RT1050: flash related issues

[flit]

Debugging a RAM-resident app on the NXP i.MX RT1050 MCU works successfully. However, programming flash and debugging an application residing in flash fails with various errors.

[AntoineZen]

Hi @flit & all,

Had also issue with the same target while flashing. The exception I have is ZeroDivisionError! I try to flash an example from NXP's SDK, nothing fancy.

Here is a sample session (with verbose flag):

$ pyocd flash --format elf led_blinky.elf -v
0000181:INFO:board:Target type is mimxrt1050_hyperflash
0000294:INFO:dap:DP IDR = 0x0bd11477
0000338:INFO:dap:AP#0 IDR = 0x04770041
0000368:INFO:rom_table:AP#0 ROM table #0 @ 0xe00fd000 (designer=00e part=88c)
0000390:INFO:rom_table:[0]<e00fe000:ROM class=1 designer=43b part=4c8>
0000390:INFO:rom_table:  AP#0 ROM table #1 @ 0xe00fe000 (designer=43b part=4c8)
0000413:INFO:rom_table:  [0]<e00ff000:ROM class=1 designer=43b part=4c7>
0000413:INFO:rom_table:    AP#0 ROM table #2 @ 0xe00ff000 (designer=43b part=4c7)
0000435:INFO:rom_table:    [0]<e000e000:SCS-M4 class=14 designer=43b part=00c>
0000451:INFO:rom_table:    [1]<e0001000:DWT class=14 designer=43b part=002>
0000466:INFO:rom_table:    [2]<e0002000:FPB class=14 designer=43b part=00e>
0000482:INFO:rom_table:    [3]<e0000000:ITM class=14 designer=43b part=001>
0000497:INFO:rom_table:  [1]<e0041000:ETM-M7 class=9 designer=43b part=975 devtype=13 archid=4a13 devid=0:0:0>
0000513:INFO:rom_table:  [2]<e0042000:CTI class=9 designer=43b part=906 devtype=14 archid=0000 devid=0:0:40800>
0000529:INFO:rom_table:[1]<e0040000:TPIU-M7 class=9 designer=43b part=9a9 devtype=11 archid=0000 devid=0:0:ca1>
0000544:INFO:rom_table:[2]<e0043000:TSGEN class=15 designer=43b part=101>
0000559:INFO:cortex_m:CPU core is Cortex-M7 r1p1
0000580:INFO:cortex_m:FPU present: FPv5-DP
0000597:INFO:dwt:4 hardware watchpoints
0000604:INFO:fpb:8 hardware breakpoints, 1 literal comparators
0000638:ERROR:__main__:uncaught exception: float division by zero
Traceback (most recent call last):
  File "/home/antoine/web-git/pyOCD/pyocd-venv/lib/python3.6/site-packages/pyocd/__main__.py", line 309, in run
    self._commands[self._args.cmd]()
  File "/home/antoine/web-git/pyOCD/pyocd-venv/lib/python3.6/site-packages/pyocd/__main__.py", line 381, in do_flash
    format=self._args.format)
  File "/home/antoine/web-git/pyOCD/pyocd-venv/lib/python3.6/site-packages/pyocd/flash/loader.py", line 135, in program
    self._loader.commit()
  File "/home/antoine/web-git/pyOCD/pyocd-venv/lib/python3.6/site-packages/pyocd/flash/loader.py", line 473, in commit
    ((totalByteCount/1024) / totalProgramTime),
ZeroDivisionError: float division by zero

Version :

$ pyocd --version
0.14.3

[flit]

@AntoineZen Thanks for the additional information!

[Hoohaha]

When will you plan to fix this issue? We really hope we can use pyocd to debug NXP RT boards.

[flit]

I'm working on it with NXP, but I'm afraid I don't have a date for you.

[AntoineZen]

I have tried you fix from #513. The Zero-division error is gone but now I'have several other exception in cascade (as you could except from what I read above). I have attached my stack-trace, if it can be useful to you.

pyocd_exception.txt

[flit]

Thanks. Unfortunately there's nothing revealing in the exception log.

[baudiffred]

I can confirm the same failure as the posted log on an i.MXRT1021 EVK board. I am seeing a failure of CMSIS transfer acknowledgement as the immediate cause of the exception.

OpenSDA 254 is running on this EVK which provides CMSIS-DAPv2 support. PyOCD is 0.22.

[danielRep]

I can confirm the same problem on an i.MXRT1020 EVK board. I am able to connect to the board and open a gdb session. When I try to "load" a FLASH resident program (led_blinky example) I have an output on pyocd terminal attesting the loading task (0019507:INFO:loader:Erased 16384 bytes (4 sectors), programmed 16384 bytes (4 pages), skipped 0 bytes (0 pages) at 8.26 kB/s). Nevertheless when I try to reset the board, the LED doesn't blink and I loose the connection to the board. To regain access I need to do a mass erase in Fuses boot mode and flash again through MCUXpressoIDE and in internal boot mode.

Any news from NXP @flit ? Thank you!

[flit]

Nothing really worth reporting yet, sorry. Recently I did get the RT1052 flash algo source code from NXP (unfortunately I'm not allowed to distribute it to others), but haven't had time to debug it. I did also get a RT1021 EVK recently, and have asked for the source for the RT1021 flash algo.

[danielRep]

Thank you for the quick feedback @flit ! I will keep track this issue to check any advances on this. If anyone has a Linux-based tool to flash the i.MXRT1020 EVK w/o depending on the MCUXpresso tools, please let me know.

[MateuszKlatecki]

@flit flash algo source codes are distributed with MCUXpresso in ide\Examples\Flashdrivers\NXP\iMXRT folder.

[flit]

@MateuszKlatecki Holy cow! Thanks for pointing that out. 😀

It's not quite the same as what I was given. This project builds a .cfx file (used by the MCUX IDE), while the one I was given builds an .FLM. The latter format is used by CMSIS-Packs and pyOCD. However, from a quick review it looks like they mostly have wrapped the .FLM API with the LPCXFlashDriverLib API.

[danielRep]

@flit , one last question.

Do you think that if I my binary is NOT FLASH resident, I could easily load the program to the board by using pyocd and a gdb a session?

[flit]

You mean RAM resident? Yes, that should work.

[Hoohaha]

Hi @flit , All NXP IMX-RT serires has same problem to load elf to flash. like EVK-MIMXRT1060, EVK-MIMXRT1015. It's good to see you get start on this issue.