dom0 crash: xen 3.4.0 / netbsd-5 domU kernel

[Florian Heigl <florian.heigl%gmail.com@localhost>]

Hi,

I just want to follow up on a list posting of 22.10.2009, with a
little bit of extra info.
Sorry for this not being in the right thread but I didnt change the
mailing list over to the new mail account.

- My system in question is a Quad Xeon 700, similar to the other poster's
- it also has 8GB ram (someone noted about bootup only being
successful under 4GB)
- I run Oracle VM 2.1 different from him.
- The system completely freezes when the domU tries to setup it's device files
- Suggestions about sysrq or debugging Oops are useless. There IS no
oops or anything like that.
- The other poster had a completely blank screen, this is due to
screen blanking being in effect, if one retries the domU boot while
the screen still has text one just sees it freezing
- my box has a button for triggering NMIs which, just as sysrq, has NO
effect after the system stops responding

i'm quite sure it IS the hypervisor that gets frozen and I dare say
that it's wrong that only a faulty dom0 kernel could crash the whole
system. I've seen it a few times in the time I use xen, i.e. with the
vmsplice exploit fired in a domU on linux you would get the same
result, I saw it a few times in NetBSD 3.99 times, too and imho the
dom0 hasn't much to say in it anyway the way xen works.

So please note:
At least on linux, using p3 xeon processors with over 8GB ram it
appears that the standard 5.0 domU kernel will fry the hypervisor.

I saw a few mails from later on that handled dom0 crashes but it
seemed to be an unrelated issue.


I can help trying to debug this and I'll retest on a current Q6600 box
soon, but beyond that it is supposedly hard, without any output :)


-- 
'Sie brauchen sich um Ihre Zukunft keine Gedanken zu machen'