LPC11U35/401 support (OKdo E1)

[sszczep]

Hey,

I have recently purchased OKdo E1 boards with LPC11U35/401 HIC and LPC55S69 target. It is visible under lsusb as 1fc9:0132 NXP Semiconductors LPC11U3x CMSIS-DAP v1.0.7. Unfortunately, I couldn't get it to work with JLink nor pyOCD (I assume it's due to old CMSIS-DAP). The only thing I found working was MCUxpresso's crt_emu_cm_redlink tool. My goal is to use Zephyr and make west flash/west debug commands work. I guess DAPLink would make it possible, however it doesn't support this HIC due to missing RAM2 memory (#809).

I would try remapping the memory layout, however it is too small (m_data + m_data_2 exceeds 8KB):

    m_interrupts:          1 KB         1 KB    100.00%
          m_text:       54400 B        59 KB     90.04%
        m_cfgrom:          10 B         4 KB      0.24%
          m_data:        7124 B       7904 B     90.13%
        m_cfgram:         142 B        256 B     55.47%
        m_data_2:          2 KB         2 KB    100.00%
       m_usb_ram:          40 B         2 KB      1.95%
   text    data     bss     dec     hex filename
  55416    1052    8316   64784    fd10 build/lpc11u35_if.elf
  55416    1052    8316   64784    fd10 (TOTALS)

Can I somehow reduce memory usage/trim functionality to make it work? I most definitely do not need DnD and other things I'm not even aware of. If not, do I have other options rather than buying other board/soldering other HIC?

I'm sorry if I'm missing something, it's my first time using debug probes. Any help appreciated.

[mbrossard]

Hi @sszczep, I also have the OKdo E1. I have tried to fit DAPLink with the LPC55S69, but as you noticed the biggest challenge with supporting the LPC11U35/401 is losing that RAM2 region. I don't know how close I got, but it was not enough.

I have made an attempt to build DAPLink without DnD for the Okdo E1 in the feature/okdo_e1 branch. I have not tested it (I need to hunt the board), but removing the drive emulation and flashing functionality saved a lot of RAM and Flash.

Memory region         Used Size  Region Size  %age Used
    m_interrupts:          1 KB         1 KB    100.00%
          m_text:       28748 B        59 KB     47.58%
        m_cfgrom:          10 B         4 KB      0.24%
          m_data:        4400 B       7904 B     55.67%
        m_cfgram:         142 B        256 B     55.47%
       m_usb_ram:          40 B         2 KB      1.95%
   text    data     bss     dec     hex filename
  29764    1008    3592   34364    863c build/lpc11u35_okdo_e1_if.elf
  29764    1008    3592   34364    863c (TOTALS)

Let me know if this works or if you can get it to work with some changes. It will need some cleanups to be mergeable.

[sszczep]

Wow, haven't expected such a fast reply. So i flashed the CRC firmware, it gets detected in lsusb: 0d28:0204 NXP ARM mbed. It also gets picked up by pyocd:

sszczep@sebastian-kde-neon:~$ pyocd list
  #   Probe/Board             Unique ID                                      Target  
-------------------------------------------------------------------------------------
  0   Arm DAPLink CMSIS-DAP   0000124a094c00000000000000000000000097969902   n/a     
      Okdo                    E1                                          

Certain pyocd commands such as reset or erase do not work (0000694 E Transfer error while reading AHB-AP#1 ROM table: Memory transfer fault @ 0xe00fefd0-0xe00fefff [ap]), however, I can successfully flash the image:

sszczep@sebastian-kde-neon:~/zephyrproject$ source .venv/bin/activate
(.venv) sszczep@sebastian-kde-neon:~/zephyrproject$ cd zephyr
(.venv) sszczep@sebastian-kde-neon:~/zephyrproject/zephyr$ west flash -r pyocd
-- west flash: rebuilding
ninja: no work to do.
-- west flash: using runner pyocd
WARNING: runners.pyocd: hex file (None) does not exist; falling back on .bin (/home/sszczep/zephyrproject/zephyr/build/zephyr/zephyr.bin). Consider enabling CONFIG_BUILD_OUTPUT_HEX.
-- runners.pyocd: Flashing file: /home/sszczep/zephyrproject/zephyr/build/zephyr/zephyr.bin
0000915 I Loading /home/sszczep/zephyrproject/zephyr/build/zephyr/zephyr.bin at 0x10000000 [load_cmd]
[==================================================] 100%
0002345 I Erased 32768 bytes (1 sector), programmed 15872 bytes (31 pages), skipped 0 bytes (0 pages) at 10.84 kB/s [loader]

Could not get debug to run (no logs after flashing samples/hello_world, might be a misconfiguration on my side). Also JLink nor MCUxpresso detects the probe. Is JLink support something achievable? Great job nonetheless.

[mbrossard]

I was pretty aggressive and chose to use hic_lpc11u35_bulk that does not have the HID end-points. This can affect the compatibility with some tools. You can try to replace with hic_lpc11u35 or add records/usb/usb-hid.yaml (the LPC11U35 has a limited number of end-points, so we typically cannot have bulk and HID at the same time, but we can in this case since we don't have MSC).

I am wondering also if would make sense to add in source/board/okdo_e1.c:

    .board_id = "0236",

Which is the board ID for the LPC55S69-EVK board so some tools might be able to infer this a LPC55S69 target.

Another option would be adding .target_cfg = &target_device, to g_board_info and adding before it:

#include "target_config.h"

target_cfg_t target_device = {
    .version                        = kTargetConfigVersion,
    .target_vendor                  = "NXP",
    .target_part_number             = "LPC55S69JBD100",
};

@flit, any suggestions regarding pyocd issues?

[mbrossard]

I updated feature/okdo_e1, with some other RAM savings changes.

Memory region         Used Size  Region Size  %age Used
    m_interrupts:          1 KB         1 KB    100.00%
          m_text:       53232 B        59 KB     88.11%
        m_cfgrom:          10 B         4 KB      0.24%
          m_data:        7856 B       7904 B     99.39%
        m_cfgram:         142 B        256 B     55.47%
       m_usb_ram:          40 B         2 KB      1.95%
arm-none-eabi-size --totals build/lpc11u35_okdo_e1_if.elf
   text    data     bss     dec     hex filename
  54248    1344    6712   62304    f360 build/lpc11u35_okdo_e1_if.elf
  54248    1344    6712   62304    f360 (TOTALS)

This way I did not have to remove MSC and added the LPC55S69 target. Drag-n-drop is not working and there is an ASSERT.TXT, but pyocd picks up the target automatically:

  #   Probe/Board             Unique ID                                      Target
-----------------------------------------------------------------------------------------------
  0   Arm DAPLink CMSIS-DAP   0b06124913e600000000000000000000000097969902   ✔︎ lpc55s69jbd100
      Okdo                    E1

@flit, when I try to flash from the CMSIS Pack target, it fails:

(local) $ pyocd flash LPCXpresso55S69.hex
0000982 W Board ID 0b06 is not recognized [mbed_board]
0001849 I Loading LPCXpresso55S69.hex [load_cmd]
0001861 W Failed to add data chunk: memory region at address 0x00000000 is not writable [file_programmer]
0001861 I Erased 0 bytes (0 sectors), programmed 0 bytes (0 pages), skipped 0 bytes (0 pages) at 0.00 kB/s [loader]

But when I use the internal LPC55S69 target it works:

(local) $ pyocd flash -t lpc55s69 LPCXpresso55S69.hex
0000981 W Board ID 0b06 is not recognized [mbed_board]
0001343 I Loading LPCXpresso55S69.hex [load_cmd]
[==================================================] 100%
0002703 I Erased 32768 bytes (1 sector), programmed 14848 bytes (29 pages), skipped 0 bytes (0 pages) at 10.76 kB/s [loader]

[sszczep]

That's a massive update. Unfortunately I had to abandon my project for the time being, but I will make sure to check it out.

Thanks!

[sszczep]

Hey @mbrossard, I've tried building the new version, however I can't get it to work - m_data exceeds 8000b. Have you used different flags by any chance? I simply cloned your fork and changed the branch to feature/okdo_e1.

[mbrossard]

It was very close for me as well. I guess one possible explanation could I was using a newer GCC version?

$ arm-none-eabi-gcc -v
[...]
gcc version 10.3.1 20210621 (release) (GNU Arm Embedded Toolchain 10.3-2021.07)

[sszczep]

Yes, that's what I've come up with. After upgrading to the latest version (12.2 I guess?) the size shrinked significantly. I had issues with undefined references to functions such as _write, _read, but was able to compile the project disabling fatal warnings in linker. Pyocd detects the lpc55s69jbd100 so it seems to work. As you mentioned previously, I think it should be detected as lpc55s69 instead, not to manually specify the target. I couldn't get it to work with MCUxpresso but I think we are on the right track