Skip to contentSkip to conversation

iio: gyro: ssp_gyro_sensor: Use devm_iio_device_register

PATCH In review Vaishali Thakkar 4 participants 10 messages Sep 14, 2015 – Sep 29, 2015 IIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 iio: gyro: ssp_gyro_sensor: Use devm_iio_device_register

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
drivers/iio/gyro/ssp_gyro_sensor.c +1 -11
index 0a8afdd..ac88de7 100644
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, indio_dev);
 
-	ret = iio_device_register(indio_dev);
+	ret = devm_iio_device_register(&pdev->dev, indio_dev);
 	if (ret < 0)
 		return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-	struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-	iio_device_unregister(indio_dev);
-
-	return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
 	.driver = {
 		.name = SSP_GYROSCOPE_NAME,
 	},
 	.probe = ssp_gyro_probe,
-	.remove = ssp_gyro_remove,
 };
 
 module_platform_driver(ssp_gyro_driver);
Signature · 1 line
1.9.1
maintainer
#1 lore raw
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors common module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a possible null pointer de reference.

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Jonathan

Quoted 40 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
drivers/iio/gyro/ssp_gyro_sensor.c +1 -11
index 0a8afdd..ac88de7 100644
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, indio_dev);
 
-	ret = iio_device_register(indio_dev);
+	ret = devm_iio_device_register(&pdev->dev, indio_dev);
 	if (ret < 0)
 		return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-	struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-	iio_device_unregister(indio_dev);
-
-	return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
 	.driver = {
 		.name = SSP_GYROSCOPE_NAME,
 	},
 	.probe = ssp_gyro_probe,
-	.remove = ssp_gyro_remove,
 };
 
 module_platform_driver(ssp_gyro_driver);
#2 lore raw
Quoted 5 lines · Jonathan Cameron
Quoted 4 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

Quoted 7 lines
Quoted 1 line

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors common module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a possible null pointer de reference.

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func" should stay. Of course we can switch to devm iio register.

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 60 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Jonathan

Quoted 30 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
drivers/iio/gyro/ssp_gyro_sensor.c +1 -9
index 0a8afdd..ac88de7 100644
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, indio_dev);
 
-	ret = iio_device_register(indio_dev);
+	ret = devm_iio_device_register(&pdev->dev, indio_dev);
 	if (ret < 0)
 		return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct platform_device *pdev)
 	return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-	struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-	iio_device_unregister(indio_dev);
-
-	return 0;
-}

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device *pdev) · line 154

Quoted 10 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -2
-
 static struct platform_driver ssp_gyro_driver = {
 	.driver = {
 		.name = SSP_GYROSCOPE_NAME,
 	},
 	.probe = ssp_gyro_probe,
-	.remove = ssp_gyro_remove,
 };
 
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw
Quoted 13 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

Quoted 6 lines

possible

Quoted 1 line

null pointer de reference.

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

Quoted 70 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
 	platform_set_drvdata(pdev, indio_dev);
 
-	ret = iio_device_register(indio_dev);
+	ret = devm_iio_device_register(&pdev->dev, indio_dev);
 	if (ret < 0)
 		return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
 	return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-	struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-	iio_device_unregister(indio_dev);
-
-	return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
 	.driver = {
 		.name = SSP_GYROSCOPE_NAME,
 	},
 	.probe = ssp_gyro_probe,
-	.remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device *pdev) · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 1 line
Sent from my Android device with K-9 Mail. Please excuse my brevity.
#4 lore raw
Quoted 16 lines · Jonathan Cameron
Quoted 13 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

Quoted 7 lines
Quoted 6 lines

possible

Quoted 1 line

null pointer de reference.

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

So better leave it out:
iio_device_unregister(indio_dev) will disable the buffers and through ops disables the sensor and than remove iio dev ptr from the internal table.

Quoted 71 lines
Quoted 70 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
 	platform_set_drvdata(pdev, indio_dev);
 
-	ret = iio_device_register(indio_dev);
+	ret = devm_iio_device_register(&pdev->dev, indio_dev);
 	if (ret < 0)
 		return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
 	return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-	struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-	iio_device_unregister(indio_dev);
-
-	return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
 	.driver = {
 		.name = SSP_GYROSCOPE_NAME,
 	},
 	.probe = ssp_gyro_probe,
-	.remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device *pdev) · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#5 lore raw
Quoted 19 lines · Karol Wrona
Quoted 18 lines · Jonathan Cameron
Quoted 15 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

possible

Quoted 1 line

null pointer de reference.

Quoted 9 lines
Quoted 5 lines
Quoted 4 lines

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

So better leave it out:
iio_device_unregister(indio_dev) will disable the buffers and through ops disables the sensor and than remove iio dev ptr from the internal table.

Sorry for the late reply. Yes, I guess I missed the point that 'ssp_register_consumer' is
used in probe function. And I believe that devm_iio_register is useful only when we
can convert all other functions to their devm counterparts and remove function will go
away.

But then why don't we need ssp_deregister_consumer here in the remove function? Is it automatically handled by iio_device_unregister? I guess Karol tried to explain
the same thing but I am still confused. Same case applies for ssp_sensors/ssp_dev.c.

Quoted 72 lines
Quoted 71 lines
Quoted 70 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
   platform_set_drvdata(pdev, indio_dev);
 
-  ret = iio_device_register(indio_dev);
+  ret = devm_iio_device_register(&pdev->dev, indio_dev);
   if (ret < 0)
           return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
   return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-  struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-  iio_device_unregister(indio_dev);
-
-  return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
   .driver = {
           .name = SSP_GYROSCOPE_NAME,
   },
   .probe = ssp_gyro_probe,
-  .remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device *pdev) · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 1 line
Vaishali
maintainer
#6 lore raw
Quoted 30 lines · Vaishali Thakkar
Quoted 19 lines · Karol Wrona
Quoted 18 lines · Jonathan Cameron
Quoted 15 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

possible

Quoted 1 line

null pointer de reference.

Quoted 9 lines
Quoted 5 lines
Quoted 4 lines

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

So better leave it out:
iio_device_unregister(indio_dev) will disable the buffers and through ops disables the sensor and than remove iio dev ptr from the internal table.

Quoted 7 lines

Sorry for the late reply. Yes, I guess I missed the point that 'ssp_register_consumer' is
used in probe function. And I believe that devm_iio_register is useful only when we
can convert all other functions to their devm counterparts and remove function will go
away.

Yes. Exactly.

Quoted 6 lines

But then why don't we need ssp_deregister_consumer here in the remove function? Is it automatically handled by iio_device_unregister? I guess Karol tried to explain
the same thing but I am still confused. Same case applies for ssp_sensors/ssp_dev.c.

We do indeed need an ssp_deregister_consumer function to be called in the remove. I don't think one currently exists, so that needs fixing.

Quoted 76 lines
Quoted 73 lines
Quoted 72 lines
Quoted 71 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
   platform_set_drvdata(pdev, indio_dev);
 
-  ret = iio_device_register(indio_dev);
+  ret = devm_iio_device_register(&pdev->dev, indio_dev);
   if (ret < 0)
           return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
   return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-  struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-  iio_device_unregister(indio_dev);
-
-  return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
   .driver = {
           .name = SSP_GYROSCOPE_NAME,
   },
   .probe = ssp_gyro_probe,
-  .remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device *pdev) · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#7 lore raw
Quoted 40 lines · Jonathan Cameron
Quoted 30 lines · Vaishali Thakkar
Quoted 19 lines · Karol Wrona
Quoted 18 lines · Jonathan Cameron
Quoted 15 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

possible

Quoted 1 line

null pointer de reference.

Quoted 9 lines
Quoted 5 lines
Quoted 4 lines

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

So better leave it out:
iio_device_unregister(indio_dev) will disable the buffers and through ops disables the sensor and than remove iio dev ptr from the internal table.

Quoted 7 lines

Sorry for the late reply. Yes, I guess I missed the point that 'ssp_register_consumer' is
used in probe function. And I believe that devm_iio_register is useful only when we
can convert all other functions to their devm counterparts and remove function will go
away.

Yes. Exactly.

Quoted 1 line
Quoted 7 lines
Quoted 5 lines

But then why don't we need ssp_deregister_consumer here in the remove function? Is it automatically handled by iio_device_unregister? I guess Karol tried to explain
the same thing but I am still confused. Same case applies for ssp_sensors/ssp_dev.c.

We do indeed need an ssp_deregister_consumer function to be called in the remove. I don't think one currently exists, so that needs fixing.

Ok. Then I'll send patches for both of these files.

Quoted 76 lines
Quoted 75 lines
Quoted 72 lines
Quoted 71 lines
Quoted 70 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
   platform_set_drvdata(pdev, indio_dev);
 
-  ret = iio_device_register(indio_dev);
+  ret = devm_iio_device_register(&pdev->dev, indio_dev);
   if (ret < 0)
           return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
   return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-  struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-  iio_device_unregister(indio_dev);
-
-  return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
   .driver = {
           .name = SSP_GYROSCOPE_NAME,
   },
   .probe = ssp_gyro_probe,
-  .remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 1 line
Vaishali
#8 lore raw
Quoted 46 lines · Vaishali Thakkar
Quoted 45 lines · Jonathan Cameron
Quoted 30 lines · Vaishali Thakkar
Quoted 19 lines · Karol Wrona
Quoted 18 lines · Jonathan Cameron
Quoted 15 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

possible

Quoted 1 line

null pointer de reference.

Quoted 9 lines
Quoted 5 lines
Quoted 4 lines

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

So better leave it out:
iio_device_unregister(indio_dev) will disable the buffers and through ops disables the sensor and than remove iio dev ptr from the internal table.

Quoted 7 lines

Sorry for the late reply. Yes, I guess I missed the point that 'ssp_register_consumer' is
used in probe function. And I believe that devm_iio_register is useful only when we
can convert all other functions to their devm counterparts and remove function will go
away.

Yes. Exactly.

Quoted 6 lines

But then why don't we need ssp_deregister_consumer here in the remove function? Is it automatically handled by iio_device_unregister? I guess Karol tried to explain
the same thing but I am still confused. Same case applies for ssp_sensors/ssp_dev.c.

Quoted 4 lines
Quoted 2 lines

We do indeed need an ssp_deregister_consumer function to be called in the remove. I don't think one currently exists, so that needs fixing.

Ok. Then I'll send patches for both of these files.

Sorry, probably I was not clear enough in my last mail. I mean I'll send patches patches
once we will have something like 'ssp_deregister_consumer'. But is there anyone who
is working on this? Is it possible to introduce devm counterpart of the same? How much will it be useful [taking note that there are only 3-4 files which are using it]?

Quoted 81 lines
Quoted 67 lines
Quoted 67 lines
Quoted 67 lines
Quoted 67 lines
Quoted 67 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
   platform_set_drvdata(pdev, indio_dev);
 
-  ret = iio_device_register(indio_dev);
+  ret = devm_iio_device_register(&pdev->dev, indio_dev);
   if (ret < 0)
           return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
   return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-  struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-  iio_device_unregister(indio_dev);
-
-  return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
   .driver = {
           .name = SSP_GYROSCOPE_NAME,
   },
   .probe = ssp_gyro_probe,
-  .remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 9 lines
Quoted 8 lines
Quoted 5 lines
Quoted 4 lines
Quoted 3 lines
Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

--
Vaishali

Signature · 1 line
Vaishali
#9 lore raw
Quoted 51 lines · Vaishali Thakkar
Quoted 49 lines · Vaishali Thakkar
Quoted 47 lines · Jonathan Cameron
Quoted 30 lines · Vaishali Thakkar
Quoted 19 lines · Karol Wrona
Quoted 18 lines · Jonathan Cameron
Quoted 15 lines · Karol Wrona
Quoted 8 lines · Jonathan Cameron
Quoted 5 lines · Vaishali Thakkar

Use resourced managed function devm_iio_device_register to make error path simpler. To be compatible with the change, the remove function is removed as it is now redundant.

This patch is reasonable, but makes me wonder if there is an issue in the remove path for this driver. It relies on the ssp_sensors

common

Quoted 3 lines

module. That in ssp_spi.c uses the fact ssp_register_consumer has saved the struct iio_dev into a local array in the core driver. I think this means that a remove of this function will leave a

possible

Quoted 1 line

null pointer de reference.

Quoted 9 lines
Quoted 5 lines
Quoted 4 lines

You are right. In this case we need something like ssp_deregister_consumer(...) in ssp_dev.c. So I think that "remove func"
should stay. Of course we can switch to devm iio register.

Not if you want to remove the userspace interface before doing you new function call. Doing so gives a nasty race.

So better leave it out:
iio_device_unregister(indio_dev) will disable the buffers and through ops disables the sensor and than remove iio dev ptr from the internal table.

Quoted 7 lines

Sorry for the late reply. Yes, I guess I missed the point that 'ssp_register_consumer' is
used in probe function. And I believe that devm_iio_register is useful only when we
can convert all other functions to their devm counterparts and remove function will go
away.

Yes. Exactly.

Quoted 6 lines

But then why don't we need ssp_deregister_consumer here in the remove function? Is it automatically handled by iio_device_unregister? I guess Karol tried to explain
the same thing but I am still confused. Same case applies for ssp_sensors/ssp_dev.c.

We do indeed need an ssp_deregister_consumer function to be called in the remove. I don't think one currently exists, so that needs fixing.

Quoted 7 lines
Quoted 1 line

Ok. Then I'll send patches for both of these files.

Sorry, probably I was not clear enough in my last mail. I mean I'll send patches patches
once we will have something like 'ssp_deregister_consumer'. But is there anyone who
is working on this? Is it possible to introduce devm counterpart of the same?

If you want you can simply add ssp_deregister_consumer (to ssp_dev.c) function and use it in the remove (in ssp sensor platform driver). I will test it.

We can use devm_iio_device_register for iio dev itself but better check what is called first - driver remove callback or the devm removes. If we deregister internally the iio device from ssp we should be quite safe.

Quoted 87 lines

How much will it be useful [taking note that there are only 3-4 files which are using it]?

Quoted 81 lines
Quoted 67 lines
Quoted 67 lines
Quoted 67 lines
Quoted 67 lines
Quoted 67 lines

Also the same can (rather should) be done for accelerometer sensor (ssp_accel_sensor.c).

Vaishali, if you want please feel free and send patch.

Quoted 13 lines

Now I suspect that case doesn't actually occur because the relevant device elements are disabled whenever this module is removed. Having said that we might expect an ssp_unregister_consumer function that sets the relevant pointer back to null on removal so as to avoid any possible race conditions around driver removal / reprobing. A spot of defensive programming rather than necessarily a bug to be fixed!

One little process thing. This driver was written by Karol so patches should probably always cc Karol as well as the more general maintainer / reviewers for IIO. Added cc.

Quoted 6 lines

Jonathan

Review

Quoted 5 lines
---
 drivers/iio/gyro/ssp_gyro_sensor.c | 12 +-----------
 1 file changed, 1 insertion(+), 11 deletions(-)
(quoted diff) +0 -0

b/drivers/iio/gyro/ssp_gyro_sensor.c

Quoted 4 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct platform_device

Quoted 4 lines

index 0a8afdd..ac88de7 100644

drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
@@ -134,7 +134,7 @@ static int ssp_gyro_probe(struct platform_device

*pdev)

Quoted 9 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct

Quoted 9 lines
drivers/iio/gyro/ssp_gyro_sensor.c +1 -1
   platform_set_drvdata(pdev, indio_dev);
 
-  ret = iio_device_register(indio_dev);
+  ret = devm_iio_device_register(&pdev->dev, indio_dev);
   if (ret < 0)
           return ret;
 
@@ -144,21 +144,11 @@ static int ssp_gyro_probe(struct

platform_device *pdev)

Quoted 20 lines
Quoted 20 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -10
   return 0;
 }
 
-static int ssp_gyro_remove(struct platform_device *pdev)
-{
-  struct iio_dev *indio_dev = platform_get_drvdata(pdev);
-
-  iio_device_unregister(indio_dev);
-
-  return 0;
-}
-
 static struct platform_driver ssp_gyro_driver = {
   .driver = {
           .name = SSP_GYROSCOPE_NAME,
   },
   .probe = ssp_gyro_probe,
-  .remove = ssp_gyro_remove,
 };
 
Quoted 5 lines

drivers/iio/gyro/ssp_gyro_sensor.c static int ssp_gyro_probe(struct · line 154

Quoted 2 lines
drivers/iio/gyro/ssp_gyro_sensor.c +0 -0
 module_platform_driver(ssp_gyro_driver);

--
To unsubscribe from this list: send the line "unsubscribe linux-iio"

in

Quoted 9 lines
Quoted 8 lines
Quoted 5 lines
Quoted 4 lines
Quoted 3 lines
Quoted 3 lines

the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

--
Vaishali