Re: [Xen-devel] VT-D CPU ressources

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 03.07.12 at 13:09, "amiard brice" <perso@xxxxxxxxxxxxxxx> wrote:
> I've post a message on  <mailto:xen-users@xxxxxxxxxxxx>
> xen-users@xxxxxxxxxxxx , according to a member of this list, I've decide to
> post it on xen-devel.

While re-posting here with this sort of problem is fine, you
should have done so in a way that's easy to read and understand.
Neither does top-posting help this, nor do the awful extra blank
lines that must have got inserted somewhere.

Further, while you attach quite a bit of information, you left us
in the dark as to what device combinations you want to assign to
individual guests (and what co-assignment error messages you
specifically get). From what I can guess so far, there's no
obvious 2-element groups that could be assigned to individual
guests. That's because devices behind PCI bridges need to all
be assigned to the same guest, as otherwise the IOMMU can't
tell apart the origin of requests from the different devices.

You could try with PV guests (and that may additionally require
turning off the IOMMU). There's also a "pci-passthrough-strict-check"
config option to suppress these checks, but using it would
make your entire setup insecure (between the individual
guests).

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel