On Wed, 16 Apr 2014, Krzysztof Kozlowski wrote:
> On wto, 2014-04-15 at 18:20 +0200, Thomas Gleixner wrote:
> > Here is a complete solution to the problem. We really want the drivers
> > to be fast and clean and not work around such issues.
> >
> > I'm quite happy that I kept the 'force' argument of set_affinity
> > callbacks. I knew that I'd need it at some point.
> >
> > So with the flag set we can disable the online mask check and force
> > the interrupt to the proper cpu.
>
> Thanks for the solution.
>
> I tested your patch on Exynos 3250 and it is still not sufficient. After
> hotplugging CPU1 ~10 times the stall happens (set_next_event is called
> on wrong CPU).
>
> The patch 3/3 ("clocksource: exynos_mct: Fix too early ISR fire up on
> wrong CPU") is needed as the clockevents_config_and_register should be
> called a little later.
Ok. That makes sense.
> Do you have rest of patches (2/3 and 3/3) or should I resend them?
>
> And one minor nit in your patch: 'cpu' local variable in
> exynos4_mct_cpu_notify() is no longer used so it can be removed.
Right.
I'm going to create a proper patch for the interrupt side and stick
that into irq/urgent. It's trivial enough to go in right away and I'll
tag it stable as well. You can build your changes on top then.
Thanks,
tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
