d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Make sure the state variable is set atomically instead of using a mutex in mlx5core.

Browse Source 
Device detach and setting error state may deadlock over the interface mutex
like this:

a) Detach code in mlx5en waits until error state is set while the interface
mutex is locked.
b) The set error handler needs to lock the interface mutex before it can
set the error state.

The solution is to use atomics to set the error state.

MFC after:		1 week
Sponsored by:		Mellanox Technologies
This commit is contained in:
Hans Petter Selasky 2018-07-17 10:20:01 +00:00
parent b575d8c850
commit 6d54b22db7
1 changed files with 5 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
sys/dev/mlx5/mlx5_core/mlx5_health.c
View File

@ -219,21 +219,19 @@ void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force)
u32 fatal_error;
int lock = -EBUSY;
mutex_lock(&dev->intf_state_mutex);
if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR) {
goto unlock;
return;
}
fatal_error = check_fatal_sensors(dev);
if (fatal_error || force) {
if (xchg(&dev->state, MLX5_DEVICE_STATE_INTERNAL_ERROR) ==
MLX5_DEVICE_STATE_INTERNAL_ERROR)
return;
if (!force)
mlx5_core_err(dev, "internal state error detected\n");
dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
mlx5_trigger_cmd_completions(dev);
}
mutex_lock(&dev->intf_state_mutex);
if (force)
goto err_state_done;
@ -272,7 +270,6 @@ void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force)
err_state_done:
mlx5_core_event(dev, MLX5_DEV_EVENT_SYS_ERROR, 0);
unlock:
mutex_unlock(&dev->intf_state_mutex);
}