x86/hpet: Prevent might sleep splat on resume
authorThomas Gleixner <tglx@linutronix.de>
Wed, 1 Mar 2017 20:10:17 +0000 (21:10 +0100)
committerThomas Gleixner <tglx@linutronix.de>
Thu, 2 Mar 2017 08:33:47 +0000 (09:33 +0100)
Sergey reported a might sleep warning triggered from the hpet resume
path. It's caused by the call to disable_irq() from interrupt disabled
context.

The problem with the low level resume code is that it is not accounted as a
special system_state like we do during the boot process. Calling the same
code during system boot would not trigger the warning. That's inconsistent
at best.

In this particular case it's trivial to replace the disable_irq() with
disable_hardirq() because this particular code path is solely used from
system resume and the involved hpet interrupts can never be force threaded.

Reported-and-tested-by: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1703012108460.3684@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
arch/x86/kernel/hpet.c

index dc6ba5bda9fc83630c773a80c4adea6871db0a59..89ff7af2de508ba0c34c2ce24c227dea5bf6d973 100644 (file)
@@ -354,7 +354,7 @@ static int hpet_resume(struct clock_event_device *evt, int timer)
 
                irq_domain_deactivate_irq(irq_get_irq_data(hdev->irq));
                irq_domain_activate_irq(irq_get_irq_data(hdev->irq));
-               disable_irq(hdev->irq);
+               disable_hardirq(hdev->irq);
                irq_set_affinity(hdev->irq, cpumask_of(hdev->cpu));
                enable_irq(hdev->irq);
        }