Re: [PATCH 4.14 61/81] ALSA: seq: Protect in-kernel ioctl calls with mutex

[Sasha Levin]

On Thu, Jun 13, 2019 at 05:44:15PM +0200, Takashi Iwai wrote:
> On Thu, 13 Jun 2019 17:39:46 +0200,
> Sasha Levin wrote:
> > On Thu, Jun 13, 2019 at 11:13:55AM +0200, Takashi Iwai wrote:
> > >On Thu, 13 Jun 2019 11:11:22 +0200,
> > >Greg Kroah-Hartman wrote:
> > >>
> > >> On Thu, Jun 13, 2019 at 11:02:54AM +0200, Takashi Iwai wrote:
> > >> > On Thu, 13 Jun 2019 10:33:44 +0200,
> > >> > Greg Kroah-Hartman wrote:
> > >> > >
> > >> > > [ Upstream commit feb689025fbb6f0aa6297d3ddf97de945ea4ad32 ]
> > >> > >
> > >> > > ALSA OSS sequencer calls the ioctl function indirectly via
> > >> > > snd_seq_kernel_client_ctl().  While we already applied the protection
> > >> > > against races between the normal ioctls and writes via the client's
> > >> > > ioctl_mutex, this code path was left untouched.  And this seems to be
> > >> > > the cause of still remaining some rare UAF as spontaneously triggered
> > >> > > by syzkaller.
> > >> > >
> > >> > > For the sake of robustness, wrap the ioctl_mutex also for the call via
> > >> > > snd_seq_kernel_client_ctl(), too.
> > >> > >
> > >> > > Reported-by: syzbot+e4c8abb920efa77bace9@xxxxxxxxxxxxxxxxxxxxxxxxx
> > >> > > Signed-off-by: Takashi Iwai <tiwai@xxxxxxx>
> > >> > > Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
> > >> >
> > >> > This commit is reverted later by commit f0654ba94e33.
> > >> > So please drop this.  The proper fix is done later by commit
> > >> > 7c32ae35fbf9 ("ALSA: seq: Cover unsubscribe_port() in list_mutex")
> > >> >
> > >> > Ditto for 4.19.y and 5.1.y.
> > >>
> > >> Now dropped everywhere, and I added 7c32ae35fbf9 ("ALSA: seq: Cover
> > >> unsubscribe_port() in list_mutex") everywhere instead.
> > >
> > >Thanks!
> > >
> > >BTW, do we have a systematic check whether the selected stable commit
> > >is reverted in a later commit?  At least, you can track it as long as
> > >Fixes tag is properly set.
> >
> > I have that scripting in place, and I usually check it once before I
> > send the initial reviews and then once Greg does the -rc release.
>
> OK, good to hear.  So this time must be some exceptional error.

I still run these manually (I guess it's time to automate that too), so
it happens sometimes that folks notice these faster :)

--
Thanks,
Sasha