[PATCH net-next 0/6] net/mlx5: misc changes 2025-07-14

[PATCH net-next 0/6] net/mlx5: misc changes 2025-07-14

[Tariq Toukan]

Hi,

This series contains misc enhancements to the mlx5 driver.

Regards,
Tariq


Lama Kayal (4):
  net/mlx5: HWS, Enable IPSec hardware offload in legacy mode
  net/mlx5e: SHAMPO, Cleanup reservation size formula
  net/mlx5e: SHAMPO, Remove mlx5e_shampo_get_log_hd_entry_size()
  net/mlx5e: Remove duplicate mkey from SHAMPO header

Leon Romanovsky (1):
  net/mlx5e: Properly access RCU protected qdisc_sleeping variable

Moshe Shemesh (1):
  net/mlx5e: fix kdoc warning on eswitch.h

 drivers/net/ethernet/mellanox/mlx5/core/en.h  | 10 ++---
 .../ethernet/mellanox/mlx5/core/en/params.c   | 43 ++++++-------------
 .../ethernet/mellanox/mlx5/core/en/params.h   |  6 ---
 .../net/ethernet/mellanox/mlx5/core/en/qos.c  |  2 +-
 .../mlx5/core/en_accel/ipsec_offload.c        |  3 +-
 .../net/ethernet/mellanox/mlx5/core/en_main.c | 27 ++++++++----
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  2 +-
 .../net/ethernet/mellanox/mlx5/core/eswitch.h |  2 +-
 8 files changed, 41 insertions(+), 54 deletions(-)


base-commit: a52f9f0d77f20efc285908a28b5697603b6597c7
-- 
2.40.1

[PATCH net-next 1/6] net/mlx5: HWS, Enable IPSec hardware offload in legacy mode

[Tariq Toukan]

From: Lama Kayal <lkayal@nvidia.com>

IPSec hardware offload in legacy mode should not be affected by the
steering mode, hence it should also work properly with hmfs mode.

Remove steering mode validation when calculating the cap for packet
offload, this will also enable the missing cap MLX5_IPSEC_CAP_PRIO
needed for crypto offload.

Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 .../net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c   | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
index 820debf3fbbf..ef7322d381af 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
@@ -42,8 +42,7 @@ u32 mlx5_ipsec_device_caps(struct mlx5_core_dev *mdev)
 
 	if (MLX5_CAP_IPSEC(mdev, ipsec_full_offload) &&
 	    (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_DMFS ||
-	     (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_SMFS &&
-	     is_mdev_legacy_mode(mdev)))) {
+	     is_mdev_legacy_mode(mdev))) {
 		if (MLX5_CAP_FLOWTABLE_NIC_TX(mdev,
 					      reformat_add_esp_trasport) &&
 		    MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
-- 
2.40.1

Re: [PATCH net-next 1/6] net/mlx5: HWS, Enable IPSec hardware offload in legacy mode

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 08:39:40AM +0300, Tariq Toukan wrote:
> From: Lama Kayal <lkayal@nvidia.com>
> 
> IPSec hardware offload in legacy mode should not be affected by the
> steering mode, hence it should also work properly with hmfs mode.

What about dmfs mode? I am not sure, if you didn't remove it because it
is still needed or just forgot about removing it.

In case it is ok as it is:
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

Thanks

> 
> Remove steering mode validation when calculating the cap for packet
> offload, this will also enable the missing cap MLX5_IPSEC_CAP_PRIO
> needed for crypto offload.
> 
> Signed-off-by: Lama Kayal <lkayal@nvidia.com>
> Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>  .../net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c   | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
> index 820debf3fbbf..ef7322d381af 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
> @@ -42,8 +42,7 @@ u32 mlx5_ipsec_device_caps(struct mlx5_core_dev *mdev)
>  
>  	if (MLX5_CAP_IPSEC(mdev, ipsec_full_offload) &&
>  	    (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_DMFS ||
> -	     (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_SMFS &&
> -	     is_mdev_legacy_mode(mdev)))) {
> +	     is_mdev_legacy_mode(mdev))) {
>  		if (MLX5_CAP_FLOWTABLE_NIC_TX(mdev,
>  					      reformat_add_esp_trasport) &&
>  		    MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
> -- 
> 2.40.1
> 

Re: [PATCH net-next 1/6] net/mlx5: HWS, Enable IPSec hardware offload in legacy mode

[Jianbo Liu]

On 7/14/2025 2:43 PM, Michal Swiatkowski wrote:
> On Mon, Jul 14, 2025 at 08:39:40AM +0300, Tariq Toukan wrote:
>> From: Lama Kayal <lkayal@nvidia.com>
>>
>> IPSec hardware offload in legacy mode should not be affected by the
>> steering mode, hence it should also work properly with hmfs mode.
> 
> What about dmfs mode? I am not sure, if you didn't remove it because it
> is still needed or just forgot about removing it.
>

It is still needed.
We support packet offload for all steering modes in legacy, and only 
dmfs in switchdev. This is the logic we added before:

             dmfs    smfs
legacy       Y       Y
switchdev    Y       N

Now we support hmfs. It is the same as smfs. So the table becomes:
             dmfs    smfs   hmfs
legacy       Y       Y      Y
switchdev    Y       N      N

Instead of adding "mdev->priv.steering->mode == 
MLX5_FLOW_STEERING_MODE_HMFS", We removed "mdev->priv.steering->mode == 
MLX5_FLOW_STEERING_MODE_SMFS", and the code is simpler and clean.


> In case it is ok as it is:
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
>

Yes, it's ok. Thanks for the review.

Jianbo

> Thanks
> 
>>
>> Remove steering mode validation when calculating the cap for packet
>> offload, this will also enable the missing cap MLX5_IPSEC_CAP_PRIO
>> needed for crypto offload.
>>
>> Signed-off-by: Lama Kayal <lkayal@nvidia.com>
>> Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
>> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
>> ---
>>   .../net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c   | 3 +--
>>   1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
>> index 820debf3fbbf..ef7322d381af 100644
>> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
>> @@ -42,8 +42,7 @@ u32 mlx5_ipsec_device_caps(struct mlx5_core_dev *mdev)
>>   
>>   	if (MLX5_CAP_IPSEC(mdev, ipsec_full_offload) &&
>>   	    (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_DMFS ||
>> -	     (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_SMFS &&
>> -	     is_mdev_legacy_mode(mdev)))) {
>> +	     is_mdev_legacy_mode(mdev))) {
>>   		if (MLX5_CAP_FLOWTABLE_NIC_TX(mdev,
>>   					      reformat_add_esp_trasport) &&
>>   		    MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
>> -- 
>> 2.40.1
>>
> 

Re: [PATCH net-next 1/6] net/mlx5: HWS, Enable IPSec hardware offload in legacy mode

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 11:04:54PM +0800, Jianbo Liu wrote:
> 
> 
> On 7/14/2025 2:43 PM, Michal Swiatkowski wrote:
> > On Mon, Jul 14, 2025 at 08:39:40AM +0300, Tariq Toukan wrote:
> > > From: Lama Kayal <lkayal@nvidia.com>
> > > 
> > > IPSec hardware offload in legacy mode should not be affected by the
> > > steering mode, hence it should also work properly with hmfs mode.
> > 
> > What about dmfs mode? I am not sure, if you didn't remove it because it
> > is still needed or just forgot about removing it.
> > 
> 
> It is still needed.
> We support packet offload for all steering modes in legacy, and only dmfs in
> switchdev. This is the logic we added before:
> 
>             dmfs    smfs
> legacy       Y       Y
> switchdev    Y       N
> 
> Now we support hmfs. It is the same as smfs. So the table becomes:
>             dmfs    smfs   hmfs
> legacy       Y       Y      Y
> switchdev    Y       N      N
> 
> Instead of adding "mdev->priv.steering->mode ==
> MLX5_FLOW_STEERING_MODE_HMFS", We removed "mdev->priv.steering->mode ==
> MLX5_FLOW_STEERING_MODE_SMFS", and the code is simpler and clean.
> 

Ok, got it, thanks.

> 
> > In case it is ok as it is:
> > Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> > 
> 
> Yes, it's ok. Thanks for the review.
> 
> Jianbo
> 
> > Thanks
> > 
> > > 
> > > Remove steering mode validation when calculating the cap for packet
> > > offload, this will also enable the missing cap MLX5_IPSEC_CAP_PRIO
> > > needed for crypto offload.
> > > 
> > > Signed-off-by: Lama Kayal <lkayal@nvidia.com>
> > > Reviewed-by: Jianbo Liu <jianbol@nvidia.com>
> > > Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> > > ---
> > >   .../net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c   | 3 +--
> > >   1 file changed, 1 insertion(+), 2 deletions(-)
> > > 
> > > diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
> > > index 820debf3fbbf..ef7322d381af 100644
> > > --- a/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
> > > +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_accel/ipsec_offload.c
> > > @@ -42,8 +42,7 @@ u32 mlx5_ipsec_device_caps(struct mlx5_core_dev *mdev)
> > >   	if (MLX5_CAP_IPSEC(mdev, ipsec_full_offload) &&
> > >   	    (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_DMFS ||
> > > -	     (mdev->priv.steering->mode == MLX5_FLOW_STEERING_MODE_SMFS &&
> > > -	     is_mdev_legacy_mode(mdev)))) {
> > > +	     is_mdev_legacy_mode(mdev))) {
> > >   		if (MLX5_CAP_FLOWTABLE_NIC_TX(mdev,
> > >   					      reformat_add_esp_trasport) &&
> > >   		    MLX5_CAP_FLOWTABLE_NIC_RX(mdev,
> > > -- 
> > > 2.40.1
> > > 
> > 
> 

[PATCH net-next 2/6] net/mlx5e: fix kdoc warning on eswitch.h

[Tariq Toukan]

From: Moshe Shemesh <moshe@nvidia.com>

Fix the following kdoc warning:
git ls-files *.[ch] | egrep drivers/net/ethernet/mellanox/mlx5/core/ |\
xargs scripts/kernel-doc --none
drivers/net/ethernet/mellanox/mlx5/core/eswitch.h:824: warning: cannot
understand function prototype: 'struct mlx5_esw_event_info '

Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/eswitch.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
index d59fdcb29cb8..b0b8ef3ec3c4 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
@@ -827,7 +827,7 @@ void mlx5_esw_vport_vhca_id_clear(struct mlx5_eswitch *esw, u16 vport_num);
 int mlx5_eswitch_vhca_id_to_vport(struct mlx5_eswitch *esw, u16 vhca_id, u16 *vport_num);
 
 /**
- * mlx5_esw_event_info - Indicates eswitch mode changed/changing.
+ * struct mlx5_esw_event_info - Indicates eswitch mode changed/changing.
  *
  * @new_mode: New mode of eswitch.
  */
-- 
2.40.1

Re: [PATCH net-next 2/6] net/mlx5e: fix kdoc warning on eswitch.h

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 08:39:41AM +0300, Tariq Toukan wrote:
> From: Moshe Shemesh <moshe@nvidia.com>
> 
> Fix the following kdoc warning:
> git ls-files *.[ch] | egrep drivers/net/ethernet/mellanox/mlx5/core/ |\
> xargs scripts/kernel-doc --none
> drivers/net/ethernet/mellanox/mlx5/core/eswitch.h:824: warning: cannot
> understand function prototype: 'struct mlx5_esw_event_info '
> 
> Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>  drivers/net/ethernet/mellanox/mlx5/core/eswitch.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
> index d59fdcb29cb8..b0b8ef3ec3c4 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/eswitch.h
> @@ -827,7 +827,7 @@ void mlx5_esw_vport_vhca_id_clear(struct mlx5_eswitch *esw, u16 vport_num);
>  int mlx5_eswitch_vhca_id_to_vport(struct mlx5_eswitch *esw, u16 vhca_id, u16 *vport_num);
>  
>  /**
> - * mlx5_esw_event_info - Indicates eswitch mode changed/changing.
> + * struct mlx5_esw_event_info - Indicates eswitch mode changed/changing.
>   *
>   * @new_mode: New mode of eswitch.
>   */

Looks like it is the only one structure in this file described using
kdoc.

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

> -- 
> 2.40.1

[PATCH net-next 3/6] net/mlx5e: Properly access RCU protected qdisc_sleeping variable

[Tariq Toukan]

From: Leon Romanovsky <leonro@nvidia.com>

qdisc_sleeping variable is declared as "struct Qdisc __rcu" and
as such needs proper annotation while accessing it.

Without rtnl_dereference(), the following error is generated by sparse:

drivers/net/ethernet/mellanox/mlx5/core/en/qos.c:377:40: warning:
  incorrect type in initializer (different address spaces)
drivers/net/ethernet/mellanox/mlx5/core/en/qos.c:377:40:    expected
  struct Qdisc *qdisc
drivers/net/ethernet/mellanox/mlx5/core/en/qos.c:377:40:    got struct
  Qdisc [noderef] __rcu *qdisc_sleeping

Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en/qos.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
index f0744a45db92..4e461cb03b83 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
@@ -374,7 +374,7 @@ void mlx5e_reactivate_qos_sq(struct mlx5e_priv *priv, u16 qid, struct netdev_que
 void mlx5e_reset_qdisc(struct net_device *dev, u16 qid)
 {
 	struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, qid);
-	struct Qdisc *qdisc = dev_queue->qdisc_sleeping;
+	struct Qdisc *qdisc = rtnl_dereference(dev_queue->qdisc_sleeping);
 
 	if (!qdisc)
 		return;
-- 
2.40.1

Re: [PATCH net-next 3/6] net/mlx5e: Properly access RCU protected qdisc_sleeping variable

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 08:39:42AM +0300, Tariq Toukan wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> qdisc_sleeping variable is declared as "struct Qdisc __rcu" and
> as such needs proper annotation while accessing it.
> 
> Without rtnl_dereference(), the following error is generated by sparse:
> 
> drivers/net/ethernet/mellanox/mlx5/core/en/qos.c:377:40: warning:
>   incorrect type in initializer (different address spaces)
> drivers/net/ethernet/mellanox/mlx5/core/en/qos.c:377:40:    expected
>   struct Qdisc *qdisc
> drivers/net/ethernet/mellanox/mlx5/core/en/qos.c:377:40:    got struct
>   Qdisc [noderef] __rcu *qdisc_sleeping
> 
> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>  drivers/net/ethernet/mellanox/mlx5/core/en/qos.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
> index f0744a45db92..4e461cb03b83 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
> @@ -374,7 +374,7 @@ void mlx5e_reactivate_qos_sq(struct mlx5e_priv *priv, u16 qid, struct netdev_que
>  void mlx5e_reset_qdisc(struct net_device *dev, u16 qid)
>  {
>  	struct netdev_queue *dev_queue = netdev_get_tx_queue(dev, qid);
> -	struct Qdisc *qdisc = dev_queue->qdisc_sleeping;
> +	struct Qdisc *qdisc = rtnl_dereference(dev_queue->qdisc_sleeping);
>  
>  	if (!qdisc)
>  		return;

Good catch, other acesses to the qdisc_sleeping are fine.

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

> -- 
> 2.40.1

[PATCH net-next 4/6] net/mlx5e: SHAMPO, Cleanup reservation size formula

[Tariq Toukan]

From: Lama Kayal <lkayal@nvidia.com>

The reservation size formula can be reduced to a simple evaluation of
MLX5E_SHAMPO_WQ_RESRV_SIZE. This leaves mlx5e_shampo_get_log_rsrv_size()
with one single use, which can be replaced with a macro for simplicity.

Also, function mlx5e_shampo_get_log_rsrv_size() is used only throughout
params.c, make it static.

Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  5 +--
 .../ethernet/mellanox/mlx5/core/en/params.c   | 34 +++++++------------
 .../ethernet/mellanox/mlx5/core/en/params.h   |  4 ---
 3 files changed, 15 insertions(+), 28 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 64e69e616b1f..019bc6ca4455 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -85,8 +85,9 @@ struct page_pool;
 #define MLX5E_SHAMPO_WQ_HEADER_PER_PAGE (PAGE_SIZE >> MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
 #define MLX5E_SHAMPO_LOG_WQ_HEADER_PER_PAGE (PAGE_SHIFT - MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
 #define MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE (64)
-#define MLX5E_SHAMPO_WQ_RESRV_SIZE (64 * 1024)
-#define MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE (4096)
+#define MLX5E_SHAMPO_WQ_RESRV_SIZE_BASE_SHIFT (12)
+#define MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE (16)
+#define MLX5E_SHAMPO_WQ_RESRV_SIZE BIT(MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE)
 
 #define MLX5_MPWRQ_MIN_LOG_STRIDE_SZ(mdev) \
 	(6 + MLX5_CAP_GEN(mdev, cache_line_128byte)) /* HW restriction */
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
index fc945bce933a..616251ec6d69 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
@@ -420,19 +420,10 @@ u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
 	return order_base_2(DIV_ROUND_UP(MLX5E_RX_MAX_HEAD, MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE));
 }
 
-u8 mlx5e_shampo_get_log_rsrv_size(struct mlx5_core_dev *mdev,
-				  struct mlx5e_params *params)
+static u8 mlx5e_shampo_get_log_pkt_per_rsrv(struct mlx5e_params *params)
 {
-	return order_base_2(MLX5E_SHAMPO_WQ_RESRV_SIZE / MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE);
-}
-
-u8 mlx5e_shampo_get_log_pkt_per_rsrv(struct mlx5_core_dev *mdev,
-				     struct mlx5e_params *params)
-{
-	u32 resrv_size = BIT(mlx5e_shampo_get_log_rsrv_size(mdev, params)) *
-			 MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE;
-
-	return order_base_2(DIV_ROUND_UP(resrv_size, params->sw_mtu));
+	return order_base_2(DIV_ROUND_UP(MLX5E_SHAMPO_WQ_RESRV_SIZE,
+					 params->sw_mtu));
 }
 
 u8 mlx5e_mpwqe_get_log_stride_size(struct mlx5_core_dev *mdev,
@@ -834,10 +825,9 @@ static u32 mlx5e_shampo_get_log_cq_size(struct mlx5_core_dev *mdev,
 					struct mlx5e_params *params,
 					struct mlx5e_xsk_param *xsk)
 {
-	int rsrv_size = BIT(mlx5e_shampo_get_log_rsrv_size(mdev, params)) *
-		MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE;
+	int rsrv_size = MLX5E_SHAMPO_WQ_RESRV_SIZE;
 	u16 num_strides = BIT(mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk));
-	int pkt_per_rsrv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(mdev, params));
+	int pkt_per_rsrv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(params));
 	u8 log_stride_sz = mlx5e_mpwqe_get_log_stride_size(mdev, params, xsk);
 	int wq_size = BIT(mlx5e_mpwqe_get_log_rq_size(mdev, params, xsk));
 	int wqe_size = BIT(log_stride_sz) * num_strides;
@@ -932,10 +922,11 @@ int mlx5e_build_rq_param(struct mlx5_core_dev *mdev,
 
 		MLX5_SET(wq, wq, shampo_enable, true);
 		MLX5_SET(wq, wq, log_reservation_size,
-			 mlx5e_shampo_get_log_rsrv_size(mdev, params));
+			 MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE -
+			 MLX5E_SHAMPO_WQ_RESRV_SIZE_BASE_SHIFT);
 		MLX5_SET(wq, wq,
 			 log_max_num_of_packets_per_reservation,
-			 mlx5e_shampo_get_log_pkt_per_rsrv(mdev, params));
+			 mlx5e_shampo_get_log_pkt_per_rsrv(params));
 		MLX5_SET(wq, wq, log_headers_entry_size,
 			 mlx5e_shampo_get_log_hd_entry_size(mdev, params));
 		lro_timeout =
@@ -1048,18 +1039,17 @@ u32 mlx5e_shampo_hd_per_wqe(struct mlx5_core_dev *mdev,
 			    struct mlx5e_params *params,
 			    struct mlx5e_rq_param *rq_param)
 {
-	int resv_size = BIT(mlx5e_shampo_get_log_rsrv_size(mdev, params)) *
-		MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE;
+	int rsrv_size = MLX5E_SHAMPO_WQ_RESRV_SIZE;
 	u16 num_strides = BIT(mlx5e_mpwqe_get_log_num_strides(mdev, params, NULL));
-	int pkt_per_resv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(mdev, params));
+	int pkt_per_resv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(params));
 	u8 log_stride_sz = mlx5e_mpwqe_get_log_stride_size(mdev, params, NULL);
 	int wqe_size = BIT(log_stride_sz) * num_strides;
 	u32 hd_per_wqe;
 
 	/* Assumption: hd_per_wqe % 8 == 0. */
-	hd_per_wqe = (wqe_size / resv_size) * pkt_per_resv;
+	hd_per_wqe = (wqe_size / rsrv_size) * pkt_per_resv;
 	mlx5_core_dbg(mdev, "%s hd_per_wqe = %d rsrv_size = %d wqe_size = %d pkt_per_resv = %d\n",
-		      __func__, hd_per_wqe, resv_size, wqe_size, pkt_per_resv);
+		      __func__, hd_per_wqe, rsrv_size, wqe_size, pkt_per_resv);
 	return hd_per_wqe;
 }
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
index bd5877acc5b1..919895f64dcd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
@@ -97,10 +97,6 @@ u8 mlx5e_mpwqe_get_log_rq_size(struct mlx5_core_dev *mdev,
 			       struct mlx5e_xsk_param *xsk);
 u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
 				      struct mlx5e_params *params);
-u8 mlx5e_shampo_get_log_rsrv_size(struct mlx5_core_dev *mdev,
-				  struct mlx5e_params *params);
-u8 mlx5e_shampo_get_log_pkt_per_rsrv(struct mlx5_core_dev *mdev,
-				     struct mlx5e_params *params);
 u32 mlx5e_shampo_hd_per_wqe(struct mlx5_core_dev *mdev,
 			    struct mlx5e_params *params,
 			    struct mlx5e_rq_param *rq_param);
-- 
2.40.1

Re: [PATCH net-next 4/6] net/mlx5e: SHAMPO, Cleanup reservation size formula

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 08:39:43AM +0300, Tariq Toukan wrote:
> From: Lama Kayal <lkayal@nvidia.com>
> 
> The reservation size formula can be reduced to a simple evaluation of
> MLX5E_SHAMPO_WQ_RESRV_SIZE. This leaves mlx5e_shampo_get_log_rsrv_size()
> with one single use, which can be replaced with a macro for simplicity.
> 
> Also, function mlx5e_shampo_get_log_rsrv_size() is used only throughout
> params.c, make it static.
> 
> Signed-off-by: Lama Kayal <lkayal@nvidia.com>
> Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>  drivers/net/ethernet/mellanox/mlx5/core/en.h  |  5 +--
>  .../ethernet/mellanox/mlx5/core/en/params.c   | 34 +++++++------------
>  .../ethernet/mellanox/mlx5/core/en/params.h   |  4 ---
>  3 files changed, 15 insertions(+), 28 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
> index 64e69e616b1f..019bc6ca4455 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
> @@ -85,8 +85,9 @@ struct page_pool;
>  #define MLX5E_SHAMPO_WQ_HEADER_PER_PAGE (PAGE_SIZE >> MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
>  #define MLX5E_SHAMPO_LOG_WQ_HEADER_PER_PAGE (PAGE_SHIFT - MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
>  #define MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE (64)
> -#define MLX5E_SHAMPO_WQ_RESRV_SIZE (64 * 1024)
> -#define MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE (4096)
> +#define MLX5E_SHAMPO_WQ_RESRV_SIZE_BASE_SHIFT (12)
> +#define MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE (16)
> +#define MLX5E_SHAMPO_WQ_RESRV_SIZE BIT(MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE)
>  
>  #define MLX5_MPWRQ_MIN_LOG_STRIDE_SZ(mdev) \
>  	(6 + MLX5_CAP_GEN(mdev, cache_line_128byte)) /* HW restriction */
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
> index fc945bce933a..616251ec6d69 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c

[...]

>  
>  u8 mlx5e_mpwqe_get_log_stride_size(struct mlx5_core_dev *mdev,
> @@ -834,10 +825,9 @@ static u32 mlx5e_shampo_get_log_cq_size(struct mlx5_core_dev *mdev,
>  					struct mlx5e_params *params,
>  					struct mlx5e_xsk_param *xsk)
>  {
> -	int rsrv_size = BIT(mlx5e_shampo_get_log_rsrv_size(mdev, params)) *
> -		MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE;
> +	int rsrv_size = MLX5E_SHAMPO_WQ_RESRV_SIZE;

Broken RCT, you can use MLX5E_SHAMPO_WQ_RESRV_SIZE directly in
order_base_2() call.

>  	u16 num_strides = BIT(mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk));
> -	int pkt_per_rsrv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(mdev, params));
> +	int pkt_per_rsrv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(params));
>  	u8 log_stride_sz = mlx5e_mpwqe_get_log_stride_size(mdev, params, xsk);
>  	int wq_size = BIT(mlx5e_mpwqe_get_log_rq_size(mdev, params, xsk));
>  	int wqe_size = BIT(log_stride_sz) * num_strides;
> @@ -932,10 +922,11 @@ int mlx5e_build_rq_param(struct mlx5_core_dev *mdev,
>  
>  		MLX5_SET(wq, wq, shampo_enable, true);
>  		MLX5_SET(wq, wq, log_reservation_size,
> -			 mlx5e_shampo_get_log_rsrv_size(mdev, params));
> +			 MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE -
> +			 MLX5E_SHAMPO_WQ_RESRV_SIZE_BASE_SHIFT);
>  		MLX5_SET(wq, wq,
>  			 log_max_num_of_packets_per_reservation,
> -			 mlx5e_shampo_get_log_pkt_per_rsrv(mdev, params));
> +			 mlx5e_shampo_get_log_pkt_per_rsrv(params));
>  		MLX5_SET(wq, wq, log_headers_entry_size,
>  			 mlx5e_shampo_get_log_hd_entry_size(mdev, params));
>  		lro_timeout =
> @@ -1048,18 +1039,17 @@ u32 mlx5e_shampo_hd_per_wqe(struct mlx5_core_dev *mdev,
>  			    struct mlx5e_params *params,
>  			    struct mlx5e_rq_param *rq_param)
>  {
> -	int resv_size = BIT(mlx5e_shampo_get_log_rsrv_size(mdev, params)) *
> -		MLX5E_SHAMPO_WQ_BASE_RESRV_SIZE;
> +	int rsrv_size = MLX5E_SHAMPO_WQ_RESRV_SIZE;

Can be moved down to have RCT.

>  	u16 num_strides = BIT(mlx5e_mpwqe_get_log_num_strides(mdev, params, NULL));
> -	int pkt_per_resv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(mdev, params));
> +	int pkt_per_resv = BIT(mlx5e_shampo_get_log_pkt_per_rsrv(params));
>  	u8 log_stride_sz = mlx5e_mpwqe_get_log_stride_size(mdev, params, NULL);
>  	int wqe_size = BIT(log_stride_sz) * num_strides;
>  	u32 hd_per_wqe;
>  
>  	/* Assumption: hd_per_wqe % 8 == 0. */
> -	hd_per_wqe = (wqe_size / resv_size) * pkt_per_resv;
> +	hd_per_wqe = (wqe_size / rsrv_size) * pkt_per_resv;
>  	mlx5_core_dbg(mdev, "%s hd_per_wqe = %d rsrv_size = %d wqe_size = %d pkt_per_resv = %d\n",
> -		      __func__, hd_per_wqe, resv_size, wqe_size, pkt_per_resv);
> +		      __func__, hd_per_wqe, rsrv_size, wqe_size, pkt_per_resv);
>  	return hd_per_wqe;
>  }
>  
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
> index bd5877acc5b1..919895f64dcd 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
> @@ -97,10 +97,6 @@ u8 mlx5e_mpwqe_get_log_rq_size(struct mlx5_core_dev *mdev,
>  			       struct mlx5e_xsk_param *xsk);
>  u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
>  				      struct mlx5e_params *params);
> -u8 mlx5e_shampo_get_log_rsrv_size(struct mlx5_core_dev *mdev,
> -				  struct mlx5e_params *params);
> -u8 mlx5e_shampo_get_log_pkt_per_rsrv(struct mlx5_core_dev *mdev,
> -				     struct mlx5e_params *params);
>  u32 mlx5e_shampo_hd_per_wqe(struct mlx5_core_dev *mdev,
>  			    struct mlx5e_params *params,
>  			    struct mlx5e_rq_param *rq_param);

Just small nits, otherwise:
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

> -- 
> 2.40.1

[PATCH net-next 5/6] net/mlx5e: SHAMPO, Remove mlx5e_shampo_get_log_hd_entry_size()

[Tariq Toukan]

From: Lama Kayal <lkayal@nvidia.com>

Refactor mlx5e_shampo_get_log_hd_entry_size() as macro, for more
simplicity.

Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h        | 2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en/params.c | 9 ++-------
 drivers/net/ethernet/mellanox/mlx5/core/en/params.h | 2 --
 3 files changed, 3 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 019bc6ca4455..22098c852570 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -84,7 +84,7 @@ struct page_pool;
 #define MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE (9)
 #define MLX5E_SHAMPO_WQ_HEADER_PER_PAGE (PAGE_SIZE >> MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
 #define MLX5E_SHAMPO_LOG_WQ_HEADER_PER_PAGE (PAGE_SHIFT - MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
-#define MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE (64)
+#define MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE_SHIFT (6)
 #define MLX5E_SHAMPO_WQ_RESRV_SIZE_BASE_SHIFT (12)
 #define MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE (16)
 #define MLX5E_SHAMPO_WQ_RESRV_SIZE BIT(MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
index 616251ec6d69..de5c97ea4dd8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
@@ -414,12 +414,6 @@ u8 mlx5e_mpwqe_get_log_rq_size(struct mlx5_core_dev *mdev,
 	return params->log_rq_mtu_frames - log_pkts_per_wqe;
 }
 
-u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
-				      struct mlx5e_params *params)
-{
-	return order_base_2(DIV_ROUND_UP(MLX5E_RX_MAX_HEAD, MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE));
-}
-
 static u8 mlx5e_shampo_get_log_pkt_per_rsrv(struct mlx5e_params *params)
 {
 	return order_base_2(DIV_ROUND_UP(MLX5E_SHAMPO_WQ_RESRV_SIZE,
@@ -928,7 +922,8 @@ int mlx5e_build_rq_param(struct mlx5_core_dev *mdev,
 			 log_max_num_of_packets_per_reservation,
 			 mlx5e_shampo_get_log_pkt_per_rsrv(params));
 		MLX5_SET(wq, wq, log_headers_entry_size,
-			 mlx5e_shampo_get_log_hd_entry_size(mdev, params));
+			 MLX5E_SHAMPO_LOG_HEADER_ENTRY_SIZE -
+			 MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE_SHIFT);
 		lro_timeout =
 			mlx5e_choose_lro_timeout(mdev,
 						 MLX5E_DEFAULT_SHAMPO_TIMEOUT);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
index 919895f64dcd..488ccdbc1e2c 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
@@ -95,8 +95,6 @@ bool mlx5e_rx_mpwqe_is_linear_skb(struct mlx5_core_dev *mdev,
 u8 mlx5e_mpwqe_get_log_rq_size(struct mlx5_core_dev *mdev,
 			       struct mlx5e_params *params,
 			       struct mlx5e_xsk_param *xsk);
-u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
-				      struct mlx5e_params *params);
 u32 mlx5e_shampo_hd_per_wqe(struct mlx5_core_dev *mdev,
 			    struct mlx5e_params *params,
 			    struct mlx5e_rq_param *rq_param);
-- 
2.40.1

Re: [PATCH net-next 5/6] net/mlx5e: SHAMPO, Remove mlx5e_shampo_get_log_hd_entry_size()

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 08:39:44AM +0300, Tariq Toukan wrote:
> From: Lama Kayal <lkayal@nvidia.com>
> 
> Refactor mlx5e_shampo_get_log_hd_entry_size() as macro, for more
> simplicity.
> 
> Signed-off-by: Lama Kayal <lkayal@nvidia.com>
> Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>  drivers/net/ethernet/mellanox/mlx5/core/en.h        | 2 +-
>  drivers/net/ethernet/mellanox/mlx5/core/en/params.c | 9 ++-------
>  drivers/net/ethernet/mellanox/mlx5/core/en/params.h | 2 --
>  3 files changed, 3 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
> index 019bc6ca4455..22098c852570 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
> @@ -84,7 +84,7 @@ struct page_pool;
>  #define MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE (9)
>  #define MLX5E_SHAMPO_WQ_HEADER_PER_PAGE (PAGE_SIZE >> MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
>  #define MLX5E_SHAMPO_LOG_WQ_HEADER_PER_PAGE (PAGE_SHIFT - MLX5E_SHAMPO_LOG_MAX_HEADER_ENTRY_SIZE)
> -#define MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE (64)
> +#define MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE_SHIFT (6)
>  #define MLX5E_SHAMPO_WQ_RESRV_SIZE_BASE_SHIFT (12)
>  #define MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE (16)
>  #define MLX5E_SHAMPO_WQ_RESRV_SIZE BIT(MLX5E_SHAMPO_WQ_LOG_RESRV_SIZE)
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
> index 616251ec6d69..de5c97ea4dd8 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
> @@ -414,12 +414,6 @@ u8 mlx5e_mpwqe_get_log_rq_size(struct mlx5_core_dev *mdev,
>  	return params->log_rq_mtu_frames - log_pkts_per_wqe;
>  }
>  
> -u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
> -				      struct mlx5e_params *params)
> -{
> -	return order_base_2(DIV_ROUND_UP(MLX5E_RX_MAX_HEAD, MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE));
> -}
> -
>  static u8 mlx5e_shampo_get_log_pkt_per_rsrv(struct mlx5e_params *params)
>  {
>  	return order_base_2(DIV_ROUND_UP(MLX5E_SHAMPO_WQ_RESRV_SIZE,
> @@ -928,7 +922,8 @@ int mlx5e_build_rq_param(struct mlx5_core_dev *mdev,
>  			 log_max_num_of_packets_per_reservation,
>  			 mlx5e_shampo_get_log_pkt_per_rsrv(params));
>  		MLX5_SET(wq, wq, log_headers_entry_size,
> -			 mlx5e_shampo_get_log_hd_entry_size(mdev, params));
> +			 MLX5E_SHAMPO_LOG_HEADER_ENTRY_SIZE -
> +			 MLX5E_SHAMPO_WQ_BASE_HEAD_ENTRY_SIZE_SHIFT);

We had 2 here, now it is 8 - 6, so it is fine, it wasn't obvious for me.
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

>  		lro_timeout =
>  			mlx5e_choose_lro_timeout(mdev,
>  						 MLX5E_DEFAULT_SHAMPO_TIMEOUT);
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
> index 919895f64dcd..488ccdbc1e2c 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.h
> @@ -95,8 +95,6 @@ bool mlx5e_rx_mpwqe_is_linear_skb(struct mlx5_core_dev *mdev,
>  u8 mlx5e_mpwqe_get_log_rq_size(struct mlx5_core_dev *mdev,
>  			       struct mlx5e_params *params,
>  			       struct mlx5e_xsk_param *xsk);
> -u8 mlx5e_shampo_get_log_hd_entry_size(struct mlx5_core_dev *mdev,
> -				      struct mlx5e_params *params);
>  u32 mlx5e_shampo_hd_per_wqe(struct mlx5_core_dev *mdev,
>  			    struct mlx5e_params *params,
>  			    struct mlx5e_rq_param *rq_param);
> -- 
> 2.40.1
> 

[PATCH net-next 6/6] net/mlx5e: Remove duplicate mkey from SHAMPO header

[Tariq Toukan]

From: Lama Kayal <lkayal@nvidia.com>

SHAMPO structure holds two variations of the mkey, which is unnecessary,
a duplication that's repeated per rq.

Remove duplicate mkey information and keep only one version, the one
used in the fast path, rename field to reflect field type clearly.

Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  3 +--
 .../net/ethernet/mellanox/mlx5/core/en_main.c | 27 ++++++++++++-------
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  2 +-
 3 files changed, 20 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 22098c852570..2f9fea076c00 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -630,14 +630,13 @@ struct mlx5e_dma_info {
 };
 
 struct mlx5e_shampo_hd {
-	u32 mkey;
 	struct mlx5e_frag_page *pages;
 	u32 hd_per_wq;
 	u16 hd_per_wqe;
 	unsigned long *bitmap;
 	u16 pi;
 	u16 ci;
-	__be32 key;
+	__be32 mkey_be;
 };
 
 struct mlx5e_hw_gro_data {
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index fee323ade522..cc1e134f9734 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -545,18 +545,26 @@ static int mlx5e_create_rq_umr_mkey(struct mlx5_core_dev *mdev, struct mlx5e_rq
 }
 
 static int mlx5e_create_rq_hd_umr_mkey(struct mlx5_core_dev *mdev,
-				       u16 hd_per_wq, u32 *umr_mkey)
+				       u16 hd_per_wq, __be32 *umr_mkey)
 {
 	u32 max_ksm_size = BIT(MLX5_CAP_GEN(mdev, log_max_klm_list_size));
+	u32 mkey;
+	int err;
 
 	if (max_ksm_size < hd_per_wq) {
 		mlx5_core_err(mdev, "max ksm list size 0x%x is smaller than shampo header buffer list size 0x%x\n",
 			      max_ksm_size, hd_per_wq);
 		return -EINVAL;
 	}
-	return mlx5e_create_umr_ksm_mkey(mdev, hd_per_wq,
-					 MLX5E_SHAMPO_LOG_HEADER_ENTRY_SIZE,
-					 umr_mkey);
+
+	err = mlx5e_create_umr_ksm_mkey(mdev, hd_per_wq,
+					MLX5E_SHAMPO_LOG_HEADER_ENTRY_SIZE,
+					&mkey);
+	if (err)
+		return err;
+
+	*umr_mkey = cpu_to_be32(mkey);
+	return 0;
 }
 
 static void mlx5e_init_frags_partition(struct mlx5e_rq *rq)
@@ -782,11 +790,10 @@ static int mlx5_rq_shampo_alloc(struct mlx5_core_dev *mdev,
 		goto err_shampo_hd_info_alloc;
 
 	err = mlx5e_create_rq_hd_umr_mkey(mdev, hd_per_wq,
-					  &rq->mpwqe.shampo->mkey);
+					  &rq->mpwqe.shampo->mkey_be);
 	if (err)
 		goto err_umr_mkey;
 
-	rq->mpwqe.shampo->key = cpu_to_be32(rq->mpwqe.shampo->mkey);
 	rq->mpwqe.shampo->hd_per_wqe =
 		mlx5e_shampo_hd_per_wqe(mdev, params, rqp);
 	wq_size = BIT(MLX5_GET(wq, wqc, log_wq_sz));
@@ -831,7 +838,7 @@ static int mlx5_rq_shampo_alloc(struct mlx5_core_dev *mdev,
 err_hw_gro_data:
 	page_pool_destroy(rq->hd_page_pool);
 err_hds_page_pool:
-	mlx5_core_destroy_mkey(mdev, rq->mpwqe.shampo->mkey);
+	mlx5_core_destroy_mkey(mdev, be32_to_cpu(rq->mpwqe.shampo->mkey_be));
 err_umr_mkey:
 	mlx5e_rq_shampo_hd_info_free(rq);
 err_shampo_hd_info_alloc:
@@ -848,7 +855,8 @@ static void mlx5e_rq_free_shampo(struct mlx5e_rq *rq)
 	if (rq->hd_page_pool != rq->page_pool)
 		page_pool_destroy(rq->hd_page_pool);
 	mlx5e_rq_shampo_hd_info_free(rq);
-	mlx5_core_destroy_mkey(rq->mdev, rq->mpwqe.shampo->mkey);
+	mlx5_core_destroy_mkey(rq->mdev,
+			       be32_to_cpu(rq->mpwqe.shampo->mkey_be));
 	kvfree(rq->mpwqe.shampo);
 }
 
@@ -1121,7 +1129,8 @@ int mlx5e_create_rq(struct mlx5e_rq *rq, struct mlx5e_rq_param *param, u16 q_cou
 	if (test_bit(MLX5E_RQ_STATE_SHAMPO, &rq->state)) {
 		MLX5_SET(wq, wq, log_headers_buffer_entry_num,
 			 order_base_2(rq->mpwqe.shampo->hd_per_wq));
-		MLX5_SET(wq, wq, headers_mkey, rq->mpwqe.shampo->mkey);
+		MLX5_SET(wq, wq, headers_mkey,
+			 be32_to_cpu(rq->mpwqe.shampo->mkey_be));
 	}
 
 	mlx5_fill_page_frag_array(&rq->wq_ctrl.buf,
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index 2bb32082bfcc..78159a5e7bbd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -676,7 +676,7 @@ static int mlx5e_build_shampo_hd_umr(struct mlx5e_rq *rq,
 	wqe_bbs = MLX5E_KSM_UMR_WQEBBS(ksm_entries);
 	pi = mlx5e_icosq_get_next_pi(sq, wqe_bbs);
 	umr_wqe = mlx5_wq_cyc_get_wqe(&sq->wq, pi);
-	build_ksm_umr(sq, umr_wqe, shampo->key, index, ksm_entries);
+	build_ksm_umr(sq, umr_wqe, shampo->mkey_be, index, ksm_entries);
 
 	WARN_ON_ONCE(ksm_entries & (MLX5E_SHAMPO_WQ_HEADER_PER_PAGE - 1));
 	while (i < ksm_entries) {
-- 
2.40.1

Re: [PATCH net-next 6/6] net/mlx5e: Remove duplicate mkey from SHAMPO header

[Michal Swiatkowski]

On Mon, Jul 14, 2025 at 08:39:45AM +0300, Tariq Toukan wrote:
> From: Lama Kayal <lkayal@nvidia.com>
> 
> SHAMPO structure holds two variations of the mkey, which is unnecessary,
> a duplication that's repeated per rq.
> 
> Remove duplicate mkey information and keep only one version, the one
> used in the fast path, rename field to reflect field type clearly.
> 
> Signed-off-by: Lama Kayal <lkayal@nvidia.com>
> Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
> Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
> ---
>  drivers/net/ethernet/mellanox/mlx5/core/en.h  |  3 +--
>  .../net/ethernet/mellanox/mlx5/core/en_main.c | 27 ++++++++++++-------
>  .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  2 +-
>  3 files changed, 20 insertions(+), 12 deletions(-)

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

Thanks

> -- 
> 2.40.1
>