Re: [PATCH] tick: prefer a lower rating device only if it's CPU local device

[Sudeep Holla]

On Tue, Jul 03, 2018 at 08:04:37AM -0700, Kevin Hilman wrote:
> On Tue, Jul 3, 2018 at 3:54 AM Sudeep Holla <sudeep.holla@xxxxxxx> wrote:
> >
> > On Mon, Jul 02, 2018 at 04:44:33PM -0700, Kevin Hilman wrote:
> > > Hi Sudeep,
> > >
> > > On Wed, May 9, 2018 at 9:02 AM Sudeep Holla <sudeep.holla@xxxxxxx> wrote:
> > > >
> > > > Checking the equality of cpumask for both new and old tick device doesn't
> > > > ensure that it's CPU local device. This will cause issue if a low rating
> > > > clockevent tick device is registered first followed by the registration
> > > > of higher rating clockevent tick device.
> > > >
> > > > In such case, clockevents_released list will never get emptied as both
> > > > the devices get selected as preferred one and we will loop forever in
> > > > clockevents_notify_released.
> > > >
> > > > Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
> > > > Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > > > Signed-off-by: Sudeep Holla <sudeep.holla@xxxxxxx>
> > >
> > > I've got a arm32 board (meson8b-odroidc1) that's been failing in
> > > kernelCI.org since the merge window (boot log[1]), and I finally got
> > > around to bisecting it[2].  Unfortunately, the bisect pointed at a
> > > merge commit, but with some trial and error (and a suggestion by Arnd)
> > > I was able to test that revering $SUBJECT commit[3], my problem goes
> > > away.
> > >
> >
> > Interesting. Sorry for causing the regression.
> >
> > > Another interesting data point is that disabling SMP (either by
> > > "nosmp" on the command-line or CONFIG_SMP=n) also makes the problem go
> > > away, without needing to revert this patch.
> > >
> >
> > I am not sure of nosmp, but with CONFIG_SMP=n, TICK_BROADCAST also gets
> > disabled. dummy_timer won't be registered I assume.
> >
> > I am not sure if dummy_timer is selected as it's per_cpu but the rating
> > is low anyways.
> 
> > > AFAICT, this platform, is using a single timer as a clocksource
> > > ("amlogic,meson6-timer") which is not a per-CPU timer.
> > >
> >
> > Yes that's what I could gather from DT. But this is A5 right ? It may
> > have per CPU TWD(watchdof timer) but DT doesn't specify it, so should be
> > fine.
> >
> > > I ran out of time to keep digging on this issue, and I'm still not
> > > sure exactly what's going on, but I wanted to report it in case anyone
> > > else has any ideas, and so we can hopefully get it fixed during the
> > > -rc cycle.
> > >
> >
> > From the log, it looks like the platform has booted to userspace. Any chance
> > we can have a look at:
> > $ grep "" /sys/devices/system/clock*/{broadcast,clock*}/{available,current}_*
> 
> In the failing case, it doesn't boot to a shell, so I can't do that,
> but after I revert the patch, I have this:
>

Ah ok, does it hang when it registers clockevents ?

> / # ls -l /sys/devices/system/clocksource
> total 0
> drwxr-xr-x    3 root     root             0 Jan  1 00:00 clocksource0
> drwxr-xr-x    2 root     root             0 Jan  1 00:00 power
> -rw-r--r--    1 root     root          4096 Jan  1 00:00 uevent
> / # cat /sys/devices/system/clocksource/clocksource0/available_clocksource
> timer jiffies

Looks good.

> / # cat /sys/devices/system/clocksource/clocksource0/current_clocksource
> timer
>

OK, meson6 clocksource is active

> / # cat /sys/devices/system/clockevents/broadcast/current_device
> meson6_tick

OK, it can support broadcast

> / # cat /sys/devices/system/clockevents/clockevent0/current_device
> dummy_timer
> / # cat /sys/devices/system/clockevents/clockevent1/current_device
> dummy_timer
> / # cat /sys/devices/system/clockevents/clockevent2/current_device
> dummy_timer

But I can't understand why is dummy_timer the active event source and
not meson6_tick. And you say this is working case ? Looks suspicious.

If dummy_timer was getting used, I think meson6_tick was never utilised
before as I see this platform doesn't have cpuidle(at-least from DT)

--
Regards,
Sudeep