Raspberry Pi 4B & 5 baremetal blink example

[iCMDdev]

Currently just a draft for a baremetal Raspberry Pi 4B 64-bit blink example (-target aarch64-none-none-elf). I'll add the code soon - I'm making this draft public now to receive suggestions & feedback as I work on this example.

I already got Embedded Swift working on a 4B (blinking a led with Swift-MMIO), but I need to do some cleanup & better organize the code & build process. I'll upload it here after I do that (it will be in a Swift Package format).

I'm currently using llvm-objcopy as a last step to convert the resulted ELF file, with something like $(LLVMPATH)/llvm-objcopy -O binary kernel8.elf kernel8.img. From what I've seen, there doesn't seem to be any viable alternative, so users will need to install llvm with brew, since llvm-objcopy is not installed by default.

[iCMDdev]

This doesn't need to be necesarily specific to the 4B, I might as well adapt it to the 5B. Some register values might be different, especially because the 5's ACT (green) LED (and the whole GPIO) is not directly connected to the main processor, but via the RP1 chip, perhaps making it a bit harder to blink directly.

Still, one could blink another LED connected directly to the GPIO by changing the pin(currently hardcoded in Swift MMIO registers)

Edit: I think I'm partially wrong, It appears that the green (ACT) led is connacted to the always-on GPIO pin 9 (the BCM2712's GPIO) on the RPi 5 according to this.

Edit 2: Yes, the last edit is accurate. I got this working.

[iCMDdev]

That was faster than expected, I think it's ready for review now!

[iCMDdev]

boot.S and link.ld are borrowed from this bare-metal code example for Raspberry Pi 4 (CC0-licensed).

[iCMDdev]

I also got this working on a Pi 5! I managed to gather the relevant register details (adresses & necessary bit values).
Now, I'm wondering whether these should be 2 separate examples, or if I should merge them into one (and use compiler directives to specify which board is used).

[rauhul]

I think the example makes sense to add. However Im trying to get CI setup for this repo before adding additional examples, so avoid hosting broken code!

[iCMDdev]

However Im trying to get CI setup for this repo before adding additional examples

Make sense, alright!

so avoid hosting broken code!

I'm not sure what you mean by that, do you mean that this example is broken somewhere? I was able to build & run it just fine, but maybe that's on my machine only. Let me know if you had any problems if you tested this.

Either way, as you suggested, we should popstone this example and come back to it once we get CI integration. That's a good idea.

[rauhul]

I mean the examples silently breaking due to changes in the swift compiler or other example dependencies, e.g. the SDKs where we depend on a branch.

[iCMDdev]

I mean the examples silently breaking due to changes in the swift compiler or other example dependencies, e.g. the SDKs where we depend on a branch.

Oh, right, yeah. Great idea.

[rauhul]

@iCMDdev we now have some basic CI for the examples (excluding the current ones that only build for macho) Can you try to extend this CI to the new content in this PR?

[iCMDdev]

Sure! I think I'll do this over the weekend.

[iCMDdev]

I added the CI example, it should be working now. I'll now work on fixing the lint errors.

Sorry for the many commits, taking advantage of the squash-on-merge policy... 😅

[iCMDdev]

Seems to be fine now! I fixed the styling issues as well.

[iCMDdev]

Thanks for running the checks! I’m not sure how I missed installing make, I know I had it in the script… Maybe I pushed the change to the wrong branch.

Either way, I’ll fix it asap.

Edit: yes, I think that’s exactly what happened. Hopefully it’s fine now.