Before actually destrying a workqueue, destroy_workqueue() checks whether it's actually idle. If it isn't, it prints out a bunch of warning messages and leaves the workqueue dangling. It unfortunately has a couple issues. * Mayday list queueing increments pwq's refcnts which gets detected as busy and fails the sanity checks. However, because mayday list queueing is asynchronous, this condition can happen without any actual work items left in the workqueue. * Sanity check failure leaves the sysfs interface behind too which can lead to init failure of newer instances of the workqueue. This patch fixes the above two by * If a workqueue has a rescuer, disable and kill the rescuer before sanity checks. Disabling and killing is guaranteed to flush the existing mayday list. * Remove sysfs interface before sanity checks.
Showing the changes to kernel/workqueue.c (1 of 1 file). Show all
Unified · Split
| @@ -4325,9 +4325,28 @@ void destroy_workqueue(struct workqueue_struct *wq) · 4324 unmodified lines | ||
| 4325 | 4325 | struct pool_workqueue *pwq; |
| 4326 | 4326 | int node; |
| 4327 | 4327 | |
| 4328 | + /* | |
| 4329 | + * Remove it from sysfs first so that sanity check failure doesn't | |
| 4330 | + * lead to sysfs name conflicts. | |
| 4331 | + */ | |
| 4332 | + workqueue_sysfs_unregister(wq); | |
| 4333 | + | |
| 4328 | 4334 | /* drain it before proceeding with destruction */ |
| 4329 | 4335 | drain_workqueue(wq); |
| 4330 | 4336 | |
| 4337 | + /* kill rescuer, if sanity checks fail, leave it w/o rescuer */ | |
| 4338 | + if (wq->rescuer) { | |
| 4339 | + struct worker *rescuer = wq->rescuer; | |
| 4340 | + | |
| 4341 | + /* this prevents new queueing */ | |
| 4342 | + spin_lock_irq(&wq_mayday_lock); | |
| 4343 | + wq->rescuer = NULL; | |
| 4344 | + spin_unlock_irq(&wq_mayday_lock); | |
| 4345 | + | |
| 4346 | + /* rescuer will empty maydays list before exiting */ | |
| 4347 | + kthread_stop(rescuer->task); | |
| 4348 | + } | |
| 4349 | + | |
| 4331 | 4350 | /* sanity checks */ |
| 4332 | 4351 | mutex_lock(&wq->mutex); |
| 4333 | 4352 | for_each_pwq(pwq, wq) { |
| @@ -4359,11 +4378,6 @@ void destroy_workqueue(struct workqueue_struct *wq) · 25 unmodified lines | ||
| 4359 | 4378 | list_del_rcu(&wq->list); |
| 4360 | 4379 | mutex_unlock(&wq_pool_mutex); |
| 4361 | 4380 | |
| 4362 | − workqueue_sysfs_unregister(wq); | |
| 4363 | − | |
| 4364 | − if (wq->rescuer) | |
| 4365 | − kthread_stop(wq->rescuer->task); | |
| 4366 | − | |
| 4367 | 4381 | if (!(wq->flags & WQ_UNBOUND)) { |
| 4368 | 4382 | wq_unregister_lockdep(wq); |
| 4369 | 4383 | /* |