Re: [PATCH V2] i2c: mux: pca954x: Add ACPI support for pca954x

[Peter Rosin]

On 2016-11-23 09:52, tnhuynh@xxxxxxx wrote:
> From: Tin Huynh <tnhuynh@xxxxxxx>
> 
> This patch enables ACPI support for mux-pca954x driver.
> 
> Signed-off-by: Tin Huynh <tnhuynh@xxxxxxx>
> 
> Change from v1 :
>  -Don't shadow id variable.
>  -Include sorted header.
>  -Redefine acpi_device_id.
>  -Add CONFIG_ACPI.

Please put these extra comment that do not belong in the changelog
below the following three dashes. Then git will ignore these comments
automatically, which simplifies applying the patch.

> ---
>  drivers/i2c/muxes/i2c-mux-pca954x.c |   29 ++++++++++++++++++++++++++++-
>  1 files changed, 28 insertions(+), 1 deletions(-)

The customary position for extra comments is here, below the diffstat.

> diff --git a/drivers/i2c/muxes/i2c-mux-pca954x.c b/drivers/i2c/muxes/i2c-mux-pca954x.c
> index 1091346..1ad67cb 100644
> --- a/drivers/i2c/muxes/i2c-mux-pca954x.c
> +++ b/drivers/i2c/muxes/i2c-mux-pca954x.c
> @@ -35,6 +35,7 @@
>   * warranty of any kind, whether express or implied.
>   */
>  
> +#include <linux/acpi.h>
>  #include <linux/device.h>
>  #include <linux/gpio/consumer.h>
>  #include <linux/i2c.h>
> @@ -120,6 +121,21 @@ struct pca954x {
>  };
>  MODULE_DEVICE_TABLE(i2c, pca954x_id);
>  
> +#ifdef CONFIG_ACPI
> +static const struct acpi_device_id pca954x_acpi_ids[] = {
> +	{ .id = "PCA9540", .driver_data = pca_9540 },
> +	{ .id = "PCA9542", .driver_data = pca_9540 },
> +	{ .id = "PCA9543", .driver_data = pca_9543 },
> +	{ .id = "PCA9544", .driver_data = pca_9544 },
> +	{ .id = "PCA9545", .driver_data = pca_9545 },
> +	{ .id = "PCA9546", .driver_data = pca_9545 },
> +	{ .id = "PCA9547", .driver_data = pca_9547 },
> +	{ .id = "PCA9548", .driver_data = pca_9548 },
> +	{ }
> +};
> +MODULE_DEVICE_TABLE(acpi, pca954x_acpi_ids);
> +#endif
> +
>  #ifdef CONFIG_OF
>  static const struct of_device_id pca954x_of_match[] = {
>  	{ .compatible = "nxp,pca9540", .data = &chips[pca_9540] },
> @@ -245,8 +261,18 @@ static int pca954x_probe(struct i2c_client *client,
>  	match = of_match_device(of_match_ptr(pca954x_of_match), &client->dev);
>  	if (match)
>  		data->chip = of_device_get_match_data(&client->dev);
> -	else
> +	else if (id)
>  		data->chip = &chips[id->driver_data];
> +#ifdef CONFIG_ACPI

No, this #ifdef is not desired. If ACPI is not configured, acpi_match_device
will return NULL unconditionally and the compiler should compile the whole
else branch to be an unconditional "return -ENODEV;" and skip the dead code
at the end of the block. This is better than leaving data->chip as NULL and
oops later. It is also less cluttery to not have this extra #ifdef.

Cheers,
Peter

> +	else {
> +		const struct acpi_device_id *acpi_id;
> +
> +		acpi_id = acpi_match_device(pca954x_acpi_ids, &client->dev);
> +		if (!acpi_id)
> +			return -ENODEV;
> +		data->chip = &chips[acpi_id->driver_data];
> +	}
> +#endif
>  
>  	data->last_chan = 0;		   /* force the first selection */
>  
> @@ -321,6 +347,7 @@ static int pca954x_resume(struct device *dev)
>  		.name	= "pca954x",
>  		.pm	= &pca954x_pm,
>  		.of_match_table = of_match_ptr(pca954x_of_match),
> +		.acpi_match_table = ACPI_PTR(pca954x_acpi_ids),
>  	},
>  	.probe		= pca954x_probe,
>  	.remove		= pca954x_remove,
>