Data Watchpoint on Cortex M3

[kr_rahul]

Posted on August 15, 2009 at 11:59

Data Watchpoint on Cortex M3

[kr_rahul]

Posted on May 17, 2011 at 13:19

Hi,

I am trying to place a data (access) watch-point on a global variable using OpenOCD which I have connected to STM32 through Amontec JTAG. However the watch-point is never hit (i.e. the program execution doesn't halt even when the variable is read/written or loaded/stored). Hence, I was wondering whether anyone has tried using the data-watch-point feature of Cortex M3 on STM32. Thanks and regards, Rahul

[st3]

Posted on May 17, 2011 at 13:19

Quote:

I was wondering whether anyone has tried using the data-watch-point feature of Cortex M3 on STM32.

Yes, I have had this working with Keil; so it must be down to the tools and/or your usage of them - it's not a limitation of the chip itself

[kr_rahul]

Posted on May 17, 2011 at 13:19

Dear st7,

Thanks for your reply. Could you please elaborate as to which Menu-option in Keil, should I use for setting hardware data watch-points (read, write OR access). My intention is to set an access watchpoint data watchpoint on a global variable. Through the watch-window I can verify that the global variable (which is a loop_count) is changing with every iteration of the loop. Further I can get the memory location of the global variable (as 0x200000e8) from the map file. I can also see that in the memory window that this memory location is getting updated regularly. However I haven't been able to place a h/w watch-point for read/write/access on this variable. I have got ver-3.72 of Keil-uVision which I can connect to STM32 using ST-LINK JTAG debugger, through the DLLs provided by Keil. Thanks again, Best Regards, Rahul

[sjo]

Posted on May 17, 2011 at 13:19

It can be done in openocd:

# break when variable @ adr 0x20000000 is 5 mon wp 0x20000000 4 a 5 0 if you are not using gdb then remove the mon (monitor) cmd Cheers sjo

[st3]

Posted on May 17, 2011 at 13:19

http://www.keil.com/support/man/docs/uv3/uv3_db_dbg_breakpnts.htm

[kr_rahul]

Posted on May 17, 2011 at 13:19

Dear st7,

Thanks again for your quick response. I can get the breakpoint-window (for entering watch-points) through the debug menu in the simulator mode. However when I am debugging on the actual hardware (with code in flash and data in RAM) this window doesn't appear. I have tried clicking on the ''Breakpoints'' icon as well as pressing ''Cntrl+B''. Hence could you please clarify which JTAG debugger you were using for testing watchpoints. I am using ST-LINK JTAG debugger. The associated dll files supplied by keil are: a) ST-LINKIII-KEIL.dll b) STLinkIIIUSBDriver.dll Thanks again for your help Regards, Rahul

[sjo]

Posted on May 17, 2011 at 13:19

My reply we in relation to openocd, as this was your original question?

Cheers sjo

[kr_rahul]

Posted on May 17, 2011 at 13:19

Dear sjo,

Thanks for your suggestion regarding the monitor command. I did try the your suggestion with regards to openOCD (connected to stm32 through amontec JTAG). I am using gdb client (yagarto) for driving openOCD. The global variable which I need to track is located at 0x200000e8 (in RAM). I was able to send monitor command at this address: (gdb) mon wp 0x200000e8 4 a 5 0 (gdb) c Continuing. Cannot watch data values From the console output, I felt that there were some issues in handling this command. I tried the above command on GDB client without mon, but that gave the following error: (gdb) wp 0x200000e8 4 a 5 0 Undefined command: ''wp''. Try ''help''. Previously, I tried placing a data watch-point using awatch command, However that too didn't cause the program to halt (in the case below I had to stop the program by a Ctrl-C (gdb) awatch loop_count Hardware access (read/write) watchpoint 1: loop_count (gdb) c Continuing. Program received signal SIGINT, Interrupt. From the transaction-logs, the RSP packet sent for the above command was: z4,200000e8,4 My open-ocd version is the following: Open On-Chip Debugger 0.1.0 (2009-01-21-21:15) Release Do you think the watchpoint feature would work with the version or should I shift to a newer version -thanks and regards, Rahul

[sjo]

Posted on May 17, 2011 at 13:19

Just had a quick look through the code, looks like it has not been implemented for cortex_m3 yet.

I will look into adding it. Cheers sjo