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

Re: [Xen-devel] Intel VT-d/Neocleus 1:1 mreged code for PCI passthrough

On 01/09/2009 10:41, "Jun Koi" <junkoi2004@xxxxxxxxx> wrote:

> On Sun, Aug 30, 2009 at 4:50 PM, Keir Fraser<keir.fraser@xxxxxxxxxxxxx> wrote:
>> For passthrough to HVM guests without hardware IOMMU (e.g., VT-d), I
>> developed the attached prototype patch. I think this may already be in the
>> XCI tree. It also applies directly to Xen 3.4 branch, and could be applied
>> quite easily with a small amount of manual intervention to xen-unstable. If
>> passthough on 'old' system without hardware assistance is what you're after,
>> you might want to give the patch a go.
> Could you confirm that this patch alone is enough?
> I applied the patch against unstable code, and Xen hang up (or seems
> to hang up) in the middle of booting up Xen, with a black screen.

Try the attached patch. Also note that this approach only works for one HVM
domain (the first you create) and that this domain needs to be allocated all
of low physical memory -- this means for example that dom0 cannot allocate
this memory for a swiotlb and you may need 'swiotlb=off' on dom0's kernel
command line.

You can check in /var/log/xen/xend.log to see whether you created a
passthru-capable HVM guest successfully. There should be a message saying
'Domain <foo> is nativedom.' or 'Domain <foo> is NOT nativedom.'

 -- Keir

Attachment: 00-nativedom-for-unstable
Description: Binary data

Xen-devel mailing list



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