diff mbox series

[1/3] pwm: cros-ec: Don't care about consumers in .get_state()

Message ID 20240607084416.897777-6-u.kleine-koenig@baylibre.com (mailing list archive)
State Handled Elsewhere
Headers show
Series pwm: cros-ec: Some simplifications | expand

Commit Message

Uwe Kleine-König June 7, 2024, 8:44 a.m. UTC 
The get_state() callback is never called (in a visible way) after there
is a consumer for a pwm device. The core handles loosing the information
about duty_cycle just fine.

Simplify the driver accordingly.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
---
 drivers/pwm/pwm-cros-ec.c | 33 +--------------------------------
 1 file changed, 1 insertion(+), 32 deletions(-)

Comments

Uwe Kleine-König June 7, 2024, 4:43 p.m. UTC | #1 
On Fri, Jun 07, 2024 at 10:44:15AM +0200, Uwe Kleine-König wrote:
> The get_state() callback is never called (in a visible way) after there
> is a consumer for a pwm device. The core handles loosing the information
> about duty_cycle just fine.
> 
> Simplify the driver accordingly.
> 
> Signed-off-by: Uwe Kleine-König <u.kleine-koenig@baylibre.com>
> ---
>  drivers/pwm/pwm-cros-ec.c | 33 +--------------------------------
>  1 file changed, 1 insertion(+), 32 deletions(-)
> 
> diff --git a/drivers/pwm/pwm-cros-ec.c b/drivers/pwm/pwm-cros-ec.c
> index 606ccfdaf4cc..ba4ee0b507b7 100644
> --- a/drivers/pwm/pwm-cros-ec.c
> +++ b/drivers/pwm/pwm-cros-ec.c
> @@ -25,15 +25,6 @@
>  struct cros_ec_pwm_device {
>  	struct cros_ec_device *ec;
>  	bool use_pwm_type;
> -	struct cros_ec_pwm *channel;
> -};

I forgot to drop the kernel doc comment. Unless I get some feedback that
makes me send a v2, I'll squash the following into this patch when
applying:

diff --git a/drivers/pwm/pwm-cros-ec.c b/drivers/pwm/pwm-cros-ec.c
index fcc33a2cb878..189301dc395e 100644
--- a/drivers/pwm/pwm-cros-ec.c
+++ b/drivers/pwm/pwm-cros-ec.c
@@ -20,7 +20,6 @@
  *
  * @ec: Pointer to EC device
  * @use_pwm_type: Use PWM types instead of generic channels
- * @channel: array with per-channel data
  */
 struct cros_ec_pwm_device {
 	struct cros_ec_device *ec;

Best regards
Uwe
kernel test robot June 8, 2024, 2:24 p.m. UTC | #2 
Hi Uwe,

kernel test robot noticed the following build warnings:

[auto build test WARNING on 1613e604df0cd359cf2a7fbd9be7a0bcfacfabd0]

url:    https://github.com/intel-lab-lkp/linux/commits/Uwe-Kleine-K-nig/pwm-cros-ec-Don-t-care-about-consumers-in-get_state/20240607-164747
base:   1613e604df0cd359cf2a7fbd9be7a0bcfacfabd0
patch link:    https://lore.kernel.org/r/20240607084416.897777-6-u.kleine-koenig%40baylibre.com
patch subject: [PATCH 1/3] pwm: cros-ec: Don't care about consumers in .get_state()
config: arm64-defconfig (https://download.01.org/0day-ci/archive/20240608/202406082139.KG4VcZiF-lkp@intel.com/config)
compiler: aarch64-linux-gcc (GCC) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240608/202406082139.KG4VcZiF-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202406082139.KG4VcZiF-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> drivers/pwm/pwm-cros-ec.c:28: warning: Excess struct member 'channel' description in 'cros_ec_pwm_device'


vim +28 drivers/pwm/pwm-cros-ec.c

3d593b6e80ad2c Fabio Baltieri   2022-04-28  17  
1f0d3bb02785f6 Brian Norris     2016-07-15  18  /**
1f0d3bb02785f6 Brian Norris     2016-07-15  19   * struct cros_ec_pwm_device - Driver data for EC PWM
1f0d3bb02785f6 Brian Norris     2016-07-15  20   *
1f0d3bb02785f6 Brian Norris     2016-07-15  21   * @ec: Pointer to EC device
3d593b6e80ad2c Fabio Baltieri   2022-04-28  22   * @use_pwm_type: Use PWM types instead of generic channels
82adc1b2688b02 Uwe Kleine-König 2023-07-05  23   * @channel: array with per-channel data
1f0d3bb02785f6 Brian Norris     2016-07-15  24   */
1f0d3bb02785f6 Brian Norris     2016-07-15  25  struct cros_ec_pwm_device {
1f0d3bb02785f6 Brian Norris     2016-07-15  26  	struct cros_ec_device *ec;
3d593b6e80ad2c Fabio Baltieri   2022-04-28  27  	bool use_pwm_type;
1db37f9561b2b3 Thierry Reding   2019-10-17 @28  };
1db37f9561b2b3 Thierry Reding   2019-10-17  29
Tzung-Bi Shih June 11, 2024, 8:50 a.m. UTC | #3 
On Fri, Jun 07, 2024 at 10:44:15AM +0200, Uwe Kleine-König wrote:
> The get_state() callback is never called (in a visible way) after there
> is a consumer for a pwm device. The core handles loosing the information
> about duty_cycle just fine.

ChromeOS EC has no separated "enabled" state, it sees `duty == 0` as
"disabled"[1].  1db37f9561b2 ("pwm: cros-ec: Cache duty cycle value")
caches the value in kernel side so that it can retrieve the original duty
value even if (struct pwm_state *)->enabled is false.

To make sure I understand, did you mean the original duty value could be less
important because:
- We are less caring as it is in a debug context at [2]?
- At [3], the PWM device is still initializing.

[1]: https://crrev.com/0e16954460a08133b2557150e0897014ea2b9672/common/pwm.c#66
[2]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L52
[3]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L371
Uwe Kleine-König June 11, 2024, 10:39 a.m. UTC | #4 
Hello Tzung,

On Tue, Jun 11, 2024 at 08:50:44AM +0000, Tzung-Bi Shih wrote:
> On Fri, Jun 07, 2024 at 10:44:15AM +0200, Uwe Kleine-König wrote:
> > The get_state() callback is never called (in a visible way) after there
> > is a consumer for a pwm device. The core handles loosing the information
> > about duty_cycle just fine.
> 
> ChromeOS EC has no separated "enabled" state, it sees `duty == 0` as
> "disabled"[1].  1db37f9561b2 ("pwm: cros-ec: Cache duty cycle value")
> caches the value in kernel side so that it can retrieve the original duty
> value even if (struct pwm_state *)->enabled is false.

There is no need to cache, so the following would work:

diff --git a/drivers/pwm/pwm-cros-ec.c b/drivers/pwm/pwm-cros-ec.c
index 606ccfdaf4cc..2b72468767f4 100644
--- a/drivers/pwm/pwm-cros-ec.c
+++ b/drivers/pwm/pwm-cros-ec.c
@@ -25,15 +25,6 @@
 struct cros_ec_pwm_device {
 	struct cros_ec_device *ec;
 	bool use_pwm_type;
-	struct cros_ec_pwm *channel;
-};
-
-/**
- * struct cros_ec_pwm - per-PWM driver data
- * @duty_cycle: cached duty cycle
- */
-struct cros_ec_pwm {
-	u16 duty_cycle;
 };
 
 static inline struct cros_ec_pwm_device *pwm_to_cros_ec_pwm(struct pwm_chip *chip)
@@ -135,37 +126,33 @@ static int cros_ec_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
 			     const struct pwm_state *state)
 {
 	struct cros_ec_pwm_device *ec_pwm = pwm_to_cros_ec_pwm(chip);
-	struct cros_ec_pwm *channel = &ec_pwm->channel[pwm->hwpwm];
 	u16 duty_cycle;
-	int ret;
 
-	/* The EC won't let us change the period */
-	if (state->period != EC_PWM_MAX_DUTY)
-		return -EINVAL;
+	if (state->enabled) {
 
-	if (state->polarity != PWM_POLARITY_NORMAL)
-		return -EINVAL;
+		/* The EC only supports period = EC_PWM_MAX_DUTY */
+		if (state->period < EC_PWM_MAX_DUTY ||
+		    state->polarity != PWM_POLARITY_NORMAL)
+			return -EINVAL;
 
-	/*
-	 * EC doesn't separate the concept of duty cycle and enabled, but
-	 * kernel does. Translate.
-	 */
-	duty_cycle = state->enabled ? state->duty_cycle : 0;
+		duty_cycle = min(state->duty_cycle, (u64)EC_PWM_MAX_DUTY);
 
-	ret = cros_ec_pwm_set_duty(ec_pwm, pwm->hwpwm, duty_cycle);
-	if (ret < 0)
-		return ret;
+	} else {
+		/*
+		 * The hardware has no possibility to disable and so save power.
+		 * Many consumers expect the PWM to at least stop to oscilate, so just
+		 * configure for duty_cycle = 0.
+		 */
+		duty_cycle = 0;
+	}
 
-	channel->duty_cycle = state->duty_cycle;
-
-	return 0;
+	return cros_ec_pwm_set_duty(ec_pwm, pwm->hwpwm, duty_cycle);
 }
 
 static int cros_ec_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm,
 				 struct pwm_state *state)
 {
 	struct cros_ec_pwm_device *ec_pwm = pwm_to_cros_ec_pwm(chip);
-	struct cros_ec_pwm *channel = &ec_pwm->channel[pwm->hwpwm];
 	int ret;
 
 	ret = cros_ec_pwm_get_duty(ec_pwm->ec, ec_pwm->use_pwm_type, pwm->hwpwm);
@@ -175,23 +162,10 @@ static int cros_ec_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm,
 	}
 
 	state->enabled = (ret > 0);
+	state->duty_cycle = ret;
 	state->period = EC_PWM_MAX_DUTY;
 	state->polarity = PWM_POLARITY_NORMAL;
 
-	/*
-	 * Note that "disabled" and "duty cycle == 0" are treated the same. If
-	 * the cached duty cycle is not zero, used the cached duty cycle. This
-	 * ensures that the configured duty cycle is kept across a disable and
-	 * enable operation and avoids potentially confusing consumers.
-	 *
-	 * For the case of the initial hardware readout, channel->duty_cycle
-	 * will be 0 and the actual duty cycle read from the EC is used.
-	 */
-	if (ret == 0 && channel->duty_cycle > 0)
-		state->duty_cycle = channel->duty_cycle;
-	else
-		state->duty_cycle = ret;
-
 	return 0;
 }
 
@@ -291,11 +265,6 @@ static int cros_ec_pwm_probe(struct platform_device *pdev)
 	chip->ops = &cros_ec_pwm_ops;
 	chip->of_xlate = cros_ec_pwm_xlate;
 
-	ec_pwm->channel = devm_kcalloc(dev, chip->npwm, sizeof(*ec_pwm->channel),
-					GFP_KERNEL);
-	if (!ec_pwm->channel)
-		return -ENOMEM;
-
 	dev_dbg(dev, "Probed %u PWMs\n", chip->npwm);
 
 	ret = devm_pwmchip_add(dev, chip);

> To make sure I understand, did you mean the original duty value could be less
> important because:
> - We are less caring as it is in a debug context at [2]?
> - At [3], the PWM device is still initializing.

It doesn't really matter that this is about debug or initialisation. The
key here is that the core can handle the PWM using duty_cycle 0 (or
anything else) when it was requested to be disabled.

Best regards
Uwe

> [1]: https://crrev.com/0e16954460a08133b2557150e0897014ea2b9672/common/pwm.c#66
> [2]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L52
> [3]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L371
Tzung-Bi Shih June 12, 2024, 6:27 a.m. UTC | #5 
On Tue, Jun 11, 2024 at 12:39:44PM +0200, Uwe Kleine-König wrote:
> On Tue, Jun 11, 2024 at 08:50:44AM +0000, Tzung-Bi Shih wrote:
> > On Fri, Jun 07, 2024 at 10:44:15AM +0200, Uwe Kleine-König wrote:
> > > The get_state() callback is never called (in a visible way) after there
> > > is a consumer for a pwm device. The core handles loosing the information
> > > about duty_cycle just fine.
> > 
> > ChromeOS EC has no separated "enabled" state, it sees `duty == 0` as
> > "disabled"[1].  1db37f9561b2 ("pwm: cros-ec: Cache duty cycle value")
> > caches the value in kernel side so that it can retrieve the original duty
> > value even if (struct pwm_state *)->enabled is false.
> 
> There is no need to cache, so the following would work:

Ack.

> > To make sure I understand, did you mean the original duty value could be less
> > important because:
> > - We are less caring as it is in a debug context at [2]?
> > - At [3], the PWM device is still initializing.
> 
> It doesn't really matter that this is about debug or initialisation. The
> key here is that the core can handle the PWM using duty_cycle 0 (or
> anything else) when it was requested to be disabled.
> 
> 
> > [1]: https://crrev.com/0e16954460a08133b2557150e0897014ea2b9672/common/pwm.c#66
> > [2]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L52
> > [3]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L371

I was trying to understand the description in the commit message:
: The get_state() callback is never called (in a visible way) after there
: is a consumer for a pwm device.

I guess I understood; the core reads the duty value via get_state() when:
- Initializing the device for the intial value.
- Debugging for checking if apply() really takes effect.

What 1db37f9561b2 worried about is already addressed by the core[4].

[4]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L495

Reviewed-by: Tzung-Bi Shih <tzungbi@kernel.org>
Uwe Kleine-König June 13, 2024, 6:07 a.m. UTC | #6 
Hello,

On Wed, Jun 12, 2024 at 06:27:14AM +0000, Tzung-Bi Shih wrote:
> On Tue, Jun 11, 2024 at 12:39:44PM +0200, Uwe Kleine-König wrote:
> > On Tue, Jun 11, 2024 at 08:50:44AM +0000, Tzung-Bi Shih wrote:
> > > On Fri, Jun 07, 2024 at 10:44:15AM +0200, Uwe Kleine-König wrote:
> > > > The get_state() callback is never called (in a visible way) after there
> > > > is a consumer for a pwm device. The core handles loosing the information
> > > > about duty_cycle just fine.
> > > 
> > > ChromeOS EC has no separated "enabled" state, it sees `duty == 0` as
> > > "disabled"[1].  1db37f9561b2 ("pwm: cros-ec: Cache duty cycle value")
> > > caches the value in kernel side so that it can retrieve the original duty
> > > value even if (struct pwm_state *)->enabled is false.
> > 
> > There is no need to cache, so the following would work:
> 
> Ack.
> 
> > > To make sure I understand, did you mean the original duty value could be less
> > > important because:
> > > - We are less caring as it is in a debug context at [2]?
> > > - At [3], the PWM device is still initializing.
> > 
> > It doesn't really matter that this is about debug or initialisation. The
> > key here is that the core can handle the PWM using duty_cycle 0 (or
> > anything else) when it was requested to be disabled.
> > 
> > 
> > > [1]: https://crrev.com/0e16954460a08133b2557150e0897014ea2b9672/common/pwm.c#66
> > > [2]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L52
> > > [3]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L371
> 
> I was trying to understand the description in the commit message:
> : The get_state() callback is never called (in a visible way) after there
> : is a consumer for a pwm device.
> 
> I guess I understood; the core reads the duty value via get_state() when:
> - Initializing the device for the intial value.

Yes, that a consumer cannot do any assumptions about a PWM just aquired,
so there is no danger for confusion.

> - Debugging for checking if apply() really takes effect.

Right, and the read pwm_state is only used in the core. The core won't
be confused about disabled vs. duty=0. Consumers never see the result of
the .get_state callback.

> What 1db37f9561b2 worried about is already addressed by the core[4].
> [4]: https://elixir.bootlin.com/linux/v6.10-rc3/source/drivers/pwm/core.c#L495

What 1db37f9561b2 worried about is bogus, because pwm_get_state()
doesn't call the .get_state() callback:

	$ git show 1db37f9561b2:include/linux/pwm.h | awk '/void pwm_get_state/,/^$/'
	static inline void pwm_get_state(const struct pwm_device *pwm,
					 struct pwm_state *state)
	{
		*state = pwm->state;
	}

> Reviewed-by: Tzung-Bi Shih <tzungbi@kernel.org>

I'll apply the patch under discussion and create a proper patch for the
ad-hoc change from my previous mail.

Thanks
Uwe
diff mbox series

Patch

diff --git a/drivers/pwm/pwm-cros-ec.c b/drivers/pwm/pwm-cros-ec.c
index 606ccfdaf4cc..ba4ee0b507b7 100644
--- a/drivers/pwm/pwm-cros-ec.c
+++ b/drivers/pwm/pwm-cros-ec.c
@@ -25,15 +25,6 @@ 
 struct cros_ec_pwm_device {
 	struct cros_ec_device *ec;
 	bool use_pwm_type;
-	struct cros_ec_pwm *channel;
-};
-
-/**
- * struct cros_ec_pwm - per-PWM driver data
- * @duty_cycle: cached duty cycle
- */
-struct cros_ec_pwm {
-	u16 duty_cycle;
 };
 
 static inline struct cros_ec_pwm_device *pwm_to_cros_ec_pwm(struct pwm_chip *chip)
@@ -135,7 +126,6 @@  static int cros_ec_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
 			     const struct pwm_state *state)
 {
 	struct cros_ec_pwm_device *ec_pwm = pwm_to_cros_ec_pwm(chip);
-	struct cros_ec_pwm *channel = &ec_pwm->channel[pwm->hwpwm];
 	u16 duty_cycle;
 	int ret;
 
@@ -156,8 +146,6 @@  static int cros_ec_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
 	if (ret < 0)
 		return ret;
 
-	channel->duty_cycle = state->duty_cycle;
-
 	return 0;
 }
 
@@ -165,7 +153,6 @@  static int cros_ec_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm,
 				 struct pwm_state *state)
 {
 	struct cros_ec_pwm_device *ec_pwm = pwm_to_cros_ec_pwm(chip);
-	struct cros_ec_pwm *channel = &ec_pwm->channel[pwm->hwpwm];
 	int ret;
 
 	ret = cros_ec_pwm_get_duty(ec_pwm->ec, ec_pwm->use_pwm_type, pwm->hwpwm);
@@ -175,23 +162,10 @@  static int cros_ec_pwm_get_state(struct pwm_chip *chip, struct pwm_device *pwm,
 	}
 
 	state->enabled = (ret > 0);
+	state->duty_cycle = ret;
 	state->period = EC_PWM_MAX_DUTY;
 	state->polarity = PWM_POLARITY_NORMAL;
 
-	/*
-	 * Note that "disabled" and "duty cycle == 0" are treated the same. If
-	 * the cached duty cycle is not zero, used the cached duty cycle. This
-	 * ensures that the configured duty cycle is kept across a disable and
-	 * enable operation and avoids potentially confusing consumers.
-	 *
-	 * For the case of the initial hardware readout, channel->duty_cycle
-	 * will be 0 and the actual duty cycle read from the EC is used.
-	 */
-	if (ret == 0 && channel->duty_cycle > 0)
-		state->duty_cycle = channel->duty_cycle;
-	else
-		state->duty_cycle = ret;
-
 	return 0;
 }
 
@@ -291,11 +265,6 @@  static int cros_ec_pwm_probe(struct platform_device *pdev)
 	chip->ops = &cros_ec_pwm_ops;
 	chip->of_xlate = cros_ec_pwm_xlate;
 
-	ec_pwm->channel = devm_kcalloc(dev, chip->npwm, sizeof(*ec_pwm->channel),
-					GFP_KERNEL);
-	if (!ec_pwm->channel)
-		return -ENOMEM;
-
 	dev_dbg(dev, "Probed %u PWMs\n", chip->npwm);
 
 	ret = devm_pwmchip_add(dev, chip);