I've been trying without much success to get NetBSD/xen booting as
domain0 on a VIA EPIA-5000 motherboard. The machine launches GRUB,
GRUB appears to successfully load xen.gz, and then seemingly as soon
as my xen0 kernel is accessed I get a reboot. I'm pretty confident
that my installation is sound: if I take the system and swap only the
VIA motherboard/cpu for a generic Intel Mobo/Intel P3 combo, the same
disk boots in to the Xen0 kernel straight away, so I am guessing a
problem with the VIA C3 CPU?

This is actually my first experience with NetBSD in any capacity, so
excuse any obvious blunders I've been making.

I think the problem is with the XEN0 kernel, and not with the xen.gz,
an assumption I arrived at when I deleted the kernel and watched the
xen.gz get as far as complaining it couldn't find the xen0 kernel.
Does this really mean the xen.gz loaded and executed without issue?

The EPIA 5000 has a 686 class CPU but I know it doesn't have the CMOV
instruction, so I attempted to compile a XEN0 kernel with I586_CPU
specified in the kernel config, thinking that might solve it. It
didn't. I tried moving back to I486_CPU & I386_CPU but then I couldn't
get the kernel compiled. It fails with:

/usr/src/sys/arch/xen/i386/identcpu.c:1166: undefined reference to
`cc_microtime'
/usr/src/sys/arch/xen/i386/pmap.c:1305: undefined reference to
`pentium_idt_vaddr'
/usr/src/sys/arch/xen/xen/clock.c:270: undefined reference to `cc_microset_=
time'
/usr/src/sys/arch/xen/xen/clock.c:271: undefined reference to `cc_microset'

... given my lack of NetBSD experience and the fact I'm not a
particularly accomplished C coder I haven't gone much further down
this path yet, but it seems like a >=3D 586 CPU might be  required
anyway?

One point probably worth mentioning is that when I tried the original
standard NetBSD installation on this system I could not get any
current NetBSD boot floppies to load, except for the ones at
netbsd-3/200508120000Z/i386/installation/floppy. The rest all caused a
reboot shortly after startup, coincidentally in a rather similar way
to my rebooting xen kernel. Anyway, this install issue is documented
quite well in PR26007. I see the patch mentiond in the PR at
http://frog.coolfactor.org/~gavan/epia_hack.diff is present in the
source I'm compiling from
(ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-current/tar_files/src/sys.tar.gz)
, and I have tried my compile with "options EPIA_HACK" in the config
file, but it hasn't made any difference.

I'd appreciate any thoughts, comments or advice!

Regards

Chris