DPDK patches and discussions
 help / color / mirror / Atom feed
Search results ordered by [date|relevance]  view[summary|nested|Atom feed]
thread overview below | download: 
* [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check
    @ 2016-03-10 10:53 15% ` Panu Matilainen
  2016-03-10 12:22  4%   ` Ferruh Yigit
  2016-03-10 12:52  7% ` [dpdk-dev] [PATCH 1/3] scripts: support parallel building in validate-abi.sh via -j[N] option Ferruh Yigit
  2 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-03-10 10:53 UTC (permalink / raw)
  To: dev

When doing multiple runs of validate-abi.sh, the git status check
will more often than not unnecessarily fail with "Working directory not
clean" error because of the compat_result and compile target directories
from the previous run. Filter out the self-generated directories
when checking.

Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
---
 scripts/validate-abi.sh | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
index ea60639..a21f883 100755
--- a/scripts/validate-abi.sh
+++ b/scripts/validate-abi.sh
@@ -163,8 +163,7 @@ log "INFO" "against DPDK DSOs built from version $TAG2."
 log "INFO" ""
 
 # Check to make sure we have a clean tree
-git status | grep -q clean
-if [ $? -ne 0 ]
+if [ $(git status --porcelain | grep -vE "($TARGET|compat_reports)" | wc -l)  -ne 0 ]
 then
 	log "WARN" "Working directory not clean, aborting"
 	cleanup_and_exit 1
-- 
2.5.0

^ permalink raw reply	[relevance 15%]

* [dpdk-dev] [PATCH v3 0/2] add support for buffered tx to ethdev
  @ 2016-03-10 10:57  4% ` Tomasz Kulasek
  2016-03-10 11:31  0%   ` Ananyev, Konstantin
  2016-03-10 17:19  4%   ` [dpdk-dev] [PATCH v4 " Tomasz Kulasek
  0 siblings, 2 replies; 200+ results
From: Tomasz Kulasek @ 2016-03-10 10:57 UTC (permalink / raw)
  To: dev

Many sample apps include internal buffering for single-packet-at-a-time
operation. Since this is such a common paradigm, this functionality is
better suited to being implemented in the ethdev API.

The new APIs in the ethdev library are:
* rte_eth_tx_buffer_init - initialize buffer
* rte_eth_tx_buffer - buffer up a single packet for future transmission
* rte_eth_tx_buffer_flush - flush any unsent buffered packets
* rte_eth_tx_buffer_set_err_callback - set up a callback to be called in
  case transmitting a buffered burst fails. By default, we just free the
  unsent packets.

As well as these, an additional reference callbacks are provided, which
frees the packets:

* rte_eth_tx_buffer_drop_callback - silently drop packets (default
  behavior)
* rte_eth_tx_buffer_count_callback - drop and update user-provided counter
  to track the number of dropped packets

Due to the feedback from mailing list, that buffer management facilities
in the user application are more preferable than API simplicity, we decided
to move internal buffer table, as well as callback functions and user data,
from rte_eth_dev/rte_eth_dev_data to the application space.
It prevents ABI breakage and gives some more flexibility in the buffer's
management such as allocation, dynamical size change, reuse buffers on many
ports or after fail, and so on.


The following steps illustrate how tx buffers can be used in application:

1) Initialization

a) Allocate memory for a buffer

   struct rte_eth_dev_tx_buffer *buffer = rte_zmalloc_socket("tx_buffer",
           RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0, socket_id);

   RTE_ETH_TX_BUFFER_SIZE(size) macro computes memory required to store
   "size" packets in buffer.

b) Initialize allocated memory and set up default values. Threshold level
   must be lower than or equal to the MAX_PKT_BURST from 1a)

   rte_eth_tx_buffer_init(buffer, threshold);


c) Set error callback (optional)

   rte_eth_tx_buffer_set_err_callback(buffer, callback_fn, userdata);


2) Store packet "pkt" in buffer and send them all to the queue_id on
   port_id when number of packets reaches threshold level set up in 1b)

   rte_eth_tx_buffer(port_id, queue_id, buffer, pkt);


3) Send all stored packets to the queue_id on port_id

   rte_eth_tx_buffer_flush(port_id, queue_id, buffer);


4) Flush buffer and free memory

   rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
   ...
   rte_free(buffer);

v3 changes:
 - error counter removed from tx buffer structure, now default behavior is
   silent drop of unsent packets
 - some names was changed in tx buffer structure to be more descriptive
 - two default calbacks are provided: rte_eth_tx_buffer_drop_callback and
   rte_eth_tx_buffer_count_callback

v2 changes:
 - reworked to use new buffer model
 - buffer data and callbacks are removed from rte_eth_dev/rte_eth_dev_data,
   so this patch doesn't brake an ABI anymore
 - introduced RTE_ETH_TX_BUFFER macro and rte_eth_tx_buffer_init
 - buffers are not attached to the port-queue
 - buffers can be allocated dynamically during application work
 - size of buffer can be changed without port restart

Tomasz Kulasek (2):
  ethdev: add buffered tx api
  examples: rework to use buffered tx

 examples/l2fwd-jobstats/main.c                     |  104 ++++------
 examples/l2fwd-keepalive/main.c                    |  100 ++++------
 examples/l2fwd/main.c                              |  104 ++++------
 examples/l3fwd-acl/main.c                          |   92 ++++-----
 examples/l3fwd-power/main.c                        |   89 ++++-----
 examples/link_status_interrupt/main.c              |  107 ++++------
 .../client_server_mp/mp_client/client.c            |  101 ++++++----
 examples/multi_process/l2fwd_fork/main.c           |   97 ++++-----
 examples/packet_ordering/main.c                    |  122 ++++++++----
 examples/qos_meter/main.c                          |   61 ++----
 lib/librte_ether/rte_ethdev.c                      |   46 +++++
 lib/librte_ether/rte_ethdev.h                      |  205 +++++++++++++++++++-
 lib/librte_ether/rte_ether_version.map             |   10 +
 13 files changed, 696 insertions(+), 542 deletions(-)

-- 
1.7.9.5

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH v3 0/2] add support for buffered tx to ethdev
  2016-03-10 10:57  4% ` [dpdk-dev] [PATCH v3 " Tomasz Kulasek
@ 2016-03-10 11:31  0%   ` Ananyev, Konstantin
  2016-03-10 16:01  0%     ` Jastrzebski, MichalX K
  2016-03-10 17:19  4%   ` [dpdk-dev] [PATCH v4 " Tomasz Kulasek
  1 sibling, 1 reply; 200+ results
From: Ananyev, Konstantin @ 2016-03-10 11:31 UTC (permalink / raw)
  To: Kulasek, TomaszX, dev

> Many sample apps include internal buffering for single-packet-at-a-time
> operation. Since this is such a common paradigm, this functionality is
> better suited to being implemented in the ethdev API.
> 
> The new APIs in the ethdev library are:
> * rte_eth_tx_buffer_init - initialize buffer
> * rte_eth_tx_buffer - buffer up a single packet for future transmission
> * rte_eth_tx_buffer_flush - flush any unsent buffered packets
> * rte_eth_tx_buffer_set_err_callback - set up a callback to be called in
>   case transmitting a buffered burst fails. By default, we just free the
>   unsent packets.
> 
> As well as these, an additional reference callbacks are provided, which
> frees the packets:
> 
> * rte_eth_tx_buffer_drop_callback - silently drop packets (default
>   behavior)
> * rte_eth_tx_buffer_count_callback - drop and update user-provided counter
>   to track the number of dropped packets
> 
> Due to the feedback from mailing list, that buffer management facilities
> in the user application are more preferable than API simplicity, we decided
> to move internal buffer table, as well as callback functions and user data,
> from rte_eth_dev/rte_eth_dev_data to the application space.
> It prevents ABI breakage and gives some more flexibility in the buffer's
> management such as allocation, dynamical size change, reuse buffers on many
> ports or after fail, and so on.
> 
> 
> The following steps illustrate how tx buffers can be used in application:
> 
> 1) Initialization
> 
> a) Allocate memory for a buffer
> 
>    struct rte_eth_dev_tx_buffer *buffer = rte_zmalloc_socket("tx_buffer",
>            RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0, socket_id);
> 
>    RTE_ETH_TX_BUFFER_SIZE(size) macro computes memory required to store
>    "size" packets in buffer.
> 
> b) Initialize allocated memory and set up default values. Threshold level
>    must be lower than or equal to the MAX_PKT_BURST from 1a)
> 
>    rte_eth_tx_buffer_init(buffer, threshold);
> 
> 
> c) Set error callback (optional)
> 
>    rte_eth_tx_buffer_set_err_callback(buffer, callback_fn, userdata);
> 
> 
> 2) Store packet "pkt" in buffer and send them all to the queue_id on
>    port_id when number of packets reaches threshold level set up in 1b)
> 
>    rte_eth_tx_buffer(port_id, queue_id, buffer, pkt);
> 
> 
> 3) Send all stored packets to the queue_id on port_id
> 
>    rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
> 
> 
> 4) Flush buffer and free memory
> 
>    rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
>    ...
>    rte_free(buffer);
> 
> v3 changes:
>  - error counter removed from tx buffer structure, now default behavior is
>    silent drop of unsent packets
>  - some names was changed in tx buffer structure to be more descriptive
>  - two default calbacks are provided: rte_eth_tx_buffer_drop_callback and
>    rte_eth_tx_buffer_count_callback
> 
> v2 changes:
>  - reworked to use new buffer model
>  - buffer data and callbacks are removed from rte_eth_dev/rte_eth_dev_data,
>    so this patch doesn't brake an ABI anymore
>  - introduced RTE_ETH_TX_BUFFER macro and rte_eth_tx_buffer_init
>  - buffers are not attached to the port-queue
>  - buffers can be allocated dynamically during application work
>  - size of buffer can be changed without port restart
> 
> Tomasz Kulasek (2):
>   ethdev: add buffered tx api
>   examples: rework to use buffered tx
> 
>  examples/l2fwd-jobstats/main.c                     |  104 ++++------
>  examples/l2fwd-keepalive/main.c                    |  100 ++++------
>  examples/l2fwd/main.c                              |  104 ++++------
>  examples/l3fwd-acl/main.c                          |   92 ++++-----
>  examples/l3fwd-power/main.c                        |   89 ++++-----
>  examples/link_status_interrupt/main.c              |  107 ++++------
>  .../client_server_mp/mp_client/client.c            |  101 ++++++----
>  examples/multi_process/l2fwd_fork/main.c           |   97 ++++-----
>  examples/packet_ordering/main.c                    |  122 ++++++++----
>  examples/qos_meter/main.c                          |   61 ++----
>  lib/librte_ether/rte_ethdev.c                      |   46 +++++
>  lib/librte_ether/rte_ethdev.h                      |  205 +++++++++++++++++++-
>  lib/librte_ether/rte_ether_version.map             |   10 +
>  13 files changed, 696 insertions(+), 542 deletions(-)
> 
> --

Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>

> 1.7.9.5

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
@ 2016-03-10 11:55  7% David Hunt
  2016-03-10 12:37  4% ` Olivier MATZ
  2016-03-10 16:23  4% ` Mcnamara, John
  0 siblings, 2 replies; 200+ results
From: David Hunt @ 2016-03-10 11:55 UTC (permalink / raw)
  To: dev

Announce the ABI breakage due to addition of external mempool
manager functionality which requires changes to rte_mempool
structure.

Signed-off-by: David Hunt <david.hunt@intel.com>
---
 doc/guides/rel_notes/deprecation.rst | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 9930b5a..263e652 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -39,3 +39,13 @@ Deprecation Notices
   and table action handlers will be updated:
   the pipeline parameter will be added, the packets mask parameter will be
   either removed (for input port action handler) or made input-only.
+
+* librte_mempool: 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.
+
-- 
2.5.0

^ permalink raw reply	[relevance 7%]

* Re: [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check
  2016-03-10 10:53 15% ` [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check Panu Matilainen
@ 2016-03-10 12:22  4%   ` Ferruh Yigit
  2016-03-10 12:29  4%     ` Panu Matilainen
  0 siblings, 1 reply; 200+ results
From: Ferruh Yigit @ 2016-03-10 12:22 UTC (permalink / raw)
  To: Panu Matilainen, dev

On 3/10/2016 10:53 AM, Panu Matilainen wrote:
> When doing multiple runs of validate-abi.sh, the git status check
> will more often than not unnecessarily fail with "Working directory not
> clean" error because of the compat_result and compile target directories
> from the previous run. Filter out the self-generated directories
> when checking.
> 
> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
> ---
>  scripts/validate-abi.sh | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
> index ea60639..a21f883 100755
> --- a/scripts/validate-abi.sh
> +++ b/scripts/validate-abi.sh
> @@ -163,8 +163,7 @@ log "INFO" "against DPDK DSOs built from version $TAG2."
>  log "INFO" ""
>  
>  # Check to make sure we have a clean tree
> -git status | grep -q clean
> -if [ $? -ne 0 ]
> +if [ $(git status --porcelain | grep -vE "($TARGET|compat_reports)" | wc -l)  -ne 0 ]
>  then
>  	log "WARN" "Working directory not clean, aborting"
>  	cleanup_and_exit 1
> 
Hi Panu,

This check catches untracked files too, does it makes sense to limit
error only to modified files (local or staged)?

This also prevents specific "compat_reports" folder check.

And of course mentioned change requires "git clean -fd" removed, or
replaced with "make clean"

Thanks,
ferruh

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 2/3] scripts: avoid editing defconfig_* files in validate-abi.sh
  @ 2016-03-10 12:25  4%   ` Ferruh Yigit
  2016-03-10 12:36  4%     ` Panu Matilainen
  0 siblings, 1 reply; 200+ results
From: Ferruh Yigit @ 2016-03-10 12:25 UTC (permalink / raw)
  To: Panu Matilainen, dev

On 3/10/2016 10:53 AM, Panu Matilainen wrote:
> The defconfig_* files are templates which are not supposed to be
> edited, and doing so tends to leave unwanted cruft behind. Edit
> the "working copy" config instead, which is the intended DPDK way.
> 
> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
> ---
>  scripts/validate-abi.sh | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
> 
> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
> index f094582..ea60639 100755
> --- a/scripts/validate-abi.sh
> +++ b/scripts/validate-abi.sh
> @@ -90,11 +90,11 @@ cleanup_and_exit() {
>  # Make sure we configure SHARED libraries
>  # Also turn off IGB and KNI as those require kernel headers to build
>  fixup_config() {
> -	sed -i -e"$ a\CONFIG_RTE_BUILD_SHARED_LIB=y" config/defconfig_$TARGET
> -	sed -i -e"$ a\CONFIG_RTE_NEXT_ABI=n" config/defconfig_$TARGET
> -	sed -i -e"$ a\CONFIG_RTE_EAL_IGB_UIO=n" config/defconfig_$TARGET
> -	sed -i -e"$ a\CONFIG_RTE_LIBRTE_KNI=n" config/defconfig_$TARGET
> -	sed -i -e"$ a\CONFIG_RTE_KNI_KMOD=n" config/defconfig_$TARGET
> +	sed -i -e"$ a\CONFIG_RTE_BUILD_SHARED_LIB=y" $TARGET/.config
> +	sed -i -e"$ a\CONFIG_RTE_NEXT_ABI=n" $TARGET/.config
> +	sed -i -e"$ a\CONFIG_RTE_EAL_IGB_UIO=n" $TARGET/.config
> +	sed -i -e"$ a\CONFIG_RTE_LIBRTE_KNI=n" $TARGET/.config
> +	sed -i -e"$ a\CONFIG_RTE_KNI_KMOD=n" $TARGET/.config
>  }
>  
Since working .config copy updated, is there a benefit to save and
restore the original .config file (if exists)?

And is "git reset --hard" still required in the script?

Thanks,
ferruh

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check
  2016-03-10 12:22  4%   ` Ferruh Yigit
@ 2016-03-10 12:29  4%     ` Panu Matilainen
  2016-03-10 12:34  4%       ` Ferruh Yigit
  0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-03-10 12:29 UTC (permalink / raw)
  To: Ferruh Yigit, dev

On 03/10/2016 02:22 PM, Ferruh Yigit wrote:
> On 3/10/2016 10:53 AM, Panu Matilainen wrote:
>> When doing multiple runs of validate-abi.sh, the git status check
>> will more often than not unnecessarily fail with "Working directory not
>> clean" error because of the compat_result and compile target directories
>> from the previous run. Filter out the self-generated directories
>> when checking.
>>
>> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
>> ---
>>   scripts/validate-abi.sh | 3 +--
>>   1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
>> index ea60639..a21f883 100755
>> --- a/scripts/validate-abi.sh
>> +++ b/scripts/validate-abi.sh
>> @@ -163,8 +163,7 @@ log "INFO" "against DPDK DSOs built from version $TAG2."
>>   log "INFO" ""
>>
>>   # Check to make sure we have a clean tree
>> -git status | grep -q clean
>> -if [ $? -ne 0 ]
>> +if [ $(git status --porcelain | grep -vE "($TARGET|compat_reports)" | wc -l)  -ne 0 ]
>>   then
>>   	log "WARN" "Working directory not clean, aborting"
>>   	cleanup_and_exit 1
>>
> Hi Panu,
>
> This check catches untracked files too, does it makes sense to limit
> error only to modified files (local or staged)?

I did ponder about that, untracked files seem mostly harmless in this 
picture but erred on the side of caution.

>
> This also prevents specific "compat_reports" folder check.
>
> And of course mentioned change requires "git clean -fd" removed, or
> replaced with "make clean"

Sorry, I dont understand you mean by these two comments.

	- Panu -

> Thanks,
> ferruh
>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check
  2016-03-10 12:29  4%     ` Panu Matilainen
@ 2016-03-10 12:34  4%       ` Ferruh Yigit
  2016-03-10 12:39  4%         ` Panu Matilainen
  0 siblings, 1 reply; 200+ results
From: Ferruh Yigit @ 2016-03-10 12:34 UTC (permalink / raw)
  To: Panu Matilainen, dev

On 3/10/2016 12:29 PM, Panu Matilainen wrote:
> On 03/10/2016 02:22 PM, Ferruh Yigit wrote:
>> On 3/10/2016 10:53 AM, Panu Matilainen wrote:
>>> When doing multiple runs of validate-abi.sh, the git status check
>>> will more often than not unnecessarily fail with "Working directory not
>>> clean" error because of the compat_result and compile target directories
>>> from the previous run. Filter out the self-generated directories
>>> when checking.
>>>
>>> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
>>> ---
>>>   scripts/validate-abi.sh | 3 +--
>>>   1 file changed, 1 insertion(+), 2 deletions(-)
>>>
>>> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
>>> index ea60639..a21f883 100755
>>> --- a/scripts/validate-abi.sh
>>> +++ b/scripts/validate-abi.sh
>>> @@ -163,8 +163,7 @@ log "INFO" "against DPDK DSOs built from version $TAG2."
>>>   log "INFO" ""
>>>
>>>   # Check to make sure we have a clean tree
>>> -git status | grep -q clean
>>> -if [ $? -ne 0 ]
>>> +if [ $(git status --porcelain | grep -vE "($TARGET|compat_reports)" | wc -l)  -ne 0 ]
>>>   then
>>>   	log "WARN" "Working directory not clean, aborting"
>>>   	cleanup_and_exit 1
>>>
>> Hi Panu,
>>
>> This check catches untracked files too, does it makes sense to limit
>> error only to modified files (local or staged)?
> 
> I did ponder about that, untracked files seem mostly harmless in this 
> picture but erred on the side of caution.
> 
This is something prevents me running script from working tree, and
forces to create a new clone.

>>
>> This also prevents specific "compat_reports" folder check.
>>
>> And of course mentioned change requires "git clean -fd" removed, or
>> replaced with "make clean"
> 
> Sorry, I dont understand you mean by these two comments.
> 
If untracked files accepted by script, "compat_reports" exclusion is no
more required, and "git clean -fd" needs to removed from script.

Regards,
ferruh

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 2/3] scripts: avoid editing defconfig_* files in validate-abi.sh
  2016-03-10 12:25  4%   ` Ferruh Yigit
@ 2016-03-10 12:36  4%     ` Panu Matilainen
  0 siblings, 0 replies; 200+ results
From: Panu Matilainen @ 2016-03-10 12:36 UTC (permalink / raw)
  To: Ferruh Yigit, dev

On 03/10/2016 02:25 PM, Ferruh Yigit wrote:
> On 3/10/2016 10:53 AM, Panu Matilainen wrote:
>> The defconfig_* files are templates which are not supposed to be
>> edited, and doing so tends to leave unwanted cruft behind. Edit
>> the "working copy" config instead, which is the intended DPDK way.
>>
>> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
>> ---
>>   scripts/validate-abi.sh | 18 +++++++++---------
>>   1 file changed, 9 insertions(+), 9 deletions(-)
>>
>> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
>> index f094582..ea60639 100755
>> --- a/scripts/validate-abi.sh
>> +++ b/scripts/validate-abi.sh
>> @@ -90,11 +90,11 @@ cleanup_and_exit() {
>>   # Make sure we configure SHARED libraries
>>   # Also turn off IGB and KNI as those require kernel headers to build
>>   fixup_config() {
>> -	sed -i -e"$ a\CONFIG_RTE_BUILD_SHARED_LIB=y" config/defconfig_$TARGET
>> -	sed -i -e"$ a\CONFIG_RTE_NEXT_ABI=n" config/defconfig_$TARGET
>> -	sed -i -e"$ a\CONFIG_RTE_EAL_IGB_UIO=n" config/defconfig_$TARGET
>> -	sed -i -e"$ a\CONFIG_RTE_LIBRTE_KNI=n" config/defconfig_$TARGET
>> -	sed -i -e"$ a\CONFIG_RTE_KNI_KMOD=n" config/defconfig_$TARGET
>> +	sed -i -e"$ a\CONFIG_RTE_BUILD_SHARED_LIB=y" $TARGET/.config
>> +	sed -i -e"$ a\CONFIG_RTE_NEXT_ABI=n" $TARGET/.config
>> +	sed -i -e"$ a\CONFIG_RTE_EAL_IGB_UIO=n" $TARGET/.config
>> +	sed -i -e"$ a\CONFIG_RTE_LIBRTE_KNI=n" $TARGET/.config
>> +	sed -i -e"$ a\CONFIG_RTE_KNI_KMOD=n" $TARGET/.config
>>   }
>>
> Since working .config copy updated, is there a benefit to save and
> restore the original .config file (if exists)?

Maybe, but it already is being modified and not restored, so that 
behavior is not new in this patch.

> And is "git reset --hard" still required in the script?

Perhaps not, strictly speaking. It does ensure there are no other 
artifacts left around though.

	- Panu -

> Thanks,
> ferruh
>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 11:55  7% [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice David Hunt
@ 2016-03-10 12:37  4% ` Olivier MATZ
  2016-03-10 13:15  4%   ` Bruce Richardson
  2016-03-10 16:23  4% ` Mcnamara, John
  1 sibling, 1 reply; 200+ results
From: Olivier MATZ @ 2016-03-10 12:37 UTC (permalink / raw)
  To: David Hunt, dev

Hi David,

On 03/10/2016 12:55 PM, David Hunt wrote:
> Announce the ABI breakage due to addition of external mempool
> manager functionality which requires changes to rte_mempool
> structure.
> 
> Signed-off-by: David Hunt <david.hunt@intel.com>

Acked-by: Olivier Matz <olivier.matz@6wind.com>

Please note we need not 2 more acks for this kind of deprecation
notices.

Thanks,
Olivier

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check
  2016-03-10 12:34  4%       ` Ferruh Yigit
@ 2016-03-10 12:39  4%         ` Panu Matilainen
  2016-03-10 12:47  4%           ` Ferruh Yigit
  0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-03-10 12:39 UTC (permalink / raw)
  To: Ferruh Yigit, dev

On 03/10/2016 02:34 PM, Ferruh Yigit wrote:
> On 3/10/2016 12:29 PM, Panu Matilainen wrote:
>> On 03/10/2016 02:22 PM, Ferruh Yigit wrote:
>>> On 3/10/2016 10:53 AM, Panu Matilainen wrote:
>>>> When doing multiple runs of validate-abi.sh, the git status check
>>>> will more often than not unnecessarily fail with "Working directory not
>>>> clean" error because of the compat_result and compile target directories
>>>> from the previous run. Filter out the self-generated directories
>>>> when checking.
>>>>
>>>> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
>>>> ---
>>>>    scripts/validate-abi.sh | 3 +--
>>>>    1 file changed, 1 insertion(+), 2 deletions(-)
>>>>
>>>> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
>>>> index ea60639..a21f883 100755
>>>> --- a/scripts/validate-abi.sh
>>>> +++ b/scripts/validate-abi.sh
>>>> @@ -163,8 +163,7 @@ log "INFO" "against DPDK DSOs built from version $TAG2."
>>>>    log "INFO" ""
>>>>
>>>>    # Check to make sure we have a clean tree
>>>> -git status | grep -q clean
>>>> -if [ $? -ne 0 ]
>>>> +if [ $(git status --porcelain | grep -vE "($TARGET|compat_reports)" | wc -l)  -ne 0 ]
>>>>    then
>>>>    	log "WARN" "Working directory not clean, aborting"
>>>>    	cleanup_and_exit 1
>>>>
>>> Hi Panu,
>>>
>>> This check catches untracked files too, does it makes sense to limit
>>> error only to modified files (local or staged)?
>>
>> I did ponder about that, untracked files seem mostly harmless in this
>> picture but erred on the side of caution.
>>
> This is something prevents me running script from working tree, and
> forces to create a new clone.

Hmm, what untracked files you typically have in your working tree then?

>>>
>>> This also prevents specific "compat_reports" folder check.
>>>
>>> And of course mentioned change requires "git clean -fd" removed, or
>>> replaced with "make clean"
>>
>> Sorry, I dont understand you mean by these two comments.
>>
> If untracked files accepted by script, "compat_reports" exclusion is no
> more required, and "git clean -fd" needs to removed from script.

Ah, sure. Thanks for clarifying.

	- Panu -

>
> Regards,
> ferruh
>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check
  2016-03-10 12:39  4%         ` Panu Matilainen
@ 2016-03-10 12:47  4%           ` Ferruh Yigit
  0 siblings, 0 replies; 200+ results
From: Ferruh Yigit @ 2016-03-10 12:47 UTC (permalink / raw)
  To: Panu Matilainen, dev

On 3/10/2016 12:39 PM, Panu Matilainen wrote:
> On 03/10/2016 02:34 PM, Ferruh Yigit wrote:
>> On 3/10/2016 12:29 PM, Panu Matilainen wrote:
>>> On 03/10/2016 02:22 PM, Ferruh Yigit wrote:
>>>> On 3/10/2016 10:53 AM, Panu Matilainen wrote:
>>>>> When doing multiple runs of validate-abi.sh, the git status check
>>>>> will more often than not unnecessarily fail with "Working directory not
>>>>> clean" error because of the compat_result and compile target directories
>>>>> from the previous run. Filter out the self-generated directories
>>>>> when checking.
>>>>>
>>>>> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
>>>>> ---
>>>>>    scripts/validate-abi.sh | 3 +--
>>>>>    1 file changed, 1 insertion(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
>>>>> index ea60639..a21f883 100755
>>>>> --- a/scripts/validate-abi.sh
>>>>> +++ b/scripts/validate-abi.sh
>>>>> @@ -163,8 +163,7 @@ log "INFO" "against DPDK DSOs built from version $TAG2."
>>>>>    log "INFO" ""
>>>>>
>>>>>    # Check to make sure we have a clean tree
>>>>> -git status | grep -q clean
>>>>> -if [ $? -ne 0 ]
>>>>> +if [ $(git status --porcelain | grep -vE "($TARGET|compat_reports)" | wc -l)  -ne 0 ]
>>>>>    then
>>>>>    	log "WARN" "Working directory not clean, aborting"
>>>>>    	cleanup_and_exit 1
>>>>>
>>>> Hi Panu,
>>>>
>>>> This check catches untracked files too, does it makes sense to limit
>>>> error only to modified files (local or staged)?
>>>
>>> I did ponder about that, untracked files seem mostly harmless in this
>>> picture but erred on the side of caution.
>>>
>> This is something prevents me running script from working tree, and
>> forces to create a new clone.
> 
> Hmm, what untracked files you typically have in your working tree then?
> 

cscope.out, various sym links, perf.data, and some more J, I want to
keep in working directory.

>>>>
>>>> This also prevents specific "compat_reports" folder check.
>>>>
>>>> And of course mentioned change requires "git clean -fd" removed, or
>>>> replaced with "make clean"
>>>
>>> Sorry, I dont understand you mean by these two comments.
>>>
>> If untracked files accepted by script, "compat_reports" exclusion is no
>> more required, and "git clean -fd" needs to removed from script.
> 
> Ah, sure. Thanks for clarifying.
> 
> 	- Panu -
> 
>>
>> Regards,
>> ferruh
>>
> 

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 1/3] scripts: support parallel building in validate-abi.sh via -j[N] option
      2016-03-10 10:53 15% ` [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check Panu Matilainen
@ 2016-03-10 12:52  7% ` Ferruh Yigit
  2 siblings, 0 replies; 200+ results
From: Ferruh Yigit @ 2016-03-10 12:52 UTC (permalink / raw)
  To: Panu Matilainen, dev

On 3/10/2016 10:53 AM, Panu Matilainen wrote:
> Signed-off-by: Panu Matilainen <pmatilai@redhat.com>
> ---
>  doc/guides/contributing/versioning.rst |  4 +++-
>  scripts/validate-abi.sh                | 13 ++++++++++---
>  2 files changed, 13 insertions(+), 4 deletions(-)
> 
> diff --git a/doc/guides/contributing/versioning.rst b/doc/guides/contributing/versioning.rst
> index ae10a98..33b03a1 100644
> --- a/doc/guides/contributing/versioning.rst
> +++ b/doc/guides/contributing/versioning.rst
> @@ -469,11 +469,13 @@ utilities which can be installed via a package manager. For example::
>  
>  The syntax of the ``validate-abi.sh`` utility is::
>  
> -   ./scripts/validate-abi.sh <REV1> <REV2> <TARGET>
> +   ./scripts/validate-abi.sh [-j[N]] <REV1> <REV2> <TARGET>
>  
>  Where ``REV1`` and ``REV2`` are valid gitrevisions(7)
>  https://www.kernel.org/pub/software/scm/git/docs/gitrevisions.html
>  on the local repo and target is the usual DPDK compilation target.
> +The optional -j[N] switch enables parallel building with at most
> +N simultaneous jobs, ie the same as -j option of ``make``.
>  
>  For example:
>  
> diff --git a/scripts/validate-abi.sh b/scripts/validate-abi.sh
> index c36ad61..f094582 100755
> --- a/scripts/validate-abi.sh
> +++ b/scripts/validate-abi.sh
> @@ -27,13 +27,20 @@
>  #   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
>  #   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
>  
> +case "$1" in
> +    -j*)
> +        MAKEJOBS="$1"
> +        shift
> +        ;;
> +esac
> +
>  TAG1=$1
>  TAG2=$2
>  TARGET=$3
>  ABI_DIR=`mktemp -d -p /tmp ABI.XXXXXX`
>  
>  usage() {
> -	echo "$0 <REV1> <REV2> <TARGET>"
> +	echo "$0 [-j[N]] <REV1> <REV2> <TARGET>"
>  }
>  
>  log() {
> @@ -183,7 +190,7 @@ 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 $MAKEJOBS O=$TARGET > $VERBOSE 2>&1
>  
>  if [ $? -ne 0 ]
>  then
> @@ -214,7 +221,7 @@ 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 $MAKEJOBS O=$TARGET > $VERBOSE 2>&1
>  
>  if [ $? -ne 0 ]
>  then
> 

This is something good to have.

I also found following is also working, not sure if need to document in
script:
"MAKEFLAGS="-j32" ./validate-abi.sh X Y Z"

And was just thinking if needs to address "-j N" usage (space after -j),
it is your call. Other than this,

Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>

^ permalink raw reply	[relevance 7%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 12:37  4% ` Olivier MATZ
@ 2016-03-10 13:15  4%   ` Bruce Richardson
  2016-03-10 13:56  4%     ` Wiles, Keith
  0 siblings, 1 reply; 200+ results
From: Bruce Richardson @ 2016-03-10 13:15 UTC (permalink / raw)
  To: Olivier MATZ; +Cc: dev

On Thu, Mar 10, 2016 at 01:37:27PM +0100, Olivier MATZ wrote:
> Hi David,
> 
> On 03/10/2016 12:55 PM, David Hunt wrote:
> > Announce the ABI breakage due to addition of external mempool
> > manager functionality which requires changes to rte_mempool
> > structure.
> > 
> > Signed-off-by: David Hunt <david.hunt@intel.com>
> 
> Acked-by: Olivier Matz <olivier.matz@6wind.com>
> 
Acked-by: Bruce Richardson <bruce.richardson@intel.com>

/Bruce

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH 0/3] ethdev: add helper functions to get eth_dev and dev private data
  @ 2016-03-10 13:37  0%   ` Ferruh Yigit
  0 siblings, 0 replies; 200+ results
From: Ferruh Yigit @ 2016-03-10 13:37 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev

On 3/10/2016 12:00 AM, Thomas Monjalon wrote:
> 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.
>>
>> There are a few more usage in drivers but they left as it is because they
>> are in fast path code.
> 
> What is the benefit of these functions if you do not plan to remove the
> global variables later?
> 
- Better discipline to abstract, function call can be costly than direct
access, but I believe this is no problem outside of the fast path.
- Helper functions for common tasks, like get private data by port.

> Anyway this discussion targets the release 16.07.
> 
There is no urgency for these.

Thanks,
ferruh

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 13:15  4%   ` Bruce Richardson
@ 2016-03-10 13:56  4%     ` Wiles, Keith
  2016-03-10 14:55  4%       ` Thomas Monjalon
  2016-04-04 14:49  4%       ` Thomas Monjalon
  0 siblings, 2 replies; 200+ results
From: Wiles, Keith @ 2016-03-10 13:56 UTC (permalink / raw)
  To: Richardson, Bruce, Olivier MATZ; +Cc: dev

>On Thu, Mar 10, 2016 at 01:37:27PM +0100, Olivier MATZ wrote:
>> Hi David,
>> 
>> On 03/10/2016 12:55 PM, David Hunt wrote:
>> > Announce the ABI breakage due to addition of external mempool
>> > manager functionality which requires changes to rte_mempool
>> > structure.
>> > 
>> > Signed-off-by: David Hunt <david.hunt@intel.com>
>> 
>> Acked-by: Olivier Matz <olivier.matz@6wind.com>
>> 
>Acked-by: Bruce Richardson <bruce.richardson@intel.com>

Asked-by: Keith Wiles <keith.wiles@intel.com>
>
>/Bruce
>


Regards,
Keith





^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 13:56  4%     ` Wiles, Keith
@ 2016-03-10 14:55  4%       ` Thomas Monjalon
  2016-04-05 14:06  4%         ` Wiles, Keith
  2016-04-04 14:49  4%       ` Thomas Monjalon
  1 sibling, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-03-10 14:55 UTC (permalink / raw)
  To: Wiles, Keith; +Cc: dev

2016-03-10 13:56, Wiles, Keith:
> >On Thu, Mar 10, 2016 at 01:37:27PM +0100, Olivier MATZ wrote:
> >> Hi David,
> >> 
> >> On 03/10/2016 12:55 PM, David Hunt wrote:
> >> > Announce the ABI breakage due to addition of external mempool
> >> > manager functionality which requires changes to rte_mempool
> >> > structure.
> >> > 
> >> > Signed-off-by: David Hunt <david.hunt@intel.com>
> >> 
> >> Acked-by: Olivier Matz <olivier.matz@6wind.com>
> >> 
> >Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> Asked-by: Keith Wiles <keith.wiles@intel.com>

Is it on purpose, Keith, or a typo? Do you have asked this notice?

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH v3 0/2] add support for buffered tx to ethdev
  2016-03-10 11:31  0%   ` Ananyev, Konstantin
@ 2016-03-10 16:01  0%     ` Jastrzebski, MichalX K
  0 siblings, 0 replies; 200+ results
From: Jastrzebski, MichalX K @ 2016-03-10 16:01 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ananyev,
> Konstantin
> Sent: Thursday, March 10, 2016 12:32 PM
> To: Kulasek, TomaszX <tomaszx.kulasek@intel.com>; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v3 0/2] add support for buffered tx to
> ethdev
> 
> > Many sample apps include internal buffering for single-packet-at-a-time
> > operation. Since this is such a common paradigm, this functionality is
> > better suited to being implemented in the ethdev API.
> >
> > The new APIs in the ethdev library are:
> > * rte_eth_tx_buffer_init - initialize buffer
> > * rte_eth_tx_buffer - buffer up a single packet for future transmission
> > * rte_eth_tx_buffer_flush - flush any unsent buffered packets
> > * rte_eth_tx_buffer_set_err_callback - set up a callback to be called in
> >   case transmitting a buffered burst fails. By default, we just free the
> >   unsent packets.
> >
> > As well as these, an additional reference callbacks are provided, which
> > frees the packets:
> >
> > * rte_eth_tx_buffer_drop_callback - silently drop packets (default
> >   behavior)
> > * rte_eth_tx_buffer_count_callback - drop and update user-provided
> counter
> >   to track the number of dropped packets
> >
> > Due to the feedback from mailing list, that buffer management facilities
> > in the user application are more preferable than API simplicity, we
> decided
> > to move internal buffer table, as well as callback functions and user data,
> > from rte_eth_dev/rte_eth_dev_data to the application space.
> > It prevents ABI breakage and gives some more flexibility in the buffer's
> > management such as allocation, dynamical size change, reuse buffers on
> many
> > ports or after fail, and so on.
> >
> >
> > The following steps illustrate how tx buffers can be used in application:
> >
> > 1) Initialization
> >
> > a) Allocate memory for a buffer
> >
> >    struct rte_eth_dev_tx_buffer *buffer = rte_zmalloc_socket("tx_buffer",
> >            RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0, socket_id);
> >
> >    RTE_ETH_TX_BUFFER_SIZE(size) macro computes memory required to
> store
> >    "size" packets in buffer.
> >
> > b) Initialize allocated memory and set up default values. Threshold level
> >    must be lower than or equal to the MAX_PKT_BURST from 1a)
> >
> >    rte_eth_tx_buffer_init(buffer, threshold);
> >
> >
> > c) Set error callback (optional)
> >
> >    rte_eth_tx_buffer_set_err_callback(buffer, callback_fn, userdata);
> >
> >
> > 2) Store packet "pkt" in buffer and send them all to the queue_id on
> >    port_id when number of packets reaches threshold level set up in 1b)
> >
> >    rte_eth_tx_buffer(port_id, queue_id, buffer, pkt);
> >
> >
> > 3) Send all stored packets to the queue_id on port_id
> >
> >    rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
> >
> >
> > 4) Flush buffer and free memory
> >
> >    rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
> >    ...
> >    rte_free(buffer);
> >
> > v3 changes:
> >  - error counter removed from tx buffer structure, now default behavior is
> >    silent drop of unsent packets
> >  - some names was changed in tx buffer structure to be more descriptive
> >  - two default calbacks are provided: rte_eth_tx_buffer_drop_callback and
> >    rte_eth_tx_buffer_count_callback
> >
> > v2 changes:
> >  - reworked to use new buffer model
> >  - buffer data and callbacks are removed from
> rte_eth_dev/rte_eth_dev_data,
> >    so this patch doesn't brake an ABI anymore
> >  - introduced RTE_ETH_TX_BUFFER macro and rte_eth_tx_buffer_init
> >  - buffers are not attached to the port-queue
> >  - buffers can be allocated dynamically during application work
> >  - size of buffer can be changed without port restart
> >
> > Tomasz Kulasek (2):
> >   ethdev: add buffered tx api
> >   examples: rework to use buffered tx
> >
> >  examples/l2fwd-jobstats/main.c                     |  104 ++++------
> >  examples/l2fwd-keepalive/main.c                    |  100 ++++------
> >  examples/l2fwd/main.c                              |  104 ++++------
> >  examples/l3fwd-acl/main.c                          |   92 ++++-----
> >  examples/l3fwd-power/main.c                        |   89 ++++-----
> >  examples/link_status_interrupt/main.c              |  107 ++++------
> >  .../client_server_mp/mp_client/client.c            |  101 ++++++----
> >  examples/multi_process/l2fwd_fork/main.c           |   97 ++++-----
> >  examples/packet_ordering/main.c                    |  122 ++++++++----
> >  examples/qos_meter/main.c                          |   61 ++----
> >  lib/librte_ether/rte_ethdev.c                      |   46 +++++
> >  lib/librte_ether/rte_ethdev.h                      |  205 +++++++++++++++++++-
> >  lib/librte_ether/rte_ether_version.map             |   10 +
> >  13 files changed, 696 insertions(+), 542 deletions(-)
> >
> > --
> 
> Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
> 
> > 1.7.9.5

Hi Thomas,
Could You write please does this patch meet Your requirements and 
does it have a green light to be applied?

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 11:55  7% [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice David Hunt
  2016-03-10 12:37  4% ` Olivier MATZ
@ 2016-03-10 16:23  4% ` Mcnamara, John
  1 sibling, 0 replies; 200+ results
From: Mcnamara, John @ 2016-03-10 16:23 UTC (permalink / raw)
  To: Hunt, David, dev

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of David Hunt
> Sent: Thursday, March 10, 2016 11:56 AM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
> 
> Announce the ABI breakage due to addition of external mempool manager
> functionality which requires changes to rte_mempool structure.
> 
> Signed-off-by: David Hunt <david.hunt@intel.com>

Acked-by: John McNamara <john.mcnamara@intel.com>

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v3 0/2] i40e setting ether type of VLANs
  @ 2016-03-10 16:36  4% ` Helin Zhang
  2016-03-10 16:36  7%   ` [dpdk-dev] [PATCH v3 1/2] ethdev: add vlan type for setting ether type Helin Zhang
                     ` (2 more replies)
  0 siblings, 3 replies; 200+ results
From: Helin Zhang @ 2016-03-10 16:36 UTC (permalink / raw)
  To: dev

It adds setting ether type of both single VLAN(inner VLAN) and
outer VLAN for i40e. For ixgbe and e1000/igb, it supports setting
single VLAN(inner VLAN) only, and can be extended in the future.

The patch set was branched off rel_16_04 of repo dpdk-next-net,
on below commit.
commit 5cfa5d194a8a45176e70af05719f7e3b136868be
Author: Zhe Tao <zhe.tao@intel.com>
Date:   Thu Mar 10 15:26:22 2016 +0000
    ixgbe: fix ixgbevf RX/TX function assignment

v3:
 - Used versioning mechanism to avoid ABI issue.
 - re-organized the patch set.

v2:
 - Used RTE_NEXT_ABI to avoid ABI change issue.
 - Reworked the announcement of ABI change for release 16.07.
 - Fixed a i40e overflow issue.

Helin Zhang (2):
  ethdev: add vlan type for setting ether type
  i40e: fix the overflow issue

 app/test-pmd/cmdline.c                 | 30 +++++++++++-----
 app/test-pmd/config.c                  |  9 ++---
 app/test-pmd/testpmd.h                 |  3 +-
 doc/guides/rel_notes/release_16_04.rst |  4 +++
 drivers/net/e1000/igb_ethdev.c         | 21 +++++++++---
 drivers/net/i40e/i40e_ethdev.c         | 63 +++++++++++++++++++++++++++++++---
 drivers/net/i40e/i40e_rxtx.c           |  4 +--
 drivers/net/ixgbe/ixgbe_ethdev.c       | 19 +++++++---
 lib/librte_ether/rte_ethdev.c          | 25 ++++++++++++--
 lib/librte_ether/rte_ethdev.h          | 23 +++++++++++--
 lib/librte_ether/rte_ether_version.map |  7 ++++
 11 files changed, 175 insertions(+), 33 deletions(-)

-- 
2.5.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v3 1/2] ethdev: add vlan type for setting ether type
  2016-03-10 16:36  4% ` [dpdk-dev] [PATCH v3 0/2] " Helin Zhang
@ 2016-03-10 16:36  7%   ` Helin Zhang
  2016-03-11  2:36  0%   ` [dpdk-dev] [PATCH v3 0/2] i40e setting ether type of VLANs Lu, Wenzhuo
  2016-03-11  8:49  4%   ` [dpdk-dev] [PATCH v4 " Helin Zhang
  2 siblings, 0 replies; 200+ results
From: Helin Zhang @ 2016-03-10 16:36 UTC (permalink / raw)
  To: dev

In order to set ether type of VLAN for single VLAN, inner
and outer VLAN, the VLAN type as an input parameter is added
to 'rte_eth_dev_set_vlan_ether_type()'.
In addition, corresponding changes in e1000, ixgbe and i40e
are also added.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
---
 app/test-pmd/cmdline.c                 | 30 +++++++++++-----
 app/test-pmd/config.c                  |  9 ++---
 app/test-pmd/testpmd.h                 |  3 +-
 doc/guides/rel_notes/release_16_04.rst |  4 +++
 drivers/net/e1000/igb_ethdev.c         | 21 +++++++++---
 drivers/net/i40e/i40e_ethdev.c         | 63 +++++++++++++++++++++++++++++++---
 drivers/net/ixgbe/ixgbe_ethdev.c       | 19 +++++++---
 lib/librte_ether/rte_ethdev.c          | 25 ++++++++++++--
 lib/librte_ether/rte_ethdev.h          | 23 +++++++++++--
 lib/librte_ether/rte_ether_version.map |  7 ++++
 10 files changed, 173 insertions(+), 31 deletions(-)

v3:
 - Used versioning mechanism to avoid ABI issue.
 - re-organized the patch set.

v2:
 - Used RTE_NEXT_ABI to avoid ABI change issue.
 - Reworked the announcement of ABI change for release 16.07.

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 52e9f5f..1eeb8a8 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -277,8 +277,8 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Set the VLAN QinQ (extended queue in queue)"
 			" on a port.\n\n"
 
-			"vlan set tpid (value) (port_id)\n"
-			"    Set the outer VLAN TPID for Packet Filtering on"
+			"vlan set (inner|outer) tpid (value) (port_id)\n"
+			"    Set the VLAN TPID for Packet Filtering on"
 			" a port\n\n"
 
 			"rx_vlan add (vlan_id|all) (port_id)\n"
@@ -297,10 +297,6 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Remove a vlan_id, to the set of VLAN identifiers"
 			"filtered for VF(s) from port_id.\n\n"
 
-			"rx_vlan set tpid (value) (port_id)\n"
-			"    Set the outer VLAN TPID for Packet Filtering on"
-			" a port\n\n"
-
 			"tunnel_filter add (port_id) (outer_mac) (inner_mac) (ip_addr) "
 			"(inner_vlan) (vxlan|nvgre) (filter_type) (tenant_id) (queue_id)\n"
 			"   add a tunnel filter of a port.\n\n"
@@ -2747,6 +2743,7 @@ cmdline_parse_inst_t cmd_rx_vlan_filter_all = {
 struct cmd_vlan_offload_result {
 	cmdline_fixed_string_t vlan;
 	cmdline_fixed_string_t set;
+	cmdline_fixed_string_t vlan_type;
 	cmdline_fixed_string_t what;
 	cmdline_fixed_string_t on;
 	cmdline_fixed_string_t port_id;
@@ -2847,6 +2844,7 @@ cmdline_parse_inst_t cmd_vlan_offload = {
 struct cmd_vlan_tpid_result {
 	cmdline_fixed_string_t vlan;
 	cmdline_fixed_string_t set;
+	cmdline_fixed_string_t vlan_type;
 	cmdline_fixed_string_t what;
 	uint16_t tp_id;
 	uint8_t port_id;
@@ -2858,8 +2856,17 @@ cmd_vlan_tpid_parsed(void *parsed_result,
 			  __attribute__((unused)) void *data)
 {
 	struct cmd_vlan_tpid_result *res = parsed_result;
-	vlan_tpid_set(res->port_id, res->tp_id);
-	return;
+	enum rte_vlan_type vlan_type;
+
+	if (!strcmp(res->vlan_type, "inner"))
+		vlan_type = ETH_VLAN_TYPE_INNER;
+	else if (!strcmp(res->vlan_type, "outer"))
+		vlan_type = ETH_VLAN_TYPE_OUTER;
+	else {
+		printf("Unknown vlan type\n");
+		return;
+	}
+	vlan_tpid_set(res->port_id, vlan_type, res->tp_id);
 }
 
 cmdline_parse_token_string_t cmd_vlan_tpid_vlan =
@@ -2868,6 +2875,9 @@ cmdline_parse_token_string_t cmd_vlan_tpid_vlan =
 cmdline_parse_token_string_t cmd_vlan_tpid_set =
 	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
 				 set, "set");
+cmdline_parse_token_string_t cmd_vlan_type =
+	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
+				 vlan_type, "inner#outer");
 cmdline_parse_token_string_t cmd_vlan_tpid_what =
 	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
 				 what, "tpid");
@@ -2881,10 +2891,12 @@ cmdline_parse_token_num_t cmd_vlan_tpid_portid =
 cmdline_parse_inst_t cmd_vlan_tpid = {
 	.f = cmd_vlan_tpid_parsed,
 	.data = NULL,
-	.help_str = "set tpid tp_id port_id, set the Outer VLAN Ether type",
+	.help_str = "set inner|outer tpid tp_id port_id, set the VLAN "
+		    "Ether type",
 	.tokens = {
 		(void *)&cmd_vlan_tpid_vlan,
 		(void *)&cmd_vlan_tpid_set,
+		(void *)&cmd_vlan_type,
 		(void *)&cmd_vlan_tpid_what,
 		(void *)&cmd_vlan_tpid_tpid,
 		(void *)&cmd_vlan_tpid_portid,
diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 0062484..5bb09a5 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -1821,19 +1821,20 @@ rx_vlan_all_filter_set(portid_t port_id, int on)
 }
 
 void
-vlan_tpid_set(portid_t port_id, uint16_t tp_id)
+vlan_tpid_set(portid_t port_id, enum rte_vlan_type vlan_type, uint16_t tp_id)
 {
 	int diag;
+
 	if (port_id_is_invalid(port_id, ENABLED_WARN))
 		return;
 
-	diag = rte_eth_dev_set_vlan_ether_type(port_id, tp_id);
+	diag = rte_eth_dev_set_vlan_ether_type(port_id, vlan_type, tp_id);
 	if (diag == 0)
 		return;
 
-	printf("tx_vlan_tpid_set(port_pi=%d, tpid=%d) failed "
+	printf("tx_vlan_tpid_set(port_pi=%d, vlan_type=%d, tpid=%d) failed "
 	       "diag=%d\n",
-	       port_id, tp_id, diag);
+	       port_id, vlan_type, tp_id, diag);
 }
 
 void
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index b618998..0f72ca1 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -508,7 +508,8 @@ void rx_vlan_filter_set(portid_t port_id, int on);
 void rx_vlan_all_filter_set(portid_t port_id, int on);
 int rx_vft_set(portid_t port_id, uint16_t vlan_id, int on);
 void vlan_extend_set(portid_t port_id, int on);
-void vlan_tpid_set(portid_t port_id, uint16_t tp_id);
+void vlan_tpid_set(portid_t port_id, enum rte_vlan_type vlan_type,
+		   uint16_t tp_id);
 void tx_vlan_set(portid_t port_id, uint16_t vlan_id);
 void tx_qinq_set(portid_t port_id, uint16_t vlan_id, uint16_t vlan_id_outer);
 void tx_vlan_reset(portid_t port_id);
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 5abf48a..4729019 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -119,6 +119,10 @@ This section should contain new features added in this release. Sample format:
 
   Only available with Mellanox OFED >= 3.2.
 
+* **Added modifying ether type of both single and double VLAN for i40e**
+
+  Macro of RTE_NEXT_ABI was introduced, as ABI change involved.
+
 
 Resolved Issues
 ---------------
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index 41c107d..36d50c9 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -126,7 +126,9 @@ static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
 		uint16_t vlan_id, int on);
-static void eth_igb_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
+static void eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
+				  enum rte_vlan_type vlan_type,
+				  uint16_t tpid_id);
 static void eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 
 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
@@ -2194,14 +2196,23 @@ eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 }
 
 static void
-eth_igb_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
+eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
+		      enum rte_vlan_type vlan_type,
+		      uint16_t tpid)
 {
 	struct e1000_hw *hw =
 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	uint32_t reg = ETHER_TYPE_VLAN ;
+	uint32_t reg = ETHER_TYPE_VLAN;
 
-	reg |= (tpid << 16);
-	E1000_WRITE_REG(hw, E1000_VET, reg);
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_INNER:
+		reg |= (tpid << 16);
+		E1000_WRITE_REG(hw, E1000_VET, reg);
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d\n", vlan_type);
+		break;
+	}
 }
 
 static void
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index 0c87ec1..e3fc0bc 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -266,6 +266,11 @@
 #define I40E_INSET_IPV6_TC_MASK       0x0009F00FUL
 #define I40E_INSET_IPV6_NEXT_HDR_MASK 0x000C00FFUL
 
+#define I40E_GL_SWT_L2TAGCTRL(_i)             (0x001C0A70 + ((_i) * 4))
+#define I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT 16
+#define I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK  \
+	I40E_MASK(0xFFFF, I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT)
+
 static int eth_i40e_dev_init(struct rte_eth_dev *eth_dev);
 static int eth_i40e_dev_uninit(struct rte_eth_dev *eth_dev);
 static int i40e_dev_configure(struct rte_eth_dev *dev);
@@ -292,7 +297,9 @@ static void i40e_dev_info_get(struct rte_eth_dev *dev,
 static int i40e_vlan_filter_set(struct rte_eth_dev *dev,
 				uint16_t vlan_id,
 				int on);
-static void i40e_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid);
+static void i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+			       enum rte_vlan_type vlan_type,
+			       uint16_t tpid);
 static void i40e_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 static void i40e_vlan_strip_queue_set(struct rte_eth_dev *dev,
 				      uint16_t queue,
@@ -2313,10 +2320,52 @@ i40e_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 }
 
 static void
-i40e_vlan_tpid_set(__rte_unused struct rte_eth_dev *dev,
-		   __rte_unused uint16_t tpid)
+i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+		   enum rte_vlan_type vlan_type,
+		   uint16_t tpid)
 {
-	PMD_INIT_FUNC_TRACE();
+	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	uint64_t reg_r = 0, reg_w = 0;
+	uint16_t reg_id = 0;
+	int ret;
+
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_OUTER:
+		reg_id = 2;
+		break;
+	case ETH_VLAN_TYPE_INNER:
+		reg_id = 3;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d", vlan_type);
+		return;
+	}
+	ret = i40e_aq_debug_read_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
+					  &reg_r, NULL);
+	if (ret != I40E_SUCCESS) {
+		PMD_DRV_LOG(ERR, "Fail to debug read from "
+			    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_id);
+		return;
+	}
+	PMD_DRV_LOG(DEBUG, "Debug read from I40E_GL_SWT_L2TAGCTRL[%d]: "
+		    "0x%08"PRIx64"", reg_id, reg_r);
+
+	reg_w = reg_r & (~(I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK));
+	reg_w |= ((uint64_t)tpid << I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT);
+	if (reg_r == reg_w) {
+		PMD_DRV_LOG(DEBUG, "No need to write");
+		return;
+	}
+
+	ret = i40e_aq_debug_write_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
+					   reg_w, NULL);
+	if (ret != I40E_SUCCESS) {
+		PMD_DRV_LOG(ERR, "Fail to debug write to "
+			    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_id);
+		return;
+	}
+	PMD_DRV_LOG(DEBUG, "Debug write 0x%08"PRIx64" to "
+		    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_w, reg_id);
 }
 
 static void
@@ -7332,11 +7381,17 @@ i40e_dev_filter_ctrl(struct rte_eth_dev *dev,
 static void
 i40e_hw_init(struct i40e_hw *hw)
 {
+	struct rte_eth_dev *dev = ((struct i40e_adapter *)(hw->back))->eth_dev;
+
 	/* clear the PF Queue Filter control register */
 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, 0);
 
 	/* Disable symmetric hash per port */
 	i40e_set_symmetric_hash_enable_per_port(hw, 0);
+
+	/* Set the 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);
 }
 
 enum i40e_filter_pctype
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a9a1583..ce3ce9f 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -178,7 +178,9 @@ static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
 		uint16_t vlan_id, int on);
-static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
+static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
+				enum rte_vlan_type vlan_type,
+				uint16_t tpid_id);
 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
 		uint16_t queue, bool on);
 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
@@ -1543,13 +1545,22 @@ ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
 }
 
 static void
-ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
+ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
+		    enum rte_vlan_type vlan_type,
+		    uint16_t tpid)
 {
 	struct ixgbe_hw *hw =
 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 
-	/* Only the high 16-bits is valid */
-	IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_INNER:
+		/* Only the high 16-bits is valid */
+		IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d\n", vlan_type);
+		break;
+	}
 }
 
 void
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index a6e83c1..bb0c3eb 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -65,6 +65,7 @@
 #include <rte_errno.h>
 #include <rte_spinlock.h>
 #include <rte_string_fns.h>
+#include <rte_compat.h>
 
 #include "rte_ether.h"
 #include "rte_ethdev.h"
@@ -1697,17 +1698,37 @@ rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id, int o
 }
 
 int
-rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tpid)
+rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t tpid)
 {
 	struct rte_eth_dev *dev;
 
 	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
 	dev = &rte_eth_devices[port_id];
 	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->vlan_tpid_set, -ENOTSUP);
-	(*dev->dev_ops->vlan_tpid_set)(dev, tpid);
+	(*dev->dev_ops->vlan_tpid_set)(dev, ETH_VLAN_TYPE_INNER, tpid);
 
 	return 0;
 }
+VERSION_SYMBOL(rte_eth_dev_set_vlan_ether_type, _v22, 2.2);
+
+int
+rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
+				      enum rte_vlan_type vlan_type,
+				      uint16_t tpid)
+{
+	struct rte_eth_dev *dev;
+
+	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
+	dev = &rte_eth_devices[port_id];
+	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->vlan_tpid_set, -ENOTSUP);
+	(*dev->dev_ops->vlan_tpid_set)(dev, vlan_type, tpid);
+
+	return 0;
+}
+BIND_DEFAULT_SYMBOL(rte_eth_dev_set_vlan_ether_type, _v1604, 16.04);
+MAP_STATIC_SYMBOL(int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
+		  enum rte_vlan_type vlan_type, uint16_t tpid),
+		  rte_eth_dev_set_vlan_ether_type_v1604);
 
 int
 rte_eth_dev_set_vlan_offload(uint8_t port_id, int offload_mask)
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index e2893ba..21db144 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -351,6 +351,17 @@ struct rte_eth_rxmode {
 };
 
 /**
+ * VLAN types to indicate if it is for single VLAN, inner VLAN or outer VLAN.
+ * Note that single VLAN is treated the same as inner VLAN.
+ */
+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_MAX,
+};
+
+/**
  * A structure used to configure the Receive Side Scaling (RSS) feature
  * of an Ethernet port.
  * If not NULL, the *rss_key* pointer of the *rss_conf* structure points
@@ -1077,7 +1088,7 @@ typedef int (*vlan_filter_set_t)(struct rte_eth_dev *dev,
 /**< @internal filtering of a VLAN Tag Identifier by an Ethernet device. */
 
 typedef void (*vlan_tpid_set_t)(struct rte_eth_dev *dev,
-				  uint16_t tpid);
+				enum rte_vlan_type type, uint16_t tpid);
 /**< @internal set the outer VLAN-TPID by an Ethernet device. */
 
 typedef void (*vlan_offload_set_t)(struct rte_eth_dev *dev, int mask);
@@ -2346,6 +2357,8 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
  *
  * @param port_id
  *   The port identifier of the Ethernet device.
+ * @vlan_type
+ *   The vlan type.
  * @param tag_type
  *   The Tag Protocol ID
  * @return
@@ -2353,7 +2366,13 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
  *   - (-ENOSUP) if hardware-assisted VLAN TPID setup is not supported.
  *   - (-ENODEV) if *port_id* invalid.
  */
-int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
+				    enum rte_vlan_type vlan_type,
+				    uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
+					  enum rte_vlan_type vlan_type,
+					  uint16_t tag_type);
 
 /**
  * Set VLAN offload configuration on an Ethernet device
diff --git a/lib/librte_ether/rte_ether_version.map b/lib/librte_ether/rte_ether_version.map
index d8db24d..6098de5 100644
--- a/lib/librte_ether/rte_ether_version.map
+++ b/lib/librte_ether/rte_ether_version.map
@@ -117,3 +117,10 @@ DPDK_2.2 {
 
 	local: *;
 };
+
+DPDK_16.04 {
+	global:
+
+	rte_eth_dev_set_vlan_ether_type;
+
+} DPDK_2.2;
-- 
2.5.0

^ permalink raw reply	[relevance 7%]

* [dpdk-dev] [PATCH v4 0/2] add support for buffered tx to ethdev
  2016-03-10 10:57  4% ` [dpdk-dev] [PATCH v3 " Tomasz Kulasek
  2016-03-10 11:31  0%   ` Ananyev, Konstantin
@ 2016-03-10 17:19  4%   ` Tomasz Kulasek
  1 sibling, 0 replies; 200+ results
From: Tomasz Kulasek @ 2016-03-10 17:19 UTC (permalink / raw)
  To: dev

Many sample apps include internal buffering for single-packet-at-a-time
operation. Since this is such a common paradigm, this functionality is
better suited to being implemented in the ethdev API.

The new APIs in the ethdev library are:
* rte_eth_tx_buffer_init - initialize buffer
* rte_eth_tx_buffer - buffer up a single packet for future transmission
* rte_eth_tx_buffer_flush - flush any unsent buffered packets
* rte_eth_tx_buffer_set_err_callback - set up a callback to be called in
  case transmitting a buffered burst fails. By default, we just free the
  unsent packets.

As well as these, an additional reference callbacks are provided, which
frees the packets:

* rte_eth_tx_buffer_drop_callback - silently drop packets (default
  behavior)
* rte_eth_tx_buffer_count_callback - drop and update user-provided counter
  to track the number of dropped packets

Due to the feedback from mailing list, that buffer management facilities
in the user application are more preferable than API simplicity, we decided
to move internal buffer table, as well as callback functions and user data,
from rte_eth_dev/rte_eth_dev_data to the application space.
It prevents ABI breakage and gives some more flexibility in the buffer's
management such as allocation, dynamical size change, reuse buffers on many
ports or after fail, and so on.


The following steps illustrate how tx buffers can be used in application:

1) Initialization

a) Allocate memory for a buffer

   struct rte_eth_dev_tx_buffer *buffer = rte_zmalloc_socket("tx_buffer",
           RTE_ETH_TX_BUFFER_SIZE(MAX_PKT_BURST), 0, socket_id);

   RTE_ETH_TX_BUFFER_SIZE(size) macro computes memory required to store
   "size" packets in buffer.

b) Initialize allocated memory and set up default values. Threshold level
   must be lower than or equal to the MAX_PKT_BURST from 1a)

   rte_eth_tx_buffer_init(buffer, threshold);


c) Set error callback (optional)

   rte_eth_tx_buffer_set_err_callback(buffer, callback_fn, userdata);


2) Store packet "pkt" in buffer and send them all to the queue_id on
   port_id when number of packets reaches threshold level set up in 1b)

   rte_eth_tx_buffer(port_id, queue_id, buffer, pkt);


3) Send all stored packets to the queue_id on port_id

   rte_eth_tx_buffer_flush(port_id, queue_id, buffer);


4) Flush buffer and free memory

   rte_eth_tx_buffer_flush(port_id, queue_id, buffer);
   ...
   rte_free(buffer);

v4 changes:
 - added comments
 - chaged names of error callback and user data
 - changed order of function names in map file

v3 changes:
 - error counter removed from tx buffer structure, now default behavior is
   silent drop of unsent packets
 - some names was changed in tx buffer structure to be more descriptive
 - two default calbacks are provided: rte_eth_tx_buffer_drop_callback and
   rte_eth_tx_buffer_count_callback

v2 changes:
 - reworked to use new buffer model
 - buffer data and callbacks are removed from rte_eth_dev/rte_eth_dev_data,
   so this patch doesn't brake an ABI anymore
 - introduced RTE_ETH_TX_BUFFER macro and rte_eth_tx_buffer_init
 - buffers are not attached to the port-queue
 - buffers can be allocated dynamically during application work
 - size of buffer can be changed without port restart

Tomasz Kulasek (2):
  ethdev: add buffered tx api
  examples: rework to use buffered tx

 examples/l2fwd-jobstats/main.c                     |  104 ++++------
 examples/l2fwd-keepalive/main.c                    |  100 ++++------
 examples/l2fwd/main.c                              |  104 ++++------
 examples/l3fwd-acl/main.c                          |   92 ++++-----
 examples/l3fwd-power/main.c                        |   89 ++++-----
 examples/link_status_interrupt/main.c              |  107 ++++------
 .../client_server_mp/mp_client/client.c            |  101 ++++++----
 examples/multi_process/l2fwd_fork/main.c           |   97 ++++-----
 examples/packet_ordering/main.c                    |  122 ++++++++----
 examples/qos_meter/main.c                          |   61 ++----
 lib/librte_ether/rte_ethdev.c                      |   46 +++++
 lib/librte_ether/rte_ethdev.h                      |  206 +++++++++++++++++++-
 lib/librte_ether/rte_ether_version.map             |   10 +
 13 files changed, 697 insertions(+), 542 deletions(-)

-- 
1.7.9.5

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH v3 0/2] i40e setting ether type of VLANs
  2016-03-10 16:36  4% ` [dpdk-dev] [PATCH v3 0/2] " Helin Zhang
  2016-03-10 16:36  7%   ` [dpdk-dev] [PATCH v3 1/2] ethdev: add vlan type for setting ether type Helin Zhang
@ 2016-03-11  2:36  0%   ` Lu, Wenzhuo
  2016-03-11  8:49  4%   ` [dpdk-dev] [PATCH v4 " Helin Zhang
  2 siblings, 0 replies; 200+ results
From: Lu, Wenzhuo @ 2016-03-11  2:36 UTC (permalink / raw)
  To: Zhang, Helin, dev

Hi,

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Helin Zhang
> Sent: Friday, March 11, 2016 12:37 AM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH v3 0/2] i40e setting ether type of VLANs
> 
> It adds setting ether type of both single VLAN(inner VLAN) and outer VLAN
> for i40e. For ixgbe and e1000/igb, it supports setting single VLAN(inner VLAN)
> only, and can be extended in the future.
> 
> The patch set was branched off rel_16_04 of repo dpdk-next-net, on below
> commit.
> commit 5cfa5d194a8a45176e70af05719f7e3b136868be
> Author: Zhe Tao <zhe.tao@intel.com>
> Date:   Thu Mar 10 15:26:22 2016 +0000
>     ixgbe: fix ixgbevf RX/TX function assignment
> 
> v3:
>  - Used versioning mechanism to avoid ABI issue.
>  - re-organized the patch set.
> 
> v2:
>  - Used RTE_NEXT_ABI to avoid ABI change issue.
>  - Reworked the announcement of ABI change for release 16.07.
>  - Fixed a i40e overflow issue.
> 
> Helin Zhang (2):
>   ethdev: add vlan type for setting ether type
>   i40e: fix the overflow issue
> 
>  app/test-pmd/cmdline.c                 | 30 +++++++++++-----
>  app/test-pmd/config.c                  |  9 ++---
>  app/test-pmd/testpmd.h                 |  3 +-
>  doc/guides/rel_notes/release_16_04.rst |  4 +++
>  drivers/net/e1000/igb_ethdev.c         | 21 +++++++++---
>  drivers/net/i40e/i40e_ethdev.c         | 63
> +++++++++++++++++++++++++++++++---
>  drivers/net/i40e/i40e_rxtx.c           |  4 +--
>  drivers/net/ixgbe/ixgbe_ethdev.c       | 19 +++++++---
>  lib/librte_ether/rte_ethdev.c          | 25 ++++++++++++--
>  lib/librte_ether/rte_ethdev.h          | 23 +++++++++++--
>  lib/librte_ether/rte_ether_version.map |  7 ++++
>  11 files changed, 175 insertions(+), 33 deletions(-)
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
If you can update the doc/guides/testpmd_app_ug/testpmd_funcs.rst for the CLI change.

> 
> --
> 2.5.0

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v4 0/2] i40e setting ether type of VLANs
  2016-03-10 16:36  4% ` [dpdk-dev] [PATCH v3 0/2] " Helin Zhang
  2016-03-10 16:36  7%   ` [dpdk-dev] [PATCH v3 1/2] ethdev: add vlan type for setting ether type Helin Zhang
  2016-03-11  2:36  0%   ` [dpdk-dev] [PATCH v3 0/2] i40e setting ether type of VLANs Lu, Wenzhuo
@ 2016-03-11  8:49  4%   ` Helin Zhang
  2016-03-11  8:49  7%     ` [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type Helin Zhang
  2016-03-11 16:50  5%     ` [dpdk-dev] [PATCH v5 0/2] i40e setting ether type of VLANs Helin Zhang
  2 siblings, 2 replies; 200+ results
From: Helin Zhang @ 2016-03-11  8:49 UTC (permalink / raw)
  To: dev

It adds setting ether type of both single VLAN(inner VLAN)
and outer VLAN for i40e. For ixgbe and e1000/igb, it supports
setting single VLAN(inner VLAN) only, and can be extended in
the future.

The patch set was branched off rel_16_04 of repo dpdk-next-net,
on below commit.
commit 5721e6447b5c20208a32c919a509e89d78c3e68d
Author: Robin Jarry <robin.jarry@6wind.com>
Date:   Thu Mar 3 15:27:40 2016 +0100
    mlx4: ensure number of RX queues is a power of 2

v4:
 - Updated the doc of testpmd guide.

v3:
 - Used versioning mechanism to avoid ABI issue.
 - Re-organized the patch set.

v2:
 - Used RTE_NEXT_ABI to avoid ABI change issue.
 - Reworked the announcement of ABI change for release 16.07.
 - Fixed a i40e overflow issue.

Helin Zhang (2):
  ethdev: add vlan type for setting ether type
  i40e: fix the overflow issue

 app/test-pmd/cmdline.c                      | 30 +++++++++-----
 app/test-pmd/config.c                       |  9 +++--
 app/test-pmd/testpmd.h                      |  3 +-
 doc/guides/rel_notes/release_16_04.rst      |  4 ++
 doc/guides/testpmd_app_ug/testpmd_funcs.rst | 11 +----
 drivers/net/e1000/igb_ethdev.c              | 21 +++++++---
 drivers/net/i40e/i40e_ethdev.c              | 63 +++++++++++++++++++++++++++--
 drivers/net/i40e/i40e_rxtx.c                |  4 +-
 drivers/net/ixgbe/ixgbe_ethdev.c            | 19 +++++++--
 lib/librte_ether/rte_ethdev.c               | 25 +++++++++++-
 lib/librte_ether/rte_ethdev.h               | 23 ++++++++++-
 lib/librte_ether/rte_ether_version.map      |  7 ++++
 12 files changed, 177 insertions(+), 42 deletions(-)

-- 
2.5.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type
  2016-03-11  8:49  4%   ` [dpdk-dev] [PATCH v4 " Helin Zhang
@ 2016-03-11  8:49  7%     ` Helin Zhang
  2016-03-11 11:19  3%       ` Panu Matilainen
  2016-03-11 16:50  5%     ` [dpdk-dev] [PATCH v5 0/2] i40e setting ether type of VLANs Helin Zhang
  1 sibling, 1 reply; 200+ results
From: Helin Zhang @ 2016-03-11  8:49 UTC (permalink / raw)
  To: dev

In order to set ether type of VLAN for single VLAN, inner
and outer VLAN, the VLAN type as an input parameter is added
to 'rte_eth_dev_set_vlan_ether_type()'.
In addition, corresponding changes in e1000, ixgbe and i40e
are also added.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
---
 app/test-pmd/cmdline.c                      | 30 +++++++++-----
 app/test-pmd/config.c                       |  9 +++--
 app/test-pmd/testpmd.h                      |  3 +-
 doc/guides/rel_notes/release_16_04.rst      |  4 ++
 doc/guides/testpmd_app_ug/testpmd_funcs.rst | 11 +----
 drivers/net/e1000/igb_ethdev.c              | 21 +++++++---
 drivers/net/i40e/i40e_ethdev.c              | 63 +++++++++++++++++++++++++++--
 drivers/net/ixgbe/ixgbe_ethdev.c            | 19 +++++++--
 lib/librte_ether/rte_ethdev.c               | 25 +++++++++++-
 lib/librte_ether/rte_ethdev.h               | 23 ++++++++++-
 lib/librte_ether/rte_ether_version.map      |  7 ++++
 11 files changed, 175 insertions(+), 40 deletions(-)

v4:
 - Updated the doc of testpmd guide.

v3:
 - Used versioning mechanism to avoid ABI issue.
 - Re-organized the patch set.

v2:
 - Used RTE_NEXT_ABI to avoid ABI change issue.
 - Reworked the announcement of ABI change for release 16.07.

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 52e9f5f..1eeb8a8 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -277,8 +277,8 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Set the VLAN QinQ (extended queue in queue)"
 			" on a port.\n\n"
 
-			"vlan set tpid (value) (port_id)\n"
-			"    Set the outer VLAN TPID for Packet Filtering on"
+			"vlan set (inner|outer) tpid (value) (port_id)\n"
+			"    Set the VLAN TPID for Packet Filtering on"
 			" a port\n\n"
 
 			"rx_vlan add (vlan_id|all) (port_id)\n"
@@ -297,10 +297,6 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Remove a vlan_id, to the set of VLAN identifiers"
 			"filtered for VF(s) from port_id.\n\n"
 
-			"rx_vlan set tpid (value) (port_id)\n"
-			"    Set the outer VLAN TPID for Packet Filtering on"
-			" a port\n\n"
-
 			"tunnel_filter add (port_id) (outer_mac) (inner_mac) (ip_addr) "
 			"(inner_vlan) (vxlan|nvgre) (filter_type) (tenant_id) (queue_id)\n"
 			"   add a tunnel filter of a port.\n\n"
@@ -2747,6 +2743,7 @@ cmdline_parse_inst_t cmd_rx_vlan_filter_all = {
 struct cmd_vlan_offload_result {
 	cmdline_fixed_string_t vlan;
 	cmdline_fixed_string_t set;
+	cmdline_fixed_string_t vlan_type;
 	cmdline_fixed_string_t what;
 	cmdline_fixed_string_t on;
 	cmdline_fixed_string_t port_id;
@@ -2847,6 +2844,7 @@ cmdline_parse_inst_t cmd_vlan_offload = {
 struct cmd_vlan_tpid_result {
 	cmdline_fixed_string_t vlan;
 	cmdline_fixed_string_t set;
+	cmdline_fixed_string_t vlan_type;
 	cmdline_fixed_string_t what;
 	uint16_t tp_id;
 	uint8_t port_id;
@@ -2858,8 +2856,17 @@ cmd_vlan_tpid_parsed(void *parsed_result,
 			  __attribute__((unused)) void *data)
 {
 	struct cmd_vlan_tpid_result *res = parsed_result;
-	vlan_tpid_set(res->port_id, res->tp_id);
-	return;
+	enum rte_vlan_type vlan_type;
+
+	if (!strcmp(res->vlan_type, "inner"))
+		vlan_type = ETH_VLAN_TYPE_INNER;
+	else if (!strcmp(res->vlan_type, "outer"))
+		vlan_type = ETH_VLAN_TYPE_OUTER;
+	else {
+		printf("Unknown vlan type\n");
+		return;
+	}
+	vlan_tpid_set(res->port_id, vlan_type, res->tp_id);
 }
 
 cmdline_parse_token_string_t cmd_vlan_tpid_vlan =
@@ -2868,6 +2875,9 @@ cmdline_parse_token_string_t cmd_vlan_tpid_vlan =
 cmdline_parse_token_string_t cmd_vlan_tpid_set =
 	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
 				 set, "set");
+cmdline_parse_token_string_t cmd_vlan_type =
+	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
+				 vlan_type, "inner#outer");
 cmdline_parse_token_string_t cmd_vlan_tpid_what =
 	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
 				 what, "tpid");
@@ -2881,10 +2891,12 @@ cmdline_parse_token_num_t cmd_vlan_tpid_portid =
 cmdline_parse_inst_t cmd_vlan_tpid = {
 	.f = cmd_vlan_tpid_parsed,
 	.data = NULL,
-	.help_str = "set tpid tp_id port_id, set the Outer VLAN Ether type",
+	.help_str = "set inner|outer tpid tp_id port_id, set the VLAN "
+		    "Ether type",
 	.tokens = {
 		(void *)&cmd_vlan_tpid_vlan,
 		(void *)&cmd_vlan_tpid_set,
+		(void *)&cmd_vlan_type,
 		(void *)&cmd_vlan_tpid_what,
 		(void *)&cmd_vlan_tpid_tpid,
 		(void *)&cmd_vlan_tpid_portid,
diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 0062484..5bb09a5 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -1821,19 +1821,20 @@ rx_vlan_all_filter_set(portid_t port_id, int on)
 }
 
 void
-vlan_tpid_set(portid_t port_id, uint16_t tp_id)
+vlan_tpid_set(portid_t port_id, enum rte_vlan_type vlan_type, uint16_t tp_id)
 {
 	int diag;
+
 	if (port_id_is_invalid(port_id, ENABLED_WARN))
 		return;
 
-	diag = rte_eth_dev_set_vlan_ether_type(port_id, tp_id);
+	diag = rte_eth_dev_set_vlan_ether_type(port_id, vlan_type, tp_id);
 	if (diag == 0)
 		return;
 
-	printf("tx_vlan_tpid_set(port_pi=%d, tpid=%d) failed "
+	printf("tx_vlan_tpid_set(port_pi=%d, vlan_type=%d, tpid=%d) failed "
 	       "diag=%d\n",
-	       port_id, tp_id, diag);
+	       port_id, vlan_type, tp_id, diag);
 }
 
 void
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index b618998..0f72ca1 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -508,7 +508,8 @@ void rx_vlan_filter_set(portid_t port_id, int on);
 void rx_vlan_all_filter_set(portid_t port_id, int on);
 int rx_vft_set(portid_t port_id, uint16_t vlan_id, int on);
 void vlan_extend_set(portid_t port_id, int on);
-void vlan_tpid_set(portid_t port_id, uint16_t tp_id);
+void vlan_tpid_set(portid_t port_id, enum rte_vlan_type vlan_type,
+		   uint16_t tp_id);
 void tx_vlan_set(portid_t port_id, uint16_t vlan_id);
 void tx_qinq_set(portid_t port_id, uint16_t vlan_id, uint16_t vlan_id_outer);
 void tx_vlan_reset(portid_t port_id);
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 7c09fef..c6e5ef0 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -125,6 +125,10 @@ This section should contain new features added in this release. Sample format:
 
   Af_packet device can now be detached using API, like other PMD devices.
 
+* **Added modifying ether type of both single and double VLAN for i40e**
+
+  Versioning mechanism was introduced, to avoid any ABI issue.
+
 
 Resolved Issues
 ---------------
diff --git a/doc/guides/testpmd_app_ug/testpmd_funcs.rst b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
index a520cc5..e2b2224 100644
--- a/doc/guides/testpmd_app_ug/testpmd_funcs.rst
+++ b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
@@ -498,9 +498,9 @@ Set the VLAN QinQ (extended queue in queue) on for a port::
 vlan set tpid
 ~~~~~~~~~~~~~
 
-Set the outer VLAN TPID for packet filtering on a port::
+Set the inner or outer VLAN TPID for packet filtering on a port::
 
-   testpmd> vlan set tpid (value) (port_id)
+   testpmd> vlan set (inner|outer) tpid (value) (port_id)
 
 .. note::
 
@@ -540,13 +540,6 @@ Remove a VLAN ID, from the set of VLAN identifiers filtered for VF(s) for port I
 
    testpmd> rx_vlan rm (vlan_id) port (port_id) vf (vf_mask)
 
-rx_vlan set tpid
-~~~~~~~~~~~~~~~~
-
-Set the outer VLAN TPID for packet filtering on a port::
-
-   testpmd> rx_vlan set tpid (value) (port_id)
-
 tunnel_filter add
 ~~~~~~~~~~~~~~~~~
 
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index f889876..4549e0a 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -126,7 +126,9 @@ static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
 		uint16_t vlan_id, int on);
-static void eth_igb_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
+static void eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
+				  enum rte_vlan_type vlan_type,
+				  uint16_t tpid_id);
 static void eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 
 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
@@ -2194,14 +2196,23 @@ eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 }
 
 static void
-eth_igb_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
+eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
+		      enum rte_vlan_type vlan_type,
+		      uint16_t tpid)
 {
 	struct e1000_hw *hw =
 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	uint32_t reg = ETHER_TYPE_VLAN ;
+	uint32_t reg = ETHER_TYPE_VLAN;
 
-	reg |= (tpid << 16);
-	E1000_WRITE_REG(hw, E1000_VET, reg);
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_INNER:
+		reg |= (tpid << 16);
+		E1000_WRITE_REG(hw, E1000_VET, reg);
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d\n", vlan_type);
+		break;
+	}
 }
 
 static void
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index 0c87ec1..e3fc0bc 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -266,6 +266,11 @@
 #define I40E_INSET_IPV6_TC_MASK       0x0009F00FUL
 #define I40E_INSET_IPV6_NEXT_HDR_MASK 0x000C00FFUL
 
+#define I40E_GL_SWT_L2TAGCTRL(_i)             (0x001C0A70 + ((_i) * 4))
+#define I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT 16
+#define I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK  \
+	I40E_MASK(0xFFFF, I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT)
+
 static int eth_i40e_dev_init(struct rte_eth_dev *eth_dev);
 static int eth_i40e_dev_uninit(struct rte_eth_dev *eth_dev);
 static int i40e_dev_configure(struct rte_eth_dev *dev);
@@ -292,7 +297,9 @@ static void i40e_dev_info_get(struct rte_eth_dev *dev,
 static int i40e_vlan_filter_set(struct rte_eth_dev *dev,
 				uint16_t vlan_id,
 				int on);
-static void i40e_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid);
+static void i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+			       enum rte_vlan_type vlan_type,
+			       uint16_t tpid);
 static void i40e_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 static void i40e_vlan_strip_queue_set(struct rte_eth_dev *dev,
 				      uint16_t queue,
@@ -2313,10 +2320,52 @@ i40e_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 }
 
 static void
-i40e_vlan_tpid_set(__rte_unused struct rte_eth_dev *dev,
-		   __rte_unused uint16_t tpid)
+i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+		   enum rte_vlan_type vlan_type,
+		   uint16_t tpid)
 {
-	PMD_INIT_FUNC_TRACE();
+	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	uint64_t reg_r = 0, reg_w = 0;
+	uint16_t reg_id = 0;
+	int ret;
+
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_OUTER:
+		reg_id = 2;
+		break;
+	case ETH_VLAN_TYPE_INNER:
+		reg_id = 3;
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d", vlan_type);
+		return;
+	}
+	ret = i40e_aq_debug_read_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
+					  &reg_r, NULL);
+	if (ret != I40E_SUCCESS) {
+		PMD_DRV_LOG(ERR, "Fail to debug read from "
+			    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_id);
+		return;
+	}
+	PMD_DRV_LOG(DEBUG, "Debug read from I40E_GL_SWT_L2TAGCTRL[%d]: "
+		    "0x%08"PRIx64"", reg_id, reg_r);
+
+	reg_w = reg_r & (~(I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK));
+	reg_w |= ((uint64_t)tpid << I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT);
+	if (reg_r == reg_w) {
+		PMD_DRV_LOG(DEBUG, "No need to write");
+		return;
+	}
+
+	ret = i40e_aq_debug_write_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
+					   reg_w, NULL);
+	if (ret != I40E_SUCCESS) {
+		PMD_DRV_LOG(ERR, "Fail to debug write to "
+			    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_id);
+		return;
+	}
+	PMD_DRV_LOG(DEBUG, "Debug write 0x%08"PRIx64" to "
+		    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_w, reg_id);
 }
 
 static void
@@ -7332,11 +7381,17 @@ i40e_dev_filter_ctrl(struct rte_eth_dev *dev,
 static void
 i40e_hw_init(struct i40e_hw *hw)
 {
+	struct rte_eth_dev *dev = ((struct i40e_adapter *)(hw->back))->eth_dev;
+
 	/* clear the PF Queue Filter control register */
 	i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, 0);
 
 	/* Disable symmetric hash per port */
 	i40e_set_symmetric_hash_enable_per_port(hw, 0);
+
+	/* Set the 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);
 }
 
 enum i40e_filter_pctype
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a9a1583..ce3ce9f 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -178,7 +178,9 @@ static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
 		uint16_t vlan_id, int on);
-static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
+static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
+				enum rte_vlan_type vlan_type,
+				uint16_t tpid_id);
 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
 		uint16_t queue, bool on);
 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
@@ -1543,13 +1545,22 @@ ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
 }
 
 static void
-ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
+ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
+		    enum rte_vlan_type vlan_type,
+		    uint16_t tpid)
 {
 	struct ixgbe_hw *hw =
 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 
-	/* Only the high 16-bits is valid */
-	IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_INNER:
+		/* Only the high 16-bits is valid */
+		IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
+		break;
+	default:
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d\n", vlan_type);
+		break;
+	}
 }
 
 void
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index a6e83c1..bb0c3eb 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -65,6 +65,7 @@
 #include <rte_errno.h>
 #include <rte_spinlock.h>
 #include <rte_string_fns.h>
+#include <rte_compat.h>
 
 #include "rte_ether.h"
 #include "rte_ethdev.h"
@@ -1697,17 +1698,37 @@ rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id, int o
 }
 
 int
-rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tpid)
+rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t tpid)
 {
 	struct rte_eth_dev *dev;
 
 	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
 	dev = &rte_eth_devices[port_id];
 	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->vlan_tpid_set, -ENOTSUP);
-	(*dev->dev_ops->vlan_tpid_set)(dev, tpid);
+	(*dev->dev_ops->vlan_tpid_set)(dev, ETH_VLAN_TYPE_INNER, tpid);
 
 	return 0;
 }
+VERSION_SYMBOL(rte_eth_dev_set_vlan_ether_type, _v22, 2.2);
+
+int
+rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
+				      enum rte_vlan_type vlan_type,
+				      uint16_t tpid)
+{
+	struct rte_eth_dev *dev;
+
+	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
+	dev = &rte_eth_devices[port_id];
+	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->vlan_tpid_set, -ENOTSUP);
+	(*dev->dev_ops->vlan_tpid_set)(dev, vlan_type, tpid);
+
+	return 0;
+}
+BIND_DEFAULT_SYMBOL(rte_eth_dev_set_vlan_ether_type, _v1604, 16.04);
+MAP_STATIC_SYMBOL(int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
+		  enum rte_vlan_type vlan_type, uint16_t tpid),
+		  rte_eth_dev_set_vlan_ether_type_v1604);
 
 int
 rte_eth_dev_set_vlan_offload(uint8_t port_id, int offload_mask)
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index e2893ba..21db144 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -351,6 +351,17 @@ struct rte_eth_rxmode {
 };
 
 /**
+ * VLAN types to indicate if it is for single VLAN, inner VLAN or outer VLAN.
+ * Note that single VLAN is treated the same as inner VLAN.
+ */
+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_MAX,
+};
+
+/**
  * A structure used to configure the Receive Side Scaling (RSS) feature
  * of an Ethernet port.
  * If not NULL, the *rss_key* pointer of the *rss_conf* structure points
@@ -1077,7 +1088,7 @@ typedef int (*vlan_filter_set_t)(struct rte_eth_dev *dev,
 /**< @internal filtering of a VLAN Tag Identifier by an Ethernet device. */
 
 typedef void (*vlan_tpid_set_t)(struct rte_eth_dev *dev,
-				  uint16_t tpid);
+				enum rte_vlan_type type, uint16_t tpid);
 /**< @internal set the outer VLAN-TPID by an Ethernet device. */
 
 typedef void (*vlan_offload_set_t)(struct rte_eth_dev *dev, int mask);
@@ -2346,6 +2357,8 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
  *
  * @param port_id
  *   The port identifier of the Ethernet device.
+ * @vlan_type
+ *   The vlan type.
  * @param tag_type
  *   The Tag Protocol ID
  * @return
@@ -2353,7 +2366,13 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
  *   - (-ENOSUP) if hardware-assisted VLAN TPID setup is not supported.
  *   - (-ENODEV) if *port_id* invalid.
  */
-int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
+				    enum rte_vlan_type vlan_type,
+				    uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
+					  enum rte_vlan_type vlan_type,
+					  uint16_t tag_type);
 
 /**
  * Set VLAN offload configuration on an Ethernet device
diff --git a/lib/librte_ether/rte_ether_version.map b/lib/librte_ether/rte_ether_version.map
index d8db24d..6098de5 100644
--- a/lib/librte_ether/rte_ether_version.map
+++ b/lib/librte_ether/rte_ether_version.map
@@ -117,3 +117,10 @@ DPDK_2.2 {
 
 	local: *;
 };
+
+DPDK_16.04 {
+	global:
+
+	rte_eth_dev_set_vlan_ether_type;
+
+} DPDK_2.2;
-- 
2.5.0

^ permalink raw reply	[relevance 7%]

* Re: [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type
  2016-03-11  8:49  7%     ` [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type Helin Zhang
@ 2016-03-11 11:19  3%       ` Panu Matilainen
  2016-03-11 11:20  0%         ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-03-11 11:19 UTC (permalink / raw)
  To: Helin Zhang, dev

On 03/11/2016 10:49 AM, Helin Zhang wrote:
> In order to set ether type of VLAN for single VLAN, inner
> and outer VLAN, the VLAN type as an input parameter is added
> to 'rte_eth_dev_set_vlan_ether_type()'.
> In addition, corresponding changes in e1000, ixgbe and i40e
> are also added.
>
> Signed-off-by: Helin Zhang <helin.zhang@intel.com>
> Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
> ---
>   app/test-pmd/cmdline.c                      | 30 +++++++++-----
>   app/test-pmd/config.c                       |  9 +++--
>   app/test-pmd/testpmd.h                      |  3 +-
>   doc/guides/rel_notes/release_16_04.rst      |  4 ++
>   doc/guides/testpmd_app_ug/testpmd_funcs.rst | 11 +----
>   drivers/net/e1000/igb_ethdev.c              | 21 +++++++---
>   drivers/net/i40e/i40e_ethdev.c              | 63 +++++++++++++++++++++++++++--
>   drivers/net/ixgbe/ixgbe_ethdev.c            | 19 +++++++--
>   lib/librte_ether/rte_ethdev.c               | 25 +++++++++++-
>   lib/librte_ether/rte_ethdev.h               | 23 ++++++++++-
>   lib/librte_ether/rte_ether_version.map      |  7 ++++
>   11 files changed, 175 insertions(+), 40 deletions(-)
>
> v4:
>   - Updated the doc of testpmd guide.
>
> v3:
>   - Used versioning mechanism to avoid ABI issue.
>   - Re-organized the patch set.
>
> v2:
>   - Used RTE_NEXT_ABI to avoid ABI change issue.
>   - Reworked the announcement of ABI change for release 16.07.
>
[...]
> @@ -1077,7 +1088,7 @@ typedef int (*vlan_filter_set_t)(struct rte_eth_dev *dev,
>   /**< @internal filtering of a VLAN Tag Identifier by an Ethernet device. */
>
>   typedef void (*vlan_tpid_set_t)(struct rte_eth_dev *dev,
> -				  uint16_t tpid);
> +				enum rte_vlan_type type, uint16_t tpid);
>   /**< @internal set the outer VLAN-TPID by an Ethernet device. */
>
>   typedef void (*vlan_offload_set_t)(struct rte_eth_dev *dev, int mask);
> @@ -2346,6 +2357,8 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
>    *
>    * @param port_id
>    *   The port identifier of the Ethernet device.
> + * @vlan_type
> + *   The vlan type.
>    * @param tag_type
>    *   The Tag Protocol ID
>    * @return
> @@ -2353,7 +2366,13 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
>    *   - (-ENOSUP) if hardware-assisted VLAN TPID setup is not supported.
>    *   - (-ENODEV) if *port_id* invalid.
>    */
> -int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tag_type);
> +int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
> +				    enum rte_vlan_type vlan_type,
> +				    uint16_t tag_type);
> +int rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t tag_type);
> +int rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
> +					  enum rte_vlan_type vlan_type,
> +					  uint16_t tag_type);
>
>   /**
>    * Set VLAN offload configuration on an Ethernet device

Its nice to see people actually trying to be compatible on occasion :)

However in this case there's not much point in doing so, because 
libethdev ABI has already been broken in this cycle:
http://dpdk.org/browse/dpdk/commit/?id=cfd2279ea6299826fe992028f1dffaf9fa7e7d0a

In other words, the compatibility versions can never get invoked because 
all software built against libethdev needs to be rebuilt anyway because 
of the soname bump. Just drop the compat versions, no point carrying 
around something that cannot possibly get used.

	- Panu -

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type
  2016-03-11 11:19  3%       ` Panu Matilainen
@ 2016-03-11 11:20  0%         ` Thomas Monjalon
  2016-03-11 14:17  4%           ` Zhang, Helin
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-03-11 11:20 UTC (permalink / raw)
  To: Panu Matilainen, Helin Zhang; +Cc: dev

2016-03-11 13:19, Panu Matilainen:
> On 03/11/2016 10:49 AM, Helin Zhang wrote:
> > -int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tag_type);
> > +int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
> > +				    enum rte_vlan_type vlan_type,
> > +				    uint16_t tag_type);
> > +int rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t tag_type);
> > +int rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
> > +					  enum rte_vlan_type vlan_type,
> > +					  uint16_t tag_type);
> >
> >   /**
> >    * Set VLAN offload configuration on an Ethernet device
> 
> Its nice to see people actually trying to be compatible on occasion :)
> 
> However in this case there's not much point in doing so, because 
> libethdev ABI has already been broken in this cycle:
> http://dpdk.org/browse/dpdk/commit/?id=cfd2279ea6299826fe992028f1dffaf9fa7e7d0a
> 
> In other words, the compatibility versions can never get invoked because 
> all software built against libethdev needs to be rebuilt anyway because 
> of the soname bump. Just drop the compat versions, no point carrying 
> around something that cannot possibly get used.

Oh yes, you are right.
Sorry Helin for having required that extra work.
On the good side, you have learnt how to do it ;)

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type
  2016-03-11 11:20  0%         ` Thomas Monjalon
@ 2016-03-11 14:17  4%           ` Zhang, Helin
  2016-03-11 14:20  3%             ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Zhang, Helin @ 2016-03-11 14:17 UTC (permalink / raw)
  To: Thomas Monjalon, Panu Matilainen; +Cc: dev



> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Friday, March 11, 2016 7:20 PM
> To: Panu Matilainen; Zhang, Helin
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether
> type
> 
> 2016-03-11 13:19, Panu Matilainen:
> > On 03/11/2016 10:49 AM, Helin Zhang wrote:
> > > -int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t
> > > tag_type);
> > > +int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
> > > +				    enum rte_vlan_type vlan_type,
> > > +				    uint16_t tag_type);
> > > +int rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t
> > > +tag_type); int rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
> > > +					  enum rte_vlan_type vlan_type,
> > > +					  uint16_t tag_type);
> > >
> > >   /**
> > >    * Set VLAN offload configuration on an Ethernet device
> >
> > Its nice to see people actually trying to be compatible on occasion :)
> >
> > However in this case there's not much point in doing so, because
> > libethdev ABI has already been broken in this cycle:
> > http://dpdk.org/browse/dpdk/commit/?id=cfd2279ea6299826fe992028f1dffaf
> > 9fa7e7d0a
> >
> > In other words, the compatibility versions can never get invoked
> > because all software built against libethdev needs to be rebuilt
> > anyway because of the soname bump. Just drop the compat versions, no
> > point carrying around something that cannot possibly get used.
> 
> Oh yes, you are right.
> Sorry Helin for having required that extra work.
> On the good side, you have learnt how to do it ;)

Yes, as Thomas said, at leat l know how to do that, and the extra work was not too big.
Thomas, Panu, thank you very much for the great comments!

Thomas, does that mean I just need to work out a new version and just let the ABI
changes as is. No ABI annoucenment will be requried? No RTE_NEXT_ABI will be used?

Regards,
Helin

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type
  2016-03-11 14:17  4%           ` Zhang, Helin
@ 2016-03-11 14:20  3%             ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-03-11 14:20 UTC (permalink / raw)
  To: Zhang, Helin; +Cc: dev

2016-03-11 14:17, Zhang, Helin:
> 
> > -----Original Message-----
> > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > Sent: Friday, March 11, 2016 7:20 PM
> > To: Panu Matilainen; Zhang, Helin
> > Cc: dev@dpdk.org
> > Subject: Re: [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether
> > type
> > 
> > 2016-03-11 13:19, Panu Matilainen:
> > > On 03/11/2016 10:49 AM, Helin Zhang wrote:
> > > > -int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t
> > > > tag_type);
> > > > +int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
> > > > +				    enum rte_vlan_type vlan_type,
> > > > +				    uint16_t tag_type);
> > > > +int rte_eth_dev_set_vlan_ether_type_v22(uint8_t port_id, uint16_t
> > > > +tag_type); int rte_eth_dev_set_vlan_ether_type_v1604(uint8_t port_id,
> > > > +					  enum rte_vlan_type vlan_type,
> > > > +					  uint16_t tag_type);
> > > >
> > > >   /**
> > > >    * Set VLAN offload configuration on an Ethernet device
> > >
> > > Its nice to see people actually trying to be compatible on occasion :)
> > >
> > > However in this case there's not much point in doing so, because
> > > libethdev ABI has already been broken in this cycle:
> > > http://dpdk.org/browse/dpdk/commit/?id=cfd2279ea6299826fe992028f1dffaf
> > > 9fa7e7d0a
> > >
> > > In other words, the compatibility versions can never get invoked
> > > because all software built against libethdev needs to be rebuilt
> > > anyway because of the soname bump. Just drop the compat versions, no
> > > point carrying around something that cannot possibly get used.
> > 
> > Oh yes, you are right.
> > Sorry Helin for having required that extra work.
> > On the good side, you have learnt how to do it ;)
> 
> Yes, as Thomas said, at leat l know how to do that, and the extra work was not too big.
> Thomas, Panu, thank you very much for the great comments!
> 
> Thomas, does that mean I just need to work out a new version and just let the ABI
> changes as is. No ABI annoucenment will be requried? No RTE_NEXT_ABI will be used?

You just need an entry in "API changes" section of the release notes.
You can state that this API change imply an ABI change.
Thanks

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH v5 0/2] i40e setting ether type of VLANs
  2016-03-11  8:49  4%   ` [dpdk-dev] [PATCH v4 " Helin Zhang
  2016-03-11  8:49  7%     ` [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type Helin Zhang
@ 2016-03-11 16:50  5%     ` Helin Zhang
  2016-03-11 16:50  7%       ` [dpdk-dev] [PATCH v5 1/2] ethdev: add vlan type for setting ether type Helin Zhang
  1 sibling, 1 reply; 200+ results
From: Helin Zhang @ 2016-03-11 16:50 UTC (permalink / raw)
  To: dev

It adds setting ether type of both single VLAN(inner VLAN) and
outer VLAN for i40e. For ixgbe and e1000/igb, it supports setting
single VLAN(inner VLAN) only, and can be extended in the future.

The patch set was branched off rel_16_04 of repo dpdk-next-net,
on below commit.
commit 5721e6447b5c20208a32c919a509e89d78c3e68d
Author: Robin Jarry <robin.jarry@6wind.com>
Date:   Thu Mar 3 15:27:40 2016 +0100
    mlx4: ensure number of RX queues is a power of 2 

v5:
 - Removed the versioning mechanism, as ABI broken is already
   there allowed.

v4:
 - Updated the doc of testpmd guide.

v3:
 - Used versioning mechanism to avoid ABI issue.
 - Re-organized the patch set.

v2:
 - Used RTE_NEXT_ABI to avoid ABI change issue.
 - Reworked the announcement of ABI change for release 16.07.
 - Fixed a i40e overflow issue.

Helin Zhang (2):
  ethdev: add vlan type for setting ether type
  i40e: fix the overflow issue

 app/test-pmd/cmdline.c                      | 30 +++++++----
 app/test-pmd/config.c                       |  9 ++--
 app/test-pmd/testpmd.h                      |  3 +-
 doc/guides/rel_notes/release_16_04.rst      |  5 ++
 doc/guides/testpmd_app_ug/testpmd_funcs.rst | 11 +---
 drivers/net/e1000/igb_ethdev.c              | 27 +++++++---
 drivers/net/i40e/i40e_ethdev.c              | 79 +++++++++++++++++++++++++++--
 drivers/net/i40e/i40e_rxtx.c                |  4 +-
 drivers/net/ixgbe/ixgbe_ethdev.c            | 25 +++++++--
 lib/librte_ether/rte_ethdev.c               |  7 +--
 lib/librte_ether/rte_ethdev.h               | 21 ++++++--
 lib/librte_ether/rte_ether_version.map      |  7 +++
 12 files changed, 181 insertions(+), 47 deletions(-)

-- 
2.5.0

^ permalink raw reply	[relevance 5%]

* [dpdk-dev] [PATCH v5 1/2] ethdev: add vlan type for setting ether type
  2016-03-11 16:50  5%     ` [dpdk-dev] [PATCH v5 0/2] i40e setting ether type of VLANs Helin Zhang
@ 2016-03-11 16:50  7%       ` Helin Zhang
  0 siblings, 0 replies; 200+ results
From: Helin Zhang @ 2016-03-11 16:50 UTC (permalink / raw)
  To: dev

In order to set ether type of VLAN for single VLAN, inner
and outer VLAN, the VLAN type as an input parameter is added
to 'rte_eth_dev_set_vlan_ether_type()'.
In addition, corresponding changes in e1000, ixgbe and i40e
are also added.

Signed-off-by: Helin Zhang <helin.zhang@intel.com>
Acked-by: Wenzhuo Lu <wenzhuo.lu@intel.com>
---
 app/test-pmd/cmdline.c                      | 30 +++++++----
 app/test-pmd/config.c                       |  9 ++--
 app/test-pmd/testpmd.h                      |  3 +-
 doc/guides/rel_notes/release_16_04.rst      |  5 ++
 doc/guides/testpmd_app_ug/testpmd_funcs.rst | 11 +---
 drivers/net/e1000/igb_ethdev.c              | 27 +++++++---
 drivers/net/i40e/i40e_ethdev.c              | 79 +++++++++++++++++++++++++++--
 drivers/net/ixgbe/ixgbe_ethdev.c            | 25 +++++++--
 lib/librte_ether/rte_ethdev.c               |  7 +--
 lib/librte_ether/rte_ethdev.h               | 21 ++++++--
 lib/librte_ether/rte_ether_version.map      |  7 +++
 11 files changed, 179 insertions(+), 45 deletions(-)

v5:
 - Removed the versioning mechanism, as ABI broken is already
   there allowed.

v4:
 - Updated the doc of testpmd guide.

v3:
 - Used versioning mechanism to avoid ABI issue.
 - Re-organized the patch set.

v2:
 - Used RTE_NEXT_ABI to avoid ABI change issue.
 - Reworked the announcement of ABI change for release 16.07.

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 52e9f5f..1eeb8a8 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -277,8 +277,8 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Set the VLAN QinQ (extended queue in queue)"
 			" on a port.\n\n"
 
-			"vlan set tpid (value) (port_id)\n"
-			"    Set the outer VLAN TPID for Packet Filtering on"
+			"vlan set (inner|outer) tpid (value) (port_id)\n"
+			"    Set the VLAN TPID for Packet Filtering on"
 			" a port\n\n"
 
 			"rx_vlan add (vlan_id|all) (port_id)\n"
@@ -297,10 +297,6 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Remove a vlan_id, to the set of VLAN identifiers"
 			"filtered for VF(s) from port_id.\n\n"
 
-			"rx_vlan set tpid (value) (port_id)\n"
-			"    Set the outer VLAN TPID for Packet Filtering on"
-			" a port\n\n"
-
 			"tunnel_filter add (port_id) (outer_mac) (inner_mac) (ip_addr) "
 			"(inner_vlan) (vxlan|nvgre) (filter_type) (tenant_id) (queue_id)\n"
 			"   add a tunnel filter of a port.\n\n"
@@ -2747,6 +2743,7 @@ cmdline_parse_inst_t cmd_rx_vlan_filter_all = {
 struct cmd_vlan_offload_result {
 	cmdline_fixed_string_t vlan;
 	cmdline_fixed_string_t set;
+	cmdline_fixed_string_t vlan_type;
 	cmdline_fixed_string_t what;
 	cmdline_fixed_string_t on;
 	cmdline_fixed_string_t port_id;
@@ -2847,6 +2844,7 @@ cmdline_parse_inst_t cmd_vlan_offload = {
 struct cmd_vlan_tpid_result {
 	cmdline_fixed_string_t vlan;
 	cmdline_fixed_string_t set;
+	cmdline_fixed_string_t vlan_type;
 	cmdline_fixed_string_t what;
 	uint16_t tp_id;
 	uint8_t port_id;
@@ -2858,8 +2856,17 @@ cmd_vlan_tpid_parsed(void *parsed_result,
 			  __attribute__((unused)) void *data)
 {
 	struct cmd_vlan_tpid_result *res = parsed_result;
-	vlan_tpid_set(res->port_id, res->tp_id);
-	return;
+	enum rte_vlan_type vlan_type;
+
+	if (!strcmp(res->vlan_type, "inner"))
+		vlan_type = ETH_VLAN_TYPE_INNER;
+	else if (!strcmp(res->vlan_type, "outer"))
+		vlan_type = ETH_VLAN_TYPE_OUTER;
+	else {
+		printf("Unknown vlan type\n");
+		return;
+	}
+	vlan_tpid_set(res->port_id, vlan_type, res->tp_id);
 }
 
 cmdline_parse_token_string_t cmd_vlan_tpid_vlan =
@@ -2868,6 +2875,9 @@ cmdline_parse_token_string_t cmd_vlan_tpid_vlan =
 cmdline_parse_token_string_t cmd_vlan_tpid_set =
 	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
 				 set, "set");
+cmdline_parse_token_string_t cmd_vlan_type =
+	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
+				 vlan_type, "inner#outer");
 cmdline_parse_token_string_t cmd_vlan_tpid_what =
 	TOKEN_STRING_INITIALIZER(struct cmd_vlan_tpid_result,
 				 what, "tpid");
@@ -2881,10 +2891,12 @@ cmdline_parse_token_num_t cmd_vlan_tpid_portid =
 cmdline_parse_inst_t cmd_vlan_tpid = {
 	.f = cmd_vlan_tpid_parsed,
 	.data = NULL,
-	.help_str = "set tpid tp_id port_id, set the Outer VLAN Ether type",
+	.help_str = "set inner|outer tpid tp_id port_id, set the VLAN "
+		    "Ether type",
 	.tokens = {
 		(void *)&cmd_vlan_tpid_vlan,
 		(void *)&cmd_vlan_tpid_set,
+		(void *)&cmd_vlan_type,
 		(void *)&cmd_vlan_tpid_what,
 		(void *)&cmd_vlan_tpid_tpid,
 		(void *)&cmd_vlan_tpid_portid,
diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 0062484..5bb09a5 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -1821,19 +1821,20 @@ rx_vlan_all_filter_set(portid_t port_id, int on)
 }
 
 void
-vlan_tpid_set(portid_t port_id, uint16_t tp_id)
+vlan_tpid_set(portid_t port_id, enum rte_vlan_type vlan_type, uint16_t tp_id)
 {
 	int diag;
+
 	if (port_id_is_invalid(port_id, ENABLED_WARN))
 		return;
 
-	diag = rte_eth_dev_set_vlan_ether_type(port_id, tp_id);
+	diag = rte_eth_dev_set_vlan_ether_type(port_id, vlan_type, tp_id);
 	if (diag == 0)
 		return;
 
-	printf("tx_vlan_tpid_set(port_pi=%d, tpid=%d) failed "
+	printf("tx_vlan_tpid_set(port_pi=%d, vlan_type=%d, tpid=%d) failed "
 	       "diag=%d\n",
-	       port_id, tp_id, diag);
+	       port_id, vlan_type, tp_id, diag);
 }
 
 void
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index b618998..0f72ca1 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -508,7 +508,8 @@ void rx_vlan_filter_set(portid_t port_id, int on);
 void rx_vlan_all_filter_set(portid_t port_id, int on);
 int rx_vft_set(portid_t port_id, uint16_t vlan_id, int on);
 void vlan_extend_set(portid_t port_id, int on);
-void vlan_tpid_set(portid_t port_id, uint16_t tp_id);
+void vlan_tpid_set(portid_t port_id, enum rte_vlan_type vlan_type,
+		   uint16_t tp_id);
 void tx_vlan_set(portid_t port_id, uint16_t vlan_id);
 void tx_qinq_set(portid_t port_id, uint16_t vlan_id, uint16_t vlan_id_outer);
 void tx_vlan_reset(portid_t port_id);
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 7c09fef..077f67b 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -125,6 +125,8 @@ This section should contain new features added in this release. Sample format:
 
   Af_packet device can now be detached using API, like other PMD devices.
 
+* **Added modifying ether type of both single and double VLAN for i40e**
+
 
 Resolved Issues
 ---------------
@@ -256,6 +258,9 @@ This section should contain API changes. Sample format:
 * Af_packet device init function is no longer public. Device should be attached
   with API.
 
+* Add one more parameter for ethdev public interface
+  of ``rte_eth_dev_set_vlan_ether_type``.
+
 
 ABI Changes
 -----------
diff --git a/doc/guides/testpmd_app_ug/testpmd_funcs.rst b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
index a520cc5..e2b2224 100644
--- a/doc/guides/testpmd_app_ug/testpmd_funcs.rst
+++ b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
@@ -498,9 +498,9 @@ Set the VLAN QinQ (extended queue in queue) on for a port::
 vlan set tpid
 ~~~~~~~~~~~~~
 
-Set the outer VLAN TPID for packet filtering on a port::
+Set the inner or outer VLAN TPID for packet filtering on a port::
 
-   testpmd> vlan set tpid (value) (port_id)
+   testpmd> vlan set (inner|outer) tpid (value) (port_id)
 
 .. note::
 
@@ -540,13 +540,6 @@ Remove a VLAN ID, from the set of VLAN identifiers filtered for VF(s) for port I
 
    testpmd> rx_vlan rm (vlan_id) port (port_id) vf (vf_mask)
 
-rx_vlan set tpid
-~~~~~~~~~~~~~~~~
-
-Set the outer VLAN TPID for packet filtering on a port::
-
-   testpmd> rx_vlan set tpid (value) (port_id)
-
 tunnel_filter add
 ~~~~~~~~~~~~~~~~~
 
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index f889876..9a34d01 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -126,7 +126,9 @@ static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
 		uint16_t vlan_id, int on);
-static void eth_igb_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
+static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
+				 enum rte_vlan_type vlan_type,
+				 uint16_t tpid_id);
 static void eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 
 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
@@ -2193,15 +2195,28 @@ eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 	return 0;
 }
 
-static void
-eth_igb_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
+static int
+eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
+		      enum rte_vlan_type vlan_type,
+		      uint16_t tpid)
 {
 	struct e1000_hw *hw =
 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-	uint32_t reg = ETHER_TYPE_VLAN ;
+	uint32_t reg = ETHER_TYPE_VLAN;
+	int ret = 0;
+
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_INNER:
+		reg |= (tpid << 16);
+		E1000_WRITE_REG(hw, E1000_VET, reg);
+		break;
+	default:
+		ret = -EINVAL;
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d\n", vlan_type);
+		break;
+	}
 
-	reg |= (tpid << 16);
-	E1000_WRITE_REG(hw, E1000_VET, reg);
+	return ret;
 }
 
 static void
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index 0c87ec1..7605a68 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -266,6 +266,11 @@
 #define I40E_INSET_IPV6_TC_MASK       0x0009F00FUL
 #define I40E_INSET_IPV6_NEXT_HDR_MASK 0x000C00FFUL
 
+#define I40E_GL_SWT_L2TAGCTRL(_i)             (0x001C0A70 + ((_i) * 4))
+#define I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT 16
+#define I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK  \
+	I40E_MASK(0xFFFF, I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT)
+
 static int eth_i40e_dev_init(struct rte_eth_dev *eth_dev);
 static int eth_i40e_dev_uninit(struct rte_eth_dev *eth_dev);
 static int i40e_dev_configure(struct rte_eth_dev *dev);
@@ -292,7 +297,9 @@ static void i40e_dev_info_get(struct rte_eth_dev *dev,
 static int i40e_vlan_filter_set(struct rte_eth_dev *dev,
 				uint16_t vlan_id,
 				int on);
-static void i40e_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid);
+static int i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+			      enum rte_vlan_type vlan_type,
+			      uint16_t tpid);
 static void i40e_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 static void i40e_vlan_strip_queue_set(struct rte_eth_dev *dev,
 				      uint16_t queue,
@@ -865,6 +872,20 @@ eth_i40e_dev_init(struct rte_eth_dev *dev)
 	hw->fc.requested_mode = I40E_FC_NONE;
 	i40e_set_fc(hw, &aq_fail, TRUE);
 
+	/* Set the global registers with default ether type value */
+	ret = i40e_vlan_tpid_set(dev, ETH_VLAN_TYPE_OUTER, 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;
+	}
+	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);
 	if (ret) {
@@ -2312,11 +2333,59 @@ i40e_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 		return i40e_vsi_delete_vlan(vsi, vlan_id);
 }
 
-static void
-i40e_vlan_tpid_set(__rte_unused struct rte_eth_dev *dev,
-		   __rte_unused uint16_t tpid)
+static int
+i40e_vlan_tpid_set(struct rte_eth_dev *dev,
+		   enum rte_vlan_type vlan_type,
+		   uint16_t tpid)
 {
-	PMD_INIT_FUNC_TRACE();
+	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	uint64_t reg_r = 0, reg_w = 0;
+	uint16_t reg_id = 0;
+	int ret = 0;
+
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_OUTER:
+		reg_id = 2;
+		break;
+	case ETH_VLAN_TYPE_INNER:
+		reg_id = 3;
+		break;
+	default:
+		ret = -EINVAL;
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d", vlan_type);
+		return ret;
+	}
+	ret = i40e_aq_debug_read_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
+					  &reg_r, NULL);
+	if (ret != I40E_SUCCESS) {
+		PMD_DRV_LOG(ERR, "Fail to debug read from "
+			    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_id);
+		ret = -EIO;
+		return ret;
+	}
+	PMD_DRV_LOG(DEBUG, "Debug read from I40E_GL_SWT_L2TAGCTRL[%d]: "
+		    "0x%08"PRIx64"", reg_id, reg_r);
+
+	reg_w = reg_r & (~(I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_MASK));
+	reg_w |= ((uint64_t)tpid << I40E_GL_SWT_L2TAGCTRL_ETHERTYPE_SHIFT);
+	if (reg_r == reg_w) {
+		ret = 0;
+		PMD_DRV_LOG(DEBUG, "No need to write");
+		return ret;
+	}
+
+	ret = i40e_aq_debug_write_register(hw, I40E_GL_SWT_L2TAGCTRL(reg_id),
+					   reg_w, NULL);
+	if (ret != I40E_SUCCESS) {
+		ret = -EIO;
+		PMD_DRV_LOG(ERR, "Fail to debug write to "
+			    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_id);
+		return ret;
+	}
+	PMD_DRV_LOG(DEBUG, "Debug write 0x%08"PRIx64" to "
+		    "I40E_GL_SWT_L2TAGCTRL[%d]", reg_w, reg_id);
+
+	return ret;
 }
 
 static void
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a9a1583..0dc050d 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -178,7 +178,9 @@ static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
 		uint16_t vlan_id, int on);
-static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
+static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
+			       enum rte_vlan_type vlan_type,
+			       uint16_t tpid_id);
 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
 		uint16_t queue, bool on);
 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
@@ -1542,14 +1544,27 @@ ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
 		ixgbe_vlan_hw_strip_disable(dev, queue);
 }
 
-static void
-ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
+static int
+ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
+		    enum rte_vlan_type vlan_type,
+		    uint16_t tpid)
 {
 	struct ixgbe_hw *hw =
 		IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+	int ret = 0;
+
+	switch (vlan_type) {
+	case ETH_VLAN_TYPE_INNER:
+		/* Only the high 16-bits is valid */
+		IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
+		break;
+	default:
+		ret = -EINVAL;
+		PMD_DRV_LOG(ERR, "Unsupported vlan type %d\n", vlan_type);
+		break;
+	}
 
-	/* Only the high 16-bits is valid */
-	IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
+	return ret;
 }
 
 void
diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.c
index a6e83c1..dedb36a 100644
--- a/lib/librte_ether/rte_ethdev.c
+++ b/lib/librte_ether/rte_ethdev.c
@@ -1697,16 +1697,17 @@ rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id, int o
 }
 
 int
-rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tpid)
+rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
+				enum rte_vlan_type vlan_type,
+				uint16_t tpid)
 {
 	struct rte_eth_dev *dev;
 
 	RTE_ETH_VALID_PORTID_OR_ERR_RET(port_id, -ENODEV);
 	dev = &rte_eth_devices[port_id];
 	RTE_FUNC_PTR_OR_ERR_RET(*dev->dev_ops->vlan_tpid_set, -ENOTSUP);
-	(*dev->dev_ops->vlan_tpid_set)(dev, tpid);
 
-	return 0;
+	return (*dev->dev_ops->vlan_tpid_set)(dev, vlan_type, tpid);
 }
 
 int
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index e2893ba..bc3f42e 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -351,6 +351,17 @@ struct rte_eth_rxmode {
 };
 
 /**
+ * VLAN types to indicate if it is for single VLAN, inner VLAN or outer VLAN.
+ * Note that single VLAN is treated the same as inner VLAN.
+ */
+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_MAX,
+};
+
+/**
  * A structure used to configure the Receive Side Scaling (RSS) feature
  * of an Ethernet port.
  * If not NULL, the *rss_key* pointer of the *rss_conf* structure points
@@ -1076,8 +1087,8 @@ typedef int (*vlan_filter_set_t)(struct rte_eth_dev *dev,
 				  int on);
 /**< @internal filtering of a VLAN Tag Identifier by an Ethernet device. */
 
-typedef void (*vlan_tpid_set_t)(struct rte_eth_dev *dev,
-				  uint16_t tpid);
+typedef int (*vlan_tpid_set_t)(struct rte_eth_dev *dev,
+			       enum rte_vlan_type type, uint16_t tpid);
 /**< @internal set the outer VLAN-TPID by an Ethernet device. */
 
 typedef void (*vlan_offload_set_t)(struct rte_eth_dev *dev, int mask);
@@ -2346,6 +2357,8 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
  *
  * @param port_id
  *   The port identifier of the Ethernet device.
+ * @vlan_type
+ *   The vlan type.
  * @param tag_type
  *   The Tag Protocol ID
  * @return
@@ -2353,7 +2366,9 @@ int rte_eth_dev_set_vlan_strip_on_queue(uint8_t port_id, uint16_t rx_queue_id,
  *   - (-ENOSUP) if hardware-assisted VLAN TPID setup is not supported.
  *   - (-ENODEV) if *port_id* invalid.
  */
-int rte_eth_dev_set_vlan_ether_type(uint8_t port_id, uint16_t tag_type);
+int rte_eth_dev_set_vlan_ether_type(uint8_t port_id,
+				    enum rte_vlan_type vlan_type,
+				    uint16_t tag_type);
 
 /**
  * Set VLAN offload configuration on an Ethernet device
diff --git a/lib/librte_ether/rte_ether_version.map b/lib/librte_ether/rte_ether_version.map
index d8db24d..6098de5 100644
--- a/lib/librte_ether/rte_ether_version.map
+++ b/lib/librte_ether/rte_ether_version.map
@@ -117,3 +117,10 @@ DPDK_2.2 {
 
 	local: *;
 };
+
+DPDK_16.04 {
+	global:
+
+	rte_eth_dev_set_vlan_ether_type;
+
+} DPDK_2.2;
-- 
2.5.0

^ permalink raw reply	[relevance 7%]

* Re: [dpdk-dev] [PATCH v9 1/5] lib/librte_ether: change function name of tunnel port config
  @ 2016-03-11 23:02  3%     ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-03-11 23:02 UTC (permalink / raw)
  To: Wenzhuo Lu; +Cc: dev

2016-03-10 10:42, Wenzhuo Lu:
> The names of function for tunnel port configuration are not
> accurate. They're tunnel_add/del, better change them to
> tunnel_port_add/del.
> As it may be an ABI change if change the names directly, the
> new functions are added but not remove the old ones. The old
> ones will be removed in the next release after an ABI change
> announcement.

As the API/ABI compatibility is already broken in this release,
I suggest to just rename the functions without keeping the old ones.
Then this patch should be merge with the next one.
I plan do it before applying.

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] vhost: remove lockless enqueue to the virtio ring
  @ 2016-03-14 23:13  0%   ` Thomas Monjalon
  2016-03-16  8:20  0%     ` Xie, Huawei
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-03-14 23:13 UTC (permalink / raw)
  To: Xie, Huawei, yuanhan.liu; +Cc: dev, ann.zhuangyanying

2016-01-05 07:16, Xie, Huawei:
> On 1/5/2016 2:42 PM, Xie, Huawei wrote:
> > This patch removes the internal lockless enqueue implmentation.
> > DPDK doesn't support receiving/transmitting packets from/to the same
> > queue. Vhost PMD wraps vhost device as normal DPDK port. DPDK
> > applications normally have their own lock implmentation when enqueue
> > packets to the same queue of a port.
> >
> > The atomic cmpset is a costly operation. This patch should help
> > performance a bit.
> >
> > Signed-off-by: Huawei Xie <huawei.xie@intel.com>
> This patch modifies the API's behavior, which is also a trivial ABI
> change. In my opinion, application shouldn't rely on previous behavior.
> Anyway, i am checking how to declare the ABI change.

I guess this patch is now obsolete?

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] vhost: remove lockless enqueue to the virtio ring
  2016-03-14 23:13  0%   ` Thomas Monjalon
@ 2016-03-16  8:20  0%     ` Xie, Huawei
  2016-03-16  8:30  0%       ` Yuanhan Liu
  0 siblings, 1 reply; 200+ results
From: Xie, Huawei @ 2016-03-16  8:20 UTC (permalink / raw)
  To: Thomas Monjalon, yuanhan.liu; +Cc: dev, ann.zhuangyanying

On 3/15/2016 7:14 AM, Thomas Monjalon wrote:
> 2016-01-05 07:16, Xie, Huawei:
>> On 1/5/2016 2:42 PM, Xie, Huawei wrote:
>>> This patch removes the internal lockless enqueue implmentation.
>>> DPDK doesn't support receiving/transmitting packets from/to the same
>>> queue. Vhost PMD wraps vhost device as normal DPDK port. DPDK
>>> applications normally have their own lock implmentation when enqueue
>>> packets to the same queue of a port.
>>>
>>> The atomic cmpset is a costly operation. This patch should help
>>> performance a bit.
>>>
>>> Signed-off-by: Huawei Xie <huawei.xie@intel.com>
>> This patch modifies the API's behavior, which is also a trivial ABI
>> change. In my opinion, application shouldn't rely on previous behavior.
>> Anyway, i am checking how to declare the ABI change.
> I guess this patch is now obsolete?

How about we delay this to next release after more considerations,
whether we should keep this behavior, and what is the best way for
concurrency in vhost.



^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] vhost: remove lockless enqueue to the virtio ring
  2016-03-16  8:20  0%     ` Xie, Huawei
@ 2016-03-16  8:30  0%       ` Yuanhan Liu
  0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-03-16  8:30 UTC (permalink / raw)
  To: Xie, Huawei; +Cc: Thomas Monjalon, dev, ann.zhuangyanying

On Wed, Mar 16, 2016 at 08:20:37AM +0000, Xie, Huawei wrote:
> On 3/15/2016 7:14 AM, Thomas Monjalon wrote:
> > 2016-01-05 07:16, Xie, Huawei:
> >> On 1/5/2016 2:42 PM, Xie, Huawei wrote:
> >>> This patch removes the internal lockless enqueue implmentation.
> >>> DPDK doesn't support receiving/transmitting packets from/to the same
> >>> queue. Vhost PMD wraps vhost device as normal DPDK port. DPDK
> >>> applications normally have their own lock implmentation when enqueue
> >>> packets to the same queue of a port.
> >>>
> >>> The atomic cmpset is a costly operation. This patch should help
> >>> performance a bit.
> >>>
> >>> Signed-off-by: Huawei Xie <huawei.xie@intel.com>
> >> This patch modifies the API's behavior, which is also a trivial ABI
> >> change. In my opinion, application shouldn't rely on previous behavior.
> >> Anyway, i am checking how to declare the ABI change.
> > I guess this patch is now obsolete?
> 
> How about we delay this to next release after more considerations,

I'd suggest so.

> whether we should keep this behavior, and what is the best way for
> concurrency in vhost.

I'm wondering should we do an announcement first, to notify user the
behaviour change?

	--yliu

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07
  @ 2016-03-17  9:05 15% ` Olivier Matz
  2016-04-04 14:38  4%   ` Thomas Monjalon
  2016-04-14 10:19  2% ` [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation Olivier Matz
  1 sibling, 1 reply; 200+ results
From: Olivier Matz @ 2016-03-17  9:05 UTC (permalink / raw)
  To: dev

Add a deprecation notice for coming changes in mempool for 16.07.

Signed-off-by: Olivier Matz <olivier.matz@6wind.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 252a096..3e8e327 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -33,3 +33,11 @@ Deprecation Notices
 * 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.
+
+* librte_mempool: new fixes and features will be added in 16.07:
+  allocation of large mempool in several virtual memory chunks, new API
+  to populate a mempool, new API to free a mempool, allocation in
+  anonymous mapping, drop of specific dom0 code. These changes will
+  induce a modification of the rte_mempool structure, plus a
+  modification of the API of rte_mempool_obj_iter(), implying a breakage
+  of the ABI.
-- 
2.1.4

^ permalink raw reply	[relevance 15%]

* [dpdk-dev] [PATCH v5 0/9] extend flow director fields in i40e driver
  @ 2016-03-21  6:18  4% ` Jingjing Wu
  2016-03-21  6:18 18%   ` [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection Jingjing Wu
  2016-03-23 13:07  4%   ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Jingjing Wu
  0 siblings, 2 replies; 200+ results
From: Jingjing Wu @ 2016-03-21  6:18 UTC (permalink / raw)
  To: dev; +Cc: jingjing.wu, helin.zhang

v5 changes:
 - remove the reorganizing of struct rte_eth_fdir_flow
 - remove fdir supporting on Tunnel Id
 - rebase to latest dpdk/master

v4 changes:
 - rebase to latest dpdk-next-net/rel_16_04.
 - comments on new fields in API structure.

v3 changes:
 - rebase to latest dpdk-next-net/rel_16_04(commit: 0f9564a0e4f2)
 - use AQ rx control register read/write for some registers
 - remove few useless lines
 - patch title rewording

v2 changes:
 - rebase on dpdk-next-net/rel_16_04
 - comments rewording.
 - redefine the value of RTE_ETH_INPUT_SET_L3_IP4_TTL to
   avoid ABI breaking.
 - remove ABI announce in Deprecation.
 - fix the ethertype setting when program filter in v1 patch set.

This patch set extends flow director to support filtering by
additional fields below in i40e driver:
 - TOS, Protocol and TTL in IP header
 - single vlan or inner vlan 




Andrey Chilikin (1):
  i40e: fix VLAN bitmasks for input set

Jingjing Wu (8):
  ethdev: extend flow director for input selection
  i40e: split function for hash and fdir input
  i40e: remove flex payload from input selection
  i40e: restore default setting on input set
  i40e: extend flow director to filter by IP Header
  testpmd: extend input set related commands
  i40e: extend flow director to filter by vlan id
  testpmd: extend flow director commands

 app/test-pmd/cmdline.c                      | 100 +++++--
 doc/guides/rel_notes/deprecation.rst        |   4 -
 doc/guides/rel_notes/release_16_04.rst      |   4 +
 doc/guides/testpmd_app_ug/testpmd_funcs.rst |  42 +--
 drivers/net/i40e/i40e_ethdev.c              | 393 ++++++++++++++++------------
 drivers/net/i40e/i40e_ethdev.h              |  11 +-
 drivers/net/i40e/i40e_fdir.c                | 100 ++++---
 lib/librte_ether/rte_eth_ctrl.h             |   8 +
 8 files changed, 416 insertions(+), 246 deletions(-)

-- 
2.4.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection
  2016-03-21  6:18  4% ` [dpdk-dev] [PATCH v5 0/9] " Jingjing Wu
@ 2016-03-21  6:18 18%   ` Jingjing Wu
  2016-03-22 22:05  0%     ` Thomas Monjalon
  2016-03-23 13:07  4%   ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Jingjing Wu
  1 sibling, 1 reply; 200+ results
From: Jingjing Wu @ 2016-03-21  6:18 UTC (permalink / raw)
  To: dev; +Cc: jingjing.wu, helin.zhang

This patch added RTE_ETH_INPUT_SET_L3_IP4_TTL,
RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS input field type and extended
struct rte_eth_ipv4_flow and rte_eth_ipv6_flow to support filtering
by tos, protocol and ttl.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
---
 doc/guides/rel_notes/deprecation.rst   | 4 ----
 doc/guides/rel_notes/release_16_04.rst | 2 ++
 lib/librte_ether/rte_eth_ctrl.h        | 8 ++++++++
 3 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 252a096..e7a7c7f 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -22,10 +22,6 @@ Deprecation Notices
 * The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
   must be updated to support 100G link and to have a cleaner link speed API.
 
-* ABI changes are planned for struct rte_eth_fdir_flow in order to support
-  extend flow director's input set. The release 2.2 does not contain these ABI
-  changes, but release 2.3 will, and no backwards compatibility is planned.
-
 * ABI changes are planned for rte_eth_ipv4_flow and rte_eth_ipv6_flow to
   include more fields to be matched against. The release 2.2 does not
   contain these ABI changes, but release 2.3 will.
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 2785b29..5803684 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -435,6 +435,8 @@ ABI Changes
 
 * The cmdline buffer size has been increase from 256 to 512.
 
+* The ethdev flow director structure ``rte_eth_fdir_flow`` structure was
+  changed. New fields were added to extend flow director's input set.
 
 Shared Library Versions
 -----------------------
diff --git a/lib/librte_ether/rte_eth_ctrl.h b/lib/librte_ether/rte_eth_ctrl.h
index 6e2f617..aabd724 100644
--- a/lib/librte_ether/rte_eth_ctrl.h
+++ b/lib/librte_ether/rte_eth_ctrl.h
@@ -343,6 +343,8 @@ enum rte_eth_input_set_field {
 	RTE_ETH_INPUT_SET_L3_IP4_PROTO,
 	RTE_ETH_INPUT_SET_L3_IP6_TC,
 	RTE_ETH_INPUT_SET_L3_IP6_NEXT_HEADER,
+	RTE_ETH_INPUT_SET_L3_IP4_TTL,
+	RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS,
 
 	/* L4 */
 	RTE_ETH_INPUT_SET_L4_UDP_SRC_PORT = 257,
@@ -410,6 +412,9 @@ struct rte_eth_l2_flow {
 struct rte_eth_ipv4_flow {
 	uint32_t src_ip;      /**< IPv4 source address to match. */
 	uint32_t dst_ip;      /**< IPv4 destination address to match. */
+	uint8_t  tos;         /**< Type of service to match. */
+	uint8_t  ttl;         /**< Time to live to match. */
+	uint8_t  proto;       /**< Protocol, next header to match. */
 };
 
 /**
@@ -446,6 +451,9 @@ struct rte_eth_sctpv4_flow {
 struct rte_eth_ipv6_flow {
 	uint32_t src_ip[4];      /**< IPv6 source address to match. */
 	uint32_t dst_ip[4];      /**< IPv6 destination address to match. */
+	uint8_t  tc;             /**< Traffic class to match. */
+	uint8_t  proto;          /**< Protocol, next header to match. */
+	uint8_t  hop_limits;     /**< Hop limits to match. */
 };
 
 /**
-- 
2.4.0

^ permalink raw reply	[relevance 18%]

* Re: [dpdk-dev] [PATCH] mempool: allow for user-owned mempool caches
  @ 2016-03-21 12:22  4% ` Olivier Matz
  2016-03-21 13:49  3%   ` Wiles, Keith
  0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-03-21 12:22 UTC (permalink / raw)
  To: Lazaros Koromilas, dev

Hi Lazaros,

Thanks for this patch. To me, this is a valuable enhancement.
Please find some comments inline.

On 03/10/2016 03:44 PM, Lazaros Koromilas wrote:
> The mempool cache is only available to EAL threads as a per-lcore
> resource. Change this so that the user can create and provide their own
> cache on mempool get and put operations. This works with non-EAL threads
> too. This commit introduces new API calls with the 'with_cache' suffix,
> while the current ones default to the per-lcore local cache.
> 
> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
> ---
>  lib/librte_mempool/rte_mempool.c |  65 +++++-
>  lib/librte_mempool/rte_mempool.h | 442 ++++++++++++++++++++++++++++++++++++---
>  2 files changed, 467 insertions(+), 40 deletions(-)
> 
> diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
> index f8781e1..cebc2b7 100644
> --- a/lib/librte_mempool/rte_mempool.c
> +++ b/lib/librte_mempool/rte_mempool.c
> @@ -375,6 +375,43 @@ rte_mempool_xmem_usage(void *vaddr, uint32_t elt_num, size_t elt_sz,
>  	return usz;
>  }
>  
> +#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0

I wonder if this wouldn't cause a conflict with Keith's patch
that removes some #ifdefs RTE_MEMPOOL_CACHE_MAX_SIZE.
See: http://www.dpdk.org/dev/patchwork/patch/10492/

As this patch is already acked for 16.07, I think that your v2
could be rebased on top of it to avoid conflicts when Thomas will apply
it.

By the way, I also encourage you to have a look at other works in
progress in mempool:
http://www.dpdk.org/ml/archives/dev/2016-March/035107.html
http://www.dpdk.org/ml/archives/dev/2016-March/035201.html


> +static void
> +mempool_cache_init(struct rte_mempool_cache *cache, uint32_t size)
> +{
> +	cache->size = size;
> +	cache->flushthresh = CALC_CACHE_FLUSHTHRESH(size);
> +	cache->len = 0;
> +}
> +
> +/*
> + * Creates and initializes a cache for objects that are retrieved from and
> + * returned to an underlying mempool. This structure is identical to the
> + * structure included inside struct rte_mempool.
> + */

On top of Keith's patch, this comment may be reworked as the cache
structure is not included in the mempool structure anymore.

nit: I think the imperative form is preferred


> +struct rte_mempool_cache *
> +rte_mempool_cache_create(uint32_t size)
> +{
> +	struct rte_mempool_cache *cache;
> +
> +	if (size > RTE_MEMPOOL_CACHE_MAX_SIZE) {
> +		rte_errno = EINVAL;
> +		return NULL;
> +	}
> +
> +	cache = rte_zmalloc("MEMPOOL_CACHE", sizeof(*cache), RTE_CACHE_LINE_SIZE);
> +	if (cache == NULL) {
> +		RTE_LOG(ERR, MEMPOOL, "Cannot allocate mempool cache!\n");

I would remove the '!'



> @@ -587,10 +624,18 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
>  	mp->elt_size = objsz.elt_size;
>  	mp->header_size = objsz.header_size;
>  	mp->trailer_size = objsz.trailer_size;
> -	mp->cache_size = cache_size;
> -	mp->cache_flushthresh = CALC_CACHE_FLUSHTHRESH(cache_size);
> +	mp->cache_size = cache_size; /* Keep this for backwards compat. */

I'm wondering if this should be kept for compat or if it makes sense
to keep it. The question is: do we want the cache_size to be a parameter
of the mempool or should it be a parameter of the cache?

I think we could remove this field from the mempool structure.


> @@ -673,13 +718,17 @@ rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
>  #if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>  	unsigned lcore_id;
>  	unsigned count = 0;
> +	unsigned cache_size;
>  	unsigned cache_count;
>  
>  	fprintf(f, "  cache infos:\n");
> -	fprintf(f, "    cache_size=%"PRIu32"\n", mp->cache_size);
>  	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
> +		cache_size = mp->local_cache[lcore_id].size;
> +		fprintf(f, "    cache_size[%u]=%"PRIu32"\n",
> +			lcore_id, cache_size);
>  		cache_count = mp->local_cache[lcore_id].len;
> -		fprintf(f, "    cache_count[%u]=%u\n", lcore_id, cache_count);
> +		fprintf(f, "    cache_count[%u]=%"PRIu32"\n",
> +			lcore_id, cache_count);
>  		count += cache_count;

Does it still make sense to dump the content of the cache as some
external caches may exist?

If we want to list all caches, we could imagine a sort of cache
registration before using a cache structure. Example:

   cache = rte_mempool_add_cache()
   rte_mempool_del_cache()

All the caches could be browsed internally using a list.

Thoughts?


> --- a/lib/librte_mempool/rte_mempool.h
> +++ b/lib/librte_mempool/rte_mempool.h
> @@ -95,19 +95,19 @@ struct rte_mempool_debug_stats {
>  } __rte_cache_aligned;
>  #endif
>  
> -#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>  /**
>   * A structure that stores a per-core object cache.
>   */
>  struct rte_mempool_cache {
> -	unsigned len; /**< Cache len */
> +	uint32_t size;        /**< Size of the cache */
> +	uint32_t flushthresh; /**< Threshold before we flush excess elements */
> +	uint32_t len;         /**< Current cache count */
>  	/*
>  	 * Cache is allocated to this size to allow it to overflow in certain
>  	 * cases to avoid needless emptying of cache.
>  	 */
>  	void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 3]; /**< Cache objects */
>  } __rte_cache_aligned;
> -#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
>  
>  /**
>   * A structure that stores the size of mempool elements.
> @@ -185,8 +185,6 @@ struct rte_mempool {
>  	int flags;                       /**< Flags of the mempool. */
>  	uint32_t size;                   /**< Size of the mempool. */
>  	uint32_t cache_size;             /**< Size of per-lcore local cache. */
> -	uint32_t cache_flushthresh;
> -	/**< Threshold before we flush excess elements. */
>  
>  	uint32_t elt_size;               /**< Size of an element. */
>  	uint32_t header_size;            /**< Size of header (before elt). */

Adding and removing these fields in the structure will break
the ABI. Could you please send a deprecation notice for it for
16.04, in the same model as http://dpdk.org/dev/patchwork/patch/11553/

The process is to get 3 acks for this deprecation notice, but I think
this won't be an issue as there are already several changes scheduled
in mempool that will break the ABI in 16.07, so it's the right time to
do it.



>  /**
> + * Put several objects back in the mempool (multi-producers safe).
> + * Use a user-provided mempool cache.
> + *
> + * @param mp
> + *   A pointer to the mempool structure.
> + * @param obj_table
> + *   A pointer to a table of void * pointers (objects).
> + * @param n
> + *   The number of objects to add in the mempool from the obj_table.
> + * @param cache
> + *   A pointer to a mempool cache structure. May be NULL if not needed.
> + */
> +static inline void __attribute__((always_inline))
> +rte_mempool_mp_put_bulk_with_cache(struct rte_mempool *mp,
> +				   void * const *obj_table, unsigned n,
> +				   struct rte_mempool_cache *cache)
> +{
> +	__mempool_check_cookies(mp, obj_table, n, 0);
> +	__mempool_put_bulk_with_cache(mp, obj_table, n, cache, 1);
> +}
> +
> +/**
> + * Put several objects back in the mempool (NOT multi-producers safe).
> + * Use a user-provided mempool cache.
> + *
> + * @param mp
> + *   A pointer to the mempool structure.
> + * @param obj_table
> + *   A pointer to a table of void * pointers (objects).
> + * @param n
> + *   The number of objects to add in the mempool from obj_table.
> + * @param cache
> + *   A pointer to a mempool cache structure. May be NULL if not needed.
> + */
> +static inline void
> +rte_mempool_sp_put_bulk_with_cache(struct rte_mempool *mp,
> +				   void * const *obj_table, unsigned n,
> +				   struct rte_mempool_cache *cache)
> +{
> +	__mempool_check_cookies(mp, obj_table, n, 0);
> +	__mempool_put_bulk_with_cache(mp, obj_table, n, cache, 0);
> +}
>
> [...]

Many functions are added here. As all of these functions are inline,
what would you think to have instead one new functions that would
match all cases:

static inline void
rte_mempool_generic_put(struct rte_mempool *mp,
		void * const *obj_table, unsigned n,
		struct rte_mempool_cache *cache,
		unsigned flags)

We could then consider the deprecation of some functions. Today,
without your patch, we have for put():

	rte_mempool_mp_put_bulk(mp, obj_table, n)
	rte_mempool_sp_put_bulk(mp, obj_table, n)
	rte_mempool_put_bulk(mp, obj_table, n)
	rte_mempool_mp_put(mp, obj)
	rte_mempool_sp_put(mp, obj)
	rte_mempool_put(mp, obj)

	__mempool_put_bulk(mp, obj_table, n, is_mp)  /* internal */

Maybe we should only keep:

	rte_mempool_put()
	rte_mempool_put_bulk(mp, obj_table, n)
	rte_mempool_generic_put(mp, obj_table, n, cache, flags)

and same for *get().



> +/**
> + * Create a user-owned mempool cache. This can be used by non-EAL threads
> + * to enable caching when they interact with a mempool.

nit: the doxygen format needs a one-line title

> + *
> + * @param size
> + *   The size of the mempool cache. See rte_mempool_create()'s cache_size
> + *   parameter description for more information. The same limits and
> + *   considerations apply here too.
> + */
> +struct rte_mempool_cache *
> +rte_mempool_cache_create(uint32_t size);
> +

I think we should also consider a free() function.

Maybe we should explain a bit more in the help of this function that
a mempool embeds a per-lcore cache by default.

Also, it would be great to have a test in app/test_mempool.c to validate
the feature and have an example of use.

And last thing, this is probably out of scope for now, but I'm
wondering if in the future this external cache could be used to
transparently share a pool (typically a mbuf pool), for instance
in case of a secondary process. Each process would have its own cache,
referencing the same shared mempool structure. This would probably
require to update the ethdev and mbuf API, so it's certainly a quite
large modification. If you have any ideas or plans going in this
direction, I would be interested.


Regards,
Olivier

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] mempool: allow for user-owned mempool caches
  2016-03-21 12:22  4% ` Olivier Matz
@ 2016-03-21 13:49  3%   ` Wiles, Keith
  2016-03-24 14:35  0%     ` Lazaros Koromilas
  0 siblings, 1 reply; 200+ results
From: Wiles, Keith @ 2016-03-21 13:49 UTC (permalink / raw)
  To: Olivier Matz, Lazaros Koromilas, dev

>Hi Lazaros,
>
>Thanks for this patch. To me, this is a valuable enhancement.
>Please find some comments inline.
>
>On 03/10/2016 03:44 PM, Lazaros Koromilas wrote:
>> The mempool cache is only available to EAL threads as a per-lcore
>> resource. Change this so that the user can create and provide their own
>> cache on mempool get and put operations. This works with non-EAL threads
>> too. This commit introduces new API calls with the 'with_cache' suffix,
>> while the current ones default to the per-lcore local cache.
>> 
>> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
>> ---
>>  lib/librte_mempool/rte_mempool.c |  65 +++++-
>>  lib/librte_mempool/rte_mempool.h | 442 ++++++++++++++++++++++++++++++++++++---
>>  2 files changed, 467 insertions(+), 40 deletions(-)
>> 
>> diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
>> index f8781e1..cebc2b7 100644
>> --- a/lib/librte_mempool/rte_mempool.c
>> +++ b/lib/librte_mempool/rte_mempool.c
>> @@ -375,6 +375,43 @@ rte_mempool_xmem_usage(void *vaddr, uint32_t elt_num, size_t elt_sz,
>>  	return usz;
>>  }
>>  
>> +#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>
>I wonder if this wouldn't cause a conflict with Keith's patch
>that removes some #ifdefs RTE_MEMPOOL_CACHE_MAX_SIZE.
>See: http://www.dpdk.org/dev/patchwork/patch/10492/

Hi Lazaros,

The patch I submitted keeps the mempool cache structure (pointers and variables) and only allocates the cache if specified by the caller to use a cache. This means to me the caller could fill in the cache pointer and values into the mempool structure to get a cache without a lot of extra code. If we added a set of APIs to fill in these structure variables would that not give you the external cache support. I have not really looked at the patch to verify this will work, but it sure seems like it.

So my suggestion the caller can just create a mempool without a cache and then call a set of APIs to fill in his cache values, does that not work?

If we can do this it reduces the API and possible the ABI changes to mempool as the new cache create routines and APIs could be in a new file I think, which just updates the mempool structure correctly.

>
>As this patch is already acked for 16.07, I think that your v2
>could be rebased on top of it to avoid conflicts when Thomas will apply
>it.
>
>By the way, I also encourage you to have a look at other works in
>progress in mempool:
>http://www.dpdk.org/ml/archives/dev/2016-March/035107.html
>http://www.dpdk.org/ml/archives/dev/2016-March/035201.html
>
>

Regards,
Keith





^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] DPDK and HW offloads
  @ 2016-03-21 15:26  3%         ` Kyle Larose
  2016-03-22  5:50  0%           ` Qiu, Michael
  0 siblings, 1 reply; 200+ results
From: Kyle Larose @ 2016-03-21 15:26 UTC (permalink / raw)
  To: Bruce Richardson; +Cc: Thomas Monjalon, Zhang, Helin, Stephen Hemminger, dev

On Mon, Mar 21, 2016 at 10:52 AM, Bruce Richardson
<bruce.richardson@intel.com> wrote:
> On Sun, Mar 20, 2016 at 08:18:57PM +0100, Thomas Monjalon wrote:
>> 2016-03-20 14:17, Zhang, Helin:
>> > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
>> > > 2016-03-18 10:16, Stephen Hemminger:
>> > > > Right now, all those offload features are pretty much unusable in a
>> > > > real product without lots and lots of extra codes and huge bug
>> > > > surface. It bothers me enough that I would recommend removing much of the
>> > > filter/offload/ptype stuff from DPDK!
>> > >
>> > > One of the biggest challenge is to think about a good filtering API.
>> > > The offloading has some interaction with the mbuf struct.
>> > >
>> > > I would like to suggest rewriting ethdev API by keeping it as is for some time for
>> > > compatibility while creating a new one. What about the prefix dpdk_netdev_ to
>> > > progressively replace rte_eth_dev?
>> >
>> > I totally agree with to add new and generic APIs for user applications. But I don't
>> > think we need to remove all current APIs. Generic APIs may not support all advanced
>> > hardware features, while specific APIs can. Why not support all? One generic APIs for
>> > common users, and others APIs for advanced users.
>>
>> Yes we cannot access to every features of a device through generic API.
>> Until now we were trying to add an ethdev API for every features even if it
>> is used by only one driver.
>> I think we should allow a direct access to the driver by the applications and
>> work on generic API only for common features.
>
> Definite +1.
> I think that we need to start pushing driver-specific functionality to get exposed
> via a driver's header files. That allow users who want to extract the max
> functionality from a particular NIC to do so via those APIs calls, while not
> polluting the generic ethdev layer.
>

What sort of requirements on ABI/API compatibility would this place on
the drivers? I would hope that it would be treated like any other
public API within DPDK. I don't think this would be too onerous, but
it would require that the drivers be designed to deal with it. (I.e.
don't just expose any old internal driver function).

> On the other hand, I don't like the idea of dpdk_netdev. I think we can work
> within the existing rte_eth_dev framework.
>
> /Bruce
>

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] DPDK and HW offloads
  2016-03-21 15:26  3%         ` Kyle Larose
@ 2016-03-22  5:50  0%           ` Qiu, Michael
  2016-03-22 10:19  0%             ` Bruce Richardson
  0 siblings, 1 reply; 200+ results
From: Qiu, Michael @ 2016-03-22  5:50 UTC (permalink / raw)
  To: Kyle Larose, Richardson, Bruce
  Cc: Thomas Monjalon, Zhang, Helin, Stephen Hemminger, dev

On 3/21/2016 11:27 PM, Kyle Larose wrote:
> On Mon, Mar 21, 2016 at 10:52 AM, Bruce Richardson
> <bruce.richardson@intel.com> wrote:
>> On Sun, Mar 20, 2016 at 08:18:57PM +0100, Thomas Monjalon wrote:
>>> 2016-03-20 14:17, Zhang, Helin:
>>>> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
>>>>> 2016-03-18 10:16, Stephen Hemminger:
>>>>>> Right now, all those offload features are pretty much unusable in a
>>>>>> real product without lots and lots of extra codes and huge bug
>>>>>> surface. It bothers me enough that I would recommend removing much of the
>>>>> filter/offload/ptype stuff from DPDK!
>>>>>
>>>>> One of the biggest challenge is to think about a good filtering API.
>>>>> The offloading has some interaction with the mbuf struct.
>>>>>
>>>>> I would like to suggest rewriting ethdev API by keeping it as is for some time for
>>>>> compatibility while creating a new one. What about the prefix dpdk_netdev_ to
>>>>> progressively replace rte_eth_dev?
>>>> I totally agree with to add new and generic APIs for user applications. But I don't
>>>> think we need to remove all current APIs. Generic APIs may not support all advanced
>>>> hardware features, while specific APIs can. Why not support all? One generic APIs for
>>>> common users, and others APIs for advanced users.
>>> Yes we cannot access to every features of a device through generic API.
>>> Until now we were trying to add an ethdev API for every features even if it
>>> is used by only one driver.
>>> I think we should allow a direct access to the driver by the applications and
>>> work on generic API only for common features.
>> Definite +1.
>> I think that we need to start pushing driver-specific functionality to get exposed
>> via a driver's header files. That allow users who want to extract the max
>> functionality from a particular NIC to do so via those APIs calls, while not
>> polluting the generic ethdev layer.
>>
> What sort of requirements on ABI/API compatibility would this place on
> the drivers? I would hope that it would be treated like any other
> public API within DPDK. I don't think this would be too onerous, but
> it would require that the drivers be designed to deal with it. (I.e.
> don't just expose any old internal driver function).

Why not to implement one simple API with variable arguments, just like
syscall ioctl() does. And drivers implement it's specific hardware
features with a feature bit param, and other needed variable arguments.

Thanks,
Michael
>> On the other hand, I don't like the idea of dpdk_netdev. I think we can work
>> within the existing rte_eth_dev framework.
>>
>> /Bruce
>>


^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] DPDK and HW offloads
  2016-03-22  5:50  0%           ` Qiu, Michael
@ 2016-03-22 10:19  0%             ` Bruce Richardson
  2016-03-23  2:47  0%               ` Qiu, Michael
  0 siblings, 1 reply; 200+ results
From: Bruce Richardson @ 2016-03-22 10:19 UTC (permalink / raw)
  To: Qiu, Michael
  Cc: Kyle Larose, Thomas Monjalon, Zhang, Helin, Stephen Hemminger, dev

On Tue, Mar 22, 2016 at 05:50:28AM +0000, Qiu, Michael wrote:
> On 3/21/2016 11:27 PM, Kyle Larose wrote:
> > On Mon, Mar 21, 2016 at 10:52 AM, Bruce Richardson
> > <bruce.richardson@intel.com> wrote:
> >> On Sun, Mar 20, 2016 at 08:18:57PM +0100, Thomas Monjalon wrote:
> >>> 2016-03-20 14:17, Zhang, Helin:
> >>>> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> >>>>> 2016-03-18 10:16, Stephen Hemminger:
> >>>>>> Right now, all those offload features are pretty much unusable in a
> >>>>>> real product without lots and lots of extra codes and huge bug
> >>>>>> surface. It bothers me enough that I would recommend removing much of the
> >>>>> filter/offload/ptype stuff from DPDK!
> >>>>>
> >>>>> One of the biggest challenge is to think about a good filtering API.
> >>>>> The offloading has some interaction with the mbuf struct.
> >>>>>
> >>>>> I would like to suggest rewriting ethdev API by keeping it as is for some time for
> >>>>> compatibility while creating a new one. What about the prefix dpdk_netdev_ to
> >>>>> progressively replace rte_eth_dev?
> >>>> I totally agree with to add new and generic APIs for user applications. But I don't
> >>>> think we need to remove all current APIs. Generic APIs may not support all advanced
> >>>> hardware features, while specific APIs can. Why not support all? One generic APIs for
> >>>> common users, and others APIs for advanced users.
> >>> Yes we cannot access to every features of a device through generic API.
> >>> Until now we were trying to add an ethdev API for every features even if it
> >>> is used by only one driver.
> >>> I think we should allow a direct access to the driver by the applications and
> >>> work on generic API only for common features.
> >> Definite +1.
> >> I think that we need to start pushing driver-specific functionality to get exposed
> >> via a driver's header files. That allow users who want to extract the max
> >> functionality from a particular NIC to do so via those APIs calls, while not
> >> polluting the generic ethdev layer.
> >>
> > What sort of requirements on ABI/API compatibility would this place on
> > the drivers? I would hope that it would be treated like any other
> > public API within DPDK. I don't think this would be too onerous, but
> > it would require that the drivers be designed to deal with it. (I.e.
> > don't just expose any old internal driver function).
> 
> Why not to implement one simple API with variable arguments, just like
> syscall ioctl() does. And drivers implement it's specific hardware
> features with a feature bit param, and other needed variable arguments.
> 
> Thanks,
> Michael

A very much dislike that idea. 
* It makes the code much harder to read as you have to closely examine all the
  parameters to work out what a function call is actually meant to do.
* It makes it much harder to see that you have an implicit dependency on a
  specific device. Having to include a driver specific header file e.g. i40e.h,
  and call a function named e.g. i40e_do_magic_stuff(), makes it pretty explicit
  that you have a dependency on i40e-based hardware
* It prevents the compiler from doing type-checking on parameters and informing
  you of little inconsistencies.

For all these reasons, I prefer the device-specific functions option. However,
at the same time, we also need to ensure we have a reasonable set of generic
APIs so that the cases where users are forced to drop down to the lower-level
device-specific primitives are reduced.

Regards,
/Bruce

> >> On the other hand, I don't like the idea of dpdk_netdev. I think we can work
> >> within the existing rte_eth_dev framework.
> >>
> >> /Bruce
> >>
> 
> 

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection
  2016-03-21  6:18 18%   ` [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection Jingjing Wu
@ 2016-03-22 22:05  0%     ` Thomas Monjalon
  2016-03-23  0:42  0%       ` Wu, Jingjing
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-03-22 22:05 UTC (permalink / raw)
  To: Jingjing Wu; +Cc: dev, helin.zhang

2016-03-21 14:18, Jingjing Wu:
> This patch added RTE_ETH_INPUT_SET_L3_IP4_TTL,
> RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS input field type and extended
> struct rte_eth_ipv4_flow and rte_eth_ipv6_flow to support filtering
> by tos, protocol and ttl.
[...]
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -22,10 +22,6 @@ Deprecation Notices
>  * The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
>    must be updated to support 100G link and to have a cleaner link speed API.
>  
> -* ABI changes are planned for struct rte_eth_fdir_flow in order to support
> -  extend flow director's input set. The release 2.2 does not contain these ABI
> -  changes, but release 2.3 will, and no backwards compatibility is planned.

The changed structures are part of rte_eth_fdir_flow.
So this deprecation notice apply to this patch.

>  * ABI changes are planned for rte_eth_ipv4_flow and rte_eth_ipv6_flow to
>    include more fields to be matched against. The release 2.2 does not
>    contain these ABI changes, but release 2.3 will.

These are the structures changed in this patch.
I think this section must be also removed.

> --- a/doc/guides/rel_notes/release_16_04.rst
> +++ b/doc/guides/rel_notes/release_16_04.rst
> @@ -435,6 +435,8 @@ ABI Changes
>  
>  * The cmdline buffer size has been increase from 256 to 512.
>  
> +* The ethdev flow director structure ``rte_eth_fdir_flow`` structure was
> +  changed. New fields were added to extend flow director's input set.

For reading ease, it's better to group ethdev changes (before cmdline change).

> --- a/lib/librte_ether/rte_eth_ctrl.h
> +++ b/lib/librte_ether/rte_eth_ctrl.h
> @@ -343,6 +343,8 @@ enum rte_eth_input_set_field {
>  	RTE_ETH_INPUT_SET_L3_IP4_PROTO,
>  	RTE_ETH_INPUT_SET_L3_IP6_TC,
>  	RTE_ETH_INPUT_SET_L3_IP6_NEXT_HEADER,
> +	RTE_ETH_INPUT_SET_L3_IP4_TTL,
> +	RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS,
>  
>  	/* L4 */
>  	RTE_ETH_INPUT_SET_L4_UDP_SRC_PORT = 257,
> @@ -410,6 +412,9 @@ struct rte_eth_l2_flow {
>  struct rte_eth_ipv4_flow {
>  	uint32_t src_ip;      /**< IPv4 source address to match. */
>  	uint32_t dst_ip;      /**< IPv4 destination address to match. */
> +	uint8_t  tos;         /**< Type of service to match. */
> +	uint8_t  ttl;         /**< Time to live to match. */
> +	uint8_t  proto;       /**< Protocol, next header to match. */
>  };
>  
>  /**
> @@ -446,6 +451,9 @@ struct rte_eth_sctpv4_flow {
>  struct rte_eth_ipv6_flow {
>  	uint32_t src_ip[4];      /**< IPv6 source address to match. */
>  	uint32_t dst_ip[4];      /**< IPv6 destination address to match. */
> +	uint8_t  tc;             /**< Traffic class to match. */
> +	uint8_t  proto;          /**< Protocol, next header to match. */
> +	uint8_t  hop_limits;     /**< Hop limits to match. */
>  };

This extension of the existing API looks OK.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection
  2016-03-22 22:05  0%     ` Thomas Monjalon
@ 2016-03-23  0:42  0%       ` Wu, Jingjing
  2016-03-23  8:45  0%         ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Wu, Jingjing @ 2016-03-23  0:42 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev, Zhang, Helin

Hi, Thomas

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Wednesday, March 23, 2016 6:06 AM
> To: Wu, Jingjing
> Cc: dev@dpdk.org; Zhang, Helin
> Subject: Re: [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for
> input selection
> 
> 2016-03-21 14:18, Jingjing Wu:
> > This patch added RTE_ETH_INPUT_SET_L3_IP4_TTL,
> > RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS input field type and extended
> > struct rte_eth_ipv4_flow and rte_eth_ipv6_flow to support filtering by
> > tos, protocol and ttl.
> [...]
> > --- a/doc/guides/rel_notes/deprecation.rst
> > +++ b/doc/guides/rel_notes/deprecation.rst
> > @@ -22,10 +22,6 @@ Deprecation Notices
> >  * The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
> >    must be updated to support 100G link and to have a cleaner link speed
> API.
> >
> > -* ABI changes are planned for struct rte_eth_fdir_flow in order to
> > support
> > -  extend flow director's input set. The release 2.2 does not contain
> > these ABI
> > -  changes, but release 2.3 will, and no backwards compatibility is planned.
> 
> The changed structures are part of rte_eth_fdir_flow.
> So this deprecation notice apply to this patch.
> 
> >  * ABI changes are planned for rte_eth_ipv4_flow and rte_eth_ipv6_flow
> to
> >    include more fields to be matched against. The release 2.2 does not
> >    contain these ABI changes, but release 2.3 will.
> 
> These are the structures changed in this patch.
> I think this section must be also removed.
> 


This deprecation notice is not raised by me. It is raised by rahul.lakkireddy@chelsio.com at commit 954f1545a1ab.
So, I'm not sure if it is OK for me to remove it in my patch.

/Jingjing

> > --- a/doc/guides/rel_notes/release_16_04.rst
> > +++ b/doc/guides/rel_notes/release_16_04.rst
> > @@ -435,6 +435,8 @@ ABI Changes
> >
> >  * The cmdline buffer size has been increase from 256 to 512.
> >
> > +* The ethdev flow director structure ``rte_eth_fdir_flow`` structure
> > +was
> > +  changed. New fields were added to extend flow director's input set.
> 
> For reading ease, it's better to group ethdev changes (before cmdline
> change).

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] DPDK and HW offloads
  2016-03-22 10:19  0%             ` Bruce Richardson
@ 2016-03-23  2:47  0%               ` Qiu, Michael
  0 siblings, 0 replies; 200+ results
From: Qiu, Michael @ 2016-03-23  2:47 UTC (permalink / raw)
  To: Richardson, Bruce
  Cc: Kyle Larose, Thomas Monjalon, Zhang, Helin, Stephen Hemminger, dev

On 3/22/2016 6:20 PM, Richardson, Bruce wrote:
> On Tue, Mar 22, 2016 at 05:50:28AM +0000, Qiu, Michael wrote:
>> On 3/21/2016 11:27 PM, Kyle Larose wrote:
>>> On Mon, Mar 21, 2016 at 10:52 AM, Bruce Richardson
>>> <bruce.richardson@intel.com> wrote:
>>>> On Sun, Mar 20, 2016 at 08:18:57PM +0100, Thomas Monjalon wrote:
>>>>> 2016-03-20 14:17, Zhang, Helin:
>>>>>> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
>>>>>>> 2016-03-18 10:16, Stephen Hemminger:
>>>>>>>> Right now, all those offload features are pretty much unusable in a
>>>>>>>> real product without lots and lots of extra codes and huge bug
>>>>>>>> surface. It bothers me enough that I would recommend removing much of the
>>>>>>> filter/offload/ptype stuff from DPDK!
>>>>>>>
>>>>>>> One of the biggest challenge is to think about a good filtering API.
>>>>>>> The offloading has some interaction with the mbuf struct.
>>>>>>>
>>>>>>> I would like to suggest rewriting ethdev API by keeping it as is for some time for
>>>>>>> compatibility while creating a new one. What about the prefix dpdk_netdev_ to
>>>>>>> progressively replace rte_eth_dev?
>>>>>> I totally agree with to add new and generic APIs for user applications. But I don't
>>>>>> think we need to remove all current APIs. Generic APIs may not support all advanced
>>>>>> hardware features, while specific APIs can. Why not support all? One generic APIs for
>>>>>> common users, and others APIs for advanced users.
>>>>> Yes we cannot access to every features of a device through generic API.
>>>>> Until now we were trying to add an ethdev API for every features even if it
>>>>> is used by only one driver.
>>>>> I think we should allow a direct access to the driver by the applications and
>>>>> work on generic API only for common features.
>>>> Definite +1.
>>>> I think that we need to start pushing driver-specific functionality to get exposed
>>>> via a driver's header files. That allow users who want to extract the max
>>>> functionality from a particular NIC to do so via those APIs calls, while not
>>>> polluting the generic ethdev layer.
>>>>
>>> What sort of requirements on ABI/API compatibility would this place on
>>> the drivers? I would hope that it would be treated like any other
>>> public API within DPDK. I don't think this would be too onerous, but
>>> it would require that the drivers be designed to deal with it. (I.e.
>>> don't just expose any old internal driver function).
>> Why not to implement one simple API with variable arguments, just like
>> syscall ioctl() does. And drivers implement it's specific hardware
>> features with a feature bit param, and other needed variable arguments.
>>
>> Thanks,
>> Michael
> A very much dislike that idea. 
> * It makes the code much harder to read as you have to closely examine all the
>   parameters to work out what a function call is actually meant to do.

It's not a big deal, if we have a document.

> * It makes it much harder to see that you have an implicit dependency on a
>   specific device. Having to include a driver specific header file e.g. i40e.h,
>   and call a function named e.g. i40e_do_magic_stuff(), makes it pretty explicit
>   that you have a dependency on i40e-based hardware

Software does not want to bind to specific hardware I think, what about
the transportability?

> * It prevents the compiler from doing type-checking on parameters and informing
>   you of little inconsistencies.

Maybe, we could do self-check for the parameters I think.

>
> For all these reasons, I prefer the device-specific functions option. However,
> at the same time, we also need to ensure we have a reasonable set of generic
> APIs so that the cases where users are forced to drop down to the lower-level
> device-specific primitives are reduced.

For software, it do not care which hardware it is, it only cares about
what ability you have.

Thanks,
Michael

> Regards,
> /Bruce
>
>>>> On the other hand, I don't like the idea of dpdk_netdev. I think we can work
>>>> within the existing rte_eth_dev framework.
>>>>
>>>> /Bruce
>>>>
>>


^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection
  2016-03-23  0:42  0%       ` Wu, Jingjing
@ 2016-03-23  8:45  0%         ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-03-23  8:45 UTC (permalink / raw)
  To: Wu, Jingjing; +Cc: dev, Zhang, Helin

2016-03-23 00:42, Wu, Jingjing:
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > 2016-03-21 14:18, Jingjing Wu:
> > > This patch added RTE_ETH_INPUT_SET_L3_IP4_TTL,
> > > RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS input field type and extended
> > > struct rte_eth_ipv4_flow and rte_eth_ipv6_flow to support filtering by
> > > tos, protocol and ttl.
> > [...]
> > > --- a/doc/guides/rel_notes/deprecation.rst
> > > +++ b/doc/guides/rel_notes/deprecation.rst
> > > @@ -22,10 +22,6 @@ Deprecation Notices
> > >  * The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
> > >    must be updated to support 100G link and to have a cleaner link speed
> > API.
> > >
> > > -* ABI changes are planned for struct rte_eth_fdir_flow in order to
> > > support
> > > -  extend flow director's input set. The release 2.2 does not contain
> > > these ABI
> > > -  changes, but release 2.3 will, and no backwards compatibility is planned.
> > 
> > The changed structures are part of rte_eth_fdir_flow.
> > So this deprecation notice apply to this patch.
> > 
> > >  * ABI changes are planned for rte_eth_ipv4_flow and rte_eth_ipv6_flow
> > to
> > >    include more fields to be matched against. The release 2.2 does not
> > >    contain these ABI changes, but release 2.3 will.
> > 
> > These are the structures changed in this patch.
> > I think this section must be also removed.
> 
> This deprecation notice is not raised by me. It is raised by rahul.lakkireddy@chelsio.com at commit 954f1545a1ab.
> So, I'm not sure if it is OK for me to remove it in my patch.

It does not matter who raised the notice.
Your patch match with it. Please remove it.

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver
  2016-03-21  6:18  4% ` [dpdk-dev] [PATCH v5 0/9] " Jingjing Wu
  2016-03-21  6:18 18%   ` [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection Jingjing Wu
@ 2016-03-23 13:07  4%   ` Jingjing Wu
  2016-03-23 13:07 19%     ` [dpdk-dev] [PATCH v6 1/9] ethdev: extend flow director for input selection Jingjing Wu
  2016-03-23 14:46  0%     ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Bruce Richardson
  1 sibling, 2 replies; 200+ results
From: Jingjing Wu @ 2016-03-23 13:07 UTC (permalink / raw)
  To: bruce.richardson; +Cc: dev, jingjing.wu, helin.zhang

v6 changes:
 - remove ABI announce on rte_eth_ipv4_flow and rte_eth_ipv6_flow in deprecation.
 - reword commit logs. 

v5 changes:
 - remove the reorganizing of struct rte_eth_fdir_flow
 - remove fdir supporting on Tunnel Id
 - rebase to latest dpdk/master

v4 changes:
 - rebase to latest dpdk-next-net/rel_16_04.
 - comments on new fields in API structure.

v3 changes:
 - rebase to latest dpdk-next-net/rel_16_04(commit: 0f9564a0e4f2)
 - use AQ rx control register read/write for some registers
 - remove few useless lines
 - patch title rewording

v2 changes:
 - rebase on dpdk-next-net/rel_16_04
 - comments rewording.
 - redefine the value of RTE_ETH_INPUT_SET_L3_IP4_TTL to
   avoid ABI breaking.
 - remove ABI announce in Deprecation.
 - fix the ethertype setting when program filter in v1 patch set.

This patch set extends flow director to support filtering by
additional fields below in i40e driver:
 - TOS, Protocol and TTL in IP header
 - single vlan or inner vlan


Andrey Chilikin (1):
  i40e: fix VLAN bitmasks for input set

Jingjing Wu (8):
  ethdev: extend flow director for input selection
  i40e: split function for hash and fdir input
  i40e: remove flex payload from input selection
  i40e: restore default setting on input set
  i40e: extend flow director to filter by IP Header
  testpmd: extend input set related commands
  i40e: extend flow director to filter by vlan id
  testpmd: extend flow director commands

 app/test-pmd/cmdline.c                      | 100 +++++--
 doc/guides/rel_notes/deprecation.rst        |   8 -
 doc/guides/rel_notes/release_16_04.rst      |   5 +
 doc/guides/testpmd_app_ug/testpmd_funcs.rst |  42 +--
 drivers/net/i40e/i40e_ethdev.c              | 393 ++++++++++++++++------------
 drivers/net/i40e/i40e_ethdev.h              |  11 +-
 drivers/net/i40e/i40e_fdir.c                | 100 ++++---
 lib/librte_ether/rte_eth_ctrl.h             |   8 +
 8 files changed, 417 insertions(+), 250 deletions(-)

-- 
2.4.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v6 1/9] ethdev: extend flow director for input selection
  2016-03-23 13:07  4%   ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Jingjing Wu
@ 2016-03-23 13:07 19%     ` Jingjing Wu
  2016-03-23 14:46  0%     ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Bruce Richardson
  1 sibling, 0 replies; 200+ results
From: Jingjing Wu @ 2016-03-23 13:07 UTC (permalink / raw)
  To: bruce.richardson; +Cc: dev, jingjing.wu, helin.zhang

This patch adds RTE_ETH_INPUT_SET_L3_IP4_TTL,
RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS input field types and extends
struct rte_eth_ipv4_flow and rte_eth_ipv6_flow to support filtering
by tos, protocol and ttl.

Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
Acked-by: Helin Zhang <helin.zhang@intel.com>
---
 doc/guides/rel_notes/deprecation.rst   | 8 --------
 doc/guides/rel_notes/release_16_04.rst | 3 +++
 lib/librte_ether/rte_eth_ctrl.h        | 8 ++++++++
 3 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 252a096..bdbac15 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -22,14 +22,6 @@ Deprecation Notices
 * The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
   must be updated to support 100G link and to have a cleaner link speed API.
 
-* ABI changes are planned for struct rte_eth_fdir_flow in order to support
-  extend flow director's input set. The release 2.2 does not contain these ABI
-  changes, but release 2.3 will, and no backwards compatibility is planned.
-
-* ABI changes are planned for rte_eth_ipv4_flow and rte_eth_ipv6_flow to
-  include more fields to be matched against. The release 2.2 does not
-  contain these ABI changes, but release 2.3 will.
-
 * 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.
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 9922bcb..6e368f7 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -437,6 +437,9 @@ ABI Changes
 * The RETA entry size in ``rte_eth_rss_reta_entry64`` has been increased
   from 8-bit to 16-bit.
 
+* The ethdev flow director structure ``rte_eth_fdir_flow`` structure was
+  changed. New fields were added to extend flow director's input set.
+
 * The cmdline buffer size has been increase from 256 to 512.
 
 
diff --git a/lib/librte_ether/rte_eth_ctrl.h b/lib/librte_ether/rte_eth_ctrl.h
index 6e2f617..aabd724 100644
--- a/lib/librte_ether/rte_eth_ctrl.h
+++ b/lib/librte_ether/rte_eth_ctrl.h
@@ -343,6 +343,8 @@ enum rte_eth_input_set_field {
 	RTE_ETH_INPUT_SET_L3_IP4_PROTO,
 	RTE_ETH_INPUT_SET_L3_IP6_TC,
 	RTE_ETH_INPUT_SET_L3_IP6_NEXT_HEADER,
+	RTE_ETH_INPUT_SET_L3_IP4_TTL,
+	RTE_ETH_INPUT_SET_L3_IP6_HOP_LIMITS,
 
 	/* L4 */
 	RTE_ETH_INPUT_SET_L4_UDP_SRC_PORT = 257,
@@ -410,6 +412,9 @@ struct rte_eth_l2_flow {
 struct rte_eth_ipv4_flow {
 	uint32_t src_ip;      /**< IPv4 source address to match. */
 	uint32_t dst_ip;      /**< IPv4 destination address to match. */
+	uint8_t  tos;         /**< Type of service to match. */
+	uint8_t  ttl;         /**< Time to live to match. */
+	uint8_t  proto;       /**< Protocol, next header to match. */
 };
 
 /**
@@ -446,6 +451,9 @@ struct rte_eth_sctpv4_flow {
 struct rte_eth_ipv6_flow {
 	uint32_t src_ip[4];      /**< IPv6 source address to match. */
 	uint32_t dst_ip[4];      /**< IPv6 destination address to match. */
+	uint8_t  tc;             /**< Traffic class to match. */
+	uint8_t  proto;          /**< Protocol, next header to match. */
+	uint8_t  hop_limits;     /**< Hop limits to match. */
 };
 
 /**
-- 
2.4.0

^ permalink raw reply	[relevance 19%]

* Re: [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver
  2016-03-23 13:07  4%   ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Jingjing Wu
  2016-03-23 13:07 19%     ` [dpdk-dev] [PATCH v6 1/9] ethdev: extend flow director for input selection Jingjing Wu
@ 2016-03-23 14:46  0%     ` Bruce Richardson
  1 sibling, 0 replies; 200+ results
From: Bruce Richardson @ 2016-03-23 14:46 UTC (permalink / raw)
  To: Jingjing Wu; +Cc: dev, helin.zhang

On Wed, Mar 23, 2016 at 09:07:03PM +0800, Jingjing Wu wrote:
> v6 changes:
>  - remove ABI announce on rte_eth_ipv4_flow and rte_eth_ipv6_flow in deprecation.
>  - reword commit logs. 
> 
> v5 changes:
>  - remove the reorganizing of struct rte_eth_fdir_flow
>  - remove fdir supporting on Tunnel Id
>  - rebase to latest dpdk/master
> 
> v4 changes:
>  - rebase to latest dpdk-next-net/rel_16_04.
>  - comments on new fields in API structure.
> 
> v3 changes:
>  - rebase to latest dpdk-next-net/rel_16_04(commit: 0f9564a0e4f2)
>  - use AQ rx control register read/write for some registers
>  - remove few useless lines
>  - patch title rewording
> 
> v2 changes:
>  - rebase on dpdk-next-net/rel_16_04
>  - comments rewording.
>  - redefine the value of RTE_ETH_INPUT_SET_L3_IP4_TTL to
>    avoid ABI breaking.
>  - remove ABI announce in Deprecation.
>  - fix the ethertype setting when program filter in v1 patch set.
> 
> This patch set extends flow director to support filtering by
> additional fields below in i40e driver:
>  - TOS, Protocol and TTL in IP header
>  - single vlan or inner vlan
> 
> 
> Andrey Chilikin (1):
>   i40e: fix VLAN bitmasks for input set
> 
> Jingjing Wu (8):
>   ethdev: extend flow director for input selection
>   i40e: split function for hash and fdir input
>   i40e: remove flex payload from input selection
>   i40e: restore default setting on input set
>   i40e: extend flow director to filter by IP Header
>   testpmd: extend input set related commands
>   i40e: extend flow director to filter by vlan id
>   testpmd: extend flow director commands
>
Patchset applied to dpdk-next-net/rel_16_04.

Thanks,
/Bruce

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] mempool: allow for user-owned mempool caches
  2016-03-21 13:49  3%   ` Wiles, Keith
@ 2016-03-24 14:35  0%     ` Lazaros Koromilas
  2016-03-24 14:58  0%       ` Venkatesan, Venky
  2016-03-24 15:03  0%       ` Wiles, Keith
  0 siblings, 2 replies; 200+ results
From: Lazaros Koromilas @ 2016-03-24 14:35 UTC (permalink / raw)
  To: Wiles, Keith; +Cc: Olivier Matz, dev

On Mon, Mar 21, 2016 at 3:49 PM, Wiles, Keith <keith.wiles@intel.com> wrote:
>>Hi Lazaros,
>>
>>Thanks for this patch. To me, this is a valuable enhancement.
>>Please find some comments inline.
>>
>>On 03/10/2016 03:44 PM, Lazaros Koromilas wrote:
>>> The mempool cache is only available to EAL threads as a per-lcore
>>> resource. Change this so that the user can create and provide their own
>>> cache on mempool get and put operations. This works with non-EAL threads
>>> too. This commit introduces new API calls with the 'with_cache' suffix,
>>> while the current ones default to the per-lcore local cache.
>>>
>>> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
>>> ---
>>>  lib/librte_mempool/rte_mempool.c |  65 +++++-
>>>  lib/librte_mempool/rte_mempool.h | 442 ++++++++++++++++++++++++++++++++++++---
>>>  2 files changed, 467 insertions(+), 40 deletions(-)
>>>
>>> diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
>>> index f8781e1..cebc2b7 100644
>>> --- a/lib/librte_mempool/rte_mempool.c
>>> +++ b/lib/librte_mempool/rte_mempool.c
>>> @@ -375,6 +375,43 @@ rte_mempool_xmem_usage(void *vaddr, uint32_t elt_num, size_t elt_sz,
>>>      return usz;
>>>  }
>>>
>>> +#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>>
>>I wonder if this wouldn't cause a conflict with Keith's patch
>>that removes some #ifdefs RTE_MEMPOOL_CACHE_MAX_SIZE.
>>See: http://www.dpdk.org/dev/patchwork/patch/10492/
>
> Hi Lazaros,
>
> The patch I submitted keeps the mempool cache structure (pointers and variables) and only allocates the cache if specified by the caller to use a cache. This means to me the caller could fill in the cache pointer and values into the mempool structure to get a cache without a lot of extra code. If we added a set of APIs to fill in these structure variables would that not give you the external cache support. I have not really looked at the patch to verify this will work, but it sure seems like it.
>
> So my suggestion the caller can just create a mempool without a cache and then call a set of APIs to fill in his cache values, does that not work?
>
> If we can do this it reduces the API and possible the ABI changes to mempool as the new cache create routines and APIs could be in a new file I think, which just updates the mempool structure correctly.

Hi Keith,

The main benefit of having an external cache is to allow mempool users
(threads) to maintain a local cache even though they don't have a
valid lcore_id (non-EAL threads). The fact that cache access is done
by indexing with the lcore_id is what makes it difficult...

What could happen is only have external caches somehow, but that hurts
the common case where you want an automatic cache.
Or a cache registration mechanism (overkill?).

So, I'm going to work on the comments and send out a v2 asap. Thanks everyone!

Lazaros.

>
>>
>>As this patch is already acked for 16.07, I think that your v2
>>could be rebased on top of it to avoid conflicts when Thomas will apply
>>it.
>>
>>By the way, I also encourage you to have a look at other works in
>>progress in mempool:
>>http://www.dpdk.org/ml/archives/dev/2016-March/035107.html
>>http://www.dpdk.org/ml/archives/dev/2016-March/035201.html
>>
>>
>
> Regards,
> Keith
>
>
>
>

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] mempool: allow for user-owned mempool caches
  2016-03-24 14:35  0%     ` Lazaros Koromilas
@ 2016-03-24 14:58  0%       ` Venkatesan, Venky
  2016-03-24 15:03  0%       ` Wiles, Keith
  1 sibling, 0 replies; 200+ results
From: Venkatesan, Venky @ 2016-03-24 14:58 UTC (permalink / raw)
  To: Lazaros Koromilas, Wiles, Keith; +Cc: Olivier Matz, dev



> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Lazaros Koromilas
> Sent: Thursday, March 24, 2016 7:36 AM
> To: Wiles, Keith <keith.wiles@intel.com>
> Cc: Olivier Matz <olivier.matz@6wind.com>; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] mempool: allow for user-owned mempool
> caches
> 
> On Mon, Mar 21, 2016 at 3:49 PM, Wiles, Keith <keith.wiles@intel.com>
> wrote:
> >>Hi Lazaros,
> >>
> >>Thanks for this patch. To me, this is a valuable enhancement.
> >>Please find some comments inline.
> >>
> >>On 03/10/2016 03:44 PM, Lazaros Koromilas wrote:
> >>> The mempool cache is only available to EAL threads as a per-lcore
> >>> resource. Change this so that the user can create and provide their
> >>> own cache on mempool get and put operations. This works with non-EAL
> >>> threads too. This commit introduces new API calls with the
> >>> 'with_cache' suffix, while the current ones default to the per-lcore local
> cache.
> >>>
> >>> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
> >>> ---
> >>>  lib/librte_mempool/rte_mempool.c |  65 +++++-
> >>> lib/librte_mempool/rte_mempool.h | 442
> >>> ++++++++++++++++++++++++++++++++++++---
> >>>  2 files changed, 467 insertions(+), 40 deletions(-)
> >>>
> >>> diff --git a/lib/librte_mempool/rte_mempool.c
> >>> b/lib/librte_mempool/rte_mempool.c
> >>> index f8781e1..cebc2b7 100644
> >>> --- a/lib/librte_mempool/rte_mempool.c
> >>> +++ b/lib/librte_mempool/rte_mempool.c
> >>> @@ -375,6 +375,43 @@ rte_mempool_xmem_usage(void *vaddr,
> uint32_t elt_num, size_t elt_sz,
> >>>      return usz;
> >>>  }
> >>>
> >>> +#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> >>
> >>I wonder if this wouldn't cause a conflict with Keith's patch that
> >>removes some #ifdefs RTE_MEMPOOL_CACHE_MAX_SIZE.
> >>See: http://www.dpdk.org/dev/patchwork/patch/10492/
> >
> > Hi Lazaros,
> >
> > The patch I submitted keeps the mempool cache structure (pointers and
> variables) and only allocates the cache if specified by the caller to use a
> cache. This means to me the caller could fill in the cache pointer and values
> into the mempool structure to get a cache without a lot of extra code. If we
> added a set of APIs to fill in these structure variables would that not give you
> the external cache support. I have not really looked at the patch to verify this
> will work, but it sure seems like it.
> >
> > So my suggestion the caller can just create a mempool without a cache and
> then call a set of APIs to fill in his cache values, does that not work?
> >
> > If we can do this it reduces the API and possible the ABI changes to
> mempool as the new cache create routines and APIs could be in a new file I
> think, which just updates the mempool structure correctly.
> 
> Hi Keith,
> 
> The main benefit of having an external cache is to allow mempool users
> (threads) to maintain a local cache even though they don't have a valid
> lcore_id (non-EAL threads). The fact that cache access is done by indexing
> with the lcore_id is what makes it difficult...

Hi Lazaros, 

Alternative suggestion: This could actually be very simply done via creating an EAL API to register and return an lcore_id for a thread wanting to use DPDK services. That way, you could simply create your pthread, call the eal_register_thread() function that assigns an lcore_id to the caller (and internally sets up the per_lcore variable. 

The advantage of doing it this way is that you could extend it to other things other than the mempool that may need an lcore_id setup.

Regards,
-Venky

> 
> What could happen is only have external caches somehow, but that hurts the
> common case where you want an automatic cache.
> Or a cache registration mechanism (overkill?).
> 
> So, I'm going to work on the comments and send out a v2 asap. Thanks
> everyone!
> 
> Lazaros.
> 
> >
> >>
> >>As this patch is already acked for 16.07, I think that your v2 could
> >>be rebased on top of it to avoid conflicts when Thomas will apply it.
> >>
> >>By the way, I also encourage you to have a look at other works in
> >>progress in mempool:
> >>http://www.dpdk.org/ml/archives/dev/2016-March/035107.html
> >>http://www.dpdk.org/ml/archives/dev/2016-March/035201.html
> >>
> >>
> >
> > Regards,
> > Keith
> >
> >
> >
> >

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] mempool: allow for user-owned mempool caches
  2016-03-24 14:35  0%     ` Lazaros Koromilas
  2016-03-24 14:58  0%       ` Venkatesan, Venky
@ 2016-03-24 15:03  0%       ` Wiles, Keith
  1 sibling, 0 replies; 200+ results
From: Wiles, Keith @ 2016-03-24 15:03 UTC (permalink / raw)
  To: Lazaros Koromilas; +Cc: Olivier Matz, dev

>On Mon, Mar 21, 2016 at 3:49 PM, Wiles, Keith <keith.wiles@intel.com> wrote:
>>>Hi Lazaros,
>>>
>>>Thanks for this patch. To me, this is a valuable enhancement.
>>>Please find some comments inline.
>>>
>>>On 03/10/2016 03:44 PM, Lazaros Koromilas wrote:
>>>> The mempool cache is only available to EAL threads as a per-lcore
>>>> resource. Change this so that the user can create and provide their own
>>>> cache on mempool get and put operations. This works with non-EAL threads
>>>> too. This commit introduces new API calls with the 'with_cache' suffix,
>>>> while the current ones default to the per-lcore local cache.
>>>>
>>>> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
>>>> ---
>>>>  lib/librte_mempool/rte_mempool.c |  65 +++++-
>>>>  lib/librte_mempool/rte_mempool.h | 442 ++++++++++++++++++++++++++++++++++++---
>>>>  2 files changed, 467 insertions(+), 40 deletions(-)
>>>>
>>>> diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
>>>> index f8781e1..cebc2b7 100644
>>>> --- a/lib/librte_mempool/rte_mempool.c
>>>> +++ b/lib/librte_mempool/rte_mempool.c
>>>> @@ -375,6 +375,43 @@ rte_mempool_xmem_usage(void *vaddr, uint32_t elt_num, size_t elt_sz,
>>>>      return usz;
>>>>  }
>>>>
>>>> +#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>>>
>>>I wonder if this wouldn't cause a conflict with Keith's patch
>>>that removes some #ifdefs RTE_MEMPOOL_CACHE_MAX_SIZE.
>>>See: http://www.dpdk.org/dev/patchwork/patch/10492/
>>
>> Hi Lazaros,
>>
>> The patch I submitted keeps the mempool cache structure (pointers and variables) and only allocates the cache if specified by the caller to use a cache. This means to me the caller could fill in the cache pointer and values into the mempool structure to get a cache without a lot of extra code. If we added a set of APIs to fill in these structure variables would that not give you the external cache support. I have not really looked at the patch to verify this will work, but it sure seems like it.
>>
>> So my suggestion the caller can just create a mempool without a cache and then call a set of APIs to fill in his cache values, does that not work?
>>
>> If we can do this it reduces the API and possible the ABI changes to mempool as the new cache create routines and APIs could be in a new file I think, which just updates the mempool structure correctly.
>
>Hi Keith,
>
>The main benefit of having an external cache is to allow mempool users
>(threads) to maintain a local cache even though they don't have a
>valid lcore_id (non-EAL threads). The fact that cache access is done
>by indexing with the lcore_id is what makes it difficult...
>
>What could happen is only have external caches somehow, but that hurts
>the common case where you want an automatic cache.
>Or a cache registration mechanism (overkill?).
>
>So, I'm going to work on the comments and send out a v2 asap. Thanks everyone!

Hi Lazaros,

I look forward seeing the next version as I have some concerns, but I do not think I am seeing the big picture or reason for the change.

Part of my goal with the patch I sent was to remove ifdefs in the code as much as possible, so please try to reduce the ifdefs too.

>
>Lazaros.
>
>>
>>>
>>>As this patch is already acked for 16.07, I think that your v2
>>>could be rebased on top of it to avoid conflicts when Thomas will apply
>>>it.
>>>
>>>By the way, I also encourage you to have a look at other works in
>>>progress in mempool:
>>>http://www.dpdk.org/ml/archives/dev/2016-March/035107.html
>>>http://www.dpdk.org/ml/archives/dev/2016-March/035201.html
>>>
>>>
>>
>> Regards,
>> Keith
>>
>>
>>
>>
>


Regards,
Keith





^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH] doc: postpone flow director changes planned for cxgbe
@ 2016-03-25 16:29  5% Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-03-25 16:29 UTC (permalink / raw)
  To: rahul.lakkireddy; +Cc: dev

It will be tried to find a better solution.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
 doc/guides/rel_notes/deprecation.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index bdbac15..179e30f 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -24,4 +24,4 @@ Deprecation Notices
 
 * 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.
+  but release 2.3 will. [postponed]
-- 
2.7.0

^ permalink raw reply	[relevance 5%]

* [dpdk-dev] [PATCH v12 6/8] ethdev: redesign link speed config
  @ 2016-03-25 19:42  3%   ` Thomas Monjalon
    1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-03-25 19:42 UTC (permalink / raw)
  To: marcdevel, bruce.richardson, declan.doherty, konstantin.ananyev,
	wenzhuo.lu, helin.zhang, jing.d.chen, harish.patil,
	rahul.lakkireddy, johndale, vido, adrien.mazarguil,
	alejandro.lucero
  Cc: dev

From: Marc Sune <marcdevel@gmail.com>

This patch redesigns the API to set the link speed/s configuration
of an ethernet port. Specifically:

- it allows to define a set of advertised speeds for
  auto-negociation.
- it allows to disable link auto-negociation (single fixed speed).
- default: auto-negociate all supported speeds.

A flag autoneg in struct rte_eth_link indicates if link speed was a
result of auto-negociation or was fixed by configuration.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Tested-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---

PLEASE REVIEW CAREFULLY THIS PATCH

 app/test-pmd/cmdline.c                    | 26 ++++----
 doc/guides/rel_notes/deprecation.rst      |  3 -
 doc/guides/rel_notes/release_16_04.rst    |  9 +++
 drivers/net/af_packet/rte_eth_af_packet.c |  1 +
 drivers/net/bnx2x/bnx2x_ethdev.c          |  4 +-
 drivers/net/bonding/rte_eth_bond_8023ad.c |  2 +-
 drivers/net/e1000/em_ethdev.c             | 99 +++++++++++++++----------------
 drivers/net/e1000/igb_ethdev.c            | 94 +++++++++++++++--------------
 drivers/net/i40e/i40e_ethdev.c            | 48 +++++++--------
 drivers/net/i40e/i40e_ethdev_vf.c         |  7 ++-
 drivers/net/ixgbe/ixgbe_ethdev.c          | 46 ++++++--------
 drivers/net/mlx4/mlx4.c                   |  2 +
 drivers/net/mpipe/mpipe_tilegx.c          |  2 +
 drivers/net/null/rte_eth_null.c           |  1 +
 drivers/net/pcap/rte_eth_pcap.c           |  1 +
 drivers/net/ring/rte_eth_ring.c           |  1 +
 drivers/net/szedata2/rte_eth_szedata2.c   |  2 +
 drivers/net/vmxnet3/vmxnet3_ethdev.c      |  1 +
 drivers/net/xenvirt/rte_eth_xenvirt.c     |  1 +
 examples/ip_pipeline/config_parse.c       |  3 +-
 lib/librte_ether/rte_ethdev.h             | 29 +++++----
 21 files changed, 196 insertions(+), 186 deletions(-)

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 815b53b..741cac3 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -989,7 +989,7 @@ struct cmd_config_speed_all {
 };
 
 static int
-parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
+parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint32_t *speed)
 {
 
 	int duplex;
@@ -1006,20 +1006,22 @@ parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
 	}
 
 	if (!strcmp(speedstr, "10")) {
-		*speed = ETH_SPEED_NUM_10M;
+		*speed = (duplex == ETH_LINK_HALF_DUPLEX) ?
+				ETH_LINK_SPEED_10M_HD : ETH_LINK_SPEED_10M;
 	} else if (!strcmp(speedstr, "100")) {
-		*speed = ETH_SPEED_NUM_100M;
+		*speed = (duplex == ETH_LINK_HALF_DUPLEX) ?
+				ETH_LINK_SPEED_100M_HD : ETH_LINK_SPEED_100M;
 	} else {
 		if (duplex != ETH_LINK_FULL_DUPLEX) {
 			printf("Invalid speed/duplex parameters\n");
 			return -1;
 		}
 		if (!strcmp(speedstr, "1000")) {
-			*speed = ETH_SPEED_NUM_1G;
+			*speed = ETH_LINK_SPEED_1G;
 		} else if (!strcmp(speedstr, "10000")) {
-			*speed = ETH_SPEED_NUM_10G;
+			*speed = ETH_LINK_SPEED_10G;
 		} else if (!strcmp(speedstr, "40000")) {
-			*speed = ETH_SPEED_NUM_40G;
+			*speed = ETH_LINK_SPEED_40G;
 		} else if (!strcmp(speedstr, "auto")) {
 			*speed = ETH_LINK_SPEED_AUTONEG;
 		} else {
@@ -1037,8 +1039,7 @@ cmd_config_speed_all_parsed(void *parsed_result,
 			__attribute__((unused)) void *data)
 {
 	struct cmd_config_speed_all *res = parsed_result;
-	uint16_t link_speed = ETH_LINK_SPEED_AUTONEG;
-	uint16_t link_duplex = 0;
+	uint32_t link_speed;
 	portid_t pid;
 
 	if (!all_ports_stopped()) {
@@ -1051,8 +1052,7 @@ cmd_config_speed_all_parsed(void *parsed_result,
 		return;
 
 	FOREACH_PORT(pid, ports) {
-		ports[pid].dev_conf.link_speed = link_speed;
-		ports[pid].dev_conf.link_duplex = link_duplex;
+		ports[pid].dev_conf.link_speeds = link_speed;
 	}
 
 	cmd_reconfig_device_queue(RTE_PORT_ALL, 1, 1);
@@ -1110,8 +1110,7 @@ cmd_config_speed_specific_parsed(void *parsed_result,
 				__attribute__((unused)) void *data)
 {
 	struct cmd_config_speed_specific *res = parsed_result;
-	uint16_t link_speed = ETH_LINK_SPEED_AUTONEG;
-	uint16_t link_duplex = 0;
+	uint32_t link_speed;
 
 	if (!all_ports_stopped()) {
 		printf("Please stop all ports first\n");
@@ -1125,8 +1124,7 @@ cmd_config_speed_specific_parsed(void *parsed_result,
 			&link_speed) < 0)
 		return;
 
-	ports[res->id].dev_conf.link_speed = link_speed;
-	ports[res->id].dev_conf.link_duplex = link_duplex;
+	ports[res->id].dev_conf.link_speeds = link_speed;
 
 	cmd_reconfig_device_queue(RTE_PORT_ALL, 1, 1);
 }
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 179e30f..c47610d 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -19,9 +19,6 @@ Deprecation Notices
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff
 
-* The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
-  must be updated to support 100G link and to have a cleaner link speed API.
-
 * 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]
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 9e7b0b7..c891a55 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -52,6 +52,12 @@ This section should contain new features added in this release. Sample format:
   The structure ``rte_eth_dev_info`` has now a ``speed_capa`` bitmap, which
   allows the application to know the supported speeds of each device.
 
+* **Added bitmap of link speeds to advertise.**
+
+  Allow defining a set of advertised speeds for auto-negotiation,
+  explicitly disabling link auto-negotiation (single speed)
+  and full auto-negotiation.
+
 * **Added new poll-mode driver for Amazon Elastic Network Adapters (ENA).**
 
   The driver operates variety of ENA adapters through feature negotiation
@@ -464,6 +470,9 @@ This section should contain API changes. Sample format:
 * The ethdev structure ``rte_eth_dev_info`` was changed to support device
   speed capabilities.
 
+* The ethdev structures ``rte_eth_link`` and ``rte_eth_conf`` were changed to
+  support the new link API.
+
 * The functions ``rte_eth_dev_udp_tunnel_add`` and ``rte_eth_dev_udp_tunnel_delete``
   have been renamed into ``rte_eth_dev_udp_tunnel_port_add`` and
   ``rte_eth_dev_udp_tunnel_port_delete``.
diff --git a/drivers/net/af_packet/rte_eth_af_packet.c b/drivers/net/af_packet/rte_eth_af_packet.c
index 641f849..f17bd7e 100644
--- a/drivers/net/af_packet/rte_eth_af_packet.c
+++ b/drivers/net/af_packet/rte_eth_af_packet.c
@@ -119,6 +119,7 @@ static struct rte_eth_link pmd_link = {
 	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
+	.link_autoneg = ETH_LINK_SPEED_AUTONEG
 };
 
 static uint16_t
diff --git a/drivers/net/bnx2x/bnx2x_ethdev.c b/drivers/net/bnx2x/bnx2x_ethdev.c
index 897081f..af84175 100644
--- a/drivers/net/bnx2x/bnx2x_ethdev.c
+++ b/drivers/net/bnx2x/bnx2x_ethdev.c
@@ -44,9 +44,9 @@ bnx2x_link_update(struct rte_eth_dev *dev)
 		case DUPLEX_HALF:
 			dev->data->dev_link.link_duplex = ETH_LINK_HALF_DUPLEX;
 			break;
-		default:
-			dev->data->dev_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
 	}
+	dev->data->dev_link.link_autoneg = (dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_AUTONEG);
 	dev->data->dev_link.link_status = sc->link_vars.link_up;
 }
 
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index ac8306f..cca7cc3 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -708,7 +708,7 @@ link_speed_key(uint16_t speed) {
 	uint16_t key_speed;
 
 	switch (speed) {
-	case ETH_LINK_SPEED_AUTONEG:
+	case ETH_SPEED_NUM_NONE:
 		key_speed = 0x00;
 		break;
 	case ETH_SPEED_NUM_10M:
diff --git a/drivers/net/e1000/em_ethdev.c b/drivers/net/e1000/em_ethdev.c
index d5f8c7f..3e26ab0 100644
--- a/drivers/net/e1000/em_ethdev.c
+++ b/drivers/net/e1000/em_ethdev.c
@@ -538,6 +538,9 @@ eth_em_start(struct rte_eth_dev *dev)
 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
 	int ret, mask;
 	uint32_t intr_vector = 0;
+	uint32_t *speeds;
+	int num_speeds;
+	bool autoneg;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -612,56 +615,46 @@ eth_em_start(struct rte_eth_dev *dev)
 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
 
 	/* Setup link speed and duplex */
-	switch (dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_HALF_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_FULL_DUPLEX;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_HALF;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_FULL;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_100M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_HALF;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_FULL;
-		else
+	speeds = &dev->data->dev_conf.link_speeds;
+	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
+	} else {
+		num_speeds = 0;
+		autoneg = (*speeds & ETH_LINK_SPEED_AUTONEG);
+
+		/* Reset */
+		hw->phy.autoneg_advertised = 0;
+
+		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
+				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G)) {
+			num_speeds = -1;
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_1G:
-		if ((dev->data->dev_conf.link_duplex ==
-				ETH_LINK_AUTONEG_DUPLEX) ||
-			(dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX))
-			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
-		else
+		}
+		if (*speeds & ETH_LINK_SPEED_10M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_10M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_1G) {
+			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
+			num_speeds++;
+		}
+		if (num_speeds == 0 || (!autoneg && (num_speeds > 2)))
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10G:
-	default:
-		goto error_invalid_config;
 	}
+
 	e1000_setup_link(hw);
 
 	if (rte_intr_allow_others(intr_handle)) {
@@ -694,9 +687,8 @@ eth_em_start(struct rte_eth_dev *dev)
 	return 0;
 
 error_invalid_config:
-	PMD_INIT_LOG(ERR, "Invalid link_speed/link_duplex (%u/%u) for port %u",
-		     dev->data->dev_conf.link_speed,
-		     dev->data->dev_conf.link_duplex, dev->data->port_id);
+	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
+		     dev->data->dev_conf.link_speeds, dev->data->port_id);
 	em_dev_clear_queues(dev);
 	return -EINVAL;
 }
@@ -1106,8 +1098,11 @@ eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	/* Now we check if a transition has happened */
 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
-		hw->mac.ops.get_link_up_info(hw, &link.link_speed,
-			&link.link_duplex);
+		uint16_t duplex, speed;
+		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
+		link.link_duplex = duplex ? ETH_LINK_FULL_DUPLEX :
+				ETH_LINK_HALF_DUPLEX;
+		link.link_speed = speed;
 		link.link_status = ETH_LINK_UP;
 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
 		link.link_speed = 0;
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index 95d1711..ced864c 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -1133,6 +1133,9 @@ eth_igb_start(struct rte_eth_dev *dev)
 	int ret, mask;
 	uint32_t intr_vector = 0;
 	uint32_t ctrl_ext;
+	uint32_t *speeds;
+	int num_speeds;
+	bool autoneg;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -1233,48 +1236,46 @@ eth_igb_start(struct rte_eth_dev *dev)
 	}
 
 	/* Setup link speed and duplex */
-	switch (dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_HALF_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_FULL_DUPLEX;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_HALF;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_FULL;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_100M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_HALF;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_FULL;
-		else
+	speeds = &dev->data->dev_conf.link_speeds;
+	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
+	} else {
+		num_speeds = 0;
+		autoneg = (*speeds & ETH_LINK_SPEED_AUTONEG);
+
+		/* Reset */
+		hw->phy.autoneg_advertised = 0;
+
+		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
+				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G)) {
+			num_speeds = -1;
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_1G:
-		if ((dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX) ||
-				(dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX))
-			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
-		else
+		}
+		if (*speeds & ETH_LINK_SPEED_10M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_10M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_1G) {
+			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
+			num_speeds++;
+		}
+		if (num_speeds == 0 || (!autoneg && (num_speeds > 2)))
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10G:
-	default:
-		goto error_invalid_config;
 	}
+
 	e1000_setup_link(hw);
 
 	if (rte_intr_allow_others(intr_handle)) {
@@ -1306,9 +1307,8 @@ eth_igb_start(struct rte_eth_dev *dev)
 	return 0;
 
 error_invalid_config:
-	PMD_INIT_LOG(ERR, "Invalid link_speed/link_duplex (%u/%u) for port %u",
-		     dev->data->dev_conf.link_speed,
-		     dev->data->dev_conf.link_duplex, dev->data->port_id);
+	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
+		     dev->data->dev_conf.link_speeds, dev->data->port_id);
 	igb_dev_clear_queues(dev);
 	return -EINVAL;
 }
@@ -2061,13 +2061,19 @@ eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	/* Now we check if a transition has happened */
 	if (link_check) {
-		hw->mac.ops.get_link_up_info(hw, &link.link_speed,
-					  &link.link_duplex);
+		uint16_t duplex, speed;
+		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
+		link.link_duplex = duplex ? ETH_LINK_FULL_DUPLEX :
+				ETH_LINK_HALF_DUPLEX;
+		link.link_speed = speed;
 		link.link_status = ETH_LINK_UP;
+		link.link_autoneg = (dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_AUTONEG);
 	} else if (!link_check) {
 		link.link_speed = 0;
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
 		link.link_status = ETH_LINK_DOWN;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 	rte_igb_dev_atomic_write_link_status(dev, &link);
 
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index dce31db..87bc767 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -1381,27 +1381,20 @@ i40e_vsi_disable_queues_intr(struct i40e_vsi *vsi)
 }
 
 static inline uint8_t
-i40e_parse_link_speed(uint16_t eth_link_speed)
+i40e_parse_link_speeds(uint16_t link_speeds)
 {
 	uint8_t link_speed = I40E_LINK_SPEED_UNKNOWN;
 
-	switch (eth_link_speed) {
-	case ETH_SPEED_NUM_40G:
-		link_speed = I40E_LINK_SPEED_40GB;
-		break;
-	case ETH_SPEED_NUM_20G:
-		link_speed = I40E_LINK_SPEED_20GB;
-		break;
-	case ETH_SPEED_NUM_10G:
-		link_speed = I40E_LINK_SPEED_10GB;
-		break;
-	case ETH_SPEED_NUM_1G:
-		link_speed = I40E_LINK_SPEED_1GB;
-		break;
-	case ETH_SPEED_NUM_100M:
-		link_speed = I40E_LINK_SPEED_100MB;
-		break;
-	}
+	if (link_speeds & ETH_LINK_SPEED_40G)
+		link_speed |= I40E_LINK_SPEED_40GB;
+	if (link_speeds & ETH_LINK_SPEED_20G)
+		link_speed |= I40E_LINK_SPEED_20GB;
+	if (link_speeds & ETH_LINK_SPEED_10G)
+		link_speed |= I40E_LINK_SPEED_10GB;
+	if (link_speeds & ETH_LINK_SPEED_1G)
+		link_speed |= I40E_LINK_SPEED_1GB;
+	if (link_speeds & ETH_LINK_SPEED_100M)
+		link_speed |= I40E_LINK_SPEED_100MB;
 
 	return link_speed;
 }
@@ -1427,9 +1420,9 @@ i40e_apply_link_speed(struct rte_eth_dev *dev)
 	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 	struct rte_eth_conf *conf = &dev->data->dev_conf;
 
-	speed = i40e_parse_link_speed(conf->link_speed);
+	speed = i40e_parse_link_speeds(conf->link_speeds);
 	abilities |= I40E_AQ_PHY_ENABLE_ATOMIC_LINK;
-	if (conf->link_speed == ETH_LINK_SPEED_AUTONEG)
+	if (conf->link_speeds & ETH_LINK_SPEED_AUTONEG)
 		abilities |= I40E_AQ_PHY_AN_ENABLED;
 	else
 		abilities |= I40E_AQ_PHY_LINK_ENABLED;
@@ -1449,10 +1442,8 @@ i40e_dev_start(struct rte_eth_dev *dev)
 
 	hw->adapter_stopped = 0;
 
-	if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
-		(dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
-		PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
-			     dev->data->dev_conf.link_duplex,
+	if (!(dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_AUTONEG)) {
+		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; autonegotiation disabled",
 			     dev->data->port_id);
 		return -EINVAL;
 	}
@@ -1525,6 +1516,12 @@ i40e_dev_start(struct rte_eth_dev *dev)
 	}
 
 	/* Apply link configure */
+	if (dev->data->dev_conf.link_speeds & ~(ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G |
+				ETH_LINK_SPEED_20G | ETH_LINK_SPEED_40G)) {
+		PMD_DRV_LOG(ERR, "Invalid link setting");
+		goto err_up;
+	}
 	ret = i40e_apply_link_speed(dev);
 	if (I40E_SUCCESS != ret) {
 		PMD_DRV_LOG(ERR, "Fail to apply link setting");
@@ -1809,6 +1806,9 @@ i40e_dev_link_update(struct rte_eth_dev *dev,
 		break;
 	}
 
+	link.link_autoneg = (dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_AUTONEG);
+
 out:
 	rte_i40e_dev_atomic_write_link_status(dev, &link);
 	if (link.link_status == old.link_status)
diff --git a/drivers/net/i40e/i40e_ethdev_vf.c b/drivers/net/i40e/i40e_ethdev_vf.c
index 295dcd2..8cf22ee 100644
--- a/drivers/net/i40e/i40e_ethdev_vf.c
+++ b/drivers/net/i40e/i40e_ethdev_vf.c
@@ -2121,12 +2121,13 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
 	 * DPDK pf host provide interfacet to acquire link status
 	 * while Linux driver does not
 	 */
-	if (vf->version_major == I40E_DPDK_VERSION_MAJOR)
+	if (vf->version_major == I40E_DPDK_VERSION_MAJOR) {
 		i40evf_get_link_status(dev, &new_link);
-	else {
+	} else {
 		/* Always assume it's up, for Linux driver PF host */
-		new_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
 		new_link.link_speed  = ETH_SPEED_NUM_10G;
+		new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
+		new_link.link_autoneg = ETH_LINK_SPEED_AUTONEG;
 		new_link.link_status = ETH_LINK_UP;
 	}
 	i40evf_dev_atomic_write_link_status(dev, &new_link);
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a98e8eb..ff23d7d 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -2094,14 +2094,13 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 	int mask = 0;
 	int status;
 	uint16_t vf, idx;
+	uint32_t *link_speeds;
 
 	PMD_INIT_FUNC_TRACE();
 
 	/* IXGBE devices don't support half duplex */
-	if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
-			(dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
-		PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
-			     dev->data->dev_conf.link_duplex,
+	if (!(dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_AUTONEG)) {
+		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; autonegotiation disabled",
 			     dev->data->port_id);
 		return -EINVAL;
 	}
@@ -2193,32 +2192,21 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 	if (err)
 		goto error;
 
-	switch(dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		speed = (hw->mac.type != ixgbe_mac_82598EB) ?
-				IXGBE_LINK_SPEED_82599_AUTONEG :
-				IXGBE_LINK_SPEED_82598_AUTONEG;
-		break;
-	case ETH_SPEED_NUM_100M:
-		/*
-		 * Invalid for 82598 but error will be detected by
-		 * ixgbe_setup_link()
-		 */
-		speed = IXGBE_LINK_SPEED_100_FULL;
-		break;
-	case ETH_SPEED_NUM_1G:
-		speed = IXGBE_LINK_SPEED_1GB_FULL;
-		break;
-	case ETH_SPEED_NUM_10G:
-		speed = IXGBE_LINK_SPEED_10GB_FULL;
-		break;
-	default:
-		PMD_INIT_LOG(ERR, "Invalid link_speed (%hu) for port %hhu",
-			     dev->data->dev_conf.link_speed,
-			     dev->data->port_id);
+	link_speeds = &dev->data->dev_conf.link_speeds;
+	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
+			ETH_LINK_SPEED_10G)) {
+		PMD_INIT_LOG(ERR, "Invalid link setting");
 		goto error;
 	}
 
+	speed = 0x0;
+	if (*link_speeds & ETH_LINK_SPEED_10G)
+		speed |= IXGBE_LINK_SPEED_10GB_FULL;
+	if (*link_speeds & ETH_LINK_SPEED_1G)
+		speed |= IXGBE_LINK_SPEED_1GB_FULL;
+	if (*link_speeds & ETH_LINK_SPEED_100M)
+		speed |= IXGBE_LINK_SPEED_100_FULL;
+
 	err = ixgbe_setup_link(hw, speed, link_up);
 	if (err)
 		goto error;
@@ -3083,7 +3071,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	if (diag != 0) {
 		link.link_speed = ETH_SPEED_NUM_100M;
-		link.link_duplex = ETH_LINK_HALF_DUPLEX;
+		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
 		if (link.link_status == old.link_status)
 			return -1;
@@ -3102,7 +3090,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 	switch (link_speed) {
 	default:
 	case IXGBE_LINK_SPEED_UNKNOWN:
-		link.link_duplex = ETH_LINK_HALF_DUPLEX;
+		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 
diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c
index 59ac423..43ac763 100644
--- a/drivers/net/mlx4/mlx4.c
+++ b/drivers/net/mlx4/mlx4.c
@@ -4721,6 +4721,8 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
 		dev_link.link_speed = link_speed;
 	dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
 				ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+	dev_link.link_autoneg = (dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_AUTONEG);
 	if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
 		/* Link status changed. */
 		dev->data->dev_link = dev_link;
diff --git a/drivers/net/mpipe/mpipe_tilegx.c b/drivers/net/mpipe/mpipe_tilegx.c
index 1a77c7a..adcbc19 100644
--- a/drivers/net/mpipe/mpipe_tilegx.c
+++ b/drivers/net/mpipe/mpipe_tilegx.c
@@ -394,6 +394,8 @@ mpipe_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 		speed = state & GXIO_MPIPE_LINK_SPEED_MASK;
 
+		new.link_autoneg = (dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_AUTONEG);
 		if (speed == GXIO_MPIPE_LINK_1G) {
 			new.link_speed = ETH_SPEED_NUM_1G;
 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c
index 5640585..5e8e203 100644
--- a/drivers/net/null/rte_eth_null.c
+++ b/drivers/net/null/rte_eth_null.c
@@ -92,6 +92,7 @@ static struct rte_eth_link pmd_link = {
 	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
+	.link_autoneg = ETH_LINK_SPEED_AUTONEG,
 };
 
 static uint16_t
diff --git a/drivers/net/pcap/rte_eth_pcap.c b/drivers/net/pcap/rte_eth_pcap.c
index c657951..c98e234 100644
--- a/drivers/net/pcap/rte_eth_pcap.c
+++ b/drivers/net/pcap/rte_eth_pcap.c
@@ -126,6 +126,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_FIXED,
 };
 
 static int
diff --git a/drivers/net/ring/rte_eth_ring.c b/drivers/net/ring/rte_eth_ring.c
index 58685e9..b1783c3 100644
--- a/drivers/net/ring/rte_eth_ring.c
+++ b/drivers/net/ring/rte_eth_ring.c
@@ -80,6 +80,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_AUTONEG
 };
 
 static uint16_t
diff --git a/drivers/net/szedata2/rte_eth_szedata2.c b/drivers/net/szedata2/rte_eth_szedata2.c
index dd1ae9e..ee97a4e 100644
--- a/drivers/net/szedata2/rte_eth_szedata2.c
+++ b/drivers/net/szedata2/rte_eth_szedata2.c
@@ -1174,6 +1174,8 @@ eth_link_update(struct rte_eth_dev *dev,
 	link.link_status = (cgmii_ibuf_is_enabled(ibuf) &&
 			cgmii_ibuf_is_link_up(ibuf)) ? ETH_LINK_UP : ETH_LINK_DOWN;
 
+	link.link_autoneg = ETH_LINK_SPEED_FIXED;
+
 	rte_atomic64_cmpset((uint64_t *)dev_link, *(uint64_t *)dev_link,
 			*(uint64_t *)link_ptr);
 
diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
index 6afa14e..94d1b2c 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
@@ -780,6 +780,7 @@ vmxnet3_dev_link_update(struct rte_eth_dev *dev, __attribute__((unused)) int wai
 		link.link_status = ETH_LINK_UP;
 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		link.link_speed = ETH_SPEED_NUM_10G;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 
 	vmxnet3_dev_atomic_write_link_status(dev, &link);
diff --git a/drivers/net/xenvirt/rte_eth_xenvirt.c b/drivers/net/xenvirt/rte_eth_xenvirt.c
index 77d3ba1..b9638d9 100644
--- a/drivers/net/xenvirt/rte_eth_xenvirt.c
+++ b/drivers/net/xenvirt/rte_eth_xenvirt.c
@@ -73,6 +73,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_FIXED
 };
 
 static void
diff --git a/examples/ip_pipeline/config_parse.c b/examples/ip_pipeline/config_parse.c
index 152889d..2cd5707 100644
--- a/examples/ip_pipeline/config_parse.c
+++ b/examples/ip_pipeline/config_parse.c
@@ -87,8 +87,7 @@ static const struct app_link_params link_params_default = {
 	.pci_bdf = {0},
 
 	.conf = {
-		.link_speed = 0,
-		.link_duplex = 0,
+		.link_speeds = 0,
 		.rxmode = {
 			.mq_mode = ETH_MQ_RX_NONE,
 
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 49fdcb7..03d3278 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -244,6 +244,8 @@ struct rte_eth_stats {
 /**
  * Device supported speeds bitmap flags
  */
+#define ETH_LINK_SPEED_FIXED    (0 <<  0)  /**< Disable autoneg (fixed speed) */
+#define ETH_LINK_SPEED_AUTONEG  (1 <<  0)  /**< Autonegotiate (all speeds) */
 #define ETH_LINK_SPEED_10M_HD   (1 <<  1)  /**<  10 Mbps half-duplex */
 #define ETH_LINK_SPEED_10M      (1 <<  2)  /**<  10 Mbps full-duplex */
 #define ETH_LINK_SPEED_100M_HD  (1 <<  3)  /**< 100 Mbps half-duplex */
@@ -261,7 +263,7 @@ struct rte_eth_stats {
 /**
  * Ethernet numeric link speeds in Mbps
  */
-#define ETH_LINK_SPEED_AUTONEG     0 /**< Auto-negotiate link speed. */
+#define ETH_SPEED_NUM_NONE         0 /**< Not defined */
 #define ETH_SPEED_NUM_10M         10 /**<  10 Mbps */
 #define ETH_SPEED_NUM_100M       100 /**< 100 Mbps */
 #define ETH_SPEED_NUM_1G        1000 /**<   1 Gbps */
@@ -278,15 +280,15 @@ struct rte_eth_stats {
  * A structure used to retrieve link-level information of an Ethernet port.
  */
 struct rte_eth_link {
-	uint16_t link_speed;      /**< ETH_SPEED_NUM_ */
-	uint16_t link_duplex;     /**< ETH_LINK_[HALF/FULL]_DUPLEX */
-	uint8_t  link_status : 1; /**< ETH_LINK_[DOWN/UP] */
-}__attribute__((aligned(8)));     /**< aligned for atomic64 read/write */
+	uint16_t link_speed;        /**< ETH_SPEED_NUM_ */
+	uint16_t link_duplex  : 1;  /**< ETH_LINK_[HALF/FULL]_DUPLEX */
+	uint16_t link_autoneg : 1;  /**< ETH_LINK_SPEED_[AUTONEG/FIXED] */
+	uint16_t link_status  : 1;  /**< ETH_LINK_[DOWN/UP] */
+} __attribute__((aligned(8)));      /**< aligned for atomic64 read/write */
 
 /* Utility constants */
-#define ETH_LINK_AUTONEG_DUPLEX 0       /**< Auto-negotiate duplex. */
-#define ETH_LINK_HALF_DUPLEX    1       /**< Half-duplex connection. */
-#define ETH_LINK_FULL_DUPLEX    2       /**< Full-duplex connection. */
+#define ETH_LINK_HALF_DUPLEX    0 /**< Half-duplex connection. */
+#define ETH_LINK_FULL_DUPLEX    1 /**< Full-duplex connection. */
 #define ETH_LINK_DOWN           0 /**< Link is down. */
 #define ETH_LINK_UP             1 /**< Link is up. */
 
@@ -802,10 +804,13 @@ struct rte_intr_conf {
  * configuration settings may be needed.
  */
 struct rte_eth_conf {
-	uint16_t link_speed;
-	/**< ETH_SPEED_NUM_ or 0 for autonegotiation */
-	uint16_t link_duplex;
-	/**< ETH_LINK_[HALF_DUPLEX|FULL_DUPLEX], or 0 for autonegotation */
+	uint32_t link_speeds; /**< bitmap of ETH_LINK_SPEED_XXX of speeds to be
+				used. ETH_LINK_SPEED_FIXED disables link
+				autonegotiation, and a unique speed shall be
+				set. Otherwise, the bitmap defines the set of
+				speeds to be advertised. If the special value
+				ETH_LINK_SPEED_AUTONEG (0) is used, all speeds
+				supported are advertised. */
 	struct rte_eth_rxmode rxmode; /**< Port RX configuration. */
 	struct rte_eth_txmode txmode; /**< Port TX configuration. */
 	uint32_t lpbk_mode; /**< Loopback operation mode. By default the value
-- 
2.7.0

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH v13 6/8] ethdev: redesign link speed config
    @ 2016-03-26  1:27  3%     ` Marc Sune
    2 siblings, 0 replies; 200+ results
From: Marc Sune @ 2016-03-26  1:27 UTC (permalink / raw)
  To: Thomas Monjalon, Xu, Qian Q, Xing, Beilei, dev, Ananyev,
	Konstantin, Lu, Wenzhuo, Richardson, Bruce, Glynn, Michael J
  Cc: Marc Sune

This patch redesigns the API to set the link speed/s configuration
of an ethernet port. Specifically:

- it allows to define a set of advertised speeds for
  auto-negociation.
- it allows to disable link auto-negociation (single fixed speed).
- default: auto-negociate all supported speeds.

A flag autoneg in struct rte_eth_link indicates if link speed was a
result of auto-negociation or was fixed by configuration.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Tested-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
 app/test-pmd/cmdline.c                    |  26 ++++----
 doc/guides/rel_notes/deprecation.rst      |   3 -
 doc/guides/rel_notes/release_16_04.rst    |   9 +++
 drivers/net/af_packet/rte_eth_af_packet.c |   1 +
 drivers/net/bnx2x/bnx2x_ethdev.c          |   4 +-
 drivers/net/bonding/rte_eth_bond_8023ad.c |   2 +-
 drivers/net/e1000/em_ethdev.c             | 103 +++++++++++++++---------------
 drivers/net/e1000/igb_ethdev.c            |  95 ++++++++++++++-------------
 drivers/net/i40e/i40e_ethdev.c            |  48 +++++++-------
 drivers/net/i40e/i40e_ethdev_vf.c         |   7 +-
 drivers/net/ixgbe/ixgbe_ethdev.c          |  51 ++++++---------
 drivers/net/mlx4/mlx4.c                   |   2 +
 drivers/net/mlx5/mlx5_ethdev.c            |   2 +
 drivers/net/mpipe/mpipe_tilegx.c          |   2 +
 drivers/net/null/rte_eth_null.c           |   1 +
 drivers/net/pcap/rte_eth_pcap.c           |   1 +
 drivers/net/ring/rte_eth_ring.c           |   1 +
 drivers/net/szedata2/rte_eth_szedata2.c   |   2 +
 drivers/net/vmxnet3/vmxnet3_ethdev.c      |   1 +
 drivers/net/xenvirt/rte_eth_xenvirt.c     |   1 +
 examples/ip_pipeline/config_parse.c       |   3 +-
 lib/librte_ether/rte_ethdev.h             |  29 +++++----
 22 files changed, 207 insertions(+), 187 deletions(-)

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 815b53b..741cac3 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -989,7 +989,7 @@ struct cmd_config_speed_all {
 };
 
 static int
-parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
+parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint32_t *speed)
 {
 
 	int duplex;
@@ -1006,20 +1006,22 @@ parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
 	}
 
 	if (!strcmp(speedstr, "10")) {
-		*speed = ETH_SPEED_NUM_10M;
+		*speed = (duplex == ETH_LINK_HALF_DUPLEX) ?
+				ETH_LINK_SPEED_10M_HD : ETH_LINK_SPEED_10M;
 	} else if (!strcmp(speedstr, "100")) {
-		*speed = ETH_SPEED_NUM_100M;
+		*speed = (duplex == ETH_LINK_HALF_DUPLEX) ?
+				ETH_LINK_SPEED_100M_HD : ETH_LINK_SPEED_100M;
 	} else {
 		if (duplex != ETH_LINK_FULL_DUPLEX) {
 			printf("Invalid speed/duplex parameters\n");
 			return -1;
 		}
 		if (!strcmp(speedstr, "1000")) {
-			*speed = ETH_SPEED_NUM_1G;
+			*speed = ETH_LINK_SPEED_1G;
 		} else if (!strcmp(speedstr, "10000")) {
-			*speed = ETH_SPEED_NUM_10G;
+			*speed = ETH_LINK_SPEED_10G;
 		} else if (!strcmp(speedstr, "40000")) {
-			*speed = ETH_SPEED_NUM_40G;
+			*speed = ETH_LINK_SPEED_40G;
 		} else if (!strcmp(speedstr, "auto")) {
 			*speed = ETH_LINK_SPEED_AUTONEG;
 		} else {
@@ -1037,8 +1039,7 @@ cmd_config_speed_all_parsed(void *parsed_result,
 			__attribute__((unused)) void *data)
 {
 	struct cmd_config_speed_all *res = parsed_result;
-	uint16_t link_speed = ETH_LINK_SPEED_AUTONEG;
-	uint16_t link_duplex = 0;
+	uint32_t link_speed;
 	portid_t pid;
 
 	if (!all_ports_stopped()) {
@@ -1051,8 +1052,7 @@ cmd_config_speed_all_parsed(void *parsed_result,
 		return;
 
 	FOREACH_PORT(pid, ports) {
-		ports[pid].dev_conf.link_speed = link_speed;
-		ports[pid].dev_conf.link_duplex = link_duplex;
+		ports[pid].dev_conf.link_speeds = link_speed;
 	}
 
 	cmd_reconfig_device_queue(RTE_PORT_ALL, 1, 1);
@@ -1110,8 +1110,7 @@ cmd_config_speed_specific_parsed(void *parsed_result,
 				__attribute__((unused)) void *data)
 {
 	struct cmd_config_speed_specific *res = parsed_result;
-	uint16_t link_speed = ETH_LINK_SPEED_AUTONEG;
-	uint16_t link_duplex = 0;
+	uint32_t link_speed;
 
 	if (!all_ports_stopped()) {
 		printf("Please stop all ports first\n");
@@ -1125,8 +1124,7 @@ cmd_config_speed_specific_parsed(void *parsed_result,
 			&link_speed) < 0)
 		return;
 
-	ports[res->id].dev_conf.link_speed = link_speed;
-	ports[res->id].dev_conf.link_duplex = link_duplex;
+	ports[res->id].dev_conf.link_speeds = link_speed;
 
 	cmd_reconfig_device_queue(RTE_PORT_ALL, 1, 1);
 }
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 179e30f..c47610d 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -19,9 +19,6 @@ Deprecation Notices
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff
 
-* The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
-  must be updated to support 100G link and to have a cleaner link speed API.
-
 * 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]
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 9e7b0b7..c891a55 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -52,6 +52,12 @@ This section should contain new features added in this release. Sample format:
   The structure ``rte_eth_dev_info`` has now a ``speed_capa`` bitmap, which
   allows the application to know the supported speeds of each device.
 
+* **Added bitmap of link speeds to advertise.**
+
+  Allow defining a set of advertised speeds for auto-negotiation,
+  explicitly disabling link auto-negotiation (single speed)
+  and full auto-negotiation.
+
 * **Added new poll-mode driver for Amazon Elastic Network Adapters (ENA).**
 
   The driver operates variety of ENA adapters through feature negotiation
@@ -464,6 +470,9 @@ This section should contain API changes. Sample format:
 * The ethdev structure ``rte_eth_dev_info`` was changed to support device
   speed capabilities.
 
+* The ethdev structures ``rte_eth_link`` and ``rte_eth_conf`` were changed to
+  support the new link API.
+
 * The functions ``rte_eth_dev_udp_tunnel_add`` and ``rte_eth_dev_udp_tunnel_delete``
   have been renamed into ``rte_eth_dev_udp_tunnel_port_add`` and
   ``rte_eth_dev_udp_tunnel_port_delete``.
diff --git a/drivers/net/af_packet/rte_eth_af_packet.c b/drivers/net/af_packet/rte_eth_af_packet.c
index 641f849..f17bd7e 100644
--- a/drivers/net/af_packet/rte_eth_af_packet.c
+++ b/drivers/net/af_packet/rte_eth_af_packet.c
@@ -119,6 +119,7 @@ static struct rte_eth_link pmd_link = {
 	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
+	.link_autoneg = ETH_LINK_SPEED_AUTONEG
 };
 
 static uint16_t
diff --git a/drivers/net/bnx2x/bnx2x_ethdev.c b/drivers/net/bnx2x/bnx2x_ethdev.c
index 897081f..071b44f 100644
--- a/drivers/net/bnx2x/bnx2x_ethdev.c
+++ b/drivers/net/bnx2x/bnx2x_ethdev.c
@@ -44,9 +44,9 @@ bnx2x_link_update(struct rte_eth_dev *dev)
 		case DUPLEX_HALF:
 			dev->data->dev_link.link_duplex = ETH_LINK_HALF_DUPLEX;
 			break;
-		default:
-			dev->data->dev_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
 	}
+	dev->data->dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
 	dev->data->dev_link.link_status = sc->link_vars.link_up;
 }
 
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index ac8306f..cca7cc3 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -708,7 +708,7 @@ link_speed_key(uint16_t speed) {
 	uint16_t key_speed;
 
 	switch (speed) {
-	case ETH_LINK_SPEED_AUTONEG:
+	case ETH_SPEED_NUM_NONE:
 		key_speed = 0x00;
 		break;
 	case ETH_SPEED_NUM_10M:
diff --git a/drivers/net/e1000/em_ethdev.c b/drivers/net/e1000/em_ethdev.c
index d5f8c7f..9fb59a2 100644
--- a/drivers/net/e1000/em_ethdev.c
+++ b/drivers/net/e1000/em_ethdev.c
@@ -538,6 +538,9 @@ eth_em_start(struct rte_eth_dev *dev)
 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
 	int ret, mask;
 	uint32_t intr_vector = 0;
+	uint32_t *speeds;
+	int num_speeds;
+	bool autoneg;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -612,56 +615,46 @@ eth_em_start(struct rte_eth_dev *dev)
 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
 
 	/* Setup link speed and duplex */
-	switch (dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_HALF_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_FULL_DUPLEX;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_HALF;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_FULL;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_100M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_HALF;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_FULL;
-		else
+	speeds = &dev->data->dev_conf.link_speeds;
+	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
+	} else {
+		num_speeds = 0;
+		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
+
+		/* Reset */
+		hw->phy.autoneg_advertised = 0;
+
+		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
+				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
+			num_speeds = -1;
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_1G:
-		if ((dev->data->dev_conf.link_duplex ==
-				ETH_LINK_AUTONEG_DUPLEX) ||
-			(dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX))
-			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
-		else
+		}
+		if (*speeds & ETH_LINK_SPEED_10M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_10M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_1G) {
+			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
+			num_speeds++;
+		}
+		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10G:
-	default:
-		goto error_invalid_config;
 	}
+
 	e1000_setup_link(hw);
 
 	if (rte_intr_allow_others(intr_handle)) {
@@ -694,9 +687,8 @@ eth_em_start(struct rte_eth_dev *dev)
 	return 0;
 
 error_invalid_config:
-	PMD_INIT_LOG(ERR, "Invalid link_speed/link_duplex (%u/%u) for port %u",
-		     dev->data->dev_conf.link_speed,
-		     dev->data->dev_conf.link_duplex, dev->data->port_id);
+	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
+		     dev->data->dev_conf.link_speeds, dev->data->port_id);
 	em_dev_clear_queues(dev);
 	return -EINVAL;
 }
@@ -1106,13 +1098,20 @@ eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	/* Now we check if a transition has happened */
 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
-		hw->mac.ops.get_link_up_info(hw, &link.link_speed,
-			&link.link_duplex);
+		uint16_t duplex, speed;
+		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
+		link.link_duplex = (duplex == FULL_DUPLEX) ?
+				ETH_LINK_FULL_DUPLEX :
+				ETH_LINK_HALF_DUPLEX;
+		link.link_speed = speed;
 		link.link_status = ETH_LINK_UP;
+		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_FIXED);
 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
 		link.link_speed = 0;
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
 		link.link_status = ETH_LINK_DOWN;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 	rte_em_dev_atomic_write_link_status(dev, &link);
 
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index 95d1711..e0053fe 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -1133,6 +1133,9 @@ eth_igb_start(struct rte_eth_dev *dev)
 	int ret, mask;
 	uint32_t intr_vector = 0;
 	uint32_t ctrl_ext;
+	uint32_t *speeds;
+	int num_speeds;
+	bool autoneg;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -1233,48 +1236,46 @@ eth_igb_start(struct rte_eth_dev *dev)
 	}
 
 	/* Setup link speed and duplex */
-	switch (dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_HALF_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_FULL_DUPLEX;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_HALF;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_FULL;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_100M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_HALF;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_FULL;
-		else
+	speeds = &dev->data->dev_conf.link_speeds;
+	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
+	} else {
+		num_speeds = 0;
+		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
+
+		/* Reset */
+		hw->phy.autoneg_advertised = 0;
+
+		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
+				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
+			num_speeds = -1;
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_1G:
-		if ((dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX) ||
-				(dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX))
-			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
-		else
+		}
+		if (*speeds & ETH_LINK_SPEED_10M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_10M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_1G) {
+			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
+			num_speeds++;
+		}
+		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10G:
-	default:
-		goto error_invalid_config;
 	}
+
 	e1000_setup_link(hw);
 
 	if (rte_intr_allow_others(intr_handle)) {
@@ -1306,9 +1307,8 @@ eth_igb_start(struct rte_eth_dev *dev)
 	return 0;
 
 error_invalid_config:
-	PMD_INIT_LOG(ERR, "Invalid link_speed/link_duplex (%u/%u) for port %u",
-		     dev->data->dev_conf.link_speed,
-		     dev->data->dev_conf.link_duplex, dev->data->port_id);
+	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
+		     dev->data->dev_conf.link_speeds, dev->data->port_id);
 	igb_dev_clear_queues(dev);
 	return -EINVAL;
 }
@@ -2061,13 +2061,20 @@ eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	/* Now we check if a transition has happened */
 	if (link_check) {
-		hw->mac.ops.get_link_up_info(hw, &link.link_speed,
-					  &link.link_duplex);
+		uint16_t duplex, speed;
+		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
+		link.link_duplex = (duplex == FULL_DUPLEX) ?
+				ETH_LINK_FULL_DUPLEX :
+				ETH_LINK_HALF_DUPLEX;
+		link.link_speed = speed;
 		link.link_status = ETH_LINK_UP;
+		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_FIXED);
 	} else if (!link_check) {
 		link.link_speed = 0;
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
 		link.link_status = ETH_LINK_DOWN;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 	rte_igb_dev_atomic_write_link_status(dev, &link);
 
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index dce31db..c9ef417 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -1381,27 +1381,20 @@ i40e_vsi_disable_queues_intr(struct i40e_vsi *vsi)
 }
 
 static inline uint8_t
-i40e_parse_link_speed(uint16_t eth_link_speed)
+i40e_parse_link_speeds(uint16_t link_speeds)
 {
 	uint8_t link_speed = I40E_LINK_SPEED_UNKNOWN;
 
-	switch (eth_link_speed) {
-	case ETH_SPEED_NUM_40G:
-		link_speed = I40E_LINK_SPEED_40GB;
-		break;
-	case ETH_SPEED_NUM_20G:
-		link_speed = I40E_LINK_SPEED_20GB;
-		break;
-	case ETH_SPEED_NUM_10G:
-		link_speed = I40E_LINK_SPEED_10GB;
-		break;
-	case ETH_SPEED_NUM_1G:
-		link_speed = I40E_LINK_SPEED_1GB;
-		break;
-	case ETH_SPEED_NUM_100M:
-		link_speed = I40E_LINK_SPEED_100MB;
-		break;
-	}
+	if (link_speeds & ETH_LINK_SPEED_40G)
+		link_speed |= I40E_LINK_SPEED_40GB;
+	if (link_speeds & ETH_LINK_SPEED_20G)
+		link_speed |= I40E_LINK_SPEED_20GB;
+	if (link_speeds & ETH_LINK_SPEED_10G)
+		link_speed |= I40E_LINK_SPEED_10GB;
+	if (link_speeds & ETH_LINK_SPEED_1G)
+		link_speed |= I40E_LINK_SPEED_1GB;
+	if (link_speeds & ETH_LINK_SPEED_100M)
+		link_speed |= I40E_LINK_SPEED_100MB;
 
 	return link_speed;
 }
@@ -1427,9 +1420,9 @@ i40e_apply_link_speed(struct rte_eth_dev *dev)
 	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 	struct rte_eth_conf *conf = &dev->data->dev_conf;
 
-	speed = i40e_parse_link_speed(conf->link_speed);
+	speed = i40e_parse_link_speeds(conf->link_speeds);
 	abilities |= I40E_AQ_PHY_ENABLE_ATOMIC_LINK;
-	if (conf->link_speed == ETH_LINK_SPEED_AUTONEG)
+	if (!(conf->link_speeds & ETH_LINK_SPEED_FIXED))
 		abilities |= I40E_AQ_PHY_AN_ENABLED;
 	else
 		abilities |= I40E_AQ_PHY_LINK_ENABLED;
@@ -1449,10 +1442,8 @@ i40e_dev_start(struct rte_eth_dev *dev)
 
 	hw->adapter_stopped = 0;
 
-	if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
-		(dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
-		PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
-			     dev->data->dev_conf.link_duplex,
+	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
+		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; autonegotiation disabled",
 			     dev->data->port_id);
 		return -EINVAL;
 	}
@@ -1525,6 +1516,12 @@ i40e_dev_start(struct rte_eth_dev *dev)
 	}
 
 	/* Apply link configure */
+	if (dev->data->dev_conf.link_speeds & ~(ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G |
+				ETH_LINK_SPEED_20G | ETH_LINK_SPEED_40G)) {
+		PMD_DRV_LOG(ERR, "Invalid link setting");
+		goto err_up;
+	}
 	ret = i40e_apply_link_speed(dev);
 	if (I40E_SUCCESS != ret) {
 		PMD_DRV_LOG(ERR, "Fail to apply link setting");
@@ -1809,6 +1806,9 @@ i40e_dev_link_update(struct rte_eth_dev *dev,
 		break;
 	}
 
+	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
+
 out:
 	rte_i40e_dev_atomic_write_link_status(dev, &link);
 	if (link.link_status == old.link_status)
diff --git a/drivers/net/i40e/i40e_ethdev_vf.c b/drivers/net/i40e/i40e_ethdev_vf.c
index 295dcd2..8cf22ee 100644
--- a/drivers/net/i40e/i40e_ethdev_vf.c
+++ b/drivers/net/i40e/i40e_ethdev_vf.c
@@ -2121,12 +2121,13 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
 	 * DPDK pf host provide interfacet to acquire link status
 	 * while Linux driver does not
 	 */
-	if (vf->version_major == I40E_DPDK_VERSION_MAJOR)
+	if (vf->version_major == I40E_DPDK_VERSION_MAJOR) {
 		i40evf_get_link_status(dev, &new_link);
-	else {
+	} else {
 		/* Always assume it's up, for Linux driver PF host */
-		new_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
 		new_link.link_speed  = ETH_SPEED_NUM_10G;
+		new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
+		new_link.link_autoneg = ETH_LINK_SPEED_AUTONEG;
 		new_link.link_status = ETH_LINK_UP;
 	}
 	i40evf_dev_atomic_write_link_status(dev, &new_link);
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index a98e8eb..6cc2da0 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -2094,14 +2094,16 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 	int mask = 0;
 	int status;
 	uint16_t vf, idx;
+	uint32_t *link_speeds;
 
 	PMD_INIT_FUNC_TRACE();
 
-	/* IXGBE devices don't support half duplex */
-	if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
-			(dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
-		PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
-			     dev->data->dev_conf.link_duplex,
+	/* IXGBE devices don't support:
+	*    - half duplex (checked afterwards for valid speeds)
+	*    - fixed speed: TODO implement
+	*/
+	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
+		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; fix speed not supported",
 			     dev->data->port_id);
 		return -EINVAL;
 	}
@@ -2193,32 +2195,21 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 	if (err)
 		goto error;
 
-	switch(dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		speed = (hw->mac.type != ixgbe_mac_82598EB) ?
-				IXGBE_LINK_SPEED_82599_AUTONEG :
-				IXGBE_LINK_SPEED_82598_AUTONEG;
-		break;
-	case ETH_SPEED_NUM_100M:
-		/*
-		 * Invalid for 82598 but error will be detected by
-		 * ixgbe_setup_link()
-		 */
-		speed = IXGBE_LINK_SPEED_100_FULL;
-		break;
-	case ETH_SPEED_NUM_1G:
-		speed = IXGBE_LINK_SPEED_1GB_FULL;
-		break;
-	case ETH_SPEED_NUM_10G:
-		speed = IXGBE_LINK_SPEED_10GB_FULL;
-		break;
-	default:
-		PMD_INIT_LOG(ERR, "Invalid link_speed (%hu) for port %hhu",
-			     dev->data->dev_conf.link_speed,
-			     dev->data->port_id);
+	link_speeds = &dev->data->dev_conf.link_speeds;
+	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
+			ETH_LINK_SPEED_10G)) {
+		PMD_INIT_LOG(ERR, "Invalid link setting");
 		goto error;
 	}
 
+	speed = 0x0;
+	if (*link_speeds & ETH_LINK_SPEED_10G)
+		speed |= IXGBE_LINK_SPEED_10GB_FULL;
+	if (*link_speeds & ETH_LINK_SPEED_1G)
+		speed |= IXGBE_LINK_SPEED_1GB_FULL;
+	if (*link_speeds & ETH_LINK_SPEED_100M)
+		speed |= IXGBE_LINK_SPEED_100_FULL;
+
 	err = ixgbe_setup_link(hw, speed, link_up);
 	if (err)
 		goto error;
@@ -3083,7 +3074,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	if (diag != 0) {
 		link.link_speed = ETH_SPEED_NUM_100M;
-		link.link_duplex = ETH_LINK_HALF_DUPLEX;
+		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
 		if (link.link_status == old.link_status)
 			return -1;
@@ -3102,7 +3093,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 	switch (link_speed) {
 	default:
 	case IXGBE_LINK_SPEED_UNKNOWN:
-		link.link_duplex = ETH_LINK_HALF_DUPLEX;
+		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 
diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c
index 59ac423..81528c9 100644
--- a/drivers/net/mlx4/mlx4.c
+++ b/drivers/net/mlx4/mlx4.c
@@ -4721,6 +4721,8 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
 		dev_link.link_speed = link_speed;
 	dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
 				ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+	dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
 	if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
 		/* Link status changed. */
 		dev->data->dev_link = dev_link;
diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
index d7a0eea..beecc63 100644
--- a/drivers/net/mlx5/mlx5_ethdev.c
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -629,6 +629,8 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
 		dev_link.link_speed = link_speed;
 	dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
 				ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+	dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
 	if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
 		/* Link status changed. */
 		dev->data->dev_link = dev_link;
diff --git a/drivers/net/mpipe/mpipe_tilegx.c b/drivers/net/mpipe/mpipe_tilegx.c
index 1a77c7a..adcbc19 100644
--- a/drivers/net/mpipe/mpipe_tilegx.c
+++ b/drivers/net/mpipe/mpipe_tilegx.c
@@ -394,6 +394,8 @@ mpipe_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 		speed = state & GXIO_MPIPE_LINK_SPEED_MASK;
 
+		new.link_autoneg = (dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_AUTONEG);
 		if (speed == GXIO_MPIPE_LINK_1G) {
 			new.link_speed = ETH_SPEED_NUM_1G;
 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c
index 5640585..5e8e203 100644
--- a/drivers/net/null/rte_eth_null.c
+++ b/drivers/net/null/rte_eth_null.c
@@ -92,6 +92,7 @@ static struct rte_eth_link pmd_link = {
 	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
+	.link_autoneg = ETH_LINK_SPEED_AUTONEG,
 };
 
 static uint16_t
diff --git a/drivers/net/pcap/rte_eth_pcap.c b/drivers/net/pcap/rte_eth_pcap.c
index c657951..c98e234 100644
--- a/drivers/net/pcap/rte_eth_pcap.c
+++ b/drivers/net/pcap/rte_eth_pcap.c
@@ -126,6 +126,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_FIXED,
 };
 
 static int
diff --git a/drivers/net/ring/rte_eth_ring.c b/drivers/net/ring/rte_eth_ring.c
index 58685e9..b1783c3 100644
--- a/drivers/net/ring/rte_eth_ring.c
+++ b/drivers/net/ring/rte_eth_ring.c
@@ -80,6 +80,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_AUTONEG
 };
 
 static uint16_t
diff --git a/drivers/net/szedata2/rte_eth_szedata2.c b/drivers/net/szedata2/rte_eth_szedata2.c
index dd1ae9e..ee97a4e 100644
--- a/drivers/net/szedata2/rte_eth_szedata2.c
+++ b/drivers/net/szedata2/rte_eth_szedata2.c
@@ -1174,6 +1174,8 @@ eth_link_update(struct rte_eth_dev *dev,
 	link.link_status = (cgmii_ibuf_is_enabled(ibuf) &&
 			cgmii_ibuf_is_link_up(ibuf)) ? ETH_LINK_UP : ETH_LINK_DOWN;
 
+	link.link_autoneg = ETH_LINK_SPEED_FIXED;
+
 	rte_atomic64_cmpset((uint64_t *)dev_link, *(uint64_t *)dev_link,
 			*(uint64_t *)link_ptr);
 
diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
index 6afa14e..94d1b2c 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
@@ -780,6 +780,7 @@ vmxnet3_dev_link_update(struct rte_eth_dev *dev, __attribute__((unused)) int wai
 		link.link_status = ETH_LINK_UP;
 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		link.link_speed = ETH_SPEED_NUM_10G;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 
 	vmxnet3_dev_atomic_write_link_status(dev, &link);
diff --git a/drivers/net/xenvirt/rte_eth_xenvirt.c b/drivers/net/xenvirt/rte_eth_xenvirt.c
index 77d3ba1..b9638d9 100644
--- a/drivers/net/xenvirt/rte_eth_xenvirt.c
+++ b/drivers/net/xenvirt/rte_eth_xenvirt.c
@@ -73,6 +73,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_FIXED
 };
 
 static void
diff --git a/examples/ip_pipeline/config_parse.c b/examples/ip_pipeline/config_parse.c
index 152889d..2cd5707 100644
--- a/examples/ip_pipeline/config_parse.c
+++ b/examples/ip_pipeline/config_parse.c
@@ -87,8 +87,7 @@ static const struct app_link_params link_params_default = {
 	.pci_bdf = {0},
 
 	.conf = {
-		.link_speed = 0,
-		.link_duplex = 0,
+		.link_speeds = 0,
 		.rxmode = {
 			.mq_mode = ETH_MQ_RX_NONE,
 
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 49fdcb7..9a1466b 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -244,6 +244,8 @@ struct rte_eth_stats {
 /**
  * Device supported speeds bitmap flags
  */
+#define ETH_LINK_SPEED_AUTONEG  (0 <<  0)  /**< Autonegotiate (all speeds) */
+#define ETH_LINK_SPEED_FIXED    (1 <<  0)  /**< Disable autoneg (fixed speed) */
 #define ETH_LINK_SPEED_10M_HD   (1 <<  1)  /**<  10 Mbps half-duplex */
 #define ETH_LINK_SPEED_10M      (1 <<  2)  /**<  10 Mbps full-duplex */
 #define ETH_LINK_SPEED_100M_HD  (1 <<  3)  /**< 100 Mbps half-duplex */
@@ -261,7 +263,7 @@ struct rte_eth_stats {
 /**
  * Ethernet numeric link speeds in Mbps
  */
-#define ETH_LINK_SPEED_AUTONEG     0 /**< Auto-negotiate link speed. */
+#define ETH_SPEED_NUM_NONE         0 /**< Not defined */
 #define ETH_SPEED_NUM_10M         10 /**<  10 Mbps */
 #define ETH_SPEED_NUM_100M       100 /**< 100 Mbps */
 #define ETH_SPEED_NUM_1G        1000 /**<   1 Gbps */
@@ -278,15 +280,15 @@ struct rte_eth_stats {
  * A structure used to retrieve link-level information of an Ethernet port.
  */
 struct rte_eth_link {
-	uint16_t link_speed;      /**< ETH_SPEED_NUM_ */
-	uint16_t link_duplex;     /**< ETH_LINK_[HALF/FULL]_DUPLEX */
-	uint8_t  link_status : 1; /**< ETH_LINK_[DOWN/UP] */
-}__attribute__((aligned(8)));     /**< aligned for atomic64 read/write */
+	uint16_t link_speed;        /**< ETH_SPEED_NUM_ */
+	uint16_t link_duplex  : 1;  /**< ETH_LINK_[HALF/FULL]_DUPLEX */
+	uint16_t link_autoneg : 1;  /**< ETH_LINK_SPEED_[AUTONEG/FIXED] */
+	uint16_t link_status  : 1;  /**< ETH_LINK_[DOWN/UP] */
+} __attribute__((aligned(8)));      /**< aligned for atomic64 read/write */
 
 /* Utility constants */
-#define ETH_LINK_AUTONEG_DUPLEX 0       /**< Auto-negotiate duplex. */
-#define ETH_LINK_HALF_DUPLEX    1       /**< Half-duplex connection. */
-#define ETH_LINK_FULL_DUPLEX    2       /**< Full-duplex connection. */
+#define ETH_LINK_HALF_DUPLEX    0 /**< Half-duplex connection. */
+#define ETH_LINK_FULL_DUPLEX    1 /**< Full-duplex connection. */
 #define ETH_LINK_DOWN           0 /**< Link is down. */
 #define ETH_LINK_UP             1 /**< Link is up. */
 
@@ -802,10 +804,13 @@ struct rte_intr_conf {
  * configuration settings may be needed.
  */
 struct rte_eth_conf {
-	uint16_t link_speed;
-	/**< ETH_SPEED_NUM_ or 0 for autonegotiation */
-	uint16_t link_duplex;
-	/**< ETH_LINK_[HALF_DUPLEX|FULL_DUPLEX], or 0 for autonegotation */
+	uint32_t link_speeds; /**< bitmap of ETH_LINK_SPEED_XXX of speeds to be
+				used. ETH_LINK_SPEED_FIXED disables link
+				autonegotiation, and a unique speed shall be
+				set. Otherwise, the bitmap defines the set of
+				speeds to be advertised. If the special value
+				ETH_LINK_SPEED_AUTONEG (0) is used, all speeds
+				supported are advertised. */
 	struct rte_eth_rxmode rxmode; /**< Port RX configuration. */
 	struct rte_eth_txmode txmode; /**< Port TX configuration. */
 	uint32_t lpbk_mode; /**< Loopback operation mode. By default the value
-- 
2.1.4

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] 16.07 Roadmap
@ 2016-03-31  8:21  2% O'Driscoll, Tim
  0 siblings, 0 replies; 200+ results
From: O'Driscoll, Tim @ 2016-03-31  8:21 UTC (permalink / raw)
  To: dev

As we're nearing the completion of the 16.04 release, I'd like to let the community know our plans for 16.07. It would be good if others are also willing to share their plans so that we can build up a complete picture of what's targeted for 16.07.

These are the features that we're planning to submit:

Vhost/Virtio Performance Loopback Utility: A tool will be provided which will allow virtio/vhost performance testing without the need for NIC traffic.

Virtio Code Refactoring for Rx/TX Split: The Rx and Tx queues will be split as they have different information to maintain apart from the common vring. Other cleanups will be made to make the queues more friendly for optimization.

Virtio Descriptor Index Update: The virtio descriptor index will be optimized for cache2cache transfer in the virtio PMD. The performance increase is expected to be below 10%.

Virtio in Containers: Support will be added for virtio in containers (see http://dpdk.org/ml/archives/dev/2016-February/032786.html). Multi-queue support will also be added.

I40e NSH: This includes: 1. Recognize the Network Services Header packet type; 2. Direct traffic to queues based on service path header and service index (dependent on firmware change so may not make 16.07); 3. Checksum offload.

I40e Floating VEB: Deferred from 16.04. See http://dpdk.org/ml/archives/dev/2016-March/036470.html for details.

Automatic VF Reset From PF (i40e/ixgbe): Currently, when a PF notifies a VF that a reset is required, DPDK just reports this event to the application, which then needs to restart the VF port. A more user-friendly mechanism will be implemented where DPDK will reset the VF port directly. The application will still be notified, but will not need to handle the reset of the VF port.

Software Implementation of the KASUMI Algorithm: Under the cryptodev API, a software implementation of the KASUMI algorithm will be supported. KASUMI is widely used in mobile communications systems.

Bit-Level Support for SNOW 3G: Support for the SNOW 3G algorithm is being added in the 16.04 release. In 16.07, this will be enhanced so that offsets and lengths can be specified in bits instead of bytes (so, you could encrypt 50 bits of a stream starting from the 5th bit for example).

IPsec Sample App Enhancements: Support for IPv6 and Transport Mode will be added to the IPsec sample application that was submitted in 16.04.

XStats Enhancements: Improve the extended NIC stats API to use id value pairs instead of string value pairs. Remap stats registers to use standard interface MIB naming and sizing.

Keep-Alive Enhancements: Improve DPDK keep-alive to use the DPDK alarm/interrupt API instead of using callbacks.

Live Migration for SRIOV: Support for live migration for vhost-user is being added to 16.04. This will be further enhanced to support live migration for SR-IOV by using link bonding to bond an SRIOV interface with a virtio interface.

IP Pipeline Enhancements: This includes: 1. Configure the MAC address in the routing pipeline; 2. Enable RSS per network interface through the configuration file; 3. Streamline the CLI code of the IP pipeline application.

Packet Capture Framework: In 16.04, there was lots of discussion on requirements for tcpdump support in DPDK (see http://dpdk.org/ml/archives/dev/2016-March/035592.html). For 16.07, we plan to submit a packet capture framework which will support hooks for filtering capabilities such as BPF. Our specific use case for this is for low rate packet capture for debug purposes. It should be possible for others to extend the framework to support high rate packet capture if they require that capability.

External Mempool Manager: This was originally submitted for 16.04 but had to be deferred due to ABI changes. See http://dpdk.org/ml/archives/dev/2016-March/035107.html for details.


In addition, there are some features that we're working on now but which we know won't make 16.07, either because time is too tight or because of external dependencies. These include:

QEMU vHost Back-End Reconnect: Currently, if a vswitch is connected to VMs via vhost-user and the vswitch is restarted, then when it comes back up again it cannot reconnect to the existing VMs. To address this, both QEMU and vhost-user need to support client mode (currently only server mode is supported), which implements reconnection messages that allow the vswitch to reconnect to the VMs. Changes are required in QEMU as well as in DPDK, so this change will need to be coordinated with the QEMU community.

Delay Packet Copy in vHost-User Dequeue: It may be possible to increase vhost-user performance by delaying the packet copy until a point where we know for certain whether the copy is required or not. This would avoid copying the packet in cases where it is not definitely required. Further investigation is required to determine how much of a performance gain can be achieved.


Tim

^ permalink raw reply	[relevance 2%]

* [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
@ 2016-03-31 13:28 16% Fan Zhang
  0 siblings, 0 replies; 200+ results
From: Fan Zhang @ 2016-03-31 13:28 UTC (permalink / raw)
  To: dev

Several new fields will be added to structure rte_port_source_params for
source port enhancement with pcap file reading support.

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>

---
 doc/guides/rel_notes/deprecation.rst | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 179e30f..9b4fc9d 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -25,3 +25,8 @@ Deprecation Notices
 * 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]
+
+* ABI changes are planned for struct rte_port_source_params in order to
+  support PCAP file reading feature. The release 16.04 contains this ABI
+  change wrapped by RTE_NEXT_ABI macro. Release 16.07 will contain this
+  change, and no backwards compatibility is planned.
-- 
2.5.0

^ permalink raw reply	[relevance 16%]

* [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
@ 2016-03-31 13:29 16% Fan Zhang
  2016-04-05 15:45  7% ` Thomas Monjalon
  2016-04-05 21:16  4% ` Singh, Jasvinder
  0 siblings, 2 replies; 200+ results
From: Fan Zhang @ 2016-03-31 13:29 UTC (permalink / raw)
  To: dev

Several new fields will be added to structure rte_port_source_params for
source port enhancement with pcap file reading support.

Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>

---
 doc/guides/rel_notes/deprecation.rst | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 179e30f..9b4fc9d 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -25,3 +25,8 @@ Deprecation Notices
 * 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]
+
+* ABI changes are planned for struct rte_port_source_params in order to
+  support PCAP file reading feature. The release 16.04 contains this ABI
+  change wrapped by RTE_NEXT_ABI macro. Release 16.07 will contain this
+  change, and no backwards compatibility is planned.
-- 
2.5.0

^ permalink raw reply	[relevance 16%]

* [dpdk-dev] [PATCH v14 6/8] ethdev: redesign link speed config
  @ 2016-03-31 22:12  3%       ` Marc Sune
  0 siblings, 0 replies; 200+ results
From: Marc Sune @ 2016-03-31 22:12 UTC (permalink / raw)
  To: Thomas Monjalon, Xu, Qian Q, Xing, Beilei, dev, Ananyev,
	Konstantin, Lu, Wenzhuo, Richardson, Bruce, Glynn, Michael J
  Cc: Marc Sune

This patch redesigns the API to set the link speed/s configuration
of an ethernet port. Specifically:

- it allows to define a set of advertised speeds for
  auto-negociation.
- it allows to disable link auto-negociation (single fixed speed).
- default: auto-negociate all supported speeds.

A flag autoneg in struct rte_eth_link indicates if link speed was a
result of auto-negociation or was fixed by configuration.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
Tested-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>
Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
 app/test-pmd/cmdline.c                    |  26 ++++----
 doc/guides/rel_notes/deprecation.rst      |   3 -
 doc/guides/rel_notes/release_16_04.rst    |   9 +++
 drivers/net/af_packet/rte_eth_af_packet.c |   1 +
 drivers/net/bnx2x/bnx2x_ethdev.c          |   4 +-
 drivers/net/bonding/rte_eth_bond_8023ad.c |   2 +-
 drivers/net/e1000/em_ethdev.c             | 103 +++++++++++++++---------------
 drivers/net/e1000/igb_ethdev.c            |  95 ++++++++++++++-------------
 drivers/net/i40e/i40e_ethdev.c            |  48 +++++++-------
 drivers/net/i40e/i40e_ethdev_vf.c         |   7 +-
 drivers/net/ixgbe/ixgbe_ethdev.c          |  53 ++++++++-------
 drivers/net/mlx4/mlx4.c                   |   2 +
 drivers/net/mlx5/mlx5_ethdev.c            |   2 +
 drivers/net/mpipe/mpipe_tilegx.c          |   2 +
 drivers/net/null/rte_eth_null.c           |   1 +
 drivers/net/pcap/rte_eth_pcap.c           |   1 +
 drivers/net/ring/rte_eth_ring.c           |   1 +
 drivers/net/szedata2/rte_eth_szedata2.c   |   2 +
 drivers/net/vmxnet3/vmxnet3_ethdev.c      |   1 +
 drivers/net/xenvirt/rte_eth_xenvirt.c     |   1 +
 examples/ip_pipeline/config_parse.c       |   3 +-
 lib/librte_ether/rte_ethdev.h             |  31 +++++----
 22 files changed, 213 insertions(+), 185 deletions(-)

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 815b53b..741cac3 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -989,7 +989,7 @@ struct cmd_config_speed_all {
 };
 
 static int
-parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
+parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint32_t *speed)
 {
 
 	int duplex;
@@ -1006,20 +1006,22 @@ parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
 	}
 
 	if (!strcmp(speedstr, "10")) {
-		*speed = ETH_SPEED_NUM_10M;
+		*speed = (duplex == ETH_LINK_HALF_DUPLEX) ?
+				ETH_LINK_SPEED_10M_HD : ETH_LINK_SPEED_10M;
 	} else if (!strcmp(speedstr, "100")) {
-		*speed = ETH_SPEED_NUM_100M;
+		*speed = (duplex == ETH_LINK_HALF_DUPLEX) ?
+				ETH_LINK_SPEED_100M_HD : ETH_LINK_SPEED_100M;
 	} else {
 		if (duplex != ETH_LINK_FULL_DUPLEX) {
 			printf("Invalid speed/duplex parameters\n");
 			return -1;
 		}
 		if (!strcmp(speedstr, "1000")) {
-			*speed = ETH_SPEED_NUM_1G;
+			*speed = ETH_LINK_SPEED_1G;
 		} else if (!strcmp(speedstr, "10000")) {
-			*speed = ETH_SPEED_NUM_10G;
+			*speed = ETH_LINK_SPEED_10G;
 		} else if (!strcmp(speedstr, "40000")) {
-			*speed = ETH_SPEED_NUM_40G;
+			*speed = ETH_LINK_SPEED_40G;
 		} else if (!strcmp(speedstr, "auto")) {
 			*speed = ETH_LINK_SPEED_AUTONEG;
 		} else {
@@ -1037,8 +1039,7 @@ cmd_config_speed_all_parsed(void *parsed_result,
 			__attribute__((unused)) void *data)
 {
 	struct cmd_config_speed_all *res = parsed_result;
-	uint16_t link_speed = ETH_LINK_SPEED_AUTONEG;
-	uint16_t link_duplex = 0;
+	uint32_t link_speed;
 	portid_t pid;
 
 	if (!all_ports_stopped()) {
@@ -1051,8 +1052,7 @@ cmd_config_speed_all_parsed(void *parsed_result,
 		return;
 
 	FOREACH_PORT(pid, ports) {
-		ports[pid].dev_conf.link_speed = link_speed;
-		ports[pid].dev_conf.link_duplex = link_duplex;
+		ports[pid].dev_conf.link_speeds = link_speed;
 	}
 
 	cmd_reconfig_device_queue(RTE_PORT_ALL, 1, 1);
@@ -1110,8 +1110,7 @@ cmd_config_speed_specific_parsed(void *parsed_result,
 				__attribute__((unused)) void *data)
 {
 	struct cmd_config_speed_specific *res = parsed_result;
-	uint16_t link_speed = ETH_LINK_SPEED_AUTONEG;
-	uint16_t link_duplex = 0;
+	uint32_t link_speed;
 
 	if (!all_ports_stopped()) {
 		printf("Please stop all ports first\n");
@@ -1125,8 +1124,7 @@ cmd_config_speed_specific_parsed(void *parsed_result,
 			&link_speed) < 0)
 		return;
 
-	ports[res->id].dev_conf.link_speed = link_speed;
-	ports[res->id].dev_conf.link_duplex = link_duplex;
+	ports[res->id].dev_conf.link_speeds = link_speed;
 
 	cmd_reconfig_device_queue(RTE_PORT_ALL, 1, 1);
 }
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 179e30f..c47610d 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -19,9 +19,6 @@ Deprecation Notices
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff
 
-* The ethdev structures rte_eth_link, rte_eth_dev_info and rte_eth_conf
-  must be updated to support 100G link and to have a cleaner link speed API.
-
 * 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]
diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index 9e7b0b7..c891a55 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -52,6 +52,12 @@ This section should contain new features added in this release. Sample format:
   The structure ``rte_eth_dev_info`` has now a ``speed_capa`` bitmap, which
   allows the application to know the supported speeds of each device.
 
+* **Added bitmap of link speeds to advertise.**
+
+  Allow defining a set of advertised speeds for auto-negotiation,
+  explicitly disabling link auto-negotiation (single speed)
+  and full auto-negotiation.
+
 * **Added new poll-mode driver for Amazon Elastic Network Adapters (ENA).**
 
   The driver operates variety of ENA adapters through feature negotiation
@@ -464,6 +470,9 @@ This section should contain API changes. Sample format:
 * The ethdev structure ``rte_eth_dev_info`` was changed to support device
   speed capabilities.
 
+* The ethdev structures ``rte_eth_link`` and ``rte_eth_conf`` were changed to
+  support the new link API.
+
 * The functions ``rte_eth_dev_udp_tunnel_add`` and ``rte_eth_dev_udp_tunnel_delete``
   have been renamed into ``rte_eth_dev_udp_tunnel_port_add`` and
   ``rte_eth_dev_udp_tunnel_port_delete``.
diff --git a/drivers/net/af_packet/rte_eth_af_packet.c b/drivers/net/af_packet/rte_eth_af_packet.c
index 641f849..f17bd7e 100644
--- a/drivers/net/af_packet/rte_eth_af_packet.c
+++ b/drivers/net/af_packet/rte_eth_af_packet.c
@@ -119,6 +119,7 @@ static struct rte_eth_link pmd_link = {
 	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
+	.link_autoneg = ETH_LINK_SPEED_AUTONEG
 };
 
 static uint16_t
diff --git a/drivers/net/bnx2x/bnx2x_ethdev.c b/drivers/net/bnx2x/bnx2x_ethdev.c
index 897081f..071b44f 100644
--- a/drivers/net/bnx2x/bnx2x_ethdev.c
+++ b/drivers/net/bnx2x/bnx2x_ethdev.c
@@ -44,9 +44,9 @@ bnx2x_link_update(struct rte_eth_dev *dev)
 		case DUPLEX_HALF:
 			dev->data->dev_link.link_duplex = ETH_LINK_HALF_DUPLEX;
 			break;
-		default:
-			dev->data->dev_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
 	}
+	dev->data->dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
 	dev->data->dev_link.link_status = sc->link_vars.link_up;
 }
 
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index ac8306f..cca7cc3 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -708,7 +708,7 @@ link_speed_key(uint16_t speed) {
 	uint16_t key_speed;
 
 	switch (speed) {
-	case ETH_LINK_SPEED_AUTONEG:
+	case ETH_SPEED_NUM_NONE:
 		key_speed = 0x00;
 		break;
 	case ETH_SPEED_NUM_10M:
diff --git a/drivers/net/e1000/em_ethdev.c b/drivers/net/e1000/em_ethdev.c
index 890c5c3..653be09 100644
--- a/drivers/net/e1000/em_ethdev.c
+++ b/drivers/net/e1000/em_ethdev.c
@@ -538,6 +538,9 @@ eth_em_start(struct rte_eth_dev *dev)
 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
 	int ret, mask;
 	uint32_t intr_vector = 0;
+	uint32_t *speeds;
+	int num_speeds;
+	bool autoneg;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -612,56 +615,46 @@ eth_em_start(struct rte_eth_dev *dev)
 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
 
 	/* Setup link speed and duplex */
-	switch (dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_HALF_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_FULL_DUPLEX;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_HALF;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_FULL;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_100M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_HALF;
-		else if (dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_FULL;
-		else
+	speeds = &dev->data->dev_conf.link_speeds;
+	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
+	} else {
+		num_speeds = 0;
+		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
+
+		/* Reset */
+		hw->phy.autoneg_advertised = 0;
+
+		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
+				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
+			num_speeds = -1;
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_1G:
-		if ((dev->data->dev_conf.link_duplex ==
-				ETH_LINK_AUTONEG_DUPLEX) ||
-			(dev->data->dev_conf.link_duplex ==
-					ETH_LINK_FULL_DUPLEX))
-			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
-		else
+		}
+		if (*speeds & ETH_LINK_SPEED_10M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_10M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_1G) {
+			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
+			num_speeds++;
+		}
+		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10G:
-	default:
-		goto error_invalid_config;
 	}
+
 	e1000_setup_link(hw);
 
 	if (rte_intr_allow_others(intr_handle)) {
@@ -695,9 +688,8 @@ eth_em_start(struct rte_eth_dev *dev)
 	return 0;
 
 error_invalid_config:
-	PMD_INIT_LOG(ERR, "Invalid link_speed/link_duplex (%u/%u) for port %u",
-		     dev->data->dev_conf.link_speed,
-		     dev->data->dev_conf.link_duplex, dev->data->port_id);
+	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
+		     dev->data->dev_conf.link_speeds, dev->data->port_id);
 	em_dev_clear_queues(dev);
 	return -EINVAL;
 }
@@ -1107,13 +1099,20 @@ eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	/* Now we check if a transition has happened */
 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
-		hw->mac.ops.get_link_up_info(hw, &link.link_speed,
-			&link.link_duplex);
+		uint16_t duplex, speed;
+		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
+		link.link_duplex = (duplex == FULL_DUPLEX) ?
+				ETH_LINK_FULL_DUPLEX :
+				ETH_LINK_HALF_DUPLEX;
+		link.link_speed = speed;
 		link.link_status = ETH_LINK_UP;
+		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_FIXED);
 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
 		link.link_speed = 0;
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
 		link.link_status = ETH_LINK_DOWN;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 	rte_em_dev_atomic_write_link_status(dev, &link);
 
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index 95d1711..e0053fe 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -1133,6 +1133,9 @@ eth_igb_start(struct rte_eth_dev *dev)
 	int ret, mask;
 	uint32_t intr_vector = 0;
 	uint32_t ctrl_ext;
+	uint32_t *speeds;
+	int num_speeds;
+	bool autoneg;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -1233,48 +1236,46 @@ eth_igb_start(struct rte_eth_dev *dev)
 	}
 
 	/* Setup link speed and duplex */
-	switch (dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_HALF_DUPLEX;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_FULL_DUPLEX;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_HALF;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_10_FULL;
-		else
-			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_100M:
-		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
-			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_HALF;
-		else if (dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX)
-			hw->phy.autoneg_advertised = ADVERTISE_100_FULL;
-		else
+	speeds = &dev->data->dev_conf.link_speeds;
+	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
+	} else {
+		num_speeds = 0;
+		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
+
+		/* Reset */
+		hw->phy.autoneg_advertised = 0;
+
+		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
+				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
+			num_speeds = -1;
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_1G:
-		if ((dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX) ||
-				(dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX))
-			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
-		else
+		}
+		if (*speeds & ETH_LINK_SPEED_10M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_10M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M_HD) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_100M) {
+			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
+			num_speeds++;
+		}
+		if (*speeds & ETH_LINK_SPEED_1G) {
+			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
+			num_speeds++;
+		}
+		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
 			goto error_invalid_config;
-		break;
-	case ETH_SPEED_NUM_10G:
-	default:
-		goto error_invalid_config;
 	}
+
 	e1000_setup_link(hw);
 
 	if (rte_intr_allow_others(intr_handle)) {
@@ -1306,9 +1307,8 @@ eth_igb_start(struct rte_eth_dev *dev)
 	return 0;
 
 error_invalid_config:
-	PMD_INIT_LOG(ERR, "Invalid link_speed/link_duplex (%u/%u) for port %u",
-		     dev->data->dev_conf.link_speed,
-		     dev->data->dev_conf.link_duplex, dev->data->port_id);
+	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
+		     dev->data->dev_conf.link_speeds, dev->data->port_id);
 	igb_dev_clear_queues(dev);
 	return -EINVAL;
 }
@@ -2061,13 +2061,20 @@ eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	/* Now we check if a transition has happened */
 	if (link_check) {
-		hw->mac.ops.get_link_up_info(hw, &link.link_speed,
-					  &link.link_duplex);
+		uint16_t duplex, speed;
+		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
+		link.link_duplex = (duplex == FULL_DUPLEX) ?
+				ETH_LINK_FULL_DUPLEX :
+				ETH_LINK_HALF_DUPLEX;
+		link.link_speed = speed;
 		link.link_status = ETH_LINK_UP;
+		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_FIXED);
 	} else if (!link_check) {
 		link.link_speed = 0;
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
 		link.link_status = ETH_LINK_DOWN;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 	rte_igb_dev_atomic_write_link_status(dev, &link);
 
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index cde314d..bc28d3c 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -1381,27 +1381,20 @@ i40e_vsi_disable_queues_intr(struct i40e_vsi *vsi)
 }
 
 static inline uint8_t
-i40e_parse_link_speed(uint16_t eth_link_speed)
+i40e_parse_link_speeds(uint16_t link_speeds)
 {
 	uint8_t link_speed = I40E_LINK_SPEED_UNKNOWN;
 
-	switch (eth_link_speed) {
-	case ETH_SPEED_NUM_40G:
-		link_speed = I40E_LINK_SPEED_40GB;
-		break;
-	case ETH_SPEED_NUM_20G:
-		link_speed = I40E_LINK_SPEED_20GB;
-		break;
-	case ETH_SPEED_NUM_10G:
-		link_speed = I40E_LINK_SPEED_10GB;
-		break;
-	case ETH_SPEED_NUM_1G:
-		link_speed = I40E_LINK_SPEED_1GB;
-		break;
-	case ETH_SPEED_NUM_100M:
-		link_speed = I40E_LINK_SPEED_100MB;
-		break;
-	}
+	if (link_speeds & ETH_LINK_SPEED_40G)
+		link_speed |= I40E_LINK_SPEED_40GB;
+	if (link_speeds & ETH_LINK_SPEED_20G)
+		link_speed |= I40E_LINK_SPEED_20GB;
+	if (link_speeds & ETH_LINK_SPEED_10G)
+		link_speed |= I40E_LINK_SPEED_10GB;
+	if (link_speeds & ETH_LINK_SPEED_1G)
+		link_speed |= I40E_LINK_SPEED_1GB;
+	if (link_speeds & ETH_LINK_SPEED_100M)
+		link_speed |= I40E_LINK_SPEED_100MB;
 
 	return link_speed;
 }
@@ -1427,9 +1420,9 @@ i40e_apply_link_speed(struct rte_eth_dev *dev)
 	struct i40e_hw *hw = I40E_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 	struct rte_eth_conf *conf = &dev->data->dev_conf;
 
-	speed = i40e_parse_link_speed(conf->link_speed);
+	speed = i40e_parse_link_speeds(conf->link_speeds);
 	abilities |= I40E_AQ_PHY_ENABLE_ATOMIC_LINK;
-	if (conf->link_speed == ETH_LINK_SPEED_AUTONEG)
+	if (!(conf->link_speeds & ETH_LINK_SPEED_FIXED))
 		abilities |= I40E_AQ_PHY_AN_ENABLED;
 	else
 		abilities |= I40E_AQ_PHY_LINK_ENABLED;
@@ -1449,10 +1442,8 @@ i40e_dev_start(struct rte_eth_dev *dev)
 
 	hw->adapter_stopped = 0;
 
-	if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
-		(dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
-		PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
-			     dev->data->dev_conf.link_duplex,
+	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
+		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; autonegotiation disabled",
 			     dev->data->port_id);
 		return -EINVAL;
 	}
@@ -1525,6 +1516,12 @@ i40e_dev_start(struct rte_eth_dev *dev)
 	}
 
 	/* Apply link configure */
+	if (dev->data->dev_conf.link_speeds & ~(ETH_LINK_SPEED_100M |
+				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G |
+				ETH_LINK_SPEED_20G | ETH_LINK_SPEED_40G)) {
+		PMD_DRV_LOG(ERR, "Invalid link setting");
+		goto err_up;
+	}
 	ret = i40e_apply_link_speed(dev);
 	if (I40E_SUCCESS != ret) {
 		PMD_DRV_LOG(ERR, "Fail to apply link setting");
@@ -1809,6 +1806,9 @@ i40e_dev_link_update(struct rte_eth_dev *dev,
 		break;
 	}
 
+	link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
+
 out:
 	rte_i40e_dev_atomic_write_link_status(dev, &link);
 	if (link.link_status == old.link_status)
diff --git a/drivers/net/i40e/i40e_ethdev_vf.c b/drivers/net/i40e/i40e_ethdev_vf.c
index 295dcd2..8cf22ee 100644
--- a/drivers/net/i40e/i40e_ethdev_vf.c
+++ b/drivers/net/i40e/i40e_ethdev_vf.c
@@ -2121,12 +2121,13 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
 	 * DPDK pf host provide interfacet to acquire link status
 	 * while Linux driver does not
 	 */
-	if (vf->version_major == I40E_DPDK_VERSION_MAJOR)
+	if (vf->version_major == I40E_DPDK_VERSION_MAJOR) {
 		i40evf_get_link_status(dev, &new_link);
-	else {
+	} else {
 		/* Always assume it's up, for Linux driver PF host */
-		new_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
 		new_link.link_speed  = ETH_SPEED_NUM_10G;
+		new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
+		new_link.link_autoneg = ETH_LINK_SPEED_AUTONEG;
 		new_link.link_status = ETH_LINK_UP;
 	}
 	i40evf_dev_atomic_write_link_status(dev, &new_link);
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index 894278f..3f1ebc1 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -2094,14 +2094,16 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 	int mask = 0;
 	int status;
 	uint16_t vf, idx;
+	uint32_t *link_speeds;
 
 	PMD_INIT_FUNC_TRACE();
 
-	/* IXGBE devices don't support half duplex */
-	if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
-			(dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
-		PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
-			     dev->data->dev_conf.link_duplex,
+	/* IXGBE devices don't support:
+	*    - half duplex (checked afterwards for valid speeds)
+	*    - fixed speed: TODO implement
+	*/
+	if (dev->data->dev_conf.link_speeds & ETH_LINK_SPEED_FIXED) {
+		PMD_INIT_LOG(ERR, "Invalid link_speeds for port %hhu; fix speed not supported",
 			     dev->data->port_id);
 		return -EINVAL;
 	}
@@ -2198,30 +2200,25 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 	if (err)
 		goto error;
 
-	switch(dev->data->dev_conf.link_speed) {
-	case ETH_LINK_SPEED_AUTONEG:
+	link_speeds = &dev->data->dev_conf.link_speeds;
+	if (*link_speeds & ~(ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
+			ETH_LINK_SPEED_10G)) {
+		PMD_INIT_LOG(ERR, "Invalid link setting");
+		goto error;
+	}
+
+	speed = 0x0;
+	if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
 		speed = (hw->mac.type != ixgbe_mac_82598EB) ?
 				IXGBE_LINK_SPEED_82599_AUTONEG :
 				IXGBE_LINK_SPEED_82598_AUTONEG;
-		break;
-	case ETH_SPEED_NUM_100M:
-		/*
-		 * Invalid for 82598 but error will be detected by
-		 * ixgbe_setup_link()
-		 */
-		speed = IXGBE_LINK_SPEED_100_FULL;
-		break;
-	case ETH_SPEED_NUM_1G:
-		speed = IXGBE_LINK_SPEED_1GB_FULL;
-		break;
-	case ETH_SPEED_NUM_10G:
-		speed = IXGBE_LINK_SPEED_10GB_FULL;
-		break;
-	default:
-		PMD_INIT_LOG(ERR, "Invalid link_speed (%hu) for port %hhu",
-			     dev->data->dev_conf.link_speed,
-			     dev->data->port_id);
-		goto error;
+	} else {
+		if (*link_speeds & ETH_LINK_SPEED_10G)
+			speed |= IXGBE_LINK_SPEED_10GB_FULL;
+		if (*link_speeds & ETH_LINK_SPEED_1G)
+			speed |= IXGBE_LINK_SPEED_1GB_FULL;
+		if (*link_speeds & ETH_LINK_SPEED_100M)
+			speed |= IXGBE_LINK_SPEED_100_FULL;
 	}
 
 	err = ixgbe_setup_link(hw, speed, link_up);
@@ -3088,7 +3085,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 	if (diag != 0) {
 		link.link_speed = ETH_SPEED_NUM_100M;
-		link.link_duplex = ETH_LINK_HALF_DUPLEX;
+		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
 		if (link.link_status == old.link_status)
 			return -1;
@@ -3107,7 +3104,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 	switch (link_speed) {
 	default:
 	case IXGBE_LINK_SPEED_UNKNOWN:
-		link.link_duplex = ETH_LINK_HALF_DUPLEX;
+		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 
diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c
index ed2f488..4f21dbe 100644
--- a/drivers/net/mlx4/mlx4.c
+++ b/drivers/net/mlx4/mlx4.c
@@ -4736,6 +4736,8 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
 		dev_link.link_speed = link_speed;
 	dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
 				ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+	dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
 	if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
 		/* Link status changed. */
 		dev->data->dev_link = dev_link;
diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
index d7a0eea..beecc63 100644
--- a/drivers/net/mlx5/mlx5_ethdev.c
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -629,6 +629,8 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
 		dev_link.link_speed = link_speed;
 	dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
 				ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+	dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+			ETH_LINK_SPEED_FIXED);
 	if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
 		/* Link status changed. */
 		dev->data->dev_link = dev_link;
diff --git a/drivers/net/mpipe/mpipe_tilegx.c b/drivers/net/mpipe/mpipe_tilegx.c
index 1a77c7a..adcbc19 100644
--- a/drivers/net/mpipe/mpipe_tilegx.c
+++ b/drivers/net/mpipe/mpipe_tilegx.c
@@ -394,6 +394,8 @@ mpipe_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 
 		speed = state & GXIO_MPIPE_LINK_SPEED_MASK;
 
+		new.link_autoneg = (dev->data->dev_conf.link_speeds &
+				ETH_LINK_SPEED_AUTONEG);
 		if (speed == GXIO_MPIPE_LINK_1G) {
 			new.link_speed = ETH_SPEED_NUM_1G;
 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c
index 5640585..5e8e203 100644
--- a/drivers/net/null/rte_eth_null.c
+++ b/drivers/net/null/rte_eth_null.c
@@ -92,6 +92,7 @@ static struct rte_eth_link pmd_link = {
 	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
+	.link_autoneg = ETH_LINK_SPEED_AUTONEG,
 };
 
 static uint16_t
diff --git a/drivers/net/pcap/rte_eth_pcap.c b/drivers/net/pcap/rte_eth_pcap.c
index c657951..c98e234 100644
--- a/drivers/net/pcap/rte_eth_pcap.c
+++ b/drivers/net/pcap/rte_eth_pcap.c
@@ -126,6 +126,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_FIXED,
 };
 
 static int
diff --git a/drivers/net/ring/rte_eth_ring.c b/drivers/net/ring/rte_eth_ring.c
index 58685e9..b1783c3 100644
--- a/drivers/net/ring/rte_eth_ring.c
+++ b/drivers/net/ring/rte_eth_ring.c
@@ -80,6 +80,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_AUTONEG
 };
 
 static uint16_t
diff --git a/drivers/net/szedata2/rte_eth_szedata2.c b/drivers/net/szedata2/rte_eth_szedata2.c
index dd1ae9e..ee97a4e 100644
--- a/drivers/net/szedata2/rte_eth_szedata2.c
+++ b/drivers/net/szedata2/rte_eth_szedata2.c
@@ -1174,6 +1174,8 @@ eth_link_update(struct rte_eth_dev *dev,
 	link.link_status = (cgmii_ibuf_is_enabled(ibuf) &&
 			cgmii_ibuf_is_link_up(ibuf)) ? ETH_LINK_UP : ETH_LINK_DOWN;
 
+	link.link_autoneg = ETH_LINK_SPEED_FIXED;
+
 	rte_atomic64_cmpset((uint64_t *)dev_link, *(uint64_t *)dev_link,
 			*(uint64_t *)link_ptr);
 
diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
index 95d533f..bd7a2bb 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
@@ -779,6 +779,7 @@ vmxnet3_dev_link_update(struct rte_eth_dev *dev, __attribute__((unused)) int wai
 		link.link_status = ETH_LINK_UP;
 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
 		link.link_speed = ETH_SPEED_NUM_10G;
+		link.link_autoneg = ETH_LINK_SPEED_FIXED;
 	}
 
 	vmxnet3_dev_atomic_write_link_status(dev, &link);
diff --git a/drivers/net/xenvirt/rte_eth_xenvirt.c b/drivers/net/xenvirt/rte_eth_xenvirt.c
index 77d3ba1..b9638d9 100644
--- a/drivers/net/xenvirt/rte_eth_xenvirt.c
+++ b/drivers/net/xenvirt/rte_eth_xenvirt.c
@@ -73,6 +73,7 @@ static struct rte_eth_link pmd_link = {
 		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
+		.link_autoneg = ETH_LINK_SPEED_FIXED
 };
 
 static void
diff --git a/examples/ip_pipeline/config_parse.c b/examples/ip_pipeline/config_parse.c
index 152889d..2cd5707 100644
--- a/examples/ip_pipeline/config_parse.c
+++ b/examples/ip_pipeline/config_parse.c
@@ -87,8 +87,7 @@ static const struct app_link_params link_params_default = {
 	.pci_bdf = {0},
 
 	.conf = {
-		.link_speed = 0,
-		.link_duplex = 0,
+		.link_speeds = 0,
 		.rxmode = {
 			.mq_mode = ETH_MQ_RX_NONE,
 
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 49fdcb7..b88300d 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -244,6 +244,8 @@ struct rte_eth_stats {
 /**
  * Device supported speeds bitmap flags
  */
+#define ETH_LINK_SPEED_AUTONEG  (0 <<  0)  /**< Autonegotiate (all speeds) */
+#define ETH_LINK_SPEED_FIXED    (1 <<  0)  /**< Disable autoneg (fixed speed) */
 #define ETH_LINK_SPEED_10M_HD   (1 <<  1)  /**<  10 Mbps half-duplex */
 #define ETH_LINK_SPEED_10M      (1 <<  2)  /**<  10 Mbps full-duplex */
 #define ETH_LINK_SPEED_100M_HD  (1 <<  3)  /**< 100 Mbps half-duplex */
@@ -261,7 +263,7 @@ struct rte_eth_stats {
 /**
  * Ethernet numeric link speeds in Mbps
  */
-#define ETH_LINK_SPEED_AUTONEG     0 /**< Auto-negotiate link speed. */
+#define ETH_SPEED_NUM_NONE         0 /**< Not defined */
 #define ETH_SPEED_NUM_10M         10 /**<  10 Mbps */
 #define ETH_SPEED_NUM_100M       100 /**< 100 Mbps */
 #define ETH_SPEED_NUM_1G        1000 /**<   1 Gbps */
@@ -278,17 +280,19 @@ struct rte_eth_stats {
  * A structure used to retrieve link-level information of an Ethernet port.
  */
 struct rte_eth_link {
-	uint16_t link_speed;      /**< ETH_SPEED_NUM_ */
-	uint16_t link_duplex;     /**< ETH_LINK_[HALF/FULL]_DUPLEX */
-	uint8_t  link_status : 1; /**< ETH_LINK_[DOWN/UP] */
-}__attribute__((aligned(8)));     /**< aligned for atomic64 read/write */
+	uint16_t link_speed;        /**< ETH_SPEED_NUM_ */
+	uint16_t link_duplex  : 1;  /**< ETH_LINK_[HALF/FULL]_DUPLEX */
+	uint16_t link_autoneg : 1;  /**< ETH_LINK_SPEED_[AUTONEG/FIXED] */
+	uint16_t link_status  : 1;  /**< ETH_LINK_[DOWN/UP] */
+} __attribute__((aligned(8)));      /**< aligned for atomic64 read/write */
 
 /* Utility constants */
-#define ETH_LINK_AUTONEG_DUPLEX 0       /**< Auto-negotiate duplex. */
-#define ETH_LINK_HALF_DUPLEX    1       /**< Half-duplex connection. */
-#define ETH_LINK_FULL_DUPLEX    2       /**< Full-duplex connection. */
+#define ETH_LINK_HALF_DUPLEX    0 /**< Half-duplex connection. */
+#define ETH_LINK_FULL_DUPLEX    1 /**< Full-duplex connection. */
 #define ETH_LINK_DOWN           0 /**< Link is down. */
 #define ETH_LINK_UP             1 /**< Link is up. */
+#define ETH_LINK_FIXED          0 /**< No autonegotiation. */
+#define ETH_LINK_AUTONEG        1 /**< Autonegotiated. */
 
 /**
  * A structure used to configure the ring threshold registers of an RX/TX
@@ -802,10 +806,13 @@ struct rte_intr_conf {
  * configuration settings may be needed.
  */
 struct rte_eth_conf {
-	uint16_t link_speed;
-	/**< ETH_SPEED_NUM_ or 0 for autonegotiation */
-	uint16_t link_duplex;
-	/**< ETH_LINK_[HALF_DUPLEX|FULL_DUPLEX], or 0 for autonegotation */
+	uint32_t link_speeds; /**< bitmap of ETH_LINK_SPEED_XXX of speeds to be
+				used. ETH_LINK_SPEED_FIXED disables link
+				autonegotiation, and a unique speed shall be
+				set. Otherwise, the bitmap defines the set of
+				speeds to be advertised. If the special value
+				ETH_LINK_SPEED_AUTONEG (0) is used, all speeds
+				supported are advertised. */
 	struct rte_eth_rxmode rxmode; /**< Port RX configuration. */
 	struct rte_eth_txmode txmode; /**< Port TX configuration. */
 	uint32_t lpbk_mode; /**< Loopback operation mode. By default the value
-- 
2.1.4

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07
  2016-03-17  9:05 15% ` [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07 Olivier Matz
@ 2016-04-04 14:38  4%   ` Thomas Monjalon
  2016-04-05  9:27  4%     ` Hunt, David
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-04 14:38 UTC (permalink / raw)
  To: dev; +Cc: Olivier Matz

2016-03-17 10:05, Olivier Matz:
> Add a deprecation notice for coming changes in mempool for 16.07.
[...]
> +* librte_mempool: new fixes and features will be added in 16.07:
> +  allocation of large mempool in several virtual memory chunks, new API
> +  to populate a mempool, new API to free a mempool, allocation in
> +  anonymous mapping, drop of specific dom0 code. These changes will
> +  induce a modification of the rte_mempool structure, plus a
> +  modification of the API of rte_mempool_obj_iter(), implying a breakage
> +  of the ABI.

Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>

Other people involved in the discussion wanting to bring their support?

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 13:56  4%     ` Wiles, Keith
  2016-03-10 14:55  4%       ` Thomas Monjalon
@ 2016-04-04 14:49  4%       ` Thomas Monjalon
  1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-04 14:49 UTC (permalink / raw)
  To: david.hunt; +Cc: dev, Wiles, Keith, Richardson, Bruce, Olivier MATZ

> >> > Announce the ABI breakage due to addition of external mempool
> >> > manager functionality which requires changes to rte_mempool
> >> > structure.
> >> > 
> >> > Signed-off-by: David Hunt <david.hunt@intel.com>
> >> 
> >> Acked-by: Olivier Matz <olivier.matz@6wind.com>
> >> 
> >Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> Acked-by: John McNamara <john.mcnamara@intel.com>
> Asked-by: Keith Wiles <keith.wiles@intel.com>

Assumed it is an ack.
Applied, thanks

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local.
  @ 2016-04-05  5:47  4%     ` Yuanhan Liu
  2016-04-05  8:37  0%       ` Thomas Monjalon
  2016-04-06  4:14  0%       ` Flavio Leitner
  0 siblings, 2 replies; 200+ results
From: Yuanhan Liu @ 2016-04-05  5:47 UTC (permalink / raw)
  To: Ilya Maximets
  Cc: dev, Dyasly Sergey, Thomas Monjalon, Flavio Leitner, Xie, Huawei

On Fri, Feb 19, 2016 at 03:06:50PM +0800, Yuanhan Liu wrote:
> On Fri, Feb 19, 2016 at 09:32:41AM +0300, Ilya Maximets wrote:
> > Array of buf_vector's is just an array for temporary storing information
> > about available descriptors. It used only locally in virtio_dev_merge_rx()
> > and there is no reason for that array to be shared.
> > 
> > Fix that by allocating local buf_vec inside virtio_dev_merge_rx().
> > 
> > Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
> > ---
> >  lib/librte_vhost/rte_virtio_net.h |  1 -
> >  lib/librte_vhost/vhost_rxtx.c     | 45 ++++++++++++++++++++-------------------
> >  2 files changed, 23 insertions(+), 23 deletions(-)
> > 
> > diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
> > index 10dcb90..ae1e4fb 100644
> > --- a/lib/librte_vhost/rte_virtio_net.h
> > +++ b/lib/librte_vhost/rte_virtio_net.h
> > @@ -91,7 +91,6 @@ struct vhost_virtqueue {
> >  	int			kickfd;			/**< Currently unused as polling mode is enabled. */
> >  	int			enabled;
> >  	uint64_t		reserved[16];		/**< Reserve some spaces for future extension. */
> > -	struct buf_vector	buf_vec[BUF_VECTOR_MAX];	/**< for scatter RX. */
> >  } __rte_cache_aligned;
> 
> I like this kind of cleanup, however, it breaks ABI.

So, I was considering to add vhost-user Tx delayed-copy (or zero copy)
support recently, which comes to yet another ABI violation, as we need
add a new field to virtio_memory_regions struct to do guest phys addr
to host phys addr translation. You may ask, however, that why do we need
expose virtio_memory_regions struct to users at all?

You are right, we don't have to. And here is the thing: we exposed way
too many fields (or even structures) than necessary. Say, vhost_virtqueue
struct should NOT be exposed to user at all: application just need to
tell the right queue id to locate a specific queue, and that's all.
The structure should be defined in an internal header file. With that,
we could do any changes to it we want, without worrying about that we
may offense the painful ABI rules.

Similar changes could be done to virtio_net struct as well, just exposing
very few fields that are necessary and moving all others to an internal
structure.

Huawei then suggested a more radical yet much cleaner one: just exposing
a virtio_net handle to application, just like the way kernel exposes an
fd to user for locating a specific file. However, it's more than an ABI
change; it's also an API change: some fields are referenced by applications,
such as flags, virt_qp_nb. We could expose some new functions to access
them though.

I'd vote for this one, as it sounds very clean to me. This would also
solve the block issue of this patch. Though it would break OVS, I'm thinking
that'd be okay, as OVS has dependence on DPDK version: what we need to
do is just to send few patches to OVS, and let it points to next release,
say DPDK v16.07. Flavio, please correct me if I'm wrong.

Thoughts/comments?

	--yliu

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local.
  2016-04-05  5:47  4%     ` [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local Yuanhan Liu
@ 2016-04-05  8:37  0%       ` Thomas Monjalon
  2016-04-05 14:06  0%         ` Yuanhan Liu
  2016-04-06  4:14  0%       ` Flavio Leitner
  1 sibling, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-05  8:37 UTC (permalink / raw)
  To: Yuanhan Liu
  Cc: dev, Ilya Maximets, Dyasly Sergey, Flavio Leitner, Xie, Huawei

2016-04-05 13:47, Yuanhan Liu:
> So, I was considering to add vhost-user Tx delayed-copy (or zero copy)
> support recently, which comes to yet another ABI violation, as we need
> add a new field to virtio_memory_regions struct to do guest phys addr
> to host phys addr translation. You may ask, however, that why do we need
> expose virtio_memory_regions struct to users at all?
> 
> You are right, we don't have to. And here is the thing: we exposed way
> too many fields (or even structures) than necessary. Say, vhost_virtqueue
> struct should NOT be exposed to user at all: application just need to
> tell the right queue id to locate a specific queue, and that's all.
> The structure should be defined in an internal header file. With that,
> we could do any changes to it we want, without worrying about that we
> may offense the painful ABI rules.
> 
> Similar changes could be done to virtio_net struct as well, just exposing
> very few fields that are necessary and moving all others to an internal
> structure.
> 
> Huawei then suggested a more radical yet much cleaner one: just exposing
> a virtio_net handle to application, just like the way kernel exposes an
> fd to user for locating a specific file. However, it's more than an ABI
> change; it's also an API change: some fields are referenced by applications,
> such as flags, virt_qp_nb. We could expose some new functions to access
> them though.
> 
> I'd vote for this one, as it sounds very clean to me. This would also
> solve the block issue of this patch. Though it would break OVS, I'm thinking
> that'd be okay, as OVS has dependence on DPDK version: what we need to
> do is just to send few patches to OVS, and let it points to next release,
> say DPDK v16.07. Flavio, please correct me if I'm wrong.
> 
> Thoughts/comments?

Do you plan to send a deprecation notice to change API in 16.07?

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH] doc: announce ABI changes for user-owned mempool caches
@ 2016-04-05  9:23  9% Lazaros Koromilas
  2016-04-05 15:42  4% ` Olivier Matz
  0 siblings, 1 reply; 200+ results
From: Lazaros Koromilas @ 2016-04-05  9:23 UTC (permalink / raw)
  To: dev

Deprecation notice for 16.04 for changes targeting release 16.07.
The changes affect struct rte_mempool, rte_mempool_cache and the
mempool API.

Signed-off-by: Lazaros Koromilas <l@nofutznetworks.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 ad31355..6ccabcb 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -40,3 +40,10 @@ Deprecation Notices
   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.
+
+* ABI change for rte_mempool struct to move the cache-related fields
+  to the more appropriate rte_mempool_cache struct. The mempool API is
+  also changed to enable external cache management that is not tied to EAL
+  threads. Some mempool get and put calls are removed in favor of a more
+  compact API. The ones that remain are backwards compatible and use the
+  per-lcore default cache if available. This change targets release 16.07.
-- 
1.9.1

^ permalink raw reply	[relevance 9%]

* Re: [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07
  2016-04-04 14:38  4%   ` Thomas Monjalon
@ 2016-04-05  9:27  4%     ` Hunt, David
  2016-04-05 14:08  4%       ` Wiles, Keith
  0 siblings, 1 reply; 200+ results
From: Hunt, David @ 2016-04-05  9:27 UTC (permalink / raw)
  To: Thomas Monjalon, dev; +Cc: Olivier Matz


On 4/4/2016 3:38 PM, Thomas Monjalon wrote:
> 2016-03-17 10:05, Olivier Matz:
>> Add a deprecation notice for coming changes in mempool for 16.07.
> [...]
>> +* librte_mempool: new fixes and features will be added in 16.07:
>> +  allocation of large mempool in several virtual memory chunks, new API
>> +  to populate a mempool, new API to free a mempool, allocation in
>> +  anonymous mapping, drop of specific dom0 code. These changes will
>> +  induce a modification of the rte_mempool structure, plus a
>> +  modification of the API of rte_mempool_obj_iter(), implying a breakage
>> +  of the ABI.
> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
>
> Other people involved in the discussion wanting to bring their support?

Acked-by: David Hunt<david.hunt@intel.com>


Regards,
David.

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local.
  2016-04-05  8:37  0%       ` Thomas Monjalon
@ 2016-04-05 14:06  0%         ` Yuanhan Liu
  0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-04-05 14:06 UTC (permalink / raw)
  To: Thomas Monjalon
  Cc: dev, Ilya Maximets, Dyasly Sergey, Flavio Leitner, Xie, Huawei

On Tue, Apr 05, 2016 at 10:37:13AM +0200, Thomas Monjalon wrote:
> 2016-04-05 13:47, Yuanhan Liu:
> > So, I was considering to add vhost-user Tx delayed-copy (or zero copy)
> > support recently, which comes to yet another ABI violation, as we need
> > add a new field to virtio_memory_regions struct to do guest phys addr
> > to host phys addr translation. You may ask, however, that why do we need
> > expose virtio_memory_regions struct to users at all?
> > 
> > You are right, we don't have to. And here is the thing: we exposed way
> > too many fields (or even structures) than necessary. Say, vhost_virtqueue
> > struct should NOT be exposed to user at all: application just need to
> > tell the right queue id to locate a specific queue, and that's all.
> > The structure should be defined in an internal header file. With that,
> > we could do any changes to it we want, without worrying about that we
> > may offense the painful ABI rules.
> > 
> > Similar changes could be done to virtio_net struct as well, just exposing
> > very few fields that are necessary and moving all others to an internal
> > structure.
> > 
> > Huawei then suggested a more radical yet much cleaner one: just exposing
> > a virtio_net handle to application, just like the way kernel exposes an
> > fd to user for locating a specific file. However, it's more than an ABI
> > change; it's also an API change: some fields are referenced by applications,
> > such as flags, virt_qp_nb. We could expose some new functions to access
> > them though.
> > 
> > I'd vote for this one, as it sounds very clean to me. This would also
> > solve the block issue of this patch. Though it would break OVS, I'm thinking
> > that'd be okay, as OVS has dependence on DPDK version: what we need to
> > do is just to send few patches to OVS, and let it points to next release,
> > say DPDK v16.07. Flavio, please correct me if I'm wrong.
> > 
> > Thoughts/comments?
> 
> Do you plan to send a deprecation notice to change API in 16.07?

Yes, I planned to, shortly. Before that, I'd ask for comments first.

	--yliu

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice
  2016-03-10 14:55  4%       ` Thomas Monjalon
@ 2016-04-05 14:06  4%         ` Wiles, Keith
  0 siblings, 0 replies; 200+ results
From: Wiles, Keith @ 2016-04-05 14:06 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev, Richardson, Bruce, Olivier MATZ

>2016-03-10 13:56, Wiles, Keith:
>> >On Thu, Mar 10, 2016 at 01:37:27PM +0100, Olivier MATZ wrote:
>> >> Hi David,
>> >> 
>> >> On 03/10/2016 12:55 PM, David Hunt wrote:
>> >> > Announce the ABI breakage due to addition of external mempool
>> >> > manager functionality which requires changes to rte_mempool
>> >> > structure.
>> >> > 
>> >> > Signed-off-by: David Hunt <david.hunt@intel.com>
>> >> 
>> >> Acked-by: Olivier Matz <olivier.matz@6wind.com>
>> >> 
>> >Acked-by: Bruce Richardson <bruce.richardson@intel.com>
>> 
>> Asked-by: Keith Wiles <keith.wiles@intel.com>
>
>Is it on purpose, Keith, or a typo? Do you have asked this notice?

Sorry, autocorrect :-(

Acked-by: Keith Wiles <keith.wiles@intel.com>
>
>


Regards,
Keith





^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07
  2016-04-05  9:27  4%     ` Hunt, David
@ 2016-04-05 14:08  4%       ` Wiles, Keith
  2016-04-05 15:17  4%         ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Wiles, Keith @ 2016-04-05 14:08 UTC (permalink / raw)
  To: Hunt, David, Thomas Monjalon, dev; +Cc: Olivier Matz

>
>On 4/4/2016 3:38 PM, Thomas Monjalon wrote:
>> 2016-03-17 10:05, Olivier Matz:
>>> Add a deprecation notice for coming changes in mempool for 16.07.
>> [...]
>>> +* librte_mempool: new fixes and features will be added in 16.07:
>>> +  allocation of large mempool in several virtual memory chunks, new API
>>> +  to populate a mempool, new API to free a mempool, allocation in
>>> +  anonymous mapping, drop of specific dom0 code. These changes will
>>> +  induce a modification of the rte_mempool structure, plus a
>>> +  modification of the API of rte_mempool_obj_iter(), implying a breakage
>>> +  of the ABI.
>> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
>>
>> Other people involved in the discussion wanting to bring their support?
>
>Acked-by: David Hunt<david.hunt@intel.com>

Acked-by: Keith Wiles <keith.wiles@intel.com>
>
>
>Regards,
>David.
>


Regards,
Keith





^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] DPDK namespace
  @ 2016-04-05 14:13  3% ` Trahe, Fiona
  2016-04-05 14:31  0%   ` Trahe, Fiona
  2016-04-05 14:31  0%   ` Arnon Warshavsky
    1 sibling, 2 replies; 200+ results
From: Trahe, Fiona @ 2016-04-05 14:13 UTC (permalink / raw)
  To: Thomas Monjalon, dev; +Cc: Trahe, Fiona



> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> Sent: Tuesday, April 05, 2016 2:57 PM
> To: dev@dpdk.org
> Subject: [dpdk-dev] DPDK namespace
> 
> DPDK is going to be more popular in Linux distributions.
> It means people will have some DPDK files in their /usr/include and some DPDK
> libraries on their system.
> 
> Let's imagine someone trying to compile an application which needs
> rte_ethdev.h. He has to figure out that this "rte header" is provided by the DPDK.
> Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
> Then someone else will try to run a binary without having installed the DPDK
> libraries. The linker will require libethdev.so (no prefix here).
> StackOverflow will probably have another good answer (among wrong ones):
> "Hey Sherlock Holmes, have you tried to install the DPDK library?"
> Followed by an insight: "You know, the DPDK naming is weird..."
> And we could continue the story with developers having some naming clash
> because of some identifiers not prefixed at all.
> 
> The goal of this email is to get some feedback on how important it is to fix the
> DPDK namespace.
> 
> If there is enough agreement that we should do something, I suggest to
> introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
> during some time.
> We could start using the new prefix for the new APIs (example: crypto) or when
> there is a significant API break (example: mempool).
> 
> Opinions welcome!
I don't have an opinion on how important it is to fix the namespace, though it does seem like a good idea. 
However if it's to be done, in my opinion it should be completed quickly or will just cause more confusion.
So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should follow in next release or two, with 
the resulting ABI compatibility handling. Maybe with dual naming handled for several releases, but a 
clear end date when all are converted.
Else there will be many years with a mix of rte_ and dpdk_ 

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-05 14:13  3% ` Trahe, Fiona
  2016-04-05 14:31  0%   ` Trahe, Fiona
@ 2016-04-05 14:31  0%   ` Arnon Warshavsky
  2016-04-06  5:26  0%     ` Yuanhan Liu
  1 sibling, 1 reply; 200+ results
From: Arnon Warshavsky @ 2016-04-05 14:31 UTC (permalink / raw)
  To: Trahe, Fiona; +Cc: Thomas Monjalon, dev

On Tue, Apr 5, 2016 at 5:13 PM, Trahe, Fiona <fiona.trahe@intel.com> wrote:

>
>
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> > Sent: Tuesday, April 05, 2016 2:57 PM
> > To: dev@dpdk.org
> > Subject: [dpdk-dev] DPDK namespace
> >
> > DPDK is going to be more popular in Linux distributions.
> > It means people will have some DPDK files in their /usr/include and some
> DPDK
> > libraries on their system.
> >
> > Let's imagine someone trying to compile an application which needs
> > rte_ethdev.h. He has to figure out that this "rte header" is provided by
> the DPDK.
> > Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
> > Then someone else will try to run a binary without having installed the
> DPDK
> > libraries. The linker will require libethdev.so (no prefix here).
> > StackOverflow will probably have another good answer (among wrong ones):
> > "Hey Sherlock Holmes, have you tried to install the DPDK library?"
> > Followed by an insight: "You know, the DPDK naming is weird..."
> > And we could continue the story with developers having some naming clash
> > because of some identifiers not prefixed at all.
> >
> > The goal of this email is to get some feedback on how important it is to
> fix the
> > DPDK namespace.
> >
> > If there is enough agreement that we should do something, I suggest to
> > introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
> > during some time.
> > We could start using the new prefix for the new APIs (example: crypto)
> or when
> > there is a significant API break (example: mempool).
> >
> > Opinions welcome!
> I don't have an opinion on how important it is to fix the namespace,
> though it does seem like a good idea.
> However if it's to be done, in my opinion it should be completed quickly
> or will just cause more confusion.
> So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should
> follow in next release or two, with
> the resulting ABI compatibility handling. Maybe with dual naming handled
> for several releases, but a
> clear end date when all are converted.
> Else there will be many years with a mix of rte_ and dpdk_
>
>

Googling rte functions or error codes usually takes you to dpdk dev email
archive so I don't think it is that much difficult to figure out where rte
comes from.
Other than that , except for my own refactoring pains when replacing a dpdk
version, I do not see a major reason why not.
If Going for dpdk_ prefix, I agree with the quick death approach.

/Arnon

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-05 14:13  3% ` Trahe, Fiona
@ 2016-04-05 14:31  0%   ` Trahe, Fiona
  2016-04-05 14:31  0%   ` Arnon Warshavsky
  1 sibling, 0 replies; 200+ results
From: Trahe, Fiona @ 2016-04-05 14:31 UTC (permalink / raw)
  To: Thomas Monjalon, dev; +Cc: Trahe, Fiona



> -----Original Message-----
> From: Trahe, Fiona
> Sent: Tuesday, April 05, 2016 3:13 PM
> To: Thomas Monjalon; dev@dpdk.org
> Cc: Trahe, Fiona
> Subject: RE: [dpdk-dev] DPDK namespace
> 
> 
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> > Sent: Tuesday, April 05, 2016 2:57 PM
> > To: dev@dpdk.org
> > Subject: [dpdk-dev] DPDK namespace
> >
> > DPDK is going to be more popular in Linux distributions.
> > It means people will have some DPDK files in their /usr/include and
> > some DPDK libraries on their system.
> >
> > Let's imagine someone trying to compile an application which needs
> > rte_ethdev.h. He has to figure out that this "rte header" is provided by the
> DPDK.
> > Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
> > Then someone else will try to run a binary without having installed
> > the DPDK libraries. The linker will require libethdev.so (no prefix here).
> > StackOverflow will probably have another good answer (among wrong ones):
> > "Hey Sherlock Holmes, have you tried to install the DPDK library?"
> > Followed by an insight: "You know, the DPDK naming is weird..."
> > And we could continue the story with developers having some naming
> > clash because of some identifiers not prefixed at all.
> >
> > The goal of this email is to get some feedback on how important it is
> > to fix the DPDK namespace.
> >
> > If there is enough agreement that we should do something, I suggest to
> > introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
> > during some time.
> > We could start using the new prefix for the new APIs (example: crypto)
> > or when there is a significant API break (example: mempool).
> >
> > Opinions welcome!
> I don't have an opinion on how important it is to fix the namespace, though it
> does seem like a good idea.
> However if it's to be done, in my opinion it should be completed quickly or will
> just cause more confusion.
> So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should follow in
> next release or two, with the resulting ABI compatibility handling. Maybe with
> dual naming handled for several releases, but a clear end date when all are
> converted.
> Else there will be many years with a mix of rte_ and dpdk_

An alternative: Would it not be better to do this as one specific 
namespace-change-only release, e.g. an extra 16.06 release, rather than bundling with functional changes?

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07
  2016-04-05 14:08  4%       ` Wiles, Keith
@ 2016-04-05 15:17  4%         ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-05 15:17 UTC (permalink / raw)
  To: Olivier Matz; +Cc: Wiles, Keith, Hunt, David, dev

> >>> Add a deprecation notice for coming changes in mempool for 16.07.
> >> [...]
> >>> +* librte_mempool: new fixes and features will be added in 16.07:
> >>> +  allocation of large mempool in several virtual memory chunks, new API
> >>> +  to populate a mempool, new API to free a mempool, allocation in
> >>> +  anonymous mapping, drop of specific dom0 code. These changes will
> >>> +  induce a modification of the rte_mempool structure, plus a
> >>> +  modification of the API of rte_mempool_obj_iter(), implying a breakage
> >>> +  of the ABI.
> >> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> >>
> >> Other people involved in the discussion wanting to bring their support?
> >
> >Acked-by: David Hunt<david.hunt@intel.com>
> 
> Acked-by: Keith Wiles <keith.wiles@intel.com>

Applied, thanks

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH v3] PCI: ABI change request for adding new field in rte_pci_id structure
  @ 2016-04-05 15:31  4%     ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-05 15:31 UTC (permalink / raw)
  To: Ziye Yang; +Cc: dev, Bruce Richardson

> > The purpose of this patch is used to add a new field
> > "class" in rte_pci_id structure. The new class field includes
> > class_id, subcalss_id, programming interface of a pci device.
> > With this field, we can identify pci device by its class info,
> > which can be more flexible instead of probing the device by
> > vendor_id OR device_id OR subvendor_id OR subdevice_id.
> > For example, we can probe all nvme devices by class field, which
> > can be quite convenient.
> > 
> > Signed-off-by: Ziye Yang <ziye.yang@intel.com>
> 
> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> Acked-by: Helin Zhang <helin.zhang@intel.com>
> Acked-by: Cunming Liang <cunming.liang@intel.com>

Applied, thanks

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI changes for user-owned mempool caches
  2016-04-05  9:23  9% [dpdk-dev] [PATCH] doc: announce ABI changes for user-owned mempool caches Lazaros Koromilas
@ 2016-04-05 15:42  4% ` Olivier Matz
  2016-04-08 14:01  4%   ` Hunt, David
  0 siblings, 1 reply; 200+ results
From: Olivier Matz @ 2016-04-05 15:42 UTC (permalink / raw)
  To: Lazaros Koromilas, dev


On 04/05/2016 11:23 AM, Lazaros Koromilas wrote:
> Deprecation notice for 16.04 for changes targeting release 16.07.
> The changes affect struct rte_mempool, rte_mempool_cache and the
> mempool API.
> 
> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>

Acked-by: Olivier Matz <olivier.matz@6wind.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
  2016-03-31 13:29 16% Fan Zhang
@ 2016-04-05 15:45  7% ` Thomas Monjalon
  2016-04-05 21:16  4% ` Singh, Jasvinder
  1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-05 15:45 UTC (permalink / raw)
  To: dev; +Cc: Fan Zhang

2016-03-31 14:29, Fan Zhang:
> Several new fields will be added to structure rte_port_source_params for
> source port enhancement with pcap file reading support.
> 
> Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>

Anyone interested or against this ABI break in rte_port?
It will be accepted when having 3 acks without any nack.

^ permalink raw reply	[relevance 7%]

* [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
@ 2016-04-05 17:58  9% Harry van Haaren
  2016-04-05 18:45  0% ` Thomas Monjalon
  2016-04-06 14:00  0% ` David Harton (dharton)
  0 siblings, 2 replies; 200+ results
From: Harry van Haaren @ 2016-04-05 17:58 UTC (permalink / raw)
  To: dev; +Cc: maryam.tahhan, Harry van Haaren

This patch adds a notice that the API for the xstats
functionality will be modified in the 16.07 release, with
no backwards compatibility planned as it would require
code duplication in each PMD that supports xstats.

Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
---
 doc/guides/rel_notes/deprecation.rst | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 98d5529..13c3a95 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -54,3 +54,8 @@ Deprecation Notices
   induce a modification of the rte_mempool structure, plus a
   modification of the API of rte_mempool_obj_iter(), implying a breakage
   of the ABI.
+
+* ABI change is planned for the xstats API and rte_eth_xstats struct, to
+  facilitate updating to an API that allows retrieval of values without any
+  string copies or parsing. No backwards compatibility is planned, as it would
+  require code duplication in every PMD that supports xstats.
-- 
2.5.0

^ permalink raw reply	[relevance 9%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-05 17:58  9% [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07 Harry van Haaren
@ 2016-04-05 18:45  0% ` Thomas Monjalon
  2016-04-06  9:02  0%   ` Van Haaren, Harry
  2016-04-06 14:00  0% ` David Harton (dharton)
  1 sibling, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-05 18:45 UTC (permalink / raw)
  To: Harry van Haaren; +Cc: dev, maryam.tahhan

2016-04-05 18:58, Harry van Haaren:
> +* ABI change is planned for the xstats API and rte_eth_xstats struct, to
> +  facilitate updating to an API that allows retrieval of values without any
> +  string copies or parsing. No backwards compatibility is planned, as it would
> +  require code duplication in every PMD that supports xstats.

Have you already submitted a RFC patch to let us have an opinion on the change?
We need, at least, to see the structure changes.
Thanks

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
  2016-03-31 13:29 16% Fan Zhang
  2016-04-05 15:45  7% ` Thomas Monjalon
@ 2016-04-05 21:16  4% ` Singh, Jasvinder
  2016-04-06  8:51  4%   ` Azarewicz, PiotrX T
  1 sibling, 1 reply; 200+ results
From: Singh, Jasvinder @ 2016-04-05 21:16 UTC (permalink / raw)
  To: Zhang, Roy Fan, dev

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Fan Zhang
> Sent: Thursday, March 31, 2016 2:29 PM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH] doc: announce ABI change for
> rte_port_source_params structure
> 
> Several new fields will be added to structure rte_port_source_params for
> source port enhancement with pcap file reading support.
> 
> Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>

Acked-by: Jasvinder Singh <jasvinder.singh@intel.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local.
  2016-04-05  5:47  4%     ` [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local Yuanhan Liu
  2016-04-05  8:37  0%       ` Thomas Monjalon
@ 2016-04-06  4:14  0%       ` Flavio Leitner
  1 sibling, 0 replies; 200+ results
From: Flavio Leitner @ 2016-04-06  4:14 UTC (permalink / raw)
  To: Yuanhan Liu
  Cc: Ilya Maximets, dev, Dyasly Sergey, Thomas Monjalon, Xie, Huawei

On Tue, Apr 05, 2016 at 01:47:33PM +0800, Yuanhan Liu wrote:
> On Fri, Feb 19, 2016 at 03:06:50PM +0800, Yuanhan Liu wrote:
> > On Fri, Feb 19, 2016 at 09:32:41AM +0300, Ilya Maximets wrote:
> > > Array of buf_vector's is just an array for temporary storing information
> > > about available descriptors. It used only locally in virtio_dev_merge_rx()
> > > and there is no reason for that array to be shared.
> > > 
> > > Fix that by allocating local buf_vec inside virtio_dev_merge_rx().
> > > 
> > > Signed-off-by: Ilya Maximets <i.maximets@samsung.com>
> > > ---
> > >  lib/librte_vhost/rte_virtio_net.h |  1 -
> > >  lib/librte_vhost/vhost_rxtx.c     | 45 ++++++++++++++++++++-------------------
> > >  2 files changed, 23 insertions(+), 23 deletions(-)
> > > 
> > > diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
> > > index 10dcb90..ae1e4fb 100644
> > > --- a/lib/librte_vhost/rte_virtio_net.h
> > > +++ b/lib/librte_vhost/rte_virtio_net.h
> > > @@ -91,7 +91,6 @@ struct vhost_virtqueue {
> > >  	int			kickfd;			/**< Currently unused as polling mode is enabled. */
> > >  	int			enabled;
> > >  	uint64_t		reserved[16];		/**< Reserve some spaces for future extension. */
> > > -	struct buf_vector	buf_vec[BUF_VECTOR_MAX];	/**< for scatter RX. */
> > >  } __rte_cache_aligned;
> > 
> > I like this kind of cleanup, however, it breaks ABI.
> 
> So, I was considering to add vhost-user Tx delayed-copy (or zero copy)
> support recently, which comes to yet another ABI violation, as we need
> add a new field to virtio_memory_regions struct to do guest phys addr
> to host phys addr translation. You may ask, however, that why do we need
> expose virtio_memory_regions struct to users at all?
> 
> You are right, we don't have to. And here is the thing: we exposed way
> too many fields (or even structures) than necessary. Say, vhost_virtqueue
> struct should NOT be exposed to user at all: application just need to
> tell the right queue id to locate a specific queue, and that's all.
> The structure should be defined in an internal header file. With that,
> we could do any changes to it we want, without worrying about that we
> may offense the painful ABI rules.
> 
> Similar changes could be done to virtio_net struct as well, just exposing
> very few fields that are necessary and moving all others to an internal
> structure.
> 
> Huawei then suggested a more radical yet much cleaner one: just exposing
> a virtio_net handle to application, just like the way kernel exposes an
> fd to user for locating a specific file. However, it's more than an ABI
> change; it's also an API change: some fields are referenced by applications,
> such as flags, virt_qp_nb. We could expose some new functions to access
> them though.
> 
> I'd vote for this one, as it sounds very clean to me. This would also
> solve the block issue of this patch. Though it would break OVS, I'm thinking
> that'd be okay, as OVS has dependence on DPDK version: what we need to
> do is just to send few patches to OVS, and let it points to next release,
> say DPDK v16.07. Flavio, please correct me if I'm wrong.

There is a plan to use vHost PMD, so from OVS point of view the virtio
stuff would be hidden because vhost PMD would look like just as a
regular ethernet, right?

I think we are waiting for 16.04 to be released with that so we can
start push changes to OVS as well.

-- 
fbl

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local.
@ 2016-04-06  5:11  0% Ilya Maximets
  0 siblings, 0 replies; 200+ results
From: Ilya Maximets @ 2016-04-06  5:11 UTC (permalink / raw)
  To: Flavio Leitner, Yuanhan Liu
  Cc: dev, Sergey Dyasly, Thomas Monjalon, Xie, Huawei

------- Original Message -------
Sender : Flavio Leitner<fbl@sysclose.org>
Date : Apr 06, 2016 07:14 (GMT+03:00)
Title : Re: [RFC] vhost-user public struct refactor (was Re: [dpdk-dev] [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local.

On Tue, Apr 05, 2016 at 01:47:33PM +0800, Yuanhan Liu wrote:
> On Fri, Feb 19, 2016 at 03:06:50PM +0800, Yuanhan Liu wrote:
> > On Fri, Feb 19, 2016 at 09:32:41AM +0300, Ilya Maximets wrote:
> > > Array of buf_vector's is just an array for temporary storing information
> > > about available descriptors. It used only locally in virtio_dev_merge_rx()
> > > and there is no reason for that array to be shared.
> > > 
> > > Fix that by allocating local buf_vec inside virtio_dev_merge_rx().
> > > 
> > > Signed-off-by: Ilya Maximets 
> > > ---
> > >  lib/librte_vhost/rte_virtio_net.h |  1 -
> > >  lib/librte_vhost/vhost_rxtx.c     | 45 ++++++++++++++++++++-------------------
> > >  2 files changed, 23 insertions(+), 23 deletions(-)
> > > 
> > > diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
> > > index 10dcb90..ae1e4fb 100644
> > > --- a/lib/librte_vhost/rte_virtio_net.h
> > > +++ b/lib/librte_vhost/rte_virtio_net.h
> > > @@ -91,7 +91,6 @@ struct vhost_virtqueue {
> > >   int kickfd; /**< Currently unused as polling mode is enabled. */
> > >   int enabled;
> > >   uint64_t reserved[16]; /**< Reserve some spaces for future extension. */
> > > - struct buf_vector buf_vec[BUF_VECTOR_MAX]; /**< for scatter RX. */
> > >  } __rte_cache_aligned;
> > 
> > I like this kind of cleanup, however, it breaks ABI.
> 
> So, I was considering to add vhost-user Tx delayed-copy (or zero copy)
> support recently, which comes to yet another ABI violation, as we need
> add a new field to virtio_memory_regions struct to do guest phys addr
> to host phys addr translation. You may ask, however, that why do we need
> expose virtio_memory_regions struct to users at all?
> 
> You are right, we don't have to. And here is the thing: we exposed way
> too many fields (or even structures) than necessary. Say, vhost_virtqueue
> struct should NOT be exposed to user at all: application just need to
> tell the right queue id to locate a specific queue, and that's all.
> The structure should be defined in an internal header file. With that,
> we could do any changes to it we want, without worrying about that we
> may offense the painful ABI rules.
> 
> Similar changes could be done to virtio_net struct as well, just exposing
> very few fields that are necessary and moving all others to an internal
> structure.
> 
> Huawei then suggested a more radical yet much cleaner one: just exposing
> a virtio_net handle to application, just like the way kernel exposes an
> fd to user for locating a specific file. However, it's more than an ABI
> change; it's also an API change: some fields are referenced by applications,
> such as flags, virt_qp_nb. We could expose some new functions to access
> them though.
> 
> I'd vote for this one, as it sounds very clean to me. This would also
> solve the block issue of this patch. Though it would break OVS, I'm thinking
> that'd be okay, as OVS has dependence on DPDK version: what we need to
> do is just to send few patches to OVS, and let it points to next release,
> say DPDK v16.07. Flavio, please correct me if I'm wrong.

> There is a plan to use vHost PMD, so from OVS point of view the virtio
> stuff would be hidden because vhost PMD would look like just as a
> regular ethernet, right?

But we still need to have access to virtqueue_enabe/disable notifications to
work properly. How this will be done if virtqueue will be hidden from user?

Best regards, Ilya Maximets.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-05 14:31  0%   ` Arnon Warshavsky
@ 2016-04-06  5:26  0%     ` Yuanhan Liu
  2016-04-06 12:07  0%       ` Panu Matilainen
  0 siblings, 1 reply; 200+ results
From: Yuanhan Liu @ 2016-04-06  5:26 UTC (permalink / raw)
  To: Arnon Warshavsky; +Cc: Trahe, Fiona, Thomas Monjalon, dev

On Tue, Apr 05, 2016 at 05:31:22PM +0300, Arnon Warshavsky wrote:
> On Tue, Apr 5, 2016 at 5:13 PM, Trahe, Fiona <fiona.trahe@intel.com> wrote:
> 
> >
> >
> > > -----Original Message-----
> > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> > > Sent: Tuesday, April 05, 2016 2:57 PM
> > > To: dev@dpdk.org
> > > Subject: [dpdk-dev] DPDK namespace
> > >
> > > DPDK is going to be more popular in Linux distributions.
> > > It means people will have some DPDK files in their /usr/include and some
> > DPDK
> > > libraries on their system.
> > >
> > > Let's imagine someone trying to compile an application which needs
> > > rte_ethdev.h. He has to figure out that this "rte header" is provided by
> > the DPDK.
> > > Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
> > > Then someone else will try to run a binary without having installed the
> > DPDK
> > > libraries. The linker will require libethdev.so (no prefix here).
> > > StackOverflow will probably have another good answer (among wrong ones):
> > > "Hey Sherlock Holmes, have you tried to install the DPDK library?"
> > > Followed by an insight: "You know, the DPDK naming is weird..."
> > > And we could continue the story with developers having some naming clash
> > > because of some identifiers not prefixed at all.
> > >
> > > The goal of this email is to get some feedback on how important it is to
> > fix the
> > > DPDK namespace.
> > >
> > > If there is enough agreement that we should do something, I suggest to
> > > introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
> > > during some time.
> > > We could start using the new prefix for the new APIs (example: crypto)
> > or when
> > > there is a significant API break (example: mempool).
> > >
> > > Opinions welcome!
> > I don't have an opinion on how important it is to fix the namespace,
> > though it does seem like a good idea.
> > However if it's to be done, in my opinion it should be completed quickly
> > or will just cause more confusion.
> > So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should
> > follow in next release or two, with
> > the resulting ABI compatibility handling. Maybe with dual naming handled
> > for several releases, but a
> > clear end date when all are converted.
> > Else there will be many years with a mix of rte_ and dpdk_
> >
> >
> 
> Googling rte functions or error codes usually takes you to dpdk dev email
> archive so I don't think it is that much difficult to figure out where rte
> comes from.
> Other than that , except for my own refactoring pains when replacing a dpdk
> version, I do not see a major reason why not.
> If Going for dpdk_ prefix, I agree with the quick death approach.

+1: it's a bit weird to keep both, especially for a long while, that
every time we turn a rte_ prefix to dpdk_ prefix, we break applications.
Instead of breaking applications many times, I'd prefer to break once.
Therefore, applications could do a simple global rte_ -> dpdk_ substitute:
it doesn't sound that painful then.

And here are few more comments:

- we should add rte_/dpdk_ prefix to all public structures as well.

  I'm thinking we are doing well here. I'm just aware that vhost lib
  does a bad job, which is something I proposed to fix in next release.

- If we do the whole change once, I'd suggest to do it ASAP when this
  release is over.

  It should be a HUGE change that touches a lot of code, if we do it
  later, at a stage that a lot of patches for new features have been
  made or sent out, all of them need rebase. That'd be painful.

	--yliu

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH] vhost: ABI/API change announcement due to refactor
@ 2016-04-06  6:53 15% Yuanhan Liu
  2016-04-07  7:12  7% ` Panu Matilainen
  0 siblings, 1 reply; 200+ results
From: Yuanhan Liu @ 2016-04-06  6:53 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Thomas Monjalon, Yuanhan Liu, Ilya Maximets

We currently exposed way too many fields (or even structures) than
necessary. For example, vhost_virtqueue struct should NOT be exposed
to user at all: application just need to tell the right queue id to
locate a specific queue, and that's all. Instead, the structure should
be defined in an internal header file. With that, we could do any changes
to it we want, without worrying about that we may offense the painful
ABI rules.

Similar changes could be done to virtio_net struct as well, just exposing
very few fields that are necessary and moving all others to an internal
structure.

Huawei then suggested a more radical yet much cleaner one: just exposing
a virtio_net handle to application, just like the way kernel exposes an
fd to user for locating a specific file, and exposing some new functions
to access those old fields, such as flags, virt_qp_nb.

With this change, we're likely to be free from ABI violations forever
(well, except when we have to extend the virtio_net_device_ops struct).
For example, following nice cleanup would not be a blocking one then:

    http://dpdk.org/ml/archives/dev/2016-February/033528.html

Suggested-by: Huawei Xie <huawei.xie@intel.com>
Cc: Ilya Maximets <i.maximets@samsung.com>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.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 ad31355..7d16d86 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -40,3 +40,10 @@ Deprecation Notices
   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
+  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.
-- 
1.9.0

^ permalink raw reply	[relevance 15%]

* Re: [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants
  @ 2016-04-06  8:34  1%       ` Weglicki, MichalX
  2016-04-06  8:52  0%         ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Weglicki, MichalX @ 2016-04-06  8:34 UTC (permalink / raw)
  To: Marc Sune, Thomas Monjalon, Xu, Qian Q, Xing, Beilei, dev,
	Ananyev, Konstantin, Lu, Wenzhuo, Richardson, Bruce, Glynn,
	Michael J, Gray, Mark D

Hello, 

I have a question about this patch. 

As far as I see changing ETH_LINK_SPEED_ to ETH_SPEED_NUM_ is rather cosmetic change, am I right? 

Disadvantage of that is that it breaks compatibility with OVS (compilation) and thus it also breaks backward compatibility between OVS & DPDK. Is it really necessary? 

It would be great to keep ABI & API stable if possible. 

Br, 
Michal. 

-----Original Message-----
From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Marc Sune
Sent: Saturday, March 26, 2016 1:27 AM
To: Thomas Monjalon <thomas.monjalon@6wind.com>; Xu, Qian Q <qian.q.xu@intel.com>; Xing, Beilei <beilei.xing@intel.com>; dev@dpdk.org; Ananyev, Konstantin <konstantin.ananyev@intel.com>; Lu, Wenzhuo <wenzhuo.lu@intel.com>; Richardson, Bruce <bruce.richardson@intel.com>; Glynn, Michael J <michael.j.glynn@intel.com>
Cc: Marc Sune <marcdevel@gmail.com>
Subject: [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants

The speed numbers ETH_LINK_SPEED_ are renamed ETH_SPEED_NUM_.
The prefix ETH_LINK_SPEED_ is kept for AUTONEG and will be used
for bit flags in next patch.

Signed-off-by: Marc Sune <marcdevel@gmail.com>
---
 app/test-pmd/cmdline.c                    | 10 +++++-----
 app/test/virtual_pmd.c                    |  2 +-
 drivers/net/af_packet/rte_eth_af_packet.c |  2 +-
 drivers/net/bonding/rte_eth_bond_8023ad.c | 12 ++++++------
 drivers/net/cxgbe/base/t4_hw.c            |  8 ++++----
 drivers/net/e1000/em_ethdev.c             |  8 ++++----
 drivers/net/e1000/igb_ethdev.c            |  8 ++++----
 drivers/net/ena/ena_ethdev.c              |  2 +-
 drivers/net/i40e/i40e_ethdev.c            | 30 +++++++++++++++---------------
 drivers/net/i40e/i40e_ethdev_vf.c         |  2 +-
 drivers/net/ixgbe/ixgbe_ethdev.c          | 22 +++++++++++-----------
 drivers/net/mpipe/mpipe_tilegx.c          |  4 ++--
 drivers/net/nfp/nfp_net.c                 |  2 +-
 drivers/net/null/rte_eth_null.c           |  2 +-
 drivers/net/pcap/rte_eth_pcap.c           |  2 +-
 drivers/net/ring/rte_eth_ring.c           |  2 +-
 drivers/net/szedata2/rte_eth_szedata2.c   |  8 ++++----
 drivers/net/vmxnet3/vmxnet3_ethdev.c      |  2 +-
 drivers/net/xenvirt/rte_eth_xenvirt.c     |  2 +-
 lib/librte_ether/rte_ethdev.h             | 29 ++++++++++++++++++-----------
 20 files changed, 83 insertions(+), 76 deletions(-)

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index eb7bbb4..815b53b 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -1006,20 +1006,20 @@ parse_and_check_speed_duplex(char *speedstr, char *duplexstr, uint16_t *speed)
 	}
 
 	if (!strcmp(speedstr, "10")) {
-		*speed = ETH_LINK_SPEED_10;
+		*speed = ETH_SPEED_NUM_10M;
 	} else if (!strcmp(speedstr, "100")) {
-		*speed = ETH_LINK_SPEED_100;
+		*speed = ETH_SPEED_NUM_100M;
 	} else {
 		if (duplex != ETH_LINK_FULL_DUPLEX) {
 			printf("Invalid speed/duplex parameters\n");
 			return -1;
 		}
 		if (!strcmp(speedstr, "1000")) {
-			*speed = ETH_LINK_SPEED_1000;
+			*speed = ETH_SPEED_NUM_1G;
 		} else if (!strcmp(speedstr, "10000")) {
-			*speed = ETH_LINK_SPEED_10G;
+			*speed = ETH_SPEED_NUM_10G;
 		} else if (!strcmp(speedstr, "40000")) {
-			*speed = ETH_LINK_SPEED_40G;
+			*speed = ETH_SPEED_NUM_40G;
 		} else if (!strcmp(speedstr, "auto")) {
 			*speed = ETH_LINK_SPEED_AUTONEG;
 		} else {
diff --git a/app/test/virtual_pmd.c b/app/test/virtual_pmd.c
index b1d40d7..b4bd2f2 100644
--- a/app/test/virtual_pmd.c
+++ b/app/test/virtual_pmd.c
@@ -604,7 +604,7 @@ virtual_ethdev_create(const char *name, struct ether_addr *mac_addr,
 	TAILQ_INIT(&(eth_dev->link_intr_cbs));
 
 	eth_dev->data->dev_link.link_status = ETH_LINK_DOWN;
-	eth_dev->data->dev_link.link_speed = ETH_LINK_SPEED_10000;
+	eth_dev->data->dev_link.link_speed = ETH_SPEED_NUM_10G;
 	eth_dev->data->dev_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 
 	eth_dev->data->mac_addrs = rte_zmalloc(name, ETHER_ADDR_LEN, 0);
diff --git a/drivers/net/af_packet/rte_eth_af_packet.c b/drivers/net/af_packet/rte_eth_af_packet.c
index dee7b59..641f849 100644
--- a/drivers/net/af_packet/rte_eth_af_packet.c
+++ b/drivers/net/af_packet/rte_eth_af_packet.c
@@ -116,7 +116,7 @@ static const char *valid_arguments[] = {
 static const char *drivername = "AF_PACKET PMD";
 
 static struct rte_eth_link pmd_link = {
-	.link_speed = 10000,
+	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
 };
diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c b/drivers/net/bonding/rte_eth_bond_8023ad.c
index 1b7e93a..ac8306f 100644
--- a/drivers/net/bonding/rte_eth_bond_8023ad.c
+++ b/drivers/net/bonding/rte_eth_bond_8023ad.c
@@ -711,22 +711,22 @@ link_speed_key(uint16_t speed) {
 	case ETH_LINK_SPEED_AUTONEG:
 		key_speed = 0x00;
 		break;
-	case ETH_LINK_SPEED_10:
+	case ETH_SPEED_NUM_10M:
 		key_speed = BOND_LINK_SPEED_KEY_10M;
 		break;
-	case ETH_LINK_SPEED_100:
+	case ETH_SPEED_NUM_100M:
 		key_speed = BOND_LINK_SPEED_KEY_100M;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		key_speed = BOND_LINK_SPEED_KEY_1000M;
 		break;
-	case ETH_LINK_SPEED_10G:
+	case ETH_SPEED_NUM_10G:
 		key_speed = BOND_LINK_SPEED_KEY_10G;
 		break;
-	case ETH_LINK_SPEED_20G:
+	case ETH_SPEED_NUM_20G:
 		key_speed = BOND_LINK_SPEED_KEY_20G;
 		break;
-	case ETH_LINK_SPEED_40G:
+	case ETH_SPEED_NUM_40G:
 		key_speed = BOND_LINK_SPEED_KEY_40G;
 		break;
 	default:
diff --git a/drivers/net/cxgbe/base/t4_hw.c b/drivers/net/cxgbe/base/t4_hw.c
index 884d2cf..79af806 100644
--- a/drivers/net/cxgbe/base/t4_hw.c
+++ b/drivers/net/cxgbe/base/t4_hw.c
@@ -2159,13 +2159,13 @@ int t4_handle_fw_rpl(struct adapter *adap, const __be64 *rpl)
 		if (stat & F_FW_PORT_CMD_TXPAUSE)
 			fc |= PAUSE_TX;
 		if (stat & V_FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_100M))
-			speed = ETH_LINK_SPEED_100;
+			speed = ETH_SPEED_NUM_100M;
 		else if (stat & V_FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_1G))
-			speed = ETH_LINK_SPEED_1000;
+			speed = ETH_SPEED_NUM_1G;
 		else if (stat & V_FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_10G))
-			speed = ETH_LINK_SPEED_10000;
+			speed = ETH_SPEED_NUM_10G;
 		else if (stat & V_FW_PORT_CMD_LSPEED(FW_PORT_CAP_SPEED_40G))
-			speed = ETH_LINK_SPEED_40G;
+			speed = ETH_SPEED_NUM_40G;
 
 		for_each_port(adap, i) {
 			pi = adap2pinfo(adap, i);
diff --git a/drivers/net/e1000/em_ethdev.c b/drivers/net/e1000/em_ethdev.c
index fad8f2f..473d77f 100644
--- a/drivers/net/e1000/em_ethdev.c
+++ b/drivers/net/e1000/em_ethdev.c
@@ -625,7 +625,7 @@ eth_em_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_10:
+	case ETH_SPEED_NUM_10M:
 		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
 			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
 		else if (dev->data->dev_conf.link_duplex ==
@@ -637,7 +637,7 @@ eth_em_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_100:
+	case ETH_SPEED_NUM_100M:
 		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
 			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
 		else if (dev->data->dev_conf.link_duplex ==
@@ -649,7 +649,7 @@ eth_em_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		if ((dev->data->dev_conf.link_duplex ==
 				ETH_LINK_AUTONEG_DUPLEX) ||
 			(dev->data->dev_conf.link_duplex ==
@@ -658,7 +658,7 @@ eth_em_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_10000:
+	case ETH_SPEED_NUM_10G:
 	default:
 		goto error_invalid_config;
 	}
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index 4dfa7e3..86f25f6 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -1244,7 +1244,7 @@ eth_igb_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_10:
+	case ETH_SPEED_NUM_10M:
 		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
 			hw->phy.autoneg_advertised = E1000_ALL_10_SPEED;
 		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
@@ -1254,7 +1254,7 @@ eth_igb_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_100:
+	case ETH_SPEED_NUM_100M:
 		if (dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX)
 			hw->phy.autoneg_advertised = E1000_ALL_100_SPEED;
 		else if (dev->data->dev_conf.link_duplex == ETH_LINK_HALF_DUPLEX)
@@ -1264,14 +1264,14 @@ eth_igb_start(struct rte_eth_dev *dev)
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		if ((dev->data->dev_conf.link_duplex == ETH_LINK_AUTONEG_DUPLEX) ||
 				(dev->data->dev_conf.link_duplex == ETH_LINK_FULL_DUPLEX))
 			hw->phy.autoneg_advertised = ADVERTISE_1000_FULL;
 		else
 			goto error_invalid_config;
 		break;
-	case ETH_LINK_SPEED_10000:
+	case ETH_SPEED_NUM_10G:
 	default:
 		goto error_invalid_config;
 	}
diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 325c513..1046286 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -504,7 +504,7 @@ static int ena_link_update(struct rte_eth_dev *dev,
 	struct rte_eth_link *link = &dev->data->dev_link;
 
 	link->link_status = 1;
-	link->link_speed = ETH_LINK_SPEED_10G;
+	link->link_speed = ETH_SPEED_NUM_10G;
 	link->link_duplex = ETH_LINK_FULL_DUPLEX;
 
 	return 0;
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index 05126e8..cce9e6f 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -1386,19 +1386,19 @@ i40e_parse_link_speed(uint16_t eth_link_speed)
 	uint8_t link_speed = I40E_LINK_SPEED_UNKNOWN;
 
 	switch (eth_link_speed) {
-	case ETH_LINK_SPEED_40G:
+	case ETH_SPEED_NUM_40G:
 		link_speed = I40E_LINK_SPEED_40GB;
 		break;
-	case ETH_LINK_SPEED_20G:
+	case ETH_SPEED_NUM_20G:
 		link_speed = I40E_LINK_SPEED_20GB;
 		break;
-	case ETH_LINK_SPEED_10G:
+	case ETH_SPEED_NUM_10G:
 		link_speed = I40E_LINK_SPEED_10GB;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		link_speed = I40E_LINK_SPEED_1GB;
 		break;
-	case ETH_LINK_SPEED_100:
+	case ETH_SPEED_NUM_100M:
 		link_speed = I40E_LINK_SPEED_100MB;
 		break;
 	}
@@ -1768,7 +1768,7 @@ i40e_dev_link_update(struct rte_eth_dev *dev,
 		/* Get link status information from hardware */
 		status = i40e_aq_get_link_info(hw, false, &link_status, NULL);
 		if (status != I40E_SUCCESS) {
-			link.link_speed = ETH_LINK_SPEED_100;
+			link.link_speed = ETH_SPEED_NUM_100M;
 			link.link_duplex = ETH_LINK_FULL_DUPLEX;
 			PMD_DRV_LOG(ERR, "Failed to get link info");
 			goto out;
@@ -1790,22 +1790,22 @@ i40e_dev_link_update(struct rte_eth_dev *dev,
 	/* Parse the link status */
 	switch (link_status.link_speed) {
 	case I40E_LINK_SPEED_100MB:
-		link.link_speed = ETH_LINK_SPEED_100;
+		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 	case I40E_LINK_SPEED_1GB:
-		link.link_speed = ETH_LINK_SPEED_1000;
+		link.link_speed = ETH_SPEED_NUM_1G;
 		break;
 	case I40E_LINK_SPEED_10GB:
-		link.link_speed = ETH_LINK_SPEED_10G;
+		link.link_speed = ETH_SPEED_NUM_10G;
 		break;
 	case I40E_LINK_SPEED_20GB:
-		link.link_speed = ETH_LINK_SPEED_20G;
+		link.link_speed = ETH_SPEED_NUM_20G;
 		break;
 	case I40E_LINK_SPEED_40GB:
-		link.link_speed = ETH_LINK_SPEED_40G;
+		link.link_speed = ETH_SPEED_NUM_40G;
 		break;
 	default:
-		link.link_speed = ETH_LINK_SPEED_100;
+		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 	}
 
@@ -8158,15 +8158,15 @@ i40e_start_timecounters(struct rte_eth_dev *dev)
 	rte_i40e_dev_atomic_read_link_status(dev, &link);
 
 	switch (link.link_speed) {
-	case ETH_LINK_SPEED_40G:
+	case ETH_SPEED_NUM_40G:
 		tsync_inc_l = I40E_PTP_40GB_INCVAL & 0xFFFFFFFF;
 		tsync_inc_h = I40E_PTP_40GB_INCVAL >> 32;
 		break;
-	case ETH_LINK_SPEED_10G:
+	case ETH_SPEED_NUM_10G:
 		tsync_inc_l = I40E_PTP_10GB_INCVAL & 0xFFFFFFFF;
 		tsync_inc_h = I40E_PTP_10GB_INCVAL >> 32;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		tsync_inc_l = I40E_PTP_1GB_INCVAL & 0xFFFFFFFF;
 		tsync_inc_h = I40E_PTP_1GB_INCVAL >> 32;
 		break;
diff --git a/drivers/net/i40e/i40e_ethdev_vf.c b/drivers/net/i40e/i40e_ethdev_vf.c
index 91df13b..295dcd2 100644
--- a/drivers/net/i40e/i40e_ethdev_vf.c
+++ b/drivers/net/i40e/i40e_ethdev_vf.c
@@ -2126,7 +2126,7 @@ i40evf_dev_link_update(struct rte_eth_dev *dev,
 	else {
 		/* Always assume it's up, for Linux driver PF host */
 		new_link.link_duplex = ETH_LINK_AUTONEG_DUPLEX;
-		new_link.link_speed  = ETH_LINK_SPEED_10000;
+		new_link.link_speed  = ETH_SPEED_NUM_10G;
 		new_link.link_status = ETH_LINK_UP;
 	}
 	i40evf_dev_atomic_write_link_status(dev, &new_link);
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index 21a3b8c..a0179d2 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -2199,17 +2199,17 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
 				IXGBE_LINK_SPEED_82599_AUTONEG :
 				IXGBE_LINK_SPEED_82598_AUTONEG;
 		break;
-	case ETH_LINK_SPEED_100:
+	case ETH_SPEED_NUM_100M:
 		/*
 		 * Invalid for 82598 but error will be detected by
 		 * ixgbe_setup_link()
 		 */
 		speed = IXGBE_LINK_SPEED_100_FULL;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		speed = IXGBE_LINK_SPEED_1GB_FULL;
 		break;
-	case ETH_LINK_SPEED_10000:
+	case ETH_SPEED_NUM_10G:
 		speed = IXGBE_LINK_SPEED_10GB_FULL;
 		break;
 	default:
@@ -3074,7 +3074,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 		diag = ixgbe_check_link(hw, &link_speed, &link_up, 1);
 
 	if (diag != 0) {
-		link.link_speed = ETH_LINK_SPEED_100;
+		link.link_speed = ETH_SPEED_NUM_100M;
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
 		rte_ixgbe_dev_atomic_write_link_status(dev, &link);
 		if (link.link_status == old.link_status)
@@ -3095,19 +3095,19 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 	default:
 	case IXGBE_LINK_SPEED_UNKNOWN:
 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
-		link.link_speed = ETH_LINK_SPEED_100;
+		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 
 	case IXGBE_LINK_SPEED_100_FULL:
-		link.link_speed = ETH_LINK_SPEED_100;
+		link.link_speed = ETH_SPEED_NUM_100M;
 		break;
 
 	case IXGBE_LINK_SPEED_1GB_FULL:
-		link.link_speed = ETH_LINK_SPEED_1000;
+		link.link_speed = ETH_SPEED_NUM_1G;
 		break;
 
 	case IXGBE_LINK_SPEED_10GB_FULL:
-		link.link_speed = ETH_LINK_SPEED_10000;
+		link.link_speed = ETH_SPEED_NUM_10G;
 		break;
 	}
 	rte_ixgbe_dev_atomic_write_link_status(dev, &link);
@@ -5909,15 +5909,15 @@ ixgbe_start_timecounters(struct rte_eth_dev *dev)
 	rte_ixgbe_dev_atomic_read_link_status(dev, &link);
 
 	switch (link.link_speed) {
-	case ETH_LINK_SPEED_100:
+	case ETH_SPEED_NUM_100M:
 		incval = IXGBE_INCVAL_100;
 		shift = IXGBE_INCVAL_SHIFT_100;
 		break;
-	case ETH_LINK_SPEED_1000:
+	case ETH_SPEED_NUM_1G:
 		incval = IXGBE_INCVAL_1GB;
 		shift = IXGBE_INCVAL_SHIFT_1GB;
 		break;
-	case ETH_LINK_SPEED_10000:
+	case ETH_SPEED_NUM_10G:
 	default:
 		incval = IXGBE_INCVAL_10GB;
 		shift = IXGBE_INCVAL_SHIFT_10GB;
diff --git a/drivers/net/mpipe/mpipe_tilegx.c b/drivers/net/mpipe/mpipe_tilegx.c
index d93ab7e..1a77c7a 100644
--- a/drivers/net/mpipe/mpipe_tilegx.c
+++ b/drivers/net/mpipe/mpipe_tilegx.c
@@ -395,11 +395,11 @@ mpipe_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 		speed = state & GXIO_MPIPE_LINK_SPEED_MASK;
 
 		if (speed == GXIO_MPIPE_LINK_1G) {
-			new.link_speed = ETH_LINK_SPEED_1000;
+			new.link_speed = ETH_SPEED_NUM_1G;
 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
 			new.link_status = ETH_LINK_UP;
 		} else if (speed == GXIO_MPIPE_LINK_10G) {
-			new.link_speed = ETH_LINK_SPEED_10000;
+			new.link_speed = ETH_SPEED_NUM_10G;
 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
 			new.link_status = ETH_LINK_UP;
 		}
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index 80dda85..18ea0f4 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -821,7 +821,7 @@ nfp_net_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
 
 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
 	/* Other cards can limit the tx and rx rate per VF */
-	link.link_speed = ETH_LINK_SPEED_40G;
+	link.link_speed = ETH_SPEED_NUM_40G;
 
 	if (old.link_status != link.link_status) {
 		nfp_net_dev_atomic_write_link_status(dev, &link);
diff --git a/drivers/net/null/rte_eth_null.c b/drivers/net/null/rte_eth_null.c
index 6adea91..5640585 100644
--- a/drivers/net/null/rte_eth_null.c
+++ b/drivers/net/null/rte_eth_null.c
@@ -89,7 +89,7 @@ struct pmd_internals {
 static struct ether_addr eth_addr = { .addr_bytes = {0} };
 static const char *drivername = "Null PMD";
 static struct rte_eth_link pmd_link = {
-	.link_speed = 10000,
+	.link_speed = ETH_SPEED_NUM_10G,
 	.link_duplex = ETH_LINK_FULL_DUPLEX,
 	.link_status = ETH_LINK_DOWN,
 };
diff --git a/drivers/net/pcap/rte_eth_pcap.c b/drivers/net/pcap/rte_eth_pcap.c
index b90c725..c657951 100644
--- a/drivers/net/pcap/rte_eth_pcap.c
+++ b/drivers/net/pcap/rte_eth_pcap.c
@@ -123,7 +123,7 @@ static int open_single_iface(const char *iface, pcap_t **pcap);
 static struct ether_addr eth_addr = { .addr_bytes = { 0, 0, 0, 0x1, 0x2, 0x3 } };
 static const char *drivername = "Pcap PMD";
 static struct rte_eth_link pmd_link = {
-		.link_speed = 10000,
+		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
 };
diff --git a/drivers/net/ring/rte_eth_ring.c b/drivers/net/ring/rte_eth_ring.c
index 4335c6a..58685e9 100644
--- a/drivers/net/ring/rte_eth_ring.c
+++ b/drivers/net/ring/rte_eth_ring.c
@@ -77,7 +77,7 @@ struct pmd_internals {
 
 static const char *drivername = "Rings PMD";
 static struct rte_eth_link pmd_link = {
-		.link_speed = 10000,
+		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
 };
diff --git a/drivers/net/szedata2/rte_eth_szedata2.c b/drivers/net/szedata2/rte_eth_szedata2.c
index 47aa7e3..dd1ae9e 100644
--- a/drivers/net/szedata2/rte_eth_szedata2.c
+++ b/drivers/net/szedata2/rte_eth_szedata2.c
@@ -1149,10 +1149,10 @@ eth_link_update(struct rte_eth_dev *dev,
 
 	switch (cgmii_link_speed(ibuf)) {
 	case SZEDATA2_LINK_SPEED_10G:
-		link.link_speed = ETH_LINK_SPEED_10G;
+		link.link_speed = ETH_SPEED_NUM_10G;
 		break;
 	case SZEDATA2_LINK_SPEED_40G:
-		link.link_speed = ETH_LINK_SPEED_40G;
+		link.link_speed = ETH_SPEED_NUM_40G;
 		break;
 	case SZEDATA2_LINK_SPEED_100G:
 		/*
@@ -1161,10 +1161,10 @@ eth_link_update(struct rte_eth_dev *dev,
 		 * will be changed to support 100Gbps speed change
 		 * this value to 100G.
 		 */
-		link.link_speed = ETH_LINK_SPEED_10G;
+		link.link_speed = ETH_SPEED_NUM_10G;
 		break;
 	default:
-		link.link_speed = ETH_LINK_SPEED_10G;
+		link.link_speed = ETH_SPEED_NUM_10G;
 		break;
 	}
 
diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
index 3f26217..6afa14e 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
@@ -779,7 +779,7 @@ vmxnet3_dev_link_update(struct rte_eth_dev *dev, __attribute__((unused)) int wai
 	if (ret & 0x1) {
 		link.link_status = ETH_LINK_UP;
 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
-		link.link_speed = ETH_LINK_SPEED_10000;
+		link.link_speed = ETH_SPEED_NUM_10G;
 	}
 
 	vmxnet3_dev_atomic_write_link_status(dev, &link);
diff --git a/drivers/net/xenvirt/rte_eth_xenvirt.c b/drivers/net/xenvirt/rte_eth_xenvirt.c
index 9453a06..77d3ba1 100644
--- a/drivers/net/xenvirt/rte_eth_xenvirt.c
+++ b/drivers/net/xenvirt/rte_eth_xenvirt.c
@@ -70,7 +70,7 @@ static int virtio_idx = 0;
 static const char *drivername = "xen virtio PMD";
 
 static struct rte_eth_link pmd_link = {
-		.link_speed = 10000,
+		.link_speed = ETH_SPEED_NUM_10G,
 		.link_duplex = ETH_LINK_FULL_DUPLEX,
 		.link_status = ETH_LINK_DOWN,
 };
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 2d13f92..bc7d607 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -242,23 +242,30 @@ struct rte_eth_stats {
 };
 
 /**
+ * Ethernet numeric link speeds in Mbps
+ */
+#define ETH_LINK_SPEED_AUTONEG     0 /**< Auto-negotiate link speed. */
+#define ETH_SPEED_NUM_10M         10 /**<  10 Mbps */
+#define ETH_SPEED_NUM_100M       100 /**< 100 Mbps */
+#define ETH_SPEED_NUM_1G        1000 /**<   1 Gbps */
+#define ETH_SPEED_NUM_2_5G      2500 /**< 2.5 Gbps */
+#define ETH_SPEED_NUM_5G        5000 /**<   5 Gbps */
+#define ETH_SPEED_NUM_10G      10000 /**<  10 Gbps */
+#define ETH_SPEED_NUM_20G      20000 /**<  20 Gbps */
+#define ETH_SPEED_NUM_25G      25000 /**<  25 Gbps */
+#define ETH_SPEED_NUM_40G      40000 /**<  40 Gbps */
+#define ETH_SPEED_NUM_50G      50000 /**<  50 Gbps */
+#define ETH_SPEED_NUM_56G      56000 /**<  56 Gbps */
+
+/**
  * A structure used to retrieve link-level information of an Ethernet port.
  */
 struct rte_eth_link {
-	uint16_t link_speed;      /**< ETH_LINK_SPEED_[10, 100, 1000, 10000] */
+	uint16_t link_speed;      /**< ETH_SPEED_NUM_ */
 	uint16_t link_duplex;     /**< ETH_LINK_[HALF/FULL]_DUPLEX */
 	uint8_t  link_status : 1; /**< ETH_LINK_[DOWN/UP] */
 }__attribute__((aligned(8)));     /**< aligned for atomic64 read/write */
 
-#define ETH_LINK_SPEED_AUTONEG  0       /**< Auto-negotiate link speed. */
-#define ETH_LINK_SPEED_10       10      /**< 10 megabits/second. */
-#define ETH_LINK_SPEED_100      100     /**< 100 megabits/second. */
-#define ETH_LINK_SPEED_1000     1000    /**< 1 gigabits/second. */
-#define ETH_LINK_SPEED_10000    10000   /**< 10 gigabits/second. */
-#define ETH_LINK_SPEED_10G      10000   /**< alias of 10 gigabits/second. */
-#define ETH_LINK_SPEED_20G      20000   /**< 20 gigabits/second. */
-#define ETH_LINK_SPEED_40G      40000   /**< 40 gigabits/second. */
-
 /* Utility constants */
 #define ETH_LINK_AUTONEG_DUPLEX 0       /**< Auto-negotiate duplex. */
 #define ETH_LINK_HALF_DUPLEX    1       /**< Half-duplex connection. */
@@ -779,7 +786,7 @@ struct rte_intr_conf {
  */
 struct rte_eth_conf {
 	uint16_t link_speed;
-	/**< ETH_LINK_SPEED_10[0|00|000], or 0 for autonegotation */
+	/**< ETH_SPEED_NUM_ or 0 for autonegotiation */
 	uint16_t link_duplex;
 	/**< ETH_LINK_[HALF_DUPLEX|FULL_DUPLEX], or 0 for autonegotation */
 	struct rte_eth_rxmode rxmode; /**< Port RX configuration. */
-- 
2.1.4

--------------------------------------------------------------
Intel Research and Development Ireland Limited
Registered in Ireland
Registered Office: Collinstown Industrial Park, Leixlip, County Kildare
Registered Number: 308263


This e-mail and any attachments may contain confidential material for the sole
use of the intended recipient(s). Any review or distribution by others is
strictly prohibited. If you are not the intended recipient, please contact the
sender and delete all copies.

^ permalink raw reply	[relevance 1%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
  2016-04-05 21:16  4% ` Singh, Jasvinder
@ 2016-04-06  8:51  4%   ` Azarewicz, PiotrX T
  2016-04-07 21:24  4%     ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Azarewicz, PiotrX T @ 2016-04-06  8:51 UTC (permalink / raw)
  To: Singh, Jasvinder, Zhang, Roy Fan, dev

> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Fan Zhang
> > Sent: Thursday, March 31, 2016 2:29 PM
> > To: dev@dpdk.org
> > Subject: [dpdk-dev] [PATCH] doc: announce ABI change for
> > rte_port_source_params structure
> >
> > Several new fields will be added to structure rte_port_source_params
> > for source port enhancement with pcap file reading support.
> >
> > Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> > Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
> 
> Acked-by: Jasvinder Singh <jasvinder.singh@intel.com>

Acked-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants
  2016-04-06  8:34  1%       ` Weglicki, MichalX
@ 2016-04-06  8:52  0%         ` Thomas Monjalon
  2016-04-06  9:16  3%           ` Weglicki, MichalX
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-06  8:52 UTC (permalink / raw)
  To: Weglicki, MichalX
  Cc: Marc Sune, Xu, Qian Q, Xing, Beilei, dev, Ananyev, Konstantin,
	Lu, Wenzhuo, Richardson, Bruce, Glynn, Michael J, Gray, Mark D

2016-04-06 08:34, Weglicki, MichalX:
> Hello, 
> 
> I have a question about this patch. 
> 
> As far as I see changing ETH_LINK_SPEED_ to ETH_SPEED_NUM_ is rather cosmetic change, am I right? 

No.
ETH_LINK_SPEED was used for configuration and reported speed.
Now the configuration is done with a bitmap filled with new ETH_LINK_SPEED
and the old numerical values are kept for other usages as ETH_SPEED_NUM.

> Disadvantage of that is that it breaks compatibility with OVS (compilation) and thus it also breaks backward compatibility between OVS & DPDK. Is it really necessary? 

Yes that's why this change was discussed 9 months ago and announced in the
previous release notes.

> It would be great to keep ABI & API stable if possible.

We try to keep it stable when possible and continue to bring some improvements.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-05 18:45  0% ` Thomas Monjalon
@ 2016-04-06  9:02  0%   ` Van Haaren, Harry
  2016-04-06  9:22  0%     ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Van Haaren, Harry @ 2016-04-06  9:02 UTC (permalink / raw)
  To: Thomas Monjalon, 'David Harton (dharton)'; +Cc: dev, Tahhan, Maryam

+ David Harton,

> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Subject: Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
> 2016-04-05 18:58, Harry van Haaren:
> > +* ABI change is planned for the xstats API

> Have you already submitted a RFC patch to let us have an opinion on the change?
> We need, at least, to see the structure changes.

This API break is to allow changing the API of xstats given the conversation that was on-list, as discussed in this thread:

http://thread.gmane.org/gmane.comp.networking.dpdk.devel/31728/focus=31903

The issue we are going to fix is that currently PMDs copy strings when retrieving statistics, which causes unnecessary overhead. The implementation is not decided yet, but using an int->value mapping seems logical.

The rte_eth_xstats struct size may be modified, and the API to retrieve statistics will be modified.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants
  2016-04-06  8:52  0%         ` Thomas Monjalon
@ 2016-04-06  9:16  3%           ` Weglicki, MichalX
  2016-04-06  9:34  3%             ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Weglicki, MichalX @ 2016-04-06  9:16 UTC (permalink / raw)
  To: Thomas Monjalon
  Cc: Marc Sune, Xu, Qian Q, Xing, Beilei, dev, Ananyev, Konstantin,
	Lu, Wenzhuo, Richardson, Bruce, Glynn, Michael J, Gray, Mark D

Hello, 

Thank you for your answer. 

I didn't mention that patch is cosmetic, rather naming. 

But all clear, I somehow missed it 9 months ago, and couldn't find it. 

Is there any official place where I can find upcoming ABI & API changes, or I have to dig through mailing list? 

Thank you in advance. 

Br, 
Michal. 

-----Original Message-----
From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com] 
Sent: Wednesday, April 6, 2016 9:53 AM
To: Weglicki, MichalX <michalx.weglicki@intel.com>
Cc: Marc Sune <marcdevel@gmail.com>; Xu, Qian Q <qian.q.xu@intel.com>; Xing, Beilei <beilei.xing@intel.com>; dev@dpdk.org; Ananyev, Konstantin <konstantin.ananyev@intel.com>; Lu, Wenzhuo <wenzhuo.lu@intel.com>; Richardson, Bruce <bruce.richardson@intel.com>; Glynn, Michael J <michael.j.glynn@intel.com>; Gray, Mark D <mark.d.gray@intel.com>
Subject: Re: [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants

2016-04-06 08:34, Weglicki, MichalX:
> Hello, 
> 
> I have a question about this patch. 
> 
> As far as I see changing ETH_LINK_SPEED_ to ETH_SPEED_NUM_ is rather cosmetic change, am I right? 

No.
ETH_LINK_SPEED was used for configuration and reported speed.
Now the configuration is done with a bitmap filled with new ETH_LINK_SPEED
and the old numerical values are kept for other usages as ETH_SPEED_NUM.

> Disadvantage of that is that it breaks compatibility with OVS (compilation) and thus it also breaks backward compatibility between OVS & DPDK. Is it really necessary? 

Yes that's why this change was discussed 9 months ago and announced in the
previous release notes.

> It would be great to keep ABI & API stable if possible.

We try to keep it stable when possible and continue to bring some improvements.

--------------------------------------------------------------
Intel Research and Development Ireland Limited
Registered in Ireland
Registered Office: Collinstown Industrial Park, Leixlip, County Kildare
Registered Number: 308263


This e-mail and any attachments may contain confidential material for the sole
use of the intended recipient(s). Any review or distribution by others is
strictly prohibited. If you are not the intended recipient, please contact the
sender and delete all copies.

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-06  9:02  0%   ` Van Haaren, Harry
@ 2016-04-06  9:22  0%     ` Thomas Monjalon
  2016-04-06 11:16  3%       ` Van Haaren, Harry
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-06  9:22 UTC (permalink / raw)
  To: Van Haaren, Harry
  Cc: 'David Harton (dharton)', dev, Tahhan, Maryam, olivier.matz

2016-04-06 09:02, Van Haaren, Harry:
> + David Harton,
> 
> > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > Subject: Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
> > 2016-04-05 18:58, Harry van Haaren:
> > > +* ABI change is planned for the xstats API
> 
> > Have you already submitted a RFC patch to let us have an opinion on the change?
> > We need, at least, to see the structure changes.
> 
> This API break is to allow changing the API of xstats given the conversation that was on-list, as discussed in this thread:
> 
> http://thread.gmane.org/gmane.comp.networking.dpdk.devel/31728/focus=31903
> 
> The issue we are going to fix is that currently PMDs copy strings when retrieving statistics, which causes unnecessary overhead. The implementation is not decided yet, but using an int->value mapping seems logical.

I am not sure performance is so much critical when retrieving statistics.
The extended stats can be infinitely extended. So a string identifier seems
a lot more natural.
I do not agree to add a new numeric identifier in the API each time a driver
wants to report a specific statistic for debugging purpose.

> The rte_eth_xstats struct size may be modified, and the API to retrieve statistics will be modified.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants
  2016-04-06  9:16  3%           ` Weglicki, MichalX
@ 2016-04-06  9:34  3%             ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-06  9:34 UTC (permalink / raw)
  To: Weglicki, MichalX
  Cc: Marc Sune, Xu, Qian Q, Xing, Beilei, dev, Ananyev, Konstantin,
	Lu, Wenzhuo, Richardson, Bruce, Glynn, Michael J, Gray, Mark D

2016-04-06 09:16, Weglicki, MichalX:
> Hello, 
> 
> Thank you for your answer. 
> 
> I didn't mention that patch is cosmetic, rather naming. 
> 
> But all clear, I somehow missed it 9 months ago, and couldn't find it. 
> 
> Is there any official place where I can find upcoming ABI & API changes, or I have to dig through mailing list? 

Yes, the API and ABI changes are described in the release notes:
	http://dpdk.org/browse/dpdk/tree/doc/guides/rel_notes/release_16_04.rst#n471
And the next deprecations are announced in another section:
	http://dpdk.org/browse/dpdk/tree/doc/guides/rel_notes/deprecation.rst

PS: please answer inline

-----------------
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com] 
> 2016-04-06 08:34, Weglicki, MichalX:
> > Hello, 
> > 
> > I have a question about this patch. 
> > 
> > As far as I see changing ETH_LINK_SPEED_ to ETH_SPEED_NUM_ is rather cosmetic change, am I right? 
> 
> No.
> ETH_LINK_SPEED was used for configuration and reported speed.
> Now the configuration is done with a bitmap filled with new ETH_LINK_SPEED
> and the old numerical values are kept for other usages as ETH_SPEED_NUM.
> 
> > Disadvantage of that is that it breaks compatibility with OVS (compilation) and thus it also breaks backward compatibility between OVS & DPDK. Is it really necessary? 
> 
> Yes that's why this change was discussed 9 months ago and announced in the
> previous release notes.
> 
> > It would be great to keep ABI & API stable if possible.
> 
> We try to keep it stable when possible and continue to bring some improvements.

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] Fw: dpdk-armv7-testing - Build # 43 - Failure!
  @ 2016-04-06 10:18  4%   ` Jan Viktorin
  0 siblings, 0 replies; 200+ results
From: Jan Viktorin @ 2016-04-06 10:18 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev, Adrien Mazarguil

On Wed, 06 Apr 2016 12:05:31 +0200
Thomas Monjalon <thomas.monjalon@6wind.com> wrote:

> Hi Jan,
> 
> 2016-04-06 11:53, Jan Viktorin:
> > Please, see the attached log file.  
> 
> The text attachments are filtered out on the mailing list because
> it is not convenient or impossible to read in the archives.
> Please use inline text, next time. Thanks

I am sorry, didn't notice that. Nobody has pointed to this before.

dpdk-armv7-testing - Build # 44 - Still Failing

See the following log:

[...]
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_parse.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_parse.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_parse_num.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_parse_num.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_parse_ipaddr.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_parse_ipaddr.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_parse_etheraddr.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_parse_etheraddr.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_parse_string.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_parse_string.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_rdline.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_rdline.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_vt100.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_vt100.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_socket.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_socket.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_cirbuf.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_cirbuf.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  SYMLINK-FILE include/cmdline_parse_portlist.h
ln -nsf `/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/scripts/relpath.sh /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_cmdline/cmdline_parse_portlist.h /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include` /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include
  INSTALL-LIB librte_cmdline.a
cp -f librte_cmdline.a /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/lib
== Build lib/librte_ether
/opt/gcc/br2-arm32-glibc-4.9.x/usr/bin/arm-buildroot-linux-gnueabi-gcc -Wp,-MD,./.rte_ethdev.o.d.tmp -mfloat-abi=softfp -mfloat-abi=softfp -mfloat-abi=softfp -pthread  -march=armv7-a -mtune=cortex-a9 -mfpu=neon -DRTE_MACHINE_CPUFLAG_NEON  -I/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include -include /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/build/include/rte_config.h -O3 -W -Wall -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations -Wold-style-definition -Wpointer-arith -Wcast-align -Wnested-externs -Wcast-qual -Wformat-nonliteral -Wformat-security -Wundef -Wwrite-strings -Werror -Wno-error=cast-align   -o rte_ethdev.o -c /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_ether/rte_ethdev.c 
In file included from /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_ether/rte_ethdev.h:186:0,
                 from /var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_ether/rte_ethdev.c:70:
/var/lib/jenkins/jobs/dpdk-armv7-testing/workspace/lib/librte_ether/rte_eth_ctrl.h:39:23: fatal error: rte_ether.h: No such file or directory
 #include <rte_ether.h>
                       ^
compilation terminated.
make[3]: *** [rte_ethdev.o] Error 1
make[2]: *** [librte_ether] Error 2
make[1]: *** [lib] Error 2
make: *** [all] Error 2
Build step 'Execute shell' marked build as failure
[WARNINGS] Skipping publisher since build result is FAILURE
Skipped archiving because build is not successful
Email was triggered for: Failure - Any
Sending email for trigger: Failure - Any


-- 
   Jan Viktorin                  E-mail: Viktorin@RehiveTech.com
   System Architect              Web:    www.RehiveTech.com
   RehiveTech
   Brno, Czech Republic

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-06  9:22  0%     ` Thomas Monjalon
@ 2016-04-06 11:16  3%       ` Van Haaren, Harry
  2016-04-06 12:14  0%         ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Van Haaren, Harry @ 2016-04-06 11:16 UTC (permalink / raw)
  To: Thomas Monjalon
  Cc: 'David Harton (dharton)', dev, Tahhan, Maryam, olivier.matz

> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Subject: Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
> > The issue we are going to fix is that currently PMDs copy strings when retrieving
> statistics, which causes unnecessary overhead. The implementation is not decided yet, but
> using an int->value mapping seems logical.

> I am not sure performance is so much critical when retrieving statistics.

In the previous discussion David was concerned about performance impact
of string copies, are those concerns still present David?

> The extended stats can be infinitely extended. So a string identifier seems
> a lot more natural.

I'm not suggesting that the string identifier is removed totally.

> I do not agree to add a new numeric identifier in the API each time a driver
> wants to report a specific statistic for debugging purpose. 

And I agree - the ints are just an index to xstats arrays, no eth-dev wide enums here.
The proposal is to make the API more flexible, see example:
http://thread.gmane.org/gmane.comp.networking.dpdk.devel/31728/focus=32795

This more flexible API would allow other types of information about
statistics be retrieved too.

For now, the sent patch announces that the API/ABI may change, and we can
discuss details of API as development starts.

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-06  5:26  0%     ` Yuanhan Liu
@ 2016-04-06 12:07  0%       ` Panu Matilainen
  2016-04-06 12:34  0%         ` Ananyev, Konstantin
                           ` (2 more replies)
  0 siblings, 3 replies; 200+ results
From: Panu Matilainen @ 2016-04-06 12:07 UTC (permalink / raw)
  To: Yuanhan Liu, Arnon Warshavsky; +Cc: Trahe, Fiona, Thomas Monjalon, dev

On 04/06/2016 08:26 AM, Yuanhan Liu wrote:
> On Tue, Apr 05, 2016 at 05:31:22PM +0300, Arnon Warshavsky wrote:
>> On Tue, Apr 5, 2016 at 5:13 PM, Trahe, Fiona <fiona.trahe@intel.com> wrote:
>>
>>>
>>>
>>>> -----Original Message-----
>>>> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
>>>> Sent: Tuesday, April 05, 2016 2:57 PM
>>>> To: dev@dpdk.org
>>>> Subject: [dpdk-dev] DPDK namespace
>>>>
>>>> DPDK is going to be more popular in Linux distributions.
>>>> It means people will have some DPDK files in their /usr/include and some
>>> DPDK
>>>> libraries on their system.
>>>>
>>>> Let's imagine someone trying to compile an application which needs
>>>> rte_ethdev.h. He has to figure out that this "rte header" is provided by
>>> the DPDK.
>>>> Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
>>>> Then someone else will try to run a binary without having installed the
>>> DPDK
>>>> libraries. The linker will require libethdev.so (no prefix here).
>>>> StackOverflow will probably have another good answer (among wrong ones):
>>>> "Hey Sherlock Holmes, have you tried to install the DPDK library?"
>>>> Followed by an insight: "You know, the DPDK naming is weird..."
>>>> And we could continue the story with developers having some naming clash
>>>> because of some identifiers not prefixed at all.
>>>>
>>>> The goal of this email is to get some feedback on how important it is to
>>> fix the
>>>> DPDK namespace.
>>>>
>>>> If there is enough agreement that we should do something, I suggest to
>>>> introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
>>>> during some time.
>>>> We could start using the new prefix for the new APIs (example: crypto)
>>> or when
>>>> there is a significant API break (example: mempool).
>>>>
>>>> Opinions welcome!
>>> I don't have an opinion on how important it is to fix the namespace,
>>> though it does seem like a good idea.
>>> However if it's to be done, in my opinion it should be completed quickly
>>> or will just cause more confusion.
>>> So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should
>>> follow in next release or two, with
>>> the resulting ABI compatibility handling. Maybe with dual naming handled
>>> for several releases, but a
>>> clear end date when all are converted.
>>> Else there will be many years with a mix of rte_ and dpdk_
>>>
>>>
>>
>> Googling rte functions or error codes usually takes you to dpdk dev email
>> archive so I don't think it is that much difficult to figure out where rte
>> comes from.
>> Other than that , except for my own refactoring pains when replacing a dpdk
>> version, I do not see a major reason why not.
>> If Going for dpdk_ prefix, I agree with the quick death approach.
>
> +1: it's a bit weird to keep both, especially for a long while, that
> every time we turn a rte_ prefix to dpdk_ prefix, we break applications.
> Instead of breaking applications many times, I'd prefer to break once.
> Therefore, applications could do a simple global rte_ -> dpdk_ substitute:
> it doesn't sound that painful then.

I concur. If (and I think that should be a pretty big IF) the prefix is 
to be changed then its better done in one fast sweep than gradually.

Gratuitious (or nearly so) change is always extremely annoying, and the 
longer it takes the more painful it is. Application developers wont much 
care what the prefix is as long as its consistent, but if they're forced 
to track prefix changes across several releases with different libraries 
moving at different pace, they WILL be calling for bloody murder :)

As for rte_ being strange for DPDK - yes it is, but it takes like 5 
minutes to get over it. It would help to have it explained on dpdk.org 
FAQ: "Due to historical reasons, DPDK libraries are prefixed rte_ 
instead of dpdk_ because <insert excuse here, probably early project 
name> and changing it is unnecessarily painful."

>
> And here are few more comments:
>
> - we should add rte_/dpdk_ prefix to all public structures as well.
>
>    I'm thinking we are doing well here. I'm just aware that vhost lib
>    does a bad job, which is something I proposed to fix in next release.

Yup, all public symbols should be prefixed. What the exact prefix is 
isn't that important really.

>
> - If we do the whole change once, I'd suggest to do it ASAP when this
>    release is over.
>
>    It should be a HUGE change that touches a lot of code, if we do it
>    later, at a stage that a lot of patches for new features have been
>    made or sent out, all of them need rebase. That'd be painful.

Nod, that's yet another aspect to consider.

So to summarize, I'm not strongly opposed to doing a one-time mass rte_ 
-> dpdk_ prefix change, but it needs to be one big sweep all at once, or 
not do it at all. Gradual change is a suicide.

Keeping rte_ is not the end of the world by any means, especially when 
applied consistently and explained someplace.

	- Panu -

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-06 11:16  3%       ` Van Haaren, Harry
@ 2016-04-06 12:14  0%         ` Thomas Monjalon
  2016-04-06 13:49  0%           ` David Harton (dharton)
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-06 12:14 UTC (permalink / raw)
  To: Van Haaren, Harry
  Cc: 'David Harton (dharton)', dev, Tahhan, Maryam, olivier.matz

2016-04-06 11:16, Van Haaren, Harry:
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > > The issue we are going to fix is that currently PMDs copy strings when retrieving
> > statistics, which causes unnecessary overhead. The implementation is not decided yet, but
> > using an int->value mapping seems logical.
> 
> > I am not sure performance is so much critical when retrieving statistics.
> 
> In the previous discussion David was concerned about performance impact
> of string copies, are those concerns still present David?
> 
> > The extended stats can be infinitely extended. So a string identifier seems
> > a lot more natural.
> 
> I'm not suggesting that the string identifier is removed totally.
> 
> > I do not agree to add a new numeric identifier in the API each time a driver
> > wants to report a specific statistic for debugging purpose. 
> 
> And I agree - the ints are just an index to xstats arrays, no eth-dev wide enums here.
> The proposal is to make the API more flexible, see example:
> http://thread.gmane.org/gmane.comp.networking.dpdk.devel/31728/focus=32795
> 
> This more flexible API would allow other types of information about
> statistics be retrieved too.

OK I think I start to understand.

> For now, the sent patch announces that the API/ABI may change, and we can
> discuss details of API as development starts.

This should not be the normal process.
It is important to understand what should be the changes to decide of
announcing or not a deprecation.
In the case of the mempool reworks, the patch have been sent and discussed
on the mailing list.
Given the previous explanations (and knowing you did good job on stats),
I give my
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-06 12:07  0%       ` Panu Matilainen
@ 2016-04-06 12:34  0%         ` Ananyev, Konstantin
  2016-04-06 14:36  0%         ` Wiles, Keith
    2 siblings, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-04-06 12:34 UTC (permalink / raw)
  To: Panu Matilainen, Yuanhan Liu, Arnon Warshavsky
  Cc: Trahe, Fiona, Thomas Monjalon, dev



> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Panu Matilainen
> Sent: Wednesday, April 06, 2016 1:08 PM
> To: Yuanhan Liu; Arnon Warshavsky
> Cc: Trahe, Fiona; Thomas Monjalon; dev@dpdk.org
> Subject: Re: [dpdk-dev] DPDK namespace
> 
> On 04/06/2016 08:26 AM, Yuanhan Liu wrote:
> > On Tue, Apr 05, 2016 at 05:31:22PM +0300, Arnon Warshavsky wrote:
> >> On Tue, Apr 5, 2016 at 5:13 PM, Trahe, Fiona <fiona.trahe@intel.com> wrote:
> >>
> >>>
> >>>
> >>>> -----Original Message-----
> >>>> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> >>>> Sent: Tuesday, April 05, 2016 2:57 PM
> >>>> To: dev@dpdk.org
> >>>> Subject: [dpdk-dev] DPDK namespace
> >>>>
> >>>> DPDK is going to be more popular in Linux distributions.
> >>>> It means people will have some DPDK files in their /usr/include and some
> >>> DPDK
> >>>> libraries on their system.
> >>>>
> >>>> Let's imagine someone trying to compile an application which needs
> >>>> rte_ethdev.h. He has to figure out that this "rte header" is provided by
> >>> the DPDK.
> >>>> Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
> >>>> Then someone else will try to run a binary without having installed the
> >>> DPDK
> >>>> libraries. The linker will require libethdev.so (no prefix here).
> >>>> StackOverflow will probably have another good answer (among wrong ones):
> >>>> "Hey Sherlock Holmes, have you tried to install the DPDK library?"
> >>>> Followed by an insight: "You know, the DPDK naming is weird..."
> >>>> And we could continue the story with developers having some naming clash
> >>>> because of some identifiers not prefixed at all.
> >>>>
> >>>> The goal of this email is to get some feedback on how important it is to
> >>> fix the
> >>>> DPDK namespace.
> >>>>
> >>>> If there is enough agreement that we should do something, I suggest to
> >>>> introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
> >>>> during some time.
> >>>> We could start using the new prefix for the new APIs (example: crypto)
> >>> or when
> >>>> there is a significant API break (example: mempool).
> >>>>
> >>>> Opinions welcome!
> >>> I don't have an opinion on how important it is to fix the namespace,
> >>> though it does seem like a good idea.
> >>> However if it's to be done, in my opinion it should be completed quickly
> >>> or will just cause more confusion.
> >>> So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should
> >>> follow in next release or two, with
> >>> the resulting ABI compatibility handling. Maybe with dual naming handled
> >>> for several releases, but a
> >>> clear end date when all are converted.
> >>> Else there will be many years with a mix of rte_ and dpdk_
> >>>
> >>>
> >>
> >> Googling rte functions or error codes usually takes you to dpdk dev email
> >> archive so I don't think it is that much difficult to figure out where rte
> >> comes from.
> >> Other than that , except for my own refactoring pains when replacing a dpdk
> >> version, I do not see a major reason why not.
> >> If Going for dpdk_ prefix, I agree with the quick death approach.
> >
> > +1: it's a bit weird to keep both, especially for a long while, that
> > every time we turn a rte_ prefix to dpdk_ prefix, we break applications.
> > Instead of breaking applications many times, I'd prefer to break once.
> > Therefore, applications could do a simple global rte_ -> dpdk_ substitute:
> > it doesn't sound that painful then.
> 
> I concur. If (and I think that should be a pretty big IF) the prefix is
> to be changed then its better done in one fast sweep than gradually.
> 
> Gratuitious (or nearly so) change is always extremely annoying, and the
> longer it takes the more painful it is. Application developers wont much
> care what the prefix is as long as its consistent, but if they're forced
> to track prefix changes across several releases with different libraries
> moving at different pace, they WILL be calling for bloody murder :)
> 
> As for rte_ being strange for DPDK - yes it is, but it takes like 5
> minutes to get over it. It would help to have it explained on dpdk.org
> FAQ: "Due to historical reasons, DPDK libraries are prefixed rte_
> instead of dpdk_ because <insert excuse here, probably early project
> name> and changing it is unnecessarily painful."
> 
> >
> > And here are few more comments:
> >
> > - we should add rte_/dpdk_ prefix to all public structures as well.
> >
> >    I'm thinking we are doing well here. I'm just aware that vhost lib
> >    does a bad job, which is something I proposed to fix in next release.
> 
> Yup, all public symbols should be prefixed. What the exact prefix is
> isn't that important really.
> 
> >
> > - If we do the whole change once, I'd suggest to do it ASAP when this
> >    release is over.
> >
> >    It should be a HUGE change that touches a lot of code, if we do it
> >    later, at a stage that a lot of patches for new features have been
> >    made or sent out, all of them need rebase. That'd be painful.
> 
> Nod, that's yet another aspect to consider.
> 
> So to summarize, I'm not strongly opposed to doing a one-time mass rte_
> -> dpdk_ prefix change, but it needs to be one big sweep all at once, or
> not do it at all. Gradual change is a suicide.
> 
> Keeping rte_ is not the end of the world by any means, especially when
> applied consistently and explained someplace.
> 

Yep, I have exactly the same thoughts:
1. Yes, dpdk_'  prefix is a better naming approach than 'rte_',
but for me not that better to overweight all the pain of such big change.
2. If we still decide to do that change - my preference would be to do it in one go. 
I personally don't care that much what the prefix would be, as long as it is consistent
across the whole codebase. 
Konstantin


^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-06 12:14  0%         ` Thomas Monjalon
@ 2016-04-06 13:49  0%           ` David Harton (dharton)
  0 siblings, 0 replies; 200+ results
From: David Harton (dharton) @ 2016-04-06 13:49 UTC (permalink / raw)
  To: Thomas Monjalon, Van Haaren, Harry; +Cc: dev, Tahhan, Maryam, olivier.matz



> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Wednesday, April 06, 2016 8:14 AM
> To: Van Haaren, Harry <harry.van.haaren@intel.com>
> Cc: David Harton (dharton) <dharton@cisco.com>; dev@dpdk.org; Tahhan,
> Maryam <maryam.tahhan@intel.com>; olivier.matz@6wind.com
> Subject: Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
> 
> 2016-04-06 11:16, Van Haaren, Harry:
> > From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > > > The issue we are going to fix is that currently PMDs copy strings
> > > > when retrieving
> > > statistics, which causes unnecessary overhead. The implementation is
> > > not decided yet, but using an int->value mapping seems logical.
> >
> > > I am not sure performance is so much critical when retrieving
> statistics.
> >
> > In the previous discussion David was concerned about performance
> > impact of string copies, are those concerns still present David?
> >
> > > The extended stats can be infinitely extended. So a string
> > > identifier seems a lot more natural.
> >
> > I'm not suggesting that the string identifier is removed totally.
> >
> > > I do not agree to add a new numeric identifier in the API each time
> > > a driver wants to report a specific statistic for debugging purpose.
> >
> > And I agree - the ints are just an index to xstats arrays, no eth-dev
> wide enums here.

Yes, I abandoned the idea of a set of stats ids.  I can see where registration will be problematic and cumbersome to driver developers.

> > The proposal is to make the API more flexible, see example:
> > http://thread.gmane.org/gmane.comp.networking.dpdk.devel/31728/focus=3
> > 2795
> >
> > This more flexible API would allow other types of information about
> > statistics be retrieved too.

I have prototyped this.  If there is interest/acceptance I can work on making an official patch to share back to the community.

Using this method still gives the flexibility the current API desires while giving the user the control to only obtain the counters.  This of course assumes that the counters per device are static but that seems a safe bet.

> 
> OK I think I start to understand.
> 
> > For now, the sent patch announces that the API/ABI may change, and we
> > can discuss details of API as development starts.
> 
> This should not be the normal process.
> It is important to understand what should be the changes to decide of
> announcing or not a deprecation.
> In the case of the mempool reworks, the patch have been sent and discussed
> on the mailing list.
> Given the previous explanations (and knowing you did good job on stats), I
> give my
> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>

Thanks for considering this.

Regards,
Dave

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
  2016-04-05 17:58  9% [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07 Harry van Haaren
  2016-04-05 18:45  0% ` Thomas Monjalon
@ 2016-04-06 14:00  0% ` David Harton (dharton)
  1 sibling, 0 replies; 200+ results
From: David Harton (dharton) @ 2016-04-06 14:00 UTC (permalink / raw)
  To: Harry van Haaren, dev; +Cc: maryam.tahhan


> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Harry van Haaren
> Sent: Tuesday, April 05, 2016 1:58 PM
> To: dev@dpdk.org
> Cc: maryam.tahhan@intel.com; Harry van Haaren <harry.van.haaren@intel.com>
> Subject: [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07
> 
> This patch adds a notice that the API for the xstats functionality will be
> modified in the 16.07 release, with no backwards compatibility planned as
> it would require code duplication in each PMD that supports xstats.
> 
> Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
> ---
>  doc/guides/rel_notes/deprecation.rst | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/doc/guides/rel_notes/deprecation.rst
> b/doc/guides/rel_notes/deprecation.rst
> index 98d5529..13c3a95 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -54,3 +54,8 @@ Deprecation Notices
>    induce a modification of the rte_mempool structure, plus a
>    modification of the API of rte_mempool_obj_iter(), implying a breakage
>    of the ABI.
> +
> +* ABI change is planned for the xstats API and rte_eth_xstats struct,
> +to
> +  facilitate updating to an API that allows retrieval of values without
> +any
> +  string copies or parsing. No backwards compatibility is planned, as
> +it would
> +  require code duplication in every PMD that supports xstats.
> --
> 2.5.0

Acked-by: David Harton <dharton@cisco.com>

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-06 12:07  0%       ` Panu Matilainen
  2016-04-06 12:34  0%         ` Ananyev, Konstantin
@ 2016-04-06 14:36  0%         ` Wiles, Keith
    2 siblings, 0 replies; 200+ results
From: Wiles, Keith @ 2016-04-06 14:36 UTC (permalink / raw)
  To: Panu Matilainen, Yuanhan Liu, Arnon Warshavsky
  Cc: Trahe, Fiona, Thomas Monjalon, dev

>On 04/06/2016 08:26 AM, Yuanhan Liu wrote:
>> On Tue, Apr 05, 2016 at 05:31:22PM +0300, Arnon Warshavsky wrote:
>>> On Tue, Apr 5, 2016 at 5:13 PM, Trahe, Fiona <fiona.trahe@intel.com> wrote:
>>>
>>>>
>>>>
>>>>> -----Original Message-----
>>>>> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
>>>>> Sent: Tuesday, April 05, 2016 2:57 PM
>>>>> To: dev@dpdk.org
>>>>> Subject: [dpdk-dev] DPDK namespace
>>>>>
>>>>> DPDK is going to be more popular in Linux distributions.
>>>>> It means people will have some DPDK files in their /usr/include and some
>>>> DPDK
>>>>> libraries on their system.
>>>>>
>>>>> Let's imagine someone trying to compile an application which needs
>>>>> rte_ethdev.h. He has to figure out that this "rte header" is provided by
>>>> the DPDK.
>>>>> Hopefully it will be explained on StackOverflow that RTE stands for DPDK.
>>>>> Then someone else will try to run a binary without having installed the
>>>> DPDK
>>>>> libraries. The linker will require libethdev.so (no prefix here).
>>>>> StackOverflow will probably have another good answer (among wrong ones):
>>>>> "Hey Sherlock Holmes, have you tried to install the DPDK library?"
>>>>> Followed by an insight: "You know, the DPDK naming is weird..."
>>>>> And we could continue the story with developers having some naming clash
>>>>> because of some identifiers not prefixed at all.
>>>>>
>>>>> The goal of this email is to get some feedback on how important it is to
>>>> fix the
>>>>> DPDK namespace.
>>>>>
>>>>> If there is enough agreement that we should do something, I suggest to
>>>>> introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
>>>>> during some time.
>>>>> We could start using the new prefix for the new APIs (example: crypto)
>>>> or when
>>>>> there is a significant API break (example: mempool).
>>>>>
>>>>> Opinions welcome!
>>>> I don't have an opinion on how important it is to fix the namespace,
>>>> though it does seem like a good idea.
>>>> However if it's to be done, in my opinion it should be completed quickly
>>>> or will just cause more confusion.
>>>> So if rte_cryptoxxx becomes dpdk_cryptoxxx all other libraries should
>>>> follow in next release or two, with
>>>> the resulting ABI compatibility handling. Maybe with dual naming handled
>>>> for several releases, but a
>>>> clear end date when all are converted.
>>>> Else there will be many years with a mix of rte_ and dpdk_
>>>>
>>>>
>>>
>>> Googling rte functions or error codes usually takes you to dpdk dev email
>>> archive so I don't think it is that much difficult to figure out where rte
>>> comes from.
>>> Other than that , except for my own refactoring pains when replacing a dpdk
>>> version, I do not see a major reason why not.
>>> If Going for dpdk_ prefix, I agree with the quick death approach.
>>
>> +1: it's a bit weird to keep both, especially for a long while, that
>> every time we turn a rte_ prefix to dpdk_ prefix, we break applications.
>> Instead of breaking applications many times, I'd prefer to break once.
>> Therefore, applications could do a simple global rte_ -> dpdk_ substitute:
>> it doesn't sound that painful then.
>
>I concur. If (and I think that should be a pretty big IF) the prefix is 
>to be changed then its better done in one fast sweep than gradually.
>
>Gratuitious (or nearly so) change is always extremely annoying, and the 
>longer it takes the more painful it is. Application developers wont much 
>care what the prefix is as long as its consistent, but if they're forced 
>to track prefix changes across several releases with different libraries 
>moving at different pace, they WILL be calling for bloody murder :)
>
>As for rte_ being strange for DPDK - yes it is, but it takes like 5 
>minutes to get over it. It would help to have it explained on dpdk.org 
>FAQ: "Due to historical reasons, DPDK libraries are prefixed rte_ 
>instead of dpdk_ because <insert excuse here, probably early project 
>name> and changing it is unnecessarily painful."

As I understand RTE is from the “Run Time Environment” which was the primary set of API’s at the time and it just kept getting propagated :-)

>
>>
>> And here are few more comments:
>>
>> - we should add rte_/dpdk_ prefix to all public structures as well.
>>
>>    I'm thinking we are doing well here. I'm just aware that vhost lib
>>    does a bad job, which is something I proposed to fix in next release.
>
>Yup, all public symbols should be prefixed. What the exact prefix is 
>isn't that important really.
>
>>
>> - If we do the whole change once, I'd suggest to do it ASAP when this
>>    release is over.
>>
>>    It should be a HUGE change that touches a lot of code, if we do it
>>    later, at a stage that a lot of patches for new features have been
>>    made or sent out, all of them need rebase. That'd be painful.
>
>Nod, that's yet another aspect to consider.
>
>So to summarize, I'm not strongly opposed to doing a one-time mass rte_ 
>-> dpdk_ prefix change, but it needs to be one big sweep all at once, or 
>not do it at all. Gradual change is a suicide.
>
>Keeping rte_ is not the end of the world by any means, especially when 
>applied consistently and explained someplace.

To me rte_ is just fine, plus we have to change the structures names and defines names. I am sure we can figure out a script to convert any app for the developer, but why change. The rte_ prefix is something which can be explained and dpdk_ adds one character to type :-)

>
>	- Panu -
>


Regards,
Keith





^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] vhost: ABI/API change announcement due to refactor
  2016-04-06  6:53 15% [dpdk-dev] [PATCH] vhost: ABI/API change announcement due to refactor Yuanhan Liu
@ 2016-04-07  7:12  7% ` Panu Matilainen
  2016-04-10  9:58  4%   ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-04-07  7:12 UTC (permalink / raw)
  To: Yuanhan Liu, dev; +Cc: huawei.xie, Thomas Monjalon, Ilya Maximets

On 04/06/2016 09:53 AM, Yuanhan Liu wrote:
> We currently exposed way too many fields (or even structures) than
> necessary. For example, vhost_virtqueue struct should NOT be exposed
> to user at all: application just need to tell the right queue id to
> locate a specific queue, and that's all. Instead, the structure should
> be defined in an internal header file. With that, we could do any changes
> to it we want, without worrying about that we may offense the painful
> ABI rules.
>
> Similar changes could be done to virtio_net struct as well, just exposing
> very few fields that are necessary and moving all others to an internal
> structure.
>
> Huawei then suggested a more radical yet much cleaner one: just exposing
> a virtio_net handle to application, just like the way kernel exposes an
> fd to user for locating a specific file, and exposing some new functions
> to access those old fields, such as flags, virt_qp_nb.
>
> With this change, we're likely to be free from ABI violations forever
> (well, except when we have to extend the virtio_net_device_ops struct).
> For example, following nice cleanup would not be a blocking one then:
>
>      http://dpdk.org/ml/archives/dev/2016-February/033528.html
>
> Suggested-by: Huawei Xie <huawei.xie@intel.com>
> Cc: Ilya Maximets <i.maximets@samsung.com>
> Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.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 ad31355..7d16d86 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -40,3 +40,10 @@ Deprecation Notices
>     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
> +  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.
>

Acked-by: Panu Matilainen <pmatilai@redhat.com>

I applaud the initiative, public structs are by far the worst offender 
when trying to maintain a stable ABI because they're so hard to 
correctly version that hardly anybody besides glibc bothers.

	- Panu -

^ permalink raw reply	[relevance 7%]

* Re: [dpdk-dev] DPDK namespace
  @ 2016-04-07  8:22  3%           ` Marc
  0 siblings, 0 replies; 200+ results
From: Marc @ 2016-04-07  8:22 UTC (permalink / raw)
  To: Dave Neary
  Cc: Panu Matilainen, Yuanhan Liu, Arnon Warshavsky, Trahe, Fiona,
	Thomas Monjalon, dev

On 6 April 2016 at 22:21, Dave Neary <dneary@redhat.com> wrote:

> Hi,
>
> On 04/06/2016 08:07 AM, Panu Matilainen wrote:
> >> +1: it's a bit weird to keep both, especially for a long while, that
> >> every time we turn a rte_ prefix to dpdk_ prefix, we break applications.
> >> Instead of breaking applications many times, I'd prefer to break once.
> >> Therefore, applications could do a simple global rte_ -> dpdk_
> >> substitute:
> >> it doesn't sound that painful then.
>

+1

Either all types and symbols use dpdk_ or rte_. It probably makes more
sense dpdk_, but to me it is not that important.

If it has to be changed, it might be a good idea to do it in this release,
now that version numbering format also changes.

>
> > I concur. If (and I think that should be a pretty big IF) the prefix is
> > to be changed then its better done in one fast sweep than gradually.
> >
> > Gratuitious (or nearly so) change is always extremely annoying, and the
> > longer it takes the more painful it is. Application developers wont much
> > care what the prefix is as long as its consistent, but if they're forced
> > to track prefix changes across several releases with different libraries
> > moving at different pace, they WILL be calling for bloody murder :)
>
> How about the idea of creating (at switch over time) an optionally
> installable dpdk_compat package that just has a list of #defines for the
> old symbols pointing them at the new symbols? That would also allow
> people with old applications to update DPDK without having to modify
> their applications.
>

You would also have to add all typedefs for type names.

Why bothering? Moving from 2.2 to 16.04 requires recompiling your
application (ABI changes), and is as simple as sed -e 's/rte_/dpdk_/g' in
all the application code base.

Marc


>
> Thanks,
> Dave.
>
> --
> Dave Neary - NFV/SDN Community Strategy
> Open Source and Standards, Red Hat - http://community.redhat.com
> Ph: +1-978-399-2182 / Cell: +1-978-799-3338
>

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] DPDK namespace
  @ 2016-04-07  9:33  3%   ` Panu Matilainen
  2016-04-07 10:16  5%     ` Marc Sune
  0 siblings, 1 reply; 200+ results
From: Panu Matilainen @ 2016-04-07  9:33 UTC (permalink / raw)
  To: Thomas Monjalon, dev; +Cc: techboard

On 04/07/2016 12:18 PM, Thomas Monjalon wrote:
> Thank you everyone for the feedbacks.
>
> 2016-04-05 15:56, Thomas Monjalon:
>> The goal of this email is to get some feedback on how important it is
>> to fix the DPDK namespace.
>
> Everybody agree every symbols must be prefixed. Checking and fixing the
> namespace consistency will be in the roadmap.
>
> It seems most of you agree renaming would be a nice improvement but not
> so important.
> The main drawback is the induced backporting pain, even if we have
> some scripts to convert the patches to the old namespace.
> Note: the backports can be in DPDK itself or in the applications.
>
>> If there is enough agreement that we should do something, I suggest to
>> introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
>> during some time.
>> We could start using the new prefix for the new APIs (example: crypto)
>> or when there is a significant API break (example: mempool).
>
> The slow change has been clearly rejected in favor of a complete change
> in one patch.
> The timing was also discussed as it could impact the pending patches.
> So it would be done at the end or the beginning of a release.
> Marc suggests to do it for 16.04 as the numbering scheme has changed.

Just noting that it cannot be done in 16.04 because the ABI policy 
requires a deprecation cycle of at least one major release for every 
breakage. And we're discussing a total 100% breakage of everything here, 
even if its just a simple rename.

	- Panu -

> There is no strong conclusion at this point because we need to decide
> wether the renaming deserves to be done or never.
> I suggest to take the inputs from the technical board.
>
> Do not hesitate to comment. Thanks
>

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-07  9:33  3%   ` Panu Matilainen
@ 2016-04-07 10:16  5%     ` Marc Sune
  2016-04-07 11:51  9%       ` [dpdk-dev] On DPDK ABI policy Panu Matilainen
  2016-04-07 21:48  0%       ` [dpdk-dev] DPDK namespace Matthew Hall
  0 siblings, 2 replies; 200+ results
From: Marc Sune @ 2016-04-07 10:16 UTC (permalink / raw)
  To: Panu Matilainen; +Cc: Thomas Monjalon, dev, techboard

2016-04-07 11:33 GMT+02:00 Panu Matilainen <pmatilai@redhat.com>:

> On 04/07/2016 12:18 PM, Thomas Monjalon wrote:
>
>> Thank you everyone for the feedbacks.
>>
>> 2016-04-05 15:56, Thomas Monjalon:
>>
>>> The goal of this email is to get some feedback on how important it is
>>> to fix the DPDK namespace.
>>>
>>
>> Everybody agree every symbols must be prefixed. Checking and fixing the
>> namespace consistency will be in the roadmap.
>>
>> It seems most of you agree renaming would be a nice improvement but not
>> so important.
>> The main drawback is the induced backporting pain, even if we have
>> some scripts to convert the patches to the old namespace.
>> Note: the backports can be in DPDK itself or in the applications.
>>
>> If there is enough agreement that we should do something, I suggest to
>>> introduce the "dpdk_" prefix slowly and live with both "rte_" and "dpdk_"
>>> during some time.
>>> We could start using the new prefix for the new APIs (example: crypto)
>>> or when there is a significant API break (example: mempool).
>>>
>>
>> The slow change has been clearly rejected in favor of a complete change
>> in one patch.
>> The timing was also discussed as it could impact the pending patches.
>> So it would be done at the end or the beginning of a release.
>> Marc suggests to do it for 16.04 as the numbering scheme has changed.
>>
>
> Just noting that it cannot be done in 16.04 because the ABI policy
> requires a deprecation cycle of at least one major release for every
> breakage. And we're discussing a total 100% breakage of everything here,
> even if its just a simple rename.


I keep not understanding the ABI policy, and particularly why ABI changes
have to be announced once cycle before _if_ there is already at least one
ABI change proposed. DPDK applications will have to recompile anyway.

This aspect of the policy only slows down DPDK development and it pollutes
the repository with commits announcing ABI changes that are irrelevant
after 2 cycles, as (code) diffs show that already (not mentioning NEXT_ABI
complexity and extra LOCs).

Maintaining LTS releases, and enforcing bug fixing in old LTS first,
upstreaming bugfixes is to me a much better approach to solve backwards
compatibility issues.

But this is probably another discussion.

Marc


>
>         - Panu -
>
>
> There is no strong conclusion at this point because we need to decide
>> wether the renaming deserves to be done or never.
>> I suggest to take the inputs from the technical board.
>>
>> Do not hesitate to comment. Thanks
>>
>>
>

^ permalink raw reply	[relevance 5%]

* [dpdk-dev] On DPDK ABI policy
  2016-04-07 10:16  5%     ` Marc Sune
@ 2016-04-07 11:51  9%       ` Panu Matilainen
  2016-04-07 21:52  4%         ` Matthew Hall
  2016-04-08  8:47  9%         ` Marc Sune
  2016-04-07 21:48  0%       ` [dpdk-dev] DPDK namespace Matthew Hall
  1 sibling, 2 replies; 200+ results
From: Panu Matilainen @ 2016-04-07 11:51 UTC (permalink / raw)
  To: Marc Sune; +Cc: Thomas Monjalon, dev

[ change of subject since this is about ABI policy, not namespacing ]

On 04/07/2016 01:16 PM, Marc Sune wrote:
>
>
> 2016-04-07 11:33 GMT+02:00 Panu Matilainen <pmatilai@redhat.com
> <mailto:pmatilai@redhat.com>>:
>
>     On 04/07/2016 12:18 PM, Thomas Monjalon wrote:
>
>         Thank you everyone for the feedbacks.
>
>         2016-04-05 15:56, Thomas Monjalon:
>
>             The goal of this email is to get some feedback on how
>             important it is
>             to fix the DPDK namespace.
>
>
>         Everybody agree every symbols must be prefixed. Checking and
>         fixing the
>         namespace consistency will be in the roadmap.
>
>         It seems most of you agree renaming would be a nice improvement
>         but not
>         so important.
>         The main drawback is the induced backporting pain, even if we have
>         some scripts to convert the patches to the old namespace.
>         Note: the backports can be in DPDK itself or in the applications.
>
>             If there is enough agreement that we should do something, I
>             suggest to
>             introduce the "dpdk_" prefix slowly and live with both
>             "rte_" and "dpdk_"
>             during some time.
>             We could start using the new prefix for the new APIs
>             (example: crypto)
>             or when there is a significant API break (example: mempool).
>
>
>         The slow change has been clearly rejected in favor of a complete
>         change
>         in one patch.
>         The timing was also discussed as it could impact the pending
>         patches.
>         So it would be done at the end or the beginning of a release.
>         Marc suggests to do it for 16.04 as the numbering scheme has
>         changed.
>
>
>     Just noting that it cannot be done in 16.04 because the ABI policy
>     requires a deprecation cycle of at least one major release for every
>     breakage. And we're discussing a total 100% breakage of everything
>     here, even if its just a simple rename.
>
>
> I keep not understanding the ABI policy, and particularly why ABI
> changes have to be announced once cycle before _if_ there is already at
> least one ABI change proposed. DPDK applications will have to recompile
> anyway.

The point is to allow API/ABI consumers to assess in advance what sort 
of pains can they expect when moving their applications from one version 
to another. Otherwise all sorts of massive changes could ride the wave 
of whatever "change 16bit struct member to 32bit" trivialities that are 
nevertheless ABI breaks.

There have already been quite a few exceptions to the rule when the ABI 
is already being broken, so its not entirely rigid. Another point that 
migth warrant some tweaking to the policy is the "core" libraries 
depending on each other so an ABI break in any one of them forces 
recompile of everything anyway.

> This aspect of the policy only slows down DPDK development and it

One could also think that slowing down development and forcing people to 
think ahead are not entirely unintentional or unwanted side-effects :)

Look at the latest librte_vhost initiative to remove unnecessarily 
exposed structs to avoid having to deal with ABI breakages all the time: 
the policy is effectively encouraging people into better library design.

> pollutes the repository with commits announcing ABI changes that are
> irrelevant after 2 cycles, as (code) diffs show that already (not
> mentioning NEXT_ABI complexity and extra LOCs).

Fully agreed on NEXT_ABI, I never liked it at all.

> Maintaining LTS releases, and enforcing bug fixing in old LTS first,
> upstreaming bugfixes is to me a much better approach to solve backwards
> compatibility issues.

LTS releases could help the situation somewhat, but then again people 
tend to still want those new fancy things backported (you know, have the 
cake and eat it too) but that can't be done because of ABI breakage, so 
they're forced to run the latest version anyway.

> But this is probably another discussion.

Yup, changed subject to avoid mixing it up with the namespace discussion 
too much.

	- Panu -

^ permalink raw reply	[relevance 9%]

* [dpdk-dev] [PATCH] doc: announce API changes for device objects
@ 2016-04-07 15:33  5% David Marchand
  2016-04-07 15:46  0% ` Jan Viktorin
  0 siblings, 1 reply; 200+ results
From: David Marchand @ 2016-04-07 15:33 UTC (permalink / raw)
  To: dev; +Cc: viktorin, olivier.matz, thomas.monjalon

Following discussions with Jan, here is a deprecation notice to prepare for
hotplug and rte_device changes to come in 16.07.

Signed-off-by: David Marchand <david.marchand@6wind.com>
---
 doc/guides/rel_notes/deprecation.rst | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 98d5529..d749e5d 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -8,6 +8,18 @@ API and ABI deprecation notices are to be posted here.
 Deprecation Notices
 -------------------
 
+* 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
+  so that it only cares about hotplugging.
+
+* Structures embodying pci and vdev devices are going to be reworked to
+  integrate new common rte_device / rte_driver objects (see
+  http://dpdk.org/ml/archives/dev/2016-January/031390.html).
+  ethdev and crypto libraries will then only handle those objects so that they
+  do not need to care about the kind of devices that are being used, making it
+  easier to add new buses later.
+
 * The EAL function pci_config_space_set is deprecated in release 16.04
   and will be removed from 16.07.
   Macros CONFIG_RTE_PCI_CONFIG, CONFIG_RTE_PCI_EXTENDED_TAG and
-- 
1.9.1

^ permalink raw reply	[relevance 5%]

* Re: [dpdk-dev] [PATCH] doc: announce API changes for device objects
  2016-04-07 15:33  5% [dpdk-dev] [PATCH] doc: announce API changes for device objects David Marchand
@ 2016-04-07 15:46  0% ` Jan Viktorin
  2016-04-07 17:00  0%   ` David Marchand
  0 siblings, 1 reply; 200+ results
From: Jan Viktorin @ 2016-04-07 15:46 UTC (permalink / raw)
  To: David Marchand; +Cc: dev, olivier.matz, thomas.monjalon

On Thu,  7 Apr 2016 17:33:17 +0200
David Marchand <david.marchand@6wind.com> wrote:

> Following discussions with Jan, here is a deprecation notice to prepare for
> hotplug and rte_device changes to come in 16.07.
> 
> Signed-off-by: David Marchand <david.marchand@6wind.com>
> ---
>  doc/guides/rel_notes/deprecation.rst | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index 98d5529..d749e5d 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -8,6 +8,18 @@ API and ABI deprecation notices are to be posted here.
>  Deprecation Notices
>  -------------------
>  
> +* 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
> +  so that it only cares about hotplugging.
> +
> +* Structures embodying pci and vdev devices are going to be reworked to
> +  integrate new common rte_device / rte_driver objects (see
> +  http://dpdk.org/ml/archives/dev/2016-January/031390.html).
> +  ethdev and crypto libraries will then only handle those objects so that they
> +  do not need to care about the kind of devices that are being used, making it
> +  easier to add new buses later.

As a result, the current rte_driver structure will be renamed to
rte_module and probably reworked in some way due to its semantics and
potential name clash with the new rte_driver struct.

Regards
Jan

> +
>  * The EAL function pci_config_space_set is deprecated in release 16.04
>    and will be removed from 16.07.
>    Macros CONFIG_RTE_PCI_CONFIG, CONFIG_RTE_PCI_EXTENDED_TAG and



-- 
   Jan Viktorin                  E-mail: Viktorin@RehiveTech.com
   System Architect              Web:    www.RehiveTech.com
   RehiveTech
   Brno, Czech Republic

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v1] doc: fix release notes for 16.04
@ 2016-04-07 16:02  8% John McNamara
  0 siblings, 0 replies; 200+ results
From: John McNamara @ 2016-04-07 16:02 UTC (permalink / raw)
  To: dev; +Cc: thomas.monjalon, John McNamara

Fix grammar, spelling and formatting of DPDK 16.04 release notes.

Signed-off-by: John McNamara <john.mcnamara@intel.com>
---
 doc/guides/rel_notes/release_16_04.rst | 266 +++++++++++++--------------------
 1 file changed, 104 insertions(+), 162 deletions(-)

diff --git a/doc/guides/rel_notes/release_16_04.rst b/doc/guides/rel_notes/release_16_04.rst
index e293960..200053c 100644
--- a/doc/guides/rel_notes/release_16_04.rst
+++ b/doc/guides/rel_notes/release_16_04.rst
@@ -2,39 +2,9 @@ DPDK Release 16.04
 ==================
 
 
-**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_04.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.
-
 * **Added function to check primary process state.**
 
   A new function ``rte_eal_primary_proc_alive()`` has been added
@@ -45,141 +15,145 @@ This section should contain new features added in this release. Sample format:
 * **Enabled bulk allocation of mbufs.**
 
   A new function ``rte_pktmbuf_alloc_bulk()`` has been added to allow the user
-  to allocate a bulk of mbufs.
+  to bulk allocate mbufs.
 
 * **Added device link speed capabilities.**
 
-  The structure ``rte_eth_dev_info`` has now a ``speed_capa`` bitmap, which
-  allows the application to know the supported speeds of each device.
+  The structure ``rte_eth_dev_info`` now has a ``speed_capa`` bitmap, which
+  allows the application to determine the supported speeds of each device.
 
 * **Added bitmap of link speeds to advertise.**
 
-  Allow defining a set of advertised speeds for auto-negotiation,
+  Added a feature to allow the definition of a set of advertised speeds for auto-negotiation,
   explicitly disabling link auto-negotiation (single speed)
   and full auto-negotiation.
 
 * **Added new poll-mode driver for Amazon Elastic Network Adapters (ENA).**
 
-  The driver operates variety of ENA adapters through feature negotiation
+  The driver operates for a variety of ENA adapters through feature negotiation
   with the adapter and upgradable commands set.
-  ENA driver handles PCI Physical and Virtual ENA functions.
+  The ENA driver handles PCI Physical and Virtual ENA functions.
 
-* **Restored vmxnet3 Tx data ring.**
+* **Restored vmxnet3 TX data ring.**
 
-  Tx data ring has been shown to improve small pkt forwarding performance
-  on vSphere environment.
+  TX data ring has been shown to improve small packet forwarding performance
+  on the vSphere environment.
 
-* **Added vmxnet3 Tx L4 checksum offload.**
+* **Added vmxnet3 TX L4 checksum offload.**
 
-  Support TCP/UDP checksum offload.
+  Added support for TCP/UDP checksum offload to vmxnet3.
 
 * **Added vmxnet3 TSO support.**
 
+  Added support for TSO to vmxnet3.
+
 * **Added vmxnet3 support for jumbo frames.**
 
   Added support for linking multi-segment buffers together to
   handle Jumbo packets.
 
-* **Virtio 1.0.**
+* **Enabled Virtio 1.0 support.**
 
-  Enabled virtio 1.0 support for virtio pmd driver.
+  Enabled Virtio 1.0 support for Virtio pmd driver.
 
-* **Supported virtio for ARM.**
+* **Supported Virtio for ARM.**
 
-  Enabled virtio support for armv7/v8. Tested for arm64.
-  Virtio for arm support VFIO-noiommu mode only.
-  Virtio can work with other non-x86 arch too like powerpc.
+  Enabled Virtio support for ARMv7/v8. Tested for ARM64.
+  Virtio for ARM supports VFIO-noiommu mode only.
+  Virtio can work with other non-x86 architectures as well, like PowerPC.
 
-* **Supported virtio offload in vhost-user.**
+* **Supported Virtio offload in vhost-user.**
 
-  Add the offload and negotiation of checksum and TSO between vhost-user and
-  vanilla Linux virtio guest.
+  Added the offload and negotiation of checksum and TSO between vhost-user and
+  vanilla Linux Virtio guest.
 
 * **Added vhost-user live migration support.**
 
 * **Added vhost driver.**
 
-  Added virtual PMD that wraps librte_vhost.
+  Added a virtual PMD that wraps ``librte_vhost``.
 
 * **Added multicast promiscuous mode support on VF for ixgbe.**
 
-  Added multicast promiscuous mode support on ixgbe VF driver. So all the VFs
+  Added multicast promiscuous mode support for the ixgbe VF driver so all VFs
   can receive the multicast packets.
 
-  Please note if we want to use this promiscuous mode, we need both PF and VF
-  driver to support it. The reason is this VF feature is configured on PF.
-  If use kernel PF driver + dpdk VF driver, make sure kernel PF driver support
-  VF multicast promiscuous mode. If use dpdk PF + dpdk VF, better make sure PF
-  driver is the same version as VF.
+  Please note if you want to use this promiscuous mode, you need both PF and VF
+  driver to support it. The reason is that this VF feature is configured in the PF.
+  If you use kernel PF driver and the dpdk VF driver, make sure the kernel PF driver supports
+  VF multicast promiscuous mode. If you use dpdk PF and  dpdk VF ensure the PF
+  driver is the same version as the VF.
 
 * **Added support for E-tag on X550.**
 
-  E-tag is defined in 802.1br. Please reference
-  http://www.ieee802.org/1/pages/802.1br.html.
+  E-tag is defined in `802.1BR - Bridge Port Extension <http://www.ieee802.org/1/pages/802.1br.html>`_.
 
-  This feature is for VF, but the settings are on PF. It means
-  the CLIs should be used on PF, but some of their effects will be shown on VF.
-  The forwarding of E-tag packets based on GRP and E-CID_base will have effect
-  on PF. Theoretically the E-tag packets can be forwarded to any pool/queue.
-  But normally we'd like to forward the packets to the pools/queues belonging
-  to the VFs. And E-tag insertion and stripping will have effect on VFs. When
-  VF receives E-tag packets, it should strip the E-tag. When VF transmits
-  packets, it should insert the E-tag. Both can be offloaded.
+  This feature is for the VF, but the settings are on the PF. It means
+  the CLIs should be used on the PF, but some of their effects will be shown on the VF.
+  The forwarding of E-tag packets based on GRP and E-CID_base will have an effect
+  on the PF. Theoretically, the E-tag packets can be forwarded to any pool/queue
+  but normally we'd like to forward the packets to the pools/queues belonging
+  to the VFs. And E-tag insertion and stripping will have an effect on VFs. When
+  a VF receives E-tag packets it should strip the E-tag. When the VF transmits
+  packets, it should insert the E-tag. Both actions can be offloaded.
 
   When we want to use this E-tag support feature, the forwarding should be
-  enabled to forward the packets received by PF to indicated VFs. And insertion
-  and stripping should be enabled for VFs to offload the effort to HW.
+  enabled to forward the packets received by the PF to the indicated VFs. And insertion
+  and stripping should be enabled for VFs to offload the effort to hardware.
+
+  Features added:
 
   * Support E-tag offloading of insertion and stripping.
   * Support Forwarding E-tag packets to pools based on
     GRP and E-CID_base.
 
-* **Added support for VxLAN & NVGRE checksum off-load on X550.**
+* **Added support for VxLAN and NVGRE checksum off-load on X550.**
 
-  * Added support for VxLAN & NVGRE RX/TX checksum off-load on
+  * Added support for VxLAN and NVGRE RX/TX checksum off-load on
     X550. RX/TX checksum off-load is provided on both inner and
     outer IP header and TCP header.
   * Added functions to support VxLAN port configuration. The
     default VxLAN port number is 4789 but this can be updated
     programmatically.
 
-* **Added new X550EM_a devices.**
+* **Added support for new X550EM_a devices.**
 
-  Added new X550EM_a devices and their mac types, X550EM_a and X550EM_a_vf.
-  Updated the code to use the new devices and mac types.
+  Added support for new X550EM_a devices and their MAC types, X550EM_a and X550EM_a_vf.
+  Updated the relevant PMD to use the new devices and MAC types.
 
 * **Added x550em_x V2 device support.**
 
-  Only x550em_x V1 was supported before. Now V2 is supported.
+  Added support for x550em_x V2 device. Only x550em_x V1 was supported before.
   A mask for V1 and V2 is defined and used to support both.
 
 * **Supported link speed auto-negotiation on X550EM_X**
 
-  Normally the auto-negotiation is supported by FW. SW need not care about
-  that. But on x550em_x, FW doesn't support auto-neg. As the ports of x550em_x
-  are 10G, if we connect the port with a peer which is 1G, the link will always
+  Normally the auto-negotiation is supported by firmware and software doesn't care about
+  it. But on x550em_x, firmware doesn't support auto-negotiation. As the ports of x550em_x
+  are 10GbE, if we connect the port with a peer which is 1GbE, the link will always
   be down.
-  We added the support of auto-neg by SW to avoid this link down issue.
+  We added the support for auto-negotiation by software to avoid this link down issue.
 
-* **Added sw-firmware sync on X550EM_a.**
+* **Added software-firmware sync on X550EM_a.**
 
-  Added support for sw-firmware sync for resource sharing.
-  Use the PHY token, shared between sw-fw for PHY access on X550EM_a.
+  Added support for software-firmware sync for resource sharing.
+  Use the PHY token, shared between software-firmware for PHY access on X550EM_a.
 
 * **Updated the i40e base driver.**
 
   The i40e base driver was updated with changes including the
   following:
 
-  * Use Rx control AQ commands to read/write Rx control registers.
+  * Use RX control AQ commands to read/write RX control registers.
   * Add new X722 device IDs, and removed X710 one was never used.
   * Expose registers for HASH/FD input set configuring.
 
 * **Enabled PCI extended tag for i40e.**
 
-  It enabled extended tag by checking and writing corresponding PCI config
-  space bytes, to boost the performance. In the meanwhile, it deprecated the
-  legacy way via reading/writing sysfile supported by kernel module igb_uio.
+  Enabled extended tag for i40e by checking and writing corresponding PCI config
+  space bytes, to boost the performance.
+  The legacy method of reading/writing sysfile supported by kernel module igb_uio
+  is now deprecated.
 
 * **Added i40e support for setting mac addresses.**
 
@@ -197,22 +171,22 @@ This section should contain new features added in this release. Sample format:
 
 * **Added PF reset event reporting in i40e VF driver.**
 
-* **Added fm10k Rx interrupt support.**
+* **Added fm10k RX interrupt support.**
 
-* **Optimized fm10k Tx.**
+* **Optimized fm10k TX.**
 
-  * Free multiple mbufs at a time to reduce freeing mbuf cycles.
+  Optimized fm10k TX by freeing multiple mbufs at a time.
 
-* **Handled error flags in fm10k vector Rx.**
+* **Handled error flags in fm10k vector RX.**
 
-  Parse err flags in Rx desc and set error bits in mbuf with vector instructions.
+  Parse error flags in RX descriptor and set error bits in mbuf with vector instructions.
 
 * **Added fm10k FTAG based forwarding support.**
 
 * **Added mlx5 flow director support.**
 
-  Added flow director support (RTE_FDIR_MODE_PERFECT and
-  RTE_FDIR_MODE_PERFECT_MAC_VLAN).
+  Added flow director support (``RTE_FDIR_MODE_PERFECT`` and
+  ``RTE_FDIR_MODE_PERFECT_MAC_VLAN``).
 
   Only available with Mellanox OFED >= 3.2.
 
@@ -238,7 +212,7 @@ This section should contain new features added in this release. Sample format:
 
 * **Added mlx5 optional packet padding by HW.**
 
-  Added an option to make PCI bus transactions rounded to multiple of a
+  Added an option to make PCI bus transactions rounded to a multiple of a
   cache line size for better alignment.
 
   Only available with Mellanox OFED >= 3.2.
@@ -249,10 +223,10 @@ This section should contain new features added in this release. Sample format:
 
   Only available with Mellanox OFED >= 3.2.
 
-* **Changed szedata2 type of driver from vdev to pdev.**
+* **Changed szedata2 driver type from vdev to pdev.**
 
   Previously szedata2 device had to be added by ``--vdev`` option.
-  Now szedata2 PMD recognises the device automatically during EAL
+  Now szedata2 PMD recognizes the device automatically during EAL
   initialization.
 
 * **Added szedata2 functions for setting link up/down.**
@@ -261,17 +235,17 @@ This section should contain new features added in this release. Sample format:
 
 * **Added af_packet dynamic removal function.**
 
-  Af_packet device can now be detached using API, like other PMD devices.
+  An af_packet device can now be detached using the API, like other PMD devices.
 
 * **Increased number of next hops for LPM IPv4 to 2^24.**
 
-  The next_hop field is extended from 8 bits to 24 bits for IPv4.
+  The ``next_hop`` field has been extended from 8 bits to 24 bits for IPv4.
 
 * **Added support of SNOW 3G (UEA2 and UIA2) for Intel Quick Assist devices.**
 
-  Enabled support for SNOW 3G wireless algorithm for Intel Quick Assist devices.
-  Support for cipher only, hash only is also provided
-  along with alg-chaining operations.
+  Enabled support for the SNOW 3G wireless algorithm for Intel Quick Assist devices.
+  Support for cipher-only and  hash-only is also provided
+  along with algorithm-chaining operations.
 
 * **Added SNOW3G SW PMD.**
 
@@ -281,31 +255,29 @@ This section should contain new features added in this release. Sample format:
 * **Added AES GCM PMD.**
 
   Added new Crypto PMD to support AES-GCM authenticated encryption and
-  authenticated decryption in SW.
+  authenticated decryption in software.
 
 * **Added NULL Crypto PMD**
 
-  Added new Crypto PMD to support null crypto operations in SW.
+  Added new Crypto PMD to support null crypto operations in software.
 
 * **Improved IP Pipeline Application.**
 
   The following features have been added to ip_pipeline application;
 
   * Added CPU utilization measurement and idle cycle rate computation.
-  * Added link idenfication support through existing port-mask option or by
+  * Added link identification support through existing port-mask option or by
     specifying PCI device in every LINK section in the configuration file.
   * Added load balancing support in passthrough pipeline.
 
 * **Added IPsec security gateway example.**
 
-  New application implementing an IPsec Security Gateway.
+  Added a new application implementing an IPsec Security Gateway.
 
 
 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.
@@ -313,21 +285,17 @@ This section should contain bug fixes added to the relevant sections. Sample for
   Add the entries in alphabetic order in the relevant sections below.
 
 
-EAL
-~~~
-
-
 Drivers
 ~~~~~~~
 
 * **ethdev: Fixed overflow for 100Gbps.**
 
-  100Gbps in Mbps (100000) was exceeding 16-bit max value of ``link_speed``
+  100Gbps in Mbps (100000) was exceeding the 16-bit max value of ``link_speed``
   in ``rte_eth_link``.
 
 * **ethdev: Fixed byte order consistency between fdir flow and mask.**
 
-  Fixed issue in ethdev library that the structure for setting
+  Fixed issue in ethdev library where the structure for setting
   fdir's mask and flow entry was not consistent in byte ordering.
 
 * **cxgbe: Fixed crash due to incorrect size allocated for RSS table.**
@@ -338,12 +306,12 @@ Drivers
 
 * **cxgbe: Fixed setting wrong device MTU.**
 
-  Fixed an incorrect device MTU being set due to ethernet header and
+  Fixed an incorrect device MTU being set due to the Ethernet header and
   CRC lengths being added twice.
 
 * **ixgbe: Fixed zeroed VF mac address.**
 
-  Resolved an issue where VF mac address is zeroed out in cases where the VF
+  Resolved an issue where the VF MAC address is zeroed out in cases where the VF
   driver is loaded while the PF interface is down.
   The solution is to only set it when we get an ACK from the PF.
 
@@ -370,7 +338,7 @@ Drivers
   It generates a MAC address for each VFs during PF host initialization,
   and keeps the VF MAC address the same among different VF launch.
 
-* **i40e: Fixed failure of reading/writing Rx control registers.**
+* **i40e: Fixed failure of reading/writing RX control registers.**
 
   Fixed i40e issue of failing to read/write rx control registers when
   under stress with traffic, which might result in application launch
@@ -378,14 +346,14 @@ Drivers
 
 * **i40e: Enabled vector driver by default.**
 
-  Previously, vector driver is disabled by default as it cannot fill packet type
-  info for l3fwd to work well. Now there is an option for l3fwd to analysis
-  packet type softly, so enable vector driver by default.
+  Previously, vector driver was disabled by default as it couldn't fill packet type
+  info for l3fwd to work well. Now there is an option for l3fwd to analyze
+  the packet type so the vector driver is enabled by default.
 
 * **i40e: Fixed link info of VF.**
 
-  Previously, the VF's link speed kept as 10G and status always was up.
-  It did not change even the physical link's status changed.
+  Previously, the VF's link speed stayed at 10GbE and status always was up.
+  It did not change even when the physical link's status changed.
   Now this issue is fixed to make VF's link info consistent with physical link.
 
 * **mlx5: Fixed possible crash during initialization.**
@@ -394,7 +362,7 @@ Drivers
 
 * **mlx5: Added port type check.**
 
-  Done to prevent port initialization on non-Ethernet link layers and
+  Added port type check to prevent port initialization on non-Ethernet link layers and
   to report an error.
 
 * **mlx5: Applied VLAN filtering to broadcast and IPv6 multicast flows.**
@@ -407,10 +375,10 @@ Drivers
 
 * **aesni_mb: Fixed wrong return value when creating a device.**
 
-  cryptodev_aesni_mb_init() was returning the device id of the device created,
-  instead of 0 (when success), that rte_eal_vdev_init() expects.
-  This made impossible the creation of more than one aesni_mb device
-  from command line.
+  The ``cryptodev_aesni_mb_init()`` function was returning the device id of the device created,
+  instead of 0 (on success) that ``rte_eal_vdev_init()`` expects.
+  This made it impossible to create more than one aesni_mb device
+  from the command line.
 
 * **qat: Fixed AES GCM decryption.**
 
@@ -424,7 +392,7 @@ Libraries
 * **hash: Fixed CRC32c hash computation for non multiple of 4 bytes sizes.**
 
   Fix crc32c hash functions to return a valid crc32c value for data lengths
-  not multiple of 4 bytes.
+  not a multiple of 4 bytes.
 
 * **hash: Fixed hash library to support multi-process mode.**
 
@@ -440,7 +408,7 @@ Libraries
   ``rte_errno`` to ``EEXIST`` when the object name already exists. This is
   the behavior described in the API documentation in the header file.
   The previous behavior was to return a pointer to the existing object in
-  that case, preventing the caller to know if the object had to be freed
+  that case, preventing the caller from knowing if the object had to be freed
   or not.
 
 * **lpm: Fixed return value when allocating an existing object.**
@@ -449,7 +417,7 @@ Libraries
   ``rte_errno`` to ``EEXIST`` when the object name already exists. This is
   the behavior described in the API documentation in the header file.
   The previous behavior was to return a pointer to the existing object in
-  that case, preventing the caller to know if the object had to be freed
+  that case, preventing the caller from knowing if the object had to be freed
   or not.
 
 * **librte_port: Fixed segmentation fault for ring and ethdev writer nodrop.**
@@ -468,39 +436,19 @@ Examples
 
 * **l3fwd: Fixed using packet type blindly.**
 
-  l3fwd makes use of packet type information without even query if devices or PMDs
-  really set it. For those don't set ptypes, add an option to parse it softly.
+  l3fwd makes use of packet type information without querying if devices or PMDs
+  really set it. For those devices that don't set ptypes, add an option to parse it.
 
 * **examples/vhost: Fixed frequent mbuf allocation failure.**
 
-  vhost-switch often fails to allocate mbuf when dequeue from vring because it
+  The vhost-switch often fails to allocate mbuf when dequeue from vring because it
   wrongly calculates the number of mbufs needed.
 
 
-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.
-
-* The ethdev statistics counter imissed is considered to be independent of ierrors.
+* The ethdev statistics counter ``imissed`` is considered to be independent of ``ierrors``.
   All drivers are now counting the missed packets only once, i.e. drivers will
   not increment ierrors anymore for missed packets.
 
@@ -524,13 +472,13 @@ This section should contain API changes. Sample format:
 * A parameter ``vlan_type`` has been added to the function
   ``rte_eth_dev_set_vlan_ether_type``.
 
-* AF_packet device init function is no longer public. Device should be attached
-  with API.
+* The af_packet device init function is no longer public. The device should be attached
+  via the API.
 
 * The LPM ``next_hop`` field is extended from 8 bits to 24 bits for IPv4
   while keeping ABI compatibility.
 
-* A new ``rte_lpm_config`` structure is used so LPM library will allocate
+* A new ``rte_lpm_config`` structure is used so the LPM library will allocate
   exactly the amount of memory which is necessary to hold application’s rules.
   The previous ABI is kept for compatibility.
 
@@ -542,10 +490,6 @@ This section should contain API changes. Sample format:
 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.
-
 * The RETA entry size in ``rte_eth_rss_reta_entry64`` has been increased
   from 8-bit to 16-bit.
 
@@ -558,8 +502,6 @@ ABI Changes
 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
-- 
2.5.0

^ permalink raw reply	[relevance 8%]

* Re: [dpdk-dev] [PATCH] doc: announce API changes for device objects
  2016-04-07 15:46  0% ` Jan Viktorin
@ 2016-04-07 17:00  0%   ` David Marchand
  2016-04-07 17:09  3%     ` Jan Viktorin
  0 siblings, 1 reply; 200+ results
From: David Marchand @ 2016-04-07 17:00 UTC (permalink / raw)
  To: Jan Viktorin; +Cc: dev, Olivier Matz, Thomas Monjalon

On Thu, Apr 7, 2016 at 5:46 PM, Jan Viktorin <viktorin@rehivetech.com> wrote:
> On Thu,  7 Apr 2016 17:33:17 +0200
> David Marchand <david.marchand@6wind.com> wrote:
>
>> Following discussions with Jan, here is a deprecation notice to prepare for
>> hotplug and rte_device changes to come in 16.07.
>>
>> Signed-off-by: David Marchand <david.marchand@6wind.com>
>> ---
>>  doc/guides/rel_notes/deprecation.rst | 12 ++++++++++++
>>  1 file changed, 12 insertions(+)
>>
>> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
>> index 98d5529..d749e5d 100644
>> --- a/doc/guides/rel_notes/deprecation.rst
>> +++ b/doc/guides/rel_notes/deprecation.rst
>> @@ -8,6 +8,18 @@ API and ABI deprecation notices are to be posted here.
>>  Deprecation Notices
>>  -------------------
>>
>> +* 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
>> +  so that it only cares about hotplugging.
>> +
>> +* Structures embodying pci and vdev devices are going to be reworked to
>> +  integrate new common rte_device / rte_driver objects (see
>> +  http://dpdk.org/ml/archives/dev/2016-January/031390.html).
>> +  ethdev and crypto libraries will then only handle those objects so that they
>> +  do not need to care about the kind of devices that are being used, making it
>> +  easier to add new buses later.
>
> As a result, the current rte_driver structure will be renamed to
> rte_module and probably reworked in some way due to its semantics and
> potential name clash with the new rte_driver struct.

If we just introduce some macros like RTE_MODULE_INIT() /
RTE_MODULE_EXIT(), we don't need a rte_module object at the moment ?

-- 
David Marchand

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce API changes for device objects
  2016-04-07 17:00  0%   ` David Marchand
@ 2016-04-07 17:09  3%     ` Jan Viktorin
  2016-04-07 17:24  0%       ` David Marchand
  0 siblings, 1 reply; 200+ results
From: Jan Viktorin @ 2016-04-07 17:09 UTC (permalink / raw)
  To: David Marchand; +Cc: dev, Olivier Matz, Thomas Monjalon

On Thu, 7 Apr 2016 19:00:43 +0200
David Marchand <david.marchand@6wind.com> wrote:

> On Thu, Apr 7, 2016 at 5:46 PM, Jan Viktorin <viktorin@rehivetech.com> wrote:
> > On Thu,  7 Apr 2016 17:33:17 +0200
> > David Marchand <david.marchand@6wind.com> wrote:
> >  
> >> Following discussions with Jan, here is a deprecation notice to prepare for
> >> hotplug and rte_device changes to come in 16.07.
> >>
> >> Signed-off-by: David Marchand <david.marchand@6wind.com>
> >> ---
> >>  doc/guides/rel_notes/deprecation.rst | 12 ++++++++++++
> >>  1 file changed, 12 insertions(+)
> >>
> >> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> >> index 98d5529..d749e5d 100644
> >> --- a/doc/guides/rel_notes/deprecation.rst
> >> +++ b/doc/guides/rel_notes/deprecation.rst
> >> @@ -8,6 +8,18 @@ API and ABI deprecation notices are to be posted here.
> >>  Deprecation Notices
> >>  -------------------
> >>
> >> +* 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
> >> +  so that it only cares about hotplugging.
> >> +
> >> +* Structures embodying pci and vdev devices are going to be reworked to
> >> +  integrate new common rte_device / rte_driver objects (see
> >> +  http://dpdk.org/ml/archives/dev/2016-January/031390.html).
> >> +  ethdev and crypto libraries will then only handle those objects so that they
> >> +  do not need to care about the kind of devices that are being used, making it
> >> +  easier to add new buses later.  
> >
> > As a result, the current rte_driver structure will be renamed to
> > rte_module and probably reworked in some way due to its semantics and
> > potential name clash with the new rte_driver struct.  
> 
> If we just introduce some macros like RTE_MODULE_INIT() /
> RTE_MODULE_EXIT(), we don't need a rte_module object at the moment ?
> 

Well, possibly, we don't need it. At least, it might be hidden and not
being a part of the API/ABI. Do you need an ack for this?

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] doc: announce API changes for device objects
  2016-04-07 17:09  3%     ` Jan Viktorin
@ 2016-04-07 17:24  0%       ` David Marchand
  0 siblings, 0 replies; 200+ results
From: David Marchand @ 2016-04-07 17:24 UTC (permalink / raw)
  To: Jan Viktorin; +Cc: dev, Olivier Matz, Thomas Monjalon

On Thu, Apr 7, 2016 at 7:09 PM, Jan Viktorin <viktorin@rehivetech.com> wrote:
> On Thu, 7 Apr 2016 19:00:43 +0200
> David Marchand <david.marchand@6wind.com> wrote:
>> >> Following discussions with Jan, here is a deprecation notice to prepare for
>> >> hotplug and rte_device changes to come in 16.07.
>> > As a result, the current rte_driver structure will be renamed to
>> > rte_module and probably reworked in some way due to its semantics and
>> > potential name clash with the new rte_driver struct.
>>
>> If we just introduce some macros like RTE_MODULE_INIT() /
>> RTE_MODULE_EXIT(), we don't need a rte_module object at the moment ?
>>
>
> Well, possibly, we don't need it. At least, it might be hidden and not
> being a part of the API/ABI. Do you need an ack for this?

Yes, please, the process requires 3 acks for this kind of changes.

-- 
David Marchand

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
  2016-04-06  8:51  4%   ` Azarewicz, PiotrX T
@ 2016-04-07 21:24  4%     ` Thomas Monjalon
  2016-04-12 12:39  4%       ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-07 21:24 UTC (permalink / raw)
  To: Zhang, Roy Fan; +Cc: dev, Azarewicz, PiotrX T, Singh, Jasvinder

> > > Several new fields will be added to structure rte_port_source_params
> > > for source port enhancement with pcap file reading support.
> > >
> > > Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> > > Acked-by: Cristian Dumitrescu <cristian.dumitrescu@intel.com>
> > 
> > Acked-by: Jasvinder Singh <jasvinder.singh@intel.com>
> 
> Acked-by: Piotr Azarewicz <piotrx.t.azarewicz@intel.com>

Applied, thanks

Please send a patch to remove NEXT_ABI early in the 16.07 cycle.

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] DPDK namespace
  2016-04-07 10:16  5%     ` Marc Sune
  2016-04-07 11:51  9%       ` [dpdk-dev] On DPDK ABI policy Panu Matilainen
@ 2016-04-07 21:48  0%       ` Matthew Hall
  1 sibling, 0 replies; 200+ results
From: Matthew Hall @ 2016-04-07 21:48 UTC (permalink / raw)
  To: Marc Sune; +Cc: Panu Matilainen, Thomas Monjalon, dev, techboard

On Thu, Apr 07, 2016 at 12:16:34PM +0200, Marc Sune wrote:
> I keep not understanding the ABI policy, and particularly why ABI changes
> have to be announced once cycle before _if_ there is already at least one
> ABI change proposed. DPDK applications will have to recompile anyway.
> 
> This aspect of the policy only slows down DPDK development and it pollutes
> the repository with commits announcing ABI changes that are irrelevant
> after 2 cycles, as (code) diffs show that already (not mentioning NEXT_ABI
> complexity and extra LOCs).
> 
> Maintaining LTS releases, and enforcing bug fixing in old LTS first,
> upstreaming bugfixes is to me a much better approach to solve backwards
> compatibility issues.
> 
> But this is probably another discussion.

Yes, separate discussion. But I agree 100,000%. As a community member in my 
spare time I get tripped up by NEXT_ABI pollution just trying to submit 
trivial patches all the time, then I don't really have any good idea how to 
fix it, and I have to annoy Thomas with dumb questions across the time zones.

I would really prefer to dump all the drama about ABIs and make a maintenance 
only LTS release which only gets bug fixes people specifically need and not 
random fixes or features.

Matthew.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] On DPDK ABI policy
  2016-04-07 11:51  9%       ` [dpdk-dev] On DPDK ABI policy Panu Matilainen
@ 2016-04-07 21:52  4%         ` Matthew Hall
  2016-04-08  8:29  4%           ` Marc Sune
  2016-04-08  8:47  9%         ` Marc Sune
  1 sibling, 1 reply; 200+ results
From: Matthew Hall @ 2016-04-07 21:52 UTC (permalink / raw)
  To: Panu Matilainen; +Cc: Marc Sune, Thomas Monjalon, dev

On Thu, Apr 07, 2016 at 02:51:35PM +0300, Panu Matilainen wrote:
> LTS releases could help the situation somewhat, but then again
> people tend to still want those new fancy things backported (you
> know, have the cake and eat it too) but that can't be done because
> of ABI breakage, so they're forced to run the latest version anyway.

RH and Debian / Ubuntu don't put features in LTS except extremely rarely. 
Generally only if there's severe functionality breakage or security issues and 
the rest is ignored, and for good reason, as this is much more reliable and 
simple and predictable.

If people are so irrational they can't deal with that simple of a policy, 
NEXT_ABI, LTS, etc. is never going to help them.

If people like to have backported stuff, yes of course we can make trees and 
branches for this, they are basically free in Git. But at that point community 
people in need of LTS forks of features need to step up to the plate to help 
out.

Matthew.

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] On DPDK ABI policy
  2016-04-07 21:52  4%         ` Matthew Hall
@ 2016-04-08  8:29  4%           ` Marc Sune
  0 siblings, 0 replies; 200+ results
From: Marc Sune @ 2016-04-08  8:29 UTC (permalink / raw)
  To: Matthew Hall; +Cc: Panu Matilainen, Thomas Monjalon, dev

2016-04-07 23:52 GMT+02:00 Matthew Hall <mhall@mhcomputing.net>:

> On Thu, Apr 07, 2016 at 02:51:35PM +0300, Panu Matilainen wrote:
> > LTS releases could help the situation somewhat, but then again
> > people tend to still want those new fancy things backported (you
> > know, have the cake and eat it too) but that can't be done because
> > of ABI breakage, so they're forced to run the latest version anyway.
>
> RH and Debian / Ubuntu don't put features in LTS except extremely rarely.
> Generally only if there's severe functionality breakage or security issues
> and
> the rest is ignored, and for good reason, as this is much more reliable and
> simple and predictable.
>
> If people are so irrational they can't deal with that simple of a policy,
> NEXT_ABI, LTS, etc. is never going to help them.
>
> If people like to have backported stuff, yes of course we can make trees
> and
> branches for this, they are basically free in Git. But at that point
> community
> people in need of LTS forks of features need to step up to the plate to
> help
> out.
>

Completely agree.

Marc


>
> Matthew.
>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] On DPDK ABI policy
  2016-04-07 11:51  9%       ` [dpdk-dev] On DPDK ABI policy Panu Matilainen
  2016-04-07 21:52  4%         ` Matthew Hall
@ 2016-04-08  8:47  9%         ` Marc Sune
  1 sibling, 0 replies; 200+ results
From: Marc Sune @ 2016-04-08  8:47 UTC (permalink / raw)
  To: Panu Matilainen, Matthew Hall; +Cc: Thomas Monjalon, dev

2016-04-07 13:51 GMT+02:00 Panu Matilainen <pmatilai@redhat.com>:

> [ change of subject since this is about ABI policy, not namespacing ]
>
> On 04/07/2016 01:16 PM, Marc Sune wrote:
>
>>
>>
>> 2016-04-07 11:33 GMT+02:00 Panu Matilainen <pmatilai@redhat.com
>> <mailto:pmatilai@redhat.com>>:
>>
>>     On 04/07/2016 12:18 PM, Thomas Monjalon wrote:
>>
>>         Thank you everyone for the feedbacks.
>>
>>         2016-04-05 15:56, Thomas Monjalon:
>>
>>             The goal of this email is to get some feedback on how
>>             important it is
>>             to fix the DPDK namespace.
>>
>>
>>         Everybody agree every symbols must be prefixed. Checking and
>>         fixing the
>>         namespace consistency will be in the roadmap.
>>
>>         It seems most of you agree renaming would be a nice improvement
>>         but not
>>         so important.
>>         The main drawback is the induced backporting pain, even if we have
>>         some scripts to convert the patches to the old namespace.
>>         Note: the backports can be in DPDK itself or in the applications.
>>
>>             If there is enough agreement that we should do something, I
>>             suggest to
>>             introduce the "dpdk_" prefix slowly and live with both
>>             "rte_" and "dpdk_"
>>             during some time.
>>             We could start using the new prefix for the new APIs
>>             (example: crypto)
>>             or when there is a significant API break (example: mempool).
>>
>>
>>         The slow change has been clearly rejected in favor of a complete
>>         change
>>         in one patch.
>>         The timing was also discussed as it could impact the pending
>>         patches.
>>         So it would be done at the end or the beginning of a release.
>>         Marc suggests to do it for 16.04 as the numbering scheme has
>>         changed.
>>
>>
>>     Just noting that it cannot be done in 16.04 because the ABI policy
>>     requires a deprecation cycle of at least one major release for every
>>     breakage. And we're discussing a total 100% breakage of everything
>>     here, even if its just a simple rename.
>>
>>
>> I keep not understanding the ABI policy, and particularly why ABI
>> changes have to be announced once cycle before _if_ there is already at
>> least one ABI change proposed. DPDK applications will have to recompile
>> anyway.
>>
>
> The point is to allow API/ABI consumers to assess in advance what sort of
> pains can they expect when moving their applications from one version to
> another. Otherwise all sorts of massive changes could ride the wave of
> whatever "change 16bit struct member to 32bit" trivialities that are
> nevertheless ABI breaks.
>
> There have already been quite a few exceptions to the rule when the ABI is
> already being broken, so its not entirely rigid. Another point that migth
> warrant some tweaking to the policy is the "core" libraries depending on
> each other so an ABI break in any one of them forces recompile of
> everything anyway.
>

In addition to what Matthew said:

I don't understand which sort of pains an announcement saying "we are going
to change this structure and this other, for those high level reasons, but
we don't know exactly how yet, because it is not fully implemented" can be
of any help to a DPDK user. At least it does not to me. This information
has to be in the release notes and users can read that before deciding to
upgrade to a new release.

On the other side, bug fixes still go to NEXT_VERSION. So in 1 ouf ot 2
cases (so far), next release is breaking ABI, so you will have to anyway
recompile your application.

And about ABI breakages; DPDK is not a standard library/library set. For
performance reasons it has a lot of inline code and other optimizations, so
even for small bug fixings can brake ABI, or to be precise, some bug fixes
in inline functions may be silently ignored. I don't know how many users
really use dynamic libraries for DPDK and if there is some warning
somewhere in the documentation for that.


> This aspect of the policy only slows down DPDK development and it
>>
>
> One could also think that slowing down development and forcing people to
> think ahead are not entirely unintentional or unwanted side-effects :)
>
> Look at the latest librte_vhost initiative to remove unnecessarily exposed
> structs to avoid having to deal with ABI breakages all the time: the policy
> is effectively encouraging people into better library design.
>
> pollutes the repository with commits announcing ABI changes that are
>> irrelevant after 2 cycles, as (code) diffs show that already (not
>> mentioning NEXT_ABI complexity and extra LOCs).
>>
>
> Fully agreed on NEXT_ABI, I never liked it at all.
>
> Maintaining LTS releases, and enforcing bug fixing in old LTS first,
>> upstreaming bugfixes is to me a much better approach to solve backwards
>> compatibility issues.
>>
>
> LTS releases could help the situation somewhat, but then again people tend
> to still want those new fancy things backported (you know, have the cake
> and eat it too) but that can't be done because of ABI breakage, so they're
> forced to run the latest version anyway.


> But this is probably another discussion.
>>
>
> Yup, changed subject to avoid mixing it up with the namespace discussion
> too much.


Yes, thanks for that
Marc


>
>
>         - Panu -
>
>

^ permalink raw reply	[relevance 9%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI changes for user-owned mempool caches
  2016-04-05 15:42  4% ` Olivier Matz
@ 2016-04-08 14:01  4%   ` Hunt, David
  2016-04-10  9:55  4%     ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Hunt, David @ 2016-04-08 14:01 UTC (permalink / raw)
  To: Olivier Matz, Lazaros Koromilas, dev



On 4/5/2016 4:42 PM, Olivier Matz wrote:
> On 04/05/2016 11:23 AM, Lazaros Koromilas wrote:
>> Deprecation notice for 16.04 for changes targeting release 16.07.
>> The changes affect struct rte_mempool, rte_mempool_cache and the
>> mempool API.
>>
>> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
> Acked-by: Olivier Matz <olivier.matz@6wind.com>
>

Acked-by: David Hunt<david.hunt@intel.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI changes for user-owned mempool caches
  2016-04-08 14:01  4%   ` Hunt, David
@ 2016-04-10  9:55  4%     ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-10  9:55 UTC (permalink / raw)
  To: Lazaros Koromilas; +Cc: dev, Hunt, David, Olivier Matz

> >> Deprecation notice for 16.04 for changes targeting release 16.07.
> >> The changes affect struct rte_mempool, rte_mempool_cache and the
> >> mempool API.
> >>
> >> Signed-off-by: Lazaros Koromilas <l@nofutznetworks.com>
> > Acked-by: Olivier Matz <olivier.matz@6wind.com>
> Acked-by: David Hunt<david.hunt@intel.com>

It is the fourth change announced for rte_mempool in 16.07.

Applied, thanks

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] vhost: ABI/API change announcement due to refactor
  2016-04-07  7:12  7% ` Panu Matilainen
@ 2016-04-10  9:58  4%   ` Thomas Monjalon
  2016-04-10 10:02  4%     ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-10  9:58 UTC (permalink / raw)
  To: Yuanhan Liu; +Cc: Panu Matilainen, dev, huawei.xie, Ilya Maximets

2016-04-07 10:12, Panu Matilainen:
> On 04/06/2016 09:53 AM, Yuanhan Liu wrote:
> > +* 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
> > +  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.
> 
> Acked-by: Panu Matilainen <pmatilai@redhat.com>
> 
> I applaud the initiative, public structs are by far the worst offender 
> when trying to maintain a stable ABI because they're so hard to 
> correctly version that hardly anybody besides glibc bothers.

Yes, nice cleanup to do.

Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] vhost: ABI/API change announcement due to refactor
  2016-04-10  9:58  4%   ` Thomas Monjalon
@ 2016-04-10 10:02  4%     ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-10 10:02 UTC (permalink / raw)
  To: Yuanhan Liu; +Cc: Panu Matilainen, dev, huawei.xie, Ilya Maximets

> > > +* 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
> > > +  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.
> > 
> > Acked-by: Panu Matilainen <pmatilai@redhat.com>
> > 
> > I applaud the initiative, public structs are by far the worst offender 
> > when trying to maintain a stable ABI because they're so hard to 
> > correctly version that hardly anybody besides glibc bothers.
> 
> Yes, nice cleanup to do.
> 
> Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>

Applied, thanks

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v1] doc: add template release notes for 16.11
@ 2016-04-12 12:01  6% John McNamara
  0 siblings, 0 replies; 200+ results
From: John McNamara @ 2016-04-12 12:01 UTC (permalink / raw)
  To: dev; +Cc: thomas.monjalon, John McNamara

Added template release notes for DPDK 16.11 with inline
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 | 160 +++++++++++++++++++++++++++++++++
 2 files changed, 161 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 84317b8..b38a58c 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_04
     release_2_2
     release_2_1
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..1b2ca1b
--- /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_04.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.1
+     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.
-- 
2.5.0

^ permalink raw reply	[relevance 6%]

* Re: [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure
  2016-04-07 21:24  4%     ` Thomas Monjalon
@ 2016-04-12 12:39  4%       ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-12 12:39 UTC (permalink / raw)
  To: Zhang, Roy Fan; +Cc: dev, Azarewicz, PiotrX T, Singh, Jasvinder

Hi Fan Zhang,

2016-04-07 23:24, Thomas Monjalon:
> Please send a patch to remove NEXT_ABI early in the 16.07 cycle.

Please could you prepare a patch to remove NEXT_ABI from rte_port?
Thanks

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v1] doc: add template release notes for 16.07
@ 2016-04-12 12:55  6% John McNamara
  0 siblings, 0 replies; 200+ results
From: John McNamara @ 2016-04-12 12:55 UTC (permalink / raw)
  To: dev; +Cc: thomas.monjalon, John McNamara

Added template release notes for DPDK 16.07 with inline
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_07.rst | 160 +++++++++++++++++++++++++++++++++
 2 files changed, 161 insertions(+)
 create mode 100644 doc/guides/rel_notes/release_16_07.rst

diff --git a/doc/guides/rel_notes/index.rst b/doc/guides/rel_notes/index.rst
index 84317b8..52c63b4 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_07
     release_16_04
     release_2_2
     release_2_1
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
new file mode 100644
index 0000000..701e827
--- /dev/null
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -0,0 +1,160 @@
+DPDK Release 16.07
+==================
+
+**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_07.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.1
+     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.
-- 
2.5.0

^ permalink raw reply	[relevance 6%]

* [dpdk-dev] [PATCH v5] mempool: reduce rte_mempool structure size
  @ 2016-04-14  9:42  2% ` Olivier Matz
  2016-04-14 13:28  0%   ` Wiles, Keith
                     ` (2 more replies)
  0 siblings, 3 replies; 200+ results
From: Olivier Matz @ 2016-04-14  9:42 UTC (permalink / raw)
  To: dev, keith.wiles; +Cc: thomas.monjalon, pmatilai

From: Keith Wiles <keith.wiles@intel.com>

The rte_mempool structure is changed, which will cause an ABI change
for this structure. Providing backward compat is not reasonable
here as this structure is used in multiple defines/inlines.

Allow mempool cache support to be dynamic depending on if the
mempool being created needs cache support. Saves about 1.5M of
memory used by the rte_mempool structure.

Allocating small mempools which do not require cache can consume
larges amounts of memory if you have a number of these mempools.

Change to be effective in release 16.07.

Signed-off-by: Keith Wiles <keith.wiles@intel.com>
Acked-by: Olivier Matz <olivier.matz@6wind.com>
---

Changes in v5:

- use RTE_PTR_ADD() instead of cast to (char *) to fix compilation on tilera.
  Error log was:

  rte_mempool.c: In function ‘rte_mempool_xmem_create’:
  rte_mempool.c:595: error: cast increases required alignment of target type


 app/test/test_mempool.c          |  4 +--
 lib/librte_mempool/rte_mempool.c | 55 ++++++++++++++++++----------------------
 lib/librte_mempool/rte_mempool.h | 29 ++++++++++-----------
 3 files changed, 40 insertions(+), 48 deletions(-)

diff --git a/app/test/test_mempool.c b/app/test/test_mempool.c
index f0f823b..10e1fa4 100644
--- a/app/test/test_mempool.c
+++ b/app/test/test_mempool.c
@@ -122,8 +122,8 @@ test_mempool_basic(void)
 		return -1;
 
 	printf("get private data\n");
-	if (rte_mempool_get_priv(mp) !=
-			(char*) mp + MEMPOOL_HEADER_SIZE(mp, mp->pg_num))
+	if (rte_mempool_get_priv(mp) != (char *)mp +
+			MEMPOOL_HEADER_SIZE(mp, mp->pg_num, mp->cache_size))
 		return -1;
 
 	printf("get physical address of an object\n");
diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
index f8781e1..7a0e07e 100644
--- a/lib/librte_mempool/rte_mempool.c
+++ b/lib/librte_mempool/rte_mempool.c
@@ -452,12 +452,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
 	/* compilation-time checks */
 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool) &
 			  RTE_CACHE_LINE_MASK) != 0);
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool_cache) &
 			  RTE_CACHE_LINE_MASK) != 0);
-	RTE_BUILD_BUG_ON((offsetof(struct rte_mempool, local_cache) &
-			  RTE_CACHE_LINE_MASK) != 0);
-#endif
 #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool_debug_stats) &
 			  RTE_CACHE_LINE_MASK) != 0);
@@ -527,9 +523,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
 		 */
 		int head = sizeof(struct rte_mempool);
 		int new_size = (private_data_size + head) % page_size;
-		if (new_size) {
+		if (new_size)
 			private_data_size += page_size - new_size;
-		}
 	}
 
 	/* try to allocate tailq entry */
@@ -544,7 +539,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
 	 * store mempool objects. Otherwise reserve a memzone that is large
 	 * enough to hold mempool header and metadata plus mempool objects.
 	 */
-	mempool_size = MEMPOOL_HEADER_SIZE(mp, pg_num) + private_data_size;
+	mempool_size = MEMPOOL_HEADER_SIZE(mp, pg_num, cache_size);
+	mempool_size += private_data_size;
 	mempool_size = RTE_ALIGN_CEIL(mempool_size, RTE_MEMPOOL_ALIGN);
 	if (vaddr == NULL)
 		mempool_size += (size_t)objsz.total_size * n;
@@ -591,8 +587,15 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
 	mp->cache_flushthresh = CALC_CACHE_FLUSHTHRESH(cache_size);
 	mp->private_data_size = private_data_size;
 
+	/*
+	 * local_cache pointer is set even if cache_size is zero.
+	 * The local_cache points to just past the elt_pa[] array.
+	 */
+	mp->local_cache = (struct rte_mempool_cache *)
+		RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, pg_num, 0));
+
 	/* calculate address of the first element for continuous mempool. */
-	obj = (char *)mp + MEMPOOL_HEADER_SIZE(mp, pg_num) +
+	obj = (char *)mp + MEMPOOL_HEADER_SIZE(mp, pg_num, cache_size) +
 		private_data_size;
 	obj = RTE_PTR_ALIGN_CEIL(obj, RTE_MEMPOOL_ALIGN);
 
@@ -606,9 +609,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
 		mp->elt_va_start = (uintptr_t)obj;
 		mp->elt_pa[0] = mp->phys_addr +
 			(mp->elt_va_start - (uintptr_t)mp);
-
-	/* mempool elements in a separate chunk of memory. */
 	} else {
+		/* mempool elements in a separate chunk of memory. */
 		mp->elt_va_start = (uintptr_t)vaddr;
 		memcpy(mp->elt_pa, paddr, sizeof (mp->elt_pa[0]) * pg_num);
 	}
@@ -643,19 +645,15 @@ unsigned
 rte_mempool_count(const struct rte_mempool *mp)
 {
 	unsigned count;
+	unsigned lcore_id;
 
 	count = rte_ring_count(mp->ring);
 
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
-	{
-		unsigned lcore_id;
-		if (mp->cache_size == 0)
-			return count;
+	if (mp->cache_size == 0)
+		return count;
 
-		for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
-			count += mp->local_cache[lcore_id].len;
-	}
-#endif
+	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
+		count += mp->local_cache[lcore_id].len;
 
 	/*
 	 * due to race condition (access to len is not locked), the
@@ -670,13 +668,16 @@ rte_mempool_count(const struct rte_mempool *mp)
 static unsigned
 rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
 {
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 	unsigned lcore_id;
 	unsigned count = 0;
 	unsigned cache_count;
 
 	fprintf(f, "  cache infos:\n");
 	fprintf(f, "    cache_size=%"PRIu32"\n", mp->cache_size);
+
+	if (mp->cache_size == 0)
+		return count;
+
 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
 		cache_count = mp->local_cache[lcore_id].len;
 		fprintf(f, "    cache_count[%u]=%u\n", lcore_id, cache_count);
@@ -684,11 +685,6 @@ rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
 	}
 	fprintf(f, "    total_cache_count=%u\n", count);
 	return count;
-#else
-	RTE_SET_USED(mp);
-	fprintf(f, "  cache disabled\n");
-	return 0;
-#endif
 }
 
 #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
@@ -753,13 +749,16 @@ mempool_audit_cookies(const struct rte_mempool *mp)
 #define mempool_audit_cookies(mp) do {} while(0)
 #endif
 
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 /* check cookies before and after objects */
 static void
 mempool_audit_cache(const struct rte_mempool *mp)
 {
 	/* check cache size consistency */
 	unsigned lcore_id;
+
+	if (mp->cache_size == 0)
+		return;
+
 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
 		if (mp->local_cache[lcore_id].len > mp->cache_flushthresh) {
 			RTE_LOG(CRIT, MEMPOOL, "badness on cache[%u]\n",
@@ -768,10 +767,6 @@ mempool_audit_cache(const struct rte_mempool *mp)
 		}
 	}
 }
-#else
-#define mempool_audit_cache(mp) do {} while(0)
-#endif
-
 
 /* check the consistency of mempool (size, cookies, ...) */
 void
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 9745bf0..8595e77 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -95,7 +95,6 @@ struct rte_mempool_debug_stats {
 } __rte_cache_aligned;
 #endif
 
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 /**
  * A structure that stores a per-core object cache.
  */
@@ -107,7 +106,6 @@ struct rte_mempool_cache {
 	 */
 	void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 3]; /**< Cache objects */
 } __rte_cache_aligned;
-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
 
 /**
  * A structure that stores the size of mempool elements.
@@ -194,10 +192,7 @@ struct rte_mempool {
 
 	unsigned private_data_size;      /**< Size of private data. */
 
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
-	/** Per-lcore local cache. */
-	struct rte_mempool_cache local_cache[RTE_MAX_LCORE];
-#endif
+	struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
 
 #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
 	/** Per-lcore statistics. */
@@ -247,6 +242,13 @@ struct rte_mempool {
 #endif
 
 /**
+ * Size of elt_pa array size based on number of pages. (Internal use)
+ */
+#define __PA_SIZE(mp, pgn) \
+	RTE_ALIGN_CEIL((((pgn) - RTE_DIM((mp)->elt_pa)) * \
+	sizeof((mp)->elt_pa[0])), RTE_CACHE_LINE_SIZE)
+
+/**
  * Calculate the size of the mempool header.
  *
  * @param mp
@@ -254,9 +256,9 @@ struct rte_mempool {
  * @param pgn
  *   Number of pages used to store mempool objects.
  */
-#define	MEMPOOL_HEADER_SIZE(mp, pgn)	(sizeof(*(mp)) + \
-	RTE_ALIGN_CEIL(((pgn) - RTE_DIM((mp)->elt_pa)) * \
-	sizeof ((mp)->elt_pa[0]), RTE_CACHE_LINE_SIZE))
+#define MEMPOOL_HEADER_SIZE(mp, pgn, cs) \
+	(sizeof(*(mp)) + __PA_SIZE(mp, pgn) + (((cs) == 0) ? 0 : \
+	(sizeof(struct rte_mempool_cache) * RTE_MAX_LCORE)))
 
 /**
  * Return true if the whole mempool is in contiguous memory.
@@ -755,19 +757,16 @@ static inline void __attribute__((always_inline))
 __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
 		    unsigned n, int is_mp)
 {
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 	struct rte_mempool_cache *cache;
 	uint32_t index;
 	void **cache_objs;
 	unsigned lcore_id = rte_lcore_id();
 	uint32_t cache_size = mp->cache_size;
 	uint32_t flushthresh = mp->cache_flushthresh;
-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
 
 	/* increment stat now, adding in mempool always success */
 	__MEMPOOL_STAT_ADD(mp, put, n);
 
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 	/* cache is not enabled or single producer or non-EAL thread */
 	if (unlikely(cache_size == 0 || is_mp == 0 ||
 		     lcore_id >= RTE_MAX_LCORE))
@@ -802,7 +801,6 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
 	return;
 
 ring_enqueue:
-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
 
 	/* push remaining objects in ring */
 #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
@@ -946,7 +944,6 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
 		   unsigned n, int is_mc)
 {
 	int ret;
-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
 	struct rte_mempool_cache *cache;
 	uint32_t index, len;
 	void **cache_objs;
@@ -992,7 +989,6 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
 	return 0;
 
 ring_dequeue:
-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
 
 	/* get remaining objects from ring */
 	if (is_mc)
@@ -1293,7 +1289,8 @@ void rte_mempool_audit(const struct rte_mempool *mp);
  */
 static inline void *rte_mempool_get_priv(struct rte_mempool *mp)
 {
-	return (char *)mp + MEMPOOL_HEADER_SIZE(mp, mp->pg_num);
+	return (char *)mp +
+		MEMPOOL_HEADER_SIZE(mp, mp->pg_num, mp->cache_size);
 }
 
 /**
-- 
2.1.4

^ permalink raw reply	[relevance 2%]

* [dpdk-dev] [RFC 0/2] add new fields to rte_eth_dev_info structure
@ 2016-04-14  9:44  4% Reshma Pattan
  2016-04-14  9:44  9% ` [dpdk-dev] [RFC 1/2] doc: announce ABI change for " Reshma Pattan
    0 siblings, 2 replies; 200+ results
From: Reshma Pattan @ 2016-04-14  9:44 UTC (permalink / raw)
  To: dev

New fields nb_rx_queues and nb_tx_queues are added to rte_eth_dev_info structure.
Changes to API rte_eth_dev_info_get() are done to update these new fields to rte_eth_dev_info object.

These changes are ABI breakage and we are late to announce deprecation notice for 16.07,
however the rte_ether library is already subject to a deprecation notice in 16.07.

Reshma Pattan (2):
  doc: announce ABI change for rte_eth_dev_info structure
  librte_ether: add new fields to rte_eth_dev_info struct

 doc/guides/rel_notes/deprecation.rst | 6 ++++++
 lib/librte_ether/rte_ethdev.c        | 2 ++
 lib/librte_ether/rte_ethdev.h        | 3 +++
 3 files changed, 11 insertions(+)

-- 
2.5.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [RFC 1/2] doc: announce ABI change for rte_eth_dev_info structure
  2016-04-14  9:44  4% [dpdk-dev] [RFC 0/2] add new fields to rte_eth_dev_info structure Reshma Pattan
@ 2016-04-14  9:44  9% ` Reshma Pattan
  2016-04-15  9:42  4%   ` Mcnamara, John
  2016-04-15 10:02  8%   ` Thomas Monjalon
    1 sibling, 2 replies; 200+ results
From: Reshma Pattan @ 2016-04-14  9:44 UTC (permalink / raw)
  To: dev

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 327fc2b..78cedb7 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -90,3 +90,9 @@ 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 00/36] mempool: rework memory allocation
    2016-03-17  9:05 15% ` [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07 Olivier Matz
@ 2016-04-14 10:19  2% ` Olivier Matz
  2016-04-14 13:50  0%   ` Wiles, Keith
  2016-05-18 11:04  2%   ` [dpdk-dev] [PATCH v3 00/35] " Olivier Matz
  1 sibling, 2 replies; 200+ results
From: Olivier Matz @ 2016-04-14 10:19 UTC (permalink / raw)
  To: dev; +Cc: bruce.richardson, stephen

This series is a rework of mempool. For those who don't want to read
all the cover letter, here is a sumary:

- it is not possible to allocate large mempools if there is not enough
  contiguous memory, this series solves this issue
- introduce new APIs with less arguments: "create, populate, obj_init"
- allow to free a mempool
- split code in smaller functions, will ease the introduction of ext_handler
- remove test-pmd anonymous mempool creation
- remove most of dom0-specific mempool code
- opens the door for a eal_memory rework: we probably don't need large
  contiguous memory area anymore, working with pages would work.

This breaks the ABI as it was indicated in the deprecation for 16.04.
The API stays almost the same, no modification is needed in examples app
or in test-pmd. Only kni and mellanox drivers are slightly modified.

This patch applies on top of 16.04 + v5 of Keith's patch:
"mempool: reduce rte_mempool structure size"

Changes RFC -> v1:

- remove the rte_deconst macro, and remove some const qualifier in
  dump/audit functions
- rework modifications in mellanox drivers to ensure the mempool is
  virtually contiguous
- fix mempool memory chunk iteration (bad pointer was used)
- fix compilation on freebsd: replace MAP_LOCKED flag by mlock()
- fix compilation on tilera (pointer arithmetics)
- slightly rework and clean the mempool autotest
- fix mempool autotest on bsd
- more validation (especially mellanox drivers and kni that were not
  tested in RFC)
- passed autotests (x86_64-native-linuxapp-gcc and x86_64-native-bsdapp-gcc)
- rebase on head, reorder the patches a bit and fix minor split issues


Description of the initial issue
--------------------------------

The allocation of mbuf pool can fail even if there is enough memory.
The problem is related to the way the memory is allocated and used in
dpdk. It is particularly annoying with mbuf pools, but it can also fail
in other use cases allocating a large amount of memory.

- rte_malloc() allocates physically contiguous memory, which is needed
  for mempools, but useless most of the time.

  Allocating a large physically contiguous zone is often impossible
  because the system provide hugepages which may not be contiguous.

- rte_mempool_create() (and therefore rte_pktmbuf_pool_create())
  requires a physically contiguous zone.

- rte_mempool_xmem_create() does not solve the issue as it still
  needs the memory to be virtually contiguous, and there is no
  way in dpdk to allocate a virtually contiguous memory that is
  not also physically contiguous.

How to reproduce the issue
--------------------------

- start the dpdk with some 2MB hugepages (it can also occur with 1GB)
- allocate a large mempool
- even if there is enough memory, the allocation can fail

Example:

  git clone http://dpdk.org/git/dpdk
  cd dpdk
  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

  # we try to allocate a mempool whose size is ~450MB, it fails
  ./build/app/testpmd -l 2,4 -- --total-num-mbufs=200000 -i

The EAL logs "EAL: Virtual area found at..." shows that there are
several zones, but all smaller than 450MB.

Workarounds:

- Use 1GB hugepages: it sometimes work, but for very large
  pools (millions of mbufs) there is the same issue. Moreover,
  it would consume 1GB memory at least which can be a lot
  in some cases.

- Reboot the machine or allocate hugepages at boot time: this increases
  the chances to have more contiguous memory, but does not completely
  solve the issue

Solutions
---------

Below is a list of proposed solutions. I implemented a quick and dirty
PoC of solution 1, but it's not working in all conditions and it's
really an ugly hack.  This series implement the solution 4 which looks
the best to me, knowing it does not prevent to do more enhancements
in dpdk memory in the future (solution 3 for instance).

Solution 1: in application
--------------------------

- allocate several hugepages using rte_malloc() or rte_memzone_reserve()
  (only keeping complete hugepages)
- parse memsegs and /proc/maps to check which files mmaps these pages
- mmap the files in a contiguous virtual area
- use rte_mempool_xmem_create()

Cons:

- 1a. parsing the memsegs of rte config in the application does not
  use a public API, and can be broken if internal dpdk code changes
- 1b. some memory is lost due to malloc headers. Also, if the memory is
  very fragmented (ex: all 2MB pages are physically separated), it does
  not work at all because we cannot get any complete page. It is not
  possible to use a lower level allocator since commit fafcc11985a.
- 1c. we cannot use rte_pktmbuf_pool_create(), so we need to use mempool
  api and do a part of the job manually
- 1d. it breaks secondary processes as the virtual addresses won't be
  mmap'd at the same place in secondary process
- 1e. it only fixes the issue for the mbuf pool of the application,
  internal pools in dpdk libraries are not modified
- 1f. this is a pure linux solution (rte_map files)
- 1g. The application has to be aware of RTE_EAL_SINGLE_SEGMENTS option
  that changes the way hugepages are mapped. By the way, it's strange
  to have such a compile-time option, we should probably have only
  one behavior that works all the time.

Solution 2: in dpdk memory allocator
------------------------------------

- do the same than solution 1 in a new function rte_malloc_non_contig():
  allocate several chunks and mmap them in a contiguous virtual memory
- a flag has to be added in malloc header to do the proper cleanup in
  rte_free() (free all the chunks, munmap the memory)
- introduce a new rte_mem_get_physmap(*physmap,addr, len) that returns
  the virt2phys mapping of a virtual area in dpdk
- add a mempool flag MEMPOOL_F_NON_PHYS_CONTIG to use
  rte_malloc_non_contig() to allocate the area storing the objects

Cons:

- 2a. same than 1b: it breaks secondary processes if the mempool flag is
  used.
- 2b. same as 1d: some memory is lost due to malloc headers, and it
  cannot work if memory is too fragmented.
- 2c. rte_malloc_virt2phy() cannot be used on these zones. It would
  return the physical address of the first page. It would be better to
  return an error in this case.
- 2d. need to check how to implement this on bsd (TBD)

Solution 3: in dpdk eal memory
------------------------------

- Rework the way hugepages are mmap'd in dpdk: instead of having several
  rte_map* files, just mmap one file per node. It may drastically
  simplify EAL memory management in dpdk.
- An API should be added to retrieve the physical mapping of a virtual
  area (ex: rte_mem_get_physmap(*physmap, addr, len))
- rte_malloc() and rte_memzone_reserve() won't allocate physically
  contiguous memory anymore (TBD)
- Update mempool to always use the rte_mempool_xmem_create() version

Cons:

- 3a. lot of rework in eal memory, it will induce some behavior changes
  and maybe api changes
- 3b. possible conflicts with xen_dom0 mempool

Solution 4: in mempool
----------------------

- Introduce a new API to fill a mempool with zones that are not
  virtually contiguous. It requires to add new functions to create and
  populate a mempool. Example (TBD):

  - rte_mempool_create_empty(name, n, elt_size, cache_size, priv_size)
  - rte_mempool_populate(mp, addr, len): add virtual memory for objects
  - rte_mempool_mempool_obj_iter(mp, obj_cb, arg): call a cb for each object

- update rte_mempool_create() to allocate objects in several memory
  chunks by default if there is no large enough physically contiguous
  memory.

Tests done
----------

Compilation
~~~~~~~~~~~

The following targets:

 x86_64-native-linuxapp-gcc
 i686-native-linuxapp-gcc
 x86_x32-native-linuxapp-gcc
 x86_64-native-linuxapp-clang
 x86_64-native-bsdapp-gcc
 ppc_64-power8-linuxapp-gcc
 tile-tilegx-linuxapp-gcc (only the mempool files, the target does not compile)

Libraries with and without debug, in static and shared mode + examples.

autotests
~~~~~~~~~

Passed all autotests on x86_64-native-linuxapp-gcc (including kni) and
mempool-related autotests on x86_64-native-bsdapp-gcc.

test-pmd
~~~~~~~~

# now starts fine, was failing before if mempool was too fragmented
./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -- -i --port-topology=chained

# still ok
./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -m 256 -- -i --port-topology=chained --mp-anon
set fwd txonly
start
stop

# fail, but was failing before too. The problem is because the physical
# addresses are not properly set when using --no-huge. The mempool phys addr
# are now correct, but the zones allocated through memzone_reserve() are
# still wrong. This could be fixed in a future series.
./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -m 256 --no-huge -- -i ---port-topology=chained
set fwd txonly
start
stop


Olivier Matz (36):
  mempool: fix comments and style
  mempool: replace elt_size by total_elt_size
  mempool: uninline function to check cookies
  mempool: use sizeof to get the size of header and trailer
  mempool: rename mempool_obj_ctor_t as mempool_obj_cb_t
  mempool: update library version
  mempool: list objects when added in the mempool
  mempool: remove const attribute in mempool_walk
  mempool: remove const qualifier in dump and audit
  mempool: use the list to iterate the mempool elements
  mempool: use the list to audit all elements
  mempool: use the list to initialize mempool objects
  mempool: create the internal ring in a specific function
  mempool: store physaddr in mempool objects
  mempool: remove MEMPOOL_IS_CONTIG()
  mempool: store memory chunks in a list
  mempool: new function to iterate the memory chunks
  mempool: simplify xmem_usage
  mempool: introduce a free callback for memory chunks
  mempool: make page size optional when getting xmem size
  mempool: default allocation in several memory chunks
  eal: lock memory when using no-huge
  mempool: support no-hugepage mode
  mempool: replace mempool physaddr by a memzone pointer
  mempool: introduce a function to free a mempool
  mempool: introduce a function to create an empty mempool
  eal/xen: return machine address without knowing memseg id
  mempool: rework support of xen dom0
  mempool: create the internal ring when populating
  mempool: populate a mempool with anonymous memory
  mempool: make mempool populate and free api public
  test-pmd: remove specific anon mempool code
  mem: avoid memzone/mempool/ring name truncation
  mempool: new flag when phys contig mem is not needed
  app/test: rework mempool test
  mempool: update copyright

 app/test-pmd/Makefile                        |    4 -
 app/test-pmd/mempool_anon.c                  |  201 -----
 app/test-pmd/mempool_osdep.h                 |   54 --
 app/test-pmd/testpmd.c                       |   23 +-
 app/test/test_mempool.c                      |  243 +++---
 doc/guides/rel_notes/release_16_04.rst       |    2 +-
 drivers/net/mlx4/mlx4.c                      |  140 ++--
 drivers/net/mlx5/mlx5_rxtx.c                 |  140 ++--
 drivers/net/mlx5/mlx5_rxtx.h                 |    4 +-
 drivers/net/xenvirt/rte_eth_xenvirt.h        |    2 +-
 drivers/net/xenvirt/rte_mempool_gntalloc.c   |    4 +-
 lib/librte_eal/common/eal_common_log.c       |    2 +-
 lib/librte_eal/common/eal_common_memzone.c   |   10 +-
 lib/librte_eal/common/include/rte_memory.h   |   11 +-
 lib/librte_eal/linuxapp/eal/eal_memory.c     |    2 +-
 lib/librte_eal/linuxapp/eal/eal_xen_memory.c |   17 +-
 lib/librte_kni/rte_kni.c                     |   12 +-
 lib/librte_mempool/Makefile                  |    5 +-
 lib/librte_mempool/rte_dom0_mempool.c        |  133 ----
 lib/librte_mempool/rte_mempool.c             | 1042 +++++++++++++++++---------
 lib/librte_mempool/rte_mempool.h             |  594 +++++++--------
 lib/librte_mempool/rte_mempool_version.map   |   18 +-
 lib/librte_ring/rte_ring.c                   |   16 +-
 23 files changed, 1377 insertions(+), 1302 deletions(-)
 delete mode 100644 app/test-pmd/mempool_anon.c
 delete mode 100644 app/test-pmd/mempool_osdep.h
 delete mode 100644 lib/librte_mempool/rte_dom0_mempool.c

-- 
2.1.4

^ permalink raw reply	[relevance 2%]

* Re: [dpdk-dev] [PATCH v5] mempool: reduce rte_mempool structure size
  2016-04-14  9:42  2% ` [dpdk-dev] [PATCH v5] " Olivier Matz
@ 2016-04-14 13:28  0%   ` Wiles, Keith
  2016-04-14 13:53  0%   ` Wiles, Keith
  2016-05-17  5:31  0%   ` Thomas Monjalon
  2 siblings, 0 replies; 200+ results
From: Wiles, Keith @ 2016-04-14 13:28 UTC (permalink / raw)
  To: Olivier Matz, dev; +Cc: thomas.monjalon, pmatilai

>From: Keith Wiles <keith.wiles@intel.com>
>
>The rte_mempool structure is changed, which will cause an ABI change
>for this structure. Providing backward compat is not reasonable
>here as this structure is used in multiple defines/inlines.
>
>Allow mempool cache support to be dynamic depending on if the
>mempool being created needs cache support. Saves about 1.5M of
>memory used by the rte_mempool structure.
>
>Allocating small mempools which do not require cache can consume
>larges amounts of memory if you have a number of these mempools.
>
>Change to be effective in release 16.07.
>
>Signed-off-by: Keith Wiles <keith.wiles@intel.com>
>Acked-by: Olivier Matz <olivier.matz@6wind.com>
>---
>
>Changes in v5:
>
>- use RTE_PTR_ADD() instead of cast to (char *) to fix compilation on tilera.
>  Error log was:
>
>  rte_mempool.c: In function ‘rte_mempool_xmem_create’:
>  rte_mempool.c:595: error: cast increases required alignment of target type
>
>
> app/test/test_mempool.c          |  4 +--
> lib/librte_mempool/rte_mempool.c | 55 ++++++++++++++++++----------------------
> lib/librte_mempool/rte_mempool.h | 29 ++++++++++-----------
> 3 files changed, 40 insertions(+), 48 deletions(-)
>
>diff --git a/app/test/test_mempool.c b/app/test/test_mempool.c
>index f0f823b..10e1fa4 100644
>--- a/app/test/test_mempool.c
>+++ b/app/test/test_mempool.c
>@@ -122,8 +122,8 @@ test_mempool_basic(void)
> 		return -1;
> 
> 	printf("get private data\n");
>-	if (rte_mempool_get_priv(mp) !=
>-			(char*) mp + MEMPOOL_HEADER_SIZE(mp, mp->pg_num))
>+	if (rte_mempool_get_priv(mp) != (char *)mp +
>+			MEMPOOL_HEADER_SIZE(mp, mp->pg_num, mp->cache_size))

Should we not add the RTE_PTR_ADD() here as well?

> 		return -1;
> 
> 	printf("get physical address of an object\n");
>diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
>index f8781e1..7a0e07e 100644
>--- a/lib/librte_mempool/rte_mempool.c
>+++ b/lib/librte_mempool/rte_mempool.c
>@@ -452,12 +452,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 	/* compilation-time checks */
> 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool) &
> 			  RTE_CACHE_LINE_MASK) != 0);
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool_cache) &
> 			  RTE_CACHE_LINE_MASK) != 0);
>-	RTE_BUILD_BUG_ON((offsetof(struct rte_mempool, local_cache) &
>-			  RTE_CACHE_LINE_MASK) != 0);
>-#endif
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
> 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool_debug_stats) &
> 			  RTE_CACHE_LINE_MASK) != 0);
>@@ -527,9 +523,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 		 */
> 		int head = sizeof(struct rte_mempool);
> 		int new_size = (private_data_size + head) % page_size;
>-		if (new_size) {
>+		if (new_size)
> 			private_data_size += page_size - new_size;
>-		}
> 	}
> 
> 	/* try to allocate tailq entry */
>@@ -544,7 +539,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 	 * store mempool objects. Otherwise reserve a memzone that is large
> 	 * enough to hold mempool header and metadata plus mempool objects.
> 	 */
>-	mempool_size = MEMPOOL_HEADER_SIZE(mp, pg_num) + private_data_size;
>+	mempool_size = MEMPOOL_HEADER_SIZE(mp, pg_num, cache_size);
>+	mempool_size += private_data_size;
> 	mempool_size = RTE_ALIGN_CEIL(mempool_size, RTE_MEMPOOL_ALIGN);
> 	if (vaddr == NULL)
> 		mempool_size += (size_t)objsz.total_size * n;
>@@ -591,8 +587,15 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 	mp->cache_flushthresh = CALC_CACHE_FLUSHTHRESH(cache_size);
> 	mp->private_data_size = private_data_size;
> 
>+	/*
>+	 * local_cache pointer is set even if cache_size is zero.
>+	 * The local_cache points to just past the elt_pa[] array.
>+	 */
>+	mp->local_cache = (struct rte_mempool_cache *)
>+		RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, pg_num, 0));
>+
> 	/* calculate address of the first element for continuous mempool. */
>-	obj = (char *)mp + MEMPOOL_HEADER_SIZE(mp, pg_num) +
>+	obj = (char *)mp + MEMPOOL_HEADER_SIZE(mp, pg_num, cache_size) +
> 		private_data_size;
> 	obj = RTE_PTR_ALIGN_CEIL(obj, RTE_MEMPOOL_ALIGN);
> 
>@@ -606,9 +609,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 		mp->elt_va_start = (uintptr_t)obj;
> 		mp->elt_pa[0] = mp->phys_addr +
> 			(mp->elt_va_start - (uintptr_t)mp);
>-
>-	/* mempool elements in a separate chunk of memory. */
> 	} else {
>+		/* mempool elements in a separate chunk of memory. */
> 		mp->elt_va_start = (uintptr_t)vaddr;
> 		memcpy(mp->elt_pa, paddr, sizeof (mp->elt_pa[0]) * pg_num);
> 	}
>@@ -643,19 +645,15 @@ unsigned
> rte_mempool_count(const struct rte_mempool *mp)
> {
> 	unsigned count;
>+	unsigned lcore_id;
> 
> 	count = rte_ring_count(mp->ring);
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>-	{
>-		unsigned lcore_id;
>-		if (mp->cache_size == 0)
>-			return count;
>+	if (mp->cache_size == 0)
>+		return count;
> 
>-		for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
>-			count += mp->local_cache[lcore_id].len;
>-	}
>-#endif
>+	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
>+		count += mp->local_cache[lcore_id].len;
> 
> 	/*
> 	 * due to race condition (access to len is not locked), the
>@@ -670,13 +668,16 @@ rte_mempool_count(const struct rte_mempool *mp)
> static unsigned
> rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
> {
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	unsigned lcore_id;
> 	unsigned count = 0;
> 	unsigned cache_count;
> 
> 	fprintf(f, "  cache infos:\n");
> 	fprintf(f, "    cache_size=%"PRIu32"\n", mp->cache_size);
>+
>+	if (mp->cache_size == 0)
>+		return count;
>+
> 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
> 		cache_count = mp->local_cache[lcore_id].len;
> 		fprintf(f, "    cache_count[%u]=%u\n", lcore_id, cache_count);
>@@ -684,11 +685,6 @@ rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
> 	}
> 	fprintf(f, "    total_cache_count=%u\n", count);
> 	return count;
>-#else
>-	RTE_SET_USED(mp);
>-	fprintf(f, "  cache disabled\n");
>-	return 0;
>-#endif
> }
> 
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
>@@ -753,13 +749,16 @@ mempool_audit_cookies(const struct rte_mempool *mp)
> #define mempool_audit_cookies(mp) do {} while(0)
> #endif
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> /* check cookies before and after objects */
> static void
> mempool_audit_cache(const struct rte_mempool *mp)
> {
> 	/* check cache size consistency */
> 	unsigned lcore_id;
>+
>+	if (mp->cache_size == 0)
>+		return;
>+
> 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
> 		if (mp->local_cache[lcore_id].len > mp->cache_flushthresh) {
> 			RTE_LOG(CRIT, MEMPOOL, "badness on cache[%u]\n",
>@@ -768,10 +767,6 @@ mempool_audit_cache(const struct rte_mempool *mp)
> 		}
> 	}
> }
>-#else
>-#define mempool_audit_cache(mp) do {} while(0)
>-#endif
>-
> 
> /* check the consistency of mempool (size, cookies, ...) */
> void
>diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
>index 9745bf0..8595e77 100644
>--- a/lib/librte_mempool/rte_mempool.h
>+++ b/lib/librte_mempool/rte_mempool.h
>@@ -95,7 +95,6 @@ struct rte_mempool_debug_stats {
> } __rte_cache_aligned;
> #endif
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> /**
>  * A structure that stores a per-core object cache.
>  */
>@@ -107,7 +106,6 @@ struct rte_mempool_cache {
> 	 */
> 	void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 3]; /**< Cache objects */
> } __rte_cache_aligned;
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> /**
>  * A structure that stores the size of mempool elements.
>@@ -194,10 +192,7 @@ struct rte_mempool {
> 
> 	unsigned private_data_size;      /**< Size of private data. */
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>-	/** Per-lcore local cache. */
>-	struct rte_mempool_cache local_cache[RTE_MAX_LCORE];
>-#endif
>+	struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
> 
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
> 	/** Per-lcore statistics. */
>@@ -247,6 +242,13 @@ struct rte_mempool {
> #endif
> 
> /**
>+ * Size of elt_pa array size based on number of pages. (Internal use)
>+ */
>+#define __PA_SIZE(mp, pgn) \
>+	RTE_ALIGN_CEIL((((pgn) - RTE_DIM((mp)->elt_pa)) * \
>+	sizeof((mp)->elt_pa[0])), RTE_CACHE_LINE_SIZE)
>+
>+/**
>  * Calculate the size of the mempool header.
>  *
>  * @param mp
>@@ -254,9 +256,9 @@ struct rte_mempool {
>  * @param pgn
>  *   Number of pages used to store mempool objects.
>  */
>-#define	MEMPOOL_HEADER_SIZE(mp, pgn)	(sizeof(*(mp)) + \
>-	RTE_ALIGN_CEIL(((pgn) - RTE_DIM((mp)->elt_pa)) * \
>-	sizeof ((mp)->elt_pa[0]), RTE_CACHE_LINE_SIZE))
>+#define MEMPOOL_HEADER_SIZE(mp, pgn, cs) \
>+	(sizeof(*(mp)) + __PA_SIZE(mp, pgn) + (((cs) == 0) ? 0 : \
>+	(sizeof(struct rte_mempool_cache) * RTE_MAX_LCORE)))
> 
> /**
>  * Return true if the whole mempool is in contiguous memory.
>@@ -755,19 +757,16 @@ static inline void __attribute__((always_inline))
> __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
> 		    unsigned n, int is_mp)
> {
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	struct rte_mempool_cache *cache;
> 	uint32_t index;
> 	void **cache_objs;
> 	unsigned lcore_id = rte_lcore_id();
> 	uint32_t cache_size = mp->cache_size;
> 	uint32_t flushthresh = mp->cache_flushthresh;
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> 	/* increment stat now, adding in mempool always success */
> 	__MEMPOOL_STAT_ADD(mp, put, n);
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	/* cache is not enabled or single producer or non-EAL thread */
> 	if (unlikely(cache_size == 0 || is_mp == 0 ||
> 		     lcore_id >= RTE_MAX_LCORE))
>@@ -802,7 +801,6 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
> 	return;
> 
> ring_enqueue:
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> 	/* push remaining objects in ring */
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
>@@ -946,7 +944,6 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
> 		   unsigned n, int is_mc)
> {
> 	int ret;
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	struct rte_mempool_cache *cache;
> 	uint32_t index, len;
> 	void **cache_objs;
>@@ -992,7 +989,6 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
> 	return 0;
> 
> ring_dequeue:
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> 	/* get remaining objects from ring */
> 	if (is_mc)
>@@ -1293,7 +1289,8 @@ void rte_mempool_audit(const struct rte_mempool *mp);
>  */
> static inline void *rte_mempool_get_priv(struct rte_mempool *mp)
> {
>-	return (char *)mp + MEMPOOL_HEADER_SIZE(mp, mp->pg_num);
>+	return (char *)mp +
>+		MEMPOOL_HEADER_SIZE(mp, mp->pg_num, mp->cache_size);

And here?

> }
> 
> /**
>-- 
>2.1.4
>
>


Regards,
Keith





^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation
  2016-04-14 10:19  2% ` [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation Olivier Matz
@ 2016-04-14 13:50  0%   ` Wiles, Keith
  2016-04-14 14:01  0%     ` Olivier MATZ
  2016-05-18 11:04  2%   ` [dpdk-dev] [PATCH v3 00/35] " Olivier Matz
  1 sibling, 1 reply; 200+ results
From: Wiles, Keith @ 2016-04-14 13:50 UTC (permalink / raw)
  To: Olivier Matz, dev; +Cc: Richardson, Bruce, stephen

>This series is a rework of mempool. For those who don't want to read
>all the cover letter, here is a sumary:
>
>- it is not possible to allocate large mempools if there is not enough
>  contiguous memory, this series solves this issue
>- introduce new APIs with less arguments: "create, populate, obj_init"
>- allow to free a mempool
>- split code in smaller functions, will ease the introduction of ext_handler
>- remove test-pmd anonymous mempool creation
>- remove most of dom0-specific mempool code
>- opens the door for a eal_memory rework: we probably don't need large
>  contiguous memory area anymore, working with pages would work.
>
>This breaks the ABI as it was indicated in the deprecation for 16.04.
>The API stays almost the same, no modification is needed in examples app
>or in test-pmd. Only kni and mellanox drivers are slightly modified.
>
>This patch applies on top of 16.04 + v5 of Keith's patch:
>"mempool: reduce rte_mempool structure size"

I have not digested this complete patch yet, but this one popped out at me as the External Memory Manager support is setting in the wings for 16.07 release. If this causes the EMM patch to be rewritten or updated that seems like a problem to me. Does this patch add the External Memory Manager support?
http://thread.gmane.org/gmane.comp.networking.dpdk.devel/32015/focus=35107


>
>Changes RFC -> v1:
>
>- remove the rte_deconst macro, and remove some const qualifier in
>  dump/audit functions
>- rework modifications in mellanox drivers to ensure the mempool is
>  virtually contiguous
>- fix mempool memory chunk iteration (bad pointer was used)
>- fix compilation on freebsd: replace MAP_LOCKED flag by mlock()
>- fix compilation on tilera (pointer arithmetics)
>- slightly rework and clean the mempool autotest
>- fix mempool autotest on bsd
>- more validation (especially mellanox drivers and kni that were not
>  tested in RFC)
>- passed autotests (x86_64-native-linuxapp-gcc and x86_64-native-bsdapp-gcc)
>- rebase on head, reorder the patches a bit and fix minor split issues
>
>
>Description of the initial issue
>--------------------------------
>
>The allocation of mbuf pool can fail even if there is enough memory.
>The problem is related to the way the memory is allocated and used in
>dpdk. It is particularly annoying with mbuf pools, but it can also fail
>in other use cases allocating a large amount of memory.
>
>- rte_malloc() allocates physically contiguous memory, which is needed
>  for mempools, but useless most of the time.
>
>  Allocating a large physically contiguous zone is often impossible
>  because the system provide hugepages which may not be contiguous.
>
>- rte_mempool_create() (and therefore rte_pktmbuf_pool_create())
>  requires a physically contiguous zone.
>
>- rte_mempool_xmem_create() does not solve the issue as it still
>  needs the memory to be virtually contiguous, and there is no
>  way in dpdk to allocate a virtually contiguous memory that is
>  not also physically contiguous.
>
>How to reproduce the issue
>--------------------------
>
>- start the dpdk with some 2MB hugepages (it can also occur with 1GB)
>- allocate a large mempool
>- even if there is enough memory, the allocation can fail
>
>Example:
>
>  git clone http://dpdk.org/git/dpdk
>  cd dpdk
>  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
>
>  # we try to allocate a mempool whose size is ~450MB, it fails
>  ./build/app/testpmd -l 2,4 -- --total-num-mbufs=200000 -i
>
>The EAL logs "EAL: Virtual area found at..." shows that there are
>several zones, but all smaller than 450MB.
>
>Workarounds:
>
>- Use 1GB hugepages: it sometimes work, but for very large
>  pools (millions of mbufs) there is the same issue. Moreover,
>  it would consume 1GB memory at least which can be a lot
>  in some cases.
>
>- Reboot the machine or allocate hugepages at boot time: this increases
>  the chances to have more contiguous memory, but does not completely
>  solve the issue
>
>Solutions
>---------
>
>Below is a list of proposed solutions. I implemented a quick and dirty
>PoC of solution 1, but it's not working in all conditions and it's
>really an ugly hack.  This series implement the solution 4 which looks
>the best to me, knowing it does not prevent to do more enhancements
>in dpdk memory in the future (solution 3 for instance).
>
>Solution 1: in application
>--------------------------
>
>- allocate several hugepages using rte_malloc() or rte_memzone_reserve()
>  (only keeping complete hugepages)
>- parse memsegs and /proc/maps to check which files mmaps these pages
>- mmap the files in a contiguous virtual area
>- use rte_mempool_xmem_create()
>
>Cons:
>
>- 1a. parsing the memsegs of rte config in the application does not
>  use a public API, and can be broken if internal dpdk code changes
>- 1b. some memory is lost due to malloc headers. Also, if the memory is
>  very fragmented (ex: all 2MB pages are physically separated), it does
>  not work at all because we cannot get any complete page. It is not
>  possible to use a lower level allocator since commit fafcc11985a.
>- 1c. we cannot use rte_pktmbuf_pool_create(), so we need to use mempool
>  api and do a part of the job manually
>- 1d. it breaks secondary processes as the virtual addresses won't be
>  mmap'd at the same place in secondary process
>- 1e. it only fixes the issue for the mbuf pool of the application,
>  internal pools in dpdk libraries are not modified
>- 1f. this is a pure linux solution (rte_map files)
>- 1g. The application has to be aware of RTE_EAL_SINGLE_SEGMENTS option
>  that changes the way hugepages are mapped. By the way, it's strange
>  to have such a compile-time option, we should probably have only
>  one behavior that works all the time.
>
>Solution 2: in dpdk memory allocator
>------------------------------------
>
>- do the same than solution 1 in a new function rte_malloc_non_contig():
>  allocate several chunks and mmap them in a contiguous virtual memory
>- a flag has to be added in malloc header to do the proper cleanup in
>  rte_free() (free all the chunks, munmap the memory)
>- introduce a new rte_mem_get_physmap(*physmap,addr, len) that returns
>  the virt2phys mapping of a virtual area in dpdk
>- add a mempool flag MEMPOOL_F_NON_PHYS_CONTIG to use
>  rte_malloc_non_contig() to allocate the area storing the objects
>
>Cons:
>
>- 2a. same than 1b: it breaks secondary processes if the mempool flag is
>  used.
>- 2b. same as 1d: some memory is lost due to malloc headers, and it
>  cannot work if memory is too fragmented.
>- 2c. rte_malloc_virt2phy() cannot be used on these zones. It would
>  return the physical address of the first page. It would be better to
>  return an error in this case.
>- 2d. need to check how to implement this on bsd (TBD)
>
>Solution 3: in dpdk eal memory
>------------------------------
>
>- Rework the way hugepages are mmap'd in dpdk: instead of having several
>  rte_map* files, just mmap one file per node. It may drastically
>  simplify EAL memory management in dpdk.
>- An API should be added to retrieve the physical mapping of a virtual
>  area (ex: rte_mem_get_physmap(*physmap, addr, len))
>- rte_malloc() and rte_memzone_reserve() won't allocate physically
>  contiguous memory anymore (TBD)
>- Update mempool to always use the rte_mempool_xmem_create() version
>
>Cons:
>
>- 3a. lot of rework in eal memory, it will induce some behavior changes
>  and maybe api changes
>- 3b. possible conflicts with xen_dom0 mempool
>
>Solution 4: in mempool
>----------------------
>
>- Introduce a new API to fill a mempool with zones that are not
>  virtually contiguous. It requires to add new functions to create and
>  populate a mempool. Example (TBD):
>
>  - rte_mempool_create_empty(name, n, elt_size, cache_size, priv_size)
>  - rte_mempool_populate(mp, addr, len): add virtual memory for objects
>  - rte_mempool_mempool_obj_iter(mp, obj_cb, arg): call a cb for each object
>
>- update rte_mempool_create() to allocate objects in several memory
>  chunks by default if there is no large enough physically contiguous
>  memory.
>
>Tests done
>----------
>
>Compilation
>~~~~~~~~~~~
>
>The following targets:
>
> x86_64-native-linuxapp-gcc
> i686-native-linuxapp-gcc
> x86_x32-native-linuxapp-gcc
> x86_64-native-linuxapp-clang
> x86_64-native-bsdapp-gcc
> ppc_64-power8-linuxapp-gcc
> tile-tilegx-linuxapp-gcc (only the mempool files, the target does not compile)
>
>Libraries with and without debug, in static and shared mode + examples.
>
>autotests
>~~~~~~~~~
>
>Passed all autotests on x86_64-native-linuxapp-gcc (including kni) and
>mempool-related autotests on x86_64-native-bsdapp-gcc.
>
>test-pmd
>~~~~~~~~
>
># now starts fine, was failing before if mempool was too fragmented
>./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -- -i --port-topology=chained
>
># still ok
>./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -m 256 -- -i --port-topology=chained --mp-anon
>set fwd txonly
>start
>stop
>
># fail, but was failing before too. The problem is because the physical
># addresses are not properly set when using --no-huge. The mempool phys addr
># are now correct, but the zones allocated through memzone_reserve() are
># still wrong. This could be fixed in a future series.
>./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -m 256 --no-huge -- -i ---port-topology=chained
>set fwd txonly
>start
>stop
>
>
>Olivier Matz (36):
>  mempool: fix comments and style
>  mempool: replace elt_size by total_elt_size
>  mempool: uninline function to check cookies
>  mempool: use sizeof to get the size of header and trailer
>  mempool: rename mempool_obj_ctor_t as mempool_obj_cb_t
>  mempool: update library version
>  mempool: list objects when added in the mempool
>  mempool: remove const attribute in mempool_walk
>  mempool: remove const qualifier in dump and audit
>  mempool: use the list to iterate the mempool elements
>  mempool: use the list to audit all elements
>  mempool: use the list to initialize mempool objects
>  mempool: create the internal ring in a specific function
>  mempool: store physaddr in mempool objects
>  mempool: remove MEMPOOL_IS_CONTIG()
>  mempool: store memory chunks in a list
>  mempool: new function to iterate the memory chunks
>  mempool: simplify xmem_usage
>  mempool: introduce a free callback for memory chunks
>  mempool: make page size optional when getting xmem size
>  mempool: default allocation in several memory chunks
>  eal: lock memory when using no-huge
>  mempool: support no-hugepage mode
>  mempool: replace mempool physaddr by a memzone pointer
>  mempool: introduce a function to free a mempool
>  mempool: introduce a function to create an empty mempool
>  eal/xen: return machine address without knowing memseg id
>  mempool: rework support of xen dom0
>  mempool: create the internal ring when populating
>  mempool: populate a mempool with anonymous memory
>  mempool: make mempool populate and free api public
>  test-pmd: remove specific anon mempool code
>  mem: avoid memzone/mempool/ring name truncation
>  mempool: new flag when phys contig mem is not needed
>  app/test: rework mempool test
>  mempool: update copyright
>
> app/test-pmd/Makefile                        |    4 -
> app/test-pmd/mempool_anon.c                  |  201 -----
> app/test-pmd/mempool_osdep.h                 |   54 --
> app/test-pmd/testpmd.c                       |   23 +-
> app/test/test_mempool.c                      |  243 +++---
> doc/guides/rel_notes/release_16_04.rst       |    2 +-
> drivers/net/mlx4/mlx4.c                      |  140 ++--
> drivers/net/mlx5/mlx5_rxtx.c                 |  140 ++--
> drivers/net/mlx5/mlx5_rxtx.h                 |    4 +-
> drivers/net/xenvirt/rte_eth_xenvirt.h        |    2 +-
> drivers/net/xenvirt/rte_mempool_gntalloc.c   |    4 +-
> lib/librte_eal/common/eal_common_log.c       |    2 +-
> lib/librte_eal/common/eal_common_memzone.c   |   10 +-
> lib/librte_eal/common/include/rte_memory.h   |   11 +-
> lib/librte_eal/linuxapp/eal/eal_memory.c     |    2 +-
> lib/librte_eal/linuxapp/eal/eal_xen_memory.c |   17 +-
> lib/librte_kni/rte_kni.c                     |   12 +-
> lib/librte_mempool/Makefile                  |    5 +-
> lib/librte_mempool/rte_dom0_mempool.c        |  133 ----
> lib/librte_mempool/rte_mempool.c             | 1042 +++++++++++++++++---------
> lib/librte_mempool/rte_mempool.h             |  594 +++++++--------
> lib/librte_mempool/rte_mempool_version.map   |   18 +-
> lib/librte_ring/rte_ring.c                   |   16 +-
> 23 files changed, 1377 insertions(+), 1302 deletions(-)
> delete mode 100644 app/test-pmd/mempool_anon.c
> delete mode 100644 app/test-pmd/mempool_osdep.h
> delete mode 100644 lib/librte_mempool/rte_dom0_mempool.c
>
>-- 
>2.1.4
>
>


Regards,
Keith





^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH v5] mempool: reduce rte_mempool structure size
  2016-04-14  9:42  2% ` [dpdk-dev] [PATCH v5] " Olivier Matz
  2016-04-14 13:28  0%   ` Wiles, Keith
@ 2016-04-14 13:53  0%   ` Wiles, Keith
  2016-05-17  5:31  0%   ` Thomas Monjalon
  2 siblings, 0 replies; 200+ results
From: Wiles, Keith @ 2016-04-14 13:53 UTC (permalink / raw)
  To: Olivier Matz, dev; +Cc: thomas.monjalon, pmatilai

>From: Keith Wiles <keith.wiles@intel.com>
>
>The rte_mempool structure is changed, which will cause an ABI change
>for this structure. Providing backward compat is not reasonable
>here as this structure is used in multiple defines/inlines.
>
>Allow mempool cache support to be dynamic depending on if the
>mempool being created needs cache support. Saves about 1.5M of
>memory used by the rte_mempool structure.
>
>Allocating small mempools which do not require cache can consume
>larges amounts of memory if you have a number of these mempools.
>
>Change to be effective in release 16.07.
>
>Signed-off-by: Keith Wiles <keith.wiles@intel.com>
>Acked-by: Olivier Matz <olivier.matz@6wind.com>

For the change to this patch:
Acked-by: Keith Wiles <keith.wiles@intel.com>

>---
>
>Changes in v5:
>
>- use RTE_PTR_ADD() instead of cast to (char *) to fix compilation on tilera.
>  Error log was:
>
>  rte_mempool.c: In function ‘rte_mempool_xmem_create’:
>  rte_mempool.c:595: error: cast increases required alignment of target type
>
>
> app/test/test_mempool.c          |  4 +--
> lib/librte_mempool/rte_mempool.c | 55 ++++++++++++++++++----------------------
> lib/librte_mempool/rte_mempool.h | 29 ++++++++++-----------
> 3 files changed, 40 insertions(+), 48 deletions(-)
>
>diff --git a/app/test/test_mempool.c b/app/test/test_mempool.c
>index f0f823b..10e1fa4 100644
>--- a/app/test/test_mempool.c
>+++ b/app/test/test_mempool.c
>@@ -122,8 +122,8 @@ test_mempool_basic(void)
> 		return -1;
> 
> 	printf("get private data\n");
>-	if (rte_mempool_get_priv(mp) !=
>-			(char*) mp + MEMPOOL_HEADER_SIZE(mp, mp->pg_num))
>+	if (rte_mempool_get_priv(mp) != (char *)mp +
>+			MEMPOOL_HEADER_SIZE(mp, mp->pg_num, mp->cache_size))
> 		return -1;
> 
> 	printf("get physical address of an object\n");
>diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
>index f8781e1..7a0e07e 100644
>--- a/lib/librte_mempool/rte_mempool.c
>+++ b/lib/librte_mempool/rte_mempool.c
>@@ -452,12 +452,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 	/* compilation-time checks */
> 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool) &
> 			  RTE_CACHE_LINE_MASK) != 0);
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool_cache) &
> 			  RTE_CACHE_LINE_MASK) != 0);
>-	RTE_BUILD_BUG_ON((offsetof(struct rte_mempool, local_cache) &
>-			  RTE_CACHE_LINE_MASK) != 0);
>-#endif
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
> 	RTE_BUILD_BUG_ON((sizeof(struct rte_mempool_debug_stats) &
> 			  RTE_CACHE_LINE_MASK) != 0);
>@@ -527,9 +523,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 		 */
> 		int head = sizeof(struct rte_mempool);
> 		int new_size = (private_data_size + head) % page_size;
>-		if (new_size) {
>+		if (new_size)
> 			private_data_size += page_size - new_size;
>-		}
> 	}
> 
> 	/* try to allocate tailq entry */
>@@ -544,7 +539,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 	 * store mempool objects. Otherwise reserve a memzone that is large
> 	 * enough to hold mempool header and metadata plus mempool objects.
> 	 */
>-	mempool_size = MEMPOOL_HEADER_SIZE(mp, pg_num) + private_data_size;
>+	mempool_size = MEMPOOL_HEADER_SIZE(mp, pg_num, cache_size);
>+	mempool_size += private_data_size;
> 	mempool_size = RTE_ALIGN_CEIL(mempool_size, RTE_MEMPOOL_ALIGN);
> 	if (vaddr == NULL)
> 		mempool_size += (size_t)objsz.total_size * n;
>@@ -591,8 +587,15 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 	mp->cache_flushthresh = CALC_CACHE_FLUSHTHRESH(cache_size);
> 	mp->private_data_size = private_data_size;
> 
>+	/*
>+	 * local_cache pointer is set even if cache_size is zero.
>+	 * The local_cache points to just past the elt_pa[] array.
>+	 */
>+	mp->local_cache = (struct rte_mempool_cache *)
>+		RTE_PTR_ADD(mp, MEMPOOL_HEADER_SIZE(mp, pg_num, 0));
>+
> 	/* calculate address of the first element for continuous mempool. */
>-	obj = (char *)mp + MEMPOOL_HEADER_SIZE(mp, pg_num) +
>+	obj = (char *)mp + MEMPOOL_HEADER_SIZE(mp, pg_num, cache_size) +
> 		private_data_size;
> 	obj = RTE_PTR_ALIGN_CEIL(obj, RTE_MEMPOOL_ALIGN);
> 
>@@ -606,9 +609,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
> 		mp->elt_va_start = (uintptr_t)obj;
> 		mp->elt_pa[0] = mp->phys_addr +
> 			(mp->elt_va_start - (uintptr_t)mp);
>-
>-	/* mempool elements in a separate chunk of memory. */
> 	} else {
>+		/* mempool elements in a separate chunk of memory. */
> 		mp->elt_va_start = (uintptr_t)vaddr;
> 		memcpy(mp->elt_pa, paddr, sizeof (mp->elt_pa[0]) * pg_num);
> 	}
>@@ -643,19 +645,15 @@ unsigned
> rte_mempool_count(const struct rte_mempool *mp)
> {
> 	unsigned count;
>+	unsigned lcore_id;
> 
> 	count = rte_ring_count(mp->ring);
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>-	{
>-		unsigned lcore_id;
>-		if (mp->cache_size == 0)
>-			return count;
>+	if (mp->cache_size == 0)
>+		return count;
> 
>-		for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
>-			count += mp->local_cache[lcore_id].len;
>-	}
>-#endif
>+	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
>+		count += mp->local_cache[lcore_id].len;
> 
> 	/*
> 	 * due to race condition (access to len is not locked), the
>@@ -670,13 +668,16 @@ rte_mempool_count(const struct rte_mempool *mp)
> static unsigned
> rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
> {
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	unsigned lcore_id;
> 	unsigned count = 0;
> 	unsigned cache_count;
> 
> 	fprintf(f, "  cache infos:\n");
> 	fprintf(f, "    cache_size=%"PRIu32"\n", mp->cache_size);
>+
>+	if (mp->cache_size == 0)
>+		return count;
>+
> 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
> 		cache_count = mp->local_cache[lcore_id].len;
> 		fprintf(f, "    cache_count[%u]=%u\n", lcore_id, cache_count);
>@@ -684,11 +685,6 @@ rte_mempool_dump_cache(FILE *f, const struct rte_mempool *mp)
> 	}
> 	fprintf(f, "    total_cache_count=%u\n", count);
> 	return count;
>-#else
>-	RTE_SET_USED(mp);
>-	fprintf(f, "  cache disabled\n");
>-	return 0;
>-#endif
> }
> 
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
>@@ -753,13 +749,16 @@ mempool_audit_cookies(const struct rte_mempool *mp)
> #define mempool_audit_cookies(mp) do {} while(0)
> #endif
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> /* check cookies before and after objects */
> static void
> mempool_audit_cache(const struct rte_mempool *mp)
> {
> 	/* check cache size consistency */
> 	unsigned lcore_id;
>+
>+	if (mp->cache_size == 0)
>+		return;
>+
> 	for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
> 		if (mp->local_cache[lcore_id].len > mp->cache_flushthresh) {
> 			RTE_LOG(CRIT, MEMPOOL, "badness on cache[%u]\n",
>@@ -768,10 +767,6 @@ mempool_audit_cache(const struct rte_mempool *mp)
> 		}
> 	}
> }
>-#else
>-#define mempool_audit_cache(mp) do {} while(0)
>-#endif
>-
> 
> /* check the consistency of mempool (size, cookies, ...) */
> void
>diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
>index 9745bf0..8595e77 100644
>--- a/lib/librte_mempool/rte_mempool.h
>+++ b/lib/librte_mempool/rte_mempool.h
>@@ -95,7 +95,6 @@ struct rte_mempool_debug_stats {
> } __rte_cache_aligned;
> #endif
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> /**
>  * A structure that stores a per-core object cache.
>  */
>@@ -107,7 +106,6 @@ struct rte_mempool_cache {
> 	 */
> 	void *objs[RTE_MEMPOOL_CACHE_MAX_SIZE * 3]; /**< Cache objects */
> } __rte_cache_aligned;
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> /**
>  * A structure that stores the size of mempool elements.
>@@ -194,10 +192,7 @@ struct rte_mempool {
> 
> 	unsigned private_data_size;      /**< Size of private data. */
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
>-	/** Per-lcore local cache. */
>-	struct rte_mempool_cache local_cache[RTE_MAX_LCORE];
>-#endif
>+	struct rte_mempool_cache *local_cache; /**< Per-lcore local cache */
> 
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
> 	/** Per-lcore statistics. */
>@@ -247,6 +242,13 @@ struct rte_mempool {
> #endif
> 
> /**
>+ * Size of elt_pa array size based on number of pages. (Internal use)
>+ */
>+#define __PA_SIZE(mp, pgn) \
>+	RTE_ALIGN_CEIL((((pgn) - RTE_DIM((mp)->elt_pa)) * \
>+	sizeof((mp)->elt_pa[0])), RTE_CACHE_LINE_SIZE)
>+
>+/**
>  * Calculate the size of the mempool header.
>  *
>  * @param mp
>@@ -254,9 +256,9 @@ struct rte_mempool {
>  * @param pgn
>  *   Number of pages used to store mempool objects.
>  */
>-#define	MEMPOOL_HEADER_SIZE(mp, pgn)	(sizeof(*(mp)) + \
>-	RTE_ALIGN_CEIL(((pgn) - RTE_DIM((mp)->elt_pa)) * \
>-	sizeof ((mp)->elt_pa[0]), RTE_CACHE_LINE_SIZE))
>+#define MEMPOOL_HEADER_SIZE(mp, pgn, cs) \
>+	(sizeof(*(mp)) + __PA_SIZE(mp, pgn) + (((cs) == 0) ? 0 : \
>+	(sizeof(struct rte_mempool_cache) * RTE_MAX_LCORE)))
> 
> /**
>  * Return true if the whole mempool is in contiguous memory.
>@@ -755,19 +757,16 @@ static inline void __attribute__((always_inline))
> __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
> 		    unsigned n, int is_mp)
> {
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	struct rte_mempool_cache *cache;
> 	uint32_t index;
> 	void **cache_objs;
> 	unsigned lcore_id = rte_lcore_id();
> 	uint32_t cache_size = mp->cache_size;
> 	uint32_t flushthresh = mp->cache_flushthresh;
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> 	/* increment stat now, adding in mempool always success */
> 	__MEMPOOL_STAT_ADD(mp, put, n);
> 
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	/* cache is not enabled or single producer or non-EAL thread */
> 	if (unlikely(cache_size == 0 || is_mp == 0 ||
> 		     lcore_id >= RTE_MAX_LCORE))
>@@ -802,7 +801,6 @@ __mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
> 	return;
> 
> ring_enqueue:
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> 	/* push remaining objects in ring */
> #ifdef RTE_LIBRTE_MEMPOOL_DEBUG
>@@ -946,7 +944,6 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
> 		   unsigned n, int is_mc)
> {
> 	int ret;
>-#if RTE_MEMPOOL_CACHE_MAX_SIZE > 0
> 	struct rte_mempool_cache *cache;
> 	uint32_t index, len;
> 	void **cache_objs;
>@@ -992,7 +989,6 @@ __mempool_get_bulk(struct rte_mempool *mp, void **obj_table,
> 	return 0;
> 
> ring_dequeue:
>-#endif /* RTE_MEMPOOL_CACHE_MAX_SIZE > 0 */
> 
> 	/* get remaining objects from ring */
> 	if (is_mc)
>@@ -1293,7 +1289,8 @@ void rte_mempool_audit(const struct rte_mempool *mp);
>  */
> static inline void *rte_mempool_get_priv(struct rte_mempool *mp)
> {
>-	return (char *)mp + MEMPOOL_HEADER_SIZE(mp, mp->pg_num);
>+	return (char *)mp +
>+		MEMPOOL_HEADER_SIZE(mp, mp->pg_num, mp->cache_size);
> }
> 
> /**
>-- 
>2.1.4
>
>


Regards,
Keith





^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v4 0/3] external mempool manager
  @ 2016-04-14 13:57  2% ` Olivier Matz
  0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-04-14 13:57 UTC (permalink / raw)
  To: dev, david.hunt; +Cc: yuanhan.liu, pmatilai

Here's a reworked version of the patch initially sent by David Hunt.
The main change is that it is rebased on top of the "mempool: rework
memory allocation" series [1], which simplifies a lot the first patch.

[1] http://dpdk.org/ml/archives/dev/2016-April/037464.html

v4 changes:
 * 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

Things that should still be discussed:

- Panu pointed out that having a compile-time configuration
  option for selecting the default mbuf handler is not a good idea.
  I mostly agree, except in one case (and that's why I kept this patch):
  if a specific architecture has its own way to provide an efficient
  pool handler for mbufs, it could be the proper place to have this
  option. But as far as I know, there is no such architecture today
  in dpdk.

- The other question I would like to raise is about the use cases.
  The cover letter below could be a bit more explicit about what this
  feature will be used for.



This is the initial unmodified cover letter from David Hunt:

Hi list.

Here's the v3 version patch for an external mempool manager

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 handler. This is achieved by adding a
     new mempool handler source file into the librte_mempool library, and
     using the REGISTER_MEMPOOL_HANDLER macro.
  2. Using the new API to call rte_mempool_create_ext to create a new mempool
     using the name parameter to identify which handler to use.

New API calls added
 1. A new mempool 'create' function which accepts mempool handler name.
 2. A new mempool 'rte_get_mempool_handler' function which accepts mempool
    handler name, and returns the index to the relevant set of callbacks for
    that mempool handler

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 handler 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 handlers may limit performance.

The new APIs are as follows:

1. rte_mempool_create_ext

struct rte_mempool *
rte_mempool_create_ext(const char * name, unsigned n,
        unsigned cache_size, unsigned private_data_size,
        int socket_id, unsigned flags,
        const char * handler_name);

2. rte_mempool_get_handler_name

char *
rte_mempool_get_handler_name(struct rte_mempool *mp);

Please see rte_mempool.h for further information on the parameters.


The important thing to note is that the mempool handler is passed by name
to rte_mempool_create_ext, and that in turn calls rte_get_mempool_handler to
get the handler index, which is stored in the rte_memool structure. This
allow multiple processes to use the same mempool, as the function pointers
are accessed via handler index.

The mempool handler structure contains callbacks to the implementation of
the handler, and is set up for registration as follows:

static struct rte_mempool_handler handler_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 handler in the array of handlers

REGISTER_MEMPOOL_HANDLER(handler_mp_mc);

For and example of a simple malloc based mempool manager, see
lib/librte_mempool/custom_mempool.c

For an example of API usage, please see app/test/test_ext_mempool.c, which
implements a rudimentary 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 handler
  mbuf: get default mempool handler from configuration

Olivier Matz (1):
  app/test: test external mempool handler

 app/test/test_mempool.c                    | 113 +++++++++++++++
 app/test/test_mempool_perf.c               |   1 -
 config/common_base                         |   1 +
 lib/librte_mbuf/rte_mbuf.c                 |  21 ++-
 lib/librte_mempool/Makefile                |   2 +
 lib/librte_mempool/rte_mempool.c           |  72 ++++------
 lib/librte_mempool/rte_mempool.h           | 212 +++++++++++++++++++++++++----
 lib/librte_mempool/rte_mempool_default.c   | 147 ++++++++++++++++++++
 lib/librte_mempool/rte_mempool_handler.c   | 139 +++++++++++++++++++
 lib/librte_mempool/rte_mempool_version.map |   4 +
 10 files changed, 637 insertions(+), 75 deletions(-)
 create mode 100644 lib/librte_mempool/rte_mempool_default.c
 create mode 100644 lib/librte_mempool/rte_mempool_handler.c

-- 
2.1.4

^ permalink raw reply	[relevance 2%]

* Re: [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation
  2016-04-14 13:50  0%   ` Wiles, Keith
@ 2016-04-14 14:01  0%     ` Olivier MATZ
  2016-04-14 14:03  0%       ` Wiles, Keith
  0 siblings, 1 reply; 200+ results
From: Olivier MATZ @ 2016-04-14 14:01 UTC (permalink / raw)
  To: Wiles, Keith, dev; +Cc: Richardson, Bruce, stephen



On 04/14/2016 03:50 PM, Wiles, Keith wrote:
>> This series is a rework of mempool. For those who don't want to read
>> all the cover letter, here is a sumary:
>>
>> - it is not possible to allocate large mempools if there is not enough
>>   contiguous memory, this series solves this issue
>> - introduce new APIs with less arguments: "create, populate, obj_init"
>> - allow to free a mempool
>> - split code in smaller functions, will ease the introduction of ext_handler
>> - remove test-pmd anonymous mempool creation
>> - remove most of dom0-specific mempool code
>> - opens the door for a eal_memory rework: we probably don't need large
>>   contiguous memory area anymore, working with pages would work.
>>
>> This breaks the ABI as it was indicated in the deprecation for 16.04.
>> The API stays almost the same, no modification is needed in examples app
>> or in test-pmd. Only kni and mellanox drivers are slightly modified.
>>
>> This patch applies on top of 16.04 + v5 of Keith's patch:
>> "mempool: reduce rte_mempool structure size"
>
> I have not digested this complete patch yet, but this one popped out at me as the External Memory Manager support is setting in the wings for 16.07 release. If this causes the EMM patch to be rewritten or updated that seems like a problem to me. Does this patch add the External Memory Manager support?
> http://thread.gmane.org/gmane.comp.networking.dpdk.devel/32015/focus=35107

I've reworked the series you are referring to, and rebased it on top
of this series. Please see:
http://dpdk.org/ml/archives/dev/2016-April/037509.html

Regards,
Olivier

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation
  2016-04-14 14:01  0%     ` Olivier MATZ
@ 2016-04-14 14:03  0%       ` Wiles, Keith
  0 siblings, 0 replies; 200+ results
From: Wiles, Keith @ 2016-04-14 14:03 UTC (permalink / raw)
  To: Olivier MATZ, dev; +Cc: Richardson, Bruce, stephen

>
>
>On 04/14/2016 03:50 PM, Wiles, Keith wrote:
>>> This series is a rework of mempool. For those who don't want to read
>>> all the cover letter, here is a sumary:
>>>
>>> - it is not possible to allocate large mempools if there is not enough
>>>   contiguous memory, this series solves this issue
>>> - introduce new APIs with less arguments: "create, populate, obj_init"
>>> - allow to free a mempool
>>> - split code in smaller functions, will ease the introduction of ext_handler
>>> - remove test-pmd anonymous mempool creation
>>> - remove most of dom0-specific mempool code
>>> - opens the door for a eal_memory rework: we probably don't need large
>>>   contiguous memory area anymore, working with pages would work.
>>>
>>> This breaks the ABI as it was indicated in the deprecation for 16.04.
>>> The API stays almost the same, no modification is needed in examples app
>>> or in test-pmd. Only kni and mellanox drivers are slightly modified.
>>>
>>> This patch applies on top of 16.04 + v5 of Keith's patch:
>>> "mempool: reduce rte_mempool structure size"
>>
>> I have not digested this complete patch yet, but this one popped out at me as the External Memory Manager support is setting in the wings for 16.07 release. If this causes the EMM patch to be rewritten or updated that seems like a problem to me. Does this patch add the External Memory Manager support?
>> http://thread.gmane.org/gmane.comp.networking.dpdk.devel/32015/focus=35107
>
>I've reworked the series you are referring to, and rebased it on top
>of this series. Please see:
>http://dpdk.org/ml/archives/dev/2016-April/037509.html

Thanks I just saw that update :-)

>
>Regards,
>Olivier
>


Regards,
Keith





^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH] port: bump ABI for pcap file support
@ 2016-04-14 18:33 25% Thomas Monjalon
  2016-04-15 10:32  4% ` Dumitrescu, Cristian
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-14 18:33 UTC (permalink / raw)
  To: roy.fan.zhang; +Cc: dev, cristian.dumitrescu, jasvinder.singh

Support of PCAP file has been added to rte_port in release 16.04
as NEXT_ABI. It is in the standard ABI of the release 16.07.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
 doc/guides/rel_notes/deprecation.rst   |  5 -----
 doc/guides/rel_notes/release_16_07.rst |  5 ++++-
 examples/ip_pipeline/init.c            |  4 ----
 lib/librte_port/Makefile               |  2 +-
 lib/librte_port/rte_port_source_sink.c | 14 --------------
 lib/librte_port/rte_port_source_sink.h |  3 ---
 6 files changed, 5 insertions(+), 28 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 327fc2b..a3fdbb1 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -79,11 +79,6 @@ Deprecation Notices
   modification of the API of rte_mempool_obj_iter(), implying a breakage
   of the ABI.
 
-* ABI changes are planned for struct rte_port_source_params in order to
-  support PCAP file reading feature. The release 16.04 contains this ABI
-  change wrapped by RTE_NEXT_ABI macro. Release 16.07 will contain this
-  change, and no backwards compatibility is planned.
-
 * 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/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 701e827..001888f 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -94,6 +94,9 @@ 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.
+  It was already in release 16.04 with ``RTE_NEXT_ABI`` flag.
+
 
 Shared Library Versions
 -----------------------
@@ -123,7 +126,7 @@ The libraries prepended with a plus sign were incremented in this version.
      librte_pipeline.so.3
      librte_pmd_bond.so.1
      librte_pmd_ring.so.2
-     librte_port.so.2
+   + librte_port.so.3
      librte_power.so.1
      librte_reorder.so.1
      librte_ring.so.1
diff --git a/examples/ip_pipeline/init.c b/examples/ip_pipeline/init.c
index 83422e8..02351f6 100644
--- a/examples/ip_pipeline/init.c
+++ b/examples/ip_pipeline/init.c
@@ -1221,8 +1221,6 @@ static void app_pipeline_params_get(struct app_params *app,
 			out->type = PIPELINE_PORT_IN_SOURCE;
 			out->params.source.mempool = app->mempool[mempool_id];
 			out->burst_size = app->source_params[in->id].burst;
-
-#ifdef RTE_NEXT_ABI
 			if (app->source_params[in->id].file_name
 				!= NULL) {
 				out->params.source.file_name = strdup(
@@ -1237,8 +1235,6 @@ static void app_pipeline_params_get(struct app_params *app,
 					app->source_params[in->id].
 					n_bytes_per_pkt;
 			}
-#endif
-
 			break;
 		default:
 			break;
diff --git a/lib/librte_port/Makefile b/lib/librte_port/Makefile
index 2c0ccbe..d4de5af 100644
--- a/lib/librte_port/Makefile
+++ b/lib/librte_port/Makefile
@@ -44,7 +44,7 @@ CFLAGS += $(WERROR_FLAGS)
 
 EXPORT_MAP := rte_port_version.map
 
-LIBABIVER := 2
+LIBABIVER := 3
 
 #
 # all source are stored in SRCS-y
diff --git a/lib/librte_port/rte_port_source_sink.c b/lib/librte_port/rte_port_source_sink.c
index 056c975..4cad710 100644
--- a/lib/librte_port/rte_port_source_sink.c
+++ b/lib/librte_port/rte_port_source_sink.c
@@ -38,17 +38,11 @@
 #include <rte_malloc.h>
 #include <rte_memcpy.h>
 
-#ifdef RTE_NEXT_ABI
-
 #ifdef RTE_PORT_PCAP
 #include <rte_ether.h>
 #include <pcap.h>
 #endif
 
-#else
-#undef RTE_PORT_PCAP
-#endif
-
 #include "rte_port_source_sink.h"
 
 /*
@@ -81,8 +75,6 @@ struct rte_port_source {
 	uint32_t pkt_index;
 };
 
-#ifdef RTE_NEXT_ABI
-
 #ifdef RTE_PORT_PCAP
 
 static int
@@ -232,8 +224,6 @@ error_exit:
 
 #endif /* RTE_PORT_PCAP */
 
-#endif /* RTE_NEXT_ABI */
-
 static void *
 rte_port_source_create(void *params, int socket_id)
 {
@@ -258,8 +248,6 @@ rte_port_source_create(void *params, int socket_id)
 	/* Initialization */
 	port->mempool = (struct rte_mempool *) p->mempool;
 
-#ifdef RTE_NEXT_ABI
-
 	if (p->file_name) {
 		int status = PCAP_SOURCE_LOAD(port, p->file_name,
 			p->n_bytes_per_pkt, socket_id);
@@ -270,8 +258,6 @@ rte_port_source_create(void *params, int socket_id)
 		}
 	}
 
-#endif
-
 	return port;
 }
 
diff --git a/lib/librte_port/rte_port_source_sink.h b/lib/librte_port/rte_port_source_sink.h
index 917abe4..4db8a8a 100644
--- a/lib/librte_port/rte_port_source_sink.h
+++ b/lib/librte_port/rte_port_source_sink.h
@@ -53,7 +53,6 @@ extern "C" {
 struct rte_port_source_params {
 	/** Pre-initialized buffer pool */
 	struct rte_mempool *mempool;
-#ifdef RTE_NEXT_ABI
 
 	/** The full path of the pcap file to read packets from */
 	char *file_name;
@@ -62,8 +61,6 @@ struct rte_port_source_params {
 	 *  if it is bigger than packet size, the generated packets
 	 *  will contain the whole packet */
 	uint32_t n_bytes_per_pkt;
-
-#endif
 };
 
 /** source port operations */
-- 
2.7.0

^ permalink raw reply	[relevance 25%]

* [dpdk-dev] [PATCH] pci: remove deprecated specific config
@ 2016-04-14 21:33  4% Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-14 21:33 UTC (permalink / raw)
  To: helin.zhang; +Cc: dev

The driver i40e was using a specific PCI config before the release 16.04.
>From 16.04, it is always enabled in i40e (commit 56465cfaf).
The API has been deprecated in the commit 68f77593823cab.
The igb_uio implementation has been deprecated in commit b7cf8e155.
The config helper - through igb_uio sysfs entries - is now removed.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
 config/common_base                        |  8 -----
 doc/guides/linux_gsg/enable_func.rst      | 15 ---------
 doc/guides/rel_notes/deprecation.rst      |  7 -----
 lib/librte_eal/common/include/rte_pci.h   | 14 ---------
 lib/librte_eal/linuxapp/eal/eal_pci.c     | 12 -------
 lib/librte_eal/linuxapp/igb_uio/igb_uio.c | 52 -------------------------------
 6 files changed, 108 deletions(-)

diff --git a/config/common_base b/config/common_base
index 0124e86..1a54e4c 100644
--- a/config/common_base
+++ b/config/common_base
@@ -101,14 +101,6 @@ CONFIG_RTE_MALLOC_DEBUG=n
 CONFIG_RTE_EAL_PMD_PATH=""
 
 #
-# Special configurations in PCI Config Space for high performance
-# They are all deprecated, and will be removed later.
-#
-CONFIG_RTE_PCI_CONFIG=n
-CONFIG_RTE_PCI_EXTENDED_TAG=""
-CONFIG_RTE_PCI_MAX_READ_REQUEST_SIZE=0
-
-#
 # Compile Environment Abstraction Layer to support Vmware TSC map
 #
 CONFIG_RTE_LIBRTE_EAL_VMWARE_TSC_MAP_SUPPORT=y
diff --git a/doc/guides/linux_gsg/enable_func.rst b/doc/guides/linux_gsg/enable_func.rst
index 076770f..ec0e04d 100644
--- a/doc/guides/linux_gsg/enable_func.rst
+++ b/doc/guides/linux_gsg/enable_func.rst
@@ -186,21 +186,6 @@ Check with the local Intel's Network Division application engineers for firmware
 The base driver to support firmware version of FVL3E will be integrated in the next
 DPDK release, so currently the validated firmware version is 4.2.6.
 
-Enabling Extended Tag
-~~~~~~~~~~~~~~~~~~~~~
-
-PCI configuration of ``extended_tag`` has big impact on small packet size
-performance of 40G ports. Enabling ``extended_tag`` can help 40G port to
-achieve the best performance, especially for small packet size.
-
-* Disabling/enabling ``extended_tag`` can be done in some BIOS implementations.
-
-* If BIOS does not enable it, and does not support changing it, tools
-  (e.g. ``setpci`` on Linux) can be used to enable or disable ``extended_tag``.
-
-* From release 16.04, ``extended_tag`` is enabled by default during port
-  initialization, users don't need to care about that anymore.
-
 Use 16 Bytes RX Descriptor Size
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index a3fdbb1..c78cde7 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -20,13 +20,6 @@ Deprecation Notices
   do not need to care about the kind of devices that are being used, making it
   easier to add new buses later.
 
-* The EAL function pci_config_space_set is deprecated in release 16.04
-  and will be removed from 16.07.
-  Macros CONFIG_RTE_PCI_CONFIG, CONFIG_RTE_PCI_EXTENDED_TAG and
-  CONFIG_RTE_PCI_MAX_READ_REQUEST_SIZE will be removed.
-  The /sys entries extended_tag and max_read_request_size created by igb_uio
-  will be removed.
-
 * ABI changes are planned for struct rte_pci_id, i.e., add new field ``class``.
   This new added ``class`` field can be used to probe pci device by class
   related info. This change should impact size of struct rte_pci_id and struct
diff --git a/lib/librte_eal/common/include/rte_pci.h b/lib/librte_eal/common/include/rte_pci.h
index e692094..9f2301d 100644
--- a/lib/librte_eal/common/include/rte_pci.h
+++ b/lib/librte_eal/common/include/rte_pci.h
@@ -577,20 +577,6 @@ void rte_eal_pci_ioport_read(struct rte_pci_ioport *p,
 void rte_eal_pci_ioport_write(struct rte_pci_ioport *p,
 			      const void *data, size_t len, off_t offset);
 
-#ifdef RTE_PCI_CONFIG
-#include <rte_common.h>
-/**
- * Set special config space registers for performance purpose.
- * It is deprecated, as all configurations have been moved into
- * each PMDs respectively.
- *
- * @param dev
- *   A pointer to a rte_pci_device structure describing the device
- *   to use
- */
-void pci_config_space_set(struct rte_pci_device *dev) __rte_deprecated;
-#endif /* RTE_PCI_CONFIG */
-
 #ifdef __cplusplus
 }
 #endif
diff --git a/lib/librte_eal/linuxapp/eal/eal_pci.c b/lib/librte_eal/linuxapp/eal/eal_pci.c
index dbf12a8..bdc08a0 100644
--- a/lib/librte_eal/linuxapp/eal/eal_pci.c
+++ b/lib/librte_eal/linuxapp/eal/eal_pci.c
@@ -481,18 +481,6 @@ error:
 	return -1;
 }
 
-#ifdef RTE_PCI_CONFIG
-/*
- * It is deprecated, all its configurations have been moved into
- * each PMD respectively.
- */
-void
-pci_config_space_set(__rte_unused struct rte_pci_device *dev)
-{
-	RTE_LOG(DEBUG, EAL, "Nothing here, as it is deprecated\n");
-}
-#endif
-
 /* Read PCI config space. */
 int rte_eal_pci_read_config(const struct rte_pci_device *device,
 			    void *buf, size_t len, off_t offset)
diff --git a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
index 72b2692..45a5720 100644
--- a/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
+++ b/lib/librte_eal/linuxapp/igb_uio/igb_uio.c
@@ -81,62 +81,10 @@ store_max_vfs(struct device *dev, struct device_attribute *attr,
 	return err ? err : count;
 }
 
-#ifdef RTE_PCI_CONFIG
-static ssize_t
-show_extended_tag(struct device *dev, struct device_attribute *attr, char *buf)
-{
-	dev_info(dev, "Deprecated\n");
-
-	return 0;
-}
-
-static ssize_t
-store_extended_tag(struct device *dev,
-		   struct device_attribute *attr,
-		   const char *buf,
-		   size_t count)
-{
-	dev_info(dev, "Deprecated\n");
-
-	return 0;
-}
-
-static ssize_t
-show_max_read_request_size(struct device *dev,
-			   struct device_attribute *attr,
-			   char *buf)
-{
-	dev_info(dev, "Deprecated\n");
-
-	return 0;
-}
-
-static ssize_t
-store_max_read_request_size(struct device *dev,
-			    struct device_attribute *attr,
-			    const char *buf,
-			    size_t count)
-{
-	dev_info(dev, "Deprecated\n");
-
-	return 0;
-}
-#endif
-
 static DEVICE_ATTR(max_vfs, S_IRUGO | S_IWUSR, show_max_vfs, store_max_vfs);
-#ifdef RTE_PCI_CONFIG
-static DEVICE_ATTR(extended_tag, S_IRUGO | S_IWUSR, show_extended_tag,
-	store_extended_tag);
-static DEVICE_ATTR(max_read_request_size, S_IRUGO | S_IWUSR,
-	show_max_read_request_size, store_max_read_request_size);
-#endif
 
 static struct attribute *dev_attrs[] = {
 	&dev_attr_max_vfs.attr,
-#ifdef RTE_PCI_CONFIG
-	&dev_attr_extended_tag.attr,
-	&dev_attr_max_read_request_size.attr,
-#endif
 	NULL,
 };
 
-- 
2.7.0

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [RFC 1/2] doc: announce ABI change for rte_eth_dev_info structure
  2016-04-14  9:44  9% ` [dpdk-dev] [RFC 1/2] doc: announce ABI change for " Reshma Pattan
@ 2016-04-15  9:42  4%   ` Mcnamara, John
  2016-04-15 10:02  8%   ` Thomas Monjalon
  1 sibling, 0 replies; 200+ results
From: Mcnamara, John @ 2016-04-15  9:42 UTC (permalink / raw)
  To: Pattan, Reshma, dev



> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Reshma Pattan
> Sent: Thursday, April 14, 2016 10:45 AM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [RFC 1/2] 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>

Acked-by: John McNamara <john.mcnamara@intel.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [RFC 1/2] doc: announce ABI change for rte_eth_dev_info structure
  2016-04-14  9:44  9% ` [dpdk-dev] [RFC 1/2] doc: announce ABI change for " Reshma Pattan
  2016-04-15  9:42  4%   ` Mcnamara, John
@ 2016-04-15 10:02  8%   ` Thomas Monjalon
  1 sibling, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-15 10:02 UTC (permalink / raw)
  To: Reshma Pattan; +Cc: dev

2016-04-14 10:44, 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>

In general the Signed-off lines are the same as the From: field.
Here it would be:
Signed-off-by: Reshma Pattan <reshma.pattan@intel.com>
(note the spaces and the uppercase)

> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -90,3 +90,9 @@ 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.

It is too late for this announce as it won't appear in the doc downloaded for
version 16.04. So it is obviously rejected.
The question here is: are you allowed to do a small ABI change given that
the ABI will be broken in this version?
I would say there can be some exceptional tolerance.
I have no strong opinion myself but maybe others will have one.

By the way, I have some comments about the patch.

^ permalink raw reply	[relevance 8%]

* Re: [dpdk-dev] [PATCH] port: bump ABI for pcap file support
  2016-04-14 18:33 25% [dpdk-dev] [PATCH] port: bump ABI for pcap file support Thomas Monjalon
@ 2016-04-15 10:32  4% ` Dumitrescu, Cristian
  2016-04-20  9:55  4%   ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Dumitrescu, Cristian @ 2016-04-15 10:32 UTC (permalink / raw)
  To: Thomas Monjalon, Zhang, Roy Fan; +Cc: dev, Singh, Jasvinder



> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Thursday, April 14, 2016 7:34 PM
> To: Zhang, Roy Fan <roy.fan.zhang@intel.com>
> Cc: dev@dpdk.org; Dumitrescu, Cristian <cristian.dumitrescu@intel.com>;
> Singh, Jasvinder <jasvinder.singh@intel.com>
> Subject: [PATCH] port: bump ABI for pcap file support
> 
> Support of PCAP file has been added to rte_port in release 16.04
> as NEXT_ABI. It is in the standard ABI of the release 16.07.
> 
> Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> ---
>  doc/guides/rel_notes/deprecation.rst   |  5 -----
>  doc/guides/rel_notes/release_16_07.rst |  5 ++++-
>  examples/ip_pipeline/init.c            |  4 ----
>  lib/librte_port/Makefile               |  2 +-
>  lib/librte_port/rte_port_source_sink.c | 14 --------------
>  lib/librte_port/rte_port_source_sink.h |  3 ---
>  6 files changed, 5 insertions(+), 28 deletions(-)
> 


Acked-by: Cristian Dumitrescu <Cristian.Dumitrescu@intel.com>

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [RFC 2/2] librte_ether: add new fields to rte_eth_dev_info struct
  @ 2016-04-15 10:36  3%   ` Thomas Monjalon
  2016-04-15 11:32  0%     ` Ananyev, Konstantin
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-15 10:36 UTC (permalink / raw)
  To: Reshma Pattan; +Cc: dev

2016-04-14 10:44, Reshma Pattan:
> --- a/lib/librte_ether/rte_ethdev.h
> +++ b/lib/librte_ether/rte_ethdev.h
> @@ -908,6 +908,9 @@ 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_). */
> +	/** number of queues configured by software*/
> +	uint16_t nb_rx_queues; /**< Number of RX queues. */
> +	uint16_t nb_tx_queues; /**< Number of TX queues. */
>  };

I think the ethdev design is strange for these structures.
struct rte_eth_dev is internal to be used inside the ethdev API
or by the drivers.
It contains struct rte_eth_dev_data which can be of interest for
the application, except the dev_private part (which could be
directly in struct rte_eth_dev).

So the global question is: how to share the device data with the
application?
Instead of giving a pointer or a copy of struct rte_eth_dev_data,
we have some different accessors:
	- rte_eth_dev_info_get() with a specific struct rte_eth_dev_info
which gathers a lot of info, not only from struct rte_eth_dev_data
	- rte_eth_macaddr_get()
	- rte_eth_dev_socket_id()
	- rte_eth_link_get() which is more than an accessor

I think having some specialized accessors is good.
But the rte_eth_dev_info_get() looks like to be a big request
without precise goal and going to break ABI really often.
There are some queues informations, some (not so precise)
offload capabilities, some steering (RSS/VMDq) informations,
the default configuration of some Intel NIC thresholds,
the speed capabilities, etc.

Shouldn't we try to streamline this API?

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [RFC 2/2] librte_ether: add new fields to rte_eth_dev_info struct
  2016-04-15 10:36  3%   ` Thomas Monjalon
@ 2016-04-15 11:32  0%     ` Ananyev, Konstantin
  0 siblings, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-04-15 11:32 UTC (permalink / raw)
  To: Thomas Monjalon, Pattan, Reshma; +Cc: dev

Hi everyone,

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Thomas Monjalon
> Sent: Friday, April 15, 2016 11:36 AM
> To: Pattan, Reshma
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [RFC 2/2] librte_ether: add new fields to rte_eth_dev_info struct
> 
> 2016-04-14 10:44, Reshma Pattan:
> > --- a/lib/librte_ether/rte_ethdev.h
> > +++ b/lib/librte_ether/rte_ethdev.h
> > @@ -908,6 +908,9 @@ 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_). */
> > +	/** number of queues configured by software*/
> > +	uint16_t nb_rx_queues; /**< Number of RX queues. */
> > +	uint16_t nb_tx_queues; /**< Number of TX queues. */
> >  };
> 
> I think the ethdev design is strange for these structures.
> struct rte_eth_dev is internal to be used inside the ethdev API
> or by the drivers.
> It contains struct rte_eth_dev_data which can be of interest for
> the application, except the dev_private part (which could be
> directly in struct rte_eth_dev).
> 
> So the global question is: how to share the device data with the
> application?
> Instead of giving a pointer or a copy of struct rte_eth_dev_data,
> we have some different accessors:
> 	- rte_eth_dev_info_get() with a specific struct rte_eth_dev_info
> which gathers a lot of info, not only from struct rte_eth_dev_data
> 	- rte_eth_macaddr_get()
> 	- rte_eth_dev_socket_id()
> 	- rte_eth_link_get() which is more than an accessor
> 
> I think having some specialized accessors is good.
> But the rte_eth_dev_info_get() looks like to be a big request
> without precise goal and going to break ABI really often.
> There are some queues informations, some (not so precise)
> offload capabilities, some steering (RSS/VMDq) informations,
> the default configuration of some Intel NIC thresholds,
> the speed capabilities, etc.
> 
> Shouldn't we try to streamline this API?

I think in general it is a good idea  to split dev_info into some smaller sub-pieces.
But introduce a new API just for these 2 fields seems like an overkill to me.
My vote would be to allow nb_rx/tx_queues into dev_info,
If we'll decide to split dev_info - I think it needs to be a subject for a separate   
patch/discussion.
Konstantin

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats
@ 2016-04-15 14:44  3% Remy Horton
  2016-04-20 16:03  0% ` David Harton (dharton)
  2016-04-29 12:52  0% ` David Harton (dharton)
  0 siblings, 2 replies; 200+ results
From: Remy Horton @ 2016-04-15 14:44 UTC (permalink / raw)
  To: dev

The current extended ethernet statistics fetching involve doing several
string operations, which causes performance issues if there are lots of
statistics and/or network interfaces. This RFC patchset changes the API
for xstats to use integer identifiers instead of strings and implements
this new API for the ixgbe driver. Others drivers to follow.

--

Since this will involve API & ABI breakage as previously advertised,
there are several design assumptions that need consideration:

*) id-name & id-value pairs for both lookup and query
Permits out-of-order and non-contigious returning of names/ids/values,
even though expected implmentations would in practice return items in
sorted order by id. Is this sufficent/desirable future proofing? Idea
is to allow possibility of drivers returning partial statistics.

*) Bulk name-id mapping lookup only
At the moment individual lookup is not supported, as this would impose
extra overheads on drivers. The assumption is that any end user would
fetch all this data once on startup and then cache the mappings.

*) Replacement or additional API
This patch replaces the current xstats API, but there is no inherant
reason beyond maintainability why this funtionality could not be in
addition rather than a replacement. What is consensus on this?

Comments welcome.

Remy Horton (3):
  rte: change xstats to use integer keys
  drivers/net/ixgbe: change xstats to use integer keys
  examples/ethtool: add xstats display command

 drivers/net/ixgbe/ixgbe_ethdev.c      | 87 +++++++++++++++++++++++++++++++----
 examples/ethtool/ethtool-app/ethapp.c | 57 +++++++++++++++++++++++
 lib/librte_ether/rte_ethdev.c         | 87 +++++++++++++++++++++++++++++++----
 lib/librte_ether/rte_ethdev.h         | 38 +++++++++++++++
 4 files changed, 252 insertions(+), 17 deletions(-)

-- 
2.5.5

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH] mk: do not enforce any specific ARM ABI
@ 2016-04-15 22:33 15% Jan Viktorin
  2016-05-02 15:47  4% ` Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Jan Viktorin @ 2016-04-15 22:33 UTC (permalink / raw)
  To: dev; +Cc: Jan Viktorin

The dpdk build system passes -mfloat-abi=softfp, which makes the build fail
when the selected ABI is EABIhf. The dpdk build system should not make
assumptions on the selected ARM ABI.

Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 mk/machine/armv7a/rte.vars.mk | 2 --
 1 file changed, 2 deletions(-)

diff --git a/mk/machine/armv7a/rte.vars.mk b/mk/machine/armv7a/rte.vars.mk
index abdb15e..36fa3de 100644
--- a/mk/machine/armv7a/rte.vars.mk
+++ b/mk/machine/armv7a/rte.vars.mk
@@ -54,8 +54,6 @@
 # CPU_LDFLAGS =
 # CPU_ASFLAGS =
 
-CPU_CFLAGS += -mfloat-abi=softfp
-
 MACHINE_CFLAGS += -march=armv7-a
 
 ifdef CONFIG_RTE_ARCH_ARM_TUNE
-- 
2.8.0

^ permalink raw reply	[relevance 15%]

* [dpdk-dev] [PATCH] ethdev: remove deprecated statistics
@ 2016-04-19 14:03  4% Thomas Monjalon
  2016-04-20  9:47  4% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
  0 siblings, 1 reply; 200+ results
From: Thomas Monjalon @ 2016-04-19 14:03 UTC (permalink / raw)
  To: maryam.tahhan, harry.van.haaren; +Cc: dev

Some statistics were deprecated since release 2.1 (49f386542af4).

The last deprecated counter to be used was imcasts.

The new counters should be added to extended statistics.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
 doc/guides/rel_notes/deprecation.rst   |  4 ----
 doc/guides/rel_notes/release_16_07.rst |  6 +++++-
 drivers/net/bonding/rte_eth_bond_pmd.c |  1 -
 drivers/net/cxgbe/cxgbe_ethdev.c       |  1 -
 drivers/net/e1000/igb_ethdev.c         |  1 -
 drivers/net/ena/ena_ethdev.c           |  2 --
 drivers/net/ena/ena_ethdev.h           |  1 -
 drivers/net/enic/enic_main.c           |  1 -
 drivers/net/i40e/i40e_ethdev.c         |  1 -
 drivers/net/ixgbe/ixgbe_ethdev.c       |  4 ----
 drivers/net/nfp/nfp_net.c              | 18 ------------------
 drivers/net/vmxnet3/vmxnet3_ethdev.c   |  1 -
 lib/librte_ether/Makefile              |  2 +-
 lib/librte_ether/rte_ethdev.h          | 18 ------------------
 14 files changed, 6 insertions(+), 55 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index c78cde7..fffe9c7 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -26,10 +26,6 @@ Deprecation Notices
   rte_pci_device. The release 16.04 does not contain these ABI changes, but
   release 16.07 will.
 
-* The following fields have been deprecated in rte_eth_stats:
-  ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
-  tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff
-
 * The xstats API and rte_eth_xstats struct will be changed to allow retrieval
   of values without any string copies or parsing.
   No backwards compatibility is planned, as it would require code duplication
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 001888f..83c841b 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -86,6 +86,10 @@ 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.
 
+* 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.
+
 
 ABI Changes
 -----------
@@ -107,7 +111,7 @@ The libraries prepended with a plus sign were incremented in this version.
 
 .. code-block:: diff
 
-     libethdev.so.3
+   + libethdev.so.4
      librte_acl.so.2
      librte_cfgfile.so.2
      librte_cmdline.so.2
diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c b/drivers/net/bonding/rte_eth_bond_pmd.c
index 54788cf..c897146 100644
--- a/drivers/net/bonding/rte_eth_bond_pmd.c
+++ b/drivers/net/bonding/rte_eth_bond_pmd.c
@@ -1836,7 +1836,6 @@ bond_ethdev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 		stats->imissed += slave_stats.imissed;
 		stats->ierrors += slave_stats.ierrors;
 		stats->oerrors += slave_stats.oerrors;
-		stats->imcasts += slave_stats.imcasts;
 		stats->rx_nombuf += slave_stats.rx_nombuf;
 
 		for (j = 0; j < RTE_ETHDEV_QUEUE_STAT_CNTRS; j++) {
diff --git a/drivers/net/cxgbe/cxgbe_ethdev.c b/drivers/net/cxgbe/cxgbe_ethdev.c
index bb134e5..04eddaf 100644
--- a/drivers/net/cxgbe/cxgbe_ethdev.c
+++ b/drivers/net/cxgbe/cxgbe_ethdev.c
@@ -656,7 +656,6 @@ static void cxgbe_dev_stats_get(struct rte_eth_dev *eth_dev,
 	/* RX Stats */
 	eth_stats->ipackets = ps.rx_frames;
 	eth_stats->ibytes   = ps.rx_octets;
-	eth_stats->imcasts  = ps.rx_mcast_frames;
 	eth_stats->imissed  = ps.rx_ovflow0 + ps.rx_ovflow1 +
 			      ps.rx_ovflow2 + ps.rx_ovflow3 +
 			      ps.rx_trunc0 + ps.rx_trunc1 +
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index e0053fe..e7682da 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -1805,7 +1805,6 @@ eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
 	rte_stats->ibytes = hw_stats->gorc;
 	rte_stats->opackets = hw_stats->gptc;
 	rte_stats->obytes = hw_stats->gotc;
-	rte_stats->imcasts = hw_stats->mprc;
 	rte_stats->ilbpackets = hw_stats->gprlbc;
 	rte_stats->ilbbytes = hw_stats->gorlbc;
 	rte_stats->olbpackets = hw_stats->gptlbc;
diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 02af67a..e157587 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -605,7 +605,6 @@ static void ena_stats_restart(struct rte_eth_dev *dev)
 
 	rte_atomic64_init(&adapter->drv_stats->ierrors);
 	rte_atomic64_init(&adapter->drv_stats->oerrors);
-	rte_atomic64_init(&adapter->drv_stats->imcasts);
 	rte_atomic64_init(&adapter->drv_stats->rx_nombuf);
 }
 
@@ -643,7 +642,6 @@ static void ena_stats_get(struct rte_eth_dev *dev,
 	/* Driver related stats */
 	stats->ierrors = rte_atomic64_read(&adapter->drv_stats->ierrors);
 	stats->oerrors = rte_atomic64_read(&adapter->drv_stats->oerrors);
-	stats->imcasts = rte_atomic64_read(&adapter->drv_stats->imcasts);
 	stats->rx_nombuf = rte_atomic64_read(&adapter->drv_stats->rx_nombuf);
 }
 
diff --git a/drivers/net/ena/ena_ethdev.h b/drivers/net/ena/ena_ethdev.h
index ba6f01e..aca853c 100644
--- a/drivers/net/ena/ena_ethdev.h
+++ b/drivers/net/ena/ena_ethdev.h
@@ -121,7 +121,6 @@ enum ena_adapter_state {
 struct ena_driver_stats {
 	rte_atomic64_t ierrors;
 	rte_atomic64_t oerrors;
-	rte_atomic64_t imcasts;
 	rte_atomic64_t rx_nombuf;
 };
 
diff --git a/drivers/net/enic/enic_main.c b/drivers/net/enic/enic_main.c
index e3da51d..60fe765 100644
--- a/drivers/net/enic/enic_main.c
+++ b/drivers/net/enic/enic_main.c
@@ -248,7 +248,6 @@ void enic_dev_stats_get(struct enic *enic, struct rte_eth_stats *r_stats)
 
 	r_stats->imissed = stats->rx.rx_drop;
 
-	r_stats->imcasts = stats->rx.rx_multicast_frames_ok;
 	r_stats->rx_nombuf = stats->rx.rx_no_bufs;
 }
 
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index bc28d3c..d8b6bd7 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -2099,7 +2099,6 @@ i40e_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	stats->obytes   = ns->eth.tx_bytes;
 	stats->oerrors  = ns->eth.tx_errors +
 			pf->main_vsi->eth_stats.tx_errors;
-	stats->imcasts  = pf->main_vsi->eth_stats.rx_multicast;
 
 	/* Rx Errors */
 	stats->imissed  = ns->eth.rx_discards +
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index 3f1ebc1..eec607c 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -2852,8 +2852,6 @@ ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	stats->ibytes = hw_stats->vfgorc;
 	stats->opackets = hw_stats->vfgptc;
 	stats->obytes = hw_stats->vfgotc;
-	stats->imcasts = hw_stats->vfmprc;
-	/* stats->imcasts should be removed as imcasts is deprecated */
 }
 
 static void
@@ -2870,8 +2868,6 @@ ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
 	hw_stats->vfgorc = 0;
 	hw_stats->vfgptc = 0;
 	hw_stats->vfgotc = 0;
-	hw_stats->vfmprc = 0;
-
 }
 
 static void
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index bcf5fa9..bc0a3d8 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -902,11 +902,6 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 	nfp_dev_stats.obytes -= hw->eth_stats_base.obytes;
 
-	nfp_dev_stats.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
-	nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
-
 	/* reading general device stats */
 	nfp_dev_stats.ierrors =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
@@ -918,12 +913,6 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 	nfp_dev_stats.oerrors -= hw->eth_stats_base.oerrors;
 
-	/* Multicast frames received */
-	nfp_dev_stats.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
-	nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
-
 	/* RX ring mbuf allocation failures */
 	nfp_dev_stats.rx_nombuf = dev->data->rx_mbuf_alloc_failed;
 
@@ -985,9 +974,6 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	hw->eth_stats_base.obytes =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_OCTETS);
 
-	hw->eth_stats_base.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
 	/* reading general device stats */
 	hw->eth_stats_base.ierrors =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
@@ -995,10 +981,6 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	hw->eth_stats_base.oerrors =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_ERRORS);
 
-	/* Multicast frames received */
-	hw->eth_stats_base.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
 	/* RX ring mbuf allocation failures */
 	dev->data->rx_mbuf_alloc_failed = 0;
 
diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
index bd7a2bb..29b469c 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
@@ -694,7 +694,6 @@ vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 		stats->q_errors[i] = rxStats->pktsRxError;
 		stats->ierrors += rxStats->pktsRxError;
-		stats->imcasts += rxStats->mcastPktsRxOK;
 		stats->rx_nombuf += rxStats->pktsRxOutOfBuf;
 	}
 }
diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
index e810284..0bb5dc9 100644
--- a/lib/librte_ether/Makefile
+++ b/lib/librte_ether/Makefile
@@ -41,7 +41,7 @@ CFLAGS += $(WERROR_FLAGS)
 
 EXPORT_MAP := rte_ether_version.map
 
-LIBABIVER := 3
+LIBABIVER := 4
 
 SRCS-y += rte_ethdev.c
 
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 022733e..d749980 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -200,27 +200,9 @@ struct rte_eth_stats {
 	/**< Total of RX packets dropped by the HW,
 	 * because there are no available mbufs (i.e. RX queues are full).
 	 */
-	uint64_t ibadcrc __rte_deprecated;
-	/**< Deprecated; Total of RX packets with CRC error. */
-	uint64_t ibadlen __rte_deprecated;
-	/**< Deprecated; Total of RX packets with bad length. */
 	uint64_t ierrors;   /**< Total number of erroneous received packets. */
 	uint64_t oerrors;   /**< Total number of failed transmitted packets. */
-	uint64_t imcasts;
-	/**< Deprecated; Total number of multicast received packets. */
 	uint64_t rx_nombuf; /**< Total number of RX mbuf allocation failures. */
-	uint64_t fdirmatch __rte_deprecated;
-	/**< Deprecated; Total number of RX packets matching a filter. */
-	uint64_t fdirmiss __rte_deprecated;
-	/**< Deprecated; Total number of RX packets not matching any filter. */
-	uint64_t tx_pause_xon __rte_deprecated;
-	 /**< Deprecated; Total nb. of XON pause frame sent. */
-	uint64_t rx_pause_xon __rte_deprecated;
-	/**< Deprecated; Total nb. of XON pause frame received. */
-	uint64_t tx_pause_xoff __rte_deprecated;
-	/**< Deprecated; Total nb. of XOFF pause frame sent. */
-	uint64_t rx_pause_xoff __rte_deprecated;
-	/**< Deprecated; Total nb. of XOFF pause frame received. */
 	uint64_t q_ipackets[RTE_ETHDEV_QUEUE_STAT_CNTRS];
 	/**< Total number of queue RX packets. */
 	uint64_t q_opackets[RTE_ETHDEV_QUEUE_STAT_CNTRS];
-- 
2.7.0

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] perfomance of rte_lpm rule subsystem
  @ 2016-04-19 15:46  3% ` Stephen Hemminger
  2016-04-19 20:46  0%   ` Vladimir Medvedkin
  2016-05-02 19:38  0%   ` Александр Киселев
  0 siblings, 2 replies; 200+ results
From: Stephen Hemminger @ 2016-04-19 15:46 UTC (permalink / raw)
  To: Александр
	Киселев
  Cc: dev

On Tue, 19 Apr 2016 14:11:11 +0300
Александр Киселев <kiselev99@gmail.com> wrote:

> Hi.
> 
> Doing some test with rte_lpm (adding/deleting bgp full table rules) I
> noticed that
> rule subsystem is very slow even considering that probably it was never
> designed for using
> in a data forwarding plane. So I want to propose some changes to the "rule"
> subsystem.
> 
> I reimplemented rule part ot the lib using rte_hash, and perfomance of
> adding/deleted routes have increased dramatically.
> If increasing speed of adding deleting routes makes sence for anybody else
> I would like to discuss my patch.
> The patch also include changes that make next_hop 64 bit, so please just
> ignore them. The rule changes are in the following
> functions only:
> 
> rte_lpm2_create
> 
> rule_find
> rule_add
> rule_delete
> find_previous_rule
> delete_depth_small
> delete_depth_big
> 
> rte_lpm2_add
> rte_lpm2_delete
> rte_lpm2_is_rule_present
> rte_lpm2_delete_all
> 

We forked LPM back several versions ago.
I sent the patches to use BSD red-black tree for rules but the patches were
ignored. mostly because it broke ABI.

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] perfomance of rte_lpm rule subsystem
  2016-04-19 15:46  3% ` Stephen Hemminger
@ 2016-04-19 20:46  0%   ` Vladimir Medvedkin
  2016-05-02 19:38  0%   ` Александр Киселев
  1 sibling, 0 replies; 200+ results
From: Vladimir Medvedkin @ 2016-04-19 20:46 UTC (permalink / raw)
  To: Stephen Hemminger
  Cc: Александр
	Киселев,
	dev

Hi Alexander,

Why next_hop is 64 bit long?


2016-04-19 18:46 GMT+03:00 Stephen Hemminger <stephen@networkplumber.org>:

> On Tue, 19 Apr 2016 14:11:11 +0300
> Александр Киселев <kiselev99@gmail.com> wrote:
>
> > Hi.
> >
> > Doing some test with rte_lpm (adding/deleting bgp full table rules) I
> > noticed that
> > rule subsystem is very slow even considering that probably it was never
> > designed for using
> > in a data forwarding plane. So I want to propose some changes to the
> "rule"
> > subsystem.
> >
> > I reimplemented rule part ot the lib using rte_hash, and perfomance of
> > adding/deleted routes have increased dramatically.
> > If increasing speed of adding deleting routes makes sence for anybody
> else
> > I would like to discuss my patch.
> > The patch also include changes that make next_hop 64 bit, so please just
> > ignore them. The rule changes are in the following
> > functions only:
> >
> > rte_lpm2_create
> >
> > rule_find
> > rule_add
> > rule_delete
> > find_previous_rule
> > delete_depth_small
> > delete_depth_big
> >
> > rte_lpm2_add
> > rte_lpm2_delete
> > rte_lpm2_is_rule_present
> > rte_lpm2_delete_all
> >
>
> We forked LPM back several versions ago.
> I sent the patches to use BSD red-black tree for rules but the patches were
> ignored. mostly because it broke ABI.
>



-- 
Regards,
Vladimir

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v2] ethdev: remove deprecated statistics
  2016-04-19 14:03  4% [dpdk-dev] [PATCH] ethdev: remove deprecated statistics Thomas Monjalon
@ 2016-04-20  9:47  4% ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-20  9:47 UTC (permalink / raw)
  To: maryam.tahhan, harry.van.haaren; +Cc: dev

Some statistics were deprecated since release 2.1 (49f386542af4).
The last deprecated counter to be used was imcasts.

The VF loopback statistics are also removed as they are used only
in igb and duplicated in extended statistics.

The new counters should be added to extended statistics.

Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
---
v2:
- remove VF loopback stats
- remove comment in ethtool example

 doc/guides/rel_notes/deprecation.rst   |  4 ----
 doc/guides/rel_notes/release_16_07.rst |  6 +++++-
 drivers/net/bonding/rte_eth_bond_pmd.c |  1 -
 drivers/net/cxgbe/cxgbe_ethdev.c       |  1 -
 drivers/net/e1000/igb_ethdev.c         |  5 -----
 drivers/net/ena/ena_ethdev.c           |  2 --
 drivers/net/ena/ena_ethdev.h           |  1 -
 drivers/net/enic/enic_main.c           |  1 -
 drivers/net/i40e/i40e_ethdev.c         |  1 -
 drivers/net/ixgbe/ixgbe_ethdev.c       |  4 ----
 drivers/net/nfp/nfp_net.c              | 18 ------------------
 drivers/net/vmxnet3/vmxnet3_ethdev.c   |  1 -
 examples/ethtool/ethtool-app/ethapp.c  |  1 -
 lib/librte_ether/Makefile              |  2 +-
 lib/librte_ether/rte_ethdev.h          | 26 --------------------------
 15 files changed, 6 insertions(+), 68 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index c78cde7..fffe9c7 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -26,10 +26,6 @@ Deprecation Notices
   rte_pci_device. The release 16.04 does not contain these ABI changes, but
   release 16.07 will.
 
-* The following fields have been deprecated in rte_eth_stats:
-  ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
-  tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff
-
 * The xstats API and rte_eth_xstats struct will be changed to allow retrieval
   of values without any string copies or parsing.
   No backwards compatibility is planned, as it would require code duplication
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 001888f..83c841b 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -86,6 +86,10 @@ 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.
 
+* 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.
+
 
 ABI Changes
 -----------
@@ -107,7 +111,7 @@ The libraries prepended with a plus sign were incremented in this version.
 
 .. code-block:: diff
 
-     libethdev.so.3
+   + libethdev.so.4
      librte_acl.so.2
      librte_cfgfile.so.2
      librte_cmdline.so.2
diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c b/drivers/net/bonding/rte_eth_bond_pmd.c
index 54788cf..c897146 100644
--- a/drivers/net/bonding/rte_eth_bond_pmd.c
+++ b/drivers/net/bonding/rte_eth_bond_pmd.c
@@ -1836,7 +1836,6 @@ bond_ethdev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 		stats->imissed += slave_stats.imissed;
 		stats->ierrors += slave_stats.ierrors;
 		stats->oerrors += slave_stats.oerrors;
-		stats->imcasts += slave_stats.imcasts;
 		stats->rx_nombuf += slave_stats.rx_nombuf;
 
 		for (j = 0; j < RTE_ETHDEV_QUEUE_STAT_CNTRS; j++) {
diff --git a/drivers/net/cxgbe/cxgbe_ethdev.c b/drivers/net/cxgbe/cxgbe_ethdev.c
index bb134e5..04eddaf 100644
--- a/drivers/net/cxgbe/cxgbe_ethdev.c
+++ b/drivers/net/cxgbe/cxgbe_ethdev.c
@@ -656,7 +656,6 @@ static void cxgbe_dev_stats_get(struct rte_eth_dev *eth_dev,
 	/* RX Stats */
 	eth_stats->ipackets = ps.rx_frames;
 	eth_stats->ibytes   = ps.rx_octets;
-	eth_stats->imcasts  = ps.rx_mcast_frames;
 	eth_stats->imissed  = ps.rx_ovflow0 + ps.rx_ovflow1 +
 			      ps.rx_ovflow2 + ps.rx_ovflow3 +
 			      ps.rx_trunc0 + ps.rx_trunc1 +
diff --git a/drivers/net/e1000/igb_ethdev.c b/drivers/net/e1000/igb_ethdev.c
index e0053fe..f0921ee 100644
--- a/drivers/net/e1000/igb_ethdev.c
+++ b/drivers/net/e1000/igb_ethdev.c
@@ -1805,11 +1805,6 @@ eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
 	rte_stats->ibytes = hw_stats->gorc;
 	rte_stats->opackets = hw_stats->gptc;
 	rte_stats->obytes = hw_stats->gotc;
-	rte_stats->imcasts = hw_stats->mprc;
-	rte_stats->ilbpackets = hw_stats->gprlbc;
-	rte_stats->ilbbytes = hw_stats->gorlbc;
-	rte_stats->olbpackets = hw_stats->gptlbc;
-	rte_stats->olbbytes = hw_stats->gotlbc;
 }
 
 static void
diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 02af67a..e157587 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -605,7 +605,6 @@ static void ena_stats_restart(struct rte_eth_dev *dev)
 
 	rte_atomic64_init(&adapter->drv_stats->ierrors);
 	rte_atomic64_init(&adapter->drv_stats->oerrors);
-	rte_atomic64_init(&adapter->drv_stats->imcasts);
 	rte_atomic64_init(&adapter->drv_stats->rx_nombuf);
 }
 
@@ -643,7 +642,6 @@ static void ena_stats_get(struct rte_eth_dev *dev,
 	/* Driver related stats */
 	stats->ierrors = rte_atomic64_read(&adapter->drv_stats->ierrors);
 	stats->oerrors = rte_atomic64_read(&adapter->drv_stats->oerrors);
-	stats->imcasts = rte_atomic64_read(&adapter->drv_stats->imcasts);
 	stats->rx_nombuf = rte_atomic64_read(&adapter->drv_stats->rx_nombuf);
 }
 
diff --git a/drivers/net/ena/ena_ethdev.h b/drivers/net/ena/ena_ethdev.h
index ba6f01e..aca853c 100644
--- a/drivers/net/ena/ena_ethdev.h
+++ b/drivers/net/ena/ena_ethdev.h
@@ -121,7 +121,6 @@ enum ena_adapter_state {
 struct ena_driver_stats {
 	rte_atomic64_t ierrors;
 	rte_atomic64_t oerrors;
-	rte_atomic64_t imcasts;
 	rte_atomic64_t rx_nombuf;
 };
 
diff --git a/drivers/net/enic/enic_main.c b/drivers/net/enic/enic_main.c
index e3da51d..60fe765 100644
--- a/drivers/net/enic/enic_main.c
+++ b/drivers/net/enic/enic_main.c
@@ -248,7 +248,6 @@ void enic_dev_stats_get(struct enic *enic, struct rte_eth_stats *r_stats)
 
 	r_stats->imissed = stats->rx.rx_drop;
 
-	r_stats->imcasts = stats->rx.rx_multicast_frames_ok;
 	r_stats->rx_nombuf = stats->rx.rx_no_bufs;
 }
 
diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethdev.c
index bc28d3c..d8b6bd7 100644
--- a/drivers/net/i40e/i40e_ethdev.c
+++ b/drivers/net/i40e/i40e_ethdev.c
@@ -2099,7 +2099,6 @@ i40e_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	stats->obytes   = ns->eth.tx_bytes;
 	stats->oerrors  = ns->eth.tx_errors +
 			pf->main_vsi->eth_stats.tx_errors;
-	stats->imcasts  = pf->main_vsi->eth_stats.rx_multicast;
 
 	/* Rx Errors */
 	stats->imissed  = ns->eth.rx_discards +
diff --git a/drivers/net/ixgbe/ixgbe_ethdev.c b/drivers/net/ixgbe/ixgbe_ethdev.c
index 3f1ebc1..eec607c 100644
--- a/drivers/net/ixgbe/ixgbe_ethdev.c
+++ b/drivers/net/ixgbe/ixgbe_ethdev.c
@@ -2852,8 +2852,6 @@ ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 	stats->ibytes = hw_stats->vfgorc;
 	stats->opackets = hw_stats->vfgptc;
 	stats->obytes = hw_stats->vfgotc;
-	stats->imcasts = hw_stats->vfmprc;
-	/* stats->imcasts should be removed as imcasts is deprecated */
 }
 
 static void
@@ -2870,8 +2868,6 @@ ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
 	hw_stats->vfgorc = 0;
 	hw_stats->vfgptc = 0;
 	hw_stats->vfgotc = 0;
-	hw_stats->vfmprc = 0;
-
 }
 
 static void
diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index bcf5fa9..bc0a3d8 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -902,11 +902,6 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 	nfp_dev_stats.obytes -= hw->eth_stats_base.obytes;
 
-	nfp_dev_stats.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
-	nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
-
 	/* reading general device stats */
 	nfp_dev_stats.ierrors =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
@@ -918,12 +913,6 @@ nfp_net_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 	nfp_dev_stats.oerrors -= hw->eth_stats_base.oerrors;
 
-	/* Multicast frames received */
-	nfp_dev_stats.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
-	nfp_dev_stats.imcasts -= hw->eth_stats_base.imcasts;
-
 	/* RX ring mbuf allocation failures */
 	nfp_dev_stats.rx_nombuf = dev->data->rx_mbuf_alloc_failed;
 
@@ -985,9 +974,6 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	hw->eth_stats_base.obytes =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_OCTETS);
 
-	hw->eth_stats_base.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
 	/* reading general device stats */
 	hw->eth_stats_base.ierrors =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_ERRORS);
@@ -995,10 +981,6 @@ nfp_net_stats_reset(struct rte_eth_dev *dev)
 	hw->eth_stats_base.oerrors =
 		nn_cfg_readq(hw, NFP_NET_CFG_STATS_TX_ERRORS);
 
-	/* Multicast frames received */
-	hw->eth_stats_base.imcasts =
-		nn_cfg_readq(hw, NFP_NET_CFG_STATS_RX_MC_FRAMES);
-
 	/* RX ring mbuf allocation failures */
 	dev->data->rx_mbuf_alloc_failed = 0;
 
diff --git a/drivers/net/vmxnet3/vmxnet3_ethdev.c b/drivers/net/vmxnet3/vmxnet3_ethdev.c
index bd7a2bb..29b469c 100644
--- a/drivers/net/vmxnet3/vmxnet3_ethdev.c
+++ b/drivers/net/vmxnet3/vmxnet3_ethdev.c
@@ -694,7 +694,6 @@ vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 
 		stats->q_errors[i] = rxStats->pktsRxError;
 		stats->ierrors += rxStats->pktsRxError;
-		stats->imcasts += rxStats->mcastPktsRxOK;
 		stats->rx_nombuf += rxStats->pktsRxOutOfBuf;
 	}
 }
diff --git a/examples/ethtool/ethtool-app/ethapp.c b/examples/ethtool/ethtool-app/ethapp.c
index 2ed4796..38e466c 100644
--- a/examples/ethtool/ethtool-app/ethapp.c
+++ b/examples/ethtool/ethtool-app/ethapp.c
@@ -535,7 +535,6 @@ static void pcmd_portstats_callback(__rte_unused void *ptr_params,
 	}
 	stat = rte_ethtool_net_get_stats64(params->port, &stat_info);
 	if (stat == 0) {
-		/* Most of rte_eth_stats is deprecated.. */
 		printf("Port %i stats\n", params->port);
 		printf("   In: %" PRIu64 " (%" PRIu64 " bytes)\n"
 			"  Out: %"PRIu64" (%"PRIu64 " bytes)\n"
diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
index e810284..0bb5dc9 100644
--- a/lib/librte_ether/Makefile
+++ b/lib/librte_ether/Makefile
@@ -41,7 +41,7 @@ CFLAGS += $(WERROR_FLAGS)
 
 EXPORT_MAP := rte_ether_version.map
 
-LIBABIVER := 3
+LIBABIVER := 4
 
 SRCS-y += rte_ethdev.c
 
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 022733e..2757510 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -200,27 +200,9 @@ struct rte_eth_stats {
 	/**< Total of RX packets dropped by the HW,
 	 * because there are no available mbufs (i.e. RX queues are full).
 	 */
-	uint64_t ibadcrc __rte_deprecated;
-	/**< Deprecated; Total of RX packets with CRC error. */
-	uint64_t ibadlen __rte_deprecated;
-	/**< Deprecated; Total of RX packets with bad length. */
 	uint64_t ierrors;   /**< Total number of erroneous received packets. */
 	uint64_t oerrors;   /**< Total number of failed transmitted packets. */
-	uint64_t imcasts;
-	/**< Deprecated; Total number of multicast received packets. */
 	uint64_t rx_nombuf; /**< Total number of RX mbuf allocation failures. */
-	uint64_t fdirmatch __rte_deprecated;
-	/**< Deprecated; Total number of RX packets matching a filter. */
-	uint64_t fdirmiss __rte_deprecated;
-	/**< Deprecated; Total number of RX packets not matching any filter. */
-	uint64_t tx_pause_xon __rte_deprecated;
-	 /**< Deprecated; Total nb. of XON pause frame sent. */
-	uint64_t rx_pause_xon __rte_deprecated;
-	/**< Deprecated; Total nb. of XON pause frame received. */
-	uint64_t tx_pause_xoff __rte_deprecated;
-	/**< Deprecated; Total nb. of XOFF pause frame sent. */
-	uint64_t rx_pause_xoff __rte_deprecated;
-	/**< Deprecated; Total nb. of XOFF pause frame received. */
 	uint64_t q_ipackets[RTE_ETHDEV_QUEUE_STAT_CNTRS];
 	/**< Total number of queue RX packets. */
 	uint64_t q_opackets[RTE_ETHDEV_QUEUE_STAT_CNTRS];
@@ -231,14 +213,6 @@ struct rte_eth_stats {
 	/**< Total number of successfully transmitted queue bytes. */
 	uint64_t q_errors[RTE_ETHDEV_QUEUE_STAT_CNTRS];
 	/**< Total number of queue packets received that are dropped. */
-	uint64_t ilbpackets;
-	/**< Total number of good packets received from loopback,VF Only */
-	uint64_t olbpackets;
-	/**< Total number of good packets transmitted to loopback,VF Only */
-	uint64_t ilbbytes;
-	/**< Total number of good bytes received from loopback,VF Only */
-	uint64_t olbbytes;
-	/**< Total number of good bytes transmitted to loopback,VF Only */
 };
 
 /**
-- 
2.7.0

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] [PATCH] port: bump ABI for pcap file support
  2016-04-15 10:32  4% ` Dumitrescu, Cristian
@ 2016-04-20  9:55  4%   ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-04-20  9:55 UTC (permalink / raw)
  To: Dumitrescu, Cristian; +Cc: Zhang, Roy Fan, dev, Singh, Jasvinder

> > Support of PCAP file has been added to rte_port in release 16.04
> > as NEXT_ABI. It is in the standard ABI of the release 16.07.
> > 
> > Signed-off-by: Thomas Monjalon <thomas.monjalon@6wind.com>
> 
> Acked-by: Cristian Dumitrescu <Cristian.Dumitrescu@intel.com>

Applied

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v2 04/17] eal: remove duplicate function declaration
  @ 2016-04-20 11:44  3%   ` David Marchand
  0 siblings, 0 replies; 200+ results
From: David Marchand @ 2016-04-20 11:44 UTC (permalink / raw)
  To: dev; +Cc: thomas.monjalon, viktorin

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 2342fa1..855fd25 100644
--- a/lib/librte_eal/common/eal_private.h
+++ b/lib/librte_eal/common/eal_private.h
@@ -262,13 +262,6 @@ 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 8aafd51..f26f8d3 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>
-- 
1.9.1

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats
  2016-04-15 14:44  3% [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats Remy Horton
@ 2016-04-20 16:03  0% ` David Harton (dharton)
  2016-04-28 14:56  0%   ` Tahhan, Maryam
  2016-04-29 12:52  0% ` David Harton (dharton)
  1 sibling, 1 reply; 200+ results
From: David Harton (dharton) @ 2016-04-20 16:03 UTC (permalink / raw)
  To: Remy Horton, dev

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Remy Horton
> Sent: Friday, April 15, 2016 10:44 AM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from
> xstats
> 
> The current extended ethernet statistics fetching involve doing several
> string operations, which causes performance issues if there are lots of
> statistics and/or network interfaces. This RFC patchset changes the API
> for xstats to use integer identifiers instead of strings and implements
> this new API for the ixgbe driver. Others drivers to follow.
> 
> --
> 
> Since this will involve API & ABI breakage as previously advertised, there
> are several design assumptions that need consideration:
> 
> *) id-name & id-value pairs for both lookup and query Permits out-of-order
> and non-contigious returning of names/ids/values, even though expected
> implmentations would in practice return items in sorted order by id. Is
> this sufficent/desirable future proofing? Idea is to allow possibility of
> drivers returning partial statistics.

I believe forcing drivers to match to a common id-space will become 
burdensome.  If the stats id-space isn't common then matching strings is 
probably just as sufficient as long as drivers don't add/remove stats
ad hoc between the time the device is initialized and removed.

> 
> *) Bulk name-id mapping lookup only
> At the moment individual lookup is not supported, as this would impose
> extra overheads on drivers. The assumption is that any end user would
> fetch all this data once on startup and then cache the mappings.

I'm not sure I see the value of looking up a single stat from a user
perspective.  I can see where the drivers might say that some stats
are less disruptive/etc but the user doesn't have that knowledge and
wouldn't know how to take advantage.  Usually all stats are grabbed
multiple times and the changes noted during debug sessions.

> 
> *) Replacement or additional API
> This patch replaces the current xstats API, but there is no inherant
> reason beyond maintainability why this funtionality could not be in
> addition rather than a replacement. What is consensus on this?

I came to the conclusion that replacing the existing API isn't necessary
but rather extending it so backwards compatibility could be maintained
during the previous discussions on this topic.  However, if we want to 
go forward with cleaning up in order to reduce the support drivers
provide I'm all for it.

I still believe the API we develop should follow an "ethtool stats like"
format as suggested earlier this year:

extern int rte_eth_xstats_names_get(uint8_t port_id,
        struct rte_eth_xstats_name *names, unsigned n);
extern int rte_eth_xstats_values_get(uint8_t port_id,
        uint64_t *values, unsigned n);

Again, these could be provided alongside the existing API or replace it.

I also like the idea you provided of a separate API to obtain the
xstats count rather than deriving the count by calling one of the
above functions with "dummy" values.

Again, I can provide the patches for the changes I've made that align
with this proposed API.  I just never got any feedback on it when
requested previously.

Regards,
Dave

> 
> Comments welcome.
> 
> Remy Horton (3):
>   rte: change xstats to use integer keys
>   drivers/net/ixgbe: change xstats to use integer keys
>   examples/ethtool: add xstats display command
> 
>  drivers/net/ixgbe/ixgbe_ethdev.c      | 87
> +++++++++++++++++++++++++++++++----
>  examples/ethtool/ethtool-app/ethapp.c | 57 +++++++++++++++++++++++
>  lib/librte_ether/rte_ethdev.c         | 87
> +++++++++++++++++++++++++++++++----
>  lib/librte_ether/rte_ethdev.h         | 38 +++++++++++++++
>  4 files changed, 252 insertions(+), 17 deletions(-)
> 
> --
> 2.5.5

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [RFC] eal: provide option to set vhost_user socket owner/permissions
@ 2016-04-25  9:18  2% Christian Ehrhardt
  2016-04-26  4:16  0% ` Yuanhan Liu
  0 siblings, 1 reply; 200+ results
From: Christian Ehrhardt @ 2016-04-25  9:18 UTC (permalink / raw)
  To: christian.ehrhardt, Aaron Conole, dev

The API doesn't hold a way to specify a owner/permission set for vhost_user
created sockets. I don't even think an API change would make that much sense.

Projects consuming DPDK start to do 'their own workarounds' like openvswitch
https://patchwork.ozlabs.org/patch/559043/
https://patchwork.ozlabs.org/patch/559045/
But for this specific example they are blocked/stalled behind a bigger
rework (https://patchwork.ozlabs.org/patch/604898/).
Also one could ask why each project would need their own workaround.

At the same time - as I want it for existing code linking against DPDK I
wanted to avoid changing API/ABI. That way I want to provide something existing
users could utilize. So I created a DPDK EAL commandline option based ideas in
the former patches.

For myself I consider this a nice interim solution for existing released
Openvswitch+DPDK solution. And I intend to put it as delta into the DPDK 2.2
currently packaged in Ubuntu to get it working more smoothly with
openvswitch 2.5.

But I'd be interested if DPDK in general would be interested in:
a) an approach like this?
b) would prefer a change of the API?
c) consider it an issue of consuming projects and let them take care?

Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
---
 doc/guides/testpmd_app_ug/run_app.rst        |  19 +++
 lib/librte_eal/common/eal_common_options.c   |   4 +
 lib/librte_eal/common/eal_internal_cfg.h     |   2 +
 lib/librte_eal/common/eal_options.h          |   4 +
 lib/librte_eal/common/include/rte_eal.h      |   5 +
 lib/librte_eal/linuxapp/eal/eal.c            | 182 +++++++++++++++++++++++++++
 lib/librte_vhost/vhost_user/vhost-net-user.c |   4 +
 7 files changed, 220 insertions(+)

diff --git a/doc/guides/testpmd_app_ug/run_app.rst b/doc/guides/testpmd_app_ug/run_app.rst
index f605564..24c9c01 100644
--- a/doc/guides/testpmd_app_ug/run_app.rst
+++ b/doc/guides/testpmd_app_ug/run_app.rst
@@ -156,6 +156,25 @@ See the DPDK Getting Started Guides for more information on these options.
 
     Use malloc instead of hugetlbfs.
 
+*   ``--vhost-owner``
+
+     When creating vhost_user sockets change owner and group to the specified value.
+     This can be given as ``user:group``, but also only ``user`` or ``:group`` are supported.
+
+    Examples::
+
+       --vhost-owner 'libvirt-qemu:kvm'
+       --vhost-owner 'libvirt-qemu'
+       --vhost-owner ':kvm'
+
+*   ``--vhost-perm``
+
+    When creating vhost_user sockets set them up with these permissions.
+
+    For example::
+
+       --vhost-perm '0664'
+
 
 Testpmd Command-line Options
 ----------------------------
diff --git a/lib/librte_eal/common/eal_common_options.c b/lib/librte_eal/common/eal_common_options.c
index 2b418d5..073198b 100644
--- a/lib/librte_eal/common/eal_common_options.c
+++ b/lib/librte_eal/common/eal_common_options.c
@@ -95,6 +95,8 @@ eal_long_options[] = {
 	{OPT_VFIO_INTR,         1, NULL, OPT_VFIO_INTR_NUM        },
 	{OPT_VMWARE_TSC_MAP,    0, NULL, OPT_VMWARE_TSC_MAP_NUM   },
 	{OPT_XEN_DOM0,          0, NULL, OPT_XEN_DOM0_NUM         },
+	{OPT_VHOST_OWNER,       1, NULL, OPT_VHOST_OWNER_NUM      },
+	{OPT_VHOST_PERM,        1, NULL, OPT_VHOST_PERM_NUM       },
 	{0,                     0, NULL, 0                        }
 };
 
@@ -153,6 +155,8 @@ eal_reset_internal_config(struct internal_config *internal_cfg)
 #endif
 	internal_cfg->vmware_tsc_map = 0;
 	internal_cfg->create_uio_dev = 0;
+	internal_cfg->vhost_sock_owner = NULL;
+	internal_cfg->vhost_sock_perm = NULL;
 }
 
 static int
diff --git a/lib/librte_eal/common/eal_internal_cfg.h b/lib/librte_eal/common/eal_internal_cfg.h
index 5f1367e..bdf34e3 100644
--- a/lib/librte_eal/common/eal_internal_cfg.h
+++ b/lib/librte_eal/common/eal_internal_cfg.h
@@ -83,6 +83,8 @@ struct internal_config {
 	volatile enum rte_intr_mode vfio_intr_mode;
 	const char *hugefile_prefix;      /**< the base filename of hugetlbfs files */
 	const char *hugepage_dir;         /**< specific hugetlbfs directory to use */
+	const char *vhost_sock_owner;     /**< owner:group of vhost_user sockets */
+	const char *vhost_sock_perm;      /**< permissions of vhost_user sockets */
 
 	unsigned num_hugepage_sizes;      /**< how many sizes on this system */
 	struct hugepage_info hugepage_info[MAX_HUGEPAGE_SIZES];
diff --git a/lib/librte_eal/common/eal_options.h b/lib/librte_eal/common/eal_options.h
index a881c62..1161083 100644
--- a/lib/librte_eal/common/eal_options.h
+++ b/lib/librte_eal/common/eal_options.h
@@ -83,6 +83,10 @@ enum {
 	OPT_VMWARE_TSC_MAP_NUM,
 #define OPT_XEN_DOM0          "xen-dom0"
 	OPT_XEN_DOM0_NUM,
+#define OPT_VHOST_OWNER       "vhost-owner"
+	OPT_VHOST_OWNER_NUM,
+#define OPT_VHOST_PERM        "vhost-perm"
+	OPT_VHOST_PERM_NUM,
 	OPT_LONG_MAX_NUM
 };
 
diff --git a/lib/librte_eal/common/include/rte_eal.h b/lib/librte_eal/common/include/rte_eal.h
index a71d6f5..506cf24 100644
--- a/lib/librte_eal/common/include/rte_eal.h
+++ b/lib/librte_eal/common/include/rte_eal.h
@@ -252,6 +252,11 @@ static inline int rte_gettid(void)
 	return RTE_PER_LCORE(_thread_id);
 }
 
+/**
+ * Set owner/permissions on sockets if requested on EAL commandline
+ */
+void rte_eal_set_socket_permissions(const char *);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index 8aafd51..3d0b709 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -53,6 +53,9 @@
 #if defined(RTE_ARCH_X86)
 #include <sys/io.h>
 #endif
+#include <sys/types.h>
+#include <pwd.h>
+#include <grp.h>
 
 #include <rte_common.h>
 #include <rte_debug.h>
@@ -343,6 +346,8 @@ eal_usage(const char *prgname)
 	       "  --"OPT_CREATE_UIO_DEV"    Create /dev/uioX (usually done by hotplug)\n"
 	       "  --"OPT_VFIO_INTR"         Interrupt mode for VFIO (legacy|msi|msix)\n"
 	       "  --"OPT_XEN_DOM0"          Support running on Xen dom0 without hugetlbfs\n"
+	       "  --"OPT_VHOST_OWNER"       Create vhost-user sockets with this owner:group\n"
+	       "  --"OPT_VHOST_PERM"        Create vhost-user sockets with these permissions\n"
 	       "\n");
 	/* Allow the application to print its usage message too if hook is set */
 	if ( rte_application_usage_hook ) {
@@ -618,6 +623,14 @@ eal_parse_args(int argc, char **argv)
 			internal_config.create_uio_dev = 1;
 			break;
 
+		case OPT_VHOST_OWNER_NUM:
+			internal_config.vhost_sock_owner = optarg;
+			break;
+
+		case OPT_VHOST_PERM_NUM:
+			internal_config.vhost_sock_perm = optarg;
+			break;
+
 		default:
 			if (opt < OPT_LONG_MIN_NUM && isprint(opt)) {
 				RTE_LOG(ERR, EAL, "Option %c is not supported "
@@ -934,3 +947,172 @@ rte_eal_check_module(const char *module_name)
 	/* Module has been found */
 	return 1;
 }
+
+/* Try to double the size of '*buf', return true
+ * if successful, and '*sizep' will be updated with
+ * the new size. Otherwise, return false.  */
+static int
+enlarge_buffer(char **buf, size_t *sizep)
+{
+    size_t newsize = *sizep * 2;
+
+    if (newsize > *sizep) {
+        *buf = realloc(*buf, newsize);
+        *sizep = newsize;
+        return 1;
+    }
+
+    return 0;
+}
+
+static int
+get_owners_from_str(const char *user_spec, uid_t *uid, gid_t *gid)
+{
+	size_t bufsize = 4096;
+
+	char *pos = strchr(user_spec, ':');
+	user_spec += strspn(user_spec, " \t\r\n");
+	size_t len = pos ? (size_t)(pos - user_spec) : strlen(user_spec);
+
+	char *buf = NULL;
+	struct passwd pwd, *res;
+	int e;
+
+	buf = malloc(bufsize);
+	char *user_search = NULL;
+	if (len) {
+		user_search = malloc(len + 1);
+		memcpy(user_search, user_spec, len);
+		user_search[len] = '\0';
+		while ((e = getpwnam_r(user_search, &pwd, buf, bufsize, &res)) == ERANGE) {
+			if (!enlarge_buffer(&buf, &bufsize)) {
+				break;
+			}
+		}
+
+		if (e != 0) {
+			RTE_LOG(ERR, EAL,"Failed to retrive user %s's uid (%s), aborting.",
+				user_search, strerror(e));
+			goto release;
+		}
+		if (res == NULL) {
+			RTE_LOG(ERR, EAL,"user %s not found,  aborting.",
+				user_search);
+			e = -1;
+			goto release;
+		}
+	} else {
+		/* User name is not specified, use current user.  */
+		while ((e = getpwuid_r(getuid(), &pwd, buf, bufsize, &res)) == ERANGE) {
+			if (!enlarge_buffer(&buf, &bufsize)) {
+				break;
+			}
+		}
+
+		if (e != 0) {
+			RTE_LOG(ERR, EAL,"Failed to retrive current user's uid "
+				"(%s), aborting.", strerror(e));
+			goto release;
+		}
+		user_search = strdup(pwd.pw_name);
+	}
+
+	if (uid)
+		*uid = pwd.pw_uid;
+
+	free(buf);
+	buf = NULL;
+
+	if (pos) {
+		char *grpstr = pos + 1;
+		grpstr += strspn(grpstr, " \t\r\n");
+
+		if (*grpstr) {
+			struct group grp, *res;
+
+			bufsize = 4096;
+			buf = malloc(bufsize);
+			while ((e = getgrnam_r(grpstr, &grp, buf, bufsize, &res))
+					 == ERANGE) {
+				if (!enlarge_buffer(&buf, &bufsize)) {
+					break;
+				}
+			}
+
+			if (e) {
+				RTE_LOG(ERR, EAL,"Failed to get group entry for %s, "
+					"(%s), aborting.", grpstr,
+					strerror(e));
+				goto release;
+			}
+			if (res == NULL) {
+				RTE_LOG(ERR, EAL,"Group %s not found, aborting.",
+					grpstr);
+				e = -1;
+				goto release;
+			}
+
+			if (gid)
+				*gid = grp.gr_gid;
+		}
+	}
+
+ release:
+	free(buf);
+	free(user_search);
+	return e;
+}
+
+static void
+vhost_set_permissions(const char *vhost_sock_location)
+{
+	unsigned long int mode = strtoul(internal_config.vhost_sock_perm, NULL, 0);
+	int err = chmod(vhost_sock_location, (mode_t)mode);
+	if (err) {
+		RTE_LOG(ERR, EAL,"vhost-user socket cannot set"
+			" permissions to %s (%s).\n",
+			internal_config.vhost_sock_perm, strerror(err));
+		return;
+	}
+	RTE_LOG(INFO, EAL,"Socket %s changed permissions"
+			" to %s\n", vhost_sock_location,
+			internal_config.vhost_sock_perm);
+}
+
+static void
+vhost_set_ownership(const char *vhost_sock_location)
+{
+	uid_t vhuid=0;
+	gid_t vhgid=0;
+
+	if (get_owners_from_str(internal_config.vhost_sock_owner, &vhuid, &vhgid)) {
+		RTE_LOG(ERR, EAL,"vhost-user socket unable to get"
+			" specified user/group: %s\n",
+			internal_config.vhost_sock_owner);
+		return;
+	}
+
+	int err = chown(vhost_sock_location, vhuid, vhgid);
+	if (err) {
+		RTE_LOG(ERR, EAL,"vhost-user socket unable to set"
+			" ownership to %s (%s).\n",
+			internal_config.vhost_sock_owner, strerror(err));
+		return;
+	}
+
+	RTE_LOG(INFO, EAL,"Socket %s changed ownership"
+			" to %s.\n", vhost_sock_location,
+			internal_config.vhost_sock_owner);
+}
+
+void
+rte_eal_set_socket_permissions(const char *path)
+{
+	if (internal_config.vhost_sock_perm) {
+		vhost_set_permissions(path);
+	}
+
+	if (internal_config.vhost_sock_owner) {
+		vhost_set_ownership(path);
+	}
+}
diff --git a/lib/librte_vhost/vhost_user/vhost-net-user.c b/lib/librte_vhost/vhost_user/vhost-net-user.c
index df2bd64..ea8dee9 100644
--- a/lib/librte_vhost/vhost_user/vhost-net-user.c
+++ b/lib/librte_vhost/vhost_user/vhost-net-user.c
@@ -51,6 +51,8 @@
 #include "vhost-net.h"
 #include "virtio-net-user.h"
 
+#include <rte_eal.h>
+
 #define MAX_VIRTIO_BACKLOG 128
 
 static void vserver_new_vq_conn(int fd, void *data, int *remove);
@@ -476,6 +478,8 @@ rte_vhost_driver_register(const char *path)
 		return -1;
 	}
 
+	rte_eal_set_socket_permissions(path);
+
 	vserver->path = strdup(path);
 
 	fdset_add(&g_vhost_server.fdset, vserver->listenfd,
-- 
2.7.4

^ permalink raw reply	[relevance 2%]

* Re: [dpdk-dev] [RFC] eal: provide option to set vhost_user socket owner/permissions
  2016-04-25  9:18  2% [dpdk-dev] [RFC] eal: provide option to set vhost_user socket owner/permissions Christian Ehrhardt
@ 2016-04-26  4:16  0% ` Yuanhan Liu
  2016-04-26  7:24  0%   ` Christian Ehrhardt
  0 siblings, 1 reply; 200+ results
From: Yuanhan Liu @ 2016-04-26  4:16 UTC (permalink / raw)
  To: Christian Ehrhardt
  Cc: Aaron Conole, dev, Xie, Huawei, Thomas Monjalon, David Marchand,
	Panu Matilainen

On Mon, Apr 25, 2016 at 11:18:16AM +0200, Christian Ehrhardt wrote:
> The API doesn't hold a way to specify a owner/permission set for vhost_user
> created sockets.

Yes, it's kind of like a known issue. So, thanks for bringing it, with
a solution, for dicussion (cc'ed more people).

> I don't even think an API change would make that much sense.
> 
> Projects consuming DPDK start to do 'their own workarounds' like openvswitch
> https://patchwork.ozlabs.org/patch/559043/
> https://patchwork.ozlabs.org/patch/559045/
> But for this specific example they are blocked/stalled behind a bigger
> rework (https://patchwork.ozlabs.org/patch/604898/).
> Also one could ask why each project would need their own workaround.
> 
> At the same time - as I want it for existing code linking against DPDK I
> wanted to avoid changing API/ABI. That way I want to provide something existing
> users could utilize. So I created a DPDK EAL commandline option based ideas in
> the former patches.
> 
> For myself I consider this a nice interim solution for existing released
> Openvswitch+DPDK solution. And I intend to put it as delta into the DPDK 2.2
> currently packaged in Ubuntu to get it working more smoothly with
> openvswitch 2.5.
> 
> But I'd be interested if DPDK in general would be interested in:
> a) an approach like this?

You were trying to add a vhost specific stuff as EAL command option,
which is something we might should try to avoid.

> b) would prefer a change of the API?

Adding a new option to the current register API might will not work well,
either. It gives you no ability to do a dynamic change later. I mean,
taking OVS as an example, OVS provides you the flexible ability to do all
kinds of configuration in a dynamic way, say number of rx queues. If we
do the permissions setup in the register time, there would be no way to
change it later, right?

So, I'm thinking that we may could add a new API for that? It then would
allow applications to change it at anytime.

> c) consider it an issue of consuming projects and let them take care?

It's not exactly an issue of consuming projects; we created the socket
file after all.

And I'd like to hear what others would say.

Thanks.

	--yliu

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix
@ 2016-04-26  4:45  3% Yuanhan Liu
  2016-04-26  4:45  2% ` [dpdk-dev] [PATCH 7/7] examples/vhost: switch_worker cleanup Yuanhan Liu
                   ` (2 more replies)
  0 siblings, 3 replies; 200+ results
From: Yuanhan Liu @ 2016-04-26  4:45 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Yuanhan Liu

I'm starting to work on the vhost ABI refactoring, that I also have to
touch the vhost example code, to make it work. The vhost example code,
however, is very messy, full of __very__ long lines. This would make
a later diff to apply the new vhost API be very ugly, therefore, not
friendly for review. This is how this cleanup comes.

Besides that, there is one enhancement patch, which handles the broadcast
packets so that we could rely the ARP request packet, to let vhost-switch
be more like a real switch. There is another patch that (hopefully) would
fix the mbuf allocation failure ultimately. I also added some guidelines
there as comments to show how to count how many mbuf entries is enough for
our usage.

---
Yuanhan Liu (7):
  examples/vhost: remove the non-working zero copy code
  examples/vhost: remove unused macro and struct
  examples/vhost: use tailq to link vhost devices
  examples/vhost: use mac compare helper function directly
  examples/vhost: handle broadcast packet
  examples/vhost: fix mbuf allocation failures
  examples/vhost: switch_worker cleanup

 doc/guides/sample_app_ug/vhost.rst |   36 +-
 examples/vhost/main.c              | 2319 ++++++------------------------------
 examples/vhost/main.h              |   49 +-
 3 files changed, 375 insertions(+), 2029 deletions(-)

-- 
1.9.0

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH 7/7] examples/vhost: switch_worker cleanup
  2016-04-26  4:45  3% [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix Yuanhan Liu
@ 2016-04-26  4:45  2% ` Yuanhan Liu
  2016-04-28  5:45  0% ` [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix Wang, Zhihong
       [not found]     ` <1462224230-19460-1-git-send-email-yuanhan.liu@linux.intel.com>
  2 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-04-26  4:45 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Yuanhan Liu

switch_worker() is the last piece of code that is messy yet it touches
virtio/vhost device.

Here do a cleanup, so that we will be less painful for later vhost ABI
refactoring.

The cleanup is straigforward: break long lines, move some code into
functions, the last, comment a bit on switch_worker().

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 examples/vhost/main.c | 255 +++++++++++++++++++++++++++-----------------------
 1 file changed, 138 insertions(+), 117 deletions(-)

diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index 6a69f34..96d6ab5 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -213,6 +213,8 @@ struct mbuf_table {
 /* TX queue for each data core. */
 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
 
+#define MBUF_TABLE_DRAIN_TSC	((rte_get_tsc_hz() + US_PER_S - 1) \
+				 / US_PER_S * BURST_TX_DRAIN_US)
 #define VLAN_HLEN       4
 
 /* Per-device statistics struct */
@@ -945,16 +947,35 @@ static void virtio_tx_offload(struct rte_mbuf *m)
 	tcp_hdr->cksum = get_psd_sum(l3_hdr, m->ol_flags);
 }
 
+static inline void
+free_pkts(struct rte_mbuf **pkts, uint16_t n)
+{
+	while (n--)
+		rte_pktmbuf_free(pkts[n]);
+}
+
+static inline void __attribute__((always_inline))
+do_drain_mbuf_table(struct mbuf_table *tx_q)
+{
+	uint16_t count;
+
+	count = rte_eth_tx_burst(ports[0], tx_q->txq_id,
+				 tx_q->m_table, tx_q->len);
+	if (unlikely(count < tx_q->len))
+		free_pkts(&tx_q->m_table[count], tx_q->len - count);
+
+	tx_q->len = 0;
+}
+
 /*
- * This function routes the TX packet to the correct interface. This may be a local device
- * or the physical port.
+ * This function routes the TX packet to the correct interface. This
+ * may be a local device or the physical port.
  */
 static inline void __attribute__((always_inline))
 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
 {
 	struct mbuf_table *tx_q;
-	struct rte_mbuf **m_table;
-	unsigned len, ret, offset = 0;
+	unsigned offset = 0;
 	const uint16_t lcore_id = rte_lcore_id();
 	struct virtio_net *dev = vdev->dev;
 	struct ether_hdr *nh;
@@ -990,7 +1011,6 @@ queue2nic:
 
 	/*Add packet to the port tx queue*/
 	tx_q = &lcore_tx_queue[lcore_id];
-	len = tx_q->len;
 
 	nh = rte_pktmbuf_mtod(m, struct ether_hdr *);
 	if (unlikely(nh->ether_type == rte_cpu_to_be_16(ETHER_TYPE_VLAN))) {
@@ -1028,55 +1048,132 @@ queue2nic:
 	if (m->ol_flags & PKT_TX_TCP_SEG)
 		virtio_tx_offload(m);
 
-	tx_q->m_table[len] = m;
-	len++;
+	tx_q->m_table[tx_q->len++] = m;
 	if (enable_stats) {
 		dev_statistics[dev->device_fh].tx_total++;
 		dev_statistics[dev->device_fh].tx++;
 	}
 
-	if (unlikely(len == MAX_PKT_BURST)) {
-		m_table = (struct rte_mbuf **)tx_q->m_table;
-		ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id, m_table, (uint16_t) len);
-		/* Free any buffers not handled by TX and update the port stats. */
-		if (unlikely(ret < len)) {
-			do {
-				rte_pktmbuf_free(m_table[ret]);
-			} while (++ret < len);
+	if (unlikely(tx_q->len == MAX_PKT_BURST))
+		do_drain_mbuf_table(tx_q);
+
+	return;
+}
+
+
+static inline void __attribute__((always_inline))
+drain_mbuf_table(struct mbuf_table *tx_q)
+{
+	static uint64_t prev_tsc = 0;
+	uint64_t cur_tsc;
+
+	if (tx_q->len == 0)
+		return;
+
+	cur_tsc = rte_rdtsc();
+	if (unlikely(cur_tsc - prev_tsc > MBUF_TABLE_DRAIN_TSC)) {
+		prev_tsc = cur_tsc;
+
+		RTE_LOG(DEBUG, VHOST_DATA,
+			"TX queue drained after timeout with burst size %u\n",
+			tx_q->len);
+		do_drain_mbuf_table(tx_q);
+	}
+}
+
+static inline void __attribute__((always_inline))
+drain_eth_rx(struct vhost_dev *vdev)
+{
+	uint16_t rx_count, enqueue_count;
+	struct virtio_net *dev = vdev->dev;
+	struct rte_mbuf *pkts[MAX_PKT_BURST];
+
+	rx_count = rte_eth_rx_burst(ports[0], vdev->vmdq_rx_q,
+				    pkts, MAX_PKT_BURST);
+	if (!rx_count)
+		return;
+
+	/*
+	 * When "enable_retry" is set, here we wait and retry when there
+	 * is no enough free slots in the queue to hold @rx_count packets,
+	 * to diminish packet loss.
+	 */
+	if (enable_retry &&
+	    unlikely(rx_count > rte_vring_available_entries(dev,
+			VIRTIO_RXQ))) {
+		uint32_t retry;
+
+		for (retry = 0; retry < burst_rx_retry_num; retry++) {
+			rte_delay_us(burst_rx_delay_time);
+			if (rx_count <= rte_vring_available_entries(dev,
+					VIRTIO_RXQ))
+				break;
 		}
+	}
 
-		len = 0;
+	enqueue_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ,
+						pkts, rx_count);
+	if (enable_stats) {
+		uint64_t fh = dev->device_fh;
+
+		rte_atomic64_add(&dev_statistics[fh].rx_total_atomic, rx_count);
+		rte_atomic64_add(&dev_statistics[fh].rx_atomic, enqueue_count);
 	}
 
-	tx_q->len = len;
-	return;
+	free_pkts(pkts, rx_count);
 }
+
+static inline void __attribute__((always_inline))
+drain_virtio_tx(struct vhost_dev *vdev)
+{
+	struct rte_mbuf *pkts[MAX_PKT_BURST];
+	uint16_t count;
+	uint16_t i;
+
+	count = rte_vhost_dequeue_burst(vdev->dev, VIRTIO_TXQ, mbuf_pool,
+					pkts, MAX_PKT_BURST);
+
+	/* setup VMDq for the first packet */
+	if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && count) {
+		if (vdev->remove || link_vmdq(vdev, pkts[0]) == -1)
+			free_pkts(pkts, count);
+	}
+
+	for (i = 0; i < count; ++i) {
+		virtio_tx_route(vdev, pkts[i],
+			vlan_tags[(uint16_t)vdev->dev->device_fh]);
+	}
+}
+
 /*
- * This function is called by each data core. It handles all RX/TX registered with the
- * core. For TX the specific lcore linked list is used. For RX, MAC addresses are compared
- * with all devices in the main linked list.
+ * Main function of vhost-switch. It basically does:
+ *
+ * for each vhost device {
+ *    - drain_eth_rx()
+ *
+ *      Which drains the host eth Rx queue linked to the vhost device,
+ *      and deliver all of them to guest virito Rx ring associated with
+ *      this vhost device.
+ *
+ *    - drain_virtio_tx()
+ *
+ *      Which drains the guest virtio Tx queue and deliver all of them
+ *      to the target, which could be another vhost device, or the
+ *      physical eth dev. The route is done in function "virtio_tx_route".
+ * }
  */
 static int
-switch_worker(__attribute__((unused)) void *arg)
+switch_worker(void *arg __rte_unused)
 {
-	struct virtio_net *dev = NULL;
-	struct vhost_dev *vdev = NULL;
-	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+	unsigned i;
+	unsigned lcore_id = rte_lcore_id();
+	struct vhost_dev *vdev;
 	struct mbuf_table *tx_q;
-	const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
-	uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
-	unsigned ret, i;
-	const uint16_t lcore_id = rte_lcore_id();
-	const uint16_t num_cores = (uint16_t)rte_lcore_count();
-	uint16_t rx_count = 0;
-	uint16_t tx_count;
-	uint32_t retry = 0;
 
 	RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
-	prev_tsc = 0;
 
 	tx_q = &lcore_tx_queue[lcore_id];
-	for (i = 0; i < num_cores; i ++) {
+	for (i = 0; i < rte_lcore_count(); i ++) {
 		if (lcore_ids[i] == lcore_id) {
 			tx_q->txq_id = i;
 			break;
@@ -1084,34 +1181,7 @@ switch_worker(__attribute__((unused)) void *arg)
 	}
 
 	while(1) {
-		cur_tsc = rte_rdtsc();
-		/*
-		 * TX burst queue drain
-		 */
-		diff_tsc = cur_tsc - prev_tsc;
-		if (unlikely(diff_tsc > drain_tsc)) {
-
-			if (tx_q->len) {
-				RTE_LOG(DEBUG, VHOST_DATA,
-					"TX queue drained after timeout with burst size %u\n",
-					tx_q->len);
-
-				/*Tx any packets in the queue*/
-				ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id,
-									   (struct rte_mbuf **)tx_q->m_table,
-									   (uint16_t)tx_q->len);
-				if (unlikely(ret < tx_q->len)) {
-					do {
-						rte_pktmbuf_free(tx_q->m_table[ret]);
-					} while (++ret < tx_q->len);
-				}
-
-				tx_q->len = 0;
-			}
-
-			prev_tsc = cur_tsc;
-
-		}
+		drain_mbuf_table(tx_q);
 
 		/*
 		 * Inform the configuration core that we have exited the
@@ -1121,69 +1191,20 @@ switch_worker(__attribute__((unused)) void *arg)
 			lcore_info[lcore_id].dev_removal_flag = ACK_DEV_REMOVAL;
 
 		/*
-		 * Process devices
+		 * Process vhost devices
 		 */
 		TAILQ_FOREACH(vdev, &lcore_info[lcore_id].vdev_list, next) {
-			uint64_t fh;
-
-			dev = vdev->dev;
-			fh  = dev->device_fh;
-
 			if (unlikely(vdev->remove)) {
 				unlink_vmdq(vdev);
 				vdev->ready = DEVICE_SAFE_REMOVE;
 				continue;
 			}
 
-			if (likely(vdev->ready == DEVICE_RX)) {
-				/*Handle guest RX*/
-				rx_count = rte_eth_rx_burst(ports[0],
-					vdev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
-
-				if (rx_count) {
-					/*
-					* Retry is enabled and the queue is full then we wait and retry to avoid packet loss
-					* Here MAX_PKT_BURST must be less than virtio queue size
-					*/
-					if (enable_retry && unlikely(rx_count > rte_vring_available_entries(dev, VIRTIO_RXQ))) {
-						for (retry = 0; retry < burst_rx_retry_num; retry++) {
-							rte_delay_us(burst_rx_delay_time);
-							if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
-								break;
-						}
-					}
-					ret_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ, pkts_burst, rx_count);
-					if (enable_stats) {
-						rte_atomic64_add(
-							&dev_statistics[fh].rx_total_atomic,
-							rx_count);
-						rte_atomic64_add(
-							&dev_statistics[fh].rx_atomic,
-							ret_count);
-					}
-					while (likely(rx_count)) {
-						rx_count--;
-						rte_pktmbuf_free(pkts_burst[rx_count]);
-					}
-
-				}
-			}
+			if (likely(vdev->ready == DEVICE_RX))
+				drain_eth_rx(vdev);
 
-			if (likely(!vdev->remove)) {
-				/* Handle guest TX*/
-				tx_count = rte_vhost_dequeue_burst(dev, VIRTIO_TXQ, mbuf_pool, pkts_burst, MAX_PKT_BURST);
-				/* If this is the first received packet we need to learn the MAC and setup VMDQ */
-				if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
-					if (vdev->remove || (link_vmdq(vdev, pkts_burst[0]) == -1)) {
-						while (tx_count)
-							rte_pktmbuf_free(pkts_burst[--tx_count]);
-					}
-				}
-				for (i = 0; i < tx_count; ++i) {
-					virtio_tx_route(vdev, pkts_burst[i],
-						vlan_tags[(uint16_t)dev->device_fh]);
-				}
-			}
+			if (likely(!vdev->remove))
+				drain_virtio_tx(vdev);
 		}
 	}
 
-- 
1.9.0

^ permalink raw reply	[relevance 2%]

* Re: [dpdk-dev] [RFC] eal: provide option to set vhost_user socket owner/permissions
  2016-04-26  4:16  0% ` Yuanhan Liu
@ 2016-04-26  7:24  0%   ` Christian Ehrhardt
  0 siblings, 0 replies; 200+ results
From: Christian Ehrhardt @ 2016-04-26  7:24 UTC (permalink / raw)
  To: Yuanhan Liu
  Cc: Aaron Conole, dev, Xie, Huawei, Thomas Monjalon, David Marchand,
	Panu Matilainen

Thanks,
great that you added more on CC for a wider discussion - I think that is
the only right way to go.

Just to "defend" a bit - solution a) was created under the special
circumstance that I wanted a workaround that would work today.
But that is/was special to what I package with DPDK 2.2 + OVS 2.5 as of
today - and therefore was the right place for a fast interim fix for me.
I totally agree that the A in EAL was meant for abstraction and we might
want to avoid vhost specific things in there that in the long run.

I like your suggestion of a new API as a proper long term solution, but I
don't feel deeply enough involved yet on the API level to give it any
judgement.
So I look forward for more opinions on it.

P.S. the patch bot hit me hard with 2 pages of space/bracket issues, sorry
for that - but it was only meant as RFC after all :-)


Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd

On Tue, Apr 26, 2016 at 6:16 AM, Yuanhan Liu <yuanhan.liu@linux.intel.com>
wrote:

> On Mon, Apr 25, 2016 at 11:18:16AM +0200, Christian Ehrhardt wrote:
> > The API doesn't hold a way to specify a owner/permission set for
> vhost_user
> > created sockets.
>
> Yes, it's kind of like a known issue. So, thanks for bringing it, with
> a solution, for dicussion (cc'ed more people).
>
> > I don't even think an API change would make that much sense.
> >
> > Projects consuming DPDK start to do 'their own workarounds' like
> openvswitch
> > https://patchwork.ozlabs.org/patch/559043/
> > https://patchwork.ozlabs.org/patch/559045/
> > But for this specific example they are blocked/stalled behind a bigger
> > rework (https://patchwork.ozlabs.org/patch/604898/).
> > Also one could ask why each project would need their own workaround.
> >
> > At the same time - as I want it for existing code linking against DPDK I
> > wanted to avoid changing API/ABI. That way I want to provide something
> existing
> > users could utilize. So I created a DPDK EAL commandline option based
> ideas in
> > the former patches.
> >
> > For myself I consider this a nice interim solution for existing released
> > Openvswitch+DPDK solution. And I intend to put it as delta into the DPDK
> 2.2
> > currently packaged in Ubuntu to get it working more smoothly with
> > openvswitch 2.5.
> >
> > But I'd be interested if DPDK in general would be interested in:
> > a) an approach like this?
>
> You were trying to add a vhost specific stuff as EAL command option,
> which is something we might should try to avoid.
>
> > b) would prefer a change of the API?
>
> Adding a new option to the current register API might will not work well,
> either. It gives you no ability to do a dynamic change later. I mean,
> taking OVS as an example, OVS provides you the flexible ability to do all
> kinds of configuration in a dynamic way, say number of rx queues. If we
> do the permissions setup in the register time, there would be no way to
> change it later, right?
>
> So, I'm thinking that we may could add a new API for that? It then would
> allow applications to change it at anytime.
>
> > c) consider it an issue of consuming projects and let them take care?
>
> It's not exactly an issue of consuming projects; we created the socket
> file after all.
>
> And I'd like to hear what others would say.
>
> Thanks.
>
>         --yliu
>

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix
  2016-04-26  4:45  3% [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix Yuanhan Liu
  2016-04-26  4:45  2% ` [dpdk-dev] [PATCH 7/7] examples/vhost: switch_worker cleanup Yuanhan Liu
@ 2016-04-28  5:45  0% ` Wang, Zhihong
  2016-04-28  6:09  0%   ` Yuanhan Liu
       [not found]     ` <1462224230-19460-1-git-send-email-yuanhan.liu@linux.intel.com>
  2 siblings, 1 reply; 200+ results
From: Wang, Zhihong @ 2016-04-28  5:45 UTC (permalink / raw)
  To: Yuanhan Liu, dev; +Cc: Xie, Huawei


> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Yuanhan Liu
> Sent: Tuesday, April 26, 2016 12:46 PM
> To: dev@dpdk.org
> Cc: Xie, Huawei <huawei.xie@intel.com>; Yuanhan Liu
> <yuanhan.liu@linux.intel.com>
> Subject: [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix
> 
> I'm starting to work on the vhost ABI refactoring, that I also have to
> touch the vhost example code, to make it work. The vhost example code,
> however, is very messy, full of __very__ long lines. This would make
> a later diff to apply the new vhost API be very ugly, therefore, not
> friendly for review. This is how this cleanup comes.


I think this patch is great effort to clean the messy code and make clearer
logic, only one suggestion: do you think a complete cleanup would help more?
in terms of code style and function organization. Since there'll be further work
on it, and it's a small file anyway. Currently some parts still seem messy to me,
which compromises the effort of this patch.


> 
> Besides that, there is one enhancement patch, which handles the broadcast
> packets so that we could rely the ARP request packet, to let vhost-switch
> be more like a real switch. There is another patch that (hopefully) would
> fix the mbuf allocation failure ultimately. I also added some guidelines
> there as comments to show how to count how many mbuf entries is enough for
> our usage.
> 
> ---
> Yuanhan Liu (7):
>   examples/vhost: remove the non-working zero copy code
>   examples/vhost: remove unused macro and struct
>   examples/vhost: use tailq to link vhost devices
>   examples/vhost: use mac compare helper function directly
>   examples/vhost: handle broadcast packet
>   examples/vhost: fix mbuf allocation failures
>   examples/vhost: switch_worker cleanup
> 
>  doc/guides/sample_app_ug/vhost.rst |   36 +-
>  examples/vhost/main.c              | 2319 ++++++------------------------------
>  examples/vhost/main.h              |   49 +-
>  3 files changed, 375 insertions(+), 2029 deletions(-)
> 
> --
> 1.9.0

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix
  2016-04-28  5:45  0% ` [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix Wang, Zhihong
@ 2016-04-28  6:09  0%   ` Yuanhan Liu
  0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-04-28  6:09 UTC (permalink / raw)
  To: Wang, Zhihong; +Cc: dev, Xie, Huawei

On Thu, Apr 28, 2016 at 05:45:16AM +0000, Wang, Zhihong wrote:
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Yuanhan Liu
> > Sent: Tuesday, April 26, 2016 12:46 PM
> > To: dev@dpdk.org
> > Cc: Xie, Huawei <huawei.xie@intel.com>; Yuanhan Liu
> > <yuanhan.liu@linux.intel.com>
> > Subject: [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix
> > 
> > I'm starting to work on the vhost ABI refactoring, that I also have to
> > touch the vhost example code, to make it work. The vhost example code,
> > however, is very messy, full of __very__ long lines. This would make
> > a later diff to apply the new vhost API be very ugly, therefore, not
> > friendly for review. This is how this cleanup comes.
> 
> 
> I think this patch is great effort to clean the messy code and make clearer
> logic, only one suggestion: do you think a complete cleanup would help more?

Yes, but I will stop here, and maybe do the left in near future, as I
have more important thing to do now. I even thought about to make the
VMDq and VLAN stuff optional, to not let our example connect with those
hardware feature that tight.

So, feel free to make patches to clean it further if you have time.

	--yliu

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats
  2016-04-20 16:03  0% ` David Harton (dharton)
@ 2016-04-28 14:56  0%   ` Tahhan, Maryam
  2016-04-28 15:58  0%     ` David Harton (dharton)
  0 siblings, 1 reply; 200+ results
From: Tahhan, Maryam @ 2016-04-28 14:56 UTC (permalink / raw)
  To: David Harton (dharton), Horton, Remy, dev
  Cc: Mcnamara, John, Van Haaren, Harry

> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of David Harton
> (dharton)
> Sent: Wednesday, April 20, 2016 5:04 PM
> To: Horton, Remy <remy.horton@intel.com>; dev@dpdk.org
> Subject: Re: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations
> from xstats
> 
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Remy Horton
> > Sent: Friday, April 15, 2016 10:44 AM
> > To: dev@dpdk.org
> > Subject: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from
> > xstats
> >
> > The current extended ethernet statistics fetching involve doing
> > several string operations, which causes performance issues if there
> > are lots of statistics and/or network interfaces. This RFC patchset
> > changes the API for xstats to use integer identifiers instead of
> > strings and implements this new API for the ixgbe driver. Others
> drivers to follow.
> >
> > --
> >
> > Since this will involve API & ABI breakage as previously advertised,
> > there are several design assumptions that need consideration:
> >
> > *) id-name & id-value pairs for both lookup and query Permits
> > out-of-order and non-contigious returning of names/ids/values, even
> > though expected implmentations would in practice return items in
> > sorted order by id. Is this sufficent/desirable future proofing? Idea
> > is to allow possibility of drivers returning partial statistics.
> 
> I believe forcing drivers to match to a common id-space will become
> burdensome.  If the stats id-space isn't common then matching strings is
> probably just as sufficient as long as drivers don't add/remove stats ad
> hoc between the time the device is initialized and removed.

I'm not aware of drivers adding/removing the stats ad hoc? The idea is to have a common-id space otherwise it will be a free for all and we won't have alignment across the drivers. I don't see it being any more burdensome than having a common register naming across the board which is what is there today. The advantage being that you don't have to pull the strings every time.

> 
> >
> > *) Bulk name-id mapping lookup only
> > At the moment individual lookup is not supported, as this would impose
> > extra overheads on drivers. The assumption is that any end user would
> > fetch all this data once on startup and then cache the mappings.
> 
> I'm not sure I see the value of looking up a single stat from a user
> perspective.  I can see where the drivers might say that some stats are
> less disruptive/etc but the user doesn't have that knowledge and
> wouldn't know how to take advantage.  Usually all stats are grabbed
> multiple times and the changes noted during debug sessions.
> 

I believe Remy's change doesn't suggest/support individual lookup. It is just a statement that we don't want to burden drivers with individual stats lookups.

> >
> > *) Replacement or additional API
> > This patch replaces the current xstats API, but there is no inherant
> > reason beyond maintainability why this funtionality could not be in
> > addition rather than a replacement. What is consensus on this?
> 
> I came to the conclusion that replacing the existing API isn't necessary
> but rather extending it so backwards compatibility could be maintained
> during the previous discussions on this topic.  However, if we want to go
> forward with cleaning up in order to reduce the support drivers provide
> I'm all for it.
> 
> I still believe the API we develop should follow an "ethtool stats like"
> format as suggested earlier this year:
> 
> extern int rte_eth_xstats_names_get(uint8_t port_id,
>         struct rte_eth_xstats_name *names, unsigned n); extern int
> rte_eth_xstats_values_get(uint8_t port_id,
>         uint64_t *values, unsigned n);
> 
> Again, these could be provided alongside the existing API or replace it.

I'm struggling a bit here. This is really what Remy has posted http://dpdk.org/dev/patchwork/patch/12094/ or am I missing something obvious?

> 
> I also like the idea you provided of a separate API to obtain the xstats
> count rather than deriving the count by calling one of the above
> functions with "dummy" values.

+1 

> 
> Again, I can provide the patches for the changes I've made that align
> with this proposed API.  I just never got any feedback on it when
> requested previously.

I believe time is not in our favour on this front. If you have patches can you post them, otherwise can you please review the patchset that Remy has posted?
Thanks in advance.

BR
Maryam

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats
  2016-04-28 14:56  0%   ` Tahhan, Maryam
@ 2016-04-28 15:58  0%     ` David Harton (dharton)
  0 siblings, 0 replies; 200+ results
From: David Harton (dharton) @ 2016-04-28 15:58 UTC (permalink / raw)
  To: Tahhan, Maryam, Horton, Remy, dev; +Cc: Mcnamara, John, Van Haaren, Harry

> -----Original Message-----
> From: Tahhan, Maryam [mailto:maryam.tahhan@intel.com]
> Sent: Thursday, April 28, 2016 10:56 AM
> To: David Harton (dharton) <dharton@cisco.com>; Horton, Remy
> <remy.horton@intel.com>; dev@dpdk.org
> Cc: Mcnamara, John <john.mcnamara@intel.com>; Van Haaren, Harry
> <harry.van.haaren@intel.com>
> Subject: RE: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from
> xstats
> 
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of David Harton
> > (dharton)
> > Sent: Wednesday, April 20, 2016 5:04 PM
> > To: Horton, Remy <remy.horton@intel.com>; dev@dpdk.org
> > Subject: Re: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations
> > from xstats
> >
> > > -----Original Message-----
> > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Remy Horton
> > > Sent: Friday, April 15, 2016 10:44 AM
> > > To: dev@dpdk.org
> > > Subject: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from
> > > xstats
> > >
> > > The current extended ethernet statistics fetching involve doing
> > > several string operations, which causes performance issues if there
> > > are lots of statistics and/or network interfaces. This RFC patchset
> > > changes the API for xstats to use integer identifiers instead of
> > > strings and implements this new API for the ixgbe driver. Others
> > drivers to follow.
> > >
> > > --
> > >
> > > Since this will involve API & ABI breakage as previously advertised,
> > > there are several design assumptions that need consideration:
> > >
> > > *) id-name & id-value pairs for both lookup and query Permits
> > > out-of-order and non-contigious returning of names/ids/values, even
> > > though expected implmentations would in practice return items in
> > > sorted order by id. Is this sufficent/desirable future proofing?
> > > Idea is to allow possibility of drivers returning partial statistics.
> >
> > I believe forcing drivers to match to a common id-space will become
> > burdensome.  If the stats id-space isn't common then matching strings
> > is probably just as sufficient as long as drivers don't add/remove
> > stats ad hoc between the time the device is initialized and removed.
> 
> I'm not aware of drivers adding/removing the stats ad hoc? The idea is to
> have a common-id space otherwise it will be a free for all and we won't
> have alignment across the drivers. I don't see it being any more
> burdensome than having a common register naming across the board which is
> what is there today. The advantage being that you don't have to pull the
> strings every time.
> 
> >
> > >
> > > *) Bulk name-id mapping lookup only
> > > At the moment individual lookup is not supported, as this would
> > > impose extra overheads on drivers. The assumption is that any end
> > > user would fetch all this data once on startup and then cache the
> mappings.
> >
> > I'm not sure I see the value of looking up a single stat from a user
> > perspective.  I can see where the drivers might say that some stats
> > are less disruptive/etc but the user doesn't have that knowledge and
> > wouldn't know how to take advantage.  Usually all stats are grabbed
> > multiple times and the changes noted during debug sessions.
> >
> 
> I believe Remy's change doesn't suggest/support individual lookup. It is
> just a statement that we don't want to burden drivers with individual
> stats lookups.
> 
> > >
> > > *) Replacement or additional API
> > > This patch replaces the current xstats API, but there is no inherant
> > > reason beyond maintainability why this funtionality could not be in
> > > addition rather than a replacement. What is consensus on this?
> >
> > I came to the conclusion that replacing the existing API isn't
> > necessary but rather extending it so backwards compatibility could be
> > maintained during the previous discussions on this topic.  However, if
> > we want to go forward with cleaning up in order to reduce the support
> > drivers provide I'm all for it.
> >
> > I still believe the API we develop should follow an "ethtool stats like"
> > format as suggested earlier this year:
> >
> > extern int rte_eth_xstats_names_get(uint8_t port_id,
> >         struct rte_eth_xstats_name *names, unsigned n); extern int
> > rte_eth_xstats_values_get(uint8_t port_id,
> >         uint64_t *values, unsigned n);
> >
> > Again, these could be provided alongside the existing API or replace it.
> 
> I'm struggling a bit here. This is really what Remy has posted
> http://dpdk.org/dev/patchwork/patch/12094/ or am I missing something
> obvious?

Maybe I misread the patch series or missed one but I don't see where 
stats can be obtained without copying strings?  This is the real issue I 
raised originally.  

Having the ability to get the names without stats is useful, but,
the real gain would be obtaining the stats without the names.

> 
> >
> > I also like the idea you provided of a separate API to obtain the
> > xstats count rather than deriving the count by calling one of the
> > above functions with "dummy" values.
> 
> +1
> 
> >
> > Again, I can provide the patches for the changes I've made that align
> > with this proposed API.  I just never got any feedback on it when
> > requested previously.
> 
> I believe time is not in our favour on this front. If you have patches can
> you post them, otherwise can you please review the patchset that Remy has
> posted?

I'm working on it but I have some process I'm navigating.  I'm hopeful 
I'll have the green light within a week if not sooner.  I apologize...
I'm pushing as hard as I can.  If you need to proceed go ahead I 
completely understand.  

All I can say is that I have implemented the API above and converted all 
drivers that supported xstats in v2.2.  Any new drivers that have added 
xstats support since would need to be added.

I did not add "get the count" because it wasn't provided in the current API
and instead followed the convention but I do believe overtly getting the
count it is the better approach.

Thanks,
Dave

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats
  2016-04-15 14:44  3% [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats Remy Horton
  2016-04-20 16:03  0% ` David Harton (dharton)
@ 2016-04-29 12:52  0% ` David Harton (dharton)
  1 sibling, 0 replies; 200+ results
From: David Harton (dharton) @ 2016-04-29 12:52 UTC (permalink / raw)
  To: Remy Horton, dev


> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Remy Horton
> Sent: Friday, April 15, 2016 10:44 AM
> To: dev@dpdk.org
> Subject: [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from
> xstats
> 
> The current extended ethernet statistics fetching involve doing several
> string operations, which causes performance issues if there are lots of
> statistics and/or network interfaces. This RFC patchset changes the API
> for xstats to use integer identifiers instead of strings and implements
> this new API for the ixgbe driver. Others drivers to follow.
> 
> --
> 
> Since this will involve API & ABI breakage as previously advertised,
> there are several design assumptions that need consideration:
> 
> *) id-name & id-value pairs for both lookup and query
> Permits out-of-order and non-contigious returning of names/ids/values,
> even though expected implmentations would in practice return items in
> sorted order by id. Is this sufficent/desirable future proofing? Idea
> is to allow possibility of drivers returning partial statistics.

I think the key is that the order of the stats must always be honored 
and if that's the case then an id isn't necessary.  However, if others 
want an id certainly doesn't hurt.

I don't see drivers autonomously providing a subset of stats and users 
can filter out stats they don't want to their presentation layers.

> 
> *) Bulk name-id mapping lookup only
> At the moment individual lookup is not supported, as this would impose
> extra overheads on drivers. The assumption is that any end user would
> fetch all this data once on startup and then cache the mappings.

Agreed.  Similarly there is no need to return a partial list of stats 
as the presentation layers can filter.

> 
> *) Replacement or additional API
> This patch replaces the current xstats API, but there is no inherant
> reason beyond maintainability why this funtionality could not be in
> addition rather than a replacement. What is consensus on this?

I suggest 3 new functions are added:
- get number of xstats
- get xstats names
- get xstats values

This facilitates:
- parallel development within the release without breaking current usage
- possibility of removing rte_eth_xstats_get() in following release

Thanks for moving this forward,
Dave

> 
> Comments welcome.
> 
> Remy Horton (3):
>   rte: change xstats to use integer keys
>   drivers/net/ixgbe: change xstats to use integer keys
>   examples/ethtool: add xstats display command
> 
>  drivers/net/ixgbe/ixgbe_ethdev.c      | 87
> +++++++++++++++++++++++++++++++----
>  examples/ethtool/ethtool-app/ethapp.c | 57 +++++++++++++++++++++++
>  lib/librte_ether/rte_ethdev.c         | 87
> +++++++++++++++++++++++++++++++----
>  lib/librte_ether/rte_ethdev.h         | 38 +++++++++++++++
>  4 files changed, 252 insertions(+), 17 deletions(-)
> 
> --
> 2.5.5

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH] mk: do not enforce any specific ARM ABI
  2016-04-15 22:33 15% [dpdk-dev] [PATCH] mk: do not enforce any specific ARM ABI Jan Viktorin
@ 2016-05-02 15:47  4% ` Thomas Monjalon
  0 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-05-02 15:47 UTC (permalink / raw)
  To: Jan Viktorin; +Cc: dev

2016-04-16 00:33, Jan Viktorin:
> The dpdk build system passes -mfloat-abi=softfp, which makes the build fail
> when the selected ABI is EABIhf. The dpdk build system should not make
> assumptions on the selected ARM ABI.
> 
> Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
> Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Applied, thanks

^ permalink raw reply	[relevance 4%]

* Re: [dpdk-dev] ovs crash when running traffic from VM to VM over DPDK and vhostuser
  @ 2016-05-02 17:40  3% ` Yi Ba
  0 siblings, 0 replies; 200+ results
From: Yi Ba @ 2016-05-02 17:40 UTC (permalink / raw)
  To: Yuanhan Liu; +Cc: dev

Running with dpdk 16.04 and latest ovs from git, and removing "mrg_rxbuf=off" from virtio params, the crash is no longer observed. However, we are wittnessing ovs gets stuck, and will post to ovs mailing list:2016-05-02T17:26:18.804Z|00111|ovs_rcu|WARN|blocked 1000 ms waiting for pmd145 to quiesce
2016-05-02T17:26:19.805Z|00112|ovs_rcu|WARN|blocked 2001 ms waiting for pmd145 to quiesce
2016-05-02T17:26:21.804Z|00113|ovs_rcu|WARN|blocked 4000 ms waiting for pmd145 to quiesce
2016-05-02T17:26:25.805Z|00114|ovs_rcu|WARN|blocked 8001 ms waiting for pmd145 to quiesce
2016-05-02T17:26:33.805Z|00115|ovs_rcu|WARN|blocked 16001 ms waiting for pmd145 to quiesce
2016-05-02T17:26:49.805Z|00116|ovs_rcu|WARN|blocked 32001 ms waiting for pmd145 to quiesce
2016-05-02T17:27:14.354Z|00072|ovs_rcu(vhost_thread2)|WARN|blocked 128000 ms waiting for pmd145 to quiesce
2016-05-02T17:27:15.841Z|00008|ovs_rcu(urcu3)|WARN|blocked 128001 ms waiting for pmd145 to quiesce
2016-05-02T17:27:21.805Z|00117|ovs_rcu|WARN|blocked 64000 ms waiting for pmd145 to quiesce
2016-05-02T17:28:25.804Z|00118|ovs_rcu|WARN|blocked 128000 ms waiting for pmd145 to quiesce
 

    On Wednesday, 6 April 2016 10:56 AM, Yuanhan Liu <yuanhan.liu@linux.intel.com> wrote:
 

 On Tue, Apr 05, 2016 at 08:36:19PM +0000, Yi Ba wrote:
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7ff1ddffb700 (LWP 21287)]
> 0x0000000000450da7 in update_secure_len (vec_idx=0x7ff1ddff27f8, secure_len=0x7ff1ddff27fc, id=13948, vq=0x7fe7992c8940)
>    at /home/stack/ovs-dpdk/dpdk-2.2.0/lib/librte_vhost/vhost_rxtx.c:452
> 452    /home/stack/ovs-dpdk/dpdk-2.2.0/lib/librte_vhost/vhost_rxtx.c: No such file or directory.
> (gdb) bt
> #0  0x0000000000450da7 in update_secure_len (vec_idx=0x7ff1ddff27f8, secure_len=0x7ff1ddff27fc, id=13948, vq=0x7fe7992c8940)

It looks like a known issue, which has been fixed in this release. So,
could you please just try again with the latest DPDK code? It should
be able to solve your issue.

    --yliu



From yuanhan.liu@linux.intel.com  Mon May  2 23:20:28 2016
Return-Path: <yuanhan.liu@linux.intel.com>
Received: from mga04.intel.com (mga04.intel.com [192.55.52.120])
 by dpdk.org (Postfix) with ESMTP id 1A22639EA
 for <dev@dpdk.org>; Mon,  2 May 2016 23:20:27 +0200 (CEST)
Received: from fmsmga004.fm.intel.com ([10.253.24.48])
 by fmsmga104.fm.intel.com with ESMTP; 02 May 2016 14:20:28 -0700
X-ExtLoop1: 1
X-IronPort-AV: E=Sophos;i="5.24,569,1455004800"; d="scan'208";a="95880071"
Received: from yliu-dev.sh.intel.com ([10.239.67.162])
 by fmsmga004.fm.intel.com with ESMTP; 02 May 2016 14:20:26 -0700
From: Yuanhan Liu <yuanhan.liu@linux.intel.com>
To: dev@dpdk.org
Cc: huawei.xie@intel.com,
	Yuanhan Liu <yuanhan.liu@linux.intel.com>
Date: Mon,  2 May 2016 14:23:42 -0700
Message-Id: <1462224230-19460-1-git-send-email-yuanhan.liu@linux.intel.com>
X-Mailer: git-send-email 1.9.0
In-Reply-To: <1461645951-14603-1-git-send-email-yuanhan.liu@linux.intel.com>
References: <1461645951-14603-1-git-send-email-yuanhan.liu@linux.intel.com>
Subject: [dpdk-dev] [PATCH v2 0/8] vhost/example cleanup/fix
X-BeenThere: dev@dpdk.org
X-Mailman-Version: 2.1.15
Precedence: list
List-Id: patches and discussions about DPDK <dev.dpdk.org>
List-Unsubscribe: <http://dpdk.org/ml/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://dpdk.org/ml/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <http://dpdk.org/ml/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
X-List-Received-Date: Mon, 02 May 2016 21:20:28 -0000

I'm starting to work on the vhost ABI refactoring, that I also have to
touch the vhost example code. The vhost example code, however, is very
messy, full of __very__ long lines. This would make a later diff to
apply the new vhost API be very ugly, therefore, not friendly for review.
This is how this cleanup comes.

Besides that, there is one enhancement patch, which handles the broadcast
packets so that we could rely the ARP request packet, to let vhost-switch
be more like a real switch. There is another patch that (hopefully) would
fix the mbuf allocation failure ultimately. I also added some guidelines
there as comments to show how to count how many mbuf entries is enough for
our usage.

In another word, an example is meant to be clean/simple and with good
coding style so that people can get the usage easily. So, one way or
another, this patch is good to have, even without this ABI refactoring
stuff.

Note that I'm going to apply it before the end of this week, if no objections.


v2: - some checkpatch fixes

    - cleaned the code about device statistics

---
Yuanhan Liu (8):
  examples/vhost: remove the non-working zero copy code
  examples/vhost: remove unused macro and struct
  examples/vhost: use tailq to link vhost devices
  examples/vhost: use mac compare helper function directly
  examples/vhost: handle broadcast packet
  examples/vhost: fix mbuf allocation failure
  examples/vhost: switch_worker cleanup
  examples/vhost: embed statistics into vhost_dev struct

 doc/guides/sample_app_ug/vhost.rst |   36 +-
 examples/vhost/main.c              | 2394 ++++++------------------------------
 examples/vhost/main.h              |   56 +-
 3 files changed, 391 insertions(+), 2095 deletions(-)

--
1.9.3

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] perfomance of rte_lpm rule subsystem
  2016-04-19 15:46  3% ` Stephen Hemminger
  2016-04-19 20:46  0%   ` Vladimir Medvedkin
@ 2016-05-02 19:38  0%   ` Александр Киселев
  1 sibling, 0 replies; 200+ results
From: Александр Киселев @ 2016-05-02 19:38 UTC (permalink / raw)
  To: Stephen Hemminger; +Cc: dev

Stephen, what was the main reason you use red-black tree instead of dir-24-8?
Did you switch to using trees because of too big memory working set of
dir-24-8 algorithm?


2016-04-19 18:46 GMT+03:00 Stephen Hemminger <stephen@networkplumber.org>:

> On Tue, 19 Apr 2016 14:11:11 +0300
> Александр Киселев <kiselev99@gmail.com> wrote:
>
> > Hi.
> >
> > Doing some test with rte_lpm (adding/deleting bgp full table rules) I
> > noticed that
> > rule subsystem is very slow even considering that probably it was never
> > designed for using
> > in a data forwarding plane. So I want to propose some changes to the
> "rule"
> > subsystem.
> >
> > I reimplemented rule part ot the lib using rte_hash, and perfomance of
> > adding/deleted routes have increased dramatically.
> > If increasing speed of adding deleting routes makes sence for anybody
> else
> > I would like to discuss my patch.
> > The patch also include changes that make next_hop 64 bit, so please just
> > ignore them. The rule changes are in the following
> > functions only:
> >
> > rte_lpm2_create
> >
> > rule_find
> > rule_add
> > rule_delete
> > find_previous_rule
> > delete_depth_small
> > delete_depth_big
> >
> > rte_lpm2_add
> > rte_lpm2_delete
> > rte_lpm2_is_rule_present
> > rte_lpm2_delete_all
> >
>
> We forked LPM back several versions ago.
> I sent the patches to use BSD red-black tree for rules but the patches were
> ignored. mostly because it broke ABI.
>



-- 
--
Kiselev Alexander

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v2 7/8] examples/vhost: switch_worker cleanup
       [not found]     ` <1462224230-19460-1-git-send-email-yuanhan.liu@linux.intel.com>
@ 2016-05-02 21:23  2%   ` Yuanhan Liu
  2016-05-09 18:06  0%   ` [dpdk-dev] [PATCH v2 0/8] vhost/example cleanup/fix Yuanhan Liu
  1 sibling, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-02 21:23 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Yuanhan Liu

switch_worker() is the last piece of code that is messy yet it touches
virtio/vhost device.

Here do a cleanup, so that we will be less painful for later vhost ABI
refactoring.

The cleanup is straigforward: break long lines, move some code into
functions. The last, comment a bit on switch_worker().

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 examples/vhost/main.c | 253 +++++++++++++++++++++++++++-----------------------
 1 file changed, 136 insertions(+), 117 deletions(-)

diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index dbb42ee..66d3bf2 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -213,6 +213,8 @@ struct mbuf_table {
 /* TX queue for each data core. */
 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
 
+#define MBUF_TABLE_DRAIN_TSC	((rte_get_tsc_hz() + US_PER_S - 1) \
+				 / US_PER_S * BURST_TX_DRAIN_US)
 #define VLAN_HLEN       4
 
 /* Per-device statistics struct */
@@ -915,16 +917,35 @@ static void virtio_tx_offload(struct rte_mbuf *m)
 	tcp_hdr->cksum = get_psd_sum(l3_hdr, m->ol_flags);
 }
 
+static inline void
+free_pkts(struct rte_mbuf **pkts, uint16_t n)
+{
+	while (n--)
+		rte_pktmbuf_free(pkts[n]);
+}
+
+static inline void __attribute__((always_inline))
+do_drain_mbuf_table(struct mbuf_table *tx_q)
+{
+	uint16_t count;
+
+	count = rte_eth_tx_burst(ports[0], tx_q->txq_id,
+				 tx_q->m_table, tx_q->len);
+	if (unlikely(count < tx_q->len))
+		free_pkts(&tx_q->m_table[count], tx_q->len - count);
+
+	tx_q->len = 0;
+}
+
 /*
- * This function routes the TX packet to the correct interface. This may be a local device
- * or the physical port.
+ * This function routes the TX packet to the correct interface. This
+ * may be a local device or the physical port.
  */
 static inline void __attribute__((always_inline))
 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
 {
 	struct mbuf_table *tx_q;
-	struct rte_mbuf **m_table;
-	unsigned len, ret, offset = 0;
+	unsigned offset = 0;
 	const uint16_t lcore_id = rte_lcore_id();
 	struct virtio_net *dev = vdev->dev;
 	struct ether_hdr *nh;
@@ -960,7 +981,6 @@ queue2nic:
 
 	/*Add packet to the port tx queue*/
 	tx_q = &lcore_tx_queue[lcore_id];
-	len = tx_q->len;
 
 	nh = rte_pktmbuf_mtod(m, struct ether_hdr *);
 	if (unlikely(nh->ether_type == rte_cpu_to_be_16(ETHER_TYPE_VLAN))) {
@@ -998,55 +1018,130 @@ queue2nic:
 	if (m->ol_flags & PKT_TX_TCP_SEG)
 		virtio_tx_offload(m);
 
-	tx_q->m_table[len] = m;
-	len++;
+	tx_q->m_table[tx_q->len++] = m;
 	if (enable_stats) {
 		dev_statistics[dev->device_fh].tx_total++;
 		dev_statistics[dev->device_fh].tx++;
 	}
 
-	if (unlikely(len == MAX_PKT_BURST)) {
-		m_table = (struct rte_mbuf **)tx_q->m_table;
-		ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id, m_table, (uint16_t) len);
-		/* Free any buffers not handled by TX and update the port stats. */
-		if (unlikely(ret < len)) {
-			do {
-				rte_pktmbuf_free(m_table[ret]);
-			} while (++ret < len);
+	if (unlikely(tx_q->len == MAX_PKT_BURST))
+		do_drain_mbuf_table(tx_q);
+}
+
+
+static inline void __attribute__((always_inline))
+drain_mbuf_table(struct mbuf_table *tx_q)
+{
+	static uint64_t prev_tsc;
+	uint64_t cur_tsc;
+
+	if (tx_q->len == 0)
+		return;
+
+	cur_tsc = rte_rdtsc();
+	if (unlikely(cur_tsc - prev_tsc > MBUF_TABLE_DRAIN_TSC)) {
+		prev_tsc = cur_tsc;
+
+		RTE_LOG(DEBUG, VHOST_DATA,
+			"TX queue drained after timeout with burst size %u\n",
+			tx_q->len);
+		do_drain_mbuf_table(tx_q);
+	}
+}
+
+static inline void __attribute__((always_inline))
+drain_eth_rx(struct vhost_dev *vdev)
+{
+	uint16_t rx_count, enqueue_count;
+	struct virtio_net *dev = vdev->dev;
+	struct rte_mbuf *pkts[MAX_PKT_BURST];
+
+	rx_count = rte_eth_rx_burst(ports[0], vdev->vmdq_rx_q,
+				    pkts, MAX_PKT_BURST);
+	if (!rx_count)
+		return;
+
+	/*
+	 * When "enable_retry" is set, here we wait and retry when there
+	 * is no enough free slots in the queue to hold @rx_count packets,
+	 * to diminish packet loss.
+	 */
+	if (enable_retry &&
+	    unlikely(rx_count > rte_vring_available_entries(dev,
+			VIRTIO_RXQ))) {
+		uint32_t retry;
+
+		for (retry = 0; retry < burst_rx_retry_num; retry++) {
+			rte_delay_us(burst_rx_delay_time);
+			if (rx_count <= rte_vring_available_entries(dev,
+					VIRTIO_RXQ))
+				break;
 		}
+	}
+
+	enqueue_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ,
+						pkts, rx_count);
+	if (enable_stats) {
+		uint64_t fh = dev->device_fh;
+
+		rte_atomic64_add(&dev_statistics[fh].rx_total_atomic, rx_count);
+		rte_atomic64_add(&dev_statistics[fh].rx_atomic, enqueue_count);
+	}
 
-		len = 0;
+	free_pkts(pkts, rx_count);
+}
+
+static inline void __attribute__((always_inline))
+drain_virtio_tx(struct vhost_dev *vdev)
+{
+	struct rte_mbuf *pkts[MAX_PKT_BURST];
+	uint16_t count;
+	uint16_t i;
+
+	count = rte_vhost_dequeue_burst(vdev->dev, VIRTIO_TXQ, mbuf_pool,
+					pkts, MAX_PKT_BURST);
+
+	/* setup VMDq for the first packet */
+	if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && count) {
+		if (vdev->remove || link_vmdq(vdev, pkts[0]) == -1)
+			free_pkts(pkts, count);
 	}
 
-	tx_q->len = len;
-	return;
+	for (i = 0; i < count; ++i) {
+		virtio_tx_route(vdev, pkts[i],
+			vlan_tags[(uint16_t)vdev->dev->device_fh]);
+	}
 }
+
 /*
- * This function is called by each data core. It handles all RX/TX registered with the
- * core. For TX the specific lcore linked list is used. For RX, MAC addresses are compared
- * with all devices in the main linked list.
+ * Main function of vhost-switch. It basically does:
+ *
+ * for each vhost device {
+ *    - drain_eth_rx()
+ *
+ *      Which drains the host eth Rx queue linked to the vhost device,
+ *      and deliver all of them to guest virito Rx ring associated with
+ *      this vhost device.
+ *
+ *    - drain_virtio_tx()
+ *
+ *      Which drains the guest virtio Tx queue and deliver all of them
+ *      to the target, which could be another vhost device, or the
+ *      physical eth dev. The route is done in function "virtio_tx_route".
+ * }
  */
 static int
-switch_worker(__attribute__((unused)) void *arg)
+switch_worker(void *arg __rte_unused)
 {
-	struct virtio_net *dev = NULL;
-	struct vhost_dev *vdev = NULL;
-	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+	unsigned i;
+	unsigned lcore_id = rte_lcore_id();
+	struct vhost_dev *vdev;
 	struct mbuf_table *tx_q;
-	const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
-	uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
-	unsigned ret, i;
-	const uint16_t lcore_id = rte_lcore_id();
-	const uint16_t num_cores = (uint16_t)rte_lcore_count();
-	uint16_t rx_count = 0;
-	uint16_t tx_count;
-	uint32_t retry = 0;
 
 	RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
-	prev_tsc = 0;
 
 	tx_q = &lcore_tx_queue[lcore_id];
-	for (i = 0; i < num_cores; i ++) {
+	for (i = 0; i < rte_lcore_count(); i++) {
 		if (lcore_ids[i] == lcore_id) {
 			tx_q->txq_id = i;
 			break;
@@ -1054,34 +1149,7 @@ switch_worker(__attribute__((unused)) void *arg)
 	}
 
 	while(1) {
-		cur_tsc = rte_rdtsc();
-		/*
-		 * TX burst queue drain
-		 */
-		diff_tsc = cur_tsc - prev_tsc;
-		if (unlikely(diff_tsc > drain_tsc)) {
-
-			if (tx_q->len) {
-				RTE_LOG(DEBUG, VHOST_DATA,
-					"TX queue drained after timeout with burst size %u\n",
-					tx_q->len);
-
-				/*Tx any packets in the queue*/
-				ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id,
-									   (struct rte_mbuf **)tx_q->m_table,
-									   (uint16_t)tx_q->len);
-				if (unlikely(ret < tx_q->len)) {
-					do {
-						rte_pktmbuf_free(tx_q->m_table[ret]);
-					} while (++ret < tx_q->len);
-				}
-
-				tx_q->len = 0;
-			}
-
-			prev_tsc = cur_tsc;
-
-		}
+		drain_mbuf_table(tx_q);
 
 		/*
 		 * Inform the configuration core that we have exited the
@@ -1091,69 +1159,20 @@ switch_worker(__attribute__((unused)) void *arg)
 			lcore_info[lcore_id].dev_removal_flag = ACK_DEV_REMOVAL;
 
 		/*
-		 * Process devices
+		 * Process vhost devices
 		 */
 		TAILQ_FOREACH(vdev, &lcore_info[lcore_id].vdev_list, next) {
-			uint64_t fh;
-
-			dev = vdev->dev;
-			fh  = dev->device_fh;
-
 			if (unlikely(vdev->remove)) {
 				unlink_vmdq(vdev);
 				vdev->ready = DEVICE_SAFE_REMOVE;
 				continue;
 			}
 
-			if (likely(vdev->ready == DEVICE_RX)) {
-				/*Handle guest RX*/
-				rx_count = rte_eth_rx_burst(ports[0],
-					vdev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
-
-				if (rx_count) {
-					/*
-					* Retry is enabled and the queue is full then we wait and retry to avoid packet loss
-					* Here MAX_PKT_BURST must be less than virtio queue size
-					*/
-					if (enable_retry && unlikely(rx_count > rte_vring_available_entries(dev, VIRTIO_RXQ))) {
-						for (retry = 0; retry < burst_rx_retry_num; retry++) {
-							rte_delay_us(burst_rx_delay_time);
-							if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
-								break;
-						}
-					}
-					ret_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ, pkts_burst, rx_count);
-					if (enable_stats) {
-						rte_atomic64_add(
-							&dev_statistics[fh].rx_total_atomic,
-							rx_count);
-						rte_atomic64_add(
-							&dev_statistics[fh].rx_atomic,
-							ret_count);
-					}
-					while (likely(rx_count)) {
-						rx_count--;
-						rte_pktmbuf_free(pkts_burst[rx_count]);
-					}
-
-				}
-			}
+			if (likely(vdev->ready == DEVICE_RX))
+				drain_eth_rx(vdev);
 
-			if (likely(!vdev->remove)) {
-				/* Handle guest TX*/
-				tx_count = rte_vhost_dequeue_burst(dev, VIRTIO_TXQ, mbuf_pool, pkts_burst, MAX_PKT_BURST);
-				/* If this is the first received packet we need to learn the MAC and setup VMDQ */
-				if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
-					if (vdev->remove || (link_vmdq(vdev, pkts_burst[0]) == -1)) {
-						while (tx_count)
-							rte_pktmbuf_free(pkts_burst[--tx_count]);
-					}
-				}
-				for (i = 0; i < tx_count; ++i) {
-					virtio_tx_route(vdev, pkts_burst[i],
-						vlan_tags[(uint16_t)dev->device_fh]);
-				}
-			}
+			if (likely(!vdev->remove))
+				drain_virtio_tx(vdev);
 		}
 	}
 
-- 
1.9.3

^ permalink raw reply	[relevance 2%]

* [dpdk-dev] [PATCH 00/16] vhost ABI/API refactoring
@ 2016-05-02 22:25  8% Yuanhan Liu
  2016-05-02 22:25  2% ` [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications Yuanhan Liu
                   ` (2 more replies)
  0 siblings, 3 replies; 200+ results
From: Yuanhan Liu @ 2016-05-02 22:25 UTC (permalink / raw)
  To: dev
  Cc: huawei.xie, Thomas Monjalon, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Yuanhan Liu

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, thefeore, 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.

Note that this refactoring breaks the tep_termination example.
Well, it's just another copy of the original messy vhost example,
and I have no interest to cleanup it again. Therefore, I might
consider to remove that example later, and add the vxlan bits
into vhost example.

Few more TODOs: update release note, update lib version, update
version.map

Thanks.

	--yliu

---
Ilya Maximets (1):
  vhost: make buf vector for scatter Rx local

Yuanhan Liu (15):
  vhost: declare backend with int type
  vhost: set/reset dev flags internally
  vhost: declare device_fh as int
  example/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: query pmd internal by vid
  vhost: add few more functions
  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
  vhost: per device vhost_hlen

 drivers/net/vhost/rte_eth_vhost.c             |  86 ++++-------
 examples/vhost/main.c                         | 126 ++++++++-------
 examples/vhost/main.h                         |   1 +
 lib/librte_vhost/rte_virtio_net.h             | 197 ++++++------------------
 lib/librte_vhost/vhost-net.h                  | 195 +++++++++++++++++++----
 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/virtio-net-user.c |  64 ++++----
 lib/librte_vhost/vhost_user/virtio-net-user.h |  18 +--
 lib/librte_vhost/virtio-net.c                 | 213 ++++++++++++++++----------
 lib/librte_vhost/virtio-net.h                 |  43 ------
 14 files changed, 644 insertions(+), 610 deletions(-)
 delete mode 100644 lib/librte_vhost/virtio-net.h

-- 
1.9.0

^ permalink raw reply	[relevance 8%]

* [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications
  2016-05-02 22:25  8% [dpdk-dev] [PATCH 00/16] vhost ABI/API refactoring Yuanhan Liu
@ 2016-05-02 22:25  2% ` Yuanhan Liu
  2016-05-10 16:17  0%   ` Rich Lane
  2016-05-02 22:25  4% ` [dpdk-dev] [PATCH 14/16] vhost: reserve few more space for future extension Yuanhan Liu
  2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
  2 siblings, 1 reply; 200+ results
From: Yuanhan Liu @ 2016-05-02 22:25 UTC (permalink / raw)
  To: dev
  Cc: huawei.xie, Thomas Monjalon, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Yuanhan Liu

With all the previous prepare works, we are just one step away from
the final ABI refactoring. That is, to change current API to let them
stick to vid instead of the old virtio_net dev.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 drivers/net/vhost/rte_eth_vhost.c             | 61 ++++++++++-----------------
 examples/vhost/main.c                         | 41 ++++++++++++------
 lib/librte_vhost/rte_virtio_net.h             | 30 +++++--------
 lib/librte_vhost/vhost_rxtx.c                 | 15 ++++++-
 lib/librte_vhost/vhost_user/virtio-net-user.c | 14 +++---
 lib/librte_vhost/virtio-net.c                 | 17 +++++---
 6 files changed, 91 insertions(+), 87 deletions(-)

diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c
index 9763cd4..a9dada5 100644
--- a/drivers/net/vhost/rte_eth_vhost.c
+++ b/drivers/net/vhost/rte_eth_vhost.c
@@ -68,9 +68,9 @@ static struct ether_addr base_eth_addr = {
 };
 
 struct vhost_queue {
+	int vid;
 	rte_atomic32_t allow_queuing;
 	rte_atomic32_t while_queuing;
-	struct virtio_net *device;
 	struct pmd_internal *internal;
 	struct rte_mempool *mb_pool;
 	uint8_t port;
@@ -137,7 +137,7 @@ eth_vhost_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 		goto out;
 
 	/* Dequeue packets from guest TX queue */
-	nb_rx = rte_vhost_dequeue_burst(r->device,
+	nb_rx = rte_vhost_dequeue_burst(r->vid,
 			r->virtqueue_id, r->mb_pool, bufs, nb_bufs);
 
 	r->rx_pkts += nb_rx;
@@ -168,7 +168,7 @@ eth_vhost_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 		goto out;
 
 	/* Enqueue packets to guest RX queue */
-	nb_tx = rte_vhost_enqueue_burst(r->device,
+	nb_tx = rte_vhost_enqueue_burst(r->vid,
 			r->virtqueue_id, bufs, nb_bufs);
 
 	r->tx_pkts += nb_tx;
@@ -217,7 +217,7 @@ find_internal_resource(int vid)
 }
 
 static int
-new_device(struct virtio_net *dev)
+new_device(int vid)
 {
 	struct rte_eth_dev *eth_dev;
 	struct internal_list *list;
@@ -228,23 +228,17 @@ new_device(struct virtio_net *dev)
 	int newnode;
 #endif
 
-	if (dev == NULL) {
-		RTE_LOG(INFO, PMD, "Invalid argument\n");
-		return -1;
-	}
-
-	list = find_internal_resource(dev->vid);
+	list = find_internal_resource(vid);
 	if (list == NULL) {
-		RTE_LOG(INFO, PMD, "Invalid vid %d\n", dev->vid);
+		RTE_LOG(INFO, PMD, "Invalid vid %d\n", vid);
 		return -1;
 	}
 
 	eth_dev = list->eth_dev;
 	internal = eth_dev->data->dev_private;
-	internal->vid = dev->vid;
 
 #ifdef RTE_LIBRTE_VHOST_NUMA
-	newnode = rte_vhost_get_numa_node(dev->vid);
+	newnode = rte_vhost_get_numa_node(vid);
 	if (newnode > 0)
 		eth_dev->data->numa_node = newnode;
 #endif
@@ -253,7 +247,7 @@ new_device(struct virtio_net *dev)
 		vq = eth_dev->data->rx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = dev;
+		vq->vid = vid;
 		vq->internal = internal;
 		vq->port = eth_dev->data->port_id;
 	}
@@ -261,15 +255,14 @@ new_device(struct virtio_net *dev)
 		vq = eth_dev->data->tx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = dev;
+		vq->vid = vid;
 		vq->internal = internal;
 		vq->port = eth_dev->data->port_id;
 	}
 
-	for (i = 0; i < rte_vhost_get_queue_num(dev->vid) * VIRTIO_QNUM; i++)
-		rte_vhost_enable_guest_notification(dev, i, 0);
+	for (i = 0; i < rte_vhost_get_queue_num(vid) * VIRTIO_QNUM; i++)
+		rte_vhost_enable_guest_notification(vid, i, 0);
 
-	dev->priv = eth_dev;
 	eth_dev->data->dev_link.link_status = ETH_LINK_UP;
 
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
@@ -293,22 +286,19 @@ new_device(struct virtio_net *dev)
 }
 
 static void
-destroy_device(volatile struct virtio_net *dev)
+destroy_device(int vid)
 {
+	struct internal_list *list;
 	struct rte_eth_dev *eth_dev;
 	struct vhost_queue *vq;
 	unsigned i;
 
-	if (dev == NULL) {
-		RTE_LOG(INFO, PMD, "Invalid argument\n");
-		return;
-	}
-
-	eth_dev = (struct rte_eth_dev *)dev->priv;
-	if (eth_dev == NULL) {
-		RTE_LOG(INFO, PMD, "Failed to find a ethdev\n");
+	list = find_internal_resource(vid);
+	if (list == NULL) {
+		RTE_LOG(INFO, PMD, "Invalid vid %d\n", vid);
 		return;
 	}
+	eth_dev = list->eth_dev;
 
 	/* Wait until rx/tx_pkt_burst stops accessing vhost device */
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
@@ -330,19 +320,17 @@ destroy_device(volatile struct virtio_net *dev)
 
 	eth_dev->data->dev_link.link_status = ETH_LINK_DOWN;
 
-	dev->priv = NULL;
-
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
 		vq = eth_dev->data->rx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = NULL;
+		vq->vid = -1;
 	}
 	for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
 		vq = eth_dev->data->tx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = NULL;
+		vq->vid = -1;
 	}
 
 	RTE_LOG(INFO, PMD, "Connection closed\n");
@@ -351,20 +339,15 @@ destroy_device(volatile struct virtio_net *dev)
 }
 
 static int
-vring_state_changed(struct virtio_net *dev, uint16_t vring, int enable)
+vring_state_changed(int vid, uint16_t vring, int enable)
 {
 	struct rte_vhost_vring_state *state;
 	struct rte_eth_dev *eth_dev;
 	struct internal_list *list;
 
-	if (dev == NULL) {
-		RTE_LOG(ERR, PMD, "Invalid argument\n");
-		return -1;
-	}
-
-	list = find_internal_resource(dev->vid);
+	list = find_internal_resource(vid);
 	if (list == NULL) {
-		RTE_LOG(ERR, PMD, "Invalid vid %d\n", dev->vid);
+		RTE_LOG(ERR, PMD, "Invalid vid %d\n", vid);
 		return -1;
 	}
 
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index 145fa6f..bbf0d28 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -701,6 +701,19 @@ find_vhost_dev(struct ether_addr *mac)
 	return NULL;
 }
 
+static inline struct vhost_dev *__attribute__((always_inline))
+find_vhost_dev_by_vid(int vid)
+{
+	struct vhost_dev *vdev;
+
+	TAILQ_FOREACH(vdev, &vhost_dev_list, next) {
+		if (vdev->ready == DEVICE_RX && vdev->vid == vid)
+			return vdev;
+	}
+
+	return NULL;
+}
+
 /*
  * This function learns the MAC address of the device and registers this along with a
  * vlan tag to a VMDQ.
@@ -796,7 +809,7 @@ virtio_xmit(struct vhost_dev *dst_vdev, struct vhost_dev *src_vdev,
 {
 	uint16_t ret;
 
-	ret = rte_vhost_enqueue_burst(dst_vdev->dev, VIRTIO_RXQ, &m, 1);
+	ret = rte_vhost_enqueue_burst(dst_vdev->vid, VIRTIO_RXQ, &m, 1);
 	if (enable_stats) {
 		rte_atomic64_inc(&dst_vdev->stats.rx_total_atomic);
 		rte_atomic64_add(&dst_vdev->stats.rx_atomic, ret);
@@ -1042,7 +1055,6 @@ static inline void __attribute__((always_inline))
 drain_eth_rx(struct vhost_dev *vdev)
 {
 	uint16_t rx_count, enqueue_count;
-	struct virtio_net *dev = vdev->dev;
 	struct rte_mbuf *pkts[MAX_PKT_BURST];
 
 	rx_count = rte_eth_rx_burst(ports[0], vdev->vmdq_rx_q,
@@ -1068,7 +1080,7 @@ drain_eth_rx(struct vhost_dev *vdev)
 		}
 	}
 
-	enqueue_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ,
+	enqueue_count = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
 						pkts, rx_count);
 	if (enable_stats) {
 		rte_atomic64_add(&vdev->stats.rx_total_atomic, rx_count);
@@ -1085,7 +1097,7 @@ drain_virtio_tx(struct vhost_dev *vdev)
 	uint16_t count;
 	uint16_t i;
 
-	count = rte_vhost_dequeue_burst(vdev->dev, VIRTIO_TXQ, mbuf_pool,
+	count = rte_vhost_dequeue_burst(vdev->vid, VIRTIO_TXQ, mbuf_pool,
 					pkts, MAX_PKT_BURST);
 
 	/* setup VMDq for the first packet */
@@ -1171,13 +1183,17 @@ switch_worker(void *arg __rte_unused)
  * of dev->remove=1 which can cause an infinite loop in the rte_pause loop.
  */
 static void
-destroy_device (volatile struct virtio_net *dev)
+destroy_device(int vid)
 {
 	struct vhost_dev *vdev;
 	int lcore;
 
-	vdev = (struct vhost_dev *)dev->priv;
-	/*set the remove flag. */
+	vdev = find_vhost_dev_by_vid(vid);
+	if (!vdev) {
+		RTE_LOG(ERR, VHOST_CONFIG, "(%d) failed to find device\n", vid);
+		return;
+	}
+
 	vdev->remove = 1;
 	while(vdev->ready != DEVICE_SAFE_REMOVE) {
 		rte_pause();
@@ -1214,12 +1230,11 @@ destroy_device (volatile struct virtio_net *dev)
  * and the allocated to a specific data core.
  */
 static int
-new_device (struct virtio_net *dev)
+new_device(int vid)
 {
 	int lcore, core_add = 0;
 	uint32_t device_num_min = num_devices;
 	struct vhost_dev *vdev;
-	int vid = dev->vid;
 
 	vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
 	if (vdev == NULL) {
@@ -1228,8 +1243,6 @@ new_device (struct virtio_net *dev)
 			vid);
 		return -1;
 	}
-	vdev->dev = dev;
-	dev->priv = vdev;
 	vdev->vid = vid;
 
 	TAILQ_INSERT_TAIL(&vhost_dev_list, vdev, next);
@@ -1252,8 +1265,8 @@ new_device (struct virtio_net *dev)
 	lcore_info[vdev->coreid].device_num++;
 
 	/* Disable notifications. */
-	rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
-	rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
+	rte_vhost_enable_guest_notification(vid, VIRTIO_RXQ, 0);
+	rte_vhost_enable_guest_notification(vid, VIRTIO_TXQ, 0);
 
 	RTE_LOG(INFO, VHOST_DATA,
 		"(%d) device has been added to data core %d\n",
@@ -1309,7 +1322,7 @@ print_stats(void)
 				"RX total:              %" PRIu64 "\n"
 				"RX dropped:            %" PRIu64 "\n"
 				"RX successful:         %" PRIu64 "\n",
-				vdev->dev->vid,
+				vdev->vid,
 				tx_total, tx_dropped, tx,
 				rx_total, rx_dropped, rx);
 		}
diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
index 27f6847..0a26df9 100644
--- a/lib/librte_vhost/rte_virtio_net.h
+++ b/lib/librte_vhost/rte_virtio_net.h
@@ -178,23 +178,12 @@ struct virtio_memory {
  *
  */
 struct virtio_net_device_ops {
-	int (*new_device)(struct virtio_net *);	/**< Add device. */
-	void (*destroy_device)(volatile struct virtio_net *);	/**< Remove device. */
+	int (*new_device)(int vid);		/**< Add device. */
+	void (*destroy_device)(int vid);	/**< Remove device. */
 
-	int (*vring_state_changed)(struct virtio_net *dev, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
+	int (*vring_state_changed)(int vid, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
 };
 
-static inline uint16_t __attribute__((always_inline))
-rte_vring_available_entries(struct virtio_net *dev, uint16_t queue_id)
-{
-	struct vhost_virtqueue *vq = dev->virtqueue[queue_id];
-
-	if (!vq->enabled)
-		return 0;
-
-	return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx_res;
-}
-
 /**
  * Function to convert guest physical addresses to vhost virtual addresses.
  * This is used to convert guest virtio buffer addresses.
@@ -231,7 +220,7 @@ int rte_vhost_feature_enable(uint64_t feature_mask);
 /* Returns currently supported vhost features */
 uint64_t rte_vhost_feature_get(void);
 
-int rte_vhost_enable_guest_notification(struct virtio_net *dev, uint16_t queue_id, int enable);
+int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
 
 /* Register vhost driver. dev_name could be different for multiple instance support. */
 int rte_vhost_driver_register(const char *dev_name);
@@ -286,8 +275,8 @@ int rte_vhost_get_numa_node(int vid);
  * be received from the physical port or from another virtual device. A packet
  * count is returned to indicate the number of packets that were succesfully
  * added to the RX queue.
- * @param dev
- *  virtio-net device
+ * @param vid
+ *  virtio-net device ID
  * @param queue_id
  *  virtio queue index in mq case
  * @param pkts
@@ -297,14 +286,14 @@ int rte_vhost_get_numa_node(int vid);
  * @return
  *  num of packets enqueued
  */
-uint16_t rte_vhost_enqueue_burst(struct virtio_net *dev, uint16_t queue_id,
+uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
 	struct rte_mbuf **pkts, uint16_t count);
 
 /**
  * This function gets guest buffers from the virtio device TX virtqueue,
  * construct host mbufs, copies guest buffer content to host mbufs and
  * store them in pkts to be processed.
- * @param dev
+ * @param vid
  *  virtio-net device
  * @param queue_id
  *  virtio queue index in mq case
@@ -317,7 +306,8 @@ uint16_t rte_vhost_enqueue_burst(struct virtio_net *dev, uint16_t queue_id,
  * @return
  *  num of packets dequeued
  */
-uint16_t rte_vhost_dequeue_burst(struct virtio_net *dev, uint16_t queue_id,
+uint16_t rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count);
 
+
 #endif /* _VIRTIO_NET_H_ */
diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
index 8d87508..08cab08 100644
--- a/lib/librte_vhost/vhost_rxtx.c
+++ b/lib/librte_vhost/vhost_rxtx.c
@@ -46,6 +46,7 @@
 #include <rte_arp.h>
 
 #include "vhost-net.h"
+#include "virtio-net.h"
 
 #define MAX_PKT_BURST 32
 #define VHOST_LOG_PAGE	4096
@@ -587,9 +588,14 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id,
 }
 
 uint16_t
-rte_vhost_enqueue_burst(struct virtio_net *dev, uint16_t queue_id,
+rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
 	struct rte_mbuf **pkts, uint16_t count)
 {
+	struct virtio_net *dev = get_device(vid);
+
+	if (!dev)
+		return 0;
+
 	if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF))
 		return virtio_dev_merge_rx(dev, queue_id, pkts, count);
 	else
@@ -815,9 +821,10 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 }
 
 uint16_t
-rte_vhost_dequeue_burst(struct virtio_net *dev, uint16_t queue_id,
+rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count)
 {
+	struct virtio_net *dev;
 	struct rte_mbuf *rarp_mbuf = NULL;
 	struct vhost_virtqueue *vq;
 	uint32_t desc_indexes[MAX_PKT_BURST];
@@ -826,6 +833,10 @@ rte_vhost_dequeue_burst(struct virtio_net *dev, uint16_t queue_id,
 	uint16_t free_entries;
 	uint16_t avail_idx;
 
+	dev = get_device(vid);
+	if (!dev)
+		return 0;
+
 	if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->virt_qp_nb))) {
 		RTE_LOG(ERR, VHOST_DATA, "(%d) %s: invalid virtqueue idx %d.\n",
 			dev->vid, __func__, queue_id);
diff --git a/lib/librte_vhost/vhost_user/virtio-net-user.c b/lib/librte_vhost/vhost_user/virtio-net-user.c
index 9385af1..7fa69a7 100644
--- a/lib/librte_vhost/vhost_user/virtio-net-user.c
+++ b/lib/librte_vhost/vhost_user/virtio-net-user.c
@@ -117,7 +117,7 @@ user_set_mem_table(int vid, struct VhostUserMsg *pmsg)
 	/* Remove from the data plane. */
 	if (dev->flags & VIRTIO_DEV_RUNNING) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	if (dev->mem) {
@@ -279,6 +279,9 @@ user_set_vring_kick(int vid, struct VhostUserMsg *pmsg)
 	struct vhost_vring_file file;
 	struct virtio_net *dev = get_device(vid);
 
+	if (!dev)
+		return;
+
 	file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
 	if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
 		file.fd = VIRTIO_INVALID_EVENTFD;
@@ -289,7 +292,7 @@ user_set_vring_kick(int vid, struct VhostUserMsg *pmsg)
 	vhost_set_vring_kick(vid, &file);
 
 	if (virtio_is_ready(dev) && !(dev->flags & VIRTIO_DEV_RUNNING)) {
-		if (notify_ops->new_device(dev) == 0)
+		if (notify_ops->new_device(vid) == 0)
 			dev->flags |= VIRTIO_DEV_RUNNING;
 	}
 }
@@ -307,7 +310,7 @@ user_get_vring_base(int vid,
 		return -1;
 	/* We have to stop the queue (virtio) if it is running. */
 	if (dev->flags & VIRTIO_DEV_RUNNING)
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 
 	/* Here we are safe to get the last used index */
 	vhost_get_vring_base(vid, state->index, state);
@@ -342,9 +345,8 @@ user_set_vring_enable(int vid,
 		"set queue enable: %d to qp idx: %d\n",
 		enable, state->index);
 
-	if (notify_ops->vring_state_changed) {
-		notify_ops->vring_state_changed(dev, state->index, enable);
-	}
+	if (notify_ops->vring_state_changed)
+		notify_ops->vring_state_changed(vid, state->index, enable);
 
 	dev->virtqueue[state->index]->enabled = enable;
 
diff --git a/lib/librte_vhost/virtio-net.c b/lib/librte_vhost/virtio-net.c
index 6bf4d87..9fd80a8 100644
--- a/lib/librte_vhost/virtio-net.c
+++ b/lib/librte_vhost/virtio-net.c
@@ -296,7 +296,7 @@ vhost_destroy_device(int vid)
 
 	if (dev->flags & VIRTIO_DEV_RUNNING) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	cleanup_device(dev, 1);
@@ -353,7 +353,7 @@ vhost_reset_owner(int vid)
 
 	if (dev->flags & VIRTIO_DEV_RUNNING) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	cleanup_device(dev, 0);
@@ -717,21 +717,26 @@ vhost_set_backend(int vid, struct vhost_vring_file *file)
 	if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
 		if (dev->virtqueue[VIRTIO_TXQ]->backend != VIRTIO_DEV_STOPPED &&
 		    dev->virtqueue[VIRTIO_RXQ]->backend != VIRTIO_DEV_STOPPED) {
-			if (notify_ops->new_device(dev) < 0)
+			if (notify_ops->new_device(vid) < 0)
 				return -1;
 			dev->flags |= VIRTIO_DEV_RUNNING;
 		}
 	} else if (file->fd == VIRTIO_DEV_STOPPED) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	return 0;
 }
 
-int rte_vhost_enable_guest_notification(struct virtio_net *dev,
-	uint16_t queue_id, int enable)
+int
+rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
 {
+	struct virtio_net *dev = get_device(vid);
+
+	if (!dev)
+		return -1;
+
 	if (enable) {
 		RTE_LOG(ERR, VHOST_CONFIG,
 			"guest notification isn't supported.\n");
-- 
1.9.0

^ permalink raw reply	[relevance 2%]

* [dpdk-dev] [PATCH 14/16] vhost: reserve few more space for future extension
  2016-05-02 22:25  8% [dpdk-dev] [PATCH 00/16] vhost ABI/API refactoring Yuanhan Liu
  2016-05-02 22:25  2% ` [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications Yuanhan Liu
@ 2016-05-02 22:25  4% ` Yuanhan Liu
  2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
  2 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-02 22:25 UTC (permalink / raw)
  To: dev
  Cc: huawei.xie, Thomas Monjalon, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Yuanhan Liu

"virtio_net_device_ops" is the only left open struct that an application
can access, therefore, it's the only place that might introduce potential
ABI break in future for extension.

So, do some reservation for it. 5 should be pretty enough, considering
that we have barely touched it for a long while. Another reason to
choose 5 is for cache alignment: 5 makes the struct 64 bytes for 64 bit
machine.

With this, it's confidence to say that we might be able to be free from
the ABI violation forever.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 lib/librte_vhost/rte_virtio_net.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
index 388621e..4e50425 100644
--- a/lib/librte_vhost/rte_virtio_net.h
+++ b/lib/librte_vhost/rte_virtio_net.h
@@ -77,6 +77,8 @@ struct virtio_net_device_ops {
 	void (*destroy_device)(int vid);	/**< Remove device. */
 
 	int (*vring_state_changed)(int vid, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
+
+	void *reserved[5]; /**< Reserved for future extension */
 };
 
 /**
-- 
1.9.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH 0/3] [RFC] vhost: micro vhost optimization
@ 2016-05-03  0:46  3% Yuanhan Liu
  2016-05-10 21:49  0% ` Rich Lane
  0 siblings, 1 reply; 200+ results
From: Yuanhan Liu @ 2016-05-03  0:46 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Yuanhan Liu

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 3%]

* [dpdk-dev] [PATCH 5/5] doc: update doc for packet capture framework
  @ 2016-05-06 10:55  6% ` Reshma Pattan
    1 sibling, 0 replies; 200+ results
From: Reshma Pattan @ 2016-05-06 10:55 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 | 121 ++++++++++++++++++++++++++++++++
 doc/guides/rel_notes/release_16_07.rst  |   7 ++
 doc/guides/sample_app_ug/index.rst      |   1 +
 doc/guides/sample_app_ug/pdump.rst      | 109 ++++++++++++++++++++++++++++
 6 files changed, 242 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 b6a39c7..6ddc818 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -432,6 +432,9 @@ 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..6af77b9
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,121 @@
+..  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:
+
+pdump Library
+=============
+
+Pdump library provides framework for packet capturing on DPDK.
+
+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 un initialize the packet capture framework.
+
+Pdump library works on server and client 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.
+
+Library provides below APIs
+
+``rte_pdump_init()``
+This API initializes the packet capture framework.
+
+``rte_pdump_enable()``
+This API enables the packet capturing on a given port and queue.
+Note: filter option in the API is place holder for future use.
+
+``rte_pdump_enable_by_deviceid()``
+This API enables the packet capturing on a given device id
+(device name or pci address) and queue.
+Note: filter option in the API is place holder for future use.
+
+``rte_pdump_disable()``
+This API disables the packet capturing on a given port and queue.
+
+``rte_pdump_disable_by_deviceid()``
+This API disables the packet capturing on a given device_id and queue.
+
+``rte_pdump_uninit()``
+This API un initializes the packet capture framework.
+
+
+Implementation Details
+----------------------
+
+On a call to library API ``rte_pdump_init()``, library creates pthread and server socket.
+Server socket in pthread context will be listening to the client requests to enable/disable
+the packet capture.
+
+Who ever calls this API first will have server socket created,
+subsequent calls to this APIs will not create any further server sockets. i.e only one server
+socket is allowed.
+
+On each call to library APIs ``rte_pdump_enable()/rte_pdump_enable_by_deviceid()``
+to enable the packet capture, library creates separate client sockets,
+builds up enable request and sends the request to the server.
+Server listening on the socket will serve the request, enable the packet capture
+by registering ethernet rx/tx callbacks for the given port/device_id and queue combinations.
+Server mirrors the packets to new mempool and enqueue them to the ring that clients has passed
+in these APIs.
+Server sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, client sockets will be closed.
+
+On each call to library APIs ``rte_pdump_disable()/rte_pdump_disable_by_deviceid()``
+to disable packet capture, library creates separate client sockets,
+builds up disable request and sends the request to the server.
+Server listening on the socket will serve the request, disable the packet capture
+by removing the ethernet rx/tx callbacks for the given port/device_id and queue combinations.
+Server sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, client sockets will be closed.
+
+On a call to library API ``rte_pdump_uninit()``, library closes the pthread and the server socket.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+app/pdump tool is developed based on this library to capture the packets
+in DPDK.
+Users can develop their own packet capturing application using new library
+if they wish to do so.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 83c841b..4d6ab10 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -34,6 +34,10 @@ This section should contain new features added in this release. Sample format:
 
   Refer to the previous release notes for examples.
 
+* **Added packet capturing support.**
+
+  Now users have facility to capture packets on dpdk ports using librte_pdump
+  and app/pdump tool.
 
 Resolved Issues
 ---------------
@@ -90,6 +94,7 @@ This section should contain API changes. Sample format:
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
 
+* Now function ``rte_eth_dev_get_port_by_name`` changed to public API.
 
 ABI Changes
 -----------
@@ -101,6 +106,8 @@ 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..c185550
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,109 @@
+
+..  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 capturing
+on dpdk ports and capturing the packets.
+
+Running the Application
+-----------------------
+The application has a pdump command line option with various sub arguments inside:
+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.
+
+.. code-block:: console
+
+   ./$(RTE_TARGET)/app/pdump -- --pdump '(port=<port_id> |
+   device_id=<pci address or device name>),
+   (queue=2), (rx-dev=<iface/path to pcap file> |
+   tx-dev=<iface/path to pcap file> |
+   rxtx-dev=<iface/path to pcap file>),
+   [ring-size=1024], [mbuf-size=2048], [total-num-mbufs=8191]'
+
+Parameters
+~~~~~~~~~~
+``--pdump``: Specifies arguments needed for packet capturing.
+
+``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.
+
+``queue``
+Queue id of the eth device on which packets should be captured.
+User can pass on queue value as ‘*’ if packets capturing has to be enabled
+on all queues of the eth device.
+
+``rx-dev``
+Can be either pcap file name or any linux iface onto which ingress side packets of
+dpdk eth device will be sent on for users to view.
+
+``tx-dev``
+Can be either pcap file name or any linux iface onto which egress side packets of
+dpdk eth device will be sent on for users to view.
+
+``rxtx-dev``
+Can be either pcap file name or any linux iface onto which both ingress &
+egress side packets of dpdk eth device will be sent on for users to view.
+
+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 should pass on both rx-dev and tx-dev.
+To receive both ingress and egress packets on same device, should pass only rxtx-dev.
+
+Pdump tool uses these devices internally to create PCAPPMD vdev having ``tx_stream``
+as either of these devices.
+
+``ring-size``
+Size of the ring. This value is used internally for ring creation.
+The ring will be used to enqueue the packets from primary application to secondary.
+
+``mbuf-size``
+Size of the mbuf data room size. This is used internally for mempool creation.
+Ideally this value must be same as primary application's mempool which is used for
+packet rx.
+
+``total-num-mbufs``
+Total number mbufs in mempool. This is used internally for mempool creation.
+
+Example
+-------
+
+.. code-block:: console
+
+        $ sudo ./x86_64-native-linuxapp-gcc/app/dpdk_pdump -- --pdump 'device_id=0000:02:00.0,queue=*,rx-dev=/tmp/rx-file.pcap,tx-dev=/tmp/tx-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=16384' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/rx2-file.pcap,tx-dev=/tmp/tx2-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
-- 
2.5.0

^ permalink raw reply	[relevance 6%]

* Re: [dpdk-dev] [PATCH v1] hash: add tsx support for cuckoo hash
  @ 2016-05-07  4:56  3% ` Stephen Hemminger
  2016-05-09 16:51  4%   ` Shen, Wei1
  0 siblings, 1 reply; 200+ results
From: Stephen Hemminger @ 2016-05-07  4:56 UTC (permalink / raw)
  To: Shen Wei; +Cc: dev, pablo.de.lara.guarch, christian.maciocco, Sameh Gobriel

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_QUEUE_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 3%]

* [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability
@ 2016-05-07  6:40  3% Yuanhan Liu
  2016-05-10  3:23  3% ` Xu, Qian Q
  2016-05-13  6:16  3% ` [dpdk-dev] [PATCH v2 " Yuanhan Liu
  0 siblings, 2 replies; 200+ results
From: Yuanhan Liu @ 2016-05-07  6:40 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Yuanhan Liu

Both the vhost-user backend (DPDK here) and frontend (QEMU) could be
server, as well as client. DPDK just acts as server so far. This patch
set would make it possible to act as both.

A new arg (flags) is introduced 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.

With the DPDK as client, it's easier to implement the "reconnect" ability,
which means we could still make vhost-user work after DPDK restarts.


---
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 and reconnect option
  vhost: add pmd client and reconnect option

 drivers/net/vhost/rte_eth_vhost.c            |  54 +++-
 examples/vhost/main.c                        |  23 +-
 lib/librte_vhost/rte_virtio_net.h            |  12 +-
 lib/librte_vhost/vhost_user/vhost-net-user.c | 355 ++++++++++++++++++---------
 lib/librte_vhost/vhost_user/vhost-net-user.h |   6 -
 lib/librte_vhost/virtio-net.c                |   8 +
 6 files changed, 313 insertions(+), 145 deletions(-)

-- 
1.9.0

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH v1] hash: add tsx support for cuckoo hash
  2016-05-07  4:56  3% ` Stephen Hemminger
@ 2016-05-09 16:51  4%   ` Shen, Wei1
  0 siblings, 0 replies; 200+ results
From: Shen, Wei1 @ 2016-05-09 16:51 UTC (permalink / raw)
  To: Stephen Hemminger
  Cc: dev, De Lara Guarch, Pablo, Maciocco, Christian, Gobriel, Sameh

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.

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 4%]

* Re: [dpdk-dev] [PATCH v2 0/8] vhost/example cleanup/fix
       [not found]     ` <1462224230-19460-1-git-send-email-yuanhan.liu@linux.intel.com>
  2016-05-02 21:23  2%   ` [dpdk-dev] [PATCH v2 7/8] examples/vhost: switch_worker cleanup Yuanhan Liu
@ 2016-05-09 18:06  0%   ` Yuanhan Liu
  1 sibling, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-09 18:06 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie

Series applied to dpdk-next-virtio.

	--yliu

On Mon, May 02, 2016 at 02:23:42PM -0700, Yuanhan Liu wrote:
> I'm starting to work on the vhost ABI refactoring, that I also have to
> touch the vhost example code. The vhost example code, however, is very
> messy, full of __very__ long lines. This would make a later diff to
> apply the new vhost API be very ugly, therefore, not friendly for review.
> This is how this cleanup comes.
> 
> Besides that, there is one enhancement patch, which handles the broadcast
> packets so that we could rely the ARP request packet, to let vhost-switch
> be more like a real switch. There is another patch that (hopefully) would
> fix the mbuf allocation failure ultimately. I also added some guidelines
> there as comments to show how to count how many mbuf entries is enough for
> our usage.
> 
> In another word, an example is meant to be clean/simple and with good
> coding style so that people can get the usage easily. So, one way or
> another, this patch is good to have, even without this ABI refactoring
> stuff.
> 
> Note that I'm going to apply it before the end of this week, if no objections.
> 
> 
> v2: - some checkpatch fixes
> 
>     - cleaned the code about device statistics
> 
> ---
> Yuanhan Liu (8):
>   examples/vhost: remove the non-working zero copy code
>   examples/vhost: remove unused macro and struct
>   examples/vhost: use tailq to link vhost devices
>   examples/vhost: use mac compare helper function directly
>   examples/vhost: handle broadcast packet
>   examples/vhost: fix mbuf allocation failure
>   examples/vhost: switch_worker cleanup
>   examples/vhost: embed statistics into vhost_dev struct
> 
>  doc/guides/sample_app_ug/vhost.rst |   36 +-
>  examples/vhost/main.c              | 2394 ++++++------------------------------
>  examples/vhost/main.h              |   56 +-
>  3 files changed, 391 insertions(+), 2095 deletions(-)
> 
> -- 
> 1.9.3

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability
  2016-05-07  6:40  3% [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability Yuanhan Liu
@ 2016-05-10  3:23  3% ` Xu, Qian Q
  2016-05-10 17:41  0%   ` Yuanhan Liu
  2016-05-13  6:16  3% ` [dpdk-dev] [PATCH v2 " Yuanhan Liu
  1 sibling, 1 reply; 200+ results
From: Xu, Qian Q @ 2016-05-10  3:23 UTC (permalink / raw)
  To: Yuanhan Liu, dev; +Cc: Xie, Huawei

Do we need patch qemu for the reconnect case?  

Thanks
Qian


-----Original Message-----
From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Yuanhan Liu
Sent: Saturday, May 07, 2016 2:40 PM
To: dev@dpdk.org
Cc: Xie, Huawei; Yuanhan Liu
Subject: [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability

Both the vhost-user backend (DPDK here) and frontend (QEMU) could be server, as well as client. DPDK just acts as server so far. This patch set would make it possible to act as both.

A new arg (flags) is introduced 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.

With the DPDK as client, it's easier to implement the "reconnect" ability, which means we could still make vhost-user work after DPDK restarts.


---
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 and reconnect option
  vhost: add pmd client and reconnect option

 drivers/net/vhost/rte_eth_vhost.c            |  54 +++-
 examples/vhost/main.c                        |  23 +-
 lib/librte_vhost/rte_virtio_net.h            |  12 +-
 lib/librte_vhost/vhost_user/vhost-net-user.c | 355 ++++++++++++++++++---------
 lib/librte_vhost/vhost_user/vhost-net-user.h |   6 -
 lib/librte_vhost/virtio-net.c                |   8 +
 6 files changed, 313 insertions(+), 145 deletions(-)

--
1.9.0

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCHv2 5/5] doc: update doc for packet capture framework
  @ 2016-05-10  9:40  6%   ` Reshma Pattan
    1 sibling, 0 replies; 200+ results
From: Reshma Pattan @ 2016-05-10  9:40 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 | 121 ++++++++++++++++++++++++++++++++
 doc/guides/rel_notes/release_16_07.rst  |   7 ++
 doc/guides/sample_app_ug/index.rst      |   1 +
 doc/guides/sample_app_ug/pdump.rst      | 109 ++++++++++++++++++++++++++++
 6 files changed, 242 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 b6a39c7..6ddc818 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -432,6 +432,9 @@ 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..6af77b9
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,121 @@
+..  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:
+
+pdump Library
+=============
+
+Pdump library provides framework for packet capturing on DPDK.
+
+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 un initialize the packet capture framework.
+
+Pdump library works on server and client 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.
+
+Library provides below APIs
+
+``rte_pdump_init()``
+This API initializes the packet capture framework.
+
+``rte_pdump_enable()``
+This API enables the packet capturing on a given port and queue.
+Note: filter option in the API is place holder for future use.
+
+``rte_pdump_enable_by_deviceid()``
+This API enables the packet capturing on a given device id
+(device name or pci address) and queue.
+Note: filter option in the API is place holder for future use.
+
+``rte_pdump_disable()``
+This API disables the packet capturing on a given port and queue.
+
+``rte_pdump_disable_by_deviceid()``
+This API disables the packet capturing on a given device_id and queue.
+
+``rte_pdump_uninit()``
+This API un initializes the packet capture framework.
+
+
+Implementation Details
+----------------------
+
+On a call to library API ``rte_pdump_init()``, library creates pthread and server socket.
+Server socket in pthread context will be listening to the client requests to enable/disable
+the packet capture.
+
+Who ever calls this API first will have server socket created,
+subsequent calls to this APIs will not create any further server sockets. i.e only one server
+socket is allowed.
+
+On each call to library APIs ``rte_pdump_enable()/rte_pdump_enable_by_deviceid()``
+to enable the packet capture, library creates separate client sockets,
+builds up enable request and sends the request to the server.
+Server listening on the socket will serve the request, enable the packet capture
+by registering ethernet rx/tx callbacks for the given port/device_id and queue combinations.
+Server mirrors the packets to new mempool and enqueue them to the ring that clients has passed
+in these APIs.
+Server sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, client sockets will be closed.
+
+On each call to library APIs ``rte_pdump_disable()/rte_pdump_disable_by_deviceid()``
+to disable packet capture, library creates separate client sockets,
+builds up disable request and sends the request to the server.
+Server listening on the socket will serve the request, disable the packet capture
+by removing the ethernet rx/tx callbacks for the given port/device_id and queue combinations.
+Server sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, client sockets will be closed.
+
+On a call to library API ``rte_pdump_uninit()``, library closes the pthread and the server socket.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+app/pdump tool is developed based on this library to capture the packets
+in DPDK.
+Users can develop their own packet capturing application using new library
+if they wish to do so.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 83c841b..4d6ab10 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -34,6 +34,10 @@ This section should contain new features added in this release. Sample format:
 
   Refer to the previous release notes for examples.
 
+* **Added packet capturing support.**
+
+  Now users have facility to capture packets on dpdk ports using librte_pdump
+  and app/pdump tool.
 
 Resolved Issues
 ---------------
@@ -90,6 +94,7 @@ This section should contain API changes. Sample format:
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
 
+* Now function ``rte_eth_dev_get_port_by_name`` changed to public API.
 
 ABI Changes
 -----------
@@ -101,6 +106,8 @@ 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..c185550
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,109 @@
+
+..  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 capturing
+on dpdk ports and capturing the packets.
+
+Running the Application
+-----------------------
+The application has a pdump command line option with various sub arguments inside:
+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.
+
+.. code-block:: console
+
+   ./$(RTE_TARGET)/app/pdump -- --pdump '(port=<port_id> |
+   device_id=<pci address or device name>),
+   (queue=2), (rx-dev=<iface/path to pcap file> |
+   tx-dev=<iface/path to pcap file> |
+   rxtx-dev=<iface/path to pcap file>),
+   [ring-size=1024], [mbuf-size=2048], [total-num-mbufs=8191]'
+
+Parameters
+~~~~~~~~~~
+``--pdump``: Specifies arguments needed for packet capturing.
+
+``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.
+
+``queue``
+Queue id of the eth device on which packets should be captured.
+User can pass on queue value as ‘*’ if packets capturing has to be enabled
+on all queues of the eth device.
+
+``rx-dev``
+Can be either pcap file name or any linux iface onto which ingress side packets of
+dpdk eth device will be sent on for users to view.
+
+``tx-dev``
+Can be either pcap file name or any linux iface onto which egress side packets of
+dpdk eth device will be sent on for users to view.
+
+``rxtx-dev``
+Can be either pcap file name or any linux iface onto which both ingress &
+egress side packets of dpdk eth device will be sent on for users to view.
+
+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 should pass on both rx-dev and tx-dev.
+To receive both ingress and egress packets on same device, should pass only rxtx-dev.
+
+Pdump tool uses these devices internally to create PCAPPMD vdev having ``tx_stream``
+as either of these devices.
+
+``ring-size``
+Size of the ring. This value is used internally for ring creation.
+The ring will be used to enqueue the packets from primary application to secondary.
+
+``mbuf-size``
+Size of the mbuf data room size. This is used internally for mempool creation.
+Ideally this value must be same as primary application's mempool which is used for
+packet rx.
+
+``total-num-mbufs``
+Total number mbufs in mempool. This is used internally for mempool creation.
+
+Example
+-------
+
+.. code-block:: console
+
+        $ sudo ./x86_64-native-linuxapp-gcc/app/dpdk_pdump -- --pdump 'device_id=0000:02:00.0,queue=*,rx-dev=/tmp/rx-file.pcap,tx-dev=/tmp/tx-file.pcap,ring-size=8192,mbuf-size=2176,total-num-mbufs=16384' --pdump 'device_id=0000:01:00.0,queue=*,rx-dev=/tmp/rx2-file.pcap,tx-dev=/tmp/tx2-file.pcap,ring-size=16384,mbuf-size=2176,total-num-mbufs=32768'
-- 
2.5.0

^ permalink raw reply	[relevance 6%]

* Re: [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications
  2016-05-02 22:25  2% ` [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications Yuanhan Liu
@ 2016-05-10 16:17  0%   ` Rich Lane
  2016-05-10 16:39  0%     ` Yuanhan Liu
  0 siblings, 1 reply; 200+ results
From: Rich Lane @ 2016-05-10 16:17 UTC (permalink / raw)
  To: Yuanhan Liu
  Cc: dev, huawei.xie, Thomas Monjalon, Panu Matilainen,
	Tetsuya Mukawa, Traynor Kevin

On Mon, May 2, 2016 at 3:25 PM, Yuanhan Liu <yuanhan.liu@linux.intel.com>
wrote:

> With all the previous prepare works, we are just one step away from
> the final ABI refactoring. That is, to change current API to let them
> stick to vid instead of the old virtio_net dev.
>

This patch removes the only assignment to internal->vid in the PMD. It's
initialized to zero, so only the first vhost connection will work.

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications
  2016-05-10 16:17  0%   ` Rich Lane
@ 2016-05-10 16:39  0%     ` Yuanhan Liu
  0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-10 16:39 UTC (permalink / raw)
  To: Rich Lane
  Cc: dev, huawei.xie, Thomas Monjalon, Panu Matilainen,
	Tetsuya Mukawa, Traynor Kevin

On Tue, May 10, 2016 at 09:17:23AM -0700, Rich Lane wrote:
> On Mon, May 2, 2016 at 3:25 PM, Yuanhan Liu <yuanhan.liu@linux.intel.com>
> wrote:
> 
>     With all the previous prepare works, we are just one step away from
>     the final ABI refactoring. That is, to change current API to let them
>     stick to vid instead of the old virtio_net dev.
> 
> 
> This patch removes the only assignment to internal->vid in the PMD. It's
> initialized to zero, so only the first vhost connection will work.

I assume you meant to following diff:

-       if (dev == NULL) {
-               RTE_LOG(INFO, PMD, "Invalid argument\n");
-               return -1;
-       }
-
-       list = find_internal_resource(dev->vid);
+       list = find_internal_resource(vid);
        if (list == NULL) {
-               RTE_LOG(INFO, PMD, "Invalid vid %d\n", dev->vid);
+               RTE_LOG(INFO, PMD, "Invalid vid %d\n", vid);
                return -1;
        }

        eth_dev = list->eth_dev;
        internal = eth_dev->data->dev_private;
-       internal->vid = dev->vid;

Then yes, I have no idea why I did that; it's a careless and
hard-to-catch issue. So, thanks a lot for catching it!

Rich, would you help try by adding following line there and
do a test? It would be great if this patch has your Tested-by :)

    internal->vid = vid;

Thanks.

	--yliu

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability
  2016-05-10  3:23  3% ` Xu, Qian Q
@ 2016-05-10 17:41  0%   ` Yuanhan Liu
  0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-10 17:41 UTC (permalink / raw)
  To: Xu, Qian Q; +Cc: dev, Xie, Huawei, marcandre.lureau

On Tue, May 10, 2016 at 03:23:15AM +0000, Xu, Qian Q wrote:
> Do we need patch qemu for the reconnect case?  

Yes, we need some support from QEMU: currently QEMU will not be able
to detect disconnection and hence will not establish the connection
when DPDK vhost restarts.

Following patchset from Marc resolves above issue.

    http://lists.nongnu.org/archive/html/qemu-devel/2016-05/msg01507.html

And note that unlike the vhost-user multiple queue support that depends
on some new vhost-uesr message from QEMU, this patchset does not depond
on QEMU.

	--yliu
> 
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Yuanhan Liu
> Sent: Saturday, May 07, 2016 2:40 PM
> To: dev@dpdk.org
> Cc: Xie, Huawei; Yuanhan Liu
> Subject: [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability
> 
> Both the vhost-user backend (DPDK here) and frontend (QEMU) could be server, as well as client. DPDK just acts as server so far. This patch set would make it possible to act as both.
> 
> A new arg (flags) is introduced 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.
> 
> With the DPDK as client, it's easier to implement the "reconnect" ability, which means we could still make vhost-user work after DPDK restarts.
> 
> 
> ---
> 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 and reconnect option
>   vhost: add pmd client and reconnect option
> 
>  drivers/net/vhost/rte_eth_vhost.c            |  54 +++-
>  examples/vhost/main.c                        |  23 +-
>  lib/librte_vhost/rte_virtio_net.h            |  12 +-
>  lib/librte_vhost/vhost_user/vhost-net-user.c | 355 ++++++++++++++++++---------
>  lib/librte_vhost/vhost_user/vhost-net-user.h |   6 -
>  lib/librte_vhost/virtio-net.c                |   8 +
>  6 files changed, 313 insertions(+), 145 deletions(-)
> 
> --
> 1.9.0

^ permalink raw reply	[relevance 0%]

* Re: [dpdk-dev] [PATCH 0/3] [RFC] vhost: micro vhost optimization
  2016-05-03  0:46  3% [dpdk-dev] [PATCH 0/3] [RFC] vhost: micro vhost optimization Yuanhan Liu
@ 2016-05-10 21:49  0% ` Rich Lane
  2016-05-10 22:08  0%   ` Yuanhan Liu
  0 siblings, 1 reply; 200+ results
From: Rich Lane @ 2016-05-10 21:49 UTC (permalink / raw)
  To: Yuanhan Liu; +Cc: dev, huawei.xie

I see a significant performance improvement with these patches, around 5%
at 64 bytes.

The one patch that didn't give any performance boost for me was "vhost:
arrange virtio_net fields for better cache sharing".

Tested-by: Rich Lane <rich.lane@bigswitch.com>

On Mon, May 2, 2016 at 5:46 PM, Yuanhan Liu <yuanhan.liu@linux.intel.com>
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 0/3] [RFC] vhost: micro vhost optimization
  2016-05-10 21:49  0% ` Rich Lane
@ 2016-05-10 22:08  0%   ` Yuanhan Liu
  0 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-10 22:08 UTC (permalink / raw)
  To: Rich Lane; +Cc: dev, huawei.xie

On Tue, May 10, 2016 at 02:49:43PM -0700, Rich Lane wrote:
> I see a significant performance improvement with these patches, around 5% at 64
> bytes.

Thanks for testing.

> 
> The one patch that didn't give any performance boost for me was "vhost: arrange
> virtio_net fields for better cache sharing".

Yeah, same here from my test. I mean, in theory, it should give us a
tiny boost, it doesn't in real life though. And since it (should) do
no harm, I would still include this patch in this set. Maybe I should
have noted at first that no real perf gain from the 3rd patch.

	--yliu

> 
> Tested-by: Rich Lane <rich.lane@bigswitch.com>
> 
> On Mon, May 2, 2016 at 5:46 PM, Yuanhan Liu <yuanhan.liu@linux.intel.com>
> 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] pci: Add the class_id support in pci probe
  @ 2016-05-11 15:21  3% ` Stephen Hemminger
  2016-05-11 15:34  3%   ` Richardson, Bruce
  0 siblings, 1 reply; 200+ results
From: Stephen Hemminger @ 2016-05-11 15:21 UTC (permalink / raw)
  To: Ziye Yang; +Cc: dev

On Wed, 11 May 2016 14:08:15 +0800
Ziye Yang <ziye.yang@intel.com> wrote:

> 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>

I like this, and it is necessary but since rte_pci_id is a visible
data structure it causes ABI breakage.

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] pci: Add the class_id support in pci probe
  2016-05-11 15:21  3% ` Stephen Hemminger
@ 2016-05-11 15:34  3%   ` Richardson, Bruce
  0 siblings, 0 replies; 200+ results
From: Richardson, Bruce @ 2016-05-11 15:34 UTC (permalink / raw)
  To: Stephen Hemminger, Yang, Ziye; +Cc: dev



> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Stephen Hemminger
> Sent: Wednesday, May 11, 2016 4:21 PM
> To: Yang, Ziye <ziye.yang@intel.com>
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] pci: Add the class_id support in pci probe
> 
> On Wed, 11 May 2016 14:08:15 +0800
> Ziye Yang <ziye.yang@intel.com> wrote:
> 
> > 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>
> 
> I like this, and it is necessary but since rte_pci_id is a visible data
> structure it causes ABI breakage.

A notice was published for this change in 16.04, so we should be ok ABI-wise.

/Bruce

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring
  2016-05-02 22:25  8% [dpdk-dev] [PATCH 00/16] vhost ABI/API refactoring Yuanhan Liu
  2016-05-02 22:25  2% ` [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications Yuanhan Liu
  2016-05-02 22:25  4% ` [dpdk-dev] [PATCH 14/16] vhost: reserve few more space for future extension Yuanhan Liu
@ 2016-05-13  5:24  8% ` Yuanhan Liu
  2016-05-13  5:25  7%   ` [dpdk-dev] [PATCH v2 11/19] vhost: introduce new API to export queue free entries Yuanhan Liu
                     ` (3 more replies)
  2 siblings, 4 replies; 200+ results
From: Yuanhan Liu @ 2016-05-13  5:24 UTC (permalink / raw)
  To: dev
  Cc: Thomas Monjalon, huawei.xie, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Rich Lane, Yuanhan Liu

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

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.

Note that this refactoring breaks the tep_termination example.
Well, it's just another copy of the original messy vhost example,
and I have no interest to cleanup it again. Therefore, I might
consider to remove that example later, and add the vxlan bits
into vhost example.

Thanks.

	--yliu

---
Ilya Maximets (1):
  vhost: make buf vector for scatter Rx local

Yuanhan Liu (18):
  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
  vhost: per device virtio net header len

 doc/guides/rel_notes/release_16_07.rst        |   9 +
 drivers/net/vhost/rte_eth_vhost.c             |  79 ++++-----
 examples/vhost/main.c                         | 124 +++++++-------
 examples/vhost/main.h                         |   1 +
 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 -----
 17 files changed, 702 insertions(+), 612 deletions(-)
 delete mode 100644 lib/librte_vhost/virtio-net.h

-- 
1.9.0

^ permalink raw reply	[relevance 8%]

* [dpdk-dev] [PATCH v2 11/19] vhost: introduce new API to export queue free entries
  2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
@ 2016-05-13  5:25  7%   ` Yuanhan Liu
  2016-05-13  5:25  3%   ` [dpdk-dev] [PATCH v2 12/19] vhost: remove dependency on priv field Yuanhan Liu
                     ` (2 subsequent siblings)
  3 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-13  5:25 UTC (permalink / raw)
  To: dev
  Cc: Thomas Monjalon, huawei.xie, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Rich Lane, Yuanhan Liu

The new API rte_vhost_avail_entries() is actually a rename of
rte_vring_available_entries(), with the "vring" to "vhost" name
change to keep the consistency of other vhost exported APIs.

This change could let us avoid the dependency of "virtio_net"
struct, to prepare for the ABI refactoring.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 doc/guides/rel_notes/release_16_07.rst |  2 ++
 examples/vhost/main.c                  |  4 ++--
 lib/librte_vhost/rte_vhost_version.map |  1 +
 lib/librte_vhost/rte_virtio_net.h      | 24 +++++++++++++-----------
 lib/librte_vhost/virtio-net.c          | 17 +++++++++++++++++
 5 files changed, 35 insertions(+), 13 deletions(-)

diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index f6d543c..d293eda 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -97,6 +97,8 @@ This section should contain API changes. Sample format:
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
 
+* ``rte_vring_available_entries`` is renamed to ``rte_vhost_avail_entries``.
+
 
 ABI Changes
 -----------
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index cb04585..67ef0ad 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -1055,13 +1055,13 @@ drain_eth_rx(struct vhost_dev *vdev)
 	 * to diminish packet loss.
 	 */
 	if (enable_retry &&
-	    unlikely(rx_count > rte_vring_available_entries(dev,
+	    unlikely(rx_count > rte_vhost_avail_entries(dev->vid,
 			VIRTIO_RXQ))) {
 		uint32_t retry;
 
 		for (retry = 0; retry < burst_rx_retry_num; retry++) {
 			rte_delay_us(burst_rx_delay_time);
-			if (rx_count <= rte_vring_available_entries(dev,
+			if (rx_count <= rte_vhost_avail_entries(dev->vid,
 					VIRTIO_RXQ))
 				break;
 		}
diff --git a/lib/librte_vhost/rte_vhost_version.map b/lib/librte_vhost/rte_vhost_version.map
index 4608e3b..93f1188 100644
--- a/lib/librte_vhost/rte_vhost_version.map
+++ b/lib/librte_vhost/rte_vhost_version.map
@@ -24,6 +24,7 @@ DPDK_2.1 {
 DPDK_16.07 {
 	global:
 
+	rte_vhost_avail_entries;
 	rte_vhost_get_ifname;
 	rte_vhost_get_numa_node;
 	rte_vhost_get_queue_num;
diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
index 0898e8b..0427461 100644
--- a/lib/librte_vhost/rte_virtio_net.h
+++ b/lib/librte_vhost/rte_virtio_net.h
@@ -184,17 +184,6 @@ struct virtio_net_device_ops {
 	int (*vring_state_changed)(struct virtio_net *dev, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
 };
 
-static inline uint16_t __attribute__((always_inline))
-rte_vring_available_entries(struct virtio_net *dev, uint16_t queue_id)
-{
-	struct vhost_virtqueue *vq = dev->virtqueue[queue_id];
-
-	if (!vq->enabled)
-		return 0;
-
-	return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx_res;
-}
-
 /**
  * Function to convert guest physical addresses to vhost virtual addresses.
  * This is used to convert guest virtio buffer addresses.
@@ -285,6 +274,19 @@ uint32_t rte_vhost_get_queue_num(int vid);
 int rte_vhost_get_ifname(int vid, char *buf, size_t len);
 
 /**
+ * Get how many avail entries are left in the queue
+ *
+ * @param vid
+ *  virtio-net device ID
+ * @param queue_id
+ *  virtio queue index
+ *
+ * @return
+ *  num of avail entires left
+ */
+uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
+
+/**
  * This function adds buffers to the virtio devices RX virtqueue. Buffers can
  * be received from the physical port or from another virtual device. A packet
  * count is returned to indicate the number of packets that were succesfully
diff --git a/lib/librte_vhost/virtio-net.c b/lib/librte_vhost/virtio-net.c
index 375c9d4..115eba4 100644
--- a/lib/librte_vhost/virtio-net.c
+++ b/lib/librte_vhost/virtio-net.c
@@ -783,6 +783,23 @@ rte_vhost_get_ifname(int vid, char *buf, size_t len)
 	return 0;
 }
 
+uint16_t
+rte_vhost_avail_entries(int vid, uint16_t queue_id)
+{
+	struct virtio_net *dev;
+	struct vhost_virtqueue *vq;
+
+	dev = get_device(vid);
+	if (!dev)
+		return 0;
+
+	vq = dev->virtqueue[queue_id];
+	if (!vq->enabled)
+		return 0;
+
+	return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx_res;
+}
+
 int rte_vhost_enable_guest_notification(struct virtio_net *dev,
 	uint16_t queue_id, int enable)
 {
-- 
1.9.0

^ permalink raw reply	[relevance 7%]

* [dpdk-dev] [PATCH v2 12/19] vhost: remove dependency on priv field
  2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
  2016-05-13  5:25  7%   ` [dpdk-dev] [PATCH v2 11/19] vhost: introduce new API to export queue free entries Yuanhan Liu
@ 2016-05-13  5:25  3%   ` Yuanhan Liu
  2016-05-13  5:25 12%   ` [dpdk-dev] [PATCH v2 13/19] vhost: export vid as the only interface to applications Yuanhan Liu
  2016-05-13  5:25  4%   ` [dpdk-dev] [PATCH v2 17/19] vhost: reserve few more space for future extension Yuanhan Liu
  3 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-13  5:25 UTC (permalink / raw)
  To: dev
  Cc: Thomas Monjalon, huawei.xie, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Rich Lane, Yuanhan Liu

This change could let us avoid the dependency of "virtio_net"
struct, to prepare for the ABI refactoring.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 drivers/net/vhost/rte_eth_vhost.c | 13 +++++++------
 examples/vhost/main.c             | 18 ++++++++++++++++--
 2 files changed, 23 insertions(+), 8 deletions(-)

diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c
index 6fa9f6b..de0f25e 100644
--- a/drivers/net/vhost/rte_eth_vhost.c
+++ b/drivers/net/vhost/rte_eth_vhost.c
@@ -275,7 +275,6 @@ new_device(struct virtio_net *dev)
 	for (i = 0; i < rte_vhost_get_queue_num(dev->vid) * VIRTIO_QNUM; i++)
 		rte_vhost_enable_guest_notification(dev, i, 0);
 
-	dev->priv = eth_dev;
 	eth_dev->data->dev_link.link_status = ETH_LINK_UP;
 
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
@@ -303,6 +302,8 @@ destroy_device(volatile struct virtio_net *dev)
 {
 	struct rte_eth_dev *eth_dev;
 	struct vhost_queue *vq;
+	struct internal_list *list;
+	char ifname[PATH_MAX];
 	unsigned i;
 
 	if (dev == NULL) {
@@ -310,11 +311,13 @@ destroy_device(volatile struct virtio_net *dev)
 		return;
 	}
 
-	eth_dev = (struct rte_eth_dev *)dev->priv;
-	if (eth_dev == NULL) {
-		RTE_LOG(INFO, PMD, "Failed to find a ethdev\n");
+	rte_vhost_get_ifname(dev->vid, ifname, sizeof(ifname));
+	list = find_internal_resource(ifname);
+	if (list == NULL) {
+		RTE_LOG(ERR, PMD, "Invalid interface name: %s\n", ifname);
 		return;
 	}
+	eth_dev = list->eth_dev;
 
 	/* Wait until rx/tx_pkt_burst stops accessing vhost device */
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
@@ -336,8 +339,6 @@ destroy_device(volatile struct virtio_net *dev)
 
 	eth_dev->data->dev_link.link_status = ETH_LINK_DOWN;
 
-	dev->priv = NULL;
-
 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
 		vq = eth_dev->data->rx_queues[i];
 		if (vq == NULL)
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index 67ef0ad..c408577 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -700,6 +700,19 @@ find_vhost_dev(struct ether_addr *mac)
 	return NULL;
 }
 
+static inline struct vhost_dev *__attribute__((always_inline))
+find_vhost_dev_by_vid(int vid)
+{
+	struct vhost_dev *vdev;
+
+	TAILQ_FOREACH(vdev, &vhost_dev_list, next) {
+		if (vdev->ready == DEVICE_RX && vdev->vid == vid)
+			return vdev;
+	}
+
+	return NULL;
+}
+
 /*
  * This function learns the MAC address of the device and registers this along with a
  * vlan tag to a VMDQ.
@@ -1175,7 +1188,9 @@ destroy_device (volatile struct virtio_net *dev)
 	struct vhost_dev *vdev;
 	int lcore;
 
-	vdev = (struct vhost_dev *)dev->priv;
+	vdev = find_vhost_dev_by_vid(dev->vid);
+	if (!vdev)
+		return;
 	/*set the remove flag. */
 	vdev->remove = 1;
 	while(vdev->ready != DEVICE_SAFE_REMOVE) {
@@ -1228,7 +1243,6 @@ new_device (struct virtio_net *dev)
 		return -1;
 	}
 	vdev->dev = dev;
-	dev->priv = vdev;
 	vdev->vid = vid;
 
 	TAILQ_INSERT_TAIL(&vhost_dev_list, vdev, next);
-- 
1.9.0

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH v2 13/19] vhost: export vid as the only interface to applications
  2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
  2016-05-13  5:25  7%   ` [dpdk-dev] [PATCH v2 11/19] vhost: introduce new API to export queue free entries Yuanhan Liu
  2016-05-13  5:25  3%   ` [dpdk-dev] [PATCH v2 12/19] vhost: remove dependency on priv field Yuanhan Liu
@ 2016-05-13  5:25 12%   ` Yuanhan Liu
  2016-05-13  5:25  4%   ` [dpdk-dev] [PATCH v2 17/19] vhost: reserve few more space for future extension Yuanhan Liu
  3 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-13  5:25 UTC (permalink / raw)
  To: dev
  Cc: Thomas Monjalon, huawei.xie, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Rich Lane, Yuanhan Liu

With all the previous prepare works, we are just one step away from
the final ABI refactoring. That is, to change current API to let them
stick to vid instead of the old virtio_net dev.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---

v2: update release note
---
 doc/guides/rel_notes/release_16_07.rst        |  7 ++++
 drivers/net/vhost/rte_eth_vhost.c             | 47 +++++++++------------------
 examples/vhost/main.c                         | 25 +++++++-------
 lib/librte_vhost/rte_virtio_net.h             | 18 +++++-----
 lib/librte_vhost/vhost_rxtx.c                 | 15 +++++++--
 lib/librte_vhost/vhost_user/virtio-net-user.c | 14 ++++----
 lib/librte_vhost/virtio-net.c                 | 17 ++++++----
 7 files changed, 75 insertions(+), 68 deletions(-)

diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index d293eda..ebc507b 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -99,6 +99,10 @@ This section should contain API changes. Sample format:
 
 * ``rte_vring_available_entries`` is renamed to ``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".
+
 
 ABI Changes
 -----------
@@ -110,6 +114,9 @@ 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.
 
+* 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.
 
 Shared Library Versions
 -----------------------
diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c
index de0f25e..56c1c36 100644
--- a/drivers/net/vhost/rte_eth_vhost.c
+++ b/drivers/net/vhost/rte_eth_vhost.c
@@ -71,9 +71,9 @@ static struct ether_addr base_eth_addr = {
 };
 
 struct vhost_queue {
+	int vid;
 	rte_atomic32_t allow_queuing;
 	rte_atomic32_t while_queuing;
-	struct virtio_net *device;
 	struct pmd_internal *internal;
 	struct rte_mempool *mb_pool;
 	uint8_t port;
@@ -139,7 +139,7 @@ eth_vhost_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 		goto out;
 
 	/* Dequeue packets from guest TX queue */
-	nb_rx = rte_vhost_dequeue_burst(r->device,
+	nb_rx = rte_vhost_dequeue_burst(r->vid,
 			r->virtqueue_id, r->mb_pool, bufs, nb_bufs);
 
 	r->rx_pkts += nb_rx;
@@ -170,7 +170,7 @@ eth_vhost_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 		goto out;
 
 	/* Enqueue packets to guest RX queue */
-	nb_tx = rte_vhost_enqueue_burst(r->device,
+	nb_tx = rte_vhost_enqueue_burst(r->vid,
 			r->virtqueue_id, bufs, nb_bufs);
 
 	r->tx_pkts += nb_tx;
@@ -222,7 +222,7 @@ find_internal_resource(char *ifname)
 }
 
 static int
-new_device(struct virtio_net *dev)
+new_device(int vid)
 {
 	struct rte_eth_dev *eth_dev;
 	struct internal_list *list;
@@ -234,12 +234,7 @@ new_device(struct virtio_net *dev)
 	int newnode;
 #endif
 
-	if (dev == NULL) {
-		RTE_LOG(INFO, PMD, "Invalid argument\n");
-		return -1;
-	}
-
-	rte_vhost_get_ifname(dev->vid, ifname, sizeof(ifname));
+	rte_vhost_get_ifname(vid, ifname, sizeof(ifname));
 	list = find_internal_resource(ifname);
 	if (list == NULL) {
 		RTE_LOG(INFO, PMD, "Invalid device name: %s\n", ifname);
@@ -250,7 +245,7 @@ new_device(struct virtio_net *dev)
 	internal = eth_dev->data->dev_private;
 
 #ifdef RTE_LIBRTE_VHOST_NUMA
-	newnode = rte_vhost_get_numa_node(dev->vid);
+	newnode = rte_vhost_get_numa_node(vid);
 	if (newnode > 0)
 		eth_dev->data->numa_node = newnode;
 #endif
@@ -259,7 +254,7 @@ new_device(struct virtio_net *dev)
 		vq = eth_dev->data->rx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = dev;
+		vq->vid = vid;
 		vq->internal = internal;
 		vq->port = eth_dev->data->port_id;
 	}
@@ -267,13 +262,13 @@ new_device(struct virtio_net *dev)
 		vq = eth_dev->data->tx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = dev;
+		vq->vid = vid;
 		vq->internal = internal;
 		vq->port = eth_dev->data->port_id;
 	}
 
-	for (i = 0; i < rte_vhost_get_queue_num(dev->vid) * VIRTIO_QNUM; i++)
-		rte_vhost_enable_guest_notification(dev, i, 0);
+	for (i = 0; i < rte_vhost_get_queue_num(vid) * VIRTIO_QNUM; i++)
+		rte_vhost_enable_guest_notification(vid, i, 0);
 
 	eth_dev->data->dev_link.link_status = ETH_LINK_UP;
 
@@ -298,7 +293,7 @@ new_device(struct virtio_net *dev)
 }
 
 static void
-destroy_device(volatile struct virtio_net *dev)
+destroy_device(int vid)
 {
 	struct rte_eth_dev *eth_dev;
 	struct vhost_queue *vq;
@@ -306,12 +301,7 @@ destroy_device(volatile struct virtio_net *dev)
 	char ifname[PATH_MAX];
 	unsigned i;
 
-	if (dev == NULL) {
-		RTE_LOG(INFO, PMD, "Invalid argument\n");
-		return;
-	}
-
-	rte_vhost_get_ifname(dev->vid, ifname, sizeof(ifname));
+	rte_vhost_get_ifname(vid, ifname, sizeof(ifname));
 	list = find_internal_resource(ifname);
 	if (list == NULL) {
 		RTE_LOG(ERR, PMD, "Invalid interface name: %s\n", ifname);
@@ -343,13 +333,13 @@ destroy_device(volatile struct virtio_net *dev)
 		vq = eth_dev->data->rx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = NULL;
+		vq->vid = -1;
 	}
 	for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
 		vq = eth_dev->data->tx_queues[i];
 		if (vq == NULL)
 			continue;
-		vq->device = NULL;
+		vq->vid = -1;
 	}
 
 	RTE_LOG(INFO, PMD, "Connection closed\n");
@@ -358,19 +348,14 @@ destroy_device(volatile struct virtio_net *dev)
 }
 
 static int
-vring_state_changed(struct virtio_net *dev, uint16_t vring, int enable)
+vring_state_changed(int vid, uint16_t vring, int enable)
 {
 	struct rte_vhost_vring_state *state;
 	struct rte_eth_dev *eth_dev;
 	struct internal_list *list;
 	char ifname[PATH_MAX];
 
-	if (dev == NULL) {
-		RTE_LOG(ERR, PMD, "Invalid argument\n");
-		return -1;
-	}
-
-	rte_vhost_get_ifname(dev->vid, ifname, sizeof(ifname));
+	rte_vhost_get_ifname(vid, ifname, sizeof(ifname));
 	list = find_internal_resource(ifname);
 	if (list == NULL) {
 		RTE_LOG(ERR, PMD, "Invalid interface name: %s\n", ifname);
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index c408577..f3a6277 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -808,7 +808,7 @@ virtio_xmit(struct vhost_dev *dst_vdev, struct vhost_dev *src_vdev,
 {
 	uint16_t ret;
 
-	ret = rte_vhost_enqueue_burst(dst_vdev->dev, VIRTIO_RXQ, &m, 1);
+	ret = rte_vhost_enqueue_burst(dst_vdev->vid, VIRTIO_RXQ, &m, 1);
 	if (enable_stats) {
 		rte_atomic64_inc(&dst_vdev->stats.rx_total_atomic);
 		rte_atomic64_add(&dst_vdev->stats.rx_atomic, ret);
@@ -1054,7 +1054,6 @@ static inline void __attribute__((always_inline))
 drain_eth_rx(struct vhost_dev *vdev)
 {
 	uint16_t rx_count, enqueue_count;
-	struct virtio_net *dev = vdev->dev;
 	struct rte_mbuf *pkts[MAX_PKT_BURST];
 
 	rx_count = rte_eth_rx_burst(ports[0], vdev->vmdq_rx_q,
@@ -1068,19 +1067,19 @@ drain_eth_rx(struct vhost_dev *vdev)
 	 * to diminish packet loss.
 	 */
 	if (enable_retry &&
-	    unlikely(rx_count > rte_vhost_avail_entries(dev->vid,
+	    unlikely(rx_count > rte_vhost_avail_entries(vdev->vid,
 			VIRTIO_RXQ))) {
 		uint32_t retry;
 
 		for (retry = 0; retry < burst_rx_retry_num; retry++) {
 			rte_delay_us(burst_rx_delay_time);
-			if (rx_count <= rte_vhost_avail_entries(dev->vid,
+			if (rx_count <= rte_vhost_avail_entries(vdev->vid,
 					VIRTIO_RXQ))
 				break;
 		}
 	}
 
-	enqueue_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ,
+	enqueue_count = rte_vhost_enqueue_burst(vdev->vid, VIRTIO_RXQ,
 						pkts, rx_count);
 	if (enable_stats) {
 		rte_atomic64_add(&vdev->stats.rx_total_atomic, rx_count);
@@ -1097,7 +1096,7 @@ drain_virtio_tx(struct vhost_dev *vdev)
 	uint16_t count;
 	uint16_t i;
 
-	count = rte_vhost_dequeue_burst(vdev->dev, VIRTIO_TXQ, mbuf_pool,
+	count = rte_vhost_dequeue_burst(vdev->vid, VIRTIO_TXQ, mbuf_pool,
 					pkts, MAX_PKT_BURST);
 
 	/* setup VMDq for the first packet */
@@ -1183,12 +1182,12 @@ switch_worker(void *arg __rte_unused)
  * of dev->remove=1 which can cause an infinite loop in the rte_pause loop.
  */
 static void
-destroy_device (volatile struct virtio_net *dev)
+destroy_device(int vid)
 {
 	struct vhost_dev *vdev;
 	int lcore;
 
-	vdev = find_vhost_dev_by_vid(dev->vid);
+	vdev = find_vhost_dev_by_vid(vid);
 	if (!vdev)
 		return;
 	/*set the remove flag. */
@@ -1228,12 +1227,11 @@ destroy_device (volatile struct virtio_net *dev)
  * and the allocated to a specific data core.
  */
 static int
-new_device (struct virtio_net *dev)
+new_device(int vid)
 {
 	int lcore, core_add = 0;
 	uint32_t device_num_min = num_devices;
 	struct vhost_dev *vdev;
-	int vid = dev->vid;
 
 	vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
 	if (vdev == NULL) {
@@ -1242,7 +1240,6 @@ new_device (struct virtio_net *dev)
 			vid);
 		return -1;
 	}
-	vdev->dev = dev;
 	vdev->vid = vid;
 
 	TAILQ_INSERT_TAIL(&vhost_dev_list, vdev, next);
@@ -1265,8 +1262,8 @@ new_device (struct virtio_net *dev)
 	lcore_info[vdev->coreid].device_num++;
 
 	/* Disable notifications. */
-	rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
-	rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
+	rte_vhost_enable_guest_notification(vid, VIRTIO_RXQ, 0);
+	rte_vhost_enable_guest_notification(vid, VIRTIO_TXQ, 0);
 
 	RTE_LOG(INFO, VHOST_DATA,
 		"(%d) device has been added to data core %d\n",
@@ -1322,7 +1319,7 @@ print_stats(void)
 				"RX total:              %" PRIu64 "\n"
 				"RX dropped:            %" PRIu64 "\n"
 				"RX successful:         %" PRIu64 "\n",
-				vdev->dev->vid,
+				vdev->vid,
 				tx_total, tx_dropped, tx,
 				rx_total, rx_dropped, rx);
 		}
diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
index 0427461..370345e 100644
--- a/lib/librte_vhost/rte_virtio_net.h
+++ b/lib/librte_vhost/rte_virtio_net.h
@@ -178,10 +178,10 @@ struct virtio_memory {
  *
  */
 struct virtio_net_device_ops {
-	int (*new_device)(struct virtio_net *);	/**< Add device. */
-	void (*destroy_device)(volatile struct virtio_net *);	/**< Remove device. */
+	int (*new_device)(int vid);		/**< Add device. */
+	void (*destroy_device)(int vid);	/**< Remove device. */
 
-	int (*vring_state_changed)(struct virtio_net *dev, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
+	int (*vring_state_changed)(int vid, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
 };
 
 /**
@@ -220,7 +220,7 @@ int rte_vhost_feature_enable(uint64_t feature_mask);
 /* Returns currently supported vhost features */
 uint64_t rte_vhost_feature_get(void);
 
-int rte_vhost_enable_guest_notification(struct virtio_net *dev, uint16_t queue_id, int enable);
+int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
 
 /* Register vhost driver. dev_name could be different for multiple instance support. */
 int rte_vhost_driver_register(const char *dev_name);
@@ -291,8 +291,8 @@ uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
  * be received from the physical port or from another virtual device. A packet
  * count is returned to indicate the number of packets that were succesfully
  * added to the RX queue.
- * @param dev
- *  virtio-net device
+ * @param vid
+ *  virtio-net device ID
  * @param queue_id
  *  virtio queue index in mq case
  * @param pkts
@@ -302,14 +302,14 @@ uint16_t rte_vhost_avail_entries(int vid, uint16_t queue_id);
  * @return
  *  num of packets enqueued
  */
-uint16_t rte_vhost_enqueue_burst(struct virtio_net *dev, uint16_t queue_id,
+uint16_t rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
 	struct rte_mbuf **pkts, uint16_t count);
 
 /**
  * This function gets guest buffers from the virtio device TX virtqueue,
  * construct host mbufs, copies guest buffer content to host mbufs and
  * store them in pkts to be processed.
- * @param dev
+ * @param vid
  *  virtio-net device
  * @param queue_id
  *  virtio queue index in mq case
@@ -322,7 +322,7 @@ uint16_t rte_vhost_enqueue_burst(struct virtio_net *dev, uint16_t queue_id,
  * @return
  *  num of packets dequeued
  */
-uint16_t rte_vhost_dequeue_burst(struct virtio_net *dev, uint16_t queue_id,
+uint16_t rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count);
 
 #endif /* _VIRTIO_NET_H_ */
diff --git a/lib/librte_vhost/vhost_rxtx.c b/lib/librte_vhost/vhost_rxtx.c
index 8d87508..08cab08 100644
--- a/lib/librte_vhost/vhost_rxtx.c
+++ b/lib/librte_vhost/vhost_rxtx.c
@@ -46,6 +46,7 @@
 #include <rte_arp.h>
 
 #include "vhost-net.h"
+#include "virtio-net.h"
 
 #define MAX_PKT_BURST 32
 #define VHOST_LOG_PAGE	4096
@@ -587,9 +588,14 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id,
 }
 
 uint16_t
-rte_vhost_enqueue_burst(struct virtio_net *dev, uint16_t queue_id,
+rte_vhost_enqueue_burst(int vid, uint16_t queue_id,
 	struct rte_mbuf **pkts, uint16_t count)
 {
+	struct virtio_net *dev = get_device(vid);
+
+	if (!dev)
+		return 0;
+
 	if (dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF))
 		return virtio_dev_merge_rx(dev, queue_id, pkts, count);
 	else
@@ -815,9 +821,10 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 }
 
 uint16_t
-rte_vhost_dequeue_burst(struct virtio_net *dev, uint16_t queue_id,
+rte_vhost_dequeue_burst(int vid, uint16_t queue_id,
 	struct rte_mempool *mbuf_pool, struct rte_mbuf **pkts, uint16_t count)
 {
+	struct virtio_net *dev;
 	struct rte_mbuf *rarp_mbuf = NULL;
 	struct vhost_virtqueue *vq;
 	uint32_t desc_indexes[MAX_PKT_BURST];
@@ -826,6 +833,10 @@ rte_vhost_dequeue_burst(struct virtio_net *dev, uint16_t queue_id,
 	uint16_t free_entries;
 	uint16_t avail_idx;
 
+	dev = get_device(vid);
+	if (!dev)
+		return 0;
+
 	if (unlikely(!is_valid_virt_queue_idx(queue_id, 1, dev->virt_qp_nb))) {
 		RTE_LOG(ERR, VHOST_DATA, "(%d) %s: invalid virtqueue idx %d.\n",
 			dev->vid, __func__, queue_id);
diff --git a/lib/librte_vhost/vhost_user/virtio-net-user.c b/lib/librte_vhost/vhost_user/virtio-net-user.c
index 9385af1..7fa69a7 100644
--- a/lib/librte_vhost/vhost_user/virtio-net-user.c
+++ b/lib/librte_vhost/vhost_user/virtio-net-user.c
@@ -117,7 +117,7 @@ user_set_mem_table(int vid, struct VhostUserMsg *pmsg)
 	/* Remove from the data plane. */
 	if (dev->flags & VIRTIO_DEV_RUNNING) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	if (dev->mem) {
@@ -279,6 +279,9 @@ user_set_vring_kick(int vid, struct VhostUserMsg *pmsg)
 	struct vhost_vring_file file;
 	struct virtio_net *dev = get_device(vid);
 
+	if (!dev)
+		return;
+
 	file.index = pmsg->payload.u64 & VHOST_USER_VRING_IDX_MASK;
 	if (pmsg->payload.u64 & VHOST_USER_VRING_NOFD_MASK)
 		file.fd = VIRTIO_INVALID_EVENTFD;
@@ -289,7 +292,7 @@ user_set_vring_kick(int vid, struct VhostUserMsg *pmsg)
 	vhost_set_vring_kick(vid, &file);
 
 	if (virtio_is_ready(dev) && !(dev->flags & VIRTIO_DEV_RUNNING)) {
-		if (notify_ops->new_device(dev) == 0)
+		if (notify_ops->new_device(vid) == 0)
 			dev->flags |= VIRTIO_DEV_RUNNING;
 	}
 }
@@ -307,7 +310,7 @@ user_get_vring_base(int vid,
 		return -1;
 	/* We have to stop the queue (virtio) if it is running. */
 	if (dev->flags & VIRTIO_DEV_RUNNING)
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 
 	/* Here we are safe to get the last used index */
 	vhost_get_vring_base(vid, state->index, state);
@@ -342,9 +345,8 @@ user_set_vring_enable(int vid,
 		"set queue enable: %d to qp idx: %d\n",
 		enable, state->index);
 
-	if (notify_ops->vring_state_changed) {
-		notify_ops->vring_state_changed(dev, state->index, enable);
-	}
+	if (notify_ops->vring_state_changed)
+		notify_ops->vring_state_changed(vid, state->index, enable);
 
 	dev->virtqueue[state->index]->enabled = enable;
 
diff --git a/lib/librte_vhost/virtio-net.c b/lib/librte_vhost/virtio-net.c
index 115eba4..ea216c0 100644
--- a/lib/librte_vhost/virtio-net.c
+++ b/lib/librte_vhost/virtio-net.c
@@ -296,7 +296,7 @@ vhost_destroy_device(int vid)
 
 	if (dev->flags & VIRTIO_DEV_RUNNING) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	cleanup_device(dev, 1);
@@ -354,7 +354,7 @@ vhost_reset_owner(int vid)
 
 	if (dev->flags & VIRTIO_DEV_RUNNING) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	cleanup_device(dev, 0);
@@ -718,13 +718,13 @@ vhost_set_backend(int vid, struct vhost_vring_file *file)
 	if (!(dev->flags & VIRTIO_DEV_RUNNING)) {
 		if (dev->virtqueue[VIRTIO_TXQ]->backend != VIRTIO_DEV_STOPPED &&
 		    dev->virtqueue[VIRTIO_RXQ]->backend != VIRTIO_DEV_STOPPED) {
-			if (notify_ops->new_device(dev) < 0)
+			if (notify_ops->new_device(vid) < 0)
 				return -1;
 			dev->flags |= VIRTIO_DEV_RUNNING;
 		}
 	} else if (file->fd == VIRTIO_DEV_STOPPED) {
 		dev->flags &= ~VIRTIO_DEV_RUNNING;
-		notify_ops->destroy_device(dev);
+		notify_ops->destroy_device(vid);
 	}
 
 	return 0;
@@ -800,9 +800,14 @@ rte_vhost_avail_entries(int vid, uint16_t queue_id)
 	return *(volatile uint16_t *)&vq->avail->idx - vq->last_used_idx_res;
 }
 
-int rte_vhost_enable_guest_notification(struct virtio_net *dev,
-	uint16_t queue_id, int enable)
+int
+rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable)
 {
+	struct virtio_net *dev = get_device(vid);
+
+	if (dev == NULL)
+		return -1;
+
 	if (enable) {
 		RTE_LOG(ERR, VHOST_CONFIG,
 			"guest notification isn't supported.\n");
-- 
1.9.0

^ permalink raw reply	[relevance 12%]

* [dpdk-dev] [PATCH v2 17/19] vhost: reserve few more space for future extension
  2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
                     ` (2 preceding siblings ...)
  2016-05-13  5:25 12%   ` [dpdk-dev] [PATCH v2 13/19] vhost: export vid as the only interface to applications Yuanhan Liu
@ 2016-05-13  5:25  4%   ` Yuanhan Liu
  3 siblings, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-13  5:25 UTC (permalink / raw)
  To: dev
  Cc: Thomas Monjalon, huawei.xie, Panu Matilainen, Tetsuya Mukawa,
	Traynor Kevin, Rich Lane, Yuanhan Liu

"virtio_net_device_ops" is the only left open struct that an application
can access, therefore, it's the only place that might introduce potential
ABI break in future for extension.

So, do some reservation for it. 5 should be pretty enough, considering
that we have barely touched it for a long while. Another reason to
choose 5 is for cache alignment: 5 makes the struct 64 bytes for 64 bit
machine.

With this, it's confidence to say that we might be able to be free from
the ABI violation forever.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
---
 lib/librte_vhost/rte_virtio_net.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/lib/librte_vhost/rte_virtio_net.h b/lib/librte_vhost/rte_virtio_net.h
index fc1d799..bc2b74b 100644
--- a/lib/librte_vhost/rte_virtio_net.h
+++ b/lib/librte_vhost/rte_virtio_net.h
@@ -66,6 +66,8 @@ struct virtio_net_device_ops {
 	void (*destroy_device)(int vid);	/**< Remove device. */
 
 	int (*vring_state_changed)(int vid, uint16_t queue_id, int enable);	/**< triggered when a vring is enabled or disabled */
+
+	void *reserved[5]; /**< Reserved for future extension */
 };
 
 /**
-- 
1.9.0

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH v2 0/6] vhost: add vhost-user client mode and reconnect ability
  2016-05-07  6:40  3% [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability Yuanhan Liu
  2016-05-10  3:23  3% ` Xu, Qian Q
@ 2016-05-13  6:16  3% ` Yuanhan Liu
  1 sibling, 0 replies; 200+ results
From: Yuanhan Liu @ 2016-05-13  6:16 UTC (permalink / raw)
  To: dev; +Cc: huawei.xie, Traynor Kevin, marcandre.lureau, Yuanhan Liu

v2: - added release doc
    - do not remove socket file for the client mode
    - create one thread ony to handle all reconnects

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: DPDK vhost-user as the client,
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

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 and reconnect option
  vhost: add pmd client and reconnect option

 doc/guides/rel_notes/release_16_07.rst       |  23 ++
 drivers/net/vhost/rte_eth_vhost.c            |  54 +++-
 examples/vhost/main.c                        |  23 +-
 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, 390 insertions(+), 148 deletions(-)

-- 
1.9.0

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] sched: fix useless call
  @ 2016-05-13 11:04  4%       ` Dumitrescu, Cristian
  0 siblings, 0 replies; 200+ results
From: Dumitrescu, Cristian @ 2016-05-13 11:04 UTC (permalink / raw)
  To: Thomas Monjalon; +Cc: dev, Yigit, Ferruh, Mrzyglod, DanielX T



> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Friday, May 13, 2016 11:12 AM
> To: Dumitrescu, Cristian <cristian.dumitrescu@intel.com>
> Cc: dev@dpdk.org; Yigit, Ferruh <ferruh.yigit@intel.com>; Mrzyglod, DanielX
> T <danielx.t.mrzyglod@intel.com>
> Subject: Re: [dpdk-dev] [PATCH] sched: fix useless call
> 
> 2016-05-11 10:46, Ferruh Yigit:
> > On 5/10/2016 6:18 PM, Dumitrescu, Cristian wrote:
> > > As previously discussed on this email list, the rte_bitmap_free() is an API
> function that works as a placeholder for any resource freeing that needs to
> be done for the bitmap. The API function should not be removed and also
> the call to this function from the rte_sched_port_free() should not be
> removed either.
> > >
> >
> > Right now it isn't required and doesn't do anything.
> > Why not add this function when it is required?
> 
> I don't understand why we keep a function which does nothing.

Every data type/class/object should have a constructor/create and destructor/free function. This is standard programming practice, right?

This API function is the free function for the bitmap object. Right now there are no internally allocated resources to be freed, but as code evolves, some other internal memory could be allocated by the bitmap, which needs to be freed in the bitmap free function.

This function should be kept in order to have a stable API. We should not go back and forth with adding / removing API functions as code evolves. It does not make any sense to go through the ABI change process to remove this API function now just to come back later on and go again through ABI change to add back this API function later.

I think each DPDK object should have its create and free functions clearly identified in the API.

^ permalink raw reply	[relevance 4%]

* [dpdk-dev] [PATCH] doc: move rel_notes instructions as comments
@ 2016-05-13 13:27 16% Olivier Matz
  0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-05-13 13:27 UTC (permalink / raw)
  To: dev; +Cc: john.mcnamara

We don't want to have this instructions in the generated docs, so use
comments. It's also less confusing for people adding entries in the
documentation.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
 doc/guides/rel_notes/release_16_07.rst | 86 +++++++++++++++++-----------------
 1 file changed, 43 insertions(+), 43 deletions(-)

diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index f6d543c..71d3540 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -1,50 +1,50 @@
 DPDK Release 16.07
 ==================
 
-**Read this first.**
+.. **Read this first.**
 
-The text below explains how to update the release notes.
+   The text below explains how to update the release notes.
 
-Use proper spelling, capitalization and punctuation in all sections.
+   Use proper spelling, capitalization and punctuation in all sections.
 
-Variable and config names should be quoted as fixed width text: ``LIKE_THIS``.
+   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::
+   Build the docs and view the output file to ensure the changes are correct::
 
-   make doc-guides-html
+      make doc-guides-html
 
-   firefox build/doc/html/guides/rel_notes/release_16_07.html
+      firefox build/doc/html/guides/rel_notes/release_16_07.html
 
 
 New Features
 ------------
 
-This section should contain new features added in this release. Sample format:
+.. 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 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.
+     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.
+     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.
+     * Added feature foo to do something.
+     * Enhanced feature bar to do something else.
 
-  Refer to the previous release notes for examples.
+     Refer to the previous release notes for examples.
 
 
 Resolved Issues
 ---------------
 
-This section should contain bug fixes added to the relevant sections. Sample format:
+.. 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.**
+   * **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.
+     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
@@ -77,21 +77,21 @@ Other
 Known Issues
 ------------
 
-This section should contain new known issues in this release. Sample format:
+.. This section should contain new known issues in this release. Sample format:
 
-* **Add title in present tense with full stop.**
+   * **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.
+     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:
+.. 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.
+   * 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,
@@ -101,9 +101,9 @@ This section should contain API changes. Sample format:
 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.
+.. * 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.
 
 * 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.
@@ -112,7 +112,7 @@ ABI Changes
 Shared Library Versions
 -----------------------
 
-Update any library version updated in this release and prepend with a ``+`` sign.
+.. 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.
 
@@ -150,25 +150,25 @@ The libraries prepended with a plus sign were incremented in this version.
 Tested Platforms
 ----------------
 
-This section should contain a list of platforms that were tested with this
-release.
+.. This section should contain a list of platforms that were tested with this
+   release.
 
-The format is:
+   The format is:
 
-#. Platform name.
+   #. Platform name.
 
-   - Platform details.
-   - Platform details.
+      - Platform details.
+      - Platform details.
 
 
 Tested NICs
 -----------
 
-This section should contain a list of NICs that were tested with this release.
+.. This section should contain a list of NICs that were tested with this release.
 
-The format is:
+   The format is:
 
-#. NIC name.
+   #. NIC name.
 
-   - NIC details.
-   - NIC details.
+      - NIC details.
+      - NIC details.
-- 
2.8.0.rc3

^ permalink raw reply	[relevance 16%]

* Re: [dpdk-dev] Ring PMD: why are stats counters atomic?
  @ 2016-05-16 13:16  3%     ` Bruce Richardson
  0 siblings, 0 replies; 200+ results
From: Bruce Richardson @ 2016-05-16 13:16 UTC (permalink / raw)
  To: Mauricio Vásquez; +Cc: dev

On Mon, May 16, 2016 at 03:12:10PM +0200, Mauricio Vásquez wrote:
> Hello Bruce,
> 
> Although having this support does not harm anyone, I am not convinced that
> it is useful, mainly because there exists the single-thread limitation in
> other PMDs. Then, if an application has to use different kind of NICs (i.e,
> different PMDs) it has to implement the locking strategies. On the other
> hand, if an application  only uses rte_rings, it could just use the
> rte_ring library.
> 
> Thanks, Mauricio V
> 
I agree. 
If you want, please submit a patch to remove this behaviour and see 
if anyone objects to it. If there are no objections, I have no problem accepting
the patch.

However, since this is a behaviour change to existing functionality, we may
need to implement function versionning for this for ABI compatibility. Please
take that into account when drafting any patch.

Regards,
/Bruce

> On Tue, May 10, 2016 at 11:36 AM, Bruce Richardson <
> bruce.richardson@intel.com> wrote:
> 
> > On Tue, May 10, 2016 at 11:13:08AM +0200, Mauricio Vásquez wrote:
> > > Hello,
> > >
> > > Per-queue stats counters are defined as rte_atomic64_t, in the tx/rx
> > > functions, they are atomically increased if the rings have the multiple
> > > consumers/producer flag enabled.
> > >
> > > According to the design principles, the application should not invoke
> > those
> > > functions on the same queue on different cores, then I think that atomic
> > > increasing is not necessary.
> > >
> > > Is there something wrong with my reasoning?, If not, I am willing to
> > send a
> > > patch.
> > >
> > > Thank you very much,
> > >
> > Since the rte_rings, on which the ring pmd is obviously based, have
> > multi-producer
> > and multi-consumer support built-in, I thought it might be useful in the
> > ring
> > PMD itself to allow multiple threads to access the ring queues at the same
> > time,
> > if the underlying rings are marked as MP/MC safe. When doing enqueues and
> > dequeue
> > from the ring, the stats are either incremented atomically, or
> > non-atomically,
> > depending on the underlying queue type.
> >
> >         const uint16_t nb_rx = (uint16_t)rte_ring_dequeue_burst(r->rng,
> >                         ptrs, nb_bufs);
> >         if (r->rng->flags & RING_F_SC_DEQ)
> >                 r->rx_pkts.cnt += nb_rx;
> >         else
> >                 rte_atomic64_add(&(r->rx_pkts), nb_rx);
> >
> > If people don't think this behaviour is worthwhile keeping, I'm ok with
> > removing
> > it, since all other PMDs have the restriction that the queues are
> > single-thread
> > only.
> >
> > Regards,
> > /Bruce
> >

^ permalink raw reply	[relevance 3%]

* [dpdk-dev] [PATCH 1/2] doc: announce ABI change of struct rte_port_source_params
  2016-05-16 13:18  9% [dpdk-dev] [PATCH 0/2] doc: announce ABI change of struct rte_port_source_params Fan Zhang
@ 2016-05-16 13:18 18% ` Fan Zhang
  2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 2/2] doc: announce ABI change of struct rte_port_sink_params Fan Zhang
  1 sibling, 0 replies; 200+ results
From: Fan Zhang @ 2016-05-16 13:18 UTC (permalink / raw)
  To: dev

The ABI changes are planned for rte_port_source_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 | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index fffe9c7..d228bae 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -74,3 +74,7 @@ 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.
-- 
2.5.5

^ permalink raw reply	[relevance 18%]

* [dpdk-dev] [PATCH 0/2] doc: announce ABI change of struct rte_port_source_params
@ 2016-05-16 13:18  9% Fan Zhang
  2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 1/2] " Fan Zhang
  2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 2/2] doc: announce ABI change of struct rte_port_sink_params Fan Zhang
  0 siblings, 2 replies; 200+ results
From: Fan Zhang @ 2016-05-16 13:18 UTC (permalink / raw)
  To: dev

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.

Fan Zhang (2):
  doc: announce ABI change of struct rte_port_source_params
  doc: announce ABI change of struct rte_port_sink_params

 doc/guides/rel_notes/deprecation.rst | 8 ++++++++
 1 file changed, 8 insertions(+)

-- 
2.5.5

^ permalink raw reply	[relevance 9%]

* [dpdk-dev] [PATCH 2/2] doc: announce ABI change of struct rte_port_sink_params
  2016-05-16 13:18  9% [dpdk-dev] [PATCH 0/2] doc: announce ABI change of struct rte_port_source_params Fan Zhang
  2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 1/2] " Fan Zhang
@ 2016-05-16 13:18 18% ` Fan Zhang
  2016-05-16 13:57  9%   ` Panu Matilainen
  1 sibling, 1 reply; 200+ results
From: Fan Zhang @ 2016-05-16 13:18 UTC (permalink / raw)
  To: dev

The ABI changes are planned for 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 | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index d228bae..d2f7306 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -78,3 +78,7 @@ Deprecation Notices
 * 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 18%]

* Re: [dpdk-dev] [PATCH 2/2] doc: announce ABI change of struct rte_port_sink_params
  2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 2/2] doc: announce ABI change of struct rte_port_sink_params Fan Zhang
@ 2016-05-16 13:57  9%   ` Panu Matilainen
  0 siblings, 0 replies; 200+ results
From: Panu Matilainen @ 2016-05-16 13:57 UTC (permalink / raw)
  To: Fan Zhang, dev

On 05/16/2016 04:18 PM, Fan Zhang wrote:
> The ABI changes are planned for 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 | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
> index d228bae..d2f7306 100644
> --- a/doc/guides/rel_notes/deprecation.rst
> +++ b/doc/guides/rel_notes/deprecation.rst
> @@ -78,3 +78,7 @@ Deprecation Notices
>  * 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.
>

Surely such a minor change doesn't require two separate announcements or 
patches for that matter. In fact I doubt it's an ABI change at all 
(although it is an API change certainly).

However to me the bigger issue is that a change like this alone doesn't 
seem like worth breaking the ABI. The ABI was just broken in 16.04 to 
introduce these struct members (among other things) and to break the ABI 
again just to fixup a const-correctness issue seems a bit much. Could it 
maybe wait until there's some actually compelling reason to break the ABI?

Note that I'm not against the change as such, const-correctness is a 
good thing.

	- Panu -

^ permalink raw reply	[relevance 9%]

* [dpdk-dev] [PATCH 1/4] pmdinfo: Add buildtools and pmdinfo utility
  @ 2016-05-16 20:41  2% ` Neil Horman
  0 siblings, 0 replies; 200+ results
From: Neil Horman @ 2016-05-16 20:41 UTC (permalink / raw)
  To: dev
  Cc: Neil Horman, Bruce Richardson, Thomas Monjalon,
	Stephen Hemminger, Panu Matilainen

pmdinfo 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.  pmdinfo
looks for the non-exported symbol names this_pmd_name<n> and this_pmd_tbl<n>
(where n is a integer counter).  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 *<pmd_name>_driver_info[] __attribute__((used)) = " \
	PMD_DRIVER_INFO=<json string>";

Where <pmd_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.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: Bruce Richardson <bruce.richardson@intel.com>
CC: Thomas Monjalon <thomas.monjalon@6wind.com>
CC: Stephen Hemminger <stephen@networkplumber.org>
CC: Panu Matilainen <pmatilai@redhat.com>
---
 GNUmakefile                  |   2 +-
 buildtools/Makefile          |  36 ++++
 buildtools/pmdinfo/Makefile  |  48 +++++
 buildtools/pmdinfo/pmdinfo.c | 435 +++++++++++++++++++++++++++++++++++++++++++
 buildtools/pmdinfo/pmdinfo.h | 210 +++++++++++++++++++++
 mk/rte.buildtools.mk         | 148 +++++++++++++++
 mk/rte.sdkbuild.mk           |   3 +-
 7 files changed, 880 insertions(+), 2 deletions(-)
 create mode 100644 buildtools/Makefile
 create mode 100644 buildtools/pmdinfo/Makefile
 create mode 100644 buildtools/pmdinfo/pmdinfo.c
 create mode 100644 buildtools/pmdinfo/pmdinfo.h
 create mode 100644 mk/rte.buildtools.mk

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/buildtools/Makefile b/buildtools/Makefile
new file mode 100644
index 0000000..0f15d58
--- /dev/null
+++ b/buildtools/Makefile
@@ -0,0 +1,36 @@
+#   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
+
+DIRS-y += pmdinfo 
+
+include $(RTE_SDK)/mk/rte.subdir.mk
diff --git a/buildtools/pmdinfo/Makefile b/buildtools/pmdinfo/Makefile
new file mode 100644
index 0000000..3dea68b
--- /dev/null
+++ b/buildtools/pmdinfo/Makefile
@@ -0,0 +1,48 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2015 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
+#
+APP = pmdinfo
+
+#
+# all sources are stored in SRCS-y
+#
+SRCS-y += pmdinfo.c
+
+#CFLAGS += $(WERROR_FLAGS) -g
+CFLAGS += -g
+
+include $(RTE_SDK)/mk/rte.buildtools.mk
+
diff --git a/buildtools/pmdinfo/pmdinfo.c b/buildtools/pmdinfo/pmdinfo.c
new file mode 100644
index 0000000..5e705ab
--- /dev/null
+++ b/buildtools/pmdinfo/pmdinfo.c
@@ -0,0 +1,435 @@
+/* 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 "pmdinfo.h"
+
+
+static const char *sym_name(struct elf_info *elf, Elf_Sym *sym)
+{
+	if (sym)
+		return elf->strtab + sym->st_name;
+	else
+		return "(unknown)";
+}
+
+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.
+  **/
+char *get_next_line(unsigned long *pos, void *file, unsigned long size)
+{
+	static char line[4096];
+	int skip = 1;
+	size_t len = 0;
+	signed char *p = (signed char *)file + *pos;
+	char *s = line;
+
+	for (; *pos < size ; (*pos)++) {
+		if (skip && isspace(*p)) {
+			p++;
+			continue;
+		}
+		skip = 0;
+		if (*p != '\n' && (*pos < size)) {
+			len++;
+			*s++ = *p++;
+			if (len > 4095)
+				break; /* Too long, stop */
+		} else {
+			/* End of string */
+			*s = '\0';
+			return line;
+		}
+	}
+	/* End of buffer */
+	return NULL;
+}
+
+void release_file(void *file, unsigned long size)
+{
+	munmap(file, size);
+}
+
+
+static void *get_sym_value(struct elf_info *info, const Elf_Sym *sym)
+{
+	void *ptr = (void *)info->hdr + info->sechdrs[sym->st_shndx].sh_offset;
+
+	return (void *)(ptr + 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 = sym_name(info, idx);
+		if (!strncmp(n, name, strlen(name)))
+			return idx;
+	}
+	return NULL;
+}
+
+static int parse_elf(struct elf_info *info, const char *filename)
+{
+	unsigned int i;
+	Elf_Ehdr *hdr;
+	Elf_Shdr *sechdrs;
+	Elf_Sym  *sym;
+	const char *secstrings;
+	unsigned int symtab_idx = ~0U, symtab_shndx_idx = ~0U;
+
+	hdr = grab_file(filename, &info->size);
+	if (!hdr) {
+		perror(filename);
+		exit(1);
+	}
+	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;
+	}
+	/* Fix endianness in ELF header */
+	hdr->e_type      = TO_NATIVE(hdr->e_type);
+	hdr->e_machine   = TO_NATIVE(hdr->e_machine);
+	hdr->e_version   = TO_NATIVE(hdr->e_version);
+	hdr->e_entry     = TO_NATIVE(hdr->e_entry);
+	hdr->e_phoff     = TO_NATIVE(hdr->e_phoff);
+	hdr->e_shoff     = TO_NATIVE(hdr->e_shoff);
+	hdr->e_flags     = TO_NATIVE(hdr->e_flags);
+	hdr->e_ehsize    = TO_NATIVE(hdr->e_ehsize);
+	hdr->e_phentsize = TO_NATIVE(hdr->e_phentsize);
+	hdr->e_phnum     = TO_NATIVE(hdr->e_phnum);
+	hdr->e_shentsize = TO_NATIVE(hdr->e_shentsize);
+	hdr->e_shnum     = TO_NATIVE(hdr->e_shnum);
+	hdr->e_shstrndx  = TO_NATIVE(hdr->e_shstrndx);
+	sechdrs = (void *)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(sechdrs[0].sh_size);
+	}
+	else {
+		info->num_sections = hdr->e_shnum;
+	}
+	if (hdr->e_shstrndx == SHN_XINDEX) {
+		info->secindex_strings = TO_NATIVE(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(sechdrs[i].sh_name);
+		sechdrs[i].sh_type      = TO_NATIVE(sechdrs[i].sh_type);
+		sechdrs[i].sh_flags     = TO_NATIVE(sechdrs[i].sh_flags);
+		sechdrs[i].sh_addr      = TO_NATIVE(sechdrs[i].sh_addr);
+		sechdrs[i].sh_offset    = TO_NATIVE(sechdrs[i].sh_offset);
+		sechdrs[i].sh_size      = TO_NATIVE(sechdrs[i].sh_size);
+		sechdrs[i].sh_link      = TO_NATIVE(sechdrs[i].sh_link);
+		sechdrs[i].sh_info      = TO_NATIVE(sechdrs[i].sh_info);
+		sechdrs[i].sh_addralign = TO_NATIVE(sechdrs[i].sh_addralign);
+		sechdrs[i].sh_entsize   = TO_NATIVE(sechdrs[i].sh_entsize);
+	}
+	/* Find symbol table. */
+	secstrings = (void *)hdr + sechdrs[info->secindex_strings].sh_offset;
+	for (i = 1; i < info->num_sections; i++) {
+		const char *secname;
+		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;
+		}
+		secname = secstrings + sechdrs[i].sh_name;
+		if (strcmp(secname, ".modinfo") == 0) {
+			if (nobits)
+				fprintf(stderr, "%s has NOBITS .modinfo\n", filename);
+			info->modinfo = (void *)hdr + sechdrs[i].sh_offset;
+			info->modinfo_len = sechdrs[i].sh_size;
+		} else if (strcmp(secname, "__ksymtab") == 0)
+			info->export_sec = i;
+		else if (strcmp(secname, "__ksymtab_unused") == 0)
+			info->export_unused_sec = i;
+		else if (strcmp(secname, "__ksymtab_gpl") == 0)
+			info->export_gpl_sec = i;
+		else if (strcmp(secname, "__ksymtab_unused_gpl") == 0)
+			info->export_unused_gpl_sec = i;
+		else if (strcmp(secname, "__ksymtab_gpl_future") == 0)
+			info->export_gpl_future_sec = i;
+
+		if (sechdrs[i].sh_type == SHT_SYMTAB) {
+			unsigned int sh_link_idx;
+			symtab_idx = i;
+			info->symtab_start = (void *)hdr +
+			    sechdrs[i].sh_offset;
+			info->symtab_stop  = (void *)hdr +
+			    sechdrs[i].sh_offset + sechdrs[i].sh_size;
+			sh_link_idx = sechdrs[i].sh_link;
+			info->strtab       = (void *)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 = (void *)hdr +
+			    sechdrs[i].sh_offset;
+			info->symtab_shndx_stop  = (void *)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(sym->st_shndx);
+		sym->st_name  = TO_NATIVE(sym->st_name);
+		sym->st_value = TO_NATIVE(sym->st_value);
+		sym->st_size  = TO_NATIVE(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(*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 *sec_name(struct elf_info *elf, int secindex)
+{
+	Elf_Shdr *sechdrs = elf->sechdrs;
+	return (void *)elf->hdr +
+		elf->sechdrs[elf->secindex_strings].sh_offset +
+		sechdrs[secindex].sh_name;
+}
+
+static int get_symbol_index(struct elf_info *info, Elf64_Sym *sym)
+{
+	const char *name =  sym_name(info, sym);
+	const char *idx;
+
+	idx = name;
+	while (idx) {
+		if (isdigit(*idx))
+			return atoi(idx);
+		idx++;
+	}
+	return -1;
+}
+
+static int complete_pmd_entry(struct elf_info *info, struct pmd_driver *drv)
+{
+	const char *tname;
+	int i = get_symbol_index(info, drv->name_sym);
+	char drvsym[128];
+
+	if (i == -1)
+		return -ENOENT;
+
+	drv->name = get_sym_value(info, drv->name_sym);
+
+	sprintf(drvsym, "this_pmd_driver%d", i);
+
+	drv->driver = find_sym_in_symtab(info, drvsym, NULL);
+
+	/*
+ 	 * If this returns NULL, then this is a PMD_VDEV, because
+ 	 * it has no pci table reference
+ 	 */
+	if (!drv->driver) {
+		drv->pci_tbl = NULL;
+		return 0;
+	}
+
+	tname = get_sym_value(info, drv->driver);
+	drv->pci_tbl_sym = find_sym_in_symtab(info, tname, NULL);
+
+	if (!drv->pci_tbl_sym)
+		return -ENOENT;
+
+	drv->pci_tbl = (struct rte_pci_id *)get_sym_value(info, drv->pci_tbl_sym);
+	if (!drv->pci_tbl)
+		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 = malloc(sizeof(struct pmd_driver));
+		new->name_sym = find_sym_in_symtab(info, "this_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);
+			} else {
+				new->next = info->drivers;
+				info->drivers = new;
+			}
+		}
+	} while (last);
+}
+
+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);
+		fprintf(ofd,"\\\"type\\\" : \\\"%s\\\", ", drv->pci_tbl ? "PMD_PDEV" : "PMD_VDEV");
+ 
+		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 = 1;
+
+	if (argc < 3) {
+		fprintf(stderr, "usage: pmdinfo <object file> <c output file>\n");
+		exit(127);
+	}
+	parse_elf(&info, argv[1]);
+
+	locate_pmd_entries(&info);
+
+	if (info.drivers) {
+		output_pmd_info_string(&info, argv[2]);
+		rc = 0;
+	} else {
+		fprintf(stderr, "Hmm, Appears to be a driver but no drivers registered\n");
+	}
+
+	parse_elf_finish(&info);
+	exit(rc);
+}
diff --git a/buildtools/pmdinfo/pmdinfo.h b/buildtools/pmdinfo/pmdinfo.h
new file mode 100644
index 0000000..5dafb67
--- /dev/null
+++ b/buildtools/pmdinfo/pmdinfo.h
@@ -0,0 +1,210 @@
+#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>
+
+
+/* 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
+
+#if 0 
+
+#define Elf_Ehdr    Elf32_Ehdr
+#define Elf_Shdr    Elf32_Shdr
+#define Elf_Sym     Elf32_Sym
+#define Elf_Addr    Elf32_Addr
+#define Elf_Sword   Elf64_Sword
+#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
+#else
+
+#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
+#endif
+
+/* The 64-bit MIPS ELF ABI uses an unusual reloc format. */
+typedef struct
+{
+	Elf32_Word    r_sym;	/* Symbol index */
+	unsigned char r_ssym;	/* Special symbol for 2nd relocation */
+	unsigned char r_type3;	/* 3rd relocation type */
+	unsigned char r_type2;	/* 2nd relocation type */
+	unsigned char r_type1;	/* 1st relocation type */
+} _Elf64_Mips_R_Info;
+
+typedef union
+{
+	Elf64_Xword		r_info_number;
+	_Elf64_Mips_R_Info	r_info_fields;
+} _Elf64_Mips_R_Info_union;
+
+#define ELF64_MIPS_R_SYM(i) \
+  ((__extension__ (_Elf64_Mips_R_Info_union)(i)).r_info_fields.r_sym)
+
+#define ELF64_MIPS_R_TYPE(i) \
+  ((__extension__ (_Elf64_Mips_R_Info_union)(i)).r_info_fields.r_type1)
+
+#if 0 
+
+static inline void __endian(const void *src, void *dest, unsigned int size)
+{
+	unsigned int i;
+	for (i = 0; i < size; i++)
+		((unsigned char*)dest)[i] = ((unsigned char*)src)[size - i-1];
+}
+
+#define TO_NATIVE(x)						\
+({								\
+	typeof(x) __x;						\
+	__endian(&(x), &(__x), sizeof(__x));			\
+	__x;							\
+})
+
+#else /* endianness matches */
+
+#define TO_NATIVE(x) (x)
+
+#endif
+
+#define NOFAIL(ptr)   do_nofail((ptr), #ptr)
+void *do_nofail(void *ptr, const char *expr);
+
+struct buffer {
+	char *p;
+	int pos;
+	int size;
+};
+
+void __attribute__((format(printf, 2, 3)))
+buf_printf(struct buffer *buf, const char *fmt, ...);
+
+void
+buf_write(struct buffer *buf, const char *s, int len);
+
+struct module {
+	struct module *next;
+	const char *name;
+	int gpl_compatible;
+	struct symbol *unres;
+	int seen;
+	int skip;
+	int has_init;
+	int has_cleanup;
+	struct buffer dev_table_buf;
+	char	     srcversion[25];
+	int is_dot_o;
+};
+
+struct rte_pci_id {
+	uint16_t vendor_id;           /**< Vendor ID or PCI_ANY_ID. */
+	uint16_t device_id;           /**< Device ID or PCI_ANY_ID. */
+	uint16_t subsystem_vendor_id; /**< Subsystem vendor ID or PCI_ANY_ID. */
+	uint16_t subsystem_device_id; /**< Subsystem device ID or PCI_ANY_ID. */
+};
+
+struct pmd_driver {
+	Elf_Sym *driver;
+	Elf_Sym *pci_tbl_sym;
+	Elf_Sym *name_sym;
+	struct rte_pci_id *pci_tbl;
+	struct pmd_driver *next;
+	const char *name;
+};
+
+struct elf_info {
+	unsigned long size;
+	Elf_Ehdr     *hdr;
+	Elf_Shdr     *sechdrs;
+	Elf_Sym      *symtab_start;
+	Elf_Sym      *symtab_stop;
+	Elf_Section  export_sec;
+	Elf_Section  export_unused_sec;
+	Elf_Section  export_gpl_sec;
+	Elf_Section  export_unused_gpl_sec;
+	Elf_Section  export_gpl_future_sec;
+	char         *strtab;
+	char	     *modinfo;
+	unsigned int modinfo_len;
+
+	/* 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;
+};
+
+static inline int is_shndx_special(unsigned int i)
+{
+	return i != SHN_XINDEX && i >= SHN_LORESERVE && i <= SHN_HIRESERVE;
+}
+
+/*
+ * Move reserved section indices SHN_LORESERVE..SHN_HIRESERVE out of
+ * the way to -256..-1, to avoid conflicting with real section
+ * indices.
+ */
+#define SPECIAL(i) ((i) - (SHN_HIRESERVE + 1))
+
+/* Accessor for sym->st_shndx, hides ugliness of "64k sections" */
+static inline unsigned int get_secindex(const struct elf_info *info,
+					const Elf_Sym *sym)
+{
+	if (is_shndx_special(sym->st_shndx))
+		return SPECIAL(sym->st_shndx);
+	if (sym->st_shndx != SHN_XINDEX)
+		return sym->st_shndx;
+	return info->symtab_shndx_start[sym - info->symtab_start];
+}
+
+/* file2alias.c */
+extern unsigned int cross_build;
+void handle_moddevtable(struct module *mod, struct elf_info *info,
+			Elf_Sym *sym, const char *symname);
+void add_moddevtable(struct buffer *buf, struct module *mod);
+
+/* sumversion.c */
+void maybe_frob_rcs_version(const char *modfilename,
+			    char *version,
+			    void *modinfo,
+			    unsigned long modinfo_offset);
+void get_src_version(const char *modname, char sum[], unsigned sumlen);
+
+/* from modpost.c */
+void *grab_file(const char *filename, unsigned long *size);
+char* get_next_line(unsigned long *pos, void *file, unsigned long size);
+void release_file(void *file, unsigned long size);
+
+void fatal(const char *fmt, ...);
+void warn(const char *fmt, ...);
+void merror(const char *fmt, ...);
diff --git a/mk/rte.buildtools.mk b/mk/rte.buildtools.mk
new file mode 100644
index 0000000..e8bfcef
--- /dev/null
+++ b/mk/rte.buildtools.mk
@@ -0,0 +1,148 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
+#   Copyright(c) 2014-2015 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 $(RTE_SDK)/mk/internal/rte.compile-pre.mk
+include $(RTE_SDK)/mk/internal/rte.install-pre.mk
+include $(RTE_SDK)/mk/internal/rte.clean-pre.mk
+include $(RTE_SDK)/mk/internal/rte.build-pre.mk
+include $(RTE_SDK)/mk/internal/rte.depdirs-pre.mk
+
+# VPATH contains at least SRCDIR
+VPATH += $(SRCDIR)
+
+_BUILD = $(APP)
+_INSTALL = $(INSTALL-FILES-y) $(SYMLINK-FILES-y)
+_INSTALL += $(RTE_OUTPUT)/buildtools/$(APP) $(RTE_OUTPUT)/buildtools/$(APP).map
+POSTINSTALL += target-appinstall
+_CLEAN = doclean
+POSTCLEAN += target-appclean
+
+.PHONY: all
+all: install
+
+.PHONY: install
+install: build _postinstall
+
+_postinstall: build
+
+.PHONY: build
+build: _postbuild
+
+exe2cmd = $(strip $(call dotfile,$(patsubst %,%.cmd,$(1))))
+
+ifeq ($(LINK_USING_CC),1)
+override EXTRA_LDFLAGS := $(call linkerprefix,$(EXTRA_LDFLAGS))
+O_TO_EXE = $(CC) $(CFLAGS) $(LDFLAGS_$(@)) \
+	-Wl,-Map=$(@).map,--cref -o $@ $(OBJS-y) $(call linkerprefix,$(LDFLAGS)) \
+	$(EXTRA_LDFLAGS) $(call linkerprefix,$(LDLIBS))
+else
+O_TO_EXE = $(LD) $(LDFLAGS) $(LDFLAGS_$(@)) $(EXTRA_LDFLAGS) \
+	-Map=$(@).map --cref -o $@ $(OBJS-y) $(LDLIBS)
+endif
+O_TO_EXE_STR = $(subst ','\'',$(O_TO_EXE)) #'# fix syntax highlight
+O_TO_EXE_DISP = $(if $(V),"$(O_TO_EXE_STR)","  LD $(@)")
+O_TO_EXE_CMD = "cmd_$@ = $(O_TO_EXE_STR)"
+O_TO_EXE_DO = @set -e; \
+	echo $(O_TO_EXE_DISP); \
+	$(O_TO_EXE) && \
+	echo $(O_TO_EXE_CMD) > $(call exe2cmd,$(@))
+
+-include .$(APP).cmd
+
+# path where libraries are retrieved
+LDLIBS_PATH := $(subst -Wl$(comma)-L,,$(filter -Wl$(comma)-L%,$(LDLIBS)))
+LDLIBS_PATH += $(subst -L,,$(filter -L%,$(LDLIBS)))
+
+# list of .a files that are linked to this application
+LDLIBS_NAMES := $(patsubst -l%,lib%.a,$(filter -l%,$(LDLIBS)))
+LDLIBS_NAMES += $(patsubst -Wl$(comma)-l%,lib%.a,$(filter -Wl$(comma)-l%,$(LDLIBS)))
+
+# list of found libraries files (useful for deps). If not found, the
+# library is silently ignored and dep won't be checked
+LDLIBS_FILES := $(wildcard $(foreach dir,$(LDLIBS_PATH),\
+	$(addprefix $(dir)/,$(LDLIBS_NAMES))))
+
+#
+# Compile executable file if needed
+#
+$(APP): $(OBJS-y) $(LDLIBS_FILES) $(DEP_$(APP)) $(LDSCRIPT) FORCE
+	@[ -d $(dir $@) ] || mkdir -p $(dir $@)
+	$(if $(D),\
+		@echo -n "$< -> $@ " ; \
+		echo -n "file_missing=$(call boolean,$(file_missing)) " ; \
+		echo -n "cmdline_changed=$(call boolean,$(call cmdline_changed,$(O_TO_EXE_STR))) " ; \
+		echo -n "depfile_missing=$(call boolean,$(depfile_missing)) " ; \
+		echo "depfile_newer=$(call boolean,$(depfile_newer)) ")
+	$(if $(or \
+		$(file_missing),\
+		$(call cmdline_changed,$(O_TO_EXE_STR)),\
+		$(depfile_missing),\
+		$(depfile_newer)),\
+		$(O_TO_EXE_DO))
+
+#
+# install app in $(RTE_OUTPUT)/app
+#
+$(RTE_OUTPUT)/buildtools/$(APP): $(APP)
+	@echo "  INSTALL-APP $(APP)"
+	@[ -d $(RTE_OUTPUT)/buildtools ] || mkdir -p $(RTE_OUTPUT)/buildtools
+	$(Q)cp -f $(APP) $(RTE_OUTPUT)/buildtools
+
+#
+# install app map file in $(RTE_OUTPUT)/app
+#
+$(RTE_OUTPUT)/buildtools/$(APP).map: $(APP)
+	@echo "  INSTALL-MAP $(APP).map"
+	@[ -d $(RTE_OUTPUT)/buildtools ] || mkdir -p $(RTE_OUTPUT)/buildtools
+	$(Q)cp -f $(APP).map $(RTE_OUTPUT)/buildtools
+
+#
+# Clean all generated files
+#
+.PHONY: clean
+clean: _postclean
+	$(Q)rm -f $(_BUILD_TARGETS) $(_INSTALL_TARGETS) $(_CLEAN_TARGETS)
+
+.PHONY: doclean
+doclean:
+	$(Q)rm -rf $(APP) $(OBJS-all) $(DEPS-all) $(DEPSTMP-all) \
+	  $(CMDS-all) $(INSTALL-FILES-all) .$(APP).cmd
+
+
+include $(RTE_SDK)/mk/internal/rte.compile-post.mk
+include $(RTE_SDK)/mk/internal/rte.install-post.mk
+include $(RTE_SDK)/mk/internal/rte.clean-post.mk
+include $(RTE_SDK)/mk/internal/rte.build-post.mk
+include $(RTE_SDK)/mk/internal/rte.depdirs-post.mk
+
+.PHONY: FORCE
+FORCE:
diff --git a/mk/rte.sdkbuild.mk b/mk/rte.sdkbuild.mk
index eec5241..fb68af2 100644
--- a/mk/rte.sdkbuild.mk
+++ b/mk/rte.sdkbuild.mk
@@ -64,7 +64,8 @@ build: $(ROOTDIRS-y)
 clean: $(CLEANDIRS)
 	@rm -rf $(RTE_OUTPUT)/include $(RTE_OUTPUT)/app \
 		$(RTE_OUTPUT)/hostapp $(RTE_OUTPUT)/lib \
-		$(RTE_OUTPUT)/hostlib $(RTE_OUTPUT)/kmod
+		$(RTE_OUTPUT)/hostlib $(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
-- 
2.5.5

^ permalink raw reply	[relevance 2%]

* Re: [dpdk-dev] [PATCH v5] mempool: reduce rte_mempool structure size
  2016-04-14  9:42  2% ` [dpdk-dev] [PATCH v5] " Olivier Matz
  2016-04-14 13:28  0%   ` Wiles, Keith
  2016-04-14 13:53  0%   ` Wiles, Keith
@ 2016-05-17  5:31  0%   ` Thomas Monjalon
  2 siblings, 0 replies; 200+ results
From: Thomas Monjalon @ 2016-05-17  5:31 UTC (permalink / raw)
  To: Olivier Matz, keith.wiles; +Cc: dev, pmatilai

2016-04-14 11:42, Olivier Matz:
> From: Keith Wiles <keith.wiles@intel.com>
> 
> The rte_mempool structure is changed, which will cause an ABI change
> for this structure. Providing backward compat is not reasonable
> here as this structure is used in multiple defines/inlines.

The deprecation notice must be removed by this patch.

[...]
> +/**
>   * Calculate the size of the mempool header.
>   *
>   * @param mp
> @@ -254,9 +256,9 @@ struct rte_mempool {
>   * @param pgn
>   *   Number of pages used to store mempool objects.

A new parameter has been forgotten:
 * @param cs
 *   Size of the per-lcore cache.                                                                                

>   */
> -#define	MEMPOOL_HEADER_SIZE(mp, pgn)	(sizeof(*(mp)) + \
> -	RTE_ALIGN_CEIL(((pgn) - RTE_DIM((mp)->elt_pa)) * \
> -	sizeof ((mp)->elt_pa[0]), RTE_CACHE_LINE_SIZE))
> +#define MEMPOOL_HEADER_SIZE(mp, pgn, cs) \
> +	(sizeof(*(mp)) + __PA_SIZE(mp, pgn) + (((cs) == 0) ? 0 : \
> +	(sizeof(struct rte_mempool_cache) * RTE_MAX_LCORE)))

Applied with above changes

^ permalink raw reply	[relevance 0%]

* [dpdk-dev] [PATCH v3 8/8] doc: update doc for packet capture framework
  @ 2016-05-17 16:37  6%     ` Reshma Pattan
  0 siblings, 0 replies; 200+ results
From: Reshma Pattan @ 2016-05-17 16:37 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 | 121 ++++++++++++++++++++++++++++++++
 doc/guides/rel_notes/release_16_07.rst  |   6 ++
 doc/guides/sample_app_ug/index.rst      |   1 +
 doc/guides/sample_app_ug/pdump.rst      | 109 ++++++++++++++++++++++++++++
 6 files changed, 241 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 58f5ba4..d4d0630 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -439,6 +439,9 @@ 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..6af77b9
--- /dev/null
+++ b/doc/guides/prog_guide/pdump_library.rst
@@ -0,0 +1,121 @@
+..  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:
+
+pdump Library
+=============
+
+Pdump library provides framework for packet capturing on DPDK.
+
+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 un initialize the packet capture framework.
+
+Pdump library works on server and client 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.
+
+Library provides below APIs
+
+``rte_pdump_init()``
+This API initializes the packet capture framework.
+
+``rte_pdump_enable()``
+This API enables the packet capturing on a given port and queue.
+Note: filter option in the API is place holder for future use.
+
+``rte_pdump_enable_by_deviceid()``
+This API enables the packet capturing on a given device id
+(device name or pci address) and queue.
+Note: filter option in the API is place holder for future use.
+
+``rte_pdump_disable()``
+This API disables the packet capturing on a given port and queue.
+
+``rte_pdump_disable_by_deviceid()``
+This API disables the packet capturing on a given device_id and queue.
+
+``rte_pdump_uninit()``
+This API un initializes the packet capture framework.
+
+
+Implementation Details
+----------------------
+
+On a call to library API ``rte_pdump_init()``, library creates pthread and server socket.
+Server socket in pthread context will be listening to the client requests to enable/disable
+the packet capture.
+
+Who ever calls this API first will have server socket created,
+subsequent calls to this APIs will not create any further server sockets. i.e only one server
+socket is allowed.
+
+On each call to library APIs ``rte_pdump_enable()/rte_pdump_enable_by_deviceid()``
+to enable the packet capture, library creates separate client sockets,
+builds up enable request and sends the request to the server.
+Server listening on the socket will serve the request, enable the packet capture
+by registering ethernet rx/tx callbacks for the given port/device_id and queue combinations.
+Server mirrors the packets to new mempool and enqueue them to the ring that clients has passed
+in these APIs.
+Server sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, client sockets will be closed.
+
+On each call to library APIs ``rte_pdump_disable()/rte_pdump_disable_by_deviceid()``
+to disable packet capture, library creates separate client sockets,
+builds up disable request and sends the request to the server.
+Server listening on the socket will serve the request, disable the packet capture
+by removing the ethernet rx/tx callbacks for the given port/device_id and queue combinations.
+Server sends the response back to the client about the status of the request that was processed.
+After the response is received from the server, client sockets will be closed.
+
+On a call to library API ``rte_pdump_uninit()``, library closes the pthread and the server socket.
+
+
+Use Case: Packet Capturing
+--------------------------
+
+app/pdump tool is developed based on this library to capture the packets
+in DPDK.
+Users can develop their own packet capturing application using new library
+if they wish to do so.
diff --git a/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 58c8ef9..7275a35 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -38,6 +38,9 @@ New Features
 
   The size of the mempool structure is reduced if the per-lcore cache is disabled.
 
+* **Added packet capturing support.**
+  Now users have facility to capture packets on dpdk ports using librte_pdump
+  and app/pdump tool.
 
 Resolved Issues
 ---------------
@@ -101,6 +104,7 @@ API Changes
   ibadcrc, ibadlen, imcasts, fdirmatch, fdirmiss,
   tx_pause_xon, rx_pause_xon, tx_pause_xoff, rx_pause_xoff.
 
+* Now function ``rte_eth_dev_get_port_by_name`` changed to public API.
 
 ABI Changes
 -----------
@@ -112,6 +116,8 @@ 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..60ea1f6
--- /dev/null
+++ b/doc/guides/sample_app_ug/pdump.rst
@@ -0,0 +1,109 @@
+
+..  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 capturing
+on dpdk ports and capturing the packets.
+
+Running the Application
+-----------------------
+The application has a pdump command line option with various sub arguments inside:
+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.
+
+.. code-block:: console
+
+   ./$(RTE_TARGET)/app/pdump -- --pdump '(port=<port_id> |
+   device_id=<pci address or device name>),
+   (queue=2), (rx-dev=<iface/path to pcap file> |
+   tx-dev=<iface/path to pcap file> |
+   rxtx-dev=<iface/path to pcap file>),
+   [ring-size=1024], [mbuf-size=2048], [total-num-mbufs=8191]'
+
+Parameters
+~~~~~~~~~~
+``--pdump``: Specifies arguments needed for packet capturing.
+
+``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.
+
+``queue``
+Queue id of the eth device on which packets should be captured.
+User can pass on queue value as ‘*’ if packets capturing has to be enabled
+on all queues of the eth device.
+
+``rx-dev``
+Can be either pcap file name or any linux iface onto which ingress side packets of
+dpdk eth device will be sent on for users to view.
+
+``tx-dev``
+Can be either pcap file name or any linux iface onto which egress side packets of
+dpdk eth device will be sent on for users to view.
+
+``rxtx-dev``
+Can be either pcap file name or any linux iface onto which both ingress &
+egress side packets of dpdk eth device will be sent on for users to view.
+
+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 should pass on both rx-dev and tx-dev.
+To receive both ingress and egress packets on same device, should pass only rxtx-dev.
+
+Pdump tool uses these devices internally to create PCAPPMD vdev having ``tx_stream``
+as either of these devices.
+
+``ring-size``
+Size of the ring. This value is used internally for ring creation.
+The ring will be used to enqueue the packets from primary application to secondary.
+
+``mbuf-size``
+Size of the mbuf data room size. This is used internally for mempool creation.
+Ideally this value must be same as primary application's mempool which is used for
+packet rx.
+
+``total-num-mbufs``
+Total number mbufs in mempool. This is used internally for mempool creation.
+
+Example
+-------
+
+.. code-block:: console
+
+        $ sudo ./x86_64-native-linuxapp-gcc/app/dpdk_pdump -- --pdump 'port=0,queue=*,rxtx-dev=/tmp/rxtx-file.pcap'
-- 
2.5.0

^ permalink raw reply	[relevance 6%]

* [dpdk-dev] [PATCH v3 00/35] mempool: rework memory allocation
  2016-04-14 10:19  2% ` [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation Olivier Matz
  2016-04-14 13:50  0%   ` Wiles, Keith
@ 2016-05-18 11:04  2%   ` Olivier Matz
  2016-05-18 11:04 10%     ` [dpdk-dev] [PATCH v3 35/35] doc: update release notes about mempool allocation Olivier Matz
  1 sibling, 1 reply; 200+ results
From: Olivier Matz @ 2016-05-18 11:04 UTC (permalink / raw)
  To: dev; +Cc: bruce.richardson, stephen, keith.wiles

This series is a rework of mempool. For those who don't want to read
all the cover letter, here is a sumary:

- it is not possible to allocate large mempools if there is not enough
  contiguous memory, this series solves this issue
- introduce new APIs with less arguments: "create, populate, obj_init"
- allow to free a mempool
- split code in smaller functions, will ease the introduction of ext_handler
- remove test-pmd anonymous mempool creation
- remove most of dom0-specific mempool code
- opens the door for a eal_memory rework: we probably don't need large
  contiguous memory area anymore, working with pages would work.

This breaks the ABI as it was indicated in the deprecation for 16.04.
The API stays almost the same, no modification is needed in examples app
or in test-pmd. Only kni and mellanox drivers are slightly modified.

Changes v2 -> v3:
- fix some checkpatch issues
- rework titles and commit logs
- fix compilation with debug + shared libraries:
  rte_mempool_check_cookies() must be exported
- rebase on head

Changes v1 -> v2:
- do not change the place of __rte_unused in txq_mp2mr_mbuf_check(),
  as suggested by Keith.

Changes RFC -> v1:
- remove the rte_deconst macro, and remove some const qualifier in
  dump/audit functions
- rework modifications in mellanox drivers to ensure the mempool is
  virtually contiguous
- fix mempool memory chunk iteration (bad pointer was used)
- fix compilation on freebsd: replace MAP_LOCKED flag by mlock()
- fix compilation on tilera (pointer arithmetics)
- slightly rework and clean the mempool autotest
- fix mempool autotest on bsd
- more validation (especially mellanox drivers and kni that were not
  tested in RFC)
- passed autotests (x86_64-native-linuxapp-gcc and x86_64-native-bsdapp-gcc)
- rebase on head, reorder the patches a bit and fix minor split issues


Description of the initial issue
--------------------------------

The allocation of mbuf pool can fail even if there is enough memory.
The problem is related to the way the memory is allocated and used in
dpdk. It is particularly annoying with mbuf pools, but it can also fail
in other use cases allocating a large amount of memory.

- rte_malloc() allocates physically contiguous memory, which is needed
  for mempools, but useless most of the time.

  Allocating a large physically contiguous zone is often impossible
  because the system provide hugepages which may not be contiguous.

- rte_mempool_create() (and therefore rte_pktmbuf_pool_create())
  requires a physically contiguous zone.

- rte_mempool_xmem_create() does not solve the issue as it still
  needs the memory to be virtually contiguous, and there is no
  way in dpdk to allocate a virtually contiguous memory that is
  not also physically contiguous.

How to reproduce the issue
--------------------------

- start the dpdk with some 2MB hugepages (it can also occur with 1GB)
- allocate a large mempool
- even if there is enough memory, the allocation can fail

Example:

  git clone http://dpdk.org/git/dpdk
  cd dpdk
  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

  # we try to allocate a mempool whose size is ~450MB, it fails
  ./build/app/testpmd -l 2,4 -- --total-num-mbufs=200000 -i

The EAL logs "EAL: Virtual area found at..." shows that there are
several zones, but all smaller than 450MB.

Workarounds:

- Use 1GB hugepages: it sometimes work, but for very large
  pools (millions of mbufs) there is the same issue. Moreover,
  it would consume 1GB memory at least which can be a lot
  in some cases.

- Reboot the machine or allocate hugepages at boot time: this increases
  the chances to have more contiguous memory, but does not completely
  solve the issue

Solutions
---------

Below is a list of proposed solutions. I implemented a quick and dirty
PoC of solution 1, but it's not working in all conditions and it's
really an ugly hack.  This series implement the solution 4 which looks
the best to me, knowing it does not prevent to do more enhancements
in dpdk memory in the future (solution 3 for instance).

Solution 1: in application
--------------------------

- allocate several hugepages using rte_malloc() or rte_memzone_reserve()
  (only keeping complete hugepages)
- parse memsegs and /proc/maps to check which files mmaps these pages
- mmap the files in a contiguous virtual area
- use rte_mempool_xmem_create()

Cons:

- 1a. parsing the memsegs of rte config in the application does not
  use a public API, and can be broken if internal dpdk code changes
- 1b. some memory is lost due to malloc headers. Also, if the memory is
  very fragmented (ex: all 2MB pages are physically separated), it does
  not work at all because we cannot get any complete page. It is not
  possible to use a lower level allocator since commit fafcc11985a.
- 1c. we cannot use rte_pktmbuf_pool_create(), so we need to use mempool
  api and do a part of the job manually
- 1d. it breaks secondary processes as the virtual addresses won't be
  mmap'd at the same place in secondary process
- 1e. it only fixes the issue for the mbuf pool of the application,
  internal pools in dpdk libraries are not modified
- 1f. this is a pure linux solution (rte_map files)
- 1g. The application has to be aware of RTE_EAL_SINGLE_SEGMENTS option
  that changes the way hugepages are mapped. By the way, it's strange
  to have such a compile-time option, we should probably have only
  one behavior that works all the time.

Solution 2: in dpdk memory allocator
------------------------------------

- do the same than solution 1 in a new function rte_malloc_non_contig():
  allocate several chunks and mmap them in a contiguous virtual memory
- a flag has to be added in malloc header to do the proper cleanup in
  rte_free() (free all the chunks, munmap the memory)
- introduce a new rte_mem_get_physmap(*physmap,addr, len) that returns
  the virt2phys mapping of a virtual area in dpdk
- add a mempool flag MEMPOOL_F_NON_PHYS_CONTIG to use
  rte_malloc_non_contig() to allocate the area storing the objects

Cons:

- 2a. same than 1b: it breaks secondary processes if the mempool flag is
  used.
- 2b. same as 1d: some memory is lost due to malloc headers, and it
  cannot work if memory is too fragmented.
- 2c. rte_malloc_virt2phy() cannot be used on these zones. It would
  return the physical address of the first page. It would be better to
  return an error in this case.
- 2d. need to check how to implement this on bsd (TBD)

Solution 3: in dpdk eal memory
------------------------------

- Rework the way hugepages are mmap'd in dpdk: instead of having several
  rte_map* files, just mmap one file per node. It may drastically
  simplify EAL memory management in dpdk.
- An API should be added to retrieve the physical mapping of a virtual
  area (ex: rte_mem_get_physmap(*physmap, addr, len))
- rte_malloc() and rte_memzone_reserve() won't allocate physically
  contiguous memory anymore (TBD)
- Update mempool to always use the rte_mempool_xmem_create() version

Cons:

- 3a. lot of rework in eal memory, it will induce some behavior changes
  and maybe api changes
- 3b. possible conflicts with xen_dom0 mempool

Solution 4: in mempool
----------------------

- Introduce a new API to fill a mempool with zones that are not
  virtually contiguous. It requires to add new functions to create and
  populate a mempool. Example (TBD):

  - rte_mempool_create_empty(name, n, elt_size, cache_size, priv_size)
  - rte_mempool_populate(mp, addr, len): add virtual memory for objects
  - rte_mempool_mempool_obj_iter(mp, obj_cb, arg): call a cb for each object

- update rte_mempool_create() to allocate objects in several memory
  chunks by default if there is no large enough physically contiguous
  memory.

Tests done
----------

Compilation
~~~~~~~~~~~

The following targets:

 x86_64-native-linuxapp-gcc
 i686-native-linuxapp-gcc
 x86_x32-native-linuxapp-gcc
 x86_64-native-linuxapp-clang
 x86_64-native-bsdapp-gcc
 ppc_64-power8-linuxapp-gcc
 tile-tilegx-linuxapp-gcc (only the mempool files, the target does not compile)

Libraries with and without debug, in static and shared mode + examples.

autotests
~~~~~~~~~

Passed all autotests on x86_64-native-linuxapp-gcc (including kni) and
mempool-related autotests on x86_64-native-bsdapp-gcc.

test-pmd
~~~~~~~~

# now starts fine, was failing before if mempool was too fragmented
./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -- -i --port-topology=chained

# still ok
./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -m 256 -- -i --port-topology=chained --mp-anon
set fwd txonly
start
stop

# fail, but was failing before too. The problem is because the physical
# addresses are not properly set when using --no-huge. The mempool phys addr
# are now correct, but the zones allocated through memzone_reserve() are
# still wrong. This could be fixed in a future series.
./x86_64-native-linuxapp-gcc/app/testpmd -l 0,2,4 -n 4 -m 256 --no-huge -- -i ---port-topology=chained
set fwd txonly
start
stop

*** BLURB HERE ***

Olivier Matz (35):
  mempool: rework comments and style
  mempool: rename element size variables
  mempool: uninline function to check cookies
  mempool: use sizeof to get the size of header and trailer
  mempool: rename object constructor typedef
  mempool: list objects when added
  mempool: remove const qualifier when browsing pools
  mempool: remove const qualifier in dump and audit
  mempool: use the list to iterate the elements
  mempool: use the list to audit all elements
  mempool: use the list to initialize objects
  mempool: create internal ring in a specific function
  mempool: store physical address in objects
  mempool: remove macro to check if contiguous
  mempool: store memory chunks in a list
  mempool: add function to iterate the memory chunks
  mempool: simplify the memory usage calculation
  mempool: introduce a free callback for memory chunks
  mempool: get memory size with unspecified page size
  mempool: allocate in several memory chunks by default
  eal: lock memory when not using hugepages
  mempool: support no hugepage mode
  mempool: replace physical address by a memzone pointer
  mempool: introduce a function to free a pool
  mempool: introduce a function to create an empty pool
  eal/xen: return machine address without knowing memseg id
  mempool: rework support of Xen dom0
  mempool: create the internal ring when populating
  mempool: populate with anonymous memory
  mempool: make mempool populate and free api public
  app/testpmd: remove anonymous mempool code
  mem: avoid memzone/mempool/ring name truncation
  mempool: add flag for removing phys contiguous constraint
  app/test: rework mempool test
  doc: update release notes about mempool allocation

 app/test-pmd/Makefile                        |    4 -
 app/test-pmd/mempool_anon.c                  |  201 -----
 app/test-pmd/mempool_osdep.h                 |   54 --
 app/test-pmd/testpmd.c                       |   23 +-
 app/test/test_mempool.c                      |  243 +++---
 doc/guides/rel_notes/deprecation.rst         |    8 -
 doc/guides/rel_notes/release_16_07.rst       |    9 +
 drivers/net/mlx4/mlx4.c                      |  140 ++--
 drivers/net/mlx5/mlx5_rxtx.c                 |  140 ++--
 drivers/net/mlx5/mlx5_rxtx.h                 |    4 +-
 drivers/net/xenvirt/rte_eth_xenvirt.h        |    2 +-
 drivers/net/xenvirt/rte_mempool_gntalloc.c   |    4 +-
 lib/librte_eal/common/eal_common_log.c       |    2 +-
 lib/librte_eal/common/eal_common_memzone.c   |   10 +-
 lib/librte_eal/common/include/rte_memory.h   |   11 +-
 lib/librte_eal/linuxapp/eal/eal_memory.c     |    2 +-
 lib/librte_eal/linuxapp/eal/eal_xen_memory.c |   18 +-
 lib/librte_kni/rte_kni.c                     |   12 +-
 lib/librte_mempool/Makefile                  |    3 -
 lib/librte_mempool/rte_dom0_mempool.c        |  133 ----
 lib/librte_mempool/rte_mempool.c             | 1061 +++++++++++++++++---------
 lib/librte_mempool/rte_mempool.h             |  597 +++++++--------
 lib/librte_mempool/rte_mempool_version.map   |   19 +-
 lib/librte_ring/rte_ring.c                   |   16 +-
 24 files changed, 1402 insertions(+), 1314 deletions(-)
 delete mode 100644 app/test-pmd/mempool_anon.c
 delete mode 100644 app/test-pmd/mempool_osdep.h
 delete mode 100644 lib/librte_mempool/rte_dom0_mempool.c

-- 
2.8.0.rc3

^ permalink raw reply	[relevance 2%]

* [dpdk-dev] [PATCH v3 35/35] doc: update release notes about mempool allocation
  2016-05-18 11:04  2%   ` [dpdk-dev] [PATCH v3 00/35] " Olivier Matz
@ 2016-05-18 11:04 10%     ` Olivier Matz
  0 siblings, 0 replies; 200+ results
From: Olivier Matz @ 2016-05-18 11:04 UTC (permalink / raw)
  To: dev; +Cc: bruce.richardson, stephen, keith.wiles

Remove the deprecation notice and add an entry in the release note
for the changes in mempool allocation.

Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
---
 doc/guides/rel_notes/deprecation.rst   | 8 --------
 doc/guides/rel_notes/release_16_07.rst | 9 +++++++++
 2 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/doc/guides/rel_notes/deprecation.rst b/doc/guides/rel_notes/deprecation.rst
index 7d94ba5..ad05eba 100644
--- a/doc/guides/rel_notes/deprecation.rst
+++ b/doc/guides/rel_notes/deprecation.rst
@@ -51,14 +51,6 @@ Deprecation Notices
   functions added to facilitate the creation of mempools using an external
   handler. The 16.07 release will contain these changes.
 
-* The rte_mempool allocation will be changed in 16.07:
-  allocation of large mempool in several virtual memory chunks, new API
-  to populate a mempool, new API to free a mempool, allocation in
-  anonymous mapping, drop of specific dom0 code. These changes will
-  induce a modification of the rte_mempool structure, plus a
-  modification of the API of rte_mempool_obj_iter(), implying a breakage
-  of the ABI.
-
 * 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/doc/guides/rel_notes/release_16_07.rst b/doc/guides/rel_notes/release_16_07.rst
index 58c8ef9..6cb5304 100644
--- a/doc/guides/rel_notes/release_16_07.rst
+++ b/doc/guides/rel_notes/release_16_07.rst
@@ -38,6 +38,15 @@ New Features
 
   The size of the mempool structure is reduced if the per-lcore cache is disabled.
 
+* **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.
+  * Dropped specific Xen Dom0 code.
+  * Dropped specific anonymous mempool code in testpmd.
+
 
 Resolved Issues
 ---------------
-- 
2.8.0.rc3

^ permalink raw reply	[relevance 10%]

* Re: [dpdk-dev] [PATCH] mbuf: make rearm_data address naturally aligned
  @ 2016-05-19  8:50  3%     ` Bruce Richardson
  2016-05-19 11:54  0%       ` Jan Viktorin
  2016-05-19 12:18  0%       ` Ananyev, Konstantin
  0 siblings, 2 replies; 200+ results
From: Bruce Richardson @ 2016-05-19  8:50 UTC (permalink / raw)
  To: Jerin Jacob
  Cc: dev, thomas.monjalon, konstantin.ananyev, viktorin, jianbo.liu

On Thu, May 19, 2016 at 12:20:16AM +0530, Jerin Jacob wrote:
> On Wed, May 18, 2016 at 05:43:00PM +0100, Bruce Richardson wrote:
> > On Wed, May 18, 2016 at 07:27:43PM +0530, Jerin Jacob wrote:
> > > To avoid multiple stores on fast path, Ethernet drivers
> > > aggregate the writes to data_off, refcnt, nb_segs and port
> > > to an uint64_t data and write the data in one shot
> > > with uint64_t* at &mbuf->rearm_data address.
> > > 
> > > Some of the non-IA platforms have store operation overhead
> > > if the store address is not naturally aligned.This patch
> > > fixes the performance issue on those targets.
> > > 
> > > Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
> > > ---
> > > 
> > > Tested this patch on IA and non-IA(ThunderX) platforms.
> > > This patch shows 400Kpps/core improvement on ThunderX + ixgbe + vector environment.
> > > and this patch does not have any overhead on IA platform.
> > > 
> > > Have tried an another similar approach by replacing "buf_len" with "pad"
> > > (in this patch context),
> > > Since it has additional overhead on read and then mask to keep "buf_len" intact,
> > > not much improvement is not shown.
> > > ref: http://dpdk.org/ml/archives/dev/2016-May/038914.html
> > > 
> > > ---
> > While this will work and from your tests doesn't seem to have a performance
> > impact, I'm not sure I particularly like it. It's extending out the end of
> > cacheline0 of the mbuf by 16 bytes, though I suppose it's not technically using
> > up any more space of it.
> 
> Extending by 2 bytes. Right ?. Yes, I guess, Now we using only 56 out of 64 bytes
> in the first 64-byte cache line.
> 
> > 
> > What I'm wondering about though, is do we have any usecases where we need a
> > variable buf_len for packets for RX. These mbufs come directly from a mempool,
> > which is generally understood to be a set of fixed-sized buffers. I realise that
> > this change was made in the past after some discussion, but one of the key points
> > there [at least to my reading] was that - even though nobody actually made a
> > concrete case where they had variable-sized buffers - having support for them
> > made no performance difference.
> > 
> > The latter part of that has now changed, and supporting variable-sized mbufs
> > from an mbuf pool has a perf impact. Do we definitely need that functionality,
> > because the easiest fix here is just to move the rxrearm marker back above
> > mbuf_len as it was originally in releases like 1.8?
> 
> And initialize the buf_len with mp->elt_size - sizeof(struct rte_mbuf).
> Right?
> 
> I don't have a strong opinion on this, I can do this if there is no
> objection on this. Let me know.
> 
> However, I do see in future, "buf_len" may belong at the end of the first 64 byte
> cache line as currently "port" is defined as uint8_t, IMO, that is less.
> We may need to increase that uint16_t. The reason why I think that
> because, Currently in ThunderX HW, we do have 128VFs per socket for
> built-in NIC, So, the two node configuration and one external PCIe NW card
> configuration can easily go beyond 256 ports.
> 
Ok, good point. If you think it's needed, and if we are changing the mbuf
structure, it might be a good time to extend that field while you are at it, save
a second ABI break later on.

/Bruce

> > 
> > Regards,
> > /Bruce
> > 
> > Ref: http://dpdk.org/ml/archives/dev/2014-December/009432.html
> > 

^ permalink raw reply	[relevance 3%]

* Re: [dpdk-dev] [PATCH] mbuf: make rearm_data address naturally aligned
  2016-05-19  8:50  3%     ` Bruce Richardson
@ 2016-05-19 11:54  0%       ` Jan Viktorin
  2016-05-19 12:18  0%       ` Ananyev, Konstantin
  1 sibling, 0 replies; 200+ results
From: Jan Viktorin @ 2016-05-19 11:54 UTC (permalink / raw)
  To: Bruce Richardson
  Cc: Jerin Jacob, dev, thomas.monjalon, konstantin.ananyev, jianbo.liu

On Thu, 19 May 2016 09:50:48 +0100
Bruce Richardson <bruce.richardson@intel.com> wrote:

> On Thu, May 19, 2016 at 12:20:16AM +0530, Jerin Jacob wrote:
> > On Wed, May 18, 2016 at 05:43:00PM +0100, Bruce Richardson wrote:  
> > > On Wed, May 18, 2016 at 07:27:43PM +0530, Jerin Jacob wrote:  
> > > > To avoid multiple stores on fast path, Ethernet drivers
> > > > aggregate the writes to data_off, refcnt, nb_segs and port
> > > > to an uint64_t data and write the data in one shot
> > > > with uint64_t* at &mbuf->rearm_data address.
> > > > 
> > > > Some of the non-IA platforms have store operation overhead
> > > > if the store address is not naturally aligned.This patch
> > > > fixes the performance issue on those targets.
> > > > 
> > > > Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
> > > > ---
> > > > 
> > > > Tested this patch on IA and non-IA(ThunderX) platforms.
> > > > This patch shows 400Kpps/core improvement on ThunderX + ixgbe + vector environment.
> > > > and this patch does not have any overhead on IA platform.

Hello,

I can confirm a very small improvement in our synthetic tests based on the PMD
null (ARM Cortex-A9). For a single-core (1C) test, there is now a lower overhead
and it is more stable with different packet lengths. However, when running dual-core
(2C), the result is slightly slower but again, it seems to be more stable.

Without this patch (cycles per packet):

 length:   64     128     256     512    1024    1280    1518
  1C      488     544     487     454     543     488     515
  2C      433     433     431     433     433     461     443

Applied this patch (cycles per packet):

 length:   64     128     256     512    1024    1280    1518
  1C      472     472     472     472     473     472     473
  2C      435     435     435     435     436     436     436

Regards
Jan

> > > > 
> > > > Have tried an another similar approach by replacing "buf_len" with "pad"
> > > > (in this patch context),
> > > > Since it has additional overhead on read and then mask to keep "buf_len" intact,
> > > > not much improvement is not shown.
> > > > ref: http://dpdk.org/ml/archives/dev/2016-May/038914.html
> > > > 
> > > > ---  
> > > While this will work and from your tests doesn't seem to have a performance
> > > impact, I'm not sure I particularly like it. It's extending out the end of
> > > cacheline0 of the mbuf by 16 bytes, though I suppose it's not technically using
> > > up any more space of it.  
> > 
> > Extending by 2 bytes. Right ?. Yes, I guess, Now we using only 56 out of 64 bytes
> > in the first 64-byte cache line.
> >   
> > > 
> > > What I'm wondering about though, is do we have any usecases where we need a
> > > variable buf_len for packets for RX. These mbufs come directly from a mempool,
> > > which is generally understood to be a set of fixed-sized buffers. I realise that
> > > this change was made in the past after some discussion, but one of the key points
> > > there [at least to my reading] was that - even though nobody actually made a
> > > concrete case where they had variable-sized buffers - having support for them
> > > made no performance difference.
> > > 
> > > The latter part of that has now changed, and supporting variable-sized mbufs
> > > from an mbuf pool has a perf impact. Do we definitely need that functionality,
> > > because the easiest fix here is just to move the rxrearm marker back above
> > > mbuf_len as it was originally in releases like 1.8?  
> > 
> > And initialize the buf_len with mp->elt_size - sizeof(struct rte_mbuf).
> > Right?
> > 
> > I don't have a strong opinion on this, I can do this if there is no
> > objection on this. Let me know.
> > 
> > However, I do see in future, "buf_len" may belong at the end of the first 64 byte
> > cache line as currently "port" is defined as uint8_t, IMO, that is less.
> > We may need to increase that uint16_t. The reason why I think that
> > because, Currently in ThunderX HW, we do have 128VFs per socket for
> > built-in NIC, So, the two node configuration and one external PCIe NW card
> > configuration can easily go beyond 256 ports.
> >   
> Ok, good point. If you think it's needed, and if we are changing the mbuf
> structure, it might be a good time to extend that field while you are at it, save
> a second ABI break later on.

> 
> /Bruce
> 
> > > 
> > > Regards,
> > > /Bruce
> > > 
> > > Ref: http://dpdk.org/ml/archives/dev/2014-December/009432.html
> > >   



-- 
   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] mbuf: make rearm_data address naturally aligned
  2016-05-19  8:50  3%     ` Bruce Richardson
  2016-05-19 11:54  0%       ` Jan Viktorin
@ 2016-05-19 12:18  0%       ` Ananyev, Konstantin
  1 sibling, 0 replies; 200+ results
From: Ananyev, Konstantin @ 2016-05-19 12:18 UTC (permalink / raw)
  To: Richardson, Bruce, Jerin Jacob; +Cc: dev, thomas.monjalon, viktorin, jianbo.liu


Hi everyone,
 
> On Thu, May 19, 2016 at 12:20:16AM +0530, Jerin Jacob wrote:
> > On Wed, May 18, 2016 at 05:43:00PM +0100, Bruce Richardson wrote:
> > > On Wed, May 18, 2016 at 07:27:43PM +0530, Jerin Jacob wrote:
> > > > To avoid multiple stores on fast path, Ethernet drivers
> > > > aggregate the writes to data_off, refcnt, nb_segs and port
> > > > to an uint64_t data and write the data in one shot
> > > > with uint64_t* at &mbuf->rearm_data address.
> > > >
> > > > Some of the non-IA platforms have store operation overhead
> > > > if the store address is not naturally aligned.This patch
> > > > fixes the performance issue on those targets.
> > > >
> > > > Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
> > > > ---
> > > >
> > > > Tested this patch on IA and non-IA(ThunderX) platforms.
> > > > This patch shows 400Kpps/core improvement on ThunderX + ixgbe + vector environment.
> > > > and this patch does not have any overhead on IA platform.
> > > >
> > > > Have tried an another similar approach by replacing "buf_len" with "pad"
> > > > (in this patch context),
> > > > Since it has additional overhead on read and then mask to keep "buf_len" intact,
> > > > not much improvement is not shown.
> > > > ref: http://dpdk.org/ml/archives/dev/2016-May/038914.html
> > > >
> > > > ---
> > > While this will work and from your tests doesn't seem to have a performance
> > > impact, I'm not sure I particularly like it. It's extending out the end of
> > > cacheline0 of the mbuf by 16 bytes, though I suppose it's not technically using
> > > up any more space of it.
> >
> > Extending by 2 bytes. Right ?. Yes, I guess, Now we using only 56 out of 64 bytes
> > in the first 64-byte cache line.
> >
> > >
> > > What I'm wondering about though, is do we have any usecases where we need a
> > > variable buf_len for packets for RX. These mbufs come directly from a mempool,
> > > which is generally understood to be a set of fixed-sized buffers. I realise that
> > > this change was made in the past after some discussion, but one of the key points
> > > there [at least to my reading] was that - even though nobody actually made a
> > > concrete case where they had variable-sized buffers - having support for them
> > > made no performance difference.

I was going to point to vhost zcp support, but as Thomas pointed out
that functionality was removed  from dpdk.org recently.
So I am not aware does such case exist right now in the 'real world' or not.
Though I still think RX function should leave buf_len field intact. 

> > >
> > > The latter part of that has now changed, and supporting variable-sized mbufs
> > > from an mbuf pool has a perf impact. Do we definitely need that functionality,
> > > because the easiest fix here is just to move the rxrearm marker back above
> > > mbuf_len as it was originally in releases like 1.8?
> >
> > And initialize the buf_len with mp->elt_size - sizeof(struct rte_mbuf).
> > Right?
> >
> > I don't have a strong opinion on this, I can do this if there is no
> > objection on this. Let me know.
> >
> > However, I do see in future, "buf_len" may belong at the end of the first 64 byte
> > cache line as currently "port" is defined as uint8_t, IMO, that is less.
> > We may need to increase that uint16_t. The reason why I think that
> > because, Currently in ThunderX HW, we do have 128VFs per socket for
> > built-in NIC, So, the two node configuration and one external PCIe NW card
> > configuration can easily go beyond 256 ports.

I wonder does anyone really use mbuf port field?
My though was - could we to drop it completely?
Actually, after discussing it with Bruce offline, an interesting idea came out:
if we'll drop port and make mbuf_prefree() to reset nb_segs=1, then
we can reduce RX rearm_data to 4B. So with that layout:

struct rte_mbuf {

         MARKER cacheline0;

        void *buf_addr;           
        phys_addr_t buf_physaddr; 
        uint16_t buf_len;
        uint8_t nb_segs;
        uint8_t reserved_1byte;   /* former port */
        
        MARKER32 rearm_data;
        uint16_t data_off;
       uint16_t refcnt;
       
        uint64_t ol_flags;
        ...

We can keep buf_len at its place and avoid 2B gap, while making rearm_data
4B long and 4B aligned.

Another similar alternative, is to make mbuf_prefree() to set refcnt=1
(as it update it anyway). Then we can remove refcnt from the RX rearm_data,
and again make rearm_data 4B long and 4B aligned:

struct rte_mbuf {

         MARKER cacheline0;

        void *buf_addr;           
        phys_addr_t buf_physaddr; 
        uint16_t buf_len;
        uint16_t refcnt;

        MARKER32 rearm_data;
        uint16_t data_off;
        uint8_t nb_segs;
        uint8_t port;
        
        uint64_t ol_flags;
         ..

As additional plus, __rte_mbuf_raw_alloc() wouldn't need to modify mbuf contents at all -
which probably is a good thing.
As a drawback - we'll have a free mbufs in pool with refcnt==1, which probably reduce
debug ability of the mbuf code.  

Konstantin

> >
> Ok, good point. If you think it's needed, and if we are changing the mbuf
> structure, it might be a good time to extend that field while you are at it, save
> a second ABI break later on.
> 
> /Bruce
> 
> > >
> > > Regards,
> > > /Bruce
> > >
> > > Ref: http://dpdk.org/ml/archives/dev/2014-December/009432.html
> > >

^ permalink raw reply	[relevance 0%]

Results 1801-2000 of ~18000   |  | reverse | sort options + mbox downloads above
-- links below jump to the message on this page --
2016-01-04 14:46     [dpdk-dev] [PATCH] vhost: remove lockless enqueue to the virtio ring Huawei Xie
2016-01-05  7:16     ` Xie, Huawei
2016-03-14 23:13  0%   ` Thomas Monjalon
2016-03-16  8:20  0%     ` Xie, Huawei
2016-03-16  8:30  0%       ` Yuanhan Liu
2016-01-11  7:07     [dpdk-dev] [PATCH 0/4] Support VxLAN & NVGRE checksum off-load on X550 Wenzhuo Lu
2016-03-10  2:42     ` [dpdk-dev] [PATCH v9 0/5] " Wenzhuo Lu
2016-03-10  2:42       ` [dpdk-dev] [PATCH v9 1/5] lib/librte_ether: change function name of tunnel port config Wenzhuo Lu
2016-03-11 23:02  3%     ` Thomas Monjalon
2016-01-29 14:08     [dpdk-dev] [PATCH 0/9] prepare for rte_device / rte_driver David Marchand
2016-04-20 11:44     ` [dpdk-dev] [PATCH v2 00/17] " David Marchand
2016-04-20 11:44  3%   ` [dpdk-dev] [PATCH v2 04/17] eal: remove duplicate function declaration David Marchand
2016-02-12 18:36     [dpdk-dev] [PATCH v4] mempool: reduce rte_mempool structure size Keith Wiles
2016-04-14  9:42  2% ` [dpdk-dev] [PATCH v5] " Olivier Matz
2016-04-14 13:28  0%   ` Wiles, Keith
2016-04-14 13:53  0%   ` Wiles, Keith
2016-05-17  5:31  0%   ` Thomas Monjalon
2016-02-16  4:15     [dpdk-dev] [PATCH v2] PCI: ABI change request for adding new field in rte_pci_id structure Ziye Yang
2016-02-17  1:54     ` [dpdk-dev] [PATCH v3] " Ziye Yang
2016-02-17 10:14       ` Bruce Richardson
2016-04-05 15:31  4%     ` Thomas Monjalon
2016-02-17 14:20     [dpdk-dev] [PATCH 0/3] ethdev: add helper functions to get eth_dev and dev private data Ferruh Yigit
2016-03-10  0:00     ` Thomas Monjalon
2016-03-10 13:37  0%   ` Ferruh Yigit
2016-02-19  6:32     [dpdk-dev] [PATCH RFC 0/4] Thread safe rte_vhost_enqueue_burst() Ilya Maximets
2016-02-19  6:32     ` [dpdk-dev] [PATCH RFC 2/4] vhost: make buf vector for scatter RX local Ilya Maximets
2016-02-19  7:06       ` Yuanhan Liu
2016-04-05  5:47  4%     ` [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local Yuanhan Liu
2016-04-05  8:37  0%       ` Thomas Monjalon
2016-04-05 14:06  0%         ` Yuanhan Liu
2016-04-06  4:14  0%       ` Flavio Leitner
2016-02-24 17:08     [dpdk-dev] [PATCH v2 0/2] add support for buffered tx to ethdev Tomasz Kulasek
2016-03-10 10:57  4% ` [dpdk-dev] [PATCH v3 " Tomasz Kulasek
2016-03-10 11:31  0%   ` Ananyev, Konstantin
2016-03-10 16:01  0%     ` Jastrzebski, MichalX K
2016-03-10 17:19  4%   ` [dpdk-dev] [PATCH v4 " Tomasz Kulasek
2016-03-07  8:12     [dpdk-dev] [PATCH v2 0/3] i40e setting ether type of VLANs Helin Zhang
2016-03-10 16:36  4% ` [dpdk-dev] [PATCH v3 0/2] " Helin Zhang
2016-03-10 16:36  7%   ` [dpdk-dev] [PATCH v3 1/2] ethdev: add vlan type for setting ether type Helin Zhang
2016-03-11  2:36  0%   ` [dpdk-dev] [PATCH v3 0/2] i40e setting ether type of VLANs Lu, Wenzhuo
2016-03-11  8:49  4%   ` [dpdk-dev] [PATCH v4 " Helin Zhang
2016-03-11  8:49  7%     ` [dpdk-dev] [PATCH v4 1/2] ethdev: add vlan type for setting ether type Helin Zhang
2016-03-11 11:19  3%       ` Panu Matilainen
2016-03-11 11:20  0%         ` Thomas Monjalon
2016-03-11 14:17  4%           ` Zhang, Helin
2016-03-11 14:20  3%             ` Thomas Monjalon
2016-03-11 16:50  5%     ` [dpdk-dev] [PATCH v5 0/2] i40e setting ether type of VLANs Helin Zhang
2016-03-11 16:50  7%       ` [dpdk-dev] [PATCH v5 1/2] ethdev: add vlan type for setting ether type Helin Zhang
2016-03-09  9:50     [dpdk-dev] [PATCH v3 0/4] external mempool manager David Hunt
2016-04-14 13:57  2% ` [dpdk-dev] [PATCH v4 0/3] " Olivier Matz
2016-03-09 16:19     [dpdk-dev] [RFC 00/35] mempool: rework memory allocation Olivier Matz
2016-03-17  9:05 15% ` [dpdk-dev] [PATCH] doc: mempool ABI deprecation notice for 16.07 Olivier Matz
2016-04-04 14:38  4%   ` Thomas Monjalon
2016-04-05  9:27  4%     ` Hunt, David
2016-04-05 14:08  4%       ` Wiles, Keith
2016-04-05 15:17  4%         ` Thomas Monjalon
2016-04-14 10:19  2% ` [dpdk-dev] [PATCH 00/36] mempool: rework memory allocation Olivier Matz
2016-04-14 13:50  0%   ` Wiles, Keith
2016-04-14 14:01  0%     ` Olivier MATZ
2016-04-14 14:03  0%       ` Wiles, Keith
2016-05-18 11:04  2%   ` [dpdk-dev] [PATCH v3 00/35] " Olivier Matz
2016-05-18 11:04 10%     ` [dpdk-dev] [PATCH v3 35/35] doc: update release notes about mempool allocation Olivier Matz
2016-03-10  3:25     [dpdk-dev] [PATCH v4 00/12] extend flow director fields in i40e driver Jingjing Wu
2016-03-21  6:18  4% ` [dpdk-dev] [PATCH v5 0/9] " Jingjing Wu
2016-03-21  6:18 18%   ` [dpdk-dev] [PATCH v5 1/9] ethdev: extend flow director for input selection Jingjing Wu
2016-03-22 22:05  0%     ` Thomas Monjalon
2016-03-23  0:42  0%       ` Wu, Jingjing
2016-03-23  8:45  0%         ` Thomas Monjalon
2016-03-23 13:07  4%   ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Jingjing Wu
2016-03-23 13:07 19%     ` [dpdk-dev] [PATCH v6 1/9] ethdev: extend flow director for input selection Jingjing Wu
2016-03-23 14:46  0%     ` [dpdk-dev] [PATCH v6 0/9] extend flow director fields in i40e driver Bruce Richardson
2016-03-10 10:53     [dpdk-dev] [PATCH 1/3] scripts: support parallel building in validate-abi.sh via -j[N] option Panu Matilainen
2016-03-10 10:53     ` [dpdk-dev] [PATCH 2/3] scripts: avoid editing defconfig_* files in validate-abi.sh Panu Matilainen
2016-03-10 12:25  4%   ` Ferruh Yigit
2016-03-10 12:36  4%     ` Panu Matilainen
2016-03-10 10:53 15% ` [dpdk-dev] [PATCH 3/3] scripts: ignore self-generated directories in validate-abi startup check Panu Matilainen
2016-03-10 12:22  4%   ` Ferruh Yigit
2016-03-10 12:29  4%     ` Panu Matilainen
2016-03-10 12:34  4%       ` Ferruh Yigit
2016-03-10 12:39  4%         ` Panu Matilainen
2016-03-10 12:47  4%           ` Ferruh Yigit
2016-03-10 12:52  7% ` [dpdk-dev] [PATCH 1/3] scripts: support parallel building in validate-abi.sh via -j[N] option Ferruh Yigit
2016-03-10 11:55  7% [dpdk-dev] [PATCH] doc: add mempool mgr ABI deprication notice David Hunt
2016-03-10 12:37  4% ` Olivier MATZ
2016-03-10 13:15  4%   ` Bruce Richardson
2016-03-10 13:56  4%     ` Wiles, Keith
2016-03-10 14:55  4%       ` Thomas Monjalon
2016-04-05 14:06  4%         ` Wiles, Keith
2016-04-04 14:49  4%       ` Thomas Monjalon
2016-03-10 16:23  4% ` Mcnamara, John
2016-03-10 14:44     [dpdk-dev] [PATCH] mempool: allow for user-owned mempool caches Lazaros Koromilas
2016-03-21 12:22  4% ` Olivier Matz
2016-03-21 13:49  3%   ` Wiles, Keith
2016-03-24 14:35  0%     ` Lazaros Koromilas
2016-03-24 14:58  0%       ` Venkatesan, Venky
2016-03-24 15:03  0%       ` Wiles, Keith
2016-03-17 18:08     [dpdk-dev] [PATCH v11 0/8] ethdev: 100G and link speed API refactoring Thomas Monjalon
2016-03-25 19:42     ` [dpdk-dev] [PATCH v12 " Thomas Monjalon
2016-03-25 19:42  3%   ` [dpdk-dev] [PATCH v12 6/8] ethdev: redesign link speed config Thomas Monjalon
2016-03-26  1:27       ` [dpdk-dev] [PATCH v13 0/8] ethdev: 100G and link speed API refactoring Marc Sune
2016-03-26  1:27         ` [dpdk-dev] [PATCH v13 4/8] ethdev: rename link speed constants Marc Sune
2016-04-06  8:34  1%       ` Weglicki, MichalX
2016-04-06  8:52  0%         ` Thomas Monjalon
2016-04-06  9:16  3%           ` Weglicki, MichalX
2016-04-06  9:34  3%             ` Thomas Monjalon
2016-03-26  1:27  3%     ` [dpdk-dev] [PATCH v13 6/8] ethdev: redesign link speed config Marc Sune
2016-03-31 22:12         ` [dpdk-dev] [PATCH v14 0/8] ethdev: 100G and link speed API refactoring Marc Sune
2016-03-31 22:12  3%       ` [dpdk-dev] [PATCH v14 6/8] ethdev: redesign link speed config Marc Sune
2016-03-18 17:16     [dpdk-dev] DPDK and HW offloads Stephen Hemminger
2016-03-18 18:00     ` Thomas Monjalon
2016-03-20 14:17       ` Zhang, Helin
2016-03-20 19:18         ` Thomas Monjalon
2016-03-21 14:52           ` Bruce Richardson
2016-03-21 15:26  3%         ` Kyle Larose
2016-03-22  5:50  0%           ` Qiu, Michael
2016-03-22 10:19  0%             ` Bruce Richardson
2016-03-23  2:47  0%               ` Qiu, Michael
2016-03-25 16:29  5% [dpdk-dev] [PATCH] doc: postpone flow director changes planned for cxgbe Thomas Monjalon
2016-03-31  8:21  2% [dpdk-dev] 16.07 Roadmap O'Driscoll, Tim
2016-03-31 13:28 16% [dpdk-dev] [PATCH] doc: announce ABI change for rte_port_source_params structure Fan Zhang
2016-03-31 13:29 16% Fan Zhang
2016-04-05 15:45  7% ` Thomas Monjalon
2016-04-05 21:16  4% ` Singh, Jasvinder
2016-04-06  8:51  4%   ` Azarewicz, PiotrX T
2016-04-07 21:24  4%     ` Thomas Monjalon
2016-04-12 12:39  4%       ` Thomas Monjalon
2016-04-05  9:23  9% [dpdk-dev] [PATCH] doc: announce ABI changes for user-owned mempool caches Lazaros Koromilas
2016-04-05 15:42  4% ` Olivier Matz
2016-04-08 14:01  4%   ` Hunt, David
2016-04-10  9:55  4%     ` Thomas Monjalon
2016-04-05 13:56     [dpdk-dev] DPDK namespace Thomas Monjalon
2016-04-05 14:13  3% ` Trahe, Fiona
2016-04-05 14:31  0%   ` Trahe, Fiona
2016-04-05 14:31  0%   ` Arnon Warshavsky
2016-04-06  5:26  0%     ` Yuanhan Liu
2016-04-06 12:07  0%       ` Panu Matilainen
2016-04-06 12:34  0%         ` Ananyev, Konstantin
2016-04-06 14:36  0%         ` Wiles, Keith
2016-04-06 20:21             ` Dave Neary
2016-04-07  8:22  3%           ` Marc
2016-04-07  9:18     ` Thomas Monjalon
2016-04-07  9:33  3%   ` Panu Matilainen
2016-04-07 10:16  5%     ` Marc Sune
2016-04-07 11:51  9%       ` [dpdk-dev] On DPDK ABI policy Panu Matilainen
2016-04-07 21:52  4%         ` Matthew Hall
2016-04-08  8:29  4%           ` Marc Sune
2016-04-08  8:47  9%         ` Marc Sune
2016-04-07 21:48  0%       ` [dpdk-dev] DPDK namespace Matthew Hall
2016-04-05 17:58  9% [dpdk-dev] [PATCH] doc: announce xstats api change for 16.07 Harry van Haaren
2016-04-05 18:45  0% ` Thomas Monjalon
2016-04-06  9:02  0%   ` Van Haaren, Harry
2016-04-06  9:22  0%     ` Thomas Monjalon
2016-04-06 11:16  3%       ` Van Haaren, Harry
2016-04-06 12:14  0%         ` Thomas Monjalon
2016-04-06 13:49  0%           ` David Harton (dharton)
2016-04-06 14:00  0% ` David Harton (dharton)
2016-04-06  5:11  0% [dpdk-dev] [RFC] vhost-user public struct refactor (was Re: [PATCH RFC 2/4] vhost: make buf vector for scatter RX) local Ilya Maximets
2016-04-06  6:53 15% [dpdk-dev] [PATCH] vhost: ABI/API change announcement due to refactor Yuanhan Liu
2016-04-07  7:12  7% ` Panu Matilainen
2016-04-10  9:58  4%   ` Thomas Monjalon
2016-04-10 10:02  4%     ` Thomas Monjalon
2016-04-06  9:53     [dpdk-dev] Fw: dpdk-armv7-testing - Build # 43 - Failure! Jan Viktorin
2016-04-06 10:05     ` Thomas Monjalon
2016-04-06 10:18  4%   ` Jan Viktorin
2016-04-06 16:58     [dpdk-dev] ovs crash when running traffic from VM to VM over DPDK and vhostuser Yuanhan Liu
2016-05-02 17:40  3% ` Yi Ba
2016-04-07 15:33  5% [dpdk-dev] [PATCH] doc: announce API changes for device objects David Marchand
2016-04-07 15:46  0% ` Jan Viktorin
2016-04-07 17:00  0%   ` David Marchand
2016-04-07 17:09  3%     ` Jan Viktorin
2016-04-07 17:24  0%       ` David Marchand
2016-04-07 16:02  8% [dpdk-dev] [PATCH v1] doc: fix release notes for 16.04 John McNamara
2016-04-12 12:01  6% [dpdk-dev] [PATCH v1] doc: add template release notes for 16.11 John McNamara
2016-04-12 12:55  6% [dpdk-dev] [PATCH v1] doc: add template release notes for 16.07 John McNamara
2016-04-14  9:44  4% [dpdk-dev] [RFC 0/2] add new fields to rte_eth_dev_info structure Reshma Pattan
2016-04-14  9:44  9% ` [dpdk-dev] [RFC 1/2] doc: announce ABI change for " Reshma Pattan
2016-04-15  9:42  4%   ` Mcnamara, John
2016-04-15 10:02  8%   ` Thomas Monjalon
2016-04-14  9:44     ` [dpdk-dev] [RFC 2/2] librte_ether: add new fields to rte_eth_dev_info struct Reshma Pattan
2016-04-15 10:36  3%   ` Thomas Monjalon
2016-04-15 11:32  0%     ` Ananyev, Konstantin
2016-04-14 18:33 25% [dpdk-dev] [PATCH] port: bump ABI for pcap file support Thomas Monjalon
2016-04-15 10:32  4% ` Dumitrescu, Cristian
2016-04-20  9:55  4%   ` Thomas Monjalon
2016-04-14 21:33  4% [dpdk-dev] [PATCH] pci: remove deprecated specific config Thomas Monjalon
2016-04-15 14:44  3% [dpdk-dev] [RFC PATCH v1 0/3] Remove string operations from xstats Remy Horton
2016-04-20 16:03  0% ` David Harton (dharton)
2016-04-28 14:56  0%   ` Tahhan, Maryam
2016-04-28 15:58  0%     ` David Harton (dharton)
2016-04-29 12:52  0% ` David Harton (dharton)
2016-04-15 22:33 15% [dpdk-dev] [PATCH] mk: do not enforce any specific ARM ABI Jan Viktorin
2016-05-02 15:47  4% ` Thomas Monjalon
2016-04-19 11:11     [dpdk-dev] perfomance of rte_lpm rule subsystem Александр Киселев
2016-04-19 15:46  3% ` Stephen Hemminger
2016-04-19 20:46  0%   ` Vladimir Medvedkin
2016-05-02 19:38  0%   ` Александр Киселев
2016-04-19 14:03  4% [dpdk-dev] [PATCH] ethdev: remove deprecated statistics Thomas Monjalon
2016-04-20  9:47  4% ` [dpdk-dev] [PATCH v2] " Thomas Monjalon
2016-04-25  9:18  2% [dpdk-dev] [RFC] eal: provide option to set vhost_user socket owner/permissions Christian Ehrhardt
2016-04-26  4:16  0% ` Yuanhan Liu
2016-04-26  7:24  0%   ` Christian Ehrhardt
2016-04-26  4:45  3% [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix Yuanhan Liu
2016-04-26  4:45  2% ` [dpdk-dev] [PATCH 7/7] examples/vhost: switch_worker cleanup Yuanhan Liu
2016-04-28  5:45  0% ` [dpdk-dev] [PATCH 0/7] vhost/example cleanup/fix Wang, Zhihong
2016-04-28  6:09  0%   ` Yuanhan Liu
     [not found]     ` <1462224230-19460-1-git-send-email-yuanhan.liu@linux.intel.com>
2016-05-02 21:23  2%   ` [dpdk-dev] [PATCH v2 7/8] examples/vhost: switch_worker cleanup Yuanhan Liu
2016-05-09 18:06  0%   ` [dpdk-dev] [PATCH v2 0/8] vhost/example cleanup/fix Yuanhan Liu
2016-05-02 22:25  8% [dpdk-dev] [PATCH 00/16] vhost ABI/API refactoring Yuanhan Liu
2016-05-02 22:25  2% ` [dpdk-dev] [PATCH 10/16] vhost: export vid as the only interface to applications Yuanhan Liu
2016-05-10 16:17  0%   ` Rich Lane
2016-05-10 16:39  0%     ` Yuanhan Liu
2016-05-02 22:25  4% ` [dpdk-dev] [PATCH 14/16] vhost: reserve few more space for future extension Yuanhan Liu
2016-05-13  5:24  8% ` [dpdk-dev] [PATCH v2 00/19] vhost ABI/API refactoring Yuanhan Liu
2016-05-13  5:25  7%   ` [dpdk-dev] [PATCH v2 11/19] vhost: introduce new API to export queue free entries Yuanhan Liu
2016-05-13  5:25  3%   ` [dpdk-dev] [PATCH v2 12/19] vhost: remove dependency on priv field Yuanhan Liu
2016-05-13  5:25 12%   ` [dpdk-dev] [PATCH v2 13/19] vhost: export vid as the only interface to applications Yuanhan Liu
2016-05-13  5:25  4%   ` [dpdk-dev] [PATCH v2 17/19] vhost: reserve few more space for future extension Yuanhan Liu
2016-05-03  0:46  3% [dpdk-dev] [PATCH 0/3] [RFC] vhost: micro vhost optimization Yuanhan Liu
2016-05-10 21:49  0% ` Rich Lane
2016-05-10 22:08  0%   ` Yuanhan Liu
2016-05-06 10:55     [dpdk-dev] [PATCH 0/5] add packet capture framework Reshma Pattan
2016-05-06 10:55  6% ` [dpdk-dev] [PATCH 5/5] doc: update doc for " Reshma Pattan
2016-05-10  9:39     ` [dpdk-dev] [PATCHv2 0/5] add " Reshma Pattan
2016-05-10  9:40  6%   ` [dpdk-dev] [PATCHv2 5/5] doc: update doc for " Reshma Pattan
2016-05-17 16:37       ` [dpdk-dev] [PATCH v3 0/8] add " Reshma Pattan
2016-05-17 16:37  6%     ` [dpdk-dev] [PATCH v3 8/8] doc: update doc for " Reshma Pattan
2016-05-06 20:05     [dpdk-dev] [PATCH v1] hash: add tsx support for cuckoo hash Shen Wei
2016-05-07  4:56  3% ` Stephen Hemminger
2016-05-09 16:51  4%   ` Shen, Wei1
2016-05-07  6:40  3% [dpdk-dev] [PATCH 0/6] vhost: add vhost-user client mode and reconnect ability Yuanhan Liu
2016-05-10  3:23  3% ` Xu, Qian Q
2016-05-10 17:41  0%   ` Yuanhan Liu
2016-05-13  6:16  3% ` [dpdk-dev] [PATCH v2 " Yuanhan Liu
2016-05-10  9:13     [dpdk-dev] Ring PMD: why are stats counters atomic? Mauricio Vásquez
2016-05-10  9:36     ` Bruce Richardson
2016-05-16 13:12       ` Mauricio Vásquez
2016-05-16 13:16  3%     ` Bruce Richardson
2016-05-10 10:11     [dpdk-dev] [PATCH] sched: fix useless call Daniel Mrzyglod
2016-05-10 17:18     ` Dumitrescu, Cristian
2016-05-11  9:46       ` Ferruh Yigit
2016-05-13 10:12         ` Thomas Monjalon
2016-05-13 11:04  4%       ` Dumitrescu, Cristian
2016-05-11  6:08     [dpdk-dev] [PATCH] pci: Add the class_id support in pci probe Ziye Yang
2016-05-11 15:21  3% ` Stephen Hemminger
2016-05-11 15:34  3%   ` Richardson, Bruce
2016-05-13 13:27 16% [dpdk-dev] [PATCH] doc: move rel_notes instructions as comments Olivier Matz
2016-05-16 13:18  9% [dpdk-dev] [PATCH 0/2] doc: announce ABI change of struct rte_port_source_params Fan Zhang
2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 1/2] " Fan Zhang
2016-05-16 13:18 18% ` [dpdk-dev] [PATCH 2/2] doc: announce ABI change of struct rte_port_sink_params Fan Zhang
2016-05-16 13:57  9%   ` Panu Matilainen
2016-05-16 20:41     [dpdk-dev] [PATCH 0/4] Implement pmd hardware support exports Neil Horman
2016-05-16 20:41  2% ` [dpdk-dev] [PATCH 1/4] pmdinfo: Add buildtools and pmdinfo utility Neil Horman
2016-05-18 13:57     [dpdk-dev] [PATCH] mbuf: make rearm_data address naturally aligned Jerin Jacob
2016-05-18 16:43     ` Bruce Richardson
2016-05-18 18:50       ` Jerin Jacob
2016-05-19  8:50  3%     ` Bruce Richardson
2016-05-19 11:54  0%       ` Jan Viktorin
2016-05-19 12:18  0%       ` Ananyev, Konstantin

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).