thermal: core: Fix the handling of invalid trip points

Commit 9ad18043fb35 ("thermal: core: Send trip crossing notifications
at init time if needed") overlooked the case when a trip point that
has started as invalid is set to a valid temperature later.  Namely,
the initial threshold value for all trips is zero, so if a previously
invalid trip becomes valid and its (new) low temperature is above the
zone temperature, a spurious trip crossing notification will occur and
it may trigger the WARN_ON() in handle_thermal_trip().

To address this, set the initial threshold for all trips to INT_MAX.

There is also the case when a valid writable trip becomes invalid that
requires special handling.  First, in accordance with the change
mentioned above, the trip's threshold needs to be set to INT_MAX to
avoid the same issue.  Second, if the trip in question is passive and
it has been crossed by the thermal zone temperature on the way up, the
zone's passive count has been incremented and it is in the passive
polling mode, so its passive count needs to be adjusted to allow the
passive polling to be turned off eventually.

Fixes: 9ad18043fb35 ("thermal: core: Send trip crossing notifications at init time if needed")
Fixes: 042a3d80f118 ("thermal: core: Move passive polling management to the core")
Reported-by: Zhang Rui <zhang.rui@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Wendy Wang <wendy.wang@intel.com>

diff --git a/drivers/thermal/thermal_core.c b/drivers/thermal/thermal_core.c
index 11750a145d74c8d39033e4dc223d22d8b6b8e926..54cce4e523bc510a0b25621317f2a56a31c32e97 100644 (file)
--- a/drivers/thermal/thermal_core.c
+++ b/drivers/thermal/thermal_core.c
@@ -1398,8 +1398,15 @@ thermal_zone_device_register_with_trips(const char *type,
        tz->device.class = thermal_class;
        tz->devdata = devdata;
        tz->num_trips = num_trips;
-       for_each_trip_desc(tz, td)
+       for_each_trip_desc(tz, td) {
                td->trip = *trip++;
+               /*
+                * Mark all thresholds as invalid to start with even though
+                * this only matters for the trips that start as invalid and
+                * become valid later.
+                */
+               td->threshold = INT_MAX;
+       }
 
        thermal_set_delay_jiffies(&tz->passive_delay_jiffies, passive_delay);
        thermal_set_delay_jiffies(&tz->polling_delay_jiffies, polling_delay);

diff --git a/drivers/thermal/thermal_trip.c b/drivers/thermal/thermal_trip.c
index 21ece83999971f695383351e13628dc67b73d30d..d6a6acc78ddb6c1586bd32c878e59a128933464f 100644 (file)
--- a/drivers/thermal/thermal_trip.c
+++ b/drivers/thermal/thermal_trip.c
@@ -152,6 +152,24 @@ void thermal_zone_set_trip_temp(struct thermal_zone_device *tz,
        if (trip->temperature == temp)
                return;
 
+       if (temp == THERMAL_TEMP_INVALID) {
+               struct thermal_trip_desc *td = trip_to_trip_desc(trip);
+
+               if (trip->type == THERMAL_TRIP_PASSIVE &&
+                   tz->temperature >= td->threshold) {
+                       /*
+                        * The trip has been crossed, so the thermal zone's
+                        * passive count needs to be adjusted.
+                        */
+                       tz->passive--;
+                       WARN_ON_ONCE(tz->passive < 0);
+               }
+               /*
+                * Invalidate the threshold to avoid triggering a spurious
+                * trip crossing notification when the trip becomes valid.
+                */
+               td->threshold = INT_MAX;
+       }
        trip->temperature = temp;
        thermal_notify_tz_trip_change(tz, trip);
 }