Re: [PATCH v2 12/14] mfd: pm8008: rework driver

[Andy Shevchenko <andy.shevchenko@xxxxxxxxx>]

On Wed, May 29, 2024 at 7:30 PM Johan Hovold <johan+linaro@xxxxxxxxxx> wrote:
>
> Rework the pm8008 driver to match the new binding which no longer
> describes internal details like interrupts and register offsets
> (including which of the two consecutive I2C addresses the registers
> belong to).
>
> Instead make the interrupt controller implementation internal and pass
> interrupts to the subdrivers using MFD cell resources.
>
> Note that subdrivers may either get their resources, like register block
> offsets, from the parent MFD or this can be included in the subdrivers
> directly.
>
> In the current implementation, the temperature alarm driver is generic
> enough to just get its base address and alarm interrupt from the parent
> driver, which already uses this information to implement the interrupt
> controller.
>
> The regulator driver, however, needs additional information like parent
> supplies and regulator characteristics so in that case it is easier to
> just augment its table with the regulator register base addresses.
>
> Similarly, the current GPIO driver already holds the number of pins and
> that lookup table can therefore also be extended with register offsets.
>
> Note that subdrivers can now access the two regmaps by name, even if the
> primary regmap is registered last so that it is returned by default when
> no name is provided in lookups.
>
> Finally, note that the temperature alarm and GPIO subdrivers need some
> minor rework before they can be used with non-SPMI devices like the
> PM8008. The temperature alarm MFD cell name specifically uses a "qpnp"
> rather than "spmi" prefix to prevent binding until the driver has been
> updated.

...

> +       dummy = devm_i2c_new_dummy_device(dev, client->adapter, client->addr + 1);
> +       if (IS_ERR(dummy)) {
> +               ret = PTR_ERR(dummy);
> +               dev_err(dev, "failed to claim second address: %d\n", ret);
> +               return ret;
> +       }


> +       ret = devm_regmap_add_irq_chip_fwnode(dev, fwnode, regmap, client->irq,
>                                 IRQF_SHARED, 0, &pm8008_irq_chip, &irq_data);
> +       if (ret) {
> +               dev_err(dev, "failed to add IRQ chip: %d\n", ret);
> +               return ret;
>         }

I believe there is no harm to use

  return dev_err_probe(...);

for these. But it seems you don't like that API. Whatever, no-one will
die, just additional work for the future :-)

-- 
With Best Regards,
Andy Shevchenko