Contributed by pitrh on from the what, octopus? dept.
Next up in the c2k15 reports series is Paul Irofti (pirofti@), who writes:
I arrived in Calgary with two projects in mind: writing an ACPI WMI driver and adding Flash driver support for the Octeon D-Link DSR-500 machine. This was obviously naive of me as I have not even started working on the former and barely managed to finish the later.
A good thing is that the resulting amdcf(4) driver was built following the Common Flash Interface (CFI) specification which will allow in the future to have out of the box support for any Flash memory device that OpenBSD might encounter. This is often the case on embedded systems such as MIPS and ARM.
For now the driver is enabled in read-only mode and kept private in the Octeon tree. Plans are to make it generic enough so that it can be relocated to the machine-independent part of the kernel.
There have been talks about turning it into a fake SCSI driver following the semantics set by David Gwynne (dlg@) but that is not on top of my priority list. Allowing the driver to speak CF+, which would expose the device as an ATA(IDE) disk, would probably favour this change. In the end this might turn out to be unfeasible due to the differences between CommonFlash and CompactFlash devices. Time will tell.
As my first disk driver I must say I had very much fun writing it. The peak moment was when I was first able to dump the contents of the puny 32M flash found on the DSR-500: I could see U-Boot and Linux ELF binaries gently flowing through the hexdump output.
In between kernel builds I reviewed and discussed Octeon diffs with miod@, jasper@ and visa@, updated ports like youtube-dl and postgresql-odbc and checked on ajacoutot@ that was nice enough to do all the hard work involving updating SuiteSparse for me.
I want to thank Nayden Markatchev, the OpenBSD Foundation and Theo de Raadt for making this event possible! It was great!
And thank you for the work and the report, Paul!
(Comments are closed)