[PATCH] net/mlx5: query hardware capability for max lro size

[PATCH] net/mlx5: query hardware capability for max lro size

[Rayane Boussanni]

Resolve a FIXME in mlx5_dev_infos_get() by dynamically checking the
hardware lro_cap flag instead of unconditionally advertising
MLX5_MAX_LRO_SIZE. Retains the standard 32-byte fallback for
min_rx_bufsize.

Signed-off-by: Rayane Boussanni <rboussanni@gmail.com>
---
 drivers/net/mlx5/mlx5_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
index a29cdeeb50..fb51de031b 100644
--- a/drivers/net/mlx5/mlx5_ethdev.c
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -363,9 +363,9 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
 	info->min_mtu = priv->min_mtu;
 	info->max_mtu = priv->max_mtu;
 	info->max_rx_pktlen = info->max_mtu + MLX5_ETH_OVERHEAD;
-	/* FIXME: we should ask the device for these values. */
 	info->min_rx_bufsize = 32;
-	info->max_lro_pkt_size = MLX5_MAX_LRO_SIZE;
+	info->max_lro_pkt_size = priv->sh->cdev->config.hca_attr.lro_cap ?
+				 MLX5_MAX_LRO_SIZE : 0;
 	/*
 	 * Since we need one CQ per QP, the limit is the minimum number
 	 * between the two values.
-- 
2.34.1

Re: [PATCH] net/mlx5: query hardware capability for max lro size

[Dariusz Sosnowski]

Thank you for the contribution.

+cc mlx5 maintainers

On Mon, Apr 20, 2026 at 02:48:08PM -0400, Rayane Boussanni wrote:
> Resolve a FIXME in mlx5_dev_infos_get() by dynamically checking the
> hardware lro_cap flag instead of unconditionally advertising
> MLX5_MAX_LRO_SIZE. Retains the standard 32-byte fallback for
> min_rx_bufsize.
> 
> Signed-off-by: Rayane Boussanni <rboussanni@gmail.com>
> ---
>  drivers/net/mlx5/mlx5_ethdev.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
> index a29cdeeb50..fb51de031b 100644
> --- a/drivers/net/mlx5/mlx5_ethdev.c
> +++ b/drivers/net/mlx5/mlx5_ethdev.c
> @@ -363,9 +363,9 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
>  	info->min_mtu = priv->min_mtu;
>  	info->max_mtu = priv->max_mtu;
>  	info->max_rx_pktlen = info->max_mtu + MLX5_ETH_OVERHEAD;
> -	/* FIXME: we should ask the device for these values. */
>  	info->min_rx_bufsize = 32;
> -	info->max_lro_pkt_size = MLX5_MAX_LRO_SIZE;
> +	info->max_lro_pkt_size = priv->sh->cdev->config.hca_attr.lro_cap ?
> +				 MLX5_MAX_LRO_SIZE : 0;

Could you please change the condition to use priv->sh->config.lro_allowed?
This field is used in other cases where LRO capability is checked,
since it covers both hca_attr and DevX support checks.

>  	/*
>  	 * Since we need one CQ per QP, the limit is the minimum number
>  	 * between the two values.

Best regards,
Dariusz Sosnowski

[PATCH v2] net/mlx5: query hardware capability for max lro size

[Rayane Boussanni]

Resolve a FIXME in mlx5_dev_infos_get() by dynamically checking the
lro_allowed flag instead of unconditionally advertising
MLX5_MAX_LRO_SIZE.

Signed-off-by: Rayane Boussanni <rboussanni@gmail.com>
---

Thank you Dariusz for the review!

 drivers/net/mlx5/mlx5_ethdev.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c
index a29cdeeb50..734ce0e60d 100644
--- a/drivers/net/mlx5/mlx5_ethdev.c
+++ b/drivers/net/mlx5/mlx5_ethdev.c
@@ -363,9 +363,9 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
 	info->min_mtu = priv->min_mtu;
 	info->max_mtu = priv->max_mtu;
 	info->max_rx_pktlen = info->max_mtu + MLX5_ETH_OVERHEAD;
-	/* FIXME: we should ask the device for these values. */
 	info->min_rx_bufsize = 32;
-	info->max_lro_pkt_size = MLX5_MAX_LRO_SIZE;
+	info->max_lro_pkt_size = priv->sh->config.lro_allowed ?
+				 MLX5_MAX_LRO_SIZE : 0;
 	/*
 	 * Since we need one CQ per QP, the limit is the minimum number
 	 * between the two values.
-- 
2.43.0.windows.1

Re: [PATCH v2] net/mlx5: query hardware capability for max lro size

[Dariusz Sosnowski]

On Fri, Apr 24, 2026 at 10:27:32AM -0400, Rayane Boussanni wrote:
> Resolve a FIXME in mlx5_dev_infos_get() by dynamically checking the
> lro_allowed flag instead of unconditionally advertising
> MLX5_MAX_LRO_SIZE.
> 
> Signed-off-by: Rayane Boussanni <rboussanni@gmail.com>

Acked-by: Dariusz Sosnowski <dsosnowski@nvidia.com>