IIO: Runtime PM related cleanups.
Subsystems (2)
Jonathan Cameron
maintainerv2 cover IIO: Runtime PM related cleanups.
From: Jonathan Cameron <[email protected]>
Changes since v1:
* Drop all patches since applied.
* Clean up some more unbalanced cases that Mauro pointed out.
* One additional fix for lack of put in an error path.
* Dropped zpa2326 patch for now. The use of runtime pm in that driver is sufficiently odd that it needs a closer look before any changes.
Note the unbalanced put is not a bug as such, because the runtime pm core protects against the reference count going negative. However, it is a bad pattern to have copied into new drivers (as it confuses me) so I'd like to clear it out.
Inspired by Mauro's work on similar issues in media and Julia's coccicheck script.
There will be at least one more set of these once this first set have flushed out any mistakes I may have made.
I checked my assumptions around the excess pm_runtime_put_noidle calls in QEMU but may well be missing something even so.
Jonathan Cameron (6):
iio: imu: mpu6050: Balance runtime pm + use
pm_runtime_resume_and_get()
iio: adc: ads1015: Balance runtime pm + pm_runtime_resume_and_get()
iio: chemical: atlas-sensor: Balance runtime pm +
pm_runtime_resume_and_get()
iio: prox: pulsed-light-v2: Fix misbalance runtime pm in error path
iio: prox: pulsed-light-v2: Use pm_runtime_resume_and_get()
iio: pressure: icp10100: Balance runtime pm + use
pm_runtime_resume_and_get()
drivers/iio/adc/ti-ads1015.c | 5 +---- drivers/iio/chemical/atlas-sensor.c | 13 ++++--------- drivers/iio/imu/inv_mpu6050/inv_mpu_core.c | 19 ++++++------------- drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c | 6 ++---- drivers/iio/pressure/icp10100.c | 5 +++-- .../iio/proximity/pulsedlight-lidar-lite-v2.c | 7 +++++-- 6 files changed, 21 insertions(+), 34 deletions(-)
Signature · 1 line
2.31.1
Jonathan Cameron
maintainerPATCH v2 1/6 iio: imu: mpu6050: Balance runtime pm + use pm_runtime_resume_and_get()
From: Jonathan Cameron <[email protected]>
Remove an unblanced pm_runtime_put_sync_suspend() call
in inv_pu_pm_disable(). Not this call is not a bug, because the runtime pm core will not allow the reference counter to go negative. It is however confusing and serves no purpose.
pm_runtime_resume_and_get() case found using coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
pm_runtime_resume_and_get() returns <= 0 only so simplify related checks to bring this more inline with nearby calls.
This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
--- drivers/iio/imu/inv_mpu6050/inv_mpu_core.c | 19 ++++++------------- drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c | 6 ++---- 2 files changed, 8 insertions(+), 17 deletions(-)
drivers/iio/imu/inv_mpu6050/inv_mpu_core.c +6 -13
@@ -570,11 +570,9 @@ static int inv_mpu6050_read_channel_data(struct iio_dev *indio_dev, freq_hz =INV_MPU6050_DIVIDER_TO_FIFO_RATE (st->chip_config.divider); period_us =1000000 / freq_hz; - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result)return result; - }switch (chan->type) {case IIO_ANGL_VEL :
@@ -812,11 +810,9 @@ static int inv_mpu6050_write_raw(struct iio_dev *indio_dev,return result;mutex_lock (&st->lock); - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result) goto error_write_raw_unlock; - }switch (mask) {case IIO_CHAN_INFO_SCALE :
@@ -930,11 +926,9 @@ inv_mpu6050_fifo_rate_store(struct device *dev, struct device_attribute *attr, result =0 ; goto fifo_rate_fail_unlock; } - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result) goto fifo_rate_fail_unlock; - } result =regmap_write (st->map, st->reg->sample_rate_div, d);if (result)
@@ -1422,7 +1416,6 @@ static void inv_mpu_pm_disable(void *data) {struct device *dev = data; -pm_runtime_put_sync_suspend (dev);pm_runtime_disable (dev); }
drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c +2 -4
@@ -173,11 +173,9 @@ static int inv_mpu6050_set_enable(struct iio_dev *indio_dev, bool enable)if (enable) { scan =inv_scan_query (indio_dev); - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result)return result; - }/* * In case autosuspend didn't trigger, turn off first not * required sensors.
Signature · 1 line
2.31.1
Jonathan Cameron
maintainerPATCH v2 2/6 iio: adc: ads1015: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Jonathan CameronFrom: Jonathan Cameron <[email protected]>
The call to pm_runtime_put_noidle() in remove() is not balancing a counter increment. Note this doesn't matter as the runtime pm core will not allow the counter to go negative. However, it is confusing to the reader so let's remove it.
The pm_runtime_resume_and_get() replacement was found using coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
--- drivers/iio/adc/ti-ads1015.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)
drivers/iio/adc/ti-ads1015.c +1 -4
@@ -323,9 +323,7 @@ static int ads1015_set_power_state(struct ads1015_data *data, bool on)struct device *dev =regmap_get_device (data->regmap);if (on) { - ret =pm_runtime_get_sync (dev); -if (ret <0 ) -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); }else {pm_runtime_mark_last_busy (dev); ret =pm_runtime_put_autosuspend (dev);
@@ -1070,7 +1068,6 @@ static int ads1015_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);/* power down single shot mode */ return ads1015_set_conv_mode (data,ADS1015_SINGLESHOT );
Signature · 1 line
2.31.1
Jonathan Cameron
maintainerPATCH v2 3/6 iio: chemical: atlas-sensor: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Jonathan CameronFrom: Jonathan Cameron <[email protected]>
The pm_runtime_put_noidle() call in remove isn't balanced with any get, so drop it. Note this isn't a bug as the runtime pm core will not allow the reference count to go negative, making this a noop. However, it is confusing to the reader so let's drop it.
pm_runtime_resume_and_get() replacement found using the coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
As pm_runtime_resume_and_get() returns <= 0 take advantage of that to change the error checking to if (ret) which is more in keeping with the rest of this driver.
This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
--- drivers/iio/chemical/atlas-sensor.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-)
drivers/iio/chemical/atlas-sensor.c +4 -9
@@ -410,11 +410,9 @@ static int atlas_buffer_postenable(struct iio_dev *indio_dev)struct atlas_data *data =iio_priv (indio_dev);int ret; - ret =pm_runtime_get_sync (&data->client->dev); -if (ret <0 ) { -pm_runtime_put_noidle (&data->client->dev); + ret =pm_runtime_resume_and_get (&data->client->dev); +if (ret)return ret; - }return atlas_set_interrupt (data, true); }
@@ -487,11 +485,9 @@ static int atlas_read_measurement(struct atlas_data *data, int reg, __be32 *val)int suspended =pm_runtime_suspended (dev);int ret; - ret =pm_runtime_get_sync (dev); -if (ret <0 ) { -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); +if (ret)return ret; - }if (suspended)msleep (data->chip->delay);
@@ -741,7 +737,6 @@ static int atlas_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);return atlas_set_powermode (data,0 ); }
Signature · 1 line
2.31.1
Jonathan Cameron
maintainerPATCH v2 4/6 iio: prox: pulsed-light-v2: Fix misbalance runtime pm in error path
In reply to Jonathan CameronFrom: Jonathan Cameron <[email protected]>
There is one path in which we don't do a runtime pm put and so leave the device enabled for ever more.
--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +2 -1
@@ -164,7 +164,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE );if (ret <0 ) {dev_err (&client->dev,"cannot send start measurement command" ); -return ret; + goto err; }while (tries--) {
@@ -188,6 +188,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) } ret = -EIO ; } +err:pm_runtime_mark_last_busy (&client->dev);pm_runtime_put_autosuspend (&client->dev);
Signature · 1 line
2.31.1
Jonathan Cameron
maintainerPATCH v2 5/6 iio: prox: pulsed-light-v2: Use pm_runtime_resume_and_get()
In reply to Jonathan CameronFrom: Jonathan Cameron <[email protected]>
Using this new call makes it easy to handle any errors as a result of runtime resume as it exits without leaving the reference count elevated.
--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +3 -1
@@ -158,7 +158,9 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg)int tries =10 ;int ret; -pm_runtime_get_sync (&client->dev); + ret =pm_runtime_resume_and_get (&client->dev); +if (ret <0 ) +return ret;/* start sample */ ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE );
Signature · 1 line
2.31.1
Jonathan Cameron
maintainerPATCH v2 6/6 iio: pressure: icp10100: Balance runtime pm + use pm_runtime_resume_and_get()
In reply to Jonathan CameronFrom: Jonathan Cameron <[email protected]>
The devm_ handled runtime pm disable calls pm_runtime_put_sync_suspend() which isn't balancing a matching get call. It isn't a bug as such, because the runtime pm core doesn't decrement the reference count below zero, but it is missleading so let's drop it.
Using pm_runtime_resume_and_get() new call makes it easy to handle failures in resume as it doesn't hold a reference count if it exits with an error.
--- drivers/iio/pressure/icp10100.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/iio/pressure/icp10100.c +3 -2
@@ -250,7 +250,9 @@ static int icp10100_get_measures(struct icp10100_state *st,__be16 measures[3 ];int ret; -pm_runtime_get_sync (&st->client->dev); + ret =pm_runtime_resume_and_get (&st->client->dev); +if (ret <0 ) +return ret;mutex_lock (&st->lock); cmd = &icp10100_cmd_measure[st->mode];
@@ -525,7 +527,6 @@ static void icp10100_pm_disable(void *data) {struct device *dev = data; -pm_runtime_put_sync_suspend (dev);pm_runtime_disable (dev); }
Signature · 1 line
2.31.1
Re: [PATCH v2 4/6] iio: prox: pulsed-light-v2: Fix misbalance runtime pm in error path
In reply to Jonathan CameronQuoted 5 lines · Jonathan Cameron
From: Jonathan Cameron <[email protected]>
There is one path in which we don't do a runtime pm put and so leave the device enabled for ever more.
Quoted 33 lines
Reported-by Mauro Carvalho Chehab <[email protected]>Fixes 4ac4e086fd8c ("iio: pulsedlight-lidar-lite: add runtime PM")Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +2
-2@@ -164,7 +164,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE );if (ret <0 ) {dev_err (&client->dev,"cannot send start measurement command" ); -return ret; + goto err; }while (tries--) {@@ -188,6 +188,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) } ret = -EIO ; } +err:pm_runtime_mark_last_busy (&client->dev);pm_runtime_put_autosuspend (&client->dev); --2.31.1
Re: [PATCH v2 3/6] iio: chemical: atlas-sensor: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Jonathan CameronQuoted 12 lines · Jonathan Cameron
From: Jonathan Cameron <[email protected]>
The
pm_runtime_put_noidle()call in remove isn't balanced with any get, so drop it. Note this isn't a bug as the runtime pm core will not allow the reference count to go negative, making this a noop. However, it is confusing to the reader so let's drop it.
pm_runtime_resume_and_get()replacement found using the coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
Quoted 7 lines
As
pm_runtime_resume_and_get()returns <= 0 take advantage of that to change the error checking to if (ret) which is more in keeping with the rest of this driver.This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Quoted 49 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>--- drivers/iio/chemical/atlas-sensor.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-)drivers/iio/chemical/atlas-sensor.c +4
-10@@ -410,11 +410,9 @@ static int atlas_buffer_postenable(struct iio_dev *indio_dev)struct atlas_data *data =iio_priv (indio_dev);int ret; - ret =pm_runtime_get_sync (&data->client->dev); -if (ret <0 ) { -pm_runtime_put_noidle (&data->client->dev); + ret =pm_runtime_resume_and_get (&data->client->dev); +if (ret)return ret; - }return atlas_set_interrupt (data, true); }@@ -487,11 +485,9 @@ static int atlas_read_measurement(struct atlas_data *data, int reg, __be32 *val)int suspended =pm_runtime_suspended (dev);int ret; - ret =pm_runtime_get_sync (dev); -if (ret <0 ) { -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); +if (ret)return ret; - }if (suspended)msleep (data->chip->delay);@@ -741,7 +737,6 @@ static int atlas_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);return atlas_set_powermode (data,0 ); } --2.31.1
Re: [PATCH v2 5/6] iio: prox: pulsed-light-v2: Use pm_runtime_resume_and_get()
In reply to Jonathan CameronQuoted 7 lines · Jonathan Cameron
From: Jonathan Cameron <[email protected]>
Using this new call makes it easy to handle any errors as a result of runtime resume as it exits without leaving the reference count elevated.
Quoted 24 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +3
-2@@ -158,7 +158,9 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg)int tries =10 ;int ret; -pm_runtime_get_sync (&client->dev); + ret =pm_runtime_resume_and_get (&client->dev); +if (ret <0 ) +return ret;/* start sample */ ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE ); --2.31.1
Re: [PATCH v2 1/6] iio: imu: mpu6050: Balance runtime pm + use pm_runtime_resume_and_get()
In reply to Jonathan CameronEm Sun, 16 May 2021 17:20:58 +0100
Jonathan Cameron <[email protected]> escreveu:
Quoted 11 lines
From: Jonathan Cameron <[email protected]>
Remove an unblanced
pm_runtime_put_sync_suspend()call
ininv_pu_pm_disable(). Not this call is not a bug, because the runtime pm core will not allow the reference counter to go negative. It is however confusing and serves no purpose.
pm_runtime_resume_and_get()case found using coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
Quoted 8 lines
pm_runtime_resume_and_get()returns <= 0 only so simplify related checks to bring this more inline with nearby calls.This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Signed-off-by Jonathan Cameron <[email protected]>Cc Jean-Baptiste Maneyrol <[email protected]>
LGTM.
Quoted 67 lines
--- drivers/iio/imu/inv_mpu6050/inv_mpu_core.c | 19 ++++++------------- drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c | 6 ++---- 2 files changed, 8 insertions(+), 17 deletions(-)drivers/iio/imu/inv_mpu6050/inv_mpu_core.c +6
-13@@ -570,11 +570,9 @@ static int inv_mpu6050_read_channel_data(struct iio_dev *indio_dev, freq_hz =INV_MPU6050_DIVIDER_TO_FIFO_RATE (st->chip_config.divider); period_us =1000000 / freq_hz; - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result)return result; - }switch (chan->type) {case IIO_ANGL_VEL :@@ -812,11 +810,9 @@ static int inv_mpu6050_write_raw(struct iio_dev *indio_dev,return result;mutex_lock (&st->lock); - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result) goto error_write_raw_unlock; - }switch (mask) {case IIO_CHAN_INFO_SCALE :@@ -930,11 +926,9 @@ inv_mpu6050_fifo_rate_store(struct device *dev, struct device_attribute *attr, result =0 ; goto fifo_rate_fail_unlock; } - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result) goto fifo_rate_fail_unlock; - } result =regmap_write (st->map, st->reg->sample_rate_div, d);if (result)@@ -1422,7 +1416,6 @@ static void inv_mpu_pm_disable(void *data) {struct device *dev = data; -pm_runtime_put_sync_suspend (dev);pm_runtime_disable (dev); }drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c +0
-0@@ -173,11 +173,9 @@ static int inv_mpu6050_set_enable(struct iio_dev *indio_dev, bool enable)if (enable) { scan =inv_scan_query (indio_dev);
drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c static int inv_mpu6050_set_enable(struct iio_dev *indio_dev, bool enable) · line 181
Quoted 10 lines
drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c +2
-4- result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result)return result; - }/* * In case autosuspend didn't trigger, turn off first not * required sensors.
Thanks,
Mauro
Re: [PATCH v2 2/6] iio: adc: ads1015: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Jonathan CameronEm Sun, 16 May 2021 17:20:59 +0100
Jonathan Cameron <[email protected]> escreveu:
Quoted 7 lines
From: Jonathan Cameron <[email protected]>
The call to
pm_runtime_put_noidle()inremove()is not balancing a counter increment. Note this doesn't matter as the runtime pm core will not allow the counter to go negative. However, it is confusing to the reader so let's remove it.
Quoted 8 lines
The
pm_runtime_resume_and_get()replacement was found using coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Signed-off-by Jonathan Cameron <[email protected]>
LGTM.
Quoted 20 lines
--- drivers/iio/adc/ti-ads1015.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)drivers/iio/adc/ti-ads1015.c +1
-3@@ -323,9 +323,7 @@ static int ads1015_set_power_state(struct ads1015_data *data, bool on)struct device *dev =regmap_get_device (data->regmap);if (on) { - ret =pm_runtime_get_sync (dev); -if (ret <0 ) -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); }else {pm_runtime_mark_last_busy (dev); ret =pm_runtime_put_autosuspend (dev);
drivers/iio/adc/ti-ads1015.c static int ads1015_remove(struct i2c_client *client) · line 1073
Quoted 7 lines
drivers/iio/adc/ti-ads1015.c +0
-1@@ -1070,7 +1068,6 @@ static int ads1015_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);/* power down single shot mode */ return ads1015_set_conv_mode (data,ADS1015_SINGLESHOT );
Thanks,
Mauro
Re: [PATCH v2 3/6] iio: chemical: atlas-sensor: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Jonathan CameronEm Sun, 16 May 2021 17:21:00 +0100
Jonathan Cameron <[email protected]> escreveu:
Quoted 15 lines
From: Jonathan Cameron <[email protected]>
The
pm_runtime_put_noidle()call in remove isn't balanced with any get, so drop it. Note this isn't a bug as the runtime pm core will not allow the reference count to go negative, making this a noop. However, it is confusing to the reader so let's drop it.
pm_runtime_resume_and_get()replacement found using the coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/As
pm_runtime_resume_and_get()returns <= 0 take advantage of that to change the error checking to if (ret) which is more in keeping with the rest of this driver.
Quoted 5 lines
This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>
LGTM.
Quoted 37 lines
--- drivers/iio/chemical/atlas-sensor.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-)drivers/iio/chemical/atlas-sensor.c +4
-8@@ -410,11 +410,9 @@ static int atlas_buffer_postenable(struct iio_dev *indio_dev)struct atlas_data *data =iio_priv (indio_dev);int ret; - ret =pm_runtime_get_sync (&data->client->dev); -if (ret <0 ) { -pm_runtime_put_noidle (&data->client->dev); + ret =pm_runtime_resume_and_get (&data->client->dev); +if (ret)return ret; - }return atlas_set_interrupt (data, true); }@@ -487,11 +485,9 @@ static int atlas_read_measurement(struct atlas_data *data, int reg, __be32 *val)int suspended =pm_runtime_suspended (dev);int ret; - ret =pm_runtime_get_sync (dev); -if (ret <0 ) { -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); +if (ret)return ret; - }if (suspended)msleep (data->chip->delay);
drivers/iio/chemical/atlas-sensor.c static int atlas_remove(struct i2c_client *client) · line 742
Quoted 7 lines
drivers/iio/chemical/atlas-sensor.c +0
-1@@ -741,7 +737,6 @@ static int atlas_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);return atlas_set_powermode (data,0 ); }
Thanks,
Mauro
Re: [PATCH v2 6/6] iio: pressure: icp10100: Balance runtime pm + use pm_runtime_resume_and_get()
In reply to Jonathan CameronEm Sun, 16 May 2021 17:21:03 +0100
Jonathan Cameron <[email protected]> escreveu:
Quoted 7 lines
From: Jonathan Cameron <[email protected]>
The
devm_handled runtime pm disable callspm_runtime_put_sync_suspend()which isn't balancing a matching get call. It isn't a bug as such, because the runtime pm core doesn't decrement the reference count below zero, but it is missleading so let's drop it.
Quoted 6 lines
Using
pm_runtime_resume_and_get()new call makes it easy to handle failures in resume as it doesn't hold a reference count if it exits with an error.Signed-off-by Jonathan Cameron <[email protected]>Cc Jean-Baptiste Maneyrol <[email protected]>
LGTM.
Quoted 20 lines
--- drivers/iio/pressure/icp10100.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)drivers/iio/pressure/icp10100.c +3
-1@@ -250,7 +250,9 @@ static int icp10100_get_measures(struct icp10100_state *st,__be16 measures[3 ];int ret; -pm_runtime_get_sync (&st->client->dev); + ret =pm_runtime_resume_and_get (&st->client->dev); +if (ret <0 ) +return ret;mutex_lock (&st->lock); cmd = &icp10100_cmd_measure[st->mode];
drivers/iio/pressure/icp10100.c static void icp10100_pm_disable(void *data) · line 532
Quoted 7 lines
drivers/iio/pressure/icp10100.c +0
-1@@ -525,7 +527,6 @@ static void icp10100_pm_disable(void *data) {struct device *dev = data; -pm_runtime_put_sync_suspend (dev);pm_runtime_disable (dev); }
Thanks,
Mauro
Re: [PATCH v2 4/6] iio: prox: pulsed-light-v2: Fix misbalance runtime pm in error path
In reply to Jonathan CameronEm Sun, 16 May 2021 17:21:01 +0100
Jonathan Cameron <[email protected]> escreveu:
Quoted 9 lines
From: Jonathan Cameron <[email protected]>
There is one path in which we don't do a runtime pm put and so leave the device enabled for ever more.
Reported-by Mauro Carvalho Chehab <[email protected]>Fixes 4ac4e086fd8c ("iio: pulsedlight-lidar-lite: add runtime PM")Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>
I was unable to apply it on the top of next-20210615, although the patch looks good to me.
Quoted 18 lines
--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +1
-1@@ -164,7 +164,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE );if (ret <0 ) {dev_err (&client->dev,"cannot send start measurement command" ); -return ret; + goto err; }while (tries--) {
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c static int lidar_get_measurement(struct lidar_data *data, u16 *reg) · line 194
Quoted 7 lines
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +1
-0@@ -188,6 +188,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) } ret = -EIO ; } +err:pm_runtime_mark_last_busy (&client->dev);pm_runtime_put_autosuspend (&client->dev);
Thanks,
Mauro
Mauro Carvalho Chehab
Reviewed-byRe: [PATCH v2 5/6] iio: prox: pulsed-light-v2: Use pm_runtime_resume_and_get()
In reply to Matt RanostayEm Sun, 16 May 2021 13:01:29 -0700
Matt Ranostay <[email protected]> escreveu:
Quoted 10 lines · Jonathan Cameron
Quoted 7 lines
From: Jonathan Cameron <[email protected]>
Using this new call makes it easy to handle any errors as a result of runtime resume as it exits without leaving the reference count elevated.
Acked-by Matt Ranostay <[email protected]>
LGTM.
Quoted 20 lines
Quoted 20 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +3
-1@@ -158,7 +158,9 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg)int tries =10 ;int ret; -pm_runtime_get_sync (&client->dev); + ret =pm_runtime_resume_and_get (&client->dev); +if (ret <0 ) +return ret;
Quoted 5 lines
Quoted 5 lines
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +0
-1/* start sample */ ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE ); --2.31.1
Thanks,
Mauro
Jonathan Cameron
maintainerRe: [PATCH v2 1/6] iio: imu: mpu6050: Balance runtime pm + use pm_runtime_resume_and_get()
In reply to Mauro Carvalho ChehabQuoted 20 lines · On Wed, 16 Jun 2021 09:15:23 +0200 Mauro Carvalho Chehab…
Em Sun, 16 May 2021 17:20:58 +0100
Jonathan Cameron <[email protected]> escreveu:Quoted 17 lines
From: Jonathan Cameron <[email protected]>
Remove an unblanced
pm_runtime_put_sync_suspend()call
ininv_pu_pm_disable(). Not this call is not a bug, because the runtime pm core will not allow the reference counter to go negative. It is however confusing and serves no purpose.
pm_runtime_resume_and_get()case found using coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
pm_runtime_resume_and_get()returns <= 0 only so simplify related checks to bring this more inline with nearby calls.This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Quoted 6 lines
Quoted 2 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Jean-Baptiste Maneyrol <[email protected]>LGTM.
Reviewed-by Mauro Carvalho Chehab <[email protected]>
Thanks!
Applied to the togreg branch of iio.git and pushed out as testing for 0-day to see if it can find anything we missed.
Jonathan
Quoted 84 lines
Quoted 69 lines
--- drivers/iio/imu/inv_mpu6050/inv_mpu_core.c | 19 ++++++------------- drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c | 6 ++---- 2 files changed, 8 insertions(+), 17 deletions(-)drivers/iio/imu/inv_mpu6050/inv_mpu_core.c +6
-13@@ -570,11 +570,9 @@ static int inv_mpu6050_read_channel_data(struct iio_dev *indio_dev, freq_hz =INV_MPU6050_DIVIDER_TO_FIFO_RATE (st->chip_config.divider); period_us =1000000 / freq_hz; - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result)return result; - }switch (chan->type) {case IIO_ANGL_VEL :@@ -812,11 +810,9 @@ static int inv_mpu6050_write_raw(struct iio_dev *indio_dev,return result;mutex_lock (&st->lock); - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result) goto error_write_raw_unlock; - }switch (mask) {case IIO_CHAN_INFO_SCALE :@@ -930,11 +926,9 @@ inv_mpu6050_fifo_rate_store(struct device *dev, struct device_attribute *attr, result =0 ; goto fifo_rate_fail_unlock; } - result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result) goto fifo_rate_fail_unlock; - } result =regmap_write (st->map, st->reg->sample_rate_div, d);if (result)@@ -1422,7 +1416,6 @@ static void inv_mpu_pm_disable(void *data) {struct device *dev = data; -pm_runtime_put_sync_suspend (dev);pm_runtime_disable (dev); }drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c +0
-0@@ -173,11 +173,9 @@ static int inv_mpu6050_set_enable(struct iio_dev *indio_dev, bool enable)if (enable) { scan =inv_scan_query (indio_dev);drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c static int inv_mpu6050_set_enable(struct iio_dev *indio_dev, bool enable) · line 181
Quoted 10 lines
drivers/iio/imu/inv_mpu6050/inv_mpu_trigger.c +2
-4- result =pm_runtime_get_sync (pdev); -if (result <0 ) { -pm_runtime_put_noidle (pdev); + result =pm_runtime_resume_and_get (pdev); +if (result)return result; - }/* * In case autosuspend didn't trigger, turn off first not * required sensors. Thanks,
Mauro
Re: [PATCH v2 2/6] iio: adc: ads1015: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Mauro Carvalho ChehabQuoted 14 lines · On Wed, 16 Jun 2021 09:15:55 +0200 Mauro Carvalho Chehab…
Em Sun, 16 May 2021 17:20:59 +0100
Jonathan Cameron <[email protected]> escreveu:Quoted 11 lines
From: Jonathan Cameron <[email protected]>
The call to
pm_runtime_put_noidle()inremove()is not balancing a counter increment. Note this doesn't matter as the runtime pm core will not allow the counter to go negative. However, it is confusing to the reader so let's remove it.The
pm_runtime_resume_and_get()replacement was found using coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/
Quoted 8 lines
Quoted 4 lines
This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Signed-off-by Jonathan Cameron <[email protected]>LGTM.
Reviewed-by Mauro Carvalho Chehab <[email protected]>
Applied
Quoted 33 lines
Quoted 21 lines
--- drivers/iio/adc/ti-ads1015.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)drivers/iio/adc/ti-ads1015.c +1
-3@@ -323,9 +323,7 @@ static int ads1015_set_power_state(struct ads1015_data *data, bool on)struct device *dev =regmap_get_device (data->regmap);if (on) { - ret =pm_runtime_get_sync (dev); -if (ret <0 ) -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); }else {pm_runtime_mark_last_busy (dev); ret =pm_runtime_put_autosuspend (dev);drivers/iio/adc/ti-ads1015.c static int ads1015_remove(struct i2c_client *client) · line 1073
Quoted 7 lines
drivers/iio/adc/ti-ads1015.c +0
-1@@ -1070,7 +1068,6 @@ static int ads1015_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);/* power down single shot mode */ return ads1015_set_conv_mode (data,ADS1015_SINGLESHOT );Thanks,
Mauro
Re: [PATCH v2 3/6] iio: chemical: atlas-sensor: Balance runtime pm + pm_runtime_resume_and_get()
In reply to Mauro Carvalho ChehabQuoted 21 lines · On Wed, 16 Jun 2021 09:16:20 +0200 Mauro Carvalho Chehab…
Em Sun, 16 May 2021 17:21:00 +0100
Jonathan Cameron <[email protected]> escreveu:Quoted 18 lines
From: Jonathan Cameron <[email protected]>
The
pm_runtime_put_noidle()call in remove isn't balanced with any get, so drop it. Note this isn't a bug as the runtime pm core will not allow the reference count to go negative, making this a noop. However, it is confusing to the reader so let's drop it.
pm_runtime_resume_and_get()replacement found using the coccicheck script under review at:
https://lore.kernel.org/lkml/[email protected]/As
pm_runtime_resume_and_get()returns <= 0 take advantage of that to change the error checking to if (ret) which is more in keeping with the rest of this driver.This is a prequel to taking a closer look at the runtime pm in IIO drivers in general.
Quoted 6 lines
Quoted 2 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>LGTM.
Reviewed-by Mauro Carvalho Chehab <[email protected]>
Applied
Quoted 50 lines
Quoted 38 lines
--- drivers/iio/chemical/atlas-sensor.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-)drivers/iio/chemical/atlas-sensor.c +4
-8@@ -410,11 +410,9 @@ static int atlas_buffer_postenable(struct iio_dev *indio_dev)struct atlas_data *data =iio_priv (indio_dev);int ret; - ret =pm_runtime_get_sync (&data->client->dev); -if (ret <0 ) { -pm_runtime_put_noidle (&data->client->dev); + ret =pm_runtime_resume_and_get (&data->client->dev); +if (ret)return ret; - }return atlas_set_interrupt (data, true); }@@ -487,11 +485,9 @@ static int atlas_read_measurement(struct atlas_data *data, int reg, __be32 *val)int suspended =pm_runtime_suspended (dev);int ret; - ret =pm_runtime_get_sync (dev); -if (ret <0 ) { -pm_runtime_put_noidle (dev); + ret =pm_runtime_resume_and_get (dev); +if (ret)return ret; - }if (suspended)msleep (data->chip->delay);drivers/iio/chemical/atlas-sensor.c static int atlas_remove(struct i2c_client *client) · line 742
Quoted 7 lines
drivers/iio/chemical/atlas-sensor.c +0
-1@@ -741,7 +737,6 @@ static int atlas_remove(struct i2c_client *client)pm_runtime_disable (&client->dev);pm_runtime_set_suspended (&client->dev); -pm_runtime_put_noidle (&client->dev);return atlas_set_powermode (data,0 ); }Thanks,
Mauro
Jonathan Cameron
maintainerRe: [PATCH v2 4/6] iio: prox: pulsed-light-v2: Fix misbalance runtime pm in error path
In reply to Mauro Carvalho ChehabQuoted 13 lines · On Wed, 16 Jun 2021 09:18:42 +0200 Mauro Carvalho Chehab…
Em Sun, 16 May 2021 17:21:01 +0100
Jonathan Cameron <[email protected]> escreveu:Quoted 9 lines
From: Jonathan Cameron <[email protected]>
There is one path in which we don't do a runtime pm put and so leave the device enabled for ever more.
Reported-by Mauro Carvalho Chehab <[email protected]>Fixes 4ac4e086fd8c ("iio: pulsedlight-lidar-lite: add runtime PM")Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>
Quoted 4 lines
I was unable to apply it on the top of next-20210615, although the patch looks good to me.
Reviewed-by Mauro Carvalho Chehab <[email protected]>
Ah. This crossed with another fix for the same issue.
I'll take another look at whether the alternate fix works as well as this and revisit this one at a future date.
Thanks,
Jonathan
Quoted 31 lines
Quoted 19 lines
--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +1
-1@@ -164,7 +164,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE );if (ret <0 ) {dev_err (&client->dev,"cannot send start measurement command" ); -return ret; + goto err; }while (tries--) {drivers/iio/proximity/pulsedlight-lidar-lite-v2.c static int lidar_get_measurement(struct lidar_data *data, u16 *reg) · line 194
Quoted 7 lines
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +1
-0@@ -188,6 +188,7 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg) } ret = -EIO ; } +err:pm_runtime_mark_last_busy (&client->dev);pm_runtime_put_autosuspend (&client->dev);Thanks,
Mauro
Re: [PATCH v2 5/6] iio: prox: pulsed-light-v2: Use pm_runtime_resume_and_get()
In reply to Mauro Carvalho ChehabQuoted 12 lines · On Wed, 16 Jun 2021 09:19:13 +0200 Mauro Carvalho Chehab…
Em Sun, 16 May 2021 13:01:29 -0700
Matt Ranostay <[email protected]> escreveu:Quoted 9 lines · Jonathan Cameron
Quoted 7 lines
From: Jonathan Cameron <[email protected]>
Using this new call makes it easy to handle any errors as a result of runtime resume as it exits without leaving the reference count elevated.
Quoted 5 lines
Quoted 1 line
Acked-by Matt Ranostay <[email protected]>LGTM.
Reviewed-by Mauro Carvalho Chehab <[email protected]>
Applied.
Quoted 31 lines
Quoted 21 lines
Quoted 21 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Matt Ranostay <[email protected]>--- drivers/iio/proximity/pulsedlight-lidar-lite-v2.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +3
-1@@ -158,7 +158,9 @@ static int lidar_get_measurement(struct lidar_data *data, u16 *reg)int tries =10 ;int ret; -pm_runtime_get_sync (&client->dev); + ret =pm_runtime_resume_and_get (&client->dev); +if (ret <0 ) +return ret;Quoted 5 lines
Quoted 5 lines
drivers/iio/proximity/pulsedlight-lidar-lite-v2.c +0
-1/* start sample */ ret =lidar_write_control (data,LIDAR_REG_CONTROL_ACQUIRE ); --2.31.1
Thanks,
Mauro
Re: [PATCH v2 6/6] iio: pressure: icp10100: Balance runtime pm + use pm_runtime_resume_and_get()
In reply to Mauro Carvalho ChehabQuoted 14 lines · On Wed, 16 Jun 2021 09:16:38 +0200 Mauro Carvalho Chehab…
Em Sun, 16 May 2021 17:21:03 +0100
Jonathan Cameron <[email protected]> escreveu:Quoted 11 lines
From: Jonathan Cameron <[email protected]>
The
devm_handled runtime pm disable callspm_runtime_put_sync_suspend()which isn't balancing a matching get call. It isn't a bug as such, because the runtime pm core doesn't decrement the reference count below zero, but it is missleading so let's drop it.Using
pm_runtime_resume_and_get()new call makes it easy to handle failures in resume as it doesn't hold a reference count if it exits with an error.
Quoted 6 lines
Quoted 2 lines
Signed-off-by Jonathan Cameron <[email protected]>Cc Jean-Baptiste Maneyrol <[email protected]>LGTM.
Reviewed-by Mauro Carvalho Chehab <[email protected]>
Applied.
Quoted 33 lines
Quoted 21 lines
--- drivers/iio/pressure/icp10100.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)drivers/iio/pressure/icp10100.c +3
-1@@ -250,7 +250,9 @@ static int icp10100_get_measures(struct icp10100_state *st,__be16 measures[3 ];int ret; -pm_runtime_get_sync (&st->client->dev); + ret =pm_runtime_resume_and_get (&st->client->dev); +if (ret <0 ) +return ret;mutex_lock (&st->lock); cmd = &icp10100_cmd_measure[st->mode];drivers/iio/pressure/icp10100.c static void icp10100_pm_disable(void *data) · line 532
Quoted 7 lines
drivers/iio/pressure/icp10100.c +0
-1@@ -525,7 +527,6 @@ static void icp10100_pm_disable(void *data) {struct device *dev = data; -pm_runtime_put_sync_suspend (dev);pm_runtime_disable (dev); }Thanks,
Mauro