CentOS virt - Nov 2009 - Xen pci passthru problems with kernel -164.6.1

I previously reported this on the centos mailing list:

http://lists.centos.org/pipermail/centos/2009-November/085672.html

And I've found out that Red Hat has backported the VT-d support from Xen 
3.3 to RHEL 5.4.

http://lists.centos.org/pipermail/centos/2009-November/085677.html

It seams to me that the "classic" Xen pci-passthru (up to Xen 3.2)
works
only on some minor cases as described here (when using the new kernel 
and hypervisor):
https://bugzilla.redhat.com/show_bug.cgi?id=514458#c4

The recommendation seams to be (although not stated at any 
documentation) to use the VT-d support for various reasons (better 
security for the guests when accessing the hardware I guess).

My concern is that the hardware I use does not support VT-d (it is a 
Intel 5000P chipset, ~2 years old) so I believe I'm kind of screwed. Or 
keep using the kernel and Xen packages from 5.3 (not a good option either).

Am I the only one bitten by this?

On Tue, Nov 10, 2009 at 01:57:42PM -0200, Miguel Di Ciurcio Filho
wrote:
> I previously reported this on the centos mailing list:
> 
> http://lists.centos.org/pipermail/centos/2009-November/085672.html
> 
> And I've found out that Red Hat has backported the VT-d support from
Xen
> 3.3 to RHEL 5.4.
> 
> http://lists.centos.org/pipermail/centos/2009-November/085677.html
> 
> It seams to me that the "classic" Xen pci-passthru (up to Xen
3.2) works
> only on some minor cases as described here (when using the new kernel 
> and hypervisor):
> https://bugzilla.redhat.com/show_bug.cgi?id=514458#c4
> 
> The recommendation seams to be (although not stated at any 
> documentation) to use the VT-d support for various reasons (better 
> security for the guests when accessing the hardware I guess).
> 
> My concern is that the hardware I use does not support VT-d (it is a 
> Intel 5000P chipset, ~2 years old) so I believe I'm kind of screwed. Or
> keep using the kernel and Xen packages from 5.3 (not a good option either).
> 
Yeah.. VT-d support is only on most recent chipsets, and many BIOSes
still have broken implementations of it :(
> Am I the only one bitten by this?
>
Yeah I guess.. so far..

-- Pasi