Debug cv_destroy() with mutex held

There still appears to be a race in the condition variables where
->cv_mutex is set after we are woken from the cv_destroy wait queue.
This might be possible when cv_destroy() is called immediately after
cv_broadcast().  We had some troubles with this previously but
there may still be a small race, see commit d599e4f.

The following patch closes one small race and improves the ASSERTs
such that they log the offending value.

Signed-off-by: Brian Behlendorf <[email protected]>
zfsonlinux/zfs#943

1 files changed, 5 insertions, 4 deletions

diff --git a/module/spl/spl-condvar.c b/module/spl/spl-condvar.c
index 52131c1e8..e9f727d72 100644
--- a/module/spl/spl-condvar.c
+++ b/module/spl/spl-condvar.c
@@ -63,7 +63,8 @@ EXPORT_SYMBOL(__cv_init);
 static int
 cv_destroy_wakeup(kcondvar_t *cvp)
 {
-	if ((waitqueue_active(&cvp->cv_event)) ||
+	if ((cvp->cv_mutex != NULL) ||
+	    (waitqueue_active(&cvp->cv_event)) ||
 	    (atomic_read(&cvp->cv_waiters) > 0))
 		return 0;
 
@@ -81,9 +82,9 @@ __cv_destroy(kcondvar_t *cvp)
 	while (cv_destroy_wakeup(cvp) == 0)
 		wait_event_timeout(cvp->cv_destroy, cv_destroy_wakeup(cvp), 1);
 
-	ASSERT(cvp->cv_mutex == NULL);
-	ASSERT(atomic_read(&cvp->cv_waiters) == 0);
-	ASSERT(!waitqueue_active(&cvp->cv_event));
+	ASSERT3P(cvp->cv_mutex, ==, NULL);
+	ASSERT3S(atomic_read(&cvp->cv_waiters), ==, 0);
+	ASSERT3S(waitqueue_active(&cvp->cv_event), ==, 0);
 
 	SEXIT;
 }