hello,

On Tue, Aug 21, 2012 at 05:59:00AM +0000, R, Durgadoss wrote:

Hi Eduardo,

quoted

quoted

+static long get_target_state(struct thermal_zone_device *tz,
+		struct thermal_cooling_device *cdev, int weight, int level)
+{
+	unsigned long max_state;
+
+	cdev->ops->get_max_state(cdev, &max_state);
+
+	return (long)(weight * level * max_state) / (100 * tz->trips);
+}
+
+static void thermal_cdev_update(struct thermal_cooling_device *cdev)
+{
+	struct thermal_instance *instance;
+	unsigned long target = 0;
+
+	mutex_lock(&cdev->lock);
+
+	/* Make sure cdev enters the deepest cooling state */
+	list_for_each_entry(instance, &cdev->thermal_instances,

cdev_node) {

quoted

+		if (instance->target > target)
+			target = instance->target;
+	}
+
+	mutex_unlock(&cdev->lock);
+
+	cdev->ops->set_cur_state(cdev, target);
+}

I believe Rui has already provided an arbitrator, can we reuse it here as well?

I thought about this. I was of the opinion a governor can choose to do arbitration
in its own specific way. For example, The fair_share does not check 'cdev->updated'
value whenever it tries to update the state of a cooling device.

I see. the arbitration starts to get more and more complex. not to talk about constraints
coming from performance domain and not thermal.

I'd still try to find a central point for doing the arbitration, this way we grow the code
in a scalable way...

Thanks,
Durga