NRF_ECB peripheral does not always generate ENDECB event?

Question

Hi, 
 We use the ECB peripheral to encrypt and decrypt data with a proprietary driver implementation (which is not that different from the one in the SDK). It is working fine most of the time, but on rare occasions it appears to hang while waiting for the ENDECB event. The event appears not to be generated. The ERRORECB event is not being generated either. The STOP task is not used in our sources, and we have no concurrent use of the ECB peripheral that could throw things off. The AAR and CCM peripherals are not in use. I did not find any PAN on the ECB peripheral. 
 The SDK implementation in nrf_ecb.c applies a counter to prevent the "while (NRF_ECB->EVENTS_ENDECB == 0)" to hang indefinitely. I'm curious about the start value of 0x1000000 since the product specification lists a maximum value for T(startecb, endecb) of 17 microseconds and the countdown in the loop takes longer than that. 
 Is it possible that the ECB peripheral will not generate the ENDECB or ERRORECB events after a START task is issued? If so, what are the circumstances? And is there way (other than the counter) to either prevent, detect or correct this situation in software? 
 Thanks

Einar Thorsrud · Answer

Hi, 
 This should not happen, and I have not found any previous reports of this. The expected behavior is that ENDECB is always (and only) generated after the START task was triggered and block encryption has successfully been run to the end without being interrupted or corrupted. If not, ERRORECB should be generated. (Always one or the other). 
 I am not aware of any other way of handling this since you write that you don't get the ERRORECB vent in this case (which should not happen). But you should get events every time. Could it be that you lose it in your firmware somehow? Missing consecutive interrupts, or putting the event in a queue with some issues or something similar? 
 Do you have some code and a procedure that reproduces this on the DK?