Re: [linux-lvm] booting a dm+lvm2 kernel

[Christophe Saout <christophe saout de>]

Am Fr, den 02.01.2004 schrieb Luca Berra um 00:39:

> >I thought you were talking about the lvmcreate_initrd script that came
> i actually never used lvmcreate_initrd that came with LVM1,

I did. It was quite simple to use and worked practically everywhere (if
you didn't use an initrd to do anything else).

> >I have attached the script. I parse /proc/cmdline, try to remove root=
> >using a shell function and if it actually removed something I know that
> >the rest has to be a device name. :)
> uhm, you should be wary of the case when a root= is in lilo.conf and
> luser wants to override it from cmdline. you stop a t the first
> occurrence.

Ah. You're probably right. I assumed the boot loader would replace it. I
never tested that though. My mistake.

> >With udev I just created /dev/console and some others. Doing everything
> >from the script seems unnecessary.
> i have to, because of tmpfs dev

Right. You mentioned that later.

> I strongly believe that raid autodetection should be done in user space,
> possibly with mdadm. There is a big advantage in being able to control
> what gets activated and when, besides distribution kernels use modular
> scsi drive modules, so probably autodetection won't work anyway.
> I also don't have the faintest idea if the kernel deals with stacked
> raid devices

One day everything will be done in userspace, even detection of
partitions. So you're right. But it's a thing to do in a generic initrd,
not in an LVM-only one. There are a lot of other things you should
probably think of in a generic initrd. EVMS users, or perhaps encrypted
root volume or something...?

> wasted memory and disk space, being able to fit kernel+initrd on a
> single floppy is a good reason

Yes, but does cramfs save a lot compared to a compressed ext2
filesystem?

> >I would love making the initrd an initramfs. But that doesn't currently
> i think initramfs is not mature yet.

It simply not finished.