[i2c] [patch 2/5] i2c can remove()

[David Brownell]

On Friday 02 March 2007 12:32 pm, Jean Delvare wrote:

> > > Actually my current i2c stack no longer has the faulty i2c_adapter
> > > driver, so it should be possible to remove the duplicate list
> > > management. Could you provide a patch doing this? Would it let you
> > > make this one patch cleaner?
> > 
> > I'd much rather get this patch series merged first, then clean up those
> > superfluous lists later.  Removing those lists first implies respinning
> > most patches in this series (not just this one), extra testing time, and
> > further magnifies the delays involved.  Simpler to remove them later.
> 
> OK, fine with me. But don't wait too much after this series is reviewed
> and pushed to -mm before sending the one cleaning up the list,
> otherwise I fear it'll never happen.

Fair enough.  There are other driver model related cleanups that should
happen at some point too.  Refcounting is a bit wierd, driver register
should use the driver's module, and so forth.


> > Maybe it would be better if i2c_del_driver() were legacy-only, and
> > new-style drivers used a new i2c_unregister_driver() call.  (And if
> > we go that route, maybe similarly on the registration side.)  What
> > would you think of that approach?
> 
> Yes, after thinking about it some more, if there is that little
> common code, having separate functions makes sense. And this will
> pretty much guarantee that your patches don't affect the current
> drivers.

OK, done.  Added a bit of kerneldoc too.  :)

- Dave