Re: [PATCH 11/29] rtc: pcf2127: set RTC_FEATURE_ALARM_RES_2S

[Alexandre Belloni <alexandre.belloni@bootlin.com>]

Hi,

On 17/03/2022 11:28:20-0400, Hugo Villeneuve wrote:
> On Thu, 10 Mar 2022 21:58:49 +0100
> Alexandre Belloni <alexandre.belloni@bootlin.com> wrote:
> 
> > On 10/03/2022 11:09:18-0500, Hugo Villeneuve wrote:
> > > On Wed,  9 Mar 2022 17:22:42 +0100
> > > Alexandre Belloni <alexandre.belloni@bootlin.com> wrote:
> > > 
> > > > The PCF2127 doesn't support UIE because setting an alarm to fire every
> > > > second confuses the chip and the fastest we can go is an alarm every 2
> > > > seconds.
> > > 
> > > Hi Alexandre,
> > > can you describe what "confuses the chip" means?
> > > 
> > > In my experimental PCF2131 driver, I activated UIE and it seems to be working fine at 1s intervals, but since it is similar to PCF2127, maybe there is still a problem and I just didn't see it.
> > > 
> > 
> > Did you remove uie_unsupported? Else, you may have been using uie
> > emulation. In my tests last year, the pcf2127 was failing to reassert
> > the interrupt if an alarm was set every second. The same happens on
> > other NXP based RTCs (i.e. including microcrystal ones).
> > 
> > I'm going to test again soon (and also reply to your series).
> 
> Hi,
> I have now access to a board with a PCF2129T.
> 

I actually did the test again this afternoon with a pcf2127...

> I have been able to test with it by sending ioctl RTC_UIE_ON and
> confirm that it exhibits the problem you reported. Basically, the
> first alarm triggers the IRQ after 1s, as expected, but the next
> configured alarm 1s into the future never triggers the IRQ again. But
> the time/date registers seem to be updating (incrementin) correctly
> after that.
> 

...and so we agree on what we observe. My guess is that the alarm
comparator has a latch that triggers only every second update and so you
need 2 seconds between each alarms.

[...]

> I even did some tests where, in pcf2127_rtc_set_alarm(), I disabled
> all the alarm registers (set to 0x80), except for the SECONDS alarm
> register. In this case, I observed that the IRQ is triggered after
> 61s. This seems to indicate that the alarm detection circuitry is
> still somewhat operating (AF flag / INT pin), but there is probably a
> bug in the IC when the configured alarm is close to current time after
> AF is set once...
> 
> I also have done other tests by disabling then re-enabling AIE,
> stopping and restarting the oscillator (STOP bit), etc, but to no
> avail.
> 
> However, in pcf2127_rtc_set_alarm(), if I set the alarm 2s past the
> current time, instead of 1s, then the IRQ is always triggered after
> 2s.
> 
> I have opened a support case with NXP to investigate this strange
> behavior.

Let me know if you ever get a reply.

> 
> Hugo.
> 
> 
> > > > Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
> > > > ---
> > > >  drivers/rtc/rtc-pcf2127.c | 1 +
> > > >  1 file changed, 1 insertion(+)
> > > > 
> > > > diff --git a/drivers/rtc/rtc-pcf2127.c b/drivers/rtc/rtc-pcf2127.c
> > > > index f8469b134411..63b275b014bd 100644
> > > > --- a/drivers/rtc/rtc-pcf2127.c
> > > > +++ b/drivers/rtc/rtc-pcf2127.c
> > > > @@ -656,6 +656,7 @@ static int pcf2127_probe(struct device *dev, struct regmap *regmap,
> > > >  	pcf2127->rtc->range_min = RTC_TIMESTAMP_BEGIN_2000;
> > > >  	pcf2127->rtc->range_max = RTC_TIMESTAMP_END_2099;
> > > >  	pcf2127->rtc->set_start_time = true; /* Sets actual start to 1970 */
> > > > +	set_bit(RTC_FEATURE_ALARM_RES_2S, pcf2127->rtc->features);
> > > >  	clear_bit(RTC_FEATURE_UPDATE_INTERRUPT, pcf2127->rtc->features);
> > > >  	clear_bit(RTC_FEATURE_ALARM, pcf2127->rtc->features);
> > > >  
> > > > -- 
> > > > 2.35.1
> > > > 
> > > 
> > > 
> > > -- 
> > > Hugo Villeneuve <hugo@hugovil.com>
> > 
> > -- 
> > Alexandre Belloni, co-owner and COO, Bootlin
> > Embedded Linux and Kernel engineering
> > https://bootlin.com
> > 
> 
> 
> -- 
> Hugo Villeneuve <hugo@hugovil.com>

-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com