Re: [PATCH RFD] alternative kobject release wait mechanism

["Dmitry Torokhov" <dmitry.torokhov@gmail.com>]

On 4/19/07, Cornelia Huck <cornelia.huck@de.ibm.com> wrote:
> On Thu, 19 Apr 2007 09:13:43 -0400,
> "Dmitry Torokhov" <dmitry.torokhov@gmail.com> wrote:
>
> > Because they are managed by 2 different entities. the struct device
> > objects are managed by device core and driver-specific objects are
> > managed by their respective driver.
>
> Not sure if I understand you here. My view of this was always that the
> embedding object was kind of an extended device and that the relevant
> driver/subsystem managed it through the driver core infrastructure.
>

I am not sure if I agree with this point of view. Driver (or
subsystem) provides an instance of struct device for the rest of the
system to iteract uniformly with (suspend/resume/tree
visualization/etc) i.e. struct device implement an interface for
subsystems. However most of the system use their own mechanisms to
manage their devices. They can rely on the driver core to a certain
degree but driver core is mostly a carries out helper functions, not
the meat.

> >
> > > > Pretty much drivers have 2 options:
> > > >
> > > > struct my_device {
> > > >         void *private_data;
> > > >         struct device dev;
> > > > };
> > > >
> > > > In this case ->release must live in a subsystem code; individual
> > > > drivers kfree(my_dev->private) and do any additional cleanup after
> > > > calling device_unregister(&my_dev->dev);
> > >
> > > They must do this in the ->remove callback.
> >
> > Why? If the driver truly stops hardware then any driver-specific data
> > is not needed. With sysfs severing access to removed attributes there
> > is no need to gave "global release", cleanup can be done in stages.
>
> I think I meant the same thing :) Freeing the data in the ->release
> callback is obviously too late. Freeing it in the ->remove callback
> means that the device is no longer really used (and can't be looked up
> any more); only some further refrences may linger (and those are of no
> consequence with the sysfs disconnect).
>

Ah, right, I confused ->remove() with ->release() in your post. Sorry
about that.

-- 
Dmitry