|
Jan Kara |
f92b2d |
From e86807862e6880809f191c4cea7f88a489f0ed34 Mon Sep 17 00:00:00 2001
|
|
Jan Kara |
f92b2d |
From: Theodore Ts'o <tytso@mit.edu>
|
|
Jan Kara |
f92b2d |
Date: Sun, 30 Dec 2018 23:20:39 -0500
|
|
Jan Kara |
f92b2d |
Subject: [PATCH] ext4: avoid kernel warning when writing the superblock to a
|
|
Jan Kara |
f92b2d |
dead device
|
|
Jan Kara |
f92b2d |
Git-commit: e86807862e6880809f191c4cea7f88a489f0ed34
|
|
Jan Kara |
f92b2d |
Patch-mainline: v5.0-rc1
|
|
Jan Kara |
f92b2d |
References: bsc#1124981
|
|
Jan Kara |
f92b2d |
|
|
Jan Kara |
f92b2d |
The xfstests generic/475 test switches the underlying device with
|
|
Jan Kara |
f92b2d |
dm-error while running a stress test. This results in a large number
|
|
Jan Kara |
f92b2d |
of file system errors, and since we can't lock the buffer head when
|
|
Jan Kara |
f92b2d |
marking the superblock dirty in the ext4_grp_locked_error() case, it's
|
|
Jan Kara |
f92b2d |
possible the superblock to be !buffer_uptodate() without
|
|
Jan Kara |
f92b2d |
buffer_write_io_error() being true.
|
|
Jan Kara |
f92b2d |
|
|
Jan Kara |
f92b2d |
We need to set buffer_uptodate() before we call mark_buffer_dirty() or
|
|
Jan Kara |
f92b2d |
this will trigger a WARN_ON. It's safe to do this since the
|
|
Jan Kara |
f92b2d |
superblock must have been properly read into memory or the mount would
|
|
Jan Kara |
f92b2d |
have been successful. So if buffer_uptodate() is not set, we can
|
|
Jan Kara |
f92b2d |
safely assume that this happened due to a failed attempt to write the
|
|
Jan Kara |
f92b2d |
superblock.
|
|
Jan Kara |
f92b2d |
|
|
Jan Kara |
f92b2d |
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
|
|
Jan Kara |
f92b2d |
Cc: stable@vger.kernel.org
|
|
Jan Kara |
f92b2d |
Acked-by: Jan Kara <jack@suse.cz>
|
|
Jan Kara |
f92b2d |
|
|
Jan Kara |
f92b2d |
---
|
|
Jan Kara |
f92b2d |
fs/ext4/super.c | 2 +-
|
|
Jan Kara |
f92b2d |
1 file changed, 1 insertion(+), 1 deletion(-)
|
|
Jan Kara |
f92b2d |
|
|
Jan Kara |
f92b2d |
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
|
|
Jan Kara |
f92b2d |
index d6c142d73d99..fb12d3c17c1b 100644
|
|
Jan Kara |
f92b2d |
--- a/fs/ext4/super.c
|
|
Jan Kara |
f92b2d |
+++ b/fs/ext4/super.c
|
|
Jan Kara |
f92b2d |
@@ -4902,7 +4902,7 @@ static int ext4_commit_super(struct super_block *sb, int sync)
|
|
Jan Kara |
f92b2d |
ext4_superblock_csum_set(sb);
|
|
Jan Kara |
f92b2d |
if (sync)
|
|
Jan Kara |
f92b2d |
lock_buffer(sbh);
|
|
Jan Kara |
f92b2d |
- if (buffer_write_io_error(sbh)) {
|
|
Jan Kara |
f92b2d |
+ if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
|
|
Jan Kara |
f92b2d |
/*
|
|
Jan Kara |
f92b2d |
* Oh, dear. A previous attempt to write the
|
|
Jan Kara |
f92b2d |
* superblock failed. This could happen because the
|
|
Jan Kara |
f92b2d |
--
|
|
Jan Kara |
f92b2d |
2.16.4
|
|
Jan Kara |
f92b2d |
|