Re: VM hangs during boot on 3.0.4 dom0 kernel, works on alternative hardware

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Tue, Sep 13, 2011 at 10:52:40AM +0100, Anthony Wright wrote:
> 
> 
> ----- Original Message -----
> > On 06/09/2011 17:16, Konrad Rzeszutek Wilk wrote:
> > > On Mon, Sep 05, 2011 at 11:31:16AM +0100, Anthony Wright wrote:
> > >> I have two machines with identical Dom0's and DomUs, but different
> > >> hardware. The Dom0 has a patch which I produced myself based on the
> > >> "Re:
> > >> [Xen-devel] Kernel bug from 3.0 (was phy disks and vifs timing out
> > >> in
> > >> DomU)" thread. The patch calls vmalloc_sync_all after every
> > >> alloc_vm_area, and I realise this isn't the best solution, but it
> > >> allowed me to move forward.
> > > <nods>
> > >> The patch fixes the problem I had on one machine, so that now the
> > >> VMs
> > >> boot correctly, but I have another system with an identical setup
> > >> (identical Dom0 & DomU kernels, identical startup for DomU) and the
> > >> VM
> > >> fails to start. I have attached a copy of the console log from the
> > >> good
> > >> VM and the bad VM.
> > > And what does the dom0 and xen hypervisor log give you?
> > >
> > > It looks to be hanging at identifying the CPU - is the hardware
> > > quite different from one setup to another? Have you toyed with
> > > using the cpuid flag in the guest to mimic the lowest CPU type?
> > I've found a solution to my problem. The PV DomU was running a 32 bit
> > linux kernel version 2.6.30.1. I hoped that a new kernel would be able
> > to handle to CPU more effectively, and so I upgraded my DomU kernel to
> > 3.0.4, and now the DomU boots.
> > 
> > This solves the problem for me, but I'm slightly concerned that my
> > original kernel wouldn't work with this CPU (the CPU type for the
> > kernel
> > builds is set to Pentium Pro). I'll try the 2.6.30.1 kernel on bare
> > metal on monday to try to see if this is a xen/Dom0 kernel issue or a
> > CPU/DomU issue.
> > 
> > Anthony
> I've tried the 2.6.30.1 kernel on bare metal and it boots correctly, so it looks like some xen/Dom0 interaction is causing it to fail when run as a DomU. I have a work around (upgrading to 3.0.4 kernel), so it's not affecting me but I'm happy to do any diagnostics to help identify the cause of the problem.
> 

That is OK. I don't we would have the time to actually figure out such
and old kernel, or to backport the patch.