[PATCH v2 3/5] regulator: helper routine to extract regulator_init_data

Grant Likely grant.likely at secretlab.ca
Tue Oct 25 09:21:13 EST 2011


On Mon, Oct 24, 2011 at 11:51:33PM +0800, Shawn Guo wrote:
> On Mon, Oct 24, 2011 at 04:56:31PM +0200, Grant Likely wrote:
> > On Mon, Oct 24, 2011 at 10:51:40PM +0800, Shawn Guo wrote:
> > > On Mon, Oct 24, 2011 at 03:59:50PM +0200, Grant Likely wrote:
> > > > On Mon, Oct 24, 2011 at 09:40:26PM +0800, Shawn Guo wrote:
> > > > > diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
> > > > > index 8fe132d..29dcf90 100644
> > > > > --- a/drivers/regulator/core.c
> > > > > +++ b/drivers/regulator/core.c
> > > > > @@ -2673,7 +2673,8 @@ struct regulator_dev *regulator_register(struct regulator_desc *regulator_desc,
> > > > >         BLOCKING_INIT_NOTIFIER_HEAD(&rdev->notifier);
> > > > > 
> > > > >         /* find device_node and attach it */
> > > > > -       rdev->dev.of_node = of_find_node_by_name(NULL, regulator_desc->name);
> > > > > +       rdev->dev.of_node = of_find_node_by_name(dev->parent->of_node,
> > > > > +                                                regulator_desc->name);
> > > > 
> > > > of_find_node_by_name() doesn't work that way.  The first argument is a
> > > > starting point, but it doesn't restrict the search to children of a
> > > > node.
> > > > 
> > > > for_each_child_of_node() is what you want to use when iterating over
> > > > the children which unfortunately changes the structure of this
> > > > function.
> > > > 
> > > The dev->parent->of_node is meant to point to node 'pmic: mc13892 at 0'.
> > > And the intention here is not to iterate over the children, but to
> > > start a search from a reasonable point rather than the top root node.
> > 
> > It is always better to attach the of_node at struct device
> > registration time instead of searching the tree in common code.  The
> > of_node should already be assigned by the time regulator_register() is
> > called.
> 
> That's the problem we have.  There is no 'struct dev' to attach of_node
> for each regulator by the time regulator_register() is called, because
> the 'struct dev' for each regulator is created inside
> regulator_register() as wrapped by 'struct regulator_dev'.

Then regulator_register must be split in half.  This is a common use
case, and other libraries have had to do the same.  One half to
allocate the structure, and then a second half for registration, and
regulator_register() becomes a simpler helper that calls both.  That
way a caller can insert extra initialization between the two steps.

g.



More information about the devicetree-discuss mailing list