thermal: core: Synchronize suspend-prepare and post-suspend actions

BugLink: https://bugs.launchpad.net/bugs/2101915

After commit 5a5efdaffd ("thermal: core: Resume thermal zones
asynchronously") it is theoretically possible that, if a system suspend
starts immediately after a system resume, thermal_zone_device_resume()
spawned by the thermal PM notifier for one of the thermal zones at the
end of the system resume will run after the PM thermal notifier for the
suspend-prepare action.  If that happens, tz->suspended set by the latter
will be reset by the former which may lead to unexpected consequences.

To avoid that race, synchronize thermal_zone_device_resume() with the
suspend-prepare thermal PM notifier with the help of additional bool
field and completion in struct thermal_zone_device.

Note that this also ensures running __thermal_zone_device_update() at
least once for each thermal zone between system resume and the following
system suspend in case it is needed to start thermal mitigation.

Fixes: 5a5efdaffd ("thermal: core: Resume thermal zones asynchronously")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
(cherry picked from commit d2278f3533a8c4933c52f85784ffa73e8250c524)
[koichiroden: Pulled from v6.10 as a dependency for the subsequent
commit, applied changes for drivers/thermal/thermal_core.h to
include/linux/thermal.h due to the missing backport of:
b1ae92dcfa8e ("thermal: core: Make struct thermal_zone_device definition internal")
with adjustments due to the absent commit:
9b0a62758665 ("thermal: core: Store zone trips table in struct thermal_zone_device").
N.B. this also fixes a K6.8+ issue]
Signed-off-by: Koichiro Den <koichiro.den@canonical.com>
Signed-off-by: Stefan Bader <stefan.bader@canonical.com>

drivers/thermal/thermal_core.c

@@ -1332,6 +1332,7 @@ thermal_zone_device_register_with_trips(const char *type, struct thermal_trip *t
 	ida_init(&tz->ida);
 	mutex_init(&tz->lock);
 	init_completion(&tz->removal);
+	init_completion(&tz->resume);
 	id = ida_alloc(&thermal_tz_ida, GFP_KERNEL);
 	if (id < 0) {
 		result = id;
@@ -1577,6 +1578,9 @@ static void thermal_zone_device_resume(struct work_struct *work)
 	thermal_zone_device_init(tz);
 	__thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
 
+	complete(&tz->resume);
+	tz->resuming = false;
+
 	mutex_unlock(&tz->lock);
 }
 
@@ -1594,6 +1598,20 @@ static int thermal_pm_notify(struct notifier_block *nb,
 		list_for_each_entry(tz, &thermal_tz_list, node) {
 			mutex_lock(&tz->lock);
 
+			if (tz->resuming) {
+				/*
+				 * thermal_zone_device_resume() queued up for
+				 * this zone has not acquired the lock yet, so
+				 * release it to let the function run and wait
+				 * util it has done the work.
+				 */
+				mutex_unlock(&tz->lock);
+
+				wait_for_completion(&tz->resume);
+
+				mutex_lock(&tz->lock);
+			}
+
 			tz->suspended = true;
 
 			mutex_unlock(&tz->lock);
@@ -1611,6 +1629,9 @@ static int thermal_pm_notify(struct notifier_block *nb,
 
 			cancel_delayed_work(&tz->poll_queue);
 
+			reinit_completion(&tz->resume);
+			tz->resuming = true;
+
 			/*
 			 * Replace the work function with the resume one, which
 			 * will restore the original work function and schedule

include/linux/thermal.h

@@ -125,6 +125,7 @@ struct thermal_cooling_device {
  * @type:	the thermal zone device type
  * @device:	&struct device for this thermal zone
  * @removal:	removal completion
+ * @resume:	resume completion
  * @trip_temp_attrs:	attributes for trip points for sysfs: trip temperature
  * @trip_type_attrs:	attributes for trip points for sysfs: trip type
  * @trip_hyst_attrs:	attributes for trip points for sysfs: trip hysteresis
@@ -160,12 +161,14 @@ struct thermal_cooling_device {
  * @poll_queue:	delayed work for polling
  * @notify_event: Last notification event
  * @suspended: thermal zone suspend indicator
+ * @resuming:	indicates whether or not thermal zone resume is in progress
  */
 struct thermal_zone_device {
 	int id;
 	char type[THERMAL_NAME_LENGTH];
 	struct device device;
 	struct completion removal;
+	struct completion resume;
 	struct attribute_group trips_attribute_group;
 	struct thermal_attr *trip_temp_attrs;
 	struct thermal_attr *trip_type_attrs;
@@ -197,6 +200,7 @@ struct thermal_zone_device {
 	struct thermal_debugfs *debugfs;
 #endif
 	bool suspended;
+	bool resuming;
 };
 
 /**