[Xen-users] VT-d passthrough strangeness

[Paul Stimpson <paul@xxxxxxxxxxxxxxxxxxxx>]

Hi,

I'm working on an HP Z420 workstation. I have an Ubuntu hypervisor with 
a home-compiled copy of Xen 4.2.0. I'm trying to pass through a pile of 
hardware into a Windows Server 2008 R2 guest.

    * (00:1a.0) - Motherboard USB 1/2 controller
    * (00:1b.0) - Motherboard audio
    * (00:1d.0) - Motherboard USB 1/2 controller
    * (05:00.0) - Second AMD Firepro V3900 graphics card
    * (05:00.1) - The HDMI audio output on the above card
    * (06:00.0) - Video playback card
    * (08:00.0) - Motherboard USB 3 controller

The 00:1b.0, 05:00.0, 05:00.1 and 06:00.0 go through to the guest fine. 
The USB controllers are more problematic and don't get passed through 
when the Windows guest starts in /etc/xen/auto.
I've tried everything I can think of:
    * I put pciback into initrd.
    * The hide is listed both in the kernel boot line and in 
/etc/modprobe.d/pciback
    * I listed the unique IDs (obtained from 
/sys/bus/pci/devices/..../modalias) in /etc/modprobe.d/blacklist
    * I created a new script in /etc/rc5.d between xencommons and 
xendomains that both echoes the PCI BDF into 
/sys/bus/pci/devices/....../driver/unbind and does an xl 
pci-assignable-add on the BDF.
If I stop the Windows guest, do an xl pci-assignable-add on each of the 
devices then restart the guest, they go through fine.
I've been forbidden from patching Xen or recompiling the kernel by 
someone I must listen to.
I would be really grateful if you could give me an idea on how to make 
this thing work as I'm starting to feel really stupid.
Thanks,
Paul.


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