You need to create a console device node in dev on / behind the tmpfs mount that udev uses: # mkdir /tmp/root # mount --bind / /tmp/root # mknod /tmp/root/dev/console c 5 1That got me past the missing console and I'm up and running. I don't know how to turn this into a permanent solution, but perhaps the problem only occurs on the first boot.
Dave David Liontooth wrote:
Hi Török,Thank you, that's very helpful. I ended up just installing a new Debian on a different partition; once that was done, the installer agreed to rewrite the MBR, and my old installation showed up in grub. I would much prefer being allowed to rewrite the MBR without first having to install another OS.Once I got that far, I could boot into the OS I'm trying to rescue; however, I now ran into the dreaded "warning: unable to open initial console". This is also /dev/ related, and this suggests I have a general dev problem with the partition I cloned. Now that I have a parallel partition that is fully operational, is there a way I can repopulate my /dev directory from chroot, or through a script?Using "mount -o bind /dev /root2/dev", I can now chroot into /root2 and see I have a complete /dev directory, so that for instance I can mount /dev/hda1 on a chrooted /boot. How do I make this happen when I boot directly into that partition? I added this to /etc/fstab:devpts /dev/pts devpts mode=0620,gid=5 0 0But it didn't solve the problem. This seems to be a well-defined problem that should have a straightforward solution.I appreciate your help with this; the partition I'm trying to rescue is in a stable functional state that is no longer possible to duplicate.Dave Török Edvin wrote:On 3/24/06, David Liontooth <firstname.lastname@example.org> wrote:I cloned a drive, starting with the MBR: dd if=/dev/hda of=/dev/hdc bs=446 count=1However, after installing the cloned drive in the new machine, all I getis a scrolling GRUB filling the screen. I downloaded an amd64 netboot CD, mounted / and /boot in /target,Did you do a mount -o bind /dev /target/dev ?[...] In the installer shell, I get ~ # df Filesystem 1k-blocks Used Available Use% Mounted on tmpfs 102400 30248 72152 30% / /dev/cdroms/cdrom0 86390 86390 0 100% /cdrom/dev/ide/host0/bus0/target0/lun0/part2 7740384 3197252 4149944 44%/target/dev/ide/host0/bus0/target0/lun0/part1 97826 9736 82871 11%/target/bootSo /dev is not bind mounted in target, try bind mounting itSo it's using devfs, which I thought was deprecated?http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=274511 AFAIK Sarge CD uses devfs, but the etch installer uses udev, quote from that bugreport: "d-i uses the device naming scheme originally used by devfs (although we use udev now)".Use devfs naming, /dev/ide/host0/...., see where is the equivalent of hda.What do I tell grub-installer?[...] klogd: segfault at 000000000000003e rip 0000002a9568b94e rsp 0000007fbfffe2a0 error 6How did this segfault happen?[...]What can I do? All I want is to rewrite the MBR, the installation itselfis fine.If the netboot CD isnt working for you, could you try using a LiveCD, like Knoppix?