Re: [PATCH v3 2/3] ARM: rockchip: ensure CPU to enter WFI/WFE state

[Caesar Wang]

å 2015å06æ06æ 04:24, Doug Anderson åé:
> Russell,
>
> On Fri, Jun 5, 2015 at 11:29 AM, Russell King - ARM Linux
> <linux@xxxxxxxxxxxxxxxx> wrote:
> > 1) v7_coherency_exit() is specific to v7 CPUs and can't be used by
> >     generic code.
> Oh, I see.  So (I think) you're saying that perhaps the reason that
> Caesar needed his patch was that he needed the dying processor to
> execute v7_exit_coherency_flush(), NOT that he needed the dying
> processor to be in WFI/WFE.  That actually makes a lot more sense to
> me!  :)  Thanks a lot for pointing that out, it's very helpful.
>
>
> > So, we're actually in a very sticky position over taking CPUs offline.
> > It seems to be something that the ARM architecture and kernel
> > architecture doesn't actually allow to be done safely.  So much so,
> > that in a similar way to the original Keystone 2 physical address
> > switch, I'm tempted to make taking a CPU offline taint the kernel!
> Wow, that's going to suck.  So if you want to suspend / resume you
> need to taint your kernel.  So much for saving the planet by going
> into suspend...  ...or are you thinking that it won't taint the kernel
> when the kernel takes CPUs offline for suspend/resume purposes?  ...or
> are you thinking you've some solution that works for suspend/resume
> that doesn't work for the general cpu offlining problem?  I'd be very
> interested to hear...
>
>
> I know I'm not a maintainer, but if I were and I knew that lots of
> smart people had thought about the problem of CPU offlining and they
> didn't have a solution and I could make my platform 99.99999999%
> reliable by allowing a very safe mdelay(1) where I had a pretty strong
> guarantee that the 1ms was enough time, I would probably accept that
> code...
>
>
> So since I'm not a maintainer and I certainly couldn't ack such code,
> I would certainly be happy to add my Reviewed-by to Caesar's patch if
> he changed it mention that he needed to make sure that
> v7_exit_coherency_flush() in rockchip_cpu_die() executed in time.

OK.
The dying processor to execute v7_exit_coherency_flush(),not that the 
dying processor to be in WFI/WFE.

It's needed to enter WFI/WFE state from the ARM refer document when CPU 
down.

But......

Here is my test: (won't to enter the WFI state)
@@ -331,8 +331,8 @@ static int rockchip_cpu_kill(unsigned int cpu)
 static void rockchip_cpu_die(unsigned int cpu)
 {
        v7_exit_coherency_flush(louis);
-       while (1)
-               cpu_do_idle();
+       while (1);
+               //cpu_do_idle();
 }

echo 0 > /sys/module/printk/parameters/console_suspend
echo 1 > /sys/power/pm_print_times
echo mem > sys/power/state

You can play anything
or do some test for CPU up/down:

cd /sys/devices/system/cpu/
for i in $(seq 10000); do
  echo "================= $i ============"
  for j in $(seq 100); do
    while [[ "$(cat cpu1/online)$(cat cpu2/online)$(cat cpu3/online)" 
!= "000" ]]; do

      echo 0 > cpu1/online
      echo 0 > cpu2/online
      echo 0 > cpu3/online
    done
    while [[ "$(cat cpu1/online)$(cat cpu2/online)$(cat cpu3/online)" 
!= "111" ]]; do
      echo 1 > cpu1/online
      echo 1 > cpu2/online
      echo 1 > cpu3/online
    done
  done
done
Sometimes,the system will be restart when do the about test.
I'm no sure what's happen, That maybe abnormal won't to enter the WFI 
state.

> -Doug

--
Thanks,
- Caesar


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/