[PATCH v3 1/2] GPIO: fxl6408: Add support for FXL6408 GPIO expander

Oleksandr Suvorov cryosay at gmail.com
Wed Sep 15 20:35:32 CEST 2021


Hi Michal,

On Fri, Sep 10, 2021 at 9:35 AM Michal Simek <michal.simek at xilinx.com> wrote:
>
>
>
> On 9/9/21 10:44 PM, Oleksandr Suvorov wrote:
> > From: Oleksandr Suvorov <oleksandr.suvorov at toradex.com>
> >
> > Initial support for Fairchild's 8 bit I2C gpio expander FXL6408.
> > The CONFIG_FXL6408_GPIO define enables support for such devices.
> >
> > Based on: https://patchwork.kernel.org/patch/9148419/
> >
> > Signed-off-by: Oleksandr Suvorov <oleksandr.suvorov at toradex.com>
> > Reviewed-by: Simon Glass <sjg at chromium.org>
> > Tested-by: Marcel Ziswiler <marcel.ziswiler at toradex.com>
> > Signed-off-by: Oleksandr Suvorov <cryosay at gmail.com>
> > Signed-off-by: Oleksandr Suvorov <oleksandr.suvorov at foundries.io>
>
> 3 emails for you. Would be the best to choose only one.

Thanks, fixed. That habit to always add '-s' when committing changes :)
>
> > ---
> >
> > Changes in v3:
> > - fix a warning:
> >     ...drivers/gpio/gpio-fxl6408.c:348:15: warning: format ‘%ld’
> >     expects argument of type ‘long int’, but argument 3 has
> >     type ‘int’ [-Wformat=]...
> > - add Tested-by record.
> >
> >  drivers/gpio/Kconfig        |   7 +
> >  drivers/gpio/Makefile       |   1 +
> >  drivers/gpio/gpio-fxl6408.c | 366 ++++++++++++++++++++++++++++++++++++
> >  3 files changed, 374 insertions(+)
> >  create mode 100644 drivers/gpio/gpio-fxl6408.c
> >
> > diff --git a/drivers/gpio/Kconfig b/drivers/gpio/Kconfig
> > index 4a89c1a62b..f56e4cc261 100644
> > --- a/drivers/gpio/Kconfig
> > +++ b/drivers/gpio/Kconfig
> > @@ -123,6 +123,13 @@ config DA8XX_GPIO
> >       help
> >         This driver supports the DA8xx GPIO controller
> >
> > +config FXL6408_GPIO
> > +     bool "FXL6408 I2C GPIO expander driver"
> > +     depends on DM_GPIO && DM_I2C
> > +     help
> > +       This driver supports the Fairchild FXL6408 device. FXL6408 is a
> > +       fully configurable 8-bit I2C-controlled GPIO expander.
> > +
> >  config INTEL_BROADWELL_GPIO
> >       bool "Intel Broadwell GPIO driver"
> >       depends on DM
> > diff --git a/drivers/gpio/Makefile b/drivers/gpio/Makefile
> > index 58f4704f6b..83d8b5c9d8 100644
> > --- a/drivers/gpio/Makefile
> > +++ b/drivers/gpio/Makefile
> > @@ -16,6 +16,7 @@ obj-$(CONFIG_AT91_GPIO)     += at91_gpio.o
> >  obj-$(CONFIG_ATMEL_PIO4)     += atmel_pio4.o
> >  obj-$(CONFIG_BCM6345_GPIO)   += bcm6345_gpio.o
> >  obj-$(CONFIG_CORTINA_GPIO)      += cortina_gpio.o
> > +obj-$(CONFIG_FXL6408_GPIO)   += gpio-fxl6408.o
> >  obj-$(CONFIG_INTEL_GPIO)     += intel_gpio.o
> >  obj-$(CONFIG_INTEL_ICH6_GPIO)        += intel_ich6_gpio.o
> >  obj-$(CONFIG_INTEL_BROADWELL_GPIO)   += intel_broadwell_gpio.o
> > diff --git a/drivers/gpio/gpio-fxl6408.c b/drivers/gpio/gpio-fxl6408.c
> > new file mode 100644
> > index 0000000000..dbc68618f9
> > --- /dev/null
> > +++ b/drivers/gpio/gpio-fxl6408.c
> > @@ -0,0 +1,366 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + *  Copyright (C) 2021 Toradex
> > + *  Copyright (C) 2016 Broadcom
> > + */
> > +
> > +/**
>
> Below is not kernel-doc format

I don't think so. And without formatting comments of this block as a
kernel-doc, the kernel-doc tool shows the warning:
drivers/gpio/gpio-fxl6408.c:1: warning: no structured comments found

Otherwise, it generates the correct description of the file:
**FXL6408 I2C to GPIO expander.**


This chip has 8 GPIO lines out of it, and is controlled by an I2C
bus (a pair of lines), providing 4x expansion of GPIO lines. It
also provides an interrupt line out for notifying of state changes.
...

> > + * DOC: FXL6408 I2C to GPIO expander.
> > + *
> > + * This chip has 8 GPIO lines out of it, and is controlled by an I2C
> > + * bus (a pair of lines), providing 4x expansion of GPIO lines. It
> > + * also provides an interrupt line out for notifying of state changes.
> > + *
> > + * Any preconfigured state will be left in place until the GPIO lines
> > + * get activated. At power on, everything is treated as an input,
> > + * default input is HIGH and pulled-up, all interrupts are masked.
> > + *
> > + * Documentation can be found at:
> > + * https://www.fairchildsemi.com/datasheets/FX/FXL6408.pdf
> > + *
> > + * This driver bases on:
> > + * - the original driver by Eric Anholt <eric at anholt.net>:
> > + *   https://patchwork.kernel.org/patch/9148419/
> > + * - the Toradex version by Max Krummenacher <max.krummenacher at toradex.com>:
> > + *   http://git.toradex.com/cgit/linux-toradex.git/tree/drivers/gpio/gpio-fxl6408.c?h=toradex_5.4-2.3.x-imx
> > + * - the U-boot PCA953x driver by Peng Fan <van.freenix at gmail.com>:
> > + *   drivers/gpio/pca953x_gpio.c
> > + *
> > + * TODO:
> > + *   - Add interrupts support
> > + *   - Replace deprecated callbacks direction_input/output() with set_flags()
> > + */
> > +
> > +#include <common.h>
> > +#include <dm.h>
> > +#include <dm/device_compat.h>
> > +#include <i2c.h>
> > +#include <log.h>
> > +#include <asm-generic/gpio.h>
> > +#include <asm/global_data.h>
> > +#include <linux/bitops.h>
> > +#include <dt-bindings/gpio/gpio.h>
>
> please sort it.

Done.

>
> > +
> > +#define REG_DEVID_CTRL               0x01
>
> 0x1

Here and for similar cases below: thanks, I'll fix it in the next revision.

> > +# define SW_RST                      BIT(0)
> > +# define RST_INT             BIT(1)
> > +/* 0b101 is the Manufacturer's ID assigned to Fairchild by Nokia */
> > +# define MF_ID_FAIRCHILD     5
> > +
> > +/* Bits set here indicate that the GPIO is an output */
> > +#define REG_IO_DIR           0x03
>
> 0x3
>
> > +
> > +/*
> > + * Bits set here, when the corresponding bit of REG_IO_DIR is set, drive
> > + * the output high instead of low.
> > + */
> > +#define REG_OUT_STATE                0x05
>
> ditto
>
> > +
> > +/* Bits here make the output High-Z, instead of the OUTPUT value */
> > +#define REG_OUT_HIGH_Z               0x07
>
> ditto
>
>
> > +
> > +/*
> > + * Bits here define the expected input state of the GPIO.
> > + * INTERRUPT_STATUS bits will be set when the INPUT transitions away
> > + * from this value.
> > + */
> > +#define REG_IN_DEFAULT_STATE 0x09
>
> ditto.
>
> > +
> > +/*
> > + * Bits here enable either pull up or pull down according to
> > + * REG_PULL_MODE.
> > + */
> > +#define REG_PULL_ENABLE              0x0b
>
> ditto.
>
> > +
> > +/*
> > + * Bits set here selects a pull-up/pull-down state of pin, which
> > + * is configured as Input and the corresponding REG_PULL_ENABLE bit is
> > + * set.
> > + */
> > +#define REG_PULL_MODE                0x0d
>
> ditto
>
> > +
> > +/* Returns the current status (1 = HIGH) of the input pins */
> > +#define REG_IN_STATUS                0x0f
>
> ditto
>
> > +
> > +/* Mask of pins which can generate interrupts */
> > +#define REG_INT_MASK         0x11
> > +
> > +/* Mask of pins which have generated an interrupt. Cleared on read */
> > +#define REG_INT_STATUS               0x13
> > +
> > +/* Manufacturer's ID getting from Device ID & Ctrl register */
> > +enum {
> > +     MF_ID_MASK = GENMASK(7, 5),
> > +     MF_ID_SHIFT = 5,
>
> This smells. Take a look at include/linux/bitfield.h
> you should be able to use that macros to get fields you want.

>
> > +};
> > +
> > +/* Firmware revision getting from Device ID & Ctrl register */
> > +enum {
> > +     FW_REV_MASK = GENMASK(4, 2),
> > +     FW_REV_SHIFT = 2,
>
> ditto.
>
> > +};
> > +
> > +enum io_direction {
> > +     DIR_IN,
> > +     DIR_OUT,
>
> good style is to initialized them.

Thanks, I'll fix it in the next revision.
>
> > +};
> > +
> > +/*
>
> /** this should be kernel doc.
>
> Run this
> ./scripts/kernel-doc -man -v 1>/dev/null *.c
>
> to check that all values are right.

The kernel-doc doesn't found errors.

> > + * struct fxl6408_info - Data for fxl6408
> > + *
> > + * @dev: udevice structure for the device
> > + * @addr: i2c slave address
> > + * @reg_io_dir: hold the value of direction register
> > + * @reg_output: hold the value of output register
> > + */
> > +struct fxl6408_info {
> > +     struct udevice *dev;
> > +     int addr;
> > +     u8 device_id;
> > +     u8 reg_io_dir;
> > +     u8 reg_output;
> > +};
> > +
> > +static inline int fxl6408_write(struct udevice *dev, int reg, u8 val)
> > +{
> > +     return dm_i2c_write(dev, reg, &val, 1);
> > +}
> > +
> > +static int fxl6408_read(struct udevice *dev, int reg)
> > +{
> > +     int ret;
> > +     u8 tmp;
> > +
> > +     ret = dm_i2c_read(dev, reg, &tmp, 1);
> > +     if (!ret)
> > +             ret = tmp;
>
> This looks completely wrong. What value are you returning in case of error.

Nope. In case of error, I return an error value stored in "ret".

> If ops->xfer is not defined dm_i2c_read returns -ENOSYS. tmp is not
> initialized and can have random value that's why here in case or error
> you will return ramdom value.

!(-ENOSYS) == false, thus "if" op fails and doesn't perform "ret = tmp".
I don't see any errors here.

> > +
> > +     return ret;
> > +}
> > +
> > +/*
>
> /**
>
> > + * fxl6408_is_output() - check whether the gpio configures as either
> > + *                    output or input.
> > + * Return: false - input, true - output.
> > + */
> > +static bool fxl6408_is_output(struct udevice *dev, int offset)
> > +{
> > +     struct fxl6408_info *info = dev_get_plat(dev);
> > +
> > +     return info->reg_io_dir & BIT(offset);
> > +}
> > +
> > +static int fxl6408_get_value(struct udevice *dev, uint offset)
> > +{
> > +     int ret, reg = fxl6408_is_output(dev, offset) ? REG_OUT_STATE : REG_IN_STATUS;
> > +
> > +     ret = fxl6408_read(dev, reg);
> > +     if (ret < 0)
> > +             return ret;
> > +
> > +     return !!(ret & BIT(offset));
> > +}
> > +
> > +static int fxl6408_set_value(struct udevice *dev, uint offset, int value)
> > +{
> > +     struct fxl6408_info *info = dev_get_plat(dev);
> > +     u8 val;
> > +     int ret;
> > +
> > +     if (value)
> > +             val = info->reg_output | BIT(offset);
> > +     else
> > +             val = info->reg_output & ~BIT(offset);
> > +
> > +     ret = fxl6408_write(dev, REG_OUT_STATE, val);
> > +     if (ret < 0)
> > +             return ret;
> > +
> > +     info->reg_output = val;
> > +
> > +     return 0;
> > +}
> > +
> > +static int fxl6408_set_direction(struct udevice *dev, uint offset,
> > +                              enum io_direction dir)
> > +{
> > +     struct fxl6408_info *info = dev_get_plat(dev);
> > +     u8 val;
> > +     int ret;
> > +
> > +     if (dir == DIR_IN)
> > +             val = info->reg_io_dir & ~BIT(offset);
> > +     else
> > +             val = info->reg_io_dir | BIT(offset);
> > +
> > +     ret = fxl6408_write(dev, REG_IO_DIR, val);
> > +     if (ret < 0)
> > +             return ret;
> > +
> > +     info->reg_io_dir = val;
> > +
> > +     return 0;
> > +}
> > +
> > +static int fxl6408_direction_input(struct udevice *dev, uint offset)
> > +{
> > +     return fxl6408_set_direction(dev, offset, DIR_IN);
> > +}
> > +
> > +static int fxl6408_direction_output(struct udevice *dev, uint offset, int value)
> > +{
> > +     int ret;
> > +
> > +     /* Configure output value */
> > +     ret = fxl6408_set_value(dev, offset, value);
> > +     if (ret < 0)
> > +             return ret;
>
> newline

Done.

>
> > +     /* Configure direction as output */
> > +     fxl6408_set_direction(dev, offset, DIR_OUT);
> > +
> > +     return 0;
> > +}
> > +
> > +static int fxl6408_get_function(struct udevice *dev, uint offset)
> > +{
> > +     if (fxl6408_is_output(dev, offset))
> > +             return GPIOF_OUTPUT;
> > +     else
>
> you can remove this else

Sure, thank you.

> > +             return GPIOF_INPUT;
> > +}
> > +
> > +static int fxl6408_xlate(struct udevice *dev, struct gpio_desc *desc,
> > +                      struct ofnode_phandle_args *args)
> > +{
> > +     desc->offset = args->args[0];
> > +     desc->flags = args->args[1] & GPIO_ACTIVE_LOW ? GPIOD_ACTIVE_LOW : 0;
> > +
> > +     return 0;
> > +}
> > +
> > +static const struct dm_gpio_ops fxl6408_ops = {
> > +     .direction_input        = fxl6408_direction_input,
> > +     .direction_output       = fxl6408_direction_output,
> > +     .get_value              = fxl6408_get_value,
> > +     .set_value              = fxl6408_set_value,
> > +     .get_function           = fxl6408_get_function,
> > +     .xlate                  = fxl6408_xlate,
> > +};
> > +
> > +static int fxl6408_probe(struct udevice *dev)
> > +{
> > +     struct fxl6408_info *info = dev_get_plat(dev);
> > +     struct gpio_dev_priv *uc_priv = dev_get_uclass_priv(dev);
> > +     char name[32], label[32], *str;
> > +     int addr;
> > +     int ret;
> > +     int size;
>
> put them on the same line.

Done, thanks.

> > +     const u8 *tmp;
> > +     u32 val32;
> > +
> > +     addr = dev_read_addr(dev);
> > +     if (addr == 0)
> > +             return -EINVAL;
>
> newline

Done.

> > +     info->addr = addr;
> > +
> > +     /*
> > +      * Check the device ID register to see if it's responding.
> > +      * This also clears RST_INT as a side effect, so we won't get
> > +      * the "we've been power cycled" interrupt once interrupts
> > +      * being enabled.
> > +      */
> > +     ret = fxl6408_read(dev, REG_DEVID_CTRL);
> > +     if (ret < 0) {
> > +             dev_err(dev, "FXL6408 probe returned %d\n", ret);
> > +             return ret;
> > +     }
> > +
> > +     if ((ret & MF_ID_MASK) >> MF_ID_SHIFT != MF_ID_FAIRCHILD) {
> > +             dev_err(dev, "FXL6408 probe: wrong Manufacturer's ID: 0x%02x\n", ret);
> > +             return -ENXIO;
> > +     }
> > +     info->device_id = ret;
> > +
> > +     /*
> > +      * Disable High-Z of outputs, so that the OUTPUT updates
> > +      * actually take effect.
> > +      */
> > +     ret = fxl6408_write(dev, REG_OUT_HIGH_Z, (u8)0);
> > +     if (ret < 0) {
> > +             dev_err(dev, "Error writing High-Z register\n");
> > +             return ret;
> > +     }
> > +
> > +     /*
> > +      * If configured, set initial output state and direction,
> > +      * otherwise read them from the chip.
> > +      */
> > +     if (dev_read_u32(dev, "initial_io_dir", &val32)) {
>
> Where do you have dt binding document for this chip? I can't see
> anything in the linux kernel or in your series.
>
> It is good that you read values from dt but you shouldn't do it in
> probe. You should create platdata and fill it by information from DT and
> then in probe just read them from platdata structure.
> This applied to all dt read in probe.
>
>
> > +             ret = fxl6408_read(dev, REG_IO_DIR);
> > +             if (ret < 0) {
> > +                     dev_err(dev, "Error reading direction register\n");
> > +                     return ret;
> > +             }
> > +             info->reg_io_dir = ret;
> > +     } else {
> > +             info->reg_io_dir = val32 & 0xFF;
>
> val32 is not initialized above. dev_read_u32 above fails that's why
> val32 wont't be initialized and here you do calculation with it.
> That's quite weird behavior.  The same pattern visible below.

Nope. dev_read_u32() returns 0 on success. The logic is correct.

> > +             ret = fxl6408_write(dev, REG_IO_DIR, info->reg_io_dir);
> > +             if (ret < 0) {
> > +                     dev_err(dev, "Error setting direction register\n");
> > +                     return ret;
> > +             }
> > +     }
> > +
> > +     if (dev_read_u32(dev, "initial_output", &val32)) {
> > +             ret = fxl6408_read(dev, REG_OUT_STATE);
> > +             if (ret < 0) {
> > +                     dev_err(dev, "Error reading output register\n");
> > +                     return ret;
> > +             }
> > +             info->reg_output = ret;
> > +     } else {
> > +             info->reg_output = val32 & 0xFF;
> > +             ret = fxl6408_write(dev, REG_OUT_STATE, info->reg_output);
> > +             if (ret < 0) {
> > +                     dev_err(dev, "Error setting output register\n");
> > +                     return ret;
> > +             }
> > +     }
> > +
> > +     tmp = dev_read_prop(dev, "label", &size);
> > +     if (tmp) {
> > +             size = min(size, (int)sizeof(label) - 1);
> > +             memcpy(label, tmp, size);
> > +             label[size] = '\0';
> > +             snprintf(name, sizeof(name), "%s@%x_", label, info->addr);
> > +     } else {
> > +             snprintf(name, sizeof(name), "gpio@%x_", info->addr);
> > +     }
>
> I see some code around labels in gpio-uclass. I would be surprised if it
> is not there already because it should be core functionality not driver
> when labels are defined.

This code is not about gpios labeling. It's about the bank name, it
just sets the uc_priv->bank_name which is used in gpio-uclass.

> > +
> > +     str = strdup(name);
> > +     if (!str)
> > +             return -ENOMEM;
> > +
> > +     uc_priv->bank_name = str;
> > +     uc_priv->gpio_count = dev_get_driver_data(dev);
> > +     uc_priv->gpio_base = -1;
> > +
> > +     dev_dbg(dev, "%s (FW rev. %d) is ready\n", str,
> > +             (info->device_id & MF_ID_MASK) >> MF_ID_SHIFT);
> > +
> > +     return 0;
> > +}
> > +
> > +static const struct udevice_id fxl6408_ids[] = {
> > +     { .compatible = "fcs,fxl6408", .data = 8 },
>
> nit: Are you going to support more chip that you have data here?

I can't be sure - I don't know plans of Fairchild or other manufacturers about
gpio devices similar to that one :)

>
> > +     { }
> > +};
> > +
> > +U_BOOT_DRIVER(fxl6408_gpio) = {
> > +     .name = "fxl6408_gpio",
> > +     .id = UCLASS_GPIO,
> > +     .ops = &fxl6408_ops,
> > +     .probe = fxl6408_probe,
> > +     .of_match = fxl6408_ids,
> > +     .plat_auto = sizeof(struct fxl6408_info),
> > +};
> >
>
> Thanks,
> Michal

Thanks for such a detailed and helpful review!

--
Best regards
Oleksandr


Oleksandr Suvorov
cryosay at gmail.com


More information about the U-Boot mailing list