RISCV issues

[MabezDev]

Finally getting around to trying out debugging with probe-rs! It was pretty easy to setup, however I'm running into a few issues.

probe-rs is correctly stopping on breakpoints (see bottom left) but the UI is not updating, nor is it filling the variables table.

[Yatekii]

cc @noppej

[noppej]

Hi Scott,

I'm sorry that you're running into problems with this. The UI is actually updating (visible by fact that the Call Stack view reports a breakpoint, and that Variable view reports "No variable information available". I realize this isn't useful information, but it does help to be precise in helping describe the potential source of the problem.

The fact that the Call Stack reports it has halted on an <unknown function> versus your atomic_emulation_trap suggests that the probe-rs stack trace is not returning the information we would expect.

As far as I know, this has never been tested with RISCV based boards, so it will probably require some debugging effort in the probe-rs lib.

I'd be willing to help but will need to acquire a similar dev board. What chip/board are you using?

[MabezDev]

Hi Jack,

Thanks for the fast reponse! I'm happy to help in anyway I can, I understand I'm often venturing into uncharted territory with the riscv stuff :D.

I'd be willing to help but will need to acquire a similar dev board. What chip/board are you using?

I'm using a esp32c3, specifically its this board.

[noppej]

@Yatekii and @Tiwalun or @Dirbaio ... do any of you have experience with these boards? Am I barking up the wrong tree thinking I can get one of these and find out why the debug isn't working? Or do you know of foundational stuff missing in the probe-rs lib that would prevent support?

I am just looking for a quick sanity check and whatever pointers your experience might suggest would be useful to proceed.

[noppej]

@MabezDev In the meantime, can you please try the following two steps:

1. In your launch.json, add the option "wire_protocol":"Jtag"
2. In your launch.json, add the option "console_log_level": "Debug", then send me a copy of the output generated in your VSCode Debug Console (I suggest start with a clean console, then connect, and as soon as the device stops at the first breakpoint, end the debug session, then send the output.

[MabezDev]

FYI I'm working around another bug, where "wire_protocol":"Jtag" is not applied (atleast not to my Jlink). It defaults to SWD regardless of the value of wire_protocol. Currently hardcoding it to use Jtag in the code.

Here is the trace you asked for:

probe-rs-debug-trace-esp32c3.txt

[MabezDev]

Seems to be having trouble unwinding the stack frames from what I can tell (StackFrameIterator is returning None), here's the elf file I'm flashing in case that helps.

[noppej]

The elf file is a good resource ... I will dig through it and get back to you.

[noppej]

@MabezDev ... can you please send me a copy of your Cargo.toml as well as the instructions you use to build the binary?

[MabezDev]

Here is the repo: https://github.com/MabezDev/esp32c3-experiments that includes instructions.

[noppej]

@MabezDev ... FYI, I have ordered one of these boards, but none were available in Canada, so I'm waiting on international shipment. Will update this once it arrives and I can reproduce the issue.

[Tiwalun]

@Yatekii and @Tiwalun or @Dirbaio ... do any of you have experience with these boards? Am I barking up the wrong tree thinking I can get one of these and find out why the debug isn't working? Or do you know of foundational stuff missing in the probe-rs lib that would prevent support?

I'm pretty sure the debugger / unwinding code in probe-rs has some hardcoded assumptions about ARM / Cortex-M right now, especially the stack unwinding code. So I would assume that it doesn't work properly on any RISCV chips, and not that it's related to the specific chip.

[noppej]

@Tiwalun ... Thanks for this. While working on the debugger, I got quite familiar with this part of the code, so I'd be happy to help make it more generic. Do you have much experience with RISCV boards? Can you recommend a 'better' (more available and more widely used) one for me to invest in?

[Tiwalun]

I've used the following ones, but not sure if they are 'better':

• HiFive1 Rev B
• Pinecone BL602
• Sipeed Longan Nano

And availability is probably a general problem right now...

[MabezDev]

I've spotted one issue and logged it under the main repo probe-rs/probe-rs#851.