aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2015-01-21 07:51:46 +1200
committerLinus Torvalds <torvalds@linux-foundation.org>2015-01-21 07:51:46 +1200
commitd4b2d0061d76e43f614e23eae3017f43a1a7c6c1 (patch)
treee33fcf87626f3c5f35c5ae4e986f1eaf3221a47c
parent06efe0e54018cb19cf0807447dc3ac747ffcfd1c (diff)
parent29187a9eeaf362d8422e62e17a22a6e115277a49 (diff)
downloadlinux-d4b2d0061d76e43f614e23eae3017f43a1a7c6c1.tar.gz
Merge branch 'for-3.19-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Pull workqueue fix from Tejun Heo: "The xfs folks have been running into weird and very rare lockups for some time now. I didn't think this could have been from workqueue side because no one else was reporting it. This time, Eric had a kdump which we looked into and it turned out this actually was a workqueue bug and the bug has been there since the beginning of concurrency managed workqueue. A worker pool ensures forward progress of the workqueues associated with it by always having at least one worker reserved from executing work items. When the pool is under contention, the idle one tries to create more workers for the pool and if that doesn't succeed quickly enough, it calls the rescuers to the pool. This logic had a subtle race condition in an early exit path. When a worker invokes this manager function, the function may return %false indicating that the caller may proceed to executing work items either because another worker is already performing the role or conditions have changed and the pool is no longer under contention. The latter part depended on the assumption that whether more workers are necessary or not remains stable while the pool is locked; however, pool->nr_running (concurrency count) may change asynchronously and it getting bumped from zero asynchronously could send off the last idle worker to execute work items. The race window is fairly narrow, and, even when it gets triggered, the pool deadlocks iff if all work items get blocked on pending work items of the pool, which is highly unlikely but can be triggered by xfs. The patch removes the race window by removing the early exit path, which doesn't server any purpose anymore anyway" * 'for-3.19-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq: workqueue: fix subtle pool management issue which can stall whole worker_pool
-rw-r--r--kernel/workqueue.c25
1 files changed, 8 insertions, 17 deletions
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 6202b08f193..beeeac9e0e3 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -1841,17 +1841,11 @@ static void pool_mayday_timeout(unsigned long __pool)
* spin_lock_irq(pool->lock) which may be released and regrabbed
* multiple times. Does GFP_KERNEL allocations. Called only from
* manager.
- *
- * Return:
- * %false if no action was taken and pool->lock stayed locked, %true
- * otherwise.
*/
-static bool maybe_create_worker(struct worker_pool *pool)
+static void maybe_create_worker(struct worker_pool *pool)
__releases(&pool->lock)
__acquires(&pool->lock)
{
- if (!need_to_create_worker(pool))
- return false;
restart:
spin_unlock_irq(&pool->lock);
@@ -1877,7 +1871,6 @@ restart:
*/
if (need_to_create_worker(pool))
goto restart;
- return true;
}
/**
@@ -1897,16 +1890,14 @@ restart:
* multiple times. Does GFP_KERNEL allocations.
*
* Return:
- * %false if the pool don't need management and the caller can safely start
- * processing works, %true indicates that the function released pool->lock
- * and reacquired it to perform some management function and that the
- * conditions that the caller verified while holding the lock before
- * calling the function might no longer be true.
+ * %false if the pool doesn't need management and the caller can safely
+ * start processing works, %true if management function was performed and
+ * the conditions that the caller verified before calling the function may
+ * no longer be true.
*/
static bool manage_workers(struct worker *worker)
{
struct worker_pool *pool = worker->pool;
- bool ret = false;
/*
* Anyone who successfully grabs manager_arb wins the arbitration
@@ -1919,12 +1910,12 @@ static bool manage_workers(struct worker *worker)
* actual management, the pool may stall indefinitely.
*/
if (!mutex_trylock(&pool->manager_arb))
- return ret;
+ return false;
- ret |= maybe_create_worker(pool);
+ maybe_create_worker(pool);
mutex_unlock(&pool->manager_arb);
- return ret;
+ return true;
}
/**