Restoring the configuration register on device removal has the side effect of also resetting the hysteresis value. This is inconsistent as the other limits are not reset, only hysteresis. So, following the principle of least surprise, preserve the hysteresis value when restoring the configuration register. Signed-off-by: Jean Delvare <khali@xxxxxxxxxxxx> Cc: Guenter Roeck <linux@xxxxxxxxxxxx> --- Changes since v1: * Fix |-for-& thinko. * Use the new definition of JC42_CFG_HYST_MASK, makes the code more readable. drivers/hwmon/jc42.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) --- linux-3.5.orig/drivers/hwmon/jc42.c 2012-07-26 22:04:46.911073234 +0200 +++ linux-3.5/drivers/hwmon/jc42.c 2012-07-26 22:11:44.918891533 +0200 @@ -534,9 +534,16 @@ static int jc42_remove(struct i2c_client struct jc42_data *data = i2c_get_clientdata(client); hwmon_device_unregister(data->hwmon_dev); sysfs_remove_group(&client->dev.kobj, &jc42_group); - if (data->config != data->orig_config) - i2c_smbus_write_word_swapped(client, JC42_REG_CONFIG, - data->orig_config); + + /* Restore original configuration except hysteresis */ + if ((data->config & ~JC42_CFG_HYST_MASK) != + (data->orig_config & ~JC42_CFG_HYST_MASK)) { + int config; + + config = (data->orig_config & ~JC42_CFG_HYST_MASK) + | (data->config & JC42_CFG_HYST_MASK); + i2c_smbus_write_word_swapped(client, JC42_REG_CONFIG, config); + } return 0; } -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors