On Thu, May 11, 2006, Trent Piepho wrote:
> On Fri, 12 May 2006, Johannes Stezenbach wrote:
> > You know, instead of messing with *all* frontend drivers
> > and the internals of dvb_unregister_frontend(), I think
> > it would be much better to:
>
> I agree with you, it would be better, but I can't think of a way to do it.
>
> > - add dvb_attach() and dvb_detach() functions
> > - leave frontend drivers unmodified
> > - so nothing changes for out-of-tree STB like users like tuxbox
>
> How can dvb_detach() work, satisfy your three requirements here, AND still
> work with chained front-ends?
>
> Without chained frontends, it's easy enough (*):
>
> int dvb_detach(struct dvb_frontend* fe)
> {
> void *addr = (void*)fe->ops->release;
> dvb_unregister_frontend(fe);
> symbol_put_addr(addr);
> }
>
> But this doesn't work at all with chained frontends.
Looks like I didn't get something...
What is a "chained frontend"? How is this handled currently
without dvb_attach()? Why does it need to be handled
differently with dvb_attach()?
Did you/Andrew mix two different concepts (chaining vs. dvb_attach)
in the same patchset?
Please explain.
> (*) This assumes, currently correctly, that foo_attach() and foo_release() are
> in the same module
why? dvb_attach() incs the frontend module ref count, dvb_detach()
decs it -- no matter where they are called from
> and that symbol_put/get does module level ref counting and
> not symbol level.
What a crazy idea to do symbol level ref counting...
The kernel doesn't do that.
Johannes
_______________________________________________
linux-dvb@xxxxxxxxxxx
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
