diff options
| -rw-r--r-- | block/blk-ioc.c | 21 |
1 files changed, 14 insertions, 7 deletions
diff --git a/block/blk-ioc.c b/block/blk-ioc.c index 9884fd7427f..8b782a63c29 100644 --- a/block/blk-ioc.c +++ b/block/blk-ioc.c @@ -80,8 +80,15 @@ static void ioc_release_fn(struct work_struct *work) struct io_context *ioc = container_of(work, struct io_context, release_work); struct request_queue *last_q = NULL; + unsigned long flags; - spin_lock_irq(&ioc->lock); + /* + * Exiting icq may call into put_io_context() through elevator + * which will trigger lockdep warning. The ioc's are guaranteed to + * be different, use a different locking subclass here. Use + * irqsave variant as there's no spin_lock_irq_nested(). + */ + spin_lock_irqsave_nested(&ioc->lock, flags, 1); while (!hlist_empty(&ioc->icq_list)) { struct io_cq *icq = hlist_entry(ioc->icq_list.first, @@ -103,15 +110,15 @@ static void ioc_release_fn(struct work_struct *work) */ if (last_q) { spin_unlock(last_q->queue_lock); - spin_unlock_irq(&ioc->lock); + spin_unlock_irqrestore(&ioc->lock, flags); blk_put_queue(last_q); } else { - spin_unlock_irq(&ioc->lock); + spin_unlock_irqrestore(&ioc->lock, flags); } last_q = this_q; - spin_lock_irq(this_q->queue_lock); - spin_lock(&ioc->lock); + spin_lock_irqsave(this_q->queue_lock, flags); + spin_lock_nested(&ioc->lock, 1); continue; } ioc_exit_icq(icq); @@ -119,10 +126,10 @@ static void ioc_release_fn(struct work_struct *work) if (last_q) { spin_unlock(last_q->queue_lock); - spin_unlock_irq(&ioc->lock); + spin_unlock_irqrestore(&ioc->lock, flags); blk_put_queue(last_q); } else { - spin_unlock_irq(&ioc->lock); + spin_unlock_irqrestore(&ioc->lock, flags); } kmem_cache_free(iocontext_cachep, ioc); |