Re: [RFC Patch] gpio: add GPIO hogging mechanism

From: Alexandre Courbot
Date: Wed Oct 29 2014 - 03:10:39 EST


Sorry for the delay in reviewing. Adding Jiri and Pantelis who might
want to extend over this patch and thus will likely have interesting
comments to make.

On Wed, Oct 22, 2014 at 5:09 AM, Benoit Parrot <bparrot@xxxxxx> wrote:
> Based on Boris Brezillion work this is a reworked patch
> of his initial GPIO hogging mechanism.
> This patch provides a way to initally configure specific GPIO
> when the gpio controller is probe.

Typo nit: s/probe/probed

>
> The actual DT scanning to collect the GPIO specific data is performed
> as part of the gpiochip_add().
>
> The purpose of this is to allows specific GPIOs to be configured
> without any driver specific code.
> This particularly usueful because board design are getting

s/suseful/useful

> increassingly complex and given SoC pins can now have upward

s/increassingly/increasingly

> of 10 mux values a lot of connections are now dependent on
> external IO muxes to switch various modes and combination.
>
> Specific drivers should not necessarily need to be aware of
> what accounts to a specific board implementation. This board level
> "description" should be best kept as part of the dts file.
>
> Signed-off-by: Benoit Parrot <bparrot@xxxxxx>
> ---
> Documentation/devicetree/bindings/gpio/gpio.txt | 33 +++++++++
> drivers/gpio/gpiolib-of.c | 99 +++++++++++++++++++++++++
> drivers/gpio/gpiolib.c | 81 ++++++++++++++++++++
> include/linux/of_gpio.h | 11 +++
> 4 files changed, 224 insertions(+)
>
> diff --git a/Documentation/devicetree/bindings/gpio/gpio.txt b/Documentation/devicetree/bindings/gpio/gpio.txt
> index 3fb8f53..a9700d8 100644
> --- a/Documentation/devicetree/bindings/gpio/gpio.txt
> +++ b/Documentation/devicetree/bindings/gpio/gpio.txt

Note: changes to DT bindings documentation should generally come as a
separate patch.

> @@ -103,6 +103,26 @@ Every GPIO controller node must contain both an empty "gpio-controller"
> property, and a #gpio-cells integer property, which indicates the number of
> cells in a gpio-specifier.
>
> +It might contain GPIO hog definitions. GPIO hogging is a mechanism providing
> +automatic GPIO request and configuration as part of the gpio-controller's
> +driver probe function.
> +
> +Each GPIO hog definition is represented as a child node of the GPIO controller.
> +Required properties:
> +- gpios: store the gpio information (id, flags, ...). Shall contain the
> + number of cells specified in its parent node (GPIO controller node).

If would be nice to be able to specify several GPIO references to that
they can be all set to the same configuration in one row instead of
requiring one node each.

> +- input: a property specifying to set the GPIO direction as input.
> +- output-high: a property specifying to set the GPIO direction to output with
> + the value high.
> +- output-low: a property specifying to set the GPIO direction to output with
> + the value low.

Wouldn't a "direction" property taking one of the values "input",
"output-low" or "output-high" be easier to parse and generally
clearer?

> +
> +Optional properties:
> +- line-name: the GPIO label name. If not present the node name is used.

I guess we also could use this for naming the GPIO during its export
if we decide to allow this in a near-future patch.

> +
> +A GPIO controller can request GPIO hogs using the "gpio-hogs" property, which
> +encodes a list of requested GPIO hogs.
> +
> Example of two SOC GPIO banks defined as gpio-controller nodes:
>
> qe_pio_a: gpio-controller@1400 {
> @@ -110,6 +130,19 @@ Example of two SOC GPIO banks defined as gpio-controller nodes:
> reg = <0x1400 0x18>;
> gpio-controller;
> #gpio-cells = <2>;
> + gpio-hogs = <&line_b>;
> +
> + /* line_a hog is defined but not enabled in this example*/
> + line_a: line_a {
> + gpios = <5 0>;
> + input;
> + };
> +
> + line_b: line_b {
> + gpios = <6 0>;
> + output-low;
> + line-name = "foo-bar-gpio";
> + };

I think it might be good to group the hog nodes such as to allow
complex configurations to be set in one go, Ã la pinmux:

gpio-controller;
#gpio-cells = <2>;
+ gpio-hogs = <&line_b>;
+
+ line_a: line_a {
+ line_a {
+ gpios = <5 0>;
+ input;
+ };
+ line_c {
+ gpios = <7 0>;
+ inputl
+ };
+ };
+
+ line_b: line_b {
+ line_b {
+ gpios = <6 0>;
+ output-low;
+ line-name = "foo-bar-gpio";
+ };
+ };

Here if you want to enable the first configuration you don't need to
specify all the lines one by one, you just need to select the right
group.

I wonder if such usage of child nodes could not interfere with GPIO
drivers (existing or to be) that need to use child nodes for other
purposes. Right now there is no way to distinguish a hog node from a
node that serves another purpose, and that might become a problem in
the future. Not sure how that should be addressed though - maybe by
enforcing a "-hog" suffix for the group nodes, e.g. line_a-hog?
Comments from people more experienced with DT would be nice.

> };
>
> qe_pio_e: gpio-controller@1460 {
> diff --git a/drivers/gpio/gpiolib-of.c b/drivers/gpio/gpiolib-of.c
> index 604dbe6..7308dfc 100644
> --- a/drivers/gpio/gpiolib-of.c
> +++ b/drivers/gpio/gpiolib-of.c
> @@ -111,6 +111,105 @@ int of_get_named_gpio_flags(struct device_node *np, const char *list_name,
> EXPORT_SYMBOL(of_get_named_gpio_flags);
>
> /**
> + * of_get_gpio_hog() - Get a GPIO hog descriptor, names and flags for GPIO API
> + * @np: device node to get GPIO from
> + * @index: index of the GPIO
> + * @name: GPIO line name
> + * @flags: a flags pointer to fill in
> + *
> + * Returns GPIO descriptor to use with Linux GPIO API, or one of the errno
> + * value on the error condition.
> + */
> +struct gpio_desc *of_get_gpio_hog(struct device_node *np, int index,
> + const char **name, unsigned long *flags)
> +{
> + struct device_node *cfg_np, *chip_np;
> + enum of_gpio_flags xlate_flags;
> + unsigned long req_flags = 0;
> + struct gpio_desc *desc;
> + struct gg_data gg_data = {
> + .flags = &xlate_flags,
> + .out_gpio = NULL,
> + };
> + u32 tmp;
> + int i;
> + int ret;
> +
> + cfg_np = of_parse_phandle(np, "gpio-hogs", index);
> + if (!cfg_np)
> + return ERR_PTR(-EINVAL);
> +
> + chip_np = cfg_np->parent;
> + if (!chip_np) {
> + desc = ERR_PTR(-EINVAL);
> + goto out;
> + }
> +
> + ret = of_property_read_u32(chip_np, "#gpio-cells", &tmp);
> + if (ret) {
> + desc = ERR_PTR(ret);
> + goto out;
> + }
> +
> + if (tmp > MAX_PHANDLE_ARGS) {
> + desc = ERR_PTR(-EINVAL);
> + goto out;
> + }
> +
> + gg_data.gpiospec.args_count = tmp;
> + gg_data.gpiospec.np = chip_np;
> + for (i = 0; i < tmp; i++) {
> + ret = of_property_read_u32(cfg_np, "gpios",
> + &gg_data.gpiospec.args[i]);
> + if (ret) {
> + desc = ERR_PTR(ret);
> + goto out;
> + }
> + }
> +
> + gpiochip_find(&gg_data, of_gpiochip_find_and_xlate);
> + if (!gg_data.out_gpio) {
> + if (cfg_np->parent == np)
> + desc = ERR_PTR(-ENXIO);
> + else
> + desc = ERR_PTR(-EPROBE_DEFER);
> +
> + goto out;
> + }
> +
> + if (xlate_flags & OF_GPIO_ACTIVE_LOW)
> + req_flags |= GPIOF_ACTIVE_LOW;
> +
> + if (of_property_read_bool(cfg_np, "input")) {
> + req_flags |= GPIOF_DIR_IN;
> + } else if (of_property_read_bool(cfg_np, "output-high")) {
> + req_flags |= GPIOF_INIT_HIGH;
> + } else if (!of_property_read_bool(cfg_np, "output-low")) {
> + desc = ERR_PTR(-EINVAL);
> + goto out;
> + }

That's why I would prefer a "direction" property instead of these 3
ones: because if you have the following node:

line_foo {
gpios = <1 GPIO_ACTIVE_HIGH>;
input;
output-high;
};

Then this code will not trigger an error and will just configure the
GPIO as input.

> +
> + if (of_property_read_bool(cfg_np, "open-drain-line"))
> + req_flags |= GPIOF_OPEN_DRAIN;
> +
> + if (of_property_read_bool(cfg_np, "open-source-line"))
> + req_flags |= GPIOF_OPEN_DRAIN;
> +
> + if (name && of_property_read_string(cfg_np, "line-name", name))
> + *name = cfg_np->name;
> +
> + if (flags)
> + *flags = req_flags;
> +
> + desc = gg_data.out_gpio;
> +
> +out:
> + of_node_put(cfg_np);
> +
> + return desc;
> +}
> +
> +/**
> * of_gpio_simple_xlate - translate gpio_spec to the GPIO number and flags
> * @gc: pointer to the gpio_chip structure
> * @np: device node of the GPIO chip
> diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
> index e8e98ca..b6f5bdb 100644
> --- a/drivers/gpio/gpiolib.c
> +++ b/drivers/gpio/gpiolib.c
> @@ -55,6 +55,9 @@ static DEFINE_MUTEX(gpio_lookup_lock);
> static LIST_HEAD(gpio_lookup_list);
> LIST_HEAD(gpio_chips);
>
> +struct gpio_desc *__must_check gpiod_get_hog_index(struct device *dev,
> + unsigned int idx);
> +
> static inline void desc_set_label(struct gpio_desc *d, const char *label)
> {
> d->label = label;
> @@ -226,6 +229,8 @@ int gpiochip_add(struct gpio_chip *chip)
> int status = 0;
> unsigned id;
> int base = chip->base;
> + struct gpio_desc *desc;
> + int i;
>
> if ((!gpio_is_valid(base) || !gpio_is_valid(base + chip->ngpio - 1))
> && base >= 0) {
> @@ -275,6 +280,16 @@ int gpiochip_add(struct gpio_chip *chip)
> of_gpiochip_add(chip);
> acpi_gpiochip_add(chip);
>
> + /*
> + * Instantiate GPIO hogs
> + * There shouldn't be mores hogs then gpio available on this chip

s/then/than

> + */
> + for (i = 0; i < chip->ngpio; i++) {
> + desc = gpiod_get_hog_index(chip->dev, i);
> + if (IS_ERR(desc))
> + break;
> + }

chip->ngpio? Isn't there a better way to know the number of phandles
in your gpio-hogs property?

Also you may have an error returned either because you reached the end
of the list, or because the hog configuration itself failed. In the
latter case don't you want to continue to go through the list?

Finally your desc variable should be declared within this loop since
it is not used elsewhere.

> +
> if (status)
> goto fail;
>
> @@ -1742,6 +1757,72 @@ struct gpio_desc *__must_check __gpiod_get_index_optional(struct device *dev,
> EXPORT_SYMBOL_GPL(__gpiod_get_index_optional);
>
> /**
> + * gpiod_get_hog_index - obtain a GPIO from a multi-index GPIO hog
> + * @dev: GPIO consumer
> + * @idx: index of the GPIO to obtain
> + *
> + * This should only be used by the gpiochip_add to request/set GPIO initial
> + * configuration.
> + *
> + * Return a valid GPIO descriptor, or an IS_ERR() condition in case of error.
> + */
> +struct gpio_desc *__must_check gpiod_get_hog_index(struct device *dev,
> + unsigned int idx)
> +{
> + struct gpio_desc *desc = NULL;
> + int err;
> + unsigned long flags;
> + const char *name;
> +
> + /* Using device tree? */
> + if (IS_ENABLED(CONFIG_OF) && dev && dev->of_node)
> + desc = of_get_gpio_hog(dev->of_node, idx, &name, &flags);
> +
> + if (!desc)
> + return ERR_PTR(-ENOTSUPP);
> + else if (IS_ERR(desc))
> + return desc;
> +
> + dev_dbg(dev, "gpio-hog: GPIO:%d (%s) as %s%s\n",
> + desc_to_gpio(desc), name, (flags&GPIOF_DIR_IN)?"input":"output",
> + (flags&GPIOF_DIR_IN)?"":(flags&GPIOF_INIT_HIGH)?"/high":"/low");
> +

...

> + err = gpiod_request(desc, name);
> + if (err)
> + return ERR_PTR(err);
> +
> + if (flags & GPIOF_OPEN_DRAIN)
> + set_bit(FLAG_OPEN_DRAIN, &desc->flags);
> +
> + if (flags & GPIOF_OPEN_SOURCE)
> + set_bit(FLAG_OPEN_SOURCE, &desc->flags);
> +
> + if (flags & GPIOF_ACTIVE_LOW)
> + set_bit(FLAG_ACTIVE_LOW, &desc->flags);
> +
> + if (flags & GPIOF_DIR_IN)
> + err = gpiod_direction_input(desc);
> + else
> + err = gpiod_direction_output_raw(desc,
> + (flags & GPIOF_INIT_HIGH) ? 1 : 0);
> +

This part of the code is very similar to what is found in
__gpiod_get_index. Could you maybe try to extract the common code into
its own function and call it from both sites instead of duplicating
code?

> + if (err)
> + goto free_gpio;
> +
> + if (flags & GPIOF_EXPORT) {
> + err = gpiod_export(desc, flags & GPIOF_EXPORT_CHANGEABLE);
> + if (err)
> + goto free_gpio;

Right now export is not possible so I don't think you need that block.

> + }
> +
> + return desc;
> +
> +free_gpio:
> + gpiod_free(desc);
> + return ERR_PTR(err);
> +}
> +
> +/**
> * gpiod_put - dispose of a GPIO descriptor
> * @desc: GPIO descriptor to dispose of
> *
> diff --git a/include/linux/of_gpio.h b/include/linux/of_gpio.h
> index 38fc050..52d154c 100644
> --- a/include/linux/of_gpio.h
> +++ b/include/linux/of_gpio.h
> @@ -59,6 +59,9 @@ extern int of_gpio_simple_xlate(struct gpio_chip *gc,
> const struct of_phandle_args *gpiospec,
> u32 *flags);
>
> +extern struct gpio_desc *of_get_gpio_hog(struct device_node *np, int index,
> + const char **name,
> + unsigned long *flags);

This function is gpiolib-private, therefore it should be declared in
drivers/gpio/gpiolib.h alongside with the declaration of
of_get_named_gpiod_flags.

If I understood the latest discussions correctly we might want to add
an export (and named export) feature on top of this patch. Linus, am I
correct to understand that you are not opposed to both features? In
this case, we might want to go ahead with the merging of one of the
named GPIOs patches, so that Jiri and Pantelis can implement export
based on this patch.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/