When trying to debug my non-secure app, I can hit the break point once, but after that, any stepping or continuing results in tfm_hal_system_halt via MPC_Handler(). I'm using Cortex Debug extension with JLink gdb server.
I tracked down the issue to flash read, which I think the debugger is doing. Although it's unclear why it's trying to read anything in the TFM secure area. For reference, we had the same application running on nrf91 w/ tfm and never had this issue.
I was able to get passed this issue by setting .gdbinit with
set mem inaccessible-by-default on
# Secure flash (TF-M) - no touch!
mem 0x0000d000 0x0002d000 wo
# Non-secure flash (app)
mem 0x0002d000 0x0016a000 rw
to prevent read access of TFM flash. However, this seems like a duct tape solution. Is there some Kconfig or other I can use to prevent this issue?
My launch.json config:
{ "name": "Cortex Debug nRF54", "type": "cortex-debug", "request": "attach", "cwd": "${workspaceFolder}", "servertype": "jlink", "toolchainPrefix": "arm-zephyr-eabi", "device": "cortex-m33", "executable": "./agmuster/build/agmuster/zephyr/zephyr.elf", "rtos": "/Applications/SEGGER/JLink/GDBServer/RTOSPlugin_Zephyr.dylib", "preAttachCommands": [ "source ${workspaceFolder}/agmuster/.gdbinit", ] },Thanks,
Konstantin
