[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAOX-t5641GnkVBa2JArL4b4jRDBcu9twMpZbj+0Y71x5mcvJbw@mail.gmail.com>
Date: Fri, 18 Mar 2022 17:18:46 +0800
From: hammer hsieh <hammerh0314@...il.com>
To: Uwe Kleine-König
<u.kleine-koenig@...gutronix.de>
Cc: thierry.reding@...il.com, lee.jones@...aro.org, robh+dt@...nel.org,
linux-pwm@...r.kernel.org, devicetree@...r.kernel.org,
linux-kernel@...r.kernel.org, wells.lu@...plus.com,
"hammer.hsieh" <hammer.hsieh@...plus.com>
Subject: Re: [PATCH v3 2/2] pwm: sunplus-pwm: Add Sunplus SoC SP7021 PWM Driver
Uwe Kleine-König <u.kleine-koenig@...gutronix.de> 於 2022年3月17日 週四 下午6:33寫道:
>
> On Mon, Mar 14, 2022 at 01:51:37PM +0800, Hammer Hsieh wrote:
> > Add Sunplus SoC SP7021 PWM Driver
> >
> > Signed-off-by: Hammer Hsieh <hammerh0314@...il.com>
> > ---
> > Changes in v3:
> > - Addressed all comments from Uwe Kleine-König.
> >
> > MAINTAINERS | 1 +
> > drivers/pwm/Kconfig | 11 +++
> > drivers/pwm/Makefile | 1 +
> > drivers/pwm/pwm-sunplus.c | 232 ++++++++++++++++++++++++++++++++++++++++++++++
> > 4 files changed, 245 insertions(+)
> > create mode 100644 drivers/pwm/pwm-sunplus.c
> >
> > diff --git a/MAINTAINERS b/MAINTAINERS
> > index e1cb7eb..6644bae 100644
> > --- a/MAINTAINERS
> > +++ b/MAINTAINERS
> > @@ -18535,6 +18535,7 @@ SUNPLUS PWM DRIVER
> > M: Hammer Hsieh <hammerh0314@...il.com>
> > S: Maintained
> > F: Documentation/devicetree/bindings/pwm/sunplus,sp7021-pwm.yaml
> > +F: drivers/pwm/pwm-sunplus.c
> >
> > SUNPLUS RTC DRIVER
> > M: Vincent Shih <vincent.sunplus@...il.com>
> > diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig
> > index 21e3b05..54cfb50 100644
> > --- a/drivers/pwm/Kconfig
> > +++ b/drivers/pwm/Kconfig
> > @@ -572,6 +572,17 @@ config PWM_SUN4I
> > To compile this driver as a module, choose M here: the module
> > will be called pwm-sun4i.
> >
> > +config PWM_SUNPLUS
> > + tristate "Sunplus PWM support"
> > + depends on ARCH_SUNPLUS || COMPILE_TEST
> > + depends on HAS_IOMEM && OF
> > + help
> > + Generic PWM framework driver for the PWM controller on
> > + Sunplus SoCs.
> > +
> > + To compile this driver as a module, choose M here: the module
> > + will be called pwm-sunplus.
> > +
> > config PWM_TEGRA
> > tristate "NVIDIA Tegra PWM support"
> > depends on ARCH_TEGRA || COMPILE_TEST
> > diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile
> > index 708840b..be58616 100644
> > --- a/drivers/pwm/Makefile
> > +++ b/drivers/pwm/Makefile
> > @@ -53,6 +53,7 @@ obj-$(CONFIG_PWM_STM32) += pwm-stm32.o
> > obj-$(CONFIG_PWM_STM32_LP) += pwm-stm32-lp.o
> > obj-$(CONFIG_PWM_STMPE) += pwm-stmpe.o
> > obj-$(CONFIG_PWM_SUN4I) += pwm-sun4i.o
> > +obj-$(CONFIG_PWM_SUNPLUS) += pwm-sunplus.o
> > obj-$(CONFIG_PWM_TEGRA) += pwm-tegra.o
> > obj-$(CONFIG_PWM_TIECAP) += pwm-tiecap.o
> > obj-$(CONFIG_PWM_TIEHRPWM) += pwm-tiehrpwm.o
> > diff --git a/drivers/pwm/pwm-sunplus.c b/drivers/pwm/pwm-sunplus.c
> > new file mode 100644
> > index 0000000..b6ab077
> > --- /dev/null
> > +++ b/drivers/pwm/pwm-sunplus.c
> > @@ -0,0 +1,232 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * PWM device driver for SUNPLUS SP7021 SoC
> > + *
> > + * Links:
> > + * Reference Manual:
> > + * https://sunplus-tibbo.atlassian.net/wiki/spaces/doc/overview
> > + *
> > + * Reference Manual(PWM module):
> > + * https://sunplus.atlassian.net/wiki/spaces/doc/pages/461144198/12.+Pulse+Width+Modulation+PWM
>
> On that wiki page someone wants to make s/desable/disable/
>
wiki page desable typo fixed.
> > + *
> > + * Limitations:
> > + * - Only supports normal polarity.
> > + * - It output low when PWM channel disabled.
> > + * - When the parameters change, current running period will not be completed
> > + * and run new settings immediately.
> > + * - In .apply() PWM output need to write register FREQ and DUTY. When first write FREQ
> > + * done and not yet write DUTY, it has short timing gap use new FREQ and old DUTY.
>
> good
>
> > + *
> > + * Author: Hammer Hsieh <hammerh0314@...il.com>
> > + */
> > +#include <linux/clk.h>
> > +#include <linux/io.h>
> > +#include <linux/kernel.h>
> > +#include <linux/module.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/pwm.h>
> > +
> > +#define SP7021_PWM_CONTROL0 0x000
> > +#define SP7021_PWM_CONTROL1 0x004
>
> The link above calls these PWM_MODE0 and PWM_MODE1, also the other
> register names don't match.
>
> > +#define SP7021_PWM_FREQ(ch) (0x008 + 4 * (ch))
> > +#define SP7021_PWM_DUTY(ch) (0x018 + 4 * (ch))
> > +#define SP7021_PWM_FREQ_MAX GENMASK(15, 0)
> > +#define SP7021_PWM_DUTY_MAX GENMASK(7, 0)
> > +#define SP7021_PWM_CONTROL_EN(ch) BIT(ch)
>
> I'm a big fan of consistently naming register defines. I'd do something
> like:
>
> #define SP7021_PWM_MODE0 0x000
> #define SP7021_PWM_MODE0_PWMEN(ch) BIT(ch)
> #define SP7021_PWM_MODE0_BYPASS(ch) BIT(8 + (ch))
>
> #define SP7021_PWM_MODE1 0x004
> #define SP7021_PWM_MODE1_CNTx_EN(ch) BIT(ch)
> ...
>
> such that register names match the manual and register fields have the
> register as a prefix. That way its easier spotable when there is a
> mismatch. (e.g. someone tries to set SP7021_PWM_MODE1_CNTx_EN(1) in
> SP7021_PWM_MODE0.)
>
>
> > +#define SP7021_PWM_NUM 4
> > +#define SP7021_PWM_BYPASS_BIT_SHIFT 8
> > +#define SP7021_PWM_DD_SEL_BIT_SHIFT 8
>
> When you use the bit masks and FIELD_PREP you never should need a define
> for a shift.
>
ok, all define will be like below in next patch.
and all u32 control0,control1;
will change to u32 mode0, mode1;
#define SP7021_PWM_MODE0 0x000
#define SP7021_PWM_MODE0_PWMEN(ch) BIT(ch)
#define SP7021_PWM_MODE0_BYPASS(ch) BIT(8 + (ch))
#define SP7021_PWM_MODE1 0x004
#define SP7021_PWM_MODE1_CNTx_EN(ch) BIT(ch)
#define SP7021_PWM_FREQ(ch) (0x008 + 4 * (ch))
#define SP7021_PWM_FREQ_MAX GENMASK(15, 0)
#define SP7021_PWM_DUTY(ch) (0x018 + 4 * (ch))
#define SP7021_PWM_DUTY_DD_SEL(ch) FIELD_PREP(GENMASK(9, 8), ch)
#define SP7021_PWM_DUTY_MAX GENMASK(7, 0)
#define SP7021_PWM_DUTY_MASK SP7021_PWM_DUTY_MAX
#define SP7021_PWM_FREQ_SCALER 256
#define SP7021_PWM_NUM 4
> > +#define SP7021_PWM_FREQ_SCALER 256
> > +
> > +struct sunplus_pwm {
> > + struct pwm_chip chip;
> > + void __iomem *base;
> > + struct clk *clk;
> > +};
> > +
> > +static inline struct sunplus_pwm *to_sunplus_pwm(struct pwm_chip *chip)
> > +{
> > + return container_of(chip, struct sunplus_pwm, chip);
> > +}
> > +
> > +static int sunplus_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
> > + const struct pwm_state *state)
> > +{
> > + struct sunplus_pwm *priv = to_sunplus_pwm(chip);
> > + u32 dd_freq, duty, control0, control1;
> > + u64 max_period, period_ns, duty_ns, clk_rate;
> > +
> > + if (state->polarity != pwm->state.polarity)
> > + return -EINVAL;
> > +
> > + if (!state->enabled) {
> > + /* disable pwm channel output */
> > + control0 = readl(priv->base + SP7021_PWM_CONTROL0);
> > + control0 &= ~SP7021_PWM_CONTROL_EN(pwm->hwpwm);
> > + writel(control0, priv->base + SP7021_PWM_CONTROL0);
> > + /* disable pwm channel clk source */
> > + control1 = readl(priv->base + SP7021_PWM_CONTROL1);
> > + control1 &= ~SP7021_PWM_CONTROL_EN(pwm->hwpwm);
> > + writel(control1, priv->base + SP7021_PWM_CONTROL1);
> > + return 0;
> > + }
> > +
> > + clk_rate = clk_get_rate(priv->clk);
> > + /*
> > + * SP7021_PWM_FREQ_MAX 16 bits, SP7021_PWM_FREQ_SCALER 8 bits
> > + * NSEC_PER_SEC 30 bits, won't overflow.
> > + */
> > + max_period = mul_u64_u64_div_u64(SP7021_PWM_FREQ_MAX, (u64)SP7021_PWM_FREQ_SCALER
> > + * NSEC_PER_SEC, clk_rate);
> > +
> > + period_ns = min(state->period, max_period);
> > + duty_ns = state->duty_cycle;
>
> duty_ns = min(state->duty_cycle, period_ns);
>
ok, but I think I will remove max_period calculation code.
And take your another recomanded code.
> > +
> > + /*
> > + * cal pwm freq and check value under range
> > + * clk_rate 202.5MHz 28 bits, period_ns max 82849185 27 bits, won't overflow.
> > + */
> > + dd_freq = mul_u64_u64_div_u64(clk_rate, period_ns, (u64)SP7021_PWM_FREQ_SCALER
> > + * NSEC_PER_SEC);
> > +
> > + if (dd_freq == 0)
> > + return -EINVAL;
> > +
> > + if (dd_freq > SP7021_PWM_FREQ_MAX)
> > + dd_freq = SP7021_PWM_FREQ_MAX;
>
> This cannot happen after period_ns was limited to max_period, can it?
> I wonder if there is a max_period value that is cheaper to calculate
> (e.g. no division) and still is good enough to ensure that the
> calculation for dd_freq doesn't overflow. The reasoning there includes
> clk_rate = 202.5 MHz. So maybe something like:
>
> clk_rate = clk_get_rate(priv->clk);
>
> /*
> * The following calculations might overflow if clk is bigger
> * than 256 GHz. In practise it's 202.5MHz, so this limitation
> * is only theoretic.
> */
> if (clk_rate > (u64)SP7021_PWM_FREQ_SCALER * NSEC_PER_SEC)
> return -EINVAL;
>
> /*
> * With clk_rate limited above we have dd_freq <= state->period,
> * so this cannot overflow.
> */
> dd_freq = mul_u64_u64_div_u64(clk_rate, state->period,
> (u64)SP7021_PWM_FREQ_SCALER * NSEC_PER_SEC);
>
> if (dd_freq == 0)
> return -EINVAL;
>
> if (dd_freq > SP7021_PWM_FREQ_MAX)
> dd_freq = SP7021_PWM_FREQ_MAX;
>
ok, will modify it as your recomanded code.
>
> > + writel(dd_freq, priv->base + SP7021_PWM_FREQ(pwm->hwpwm));
> > +
> > + /* cal and set pwm duty */
> > + control0 = readl(priv->base + SP7021_PWM_CONTROL0);
> > + control0 |= SP7021_PWM_CONTROL_EN(pwm->hwpwm);
> > + control1 = readl(priv->base + SP7021_PWM_CONTROL1);
> > + control1 |= SP7021_PWM_CONTROL_EN(pwm->hwpwm);
> > + if (duty_ns == period_ns) {
> > + /* PWM channel output = high */
> > + control0 |= SP7021_PWM_CONTROL_EN(pwm->hwpwm + SP7021_PWM_BYPASS_BIT_SHIFT);
> > + duty = SP7021_PWM_DUTY_MAX;
> > + } else {
> > + control0 &= ~SP7021_PWM_CONTROL_EN(pwm->hwpwm + SP7021_PWM_BYPASS_BIT_SHIFT);
> > + /*
> > + * duty_ns <= period_ns 27 bits, SP7021_PWM_FREQ_SCALER 8 bits
> > + * won't overflow.
> > + */
> > + duty = mul_u64_u64_div_u64(duty_ns, (u64)SP7021_PWM_FREQ_SCALER,
> > + period_ns);
>
> Note this might configure a duty cycle that is too small.
> Consider:
>
> clk_rate = 202500000
> period = 3333643
> duty_cycle = 3306391
>
> Then we get dd_freq = 2636 and duty = 253.
>
> With dd_freq = 2636 and duty = 254 the resulting duty_cycle is
>
> 2636 * 1000000000 * 254 / 202500000 = 3306390.12345679
>
> so 254 would be the better value. The problem is that you use period_ns
> in the division which however is a bit of as the real period is a tad
> smaller.
>
> So the right thing to do here is:
>
> duty = duty_ns * clk / (dd_freq * NSEC_PER_SEC)
>
ok, duty calculation have two method
duty = duty_ns *256 / period_ns
duty = duty_ns * clk / (dd_freq * NSEC_PER_SEC)
In this case , it is better with duty = duty_ns * clk / (dd_freq * NSEC_PER_SEC)
will modify it.
> > + duty |= (pwm->hwpwm << SP7021_PWM_DD_SEL_BIT_SHIFT);
> > + }
> > + writel(duty, priv->base + SP7021_PWM_DUTY(pwm->hwpwm));
>
> I don't understand the DDx SEL bitfield in this register. Is it right
> that it is 0 for all 4 PWMs?
>
PWM0 can select DD0 ~DD3, 0x9c007a18 dd_sel[9:8] and pwm0 duty[7:0]
PWM1 can select DD0 ~DD3, 0x9c007a1c dd_sel[9:8] and pwm1 duty[7:0]
PWM2 can select DD0 ~DD3, 0x9c007a20 dd_sel[9:8] and pwm2 duty[7:0]
PWM3 can select DD0 ~DD3, 0x9c007a24 dd_sel[9:8] and pwm3 duty[7:0]
I will design the driver settings as
PWM0 select DD0
PWM1 select DD1
PWM2 select DD2
PWM3 select DD3
PWM DUTY REG contains dd_sel[9:8] and duty[7:0] for each pwm channel.
once duty calculation done, must conbine dd_sel[9:8] = pwm->hwpwm (0
or 1 or 2 or 3)
then write it to PWM DUTY REG.
before
duty |= (pwm->hwpwm << SP7021_PWM_DD_SEL_BIT_SHIFT);
after
#define SP7021_PWM_DUTY_DD_SEL(ch) FIELD_PREP(GENMASK(9, 8), ch)
duty = SP7021_PWM_DUTY_DD_SEL(pwm->hwpwm) | duty;
> > + writel(control1, priv->base + SP7021_PWM_CONTROL1);
> > + writel(control0, priv->base + SP7021_PWM_CONTROL0);
> > +
> > + return 0;
> > +}
> > +
> > +static void sunplus_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm,
> > + struct pwm_state *state)
> > +{
> > + struct sunplus_pwm *priv = to_sunplus_pwm(chip);
> > + u32 control0, freq, duty;
> > + u64 clk_rate;
> > +
> > + control0 = readl(priv->base + SP7021_PWM_CONTROL0);
> > +
> > + if (control0 & BIT(pwm->hwpwm)) {
> > + clk_rate = clk_get_rate(priv->clk);
> > + freq = readl(priv->base + SP7021_PWM_FREQ(pwm->hwpwm));
>
> I'd call this dd_freq to match the variable name in .apply().
>
ok, will modify it.
> > + duty = readl(priv->base + SP7021_PWM_DUTY(pwm->hwpwm));
> > + duty &= ~GENMASK(9, 8);
>
> That looks wrong, The bit field 9:8 is the divisor source select. Also
> please introduce a define for GENMASK(9,8).
>
ok, will modify it.
#define SP7021_PWM_DUTY_MAX GENMASK(7, 0)
#define SP7021_PWM_DUTY_MASK SP7021_PWM_DUTY_MAX
For duty_cycle calculation, must mask dd_sel[9:8] and only get duty[7:0] value.
duty = readl(priv->base + SP7021_PWM_DUTY(pwm->hwpwm));
before
duty &= ~GENMASK(9, 8);
after
duty = FIELD_GET(SP7021_PWM_DUTY_MASK, duty);
> > + /*
> > + * freq 16 bits, SP7021_PWM_FREQ_SCALER 8 bits
> > + * NSEC_PER_SEC 30 bits, won't overflow.
> > + */
> > + state->period = DIV64_U64_ROUND_UP((u64)freq * (u64)SP7021_PWM_FREQ_SCALER
> > + * NSEC_PER_SEC, clk_rate);
> > + /*
> > + * freq 16 bits, duty 8 bits, NSEC_PER_SEC 30 bits, won't overflow.
> > + */
> > + state->duty_cycle = DIV64_U64_ROUND_UP((u64)freq * (u64)duty * NSEC_PER_SEC,
> > + clk_rate);
> > + state->enabled = true;
> > + } else {
> > + state->enabled = false;
> > + }
> > +
> > + state->polarity = PWM_POLARITY_NORMAL;
> > +}
> > +
> > +static const struct pwm_ops sunplus_pwm_ops = {
> > + .apply = sunplus_pwm_apply,
> > + .get_state = sunplus_pwm_get_state,
> > + .owner = THIS_MODULE,
> > +};
> > +
> > +static void sunplus_pwm_clk_release(void *data)
> > +{
> > + struct clk *clk = data;
> > +
> > + clk_disable_unprepare(clk);
> > +}
> > +
> > +static int sunplus_pwm_probe(struct platform_device *pdev)
> > +{
> > + struct device *dev = &pdev->dev;
> > + struct sunplus_pwm *priv;
> > + int ret;
> > +
> > + priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> > + if (!priv)
> > + return -ENOMEM;
> > +
> > + priv->base = devm_platform_ioremap_resource(pdev, 0);
> > + if (IS_ERR(priv->base))
> > + return PTR_ERR(priv->base);
> > +
> > + priv->clk = devm_clk_get(dev, NULL);
> > + if (IS_ERR(priv->clk))
> > + return dev_err_probe(dev, PTR_ERR(priv->clk),
> > + "get pwm clock failed\n");
> > +
> > + ret = clk_prepare_enable(priv->clk);
> > + if (ret < 0) {
> > + dev_err(dev, "failed to enable clock: %d\n", ret);
> > + return ret;
> > + }
> > +
> > + ret = devm_add_action_or_reset(dev, sunplus_pwm_clk_release, priv->clk);
> > + if (ret < 0) {
> > + dev_err(dev, "failed to release clock: %d\n", ret);
> > + return ret;
> > + }
> > +
> > + priv->chip.dev = dev;
> > + priv->chip.ops = &sunplus_pwm_ops;
> > + priv->chip.npwm = SP7021_PWM_NUM;
> > +
> > + ret = devm_pwmchip_add(dev, &priv->chip);
> > + if (ret < 0)
> > + return dev_err_probe(dev, ret, "Cannot register sunplus PWM\n");
> > +
> > + return 0;
> > +}
> > +
> > +static const struct of_device_id sunplus_pwm_of_match[] = {
> > + { .compatible = "sunplus,sp7021-pwm", },
> > + {}
> > +};
> > +MODULE_DEVICE_TABLE(of, sunplus_pwm_of_match);
> > +
> > +static struct platform_driver sunplus_pwm_driver = {
> > + .probe = sunplus_pwm_probe,
> > + .driver = {
> > + .name = "sunplus-pwm",
> > + .of_match_table = sunplus_pwm_of_match,
> > + },
> > +};
> > +module_platform_driver(sunplus_pwm_driver);
> > +
> > +MODULE_DESCRIPTION("Sunplus SoC PWM Driver");
> > +MODULE_AUTHOR("Hammer Hsieh <hammerh0314@...il.com>");
> > +MODULE_LICENSE("GPL");
>
> Best regards
> Uwe
>
> --
> Pengutronix e.K. | Uwe Kleine-König |
> Industrial Linux Solutions | https://www.pengutronix.de/ |
Powered by blists - more mailing lists