Skip to contentSkip to conversation

Improve phy-mapphone pm

PATCH In review Tony Lindgren 5 participants 10 messages Nov 17, 2018 – Jun 2, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover Improve phy-mapphone pm

Hi,

Here are two non-urgent patches to get phy-mapphone runtime PM working better. Currently it only works when toggled from the userspace via the sysfs entry for for "auto" and "on".

Regards,

Tony

Tony Lindgren (2):
phy: core: Add phy_pm_runtime_enabled
phy: mapphone-mdm6600: Improve phy related runtime PM calls

 drivers/phy/motorola/phy-mapphone-mdm6600.c | 71 +++++++++++++++------
 drivers/phy/phy-core.c                      |  9 +++
 include/linux/phy/phy.h                     |  6 ++
 3 files changed, 66 insertions(+), 20 deletions(-)
Signature · 1 line
2.19.1
#1 lore raw

PATCH 1/2 phy: core: Add phy_pm_runtime_enabled

The phy driver may need to check phy_pm_runtime_enabled() in suspend as PM runtime for phy may be already disabled when phy power_off() is called.

Cc Pavel Machek <[email protected]>
Cc Sebastian Reichel <[email protected]>
---
 drivers/phy/phy-core.c  | 9 +++++++++
 include/linux/phy/phy.h | 6 ++++++
 2 files changed, 15 insertions(+)
drivers/phy/phy-core.c +9 -0
@@ -209,6 +209,15 @@ int phy_pm_runtime_put_sync(struct phy *phy)
 }
 EXPORT_SYMBOL_GPL(phy_pm_runtime_put_sync);
 
+bool phy_pm_runtime_enabled(struct phy *phy)
+{
+	if (!phy)
+		return false;
+
+	return pm_runtime_enabled(&phy->dev);
+}
+EXPORT_SYMBOL_GPL(phy_pm_runtime_enabled);
+
 void phy_pm_runtime_allow(struct phy *phy)
 {
 	if (!phy)
include/linux/phy/phy.h +6 -0
@@ -158,6 +158,7 @@ int phy_pm_runtime_get(struct phy *phy);
 int phy_pm_runtime_get_sync(struct phy *phy);
 int phy_pm_runtime_put(struct phy *phy);
 int phy_pm_runtime_put_sync(struct phy *phy);
+bool phy_pm_runtime_enabled(struct phy *phy);
 void phy_pm_runtime_allow(struct phy *phy);
 void phy_pm_runtime_forbid(struct phy *phy);
 int phy_init(struct phy *phy);
@@ -240,6 +241,11 @@ static inline int phy_pm_runtime_put_sync(struct phy *phy)
 	return -ENOSYS;
 }
 
+static inline bool phy_pm_runtime_enabled(struct phy *phy)
+{
+	return false
+}
+
 static inline void phy_pm_runtime_allow(struct phy *phy)
 {
 	return;
Signature · 1 line
2.19.1
#2 lore raw

Re: [PATCH 1/2] phy: core: Add phy_pm_runtime_enabled

Quoted 27 lines · Tony Lindgren

The phy driver may need to check phy_pm_runtime_enabled() in suspend as PM runtime for phy may be already disabled when phy power_off() is called.

Cc Pavel Machek <[email protected]>
Cc Sebastian Reichel <[email protected]>
---
 drivers/phy/phy-core.c  | 9 +++++++++
 include/linux/phy/phy.h | 6 ++++++
 2 files changed, 15 insertions(+)
drivers/phy/phy-core.c +0 -0
include/linux/phy/phy.h +1 -0
@@ -158,6 +158,7 @@ int phy_pm_runtime_get(struct phy *phy);
 int phy_pm_runtime_get_sync(struct phy *phy);
 int phy_pm_runtime_put(struct phy *phy);
 int phy_pm_runtime_put_sync(struct phy *phy);
+bool phy_pm_runtime_enabled(struct phy *phy);
 void phy_pm_runtime_allow(struct phy *phy);
 void phy_pm_runtime_forbid(struct phy *phy);
 int phy_init(struct phy *phy);

include/linux/phy/phy.h static inline int phy_pm_runtime_put_sync(struct phy *phy) · line 246

Quoted 6 lines
include/linux/phy/phy.h +3 -0
@@ -240,6 +241,11 @@ static inline int phy_pm_runtime_put_sync(struct phy *phy)
 	return -ENOSYS;
 }
 
+static inline bool phy_pm_runtime_enabled(struct phy *phy)
+{
+	return false

Missing semicolon.

include/linux/phy/phy.h static inline int phy_pm_runtime_put_sync(struct phy *phy) · line 251

Quoted 5 lines
include/linux/phy/phy.h +2 -0
+}
+
 static inline void phy_pm_runtime_allow(struct phy *phy)
 {
 	return;

Johan

#3 lore raw

Re: [PATCH 1/2] phy: core: Add phy_pm_runtime_enabled

Quoted 31 lines · Johan Hovold
Quoted 30 lines · Tony Lindgren

The phy driver may need to check phy_pm_runtime_enabled() in suspend as PM runtime for phy may be already disabled when phy power_off() is called.

Cc Pavel Machek <[email protected]>
Cc Sebastian Reichel <[email protected]>
---
 drivers/phy/phy-core.c  | 9 +++++++++
 include/linux/phy/phy.h | 6 ++++++
 2 files changed, 15 insertions(+)
drivers/phy/phy-core.c +0 -0
include/linux/phy/phy.h +1 -0
@@ -158,6 +158,7 @@ int phy_pm_runtime_get(struct phy *phy);
 int phy_pm_runtime_get_sync(struct phy *phy);
 int phy_pm_runtime_put(struct phy *phy);
 int phy_pm_runtime_put_sync(struct phy *phy);
+bool phy_pm_runtime_enabled(struct phy *phy);
 void phy_pm_runtime_allow(struct phy *phy);
 void phy_pm_runtime_forbid(struct phy *phy);
 int phy_init(struct phy *phy);
@@ -240,6 +241,11 @@ static inline int phy_pm_runtime_put_sync(struct phy *phy)
 	return -ENOSYS;
 }
 
Quoted 5 lines

include/linux/phy/phy.h static inline int phy_pm_runtime_put_sync(struct phy *phy) · line 246

Quoted 3 lines
include/linux/phy/phy.h +3 -0
+static inline bool phy_pm_runtime_enabled(struct phy *phy)
+{
+	return false

Missing semicolon.

Oops thanks for catching that. I guess I did not try building without CONFIG_GENERIC_PHY. Will fix and repost.

Regards,

Tony

#4 lore raw

Re: [PATCH 1/2] phy: core: Add phy_pm_runtime_enabled

Hi!

Quoted 30 lines
Quoted 30 lines
Quoted 30 lines

The phy driver may need to check phy_pm_runtime_enabled() in suspend as PM runtime for phy may be already disabled when phy power_off() is called.

Cc Pavel Machek <[email protected]>
Cc Sebastian Reichel <[email protected]>
---
 drivers/phy/phy-core.c  | 9 +++++++++
 include/linux/phy/phy.h | 6 ++++++
 2 files changed, 15 insertions(+)
drivers/phy/phy-core.c +0 -0
include/linux/phy/phy.h +1 -0
@@ -158,6 +158,7 @@ int phy_pm_runtime_get(struct phy *phy);
 int phy_pm_runtime_get_sync(struct phy *phy);
 int phy_pm_runtime_put(struct phy *phy);
 int phy_pm_runtime_put_sync(struct phy *phy);
+bool phy_pm_runtime_enabled(struct phy *phy);
 void phy_pm_runtime_allow(struct phy *phy);
 void phy_pm_runtime_forbid(struct phy *phy);
 int phy_init(struct phy *phy);
@@ -240,6 +241,11 @@ static inline int phy_pm_runtime_put_sync(struct phy *phy)
 	return -ENOSYS;
 }
 
Quoted 8 lines
Quoted 5 lines

include/linux/phy/phy.h static inline int phy_pm_runtime_put_sync(struct phy *phy) · line 246

Quoted 3 lines
include/linux/phy/phy.h +3 -0
+static inline bool phy_pm_runtime_enabled(struct phy *phy)
+{
+	return false

Missing semicolon.

Oops thanks for catching that. I guess I did not try building without CONFIG_GENERIC_PHY. Will fix and repost.

Did this series get lost/forgotten somewhere? Is it still needed? Any way I can help?
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#5 lore raw

Re: [PATCH 1/2] phy: core: Add phy_pm_runtime_enabled

Hi!

Quoted 10 lines
Quoted 10 lines
Quoted 9 lines

include/linux/phy/phy.h static inline int phy_pm_runtime_put_sync(struct phy *phy) · line 246

Quoted 7 lines
include/linux/phy/phy.h +3 -0
@@ -240,6 +241,11 @@ static inline int phy_pm_runtime_put_sync(struct phy *phy)
 	return -ENOSYS;
 }
 
+static inline bool phy_pm_runtime_enabled(struct phy *phy)
+{
+	return false

Missing semicolon.

Quoted 5 lines
Quoted 2 lines

Oops thanks for catching that. I guess I did not try building without CONFIG_GENERIC_PHY. Will fix and repost.

Did this series get lost/forgotten somewhere? Is it still needed? Any way I can help?

Aha, it seems v2 of the series was applied, which touches different files and that's why I did not detect it. Sorry for the noise.

Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#6 lore raw

Re: [PATCH 1/2] phy: core: Add phy_pm_runtime_enabled

In reply to Tony Lindgren
I love your patch! Yet something to improve:

Hi Tony,

I love your patch! Yet something to improve:

[auto build test ERROR on phy/next]
[also build test ERROR on v4.20-rc3 next-20181119]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Tony-Lindgren/Improve-phy-mapphone-pm/20181119-222603
base:   https://git.kernel.org/pub/scm/linux/kernel/git/kishon/linux-phy.git next
config: riscv-tinyconfig (attached as .config)
compiler: riscv64-linux-gcc (GCC) 8.1.0

reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
GCC_VERSION=8.1.0 make.cross ARCH=riscv

All errors (new ones prefixed by >>):

In file included from include/linux/usb/otg.h:13,

                    from include/linux/usb/of.h:12,
                    from drivers/usb/phy/of.c:9:
   include/linux/phy/phy.h: In function 'phy_pm_runtime_enabled':
>> include/linux/phy/phy.h:246:14: error: expected ';' before '}' token
     return false
                 ^
                 ;
    }
    ~             

vim +246 include/linux/phy/phy.h

243
244 static inline bool phy_pm_runtime_enabled(struct phy *phy)
245 {

 > 246		return false

247 }
248

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

#7 lore raw

PATCH 2/2 phy: mapphone-mdm6600: Improve phy related runtime PM calls

In reply to Tony Lindgren

I noticed that phy_pm_runtime_get_sync() and phy_pm_runtime_put() are not currently doing anything for phy-mapphone-mdm6600, only the sysfs interface for works for "auto" and "on".

This is because of the shared GPIO pins between mdm6600 USB port and n_gsm port. We have not enabled runtime PM for the phy driver until after we've booted up mdm6600 properly to the USB mode. Otherwise phy_create() would have called pm_runtime_enable() and pm_runtime_no_callbacks() automatically on init.

Let's fix this by registering the phy a bit later after we've powered up the mdm6600 USB port.

And as the PM runtime support is only needed for the n_gsm mode and not for USB, we can allow the device to idle between phy_mdm6600_power_on() and phy_mdm6600_power_off(). Note that for suspend, runtime_pm is already disabled for the phy so we need to check for phy_pm_runtime_enabled().

Cc Pavel Machek <[email protected]>
Cc Sebastian Reichel <[email protected]>
---
 drivers/phy/motorola/phy-mapphone-mdm6600.c | 71 +++++++++++++++------
 1 file changed, 51 insertions(+), 20 deletions(-)
drivers/phy/motorola/phy-mapphone-mdm6600.c +51 -20
@@ -16,6 +16,7 @@
 #include <linux/gpio/consumer.h>
 #include <linux/of_platform.h>
 #include <linux/phy/phy.h>
+#include <linux/pinctrl/consumer.h>
 
 #define PHY_MDM6600_PHY_DELAY_MS	4000	/* PHY enable 2.2s to 3.5s */
 #define PHY_MDM6600_ENABLED_DELAY_MS	8000	/* 8s more total for MDM6600 */
@@ -124,12 +125,22 @@ static int phy_mdm6600_power_on(struct phy *x)
 {
 	struct phy_mdm6600 *ddata = phy_get_drvdata(x);
 	struct gpio_desc *enable_gpio = ddata->ctrl_gpios[PHY_MDM6600_ENABLE];
+	int error;
 
 	if (!ddata->enabled)
 		return -ENODEV;
 
+	error = pinctrl_pm_select_default_state(ddata->dev);
+	if (error)
+		dev_warn(ddata->dev, "%s: error with default_state: %i\n",
+			 __func__, error);
+
 	gpiod_set_value_cansleep(enable_gpio, 1);
 
+	/* Allow aggressive PM for USB, it's only needed for n_gsm port */
+	if (phy_pm_runtime_enabled(ddata->generic_phy))
+		phy_pm_runtime_put(ddata->generic_phy);
+
 	return 0;
 }
 
@@ -137,12 +148,26 @@ static int phy_mdm6600_power_off(struct phy *x)
 {
 	struct phy_mdm6600 *ddata = phy_get_drvdata(x);
 	struct gpio_desc *enable_gpio = ddata->ctrl_gpios[PHY_MDM6600_ENABLE];
+	int error;
 
 	if (!ddata->enabled)
 		return -ENODEV;
 
+	/* Paired with phy_pm_runtime_put() in phy_mdm6600_power_on() */
+	if (phy_pm_runtime_enabled(ddata->generic_phy)) {
+		error = phy_pm_runtime_get(ddata->generic_phy);
+		if (error < 0 && error != -EINPROGRESS)
+			dev_warn(ddata->dev, "%s: phy_pm_runtime_get: %i\n",
+				 __func__, error);
+	}
+
 	gpiod_set_value_cansleep(enable_gpio, 0);
 
+	error = pinctrl_pm_select_sleep_state(ddata->dev);
+	if (error)
+		dev_warn(ddata->dev, "%s: error with sleep_state: %i\n",
+			 __func__, error);
+
 	return 0;
 }
 
@@ -555,28 +580,17 @@ static int phy_mdm6600_probe(struct platform_device *pdev)
 	ddata->dev = &pdev->dev;
 	platform_set_drvdata(pdev, ddata);
 
+	/* Active state selected in phy_mdm6600_power_on() */
+	error = pinctrl_pm_select_sleep_state(ddata->dev);
+	if (error)
+		dev_warn(ddata->dev, "%s: error with sleep_state: %i\n",
+			 __func__, error);
+
 	error = phy_mdm6600_init_lines(ddata);
 	if (error)
 		return error;
 
 	phy_mdm6600_init_irq(ddata);
-
-	ddata->generic_phy = devm_phy_create(ddata->dev, NULL, &gpio_usb_ops);
-	if (IS_ERR(ddata->generic_phy)) {
-		error = PTR_ERR(ddata->generic_phy);
-		goto cleanup;
-	}
-
-	phy_set_drvdata(ddata->generic_phy, ddata);
-
-	ddata->phy_provider =
-		devm_of_phy_provider_register(ddata->dev,
-					      of_phy_simple_xlate);
-	if (IS_ERR(ddata->phy_provider)) {
-		error = PTR_ERR(ddata->phy_provider);
-		goto cleanup;
-	}
-
 	schedule_delayed_work(&ddata->bootup_work, 0);
 
 	/*
@@ -600,14 +614,31 @@ static int phy_mdm6600_probe(struct platform_device *pdev)
 	if (error < 0) {
 		dev_warn(ddata->dev, "failed to wake modem: %i\n", error);
 		pm_runtime_put_noidle(ddata->dev);
+		goto cleanup;
 	}
+
+	ddata->generic_phy = devm_phy_create(ddata->dev, NULL, &gpio_usb_ops);
+	if (IS_ERR(ddata->generic_phy)) {
+		error = PTR_ERR(ddata->generic_phy);
+		goto idle;
+	}
+
+	phy_set_drvdata(ddata->generic_phy, ddata);
+
+	ddata->phy_provider =
+		devm_of_phy_provider_register(ddata->dev,
+					      of_phy_simple_xlate);
+	if (IS_ERR(ddata->phy_provider))
+		error = PTR_ERR(ddata->phy_provider);
+
+idle:
 	pm_runtime_mark_last_busy(ddata->dev);
 	pm_runtime_put_autosuspend(ddata->dev);
 
-	return 0;
-
 cleanup:
-	phy_mdm6600_device_power_off(ddata);
+	if (error < 0)
+		phy_mdm6600_device_power_off(ddata);
+
 	return error;
 }
 
Signature · 1 line
2.19.1
#8 lore raw

Re: [PATCH 2/2] phy: mapphone-mdm6600: Improve phy related runtime PM calls

Hi Tony,

Quoted 45 lines · Tony Lindgren

I noticed that phy_pm_runtime_get_sync() and phy_pm_runtime_put() are not currently doing anything for phy-mapphone-mdm6600, only the sysfs interface for works for "auto" and "on".

This is because of the shared GPIO pins between mdm6600 USB port and n_gsm port. We have not enabled runtime PM for the phy driver until after we've booted up mdm6600 properly to the USB mode. Otherwise phy_create() would have called pm_runtime_enable() and pm_runtime_no_callbacks() automatically on init.

Let's fix this by registering the phy a bit later after we've powered up the mdm6600 USB port.

And as the PM runtime support is only needed for the n_gsm mode and not for USB, we can allow the device to idle between phy_mdm6600_power_on() and phy_mdm6600_power_off(). Note that for suspend, runtime_pm is already disabled for the phy so we need to check for phy_pm_runtime_enabled().

Cc Pavel Machek <[email protected]>
Cc Sebastian Reichel <[email protected]>
---
 drivers/phy/motorola/phy-mapphone-mdm6600.c | 71 +++++++++++++++------
 1 file changed, 51 insertions(+), 20 deletions(-)
drivers/phy/motorola/phy-mapphone-mdm6600.c +2 -0
@@ -16,6 +16,7 @@
 #include <linux/gpio/consumer.h>
 #include <linux/of_platform.h>
 #include <linux/phy/phy.h>
+#include <linux/pinctrl/consumer.h>
 
 #define PHY_MDM6600_PHY_DELAY_MS	4000	/* PHY enable 2.2s to 3.5s */
 #define PHY_MDM6600_ENABLED_DELAY_MS	8000	/* 8s more total for MDM6600 */
@@ -124,12 +125,22 @@ static int phy_mdm6600_power_on(struct phy *x)
 {
 	struct phy_mdm6600 *ddata = phy_get_drvdata(x);
 	struct gpio_desc *enable_gpio = ddata->ctrl_gpios[PHY_MDM6600_ENABLE];
+	int error;
 
 	if (!ddata->enabled)
 		return -ENODEV;
 

drivers/phy/motorola/phy-mapphone-mdm6600.c static int phy_mdm6600_power_on(struct phy *x) · line 142

Quoted 10 lines
drivers/phy/motorola/phy-mapphone-mdm6600.c +8 -0
+	error = pinctrl_pm_select_default_state(ddata->dev);
+	if (error)
+		dev_warn(ddata->dev, "%s: error with default_state: %i\n",
+			 __func__, error);
+
 	gpiod_set_value_cansleep(enable_gpio, 1);
 
+	/* Allow aggressive PM for USB, it's only needed for n_gsm port */
+	if (phy_pm_runtime_enabled(ddata->generic_phy))
+		phy_pm_runtime_put(ddata->generic_phy);

phy_*() API's are generally added to be used by the consumer driver. I guess in this case we can directly use pm_runtime_enabled(x).

Thanks
Kishon

#9 lore raw

Re: [PATCH 2/2] phy: mapphone-mdm6600: Improve phy related runtime PM calls

Quoted 7 lines · Kishon Vijay Abraham I

drivers/phy/motorola/phy-mapphone-mdm6600.c static int phy_mdm6600_power_on(struct phy *x) · line 142

Quoted 3 lines · Tony Lindgren
drivers/phy/motorola/phy-mapphone-mdm6600.c +3 -0
+	/* Allow aggressive PM for USB, it's only needed for n_gsm port */
+	if (phy_pm_runtime_enabled(ddata->generic_phy))
+		phy_pm_runtime_put(ddata->generic_phy);

phy_*() API's are generally added to be used by the consumer driver. I guess in this case we can directly use pm_runtime_enabled(x).

OK sure that works for me too.

Regards,

Tony