Re: [coLinux-devel] Initrd support?

[Eric A. <ea...@ma...>]

Correct. I've basically bypassed this failing by loading it as a block
device, though. In order to make it act as an initrd in this case, you
have to add the line:
exec chroot . /sbin/init <dev/console >dev/console 2>&1

*after* your pivot_root, otherwise when your script ends, so does your
kernel's lifeline.

I've basically finished my initrd work, and it appears to working well.
During this configuration process, though, I've noticed a disturbing
problem that I can confirm occurs on two different windows machines (one
running Server 2003, the other XP Pro): The first time through, coLinux
runs happily, until the point where init effectively dies (not the
problem, simply unfinished configuration). So, I close coLinux, and
restart it. This time coLinux hangs the entire system sometime before
the expected dying of init (right after setting the system clock, but I
doubt that is even relevant).

I'm using the 20040304 development snapshot. I'm not sure what the cause
of this problem is, but my guess would be something is not being
properly cleaned out when coLinux closes the first time around. There
was a single time where it did not occur until the third restart of
coLinux.

Memory usage differed by about 3-8mb before and after coLinux execution,
but that could simply be buffers/cache.

-- Eric Andresen


On Wed, 2004-03-03 at 19:07, Pat Erley wrote:
> On Thu, 4 Mar 2004 01:02:46 +0100 (CET)
> ch...@to... wrote:
> 
> > did you create a new kernel with initrd support?
> > 
> > did you read /usr/src/Documentation/initrd.txt
> > I think root=/dev/cobd0 init=/linuxrc
> > needs the explicit change root commands at the end of linuxrc
> > where as intrd=  can use a linuxrc that will automaticly pivot root at the
> > end.
> > 
> > if you checked out all this you could privatly e-mail me your linuxrc and
> > .config for your kernel and I can look it over. I could also do some
> > further experiments with what I have tomorow. I think I stopped at the
> > same place you are due to not having the commands needed on my initrd and
> > realizing that I had stripped my linuxrc down to almost nothing and I came
> > up with another way to do it.
> 
> actually, it'd be colinux' job to handle loading the initrd into ram.  the entire point of initrd is to load the initrd into ram and then pass vmlinuz the location of it and let it start processing it, having not had to do any disk io.
> 
> Pat Erley