On Mon, Nov 30, 2015 at 03:53:56PM +0100, Martin Kletzander wrote: > On Mon, Nov 30, 2015 at 01:25:25PM +0100, Pavel Hrdina wrote: > >On Mon, Nov 30, 2015 at 09:14:11AM +0100, Martin Kletzander wrote: > >> On Fri, Nov 27, 2015 at 02:36:09PM +0100, Pavel Hrdina wrote: > >> >On Fri, Nov 20, 2015 at 09:59:35AM +0100, Ján Tomko wrote: > >> >> Support virtio-keyboard, virtio-mouse, virtio-tablet and virtio-input-host. > >> >> > >> >> Requires kernel 4.1+ in the guest. > >> >> > >> >> https://www.kraxel.org/blog/2015/06/new-member-in-the-virtio-family-input-devices/ > >> >> https://bugzilla.redhat.com/show_bug.cgi?id=1231114 > >> >> > >> >> Ján Tomko (8): > >> >> qemu: add capabilities for virtio input devices > >> >> conf: parse and format virtio input bus in domain XML > >> >> qemu: build command line for virtio input devices > >> >> qemu: add capability for virtio-input-host-device > >> >> conf: add XML for input device passthrough > >> >> security: label the evdev for input device passthrough > >> >> qemu: add passed-through input devs to cgroup ACL > >> >> qemu: build command line for virtio-input-host device > >> >> > >> > > >> >ACK series. > >> > > >> >I have only one thing to mention, but it's not in scope of this series, it's > >> >an existing behavior. If you define a new guest with mouse/keyboard input > >> >device with bus!=ps2 and graphics is defined, then we always add implicit mouse > >> >and keyboard with bus=ps2 which leads to having two different mouses and > >> >keyboards defined for the guest. We should probably update the code to not add > >> >any implicit mouse or keyboard if there is already one. > >> > > >> > >> I don't think it's that easy. We add it to the XML because when we > >> start QEMU with any graphics, it also automatically gains an > >> additional mouse and keyboard. We con't specify it on the command > >> line, so we're merely trying so that the XML matches what the domain > >> looks like. > >> > > > >In that case we should put the ps/2 mouse and keyboard always in the domain XML > >to represent actual state of the domain. > > Or find out how to disable those in QEMU so that we don't double them > (that apparently causes some issues, too). Well, that's the thing, we cannot disable it for QEMU [1]. The best solution is to put those implicit input devices only into live XML. This will result in config XML that contains input devices defined by user, but live XML will add those implicit devices if necessary. [1] https://www.redhat.com/archives/libvir-list/2015-November/msg01253.html -- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list