From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 6 Mar 2023 12:57:51 +0000 (-0500)
Subject: six locks: be more careful about lost wakeups
X-Git-Tag: io_uring-6.7-2023-11-10~119^2~439
X-Git-Url: https://git.kernel.dk/?a=commitdiff_plain;h=4b5b13da527b1ce02ee9a96382684496e8d83696;p=linux-block.git

six locks: be more careful about lost wakeups

This is a workaround for a lost wakeup bug we've been seeing - we still
need to discover the actual bug.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---

diff --git a/fs/bcachefs/six.c b/fs/bcachefs/six.c
index e1e9df0368b6..b54a2ac480c8 100644
--- a/fs/bcachefs/six.c
+++ b/fs/bcachefs/six.c
@@ -142,8 +142,17 @@ static int __do_six_trylock_type(struct six_lock *lock,
 		 * lock, issue a wakeup because we might have caused a
 		 * spurious trylock failure:
 		 */
+#if 0
+		/*
+		 * This code should be sufficient, but we're seeing unexplained
+		 * lost wakeups:
+		 */
 		if (old.write_locking)
 			ret = -1 - SIX_LOCK_write;
+#else
+		if (!ret)
+			ret = -1 - SIX_LOCK_write;
+#endif
 	} else if (type == SIX_LOCK_write && lock->readers) {
 		if (try) {
 			atomic64_add(__SIX_VAL(write_locking, 1),
@@ -319,11 +328,10 @@ static bool __six_relock_type(struct six_lock *lock, enum six_lock_type type,
 		 * Similar to the lock path, we may have caused a spurious write
 		 * lock fail and need to issue a wakeup:
 		 */
-		if (old.write_locking)
-			six_lock_wakeup(lock, old, SIX_LOCK_write);
-
 		if (ret)
 			six_acquire(&lock->dep_map, 1, type == SIX_LOCK_read, ip);
+		else
+			six_lock_wakeup(lock, old, SIX_LOCK_write);
 
 		return ret;
 	}