On Thu, Dec 20, 2012 at 08:15:30PM +0000, Stephen Warren wrote:

On 12/20/2012 10:59 AM, Stephen Warren wrote:

quoted

On 12/20/2012 10:36 AM, Will Deacon wrote:

quoted

I think you're better off using what we currently have and hanging your code
off platform_cpu_kill.

OK, I'll look into that. Joseph Lo just posted patches to implement
cpu_kill() on Tegra, which was needed to fix some issues in our hotplug
code anyway. Perhaps that will remove the need for any other changes...

Will,

Hi Stephen,

I just remembered one other advantage of disable_nonboot_cpus(); it
always makes the kexec happen on the boot CPU. Without this, I believe
it's random whether CPU0 or CPU1 performs the kexec. I suspect it's most
likely to work if we can always kexec on the boot CPU rather than a
random CPU?

It will be the same CPU as the one which made the kexec system call. If you
need to invoke this from a specific CPU, you currently have to do this in
userspace using something like the `taskset' utility. Linux can actually
boot on any physical CPU for ARM now (thanks to the logical map) but most
platforms have restrictions related to secure firmware.

Will