ext4: fix crashes in dioread_nolock mode
authorJan Kara <jack@suse.cz>
Fri, 19 Feb 2016 05:33:21 +0000 (00:33 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 24 Apr 2018 07:32:10 +0000 (09:32 +0200)
commitb9b98c26705b8d5fba8f15faeb923b1c6f48d223
tree0804a1601ac2ce77b087ee3d253b70d277b1a652
parentba250be92484186b63ad52b7f9bcb66662e7ff2d
ext4: fix crashes in dioread_nolock mode

commit 74dae4278546b897eb81784fdfcce872ddd8b2b8 upstream.

Competing overwrite DIO in dioread_nolock mode will just overwrite
pointer to io_end in the inode. This may result in data corruption or
extent conversion happening from IO completion interrupt because we
don't properly set buffer_defer_completion() when unlocked DIO races
with locked DIO to unwritten extent.

Since unlocked DIO doesn't need io_end for anything, just avoid
allocating it and corrupting pointer from inode for locked DIO.
A cleaner fix would be to avoid these games with io_end pointer from the
inode but that requires more intrusive changes so we leave that for
later.

Cc: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/ext4/inode.c