Dectecting the presence of a debugger on STM32F767 does not work

[Richard Peters]

Posted on June 06, 2018 at 10:00

Hello all!

I'm trying to detect whether a debugger is attached to my program. I've read in other threads (like this one: 

https://community.st.com/thread/25890?commentID=69875&sharpcomment

  ) that there is a register in the SCB named DHCSR (Debug Halt Control and Status Register) which is used by debuggers who set bit 0 when they attach. 

http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0337e/CEGCJAHJ.html

  gives more details about this register.

However, when I read out that register on my Nucleo-F767ZI board, bit 0 is always set to 1. Of course it should be 1 when I launch my program from System Workbench, but when I disconnect the debugger and reset the microcontroller, bit 1 is still set. The contents of DHCSR is different when debugging or not debugging (when debugging, it is 0x1000d, when not debugging, it is 0x1010001), but I expected the least significant bit to be 0 when not debugging.

I use this code to read out DHCSR:

volatile unsigned int* const SCB_DHCSR = (volatile unsigned int *)0xE000EDF0;

if (*SCB_DHCSR & 1)

    ...

Am I doing something wrong? Is this a peculiarity with the STM32F767?

null

[Tesla DeLorean]

Posted on June 06, 2018 at 13:37

On the CM7 you usually have to unlock access to ITM and DWT registers

Equivalent code to this for assorted LAR in debug peripherals

volatile unsigned int *DWT_LAR      = (volatile unsigned int *)0xE0001FB0; //address of the register

*DWT_LAR = 0xC5ACCE55; // unlock

Might dig into this later.

[Richard Peters]

Posted on June 06, 2018 at 13:59

Thanks for the quick response! I gave it a try, but unfortunately even with the st-link disconnected I get a read out of 0x1010001, indicating that it thinks a debugger is attached.

[Richard Peters]

Posted on June 06, 2018 at 14:46

Yes, I've removed the two jumpers connecting the SWD lines from ST-LINK to the STM32F767. By the way, I've upgraded the ST-LINK device with the J-LINK software, but since I've removed the jumpers that shouldn't matter.

[Tesla DeLorean]

Posted on June 06, 2018 at 14:11

Are you pulling the ST-LINK jumpers in the test case? I might expect the mbed firmware to probe the device

[Richard Peters]

Posted on June 12, 2018 at 07:49

I've figured out what was going on: Although the debugger sets that specific bit in SCB_DHCSR, it is never reset to 0, not even by a hard reset. When power is removed though, the bit is reset. With this behaviour, the bit indicates whether a debugger /is/ attached /or has been/ attached since the last power up. Well, works good enough for my purpose.