[PATCH] workqueue/lockdep: Explicitly initialize wq_barrier::done::map

* [PATCH] workqueue/lockdep: Explicitly initialize wq_barrier::done::map
@ 2017-08-17  9:46 Boqun Feng
  2017-08-17 10:22 ` [tip:locking/core] locking/lockdep: " tip-bot for Boqun Feng
  2017-08-17 13:26 ` [PATCH] workqueue/lockdep: " Tejun Heo
  0 siblings, 2 replies; 3+ messages in thread
From: Boqun Feng @ 2017-08-17  9:46 UTC (permalink / raw)
  To: linux-kernel
  Cc: Boqun Feng, Byungchul Park, Peter Zijlstra, Ingo Molnar,
	Tejun Heo, Lai Jiangshan

With CROSSRELEASE feature introduced, such a potential deadlock is
reported:

> Worker A : acquired of wfc.work -> wait for cpu_hotplug_lock to be released
> Task   B : acquired of cpu_hotplug_lock -> wait for lock#3 to be released
> Task   C : acquired of lock#3 -> wait for completion of barr->done
> (Task C is in lru_add_drain_all_cpuslocked())
> Worker D : wait for wfc.work to be released -> will complete barr->done

However, given this very case, such a dead lock could not happen because
Task C's barr->done and Worker D's barr->done can not be the same
instance. So this is a false positive.

The reason of this false positive is we initialize all wq_barrier::done
at insert_wq_barrier() via init_completion(), which makes them belong to
the same lock class, therefore, impossible circles are reported.

To fix this, explicitly initialize the lockdep map for wq_barrier::done
in insert_wq_barrier(), so that the lock class key of wq_barrier::done
is a subkey of the corresponding work_struct, as a result we won't build
a dependency between a wq_barrier with a unrelated work, and we can
differ wq barriers based on the related works, so the false positive
above is avoided.

Also define the empty lockdep_init_map_crosslock() for !CROSSRELEASE
to make the code simple and away from unnecessary #ifdefs.

Reported-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Cc: Byungchul Park <byungchul.park@lge.com>
---
 include/linux/lockdep.h |  1 +
 kernel/workqueue.c      | 11 ++++++++++-
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/include/linux/lockdep.h b/include/linux/lockdep.h
index 651cc61af041..fc827cab6d6e 100644
--- a/include/linux/lockdep.h
+++ b/include/linux/lockdep.h
@@ -583,6 +583,7 @@ extern void crossrelease_hist_end(enum xhlock_context_t c);
 extern void lockdep_init_task(struct task_struct *task);
 extern void lockdep_free_task(struct task_struct *task);
 #else
+#define lockdep_init_map_crosslock(m, n, k, s) do {} while (0)
 /*
  * To initialize a lockdep_map statically use this macro.
  * Note that _name must not be NULL.
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index e86733a8b344..861301c2b638 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2476,7 +2476,16 @@ static void insert_wq_barrier(struct pool_workqueue *pwq,
 	 */
 	INIT_WORK_ONSTACK(&barr->work, wq_barrier_func);
 	__set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(&barr->work));
-	init_completion(&barr->done);
+
+	/*
+	 * Explicitly init the crosslock for wq_barrier::done, make its lock
+	 * key is a subkey of the corresponding work. As a result we won't
+	 * build a dependency between a wq_barrier::done and a unrelated work.
+	 */
+	lockdep_init_map_crosslock((struct lockdep_map *)&barr->done.map,
+				   "(complete)wq_barr::done",
+				   target->lockdep_map.key, 1);
+	__init_completion(&barr->done);
 	barr->task = current;
 
 	/*
-- 
2.14.1

^ permalink raw reply related	[flat|nested] 3+ messages in thread