Re: KASAN: use-after-free Read in seccomp_notify_release (2)

[Tycho Andersen]

On Tue, Apr 23, 2019 at 02:39:52PM -0700, Kees Cook wrote:
> On Mon, Mar 25, 2019 at 1:02 AM syzbot
> <syzbot+b562969adb2e04af3442@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> >
> > Hello,
> >
> > syzbot found the following crash on:
> >
> > HEAD commit:    1bdd3dbf Merge tag 'io_uring-20190323' of git://git.kernel..
> > git tree:       upstream
> > console output: https://syzkaller.appspot.com/x/log.txt?x=12ae5b93200000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=9a31fb246de2a622
> > dashboard link: https://syzkaller.appspot.com/bug?extid=b562969adb2e04af3442
> > compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> > syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=162ca25d200000
> > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1733b53b200000
> >
> > The bug was bisected to:
> >
> > commit a799aea0988ea0d1b1f263e996fdad2f6133c680
> > Author: wenxu <wenxu@xxxxxxxxx>
> > Date:   Wed Jan 9 02:40:11 2019 +0000
> >
> >      netfilter: nft_flow_offload: Fix reverse route lookup
> 
> This bisection looks bogus?
> 
> However, I _can_ trigger the problem on this kernel version with this
> config. (And not with Linus's latest tree.)
> 
> The PoC is identical to the prior report[1] that we thought was fixed.
> Perhaps the fix didn't actually fix it? (I mean
> a811dc61559e0c8003f1086c2a4dc8e4d5ae4cb8) But it's been silent for 29
> days now, so I'm not sure what's going on.
> 
> Tycho are you able to reproduce this on the older tree?

It looks like this case is using TSYNC and NEW_LISTENER together:

syscall(__NR_seccomp, 1, 0xb, 0x20000140);

and I did fix a uaf with TSYNC and NEW_LISTENER in this series:

https://lore.kernel.org/lkml/20190306201413.14153-2-tycho@xxxxxxxx/T/#u

which seems like it's probably related. I don't think the above ever
got applied though?

Tycho