[dpdk-dev] [PATCH] net/mlx5: fix queue rollback when starting device

[Yongseok Koh <yskoh@mellanox.com>]

mlx5_rxq_start() and mlx5_rxq_stop() must be strictly paired because
internal reference counter is increased or decreased inside. Also,
mlx5_rxq_get() must be paired with mlx5_rxq_release().

Fixes: 7d6bf6b866b8 ("net/mlx5: add Multi-Packet Rx support")
Fixes: a1366b1a2be3 ("net/mlx5: add reference counter on DPDK Rx queues")
Fixes: 6e78005a9b30 ("net/mlx5: add reference counter on DPDK Tx queues")
Cc: stable@dpdk.org

Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
---
 drivers/net/mlx5/mlx5_trigger.c | 33 ++++++++++++++++++++-------------
 1 file changed, 20 insertions(+), 13 deletions(-)

diff --git a/drivers/net/mlx5/mlx5_trigger.c b/drivers/net/mlx5/mlx5_trigger.c
index 4d2078bbd..ef499d898 100644
--- a/drivers/net/mlx5/mlx5_trigger.c
+++ b/drivers/net/mlx5/mlx5_trigger.c
@@ -46,7 +46,6 @@ mlx5_txq_start(struct rte_eth_dev *dev)
 	unsigned int i;
 	int ret;
 
-	/* Add memory regions to Tx queues. */
 	for (i = 0; i != priv->txqs_n; ++i) {
 		struct mlx5_txq_ctrl *txq_ctrl = mlx5_txq_get(dev, i);
 
@@ -60,12 +59,17 @@ mlx5_txq_start(struct rte_eth_dev *dev)
 		}
 	}
 	ret = mlx5_tx_uar_remap(dev, priv->ctx->cmd_fd);
-	if (ret)
+	if (ret) {
+		/* Adjust index for rollback. */
+		i = priv->txqs_n - 1;
 		goto error;
+	}
 	return 0;
 error:
 	ret = rte_errno; /* Save rte_errno before cleanup. */
-	mlx5_txq_stop(dev);
+	do {
+		mlx5_txq_release(dev, i);
+	} while (i-- != 0);
 	rte_errno = ret; /* Restore rte_errno. */
 	return -rte_errno;
 }
@@ -103,8 +107,10 @@ mlx5_rxq_start(struct rte_eth_dev *dev)
 	int ret = 0;
 
 	/* Allocate/reuse/resize mempool for Multi-Packet RQ. */
-	if (mlx5_mprq_alloc_mp(dev))
-		goto error;
+	if (mlx5_mprq_alloc_mp(dev)) {
+		/* Should not release Rx queues but return immediately. */
+		return -rte_errno;
+	}
 	for (i = 0; i != priv->rxqs_n; ++i) {
 		struct mlx5_rxq_ctrl *rxq_ctrl = mlx5_rxq_get(dev, i);
 		struct rte_mempool *mp;
@@ -130,7 +136,9 @@ mlx5_rxq_start(struct rte_eth_dev *dev)
 	return 0;
 error:
 	ret = rte_errno; /* Save rte_errno before cleanup. */
-	mlx5_rxq_stop(dev);
+	do {
+		mlx5_rxq_release(dev, i);
+	} while (i-- != 0);
 	rte_errno = ret; /* Restore rte_errno. */
 	return -rte_errno;
 }
@@ -152,21 +160,21 @@ mlx5_dev_start(struct rte_eth_dev *dev)
 	struct priv *priv = dev->data->dev_private;
 	int ret;
 
-	dev->data->dev_started = 1;
-	DRV_LOG(DEBUG, "port %u allocating and configuring hash Rx queues",
-		dev->data->port_id);
+	DRV_LOG(DEBUG, "port %u starting device", dev->data->port_id);
 	ret = mlx5_txq_start(dev);
 	if (ret) {
 		DRV_LOG(ERR, "port %u Tx queue allocation failed: %s",
 			dev->data->port_id, strerror(rte_errno));
-		goto error;
+		return -rte_errno;
 	}
 	ret = mlx5_rxq_start(dev);
 	if (ret) {
 		DRV_LOG(ERR, "port %u Rx queue allocation failed: %s",
 			dev->data->port_id, strerror(rte_errno));
-		goto error;
+		mlx5_txq_stop(dev);
+		return -rte_errno;
 	}
+	dev->data->dev_started = 1;
 	ret = mlx5_rx_intr_vec_enable(dev);
 	if (ret) {
 		DRV_LOG(ERR, "port %u Rx interrupt vector creation failed",
@@ -221,8 +229,7 @@ mlx5_dev_stop(struct rte_eth_dev *dev)
 	dev->tx_pkt_burst = removed_tx_burst;
 	rte_wmb();
 	usleep(1000 * priv->rxqs_n);
-	DRV_LOG(DEBUG, "port %u cleaning up and destroying hash Rx queues",
-		dev->data->port_id);
+	DRV_LOG(DEBUG, "port %u stopping device", dev->data->port_id);
 	mlx5_flow_stop(dev, &priv->flows);
 	mlx5_traffic_disable(dev);
 	mlx5_rx_intr_vec_disable(dev);
-- 
2.11.0