Xorg input thread (2)

[Alan Coopersmith]

Keith Packard wrote:
> On Sat, 2007-07-07 at 11:11 +0300, Daniel Stone wrote:
>> On Fri, Jul 06, 2007 at 10:45:39PM -0700, Keith Packard wrote:
>>> Ultrix systems placed the entire input subsystem in the kernel,
>>> connecting the mouse hardware directly to the cursor position update
>>> function.
>> 'Entire input subsystem' is a no-go, obviously.  Particularly if we're
>> going to actually use this input transformation stuff ...
> 
> By "entire input sybsystem", I mean only up to the point of creating a
> root window coordinate. That's before input transformation. Effectively,
> the portion of the server currently running from the signal handler
> would run in the DRM instead. We can look at the old Ultrix code to see
> how that worked, it wasn't that complicated iirc. Transforming the
> relative mouse motion into an absolute screen position isn't that
> complicated, but it does involve acceleration and cursor limits. Much of
> the existing server input architecture is structured to handle precisely
> this separation of responsibilities.

If it's useful, I can also provide the similar code from Solaris/Xsun,
which uses the SysV kernel streams architecture to push a "hwc" module on
top of the mouse stream that sends requests directly to the kernel fb driver
to move the cursor image without the Xserver process ever having to be
scheduled.   hwc has it's own set of ioctl's for the X server to pass
down parameters like bounding boxes and acceleration.


-- 
	-Alan Coopersmith-           alan.coopersmith at sun.com
	 Sun Microsystems, Inc. - X Window System Engineering