[dpdk-dev] [PATCH 04/11] net/bnxt: fix double counting VLAN tags

Maxime Coquelin maxime.coquelin at redhat.com
Thu May 23 14:21:15 CEST 2019



On 5/21/19 11:39 PM, Ajit Khaparde wrote:
> From: Lance Richardson <lance.richardson at broadcom.com>
> 
> When determining whether scattered receive mode is needed, the
> size of a QinQ header is being added to rxmode.max_rx_pkt_len
> before comparing to the receive buffer size, however max_rx_pkt_len
> already accounts for the QinQ header size. Fix by simply comparing
> rxmode.max_rx_pkt_len against the buffer size.
> 
> Fixes: daef48efe5e5 ("net/bnxt: support set MTU")
> Signed-off-by: Lance Richardson <lance.richardson at broadcom.com>
> Signed-off-by: Ajit Khaparde <ajit.khaparde at broadcom.com>
> ---
>   drivers/net/bnxt/bnxt_ethdev.c | 3 +--
>   1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/drivers/net/bnxt/bnxt_ethdev.c b/drivers/net/bnxt/bnxt_ethdev.c
> index 064a153ec..52a6b94e0 100644
> --- a/drivers/net/bnxt/bnxt_ethdev.c
> +++ b/drivers/net/bnxt/bnxt_ethdev.c
> @@ -638,8 +638,7 @@ static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
>   
>   		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
>   				      RTE_PKTMBUF_HEADROOM);
> -		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len +
> -		    VLAN_TAG_SIZE * BNXT_NUM_VLANS > buf_size)
> +		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
>   			return 1;
>   	}
>   	return 0;
> 

Isn't it fixing patch 2?
If so, Fixes line is wrong, and it should be squashed directly into
patch 2.


More information about the dev mailing list