[PATCH] ACPI / cpuidle: fix deadlock between cpuidle_lock and cpu_hotplug.lock

From: Jiri Kosina
Date: Wed Sep 03 2014 - 09:04:36 EST


There is a following AB-BA dependency between cpu_hotplug.lock and
cpuidle_lock:

1) cpu_hotplug.lock -> cpuidle_lock
enable_nonboot_cpus()
_cpu_up()
cpu_hotplug_being()
LOCK(cpu_hotplug.lock)
cpu_notify()
...
acpi_processor_hotplug()
cpuidle_pause_and_lock()
LOCK(cpuidle_lock)

2) cpuidle_lock -> cpu_hotplug.lock
acpi_os_execute_deferred() workqueue
...
acpi_processor_cst_has_changed()
cpuidle_pause_and_lock()
LOCK(cpuidle_lock)
get_online_cpus()
LOCK(cpu_hotplug.lock)

Fix this by reversing the order acpi_processor_cst_has_changed() does
thigs -- let it first execute the protection against CPU hotplug by
calling get_online_cpus() and obtain the cpuidle lock only after that (and
perform the symmentric change when allowing CPUs hotplug again and
dropping cpuidle lock).

Spotted by lockdep.

Signed-off-by: Jiri Kosina <jkosina@xxxxxxx>
---
drivers/acpi/processor_idle.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index 3dca36d..17f9ec5 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -1071,9 +1071,9 @@ int acpi_processor_cst_has_changed(struct acpi_processor *pr)

if (pr->id == 0 && cpuidle_get_driver() == &acpi_idle_driver) {

- cpuidle_pause_and_lock();
/* Protect against cpu-hotplug */
get_online_cpus();
+ cpuidle_pause_and_lock();

/* Disable all cpuidle devices */
for_each_online_cpu(cpu) {
@@ -1100,8 +1100,8 @@ int acpi_processor_cst_has_changed(struct acpi_processor *pr)
cpuidle_enable_device(dev);
}
}
- put_online_cpus();
cpuidle_resume_and_unlock();
+ put_online_cpus();
}

return 0;
--
Jiri Kosina
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/