Re: [dpdk-dev] [PATCH] net/bnxt: fix to reset Rx next consumer index

[Ferruh Yigit <ferruh.yigit@intel.com>]

On 8/10/2021 7:07 AM, Somnath Kotur wrote:
> In bnxt_init_one_rx_ring(), reset this variable internal to the driver
> ring to 0, so that there is no mismatch with actual value in HW on
> traffic resumption.
> 

Hi Somnath,

What is the impact of the patch, what are you fixing?

If the 'rx_next_cons' (next consumer index) is not matching with HW value, what
happens, does the Rx fails completely and unrecoverable? Or does miss one or
more packet?

And when/how this happens, since driver is not new I assume this doesn't happen
in first ring init. Does this happen after uninit and init again? Providing this
information helps clarifying scope of the fix.

Can you please send a new version with updated commit log?

Thanks,
ferruh

> Fixes: 03c8f2fe111c ("net/bnxt: detect bad opaque in Rx completion")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
> Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
> Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
> ---
>  drivers/net/bnxt/bnxt_rxr.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
> index aea71703d1..73fbdd17d1 100644
> --- a/drivers/net/bnxt/bnxt_rxr.c
> +++ b/drivers/net/bnxt/bnxt_rxr.c
> @@ -1379,6 +1379,9 @@ int bnxt_init_one_rx_ring(struct bnxt_rx_queue *rxq)
>  	}
>  	PMD_DRV_LOG(DEBUG, "TPA alloc Done!\n");
>  
> +	/* Explicitly reset this driver internal tracker on a ring init */
> +	rxr->rx_next_cons = 0;
> +
>  	return 0;
>  }
>  
>