Are there any cases soft device resets system?

[Ben.Qin]

I encountered a strange problem system reset. My application runs on iOS or android. My device was connected by iPhone 6 plus(128G) with iOS 8.1.2. ANCS was turned on. I have never seen the same problem by using iPod and iPad or phones and pads with android. Hard fault and assert are excluded. Watch dog is turned off. NVIC_SystemReset() and sd_nvic_SystemReset() aren't called at another place. So could you tell me whether soft device resets system in some cases?

[Ulrich Myhre]

To my knowledge: No, the SoftDevice will not reset the system unless SystemReset() is explicitly called. It can assert if the SoftDevice discovers a logical bug in itself, and the system can HardFault if you divide by zero or give the wrong lengths when passing data to the stack, but it should not reset.

Some of the previous SDK examples used to call reset in the application error handler. Could that be the cause?

[Ben.Qin]

All SystemReset() calls were replaced by an infinite loop in the application, including error handler. Recently, I caught the information before reset. It is file: ll_lm_s0.c, line: 1128, err_code: 0xDEADBEEF. Some one encountered the same problem. I didn't find a way to solve it from this site.

[Ulrich Myhre]

Thanks, this is a stack assert, meaning the SoftDevice went into some undefined state and stopped everything. It will trigger a HardFault once the assert routine is finished. Could you please tell me which version of the SoftDevice you are using, and roughly what you were doing? I will see if there are any workarounds available. If not, Support would be more than happy to help you with this.

[Ben.Qin]

Soft device s110_nrf51822_6.0.0 is used. It is hard to describe what I were doing. I implement a wearable device step counter. If iOS doesn't reconnect the device, reset seems not to appear again. Support of Nordic semiconductor gave me the code for reconnection: adding two app timers. I suspect there is something with ANCS. Other iOS devices, such as iPod, iPad, iPhone 5s and so on, connected to the same wearable device, then I have never seen this problem. If this information does helpless I'll add more.

[Ulrich Myhre]

Asserts should not happen no matter what the peer does, so this is an issue on our side. I am personally not sure if this is fixed in later versions (S110 7.x and 8.x.x are both out), but I would contact support for this and let them find a proper workaround for your case.

[Kenneth]

I believe this issue is already described in the release notes for the S110 v6.x softdevice. Most likely the code you have received using app_timers ensure to meet the workaround for the issue (add a 50ms delay). See snippet from release notes where this issue and workaround is described. devzone.nordicsemi.com/.../6b513626e1935870508b1616df2db161