DPDK patches and discussions
 help / color / mirror / Atom feed
From: Matan Azrad <matan@mellanox.com>
To: dev@dpdk.org
Cc: Viacheslav Ovsiienko <viacheslavo@mellanox.com>,
	Maxime Coquelin <maxime.coquelin@redhat.com>
Subject: [dpdk-dev] [PATCH] vdpa/mlx5: fix completion queue arming
Date: Tue, 18 Feb 2020 12:25:21 +0000	[thread overview]
Message-ID: <1582028721-637-1-git-send-email-matan@mellanox.com> (raw)

The mlx5 vDPA driver manages QP and CQ in order to forward the HW event
to the guest by the callfd file descriptor for each virtq.

The driver arms the CQ for the next CQE index that should be
completed by the HW in order to create completion event.

In the SW completion event handler, the driver arms the CQ again for the
next index,

The CQE index in the CQ doorbell and in the CQ doorbell record was
masked incorrectly with the CQ size mask while it should be masked only
with 0xFFFFFF mask.

Remove the CQ size mask, stay only with 0xFFFFFF mask.

Fixes: 8395927cdfaf ("vdpa/mlx5: prepare HW queues")

Signed-off-by: Matan Azrad <matan@mellanox.com>
---
 drivers/vdpa/mlx5/mlx5_vdpa_event.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/vdpa/mlx5/mlx5_vdpa_event.c b/drivers/vdpa/mlx5/mlx5_vdpa_event.c
index c50e58e..17fd9dd 100644
--- a/drivers/vdpa/mlx5/mlx5_vdpa_event.c
+++ b/drivers/vdpa/mlx5/mlx5_vdpa_event.c
@@ -81,9 +81,8 @@
 static inline void
 mlx5_vdpa_cq_arm(struct mlx5_vdpa_priv *priv, struct mlx5_vdpa_cq *cq)
 {
-	const unsigned int cqe_mask = (1 << cq->log_desc_n) - 1;
 	uint32_t arm_sn = cq->arm_sn << MLX5_CQ_SQN_OFFSET;
-	uint32_t cq_ci = cq->cq_ci & MLX5_CI_MASK & cqe_mask;
+	uint32_t cq_ci = cq->cq_ci & MLX5_CI_MASK;
 	uint32_t doorbell_hi = arm_sn | MLX5_CQ_DBR_CMD_ALL | cq_ci;
 	uint64_t doorbell = ((uint64_t)doorbell_hi << 32) | cq->cq->id;
 	uint64_t db_be = rte_cpu_to_be_64(doorbell);
@@ -182,15 +181,15 @@
 {
 	struct mlx5_vdpa_event_qp *eqp =
 				container_of(cq, struct mlx5_vdpa_event_qp, cq);
-	const unsigned int cqe_size = 1 << cq->log_desc_n;
-	const unsigned int cqe_mask = cqe_size - 1;
+	const unsigned int cq_size = 1 << cq->log_desc_n;
+	const unsigned int cq_mask = cq_size - 1;
 	int ret;
 
 	do {
 		volatile struct mlx5_cqe *cqe = cq->cqes + (cq->cq_ci &
-							    cqe_mask);
+							    cq_mask);
 
-		ret = check_cqe(cqe, cqe_size, cq->cq_ci);
+		ret = check_cqe(cqe, cq_size, cq->cq_ci);
 		switch (ret) {
 		case MLX5_CQE_STATUS_ERR:
 			cq->errors++;
@@ -208,7 +207,7 @@
 	cq->db_rec[0] = rte_cpu_to_be_32(cq->cq_ci);
 	rte_io_wmb();
 	/* Ring SW QP doorbell record. */
-	eqp->db_rec[0] = rte_cpu_to_be_32(cq->cq_ci + cqe_size);
+	eqp->db_rec[0] = rte_cpu_to_be_32(cq->cq_ci + cq_size);
 }
 
 static void
-- 
1.8.3.1


             reply	other threads:[~2020-02-18 12:25 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-02-18 12:25 Matan Azrad [this message]
2020-02-19  9:15 ` Slava Ovsiienko
2020-02-19 10:02 ` Maxime Coquelin
2020-02-19 10:23 ` Maxime Coquelin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1582028721-637-1-git-send-email-matan@mellanox.com \
    --to=matan@mellanox.com \
    --cc=dev@dpdk.org \
    --cc=maxime.coquelin@redhat.com \
    --cc=viacheslavo@mellanox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).