On Wed, 2021-12-15 at 06:41 +0000, Claudiu.Beznea@xxxxxxxxxxxxx wrote: > On 15.12.2021 05:05, David Mosberger-Tang wrote: > > > +static int wilc_parse_gpios(struct wilc *wilc) > > +{ > > + struct spi_device *spi = to_spi_device(wilc->dev); > > + struct wilc_spi *spi_priv = wilc->bus_data; > > + struct wilc_gpios *gpios = &spi_priv->gpios; > > + > > + /* get ENABLE pin and deassert it (if it is defined): */ > > + gpios->enable = devm_gpiod_get_optional(&spi->dev, > > + "enable", GPIOD_OUT_LOW); > > + /* get RESET pin and assert it (if it is defined): */ > > + if (gpios->enable) { > > + /* if enable pin exists, reset must exist as well */ > > + gpios->reset = devm_gpiod_get(&spi->dev, > > + "reset", GPIOD_OUT_HIGH); > > As far as I can tell form gpiolib code the difference b/w GPIOD_OUT_HIGH > and GPIOD_OUT_LOW in gpiolib is related to the initial value for the GPIO. Yes. > Did you used GPIOD_OUT_HIGH for reset to have the chip out of reset at this > point? No, ~RESET is an active-low signal. GPIOD_OUT_LOW should really be called GPIOD_OUT_DEASSERTED or something like that. The code ensures that the chip is in RESET and ~ENABLEd after parsing the GPIOs. > > + if (IS_ERR(gpios->reset)) { > > + dev_err(&spi->dev, "missing reset gpio.\n"); > > + return PTR_ERR(gpios->reset); > > + } > > + } else { > > + gpios->reset = devm_gpiod_get_optional(&spi->dev, > > + "reset", GPIOD_OUT_HIGH); > > + } > > + return 0; > > +} > > + > > +static void wilc_wlan_power(struct wilc *wilc, bool on) > > +{ > > + struct wilc_spi *spi_priv = wilc->bus_data; > > + struct wilc_gpios *gpios = &spi_priv->gpios; > > + > > + if (on) { > > + gpiod_set_value(gpios->enable, 1); /* assert ENABLE */ > > + mdelay(5); > > + gpiod_set_value(gpios->reset, 0); /* deassert RESET */ > > From what I can tell from gpiolib code, requesting the pin from device tree > with: > > + reset-gpios = <&pioA 6 GPIO_ACTIVE_LOW>; > > makes the value written with gpiod_set_value() to be negated, thus the 0 > written here is translated to a 1 on the pin. Is there a reason you did it > like this? Yes, of course. RESET is an active-low signal, as defined in the datasheet. > Would it have been simpler to have both pins requested with > GPIO_ACTIVE_HIGH and here to do gpiod_set_value(gpio, 1) for both of the > pin. In this way, at the first read of the code one one would have been > telling that it does what datasheet specifies: for power on toggle enable > and reset gpios from 0 to 1 with a delay in between. I think you're confusing 0 and 1 with low-voltage and high-voltage. 0 means de-assert the signal, 1 means assert the signal. Whether that translates to a low voltage or a high voltage depends on whether the signal a active-low or active-high. > > > > + } else { > > + gpiod_set_value(gpios->reset, 1); /* assert RESET */ > > + gpiod_set_value(gpios->enable, 0); /* deassert ENABLE */ > > I don't usually see comments near the code line in kernel. Maybe move them > before the actual code line or remove them at all as the code is impler enough? You're kidding, right? --david