W dniu 10.12.2021 o 16:26, Alexandre Belloni pisze:

Hi,

On 25/11/2021 23:12:42+0100, Mateusz Jończyk wrote:

quoted

W dniu 24.11.2021 o 23:31, Alexandre Belloni pisze:

quoted

Hello,

By moving this patch later on in the series, you'd avoid the subsequent
refactor. I think this would be better for bisection later on.

Hi,

There are three issues I'm trying to fix in this series:

1. (less important) Insufficient locking in cmos_set_alarm()
2. misdetection of the RTC CMOS as broken on some motherboards,
3. reading / writing of the RTC alarm time during RTC update-in-progress.

Do you mean I should drop the patch
    nr 2. ("rtc-mc146818-lib: fix RTC presence check")
and instead straight away introduce mc146818_avoid_UIP() with the new approach (as in patch 3 in the series),
then modify mc146818_get_time() to use it (as in patch 4 - fixing issue nr 2),
then modify cmos_read_alarm / cmos_set_alarm to use mc146818_avoid_UIP() (patches 5-6, fixing issue no. 3)?

I was afraid this risks some confusion what is being fixed when.

I realize I never replied to this. This is fine, I'm planning to apply
the whole series once the few comments are fixed.

Great!

I've got other things mostly sorted out and tested, so I'll send a v4 shortly after some last-minute checks.

We'll probably get some breakage later on because many RTCs using this
driver are not adhering to the spec.

Thanks,

Mateusz