On Fri, Dec 20, 2013 at 10:17:22AM +0200, Laine Stump wrote: > On 12/19/2013 07:41 PM, Ján Tomko wrote: > > On 12/19/2013 06:18 PM, Jiri Denemark wrote: > >> On Thu, Dec 19, 2013 at 17:43:33 +0100, Jano Tomko wrote: > >>> Don't set the PCI address of primary video to the default > >>> if it's already occupied by another device. > >>> > >>> Without this, a primary video card that doesn't have an > >>> address gets the default one on the 1st pass of PCI address > >>> allocation, even if it's occupied. This address gets picked > >>> up by the second pass, and a conflict occurs: > >>> > >>> XML error: Attempted double use of PCI slot 0000:00:02.0 > >>> (may need "multifunction='on'" for device on function 0) > >>> > >>> Also fix the test that was supposed to catch this. > >> And does it actually work after this patch? IIRC, the primary video card > >> is forced to be at 0000:00:02.0 by QEMU. That is, if this address is > >> occupied by another device, there should be no way to add a video card. > >> > > It works without this patch too, if you explicitly specify the PCI address of > > the video card (and you have QEMU >= 1.6, because that's when we use -device > > instead of -vga - (the QEMU_CAPS_DEVICE_VIDEO_PRIMARY capability)). > > Isn't there also something about multi-display support that requires the > primary video to be at a particular address? FWIW, the only reason why I had hardcoded display originally was because we were forced to use -vga instead of -device, and so could't control the PCI port. Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list