[PATCH] dm: rtc: Avoid a race in the rtc_reset test

[Simon Glass]

Hi Rasmus,

On Tue, 2 Aug 2022 at 01:08, Rasmus Villemoes
<rasmus.villemoes at prevas.dk> wrote:
>
> On 01/08/2022 21.13, Simon Glass wrote:
> > Hi Heinrich,
> >
> > On Mon, 1 Aug 2022 at 08:58, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
> >>
> >> On 8/1/22 15:59, Simon Glass wrote:
> >>> Hi Heinrich,
> >>>
> >>> On Mon, 1 Aug 2022 at 02:11, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
> >>>>
> >>>> On 7/31/22 20:27, Simon Glass wrote:
> >>>>> Since resetting the RTC on sandbox causes it to read the base time from
> >>>>> the system, we cannot rely on this being unchanged since it was last read.
> >>>>> Allow for a one-second delay.
> >>>>>
> >>>>> Fixes: https://source.denx.de/u-boot/u-boot/-/issues/4
> >>>>> Reported-by: Bin Meng <bmeng.cn at gmail.com>
> >>>>> Reported-by: Tom Rini <trini at konsulko.com>
> >>>>> Suggested-by: Rasmus Villemoes <rasmus.villemoes at prevas.dk>
> >>>>> Signed-off-by: Simon Glass <sjg at chromium.org>
> >>>>> ---
> >>>>>
> >>>>>    test/dm/rtc.c | 11 ++++++++---
> >>>>>    1 file changed, 8 insertions(+), 3 deletions(-)
> >>>>>
> >>>>> diff --git a/test/dm/rtc.c b/test/dm/rtc.c
> >>>>> index c7f9f8f0ce7..403bf5c640a 100644
> >>>>> --- a/test/dm/rtc.c
> >>>>> +++ b/test/dm/rtc.c
> >>>>> @@ -245,16 +245,21 @@ static int dm_test_rtc_reset(struct unit_test_state *uts)
> >>>>>        ut_assertok(dm_rtc_get(dev, &now));
> >>>>>
> >>>>>        ut_assertok(i2c_emul_find(dev, &emul));
> >>>>> -     ut_assert(emul != NULL);
> >>>>> +     ut_assertnonnull(emul);
> >>>>
> >>>> This is an unrelated change. It would be preferable to describe it in
> >>>> the commit message.
> >>>>
> >>>>>
> >>>>>        old_base_time = sandbox_i2c_rtc_get_set_base_time(emul, 0);
> >>>>>
> >>>>>        ut_asserteq(0, sandbox_i2c_rtc_get_set_base_time(emul, -1));
> >>>>>
> >>>>> -     /* Resetting the RTC should put he base time back to normal */
> >>>>> +     /*
> >>>>> +      * Resetting the RTC should put the base time back to normal. Allow for
> >>>>> +      * a one-second adjustment in case the time flips over while this
> >>>>> +      * test process is pre-empted, since reset_time() in i2c_rtc_emul.c
> >>>>> +      * reads the time from the OS.
> >>>>> +      */
> >>>>>        ut_assertok(dm_rtc_reset(dev));
> >>>>>        base_time = sandbox_i2c_rtc_get_set_base_time(emul, -1);
> >>>>> -     ut_asserteq(old_base_time, base_time);
> >>>>> +     ut_assert(base_time - old_base_time <= 1);
> >>>>
> >>>> If the operating system uses daylight saving time, this may still fail
> >>>> (very rarely).
> >>>>
> >>>> How about using gmtime() instead of localtime()? But that would be a
> >>>> separate patch.
> >>>
> >>> I'm not sure how to do this, as U-Boot expects local time. Of course
> >>> we could enhance the rtc API to support both (and use gmtime for the
> >>
> >> What makes you think that U-Boot expects local time? U-Boot has no
> >> notion of time zones. Linux systems tend to use UTC on the RTC. Why
> >> should sandbox_defconfig deviate?
> >>
> >> $ sudo hwclock --show -v
> >> hwclock from util-linux 2.38
> >> System Time: 1659365754.792540
> >> Trying to open: /dev/rtc0
> >> Using the rtc interface to the clock.
> >> Assuming hardware clock is kept in UTC time.
> >
> > Well the thing is, we want to show local times in U-Boot.
>
> Who's "we"?

Me, at least. E.g. the 'date' command on sandbox shows the current local time.

>
> I'm with Heinrich. I'd much rather U-Boot showed the time actually
> stored in the RTC, and if we emulate an RTC, that RTC should emulate
> what a real RTC does, namely store UTC. If you're worried about that
> that might confuse somebody, there's no harm adding " UTC" or "+00" or
> whatever to the output-for-humans (wherever that is).
>
> Especially when doing anything else causes weird and still-fragile hacks
> to be sprinkled throughout the testing code (that DST hack is ugly, and
> the "only twice a year" isn't accurate, because something on the host
> may also change /etc/localtime at any time).

OK, we have an is_dst but not many drivers actually set it, Anyway, it
doesn't specify the offset. So if we wanted to support a DST offset we
would have to add that to the API.

The patches in this series deal with the 'second' problem. They happen
to also deal with the DST issue mentioned by Heinrich in v1.

So one option is just to remove any mention of DST, since the code
changes are the same for the 'second' problem as the DST problem.

For changing sandbox to use gmtime, we could add the concept of a
timezone offset to U-Boot, along with an option to the 'date' command
to allow GMT or local time. Then we could change sandbox to internally
use gmtime(). Would that be useful?

As I said, it doesn't really affect the code, as I see it, but I could
certainly drop the DST comments if that helps.

>
> Rasmus
>
> PS: I loved working in Iceland. Apart from the beautiful nature, their
> timezone is UTC+0 all year round, so all questions of UTC vs localtime
> vs DST were moot :)

>From what I understand in the 1800s every town/city had its own time
based on the position of the sun, so things were even worse!

I'm not sure I like DST, actually. Perhaps I should move to Arizona?

Regards,
Simon