Hello Chris, Nice to see you keeping up your quest. On Tuesday 09 January 2007 23:03, Chris Carr wrote: [...] > 4. Continue through detect and partition disks, timezone and clock, > users and passwords, install base system, install kernel Looks good so far. > 5. Switch back to console on F2 and do the following: > 5.1 cp /sbin/dmraid /target/sbin (this is necessary because the > dmraid .deb is not available for install to /target) > 5.2 chroot /target > 5.3 nano /usr/share/initramfs-tools/hooks/dmraid Here I think you are really doing the wrong thing. Instead of trying to reinvent the wheel, you'd be much better of just installing the dmraid package in the /target directory. If you already have networking setup (maybe run the apt-setup step first), this should be trivial. Just run 'apt-install dmraid' while the installer itself is showing the menu (this is important!). The advantage of using this command is that it will do some special things to set up the chroot properly. I _think_ this will work from the command line. (I tested it and it worked for me, the initrd is regenerated automatically; output will be on VT4.) The reason you should use the official dmraid package is that that should set up the initrd correctly. > 7. [At this point I still don't know how to get either grub or lilo to > install, so choose "Continue without boot loader" and finish - I am > lucky enough to have Grub already installed and working, so I can > continue to experiment] The problem is that the grub installer assumes that any dm devices that are not crypto, are LVM, which is not true if dmraid is used. So, we need a patch to distinguish dmraid devices. > Now, that all looks as if it ought to work, but for some reason it > doesn't quite. Three out of four modifications to the initrd work: the > dm-mod module is included in the build, the dmraid executable is > included in the build, and the dm-mod is successfully modprobe'd. But > the dmraid -ay command in the init-premount script does not get run, > even though /sbin/dmraid is present on the initrd, and can be run > manually from the busybox shell (where it successfully creates the > /dev/mapper/sil_* devices). Hopefully this will solve itself if you use the official dmraid package. > So why will it do that when I process init manually, and not when I use > the break option and manually execute dmraid? > > And why will dmraid not run from my init-premount script, even though > modprobe runs successfully from the same script? Does this have > something to do with 5.1 (the executable being from a udeb rather than > from a proper package)? If using the dmraid package does not work, try contacting the maintainer of the dmraid package or possibly the initramfs-tools maintainer for these questions. > I would love to get this working - it seems very close now. We will only be able to tackle this after the release of Etch I'm afraid. Cheers, FJP
Attachment:
pgpnI5FpCm_EV.pgp
Description: PGP signature