On Tue, May 28, 2024 at 09:09:28AM +0200, Thomas Weißschuh wrote:
> On 2024-05-28 06:39:25+0000, Tzung-Bi Shih wrote:
> > On Mon, May 27, 2024 at 10:58:32PM +0200, Thomas Weißschuh wrote:
> > > +struct cros_ec_hwmon_priv {
> > > + struct cros_ec_device *cros_ec;
> > > + const char *temp_sensor_names[EC_TEMP_SENSOR_ENTRIES + EC_TEMP_SENSOR_B_ENTRIES];
> > > + u8 usable_fans;
> > > +};
> > > +
> > > +static int cros_ec_hwmon_read_fan_speed(struct cros_ec_device *cros_ec, u8 index, u16 *speed)
> > > +{
> > > + u16 data;
> > > + int ret;
> > > +
> > > + ret = cros_ec_cmd_readmem(cros_ec, EC_MEMMAP_FAN + index * 2, 2, &data);
> > > + if (ret < 0)
> > > + return ret;
> > > +
> > > + data = le16_to_cpu(data);
> > > + *speed = data;
> > > +
> > > + if (data == EC_FAN_SPEED_NOT_PRESENT || data == EC_FAN_SPEED_STALLED)
> > > + return -EIO;
> >
> > `data` could be eliminated; use `*speed` instead.
>
> Then the le16 value would need to be written directly to the out
> parameter. The current usage relies on *speed (sometimes) being set even
> if ret != 0.
>
> (See next response block)
>
> >
> > > +static int cros_ec_hwmon_read(struct device *dev, enum hwmon_sensor_types type,
> > > + u32 attr, int channel, long *val)
> > > +{
> > [...]
> > > + u16 speed = 0;
> > > + u8 temp = 0;
> >
> > They don't need to initialize.
>
> They need to.
>
> The logic
>
> if (ret == -EIO && speed == EC_FAN_SPEED_STALLED)
> ret = 0;
>
> relies on -EIO and a write to speed from cros_ec_hwmon_read_fan_speed().
>
> But if cros_ec_cmd_readmem() already returns -EIO, then speed would be
> uninitialized.
I see.
