2019-02-11 14:24:19

by Thomas Petazzoni

[permalink] [raw]
Subject: [PATCH] clk: clk-gpio: add support for sleeping GPIOs in gpio-gate-clk

The current implementation of gpio-gate-clk enables/disables the clock
using the GPIO in the ->enable() and ->disable() clock callbacks. This
requires the GPIO to be configurable in atomic contexts. While it is
the case for most memory-mapped GPIO controllers, it is not the case
for GPIO expanders on I2C or SPI.

This commit extends the gpio-gate-clk to check whether the GPIO calls
require sleeping or not. If sleeping is not required, the current
implementation based on ->enable()/->disable() is kept. However, if
sleeping is needed, we instead implement the logic in the ->prepare()
and ->unprepare() hooks. Thanks to this, a gate clock connected to a
GPIO on a GPIO expander can be controlled with the existing driver.

Signed-off-by: Thomas Petazzoni <[email protected]>
---
drivers/clk/clk-gpio.c | 39 +++++++++++++++++++++++++++++++++++++--
1 file changed, 37 insertions(+), 2 deletions(-)

diff --git a/drivers/clk/clk-gpio.c b/drivers/clk/clk-gpio.c
index 6a43ce420492..f5d481713c7d 100644
--- a/drivers/clk/clk-gpio.c
+++ b/drivers/clk/clk-gpio.c
@@ -61,6 +61,35 @@ const struct clk_ops clk_gpio_gate_ops = {
};
EXPORT_SYMBOL_GPL(clk_gpio_gate_ops);

+static int clk_sleeping_gpio_gate_prepare(struct clk_hw *hw)
+{
+ struct clk_gpio *clk = to_clk_gpio(hw);
+
+ gpiod_set_value_cansleep(clk->gpiod, 1);
+
+ return 0;
+}
+
+static void clk_sleeping_gpio_gate_unprepare(struct clk_hw *hw)
+{
+ struct clk_gpio *clk = to_clk_gpio(hw);
+
+ gpiod_set_value_cansleep(clk->gpiod, 0);
+}
+
+static int clk_sleeping_gpio_gate_is_prepared(struct clk_hw *hw)
+{
+ struct clk_gpio *clk = to_clk_gpio(hw);
+
+ return gpiod_get_value_cansleep(clk->gpiod);
+}
+
+const struct clk_ops clk_sleeping_gpio_gate_ops = {
+ .prepare = clk_sleeping_gpio_gate_prepare,
+ .unprepare = clk_sleeping_gpio_gate_unprepare,
+ .is_prepared = clk_sleeping_gpio_gate_is_prepared,
+};
+
/**
* DOC: basic clock multiplexer which can be controlled with a gpio output
* Traits of this clock:
@@ -147,10 +176,16 @@ struct clk_hw *clk_hw_register_gpio_gate(struct device *dev, const char *name,
const char *parent_name, struct gpio_desc *gpiod,
unsigned long flags)
{
+ const struct clk_ops *ops;
+
+ if (gpiod_cansleep(gpiod))
+ ops = &clk_sleeping_gpio_gate_ops;
+ else
+ ops = &clk_gpio_gate_ops;
+
return clk_register_gpio(dev, name,
(parent_name ? &parent_name : NULL),
- (parent_name ? 1 : 0), gpiod, flags,
- &clk_gpio_gate_ops);
+ (parent_name ? 1 : 0), gpiod, flags, ops);
}
EXPORT_SYMBOL_GPL(clk_hw_register_gpio_gate);

--
2.20.1



2019-02-13 22:51:37

by Stephen Boyd

[permalink] [raw]
Subject: Re: [PATCH] clk: clk-gpio: add support for sleeping GPIOs in gpio-gate-clk

Quoting Thomas Petazzoni (2019-02-11 05:58:18)
> The current implementation of gpio-gate-clk enables/disables the clock
> using the GPIO in the ->enable() and ->disable() clock callbacks. This
> requires the GPIO to be configurable in atomic contexts. While it is
> the case for most memory-mapped GPIO controllers, it is not the case
> for GPIO expanders on I2C or SPI.
>
> This commit extends the gpio-gate-clk to check whether the GPIO calls
> require sleeping or not. If sleeping is not required, the current
> implementation based on ->enable()/->disable() is kept. However, if
> sleeping is needed, we instead implement the logic in the ->prepare()
> and ->unprepare() hooks. Thanks to this, a gate clock connected to a
> GPIO on a GPIO expander can be controlled with the existing driver.
>
> Signed-off-by: Thomas Petazzoni <[email protected]>

Sounds like a good idea!

> diff --git a/drivers/clk/clk-gpio.c b/drivers/clk/clk-gpio.c
> index 6a43ce420492..f5d481713c7d 100644
> --- a/drivers/clk/clk-gpio.c
> +++ b/drivers/clk/clk-gpio.c
> @@ -61,6 +61,35 @@ const struct clk_ops clk_gpio_gate_ops = {
> };
> EXPORT_SYMBOL_GPL(clk_gpio_gate_ops);
>
> +static int clk_sleeping_gpio_gate_prepare(struct clk_hw *hw)
> +{
> + struct clk_gpio *clk = to_clk_gpio(hw);
> +
> + gpiod_set_value_cansleep(clk->gpiod, 1);
> +
> + return 0;
> +}
> +
> +static void clk_sleeping_gpio_gate_unprepare(struct clk_hw *hw)
> +{
> + struct clk_gpio *clk = to_clk_gpio(hw);
> +
> + gpiod_set_value_cansleep(clk->gpiod, 0);
> +}
> +
> +static int clk_sleeping_gpio_gate_is_prepared(struct clk_hw *hw)
> +{
> + struct clk_gpio *clk = to_clk_gpio(hw);
> +
> + return gpiod_get_value_cansleep(clk->gpiod);
> +}
> +
> +const struct clk_ops clk_sleeping_gpio_gate_ops = {

static?

> + .prepare = clk_sleeping_gpio_gate_prepare,
> + .unprepare = clk_sleeping_gpio_gate_unprepare,
> + .is_prepared = clk_sleeping_gpio_gate_is_prepared,
> +};
> +
> /**
> * DOC: basic clock multiplexer which can be controlled with a gpio output
> * Traits of this clock:

2019-02-21 22:07:29

by Stephen Boyd

[permalink] [raw]
Subject: Re: [PATCH] clk: clk-gpio: add support for sleeping GPIOs in gpio-gate-clk

Quoting Stephen Boyd (2019-02-13 09:14:22)
>
> static?
>

Ok I marked it static and applied to clk-next.


2019-02-22 08:38:53

by Thomas Petazzoni

[permalink] [raw]
Subject: Re: [PATCH] clk: clk-gpio: add support for sleeping GPIOs in gpio-gate-clk

Hello,

On Thu, 21 Feb 2019 14:05:09 -0800
Stephen Boyd <[email protected]> wrote:

> Quoting Stephen Boyd (2019-02-13 09:14:22)
> >
> > static?
>
> Ok I marked it static and applied to clk-next.

Thanks, and sorry for not getting back with a v2 in a timely fashion. I
had several patch series in-flight for this project, and for several of
them, additional iterations were needed, so I kind of lost track of
this one. Sorry about that.

Thanks again,

Thomas
--
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com