Re: PANIC: double fault in fixup_bad_iret

[Peter Zijlstra]

On Fri, May 29, 2020 at 05:57:10PM +0200, Thomas Gleixner wrote:
> Dmitry,
> 
> Dmitry Vyukov <dvyukov@xxxxxxxxxx> writes:
> > On Fri, May 29, 2020 at 3:14 PM syzbot
> > <syzbot+dc1fa714cb070b184db5@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> >>
> >> Hello,
> >>
> >> syzbot found the following crash on:
> >>
> >> HEAD commit:    7b4cb0a4 Add linux-next specific files for 20200525
> >> git tree:       linux-next
> >> console output: https://syzkaller.appspot.com/x/log.txt?x=15dc34ba100000
> >> kernel config:  https://syzkaller.appspot.com/x/.config?x=47b0740d89299c10
> >> dashboard link: https://syzkaller.appspot.com/bug?extid=dc1fa714cb070b184db5
> >> compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> >> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=14678626100000
> >> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1017ef06100000
> >>
> >> IMPORTANT: if you fix the bug, please add the following tag to the commit:
> >> Reported-by: syzbot+dc1fa714cb070b184db5@xxxxxxxxxxxxxxxxxxxxxxxxx
> >
> > From the reproducer it seems to be either x86 related or ptrace
> > related.
> >
> >> RIP: 0010:fixup_bad_iret+0x24/0x170 arch/x86/kernel/traps.c:665
> 
> as a quick assumption that's related to KASAN in fixup_bad_iret() which
> is a frightenly bad idea. I'm about to verify.

Like with KCSAN, we should blanket kill KASAN/UBSAN and friends (at the
very least in arch/x86/) until they get that function attribute stuff
sorted.