Skip to contentSkip to conversation

iio: PM macro rework continued.

PATCH Applied Jonathan Cameron 3 participants 21 messages Aug 7, 2022 – Sep 18, 2022

Subsystems (2)
lore
Share & tools
lore

Jonathan Cameron

author maintainer
#0 lore raw

cover iio: PM macro rework continued.

From: Jonathan Cameron <[email protected]>

These are straight forward cases so I've grouped them together. Aim here is to move to the macros that don't need __maybe_unused markings and generally simplify the handling of different CONFIG_PM* options.

Jonathan Cameron (6):
iio: proximity: sx9310: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and

    pm_sleep_ptr()
  iio: proximity: sx9324: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and
    pm_sleep_ptr()
  iio: proximity: sx9360: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and
    pm_sleep_ptr()
  iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops
  iio: accel: bmi088: Use EXPORT_NS_GPL_RUNTIME_DEV_PM_OPS() and
    pm_ptr()

iio: light: st_uvis25: Use EXPORT_NS_SIMPLE_DEV_PM_OPS()

 drivers/iio/accel/bmi088-accel-core.c | 15 ++++++---------
 drivers/iio/accel/bmi088-accel-spi.c  |  2 +-
 drivers/iio/light/st_uvis25_core.c    |  9 +++------
 drivers/iio/light/st_uvis25_i2c.c     |  2 +-
 drivers/iio/light/st_uvis25_spi.c     |  2 +-
 drivers/iio/proximity/srf04.c         | 10 +++++-----
 drivers/iio/proximity/sx9310.c        |  8 ++++----
 drivers/iio/proximity/sx9324.c        |  8 ++++----
 drivers/iio/proximity/sx9360.c        |  8 ++++----
 9 files changed, 29 insertions(+), 35 deletions(-)
Signature · 1 line
2.37.1
author maintainer
#1 lore raw

PATCH 1/6 iio: proximity: sx9310: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and pm_sleep_ptr()

From: Jonathan Cameron <[email protected]>

These new macros avoid the need for marking the callbacks __maybe_unused whilst ensuring both callbacks and structure may be dropped by the compiler if CONFIG_PM_SLEEP is not enabled.

---
 drivers/iio/proximity/sx9310.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/iio/proximity/sx9310.c +4 -4
index ea7318b508ea..0e4747ccd3cf 100644
@@ -965,7 +965,7 @@ static int sx9310_probe(struct i2c_client *client)
 	return sx_common_probe(client, &sx9310_chip_info, &sx9310_regmap_config);
 }
 
-static int __maybe_unused sx9310_suspend(struct device *dev)
+static int sx9310_suspend(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	u8 ctrl0;
@@ -991,7 +991,7 @@ static int __maybe_unused sx9310_suspend(struct device *dev)
 	return ret;
 }
 
-static int __maybe_unused sx9310_resume(struct device *dev)
+static int sx9310_resume(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	int ret;
@@ -1013,7 +1013,7 @@ static int __maybe_unused sx9310_resume(struct device *dev)
 	return 0;
 }
 
-static SIMPLE_DEV_PM_OPS(sx9310_pm_ops, sx9310_suspend, sx9310_resume);
+static DEFINE_SIMPLE_DEV_PM_OPS(sx9310_pm_ops, sx9310_suspend, sx9310_resume);
 
 static const struct acpi_device_id sx9310_acpi_match[] = {
 	{ "STH9310", SX9310_WHOAMI_VALUE },
@@ -1041,7 +1041,7 @@ static struct i2c_driver sx9310_driver = {
 		.name	= "sx9310",
 		.acpi_match_table = sx9310_acpi_match,
 		.of_match_table = sx9310_of_match,
-		.pm = &sx9310_pm_ops,
+		.pm = pm_sleep_ptr(&sx9310_pm_ops),
 
 		/*
 		 * Lots of i2c transfers in probe + over 200 ms waiting in
Signature · 1 line
2.37.1
author maintainer Applied
#2 lore raw

Re: [PATCH 1/6] iio: proximity: sx9310: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and pm_sleep_ptr()

Quoted 8 lines · On Sun, 7 Aug 2022 19:56:13 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

These new macros avoid the need for marking the callbacks __maybe_unused whilst ensuring both callbacks and structure may be dropped by the compiler if CONFIG_PM_SLEEP is not enabled.

Applied

Quoted 44 lines
---
 drivers/iio/proximity/sx9310.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/iio/proximity/sx9310.c +4 -4
index ea7318b508ea..0e4747ccd3cf 100644
@@ -965,7 +965,7 @@ static int sx9310_probe(struct i2c_client *client)
 	return sx_common_probe(client, &sx9310_chip_info, &sx9310_regmap_config);
 }
 
-static int __maybe_unused sx9310_suspend(struct device *dev)
+static int sx9310_suspend(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	u8 ctrl0;
@@ -991,7 +991,7 @@ static int __maybe_unused sx9310_suspend(struct device *dev)
 	return ret;
 }
 
-static int __maybe_unused sx9310_resume(struct device *dev)
+static int sx9310_resume(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	int ret;
@@ -1013,7 +1013,7 @@ static int __maybe_unused sx9310_resume(struct device *dev)
 	return 0;
 }
 
-static SIMPLE_DEV_PM_OPS(sx9310_pm_ops, sx9310_suspend, sx9310_resume);
+static DEFINE_SIMPLE_DEV_PM_OPS(sx9310_pm_ops, sx9310_suspend, sx9310_resume);
 
 static const struct acpi_device_id sx9310_acpi_match[] = {
 	{ "STH9310", SX9310_WHOAMI_VALUE },
@@ -1041,7 +1041,7 @@ static struct i2c_driver sx9310_driver = {
 		.name	= "sx9310",
 		.acpi_match_table = sx9310_acpi_match,
 		.of_match_table = sx9310_of_match,
-		.pm = &sx9310_pm_ops,
+		.pm = pm_sleep_ptr(&sx9310_pm_ops),
 
 		/*
 		 * Lots of i2c transfers in probe + over 200 ms waiting in
author maintainer
#3 lore raw

PATCH 2/6 iio: proximity: sx9324: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and pm_sleep_ptr()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

These new macros avoid the need for marking the callbacks __maybe_unused whilst ensuring both callbacks and structure may be dropped by the compiler if CONFIG_PM_SLEEP is not enabled.

---
 drivers/iio/proximity/sx9324.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/iio/proximity/sx9324.c +4 -4
index edb5a2ce4e27..977cf17cec52 100644
@@ -1073,7 +1073,7 @@ static int sx9324_probe(struct i2c_client *client)
 	return sx_common_probe(client, &sx9324_chip_info, &sx9324_regmap_config);
 }
 
-static int __maybe_unused sx9324_suspend(struct device *dev)
+static int sx9324_suspend(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	unsigned int regval;
@@ -1098,7 +1098,7 @@ static int __maybe_unused sx9324_suspend(struct device *dev)
 	return ret;
 }
 
-static int __maybe_unused sx9324_resume(struct device *dev)
+static int sx9324_resume(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	int ret;
@@ -1114,7 +1114,7 @@ static int __maybe_unused sx9324_resume(struct device *dev)
 	return 0;
 }
 
-static SIMPLE_DEV_PM_OPS(sx9324_pm_ops, sx9324_suspend, sx9324_resume);
+static DEFINE_SIMPLE_DEV_PM_OPS(sx9324_pm_ops, sx9324_suspend, sx9324_resume);
 
 static const struct acpi_device_id sx9324_acpi_match[] = {
 	{ "STH9324", SX9324_WHOAMI_VALUE },
@@ -1139,7 +1139,7 @@ static struct i2c_driver sx9324_driver = {
 		.name	= "sx9324",
 		.acpi_match_table = sx9324_acpi_match,
 		.of_match_table = sx9324_of_match,
-		.pm = &sx9324_pm_ops,
+		.pm = pm_sleep_ptr(&sx9324_pm_ops),
 
 		/*
 		 * Lots of i2c transfers in probe + over 200 ms waiting in
Signature · 1 line
2.37.1
author maintainer Applied
#4 lore raw

Re: [PATCH 2/6] iio: proximity: sx9324: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and pm_sleep_ptr()

Quoted 8 lines · On Sun, 7 Aug 2022 19:56:14 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

These new macros avoid the need for marking the callbacks __maybe_unused whilst ensuring both callbacks and structure may be dropped by the compiler if CONFIG_PM_SLEEP is not enabled.

Applied

author maintainer
#5 lore raw

PATCH 3/6 iio: proximity: sx9360: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and pm_sleep_ptr()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

These new macros avoid the need for marking the callbacks __maybe_unused whilst ensuring both callbacks and structure may be dropped by the compiler if CONFIG_PM_SLEEP is not enabled.

---
 drivers/iio/proximity/sx9360.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/iio/proximity/sx9360.c +4 -4
index d9a12e6be6ca..7fa2213d23ba 100644
@@ -819,7 +819,7 @@ static int sx9360_probe(struct i2c_client *client)
 	return sx_common_probe(client, &sx9360_chip_info, &sx9360_regmap_config);
 }
 
-static int __maybe_unused sx9360_suspend(struct device *dev)
+static int sx9360_suspend(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	unsigned int regval;
@@ -844,7 +844,7 @@ static int __maybe_unused sx9360_suspend(struct device *dev)
 	return ret;
 }
 
-static int __maybe_unused sx9360_resume(struct device *dev)
+static int sx9360_resume(struct device *dev)
 {
 	struct sx_common_data *data = iio_priv(dev_get_drvdata(dev));
 	int ret;
@@ -861,7 +861,7 @@ static int __maybe_unused sx9360_resume(struct device *dev)
 	return 0;
 }
 
-static SIMPLE_DEV_PM_OPS(sx9360_pm_ops, sx9360_suspend, sx9360_resume);
+static DEFINE_SIMPLE_DEV_PM_OPS(sx9360_pm_ops, sx9360_suspend, sx9360_resume);
 
 static const struct acpi_device_id sx9360_acpi_match[] = {
 	{ "STH9360", SX9360_WHOAMI_VALUE },
@@ -886,7 +886,7 @@ static struct i2c_driver sx9360_driver = {
 		.name	= "sx9360",
 		.acpi_match_table = sx9360_acpi_match,
 		.of_match_table = sx9360_of_match,
-		.pm = &sx9360_pm_ops,
+		.pm = pm_sleep_ptr(&sx9360_pm_ops),
 
 		/*
 		 * Lots of i2c transfers in probe + over 200 ms waiting in
Signature · 1 line
2.37.1
author maintainer Applied
#6 lore raw

Re: [PATCH 3/6] iio: proximity: sx9360: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and pm_sleep_ptr()

Quoted 8 lines · On Sun, 7 Aug 2022 19:56:15 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

These new macros avoid the need for marking the callbacks __maybe_unused whilst ensuring both callbacks and structure may be dropped by the compiler if CONFIG_PM_SLEEP is not enabled.

Applied.

author maintainer
#7 lore raw

PATCH 4/6 iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

If CONFIG_PM is not set, the pm_ptr() will ensure that the struct dev_pm_ops and callbacks are removed without the need for __maybe_unused markings.

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS() because that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

---
 drivers/iio/proximity/srf04.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/iio/proximity/srf04.c +5 -5
index 05015351a34a..faf2f806ce80 100644
@@ -359,7 +359,7 @@ static int srf04_remove(struct platform_device *pdev)
 	return 0;
 }
 
-static int __maybe_unused srf04_pm_runtime_suspend(struct device *dev)
+static int  srf04_pm_runtime_suspend(struct device *dev)
 {
 	struct platform_device *pdev = container_of(dev,
 						struct platform_device, dev);
@@ -371,7 +371,7 @@ static int __maybe_unused srf04_pm_runtime_suspend(struct device *dev)
 	return 0;
 }
 
-static int __maybe_unused srf04_pm_runtime_resume(struct device *dev)
+static int srf04_pm_runtime_resume(struct device *dev)
 {
 	struct platform_device *pdev = container_of(dev,
 						struct platform_device, dev);
@@ -385,8 +385,8 @@ static int __maybe_unused srf04_pm_runtime_resume(struct device *dev)
 }
 
 static const struct dev_pm_ops srf04_pm_ops = {
-	SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
-				srf04_pm_runtime_resume, NULL)
+	RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+		       srf04_pm_runtime_resume, NULL)
 };
 
 static struct platform_driver srf04_driver = {
@@ -395,7 +395,7 @@ static struct platform_driver srf04_driver = {
 	.driver		= {
 		.name		= "srf04-gpio",
 		.of_match_table	= of_srf04_match,
-		.pm		= &srf04_pm_ops,
+		.pm		= pm_ptr(&srf04_pm_ops),
 	},
 };
 
Signature · 1 line
2.37.1
#8 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Quoted 12 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

If CONFIG_PM is not set, the pm_ptr() will ensure that the struct dev_pm_ops and callbacks are removed without the need for __maybe_unused markings.

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS() because that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

...

drivers/iio/proximity/srf04.c static int __maybe_unused srf04_pm_runtime_resume(struct device *dev) · line 390

Quoted 6 lines

static const struct dev_pm_ops srf04_pm_ops = {

drivers/iio/proximity/srf04.c +2 -2
-       SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
-                               srf04_pm_runtime_resume, NULL)
+       RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+                      srf04_pm_runtime_resume, NULL)
 };

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#9 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Hi Andy,

Quoted 24 lines · Andy Shevchenko
Quoted 13 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

If CONFIG_PM is not set, the pm_ptr() will ensure that the struct dev_pm_ops and callbacks are removed without the need for
__maybe_unused
markings.

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS() because that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

...

Quoted 6 lines

static const struct dev_pm_ops srf04_pm_ops = {
- SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
- srf04_pm_runtime_resume, NULL)
+ RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+ srf04_pm_runtime_resume, NULL)
};

Quoted 3 lines

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Read the commit message?

Cheers,
-Paul

#10 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Quoted 4 lines · Paul Cercueil
Quoted 2 lines · Andy Shevchenko

On Sun, Aug 7, 2022 at 8:46 PM Jonathan Cameron <[email protected]> wrote:

...

Quoted 15 lines
Quoted 15 lines
Quoted 5 lines

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS() because that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

...

Quoted 6 lines

static const struct dev_pm_ops srf04_pm_ops = {
- SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
- srf04_pm_runtime_resume, NULL)
+ RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+ srf04_pm_runtime_resume, NULL)
};

Quoted 5 lines
Quoted 3 lines

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Read the commit message?

Yes, and I'm not sure how that part is relevant. The callbacks won't be called if pm_ptr() equals no-op, no?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#11 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Quoted 24 lines · Andy Shevchenko
Quoted 4 lines · Paul Cercueil
Quoted 2 lines · Andy Shevchenko

On Sun, Aug 7, 2022 at 8:46 PM Jonathan Cameron <[email protected]> wrote:

...

Quoted 16 lines
Quoted 1 line
Quoted 1 line

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS()

because

Quoted 14 lines
Quoted 4 lines

that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

...

Quoted 6 lines

static const struct dev_pm_ops srf04_pm_ops = {
- SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
- srf04_pm_runtime_resume, NULL)
+ RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+ srf04_pm_runtime_resume, NULL)
};

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Read the commit message?

Yes, and I'm not sure how that part is relevant. The callbacks won't be called if pm_ptr() equals no-op, no?

Have a look at the definition of DEFINE_RUNTIME_DEV_PM_OPS(). I believe it does not do what you think it does.

What the commit message says is that using DEFINE_RUNTIME_DEV_PM_OPS() would add .suspend/.resume callbacks, which aren't provided with the current code.

Cheers,
-Paul

#12 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Quoted 39 lines · Paul Cercueil
Quoted 25 lines · Andy Shevchenko
Quoted 4 lines · Paul Cercueil
Quoted 2 lines · Andy Shevchenko

On Sun, Aug 7, 2022 at 8:46 PM Jonathan Cameron <[email protected]> wrote:

...

Quoted 16 lines
Quoted 1 line
Quoted 1 line

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS()

because

Quoted 14 lines
Quoted 4 lines

that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

...

Quoted 6 lines

static const struct dev_pm_ops srf04_pm_ops = {
- SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
- srf04_pm_runtime_resume, NULL)
+ RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+ srf04_pm_runtime_resume, NULL)
};

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Read the commit message?

Yes, and I'm not sure how that part is relevant. The callbacks won't be called if pm_ptr() equals no-op, no?

Quoted 6 lines

Have a look at the definition of DEFINE_RUNTIME_DEV_PM_OPS(). I believe it does not do what you think it does.

What the commit message says is that using DEFINE_RUNTIME_DEV_PM_OPS() would add .suspend/.resume callbacks, which aren't provided with the current code.

Effectively the use of DEFINE_RUNTIME_DEV_PM_OPS() enables system sleep with the same callbacks as runtime PM. I don't see any disadvantages of using it instead of keeping runtime PM only. That said, I don't see the commit message that describes that nuance. It rather says, as I said, something irrelevant.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#13 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Quoted 56 lines · Andy Shevchenko
Quoted 46 lines · Paul Cercueil
Quoted 1 line · Andy Shevchenko

On Mon, Aug 8, 2022 at 11:35 AM Paul Cercueil

<[email protected]>

Quoted 4 lines
Quoted 3 lines · wrote:
Quoted 1 line · Andy Shevchenko

On Sun, Aug 7, 2022 at 8:46 PM Jonathan Cameron

<[email protected]>

Quoted 8 lines
Quoted 1 line
Quoted 1 line

wrote:

...

Quoted 4 lines
Quoted 1 line
Quoted 1 line

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS()

because

Quoted 2 lines
Quoted 2 lines

that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm

functions

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

(whether the necessary GPIO is provided). It may be

possible to

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

clean that up in future by moving the checks into the

callbacks.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

...

Quoted 3 lines

static const struct dev_pm_ops srf04_pm_ops = {
- SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
- srf04_pm_runtime_resume,

NULL)

Quoted 11 lines
Quoted 9 lines
Quoted 7 lines
Quoted 3 lines

+ RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+ srf04_pm_runtime_resume, NULL)
};

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Read the commit message?

Yes, and I'm not sure how that part is relevant. The callbacks

won't

Quoted 1 line

be called if pm_ptr() equals no-op, no?

Quoted 8 lines

Have a look at the definition of DEFINE_RUNTIME_DEV_PM_OPS(). I
believe
it does not do what you think it does.

What the commit message says is that using
DEFINE_RUNTIME_DEV_PM_OPS()
would add .suspend/.resume callbacks, which aren't provided with the current code.

Quoted 5 lines

Effectively the use of DEFINE_RUNTIME_DEV_PM_OPS() enables system sleep with the same callbacks as runtime PM. I don't see any disadvantages of using it instead of keeping runtime PM only. That said, I don't see the commit message that describes that nuance. It rather says, as I said, something irrelevant.

Probably no disavantages, yes, but that would be a functional change, so I can understand why it's not done in this patchset. That doesn't mean it cannot be done later, though, but somebody would need to test it on hardware.

Cheers,
-Paul

#14 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

Quoted 62 lines · Paul Cercueil
Quoted 54 lines · Andy Shevchenko
Quoted 43 lines · Paul Cercueil
Quoted 1 line · Andy Shevchenko

On Mon, Aug 8, 2022 at 11:35 AM Paul Cercueil

<[email protected]>

Quoted 4 lines
Quoted 3 lines · wrote:
Quoted 1 line · Andy Shevchenko

On Sun, Aug 7, 2022 at 8:46 PM Jonathan Cameron

<[email protected]>

Quoted 8 lines
Quoted 1 line
Quoted 1 line

wrote:

...

Quoted 4 lines
Quoted 1 line
Quoted 1 line

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS()

because

Quoted 2 lines
Quoted 2 lines

that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm

functions

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

(whether the necessary GPIO is provided). It may be

possible to

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

clean that up in future by moving the checks into the

callbacks.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines

...

Quoted 3 lines

static const struct dev_pm_ops srf04_pm_ops = {
- SET_RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
- srf04_pm_runtime_resume,

NULL)

Quoted 11 lines
Quoted 9 lines
Quoted 7 lines
Quoted 3 lines

+ RUNTIME_PM_OPS(srf04_pm_runtime_suspend,
+ srf04_pm_runtime_resume, NULL)
};

static DEFINE_RUNTIME_DEV_PM_OPS(...);

?

Read the commit message?

Yes, and I'm not sure how that part is relevant. The callbacks

won't

Quoted 1 line

be called if pm_ptr() equals no-op, no?

Quoted 8 lines

Have a look at the definition of DEFINE_RUNTIME_DEV_PM_OPS(). I
believe
it does not do what you think it does.

What the commit message says is that using
DEFINE_RUNTIME_DEV_PM_OPS()
would add .suspend/.resume callbacks, which aren't provided with the current code.

Quoted 5 lines

Effectively the use of DEFINE_RUNTIME_DEV_PM_OPS() enables system sleep with the same callbacks as runtime PM. I don't see any disadvantages of using it instead of keeping runtime PM only. That said, I don't see the commit message that describes that nuance. It rather says, as I said, something irrelevant.

Quoted 4 lines

Probably no disavantages, yes, but that would be a functional change, so I can understand why it's not done in this patchset. That doesn't mean it cannot be done later, though, but somebody would need to test it on hardware.

This is a fair point.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author maintainer Applied
#15 lore raw

Re: [PATCH 4/6] iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops

In reply to Jonathan Cameron
Quoted 6 lines · On Sun, 7 Aug 2022 19:56:16 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

If CONFIG_PM is not set, the pm_ptr() will ensure that the struct dev_pm_ops and callbacks are removed without the need for __maybe_unused markings.

Quoted 8 lines

In this case we can't simply use DEFINE_RUNTIME_DEV_PM_OPS() because that would provide suspend and resume functions without the checks the driver is doing before calling runtime_pm functions (whether the necessary GPIO is provided). It may be possible to clean that up in future by moving the checks into the callbacks.

Some consensus reached in the discussion so even though no one gave a tag I feel comfortable taking this one.
The suggested follow up change needs hardware to boost confidence that there are no side effects.

Applied.

Jonathan

author maintainer
#16 lore raw

PATCH 5/6 iio: accel: bmi088: Use EXPORT_NS_GPL_RUNTIME_DEV_PM_OPS() and pm_ptr()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

These macros allow the compiler to remove unused pm ops functions without needing to mark them maybe unused.

Cc LI Qingwu <[email protected]>
Cc Mike Looijmans <[email protected]>
---
 drivers/iio/accel/bmi088-accel-core.c | 15 ++++++---------
 drivers/iio/accel/bmi088-accel-spi.c  |  2 +-
 2 files changed, 7 insertions(+), 10 deletions(-)
drivers/iio/accel/bmi088-accel-core.c +6 -9
index bca4cf98bf4d..84edcc78d796 100644
@@ -606,7 +606,7 @@ void bmi088_accel_core_remove(struct device *dev)
 }
 EXPORT_SYMBOL_NS_GPL(bmi088_accel_core_remove, IIO_BMI088);
 
-static int __maybe_unused bmi088_accel_runtime_suspend(struct device *dev)
+static int bmi088_accel_runtime_suspend(struct device *dev)
 {
 	struct iio_dev *indio_dev = dev_get_drvdata(dev);
 	struct bmi088_accel_data *data = iio_priv(indio_dev);
@@ -614,7 +614,7 @@ static int __maybe_unused bmi088_accel_runtime_suspend(struct device *dev)
 	return bmi088_accel_power_down(data);
 }
 
-static int __maybe_unused bmi088_accel_runtime_resume(struct device *dev)
+static int bmi088_accel_runtime_resume(struct device *dev)
 {
 	struct iio_dev *indio_dev = dev_get_drvdata(dev);
 	struct bmi088_accel_data *data = iio_priv(indio_dev);
@@ -622,13 +622,10 @@ static int __maybe_unused bmi088_accel_runtime_resume(struct device *dev)
 	return bmi088_accel_power_up(data);
 }
 
-const struct dev_pm_ops bmi088_accel_pm_ops = {
-	SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
-				pm_runtime_force_resume)
-	SET_RUNTIME_PM_OPS(bmi088_accel_runtime_suspend,
-			   bmi088_accel_runtime_resume, NULL)
-};
-EXPORT_SYMBOL_NS_GPL(bmi088_accel_pm_ops, IIO_BMI088);
+EXPORT_NS_GPL_RUNTIME_DEV_PM_OPS(bmi088_accel_pm_ops,
+				 bmi088_accel_runtime_suspend,
+				 bmi088_accel_runtime_resume, NULL,
+				 IIO_BMI088);
 
 MODULE_AUTHOR("Niek van Agt <[email protected]>");
 MODULE_LICENSE("GPL v2");
drivers/iio/accel/bmi088-accel-spi.c +1 -1
index 9e2ed3bd5661..ee540edd8412 100644
@@ -80,7 +80,7 @@ MODULE_DEVICE_TABLE(spi, bmi088_accel_id);
 static struct spi_driver bmi088_accel_driver = {
 	.driver = {
 		.name	= "bmi088_accel_spi",
-		.pm	= &bmi088_accel_pm_ops,
+		.pm	= pm_ptr(&bmi088_accel_pm_ops),
 		.of_match_table = bmi088_of_match,
 	},
 	.probe		= bmi088_accel_probe,
Signature · 1 line
2.37.1
author maintainer Applied
#17 lore raw

Re: [PATCH 5/6] iio: accel: bmi088: Use EXPORT_NS_GPL_RUNTIME_DEV_PM_OPS() and pm_ptr()

Quoted 9 lines · On Sun, 7 Aug 2022 19:56:17 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

These macros allow the compiler to remove unused pm ops functions without needing to mark them maybe unused.

Cc LI Qingwu <[email protected]>
Cc Mike Looijmans <[email protected]>

Applied.

author maintainer
#18 lore raw

PATCH 6/6 iio: light: st_uvis25: Use EXPORT_NS_SIMPLE_DEV_PM_OPS()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Using this new macro removes the need to mark the callbacks __maybe_unused. One slightly complexity in this case is that the export will exist if CONFIG_PM is set, but only be used if CONFIG_PM_SLEEP is also set. This is harmless.

---
 drivers/iio/light/st_uvis25_core.c | 9 +++------
 drivers/iio/light/st_uvis25_i2c.c  | 2 +-
 drivers/iio/light/st_uvis25_spi.c  | 2 +-
 3 files changed, 5 insertions(+), 8 deletions(-)
drivers/iio/light/st_uvis25_core.c +3 -6
index 3d4cc1180b6a..c737d3e193ae 100644
@@ -325,7 +325,7 @@ int st_uvis25_probe(struct device *dev, int irq, struct regmap *regmap)
 }
 EXPORT_SYMBOL_NS(st_uvis25_probe, IIO_UVIS25);
 
-static int __maybe_unused st_uvis25_suspend(struct device *dev)
+static int st_uvis25_suspend(struct device *dev)
 {
 	struct iio_dev *iio_dev = dev_get_drvdata(dev);
 	struct st_uvis25_hw *hw = iio_priv(iio_dev);
@@ -334,7 +334,7 @@ static int __maybe_unused st_uvis25_suspend(struct device *dev)
 				  ST_UVIS25_REG_ODR_MASK, 0);
 }
 
-static int __maybe_unused st_uvis25_resume(struct device *dev)
+static int st_uvis25_resume(struct device *dev)
 {
 	struct iio_dev *iio_dev = dev_get_drvdata(dev);
 	struct st_uvis25_hw *hw = iio_priv(iio_dev);
@@ -346,10 +346,7 @@ static int __maybe_unused st_uvis25_resume(struct device *dev)
 	return 0;
 }
 
-const struct dev_pm_ops st_uvis25_pm_ops = {
-	SET_SYSTEM_SLEEP_PM_OPS(st_uvis25_suspend, st_uvis25_resume)
-};
-EXPORT_SYMBOL_NS(st_uvis25_pm_ops, IIO_UVIS25);
+EXPORT_NS_SIMPLE_DEV_PM_OPS(st_uvis25_pm_ops, st_uvis25_suspend, st_uvis25_resume, IIO_UVIS25);
 
 MODULE_AUTHOR("Lorenzo Bianconi <[email protected]>");
 MODULE_DESCRIPTION("STMicroelectronics uvis25 sensor driver");
drivers/iio/light/st_uvis25_i2c.c +1 -1
index b06d09af28a3..c982b0b255cf 100644
@@ -55,7 +55,7 @@ MODULE_DEVICE_TABLE(i2c, st_uvis25_i2c_id_table);
 static struct i2c_driver st_uvis25_driver = {
 	.driver = {
 		.name = "st_uvis25_i2c",
-		.pm = &st_uvis25_pm_ops,
+		.pm = pm_sleep_ptr(&st_uvis25_pm_ops),
 		.of_match_table = st_uvis25_i2c_of_match,
 	},
 	.probe = st_uvis25_i2c_probe,
drivers/iio/light/st_uvis25_spi.c +1 -1
index 3a4dc6d7180c..86a232320d7d 100644
@@ -55,7 +55,7 @@ MODULE_DEVICE_TABLE(spi, st_uvis25_spi_id_table);
 static struct spi_driver st_uvis25_driver = {
 	.driver = {
 		.name = "st_uvis25_spi",
-		.pm = &st_uvis25_pm_ops,
+		.pm = pm_sleep_ptr(&st_uvis25_pm_ops),
 		.of_match_table = st_uvis25_spi_of_match,
 	},
 	.probe = st_uvis25_spi_probe,
Signature · 1 line
2.37.1
author maintainer Applied
#19 lore raw

Re: [PATCH 6/6] iio: light: st_uvis25: Use EXPORT_NS_SIMPLE_DEV_PM_OPS()

Quoted 9 lines · On Sun, 7 Aug 2022 19:56:18 +0100 Jonathan Cameron <[email protected]> wrote:

From: Jonathan Cameron <[email protected]>

Using this new macro removes the need to mark the callbacks __maybe_unused. One slightly complexity in this case is that the export will exist if CONFIG_PM is set, but only be used if CONFIG_PM_SLEEP is also set. This is harmless.

Applied

Reviewed-by on the v1 series
#20 lore raw
In reply to Jonathan Cameron
Quoted 6 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

These are straight forward cases so I've grouped them together. Aim here is to move to the macros that don't need __maybe_unused markings and generally simplify the handling of different CONFIG_PM* options.

for non-commented and, once addressed as suggested, for commented.

Quoted 26 lines

Jonathan Cameron (6):
iio: proximity: sx9310: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and

    pm_sleep_ptr()
  iio: proximity: sx9324: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and
    pm_sleep_ptr()
  iio: proximity: sx9360: Switch to DEFINE_SIMPLE_DEV_PM_OPS() and
    pm_sleep_ptr()
  iio: proximity: srf04: Use pm_ptr() to remove unused struct dev_pm_ops
  iio: accel: bmi088: Use EXPORT_NS_GPL_RUNTIME_DEV_PM_OPS() and
    pm_ptr()

iio: light: st_uvis25: Use EXPORT_NS_SIMPLE_DEV_PM_OPS()

 drivers/iio/accel/bmi088-accel-core.c | 15 ++++++---------
 drivers/iio/accel/bmi088-accel-spi.c  |  2 +-
 drivers/iio/light/st_uvis25_core.c    |  9 +++------
 drivers/iio/light/st_uvis25_i2c.c     |  2 +-
 drivers/iio/light/st_uvis25_spi.c     |  2 +-
 drivers/iio/proximity/srf04.c         | 10 +++++-----
 drivers/iio/proximity/sx9310.c        |  8 ++++----
 drivers/iio/proximity/sx9324.c        |  8 ++++----
 drivers/iio/proximity/sx9360.c        |  8 ++++----
 9 files changed, 29 insertions(+), 35 deletions(-)

--
2.37.1

Signature · 2 lines
With Best Regards,
Andy Shevchenko