Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1411363 > unrolled thread

[PATCH 1/2] rtc-cmos: Clear expired alarm after resume

Back to article view | Back to linux.kernel

Contents

  [PATCH 1/2] rtc-cmos: Clear expired alarm after resume Gabriele Mazzotta <gabriele.mzt@gmail.com> - 2016-06-01 17:50 +0200
    Re: [PATCH 1/2] rtc-cmos: Clear expired alarm after resume Gabriele Mazzotta <gabriele.mzt@gmail.com> - 2016-06-01 18:20 +0200
    Re: [PATCH 1/2] rtc-cmos: Clear expired alarm after resume Alexandre Belloni <alexandre.belloni@free-electrons.com> - 2016-06-04 16:50 +0200
      Re: [PATCH 1/2] rtc-cmos: Clear expired alarm after resume Gabriele Mazzotta <gabriele.mzt@gmail.com> - 2016-06-04 19:00 +0200
        Re: [PATCH 1/2] rtc-cmos: Clear expired alarm after resume Alexandre Belloni <alexandre.belloni@free-electrons.com> - 2016-06-04 19:50 +0200

#1411363 — [PATCH 1/2] rtc-cmos: Clear expired alarm after resume

If the system wakes up because of a wake alarm, the internal state
of the alarm is not updated. As consequence, the state no longer
reflects the actual state of the hardware and setting a new alarm
is not possible until the expired alarm is cleared.

Signed-off-by: Gabriele Mazzotta <gabriele.mzt@gmail.com>
---
 drivers/rtc/rtc-cmos.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/drivers/rtc/rtc-cmos.c b/drivers/rtc/rtc-cmos.c
index fbe9c72..fd121e3 100644
--- a/drivers/rtc/rtc-cmos.c
+++ b/drivers/rtc/rtc-cmos.c
@@ -900,11 +900,33 @@ static inline int cmos_poweroff(struct device *dev)
 
 #ifdef	CONFIG_PM_SLEEP
 
+static void cmos_check_alarm(struct device *dev)
+{
+	struct cmos_rtc *cmos = dev_get_drvdata(dev);
+	struct rtc_wkalrm alarm;
+	struct rtc_time now;
+	time64_t t_now;
+	time64_t t_expires;
+
+	cmos_read_time(dev, &now);
+	rtc_read_alarm(cmos->rtc, &alarm);
+	t_now = rtc_tm_to_time64(&now);
+	t_expires = rtc_tm_to_time64(&alarm.time);
+
+	if (t_expires <= t_now && alarm.enabled) {
+		alarm.enabled = 0;
+		cmos->suspend_ctrl &= ~RTC_AIE;
+		rtc_set_alarm(cmos->rtc, &alarm);
+	}
+}
+
 static int cmos_resume(struct device *dev)
 {
 	struct cmos_rtc	*cmos = dev_get_drvdata(dev);
 	unsigned char tmp;
 
+	cmos_check_alarm(dev);
+
 	if (cmos->enabled_wake) {
 		if (cmos->wake_off)
 			cmos->wake_off(dev);
-- 
2.8.1

[toc] | [next] | [standalone]

#1411378

On 01/06/2016 17:40, Gabriele Mazzotta wrote:
> If the system wakes up because of a wake alarm, the internal state
> of the alarm is not updated. As consequence, the state no longer
> reflects the actual state of the hardware and setting a new alarm
> is not possible until the expired alarm is cleared.
> 
> Signed-off-by: Gabriele Mazzotta <gabriele.mzt@gmail.com>
> ---
>  drivers/rtc/rtc-cmos.c | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)
> 
> diff --git a/drivers/rtc/rtc-cmos.c b/drivers/rtc/rtc-cmos.c
> index fbe9c72..fd121e3 100644
> --- a/drivers/rtc/rtc-cmos.c
> +++ b/drivers/rtc/rtc-cmos.c
> @@ -900,11 +900,33 @@ static inline int cmos_poweroff(struct device *dev)
>  
>  #ifdef	CONFIG_PM_SLEEP
>  
> +static void cmos_check_alarm(struct device *dev)
> +{
> +	struct cmos_rtc *cmos = dev_get_drvdata(dev);
> +	struct rtc_wkalrm alarm;
> +	struct rtc_time now;
> +	time64_t t_now;
> +	time64_t t_expires;
> +
> +	cmos_read_time(dev, &now);
> +	rtc_read_alarm(cmos->rtc, &alarm);

Here it should probably check the return value and exit in case of error.

> +	t_now = rtc_tm_to_time64(&now);
> +	t_expires = rtc_tm_to_time64(&alarm.time);
> +
> +	if (t_expires <= t_now && alarm.enabled) {
> +		alarm.enabled = 0;
> +		cmos->suspend_ctrl &= ~RTC_AIE;
> +		rtc_set_alarm(cmos->rtc, &alarm);

Same here.

> +	}
> +}
> +
>  static int cmos_resume(struct device *dev)
>  {
>  	struct cmos_rtc	*cmos = dev_get_drvdata(dev);
>  	unsigned char tmp;
>  
> +	cmos_check_alarm(dev);
> +
>  	if (cmos->enabled_wake) {
>  		if (cmos->wake_off)
>  			cmos->wake_off(dev);
> 

[toc] | [prev] | [next] | [standalone]

#1413769

Hi,

On 01/06/2016 at 17:40:14 +0200, Gabriele Mazzotta wrote :
> If the system wakes up because of a wake alarm, the internal state
> of the alarm is not updated. As consequence, the state no longer
> reflects the actual state of the hardware and setting a new alarm
> is not possible until the expired alarm is cleared.
> 

I'm not completely sure to understand what is happening but could you
check whether that one is solved by
2b2f5ff00f63847d95adad6289bd8b05f5983dd5 in my tree (rtc-next).


-- 
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

[toc] | [prev] | [next] | [standalone]

#1413797

2016-06-04 16:46 GMT+02:00 Alexandre Belloni
<alexandre.belloni@free-electrons.com>:
> Hi,
>
> On 01/06/2016 at 17:40:14 +0200, Gabriele Mazzotta wrote :
>> If the system wakes up because of a wake alarm, the internal state
>> of the alarm is not updated. As consequence, the state no longer
>> reflects the actual state of the hardware and setting a new alarm
>> is not possible until the expired alarm is cleared.
>>
>
> I'm not completely sure to understand what is happening but could you
> check whether that one is solved by
> 2b2f5ff00f63847d95adad6289bd8b05f5983dd5 in my tree (rtc-next).
>

I picked 2b2f5ff00f63847d95adad6289bd8b05f5983dd5 and applied it
on top of 4.7-rc1, but that didn't fix the problem.

Let me explain the problem by showing you how I reproduce it:

root@localhost:~# cat /proc/driver/rtc | grep alarm_IRQ
alarm_IRQ       : no
root@localhost:~# echo +10 > /sys/class/rtc/rtc0/wakealarm
root@localhost:~# echo mem > /sys/power/state  # wait for auto-resume
root@localhost:~# echo +10 > /sys/class/rtc/rtc0/wakealarm
bash: echo: write error: Device or resource busy
root@localhost:~# cat /proc/driver/rtc | grep alarm_IRQ
alarm_IRQ       : yes

To set another alarm, I have to first write 0 to wakealarm. After that
I can set what I want. This doesn't happen if the alarm fires while
the system is awake, it happens only if the system is suspended and
the alarm wakes it.

I actually forgot to say that maybe this problem is not limited
to rtc-cmos and that maybe a general solution could be used.

I've just looked better into what is causing this and the problem
seems to be caused by the fact that rtc_timer_do_work() is not
executed if the timer expires while the system is suspended.

The following solves this particular problem:

diff --git a/drivers/rtc/class.c b/drivers/rtc/class.c
index 74fd974..80d6a12 100644
--- a/drivers/rtc/class.c
+++ b/drivers/rtc/class.c
@@ -102,6 +102,9 @@ static int rtc_resume(struct device *dev)
     struct timespec64    sleep_time;
     int err;

+    /* A timer might have expired while suspended */
+    schedule_work(&rtc->irqwork);
+
     if (timekeeping_rtc_skipresume())
         return 0;


> --
> Alexandre Belloni, Free Electrons
> Embedded Linux, Kernel and Android engineering
> http://free-electrons.com

[toc] | [prev] | [next] | [standalone]

#1413829

On 04/06/2016 at 18:58:59 +0200, Gabriele Mazzotta wrote :
> 2016-06-04 16:46 GMT+02:00 Alexandre Belloni
> <alexandre.belloni@free-electrons.com>:
> > On 01/06/2016 at 17:40:14 +0200, Gabriele Mazzotta wrote :
> >> If the system wakes up because of a wake alarm, the internal state
> >> of the alarm is not updated. As consequence, the state no longer
> >> reflects the actual state of the hardware and setting a new alarm
> >> is not possible until the expired alarm is cleared.
> >>
> >
> > I'm not completely sure to understand what is happening but could you
> > check whether that one is solved by
> > 2b2f5ff00f63847d95adad6289bd8b05f5983dd5 in my tree (rtc-next).
> >
> 
> I picked 2b2f5ff00f63847d95adad6289bd8b05f5983dd5 and applied it
> on top of 4.7-rc1, but that didn't fix the problem.
> 
> Let me explain the problem by showing you how I reproduce it:
> 
> root@localhost:~# cat /proc/driver/rtc | grep alarm_IRQ
> alarm_IRQ       : no
> root@localhost:~# echo +10 > /sys/class/rtc/rtc0/wakealarm
> root@localhost:~# echo mem > /sys/power/state  # wait for auto-resume
> root@localhost:~# echo +10 > /sys/class/rtc/rtc0/wakealarm
> bash: echo: write error: Device or resource busy
> root@localhost:~# cat /proc/driver/rtc | grep alarm_IRQ
> alarm_IRQ       : yes
> 
> To set another alarm, I have to first write 0 to wakealarm. After that
> I can set what I want. This doesn't happen if the alarm fires while
> the system is awake, it happens only if the system is suspended and
> the alarm wakes it.
> 
> I actually forgot to say that maybe this problem is not limited
> to rtc-cmos and that maybe a general solution could be used.
> 
> I've just looked better into what is causing this and the problem
> seems to be caused by the fact that rtc_timer_do_work() is not
> executed if the timer expires while the system is suspended.
> 

I doubt this is a driver independent issues as I don't get that
behaviour on many other RTCs. I'll have a look.

-- 
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web