Re: [PATCH] x86: call machine_shutdown and stop all CPUs in native_machine_halt

[Ingo Molnar <mingo@elte.hu>]

* Ivan Vecera <ivecera@redhat.com> wrote:

> Function machine_halt (resp. native_machine_halt) is empty for x86 
> architectures. When command 'halt -f' is invoked, the message "System 
> halted." is displayed but this is not really true because all CPUs are 
> still running. There are also similar inconsistencies for other arches 
> (some uses power-off for halt or forever-loop with IRQs 
> enabled/disabled). IMO there should be used the same approach for all 
> architectures OR what does the message "System halted" really mean?

no fundamental objections, but could you please do it a bit cleaner:

>  static void native_machine_halt(void)
>  {
> +	/* stop other cpus and apics */
> +	machine_shutdown();
> +
> +	/* stop this cpu */
> +	local_irq_disable();
> +	if (hlt_works(smp_processor_id()))
> +		for (;;) halt();
> +	for (;;);
>  }

the code in arch/x86/kernel/smp.c::stop_this_cpu() is very similar to 
this and could be shared. You could move the stop_this_cpu() function to 
arch/x86/kernel/process.c (out of smp.c), so that it can be used by 
reboot.c.

furthermore, native_machine_power_off() should probably fall back to 
native_machine_halt() as well - should pm_power_off() be disabled (or if 
it fails to stop the machine).

hm?

	Ingo