Using cpu_all_mask in clockevents cpumask may result in issues while
comparing multiple clockevent devices to choose the preferred one.
On one of the platforms with 2 system(i.e. non per-CPU) timers with
different ratings, having cpu_all_mask for one of the device resulted
in boot hang with forever loop in clockevents_notify_released as both
were being selected as preferred.
In order to prevent such issues in future, let's warn if any clockevent
drivers sets cpu_all_mask as it's cpumask and just override it to use
cpu_possible_mask. All the existing occurrences of cpu_all_mask are
already replaced with cpu_possible_mask.
Cc: Thomas Gleixner <t...@linutronix.de>
Signed-off-by: Sudeep Holla <sudeep.ho...@arm.com>
---
kernel/time/clockevents.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c
index 16c027e9cc73..8c0e4092f661 100644
--- a/kernel/time/clockevents.c
+++ b/kernel/time/clockevents.c
@@ -463,6 +463,12 @@ void clockevents_register_device(struct clock_event_device
*dev)
dev->cpumask = cpumask_of(smp_processor_id());
}
+ if (dev->cpumask == cpu_all_mask) {
+ WARN(1, "%s cpumask == cpu_all_mask, using cpu_possible_mask
instead\n",
+ dev->name);
+ dev->cpumask = cpu_possible_mask;
+ }
+
raw_spin_lock_irqsave(&clockevents_lock, flags);
list_add(&dev->list, &clockevent_devices);
--
2.7.4
