Memfault for the STM32H7xx Family

memfault_sdk.mk includes an example makefile that can be included into a build system to integrate the reference port into a stm32h7 device. It expects two variables to be defined prior to including the makefile

• EMBEDDED_MFLT_SDK_ROOT - The location for the root directory of this repository
• STM32_CUBE_DIR - The location of the root directory for the STM32 Cube HAL. This demo was tested with the V1.5.0 STM32H7CUBE HAL available here

Example Steps for Adding port to ChibiOS Demo

Setup

Download the following items:

• git clone https://github.com/ChibiOS/ChibiOS.git
• cd ChibiOS/demos/STM32/RT-STM32H743I-NUCLEO144 and
  • Move the downloaded contents of STM32Cube_FW_H7_V1.5.0 (link above) to stm32_cube
  • clone the memfault firmware sdk: git clone https://github.com/memfault/memfault-firmware-sdk.git

Integrate Memfault into the ChibiOS build system. This can be achieved by running: git apply <path_to_memfault_sdk>/examples/stm32/stm32h743i/chibios-memfault-integration.patch

The patch does three things:

• includes memfault_sdk.mk into the ChibiOS build system
• defines a new rule for building the files needed for memfault (sdk, platform port, and flash stm32 hal files)
• forces an assert in the main loop within main.c for simple testing

Testing

At this point you should be able to run make and compile the code & then load the resultant elf at build/ch.elf onto your board.

For example, I'm connected to the nucleo board with JLink & GDB:

JLinkGDBServer -if swd -device STM32H753ZI -speed auto -port 2331

And run the following to flash the board:

arm-none-eabi-gdb-py --eval-command="target remote localhost:2331" --ex="mon reset" --ex="load" --ex="mon reset" --se=build/ch.elf

Soon after boot, you should hit the assert in the main loop and a breakpoint:

memfault_platform_halt_if_debugging () at ./memfault-firmware-sdk/examples/stm32/stm32h743i/platform_reference_impl/memfault_platform_core.c:17
17      __asm("bkpt");
(gdb) bt
#0  memfault_platform_halt_if_debugging () at ./memfault-firmware-sdk/examples/stm32/stm32h743i/platform_reference_impl/memfault_platform_core.c:17
#1  0x08009af2 in memfault_fault_handling_assert (pc=<optimized out>, lr=lr@entry=0x800034d <endinitloop+4>, extra=extra@entry=0) at ./memfault-firmware-sdk/components/panics/src/memfault_fault_handling_arm.c:162
#2  0x0800967e in main () at main.c:76

You'll want to step over the breakpoint and then continue. At this point the coredump saving logic will run and the next breakpoint hit will be right before reboot

(gdb) next
memfault_fault_handling_assert (pc=<optimized out>, lr=lr@entry=0x800034d <endinitloop+4>, extra=extra@entry=0) at ./memfault-firmware-sdk/components/panics/src/memfault_fault_handling_arm.c:174
174   *icsr |= nmipendset_mask;
(gdb) continue
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap.
memfault_platform_halt_if_debugging () at ./memfault-firmware-sdk/examples/stm32/stm32h743i/platform_reference_impl/memfault_platform_core.c:17
17      __asm("bkpt");
(gdb) bt
#0  memfault_platform_halt_if_debugging () at ./memfault-firmware-sdk/examples/stm32/stm32h743i/platform_reference_impl/memfault_platform_core.c:17
#1  memfault_platform_reboot () at ./memfault-firmware-sdk/examples/stm32/stm32h743i/platform_reference_impl/memfault_platform_core.c:22
#2  0x08009a24 in memfault_fault_handler (regs=<optimized out>, reason=<optimized out>) at ./memfault-firmware-sdk/components/panics/src/memfault_fault_handling_arm.c:124
#3  <signal handler called>
#4  0x00000000 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

Now you can extract the coredump via gdb. The example port is configured to save coredumps in the last sector of flash and can be extracted from GDB using:

dump binary memory memfault-core.bin 0x81E0000 0x8200000

You can then navigate to the Issues page on the memfault dashboard and click on “Manual Upload” to upload the file collected.