ACPI / processor: Fix STARTING/DYING action in acpi_cpu_soft_notify()
authorToshi Kani <toshi.kani@hp.com>
Thu, 8 May 2014 13:58:59 +0000 (07:58 -0600)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Fri, 16 May 2014 14:36:03 +0000 (16:36 +0200)
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>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/acpi/processor_driver.c

index 7f70f31..4fcbd67 100644 (file)
@@ -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;