Re: [PATCH v4 1/4] rtc: omap: Cut down the shutdown time from 2 seconds to 1 sec

[Johan Hovold <johan@kernel.org>]

On Fri, Jul 20, 2018 at 04:03:20PM +0530, Keerthy wrote:
> 
> 
> On Thursday 19 July 2018 06:16 PM, Keerthy wrote:
> > 
> > 
> > On Thursday 19 July 2018 06:06 PM, Johan Hovold wrote:
> >> On Thu, Jul 19, 2018 at 05:52:17PM +0530, Keerthy wrote:
> >>> On Thursday 19 July 2018 05:23 PM, Keerthy wrote:
> >>>> On Thursday 19 July 2018 03:32 PM, Johan Hovold wrote:
> >>>>> On Thu, Jul 12, 2018 at 10:37:37AM +0530, Keerthy wrote:
> >>
> >>>>>> @@ -470,6 +476,9 @@ static void omap_rtc_power_off(void)
> >>>>>>  	val = rtc_read(rtc, OMAP_RTC_INTERRUPTS_REG);
> >>>>>>  	rtc_writel(rtc, OMAP_RTC_INTERRUPTS_REG,
> >>>>>>  			val | OMAP_RTC_INTERRUPTS_IT_ALARM2);
> >>
> >>>>>> +	/* Our calculations started right before the rollover, try again */
> >>
> >>>>>> +	if (seconds != rtc_read(omap_rtc_power_off_rtc, OMAP_RTC_SECONDS_REG))
> >>>>>> +		goto again;
> >>>>>
> >>>>> Here the alarm may have gone off as part of the roll over, in which case
> >>>>> you shouldn't retry.
> >>>>
> >>>> Ex: We programmed at Sec = 2 and we expect ALARM2 to fire at sec = 3.
> >>>>
> >>>> In the event of Roll over before setting the
> >>>> OMAP_RTC_INTERRUPTS_IT_ALARM2 bit in the OMAP_RTC_INTERRUPTS_REG will we
> >>>> not miss the ALARM2 event? Then poweroff would fail right?
> >>
> >> Right, that would fail.
> >>
> >>>> Hence the attempt to retry the next second. This sequence would begin
> >>>> right at the beginning of a new second and we expect the full sequence
> >>>> to get over without having to retry again.
> >>>>
> >>>> Hope i am clear.
> >>
> >> Yes, sure, but my point is that could end up retrying also after the
> >> alarm has fired correctly (e.g. due to latencies in turning of the
> >> power)>
> >> It may be enough to check OMAP_RTC_STATUS_REG before retrying.
> 
> On a second thought. Status gets set only after the next second.
> 
> if ALARM2 status bit is set that surely means interrupt has fired but if
> it is not set then there are 2 possibilities
> 
> 1) ALARM2 is missed as the roll over happened
> 2) ALARM2 yet to fire as we are yet to get to the next second.
> 
> On the other hand Seconds gives me clear indication if we missed the
> interrupt or we are about to get one.

Yes, you still have to check seconds *before* retrying based on status.

That should do, right?

Johan