INFO: trying to register non-static key in try_to_wake_up

Dmitry Vyukov dvyukov at google.com
Fri Apr 3 08:57:54 UTC 2020


On Wed, Apr 1, 2020 at 11:06 AM Daniel Vetter <daniel at ffwll.ch> wrote:
> > > > On Tue, Mar 31, 2020 at 2:18 PM Bartlomiej Zolnierkiewicz
> > > > <b.zolnierkie at samsung.com> wrote:
> > > > >
> > > > >
> > > > > On 3/31/20 12:18 PM, Dmitry Vyukov wrote:
> > > > > > On Tue, Mar 31, 2020 at 11:57 AM Peter Zijlstra <peterz at infradead.org> wrote:
> > > > > >>
> > > > > >> On Mon, Mar 30, 2020 at 10:01:12PM -0700, syzbot wrote:
> > > > > >>> Hello,
> > > > > >>>
> > > > > >>> syzbot found the following crash on:
> > > > > >>>
> > > > > >>> HEAD commit:    9420e8ad Merge tag 'for-linus' of git://git.kernel.org/pub..
> > > > > >>> git tree:       upstream
> > > > > >>> console output: https://protect2.fireeye.com/url?k=0756a78d-5a9a6c49-07572cc2-0cc47a314e9a-e4dc8b657d340686&u=https://syzkaller.appspot.com/x/log.txt?x=1206ed4be00000
> > > > > >>> kernel config:  https://protect2.fireeye.com/url?k=43211072-1eeddbb6-43209b3d-0cc47a314e9a-3bd45a19932c37c8&u=https://syzkaller.appspot.com/x/.config?x=27392dd2975fd692
> > > > > >>> dashboard link: https://protect2.fireeye.com/url?k=bf7a6153-e2b6aa97-bf7bea1c-0cc47a314e9a-c64073ee605efb7b&u=https://syzkaller.appspot.com/bug?extid=e84d7ebd1361da13c356
> > > > > >>> compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> > > > > >>>
> > > > > >>> Unfortunately, I don't have any reproducer for this crash yet.
> > > > > >>>
> > > > > >>> IMPORTANT: if you fix the bug, please add the following tag to the commit:
> > > > > >>> Reported-by: syzbot+e84d7ebd1361da13c356 at syzkaller.appspotmail.com
> > > > > >>>
> > > > > >>> INFO: trying to register non-static key.
> > > > > >>> the code is fine but needs lockdep annotation.
> > > > > >>> turning off the locking correctness validator.
> > > > > >>> CPU: 1 PID: 1014 Comm: syz-executor.0 Not tainted 5.6.0-rc7-syzkaller #0
> > > > > >>> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> > > > > >>> Call Trace:
> > > > > >>>  <IRQ>
> > > > > >>>  __dump_stack lib/dump_stack.c:77 [inline]
> > > > > >>>  dump_stack+0x188/0x20d lib/dump_stack.c:118
> > > > > >>>  assign_lock_key kernel/locking/lockdep.c:880 [inline]
> > > > > >>>  register_lock_class+0x14c4/0x1540 kernel/locking/lockdep.c:1189
> > > > > >>>  __lock_acquire+0xfc/0x3ca0 kernel/locking/lockdep.c:3836
> > > > > >>>  lock_acquire+0x197/0x420 kernel/locking/lockdep.c:4484
> > > > > >>>  __raw_spin_lock_irqsave include/linux/spinlock_api_smp.h:110 [inline]
> > > > > >>>  _raw_spin_lock_irqsave+0x8c/0xbf kernel/locking/spinlock.c:159
> > > > > >>>  try_to_wake_up+0x9f/0x17c0 kernel/sched/core.c:2547
> > > > > >>
> > > > > >> That's p->pi_lock, which gets initialized in rt_mutex_init_task() in
> > > > > >> copy_process(). This should be impossible. Very odd.
> > > > > >
> > > > > > The stack mentions fbdev, which is a red flag at the moment. There are
> > > > > > a dozen of bad bugs in fbdev and around. Just few days ago Andy
> > > > > > pointed to another "impossible" crash "general protection fault in
> > > > > > do_syscall_64" which is related to dri:
> > > > > > https://protect2.fireeye.com/url?k=0cb8ad06-517466c2-0cb92649-0cc47a314e9a-a20c11191483c65b&u=https://syzkaller.appspot.com/bug?id=0ec7b2602b1ff40f0d34f38baa4ba1640727c3d9
> > > > > > https://protect2.fireeye.com/url?k=614292e3-3c8e5927-614319ac-0cc47a314e9a-aeda6d72c01a7b0e&u=https://groups.google.com/forum/#!msg/syzkaller-bugs/ePqhfYx0-8M/Q_Urt97iAAAJ
> > > > > >
> > > > > > There are probably more random manifestations of these bugs already,
> > > > > > and I guess we will be getting more.
> > > > > >
> > > > > > +fbdev maintainers
> > > > >
> > > > > Thank you for the report.
> > > > >
> > > > > fbdev is in the maintenance mode and no new features or drivers are
> > > > > being added so syzbot reports are not for a new bugs (regressions) and
> > > > > are not a priority (at least to me).
> > > >
> > > > Yup same here, I've seen a pile of syzbot reports for fbdev (and also
> > > > vt, or combinations of them since fbdev is linked to vt through fbcon)
> > > > fly by. But I really don't have to deal with these, my recommendation
> > > > to anyone who cares about security are:
> > > > - Don't enable vt
> > > > - Don't enable fbdev
> > >
> > > 1. How do we deliver this message to relevant people?
> > >
> > > Because:
> > >
> > > $ grep FBDEV syzkaller/dashboard/config/upstream-kasan.config
> > > CONFIG_DRM_FBDEV_EMULATION=y
> > > CONFIG_DRM_FBDEV_OVERALLOC=100
> > > # CONFIG_DRM_FBDEV_LEAK_PHYS_SMEM is not set
> > > CONFIG_XEN_FBDEV_FRONTEND=y
> > >
> > > and my current work machine:
> > >
> > > $ grep FBDEV /boot/config-5.2.17-1-amd64
> > > CONFIG_DRM_FBDEV_EMULATION=y
> > > CONFIG_DRM_FBDEV_OVERALLOC=100
> > > # CONFIG_DRM_FBDEV_LEAK_PHYS_SMEM is not set
> > > CONFIG_XEN_FBDEV_FRONTEND=y
> >
> > Yeah I know it's been like this since forever. In theory you could
> > build a fbdev/fbcon less distro since years (the last bit for a proof
> > of concept was kmscon/systemd-consoled), but the amount of investment
> > into classic linux desktop is so little that it's impossible to get
> > this funded. CrOS fixed this a while ago iirc though.
> >
> > I think to fix the syzbot issues all we'd need is a competent intern
> > for a few months, that should take care of the worst stuff. Obviously
> > wont include getting a test suite going, nor fixing any of the
> > fundamental issues. But duct-taping over all the bugs should be
> > possible (it's what we've been doing for well over 10 years by now in
> > fbdev/fbocn/vt code anyway). I'd be willing to help mentoring, but
> > that's about all I can do.
> >
> > Adding Matthew Garret, I have discussed with him in the past finding
> > some funding for linux desktop stuff like this.

I will keep this in mind. We _may_ get some interns this year who
_may_ be interested in fixing Linux kernel bugs (but otherwise
extending syzkaller descriptions).

FTR, there is also some follow up on twitter re extending
https://github.com/a13xp0p0v/kconfig-hardened-check to capture such
recommendations:
https://twitter.com/dvyukov/status/1245969522869309441
https://github.com/a13xp0p0v/kconfig-hardened-check/issues/38


> > > 2. What do we do with fbdev testing on syzbot? Is there a way to
> > > disable all of the unsupported stuff? But if we disable it, we don't
> > > find any regressions as well. And in the end that's what is in the
> > > mainline kernel and is still enabled in distros (at least in the 2
> > > real configs I can grep now).
> >
> > This would be bad I agree, but it's not any worse than the state of
> > things the past 10 years. That's roughly for as long as fbdev has been
> > in maintainance only mode, meaning "we'll apply patches if they come".
> > Without Bart volunteering, we wouldn't even have that much really.
>
> Oh wrt disabling fbdev: Make sure CONFIG_FB isn't set. Unfortunately a
> pile of other things select that, for convenience (like
> CONFIG_DRM_KMS_FB_HELPER).
>
> That should get rid of all the problematic fbdev code.
>
> From what I've seen in some of the syzbot mails flying by we also have
> issues in vt and console code blowing up (not just on fbcon/fbdev, but
> also e.g. on vgacon). That stuff you'll still hit. But maybe you can
> trick Greg KH into fixing the vt/console.c issues, he just claimed
> that :-P
> -Daniel
>
> >
> > > > All that code has been developed long ago, in a much more innocent
> > > > time. If someone wants to fix this you'd not just need to fix all the
> > > > syzbot stuff, but also ramp up a full testsuite for all the ioctl, and
> > > > all the corner-cases. Plus also fix some of the horrendous locking in
> > > > there, probably.
> > > >
> > > > Multi-year effort, easily.
> > > >
> > > > Regressions I'll obviously try to handle, but none of these are. It's
> > > > just syzbot has become smarter at hitting bugs in fbdev and vt
> > > > subsystems (or maybe the hw the virtual machines emulate has become
> > > > more varied, some of the reports are for fun stuff like vgacon ...).
> > > >
> > > > Cheers, Daniel
> > > >
> > > > > I have only resources to review/merge pending fbdev patches from time
> > > > > to time so any help in fixing these syzbot reports is welcomed (there
> > > > > have been a few fbdev related syzbot reports recently).
> > > > >
> > > > > Also please note that fbdev is maintained through drm-misc tree so
> > > > > patches can also be handled by other drm-misc maintainers in case I'm
> > > > > not available / busy with other things.
> > > > >
> > > > > Best regards,
> > > > > --
> > > > > Bartlomiej Zolnierkiewicz
> > > > > Samsung R&D Institute Poland
> > > > > Samsung Electronics
> > > > >
> > > > > >>>  wake_up_worker kernel/workqueue.c:836 [inline]
> > > > > >>>  insert_work+0x2ad/0x3a0 kernel/workqueue.c:1337
> > > > > >>>  __queue_work+0x50d/0x1280 kernel/workqueue.c:1488
> > > > > >>>  call_timer_fn+0x195/0x760 kernel/time/timer.c:1404
> > > > > >>>  expire_timers kernel/time/timer.c:1444 [inline]
> > > > > >>>  __run_timers kernel/time/timer.c:1773 [inline]
> > > > > >>>  __run_timers kernel/time/timer.c:1740 [inline]
> > > > > >>>  run_timer_softirq+0x412/0x1600 kernel/time/timer.c:1786
> > > > > >>>  __do_softirq+0x26c/0x99d kernel/softirq.c:292
> > > > > >>>  invoke_softirq kernel/softirq.c:373 [inline]
> > > > > >>>  irq_exit+0x192/0x1d0 kernel/softirq.c:413
> > > > > >>>  exiting_irq arch/x86/include/asm/apic.h:546 [inline]
> > > > > >>>  smp_apic_timer_interrupt+0x19e/0x600 arch/x86/kernel/apic/apic.c:1146
> > > > > >>>  apic_timer_interrupt+0xf/0x20 arch/x86/entry/entry_64.S:829
> > > > > >>>  </IRQ>
> > > > > _______________________________________________
> > > > > dri-devel mailing list
> > > > > dri-devel at lists.freedesktop.org
> > > > > https://lists.freedesktop.org/mailman/listinfo/dri-devel
> > > >
> > > >
> > > >
> > > > --
> > > > Daniel Vetter
> > > > Software Engineer, Intel Corporation
> > > > +41 (0) 79 365 57 48 - http://blog.ffwll.ch
> >
> >
> >
> > --
> > Daniel Vetter
> > Software Engineer, Intel Corporation
> > +41 (0) 79 365 57 48 - http://blog.ffwll.ch
>
>
>
> --
> Daniel Vetter
> Software Engineer, Intel Corporation
> +41 (0) 79 365 57 48 - http://blog.ffwll.ch


More information about the dri-devel mailing list