Re: [PATCH 1/3] cpufreq: qcom: fix memory leak in error path

[Viresh Kumar]

On 10-10-22, 11:25, Viresh Kumar wrote:
> On 01-10-22, 19:10, Fabien Parent wrote:
> > If for some reason the speedbin length is incorrect, then there is a
> > memory leak in the error path because we never free the speedbin buffer.
> > This commit fixes the error path to always free the speedbin buffer.
> > 
> > Signed-off-by: Fabien Parent <fabien.parent@xxxxxxxxxx>
> > ---
> >  drivers/cpufreq/qcom-cpufreq-nvmem.c | 7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/cpufreq/qcom-cpufreq-nvmem.c b/drivers/cpufreq/qcom-cpufreq-nvmem.c
> > index 863548f59c3e..3bd38acde4b9 100644
> > --- a/drivers/cpufreq/qcom-cpufreq-nvmem.c
> > +++ b/drivers/cpufreq/qcom-cpufreq-nvmem.c
> > @@ -213,6 +213,7 @@ static int qcom_cpufreq_krait_name_version(struct device *cpu_dev,
> >  	int speed = 0, pvs = 0, pvs_ver = 0;
> >  	u8 *speedbin;
> >  	size_t len;
> > +	int ret = 0;
> >  
> >  	speedbin = nvmem_cell_read(speedbin_nvmem, &len);
> >  
> > @@ -230,7 +231,8 @@ static int qcom_cpufreq_krait_name_version(struct device *cpu_dev,
> >  		break;
> >  	default:
> >  		dev_err(cpu_dev, "Unable to read nvmem data. Defaulting to 0!\n");
> > -		return -ENODEV;
> > +		ret = -ENODEV;
> > +		goto len_error;
> >  	}
> >  
> >  	snprintf(*pvs_name, sizeof("speedXX-pvsXX-vXX"), "speed%d-pvs%d-v%d",
> > @@ -238,8 +240,9 @@ static int qcom_cpufreq_krait_name_version(struct device *cpu_dev,
> >  
> >  	drv->versions = (1 << speed);
> >  
> > +len_error:
> >  	kfree(speedbin);
> > -	return 0;
> > +	return ret;
> >  }
> >  
> >  static const struct qcom_cpufreq_match_data match_data_kryo = {
> 
> Applied. Thanks.

Btw, it will be good to have a Fixes or Cc:Stable for this patch too.
I can directly add the lines myself, just let me know what you want.

-- 
viresh