On Thu, Mar 27, 2008 at 5:00 PM, Andrew Sackville-West < [EMAIL PROTECTED]> wrote:
> On Wed, Mar 26, 2008 at 08:32:50PM +0100, Benjamin Kircher wrote: > > Hi there, > > > > > > need some advice over here. Installed Etch from netinstall iso and > upgraded > > to sid, which works flawlessly. All partitions except /boot are in a LVM > > group (even my root fs is a LVM volume). Then I removed (more precise: > > purged) 2.6.18-5-486 kernel in favor of 2.6.24-1-686. After running a > > "update-initramfs -u" and a "shutdown -r now" my box fails to boot up. > It > > complains: > > > > (nah nah nah, a lot of this not and a lot of that not, neh neh ney) > > ... > > /scripts/local-top/lvm2: line 68 tr not found > > /scripts/local-top/lvm2: line 68 sed not found > > (I guess, this part of the script tries to activate my volume group) > > ... > > /proc/mounts: fopen %s failed: No such file or directory > > Failed to create lvm type filter > > Waiting for root file system ... > > > > ... which my box never gets. Hereafter it throws me to a very limited > shell. > > Indeed, /proc and /dev are empty ;( > > Hey wow! I'm seeing this too (though i can boot). I thought it was > from moving to an amd64 kernel on i386 userland. But apparently > not. It seems that the initrd environment is broken at the moment. I'm > not sure, maybe it's ash (I think that's the initrd shell) that's > broken. > > You'll notice that this limited shell (busybox) doesn't function > properly -- ls doesn't exist for example. But you can fake it with > `busybox <cmd>` to get it to execute. So, for example, to get an ls of > the current directory, use `busybox ls` and there it will be. > > So to boot this thing, you need to duplicate the processes that happen > in the initrd phase. This means you need to activate your lvm and > possible any encrypted stuff (if you have encryption as well, let me > know and we'll work through that). > > to activate lvm, do this from the busybox shell: > > vgchange -ay <name-of-volume-group> > > -ay for "activate yes" essentially. > > this will create your volume group in /dev/mapper. Check it with > > busybox ls /dev/mapper > > if that's working, then you should be able to just > > exit > > the busybox shell and your boot should continue. > > > Essentially, certain necessary functions are missing from the initrd > causing these scripts to fail. It should be fairly simple for you to > get booted, and once that's done just stay up until a fix comes > through, or keep track of what you did to successfully boot so you can > reproduce it. > > And this warrants a bug report, if one doesn't exist > already. hmmm... a little quick searching doesn't reveal anything > specific. I would file it with initramfs-tools as they know who get's > included in the initrd and can forward it on if required. > > A > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.6 (GNU/Linux) > > iD8DBQFH68SiaIeIEqwil4YRAg5cAKCUit3yyl47C5TQUEdzBbc6t4xXtwCdGwSl > VFJplL3c2mRDs2MIbP6Xw5U= > =YJYt > -----END PGP SIGNATURE----- > > Hi Andrew, I'm not able to activate lvm from within busybox, a "vgchange -ay group0" gives me: /proc/mounts: fopen %s failed: No such file or directory /proc/devices: fopen failed: No such file or directory Failed to create lvm type filter /proc and /dev (except /dev/console) are empty like the ocean. I think you're right: My initrd is really f'd up. The first lines I noticed during boot up: /init: line 11: mount: not found cat: can't open '/proc/cmdline': No such file or directory ... /scripts/init-top/framebuffer: line 100: mknod: not found ... /init: line 157: printf: not found ... I took a look at the initramfs-tools bug page and couldn't find any similar reports. I'll do some further investigations (need to know how initramfs works in detail). Benjamin
