linux-kernel - Re: [syzbot] [sound?] [usb?] general protection fault in snd_usbmidi_do

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20250927154109.879926-1-aha310510@gmail.com>
Date: Sun, 28 Sep 2025 00:41:09 +0900
From: Jeongjun Park <aha310510@...il.com>
To: tiwai@...e.de
Cc: aha310510@...il.com,
	clemens@...isch.de,
	hdanton@...a.com,
	linux-kernel@...r.kernel.org,
	linux-sound@...r.kernel.org,
	perex@...ex.cz,
	syzbot+f02665daa2abeef4a947@...kaller.appspotmail.com,
	syzkaller-bugs@...glegroups.com
Subject: Re: [syzbot] [sound?] [usb?] general protection fault in snd_usbmidi_do_output

Hi,

Takashi Iwai <tiwai@...e.de> wrote:
>
> On Sat, 27 Sep 2025 12:36:07 +0200,
> Jeongjun Park wrote:
> >
> > syzbot <syzbot+f02665daa2abeef4a947@...kaller.appspotmail.com> wrote:
> > >
> > > Hello,
> > >
> > > syzbot has tested the proposed patch but the reproducer is still triggering an issue:
> > > KASAN: slab-use-after-free Write in snd_usbmidi_in_urb_complete
> > >
> > > ==================================================================
> > > BUG: KASAN: slab-use-after-free in snd_usbmidi_in_urb_complete+0x389/0x3c0 sound/usb/midi.c:251
> > > Write of size 1 at addr ffff888074717943 by task kworker/1:3/5866
> > >
> >
> > Wow, the UAF bug still occurs?
> >
> > But... this UAF seems to be a problem with how midi handles urb rather
> > than a problem with my patch.
> >
> > Is there something wrong with the way snd_usbmidi_in_urb_complete() is
> > implemented?
>
> This can be rather a missing kill-and-cleanup in the code path.
> So the patch like below.
>
> Could you check whether this works for you instead of your fix, too?
> timer_shutdown_sync() is already called in snd_usbmidi_disconnect(),
> and the call in snd_usbmidi_free() should be superfluous after this
> change.
>

Since both my tests and syzbot tests did not cause any bugs when applying
this patch, I think the root cause of this vulnerability is the missing
kill error_timer + urb.

So, I think it would be okay to patch it this way, but in addition, since
most of the endpoint delete-related code that existed in
snd_usbmidi_free() is now done in snd_usbmidi_disconnect(),
I think it would be appropriate to modify it to only perform
kfree(ep->out) as in this patch below.

---
 sound/usb/midi.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/sound/usb/midi.c b/sound/usb/midi.c
index acb3bf92857c..97e7e7662b12 100644
--- a/sound/usb/midi.c
+++ b/sound/usb/midi.c
@@ -1522,15 +1522,14 @@ static void snd_usbmidi_free(struct snd_usb_midi *umidi)
 {
 	int i;
 
+	if (!umidi->disconnected)
+		snd_usbmidi_disconnect(&umidi->list);
+
 	for (i = 0; i < MIDI_MAX_ENDPOINTS; ++i) {
 		struct snd_usb_midi_endpoint *ep = &umidi->endpoints[i];
-		if (ep->out)
-			snd_usbmidi_out_endpoint_delete(ep->out);
-		if (ep->in)
-			snd_usbmidi_in_endpoint_delete(ep->in);
+		kfree(ep->out);
 	}
 	mutex_destroy(&umidi->mutex);
-	timer_shutdown_sync(&umidi->error_timer);
 	kfree(umidi);
 }
 
--

>
> thanks,
>
> Takashi
>
> --- a/sound/usb/midi.c
> +++ b/sound/usb/midi.c
> @@ -1522,6 +1522,9 @@ static void snd_usbmidi_free(struct snd_usb_midi *umidi)
>  {
>         int i;
>
> +       if (!umidi->disconnected)
> +               snd_usbmidi_disconnect(&umidi->list);
> +
>         for (i = 0; i < MIDI_MAX_ENDPOINTS; ++i) {
>                 struct snd_usb_midi_endpoint *ep = &umidi->endpoints[i];
>                 if (ep->out)

Regards,
Jeongjun Park