[PATCH 1/2] ALSA: pcmtest: Convert to platform remove callback returning void
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() is 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 --- sound/drivers/pcmtest.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/sound/drivers/pcmtest.c b/sound/drivers/pcmtest.c index 2ae912a64d16..1fff85feaf82 100644 --- a/sound/drivers/pcmtest.c +++ b/sound/drivers/pcmtest.c @@ -576,10 +576,9 @@ static int pcmtst_probe(struct platform_device *pdev) return 0; }
-static int pdev_remove(struct platform_device *dev) +static void pdev_remove(struct platform_device *dev) { snd_pcmtst_free(pcmtst); - return 0; }
static struct platform_device pcmtst_pdev = { @@ -589,7 +588,7 @@ static struct platform_device pcmtst_pdev = {
static struct platform_driver pcmtst_pdrv = { .probe = pcmtst_probe, - .remove = pdev_remove, + .remove_new = pdev_remove, .driver = { .name = "pcmtest", },
base-commit: 5133c9e51de41bfa902153888e11add3342ede18
While there is probably only ever a single instance of such a pcmtst device, it's still bad style to use a static variable to store per device data. Make use of platform_get_drvdata() and platform_set_drvdata() which fixes a data corruption if there should be two or more such devices (or this driver is used as a template for another driver).
Signed-off-by: Uwe Kleine-König u.kleine-koenig@pengutronix.de --- sound/drivers/pcmtest.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/sound/drivers/pcmtest.c b/sound/drivers/pcmtest.c index 1fff85feaf82..291e7fe47893 100644 --- a/sound/drivers/pcmtest.c +++ b/sound/drivers/pcmtest.c @@ -110,8 +110,6 @@ struct pcmtst_buf_iter { struct timer_list timer_instance; };
-static struct pcmtst *pcmtst; - static struct snd_pcm_hardware snd_pcmtst_hw = { .info = (SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_BLOCK_TRANSFER | @@ -552,6 +550,7 @@ static int snd_pcmtst_create(struct snd_card *card, struct platform_device *pdev static int pcmtst_probe(struct platform_device *pdev) { struct snd_card *card; + struct pcmtst *pcmtst; int err;
err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32)); @@ -573,11 +572,15 @@ static int pcmtst_probe(struct platform_device *pdev) if (err < 0) return err;
+ platform_set_drvdata(pdev, pcmtst); + return 0; }
-static void pdev_remove(struct platform_device *dev) +static void pdev_remove(struct platform_device *pdev) { + struct pcmtst *pcmtst = platform_get_drvdata(pdev); + snd_pcmtst_free(pcmtst); }
On 7/7/23 11:50, Uwe Kleine-König wrote:
While there is probably only ever a single instance of such a pcmtst device, it's still bad style to use a static variable to store per device data. Make use of platform_get_drvdata() and platform_set_drvdata() which fixes a data corruption if there should be two or more such devices (or this driver is used as a template for another driver).
Signed-off-by: Uwe Kleine-König u.kleine-koenig@pengutronix.de
sound/drivers/pcmtest.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/sound/drivers/pcmtest.c b/sound/drivers/pcmtest.c index 1fff85feaf82..291e7fe47893 100644 --- a/sound/drivers/pcmtest.c +++ b/sound/drivers/pcmtest.c @@ -110,8 +110,6 @@ struct pcmtst_buf_iter { struct timer_list timer_instance; };
-static struct pcmtst *pcmtst;
- static struct snd_pcm_hardware snd_pcmtst_hw = { .info = (SNDRV_PCM_INFO_INTERLEAVED | SNDRV_PCM_INFO_BLOCK_TRANSFER |
@@ -552,6 +550,7 @@ static int snd_pcmtst_create(struct snd_card *card, struct platform_device *pdev static int pcmtst_probe(struct platform_device *pdev) { struct snd_card *card;
struct pcmtst *pcmtst; int err;
err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
@@ -573,11 +572,15 @@ static int pcmtst_probe(struct platform_device *pdev) if (err < 0) return err;
- platform_set_drvdata(pdev, pcmtst);
- return 0; }
-static void pdev_remove(struct platform_device *dev) +static void pdev_remove(struct platform_device *pdev) {
- struct pcmtst *pcmtst = platform_get_drvdata(pdev);
- snd_pcmtst_free(pcmtst); }
Also looks good.
Acked-by: Ivan Orlov ivan.orlov0322@gmail.com
On 7/7/23 11:50, Uwe Kleine-König wrote:
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() is 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
sound/drivers/pcmtest.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/sound/drivers/pcmtest.c b/sound/drivers/pcmtest.c index 2ae912a64d16..1fff85feaf82 100644 --- a/sound/drivers/pcmtest.c +++ b/sound/drivers/pcmtest.c @@ -576,10 +576,9 @@ static int pcmtst_probe(struct platform_device *pdev) return 0; }
-static int pdev_remove(struct platform_device *dev) +static void pdev_remove(struct platform_device *dev) { snd_pcmtst_free(pcmtst);
return 0; }
static struct platform_device pcmtst_pdev = {
@@ -589,7 +588,7 @@ static struct platform_device pcmtst_pdev = {
static struct platform_driver pcmtst_pdrv = { .probe = pcmtst_probe,
- .remove = pdev_remove,
- .remove_new = pdev_remove, .driver = { .name = "pcmtest", },
base-commit: 5133c9e51de41bfa902153888e11add3342ede18
Looks good to me, thanks!
On 7/7/23 11:50, Uwe Kleine-König wrote:
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() is 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
sound/drivers/pcmtest.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/sound/drivers/pcmtest.c b/sound/drivers/pcmtest.c index 2ae912a64d16..1fff85feaf82 100644 --- a/sound/drivers/pcmtest.c +++ b/sound/drivers/pcmtest.c @@ -576,10 +576,9 @@ static int pcmtst_probe(struct platform_device *pdev) return 0; }
-static int pdev_remove(struct platform_device *dev) +static void pdev_remove(struct platform_device *dev) { snd_pcmtst_free(pcmtst);
return 0; }
static struct platform_device pcmtst_pdev = {
@@ -589,7 +588,7 @@ static struct platform_device pcmtst_pdev = {
static struct platform_driver pcmtst_pdrv = { .probe = pcmtst_probe,
- .remove = pdev_remove,
- .remove_new = pdev_remove, .driver = { .name = "pcmtest", },
base-commit: 5133c9e51de41bfa902153888e11add3342ede18
Acked-by: Ivan Orlov ivan.orlov0322@gmail.com
On Fri, 07 Jul 2023 09:50:57 +0200, Uwe Kleine-König wrote:
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() is 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
Applied both patches now. Thanks!
Takashi
participants (3)
-
Ivan Orlov
-
Takashi Iwai
-
Uwe Kleine-König