Re: [PATCH v3 2/3] pwm: Add Allwinner's D1/T113-S3/R329 SoCs PWM support

["Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 17262 bytes --]

Hello,

On Fri, Aug 04, 2023 at 10:19:26AM +0300, Александр Шубин wrote:
> пн, 10 июл. 2023 г. в 12:14, Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> > On Tue, Jun 27, 2023 at 11:23:25AM +0300, Aleksandr Shubin wrote:
> > > Allwinner's D1, T113-S3 and R329 SoCs have a quite different PWM
> > > controllers with ones supported by pwm-sun4i driver.
> > >
> > > This patch adds a PWM controller driver for Allwinner's D1,
> > > T113-S3 and R329 SoCs. The main difference between these SoCs
> > > is the number of channels defined by the DT property.
> > >
> > > Signed-off-by: Aleksandr Shubin <privatesub2@gmail.com>
> > > ---
> > >  drivers/pwm/Kconfig      |  10 ++
> > >  drivers/pwm/Makefile     |   1 +
> > >  drivers/pwm/pwm-sun20i.c | 322 +++++++++++++++++++++++++++++++++++++++
> > >  3 files changed, 333 insertions(+)
> > >  create mode 100644 drivers/pwm/pwm-sun20i.c
> > >
> > > diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig
> > > index 8df861b1f4a3..05c48a36969e 100644
> > > --- a/drivers/pwm/Kconfig
> > > +++ b/drivers/pwm/Kconfig
> > > @@ -594,6 +594,16 @@ config PWM_SUN4I
> > >         To compile this driver as a module, choose M here: the module
> > >         will be called pwm-sun4i.
> > >
> > > +config PWM_SUN20I
> > > +     tristate "Allwinner D1/T113s/R329 PWM support"
> > > +     depends on ARCH_SUNXI || COMPILE_TEST
> > > +     depends on COMMON_CLK
> > > +     help
> > > +       Generic PWM framework driver for Allwinner D1/T113s/R329 SoCs.
> > > +
> > > +       To compile this driver as a module, choose M here: the module
> > > +       will be called pwm-sun20i.
> > > +
> > >  config PWM_SUNPLUS
> > >       tristate "Sunplus PWM support"
> > >       depends on ARCH_SUNPLUS || COMPILE_TEST
> > > diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile
> > > index 19899b912e00..cea872e22c78 100644
> > > --- a/drivers/pwm/Makefile
> > > +++ b/drivers/pwm/Makefile
> > > @@ -55,6 +55,7 @@ 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-$(CONFIG_PWM_SUN20I)     += pwm-sun20i.o
> > >  obj-$(CONFIG_PWM_SUNPLUS)    += pwm-sunplus.o
> > >  obj-$(CONFIG_PWM_TEGRA)              += pwm-tegra.o
> > >  obj-$(CONFIG_PWM_TIECAP)     += pwm-tiecap.o
> > > diff --git a/drivers/pwm/pwm-sun20i.c b/drivers/pwm/pwm-sun20i.c
> > > new file mode 100644
> > > index 000000000000..63e9c64e0e18
> > > --- /dev/null
> > > +++ b/drivers/pwm/pwm-sun20i.c
> > > @@ -0,0 +1,322 @@
> > > +// SPDX-License-Identifier: GPL-2.0
> > > +/*
> > > + * PWM Controller Driver for sunxi platforms (D1, T113-S3 and R329)
> > > + *
> > > + * Limitations:
> > > + * - When the parameters change, current running period will not be completed
> > > + *   and run new settings immediately.
> > > + * - It output HIGH-Z state when PWM channel disabled.
> > > + *
> > > + * Copyright (c) 2023 Aleksandr Shubin <privatesub2@gmail.com>
> > > + */
> > > +
> > > +#include <linux/bitfield.h>
> > > +#include <linux/clk.h>
> > > +#include <linux/err.h>
> > > +#include <linux/io.h>
> > > +#include <linux/module.h>
> > > +#include <linux/of_device.h>
> > > +#include <linux/pwm.h>
> > > +#include <linux/reset.h>
> > > +
> > > +#define PWM_CLK_CFG_REG(chan)                (0x20 + (((chan) >> 1) * 0x4))
> > > +#define PWM_CLK_SRC                  GENMASK(8, 7)
> > > +#define PWM_CLK_DIV_M                        GENMASK(3, 0)
> > > +
> > > +#define PWM_CLK_GATE_REG             0x40
> > > +#define PWM_CLK_BYPASS(chan)         BIT((chan) - 16)
> > > +#define PWM_CLK_GATING(chan)         BIT(chan)
> > > +
> > > +#define PWM_ENABLE_REG                       0x80
> > > +#define PWM_EN(chan)                 BIT(chan)
> > > +
> > > +#define PWM_CTL_REG(chan)            (0x100 + (chan) * 0x20)
> > > +#define PWM_ACT_STA                  BIT(8)
> > > +#define PWM_PRESCAL_K                        GENMASK(7, 0)
> > > +
> > > +#define PWM_PERIOD_REG(chan)         (0x104 + (chan) * 0x20)
> > > +#define PWM_ENTIRE_CYCLE             GENMASK(31, 16)
> > > +#define PWM_ACT_CYCLE                        GENMASK(15, 0)
> >
> > Can you please adapt the register field names to include the register
> > name? I'd use:
> >
> > #define PWM_CTL(chan)           (0x100 + (chan) * 0x20)
> > #define PWM_CTL_ACT_STA                 BIT(8)
> > #define PWM_CTL_PRESCAL_K               GENMASK(7, 0)
> >
> > then you get a chance to spot when PWM_CLK_BYPASS(x) is written to
> > PWM_CLK_CFG.
> >
> >
> > > +struct sun20i_pwm_chip {
> > > +     struct pwm_chip chip;
> > > +     struct clk *clk_bus, *clk_hosc;
> > > +     struct reset_control *rst;
> > > +     void __iomem *base;
> > > +     /* Mutex to protect pwm apply state */
> > > +     struct mutex mutex;
> > > +};
> > > +
> > > +static inline struct sun20i_pwm_chip *to_sun20i_pwm_chip(struct pwm_chip *chip)
> > > +{
> > > +     return container_of(chip, struct sun20i_pwm_chip, chip);
> > > +}
> > > +
> > > +static inline u32 sun20i_pwm_readl(struct sun20i_pwm_chip *chip,
> > > +                                unsigned long offset)
> > > +{
> > > +     return readl(chip->base + offset);
> > > +}
> > > +
> > > +static inline void sun20i_pwm_writel(struct sun20i_pwm_chip *chip,
> > > +                                  u32 val, unsigned long offset)
> > > +{
> > > +     writel(val, chip->base + offset);
> > > +}
> > > +
> > > +static int sun20i_pwm_get_state(struct pwm_chip *chip,
> > > +                             struct pwm_device *pwm,
> > > +                             struct pwm_state *state)
> > > +{
> > > +     struct sun20i_pwm_chip *sun20i_chip = to_sun20i_pwm_chip(chip);
> > > +     u64 clk_rate, tmp;
> > > +     u32 val;
> > > +     u16 ent_cycle, act_cycle;
> > > +     u8 prescal, div_id;
> > > +
> > > +     mutex_lock(&sun20i_chip->mutex);
> > > +
> > > +     val = sun20i_pwm_readl(sun20i_chip, PWM_CLK_CFG_REG(pwm->hwpwm));
> > > +     div_id = FIELD_GET(PWM_CLK_DIV_M, val);
> > > +     if (FIELD_GET(PWM_CLK_SRC, val) == 0)
> > > +             clk_rate = clk_get_rate(sun20i_chip->clk_hosc);
> > > +     else
> > > +             clk_rate = clk_get_rate(sun20i_chip->clk_bus);
> > > +
> > > +     val = sun20i_pwm_readl(sun20i_chip, PWM_CTL_REG(pwm->hwpwm));
> > > +     state->polarity = (PWM_ACT_STA & val) ? PWM_POLARITY_NORMAL : PWM_POLARITY_INVERSED;
> > > +
> > > +     prescal = FIELD_GET(PWM_PRESCAL_K, val) + 1;
> >
> > If PWM_PRESCAL_K is 0xff, prescal ends up being 0. This isn't right, is
> > it?
> >
> > > +     val = sun20i_pwm_readl(sun20i_chip, PWM_ENABLE_REG);
> > > +     state->enabled = (PWM_EN(pwm->hwpwm) & val) ? true : false;
> > > +
> > > +     val = sun20i_pwm_readl(sun20i_chip, PWM_PERIOD_REG(pwm->hwpwm));
> > > +     act_cycle = FIELD_GET(PWM_ACT_CYCLE, val);
> > > +     ent_cycle = FIELD_GET(PWM_ENTIRE_CYCLE, val);
> > > +     if (act_cycle > ent_cycle)
> > > +             act_cycle = ent_cycle;
> > > +
> >
> > A comment that with the width of the used factors this cannot overflow
> > would be nice here.
> >
> > > +     tmp = (u64)(act_cycle) * prescal * (1U << div_id) * NSEC_PER_SEC;
> >
> > Can be simplified to:
> >
> >         tmp = (u64)act_cycle * prescal << div_id * NSEC_PER_SEC;
> >
> > > +     state->duty_cycle = DIV_ROUND_UP_ULL(tmp, clk_rate);
> > > +     tmp = (u64)(ent_cycle) * prescal * (1U << div_id) * NSEC_PER_SEC;
> > > +     state->period = DIV_ROUND_UP_ULL(tmp, clk_rate);
> > > +     mutex_unlock(&sun20i_chip->mutex);
> > > +
> > > +     return 0;
> > > +}
> > > +
> > > +static int sun20i_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
> > > +                         const struct pwm_state *state)
> > > +{
> > > +     int ret = 0;
> > > +     u32 clk_gate, clk_cfg, pwm_en, ctl, period;
> > > +     u64 bus_rate, hosc_rate, clk_div, val;
> > > +     u16 prescaler, div_m;
> > > +     bool use_bus_clk, calc_div_m;
> > > +     struct sun20i_pwm_chip *sun20i_chip = to_sun20i_pwm_chip(chip);
> > > +
> > > +     mutex_lock(&sun20i_chip->mutex);
> > > +
> > > +     pwm_en = sun20i_pwm_readl(sun20i_chip, PWM_ENABLE_REG);
> > > +
> > > +     if (state->enabled != pwm->state.enabled)
> > > +             clk_gate = sun20i_pwm_readl(sun20i_chip, PWM_CLK_GATE_REG);
> > > +
> > > +     if (state->enabled != pwm->state.enabled && !state->enabled) {
> > > +             clk_gate &= ~PWM_CLK_GATING(pwm->hwpwm);
> > > +             pwm_en &= ~PWM_EN(pwm->hwpwm);
> > > +             sun20i_pwm_writel(sun20i_chip, pwm_en, PWM_ENABLE_REG);
> > > +             sun20i_pwm_writel(sun20i_chip, clk_gate, PWM_CLK_GATE_REG);
> > > +     }
> > > +
> > > +     if (state->polarity != pwm->state.polarity ||
> > > +         state->duty_cycle != pwm->state.duty_cycle ||
> > > +         state->period != pwm->state.period) {
> > > +             ctl = sun20i_pwm_readl(sun20i_chip, PWM_CTL_REG(pwm->hwpwm));
> > > +             clk_cfg = sun20i_pwm_readl(sun20i_chip, PWM_CLK_CFG_REG(pwm->hwpwm));
> > > +             hosc_rate = clk_get_rate(sun20i_chip->clk_hosc);
> > > +             bus_rate = clk_get_rate(sun20i_chip->clk_bus);
> > > +             if (pwm_en & PWM_EN(pwm->hwpwm ^ 1)) {
> > > +                     /* if the neighbor channel is enable, check period only */
> > > +                     use_bus_clk = FIELD_GET(PWM_CLK_SRC, clk_cfg) != 0;
> > > +                     if (use_bus_clk)
> > > +                             val = state->period * bus_rate;
> > > +                     else
> > > +                             val = state->period * hosc_rate;
> > > +                     do_div(val, NSEC_PER_SEC);
> > > +
> > > +                     div_m = FIELD_GET(PWM_CLK_DIV_M, clk_cfg);
> > > +                     calc_div_m = false;
> > > +             } else {
> > > +                     /* check period and select clock source */
> > > +                     use_bus_clk = false;
> > > +                     val = state->period * hosc_rate;
> > > +                     do_div(val, NSEC_PER_SEC);
> > > +                     if (val <= 1) {
> > > +                             use_bus_clk = true;
> > > +                             val = state->period * bus_rate;
> > > +                             do_div(val, NSEC_PER_SEC);
> > > +                             if (val <= 1) {
> > > +                                     ret = -EINVAL;
> > > +                                     goto unlock_mutex;
> > > +                             }
> > > +                     }
> > > +                     div_m = 0;
> > > +                     calc_div_m = true;
> > > +
> > > +                     /* set up the CLK_DIV_M and clock CLK_SRC */
> > > +                     clk_cfg = FIELD_PREP(PWM_CLK_DIV_M, div_m);
> > > +                     clk_cfg |= FIELD_PREP(PWM_CLK_SRC, use_bus_clk ? 1 : 0);
> > > +
> > > +                     sun20i_pwm_writel(sun20i_chip, clk_cfg, PWM_CLK_CFG_REG(pwm->hwpwm));
> > > +             }
> > > +
> > > +             /* calculate prescaler, M factor, PWM entire cycle */
> > > +             clk_div = val;
> >
> > This assignment is useless as it is overwritten in the loop below, isn't
> > it?
> >
> > > +             for (prescaler = 0;; prescaler++) {
> > > +                     if (prescaler >= 256) {
> > > +                             if (calc_div_m) {
> > > +                                     prescaler = 0;
> > > +                                     div_m++;
> > > +                                     if (div_m >= 9) {
> > > +                                             ret = -EINVAL;
> > > +                                             goto unlock_mutex;
> > > +                                     }
> > > +                             } else {
> > > +                                     ret = -EINVAL;
> > > +                                     goto unlock_mutex;
> > > +                             }
> > > +                     }
> > > +
> > > +                     clk_div = val >> div_m;
> > > +                     do_div(clk_div, prescaler + 1);
> > > +                     if (clk_div <= 65534)
> > > +                             break;
> >
> > This can be calculated without a loop.
> 
> Point me please where can I see the calculation of two divisors without a loop?

for a given value of div_m you search the smallest prescaler such that

	(val >> div_m) // (prescaler + 1) ≤ 65534

(Using Python syntax where // denotes the usual
round-down-to-next-integer division and / normal exact division.)

This is equivalent to:

	  (val >> div_m) // (prescaler + 1) ≤ 65534
	⟺ (val >> div_m) ≤ 65534 * (prescaler + 1) + prescaler
	⟺ (val >> div_m) ≤ 65535 * prescaler + 65534
⟺ (val >> div_m) - 65534 ≤ 65535 * prescaler
	⟺ ((val >> div_m) - 65534) / 65535 ≤ prescaler

And as prescaler is integer, this is

	  ...
	⟺ ((val >> div_m) - 65534) // 65535 ≤ prescaler

So the prescaler value you're looking for is:

	 ((val >> div_m) - 65534) // 65535

And then you have to pick the smallest div_m such that prescaler ≤ 255:

	  ((val >> div_m) - 65534) // 65535 ≤ 255
	⟺ (val >> div_m) - 65534 ≤ 255 * 65535 + 65534
	⟺ val >> div_m ≤ 255 * 65535 + 2 * 65534
	⟺ val >> div_m ≤ 16842493
	⟺ val >> div_m < 16842494

so div_m is fls((val) / 16842494).

You might want to double check this and explain the algorithm in a
comment similar to the above calculation.

> > > +             }
> > > +
> > > +             period = FIELD_PREP(PWM_ENTIRE_CYCLE, clk_div);
> > > +
> > > +             /* set duty cycle */
> > > +             if (use_bus_clk)
> > > +                     val = state->duty_cycle * bus_rate;
> > > +             else
> > > +                     val = state->duty_cycle * hosc_rate;

maybe better use:

	if (use_bus_clk)
		rate = bus_rate;
	else
		rate = hosc_rate;

	val = state->duty_cycle * rate

> > > +             do_div(val, NSEC_PER_SEC);
> > > +             clk_div = val >> div_m;
> > > +             do_div(clk_div, prescaler + 1);
> > > +
> > > +             if (state->duty_cycle == state->period)
> > > +                     clk_div++;
> >
> > I don't understand that one. Can you explain that in a comment please?
> 
> The formula of the output period and the duty-cycle for PWM are as follows.
> T period = (PWM01_CLK / PWM0_PRESCALE_K)^-1 * (PPR0.PWM_ENTIRE_CYCLE + 1)

That can be simpler written as:

	Tperiod = PWM0_PRESCALE_K / PWM01_CLK * (PPR0.PWM_ENTIRE_CYCLE + 1)

right?

> T high-level = (PWM01_CLK / PWM0_PRESCALE_K)^-1 * PPR0.PWM_ACT_CYCLE

simlar:

	Thigh = PWM0_PRESCALE_K / PWM01_CLK * PPR0.PWM_ACT_CYCLE

> Duty-cycle = T high-level / T period
> In accordance with this formula, in order to set the duty-cycle to 100%,
> it is necessary that PWM_ACT_CYCLE >= PWM_ENTIRE_CYCLE + 1

The +1 is also relevant for duty_cycles other than state->period, right?

> > > +             period |= FIELD_PREP(PWM_ACT_CYCLE, clk_div);
> > > +             sun20i_pwm_writel(sun20i_chip, period, PWM_PERIOD_REG(pwm->hwpwm));
> > > +
> > > +             ctl = FIELD_PREP(PWM_PRESCAL_K, prescaler);
> > > +             if (state->polarity == PWM_POLARITY_NORMAL)
> > > +                     ctl |= PWM_ACT_STA;
> > > +
> > > +             sun20i_pwm_writel(sun20i_chip, ctl, PWM_CTL_REG(pwm->hwpwm));
> >
> > Is this racy? I.e. does the write to PWM_PERIOD_REG(pwm->hwpwm) above
> > already has an effect before PWM_CTL_REG(pwm->hwpwm) is written?
> >
> > > +     }
> > > +
> > > +     if (state->enabled != pwm->state.enabled && state->enabled) {
> > > +             clk_gate &= ~PWM_CLK_BYPASS(pwm->hwpwm);
> > > +             clk_gate |= PWM_CLK_GATING(pwm->hwpwm);
> > > +             pwm_en |= PWM_EN(pwm->hwpwm);
> > > +             sun20i_pwm_writel(sun20i_chip, pwm_en, PWM_ENABLE_REG);
> > > +             sun20i_pwm_writel(sun20i_chip, clk_gate, PWM_CLK_GATE_REG);
> >
> > This is (I guess) racy. If your PWM is running with
> >
> >         .period = 10000
> >         .duty_cyle = 0
> >         .enabled = true
> >
> > and you configure it to
> >
> >         .period = 10000
> >         .duty_cyle = 10000
> >         .enabled = false
> >
> > you get a short spike. For a enabled=true -> enabled=false transition
> > you should disable first before configuring duty+period (or skip the
> > latter completely).
> 
> When switching enabled=true -> enabled=false then before setting the period,
> the enable register will be written false and the pvm will become inactive.
> this is the place:
> if (state->enabled != pwm->state.enabled && !state->enabled) {

Ah, it seems I missed that (or I just don't understand any more what
meant back then :-)

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]