Version 4.3 troubles

[leb]

I've been happily coding along w JM60 (CQH in my case) firmware 4.0.  Then I built a board with the MCF52211, that was my mistake.  Legacy Codewarrior reports v4.0 firmware as stale.  So, I built and installed v4.3 (using Legacy Codewarrior 6.3) which made Codewarrior 7.2 happy.  Lots of errors testing the board with the diagnostics, but that's not the issue here.  My problem is that firmware 4.3 broke my ability to work with CFv1 targets.  Debugger reports "Cannot read memory".  Using tryUSBDM utility I find that if I attempt to read more than 0xC0 bytes using a RBLOCK command I get a USB_TRANSFER_ERROR. (specifically RBLOCK 800000 C0 succeeds, RBLOCK 800000 C4 fails).  Am I alone with this problem?

[pgo]

Dear leb,

I've been unable to reproduce the problem you have described.

I did the following:

Changed the target firmware of a USBDM_JM60_CF BDM to V 4.3 using the bootloader.  These were the compiled files posted on Sourceforge.

Ran the following script using testUSBDM (from V4.3):

settarget 3connectgsrblock 800000 c0rblock 800000 c4rblock 800000 140
exit

The target device was an almost bare MCF51CN128CLK.

All three reads were successful without any problems.

The USBDM configuration dialogue reports BDM Firmware Ver 4.3, DLL Ver 4.3.0-Debug.

The log file in %APPSDATA%/usbdm/usbdm.log had no errors.

These tests were done under Win-XP.

The error you describe is suggestive of there being a incompatibility between the DLLs and the BDM firmware.  It is possible that the V4.3 firmware source posted may be inconsistent with the DLL posted.  Is it possible to use the pre-packaged firmware files?

Finally V4.4a is a more thorougly tested version but the BDM firmware source is not available.

bye

[leb]

Thought you would like to know that using FC I discovered that the command buffer in V4.0 was 168 bytes, and in V4.3 was 254 bytes.  Dialing the command buffer size in V4.3 back to 168 bytes cured the problem.  The V4.3 firmware now works fine with Codewarrior v6.3 and CFv1 targets.  I haven't tested with CW7.2 yet, I hope I'm not set up for an unpleasant surprise there.  I'm using a MC9S08JM60CQH chip and the legacy Codewarrior to build the firmware.  Is it possible that when I created the .mcp for the project I missed a setting that made memory useage less efficient than the CWv10 project (structure alignment, etc)? 

[pgo]

Sorry leb we cross posted.

I have been building the firmware using Codewarrior (Eclipse) V10 or 10.1.

There was an increase in the buffer size for JM32/60 targets which improves the large memory read/writes especially for the JTAG based targets.  The DLLs now query the BDM for the transfer size so should adapt.

Compare the Project_JM_CF.prm file posted in the sources against the one you are using.  I am not aware of any settings that would cause the software to fail without the build failing as well.

The JM60 chip actually has more than enough RAM.  The buffer size is limited by the 1 byte USB transfer size.

bye