Messages in this thread | | | Date | Fri, 1 Jul 2022 10:52:57 -0700 | Subject | Re: [PATCH] hwmon: (asus_wmi_sensors) Save a few bytes of memory | From | Guenter Roeck <> |
| |
On 7/1/22 10:47, Christophe JAILLET wrote: > The first 'for' loop of asus_wmi_configure_sensor_setup() only computes > the number and type of sensors that exist in the system. > > Here, the 'temp_sensor' structure is only used to store the data collected > by asus_wmi_sensor_info(). There is no point in using a devm_ variant for > this first allocation and it can be freed just after this initial loop. > > So use kzalloc()/kfree() to save a few bytes of memory that would be kept > allocated for no good reason otherwise.
Then why isn't this just allocated on the stack ? The structure isn't _that_ large.
Guenter
> > Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr> > --- > drivers/hwmon/asus_wmi_sensors.c | 7 +++++-- > 1 file changed, 5 insertions(+), 2 deletions(-) > > diff --git a/drivers/hwmon/asus_wmi_sensors.c b/drivers/hwmon/asus_wmi_sensors.c > index 9e935e34c998..4873edad4996 100644 > --- a/drivers/hwmon/asus_wmi_sensors.c > +++ b/drivers/hwmon/asus_wmi_sensors.c > @@ -514,14 +514,16 @@ static int asus_wmi_configure_sensor_setup(struct device *dev, > int i, idx; > int err; > > - temp_sensor = devm_kcalloc(dev, 1, sizeof(*temp_sensor), GFP_KERNEL); > + temp_sensor = kzalloc(sizeof(*temp_sensor), GFP_KERNEL); > if (!temp_sensor) > return -ENOMEM; > > for (i = 0; i < sensor_data->wmi.sensor_count; i++) { > err = asus_wmi_sensor_info(i, temp_sensor); > - if (err) > + if (err) { > + kfree(temp_sensor); > return err; > + } > > switch (temp_sensor->data_type) { > case TEMPERATURE_C: > @@ -536,6 +538,7 @@ static int asus_wmi_configure_sensor_setup(struct device *dev, > break; > } > } > + kfree(temp_sensor); > > if (nr_count[hwmon_temp]) > nr_count[hwmon_chip]++, nr_types++;
| |