> Ahh. So it's easily repeatable for you.. Do you think you could bisect it=
?=20
quoted text
> At least a few runs? Even a partial bisection will give a nice (=3D=3D mu=
ch=20
quoted text
> smaller) range of commits to be blamed, and might tell us why it started=20
> happening..

Unfortunately, it takes forever on this machine to compile a kernel
after any bigger changes. I can do it early next week when I have access
to my bigger box again.

On the other hand, it should be easily reproducible by anyone else with
the same trick, here's what I do:

 * configure X to use /dev/input/event* devices
 * in an xterm, do something like
      rmmod usbhid ; modprobe usbhid
 * switch to a VT
 * watch kernel crash as X releases the grab on the event device

Mind you, I actually did this with the appletouch driver, but I don't
think it makes a difference since the problematic thing is the code that
X grabs the device. A trivial program that executes the grab ioctl would
probably suffice if you leave it running over a rmmod/modprobe cycle.

johannes