The RTC driver topic

[Giovanni]

In my opinion the driver should not depend on the library or we would be back to the previous situation which is exactly what I wanted to avoid, the library should be optional.

The realtime counter clock is the system clock.

Giovanni

[barthess]

Next step performed.
Conversion code moved to rtc.c(h). Now HAL contains all needed functions.

[Giovanni]

Much nicer this way I think, users will not have to search for extra files.

Giovanni

[barthess]

Much nicer this way

Agree.
I have fixed RTC testhal application for F4x.
Waking up on it works fine in opposite to F1x I still have no ideas why.

[Michail]

Hello Barthess, hi Giovanni!

Think I've stumbled upon a bug in the RTC driver in current trunk on STM32F100C8.

In halInit() function hal_lld_init() calls before rtcInit().
hal_lld_init() calls hal_lld_backup_domain_init(), it calls rtc_lld_set_prescaler(), it calls rtc_acquire_access() and at last it calls rtc_wait_write_completed()
Call stack for my code is:

Code: Select all

8 rtc_wait_write_completed() rtc_lld.c:74 0x080059aa   
7 rtc_acquire_access() rtc_lld.c:87 0x080059d8   
6 rtc_lld_set_prescaler() rtc_lld.c:191 0x08005ad0   
5 hal_lld_backup_domain_init() hal_lld.c:76 0x08001992   
4 hal_lld_init() hal_lld.c:109 0x080019f4   
3 halInit() hal.c:66 0x08005b6c   


and in rtc_wait_write_completed() in os/hal/ports/STM32/LLD/RTCv1/rtc_lld.c we can see:

Code: Select all

static void rtc_wait_write_completed(void) {
  while ((RTCD1.rtc->CRL & RTC_CRL_RTOFF) == 0)
    ;
}

But RTCD1 was not initialised yet! RTCD1.rtc points to NULL and my program hangs here.

Code: Select all

"RTCD1"   {...}   
   rtc   0x00000000   
   callback   0x00000000   


[Giovanni]

Hi,

Thanks for finding, tomorrow I will fix this.

Giovanni

[Giovanni]

I just committed a fix, could you verify functionality? I lack HW for this.

Giovanni

[Michail]

Yes, now everything works fine!

Thank you!

[steved]

Just looked at the latest change to STM32/lld/rtc2/rtc_lld.c; calls to rtc_enter_init(); and rtc_exit_init(); have been added round the calendar initialisation. These calls are already present earlier and later in the code - should they have been deleted?

[Giovanni]

It is possible, probably I misunderstood the requested change there, I will look into it again tomorrow.

Giovanni