X server can be crashed by a badly timed kill -9 of a DRI2 client

[Simon Farnsworth]

We've been hitting an issue in the lab where killing off an active
DRI2 client results in the X server going down hard, with a consistent
backtrace. A bit of investigating results in this patch, although I'm
not sure that this is the right way to fix it.

The root cause is that DRI2 is still processing a request when the
XTrans request structure is set to NULL; if the timing's just right,
which appears to be fairly frequent in our lab, the call to
LocalClient fails. The patch just works round this failure mode -
there may be a better way to fix it, but this is simple and not very
intrusive.