[PATCH] ARM: decouple CPU offlining from reboot/shutdown

[Will Deacon]

On Tue, Jun 11, 2013 at 07:41:06PM +0100, Stephen Warren wrote:
> On 06/11/2013 11:23 AM, Will Deacon wrote:
> > Hi Stephen,
> > 
> > On Mon, Jun 10, 2013 at 07:12:41PM +0100, Stephen Warren wrote:
> >> From: Stephen Warren <swarren at nvidia.com>
> >>
> >> machine_shutdown() is a hook for kexec. Add a comment saying so, since
> >> it isn't obvious from the function name.
> 
> >> diff --git a/arch/arm/kernel/process.c b/arch/arm/kernel/process.c
> >> index 282de48..dbe1692 100644
> >> --- a/arch/arm/kernel/process.c
> >> +++ b/arch/arm/kernel/process.c
> >> @@ -97,13 +97,14 @@ void soft_restart(unsigned long addr)
> >>  {
> >>  	u64 *stack = soft_restart_stack + ARRAY_SIZE(soft_restart_stack);
> >>  
> >> +	BUG_ON(num_online_cpus() > 1);
> > 
> > I think this is overkill, and we could actually scream and try to return an
> > error here (we've not yet switched stack, and the upper layers of kexec look
> > like they can do something with an error code).
> 
> Oh, that's the BUG_ON I added to soft_restart() not the one I added into
> machine_shutdown().
> 
> I added this one to make sure nobody was using soft_restart() on an SMP
> machine; Russell had asked to validate that all SMP systems provided a
> HW restart implementation.
> 
> I assume your comments re: aborting the kexec by returning an error code
> apply more to machine_shutdown() which happens a bunch earlier.

No, I did mean soft_restart! For kexec, this hangs off machine_kexec, which
seems to propogate errors correctly. For invocations via the machine
descriptor (i.e. arm_pm_restart), we'll print out `reboot failed' and go
into a while(1) loop.

Will