Skip to contentSkip to conversation

vdpa/mlx5: fix unregister kick handler order

PATCH Applied Yajun Wu 2 participants 3 messages Aug 8, 2023 – Oct 12, 2023

Subsystems (2)
lore
Share & tools
lore

Yajun Wu

author
#0 lore raw

PATCH 1/1 vdpa/mlx5: fix unregister kick handler order

The mlx5_vdpa_virtq_kick_handler function may still be running and waiting on virtq->virtq_lock while mlx5_vdpa_cqe_event_unset function is trying to re-initialize the virtq->virtq_lock.

This causes mlx5_vdpa_virtq_kick_handler thread can't be wake up and can't be unregister. Following print may loop forever when calling rte_vhost_driver_unregister(socket_path):

mlx5_vdpa: Try again to unregister fd 154 of virtq 11 interrupt
mlx5_vdpa: Try again to unregister fd 154 of virtq 11 interrupt
...

The fix is to move mlx5_vdpa_virtq_unregister_intr_handle before mlx5_vdpa_cqe_event_unset.

Fixes 057f7d2084 ("vdpa/mlx5: optimize datapath-control synchronization")
---
 drivers/vdpa/mlx5/mlx5_vdpa.c         | 1 +
 drivers/vdpa/mlx5/mlx5_vdpa_cthread.c | 1 -
 2 files changed, 1 insertion(+), 1 deletion(-)
drivers/vdpa/mlx5/mlx5_vdpa.c +1 -0
index f1737f82a8..8b1de8bd62 100644
@@ -282,6 +282,7 @@ _internal_mlx5_vdpa_dev_close(struct mlx5_vdpa_priv *priv,
 	int ret = 0;
 	int vid = priv->vid;
 
+	mlx5_vdpa_virtq_unreg_intr_handle_all(priv);
 	mlx5_vdpa_cqe_event_unset(priv);
 	if (priv->state == MLX5_VDPA_STATE_CONFIGURED) {
 		ret |= mlx5_vdpa_lm_log(priv);
drivers/vdpa/mlx5/mlx5_vdpa_cthread.c +0 -1
index 6e6624e5a3..1d84e422d4 100644
@@ -190,7 +190,6 @@ mlx5_vdpa_c_thread_handle(void *arg)
 			pthread_mutex_unlock(&virtq->virtq_lock);
 			break;
 		case MLX5_VDPA_TASK_DEV_CLOSE_NOWAIT:
-			mlx5_vdpa_virtq_unreg_intr_handle_all(priv);
 			pthread_mutex_lock(&priv->steer_update_lock);
 			mlx5_vdpa_steer_unset(priv);
 			pthread_mutex_unlock(&priv->steer_update_lock);
Signature · 1 line
2.27.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 42 lines · Yajun Wu

The mlx5_vdpa_virtq_kick_handler function may still be running and waiting on virtq->virtq_lock while mlx5_vdpa_cqe_event_unset function is trying to re-initialize the virtq->virtq_lock.

This causes mlx5_vdpa_virtq_kick_handler thread can't be wake up and can't be unregister. Following print may loop forever when calling rte_vhost_driver_unregister(socket_path):

mlx5_vdpa: Try again to unregister fd 154 of virtq 11 interrupt
mlx5_vdpa: Try again to unregister fd 154 of virtq 11 interrupt
...

The fix is to move mlx5_vdpa_virtq_unregister_intr_handle before mlx5_vdpa_cqe_event_unset.

Fixes 057f7d2084 ("vdpa/mlx5: optimize datapath-control synchronization")
---
  drivers/vdpa/mlx5/mlx5_vdpa.c         | 1 +
  drivers/vdpa/mlx5/mlx5_vdpa_cthread.c | 1 -
  2 files changed, 1 insertion(+), 1 deletion(-)
drivers/vdpa/mlx5/mlx5_vdpa.c +1 -0
index f1737f82a8..8b1de8bd62 100644
@@ -282,6 +282,7 @@ _internal_mlx5_vdpa_dev_close(struct mlx5_vdpa_priv *priv,
  	int ret = 0;
  	int vid = priv->vid;
 
+	mlx5_vdpa_virtq_unreg_intr_handle_all(priv);
  	mlx5_vdpa_cqe_event_unset(priv);
  	if (priv->state == MLX5_VDPA_STATE_CONFIGURED) {
  		ret |= mlx5_vdpa_lm_log(priv);
drivers/vdpa/mlx5/mlx5_vdpa_cthread.c +0 -0
index 6e6624e5a3..1d84e422d4 100644

drivers/vdpa/mlx5/mlx5_vdpa_cthread.c mlx5_vdpa_c_thread_handle(void *arg) · line 195

Quoted 7 lines
drivers/vdpa/mlx5/mlx5_vdpa_cthread.c +0 -1
@@ -190,7 +190,6 @@ mlx5_vdpa_c_thread_handle(void *arg)
  			pthread_mutex_unlock(&virtq->virtq_lock);
  			break;
  		case MLX5_VDPA_TASK_DEV_CLOSE_NOWAIT:
-			mlx5_vdpa_virtq_unreg_intr_handle_all(priv);
  			pthread_mutex_lock(&priv->steer_update_lock);
  			mlx5_vdpa_steer_unset(priv);
  			pthread_mutex_unlock(&priv->steer_update_lock);

Thanks,
Maxime

#2 lore raw
In reply to Yajun Wu
Quoted 18 lines · Yajun Wu

The mlx5_vdpa_virtq_kick_handler function may still be running and waiting on virtq->virtq_lock while mlx5_vdpa_cqe_event_unset function is trying to re-initialize the virtq->virtq_lock.

This causes mlx5_vdpa_virtq_kick_handler thread can't be wake up and can't be unregister. Following print may loop forever when calling rte_vhost_driver_unregister(socket_path):

mlx5_vdpa: Try again to unregister fd 154 of virtq 11 interrupt
mlx5_vdpa: Try again to unregister fd 154 of virtq 11 interrupt
...

The fix is to move mlx5_vdpa_virtq_unregister_intr_handle before mlx5_vdpa_cqe_event_unset.

Fixes 057f7d2084 ("vdpa/mlx5: optimize datapath-control synchronization")
Quoted 6 lines
---
  drivers/vdpa/mlx5/mlx5_vdpa.c         | 1 +
  drivers/vdpa/mlx5/mlx5_vdpa_cthread.c | 1 -
  2 files changed, 1 insertion(+), 1 deletion(-)

Applied to nex-virtio/for-next-net.

Thanks,
Maxime