Re: [PATCH[[2.5][3-11] update dvb subsystem core

[Christoph Hellwig <hch@infradead.org>]

On Wed, May 07, 2003 at 05:56:33PM +0200, Michael Hunold wrote:
> I won't insist on keeping code that I haven't written. My only point is 
> that we use the code in set-top-boxes, where every byte is valuable. But 
>   I suspect that there are numerous other places where we could safe 
> bytes... 8-)

This code will go away soon for both the devfs and non-devfs case..

> > Okay, you're right I should have read more of the code to get the global
> > picture.  You still wan't an owner field for at least struct dvb_device
> > device, though - but the try_module_get must go into dvb_generic_open
> > and maybe in more other places where you use the "backend" modules.
> 
> I don't get that, sorry. The backend modules have functional 
> dependencies and register/deregister upon loading/unloading. There is 
> never a call from the dvb-core to the backend modules. Do I really need 
> an owner field then?

It doesn't have to be a call.  Unless I completely misread the code
your dvb core references struct dvb_adapter in certain cases.  But
struct dvb_adapter is allocated in the actual drivers so these could be
unloaded and give you scrambled memory even when it's still in use.

So you need to acquire a reference on those backends whenever you
touch any object that logically belongs to them.