dm: fix bio length of empty flush

The commit 92986f6b4c8a ("dm: use bio_clone_fast in alloc_io/alloc_tio")
removed bio_clone_fast() call from alloc_tio() when ci->io->tio is
available. In this case, ci->bio is not copied to ci->io->tio.clone.
This is fine since init_clone_info() sets same values to ci->bio and
ci->io->tio.clone.

However, when incoming bios have REQ_PREFLUSH flag, __send_empty_flush()
prepares a zero length bio on stack and set it to ci->bio. At this time,
ci->io->tio.clone still keeps non-zero length. When alloc_tio() chooses
this ci->io->tio.clone as the bio to map, it is passed to targets as
non-empty flush bio. It causes bio length check failure in dm-zoned and
unexpected operation such as dm_accept_partial_bio() call.

To avoid the non-empty flush bio, set zero length to ci->io->tio.clone
in __send_empty_flush().

Fixes: 92986f6b4c8a ("dm: use bio_clone_fast in alloc_io/alloc_tio")
Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>

diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index fc1f9583a2710639f4fa080a2830814fbdd60f41..82957bd460e894556fed5eed0b991c14f3c0713b 100644 (file)
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -1391,6 +1391,7 @@ static void __send_empty_flush(struct clone_info *ci)
 
        ci->bio = &flush_bio;
        ci->sector_count = 0;
+       ci->io->tio.clone.bi_iter.bi_size = 0;
 
        while ((ti = dm_table_get_target(ci->map, target_nr++)))
                __send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);