On Tue, Sep 17, 2019 at 08:12:46PM +0300, Andy Shevchenko wrote:
> On Tue, Sep 17, 2019 at 09:52:45AM -0700, Dmitry Torokhov wrote:
> > On Tue, Sep 17, 2019 at 07:35:36PM +0300, Andy Shevchenko wrote:
> > > On Tue, Sep 17, 2019 at 05:58:08PM +0200, Marco Felsch wrote:
> > > > Currently the driver do not care about the regulator which supplies the
> > > > controller. This can lead into problems since we support (deep-)sleep
> > > > because the regulator can be turned off before we send the (deep-)sleep
> > > > command to the controller. Using a own regulator improves the power
> > > > consumption during sleep even more.
> > >
> > > > + tsdata->vdd = devm_regulator_get(&client->dev, "vdd");
> > > > + if (IS_ERR(tsdata->vdd)) {
> > > > + error = PTR_ERR(tsdata->vdd);
> > > > + if (error == -EPROBE_DEFER)
> > > > + return error;
> > >
> > > Before it worked w/o regulator. You have to make it optional.
> >
> > No, regulators are funky this way. If there isn't one declared in the
> > device tree then a dummy is created automatically. Optional regulators
> > are only to be used when parts of an IC can be powered up separately.
>
> It depends if platform has full constrains or not.
Full constraints is the default behavior. Do we even have platforms that
are not? ACPI and DT are fully-constrained, so that leaves legacy
boards... But there isn't a single one in the tree that uses this
driver.
Thanks.
--
Dmitry
