RP2040

[JohnK1987]

Hello,

we already discussed (Mbed's Forum) about RP2040 and unfinished PR - ARMmbed#14488
I already check that and after few corrections it is compilable, but cannot verify that because a standart output is .hex or .bin and this not enough and a format .uf2 is required.
I found all necessary things are already in PicoSDK only MinGW-w64_GCC_Tools must be installed, but I have no idea how to set this up together with MbedCE in VS Code.

Do you have any ideas guys? @JojoS62 @multiplemonomials

BR, Jan

[JojoS62]

It is interesting, although the official Arduino Core uses Mbed and so you can use Mbed classes.
For a good integration into Mbed CE it needs a new upload method for uf2. The SDK tool is running fine under Linux or in WSL, but I have also struggled with this in Windows. It could be easier when there is a python module for handling uf2.
Another tool is BMP, black magic probe and has an integrated gdbserver and good cortex-debug support. Afaik this works now also with RP2040.

[JohnK1987]

When you compare ArduinoCore-mbed/targets/TARGET_RASPBERRYPI/TARGET_RP2040 with the content of original PR it is little bit different (check the target folder in the tree) - https://github.dev/ARMmbed/mbed-os/pull/14488. Also latest ArduinoCore-mbed probably uses newer version of PicoSDK.
Moreover https://github.com/arduino/ArduinoCore-mbed/tree/master/patches

I do not have BMP but PicoProbe seems be also promising. Ofc yeah, it is just for RP.
PicoProbe:

• openOCD - https://github.com/openocd-org/openocd/releases/tag/v0.12.0-rc1
• pyOCD - https://pyocd.io/docs/debug_probes.html

BR, Jan

[multiplemonomials]

Alright, I got to the point where I can run the test suite on it!

79% tests passed, 15 tests failed out of 71

Total Test time (real) = 1920.33 sec

The following tests did not run:
          8 - mbed-hal-crc (Skipped)
         11 - mbed-hal-lp-ticker (Skipped)
         13 - mbed-hal-mpu (Skipped)
         14 - mbed-hal-ospi (Skipped)
         16 - mbed-hal-qspi (Skipped)
         22 - mbed-hal-sleep (Skipped)
         23 - mbed-hal-sleep-manager (Skipped)
         24 - mbed-hal-sleep-manager-racecondition (Skipped)
         27 - mbed-hal-trng (Skipped)
         32 - mbed-hal-verify-arduino-pinmap (Skipped)
         43 - mbed-platform-crash-reporting (Skipped)
         53 - mbed-platform-stats-cpu (Skipped)
         54 - mbed-platform-stats-heap (Skipped)
         55 - mbed-platform-stats-sys (Skipped)
         71 - mbed-connectivity-nfc-eeprom (Skipped)

The following tests FAILED:
          2 - mbed-usb-device-basic (Failed)
          3 - mbed-usb-device-hid (Failed)
          4 - mbed-usb-device-msd (Failed)
          5 - mbed-usb-device-serial (Failed)
          6 - mbed-hal-common-tickers (Failed)
          7 - mbed-hal-common-tickers-freq (Failed)
         10 - mbed-hal-flash-functional-tests (Failed)
         15 - mbed-hal-pinmap (Failed)
         17 - mbed-hal-reset-reason (Failed)
         18 - mbed-hal-rtc (Failed)
         19 - mbed-hal-rtc-reset (Failed)
         20 - mbed-hal-rtc-time (Failed)
         25 - mbed-hal-stack-size-unification (Failed)
         29 - mbed-hal-watchdog (Failed)
         31 - mbed-hal-watchdog-timing (Failed)

Looks like USB is a bit of a dumpster fire at present, every test fails or hangs. Also, the watchdog and RTC need help. (the stack-size-unification test was due to a missing constant, I just fixed that one). But all in all, not a bad start! Note that this test run was using the default serial, not BufferedSerial

[multiplemonomials]

My branch: https://github.com/mbed-ce/mbed-os/tree/RP2040

[multiplemonomials]

Good news! After a bit of debugging, I was able to fix the BufferedSerial issue. Turns out that the problem was pretty simple -- no one had implemented the ability to set and use Tx IRQs in the serial_api driver. The code was broken in such a way that it ended up delivering Rx IRQs when it meant to send Tx IRQs. This, of course, caused BufferedSerial to not work as it would get stuck in an infinite ISR handler after sending the first few bytes. I implemented the missing code and now it works -- check my branch above!

Also, I added Picotool upload method support and fleshed out the wiki page a lot.

[JohnK1987]

I know that was wrong but I was not able to find how to fix this. Good job.
I will try to fill some more info about Picoprobe into wiki page.

[JojoS62]

I've tried the RP2040 branch now also and first steps worked, a simple test program with EventQueue and Threads. Using the Picoprobe and OpenOCD under Windows.
There are a lot of warnings during compiling, signedness and missing return value, will check it again.
Great work!

[JojoS62]

that is interesting. There are already H7 multicore MCUs, how are they handled? Loading two independent projects?

PlatformIO comes with a modified version of OOCD for the pico, But you're right, it will be better to wait for an official new release.

[JohnK1987]

It seems like that - https://forums.mbed.com/t/mbed-os-new-features/11248/17

[JohnK1987]

@JojoS62 can you help me please? I am trying build BMP for ST-Link V2.1 according to this failed

[JojoS62]

I can check tomorrow. Can you build under Linux or in WSL? I remember that the build process was not Windows friendly.

[JojoS62]

still have problems with my eyes and cannot stare for long time at the PC monitor. There is a Discord channel #blackmagic where you get good support: https://discord.gg/CFDxXKZW
The instructions for compiling are here: https://black-magic.org/hacking/hacking.html
But I don't know if they are up-to-date. There was a lot of revamping in the project in the past.
When you succeeded with flashing the probe, you should try first the manual use of gdb:
https://black-magic.org/usage/gdb-commands.html
Then create a debug configuration in VSC for cortex-debug, with servertype bmp. The only settings should be path to .elf file and the com port for the probe.

[JohnK1987]

Jojo you always create new conversation, because you select wrong Write window/line probably. Try to use window with Reply button not with Comment button :)
Oh, understand, then don't strain your eyes for stupid things. I will left this for some time, maybe forever because this is really pain, I do not understand this. I believe I made some mistakes somewhere, but... I do not want to know how solve every problem on internet :)

Just a summary.
I just wanted try the product BMP on ST-link. According to https://black-magic.org/hacking/hacking.html

C:\Users\User\BMP\blackmagic>make PROBE_HOST=stlink ST_BOOTLOADER=1
! was unexpected at this time.
make: *** [all] Error 255

I did it under Windows, because I do not see any restriction for Linux ONLY!
According to your recommendation I installed WSL. Inside of WSL other things (Python, GNU...) what are already in my PC (waste of space). Different error, but result same.

root@DESKTOP-SBJJCIP:/home/johnnyk/Test/blackmagic# make PROBE_HOST=stlink ST_BOOTLOADER=1
  BUILD   lib/stm32/f1
make[2]: Nothing to be done for 'all'.
  BUILD   lib/stm32/f4
make[2]: Nothing to be done for 'all'.
  BUILD   lib/lm4f
make[2]: Nothing to be done for 'all'.
Load address 0x08004000 for original ST-LinkV2 Bootloader
  LD      blackmagic.elf
/usr/lib/gcc/arm-none-eabi/10.3.1/../../../arm-none-eabi/bin/ld: address 0x8020878 of blackmagic.elf section `.text' is not within region `rom'
/usr/lib/gcc/arm-none-eabi/10.3.1/../../../arm-none-eabi/bin/ld: blackmagic.elf section `.ARM.exidx' will not fit in region `rom'
/usr/lib/gcc/arm-none-eabi/10.3.1/../../../arm-none-eabi/bin/ld: address 0x8020878 of blackmagic.elf section `.text' is not within region `rom'
/usr/lib/gcc/arm-none-eabi/10.3.1/../../../arm-none-eabi/bin/ld: region `rom' overflowed by 2464 bytes
Memory region         Used Size  Region Size  %age Used
             rom:      133536 B       128 KB    101.88%
             ram:       11604 B        20 KB     56.66%
collect2: error: ld returned 1 exit status
make[1]: *** [Makefile:120: blackmagic.elf] Error 1
make: *** [Makefile:17: all] Error 2

Then, I found this - blackmagic-debug/blackmagic#1368 (reply in thread)
So I deleted some files (corresponding to non ST targets) from blackmagic\src\Makefile

Memory region         Used Size  Region Size  %age Used
             rom:      123144 B       128 KB     93.95%
             ram:       11920 B        20 KB     58.20%
  OBJCOPY blackmagic.bin

After hours I had binary (probably for a ST-Link) and I noticed I need another tool stlink-tool, again need to be builded, again it is not working. This time because of libusb.h: No such file or directory. I tried to solve this (I already have it installed in WSL), but I am done.

Eh... they want 68 dolars for their probe (Official hardware), when they are not able to make simple instructions for their product? No, thank you, nothing for me :)

[JohnK1987]

@multiplemonomials I added another info for the picoprobe to wiki page, hope it will be useful.

[JojoS62]

I have build the blinky for the tufty 2040 board from Pimoroni, it works. Used picotool upload in WSL2.
I had started a project using this target with PlatformIO, that was a little bit easier, but using the Earl Phil Power core. However, that is pretty fast with SPI DMA and PIO from the Bodmer/tft_eSPI Lib. So the next is to get this running also with Mbed.

[multiplemonomials]

Awesome! Yeah it would pretty awesome if SPI DMA support could be a thing for RPi Pico!

[JohnK1987]

I have Arduino nano rp2040 connect on my table, but no space in calendar :)