Tyson D Sawyer <[EMAIL PROTECTED]> writes: > Its going to take me a bit to get back up to speed on linuxbios and there may be > > things I'm missing, but here is an observation: > > The bios code itself seems to be getting pretty good. However, the bits around > the edges such as kernel patches and elf stuff is quite chaotic.
Agreed. I'm starting to focus on this stuff now. > Kernel patches are not maintained in useful chunks such as IDE spin up. I had > hopes of using an elfimage with a built in ramdisk image (with romfs) or > something like that, but inspite of recent postings for elf patches and the > like, I haven't figured out where to begin or how it works. What do I do with a > > kernel that has the latest patch applied to it? How do I use it. The place to start is with mkelfImage with an unpatched kernel. That gets you ramdisk support. The trivial ide spin up is to include a kernel command like: "hda=0,0,0 hdc=cdrom" Which just tells the kernel something is there so it attempts to spin it up. At least with recent 2.4.x kernels this works. So you can do this patch free. > Please don't take this as bitching. However, since its been about a year since > I last worked on linuxbios I have a bit of an outsiders view again and I see > some things that keep this from going prime-time in any way. Feedback is good, and very much appreciated. > Unless someone is feeling suffciently bored to organize elf image support and > kernel patches, I will probably just get my stuff updated but do it the way I > had been which was to have the bios load a compressed ram disk and have the bios > uncompress the kernel and jump to it the same as was done a year ago. That stuff is being organized, but I'm still at the chaotic fringe. The biggest problem with a lot of the kernel stuff is that we haven't had a way to say if chipset do this and if this board do that. With the LinuxBIOS tables we now have a way to pass all of that information so if need be the kernel can decide what to do at runtime. So we start making patches with everything. So far my patches for native LinuxBIOS support are mostly proof of concept patches, and starting places. But it shouldn't be too hard to go the rest of the way from there. I have just started the process to get LinuxBIOS support in the mainstream kernel. 2.4.x for a while will probably be a backport but... So on the stability side I would say that the LinuxBIOS core is very close to being in the final stretch. And for what I'm doing with clusters I'm certainly pushing the envelope of acceptance. But right now I believe we have all the big pieces and it is just a matter of dotting the i's and crossing the t's. > If anyone is interested in knowing about some places linuxbios has been, it went > into the WTC in NY on about Oct 12 and the following days. The iRobot Packbot > (see http://www.irobot.com) is the project I am working on. It contains a > custom PC running linux booted with linuxbios. We didn't find any survivors, > but several Packbots where among the robots sent into places where people didn't > want to go. Very Cool. Eric
