[OpenOCD-user] Trouble programming flash memory on ATSAMD21 chip with STLink V2 interface

[Tony D. <to...@to...>]

Hi all, I'm running into a lot of trouble programming the flash memory
of an Atmel SAMD21 chip (specifically the ATSAMD21G18A, a 256kb flash
memory Cortex M0) and am curious if anyone has pointers or experience
programming this chip with an STLink V2 programmer & OpenOCD.

Here's a little bit about my setup:
- OpenOCD version: 0.9.0 (2015-06-30-17:02)
- Target: ATSAMD21G18A (same as in the Arduino Zero board), using the
stock at91samXX.cfg target config
- Interface: STLink V2, using the stock stlink-v2.cfg
- Protocol: SWD

The problem is that I can't get any of the flash memory programming
functions to work.  I can connect just fine to the chip, halt it, read
memory, do a full chip erase (with the 'at91samd chip-erase' command),
etc.  However I've tried all the flash programming commands I can find
and each fail in a slightly different but related way.

And just to answer an obvious question, yes I've checked that any
flash write protection and bootloader protection is disabled (using
the 'at91samd bootloader 0' command to disable bootloader
protection--the 'flash info 0' command also shows all 16 sections are
not protected too).

- flash write_image throws an error "Failed to erase row containing
00000000".  Digging into the debug output (with -d 3) I see the
problem is this:

Debug: 470 272275 target.c:2288 target_write_u16(): address:
0x41004000, value: 0x0000a502
Debug: 471 272275 hla_target.c:766 adapter_write_memory():
adapter_write_memory 0x41004000 2 1
Debug: 472 272276 stlink_usb.c:399 stlink_usb_error_check(): Verify error
Debug: 473 272276 target.c:2293 target_write_u16(): failed: -4

Crawling through the code a bit it looks like after issuing the erase
row command to the NVMEM controller on the chip the STLink is throwing
back a STLINK_JTAG_WRITE_VERIF_ERROR error.  I'm kind of at a loss at
this point as to what the problem could be--anyone hit this issue and
have suggestions?

- load_image executes but silently fails.  A blank line is output, but
examining the debug output I see it's hitting a timeout error:

Debug: 513 854015 target.c:1970 target_write_buffer(): writing buffer
of 6328 byte at 0x00000000
Debug: 514 854015 hla_target.c:766 adapter_write_memory():
adapter_write_memory 0x00000000 4 1582
Debug: 515 854027 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 516 854042 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 517 854080 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 518 854098 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 519 854142 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 520 854172 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 521 854239 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 522 854318 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 523 854481 stlink_usb.c:393 stlink_usb_error_check(): wait
status SWD_DP_WAIT (0x10)
Debug: 524 854481 command.c:628 run_command(): Command failed with error code -5

Looking at the code it looks like there's a minor bug in OpenOCD error
logging here--the stlink_usb.c function stlink_usb_write_mem can fail
without logging an error if it exceeds the number of retries from a
ERROR_WAIT response (this code:
https://github.com/ntfreak/openocd/blob/master/src/jtag/drivers/stlink_usb.c#L1509-L1514).
This goes against the assumption in the comments here that commands
will print out an error if they fail:
https://github.com/ntfreak/openocd/blob/master/src/helper/command.c#L624-L629
 This is just a minor issue though and not related to my problem--I
just wanted to raise it in case anyone else has noticed commands
silently failing.

Back to the load_image problem it appears I'm hitting the timeout
while writing to the chip.  Has anyone run into this issue and have
any workarounds?  From looking at the code it looks like it's
hardcoded to a max of 8 retries over the course of about 0.25 seconds.

- Finally the program command fails with the same "Failed to erase row
containing 00000000" error as flash write_image above.  I assume
internally the program script is just calling flash write_image so
it's the same JTAG verify error.

Unfortunately I'm really stumped why I can't seem to get OpenOCD to
program this chip using a STLink V2 programmer.  I can program the
same chip just fine with a Segger J-Link and their JLink tools so it
doesn't appear to be a hardware or other physical issue.  I've also
used this same STLink V2 to program a few other chips from Nordic,
STM, etc. without issue so it should be fine.  And as mentioned
earlier OpenOCD can connect and control the chip just fine, it's only
the flash memory writes that fail.

Anyone have any pointers or idea what could be going wrong trying to
flash the ATSAMD21 with the STLink?  I'd really like to be able to use
the STLink V2 programmer here but I'm kind of hitting a brick wall
with the JTAG verify error & hard coded timeout.

Also let me know if the user list isn't appropriate for this issue and
if I should move it to the developer list instead.  Thanks!

-Tony