From: Kevin Traynor <ktraynor@redhat.com>
To: Bing Zhao <bingz@nvidia.com>
Cc: Dariusz Sosnowski <dsosnowski@nvidia.com>, dpdk stable <stable@dpdk.org>
Subject: patch 'net/mlx5: fix unneeded stub flow table allocation' has been queued to stable release 24.11.2
Date: Thu, 13 Feb 2025 09:57:49 +0000 [thread overview]
Message-ID: <20250213095933.362078-22-ktraynor@redhat.com> (raw)
In-Reply-To: <20250213095933.362078-1-ktraynor@redhat.com>
Hi,
FYI, your patch has been queued to stable release 24.11.2
Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
It will be pushed if I get no objections before 02/17/25. So please
shout if anyone has objections.
Also note that after the patch there's a diff of the upstream commit vs the
patch applied to the branch. This will indicate if there was any rebasing
needed to apply to the stable branch. If there were code changes for rebasing
(ie: not only metadata diffs), please double check that the rebase was
correctly done.
Queued patches are on a temporary branch at:
https://github.com/kevintraynor/dpdk-stable
This queued commit can be viewed at:
https://github.com/kevintraynor/dpdk-stable/commit/a3ee0fe3555b8cc0daa6bf2a68002c7f73845785
Thanks.
Kevin
---
From a3ee0fe3555b8cc0daa6bf2a68002c7f73845785 Mon Sep 17 00:00:00 2001
From: Bing Zhao <bingz@nvidia.com>
Date: Tue, 26 Nov 2024 11:25:39 +0200
Subject: [PATCH] net/mlx5: fix unneeded stub flow table allocation
[ upstream commit 3cd695c34528571c378c5f6be7ff81d3cca9a84c ]
The HWS non-template flow API is reusing some implementation of
template API to unify code logic. So for each rule creation, a stub
/ temporary table is used in order to reuse the actions construction.
Since this is temporary and used only internally, there is no need to
save the table permanently. Only parts of them are mandatory, so the
allocation / free from the heap of RTE memory is a waste and causes
a lot of overhead. By using the pre-allocated workspace and set the
needed fields expliticly will save the overhead and help to speed up
the rule insertion rate.
Fixes: 27d171b88031 ("net/mlx5: abstract flow action and enable reconfigure")
Signed-off-by: Bing Zhao <bingz@nvidia.com>
Acked-by: Dariusz Sosnowski <dsosnowski@nvidia.com>
---
drivers/net/mlx5/mlx5_flow.c | 11 +++++++++--
drivers/net/mlx5/mlx5_flow.h | 3 +++
drivers/net/mlx5/mlx5_flow_hw.c | 19 +++++++++++--------
3 files changed, 23 insertions(+), 10 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index 16ddd05448..9203643300 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -8271,6 +8271,10 @@ flow_alloc_thread_workspace(void)
size_t data_size = RTE_ALIGN(sizeof(struct mlx5_flow_workspace), sizeof(long));
size_t rss_queue_array_size = sizeof(uint16_t) * RTE_ETH_RSS_RETA_SIZE_512;
- struct mlx5_flow_workspace *data = calloc(1, data_size +
- rss_queue_array_size);
+ size_t alloc_size = data_size + rss_queue_array_size;
+#ifdef HAVE_MLX5_HWS_SUPPORT
+ /* Dummy table size for the non-template API. */
+ alloc_size += sizeof(struct rte_flow_template_table);
+#endif
+ struct mlx5_flow_workspace *data = calloc(1, alloc_size);
if (!data) {
@@ -8279,4 +8283,7 @@ flow_alloc_thread_workspace(void)
}
data->rss_desc.queue = RTE_PTR_ADD(data, data_size);
+#ifdef HAVE_MLX5_HWS_SUPPORT
+ data->table = RTE_PTR_ADD(data->rss_desc.queue, rss_queue_array_size);
+#endif
return data;
}
diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
index bcc2782460..757bbf73c1 100644
--- a/drivers/net/mlx5/mlx5_flow.h
+++ b/drivers/net/mlx5/mlx5_flow.h
@@ -1920,4 +1920,7 @@ struct mlx5_flow_workspace {
struct mlx5_flow_meter_policy *final_policy;
/* The final policy when meter policy is hierarchy. */
+#ifdef HAVE_MLX5_HWS_SUPPORT
+ struct rte_flow_template_table *table;
+#endif
uint32_t skip_matcher_reg:1;
/* Indicates if need to skip matcher register in translate. */
diff --git a/drivers/net/mlx5/mlx5_flow_hw.c b/drivers/net/mlx5/mlx5_flow_hw.c
index 1de6b889a7..2b62711413 100644
--- a/drivers/net/mlx5/mlx5_flow_hw.c
+++ b/drivers/net/mlx5/mlx5_flow_hw.c
@@ -13518,5 +13518,4 @@ flow_hw_translate_flow_actions(struct rte_eth_dev *dev,
uint32_t src_group = 0;
enum mlx5dr_table_type table_type;
- struct rte_flow_template_table *table = NULL;
struct mlx5_flow_group grp;
struct rte_flow_actions_template *at = NULL;
@@ -13532,4 +13531,8 @@ flow_hw_translate_flow_actions(struct rte_eth_dev *dev,
memset(masks, 0, sizeof(masks));
memset(mask_conf, 0, sizeof(mask_conf));
+ /* Only set the needed fields explicitly. */
+ struct mlx5_flow_workspace *wks = mlx5_flow_push_thread_workspace();
+ struct rte_flow_template_table *table;
+
/*
* Notice All direct actions will be unmasked,
@@ -13541,4 +13544,10 @@ flow_hw_translate_flow_actions(struct rte_eth_dev *dev,
* and not during action construct.
*/
+ if (!wks)
+ return rte_flow_error_set(error, ENOMEM,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+ NULL,
+ "failed to push flow workspace");
+ table = wks->table;
flow_nta_build_template_mask(actions, masks, mask_conf);
/* The group in the attribute translation was done in advance. */
@@ -13552,9 +13561,4 @@ flow_hw_translate_flow_actions(struct rte_eth_dev *dev,
else
table_type = MLX5DR_TABLE_TYPE_NIC_RX;
- /* TODO: consider to reuse the workspace per thread. */
- table = mlx5_malloc(MLX5_MEM_ZERO, sizeof(*table), 0, SOCKET_ID_ANY);
- if (!table)
- return rte_flow_error_set(error, ENOMEM, RTE_FLOW_ERROR_TYPE_ACTION,
- actions, "Failed to allocate dummy table");
at = __flow_hw_actions_template_create(dev, &template_attr, actions, masks, true, error);
if (!at) {
@@ -13593,8 +13597,7 @@ end:
else
__flow_hw_act_data_flush(dev, hw_acts);
- if (table)
- mlx5_free(table);
if (at)
mlx5_free(at);
+ mlx5_flow_pop_thread_workspace();
return ret;
}
--
2.48.1
---
Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- - 2025-02-12 17:29:36.941546295 +0000
+++ 0022-net-mlx5-fix-unneeded-stub-flow-table-allocation.patch 2025-02-12 17:29:34.258945381 +0000
@@ -1 +1 @@
-From 3cd695c34528571c378c5f6be7ff81d3cca9a84c Mon Sep 17 00:00:00 2001
+From a3ee0fe3555b8cc0daa6bf2a68002c7f73845785 Mon Sep 17 00:00:00 2001
@@ -5,0 +6,2 @@
+[ upstream commit 3cd695c34528571c378c5f6be7ff81d3cca9a84c ]
+
@@ -18 +19,0 @@
-Cc: stable@dpdk.org
@@ -54 +55 @@
-index 50b0e2ce47..93c2406abc 100644
+index bcc2782460..757bbf73c1 100644
next prev parent reply other threads:[~2025-02-13 10:00 UTC|newest]
Thread overview: 123+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-02-13 9:57 patch 'log: fix double free on cleanup' " Kevin Traynor
2025-02-13 9:57 ` patch 'test/ring: fix init with custom number of lcores' " Kevin Traynor
2025-02-13 9:57 ` patch 'vhost: clear ring addresses when getting vring base' " Kevin Traynor
2025-02-13 9:57 ` patch 'vhost: check GSO size validity' " Kevin Traynor
2025-02-13 9:57 ` patch 'vhost: fix log when setting max queue num' " Kevin Traynor
2025-02-13 9:57 ` patch 'vhost: reset packets count when not ready' " Kevin Traynor
2025-02-13 9:57 ` patch 'crypto/cnxk: fix build with GCC 15' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/thunderx/base: " Kevin Traynor
2025-02-13 9:57 ` patch 'examples/flow_filtering: " Kevin Traynor
2025-02-13 9:57 ` patch 'eal/x86: fix some intrinsics header include for Windows' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/ena: fix missing default LLQ policy' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/nfp: fix VF link speed' " Kevin Traynor
2025-02-13 9:57 ` patch 'test/bonding: fix active backup receive test' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/bonding: fix dedicated queue setup' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/hns3: fix mbuf freeing in simple Tx path' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/hns3: remove PVID info dump for VF' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/hns3: rename RAS module' " Kevin Traynor
2025-02-13 9:57 ` patch 'examples/flow_filtering: remove duplicate assignment' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/sfc: remove unnecessary " Kevin Traynor
2025-02-13 9:57 ` patch 'net/mlx5: fix polling CQEs' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/mlx5: fix leak of flow action data list' " Kevin Traynor
2025-02-13 9:57 ` Kevin Traynor [this message]
2025-02-13 9:57 ` patch 'net/ixgbe: fix E610 support in flow engine' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/ixgbe/base: fix TSAM checking return value' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/ixgbe/base: remove 2.5/5G from auto-negotiation for E610' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/ixgbe/base: fix driver hang in VM' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/ixgbe/base: add missing buffer copy for ACI' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/iavf: remove reset of Tx prepare function pointer' " Kevin Traynor
2025-02-13 9:57 ` patch 'net/ice: fix memory leak in scalar Rx' " Kevin Traynor
2025-02-13 9:57 ` patch 'common/cnxk: fix atomic load in batch ops' " Kevin Traynor
2025-02-13 9:57 ` patch 'common/cnxk: fix DPI mailbox structure' " Kevin Traynor
2025-02-13 9:57 ` patch 'eventdev: fix format string data type in log messages' " Kevin Traynor
2025-02-13 9:58 ` patch 'crypto/virtio: fix redundant queue free' " Kevin Traynor
2025-02-13 9:58 ` patch 'crypto/openssl: fix CMAC auth context update' " Kevin Traynor
2025-02-13 15:18 ` Wathsala Wathawana Vithanage
2025-02-13 9:58 ` patch 'cryptodev: fix C++ include' " Kevin Traynor
2025-02-13 9:58 ` patch 'crypto/virtio: fix data queues iteration' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/octeon_ep: remove useless assignment' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/enetfec: " Kevin Traynor
2025-02-13 9:58 ` patch 'net/gve: allocate Rx QPL pages using malloc' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/nfp: fix multiple PFs check from NSP' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/nfp: fix misuse of function return values' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/nfp: fix multi-PF control flag' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/nfp: fix firmware load from flash' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/ice: fix flows handling' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/cnxk: fix NIX send header L3 type' " Kevin Traynor
2025-02-13 9:58 ` patch 'gro: fix unprocessed IPv4 packets' " Kevin Traynor
2025-02-13 9:58 ` patch 'log: fix systemd journal socket open' " Kevin Traynor
2025-02-13 9:58 ` patch 'eal/linux: fix memseg length in legacy mem init' " Kevin Traynor
2025-02-13 9:58 ` patch 'service: fix getting service lcore attributes' " Kevin Traynor
2025-02-13 9:58 ` patch 'ptr_compress: fix build with Arm SVE' " Kevin Traynor
2025-02-13 9:58 ` patch 'buildtools: fix some Python regex syntax warnings' " Kevin Traynor
2025-02-13 9:58 ` patch 'dts: fix MTU set command' " Kevin Traynor
2025-02-13 9:58 ` patch 'dts: fix pass rate edge case in results' " Kevin Traynor
2025-02-13 9:58 ` patch 'use Python raw string notation' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/af_packet: fix socket close on device stop' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/mana: do not ring short doorbell for every mbuf alloc' " Kevin Traynor
2025-02-13 9:58 ` patch 'ethdev: fix functions available in new device event' " Kevin Traynor
2025-02-13 9:58 ` patch 'vhost: add null callback checks' " Kevin Traynor
2025-02-13 9:58 ` patch 'vhost: fix FD entries cleanup' " Kevin Traynor
2025-02-13 9:58 ` patch 'build: force GCC 15 to initialize padding bits' " Kevin Traynor
2025-02-13 9:58 ` patch 'random: defer seeding to EAL init' " Kevin Traynor
2025-02-13 9:58 ` patch 'power: defer lcore variable allocation' " Kevin Traynor
2025-02-13 9:58 ` patch 'eal/x86: defer power intrinsics " Kevin Traynor
2025-02-13 9:58 ` patch 'net/bnxt: fix indication of " Kevin Traynor
2025-02-13 9:58 ` patch 'net/bnxt: fix Rx handler' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/bnxt: fix crash when representor is re-attached' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/bnxt: fix dead code' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/bnxt: fix overflow' " Kevin Traynor
2025-02-13 9:58 ` Kevin Traynor
2025-02-13 9:58 ` patch 'net/bnxt: address uninitialized variables' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/mlx5: fix Netlink socket leak' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/mlx5: adjust actions per rule limitation' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/mlx5: fix crash in non-template metadata split' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/mlx5: fix flush of non-template flow rules' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/mlx5: fix GRE flow match with SWS' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix deadlock when writing i225 register' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix infinite loop' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix bitwise operation type' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: increase PHY power up delay' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: reset loop variable' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix LTR for i225' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix typo in LTR calculation' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix unused value' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix semaphore timeout " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix iterator type' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix MAC address hash bit shift' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix data type in MAC hash' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/igc/base: fix NVM data type in bit shift' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/e1000/base: fix semaphore timeout value' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/e1000/base: correct mPHY access logic' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/e1000/base: fix iterator type' " Kevin Traynor
2025-02-13 9:58 ` patch 'net/e1000/base: fix MAC address hash bit shift' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: fix data type in MAC hash' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: fix uninitialized variable' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: fix bitwise operation type' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: fix NVM data type in bit shift' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: fix reset for 82580' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: fix unchecked return' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/e1000/base: skip management check for 82575' " Kevin Traynor
2025-02-13 9:59 ` patch 'common/idpf: fix void function returning a value' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/intel: fix void functions " Kevin Traynor
2025-02-13 9:59 ` patch 'net/intel: fix build with icx' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/hns3: fix copper port initialization' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/hns3: fix reset timeout' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/cpfl: fix representor parsing log' " Kevin Traynor
2025-02-13 9:59 ` patch 'raw/cnxk_gpio: fix file descriptor leak' " Kevin Traynor
2025-02-13 9:59 ` patch 'eal: fix devargs layers parsing out of bounds' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/qede: fix debug messages array' " Kevin Traynor
2025-02-13 9:59 ` patch 'examples/ptpclient: fix message parsing' " Kevin Traynor
2025-02-13 9:59 ` patch 'net/hinic: fix flow type bitmask overflow' " Kevin Traynor
2025-02-13 9:59 ` patch 'crypto/dpaa2_sec: fix bitmask truncation' " Kevin Traynor
2025-02-13 9:59 ` patch 'crypto/dpaa_sec: " Kevin Traynor
2025-02-13 9:59 ` patch 'event/dpaa: " Kevin Traynor
2025-02-13 9:59 ` patch 'net/dpaa: " Kevin Traynor
2025-02-13 9:59 ` patch 'net/dpaa2: " Kevin Traynor
2025-02-13 9:59 ` patch 'net/qede: fix nested loops' " Kevin Traynor
2025-02-13 9:59 ` patch 'examples/l3fwd: fix socket ID check' " Kevin Traynor
2025-02-13 9:59 ` patch 'common/cnxk: fix null " Kevin Traynor
2025-02-13 9:59 ` patch 'eal/linux: remove useless assignments' " Kevin Traynor
2025-02-13 9:59 ` patch 'mempool: fix errno in empty create' " Kevin Traynor
2025-02-13 9:59 ` patch 'doc: add tested platforms with NVIDIA NICs in 24.11' " Kevin Traynor
2025-02-13 9:59 ` patch 'doc: add two more tested Intel " Kevin Traynor
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250213095933.362078-22-ktraynor@redhat.com \
--to=ktraynor@redhat.com \
--cc=bingz@nvidia.com \
--cc=dsosnowski@nvidia.com \
--cc=stable@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).