VFS: Kernel Panic with Slackware LUKS + LVM

slack_ · 09-03-2014, 02:12 PM

I just installed Slackware 14.1 with LUKS and LVM full disk encryption. I'm running into a problem similar to this: http://postimg.org/image/3nvdh48ex/ and can't find anything that seems to resolve the problem. This is the specific error:

Code:

VFS: Unable to mount root fs on unknown-block(253,1)

Pretty much the same as in that image. Here's my lilo.conf:

Code:

boot = /dev/sda
...
image = /boot/vmlinuz-generic-3.10.17
    initrd = /boot/initrd.gz
    root = /dev/shadowvg/root
    label = Slackware
    read-only

The mkinitrd command:

Code:

mkinitrd -c -k 3.10.17 -f ext4 -r /dev/shadowvg/root -m mbcache:jbd2:ext4 -C /dev/sda2 -L -u -o /boot/initrd.gz

Then I run lilo to reinstall. Once I reboot it starts up and I get that error. Now, something I noticed is that upon startup, the menu entry still says 'Linux' even though I specified it as Slackware in /etc/lilo.conf. So it seems to be that the system isn't even using the lilo.conf that I created. Why? What lilo.conf IS it using then? How do I fix this?

EDIT: Also, if I reboot and chroot back in to try and fix things, I find that there is nothing at all in the /boot dir. Not sure if this is normal but it doesn't seem to be.

maciuszek · 09-03-2014, 02:38 PM

You mounted your boot partition for the install?

naitso · 09-03-2014, 02:51 PM

Have you read the README file?
http://www.slackware.at/data/slackware-14.1/README_CRYPT.TXT

slack_ · 09-03-2014, 02:51 PM

Yes when it got to the part in setup where it asks where to mount everything, I created the mounts as such:

Code:

/dev/shadowvg/root    /
/dev/shadowvg/home    /home
/dev/sda1             /boot

The computer is paritioned like this:

Code:

/dev/sda1    boot         100mb
/dev/sda2    crypto_LUKS  100gb < Where system is to be installed
/dev/sda3    not used     300gb

I didn't mount /dev/sda3 anywhere because I don't want it to be part of the Linux root. I did not use /dev/sda3 at all or mount it anywhere in the fs. So that shouldn't matter at all.

EDIT: Yes, I've read the README file. I've done this before and it worked just fine. That's why I am posting here because I have no idea why it simply didn't work this time. I wrote down a bunch of notes on exactly how to do this so there doesn't seem to be much way I could screw it up if the notes worked for me the first time.

EDIT: I also noticed that everywhere it says that installing lilo the first time will fail. That doesn't happen to me. I still exit the setup but wait to reboot so I can do this: http://www.jveweb.net/en/archives/20...ypted-lvm.html

Down where it says 'Fixing lilo'.

michaelslack · 09-04-2014, 05:44 AM

I can't easily see a reason for the problem. But it seems unable to find and/or mount your root partition.

One question: did you run alienBOB's mkinitrd script to get that mkinitrd command? One thing it does is work out all the kernel modules you need to mount your root partition. I see you put a few modules there. Maybe more are needed though?

I use LVM + LUKS on external USB drives and it took me a while to get it working. One thing I needed was to add -w 5 to the mkinitrd command to give the machine time to mount the external disk. Now it ought not be needed for you (using /dev/sda) but before I did that it would give this kind of error. Maybe adding a short wait will help?

Sorry I can't think of anything else immediately.

Michael

maciuszek · 09-04-2014, 07:04 PM

Check if your boot partition is truly empty if so the partition was not mounted for the install thereby leaving the boot in your encrypted root partition.
you can move the files to the unencrypted boot and have fstab handle the mounting of it.

there won't be an issue of the kernel-* packages or any other package that include this folder since all management will be post fstab.

In the case where it is populated did you generate your initrd including all such arguments: -c -k <kernel version> -m <modules> -f <root fs type> -r /dev/<volume group name>/<root lv name> -C /dev/<physical encrypted block> -L

and of course did you add the initrd to whatever bootloaders conf?

yozh · 09-05-2014, 03:09 PM

Hi Slack..

Facing the same problem right here, right now, funny thing is that it worked once, but couldt type the disks unlock passphrase because module usbhid hid and ehci, so i rebuilt initrd.gz but got the same error youre having.

I was wondering that maybe we need to add crypto/luks and required lvm modules into the initrd.gz
i dont know the name of the modules, so if someone can hook us up with the kernels modules name would be great, anyway those are just my 2 cents, i actually dont know if it can be accomplished or not.

yozH

maciuszek · 09-05-2014, 04:46 PM

Oh.. did you add your root partition to crypttab?

Edit: wait we're talking about lvm right.. hrm i honestly just played with lvm crypt a few years back and dont use it anymore. sadly i cant remember if i added

Code:

slackluks /dev/<physical encrypted blockk> none luks

to crypttab

*slackluks being the naming used in README_CRYPT.TXT

or if it's even needed.. i can assume how it mighhht be needed, but i know nothing about the crypt lvm internals

yozh · 09-06-2014, 03:45 AM

Slack__:

Try replacing your root device "/dev/shadowvg/root" with "/dev/sda2" in mkinitrd command, for me it worked.

The problem is that initrd looks for a non existent root device in that stage yet.