B fix #1480 : RTC not starting after soft reset

[adustm]

We needed to change the location of the check of RTC_INITS flag after the PWR_CLK_ENABLE and LSE LSI oscillator configuration.
Only the HAL_RTC_ENABLE and HAL_RTC_Init are bypassed in case the RTC_INITS is already set.
This has been tested with MBED_16 test.
This fits also with the request of the issue #1403.

[0xc0170]

Are you using any GUI? Can you set it up to use pull --rebase ? It would not produce the merge commits

[0xc0170]

Why this change? Why did it fail before?

[adustm]

Hello,
Yes it used to fail since PR #1449 merge, that was included in release 111.
The issue #1480 has been created explaining that the RTC is not re-starting after a soft reset.
We have fixed this issue #1480 , taking care of keeping the fix of the previous request #1403 .
Cheers

[0xc0170]

Merged manually

[adustm]

Hello, here is a quick test passed on the merged version :
| | Result | Target | Toolchain | Test ID | Test Description | Elapsed Time (sec) | Timeout (sec) | Loops
------------ | ------------- | ------------- | ------------- | ------------- | ------------- | ------------- | -------------
OK | DISCO_F429ZI | [F4D9] | uARM | MBED_16 | RTC | 4.67 | 20 | 1/1

[star297]

If this has been merged in Mbed-dev rev: 58, I'm afraid you can't light up that cigar.

Remains the same issue.

This function needs testing correctly:

int rtc_isenabled(void)
{
    return rtc_inited;
}

To something like this (that works):

int rtc_isenabled(void)
{
    if ((RTC->ISR & RTC_ISR_INITS) ==  RTC_ISR_INITS) {
        return 1;
    } else {
        return 0;
    } 
}

Otherwise the RTC time registers are reset on POR (with bat V applied) and nRST.

[star297]

There is another problem, the RTC now runs 10 seconds per minute slow on the L476.

[0xc0170]

@star297 Besides that rtc_inited can be removed, shouldn't this be functional? It checks for ISR_INITS in the init, thus should not reset it, or did we miss something?

@adustm rtc_inited check please

[0xc0170]

@star297 I just realized this one was a fix for #1403 , which is still open, is that correct that it should not be closed (my question above - is this not functional).

[star297]

Took some tracking to find the problems, but here's the fix:

Tested using L476 Nucleo
Set up tried with 48MHz and 80MHz HSE EXTC and EXTL( I have the 8MHz crystal fitted )

rtc_inited parts need to be removed.

The rtc_isenabled part above is required otherwise Mbed does not know it has been enabled.
The Mbed RTC api checks this, if the registers indicate not enabled it calls rtc_init.

Change line 162 in the hal_rcc.h file to:
#define RCC_LSE_TIMEOUT_VALUE ((uint32_t)300)

100mS is not enough time for the LSE to stabilize, 200mS works but add some contingency.
These crystal do vary quite a bit.
This fixes the slow RTC, that was falling back to the LSI.

Once this is done you should be okay to close both #1403 & #1480.
I will run some tests on the other boards.

[0xc0170]

@adustm @bcostm

[star297]

@adustm
@0xc0170

I have loaded Mbed_dev rev: 58, it is not correct and will not work.

This part has to be changed/ included:

int rtc_isenabled(void)
{
if ((RTC->ISR & RTC_ISR_INITS) == RTC_ISR_INITS){
return 1;
} else {
return 0;
}
}

All the "rtc_inited" code needs to be removed, it's no longer required and has no use here.

I have this implemented on several targets L476, L152, F4xx, F746 and is working correctly.

If you prefer I will send an individual PR with the working code for each of the targets I have fully tested.

Each target that uses the same embedded peripheral RTC in the Mbed line up will need this change to work properly. Some of the other 'low end' targets do not have the RTC->ISR register so must not be changed.

Paul