[Cluster-devel] [PATCH] gfs2: Fix glock_hash_walk bugs

Andreas Gruenbacher agruenba at redhat.com
Thu Oct 7 14:19:49 UTC 2021


So far, glock_hash_walk took a reference on each glock it iterated over, and it
was the examiner's responsibility to drop those references.  Dropping the final
reference to a glock can sleep and the examiners are called in a RCU critical
section with spin locks held, so examiners that didn't need the extra reference
had to drop it asynchronously via gfs2_glock_queue_put or similar.  This wasn't
done correctly in thaw_glock which did call gfs2_glock_put, and not at all in
dump_glock_func.

Change glock_hash_walk to not take glock references at all.  That way, the
examiners that don't need them won't have to bother with slow asynchronous
puts, and the examiners that do need references can take them themselves.

Reported-by: Alexander Aring <aahringo at redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba at redhat.com>
---
 fs/gfs2/glock.c | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index e0eaa9cf9fb6..058d06e5d8c3 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1893,10 +1893,10 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
 	do {
 		rhashtable_walk_start(&iter);
 
-		while ((gl = rhashtable_walk_next(&iter)) && !IS_ERR(gl))
-			if (gl->gl_name.ln_sbd == sdp &&
-			    lockref_get_not_dead(&gl->gl_lockref))
+		while ((gl = rhashtable_walk_next(&iter)) && !IS_ERR(gl)) {
+			if (gl->gl_name.ln_sbd == sdp)
 				examiner(gl);
+		}
 
 		rhashtable_walk_stop(&iter);
 	} while (cond_resched(), gl == ERR_PTR(-EAGAIN));
@@ -1938,7 +1938,6 @@ static void flush_delete_work(struct gfs2_glock *gl)
 					   &gl->gl_delete, 0);
 		}
 	}
-	gfs2_glock_queue_work(gl, 0);
 }
 
 void gfs2_flush_delete_work(struct gfs2_sbd *sdp)
@@ -1955,10 +1954,10 @@ void gfs2_flush_delete_work(struct gfs2_sbd *sdp)
 
 static void thaw_glock(struct gfs2_glock *gl)
 {
-	if (!test_and_clear_bit(GLF_FROZEN, &gl->gl_flags)) {
-		gfs2_glock_put(gl);
+	if (!test_and_clear_bit(GLF_FROZEN, &gl->gl_flags))
+		return;
+	if (!lockref_get_not_dead(&gl->gl_lockref))
 		return;
-	}
 	set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
 	gfs2_glock_queue_work(gl, 0);
 }
@@ -1974,9 +1973,12 @@ static void clear_glock(struct gfs2_glock *gl)
 	gfs2_glock_remove_from_lru(gl);
 
 	spin_lock(&gl->gl_lockref.lock);
-	if (gl->gl_state != LM_ST_UNLOCKED)
-		handle_callback(gl, LM_ST_UNLOCKED, 0, false);
-	__gfs2_glock_queue_work(gl, 0);
+	if (!__lockref_is_dead(&gl->gl_lockref)) {
+		gl->gl_lockref.count++;
+		if (gl->gl_state != LM_ST_UNLOCKED)
+			handle_callback(gl, LM_ST_UNLOCKED, 0, false);
+		__gfs2_glock_queue_work(gl, 0);
+	}
 	spin_unlock(&gl->gl_lockref.lock);
 }
 
-- 
2.26.3




More information about the Cluster-devel mailing list