JLINK Disconnecting? Or just a bug?

[kevray]

I've either got a bug in my code or the JLINK connection is dying. Either way it's a real headache to pinpoint and discover how to resolve. Knowing more about extracting info from the JLINK Log would probably be helpful in these cases.

Within Crossworks I see that the JLINK has halted with an error. However, my device continues to run. Crossworks shows what looks like a dereferenced null pointer. Stack is blown. Registers are all 0. IP is at 0x00000000.

The crash/disconnect happens at random times usually within 5 minutes. The device continues to operate and internal logging of events shows no errors.

The tail of the JLINK log shows: T0B70 1112:717 JLINK_ReadMemU32(0xE000101C, 0x0001 Items, ...)-- failed - DWT_PCSR T0B70 1112:765 JLINK_ReadMemU32(0xE000EDF0, 0x0001 Items, ...)-- failed - DHCSR T0B70 1112:766 JLINK_Halt()

Many thanks for any pointers. I will followup with the solution if I find one.

Target device: nRF51822_xxAC with multiple I2C connected sensors, s130 connected via SWD on nRF51-DK. GPIOTE+TWI+BLE used in application.

JLINK_LOG_TAIL.txt

[kevray]

Updated JLINK to v512f -- (v512g crashes). Updated Crossworks. Reflashed the nrf51-DK. Problem went away.

Updated SDK9->SDK11 and found out TWI is broken again. :(

"The new SDK 11 TWI driver is not able to do two transmits without repeating the ADDRESS + Write bit byte. Hence we need to merge the MPU register address with the buffer and then transmit all as one transmission."

[MartinBL]

I'm glad you solved your problem. The TWI issue is briefly discussed here and here if someone is interested. It is related to the nRF52 hardware, but since you are using nRF51 you can work around it by using the drivers from SDK 10.

[kevray]

Thanks Martin. Can you point me to a post about swapping the driver from SDK10? Or is it just a drop in of nrf_drv_twi.* from that SDK?

[MartinBL]

Unfortunately we don't have any migration guides, but I guess you can look in the SDK 10 examples or maybe this simple example. I believe the look and feel of the drivers are quite similar, but some of the functions take some different parameters and the event handler "redesigned" when moving from SDK 10 to 11. However, in my experience it isn't that hard to port.

[kevray]

Thanks. Can we update the SDK11 known issues list with this until it's fixed?

[MartinBL]

I reported it in the thread. Although I'm a little in doubt if it belongs there as technically it is not an SDK bug, but rather a hardware bug in the nRF52. The SDK team chose (inconveniently) to remove the feature from the nRF51 part of the driver as well for consistency. However, they are aware of it and they miss the feature too. So maybe one day it will be back.