Re: [RFC PATCH v5 1/7] drivers: thermal: tsens: Add VER_0 tsens version

From: Amit Kucheria
Date: Thu Aug 13 2020 - 07:19:22 EST


On Tue, Aug 11, 2020 at 6:48 PM <ansuelsmth@xxxxxxxxx> wrote:
>
>
>
> > -----Messaggio originale-----
> > Da: Amit Kucheria <amit.kucheria@xxxxxxxxxx>

> >
> > > if (IS_ERR(priv->rf[i]))
> > > return PTR_ERR(priv->rf[i]);
> > > }
> > > @@ -775,12 +800,80 @@ int __init init_common(struct tsens_priv
> > *priv)
> > > goto err_put_device;
> > > }
> > >
> > > - priv->rf[TSENS_EN] = devm_regmap_field_alloc(dev, priv->srot_map,
> > > - priv->fields[TSENS_EN]);
> > > - if (IS_ERR(priv->rf[TSENS_EN])) {
> > > - ret = PTR_ERR(priv->rf[TSENS_EN]);
> > > - goto err_put_device;
> > > + if (tsens_version(priv) >= VER_0_1) {
> > > + priv->rf[TSENS_EN] = devm_regmap_field_alloc(
> > > + dev, priv->srot_map, priv->fields[TSENS_EN]);
> > > + if (IS_ERR(priv->rf[TSENS_EN])) {
> > > + ret = PTR_ERR(priv->rf[TSENS_EN]);
> > > + goto err_put_device;
> > > + }
> > > +
> > > + priv->rf[SENSOR_EN] = devm_regmap_field_alloc(
> > > + dev, priv->srot_map, priv->fields[SENSOR_EN]);
> > > + if (IS_ERR(priv->rf[SENSOR_EN])) {
> > > + ret = PTR_ERR(priv->rf[SENSOR_EN]);
> > > + goto err_put_device;
> > > + }
> > > + priv->rf[INT_EN] = devm_regmap_field_alloc(
> > > + dev, priv->tm_map, priv->fields[INT_EN]);
> > > + if (IS_ERR(priv->rf[INT_EN])) {
> > > + ret = PTR_ERR(priv->rf[INT_EN]);
> > > + goto err_put_device;
> > > + }
> > > + } else {
> >
> > Let's not create two big sections with if-else for 8960 and everything
> > else. For example, what is wrong with using common code for TSENS_EN?
> >
> > If the concern is memory wasted trying to allocate fields not present
> > on this older platform, perhaps consider adding a check in the loop to
> > break early in case of 8960?
> >
>
> About TSENS_EN the old platform doesn't have SROT so I need to use TM_MAP.
> Should I set the srot map to match the tm map so we can use the common function?
> Aside from this problem, I will try to remove the big if-else.

Ick. I guess srot_map and tm_map pointing to the same region is the
lesser of two evils? It makes it so this will be constrained to a
single place in init_common().