[Pkg-xen-devel] Bug#617421: Bug#617421: Bug#617421: xen-hypervisor-4.0-i386: can't migrate (or save/resume) lenny guests: BUG: recent printk recursion

[Ferenc Wagner]

Ian Campbell <ijc at hellion.org.uk> writes:

> The guest visible hypervisor ABI has been stable since (IIRC) Xen 3.0.0
> and Xen upstream maintains backwards compatibility for guest visible
> interfaces -- if an old guest was broken by a hypervisor update then
> this is most likely a bug.

Agreed.

> On the other hand, the Xen enabled kernel in Lenny turned out to be a
> bit of a disaster and I could well imagine that the new hypervisor/tools
> are simply exposing a prexisting bug in that kernel.

It wasn't entirely explicit, but I don't use (and haven't ever used) the
Xen-enabled lenny kernel: it was way too unstable for production use
from the very beginning.  The guests I'm talking about run the latest
linux-image-2.6.26-2-686-bigmem kernel (version 2.6.26-26lenny2),
sporting upstream Xen domU support, and no dom0 support.  The dom0 was
running the "latest" etch Xen kernel, linux-image-2.6.18-6-xen-686
(version 2.6.18.dfsg.1-26etch2).

> Due to that and the fact that Lenny is now oldstable you may find
> there is not much interest in tracking this issue down.

The latter part of course still applies, and I can accept that.  This
report can then serve documentation purposes at least. :)

> Ferenc should definitely make sure you are running the most recent Lenny
> kernel though, the bug report doesn't include the version of this
> package AFAICT.

Sorry, I left that out indeed.  It's the latest lenny kernel (Jan 27).

> Failing that running a Squeeze kernel from the lenny-backports suite
> is a good option IMHO.

On some machines it worked fine, on others it broke the iSCSI root setup:

[    1.267399] iscsi: registered transport (tcp)
iscsistart: transport class version 2.0-870. iscsid version 2.0-870
iscsistart: Logging into **************************************** *************************
iscsistart: can not connect to iSCSI daemon (111)!
iscsistart: initiator reported error (18 - could not communicate to iscsid)

Which is rather strange, because the point of iscsistart is that it does
not require iscsid...  Anyway, it's not Xen-related, and I'll hopefully
find a way out.
-- 
Thanks,
Feri.