[v4,0/2] Support pwm driver for aspeed ast26xx

The legacy driver of aspeed pwm is binding with tach controller and it
doesn't follow the pwm framworks usage. In addition, the pwm register
usage of the 6th generation of ast26xx has drastic change. So these
patch serials add the new aspeed pwm driver to fix up the problem above.

Changes since v3:
- Add the dt_binding for aspeed,ast2600-tach.
- Describe the pwm/tach as child-node of pwm-tach mfd.
- Complete the properties of pwm node.

Changes since v2:
- Remove the tach node, #address-cells and #size-cells from pwm-tach.yaml
- Add clocks and reset properties to pwm-tach.yaml
- Kconfig/Makfile sorted alphabetically
- pwm-aspeed-g6.c suggested by Uwe Kleine-König
  - Add more hardware descriptions at top of the driver.
  - Remove unused api request and free
  - Move the initialize settings of all pwm channel to probe.
  - Change the method of getting the approximate period.
  - Read the hardware register values to fill the state for .get_state()

Changes since v1:
- Fix the dt_binding_check fail suggested by Rob Herring
- Add depends to PWM_ASPEED_G6 configure suggested by Uwe Kleine-Konig
- pwm-aspeed-g6.c suggested by Uwe Kleine-König
  - Fix license header
  - Use bitfiled.h macro to define register fields
  - Implement .remove device function
  - Implement .get_state pwm api

Billy Tsai (2):
  dt-bindings: Add bindings for aspeed pwm-tach.
  pwm: Add Aspeed ast2600 PWM support

 .../bindings/hwmon/aspeed,ast2600-tach.yaml   |  66 ++++
 .../bindings/mfd/aspeed,ast2600-pwm-tach.yaml |  82 ++++
 .../bindings/pwm/aspeed,ast2600-pwm.yaml      |  62 +++
 drivers/pwm/Kconfig                           |   8 +
 drivers/pwm/Makefile                          |   1 +
 drivers/pwm/pwm-aspeed-g6.c                   | 368 ++++++++++++++++++
 6 files changed, 587 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/hwmon/aspeed,ast2600-tach.yaml
 create mode 100644 Documentation/devicetree/bindings/mfd/aspeed,ast2600-pwm-tach.yaml
 create mode 100644 Documentation/devicetree/bindings/pwm/aspeed,ast2600-pwm.yaml
 create mode 100644 drivers/pwm/pwm-aspeed-g6.c

Hello,

On Mon, May 10, 2021 at 03:35:11PM +0800, Billy Tsai wrote:
> This patch add the support of PWM controller which can be found at aspeed
> ast2600 soc. The pwm supoorts up to 16 channels and it's part function
> of multi-function device "pwm-tach controller".
> 
> Signed-off-by: Billy Tsai <billy_tsai@aspeedtech.com>
> ---
>  drivers/pwm/Kconfig         |   8 +
>  drivers/pwm/Makefile        |   1 +
>  drivers/pwm/pwm-aspeed-g6.c | 368 ++++++++++++++++++++++++++++++++++++
>  3 files changed, 377 insertions(+)
>  create mode 100644 drivers/pwm/pwm-aspeed-g6.c
> 
> diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig
> index 63be5362fd3a..b0964b9a3273 100644
> --- a/drivers/pwm/Kconfig
> +++ b/drivers/pwm/Kconfig
> @@ -51,6 +51,14 @@ config PWM_AB8500
>  	  To compile this driver as a module, choose M here: the module
>  	  will be called pwm-ab8500.
>  
> +config PWM_ASPEED_G6
> +	tristate "ASPEEDG6 PWM support"
> +	depends on ARCH_ASPEED || COMPILE_TEST
> +	help
> +	  Generic PWM framework driver for ASPEED G6 SoC.
> +
> +	  This driver provides support for ASPEED G6 PWM controllers.
> +
>  config PWM_ATMEL
>  	tristate "Atmel PWM support"
>  	depends on OF
> diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile
> index cbdcd55d69ee..29d22d806e68 100644
> --- a/drivers/pwm/Makefile
> +++ b/drivers/pwm/Makefile
> @@ -2,6 +2,7 @@
>  obj-$(CONFIG_PWM)		+= core.o
>  obj-$(CONFIG_PWM_SYSFS)		+= sysfs.o
>  obj-$(CONFIG_PWM_AB8500)	+= pwm-ab8500.o
> +obj-$(CONFIG_PWM_ASPEED_G6)	+= pwm-aspeed-g6.o
>  obj-$(CONFIG_PWM_ATMEL)		+= pwm-atmel.o
>  obj-$(CONFIG_PWM_ATMEL_HLCDC_PWM)	+= pwm-atmel-hlcdc.o
>  obj-$(CONFIG_PWM_ATMEL_TCB)	+= pwm-atmel-tcb.o
> diff --git a/drivers/pwm/pwm-aspeed-g6.c b/drivers/pwm/pwm-aspeed-g6.c
> new file mode 100644
> index 000000000000..f5bb982c21e6
> --- /dev/null
> +++ b/drivers/pwm/pwm-aspeed-g6.c
> @@ -0,0 +1,368 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/*
> + * Copyright (C) 2021 ASPEED Technology Inc.
> + *
> + * PWM controller driver for Aspeed ast26xx SoCs.
> + * This drivers doesn't rollback to previous version of aspeed SoCs.

Does that mean it doesn't support earlier versions of the IP?

> + *
> + * The formula of pwm frequency:
> + * PWM frequency = CLK Source / ((DIV_L + 1) * BIT(DIV_H) * (PERIOD + 1))
> + *
> + * The software driver fixes the period to 256, which causes the high-frequency
> + * precision of the PWM to be coarse, in exchange for the fineness of the duty cycle.
> + *
> + * Register usage:
> + * PIN_ENABLE: When it is unset the pwm controller will always output low to the extern.
> + * Use to determin PWM channel enable/disable.

s/determin PWM channel enable.disable/determine if the PWM channel is enabled or disabled/

> + * CLK_ENABLE: When it is unset the pwm controller will reset the duty counter to 0 and
> + * output low to the PIN_ENABLE mux after that the driver can still change the pwm period
> + * and duty and the value will apply when CLK_ENABLE be set again.
> + * Use to determin whether duty_cycle bigger than 0.
> + * PWM_ASPEED_INVERSE: When it is toggled the output value will inverse immediately.
> + *
> + * Limitations:
> + * - When changing both duty cycle and period, we cannot prevent in
> + *   software that the output might produce a period with mixed
> + *   settings.
> + *
> + * Improvements:
> + * - When changing the duty cycle or period, our pwm controller will not
> + *   generate the glitch, the configure will change at next cycle of pwm.
> + *   This improvement can disable/enable through PWM_ASPEED_DUTY_SYNC_DISABLE.
> + */
> +
> +#include <linux/clk.h>
> +#include <linux/errno.h>
> +#include <linux/delay.h>
> +#include <linux/io.h>
> +#include <linux/kernel.h>
> +#include <linux/mfd/syscon.h>
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_device.h>
> +#include <linux/platform_device.h>
> +#include <linux/sysfs.h>
> +#include <linux/reset.h>
> +#include <linux/regmap.h>
> +#include <linux/bitfield.h>
> +#include <linux/slab.h>
> +#include <linux/pwm.h>
> +
> +/* The channel number of Aspeed pwm controller */
> +#define PWM_ASPEED_NR_PWMS 16
> +
> +/* PWM Control Register */
> +#define PWM_ASPEED_CTRL_CH(ch) (((ch * 0x10) + 0x00))

Please put ch in parenthesis.

> +#define PWM_ASPEED_LOAD_SEL_RISING_AS_WDT BIT(19)
> +#define PWM_ASPEED_DUTY_LOAD_AS_WDT_ENABLE BIT(18)
> +#define PWM_ASPEED_DUTY_SYNC_DISABLE BIT(17)
> +#define PWM_ASPEED_CLK_ENABLE BIT(16)
> +#define PWM_ASPEED_LEVEL_OUTPUT BIT(15)
> +#define PWM_ASPEED_INVERSE BIT(14)
> +#define PWM_ASPEED_OPEN_DRAIN_ENABLE BIT(13)
> +#define PWM_ASPEED_PIN_ENABLE BIT(12)
> +#define PWM_ASPEED_CLK_DIV_H GENMASK(11, 8)
> +#define PWM_ASPEED_CLK_DIV_L GENMASK(7, 0)

It would be great if the bit definitions contained the name of the
register. This simplifies verifying that we're using a define for the
expected register.

> +/* PWM Duty Cycle Register */
> +#define PWM_ASPEED_DUTY_CYCLE_CH(ch) (((ch * 0x10) + 0x04))
> +#define PWM_ASPEED_PERIOD GENMASK(31, 24)
> +#define PWM_ASPEED_POINT_AS_WDT GENMASK(23, 16)
> +#define PWM_ASPEED_FALLING_POINT GENMASK(15, 8)
> +#define PWM_ASPEED_RISING_POINT GENMASK(7, 0)
> +
> +/* PWM fixed value */
> +#define PWM_ASPEED_FIXED_PERIOD 0xff
> +
> +struct aspeed_pwm_data {
> +	struct pwm_chip chip;
> +	struct clk *clk;
> +	struct regmap *regmap;
> +	struct reset_control *reset;
> +};
> +
> +static inline struct aspeed_pwm_data *
> +aspeed_pwm_chip_to_data(struct pwm_chip *c)
> +{
> +	return container_of(c, struct aspeed_pwm_data, chip);
> +}
> +
> +static void aspeed_set_pwm_clk_enable(struct regmap *regmap, u8 pwm_channel,
> +					  bool enable)

I'd call this function aspeed_pwm_set_clk_enable to use "aspeed_pwm_" as
the common prefix for all functions. Similarly for the other functions.
> +{
> +	regmap_update_bits(regmap, PWM_ASPEED_CTRL_CH(pwm_channel),
> +			   PWM_ASPEED_CLK_ENABLE,
> +			   enable ? PWM_ASPEED_CLK_ENABLE : 0);

There is only a single caller for this function, I'd recommend to fold
it into this caller.

> +}
> +
> +static u32 apseed_get_pwm_freq(struct pwm_chip *chip, struct pwm_device *pwm)
> +{
> +	struct aspeed_pwm_data *priv = aspeed_pwm_chip_to_data(chip);
> +	unsigned long rate;
> +	u32 index = pwm->hwpwm;
> +	u32 div_h, div_l, cur_freq, val;
> +
> +	rate = clk_get_rate(priv->clk);
> +	regmap_read(priv->regmap, PWM_ASPEED_CTRL_CH(index), &val);
> +	div_h = FIELD_GET(PWM_ASPEED_CLK_DIV_H, val);
> +	div_l = FIELD_GET(PWM_ASPEED_CLK_DIV_L, val);
> +
> +	cur_freq = DIV_ROUND_DOWN_ULL(rate, (BIT(div_h) * (div_l + 1) *
> +					     (PWM_ASPEED_FIXED_PERIOD + 1)));
> +	return cur_freq;
> +}
> +
> +static int aspeed_set_pwm_freq(struct pwm_chip *chip, struct pwm_device *pwm,
> +			       const struct pwm_state *state)
> +{
> +	struct device *dev = chip->dev;
> +	struct aspeed_pwm_data *priv = aspeed_pwm_chip_to_data(chip);
> +	unsigned long rate;
> +	u32 div_h, div_l, freq;
> +	u32 index = pwm->hwpwm;
> +	/* Get the smallest value for div_h  */
> +	freq = DIV_ROUND_UP_ULL(NSEC_PER_SEC, state->period);
> +	rate = clk_get_rate(priv->clk);
> +	div_h = DIV_ROUND_DOWN_ULL(rate,
> +				   ((FIELD_MAX(PWM_ASPEED_CLK_DIV_L) + 1) *
> +				    freq * (PWM_ASPEED_FIXED_PERIOD + 1)));

Please don't divide by the result of a division.

> +	div_h = order_base_2(div_h);
> +	if (div_h > 0xf)
> +		div_h = 0xf;
> +
> +	div_l = DIV_ROUND_DOWN_ULL(rate >> div_h,
> +				   (freq * (PWM_ASPEED_FIXED_PERIOD + 1)));
> +	if (div_l == 0) {
> +		dev_err(dev, "Period too small, cannot implement it");
> +		return -ERANGE;
> +	}
> +
> +	div_l -= 1;
> +
> +	if (div_l > 255)
> +		div_l = 255;
> +
> +	dev_dbg(dev, "clk source: %ld div h %x, l : %x\n", rate, div_h, div_l);
> +
> +	regmap_update_bits(priv->regmap, PWM_ASPEED_CTRL_CH(index),
> +			   (PWM_ASPEED_CLK_DIV_H | PWM_ASPEED_CLK_DIV_L),
> +			   FIELD_PREP(PWM_ASPEED_CLK_DIV_H, div_h) |
> +				   FIELD_PREP(PWM_ASPEED_CLK_DIV_L, div_l));
> +	return 0;
> +}
> +
> +static void aspeed_set_pwm_duty(struct pwm_chip *chip, struct pwm_device *pwm,
> +				const struct pwm_state *state)
> +{
> +	struct aspeed_pwm_data *priv = aspeed_pwm_chip_to_data(chip);
> +	u32 duty_pt;
> +	u32 index = pwm->hwpwm;
> +	u32 cur_freq;
> +	u64 cur_period;
> +
> +	cur_freq = apseed_get_pwm_freq(chip, pwm);

s/apseed/aspeed/

cur_freq is the result of a division ...

> +	cur_period = DIV_ROUND_DOWN_ULL(NSEC_PER_SEC, cur_freq);

... dividing through such a number loses precision.


> +	duty_pt = DIV_ROUND_DOWN_ULL(
> +		state->duty_cycle * (PWM_ASPEED_FIXED_PERIOD + 1), cur_period);
> +	if (duty_pt == 0) {
> +		aspeed_set_pwm_clk_enable(priv->regmap, index, false);
> +	} else {
> +		if (duty_pt >= (PWM_ASPEED_FIXED_PERIOD + 1))
> +			duty_pt = 0;
> +		/* When duty_pt = 0 it mean our duty cycle = 100% */

Please put this comment above the code that it explains.

> +		regmap_update_bits(
> +			priv->regmap, PWM_ASPEED_DUTY_CYCLE_CH(index),
> +			PWM_ASPEED_FALLING_POINT,
> +			FIELD_PREP(PWM_ASPEED_FALLING_POINT, duty_pt));
> +		aspeed_set_pwm_clk_enable(priv->regmap, index, true);
> +	}
> +}
> +
> +static void aspeed_set_pwm_polarity(struct pwm_chip *chip,
> +				    struct pwm_device *pwm,
> +				    const struct pwm_state *state)
> +{
> +	struct aspeed_pwm_data *priv = aspeed_pwm_chip_to_data(chip);
> +	u32 index = pwm->hwpwm;
> +
> +	regmap_update_bits(priv->regmap, PWM_ASPEED_CTRL_CH(index),
> +			   PWM_ASPEED_INVERSE,
> +			   FIELD_PREP(PWM_ASPEED_INVERSE, state->polarity));
> +}

This function also only has a single caller and can better be folded
into it.

> +static void aspeed_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm,
> +				 struct pwm_state *state)
> +{
> +	struct device *dev = chip->dev;
> +	struct aspeed_pwm_data *priv = aspeed_pwm_chip_to_data(chip);
> +	u32 index = pwm->hwpwm;
> +	bool polarity, ch_en, clk_en;
> +	u32 duty_pt, val;
> +	u32 cur_freq;
> +
> +	regmap_read(priv->regmap, PWM_ASPEED_CTRL_CH(index), &val);
> +	polarity = FIELD_GET(PWM_ASPEED_INVERSE, val);
> +	ch_en = FIELD_GET(PWM_ASPEED_PIN_ENABLE, val);
> +	clk_en = FIELD_GET(PWM_ASPEED_CLK_ENABLE, val);
> +	regmap_read(priv->regmap, PWM_ASPEED_DUTY_CYCLE_CH(index), &val);
> +	duty_pt = FIELD_GET(PWM_ASPEED_FALLING_POINT, val);
> +
> +	cur_freq = apseed_get_pwm_freq(chip, pwm);
> +
> +	state->period = DIV_ROUND_DOWN_ULL(NSEC_PER_SEC, cur_freq);

Please round up in .get_state().

> +	if (clk_en && duty_pt)
> +		state->duty_cycle = DIV_ROUND_DOWN_ULL(
> +			state->period * duty_pt, PWM_ASPEED_FIXED_PERIOD + 1);
> +	else
> +		state->duty_cycle = clk_en ? state->period : 0;
> +	state->polarity = polarity;
> +	state->enabled = ch_en;
> +	dev_dbg(dev, "get period: %lldns, duty_cycle: %lldns", state->period,
> +		state->duty_cycle);
> +}
> +
> +static int aspeed_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
> +			    const struct pwm_state *state)
> +{
> +	struct device *dev = chip->dev;
> +	struct aspeed_pwm_data *priv = aspeed_pwm_chip_to_data(chip);
> +	u32 index = pwm->hwpwm;
> +	struct pwm_state cur_state;
> +	int ret;
> +
> +	aspeed_pwm_get_state(chip, pwm, &cur_state);
> +	dev_dbg(dev, "cur period: %lldns, cur duty_cycle: %lldns",
> +		cur_state.period, cur_state.duty_cycle);
> +	dev_dbg(dev, "apply period: %lldns, duty_cycle: %lldns", state->period,
> +		state->duty_cycle);
> +	regmap_update_bits(priv->regmap, PWM_ASPEED_CTRL_CH(index),
> +			   PWM_ASPEED_PIN_ENABLE,
> +			   state->enabled ? PWM_ASPEED_PIN_ENABLE : 0);
> +	if (cur_state.period != state->period) {
> +		ret = aspeed_set_pwm_freq(chip, pwm, state);
> +		if (ret)
> +			return ret;
> +		aspeed_set_pwm_duty(chip, pwm, state);
> +	} else if (cur_state.duty_cycle != state->duty_cycle)
> +		aspeed_set_pwm_duty(chip, pwm, state);

In aspeed_set_pwm_duty() you call apseed_get_pwm_freq(), in one of the two
cases above you already know the freq. If you ask me, just drop the
cur_state.period != state->period check and calculate all the required
settings unconditionally. Then you also don't need to call
aspeed_pwm_get_state(). (Apart from that apseed_get_pwm_freq implements
the wrong rounding for .apply().)

> +	aspeed_set_pwm_polarity(chip, pwm, state);
> +	return 0;
> +}
> +
> +static const struct pwm_ops aspeed_pwm_ops = {
> +	.apply = aspeed_pwm_apply,
> +	.get_state = aspeed_pwm_get_state,
> +	.owner = THIS_MODULE,
> +};
> +
> +static int aspeed_pwm_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	int ret, index;
> +	struct aspeed_pwm_data *priv;
> +	struct device_node *np;
> +
> +	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return -ENOMEM;
> +
> +	np = pdev->dev.parent->of_node;
> +	if (!of_device_is_compatible(np, "aspeed,ast2600-pwm-tach")) {
> +		dev_err(dev, "unsupported pwm device binding\n");
> +		return -ENODEV;
> +	}
> +
> +	priv->regmap = syscon_node_to_regmap(np);
> +	if (IS_ERR(priv->regmap)) {
> +		dev_err(dev, "Couldn't get regmap\n");
> +		return -ENODEV;

Please return PTR_ERR(priv->regmap) here and use dev_err_probe for
consistency.

> +	}
> +
> +	priv->clk = of_clk_get(np, 0);

Maybe use devm_clk_get() instead? (Otherwise you have to put the clock
in the error paths.)

> +	if (IS_ERR(priv->clk))
> +		return dev_err_probe(dev, PTR_ERR(priv->clk),
> +				     "get clock failed\n");
> +
> +	ret = clk_prepare_enable(priv->clk);
> +	if (ret) {
> +		dev_err(dev, "couldn't enable clock\n");
> +		return ret;
> +	}
> +
> +	priv->reset = of_reset_control_get_shared(np, NULL);
> +	if (IS_ERR(priv->reset))
> +		return dev_err_probe(dev, PTR_ERR(priv->reset),
> +				     "get reset failed\n");
> +
> +	ret = reset_control_deassert(priv->reset);
> +	if (ret) {
> +		dev_err(dev, "cannot deassert reset control: %pe\n",
> +			ERR_PTR(ret));
> +		clk_disable_unprepare(priv->clk);

Consider using goto for error handling. Then you don't have to repeat
the same thing over and over again.

> +		return ret;
> +	}
> +
> +	priv->chip.dev = dev;
> +	priv->chip.ops = &aspeed_pwm_ops;
> +	priv->chip.npwm = PWM_ASPEED_NR_PWMS;
> +	priv->chip.of_xlate = of_pwm_xlate_with_flags;
> +	priv->chip.of_pwm_n_cells = 3;
> +
> +	/*
> +	 * Fixed the period to the max value and rising point to 0
> +	 * for high resolution and simplified frequency calculation.

There is a stray non-printable char before "simplified".

> +	 */
> +	for (index = 0; index < PWM_ASPEED_NR_PWMS; index++) {
> +		regmap_update_bits(
> +			priv->regmap, PWM_ASPEED_DUTY_CYCLE_CH(index),
> +			PWM_ASPEED_PERIOD,
> +			FIELD_PREP(PWM_ASPEED_PERIOD, PWM_ASPEED_FIXED_PERIOD));
> +		regmap_update_bits(priv->regmap,
> +				   PWM_ASPEED_DUTY_CYCLE_CH(index),
> +				   PWM_ASPEED_RISING_POINT, 0);
> +	}

Please don't modify the output in .probe. .apply is the only function
that is supposed to do that.

> +
> +	ret = pwmchip_add(&priv->chip);
> +	if (ret < 0) {
> +		dev_err(dev, "failed to add PWM chip: %pe\n", ERR_PTR(ret));
> +		reset_control_assert(priv->reset);
> +		clk_disable_unprepare(priv->clk);
> +		return ret;
> +	}
> +	dev_set_drvdata(dev, priv);
> +	return ret;

s/ret/0/

> +}
> +
> +static int aspeed_pwm_remove(struct platform_device *dev)
> +{
> +	struct aspeed_pwm_data *priv = platform_get_drvdata(dev);
> +
> +	pwmchip_remove(&priv->chip);
> +	reset_control_assert(priv->reset);
> +	clk_disable_unprepare(priv->clk);
> +
> +	return 0;
> +}
> +
> +static const struct of_device_id of_pwm_match_table[] = {
> +	{
> +		.compatible = "aspeed,ast2600-pwm",
> +	},
> +	{},
> +};
> +MODULE_DEVICE_TABLE(of, of_pwm_match_table);
> +
> +static struct platform_driver aspeed_pwm_driver = {
> +	.probe		= aspeed_pwm_probe,
> +	.remove		= aspeed_pwm_remove,
> +	.driver		= {
> +		.name	= "aspeed_pwm",
> +		.of_match_table = of_pwm_match_table,
> +	},
> +};

Please use alignment of = consistently. Either don't do it at all (my
favourite) or do it for all structs in the same way.

Best regards
Uwe