Xen users - Nov 2006 - Does HVM( AMD & Intel''s VT ) support direct device access w/o guest modification?

I wonder, does the Xen hypervisor that uses the AMD or Intel hardware
virtualization( svm or vti ) support direct device access without guest OS
modification? As I understand the pciback module requires the guest
OS''s(
domU ) drivers modification.

-- 
Slava Imameyev


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

> -----Original Message-----
> From: xen-users-bounces@lists.xensource.com 
> [mailto:xen-users-bounces@lists.xensource.com] On Behalf Of 
> Slava Imameyev
> Sent: 14 November 2006 16:13
> To: xen-users@lists.xensource.com
> Subject: [Xen-users] Does HVM( AMD & Intel''s VT ) support 
> direct device access w/o guest modification?
> 
> 
> I wonder, does the Xen hypervisor that uses the AMD or Intel 
> hardware virtualization( svm or vti ) support direct device 
> access without guest OS modification? As I understand the 
> pciback module requires the guest OS''s( domU ) drivers
modification.
Not at present. There are two issues here:
1. PCI device needs to be inserted into the device table of the HVM
guest. This is code that any decent programmer can do in a few hours. 
2. PCI devices that do direct memory access (i.e. almost all of them)
will need the address translated from the guest''s belief of what the
memory address is, compared to the actual machine address (OS believes
that memory starts at 0 and goes to 256MB, but a guest OS may live at,
say 512..768MB). The easy solution to this is to have IOMMU-hardware
(and related support in the HVM part of the hypervisor). 

--
Mats
> 
> -- 
> Slava Imameyev
> 


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