Am 19.03.2014 22:06, schrieb Sebastian Capella:
On 19 March 2014 13:47, Sebastian Capella sebastian.capella@linaro.org wrote:
diff --git a/kernel/power/hibernate.c b/kernel/power/hibernate.c index a5f702a..d96b910 100644 --- a/kernel/power/hibernate.c +++ b/kernel/power/hibernate.c @@ -594,7 +594,8 @@ static void power_down(void) case HIBERNATION_PLATFORM: hibernation_platform_enter(); case HIBERNATION_SHUTDOWN:
kernel_power_off();
if (pm_power_off)kernel_power_off(); break;This follows the behavior in the reboot syscall which does it this way also. I'm testing this now, and it seems work fine. If this looks good, I can add it as an additional patch.
BTW, one thing I would point out is that kernel_power_off and kernel_halt call the same notifier but with different parameters (SYS_POWER_OFF and SYS_HALT).
If pm_power_down is null, I dont see why we'd want to notify SYS_POWER_OFF before SYS_HALT. With the previous change I'm assuming there's no benefit, so please chime in if you know a reason.
Both states, power off and sys halt, do sound pretty final and I would assume something is broken, if power off is called before sys halt or vice versa. At least I would never expect that the reboot/poweroff/syshalt notifier may be called twice (and thats why the heartbeat-trigger may crash).
But just in case, changing that behaviour in ledtrig-heartbeat.c would be pretty easy, just remove the heartbeat_reboot_notifier (which plays nice and deregisters the trigger on reboot) and use the panic_notifier (which doesn't unregister the trigger but just turns off the led) for reboot too. Another solution would be to unregister the reboot_notifier in the reboot_nofifier itself. I've just seen one watchdog driver (drivers/rtc/rtc-m41t80.c) which does that. But I still think such shouldn't be necessary (and I haven't had a look at other reboot_notifier users).
Regards,
Alexander Holler