From: Christoph Hellwig <hch@lst.de>
Date: Wed, 26 Jan 2022 14:21:21 +0000 (+0100)
Subject: loop: only take lo_mutex for the last reference in lo_release
X-Git-Url: https://www.infradead.org/git/?a=commitdiff_plain;h=a27d6a1db4b1de01210be565edc1aa4392baf2b4;p=users%2Fhch%2Fblock.git

loop: only take lo_mutex for the last reference in lo_release

lo_refcnt is only incremented in lo_open and decremented in lo_release,
and thus protected by open_mutex.  Only take lo_mutex when lo_release
actually takes action for the final release.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Darrick J. Wong <djwong@kernel.org>
---

diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index d3a7f281ce1b..b58dc95f80d9 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -1740,10 +1740,14 @@ static void lo_release(struct gendisk *disk, fmode_t mode)
 {
 	struct loop_device *lo = disk->private_data;
 
-	mutex_lock(&lo->lo_mutex);
-	if (atomic_dec_return(&lo->lo_refcnt))
-		goto out_unlock;
+	/*
+	 * Note: this requires disk->open_mutex to protect against races
+	 * with lo_open.
+	 */
+	if (!atomic_dec_and_test(&lo->lo_refcnt))
+		return;
 
+	mutex_lock(&lo->lo_mutex);
 	if (lo->lo_flags & LO_FLAGS_AUTOCLEAR) {
 		if (lo->lo_state != Lo_bound)
 			goto out_unlock;