0x0 ~0x7FF (1st FLASH sector) is always erased, debugging with USBDM

[madisonchou]

Hi,

I am porting CW10 project to KDS and encounter difficulties when debugging with latest USBDM (V4.12.1.30). Our target is MK10DX128M7.

I have project reside on FlexNVM with interrupt vectors in 0x0~0x400 just like any normal project.   Then, I have another project that is in FLASH, but arranged to avoid using 0x0 ~ 0x7ff (1st FLASH sector) by the linker file:

  m_interrupts          (RX)  : ORIGIN = 0x00001000, LENGTH = 0x00000400

  m_flash_config        (RX)  : ORIGIN = 0x00001400, LENGTH = 0x00000010

  m_text                (RX)  : ORIGIN = 0x00001410, LENGTH = 0x0001EBF0

This arrangement works in CW V10 using USBDM.   But doesn't work in KDS using USBDM.   After code is downloaded, it is obvious that problem is due to 0x0 ~ 0x7FF being erased, even I have selected "Selective Erase (by sector)".

Is there a work-around?   Thanks!

Jei

[pgo]

Hi Jei,

The sector size is set incorrectly in the device file (arm_devices.xml).

Please try the attached file.

bye

[madisonchou]

Hi pgo,

The new device data file doesn't seem to make any difference.   After code download, 0x0 ~0x7FF remains erased.

One more piece of info.  The "ARM Programmer" is now good.  Though, I am not 100% sure because I wasn't as careful as today.  I did this yesterday and today:

Set linker to generate S-record output.   Use "ARM Programmer" to program the my target.  First, mass-erase the target.   Select EraseSelective.   Then, open a S-record of 1st project that uses the 0x0 region.  Programming is successful.   Open S-record of 2nd project that doesn't use 0x0 region.   Programming is successful.   Use "USBDM Memory Dump" and verified the bits in 0x0~0x7FF are intact, today.   (But I believe it was erased when I did the same yesterday.)

Jei

[pgo · Accepted solution]

Hi Jei,

Using the working copy of USBDM which is slightly later than the release version I was able to do what I believe you were trying but there are several caveats.

• If you are not programming the security area at 0x400 then is necessary that its present contents unsecure the chip - by default the USBDM interface will program and hence erase that area even if the image does not contain data.  You have to disable this. (see below)/
• I found it necessary to modify the standard start-up code to set the stack pointer as this will no longer be loaded from the reset vector correctly on reset.

The setup I used for the debugger is as follows:

• Change to selective erase.
• Change to using security field from image (there should NOT be one).
• Don't reset the target
• Do set the initial PC to whatever the reset handler is called.

You may have to change the above Reset_Handler depending on your code.

The example changes to the KDS startup code should be similar to:

I tested this on a MK20DX128M5 as follows:

• Programmed the entire flash image with random data (except for the security area which has FFFFFFFFFEFF etc to unsecure)
• Ran a program with the rom relocated to 0x1000.
• The debugger appear to work correctly.
• I used the memory window to check the rom contents.  They appear unchanged except for the region starting at 0x1000.
• I used the USBDM memory dump program and examined the memory image.  It appeared as expected.

I believe the above should work with the current release version with the modified device file provided earlier.  If not it would be best to wait a few days for Version 4.11.1.50 to be released.

bye

[madisonchou]

Hi pgo,

Thanks for you time.   Security Option is the trick!!

But my project runs to never land in the debugger.  So something is still wrong.   Though, that will be a different thread after I characterize it.

Thanks again!!!

Jei