Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Vratislav Bendel <vbendel@redhat.com>
To:     linux-xfs@vger.kernel.org,
        "Darrick J . Wong" <darrick.wong@oracle.com>
Cc:     Brian Foster <bfoster@redhat.com>, linux-kernel@vger.kernel.org
Subject: [PATCH] xfs: Correctly invert xfs_buftarg LRU isolation logic
Date:   Wed, 28 Feb 2018 16:49:51 +0100
Message-Id: <20180228154951.31714-1-vbendel@redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

The function xfs_buftarg_isolate() used by xfs buffer schrinkers 
to determine whether a buffer should be isolated and disposed 
from LRU list, has inverted logic.

Excerpt from xfs_buftarg_isolate():
        /*
         * Decrement the b_lru_ref count unless the value is already
         * zero. If the value is already zero, we need to reclaim the
         * buffer, otherwise it gets another trip through the LRU.
         */
        if (!atomic_add_unless(&bp->b_lru_ref, -1, 0)) {
                spin_unlock(&bp->b_lock);
                return LRU_ROTATE;
        }

However, as per documentation, atomic_add_unless() returns _zero_
if the atomic value was originally equal to the specified *unsless* value.

Ultimately causing a xfs_buffer with ->b_lru_ref == 0, to take another 
trip around LRU, while isolating buffers with non-zero b_lru_ref.

Signed-off-by: Vratislav Bendel <vbendel@redhat.com>
CC: Brian Foster <bfoster@redhat.com>
---
 fs/xfs/xfs_buf.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index d1da2ee9e6db..ac669a10c62f 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -1708,7 +1708,7 @@ xfs_buftarg_isolate(
 	 * zero. If the value is already zero, we need to reclaim the
 	 * buffer, otherwise it gets another trip through the LRU.
 	 */
-	if (!atomic_add_unless(&bp->b_lru_ref, -1, 0)) {
+	if (atomic_add_unless(&bp->b_lru_ref, -1, 0)) {
 		spin_unlock(&bp->b_lock);
 		return LRU_ROTATE;
 	}
-- 
2.14.3