DPDK patches and discussions
 help / color / mirror / Atom feed
From: Weifeng Li <liweifeng96@126.com>
To: matan@nvidia.com
Cc: dev@dpdk.org, Weifeng Li <liweifeng96@126.com>
Subject: [dpdk-dev] [PATCH v2] net/mlx5: make flow operation thread safe
Date: Sun,  8 Nov 2020 15:53:30 +0800
Message-ID: <20201108075330.1502-1-liweifeng96@126.com> (raw)
In-Reply-To: <20201108073714.1448-1-liweifeng96@126.com>

Does it need a lock for flow about below scene.
Thread1: flow_list_destroy----flow_list_create
Thread2: ---------flow_list_destroy----
Maybe the same flow can be operate at the same time.

When i start mlx5 bond and trigger LSC at the same time.
It is possible to assert in mlx5_rx_queue_release func and
print "port 4 Rx queue 0 is still used by a flow and cannot
be removed". I use dpdk-testpmd to simulate the test.

Signed-off-by: Weifeng Li <liweifeng96@126.com>
---
v2: adjust coding style issue.
---
 drivers/net/mlx5/linux/mlx5_os.c |  1 +
 drivers/net/mlx5/mlx5.h          |  1 +
 drivers/net/mlx5/mlx5_flow.c     | 13 +++++++++++--
 3 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/net/mlx5/linux/mlx5_os.c b/drivers/net/mlx5/linux/mlx5_os.c
index c78d56f..59c074e 100644
--- a/drivers/net/mlx5/linux/mlx5_os.c
+++ b/drivers/net/mlx5/linux/mlx5_os.c
@@ -1426,6 +1426,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev,
 				      MLX5_MAX_MAC_ADDRESSES);
 	priv->flows = 0;
 	priv->ctrl_flows = 0;
+	rte_spinlock_init(&priv->flow_lock);
 	rte_spinlock_init(&priv->flow_list_lock);
 	TAILQ_INIT(&priv->flow_meters);
 	TAILQ_INIT(&priv->flow_meter_profiles);
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index b43a8c9..860bf2f 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -963,6 +963,7 @@ struct mlx5_priv {
 	unsigned int reta_idx_n; /* RETA index size. */
 	struct mlx5_drop drop_queue; /* Flow drop queues. */
 	uint32_t flows; /* RTE Flow rules. */
+	rte_spinlock_t flow_lock;
 	uint32_t ctrl_flows; /* Control flow rules. */
 	rte_spinlock_t flow_list_lock;
 	struct mlx5_obj_ops obj_ops; /* HW objects operations. */
diff --git a/drivers/net/mlx5/mlx5_flow.c b/drivers/net/mlx5/mlx5_flow.c
index e9c0ddd..69d8159 100644
--- a/drivers/net/mlx5/mlx5_flow.c
+++ b/drivers/net/mlx5/mlx5_flow.c
@@ -5577,6 +5577,7 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
 				external, hairpin_flow, error);
 	if (ret < 0)
 		goto error_before_hairpin_split;
+	rte_spinlock_lock(&priv->flow_lock);
 	flow = mlx5_ipool_zmalloc(priv->sh->ipool[MLX5_IPOOL_RTE_FLOW], &idx);
 	if (!flow) {
 		rte_errno = ENOMEM;
@@ -5598,8 +5599,10 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
 	memset(rss_desc, 0, offsetof(struct mlx5_flow_rss_desc, queue));
 	rss = flow_get_rss_action(p_actions_rx);
 	if (rss) {
-		if (flow_rss_workspace_adjust(wks, rss_desc, rss->queue_num))
+		if (flow_rss_workspace_adjust(wks, rss_desc, rss->queue_num)) {
+			rte_spinlock_unlock(&priv->flow_lock);
 			return 0;
+		}
 		/*
 		 * The following information is required by
 		 * mlx5_flow_hashfields_adjust() in advance.
@@ -5723,6 +5726,7 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
 		__atomic_add_fetch(&tunnel->refctn, 1, __ATOMIC_RELAXED);
 		mlx5_free(default_miss_ctx.queue);
 	}
+	rte_spinlock_unlock(&priv->flow_lock);
 	return idx;
 error:
 	MLX5_ASSERT(flow);
@@ -5738,6 +5742,7 @@ flow_list_create(struct rte_eth_dev *dev, uint32_t *list,
 		wks->flow_nested_idx = 0;
 error_before_hairpin_split:
 	rte_free(translated_actions);
+	rte_spinlock_unlock(&priv->flow_lock);
 	return 0;
 }
 
@@ -5877,11 +5882,14 @@ flow_list_destroy(struct rte_eth_dev *dev, uint32_t *list,
 		  uint32_t flow_idx)
 {
 	struct mlx5_priv *priv = dev->data->dev_private;
+	rte_spinlock_lock(&priv->flow_lock);
 	struct rte_flow *flow = mlx5_ipool_get(priv->sh->ipool
 					       [MLX5_IPOOL_RTE_FLOW], flow_idx);
 
-	if (!flow)
+	if (!flow) {
+		rte_spinlock_unlock(&priv->flow_lock);
 		return;
+	}
 	/*
 	 * Update RX queue flags only if port is started, otherwise it is
 	 * already clean.
@@ -5908,6 +5916,7 @@ flow_list_destroy(struct rte_eth_dev *dev, uint32_t *list,
 	}
 	flow_mreg_del_copy_action(dev, flow);
 	mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_RTE_FLOW], flow_idx);
+	rte_spinlock_unlock(&priv->flow_lock);
 }
 
 /**
-- 
2.9.5


  reply	other threads:[~2020-11-08  7:57 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-11-08  7:37 [dpdk-dev] [PATCH] net/mlx5: make flow opration " Weifeng Li
2020-11-08  7:53 ` Weifeng Li [this message]
2020-11-19  1:01   ` [dpdk-dev] [PATCH v2] net/mlx5: make flow operation " Suanming Mou
2020-11-18  2:09 ` [dpdk-dev] [PATCH] net/mlx5: make flow opration " Suanming Mou

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=20201108075330.1502-1-liweifeng96@126.com \
    --to=liweifeng96@126.com \
    --cc=dev@dpdk.org \
    --cc=matan@nvidia.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

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ https://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git