Re: [PATCH v8 2/3] mfd: add Gateworks System Controller core driver

[Tim Harvey]

On Tue, Apr 28, 2020 at 11:33 PM Lee Jones <lee.jones@xxxxxxxxxx> wrote:
>
> On Tue, 28 Apr 2020, Tim Harvey wrote:
>
> > On Tue, Apr 28, 2020 at 2:44 AM Lee Jones <lee.jones@xxxxxxxxxx> wrote:
> > >
> > <snip>
> > > > +
> > > > +static int gsc_probe(struct i2c_client *client)
> > > > +{
> > > > +     struct device *dev = &client->dev;
> > > > +     struct gsc_dev *gsc;
> > > > +     int ret;
> > > > +     unsigned int reg;
> > > > +
> > > > +     gsc = devm_kzalloc(dev, sizeof(*gsc), GFP_KERNEL);
> > > > +     if (!gsc)
> > > > +             return -ENOMEM;
> > > > +
> > > > +     gsc->dev = &client->dev;
> > > > +     gsc->i2c = client;
> > > > +     i2c_set_clientdata(client, gsc);
> > > > +
> > > > +     gsc->bus.reg_write = gsc_regmap_regwrite;
> > > > +     gsc->bus.reg_read = gsc_regmap_regread;
> > >
> > > Why do you need to store these in ddata?
> >
> > Lee,
> >
> > Thanks for the review!
> >
> > I need the remap_bus* for devm_regmap_init() in the hwmon sub-module driver:
> >
> > hwmon->regmap = devm_regmap_init(dev, &gsc->bus, gsc->i2c_hwmon,
> > &gsc_hwmon_regmap_config);
> >
> > Is there something easier I'm missing?
>
> This is an odd setup.  I haven't seen one driver registering another
> driver's Regmap call-backs before, related or otherwise.  Normally the
> Regmap is setup (initialised) in the parent driver and child drivers
> just make use of it.  Here it looks like you are registering 2
> separate Regmaps, but using the same call-backs for both, which seems
> wrong to me.
>

Lee,

It is perhaps an odd setup. The hwmon sub-device is at a different i2c
slave address than the other sub-devices. The same callbacks are used
for reg read/write to take advantage of the retries due to the errata
resulting in occasional NAK'd register reads.

Tim