Btrfs: fix potential use-after-free for cloned bio
authorLiu Bo <bo.li.liu@oracle.com>
Mon, 10 Apr 2017 19:36:26 +0000 (12:36 -0700)
committerDavid Sterba <dsterba@suse.com>
Tue, 11 Apr 2017 16:49:56 +0000 (18:49 +0200)
KASAN reports that there is a use-after-free case of bio in btrfs_map_bio.

If we need to submit IOs to several disks at a time, the original bio
would get cloned and mapped to the destination disk, but we really should
use the original bio instead of a cloned bio to do the sanity check
because cloned bios are likely to be freed by its endio.

Reported-by: Diego <diegocg@gmail.com>
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/volumes.c

index 7c8c7bbee197514cb92b2bed02939df9ae124aba..7c7e0c99360f429252cb33d330cb4056d9309f76 100644 (file)
@@ -6213,7 +6213,7 @@ int btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
        for (dev_nr = 0; dev_nr < total_devs; dev_nr++) {
                dev = bbio->stripes[dev_nr].dev;
                if (!dev || !dev->bdev ||
-                   (bio_op(bio) == REQ_OP_WRITE && !dev->writeable)) {
+                   (bio_op(first_bio) == REQ_OP_WRITE && !dev->writeable)) {
                        bbio_error(bbio, first_bio, logical);
                        continue;
                }