DM6437 Exception Crash Debugging

Hi B,

Which version of BIOS are you using?

Bandeg said:

I have connected the JTAG and loaded symbols (release mode though so it wasn't compiled with symbols) and I can run this.   However the crash still doesn't tell me anything. 

In this scenario, do you see any output in the CCS standard out window?  What about in the BIOS system logs (need to know your BIOS version to help you better with that, but in short check the ROV tool).

Where is the program at when the crash occurs?  (you may need to check the dis/assembly window to see that).

Please also see these related forum posts if you haven't already, they may be helpful to you:

http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/99/t/6310.aspx

http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/99/t/41385.aspx

Steve

Okay. Below are some logs from the "Execution Graph Details." 

According to http://www.ti.com/lit/ug/spru732j/spru732j.pdf  says that Internal exception: IERR=0x1 is a fetch packet exception??    Any idea what this means?   How can I debug this further using CCS?

Also, the program stopped at: C$L1 in assembly.  What does this symbol refer to?

Thanks,

-B

You are getting an exception because some code somewhere is doing something illegal.

At this point you need to try to trace backwards in order to gain clues as to what line(s) of code are causing the exception.  The information you've found will help.

Bandeg said:

Also, the program stopped at: C$L1 in assembly.  What does this symbol refer to?

C$L1 is probably just a label within some function and you can use the disassembly window to find which function that is.  If you click the mouse on C$L1 in the disassembly window, so that the cursor is in the window, you can then use the keyboard arrows or keyboard page up/down to navigate in the disassembly window (for whatever reason, clicking with the mouse on the window scroll bar doesn't work well in the disassembly window, so please use the keyboard).

Once you have the cursor in the window at C$L1, scroll or page up until you see a symbol for a function name (you may have to move up for a while and there may be a lot of other labels).

What function symbol do you see first?  This will be the function that you are in at "C$L1".

Bandeg said:

937255228     NRP=0x5eeaff84

Can you enter this address into the disassembly window?  This is probably pointing to where the offending code instruction is.

Bandeg said:

937255225   TSK: running dynamic TSK (0x8102ec5c)

937255226   SWI: end   KNL_swi (TSK scheduler) (0x8543e2ac) state = done

937255227   EXC_exceptionHandler: EFR=0x2

Looking at the output, it looks like the above bolded task ran last.  Can you check the code of that task?  It may contain the offending code.

You should also check this forum post if you haven't already, as the problem is similar.

Steve

I doubt that NRP address is correct.  The .text region is at 0x85000000.  

The dynamic task runs the heart of our code, and it is large.  I have been using LOG_printf() to get closer.  

Can the UBM help me here? 

http://processors.wiki.ti.com/index.php/DSP_BIOS_Debugging_Tips

http://processors.wiki.ti.com/images/0/0d/Ubm-ext-02.pdf

I am using CCSv3.3 and I'd like to set-up a Watchpoint new action in the UBM window (page 9).  This window only shows me "Software Breakpoint and Hardware Breakpoint" as my options.   How do I enable all the Watchpoint option so that I can watch the stack?

Regards, B