Logbook

Framework laptop with a RISC-V main board

File under "interesting, but not likely to purchase in this go-around".

Framework makes nice, modular laptops. They are modular enough that you can swap out the main board, keep everything else, and keep on trucking. That's a good thing. Upgrades become a component swap rather than a complete new system.

DeepComputing has announced a RISC-V main board for the Framework Laptop 13, based around the StarFive JH7110 SOC. The good news is that you should have a fully functional laptop when you're done installing it. The bad news is that this SOC is roughly Raspberry Pi 3 performance, and is based on an earlier version of the RISC-V spec. (But if you wanted laptop performance you'd put in one of the perfectly serviceable AMD or Intel main boards.)

No word on pricing and availability just yet.

https://deepcomputing.io/a-risc-v-world-first-independently-developed-risc-v-mainboard-for-a-framework-laptop-from-deepcomputing/

It will be good for the RISC-V ecosystem when some people try to start to use this as a "daily driver". If you are careful about your software choices, and prefer simple monochrome text to lots of fancy graphics, you'll probably be happier. The Framework design is a good match for a development environment, but four slow cores are not going to run all of the compile and test cycles you might want.

Learning complex systems, piecemeal - this time lima, limactl, Fedora and RISC-V

The pattern runs something like this.

1. Identify a shiny, new, and ferociously complex system that you'd like to know more about.

2. Bumble your way through the initial install, taking copious notes on things that will have changed by the next time you get around to look at them.

3. Declare victory after some trivial task beyond the initial install is complete.

This is an anti-pattern in a lot of ways. Perhaps the ferociously complex system can be safely ignored for a week or a year, rather than needing attention now. If it does merit attention, then give it a proper fully automated install rather than a haphazard try. Note that this might take a few uninstall cycles, that would be fine.

As an example, I have Rancher Desktop running on my MacBook Air M1. It seems to idle just fine, taking up a bunch of memory and a couple of cores but not getting in the way of actual use of the rest of the machine. I've installed exactly one application on it (my "sisyphus" function for OpenFaaS), and even that is less than ideal because the container it's running in isn't up to date. There's some complexity in there that I just barely remember about nerdctl.

And yet - and yet - here I go again with the same pattern. Gee, it would be interesting to have a Fedora version running Tailscale SSH on my M1 Mac. I'll bet I can run that under something - "lima" it is.

limactl start --name=fedora template://fedora

Try it out, run into something of course, a fresh issue at

https://github.com/tailscale/tailscale/issues/4908

in which Fedora Server at Fedora 36 doesn't want to run Tailscale SSH because of SELinux, but there's a workaround of sorts there. And with the workaround, I can ssh into an arm64 Fedora VM on my M1.

But wait, there's more. Armed with that knowledge, and looking at the limactl templates, I see there's a "riscv64" template that turns up a VM under emulation. A failed install later, notice that the Ubuntu distribution has moved a file, and we're up (and also have Tailscale SSH up again). Notes at

https://github.com/lima-vm/lima/issues/930

Now not just one but two different VMs running, plus a third in which to run Kubernetes. An embarrassment of riches (or perhaps just an embarrassment of idle cycles). Two more diverse systems for my tailnet, and two more chips away at a complex system to understand just a little bit better.