[dpdk-dev] [PATCH] net/memif: fix missing Tx-bps stats for zero-copy

Ferruh Yigit ferruh.yigit at intel.com
Wed Apr 14 10:13:10 CEST 2021


On 4/12/2021 9:22 AM, Tianyu Li wrote:
> Fix the missing Tx-bps counter for memif zero-copy mode
> Before
>    Rx-pps:      6891450          Rx-bps:   3528438928
>    Tx-pps:      6891482          Tx-bps:            0
> After
>    Throughput (since last show)
>    Rx-pps:     11157056          Rx-bps:   5712413016
>    Tx-pps:     11157056          Tx-bps:   5712413016
> 
> Fixes: 43b815d88188 ("net/memif: support zero-copy slave")
> Cc: stable at dpdk.org
> 
> Signed-off-by: Tianyu Li <tianyu.li at arm.com>
> ---
>   drivers/net/memif/rte_eth_memif.c | 1 +
>   1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/net/memif/rte_eth_memif.c b/drivers/net/memif/rte_eth_memif.c
> index 77c95bcb7..dd2825968 100644
> --- a/drivers/net/memif/rte_eth_memif.c
> +++ b/drivers/net/memif/rte_eth_memif.c
> @@ -706,6 +706,7 @@ memif_tx_one_zc(struct pmd_process_private *proc_private, struct memif_queue *mq
>   	/* populate descriptor */
>   	d0 = &ring->desc[slot & mask];
>   	d0->length = rte_pktmbuf_data_len(mbuf);
> +	mq->n_bytes += rte_pktmbuf_data_len(mbuf);
>   	/* FIXME: get region index */
>   	d0->region = 1;
>   	d0->offset = rte_pktmbuf_mtod(mbuf, uint8_t *) -
> 

Looks good to me, but let's wait Jakub's review for it.


More information about the dev mailing list