Hello,

I was installing Slackware 14.1 64-bit in Vmware workstation with a different partition scheme than I'm used to of. I wanted LVM with xfs root and btrfs home. So my partition scheme was :-

Normal partitions :-
/boot ext2 - 100mb
swap - 2GB

LVM partitions layout :-
/dev/myvg/root - xfs - 20GB
/dev/myvg/home - btrfs - 10GB

I followed exactly the same steps as mentioned in official README_LVM.txt, except that the root FS I used was xfs and for home btrfs. And there was no 3.10.17-smp kernel, just 3.10.17 and thus created the initrd accordingly.

ftp://ftp.slackware.com/pub/slackwar...README_LVM.TXT

The LILO install was successful too without any error. However after final rebooting I got slapped with this error :-

Attachment 14835

Is it related to xfs being root on LVM or something else I totally missed and screwed up?

Regards.

It seems /dev/myvg/root is trying to mount in /mnt (which doesn't exist, since it shouldn't)... It should be mounted in '/' ...

Check that '/etc/lilo.conf' (from that partition) is correct and possibly redo your initrd image

P.S. it is ok to use the Slackware installation disk if you need a clean simple console... Basicaly, instead of running 'setup' you should mount the root partition and chroot to it ( mkdir -p /mnt && mount -t xfs /dev/myvg/root /mnt && chroot /mnt ) .. From there on, recheck lilo configuration (and maybe reinstall it) and redo the init image... Insure that the -r parameter for the mkinitrd command is '-r /dev/myvg/root'

That's the strangest thing.
I did the steps you mentioned above. The LVs were inactive so I activated the VG using vgchange -a y first.
The /etc/fstab looks totally fine:-

Attachment 14836

Previously I created initrd using :-
# $( /usr/share/mkinitrd/mkinitrd_command_generator.sh -r )

This time I did it the manual way :-
# mkinitrd -c -k 3.10.17 -m xfs -f xfs -r /dev/myvg/root -L

The lilo also looked pretty normal, however when I ran #lilo, it threw an error of /proc/partitions not found and after reboot I'm still at square 1. The same error, mounting on /mnt, why oh why?

Attachment 14837


Regards.

Did you install lilo to MBR or root? Lilo can't boot if it has been installed to a root partition formatted with XFS. make sure it is installed to MBR.

Hi gezley, I'm aware of that and thus I installed it in MBR.
why is it trying to mount root to /mnt?

Regards.

Ok, I've created an initrd and browse over the init.. It is normal to mount in /mnt and 'switch_root' after...

So now, can you please (at the console after the error) do an 'ls /', 'ls /sbin' and 'ls /dev/myvg'. Check that /mnt folder does exist and that /sbin/vgchange executable file does exist? And, ofcourse... see that /dev/myvg/root device file exists...

I think you're missing one of these for some reason...

Thanks Smokey_justme. A quick check showed out of the files/folders you mentioned , I'm missing on /dev/myvg/root device file. Rest are there.

Regards.

Ok, so for some reason you logical volume isn't recognized... hmm... at this point, since I don't really use LVM I'm kind of out of ideeas... Maybe I'll give it a shot later on a Virtual box.. Or maybe someone else knows more about this..

For now, sorry..

Not a problem mate, thanks for the help so far. I've been using all my CentOS VMs with LVM and never faced such problem. Lets see what I'm missing here.
Anybody else with any suggestion please?

Regards.

This will probably fail but still, it might be worth trying:

The mkinitrd in 14.1 has a bug which requires you to use the -L flag even if you only need to unlock a LUKS partition wit the -C flag. Just -C should suffice, but not all files are copied to initrd.gz unless you pass -L too. So maybe you could try adding -C to see if the same weird thing happens.

Hi, I'm not using LUKS. However thanks for noticing. I'll read more about the bug.

Regards.

Yes I noticed that, but as I explained above, at least in certain scenarios you need to use irrelevant flags to make it work .

I was attempting to reproduce this using libvirt and virt-manager, but lilo doesn't recognize the virtualio disk type that is used during the installation. I will give it another try this evening, but what I had attempted to do was:

1. I don't try to run lilo during the initial setup. There's no initrd created, so whatever lilo does at that point won't be correct.
2. After setup is finished and you are back at the command prompt, I did these additional steps:
   
   Code:

   bind -R /proc /mnt/proc
   bind -R /sys /mnt/sys
   bind -R /dev /mnt/dev
   chroot /mnt

   Once inside the chroot, I ran the mkinitrd_command_generator.sh script, then ran liloconfig and then ran lilo itself. Things worked as expected up to executing lilo, which failed when it didn't recognize the virtual disk's characteristics.

Quick follow-up: http://www.linux-kvm.org/page/Boot_f...o_block_device appears to tell me how to make lilo work with a virtual block device. I'll be giving that a spin shortly.

OK, I was able to install lilo on my virtual image and the system booted correctly.

liloconfig was pretty useless, however. I ran it, selected "simple", and chose reasonable answers to the questions. When it failed, it at least left /etc/lilo.conf behind that I could edit.

So, after "setup" completed, I exited setup and then ran the following commands......and then shut down the VM and restarted it since the lilo output told me that lilo worked.

FWIW, mkinitrd_command_generator.sh in my case provided the following command line:I also changed the boot device in /etc/lilo.conf to /dev/vda and added the line "device = /dev/vda bios=0x80 max-partitions=7" so that lilo would work. You probably won't have to do that.