diff mbox series

[v3,13/79] media: s5p: fix pm_runtime_get_sync() usage count

Message ID fd173b0ac00a31630bc60edaf47c2231970a87ed.1619519080.git.mchehab+huawei@kernel.org
State Superseded
Headers show
Series Address some issues with PM runtime at media subsystem | expand

Commit Message

Mauro Carvalho Chehab April 27, 2021, 10:26 a.m. UTC
The pm_runtime_get_sync() internally increments the
dev->power.usage_count without decrementing it, even on errors.
Replace it by the new pm_runtime_resume_and_get(), introduced by:
commit dd8088d5a896 ("PM: runtime: Add pm_runtime_resume_and_get to deal with usage counter")
in order to properly decrement the usage counter and avoid memory
leaks.

While here, check if the PM runtime error was caught at
s5p_cec_adap_enable().

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
---
 drivers/media/cec/platform/s5p/s5p_cec.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Comments

Marek Szyprowski April 27, 2021, 10:36 a.m. UTC | #1
On 27.04.2021 12:26, Mauro Carvalho Chehab wrote:
> The pm_runtime_get_sync() internally increments the
> dev->power.usage_count without decrementing it, even on errors.
> Replace it by the new pm_runtime_resume_and_get(), introduced by:
> commit dd8088d5a896 ("PM: runtime: Add pm_runtime_resume_and_get to deal with usage counter")
> in order to properly decrement the usage counter and avoid memory
> leaks.
>
> While here, check if the PM runtime error was caught at
> s5p_cec_adap_enable().
>
> Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
> ---
>   drivers/media/cec/platform/s5p/s5p_cec.c | 5 ++++-
>   1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/media/cec/platform/s5p/s5p_cec.c b/drivers/media/cec/platform/s5p/s5p_cec.c
> index 2a3e7ffefe0a..2250c1cbc64e 100644
> --- a/drivers/media/cec/platform/s5p/s5p_cec.c
> +++ b/drivers/media/cec/platform/s5p/s5p_cec.c
> @@ -35,10 +35,13 @@ MODULE_PARM_DESC(debug, "debug level (0-2)");
>   
>   static int s5p_cec_adap_enable(struct cec_adapter *adap, bool enable)
>   {
> +	int ret;
>   	struct s5p_cec_dev *cec = cec_get_drvdata(adap);
>   
>   	if (enable) {
> -		pm_runtime_get_sync(cec->dev);
> +		ret = pm_runtime_resume_and_get(cec->dev);
> +		if (ret < 0)
> +			return ret;
>   
>   		s5p_cec_reset(cec);
>   

Best regards
Mauro Carvalho Chehab April 28, 2021, 7:41 a.m. UTC | #2
Em Tue, 27 Apr 2021 12:51:45 +0200
Sylwester Nawrocki <s.nawrocki@samsung.com> escreveu:

> On 27.04.2021 12:26, Mauro Carvalho Chehab wrote:
> > The pm_runtime_get_sync() internally increments the
> > dev->power.usage_count without decrementing it, even on errors.
> > Replace it by the new pm_runtime_resume_and_get(), introduced by:
> > commit dd8088d5a896 ("PM: runtime: Add pm_runtime_resume_and_get to deal with usage counter")
> > in order to properly decrement the usage counter and avoid memory
> > leaks.
> > 
> > While here, check if the PM runtime error was caught at
> > s5p_cec_adap_enable().
> > 
> > Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
> > ---
> >  drivers/media/cec/platform/s5p/s5p_cec.c | 5 ++++-
> >  1 file changed, 4 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/media/cec/platform/s5p/s5p_cec.c b/drivers/media/cec/platform/s5p/s5p_cec.c
> > index 2a3e7ffefe0a..2250c1cbc64e 100644
> > --- a/drivers/media/cec/platform/s5p/s5p_cec.c
> > +++ b/drivers/media/cec/platform/s5p/s5p_cec.c
> > @@ -35,10 +35,13 @@ MODULE_PARM_DESC(debug, "debug level (0-2)");
> >  
> >  static int s5p_cec_adap_enable(struct cec_adapter *adap, bool enable)
> >  {
> > +	int ret;
> >  	struct s5p_cec_dev *cec = cec_get_drvdata(adap);
> >  
> >  	if (enable) {
> > -		pm_runtime_get_sync(cec->dev);
> > +		ret = pm_runtime_resume_and_get(cec->dev);
> > +		if (ret < 0)
> > +			return ret;
> >  
> >  		s5p_cec_reset(cec);  
> 
> Reviewed-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
> 
> Not related to this patch, it seems there is bug in the second
> 'if (enable)' branch, where pm_runtime_disable() is used
> instead of pm_runtime_put(_sync)().

Yeah. I'll add an additional patch before this series in order to
fix the bug. Thanks!

Regards,
Mauro



Thanks,
Mauro
diff mbox series

Patch

diff --git a/drivers/media/cec/platform/s5p/s5p_cec.c b/drivers/media/cec/platform/s5p/s5p_cec.c
index 2a3e7ffefe0a..2250c1cbc64e 100644
--- a/drivers/media/cec/platform/s5p/s5p_cec.c
+++ b/drivers/media/cec/platform/s5p/s5p_cec.c
@@ -35,10 +35,13 @@  MODULE_PARM_DESC(debug, "debug level (0-2)");
 
 static int s5p_cec_adap_enable(struct cec_adapter *adap, bool enable)
 {
+	int ret;
 	struct s5p_cec_dev *cec = cec_get_drvdata(adap);
 
 	if (enable) {
-		pm_runtime_get_sync(cec->dev);
+		ret = pm_runtime_resume_and_get(cec->dev);
+		if (ret < 0)
+			return ret;
 
 		s5p_cec_reset(cec);