[Pkg-xen-devel] Re: [Pkg-xen-changes] r55 - trunk/debian [u]

[Ralph Passgang]

Am Sonntag, 26. Februar 2006 00:21 schrieb Julien Danjou:
> On Sat, Feb 25, 2006 at 07:37:34PM +0100, Guido Trotter wrote:
> > Mmm... so for you it only works with udev from backports.org...  For me
> > instead it only works if I only have hotplug from sarge... Mmmm... good
> > shape... :(
>
> I also tried to remove udev and leave hotplug, and it did not work.
> But it was maybe me.

you have to reboot then? I think it's needed, but maybe I just don't know what 
to do to get "hotplug" activated after udev is  removed.

I have all my sarge setups running just with hotplug, so that should be ok.

we maybe should really conflict with udev (< 0.059) then? that would prevent 
that users even try the sarge udev version.

> > And if you didn't what happened? We cannot count on users doing that!
> > (Even if it's only from unstable...)
>
> If I did not tweak the prerm script, the upgrade cannot happen since the
> prerm script fails.
>
> That's the:
> invoke xend stop
> invoke xendomains stop
>
> that fails, I think we know why :-)
>
> > Was it a problem if it failed? Did it refuse to complete the upgrade?
>
> Yes, if a post/pre rm/inst script fails, apt stops.
>
> > In that case we should probably, if we can fix this... The best way to do
> > it would be to only start xend if the current hypervisor is a supported
> > version... Is there any way to check the version of the running
> > hypervisor?

if xend is running, then this should give out the xen hypervisor version:
# xm dmesg |grep "Xen version" | awk '{print $3}'

I am not sure what version "xm info" reports. is it the userspace or 
hypervisor version? I think userspace, but not sure

> It can be checked in python.
> Or we can also make "xend start failure" non fatal with something like:
> invoke xend start || true
>
> (We should check policy on that sorts of ugly things :-)
>
> > > However it is not working for me, it seems I have a kernel/hardware
> > > problem with the kernel I compiled... :-/ I downgraded back
> > > (successfully *g*) to Xen 2.0.6 for now... But this if off topic.
> >
> > I'm sorry! :(

if possible try to stay with the default config as far as possible. it will 
work on the most system and it will produce a kernel useable for dom0 and 
domU.

You can also try "linux-xen0-2.6" from packages.debianbase.de for testing, so 
you not need to recompile a kernel again.

I had some users reporting problems if they changed the default config and 
also the xen ml suggest to stay as close as possible to the default config.

> For the record, I solved it by booting the xen hypervizor with noapic
> and acpi=off options, and after that I had a UDP checksum error, so I
> turned off tx checksum (ethtool -K eth0 tx off). :-/
> Thanks googling.

you really checked 3.0.1? I thought at least the udp checksum problem was 
fixed with 3.0.1. hmmm. I never saw this problem on any system myself.

--Ralph