On Wed, 17 Jan 2024 14:51:13 +0200
Dumitru Ceclan <mitrutzceclan@xxxxxxxxx> wrote:
> Change the match table to use pointers instead of device ids.
> Alignment of the hmc425a_state was changed because of the const
> specifier for hmc425a_chip_info.
>
> Signed-off-by: Dumitru Ceclan <mitrutzceclan@xxxxxxxxx>
Hi Dumitru
The remaining use of type in here and deriving from structure offsets is
not nice. Add a trivial callback for the stuff in write_raw() that needs
to be different and is currently in a switch statement.
Then get rid of the type enum completely if possible.
> ---
> drivers/iio/amplifiers/hmc425a.c | 39 ++++++++++++++++++--------------
> 1 file changed, 22 insertions(+), 17 deletions(-)
>
> diff --git a/drivers/iio/amplifiers/hmc425a.c b/drivers/iio/amplifiers/hmc425a.c
> index e1162a500daf..b116b54e4206 100644
> --- a/drivers/iio/amplifiers/hmc425a.c
> +++ b/drivers/iio/amplifiers/hmc425a.c
> @@ -37,11 +37,11 @@ struct hmc425a_chip_info {
> };
>
> struct hmc425a_state {
> - struct mutex lock; /* protect sensor state */
> - struct hmc425a_chip_info *chip_info;
> - struct gpio_descs *gpios;
> - enum hmc425a_type type;
> - u32 gain;
> + struct mutex lock; /* protect sensor state */
> + const struct hmc425a_chip_info *chip_info;
> + struct gpio_descs *gpios;
> + enum hmc425a_type type;
> + u32 gain;
This illustrates why I'm not keen on manual alignment like this. Generates
churn that makes it hard to spot the actual changes. To avoid this happening
again I'd suggest a single space is fine for all lines and don't align them
at all!
> };
>
> static int hmc425a_write(struct iio_dev *indio_dev, u32 value)
> @@ -58,7 +58,7 @@ static int hmc425a_write(struct iio_dev *indio_dev, u32 value)
>
> static int hmc425a_gain_dB_to_code(struct hmc425a_state *st, int val, int val2, int *code)
> {
> - struct hmc425a_chip_info *inf = st->chip_info;
> + const struct hmc425a_chip_info *inf = st->chip_info;
> int gain, temp;
>
> if (val < 0)
> @@ -187,15 +187,6 @@ static const struct iio_chan_spec hmc425a_channels[] = {
> HMC425A_CHAN(0),
> };
>
> -/* Match table for of_platform binding */
> -static const struct of_device_id hmc425a_of_match[] = {
> - { .compatible = "adi,hmc425a", .data = (void *)ID_HMC425A },
> - { .compatible = "adi,hmc540s", .data = (void *)ID_HMC540S },
> - { .compatible = "adi,adrf5740", .data = (void *)ID_ADRF5740 },
> - {},
> -};
> -MODULE_DEVICE_TABLE(of, hmc425a_of_match);
> -
> static struct hmc425a_chip_info hmc425a_chip_info_tbl[] = {
> [ID_HMC425A] = {
> .name = "hmc425a",
> @@ -226,6 +217,18 @@ static struct hmc425a_chip_info hmc425a_chip_info_tbl[] = {
> },
> };
>
> +/* Match table for of_platform binding */
> +static const struct of_device_id hmc425a_of_match[] = {
> + { .compatible = "adi,hmc425a",
> + .data = &hmc425a_chip_info_tbl[ID_HMC425A]},
> + { .compatible = "adi,hmc540s",
> + .data = &hmc425a_chip_info_tbl[ID_HMC540S]},
> + { .compatible = "adi,adrf5740",
> + .data = &hmc425a_chip_info_tbl[ID_ADRF5740]},
> + {},
Nice to drop that trailing comma whilst here. No need for one on a 'terminator'
of such an array as by definition nothing should ever be added after it.
> +};
> +MODULE_DEVICE_TABLE(of, hmc425a_of_match);
> +
> static int hmc425a_probe(struct platform_device *pdev)
> {
> struct iio_dev *indio_dev;
> @@ -237,14 +240,16 @@ static int hmc425a_probe(struct platform_device *pdev)
> return -ENOMEM;
>
> st = iio_priv(indio_dev);
> - st->type = (uintptr_t)device_get_match_data(&pdev->dev);
>
> - st->chip_info = &hmc425a_chip_info_tbl[st->type];
> + st->chip_info = device_get_match_data(&pdev->dev);
> indio_dev->num_channels = st->chip_info->num_channels;
> indio_dev->channels = st->chip_info->channels;
> indio_dev->name = st->chip_info->name;
> st->gain = st->chip_info->default_gain;
>
> + /* Compute index of the acquired chip info in the array */
> + st->type = st->chip_info - hmc425a_chip_info_tbl;
Definitely not a good idea. If you need a type field, put it in the chip_info_tbl
but you should not need one anyway because type is rarely what matters
but rather data or behavior (via a callback) needed for a given device.
Here it looks like a callback is needed for the few lines in write_raw()
that are fiddly to express as data.
> +
> st->gpios = devm_gpiod_get_array(&pdev->dev, "ctrl", GPIOD_OUT_LOW);
> if (IS_ERR(st->gpios))
> return dev_err_probe(&pdev->dev, PTR_ERR(st->gpios),
