[Xen-devel] Re: [PATCH]x86: Fix possible S3 suspend hangs

[Keir Fraser <keir.fraser@xxxxxxxxxxxxx>]

Not doubting this race exists, but could you briefly give an example of an
offlining scenario that would hit the hang? And explain the result: is it
that the offliner thinks the operation doesn't complete and spins forever in
__cpu_die()?

 -- Keir

On 12/03/2010 03:42, "Wei, Gang" <gang.wei@xxxxxxxxx> wrote:

> X86: Fix possible S3 suspend hangs
> 
> It is possible for cpu to become offlined before irq disabled in idle loop,
> which will cause this cpu stay in C state and can't wakeup to play dead.
> 
> Signed-off-by: Wei Gang <gang.wei@xxxxxxxxx>
> Signed-off-by: Yu Ke <ke.yu@xxxxxxxxx>
> 
> diff -r 132ac04cbdba xen/arch/x86/acpi/cpu_idle.c
> --- a/xen/arch/x86/acpi/cpu_idle.c Tue Mar 09 18:18:19 2010 +0000
> +++ b/xen/arch/x86/acpi/cpu_idle.c Fri Mar 12 11:07:43 2010 +0800
> @@ -268,7 +268,8 @@ static void acpi_processor_idle(void)
>       */
>      local_irq_disable();
>  
> -    if ( softirq_pending(smp_processor_id()) )
> +    if ( softirq_pending(smp_processor_id()) ||
> +         cpu_is_offline(smp_processor_id()) )
>      {
>          local_irq_enable();
>          sched_tick_resume();
> diff -r 132ac04cbdba xen/arch/x86/domain.c
> --- a/xen/arch/x86/domain.c Tue Mar 09 18:18:19 2010 +0000
> +++ b/xen/arch/x86/domain.c Fri Mar 12 11:12:24 2010 +0800
> @@ -82,7 +82,8 @@ static void default_idle(void)
>  static void default_idle(void)
>  {
>      local_irq_disable();
> -    if ( !softirq_pending(smp_processor_id()) )
> +    if ( !softirq_pending(smp_processor_id()) &&
> +         cpu_online(smp_processor_id()) )
>          safe_halt();
>      else
>          local_irq_enable();



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel