On Wed, Jun 13, 2012 at 05:57:37PM +0200, Jean Delvare wrote: > Hi Guenter, > > On Mon, 4 Jun 2012 08:53:48 -0700, Guenter Roeck wrote: > > Fix: > > > > emc2103.c: In function set_pwm_enable: > > emc2103.c:463:12: warning: conf_reg may be used uninitialized in this function > > > > by checking the return value from read_u8_from_i2c(). This fixes a real problem, > > as conf_reg is really uninitialized if read_u8_from_i2c returns an error. > > > > Cc: Steve Glendinning <steve.glendinning@xxxxxxxx> > > Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx> > > --- > > drivers/hwmon/emc2103.c | 12 ++++++++---- > > 1 files changed, 8 insertions(+), 4 deletions(-) > > Good catch. Still I'm a little worried about... > thanks to those modern compilers ... > > > > diff --git a/drivers/hwmon/emc2103.c b/drivers/hwmon/emc2103.c > > index 9691f66..ff68563 100644 > > --- a/drivers/hwmon/emc2103.c > > +++ b/drivers/hwmon/emc2103.c > > @@ -451,11 +451,13 @@ static ssize_t set_pwm_enable(struct device *dev, struct device_attribute *da, > > data->fan_rpm_control = true; > > break; > > default: > > - mutex_unlock(&data->update_lock); > > - return -EINVAL; > > + result = -EINVAL; > > + goto err; > > } > > > > - read_u8_from_i2c(client, REG_FAN_CONF1, &conf_reg); > > + result = read_u8_from_i2c(client, REG_FAN_CONF1, &conf_reg); > > + if (result) > > + goto err; > > > > if (data->fan_rpm_control) > > conf_reg |= 0x80; > > @@ -464,8 +466,10 @@ static ssize_t set_pwm_enable(struct device *dev, struct device_attribute *da, > > > > i2c_smbus_write_byte_data(client, REG_FAN_CONF1, conf_reg); > > > > + result = count; > > ... this. Is it guaranteed that sizeof(int) >= sizeof(size_t)? Probably > it doesn't matter much in practice, but I'm still wondering. I think I > would prefer if you'd return "count" and stuff error codes in it when > needed, rather than returning "result". > > But you can also just ignore me if you think it's irrelevant. Both > cases, you have my ack. > Hi Jean, good point. sizeof(size_t) _can_ be larger than sizeof(int), and often is. Sure, it doesn't matter much in practice, as count is hopefully never that large, but it is better to be careful. I changed the code as you suggested, ie return count and assign the error code to it if needed. Thanks a lot for the review. Guenter _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
