KASAN: vmalloc-out-of-bounds Write in snd_pcm_hw_params

[Dipanjan Das]

Hi,

We would like to report the following bug which has been found by our
modified version of syzkaller.

======================================================
description: KASAN: vmalloc-out-of-bounds Write in snd_pcm_hw_params
affected file: sound/core/pcm_native.c
kernel version: 5.10.131
kernel commit: de62055f423f5dcb548f74cebd68f03c8903f73a
git tree: upstream
kernel config: https://syzkaller.appspot.com/x/.config?x=e49433cfed49b7d9
crash reproducer: attached
======================================================
Crash log:
======================================================
BUG: KASAN: vmalloc-out-of-bounds in memset include/linux/string.h:384 [inline]
BUG: KASAN: vmalloc-out-of-bounds in snd_pcm_hw_params+0x19b0/0x1db0
sound/core/pcm_native.c:799
Write of size 2097152 at addr ffffc900113b2000 by task syz-executor.5/14437

CPU: 1 PID: 14437 Comm: syz-executor.5 Tainted: G OE 5.10.131+ #3
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
1.13.0-1ubuntu1.1 04/01/2014
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0x107/0x163 lib/dump_stack.c:118
print_address_description.constprop.0.cold+0x5/0x4f7 mm/kasan/report.c:385
__kasan_report mm/kasan/report.c:545 [inline]
kasan_report.cold+0x1f/0x37 mm/kasan/report.c:562
check_memory_region_inline mm/kasan/generic.c:194 [inline]
check_memory_region+0x187/0x1e0 mm/kasan/generic.c:200
memset+0x20/0x40 mm/kasan/common.c:85
memset include/linux/string.h:384 [inline]
snd_pcm_hw_params+0x19b0/0x1db0 sound/core/pcm_native.c:799
snd_pcm_kernel_ioctl+0xd1/0x240 sound/core/pcm_native.c:3401
snd_pcm_oss_change_params_locked+0x17b6/0x3aa0 sound/core/oss/pcm_oss.c:965
snd_pcm_oss_change_params+0x76/0xd0 sound/core/oss/pcm_oss.c:1107
snd_pcm_oss_make_ready+0xb7/0x170 sound/core/oss/pcm_oss.c:1166
snd_pcm_oss_set_trigger.isra.0+0x34f/0x770 sound/core/oss/pcm_oss.c:2074
snd_pcm_oss_poll+0x679/0xb40 sound/core/oss/pcm_oss.c:2858
vfs_poll include/linux/poll.h:90 [inline]
do_pollfd fs/select.c:872 [inline]
do_poll fs/select.c:920 [inline]
do_sys_poll+0x63c/0xe40 fs/select.c:1014
__do_sys_poll fs/select.c:1079 [inline]
__se_sys_poll fs/select.c:1067 [inline]
__x64_sys_poll+0x18c/0x490 fs/select.c:1067
do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x7f095de4f4ed
Code: 02 b8 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa 48 89 f8 48
89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d
01 f0 ff ff 73 01 c3 48 c7 c1 b8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f095bdffbe8 EFLAGS: 00000246 ORIG_RAX: 0000000000000007
RAX: ffffffffffffffda RBX: 00007f095df6df60 RCX: 00007f095de4f4ed
RDX: 0000000000000009 RSI: 0000000000000001 RDI: 00000000200000c0
RBP: 00007f095bdffc40 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000001d
R13: 00007ffff286ceff R14: 00007f095df6df60 R15: 00007f095bdffd80


Memory state around the buggy address:
ffffc900115b1d00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
ffffc900115b1d80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
>ffffc900115b1e00: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
^
ffffc900115b1e80: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
ffffc900115b1f00: f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8 f8
==================================================================

--
Thanks and Regards,

Dipanjan

[Greg KH]

Wondeful, do you have a fix for this that solves the reported problem
that you have tested with the reproducer?

thanks,

greg k-h

[Takashi Iwai]

... or at least more detailed information.

The given log snippet alone doesn't help for further analysis, as it
doesn't show which device / driver is involved. The code is the
common helper and the condition for the trigger might be depending on
the driver side. The full kernel log might show which driver (IIUC,
it's /dev/adsp1) is in place.

Last but not least, you should check whether it's specific to your
5.10.x kernel or it's also seen with the latest upstream, too.


thanks,

Takashi

[Dipanjan Das]

On Sat, Jul 23, 2022 at 3:17 AM Takashi Iwai <ti...@suse.de> wrote:
>
> On Sat, 23 Jul 2022 09:00:08 +0200,
> Greg KH wrote:
> >
> > Wondeful, do you have a fix for this that solves the reported problem
> > that you have tested with the reproducer?
>
> ... or at least more detailed information.

Here is our analysis of the bug in the kernel v5.10.131.

During allocation, the `size` of the DMA buffer is not page-aligned:
https://elixir.bootlin.com/linux/v5.10.131/source/sound/core/memalloc.c#L149.
However, in sound/core/pcm_native.c:798
(https://elixir.bootlin.com/linux/v5.10.131/source/sound/core/pcm_native.c#L798),
the `size` variable is page-aligned before memset-ing the `dma_area`.
From the other BUG_ON assertions in other parts of the code, it looks
like the DMA area is not supposed to be equal to or greater than
0x200000 bytes. However, due to page-alignment, the `size` can indeed
get rounded up to 0x200000 which causes the out of bound access.

> Last but not least, you should check whether it's specific to your
> 5.10.x kernel or it's also seen with the latest upstream, too.

The bug is not reproducible on the latest mainline, because in
sound/core/memalloc.c:66
(https://github.com/torvalds/linux/blob/5de64d44968e4ae66ebdb0a2d08b443f189d3651/sound/core/memalloc.c#L66)
the allocation function `snd_dma_alloc_dir_pages()` now page-aligns
the `size` right before allocating the DMA buffer. Therefore, any
subsequent page-alignment, like the one in `snd_pcm_hw_params()` does
not cause an out of bound access.

[Lukas Bulwahn]

Great analysis!

Now, you would just need to identify the specific commit in the
mainline repository, where 'function `snd_dma_alloc_dir_pages()` now
page-aligns the `size` right before allocating the DMA buffer.', and
then ask for applying that commit to the v5.10 stable branch,
following the guide from
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html.
Greg KH and Sasha Levin are then going to let you know if that works
or needs rework to backport.


Lukas

[Takashi Iwai]

Thanks for the analysis. A good news is that, at least for the
vmalloc() case, it's a kind of false-positive; vmalloc() always takes
the full pages, so practically seen, the size is page-aligned. It's
fooling the memory checker, though.

But the similar problem could be seen with genalloc calls, and this
was fixed by the upstream commit
5c1733e33c888a3cb7f576564d8ad543d5ad4a9e
ALSA: memalloc: Align buffer allocations in page size

I suppose you can simply backport this commit to 5.10.y. Could you
confirm that this fixes your problem?


thanks,

Takashi

[Dipanjan Das]

On Tue, Jul 26, 2022 at 10:25 PM Takashi Iwai <ti...@suse.de> wrote:
>
> Thanks for the analysis. A good news is that, at least for the
> vmalloc() case, it's a kind of false-positive; vmalloc() always takes
> the full pages, so practically seen, the size is page-aligned. It's
> fooling the memory checker, though.
>
> But the similar problem could be seen with genalloc calls, and this
> was fixed by the upstream commit
> 5c1733e33c888a3cb7f576564d8ad543d5ad4a9e
> ALSA: memalloc: Align buffer allocations in page size
>
> I suppose you can simply backport this commit to 5.10.y. Could you
> confirm that this fixes your problem?

We confirm that the patch you proposed fixes the problem (blocks the
reproducer). How do we proceed with getting the issue fixed? Do we
send a patch according to the steps detailed here:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html?

[Takashi Iwai]

Don't worry, Greg already picked up the fix commit :)


thanks,

Takashi

[Greg KH]

Normally, yes, that is the correct process. But as Takashi mentioned, I
already picked it up as I happened to see this thread.

thanks,

greg k-h