Matthew Burgess wrote: > On Thu, 17 Sep 2009 16:18:16 -0500, Bruce Dubbs > <bruce.du...@gmail.com> wrote: > >> #2033 Create initramfs >> >> I think this one should be closed as won't fix. The reason for a >> initrd is for those systems that don't know in advance which >> drivers are needed for accessing the boot files.
That's not why Alexander originally filed it, though. He originally filed it so we could handle machines that used crappy software RAID cards that need dmraid. You can't autoconfigure dmraid in the kernel, considering it's a purely userspace tool... :-) >> That really isn't an issue for LFS where you build a custom Linux >> kernel. Add to that the ability of GRUB2 to access many systems >> such as LVM directly. Overall it's just not needed for LFS. GRUB2 can do LVM and maybe MD RAID, right? (Been a while since I looked.) It can probably half-do dmraid, in that a RAID1 mirror will boot up using only one half. I doubt it'll do a dmraid RAID0 setup. (Not that anyone in their right mind should use RAID0, but...) Can it do dm-crypt? (Yeah, that's a bit of feature creep for an initramfs, but it is possible... :-P ) > I think a hint for this, if it doesn't already exist, would be a good > start. I actually have one mostly-written; see below... > Once written, we could see if any parts are applicable to the book > itself, but like you, I think its a pretty specialised requirement at > the moment. I think there have been some rumblings on lkml about > requiring an initramfs for some uevent/udev changes but I think that > was either shot down or I possible misunderstood it. Yeah, I think there was some discussion several years back about making all device numbers dynamic, which would require udev to find *any* root FS. There was also some talk slightly more recently about making all device probing happen in parallel (or at least as much of it as possible), which would also require udev, which would have to use the persistent partition symlinks to mount the rootfs. I haven't seen either of these actually happen yet, though. Shortly after the 6.5 release, I realized that my CLFS-1.0.0-rc3 build was getting rather old (2006? yeah, maybe... :-) ). I also remembered that either the initramfs ticket, or discussion here, had said something about doing a hint as a first step. So I figured while I was building a new system anyway, I'd do it on LVM on dm-crypt, and keep track of where I had to deviate from the book. It was surprisingly few places; just a couple more packages at the end, and some setup in chapter 2. I have the notes mostly written up, but I haven't had a chance to clean them up and dump them into hint format yet. It's close, though (stuff at work is getting crazy for me as well). I had figured that once this was done, all that would be required is a note at the start of chapter 2 pointing to the hint and saying something like "specialized root FS setups like x, y, and z are not supported directly by the book, but this hint over here is supposed to work...". But the hint needs to be created first. Soon, I hope. :-)
signature.asc
Description: OpenPGP digital signature
-- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
