(please don't hate me for this... :-)

There are two obvious problems with the previous patch...

1. From the beginning  I tried to verbosely comment the scripts (so they 
can be self-explanatory and easy to read) and, apparently, I forgot to 
update the comments in the "uhci.agent" after reorganizing the scripts 
(to experiment with), so... the comments in /etc/hotplug/pci/uhci.agent 
are wrong, misleading and confusing (ouch)!

2. Only after posting that previous patch I realised that the way in 
which the "start" and "stop" events are passed down from policy agents 
to driver agents was, well, not so good. "start" was passed if a kernel 
module was already loaded and "stop" was passed if a module was already 
unloaded. Well, it turns out that the right thing is to handle 
start/stop in the same (almost) way that add/remove events are handled. 
The only difference is that, for start/stop, $MODULE is also exported 
and is the name of the currently loaded kernel module ("-" if none, "#" 
if undetermined).

Now, the final thing that remains for unloading to be complete is 
reference counting. I tried to do it with bash (by storing counts in 
*.ref files in the driver directories) and guess what I stumbled to... a 
race condition! Multiple hotplug events can be processed by the scripts 
at the same time. So I took it back. Isn't there a way to use modutils 
programs to play with the module's real refrence counts? It would make 
things a whole lot easier (and safer).