Skip to contentSkip to conversation

net/mlx5: HWS single flow counter support

PATCH Applied Tariq Toukan 4 participants 7 messages Jan 12 – Jan 15

Subsystems (4)
lore
Share & tools
lore

Tariq Toukan

author maintainer
#0 lore raw

cover net/mlx5: HWS single flow counter support

Hi,

This small series refactors the flow counter bulk initialization code and extends it so that single flow counters are also usable by hardware steering (HWS) rules.

Patches 1-2 refactor the bulk init path: first by factoring out common flow counter bulk initialization into mlx5_fc_bulk_init(), then by splitting the bitmap allocation into mlx5_fs_bulk_bitmap_alloc(), with no functional changes.

Patch 3 initializes bulk data for counters allocated via mlx5_fc_single_alloc(), so they can be safely used by HWS rules.

Regards,
Tariq

Mark Bloch (2):
net/mlx5: fs, factor out flow counter bulk init
net/mlx5: fs, split bulk init

Moshe Shemesh (1):
net/mlx5: Initialize bulk for single flow counters

 .../net/ethernet/mellanox/mlx5/core/fs_core.h |  3 +-
 .../ethernet/mellanox/mlx5/core/fs_counters.c | 47 ++++++++++++++-----
 .../net/ethernet/mellanox/mlx5/core/fs_pool.c | 16 ++++---
 .../net/ethernet/mellanox/mlx5/core/fs_pool.h |  5 +-
 .../mlx5/core/steering/hws/fs_hws_pools.c     |  8 +++-
 5 files changed, 55 insertions(+), 24 deletions(-)

base-commit: 60d8484c4cec811f5ceb6550655df74490d1a165

Signature · 1 line
2.31.1
author maintainer
#1 lore raw

PATCH 1/3 net/mlx5: fs, factor out flow counter bulk init

From: Mark Bloch <[email protected]>

Add mlx5_fc_bulk_init() to handle bulk initialization of flow counters. This change does not alter any logic, but refactors the code to remove duplicate initialization logic by centralizing it in a single function.

---
 .../ethernet/mellanox/mlx5/core/fs_counters.c    | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c +10 -6
index 83001eda3884..e14570a3d492 100644
@@ -421,6 +421,13 @@ static void mlx5_fc_init(struct mlx5_fc *counter, struct mlx5_fc_bulk *bulk,
 	counter->id = id;
 }
 
+static void mlx5_fc_bulk_init(struct mlx5_fc_bulk *fc_bulk, u32 base_id)
+{
+	fc_bulk->base_id = base_id;
+	refcount_set(&fc_bulk->hws_data.hws_action_refcount, 0);
+	mutex_init(&fc_bulk->hws_data.lock);
+}
+
 u32 mlx5_fc_get_base_id(struct mlx5_fc *counter)
 {
 	return counter->bulk->base_id;
@@ -447,12 +454,11 @@ static struct mlx5_fs_bulk *mlx5_fc_bulk_create(struct mlx5_core_dev *dev,
 
 	if (mlx5_cmd_fc_bulk_alloc(dev, alloc_bitmask, &base_id))
 		goto fs_bulk_cleanup;
-	fc_bulk->base_id = base_id;
+
+	mlx5_fc_bulk_init(fc_bulk, base_id);
 	for (i = 0; i < bulk_len; i++)
 		mlx5_fc_init(&fc_bulk->fcs[i], fc_bulk, base_id + i);
 
-	refcount_set(&fc_bulk->hws_data.hws_action_refcount, 0);
-	mutex_init(&fc_bulk->hws_data.lock);
 	return &fc_bulk->fs_bulk;
 
 fs_bulk_cleanup:
@@ -560,10 +566,8 @@ mlx5_fc_local_create(u32 counter_id, u32 offset, u32 bulk_size)
 
 	counter->type = MLX5_FC_TYPE_LOCAL;
 	counter->id = counter_id;
-	fc_bulk->base_id = counter_id - offset;
 	fc_bulk->fs_bulk.bulk_len = bulk_size;
-	refcount_set(&fc_bulk->hws_data.hws_action_refcount, 0);
-	mutex_init(&fc_bulk->hws_data.lock);
+	mlx5_fc_bulk_init(fc_bulk, counter_id - offset);
 	counter->bulk = fc_bulk;
 	refcount_set(&counter->fc_local_refcount, 1);
 	return counter;
Signature · 1 line
2.31.1
author maintainer
#2 lore raw

PATCH 2/3 net/mlx5: fs, split bulk init

In reply to Tariq Toukan

From: Mark Bloch <[email protected]>

Refactor mlx5_fs_bulk_init() by moving bitmap allocation logic into a new helper function mlx5_fs_bulk_bitmap_alloc(). This change does not alter any logic.

---
 .../ethernet/mellanox/mlx5/core/fs_counters.c    |  6 ++++--
 .../net/ethernet/mellanox/mlx5/core/fs_pool.c    | 16 ++++++++++------
 .../net/ethernet/mellanox/mlx5/core/fs_pool.h    |  5 +++--
 .../mlx5/core/steering/hws/fs_hws_pools.c        |  8 ++++++--
 4 files changed, 23 insertions(+), 12 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c +4 -2
index e14570a3d492..14539a20a60f 100644
@@ -449,7 +449,9 @@ static struct mlx5_fs_bulk *mlx5_fc_bulk_create(struct mlx5_core_dev *dev,
 	if (!fc_bulk)
 		return NULL;
 
-	if (mlx5_fs_bulk_init(dev, &fc_bulk->fs_bulk, bulk_len))
+	mlx5_fs_bulk_init(&fc_bulk->fs_bulk, bulk_len);
+
+	if (mlx5_fs_bulk_bitmap_alloc(dev, &fc_bulk->fs_bulk))
 		goto fc_bulk_free;
 
 	if (mlx5_cmd_fc_bulk_alloc(dev, alloc_bitmask, &base_id))
@@ -566,7 +568,7 @@ mlx5_fc_local_create(u32 counter_id, u32 offset, u32 bulk_size)
 
 	counter->type = MLX5_FC_TYPE_LOCAL;
 	counter->id = counter_id;
-	fc_bulk->fs_bulk.bulk_len = bulk_size;
+	mlx5_fs_bulk_init(&fc_bulk->fs_bulk, bulk_size);
 	mlx5_fc_bulk_init(fc_bulk, counter_id - offset);
 	counter->bulk = fc_bulk;
 	refcount_set(&counter->fc_local_refcount, 1);
drivers/net/ethernet/mellanox/mlx5/core/fs_pool.c +10 -6
index f6c226664602..faa519254316 100644
@@ -4,23 +4,27 @@
 #include <mlx5_core.h>
 #include "fs_pool.h"
 
-int mlx5_fs_bulk_init(struct mlx5_core_dev *dev, struct mlx5_fs_bulk *fs_bulk,
-		      int bulk_len)
+int mlx5_fs_bulk_bitmap_alloc(struct mlx5_core_dev *dev,
+			      struct mlx5_fs_bulk *fs_bulk)
 {
 	int i;
 
-	fs_bulk->bitmask = kvcalloc(BITS_TO_LONGS(bulk_len), sizeof(unsigned long),
-				    GFP_KERNEL);
+	fs_bulk->bitmask = kvcalloc(BITS_TO_LONGS(fs_bulk->bulk_len),
+				    sizeof(unsigned long), GFP_KERNEL);
 	if (!fs_bulk->bitmask)
 		return -ENOMEM;
 
-	fs_bulk->bulk_len = bulk_len;
-	for (i = 0; i < bulk_len; i++)
+	for (i = 0; i < fs_bulk->bulk_len; i++)
 		set_bit(i, fs_bulk->bitmask);
 
 	return 0;
 }
 
+void mlx5_fs_bulk_init(struct mlx5_fs_bulk *fs_bulk, int bulk_len)
+{
+	fs_bulk->bulk_len = bulk_len;
+}
+
 void mlx5_fs_bulk_cleanup(struct mlx5_fs_bulk *fs_bulk)
 {
 	kvfree(fs_bulk->bitmask);
drivers/net/ethernet/mellanox/mlx5/core/fs_pool.h +3 -2
index f04ec3107498..4deb66479d16 100644
@@ -39,8 +39,9 @@ struct mlx5_fs_pool {
 	int threshold;
 };
 
-int mlx5_fs_bulk_init(struct mlx5_core_dev *dev, struct mlx5_fs_bulk *fs_bulk,
-		      int bulk_len);
+void mlx5_fs_bulk_init(struct mlx5_fs_bulk *fs_bulk, int bulk_len);
+int mlx5_fs_bulk_bitmap_alloc(struct mlx5_core_dev *dev,
+			      struct mlx5_fs_bulk *fs_bulk);
 void mlx5_fs_bulk_cleanup(struct mlx5_fs_bulk *fs_bulk);
 int mlx5_fs_bulk_get_free_amount(struct mlx5_fs_bulk *bulk);
 
drivers/net/ethernet/mellanox/mlx5/core/steering/hws/fs_hws_pools.c +6 -2
index 839d71bd4216..5bc8e97ecf1c 100644
@@ -121,7 +121,9 @@ mlx5_fs_hws_pr_bulk_create(struct mlx5_core_dev *dev, void *pool_ctx)
 	if (!pr_bulk)
 		return NULL;
 
-	if (mlx5_fs_bulk_init(dev, &pr_bulk->fs_bulk, bulk_len))
+	mlx5_fs_bulk_init(&pr_bulk->fs_bulk, bulk_len);
+
+	if (mlx5_fs_bulk_bitmap_alloc(dev, &pr_bulk->fs_bulk))
 		goto free_pr_bulk;
 
 	for (i = 0; i < bulk_len; i++) {
@@ -275,7 +277,9 @@ mlx5_fs_hws_mh_bulk_create(struct mlx5_core_dev *dev, void *pool_ctx)
 	if (!mh_bulk)
 		return NULL;
 
-	if (mlx5_fs_bulk_init(dev, &mh_bulk->fs_bulk, bulk_len))
+	mlx5_fs_bulk_init(&mh_bulk->fs_bulk, bulk_len);
+
+	if (mlx5_fs_bulk_bitmap_alloc(dev, &mh_bulk->fs_bulk))
 		goto free_mh_bulk;
 
 	for (int i = 0; i < bulk_len; i++) {
Signature · 1 line
2.31.1
author maintainer
#3 lore raw

PATCH 3/3 net/mlx5: Initialize bulk for single flow counters

In reply to Tariq Toukan

From: Moshe Shemesh <[email protected]>

Ensure that flow counters allocated with mlx5_fc_single_alloc() have bulk correctly initialized so they can safely be used in HWS rules.

---
 .../net/ethernet/mellanox/mlx5/core/fs_core.h |  3 +-
 .../ethernet/mellanox/mlx5/core/fs_counters.c | 39 +++++++++++++------
 2 files changed, 29 insertions(+), 13 deletions(-)
drivers/net/ethernet/mellanox/mlx5/core/fs_core.h +2 -1
index 1c6591425260..dbaf33b537f7 100644
@@ -308,7 +308,8 @@ struct mlx5_flow_root_namespace {
 };
 
 enum mlx5_fc_type {
-	MLX5_FC_TYPE_ACQUIRED = 0,
+	MLX5_FC_TYPE_POOL_ACQUIRED = 0,
+	MLX5_FC_TYPE_SINGLE,
 	MLX5_FC_TYPE_LOCAL,
 };
 
drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c +27 -12
index 14539a20a60f..fe7caa910219 100644
@@ -153,6 +153,7 @@ static void mlx5_fc_stats_query_all_counters(struct mlx5_core_dev *dev)
 static void mlx5_fc_free(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
 {
 	mlx5_cmd_fc_free(dev, counter->id);
+	kfree(counter->bulk);
 	kfree(counter);
 }
 
@@ -163,7 +164,7 @@ static void mlx5_fc_release(struct mlx5_core_dev *dev, struct mlx5_fc *counter)
 	if (WARN_ON(counter->type == MLX5_FC_TYPE_LOCAL))
 		return;
 
-	if (counter->bulk)
+	if (counter->type == MLX5_FC_TYPE_POOL_ACQUIRED)
 		mlx5_fc_pool_release_counter(&fc_stats->fc_pool, counter);
 	else
 		mlx5_fc_free(dev, counter);
@@ -220,8 +221,16 @@ static void mlx5_fc_stats_work(struct work_struct *work)
 	mlx5_fc_stats_query_all_counters(dev);
 }
 
+static void mlx5_fc_bulk_init(struct mlx5_fc_bulk *fc_bulk, u32 base_id)
+{
+	fc_bulk->base_id = base_id;
+	refcount_set(&fc_bulk->hws_data.hws_action_refcount, 0);
+	mutex_init(&fc_bulk->hws_data.lock);
+}
+
 static struct mlx5_fc *mlx5_fc_single_alloc(struct mlx5_core_dev *dev)
 {
+	struct mlx5_fc_bulk *fc_bulk;
 	struct mlx5_fc *counter;
 	int err;
 
@@ -229,13 +238,26 @@ static struct mlx5_fc *mlx5_fc_single_alloc(struct mlx5_core_dev *dev)
 	if (!counter)
 		return ERR_PTR(-ENOMEM);
 
-	err = mlx5_cmd_fc_alloc(dev, &counter->id);
-	if (err) {
-		kfree(counter);
-		return ERR_PTR(err);
+	fc_bulk = kzalloc(sizeof(*fc_bulk), GFP_KERNEL);
+	if (!fc_bulk) {
+		err = -ENOMEM;
+		goto free_counter;
 	}
+	err = mlx5_cmd_fc_alloc(dev, &counter->id);
+	if (err)
+		goto free_bulk;
 
+	counter->type = MLX5_FC_TYPE_SINGLE;
+	mlx5_fs_bulk_init(&fc_bulk->fs_bulk, 1);
+	mlx5_fc_bulk_init(fc_bulk, counter->id);
+	counter->bulk = fc_bulk;
 	return counter;
+
+free_bulk:
+	kfree(fc_bulk);
+free_counter:
+	kfree(counter);
+	return ERR_PTR(err);
 }
 
 static struct mlx5_fc *mlx5_fc_acquire(struct mlx5_core_dev *dev, bool aging)
@@ -421,13 +443,6 @@ static void mlx5_fc_init(struct mlx5_fc *counter, struct mlx5_fc_bulk *bulk,
 	counter->id = id;
 }
 
-static void mlx5_fc_bulk_init(struct mlx5_fc_bulk *fc_bulk, u32 base_id)
-{
-	fc_bulk->base_id = base_id;
-	refcount_set(&fc_bulk->hws_data.hws_action_refcount, 0);
-	mutex_init(&fc_bulk->hws_data.lock);
-}
-
 u32 mlx5_fc_get_base_id(struct mlx5_fc *counter)
 {
 	return counter->bulk->base_id;
Signature · 1 line
2.31.1
maintainer
#4 lore raw

Re: [PATCH net-next 3/3] net/mlx5: Initialize bulk for single flow counters

drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c static void mlx5_fc_stats_work(struct work_struct *work) · line 229

Quoted 10 lines · Tariq Toukan
drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c +6 -0
@@ -220,8 +221,16 @@ static void mlx5_fc_stats_work(struct work_struct *work)
 	mlx5_fc_stats_query_all_counters(dev);
 }
 
+static void mlx5_fc_bulk_init(struct mlx5_fc_bulk *fc_bulk, u32 base_id)
+{
+	fc_bulk->base_id = base_id;
+	refcount_set(&fc_bulk->hws_data.hws_action_refcount, 0);
+	mutex_init(&fc_bulk->hws_data.lock);
+}

Not worthy a repost, but you could have avoided moving this function placing it here in patch 1/3.

/P

reviewer
#5 lore raw
In reply to Tariq Toukan
Quoted 6 lines · Tariq Toukan

Hi,

This small series refactors the flow counter bulk initialization code and extends it so that single flow counters are also usable by hardware steering (HWS) rules.

Quoted 7 lines

Patches 1-2 refactor the bulk init path: first by factoring out common flow counter bulk initialization into mlx5_fc_bulk_init(), then by splitting the bitmap allocation into mlx5_fs_bulk_bitmap_alloc(), with no functional changes.

Patch 3 initializes bulk data for counters allocated via mlx5_fc_single_alloc(), so they can be safely used by HWS rules.

Hi Tariq,

Overall this looks good to me.
Am I correct in thinking that there will
be follow-up patches to make HWS use counters?

#6 lore raw
In reply to Tariq Toukan
This series was applied to netdev/net-next.git (main) by Paolo Abeni <[email protected]>:

Hello:

This series was applied to netdev/net-next.git (main) by Paolo Abeni <[email protected]>:

Quoted 6 lines · On Mon, 12 Jan 2026 11:40:22 +0200 you wrote:

Hi,

This small series refactors the flow counter bulk initialization code and extends it so that single flow counters are also usable by hardware steering (HWS) rules.

Quoted 6 lines

Patches 1-2 refactor the bulk init path: first by factoring out common flow counter bulk initialization into mlx5_fc_bulk_init(), then by splitting the bitmap allocation into mlx5_fs_bulk_bitmap_alloc(), with no functional changes.

[...]

Here is the summary with links:
- [net-next,1/3] net/mlx5: fs, factor out flow counter bulk init

    https://git.kernel.org/netdev/net-next/c/96e89982a68c
  - [net-next,2/3] net/mlx5: fs, split bulk init
    https://git.kernel.org/netdev/net-next/c/6a6c4dd7c019
  - [net-next,3/3] net/mlx5: Initialize bulk for single flow counters
    https://git.kernel.org/netdev/net-next/c/1c8910f50350

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html