thermal: gov_step_wise: Go straight to instance->lower when mitigation is over
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>
Sat, 22 Jun 2024 12:26:33 +0000 (14:26 +0200)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Tue, 25 Jun 2024 12:37:05 +0000 (14:37 +0200)
Commit b6846826982b ("thermal: gov_step_wise: Restore passive polling
management") attempted to fix a Step-Wise thermal governor issue
introduced by commit 042a3d80f118 ("thermal: core: Move passive polling
management to the core"), which caused the governor to leave cooling
devices in high states, by partially reverting that commit.

However, this turns out to be insufficient on some systems due to
interactions between the governor code restored by commit b6846826982b
and the passive polling management in the thermal core.

For this reason, revert commit b6846826982b and make the governor set
the target cooling device state to the "lower" one as soon as the zone
temperature falls below the threshold of the trip point corresponding
to the given thermal instance, which means that thermal mitigation is
not necessary any more.

Before this change the "lower" cooling device state would be reached in
steps through the passive polling mechanism which was questionable for
three reasons: (1) cooling device were kept in high states when that was
not necessary (and it could adversely impact performance), (2) it only
worked for thermal zones with nonzero passive_delay_jiffies value, and
(3) passive polling belongs to the core and should not be hijacked by
governors for their internal purposes.

Fixes: b6846826982b ("thermal: gov_step_wise: Restore passive polling management")
Closes: https://lore.kernel.org/linux-pm/6759ce9f-281d-4fcd-bb4c-b784a1cc5f6e@oldschoolsolutions.biz
Reported-by: Jens Glathe <jens.glathe@oldschoolsolutions.biz>
Tested-by: Jens Glathe <jens.glathe@oldschoolsolutions.biz>
Link: https://patch.msgid.link/12464461.O9o76ZdvQC@rjwysocki.net
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Steev Klimaszewski <steev@kali.org>
Tested-by: Johan Hovold <johan+linaro@kernel.org>
drivers/thermal/gov_step_wise.c

index 65974fe..fd55271 100644 (file)
@@ -55,7 +55,11 @@ static unsigned long get_target_state(struct thermal_instance *instance,
                if (cur_state <= instance->lower)
                        return THERMAL_NO_TARGET;
 
-               return clamp(cur_state - 1, instance->lower, instance->upper);
+               /*
+                * If 'throttle' is false, no mitigation is necessary, so
+                * request the lower state for this instance.
+                */
+               return instance->lower;
        }
 
        return instance->target;
@@ -93,23 +97,6 @@ static void thermal_zone_trip_update(struct thermal_zone_device *tz,
                if (instance->initialized && old_target == instance->target)
                        continue;
 
-               if (trip->type == THERMAL_TRIP_PASSIVE) {
-                       /*
-                        * If the target state for this thermal instance
-                        * changes from THERMAL_NO_TARGET to something else,
-                        * ensure that the zone temperature will be updated
-                        * (assuming enabled passive cooling) until it becomes
-                        * THERMAL_NO_TARGET again, or the cooling device may
-                        * not be reset to its initial state.
-                        */
-                       if (old_target == THERMAL_NO_TARGET &&
-                           instance->target != THERMAL_NO_TARGET)
-                               tz->passive++;
-                       else if (old_target != THERMAL_NO_TARGET &&
-                                instance->target == THERMAL_NO_TARGET)
-                               tz->passive--;
-               }
-
                instance->initialized = true;
 
                mutex_lock(&instance->cdev->lock);