[syzbot] [mm?] KMSAN: uninit-value in zswap_store

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: 614da38e2f7a Merge tag 'hid-for-linus-2024051401' of git:/..
git tree: upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=113f3fd6980000
kernel config: https://syzkaller.appspot.com/x/.config?x=f5d2cbf33633f507
dashboard link: https://syzkaller.appspot.com/bug?extid=9c1fe13fcb51574b249b
compiler: Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40

Unfortunately, I don't have any reproducer for this issue yet.

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/89eafb874b71/disk-614da38e.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/356000512ad9/vmlinux-614da38e.xz
kernel image: https://storage.googleapis.com/syzbot-assets/839c73939115/bzImage-614da38e.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+9c1fe1...@syzkaller.appspotmail.com

=====================================================
BUG: KMSAN: uninit-value in zswap_is_page_same_filled mm/zswap.c:1481 [inline]
BUG: KMSAN: uninit-value in zswap_store+0x1008/0x2ca0 mm/zswap.c:1553
zswap_is_page_same_filled mm/zswap.c:1481 [inline]
zswap_store+0x1008/0x2ca0 mm/zswap.c:1553
swap_writepage+0x126/0x4c0 mm/page_io.c:198
shmem_writepage+0x1826/0x1f70 mm/shmem.c:1518
pageout mm/vmscan.c:660 [inline]
shrink_folio_list+0x4a55/0x7910 mm/vmscan.c:1323
evict_folios+0x9d7f/0xcc20 mm/vmscan.c:4537
try_to_shrink_lruvec+0x160e/0x1a50 mm/vmscan.c:4733
shrink_one+0x66f/0xd40 mm/vmscan.c:4772
shrink_many mm/vmscan.c:4835 [inline]
lru_gen_shrink_node mm/vmscan.c:4935 [inline]
shrink_node+0x4856/0x55f0 mm/vmscan.c:5894
kswapd_shrink_node mm/vmscan.c:6704 [inline]
balance_pgdat mm/vmscan.c:6895 [inline]
kswapd+0x1eba/0x4460 mm/vmscan.c:7164
kthread+0x3e2/0x540 kernel/kthread.c:389
ret_from_fork+0x6d/0x90 arch/x86/kernel/process.c:147
ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:244

Uninit was stored to memory at:
memcpy_from_iter lib/iov_iter.c:73 [inline]
iterate_bvec include/linux/iov_iter.h:122 [inline]
iterate_and_advance2 include/linux/iov_iter.h:249 [inline]
iterate_and_advance include/linux/iov_iter.h:271 [inline]
__copy_from_iter lib/iov_iter.c:249 [inline]
copy_page_from_iter_atomic+0x12b7/0x2ae0 lib/iov_iter.c:481
generic_perform_write+0x4c1/0xc60 mm/filemap.c:3982
shmem_file_write_iter+0x2bd/0x2f0 mm/shmem.c:2920
do_iter_readv_writev+0x7e6/0x960
vfs_iter_write+0x459/0xd00 fs/read_write.c:895
lo_write_bvec drivers/block/loop.c:246 [inline]
lo_write_simple drivers/block/loop.c:267 [inline]
do_req_filebacked drivers/block/loop.c:491 [inline]
loop_handle_cmd drivers/block/loop.c:1907 [inline]
loop_process_work+0x1502/0x3440 drivers/block/loop.c:1942
loop_rootcg_workfn+0x2b/0x40 drivers/block/loop.c:1973
process_one_work kernel/workqueue.c:3267 [inline]
process_scheduled_works+0xa81/0x1bd0 kernel/workqueue.c:3348
worker_thread+0xea5/0x1560 kernel/workqueue.c:3429
kthread+0x3e2/0x540 kernel/kthread.c:389
ret_from_fork+0x6d/0x90 arch/x86/kernel/process.c:147
ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:244

Uninit was created at:
__alloc_pages+0x9d6/0xe70 mm/page_alloc.c:4598
alloc_pages_mpol+0x299/0x990 mm/mempolicy.c:2264
alloc_pages mm/mempolicy.c:2335 [inline]
folio_alloc+0x1d0/0x230 mm/mempolicy.c:2342
filemap_alloc_folio+0xa6/0x440 mm/filemap.c:984
__filemap_get_folio+0xa10/0x14b0 mm/filemap.c:1926
grow_dev_folio fs/buffer.c:1042 [inline]
grow_buffers fs/buffer.c:1108 [inline]
__getblk_slow fs/buffer.c:1134 [inline]
bdev_getblk+0x39b/0xc80 fs/buffer.c:1429
__getblk include/linux/buffer_head.h:355 [inline]
sb_getblk include/linux/buffer_head.h:361 [inline]
ext4_getblk+0x3da/0xe00 fs/ext4/inode.c:843
ext4_xattr_inode_write fs/ext4/xattr.c:1421 [inline]
ext4_xattr_inode_lookup_create fs/ext4/xattr.c:1596 [inline]
ext4_xattr_set_entry+0x574d/0x6880 fs/ext4/xattr.c:1718
ext4_xattr_block_set+0xb94/0x4fb0 fs/ext4/xattr.c:2037
ext4_xattr_move_to_block fs/ext4/xattr.c:2654 [inline]
ext4_xattr_make_inode_space fs/ext4/xattr.c:2729 [inline]
ext4_expand_extra_isize_ea+0x20bd/0x3560 fs/ext4/xattr.c:2821
__ext4_expand_extra_isize+0x5dc/0x680 fs/ext4/inode.c:5789
ext4_try_to_expand_extra_isize fs/ext4/inode.c:5832 [inline]
__ext4_mark_inode_dirty+0x70d/0xa10 fs/ext4/inode.c:5910
ext4_delete_inline_entry+0x650/0x7d0 fs/ext4/inline.c:1753
ext4_delete_entry+0x13f/0x7d0 fs/ext4/namei.c:2719
__ext4_unlink+0x9b8/0x11b0 fs/ext4/namei.c:3273
ext4_unlink+0x226/0x630 fs/ext4/namei.c:3321
vfs_unlink+0x676/0xa30 fs/namei.c:4343
do_unlinkat+0x823/0xe10 fs/namei.c:4407
__do_sys_unlinkat fs/namei.c:4450 [inline]
__se_sys_unlinkat fs/namei.c:4443 [inline]
__x64_sys_unlinkat+0x17c/0x230 fs/namei.c:4443
x64_sys_call+0x846/0x3b50 arch/x86/include/generated/asm/syscalls_64.h:264
do_syscall_x64 arch/x86/entry/common.c:52 [inline]
do_syscall_64+0xcf/0x1e0 arch/x86/entry/common.c:83
entry_SYSCALL_64_after_hwframe+0x77/0x7f

CPU: 1 PID: 88 Comm: kswapd1 Not tainted 6.9.0-syzkaller-02707-g614da38e2f7a #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 04/02/2024
=====================================================


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

If the report is already addressed, let syzbot know by replying with:
#syz fix: exact-commit-title

If you want to overwrite report's subsystems, reply with:
#syz set subsystems: new-subsystem
(See the list of subsystem names on the web dashboard)

If the report is a duplicate of another one, reply with:
#syz dup: exact-subject-of-another-report

If you want to undo deduplication, reply with:
#syz undup

[Yosry Ahmed]

On Sat, Jun 8, 2024 at 2:09 PM syzbot
<syzbot+9c1fe1...@syzkaller.appspotmail.com> wrote:
>
> Hello,
>
> syzbot found the following issue on:
>
> HEAD commit: 614da38e2f7a Merge tag 'hid-for-linus-2024051401' of git:/..
> git tree: upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=113f3fd6980000
> kernel config: https://syzkaller.appspot.com/x/.config?x=f5d2cbf33633f507
> dashboard link: https://syzkaller.appspot.com/bug?extid=9c1fe13fcb51574b249b
> compiler: Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
>
> Unfortunately, I don't have any reproducer for this issue yet.
>
> Downloadable assets:
> disk image: https://storage.googleapis.com/syzbot-assets/89eafb874b71/disk-614da38e.raw.xz
> vmlinux: https://storage.googleapis.com/syzbot-assets/356000512ad9/vmlinux-614da38e.xz
> kernel image: https://storage.googleapis.com/syzbot-assets/839c73939115/bzImage-614da38e.xz
>
> IMPORTANT: if you fix the issue, please add the following tag to the commit:
> Reported-by: syzbot+9c1fe1...@syzkaller.appspotmail.com

This doesn't look like a zswap bug, and perhaps not a bug at all but I
am not sure. Zswap is reading the data in a folio to check if it is
filled with a repeated pattern, and it is encountering uninitialized
data. I am not sure if this is expected behavior or not.

Could it just be that the size of the data written by syzbot is not
divisible by PAGE_SIZE, so part of the final page is left
uninitialized? Do we keep the unwritten parts of a shmem page
uninitialized by any chance?

Hugh, do you mind taking a quick look at whether this is a real bug?

If this is expected behavior, perhaps there is some annotation we can
use in zswap that it is fine to encounter uninitialized data when
reading the folio.

[Hugh Dickins]

On Mon, 10 Jun 2024, Yosry Ahmed wrote:
> On Sat, Jun 8, 2024 at 2:09 PM syzbot
> <syzbot+9c1fe1...@syzkaller.appspotmail.com> wrote:
> >
> > Hello,
> >
> > syzbot found the following issue on:
> >
> > HEAD commit: 614da38e2f7a Merge tag 'hid-for-linus-2024051401' of git:/..
> > git tree: upstream
> > console output: https://syzkaller.appspot.com/x/log.txt?x=113f3fd6980000
> > kernel config: https://syzkaller.appspot.com/x/.config?x=f5d2cbf33633f507
> > dashboard link: https://syzkaller.appspot.com/bug?extid=9c1fe13fcb51574b249b
> > compiler: Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40
> >
> > Unfortunately, I don't have any reproducer for this issue yet.
> >
> > Downloadable assets:
> > disk image: https://storage.googleapis.com/syzbot-assets/89eafb874b71/disk-614da38e.raw.xz
> > vmlinux: https://storage.googleapis.com/syzbot-assets/356000512ad9/vmlinux-614da38e.xz
> > kernel image: https://storage.googleapis.com/syzbot-assets/839c73939115/bzImage-614da38e.xz
> >
> > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > Reported-by: syzbot+9c1fe1...@syzkaller.appspotmail.com
>
> This doesn't look like a zswap bug, and perhaps not a bug at all but I
> am not sure. Zswap is reading the data in a folio to check if it is
> filled with a repeated pattern, and it is encountering uninitialized
> data. I am not sure if this is expected behavior or not.

I agree, it doesn't look like a zswap bug: zswap is just following orders.

>
> Could it just be that the size of the data written by syzbot is not
> divisible by PAGE_SIZE, so part of the final page is left
> uninitialized? Do we keep the unwritten parts of a shmem page
> uninitialized by any chance?

Shmem can keep uninitialized data pages around, but should be zeroing
what's still uninitialized before it can reach the outside world -
in this case being written to swap. I don't know of a shmem bug here.

>
> Hugh, do you mind taking a quick look at whether this is a real bug?
>
> If this is expected behavior, perhaps there is some annotation we can
> use in zswap that it is fine to encounter uninitialized data when
> reading the folio.

I've not been faced with a KMSAN report before, so I might well be
misunderstanding its language: but this looks like an ext4 "bug" to me.

I think the story that the three KMSAN stacks are telling is this:
syzbot has an ext4 filesystem on a loop device on a tmpfs file (I do
exactly that too); ext4 is doing some ext4_xattr_inode_write() business,
in the course of which it writes back a not-fully-initialized block to
the loop device; shmem faithfully copies that data into its pagecache,
then later when under memory pressure that page gets "written" out to
zswap: where zswap_is_page_same_filled() takes an interest in the data,
and KMSAN objects because some of it was not originally initialized.

If that's a correct interpretation, then it's probably not a big deal:
it's probably the uninitialized end of a buffer that's written out,
not any part of the "disk" which ext4 would ever show to a user; but
I do agree with KMSAN that ext4 would do better to clear that area,
rather than accidentally storing someone else's super-secret info.

Hugh

[Yosry Ahmed]

Thanks for taking a look at this and stating that shmem will not pass
uninitialized memory to swap/zswap.

>
> >
> > Hugh, do you mind taking a quick look at whether this is a real bug?
> >
> > If this is expected behavior, perhaps there is some annotation we can
> > use in zswap that it is fine to encounter uninitialized data when
> > reading the folio.
>
> I've not been faced with a KMSAN report before, so I might well be
> misunderstanding its language: but this looks like an ext4 "bug" to me.
>
> I think the story that the three KMSAN stacks are telling is this:
> syzbot has an ext4 filesystem on a loop device on a tmpfs file (I do
> exactly that too); ext4 is doing some ext4_xattr_inode_write() business,
> in the course of which it writes back a not-fully-initialized block to
> the loop device; shmem faithfully copies that data into its pagecache,
> then later when under memory pressure that page gets "written" out to
> zswap: where zswap_is_page_same_filled() takes an interest in the data,
> and KMSAN objects because some of it was not originally initialized.
>
> If that's a correct interpretation, then it's probably not a big deal:
> it's probably the uninitialized end of a buffer that's written out,
> not any part of the "disk" which ext4 would ever show to a user; but
> I do agree with KMSAN that ext4 would do better to clear that area,
> rather than accidentally storing someone else's super-secret info.

Interesting, I never saw an ext4 filesystem on a loop device on a
tmpfs file :) My understanding of the KMSAN stacks seems to match
yours. Thanks for looping in ext4 folks, hopefully they will shed more
light on this from their end.

[Jan Kara]

On Tue 11-06-24 02:13:59, Hugh Dickins wrote:
> On Mon, 10 Jun 2024, Yosry Ahmed wrote:
> > Hugh, do you mind taking a quick look at whether this is a real bug?
> >
> > If this is expected behavior, perhaps there is some annotation we can
> > use in zswap that it is fine to encounter uninitialized data when
> > reading the folio.
>
> I've not been faced with a KMSAN report before, so I might well be
> misunderstanding its language: but this looks like an ext4 "bug" to me.
>
> I think the story that the three KMSAN stacks are telling is this:
> syzbot has an ext4 filesystem on a loop device on a tmpfs file (I do
> exactly that too); ext4 is doing some ext4_xattr_inode_write() business,
> in the course of which it writes back a not-fully-initialized block to
> the loop device; shmem faithfully copies that data into its pagecache,
> then later when under memory pressure that page gets "written" out to
> zswap: where zswap_is_page_same_filled() takes an interest in the data,
> and KMSAN objects because some of it was not originally initialized.
>
> If that's a correct interpretation, then it's probably not a big deal:
> it's probably the uninitialized end of a buffer that's written out,
> not any part of the "disk" which ext4 would ever show to a user; but
> I do agree with KMSAN that ext4 would do better to clear that area,
> rather than accidentally storing someone else's super-secret info.

Yes, that seems to be accurate. ext4_xattr_inode_write() stores large
extended attribute in the inode and we don't bother to zero out the tail of
the block we use since we never access bytes beyond xattr size. Frankly I
don't consider this a bug since the uninitialized bytes are never exposed
to (unpriviledged) userspace. But I agree that out of pure precaution and
because it doesn't cost much in terms of performance we could zero out the
block tail.

Honza
--
Jan Kara <ja...@suse.com>
SUSE Labs, CR

[Eric Biggers]

Writing uninitialized memory to disk is definitely a bug.

- Eric