RE: [PATCH v2 1/2] pwm: Add PWM driver for Intel Keem Bay

["G Jaya Kumaran, Vineetha" <vineetha.g.jaya.kumaran@intel.com>]

> -----Original Message-----
> From: linux-pwm-owner@vger.kernel.org <linux-pwm-
> owner@vger.kernel.org> On Behalf Of G Jaya Kumaran, Vineetha
> Sent: Monday, July 27, 2020 1:12 PM
> To: Uwe Kleine Konig <u.kleine-koenig@pengutronix.de>
> Cc: thierry.reding@gmail.com; robh+dt@kernel.org; linux-
> pwm@vger.kernel.org; devicetree@vger.kernel.org; Wan Mohamad, Wan
> Ahmad Zainie <wan.ahmad.zainie.wan.mohamad@intel.com>; Shevchenko,
> Andriy <andriy.shevchenko@intel.com>
> Subject: RE: [PATCH v2 1/2] pwm: Add PWM driver for Intel Keem Bay
> 
> Hi Uwe,
> 
> Thank you for the feedback, will work on the changes for V3.
> 
> > -----Original Message-----
> > From: linux-pwm-owner@vger.kernel.org <linux-pwm-
> > owner@vger.kernel.org> On Behalf Of Uwe Kleine-König
> > Sent: Tuesday, July 21, 2020 5:14 PM
> > To: G Jaya Kumaran, Vineetha <vineetha.g.jaya.kumaran@intel.com>
> > Cc: thierry.reding@gmail.com; robh+dt@kernel.org; linux-
> > pwm@vger.kernel.org; devicetree@vger.kernel.org; Wan Mohamad, Wan
> > Ahmad Zainie <wan.ahmad.zainie.wan.mohamad@intel.com>;
> Shevchenko,
> > Andriy <andriy.shevchenko@intel.com>
> > Subject: Re: [PATCH v2 1/2] pwm: Add PWM driver for Intel Keem Bay
> >
> > Hello,
> >
> > On Sat, Jul 18, 2020 at 10:47:07PM +0800,
> > vineetha.g.jaya.kumaran@intel.com wrote:
> > > From: "Lai, Poey Seng" <poey.seng.lai@intel.com>
> > >
> > > Enable PWM support for the Intel Keem Bay SoC.
> > >
> > > Signed-off-by: Lai, Poey Seng <poey.seng.lai@intel.com>
> > > Signed-off-by: Vineetha G. Jaya Kumaran
> > > <vineetha.g.jaya.kumaran@intel.com>
> > > ---
> > >  drivers/pwm/Kconfig       |   9 ++
> > >  drivers/pwm/Makefile      |   1 +
> > >  drivers/pwm/pwm-keembay.c | 239
> > > ++++++++++++++++++++++++++++++++++++++++++++++
> > >  3 files changed, 249 insertions(+)
> > >  create mode 100644 drivers/pwm/pwm-keembay.c
> > >
> > > diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig index
> > > cb8d739..2b0419b 100644
> > > --- a/drivers/pwm/Kconfig
> > > +++ b/drivers/pwm/Kconfig
> > > @@ -569,4 +569,13 @@ config PWM_ZX
> > >  	  To compile this driver as a module, choose M here: the module
> > >  	  will be called pwm-zx.
> > >
> > > +config PWM_KEEMBAY
> > > +	tristate "Intel Keem Bay PWM driver"
> > > +	depends on ARM64 || COMPILE_TEST
> > > +	help
> > > +	  The platform driver for Intel Keem Bay PWM controller.
> > > +
> > > +	  To compile this driver as a module, choose M here: the module
> > > +	  will be called pwm-keembay.
> > > +
> > >  endif
> > > diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile index
> > > a59c710..0c84ff2 100644
> > > --- a/drivers/pwm/Makefile
> > > +++ b/drivers/pwm/Makefile
> > > @@ -55,3 +55,4 @@ obj-$(CONFIG_PWM_TWL)		+= pwm-twl.o
> > >  obj-$(CONFIG_PWM_TWL_LED)	+= pwm-twl-led.o
> > >  obj-$(CONFIG_PWM_VT8500)	+= pwm-vt8500.o
> > >  obj-$(CONFIG_PWM_ZX)		+= pwm-zx.o
> > > +obj-$(CONFIG_PWM_KEEMBAY)	+= pwm-keembay.o
> > > diff --git a/drivers/pwm/pwm-keembay.c b/drivers/pwm/pwm-
> keembay.c
> > new
> > > file mode 100644 index 00000000..fa5fe95
> > > --- /dev/null
> > > +++ b/drivers/pwm/pwm-keembay.c
> > > @@ -0,0 +1,239 @@
> > > +// SPDX-License-Identifier: GPL-2.0
> > > +/*
> > > + * Intel Keem Bay PWM driver
> > > + *
> > > + * Copyright (C) 2020 Intel Corporation
> > > + * Authors: Lai Poey Seng <poey.seng.lai@intel.com>
> > > + *          Vineetha G. Jaya Kumaran
> <vineetha.g.jaya.kumaran@intel.com>
> > > + */
> >
> > If possible, please add a link here to documentation for this chip.
> >
> 
> There is no documentation released publicly (yet), but once there is, I will
> add it in.
> 
> > > +
> > > +#include <linux/bitfield.h>
> > > +#include <linux/clk.h>
> > > +#include <linux/io.h>
> > > +#include <linux/module.h>
> > > +#include <linux/of.h>
> > > +#include <linux/platform_device.h>
> > > +#include <linux/pwm.h>
> > > +#include <linux/regmap.h>
> > > +
> > > +#define KMB_TOTAL_PWM_CHANNELS		6
> > > +#define KMB_PWM_COUNT_MAX		65535
> >
> > If you write this as hexadecimal constant it is more obvious.
> >
> > > +#define KMB_PWM_EN_BIT			BIT(31)
> > > +
> > > +/* Mask */
> > > +#define KMB_PWM_HIGH_MASK		GENMASK(31, 16)
> > > +#define KMB_PWM_LOW_MASK		GENMASK(15, 0)
> > > +#define KMB_PWM_COUNT_MASK		GENMASK(31, 0)
> > > +
> > > +/* PWM Register offset */
> > > +#define KMB_PWM_LEADIN0_OFFSET		0x00
> > > +#define KMB_PWM_LEADIN1_OFFSET		0x04
> > > +#define KMB_PWM_LEADIN2_OFFSET		0x08
> > > +#define KMB_PWM_LEADIN3_OFFSET		0x0c
> > > +#define KMB_PWM_LEADIN4_OFFSET		0x10
> > > +#define KMB_PWM_LEADIN5_OFFSET		0x14
> >
> > All but ..LEADIN0.. are unused. Is this maybe more useful to write as:
> >
> > 	#define KMB_PWM_LEADIN_OFFSET(ch)	(0x00 + 4 * (ch))
> >
> > ?
> >
> 
> Agreed, will change this for both KMB_PWM_HIGHLOW_OFFSET and
> KMB_PWM_LEADIN_OFFSET.
> 
> > > +#define KMB_PWM_HIGHLOW0_OFFSET		0x20
> > > +#define KMB_PWM_HIGHLOW1_OFFSET		0x24
> > > +#define KMB_PWM_HIGHLOW2_OFFSET		0x28
> > > +#define KMB_PWM_HIGHLOW3_OFFSET		0x2c
> > > +#define KMB_PWM_HIGHLOW4_OFFSET		0x30
> > > +#define KMB_PWM_HIGHLOW5_OFFSET		0x34
> >
> > ditto.
> >
> > > +struct keembay_pwm {
> > > +	struct pwm_chip chip;
> > > +	struct device *dev;
> > > +	struct clk *clk;
> > > +	void __iomem *base;
> > > +};
> > > +
> > > +static inline struct keembay_pwm *to_keembay_pwm_dev(struct
> > pwm_chip
> > > +*chip) {
> > > +	return container_of(chip, struct keembay_pwm, chip); }
> > > +
> > > +static inline void keembay_pwm_update_bits(struct keembay_pwm
> > *priv, u32 mask,
> > > +					   u32 val, u32 reg, int ch)
> > > +{
> > > +	u32 buff, offset, tmp;
> > > +	void __iomem *address;
> > > +
> > > +	offset = reg + ch * 4;
> > > +	address = priv->base + offset;
> > > +	buff = readl(address);
> > > +	tmp = buff & ~mask;
> > > +	tmp |= FIELD_PREP(mask, val);
> > > +	writel(tmp, address);
> > > +}
> > > +
> > > +static void keembay_pwm_enable(struct keembay_pwm *priv, int ch) {
> > > +	keembay_pwm_update_bits(priv, KMB_PWM_EN_BIT, 1,
> > > +				KMB_PWM_LEADIN0_OFFSET, ch);
> > > +}
> > > +
> > > +static void keembay_pwm_disable(struct keembay_pwm *priv, int ch) {
> > > +	keembay_pwm_update_bits(priv, KMB_PWM_EN_BIT, 0,
> > > +				KMB_PWM_LEADIN0_OFFSET, ch);
> > > +}
> > > +
> > > +static void keembay_pwm_get_state(struct pwm_chip *chip, struct
> > pwm_device *pwm,
> > > +				  struct pwm_state *state)
> > > +{
> > > +	struct keembay_pwm *priv = to_keembay_pwm_dev(chip);
> > > +	unsigned long long pwm_h_count, pwm_l_count;
> > > +	unsigned long clk_rate;
> > > +	u32 buff;
> > > +
> > > +	clk_rate = clk_get_rate(priv->clk);
> > > +
> > > +	/* Read channel enabled status */
> > > +	buff = readl(priv->base + KMB_PWM_LEADIN0_OFFSET + pwm-
> > >hwpwm * 4);
> > > +	if (buff & KMB_PWM_EN_BIT)
> > > +		state->enabled = true;
> > > +	else
> > > +		state->enabled = false;
> > > +
> > > +	/* Read period and duty cycle*/
> >
> > Missing ' ' before closing */
> >
> > > +	buff = readl(priv->base + KMB_PWM_HIGHLOW0_OFFSET + pwm-
> > >hwpwm * 4);
> > > +	pwm_l_count = (buff & KMB_PWM_LOW_MASK) * NSEC_PER_SEC;
> > > +	pwm_h_count = ((buff & KMB_PWM_HIGH_MASK) >> 16) *
> > NSEC_PER_SEC;
> >
> > pwm_h_count = FIELD_GET(KMB_PWM_HIGH_MASK, buff) *
> NSEC_PER_SEC
> >
> > > +	state->duty_cycle = pwm_h_count / clk_rate;
> > > +	state->period = (pwm_h_count + pwm_l_count) / clk_rate;
> >
> > Please round up both values.
> >
> > > +}
> > > +
> > > +static int keembay_pwm_apply(struct pwm_chip *chip, struct
> > pwm_device *pwm,
> > > +			     const struct pwm_state *state) {
> > > +	struct keembay_pwm *priv = to_keembay_pwm_dev(chip);
> > > +	struct pwm_state current_state;
> > > +	u16 pwm_h_count, pwm_l_count;
> > > +	unsigned long long div;
> > > +	unsigned long clk_rate;
> > > +	u32 pwm_count = 0;
> > > +
> > > +	pwm_get_state(pwm, &current_state);
> >
> > Please check the hardware state, not the value cached in the PWM
> > framework.
> >
> > > +	if (!state->enabled && current_state.enabled) {
> > > +		keembay_pwm_disable(priv, pwm->hwpwm);
> > > +		return 0;
> > > +	}
> > > +
> > > +	if (state->polarity != PWM_POLARITY_NORMAL)
> > > +		return -ENOSYS;
> >
> > This must be checked before .enabled. That's because the expectation
> > on
> >
> > 	.enabled = false
> > 	.polarity = PWM_POLARITY_INVERSED
> >
> > is that the output gets constant high.
> >
> > > +	/*
> > > +	 * The upper 16 bits of the KMB_PWM_HIGHLOWx_OFFSET register
> > contain
> > > +	 * the high time of the waveform, while the last 16 bits contain
> > > +	 * the low time of the waveform, in terms of clock cycles.
> >
> > Just to be sure: Each period starts with the high time, right?
> >
> 
> Yes, each period starts with the high time.
> 
> > > +	 * high time = clock rate * duty cycle / NSEC_PER_SEC
> > > +	 * low time =  clock rate * (period - duty cycle) / NSEC_PER_SEC
> > > +	 *
> > > +	 * e.g. For period 50000ns, duty cycle 30000ns, and clock rate 500MHz:
> > > +	 * high time = (500000000 * 30000) / 1000000000 = 0x3A98
> > > +	 * low time = (500000000 * 20000) / 1000000000 = 0x2710
> > > +	 * Value written to KMB_PWM_HIGHLOWx_OFFSET = 0x3A982710
> >
> > For period = 50000ns, duty_cycle = 30000ns and clock rate 266666666 Hz
> > you have to configure:
> >
> > 	high = 7999
> > 	low = 5334
> >
> > > +	 */
> > > +
> > > +	clk_rate = clk_get_rate(priv->clk);
> > > +
> > > +	/* Configure waveform high time */
> > > +	div = clk_rate * state->duty_cycle;
> > > +	do_div(div, NSEC_PER_SEC);
> >
> > Can this overflow?
> >

Hi Uwe, sorry for missing this in my previous reply.
Yes, it may exceed the 16-bit limit for the high/low time, depending on the selected duty cycle.
There is a range check after, and -ERANGE will be returned if the high/low time range is exceeded.

> > > +	if (div > KMB_PWM_COUNT_MAX)
> > > +		return -ERANGE;
> > > +
> > > +	pwm_h_count = (u16)div;
> >
> > No need for this cast.
> >
> > > +	/* Configure waveform low time */
> > > +	div = clk_rate * (state->period - state->duty_cycle);
> > > +	do_div(div, NSEC_PER_SEC);
> >
> > Here the rounding is wrong. (See above example, currently you use low
> > =
> > 5333 here)
> >

Will fix the rounding to the nearest integer for the high/low time.

> > > +	if (div > KMB_PWM_COUNT_MAX)
> > > +		return -ERANGE;
> > > +
> > > +	pwm_l_count = (u16)div;
> > > +
> > > +	pwm_count |= pwm_h_count << 16;
> > > +	pwm_count |= pwm_l_count;
> >
> > pwm_count = FIELD_PREP(KMB_PWM_HIGH_MASK, pwm_h_count) |
> > FIELD_PREP(KMB_PWM_LOW_MASK, pwm_l_count);
> >
> > > +
> > > +	keembay_pwm_update_bits(priv, KMB_PWM_COUNT_MASK,
> > > +				pwm_count,
> > KMB_PWM_HIGHLOW0_OFFSET, pwm->hwpwm);
> >
> > here the rwm-procedure is not necessary as all 32 bits are written anyhow.
> >
> > > +	if (state->enabled && !current_state.enabled)
> > > +		keembay_pwm_enable(priv, pwm->hwpwm);
> > > +
> > > +	return 0;
> > > +}
> > > [...]
> > > +MODULE_ALIAS("platform:keembay");
> >
> > This has to match your driver's name, so use:
> >
> > 	MODULE_ALIAS("platform:pwm-keembay");
> >
> > > +MODULE_DESCRIPTION("Intel Keem Bay PWM driver");
> > MODULE_LICENSE("GPL
> > > +v2");
> >
> > In v1 you told that on reconfiguration the hardware completes the
> > currently running period. Please document this in the driver, similar
> > to e.g. pwm- sifive.c.
> >
> 
> OK, I will refer to that example and add this information in.
> 
> > You only ever write the enable bit in the leadin register. If there is
> > something != 0 at boot this influences the correct behaviour of the driver,
> right?
> >
> 
> That's right, I will add a check during probe to ensure the bit is set to 0 as
> default.
> 
> > Further things to note there:
> >
> >  - What is the behaviour on disable (usual candidates: freezes at
> >    current value, completes period and emits low, changes to High-Z)?
> >
> 
> For this case, the period does not get completed.
> As soon as the channel is disabled, the output goes low.
> 
> >  - Can it do 0% and 100% duty ratio?
> >
> 
> Yes, we can set duty cycles of 0% and 100%.
> 
> > Best regards
> > Uwe
> >
> > --
> > Pengutronix e.K.                           | Uwe Kleine-König            |
> > Industrial Linux Solutions                 | https://www.pengutronix.de/ |