On Tue, Apr 19, 2016 at 09:43:08AM +0200, H. Nikolaus Schaller wrote:
>
> > Am 18.04.2016 um 23:22 schrieb Dmitry Torokhov <dmitry.torok...@gmail.com>:
> >
> > On Mon, Apr 18, 2016 at 09:55:37PM +0200, H. Nikolaus Schaller wrote:
> >> commit e7ec014a47e4 ("Input: twl6040-vibra - update for device tree
> >> support")
> >>
> >> made the separate vibra DT node to a subnode of the twl6040.
> >>
> >> It now calls of_find_node_by_name() to locate the "vibra" subnode.
> >> This function has a side effect to call of_node_put on() for the twl6040
> >> parent node passed in as a parameter. This causes trouble later on.
> >>
> >> Solution: we must call of_node_get() before of_find_node_by_name()
> >
> > God, what messed up API.
>
> Yes, indeed. It is opposite to the usual object ownership rule that the code
> fragment that asks for a handle has to release it.
>
> Usually it does not become obvious because often CONFIG_OF_DYNAMIC=n.
> This disables all of_node refcounting completely so such bugs remain
> unnoticed.
>
> > Any chance we can make it a bit more sane and
> > not drop the reference inside it instead?
>
> Well, if you want to change ~2000 files, test on all platforms and ask Linus
> for agreement?
It's not that bad, let's see what DT maintainers say to the patch I
posted...
--
Dmitry
