multiseat (was: Re: Input Devices (was: Re: [Xgl/Xegl] Input Devices))

Joe Krahn krahn at
Thu Aug 25 10:00:21 PDT 2005

David Balažic wrote:
> On Mon, 22 Aug 2005 at 00:12:55, Adam Jackson wrote:
>>On Sunday 21 August 2005 12:09, David Balažic wrote:
>>>Here is what I would expect for multiseat in next version of X :
>>>  - improved video driver behavior in multi-card setups (are the problems
>>>for multiseat different than for "classic" multimonitor/multihead setup ?)
>>This is not a fun problem to solve, it involves lots of PCI routing ickiness,
>>but people are looking at it.
> Can someone clarify : Are those problems the same for multiseat as for "normal"
> multihead/multimonitor ? Afterall, multiple gfx cards are supported by Windows
> and X sice ages. And it usually works.
>>>  - option for X not to touch VT system (so the first seat would have
>>>classic VTs, while the other would just run X)
>>Hrm, yep, definitely have that now:
>>~ % X -help 2>&1 | grep vt
>>vtXX                   use the specified VT number
>>-novtswitch            don't immediately switch to new VT
>>-sharevts              share VTs with another X server
>>(The help text for -novtswitch is slightly misleading, it disables VT switch
>>functionality for that server entirely.)
> Are you sure ? The last time I looked into the "-novtswitch" patch, is still
> left some VT related code enabled. It also messed up the first display, when I
> started X with -novtswitch in the second.

The problem I had trying to run multiple servers some time ago was that 
they both would probe all PCI devices, even if the config pointed to 
just one. Maybe this is fixed now.

More problems arise when trying to run multiple servers sharing a single 
VT. The XInput drivers don't always release input devices to share them 
with another switched-in X server. Also, XFree86 always restored video 
hardware to "the way it was found", so if server 2 initializes while 
server 1 is active, then server 2 does not know how to correctly restore 
the VT text mode. But, maybe this has been fixed.

Joe Krahn

More information about the xorg mailing list