Re: [Xen-users] Decent motherboard with VT-d (Intel) support?

[Mark Williamson <mark.williamson@xxxxxxxxxxxx>]

> >However, I was wondering if anyone had success stories or suggestions on a
> > motherboard which support >VT-d on Intel boards or if there are any for
> > AMD CPU's for their equivalent.  (I keep seeing references to >IOMMU for
> > AMD but haven't seen any boards out supporting it in my brief search).  I
> > would like to test PCI >passthru to HVM's.
>
> I am about to do this on a new DL380G5 with the newer 5000P chipset
> supporting the Quad Cores. As far as pass through to an HVM, I understand
> that needs the pv drivers which aren't yet stable.

The idea behind passthrough to HVM guests using IOMMU is that it should be 
possible without an PV drivers being installed in the guest.  The IOMMU 
functionality takes care of making the memory layout look like a fully 
virtualised guest expects it to, protecting the rest of memory from the 
guest, etc...

There has also been code from Neocleus for HVM guest passthrough without IOMMU 
but again this didn't require PV drivers in the guest.

I think that VT-d passthrough of PCI devices to HVM guests may be possible in 
the next release (if it isn't possible in the previous one).  There were 
later patches submitted to allow the VT-d IOMMU to provide protections when 
passing PCI devices through to PV guests too (without this, giving a PV guest 
a PCI device implies giving it equivalent power and trust as dom0 itself, 
since it could abuse the device to take over the machine) but I don't recall 
the status of these.

Cheers,
Mark


-- 
Push Me Pull You - Distributed SCM tool (http://www.cl.cam.ac.uk/~maw48/pmpu/)

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users