ext4: fix unjournaled inode bitmap modification
authorEric Sandeen <sandeen@redhat.com>
Mon, 29 Oct 2012 02:24:57 +0000 (22:24 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Mon, 29 Oct 2012 02:24:57 +0000 (22:24 -0400)
commitffb5387e85d528fb6d0d924abfa3fbf0fc484071
treeb4e444479eada05eb93c3f1b2099a018c5aef116
parent8f0d8163b50e01f398b14bcd4dc039ac5ab18d64
ext4: fix unjournaled inode bitmap modification

commit 119c0d4460b001e44b41dcf73dc6ee794b98bd31 changed
ext4_new_inode() such that the inode bitmap was being modified
outside a transaction, which could lead to corruption, and was
discovered when journal_checksum found a bad checksum in the
journal during log replay.

Nix ran into this when using the journal_async_commit mount
option, which enables journal checksumming.  The ensuing
journal replay failures due to the bad checksums led to
filesystem corruption reported as the now infamous
"Apparent serious progressive ext4 data corruption bug"

[ Changed by tytso to only call ext4_journal_get_write_access() only
  when we're fairly certain that we're going to allocate the inode. ]

I've tested this by mounting with journal_checksum and
running fsstress then dropping power; I've also tested by
hacking DM to create snapshots w/o first quiescing, which
allows me to test journal replay repeatedly w/o actually
power-cycling the box.  Without the patch I hit a journal
checksum error every time.  With this fix it survives
many iterations.

Reported-by: Nix <nix@esperi.org.uk>
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@vger.kernel.org
fs/ext4/ialloc.c