Skip to contentSkip to conversation

thermal: intel: int340x: Use generic trip points table

PATCH In review Rafael J. Wysocki 3 participants 9 messages Jan 25, 2023 – Jan 26, 2023

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover thermal: intel: int340x: Use generic trip points table

Hi All,

This series replaces the following patch:

https://patchwork.kernel.org/project/linux-pm/patch/2147918.irdbgypaU6@kreacher/

but it has been almost completely rewritten, so I've dropped all tags from it.

The most significant difference is that firmware-induced trip point updates are now handled in a less controversial manner (no renumbering, just temperature updates if applicable).

Please refer to the individual patch changelogs for details.

The series is on top of this patch:

https://patchwork.kernel.org/project/linux-pm/patch/2688799.mvXUDI8C0e@kreacher/

which applies on top of the linux-next branch in linux-pm.git from today.

Thanks!

#1 lore raw

PATCH 1/3 thermal: intel: int340x: Rework updating trip points

From: Rafael J. Wysocki <[email protected]>

It is generally invalid to change the trip point indices after they have been exposed via sysfs.

Moreover, the thermal objects in the ACPI namespace cannot go away and appear on the fly. In practice, the only thing that can happen when the INT3403_PERF_TRIP_POINT_CHANGED notification is sent by the platform firmware is a change of the return values of those thermal objects.

For this reason, add a special function for updating the trip point temperatures after re-evaluating the respective ACPI thermal objects and change int3403_notify() to invoke it instead of int340x_thermal_read_trips() that would change the trip point indices on errors. Also remove the locking from the latter, because it is only called before registering the thermal zone and it cannot race with the zone's callbacks.

---
 drivers/thermal/intel/int340x_thermal/int3403_thermal.c      |    2 
 drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c |   51 +++++++++--
 drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h |    2 
 3 files changed, 47 insertions(+), 8 deletions(-)
drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c +45 -6
Index: linux-pm/drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c
===================================================================
@@ -168,13 +168,11 @@ static int int340x_thermal_get_trip_conf
 	return 0;
 }
 
-int int340x_thermal_read_trips(struct int34x_thermal_zone *int34x_zone)
+static int int340x_thermal_read_trips(struct int34x_thermal_zone *int34x_zone)
 {
 	int trip_cnt = int34x_zone->aux_trip_nr;
 	int i;
 
-	mutex_lock(&int34x_zone->trip_mutex);
-
 	int34x_zone->crt_trip_id = -1;
 	if (!int340x_thermal_get_trip_config(int34x_zone->adev->handle, "_CRT",
 					     &int34x_zone->crt_temp))
@@ -202,11 +200,8 @@ int int340x_thermal_read_trips(struct in
 		int34x_zone->act_trips[i].valid = true;
 	}
 
-	mutex_unlock(&int34x_zone->trip_mutex);
-
 	return trip_cnt;
 }
-EXPORT_SYMBOL_GPL(int340x_thermal_read_trips);
 
 static struct thermal_zone_params int340x_thermal_params = {
 	.governor_name = "user_space",
@@ -309,6 +304,50 @@ void int340x_thermal_zone_remove(struct
 }
 EXPORT_SYMBOL_GPL(int340x_thermal_zone_remove);
 
+void int340x_thermal_update_trips(struct int34x_thermal_zone *int34x_zone)
+{
+	acpi_handle zone_handle = int34x_zone->adev->handle;
+	int i, err;
+
+	mutex_lock(&int34x_zone->trip_mutex);
+
+	if (int34x_zone->crt_trip_id > 0) {
+		err = int340x_thermal_get_trip_config(zone_handle, "_CRT",
+						      &int34x_zone->crt_temp);
+		if (err)
+			int34x_zone->crt_temp = THERMAL_TEMP_INVALID;
+	}
+
+	if (int34x_zone->hot_trip_id > 0) {
+		err = int340x_thermal_get_trip_config(zone_handle, "_HOT",
+						      &int34x_zone->hot_temp);
+		if (err)
+			int34x_zone->hot_temp = THERMAL_TEMP_INVALID;
+	}
+
+	if (int34x_zone->psv_trip_id > 0) {
+		err = int340x_thermal_get_trip_config(zone_handle, "_PSV",
+						      &int34x_zone->psv_temp);
+		if (err)
+			int34x_zone->psv_temp = THERMAL_TEMP_INVALID;
+	}
+
+	for (i = 0; i < INT340X_THERMAL_MAX_ACT_TRIP_COUNT; i++) {
+		char name[5] = { '_', 'A', 'C', '0' + i, '\0' };
+
+		if (!int34x_zone->act_trips[i].valid)
+			break;
+
+		err = int340x_thermal_get_trip_config(zone_handle, name,
+						      &int34x_zone->act_trips[i].temp);
+		if (err)
+			int34x_zone->act_trips[i].temp = THERMAL_TEMP_INVALID;
+	}
+
+	mutex_unlock(&int34x_zone->trip_mutex);
+}
+EXPORT_SYMBOL_GPL(int340x_thermal_update_trips);
+
 MODULE_AUTHOR("Aaron Lu <[email protected]>");
 MODULE_AUTHOR("Srinivas Pandruvada <[email protected]>");
 MODULE_DESCRIPTION("Intel INT340x common thermal zone handler");
drivers/thermal/intel/int340x_thermal/int3403_thermal.c +1 -1
Index: linux-pm/drivers/thermal/intel/int340x_thermal/int3403_thermal.c
===================================================================
@@ -69,7 +69,7 @@ static void int3403_notify(acpi_handle h
 						   THERMAL_TRIP_VIOLATED);
 		break;
 	case INT3403_PERF_TRIP_POINT_CHANGED:
-		int340x_thermal_read_trips(obj->int340x_zone);
+		int340x_thermal_update_trips(obj->int340x_zone);
 		int340x_thermal_zone_device_update(obj->int340x_zone,
 						   THERMAL_TRIP_CHANGED);
 		break;
drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h +1 -1
Index: linux-pm/drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h
===================================================================
@@ -38,7 +38,7 @@ struct int34x_thermal_zone {
 struct int34x_thermal_zone *int340x_thermal_zone_add(struct acpi_device *,
 				int (*get_temp) (struct thermal_zone_device *, int *));
 void int340x_thermal_zone_remove(struct int34x_thermal_zone *);
-int int340x_thermal_read_trips(struct int34x_thermal_zone *int34x_zone);
+void int340x_thermal_update_trips(struct int34x_thermal_zone *int34x_zone);
 
 static inline void int340x_thermal_zone_set_priv_data(
 			struct int34x_thermal_zone *tzone, void *priv_data)
#2 lore raw

PATCH 2/3 thermal: intel: int340x: Use zone lock for synchronization

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Because the ->get_trip_temp() and ->get_trip_type() thermal zone callbacks are only invoked from __thermal_zone_get_trip() which is always called by the thermal core under the zone lock, it is sufficient for int340x_thermal_update_trips() to acquire the zone lock for mutual exclusion with those callbacks.

Accordingly, modify int340x_thermal_update_trips() to use the zone lock instead of the internal trip_mutex and drop the latter which is not necessary any more.

---
 drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c |   28 +++--------
 drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h |    1 
 2 files changed, 8 insertions(+), 21 deletions(-)
drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c +8 -20
Index: linux-pm/drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c
===================================================================
@@ -41,9 +41,7 @@ static int int340x_thermal_get_trip_temp
 					 int trip, int *temp)
 {
 	struct int34x_thermal_zone *d = zone->devdata;
-	int i, ret = 0;
-
-	mutex_lock(&d->trip_mutex);
+	int i;
 
 	if (trip < d->aux_trip_nr)
 		*temp = d->aux_trips[trip];
@@ -62,12 +60,10 @@ static int int340x_thermal_get_trip_temp
 			}
 		}
 		if (i == INT340X_THERMAL_MAX_ACT_TRIP_COUNT)
-			ret = -EINVAL;
+			return -EINVAL;
 	}
 
-	mutex_unlock(&d->trip_mutex);
-
-	return ret;
+	return 0;
 }
 
 static int int340x_thermal_get_trip_type(struct thermal_zone_device *zone,
@@ -75,9 +71,7 @@ static int int340x_thermal_get_trip_type
 					 enum thermal_trip_type *type)
 {
 	struct int34x_thermal_zone *d = zone->devdata;
-	int i, ret = 0;
-
-	mutex_lock(&d->trip_mutex);
+	int i;
 
 	if (trip < d->aux_trip_nr)
 		*type = THERMAL_TRIP_PASSIVE;
@@ -96,12 +90,10 @@ static int int340x_thermal_get_trip_type
 			}
 		}
 		if (i == INT340X_THERMAL_MAX_ACT_TRIP_COUNT)
-			ret = -EINVAL;
+			return -EINVAL;
 	}
 
-	mutex_unlock(&d->trip_mutex);
-
-	return ret;
+	return 0;
 }
 
 static int int340x_thermal_set_trip_temp(struct thermal_zone_device *zone,
@@ -222,8 +214,6 @@ struct int34x_thermal_zone *int340x_ther
 	if (!int34x_thermal_zone)
 		return ERR_PTR(-ENOMEM);
 
-	mutex_init(&int34x_thermal_zone->trip_mutex);
-
 	int34x_thermal_zone->adev = adev;
 
 	int34x_thermal_zone->ops = kmemdup(&int340x_thermal_zone_ops,
@@ -286,7 +276,6 @@ err_thermal_zone:
 err_trip_alloc:
 	kfree(int34x_thermal_zone->ops);
 err_ops_alloc:
-	mutex_destroy(&int34x_thermal_zone->trip_mutex);
 	kfree(int34x_thermal_zone);
 	return ERR_PTR(ret);
 }
@@ -299,7 +288,6 @@ void int340x_thermal_zone_remove(struct
 	acpi_lpat_free_conversion_table(int34x_thermal_zone->lpat_table);
 	kfree(int34x_thermal_zone->aux_trips);
 	kfree(int34x_thermal_zone->ops);
-	mutex_destroy(&int34x_thermal_zone->trip_mutex);
 	kfree(int34x_thermal_zone);
 }
 EXPORT_SYMBOL_GPL(int340x_thermal_zone_remove);
@@ -309,7 +297,7 @@ void int340x_thermal_update_trips(struct
 	acpi_handle zone_handle = int34x_zone->adev->handle;
 	int i, err;
 
-	mutex_lock(&int34x_zone->trip_mutex);
+	mutex_lock(&int34x_zone->zone->lock);
 
 	if (int34x_zone->crt_trip_id > 0) {
 		err = int340x_thermal_get_trip_config(zone_handle, "_CRT",
@@ -344,7 +332,7 @@ void int340x_thermal_update_trips(struct
 			int34x_zone->act_trips[i].temp = THERMAL_TEMP_INVALID;
 	}
 
-	mutex_unlock(&int34x_zone->trip_mutex);
+	mutex_unlock(&int34x_zone->zone->lock);
 }
 EXPORT_SYMBOL_GPL(int340x_thermal_update_trips);
 
drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h +0 -1
Index: linux-pm/drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h
===================================================================
@@ -32,7 +32,6 @@ struct int34x_thermal_zone {
 	struct thermal_zone_device_ops *ops;
 	void *priv_data;
 	struct acpi_lpat_conversion_table *lpat_table;
-	struct mutex trip_mutex;
 };
 
 struct int34x_thermal_zone *int340x_thermal_zone_add(struct acpi_device *,
#3 lore raw

PATCH 3/3 thermal: intel: int340x: Use generic trip points table

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Modify int340x_thermal_zone_add() to register the thermal zone along with a trip points table, which allows the trip-related zone callbacks to be dropped, because they are not needed any more.

In order to consolidate the code, use ACPI trip library functions to populate generic trip points in int340x_thermal_read_trips() and to update them in int340x_thermal_update_trips().

---
 drivers/thermal/intel/int340x_thermal/Kconfig                |    1 
 drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c |  265 +++--------
 drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h |   10 
 3 files changed, 90 insertions(+), 186 deletions(-)
drivers/thermal/intel/int340x_thermal/Kconfig +1 -0
Index: linux-pm/drivers/thermal/intel/int340x_thermal/Kconfig
===================================================================
@@ -9,6 +9,7 @@ config INT340X_THERMAL
 	select THERMAL_GOV_USER_SPACE
 	select ACPI_THERMAL_REL
 	select ACPI_FAN
+	select THERMAL_ACPI
 	select INTEL_SOC_DTS_IOSF_CORE
 	select INTEL_TCC
 	select PROC_THERMAL_MMIO_RAPL if POWERCAP
drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.c +87 -178

336 lines not on this page · raw message

drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h +2 -8
Index: linux-pm/drivers/thermal/intel/int340x_thermal/int340x_thermal_zone.h
===================================================================
@@ -10,6 +10,7 @@
 #include <acpi/acpi_lpat.h>
 
 #define INT340X_THERMAL_MAX_ACT_TRIP_COUNT	10
+#define INT340X_THERMAL_MAX_TRIP_COUNT INT340X_THERMAL_MAX_ACT_TRIP_COUNT + 3
 
 struct active_trip {
 	int temp;
@@ -19,15 +20,8 @@ struct active_trip {
 
 struct int34x_thermal_zone {
 	struct acpi_device *adev;
-	struct active_trip act_trips[INT340X_THERMAL_MAX_ACT_TRIP_COUNT];
-	unsigned long *aux_trips;
+	struct thermal_trip *trips;
 	int aux_trip_nr;
-	int psv_temp;
-	int psv_trip_id;
-	int crt_temp;
-	int crt_trip_id;
-	int hot_temp;
-	int hot_trip_id;
 	struct thermal_zone_device *zone;
 	struct thermal_zone_device_ops *ops;
 	void *priv_data;
#4 lore raw
In reply to Rafael J. Wysocki

Hi Srinivas,

Quoted 13 lines · Rafael J. Wysocki

Hi All,

This series replaces the following patch:

https://patchwork.kernel.org/project/linux-pm/patch/2147918.irdbgypaU6@kreacher/

but it has been almost completely rewritten, so I've dropped all tags from it.

The most significant difference is that firmware-induced trip point updates are now handled in a less controversial manner (no renumbering, just temperature updates if applicable).

Quoted 7 lines

Please refer to the individual patch changelogs for details.

The series is on top of this patch:

https://patchwork.kernel.org/project/linux-pm/patch/2688799.mvXUDI8C0e@kreacher/

which applies on top of the linux-next branch in linux-pm.git from today.

There are two additional branches in linux-pm.git:

thermal-intel-fixes
thermal-intel-testing

The former is just fixes to go on top of 6.2-rc5 and the latter - this series on top of those and the current thermal-intel branch I have locally with the Intel thermal drivers changes for 6.3.

I would appreciate giving each of them a go in your test setup.

Cheers!

#5 lore raw

Hi Rafael,

Quoted 15 lines · Rafael J. Wysocki

Hi Srinivas,

Quoted 11 lines · Rafael J. Wysocki

Hi All,

This series replaces the following patch:

https://patchwork.kernel.org/project/linux-pm/patch/2147918.irdbgypaU6@kreacher/

but it has been almost completely rewritten, so I've dropped all tags from it.

[...]

Quoted 10 lines
Quoted 6 lines

The series is on top of this patch:

https://patchwork.kernel.org/project/linux-pm/patch/2688799.mvXUDI8C0e@kreacher/

which applies on top of the linux-next branch in linux-pm.git from today.

There are two additional branches in linux-pm.git:

thermal-intel-fixes

On two systems test, no issues are observed.

Quoted 1 line

thermal-intel-testing

branch: thermal-intel-test

No issues, but number of trips are not same as invalid trips are not registered.
Not sure if this is correct. At boot up they may be invalid, but firmware may update later (Not aware of such scenario).

For example, the hot is not registered.

Current:

thermal_zone9/trip_point_0_type:critical
thermal_zone9/trip_point_0_temp:125050
thermal_zone9/trip_point_0_hyst:0

thermal_zone9/trip_point_1_type:hot
thermal_zone9/trip_point_1_temp:-273250
thermal_zone9/trip_point_1_hyst:0

thermal_zone9/trip_point_2_type:passive
thermal_zone9/trip_point_2_temp:103050
thermal_zone9/trip_point_2_hyst:0

thermal_zone9/trip_point_3_type:active
thermal_zone9/trip_point_3_temp:103050
thermal_zone9/trip_point_3_hyst:0

thermal_zone9/trip_point_4_type:active
thermal_zone9/trip_point_4_temp:101050
thermal_zone9/trip_point_4_hyst:0

thermal_zone9/trip_point_5_type:active
thermal_zone9/trip_point_5_temp:100050
thermal_zone9/trip_point_5_hyst:0

thermal_zone9/trip_point_6_type:active
thermal_zone9/trip_point_6_temp:98550
thermal_zone9/trip_point_6_hyst:0

thermal_zone9/trip_point_7_type:active
thermal_zone9/trip_point_7_temp:97050
thermal_zone9/trip_point_7_hyst:0

with 6.3-rc1 changes

thermal_zone9/trip_point_0_type:critical
thermal_zone9/trip_point_0_temp:125050
thermal_zone9/trip_point_0_hyst:0

thermal_zone9/trip_point_1_type:passive
thermal_zone9/trip_point_1_temp:103050
thermal_zone9/trip_point_1_hyst:0

thermal_zone9/trip_point_2_type:active
thermal_zone9/trip_point_2_temp:103050
thermal_zone9/trip_point_2_hyst:0

thermal_zone9/trip_point_3_type:active
thermal_zone9/trip_point_3_temp:101050
thermal_zone9/trip_point_3_hyst:0

thermal_zone9/trip_point_4_type:active
thermal_zone9/trip_point_4_temp:100050
thermal_zone9/trip_point_4_hyst:0

thermal_zone9/trip_point_5_type:active
thermal_zone9/trip_point_5_temp:98550
thermal_zone9/trip_point_5_hyst:0

thermal_zone9/trip_point_6_hyst:0
thermal_zone9/trip_point_6_temp:97050
thermal_zone9/trip_point_6_type:active

Thanks,
Srinivas

Quoted 9 lines

The former is just fixes to go on top of 6.2-rc5 and the latter - this
series on top of those and the current thermal-intel branch I have locally with the Intel thermal drivers changes for 6.3.

I would appreciate giving each of them a go in your test setup.

Cheers!

#6 lore raw
Quoted 26 lines · CET srinivas pandruvada

Hi Rafael,

Quoted 15 lines · Rafael J. Wysocki

Hi Srinivas,

Quoted 11 lines · Rafael J. Wysocki

Hi All,

This series replaces the following patch:

https://patchwork.kernel.org/project/linux-pm/patch/2147918.irdbgypaU6@kreacher/

but it has been almost completely rewritten, so I've dropped all tags from it.

[...]

Quoted 4 lines
Quoted 4 lines

The series is on top of this patch:

https://patchwork.kernel.org/project/linux-pm/patch/2688799.mvXUDI8C0e@kreacher/

Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

which applies on top of the linux-next branch in linux-pm.git from today.

There are two additional branches in linux-pm.git:

thermal-intel-fixes

On two systems test, no issues are observed.

Great! I'll move this to linux-next then.

Quoted 6 lines
Quoted 1 line

thermal-intel-testing

branch: thermal-intel-test

No issues, but number of trips are not same as invalid trips are not registered.
Not sure if this is correct.

It may not be. At least it is a change in behavior that is not expected to happen after these changes.

Quoted 7 lines

At boot up they may be invalid, but
firmware may update later (Not aware of such scenario).

For example, the hot is not registered.

Current:

Quoted 7 lines

thermal_zone9/trip_point_0_type:critical
thermal_zone9/trip_point_0_temp:125050
thermal_zone9/trip_point_0_hyst:0

thermal_zone9/trip_point_1_type:hot
thermal_zone9/trip_point_1_temp:-273250
thermal_zone9/trip_point_1_hyst:0

So this means that _HOT is evaluated successfully (or the trip point index would be negative), but it probably returned an invalid temperature (likely 0) that has been turned into an error by the temperature range check in the new ACPI helper introduced by the change.

OK, thanks for testing!

I've added the appended patch to the thermal-intel-test branch. Can you please check if it makes that difference in behavior go away?

---
From: Rafael J. Wysocki <[email protected]>
Subject: [PATCH] thermal: ACPI: Initialize trips if temperature is out of range

In some cases it is still useful to register a trip point if the temperature returned by the corresponding ACPI thermal object (for example, _HOT) is invalid to start with, because the same ACPI thermal object may start to return a valid temperature after a system configuration change (for example, from an AC power source to battery an vice versa).

For this reason, if the ACPI thermal object evaluated by thermal_acpi_trip_init() successfully returns a temperature value that is out of the range of values taken into account, initialize the trip point using THERMAL_TEMP_INVALID as the temperature value instead of returning an error to allow the user of the trip point to decide what to do with it.

Also update pch_wpt_add_acpi_psv_trip() to reject trip points with invalid temperature values.

Fixes 7a0e39748861 ("thermal: ACPI: Add ACPI trip point routines")
---
 drivers/thermal/intel/intel_pch_thermal.c |    2 +-
 drivers/thermal/thermal_acpi.c            |    7 ++++---
 2 files changed, 5 insertions(+), 4 deletions(-)
drivers/thermal/thermal_acpi.c +4 -3
Index: linux-pm/drivers/thermal/thermal_acpi.c
===================================================================
@@ -64,13 +64,14 @@ static int thermal_acpi_trip_init(struct
 		return -ENODATA;
 	}
 
-	if (temp < TEMP_MIN_DECIK || temp >= TEMP_MAX_DECIK) {
+	if (temp >= TEMP_MIN_DECIK && temp <= TEMP_MAX_DECIK) {
+		trip->temperature = deci_kelvin_to_millicelsius(temp);
+	} else {
 		acpi_handle_debug(adev->handle, "%s result %llu out of range\n",
 				  obj_name, temp);
-		return -ENODATA;
+		trip->temperature = THERMAL_TEMP_INVALID;
 	}
 
-	trip->temperature = deci_kelvin_to_millicelsius(temp);
 	trip->hysteresis = 0;
 	trip->type = type;
 
drivers/thermal/intel/intel_pch_thermal.c +1 -1
Index: linux-pm/drivers/thermal/intel/intel_pch_thermal.c
===================================================================
@@ -107,7 +107,7 @@ static void pch_wpt_add_acpi_psv_trip(st
 		return;
 
 	ret = thermal_acpi_trip_passive(adev, &ptd->trips[*nr_trips]);
-	if (ret)
+	if (ret || ptd->trips[*nr_trips].temperature <= 0)
 		return;
 
 	++(*nr_trips);
#7 lore raw

Hi Rafael,

Quoted 6 lines · Rafael J. Wysocki
Quoted 3 lines · CET srinivas pandruvada

Hi Rafael,

[...]

Quoted 3 lines

I've added the appended patch to the thermal-intel-test branch.  Can you please
check if it makes that difference in behavior go away?

I synced the tree again and your patch in thermal-intel-test fixes the issue.

Thanks,
Srinivas

Quoted 74 lines

---
From: Rafael J. Wysocki <[email protected]>
Subject: [PATCH] thermal: ACPI: Initialize trips if temperature is out of range

In some cases it is still useful to register a trip point if the temperature returned by the corresponding ACPI thermal object (for example, _HOT) is invalid to start with, because the same ACPI thermal object may start to return a valid temperature after a system configuration change (for example, from an AC power source to battery an vice versa).

For this reason, if the ACPI thermal object evaluated by thermal_acpi_trip_init() successfully returns a temperature value that
is out of the range of values taken into account, initialize the trip point using THERMAL_TEMP_INVALID as the temperature value instead of returning an error to allow the user of the trip point to decide what to do with it.

Also update pch_wpt_add_acpi_psv_trip() to reject trip points with invalid temperature values.

Fixes 7a0e39748861 ("thermal: ACPI: Add ACPI trip point routines")
Reported-by Srinivas Pandruvada

<[email protected]>

---
 drivers/thermal/intel/intel_pch_thermal.c |    2 +- drivers/thermal/thermal_acpi.c            |    7 ++++---

 2 files changed, 5 insertions(+), 4 deletions(-)

Index: linux-pm/drivers/thermal/thermal_acpi.c
===================================================================

drivers/thermal/thermal_acpi.c +0 -0
@@ -64,13 +64,14 @@ static int thermal_acpi_trip_init(struct
                 return -ENODATA;
         }


-       if (temp < TEMP_MIN_DECIK || temp >= TEMP_MAX_DECIK) { +       if (temp >= TEMP_MIN_DECIK && temp <= TEMP_MAX_DECIK) { +               trip->temperature = deci_kelvin_to_millicelsius(temp);
+       } else {
                acpi_handle_debug(adev->handle, "%s result %llu out of range\n",
                                  obj_name, temp);
-               return -ENODATA;
+               trip->temperature = THERMAL_TEMP_INVALID; }

-       trip->temperature = deci_kelvin_to_millicelsius(temp); trip->hysteresis = 0;
        trip->type = type;

Index: linux-pm/drivers/thermal/intel/intel_pch_thermal.c
===================================================================

drivers/thermal/intel/intel_pch_thermal.c +0 -0
@@ -107,7 +107,7 @@ static void pch_wpt_add_acpi_psv_trip(st
                 return;


        ret = thermal_acpi_trip_passive(adev, &ptd-

Quoted 1 line

trips[*nr_trips]);

-       if (ret)
+       if (ret || ptd->trips[*nr_trips].temperature <= 0) return;

        ++(*nr_trips);

#8 lore raw

Hi Srinivas,

Quoted 11 lines · srinivas pandruvada

Hi Rafael,

Quoted 6 lines · Rafael J. Wysocki
Quoted 3 lines · CET srinivas pandruvada

Hi Rafael,

Quoted 7 lines

[...]

Quoted 3 lines

I've added the appended patch to the thermal-intel-test branch. Can you please
check if it makes that difference in behavior go away?

I synced the tree again and your patch in thermal-intel-test fixes the issue.

Thanks a lot for testing and the confirmation!

In the meantime, I've merged the thermal-intel-test into the bleeding-edge branch and if 0-day reports success with building it, I'll move the patches to linux-next.

Cheers!