Re: [PATCH v2] fs/ntfs3: fix slab-out-of-bounds Read in run_unpack

[Konstantin Komarov]

On 9/23/22 14:09, Hawkins Jiawei wrote:
> Syzkaller reports slab-out-of-bounds bug as follows:
> ==================================================================
> BUG: KASAN: slab-out-of-bounds in run_unpack+0x8b7/0x970 fs/ntfs3/run.c:944
> Read of size 1 at addr ffff88801bbdff02 by task syz-executor131/3611
>
> [...]
> Call Trace:
>   <TASK>
>   __dump_stack lib/dump_stack.c:88 [inline]
>   dump_stack_lvl+0xcd/0x134 lib/dump_stack.c:106
>   print_address_description mm/kasan/report.c:317 [inline]
>   print_report.cold+0x2ba/0x719 mm/kasan/report.c:433
>   kasan_report+0xb1/0x1e0 mm/kasan/report.c:495
>   run_unpack+0x8b7/0x970 fs/ntfs3/run.c:944
>   run_unpack_ex+0xb0/0x7c0 fs/ntfs3/run.c:1057
>   ntfs_read_mft fs/ntfs3/inode.c:368 [inline]
>   ntfs_iget5+0xc20/0x3280 fs/ntfs3/inode.c:501
>   ntfs_loadlog_and_replay+0x124/0x5d0 fs/ntfs3/fsntfs.c:272
>   ntfs_fill_super+0x1eff/0x37f0 fs/ntfs3/super.c:1018
>   get_tree_bdev+0x440/0x760 fs/super.c:1323
>   vfs_get_tree+0x89/0x2f0 fs/super.c:1530
>   do_new_mount fs/namespace.c:3040 [inline]
>   path_mount+0x1326/0x1e20 fs/namespace.c:3370
>   do_mount fs/namespace.c:3383 [inline]
>   __do_sys_mount fs/namespace.c:3591 [inline]
>   __se_sys_mount fs/namespace.c:3568 [inline]
>   __x64_sys_mount+0x27f/0x300 fs/namespace.c:3568
>   do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>   do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
>   entry_SYSCALL_64_after_hwframe+0x63/0xcd
>   [...]
>   </TASK>
>
> The buggy address belongs to the physical page:
> page:ffffea00006ef600 refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x1bbd8
> head:ffffea00006ef600 order:3 compound_mapcount:0 compound_pincount:0
> flags: 0xfff00000010200(slab|head|node=0|zone=1|lastcpupid=0x7ff)
> page dumped because: kasan: bad access detected
>
> Memory state around the buggy address:
>   ffff88801bbdfe00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>   ffff88801bbdfe80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
> > ffff88801bbdff00: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>                     ^
>   ffff88801bbdff80: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>   ffff88801bbe0000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
> ==================================================================
>
> Kernel will tries to read record and parse MFT from disk in
> ntfs_read_mft().
>
> Yet the problem is that during enumerating attributes in record,
> kernel doesn't check whether run_off field loading from the disk
> is a valid value.
>
> To be more specific, if attr->nres.run_off is larger than attr->size,
> kernel will passes an invalid argument run_buf_size in
> run_unpack_ex(), which having an integer overflow. Then this invalid
> argument will triggers the slab-out-of-bounds Read bug as above.
>
> This patch solves it by adding the sanity check between
> the offset to packed runs and attribute size.
>
> link: https://lore.kernel.org/all/0000000000009145fc05e94bd5c3@xxxxxxxxxx/#t
> Reported-and-tested-by: syzbot+8d6fbb27a6aded64b25b@xxxxxxxxxxxxxxxxxxxxxxxxx
> Signed-off-by: Hawkins Jiawei <yin31149@xxxxxxxxx>
> ---
> v1 -> v2:
>    return -EINVAL when roff is out-of-bounds
>
>   fs/ntfs3/inode.c | 7 +++++++
>   1 file changed, 7 insertions(+)
>
> diff --git a/fs/ntfs3/inode.c b/fs/ntfs3/inode.c
> index 51363d4e8636..10723231e482 100644
> --- a/fs/ntfs3/inode.c
> +++ b/fs/ntfs3/inode.c
> @@ -365,6 +365,13 @@ static struct inode *ntfs_read_mft(struct inode *inode,
>   	roff = le16_to_cpu(attr->nres.run_off);
>   
>   	t64 = le64_to_cpu(attr->nres.svcn);
> +
> +	/* offset to packed runs is out-of-bounds */
> +	if (roff > asize) {
> +		err = -EINVAL;
> +		goto out;
> +	}
> +
>   	err = run_unpack_ex(run, sbi, ino, t64, le64_to_cpu(attr->nres.evcn),
>   			    t64, Add2Ptr(attr, roff), asize - roff);
>   	if (err < 0)

Thanks for patch, applied!