[dpdk-dev] [PATCH] net/mlx5: fix Rx queue initialization for scattered segment
Raslan Darawsheh
rasland at nvidia.com
Mon Oct 26 19:07:48 CET 2020
Hi,
> -----Original Message-----
> From: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
> Sent: Monday, October 26, 2020 7:18 PM
> To: dev at dpdk.org
> Cc: Raslan Darawsheh <rasland at nvidia.com>
> Subject: [PATCH] net/mlx5: fix Rx queue initialization for scattered segment
>
> During integration/rebase there was introduced the bugs:
>
> - double memory allocation for queue structure resulting in
> losing the part of configuration settings and following
> crash
>
> - the erroneous fix for the segment logarithm
>
> Fixes: 919ef3e26cff ("net/mlx5: configure Rx queue to support split")
>
> Signed-off-by: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
> ---
> drivers/net/mlx5/mlx5.h | 1 +
> drivers/net/mlx5/mlx5_rxq.c | 11 +----------
> 2 files changed, 2 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
> index 258be03..8d65828 100644
> --- a/drivers/net/mlx5/mlx5.h
> +++ b/drivers/net/mlx5/mlx5.h
> @@ -730,6 +730,7 @@ struct mlx5_ind_table_obj {
> };
>
> /* Hash Rx queue. */
> +__extension__
> struct mlx5_hrxq {
> ILIST_ENTRY(uint32_t)next; /* Index to the next element. */
> rte_atomic32_t refcnt; /* Reference counter. */
> diff --git a/drivers/net/mlx5/mlx5_rxq.c b/drivers/net/mlx5/mlx5_rxq.c
> index 1cc477a..4e17535 100644
> --- a/drivers/net/mlx5/mlx5_rxq.c
> +++ b/drivers/net/mlx5/mlx5_rxq.c
> @@ -1486,15 +1486,6 @@ struct mlx5_rxq_ctrl *
> rte_errno = ENOSPC;
> return NULL;
> }
> - tmpl = mlx5_malloc(MLX5_MEM_RTE | MLX5_MEM_ZERO,
> - sizeof(*tmpl) + desc_n * sizeof(struct rte_mbuf *) +
> - (desc >> mprq_stride_nums) * sizeof(struct mlx5_mprq_buf
> *),
> - 0, socket);
> -
> - if (!tmpl) {
> - rte_errno = ENOMEM;
> - return NULL;
> - }
> tmpl->type = MLX5_RXQ_TYPE_STANDARD;
> if (mlx5_mr_btree_init(&tmpl->rxq.mr_ctrl.cache_bh,
> MLX5_MR_BTREE_CACHE_N, socket)) {
> @@ -1560,7 +1551,7 @@ struct mlx5_rxq_ctrl *
> * Determine the number of SGEs needed for a full packet
> * and round it to the next power of two.
> */
> - sges_n = tmpl->rxq.rxseg_n;
> + sges_n = log2above(tmpl->rxq.rxseg_n);
> if (sges_n > MLX5_MAX_LOG_RQ_SEGS) {
> DRV_LOG(ERR,
> "port %u too many SGEs (%u) needed to
> handle"
> --
> 1.8.3.1
Patch squashed into relevant commit in master-net-mlx,
Kindest regards,
Raslan Darawsheh
More information about the dev
mailing list