[dpdk-stable] [PATCH 1/3] Revert "net/mlx5: fix flow priority on queue action"
Yuanhan Liu
yliu at fridaylinux.org
Tue Feb 27 10:49:46 CET 2018
On Tue, Feb 27, 2018 at 10:36:43AM +0100, Nelio Laranjeiro wrote:
> This patch depends on another one not backported. For now it should be
> reverted to backport first the one missing and then re-apply this one.
>
> This reverts commit 0ec3529b056a826cee94104e20dfd3d58d64ab67.
Series Applied to dpdk-stable/17.11.
Thanks for the backporting!
--yliu
> ---
> drivers/net/mlx5/mlx5_flow.c | 30 +++++++++++++++++++++++++++++-
> 1 file changed, 29 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
> index 82efddc5a..c18c77d48 100644
> --- a/drivers/net/mlx5/mlx5_flow.c
> +++ b/drivers/net/mlx5/mlx5_flow.c
> @@ -882,6 +882,8 @@ priv_flow_convert_items_validate(struct priv *priv,
> }
> if (parser->drop) {
> parser->drop_q.offset += cur_item->dst_sz;
> + } else if (parser->queues_n == 1) {
> + parser->queue[HASH_RXQ_ETH].offset += cur_item->dst_sz;
> } else {
> for (n = 0; n != hash_rxq_init_n; ++n)
> parser->queue[n].offset += cur_item->dst_sz;
> @@ -1124,6 +1126,19 @@ priv_flow_convert(struct priv *priv,
> if (!parser->drop_q.ibv_attr)
> return ENOMEM;
> parser->drop_q.offset = sizeof(struct ibv_flow_attr);
> + } else if (parser->queues_n == 1) {
> + unsigned int priority =
> + attr->priority +
> + hash_rxq_init[HASH_RXQ_ETH].flow_priority;
> + unsigned int offset = parser->queue[HASH_RXQ_ETH].offset;
> +
> + parser->queue[HASH_RXQ_ETH].ibv_attr =
> + priv_flow_convert_allocate(priv, priority,
> + offset, error);
> + if (!parser->queue[HASH_RXQ_ETH].ibv_attr)
> + return ENOMEM;
> + parser->queue[HASH_RXQ_ETH].offset =
> + sizeof(struct ibv_flow_attr);
> } else {
> for (i = 0; i != hash_rxq_init_n; ++i) {
> unsigned int priority =
> @@ -1173,9 +1188,22 @@ priv_flow_convert(struct priv *priv,
> * Last step. Complete missing specification to reach the RSS
> * configuration.
> */
> - if (!parser->drop) {
> + if (parser->drop) {
> + /*
> + * Drop queue priority needs to be adjusted to
> + * their most specific layer priority.
> + */
> + parser->drop_q.ibv_attr->priority =
> + attr->priority +
> + hash_rxq_init[parser->layer].flow_priority;
> + } else if (parser->queues_n > 1) {
> priv_flow_convert_finalise(priv, parser);
> } else {
> + /*
> + * Action queue have their priority overridden with
> + * Ethernet priority, this priority needs to be adjusted to
> + * their most specific layer priority.
> + */
> parser->queue[HASH_RXQ_ETH].ibv_attr->priority =
> attr->priority +
> hash_rxq_init[parser->layer].flow_priority;
> --
> 2.11.0
More information about the stable
mailing list