[PATCH 0/5] ARM: tegra: add CPU hotplug support

[josephl@nvidia.com (Joseph Lo)]

Hi Stephen,

Thanks for your review.

On Wed, 2012-08-15 at 02:32 +0800, Stephen Warren wrote:
> On 08/14/2012 02:07 AM, Joseph Lo wrote:
> > This patch set adds CPU hotplug support for Tegra20 and Tegra30 SoCs.
> > 
> > The first 3 patches make the CPU clock and reset control be a module and
> > the abstraction APIs for TegraXX SoCs. And clean up some codes for later
> > usage. Then add CPU hotplug support for Tegra20 and Tegra30.
> > 
> > Verified on Seaboard(Tegra20) and Cardhu(Tegra30)
> 
> There are two problems with this series:
> 
> 1) It doesn't compile if CONFIG_HOTPLUG_CPU isn't enabled. Please also
> make sure that Tegra20-only and Tegra30-only kernels compile (I didn't
> check this, so it may well already be the case).

I will add one more patch to enable HOTPLUG_CPU in this patch set.

I also verified the compiling and functionality of the Tegra20-only and
Tegra30-only kernels. It's OK.
> 
> 2) It prevents the "reboot" from working; the kernel hangs as it
> attempts to do the reboot. I've confirmed this isn't the case without
> this patch series, even if CONFIG_HOTPLUG_CPU is enabled.

I just found during the reboot process. It will call to
"platfor_cpu_kill" before assert the reset. But in this patch, the
"platform_cpu_kill" can't just be called along. It must be called after
"platform_cpu_die". So I will move all the SoC specific CPU shutdown
code into "platform_cpu_die" not separate into two functions. It will
prevent the system hang when calling to "smp_send_stop". (the flow for
your reference,
sys_reboot->machine_restart->smp_send_stop->smp_kill_cpus->platform_cpu_kill.)
> 
> I assume we'll want to enable CONFIG_HOTPLUG_CPU in tegra_defconfig once
> this is checked in?

Thanks,
Joseph

--
nvpublic