ntfs: check overflow when iterating ATTR_RECORDs
authorHawkins Jiawei <yin31149@gmail.com>
Wed, 31 Aug 2022 16:09:38 +0000 (00:09 +0800)
committerAndrew Morton <akpm@linux-foundation.org>
Mon, 12 Sep 2022 04:55:12 +0000 (21:55 -0700)
commit63095f4f3af59322bea984a6ae44337439348fe0
tree27fcd939c7ce47e02a3f844d77244da0f75f5403
parent36a4d82dddbbd421d2b8e79e1cab68c8126d5075
ntfs: check overflow when iterating ATTR_RECORDs

Kernel iterates over ATTR_RECORDs in mft record in ntfs_attr_find().
Because the ATTR_RECORDs are next to each other, kernel can get the next
ATTR_RECORD from end address of current ATTR_RECORD, through current
ATTR_RECORD length field.

The problem is that during iteration, when kernel calculates the end
address of current ATTR_RECORD, kernel may trigger an integer overflow bug
in executing `a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length))`.  This
may wrap, leading to a forever iteration on 32bit systems.

This patch solves it by adding some checks on calculating end address
of current ATTR_RECORD during iteration.

Link: https://lkml.kernel.org/r/20220831160935.3409-4-yin31149@gmail.com
Link: https://lore.kernel.org/all/20220827105842.GM2030@kadam/
Signed-off-by: Hawkins Jiawei <yin31149@gmail.com>
Suggested-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Anton Altaparmakov <anton@tuxera.com>
Cc: chenxiaosong (A) <chenxiaosong2@huawei.com>
Cc: syzkaller-bugs <syzkaller-bugs@googlegroups.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
fs/ntfs/attrib.c