The fact that we now explicitly set fepriv->exit = 1 when the thread is
shutting down exposed an edge case where it was not being reset back to zero
once the thread went away in some cases.  This resulted in failures in cases
where the frontend was closed, and then opened O_RDONLY, since in that case
the thread is not being restarted but it was checking the fepriv->exit flag.

Thanks to Thierry Lelegard, who and encountered and debugged a large portion
of the issue in the same twelve hours that I did (as well as testing my patch).

Cc: Thierry Lelegard <thierry.lelegard@tv-numeric.com>
Signed-off-by: Devin Heitmueller <dheitmueller@kernellabs.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>