[PATCH] ACPI/hotplug: Fix STARTING/DYING action in acpi_cpu_soft_notify()

[PATCH] ACPI/hotplug: Fix STARTING/DYING action in acpi_cpu_soft_notify()

[Toshi Kani]

During CPU online/offline testing on a large system, one of the
processors got stuck after the message "bad: scheduling from the
idle thread!".  The problem is that acpi_cpu_soft_notify() calls
acpi_bus_get_device() for all action types.  CPU_STARTING and
CPU_DYING do not allow the notify handlers to sleep.  However,
acpi_bus_get_device() can sleep in acpi_ut_acquire_mutex().

Change acpi_cpu_soft_notify() to return immediately for CPU_STARTING
and CPU_DYING as they have no action in this handler.

Signed-off-by: Toshi Kani <toshi.kani@hp.com>
---
 drivers/acpi/processor_driver.c |    7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/acpi/processor_driver.c b/drivers/acpi/processor_driver.c
index 7f70f31..4fcbd67 100644
--- a/drivers/acpi/processor_driver.c
+++ b/drivers/acpi/processor_driver.c
@@ -121,6 +121,13 @@ static int acpi_cpu_soft_notify(struct notifier_block *nfb,
 	struct acpi_processor *pr = per_cpu(processors, cpu);
 	struct acpi_device *device;
 
+	/*
+	 * CPU_STARTING and CPU_DYING must not sleep. Return here since
+	 * acpi_bus_get_device() may sleep.
+	 */
+	if (action == CPU_STARTING || action == CPU_DYING)
+		return NOTIFY_DONE;
+
 	if (!pr || acpi_bus_get_device(pr->handle, &device))
 		return NOTIFY_DONE;
 

Re: [PATCH] ACPI/hotplug: Fix STARTING/DYING action in acpi_cpu_soft_notify()

[Rafael J. Wysocki]

On Thursday, May 08, 2014 07:58:59 AM Toshi Kani wrote:
> During CPU online/offline testing on a large system, one of the
> processors got stuck after the message "bad: scheduling from the
> idle thread!".  The problem is that acpi_cpu_soft_notify() calls
> acpi_bus_get_device() for all action types.  CPU_STARTING and
> CPU_DYING do not allow the notify handlers to sleep.  However,
> acpi_bus_get_device() can sleep in acpi_ut_acquire_mutex().
> 
> Change acpi_cpu_soft_notify() to return immediately for CPU_STARTING
> and CPU_DYING as they have no action in this handler.
> 
> Signed-off-by: Toshi Kani <toshi.kani@hp.com>

Queued up for 3.16, thanks!

> ---
>  drivers/acpi/processor_driver.c |    7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/drivers/acpi/processor_driver.c b/drivers/acpi/processor_driver.c
> index 7f70f31..4fcbd67 100644
> --- a/drivers/acpi/processor_driver.c
> +++ b/drivers/acpi/processor_driver.c
> @@ -121,6 +121,13 @@ static int acpi_cpu_soft_notify(struct notifier_block *nfb,
>  	struct acpi_processor *pr = per_cpu(processors, cpu);
>  	struct acpi_device *device;
>  
> +	/*
> +	 * CPU_STARTING and CPU_DYING must not sleep. Return here since
> +	 * acpi_bus_get_device() may sleep.
> +	 */
> +	if (action == CPU_STARTING || action == CPU_DYING)
> +		return NOTIFY_DONE;
> +
>  	if (!pr || acpi_bus_get_device(pr->handle, &device))
>  		return NOTIFY_DONE;
>  
> --
> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

Re: [PATCH] ACPI/hotplug: Fix STARTING/DYING action in acpi_cpu_soft_notify()

[Toshi Kani]

On Tue, 2014-05-20 at 00:55 +0200, Rafael J. Wysocki wrote:
> On Thursday, May 08, 2014 07:58:59 AM Toshi Kani wrote:
> > During CPU online/offline testing on a large system, one of the
> > processors got stuck after the message "bad: scheduling from the
> > idle thread!".  The problem is that acpi_cpu_soft_notify() calls
> > acpi_bus_get_device() for all action types.  CPU_STARTING and
> > CPU_DYING do not allow the notify handlers to sleep.  However,
> > acpi_bus_get_device() can sleep in acpi_ut_acquire_mutex().
> > 
> > Change acpi_cpu_soft_notify() to return immediately for CPU_STARTING
> > and CPU_DYING as they have no action in this handler.
> > 
> > Signed-off-by: Toshi Kani <toshi.kani@hp.com>
> 
> Queued up for 3.16, thanks!

Great! Thanks Rafael!
-Toshi