Hi Guenter,
On Fri, 18 Jul 2014 07:35:21 -0700, Guenter Roeck wrote:
> Temperature limit clamps are applied after converting the temeprature
> from milli-degrees C to degrees C, so the clamp limit needs to be
> specified in degrees C, not milli-degrees C.
>
> vrm is an u8, so the written value needs to be clamped to [0, 255].
I'd rather reject out-of-range VRM values. These are discrete values,
not continuous, so clamping makes no sense.
> Cc: Axel Lin <axel.lin@xxxxxxxxxx>
> Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> ---
> drivers/hwmon/smsc47m192.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/hwmon/smsc47m192.c b/drivers/hwmon/smsc47m192.c
> index efee4c5..e34918a 100644
> --- a/drivers/hwmon/smsc47m192.c
> +++ b/drivers/hwmon/smsc47m192.c
> @@ -86,7 +86,7 @@ static inline u8 IN_TO_REG(unsigned long val, int n)
> */
> static inline s8 TEMP_TO_REG(int val)
> {
> - return clamp_val(SCALE(val, 1, 1000), -128000, 127000);
> + return clamp_val(SCALE(val, 1, 1000), -128, 127);
> }
SCALE could overflow on large values. It would be safer to clamp first
and then scale.
>
> static inline int TEMP_FROM_REG(s8 val)
> @@ -385,7 +385,7 @@ static ssize_t set_vrm(struct device *dev, struct device_attribute *attr,
> if (err)
> return err;
>
> - data->vrm = val;
> + data->vrm = clamp_val(val, 0, 255);
> return count;
> }
> static DEVICE_ATTR(vrm, S_IRUGO | S_IWUSR, show_vrm, set_vrm);
--
Jean Delvare
SUSE L3 Support
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
