Contributed by jj on from the hunt for red Octeon dept.
Unlike previous hackathons I didn't have a clear idea what I wanted to work on, apart from "some ports" and "some octeon" stuff. Which as I came to learn is much better than having a fixed set of tasks...since you'll almost never do just that what you thought you would work on. There's always something more interesting that pops up.
After spending a few days in Toronto with pirofti@ the hackathon began on Wednesday. The first two/three days I spent mostly on ports where I worked through a backlog of outstanding updates and fixes I had in my tree. This included fixing all the Lua ports we have in tree to build and work correctly with Lua 5.2.
Of course, no hackathon would be complete without touching pkg-config for at least a short while. I ended up fixing a small buglet in dealing with corrupt files as well as fixing an issue in the parser which matthieu@ discovered with new freetype.
After finishing up the installer for Octeon some time ago I hadn't had the time to get back to Octeon, thus this hackathon provided a good opportunity to do so again. Even more so with the presence of uebayasi@ who was great to talk to and discuss the Octeon port.
After uebayasi@ brought up the large number of XXX's in the code I decided to start cleaning up some of them and I ended up cleaning some header files by removing backward compat and other unneeded cruft. As well as by making the use of sync instructions consistent across the codebase.
Together with pirofti@ I found an issue in the CPU/core enumeration code for Octeon when running an MP kernel, which would prevent the right number of cores from attaching. With that fixed bsd.mp boots, but hangs when going to userland. As this cnMIPS64 platform was designed with multiple cores in mind, it'd be a shame not to make full use of this. To be continued...
I also started working on a standalone bootloader for Octeon, as the current way of loading a kernel is "sub optimal". The installer places a kernel into the MSDOS partition, which U-Boot then directly loads. This works, but we lose the ability to properly pass arguments to the kernel for example. The work on bootloader is not complete, so finishing it will probably be material for another hackathon report.
Finally a big thank you to krw@ and mtu@ and those involved for this hackathon in awesome Toronto.
(Comments are closed)