LinuxQuestions.org - [SOLVED] Custom kernel problems after encrypting disk

Page 1 of 2

Show 50 post(s) from this thread on one page

- Slackware (https://www.linuxquestions.org/questions/slackware-14/)

- - Custom kernel problems after encrypting disk (https://www.linuxquestions.org/questions/slackware-14/custom-kernel-problems-after-encrypting-disk-670332/)

brooky9999

09-16-2008 03:28 PM

Custom kernel problems after encrypting disk

Hi all,

I followed the excellent guide to encrypting my laptop hard drive at http://slackware.osuosl.org/slackware/README_CRYPT.TXT on Slack 12.1. Everything went well, until I came to install my own 2.6.25.4 kernel.

I followed (Bruce Hill's I think) kernel compile guide, which in short consisted of the following:

Code:

Extract source

copy in custom .config file

make

cp System.map /boot/System.map-2.6.25.4

ln -sf /boot/System.map-2.6.25.4 /boot/System.map

make modules_install

cp arch/i386/boot/bzImage /boot/vmlinuz-2.6.25.4

I then added the following to /etc/lilo.conf:

Code:

image = /boot/vmlinuz-2.6.25.4

  initrd = /boot/initrd.gz

  root = /dev/cryptvg/root

  label = Linux_2.6.25.4

  read-only

The following section was already there from the cryptsetup (which works):

Code:

image = /boot/vmlinuz-generic-smp-2.6.24.5-smp

  initrd = /boot/initrd.gz

  root = /dev/cryptvg/root

  label = Linux

  read-only

I compiled in the following to my custom kernel:

Code:

CONFIG_EXT3=y

CONFIG_BLK_DEV_DM=y

CONFIG_DM_CRYPT=y

After running lilo and rebooting, my system system now fails to boot. I get the following error:

Code:

VFS: Cannot open root device "fd02" or unknown-block(253,2)

Please append a correct "root=" boot option; here are the available partitions:

0300  58605120 hda driver: ide-disk

  0301    96358 hda1

I would expect an error like this because I've fail to compile something into my custom kernel - but I thought I had everything.

What have I missed?

Do I have to create a separate initrd.gz for each kernel, for instance:

Code:

2.6.24.5-smp-initrd.gz

2.6.25.4-initrd.gz

Mark.

brodo

09-16-2008 05:16 PM

Did you correctly generate initrd.gz file and initrd-tree subdir in /boot ?
That should be done after placing your new modules in /lib/modules/ and kernel image file in /boot.

I used the following command to do it (kernel generic 2.6.24.7-smp from Patrick V.) while succesfully upgrading to current kernel:

mkinitrd -c -k 2.6.24.7-smp -m ext3 -f ext3 -r /dev/cryptvg/root -C /dev/sda3 -L

Try to compare your kernel settings with settings applied by Patrick V.

My first reboot attempt failed due to my faulty /dev/sda2 definition in the mkinitrd command sequence but I was able to correct that:
http://www.linuxquestions.org/questi...ukslvm-669700/

brooky9999

09-16-2008 05:25 PM

Hi,

Not at first - and this is where I think I went wrong.

The newer 2.6.25.4 wouldn't boot. So I did the following:

Code:

mv /boot/initrd.gz initrd-old.gz

mkinitrd -c -k 2.6.25.4 -m ext3 -f ext3 -r /dev/cryptvg/root/ -C /dev/hda2 -L

Although this wouldn't boot either. So I tried to backout by:

Code:

mv initrd.gz initrd-didntwork.gz

mv initrd-old.gz initrd.gz

My thinking was that my stock 2.6.24.5-smp kernel would at least now boot again. And it doesn't.

So now I have two kernels and neither boot. I'm pretty much resigned to the fact I'll have to do a complete rebuild, as I don't know how to get back into my encrypted drive.

Any ideas?

M.

Bruce Hill

09-16-2008 06:05 PM

Mark,

No need to "do a complete rebuild."

When you use -c with mkinitrd, it does this:

Code:

mingdao@silas:~$ man mkinitrd

<snip>

      -c    Clear the existing initrd tree (by default in /boot/initrd-tree/) first.  If this is not done, running mkinitrd will add additional

              modules to the existing initrd.

So when you issued that command the first time, your old initrd image,
and everything under /boot/initrd-tree, cease to exist. Therefore, the
subsequent copying of the old stuff was futile.

You don't need to add -L to your mkinitrd comand unless you created a LVM.

Before I write "HOW-TO recover your system", can you tell me:

1 - if you've already reinstalled

2 - if you also created a LVM (and it's parameters if you did)

brooky9999

09-17-2008 02:19 AM

Hi Bruce,

Thanks for the info on initrd. I guess simply copying the files wasn't enough.

I have not rebuilt my system yet.

I am using LVM, and I used the following commands to create my volumes:

Code:

pvcreate /dev/mapper/slackluks

vgcreate cryptvg

lvcreate -L 2G -n swap cryptvg

lvcreate -L 55.79 -n root cryptvg



vgscan --mknodes

vgchange -ay

I am now going to go read the man page for mkinitrd ;-)

Thanks,

M.

Bruce Hill

09-17-2008 06:02 AM

Okay, here's what you need to do to get back in to that Slackware system.

First, boot with Slackware CD1 of the DVD.

Second, at the boot prompt just hit enter. Then:

Code:

- Unlock the LUKS encrypted LVM:

    cryptsetup luksOpen /dev/hda2 slackluks (substitute your LVM for /dev/hda2)

- Create the device nodes for the LVM then activate the volumes:

    vgscan --mknodes

    vgchange -ay

- Mount the filesystems:

    mount /dev/cryptvg/root /mnt

    (no need to mount swap)

    chroot /mnt/

That should get you back into your system. Do a "ls -l" to see if you're in,
then you can start to fix it.

First, I'd recreate the initrd image for the Slackware default kernel, then
we can find out what's wrong with your custom kernel, once we have a
working Slackware system.

If you're in, then issue:

Code:

cd /boot

ls -l (make sure you see the default kernel -- vmlinuz-generic-smp-2.6.24.5-smp)

mkinitrd -c -k 2.6.24.5-smp -m ext3 -f ext3 -r /dev/cryptvg/root/ -C /dev/hda2 -L

less /etc/lilo.conf (make sure you still have this stanza for the default kernel:)

    image = /boot/vmlinuz-generic-smp-2.6.24.5-smp (or different if it's symlinked)

      initrd = /boot/initrd.gz

      root = /dev/cryptvg/root

      label = Slack-12.1

      read-only

lilo (if everything in /etc/lilo.conf looks correct)

When you issue the mkinitrd command, it should output something like
"9607 blocks" and nothing more (your blocks won't be the same).

When you issue "lilo" it should output something like:

Code:

root@slackware:/# lilo

Warning: LBA32 addressing assumed

Warning: '/proc/partitions' does not match '/dev' directory structure.

    Name change: '/dev/dm-0' -> '/dev/mapper/slackluks'

Warning: Name change: '/dev/dm-1' -> '/dev/cryptvg/root'

Warning: Name change: '/dev/dm-2' -> '/dev/cryptvg/home'

Warning: Name change: '/dev/dm-3' -> '/dev/cryptvg/swap'

Warning: Unable to determine video adapter in use in the present system.

Warning: Video adapter does not support VESA BIOS extensions needed for

  display of 256 colors.  Boot loader will fall back to TEXT only operation.

Added Slackware *

Added Windows

7 warnings were issued.

Those warnings are harmless -- anything else and you should record it.

Now that should boot you back into your Slackware system. If so, then we
can fix the custom kernel. First, let's just get the default to boot, so
we can always have it to fall back on. Next time we make that initrd, we
will drop the -c switch, to keep our existing modules in the initrd.gz.

brooky9999

09-17-2008 08:17 AM

Hi Bruce,

Thanks for this info... although something weird has stopped me.

I can get in, create the device nodes and mount the filesystems okay. However, if I cd to /boot, there are no files in there.

I definitely set /dev/hda1 to /boot. When I run:

Code:

mount

I get:

Code:

/dev/mapper/cryptvg-root on / type ext3 (rw)

/dev/hda1 on /boot type ext3 (rw)

(as well as other mount points).

Why would /boot show up as empty? I definitely haven't wiped it!

M.

brooky9999

09-17-2008 11:47 AM

Okay. I got in, and now I have a rescued 2.6.24.5-smp default kernel... yay!

Now I have to work out why my new custom kernel won't boot :-(

How do I create a new initrd for my custom kernel with blowing away the 2.6.24.5-smp one?

Thanks for your help so far... it's saved me a lot of time.

M.

Alien Bob

09-17-2008 12:29 PM

Quote:

Originally Posted by brooky9999 (Post 3283313)

How do I create a new initrd for my custom kernel with blowing away the 2.6.24.5-smp one?

Like Bruce already told you (and the man page as well if you read it):
If you want to use a single initrd.gz for multiple kernels, you just omit the "-c" parameter to mkinitrd.
The "-c" stands for "create" - it wipes your current /etc/initrd-tree directory (out of which your initrd.gz is created) before re-adding the requested stuff. So, the correct way to proceed is

Code:

mkinitrd -k 2.6.25.4 -m ext3 -f ext3 -r /dev/cryptvg/root/ -C /dev/hda2 -L

(not the absent "-c"). This command would add the 2.6.25.4 kernel modules to your existing initrd.gz ..

Never forget to run

Code:

lilo

everytime you have messed with an initrd.gz! LILO does not read files in your /boot filesystem like GRUB does. The lilo bootloader only knows what block of data to load from your disk because the start of that block is written in the bootloader. If you change initrd.gz then LILO will lose track of it and your computer will not boot - so you run "lilo" before you reboot.

brooky9999

09-17-2008 01:12 PM

Cheers Eric,

I omitted the -c but my custom kernel still didn't boot.

There's obviously something still missing from it which I need to look into.

M.

Bruce Hill

09-17-2008 04:02 PM

What are the errors you get when your custom kernel won't boot?

brooky9999

09-17-2008 04:35 PM

I'm getting a bunch of warnings such as:

Code:

Using /lib/modules/2.6.24.5-smp/kernel/fs/mbcache.ko

mbcache: version magic '2.6.24.5-smp SMP mod_unload 686 ' should be '2.6.25.4 SMP mod_unload CORE2 '

insmod: cannot insert '/lib/modules/2.6.24.5-smp/kernel/fs/mbcache.ko' : Invalid module format (-1): Exec format error

Using /lib/modules/2.6.24.5-smp/kernel/fs/ext3/ext3.ko

jdb: version magic '2.6.24.5-smp SMP mod_unload 686 ' should be '2.6.25.4 SMP mod_unload CORE2 '

insmod: cannot insert '/lib/modules/2.6.24.5-smp/kernel/fs/jdb/jdb.ko' : Invalid module format (-1): Exec format error

Using /lib/modules/2.6.24.5-smp/kernel/fs/ext3/ext3.ko

ext3: version magic '2.6.24.5-smp SMP mod_unload 686 ' should be '2.6.25.4 SMP mod_unload CORE2 '

insmod: cannot insert '/lib/modules/2.6.24.5-smp/kernel/fs/ext3/ext3.ko' : Invalid module format (-1): Exec format error

However, I do then get a chance to enter my LUKS passphrase.

After correctly entering it in, I get:

Code:

mount: mounting /dev/crypt/root/ on /mnt failed: Not a directory

ERROR:  No /sbin/init found for rootdev (or not mounted).  Trouble ahead.

        You can try to fix it.  Type 'exit' when things are done.

I created my initd for 2.6.24.5-smp as per your post, and then another (minus -c) for 2.6.25.4. So I'm not sure why it's trying to load the wrong versions of mbcache, jdb and ext3.

I'm guessing that those first errors are giving me the more serious one about not mounting my root volume.

Would I be about right?

M.

Alien Bob

09-17-2008 05:55 PM

Quote:

Originally Posted by brooky9999 (Post 3283632)

I'm getting a bunch of warnings such as:

Code:

Using /lib/modules/2.6.24.5-smp/kernel/fs/mbcache.ko

mbcache: version magic '2.6.24.5-smp SMP mod_unload 686 ' should be '2.6.25.4 SMP mod_unload CORE2 '

insmod: cannot insert '/lib/modules/2.6.24.5-smp/kernel/fs/mbcache.ko' : Invalid module format (-1): Exec format error

Using /lib/modules/2.6.24.5-smp/kernel/fs/ext3/ext3.ko

jdb: version magic '2.6.24.5-smp SMP mod_unload 686 ' should be '2.6.25.4 SMP mod_unload CORE2 '

insmod: cannot insert '/lib/modules/2.6.24.5-smp/kernel/fs/jdb/jdb.ko' : Invalid module format (-1): Exec format error

Using /lib/modules/2.6.24.5-smp/kernel/fs/ext3/ext3.ko

ext3: version magic '2.6.24.5-smp SMP mod_unload 686 ' should be '2.6.25.4 SMP mod_unload CORE2 '

insmod: cannot insert '/lib/modules/2.6.24.5-smp/kernel/fs/ext3/ext3.ko' : Invalid module format (-1): Exec format error

It looks as if you have edited /boot/initrd-tree/load_kernel_modules and used hard-coded kernel versions there. Normally, the initrd would not even try to load kernel modules that did not match the version of the kernel it is running on.
But these errors are harmless (though annoying).
It is not even required to use /boot/initrd-tree/load_kernel_modules by making it executable - it is there to help you if you have to go through hoops to load your modules properly (for instance when you have to pass parameters to the modules).

The other stuff you posted indicates that your initrd may have forgotten to load the driver for your ext3 filesystem. Did you really add these modules (mkinitrd -m ext3)? If so, I can only guess that you edited B]/boot/initrd-tree/load_kernel_modules[/B] (forgetting to add the modules for your new kernel) and made it executable.

Eric

Bruce Hill

09-17-2008 06:34 PM

As for the non-loading custom kernel, you're possibly missing something
in your .config. You need certain crypto cipher modules for DM_CRYPT to
work. For sure you need SHA256, AES, and CBC.

Please post your 2.6.25.4 .config file somewhere for us to see. Also
would not hurt to see:
ls -l /boot
cat /etc/lilo.conf

brodo

09-18-2008 09:27 AM

Maybe it would be easier only to upgrade your kernel/Slackware packages to PV's current state (for example 2.6.24.7-smp) as a 1st step.
That way you'll see if the procedure works.
If so, then compile your own kernel.

All times are GMT -5. The time now is 12:02 PM.

Page 1 of 2

Show 50 post(s) from this thread on one page