How to use hardfault exception library ?

[Salman Azmat]

Hi,

I'm using nrf52832 and sdk14.2 and need to handle hardfaults in my code. From what I've found on the forums and infocenter, I need to modify the "void HardFault_process(HardFault_stack_t * p_stack);" function in hardfault_implementation.c file. By default it resets the chip. So to test this function I enabled the hardfault_handler_enabled macro in sdk_config.h and generated a hard fault using following code.

int* ff = 0x80000000;
*ff = 3;

Now i get the error, but the system doesn't reset as it should acc. to the default implementation.

To further debug, i put a breakpoint inside "void HardFault_c_handler(uint32_t * p_stack_address)" function in hardfault_implementation.c file, but this function never gets called. So the hardfault library is not working properly. Is there anything that I'm missing here ?

Also, I tried to use following functions in my main.c, to generate hardfaults for testing, but i get "undefined reference" errors while compiling ?

hardfault_genhf_invalid_fp

hardfault_genhf_undefined_instr

hardfault_genhf_ldm_align

So, what am i doing wrong here ?

[Vidar Berg]

Hi,

hardfault_handler_<toolchain>.c needs to be included in order for HardFault_c_handler() to be invoked on a fault exception. Did you include one of the handler files from  \components\libraries\hardfault\nrf52\handler (different versions depending on toolchain)?

[Salman Azmat]

Hi, I included the gcc file as I'm using seggers embedded studio. But still the handlers don't get invoked. Currently my project path contains hardfault_handler_gcc.c and hardfault_implmenetation.c files, and I did "#include "hardfault.h"" in my main.c.

Though, I was able to generate hardfaults with "hardfault_genhf_invalid_fp()" function after doing #include "hardfault_genhf.h". But still no luck with the handler.

[Vidar Berg]

Segger Embedded studio halts the CPU when a fault exception occurs (shown as "vector catch" in console window). Maybe that is why the program is not reaching the hardfault handler in your application. You can disable vector catch for hardfault in the breakpoint view (press ctrl + alt + b to open)

 

 

[Salman Azmat]

Hi, I tried disbaling all the V7M breakpoints, but still it doesn't work.

[Vidar Berg]

Hi,

Did you write to address 0x80000000 to trigger this hardfault? Your screenshot shows the program being stuck in the softdevice's hardfault handler. This may happen if the app returns from app_error_fault_handler() after a softdevice assert. 

[Salman Azmat]

Hi,

I'm using hardfault_genhf_invalid_fp() function to generate the hardfault as shown in the picture in my previous comment. Also, how did you know the program is stuck in soft device's handler ? Also, I'm generating the fault before initializing the ble_stack and other modules.

[Vidar Berg]

Hi,

I've replicated the behavior you saw with hardfault_genhf_invalid_fp(). The softdevice should forward hardfault exceptions to the application, which it's not working in this particular case. The forwarding mechanism triggers another hardfault inside the fault handler that leads to a CPU lockup reset. This has been reported as a bug internally.  Note that lockup reset is disabled while the device is in debug interface mode, that is why the program hangs inside the softdevice when you try to debug it (lockup reset). 

"Also, how did you know the program is stuck in soft device's handle"

It was based on the CPU registers from your screenshot: 

Here's the example I used for test:

ble_app_hrs_hf_test.zip

[dlip]

Is there a workaround for this that will allow the application to process hardfaults?

[Vidar Berg]

Unfortunately, it is going to require changes to the interrupt forwarding implementation inside the softdevice. However, this problem only seems to occur when some specific condition(s) are met, I haven't experienced it before I tried hardfault_genhf_invalid_fp() that branches to address 0. So the fault exception should be forwarded to the application in most cases.