[9fans] Xen 3.1 w/ Plan 9 domU: plan9.img file after install not recognized.
Mike
I followed these instructions:
http://www.cs.bell-labs.com/wiki/plan9/Installing_in_Xen/index.html
for installing plan9 in a Xen domU. It appeared to go swimmingly, until
I tried booting the image file after my successful install:
216# xm create -c plan9.py
Using config file "./plan9.py".
Error: unknown image type: plan9
217# file plan9.img
plan9.img: x86 boot sector
218# ls -l plan9.img
-rw-r--r-- 1 root root 1073741824 2007-07-04 08:56 plan9.img
219#
My plan9.py looks like:
kernel = "/export/2/Guests/Plan9/9xenpcf.gz"
builder = "plan9"
memory = 256
name = "plan9"
vif = [ 'mac=aa:00:10:00:00:1b, bridge=xen-br0' ]
disk = [ 'file:/export/2/Guests/Plan9/plan9.img,loop0,w' ]
ip = "10.10.0.9/24" # <- IP you will give to xen9
#restart = 'never'
console = 999
# This is the equivalent of plan9.ini:
extra="""
nobootprompt=local
"""
My dom0 is Fedora 7 + Xen 3.1, both compiled with PAE turned off.
Apparently, Xen on F7 doesn't recognize the boot image plan9.img,
created by the plan9 install.
Any hints as to what incantations I may have missed?
Thank you,
Mike
Richard Miller
> builder = "plan9"
ron minnich
on the new lguest hypervisor as the final plan. Please google lguest
and see what you think. It's 5K lines of code, the guest support code
is trivial, and I think overall this is going to be the easiest thing
to do. The work should actually not take too long this time. Xen can
be hard to get your brain around.
Xen was notable at ottawa linux symposium mostly by its absence, and
the fact that people (almost continually) deprecated it in favor of
either KVM or lguest. The Xen BOF actually got cancelled. The Xen
Plan 9 is fine for the time being, but it is also time to move on to a
simpler way of doing things.
thanks
ron
Mike
> Try deleting this line from your config file:
>
>> builder = "plan9"
>
>
That got me further, thanks. It next dies trying to read then write the
nvram, which is /dev/sd00/nvram in my image. It apparently can't find it.
256# xm create -c plan9.py
Using config file "./plan9.py".
Plan 9
127 holes free
001e5000 04dad000 79462400
79462400 bytes free
cpu0: 1993MHz GenuineIntel PentiumIV/Xeon (cpuid: AX 0x0F27 DX 0xBFEBFBFF)
Started domain plan9
#l0: xen: 100Mbps port 0x0 irq -1: aa001000001b
256M memory: 78M kernel data, 178M user, 803M swap
cpu0: spurious interrupt 101, last 0
can't read nvram: unknown device in # filename
authid: bootes
authdom: fubar
secstore key:
password:
can't write key to nvram: fd out of range or not open
boot: can't connect to file server: '#S/sdC0/' file does not exist
panic: boot process died: unknown
panic: boot process died: unknown
dumpstack
ktrace /kernel/path 80014015 801e71f8 <<EOF
Am I missing something from my bootargs?
Thanks,
Mike
ron minnich
That should bd sd00 on Xen, not sdC0. Where is sdC0 getting set?
ron
Richard Miller
> nvram, which is /dev/sd00/nvram in my image. It apparently can't find it.
You are looking at the old Xen2 instructions on the wiki. I should have
noticed before that you said this:
> I followed these instructions:
> http://www.cs.bell-labs.com/wiki/plan9/Installing_in_Xen/index.html
The very first thing on that page is
These instructions are for Xen 2.0. See
<a href="http://www.cs.bell-labs.com/wiki/plan9/installing_in_xen_3.0">
for Xen 3.0 instructions.
You should get on a bit better with those.
Mike
Thanks, Richard - that was the ticket. I had read those instructions,
but had ended up on the wrong link after migrating around the plan9 wiki
(through no fault of the wiki - I probably hit the "back arrow" once too
many times). ;^)
>
>
Mike
Greetings, ron,
My xen startup file was 99% right - I had the bootargs set from the Xen
3.0 instructions, and the disk setting from the Xen 2.0 instructions.
Once I fixed that, I was in business.
6c6
< disk = [ 'file:/export/2/Guests/Plan9/plan9.img,loop0,w' ]
---
> disk = [ 'file:/export/2/Guests/Plan9/plan9.img,sda,w' ]
15c15
Thanks much for your help.
Mike
Richard Miller
> and see what you think. It's 5K lines of code, the guest support code
> is trivial, and I think overall this is going to be the easiest thing
> to do.
I've just had a quick browse through (the patch is actually 8977 lines,
but maybe the 5k doesn't include comments). My first impression is
"that's more like it". Way simpler.
But:
- Does some of the simplicity come from being linux-specific? Comments
say "guest and host run the same kernel". Or is it generic enough to
work with Plan 9 too?
- My interest in Xen is pragmatic: there are plenty of hosting companies
offering "virtual servers" based on Xen, and some at least are happy
to let you run a Plan 9 guest kernel. (In particular I can recommend
http://www.blackcatnetworks.co.uk from personal experience.) How long
before lguest reaches that level of maturity? Sadly, simplicity in
itself is not much of a marketing advantage in the real world.
-- Richard
ron minnich
> - Does some of the simplicity come from being linux-specific? Comments
> say "guest and host run the same kernel". Or is it generic enough to
> work with Plan 9 too?
it's working up to me calling LHINIT and then LHCRASH.
>
> - My interest in Xen is pragmatic: there are plenty of hosting companies
> offering "virtual servers" based on Xen, and some at least are happy
> to let you run a Plan 9 guest kernel. (In particular I can recommend
> http://www.blackcatnetworks.co.uk from personal experience.) How long
> before lguest reaches that level of maturity? Sadly, simplicity in
> itself is not much of a marketing advantage in the real world.
Ah, so keeping a xen port is good. That's fine. But for THX, Xen is
just too hard.
ron
Richard Miller
Now that's what I call minimalist.