Re: RHEL (CentOS, Oracle Liux) kernel 2.6.32-358 does not start at XCP

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 07/03/13 10:26, Constantin Dunayev wrote:
> Hi, everyone,
> Two days ago I have filed a bug to Red Hat
>  https://bugzilla.redhat.com/show_bug.cgi?id=918239
> The problem is, that new RHEL-6.4's kernel 2.6.32-358 does not start
> at HVM on Xen Cloud Platform 1.6 (maybe, also affected XenServer 6.1)
> I have tested kernels -358 from CentOS, Oracle and -356 from Red Hat
> both i686 and x86_64 at 2 different server hardware
> The sympthoms: kernel does not boot at all, does not produce any log;
> previous (all 2.6.32-279 and 2.6.32-220), and Oracle's kernel-uek work
> perfectly.
>
> Andrew Jones from Red Hat advised me to ask, does anybody else have
> this issue?
>
> Best regards,
> -- 
> Constantin Dunayev,
> Kharkov, UA  

To follow up on this issue.  My findings are presented in full on the BZ
ticket, but in summary:

RHEL 6.4 appears to be somewhat viridian aware

The final message on the console before the VCPU spins is

"Switching to clocksource hyperv_clocksource"

and is presumably waiting for an interrupt from a clocksource we do not
implement.

Disabling viridian for the domain appears to work around the issue:

For XenServer,XCP use:

xe vm-param-set vm=$VM platform:viridian=false

For xl/xm, ensure viridian=false in the VM configuration file

~Andrew