Re: [PATCH V2 2/4] gpio: gpio-f81504: Add Fintek F81504/508/512 PCIE-to-UART/GPIO GPIOLIB support

[Linus Walleij]

On Thu, Jan 28, 2016 at 10:20 AM, Peter Hung <hpeter@xxxxxxxxx> wrote:

> This driver is GPIOLIB driver for F81504/508/512, it'll handle the
> GPIOLIB operation of this device. This module will depend on
> MFD_FINTEK_F81504_CORE.
>
> IC function list:
>     F81504: Max 2x8 GPIOs and max 4 serial ports
>         port2/3 are multi-function
>     F81508: Max 6x8 GPIOs and max 8 serial ports
>         port2/3 are multi-function, port8/9/10/11 are gpio only
>     F81512: Max 6x8 GPIOs and max 12 serial ports
>         port2/3/8/9/10/11 are multi-function
>
> GPIO register:
> PCI Configuration space:
>     F0h: bit0~5: Enable GPIO0~5
>          bit6~7: Reserve
>     F3h: bit0~5: Multi-Functional Flag (0:GPIO/1:UART)
>          bit0: UART2 pin out for UART2 / GPIO0
>          bit1: UART3 pin out for UART3 / GPIO1
>          bit2: UART8 pin out for UART8 / GPIO2
>          bit3: UART9 pin out for UART9 / GPIO3
>          bit4: UART10 pin out for UART10 / GPIO4
>          bit5: UART11 pin out for UART11 / GPIO5
>          bit6~7: Reserve
>     F1h: IO address (LSB)
>     F2h: IO address (MSB)
>     F8h + 8 * set: Direction control (bitwise)
>          bitx: 0 - Input mode
>          bitx: 1 - Output mode
>     F9h + 8 * set: Drive ability control (bitwise)
>          bitx: 0 - Open drain (default)
>          bitx: 1 - Push Pull
>          In this driver, we only implements open drain mode.
>
> IO space:
>     (IO base + 0~5): GPIO-0x~5x in/out value (bitwise)
>
> Suggested-by: One Thousand Gnomes <gnomes@xxxxxxxxxxxxxxxxxxx>
> Suggested-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> Signed-off-by: Peter Hung <hpeter+linux_kernel@xxxxxxxxx>

Overall a very nice and interesting patch set.

> +++ b/drivers/gpio/gpio-f81504.c
(...)
> +#include <linux/platform_device.h>
> +#include <linux/gpio.h>

Drivers should just
#include <linux/gpio/driver.h>

> +static struct f81504_gpio_chip *gpio_to_f81504_chip(struct gpio_chip *chip)
> +{
> +       return container_of(chip, struct f81504_gpio_chip, chip);
> +}

Avoid this construction in new code.

Use gpiochip_get_data(chip) everywhere that gpio_to_f81504_chip()
is used and register the gpiochip with gpiochip_add_data()
and the code will be simpler.

See any other driver in drivers/gpio for examples, I converted them
all.

> +static int f81504_gpio_get_direction(struct gpio_chip *chip, unsigned offset)
> +{
> +       u8 tmp;
> +       struct f81504_gpio_chip *gc = gpio_to_f81504_chip(chip);
> +       struct platform_device *pdev = to_platform_device(chip->dev);
> +       struct pci_dev *pci_dev = to_pci_dev(pdev->dev.parent);
> +
> +       mutex_lock(&gc->locker);
> +       pci_read_config_byte(pci_dev, F81504_GPIO_START_ADDR + gc->idx *
> +                       F81504_GPIO_SET_OFFSET, &tmp);
> +       mutex_unlock(&gc->locker);
> +
> +       if (tmp & BIT(offset))
> +               return GPIOF_DIR_OUT;
> +
> +       return GPIOF_DIR_IN;
> +}

Do not use GPIOF* flags in driver code, these are for the consumer
API. Just return 0/1.

> +       status = gpiochip_add(&gc->chip);

As mentioned, use gpiochip_add_data(&gc->chip, gc);


> +static struct platform_driver f81504_gpio_driver = {
> +       .driver = {
> +               .name   = F81504_GPIO_NAME,
> +               .owner  = THIS_MODULE,

I saw coccinelle was already complaining about this.

Looking forward to v3!

Yours,
Linus Walleij