RE: [Xen-users] pci passthrough under xen-3.3 without VT-d

["Benjamin Huntsman" <BHuntsman@xxxxxxxxxxxxxxxxxxxxx>]

>Yes, as long as you understand the security implications (i.e driver
>DMA to memory it doesn't own)

No worries... this isn't production, so who cares! :)


>I am pretty sure that you don't want the iommu line in here, since you
>don't have an iommu

Okay, I'll take it out, but I had put it in originally because it wasn't 
working...

>instead of doing the modprobe below, you could instead pass the pciback.hide
>on the dom0 command line. It should give a similar effect, but still
>worth a shot.

pciback.hide only works if you have pciback built static into the kernel.  I 
don't.  I have it built as a module.  Because of this, you have to make it 
preload via the initrd, and it's config gets specified in /etc/modprobe.conf.  
I might try building it in static, though I don't really like that option... :)

>I think you are getting this error because you passed iommu=1

It was giving that error before I added iommu=1.  I added it because of this 
error in the first place, as I had read that it was required in order to handle 
PCI-passthrough.


>You can't pass to a HVM domain without VT-d. You can only pass to a PV guest.

First, I want to thank you for responding, as I appreciate your assistance.  
But, are you 100% sure of that?  I've never seen that written so 
black-and-white in any of the Xen documentation...  From what I could cull from 
the docs and mail lists, it might not work correctly if the guest doesn't know 
how to talk to Xen, the hvm qemu-dm ought to at least recognize it's got some 
extra PCI card...


>Hope that helps.
>
>Cheers,
>Todd

Indeed!  Many thanks!  Take care!

-Ben


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