[bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Wed, 28 Aug 2019 13:18:07 GMT

Hi Nick,

This doesn't seem normal.

Could you use a logic analyzer/Oscilloscope to check the pattern of the CPU active time ? You can toggle and LED near the sleep command to know when it's up.

If you check other example which use the app timer, for example ble_app_hrs, do you see the same problem ?

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Thu, 22 Aug 2019 11:02:12 GMT

Hi Nick,

I'm really sorry for the inconvenience.

You may hit the same issue as this one.

It's a limitation of app_timer, is it possible to enable the app timer after the softdevice, or stop the timer (to avoid any interrupt) when you enable the softdevice ?

Since app_timer2 is not intensively tested I would suggest you follow Markku suggestion to keep a dummy app_timer running all the time.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

mrono — Thu, 22 Aug 2019 06:39:21 GMT

[quote userid="78841" url="~/f/nordic-q-a/29144/bugreport-app_timer_stop-resets-rtc1-counter-when-app_timer_keeps_rtc_active-is-set/205264"]The actual requirement is to get a reliable timestamp with BLE stack running, an event scheduler and sleep/wake power-modes invoked. Fundamental requirement really,[/quote]

I struggled with the same situation, though I'm using SDK 15.2. The easiest workaround is to make sure you always have at least one app_timer running. So even if you don't need any app_timers, keep a dummy one running in the background.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

NickT — Wed, 21 Aug 2019 10:21:05 GMT

Thanks for the help so far.

We have been gradually breaking our product as I fixed each of the inconsistencies between app_timer2 and app_timer. We have a BLE stack enabled, and a scheduler. Using app_timer2 and app_timer_rtc from your zip file earlier with a modification to avoid the use of NRFX, we get error 4097 from nrf_sdh_enable_request if we initialize the BLE after the timers, the code locks up if we initialize the BLE before the app_timers and continually restarts if we initialise scheduler before the timers.

Obviously, since we cannot do in-circuit debugging with a BLE stack, this whole thing is working "blind".

As I have one more week in which to try to fix this, I plan to revert this branch and start over.

The actual requirement is to get a reliable timestamp with BLE stack running, an event scheduler and sleep/wake power-modes invoked. Fundamental requirement really,

Any help appreciated.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Tue, 20 Aug 2019 12:58:11 GMT

Hi Nick,

I'm sorry for the confusion. It was an old case and I wasn't following it earlier.

I had a look at the implementation we have now on SDK v15.3 and it turned out that we had little development on this over the last 2 years.

So the official app timer implementation we have in SDK v15.3 remains unchanged compare to what on SDK v13.x/14.x

The new implementation, app_timer2.c, was introduced but remains in experimental. This means it can be used as an example code but it's not extensively tested.

The solution for now is either:

- Adding a dummy timer to keep the RTC running

- Use app_timer2 which requries you to copy all the dependencies (nrfx drivers) into SDK v14.2

I will check internally with R&D if we can do anything.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

NickT — Fri, 16 Aug 2019 14:23:05 GMT

HungBui - Thanks for the rapid response.
sorry about this, but I do not completely understand what you mean

The function timer_list_remove() in apptimer.c file in SDK15 looks identical to the one in 14.2 - so I don't know how porting SDK15 timer library helps.

The apptimer2.c file in my copy of nRF5_SDK_15.3.0_59ac345 is in a subdirectory called "experimental."

Please clarify - am I supposed to replace apptimer.c in my project source list with apptimer2.c from experimental subdirectory in 15.3?

Are there any other files need replacing at the same time?

Are there other things required such as #define APP_TIMER_V2 and if so, where to define that?

What side effects of this "experimental" version are known since release?

Thanks

Edit: addition - trying to put the experimental version from 15.3 back into my 14.2 project falls over heavily on the lack of nrfx. Please advise before I get too deep into this.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Fri, 16 Aug 2019 08:43:19 GMT

Hi Nick,

You can either take the official app timer in SDK v15.x and back port it to SDK v14.2 or you can use apptime2 that I provided earlier. You would need to replace the whole file, use it as a different module to the original app timer.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

NickT — Thu, 15 Aug 2019 13:52:00 GMT

Realise this is old hat, but was there a workaround within SDK_14.2.0_17b948a ? (we are stuck with this older version for our product)
The supplied file app_timer2 is significantly different from app_timer..c and I am not sure from the above discussion whether to replace the whole file or just a few lines

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Tue, 23 Jan 2018 15:22:05 GMT

Glad to hear it fixed the problem. Unfortunately we can't talk about timeline here but it shouldn't take too long to be released. I would say 1 2 months.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Keton — Mon, 22 Jan 2018 13:48:46 GMT

@hungbui the code you've provided solves the problem. Thanks for providing support. as for nrfx this looks very promising for integration with RTOS I'm doing for another project. Do you have any rough estimate when SDK15 will be out so I can try SDK libraries which are using it?

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Fri, 12 Jan 2018 12:23:16 GMT

nrfx is something coming with SDK v15. You can have a look at the driver here.

The modification you added seems fine. Sortlist can be taken from SDK v14.2, it's fine. Please let us know if there is any issue.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Keton — Fri, 12 Jan 2018 10:24:54 GMT

@hungbui I see that app_timer2 requires nrf_sortlist module which is not included in SDK 13.1. Should I use SDK14.2 one or will you provide SDK15 variant?

EDIT: also nrfx.h file is missing both from 13.1 and 14.2 EDIT2:

since nrfx.h is missing I've had to patch app_timer_rtc.c and enable nrf_drv_clock. Now my drv_rtc_init() and drv_rtc_uninit() look like this:

ret_code_t drv_rtc_init(drv_rtc_config_t const * p_config)
{
    ret_code_t err_code = nrf_drv_clock_init();

    if (err_code != NRF_SUCCESS)
    {
        return err_code;
    }

    nrf_drv_clock_lfclk_request(NULL);
    m_irq_func = p_config->irq_func;
    nrf_rtc_prescaler_set(NRF_RTC1, p_config->prescaler);
    NVIC_SetPriority(RTC1_IRQn, p_config->irq_priority);
    NVIC_EnableIRQ(RTC1_IRQn);
    return NRF_SUCCESS;
}

void drv_rtc_uninit(void)
{
    NVIC_DisableIRQ(RTC1_IRQn);
    drv_rtc_stop();
    nrf_drv_clock_lfclk_release();
}

Could you please check if that's correct approach or maybe I should adapt rtc1_init() / uninit functions?

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Fri, 12 Jan 2018 10:04:16 GMT

Hi Keton,

I'm terribly sorry for the delay. I did send your request to our developers. But as I was afraid of we can't run a thorough test on the workaround and even though it looked fine we were not completely sure if there wasn't any corner case that can crash the timer. And as you reported, it crashed.

Our suggestion is that you can use our new implementation. It still in experimental phase but passed all our internal tests, and will come officially in SDK v15:

app_timer2.zip

The API is the same, but you may need to add some define for the log into sdk_config.h:

APP_TIMER_CONFIG_LOG_ENABLED, APP_TIMER_CONFIG_LOG_LEVEL , APP_TIMER_CONFIG_INFO_COLOR , APP_TIMER_CONFIG_DEBUG_COLOR

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Keton — Thu, 11 Jan 2018 18:03:44 GMT

@hungbui any updates on this? My quick fix has negative side effects - timers stop working after application is active for a few hours (maybe rtc counter rollover breaks things). If you aren't going to propose a fix just let me know so I can start implementing my own timer library.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Fri, 22 Dec 2017 08:58:43 GMT

I understand. We will try to get back to you as soon as possible.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Keton — Thu, 21 Dec 2017 17:57:13 GMT

As a temporary workaround it's fine but for the release I'm afraid it won't do.

Per safety standards we must be compatible with such timer is considered a 'dead code' and should be removed. The software module which uses my timestamp function for debouncing and tracking time between interrupt events is used as a library outside of our development team.

It's very likely that upstream developers will simply omit this timer either by mistake or thinking that it shouldn't be there since it does nothing. SDK patch which makes app_timer_cnt_get() robust would be a much better solution.

I understand that is's holiday time but I'd really appreciate if someone could take a look at my patch when Christmas break ends.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Hung Bui — Thu, 21 Dec 2017 09:56:11 GMT

Hi Keton, We are sorry that we are on reduced staff now due to Xmas vacation, we can't do thorough test on your workaround. But would keeping a timer running all the time cause any draw back to you ?

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

Keton — Tue, 19 Dec 2017 10:53:03 GMT

Great, but this project cannot be migrated to unreleased yet SDK15. We are near the end of our deployment window and can't make any changes. Could you please verify my fix or propose a better one?

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

MartinBL — Mon, 18 Dec 2017 10:03:47 GMT

Hi,

The SDK team is aware of the vulnerability, but they haven't looked thoroughly into it because they have been working on an entirely new implementation of the library. The new implementation is scheduled to be included in SDK 15.

RE: [bugreport] app_timer_stop() resets RTC1 counter when APP_TIMER_KEEPS_RTC_ACTIVE is set

paul — Fri, 15 Dec 2017 19:13:56 GMT

I also encountered this issue today and found the same problematic block of code that resets the timer. I've already modified the SDK13.1 app_timer code with the change mentioned in this question, and I'm hesitant to add another hack. Thinking its safer to use an always running timer instead of APP_TIMER_KEEPS_RTC_ACTIVE=1.

@Nordic, any thoughts here?