| Message ID | dfcaf3d23ea35655d8b3a9254ee00f40c20c95c4.1708692946.git.u.kleine-koenig@pengutronix.de |
|---|---|
| State | Accepted |
| Commit | ad9be163c5ba80e3e923be2b31c70d2c6fa27138 |
| Headers | show |
| Series | media: Convert to platform remove callback returning void | expand |
Hi, maybe its already been discussed but ... Le vendredi 23 février 2024 à 13:59 +0100, Uwe Kleine-König a écrit : > The .remove() callback for a platform driver returns an int which makes > many driver authors wrongly assume it's possible to do error handling by > returning an error code. However the value returned is ignored (apart > from emitting a warning) and this typically results in resource leaks. > > To improve here there is a quest to make the remove callback return > void. In the first step of this quest all drivers are converted to > .remove_new(), which already returns void. Eventually after all drivers > are converted, .remove_new() will be renamed to .remove(). > > Trivially convert this driver from always returning zero in the remove > callback to the void returning variant. > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > --- > drivers/media/platform/chips-media/wave5/wave5-vpu.c | 6 ++---- > 1 file changed, 2 insertions(+), 4 deletions(-) > > diff --git a/drivers/media/platform/chips-media/wave5/wave5-vpu.c b/drivers/media/platform/chips-media/wave5/wave5-vpu.c > index 0d90b5820bef..1b3df5b04249 100644 > --- a/drivers/media/platform/chips-media/wave5/wave5-vpu.c > +++ b/drivers/media/platform/chips-media/wave5/wave5-vpu.c > @@ -250,7 +250,7 @@ static int wave5_vpu_probe(struct platform_device *pdev) > return ret; > } > > -static int wave5_vpu_remove(struct platform_device *pdev) > +static void wave5_vpu_remove(struct platform_device *pdev) > { > struct vpu_device *dev = dev_get_drvdata(&pdev->dev); > > @@ -262,8 +262,6 @@ static int wave5_vpu_remove(struct platform_device *pdev) > v4l2_device_unregister(&dev->v4l2_dev); > wave5_vdi_release(&pdev->dev); > ida_destroy(&dev->inst_ida); > - > - return 0; > } > > static const struct wave5_match_data ti_wave521c_data = { > @@ -283,7 +281,7 @@ static struct platform_driver wave5_vpu_driver = { > .of_match_table = of_match_ptr(wave5_dt_ids), > }, > .probe = wave5_vpu_probe, > - .remove = wave5_vpu_remove, > + .remove_new = wave5_vpu_remove, What if we have another screw up ? remove_new_new ? Why not just remove the int across the tree instead ? Or perhaps its a two step plane ? Nicolas > }; > > module_platform_driver(wave5_vpu_driver);
Le vendredi 23 février 2024 à 13:19 -0500, Nicolas Dufresne a écrit : > Hi, > > maybe its already been discussed but ... > > Le vendredi 23 février 2024 à 13:59 +0100, Uwe Kleine-König a écrit : > > The .remove() callback for a platform driver returns an int which makes > > many driver authors wrongly assume it's possible to do error handling by > > returning an error code. However the value returned is ignored (apart > > from emitting a warning) and this typically results in resource leaks. > > > > To improve here there is a quest to make the remove callback return > > void. In the first step of this quest all drivers are converted to > > .remove_new(), which already returns void. Eventually after all drivers > > are converted, .remove_new() will be renamed to .remove(). Please ignore my comment, sorry for the noise. Nicolas > > > > Trivially convert this driver from always returning zero in the remove > > callback to the void returning variant. > > > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > --- > > drivers/media/platform/chips-media/wave5/wave5-vpu.c | 6 ++---- > > 1 file changed, 2 insertions(+), 4 deletions(-) > > > > diff --git a/drivers/media/platform/chips-media/wave5/wave5-vpu.c b/drivers/media/platform/chips-media/wave5/wave5-vpu.c > > index 0d90b5820bef..1b3df5b04249 100644 > > --- a/drivers/media/platform/chips-media/wave5/wave5-vpu.c > > +++ b/drivers/media/platform/chips-media/wave5/wave5-vpu.c > > @@ -250,7 +250,7 @@ static int wave5_vpu_probe(struct platform_device *pdev) > > return ret; > > } > > > > -static int wave5_vpu_remove(struct platform_device *pdev) > > +static void wave5_vpu_remove(struct platform_device *pdev) > > { > > struct vpu_device *dev = dev_get_drvdata(&pdev->dev); > > > > @@ -262,8 +262,6 @@ static int wave5_vpu_remove(struct platform_device *pdev) > > v4l2_device_unregister(&dev->v4l2_dev); > > wave5_vdi_release(&pdev->dev); > > ida_destroy(&dev->inst_ida); > > - > > - return 0; > > } > > > > static const struct wave5_match_data ti_wave521c_data = { > > @@ -283,7 +281,7 @@ static struct platform_driver wave5_vpu_driver = { > > .of_match_table = of_match_ptr(wave5_dt_ids), > > }, > > .probe = wave5_vpu_probe, > > - .remove = wave5_vpu_remove, > > + .remove_new = wave5_vpu_remove, > > What if we have another screw up ? remove_new_new ? Why not just remove the int > across the tree instead ? Or perhaps its a two step plane ? > > Nicolas > > > }; > > > > module_platform_driver(wave5_vpu_driver); >
Hello Nicolas, [reordered the mail a bit to give understandable context.] On Fri, Feb 23, 2024 at 01:20:59PM -0500, Nicolas Dufresne wrote: > Le vendredi 23 février 2024 à 13:19 -0500, Nicolas Dufresne a écrit : > > maybe its already been discussed but ... > > > > Le vendredi 23 février 2024 à 13:59 +0100, Uwe Kleine-König a écrit : > > > The .remove() callback for a platform driver returns an int which makes > > > many driver authors wrongly assume it's possible to do error handling by > > > returning an error code. However the value returned is ignored (apart > > > from emitting a warning) and this typically results in resource leaks. > > > > > > To improve here there is a quest to make the remove callback return > > > void. In the first step of this quest all drivers are converted to > > > .remove_new(), which already returns void. Eventually after all drivers > > > are converted, .remove_new() will be renamed to .remove(). > > > > What if we have another screw up ? remove_new_new ? Why not just remove the int > > across the tree instead ? Or perhaps its a two step plane ? > > Please ignore my comment, sorry for the noise. Is that an Ack? :-) Best regards Uwe
Hi, Le samedi 24 février 2024 à 09:52 +0100, Uwe Kleine-König a écrit : > Hello Nicolas, > > [reordered the mail a bit to give understandable context.] > > On Fri, Feb 23, 2024 at 01:20:59PM -0500, Nicolas Dufresne wrote: > > Le vendredi 23 février 2024 à 13:19 -0500, Nicolas Dufresne a écrit : > > > maybe its already been discussed but ... > > > > > > Le vendredi 23 février 2024 à 13:59 +0100, Uwe Kleine-König a écrit : > > > > The .remove() callback for a platform driver returns an int which makes > > > > many driver authors wrongly assume it's possible to do error handling by > > > > returning an error code. However the value returned is ignored (apart > > > > from emitting a warning) and this typically results in resource leaks. > > > > > > > > To improve here there is a quest to make the remove callback return > > > > void. In the first step of this quest all drivers are converted to > > > > .remove_new(), which already returns void. Eventually after all drivers > > > > are converted, .remove_new() will be renamed to .remove(). > > > > > > What if we have another screw up ? remove_new_new ? Why not just remove the int > > > across the tree instead ? Or perhaps its a two step plane ? > > > > Please ignore my comment, sorry for the noise. > > Is that an Ack? :-) You have my: Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Ideally I would have liked to see Nas or Jackson sends their Ack as they are maintaining that driver. But I'll leave that to Sebastian and Hans to decide. Nicolas > > Best regards > Uwe >
diff --git a/drivers/media/platform/chips-media/wave5/wave5-vpu.c b/drivers/media/platform/chips-media/wave5/wave5-vpu.c index 0d90b5820bef..1b3df5b04249 100644 --- a/drivers/media/platform/chips-media/wave5/wave5-vpu.c +++ b/drivers/media/platform/chips-media/wave5/wave5-vpu.c @@ -250,7 +250,7 @@ static int wave5_vpu_probe(struct platform_device *pdev) return ret; } -static int wave5_vpu_remove(struct platform_device *pdev) +static void wave5_vpu_remove(struct platform_device *pdev) { struct vpu_device *dev = dev_get_drvdata(&pdev->dev); @@ -262,8 +262,6 @@ static int wave5_vpu_remove(struct platform_device *pdev) v4l2_device_unregister(&dev->v4l2_dev); wave5_vdi_release(&pdev->dev); ida_destroy(&dev->inst_ida); - - return 0; } static const struct wave5_match_data ti_wave521c_data = { @@ -283,7 +281,7 @@ static struct platform_driver wave5_vpu_driver = { .of_match_table = of_match_ptr(wave5_dt_ids), }, .probe = wave5_vpu_probe, - .remove = wave5_vpu_remove, + .remove_new = wave5_vpu_remove, }; module_platform_driver(wave5_vpu_driver);
The .remove() callback for a platform driver returns an int which makes many driver authors wrongly assume it's possible to do error handling by returning an error code. However the value returned is ignored (apart from emitting a warning) and this typically results in resource leaks. To improve here there is a quest to make the remove callback return void. In the first step of this quest all drivers are converted to .remove_new(), which already returns void. Eventually after all drivers are converted, .remove_new() will be renamed to .remove(). Trivially convert this driver from always returning zero in the remove callback to the void returning variant. Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> --- drivers/media/platform/chips-media/wave5/wave5-vpu.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)