Re: [PATCH] ALSA: usb-audio: Add the number of endpoints checked was 0

[Takashi Iwai <tiwai@suse.de>]

On Mon, 09 Mar 2026 08:32:29 +0100,
Edward Adam Davis wrote:
> 
> The user constructed a corrupted USB device, causing the USB device
> enumeration phase to fail to resolve any endpoints. This resulted in
> a null pointer dereference reported in [1] when the USB sound card
> driver executed probe to initialize the mixer. 
> 
> To avoid the problem reported in [1], a check was added to ensure that
> the number of endpoints contained in the interface was 0 when checking
> the mixer status.
> 
> [1]
> KASAN: null-ptr-deref in range [0x0000000000000000-0x0000000000000007]
> RIP: 0010:usb_endpoint_num include/uapi/linux/usb/ch9.h:479 [inline]
> RIP: 0010:scarlett2_find_fc_interface sound/usb/mixer_scarlett2.c:8261 [inline]
> RIP: 0010:scarlett2_init_private sound/usb/mixer_scarlett2.c:8295 [inline]
> RIP: 0010:snd_scarlett2_controls_create sound/usb/mixer_scarlett2.c:8684 [inline]
> RIP: 0010:snd_scarlett2_init.cold+0xbad/0x6c79 sound/usb/mixer_scarlett2.c:9407
> Call Trace:
>  snd_usb_mixer_apply_create_quirk+0x1c21/0x2b80 sound/usb/mixer_quirks.c:4446
>  snd_usb_create_mixer+0x7a2/0x1910 sound/usb/mixer.c:3641
>  usb_audio_probe+0xf6d/0x3a90 sound/usb/card.c:1033
> 
> Reported-by: syzbot+ae893a8901067fde2741@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=ae893a8901067fde2741 
> Tested-by: syzbot+ae893a8901067fde2741@syzkaller.appspotmail.com
> Signed-off-by: Edward Adam Davis <eadavis@qq.com>
> ---
>  sound/usb/mixer.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/sound/usb/mixer.c b/sound/usb/mixer.c
> index ac8c71ba9483..bd28caec3004 100644
> --- a/sound/usb/mixer.c
> +++ b/sound/usb/mixer.c
> @@ -3563,8 +3563,12 @@ static int snd_usb_mixer_status_create(struct usb_mixer_interface *mixer)
>  	int buffer_length;
>  	unsigned int epnum;
>  
> +	epnum = get_iface_desc(mixer->hostif)->bNumEndpoints;
> +	if (epnum == 0)
> +		return -EINVAL;
> +
>  	/* we need one interrupt input endpoint */
> -	if (get_iface_desc(mixer->hostif)->bNumEndpoints < 1)
> +	if (epnum < 1)
>  		return 0;

The change looks wrong.  We don't want to return an error here, as
mixer->urb is optional.  And the second check introduced above is
utterly nonsense; epnum is an unsigned int, so epnum < 1 means only
epnum == 0.

The problem is rather in the scarlett2 side, and it should have a
check there, not in the core function.


thanks,

Takashi