Re: Crashes in perf_event_ctx_lock_nested

[Guenter Roeck]

On Tue, Oct 31, 2017 at 02:50:59PM -0400, Don Zickus wrote:
> On Tue, Oct 31, 2017 at 10:16:22AM -0700, Guenter Roeck wrote:
> > On Tue, Oct 31, 2017 at 02:48:50PM +0100, Peter Zijlstra wrote:
> > > On Mon, Oct 30, 2017 at 03:45:12PM -0700, Guenter Roeck wrote:
> > > > I added some logging and a long msleep() in hardlockup_detector_perf_cleanup().
> > > > Here is the result:
> > > > 
> > > > [    0.274361] NMI watchdog: ############ hardlockup_detector_perf_init
> > > > [    0.274915] NMI watchdog: ############ hardlockup_detector_event_create(0)
> > > > [    0.277049] NMI watchdog: ############ hardlockup_detector_perf_cleanup
> > > > [    0.277593] NMI watchdog: ############ hardlockup_detector_perf_enable(0)
> > > > [    0.278027] NMI watchdog: ############ hardlockup_detector_event_create(0)
> > > > [    1.312044] NMI watchdog: ############ hardlockup_detector_perf_cleanup done
> > > > [    1.385122] NMI watchdog: ############ hardlockup_detector_perf_enable(1)
> > > > [    1.386028] NMI watchdog: ############ hardlockup_detector_event_create(1)
> > > > [    1.466102] NMI watchdog: ############ hardlockup_detector_perf_enable(2)
> > > > [    1.475536] NMI watchdog: ############ hardlockup_detector_event_create(2)
> > > > [    1.535099] NMI watchdog: ############ hardlockup_detector_perf_enable(3)
> > > > [    1.535101] NMI watchdog: ############ hardlockup_detector_event_create(3)
> > > 
> > > > [    7.222816] NMI watchdog: ############ hardlockup_detector_perf_disable(0)
> > > > [    7.230567] NMI watchdog: ############ hardlockup_detector_perf_disable(1)
> > > > [    7.243138] NMI watchdog: ############ hardlockup_detector_perf_disable(2)
> > > > [    7.250966] NMI watchdog: ############ hardlockup_detector_perf_disable(3)
> > > > [    7.258826] NMI watchdog: ############ hardlockup_detector_perf_enable(1)
> > > > [    7.258827] NMI watchdog: ############ hardlockup_detector_perf_cleanup
> > > > [    7.258831] NMI watchdog: ############ hardlockup_detector_perf_enable(2)
> > > > [    7.258833] NMI watchdog: ############ hardlockup_detector_perf_enable(0)
> > > > [    7.258834] NMI watchdog: ############ hardlockup_detector_event_create(2)
> > > > [    7.258835] NMI watchdog: ############ hardlockup_detector_event_create(0)
> > > > [    7.260169] NMI watchdog: ############ hardlockup_detector_perf_enable(3)
> > > > [    7.260170] NMI watchdog: ############ hardlockup_detector_event_create(3)
> > > > [    7.494251] NMI watchdog: ############ hardlockup_detector_event_create(1)
> > > > [    8.287135] NMI watchdog: ############ hardlockup_detector_perf_cleanup done
> > > > 
> > > > Looks like there are a number of problems: hardlockup_detector_event_create()
> > > > creates the event data structure even if it is already created, 
> > > 
> > > Right, that does look dodgy. And on its own should be fairly straight
> > > forward to cure. But I'd like to understand the rest of it first.
> > > 
> > > > and hardlockup_detector_perf_cleanup() runs unprotected and in
> > > > parallel to the enable/create functions.
> > > 
> > > Well, looking at the code, cpu_maps_update_begin() aka.
> > > cpu_add_remove_lock is serializing cpu_up() and cpu_down() and _should_
> > > thereby also serialize cleanup vs the smp_hotplug_thread operations.
> > > 
> > > Your trace does indeed indicate this is not the case, but I cannot, from
> > > the code, see how this could happen.
> > > 
> > > Could you use trace_printk() instead and boot with
> > > "trace_options=stacktrace" ?
> > > 
> > Attached. Let me know if you need more information. Note this is with
> > msleep(1000) in the cleanup function to avoid the crash.
> > 
> > > > ALso, the following message is seen twice.
> > > > 
> > > > [    0.278758] NMI watchdog: Enabled. Permanently consumes one hw-PMU counter.
> > > > [    7.258838] NMI watchdog: Enabled. Permanently consumes one hw-PMU counter.
> > > > 
> > > > I don't offer a proposed patch since I have no idea how to best solve the
> > > > problem.
> > > > 
> > > > Also, is the repeated enable/disable/cleanup as part of the normal boot
> > > > really necessary ?
> > > 
> > > That's weird, I don't see that on my machines. We very much only bring
> > > up the CPUs _once_. Also note they're 7s apart. Did you do something
> > > funny like resume-from-disk or so?
> > 
> > No, just whatever Chrome OS does when it starts the kernel. The hardware
> > used in this test is a Google Pixelbook, though we have also seen the problem
> > with other Chromebooks.
> 
> Is Chrome OS, changing the default timeout from 10s to something else?
> That would explain it as a script is executed late in the boot cycle and
> explain the quick restart.
> 

Correct, Chrome OS changes the timeout from 10 to 5 seconds.

A little experiment suggests that the problem can be triggered by updating
/proc/sys/kernel/watchdog_thresh. hardlockup_detector_perf_enable() is
called while hardlockup_detector_perf_cleanup() is running.

Guenter