[PATCH] net/mlx5: improve error on completing table resize

[Bing Zhao]

Hi,

> -----Original Message-----
> From: Dariusz Sosnowski <dsosnowski at nvidia.com>
> Sent: Saturday, November 15, 2025 3:27 AM
> To: Slava Ovsiienko <viacheslavo at nvidia.com>; Bing Zhao
> <bingz at nvidia.com>; Ori Kam <orika at nvidia.com>; Suanming Mou
> <suanmingm at nvidia.com>; Matan Azrad <matan at nvidia.com>
> Cc: dev at dpdk.org; Gregory Etelson <getelson at nvidia.com>
> Subject: [PATCH] net/mlx5: improve error on completing table resize
> 
> If flow template table was created with resizable attribute, then
> completing table resize could fail for 2 user-related reasons:
> 
> - not all flow rules were yet updated to use the resized table,
> - resize was not started.
> 
> Both of these were reported with the same error message i.e., "cannot
> complete table resize".
> 
> Since PMD can distinguish these 2 cases, this patch improves the error
> reporting to report these 2 errors separately.
> 
> Also, this patch removes redundant __rte_unused on device parameter.
> 
> Signed-off-by: Dariusz Sosnowski <dsosnowski at nvidia.com>
> Acked-by: Gregory Etelson <getelson at nvidia.com>
> ---
>  drivers/net/mlx5/mlx5_flow_hw.c | 10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/mlx5/mlx5_flow_hw.c
> b/drivers/net/mlx5/mlx5_flow_hw.c index 208f50fbfd..8f3c84dec5 100644
> --- a/drivers/net/mlx5/mlx5_flow_hw.c
> +++ b/drivers/net/mlx5/mlx5_flow_hw.c
> @@ -15384,7 +15384,7 @@ flow_hw_table_resize(struct rte_eth_dev *dev,  }
> 
>  static int
> -flow_hw_table_resize_complete(__rte_unused struct rte_eth_dev *dev,
> +flow_hw_table_resize_complete(struct rte_eth_dev *dev,
>  			      struct rte_flow_template_table *table,
>  			      struct rte_flow_error *error)
>  {
> @@ -15398,12 +15398,16 @@ flow_hw_table_resize_complete(__rte_unused
> struct rte_eth_dev *dev,
>  		return rte_flow_error_set(error, EINVAL,
>  					  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
>  					  table, "no resizable attribute");
> +	if (matcher_info->matcher == NULL)
> +		return rte_flow_error_set(error, EINVAL,
> +					  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
> +					  table, "table resize was not started");
>  	matcher_refcnt = rte_atomic_load_explicit(&matcher_info->refcnt,
>  						  rte_memory_order_relaxed);
> -	if (!matcher_info->matcher || matcher_refcnt)
> +	if (matcher_refcnt > 0)
>  		return rte_flow_error_set(error, EBUSY,
>  					  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
> -					  table, "cannot complete table resize");
> +					  table, "all rules not yet updated");
>  	if (matcher_info->jump)
>  		mlx5dr_action_destroy(matcher_info->jump);
>  	ret = mlx5dr_matcher_destroy(matcher_info->matcher);
> --
> 2.39.5

Acked-by: Bing Zhao <bingz at nvidia.com>