Re: Re: Problems with guest mouse Was: More issues with xl on 4.1.0-rc

[John Haxby <john.haxby@oracle.com>]

On 08/03/11 22:03, M A Young wrote:
> On Mon, 7 Mar 2011, Stefano Stabellini wrote:
>
>> On Sun, 6 Mar 2011, M A Young wrote:
>>
>> Is there any difference in the VMs between the system that works and the
>> one that does not? In particular I am think about kernel or xorg
>> versions.
>
> I think this could be it. The mouse doesn't work on newer VMs but does
> work on older ones (I am not sure my earlier test results quite fitted
> that pattern but I could be mis-remembering or have been mis-testing).


I thought this sounded familiar.  In the bad log you have

(==) Using system config directory "/usr/share/X11/xorg.conf.d"
...
(II) config/udev: Adding input device Xen Virtual Pointer
(/dev/input/event1)
(**) Xen Virtual Pointer: Applying InputClass "evdev pointer catchall"
(II) Loading /usr/lib64/xorg/modules/input/evdev_drv.so
(**) Xen Virtual Pointer: always reports core events
(**) Xen Virtual Pointer: Device: "/dev/input/event1"
(--) Xen Virtual Pointer: Found 12 mouse buttons
(--) Xen Virtual Pointer: Found scroll wheel(s)
(--) Xen Virtual Pointer: Found relative axes
(--) Xen Virtual Pointer: Found x and y relative axes
(--) Xen Virtual Pointer: Found absolute axes
(--) Xen Virtual Pointer: Found x and y absolute axes
(II) Xen Virtual Pointer: Configuring as mouse
(II) Xen Virtual Pointer: Adding scrollwheel support
(**) Xen Virtual Pointer: YAxisMapping: buttons 4 and 5
(**) Xen Virtual Pointer: EmulateWheelButton: 4, EmulateWheelInertia:
10, EmulateWheelTimeout: 200
(**) Option "config_info" "udev:/sys/devices/virtual/input/input1/event1"
(II) XINPUT: Adding extended input device "Xen Virtual Pointer" (type:
MOUSE)
(II) Xen Virtual Pointer: initialized for relative axes.
(WW) Xen Virtual Pointer: ignoring absolute axes.
(**) Xen Virtual Pointer: (accel) keeping acceleration scheme 1
(**) Xen Virtual Pointer: (accel) acceleration profile 0
(**) Xen Virtual Pointer: (accel) acceleration factor: 2.000
(**) Xen Virtual Pointer: (accel) acceleration threshold: 4


The critical bit it "ignoring absolute axes".   If you look at the evdev
man page it says

                                                                  The X
              server cannot deal with devices that have  both  relative
              and  absolute  axes.  Evdev  tries  to guess wich axes to
              ignore given the device type and disables  absolute  axes
              for  mice and relative axes for tablets, touchscreens and
              touchpad.

Your "good" log file indicates that something was done about this in the
configuration, but the easiest thing to do is to put this in
/usr/share/X11/xorg.conf.d/xen.conf (that file doesn't exist):

Section "InputClass"
    Identifier "Xen Virtual Pointer"
    MatchProduct "Xen Virtual Pointer"
    Option "IgnoreRelativeAxes" "True"
EndSection

When you start up the X server the log file will show that this option
has been taken and all will be sweetness and light.

The underlying problem seems to be that the xen virtual mouse driver is
advertising both relative and absolute axes (duh) but the back end is
only supplying absolute axes.

I haven't had the time to look since I discovered this a week ago -- but
I think that the front end shouldn't be advertising relative axes if the
back end is supplying absolute ones.   Perhaps now I should actually
look ...

jch