Re: [PATCH] udf: Check return code from udf_update_extents

[Jan Kara]

On Fri 20-01-23 12:10:28, Vladislav Efanov wrote:
> udf_add_aext() does not create new extent and returns ENOSPC if new
> block was not created by udf_bitmap_new_block(). The caller,
> udf_insert_aext(), does not check this return code and returns OK to
> its caller(udf_update_extents). Finally the error is being lost. So
> an inconsistency in inode.i_size and extents length becomes.
> 
> Later this inconsistency leads to WARNING:
> 
> WARNING: CPU: 3 PID: 1104 at fs/udf/truncate.c:226
>         udf_truncate_extents+0x7e0/0x8e0 fs/udf/truncate.c:226
> 
> RIP: 0010:udf_truncate_extents+0x7e0/0x8e0 fs/udf/truncate.c:226
> Call Trace:
>  udf_write_failed.isra.0+0x173/0x1c0 fs/udf/inode.c:179
>  udf_write_begin+0x8d/0xb0 fs/udf/inode.c:214
>  generic_perform_write+0x20a/0x4e0 mm/filemap.c:3333
>  __generic_file_write_iter+0x252/0x610 mm/filemap.c:3462
>  udf_file_write_iter+0x2cc/0x4e0 fs/udf/file.c:168
>  call_write_iter include/linux/fs.h:1904 [inline]
>  new_sync_write+0x42c/0x660 fs/read_write.c:518
>  vfs_write+0x75b/0xa40 fs/read_write.c:605
> 
> Found by Linux Verification Center (linuxtesting.org) with syzkaller.
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Signed-off-by: Vladislav Efanov <VEfanov@xxxxxxxxx>

Thanks for the fix but I have a very similar fix already queued in my tree
in linux-next: 19fd80de0a8 ("udf: Handle error when adding extent to a
file").

								Honza
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR