sd_ble_gap_adv_stop() never returns.

RE: sd_ble_gap_adv_stop() never returns.

Hung Bui — Mon, 23 May 2016 10:25:33 GMT

@dav lion: please, open a new case, and please provide a test code that can reproduce the issue.

RE: sd_ble_gap_adv_stop() never returns.

dav lion — Fri, 20 May 2016 22:25:01 GMT

I am having similar problems calling sd_ble_gap_adv_stop. I can call it from main(), but neither from a scheduler callback nor from a ble_event handler. sdk11, softdevice130

RE: sd_ble_gap_adv_stop() never returns.

Carles — Tue, 05 Apr 2016 15:18:44 GMT

@AndersE we need more info then, I just can't explain it. Can you try removing all synchronization primitives from your scheduler, or calling sd_ble_gap_adv_stop() directly from main() without the scheduler?

RE: sd_ble_gap_adv_stop() never returns.

Hung Bui — Tue, 05 Apr 2016 15:14:18 GMT

Hmm, I'm out of idea, can you reproduce the issue on our nRF5 DK ? If you have some simple code that can replicate the issue, you can send us and we can test it here.

RE: sd_ble_gap_adv_stop() never returns.

AndersE — Tue, 05 Apr 2016 13:22:12 GMT

@Hung Bui: There is no difference in the behavior using sd_nvic_critical_region_enter() or __disable_interrupt(). The SD is still not returning when called to stop advertising.

RE: sd_ble_gap_adv_stop() never returns.

AndersE — Tue, 05 Apr 2016 08:11:19 GMT

@Hung Bui I'm changing the calls to use the critical_region function in nrf_soc.h but it was a bit trickier than I expected so I didn't have the time to finish it yesterday. I have a break point after the call to the SD and in never arrives there, so I make a manual break after a few minutes and the PC is moving around address 0x12ca8 WFE. Which I believe is the scheduler in the SD. I'll update here as soon as I get the code from nrf_soc.h in place.

RE: sd_ble_gap_adv_stop() never returns.

Hung Bui — Tue, 05 Apr 2016 07:56:56 GMT

@AndersE: Do you have update on the case? I guess you have done, but just in case, when you are debugging you should set the optimization level to 0 and when you are debugging you should not step through the softdevice API but better add the breakpoint after the call. This is because after CPU stopped for a breakpoint, the softdevice wouldn't be able to continue functional due to timing error.

RE: sd_ble_gap_adv_stop() never returns.

Carles — Mon, 04 Apr 2016 12:07:43 GMT

Disabling interrupts can cause serious issues with the SoftDevice. You should instead use the critical_region functions in nrf_soc.h. Can you try removing the disabling of interrupts to see if adv_stop() still locks up?

RE: sd_ble_gap_adv_stop() never returns.

AndersE — Mon, 04 Apr 2016 12:00:14 GMT

Yes, in the scheduler, when it is time for task switching. As all tasks are set up to be "run to completion" it will not be scheduling until after the call returns.

RE: sd_ble_gap_adv_stop() never returns.

Carles — Mon, 04 Apr 2016 11:34:26 GMT

Are you by any chance disabling interrupts somewhere in your code? __irq_disable() and family

RE: sd_ble_gap_adv_stop() never returns.

AndersE — Mon, 04 Apr 2016 11:32:29 GMT

I'm not using the macros, it is just: errCode = sd_ble_gap_adv_stop(); and it never comes back. If you by the SD assertion handler mean softdevice_assertion_handler(...) it is running and working but it never comes to that. The SD is doing something, but I can't really figure out what by just looking at the assembly code in the debugger.

RE: sd_ble_gap_adv_stop() never returns.

Carles — Mon, 04 Apr 2016 11:10:56 GMT

When you say "function call" I assume SuperVisor Call (SVC) since that's how we expose the SD API. Do you have a SD assertion handler set up? This could be the SD for some reason (that we can debug later) asserting and you not catching it.

RE: sd_ble_gap_adv_stop() never returns.

AndersE — Mon, 04 Apr 2016 11:08:48 GMT

yes, as a function call.

RE: sd_ble_gap_adv_stop() never returns.

Carles — Mon, 04 Apr 2016 11:05:14 GMT

So you're saying that the call to adv_stop() is being made from Thread Mode (aka main()) ?

RE: sd_ble_gap_adv_stop() never returns.

Carles — Mon, 04 Apr 2016 10:35:21 GMT

Can you edit your original post with the SD and SDK versions?

RE: sd_ble_gap_adv_stop() never returns.

AndersE — Mon, 04 Apr 2016 10:24:05 GMT

I can trace the code running around inside the SoftDevice, so something is happening but not the right thing to get it to return. There are no interrupts involved at all in this call.

RE: sd_ble_gap_adv_stop() never returns.

Carles — Mon, 04 Apr 2016 08:32:16 GMT

could this be a hard fault? Are you perhaps calling the SD function from a higher priority than the SVC one? If you are calling it from APP_HIGH then this could be the issue