On Mon, Nov 03, 2014 at 04:48:34PM +0100, Javier Martinez Canillas wrote:
> On 11/03/2014 04:33 PM, Mark Brown wrote:
> > On Mon, Nov 03, 2014 at 03:40:38PM +0100, Javier Martinez Canillas wrote:
> >> if (!of_node_cmp(np->name, info->desc.name)) {
> >> config->init_data =
> >> - of_get_regulator_init_data(&pdev->dev, np);
> >> + of_get_regulator_init_data(&pdev->dev, np,
> >> + NULL);
> > This looks buggy, we're not passing in a descriptor.
> The descriptor is only used when extracting the init_data to map the
> modes and since it was not a parameter before, some drivers needs to
> be refactored to pass that information.
No, it's only *currently* used for that. If we don't bother passing the
descriptor in then future additions which make use of it (including
adding mode operations to existing drivers) won't work and it might not
be obvious why.
> I thought that instead of adding intrusive changes in drivers that I
> don't have hw to test, that parameter could be optional so the patch
> that use that information check if the descriptor and the map_modes
> function pointer are not NULL.
You're already going through and modifying every single driver and all
of those I looked at already had references to the descriptor in
adjacent code or a global descriptor for the one regulator supported by
the driver.
Attachment:
signature.asc
Description: Digital signature
