Hi Eduardo

quoted

--- a/drivers/thermal/rcar_thermal.c
+++ b/drivers/thermal/rcar_thermal.c

@@ -199,9 +199,9 @@ static int rcar_thermal_update_temp(struct rcar_thermal_priv *priv)
 
 	dev_dbg(dev, "thermal%d  %d -> %d\n", priv->id, priv->ctemp, ctemp);
 
-	priv->ctemp = ctemp;
 	ret = 0;
 err_out_unlock:
+	priv->ctemp = ctemp;
 	mutex_unlock(&priv->lock);
 	return ret;

I believe the problem here is actually the lack of error
handling/propagation. Are you sure you want to write to parameter
in the fail path ?

rcar_thermal_update_temp already returns error code when it fails
to read temperature. Don't you think it would make more sense to fix the
places that call rcar_thermal_update_temp to properly handle its return
value and propagate that error code when necessary?

Thanks. But the logic is that
priv->ctemp will be 0 in error cases, and it will be reported as
dev_err() in rcar_thermal_get_current_temp()

The problem was that priv->ctemp will not be updated
in error case of rcar_thermal_update_temp().
but user can call get_temp (it doesn't call
rcar_thermal_update_temp() if interrupt was supported).
In such case, user will get old thermal.