2024-02-16 13:01:19

by Arnd Bergmann

[permalink] [raw]
Subject: [PATCH] ALSA: core: fix buffer overflow in test_format_fill_silence()

From: Arnd Bergmann <[email protected]>

KASAN caught a buffer overflow with the hardcoded 2048 byte buffer
size, when 2080 bytes are written to it:

BUG: KASAN: slab-out-of-bounds in snd_pcm_format_set_silence+0x3bc/0x3e4
Write of size 8 at addr ffff0000c8149800 by task kunit_try_catch/1297

CPU: 0 PID: 1297 Comm: kunit_try_catch Tainted: G N 6.8.0-rc4-next-20240216 #1
Hardware name: linux,dummy-virt (DT)
Call trace:
kasan_report+0x78/0xc0
__asan_report_store_n_noabort+0x1c/0x28
snd_pcm_format_set_silence+0x3bc/0x3e4
_test_fill_silence+0xdc/0x298
test_format_fill_silence+0x110/0x228
kunit_try_run_case+0x144/0x3bc
kunit_generic_run_threadfn_adapter+0x50/0x94
kthread+0x330/0x3e8
ret_from_fork+0x10/0x20

Allocated by task 1297:
__kmalloc+0x17c/0x2f0
kunit_kmalloc_array+0x2c/0x78
test_format_fill_silence+0xcc/0x228
kunit_try_run_case+0x144/0x3bc
kunit_generic_run_threadfn_adapter+0x50/0x94
kthread+0x330/0x3e8
ret_from_fork+0x10/0x20

Replace the incorrect size with the correct length of 260 64-bit samples.

Reported-by: Naresh Kamboju <[email protected]>
Fixes: 3e39acf56ede ("ALSA: core: Add sound core KUnit test")
Signed-off-by: Arnd Bergmann <[email protected]>
---
Naresh, I slightly changed the patch to make the computation more obvious,
can you test again to make sure I got this right?
---
sound/core/sound_kunit.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sound/core/sound_kunit.c b/sound/core/sound_kunit.c
index 5d5a7bf88de4..7f16101ece7a 100644
--- a/sound/core/sound_kunit.c
+++ b/sound/core/sound_kunit.c
@@ -8,7 +8,7 @@
#include <sound/core.h>
#include <sound/pcm.h>

-#define SILENCE_BUFFER_SIZE 2048
+#define SILENCE_BUFFER_SIZE (sizeof(u64) * 260)
#define SILENCE(...) { __VA_ARGS__ }
#define DEFINE_FORMAT(fmt, pbits, wd, endianness, signd, silence_arr) { \
.format = SNDRV_PCM_FORMAT_##fmt, .physical_bits = pbits, \
--
2.39.2



2024-02-16 23:22:40

by Ivan Orlov

[permalink] [raw]
Subject: Re: [PATCH] ALSA: core: fix buffer overflow in test_format_fill_silence()

On 2/16/24 13:00, Arnd Bergmann wrote:
> From: Arnd Bergmann <[email protected]>
>
> KASAN caught a buffer overflow with the hardcoded 2048 byte buffer
> size, when 2080 bytes are written to it:
>
> BUG: KASAN: slab-out-of-bounds in snd_pcm_format_set_silence+0x3bc/0x3e4
> Write of size 8 at addr ffff0000c8149800 by task kunit_try_catch/1297
>
> CPU: 0 PID: 1297 Comm: kunit_try_catch Tainted: G N 6.8.0-rc4-next-20240216 #1
> Hardware name: linux,dummy-virt (DT)
> Call trace:
> kasan_report+0x78/0xc0
> __asan_report_store_n_noabort+0x1c/0x28
> snd_pcm_format_set_silence+0x3bc/0x3e4
> _test_fill_silence+0xdc/0x298
> test_format_fill_silence+0x110/0x228
> kunit_try_run_case+0x144/0x3bc
> kunit_generic_run_threadfn_adapter+0x50/0x94
> kthread+0x330/0x3e8
> ret_from_fork+0x10/0x20
>
> Allocated by task 1297:
> __kmalloc+0x17c/0x2f0
> kunit_kmalloc_array+0x2c/0x78
> test_format_fill_silence+0xcc/0x228
> kunit_try_run_case+0x144/0x3bc
> kunit_generic_run_threadfn_adapter+0x50/0x94
> kthread+0x330/0x3e8
> ret_from_fork+0x10/0x20
>
> Replace the incorrect size with the correct length of 260 64-bit samples.
>
> Reported-by: Naresh Kamboju <[email protected]>
> Fixes: 3e39acf56ede ("ALSA: core: Add sound core KUnit test")
> Signed-off-by: Arnd Bergmann <[email protected]>
> ---
> Naresh, I slightly changed the patch to make the computation more obvious,
> can you test again to make sure I got this right?
> ---
> sound/core/sound_kunit.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/sound/core/sound_kunit.c b/sound/core/sound_kunit.c
> index 5d5a7bf88de4..7f16101ece7a 100644
> --- a/sound/core/sound_kunit.c
> +++ b/sound/core/sound_kunit.c
> @@ -8,7 +8,7 @@
> #include <sound/core.h>
> #include <sound/pcm.h>
>
> -#define SILENCE_BUFFER_SIZE 2048
> +#define SILENCE_BUFFER_SIZE (sizeof(u64) * 260)

I believe it would be good to define FILL_SILENCE_MAX_FRAMES to 260,
update the 'buf_samples' array correspondingly and define the
SILENCE_BUFFER_SIZE as (sizeof(u64) * FILL_SILENCE_MAX_FRAMES), so it
would be more clear where '260' came from.

Thank you for fixing this!
--
Kind regards,
Ivan Orlov