On Thu, Dec 01, 2016 at 10:20:47AM +0100, Christian Gromm wrote:
> On Thu, 1 Dec 2016 09:56:11 +0100
> Greg KH <gre...@linuxfoundation.org> wrote:
> > On Thu, Dec 01, 2016 at 09:50:12AM +0100, Christian Gromm wrote:
> > > On Thu, 1 Dec 2016 08:00:57 +0100
> > > Greg KH <gre...@linuxfoundation.org> wrote:
> > >
> > > > On Wed, Nov 30, 2016 at 10:48:32PM -0700, Jason Litzinger wrote:
> > > > > > This is fine, but the fact that the most subsystem feels like it
> > > > > > has to
> > > > > > have its own attribute type is the big problem with this file, that
> > > > > > should not be needed at all, and hopefully will be fixed up someday
> > > > > > soon
> > > > > > (i.e. it's a requirement before it can get out of staging...)
> > >
> > > That's interesting. We still need to set up a strategy to get this driver
> > > out of staging one day, as we decided on the ELCE 2015 in Dublin. A good
> > > starting point would be a list of requirements that need to be met to
> > > achieve this goal.
> > >
> > > Does this make sense?
> > Yes, and starting with cleaning up the kobject mess would be a good
> > first item for that list :)
> And what "kobject mess" exactly is it that you're talking about?
> The kobjects to hook things up in sysfs?
Yes, you should never be using "raw" kobjects in a driver subsystem like
this, you should tie into the "real" driver model by being a bus and
devel mailing list