Re: [stable] [patch 09/12] Fix SMP poweroff hangs

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Wednesday, 10 October 2007 12:06, Milton Miller wrote:
> 
> On Tue, Oct 9 2007 Thomas Gleixner wrote:
> >On Tue, 9 Oct 2007, Linus Torvalds wrote:
> >>On Wed, 10 Oct 2007, Thomas Gleixner wrote:
> >>>
> >>> Wrapping it into a #ifdef CONFIG_X86 would be sufficient.
> >>
> >> Well, the ppc oops seems to be a ppc bug regardless.
> >>
> >> If CPU_HOTPLUG isn't defined, the thing does nothing. And if it is
> >> defined, I don't see why/how ppc can validly oops. So I think the first
> >> thing to do is to try to figure out why it oopses, not to disable it for
> >> ppc.
> >
> >Fair enough. OTOH for the affected PPC users it's a regression and that's
> >what I'm concerned of.
> >
> >tglx
> 
> I agree this exposes a ppc arch bug (and the reason has been diagnosed),
> but might not other archs have similar problems?  The patch ties
> power-off code to suspend and hibernate cpu hotplug code.
> 
> Why not put this ACPI requirement in the ACPI code?   There is already
> a hook called at the same place this function was called that is only
> used for power off.

IMO this is a good idea.

> Also, the code is compiled for CONFIG_PM_SLEEP_SMP, which is not
> CONFIG_HOTPLUG_CPU, it requires the selection of either CONFIG_HIBERNATION
> or CONFIG_SUSPEND.  I confirmed that both of these can be disabled
> on i386 without disabling acpi.  In such a kernel the existing patch
> is insufficient.

Well, it's a known problem with this patch.

However, to resolve it we'd need to make ACPI select the CPU hotplug on SMP,
just like PM_SLEEP.  IOW, we should have a rule that if
(ACPI || PM_SLEEP) && SMP, then the CPU hotplug is selected.

Greetings,
Rafael


>  drivers/acpi/sleep/main.c |    2 ++
>  kernel/sys.c              |    1 -
>  2 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/acpi/sleep/main.c b/drivers/acpi/sleep/main.c
> index 2cbb9aa..79589bd 100644
> --- a/drivers/acpi/sleep/main.c
> +++ b/drivers/acpi/sleep/main.c
> @@ -15,6 +15,7 @@
>  #include <linux/dmi.h>
>  #include <linux/device.h>
>  #include <linux/suspend.h>
> +#include <linux/cpu.h>
>  
>  #include <asm/io.h>
>  
> @@ -382,6 +383,7 @@ static void acpi_power_off_prepare(void)
>  {
>  	/* Prepare to power off the system */
>  	acpi_sleep_prepare(ACPI_STATE_S5);
> +	disable_nonboot_cpus();
>  }
>  
>  static void acpi_power_off(void)
> diff --git a/kernel/sys.c b/kernel/sys.c
> index 8ae2e63..138f5c8 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -879,7 +879,6 @@ void kernel_power_off(void)
>  	kernel_shutdown_prepare(SYSTEM_POWER_OFF);
>  	if (pm_power_off_prepare)
>  		pm_power_off_prepare();
> -	disable_nonboot_cpus();
>  	sysdev_shutdown();
>  	printk(KERN_EMERG "Power down.\n");
>  	machine_power_off();
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 
> 

-- 
"Premature optimization is the root of all evil." - Donald Knuth