while(0)
#endif /* RTE_LIBRTE_MEMPOOL_DEBUG */
+#define RTE_MEMPOOL_OPS_NAMESIZE 32 /**< Max length of ops struct name. */
+
+/**
+ * Prototype for implementation specific data provisioning function.
+ *
+ * The function should provide the implementation specific memory for
+ * for use by the other mempool ops functions in a given mempool ops struct.
+ * E.g. the default ops provides an instance of the rte_ring for this purpose.
+ * it will most likely point to a different type of data structure, and
+ * will be transparent to the application programmer.
+ * This function should set mp->pool_data.
+ */
+typedef int (*rte_mempool_alloc_t)(struct rte_mempool *mp);
+
+/**
+ * Free the opaque private data pointed to by mp->pool_data pointer.
+ */
+typedef void (*rte_mempool_free_t)(struct rte_mempool *mp);
+
+/**
+ * Enqueue an object into the external pool.
+ */
+typedef int (*rte_mempool_enqueue_t)(struct rte_mempool *mp,
+ void * const *obj_table, unsigned int n);
+
+/**
+ * Dequeue an object from the external pool.
+ */
+typedef int (*rte_mempool_dequeue_t)(struct rte_mempool *mp,
+ void **obj_table, unsigned int n);
+
+/**
+ * Return the number of available objects in the external pool.
+ */
+typedef unsigned (*rte_mempool_get_count)(const struct rte_mempool *mp);
+
+/** Structure defining mempool operations structure */
+struct rte_mempool_ops {
+ char name[RTE_MEMPOOL_OPS_NAMESIZE]; /**< Name of mempool ops struct. */
+ rte_mempool_alloc_t alloc; /**< Allocate private data. */
+ rte_mempool_free_t free; /**< Free the external pool. */
+ rte_mempool_enqueue_t enqueue; /**< Enqueue an object. */
+ rte_mempool_dequeue_t dequeue; /**< Dequeue an object. */
+ rte_mempool_get_count get_count; /**< Get qty of available objs. */
+} __rte_cache_aligned;
+
+#define RTE_MEMPOOL_MAX_OPS_IDX 16 /**< Max registered ops structs */
+
+/**
+ * Structure storing the table of registered ops structs, each of which contain
+ * the function pointers for the mempool ops functions.
+ * Each process has its own storage for this ops struct array so that
+ * the mempools can be shared across primary and secondary processes.
+ * The indices used to access the array are valid across processes, whereas
+ * any function pointers stored directly in the mempool struct would not be.
+ * This results in us simply having "ops_index" in the mempool struct.
+ */
+struct rte_mempool_ops_table {
+ rte_spinlock_t sl; /**< Spinlock for add/delete. */
+ uint32_t num_ops; /**< Number of used ops structs in the table. */
+ /**
+ * Storage for all possible ops structs.
+ */
+ struct rte_mempool_ops ops[RTE_MEMPOOL_MAX_OPS_IDX];
+} __rte_cache_aligned;
+
+/** Array of registered ops structs. */
+extern struct rte_mempool_ops_table rte_mempool_ops_table;
+
+/**
+ * @internal Get the mempool ops struct from its index.
+ *
+ * @param ops_index
+ * The index of the ops struct in the ops struct table. It must be a valid
+ * index: (0 <= idx < num_ops).
+ * @return
+ * The pointer to the ops struct in the table.
+ */
+static inline struct rte_mempool_ops *
+rte_mempool_ops_get(int ops_index)
+{
+ RTE_VERIFY(ops_index < RTE_MEMPOOL_MAX_OPS_IDX);
+
+ return &rte_mempool_ops_table.ops[ops_index];
+}
+
+/**
+ * @internal Wrapper for mempool_ops alloc callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * - 0: Success; successfully allocated mempool pool_data.
+ * - <0: Error; code of alloc function.
+ */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp);
+
+/**
+ * @internal Wrapper for mempool_ops get callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to get.
+ * @return
+ * - 0: Success; got n objects.
+ * - <0: Error; code of get function.
+ */
+static inline int
+rte_mempool_ops_dequeue_bulk(struct rte_mempool *mp,
+ void **obj_table, unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->dequeue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops put callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to put.
+ * @return
+ * - 0: Success; n objects supplied.
+ * - <0: Error; code of put function.
+ */
+static inline int
+rte_mempool_ops_enqueue_bulk(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->enqueue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops get_count callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * The number of available objects in the external pool.
+ */
+unsigned
+rte_mempool_ops_get_count(const struct rte_mempool *mp);
+
+/**
+ * @internal wrapper for mempool_ops free callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ */
+void
+rte_mempool_ops_free(struct rte_mempool *mp);
+
+/**
+ * Set the ops of a mempool.
+ *
+ * This can only be done on a mempool that is not populated, i.e. just after
+ * a call to rte_mempool_create_empty().
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param name
+ * Name of the ops structure to use for this mempool.
+ * @param pool_config
+ * Opaque data that can be passed by the application to the ops functions.
+ * @return
+ * - 0: Success; the mempool is now using the requested ops functions.
+ * - -EINVAL - Invalid ops struct name provided.
+ * - -EEXIST - mempool already has an ops struct assigned.
+ */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config);
+
+/**
+ * Register mempool operations.
+ *
+ * @param ops
+ * Pointer to an ops structure to register.
+ * @return
+ * - >=0: Success; return the index of the ops struct in the table.
+ * - -EINVAL - some missing callbacks while registering ops struct.
+ * - -ENOSPC - the maximum number of ops structs has been reached.
+ */
+int rte_mempool_ops_register(const struct rte_mempool_ops *ops);
+
+/**
+ * Macro to statically register the ops of a mempool handler.
+ * Note that the rte_mempool_ops_register fails silently here when
+ * more then RTE_MEMPOOL_MAX_OPS_IDX is registered.
+ */
+#define MEMPOOL_REGISTER_OPS(ops) \
+ void mp_hdlr_init_##ops(void); \
+ void __attribute__((constructor, used)) mp_hdlr_init_##ops(void)\
+ { \
+ rte_mempool_ops_register(&ops); \
+ }
+
/**
* An object callback function for mempool.
*
@@ -774,7 +996,7 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
cache->len += n;
if (cache->len >= flushthresh) {
- rte_ring_mp_enqueue_bulk(mp->ring, &cache->objs[cache_size],
+ rte_mempool_ops_enqueue_bulk(mp, &cache->objs[cache_size],
cache->len - cache_size);
cache->len = cache_size;
}
@@ -785,19 +1007,10 @@ ring_enqueue:
/* push remaining objects in ring */
#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
- if (is_mp) {
- if (rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
- else {
- if (rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
+ if (rte_mempool_ops_enqueue_bulk(mp, obj_table, n) < 0)
+ rte_panic("cannot put objects in mempool\n");
#else
- if (is_mp)
- rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n);
- else
- rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n);
+ rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
#endif
}
@@ -945,7 +1158,8 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
uint32_t req = n + (cache_size - cache->len);
/* How many do we require i.e. number to fill the cache + the request */
- ret = rte_ring_mc_dequeue_bulk(mp->ring, &cache->objs[cache->len], req);
+ ret = rte_mempool_ops_dequeue_bulk(mp,
+ &cache->objs[cache->len], req);
if (unlikely(ret < 0)) {
/*
* In the offchance that we are buffer constrained,
@@ -972,10 +1186,7 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
ring_dequeue:
/* get remaining objects from ring */
- if (is_mc)
- ret = rte_ring_mc_dequeue_bulk(mp->ring, obj_table, n);
- else
- ret = rte_ring_sc_dequeue_bulk(mp->ring, obj_table, n);
+ ret = rte_mempool_ops_dequeue_bulk(mp, obj_table, n);
if (ret < 0)
__MEMPOOL_STAT_ADD(mp, get_fail, n);
diff --git a/lib/librte_mempool/rte_mempool_ops.c b/lib/librte_mempool/rte_mempool_ops.c
new file mode 100644
index 0000000..7977a14
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ops.c
@@ -0,0 +1,150 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Copyright(c) 2016 6WIND S.A.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_mempool.h>
+#include <rte_errno.h>
+
+/* indirect jump table to support external memory pools. */
+struct rte_mempool_ops_table rte_mempool_ops_table = {
+ .sl = RTE_SPINLOCK_INITIALIZER,
+ .num_ops = 0
+};
+
+/* add a new ops struct in rte_mempool_ops_table, return its index. */
+int
+rte_mempool_ops_register(const struct rte_mempool_ops *h)
+{
+ struct rte_mempool_ops *ops;
+ int16_t ops_index;
+
+ rte_spinlock_lock(&rte_mempool_ops_table.sl);
+
+ if (rte_mempool_ops_table.num_ops >=
+ RTE_MEMPOOL_MAX_OPS_IDX) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Maximum number of mempool ops structs exceeded\n");
+ return -ENOSPC;
+ }
+
+ if (h->alloc == NULL || h->enqueue == NULL ||
+ h->dequeue == NULL || h->get_count == NULL) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Missing callback while registering mempool ops\n");
+ return -EINVAL;
+ }
+
+ if (strlen(h->name) >= sizeof(ops->name) - 1) {
+ RTE_LOG(DEBUG, EAL, "%s(): mempool_ops <%s>: name too long\n",
+ __func__, h->name);
+ rte_errno = EEXIST;
+ return -EEXIST;
+ }
+
+ ops_index = rte_mempool_ops_table.num_ops++;
+ ops = &rte_mempool_ops_table.ops[ops_index];
+ snprintf(ops->name, sizeof(ops->name), "%s", h->name);
+ ops->alloc = h->alloc;
+ ops->enqueue = h->enqueue;
+ ops->dequeue = h->dequeue;
+ ops->get_count = h->get_count;
+
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+
+ return ops_index;
+}
+
+/* wrapper to allocate an external mempool's private (pool) data. */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->alloc(mp);
+}
+
+/* wrapper to free an external pool ops. */
+void
+rte_mempool_ops_free(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ if (ops->free == NULL)
+ return;
+ return ops->free(mp);
+}
+
+/* wrapper to get available objects in an external mempool. */
+unsigned int
+rte_mempool_ops_get_count(const struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->get_count(mp);
+}
+
+/* sets mempool ops previously registered by rte_mempool_ops_register. */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config)
+{
+ struct rte_mempool_ops *ops = NULL;
+ unsigned i;
+
+ /* too late, the mempool is already populated. */
+ if (mp->flags & MEMPOOL_F_POOL_CREATED)
+ return -EEXIST;
+
+ for (i = 0; i < rte_mempool_ops_table.num_ops; i++) {
+ if (!strcmp(name,
+ rte_mempool_ops_table.ops[i].name)) {
+ ops = &rte_mempool_ops_table.ops[i];
+ break;
+ }
+ }
+
+ if (ops == NULL)
+ return -EINVAL;
+
+ mp->ops_index = i;
+ mp->pool_config = pool_config;
+ return 0;
+}
diff --git a/lib/librte_mempool/rte_mempool_ring.c b/lib/librte_mempool/rte_mempool_ring.c
new file mode 100644
index 0000000..b9aa64d
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ring.c
@@ -0,0 +1,161 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_errno.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+
+static int
+common_ring_mp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_mp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_sp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_mc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_mc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_sc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static unsigned
+common_ring_get_count(const struct rte_mempool *mp)
+{
+ return rte_ring_count(mp->pool_data);
+}
+
+
+static int
+common_ring_alloc(struct rte_mempool *mp)
+{
+ int rg_flags = 0, ret;
+ char rg_name[RTE_RING_NAMESIZE];
+ struct rte_ring *r;
+
+ ret = snprintf(rg_name, sizeof(rg_name),
+ RTE_MEMPOOL_MZ_FORMAT, mp->name);
+ if (ret < 0 || ret >= (int)sizeof(rg_name)) {
+ rte_errno = ENAMETOOLONG;
+ return -rte_errno;
+ }
+
+ /* ring flags */
+ if (mp->flags & MEMPOOL_F_SP_PUT)
+ rg_flags |= RING_F_SP_ENQ;
+ if (mp->flags & MEMPOOL_F_SC_GET)
+ rg_flags |= RING_F_SC_DEQ;
+
+ /*
+ * Allocate the ring that will be used to store objects.
+ * Ring functions will return appropriate errors if we are
+ * running as a secondary process etc., so no checks made
+ * in this function for that condition.
+ */
+ r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
+ mp->socket_id, rg_flags);
+ if (r == NULL)
+ return -rte_errno;
+
+ mp->pool_data = r;
+
+ return 0;
+}
+
+static void
+common_ring_free(struct rte_mempool *mp)
+{
+ rte_ring_free(mp->pool_data);
+}
+
+/*
+ * The following 4 declarations of mempool ops structs address
+ * the need for the backward compatible mempool handlers for
+ * single/multi producers and single/multi consumers as dictated by the
+ * flags provided to the rte_mempool_create function
+ */
+static const struct rte_mempool_ops ops_mp_mc = {
+ .name = "ring_mp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_sc = {
+ .name = "ring_sp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_mp_sc = {
+ .name = "ring_mp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_mc = {
+ .name = "ring_sp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+MEMPOOL_REGISTER_OPS(ops_mp_mc);
+MEMPOOL_REGISTER_OPS(ops_sp_sc);
+MEMPOOL_REGISTER_OPS(ops_mp_sc);
+MEMPOOL_REGISTER_OPS(ops_sp_mc);
diff --git a/lib/librte_mempool/rte_mempool_version.map b/lib/librte_mempool/rte_mempool_version.map
index f63461b..6209ec2 100644
--- a/lib/librte_mempool/rte_mempool_version.map
+++ b/lib/librte_mempool/rte_mempool_version.map
@@ -20,15 +20,18 @@ DPDK_16.7 {
global:
rte_mempool_check_cookies;
- rte_mempool_obj_iter;
- rte_mempool_mem_iter;
rte_mempool_create_empty;
+ rte_mempool_free;
+ rte_mempool_mem_iter;
+ rte_mempool_obj_iter;
+ rte_mempool_ops_register;
+ rte_mempool_ops_table;
+ rte_mempool_populate_anon;
+ rte_mempool_populate_default;
rte_mempool_populate_phys;
rte_mempool_populate_phys_tab;
rte_mempool_populate_virt;
- rte_mempool_populate_default;
- rte_mempool_populate_anon;
- rte_mempool_free;
+ rte_mempool_set_ops_byname;
local: *;
} DPDK_2.0;
--
2.5.5
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v14 0/3] mempool: add mempool handler feature
2016-06-16 12:30 3% ` [dpdk-dev] [PATCH v13 " David Hunt
2016-06-16 12:30 1% ` [dpdk-dev] [PATCH v13 1/3] mempool: support external mempool operations David Hunt
@ 2016-06-17 13:53 3% ` David Hunt
2016-06-17 13:53 1% ` [dpdk-dev] [PATCH v14 1/3] mempool: support mempool handler operations David Hunt
2016-06-19 12:05 3% ` [dpdk-dev] [PATCH v15 0/3] mempool: add mempool handler feature David Hunt
1 sibling, 2 replies; 200+ results
From: David Hunt @ 2016-06-17 13:53 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the Mempool Handler feature (previously
known as the External Mempool Manager.
It's re-based on top of the latest head as of 17/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v14 changes:
* set MEMPOOL_F_RING_CREATED flag after rte_mempool_ring_create() is called.
* Changed name of feature from "external mempool manager" to "mempool handler"
and updated comments and release notes accordingly.
* Added a comment for newly added pool_config param in
rte_mempool_set_ops_byname.
v13 changes:
* Added in extra opaque data (pool_config) to mempool struct for mempool
configuration by the ops functions. For example, this can be used to pass
device names or device flags to the underlying alloc function.
* Added mempool_config param to rte_mempool_set_ops_byname()
v12 changes:
* Fixed a comment (function pram h -> ops)
* fixed a typo (callbacki)
v11 changes:
* Fixed comments (added '.' where needed for consistency)
* removed ABI breakage notice for mempool manager in deprecation.rst
* Added description of the external mempool manager functionality to
doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
* renamed rte_mempool_default.c to rte_mempool_ring.c
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The Mempool Handler feature is an extension to the mempool API that allows
users to add and use an alternative mempool handler, which allows
external memory subsystems such as external hardware memory management
systems and software based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool handler will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing.
There are two aspects to mempool handlers.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops_byname to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempools ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several mempool handlers may be used in the same application. A new
mempool can then be created by using the new rte_mempool_create_empty function,
then calling rte_mempool_set_ops_byname to point the mempool to the relevant
mempool handler callback (ops) structure.
Legacy applications will continue to use the old rte_mempool_create API call,
which uses a ring based mempool handler by default. These applications
will need to be modified to use a new mempool handler.
A mempool handler needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name
void *pool_config);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool handler using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support mempool handler operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test mempool handler
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v10 3/7] ethdev: add new fields to ethdev info struct
2016-06-15 14:06 10% ` [dpdk-dev] [PATCH v10 3/7] ethdev: add new fields to ethdev info struct Reshma Pattan
@ 2016-06-16 19:14 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-16 19:14 UTC (permalink / raw)
To: Reshma Pattan; +Cc: dev
2016-06-15 15:06, Reshma Pattan:
> The new fields nb_rx_queues and nb_tx_queues are added to the
> rte_eth_dev_info structure.
> Changes to API rte_eth_dev_info_get() are done to update these new fields
> to the rte_eth_dev_info object.
The ABI is changed, not the API.
> Release notes is updated with the changes.
[...]
> --- a/lib/librte_ether/rte_ether_version.map
> +++ b/lib/librte_ether/rte_ether_version.map
> @@ -137,4 +137,5 @@ DPDK_16.07 {
> global:
>
> rte_eth_add_first_rx_callback;
> + rte_eth_dev_info_get;
> } DPDK_16.04;
Why duplicating this symbol in 16.07?
The ABI is broken anyway.
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH v3 04/17] eal: remove duplicate function declaration
@ 2016-06-16 14:06 3% ` Shreyansh Jain
0 siblings, 0 replies; 200+ results
From: Shreyansh Jain @ 2016-06-16 14:06 UTC (permalink / raw)
To: dev; +Cc: viktorin, thomas.monjalon, David Marchand
From: David Marchand <david.marchand@6wind.com>
rte_eal_dev_init is declared in both eal_private.h and rte_dev.h since its
introduction.
This function has been exported in ABI, so remove it from eal_private.h
Fixes: e57f20e05177 ("eal: make vdev init path generic for both virtual and pci devices")
Signed-off-by: David Marchand <david.marchand@6wind.com>
---
lib/librte_eal/common/eal_private.h | 7 -------
lib/librte_eal/linuxapp/eal/eal.c | 1 +
2 files changed, 1 insertion(+), 7 deletions(-)
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 857dc3e..06a68f6 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -259,13 +259,6 @@ int rte_eal_intr_init(void);
int rte_eal_alarm_init(void);
/**
- * This function initialises any virtual devices
- *
- * This function is private to the EAL.
- */
-int rte_eal_dev_init(void);
-
-/**
* Function is to check if the kernel module(like, vfio, vfio_iommu_type1,
* etc.) loaded.
*
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index bba8fea..5ec3d4e 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -70,6 +70,7 @@
#include <rte_cpuflags.h>
#include <rte_interrupts.h>
#include <rte_pci.h>
+#include <rte_dev.h>
#include <rte_devargs.h>
#include <rte_common.h>
#include <rte_version.h>
--
2.7.4
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v13 1/3] mempool: support external mempool operations
2016-06-16 12:30 3% ` [dpdk-dev] [PATCH v13 " David Hunt
@ 2016-06-16 12:30 1% ` David Hunt
2016-06-17 13:53 3% ` [dpdk-dev] [PATCH v14 0/3] mempool: add mempool handler feature David Hunt
1 sibling, 0 replies; 200+ results
From: David Hunt @ 2016-06-16 12:30 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain, David Hunt
Until now, the objects stored in a mempool were internally stored in a
ring. This patch introduces the possibility to register external handlers
replacing the ring.
The default behavior remains unchanged, but calling the new function
rte_mempool_set_ops_byname() right after rte_mempool_create_empty() allows
the user to change the handler that will be used when populating
the mempool.
This patch also adds a set of default ops (function callbacks) based
on rte_ring.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---
app/test/test_mempool_perf.c | 1 -
doc/guides/prog_guide/mempool_lib.rst | 31 +++-
doc/guides/rel_notes/deprecation.rst | 9 -
lib/librte_mempool/Makefile | 2 +
lib/librte_mempool/rte_mempool.c | 66 +++-----
lib/librte_mempool/rte_mempool.h | 253 ++++++++++++++++++++++++++---
lib/librte_mempool/rte_mempool_ops.c | 150 +++++++++++++++++
lib/librte_mempool/rte_mempool_ring.c | 161 ++++++++++++++++++
lib/librte_mempool/rte_mempool_version.map | 13 +-
9 files changed, 605 insertions(+), 81 deletions(-)
create mode 100644 lib/librte_mempool/rte_mempool_ops.c
create mode 100644 lib/librte_mempool/rte_mempool_ring.c
diff --git a/app/test/test_mempool_perf.c b/app/test/test_mempool_perf.c
index c5e3576..c5f8455 100644
--- a/app/test/test_mempool_perf.c
+++ b/app/test/test_mempool_perf.c
@@ -161,7 +161,6 @@ per_lcore_mempool_test(__attribute__((unused)) void *arg)
n_get_bulk);
if (unlikely(ret < 0)) {
rte_mempool_dump(stdout, mp);
- rte_ring_dump(stdout, mp->ring);
/* in this case, objects are lost... */
return -1;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index c3afc2e..2e3116e 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -34,7 +34,7 @@ Mempool Library
===============
A memory pool is an allocator of a fixed-sized object.
-In the DPDK, it is identified by name and uses a ring to store free objects.
+In the DPDK, it is identified by name and uses a ring or an external mempool manager to store free objects.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
@@ -127,6 +127,35 @@ The maximum size of the cache is static and is defined at compilation time (CONF
A mempool in Memory with its Associated Ring
+External Mempool Manager
+------------------------
+
+This allows external memory subsystems, such as external hardware memory
+management systems and software based memory allocators, to be used with DPDK.
+
+There are two aspects to external mempool manager.
+
+* Adding the code for your new mempool operations (ops). This is achieved by
+ adding a new mempool ops code, and using the ``REGISTER_MEMPOOL_OPS`` macro.
+
+* Using the new API to call ``rte_mempool_create_empty()`` and
+ ``rte_mempool_set_ops_byname()`` to create a new mempool and specifying which
+ ops to use.
+
+Several external mempool managers may be used in the same application. A new
+mempool can be created by using the ``rte_mempool_create_empty()`` function,
+then using ``rte_mempool_set_ops_byname()`` to point the mempool to the
+relevant mempool manager callback (ops) structure.
+
+Legacy applications may continue to use the old ``rte_mempool_create()`` API
+call, which uses a ring based mempool manager by default. These applications
+will need to be modified to use a new external mempool manager.
+
+For applications that use ``rte_pktmbuf_create()``, there is a config setting
+(``RTE_MBUF_DEFAULT_MEMPOOL_OPS``) that allows the application to make use of
+an external mempool manager.
+
+
Use Cases
---------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 7d947ae..c415095 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -39,15 +39,6 @@ Deprecation Notices
compact API. The ones that remain are backwards compatible and use the
per-lcore default cache if available. This change targets release 16.07.
-* The rte_mempool struct will be changed in 16.07 to facilitate the new
- external mempool manager functionality.
- The ring element will be replaced with a more generic 'pool' opaque pointer
- to allow new mempool handlers to use their own user-defined mempool
- layout. Also newly added to rte_mempool is a handler index.
- The existing API will be backward compatible, but there will be new API
- functions added to facilitate the creation of mempools using an external
- handler. The 16.07 release will contain these changes.
-
* A librte_vhost public structures refactor is planned for DPDK 16.07
that requires both ABI and API change.
The proposed refactor would expose DPDK vhost dev to applications as
diff --git a/lib/librte_mempool/Makefile b/lib/librte_mempool/Makefile
index 43423e0..a4c089e 100644
--- a/lib/librte_mempool/Makefile
+++ b/lib/librte_mempool/Makefile
@@ -42,6 +42,8 @@ LIBABIVER := 2
# all source are stored in SRCS-y
SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ring.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_MEMPOOL)-include := rte_mempool.h
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index 22a5645..0fb84ad 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -148,7 +148,7 @@ mempool_add_elem(struct rte_mempool *mp, void *obj, phys_addr_t physaddr)
#endif
/* enqueue in ring */
- rte_ring_sp_enqueue(mp->ring, obj);
+ rte_mempool_ops_enqueue_bulk(mp, &obj, 1);
}
/* call obj_cb() for each mempool element */
@@ -303,40 +303,6 @@ rte_mempool_xmem_usage(__rte_unused void *vaddr, uint32_t elt_num,
return (size_t)paddr_idx << pg_shift;
}
-/* create the internal ring */
-static int
-rte_mempool_ring_create(struct rte_mempool *mp)
-{
- int rg_flags = 0, ret;
- char rg_name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
-
- ret = snprintf(rg_name, sizeof(rg_name),
- RTE_MEMPOOL_MZ_FORMAT, mp->name);
- if (ret < 0 || ret >= (int)sizeof(rg_name))
- return -ENAMETOOLONG;
-
- /* ring flags */
- if (mp->flags & MEMPOOL_F_SP_PUT)
- rg_flags |= RING_F_SP_ENQ;
- if (mp->flags & MEMPOOL_F_SC_GET)
- rg_flags |= RING_F_SC_DEQ;
-
- /* Allocate the ring that will be used to store objects.
- * Ring functions will return appropriate errors if we are
- * running as a secondary process etc., so no checks made
- * in this function for that condition.
- */
- r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
- mp->socket_id, rg_flags);
- if (r == NULL)
- return -rte_errno;
-
- mp->ring = r;
- mp->flags |= MEMPOOL_F_RING_CREATED;
- return 0;
-}
-
/* free a memchunk allocated with rte_memzone_reserve() */
static void
rte_mempool_memchunk_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
@@ -354,7 +320,7 @@ rte_mempool_free_memchunks(struct rte_mempool *mp)
void *elt;
while (!STAILQ_EMPTY(&mp->elt_list)) {
- rte_ring_sc_dequeue(mp->ring, &elt);
+ rte_mempool_ops_dequeue_bulk(mp, &elt, 1);
(void)elt;
STAILQ_REMOVE_HEAD(&mp->elt_list, next);
mp->populated_size--;
@@ -386,9 +352,9 @@ rte_mempool_populate_phys(struct rte_mempool *mp, char *vaddr,
int ret;
/* create the internal ring if not already done */
- if ((mp->flags & MEMPOOL_F_RING_CREATED) == 0) {
- ret = rte_mempool_ring_create(mp);
- if (ret < 0)
+ if ((mp->flags & MEMPOOL_F_POOL_CREATED) == 0) {
+ ret = rte_mempool_ops_alloc(mp);
+ if (ret != 0)
return ret;
}
@@ -703,7 +669,7 @@ rte_mempool_free(struct rte_mempool *mp)
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_mempool_free_memchunks(mp);
- rte_ring_free(mp->ring);
+ rte_mempool_ops_free(mp);
rte_memzone_free(mp->mz);
}
@@ -815,6 +781,7 @@ rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, 0));
te->data = mp;
+
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_INSERT_TAIL(mempool_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
@@ -844,6 +811,19 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
if (mp == NULL)
return NULL;
+ /*
+ * Since we have 4 combinations of the SP/SC/MP/MC examine the flags to
+ * set the correct index into the table of ops structs.
+ */
+ if (flags & (MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET))
+ rte_mempool_set_ops_byname(mp, "ring_sp_sc", NULL);
+ else if (flags & MEMPOOL_F_SP_PUT)
+ rte_mempool_set_ops_byname(mp, "ring_sp_mc", NULL);
+ else if (flags & MEMPOOL_F_SC_GET)
+ rte_mempool_set_ops_byname(mp, "ring_mp_sc", NULL);
+ else
+ rte_mempool_set_ops_byname(mp, "ring_mp_mc", NULL);
+
/* call the mempool priv initializer */
if (mp_init)
mp_init(mp, mp_init_arg);
@@ -930,7 +910,7 @@ rte_mempool_count(const struct rte_mempool *mp)
unsigned count;
unsigned lcore_id;
- count = rte_ring_count(mp->ring);
+ count = rte_mempool_ops_get_count(mp);
if (mp->cache_size == 0)
return count;
@@ -1119,7 +1099,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
fprintf(f, "mempool <%s>@%p\n", mp->name, mp);
fprintf(f, " flags=%x\n", mp->flags);
- fprintf(f, " ring=<%s>@%p\n", mp->ring->name, mp->ring);
+ fprintf(f, " pool=%p\n", mp->pool_data);
fprintf(f, " phys_addr=0x%" PRIx64 "\n", mp->mz->phys_addr);
fprintf(f, " nb_mem_chunks=%u\n", mp->nb_mem_chunks);
fprintf(f, " size=%"PRIu32"\n", mp->size);
@@ -1140,7 +1120,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
}
cache_count = rte_mempool_dump_cache(f, mp);
- common_count = rte_ring_count(mp->ring);
+ common_count = rte_mempool_ops_get_count(mp);
if ((cache_count + common_count) > mp->size)
common_count = mp->size - cache_count;
fprintf(f, " common_pool_count=%u\n", common_count);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 60339bd..a763fb5 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -67,6 +67,7 @@
#include <inttypes.h>
#include <sys/queue.h>
+#include <rte_spinlock.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_lcore.h>
@@ -203,10 +204,14 @@ struct rte_mempool_memhdr {
*/
struct rte_mempool {
char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
- struct rte_ring *ring; /**< Ring to store objects. */
- const struct rte_memzone *mz; /**< Memzone where pool is allocated */
+ union {
+ void *pool_data; /**< Ring or pool to store objects. */
+ uint64_t pool_id; /**< External mempool identifier. */
+ };
+ void *pool_config; /**< optional args for ops alloc. */
+ const struct rte_memzone *mz; /**< Memzone where pool is alloc'd. */
int flags; /**< Flags of the mempool. */
- int socket_id; /**< Socket id passed at mempool creation. */
+ int socket_id; /**< Socket id passed at create. */
uint32_t size; /**< Max size of the mempool. */
uint32_t cache_size; /**< Size of per-lcore local cache. */
uint32_t cache_flushthresh;
@@ -217,6 +222,14 @@ struct rte_mempool {
uint32_t trailer_size; /**< Size of trailer (after elt). */
unsigned private_data_size; /**< Size of private data. */
+ /**
+ * Index into rte_mempool_ops_table array of mempool ops
+ * structs, which contain callback function pointers.
+ * We're using an index here rather than pointers to the callbacks
+ * to facilitate any secondary processes that may want to use
+ * this mempool.
+ */
+ int32_t ops_index;
struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
@@ -235,7 +248,7 @@ struct rte_mempool {
#define MEMPOOL_F_NO_CACHE_ALIGN 0x0002 /**< Do not align objs on cache lines.*/
#define MEMPOOL_F_SP_PUT 0x0004 /**< Default put is "single-producer".*/
#define MEMPOOL_F_SC_GET 0x0008 /**< Default get is "single-consumer".*/
-#define MEMPOOL_F_RING_CREATED 0x0010 /**< Internal: ring is created */
+#define MEMPOOL_F_POOL_CREATED 0x0010 /**< Internal: pool is created. */
#define MEMPOOL_F_NO_PHYS_CONTIG 0x0020 /**< Don't need physically contiguous objs. */
/**
@@ -325,6 +338,213 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
#define __mempool_check_cookies(mp, obj_table_const, n, free) do {} while(0)
#endif /* RTE_LIBRTE_MEMPOOL_DEBUG */
+#define RTE_MEMPOOL_OPS_NAMESIZE 32 /**< Max length of ops struct name. */
+
+/**
+ * Prototype for implementation specific data provisioning function.
+ *
+ * The function should provide the implementation specific memory for
+ * for use by the other mempool ops functions in a given mempool ops struct.
+ * E.g. the default ops provides an instance of the rte_ring for this purpose.
+ * it will most likely point to a different type of data structure, and
+ * will be transparent to the application programmer.
+ * This function should set mp->pool_data.
+ */
+typedef int (*rte_mempool_alloc_t)(struct rte_mempool *mp);
+
+/**
+ * Free the opaque private data pointed to by mp->pool_data pointer.
+ */
+typedef void (*rte_mempool_free_t)(struct rte_mempool *mp);
+
+/**
+ * Enqueue an object into the external pool.
+ */
+typedef int (*rte_mempool_enqueue_t)(struct rte_mempool *mp,
+ void * const *obj_table, unsigned int n);
+
+/**
+ * Dequeue an object from the external pool.
+ */
+typedef int (*rte_mempool_dequeue_t)(struct rte_mempool *mp,
+ void **obj_table, unsigned int n);
+
+/**
+ * Return the number of available objects in the external pool.
+ */
+typedef unsigned (*rte_mempool_get_count)(const struct rte_mempool *mp);
+
+/** Structure defining mempool operations structure */
+struct rte_mempool_ops {
+ char name[RTE_MEMPOOL_OPS_NAMESIZE]; /**< Name of mempool ops struct. */
+ rte_mempool_alloc_t alloc; /**< Allocate private data. */
+ rte_mempool_free_t free; /**< Free the external pool. */
+ rte_mempool_enqueue_t enqueue; /**< Enqueue an object. */
+ rte_mempool_dequeue_t dequeue; /**< Dequeue an object. */
+ rte_mempool_get_count get_count; /**< Get qty of available objs. */
+} __rte_cache_aligned;
+
+#define RTE_MEMPOOL_MAX_OPS_IDX 16 /**< Max registered ops structs */
+
+/**
+ * Structure storing the table of registered ops structs, each of which contain
+ * the function pointers for the mempool ops functions.
+ * Each process has its own storage for this ops struct array so that
+ * the mempools can be shared across primary and secondary processes.
+ * The indices used to access the array are valid across processes, whereas
+ * any function pointers stored directly in the mempool struct would not be.
+ * This results in us simply having "ops_index" in the mempool struct.
+ */
+struct rte_mempool_ops_table {
+ rte_spinlock_t sl; /**< Spinlock for add/delete. */
+ uint32_t num_ops; /**< Number of used ops structs in the table. */
+ /**
+ * Storage for all possible ops structs.
+ */
+ struct rte_mempool_ops ops[RTE_MEMPOOL_MAX_OPS_IDX];
+} __rte_cache_aligned;
+
+/** Array of registered ops structs. */
+extern struct rte_mempool_ops_table rte_mempool_ops_table;
+
+/**
+ * @internal Get the mempool ops struct from its index.
+ *
+ * @param ops_index
+ * The index of the ops struct in the ops struct table. It must be a valid
+ * index: (0 <= idx < num_ops).
+ * @return
+ * The pointer to the ops struct in the table.
+ */
+static inline struct rte_mempool_ops *
+rte_mempool_ops_get(int ops_index)
+{
+ RTE_VERIFY(ops_index < RTE_MEMPOOL_MAX_OPS_IDX);
+
+ return &rte_mempool_ops_table.ops[ops_index];
+}
+
+/**
+ * @internal Wrapper for mempool_ops alloc callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * - 0: Success; successfully allocated mempool pool_data.
+ * - <0: Error; code of alloc function.
+ */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp);
+
+/**
+ * @internal Wrapper for mempool_ops get callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to get.
+ * @return
+ * - 0: Success; got n objects.
+ * - <0: Error; code of get function.
+ */
+static inline int
+rte_mempool_ops_dequeue_bulk(struct rte_mempool *mp,
+ void **obj_table, unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->dequeue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops put callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to put.
+ * @return
+ * - 0: Success; n objects supplied.
+ * - <0: Error; code of put function.
+ */
+static inline int
+rte_mempool_ops_enqueue_bulk(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->enqueue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops get_count callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * The number of available objects in the external pool.
+ */
+unsigned
+rte_mempool_ops_get_count(const struct rte_mempool *mp);
+
+/**
+ * @internal wrapper for mempool_ops free callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ */
+void
+rte_mempool_ops_free(struct rte_mempool *mp);
+
+/**
+ * Set the ops of a mempool.
+ *
+ * This can only be done on a mempool that is not populated, i.e. just after
+ * a call to rte_mempool_create_empty().
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param name
+ * Name of the ops structure to use for this mempool.
+ * @return
+ * - 0: Success; the mempool is now using the requested ops functions.
+ * - -EINVAL - Invalid ops struct name provided.
+ * - -EEXIST - mempool already has an ops struct assigned.
+ */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config);
+
+/**
+ * Register mempool operations.
+ *
+ * @param ops
+ * Pointer to an ops structure to register.
+ * @return
+ * - >=0: Success; return the index of the ops struct in the table.
+ * - -EINVAL - some missing callbacks while registering ops struct.
+ * - -ENOSPC - the maximum number of ops structs has been reached.
+ */
+int rte_mempool_ops_register(const struct rte_mempool_ops *ops);
+
+/**
+ * Macro to statically register the ops of an external mempool manager.
+ * Note that the rte_mempool_ops_register fails silently here when
+ * more then RTE_MEMPOOL_MAX_OPS_IDX is registered.
+ */
+#define MEMPOOL_REGISTER_OPS(ops) \
+ void mp_hdlr_init_##ops(void); \
+ void __attribute__((constructor, used)) mp_hdlr_init_##ops(void)\
+ { \
+ rte_mempool_ops_register(&ops); \
+ }
+
/**
* An object callback function for mempool.
*
@@ -774,7 +994,7 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
cache->len += n;
if (cache->len >= flushthresh) {
- rte_ring_mp_enqueue_bulk(mp->ring, &cache->objs[cache_size],
+ rte_mempool_ops_enqueue_bulk(mp, &cache->objs[cache_size],
cache->len - cache_size);
cache->len = cache_size;
}
@@ -785,19 +1005,10 @@ ring_enqueue:
/* push remaining objects in ring */
#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
- if (is_mp) {
- if (rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
- else {
- if (rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
+ if (rte_mempool_ops_enqueue_bulk(mp, obj_table, n) < 0)
+ rte_panic("cannot put objects in mempool\n");
#else
- if (is_mp)
- rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n);
- else
- rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n);
+ rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
#endif
}
@@ -945,7 +1156,8 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
uint32_t req = n + (cache_size - cache->len);
/* How many do we require i.e. number to fill the cache + the request */
- ret = rte_ring_mc_dequeue_bulk(mp->ring, &cache->objs[cache->len], req);
+ ret = rte_mempool_ops_dequeue_bulk(mp,
+ &cache->objs[cache->len], req);
if (unlikely(ret < 0)) {
/*
* In the offchance that we are buffer constrained,
@@ -972,10 +1184,7 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
ring_dequeue:
/* get remaining objects from ring */
- if (is_mc)
- ret = rte_ring_mc_dequeue_bulk(mp->ring, obj_table, n);
- else
- ret = rte_ring_sc_dequeue_bulk(mp->ring, obj_table, n);
+ ret = rte_mempool_ops_dequeue_bulk(mp, obj_table, n);
if (ret < 0)
__MEMPOOL_STAT_ADD(mp, get_fail, n);
diff --git a/lib/librte_mempool/rte_mempool_ops.c b/lib/librte_mempool/rte_mempool_ops.c
new file mode 100644
index 0000000..7977a14
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ops.c
@@ -0,0 +1,150 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Copyright(c) 2016 6WIND S.A.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_mempool.h>
+#include <rte_errno.h>
+
+/* indirect jump table to support external memory pools. */
+struct rte_mempool_ops_table rte_mempool_ops_table = {
+ .sl = RTE_SPINLOCK_INITIALIZER,
+ .num_ops = 0
+};
+
+/* add a new ops struct in rte_mempool_ops_table, return its index. */
+int
+rte_mempool_ops_register(const struct rte_mempool_ops *h)
+{
+ struct rte_mempool_ops *ops;
+ int16_t ops_index;
+
+ rte_spinlock_lock(&rte_mempool_ops_table.sl);
+
+ if (rte_mempool_ops_table.num_ops >=
+ RTE_MEMPOOL_MAX_OPS_IDX) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Maximum number of mempool ops structs exceeded\n");
+ return -ENOSPC;
+ }
+
+ if (h->alloc == NULL || h->enqueue == NULL ||
+ h->dequeue == NULL || h->get_count == NULL) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Missing callback while registering mempool ops\n");
+ return -EINVAL;
+ }
+
+ if (strlen(h->name) >= sizeof(ops->name) - 1) {
+ RTE_LOG(DEBUG, EAL, "%s(): mempool_ops <%s>: name too long\n",
+ __func__, h->name);
+ rte_errno = EEXIST;
+ return -EEXIST;
+ }
+
+ ops_index = rte_mempool_ops_table.num_ops++;
+ ops = &rte_mempool_ops_table.ops[ops_index];
+ snprintf(ops->name, sizeof(ops->name), "%s", h->name);
+ ops->alloc = h->alloc;
+ ops->enqueue = h->enqueue;
+ ops->dequeue = h->dequeue;
+ ops->get_count = h->get_count;
+
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+
+ return ops_index;
+}
+
+/* wrapper to allocate an external mempool's private (pool) data. */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->alloc(mp);
+}
+
+/* wrapper to free an external pool ops. */
+void
+rte_mempool_ops_free(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ if (ops->free == NULL)
+ return;
+ return ops->free(mp);
+}
+
+/* wrapper to get available objects in an external mempool. */
+unsigned int
+rte_mempool_ops_get_count(const struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->get_count(mp);
+}
+
+/* sets mempool ops previously registered by rte_mempool_ops_register. */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name,
+ void *pool_config)
+{
+ struct rte_mempool_ops *ops = NULL;
+ unsigned i;
+
+ /* too late, the mempool is already populated. */
+ if (mp->flags & MEMPOOL_F_POOL_CREATED)
+ return -EEXIST;
+
+ for (i = 0; i < rte_mempool_ops_table.num_ops; i++) {
+ if (!strcmp(name,
+ rte_mempool_ops_table.ops[i].name)) {
+ ops = &rte_mempool_ops_table.ops[i];
+ break;
+ }
+ }
+
+ if (ops == NULL)
+ return -EINVAL;
+
+ mp->ops_index = i;
+ mp->pool_config = pool_config;
+ return 0;
+}
diff --git a/lib/librte_mempool/rte_mempool_ring.c b/lib/librte_mempool/rte_mempool_ring.c
new file mode 100644
index 0000000..626786e
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ring.c
@@ -0,0 +1,161 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_errno.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+
+static int
+common_ring_mp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_mp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_sp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_mc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_mc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_sc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static unsigned
+common_ring_get_count(const struct rte_mempool *mp)
+{
+ return rte_ring_count(mp->pool_data);
+}
+
+
+static int
+common_ring_alloc(struct rte_mempool *mp)
+{
+ int rg_flags = 0, ret;
+ char rg_name[RTE_RING_NAMESIZE];
+ struct rte_ring *r;
+
+ ret = snprintf(rg_name, sizeof(rg_name),
+ RTE_MEMPOOL_MZ_FORMAT, mp->name);
+ if (ret < 0 || ret >= (int)sizeof(rg_name)) {
+ rte_errno = ENAMETOOLONG;
+ return -rte_errno;
+ }
+
+ /* ring flags */
+ if (mp->flags & MEMPOOL_F_SP_PUT)
+ rg_flags |= RING_F_SP_ENQ;
+ if (mp->flags & MEMPOOL_F_SC_GET)
+ rg_flags |= RING_F_SC_DEQ;
+
+ /*
+ * Allocate the ring that will be used to store objects.
+ * Ring functions will return appropriate errors if we are
+ * running as a secondary process etc., so no checks made
+ * in this function for that condition.
+ */
+ r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
+ mp->socket_id, rg_flags);
+ if (r == NULL)
+ return -rte_errno;
+
+ mp->pool_data = r;
+
+ return 0;
+}
+
+static void
+common_ring_free(struct rte_mempool *mp)
+{
+ rte_ring_free(mp->pool_data);
+}
+
+/*
+ * The following 4 declarations of mempool ops structs address
+ * the need for the backward compatible mempool managers for
+ * single/multi producers and single/multi consumers as dictated by the
+ * flags provided to the rte_mempool_create function
+ */
+static const struct rte_mempool_ops ops_mp_mc = {
+ .name = "ring_mp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_sc = {
+ .name = "ring_sp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_mp_sc = {
+ .name = "ring_mp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_mc = {
+ .name = "ring_sp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+MEMPOOL_REGISTER_OPS(ops_mp_mc);
+MEMPOOL_REGISTER_OPS(ops_sp_sc);
+MEMPOOL_REGISTER_OPS(ops_mp_sc);
+MEMPOOL_REGISTER_OPS(ops_sp_mc);
diff --git a/lib/librte_mempool/rte_mempool_version.map b/lib/librte_mempool/rte_mempool_version.map
index f63461b..6209ec2 100644
--- a/lib/librte_mempool/rte_mempool_version.map
+++ b/lib/librte_mempool/rte_mempool_version.map
@@ -20,15 +20,18 @@ DPDK_16.7 {
global:
rte_mempool_check_cookies;
- rte_mempool_obj_iter;
- rte_mempool_mem_iter;
rte_mempool_create_empty;
+ rte_mempool_free;
+ rte_mempool_mem_iter;
+ rte_mempool_obj_iter;
+ rte_mempool_ops_register;
+ rte_mempool_ops_table;
+ rte_mempool_populate_anon;
+ rte_mempool_populate_default;
rte_mempool_populate_phys;
rte_mempool_populate_phys_tab;
rte_mempool_populate_virt;
- rte_mempool_populate_default;
- rte_mempool_populate_anon;
- rte_mempool_free;
+ rte_mempool_set_ops_byname;
local: *;
} DPDK_2.0;
--
2.5.5
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v13 0/3] mempool: add external mempool manager
2016-06-15 7:47 3% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager David Hunt
2016-06-15 7:47 1% ` [dpdk-dev] [PATCH v12 1/3] mempool: support external mempool operations David Hunt
2016-06-15 10:13 0% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager Jan Viktorin
@ 2016-06-16 12:30 3% ` David Hunt
2016-06-16 12:30 1% ` [dpdk-dev] [PATCH v13 1/3] mempool: support external mempool operations David Hunt
2016-06-17 13:53 3% ` [dpdk-dev] [PATCH v14 0/3] mempool: add mempool handler feature David Hunt
2 siblings, 2 replies; 200+ results
From: David Hunt @ 2016-06-16 12:30 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the External Mempool Manager patchset.
It's re-based on top of the latest head as of 15/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v13 changes:
* Added in extra opaque data (pool_config) to mempool struct for mempool
configuration by the ops functions. For example, this can be used to pass
device names or device flags to the underlying alloc function.
* Added mempool_config param to rte_mempool_set_ops_byname()
v12 changes:
* Fixed a comment (function pram h -> ops)
* fixed a typo (callbacki)
v11 changes:
* Fixed comments (added '.' where needed for consistency)
* removed ABI breakage notice for mempool manager in deprecation.rst
* Added description of the external mempool manager functionality to
doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
* renamed rte_mempool_default.c to rte_mempool_ring.c
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The External Mempool Manager is an extension to the mempool API that allows
users to add and use an external mempool manager, which allows external memory
subsystems such as external hardware memory management systems and software
based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool manager will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing. These changes are all contained withing
RTE_NEXT_ABI defs, and the current or next code can be changed with
the CONFIG_RTE_NEXT_ABI config setting
There are two aspects to external mempool manager.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops_byname to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempool's ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several external mempool managers may be used in the same application. A new
mempool can then be created by using the new rte_mempool_create_empty function,
then calling rte_mempool_set_ops_byname to point the mempool to the relevant
mempool manager callback structure.
Legacy applications will continue to use the old rte_mempool_create API call,
which uses a ring based mempool manager by default. These applications
will need to be modified to use a new external mempool manager.
The external mempool manager needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name
void *pool_config);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool manager using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support external mempool operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test external mempool handler
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v2 5/7] eal/linux: mmap ioports on ppc64
@ 2016-06-15 16:13 3% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-15 16:13 UTC (permalink / raw)
To: Olivier Matz, Yuanhan Liu
Cc: David Marchand, dev, Chao Zhu, Xie, Huawei, Panu Matilainen
2016-05-30 10:45, Olivier Matz:
> On 05/24/2016 07:15 AM, Yuanhan Liu wrote:
> > On Mon, May 23, 2016 at 03:40:58PM +0200, Olivier Matz wrote:
> >> For reference, here is the report of the ABI checker for EAL:
> >>
> >> [−] struct rte_pci_ioport (2)
> >>
> >> 1 Field len has been added to this type.
> >> 1) This field will not be initialized by old clients.
> >> 2) Size of the inclusive type has been changed.
> >> NOTE: this field should be accessed only from the new library
> >> functions, otherwise it may result in crash or incorrect behavior
> >> of applications.
> >> 2 Size of this type has been changed from 16 bytes to 24 bytes.
> >> The fields or parameters of such data type may be incorrectly
> >> initialized or accessed by old client applications.
> >>
> >> [−] affected symbols (4)
> >> rte_eal_pci_ioport_map ( struct rte_pci_device* dev, int bar,
> >> struct rte_pci_ioport* p ) @@ DPDK_16.04
> >> 3rd parameter 'p' (pointer) has base type 'struct rte_pci_ioport'.
> >> rte_eal_pci_ioport_read ( struct rte_pci_ioport* p, void* data,
> >> size_t len, off_t offset ) @@ DPDK_16.04
> >> 1st parameter 'p' (pointer) has base type 'struct rte_pci_ioport'.
> >> rte_eal_pci_ioport_unmap ( struct rte_pci_ioport* p ) @@ DPDK_16.04
> >> 1st parameter 'p' (pointer) has base type 'struct rte_pci_ioport'.
> >> rte_eal_pci_ioport_write ( struct rte_pci_ioport* p, void const* data,
> >> size_t len, off_t offset ) @@ DPDK_16.04
> >> 1st parameter 'p' (pointer) has base type 'struct rte_pci_ioport'.
> >>
> >>
> >> My understanding of the comment for this structure is that it's
> >> internal to EAL:
> >
> > I'm not quite sure that is enough. Cc'ed Panu, the guru on ABI stuff,
> > hopefully he could shed some light on it.
> >
> >> /**
> >> * A structure used to access io resources for a pci device.
> >> * rte_pci_ioport is arch, os, driver specific, and should not be used
> >> outside
> >> * of pci ioport api.
> >> */
> >> struct rte_pci_ioport {
> >> ...
> >> }
> >>
> >> So I'd say it's ok to have it integrated for 16.07.
> >
> > I'll let Thomas to decide it :)
>
> Panu or Thomas, do you have any comment on this?
The user of this struct is virtio.
The ABI policy does not apply to drivers:
- A means Application
- external drivers must be rebuilt for each new release
Thus no problem here.
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v10 3/7] ethdev: add new fields to ethdev info struct
2016-06-15 14:06 2% ` [dpdk-dev] [PATCH v10 0/7] add " Reshma Pattan
@ 2016-06-15 14:06 10% ` Reshma Pattan
2016-06-16 19:14 4% ` Thomas Monjalon
2016-06-15 14:06 4% ` [dpdk-dev] [PATCH v10 4/7] ethdev: make get port by name and get name by port public Reshma Pattan
1 sibling, 1 reply; 200+ results
From: Reshma Pattan @ 2016-06-15 14:06 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
The new fields nb_rx_queues and nb_tx_queues are added to the
rte_eth_dev_info structure.
Changes to API rte_eth_dev_info_get() are done to update these new fields
to the rte_eth_dev_info object.
Release notes is updated with the changes.
The librte_pdump library needs to register Rx and Tx callbacks for all
the nb_rx_queues and nb_tx_queues, when application wants to capture the
packets on all the software configured number of Rx and Tx queues of the
device. So far there is no support to get nb_rx_queues and nb_tx_queues
information from the ethdev library. Hence these changes are introduced.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
---
doc/guides/rel_notes/release_16_07.rst | 6 ++++++
lib/librte_ether/rte_ethdev.c | 2 ++
lib/librte_ether/rte_ethdev.h | 3 +++
lib/librte_ether/rte_ether_version.map | 1 +
4 files changed, 12 insertions(+)
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index c0f6b02..004ecee 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -135,6 +135,9 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ in the ``rte_eth_dev_info`` object.
+
ABI Changes
-----------
@@ -146,6 +149,9 @@ ABI Changes
* The ``rte_port_source_params`` structure has new fields to support PCAP file.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ to support number of queues configured by software.
+
Shared Library Versions
-----------------------
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index 97d167e..1f634c9 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -1661,6 +1661,8 @@ rte_eth_dev_info_get(uint8_t port_id, struct rte_eth_dev_info *dev_info)
(*dev->dev_ops->dev_infos_get)(dev, dev_info);
dev_info->pci_dev = dev->pci_dev;
dev_info->driver_name = dev->data->drv_name;
+ dev_info->nb_rx_queues = dev->data->nb_rx_queues;
+ dev_info->nb_tx_queues = dev->data->nb_tx_queues;
}
int
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 237e6ef..8ad7c01 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -882,6 +882,9 @@ struct rte_eth_dev_info {
struct rte_eth_desc_lim rx_desc_lim; /**< RX descriptors limits */
struct rte_eth_desc_lim tx_desc_lim; /**< TX descriptors limits */
uint32_t speed_capa; /**< Supported speeds bitmap (ETH_LINK_SPEED_). */
+ /** Configured number of rx/tx queues */
+ uint16_t nb_rx_queues; /**< Number of RX queues. */
+ uint16_t nb_tx_queues; /**< Number of TX queues. */
};
/**
diff --git a/lib/librte_ether/rte_ether_version.map b/lib/librte_ether/rte_ether_version.map
index c990b04..d06d648 100644
--- a/lib/librte_ether/rte_ether_version.map
+++ b/lib/librte_ether/rte_ether_version.map
@@ -137,4 +137,5 @@ DPDK_16.07 {
global:
rte_eth_add_first_rx_callback;
+ rte_eth_dev_info_get;
} DPDK_16.04;
--
2.5.0
^ permalink raw reply [relevance 10%]
* [dpdk-dev] [PATCH v10 0/7] add packet capture framework
2016-06-14 9:38 2% ` [dpdk-dev] [PATCH v9 " Reshma Pattan
2016-06-14 9:38 5% ` [dpdk-dev] [PATCH v9 8/8] doc: update doc for " Reshma Pattan
@ 2016-06-15 14:06 2% ` Reshma Pattan
2016-06-15 14:06 10% ` [dpdk-dev] [PATCH v10 3/7] ethdev: add new fields to ethdev info struct Reshma Pattan
2016-06-15 14:06 4% ` [dpdk-dev] [PATCH v10 4/7] ethdev: make get port by name and get name by port public Reshma Pattan
1 sibling, 2 replies; 200+ results
From: Reshma Pattan @ 2016-06-15 14:06 UTC (permalink / raw)
To: dev
This patch set include below changes
1)Changes to librte_ether.
2)A new library librte_pdump added for packet capture framework.
3)A new app/pdump tool added for packet capturing.
4)Test pmd changes done to initialize packet capture framework.
5)Documentation update.
1)librte_pdump
==============
To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
is added.Users can develop their own packet capturing application using new library APIs.
Operation:
----------
The librte_pdump provides APIs to support packet capturing on dpdk Ethernet devices.
Library provides APIs to initialize the packet capture framework, enable/disable
the packet capture and uninitialize the packet capture framework.
The librte_pdump library works on a client/server model. The server is responsible for enabling or
disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
the packet capture.
The packet capture framework, as part of its initialization, creates the pthread and the server socket in
the pthread. The application that calls the framework initialization will have the server socket created,
either under the path that the application has passed or under the default path i.e. either ''/var/run'' for
root user or ''$HOME'' for non root user.
Applications that request enabling or disabling of the packet capture will have the client socket created
either under the path that the application has passed or under the default path i.e. either ''/var/run/''
for root users or ''$HOME'' for not root users to send the requests to the server.
The server socket will listen for client requests for enabling or disabling the packet capture.
Applications using below APIs need to pass port/device_id, queue, mempool and
ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
packets of the port for users. Users need to dequeue the rings and write the packets
to vdev(pcap/tuntap) to view the packets using any standard tools.
Note:
Mempool and Ring should be mc/mp supportable.
Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
APIs:
-----
rte_pdump_init()
rte_pdump_enable()
rte_pdump_enable_by_deviceid()
rte_pdump_disable()
rte_pdump_disable_by_deviceid()
rte_pdump_uninit()
rte_pdump_set_socket_dir()
2)app/pdump tool
================
Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
This tool by default runs as secondary process, and provides the support for
the command line options for packet capture.
./build/app/dpdk_pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue id>),
(rx-dev=<iface or pcap file> |
tx-dev=<iface or pcap file>),
[ring-size=<ring size>],
[mbuf-size=<mbuf data size>],
[total-num-mbufs=<number of mbufs>]'
Parameters inside the parenthesis represents the mandatory parameters.
Parameters inside the square brackets represents optional parameters.
User has to pass on packet capture parameters under --pdump parameters, multiples of
--pdump can be passed to capture packets on different port and queue combinations
Operation:
----------
*Tool parse the user command line arguments,
creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
of the device passed in rx-dev|tx-dev parameters.
*Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
to enable packet capturing on a specific port/device_id and queue by passing on
port|device_id, queue, mempool and ring info.
*Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
*Tool can be stopped using SIGINT, upon which tool calls
rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
Note:
CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
3)Test-pmd changes
==================
Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
Similarly any application which needs packet capture should call initialize/uninitialize APIs of
librte_pdump and use pdump tool to start the capture.
4)Packet capture flow between pdump tool and librte_pdump
=========================================================
* Pdump tool (Secondary process) requests packet capture
for specific port|device_id and queue combinations.
*Library in secondary process context creates client socket and communicates
the port|device_id, queue, ring and mempool to server.
*Library initializes server in primary process 'test-pmd' context and server serves
the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
*Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
*Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
*Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
*Library receives the disable packet capture request, communicate the info to server,
server will remove the Ethernet rxtx call-backs.
*Packet capture can be seen using tcpdump command
"tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
5)Example command line
======================
./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
v10:
fixed commit messages description.
fixed compilation issue when CONFIG_RTE_LIBRTE_PDUMP is disabled.
removed wrong config option CONFIG_RTE_EXEC_ENV_LINUXAPP inside app/Makefile
for pdump tool.
moved document changes to appropriate patches.
v9:
added a support in rte_pdump_set_socket_dir() to set server and client socket paths
==> http://dpdk.org/dev/patchwork/patch/13450/
updated the documentation for the new changes.
updated the commit messages.
v8:
added server socket argument to rte_pdump_init() API ==> http://dpdk.org/dev/patchwork/patch/13402/
added rte_pdump_set_socket_dir() API.
updated documentation for new changes.
v7:
fixed lines over 90 characters.
v6:
removed below deprecation notice patch from patch set.
http://dpdk.org/dev/patchwork/patch/13372/
v5:
addressed code review comments for below patches
http://dpdk.org/dev/patchwork/patch/12955/
http://dpdk.org/dev/patchwork/patch/12951/
v4:
added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
made doc changes as per doc guidelines.
replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
removed rxtx-dev parameter from pdump tool command line.
v3:
app/pdump: Moved cleanup code from signal handler to main.
divided librte_ether changes into multiple patches.
example command changed in app/pdump application guide
v2:
fix compilation issues for 4.8.3
fix unnecessary #includes
Reshma Pattan (7):
ethdev: use locks to protect Rx/Tx callback lists
ethdev: add new api to add Rx callback as head of the list
ethdev: add new fields to ethdev info struct
ethdev: make get port by name and get name by port public
pdump: add new library for packet capturing support
app/pdump: add pdump tool for packet capturing
app/testpmd: add pdump initialization uninitialization
MAINTAINERS | 7 +
app/Makefile | 1 +
app/pdump/Makefile | 49 ++
app/pdump/main.c | 844 +++++++++++++++++++++++++++++
app/test-pmd/testpmd.c | 12 +
config/common_base | 5 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 123 +++++
doc/guides/rel_notes/release_16_07.rst | 14 +
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 +++++
lib/Makefile | 1 +
lib/librte_ether/rte_ethdev.c | 123 +++--
lib/librte_ether/rte_ethdev.h | 60 +++
lib/librte_ether/rte_ether_version.map | 9 +
lib/librte_pdump/Makefile | 55 ++
lib/librte_pdump/rte_pdump.c | 913 ++++++++++++++++++++++++++++++++
lib/librte_pdump/rte_pdump.h | 216 ++++++++
lib/librte_pdump/rte_pdump_version.map | 13 +
mk/rte.app.mk | 1 +
20 files changed, 2526 insertions(+), 44 deletions(-)
create mode 100644 app/pdump/Makefile
create mode 100644 app/pdump/main.c
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
create mode 100644 lib/librte_pdump/Makefile
create mode 100644 lib/librte_pdump/rte_pdump.c
create mode 100644 lib/librte_pdump/rte_pdump.h
create mode 100644 lib/librte_pdump/rte_pdump_version.map
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
--
2.5.0
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH v10 4/7] ethdev: make get port by name and get name by port public
2016-06-15 14:06 2% ` [dpdk-dev] [PATCH v10 0/7] add " Reshma Pattan
2016-06-15 14:06 10% ` [dpdk-dev] [PATCH v10 3/7] ethdev: add new fields to ethdev info struct Reshma Pattan
@ 2016-06-15 14:06 4% ` Reshma Pattan
1 sibling, 0 replies; 200+ results
From: Reshma Pattan @ 2016-06-15 14:06 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
Converted rte_eth_dev_get_port_by_name to a public API.
Converted rte_eth_dev_get_name_by_port to a public API.
Updated the release notes with the changes.
The librte_pdump library provides the APIs to enable or disable the
packet capture either using the port id or pci address or device name.
So pdump library need to do a mapping from name to port and port to name
internally to validate the device name and register the Rx and Tx
callbacks for the mapped ports. So these APIs are made public for the
pdump library for doing the mentioned mappings.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
---
doc/guides/rel_notes/release_16_07.rst | 3 +++
lib/librte_ether/rte_ethdev.c | 4 ++--
lib/librte_ether/rte_ethdev.h | 29 +++++++++++++++++++++++++++++
lib/librte_ether/rte_ether_version.map | 2 ++
4 files changed, 36 insertions(+), 2 deletions(-)
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 004ecee..c6222f8 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -138,6 +138,9 @@ API Changes
* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
in the ``rte_eth_dev_info`` object.
+* Functions ``rte_eth_dev_get_port_by_name`` and ``rte_eth_dev_get_name_by_port``
+ are changed to a public APIs.
+
ABI Changes
-----------
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index 1f634c9..0b19569 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -406,7 +406,7 @@ rte_eth_dev_get_addr_by_port(uint8_t port_id, struct rte_pci_addr *addr)
return 0;
}
-static int
+int
rte_eth_dev_get_name_by_port(uint8_t port_id, char *name)
{
char *tmp;
@@ -425,7 +425,7 @@ rte_eth_dev_get_name_by_port(uint8_t port_id, char *name)
return 0;
}
-static int
+int
rte_eth_dev_get_port_by_name(const char *name, uint8_t *port_id)
{
int i;
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 8ad7c01..fab281e 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -4284,6 +4284,35 @@ rte_eth_dev_l2_tunnel_offload_set(uint8_t port_id,
uint32_t mask,
uint8_t en);
+/**
+* Get the port id from pci adrress or device name
+* Ex: 0000:2:00.0 or vdev name eth_pcap0
+*
+* @param name
+* pci address or name of the device
+* @param port_id
+* pointer to port identifier of the device
+* @return
+* - (0) if successful.
+* - (-ENODEV or -EINVAL) on failure.
+*/
+int
+rte_eth_dev_get_port_by_name(const char *name, uint8_t *port_id);
+
+/**
+* Get the device name from port id
+*
+* @param port_id
+* pointer to port identifier of the device
+* @param name
+* pci address or name of the device
+* @return
+* - (0) if successful.
+* - (-EINVAL) on failure.
+*/
+int
+rte_eth_dev_get_name_by_port(uint8_t port_id, char *name);
+
#ifdef __cplusplus
}
#endif
diff --git a/lib/librte_ether/rte_ether_version.map b/lib/librte_ether/rte_ether_version.map
index d06d648..73e730d 100644
--- a/lib/librte_ether/rte_ether_version.map
+++ b/lib/librte_ether/rte_ether_version.map
@@ -137,5 +137,7 @@ DPDK_16.07 {
global:
rte_eth_add_first_rx_callback;
+ rte_eth_dev_get_name_by_port;
+ rte_eth_dev_get_port_by_name;
rte_eth_dev_info_get;
} DPDK_16.04;
--
2.5.0
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH v3] mbuf: new flag when Vlan is stripped
@ 2016-06-15 11:48 2% ` Olivier Matz
0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-06-15 11:48 UTC (permalink / raw)
To: dev
Cc: johndale, konstantin.ananyev, helin.zhang, adrien.mazarguil,
rahul.lakkireddy, alejandro.lucero, sony.chacko
The behavior of PKT_RX_VLAN_PKT was not very well defined, resulting in
PMDs not advertising the same flags in similar conditions.
Following discussion in [1], introduce 2 new flags PKT_RX_VLAN_STRIPPED
and PKT_RX_QINQ_STRIPPED that are better defined:
PKT_RX_VLAN_STRIPPED: a vlan has been stripped by the hardware and its
tci is saved in mbuf->vlan_tci. This can only happen if vlan stripping
is enabled in the RX configuration of the PMD.
For now, the old flag PKT_RX_VLAN_PKT is kept but marked as deprecated.
It should be removed from applications and PMDs in a future revision.
This patch also updates the drivers. For PKT_RX_VLAN_PKT:
- e1000, enic, i40e, mlx5, nfp, vmxnet3: done, PKT_RX_VLAN_PKT already
had the same meaning than PKT_RX_VLAN_STRIPPED, minor update is
required.
- fm10k: done, PKT_RX_VLAN_PKT already had the same meaning than
PKT_RX_VLAN_STRIPPED, and vlan stripping is always enabled on fm10k.
- ixgbe: modification done (vector and normal), the old flag was set
when a vlan was recognized, even if vlan stripping was disabled.
- the other drivers do not support vlan stripping.
For PKT_RX_QINQ_PKT, it was only supported on i40e, and the behavior was
already correct, so we can reuse the same bit value for
PKT_RX_QINQ_STRIPPED.
[1] http://dpdk.org/ml/archives/dev/2016-April/037837.html,
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
v2 -> v3:
- remove the "deprecated" notice from PKT_RX_VLAN_PKT comment.
- use UINT8_MAX when appropriate in ixgbe_vec
- simplify check of vlan flags in RTE_BUILD_BUG_ON() in ixgbe_vec
v1 -> v2:
- fix ixgbe (vector mode) and i40e (normal and vector mode)
- store vlan flags instead of a boolean value in ixgbe rxq, as
suggested by Konstantin
- replay tests on ixgbe (normal + vector) and i40e (normal +
vector). See below.
RFC -> v1:
- fix checkpatch and check-git-log.sh issues
- add a deprecation notice for the old vlan flags
- rebase on head
app/test-pmd/rxonly.c | 4 +--
doc/guides/rel_notes/deprecation.rst | 5 ++++
drivers/net/e1000/em_rxtx.c | 3 ++-
drivers/net/e1000/igb_rxtx.c | 3 ++-
drivers/net/enic/enic_rx.c | 2 +-
drivers/net/i40e/i40e_rxtx.c | 4 +--
drivers/net/i40e/i40e_rxtx_vec.c | 2 +-
drivers/net/ixgbe/ixgbe_ethdev.c | 11 ++++++++
drivers/net/ixgbe/ixgbe_rxtx.c | 14 +++++++----
drivers/net/ixgbe/ixgbe_rxtx.h | 2 ++
drivers/net/ixgbe/ixgbe_rxtx_vec.c | 35 +++++++++++++++++---------
drivers/net/mlx5/mlx5_rxtx.c | 6 +++--
drivers/net/nfp/nfp_net.c | 2 +-
drivers/net/vmxnet3/vmxnet3_rxtx.c | 2 +-
lib/librte_mbuf/rte_mbuf.c | 2 ++
lib/librte_mbuf/rte_mbuf.h | 49 ++++++++++++++++++++++++++++++++----
16 files changed, 112 insertions(+), 34 deletions(-)
This patch is tested on ixgbe (normal + vector), i40e (normal +
vector) and igb (hardware is a 82575):
# we use scapy to send packets like this:
# Ether(src="00:01:02:03:04:05", dst="00:1B:21:AB:8F:10")/Dot1Q(vlan=0x666)/IP()/UDP()/Raw("x"*32)
cd dpdk.org/
make config T=x86_64-native-linuxapp-gcc
make -j32
mkdir -p /mnt/huge
mount -t hugetlbfs nodev /mnt/huge
echo 256 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
modprobe uio_pci_generic
# test-pmd is started with vlan stripping, using the rx-vector
# function if available (i40e and ixgbe)
./build/app/testpmd -l 2,4 -- --total-num-mbufs=65536 -i --port-topology=chained \
--disable-hw-vlan-filter
# to disable vlan stripping, add:
--disable-hw-vlan-strip
# to disable the vector mode (it can be checked in debug logs), add:
--enable-rx-cksum
# we run test-pmd in rxonly mode, displaying the packet information.
set fwd rxonly
set verbose 1
start
==== IXGBE normal rx function
# ixgbe: the behavior of the flag PKT_RX_VLAN_PKT is kept as before,
# and the new flag PKT_RX_VLAN_STRIPPED is introduced when vlan stripping
# is enabled and a vlan is stripped.
--- vlan stripping enabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4 - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - VLAN tci=0x666 - (outer) L2 type: ETHER - (outer) L3 type: IPV4 - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
PKT_RX_VLAN_PKT
PKT_RX_VLAN_STRIPPED
--- vlan stripping disabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4 - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x8100 - length=78 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4 - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
PKT_RX_VLAN_PKT
==== IXGBE vector rx function
--- vlan stripping enabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1Unknown packet type
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - VLAN tci=0x666Unknown packet type
- Receive queue=0x0
PKT_RX_VLAN_PKT
PKT_RX_VLAN_STRIPPED
--- vlan stripping disabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1Unknown packet type
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x8100 - length=78 - nb_segs=1Unknown packet type
- Receive queue=0x0
PKT_RX_VLAN_PKT
==== I40E normal rx function
--- vlan stripping enabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4_EXT_UNKNOWN - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - VLAN tci=0x666 - (outer) L2 type: ETHER - (outer) L3 type: IPV4_EXT_UNKNOWN - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
PKT_RX_VLAN_PKT
PKT_RX_VLAN_STRIPPED
--- vlan stripping disabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4_EXT_UNKNOWN - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x8100 - length=78 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4_EXT_UNKNOWN - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4
==== I40E vector rx function
--- vlan stripping enabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1Unknown packet type
- Receive queue=0x0
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - VLAN tci=0x666Unknown packet type
- Receive queue=0x0
PKT_RX_VLAN_PKT
PKT_RX_VLAN_STRIPPED
--- vlan stripping disabled
# packet without vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1Unknown packet type
- Receive queue=0x0
port 0/queue 0: received 1 packets
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x8100 - length=78 - nb_segs=1Unknown packet type
- Receive queue=0x0
==== IGB
(not retested since RFC patch, but there was no code modification)
--- vlan stripping enabled
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x8100 - length=78 - nb_segs=1 - (outer) L2 type: ETHER - (outer) L3 type: IPV4 - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
--- vlan stripping disabled
# packet with vlan
src=00:01:02:03:04:05 - dst=00:1B:21:AB:8F:10 - type=0x0800 - length=74 - nb_segs=1 - VLAN tci=0x666 - (outer) L2 type: ETHER - (outer) L3 type: IPV4 - (outer) L4 type: UDP - Tunnel type: Unknown - Inner L2 type: Unknown - Inner L3 type: Unknown - Inner L4 type: Unknown
- Receive queue=0x0
PKT_RX_VLAN_PKT
PKT_RX_VLAN_STRIPPED
diff --git a/app/test-pmd/rxonly.c b/app/test-pmd/rxonly.c
index 14555ab..c69b344 100644
--- a/app/test-pmd/rxonly.c
+++ b/app/test-pmd/rxonly.c
@@ -156,9 +156,9 @@ pkt_burst_receive(struct fwd_stream *fs)
printf("hash=0x%x ID=0x%x ",
mb->hash.fdir.hash, mb->hash.fdir.id);
}
- if (ol_flags & PKT_RX_VLAN_PKT)
+ if (ol_flags & PKT_RX_VLAN_STRIPPED)
printf(" - VLAN tci=0x%x", mb->vlan_tci);
- if (ol_flags & PKT_RX_QINQ_PKT)
+ if (ol_flags & PKT_RX_QINQ_STRIPPED)
printf(" - QinQ VLAN tci=0x%x, VLAN tci outer=0x%x",
mb->vlan_tci, mb->vlan_tci_outer);
if (mb->packet_type) {
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 7d947ae..702dfce 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -54,3 +54,8 @@ Deprecation Notices
a handle, like the way kernel exposes an fd to user for locating a
specific file, and to keep all major structures internally, so that
we are likely to be free from ABI violations in future.
+
+* The mbuf flags PKT_RX_VLAN_PKT and PKT_RX_QINQ_PKT are deprecated and
+ are respectively replaced by PKT_RX_VLAN_STRIPPED and
+ PKT_RX_QINQ_STRIPPED, that are better described. The old flags and
+ their behavior will be kept in 16.07 and will be removed in 16.11.
diff --git a/drivers/net/e1000/em_rxtx.c b/drivers/net/e1000/em_rxtx.c
index 3d36f21..6d8750a 100644
--- a/drivers/net/e1000/em_rxtx.c
+++ b/drivers/net/e1000/em_rxtx.c
@@ -629,7 +629,8 @@ rx_desc_status_to_pkt_flags(uint32_t rx_status)
uint64_t pkt_flags;
/* Check if VLAN present */
- pkt_flags = ((rx_status & E1000_RXD_STAT_VP) ? PKT_RX_VLAN_PKT : 0);
+ pkt_flags = ((rx_status & E1000_RXD_STAT_VP) ?
+ PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED : 0);
return pkt_flags;
}
diff --git a/drivers/net/e1000/igb_rxtx.c b/drivers/net/e1000/igb_rxtx.c
index 18aeead..9d80a0b 100644
--- a/drivers/net/e1000/igb_rxtx.c
+++ b/drivers/net/e1000/igb_rxtx.c
@@ -729,7 +729,8 @@ rx_desc_status_to_pkt_flags(uint32_t rx_status)
uint64_t pkt_flags;
/* Check if VLAN present */
- pkt_flags = (rx_status & E1000_RXD_STAT_VP) ? PKT_RX_VLAN_PKT : 0;
+ pkt_flags = ((rx_status & E1000_RXD_STAT_VP) ?
+ PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED : 0);
#if defined(RTE_LIBRTE_IEEE1588)
if (rx_status & E1000_RXD_STAT_TMST)
diff --git a/drivers/net/enic/enic_rx.c b/drivers/net/enic/enic_rx.c
index f92f6bc..6459e97 100644
--- a/drivers/net/enic/enic_rx.c
+++ b/drivers/net/enic/enic_rx.c
@@ -197,7 +197,7 @@ enic_cq_rx_to_pkt_flags(struct cq_desc *cqd, struct rte_mbuf *mbuf)
/* VLAN stripping */
if (bwflags & CQ_ENET_RQ_DESC_FLAGS_VLAN_STRIPPED) {
- pkt_flags |= PKT_RX_VLAN_PKT;
+ pkt_flags |= PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
mbuf->vlan_tci = enic_cq_rx_desc_vlan(cqrd);
} else {
mbuf->vlan_tci = 0;
diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index c833aa3..eea246b 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -88,7 +88,7 @@ i40e_rxd_to_vlan_tci(struct rte_mbuf *mb, volatile union i40e_rx_desc *rxdp)
{
if (rte_le_to_cpu_64(rxdp->wb.qword1.status_error_len) &
(1 << I40E_RX_DESC_STATUS_L2TAG1P_SHIFT)) {
- mb->ol_flags |= PKT_RX_VLAN_PKT;
+ mb->ol_flags |= PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
mb->vlan_tci =
rte_le_to_cpu_16(rxdp->wb.qword0.lo_dword.l2tag1);
PMD_RX_LOG(DEBUG, "Descriptor l2tag1: %u",
@@ -99,7 +99,7 @@ i40e_rxd_to_vlan_tci(struct rte_mbuf *mb, volatile union i40e_rx_desc *rxdp)
#ifndef RTE_LIBRTE_I40E_16BYTE_RX_DESC
if (rte_le_to_cpu_16(rxdp->wb.qword2.ext_status) &
(1 << I40E_RX_DESC_EXT_STATUS_L2TAG2P_SHIFT)) {
- mb->ol_flags |= PKT_RX_QINQ_PKT;
+ mb->ol_flags |= PKT_RX_QINQ_STRIPPED;
mb->vlan_tci_outer = mb->vlan_tci;
mb->vlan_tci = rte_le_to_cpu_16(rxdp->wb.qword2.l2tag2_2);
PMD_RX_LOG(DEBUG, "Descriptor l2tag2_1: %u, l2tag2_2: %u",
diff --git a/drivers/net/i40e/i40e_rxtx_vec.c b/drivers/net/i40e/i40e_rxtx_vec.c
index eef80d9..634bd39 100644
--- a/drivers/net/i40e/i40e_rxtx_vec.c
+++ b/drivers/net/i40e/i40e_rxtx_vec.c
@@ -154,7 +154,7 @@ desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
/* map rss and vlan type to rss hash and vlan flag */
const __m128i vlan_flags = _mm_set_epi8(0, 0, 0, 0,
0, 0, 0, 0,
- 0, 0, 0, PKT_RX_VLAN_PKT,
+ 0, 0, 0, PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED,
0, 0, 0, 0);
const __m128i rss_flags = _mm_set_epi8(0, 0, 0, 0,
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a2b170b..5f3e047 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -1636,6 +1636,7 @@ ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
{
struct ixgbe_hwstrip *hwstrip =
IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
+ struct ixgbe_rx_queue *rxq;
if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
return;
@@ -1644,6 +1645,16 @@ ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
IXGBE_SET_HWSTRIP(hwstrip, queue);
else
IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
+
+ if (queue >= dev->data->nb_rx_queues)
+ return;
+
+ rxq = dev->data->rx_queues[queue];
+
+ if (on)
+ rxq->vlan_flags = PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
+ else
+ rxq->vlan_flags = PKT_RX_VLAN_PKT;
}
static void
diff --git a/drivers/net/ixgbe/ixgbe_rxtx.c b/drivers/net/ixgbe/ixgbe_rxtx.c
index 9c6eaf2..5a7064c 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx.c
+++ b/drivers/net/ixgbe/ixgbe_rxtx.c
@@ -1221,7 +1221,7 @@ ixgbe_rxd_pkt_info_to_pkt_flags(uint16_t pkt_info)
}
static inline uint64_t
-rx_desc_status_to_pkt_flags(uint32_t rx_status)
+rx_desc_status_to_pkt_flags(uint32_t rx_status, uint64_t vlan_flags)
{
uint64_t pkt_flags;
@@ -1230,7 +1230,7 @@ rx_desc_status_to_pkt_flags(uint32_t rx_status)
* Do not check whether L3/L4 rx checksum done by NIC or not,
* That can be found from rte_eth_rxmode.hw_ip_checksum flag
*/
- pkt_flags = (rx_status & IXGBE_RXD_STAT_VP) ? PKT_RX_VLAN_PKT : 0;
+ pkt_flags = (rx_status & IXGBE_RXD_STAT_VP) ? vlan_flags : 0;
#ifdef RTE_LIBRTE_IEEE1588
if (rx_status & IXGBE_RXD_STAT_TMST)
@@ -1287,6 +1287,7 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
uint32_t pkt_info[LOOK_AHEAD];
int i, j, nb_rx = 0;
uint32_t status;
+ uint64_t vlan_flags = rxq->vlan_flags;
/* get references to current descriptor and S/W ring entry */
rxdp = &rxq->rx_ring[rxq->rx_tail];
@@ -1328,7 +1329,8 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
mb->vlan_tci = rte_le_to_cpu_16(rxdp[j].wb.upper.vlan);
/* convert descriptor fields to rte mbuf flags */
- pkt_flags = rx_desc_status_to_pkt_flags(s[j]);
+ pkt_flags = rx_desc_status_to_pkt_flags(s[j],
+ vlan_flags);
pkt_flags |= rx_desc_error_to_pkt_flags(s[j]);
pkt_flags |= ixgbe_rxd_pkt_info_to_pkt_flags
((uint16_t)pkt_info[j]);
@@ -1544,6 +1546,7 @@ ixgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
uint16_t nb_rx;
uint16_t nb_hold;
uint64_t pkt_flags;
+ uint64_t vlan_flags;
nb_rx = 0;
nb_hold = 0;
@@ -1551,6 +1554,7 @@ ixgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
rx_id = rxq->rx_tail;
rx_ring = rxq->rx_ring;
sw_ring = rxq->sw_ring;
+ vlan_flags = rxq->vlan_flags;
while (nb_rx < nb_pkts) {
/*
* The order of operations here is important as the DD status
@@ -1660,7 +1664,7 @@ ixgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
/* Only valid if PKT_RX_VLAN_PKT set in pkt_flags */
rxm->vlan_tci = rte_le_to_cpu_16(rxd.wb.upper.vlan);
- pkt_flags = rx_desc_status_to_pkt_flags(staterr);
+ pkt_flags = rx_desc_status_to_pkt_flags(staterr, vlan_flags);
pkt_flags = pkt_flags | rx_desc_error_to_pkt_flags(staterr);
pkt_flags = pkt_flags |
ixgbe_rxd_pkt_info_to_pkt_flags((uint16_t)pkt_info);
@@ -1753,7 +1757,7 @@ ixgbe_fill_cluster_head_buf(
*/
head->vlan_tci = rte_le_to_cpu_16(desc->wb.upper.vlan);
pkt_info = rte_le_to_cpu_32(desc->wb.lower.lo_dword.data);
- pkt_flags = rx_desc_status_to_pkt_flags(staterr);
+ pkt_flags = rx_desc_status_to_pkt_flags(staterr, rxq->vlan_flags);
pkt_flags |= rx_desc_error_to_pkt_flags(staterr);
pkt_flags |= ixgbe_rxd_pkt_info_to_pkt_flags((uint16_t)pkt_info);
head->ol_flags = pkt_flags;
diff --git a/drivers/net/ixgbe/ixgbe_rxtx.h b/drivers/net/ixgbe/ixgbe_rxtx.h
index 3691a19..2608b36 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx.h
+++ b/drivers/net/ixgbe/ixgbe_rxtx.h
@@ -146,6 +146,8 @@ struct ixgbe_rx_queue {
uint8_t crc_len; /**< 0 if CRC stripped, 4 otherwise. */
uint8_t drop_en; /**< If not 0, set SRRCTL.Drop_En. */
uint8_t rx_deferred_start; /**< not in global dev start. */
+ /** flags to set in mbuf when a vlan is detected. */
+ uint64_t vlan_flags;
/** need to alloc dummy mbuf, for wraparound when scanning hw ring */
struct rte_mbuf fake_mbuf;
/** hold packets to return to application */
diff --git a/drivers/net/ixgbe/ixgbe_rxtx_vec.c b/drivers/net/ixgbe/ixgbe_rxtx_vec.c
index e97ea82..12190d2 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx_vec.c
+++ b/drivers/net/ixgbe/ixgbe_rxtx_vec.c
@@ -140,10 +140,9 @@ ixgbe_rxq_rearm(struct ixgbe_rx_queue *rxq)
*/
#ifdef RTE_IXGBE_RX_OLFLAGS_ENABLE
-#define VTAG_SHIFT (3)
-
static inline void
-desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
+desc_to_olflags_v(__m128i descs[4], uint8_t vlan_flags,
+ struct rte_mbuf **rx_pkts)
{
__m128i ptype0, ptype1, vtag0, vtag1;
union {
@@ -151,12 +150,6 @@ desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
uint64_t dword;
} vol;
- /* pkt type + vlan olflags mask */
- const __m128i pkttype_msk = _mm_set_epi16(
- 0x0000, 0x0000, 0x0000, 0x0000,
- PKT_RX_VLAN_PKT, PKT_RX_VLAN_PKT,
- PKT_RX_VLAN_PKT, PKT_RX_VLAN_PKT);
-
/* mask everything except rss type */
const __m128i rsstype_msk = _mm_set_epi16(
0x0000, 0x0000, 0x0000, 0x0000,
@@ -168,6 +161,19 @@ desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
PKT_RX_RSS_HASH, 0, PKT_RX_RSS_HASH, 0,
PKT_RX_RSS_HASH, PKT_RX_RSS_HASH, PKT_RX_RSS_HASH, 0);
+ /* mask everything except vlan present bit */
+ const __m128i vlan_msk = _mm_set_epi16(
+ 0x0000, 0x0000,
+ 0x0000, 0x0000,
+ IXGBE_RXD_STAT_VP, IXGBE_RXD_STAT_VP,
+ IXGBE_RXD_STAT_VP, IXGBE_RXD_STAT_VP);
+ /* map vlan present (0x8) to ol_flags */
+ const __m128i vlan_map = _mm_set_epi8(
+ 0, 0, 0, 0,
+ 0, 0, 0, vlan_flags,
+ 0, 0, 0, 0,
+ 0, 0, 0, 0);
+
ptype0 = _mm_unpacklo_epi16(descs[0], descs[1]);
ptype1 = _mm_unpacklo_epi16(descs[2], descs[3]);
vtag0 = _mm_unpackhi_epi16(descs[0], descs[1]);
@@ -178,8 +184,8 @@ desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts)
ptype0 = _mm_shuffle_epi8(rss_flags, ptype0);
vtag1 = _mm_unpacklo_epi32(vtag0, vtag1);
- vtag1 = _mm_srli_epi16(vtag1, VTAG_SHIFT);
- vtag1 = _mm_and_si128(vtag1, pkttype_msk);
+ vtag1 = _mm_and_si128(vtag1, vlan_msk);
+ vtag1 = _mm_shuffle_epi8(vlan_map, vtag1);
vtag1 = _mm_or_si128(ptype0, vtag1);
vol.dword = _mm_cvtsi128_si64(vtag1);
@@ -221,6 +227,7 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts,
0, 0 /* ignore pkt_type field */
);
__m128i dd_check, eop_check;
+ uint8_t vlan_flags;
/* nb_pkts shall be less equal than RTE_IXGBE_MAX_RX_BURST */
nb_pkts = RTE_MIN(nb_pkts, RTE_IXGBE_MAX_RX_BURST);
@@ -270,6 +277,10 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts,
*/
sw_ring = &rxq->sw_ring[rxq->rx_tail];
+ /* ensure these 2 flags are in the lower 8 bits */
+ RTE_BUILD_BUG_ON((PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED) > UINT8_MAX);
+ vlan_flags = rxq->vlan_flags & UINT8_MAX;
+
/* A. load 4 packet in one loop
* [A*. mask out 4 unused dirty field in desc]
* B. copy 4 mbuf point from swring to rx_pkts
@@ -330,7 +341,7 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts,
sterr_tmp1 = _mm_unpackhi_epi32(descs[1], descs[0]);
/* set ol_flags with vlan packet type */
- desc_to_olflags_v(descs, &rx_pkts[pos]);
+ desc_to_olflags_v(descs, vlan_flags, &rx_pkts[pos]);
/* D.2 pkt 3,4 set in_port/nb_seg and remove crc */
pkt_mb4 = _mm_add_epi16(pkt_mb4, crc_adjust);
diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index 29bfcec..d5b2286 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -1051,7 +1051,8 @@ mlx5_rx_burst_sp(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
pkt_buf->ol_flags = rxq_cq_to_ol_flags(rxq, flags);
#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
if (flags & IBV_EXP_CQ_RX_CVLAN_STRIPPED_V1) {
- pkt_buf->ol_flags |= PKT_RX_VLAN_PKT;
+ pkt_buf->ol_flags |= PKT_RX_VLAN_PKT |
+ PKT_RX_VLAN_STRIPPED;
pkt_buf->vlan_tci = vlan_tci;
}
#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
@@ -1207,7 +1208,8 @@ mlx5_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n)
seg->ol_flags = rxq_cq_to_ol_flags(rxq, flags);
#ifdef HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS
if (flags & IBV_EXP_CQ_RX_CVLAN_STRIPPED_V1) {
- seg->ol_flags |= PKT_RX_VLAN_PKT;
+ seg->ol_flags |= PKT_RX_VLAN_PKT |
+ PKT_RX_VLAN_STRIPPED;
seg->vlan_tci = vlan_tci;
}
#endif /* HAVE_EXP_DEVICE_ATTR_VLAN_OFFLOADS */
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index ea5a2a3..5c9f350 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -1800,7 +1800,7 @@ nfp_net_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
if ((rxds->rxd.flags & PCIE_DESC_RX_VLAN) &&
(hw->ctrl & NFP_NET_CFG_CTRL_RXVLAN)) {
mb->vlan_tci = rte_cpu_to_le_32(rxds->rxd.vlan);
- mb->ol_flags |= PKT_RX_VLAN_PKT;
+ mb->ol_flags |= PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED;
}
/* Adding the mbuff to the mbuff array passed by the app */
diff --git a/drivers/net/vmxnet3/vmxnet3_rxtx.c b/drivers/net/vmxnet3/vmxnet3_rxtx.c
index 9fe8752..ccafc0c 100644
--- a/drivers/net/vmxnet3/vmxnet3_rxtx.c
+++ b/drivers/net/vmxnet3/vmxnet3_rxtx.c
@@ -579,7 +579,7 @@ vmxnet3_rx_offload(const Vmxnet3_RxCompDesc *rcd, struct rte_mbuf *rxm)
{
/* Check for hardware stripped VLAN tag */
if (rcd->ts) {
- rxm->ol_flags |= PKT_RX_VLAN_PKT;
+ rxm->ol_flags |= (PKT_RX_VLAN_PKT | PKT_RX_VLAN_STRIPPED);
rxm->vlan_tci = rte_le_to_cpu_16((uint16_t)rcd->tci);
}
diff --git a/lib/librte_mbuf/rte_mbuf.c b/lib/librte_mbuf/rte_mbuf.c
index eec1456..2ece742 100644
--- a/lib/librte_mbuf/rte_mbuf.c
+++ b/lib/librte_mbuf/rte_mbuf.c
@@ -258,8 +258,10 @@ const char *rte_get_rx_ol_flag_name(uint64_t mask)
/* case PKT_RX_HBUF_OVERFLOW: return "PKT_RX_HBUF_OVERFLOW"; */
/* case PKT_RX_RECIP_ERR: return "PKT_RX_RECIP_ERR"; */
/* case PKT_RX_MAC_ERR: return "PKT_RX_MAC_ERR"; */
+ case PKT_RX_VLAN_STRIPPED: return "PKT_RX_VLAN_STRIPPED";
case PKT_RX_IEEE1588_PTP: return "PKT_RX_IEEE1588_PTP";
case PKT_RX_IEEE1588_TMST: return "PKT_RX_IEEE1588_TMST";
+ case PKT_RX_QINQ_STRIPPED: return "PKT_RX_QINQ_STRIPPED";
default: return NULL;
}
}
diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index 11fa06d..8798c41 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -79,7 +79,15 @@ extern "C" {
* Keep these flags synchronized with rte_get_rx_ol_flag_name() and
* rte_get_tx_ol_flag_name().
*/
-#define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
+
+/**
+ * RX packet is a 802.1q VLAN packet. This flag was set by PMDs when
+ * the packet is recognized as a VLAN, but the behavior between PMDs
+ * was not the same. This flag is kept for some time to avoid breaking
+ * applications and should be replaced by PKT_RX_VLAN_STRIPPED.
+ */
+#define PKT_RX_VLAN_PKT (1ULL << 0)
+
#define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
#define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
#define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
@@ -89,11 +97,37 @@ extern "C" {
#define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
#define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
#define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
+
+/**
+ * A vlan has been stripped by the hardware and its tci is saved in
+ * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
+ * in the RX configuration of the PMD.
+ */
+#define PKT_RX_VLAN_STRIPPED (1ULL << 6)
+
+/* hole, some bits can be reused here */
+
#define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
#define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
#define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
#define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
-#define PKT_RX_QINQ_PKT (1ULL << 15) /**< RX packet with double VLAN stripped. */
+
+/**
+ * The 2 vlans have been stripped by the hardware and their tci are
+ * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
+ * This can only happen if vlan stripping is enabled in the RX
+ * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
+ * must also be set.
+ */
+#define PKT_RX_QINQ_STRIPPED (1ULL << 15)
+
+/**
+ * Deprecated.
+ * RX packet with double VLAN stripped.
+ * This flag is replaced by PKT_RX_QINQ_STRIPPED.
+ */
+#define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
+
/* add new RX flags here */
/* add new TX flags here */
@@ -761,7 +795,10 @@ struct rte_mbuf {
/*
* The packet type, which is the combination of outer/inner L2, L3, L4
- * and tunnel types.
+ * and tunnel types. The packet_type is about data really present in the
+ * mbuf. Example: if vlan stripping is enabled, a received vlan packet
+ * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
+ * vlan is stripped from the data.
*/
union {
uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
@@ -778,7 +815,8 @@ struct rte_mbuf {
uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
uint16_t data_len; /**< Amount of data in segment buffer. */
- uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
+ /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
+ uint16_t vlan_tci;
union {
uint32_t rss; /**< RSS hash result if RSS enabled */
@@ -804,7 +842,8 @@ struct rte_mbuf {
uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
- uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */
+ /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
+ uint16_t vlan_tci_outer;
/* second cache line - fields only used in slow path or on TX */
MARKER cacheline1 __rte_cache_min_aligned;
--
2.8.0.rc3
^ permalink raw reply [relevance 2%]
* Re: [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager
2016-06-15 7:47 3% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager David Hunt
2016-06-15 7:47 1% ` [dpdk-dev] [PATCH v12 1/3] mempool: support external mempool operations David Hunt
@ 2016-06-15 10:13 0% ` Jan Viktorin
2016-06-16 12:30 3% ` [dpdk-dev] [PATCH v13 " David Hunt
2 siblings, 0 replies; 200+ results
From: Jan Viktorin @ 2016-06-15 10:13 UTC (permalink / raw)
To: David Hunt; +Cc: dev, olivier.matz, jerin.jacob, shreyansh.jain
Hi,
I've got one last question. Initially, I was interested in creating
my own external memory provider based on a Linux Kernel driver.
So, I've got an opened file descriptor that points to a device which
can mmap a memory regions for me.
...
int fd = open("/dev/uio0" ...);
...
rte_mempool *pool = rte_mempool_create_empty(...);
rte_mempool_set_ops_byname(pool, "uio_allocator_ops");
I am not sure how to pass the file descriptor pointer. I thought it would
be possible by the rte_mempool_alloc but it's not... Is it possible
to solve this case?
The allocator is device-specific.
Regards
Jan
On Wed, 15 Jun 2016 08:47:01 +0100
David Hunt <david.hunt@intel.com> wrote:
> Here's the latest version of the External Mempool Manager patchset.
> It's re-based on top of the latest head as of 14/6/2016, including
> Olivier's 35-part patch series on mempool re-org [1]
>
> [1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
>
> v12 changes:
>
> * Fixed a comment (function pram h -> ops)
> * Fixed a typo in mempool docs (callbacki)
>
> v11 changes:
>
> * Fixed comments (added '.' where needed for consistency)
> * removed ABI breakage notice for mempool manager in deprecation.rst
> * Added description of the external mempool manager functionality to
> doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
> * renamed rte_mempool_default.c to rte_mempool_ring.c
>
> v10 changes:
>
> * changed the _put/_get op names to _enqueue/_dequeue to be consistent
> with the function names
> * some rte_errno cleanup
> * comment tweaks about when to set pool_data
> * removed an un-needed check for ops->alloc == NULL
>
> v9 changes:
>
> * added a check for NULL alloc in rte_mempool_ops_register
> * rte_mempool_alloc_t now returns int instead of void*
> * fixed some comment typo's
> * removed some unneeded typecasts
> * changed a return NULL to return -EEXIST in rte_mempool_ops_register
> * fixed rte_mempool_version.map file so builds ok as shared libs
> * moved flags check from rte_mempool_create_empty to rte_mempool_create
>
> v8 changes:
>
> * merged first three patches in the series into one.
> * changed parameters to ops callback to all be rte_mempool pointer
> rather than than pointer to opaque data or uint64.
> * comment fixes.
> * fixed parameter to _free function (was inconsistent).
> * changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
>
> v7 changes:
>
> * Changed rte_mempool_handler_table to rte_mempool_ops_table
> * Changed hander_idx to ops_index in rte_mempool struct
> * Reworked comments in rte_mempool.h around ops functions
> * Changed rte_mempool_hander.c to rte_mempool_ops.c
> * Changed all functions containing _handler_ to _ops_
> * Now there is no mention of 'handler' left
> * Other small changes out of review of mailing list
>
> v6 changes:
>
> * Moved the flags handling from rte_mempool_create_empty to
> rte_mempool_create, as it's only there for backward compatibility
> * Various comment additions and cleanup
> * Renamed rte_mempool_handler to rte_mempool_ops
> * Added a union for *pool and u64 pool_id in struct rte_mempool
> * split the original patch into a few parts for easier review.
> * rename functions with _ext_ to _ops_.
> * addressed review comments
> * renamed put and get functions to enqueue and dequeue
> * changed occurences of rte_mempool_ops to const, as they
> contain function pointers (security)
> * split out the default external mempool handler into a separate
> patch for easier review
>
> v5 changes:
> * rebasing, as it is dependent on another patch series [1]
>
> v4 changes (Olivier Matz):
> * remove the rte_mempool_create_ext() function. To change the handler, the
> user has to do the following:
> - mp = rte_mempool_create_empty()
> - rte_mempool_set_handler(mp, "my_handler")
> - rte_mempool_populate_default(mp)
> This avoids to add another function with more than 10 arguments, duplicating
> the doxygen comments
> * change the api of rte_mempool_alloc_t: only the mempool pointer is required
> as all information is available in it
> * change the api of rte_mempool_free_t: remove return value
> * move inline wrapper functions from the .c to the .h (else they won't be
> inlined). This implies to have one header file (rte_mempool.h), or it
> would have generate cross dependencies issues.
> * remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
> to the use of && instead of &)
> * fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
> * fix build with shared libraries (global handler has to be declared in
> the .map file)
> * rationalize #include order
> * remove unused function rte_mempool_get_handler_name()
> * rename some structures, fields, functions
> * remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
> from Yuanhan)
> * test the ext mempool handler in the same file than standard mempool tests,
> avoiding to duplicate the code
> * rework the custom handler in mempool_test
> * rework a bit the patch selecting default mbuf pool handler
> * fix some doxygen comments
>
> v3 changes:
> * simplified the file layout, renamed to rte_mempool_handler.[hc]
> * moved the default handlers into rte_mempool_default.c
> * moved the example handler out into app/test/test_ext_mempool.c
> * removed is_mc/is_mp change, slight perf degredation on sp cached operation
> * removed stack hanler, may re-introduce at a later date
> * Changes out of code reviews
>
> v2 changes:
> * There was a lot of duplicate code between rte_mempool_xmem_create and
> rte_mempool_create_ext. This has now been refactored and is now
> hopefully cleaner.
> * The RTE_NEXT_ABI define is now used to allow building of the library
> in a format that is compatible with binaries built against previous
> versions of DPDK.
> * Changes out of code reviews. Hopefully I've got most of them included.
>
> The External Mempool Manager is an extension to the mempool API that allows
> users to add and use an external mempool manager, which allows external memory
> subsystems such as external hardware memory management systems and software
> based memory allocators to be used with DPDK.
>
> The existing API to the internal DPDK mempool manager will remain unchanged
> and will be backward compatible. However, there will be an ABI breakage, as
> the mempool struct is changing. These changes are all contained withing
> RTE_NEXT_ABI defs, and the current or next code can be changed with
> the CONFIG_RTE_NEXT_ABI config setting
>
> There are two aspects to external mempool manager.
> 1. Adding the code for your new mempool operations (ops). This is
> achieved by adding a new mempool ops source file into the
> librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
> 2. Using the new API to call rte_mempool_create_empty and
> rte_mempool_set_ops_byname to create a new mempool
> using the name parameter to identify which ops to use.
>
> New API calls added
> 1. A new rte_mempool_create_empty() function
> 2. rte_mempool_set_ops_byname() which sets the mempool's ops (functions)
> 3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
> which populates the mempool using the relevant ops
>
> Several external mempool managers may be used in the same application. A new
> mempool can then be created by using the new rte_mempool_create_empty function,
> then calling rte_mempool_set_ops_byname to point the mempool to the relevant
> mempool manager callback structure.
>
> Legacy applications will continue to use the old rte_mempool_create API call,
> which uses a ring based mempool manager by default. These applications
> will need to be modified to use a new external mempool manager.
>
> The external mempool manager needs to provide the following functions.
> 1. alloc - allocates the mempool memory, and adds each object onto a ring
> 2. enqueue - puts an object back into the mempool once an application has
> finished with it
> 3. dequeue - gets an object from the mempool for use by the application
> 4. get_count - gets the number of available objects in the mempool
> 5. free - frees the mempool memory
>
> Every time an enqueue/dequeue/get_count is called from the application/PMD,
> the callback for that mempool is called. These functions are in the fastpath,
> and any unoptimised ops may limit performance.
>
> The new APIs are as follows:
>
> 1. rte_mempool_create_empty
>
> struct rte_mempool *
> rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
> unsigned cache_size, unsigned private_data_size,
> int socket_id, unsigned flags);
>
> 2. rte_mempool_set_ops_byname()
>
> int
> rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
>
> 3. rte_mempool_populate_default()
>
> int rte_mempool_populate_default(struct rte_mempool *mp);
>
> 4. rte_mempool_populate_anon()
>
> int rte_mempool_populate_anon(struct rte_mempool *mp);
>
> Please see rte_mempool.h for further information on the parameters.
>
>
> The important thing to note is that the mempool ops struct is passed by name
> to rte_mempool_set_ops_byname, which looks through the ops struct array to
> get the ops_index, which is then stored in the rte_memool structure. This
> allow multiple processes to use the same mempool, as the function pointers
> are accessed via ops index.
>
> The mempool ops structure contains callbacks to the implementation of
> the ops function, and is set up for registration as follows:
>
> static const struct rte_mempool_ops ops_sp_mc = {
> .name = "ring_sp_mc",
> .alloc = rte_mempool_common_ring_alloc,
> .enqueue = common_ring_sp_enqueue,
> .dequeue = common_ring_mc_dequeue,
> .get_count = common_ring_get_count,
> .free = common_ring_free,
> };
>
> And then the following macro will register the ops in the array of ops
> structures
>
> REGISTER_MEMPOOL_OPS(ops_mp_mc);
>
> For an example of API usage, please see app/test/test_mempool.c, which
> implements a rudimentary "custom_handler" mempool manager using simple mallocs
> for each mempool object. This file also contains the callbacks and self
> registration for the new handler.
>
> David Hunt (2):
> mempool: support external mempool operations
> mbuf: make default mempool ops configurable at build
>
> Olivier Matz (1):
> app/test: test external mempool manager
>
>
--
Jan Viktorin E-mail: Viktorin@RehiveTech.com
System Architect Web: www.RehiveTech.com
RehiveTech
Brno, Czech Republic
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v9 8/8] doc: update doc for packet capture framework
2016-06-15 5:44 0% ` Pattan, Reshma
@ 2016-06-15 8:24 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-15 8:24 UTC (permalink / raw)
To: Pattan, Reshma; +Cc: dev, Mcnamara, John
2016-06-15 05:44, Pattan, Reshma:
> > > ABI Changes
> > > -----------
> > > @@ -146,6 +156,9 @@ ABI Changes
> > > * The ``rte_port_source_params`` structure has new fields to support PCAP
> > file.
> > > It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
> > >
> > > +* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and
> > ``nb_tx_queues``
> > > + to support number of queues configured by software.
> >
> > There was no deprecation notice in 16.04 for this ABI change.
>
> Deprecation notice and relevant planed changes were sent as RFC during the start of 16.07 , please find the below is the link for the same.
> http://dpdk.org/dev/patchwork/patch/12033/
> http://dpdk.org/dev/patchwork/patch/12034/
Yes there was no notice in 16.04.
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v12 1/3] mempool: support external mempool operations
2016-06-15 7:47 3% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager David Hunt
@ 2016-06-15 7:47 1% ` David Hunt
2016-06-15 10:13 0% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager Jan Viktorin
2016-06-16 12:30 3% ` [dpdk-dev] [PATCH v13 " David Hunt
2 siblings, 0 replies; 200+ results
From: David Hunt @ 2016-06-15 7:47 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain, David Hunt
Until now, the objects stored in a mempool were internally stored in a
ring. This patch introduces the possibility to register external handlers
replacing the ring.
The default behavior remains unchanged, but calling the new function
rte_mempool_set_ops_byname() right after rte_mempool_create_empty() allows
the user to change the handler that will be used when populating
the mempool.
This patch also adds a set of default ops (function callbacks) based
on rte_ring.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---
app/test/test_mempool_perf.c | 1 -
doc/guides/prog_guide/mempool_lib.rst | 31 +++-
doc/guides/rel_notes/deprecation.rst | 9 --
lib/librte_mempool/Makefile | 2 +
lib/librte_mempool/rte_mempool.c | 66 +++-----
lib/librte_mempool/rte_mempool.h | 251 ++++++++++++++++++++++++++---
lib/librte_mempool/rte_mempool_ops.c | 148 +++++++++++++++++
lib/librte_mempool/rte_mempool_ring.c | 161 ++++++++++++++++++
lib/librte_mempool/rte_mempool_version.map | 13 +-
9 files changed, 601 insertions(+), 81 deletions(-)
create mode 100644 lib/librte_mempool/rte_mempool_ops.c
create mode 100644 lib/librte_mempool/rte_mempool_ring.c
diff --git a/app/test/test_mempool_perf.c b/app/test/test_mempool_perf.c
index c5e3576..c5f8455 100644
--- a/app/test/test_mempool_perf.c
+++ b/app/test/test_mempool_perf.c
@@ -161,7 +161,6 @@ per_lcore_mempool_test(__attribute__((unused)) void *arg)
n_get_bulk);
if (unlikely(ret < 0)) {
rte_mempool_dump(stdout, mp);
- rte_ring_dump(stdout, mp->ring);
/* in this case, objects are lost... */
return -1;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index c3afc2e..2e3116e 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -34,7 +34,7 @@ Mempool Library
===============
A memory pool is an allocator of a fixed-sized object.
-In the DPDK, it is identified by name and uses a ring to store free objects.
+In the DPDK, it is identified by name and uses a ring or an external mempool manager to store free objects.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
@@ -127,6 +127,35 @@ The maximum size of the cache is static and is defined at compilation time (CONF
A mempool in Memory with its Associated Ring
+External Mempool Manager
+------------------------
+
+This allows external memory subsystems, such as external hardware memory
+management systems and software based memory allocators, to be used with DPDK.
+
+There are two aspects to external mempool manager.
+
+* Adding the code for your new mempool operations (ops). This is achieved by
+ adding a new mempool ops code, and using the ``REGISTER_MEMPOOL_OPS`` macro.
+
+* Using the new API to call ``rte_mempool_create_empty()`` and
+ ``rte_mempool_set_ops_byname()`` to create a new mempool and specifying which
+ ops to use.
+
+Several external mempool managers may be used in the same application. A new
+mempool can be created by using the ``rte_mempool_create_empty()`` function,
+then using ``rte_mempool_set_ops_byname()`` to point the mempool to the
+relevant mempool manager callback (ops) structure.
+
+Legacy applications may continue to use the old ``rte_mempool_create()`` API
+call, which uses a ring based mempool manager by default. These applications
+will need to be modified to use a new external mempool manager.
+
+For applications that use ``rte_pktmbuf_create()``, there is a config setting
+(``RTE_MBUF_DEFAULT_MEMPOOL_OPS``) that allows the application to make use of
+an external mempool manager.
+
+
Use Cases
---------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 7d947ae..c415095 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -39,15 +39,6 @@ Deprecation Notices
compact API. The ones that remain are backwards compatible and use the
per-lcore default cache if available. This change targets release 16.07.
-* The rte_mempool struct will be changed in 16.07 to facilitate the new
- external mempool manager functionality.
- The ring element will be replaced with a more generic 'pool' opaque pointer
- to allow new mempool handlers to use their own user-defined mempool
- layout. Also newly added to rte_mempool is a handler index.
- The existing API will be backward compatible, but there will be new API
- functions added to facilitate the creation of mempools using an external
- handler. The 16.07 release will contain these changes.
-
* A librte_vhost public structures refactor is planned for DPDK 16.07
that requires both ABI and API change.
The proposed refactor would expose DPDK vhost dev to applications as
diff --git a/lib/librte_mempool/Makefile b/lib/librte_mempool/Makefile
index 43423e0..a4c089e 100644
--- a/lib/librte_mempool/Makefile
+++ b/lib/librte_mempool/Makefile
@@ -42,6 +42,8 @@ LIBABIVER := 2
# all source are stored in SRCS-y
SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ring.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_MEMPOOL)-include := rte_mempool.h
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index 22a5645..ac40cb3 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -148,7 +148,7 @@ mempool_add_elem(struct rte_mempool *mp, void *obj, phys_addr_t physaddr)
#endif
/* enqueue in ring */
- rte_ring_sp_enqueue(mp->ring, obj);
+ rte_mempool_ops_enqueue_bulk(mp, &obj, 1);
}
/* call obj_cb() for each mempool element */
@@ -303,40 +303,6 @@ rte_mempool_xmem_usage(__rte_unused void *vaddr, uint32_t elt_num,
return (size_t)paddr_idx << pg_shift;
}
-/* create the internal ring */
-static int
-rte_mempool_ring_create(struct rte_mempool *mp)
-{
- int rg_flags = 0, ret;
- char rg_name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
-
- ret = snprintf(rg_name, sizeof(rg_name),
- RTE_MEMPOOL_MZ_FORMAT, mp->name);
- if (ret < 0 || ret >= (int)sizeof(rg_name))
- return -ENAMETOOLONG;
-
- /* ring flags */
- if (mp->flags & MEMPOOL_F_SP_PUT)
- rg_flags |= RING_F_SP_ENQ;
- if (mp->flags & MEMPOOL_F_SC_GET)
- rg_flags |= RING_F_SC_DEQ;
-
- /* Allocate the ring that will be used to store objects.
- * Ring functions will return appropriate errors if we are
- * running as a secondary process etc., so no checks made
- * in this function for that condition.
- */
- r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
- mp->socket_id, rg_flags);
- if (r == NULL)
- return -rte_errno;
-
- mp->ring = r;
- mp->flags |= MEMPOOL_F_RING_CREATED;
- return 0;
-}
-
/* free a memchunk allocated with rte_memzone_reserve() */
static void
rte_mempool_memchunk_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
@@ -354,7 +320,7 @@ rte_mempool_free_memchunks(struct rte_mempool *mp)
void *elt;
while (!STAILQ_EMPTY(&mp->elt_list)) {
- rte_ring_sc_dequeue(mp->ring, &elt);
+ rte_mempool_ops_dequeue_bulk(mp, &elt, 1);
(void)elt;
STAILQ_REMOVE_HEAD(&mp->elt_list, next);
mp->populated_size--;
@@ -386,9 +352,9 @@ rte_mempool_populate_phys(struct rte_mempool *mp, char *vaddr,
int ret;
/* create the internal ring if not already done */
- if ((mp->flags & MEMPOOL_F_RING_CREATED) == 0) {
- ret = rte_mempool_ring_create(mp);
- if (ret < 0)
+ if ((mp->flags & MEMPOOL_F_POOL_CREATED) == 0) {
+ ret = rte_mempool_ops_alloc(mp);
+ if (ret != 0)
return ret;
}
@@ -703,7 +669,7 @@ rte_mempool_free(struct rte_mempool *mp)
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_mempool_free_memchunks(mp);
- rte_ring_free(mp->ring);
+ rte_mempool_ops_free(mp);
rte_memzone_free(mp->mz);
}
@@ -815,6 +781,7 @@ rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, 0));
te->data = mp;
+
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_INSERT_TAIL(mempool_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
@@ -844,6 +811,19 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
if (mp == NULL)
return NULL;
+ /*
+ * Since we have 4 combinations of the SP/SC/MP/MC examine the flags to
+ * set the correct index into the table of ops structs.
+ */
+ if (flags & (MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET))
+ rte_mempool_set_ops_byname(mp, "ring_sp_sc");
+ else if (flags & MEMPOOL_F_SP_PUT)
+ rte_mempool_set_ops_byname(mp, "ring_sp_mc");
+ else if (flags & MEMPOOL_F_SC_GET)
+ rte_mempool_set_ops_byname(mp, "ring_mp_sc");
+ else
+ rte_mempool_set_ops_byname(mp, "ring_mp_mc");
+
/* call the mempool priv initializer */
if (mp_init)
mp_init(mp, mp_init_arg);
@@ -930,7 +910,7 @@ rte_mempool_count(const struct rte_mempool *mp)
unsigned count;
unsigned lcore_id;
- count = rte_ring_count(mp->ring);
+ count = rte_mempool_ops_get_count(mp);
if (mp->cache_size == 0)
return count;
@@ -1119,7 +1099,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
fprintf(f, "mempool <%s>@%p\n", mp->name, mp);
fprintf(f, " flags=%x\n", mp->flags);
- fprintf(f, " ring=<%s>@%p\n", mp->ring->name, mp->ring);
+ fprintf(f, " pool=%p\n", mp->pool_data);
fprintf(f, " phys_addr=0x%" PRIx64 "\n", mp->mz->phys_addr);
fprintf(f, " nb_mem_chunks=%u\n", mp->nb_mem_chunks);
fprintf(f, " size=%"PRIu32"\n", mp->size);
@@ -1140,7 +1120,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
}
cache_count = rte_mempool_dump_cache(f, mp);
- common_count = rte_ring_count(mp->ring);
+ common_count = rte_mempool_ops_get_count(mp);
if ((cache_count + common_count) > mp->size)
common_count = mp->size - cache_count;
fprintf(f, " common_pool_count=%u\n", common_count);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 60339bd..92deb42 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -67,6 +67,7 @@
#include <inttypes.h>
#include <sys/queue.h>
+#include <rte_spinlock.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_lcore.h>
@@ -203,10 +204,13 @@ struct rte_mempool_memhdr {
*/
struct rte_mempool {
char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
- struct rte_ring *ring; /**< Ring to store objects. */
- const struct rte_memzone *mz; /**< Memzone where pool is allocated */
+ union {
+ void *pool_data; /**< Ring or pool to store objects. */
+ uint64_t pool_id; /**< External mempool identifier. */
+ };
+ const struct rte_memzone *mz; /**< Memzone where pool is alloc'd. */
int flags; /**< Flags of the mempool. */
- int socket_id; /**< Socket id passed at mempool creation. */
+ int socket_id; /**< Socket id passed at create. */
uint32_t size; /**< Max size of the mempool. */
uint32_t cache_size; /**< Size of per-lcore local cache. */
uint32_t cache_flushthresh;
@@ -217,6 +221,14 @@ struct rte_mempool {
uint32_t trailer_size; /**< Size of trailer (after elt). */
unsigned private_data_size; /**< Size of private data. */
+ /**
+ * Index into rte_mempool_ops_table array of mempool ops
+ * structs, which contain callback function pointers.
+ * We're using an index here rather than pointers to the callbacks
+ * to facilitate any secondary processes that may want to use
+ * this mempool.
+ */
+ int32_t ops_index;
struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
@@ -235,7 +247,7 @@ struct rte_mempool {
#define MEMPOOL_F_NO_CACHE_ALIGN 0x0002 /**< Do not align objs on cache lines.*/
#define MEMPOOL_F_SP_PUT 0x0004 /**< Default put is "single-producer".*/
#define MEMPOOL_F_SC_GET 0x0008 /**< Default get is "single-consumer".*/
-#define MEMPOOL_F_RING_CREATED 0x0010 /**< Internal: ring is created */
+#define MEMPOOL_F_POOL_CREATED 0x0010 /**< Internal: pool is created. */
#define MEMPOOL_F_NO_PHYS_CONTIG 0x0020 /**< Don't need physically contiguous objs. */
/**
@@ -325,6 +337,212 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
#define __mempool_check_cookies(mp, obj_table_const, n, free) do {} while(0)
#endif /* RTE_LIBRTE_MEMPOOL_DEBUG */
+#define RTE_MEMPOOL_OPS_NAMESIZE 32 /**< Max length of ops struct name. */
+
+/**
+ * Prototype for implementation specific data provisioning function.
+ *
+ * The function should provide the implementation specific memory for
+ * for use by the other mempool ops functions in a given mempool ops struct.
+ * E.g. the default ops provides an instance of the rte_ring for this purpose.
+ * it will most likely point to a different type of data structure, and
+ * will be transparent to the application programmer.
+ * This function should set mp->pool_data.
+ */
+typedef int (*rte_mempool_alloc_t)(struct rte_mempool *mp);
+
+/**
+ * Free the opaque private data pointed to by mp->pool_data pointer.
+ */
+typedef void (*rte_mempool_free_t)(struct rte_mempool *mp);
+
+/**
+ * Enqueue an object into the external pool.
+ */
+typedef int (*rte_mempool_enqueue_t)(struct rte_mempool *mp,
+ void * const *obj_table, unsigned int n);
+
+/**
+ * Dequeue an object from the external pool.
+ */
+typedef int (*rte_mempool_dequeue_t)(struct rte_mempool *mp,
+ void **obj_table, unsigned int n);
+
+/**
+ * Return the number of available objects in the external pool.
+ */
+typedef unsigned (*rte_mempool_get_count)(const struct rte_mempool *mp);
+
+/** Structure defining mempool operations structure */
+struct rte_mempool_ops {
+ char name[RTE_MEMPOOL_OPS_NAMESIZE]; /**< Name of mempool ops struct. */
+ rte_mempool_alloc_t alloc; /**< Allocate private data. */
+ rte_mempool_free_t free; /**< Free the external pool. */
+ rte_mempool_enqueue_t enqueue; /**< Enqueue an object. */
+ rte_mempool_dequeue_t dequeue; /**< Dequeue an object. */
+ rte_mempool_get_count get_count; /**< Get qty of available objs. */
+} __rte_cache_aligned;
+
+#define RTE_MEMPOOL_MAX_OPS_IDX 16 /**< Max registered ops structs */
+
+/**
+ * Structure storing the table of registered ops structs, each of which contain
+ * the function pointers for the mempool ops functions.
+ * Each process has its own storage for this ops struct array so that
+ * the mempools can be shared across primary and secondary processes.
+ * The indices used to access the array are valid across processes, whereas
+ * any function pointers stored directly in the mempool struct would not be.
+ * This results in us simply having "ops_index" in the mempool struct.
+ */
+struct rte_mempool_ops_table {
+ rte_spinlock_t sl; /**< Spinlock for add/delete. */
+ uint32_t num_ops; /**< Number of used ops structs in the table. */
+ /**
+ * Storage for all possible ops structs.
+ */
+ struct rte_mempool_ops ops[RTE_MEMPOOL_MAX_OPS_IDX];
+} __rte_cache_aligned;
+
+/** Array of registered ops structs. */
+extern struct rte_mempool_ops_table rte_mempool_ops_table;
+
+/**
+ * @internal Get the mempool ops struct from its index.
+ *
+ * @param ops_index
+ * The index of the ops struct in the ops struct table. It must be a valid
+ * index: (0 <= idx < num_ops).
+ * @return
+ * The pointer to the ops struct in the table.
+ */
+static inline struct rte_mempool_ops *
+rte_mempool_ops_get(int ops_index)
+{
+ RTE_VERIFY(ops_index < RTE_MEMPOOL_MAX_OPS_IDX);
+
+ return &rte_mempool_ops_table.ops[ops_index];
+}
+
+/**
+ * @internal Wrapper for mempool_ops alloc callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * - 0: Success; successfully allocated mempool pool_data.
+ * - <0: Error; code of alloc function.
+ */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp);
+
+/**
+ * @internal Wrapper for mempool_ops get callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to get.
+ * @return
+ * - 0: Success; got n objects.
+ * - <0: Error; code of get function.
+ */
+static inline int
+rte_mempool_ops_dequeue_bulk(struct rte_mempool *mp,
+ void **obj_table, unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->dequeue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops put callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to put.
+ * @return
+ * - 0: Success; n objects supplied.
+ * - <0: Error; code of put function.
+ */
+static inline int
+rte_mempool_ops_enqueue_bulk(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->enqueue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops get_count callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * The number of available objects in the external pool.
+ */
+unsigned
+rte_mempool_ops_get_count(const struct rte_mempool *mp);
+
+/**
+ * @internal wrapper for mempool_ops free callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ */
+void
+rte_mempool_ops_free(struct rte_mempool *mp);
+
+/**
+ * Set the ops of a mempool.
+ *
+ * This can only be done on a mempool that is not populated, i.e. just after
+ * a call to rte_mempool_create_empty().
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param name
+ * Name of the ops structure to use for this mempool.
+ * @return
+ * - 0: Success; the mempool is now using the requested ops functions.
+ * - -EINVAL - Invalid ops struct name provided.
+ * - -EEXIST - mempool already has an ops struct assigned.
+ */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
+
+/**
+ * Register mempool operations.
+ *
+ * @param ops
+ * Pointer to an ops structure to register.
+ * @return
+ * - >=0: Success; return the index of the ops struct in the table.
+ * - -EINVAL - some missing callbacks while registering ops struct.
+ * - -ENOSPC - the maximum number of ops structs has been reached.
+ */
+int rte_mempool_ops_register(const struct rte_mempool_ops *ops);
+
+/**
+ * Macro to statically register the ops of an external mempool manager.
+ * Note that the rte_mempool_ops_register fails silently here when
+ * more then RTE_MEMPOOL_MAX_OPS_IDX is registered.
+ */
+#define MEMPOOL_REGISTER_OPS(ops) \
+ void mp_hdlr_init_##ops(void); \
+ void __attribute__((constructor, used)) mp_hdlr_init_##ops(void)\
+ { \
+ rte_mempool_ops_register(&ops); \
+ }
+
/**
* An object callback function for mempool.
*
@@ -774,7 +992,7 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
cache->len += n;
if (cache->len >= flushthresh) {
- rte_ring_mp_enqueue_bulk(mp->ring, &cache->objs[cache_size],
+ rte_mempool_ops_enqueue_bulk(mp, &cache->objs[cache_size],
cache->len - cache_size);
cache->len = cache_size;
}
@@ -785,19 +1003,10 @@ ring_enqueue:
/* push remaining objects in ring */
#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
- if (is_mp) {
- if (rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
- else {
- if (rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
+ if (rte_mempool_ops_enqueue_bulk(mp, obj_table, n) < 0)
+ rte_panic("cannot put objects in mempool\n");
#else
- if (is_mp)
- rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n);
- else
- rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n);
+ rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
#endif
}
@@ -945,7 +1154,8 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
uint32_t req = n + (cache_size - cache->len);
/* How many do we require i.e. number to fill the cache + the request */
- ret = rte_ring_mc_dequeue_bulk(mp->ring, &cache->objs[cache->len], req);
+ ret = rte_mempool_ops_dequeue_bulk(mp,
+ &cache->objs[cache->len], req);
if (unlikely(ret < 0)) {
/*
* In the offchance that we are buffer constrained,
@@ -972,10 +1182,7 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
ring_dequeue:
/* get remaining objects from ring */
- if (is_mc)
- ret = rte_ring_mc_dequeue_bulk(mp->ring, obj_table, n);
- else
- ret = rte_ring_sc_dequeue_bulk(mp->ring, obj_table, n);
+ ret = rte_mempool_ops_dequeue_bulk(mp, obj_table, n);
if (ret < 0)
__MEMPOOL_STAT_ADD(mp, get_fail, n);
diff --git a/lib/librte_mempool/rte_mempool_ops.c b/lib/librte_mempool/rte_mempool_ops.c
new file mode 100644
index 0000000..9328b77
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ops.c
@@ -0,0 +1,148 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Copyright(c) 2016 6WIND S.A.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_mempool.h>
+#include <rte_errno.h>
+
+/* indirect jump table to support external memory pools. */
+struct rte_mempool_ops_table rte_mempool_ops_table = {
+ .sl = RTE_SPINLOCK_INITIALIZER,
+ .num_ops = 0
+};
+
+/* add a new ops struct in rte_mempool_ops_table, return its index. */
+int
+rte_mempool_ops_register(const struct rte_mempool_ops *h)
+{
+ struct rte_mempool_ops *ops;
+ int16_t ops_index;
+
+ rte_spinlock_lock(&rte_mempool_ops_table.sl);
+
+ if (rte_mempool_ops_table.num_ops >=
+ RTE_MEMPOOL_MAX_OPS_IDX) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Maximum number of mempool ops structs exceeded\n");
+ return -ENOSPC;
+ }
+
+ if (h->alloc == NULL || h->enqueue == NULL ||
+ h->dequeue == NULL || h->get_count == NULL) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Missing callback while registering mempool ops\n");
+ return -EINVAL;
+ }
+
+ if (strlen(h->name) >= sizeof(ops->name) - 1) {
+ RTE_LOG(DEBUG, EAL, "%s(): mempool_ops <%s>: name too long\n",
+ __func__, h->name);
+ rte_errno = EEXIST;
+ return -EEXIST;
+ }
+
+ ops_index = rte_mempool_ops_table.num_ops++;
+ ops = &rte_mempool_ops_table.ops[ops_index];
+ snprintf(ops->name, sizeof(ops->name), "%s", h->name);
+ ops->alloc = h->alloc;
+ ops->enqueue = h->enqueue;
+ ops->dequeue = h->dequeue;
+ ops->get_count = h->get_count;
+
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+
+ return ops_index;
+}
+
+/* wrapper to allocate an external mempool's private (pool) data. */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->alloc(mp);
+}
+
+/* wrapper to free an external pool ops. */
+void
+rte_mempool_ops_free(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ if (ops->free == NULL)
+ return;
+ return ops->free(mp);
+}
+
+/* wrapper to get available objects in an external mempool. */
+unsigned int
+rte_mempool_ops_get_count(const struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->get_count(mp);
+}
+
+/* sets mempool ops previously registered by rte_mempool_ops_register. */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name)
+{
+ struct rte_mempool_ops *ops = NULL;
+ unsigned i;
+
+ /* too late, the mempool is already populated. */
+ if (mp->flags & MEMPOOL_F_POOL_CREATED)
+ return -EEXIST;
+
+ for (i = 0; i < rte_mempool_ops_table.num_ops; i++) {
+ if (!strcmp(name,
+ rte_mempool_ops_table.ops[i].name)) {
+ ops = &rte_mempool_ops_table.ops[i];
+ break;
+ }
+ }
+
+ if (ops == NULL)
+ return -EINVAL;
+
+ mp->ops_index = i;
+ return 0;
+}
diff --git a/lib/librte_mempool/rte_mempool_ring.c b/lib/librte_mempool/rte_mempool_ring.c
new file mode 100644
index 0000000..626786e
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ring.c
@@ -0,0 +1,161 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_errno.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+
+static int
+common_ring_mp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_mp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_sp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_mc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_mc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_sc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static unsigned
+common_ring_get_count(const struct rte_mempool *mp)
+{
+ return rte_ring_count(mp->pool_data);
+}
+
+
+static int
+common_ring_alloc(struct rte_mempool *mp)
+{
+ int rg_flags = 0, ret;
+ char rg_name[RTE_RING_NAMESIZE];
+ struct rte_ring *r;
+
+ ret = snprintf(rg_name, sizeof(rg_name),
+ RTE_MEMPOOL_MZ_FORMAT, mp->name);
+ if (ret < 0 || ret >= (int)sizeof(rg_name)) {
+ rte_errno = ENAMETOOLONG;
+ return -rte_errno;
+ }
+
+ /* ring flags */
+ if (mp->flags & MEMPOOL_F_SP_PUT)
+ rg_flags |= RING_F_SP_ENQ;
+ if (mp->flags & MEMPOOL_F_SC_GET)
+ rg_flags |= RING_F_SC_DEQ;
+
+ /*
+ * Allocate the ring that will be used to store objects.
+ * Ring functions will return appropriate errors if we are
+ * running as a secondary process etc., so no checks made
+ * in this function for that condition.
+ */
+ r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
+ mp->socket_id, rg_flags);
+ if (r == NULL)
+ return -rte_errno;
+
+ mp->pool_data = r;
+
+ return 0;
+}
+
+static void
+common_ring_free(struct rte_mempool *mp)
+{
+ rte_ring_free(mp->pool_data);
+}
+
+/*
+ * The following 4 declarations of mempool ops structs address
+ * the need for the backward compatible mempool managers for
+ * single/multi producers and single/multi consumers as dictated by the
+ * flags provided to the rte_mempool_create function
+ */
+static const struct rte_mempool_ops ops_mp_mc = {
+ .name = "ring_mp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_sc = {
+ .name = "ring_sp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_mp_sc = {
+ .name = "ring_mp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_mc = {
+ .name = "ring_sp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+MEMPOOL_REGISTER_OPS(ops_mp_mc);
+MEMPOOL_REGISTER_OPS(ops_sp_sc);
+MEMPOOL_REGISTER_OPS(ops_mp_sc);
+MEMPOOL_REGISTER_OPS(ops_sp_mc);
diff --git a/lib/librte_mempool/rte_mempool_version.map b/lib/librte_mempool/rte_mempool_version.map
index f63461b..6209ec2 100644
--- a/lib/librte_mempool/rte_mempool_version.map
+++ b/lib/librte_mempool/rte_mempool_version.map
@@ -20,15 +20,18 @@ DPDK_16.7 {
global:
rte_mempool_check_cookies;
- rte_mempool_obj_iter;
- rte_mempool_mem_iter;
rte_mempool_create_empty;
+ rte_mempool_free;
+ rte_mempool_mem_iter;
+ rte_mempool_obj_iter;
+ rte_mempool_ops_register;
+ rte_mempool_ops_table;
+ rte_mempool_populate_anon;
+ rte_mempool_populate_default;
rte_mempool_populate_phys;
rte_mempool_populate_phys_tab;
rte_mempool_populate_virt;
- rte_mempool_populate_default;
- rte_mempool_populate_anon;
- rte_mempool_free;
+ rte_mempool_set_ops_byname;
local: *;
} DPDK_2.0;
--
2.5.5
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager
2016-06-14 15:48 3% ` [dpdk-dev] [PATCH v11 " David Hunt
2016-06-14 15:48 1% ` [dpdk-dev] [PATCH v11 1/3] mempool: support external mempool operations David Hunt
@ 2016-06-15 7:47 3% ` David Hunt
2016-06-15 7:47 1% ` [dpdk-dev] [PATCH v12 1/3] mempool: support external mempool operations David Hunt
` (2 more replies)
1 sibling, 3 replies; 200+ results
From: David Hunt @ 2016-06-15 7:47 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the External Mempool Manager patchset.
It's re-based on top of the latest head as of 14/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v12 changes:
* Fixed a comment (function pram h -> ops)
* Fixed a typo in mempool docs (callbacki)
v11 changes:
* Fixed comments (added '.' where needed for consistency)
* removed ABI breakage notice for mempool manager in deprecation.rst
* Added description of the external mempool manager functionality to
doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
* renamed rte_mempool_default.c to rte_mempool_ring.c
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The External Mempool Manager is an extension to the mempool API that allows
users to add and use an external mempool manager, which allows external memory
subsystems such as external hardware memory management systems and software
based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool manager will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing. These changes are all contained withing
RTE_NEXT_ABI defs, and the current or next code can be changed with
the CONFIG_RTE_NEXT_ABI config setting
There are two aspects to external mempool manager.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops_byname to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempool's ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several external mempool managers may be used in the same application. A new
mempool can then be created by using the new rte_mempool_create_empty function,
then calling rte_mempool_set_ops_byname to point the mempool to the relevant
mempool manager callback structure.
Legacy applications will continue to use the old rte_mempool_create API call,
which uses a ring based mempool manager by default. These applications
will need to be modified to use a new external mempool manager.
The external mempool manager needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool manager using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support external mempool operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test external mempool manager
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v9 8/8] doc: update doc for packet capture framework
2016-06-14 20:41 3% ` Thomas Monjalon
@ 2016-06-15 5:44 0% ` Pattan, Reshma
2016-06-15 8:24 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Pattan, Reshma @ 2016-06-15 5:44 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev, Mcnamara, John
> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Tuesday, June 14, 2016 9:41 PM
> To: Pattan, Reshma <reshma.pattan@intel.com>
> Cc: dev@dpdk.org; Mcnamara, John <john.mcnamara@intel.com>
> Subject: Re: [dpdk-dev] [PATCH v9 8/8] doc: update doc for packet capture
> framework
>
> When commenting previous patches, I missed these docs.
> Please move them in the appropriate patches.
>
> > +Use Case: Packet Capturing
> > +--------------------------
> > +
> > +The DPDK ``app/pdump`` tool is developed based on this library to capture
> packets in DPDK.
> > +Users can use this as an example to develop their own packet capturing
> application.
>
> Is it an example or a debugging tool?
It is a debugging tool.
> If it is an example, it should be in the examples/ directory.
>
> > ABI Changes
> > -----------
> > @@ -146,6 +156,9 @@ ABI Changes
> > * The ``rte_port_source_params`` structure has new fields to support PCAP
> file.
> > It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
> >
> > +* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and
> ``nb_tx_queues``
> > + to support number of queues configured by software.
>
> There was no deprecation notice in 16.04 for this ABI change.
Deprecation notice and relevant planed changes were sent as RFC during the start of 16.07 , please find the below is the link for the same.
http://dpdk.org/dev/patchwork/patch/12033/
http://dpdk.org/dev/patchwork/patch/12034/
Thanks,
Reshma
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v9 8/8] doc: update doc for packet capture framework
2016-06-14 9:38 5% ` [dpdk-dev] [PATCH v9 8/8] doc: update doc for " Reshma Pattan
@ 2016-06-14 20:41 3% ` Thomas Monjalon
2016-06-15 5:44 0% ` Pattan, Reshma
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-06-14 20:41 UTC (permalink / raw)
To: Reshma Pattan; +Cc: dev, john.mcnamara
When commenting previous patches, I missed these docs.
Please move them in the appropriate patches.
2016-06-14 10:38, Reshma Pattan:
> --- a/doc/guides/prog_guide/index.rst
> +++ b/doc/guides/prog_guide/index.rst
> @@ -71,6 +71,7 @@ Programmer's Guide
> writing_efficient_code
> profile_app
> glossary
> + pdump_library
There is probably a better place that after the glossary.
[...]
> +The librte_pdump Library
> +========================
> +
> +The ``librte_pdump`` library provides a framework for packet capturing in DPDK.
Here you need to explain what you mean by "packet capturing".
Doing a copy?
Slowing down the normal processing?
Which usage do you target? debugging? fast mirroring?
> +Use Case: Packet Capturing
> +--------------------------
> +
> +The DPDK ``app/pdump`` tool is developed based on this library to capture packets in DPDK.
> +Users can use this as an example to develop their own packet capturing application.
Is it an example or a debugging tool?
If it is an example, it should be in the examples/ directory.
> ABI Changes
> -----------
> @@ -146,6 +156,9 @@ ABI Changes
> * The ``rte_port_source_params`` structure has new fields to support PCAP file.
> It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
>
> +* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
> + to support number of queues configured by software.
There was no deprecation notice in 16.04 for this ABI change.
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v11 1/3] mempool: support external mempool operations
2016-06-14 15:48 3% ` [dpdk-dev] [PATCH v11 " David Hunt
@ 2016-06-14 15:48 1% ` David Hunt
2016-06-15 7:47 3% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager David Hunt
1 sibling, 0 replies; 200+ results
From: David Hunt @ 2016-06-14 15:48 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain, David Hunt
Until now, the objects stored in a mempool were internally stored in a
ring. This patch introduces the possibility to register external handlers
replacing the ring.
The default behavior remains unchanged, but calling the new function
rte_mempool_set_ops_byname() right after rte_mempool_create_empty() allows
the user to change the handler that will be used when populating
the mempool.
This patch also adds a set of default ops (function callbacks) based
on rte_ring.
Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
Signed-off-by: David Hunt <david.hunt@intel.com>
Acked-by: Shreyansh Jain <shreyansh.jain@nxp.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---
app/test/test_mempool_perf.c | 1 -
doc/guides/prog_guide/mempool_lib.rst | 31 +++-
doc/guides/rel_notes/deprecation.rst | 9 --
lib/librte_mempool/Makefile | 2 +
lib/librte_mempool/rte_mempool.c | 66 +++-----
lib/librte_mempool/rte_mempool.h | 251 ++++++++++++++++++++++++++---
lib/librte_mempool/rte_mempool_ops.c | 148 +++++++++++++++++
lib/librte_mempool/rte_mempool_ring.c | 161 ++++++++++++++++++
lib/librte_mempool/rte_mempool_version.map | 13 +-
9 files changed, 601 insertions(+), 81 deletions(-)
create mode 100644 lib/librte_mempool/rte_mempool_ops.c
create mode 100644 lib/librte_mempool/rte_mempool_ring.c
diff --git a/app/test/test_mempool_perf.c b/app/test/test_mempool_perf.c
index c5e3576..c5f8455 100644
--- a/app/test/test_mempool_perf.c
+++ b/app/test/test_mempool_perf.c
@@ -161,7 +161,6 @@ per_lcore_mempool_test(__attribute__((unused)) void *arg)
n_get_bulk);
if (unlikely(ret < 0)) {
rte_mempool_dump(stdout, mp);
- rte_ring_dump(stdout, mp->ring);
/* in this case, objects are lost... */
return -1;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index c3afc2e..6e358d5 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -34,7 +34,7 @@ Mempool Library
===============
A memory pool is an allocator of a fixed-sized object.
-In the DPDK, it is identified by name and uses a ring to store free objects.
+In the DPDK, it is identified by name and uses a ring or an external mempool manager to store free objects.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
@@ -127,6 +127,35 @@ The maximum size of the cache is static and is defined at compilation time (CONF
A mempool in Memory with its Associated Ring
+External Mempool Manager
+------------------------
+
+This allows external memory subsystems, such as external hardware memory
+management systems and software based memory allocators, to be used with DPDK.
+
+There are two aspects to external mempool manager.
+
+* Adding the code for your new mempool operations (ops). This is achieved by
+ adding a new mempool ops code, and using the ``REGISTER_MEMPOOL_OPS`` macro.
+
+* Using the new API to call ``rte_mempool_create_empty()`` and
+ ``rte_mempool_set_ops_byname()`` to create a new mempool and specifying which
+ ops to use.
+
+Several external mempool managers may be used in the same application. A new
+mempool can be created by using the ``rte_mempool_create_empty()`` function,
+then using ``rte_mempool_set_ops_byname()`` to point the mempool to the
+relevant mempool manager callbacki (ops) structure.
+
+Legacy applications may continue to use the old ``rte_mempool_create()`` API
+call, which uses a ring based mempool manager by default. These applications
+will need to be modified to use a new external mempool manager.
+
+For applications that use ``rte_pktmbuf_create()``, there is a config setting
+(``RTE_MBUF_DEFAULT_MEMPOOL_OPS``) that allows the application to make use of
+an external mempool manager.
+
+
Use Cases
---------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index bda40c1..5708eef 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -45,15 +45,6 @@ Deprecation Notices
compact API. The ones that remain are backwards compatible and use the
per-lcore default cache if available. This change targets release 16.07.
-* The rte_mempool struct will be changed in 16.07 to facilitate the new
- external mempool manager functionality.
- The ring element will be replaced with a more generic 'pool' opaque pointer
- to allow new mempool handlers to use their own user-defined mempool
- layout. Also newly added to rte_mempool is a handler index.
- The existing API will be backward compatible, but there will be new API
- functions added to facilitate the creation of mempools using an external
- handler. The 16.07 release will contain these changes.
-
* A librte_vhost public structures refactor is planned for DPDK 16.07
that requires both ABI and API change.
The proposed refactor would expose DPDK vhost dev to applications as
diff --git a/lib/librte_mempool/Makefile b/lib/librte_mempool/Makefile
index 43423e0..a4c089e 100644
--- a/lib/librte_mempool/Makefile
+++ b/lib/librte_mempool/Makefile
@@ -42,6 +42,8 @@ LIBABIVER := 2
# all source are stored in SRCS-y
SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_MEMPOOL) += rte_mempool_ring.c
# install includes
SYMLINK-$(CONFIG_RTE_LIBRTE_MEMPOOL)-include := rte_mempool.h
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index 22a5645..ac40cb3 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -148,7 +148,7 @@ mempool_add_elem(struct rte_mempool *mp, void *obj, phys_addr_t physaddr)
#endif
/* enqueue in ring */
- rte_ring_sp_enqueue(mp->ring, obj);
+ rte_mempool_ops_enqueue_bulk(mp, &obj, 1);
}
/* call obj_cb() for each mempool element */
@@ -303,40 +303,6 @@ rte_mempool_xmem_usage(__rte_unused void *vaddr, uint32_t elt_num,
return (size_t)paddr_idx << pg_shift;
}
-/* create the internal ring */
-static int
-rte_mempool_ring_create(struct rte_mempool *mp)
-{
- int rg_flags = 0, ret;
- char rg_name[RTE_RING_NAMESIZE];
- struct rte_ring *r;
-
- ret = snprintf(rg_name, sizeof(rg_name),
- RTE_MEMPOOL_MZ_FORMAT, mp->name);
- if (ret < 0 || ret >= (int)sizeof(rg_name))
- return -ENAMETOOLONG;
-
- /* ring flags */
- if (mp->flags & MEMPOOL_F_SP_PUT)
- rg_flags |= RING_F_SP_ENQ;
- if (mp->flags & MEMPOOL_F_SC_GET)
- rg_flags |= RING_F_SC_DEQ;
-
- /* Allocate the ring that will be used to store objects.
- * Ring functions will return appropriate errors if we are
- * running as a secondary process etc., so no checks made
- * in this function for that condition.
- */
- r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
- mp->socket_id, rg_flags);
- if (r == NULL)
- return -rte_errno;
-
- mp->ring = r;
- mp->flags |= MEMPOOL_F_RING_CREATED;
- return 0;
-}
-
/* free a memchunk allocated with rte_memzone_reserve() */
static void
rte_mempool_memchunk_mz_free(__rte_unused struct rte_mempool_memhdr *memhdr,
@@ -354,7 +320,7 @@ rte_mempool_free_memchunks(struct rte_mempool *mp)
void *elt;
while (!STAILQ_EMPTY(&mp->elt_list)) {
- rte_ring_sc_dequeue(mp->ring, &elt);
+ rte_mempool_ops_dequeue_bulk(mp, &elt, 1);
(void)elt;
STAILQ_REMOVE_HEAD(&mp->elt_list, next);
mp->populated_size--;
@@ -386,9 +352,9 @@ rte_mempool_populate_phys(struct rte_mempool *mp, char *vaddr,
int ret;
/* create the internal ring if not already done */
- if ((mp->flags & MEMPOOL_F_RING_CREATED) == 0) {
- ret = rte_mempool_ring_create(mp);
- if (ret < 0)
+ if ((mp->flags & MEMPOOL_F_POOL_CREATED) == 0) {
+ ret = rte_mempool_ops_alloc(mp);
+ if (ret != 0)
return ret;
}
@@ -703,7 +669,7 @@ rte_mempool_free(struct rte_mempool *mp)
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_mempool_free_memchunks(mp);
- rte_ring_free(mp->ring);
+ rte_mempool_ops_free(mp);
rte_memzone_free(mp->mz);
}
@@ -815,6 +781,7 @@ rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, 0));
te->data = mp;
+
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_INSERT_TAIL(mempool_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
@@ -844,6 +811,19 @@ rte_mempool_create(const char *name, unsigned n, unsigned elt_size,
if (mp == NULL)
return NULL;
+ /*
+ * Since we have 4 combinations of the SP/SC/MP/MC examine the flags to
+ * set the correct index into the table of ops structs.
+ */
+ if (flags & (MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET))
+ rte_mempool_set_ops_byname(mp, "ring_sp_sc");
+ else if (flags & MEMPOOL_F_SP_PUT)
+ rte_mempool_set_ops_byname(mp, "ring_sp_mc");
+ else if (flags & MEMPOOL_F_SC_GET)
+ rte_mempool_set_ops_byname(mp, "ring_mp_sc");
+ else
+ rte_mempool_set_ops_byname(mp, "ring_mp_mc");
+
/* call the mempool priv initializer */
if (mp_init)
mp_init(mp, mp_init_arg);
@@ -930,7 +910,7 @@ rte_mempool_count(const struct rte_mempool *mp)
unsigned count;
unsigned lcore_id;
- count = rte_ring_count(mp->ring);
+ count = rte_mempool_ops_get_count(mp);
if (mp->cache_size == 0)
return count;
@@ -1119,7 +1099,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
fprintf(f, "mempool <%s>@%p\n", mp->name, mp);
fprintf(f, " flags=%x\n", mp->flags);
- fprintf(f, " ring=<%s>@%p\n", mp->ring->name, mp->ring);
+ fprintf(f, " pool=%p\n", mp->pool_data);
fprintf(f, " phys_addr=0x%" PRIx64 "\n", mp->mz->phys_addr);
fprintf(f, " nb_mem_chunks=%u\n", mp->nb_mem_chunks);
fprintf(f, " size=%"PRIu32"\n", mp->size);
@@ -1140,7 +1120,7 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
}
cache_count = rte_mempool_dump_cache(f, mp);
- common_count = rte_ring_count(mp->ring);
+ common_count = rte_mempool_ops_get_count(mp);
if ((cache_count + common_count) > mp->size)
common_count = mp->size - cache_count;
fprintf(f, " common_pool_count=%u\n", common_count);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 60339bd..e429f3f 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -67,6 +67,7 @@
#include <inttypes.h>
#include <sys/queue.h>
+#include <rte_spinlock.h>
#include <rte_log.h>
#include <rte_debug.h>
#include <rte_lcore.h>
@@ -203,10 +204,13 @@ struct rte_mempool_memhdr {
*/
struct rte_mempool {
char name[RTE_MEMPOOL_NAMESIZE]; /**< Name of mempool. */
- struct rte_ring *ring; /**< Ring to store objects. */
- const struct rte_memzone *mz; /**< Memzone where pool is allocated */
+ union {
+ void *pool_data; /**< Ring or pool to store objects. */
+ uint64_t pool_id; /**< External mempool identifier. */
+ };
+ const struct rte_memzone *mz; /**< Memzone where pool is alloc'd. */
int flags; /**< Flags of the mempool. */
- int socket_id; /**< Socket id passed at mempool creation. */
+ int socket_id; /**< Socket id passed at create. */
uint32_t size; /**< Max size of the mempool. */
uint32_t cache_size; /**< Size of per-lcore local cache. */
uint32_t cache_flushthresh;
@@ -217,6 +221,14 @@ struct rte_mempool {
uint32_t trailer_size; /**< Size of trailer (after elt). */
unsigned private_data_size; /**< Size of private data. */
+ /**
+ * Index into rte_mempool_ops_table array of mempool ops
+ * structs, which contain callback function pointers.
+ * We're using an index here rather than pointers to the callbacks
+ * to facilitate any secondary processes that may want to use
+ * this mempool.
+ */
+ int32_t ops_index;
struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
@@ -235,7 +247,7 @@ struct rte_mempool {
#define MEMPOOL_F_NO_CACHE_ALIGN 0x0002 /**< Do not align objs on cache lines.*/
#define MEMPOOL_F_SP_PUT 0x0004 /**< Default put is "single-producer".*/
#define MEMPOOL_F_SC_GET 0x0008 /**< Default get is "single-consumer".*/
-#define MEMPOOL_F_RING_CREATED 0x0010 /**< Internal: ring is created */
+#define MEMPOOL_F_POOL_CREATED 0x0010 /**< Internal: pool is created. */
#define MEMPOOL_F_NO_PHYS_CONTIG 0x0020 /**< Don't need physically contiguous objs. */
/**
@@ -325,6 +337,212 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
#define __mempool_check_cookies(mp, obj_table_const, n, free) do {} while(0)
#endif /* RTE_LIBRTE_MEMPOOL_DEBUG */
+#define RTE_MEMPOOL_OPS_NAMESIZE 32 /**< Max length of ops struct name. */
+
+/**
+ * Prototype for implementation specific data provisioning function.
+ *
+ * The function should provide the implementation specific memory for
+ * for use by the other mempool ops functions in a given mempool ops struct.
+ * E.g. the default ops provides an instance of the rte_ring for this purpose.
+ * it will most likely point to a different type of data structure, and
+ * will be transparent to the application programmer.
+ * This function should set mp->pool_data.
+ */
+typedef int (*rte_mempool_alloc_t)(struct rte_mempool *mp);
+
+/**
+ * Free the opaque private data pointed to by mp->pool_data pointer.
+ */
+typedef void (*rte_mempool_free_t)(struct rte_mempool *mp);
+
+/**
+ * Enqueue an object into the external pool.
+ */
+typedef int (*rte_mempool_enqueue_t)(struct rte_mempool *mp,
+ void * const *obj_table, unsigned int n);
+
+/**
+ * Dequeue an object from the external pool.
+ */
+typedef int (*rte_mempool_dequeue_t)(struct rte_mempool *mp,
+ void **obj_table, unsigned int n);
+
+/**
+ * Return the number of available objects in the external pool.
+ */
+typedef unsigned (*rte_mempool_get_count)(const struct rte_mempool *mp);
+
+/** Structure defining mempool operations structure */
+struct rte_mempool_ops {
+ char name[RTE_MEMPOOL_OPS_NAMESIZE]; /**< Name of mempool ops struct. */
+ rte_mempool_alloc_t alloc; /**< Allocate private data. */
+ rte_mempool_free_t free; /**< Free the external pool. */
+ rte_mempool_enqueue_t enqueue; /**< Enqueue an object. */
+ rte_mempool_dequeue_t dequeue; /**< Dequeue an object. */
+ rte_mempool_get_count get_count; /**< Get qty of available objs. */
+} __rte_cache_aligned;
+
+#define RTE_MEMPOOL_MAX_OPS_IDX 16 /**< Max registered ops structs */
+
+/**
+ * Structure storing the table of registered ops structs, each of which contain
+ * the function pointers for the mempool ops functions.
+ * Each process has its own storage for this ops struct array so that
+ * the mempools can be shared across primary and secondary processes.
+ * The indices used to access the array are valid across processes, whereas
+ * any function pointers stored directly in the mempool struct would not be.
+ * This results in us simply having "ops_index" in the mempool struct.
+ */
+struct rte_mempool_ops_table {
+ rte_spinlock_t sl; /**< Spinlock for add/delete. */
+ uint32_t num_ops; /**< Number of used ops structs in the table. */
+ /**
+ * Storage for all possible ops structs.
+ */
+ struct rte_mempool_ops ops[RTE_MEMPOOL_MAX_OPS_IDX];
+} __rte_cache_aligned;
+
+/** Array of registered ops structs. */
+extern struct rte_mempool_ops_table rte_mempool_ops_table;
+
+/**
+ * @internal Get the mempool ops struct from its index.
+ *
+ * @param ops_index
+ * The index of the ops struct in the ops struct table. It must be a valid
+ * index: (0 <= idx < num_ops).
+ * @return
+ * The pointer to the ops struct in the table.
+ */
+static inline struct rte_mempool_ops *
+rte_mempool_ops_get(int ops_index)
+{
+ RTE_VERIFY(ops_index < RTE_MEMPOOL_MAX_OPS_IDX);
+
+ return &rte_mempool_ops_table.ops[ops_index];
+}
+
+/**
+ * @internal Wrapper for mempool_ops alloc callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * - 0: Success; successfully allocated mempool pool_data.
+ * - <0: Error; code of alloc function.
+ */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp);
+
+/**
+ * @internal Wrapper for mempool_ops get callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to get.
+ * @return
+ * - 0: Success; got n objects.
+ * - <0: Error; code of get function.
+ */
+static inline int
+rte_mempool_ops_dequeue_bulk(struct rte_mempool *mp,
+ void **obj_table, unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->dequeue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops put callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param obj_table
+ * Pointer to a table of void * pointers (objects).
+ * @param n
+ * Number of objects to put.
+ * @return
+ * - 0: Success; n objects supplied.
+ * - <0: Error; code of put function.
+ */
+static inline int
+rte_mempool_ops_enqueue_bulk(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->enqueue(mp, obj_table, n);
+}
+
+/**
+ * @internal wrapper for mempool_ops get_count callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @return
+ * The number of available objects in the external pool.
+ */
+unsigned
+rte_mempool_ops_get_count(const struct rte_mempool *mp);
+
+/**
+ * @internal wrapper for mempool_ops free callback.
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ */
+void
+rte_mempool_ops_free(struct rte_mempool *mp);
+
+/**
+ * Set the ops of a mempool.
+ *
+ * This can only be done on a mempool that is not populated, i.e. just after
+ * a call to rte_mempool_create_empty().
+ *
+ * @param mp
+ * Pointer to the memory pool.
+ * @param name
+ * Name of the ops structure to use for this mempool.
+ * @return
+ * - 0: Success; the mempool is now using the requested ops functions.
+ * - -EINVAL - Invalid ops struct name provided.
+ * - -EEXIST - mempool already has an ops struct assigned.
+ */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
+
+/**
+ * Register mempool operations.
+ *
+ * @param h
+ * Pointer to and ops structure to register.
+ * @return
+ * - >=0: Success; return the index of the ops struct in the table.
+ * - -EINVAL - some missing callbacks while registering ops struct.
+ * - -ENOSPC - the maximum number of ops structs has been reached.
+ */
+int rte_mempool_ops_register(const struct rte_mempool_ops *ops);
+
+/**
+ * Macro to statically register the ops of an external mempool manager.
+ * Note that the rte_mempool_ops_register fails silently here when
+ * more then RTE_MEMPOOL_MAX_OPS_IDX is registered.
+ */
+#define MEMPOOL_REGISTER_OPS(ops) \
+ void mp_hdlr_init_##ops(void); \
+ void __attribute__((constructor, used)) mp_hdlr_init_##ops(void)\
+ { \
+ rte_mempool_ops_register(&ops); \
+ }
+
/**
* An object callback function for mempool.
*
@@ -774,7 +992,7 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
cache->len += n;
if (cache->len >= flushthresh) {
- rte_ring_mp_enqueue_bulk(mp->ring, &cache->objs[cache_size],
+ rte_mempool_ops_enqueue_bulk(mp, &cache->objs[cache_size],
cache->len - cache_size);
cache->len = cache_size;
}
@@ -785,19 +1003,10 @@ ring_enqueue:
/* push remaining objects in ring */
#ifdef RTE_LIBRTE_MEMPOOL_DEBUG
- if (is_mp) {
- if (rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
- else {
- if (rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n) < 0)
- rte_panic("cannot put objects in mempool\n");
- }
+ if (rte_mempool_ops_enqueue_bulk(mp, obj_table, n) < 0)
+ rte_panic("cannot put objects in mempool\n");
#else
- if (is_mp)
- rte_ring_mp_enqueue_bulk(mp->ring, obj_table, n);
- else
- rte_ring_sp_enqueue_bulk(mp->ring, obj_table, n);
+ rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
#endif
}
@@ -945,7 +1154,8 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
uint32_t req = n + (cache_size - cache->len);
/* How many do we require i.e. number to fill the cache + the request */
- ret = rte_ring_mc_dequeue_bulk(mp->ring, &cache->objs[cache->len], req);
+ ret = rte_mempool_ops_dequeue_bulk(mp,
+ &cache->objs[cache->len], req);
if (unlikely(ret < 0)) {
/*
* In the offchance that we are buffer constrained,
@@ -972,10 +1182,7 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
ring_dequeue:
/* get remaining objects from ring */
- if (is_mc)
- ret = rte_ring_mc_dequeue_bulk(mp->ring, obj_table, n);
- else
- ret = rte_ring_sc_dequeue_bulk(mp->ring, obj_table, n);
+ ret = rte_mempool_ops_dequeue_bulk(mp, obj_table, n);
if (ret < 0)
__MEMPOOL_STAT_ADD(mp, get_fail, n);
diff --git a/lib/librte_mempool/rte_mempool_ops.c b/lib/librte_mempool/rte_mempool_ops.c
new file mode 100644
index 0000000..9328b77
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ops.c
@@ -0,0 +1,148 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2016 Intel Corporation. All rights reserved.
+ * Copyright(c) 2016 6WIND S.A.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_mempool.h>
+#include <rte_errno.h>
+
+/* indirect jump table to support external memory pools. */
+struct rte_mempool_ops_table rte_mempool_ops_table = {
+ .sl = RTE_SPINLOCK_INITIALIZER,
+ .num_ops = 0
+};
+
+/* add a new ops struct in rte_mempool_ops_table, return its index. */
+int
+rte_mempool_ops_register(const struct rte_mempool_ops *h)
+{
+ struct rte_mempool_ops *ops;
+ int16_t ops_index;
+
+ rte_spinlock_lock(&rte_mempool_ops_table.sl);
+
+ if (rte_mempool_ops_table.num_ops >=
+ RTE_MEMPOOL_MAX_OPS_IDX) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Maximum number of mempool ops structs exceeded\n");
+ return -ENOSPC;
+ }
+
+ if (h->alloc == NULL || h->enqueue == NULL ||
+ h->dequeue == NULL || h->get_count == NULL) {
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+ RTE_LOG(ERR, MEMPOOL,
+ "Missing callback while registering mempool ops\n");
+ return -EINVAL;
+ }
+
+ if (strlen(h->name) >= sizeof(ops->name) - 1) {
+ RTE_LOG(DEBUG, EAL, "%s(): mempool_ops <%s>: name too long\n",
+ __func__, h->name);
+ rte_errno = EEXIST;
+ return -EEXIST;
+ }
+
+ ops_index = rte_mempool_ops_table.num_ops++;
+ ops = &rte_mempool_ops_table.ops[ops_index];
+ snprintf(ops->name, sizeof(ops->name), "%s", h->name);
+ ops->alloc = h->alloc;
+ ops->enqueue = h->enqueue;
+ ops->dequeue = h->dequeue;
+ ops->get_count = h->get_count;
+
+ rte_spinlock_unlock(&rte_mempool_ops_table.sl);
+
+ return ops_index;
+}
+
+/* wrapper to allocate an external mempool's private (pool) data. */
+int
+rte_mempool_ops_alloc(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->alloc(mp);
+}
+
+/* wrapper to free an external pool ops. */
+void
+rte_mempool_ops_free(struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ if (ops->free == NULL)
+ return;
+ return ops->free(mp);
+}
+
+/* wrapper to get available objects in an external mempool. */
+unsigned int
+rte_mempool_ops_get_count(const struct rte_mempool *mp)
+{
+ struct rte_mempool_ops *ops;
+
+ ops = rte_mempool_ops_get(mp->ops_index);
+ return ops->get_count(mp);
+}
+
+/* sets mempool ops previously registered by rte_mempool_ops_register. */
+int
+rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name)
+{
+ struct rte_mempool_ops *ops = NULL;
+ unsigned i;
+
+ /* too late, the mempool is already populated. */
+ if (mp->flags & MEMPOOL_F_POOL_CREATED)
+ return -EEXIST;
+
+ for (i = 0; i < rte_mempool_ops_table.num_ops; i++) {
+ if (!strcmp(name,
+ rte_mempool_ops_table.ops[i].name)) {
+ ops = &rte_mempool_ops_table.ops[i];
+ break;
+ }
+ }
+
+ if (ops == NULL)
+ return -EINVAL;
+
+ mp->ops_index = i;
+ return 0;
+}
diff --git a/lib/librte_mempool/rte_mempool_ring.c b/lib/librte_mempool/rte_mempool_ring.c
new file mode 100644
index 0000000..626786e
--- /dev/null
+++ b/lib/librte_mempool/rte_mempool_ring.c
@@ -0,0 +1,161 @@
+/*-
+ * BSD LICENSE
+ *
+ * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * * Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in
+ * the documentation and/or other materials provided with the
+ * distribution.
+ * * Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_errno.h>
+#include <rte_ring.h>
+#include <rte_mempool.h>
+
+static int
+common_ring_mp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_mp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sp_enqueue(struct rte_mempool *mp, void * const *obj_table,
+ unsigned n)
+{
+ return rte_ring_sp_enqueue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_mc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_mc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static int
+common_ring_sc_dequeue(struct rte_mempool *mp, void **obj_table, unsigned n)
+{
+ return rte_ring_sc_dequeue_bulk(mp->pool_data, obj_table, n);
+}
+
+static unsigned
+common_ring_get_count(const struct rte_mempool *mp)
+{
+ return rte_ring_count(mp->pool_data);
+}
+
+
+static int
+common_ring_alloc(struct rte_mempool *mp)
+{
+ int rg_flags = 0, ret;
+ char rg_name[RTE_RING_NAMESIZE];
+ struct rte_ring *r;
+
+ ret = snprintf(rg_name, sizeof(rg_name),
+ RTE_MEMPOOL_MZ_FORMAT, mp->name);
+ if (ret < 0 || ret >= (int)sizeof(rg_name)) {
+ rte_errno = ENAMETOOLONG;
+ return -rte_errno;
+ }
+
+ /* ring flags */
+ if (mp->flags & MEMPOOL_F_SP_PUT)
+ rg_flags |= RING_F_SP_ENQ;
+ if (mp->flags & MEMPOOL_F_SC_GET)
+ rg_flags |= RING_F_SC_DEQ;
+
+ /*
+ * Allocate the ring that will be used to store objects.
+ * Ring functions will return appropriate errors if we are
+ * running as a secondary process etc., so no checks made
+ * in this function for that condition.
+ */
+ r = rte_ring_create(rg_name, rte_align32pow2(mp->size + 1),
+ mp->socket_id, rg_flags);
+ if (r == NULL)
+ return -rte_errno;
+
+ mp->pool_data = r;
+
+ return 0;
+}
+
+static void
+common_ring_free(struct rte_mempool *mp)
+{
+ rte_ring_free(mp->pool_data);
+}
+
+/*
+ * The following 4 declarations of mempool ops structs address
+ * the need for the backward compatible mempool managers for
+ * single/multi producers and single/multi consumers as dictated by the
+ * flags provided to the rte_mempool_create function
+ */
+static const struct rte_mempool_ops ops_mp_mc = {
+ .name = "ring_mp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_sc = {
+ .name = "ring_sp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_mp_sc = {
+ .name = "ring_mp_sc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_mp_enqueue,
+ .dequeue = common_ring_sc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+static const struct rte_mempool_ops ops_sp_mc = {
+ .name = "ring_sp_mc",
+ .alloc = common_ring_alloc,
+ .free = common_ring_free,
+ .enqueue = common_ring_sp_enqueue,
+ .dequeue = common_ring_mc_dequeue,
+ .get_count = common_ring_get_count,
+};
+
+MEMPOOL_REGISTER_OPS(ops_mp_mc);
+MEMPOOL_REGISTER_OPS(ops_sp_sc);
+MEMPOOL_REGISTER_OPS(ops_mp_sc);
+MEMPOOL_REGISTER_OPS(ops_sp_mc);
diff --git a/lib/librte_mempool/rte_mempool_version.map b/lib/librte_mempool/rte_mempool_version.map
index f63461b..6209ec2 100644
--- a/lib/librte_mempool/rte_mempool_version.map
+++ b/lib/librte_mempool/rte_mempool_version.map
@@ -20,15 +20,18 @@ DPDK_16.7 {
global:
rte_mempool_check_cookies;
- rte_mempool_obj_iter;
- rte_mempool_mem_iter;
rte_mempool_create_empty;
+ rte_mempool_free;
+ rte_mempool_mem_iter;
+ rte_mempool_obj_iter;
+ rte_mempool_ops_register;
+ rte_mempool_ops_table;
+ rte_mempool_populate_anon;
+ rte_mempool_populate_default;
rte_mempool_populate_phys;
rte_mempool_populate_phys_tab;
rte_mempool_populate_virt;
- rte_mempool_populate_default;
- rte_mempool_populate_anon;
- rte_mempool_free;
+ rte_mempool_set_ops_byname;
local: *;
} DPDK_2.0;
--
2.5.5
^ permalink raw reply [relevance 1%]
* [dpdk-dev] [PATCH v11 0/3] mempool: add external mempool manager
2016-06-14 9:46 2% ` [dpdk-dev] [PATCH v10 " David Hunt
@ 2016-06-14 15:48 3% ` David Hunt
2016-06-14 15:48 1% ` [dpdk-dev] [PATCH v11 1/3] mempool: support external mempool operations David Hunt
2016-06-15 7:47 3% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager David Hunt
0 siblings, 2 replies; 200+ results
From: David Hunt @ 2016-06-14 15:48 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the External Mempool Manager patchset.
It's re-based on top of the latest head as of 14/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v11 changes:
* Fixed comments (added '.' where needed for consistency)
* removed ABI breakage notice for mempool manager in deprecation.rst
* Added description of the external mempool manager functionality to
doc/guides/prog_guide/mempool_lib.rst (John Mc reviewed)
* renamed rte_mempool_default.c to rte_mempool_ring.c
* Kept the v10 ACK from Shreyansh and Olivier for v11
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The External Mempool Manager is an extension to the mempool API that allows
users to add and use an external mempool manager, which allows external memory
subsystems such as external hardware memory management systems and software
based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool manager will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing. These changes are all contained withing
RTE_NEXT_ABI defs, and the current or next code can be changed with
the CONFIG_RTE_NEXT_ABI config setting
There are two aspects to external mempool manager.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_OPS macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops_byname to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempool's ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several external mempool managers may be used in the same application. A new
mempool can then be created by using the new rte_mempool_create_empty function,
then calling rte_mempool_set_ops_byname to point the mempool to the relevant
mempool manager callback structure.
Legacy applications will continue to use the old rte_mempool_create API call,
which uses a ring based mempool manager by default. These applications
will need to be modified to use a new external mempool manager.
The external mempool manager needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool manager using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support external mempool operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test external mempool handler
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH] vhost: remove internal lockless enqueue
@ 2016-06-14 14:07 3% ` Yuanhan Liu
0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-06-14 14:07 UTC (permalink / raw)
To: Huawei Xie
Cc: dev, yuanhan.liu, jianfeng.tan, mukawa, kevin.traynor, haifeng.lin
Firstly, it's V2. So, don't forget to add version number and link it to
the previous version next time.
On Mon, Jun 13, 2016 at 07:52:12PM +0800, Huawei Xie wrote:
> All other DPDK PMDs doesn't support concurrent receiving or sending
> packets to the same queue. The upper application should deal with
> this, normally through queue and core bindings.
>
> Due to historical reason, vhost internally supports concurrent lockless
> enqueuing packets to the same virtio queue through costly cmpset operation.
> This patch removes this internal lockless implementation and should improve
> performance a bit.
>
> Luckily DPDK OVS doesn't rely on this behavior.
>
> Signed-off-by: Huawei Xie <huawei.xie@intel.com>
Applied to dpdk-next-virtio, with the rebase on top of my vhost ABI/API
changes.
FYI, I also renamed the title a bit to "remove concurrent enqueue" as
Thomas mentioned in the last version, that it's confusing to say "remove
lockless" here, since we are actually removing the lock.
Thanks.
--yliu
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH 0/3] [RFC] vhost: micro vhost optimization
@ 2016-06-14 12:42 0% ` Yuanhan Liu
0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-06-14 12:42 UTC (permalink / raw)
To: dev; +Cc: huawei.xie
Applied to dpdk-next-virtio.
--yliu
On Mon, May 02, 2016 at 05:46:15PM -0700, Yuanhan Liu wrote:
> Here is a small patch set does the micro optimization, which brings about
> 10% performance boost in my 64B packet testing, with the following topo:
>
> pkt generator <----> NIC <-----> Virtio NIC
>
> Patch 1 pre updates the used ring and update them in batch. It should be
> feasible from my understanding: there will be no issue, guest driver will
> not start processing them as far as we haven't updated the "used->idx"
> yet. I could miss something though.
>
> Patch 2 saves one check for small packets (that can be hold in one desc
> buf and mbuf).
>
> Patch 3 moves several frequently used fields into one cache line, for
> better cache sharing.
>
> Note that this patch set is based on my latest vhost ABI refactoring patchset.
>
>
> ---
> Yuanhan Liu (3):
> vhost: pre update used ring for Tx and Rx
> vhost: optimize dequeue for small packets
> vhost: arrange virtio_net fields for better cache sharing
>
> lib/librte_vhost/vhost-net.h | 8 +--
> lib/librte_vhost/vhost_rxtx.c | 110 ++++++++++++++++++++++++------------------
> 2 files changed, 68 insertions(+), 50 deletions(-)
>
> --
> 1.9.0
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v3 0/6] vhost: add vhost-user client mode and reconnect ability
@ 2016-06-14 12:00 0% ` Yuanhan Liu
0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-06-14 12:00 UTC (permalink / raw)
To: dev; +Cc: huawei.xie, Traynor Kevin, marcandre.lureau
Applied to dpdk-next-virtio.
--yliu
On Tue, Jun 07, 2016 at 12:05:02PM +0800, Yuanhan Liu wrote:
> v3: - make the "reconnect" feature be default for client mode, as it's
> good to handle guest OS restart with less effort.
> - fix var not-initilized error pointed out by Rich
>
>
> NOTE: I created a branch at dpdk.org [0] for more convenient testing:
>
> [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
>
>
> When the DPDK vhost-user application (such as OVS) restarts (due to
> crash, or update), the vhost-user connection between DPDK and QEMU
> won't be established automatically again. In another word, the virtio
> net is broken.
>
> The reason it doesn't work is that DPDK just acts as server only.
> A restart of the server needs a reconnection from the client (QEMU).
> However, reconnect from QEMU is not supported from QEMU.
>
> Adding the support of client mode and let DPDK be the client somehow
> would resolve above issue a bit easier: a restart of DPDK would naturally
> try to connect to the server (QEMU) automatically.
>
> Therefore, this patchset implements the DPDK vhost-user client mode, by
> introducing a new arg (flags) for API rte_vhost_driver_register(). And the
> client mode is enabled when RTE_VHOST_USER_CLIENT is given. Note that this
> implies an API breakage. However, since this release deals with ABI/API
> refactoring, it should not be an issue.
>
> Another interesting thing to make it work is that you not only have
> to consider that case the DPDK vhost-user app might restart, but also
> have to think that QEMU might restart as well: guest OS sometimes
> just reboots. In such case, when the server is down, the client has
> to keep reconnecting with the server until the server is back and the
> connection is established again. And that's what "reconnect" patch for.
>
> Note that current QEMU doesn't not support a second time connection
> from client, thus a restart of DPDK vhost-user will not work. This is
> because current QEMU won't be able to detect the disconnect from
> restart, thus it will not listen for later connections. Patches [1] have
> been sent, it's just not merged yet. But unlike the vhost-user mulitple
> queue case, that we have critical depends on QEMU implementation, here
> we have no such dependency, therefore, I think it's okay to make DPDK
> be ready for the "reconnect" stuff first. (note that I also mentioned
> this fact in the release doc).
>
> [1]: http://lists.nongnu.org/archive/html/qemu-devel/2016-05/msg01507.html
>
> v2: - added release doc
> - do not remove socket file for the client mode
> - create one thread ony to handle all reconnects
>
>
> Thanks.
> --yliu
>
> ---
> Yuanhan Liu (6):
> vhost: rename structs for enabling client mode
> vhost: add vhost-user client mode
> vhost: add reconnect ability
> vhost: workaround stale vring base
> examples/vhost: add client option
> vhost: add pmd client option
>
> doc/guides/rel_notes/release_16_07.rst | 21 ++
> drivers/net/vhost/rte_eth_vhost.c | 38 ++-
> examples/vhost/main.c | 12 +-
> lib/librte_vhost/rte_virtio_net.h | 12 +-
> lib/librte_vhost/vhost_cuse/vhost-net-cdev.c | 8 +-
> lib/librte_vhost/vhost_user/vhost-net-user.c | 403 ++++++++++++++++++---------
> lib/librte_vhost/vhost_user/vhost-net-user.h | 6 -
> lib/librte_vhost/virtio-net.c | 9 +
> 8 files changed, 361 insertions(+), 148 deletions(-)
>
> --
> 1.9.0
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v3 00/20] vhost ABI/API refactoring
@ 2016-06-14 12:00 4% ` Yuanhan Liu
2016-06-30 7:39 9% ` Panu Matilainen
1 sibling, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-06-14 12:00 UTC (permalink / raw)
To: dev
Cc: huawei.xie, Thomas Monjalon, Panu Matilainen, Traynor Kevin,
Rich Lane, Tetsuya Mukawa
Applied to dpdk-next-virtio.
--yliu
On Tue, Jun 07, 2016 at 11:51:50AM +0800, Yuanhan Liu wrote:
> v3: - adapted the new vhost ABI/API changes to tep_term example, to make
> sure not break build at least.
> - bumped the ABI version to 3
>
> NOTE: I created a branch at dpdk.org [0] for more conveinient testing:
>
> [0]: git://dpdk.org/next/dpdk-next-virtio for-testing
>
>
> Every time we introduce a new feature to vhost, we are likely to break
> ABI. Moreover, some cleanups (such as the one from Ilya to remove vec_buf
> from vhost_virtqueue struct) also break ABI.
>
> This patch set is meant to resolve above issue ultimately, by hiding
> virtio_net structure (as well as few others) internaly, and export the
> virtio_net dev strut to applications by a number, vid, like the way
> kernel exposes an fd to user space.
>
> Back to the patch set, the first part of this set makes some changes to
> vhost example, vhost-pmd and vhost, bit by bit, to remove the dependence
> to "virtio_net" struct. And then do the final change to make the current
> APIs to adapt to using "vid".
>
> After that, "vrtio_net_device_ops" is the only left open struct that an
> application can acces, therefore, it's the only place that might introduce
> potential ABI breakage in future for extension. Hence, I made few more
> (5) space reservation, to make sure we will not break ABI for a long time,
> and hopefuly, forever.
>
> The last bit of this patch set is some cleanups, including the one from
> Ilya.
>
> v2: - exported ifname as well to fix a vhost-pmd issue reported by Rich
> - separated the big patch that introduces several new APIs into some
> small patches.
> - updated release note
> - updated version.map
>
> Thanks.
>
> --yliu
>
> ---
> Ilya Maximets (1):
> vhost: make buf vector for scatter Rx local
>
> Yuanhan Liu (19):
> vhost: declare backend with int type
> vhost: set/reset dev flags internally
> vhost: declare device fh as int
> examples/vhost: make a copy of virtio device id
> vhost: rename device fh to vid
> vhost: get device by vid only
> vhost: move vhost device ctx to cuse
> vhost: introduce new API to export numa node
> vhost: introduce new API to export number of queues
> vhost: introduce new API to export ifname
> vhost: introduce new API to export queue free entries
> vhost: remove dependency on priv field
> vhost: export vid as the only interface to applications
> vhost: hide internal structs/macros/functions
> vhost: remove unnecessary fields
> vhost: remove virtio-net.h
> vhost: reserve few more space for future extension
> examples/tep_term: adapt to new vhost ABI/API changes
> vhost: per device virtio net header len
>
> doc/guides/rel_notes/release_16_07.rst | 11 +-
> drivers/net/vhost/rte_eth_vhost.c | 79 ++++-----
> examples/tep_termination/main.c | 83 +++++-----
> examples/tep_termination/main.h | 5 +-
> examples/tep_termination/vxlan_setup.c | 20 +--
> examples/tep_termination/vxlan_setup.h | 6 +-
> examples/vhost/main.c | 116 +++++++------
> examples/vhost/main.h | 3 +-
> lib/librte_vhost/Makefile | 2 +-
> lib/librte_vhost/rte_vhost_version.map | 10 ++
> lib/librte_vhost/rte_virtio_net.h | 223 +++++++------------------
> lib/librte_vhost/vhost-net.h | 201 ++++++++++++++++++----
> lib/librte_vhost/vhost_cuse/vhost-net-cdev.c | 83 +++++-----
> lib/librte_vhost/vhost_cuse/virtio-net-cdev.c | 30 ++--
> lib/librte_vhost/vhost_cuse/virtio-net-cdev.h | 12 +-
> lib/librte_vhost/vhost_rxtx.c | 133 ++++++++-------
> lib/librte_vhost/vhost_user/vhost-net-user.c | 53 +++---
> lib/librte_vhost/vhost_user/vhost-net-user.h | 2 +
> lib/librte_vhost/vhost_user/virtio-net-user.c | 64 +++----
> lib/librte_vhost/vhost_user/virtio-net-user.h | 18 +-
> lib/librte_vhost/virtio-net.c | 229 +++++++++++++++++---------
> lib/librte_vhost/virtio-net.h | 43 -----
> 22 files changed, 752 insertions(+), 674 deletions(-)
> delete mode 100644 lib/librte_vhost/virtio-net.h
>
> --
> 1.9.0
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH v10 0/3] mempool: add external mempool manager
2016-06-10 15:16 2% ` [dpdk-dev] [PATCH v9 0/3] " David Hunt
@ 2016-06-14 9:46 2% ` David Hunt
2016-06-14 15:48 3% ` [dpdk-dev] [PATCH v11 " David Hunt
0 siblings, 1 reply; 200+ results
From: David Hunt @ 2016-06-14 9:46 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the External Mempool Manager patchset.
It's re-based on top of the latest head as of 14/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v10 changes:
* changed the _put/_get op names to _enqueue/_dequeue to be consistent
with the function names
* some rte_errno cleanup
* comment tweaks about when to set pool_data
* removed an un-needed check for ops->alloc == NULL
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The External Mempool Manager is an extension to the mempool API that allows
users to add and use an external mempool manager, which allows external memory
subsystems such as external hardware memory management systems and software
based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool manager will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing. These changes are all contained withing
RTE_NEXT_ABI defs, and the current or next code can be changed with
the CONFIG_RTE_NEXT_ABI config setting
There are two aspects to external mempool manager.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_HANDLER macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempool's ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several external mempool managers may be used in the same application. A new
mempool can then be created by using the new 'create' function, providing the
mempool ops struct name to point the mempool to the relevant mempool manager
callback structure.
The old 'create' function can still be called by legacy programs, and will
internally work out the mempool handle based on the flags provided (single
producer, single consumer, etc). By default handles are created internally to
implement the built-in DPDK mempool manager and mempool types.
The external mempool manager needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. enqueue - puts an object back into the mempool once an application has
finished with it
3. dequeue - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time an enqueue/dequeue/get_count is called from the application/PMD,
the callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.enqueue = common_ring_sp_enqueue,
.dequeue = common_ring_mc_dequeue,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool manager using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support external mempool operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test external mempool manager
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH v9 8/8] doc: update doc for packet capture framework
2016-06-14 9:38 2% ` [dpdk-dev] [PATCH v9 " Reshma Pattan
@ 2016-06-14 9:38 5% ` Reshma Pattan
2016-06-14 20:41 3% ` Thomas Monjalon
2016-06-15 14:06 2% ` [dpdk-dev] [PATCH v10 0/7] add " Reshma Pattan
1 sibling, 1 reply; 200+ results
From: Reshma Pattan @ 2016-06-14 9:38 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
Added programmers guide for librte_pdump.
Added sample application guide for app/pdump application.
Updated release note for packet capture framework changes.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
---
MAINTAINERS | 3 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 119 +++++++++++++++++++++++++++++++
doc/guides/rel_notes/release_16_07.rst | 13 ++++
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 ++++++++++++++++++++++++++++++++
6 files changed, 259 insertions(+)
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
diff --git a/MAINTAINERS b/MAINTAINERS
index c46cf86..9a84f59 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -437,6 +437,9 @@ Pdump
M: Reshma Pattan <reshma.pattan@intel.com>
F: lib/librte_pdump/
F: app/pdump/
+F: doc/guides/prog_guide/pdump_library.rst
+F: doc/guides/sample_app_ug/pdump.rst
+
Hierarchical scheduler
M: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
diff --git a/doc/guides/prog_guide/index.rst b/doc/guides/prog_guide/index.rst
index b862d0c..4caf969 100644
--- a/doc/guides/prog_guide/index.rst
+++ b/doc/guides/prog_guide/index.rst
@@ -71,6 +71,7 @@ Programmer's Guide
writing_efficient_code
profile_app
glossary
+ pdump_library
**Figures**
diff --git a/doc/guides/prog_guide/pdump_library.rst b/doc/guides/prog_guide/pdump_library.rst
new file mode 100644
index 0000000..8781ffb
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,119 @@
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+.. _pdump_library:
+
+The librte_pdump Library
+========================
+
+The ``librte_pdump`` library provides a framework for packet capturing in DPDK.
+The library provides the following APIs to initialize the packet capture framework, to enable
+or disable the packet capture, and to uninitialize it:
+
+* ``rte_pdump_init()``:
+ This API initializes the packet capture framework.
+
+* ``rte_pdump_enable()``:
+ This API enables the packet capture on a given port and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_enable_by_deviceid()``:
+ This API enables the packet capture on a given device id (``vdev name or pci address``) and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_disable()``:
+ This API disables the packet capture on a given port and queue.
+
+* ``rte_pdump_disable_by_deviceid()``:
+ This API disables the packet capture on a given device id (``vdev name or pci address``) and queue.
+
+* ``rte_pdump_uninit()``:
+ This API uninitializes the packet capture framework.
+
+* ``rte_pdump_set_socket_dir()``:
+ This API sets the server and client socket paths.
+ Note: This API is not thread-safe.
+
+
+Operation
+---------
+
+The ``librte_pdump`` library works on a client/server model. The server is responsible for enabling or
+disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
+the packet capture.
+
+The packet capture framework, as part of its initialization, creates the pthread and the server socket in
+the pthread. The application that calls the framework initialization will have the server socket created,
+either under the path that the application has passed or under the default path i.e. either ``/var/run`` for
+root user or ``$HOME`` for non root user.
+
+Applications that request enabling or disabling of the packet capture will have the client socket created either under
+the path that the application has passed or under the default path i.e. either ``/var/run/`` for root user or ``$HOME``
+for not root user to send the requests to the server.
+The server socket will listen for client requests for enabling or disabling the packet capture.
+
+
+Implementation Details
+----------------------
+
+The library API ``rte_pdump_init()``, initializes the packet capture framework by creating the pthread and the server
+socket. The server socket in the pthread context will be listening to the client requests to enable or disable the
+packet capture.
+
+The library APIs ``rte_pdump_enable()`` and ``rte_pdump_enable_by_deviceid()`` enables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump enable" request and sends
+the request to the server. The server that is listening on the socket will take the request and enable the packet capture
+by registering the Ethernet RX and TX callbacks for the given port or device_id and queue combinations.
+Then the server will mirror the packets to the new mempool and enqueue them to the rte_ring that clients have passed
+to these APIs. The server also sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, the client socket is closed.
+
+The library APIs ``rte_pdump_disable()`` and ``rte_pdump_disable_by_deviceid()`` disables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump disable" request and sends
+the request to the server. The server that is listening on the socket will take the request and disable the packet
+capture by removing the Ethernet RX and TX callbacks for the given port or device_id and queue combinations. The server
+also sends the response back to the client about the status of the request that was processed. After the response is
+received from the server, the client socket is closed.
+
+The library API ``rte_pdump_uninit()``, uninitializes the packet capture framework by closing the pthread and the
+server socket.
+
+The library API ``rte_pdump_set_socket_dir()``, sets the given path as either server socket path
+or client socket path based on the ``type`` argument of the API.
+If the given path is ``NULL``, default path will be selected, i.e. either ``/var/run/`` for root user or ``$HOME``
+for non root user. Clients also need to call this API to set their server socket path if the server socket
+path is different from default path.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+The DPDK ``app/pdump`` tool is developed based on this library to capture packets in DPDK.
+Users can use this as an example to develop their own packet capturing application.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index c0f6b02..a4de2a2 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -66,6 +66,11 @@ New Features
* Enable RSS per network interface through the configuration file.
* Streamline the CLI code.
+* **Added packet capture framework.**
+
+ * A new library ``librte_pdump`` is added to provide packet capture APIs.
+ * A new ``app/pdump`` tool is added to capture packets in DPDK.
+
Resolved Issues
---------------
@@ -135,6 +140,11 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* Function ``rte_eth_dev_get_port_by_name`` changed to a public API.
+
+* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ in the ``rte_eth_dev_info`` object.
+
ABI Changes
-----------
@@ -146,6 +156,9 @@ ABI Changes
* The ``rte_port_source_params`` structure has new fields to support PCAP file.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ to support number of queues configured by software.
+
Shared Library Versions
-----------------------
diff --git a/doc/guides/sample_app_ug/index.rst b/doc/guides/sample_app_ug/index.rst
index 930f68c..96bb317 100644
--- a/doc/guides/sample_app_ug/index.rst
+++ b/doc/guides/sample_app_ug/index.rst
@@ -76,6 +76,7 @@ Sample Applications User Guide
ptpclient
performance_thread
ipsec_secgw
+ pdump
**Figures**
diff --git a/doc/guides/sample_app_ug/pdump.rst b/doc/guides/sample_app_ug/pdump.rst
new file mode 100644
index 0000000..96c8709
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,122 @@
+
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+dpdk_pdump Application
+======================
+
+The ``dpdk_pdump`` application is a Data Plane Development Kit (DPDK) application that runs as a DPDK secondary process and
+is capable of enabling packet capture on dpdk ports.
+
+
+Running the Application
+-----------------------
+
+The application has a ``--pdump`` command line option with various sub arguments:
+
+.. code-block:: console
+
+ ./build/app/dpdk_pdump --
+ --pdump '(port=<port id> | device_id=<pci id or vdev name>),
+ (queue=<queue_id>),
+ (rx-dev=<iface or pcap file> |
+ tx-dev=<iface or pcap file>),
+ [ring-size=<ring size>],
+ [mbuf-size=<mbuf data size>],
+ [total-num-mbufs=<number of mbufs>]'
+
+Note:
+
+* Parameters inside the parentheses represents mandatory parameters.
+
+* Parameters inside the square brackets represents optional parameters.
+
+Multiple instances of ``--pdump`` can be passed to capture packets on different port and queue combinations.
+
+
+Parameters
+~~~~~~~~~~
+
+``port``:
+Port id of the eth device on which packets should be captured.
+
+``device_id``:
+PCI address (or) name of the eth device on which packets should be captured.
+
+ .. Note::
+
+ * As of now the ``dpdk_pdump`` tool cannot capture the packets of virtual devices
+ in the primary process due to a bug in the ethdev library. Due to this bug, in a multi process context,
+ when the primary and secondary have different ports set, then the secondary process
+ (here the ``dpdk_pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
+
+``queue``:
+Queue id of the eth device on which packets should be captured. The user can pass a queue value of ``*`` to enable
+packet capture on all queues of the eth device.
+
+``rx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+``tx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+ .. Note::
+
+ * To receive ingress packets only, ``rx-dev`` should be passed.
+
+ * To receive egress packets only, ``tx-dev`` should be passed.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the different file names or the Linux iface names.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the same file names or the the Linux iface names.
+
+``ring-size``:
+Size of the ring. This value is used internally for ring creation. The ring will be used to enqueue the packets from
+the primary application to the secondary. This is an optional parameter with default size 16384.
+
+``mbuf-size``:
+Size of the mbuf data. This is used internally for mempool creation. Ideally this value must be same as
+the primary application's mempool's mbuf data size which is used for packet RX. This is an optional parameter with
+default size 2176.
+
+``total-num-mbufs``:
+Total number mbufs in mempool. This is used internally for mempool creation. This is an optional parameter with default
+value 65535.
+
+
+Example
+-------
+
+.. code-block:: console
+
+ $ sudo ./build/app/dpdk_pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
--
2.5.0
^ permalink raw reply [relevance 5%]
* [dpdk-dev] [PATCH v9 0/8] add packet capture framework
2016-06-10 16:18 2% ` [dpdk-dev] [PATCH v8 0/8] add packet capture framework Reshma Pattan
2016-06-10 16:18 5% ` [dpdk-dev] [PATCH v8 8/8] doc: update doc for " Reshma Pattan
2016-06-10 23:23 0% ` [dpdk-dev] [PATCH v8 0/8] add " Neil Horman
@ 2016-06-14 9:38 2% ` Reshma Pattan
2016-06-14 9:38 5% ` [dpdk-dev] [PATCH v9 8/8] doc: update doc for " Reshma Pattan
2016-06-15 14:06 2% ` [dpdk-dev] [PATCH v10 0/7] add " Reshma Pattan
2 siblings, 2 replies; 200+ results
From: Reshma Pattan @ 2016-06-14 9:38 UTC (permalink / raw)
To: dev
This patch set include below changes
1)Changes to librte_ether.
2)A new library librte_pdump added for packet capture framework.
3)A new app/pdump tool added for packet capturing.
4)Test pmd changes done to initialize packet capture framework.
5)Documentation update.
1)librte_pdump
==============
To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
is added.Users can develop their own packet capturing application using new library APIs.
Operation:
----------
The librte_pdump provides APIs to support packet capturing on dpdk Ethernet devices.
Library provides APIs to initialize the packet capture framework, enable/disable
the packet capture and uninitialize the packet capture framework.
The librte_pdump library works on a client/server model. The server is responsible for enabling or
disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
the packet capture.
The packet capture framework, as part of its initialization, creates the pthread and the server socket in
the pthread. The application that calls the framework initialization will have the server socket created,
either under the path that the application has passed or under the default path i.e. either ''/var/run'' for
root user or ''$HOME'' for non root user.
Applications that request enabling or disabling of the packet capture will have the client socket created
either under the path that the application has passed or under the default path i.e. either ''/var/run/''
for root users or ''$HOME'' for not root users to send the requests to the server.
The server socket will listen for client requests for enabling or disabling the packet capture.
Applications using below APIs need to pass port/device_id, queue, mempool and
ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
packets of the port for users. Users need to dequeue the rings and write the packets
to vdev(pcap/tuntap) to view the packets using any standard tools.
Note:
Mempool and Ring should be mc/mp supportable.
Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
APIs:
-----
rte_pdump_init()
rte_pdump_enable()
rte_pdump_enable_by_deviceid()
rte_pdump_disable()
rte_pdump_disable_by_deviceid()
rte_pdump_uninit()
rte_pdump_set_socket_dir()
2)app/pdump tool
================
Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
This tool by default runs as secondary process, and provides the support for
the command line options for packet capture.
./build/app/dpdk_pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue id>),
(rx-dev=<iface or pcap file> |
tx-dev=<iface or pcap file>),
[ring-size=<ring size>],
[mbuf-size=<mbuf data size>],
[total-num-mbufs=<number of mbufs>]'
Parameters inside the parenthesis represents the mandatory parameters.
Parameters inside the square brackets represents optional parameters.
User has to pass on packet capture parameters under --pdump parameters, multiples of
--pdump can be passed to capture packets on different port and queue combinations
Operation:
----------
*Tool parse the user command line arguments,
creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
of the device passed in rx-dev|tx-dev parameters.
*Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
to enable packet capturing on a specific port/device_id and queue by passing on
port|device_id, queue, mempool and ring info.
*Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
*Tool can be stopped using SIGINT, upon which tool calls
rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
Note:
CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
3)Test-pmd changes
==================
Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
Similarly any application which needs packet capture should call initialize/uninitialize APIs of
librte_pdump and use pdump tool to start the capture.
4)Packet capture flow between pdump tool and librte_pdump
=========================================================
* Pdump tool (Secondary process) requests packet capture
for specific port|device_id and queue combinations.
*Library in secondary process context creates client socket and communicates
the port|device_id, queue, ring and mempool to server.
*Library initializes server in primary process 'test-pmd' context and server serves
the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
*Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
*Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
*Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
*Library receives the disable packet capture request, communicate the info to server,
server will remove the Ethernet rxtx call-backs.
*Packet capture can be seen using tcpdump command
"tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
5)Example command line
======================
./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
v9:
added a support in rte_pdump_set_socket_dir() to set server and client socket paths
==> http://dpdk.org/dev/patchwork/patch/13450/
updated the documentation for the new changes.
updated the commit messages.
v8:
added server socket argument to rte_pdump_init() API ==> http://dpdk.org/dev/patchwork/patch/13402/
added rte_pdump_set_socket_dir() API.
updated documentation for new changes.
v7:
fixed lines over 90 characters.
v6:
removed below deprecation notice patch from patch set.
http://dpdk.org/dev/patchwork/patch/13372/
v5:
addressed code review comments for below patches
http://dpdk.org/dev/patchwork/patch/12955/
http://dpdk.org/dev/patchwork/patch/12951/
v4:
added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
made doc changes as per doc guidelines.
replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
removed rxtx-dev parameter from pdump tool command line.
v3:
app/pdump: Moved cleanup code from signal handler to main.
divided librte_ether changes into multiple patches.
example command changed in app/pdump application guide
v2:
fix compilation issues for 4.8.3
fix unnecessary #includes
Reshma Pattan (8):
ethdev: use locks to protect Rx/Tx callback lists
ethdev: add new api to add Rx callback as head of the list
ethdev: add new fields to ethdev info struct
ethdev: make get port by name and get name by port public
pdump: add new library for packet capturing support
app/pdump: add pdump tool for packet capturing
app/testpmd: add pdump initialization uninitialization
doc: update doc for packet capture framework
MAINTAINERS | 8 +
app/Makefile | 1 +
app/pdump/Makefile | 45 ++
app/pdump/main.c | 844 +++++++++++++++++++++++++++++
app/test-pmd/testpmd.c | 6 +
config/common_base | 5 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 119 +++++
doc/guides/rel_notes/release_16_07.rst | 13 +
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 +++++
lib/Makefile | 1 +
lib/librte_ether/rte_ethdev.c | 123 +++--
lib/librte_ether/rte_ethdev.h | 60 +++
lib/librte_ether/rte_ether_version.map | 9 +
lib/librte_pdump/Makefile | 55 ++
lib/librte_pdump/rte_pdump.c | 913 ++++++++++++++++++++++++++++++++
lib/librte_pdump/rte_pdump.h | 216 ++++++++
lib/librte_pdump/rte_pdump_version.map | 13 +
mk/rte.app.mk | 1 +
20 files changed, 2512 insertions(+), 44 deletions(-)
create mode 100644 app/pdump/Makefile
create mode 100644 app/pdump/main.c
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
create mode 100644 lib/librte_pdump/Makefile
create mode 100644 lib/librte_pdump/rte_pdump.c
create mode 100644 lib/librte_pdump/rte_pdump.h
create mode 100644 lib/librte_pdump/rte_pdump_version.map
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
--
2.5.0
^ permalink raw reply [relevance 2%]
* Re: [dpdk-dev] [PATCH v8 0/8] add packet capture framework
2016-06-10 23:23 0% ` [dpdk-dev] [PATCH v8 0/8] add " Neil Horman
@ 2016-06-13 8:47 0% ` Pattan, Reshma
0 siblings, 0 replies; 200+ results
From: Pattan, Reshma @ 2016-06-13 8:47 UTC (permalink / raw)
To: Neil Horman; +Cc: dev
Hi,
> -----Original Message-----
> From: Neil Horman [mailto:nhorman@tuxdriver.com]
> Sent: Saturday, June 11, 2016 12:23 AM
> To: Pattan, Reshma <reshma.pattan@intel.com>
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v8 0/8] add packet capture framework
>
> On Fri, Jun 10, 2016 at 05:18:46PM +0100, Reshma Pattan wrote:
> > This patch set include below changes
> >
> > 1)Changes to librte_ether.
> > 2)A new library librte_pdump added for packet capture framework.
> > 3)A new app/pdump tool added for packet capturing.
> > 4)Test pmd changes done to initialize packet capture framework.
> > 5)Documentation update.
> >
> > 1)librte_pdump
> > ==============
> > To support packet capturing on dpdk Ethernet devices, a new library
> > librte_pdump is added.Users can develop their own packet capturing
> application using new library APIs.
> >
> > Operation:
> > ----------
> > The librte_pdump provides APIs to support packet capturing on dpdk Ethernet
> devices.
> > Library provides APIs to initialize the packet capture framework,
> > enable/disable the packet capture and uninitialize the packet capture
> framework.
> >
> > The librte_pdump library works on a client/server model. The server is
> > responsible for enabling or disabling the packet capture and the
> > clients are responsible for requesting the enabling or disabling of the packet
> capture.
> >
> > The packet capture framework, as part of its initialization, creates
> > the pthread and the server socket in the pthread. The application that
> > calls the framework initialization will have the server socket
> > created, either under the path that the application has passed or under the
> default path i.e. either ''/var/run'' for root user or ''$HOME'' for non root user.
> >
> > Applications that request enabling or disabling of the packet capture
> > will have the client socket created either under the ''/var/run/'' for root users
> or ''$HOME'' for not root users to send the requests to the server.
> > The server socket will listen for client requests for enabling or disabling the
> packet capture.
> >
> > Applications using below APIs need to pass port/device_id, queue,
> > mempool and ring parameters. Library uses user provided ring and
> > mempool to mirror the rx/tx packets of the port for users. Users need
> > to dequeue the rings and write the packets to vdev(pcap/tuntap) to view the
> packets using any standard tools.
> >
> > Note:
> > Mempool and Ring should be mc/mp supportable.
> > Mempool mbuf size should be big enough to handle the rx/tx packets of a
> port.
> >
> > APIs:
> > -----
> > rte_pdump_init()
> > rte_pdump_enable()
> > rte_pdump_enable_by_deviceid()
> > rte_pdump_disable()
> > rte_pdump_disable_by_deviceid()
> > rte_pdump_uninit()
> > rte_pdump_set_socket_dir()
> >
> > 2)app/pdump tool
> > ================
> > Tool app/pdump is designed based on librte_pdump for packet capturing in
> DPDK.
> > This tool by default runs as secondary process, and provides the
> > support for the command line options for packet capture.
> >
> > ./build/app/dpdk_pdump --
> > --pdump '(port=<port id> | device_id=<pci id or vdev name>),
> > (queue=<queue id>),
> > (rx-dev=<iface or pcap file> |
> > tx-dev=<iface or pcap file>),
> > [ring-size=<ring size>],
> > [mbuf-size=<mbuf data size>],
> > [total-num-mbufs=<number of mbufs>]'
> >
> > Parameters inside the parenthesis represents the mandatory parameters.
> > Parameters inside the square brackets represents optional parameters.
> > User has to pass on packet capture parameters under --pdump
> > parameters, multiples of --pdump can be passed to capture packets on
> > different port and queue combinations
> >
> > Operation:
> > ----------
> > *Tool parse the user command line arguments, creates the mempool, ring
> > and the PCAP PMD vdev with 'tx_stream' as either of the device passed
> > in rx-dev|tx-dev parameters.
> >
> > *Then calls the APIs of librte_pdump i.e.
> > rte_pdump_enable()/rte_pdump_enable_by_deviceid()
> > to enable packet capturing on a specific port/device_id and queue by
> > passing on
> > port|device_id, queue, mempool and ring info.
> >
> > *Tool runs in while loop to dequeue the packets from the ring and write them
> to pcap device.
> >
> > *Tool can be stopped using SIGINT, upon which tool calls
> > rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated
> resources.
> >
> > Note:
> > CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run
> the pdump tool.
> >
> > 3)Test-pmd changes
> > ==================
> > Changes are done to test-pmd application to initialize/uninitialize the packet
> capture framework.
> > So app/pdump tool can be run to see packets of dpdk ports that are used by
> test-pmd.
> >
> > Similarly any application which needs packet capture should call
> > initialize/uninitialize APIs of librte_pdump and use pdump tool to start the
> capture.
> >
> > 4)Packet capture flow between pdump tool and librte_pdump
> > =========================================================
> > * Pdump tool (Secondary process) requests packet capture for specific
> > port|device_id and queue combinations.
> >
> > *Library in secondary process context creates client socket and
> > communicates the port|device_id, queue, ring and mempool to server.
> >
> > *Library initializes server in primary process 'test-pmd' context and
> > server serves the client request to enable Ethernet rxtx call-backs for a given
> port|device_id and queue.
> >
> > *Copy the rx/tx packets to passed mempool and enqueue the packets to ring
> for secondary process.
> >
> > *Pdump tool will dequeue the packets from ring and writes them to
> > PCAPMD vdev, so ultimately packets will be seen on the device that is passed
> in rx-dev|tx-dev.
> >
> > *Once the pdump tool is terminated with SIGINT it will disable the packet
> capturing.
> >
> > *Library receives the disable packet capture request, communicate the
> > info to server, server will remove the Ethernet rxtx call-backs.
> >
> > *Packet capture can be seen using tcpdump command "tcpdump -ni
> > <iface>" (or) "tcpdump –nr <pcapfile>"
> >
> > 5)Example command line
> > ======================
> > ./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-
> dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-
> size=2176,total-num-mbufs=32768' --pdump
> 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-
> file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
> >
> > v8:
> > added server socket argument to rte_pdump_init() API ==>
> > http://dpdk.org/dev/patchwork/patch/13402/
> > added rte_pdump_set_socket_dir() API.
> > updated documentation for new changes.
> >
> > v7:
> > fixed lines over 90 characters.
> >
> > v6:
> > removed below deprecation notice patch from patch set.
> > http://dpdk.org/dev/patchwork/patch/13372/
> >
> > v5:
> > addressed code review comments for below patches
> > http://dpdk.org/dev/patchwork/patch/12955/
> > http://dpdk.org/dev/patchwork/patch/12951/
> >
> > v4:
> > added missing deprecation notice for ABI changes of rte_eth_dev_info
> structure.
> > made doc changes as per doc guidelines.
> > replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
> > removed rxtx-dev parameter from pdump tool command line.
> >
> > v3:
> > app/pdump: Moved cleanup code from signal handler to main.
> > divided librte_ether changes into multiple patches.
> > example command changed in app/pdump application guide
> >
> > v2:
> > fix compilation issues for 4.8.3
> > fix unnecessary #includes
> >
> >
> > Reshma Pattan (8):
> > librte_ether: protect add/remove of rxtx callbacks with spinlocks
> > librte_ether: add new api rte_eth_add_first_rx_callback
> > librte_ether: add new fields to rte_eth_dev_info struct
> > librte_ether: make rte_eth_dev_get_port_by_name
> > rte_eth_dev_get_name_by_port public
> > lib/librte_pdump: add new library for packet capturing support
> > app/pdump: add pdump tool for packet capturing
> > app/test-pmd: add pdump initialization uninitialization
> > doc: update doc for packet capture framework
> >
> > MAINTAINERS | 8 +
> > app/Makefile | 1 +
> > app/pdump/Makefile | 45 ++
> > app/pdump/main.c | 844 +++++++++++++++++++++++++++++
> > app/test-pmd/testpmd.c | 6 +
> > config/common_base | 5 +
> > doc/guides/prog_guide/index.rst | 1 +
> > doc/guides/prog_guide/pdump_library.rst | 117 +++++
> > doc/guides/rel_notes/release_16_07.rst | 13 +
> > doc/guides/sample_app_ug/index.rst | 1 +
> > doc/guides/sample_app_ug/pdump.rst | 122 +++++
> > lib/Makefile | 1 +
> > lib/librte_ether/rte_ethdev.c | 123 +++--
> > lib/librte_ether/rte_ethdev.h | 60 +++
> > lib/librte_ether/rte_ether_version.map | 9 +
> > lib/librte_pdump/Makefile | 55 ++
> > lib/librte_pdump/rte_pdump.c | 904
> ++++++++++++++++++++++++++++++++
> > lib/librte_pdump/rte_pdump.h | 208 ++++++++
> > lib/librte_pdump/rte_pdump_version.map | 13 +
> > mk/rte.app.mk | 1 +
> > 20 files changed, 2493 insertions(+), 44 deletions(-) create mode
> > 100644 app/pdump/Makefile create mode 100644 app/pdump/main.c create
> > mode 100644 doc/guides/prog_guide/pdump_library.rst
> > create mode 100644 doc/guides/sample_app_ug/pdump.rst
> > create mode 100644 lib/librte_pdump/Makefile create mode 100644
> > lib/librte_pdump/rte_pdump.c create mode 100644
> > lib/librte_pdump/rte_pdump.h create mode 100644
> > lib/librte_pdump/rte_pdump_version.map
> >
> > Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> > --
> > 2.5.0
> >
> >
> This seems useful, but the pcap pmd already accepts pcap formatted files for
> input to send using the pcap library. Shouldn't this functionality be integrated
> with that pmd instead of breaking it out to its own library?
>
The librte_pdump library doesn’t' deal with any PCAP functionality, it is solely to send packets of DPDK ports to applications over the rings.
It is upto the applications to decide on which vdev pmd they would like to use to send the packets further on to the Linux devices.
In this patch set, the pdump tool(application based on librte_pdump) uses pcap pmd vdev. In future we also can replace pcap pmd vdev usage with other virtual device pmd types.
> Neil
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v2] i40e: modify the meaning of single VLAN type
@ 2016-06-13 8:03 4% ` Beilei Xing
2016-06-21 10:29 4% ` Bruce Richardson
0 siblings, 1 reply; 200+ results
From: Beilei Xing @ 2016-06-13 8:03 UTC (permalink / raw)
To: jingjing.wu; +Cc: dev, Beilei Xing
In current i40e codebase, if single VLAN header is added in a packet,
it's treated as inner VLAN. Generally, a single VLAN header is
treated as the outer VLAN header. So change corresponding register
for single VLAN.
At the meanwhile, change the meanings of inner VLAN and outer VLAN.
Signed-off-by: Beilei Xing <beilei.xing@intel.com>
---
v2 changes:
Combine corresponding i40e driver changes into this patch.
doc/guides/rel_notes/release_16_07.rst | 3 +++
drivers/net/i40e/i40e_ethdev.c | 29 ++++++++++++++++++++---------
lib/librte_ether/rte_ethdev.h | 4 ++--
3 files changed, 25 insertions(+), 11 deletions(-)
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index c0f6b02..ae02824 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -135,6 +135,9 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* The meanings of ``ETH_VLAN_TYPE_INNER`` and ``ETH_VLAN_TYPE_OUTER`` in
+ ``rte_vlan_type`` are changed.
+
ABI Changes
-----------
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index 24777d5..3672c52 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -920,12 +920,6 @@ eth_i40e_dev_init(struct rte_eth_dev *dev)
"VLAN ether type");
goto err_setup_pf_switch;
}
- ret = i40e_vlan_tpid_set(dev, ETH_VLAN_TYPE_INNER, ETHER_TYPE_VLAN);
- if (ret != I40E_SUCCESS) {
- PMD_INIT_LOG(ERR, "Failed to set the default outer "
- "VLAN ether type");
- goto err_setup_pf_switch;
- }
/* PF setup, which includes VSI setup */
ret = i40e_pf_setup(pf);
@@ -2392,13 +2386,24 @@ i40e_vlan_tpid_set(struct rte_eth_dev *dev,
uint64_t reg_r = 0, reg_w = 0;
uint16_t reg_id = 0;
int ret = 0;
+ int qinq = dev->data->dev_conf.rxmode.hw_vlan_extend;
switch (vlan_type) {
case ETH_VLAN_TYPE_OUTER:
- reg_id = 2;
+ if (qinq)
+ reg_id = 2;
+ else
+ reg_id = 3;
break;
case ETH_VLAN_TYPE_INNER:
- reg_id = 3;
+ if (qinq)
+ reg_id = 3;
+ else {
+ ret = -EINVAL;
+ PMD_DRV_LOG(ERR, "Unsupported vlan type"
+ "in single vlan.\n");
+ return ret;
+ }
break;
default:
ret = -EINVAL;
@@ -2460,8 +2465,14 @@ i40e_vlan_offload_set(struct rte_eth_dev *dev, int mask)
}
if (mask & ETH_VLAN_EXTEND_MASK) {
- if (dev->data->dev_conf.rxmode.hw_vlan_extend)
+ if (dev->data->dev_conf.rxmode.hw_vlan_extend) {
i40e_vsi_config_double_vlan(vsi, TRUE);
+ /* Set global registers with default ether type value */
+ i40e_vlan_tpid_set(dev, ETH_VLAN_TYPE_OUTER,
+ ETHER_TYPE_VLAN);
+ i40e_vlan_tpid_set(dev, ETH_VLAN_TYPE_INNER,
+ ETHER_TYPE_VLAN);
+ }
else
i40e_vsi_config_double_vlan(vsi, FALSE);
}
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 2757510..c5c29fb 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -363,8 +363,8 @@ struct rte_eth_rxmode {
*/
enum rte_vlan_type {
ETH_VLAN_TYPE_UNKNOWN = 0,
- ETH_VLAN_TYPE_INNER, /**< Single VLAN, or inner VLAN. */
- ETH_VLAN_TYPE_OUTER, /**< Outer VLAN. */
+ ETH_VLAN_TYPE_INNER, /**< Inner VLAN. */
+ ETH_VLAN_TYPE_OUTER, /**< Single VLAN, or outer VLAN. */
ETH_VLAN_TYPE_MAX,
};
--
2.5.0
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-12 23:16 0% ` Ananyev, Konstantin
@ 2016-06-13 1:06 0% ` Lu, Wenzhuo
0 siblings, 0 replies; 200+ results
From: Lu, Wenzhuo @ 2016-06-13 1:06 UTC (permalink / raw)
To: Ananyev, Konstantin, Tao, Zhe, dev
Cc: Richardson, Bruce, Chen, Jing D, Liang, Cunming, Wu, Jingjing,
Zhang, Helin
Hi Konstantin,
> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Monday, June 13, 2016 7:17 AM
> To: Lu, Wenzhuo; Tao, Zhe; dev@dpdk.org
> Cc: Richardson, Bruce; Chen, Jing D; Liang, Cunming; Wu, Jingjing; Zhang,
> Helin
> Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
>
> Hi Wenzhuo,
>
> >
> > Hi Konstantin,
> >
> >
> > > -----Original Message-----
> > > From: Ananyev, Konstantin
> > > Sent: Wednesday, June 8, 2016 5:20 PM
> > > To: Lu, Wenzhuo; Tao, Zhe; dev@dpdk.org
> > > Cc: Richardson, Bruce; Chen, Jing D; Liang, Cunming; Wu, Jingjing;
> > > Zhang, Helin
> > > Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
> > >
> > >
> > >
> > > >
> > > > Hi Konstantin,
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Ananyev, Konstantin
> > > > > Sent: Tuesday, June 7, 2016 5:59 PM
> > > > > To: Tao, Zhe; dev@dpdk.org
> > > > > Cc: Lu, Wenzhuo; Richardson, Bruce; Chen, Jing D; Liang,
> > > > > Cunming; Wu, Jingjing; Zhang, Helin
> > > > > Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock
> > > > > mode
> > > > >
> > > > >
> > > > > Hi Zhe & Wenzhuo,
> > > > >
> > > > > Please find my comments below.
> > > > > BTW, for clarification - is that patch for 16.11?
> > > > > I believe it's too late to introduce such significant change in 16.07.
> > > > > Thanks
> > > > > Konstantin
> > > > Thanks for the comments.
> > > > Honestly, our purpose is 16.07. Realizing the big impact, we use
> > > > NEXT_ABI to comment our change. So, I think although we want to
> > > > merge it in
> > > 16.07 this change will become effective after we remove NEXT_ABI in
> 16.11.
> > >
> > > I don't think it is achievable.
> > > First I think your code is not in proper shape yet, right now.
> > > Second, as you said, it is a significant change and I would like to
> > > hear opinions from the rest of the community.
> > Agree it should have risk. I mean our target is 16.07. But surely if it can be
> achieved depends on the feedback from the community.
> >
> > >
> > > >
> > > > >
> > > > > > Define lock mode for RX/TX queue. Because when resetting the
> > > > > > device we want the resetting thread to get the lock of the
> > > > > > RX/TX queue to make sure the RX/TX is stopped.
> > > > > >
> > > > > > Using next ABI macro for this ABI change as it has too much
> > > > > > impact. 7 APIs and 1 global variable are impacted.
> > > > > >
> > > > > > Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> > > > > > Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> > > > > > ---
> > > > > > lib/librte_ether/rte_ethdev.h | 62
> > > > > > +++++++++++++++++++++++++++++++++++++++++++
> > > > > > 1 file changed, 62 insertions(+)
> > > > > >
> > > > > > diff --git a/lib/librte_ether/rte_ethdev.h
> > > > > > b/lib/librte_ether/rte_ethdev.h index 74e895f..4efb5e9 100644
> > > > > > --- a/lib/librte_ether/rte_ethdev.h
> > > > > > +++ b/lib/librte_ether/rte_ethdev.h
> > > > > > @@ -354,7 +354,12 @@ struct rte_eth_rxmode {
> > > > > > jumbo_frame : 1, /**< Jumbo Frame Receipt
> enable. */
> > > > > > hw_strip_crc : 1, /**< Enable CRC stripping by
> hardware. */
> > > > > > enable_scatter : 1, /**< Enable scatter packets rx
> handler */
> > > > > > +#ifndef RTE_NEXT_ABI
> > > > > > enable_lro : 1; /**< Enable LRO */
> > > > > > +#else
> > > > > > + enable_lro : 1, /**< Enable LRO */
> > > > > > + lock_mode : 1; /**< Using lock path */
> > > > > > +#endif
> > > > > > };
> > > > > >
> > > > > > /**
> > > > > > @@ -634,11 +639,68 @@ struct rte_eth_txmode {
> > > > > > /**< If set, reject sending out tagged pkts */
> > > > > > hw_vlan_reject_untagged : 1,
> > > > > > /**< If set, reject sending out untagged pkts */
> > > > > > +#ifndef RTE_NEXT_ABI
> > > > > > hw_vlan_insert_pvid : 1;
> > > > > > /**< If set, enable port based VLAN insertion */
> > > > > > +#else
> > > > > > + hw_vlan_insert_pvid : 1,
> > > > > > + /**< If set, enable port based VLAN insertion */
> > > > > > + lock_mode : 1;
> > > > > > + /**< If set, using lock path */ #endif
> > > > > > };
> > > > > >
> > > > > > /**
> > > > > > + * The macros for the RX/TX lock mode functions */ #ifdef
> > > > > > +RTE_NEXT_ABI #define RX_LOCK_FUNCTION(dev, func) \
> > > > > > + (dev->data->dev_conf.rxmode.lock_mode ? \
> > > > > > + func ## _lock : func)
> > > > > > +
> > > > > > +#define TX_LOCK_FUNCTION(dev, func) \
> > > > > > + (dev->data->dev_conf.txmode.lock_mode ? \
> > > > > > + func ## _lock : func)
> > > > > > +#else
> > > > > > +#define RX_LOCK_FUNCTION(dev, func) func
> > > > > > +
> > > > > > +#define TX_LOCK_FUNCTION(dev, func) func #endif
> > > > > > +
> > > > > > +/* Add the lock RX/TX function for VF reset */ #define
> > > > > > +GENERATE_RX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > > > > > +*rx_queue, \
> > > > > > + struct rte_mbuf **rx_pkts, \
> > > > > > + uint16_t nb_pkts) \
> > > > > > +{ \
> > > > > > + struct nic ## _rx_queue *rxq = rx_queue; \
> > > > > > + uint16_t nb_rx = 0; \
> > > > > > + \
> > > > > > + if (rte_spinlock_trylock(&rxq->rx_lock)) { \
> > > > > > + nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
> > > > > > + rte_spinlock_unlock(&rxq->rx_lock); \
> > > > > > + } \
> > > > > > + \
> > > > > > + return nb_rx; \
> > > > > > +}
> > > > > > +
> > > > > > +#define GENERATE_TX_LOCK(func, nic) \ uint16_t func ##
> > > > > > +_lock(void *tx_queue, \
> > > > > > + struct rte_mbuf **tx_pkts, \
> > > > > > + uint16_t nb_pkts) \
> > > > > > +{ \
> > > > > > + struct nic ## _tx_queue *txq = tx_queue; \
> > > > > > + uint16_t nb_tx = 0; \
> > > > > > + \
> > > > > > + if (rte_spinlock_trylock(&txq->tx_lock)) { \
> > > > > > + nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
> > > > > > + rte_spinlock_unlock(&txq->tx_lock); \
> > > > > > + } \
> > > > > > + \
> > > > > > + return nb_tx; \
> > > > > > +}
> > > > >
> > > > > 1. As I said in off-line dicussiion, I think this locking could
> > > > > (and I think better be) impelented completely on rte_ethdev layer.
> > > > > So actual PMD code will be unaffected.
> > > > > Again that avoids us to introduce _lock version of every RX/Tx
> > > > > function in each PMD.
> > > > One purpose of implementing the lock in PMD layer is to avoid ABI
> > > > change. But we introduce the field lock_mode in struct
> > > > rte_eth_rx/txmode. So seems it's not a good reason now :) The
> > > > other purpose is we want to add a lock for every queue. But in rte
> > > > layer the queue is void *, so we add the lock in the specific
> > > > structures of the NICs. But as
> > > you mentioned below, we can add the lock as
> > > dev->data->rx_queue_state it the struct rte_eth_dev_data.
> > > > So, I prefer to add the lock in rte layer now.
> > >
> > > OK.
> > >
> > > >
> > > > >
> > > > > 2. Again, as discussed offline, I think it is better to have an
> > > > > explicit
> > > > > rte_eth_(rx|tx)_burst_lock(sync?) API, instead of add new fileds
> > > > > into RX/TX config strcutures.
> > > > > Would help to avoid any confusion, I think.
> > > > We want the users to choose the rx/tx path without lock if
> > > > they're sensitive to the performance and can handle the reset
> > > > event in their APP. After
> > > introducing new fields of config struct, users can change the config
> > > to choose the different path.
> > >
> > > I understand what you are doing.
> > >
> > > > If we introduce new API, it may be harder for the use to use it. I
> > > > mean when users want to use lock mode, they may need to replace
> > > > all the
> > > rte_eth_rx/tx_burst by rte_eth_rx/tx_burst_lock.
> > >
> > > Yes, my opinion if users would like to use locking API they need to
> > > call it explicitly.
> > >
> > >
> > > >So if we add the lock in rte layer, I still prefer adding lock_mode
> > > >in the configuration, and the rte_eth_rx/tx_burst is changed like
> > > >this, rte_eth_rx/tx_burst {
> > > > + if lock_mode
> > > > + try_lock
> > > > ......
> > > > + if lock_mode
> > > > + release_lock
> > > > }
> > >
> > > My preference is to keep existing rx/tx_burst() functions unaffected
> > > by that patch.
> > > At least for now.
> > > I suppose that will minimise the risks and help users to avoid
> > > confusion what API
> > > (locking/non-locking) is in use.
> > OK. Let me add new APIs.
> >
> > >
> > > >
> > > >
> > > > >
> > > > > 3. I thought the plan was to introduce a locking in all
> > > > > appropriate control path functions (dev_start/dev_stop etc.)
> > > > > Without that locking version of RX/TX seems a bit useless.
> > > > > Yes, I understand that you do use locking inside dev_reset, but
> > > > > I suppose the plan was to have a generic solution, no?
> > > > > Again, interrupt fire when user invokes dev_start/stop or so, so
> > > > > we still need some synchronisation between them.
> > > > >
> > > > > To be more specific, I thought about something like that:
> > > > >
> > > > > static inline uint16_t
> > > > > rte_eth_rx_burst_lock(uint8_t port_id, uint16_t queue_id,
> > > > > struct rte_mbuf **rx_pkts, const uint16_t nb_pkts) {
> > > > > struct rte_eth_dev *dev = &rte_eth_devices[port_id];
> > > > >
> > > > > #ifdef RTE_LIBRTE_ETHDEV_DEBUG
> > > > > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
> > > > > RTE_FUNC_PTR_OR_ERR_RET(*dev->rx_pkt_burst, 0);
> > > > >
> > > > > if (queue_id >= dev->data->nb_rx_queues) {
> > > > > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n",
> queue_id);
> > > > > return 0;
> > > > > }
> > > > > #endif
> > > > >
> > > > > + if
> > > > > + (rte_spinlock_trylock(&dev->data->rx_queue_state[rx_queue_id].
> > > > > + lock)
> > > == 0)
> > > > > + return 0;
> > > > > + else if (dev->data->rx_queue_state[rx_queue_id] ==
> > > > > RTE_ETH_QUEUE_STATE_STOPPED)) {
> > > > > + rte_spinlock_unlock(&dev->data-
> >rx_queue_state[rx_queue_id].unlock);
> > > > > + return 0;
> > > > > +
> > > > >
> > > > > nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
> > > > > rx_pkts, nb_pkts);
> > > > >
> > > > > + rte_spinlock_unlock(&dev->data-
> >rx_queue_state[rx_queue_id].un
> > > > > + lock
> > > > > + );
> > > > >
> > > > > ....
> > > > >
> > > > > return nb_rx;
> > > > > }
> > > > >
> > > > > And inside queue_start:
> > > > >
> > > > > int
> > > > > rte_eth_dev_rx_queue_start(uint8_t port_id, uint16_t rx_queue_id)
> {
> > > > > struct rte_eth_dev *dev;
> > > > >
> > > > > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
> > > > >
> > > > > dev = &rte_eth_devices[port_id];
> > > > > if (rx_queue_id >= dev->data->nb_rx_queues) {
> > > > > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n",
> > > rx_queue_id);
> > > > > return -EINVAL;
> > > > > }
> > > > >
> > > > > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->rx_queue_start,
> > > > > -ENOTSUP);
> > > > >
> > > > >
> > > > > rte_spinlock_lock(&dev->data->rx_queue_state[rx_queue_id].lock)
> > > > I think you add the lock here to stop the rx/tx.
> > > > But to my opinion, we should lock the rx/tx much earlier before
> > > > starting the queue. For example, when stop the port, the resource
> > > > of the
> > > queues may be released.
> > >
> > > I didn't get you here...
> > > Before releasing the queue resources, queue_stop() has to be executed,
> right?
> > Sorry, I saw your example with rte_eth_dev_rx_queue_start, I didn't
> > know you also want to change rte_eth_dev_rx_queue_stop too.
> > Agree this should work it we call queue_start/stop when reset the
> > port. But we will not call them. I find the queue_stop/start are per- queue
> functions and not supported by all NICs.
>
> But right now you do reset only for ixgbe/i40e.
Not only for ixgbe/i40e. You forget igb, which doesn't support queue_start/stop :)
> For these devices we defiantly do support queue start/stop.
> And again, it is not only about reset op.
> If we want to add rx locked (synced), I think it should be in sync with all
> control API that changes queue state.
> As I said before it is a lot of work and a lot of hassle...
> So probably the easiest (and might be safiest) way just leave things as there
> are right now:
> we allow user to setup a callback on VF reset, and it is user responsibility to
> make sure no RX/TX is active while reset operation is performed.
> Pretty much what Olivier and Stephen suggested, as I understand.
Agree. It's not a good way to add lock for just one feature. It could be tricky if we want to extend the lock to other features. A whole picture is needed.
We've sent another patch set to let the user setup a callback on VF reset. Depend on that, user can use existing rte APIs to reset the VF port. But how about your opinion if we add a specific rte_reset API? It may be easier for the user.
> Konstantin
>
> > Our solution now is stop the whole port and restart the whole port. We
> will not stop/restart queue by queue.
> >
> > >
> > > >The rx/tx cannot be executed. So I prefer to get the lock before
> > > >stopping the
> > > ports.
> > >
> > > Might be I wasn't clear enough here.
> > > What I think we need to have:
> > > -To stop/start/rx/tx the queue (or do any other action that might
> > > change the queue internal structure)
> > > you have to grab the lock.
> > > After queue is stopped it's state has to be changed to
> > > QUEUE_STATE_STOPPED (whti queue lock grabbed),
> > > so rx/tx_locked wouldn't proceed with that queue.
> > > - dev_stop() - has to stop all its queues first, i.e. it needs to
> > > call queue_stop() for all of them.
> > > So after dev_stop() had finished - all device queues have to be in
> > > QUEUE_STATE_STOPPED Same about dev_start() - after it does all other
> > > things - it will call queue_start() for all it's queues.
> > > that will bring them into QUEUE_STARTED.
> > > After that rx/tx_locked can use them again.
> > >
> > > >Maybe better to keep the spinlock in the dev_reset.
> > >
> > > Might be not :)
> > >
> > > >
> > > > >
> > > > > if (dev->data->rx_queue_state[rx_queue_id] !=
> > > > > RTE_ETH_QUEUE_STATE_STOPPED) {
> > > > > RTE_PMD_DEBUG_TRACE("Queue %" PRIu16" of device
> > > > > with port_id=%" PRIu8
> > > > > " already started\n",
> > > > > rx_queue_id, port_id);
> > > > > ret = -EINVAL 0;
> > > > > } else
> > > > > ret = dev->dev_ops->rx_queue_start(dev, rx_queue_id);
> > > > >
> > > > >
> > > > > rte_spinlock_unlock(&dev->data-
> >rx_queue_state[rx_queue_id].unlo
> > > > > ck);
> > > > >
> > > > > return ret;
> > > > > }
> > > > >
> > > > > Then again, we don't need to do explicit locking inside dev_reset().
> > > > > Does it make sense to you guys?
> > > > Please see the answer above.
> > > >
> > > > >
> > > > >
> > > > > > +
> > > > > > +/**
> > > > > > * A structure used to configure an RX ring of an Ethernet port.
> > > > > > */
> > > > > > struct rte_eth_rxconf {
> > > > > > --
> > > > > > 2.1.4
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-12 2:00 0% ` Lu, Wenzhuo
@ 2016-06-12 23:16 0% ` Ananyev, Konstantin
2016-06-13 1:06 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-06-12 23:16 UTC (permalink / raw)
To: Lu, Wenzhuo, Tao, Zhe, dev
Cc: Richardson, Bruce, Chen, Jing D, Liang, Cunming, Wu, Jingjing,
Zhang, Helin
Hi Wenzhuo,
>
> Hi Konstantin,
>
>
> > -----Original Message-----
> > From: Ananyev, Konstantin
> > Sent: Wednesday, June 8, 2016 5:20 PM
> > To: Lu, Wenzhuo; Tao, Zhe; dev@dpdk.org
> > Cc: Richardson, Bruce; Chen, Jing D; Liang, Cunming; Wu, Jingjing; Zhang, Helin
> > Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
> >
> >
> >
> > >
> > > Hi Konstantin,
> > >
> > >
> > > > -----Original Message-----
> > > > From: Ananyev, Konstantin
> > > > Sent: Tuesday, June 7, 2016 5:59 PM
> > > > To: Tao, Zhe; dev@dpdk.org
> > > > Cc: Lu, Wenzhuo; Richardson, Bruce; Chen, Jing D; Liang, Cunming;
> > > > Wu, Jingjing; Zhang, Helin
> > > > Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
> > > >
> > > >
> > > > Hi Zhe & Wenzhuo,
> > > >
> > > > Please find my comments below.
> > > > BTW, for clarification - is that patch for 16.11?
> > > > I believe it's too late to introduce such significant change in 16.07.
> > > > Thanks
> > > > Konstantin
> > > Thanks for the comments.
> > > Honestly, our purpose is 16.07. Realizing the big impact, we use
> > > NEXT_ABI to comment our change. So, I think although we want to merge it in
> > 16.07 this change will become effective after we remove NEXT_ABI in 16.11.
> >
> > I don't think it is achievable.
> > First I think your code is not in proper shape yet, right now.
> > Second, as you said, it is a significant change and I would like to hear opinions
> > from the rest of the community.
> Agree it should have risk. I mean our target is 16.07. But surely if it can be achieved depends on the feedback from the community.
>
> >
> > >
> > > >
> > > > > Define lock mode for RX/TX queue. Because when resetting the
> > > > > device we want the resetting thread to get the lock of the RX/TX
> > > > > queue to make sure the RX/TX is stopped.
> > > > >
> > > > > Using next ABI macro for this ABI change as it has too much
> > > > > impact. 7 APIs and 1 global variable are impacted.
> > > > >
> > > > > Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> > > > > Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> > > > > ---
> > > > > lib/librte_ether/rte_ethdev.h | 62
> > > > > +++++++++++++++++++++++++++++++++++++++++++
> > > > > 1 file changed, 62 insertions(+)
> > > > >
> > > > > diff --git a/lib/librte_ether/rte_ethdev.h
> > > > > b/lib/librte_ether/rte_ethdev.h index 74e895f..4efb5e9 100644
> > > > > --- a/lib/librte_ether/rte_ethdev.h
> > > > > +++ b/lib/librte_ether/rte_ethdev.h
> > > > > @@ -354,7 +354,12 @@ struct rte_eth_rxmode {
> > > > > jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
> > > > > hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
> > > > > enable_scatter : 1, /**< Enable scatter packets rx handler */
> > > > > +#ifndef RTE_NEXT_ABI
> > > > > enable_lro : 1; /**< Enable LRO */
> > > > > +#else
> > > > > + enable_lro : 1, /**< Enable LRO */
> > > > > + lock_mode : 1; /**< Using lock path */
> > > > > +#endif
> > > > > };
> > > > >
> > > > > /**
> > > > > @@ -634,11 +639,68 @@ struct rte_eth_txmode {
> > > > > /**< If set, reject sending out tagged pkts */
> > > > > hw_vlan_reject_untagged : 1,
> > > > > /**< If set, reject sending out untagged pkts */
> > > > > +#ifndef RTE_NEXT_ABI
> > > > > hw_vlan_insert_pvid : 1;
> > > > > /**< If set, enable port based VLAN insertion */
> > > > > +#else
> > > > > + hw_vlan_insert_pvid : 1,
> > > > > + /**< If set, enable port based VLAN insertion */
> > > > > + lock_mode : 1;
> > > > > + /**< If set, using lock path */ #endif
> > > > > };
> > > > >
> > > > > /**
> > > > > + * The macros for the RX/TX lock mode functions */ #ifdef
> > > > > +RTE_NEXT_ABI #define RX_LOCK_FUNCTION(dev, func) \
> > > > > + (dev->data->dev_conf.rxmode.lock_mode ? \
> > > > > + func ## _lock : func)
> > > > > +
> > > > > +#define TX_LOCK_FUNCTION(dev, func) \
> > > > > + (dev->data->dev_conf.txmode.lock_mode ? \
> > > > > + func ## _lock : func)
> > > > > +#else
> > > > > +#define RX_LOCK_FUNCTION(dev, func) func
> > > > > +
> > > > > +#define TX_LOCK_FUNCTION(dev, func) func #endif
> > > > > +
> > > > > +/* Add the lock RX/TX function for VF reset */ #define
> > > > > +GENERATE_RX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > > > > +*rx_queue, \
> > > > > + struct rte_mbuf **rx_pkts, \
> > > > > + uint16_t nb_pkts) \
> > > > > +{ \
> > > > > + struct nic ## _rx_queue *rxq = rx_queue; \
> > > > > + uint16_t nb_rx = 0; \
> > > > > + \
> > > > > + if (rte_spinlock_trylock(&rxq->rx_lock)) { \
> > > > > + nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
> > > > > + rte_spinlock_unlock(&rxq->rx_lock); \
> > > > > + } \
> > > > > + \
> > > > > + return nb_rx; \
> > > > > +}
> > > > > +
> > > > > +#define GENERATE_TX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > > > > +*tx_queue, \
> > > > > + struct rte_mbuf **tx_pkts, \
> > > > > + uint16_t nb_pkts) \
> > > > > +{ \
> > > > > + struct nic ## _tx_queue *txq = tx_queue; \
> > > > > + uint16_t nb_tx = 0; \
> > > > > + \
> > > > > + if (rte_spinlock_trylock(&txq->tx_lock)) { \
> > > > > + nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
> > > > > + rte_spinlock_unlock(&txq->tx_lock); \
> > > > > + } \
> > > > > + \
> > > > > + return nb_tx; \
> > > > > +}
> > > >
> > > > 1. As I said in off-line dicussiion, I think this locking could (and
> > > > I think better be) impelented completely on rte_ethdev layer.
> > > > So actual PMD code will be unaffected.
> > > > Again that avoids us to introduce _lock version of every RX/Tx
> > > > function in each PMD.
> > > One purpose of implementing the lock in PMD layer is to avoid ABI
> > > change. But we introduce the field lock_mode in struct
> > > rte_eth_rx/txmode. So seems it's not a good reason now :) The other
> > > purpose is we want to add a lock for every queue. But in rte layer the
> > > queue is void *, so we add the lock in the specific structures of the NICs. But as
> > you mentioned below, we can add the lock as dev->data->rx_queue_state it the
> > struct rte_eth_dev_data.
> > > So, I prefer to add the lock in rte layer now.
> >
> > OK.
> >
> > >
> > > >
> > > > 2. Again, as discussed offline, I think it is better to have an
> > > > explicit
> > > > rte_eth_(rx|tx)_burst_lock(sync?) API, instead of add new fileds
> > > > into RX/TX config strcutures.
> > > > Would help to avoid any confusion, I think.
> > > We want the users to choose the rx/tx path without lock if they're
> > > sensitive to the performance and can handle the reset event in their APP. After
> > introducing new fields of config struct, users can change the config to choose
> > the different path.
> >
> > I understand what you are doing.
> >
> > > If we introduce new API, it may be harder for the use to use it. I
> > > mean when users want to use lock mode, they may need to replace all the
> > rte_eth_rx/tx_burst by rte_eth_rx/tx_burst_lock.
> >
> > Yes, my opinion if users would like to use locking API they need to call it
> > explicitly.
> >
> >
> > >So if we add the lock in rte layer, I still prefer adding lock_mode in
> > >the configuration, and the rte_eth_rx/tx_burst is changed like this,
> > >rte_eth_rx/tx_burst {
> > > + if lock_mode
> > > + try_lock
> > > ......
> > > + if lock_mode
> > > + release_lock
> > > }
> >
> > My preference is to keep existing rx/tx_burst() functions unaffected by that
> > patch.
> > At least for now.
> > I suppose that will minimise the risks and help users to avoid confusion what API
> > (locking/non-locking) is in use.
> OK. Let me add new APIs.
>
> >
> > >
> > >
> > > >
> > > > 3. I thought the plan was to introduce a locking in all appropriate
> > > > control path functions (dev_start/dev_stop etc.) Without that
> > > > locking version of RX/TX seems a bit useless.
> > > > Yes, I understand that you do use locking inside dev_reset, but I
> > > > suppose the plan was to have a generic solution, no?
> > > > Again, interrupt fire when user invokes dev_start/stop or so, so we
> > > > still need some synchronisation between them.
> > > >
> > > > To be more specific, I thought about something like that:
> > > >
> > > > static inline uint16_t
> > > > rte_eth_rx_burst_lock(uint8_t port_id, uint16_t queue_id,
> > > > struct rte_mbuf **rx_pkts, const uint16_t nb_pkts) {
> > > > struct rte_eth_dev *dev = &rte_eth_devices[port_id];
> > > >
> > > > #ifdef RTE_LIBRTE_ETHDEV_DEBUG
> > > > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
> > > > RTE_FUNC_PTR_OR_ERR_RET(*dev->rx_pkt_burst, 0);
> > > >
> > > > if (queue_id >= dev->data->nb_rx_queues) {
> > > > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", queue_id);
> > > > return 0;
> > > > }
> > > > #endif
> > > >
> > > > + if (rte_spinlock_trylock(&dev->data->rx_queue_state[rx_queue_id].lock)
> > == 0)
> > > > + return 0;
> > > > + else if (dev->data->rx_queue_state[rx_queue_id] ==
> > > > RTE_ETH_QUEUE_STATE_STOPPED)) {
> > > > + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> > > > + return 0;
> > > > +
> > > >
> > > > nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
> > > > rx_pkts, nb_pkts);
> > > >
> > > > + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock
> > > > + );
> > > >
> > > > ....
> > > >
> > > > return nb_rx;
> > > > }
> > > >
> > > > And inside queue_start:
> > > >
> > > > int
> > > > rte_eth_dev_rx_queue_start(uint8_t port_id, uint16_t rx_queue_id) {
> > > > struct rte_eth_dev *dev;
> > > >
> > > > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
> > > >
> > > > dev = &rte_eth_devices[port_id];
> > > > if (rx_queue_id >= dev->data->nb_rx_queues) {
> > > > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n",
> > rx_queue_id);
> > > > return -EINVAL;
> > > > }
> > > >
> > > > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->rx_queue_start,
> > > > -ENOTSUP);
> > > >
> > > > rte_spinlock_lock(&dev->data->rx_queue_state[rx_queue_id].lock)
> > > I think you add the lock here to stop the rx/tx.
> > > But to my opinion, we should lock the rx/tx much earlier before
> > > starting the queue. For example, when stop the port, the resource of the
> > queues may be released.
> >
> > I didn't get you here...
> > Before releasing the queue resources, queue_stop() has to be executed, right?
> Sorry, I saw your example with rte_eth_dev_rx_queue_start, I didn't know you also want to change rte_eth_dev_rx_queue_stop
> too.
> Agree this should work it we call queue_start/stop when reset the port. But we will not call them. I find the queue_stop/start are per-
> queue functions and not supported by all NICs.
But right now you do reset only for ixgbe/i40e.
For these devices we defiantly do support queue start/stop.
And again, it is not only about reset op.
If we want to add rx locked (synced), I think it should be in sync with all control API
that changes queue state.
As I said before it is a lot of work and a lot of hassle...
So probably the easiest (and might be safiest) way just leave things as there are right now:
we allow user to setup a callback on VF reset, and it is user responsibility to make
sure no RX/TX is active while reset operation is performed.
Pretty much what Olivier and Stephen suggested, as I understand.
Konstantin
> Our solution now is stop the whole port and restart the whole port. We will not stop/restart queue by queue.
>
> >
> > >The rx/tx cannot be executed. So I prefer to get the lock before stopping the
> > ports.
> >
> > Might be I wasn't clear enough here.
> > What I think we need to have:
> > -To stop/start/rx/tx the queue (or do any other action that might change the
> > queue internal structure)
> > you have to grab the lock.
> > After queue is stopped it's state has to be changed to
> > QUEUE_STATE_STOPPED (whti queue lock grabbed),
> > so rx/tx_locked wouldn't proceed with that queue.
> > - dev_stop() - has to stop all its queues first, i.e. it needs to call queue_stop()
> > for all of them.
> > So after dev_stop() had finished - all device queues have to be in
> > QUEUE_STATE_STOPPED
> > Same about dev_start() - after it does all other things - it will call queue_start()
> > for all it's queues.
> > that will bring them into QUEUE_STARTED.
> > After that rx/tx_locked can use them again.
> >
> > >Maybe better to keep the spinlock in the dev_reset.
> >
> > Might be not :)
> >
> > >
> > > >
> > > > if (dev->data->rx_queue_state[rx_queue_id] !=
> > > > RTE_ETH_QUEUE_STATE_STOPPED) {
> > > > RTE_PMD_DEBUG_TRACE("Queue %" PRIu16" of device with
> > > > port_id=%" PRIu8
> > > > " already started\n",
> > > > rx_queue_id, port_id);
> > > > ret = -EINVAL 0;
> > > > } else
> > > > ret = dev->dev_ops->rx_queue_start(dev, rx_queue_id);
> > > >
> > > >
> > > > rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> > > >
> > > > return ret;
> > > > }
> > > >
> > > > Then again, we don't need to do explicit locking inside dev_reset().
> > > > Does it make sense to you guys?
> > > Please see the answer above.
> > >
> > > >
> > > >
> > > > > +
> > > > > +/**
> > > > > * A structure used to configure an RX ring of an Ethernet port.
> > > > > */
> > > > > struct rte_eth_rxconf {
> > > > > --
> > > > > 2.1.4
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-09 7:50 0% ` Olivier Matz
@ 2016-06-12 5:25 0% ` Lu, Wenzhuo
0 siblings, 0 replies; 200+ results
From: Lu, Wenzhuo @ 2016-06-12 5:25 UTC (permalink / raw)
To: Olivier Matz, Stephen Hemminger; +Cc: dev, Tao, Zhe
Hi Olivier,
> -----Original Message-----
> From: Olivier Matz [mailto:olivier.matz@6wind.com]
> Sent: Thursday, June 9, 2016 3:51 PM
> To: Lu, Wenzhuo; Stephen Hemminger
> Cc: dev@dpdk.org; Tao, Zhe
> Subject: Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
>
> Hi,
>
> On 06/08/2016 09:34 AM, Lu, Wenzhuo wrote:
> > Hi Stephen,
> >
> >
> >> -----Original Message-----
> >> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> >> Sent: Wednesday, June 8, 2016 10:16 AM
> >> To: Lu, Wenzhuo
> >> Cc: dev@dpdk.org; Tao, Zhe
> >> Subject: Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX
> >> lock mode
> >>
> >> On Mon, 6 Jun 2016 13:40:47 +0800
> >> Wenzhuo Lu <wenzhuo.lu@intel.com> wrote:
> >>
> >>> Define lock mode for RX/TX queue. Because when resetting the device
> >>> we want the resetting thread to get the lock of the RX/TX queue to
> >>> make sure the RX/TX is stopped.
> >>>
> >>> Using next ABI macro for this ABI change as it has too much impact.
> >>> 7 APIs and 1 global variable are impacted.
> >>>
> >>> Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> >>> Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> >>
> >> Why does this patch set make a different assumption the rest of the DPDK?
> >>
> >> The rest of the DPDK operates on the principle that the application
> >> is smart enough to stop the device before making changes. There is no
> >> equivalent to the Linux kernel RTNL mutex. The API assumes
> >> application threads are well behaved and will not try and sabotage each
> other.
> >>
> >> If you restrict the reset operation to only being available when
> >> RX/TX is stopped, then no lock is needed.
> >>
> >> The fact that it requires lots more locking inside each device driver
> >> implies to me this is not correct way to architect this.
>
> +1
>
> I'm not sure adding locks is the proper way to do.
> This is the application responsibility to ensure that:
> - control functions are not called concurrently on the same port
> - rx/tx functions are not called when the device is stopped/reset/...
>
> However, I do think the usage paradigms of the ethdev api should be better
> documented in rte_ethdev.h (ex: which functions can be called concurrently).
> This would be a first step.
>
> If we really want a helper API to do that in DPDK, the _next_ step could be to
> add them in the ethdev api to achieve this. Maybe something like (the function
> names could be better):
>
> - to be called on one control thread:
>
> rte_eth_stop_rxtx(port)
> rte_eth_start_rxtx(port)
>
> rte_eth_get_rxtx_state(port)
> -> return "running" if at least one core is inside the rx/tx code
> -> return "stopped" if all cores are outside the rx/tx code
>
> - to be called on dataplane cores:
>
> /* same than rte_eth_rx_burst(), but checks if rx/tx is allowed
> * first, else do nothing */
> rte_eth_rx_burst_interruptible()
> rte_eth_tx_burst_interruptible()
>
>
> The code of control thread could be:
>
> rte_eth_stop_rxtx(port);
> /* wait that all dataplane cores finished their processing */
> while (rte_eth_get_rxtx_state(port) != stopped)
> ;
> rte_eth_some_control_operation(port);
> rte_eth_start_rxtx(port);
>
>
> I think this could be done without any lock, just with the proper memory barriers
> and a per-core status.
>
> But this API may impose a paradigm to the application, and I'm not sure the
> DPDK should do that.
I don't quite catch your point. Seems your solution still need the APP to change the code. I think it's more complex than just letting the APP to stop the rx/tx and reset the port. Our purpose of this patch set is to let APP do less as possible. It's not a good choice if we make it more complex.
And seems it's hard to stop and start rx/tx in rte layer. Normally APP should do that. To my opinion, we have to introduce lock in rte to achieve that.
>
> Regards,
> Olivier
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-08 9:19 0% ` Ananyev, Konstantin
@ 2016-06-12 2:00 0% ` Lu, Wenzhuo
2016-06-12 23:16 0% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Lu, Wenzhuo @ 2016-06-12 2:00 UTC (permalink / raw)
To: Ananyev, Konstantin, Tao, Zhe, dev
Cc: Richardson, Bruce, Chen, Jing D, Liang, Cunming, Wu, Jingjing,
Zhang, Helin
Hi Konstantin,
> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Wednesday, June 8, 2016 5:20 PM
> To: Lu, Wenzhuo; Tao, Zhe; dev@dpdk.org
> Cc: Richardson, Bruce; Chen, Jing D; Liang, Cunming; Wu, Jingjing; Zhang, Helin
> Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
>
>
>
> >
> > Hi Konstantin,
> >
> >
> > > -----Original Message-----
> > > From: Ananyev, Konstantin
> > > Sent: Tuesday, June 7, 2016 5:59 PM
> > > To: Tao, Zhe; dev@dpdk.org
> > > Cc: Lu, Wenzhuo; Richardson, Bruce; Chen, Jing D; Liang, Cunming;
> > > Wu, Jingjing; Zhang, Helin
> > > Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
> > >
> > >
> > > Hi Zhe & Wenzhuo,
> > >
> > > Please find my comments below.
> > > BTW, for clarification - is that patch for 16.11?
> > > I believe it's too late to introduce such significant change in 16.07.
> > > Thanks
> > > Konstantin
> > Thanks for the comments.
> > Honestly, our purpose is 16.07. Realizing the big impact, we use
> > NEXT_ABI to comment our change. So, I think although we want to merge it in
> 16.07 this change will become effective after we remove NEXT_ABI in 16.11.
>
> I don't think it is achievable.
> First I think your code is not in proper shape yet, right now.
> Second, as you said, it is a significant change and I would like to hear opinions
> from the rest of the community.
Agree it should have risk. I mean our target is 16.07. But surely if it can be achieved depends on the feedback from the community.
>
> >
> > >
> > > > Define lock mode for RX/TX queue. Because when resetting the
> > > > device we want the resetting thread to get the lock of the RX/TX
> > > > queue to make sure the RX/TX is stopped.
> > > >
> > > > Using next ABI macro for this ABI change as it has too much
> > > > impact. 7 APIs and 1 global variable are impacted.
> > > >
> > > > Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> > > > Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> > > > ---
> > > > lib/librte_ether/rte_ethdev.h | 62
> > > > +++++++++++++++++++++++++++++++++++++++++++
> > > > 1 file changed, 62 insertions(+)
> > > >
> > > > diff --git a/lib/librte_ether/rte_ethdev.h
> > > > b/lib/librte_ether/rte_ethdev.h index 74e895f..4efb5e9 100644
> > > > --- a/lib/librte_ether/rte_ethdev.h
> > > > +++ b/lib/librte_ether/rte_ethdev.h
> > > > @@ -354,7 +354,12 @@ struct rte_eth_rxmode {
> > > > jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
> > > > hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
> > > > enable_scatter : 1, /**< Enable scatter packets rx handler */
> > > > +#ifndef RTE_NEXT_ABI
> > > > enable_lro : 1; /**< Enable LRO */
> > > > +#else
> > > > + enable_lro : 1, /**< Enable LRO */
> > > > + lock_mode : 1; /**< Using lock path */
> > > > +#endif
> > > > };
> > > >
> > > > /**
> > > > @@ -634,11 +639,68 @@ struct rte_eth_txmode {
> > > > /**< If set, reject sending out tagged pkts */
> > > > hw_vlan_reject_untagged : 1,
> > > > /**< If set, reject sending out untagged pkts */
> > > > +#ifndef RTE_NEXT_ABI
> > > > hw_vlan_insert_pvid : 1;
> > > > /**< If set, enable port based VLAN insertion */
> > > > +#else
> > > > + hw_vlan_insert_pvid : 1,
> > > > + /**< If set, enable port based VLAN insertion */
> > > > + lock_mode : 1;
> > > > + /**< If set, using lock path */ #endif
> > > > };
> > > >
> > > > /**
> > > > + * The macros for the RX/TX lock mode functions */ #ifdef
> > > > +RTE_NEXT_ABI #define RX_LOCK_FUNCTION(dev, func) \
> > > > + (dev->data->dev_conf.rxmode.lock_mode ? \
> > > > + func ## _lock : func)
> > > > +
> > > > +#define TX_LOCK_FUNCTION(dev, func) \
> > > > + (dev->data->dev_conf.txmode.lock_mode ? \
> > > > + func ## _lock : func)
> > > > +#else
> > > > +#define RX_LOCK_FUNCTION(dev, func) func
> > > > +
> > > > +#define TX_LOCK_FUNCTION(dev, func) func #endif
> > > > +
> > > > +/* Add the lock RX/TX function for VF reset */ #define
> > > > +GENERATE_RX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > > > +*rx_queue, \
> > > > + struct rte_mbuf **rx_pkts, \
> > > > + uint16_t nb_pkts) \
> > > > +{ \
> > > > + struct nic ## _rx_queue *rxq = rx_queue; \
> > > > + uint16_t nb_rx = 0; \
> > > > + \
> > > > + if (rte_spinlock_trylock(&rxq->rx_lock)) { \
> > > > + nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
> > > > + rte_spinlock_unlock(&rxq->rx_lock); \
> > > > + } \
> > > > + \
> > > > + return nb_rx; \
> > > > +}
> > > > +
> > > > +#define GENERATE_TX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > > > +*tx_queue, \
> > > > + struct rte_mbuf **tx_pkts, \
> > > > + uint16_t nb_pkts) \
> > > > +{ \
> > > > + struct nic ## _tx_queue *txq = tx_queue; \
> > > > + uint16_t nb_tx = 0; \
> > > > + \
> > > > + if (rte_spinlock_trylock(&txq->tx_lock)) { \
> > > > + nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
> > > > + rte_spinlock_unlock(&txq->tx_lock); \
> > > > + } \
> > > > + \
> > > > + return nb_tx; \
> > > > +}
> > >
> > > 1. As I said in off-line dicussiion, I think this locking could (and
> > > I think better be) impelented completely on rte_ethdev layer.
> > > So actual PMD code will be unaffected.
> > > Again that avoids us to introduce _lock version of every RX/Tx
> > > function in each PMD.
> > One purpose of implementing the lock in PMD layer is to avoid ABI
> > change. But we introduce the field lock_mode in struct
> > rte_eth_rx/txmode. So seems it's not a good reason now :) The other
> > purpose is we want to add a lock for every queue. But in rte layer the
> > queue is void *, so we add the lock in the specific structures of the NICs. But as
> you mentioned below, we can add the lock as dev->data->rx_queue_state it the
> struct rte_eth_dev_data.
> > So, I prefer to add the lock in rte layer now.
>
> OK.
>
> >
> > >
> > > 2. Again, as discussed offline, I think it is better to have an
> > > explicit
> > > rte_eth_(rx|tx)_burst_lock(sync?) API, instead of add new fileds
> > > into RX/TX config strcutures.
> > > Would help to avoid any confusion, I think.
> > We want the users to choose the rx/tx path without lock if they're
> > sensitive to the performance and can handle the reset event in their APP. After
> introducing new fields of config struct, users can change the config to choose
> the different path.
>
> I understand what you are doing.
>
> > If we introduce new API, it may be harder for the use to use it. I
> > mean when users want to use lock mode, they may need to replace all the
> rte_eth_rx/tx_burst by rte_eth_rx/tx_burst_lock.
>
> Yes, my opinion if users would like to use locking API they need to call it
> explicitly.
>
>
> >So if we add the lock in rte layer, I still prefer adding lock_mode in
> >the configuration, and the rte_eth_rx/tx_burst is changed like this,
> >rte_eth_rx/tx_burst {
> > + if lock_mode
> > + try_lock
> > ......
> > + if lock_mode
> > + release_lock
> > }
>
> My preference is to keep existing rx/tx_burst() functions unaffected by that
> patch.
> At least for now.
> I suppose that will minimise the risks and help users to avoid confusion what API
> (locking/non-locking) is in use.
OK. Let me add new APIs.
>
> >
> >
> > >
> > > 3. I thought the plan was to introduce a locking in all appropriate
> > > control path functions (dev_start/dev_stop etc.) Without that
> > > locking version of RX/TX seems a bit useless.
> > > Yes, I understand that you do use locking inside dev_reset, but I
> > > suppose the plan was to have a generic solution, no?
> > > Again, interrupt fire when user invokes dev_start/stop or so, so we
> > > still need some synchronisation between them.
> > >
> > > To be more specific, I thought about something like that:
> > >
> > > static inline uint16_t
> > > rte_eth_rx_burst_lock(uint8_t port_id, uint16_t queue_id,
> > > struct rte_mbuf **rx_pkts, const uint16_t nb_pkts) {
> > > struct rte_eth_dev *dev = &rte_eth_devices[port_id];
> > >
> > > #ifdef RTE_LIBRTE_ETHDEV_DEBUG
> > > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
> > > RTE_FUNC_PTR_OR_ERR_RET(*dev->rx_pkt_burst, 0);
> > >
> > > if (queue_id >= dev->data->nb_rx_queues) {
> > > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", queue_id);
> > > return 0;
> > > }
> > > #endif
> > >
> > > + if (rte_spinlock_trylock(&dev->data->rx_queue_state[rx_queue_id].lock)
> == 0)
> > > + return 0;
> > > + else if (dev->data->rx_queue_state[rx_queue_id] ==
> > > RTE_ETH_QUEUE_STATE_STOPPED)) {
> > > + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> > > + return 0;
> > > +
> > >
> > > nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
> > > rx_pkts, nb_pkts);
> > >
> > > + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock
> > > + );
> > >
> > > ....
> > >
> > > return nb_rx;
> > > }
> > >
> > > And inside queue_start:
> > >
> > > int
> > > rte_eth_dev_rx_queue_start(uint8_t port_id, uint16_t rx_queue_id) {
> > > struct rte_eth_dev *dev;
> > >
> > > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
> > >
> > > dev = &rte_eth_devices[port_id];
> > > if (rx_queue_id >= dev->data->nb_rx_queues) {
> > > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n",
> rx_queue_id);
> > > return -EINVAL;
> > > }
> > >
> > > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->rx_queue_start,
> > > -ENOTSUP);
> > >
> > > rte_spinlock_lock(&dev->data->rx_queue_state[rx_queue_id].lock)
> > I think you add the lock here to stop the rx/tx.
> > But to my opinion, we should lock the rx/tx much earlier before
> > starting the queue. For example, when stop the port, the resource of the
> queues may be released.
>
> I didn't get you here...
> Before releasing the queue resources, queue_stop() has to be executed, right?
Sorry, I saw your example with rte_eth_dev_rx_queue_start, I didn't know you also want to change rte_eth_dev_rx_queue_stop too.
Agree this should work it we call queue_start/stop when reset the port. But we will not call them. I find the queue_stop/start are per-queue functions and not supported by all NICs.
Our solution now is stop the whole port and restart the whole port. We will not stop/restart queue by queue.
>
> >The rx/tx cannot be executed. So I prefer to get the lock before stopping the
> ports.
>
> Might be I wasn't clear enough here.
> What I think we need to have:
> -To stop/start/rx/tx the queue (or do any other action that might change the
> queue internal structure)
> you have to grab the lock.
> After queue is stopped it's state has to be changed to
> QUEUE_STATE_STOPPED (whti queue lock grabbed),
> so rx/tx_locked wouldn't proceed with that queue.
> - dev_stop() - has to stop all its queues first, i.e. it needs to call queue_stop()
> for all of them.
> So after dev_stop() had finished - all device queues have to be in
> QUEUE_STATE_STOPPED
> Same about dev_start() - after it does all other things - it will call queue_start()
> for all it's queues.
> that will bring them into QUEUE_STARTED.
> After that rx/tx_locked can use them again.
>
> >Maybe better to keep the spinlock in the dev_reset.
>
> Might be not :)
>
> >
> > >
> > > if (dev->data->rx_queue_state[rx_queue_id] !=
> > > RTE_ETH_QUEUE_STATE_STOPPED) {
> > > RTE_PMD_DEBUG_TRACE("Queue %" PRIu16" of device with
> > > port_id=%" PRIu8
> > > " already started\n",
> > > rx_queue_id, port_id);
> > > ret = -EINVAL 0;
> > > } else
> > > ret = dev->dev_ops->rx_queue_start(dev, rx_queue_id);
> > >
> > >
> > > rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> > >
> > > return ret;
> > > }
> > >
> > > Then again, we don't need to do explicit locking inside dev_reset().
> > > Does it make sense to you guys?
> > Please see the answer above.
> >
> > >
> > >
> > > > +
> > > > +/**
> > > > * A structure used to configure an RX ring of an Ethernet port.
> > > > */
> > > > struct rte_eth_rxconf {
> > > > --
> > > > 2.1.4
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v8 0/8] add packet capture framework
2016-06-10 16:18 2% ` [dpdk-dev] [PATCH v8 0/8] add packet capture framework Reshma Pattan
2016-06-10 16:18 5% ` [dpdk-dev] [PATCH v8 8/8] doc: update doc for " Reshma Pattan
@ 2016-06-10 23:23 0% ` Neil Horman
2016-06-13 8:47 0% ` Pattan, Reshma
2016-06-14 9:38 2% ` [dpdk-dev] [PATCH v9 " Reshma Pattan
2 siblings, 1 reply; 200+ results
From: Neil Horman @ 2016-06-10 23:23 UTC (permalink / raw)
To: Reshma Pattan; +Cc: dev
On Fri, Jun 10, 2016 at 05:18:46PM +0100, Reshma Pattan wrote:
> This patch set include below changes
>
> 1)Changes to librte_ether.
> 2)A new library librte_pdump added for packet capture framework.
> 3)A new app/pdump tool added for packet capturing.
> 4)Test pmd changes done to initialize packet capture framework.
> 5)Documentation update.
>
> 1)librte_pdump
> ==============
> To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
> is added.Users can develop their own packet capturing application using new library APIs.
>
> Operation:
> ----------
> The librte_pdump provides APIs to support packet capturing on dpdk Ethernet devices.
> Library provides APIs to initialize the packet capture framework, enable/disable
> the packet capture and uninitialize the packet capture framework.
>
> The librte_pdump library works on a client/server model. The server is responsible for enabling or
> disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
> the packet capture.
>
> The packet capture framework, as part of its initialization, creates the pthread and the server socket in
> the pthread. The application that calls the framework initialization will have the server socket created,
> either under the path that the application has passed or under the default path i.e. either ''/var/run'' for
> root user or ''$HOME'' for non root user.
>
> Applications that request enabling or disabling of the packet capture will have the client socket created either under
> the ''/var/run/'' for root users or ''$HOME'' for not root users to send the requests to the server.
> The server socket will listen for client requests for enabling or disabling the packet capture.
>
> Applications using below APIs need to pass port/device_id, queue, mempool and
> ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
> packets of the port for users. Users need to dequeue the rings and write the packets
> to vdev(pcap/tuntap) to view the packets using any standard tools.
>
> Note:
> Mempool and Ring should be mc/mp supportable.
> Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
>
> APIs:
> -----
> rte_pdump_init()
> rte_pdump_enable()
> rte_pdump_enable_by_deviceid()
> rte_pdump_disable()
> rte_pdump_disable_by_deviceid()
> rte_pdump_uninit()
> rte_pdump_set_socket_dir()
>
> 2)app/pdump tool
> ================
> Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
> This tool by default runs as secondary process, and provides the support for
> the command line options for packet capture.
>
> ./build/app/dpdk_pdump --
> --pdump '(port=<port id> | device_id=<pci id or vdev name>),
> (queue=<queue id>),
> (rx-dev=<iface or pcap file> |
> tx-dev=<iface or pcap file>),
> [ring-size=<ring size>],
> [mbuf-size=<mbuf data size>],
> [total-num-mbufs=<number of mbufs>]'
>
> Parameters inside the parenthesis represents the mandatory parameters.
> Parameters inside the square brackets represents optional parameters.
> User has to pass on packet capture parameters under --pdump parameters, multiples of
> --pdump can be passed to capture packets on different port and queue combinations
>
> Operation:
> ----------
> *Tool parse the user command line arguments,
> creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
> of the device passed in rx-dev|tx-dev parameters.
>
> *Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
> to enable packet capturing on a specific port/device_id and queue by passing on
> port|device_id, queue, mempool and ring info.
>
> *Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
>
> *Tool can be stopped using SIGINT, upon which tool calls
> rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
>
> Note:
> CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
>
> 3)Test-pmd changes
> ==================
> Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
> So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
>
> Similarly any application which needs packet capture should call initialize/uninitialize APIs of
> librte_pdump and use pdump tool to start the capture.
>
> 4)Packet capture flow between pdump tool and librte_pdump
> =========================================================
> * Pdump tool (Secondary process) requests packet capture
> for specific port|device_id and queue combinations.
>
> *Library in secondary process context creates client socket and communicates
> the port|device_id, queue, ring and mempool to server.
>
> *Library initializes server in primary process 'test-pmd' context and server serves
> the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
>
> *Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
>
> *Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
> so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
>
> *Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
>
> *Library receives the disable packet capture request, communicate the info to server,
> server will remove the Ethernet rxtx call-backs.
>
> *Packet capture can be seen using tcpdump command
> "tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
>
> 5)Example command line
> ======================
> ./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
>
> v8:
> added server socket argument to rte_pdump_init() API ==> http://dpdk.org/dev/patchwork/patch/13402/
> added rte_pdump_set_socket_dir() API.
> updated documentation for new changes.
>
> v7:
> fixed lines over 90 characters.
>
> v6:
> removed below deprecation notice patch from patch set.
> http://dpdk.org/dev/patchwork/patch/13372/
>
> v5:
> addressed code review comments for below patches
> http://dpdk.org/dev/patchwork/patch/12955/
> http://dpdk.org/dev/patchwork/patch/12951/
>
> v4:
> added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
> made doc changes as per doc guidelines.
> replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
> removed rxtx-dev parameter from pdump tool command line.
>
> v3:
> app/pdump: Moved cleanup code from signal handler to main.
> divided librte_ether changes into multiple patches.
> example command changed in app/pdump application guide
>
> v2:
> fix compilation issues for 4.8.3
> fix unnecessary #includes
>
>
> Reshma Pattan (8):
> librte_ether: protect add/remove of rxtx callbacks with spinlocks
> librte_ether: add new api rte_eth_add_first_rx_callback
> librte_ether: add new fields to rte_eth_dev_info struct
> librte_ether: make rte_eth_dev_get_port_by_name
> rte_eth_dev_get_name_by_port public
> lib/librte_pdump: add new library for packet capturing support
> app/pdump: add pdump tool for packet capturing
> app/test-pmd: add pdump initialization uninitialization
> doc: update doc for packet capture framework
>
> MAINTAINERS | 8 +
> app/Makefile | 1 +
> app/pdump/Makefile | 45 ++
> app/pdump/main.c | 844 +++++++++++++++++++++++++++++
> app/test-pmd/testpmd.c | 6 +
> config/common_base | 5 +
> doc/guides/prog_guide/index.rst | 1 +
> doc/guides/prog_guide/pdump_library.rst | 117 +++++
> doc/guides/rel_notes/release_16_07.rst | 13 +
> doc/guides/sample_app_ug/index.rst | 1 +
> doc/guides/sample_app_ug/pdump.rst | 122 +++++
> lib/Makefile | 1 +
> lib/librte_ether/rte_ethdev.c | 123 +++--
> lib/librte_ether/rte_ethdev.h | 60 +++
> lib/librte_ether/rte_ether_version.map | 9 +
> lib/librte_pdump/Makefile | 55 ++
> lib/librte_pdump/rte_pdump.c | 904 ++++++++++++++++++++++++++++++++
> lib/librte_pdump/rte_pdump.h | 208 ++++++++
> lib/librte_pdump/rte_pdump_version.map | 13 +
> mk/rte.app.mk | 1 +
> 20 files changed, 2493 insertions(+), 44 deletions(-)
> create mode 100644 app/pdump/Makefile
> create mode 100644 app/pdump/main.c
> create mode 100644 doc/guides/prog_guide/pdump_library.rst
> create mode 100644 doc/guides/sample_app_ug/pdump.rst
> create mode 100644 lib/librte_pdump/Makefile
> create mode 100644 lib/librte_pdump/rte_pdump.c
> create mode 100644 lib/librte_pdump/rte_pdump.h
> create mode 100644 lib/librte_pdump/rte_pdump_version.map
>
> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> --
> 2.5.0
>
>
This seems useful, but the pcap pmd already accepts pcap formatted files for
input to send using the pcap library. Shouldn't this functionality be
integrated with that pmd instead of breaking it out to its own library?
Neil
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v8 8/8] doc: update doc for packet capture framework
2016-06-10 16:18 2% ` [dpdk-dev] [PATCH v8 0/8] add packet capture framework Reshma Pattan
@ 2016-06-10 16:18 5% ` Reshma Pattan
2016-06-10 23:23 0% ` [dpdk-dev] [PATCH v8 0/8] add " Neil Horman
2016-06-14 9:38 2% ` [dpdk-dev] [PATCH v9 " Reshma Pattan
2 siblings, 0 replies; 200+ results
From: Reshma Pattan @ 2016-06-10 16:18 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
Added programmers guide for librte_pdump.
Added sample application guide for app/pdump application.
Updated release note for packet capture framework changes.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
---
MAINTAINERS | 3 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 117 ++++++++++++++++++++++++++++++
doc/guides/rel_notes/release_16_07.rst | 13 ++++
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 ++++++++++++++++++++++++++++++++
6 files changed, 257 insertions(+)
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
diff --git a/MAINTAINERS b/MAINTAINERS
index a48c8de..ce7c941 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -436,6 +436,9 @@ Pdump
M: Reshma Pattan <reshma.pattan@intel.com>
F: lib/librte_pdump/
F: app/pdump/
+F: doc/guides/prog_guide/pdump_library.rst
+F: doc/guides/sample_app_ug/pdump.rst
+
Hierarchical scheduler
M: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
diff --git a/doc/guides/prog_guide/index.rst b/doc/guides/prog_guide/index.rst
index b862d0c..4caf969 100644
--- a/doc/guides/prog_guide/index.rst
+++ b/doc/guides/prog_guide/index.rst
@@ -71,6 +71,7 @@ Programmer's Guide
writing_efficient_code
profile_app
glossary
+ pdump_library
**Figures**
diff --git a/doc/guides/prog_guide/pdump_library.rst b/doc/guides/prog_guide/pdump_library.rst
new file mode 100644
index 0000000..3088063
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,117 @@
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+.. _pdump_library:
+
+The librte_pdump Library
+========================
+
+The ``librte_pdump`` library provides a framework for packet capturing in DPDK.
+The library provides the following APIs to initialize the packet capture framework, to enable
+or disable the packet capture, and to uninitialize it:
+
+* ``rte_pdump_init()``:
+ This API initializes the packet capture framework.
+
+* ``rte_pdump_enable()``:
+ This API enables the packet capture on a given port and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_enable_by_deviceid()``:
+ This API enables the packet capture on a given device id (``vdev name or pci address``) and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_disable()``:
+ This API disables the packet capture on a given port and queue.
+
+* ``rte_pdump_disable_by_deviceid()``:
+ This API disables the packet capture on a given device id (``vdev name or pci address``) and queue.
+
+* ``rte_pdump_uninit()``:
+ This API uninitializes the packet capture framework.
+
+* ``rte_pdump_set_socket_dir()``:
+ This API sets the server socket path.
+ Note: This API is not thread-safe.
+
+
+Operation
+---------
+
+The ``librte_pdump`` library works on a client/server model. The server is responsible for enabling or
+disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
+the packet capture.
+
+The packet capture framework, as part of its initialization, creates the pthread and the server socket in
+the pthread. The application that calls the framework initialization will have the server socket created,
+either under the path that the application has passed or under the default path i.e. either ``/var/run`` for
+root user or ``$HOME`` for non root user.
+
+Applications that request enabling or disabling of the packet capture will have the client socket created either under
+the ``/var/run/`` for root users or ``$HOME`` for not root users to send the requests to the server.
+The server socket will listen for client requests for enabling or disabling the packet capture.
+
+
+Implementation Details
+----------------------
+
+The library API ``rte_pdump_init()``, initializes the packet capture framework by creating the pthread and the server
+socket. The server socket in the pthread context will be listening to the client requests to enable or disable the
+packet capture.
+
+The library APIs ``rte_pdump_enable()`` and ``rte_pdump_enable_by_deviceid()`` enables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump enable" request and sends
+the request to the server. The server that is listening on the socket will take the request and enable the packet capture
+by registering the Ethernet RX and TX callbacks for the given port or device_id and queue combinations.
+Then the server will mirror the packets to the new mempool and enqueue them to the rte_ring that clients have passed
+to these APIs. The server also sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, the client socket is closed.
+
+The library APIs ``rte_pdump_disable()`` and ``rte_pdump_disable_by_deviceid()`` disables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump disable" request and sends
+the request to the server. The server that is listening on the socket will take the request and disable the packet
+capture by removing the Ethernet RX and TX callbacks for the given port or device_id and queue combinations. The server
+also sends the response back to the client about the status of the request that was processed. After the response is
+received from the server, the client socket is closed.
+
+The library API ``rte_pdump_uninit()``, uninitializes the packet capture framework by closing the pthread and the
+server socket.
+
+The library API ``rte_pdump_set_server_socket_dir()``, sets the given path as server socket path.
+If the given path is ``NULL``, default path will be selected, i.e. either ``/var/run/`` for root users or ``$HOME``
+for non root users. Clients need to call this API only when their server socket path is non default path.
+The given server socket path will be used by clients to send the pdump enable and disable requests to the server.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+The DPDK ``app/pdump`` tool is developed based on this library to capture packets in DPDK.
+Users can use this as an example to develop their own packet capturing application.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index c0f6b02..a4de2a2 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -66,6 +66,11 @@ New Features
* Enable RSS per network interface through the configuration file.
* Streamline the CLI code.
+* **Added packet capture framework.**
+
+ * A new library ``librte_pdump`` is added to provide packet capture APIs.
+ * A new ``app/pdump`` tool is added to capture packets in DPDK.
+
Resolved Issues
---------------
@@ -135,6 +140,11 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* Function ``rte_eth_dev_get_port_by_name`` changed to a public API.
+
+* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ in the ``rte_eth_dev_info`` object.
+
ABI Changes
-----------
@@ -146,6 +156,9 @@ ABI Changes
* The ``rte_port_source_params`` structure has new fields to support PCAP file.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ to support number of queues configured by software.
+
Shared Library Versions
-----------------------
diff --git a/doc/guides/sample_app_ug/index.rst b/doc/guides/sample_app_ug/index.rst
index 930f68c..96bb317 100644
--- a/doc/guides/sample_app_ug/index.rst
+++ b/doc/guides/sample_app_ug/index.rst
@@ -76,6 +76,7 @@ Sample Applications User Guide
ptpclient
performance_thread
ipsec_secgw
+ pdump
**Figures**
diff --git a/doc/guides/sample_app_ug/pdump.rst b/doc/guides/sample_app_ug/pdump.rst
new file mode 100644
index 0000000..96c8709
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,122 @@
+
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+dpdk_pdump Application
+======================
+
+The ``dpdk_pdump`` application is a Data Plane Development Kit (DPDK) application that runs as a DPDK secondary process and
+is capable of enabling packet capture on dpdk ports.
+
+
+Running the Application
+-----------------------
+
+The application has a ``--pdump`` command line option with various sub arguments:
+
+.. code-block:: console
+
+ ./build/app/dpdk_pdump --
+ --pdump '(port=<port id> | device_id=<pci id or vdev name>),
+ (queue=<queue_id>),
+ (rx-dev=<iface or pcap file> |
+ tx-dev=<iface or pcap file>),
+ [ring-size=<ring size>],
+ [mbuf-size=<mbuf data size>],
+ [total-num-mbufs=<number of mbufs>]'
+
+Note:
+
+* Parameters inside the parentheses represents mandatory parameters.
+
+* Parameters inside the square brackets represents optional parameters.
+
+Multiple instances of ``--pdump`` can be passed to capture packets on different port and queue combinations.
+
+
+Parameters
+~~~~~~~~~~
+
+``port``:
+Port id of the eth device on which packets should be captured.
+
+``device_id``:
+PCI address (or) name of the eth device on which packets should be captured.
+
+ .. Note::
+
+ * As of now the ``dpdk_pdump`` tool cannot capture the packets of virtual devices
+ in the primary process due to a bug in the ethdev library. Due to this bug, in a multi process context,
+ when the primary and secondary have different ports set, then the secondary process
+ (here the ``dpdk_pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
+
+``queue``:
+Queue id of the eth device on which packets should be captured. The user can pass a queue value of ``*`` to enable
+packet capture on all queues of the eth device.
+
+``rx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+``tx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+ .. Note::
+
+ * To receive ingress packets only, ``rx-dev`` should be passed.
+
+ * To receive egress packets only, ``tx-dev`` should be passed.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the different file names or the Linux iface names.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the same file names or the the Linux iface names.
+
+``ring-size``:
+Size of the ring. This value is used internally for ring creation. The ring will be used to enqueue the packets from
+the primary application to the secondary. This is an optional parameter with default size 16384.
+
+``mbuf-size``:
+Size of the mbuf data. This is used internally for mempool creation. Ideally this value must be same as
+the primary application's mempool's mbuf data size which is used for packet RX. This is an optional parameter with
+default size 2176.
+
+``total-num-mbufs``:
+Total number mbufs in mempool. This is used internally for mempool creation. This is an optional parameter with default
+value 65535.
+
+
+Example
+-------
+
+.. code-block:: console
+
+ $ sudo ./build/app/dpdk_pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
--
2.5.0
^ permalink raw reply [relevance 5%]
* [dpdk-dev] [PATCH v8 0/8] add packet capture framework
[not found] <1465487895-5870-1-git-send-email-reshma.pattan@intel.com>
@ 2016-06-10 16:18 2% ` Reshma Pattan
2016-06-10 16:18 5% ` [dpdk-dev] [PATCH v8 8/8] doc: update doc for " Reshma Pattan
` (2 more replies)
0 siblings, 3 replies; 200+ results
From: Reshma Pattan @ 2016-06-10 16:18 UTC (permalink / raw)
To: dev
This patch set include below changes
1)Changes to librte_ether.
2)A new library librte_pdump added for packet capture framework.
3)A new app/pdump tool added for packet capturing.
4)Test pmd changes done to initialize packet capture framework.
5)Documentation update.
1)librte_pdump
==============
To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
is added.Users can develop their own packet capturing application using new library APIs.
Operation:
----------
The librte_pdump provides APIs to support packet capturing on dpdk Ethernet devices.
Library provides APIs to initialize the packet capture framework, enable/disable
the packet capture and uninitialize the packet capture framework.
The librte_pdump library works on a client/server model. The server is responsible for enabling or
disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
the packet capture.
The packet capture framework, as part of its initialization, creates the pthread and the server socket in
the pthread. The application that calls the framework initialization will have the server socket created,
either under the path that the application has passed or under the default path i.e. either ''/var/run'' for
root user or ''$HOME'' for non root user.
Applications that request enabling or disabling of the packet capture will have the client socket created either under
the ''/var/run/'' for root users or ''$HOME'' for not root users to send the requests to the server.
The server socket will listen for client requests for enabling or disabling the packet capture.
Applications using below APIs need to pass port/device_id, queue, mempool and
ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
packets of the port for users. Users need to dequeue the rings and write the packets
to vdev(pcap/tuntap) to view the packets using any standard tools.
Note:
Mempool and Ring should be mc/mp supportable.
Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
APIs:
-----
rte_pdump_init()
rte_pdump_enable()
rte_pdump_enable_by_deviceid()
rte_pdump_disable()
rte_pdump_disable_by_deviceid()
rte_pdump_uninit()
rte_pdump_set_socket_dir()
2)app/pdump tool
================
Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
This tool by default runs as secondary process, and provides the support for
the command line options for packet capture.
./build/app/dpdk_pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue id>),
(rx-dev=<iface or pcap file> |
tx-dev=<iface or pcap file>),
[ring-size=<ring size>],
[mbuf-size=<mbuf data size>],
[total-num-mbufs=<number of mbufs>]'
Parameters inside the parenthesis represents the mandatory parameters.
Parameters inside the square brackets represents optional parameters.
User has to pass on packet capture parameters under --pdump parameters, multiples of
--pdump can be passed to capture packets on different port and queue combinations
Operation:
----------
*Tool parse the user command line arguments,
creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
of the device passed in rx-dev|tx-dev parameters.
*Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
to enable packet capturing on a specific port/device_id and queue by passing on
port|device_id, queue, mempool and ring info.
*Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
*Tool can be stopped using SIGINT, upon which tool calls
rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
Note:
CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
3)Test-pmd changes
==================
Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
Similarly any application which needs packet capture should call initialize/uninitialize APIs of
librte_pdump and use pdump tool to start the capture.
4)Packet capture flow between pdump tool and librte_pdump
=========================================================
* Pdump tool (Secondary process) requests packet capture
for specific port|device_id and queue combinations.
*Library in secondary process context creates client socket and communicates
the port|device_id, queue, ring and mempool to server.
*Library initializes server in primary process 'test-pmd' context and server serves
the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
*Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
*Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
*Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
*Library receives the disable packet capture request, communicate the info to server,
server will remove the Ethernet rxtx call-backs.
*Packet capture can be seen using tcpdump command
"tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
5)Example command line
======================
./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
v8:
added server socket argument to rte_pdump_init() API ==> http://dpdk.org/dev/patchwork/patch/13402/
added rte_pdump_set_socket_dir() API.
updated documentation for new changes.
v7:
fixed lines over 90 characters.
v6:
removed below deprecation notice patch from patch set.
http://dpdk.org/dev/patchwork/patch/13372/
v5:
addressed code review comments for below patches
http://dpdk.org/dev/patchwork/patch/12955/
http://dpdk.org/dev/patchwork/patch/12951/
v4:
added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
made doc changes as per doc guidelines.
replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
removed rxtx-dev parameter from pdump tool command line.
v3:
app/pdump: Moved cleanup code from signal handler to main.
divided librte_ether changes into multiple patches.
example command changed in app/pdump application guide
v2:
fix compilation issues for 4.8.3
fix unnecessary #includes
Reshma Pattan (8):
librte_ether: protect add/remove of rxtx callbacks with spinlocks
librte_ether: add new api rte_eth_add_first_rx_callback
librte_ether: add new fields to rte_eth_dev_info struct
librte_ether: make rte_eth_dev_get_port_by_name
rte_eth_dev_get_name_by_port public
lib/librte_pdump: add new library for packet capturing support
app/pdump: add pdump tool for packet capturing
app/test-pmd: add pdump initialization uninitialization
doc: update doc for packet capture framework
MAINTAINERS | 8 +
app/Makefile | 1 +
app/pdump/Makefile | 45 ++
app/pdump/main.c | 844 +++++++++++++++++++++++++++++
app/test-pmd/testpmd.c | 6 +
config/common_base | 5 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 117 +++++
doc/guides/rel_notes/release_16_07.rst | 13 +
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 +++++
lib/Makefile | 1 +
lib/librte_ether/rte_ethdev.c | 123 +++--
lib/librte_ether/rte_ethdev.h | 60 +++
lib/librte_ether/rte_ether_version.map | 9 +
lib/librte_pdump/Makefile | 55 ++
lib/librte_pdump/rte_pdump.c | 904 ++++++++++++++++++++++++++++++++
lib/librte_pdump/rte_pdump.h | 208 ++++++++
lib/librte_pdump/rte_pdump_version.map | 13 +
mk/rte.app.mk | 1 +
20 files changed, 2493 insertions(+), 44 deletions(-)
create mode 100644 app/pdump/Makefile
create mode 100644 app/pdump/main.c
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
create mode 100644 lib/librte_pdump/Makefile
create mode 100644 lib/librte_pdump/rte_pdump.c
create mode 100644 lib/librte_pdump/rte_pdump.h
create mode 100644 lib/librte_pdump/rte_pdump_version.map
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
--
2.5.0
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH v9 0/3] mempool: add external mempool manager
@ 2016-06-10 15:16 2% ` David Hunt
2016-06-14 9:46 2% ` [dpdk-dev] [PATCH v10 " David Hunt
0 siblings, 1 reply; 200+ results
From: David Hunt @ 2016-06-10 15:16 UTC (permalink / raw)
To: dev; +Cc: olivier.matz, viktorin, jerin.jacob, shreyansh.jain
Here's the latest version of the External Mempool Manager patchset.
It's re-based on top of the latest head as of 09/6/2016, including
Olivier's 35-part patch series on mempool re-org [1]
[1] http://dpdk.org/ml/archives/dev/2016-May/039229.html
v9 changes:
* added a check for NULL alloc in rte_mempool_ops_register
* rte_mempool_alloc_t now returns int instead of void*
* fixed some comment typo's
* removed some unneeded typecasts
* changed a return NULL to return -EEXIST in rte_mempool_ops_register
* fixed rte_mempool_version.map file so builds ok as shared libs
* moved flags check from rte_mempool_create_empty to rte_mempool_create
v8 changes:
* merged first three patches in the series into one.
* changed parameters to ops callback to all be rte_mempool pointer
rather than than pointer to opaque data or uint64.
* comment fixes.
* fixed parameter to _free function (was inconsistent).
* changed MEMPOOL_F_RING_CREATED to MEMPOOL_F_POOL_CREATED
v7 changes:
* Changed rte_mempool_handler_table to rte_mempool_ops_table
* Changed hander_idx to ops_index in rte_mempool struct
* Reworked comments in rte_mempool.h around ops functions
* Changed rte_mempool_hander.c to rte_mempool_ops.c
* Changed all functions containing _handler_ to _ops_
* Now there is no mention of 'handler' left
* Other small changes out of review of mailing list
v6 changes:
* Moved the flags handling from rte_mempool_create_empty to
rte_mempool_create, as it's only there for backward compatibility
* Various comment additions and cleanup
* Renamed rte_mempool_handler to rte_mempool_ops
* Added a union for *pool and u64 pool_id in struct rte_mempool
* split the original patch into a few parts for easier review.
* rename functions with _ext_ to _ops_.
* addressed review comments
* renamed put and get functions to enqueue and dequeue
* changed occurences of rte_mempool_ops to const, as they
contain function pointers (security)
* split out the default external mempool handler into a separate
patch for easier review
v5 changes:
* rebasing, as it is dependent on another patch series [1]
v4 changes (Olivier Matz):
* remove the rte_mempool_create_ext() function. To change the handler, the
user has to do the following:
- mp = rte_mempool_create_empty()
- rte_mempool_set_handler(mp, "my_handler")
- rte_mempool_populate_default(mp)
This avoids to add another function with more than 10 arguments, duplicating
the doxygen comments
* change the api of rte_mempool_alloc_t: only the mempool pointer is required
as all information is available in it
* change the api of rte_mempool_free_t: remove return value
* move inline wrapper functions from the .c to the .h (else they won't be
inlined). This implies to have one header file (rte_mempool.h), or it
would have generate cross dependencies issues.
* remove now unused MEMPOOL_F_INT_HANDLER (note: it was misused anyway due
to the use of && instead of &)
* fix build in debug mode (__MEMPOOL_STAT_ADD(mp, put_pool, n) remaining)
* fix build with shared libraries (global handler has to be declared in
the .map file)
* rationalize #include order
* remove unused function rte_mempool_get_handler_name()
* rename some structures, fields, functions
* remove the static in front of rte_tailq_elem rte_mempool_tailq (comment
from Yuanhan)
* test the ext mempool handler in the same file than standard mempool tests,
avoiding to duplicate the code
* rework the custom handler in mempool_test
* rework a bit the patch selecting default mbuf pool handler
* fix some doxygen comments
v3 changes:
* simplified the file layout, renamed to rte_mempool_handler.[hc]
* moved the default handlers into rte_mempool_default.c
* moved the example handler out into app/test/test_ext_mempool.c
* removed is_mc/is_mp change, slight perf degredation on sp cached operation
* removed stack hanler, may re-introduce at a later date
* Changes out of code reviews
v2 changes:
* There was a lot of duplicate code between rte_mempool_xmem_create and
rte_mempool_create_ext. This has now been refactored and is now
hopefully cleaner.
* The RTE_NEXT_ABI define is now used to allow building of the library
in a format that is compatible with binaries built against previous
versions of DPDK.
* Changes out of code reviews. Hopefully I've got most of them included.
The External Mempool Manager is an extension to the mempool API that allows
users to add and use an external mempool manager, which allows external memory
subsystems such as external hardware memory management systems and software
based memory allocators to be used with DPDK.
The existing API to the internal DPDK mempool manager will remain unchanged
and will be backward compatible. However, there will be an ABI breakage, as
the mempool struct is changing. These changes are all contained withing
RTE_NEXT_ABI defs, and the current or next code can be changed with
the CONFIG_RTE_NEXT_ABI config setting
There are two aspects to external mempool manager.
1. Adding the code for your new mempool operations (ops). This is
achieved by adding a new mempool ops source file into the
librte_mempool library, and using the REGISTER_MEMPOOL_HANDLER macro.
2. Using the new API to call rte_mempool_create_empty and
rte_mempool_set_ops to create a new mempool
using the name parameter to identify which ops to use.
New API calls added
1. A new rte_mempool_create_empty() function
2. rte_mempool_set_ops_byname() which sets the mempool's ops (functions)
3. An rte_mempool_populate_default() and rte_mempool_populate_anon() functions
which populates the mempool using the relevant ops
Several external mempool managers may be used in the same application. A new
mempool can then be created by using the new 'create' function, providing the
mempool ops struct name to point the mempool to the relevant mempool manager
callback structure.
The old 'create' function can still be called by legacy programs, and will
internally work out the mempool handle based on the flags provided (single
producer, single consumer, etc). By default handles are created internally to
implement the built-in DPDK mempool manager and mempool types.
The external mempool manager needs to provide the following functions.
1. alloc - allocates the mempool memory, and adds each object onto a ring
2. put - puts an object back into the mempool once an application has
finished with it
3. get - gets an object from the mempool for use by the application
4. get_count - gets the number of available objects in the mempool
5. free - frees the mempool memory
Every time a get/put/get_count is called from the application/PMD, the
callback for that mempool is called. These functions are in the fastpath,
and any unoptimised ops may limit performance.
The new APIs are as follows:
1. rte_mempool_create_empty
struct rte_mempool *
rte_mempool_create_empty(const char *name, unsigned n, unsigned elt_size,
unsigned cache_size, unsigned private_data_size,
int socket_id, unsigned flags);
2. rte_mempool_set_ops_byname()
int
rte_mempool_set_ops_byname(struct rte_mempool *mp, const char *name);
3. rte_mempool_populate_default()
int rte_mempool_populate_default(struct rte_mempool *mp);
4. rte_mempool_populate_anon()
int rte_mempool_populate_anon(struct rte_mempool *mp);
Please see rte_mempool.h for further information on the parameters.
The important thing to note is that the mempool ops struct is passed by name
to rte_mempool_set_ops_byname, which looks through the ops struct array to
get the ops_index, which is then stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via ops index.
The mempool ops structure contains callbacks to the implementation of
the ops function, and is set up for registration as follows:
static const struct rte_mempool_ops ops_sp_mc = {
.name = "ring_sp_mc",
.alloc = rte_mempool_common_ring_alloc,
.put = common_ring_sp_put,
.get = common_ring_mc_get,
.get_count = common_ring_get_count,
.free = common_ring_free,
};
And then the following macro will register the ops in the array of ops
structures
REGISTER_MEMPOOL_OPS(ops_mp_mc);
For an example of API usage, please see app/test/test_mempool.c, which
implements a rudimentary "custom_handler" mempool manager using simple mallocs
for each mempool object. This file also contains the callbacks and self
registration for the new handler.
David Hunt (2):
mempool: support external mempool operations
mbuf: make default mempool ops configurable at build
Olivier Matz (1):
app/test: test external mempool manager
^ permalink raw reply [relevance 2%]
* Re: [dpdk-dev] [PATCH v3] log: deprecate history dump
2016-06-10 9:50 0% ` David Marchand
@ 2016-06-10 13:09 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-10 13:09 UTC (permalink / raw)
To: dev; +Cc: David Marchand
> > The log history uses rte_mempool. In order to remove the mempool
> > dependency in EAL (and improve the build), this feature is deprecated.
> > The ABI is kept but the behaviour is now voided because it seems this
> > function was not used. The history can be read from syslog.
> >
> > Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
>
> Acked-by: David Marchand <david.marchand@6wind.com>
Applied
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v1] hash: add tsx support for cuckoo hash
@ 2016-06-10 11:09 0% ` De Lara Guarch, Pablo
0 siblings, 0 replies; 200+ results
From: De Lara Guarch, Pablo @ 2016-06-10 11:09 UTC (permalink / raw)
To: Shen, Wei1, Stephen Hemminger; +Cc: dev, Maciocco, Christian, Gobriel, Sameh
> -----Original Message-----
> From: Shen, Wei1
> Sent: Monday, May 09, 2016 5:52 PM
> To: Stephen Hemminger
> Cc: dev@dpdk.org; De Lara Guarch, Pablo; Maciocco, Christian; Gobriel,
> Sameh
> Subject: Re: [dpdk-dev] [PATCH v1] hash: add tsx support for cuckoo hash
>
> Hi Stephen,
>
> Greetings. Thanks for your great feedback. Let’s me address your concern
> here.
>
> 1) It changes ABI, so it breaks old programs
> The patch uses the extra_flag field in the rte_hash_parameters struct to set
> the default insertion behavior. Today there is only one bit used by this flag
> (RTE_HASH_EXTRA_FLAGS_TRANS_MEM_SUPPORT 0x1) and we used the
> next unused bit (RTE_HASH_EXTRA_FLAGS_MULTI_WRITER_ADD 0x2) in this
> patch. So ABI are maintained.
Agree on this. Also, if the problem is on the rte_hash (because of the change of
the cache size or the addition of a new field at the end), that should not be a problem,
as far as I understand, since it is modifying an internal structure
(rte_hash was made internal back in v2.1).
>
> 2) What about older processors, need to detect and handle them at runtime.
> Correct. This patch is based on the previous Transactional Memory patch.
> Since these previous patches already assume the user to check the presence
> of TSX so we build on top this assumption. But I personally agree with you
> that handling TSX check should be made easier.
> http://dpdk.org/ml/archives/dev/2015-June/018571.html
> http://dpdk.org/ml/archives/dev/2015-June/018566.html
>
> 3) Why can't this just be the default behavior with correct fallback to locking
> on older processors.
> This is an excellent point. We discussed this before. Our thought at that time
> is, since TSX insertion is a bit slower than without anything (TSX or other
> locks), it would benefit apps that is designed to have a single writer to the
> hash table (for instance in some master-slave model). We might need more
> feedback from user about whether making it default is more desirable if
> most the app is designed with multi-writer manner.
>
> Thanks,
>
>
> --
> Best,
>
> Wei Shen.
>
>
>
>
>
>
> On 5/6/16, 9:56 PM, "Stephen Hemminger"
> <stephen@networkplumber.org> wrote:
>
> >On Fri, 6 May 2016 21:05:02 +0100
> >Shen Wei <wei1.shen@intel.com> wrote:
> >
> >> --- a/lib/librte_hash/rte_cuckoo_hash.c
> >> +++ b/lib/librte_hash/rte_cuckoo_hash.c
> >> @@ -1,7 +1,7 @@
> >> /*-
> >> * BSD LICENSE
> >> *
> >> - * Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
> >> + * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
> >> * All rights reserved.
> >> *
> >> * Redistribution and use in source and binary forms, with or without
> >> @@ -100,7 +100,9 @@ EAL_REGISTER_TAILQ(rte_hash_tailq)
> >>
> >> #define KEY_ALIGNMENT 16
> >>
> >> -#define LCORE_CACHE_SIZE 8
> >> +#define LCORE_CACHE_SIZE 64
> >> +
> >> +#define RTE_HASH_BFS_QUEUEs_MAX_LEN 5000
> >>
> >> #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
> >> /*
> >> @@ -190,6 +192,7 @@ struct rte_hash {
> >> memory support */
> >> struct lcore_cache *local_free_slots;
> >> /**< Local cache per lcore, storing some indexes of the free slots */
> >> + uint8_t multiwrite_add; /**< Multi-write safe hash add behavior */
> >> } __rte_cache_aligned;
> >>
> >
> >I like the idea of using TSX to allow multi-writer safety, but there are
> >several problems with this patch.
> >
> >1) It changes ABI, so it breaks old programs
> >2) What about older processors, need to detect and handle them at
> runtime.
> >3) Why can't this just be the default behavior with correct
> > fallback to locking on older processors.
> >
> >Actually lock ellision in DPDK is an interesting topic in general that
> >needs to be addressed.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v3] log: deprecate history dump
2016-06-09 22:10 5% ` [dpdk-dev] [PATCH v3] " Thomas Monjalon
@ 2016-06-10 9:50 0% ` David Marchand
2016-06-10 13:09 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: David Marchand @ 2016-06-10 9:50 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev
On Fri, Jun 10, 2016 at 12:10 AM, Thomas Monjalon
<thomas.monjalon@6wind.com> wrote:
> The log history uses rte_mempool. In order to remove the mempool
> dependency in EAL (and improve the build), this feature is deprecated.
> The ABI is kept but the behaviour is now voided because it seems this
> function was not used. The history can be read from syslog.
>
> Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
Acked-by: David Marchand <david.marchand@6wind.com>
--
David Marchand
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v3] log: deprecate history dump
2016-06-09 15:06 5% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
@ 2016-06-09 22:10 5% ` Thomas Monjalon
2016-06-10 9:50 0% ` David Marchand
0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-06-09 22:10 UTC (permalink / raw)
To: david.marchand; +Cc: dev
The log history uses rte_mempool. In order to remove the mempool
dependency in EAL (and improve the build), this feature is deprecated.
The ABI is kept but the behaviour is now voided because it seems this
function was not used. The history can be read from syslog.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
v3:
- keep mempool header path in linuxapp for CONFIG_RTE_LIBRTE_IVSHMEM
v2:
- remove more mempool and log history traces
- add a warning if enabling log history
- move not related mempool includes cleanup in another patch
---
app/test-pmd/cmdline.c | 3 -
app/test/autotest_data.py | 6 --
app/test/autotest_test_funcs.py | 5 --
app/test/commands.c | 4 +-
app/test/test_logs.c | 3 -
doc/guides/prog_guide/mempool_lib.rst | 4 +-
doc/guides/rel_notes/deprecation.rst | 3 +
lib/librte_eal/bsdapp/eal/Makefile | 1 -
lib/librte_eal/bsdapp/eal/eal_debug.c | 6 --
lib/librte_eal/common/eal_common_log.c | 148 ++------------------------------
lib/librte_eal/common/eal_private.h | 3 -
lib/librte_eal/common/include/rte_log.h | 8 ++
lib/librte_eal/linuxapp/eal/eal_debug.c | 6 --
lib/librte_eal/linuxapp/eal/eal_log.c | 9 +-
lib/librte_mempool/rte_mempool.c | 4 -
15 files changed, 20 insertions(+), 193 deletions(-)
diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 1921612..fd389ac 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -7268,8 +7268,6 @@ static void cmd_dump_parsed(void *parsed_result,
rte_dump_physmem_layout(stdout);
else if (!strcmp(res->dump, "dump_memzone"))
rte_memzone_dump(stdout);
- else if (!strcmp(res->dump, "dump_log_history"))
- rte_log_dump_history(stdout);
else if (!strcmp(res->dump, "dump_struct_sizes"))
dump_struct_sizes();
else if (!strcmp(res->dump, "dump_ring"))
@@ -7284,7 +7282,6 @@ cmdline_parse_token_string_t cmd_dump_dump =
TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
"dump_physmem#"
"dump_memzone#"
- "dump_log_history#"
"dump_struct_sizes#"
"dump_ring#"
"dump_mempool#"
diff --git a/app/test/autotest_data.py b/app/test/autotest_data.py
index 78d2edd..6c87809 100644
--- a/app/test/autotest_data.py
+++ b/app/test/autotest_data.py
@@ -94,12 +94,6 @@ parallel_test_group_list = [
"Report" : None,
},
{
- "Name" : "Dump log history",
- "Command" : "dump_log_history",
- "Func" : dump_autotest,
- "Report" : None,
- },
- {
"Name" : "Dump rings",
"Command" : "dump_ring",
"Func" : dump_autotest,
diff --git a/app/test/autotest_test_funcs.py b/app/test/autotest_test_funcs.py
index b60b941..14cffd0 100644
--- a/app/test/autotest_test_funcs.py
+++ b/app/test/autotest_test_funcs.py
@@ -144,16 +144,11 @@ def logs_autotest(child, test_name):
i = 0
child.sendline(test_name)
- # logs sequence is printed twice because of history dump
log_list = [
"TESTAPP1: error message",
"TESTAPP1: critical message",
"TESTAPP2: critical message",
"TESTAPP1: error message",
- "TESTAPP1: error message",
- "TESTAPP1: critical message",
- "TESTAPP2: critical message",
- "TESTAPP1: error message",
]
for log_msg in log_list:
diff --git a/app/test/commands.c b/app/test/commands.c
index e0af8e4..2df46b0 100644
--- a/app/test/commands.c
+++ b/app/test/commands.c
@@ -150,8 +150,6 @@ static void cmd_dump_parsed(void *parsed_result,
rte_dump_physmem_layout(stdout);
else if (!strcmp(res->dump, "dump_memzone"))
rte_memzone_dump(stdout);
- else if (!strcmp(res->dump, "dump_log_history"))
- rte_log_dump_history(stdout);
else if (!strcmp(res->dump, "dump_struct_sizes"))
dump_struct_sizes();
else if (!strcmp(res->dump, "dump_ring"))
@@ -164,7 +162,7 @@ static void cmd_dump_parsed(void *parsed_result,
cmdline_parse_token_string_t cmd_dump_dump =
TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
- "dump_physmem#dump_memzone#dump_log_history#"
+ "dump_physmem#dump_memzone#"
"dump_struct_sizes#dump_ring#dump_mempool#"
"dump_devargs");
diff --git a/app/test/test_logs.c b/app/test/test_logs.c
index 05aa862..d0a9962 100644
--- a/app/test/test_logs.c
+++ b/app/test/test_logs.c
@@ -83,9 +83,6 @@ test_logs(void)
RTE_LOG(ERR, TESTAPP1, "error message\n");
RTE_LOG(ERR, TESTAPP2, "error message (not displayed)\n");
- /* print again the previous logs */
- rte_log_dump_history(stdout);
-
return 0;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index 5fae79a..c3afc2e 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -38,9 +38,7 @@ In the DPDK, it is identified by name and uses a ring to store free objects.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
-This library is used by the
-:ref:`Mbuf Library <Mbuf_Library>` and the
-:ref:`Environment Abstraction Layer <Environment_Abstraction_Layer>` (for logging history).
+This library is used by the :ref:`Mbuf Library <Mbuf_Library>`.
Cookies
-------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index ad05eba..bda40c1 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -8,6 +8,9 @@ API and ABI deprecation notices are to be posted here.
Deprecation Notices
-------------------
+* The log history is deprecated.
+ It is voided in 16.07 and will be removed in release 16.11.
+
* The ethdev hotplug API is going to be moved to EAL with a notification
mechanism added to crypto and ethdev libraries so that hotplug is now
available to both of them. This API will be stripped of the device arguments
diff --git a/lib/librte_eal/bsdapp/eal/Makefile b/lib/librte_eal/bsdapp/eal/Makefile
index 9054ad6..474651b 100644
--- a/lib/librte_eal/bsdapp/eal/Makefile
+++ b/lib/librte_eal/bsdapp/eal/Makefile
@@ -41,7 +41,6 @@ CFLAGS += -I$(SRCDIR)/include
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common/include
CFLAGS += -I$(RTE_SDK)/lib/librte_ring
-CFLAGS += -I$(RTE_SDK)/lib/librte_mempool
CFLAGS += $(WERROR_FLAGS) -O3
LDLIBS += -lexecinfo
diff --git a/lib/librte_eal/bsdapp/eal/eal_debug.c b/lib/librte_eal/bsdapp/eal/eal_debug.c
index 907fbfa..5fbc17c 100644
--- a/lib/librte_eal/bsdapp/eal/eal_debug.c
+++ b/lib/librte_eal/bsdapp/eal/eal_debug.c
@@ -77,9 +77,6 @@ void __rte_panic(const char *funcname, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
rte_log(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, "PANIC in %s():\n", funcname);
va_start(ap, format);
rte_vlog(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, format, ap);
@@ -98,9 +95,6 @@ rte_exit(int exit_code, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
if (exit_code != 0)
RTE_LOG(CRIT, EAL, "Error - exiting with code: %d\n"
" Cause: ", exit_code);
diff --git a/lib/librte_eal/common/eal_common_log.c b/lib/librte_eal/common/eal_common_log.c
index b5e37bb..7916c78 100644
--- a/lib/librte_eal/common/eal_common_log.c
+++ b/lib/librte_eal/common/eal_common_log.c
@@ -31,54 +31,16 @@
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
-#include <string.h>
#include <stdio.h>
#include <stdint.h>
#include <stdarg.h>
-#include <sys/types.h>
#include <stdlib.h>
-#include <unistd.h>
-#include <inttypes.h>
-#include <errno.h>
-#include <sys/queue.h>
#include <rte_log.h>
-#include <rte_memory.h>
-#include <rte_memzone.h>
-#include <rte_launch.h>
-#include <rte_common.h>
-#include <rte_cycles.h>
-#include <rte_eal.h>
#include <rte_per_lcore.h>
-#include <rte_lcore.h>
-#include <rte_atomic.h>
-#include <rte_debug.h>
-#include <rte_spinlock.h>
-#include <rte_branch_prediction.h>
-#include <rte_ring.h>
-#include <rte_mempool.h>
#include "eal_private.h"
-#define LOG_ELT_SIZE 2048
-
-#define LOG_HISTORY_MP_NAME "log_history"
-
-STAILQ_HEAD(log_history_list, log_history);
-
-/**
- * The structure of a message log in the log history.
- */
-struct log_history {
- STAILQ_ENTRY(log_history) next;
- unsigned size;
- char buf[0];
-};
-
-static struct rte_mempool *log_history_mp = NULL;
-static unsigned log_history_size = 0;
-static struct log_history_list log_history;
-
/* global log structure */
struct rte_logs rte_logs = {
.type = ~0,
@@ -86,10 +48,7 @@ struct rte_logs rte_logs = {
.file = NULL,
};
-static rte_spinlock_t log_dump_lock = RTE_SPINLOCK_INITIALIZER;
-static rte_spinlock_t log_list_lock = RTE_SPINLOCK_INITIALIZER;
static FILE *default_log_stream;
-static int history_enabled = 1;
/**
* This global structure stores some informations about the message
@@ -106,59 +65,16 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg, log_cur_msg);
/* default logs */
int
-rte_log_add_in_history(const char *buf, size_t size)
+rte_log_add_in_history(const char *buf __rte_unused, size_t size __rte_unused)
{
- struct log_history *hist_buf = NULL;
- static const unsigned hist_buf_size = LOG_ELT_SIZE - sizeof(*hist_buf);
- void *obj;
-
- if (history_enabled == 0)
- return 0;
-
- rte_spinlock_lock(&log_list_lock);
-
- /* get a buffer for adding in history */
- if (log_history_size > RTE_LOG_HISTORY) {
- hist_buf = STAILQ_FIRST(&log_history);
- if (hist_buf) {
- STAILQ_REMOVE_HEAD(&log_history, next);
- log_history_size--;
- }
- }
- else {
- if (rte_mempool_mc_get(log_history_mp, &obj) < 0)
- obj = NULL;
- hist_buf = obj;
- }
-
- /* no buffer */
- if (hist_buf == NULL) {
- rte_spinlock_unlock(&log_list_lock);
- return -ENOBUFS;
- }
-
- /* not enough room for msg, buffer go back in mempool */
- if (size >= hist_buf_size) {
- rte_mempool_mp_put(log_history_mp, hist_buf);
- rte_spinlock_unlock(&log_list_lock);
- return -ENOBUFS;
- }
-
- /* add in history */
- memcpy(hist_buf->buf, buf, size);
- hist_buf->buf[size] = hist_buf->buf[hist_buf_size-1] = '\0';
- hist_buf->size = size;
- STAILQ_INSERT_TAIL(&log_history, hist_buf, next);
- log_history_size++;
- rte_spinlock_unlock(&log_list_lock);
-
return 0;
}
void
rte_log_set_history(int enable)
{
- history_enabled = enable;
+ if (enable)
+ RTE_LOG(WARNING, EAL, "The log history is deprecated.\n");
}
/* Change the stream that will be used by logging system */
@@ -217,44 +133,8 @@ int rte_log_cur_msg_logtype(void)
/* Dump log history to file */
void
-rte_log_dump_history(FILE *out)
+rte_log_dump_history(FILE *out __rte_unused)
{
- struct log_history_list tmp_log_history;
- struct log_history *hist_buf;
- unsigned i;
-
- /* only one dump at a time */
- rte_spinlock_lock(&log_dump_lock);
-
- /* save list, and re-init to allow logging during dump */
- rte_spinlock_lock(&log_list_lock);
- tmp_log_history = log_history;
- STAILQ_INIT(&log_history);
- log_history_size = 0;
- rte_spinlock_unlock(&log_list_lock);
-
- for (i=0; i<RTE_LOG_HISTORY; i++) {
-
- /* remove one message from history list */
- hist_buf = STAILQ_FIRST(&tmp_log_history);
-
- if (hist_buf == NULL)
- break;
-
- STAILQ_REMOVE_HEAD(&tmp_log_history, next);
-
- /* write on stdout */
- if (fwrite(hist_buf->buf, hist_buf->size, 1, out) == 0) {
- rte_mempool_mp_put(log_history_mp, hist_buf);
- break;
- }
-
- /* put back message structure in pool */
- rte_mempool_mp_put(log_history_mp, hist_buf);
- }
- fflush(out);
-
- rte_spinlock_unlock(&log_dump_lock);
}
/*
@@ -297,29 +177,11 @@ rte_log(uint32_t level, uint32_t logtype, const char *format, ...)
}
/*
- * called by environment-specific log init function to initialize log
- * history
+ * called by environment-specific log init function
*/
int
rte_eal_common_log_init(FILE *default_log)
{
- STAILQ_INIT(&log_history);
-
- /* reserve RTE_LOG_HISTORY*2 elements, so we can dump and
- * keep logging during this time */
- log_history_mp = rte_mempool_create(LOG_HISTORY_MP_NAME, RTE_LOG_HISTORY*2,
- LOG_ELT_SIZE, 0, 0,
- NULL, NULL,
- NULL, NULL,
- SOCKET_ID_ANY, MEMPOOL_F_NO_PHYS_CONTIG);
-
- if ((log_history_mp == NULL) &&
- ((log_history_mp = rte_mempool_lookup(LOG_HISTORY_MP_NAME)) == NULL)){
- RTE_LOG(ERR, EAL, "%s(): cannot create log_history mempool\n",
- __func__);
- return -1;
- }
-
default_log_stream = default_log;
rte_openlog_stream(default_log);
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 2342fa1..857dc3e 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -49,9 +49,6 @@ int rte_eal_memzone_init(void);
/**
* Common log initialization function (private to eal).
*
- * Called by environment-specific log initialization function to initialize
- * log history.
- *
* @param default_log
* The default log stream to be used.
* @return
diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
index 2e47e7f..b1add04 100644
--- a/lib/librte_eal/common/include/rte_log.h
+++ b/lib/librte_eal/common/include/rte_log.h
@@ -42,6 +42,8 @@
* This file provides a log API to RTE applications.
*/
+#include "rte_common.h" /* for __rte_deprecated macro */
+
#ifdef __cplusplus
extern "C" {
#endif
@@ -179,22 +181,27 @@ int rte_log_cur_msg_loglevel(void);
int rte_log_cur_msg_logtype(void);
/**
+ * @deprecated
* Enable or disable the history (enabled by default)
*
* @param enable
* true to enable, or 0 to disable history.
*/
+__rte_deprecated
void rte_log_set_history(int enable);
/**
+ * @deprecated
* Dump the log history to a file
*
* @param f
* A pointer to a file for output
*/
+__rte_deprecated
void rte_log_dump_history(FILE *f);
/**
+ * @deprecated
* Add a log message to the history.
*
* This function can be called from a user-defined log stream. It adds
@@ -209,6 +216,7 @@ void rte_log_dump_history(FILE *f);
* - 0: Success.
* - (-ENOBUFS) if there is no room to store the message.
*/
+__rte_deprecated
int rte_log_add_in_history(const char *buf, size_t size);
/**
diff --git a/lib/librte_eal/linuxapp/eal/eal_debug.c b/lib/librte_eal/linuxapp/eal/eal_debug.c
index 907fbfa..5fbc17c 100644
--- a/lib/librte_eal/linuxapp/eal/eal_debug.c
+++ b/lib/librte_eal/linuxapp/eal/eal_debug.c
@@ -77,9 +77,6 @@ void __rte_panic(const char *funcname, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
rte_log(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, "PANIC in %s():\n", funcname);
va_start(ap, format);
rte_vlog(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, format, ap);
@@ -98,9 +95,6 @@ rte_exit(int exit_code, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
if (exit_code != 0)
RTE_LOG(CRIT, EAL, "Error - exiting with code: %d\n"
" Cause: ", exit_code);
diff --git a/lib/librte_eal/linuxapp/eal/eal_log.c b/lib/librte_eal/linuxapp/eal/eal_log.c
index 0b133c3..d391100 100644
--- a/lib/librte_eal/linuxapp/eal/eal_log.c
+++ b/lib/librte_eal/linuxapp/eal/eal_log.c
@@ -50,8 +50,7 @@
#include "eal_private.h"
/*
- * default log function, used once mempool (hence log history) is
- * available
+ * default log function
*/
static ssize_t
console_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
@@ -60,9 +59,6 @@ console_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
ssize_t ret;
uint32_t loglevel;
- /* add this log in history */
- rte_log_add_in_history(buf, size);
-
/* write on stdout */
ret = fwrite(buf, 1, size, stdout);
fflush(stdout);
@@ -110,8 +106,7 @@ rte_eal_log_init(const char *id, int facility)
/* early logs */
/*
- * early log function, used during boot when mempool (hence log
- * history) is not available
+ * early log function, used before rte_eal_log_init
*/
static ssize_t
early_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index b54de43..22a5645 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -1003,7 +1003,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
if (free == 0) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE1) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1012,7 +1011,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
hdr->cookie = RTE_MEMPOOL_HEADER_COOKIE2;
} else if (free == 1) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE2) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1022,7 +1020,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
} else if (free == 2) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE1 &&
cookie != RTE_MEMPOOL_HEADER_COOKIE2) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1032,7 +1029,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
tlr = __mempool_get_trailer(obj);
cookie = tlr->cookie;
if (cookie != RTE_MEMPOOL_TRAILER_COOKIE) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
--
2.7.0
^ permalink raw reply [relevance 5%]
* Re: [dpdk-dev] [PATCH v7 0/8] add packet capture framework
2016-06-09 16:10 2% ` [dpdk-dev] [PATCH v7 0/8] add " Reshma Pattan
2016-06-09 16:10 5% ` [dpdk-dev] [PATCH v7 8/8] doc: update doc for " Reshma Pattan
@ 2016-06-09 17:34 0% ` Ananyev, Konstantin
1 sibling, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-06-09 17:34 UTC (permalink / raw)
To: Pattan, Reshma, dev
>
> This patch set include below changes
>
> 1)Changes to librte_ether.
> 2)A new library librte_pdump added for packet capture framework.
> 3)A new app/pdump tool added for packet capturing.
> 4)Test pmd changes done to initialize packet capture framework.
> 5)Documentation update.
>
> 1)librte_pdump
> ==============
> To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
> is added.Users can develop their own packet capturing application using new library APIs.
>
> Operation:
> ----------
> Pdump library provides APIs to support packet capturing on dpdk Ethernet devices.
> Library provides APIs to initialize the packet capture framework, enable/disable
> the packet capture and uninitialize the packet capture framework.
>
> Pdump library works on client/server based model.
>
> Sever is responsible for enabling/disabling the packet captures.
> Clients are responsible for requesting enable/disable of the
> packet captures.
>
> As part of packet capture framework initialization, pthread and
> the server socket is created. Only one server socket is allowed on the system.
> As part of enabling/disabling the packet capture, client sockets are created
> and multiple client sockets are allowed.
> Who ever calls initialization first they will succeed with the initialization,
> next subsequent calls of initialization are not allowed. So next users can only
> request enabling/disabling the packet capture.
>
> Applications using below APIs need to pass port/device_id, queue, mempool and
> ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
> packets of the port for users. Users need to dequeue the rings and write the packets
> to vdev(pcap/tuntap) to view the packets using any standard tools.
>
> Note:
> Mempool and Ring should be mc/mp supportable.
> Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
>
> APIs:
> -----
> rte_pdump_init()
> rte_pdump_enable()
> rte_pdump_enable_by_deviceid()
> rte_pdump_disable()
> rte_pdump_disable_by_deviceid()
> rte_pdump_uninit()
>
> 2)app/pdump tool
> ================
> Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
> This tool by default runs as secondary process, and provides the support for
> the command line options for packet capture.
>
> ./build/app/dpdk_pdump --
> --pdump '(port=<port id> | device_id=<pci id or vdev name>),
> (queue=<queue id>),
> (rx-dev=<iface or pcap file> |
> tx-dev=<iface or pcap file>),
> [ring-size=<ring size>],
> [mbuf-size=<mbuf data size>],
> [total-num-mbufs=<number of mbufs>]'
>
> Parameters inside the parenthesis represents the mandatory parameters.
> Parameters inside the square brackets represents optional parameters.
> User has to pass on packet capture parameters under --pdump parameters, multiples of
> --pdump can be passed to capture packets on different port and queue combinations
>
> Operation:
> ----------
> *Tool parse the user command line arguments,
> creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
> of the device passed in rx-dev|tx-dev parameters.
>
> *Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
> to enable packet capturing on a specific port/device_id and queue by passing on
> port|device_id, queue, mempool and ring info.
>
> *Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
>
> *Tool can be stopped using SIGINT, upon which tool calls
> rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
>
> Note:
> CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
>
> 3)Test-pmd changes
> ==================
> Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
> So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
>
> Similarly any application which needs packet capture should call initialize/uninitialize APIs of
> librte_pdump and use pdump tool to start the capture.
>
> 4)Packet capture flow between pdump tool and librte_pdump
> =========================================================
> * Pdump tool (Secondary process) requests packet capture
> for specific port|device_id and queue combinations.
>
> *Library in secondary process context creates client socket and communicates
> the port|device_id, queue, ring and mempool to server.
>
> *Library initializes server in primary process 'test-pmd' context and server serves
> the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
>
> *Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
>
> *Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
> so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
>
> *Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
>
> *Library receives the disable packet capture request, communicate the info to server,
> server will remove the Ethernet rxtx call-backs.
>
> *Packet capture can be seen using tcpdump command
> "tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
>
> 5)Example command line
> ======================
> ./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-
> size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-
> dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
>
> v7:
> fixed lines over 90 characters.
>
> v6:
> removed below deprecation notice patch from patch set.
> http://dpdk.org/dev/patchwork/patch/13372/
>
> v5:
> addressed code review comments for below patches
> http://dpdk.org/dev/patchwork/patch/12955/
> http://dpdk.org/dev/patchwork/patch/12951/
>
> v4:
> added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
> made doc changes as per doc guidelines.
> replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
> removed rxtx-dev parameter from pdump tool command line.
>
> v3:
> app/pdump: Moved cleanup code from signal handler to main.
> divided librte_ether changes into multiple patches.
> example command changed in app/pdump application guide
>
> v2:
> fix compilation issues for 4.8.3
> fix unnecessary #includes
>
>
> Reshma Pattan (8):
> librte_ether: protect add/remove of rxtx callbacks with spinlocks
> librte_ether: add new api rte_eth_add_first_rx_callback
> librte_ether: add new fields to rte_eth_dev_info struct
> librte_ether: make rte_eth_dev_get_port_by_name
> rte_eth_dev_get_name_by_port public
> lib/librte_pdump: add new library for packet capturing support
> app/pdump: add pdump tool for packet capturing
> app/test-pmd: add pdump initialization uninitialization
> doc: update doc for packet capture framework
>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> --
> 2.5.0
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v7 8/8] doc: update doc for packet capture framework
2016-06-09 16:10 2% ` [dpdk-dev] [PATCH v7 0/8] add " Reshma Pattan
@ 2016-06-09 16:10 5% ` Reshma Pattan
2016-06-09 17:34 0% ` [dpdk-dev] [PATCH v7 0/8] add " Ananyev, Konstantin
1 sibling, 0 replies; 200+ results
From: Reshma Pattan @ 2016-06-09 16:10 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
Added programmers guide for librte_pdump.
Added sample application guide for app/pdump application.
Updated release note for packet capture framework changes.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
---
MAINTAINERS | 3 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 107 ++++++++++++++++++++++++++++
doc/guides/rel_notes/release_16_07.rst | 13 ++++
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 ++++++++++++++++++++++++++++++++
6 files changed, 247 insertions(+)
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
diff --git a/MAINTAINERS b/MAINTAINERS
index a48c8de..ce7c941 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -436,6 +436,9 @@ Pdump
M: Reshma Pattan <reshma.pattan@intel.com>
F: lib/librte_pdump/
F: app/pdump/
+F: doc/guides/prog_guide/pdump_library.rst
+F: doc/guides/sample_app_ug/pdump.rst
+
Hierarchical scheduler
M: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
diff --git a/doc/guides/prog_guide/index.rst b/doc/guides/prog_guide/index.rst
index b862d0c..4caf969 100644
--- a/doc/guides/prog_guide/index.rst
+++ b/doc/guides/prog_guide/index.rst
@@ -71,6 +71,7 @@ Programmer's Guide
writing_efficient_code
profile_app
glossary
+ pdump_library
**Figures**
diff --git a/doc/guides/prog_guide/pdump_library.rst b/doc/guides/prog_guide/pdump_library.rst
new file mode 100644
index 0000000..1809234
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,107 @@
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+.. _pdump_library:
+
+The librte_pdump Library
+========================
+
+The ``librte_pdump`` library provides a framework for packet capturing in DPDK.
+The library provides the following APIs to initialize the packet capture framework, to enable
+or disable the packet capture, and to uninitialize it:
+
+* ``rte_pdump_init()``:
+ This API initializes the packet capture framework.
+
+* ``rte_pdump_enable()``:
+ This API enables the packet capture on a given port and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_enable_by_deviceid()``:
+ This API enables the packet capture on a given device id (``vdev name or pci address``) and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_disable()``:
+ This API disables the packet capture on a given port and queue.
+
+* ``rte_pdump_disable_by_deviceid()``:
+ This API disables the packet capture on a given device id (``vdev name or pci address``) and queue.
+
+* ``rte_pdump_uninit()``:
+ This API uninitializes the packet capture framework.
+
+
+Operation
+---------
+
+The ``librte_pdump`` library works on a client/server model. The server is responsible for enabling or
+disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
+the packet capture.
+
+The packet capture framework, as part of its initialization, creates the pthread and the server socket in
+the pthread. The application that calls the framework initialization first will have the server socket created.
+Further calls to the framework initialization by the same application or other applications is not allowed i.e., only
+one server socket is allowed on the system. So the other applications can only request enabling or disabling of
+the packet capture at which point the client socket is created for them to send the request to the server.
+The server socket will listen for client requests for enabling or disabling the packet capture.
+
+
+Implementation Details
+----------------------
+
+The library API ``rte_pdump_init()``, initializes the packet capture framework by creating the pthread and the server
+socket. The server socket in the pthread context will be listening to the client requests to enable or disable the
+packet capture. Whoever calls this API first will have the server socket created, the subsequent calls to this APIs
+will not create any further server socket. i.e. only one server socket is allowed.
+
+The library APIs ``rte_pdump_enable()`` and ``rte_pdump_enable_by_deviceid()`` enables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump enable" request and sends
+the request to the server. The server that is listening on the socket will take the request and enable the packet capture
+by registering the Ethernet RX and TX callbacks for the given port or device_id and queue combinations.
+Then the server will mirror the packets to the new mempool and enqueue them to the rte_ring that clients have passed
+to these APIs. The server also sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, the client socket is closed.
+
+The library APIs ``rte_pdump_disable()`` and ``rte_pdump_disable_by_deviceid()`` disables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump disable" request and sends
+the request to the server. The server that is listening on the socket will take the request and disable the packet
+capture by removing the Ethernet RX and TX callbacks for the given port or device_id and queue combinations. The server
+also sends the response back to the client about the status of the request that was processed. After the response is
+received from the server, the client socket is closed.
+
+The library API ``rte_pdump_uninit()``, uninitializes the packet capture framework by closing the pthread and the
+server socket.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+The DPDK ``app/pdump`` tool is developed based on this library to capture packets in DPDK.
+Users can use this as an example to develop their own packet capturing application.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index c0f6b02..a4de2a2 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -66,6 +66,11 @@ New Features
* Enable RSS per network interface through the configuration file.
* Streamline the CLI code.
+* **Added packet capture framework.**
+
+ * A new library ``librte_pdump`` is added to provide packet capture APIs.
+ * A new ``app/pdump`` tool is added to capture packets in DPDK.
+
Resolved Issues
---------------
@@ -135,6 +140,11 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* Function ``rte_eth_dev_get_port_by_name`` changed to a public API.
+
+* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ in the ``rte_eth_dev_info`` object.
+
ABI Changes
-----------
@@ -146,6 +156,9 @@ ABI Changes
* The ``rte_port_source_params`` structure has new fields to support PCAP file.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ to support number of queues configured by software.
+
Shared Library Versions
-----------------------
diff --git a/doc/guides/sample_app_ug/index.rst b/doc/guides/sample_app_ug/index.rst
index 930f68c..96bb317 100644
--- a/doc/guides/sample_app_ug/index.rst
+++ b/doc/guides/sample_app_ug/index.rst
@@ -76,6 +76,7 @@ Sample Applications User Guide
ptpclient
performance_thread
ipsec_secgw
+ pdump
**Figures**
diff --git a/doc/guides/sample_app_ug/pdump.rst b/doc/guides/sample_app_ug/pdump.rst
new file mode 100644
index 0000000..96c8709
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,122 @@
+
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+dpdk_pdump Application
+======================
+
+The ``dpdk_pdump`` application is a Data Plane Development Kit (DPDK) application that runs as a DPDK secondary process and
+is capable of enabling packet capture on dpdk ports.
+
+
+Running the Application
+-----------------------
+
+The application has a ``--pdump`` command line option with various sub arguments:
+
+.. code-block:: console
+
+ ./build/app/dpdk_pdump --
+ --pdump '(port=<port id> | device_id=<pci id or vdev name>),
+ (queue=<queue_id>),
+ (rx-dev=<iface or pcap file> |
+ tx-dev=<iface or pcap file>),
+ [ring-size=<ring size>],
+ [mbuf-size=<mbuf data size>],
+ [total-num-mbufs=<number of mbufs>]'
+
+Note:
+
+* Parameters inside the parentheses represents mandatory parameters.
+
+* Parameters inside the square brackets represents optional parameters.
+
+Multiple instances of ``--pdump`` can be passed to capture packets on different port and queue combinations.
+
+
+Parameters
+~~~~~~~~~~
+
+``port``:
+Port id of the eth device on which packets should be captured.
+
+``device_id``:
+PCI address (or) name of the eth device on which packets should be captured.
+
+ .. Note::
+
+ * As of now the ``dpdk_pdump`` tool cannot capture the packets of virtual devices
+ in the primary process due to a bug in the ethdev library. Due to this bug, in a multi process context,
+ when the primary and secondary have different ports set, then the secondary process
+ (here the ``dpdk_pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
+
+``queue``:
+Queue id of the eth device on which packets should be captured. The user can pass a queue value of ``*`` to enable
+packet capture on all queues of the eth device.
+
+``rx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+``tx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+ .. Note::
+
+ * To receive ingress packets only, ``rx-dev`` should be passed.
+
+ * To receive egress packets only, ``tx-dev`` should be passed.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the different file names or the Linux iface names.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the same file names or the the Linux iface names.
+
+``ring-size``:
+Size of the ring. This value is used internally for ring creation. The ring will be used to enqueue the packets from
+the primary application to the secondary. This is an optional parameter with default size 16384.
+
+``mbuf-size``:
+Size of the mbuf data. This is used internally for mempool creation. Ideally this value must be same as
+the primary application's mempool's mbuf data size which is used for packet RX. This is an optional parameter with
+default size 2176.
+
+``total-num-mbufs``:
+Total number mbufs in mempool. This is used internally for mempool creation. This is an optional parameter with default
+value 65535.
+
+
+Example
+-------
+
+.. code-block:: console
+
+ $ sudo ./build/app/dpdk_pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
--
2.5.0
^ permalink raw reply [relevance 5%]
* [dpdk-dev] [PATCH v7 0/8] add packet capture framework
2016-06-09 8:50 2% ` [dpdk-dev] [PATCH v6 0/8] add packet capture framework Reshma Pattan
2016-06-09 8:50 5% ` [dpdk-dev] [PATCH v6 8/8] doc: update doc for " Reshma Pattan
@ 2016-06-09 16:10 2% ` Reshma Pattan
2016-06-09 16:10 5% ` [dpdk-dev] [PATCH v7 8/8] doc: update doc for " Reshma Pattan
2016-06-09 17:34 0% ` [dpdk-dev] [PATCH v7 0/8] add " Ananyev, Konstantin
1 sibling, 2 replies; 200+ results
From: Reshma Pattan @ 2016-06-09 16:10 UTC (permalink / raw)
To: dev
This patch set include below changes
1)Changes to librte_ether.
2)A new library librte_pdump added for packet capture framework.
3)A new app/pdump tool added for packet capturing.
4)Test pmd changes done to initialize packet capture framework.
5)Documentation update.
1)librte_pdump
==============
To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
is added.Users can develop their own packet capturing application using new library APIs.
Operation:
----------
Pdump library provides APIs to support packet capturing on dpdk Ethernet devices.
Library provides APIs to initialize the packet capture framework, enable/disable
the packet capture and uninitialize the packet capture framework.
Pdump library works on client/server based model.
Sever is responsible for enabling/disabling the packet captures.
Clients are responsible for requesting enable/disable of the
packet captures.
As part of packet capture framework initialization, pthread and
the server socket is created. Only one server socket is allowed on the system.
As part of enabling/disabling the packet capture, client sockets are created
and multiple client sockets are allowed.
Who ever calls initialization first they will succeed with the initialization,
next subsequent calls of initialization are not allowed. So next users can only
request enabling/disabling the packet capture.
Applications using below APIs need to pass port/device_id, queue, mempool and
ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
packets of the port for users. Users need to dequeue the rings and write the packets
to vdev(pcap/tuntap) to view the packets using any standard tools.
Note:
Mempool and Ring should be mc/mp supportable.
Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
APIs:
-----
rte_pdump_init()
rte_pdump_enable()
rte_pdump_enable_by_deviceid()
rte_pdump_disable()
rte_pdump_disable_by_deviceid()
rte_pdump_uninit()
2)app/pdump tool
================
Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
This tool by default runs as secondary process, and provides the support for
the command line options for packet capture.
./build/app/dpdk_pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue id>),
(rx-dev=<iface or pcap file> |
tx-dev=<iface or pcap file>),
[ring-size=<ring size>],
[mbuf-size=<mbuf data size>],
[total-num-mbufs=<number of mbufs>]'
Parameters inside the parenthesis represents the mandatory parameters.
Parameters inside the square brackets represents optional parameters.
User has to pass on packet capture parameters under --pdump parameters, multiples of
--pdump can be passed to capture packets on different port and queue combinations
Operation:
----------
*Tool parse the user command line arguments,
creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
of the device passed in rx-dev|tx-dev parameters.
*Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
to enable packet capturing on a specific port/device_id and queue by passing on
port|device_id, queue, mempool and ring info.
*Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
*Tool can be stopped using SIGINT, upon which tool calls
rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
Note:
CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
3)Test-pmd changes
==================
Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
Similarly any application which needs packet capture should call initialize/uninitialize APIs of
librte_pdump and use pdump tool to start the capture.
4)Packet capture flow between pdump tool and librte_pdump
=========================================================
* Pdump tool (Secondary process) requests packet capture
for specific port|device_id and queue combinations.
*Library in secondary process context creates client socket and communicates
the port|device_id, queue, ring and mempool to server.
*Library initializes server in primary process 'test-pmd' context and server serves
the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
*Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
*Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
*Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
*Library receives the disable packet capture request, communicate the info to server,
server will remove the Ethernet rxtx call-backs.
*Packet capture can be seen using tcpdump command
"tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
5)Example command line
======================
./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
v7:
fixed lines over 90 characters.
v6:
removed below deprecation notice patch from patch set.
http://dpdk.org/dev/patchwork/patch/13372/
v5:
addressed code review comments for below patches
http://dpdk.org/dev/patchwork/patch/12955/
http://dpdk.org/dev/patchwork/patch/12951/
v4:
added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
made doc changes as per doc guidelines.
replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
removed rxtx-dev parameter from pdump tool command line.
v3:
app/pdump: Moved cleanup code from signal handler to main.
divided librte_ether changes into multiple patches.
example command changed in app/pdump application guide
v2:
fix compilation issues for 4.8.3
fix unnecessary #includes
Reshma Pattan (8):
librte_ether: protect add/remove of rxtx callbacks with spinlocks
librte_ether: add new api rte_eth_add_first_rx_callback
librte_ether: add new fields to rte_eth_dev_info struct
librte_ether: make rte_eth_dev_get_port_by_name
rte_eth_dev_get_name_by_port public
lib/librte_pdump: add new library for packet capturing support
app/pdump: add pdump tool for packet capturing
app/test-pmd: add pdump initialization uninitialization
doc: update doc for packet capture framework
MAINTAINERS | 8 +
app/Makefile | 1 +
app/pdump/Makefile | 45 ++
app/pdump/main.c | 844 +++++++++++++++++++++++++++++++
app/test-pmd/testpmd.c | 6 +
config/common_base | 5 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 107 ++++
doc/guides/rel_notes/release_16_07.rst | 13 +
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 +++++
lib/Makefile | 1 +
lib/librte_ether/rte_ethdev.c | 123 +++--
lib/librte_ether/rte_ethdev.h | 60 +++
lib/librte_ether/rte_ether_version.map | 9 +
lib/librte_pdump/Makefile | 55 ++
lib/librte_pdump/rte_pdump.c | 872 ++++++++++++++++++++++++++++++++
lib/librte_pdump/rte_pdump.h | 186 +++++++
lib/librte_pdump/rte_pdump_version.map | 12 +
mk/rte.app.mk | 1 +
20 files changed, 2428 insertions(+), 44 deletions(-)
create mode 100644 app/pdump/Makefile
create mode 100644 app/pdump/main.c
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
create mode 100644 lib/librte_pdump/Makefile
create mode 100644 lib/librte_pdump/rte_pdump.c
create mode 100644 lib/librte_pdump/rte_pdump.h
create mode 100644 lib/librte_pdump/rte_pdump_version.map
--
2.5.0
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH v2] log: deprecate history dump
2016-06-09 14:09 6% [dpdk-dev] [PATCH] log: deprecate history dump Thomas Monjalon
2016-06-09 14:45 0% ` David Marchand
@ 2016-06-09 15:06 5% ` Thomas Monjalon
2016-06-09 22:10 5% ` [dpdk-dev] [PATCH v3] " Thomas Monjalon
1 sibling, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-06-09 15:06 UTC (permalink / raw)
To: david.marchand; +Cc: dev
The log history uses rte_mempool. In order to remove the mempool
dependency in EAL (and improve the build), this feature is deprecated.
The ABI is kept but the behaviour is now voided because it seems this
function was not used. The history can be read from syslog.
When enabling the log history, a warning is logged.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
v2:
- remove more mempool and log history traces
- add a warning if enabling log history
- move not related mempool includes cleanup in another patch
---
app/test-pmd/cmdline.c | 3 -
app/test/autotest_data.py | 6 --
app/test/autotest_test_funcs.py | 5 --
app/test/commands.c | 4 +-
app/test/test_logs.c | 3 -
doc/guides/prog_guide/mempool_lib.rst | 4 +-
doc/guides/rel_notes/deprecation.rst | 3 +
lib/librte_eal/bsdapp/eal/Makefile | 1 -
lib/librte_eal/bsdapp/eal/eal_debug.c | 6 --
lib/librte_eal/common/eal_common_log.c | 143 ++------------------------------
lib/librte_eal/common/eal_private.h | 3 -
lib/librte_eal/common/include/rte_log.h | 8 ++
lib/librte_eal/linuxapp/eal/Makefile | 1 -
lib/librte_eal/linuxapp/eal/eal_debug.c | 6 --
lib/librte_eal/linuxapp/eal/eal_log.c | 9 +-
lib/librte_mempool/rte_mempool.c | 4 -
16 files changed, 20 insertions(+), 189 deletions(-)
diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 1921612..fd389ac 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -7268,8 +7268,6 @@ static void cmd_dump_parsed(void *parsed_result,
rte_dump_physmem_layout(stdout);
else if (!strcmp(res->dump, "dump_memzone"))
rte_memzone_dump(stdout);
- else if (!strcmp(res->dump, "dump_log_history"))
- rte_log_dump_history(stdout);
else if (!strcmp(res->dump, "dump_struct_sizes"))
dump_struct_sizes();
else if (!strcmp(res->dump, "dump_ring"))
@@ -7284,7 +7282,6 @@ cmdline_parse_token_string_t cmd_dump_dump =
TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
"dump_physmem#"
"dump_memzone#"
- "dump_log_history#"
"dump_struct_sizes#"
"dump_ring#"
"dump_mempool#"
diff --git a/app/test/autotest_data.py b/app/test/autotest_data.py
index 78d2edd..6c87809 100644
--- a/app/test/autotest_data.py
+++ b/app/test/autotest_data.py
@@ -94,12 +94,6 @@ parallel_test_group_list = [
"Report" : None,
},
{
- "Name" : "Dump log history",
- "Command" : "dump_log_history",
- "Func" : dump_autotest,
- "Report" : None,
- },
- {
"Name" : "Dump rings",
"Command" : "dump_ring",
"Func" : dump_autotest,
diff --git a/app/test/autotest_test_funcs.py b/app/test/autotest_test_funcs.py
index b60b941..14cffd0 100644
--- a/app/test/autotest_test_funcs.py
+++ b/app/test/autotest_test_funcs.py
@@ -144,16 +144,11 @@ def logs_autotest(child, test_name):
i = 0
child.sendline(test_name)
- # logs sequence is printed twice because of history dump
log_list = [
"TESTAPP1: error message",
"TESTAPP1: critical message",
"TESTAPP2: critical message",
"TESTAPP1: error message",
- "TESTAPP1: error message",
- "TESTAPP1: critical message",
- "TESTAPP2: critical message",
- "TESTAPP1: error message",
]
for log_msg in log_list:
diff --git a/app/test/commands.c b/app/test/commands.c
index e0af8e4..2df46b0 100644
--- a/app/test/commands.c
+++ b/app/test/commands.c
@@ -150,8 +150,6 @@ static void cmd_dump_parsed(void *parsed_result,
rte_dump_physmem_layout(stdout);
else if (!strcmp(res->dump, "dump_memzone"))
rte_memzone_dump(stdout);
- else if (!strcmp(res->dump, "dump_log_history"))
- rte_log_dump_history(stdout);
else if (!strcmp(res->dump, "dump_struct_sizes"))
dump_struct_sizes();
else if (!strcmp(res->dump, "dump_ring"))
@@ -164,7 +162,7 @@ static void cmd_dump_parsed(void *parsed_result,
cmdline_parse_token_string_t cmd_dump_dump =
TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
- "dump_physmem#dump_memzone#dump_log_history#"
+ "dump_physmem#dump_memzone#"
"dump_struct_sizes#dump_ring#dump_mempool#"
"dump_devargs");
diff --git a/app/test/test_logs.c b/app/test/test_logs.c
index 05aa862..d0a9962 100644
--- a/app/test/test_logs.c
+++ b/app/test/test_logs.c
@@ -83,9 +83,6 @@ test_logs(void)
RTE_LOG(ERR, TESTAPP1, "error message\n");
RTE_LOG(ERR, TESTAPP2, "error message (not displayed)\n");
- /* print again the previous logs */
- rte_log_dump_history(stdout);
-
return 0;
}
diff --git a/doc/guides/prog_guide/mempool_lib.rst b/doc/guides/prog_guide/mempool_lib.rst
index 5fae79a..c3afc2e 100644
--- a/doc/guides/prog_guide/mempool_lib.rst
+++ b/doc/guides/prog_guide/mempool_lib.rst
@@ -38,9 +38,7 @@ In the DPDK, it is identified by name and uses a ring to store free objects.
It provides some other optional services such as a per-core object cache and
an alignment helper to ensure that objects are padded to spread them equally on all DRAM or DDR3 channels.
-This library is used by the
-:ref:`Mbuf Library <Mbuf_Library>` and the
-:ref:`Environment Abstraction Layer <Environment_Abstraction_Layer>` (for logging history).
+This library is used by the :ref:`Mbuf Library <Mbuf_Library>`.
Cookies
-------
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index ad05eba..bda40c1 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -8,6 +8,9 @@ API and ABI deprecation notices are to be posted here.
Deprecation Notices
-------------------
+* The log history is deprecated.
+ It is voided in 16.07 and will be removed in release 16.11.
+
* The ethdev hotplug API is going to be moved to EAL with a notification
mechanism added to crypto and ethdev libraries so that hotplug is now
available to both of them. This API will be stripped of the device arguments
diff --git a/lib/librte_eal/bsdapp/eal/Makefile b/lib/librte_eal/bsdapp/eal/Makefile
index 9054ad6..474651b 100644
--- a/lib/librte_eal/bsdapp/eal/Makefile
+++ b/lib/librte_eal/bsdapp/eal/Makefile
@@ -41,7 +41,6 @@ CFLAGS += -I$(SRCDIR)/include
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common/include
CFLAGS += -I$(RTE_SDK)/lib/librte_ring
-CFLAGS += -I$(RTE_SDK)/lib/librte_mempool
CFLAGS += $(WERROR_FLAGS) -O3
LDLIBS += -lexecinfo
diff --git a/lib/librte_eal/bsdapp/eal/eal_debug.c b/lib/librte_eal/bsdapp/eal/eal_debug.c
index 907fbfa..5fbc17c 100644
--- a/lib/librte_eal/bsdapp/eal/eal_debug.c
+++ b/lib/librte_eal/bsdapp/eal/eal_debug.c
@@ -77,9 +77,6 @@ void __rte_panic(const char *funcname, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
rte_log(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, "PANIC in %s():\n", funcname);
va_start(ap, format);
rte_vlog(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, format, ap);
@@ -98,9 +95,6 @@ rte_exit(int exit_code, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
if (exit_code != 0)
RTE_LOG(CRIT, EAL, "Error - exiting with code: %d\n"
" Cause: ", exit_code);
diff --git a/lib/librte_eal/common/eal_common_log.c b/lib/librte_eal/common/eal_common_log.c
index b5e37bb..0739331 100644
--- a/lib/librte_eal/common/eal_common_log.c
+++ b/lib/librte_eal/common/eal_common_log.c
@@ -38,47 +38,14 @@
#include <sys/types.h>
#include <stdlib.h>
#include <unistd.h>
-#include <inttypes.h>
-#include <errno.h>
-#include <sys/queue.h>
#include <rte_log.h>
-#include <rte_memory.h>
-#include <rte_memzone.h>
-#include <rte_launch.h>
#include <rte_common.h>
-#include <rte_cycles.h>
#include <rte_eal.h>
#include <rte_per_lcore.h>
-#include <rte_lcore.h>
-#include <rte_atomic.h>
-#include <rte_debug.h>
-#include <rte_spinlock.h>
-#include <rte_branch_prediction.h>
-#include <rte_ring.h>
-#include <rte_mempool.h>
#include "eal_private.h"
-#define LOG_ELT_SIZE 2048
-
-#define LOG_HISTORY_MP_NAME "log_history"
-
-STAILQ_HEAD(log_history_list, log_history);
-
-/**
- * The structure of a message log in the log history.
- */
-struct log_history {
- STAILQ_ENTRY(log_history) next;
- unsigned size;
- char buf[0];
-};
-
-static struct rte_mempool *log_history_mp = NULL;
-static unsigned log_history_size = 0;
-static struct log_history_list log_history;
-
/* global log structure */
struct rte_logs rte_logs = {
.type = ~0,
@@ -86,10 +53,7 @@ struct rte_logs rte_logs = {
.file = NULL,
};
-static rte_spinlock_t log_dump_lock = RTE_SPINLOCK_INITIALIZER;
-static rte_spinlock_t log_list_lock = RTE_SPINLOCK_INITIALIZER;
static FILE *default_log_stream;
-static int history_enabled = 1;
/**
* This global structure stores some informations about the message
@@ -106,59 +70,16 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg, log_cur_msg);
/* default logs */
int
-rte_log_add_in_history(const char *buf, size_t size)
+rte_log_add_in_history(const char *buf __rte_unused, size_t size __rte_unused)
{
- struct log_history *hist_buf = NULL;
- static const unsigned hist_buf_size = LOG_ELT_SIZE - sizeof(*hist_buf);
- void *obj;
-
- if (history_enabled == 0)
- return 0;
-
- rte_spinlock_lock(&log_list_lock);
-
- /* get a buffer for adding in history */
- if (log_history_size > RTE_LOG_HISTORY) {
- hist_buf = STAILQ_FIRST(&log_history);
- if (hist_buf) {
- STAILQ_REMOVE_HEAD(&log_history, next);
- log_history_size--;
- }
- }
- else {
- if (rte_mempool_mc_get(log_history_mp, &obj) < 0)
- obj = NULL;
- hist_buf = obj;
- }
-
- /* no buffer */
- if (hist_buf == NULL) {
- rte_spinlock_unlock(&log_list_lock);
- return -ENOBUFS;
- }
-
- /* not enough room for msg, buffer go back in mempool */
- if (size >= hist_buf_size) {
- rte_mempool_mp_put(log_history_mp, hist_buf);
- rte_spinlock_unlock(&log_list_lock);
- return -ENOBUFS;
- }
-
- /* add in history */
- memcpy(hist_buf->buf, buf, size);
- hist_buf->buf[size] = hist_buf->buf[hist_buf_size-1] = '\0';
- hist_buf->size = size;
- STAILQ_INSERT_TAIL(&log_history, hist_buf, next);
- log_history_size++;
- rte_spinlock_unlock(&log_list_lock);
-
return 0;
}
void
rte_log_set_history(int enable)
{
- history_enabled = enable;
+ if (enable)
+ RTE_LOG(WARNING, EAL, "The log history is deprecated.\n");
}
/* Change the stream that will be used by logging system */
@@ -217,44 +138,8 @@ int rte_log_cur_msg_logtype(void)
/* Dump log history to file */
void
-rte_log_dump_history(FILE *out)
+rte_log_dump_history(FILE *out __rte_unused)
{
- struct log_history_list tmp_log_history;
- struct log_history *hist_buf;
- unsigned i;
-
- /* only one dump at a time */
- rte_spinlock_lock(&log_dump_lock);
-
- /* save list, and re-init to allow logging during dump */
- rte_spinlock_lock(&log_list_lock);
- tmp_log_history = log_history;
- STAILQ_INIT(&log_history);
- log_history_size = 0;
- rte_spinlock_unlock(&log_list_lock);
-
- for (i=0; i<RTE_LOG_HISTORY; i++) {
-
- /* remove one message from history list */
- hist_buf = STAILQ_FIRST(&tmp_log_history);
-
- if (hist_buf == NULL)
- break;
-
- STAILQ_REMOVE_HEAD(&tmp_log_history, next);
-
- /* write on stdout */
- if (fwrite(hist_buf->buf, hist_buf->size, 1, out) == 0) {
- rte_mempool_mp_put(log_history_mp, hist_buf);
- break;
- }
-
- /* put back message structure in pool */
- rte_mempool_mp_put(log_history_mp, hist_buf);
- }
- fflush(out);
-
- rte_spinlock_unlock(&log_dump_lock);
}
/*
@@ -297,29 +182,11 @@ rte_log(uint32_t level, uint32_t logtype, const char *format, ...)
}
/*
- * called by environment-specific log init function to initialize log
- * history
+ * called by environment-specific log init function
*/
int
rte_eal_common_log_init(FILE *default_log)
{
- STAILQ_INIT(&log_history);
-
- /* reserve RTE_LOG_HISTORY*2 elements, so we can dump and
- * keep logging during this time */
- log_history_mp = rte_mempool_create(LOG_HISTORY_MP_NAME, RTE_LOG_HISTORY*2,
- LOG_ELT_SIZE, 0, 0,
- NULL, NULL,
- NULL, NULL,
- SOCKET_ID_ANY, MEMPOOL_F_NO_PHYS_CONTIG);
-
- if ((log_history_mp == NULL) &&
- ((log_history_mp = rte_mempool_lookup(LOG_HISTORY_MP_NAME)) == NULL)){
- RTE_LOG(ERR, EAL, "%s(): cannot create log_history mempool\n",
- __func__);
- return -1;
- }
-
default_log_stream = default_log;
rte_openlog_stream(default_log);
diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
index 2342fa1..857dc3e 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -49,9 +49,6 @@ int rte_eal_memzone_init(void);
/**
* Common log initialization function (private to eal).
*
- * Called by environment-specific log initialization function to initialize
- * log history.
- *
* @param default_log
* The default log stream to be used.
* @return
diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
index 2e47e7f..b1add04 100644
--- a/lib/librte_eal/common/include/rte_log.h
+++ b/lib/librte_eal/common/include/rte_log.h
@@ -42,6 +42,8 @@
* This file provides a log API to RTE applications.
*/
+#include "rte_common.h" /* for __rte_deprecated macro */
+
#ifdef __cplusplus
extern "C" {
#endif
@@ -179,22 +181,27 @@ int rte_log_cur_msg_loglevel(void);
int rte_log_cur_msg_logtype(void);
/**
+ * @deprecated
* Enable or disable the history (enabled by default)
*
* @param enable
* true to enable, or 0 to disable history.
*/
+__rte_deprecated
void rte_log_set_history(int enable);
/**
+ * @deprecated
* Dump the log history to a file
*
* @param f
* A pointer to a file for output
*/
+__rte_deprecated
void rte_log_dump_history(FILE *f);
/**
+ * @deprecated
* Add a log message to the history.
*
* This function can be called from a user-defined log stream. It adds
@@ -209,6 +216,7 @@ void rte_log_dump_history(FILE *f);
* - 0: Success.
* - (-ENOBUFS) if there is no room to store the message.
*/
+__rte_deprecated
int rte_log_add_in_history(const char *buf, size_t size);
/**
diff --git a/lib/librte_eal/linuxapp/eal/Makefile b/lib/librte_eal/linuxapp/eal/Makefile
index e109361..a0ea698 100644
--- a/lib/librte_eal/linuxapp/eal/Makefile
+++ b/lib/librte_eal/linuxapp/eal/Makefile
@@ -45,7 +45,6 @@ CFLAGS += -I$(SRCDIR)/include
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common
CFLAGS += -I$(RTE_SDK)/lib/librte_eal/common/include
CFLAGS += -I$(RTE_SDK)/lib/librte_ring
-CFLAGS += -I$(RTE_SDK)/lib/librte_mempool
CFLAGS += -I$(RTE_SDK)/lib/librte_ivshmem
CFLAGS += $(WERROR_FLAGS) -O3
diff --git a/lib/librte_eal/linuxapp/eal/eal_debug.c b/lib/librte_eal/linuxapp/eal/eal_debug.c
index 907fbfa..5fbc17c 100644
--- a/lib/librte_eal/linuxapp/eal/eal_debug.c
+++ b/lib/librte_eal/linuxapp/eal/eal_debug.c
@@ -77,9 +77,6 @@ void __rte_panic(const char *funcname, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
rte_log(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, "PANIC in %s():\n", funcname);
va_start(ap, format);
rte_vlog(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, format, ap);
@@ -98,9 +95,6 @@ rte_exit(int exit_code, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
if (exit_code != 0)
RTE_LOG(CRIT, EAL, "Error - exiting with code: %d\n"
" Cause: ", exit_code);
diff --git a/lib/librte_eal/linuxapp/eal/eal_log.c b/lib/librte_eal/linuxapp/eal/eal_log.c
index 0b133c3..d391100 100644
--- a/lib/librte_eal/linuxapp/eal/eal_log.c
+++ b/lib/librte_eal/linuxapp/eal/eal_log.c
@@ -50,8 +50,7 @@
#include "eal_private.h"
/*
- * default log function, used once mempool (hence log history) is
- * available
+ * default log function
*/
static ssize_t
console_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
@@ -60,9 +59,6 @@ console_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
ssize_t ret;
uint32_t loglevel;
- /* add this log in history */
- rte_log_add_in_history(buf, size);
-
/* write on stdout */
ret = fwrite(buf, 1, size, stdout);
fflush(stdout);
@@ -110,8 +106,7 @@ rte_eal_log_init(const char *id, int facility)
/* early logs */
/*
- * early log function, used during boot when mempool (hence log
- * history) is not available
+ * early log function, used before rte_eal_log_init
*/
static ssize_t
early_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index b54de43..22a5645 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -1003,7 +1003,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
if (free == 0) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE1) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1012,7 +1011,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
hdr->cookie = RTE_MEMPOOL_HEADER_COOKIE2;
} else if (free == 1) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE2) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1022,7 +1020,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
} else if (free == 2) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE1 &&
cookie != RTE_MEMPOOL_HEADER_COOKIE2) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1032,7 +1029,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
tlr = __mempool_get_trailer(obj);
cookie = tlr->cookie;
if (cookie != RTE_MEMPOOL_TRAILER_COOKIE) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
--
2.7.0
^ permalink raw reply [relevance 5%]
* Re: [dpdk-dev] [PATCH] log: deprecate history dump
2016-06-09 14:45 0% ` David Marchand
2016-06-09 15:01 0% ` Thomas Monjalon
@ 2016-06-09 15:01 0% ` Christian Ehrhardt
1 sibling, 0 replies; 200+ results
From: Christian Ehrhardt @ 2016-06-09 15:01 UTC (permalink / raw)
To: David Marchand; +Cc: Thomas Monjalon, dev
Hi,
in I totally like it - thanks Thomas for picking that up.
I just wanted to mention that the Makefile still refers to mempool, but
David beat me in time and Detail a lot.
I'll certainly try to follow and help the bit I can.
Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd
On Thu, Jun 9, 2016 at 4:45 PM, David Marchand <david.marchand@6wind.com>
wrote:
> Thomas,
>
> On Thu, Jun 9, 2016 at 4:09 PM, Thomas Monjalon
> <thomas.monjalon@6wind.com> wrote:
> > The log history uses rte_mempool. In order to remove the mempool
> > dependency in EAL (and improve the build), this feature is deprecated.
> > The ABI is kept but the behaviour is now voided because it seems this
> > function was not used. The history can be read from syslog.
>
> It does look like it is not really used.
> I am for this change unless someone complains.
>
> Comments below.
>
> > Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> > ---
> > app/test-pmd/cmdline.c | 3 -
> > app/test/autotest_test_funcs.py | 5 --
> > app/test/commands.c | 4 +-
> > app/test/test_logs.c | 3 -
> > doc/guides/rel_notes/deprecation.rst | 3 +
> > lib/librte_eal/bsdapp/eal/eal_debug.c | 6 --
> > lib/librte_eal/common/eal_common_log.c | 128
> +--------------------------
> > lib/librte_eal/common/include/rte_log.h | 8 ++
> > lib/librte_eal/linuxapp/eal/eal_debug.c | 6 --
> > lib/librte_eal/linuxapp/eal/eal_interrupts.c | 1 -
> > lib/librte_eal/linuxapp/eal/eal_ivshmem.c | 1 -
> > lib/librte_eal/linuxapp/eal/eal_log.c | 3 -
> > lib/librte_mempool/rte_mempool.c | 4 -
> > 13 files changed, 16 insertions(+), 159 deletions(-)
>
> - You missed autotest_data.py.
>
> $ git grep dump_log_history
> app/test/autotest_data.py: "Command" : "dump_log_history",
>
> - eal Makefile still refers to librte_mempool.
>
> - Since you are looking at this, what keeps us from removing the
> dependency on librte_ring ?
> I would say it was mainly because of mempool.
> Maybe ivshmem ?
>
>
> [snip]
>
> > diff --git a/doc/guides/rel_notes/deprecation.rst
> b/doc/guides/rel_notes/deprecation.rst
> > index ad05eba..f11df93 100644
> > --- a/doc/guides/rel_notes/deprecation.rst
> > +++ b/doc/guides/rel_notes/deprecation.rst
> > @@ -8,6 +8,9 @@ API and ABI deprecation notices are to be posted here.
> > Deprecation Notices
> > -------------------
> >
> > +* The log history is deprecated in release 16.07.
> > + It is voided and will be completely removed in release 16.11.
> > +
>
> It is voided in 16.07 and will be completely removed in release 16.11.
>
>
> > * The ethdev hotplug API is going to be moved to EAL with a notification
> > mechanism added to crypto and ethdev libraries so that hotplug is now
> > available to both of them. This API will be stripped of the device
> arguments
>
> > diff --git a/lib/librte_eal/common/eal_common_log.c
> b/lib/librte_eal/common/eal_common_log.c
> > index b5e37bb..94ecdd2 100644
> > --- a/lib/librte_eal/common/eal_common_log.c
> > +++ b/lib/librte_eal/common/eal_common_log.c
> > @@ -56,29 +56,11 @@
> > #include <rte_spinlock.h>
> > #include <rte_branch_prediction.h>
> > #include <rte_ring.h>
> > -#include <rte_mempool.h>
> >
> > #include "eal_private.h"
> >
> > #define LOG_ELT_SIZE 2048
>
> We don't need LOG_ELT_SIZE.
>
> >
> > -#define LOG_HISTORY_MP_NAME "log_history"
> > -
> > -STAILQ_HEAD(log_history_list, log_history);
> > -
> > -/**
> > - * The structure of a message log in the log history.
> > - */
> > -struct log_history {
> > - STAILQ_ENTRY(log_history) next;
> > - unsigned size;
> > - char buf[0];
> > -};
> > -
> > -static struct rte_mempool *log_history_mp = NULL;
> > -static unsigned log_history_size = 0;
> > -static struct log_history_list log_history;
> > -
> > /* global log structure */
> > struct rte_logs rte_logs = {
> > .type = ~0,
> > @@ -86,10 +68,7 @@ struct rte_logs rte_logs = {
> > .file = NULL,
> > };
> >
> > -static rte_spinlock_t log_dump_lock = RTE_SPINLOCK_INITIALIZER;
> > -static rte_spinlock_t log_list_lock = RTE_SPINLOCK_INITIALIZER;
> > static FILE *default_log_stream;
> > -static int history_enabled = 1;
> >
> > /**
> > * This global structure stores some informations about the message
> > @@ -106,59 +85,14 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg,
> log_cur_msg);
> > /* default logs */
> >
> > int
> > -rte_log_add_in_history(const char *buf, size_t size)
> > +rte_log_add_in_history(const char *buf __rte_unused, size_t size
> __rte_unused)
> > {
> > - struct log_history *hist_buf = NULL;
> > - static const unsigned hist_buf_size = LOG_ELT_SIZE -
> sizeof(*hist_buf);
> > - void *obj;
> > -
> > - if (history_enabled == 0)
> > - return 0;
> > -
> > - rte_spinlock_lock(&log_list_lock);
> > -
> > - /* get a buffer for adding in history */
> > - if (log_history_size > RTE_LOG_HISTORY) {
> > - hist_buf = STAILQ_FIRST(&log_history);
> > - if (hist_buf) {
> > - STAILQ_REMOVE_HEAD(&log_history, next);
> > - log_history_size--;
> > - }
> > - }
> > - else {
> > - if (rte_mempool_mc_get(log_history_mp, &obj) < 0)
> > - obj = NULL;
> > - hist_buf = obj;
> > - }
> > -
> > - /* no buffer */
> > - if (hist_buf == NULL) {
> > - rte_spinlock_unlock(&log_list_lock);
> > - return -ENOBUFS;
> > - }
> > -
> > - /* not enough room for msg, buffer go back in mempool */
> > - if (size >= hist_buf_size) {
> > - rte_mempool_mp_put(log_history_mp, hist_buf);
> > - rte_spinlock_unlock(&log_list_lock);
> > - return -ENOBUFS;
> > - }
> > -
> > - /* add in history */
> > - memcpy(hist_buf->buf, buf, size);
> > - hist_buf->buf[size] = hist_buf->buf[hist_buf_size-1] = '\0';
> > - hist_buf->size = size;
> > - STAILQ_INSERT_TAIL(&log_history, hist_buf, next);
> > - log_history_size++;
> > - rte_spinlock_unlock(&log_list_lock);
> > -
> > return 0;
> > }
> >
> > void
> > -rte_log_set_history(int enable)
> > +rte_log_set_history(int enable __rte_unused)
> > {
> > - history_enabled = enable;
> > }
>
> Maybe a warning here for the people who did not read the deprecation
> notices ?
>
>
> --
> David Marchand
>
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH] log: deprecate history dump
2016-06-09 14:45 0% ` David Marchand
@ 2016-06-09 15:01 0% ` Thomas Monjalon
2016-06-09 15:01 0% ` Christian Ehrhardt
1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-09 15:01 UTC (permalink / raw)
To: David Marchand; +Cc: dev
2016-06-09 16:45, David Marchand:
> On Thu, Jun 9, 2016 at 4:09 PM, Thomas Monjalon
> <thomas.monjalon@6wind.com> wrote:
> > The log history uses rte_mempool. In order to remove the mempool
> > dependency in EAL (and improve the build), this feature is deprecated.
> > The ABI is kept but the behaviour is now voided because it seems this
> > function was not used. The history can be read from syslog.
>
> It does look like it is not really used.
> I am for this change unless someone complains.
>
> Comments below.
All your comments will be addressed in a v2. Thanks
> - Since you are looking at this, what keeps us from removing the
> dependency on librte_ring ?
Please see this first small cleanup:
http://dpdk.org/ml/archives/dev/2016-June/040798.html
> I would say it was mainly because of mempool.
> Maybe ivshmem ?
Yes CONFIG_RTE_LIBRTE_IVSHMEM brings dependencies to rte_ring and rte_ivshmem.
This "feature" also pollute the memory allocator and makes rework harder.
That's why I would be in favor of removing CONFIG_RTE_LIBRTE_IVSHMEM.
Otherwise, as an alternative proposal, the file
lib/librte_eal/linuxapp/eal/eal_ivshmem.c
could be moved outside of EAL. Probably that lib/librte_ivshmem/
would be a good place.
The tricky operation would be to remove ivshmem init from eal:
#ifdef RTE_LIBRTE_IVSHMEM
if (rte_eal_ivshmem_init() < 0)
rte_panic("Cannot init IVSHMEM\n");
#endif
if (rte_eal_memory_init() < 0)
rte_panic("Cannot init memory\n");
/* the directories are locked during eal_hugepage_info_init */
eal_hugedirs_unlock();
if (rte_eal_memzone_init() < 0)
rte_panic("Cannot init memzone\n");
if (rte_eal_tailqs_init() < 0)
rte_panic("Cannot init tail queues for objects\n");
#ifdef RTE_LIBRTE_IVSHMEM
if (rte_eal_ivshmem_obj_init() < 0)
rte_panic("Cannot init IVSHMEM objects\n");
#endif
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH] log: deprecate history dump
2016-06-09 14:09 6% [dpdk-dev] [PATCH] log: deprecate history dump Thomas Monjalon
@ 2016-06-09 14:45 0% ` David Marchand
2016-06-09 15:01 0% ` Thomas Monjalon
2016-06-09 15:01 0% ` Christian Ehrhardt
2016-06-09 15:06 5% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
1 sibling, 2 replies; 200+ results
From: David Marchand @ 2016-06-09 14:45 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev
Thomas,
On Thu, Jun 9, 2016 at 4:09 PM, Thomas Monjalon
<thomas.monjalon@6wind.com> wrote:
> The log history uses rte_mempool. In order to remove the mempool
> dependency in EAL (and improve the build), this feature is deprecated.
> The ABI is kept but the behaviour is now voided because it seems this
> function was not used. The history can be read from syslog.
It does look like it is not really used.
I am for this change unless someone complains.
Comments below.
> Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> ---
> app/test-pmd/cmdline.c | 3 -
> app/test/autotest_test_funcs.py | 5 --
> app/test/commands.c | 4 +-
> app/test/test_logs.c | 3 -
> doc/guides/rel_notes/deprecation.rst | 3 +
> lib/librte_eal/bsdapp/eal/eal_debug.c | 6 --
> lib/librte_eal/common/eal_common_log.c | 128 +--------------------------
> lib/librte_eal/common/include/rte_log.h | 8 ++
> lib/librte_eal/linuxapp/eal/eal_debug.c | 6 --
> lib/librte_eal/linuxapp/eal/eal_interrupts.c | 1 -
> lib/librte_eal/linuxapp/eal/eal_ivshmem.c | 1 -
> lib/librte_eal/linuxapp/eal/eal_log.c | 3 -
> lib/librte_mempool/rte_mempool.c | 4 -
> 13 files changed, 16 insertions(+), 159 deletions(-)
- You missed autotest_data.py.
$ git grep dump_log_history
app/test/autotest_data.py: "Command" : "dump_log_history",
- eal Makefile still refers to librte_mempool.
- Since you are looking at this, what keeps us from removing the
dependency on librte_ring ?
I would say it was mainly because of mempool.
Maybe ivshmem ?
[snip]
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index ad05eba..f11df93 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -8,6 +8,9 @@ API and ABI deprecation notices are to be posted here.
> Deprecation Notices
> -------------------
>
> +* The log history is deprecated in release 16.07.
> + It is voided and will be completely removed in release 16.11.
> +
It is voided in 16.07 and will be completely removed in release 16.11.
> * The ethdev hotplug API is going to be moved to EAL with a notification
> mechanism added to crypto and ethdev libraries so that hotplug is now
> available to both of them. This API will be stripped of the device arguments
> diff --git a/lib/librte_eal/common/eal_common_log.c b/lib/librte_eal/common/eal_common_log.c
> index b5e37bb..94ecdd2 100644
> --- a/lib/librte_eal/common/eal_common_log.c
> +++ b/lib/librte_eal/common/eal_common_log.c
> @@ -56,29 +56,11 @@
> #include <rte_spinlock.h>
> #include <rte_branch_prediction.h>
> #include <rte_ring.h>
> -#include <rte_mempool.h>
>
> #include "eal_private.h"
>
> #define LOG_ELT_SIZE 2048
We don't need LOG_ELT_SIZE.
>
> -#define LOG_HISTORY_MP_NAME "log_history"
> -
> -STAILQ_HEAD(log_history_list, log_history);
> -
> -/**
> - * The structure of a message log in the log history.
> - */
> -struct log_history {
> - STAILQ_ENTRY(log_history) next;
> - unsigned size;
> - char buf[0];
> -};
> -
> -static struct rte_mempool *log_history_mp = NULL;
> -static unsigned log_history_size = 0;
> -static struct log_history_list log_history;
> -
> /* global log structure */
> struct rte_logs rte_logs = {
> .type = ~0,
> @@ -86,10 +68,7 @@ struct rte_logs rte_logs = {
> .file = NULL,
> };
>
> -static rte_spinlock_t log_dump_lock = RTE_SPINLOCK_INITIALIZER;
> -static rte_spinlock_t log_list_lock = RTE_SPINLOCK_INITIALIZER;
> static FILE *default_log_stream;
> -static int history_enabled = 1;
>
> /**
> * This global structure stores some informations about the message
> @@ -106,59 +85,14 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg, log_cur_msg);
> /* default logs */
>
> int
> -rte_log_add_in_history(const char *buf, size_t size)
> +rte_log_add_in_history(const char *buf __rte_unused, size_t size __rte_unused)
> {
> - struct log_history *hist_buf = NULL;
> - static const unsigned hist_buf_size = LOG_ELT_SIZE - sizeof(*hist_buf);
> - void *obj;
> -
> - if (history_enabled == 0)
> - return 0;
> -
> - rte_spinlock_lock(&log_list_lock);
> -
> - /* get a buffer for adding in history */
> - if (log_history_size > RTE_LOG_HISTORY) {
> - hist_buf = STAILQ_FIRST(&log_history);
> - if (hist_buf) {
> - STAILQ_REMOVE_HEAD(&log_history, next);
> - log_history_size--;
> - }
> - }
> - else {
> - if (rte_mempool_mc_get(log_history_mp, &obj) < 0)
> - obj = NULL;
> - hist_buf = obj;
> - }
> -
> - /* no buffer */
> - if (hist_buf == NULL) {
> - rte_spinlock_unlock(&log_list_lock);
> - return -ENOBUFS;
> - }
> -
> - /* not enough room for msg, buffer go back in mempool */
> - if (size >= hist_buf_size) {
> - rte_mempool_mp_put(log_history_mp, hist_buf);
> - rte_spinlock_unlock(&log_list_lock);
> - return -ENOBUFS;
> - }
> -
> - /* add in history */
> - memcpy(hist_buf->buf, buf, size);
> - hist_buf->buf[size] = hist_buf->buf[hist_buf_size-1] = '\0';
> - hist_buf->size = size;
> - STAILQ_INSERT_TAIL(&log_history, hist_buf, next);
> - log_history_size++;
> - rte_spinlock_unlock(&log_list_lock);
> -
> return 0;
> }
>
> void
> -rte_log_set_history(int enable)
> +rte_log_set_history(int enable __rte_unused)
> {
> - history_enabled = enable;
> }
Maybe a warning here for the people who did not read the deprecation notices ?
--
David Marchand
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH] log: deprecate history dump
@ 2016-06-09 14:09 6% Thomas Monjalon
2016-06-09 14:45 0% ` David Marchand
2016-06-09 15:06 5% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
0 siblings, 2 replies; 200+ results
From: Thomas Monjalon @ 2016-06-09 14:09 UTC (permalink / raw)
To: david.marchand; +Cc: dev
The log history uses rte_mempool. In order to remove the mempool
dependency in EAL (and improve the build), this feature is deprecated.
The ABI is kept but the behaviour is now voided because it seems this
function was not used. The history can be read from syslog.
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
app/test-pmd/cmdline.c | 3 -
app/test/autotest_test_funcs.py | 5 --
app/test/commands.c | 4 +-
app/test/test_logs.c | 3 -
doc/guides/rel_notes/deprecation.rst | 3 +
lib/librte_eal/bsdapp/eal/eal_debug.c | 6 --
lib/librte_eal/common/eal_common_log.c | 128 +--------------------------
lib/librte_eal/common/include/rte_log.h | 8 ++
lib/librte_eal/linuxapp/eal/eal_debug.c | 6 --
lib/librte_eal/linuxapp/eal/eal_interrupts.c | 1 -
lib/librte_eal/linuxapp/eal/eal_ivshmem.c | 1 -
lib/librte_eal/linuxapp/eal/eal_log.c | 3 -
lib/librte_mempool/rte_mempool.c | 4 -
13 files changed, 16 insertions(+), 159 deletions(-)
diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 1921612..fd389ac 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -7268,8 +7268,6 @@ static void cmd_dump_parsed(void *parsed_result,
rte_dump_physmem_layout(stdout);
else if (!strcmp(res->dump, "dump_memzone"))
rte_memzone_dump(stdout);
- else if (!strcmp(res->dump, "dump_log_history"))
- rte_log_dump_history(stdout);
else if (!strcmp(res->dump, "dump_struct_sizes"))
dump_struct_sizes();
else if (!strcmp(res->dump, "dump_ring"))
@@ -7284,7 +7282,6 @@ cmdline_parse_token_string_t cmd_dump_dump =
TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
"dump_physmem#"
"dump_memzone#"
- "dump_log_history#"
"dump_struct_sizes#"
"dump_ring#"
"dump_mempool#"
diff --git a/app/test/autotest_test_funcs.py b/app/test/autotest_test_funcs.py
index b60b941..14cffd0 100644
--- a/app/test/autotest_test_funcs.py
+++ b/app/test/autotest_test_funcs.py
@@ -144,16 +144,11 @@ def logs_autotest(child, test_name):
i = 0
child.sendline(test_name)
- # logs sequence is printed twice because of history dump
log_list = [
"TESTAPP1: error message",
"TESTAPP1: critical message",
"TESTAPP2: critical message",
"TESTAPP1: error message",
- "TESTAPP1: error message",
- "TESTAPP1: critical message",
- "TESTAPP2: critical message",
- "TESTAPP1: error message",
]
for log_msg in log_list:
diff --git a/app/test/commands.c b/app/test/commands.c
index e0af8e4..2df46b0 100644
--- a/app/test/commands.c
+++ b/app/test/commands.c
@@ -150,8 +150,6 @@ static void cmd_dump_parsed(void *parsed_result,
rte_dump_physmem_layout(stdout);
else if (!strcmp(res->dump, "dump_memzone"))
rte_memzone_dump(stdout);
- else if (!strcmp(res->dump, "dump_log_history"))
- rte_log_dump_history(stdout);
else if (!strcmp(res->dump, "dump_struct_sizes"))
dump_struct_sizes();
else if (!strcmp(res->dump, "dump_ring"))
@@ -164,7 +162,7 @@ static void cmd_dump_parsed(void *parsed_result,
cmdline_parse_token_string_t cmd_dump_dump =
TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
- "dump_physmem#dump_memzone#dump_log_history#"
+ "dump_physmem#dump_memzone#"
"dump_struct_sizes#dump_ring#dump_mempool#"
"dump_devargs");
diff --git a/app/test/test_logs.c b/app/test/test_logs.c
index 05aa862..d0a9962 100644
--- a/app/test/test_logs.c
+++ b/app/test/test_logs.c
@@ -83,9 +83,6 @@ test_logs(void)
RTE_LOG(ERR, TESTAPP1, "error message\n");
RTE_LOG(ERR, TESTAPP2, "error message (not displayed)\n");
- /* print again the previous logs */
- rte_log_dump_history(stdout);
-
return 0;
}
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index ad05eba..f11df93 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -8,6 +8,9 @@ API and ABI deprecation notices are to be posted here.
Deprecation Notices
-------------------
+* The log history is deprecated in release 16.07.
+ It is voided and will be completely removed in release 16.11.
+
* The ethdev hotplug API is going to be moved to EAL with a notification
mechanism added to crypto and ethdev libraries so that hotplug is now
available to both of them. This API will be stripped of the device arguments
diff --git a/lib/librte_eal/bsdapp/eal/eal_debug.c b/lib/librte_eal/bsdapp/eal/eal_debug.c
index 907fbfa..5fbc17c 100644
--- a/lib/librte_eal/bsdapp/eal/eal_debug.c
+++ b/lib/librte_eal/bsdapp/eal/eal_debug.c
@@ -77,9 +77,6 @@ void __rte_panic(const char *funcname, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
rte_log(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, "PANIC in %s():\n", funcname);
va_start(ap, format);
rte_vlog(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, format, ap);
@@ -98,9 +95,6 @@ rte_exit(int exit_code, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
if (exit_code != 0)
RTE_LOG(CRIT, EAL, "Error - exiting with code: %d\n"
" Cause: ", exit_code);
diff --git a/lib/librte_eal/common/eal_common_log.c b/lib/librte_eal/common/eal_common_log.c
index b5e37bb..94ecdd2 100644
--- a/lib/librte_eal/common/eal_common_log.c
+++ b/lib/librte_eal/common/eal_common_log.c
@@ -56,29 +56,11 @@
#include <rte_spinlock.h>
#include <rte_branch_prediction.h>
#include <rte_ring.h>
-#include <rte_mempool.h>
#include "eal_private.h"
#define LOG_ELT_SIZE 2048
-#define LOG_HISTORY_MP_NAME "log_history"
-
-STAILQ_HEAD(log_history_list, log_history);
-
-/**
- * The structure of a message log in the log history.
- */
-struct log_history {
- STAILQ_ENTRY(log_history) next;
- unsigned size;
- char buf[0];
-};
-
-static struct rte_mempool *log_history_mp = NULL;
-static unsigned log_history_size = 0;
-static struct log_history_list log_history;
-
/* global log structure */
struct rte_logs rte_logs = {
.type = ~0,
@@ -86,10 +68,7 @@ struct rte_logs rte_logs = {
.file = NULL,
};
-static rte_spinlock_t log_dump_lock = RTE_SPINLOCK_INITIALIZER;
-static rte_spinlock_t log_list_lock = RTE_SPINLOCK_INITIALIZER;
static FILE *default_log_stream;
-static int history_enabled = 1;
/**
* This global structure stores some informations about the message
@@ -106,59 +85,14 @@ static RTE_DEFINE_PER_LCORE(struct log_cur_msg, log_cur_msg);
/* default logs */
int
-rte_log_add_in_history(const char *buf, size_t size)
+rte_log_add_in_history(const char *buf __rte_unused, size_t size __rte_unused)
{
- struct log_history *hist_buf = NULL;
- static const unsigned hist_buf_size = LOG_ELT_SIZE - sizeof(*hist_buf);
- void *obj;
-
- if (history_enabled == 0)
- return 0;
-
- rte_spinlock_lock(&log_list_lock);
-
- /* get a buffer for adding in history */
- if (log_history_size > RTE_LOG_HISTORY) {
- hist_buf = STAILQ_FIRST(&log_history);
- if (hist_buf) {
- STAILQ_REMOVE_HEAD(&log_history, next);
- log_history_size--;
- }
- }
- else {
- if (rte_mempool_mc_get(log_history_mp, &obj) < 0)
- obj = NULL;
- hist_buf = obj;
- }
-
- /* no buffer */
- if (hist_buf == NULL) {
- rte_spinlock_unlock(&log_list_lock);
- return -ENOBUFS;
- }
-
- /* not enough room for msg, buffer go back in mempool */
- if (size >= hist_buf_size) {
- rte_mempool_mp_put(log_history_mp, hist_buf);
- rte_spinlock_unlock(&log_list_lock);
- return -ENOBUFS;
- }
-
- /* add in history */
- memcpy(hist_buf->buf, buf, size);
- hist_buf->buf[size] = hist_buf->buf[hist_buf_size-1] = '\0';
- hist_buf->size = size;
- STAILQ_INSERT_TAIL(&log_history, hist_buf, next);
- log_history_size++;
- rte_spinlock_unlock(&log_list_lock);
-
return 0;
}
void
-rte_log_set_history(int enable)
+rte_log_set_history(int enable __rte_unused)
{
- history_enabled = enable;
}
/* Change the stream that will be used by logging system */
@@ -217,44 +151,8 @@ int rte_log_cur_msg_logtype(void)
/* Dump log history to file */
void
-rte_log_dump_history(FILE *out)
+rte_log_dump_history(FILE *out __rte_unused)
{
- struct log_history_list tmp_log_history;
- struct log_history *hist_buf;
- unsigned i;
-
- /* only one dump at a time */
- rte_spinlock_lock(&log_dump_lock);
-
- /* save list, and re-init to allow logging during dump */
- rte_spinlock_lock(&log_list_lock);
- tmp_log_history = log_history;
- STAILQ_INIT(&log_history);
- log_history_size = 0;
- rte_spinlock_unlock(&log_list_lock);
-
- for (i=0; i<RTE_LOG_HISTORY; i++) {
-
- /* remove one message from history list */
- hist_buf = STAILQ_FIRST(&tmp_log_history);
-
- if (hist_buf == NULL)
- break;
-
- STAILQ_REMOVE_HEAD(&tmp_log_history, next);
-
- /* write on stdout */
- if (fwrite(hist_buf->buf, hist_buf->size, 1, out) == 0) {
- rte_mempool_mp_put(log_history_mp, hist_buf);
- break;
- }
-
- /* put back message structure in pool */
- rte_mempool_mp_put(log_history_mp, hist_buf);
- }
- fflush(out);
-
- rte_spinlock_unlock(&log_dump_lock);
}
/*
@@ -297,29 +195,11 @@ rte_log(uint32_t level, uint32_t logtype, const char *format, ...)
}
/*
- * called by environment-specific log init function to initialize log
- * history
+ * called by environment-specific log init function
*/
int
rte_eal_common_log_init(FILE *default_log)
{
- STAILQ_INIT(&log_history);
-
- /* reserve RTE_LOG_HISTORY*2 elements, so we can dump and
- * keep logging during this time */
- log_history_mp = rte_mempool_create(LOG_HISTORY_MP_NAME, RTE_LOG_HISTORY*2,
- LOG_ELT_SIZE, 0, 0,
- NULL, NULL,
- NULL, NULL,
- SOCKET_ID_ANY, MEMPOOL_F_NO_PHYS_CONTIG);
-
- if ((log_history_mp == NULL) &&
- ((log_history_mp = rte_mempool_lookup(LOG_HISTORY_MP_NAME)) == NULL)){
- RTE_LOG(ERR, EAL, "%s(): cannot create log_history mempool\n",
- __func__);
- return -1;
- }
-
default_log_stream = default_log;
rte_openlog_stream(default_log);
diff --git a/lib/librte_eal/common/include/rte_log.h b/lib/librte_eal/common/include/rte_log.h
index 2e47e7f..b1add04 100644
--- a/lib/librte_eal/common/include/rte_log.h
+++ b/lib/librte_eal/common/include/rte_log.h
@@ -42,6 +42,8 @@
* This file provides a log API to RTE applications.
*/
+#include "rte_common.h" /* for __rte_deprecated macro */
+
#ifdef __cplusplus
extern "C" {
#endif
@@ -179,22 +181,27 @@ int rte_log_cur_msg_loglevel(void);
int rte_log_cur_msg_logtype(void);
/**
+ * @deprecated
* Enable or disable the history (enabled by default)
*
* @param enable
* true to enable, or 0 to disable history.
*/
+__rte_deprecated
void rte_log_set_history(int enable);
/**
+ * @deprecated
* Dump the log history to a file
*
* @param f
* A pointer to a file for output
*/
+__rte_deprecated
void rte_log_dump_history(FILE *f);
/**
+ * @deprecated
* Add a log message to the history.
*
* This function can be called from a user-defined log stream. It adds
@@ -209,6 +216,7 @@ void rte_log_dump_history(FILE *f);
* - 0: Success.
* - (-ENOBUFS) if there is no room to store the message.
*/
+__rte_deprecated
int rte_log_add_in_history(const char *buf, size_t size);
/**
diff --git a/lib/librte_eal/linuxapp/eal/eal_debug.c b/lib/librte_eal/linuxapp/eal/eal_debug.c
index 907fbfa..5fbc17c 100644
--- a/lib/librte_eal/linuxapp/eal/eal_debug.c
+++ b/lib/librte_eal/linuxapp/eal/eal_debug.c
@@ -77,9 +77,6 @@ void __rte_panic(const char *funcname, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
rte_log(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, "PANIC in %s():\n", funcname);
va_start(ap, format);
rte_vlog(RTE_LOG_CRIT, RTE_LOGTYPE_EAL, format, ap);
@@ -98,9 +95,6 @@ rte_exit(int exit_code, const char *format, ...)
{
va_list ap;
- /* disable history */
- rte_log_set_history(0);
-
if (exit_code != 0)
RTE_LOG(CRIT, EAL, "Error - exiting with code: %d\n"
" Cause: ", exit_code);
diff --git a/lib/librte_eal/linuxapp/eal/eal_interrupts.c b/lib/librte_eal/linuxapp/eal/eal_interrupts.c
index 06b26a9..0bee8aa 100644
--- a/lib/librte_eal/linuxapp/eal/eal_interrupts.c
+++ b/lib/librte_eal/linuxapp/eal/eal_interrupts.c
@@ -60,7 +60,6 @@
#include <rte_ring.h>
#include <rte_debug.h>
#include <rte_log.h>
-#include <rte_mempool.h>
#include <rte_pci.h>
#include <rte_malloc.h>
#include <rte_errno.h>
diff --git a/lib/librte_eal/linuxapp/eal/eal_ivshmem.c b/lib/librte_eal/linuxapp/eal/eal_ivshmem.c
index eea0314..67b3caf 100644
--- a/lib/librte_eal/linuxapp/eal/eal_ivshmem.c
+++ b/lib/librte_eal/linuxapp/eal/eal_ivshmem.c
@@ -49,7 +49,6 @@
#include <rte_string_fns.h>
#include <rte_errno.h>
#include <rte_ring.h>
-#include <rte_mempool.h>
#include <rte_malloc.h>
#include <rte_common.h>
#include <rte_ivshmem.h>
diff --git a/lib/librte_eal/linuxapp/eal/eal_log.c b/lib/librte_eal/linuxapp/eal/eal_log.c
index 0b133c3..8464152 100644
--- a/lib/librte_eal/linuxapp/eal/eal_log.c
+++ b/lib/librte_eal/linuxapp/eal/eal_log.c
@@ -60,9 +60,6 @@ console_log_write(__attribute__((unused)) void *c, const char *buf, size_t size)
ssize_t ret;
uint32_t loglevel;
- /* add this log in history */
- rte_log_add_in_history(buf, size);
-
/* write on stdout */
ret = fwrite(buf, 1, size, stdout);
fflush(stdout);
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index b54de43..22a5645 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -1003,7 +1003,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
if (free == 0) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE1) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1012,7 +1011,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
hdr->cookie = RTE_MEMPOOL_HEADER_COOKIE2;
} else if (free == 1) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE2) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1022,7 +1020,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
} else if (free == 2) {
if (cookie != RTE_MEMPOOL_HEADER_COOKIE1 &&
cookie != RTE_MEMPOOL_HEADER_COOKIE2) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
@@ -1032,7 +1029,6 @@ void rte_mempool_check_cookies(const struct rte_mempool *mp,
tlr = __mempool_get_trailer(obj);
cookie = tlr->cookie;
if (cookie != RTE_MEMPOOL_TRAILER_COOKIE) {
- rte_log_set_history(0);
RTE_LOG(CRIT, MEMPOOL,
"obj=%p, mempool=%p, cookie=%" PRIx64 "\n",
obj, (const void *) mp, cookie);
--
2.7.0
^ permalink raw reply [relevance 6%]
* Re: [dpdk-dev] [PATCH] mbuf: remove inconsistent assert statements
2016-06-09 7:46 0% ` Olivier Matz
@ 2016-06-09 13:21 0% ` Ananyev, Konstantin
0 siblings, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-06-09 13:21 UTC (permalink / raw)
To: Olivier Matz, dev, Adrien Mazarguil; +Cc: Zhang, Helin
Hi Olivier,
> -----Original Message-----
> From: Olivier Matz [mailto:olivier.matz@6wind.com]
> Sent: Thursday, June 09, 2016 8:47 AM
> To: Ananyev, Konstantin; dev@dpdk.org; Adrien Mazarguil
> Subject: Re: [dpdk-dev] [PATCH] mbuf: remove inconsistent assert statements
>
> Hi Konstantin,
>
> >>>>>> Yes, it refcnt supposed to be set to 0 by __rte_pktmbuf_prefree_seg().
> >>>>>> Wright now, it is a user responsibility to make sure refcnt==0 before pushing
> >>>>>> mbuf back to the pool.
> >>>>>> Not sure why do you consider that wrong?
> >>>>>
> >>>>> I do not consider this wrong and I'm all for using assert() to catch
> >>>>> programming errors, however in this specific case, I think they are
> >>>>> inconsistent and misleading.
> >>>>
> >>>> Honestly, I don't understand why.
> >>>> Right now the rule of thumb is - when mbuf is in the pool, it's refcnt should be equal zero.
> >>
> >> What is the purpose of this? Is there some code relying on this?
> >
> > The whole current implementation of mbuf_free code path relies on that.
> > Straight here:
> > if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
> > m->next = NULL;
> > __rte_mbuf_raw_free(m);
> > }
> >
> > If we'll exclude indirect mbuf logic, all it does:
> > if (rte_mbuf_refcnt_update(m, -1) == 0) {
> > m->next = NULL;
> > __rte_mbuf_raw_free(m);
> > }
> >
> > I.E.:
> > decrement mbuf->refcnt.
> > If new value of refcnt is zero, then put it back into the pool.
> >
> > So having ASERT(mbuf->refcnt==0) inside
> > __rte_mbuf_raw_free()/__rte_mbuf_raw_alloc()
> > looks absolutely valid to me.
> > I *has* to be zero at that point with current implementation,
> > And if it is not then we probably have (or will have) a silent memory corruption.
>
> This explains how the refcount is used, and why it is set
> to zero before returning the mbuf to the pool with the mbuf
> free functions.
>From my point, that shows that rte_pktmbuf_free() relies on the value of the refcnt
to make a decision is it ok to put mbuf back to the pool or not.
Right now it puts mbuf to the pool *only* if it's refcnt==0.
As discussed below, we probably can change it to be refcnt==1
(if there really would be noticeable performance gain).
But I think it still should be just one predefined value of refcnt (0 or 1).
In theory it is possible to allow both (0 and 1),
but that will make it hard to debug any alloc/free issues,
plus would neglect any possible performance gain -
as in that case raw_alloc (or it's analog) should still do
mbuf->refcnt=1;
>
> It does not explain which code relies on the refcnt beeing 0
> while the mbuf is in the pool.
>
>
> >> But since [1], it is allowed to call rte_mbuf_raw_alloc() in PMDs (all
> >> PMDs were calling an internal function before).
> >
> > Yes, raw_alloc is public, NP with that.
> >
> >> We could argue that
> >> rte_mbuf_raw_free() should also be made public for PMDs.
> >
> > We could, but right now it is not.
> > Again, as I said, user could use it on his own but he obviously has to
> > obey the rules and do manually what __rte_pktmbuf_prefree_seg() does.
> > At least:
> >
> > rte_mbuf_refcnt_set(m, 0);
> > __rte_mbuf_raw_free(m);
> >
> >>
> >> As you said below, no-one promised that the free() reverts the malloc(),
> >> but given the function names, one can legitimately expect that the
> >> following code is valid:
> >>
> >> m = __rte_mbuf_raw_alloc();
> >> /* do nothing here */
> >> __rte_mbuf_raw_free(m);
> >
> > Surely people could (and would) expect various things...
> > But the reality right now is: __rte_mbuf_raw_free() is an internal
> > function and not counterpart of __rte_mbuf_raw_alloc().
> > If the people don't bother to read API docs or/and the actual code,
> > I can't see how we can help them :)
>
> Yes, of course, people should read the doc.
> This does not prevent to have a nice API that behaves in a
> natural way :)
>
> By the way, the fact that today the mbuf refcnt should be 0 while
> in a pool is not in the api doc, but in the code.
Ok, I admit there is a bug in the docs, let's add this line to the PG and fix it :)
>
> >> If no code relies on having the refcnt set to 0 when a mbuf is in
> >> the pool, I suggest to relax this constraint as Adrien proposed.
> >
> > Why not just rename it to __rte_mbuf_raw_free_dont_use_after_raw_alloc()?
> > To avoid any further confusion :)
> > Seriously speaking I would prefer to leave it as it is.
> > If you feel we have to introduce a counterpart of rte_mbuf_raw_alloc(),
> > we can make a new public one:
> >
> > rte_mbuf_raw_free(stuct rte_mbuf *m)
> > {
> > if (rte_mbuf_refcnt_update(m, -1) == 0)
> > __rte_mbuf_raw_free(m);
> > }
>
> This is an option, but I think it's not efficient to touch
> the mbuf structure when allocating/freeing. See below.
I don't think it is totally avoidable for generic case anyway.
>
> >> Then, my opinion is that the refcount should be set to 1 in
> >> rte_pktmbuf_reset().
> >
> > I don't think we need to update rte_pktmbuf_reset(),
> > it doesn't touch refcnt at all and probably better to keep it that way.
>
> Why would it be better?
First because pktmbuf_reset() is not very efficient one, and many PMDs
avoid to use it.
Second I think managing refcnt is more part of alloc/free then init procedure.
> All mbuf struct initializations are done in that function, why would
> it be different for the refcnt?
>
> > To achieve what you suggesting, we probably need to:
> > 1) update _rte_pktmbuf_prefree_seg and rte_pktmbuf_detach() to
> > set refcnt back to 1, i.e:
> >
> > static inline struct rte_mbuf* __attribute__((always_inline))
> > __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
> > {
> > __rte_mbuf_sanity_check(m, 0);
> >
> > if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
> > /* if this is an indirect mbuf, it is detached. */
> > if (RTE_MBUF_INDIRECT(m))
> > rte_pktmbuf_detach(m);
> > + rte_mbuf_refcnt_set(m, 1);
> > return m;
> > }
> > return NULL;
> > }
> >
> > 2) either:
> > a) update mbuf constructor function, so it sets refcnt=1.
> > I suppose that is easy for rte_pktmbuf_init(), but it means that all custom
> > constructors should do the same.
> > Which means possible changes in existing user code and all ABI change related hassle.
> > b) keep rte_mbuf_raw_alloc() setting mbuf->refcnt=1.
> > But then I don't see how will get any performance gain here.
> >
> > So not sure is it really worth it.
> >
> >> And rte_pktmbuf_free() should not be allowed on
> >> an uninitialized mbuf (yes, this would require some changes in PMDs).
> >
> > Not sure I understand you here...
> > free() wouldn not be allowed on mbuf whose recnf==0?
> > But it is not allowed right now anyway.
>
> Today:
>
> /* allowed */
> m = rte_pktmbuf_alloc();
> rte_pktmbuf_free(m);
>
> /* not allowed */
> m = rte_mbuf_raw_alloc();
> __rte_mbuf_raw_free(m);
>
> /* we should do instead (strange): */
> m = rte_mbuf_raw_alloc();
> rte_pktmbuf_free(m);
>
> What I suggest to have:
>
> /* allowed, no change */
> m = rte_pktmbuf_alloc();
> rte_pktmbuf_free(m);
>
> /* allowed, these functions would be symetrical */
> m = rte_mbuf_raw_alloc();
> rte_mbuf_raw_free(m);
>
> /* not allowed, m->refcnt is uninitialized */
> m = rte_mbuf_raw_alloc();
> rte_pktmbuf_free(m);
Hmm, and what it will buy us (except of symmetry)?
>
>
>
> >
> >> This would open the door for bulk allocation/free in the mbuf api.
> >
> > Hmm and what stops us having one right now?
> > As I know we do have rte_pktmbuf_alloc_bulk(), I don't see
> > why we can't have rte_pktmbuf_free_bulk() right now.
>
> I think, not touching the refcnt on raw allocation allows a PMD to
> do the following:
>
> rte_mbuf_allow_raw_bulk(pool, &mbufs, n)
> for (i = 0; i < n; i++) {
> prefetch(m[i+1]);
> rte_pktmbuf_reset(); /* including refcnt = 1 */
> do_stuff(m);
> }
You can do that part right now.
That what many PMDs do, just function names are a bit different:
rte_mempool_get_bulk(pool, &mbufs, n);
for (i = 0; i < n; i++) {
custom_rte_pktmbuf_fill(); /* including refcnt = 1 */
do_stuff(m);
}
> /* if all mbufs are not used, free the remaining */
> rte_mbuf_free_raw_bulk(&mbufs[i], n-i);
Ok and what be a performance critical use-case for that?
As I know, the only possible cases inside PMD when it allocates & free same
mbuf internally are:
- error handling, queue stop (which usually not that performance critical).
- situation when you have cut crc bytes and they are in separate segment
(again a corner case).
What is the use case you have in your mind for that?
>
> In that case, we can prefetch mbufs before using them,
You can do that now, look at rte_pktmbuf_alloc_bulk().
Nothing stops you to add prefetch() here, if it will be really faster -
extra parameter for rte_pktmbuf_alloc_bulk() or new function rte_pktmbuf_alloc_bulk_prefetch().
> and
> we can free the unused mbufs without touching the structure.
In many case this is not possible anyway:
you can't guarantee that all mbufs you get from upper layer are from the same pool,
and they all have refcnt==1.
There are probably some custom cases when you can safely assume that,
but I am not sure we should restructure all mbuf API for that particular case.
In such situations some custom function should do.
> It looks to be a good advantage.
>
> Yes, we can do that with mempool functions. But I feel having a mbuf
> API with type checking is better.
>
> >> This could be done in several steps:
> >>
> >> 1/ remove these assert(), add introduce a public rte_mbuf_raw_free()
> >> 2/ announce that rte_pktmbuf_free() won't work on uninitialized mbufs
> >> in a future version, and ensure that all PMD are inline with this
> >> requirement
> >> 3/ later, move refcount initialization in rte_pktmbuf_reset()
> >>
> >> What do you think?
> >
> > I still think that assert() is on a right place :)
> > *If* we'll change mbuf free code in a way that mbufs inside the pool
> > should have refcnt==1, then I think we'll change it to:
> > RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
> > But as I stated above, that change might cause some backward compatibility hassle: 2.a)
> > Or might not give us any performance gain: 2.b).
>
> I suggest instead to have no constraint on the value of the refcnt
> in the pool.
>
> >> Another option is to remove the rte_mbuf_raw_alloc()/rte_mbuf_raw_free()
> >> and use mempool calls.
> >
> > Don't see why we have to remove them...
> > Basically we have a bug in PMD, but instead of fixing it,
> > you guys suggest to change mbuf code.
> > Sounds a bit strange to me.
> > Why not just make for these PMDs to call rte_mempool_put() directly, if you are sure it is safe here?
>
> Yes, there are some bugs in the PMDs regarding the refcnt value when
> compiled with MBUF_DEBUG=y.
Ok, then I suggest we fix them first, then start to discuss possible mbuf optimisations.
It is really strange when invalid usage of generic API in some submodules drives the generic API change.
> By the way, looking at the i40e code, we
> have:
>
> i40e_alloc_rx_queue_mbufs()
> {
> for (...) {
> struct rte_mbuf *mbuf = rte_mbuf_raw_alloc(rxq->mp);
> ...
> rte_mbuf_refcnt_set(mbuf, 1); /* not needed */
> ...
> }
Yes, I agree it is unnecessary here, and could be safely removed,
but how it relates to our discussion?
>
> i40e_tx_free_bufs()
> {
> ...
> if (txq->txq_flags & (uint32_t)ETH_TXQ_FLAGS_NOREFCOUNT) {
> for (...) {
> /* returning a mbuf with refcnt=1 */
> rte_mempool_put(txep->mbuf->pool, txep->mbuf);
> ...
Hmm, from what I see that code is not correct anyway -
as we still can't guarantee that all mbufs we TX-ed are from the same pool.
So I think that code has to be removed.
Probably i40e maintainers have a better explanation why it is here.
Again, personally I think the ETH_TXQ_FLAGS_NOREFCOUNT flag should
also be removed (or deprecated) - I think it was left here from old days.
>
> The fact that we can find many bugs related to that in PMDs is a
> sign that the API is not understandable enough.
For me it is just a sign that we need put a better effort in writing/reviewing the code.
>
> The point is not to fix the bugs in PMDs. I think the point is
> to enhance the mbuf API.
As I said - I think bugs in PMDs shouldn't drive API changes :)
They need to be fixed first.
BTW, I looked who is using __rte_mbuf_raw_free() in the current DPDK code.
I found only two instances:
1) drivers/net/mpipe/mpipe_tilegx.c:
Inside mpipe_recv_flush_stack(struct mpipe_dev_priv *priv)
As I can see that function does set mbuf->refcnt=1 before calling __raw_free.
So we should be fine here, I think.
2) drivers/net/ena/ena_ethdev.c
Inside ena_rx_queue_release_bufs().
Again that one seems ok, as I can see from the code:
ena PMD calls rte_mempool_get_bulk() and keep mbuf->refcnt==0
till either:
- that mbuf is freed by ena_rx_queue_release_bufs()
- or filled by eth_ena_recv_pkts() - at that stage mbuf->recnt is set to 1.
So another question is what PMD is failing?
>
> Hope I have convinced you ;)
>
> >
> >> But having a mbuf wrapper does not seem a bad
> >> thing to me.
> >
> > We can add some extra wrapper then, something like:
> > #define __RTE_MBUF_PUT_FREED(m) (rte_mempool_put((m)->pool, m))
> > ?
>
> I think a wrapper should do the type checking (struct mbuf).
Sure, we can have a wrapper inline function, this is just an example.
>
> Thanks for this exchange.
> Regards,
> Olivier
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v6 0/8] add packet capture framework
2016-06-08 13:38 3% ` [dpdk-dev] [PATCH v5 " Reshma Pattan
2016-06-08 13:38 5% ` [dpdk-dev] [PATCH v5 8/9] doc: update doc for " Reshma Pattan
2016-06-08 13:38 9% ` [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for rte_eth_dev_info structure Reshma Pattan
@ 2016-06-09 8:50 2% ` Reshma Pattan
2016-06-09 8:50 5% ` [dpdk-dev] [PATCH v6 8/8] doc: update doc for " Reshma Pattan
2016-06-09 16:10 2% ` [dpdk-dev] [PATCH v7 0/8] add " Reshma Pattan
2 siblings, 2 replies; 200+ results
From: Reshma Pattan @ 2016-06-09 8:50 UTC (permalink / raw)
To: dev
This patch set include below changes
1)Changes to librte_ether.
2)A new library librte_pdump added for packet capture framework.
3)A new app/pdump tool added for packet capturing.
4)Test pmd changes done to initialize packet capture framework.
5)Documentation update.
1)librte_pdump
==============
To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
is added.Users can develop their own packet capturing application using new library APIs.
Operation:
----------
Pdump library provides APIs to support packet capturing on dpdk Ethernet devices.
Library provides APIs to initialize the packet capture framework, enable/disable
the packet capture and uninitialize the packet capture framework.
Pdump library works on client/server based model.
Sever is responsible for enabling/disabling the packet captures.
Clients are responsible for requesting enable/disable of the
packet captures.
As part of packet capture framework initialization, pthread and
the server socket is created. Only one server socket is allowed on the system.
As part of enabling/disabling the packet capture, client sockets are created
and multiple client sockets are allowed.
Who ever calls initialization first they will succeed with the initialization,
next subsequent calls of initialization are not allowed. So next users can only
request enabling/disabling the packet capture.
Applications using below APIs need to pass port/device_id, queue, mempool and
ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
packets of the port for users. Users need to dequeue the rings and write the packets
to vdev(pcap/tuntap) to view the packets using any standard tools.
Note:
Mempool and Ring should be mc/mp supportable.
Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
APIs:
-----
rte_pdump_init()
rte_pdump_enable()
rte_pdump_enable_by_deviceid()
rte_pdump_disable()
rte_pdump_disable_by_deviceid()
rte_pdump_uninit()
2)app/pdump tool
================
Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
This tool by default runs as secondary process, and provides the support for
the command line options for packet capture.
./build/app/dpdk_pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue id>),
(rx-dev=<iface or pcap file> |
tx-dev=<iface or pcap file>),
[ring-size=<ring size>],
[mbuf-size=<mbuf data size>],
[total-num-mbufs=<number of mbufs>]'
Parameters inside the parenthesis represents the mandatory parameters.
Parameters inside the square brackets represents optional parameters.
User has to pass on packet capture parameters under --pdump parameters, multiples of
--pdump can be passed to capture packets on different port and queue combinations
Operation:
----------
*Tool parse the user command line arguments,
creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
of the device passed in rx-dev|tx-dev parameters.
*Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
to enable packet capturing on a specific port/device_id and queue by passing on
port|device_id, queue, mempool and ring info.
*Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
*Tool can be stopped using SIGINT, upon which tool calls
rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
Note:
CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
3)Test-pmd changes
==================
Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
Similarly any application which needs packet capture should call initialize/uninitialize APIs of
librte_pdump and use pdump tool to start the capture.
4)Packet capture flow between pdump tool and librte_pdump
=========================================================
* Pdump tool (Secondary process) requests packet capture
for specific port|device_id and queue combinations.
*Library in secondary process context creates client socket and communicates
the port|device_id, queue, ring and mempool to server.
*Library initializes server in primary process 'test-pmd' context and server serves
the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
*Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
*Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
*Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
*Library receives the disable packet capture request, communicate the info to server,
server will remove the Ethernet rxtx call-backs.
*Packet capture can be seen using tcpdump command
"tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
5)Example command line
======================
./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
v6:
removed below deprecation notice patch from patch set.
http://dpdk.org/dev/patchwork/patch/13372/
v5:
addressed code review comments for below patches
http://dpdk.org/dev/patchwork/patch/12955/
http://dpdk.org/dev/patchwork/patch/12951/
v4:
added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
made doc changes as per doc guidelines.
replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
removed rxtx-dev parameter from pdump tool command line.
v3:
app/pdump: Moved cleanup code from signal handler to main.
divided librte_ether changes into multiple patches.
example command changed in app/pdump application guide
v2:
fix compilation issues for 4.8.3
fix unnecessary #includes
Reshma Pattan (8):
librte_ether: protect add/remove of rxtx callbacks with spinlocks
librte_ether: add new api rte_eth_add_first_rx_callback
librte_ether: add new fields to rte_eth_dev_info struct
librte_ether: make rte_eth_dev_get_port_by_name
rte_eth_dev_get_name_by_port public
lib/librte_pdump: add new library for packet capturing support
app/pdump: add pdump tool for packet capturing
app/test-pmd: add pdump initialization uninitialization
doc: update doc for packet capture framework
MAINTAINERS | 8 +
app/Makefile | 1 +
app/pdump/Makefile | 45 ++
app/pdump/main.c | 814 +++++++++++++++++++++++++++++++
app/test-pmd/testpmd.c | 6 +
config/common_base | 5 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 107 ++++
doc/guides/rel_notes/release_16_07.rst | 13 +
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 +++++
lib/Makefile | 1 +
lib/librte_ether/rte_ethdev.c | 123 +++--
lib/librte_ether/rte_ethdev.h | 59 +++
lib/librte_ether/rte_ether_version.map | 9 +
lib/librte_pdump/Makefile | 55 +++
lib/librte_pdump/rte_pdump.c | 841 ++++++++++++++++++++++++++++++++
lib/librte_pdump/rte_pdump.h | 186 +++++++
lib/librte_pdump/rte_pdump_version.map | 12 +
mk/rte.app.mk | 1 +
20 files changed, 2366 insertions(+), 44 deletions(-)
create mode 100644 app/pdump/Makefile
create mode 100644 app/pdump/main.c
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
create mode 100644 lib/librte_pdump/Makefile
create mode 100644 lib/librte_pdump/rte_pdump.c
create mode 100644 lib/librte_pdump/rte_pdump.h
create mode 100644 lib/librte_pdump/rte_pdump_version.map
--
2.5.0
^ permalink raw reply [relevance 2%]
* [dpdk-dev] [PATCH v6 8/8] doc: update doc for packet capture framework
2016-06-09 8:50 2% ` [dpdk-dev] [PATCH v6 0/8] add packet capture framework Reshma Pattan
@ 2016-06-09 8:50 5% ` Reshma Pattan
2016-06-09 16:10 2% ` [dpdk-dev] [PATCH v7 0/8] add " Reshma Pattan
1 sibling, 0 replies; 200+ results
From: Reshma Pattan @ 2016-06-09 8:50 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
Added programmers guide for librte_pdump.
Added sample application guide for app/pdump application.
Updated release note for packet capture framework changes.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
Acked-by: John McNamara <john.mcnamara@intel.com>
---
MAINTAINERS | 3 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 107 ++++++++++++++++++++++++++++
doc/guides/rel_notes/release_16_07.rst | 13 ++++
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 ++++++++++++++++++++++++++++++++
6 files changed, 247 insertions(+)
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
diff --git a/MAINTAINERS b/MAINTAINERS
index a48c8de..ce7c941 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -436,6 +436,9 @@ Pdump
M: Reshma Pattan <reshma.pattan@intel.com>
F: lib/librte_pdump/
F: app/pdump/
+F: doc/guides/prog_guide/pdump_library.rst
+F: doc/guides/sample_app_ug/pdump.rst
+
Hierarchical scheduler
M: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
diff --git a/doc/guides/prog_guide/index.rst b/doc/guides/prog_guide/index.rst
index b862d0c..4caf969 100644
--- a/doc/guides/prog_guide/index.rst
+++ b/doc/guides/prog_guide/index.rst
@@ -71,6 +71,7 @@ Programmer's Guide
writing_efficient_code
profile_app
glossary
+ pdump_library
**Figures**
diff --git a/doc/guides/prog_guide/pdump_library.rst b/doc/guides/prog_guide/pdump_library.rst
new file mode 100644
index 0000000..1809234
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,107 @@
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+.. _pdump_library:
+
+The librte_pdump Library
+========================
+
+The ``librte_pdump`` library provides a framework for packet capturing in DPDK.
+The library provides the following APIs to initialize the packet capture framework, to enable
+or disable the packet capture, and to uninitialize it:
+
+* ``rte_pdump_init()``:
+ This API initializes the packet capture framework.
+
+* ``rte_pdump_enable()``:
+ This API enables the packet capture on a given port and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_enable_by_deviceid()``:
+ This API enables the packet capture on a given device id (``vdev name or pci address``) and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_disable()``:
+ This API disables the packet capture on a given port and queue.
+
+* ``rte_pdump_disable_by_deviceid()``:
+ This API disables the packet capture on a given device id (``vdev name or pci address``) and queue.
+
+* ``rte_pdump_uninit()``:
+ This API uninitializes the packet capture framework.
+
+
+Operation
+---------
+
+The ``librte_pdump`` library works on a client/server model. The server is responsible for enabling or
+disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
+the packet capture.
+
+The packet capture framework, as part of its initialization, creates the pthread and the server socket in
+the pthread. The application that calls the framework initialization first will have the server socket created.
+Further calls to the framework initialization by the same application or other applications is not allowed i.e., only
+one server socket is allowed on the system. So the other applications can only request enabling or disabling of
+the packet capture at which point the client socket is created for them to send the request to the server.
+The server socket will listen for client requests for enabling or disabling the packet capture.
+
+
+Implementation Details
+----------------------
+
+The library API ``rte_pdump_init()``, initializes the packet capture framework by creating the pthread and the server
+socket. The server socket in the pthread context will be listening to the client requests to enable or disable the
+packet capture. Whoever calls this API first will have the server socket created, the subsequent calls to this APIs
+will not create any further server socket. i.e. only one server socket is allowed.
+
+The library APIs ``rte_pdump_enable()`` and ``rte_pdump_enable_by_deviceid()`` enables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump enable" request and sends
+the request to the server. The server that is listening on the socket will take the request and enable the packet capture
+by registering the Ethernet RX and TX callbacks for the given port or device_id and queue combinations.
+Then the server will mirror the packets to the new mempool and enqueue them to the rte_ring that clients have passed
+to these APIs. The server also sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, the client socket is closed.
+
+The library APIs ``rte_pdump_disable()`` and ``rte_pdump_disable_by_deviceid()`` disables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump disable" request and sends
+the request to the server. The server that is listening on the socket will take the request and disable the packet
+capture by removing the Ethernet RX and TX callbacks for the given port or device_id and queue combinations. The server
+also sends the response back to the client about the status of the request that was processed. After the response is
+received from the server, the client socket is closed.
+
+The library API ``rte_pdump_uninit()``, uninitializes the packet capture framework by closing the pthread and the
+server socket.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+The DPDK ``app/pdump`` tool is developed based on this library to capture packets in DPDK.
+Users can use this as an example to develop their own packet capturing application.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index c0f6b02..a4de2a2 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -66,6 +66,11 @@ New Features
* Enable RSS per network interface through the configuration file.
* Streamline the CLI code.
+* **Added packet capture framework.**
+
+ * A new library ``librte_pdump`` is added to provide packet capture APIs.
+ * A new ``app/pdump`` tool is added to capture packets in DPDK.
+
Resolved Issues
---------------
@@ -135,6 +140,11 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* Function ``rte_eth_dev_get_port_by_name`` changed to a public API.
+
+* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ in the ``rte_eth_dev_info`` object.
+
ABI Changes
-----------
@@ -146,6 +156,9 @@ ABI Changes
* The ``rte_port_source_params`` structure has new fields to support PCAP file.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ to support number of queues configured by software.
+
Shared Library Versions
-----------------------
diff --git a/doc/guides/sample_app_ug/index.rst b/doc/guides/sample_app_ug/index.rst
index 930f68c..96bb317 100644
--- a/doc/guides/sample_app_ug/index.rst
+++ b/doc/guides/sample_app_ug/index.rst
@@ -76,6 +76,7 @@ Sample Applications User Guide
ptpclient
performance_thread
ipsec_secgw
+ pdump
**Figures**
diff --git a/doc/guides/sample_app_ug/pdump.rst b/doc/guides/sample_app_ug/pdump.rst
new file mode 100644
index 0000000..96c8709
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,122 @@
+
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+dpdk_pdump Application
+======================
+
+The ``dpdk_pdump`` application is a Data Plane Development Kit (DPDK) application that runs as a DPDK secondary process and
+is capable of enabling packet capture on dpdk ports.
+
+
+Running the Application
+-----------------------
+
+The application has a ``--pdump`` command line option with various sub arguments:
+
+.. code-block:: console
+
+ ./build/app/dpdk_pdump --
+ --pdump '(port=<port id> | device_id=<pci id or vdev name>),
+ (queue=<queue_id>),
+ (rx-dev=<iface or pcap file> |
+ tx-dev=<iface or pcap file>),
+ [ring-size=<ring size>],
+ [mbuf-size=<mbuf data size>],
+ [total-num-mbufs=<number of mbufs>]'
+
+Note:
+
+* Parameters inside the parentheses represents mandatory parameters.
+
+* Parameters inside the square brackets represents optional parameters.
+
+Multiple instances of ``--pdump`` can be passed to capture packets on different port and queue combinations.
+
+
+Parameters
+~~~~~~~~~~
+
+``port``:
+Port id of the eth device on which packets should be captured.
+
+``device_id``:
+PCI address (or) name of the eth device on which packets should be captured.
+
+ .. Note::
+
+ * As of now the ``dpdk_pdump`` tool cannot capture the packets of virtual devices
+ in the primary process due to a bug in the ethdev library. Due to this bug, in a multi process context,
+ when the primary and secondary have different ports set, then the secondary process
+ (here the ``dpdk_pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
+
+``queue``:
+Queue id of the eth device on which packets should be captured. The user can pass a queue value of ``*`` to enable
+packet capture on all queues of the eth device.
+
+``rx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+``tx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+ .. Note::
+
+ * To receive ingress packets only, ``rx-dev`` should be passed.
+
+ * To receive egress packets only, ``tx-dev`` should be passed.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the different file names or the Linux iface names.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the same file names or the the Linux iface names.
+
+``ring-size``:
+Size of the ring. This value is used internally for ring creation. The ring will be used to enqueue the packets from
+the primary application to the secondary. This is an optional parameter with default size 16384.
+
+``mbuf-size``:
+Size of the mbuf data. This is used internally for mempool creation. Ideally this value must be same as
+the primary application's mempool's mbuf data size which is used for packet RX. This is an optional parameter with
+default size 2176.
+
+``total-num-mbufs``:
+Total number mbufs in mempool. This is used internally for mempool creation. This is an optional parameter with default
+value 65535.
+
+
+Example
+-------
+
+.. code-block:: console
+
+ $ sudo ./build/app/dpdk_pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
--
2.5.0
^ permalink raw reply [relevance 5%]
* Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-08 7:34 0% ` Lu, Wenzhuo
@ 2016-06-09 7:50 0% ` Olivier Matz
2016-06-12 5:25 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-06-09 7:50 UTC (permalink / raw)
To: Lu, Wenzhuo, Stephen Hemminger; +Cc: dev, Tao, Zhe
Hi,
On 06/08/2016 09:34 AM, Lu, Wenzhuo wrote:
> Hi Stephen,
>
>
>> -----Original Message-----
>> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
>> Sent: Wednesday, June 8, 2016 10:16 AM
>> To: Lu, Wenzhuo
>> Cc: dev@dpdk.org; Tao, Zhe
>> Subject: Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
>>
>> On Mon, 6 Jun 2016 13:40:47 +0800
>> Wenzhuo Lu <wenzhuo.lu@intel.com> wrote:
>>
>>> Define lock mode for RX/TX queue. Because when resetting the device we
>>> want the resetting thread to get the lock of the RX/TX queue to make
>>> sure the RX/TX is stopped.
>>>
>>> Using next ABI macro for this ABI change as it has too much impact. 7
>>> APIs and 1 global variable are impacted.
>>>
>>> Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
>>> Signed-off-by: Zhe Tao <zhe.tao@intel.com>
>>
>> Why does this patch set make a different assumption the rest of the DPDK?
>>
>> The rest of the DPDK operates on the principle that the application is smart
>> enough to stop the device before making changes. There is no equivalent to the
>> Linux kernel RTNL mutex. The API assumes application threads are well behaved
>> and will not try and sabotage each other.
>>
>> If you restrict the reset operation to only being available when RX/TX is stopped,
>> then no lock is needed.
>>
>> The fact that it requires lots more locking inside each device driver implies to me
>> this is not correct way to architect this.
+1
I'm not sure adding locks is the proper way to do.
This is the application responsibility to ensure that:
- control functions are not called concurrently on the same port
- rx/tx functions are not called when the device is stopped/reset/...
However, I do think the usage paradigms of the ethdev api should be
better documented in rte_ethdev.h (ex: which functions can be called
concurrently). This would be a first step.
If we really want a helper API to do that in DPDK, the _next_ step
could be to add them in the ethdev api to achieve this. Maybe
something like (the function names could be better):
- to be called on one control thread:
rte_eth_stop_rxtx(port)
rte_eth_start_rxtx(port)
rte_eth_get_rxtx_state(port)
-> return "running" if at least one core is inside the rx/tx code
-> return "stopped" if all cores are outside the rx/tx code
- to be called on dataplane cores:
/* same than rte_eth_rx_burst(), but checks if rx/tx is allowed
* first, else do nothing */
rte_eth_rx_burst_interruptible()
rte_eth_tx_burst_interruptible()
The code of control thread could be:
rte_eth_stop_rxtx(port);
/* wait that all dataplane cores finished their processing */
while (rte_eth_get_rxtx_state(port) != stopped)
;
rte_eth_some_control_operation(port);
rte_eth_start_rxtx(port);
I think this could be done without any lock, just with the proper
memory barriers and a per-core status.
But this API may impose a paradigm to the application, and I'm not
sure the DPDK should do that.
Regards,
Olivier
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH] mbuf: remove inconsistent assert statements
2016-06-08 16:07 3% ` Ananyev, Konstantin
@ 2016-06-09 7:46 0% ` Olivier Matz
2016-06-09 13:21 0% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-06-09 7:46 UTC (permalink / raw)
To: Ananyev, Konstantin, dev, Adrien Mazarguil
Hi Konstantin,
>>>>>> Yes, it refcnt supposed to be set to 0 by __rte_pktmbuf_prefree_seg().
>>>>>> Wright now, it is a user responsibility to make sure refcnt==0 before pushing
>>>>>> mbuf back to the pool.
>>>>>> Not sure why do you consider that wrong?
>>>>>
>>>>> I do not consider this wrong and I'm all for using assert() to catch
>>>>> programming errors, however in this specific case, I think they are
>>>>> inconsistent and misleading.
>>>>
>>>> Honestly, I don't understand why.
>>>> Right now the rule of thumb is - when mbuf is in the pool, it's refcnt should be equal zero.
>>
>> What is the purpose of this? Is there some code relying on this?
>
> The whole current implementation of mbuf_free code path relies on that.
> Straight here:
> if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
> m->next = NULL;
> __rte_mbuf_raw_free(m);
> }
>
> If we'll exclude indirect mbuf logic, all it does:
> if (rte_mbuf_refcnt_update(m, -1) == 0) {
> m->next = NULL;
> __rte_mbuf_raw_free(m);
> }
>
> I.E.:
> decrement mbuf->refcnt.
> If new value of refcnt is zero, then put it back into the pool.
>
> So having ASERT(mbuf->refcnt==0) inside
> __rte_mbuf_raw_free()/__rte_mbuf_raw_alloc()
> looks absolutely valid to me.
> I *has* to be zero at that point with current implementation,
> And if it is not then we probably have (or will have) a silent memory corruption.
This explains how the refcount is used, and why it is set
to zero before returning the mbuf to the pool with the mbuf
free functions.
It does not explain which code relies on the refcnt beeing 0
while the mbuf is in the pool.
>> But since [1], it is allowed to call rte_mbuf_raw_alloc() in PMDs (all
>> PMDs were calling an internal function before).
>
> Yes, raw_alloc is public, NP with that.
>
>> We could argue that
>> rte_mbuf_raw_free() should also be made public for PMDs.
>
> We could, but right now it is not.
> Again, as I said, user could use it on his own but he obviously has to
> obey the rules and do manually what __rte_pktmbuf_prefree_seg() does.
> At least:
>
> rte_mbuf_refcnt_set(m, 0);
> __rte_mbuf_raw_free(m);
>
>>
>> As you said below, no-one promised that the free() reverts the malloc(),
>> but given the function names, one can legitimately expect that the
>> following code is valid:
>>
>> m = __rte_mbuf_raw_alloc();
>> /* do nothing here */
>> __rte_mbuf_raw_free(m);
>
> Surely people could (and would) expect various things...
> But the reality right now is: __rte_mbuf_raw_free() is an internal
> function and not counterpart of __rte_mbuf_raw_alloc().
> If the people don't bother to read API docs or/and the actual code,
> I can't see how we can help them :)
Yes, of course, people should read the doc.
This does not prevent to have a nice API that behaves in a
natural way :)
By the way, the fact that today the mbuf refcnt should be 0 while
in a pool is not in the api doc, but in the code.
>> If no code relies on having the refcnt set to 0 when a mbuf is in
>> the pool, I suggest to relax this constraint as Adrien proposed.
>
> Why not just rename it to __rte_mbuf_raw_free_dont_use_after_raw_alloc()?
> To avoid any further confusion :)
> Seriously speaking I would prefer to leave it as it is.
> If you feel we have to introduce a counterpart of rte_mbuf_raw_alloc(),
> we can make a new public one:
>
> rte_mbuf_raw_free(stuct rte_mbuf *m)
> {
> if (rte_mbuf_refcnt_update(m, -1) == 0)
> __rte_mbuf_raw_free(m);
> }
This is an option, but I think it's not efficient to touch
the mbuf structure when allocating/freeing. See below.
>> Then, my opinion is that the refcount should be set to 1 in
>> rte_pktmbuf_reset().
>
> I don't think we need to update rte_pktmbuf_reset(),
> it doesn't touch refcnt at all and probably better to keep it that way.
Why would it be better?
All mbuf struct initializations are done in that function, why would
it be different for the refcnt?
> To achieve what you suggesting, we probably need to:
> 1) update _rte_pktmbuf_prefree_seg and rte_pktmbuf_detach() to
> set refcnt back to 1, i.e:
>
> static inline struct rte_mbuf* __attribute__((always_inline))
> __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
> {
> __rte_mbuf_sanity_check(m, 0);
>
> if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
> /* if this is an indirect mbuf, it is detached. */
> if (RTE_MBUF_INDIRECT(m))
> rte_pktmbuf_detach(m);
> + rte_mbuf_refcnt_set(m, 1);
> return m;
> }
> return NULL;
> }
>
> 2) either:
> a) update mbuf constructor function, so it sets refcnt=1.
> I suppose that is easy for rte_pktmbuf_init(), but it means that all custom
> constructors should do the same.
> Which means possible changes in existing user code and all ABI change related hassle.
> b) keep rte_mbuf_raw_alloc() setting mbuf->refcnt=1.
> But then I don't see how will get any performance gain here.
>
> So not sure is it really worth it.
>
>> And rte_pktmbuf_free() should not be allowed on
>> an uninitialized mbuf (yes, this would require some changes in PMDs).
>
> Not sure I understand you here...
> free() wouldn not be allowed on mbuf whose recnf==0?
> But it is not allowed right now anyway.
Today:
/* allowed */
m = rte_pktmbuf_alloc();
rte_pktmbuf_free(m);
/* not allowed */
m = rte_mbuf_raw_alloc();
__rte_mbuf_raw_free(m);
/* we should do instead (strange): */
m = rte_mbuf_raw_alloc();
rte_pktmbuf_free(m);
What I suggest to have:
/* allowed, no change */
m = rte_pktmbuf_alloc();
rte_pktmbuf_free(m);
/* allowed, these functions would be symetrical */
m = rte_mbuf_raw_alloc();
rte_mbuf_raw_free(m);
/* not allowed, m->refcnt is uninitialized */
m = rte_mbuf_raw_alloc();
rte_pktmbuf_free(m);
>
>> This would open the door for bulk allocation/free in the mbuf api.
>
> Hmm and what stops us having one right now?
> As I know we do have rte_pktmbuf_alloc_bulk(), I don't see
> why we can't have rte_pktmbuf_free_bulk() right now.
I think, not touching the refcnt on raw allocation allows a PMD to
do the following:
rte_mbuf_allow_raw_bulk(pool, &mbufs, n)
for (i = 0; i < n; i++) {
prefetch(m[i+1]);
rte_pktmbuf_reset(); /* including refcnt = 1 */
do_stuff(m);
}
/* if all mbufs are not used, free the remaining */
rte_mbuf_free_raw_bulk(&mbufs[i], n-i);
In that case, we can prefetch mbufs before using them, and
we can free the unused mbufs without touching the structure.
It looks to be a good advantage.
Yes, we can do that with mempool functions. But I feel having a mbuf
API with type checking is better.
>> This could be done in several steps:
>>
>> 1/ remove these assert(), add introduce a public rte_mbuf_raw_free()
>> 2/ announce that rte_pktmbuf_free() won't work on uninitialized mbufs
>> in a future version, and ensure that all PMD are inline with this
>> requirement
>> 3/ later, move refcount initialization in rte_pktmbuf_reset()
>>
>> What do you think?
>
> I still think that assert() is on a right place :)
> *If* we'll change mbuf free code in a way that mbufs inside the pool
> should have refcnt==1, then I think we'll change it to:
> RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
> But as I stated above, that change might cause some backward compatibility hassle: 2.a)
> Or might not give us any performance gain: 2.b).
I suggest instead to have no constraint on the value of the refcnt
in the pool.
>> Another option is to remove the rte_mbuf_raw_alloc()/rte_mbuf_raw_free()
>> and use mempool calls.
>
> Don't see why we have to remove them...
> Basically we have a bug in PMD, but instead of fixing it,
> you guys suggest to change mbuf code.
> Sounds a bit strange to me.
> Why not just make for these PMDs to call rte_mempool_put() directly, if you are sure it is safe here?
Yes, there are some bugs in the PMDs regarding the refcnt value when
compiled with MBUF_DEBUG=y. By the way, looking at the i40e code, we
have:
i40e_alloc_rx_queue_mbufs()
{
for (...) {
struct rte_mbuf *mbuf = rte_mbuf_raw_alloc(rxq->mp);
...
rte_mbuf_refcnt_set(mbuf, 1); /* not needed */
...
}
i40e_tx_free_bufs()
{
...
if (txq->txq_flags & (uint32_t)ETH_TXQ_FLAGS_NOREFCOUNT) {
for (...) {
/* returning a mbuf with refcnt=1 */
rte_mempool_put(txep->mbuf->pool, txep->mbuf);
...
The fact that we can find many bugs related to that in PMDs is a
sign that the API is not understandable enough.
The point is not to fix the bugs in PMDs. I think the point is
to enhance the mbuf API.
Hope I have convinced you ;)
>
>> But having a mbuf wrapper does not seem a bad
>> thing to me.
>
> We can add some extra wrapper then, something like:
> #define __RTE_MBUF_PUT_FREED(m) (rte_mempool_put((m)->pool, m))
> ?
I think a wrapper should do the type checking (struct mbuf).
Thanks for this exchange.
Regards,
Olivier
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for rte_eth_dev_info structure
2016-06-08 13:38 9% ` [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for rte_eth_dev_info structure Reshma Pattan
@ 2016-06-08 16:15 4% ` Mcnamara, John
0 siblings, 0 replies; 200+ results
From: Mcnamara, John @ 2016-06-08 16:15 UTC (permalink / raw)
To: Pattan, Reshma, dev; +Cc: Pattan, Reshma
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Reshma Pattan
> Sent: Wednesday, June 8, 2016 2:38 PM
> To: dev@dpdk.org
> Cc: Pattan, Reshma <reshma.pattan@intel.com>
> Subject: [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for
> rte_eth_dev_info structure
>
> New fields nb_rx_queues and nb_tx_queues will be added to rte_eth_dev_info
> structure.
> Changes to API rte_eth_dev_info_get() will be done to update these new
> fields to rte_eth_dev_info object.
>
> Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
> ---
> doc/guides/rel_notes/deprecation.rst | 6 ++++++
> 1 file changed, 6 insertions(+)
>
Hi,
This isn't required in this patchset since the deprecation has already occurred in this release.
John
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH v2] doc: fix code section in abi versioning doc
@ 2016-06-08 16:46 4% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-08 16:46 UTC (permalink / raw)
To: John McNamara; +Cc: dev
2016-05-20 15:08, John McNamara:
> Fix broken cosole directive in the ABI validator section of the
> ABI versioning docs.
>
> Fixes: f1ef9794f9bd ("doc: add ABI guidelines")
>
> Signed-off-by: John McNamara <john.mcnamara@intel.com>
Applied, thanks
^ permalink raw reply [relevance 4%]
* Re: [dpdk-dev] [PATCH] mbuf: remove inconsistent assert statements
@ 2016-06-08 16:07 3% ` Ananyev, Konstantin
2016-06-09 7:46 0% ` Olivier Matz
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-06-08 16:07 UTC (permalink / raw)
To: Olivier Matz, dev, Adrien Mazarguil
Hi Olivier,
>
> Hi Adrien, Konstantin,
>
> I'm jumping in this (interesting) discussion. I already talked a
> bit with Adrien in point to point, and I think its patch is valid.
> Please see some comments below.
>
>
> On 06/08/2016 03:57 PM, Adrien Mazarguil wrote:
> > On Wed, Jun 08, 2016 at 01:09:18PM +0000, Ananyev, Konstantin wrote:
> >>>
> >>> Hi Konstantin,
> >>>
> >>> On Wed, Jun 08, 2016 at 10:34:17AM +0000, Ananyev, Konstantin wrote:
> >>>> Hi Adrien,
> >>>>
> >>>>>
> >>>>> An assertion failure occurs in __rte_mbuf_raw_free() (called by a few PMDs)
> >>>>> when compiling DPDK with CONFIG_RTE_LOG_LEVEL=RTE_LOG_DEBUG and starting
> >>>>> applications with a log level high enough to trigger it.
> >>>>>
> >>>>> While rte_mbuf_raw_alloc() sets refcount to 1, __rte_mbuf_raw_free()
> >>>>> expects it to be 0.
> >>>>> Considering users are not expected to reset the
> >>>>> reference count to satisfy assert() and that raw functions are designed on
> >>>>> purpose without safety belts, remove these checks.
> >>>>
> >>>> Yes, it refcnt supposed to be set to 0 by __rte_pktmbuf_prefree_seg().
> >>>> Wright now, it is a user responsibility to make sure refcnt==0 before pushing
> >>>> mbuf back to the pool.
> >>>> Not sure why do you consider that wrong?
> >>>
> >>> I do not consider this wrong and I'm all for using assert() to catch
> >>> programming errors, however in this specific case, I think they are
> >>> inconsistent and misleading.
> >>
> >> Honestly, I don't understand why.
> >> Right now the rule of thumb is - when mbuf is in the pool, it's refcnt should be equal zero.
>
> What is the purpose of this? Is there some code relying on this?
The whole current implementation of mbuf_free code path relies on that.
Straight here:
if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
m->next = NULL;
__rte_mbuf_raw_free(m);
}
If we'll exclude indirect mbuf logic, all it does:
if (rte_mbuf_refcnt_update(m, -1) == 0) {
m->next = NULL;
__rte_mbuf_raw_free(m);
}
I.E.:
decrement mbuf->refcnt.
If new value of refcnt is zero, then put it back into the pool.
So having ASERT(mbuf->refcnt==0) inside
__rte_mbuf_raw_free()/__rte_mbuf_raw_alloc()
looks absolutely valid to me.
I *has* to be zero at that point with current implementation,
And if it is not then we probably have (or will have) a silent memory corruption.
>
> >> Yes, as you pointed below - that rule probably can be changed to:
> >> when mbuf is in the pool, it's refcnt should equal one, and that would probably allow us
> >> to speedup things a bit, but I suppose that's the matter of another aptch/discussion.
> >
> > Agreed.
>
> I agree this is somehow another discussion, but let's dive into :)
OK :)
>
> But since [1], it is allowed to call rte_mbuf_raw_alloc() in PMDs (all
> PMDs were calling an internal function before).
Yes, raw_alloc is public, NP with that.
> We could argue that
> rte_mbuf_raw_free() should also be made public for PMDs.
We could, but right now it is not.
Again, as I said, user could use it on his own but he obviously has to
obey the rules and do manually what __rte_pktmbuf_prefree_seg() does.
At least:
rte_mbuf_refcnt_set(m, 0);
__rte_mbuf_raw_free(m);
>
> As you said below, no-one promised that the free() reverts the malloc(),
> but given the function names, one can legitimately expect that the
> following code is valid:
>
> m = __rte_mbuf_raw_alloc();
> /* do nothing here */
> __rte_mbuf_raw_free(m);
Surely people could (and would) expect various things...
But the reality right now is: __rte_mbuf_raw_free() is an internal
function and not counterpart of __rte_mbuf_raw_alloc().
If the people don't bother to read API docs or/and the actual code,
I can't see how we can help them :)
>
> If no code relies on having the refcnt set to 0 when a mbuf is in
> the pool, I suggest to relax this constraint as Adrien proposed.
Why not just rename it to __rte_mbuf_raw_free_dont_use_after_raw_alloc()?
To avoid any further confusion :)
Seriously speaking I would prefer to leave it as it is.
If you feel we have to introduce a counterpart of rte_mbuf_raw_alloc(),
we can make a new public one:
rte_mbuf_raw_free(stuct rte_mbuf *m)
{
if (rte_mbuf_refcnt_update(m, -1) == 0)
__rte_mbuf_raw_free(m);
}
>
> Then, my opinion is that the refcount should be set to 1 in
> rte_pktmbuf_reset().
I don't think we need to update rte_pktmbuf_reset(),
it doesn't touch refcnt at all and probably better to keep it that way.
To achieve what you suggesting, we probably need to:
1) update _rte_pktmbuf_prefree_seg and rte_pktmbuf_detach() to
set refcnt back to 1, i.e:
static inline struct rte_mbuf* __attribute__((always_inline))
__rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
{
__rte_mbuf_sanity_check(m, 0);
if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
/* if this is an indirect mbuf, it is detached. */
if (RTE_MBUF_INDIRECT(m))
rte_pktmbuf_detach(m);
+ rte_mbuf_refcnt_set(m, 1);
return m;
}
return NULL;
}
2) either:
a) update mbuf constructor function, so it sets refcnt=1.
I suppose that is easy for rte_pktmbuf_init(), but it means that all custom
constructors should do the same.
Which means possible changes in existing user code and all ABI change related hassle.
b) keep rte_mbuf_raw_alloc() setting mbuf->refcnt=1.
But then I don't see how will get any performance gain here.
So not sure is it really worth it.
> And rte_pktmbuf_free() should not be allowed on
> an uninitialized mbuf (yes, this would require some changes in PMDs).
Not sure I understand you here...
free() wouldn not be allowed on mbuf whose recnf==0?
But it is not allowed right now anyway.
> This would open the door for bulk allocation/free in the mbuf api.
Hmm and what stops us having one right now?
As I know we do have rte_pktmbuf_alloc_bulk(), I don't see
why we can't have rte_pktmbuf_free_bulk() right now.
>
> This could be done in several steps:
>
> 1/ remove these assert(), add introduce a public rte_mbuf_raw_free()
> 2/ announce that rte_pktmbuf_free() won't work on uninitialized mbufs
> in a future version, and ensure that all PMD are inline with this
> requirement
> 3/ later, move refcount initialization in rte_pktmbuf_reset()
>
> What do you think?
I still think that assert() is on a right place :)
*If* we'll change mbuf free code in a way that mbufs inside the pool
should have refcnt==1, then I think we'll change it to:
RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
But as I stated above, that change might cause some backward compatibility hassle: 2.a)
Or might not give us any performance gain: 2.b).
>
> Another option is to remove the rte_mbuf_raw_alloc()/rte_mbuf_raw_free()
> and use mempool calls.
Don't see why we have to remove them...
Basically we have a bug in PMD, but instead of fixing it,
you guys suggest to change mbuf code.
Sounds a bit strange to me.
Why not just make for these PMDs to call rte_mempool_put() directly, if you are sure it is safe here?
> But having a mbuf wrapper does not seem a bad
> thing to me.
We can add some extra wrapper then, something like:
#define __RTE_MBUF_PUT_FREED(m) (rte_mempool_put((m)->pool, m))
?
Konstantin
>
> By the way, __rte_pktmbuf_prefree_seg() is also an internal function.
> I think we should try to update the mbuf API so that the PMDs do not
> need to call these internal functions.
>
>
> [1]
> http://dpdk.org/browse/dpdk/commit/?id=fbfd99551ca370266f4bfff58ce441cf5cb1203a
>
>
> Regards,
> Olivier
>
>
> >
> >>>> If the user calls __rte_mbuf_raw_free() manualy it is his responsibility to make
> >>>> sure mbuf's refcn==0.
> >>>
> >>> Sure, however what harm does it cause (besides assert() to fail), since the
> >>> allocation function sets refcount to 1?
> >>>
> >>> Why having the allocation function set the refcount if we are sure it is
> >>> already 0 (assert() proves it). Removing rte_mbuf_refcnt_set(m, 1) can
> >>> surely improve performance.
> >>
> >> That's' just an assert() enabled when MBUF_DEBUG is on.
> >> Its sole purpose is to help troubleshoot the bugs and help to catch situations
> >> when someone silently updates mbufs supposed to be free.
> >
> > I perfectly understand and I cannot agree more with this explanation,
> > however the fact these functions are not symmetrical remains an issue that
> > needs to be addressed somehow in my opinion.
> >
> >>>> BTW, why are you doing it?
> >>>> The comment clearly states that the function is for internal use:
> >>>> /**
> >>>> * @internal Put mbuf back into its original mempool.
> >>>> * The use of that function is reserved for RTE internal needs.
> >>>> * Please use rte_pktmbuf_free().
> >>>> *
> >>>> * @param m
> >>>> * The mbuf to be freed.
> >>>> */
> >>>> static inline void __attribute__((always_inline))
> >>>> __rte_mbuf_raw_free(struct rte_mbuf *m)
> >>>
> >>> Several PMDs are using it anyway (won't name names, but I know one of them
> >>> quite well).
> >>
> >> Then it probably is a bug in these PMDs that need to be fixed.
> >>
> >>> I chose to modify this code instead of its users for the
> >>> following reasons:
> >>>
> >>> - Considering their names, these functions should be opposites and able to
> >>> work together like malloc()/free().
> >>
> >> These are internal functions.
> >> Comments in mbuf clearly state that library users shouldn't call them directly.
> >> They are written to fit internal librte_mbuf needs, and no-one ever promised
> >> malloc/free() compatibility here.
> >
> > So it cannot be provided for the sake of not providing it or is there a good
> > reason?
> >
> > What I meant is that since PMDs already made the mistake of using these
> > functions to benefit from the improved performance, DPDK being all about
> > performance and stuff, let them use it as intended. Perhaps we should drop
> > those "__" like for rte_mbuf_raw_alloc().
> >
> >>>
> >>> - PMDs are relying on these functions for performance reasons, we can assume
> >>> they took the extra care necessary to make sure it would work properly.
> >>
> >> That just doesn't seem correct to me.
> >> The proper way to do free fo mbuf segment is:
> >>
> >> static inline void __attribute__((always_inline))
> >> rte_pktmbuf_free_seg(struct rte_mbuf *m)
> >> {
> >> if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
> >> m->next = NULL;
> >> __rte_mbuf_raw_free(m);
> >> }
> >> }
> >>
> >> If by some reason you choose not to use this function, then it is your
> >> responsibility to perform similar actions on your own before pushing mbuf into the pool.
> >> That's what some TX functions for some Intel NICs do to improve performance:
> >> they call _prefree_seg() manually and try to put mbufs into the pool in groups.
> >
> > Not anymore it seems, but in the current code base both ena and mpipe PMDs
> > (soon mlx5 as well) seem to get this wrong.
> >
> >>> - Preventing it would make these PMDs slower and is not acceptable either.
> >>
> >> I can hardly imagine that __rte_pktmbuf_prefree_seg() impact would be that severe...
> >> But ok, probably you do have some very specific case, but then why you PMD just doesn't call:
> >> rte_mempool_put(m->pool, m);
> >> directly?
> >
> > To survive the upcoming transition to mbufs backed by libc malloc() without
> > having to fix them? Joke aside, I guess the reason is to use functions with
> > "mbuf" in their names when dealing with mbufs.
> >
> >> Why instead you choose to change common functions and compromise
> >> librte_mbuf debug ability?
> >
> > No, I'm fine with keeping the debug ability, however I did not find a way to
> > both keep it and fix the consistency issue without breaking something
> > (performance or refcount assumptions I'm not familiar with elsewhere).
> >
> >>> What remains is the consistency issue, I think these statements were only
> >>> added to catch multiple frees,
> >>
> >> Yes these asserts() here to help catch bugs,
> >> and I think it is a good thing to have them here.
> >>
> >>> and those should be caught at a higher
> >>> level, where other consistency checks are also performed.
> >>
> >> Like where?
> >
> > Possibly rte_pktmbuf_free_seg().
> >
> >>>>> Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
> >>>>> ---
> >>>>> lib/librte_mbuf/rte_mbuf.h | 2 --
> >>>>> 1 file changed, 2 deletions(-)
> >>>>>
> >>>>> diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
> >>>>> index 11fa06d..7070bb8 100644
> >>>>> --- a/lib/librte_mbuf/rte_mbuf.h
> >>>>> +++ b/lib/librte_mbuf/rte_mbuf.h
> >>>>> @@ -1108,7 +1108,6 @@ static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
> >>>>> if (rte_mempool_get(mp, &mb) < 0)
> >>>>> return NULL;
> >>>>> m = (struct rte_mbuf *)mb;
> >>>>> - RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
> >>>>> rte_mbuf_refcnt_set(m, 1);
> >>>>> __rte_mbuf_sanity_check(m, 0);
> >>>>>
> >>>>> @@ -1133,7 +1132,6 @@ __rte_mbuf_raw_alloc(struct rte_mempool *mp)
> >>>>> static inline void __attribute__((always_inline))
> >>>>> __rte_mbuf_raw_free(struct rte_mbuf *m)
> >>>>> {
> >>>>> - RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
> >>>>> rte_mempool_put(m->pool, m);
> >>>>> }
> >>>>>
> >>>>> --
> >>>>> 2.1.4
> >>>>
> >>>
> >>> --
> >>> Adrien Mazarguil
> >>> 6WIND
> >
^ permalink raw reply [relevance 3%]
* [dpdk-dev] [PATCH v5 8/9] doc: update doc for packet capture framework
2016-06-08 13:38 3% ` [dpdk-dev] [PATCH v5 " Reshma Pattan
@ 2016-06-08 13:38 5% ` Reshma Pattan
2016-06-08 13:38 9% ` [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for rte_eth_dev_info structure Reshma Pattan
2016-06-09 8:50 2% ` [dpdk-dev] [PATCH v6 0/8] add packet capture framework Reshma Pattan
2 siblings, 0 replies; 200+ results
From: Reshma Pattan @ 2016-06-08 13:38 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
Added programmers guide for librte_pdump.
Added sample application guide for app/pdump application.
Updated release note for packet capture framework changes.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
---
MAINTAINERS | 3 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 107 ++++++++++++++++++++++++++++
doc/guides/rel_notes/release_16_07.rst | 13 ++++
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 ++++++++++++++++++++++++++++++++
6 files changed, 247 insertions(+)
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
diff --git a/MAINTAINERS b/MAINTAINERS
index a48c8de..ce7c941 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -436,6 +436,9 @@ Pdump
M: Reshma Pattan <reshma.pattan@intel.com>
F: lib/librte_pdump/
F: app/pdump/
+F: doc/guides/prog_guide/pdump_library.rst
+F: doc/guides/sample_app_ug/pdump.rst
+
Hierarchical scheduler
M: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
diff --git a/doc/guides/prog_guide/index.rst b/doc/guides/prog_guide/index.rst
index b862d0c..4caf969 100644
--- a/doc/guides/prog_guide/index.rst
+++ b/doc/guides/prog_guide/index.rst
@@ -71,6 +71,7 @@ Programmer's Guide
writing_efficient_code
profile_app
glossary
+ pdump_library
**Figures**
diff --git a/doc/guides/prog_guide/pdump_library.rst b/doc/guides/prog_guide/pdump_library.rst
new file mode 100644
index 0000000..1809234
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,107 @@
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+.. _pdump_library:
+
+The librte_pdump Library
+========================
+
+The ``librte_pdump`` library provides a framework for packet capturing in DPDK.
+The library provides the following APIs to initialize the packet capture framework, to enable
+or disable the packet capture, and to uninitialize it:
+
+* ``rte_pdump_init()``:
+ This API initializes the packet capture framework.
+
+* ``rte_pdump_enable()``:
+ This API enables the packet capture on a given port and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_enable_by_deviceid()``:
+ This API enables the packet capture on a given device id (``vdev name or pci address``) and queue.
+ Note: The filter option in the API is a place holder for future enhancements.
+
+* ``rte_pdump_disable()``:
+ This API disables the packet capture on a given port and queue.
+
+* ``rte_pdump_disable_by_deviceid()``:
+ This API disables the packet capture on a given device id (``vdev name or pci address``) and queue.
+
+* ``rte_pdump_uninit()``:
+ This API uninitializes the packet capture framework.
+
+
+Operation
+---------
+
+The ``librte_pdump`` library works on a client/server model. The server is responsible for enabling or
+disabling the packet capture and the clients are responsible for requesting the enabling or disabling of
+the packet capture.
+
+The packet capture framework, as part of its initialization, creates the pthread and the server socket in
+the pthread. The application that calls the framework initialization first will have the server socket created.
+Further calls to the framework initialization by the same application or other applications is not allowed i.e., only
+one server socket is allowed on the system. So the other applications can only request enabling or disabling of
+the packet capture at which point the client socket is created for them to send the request to the server.
+The server socket will listen for client requests for enabling or disabling the packet capture.
+
+
+Implementation Details
+----------------------
+
+The library API ``rte_pdump_init()``, initializes the packet capture framework by creating the pthread and the server
+socket. The server socket in the pthread context will be listening to the client requests to enable or disable the
+packet capture. Whoever calls this API first will have the server socket created, the subsequent calls to this APIs
+will not create any further server socket. i.e. only one server socket is allowed.
+
+The library APIs ``rte_pdump_enable()`` and ``rte_pdump_enable_by_deviceid()`` enables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump enable" request and sends
+the request to the server. The server that is listening on the socket will take the request and enable the packet capture
+by registering the Ethernet RX and TX callbacks for the given port or device_id and queue combinations.
+Then the server will mirror the packets to the new mempool and enqueue them to the rte_ring that clients have passed
+to these APIs. The server also sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, the client socket is closed.
+
+The library APIs ``rte_pdump_disable()`` and ``rte_pdump_disable_by_deviceid()`` disables the packet capture.
+On each call to these APIs, the library creates a separate client socket, creates the "pdump disable" request and sends
+the request to the server. The server that is listening on the socket will take the request and disable the packet
+capture by removing the Ethernet RX and TX callbacks for the given port or device_id and queue combinations. The server
+also sends the response back to the client about the status of the request that was processed. After the response is
+received from the server, the client socket is closed.
+
+The library API ``rte_pdump_uninit()``, uninitializes the packet capture framework by closing the pthread and the
+server socket.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+The DPDK ``app/pdump`` tool is developed based on this library to capture packets in DPDK.
+Users can use this as an example to develop their own packet capturing application.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 307e7c4..438b705 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -57,6 +57,11 @@ New Features
Enabled support for the AES CTR algorithm for Intel QuickAssist devices.
Provided support for algorithm-chaining operations.
+* **Added packet capture framework.**
+
+ * A new library ``librte_pdump`` is added to provide packet capture APIs.
+ * A new ``app/pdump`` tool is added to capture packets in DPDK.
+
Resolved Issues
---------------
@@ -126,6 +131,11 @@ API Changes
ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
+* Function ``rte_eth_dev_get_port_by_name`` changed to a public API.
+
+* Function ``rte_eth_dev_info_get`` updated to return new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ in the ``rte_eth_dev_info`` object.
+
ABI Changes
-----------
@@ -137,6 +147,9 @@ ABI Changes
* The ``rte_port_source_params`` structure has new fields to support PCAP file.
It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+* The ``rte_eth_dev_info`` structure has new fields ``nb_rx_queues`` and ``nb_tx_queues``
+ to support number of queues configured by software.
+
Shared Library Versions
-----------------------
diff --git a/doc/guides/sample_app_ug/index.rst b/doc/guides/sample_app_ug/index.rst
index 930f68c..96bb317 100644
--- a/doc/guides/sample_app_ug/index.rst
+++ b/doc/guides/sample_app_ug/index.rst
@@ -76,6 +76,7 @@ Sample Applications User Guide
ptpclient
performance_thread
ipsec_secgw
+ pdump
**Figures**
diff --git a/doc/guides/sample_app_ug/pdump.rst b/doc/guides/sample_app_ug/pdump.rst
new file mode 100644
index 0000000..96c8709
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,122 @@
+
+.. BSD LICENSE
+ Copyright(c) 2016 Intel Corporation. All rights reserved.
+ All rights reserved.
+
+ Redistribution and use in source and binary forms, with or without
+ modification, are permitted provided that the following conditions
+ are met:
+
+ * Redistributions of source code must retain the above copyright
+ notice, this list of conditions and the following disclaimer.
+ * Redistributions in binary form must reproduce the above copyright
+ notice, this list of conditions and the following disclaimer in
+ the documentation and/or other materials provided with the
+ distribution.
+ * Neither the name of Intel Corporation nor the names of its
+ contributors may be used to endorse or promote products derived
+ from this software without specific prior written permission.
+
+ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+dpdk_pdump Application
+======================
+
+The ``dpdk_pdump`` application is a Data Plane Development Kit (DPDK) application that runs as a DPDK secondary process and
+is capable of enabling packet capture on dpdk ports.
+
+
+Running the Application
+-----------------------
+
+The application has a ``--pdump`` command line option with various sub arguments:
+
+.. code-block:: console
+
+ ./build/app/dpdk_pdump --
+ --pdump '(port=<port id> | device_id=<pci id or vdev name>),
+ (queue=<queue_id>),
+ (rx-dev=<iface or pcap file> |
+ tx-dev=<iface or pcap file>),
+ [ring-size=<ring size>],
+ [mbuf-size=<mbuf data size>],
+ [total-num-mbufs=<number of mbufs>]'
+
+Note:
+
+* Parameters inside the parentheses represents mandatory parameters.
+
+* Parameters inside the square brackets represents optional parameters.
+
+Multiple instances of ``--pdump`` can be passed to capture packets on different port and queue combinations.
+
+
+Parameters
+~~~~~~~~~~
+
+``port``:
+Port id of the eth device on which packets should be captured.
+
+``device_id``:
+PCI address (or) name of the eth device on which packets should be captured.
+
+ .. Note::
+
+ * As of now the ``dpdk_pdump`` tool cannot capture the packets of virtual devices
+ in the primary process due to a bug in the ethdev library. Due to this bug, in a multi process context,
+ when the primary and secondary have different ports set, then the secondary process
+ (here the ``dpdk_pdump`` tool) overwrites the ``rte_eth_devices[]`` entries of the primary process.
+
+``queue``:
+Queue id of the eth device on which packets should be captured. The user can pass a queue value of ``*`` to enable
+packet capture on all queues of the eth device.
+
+``rx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+``tx-dev``:
+Can be either a pcap file name or any Linux iface.
+
+ .. Note::
+
+ * To receive ingress packets only, ``rx-dev`` should be passed.
+
+ * To receive egress packets only, ``tx-dev`` should be passed.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the different file names or the Linux iface names.
+
+ * To receive ingress and egress packets separately ``rx-dev`` and ``tx-dev``
+ should both be passed with the same file names or the the Linux iface names.
+
+``ring-size``:
+Size of the ring. This value is used internally for ring creation. The ring will be used to enqueue the packets from
+the primary application to the secondary. This is an optional parameter with default size 16384.
+
+``mbuf-size``:
+Size of the mbuf data. This is used internally for mempool creation. Ideally this value must be same as
+the primary application's mempool's mbuf data size which is used for packet RX. This is an optional parameter with
+default size 2176.
+
+``total-num-mbufs``:
+Total number mbufs in mempool. This is used internally for mempool creation. This is an optional parameter with default
+value 65535.
+
+
+Example
+-------
+
+.. code-block:: console
+
+ $ sudo ./build/app/dpdk_pdump -- --pdump 'port=0,queue=*,rx-dev=/tmp/rx.pcap'
--
2.5.0
^ permalink raw reply [relevance 5%]
* [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for rte_eth_dev_info structure
2016-06-08 13:38 3% ` [dpdk-dev] [PATCH v5 " Reshma Pattan
2016-06-08 13:38 5% ` [dpdk-dev] [PATCH v5 8/9] doc: update doc for " Reshma Pattan
@ 2016-06-08 13:38 9% ` Reshma Pattan
2016-06-08 16:15 4% ` Mcnamara, John
2016-06-09 8:50 2% ` [dpdk-dev] [PATCH v6 0/8] add packet capture framework Reshma Pattan
2 siblings, 1 reply; 200+ results
From: Reshma Pattan @ 2016-06-08 13:38 UTC (permalink / raw)
To: dev; +Cc: Reshma Pattan
New fields nb_rx_queues and nb_tx_queues will be added to
rte_eth_dev_info structure.
Changes to API rte_eth_dev_info_get() will be done to update
these new fields to rte_eth_dev_info object.
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
---
doc/guides/rel_notes/deprecation.rst | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index ad05eba..04316fb 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -57,3 +57,9 @@ Deprecation Notices
a handle, like the way kernel exposes an fd to user for locating a
specific file, and to keep all major structures internally, so that
we are likely to be free from ABI violations in future.
+
+* A librte_ether public structure ``rte_eth_dev_info`` will be changed in 16.07.
+ The proposed change will add new parameters ``nb_rx_queues``, ``nb_tx_queues``
+ to the structure. These are the number of queues configured by software.
+ Modification to definition of ``rte_eth_dev_info_get()`` will be done
+ to update new parameters to ``rte_eth_dev_info`` object.
--
2.5.0
^ permalink raw reply [relevance 9%]
* [dpdk-dev] [PATCH v5 0/9] add packet capture framework
@ 2016-06-08 13:38 3% ` Reshma Pattan
2016-06-08 13:38 5% ` [dpdk-dev] [PATCH v5 8/9] doc: update doc for " Reshma Pattan
` (2 more replies)
0 siblings, 3 replies; 200+ results
From: Reshma Pattan @ 2016-06-08 13:38 UTC (permalink / raw)
To: dev
This patch set include below changes
1)Changes to librte_ether.
2)A new library librte_pdump added for packet capture framework.
3)A new app/pdump tool added for packet capturing.
4)Test pmd changes done to initialize packet capture framework.
5)Documentation update.
1)librte_pdump
==============
To support packet capturing on dpdk Ethernet devices, a new library librte_pdump
is added.Users can develop their own packet capturing application using new library APIs.
Operation:
----------
Pdump library provides APIs to support packet capturing on dpdk Ethernet devices.
Library provides APIs to initialize the packet capture framework, enable/disable
the packet capture and uninitialize the packet capture framework.
Pdump library works on client/server based model.
Sever is responsible for enabling/disabling the packet captures.
Clients are responsible for requesting enable/disable of the
packet captures.
As part of packet capture framework initialization, pthread and
the server socket is created. Only one server socket is allowed on the system.
As part of enabling/disabling the packet capture, client sockets are created
and multiple client sockets are allowed.
Who ever calls initialization first they will succeed with the initialization,
next subsequent calls of initialization are not allowed. So next users can only
request enabling/disabling the packet capture.
Applications using below APIs need to pass port/device_id, queue, mempool and
ring parameters. Library uses user provided ring and mempool to mirror the rx/tx
packets of the port for users. Users need to dequeue the rings and write the packets
to vdev(pcap/tuntap) to view the packets using any standard tools.
Note:
Mempool and Ring should be mc/mp supportable.
Mempool mbuf size should be big enough to handle the rx/tx packets of a port.
APIs:
-----
rte_pdump_init()
rte_pdump_enable()
rte_pdump_enable_by_deviceid()
rte_pdump_disable()
rte_pdump_disable_by_deviceid()
rte_pdump_uninit()
2)app/pdump tool
================
Tool app/pdump is designed based on librte_pdump for packet capturing in DPDK.
This tool by default runs as secondary process, and provides the support for
the command line options for packet capture.
./build/app/dpdk_pdump --
--pdump '(port=<port id> | device_id=<pci id or vdev name>),
(queue=<queue id>),
(rx-dev=<iface or pcap file> |
tx-dev=<iface or pcap file>),
[ring-size=<ring size>],
[mbuf-size=<mbuf data size>],
[total-num-mbufs=<number of mbufs>]'
Parameters inside the parenthesis represents the mandatory parameters.
Parameters inside the square brackets represents optional parameters.
User has to pass on packet capture parameters under --pdump parameters, multiples of
--pdump can be passed to capture packets on different port and queue combinations
Operation:
----------
*Tool parse the user command line arguments,
creates the mempool, ring and the PCAP PMD vdev with 'tx_stream' as either
of the device passed in rx-dev|tx-dev parameters.
*Then calls the APIs of librte_pdump i.e. rte_pdump_enable()/rte_pdump_enable_by_deviceid()
to enable packet capturing on a specific port/device_id and queue by passing on
port|device_id, queue, mempool and ring info.
*Tool runs in while loop to dequeue the packets from the ring and write them to pcap device.
*Tool can be stopped using SIGINT, upon which tool calls
rte_pdump_disable()/rte_pdump_disable_by_deviceid() and free the allocated resources.
Note:
CONFIG_RTE_LIBRTE_PMD_PCAP flag should be set to yes to compile and run the pdump tool.
3)Test-pmd changes
==================
Changes are done to test-pmd application to initialize/uninitialize the packet capture framework.
So app/pdump tool can be run to see packets of dpdk ports that are used by test-pmd.
Similarly any application which needs packet capture should call initialize/uninitialize APIs of
librte_pdump and use pdump tool to start the capture.
4)Packet capture flow between pdump tool and librte_pdump
=========================================================
* Pdump tool (Secondary process) requests packet capture
for specific port|device_id and queue combinations.
*Library in secondary process context creates client socket and communicates
the port|device_id, queue, ring and mempool to server.
*Library initializes server in primary process 'test-pmd' context and server serves
the client request to enable Ethernet rxtx call-backs for a given port|device_id and queue.
*Copy the rx/tx packets to passed mempool and enqueue the packets to ring for secondary process.
*Pdump tool will dequeue the packets from ring and writes them to PCAPMD vdev,
so ultimately packets will be seen on the device that is passed in rx-dev|tx-dev.
*Once the pdump tool is terminated with SIGINT it will disable the packet capturing.
*Library receives the disable packet capture request, communicate the info to server,
server will remove the Ethernet rxtx call-backs.
*Packet capture can be seen using tcpdump command
"tcpdump -ni <iface>" (or) "tcpdump –nr <pcapfile>"
5)Example command line
======================
./build/app/dpdk_pdump -- --pdump 'device_id=0000:02:0.0,queue=*,tx-dev=/tmp/dt-file.pcap,rx-dev=/tmp/dr-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=32768' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/d-file.pcap,tx-dev=/tmp/d-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
v5:
addressed code review comments for below patches
http://dpdk.org/dev/patchwork/patch/12955/
http://dpdk.org/dev/patchwork/patch/12951/
v4:
added missing deprecation notice for ABI changes of rte_eth_dev_info structure.
made doc changes as per doc guidelines.
replaced rte_eal_vdev_init with rte_eth_dev_attach in pdump tool.
removed rxtx-dev parameter from pdump tool command line.
v3:
app/pdump: Moved cleanup code from signal handler to main.
divided librte_ether changes into multiple patches.
example command changed in app/pdump application guide
v2:
fix compilation issues for 4.8.3
fix unnecessary #includes
Reshma Pattan (9):
librte_ether: protect add/remove of rxtx callbacks with spinlocks
librte_ether: add new api rte_eth_add_first_rx_callback
librte_ether: add new fields to rte_eth_dev_info struct
librte_ether: make rte_eth_dev_get_port_by_name
rte_eth_dev_get_name_by_port public
lib/librte_pdump: add new library for packet capturing support
app/pdump: add pdump tool for packet capturing
app/test-pmd: add pdump initialization uninitialization
doc: update doc for packet capture framework
doc: announce ABI change for rte_eth_dev_info structure
MAINTAINERS | 8 +
app/Makefile | 1 +
app/pdump/Makefile | 45 ++
app/pdump/main.c | 814 +++++++++++++++++++++++++++++++
app/test-pmd/testpmd.c | 6 +
config/common_base | 5 +
doc/guides/prog_guide/index.rst | 1 +
doc/guides/prog_guide/pdump_library.rst | 107 ++++
doc/guides/rel_notes/deprecation.rst | 6 +
doc/guides/rel_notes/release_16_07.rst | 13 +
doc/guides/sample_app_ug/index.rst | 1 +
doc/guides/sample_app_ug/pdump.rst | 122 +++++
lib/Makefile | 1 +
lib/librte_ether/rte_ethdev.c | 123 +++--
lib/librte_ether/rte_ethdev.h | 59 +++
lib/librte_ether/rte_ether_version.map | 9 +
lib/librte_pdump/Makefile | 55 +++
lib/librte_pdump/rte_pdump.c | 841 ++++++++++++++++++++++++++++++++
lib/librte_pdump/rte_pdump.h | 186 +++++++
lib/librte_pdump/rte_pdump_version.map | 12 +
mk/rte.app.mk | 1 +
21 files changed, 2372 insertions(+), 44 deletions(-)
create mode 100644 app/pdump/Makefile
create mode 100644 app/pdump/main.c
create mode 100644 doc/guides/prog_guide/pdump_library.rst
create mode 100644 doc/guides/sample_app_ug/pdump.rst
create mode 100644 lib/librte_pdump/Makefile
create mode 100644 lib/librte_pdump/rte_pdump.c
create mode 100644 lib/librte_pdump/rte_pdump.h
create mode 100644 lib/librte_pdump/rte_pdump_version.map
--
2.5.0
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v3 01/10] rte: change xstats to use integer ids
2016-06-08 11:16 3% ` Remy Horton
@ 2016-06-08 12:22 0% ` Thomas Monjalon
0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-06-08 12:22 UTC (permalink / raw)
To: Remy Horton; +Cc: dev
2016-06-08 12:16, Remy Horton:
> 'noon,
>
> On 08/06/2016 10:37, Thomas Monjalon wrote:
> > 2016-05-30 11:48, Remy Horton:
> >> struct rte_eth_xstats {
> >> + /* FIXME: Remove name[] once remaining drivers converted */
> >> char name[RTE_ETH_XSTATS_NAME_SIZE];
> >
> > What is the plan? This field must be deprecated with an attribute.
> > We cannot have 2 different APIs depending of the driver.
>
> This is where it gets logistically tricky..
>
> Since there's an API/ABI breakage notice in place on this, my own
> preference would be to have the entire patchset quashed into a single
> patch. Problem is that rte/app changes (patches 1 & 7-9) are normally
> applied via master whereas driver changes (patches 2-6) go in via
> dpdk-next-net - it is not clear to me how patches should be submitted
> for this case..
Misunderstanding here. Patches are fine and will be integrated in the
main tree because they are not only some drivers changes.
I was talking about the old API with name in rte_eth_xstats.
I have not seen the patch 9 which removes it.
> >> +int rte_eth_xstats_count(uint8_t port_id);
> >
> > This function is useless because we can have the count with
> > rte_eth_xstats_get(p, NULL, 0)
> > By the way it would be more consistent to have the same behaviour
> > in rte_eth_xstats_names().
>
> Feedback I got with earlier patches was that a seperate count function
> was preferable to overloading the fetch function using *data==NULL - is
> the use of the latter specifically preferred?
I prefer the fetch/NULL style to get a count.
It also handles nicely the fetch error because of a too small buffer.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v3 01/10] rte: change xstats to use integer ids
@ 2016-06-08 11:16 3% ` Remy Horton
2016-06-08 12:22 0% ` Thomas Monjalon
0 siblings, 1 reply; 200+ results
From: Remy Horton @ 2016-06-08 11:16 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev
'noon,
On 08/06/2016 10:37, Thomas Monjalon wrote:
> 2016-05-30 11:48, Remy Horton:
>> struct rte_eth_xstats {
>> + /* FIXME: Remove name[] once remaining drivers converted */
>> char name[RTE_ETH_XSTATS_NAME_SIZE];
>
> What is the plan? This field must be deprecated with an attribute.
> We cannot have 2 different APIs depending of the driver.
This is where it gets logistically tricky..
Since there's an API/ABI breakage notice in place on this, my own
preference would be to have the entire patchset quashed into a single
patch. Problem is that rte/app changes (patches 1 & 7-9) are normally
applied via master whereas driver changes (patches 2-6) go in via
dpdk-next-net - it is not clear to me how patches should be submitted
for this case..
> What are the remaining drivers to convert?
Opps, none. All relevant drivers are converted..
> This structure and the other one (rte_eth_xstats) are badly named.
> There is only one stat in each. So they should not have the plural form.
> rte_eth_xstat and rte_eth_xstat_name would be better.
I kept rte_eth_xstats as it was the name already in use within DPDK.
Will change the other.
>> +int rte_eth_xstats_count(uint8_t port_id);
>
> This function is useless because we can have the count with
> rte_eth_xstats_get(p, NULL, 0)
> By the way it would be more consistent to have the same behaviour
> in rte_eth_xstats_names().
Feedback I got with earlier patches was that a seperate count function
was preferable to overloading the fetch function using *data==NULL - is
the use of the latter specifically preferred?
Other comments noted.
..Remy
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-08 7:24 3% ` Lu, Wenzhuo
@ 2016-06-08 9:19 0% ` Ananyev, Konstantin
2016-06-12 2:00 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-06-08 9:19 UTC (permalink / raw)
To: Lu, Wenzhuo, Tao, Zhe, dev
Cc: Richardson, Bruce, Chen, Jing D, Liang, Cunming, Wu, Jingjing,
Zhang, Helin
>
> Hi Konstantin,
>
>
> > -----Original Message-----
> > From: Ananyev, Konstantin
> > Sent: Tuesday, June 7, 2016 5:59 PM
> > To: Tao, Zhe; dev@dpdk.org
> > Cc: Lu, Wenzhuo; Richardson, Bruce; Chen, Jing D; Liang, Cunming; Wu, Jingjing;
> > Zhang, Helin
> > Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
> >
> >
> > Hi Zhe & Wenzhuo,
> >
> > Please find my comments below.
> > BTW, for clarification - is that patch for 16.11?
> > I believe it's too late to introduce such significant change in 16.07.
> > Thanks
> > Konstantin
> Thanks for the comments.
> Honestly, our purpose is 16.07. Realizing the big impact, we use NEXT_ABI to comment our change. So, I think although we want to
> merge it in 16.07 this change will become effective after we remove NEXT_ABI in 16.11.
I don't think it is achievable.
First I think your code is not in proper shape yet, right now.
Second, as you said, it is a significant change and I would like to hear opinions from the rest of the community.
>
> >
> > > Define lock mode for RX/TX queue. Because when resetting the device we
> > > want the resetting thread to get the lock of the RX/TX queue to make
> > > sure the RX/TX is stopped.
> > >
> > > Using next ABI macro for this ABI change as it has too much impact. 7
> > > APIs and 1 global variable are impacted.
> > >
> > > Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> > > Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> > > ---
> > > lib/librte_ether/rte_ethdev.h | 62
> > > +++++++++++++++++++++++++++++++++++++++++++
> > > 1 file changed, 62 insertions(+)
> > >
> > > diff --git a/lib/librte_ether/rte_ethdev.h
> > > b/lib/librte_ether/rte_ethdev.h index 74e895f..4efb5e9 100644
> > > --- a/lib/librte_ether/rte_ethdev.h
> > > +++ b/lib/librte_ether/rte_ethdev.h
> > > @@ -354,7 +354,12 @@ struct rte_eth_rxmode {
> > > jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
> > > hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
> > > enable_scatter : 1, /**< Enable scatter packets rx handler */
> > > +#ifndef RTE_NEXT_ABI
> > > enable_lro : 1; /**< Enable LRO */
> > > +#else
> > > + enable_lro : 1, /**< Enable LRO */
> > > + lock_mode : 1; /**< Using lock path */
> > > +#endif
> > > };
> > >
> > > /**
> > > @@ -634,11 +639,68 @@ struct rte_eth_txmode {
> > > /**< If set, reject sending out tagged pkts */
> > > hw_vlan_reject_untagged : 1,
> > > /**< If set, reject sending out untagged pkts */
> > > +#ifndef RTE_NEXT_ABI
> > > hw_vlan_insert_pvid : 1;
> > > /**< If set, enable port based VLAN insertion */
> > > +#else
> > > + hw_vlan_insert_pvid : 1,
> > > + /**< If set, enable port based VLAN insertion */
> > > + lock_mode : 1;
> > > + /**< If set, using lock path */
> > > +#endif
> > > };
> > >
> > > /**
> > > + * The macros for the RX/TX lock mode functions */ #ifdef
> > > +RTE_NEXT_ABI #define RX_LOCK_FUNCTION(dev, func) \
> > > + (dev->data->dev_conf.rxmode.lock_mode ? \
> > > + func ## _lock : func)
> > > +
> > > +#define TX_LOCK_FUNCTION(dev, func) \
> > > + (dev->data->dev_conf.txmode.lock_mode ? \
> > > + func ## _lock : func)
> > > +#else
> > > +#define RX_LOCK_FUNCTION(dev, func) func
> > > +
> > > +#define TX_LOCK_FUNCTION(dev, func) func #endif
> > > +
> > > +/* Add the lock RX/TX function for VF reset */ #define
> > > +GENERATE_RX_LOCK(func, nic) \ uint16_t func ## _lock(void *rx_queue,
> > > +\
> > > + struct rte_mbuf **rx_pkts, \
> > > + uint16_t nb_pkts) \
> > > +{ \
> > > + struct nic ## _rx_queue *rxq = rx_queue; \
> > > + uint16_t nb_rx = 0; \
> > > + \
> > > + if (rte_spinlock_trylock(&rxq->rx_lock)) { \
> > > + nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
> > > + rte_spinlock_unlock(&rxq->rx_lock); \
> > > + } \
> > > + \
> > > + return nb_rx; \
> > > +}
> > > +
> > > +#define GENERATE_TX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > > +*tx_queue, \
> > > + struct rte_mbuf **tx_pkts, \
> > > + uint16_t nb_pkts) \
> > > +{ \
> > > + struct nic ## _tx_queue *txq = tx_queue; \
> > > + uint16_t nb_tx = 0; \
> > > + \
> > > + if (rte_spinlock_trylock(&txq->tx_lock)) { \
> > > + nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
> > > + rte_spinlock_unlock(&txq->tx_lock); \
> > > + } \
> > > + \
> > > + return nb_tx; \
> > > +}
> >
> > 1. As I said in off-line dicussiion, I think this locking could (and I think better be)
> > impelented completely on rte_ethdev layer.
> > So actual PMD code will be unaffected.
> > Again that avoids us to introduce _lock version of every RX/Tx function in each
> > PMD.
> One purpose of implementing the lock in PMD layer is to avoid ABI change. But we introduce the field lock_mode in struct
> rte_eth_rx/txmode. So seems it's not a good reason now :)
> The other purpose is we want to add a lock for every queue. But in rte layer the queue is void *, so we add the lock in the specific
> structures of the NICs. But as you mentioned below, we can add the lock as dev->data->rx_queue_state it the struct
> rte_eth_dev_data.
> So, I prefer to add the lock in rte layer now.
OK.
>
> >
> > 2. Again, as discussed offline, I think it is better to have an explicit
> > rte_eth_(rx|tx)_burst_lock(sync?) API, instead of add new fileds into RX/TX
> > config strcutures.
> > Would help to avoid any confusion, I think.
> We want the users to choose the rx/tx path without lock if they're sensitive to the performance and can handle the reset event in
> their APP. After introducing new fields of config struct, users can change the config to choose the different path.
I understand what you are doing.
> If we introduce new API, it may be harder for the use to use it. I mean when users want to use lock mode, they may need to replace
> all the rte_eth_rx/tx_burst by rte_eth_rx/tx_burst_lock.
Yes, my opinion if users would like to use locking API they need to call it explicitly.
>So if we add the lock in rte layer, I still prefer adding lock_mode in the
> configuration, and the rte_eth_rx/tx_burst is changed like this,
> rte_eth_rx/tx_burst
> {
> + if lock_mode
> + try_lock
> ......
> + if lock_mode
> + release_lock
> }
My preference is to keep existing rx/tx_burst() functions unaffected by that patch.
At least for now.
I suppose that will minimise the risks and help users to avoid confusion what API
(locking/non-locking) is in use.
>
>
> >
> > 3. I thought the plan was to introduce a locking in all appropriate control path
> > functions (dev_start/dev_stop etc.) Without that locking version of RX/TX seems
> > a bit useless.
> > Yes, I understand that you do use locking inside dev_reset, but I suppose the
> > plan was to have a generic solution, no?
> > Again, interrupt fire when user invokes dev_start/stop or so, so we still need
> > some synchronisation between them.
> >
> > To be more specific, I thought about something like that:
> >
> > static inline uint16_t
> > rte_eth_rx_burst_lock(uint8_t port_id, uint16_t queue_id,
> > struct rte_mbuf **rx_pkts, const uint16_t nb_pkts) {
> > struct rte_eth_dev *dev = &rte_eth_devices[port_id];
> >
> > #ifdef RTE_LIBRTE_ETHDEV_DEBUG
> > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
> > RTE_FUNC_PTR_OR_ERR_RET(*dev->rx_pkt_burst, 0);
> >
> > if (queue_id >= dev->data->nb_rx_queues) {
> > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", queue_id);
> > return 0;
> > }
> > #endif
> >
> > + if (rte_spinlock_trylock(&dev->data->rx_queue_state[rx_queue_id].lock) == 0)
> > + return 0;
> > + else if (dev->data->rx_queue_state[rx_queue_id] ==
> > RTE_ETH_QUEUE_STATE_STOPPED)) {
> > + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> > + return 0;
> > +
> >
> > nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
> > rx_pkts, nb_pkts);
> >
> > + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> >
> > ....
> >
> > return nb_rx;
> > }
> >
> > And inside queue_start:
> >
> > int
> > rte_eth_dev_rx_queue_start(uint8_t port_id, uint16_t rx_queue_id) {
> > struct rte_eth_dev *dev;
> >
> > RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
> >
> > dev = &rte_eth_devices[port_id];
> > if (rx_queue_id >= dev->data->nb_rx_queues) {
> > RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", rx_queue_id);
> > return -EINVAL;
> > }
> >
> > RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->rx_queue_start, -ENOTSUP);
> >
> > rte_spinlock_lock(&dev->data->rx_queue_state[rx_queue_id].lock)
> I think you add the lock here to stop the rx/tx.
> But to my opinion, we should lock the rx/tx much earlier before starting the queue. For example, when stop the port, the resource of
> the queues may be released.
I didn't get you here...
Before releasing the queue resources, queue_stop() has to be executed, right?
>The rx/tx cannot be executed. So I prefer to get the lock before stopping the ports.
Might be I wasn't clear enough here.
What I think we need to have:
-To stop/start/rx/tx the queue (or do any other action that might change the queue internal structure)
you have to grab the lock.
After queue is stopped it's state has to be changed to QUEUE_STATE_STOPPED (whti queue lock grabbed),
so rx/tx_locked wouldn't proceed with that queue.
- dev_stop() - has to stop all its queues first, i.e. it needs to call queue_stop() for all of them.
So after dev_stop() had finished - all device queues have to be in QUEUE_STATE_STOPPED
Same about dev_start() - after it does all other things - it will call queue_start() for all it's queues.
that will bring them into QUEUE_STARTED.
After that rx/tx_locked can use them again.
>Maybe better to keep the spinlock in the dev_reset.
Might be not :)
>
> >
> > if (dev->data->rx_queue_state[rx_queue_id] !=
> > RTE_ETH_QUEUE_STATE_STOPPED) {
> > RTE_PMD_DEBUG_TRACE("Queue %" PRIu16" of device with
> > port_id=%" PRIu8
> > " already started\n",
> > rx_queue_id, port_id);
> > ret = -EINVAL 0;
> > } else
> > ret = dev->dev_ops->rx_queue_start(dev, rx_queue_id);
> >
> > rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> >
> > return ret;
> > }
> >
> > Then again, we don't need to do explicit locking inside dev_reset().
> > Does it make sense to you guys?
> Please see the answer above.
>
> >
> >
> > > +
> > > +/**
> > > * A structure used to configure an RX ring of an Ethernet port.
> > > */
> > > struct rte_eth_rxconf {
> > > --
> > > 2.1.4
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-08 2:15 0% ` Stephen Hemminger
@ 2016-06-08 7:34 0% ` Lu, Wenzhuo
2016-06-09 7:50 0% ` Olivier Matz
0 siblings, 1 reply; 200+ results
From: Lu, Wenzhuo @ 2016-06-08 7:34 UTC (permalink / raw)
To: Stephen Hemminger; +Cc: dev, Tao, Zhe
Hi Stephen,
> -----Original Message-----
> From: Stephen Hemminger [mailto:stephen@networkplumber.org]
> Sent: Wednesday, June 8, 2016 10:16 AM
> To: Lu, Wenzhuo
> Cc: dev@dpdk.org; Tao, Zhe
> Subject: Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
>
> On Mon, 6 Jun 2016 13:40:47 +0800
> Wenzhuo Lu <wenzhuo.lu@intel.com> wrote:
>
> > Define lock mode for RX/TX queue. Because when resetting the device we
> > want the resetting thread to get the lock of the RX/TX queue to make
> > sure the RX/TX is stopped.
> >
> > Using next ABI macro for this ABI change as it has too much impact. 7
> > APIs and 1 global variable are impacted.
> >
> > Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> > Signed-off-by: Zhe Tao <zhe.tao@intel.com>
>
> Why does this patch set make a different assumption the rest of the DPDK?
>
> The rest of the DPDK operates on the principle that the application is smart
> enough to stop the device before making changes. There is no equivalent to the
> Linux kernel RTNL mutex. The API assumes application threads are well behaved
> and will not try and sabotage each other.
>
> If you restrict the reset operation to only being available when RX/TX is stopped,
> then no lock is needed.
>
> The fact that it requires lots more locking inside each device driver implies to me
> this is not correct way to architect this.
It's a good question. This patch set doesn't follow the regular assumption of DPDK.
But it's a requirement we've got from some customers. The users want the driver does as much as it can. The best is the link state change is transparent to the users.
The patch set tries to provide another choice if the users don't want to stop their rx/tx to handle the reset event.
And as discussed in the other thread, most probably we will move the lock from the PMD layer to rte lay. It'll avoid the change in every device.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-07 9:58 0% ` Ananyev, Konstantin
@ 2016-06-08 7:24 3% ` Lu, Wenzhuo
2016-06-08 9:19 0% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Lu, Wenzhuo @ 2016-06-08 7:24 UTC (permalink / raw)
To: Ananyev, Konstantin, Tao, Zhe, dev
Cc: Richardson, Bruce, Chen, Jing D, Liang, Cunming, Wu, Jingjing,
Zhang, Helin
Hi Konstantin,
> -----Original Message-----
> From: Ananyev, Konstantin
> Sent: Tuesday, June 7, 2016 5:59 PM
> To: Tao, Zhe; dev@dpdk.org
> Cc: Lu, Wenzhuo; Richardson, Bruce; Chen, Jing D; Liang, Cunming; Wu, Jingjing;
> Zhang, Helin
> Subject: RE: [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
>
>
> Hi Zhe & Wenzhuo,
>
> Please find my comments below.
> BTW, for clarification - is that patch for 16.11?
> I believe it's too late to introduce such significant change in 16.07.
> Thanks
> Konstantin
Thanks for the comments.
Honestly, our purpose is 16.07. Realizing the big impact, we use NEXT_ABI to comment our change. So, I think although we want to merge it in 16.07 this change will become effective after we remove NEXT_ABI in 16.11.
>
> > Define lock mode for RX/TX queue. Because when resetting the device we
> > want the resetting thread to get the lock of the RX/TX queue to make
> > sure the RX/TX is stopped.
> >
> > Using next ABI macro for this ABI change as it has too much impact. 7
> > APIs and 1 global variable are impacted.
> >
> > Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> > Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> > ---
> > lib/librte_ether/rte_ethdev.h | 62
> > +++++++++++++++++++++++++++++++++++++++++++
> > 1 file changed, 62 insertions(+)
> >
> > diff --git a/lib/librte_ether/rte_ethdev.h
> > b/lib/librte_ether/rte_ethdev.h index 74e895f..4efb5e9 100644
> > --- a/lib/librte_ether/rte_ethdev.h
> > +++ b/lib/librte_ether/rte_ethdev.h
> > @@ -354,7 +354,12 @@ struct rte_eth_rxmode {
> > jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
> > hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
> > enable_scatter : 1, /**< Enable scatter packets rx handler */
> > +#ifndef RTE_NEXT_ABI
> > enable_lro : 1; /**< Enable LRO */
> > +#else
> > + enable_lro : 1, /**< Enable LRO */
> > + lock_mode : 1; /**< Using lock path */
> > +#endif
> > };
> >
> > /**
> > @@ -634,11 +639,68 @@ struct rte_eth_txmode {
> > /**< If set, reject sending out tagged pkts */
> > hw_vlan_reject_untagged : 1,
> > /**< If set, reject sending out untagged pkts */
> > +#ifndef RTE_NEXT_ABI
> > hw_vlan_insert_pvid : 1;
> > /**< If set, enable port based VLAN insertion */
> > +#else
> > + hw_vlan_insert_pvid : 1,
> > + /**< If set, enable port based VLAN insertion */
> > + lock_mode : 1;
> > + /**< If set, using lock path */
> > +#endif
> > };
> >
> > /**
> > + * The macros for the RX/TX lock mode functions */ #ifdef
> > +RTE_NEXT_ABI #define RX_LOCK_FUNCTION(dev, func) \
> > + (dev->data->dev_conf.rxmode.lock_mode ? \
> > + func ## _lock : func)
> > +
> > +#define TX_LOCK_FUNCTION(dev, func) \
> > + (dev->data->dev_conf.txmode.lock_mode ? \
> > + func ## _lock : func)
> > +#else
> > +#define RX_LOCK_FUNCTION(dev, func) func
> > +
> > +#define TX_LOCK_FUNCTION(dev, func) func #endif
> > +
> > +/* Add the lock RX/TX function for VF reset */ #define
> > +GENERATE_RX_LOCK(func, nic) \ uint16_t func ## _lock(void *rx_queue,
> > +\
> > + struct rte_mbuf **rx_pkts, \
> > + uint16_t nb_pkts) \
> > +{ \
> > + struct nic ## _rx_queue *rxq = rx_queue; \
> > + uint16_t nb_rx = 0; \
> > + \
> > + if (rte_spinlock_trylock(&rxq->rx_lock)) { \
> > + nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
> > + rte_spinlock_unlock(&rxq->rx_lock); \
> > + } \
> > + \
> > + return nb_rx; \
> > +}
> > +
> > +#define GENERATE_TX_LOCK(func, nic) \ uint16_t func ## _lock(void
> > +*tx_queue, \
> > + struct rte_mbuf **tx_pkts, \
> > + uint16_t nb_pkts) \
> > +{ \
> > + struct nic ## _tx_queue *txq = tx_queue; \
> > + uint16_t nb_tx = 0; \
> > + \
> > + if (rte_spinlock_trylock(&txq->tx_lock)) { \
> > + nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
> > + rte_spinlock_unlock(&txq->tx_lock); \
> > + } \
> > + \
> > + return nb_tx; \
> > +}
>
> 1. As I said in off-line dicussiion, I think this locking could (and I think better be)
> impelented completely on rte_ethdev layer.
> So actual PMD code will be unaffected.
> Again that avoids us to introduce _lock version of every RX/Tx function in each
> PMD.
One purpose of implementing the lock in PMD layer is to avoid ABI change. But we introduce the field lock_mode in struct rte_eth_rx/txmode. So seems it's not a good reason now :)
The other purpose is we want to add a lock for every queue. But in rte layer the queue is void *, so we add the lock in the specific structures of the NICs. But as you mentioned below, we can add the lock as dev->data->rx_queue_state it the struct rte_eth_dev_data.
So, I prefer to add the lock in rte layer now.
>
> 2. Again, as discussed offline, I think it is better to have an explicit
> rte_eth_(rx|tx)_burst_lock(sync?) API, instead of add new fileds into RX/TX
> config strcutures.
> Would help to avoid any confusion, I think.
We want the users to choose the rx/tx path without lock if they're sensitive to the performance and can handle the reset event in their APP. After introducing new fields of config struct, users can change the config to choose the different path.
If we introduce new API, it may be harder for the use to use it. I mean when users want to use lock mode, they may need to replace all the rte_eth_rx/tx_burst by rte_eth_rx/tx_burst_lock. So if we add the lock in rte layer, I still prefer adding lock_mode in the configuration, and the rte_eth_rx/tx_burst is changed like this,
rte_eth_rx/tx_burst
{
+ if lock_mode
+ try_lock
......
+ if lock_mode
+ release_lock
}
>
> 3. I thought the plan was to introduce a locking in all appropriate control path
> functions (dev_start/dev_stop etc.) Without that locking version of RX/TX seems
> a bit useless.
> Yes, I understand that you do use locking inside dev_reset, but I suppose the
> plan was to have a generic solution, no?
> Again, interrupt fire when user invokes dev_start/stop or so, so we still need
> some synchronisation between them.
>
> To be more specific, I thought about something like that:
>
> static inline uint16_t
> rte_eth_rx_burst_lock(uint8_t port_id, uint16_t queue_id,
> struct rte_mbuf **rx_pkts, const uint16_t nb_pkts) {
> struct rte_eth_dev *dev = &rte_eth_devices[port_id];
>
> #ifdef RTE_LIBRTE_ETHDEV_DEBUG
> RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
> RTE_FUNC_PTR_OR_ERR_RET(*dev->rx_pkt_burst, 0);
>
> if (queue_id >= dev->data->nb_rx_queues) {
> RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", queue_id);
> return 0;
> }
> #endif
>
> + if (rte_spinlock_trylock(&dev->data->rx_queue_state[rx_queue_id].lock) == 0)
> + return 0;
> + else if (dev->data->rx_queue_state[rx_queue_id] ==
> RTE_ETH_QUEUE_STATE_STOPPED)) {
> + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
> + return 0;
> +
>
> nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
> rx_pkts, nb_pkts);
>
> + rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
>
> ....
>
> return nb_rx;
> }
>
> And inside queue_start:
>
> int
> rte_eth_dev_rx_queue_start(uint8_t port_id, uint16_t rx_queue_id) {
> struct rte_eth_dev *dev;
>
> RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
>
> dev = &rte_eth_devices[port_id];
> if (rx_queue_id >= dev->data->nb_rx_queues) {
> RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", rx_queue_id);
> return -EINVAL;
> }
>
> RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->rx_queue_start, -ENOTSUP);
>
> rte_spinlock_lock(&dev->data->rx_queue_state[rx_queue_id].lock)
I think you add the lock here to stop the rx/tx.
But to my opinion, we should lock the rx/tx much earlier before starting the queue. For example, when stop the port, the resource of the queues may be released. The rx/tx cannot be executed. So I prefer to get the lock before stopping the ports. Maybe better to keep the spinlock in the dev_reset.
>
> if (dev->data->rx_queue_state[rx_queue_id] !=
> RTE_ETH_QUEUE_STATE_STOPPED) {
> RTE_PMD_DEBUG_TRACE("Queue %" PRIu16" of device with
> port_id=%" PRIu8
> " already started\n",
> rx_queue_id, port_id);
> ret = -EINVAL 0;
> } else
> ret = dev->dev_ops->rx_queue_start(dev, rx_queue_id);
>
> rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
>
> return ret;
> }
>
> Then again, we don't need to do explicit locking inside dev_reset().
> Does it make sense to you guys?
Please see the answer above.
>
>
> > +
> > +/**
> > * A structure used to configure an RX ring of an Ethernet port.
> > */
> > struct rte_eth_rxconf {
> > --
> > 2.1.4
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode
@ 2016-06-08 2:15 0% ` Stephen Hemminger
2016-06-08 7:34 0% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Stephen Hemminger @ 2016-06-08 2:15 UTC (permalink / raw)
To: Wenzhuo Lu; +Cc: dev, Zhe Tao
On Mon, 6 Jun 2016 13:40:47 +0800
Wenzhuo Lu <wenzhuo.lu@intel.com> wrote:
> Define lock mode for RX/TX queue. Because when resetting
> the device we want the resetting thread to get the lock
> of the RX/TX queue to make sure the RX/TX is stopped.
>
> Using next ABI macro for this ABI change as it has too
> much impact. 7 APIs and 1 global variable are impacted.
>
> Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> Signed-off-by: Zhe Tao <zhe.tao@intel.com>
Why does this patch set make a different assumption the rest of the DPDK?
The rest of the DPDK operates on the principle that the application
is smart enough to stop the device before making changes. There is no
equivalent to the Linux kernel RTNL mutex. The API assumes application
threads are well behaved and will not try and sabotage each other.
If you restrict the reset operation to only being available when RX/TX is stopped,
then no lock is needed.
The fact that it requires lots more locking inside each device driver implies
to me this is not correct way to architect this.
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] RFC: DPDK Long Term Support
2016-06-07 12:36 3% ` Christian Ehrhardt
@ 2016-06-07 19:39 0% ` Martinx - ジェームズ
0 siblings, 0 replies; 200+ results
From: Martinx - ジェームズ @ 2016-06-07 19:39 UTC (permalink / raw)
To: Christian Ehrhardt; +Cc: dev
On 7 June 2016 at 08:36, Christian Ehrhardt <
christian.ehrhardt@canonical.com> wrote:
> On Fri, Jun 3, 2016 at 5:07 PM, Mcnamara, John <john.mcnamara@intel.com>
> wrote:
> [...]
> >
> > LTS Version
> > ------------
> >
> > The proposed initial LTS version will be DPDK 16.07. The next versions,
> > based
> > on a 2 year cycle, will be DPDK 18.08, 20.08, etc.
> >
>
> I can see on the discussions that much more things around this have to be
> discussed and agreed, but to some extend we will also just "have to wait
> and see" how things work out.
> I fully agree to the API change argument to start with 16.07 and the 2 year
> cycle (more would be nice, but this means effort and after a while almost
> nothing is "easily" backportable).
>
> Never the less I have to ask - as I'd be personally much more happy if it
> would be the odd years autumn release that would make the LTS as it would
> match our LTS releases much much better.
> Otherwise we (Ubuntu) will always "just miss" the LTS by a few months.
> First I'd have thought on xx.02 releases, but consuming applications might
> need time to adapt and while there are the nice API/ABI guarantees
> experience tells me to better leave some kind of time-buffer.
>
> Also this would give all of us a first shot with a shorter (not so long as
> in L) LTS to see if the process we defined works out before jumping on a
> full 2 year cadence.
>
> So while I see that this is kind of "my problem" I would at least try to
> personally ask and vote for LTS being: 16.7, 17.11, 19.11, 21.11, ...
>
> Christian Ehrhardt
> Software Engineer, Ubuntu Server
> Canonical Ltd
>
+1 For DPDK LTS being:
16.7, 17.11, 19.11, 21.11, ...
This way, DPDK LTS 17.11 will be part of Ubuntu LTS 18.04... DPDK LTS 19.11
on Ubuntu LTS 20.04...
Very likely that DPDK LTS 16.07 will be available to Ubuntu 16.04 via
Ubuntu Cloud Archive Newton.
Cheers!
Thiago
^ permalink raw reply [relevance 0%]
* Re: [dpdk-dev] RFC: DPDK Long Term Support
@ 2016-06-07 16:21 3% ` Mcnamara, John
0 siblings, 0 replies; 200+ results
From: Mcnamara, John @ 2016-06-07 16:21 UTC (permalink / raw)
To: Neil Horman, Thomas Monjalon
Cc: dev, Christian Ehrhardt, Markos Chandras, Panu Matilainen
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Neil Horman
> Sent: Monday, June 6, 2016 2:48 PM
> To: Thomas Monjalon <thomas.monjalon@6wind.com>
> Cc: dev@dpdk.org; Mcnamara, John <john.mcnamara@intel.com>; Christian
> Ehrhardt <christian.ehrhardt@canonical.com>; Markos Chandras
> <mchandras@suse.de>; Panu Matilainen <pmatilai@redhat.com>
> Subject: Re: [dpdk-dev] RFC: DPDK Long Term Support
>
> While I don't disagree with that statement (LTS does provide both of those
> things if the maintainer does it properly), I'm forced to ask the
> question, before we solve this problem in a new way, lets ask why the
> existing way isn't being used. Do developers just not care about
> backwards compatibility? Is the process to hard? Something else? I
> really don't like the idea of abandoning what currently exists to replace
> it with something else, without first addressing why what we have isn't
> working.
Hi Neil,
I think these questions around why the current ABI policy isn't working
(or at least not working well) and how it can be fixed are worth raising
as a new discussion.
John.
--
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] RFC: DPDK Long Term Support
@ 2016-06-07 15:55 5% ` Mcnamara, John
1 sibling, 0 replies; 200+ results
From: Mcnamara, John @ 2016-06-07 15:55 UTC (permalink / raw)
To: Neil Horman; +Cc: dev, Christian Ehrhardt, Markos Chandras, Panu Matilainen
> -----Original Message-----
> From: Neil Horman [mailto:nhorman@tuxdriver.com]
> Sent: Sunday, June 5, 2016 7:15 PM
> To: Mcnamara, John <john.mcnamara@intel.com>
> Cc: dev <dev@dpdk.org>; Christian Ehrhardt
> <christian.ehrhardt@canonical.com>; Markos Chandras <mchandras@suse.de>;
> Panu Matilainen <pmatilai@redhat.com>
> Subject: Re: [dpdk-dev] RFC: DPDK Long Term Support
>
> >
> I'm not opposed to an LTS release, but it seems to be re-solving the issue
> of ABI breakage. That is to say, there is alreay a process in place for
> managing ABI changes to the DPDK, which is designed to help ensure that:
>
> 1) ABI changes are signaled at least 2 releases early
> 2) ABI changes whenever possible are designed such that backward
> compatibility versions can be encoded at the same time with versioning
> tags
>
> Those two mechanism are expressly intended to allow application upgrades
> of DPDK libraries without worrying about ABI breakage.
Hi,
The purpose of the LTS proposal isn't to replace or circumvent the ABI policy.
In fact backporting of patches would be very difficult without an upstream
ABI policy.
Even if the ABI policy was working perfectly there would still be a use case
for an LTS among consumers who want a fixed version with bug fixes or minor
changes. There are already several companies maintaining their own branches
like this. This purpose of this proposal is to get them to converge on a
single version (or, if there is support, versions) and combine their efforts.
> While LTS releases
> are a fine approach for some things, they sacrifice upstream efficiency
> (by creating work for backporting teams), while allowing upstream
> developers more leverage to just create ABI breaking changes on a whim,
> ignoring the existing ABI compatibility mechanism
An LTS release doesn't prevent us from maintaining upstream ABI compatibility
and it only gives developers leverage if we allow it to.
John.
--
^ permalink raw reply [relevance 5%]
* Re: [dpdk-dev] RFC: DPDK Long Term Support
@ 2016-06-07 13:17 3% ` Mcnamara, John
0 siblings, 0 replies; 200+ results
From: Mcnamara, John @ 2016-06-07 13:17 UTC (permalink / raw)
To: Thomas Monjalon; +Cc: dev, Christian Ehrhardt, Markos Chandras, Panu Matilainen
> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Friday, June 3, 2016 5:05 PM
> To: Mcnamara, John <john.mcnamara@intel.com>
> Cc: dev@dpdk.org; Christian Ehrhardt <christian.ehrhardt@canonical.com>;
> Markos Chandras <mchandras@suse.de>; Panu Matilainen <pmatilai@redhat.com>
> Subject: Re: [dpdk-dev] RFC: DPDK Long Term Support
>
> Hi,
>
> 2016-06-03 15:07, Mcnamara, John:
> > Introduction
> > ------------
> >
> > This document sets out a proposal for a DPDK Long Term Support release
> (LTS).
>
> In general, LTS refer to a longer maintenance than than regular one.
> Here we are talking to doing some maintenance as stable releases first.
> Currently we have no maintenance at all.
> So I suggest to differentiate "stable branches" and "LTS" for some stable
> branches.
Hi Thomas,
I have no argument against this. It would be great to have a series of stable
branches of which some are LTS.
But at a minimum we are committing to have a least one maintained stable branch
that will also be a LTS.
> I wonder if Yuanhan is OK to maintain every stable releases which could be
> requested/needed? Or should we have other committers for the stable
> releases that Yuanhan would not want to maintain himself?
> The Linux model is to let people declare themselves when they want to
> maintain a stable branch.
I think it is fine to have other committers.
> > The proposed duration of the LTS support is 2 years.
>
> I think we should discuss the support duration for each release
> separately.
>
> > There will only be one LTS branch being maintained at any time. At the
> > end of the 2 year cycle the maintenance on the previous LTS will be
> wound down.
>
> Seems a bit too restrictive.
> Currently, there is no maintenance at all because nobody was volunteer.
> If Yuanhan is volunteer for a stable branch every 2 years, fine.
> If someone else is volunteer for other branches, why not let him do it?
I see no problem with that. This proposal just reflects that fact that we
have only had one volunteer to date and is based on what could be reasonably
done by one person (plus the validation support). If more maintainers come
forward we can have more/more frequent stable branches.
We will, however, be constrained by the validation effort that can be offered,
unless there are other validation volunteers.
> > The proposed initial LTS version will be DPDK 16.07. The next
> > versions, based on a 2 year cycle, will be DPDK 18.08, 20.08, etc.
>
> Let's do a first run with 16.07 and see later what we want to do next.
> How long time a stable branch must be announced before its initial
> release?
Ok. The statement at the end about reviewing at the end of the first year
is meant to cover adjustments like this. I think that we will have to see
how things work out in practice and adjust as we go.
> > What changes should be backported
> > ---------------------------------
> >
> > * Bug fixes that don't break the ABI.
>
> And API?
> And behaviour (if not clearly documented in the API)?
Yes. It should say ABI and API. Undocumented but implied or existing
bahaviour should also be maintained.
> > (OSV reps please confirm.)
> >
> > * Ubuntu 16.04 LTS
> > * RHEL 7.3
> > * SuSE 11 SP4 or 12
> > * FreeBSD 10.3
>
> I'm sure there will be more validation on the field or from contributors.
Hopefully. :-)
John.
--
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] RFC: DPDK Long Term Support
@ 2016-06-07 12:53 3% ` Mcnamara, John
0 siblings, 0 replies; 200+ results
From: Mcnamara, John @ 2016-06-07 12:53 UTC (permalink / raw)
To: Matthew Hall; +Cc: dev, Christian Ehrhardt, Markos Chandras, Panu Matilainen
> -----Original Message-----
> From: Matthew Hall [mailto:mhall@mhcomputing.net]
> Sent: Friday, June 3, 2016 7:17 PM
> To: Mcnamara, John <john.mcnamara@intel.com>
> Cc: dev <dev@dpdk.org>; Christian Ehrhardt
> <christian.ehrhardt@canonical.com>; Markos Chandras <mchandras@suse.de>;
> Panu Matilainen <pmatilai@redhat.com>
> Subject: Re: [dpdk-dev] RFC: DPDK Long Term Support
>
> >
> > What changes should not be backported
> > -------------------------------------
> >
> > * API or ABI breaking changes.
>
> I think this part needs some adjusting.
>
> It seems like there should be allowance for bug fixes where the original
> does break ABI but it is possible to make a version that doesn't.
>
> A lot of DPDK bug fixes I see would fall into this category and it isn't
> discussed.
Hi Matthew,
Agreed, we should allow fix to be backported even if the patch itself cannot be, for ABI reasons.
John
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] RFC: DPDK Long Term Support
` (2 preceding siblings ...)
@ 2016-06-07 12:36 3% ` Christian Ehrhardt
2016-06-07 19:39 0% ` Martinx - ジェームズ
3 siblings, 1 reply; 200+ results
From: Christian Ehrhardt @ 2016-06-07 12:36 UTC (permalink / raw)
To: Mcnamara, John; +Cc: dev, Markos Chandras, Panu Matilainen
On Fri, Jun 3, 2016 at 5:07 PM, Mcnamara, John <john.mcnamara@intel.com>
wrote:
[...]
>
> LTS Version
> ------------
>
> The proposed initial LTS version will be DPDK 16.07. The next versions,
> based
> on a 2 year cycle, will be DPDK 18.08, 20.08, etc.
>
I can see on the discussions that much more things around this have to be
discussed and agreed, but to some extend we will also just "have to wait
and see" how things work out.
I fully agree to the API change argument to start with 16.07 and the 2 year
cycle (more would be nice, but this means effort and after a while almost
nothing is "easily" backportable).
Never the less I have to ask - as I'd be personally much more happy if it
would be the odd years autumn release that would make the LTS as it would
match our LTS releases much much better.
Otherwise we (Ubuntu) will always "just miss" the LTS by a few months.
First I'd have thought on xx.02 releases, but consuming applications might
need time to adapt and while there are the nice API/ABI guarantees
experience tells me to better leave some kind of time-buffer.
Also this would give all of us a first shot with a shorter (not so long as
in L) LTS to see if the process we defined works out before jumping on a
full 2 year cadence.
So while I see that this is kind of "my problem" I would at least try to
personally ask and vote for LTS being: 16.7, 17.11, 19.11, 21.11, ...
Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd
^ permalink raw reply [relevance 3%]
* Re: [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
2016-06-07 6:53 4% ` [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode Zhe Tao
@ 2016-06-07 9:58 0% ` Ananyev, Konstantin
2016-06-08 7:24 3% ` Lu, Wenzhuo
0 siblings, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-06-07 9:58 UTC (permalink / raw)
To: Tao, Zhe, dev
Cc: Lu, Wenzhuo, Richardson, Bruce, Chen, Jing D, Liang, Cunming, Wu,
Jingjing, Zhang, Helin
Hi Zhe & Wenzhuo,
Please find my comments below.
BTW, for clarification - is that patch for 16.11?
I believe it's too late to introduce such significant change in 16.07.
Thanks
Konstantin
> Define lock mode for RX/TX queue. Because when resetting
> the device we want the resetting thread to get the lock
> of the RX/TX queue to make sure the RX/TX is stopped.
>
> Using next ABI macro for this ABI change as it has too
> much impact. 7 APIs and 1 global variable are impacted.
>
> Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> Signed-off-by: Zhe Tao <zhe.tao@intel.com>
> ---
> lib/librte_ether/rte_ethdev.h | 62 +++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 62 insertions(+)
>
> diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
> index 74e895f..4efb5e9 100644
> --- a/lib/librte_ether/rte_ethdev.h
> +++ b/lib/librte_ether/rte_ethdev.h
> @@ -354,7 +354,12 @@ struct rte_eth_rxmode {
> jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
> hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
> enable_scatter : 1, /**< Enable scatter packets rx handler */
> +#ifndef RTE_NEXT_ABI
> enable_lro : 1; /**< Enable LRO */
> +#else
> + enable_lro : 1, /**< Enable LRO */
> + lock_mode : 1; /**< Using lock path */
> +#endif
> };
>
> /**
> @@ -634,11 +639,68 @@ struct rte_eth_txmode {
> /**< If set, reject sending out tagged pkts */
> hw_vlan_reject_untagged : 1,
> /**< If set, reject sending out untagged pkts */
> +#ifndef RTE_NEXT_ABI
> hw_vlan_insert_pvid : 1;
> /**< If set, enable port based VLAN insertion */
> +#else
> + hw_vlan_insert_pvid : 1,
> + /**< If set, enable port based VLAN insertion */
> + lock_mode : 1;
> + /**< If set, using lock path */
> +#endif
> };
>
> /**
> + * The macros for the RX/TX lock mode functions
> + */
> +#ifdef RTE_NEXT_ABI
> +#define RX_LOCK_FUNCTION(dev, func) \
> + (dev->data->dev_conf.rxmode.lock_mode ? \
> + func ## _lock : func)
> +
> +#define TX_LOCK_FUNCTION(dev, func) \
> + (dev->data->dev_conf.txmode.lock_mode ? \
> + func ## _lock : func)
> +#else
> +#define RX_LOCK_FUNCTION(dev, func) func
> +
> +#define TX_LOCK_FUNCTION(dev, func) func
> +#endif
> +
> +/* Add the lock RX/TX function for VF reset */
> +#define GENERATE_RX_LOCK(func, nic) \
> +uint16_t func ## _lock(void *rx_queue, \
> + struct rte_mbuf **rx_pkts, \
> + uint16_t nb_pkts) \
> +{ \
> + struct nic ## _rx_queue *rxq = rx_queue; \
> + uint16_t nb_rx = 0; \
> + \
> + if (rte_spinlock_trylock(&rxq->rx_lock)) { \
> + nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
> + rte_spinlock_unlock(&rxq->rx_lock); \
> + } \
> + \
> + return nb_rx; \
> +}
> +
> +#define GENERATE_TX_LOCK(func, nic) \
> +uint16_t func ## _lock(void *tx_queue, \
> + struct rte_mbuf **tx_pkts, \
> + uint16_t nb_pkts) \
> +{ \
> + struct nic ## _tx_queue *txq = tx_queue; \
> + uint16_t nb_tx = 0; \
> + \
> + if (rte_spinlock_trylock(&txq->tx_lock)) { \
> + nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
> + rte_spinlock_unlock(&txq->tx_lock); \
> + } \
> + \
> + return nb_tx; \
> +}
1. As I said in off-line dicussiion, I think this locking could
(and I think better be) impelented completely on rte_ethdev layer.
So actual PMD code will be unaffected.
Again that avoids us to introduce _lock version of every RX/Tx function
in each PMD.
2. Again, as discussed offline, I think it is better to have an explicit
rte_eth_(rx|tx)_burst_lock(sync?) API, instead of add new fileds into
RX/TX config strcutures.
Would help to avoid any confusion, I think.
3. I thought the plan was to introduce a locking in all appropriate control path
functions (dev_start/dev_stop etc.)
Without that locking version of RX/TX seems a bit useless.
Yes, I understand that you do use locking inside dev_reset, but I suppose
the plan was to have a generic solution, no?
Again, interrupt fire when user invokes dev_start/stop or so, so we still
need some synchronisation between them.
To be more specific, I thought about something like that:
static inline uint16_t
rte_eth_rx_burst_lock(uint8_t port_id, uint16_t queue_id,
struct rte_mbuf **rx_pkts, const uint16_t nb_pkts)
{
struct rte_eth_dev *dev = &rte_eth_devices[port_id];
#ifdef RTE_LIBRTE_ETHDEV_DEBUG
RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, 0);
RTE_FUNC_PTR_OR_ERR_RET(*dev->rx_pkt_burst, 0);
if (queue_id >= dev->data->nb_rx_queues) {
RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", queue_id);
return 0;
}
#endif
+ if (rte_spinlock_trylock(&dev->data->rx_queue_state[rx_queue_id].lock) == 0)
+ return 0;
+ else if (dev->data->rx_queue_state[rx_queue_id] == RTE_ETH_QUEUE_STATE_STOPPED)) {
+ rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
+ return 0;
+
nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
rx_pkts, nb_pkts);
+ rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
....
return nb_rx;
}
And inside queue_start:
int
rte_eth_dev_rx_queue_start(uint8_t port_id, uint16_t rx_queue_id)
{
struct rte_eth_dev *dev;
RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -EINVAL);
dev = &rte_eth_devices[port_id];
if (rx_queue_id >= dev->data->nb_rx_queues) {
RTE_PMD_DEBUG_TRACE("Invalid RX queue_id=%d\n", rx_queue_id);
return -EINVAL;
}
RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->rx_queue_start, -ENOTSUP);
rte_spinlock_lock(&dev->data->rx_queue_state[rx_queue_id].lock)
if (dev->data->rx_queue_state[rx_queue_id] != RTE_ETH_QUEUE_STATE_STOPPED) {
RTE_PMD_DEBUG_TRACE("Queue %" PRIu16" of device with port_id=%" PRIu8
" already started\n",
rx_queue_id, port_id);
ret = -EINVAL 0;
} else
ret = dev->dev_ops->rx_queue_start(dev, rx_queue_id);
rte_spinlock_unlock(&dev->data->rx_queue_state[rx_queue_id].unlock);
return ret;
}
Then again, we don't need to do explicit locking inside dev_reset().
Does it make sense to you guys?
> +
> +/**
> * A structure used to configure an RX ring of an Ethernet port.
> */
> struct rte_eth_rxconf {
> --
> 2.1.4
^ permalink raw reply [relevance 0%]
* [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode
@ 2016-06-07 6:53 4% ` Zhe Tao
2016-06-07 9:58 0% ` Ananyev, Konstantin
0 siblings, 1 reply; 200+ results
From: Zhe Tao @ 2016-06-07 6:53 UTC (permalink / raw)
To: dev
Cc: wenzhuo.lu, zhe.tao, konstantin.ananyev, bruce.richardson,
jing.d.chen, cunming.liang, jingjing.wu, helin.zhang
Define lock mode for RX/TX queue. Because when resetting
the device we want the resetting thread to get the lock
of the RX/TX queue to make sure the RX/TX is stopped.
Using next ABI macro for this ABI change as it has too
much impact. 7 APIs and 1 global variable are impacted.
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Signed-off-by: Zhe Tao <zhe.tao@intel.com>
---
lib/librte_ether/rte_ethdev.h | 62 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 62 insertions(+)
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 74e895f..4efb5e9 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -354,7 +354,12 @@ struct rte_eth_rxmode {
jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
enable_scatter : 1, /**< Enable scatter packets rx handler */
+#ifndef RTE_NEXT_ABI
enable_lro : 1; /**< Enable LRO */
+#else
+ enable_lro : 1, /**< Enable LRO */
+ lock_mode : 1; /**< Using lock path */
+#endif
};
/**
@@ -634,11 +639,68 @@ struct rte_eth_txmode {
/**< If set, reject sending out tagged pkts */
hw_vlan_reject_untagged : 1,
/**< If set, reject sending out untagged pkts */
+#ifndef RTE_NEXT_ABI
hw_vlan_insert_pvid : 1;
/**< If set, enable port based VLAN insertion */
+#else
+ hw_vlan_insert_pvid : 1,
+ /**< If set, enable port based VLAN insertion */
+ lock_mode : 1;
+ /**< If set, using lock path */
+#endif
};
/**
+ * The macros for the RX/TX lock mode functions
+ */
+#ifdef RTE_NEXT_ABI
+#define RX_LOCK_FUNCTION(dev, func) \
+ (dev->data->dev_conf.rxmode.lock_mode ? \
+ func ## _lock : func)
+
+#define TX_LOCK_FUNCTION(dev, func) \
+ (dev->data->dev_conf.txmode.lock_mode ? \
+ func ## _lock : func)
+#else
+#define RX_LOCK_FUNCTION(dev, func) func
+
+#define TX_LOCK_FUNCTION(dev, func) func
+#endif
+
+/* Add the lock RX/TX function for VF reset */
+#define GENERATE_RX_LOCK(func, nic) \
+uint16_t func ## _lock(void *rx_queue, \
+ struct rte_mbuf **rx_pkts, \
+ uint16_t nb_pkts) \
+{ \
+ struct nic ## _rx_queue *rxq = rx_queue; \
+ uint16_t nb_rx = 0; \
+ \
+ if (rte_spinlock_trylock(&rxq->rx_lock)) { \
+ nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
+ rte_spinlock_unlock(&rxq->rx_lock); \
+ } \
+ \
+ return nb_rx; \
+}
+
+#define GENERATE_TX_LOCK(func, nic) \
+uint16_t func ## _lock(void *tx_queue, \
+ struct rte_mbuf **tx_pkts, \
+ uint16_t nb_pkts) \
+{ \
+ struct nic ## _tx_queue *txq = tx_queue; \
+ uint16_t nb_tx = 0; \
+ \
+ if (rte_spinlock_trylock(&txq->tx_lock)) { \
+ nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
+ rte_spinlock_unlock(&txq->tx_lock); \
+ } \
+ \
+ return nb_tx; \
+}
+
+/**
* A structure used to configure an RX ring of an Ethernet port.
*/
struct rte_eth_rxconf {
--
2.1.4
^ permalink raw reply [relevance 4%]
* [dpdk-dev] [PATCH v3 2/8] lib/librte_ether: defind RX/TX lock mode
@ 2016-06-07 6:12 4% ` Zhe Tao
0 siblings, 0 replies; 200+ results
From: Zhe Tao @ 2016-06-07 6:12 UTC (permalink / raw)
To: dev
Cc: wenzhuo.lu, zhe.tao, konstantin.ananyev, bruce.richardson,
jing.d.chen, cunming.liang, jingjing.wu, helin.zhang
Define lock mode for RX/TX queue. Because when resetting
the device we want the resetting thread to get the lock
of the RX/TX queue to make sure the RX/TX is stopped.
Using next ABI macro for this ABI change as it has too
much impact. 7 APIs and 1 global variable are impacted.
Signed-off-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
Signed-off-by: Zhe Tao <zhe.tao@intel.com>
---
lib/librte_ether/rte_ethdev.h | 62 +++++++++++++++++++++++++++++++++++++++++++
1 file changed, 62 insertions(+)
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 74e895f..4efb5e9 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -354,7 +354,12 @@ struct rte_eth_rxmode {
jumbo_frame : 1, /**< Jumbo Frame Receipt enable. */
hw_strip_crc : 1, /**< Enable CRC stripping by hardware. */
enable_scatter : 1, /**< Enable scatter packets rx handler */
+#ifndef RTE_NEXT_ABI
enable_lro : 1; /**< Enable LRO */
+#else
+ enable_lro : 1, /**< Enable LRO */
+ lock_mode : 1; /**< Using lock path */
+#endif
};
/**
@@ -634,11 +639,68 @@ struct rte_eth_txmode {
/**< If set, reject sending out tagged pkts */
hw_vlan_reject_untagged : 1,
/**< If set, reject sending out untagged pkts */
+#ifndef RTE_NEXT_ABI
hw_vlan_insert_pvid : 1;
/**< If set, enable port based VLAN insertion */
+#else
+ hw_vlan_insert_pvid : 1,
+ /**< If set, enable port based VLAN insertion */
+ lock_mode : 1;
+ /**< If set, using lock path */
+#endif
};
/**
+ * The macros for the RX/TX lock mode functions
+ */
+#ifdef RTE_NEXT_ABI
+#define RX_LOCK_FUNCTION(dev, func) \
+ (dev->data->dev_conf.rxmode.lock_mode ? \
+ func ## _lock : func)
+
+#define TX_LOCK_FUNCTION(dev, func) \
+ (dev->data->dev_conf.txmode.lock_mode ? \
+ func ## _lock : func)
+#else
+#define RX_LOCK_FUNCTION(dev, func) func
+
+#define TX_LOCK_FUNCTION(dev, func) func
+#endif
+
+/* Add the lock RX/TX function for VF reset */
+#define GENERATE_RX_LOCK(func, nic) \
+uint16_t func ## _lock(void *rx_queue, \
+ struct rte_mbuf **rx_pkts, \
+ uint16_t nb_pkts) \
+{ \
+ struct nic ## _rx_queue *rxq = rx_queue; \
+ uint16_t nb_rx = 0; \
+ \
+ if (rte_spinlock_trylock(&rxq->rx_lock)) { \
+ nb_rx = func(rx_queue, rx_pkts, nb_pkts); \
+ rte_spinlock_unlock(&rxq->rx_lock); \
+ } \
+ \
+ return nb_rx; \
+}
+
+#define GENERATE_TX_LOCK(func, nic) \
+uint16_t func ## _lock(void *tx_queue, \
+ struct rte_mbuf **tx_pkts, \
+ uint16_t nb_pkts) \
+{ \
+ struct nic ## _tx_queue *txq = tx_queue; \
+ uint16_t nb_tx = 0; \
+ \
+ if (rte_spinlock_trylock(&txq->tx_lock)) { \
+ nb_tx = func(tx_queue, tx_pkts, nb_pkts); \
+ rte_spinlock_unlock(&txq->tx_lock); \
+ } \
+ \
+ return nb_tx; \
+}
+
+/**
* A structure used to configure an RX ring of an Ethernet port.
*/
struct rte_eth_rxconf {
--
2.1.4
^ permalink raw reply [relevance 4%]
Results 11801-12000 of ~18000 next (older) | prev (newer) | reverse | sort options + mbox downloads above
-- links below jump to the message on this page --
2016-01-29 14:08 [dpdk-dev] [PATCH 0/9] prepare for rte_device / rte_driver David Marchand
2016-06-16 14:06 ` [dpdk-dev] [PATCH v3 00/17] " Shreyansh Jain
2016-06-16 14:06 3% ` [dpdk-dev] [PATCH v3 04/17] eal: remove duplicate function declaration Shreyansh Jain
2016-06-21 12:02 ` [dpdk-dev] [PATCH v4 00/17] prepare for rte_device / rte_driver Shreyansh Jain
2016-06-21 12:02 3% ` [dpdk-dev] [PATCH v4 04/17] eal: remove duplicate function declaration Shreyansh Jain
2016-06-22 9:06 ` [dpdk-dev] [PATCH v5 00/17] Prepare for rte_device / rte_driver Shreyansh Jain
2016-06-22 9:06 3% ` [dpdk-dev] [PATCH v5 04/17] eal: remove duplicate function declaration Shreyansh Jain
2016-06-22 9:06 ` [dpdk-dev] [PATCH v5 09/17] crypto: get rid of crypto driver register callback Shreyansh Jain
2016-06-22 13:27 ` Neil Horman
2016-06-22 13:44 3% ` Thomas Monjalon
2016-07-12 6:01 ` [dpdk-dev] [PATCH v6 00/17] Prepare for rte_device / rte_driver Shreyansh Jain
2016-07-12 6:01 3% ` [dpdk-dev] [PATCH v6 04/17] eal: remove duplicate function declaration Shreyansh Jain
2016-07-14 17:13 0% ` viktorin
2016-02-17 14:20 [dpdk-dev] [PATCH 0/3] ethdev: add helper functions to get eth_dev and dev private data Ferruh Yigit
2016-06-22 21:47 0% ` Thomas Monjalon
2016-05-03 0:46 [dpdk-dev] [PATCH 0/3] [RFC] vhost: micro vhost optimization Yuanhan Liu
2016-06-14 12:42 0% ` Yuanhan Liu
2016-05-06 20:05 [dpdk-dev] [PATCH v1] hash: add tsx support for cuckoo hash Shen Wei
2016-05-07 4:56 ` Stephen Hemminger
2016-05-09 16:51 ` Shen, Wei1
2016-06-10 11:09 0% ` De Lara Guarch, Pablo
2016-05-13 5:24 [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
2016-06-07 3:51 ` [dpdk-dev] [PATCH v3 00/20] " Yuanhan Liu
2016-06-14 12:00 4% ` Yuanhan Liu
2016-06-30 7:39 9% ` Panu Matilainen
2016-06-30 7:57 4% ` Yuanhan Liu
2016-06-30 9:05 7% ` Panu Matilainen
2016-06-30 11:15 7% ` Mcnamara, John
2016-06-30 11:40 4% ` Thomas Monjalon
2016-05-13 12:50 [dpdk-dev] [PATCH 0/7] virtio-net support on ppc64 Olivier Matz
2016-05-24 5:15 ` [dpdk-dev] [PATCH v2 5/7] eal/linux: mmap ioports " Yuanhan Liu
2016-05-30 8:45 ` Olivier Matz
2016-06-15 16:13 3% ` Thomas Monjalon
2016-05-16 13:18 [dpdk-dev] [PATCH 0/2] doc: announce ABI change of struct rte_port_source_params Fan Zhang
2016-05-19 14:18 ` [dpdk-dev] [PATCH v2] doc: announce ABI change of struct rte_port_source_params and rte_port_sink_params Fan Zhang
2016-07-27 10:08 9% ` Dumitrescu, Cristian
2016-07-27 10:42 7% ` Thomas Monjalon
2016-07-28 18:28 4% ` Thomas Monjalon
2016-05-19 13:17 [dpdk-dev] [PATCH v3] ci: Add the class_id support in pci probe Ziye Yang
2016-05-24 12:50 ` [dpdk-dev] [PATCH v4] Pci: Add the class_id support Ziye Yang
2016-06-14 14:52 ` Thomas Monjalon
2016-07-06 11:08 3% ` Ferruh Yigit
2016-07-07 7:46 0% ` Thomas Monjalon
2016-05-19 14:48 [dpdk-dev] v2 mempool: add stack (lifo) mempool handler David Hunt
2016-06-20 13:08 ` [dpdk-dev] [PATCH v3 1/2] " David Hunt
2016-06-20 13:25 ` Jerin Jacob
2016-06-20 13:54 ` Thomas Monjalon
2016-06-20 13:58 ` Ananyev, Konstantin
2016-06-20 14:22 ` Jerin Jacob
2016-06-20 17:56 ` Ananyev, Konstantin
2016-06-21 3:35 ` Jerin Jacob
2016-06-21 9:28 ` Ananyev, Konstantin
2016-06-21 9:44 3% ` Olivier Matz
2016-05-20 14:08 [dpdk-dev] [PATCH v2] doc: fix code section in abi versioning doc John McNamara
2016-06-08 16:46 4% ` Thomas Monjalon
2016-05-23 21:38 [dpdk-dev] [PATCH v4 0/9] add packet capture framework Reshma Pattan
2016-06-08 13:38 3% ` [dpdk-dev] [PATCH v5 " Reshma Pattan
2016-06-08 13:38 5% ` [dpdk-dev] [PATCH v5 8/9] doc: update doc for " Reshma Pattan
2016-06-08 13:38 9% ` [dpdk-dev] [PATCH v5 9/9] doc: announce ABI change for rte_eth_dev_info structure Reshma Pattan
2016-06-08 16:15 4% ` Mcnamara, John
2016-06-09 8:50 2% ` [dpdk-dev] [PATCH v6 0/8] add packet capture framework Reshma Pattan
2016-06-09 8:50 5% ` [dpdk-dev] [PATCH v6 8/8] doc: update doc for " Reshma Pattan
2016-06-09 16:10 2% ` [dpdk-dev] [PATCH v7 0/8] add " Reshma Pattan
2016-06-09 16:10 5% ` [dpdk-dev] [PATCH v7 8/8] doc: update doc for " Reshma Pattan
2016-06-09 17:34 0% ` [dpdk-dev] [PATCH v7 0/8] add " Ananyev, Konstantin
2016-05-26 7:28 [dpdk-dev] [PATCH] ethdev: change comments of VLAN type Beilei Xing
2016-06-13 8:03 4% ` [dpdk-dev] [PATCH v2] i40e: modify the meaning of single " Beilei Xing
2016-06-21 10:29 4% ` Bruce Richardson
2016-06-21 11:06 0% ` Panu Matilainen
2016-06-21 11:28 0% ` Bruce Richardson
2016-05-27 14:33 [dpdk-dev] [PATCH v2] mbuf: new flag when Vlan is stripped Olivier Matz
2016-06-15 11:48 2% ` [dpdk-dev] [PATCH v3] " Olivier Matz
2016-05-30 10:48 [dpdk-dev] [PATCH v3 00/10] Remove string operations from xstats Remy Horton
2016-05-30 10:48 ` [dpdk-dev] [PATCH v3 01/10] rte: change xstats to use integer ids Remy Horton
2016-06-08 9:37 ` Thomas Monjalon
2016-06-08 11:16 3% ` Remy Horton
2016-06-08 12:22 0% ` Thomas Monjalon
2016-06-03 14:58 [dpdk-dev] [PATCH v8 0/5] mempool: add external mempool manager David Hunt
2016-06-10 15:16 2% ` [dpdk-dev] [PATCH v9 0/3] " David Hunt
2016-06-14 9:46 2% ` [dpdk-dev] [PATCH v10 " David Hunt
2016-06-14 15:48 3% ` [dpdk-dev] [PATCH v11 " David Hunt
2016-06-14 15:48 1% ` [dpdk-dev] [PATCH v11 1/3] mempool: support external mempool operations David Hunt
2016-06-15 7:47 3% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager David Hunt
2016-06-15 7:47 1% ` [dpdk-dev] [PATCH v12 1/3] mempool: support external mempool operations David Hunt
2016-06-15 10:13 0% ` [dpdk-dev] [PATCH v12 0/3] mempool: add external mempool manager Jan Viktorin
2016-06-16 12:30 3% ` [dpdk-dev] [PATCH v13 " David Hunt
2016-06-16 12:30 1% ` [dpdk-dev] [PATCH v13 1/3] mempool: support external mempool operations David Hunt
2016-06-17 13:53 3% ` [dpdk-dev] [PATCH v14 0/3] mempool: add mempool handler feature David Hunt
2016-06-17 13:53 1% ` [dpdk-dev] [PATCH v14 1/3] mempool: support mempool handler operations David Hunt
2016-06-19 12:05 3% ` [dpdk-dev] [PATCH v15 0/3] mempool: add mempool handler feature David Hunt
2016-06-19 12:05 1% ` [dpdk-dev] [PATCH v15 1/3] mempool: support mempool handler operations David Hunt
2016-06-22 9:27 3% ` [dpdk-dev] [PATCH v16 0/3] mempool: add mempool handler feature David Hunt
2016-06-22 9:27 1% ` [dpdk-dev] [PATCH v16 1/3] mempool: support mempool handler operations David Hunt
2016-06-03 15:07 [dpdk-dev] RFC: DPDK Long Term Support Mcnamara, John
2016-06-03 16:05 ` Thomas Monjalon
2016-06-07 13:17 3% ` Mcnamara, John
2016-06-03 18:17 ` Matthew Hall
2016-06-07 12:53 3% ` Mcnamara, John
2016-06-05 18:15 ` Neil Horman
2016-06-06 9:27 ` Thomas Monjalon
2016-06-06 13:47 ` Neil Horman
2016-06-07 16:21 3% ` Mcnamara, John
2016-06-07 15:55 5% ` Mcnamara, John
2016-06-07 12:36 3% ` Christian Ehrhardt
2016-06-07 19:39 0% ` Martinx - ジェームズ
2016-06-06 5:40 [dpdk-dev] [PATCH 0/8] support reset of VF link Wenzhuo Lu
2016-06-06 5:40 ` [dpdk-dev] [PATCH 2/8] lib/librte_ether: defind RX/TX lock mode Wenzhuo Lu
2016-06-08 2:15 0% ` Stephen Hemminger
2016-06-08 7:34 0% ` Lu, Wenzhuo
2016-06-09 7:50 0% ` Olivier Matz
2016-06-12 5:25 0% ` Lu, Wenzhuo
2016-06-07 4:05 [dpdk-dev] [PATCH v3 0/6] vhost: add vhost-user client mode and reconnect ability Yuanhan Liu
2016-06-14 12:00 0% ` Yuanhan Liu
2016-06-07 6:12 [dpdk-dev] [PATCH v3 0/8] support reset of VF link Zhe Tao
2016-06-07 6:12 4% ` [dpdk-dev] [PATCH v3 2/8] lib/librte_ether: defind RX/TX lock mode Zhe Tao
[not found] <1465278858-5131-1-git-send-email-zhe.tao@intel.com>
2016-06-07 6:53 ` [dpdk-dev] [PATCH v4 0/8] support reset of VF link Zhe Tao
2016-06-07 6:53 4% ` [dpdk-dev] [PATCH v4 2/8] lib/librte_ether: defind RX/TX lock mode Zhe Tao
2016-06-07 9:58 0% ` Ananyev, Konstantin
2016-06-08 7:24 3% ` Lu, Wenzhuo
2016-06-08 9:19 0% ` Ananyev, Konstantin
2016-06-12 2:00 0% ` Lu, Wenzhuo
2016-06-12 23:16 0% ` Ananyev, Konstantin
2016-06-13 1:06 0% ` Lu, Wenzhuo
2016-06-08 8:31 [dpdk-dev] [PATCH] mbuf: remove inconsistent assert statements Adrien Mazarguil
2016-06-08 10:34 ` Ananyev, Konstantin
2016-06-08 12:27 ` Adrien Mazarguil
2016-06-08 13:09 ` Ananyev, Konstantin
2016-06-08 13:57 ` Adrien Mazarguil
2016-06-08 14:11 ` Olivier Matz
2016-06-08 16:07 3% ` Ananyev, Konstantin
2016-06-09 7:46 0% ` Olivier Matz
2016-06-09 13:21 0% ` Ananyev, Konstantin
2016-06-09 14:09 6% [dpdk-dev] [PATCH] log: deprecate history dump Thomas Monjalon
2016-06-09 14:45 0% ` David Marchand
2016-06-09 15:01 0% ` Thomas Monjalon
2016-06-09 15:01 0% ` Christian Ehrhardt
2016-06-09 15:06 5% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
2016-06-09 22:10 5% ` [dpdk-dev] [PATCH v3] " Thomas Monjalon
2016-06-10 9:50 0% ` David Marchand
2016-06-10 13:09 0% ` Thomas Monjalon
[not found] <1465487895-5870-1-git-send-email-reshma.pattan@intel.com>
2016-06-10 16:18 2% ` [dpdk-dev] [PATCH v8 0/8] add packet capture framework Reshma Pattan
2016-06-10 16:18 5% ` [dpdk-dev] [PATCH v8 8/8] doc: update doc for " Reshma Pattan
2016-06-10 23:23 0% ` [dpdk-dev] [PATCH v8 0/8] add " Neil Horman
2016-06-13 8:47 0% ` Pattan, Reshma
2016-06-14 9:38 2% ` [dpdk-dev] [PATCH v9 " Reshma Pattan
2016-06-14 9:38 5% ` [dpdk-dev] [PATCH v9 8/8] doc: update doc for " Reshma Pattan
2016-06-14 20:41 3% ` Thomas Monjalon
2016-06-15 5:44 0% ` Pattan, Reshma
2016-06-15 8:24 0% ` Thomas Monjalon
2016-06-15 14:06 2% ` [dpdk-dev] [PATCH v10 0/7] add " Reshma Pattan
2016-06-15 14:06 10% ` [dpdk-dev] [PATCH v10 3/7] ethdev: add new fields to ethdev info struct Reshma Pattan
2016-06-16 19:14 4% ` Thomas Monjalon
2016-06-15 14:06 4% ` [dpdk-dev] [PATCH v10 4/7] ethdev: make get port by name and get name by port public Reshma Pattan
2016-06-13 11:52 [dpdk-dev] [PATCH] vhost: remove internal lockless enqueue Huawei Xie
2016-06-14 14:07 3% ` Yuanhan Liu
2016-06-17 10:32 [dpdk-dev] [PATCH v2 0/3] Add new KASUMI SW PMD Pablo de Lara
2016-06-20 14:40 ` [dpdk-dev] [PATCH v3 " Pablo de Lara
2016-06-20 14:40 ` [dpdk-dev] [PATCH v3 1/3] kasumi: add new KASUMI PMD Pablo de Lara
2016-07-06 11:26 3% ` Ferruh Yigit
2016-07-06 13:07 0% ` Thomas Monjalon
2016-07-06 13:22 0% ` De Lara Guarch, Pablo
2016-06-17 15:32 3% [dpdk-dev] [RFC] librte_vhost: Add unix domain socket fd registration Aaron Conole
2016-06-17 18:46 [dpdk-dev] [PATCHv8 0/6] Implement pmd hardware support exports Neil Horman
2016-07-04 1:13 ` [dpdk-dev] [PATCH v9 0/7] export PMD infos Thomas Monjalon
2016-07-04 1:14 2% ` [dpdk-dev] [PATCH v9 4/7] pmdinfogen: parse driver to generate code to export Thomas Monjalon
2016-07-04 1:14 2% ` [dpdk-dev] [PATCH v9 7/7] tools: query binaries for support information Thomas Monjalon
2016-06-30 11:57 [dpdk-dev] [RFC] mk: filter duplicate configuration entries Christian Ehrhardt
2016-06-30 12:00 ` [dpdk-dev] [PATCH v2] " Christian Ehrhardt
2016-07-05 16:47 ` Ferruh Yigit
2016-07-05 19:47 ` Thomas Monjalon
2016-07-06 5:37 3% ` Christian Ehrhardt
2016-07-05 15:41 [dpdk-dev] [PATCH 00/18] software parser for packet type Olivier Matz
2016-07-05 15:41 6% ` [dpdk-dev] [PATCH 01/18] doc: add template for release notes 16.11 Olivier Matz
2016-07-05 18:16 2% [dpdk-dev] [RFC] Generic flow director/filtering/classification API Adrien Mazarguil
2016-07-07 7:14 0% ` Lu, Wenzhuo
2016-07-07 10:26 2% ` Adrien Mazarguil
2016-07-19 8:11 0% ` Lu, Wenzhuo
2016-07-19 13:12 0% ` Adrien Mazarguil
2016-07-20 2:16 0% ` Lu, Wenzhuo
2016-07-20 10:41 2% ` Adrien Mazarguil
2016-07-21 3:18 0% ` Lu, Wenzhuo
2016-07-07 23:15 0% ` Chandran, Sugesh
2016-07-08 13:03 0% ` Adrien Mazarguil
2016-07-08 11:11 0% ` Liang, Cunming
2016-07-06 11:39 3% [dpdk-dev] [PATCH] librte_pmd_bond: fix exported symbol versioning Christian Ehrhardt
2016-07-11 11:27 3% ` [dpdk-dev] [PATCH v2] " Christian Ehrhardt
2016-07-11 12:58 0% ` Thomas Monjalon
2016-07-06 14:05 3% [dpdk-dev] [PATCH] cryptodev: move new cryptodev type to bottom of enum Pablo de Lara
2016-07-08 17:52 0% ` Thomas Monjalon
2016-07-07 15:36 [dpdk-dev] [PATCH 00/11] additions to pmdinfogen Thomas Monjalon
2016-07-07 15:36 4% ` [dpdk-dev] [PATCH 11/11] maintainers: add section for pmdinfo Thomas Monjalon
2016-07-07 16:14 0% ` Neil Horman
2016-07-08 10:14 ` [dpdk-dev] [PATCH v2 00/10] additions to pmdinfogen Thomas Monjalon
2016-07-08 10:14 4% ` [dpdk-dev] [PATCH v2 10/10] maintainers: add section for pmdinfo Thomas Monjalon
2016-07-08 14:42 ` [dpdk-dev] [PATCH v3 00/10] additions to pmdinfogen Thomas Monjalon
2016-07-08 14:42 4% ` [dpdk-dev] [PATCH v3 10/10] maintainers: add section for pmdinfo Thomas Monjalon
2016-07-14 13:29 [dpdk-dev] rte_ether: Driver-specific stats getting overwritten Remy Horton
2016-07-14 13:37 ` Thomas Monjalon
2016-07-14 13:51 ` Igor Ryzhov
2016-07-14 15:50 3% ` Remy Horton
2016-07-19 12:42 [dpdk-dev] [PATCH] rte_delay_us can be replaced with user function jozmarti
2016-07-19 13:17 3% ` Wiles, Keith
2016-07-19 13:16 13% [dpdk-dev] [PATCH v1] doc: fix release notes for 16.07 John McNamara
2016-07-19 14:01 13% [dpdk-dev] [PATCH] doc: announce ABI change for mbuf structure Olivier Matz
2016-07-19 14:40 4% ` Bruce Richardson
2016-07-19 15:04 7% ` Olivier Matz
2016-07-19 15:07 4% ` Richardson, Bruce
2016-07-19 15:28 4% ` Olivier Matz
2016-07-20 7:16 13% ` [dpdk-dev] [PATCH v2] " Olivier Matz
2016-07-20 8:54 4% ` Ferruh Yigit
2016-07-27 8:33 4% ` Thomas Monjalon
2016-07-28 18:04 4% ` Thomas Monjalon
2016-07-27 9:34 4% ` Ananyev, Konstantin
2016-07-28 2:35 4% ` John Daley (johndale)
2016-07-28 2:39 4% ` Jerin Jacob
2016-07-19 14:37 [dpdk-dev] [PATCH] mempool: fix lack of free() registration Zoltan Kiss
2016-07-19 14:37 3% ` [dpdk-dev] [PATCH] mempool: adjust name string size in related data types Zoltan Kiss
2016-07-19 15:37 4% ` Olivier Matz
2016-07-19 15:59 3% ` Zoltan Kiss
2016-07-19 16:17 0% ` Olivier Matz
2016-07-20 12:41 4% ` Zoltan Kiss
2016-07-20 13:37 4% ` Olivier Matz
2016-07-20 14:01 0% ` Richardson, Bruce
2016-07-20 17:20 0% ` Zoltan Kiss
2016-07-20 17:16 12% ` [dpdk-dev] [PATCH v2] " Zoltan Kiss
2016-07-21 13:40 0% ` Olivier Matz
2016-07-21 13:47 0% ` Zoltan Kiss
2016-07-21 14:25 0% ` Olivier Matz
2016-07-21 21:16 0% ` Thomas Monjalon
2016-07-20 14:24 3% [dpdk-dev] [PATCH] unify tools naming Thomas Monjalon
2016-07-20 14:24 13% [dpdk-dev] [PATCH] doc: announce ABI change for rte_eth_dev structure Tomasz Kulasek
2016-07-20 15:01 4% ` Thomas Monjalon
2016-07-20 15:13 7% ` Ananyev, Konstantin
2016-07-20 15:22 7% ` Thomas Monjalon
2016-07-20 15:42 4% ` Kulasek, TomaszX
2016-07-21 15:24 11% ` [dpdk-dev] [PATCH v2] " Tomasz Kulasek
2016-07-21 22:48 4% ` Ananyev, Konstantin
2016-07-27 8:59 4% ` Thomas Monjalon
2016-07-27 17:10 4% ` Jerin Jacob
2016-07-27 17:33 4% ` Ananyev, Konstantin
2016-07-27 17:41 4% ` Jerin Jacob
2016-07-27 20:51 4% ` Ananyev, Konstantin
2016-07-28 2:13 4% ` Jerin Jacob
2016-07-28 10:36 4% ` Ananyev, Konstantin
2016-07-28 11:38 4% ` Jerin Jacob
2016-07-28 12:07 4% ` Avi Kivity
2016-07-28 13:01 4% ` Ananyev, Konstantin
2016-07-28 13:58 4% ` Olivier MATZ
2016-07-28 14:21 4% ` Ananyev, Konstantin
2016-07-28 13:59 4% ` Jerin Jacob
2016-07-28 14:52 4% ` Thomas Monjalon
2016-07-28 16:25 7% ` Jerin Jacob
2016-07-28 17:07 4% ` Thomas Monjalon
2016-07-28 12:04 4% ` Avi Kivity
2016-07-31 7:46 4% ` [dpdk-dev] [PATCH] " Vlad Zolotarov
2016-07-20 16:35 [dpdk-dev] [PATCH] doc: announce ivshmem support removal Thomas Monjalon
2016-07-27 19:08 ` [dpdk-dev] " Jan Viktorin
2016-07-28 9:20 3% ` Christian Ehrhardt
2016-07-28 15:23 0% ` Mauricio Vasquez
2016-07-20 17:09 9% [dpdk-dev] [PATCH] validate_abi: build faster by augmenting make with job count Neil Horman
2016-07-20 17:40 Thomas Monjalon
2016-07-20 19:02 9% ` [dpdk-dev] [PATCH v2] " Neil Horman
2016-07-26 16:22 [dpdk-dev] [PATCH] doc: announce renaming of ethdev library Thomas Monjalon
2016-07-27 16:33 ` [dpdk-dev] " Jan Viktorin
2016-07-28 9:29 3% ` Christian Ehrhardt
2016-07-28 9:52 0% ` Thomas Monjalon
2016-07-27 9:15 3% [dpdk-dev] last days for deprecation notices Thomas Monjalon
2016-07-27 22:14 3% [dpdk-dev] [bug] dpdk-vfio: Invalid region/index assumption Alex Williamson
2016-07-28 6:54 0% ` Thomas Monjalon
2016-07-28 9:42 ` Burakov, Anatoly
2016-07-28 14:54 3% ` Alex Williamson
2016-07-28 8:06 0% ` Santosh Shukla
2016-07-28 8:29 4% [dpdk-dev] removal of old deprecation notice for Chelsio filtering Thomas Monjalon
2016-07-28 10:12 0% ` Rahul Lakkireddy
2016-07-28 10:15 7% [dpdk-dev] [PATCH] doc: remove deprecation notice related to new flow types Rahul Lakkireddy
2016-07-28 12:33 3% [dpdk-dev] DPDK Stable Releases and Long Term Support Mcnamara, John
2016-07-29 11:23 6% [dpdk-dev] [PATCH v1] doc: add template release notes for 16.11 John McNamara
2016-07-29 12:28 1% [dpdk-dev] [PATCH] ivshmem: remove integration in dpdk David Marchand
2016-07-29 13:01 9% [dpdk-dev] [PATCH] log: remove history dump Thomas Monjalon
2016-07-29 13:50 9% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
2016-07-29 13:41 8% [dpdk-dev] [PATCH] doc: postpone mempool ABI breakage Thomas Monjalon
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).