<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Jeremy Katz wrote:<br>
<blockquote
cite="mid1142269661.1963.13.camel@orodruin.boston.redhat.com"
type="cite">
<pre wrap="">
You shouldn't have to specify this. xend is configured to auto-balloon
dom0 out to something reasonable (256M, iirc) as needed to avoid the
need for specifying this
</pre>
</blockquote>
I see. I wasn't aware that xend could dynamically change this in such a
way.<br>
I will look more into it, and probably send a modified patch to the
list to reflect the changes.<br>
<blockquote
cite="mid1142269661.1963.13.camel@orodruin.boston.redhat.com"
type="cite">
<pre wrap=""></pre>
<blockquote type="cite">
<pre wrap="">2) The ramdisk_size= values specified at 10000
</pre>
</blockquote>
<pre wrap=""><!---->
This also shouldn't be needed as the "initrd" is actually an initramfs
and thus doesn't use the ramdisk code in the kernel at all.
</pre>
</blockquote>
It isn't actually. It is a small gzipped ext2 filesystem. <br>
I would like to see Kadischi move to more standard, and base procedures
like<br>
using the same format as Fedora Core does, with what I understand as an
initramfs <br>
being the gzipped cpio archive. I may be way off base, but to the best
of my knowledge it isn't.<br>
<br>
I had been fiddling with the provided mkinitrd from Fedora Core stock
and generated a few patches already<br>
that _could possibly_ make this happen. I'll provide the bugzilla
reference here:<br>
<a class="moz-txt-link-freetext" href="https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=183566">https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=183566</a><br>
A brief explanation shows that:<br>
1) The initrd for the Live discs would fall in organization with the
existing Fedora Core style.<br>
2) A hacked up livecd-mkinitrd.sh could be deprecated in favor of using
/sbin/mkinitrd from the host system.<br>
2a) This would bring the Kadischi package down a few KB.<br>
3) The patches allow for what is neccessary to be inserted to the
initrd. <br>
3a) --init= for the linuxrc (Which could, or may not be using /bin/ash
(busybox.anaconda or nash, anything is possible)<br>
3b) --root= would be specified to build an initrd for a kernel that is
not on / (/lib/modules/$(uname -r) on the host) for example.<br>
3c) --bin= for installation of any binary needed, whether statically
compiled or not (such as busybox.anaconda).<br>
This same patch set I think could use --lib= also, though I have not
written that up and attached it to that report.<br>
<br>
As I said, and to not make this mail too lengthy, these were ideas and
sound plausible.<br>
If this were to become reality and put forward to deployment the
groundwork for it is laid. <br>
Even if modifications are neccessary. If not, no big deal. <br>
<br>
In any case, if these were folded into mkinitrd even just for Kadischi,
it would benefit Kadischi and any other<br>
project that needs an initrd with some modifications.<br>
Otherwise we always make an initrd for Live media duplicating the
efforts of the existing Fedora Core mkinitrd.<br>
<br>
J. Hartline<br>
</body>
</html>