Re: [RFC] Sample kset/ktype/kobject implementation

[Greg KH <greg@kroah.com>]

On Thu, Nov 29, 2007 at 10:39:07AM +0100, Cornelia Huck wrote:
> On Wed, 28 Nov 2007 22:11:39 -0800,
> Greg KH <greg@kroah.com> wrote:
> 
> > > > 	foo->kobj.kset = example_kset;
> > > > 	foo->kobj.ktype = &foo_ktype;
> > > > 
> > > > 	/*
> > > > 	 * Register the kobject with the kernel, all the default files will
> > > > 	 * be created here and the uevent will be sent out.  If we were to
> > > > 	 * call kobject_init() and then kobject_add() we would be
> > > > 	 * responsible for sending out the initial KOBJ_ADD uevent.
> > > > 	 */
> > > > 	retval = kobject_register(&foo->kobj);
> > > > 	if (retval) {
> > > > 		kfree(foo);
> > > 
> > > kobject_put(foo) is needed since it gets you through kobject_cleanup()
> > > where the name can be freed.
> > 
> > No, kobject_register() should have handled that for us, right?
> 
> kobject_register() doesn't do a kobject_put() if kobject_add() failed.

Crap.  If I can't get this code right in an example, the API is messed
up.  Time to take Kay seriously and start to revamp the basic kobject
api :)

thanks,

greg k-h