topic Debugger Losing Connection, Symbols, or... ? in MCUXpresso General

Debugger Losing Connection, Symbols, or... ?

mattschrader — Tue, 12 Mar 2019 16:23:17 GMT

I've been working with MCUXpresso for a while now, and this problem has been a recurring one... Occasionally when debugging, the target is flashed successfully, but then the debugger looses connection or otherwise does not know what is going on. I see something like the attached screenshot. This is happening on my custom hardware as well as the development board OM13092 (LPC54608).

Until recently, I was primarily using the LPC-Link 2 with the CMSIS-DAP probe firmware. Recently, I decided to try changing to SEGGER J-Link. This problem occurs with both.

The most fustrating thing is that this behavior seems to be random. It is usually fixed by some combination of:

Power cycle LPC-Link 2
Power cycle Target
Run "Clean Up Debug" in MCUXpresso IDE
Close MCUXpresso IDE

Currently using MCUXpresso IDE v10.3.0

I'll also point out here that I recently found and followed the guidance provided in this article on thread-aware FreeRTOS debugging:

Troubleshooting Tips for FreeRTOS Thread Aware Debugging in Eclipse | MCU on Eclipse

Re: Debugger Losing Connection, Symbols, or... ?

converse — Tue, 12 Mar 2019 17:35:55 GMT

That address (0xa4010000) is in off chip ram - do you have any? Whatever, it is outside your application so the debugger has no information to display.

my guess is some memory corruption- most likely stack. You probably should closely examine your code to see if you can find where you have an overflowing buffer or an uninitialised pointer.

errors like this are really difficult to track down, because they appear random.

i think it is highly unlikely that there is a hardware or debugger problem - almost certainly it is a software problem.

good luck

Re: Debugger Losing Connection, Symbols, or... ?

mattschrader — Tue, 12 Mar 2019 23:58:09 GMT

Thank you for the post and astute observation. Yes, there is off chip RAM but it it is located at 0xa0000000 with size of 0x1000000. After further review I found this in the debugger console:

...
Downloading 10116 bytes @ address 0x1028ED10 - Verified OK
J-Link: Flash download: Bank 0 @ 0x00000000: Skipped. Contents already match
J-Link: Flash download: Bank 1 @ 0x10000000: Skipped. Contents already match
Writing register (PC = 0xa4010000)
Reading all registers
WARNING: Failed to read memory @ address 0xA4010000
Read 4 bytes @ address 0x03000CAA (Data = 0x2DE9F041)
Read 2 bytes @ address 0x03000CAA (Data = 0x2DE9)
...

For whatever reason, my launch configuration was setup for big endian. It was trying to set PC to 0xa4010000 instead of 0x1a4. I still don't understand how it would sometimes work with this incorrect setting?