Hi Eric,
On Mon, Jan 07, 2013 at 01:53:30AM +0000, Eric W. Biederman wrote:
> Russell King - ARM Linux <linux at arm.linux.org.uk> writes:
> > The problem is... on ARM we're yet again shot in the foot through the
> > unwillingness to architect certain aspects of the system: there is no
> > architecturally known way to place CPUs into a reset state once they've
> > been started up. In other words, once a CPU starts executing kernel
> > code, it needs to always execute kernel code or there must be some kind
> > of platform specific hook to disable it.
> >
> > Maybe that's the answer here: have machine_shutdown() call out to some
> > platform specific function to do the forced-takedown of secondary CPUs,
> > and if there's no such specific function, then we use our present CPU
> > stopping method of making them spin in a WFI loop with IRQs disabled.
>
> Yes. On x86 we have had the generic equivalent of disable_nonboot_cpus
> in the machine_shutdown for a long time.
Ok, we can do that that on ARM too, but we currently call machine_shutdown
from machine_{power_off, halt, restart} so we just need to fix that and we
will be doing functionally the same thing as x86.
> Please note the only code path that generically calls machine_shutdown()
> is kernel_kexec() so if you wish you can avoid duplication by
> refactoring you architecture specific code.
Exactly, that sounds like the right way forward to me. Thanks.
Will
