Re: [PATCH] Driver Core: document struct class_device properly

[Greg KH <greg@kroah.com>]

On Fri, Oct 28, 2005 at 01:54:59AM -0500, Dmitry Torokhov wrote:
> On Friday 28 October 2005 01:30, Greg KH wrote:
> > [PATCH] Driver Core: document struct class_device properly
> ...
> 
> > + * @release: pointer to a release function for this struct class_device.  If
> > + * set, this will be called instead of the class specific release function.
> > + * Only use this if you want to override the default release function, like
> > + * when you are nesting class_device structures.
> > + * @hotplug: pointer to a hotplug function for this struct class_device.  If
> > + * set, this will be called instead of the class specific hotplug function.
> > + * Only use this if you want to override the default hotplug function, like
> > + * when you are nesting class_device structures.
> 
> Greg, 
> 
> Is this solution for nesting class devices considered permanent or is it
> a stop-gap measure?

As I detalied a while ago, a stop-gap for now.

> I hope it is latter as these 2 new methods allow one
> class device walk all over class's intended interface and semantics and
> you can no longer rely that objects of the same class have similar
> characteristics/attributes and similar behavior. You already had to
> abandon using class's default attributes when dealing with nested devices,
> I think it is wrong long-term solution.
> 
> What about Kay's proposal about moving (as far as userspace concerned)
> everything into /sys/devices?

That's exactly what I am now working on.  But it will take much longer
than 2.6.15 to get there for that.  More like the next 6 months or so at
the least...

For a good description of the latest summary, see:
	http://www.kroah.com/log/linux/driver_model_changes.html
thanks,

greg k-h