Patchwork rtc: Fix unexpected poweron and unnecessary alarm disabling

login
register
mail settings
Submitter Jérémy Compostella
Date Jan. 4, 2012, 8:15 p.m.
Message ID <CA+TATJHXo2xV=hFjfa3-S759hkse_5YxWhdOBVwXbf2wCikpoQ@mail.gmail.com>
Download mbox | patch
Permalink /patch/134354/
State New
Headers show

Comments

Jérémy Compostella - Jan. 4, 2012, 8:15 p.m.
2012/1/4 Jérémy Compostella <jeremy.compostella@gmail.com>

> 2012/1/4 John Stultz <john.stultz@linaro.org>
>
>> On Wed, 2012-01-04 at 01:59 +0100, Linus Walleij wrote:
>> > 2011/12/28 Jérémy Compostella <jeremy.compostella@gmail.com>:
>> >
>> > > On my laptop with 3.2.0-rc7 I have unexpected poweron after shutdown
>> > > 300 seconds later. I found out that the "hwclock --systohw" plays with
>> > > UIE which lead to the rtc_disable_rtc() function call. This function
>> > > disables the alarm by setting the alarm 300 seconds later with the
>> > > disable flag which does not seems to be the best way since we could
>> > > use the alarm_irq_enable handler.
>> > >
>> > > I found out that the real problem is not really related to a bad RTC
>> > > management but to a bad ACPI compliance. Actually, the platform should
>> > > not wake up when the PM1_ENABLE RTC_EN ACPI flag isn't set. But it
>> does
>> > > on some machine like mine if there is a RTC CMOS alarm set in future
>> even
>> > > with the AIE disabled.
>> > >
>> > > Moreover, the rtc_timer_do_work calls the rtc_alarm_disable even if
>> > > there is no alarm to disable. Therefore, during the boot time, a
>> > > disabled alarm is set 300 seconds in the future which does not make
>> > > sense.
>> >
>> > The patch looks very similar to a patch submitted by Rabin.
>>
> I took a look at the Rabin patch. I do agree it's very similar except that
> my
> patch avoid useless alarm disabling too.
>
> FYI, I successfully passed the rtc_test program with this patch applied.
>
>  > I would prefer if John could say something about the proper solution...
>>
>> For 3.2 I've requested the problematic patch be reverted. We can try
>> again for a proper fix for 3.3.
>>
>> thanks
>> -john
>>
>
I took a look at the whole RTC drivers and some of them does not
implement the alarm_irq_enable callback and I'm not sure it makes sense
and it's realistic to implement it for these one.

So, I propose a new patch which disables the appropriate alarm instead of
the arbitrary 300 seconds in the future alarm. This patch is compatible
with the removed problematic patch and passed with success the rtctest
test suite.

Please review it,

Best regards,

Jeremy Compostella

Patch

From fa83fa95ffb2279f2af80316583b18b784537245 Mon Sep 17 00:00:00 2001
From: Jeremy Compostella <jeremy.compostella@gmail.com>
Date: Wed, 4 Jan 2012 21:03:38 +0100
Subject: [PATCH] rtc: Fix unexpected poweron and unnecessary alarm disabling

On my laptop 3.2.0-rc6 I have unexpected poweron after shutdown
300 seconds later. I found out that the "hwclock --systohw" plays with
UIE which lead to the rtc_disable_rtc() function call. This function
disables the alarm by setting the alarm 300 seconds later with the
disable flag which does not seems to be the best way since we could
disable the appropriate alarm instead.

Moreover, the rtc_timer_do_work calls the rtc_alarm_disable event if
there is no alarm to disable. Therefore, during the boot time, a
disabled alarm is set 300 seconds in the future which does not make
sense.

Signed-off-by: Jeremy Compostella <jeremy.compostella@gmail.com>
---
 drivers/rtc/interface.c |   16 ++++++----------
 1 files changed, 6 insertions(+), 10 deletions(-)

diff --git a/drivers/rtc/interface.c b/drivers/rtc/interface.c
index 3bcc7cf..7241b3f 100644
--- a/drivers/rtc/interface.c
+++ b/drivers/rtc/interface.c
@@ -776,15 +776,11 @@  static int rtc_timer_enqueue(struct rtc_device *rtc, struct rtc_timer *timer)
 	return 0;
 }
 
-static void rtc_alarm_disable(struct rtc_device *rtc)
+static void rtc_alarm_disable(struct rtc_device *rtc, ktime_t time)
 {
 	struct rtc_wkalrm alarm;
-	struct rtc_time tm;
-
-	__rtc_read_time(rtc, &tm);
 
-	alarm.time = rtc_ktime_to_tm(ktime_add(rtc_tm_to_ktime(tm),
-				     ktime_set(300, 0)));
+	alarm.time = rtc_ktime_to_tm(time);
 	alarm.enabled = 0;
 
 	___rtc_set_alarm(rtc, &alarm);
@@ -812,7 +808,7 @@  static void rtc_timer_remove(struct rtc_device *rtc, struct rtc_timer *timer)
 		int err;
 		next = timerqueue_getnext(&rtc->timerqueue);
 		if (!next) {
-			rtc_alarm_disable(rtc);
+			rtc_alarm_disable(rtc, timer->node.expires);
 			return;
 		}
 		alarm.time = rtc_ktime_to_tm(next->expires);
@@ -835,7 +831,7 @@  static void rtc_timer_remove(struct rtc_device *rtc, struct rtc_timer *timer)
  */
 void rtc_timer_do_work(struct work_struct *work)
 {
-	struct rtc_timer *timer;
+	struct rtc_timer *timer = NULL;
 	struct timerqueue_node *next;
 	ktime_t now;
 	struct rtc_time tm;
@@ -876,8 +872,8 @@  again:
 		err = __rtc_set_alarm(rtc, &alarm);
 		if (err == -ETIME)
 			goto again;
-	} else
-		rtc_alarm_disable(rtc);
+	} else if (timer && timer->enabled == 0)
+		rtc_alarm_disable(rtc, timer->node.expires);
 
 	mutex_unlock(&rtc->ops_lock);
 }
-- 
1.7.2.5