ext4: check for overlapping extents in ext4_valid_extent_entries()
commit4645e4ee32aee01a85bdc03348982a65c65ce216
authorEryu Guan <guaneryu@gmail.com>
Wed, 4 Dec 2013 02:22:21 +0000 (3 21:22 -0500)
committerBen Hutchings <ben@decadent.org.uk>
Sat, 15 Feb 2014 19:20:12 +0000 (15 19:20 +0000)
tree43466e8dd141c574a0c44c54dadf45cb6115a069
parentec94b7aba9ced72a96cfdf0cdf693b30ff604039
ext4: check for overlapping extents in ext4_valid_extent_entries()

commit 5946d089379a35dda0e531710b48fca05446a196 upstream.

A corrupted ext4 may have out of order leaf extents, i.e.

extent: lblk 0--1023, len 1024, pblk 9217, flags: LEAF UNINIT
extent: lblk 1000--2047, len 1024, pblk 10241, flags: LEAF UNINIT
             ^^^^ overlap with previous extent

Reading such extent could hit BUG_ON() in ext4_es_cache_extent().

BUG_ON(end < lblk);

The problem is that __read_extent_tree_block() tries to cache holes as
well but assumes 'lblk' is greater than 'prev' and passes underflowed
length to ext4_es_cache_extent(). Fix it by checking for overlapping
extents in ext4_valid_extent_entries().

I hit this when fuzz testing ext4, and am able to reproduce it by
modifying the on-disk extent by hand.

Also add the check for (ee_block + len - 1) in ext4_valid_extent() to
make sure the value is not overflow.

Ran xfstests on patched ext4 and no regression.

Cc: Lukáš Czerner <lczerner@redhat.com>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
fs/ext4/extents.c