[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] pci passthrough and svm



I can not pass through PCI with out VT-d or equivalent
AMD doesn't have a VT-d equivalent out yet
You must use VT-d to pass through PCI devices to an hvm
Windows must be run under hvm.

Hence I can not pass PCI devices through to Windows.

Paul Schulze wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Forgot to CC, so here is a copy of my answer.

Forwardl:

Von: Paul Schulze <avlex@xxxxxxx>
Datum: 10. September 2008 19:33:57 MESZ
An: mpower@xxxxxxxxxxxxxxxxxx
Betreff: Re: [Xen-users] pci passthrough and svm

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Mike,

The bad news (as far as my understanding of this goes, no clue if its technically correct): At the moment, there is no AMD equivalent for VT-d. AMD SVM is a CPU virtualization extension, comparable with IVT. VT-d, is an IOMMU, that probably resides with the memory controller and handles virtualization and protection of DMA. So in case of an Intel CPU on the mainboard (no clue if it needs special support on the CPU on top of that). AMD first announced to implement an equivalent to that technology quite a few years ago, but up until now there are no products I can think of that include such a feature. However, CPUs, supporting it are rumored to be released early in 2009, so there is hope (for me too, because my Xen server is built on AMD hardware as well).

And the good news: Even though there is no IOMMU support on AMD platforms, it is possible to pass PCI hardware through to a paravirtualized DomU. This feature does not depend on VT-d, with the exception of PCI passthrough to HVM domains (to Windows for example). Like I said, I am on an AMD platform and I have succeeded in passing though multiple devices, including USB controllers, audio devices and network devices (no graphics cards though). However, when passing PCI devices to a paravirtualized domain, always remember that there is no protection against faulty or malicious memory access due to the missing IOMMU, enabling the DomU to potentially access even Dom0 memory.

Cheers,


Paul.
- --
Paul Schulze
Mail: avlex  gmx  net ($1@$2.$3)
Public Key: http://solaris-net.dyndns.org/keys/key_avlex.asc

"Making mistakes is human,
but to really screw things up, you need Computers"



Am 10.09.2008 um 17:50 schrieb Mike Power:

I noticed talk of pci passthrough seems to always be in reference to VT_D which from my understanding is an intel technology. Is there equivalent and xen enabled functionality from AMD's svm technology?

Mike Power

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

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Darwin)

iD8DBQFIyAUFYDWOGtiChoARAoVQAJ47bVjaeE8qG8c7Pa52DoVQPdx+BwCeKXuX
DYlLbR27PkMtF+hTxiaJjK8=
=UEFn
-----END PGP SIGNATURE-----


- --
Paul Schulze
Mail: avlex  gmx  net ($1@$2.$3)
Public Key: http://solaris-net.dyndns.org/keys/key_avlex.asc

"Making mistakes is human,
but to really screw things up, you need Computers"

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Darwin)

iD8DBQFIyPaDYDWOGtiChoARAt5zAJ4sKWolvJ+N2D8klsFsNNm0mpkh0ACeJ4ba
ma4PGCeBLwbWYQIsYy+vwro=
=0mAj
-----END PGP SIGNATURE-----

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


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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.