Hi Stefan,
On Thu, 27 Sep 2018 at 17:44, Stefan Schmidt <stefan@xxxxxxxxxxxxxxxxxx> wrote:
>
> Hello Xue.
>
> On 31/08/2018 23:46, Xue Liu wrote:
> > The struct mcr20a_platform_data is uesed only in probe function
> > and it holds only one member. So it is not necessary to reserve it.
> >
> > Using gpiod family API to handle reset pin.
> >
> > Signed-off-by: Xue Liu <liuxuenetmail@xxxxxxxxx>
> > ---
> > drivers/net/ieee802154/mcr20a.c | 64 +++++++--------------------------
> > 1 file changed, 13 insertions(+), 51 deletions(-)
> >
> > diff --git a/drivers/net/ieee802154/mcr20a.c b/drivers/net/ieee802154/mcr20a.c
> > index 04891429a554..44de81e5f140 100644
> > --- a/drivers/net/ieee802154/mcr20a.c
> > +++ b/drivers/net/ieee802154/mcr20a.c
> > @@ -132,11 +132,6 @@ static const struct reg_sequence mar20a_iar_overwrites[] = {
> > };
> >
> > #define MCR20A_VALID_CHANNELS (0x07FFF800)
> > -
> > -struct mcr20a_platform_data {
> > - int rst_gpio;
> > -};
> > -
> > #define MCR20A_MAX_BUF (127)
> >
> > #define printdev(X) (&X->spi->dev)
> > @@ -412,7 +407,6 @@ struct mcr20a_local {
> > struct spi_device *spi;
> >
> > struct ieee802154_hw *hw;
> > - struct mcr20a_platform_data *pdata;
> > struct regmap *regmap_dar;
> > struct regmap *regmap_iar;
> >
> > @@ -976,20 +970,6 @@ static irqreturn_t mcr20a_irq_isr(int irq, void *data)
> > return IRQ_HANDLED;
> > }
> >
> > -static int mcr20a_get_platform_data(struct spi_device *spi,
> > - struct mcr20a_platform_data *pdata)
> > -{
> > - int ret = 0;
> > -
> > - if (!spi->dev.of_node)
> > - return -EINVAL;
> > -
> > - pdata->rst_gpio = of_get_named_gpio(spi->dev.of_node, "rst_b-gpio", 0);
> > - dev_dbg(&spi->dev, "rst_b-gpio: %d\n", pdata->rst_gpio);
> > -
> > - return ret;
> > -}
> > -
> > static void mcr20a_hw_setup(struct mcr20a_local *lp)
> > {
> > u8 i;
> > @@ -1249,7 +1229,7 @@ mcr20a_probe(struct spi_device *spi)
> > {
> > struct ieee802154_hw *hw;
> > struct mcr20a_local *lp;
> > - struct mcr20a_platform_data *pdata;
> > + struct gpio_desc *rst_b;
> > int irq_type;
> > int ret = -ENOMEM;
> >
> > @@ -1260,48 +1240,32 @@ mcr20a_probe(struct spi_device *spi)
> > return -EINVAL;
> > }
> >
> > - pdata = kmalloc(sizeof(*pdata), GFP_KERNEL);
> > - if (!pdata)
> > - return -ENOMEM;
> > -
> > - /* set mcr20a platform data */
> > - ret = mcr20a_get_platform_data(spi, pdata);
> > - if (ret < 0) {
> > - dev_crit(&spi->dev, "mcr20a_get_platform_data failed.\n");
> > - goto free_pdata;
> > - }
> > -
> > - /* init reset gpio */
> > - if (gpio_is_valid(pdata->rst_gpio)) {
> > - ret = devm_gpio_request_one(&spi->dev, pdata->rst_gpio,
> > - GPIOF_OUT_INIT_HIGH, "reset");
> > - if (ret)
> > - goto free_pdata;
> > + rst_b = devm_gpiod_get(&spi->dev, "rst_b", GPIOD_OUT_HIGH);
>
> I am a bit confused about the pin name here. When using
> of_get_named_gpio() the name is "rst_b-gpio" which is the same I see
> referenced in the devicetree bindings file.
>
> When switching to devm_gpiod_get() the name is shortened to "rst_b".
> Does the gpiod API implicitly add a -gpio to the name when searching for
> it in the dst?
>
Yes. The function calling tree is shown below:
--> devm_gpiod_get()
--> __gpiod_get_index()
--> of_find_gpio()
The definition of function of_find_gpio() is here
https://elixir.bootlin.com/linux/latest/source/drivers/gpio/gpiolib-of.c#L231.
The definition of struct gpio_suffixes is here
https://elixir.bootlin.com/linux/latest/source/drivers/gpio/gpiolib.h#L95.
You can see that the suffix "gpio" or "gpios" is automatically added
when it is looking for the name of gpio in the device tree.
> The reason I ask is that I would want to avoid a name change of the
> property. That would break the dts bindings already in place.
>
We don't need to change the dts bindings.
> > + if (IS_ERR(rst_b)) {
> > + ret = PTR_ERR(rst_b);
> > + if (ret != -EPROBE_DEFER)
> > + dev_err(&spi->dev, "Failed to get 'rst_b' gpio: %d", ret);
> > + return ret;
> > }
> >
> > /* reset mcr20a */
> > - if (gpio_is_valid(pdata->rst_gpio)) {
> > - usleep_range(10, 20);
> > - gpio_set_value_cansleep(pdata->rst_gpio, 0);
> > - usleep_range(10, 20);
> > - gpio_set_value_cansleep(pdata->rst_gpio, 1);
> > - usleep_range(120, 240);
> > - }
> > + usleep_range(10, 20);
> > + gpiod_set_value_cansleep(rst_b, 1);
> > + usleep_range(10, 20);
> > + gpiod_set_value_cansleep(rst_b, 0);
> > + usleep_range(120, 240);
>
> With your change you reversing the setting from ->0->1 to ->1->0. Is the
> gpiod API reverse here or did you made a copy and paste mistake? :-)
>
I am afraid both of assumptions are wrong here. The new GPIO
descriptor consumer interface uses *logical* value.
It means 0 and 1 denote GPIO deassertion and assertion. The property
of rst_b is GPIO_ACTIVE_LOW. So the value 1
means low in physical line and the value 0 means high.
Please reference
https://www.kernel.org/doc/Documentation/gpio/consumer.txt for more
information.
> >
> > /* allocate ieee802154_hw and private data */
> > hw = ieee802154_alloc_hw(sizeof(*lp), &mcr20a_hw_ops);
> > if (!hw) {
> > dev_crit(&spi->dev, "ieee802154_alloc_hw failed\n");
> > - ret = -ENOMEM;
> > - goto free_pdata;
> > + return ret;
> > }
> >
> > /* init mcr20a local data */
> > lp = hw->priv;
> > lp->hw = hw;
> > lp->spi = spi;
> > - lp->spi->dev.platform_data = pdata;
> > - lp->pdata = pdata;
> >
> > /* init ieee802154_hw */
> > hw->parent = &spi->dev;
> > @@ -1370,8 +1334,6 @@ mcr20a_probe(struct spi_device *spi)
> >
> > free_dev:
> > ieee802154_free_hw(lp->hw);
> > -free_pdata:
> > - kfree(pdata);
> >
> > return ret;
> > }
> >
>
> The rest looks good to me and I like that we can get rid of all the
> boiler code associated with the pdata.
>
> If you could clarify (and potentially fix) the two points I raised I
> would be happy to apply this.
>
> regards
> Stefan Schmidt
regards
Xue Liu
--
