Re: [Xen-API] PCI passthrough limit of device?

[James Bulpin <James.Bulpin@xxxxxxxxxxxxx>]

Melvin wrote:
> But for VT-c, not VT-d.
> Does VT-c NIC require SR-IOV in bios, or VT-c NIC can be use only with
> VT-d,
> without SR-IOV feature.

VT-c is Intel's collective name for virtualisation technologies for
networking including SR-IOV enabled NICs and VMDQ. If you are intending to
use SR-IOV then both the NIC and BIOS must support SR-IOV.

You can use VMDQ without BIOS SR-IOV support but that will only allow
multiple queues to dom0, not pass-through to VMs.

Cheers,
James

> -----Message d'origine-----
> De : James Bulpin [mailto:James.Bulpin@xxxxxxxxxxxxx]
> Envoyé : 22 octobre 2012 08:16
> À : Melvin B.
> Cc : xen-api@xxxxxxx
> Objet : RE: [Xen-API] PCI passthrough limit of device?
> 
> (Adding the list back to CC - please don't remove the list; by keeping
> the discussion public others may be able to benefit from the
> discussion.)
> 
> Melvin wrote:
> > My last question is about intel VT-c and SR-IOV NIC.
> > If SR-IOV NIC require a Motherboard feature other than VT-d ?
> 
> Your BIOS must support SR-IOV.
> 
> > And if VT-c is for device only, require only a motherboard with VT-d?
> 
> You can perform PCI pass-through to a PV guest without VT-d/IOMMU but
> that
> is unsafe (e.g. the device could DMA to other guests' memory) so VT-d
> is
> recommended. If you want to pass-through to a HVM guest you must use
> VT-d/IOMMU.
> 
> See http://wiki.xen.org/wiki/Xen_PCI_Passthrough for more detail.
> 
> Cheers,
> James


_______________________________________________
Xen-api mailing list
Xen-api@xxxxxxxxxxxxx
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api