As I work on some ports for OpenBSD, one of the issues that has been mentally nagging at me is the idea that I might have a great AMD64 port but it won’t actually work on other processor architectures. To me, one of the big draws of OpenBSD is the fact that you can take hardware that others might consider “obsolete”, put a modern operating system on it and then have a useful machine again.
With this in mind, I decided to see what I could do about getting some “other” processors into my stable.
Right off the bat, I realized I had an old G4 Mac Mini in a closet that would make a nice start. I installed 5.7 from my release media (anyone running OpenBSD should buy the media to support the project – even if you don’t use it the stickers are always cool <g>). I immediately ran into problems because the approach I took to partitioning was not creating the MSDOS partition with the bootloader. Turns out that the path of least resistance is to edit the default partition scheme and go from there. The bad idea (that I kept bumping my head into the wall over) was to either start from scratch or immediately do a “d *” from disklabel in the installer.
Once I had my PowerPC 32-bit system up and running, I realized I had another processor architecture I could try. Sitting in that same closet was an old Thinkpad T21. I pulled it out, spent 20 minutes looking for the power supply and was able to load the system pretty quickly. Now I have three processor architectures to compile and test on natively.
Well, those were some pretty tame ones. I wanted to try my hand at something more exotic so I read the docs carefully, looking for ideal machines that would be powerful enough to be useful, wouldn’t take up a lot of space and would also be affordable on the second-hand market. I found a Sparc64 based machine, the Sun Blade 100. This is not a “blade” server but instead is a workstation with a funny name. I found one for less than $30 on eBay and went for it.
Setting up OpenBSD on this hardware was surprisingly simple. I read through the documentation and then followed it pretty carefully. It was surprisingly similar to setting up a PowerPC machine and before too much longer I had it up and running.
I’m eyeing an SGI machine and possibly an Alpha. It seems that the PA-RISC machines are surprisingly expensive for even older boxes. Apparently the resale market for them is much stronger or the supply is more constrained. I’m not sure there.
At the end of the day, building and testing on these other platforms is useful to the community and the “exotic” nature of them in today’s homogeneous computing landscape lends a little spice to working on things. I’m glad my path took this turn.