ACPI: Adjust Kelvin offset to match local implementation

The exact offset between Kelvin and degree Celsius is 273.15. However ACPI handles temperature values with a single decimal place. As a consequence, some implementations use an offset of 273.1 and others use an offset of 273.2. Try to find out which one is being used, to present the most accurate and visually appealing number.

Tested on a Sony Vaio PGC-GR214EP (which uses 273.1) and a Lenovo Thinkpad T60p (which uses 273.2).

Signed-off-by Jean Delvare <[email protected]>

Cc Zhang Rui<[email protected]>

Cc Len Brown <[email protected]>

---
Blame the mess on whoever decided that ACPI temperature would be expressed in the least adapted unit in the world :(

Without this patch, I have the following sensors output:

acpitz-virtual-0
Adapter: Virtual device
temp1: +44.9 C (crit = +89.9 C)

With the patch I instead have:

acpitz-virtual-0
Adapter: Virtual device
temp1: +45.0 C (crit = +90.0 C)

Which I think is much easier to read.

 drivers/acpi/thermal.c |   42 +++++++++++++++++++++++++++++++++++-------
 1 file changed, 35 insertions(+), 7 deletions(-)

drivers/acpi/thermal.c +35 -7

@@ -193,6 +193,7 @@ struct acpi_thermal {
 	struct timer_list timer;
 	struct thermal_zone_device *thermal_zone;
 	int tz_enabled;
+	int kelvin_offset;
 	struct mutex lock;
 };
 

@@ -952,7 +953,7 @@ static void acpi_thermal_check(void *dat
 }
 
 /* sys I/F for generic thermal sysfs support */
-#define KELVIN_TO_MILLICELSIUS(t) (t * 100 - 273200)
+#define KELVIN_TO_MILLICELSIUS(t, off) (((t) - (off)) * 100)
 
 static int thermal_get_temp(struct thermal_zone_device *thermal, char *buf)
 {

@@ -966,7 +967,8 @@ static int thermal_get_temp(struct therm
 	if (result)
 		return result;
 
-	return sprintf(buf, "%ld\n", KELVIN_TO_MILLICELSIUS(tz->temperature));
+	return sprintf(buf, "%ld\n", KELVIN_TO_MILLICELSIUS(tz->temperature,
+							    tz->kelvin_offset));
 }
 
 static const char enabled[] = "kernel";

@@ -1061,21 +1063,24 @@ static int thermal_get_trip_temp(struct
 	if (tz->trips.critical.flags.valid) {
 		if (!trip)
 			return sprintf(buf, "%ld\n", KELVIN_TO_MILLICELSIUS(
-				tz->trips.critical.temperature));
+				tz->trips.critical.temperature,
+				tz->kelvin_offset));
 		trip--;
 	}
 
 	if (tz->trips.hot.flags.valid) {
 		if (!trip)
 			return sprintf(buf, "%ld\n", KELVIN_TO_MILLICELSIUS(
-					tz->trips.hot.temperature));
+					tz->trips.hot.temperature,
+					tz->kelvin_offset));
 		trip--;
 	}
 
 	if (tz->trips.passive.flags.valid) {
 		if (!trip)
 			return sprintf(buf, "%ld\n", KELVIN_TO_MILLICELSIUS(
-					tz->trips.passive.temperature));
+					tz->trips.passive.temperature,
+					tz->kelvin_offset));
 		trip--;
 	}
 

@@ -1083,7 +1088,8 @@ static int thermal_get_trip_temp(struct
 		tz->trips.active[i].flags.valid; i++) {
 		if (!trip)
 			return sprintf(buf, "%ld\n", KELVIN_TO_MILLICELSIUS(
-					tz->trips.active[i].temperature));
+					tz->trips.active[i].temperature,
+					tz->kelvin_offset));
 		trip--;
 	}
 

@@ -1096,7 +1102,8 @@ static int thermal_get_crit_temp(struct
 
 	if (tz->trips.critical.flags.valid) {
 		*temperature = KELVIN_TO_MILLICELSIUS(
-				tz->trips.critical.temperature);
+				tz->trips.critical.temperature,
+				tz->kelvin_offset);
 		return 0;
 	} else
 		return -EINVAL;

@@ -1649,6 +1656,25 @@ static int acpi_thermal_get_info(struct
 	return 0;
 }
 
+/*
+ * The exact offset between Kelvin and degree Celsius is 273.15. However ACPI
+ * handles temperature values with a single decimal place. As a consequence,
+ * some implementations use an offset of 273.1 and others use an offset of
+ * 273.2. Try to find out which one is being used, to present the most
+ * accurate and visually appealing number.
+ *
+ * The heuristic below should work for all ACPI thermal zones which have a
+ * critical trip point with a value being a multiple of 0.5 degree Celsius.
+ */
+static void acpi_thermal_guess_offset(struct acpi_thermal *tz)
+{
+	if (tz->trips.critical.flags.valid &&
+	    (tz->trips.critical.temperature % 5) == 1)
+		tz->kelvin_offset = 2731;
+	else
+		tz->kelvin_offset = 2732;
+}
+
 static int acpi_thermal_add(struct acpi_device *device)
 {
 	int result = 0;

@@ -1675,6 +1701,8 @@ static int acpi_thermal_add(struct acpi_
 	if (result)
 		goto free_memory;
 
+	acpi_thermal_guess_offset(tz);
+
 	result = acpi_thermal_register_thermal_zone(tz);
 	if (result)
 		goto free_memory;

Quoted 6 lines · Jean Delvare

The exact offset between Kelvin and degree Celsius is 273.15. However ACPI handles temperature values with a single decimal place. As a consequence, some implementations use an offset of 273.1 and others use an offset of 273.2. Try to find out which one is being used, to present the most accurate and visually appealing number.

Quoted 2 lines

Tested on a Sony Vaio PGC-GR214EP (which uses 273.1) and a Lenovo Thinkpad T60p (which uses 273.2).

Sounds reasonable.

Acked-by Zhang Rui <[email protected]>

Hi Rui,

Quoted 7 lines · Zhang Rui

Quoted 6 lines · Jean Delvare

The exact offset between Kelvin and degree Celsius is 273.15. However ACPI handles temperature values with a single decimal place. As a consequence, some implementations use an offset of 273.1 and others use an offset of 273.2. Try to find out which one is being used, to present the most accurate and visually appealing number.

Quoted 5 lines

Quoted 2 lines

Tested on a Sony Vaio PGC-GR214EP (which uses 273.1) and a Lenovo Thinkpad T60p (which uses 273.2).

Sounds reasonable.

Acked-by Zhang Rui <[email protected]>

Thanks. Is this patch queued for 2.6.30 somewhere?

Quoted 14 lines · Jean Delvare

Hi Rui,

Quoted 11 lines · Zhang Rui

Quoted 8 lines · Jean Delvare

The exact offset between Kelvin and degree Celsius is 273.15. However ACPI handles temperature values with a single decimal place. As a consequence, some implementations use an offset of 273.1 and others use an offset of 273.2. Try to find out which one is being used, to present the most accurate and visually appealing number.

Tested on a Sony Vaio PGC-GR214EP (which uses 273.1) and a Lenovo Thinkpad T60p (which uses 273.2).

Sounds reasonable.

Quoted 3 lines

Quoted 1 line

Acked-by Zhang Rui <[email protected]>

Thanks. Is this patch queued for 2.6.30 somewhere?

it was, but I had a conflict between this an another and so I dropped this one until I could look at it - which I havn't yet. If you'd like to do it for me, rebase on top of the acpi test branch.