The Question of app_error_fault_handler.

[hongpal]

Hello Nordic!

I'm developing with nrf52840, s340, and SDK 15.3.0

Sometimes, I got an error in app_error_fault_handler. 

This is information that gives me in app_error_fault_handler.

Received a fault! id: 0x00001001, pc: 0x00061c94, info: 0x00000000

SOFTDEVICE: INVALID MEMORY ACCESS

I know this problem is violate at softdevice section. 

pc location is memset fuction.  I want to know return address sins memset finised.

How to do that ? 

[Vidar Berg]

Hello, 

It sounds like your program is either passing an invalid pointer or the wrong number of bytes to memset. I did a quick test here, and it does not seem to be possible unwind the call stack to find the return address. But you may see the memset pattern in RAM if this is caused by a buffer overrun which would help you find the start pointer. The fault is triggered as soon as the app writes into the Softdevice's RAM region.

Another option is to verify the arguments each time you call memset in your code.  

[hongpal]

Thank you your answer!

But I don't understand a litle bit.

How to find the start pointer?? I don't know the started location.

[Vidar Berg]

The MEMACC assert will trigger as soon as the app writes to an address between 0x0 0xFFF so it is no point in inspecting the Softdevice region. Invalid pointers should be caught before you call memset. 

hongpal said:

Do I need to check out 0x1000 around ? What is meaning checking the global variables ?

Yes, you could try to check if there appears to be a pattern (all zeroes for instance) starting at address 0x1000. This could indicate that you have a memset() that crosses the Softdevice boundary. 

Have you reviewed your code for susceptible calls to memset? One with non-static input for instance.   

[hongpal]

If I do memset start at 0x1000, I don't need to check out and I don't get any error. is it right?

I don't understand yet. Why do the starting point indicate crossing the Softdevice boundary.

[Vidar Berg]

The starting point may point to one of your variables. Then you could check where that variable is set in your code. Anyway, have you reviewed each call to memset. And to you know the exact steps to trigger this fault? 

[hongpal]

I will try to review each call to memset next day. 

But, I don't know the exact steps to trigger fault.

[Vidar Berg]

Ok, I think it's a good start. Let me know if I can help with reviewing these calls, or debug your code.