Well, drat. Ian's and Luciano's comments are correct: on the FC2 and FC3
ISO images, the anaconda installer appears to require a successful module
load before proceeding with an installation. This is true even when
mountable block device nodes exist on the system.
I'm going to throw in the towel for now, but there are several possible
continuing approaches if anyone in the future wants to give this a shot:
1. Modify anaconda to automatically bypass the module loading. There are
several web sites that discuss modifying anaconda (to resolve specific
boot issues, such as when using new motherboards), but I unfortunately
found no good central reference on anaconda itself.
2. Load a bootstrap file system (see below), then invoke the installation
routines that would normally be called by anaconda. (Note: it is not
immediately obvious to me that this is possible without rebuilding
anaconda; there is a single "loader" executable that's called by the
installer's "init" executable, and "loader" appears to do everything by
itself.)
3. Recompile the xenU kernel to use modules for the disk and network
interfaces, then modify the installer's initrd image to recognize and
include these modules during its initial probe.
4. As per Luciano's suggestion, [i] install the anaconda RPM and run it
from the xen0 partition with the --rootpath option or [ii] use the
"debootstrap" utility with the Red Hat images. This isn't quite solving
the problem I set out to solve (installing Red Hat, wholly within xenU
domains, with no/few modifications), but this will perhaps allow the
creation of a file system without requiring both a reboot and an actual
disk partition.
> I've never tried it, but I wander if its possible to run anaconda
> directly from an installed system? If so, running it under strace
> might provide a useful clue as to what's going on.
I was able to run strace on the anaconda "loader" executable, by booting
xenU directly into the "stage2.img" image included with the ISO images:
A. Mount the DVD image in loopback mode.
B. Mount the file "Fedora/base/stage2.img" (in the DVD image) in loopback
mode. (Note, this is a "cramfs" image, which requires kernel support.)
C. Create a blank 256MB file (or partition), make an ext2 file system on
it, mount it, copy the contents of "stage2.img" into it. [I refer to this
as stage2-ext2.img below.]
D. You'll need to make some changes to stage2-ext2.img before booting from
it, depending on what you want to do below. At a minimum, you can copy
the "strace" binary into /usr/bin. To run the anaconda installer, you'll
need to make directories for /proc, /sys, /var/run, and perhaps some
others. You also may need to copy some files (for example, search for
"lang-table" and copy that file into /etc); you'll see error messages for
each of these when you try to run anaconda.
E. Start an unprivileged domain with options something like this:
disk = [ 'file:/flash/fc2-install/stage2-ext2.img,sda1,w' ]
root = "/dev/sda1"
#extra = "ro init=/usr/lib/anaconda-runtime/loader/init"
extra = "ro init=/usr/bin/bash"
Using the first "extra" line will boot into the anaconda installer, which
will ultimately fail or get into the no-module-found black hole. Using
the second "extra" line will get you a helpful bash prompt and a
reasonably populated /usr/bin directory. Unfortunately, invoking "strace"
on the "loader" executable produced a 500KB file. In lieu of posting that
here, I will forward the trace directly to Ian.
-------------------------------------------------------
SF email is sponsored by - The IT Product Guide
Read honest & candid reviews on hundreds of IT Products from real users.
Discover which products truly live up to the hype. Start reading now.
http://productguide.itmanagersjournal.com/
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/xen-devel
|
Home