drm/amd/powerplay: expose current hotspot and memory temperatures V2

Two new hwmon interfaces(temp2_input and temp3_input) are added.
They are supported on SOC15 dGPUs only.

- V2: correct thermal sensor output
Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c

@@ -1382,6 +1382,7 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev,
 {
 	struct amdgpu_device *adev = dev_get_drvdata(dev);
 	struct drm_device *ddev = adev->ddev;
+	int channel = to_sensor_dev_attr(attr)->index;
 	int r, temp, size = sizeof(temp);
 
 	/* Can't get temperature when the card is off */
@@ -1389,11 +1390,32 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	/* get the temperature */
-	r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_TEMP,
-				   (void *)&temp, &size);
-	if (r)
-		return r;
+	if (channel >= PP_TEMP_MAX)
+		return -EINVAL;
+
+	switch (channel) {
+	case PP_TEMP_JUNCTION:
+		/* get current junction temperature */
+		r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_HOTSPOT_TEMP,
+					   (void *)&temp, &size);
+		if (r)
+			return r;
+		break;
+	case PP_TEMP_EDGE:
+		/* get current edge temperature */
+		r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_EDGE_TEMP,
+					   (void *)&temp, &size);
+		if (r)
+			return r;
+		break;
+	case PP_TEMP_MEM:
+		/* get current memory temperature */
+		r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_MEM_TEMP,
+					   (void *)&temp, &size);
+		if (r)
+			return r;
+		break;
+	}
 
 	return snprintf(buf, PAGE_SIZE, "%d\n", temp);
 }
@@ -2041,7 +2063,8 @@ static ssize_t amdgpu_hwmon_show_mclk_label(struct device *dev,
  *
  * hwmon interfaces for GPU temperature:
  *
- * - temp1_input: the on die GPU temperature in millidegrees Celsius
+ * - temp[1-3]_input: the on die GPU temperature in millidegrees Celsius
+ *   - temp2_input and temp3_input are supported on SOC15 dGPUs only
  *
  * - temp[1-3]_crit: temperature critical max value in millidegrees Celsius
  *   - temp2_crit and temp3_crit are supported on SOC15 dGPUs only
@@ -2098,13 +2121,15 @@ static ssize_t amdgpu_hwmon_show_mclk_label(struct device *dev,
  *
  */
 
-static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, amdgpu_hwmon_show_temp, NULL, 0);
+static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, amdgpu_hwmon_show_temp, NULL, PP_TEMP_EDGE);
 static SENSOR_DEVICE_ATTR(temp1_crit, S_IRUGO, amdgpu_hwmon_show_temp_thresh, NULL, 0);
 static SENSOR_DEVICE_ATTR(temp1_crit_hyst, S_IRUGO, amdgpu_hwmon_show_temp_thresh, NULL, 1);
 static SENSOR_DEVICE_ATTR(temp1_emergency, S_IRUGO, amdgpu_hwmon_show_temp_emergency, NULL, PP_TEMP_EDGE);
+static SENSOR_DEVICE_ATTR(temp2_input, S_IRUGO, amdgpu_hwmon_show_temp, NULL, PP_TEMP_JUNCTION);
 static SENSOR_DEVICE_ATTR(temp2_crit, S_IRUGO, amdgpu_hwmon_show_hotspot_temp_thresh, NULL, 0);
 static SENSOR_DEVICE_ATTR(temp2_crit_hyst, S_IRUGO, amdgpu_hwmon_show_hotspot_temp_thresh, NULL, 1);
 static SENSOR_DEVICE_ATTR(temp2_emergency, S_IRUGO, amdgpu_hwmon_show_temp_emergency, NULL, PP_TEMP_JUNCTION);
+static SENSOR_DEVICE_ATTR(temp3_input, S_IRUGO, amdgpu_hwmon_show_temp, NULL, PP_TEMP_MEM);
 static SENSOR_DEVICE_ATTR(temp3_crit, S_IRUGO, amdgpu_hwmon_show_mem_temp_thresh, NULL, 0);
 static SENSOR_DEVICE_ATTR(temp3_crit_hyst, S_IRUGO, amdgpu_hwmon_show_mem_temp_thresh, NULL, 1);
 static SENSOR_DEVICE_ATTR(temp3_emergency, S_IRUGO, amdgpu_hwmon_show_temp_emergency, NULL, PP_TEMP_MEM);
@@ -2134,8 +2159,10 @@ static struct attribute *hwmon_attributes[] = {
 	&sensor_dev_attr_temp1_input.dev_attr.attr,
 	&sensor_dev_attr_temp1_crit.dev_attr.attr,
 	&sensor_dev_attr_temp1_crit_hyst.dev_attr.attr,
+	&sensor_dev_attr_temp2_input.dev_attr.attr,
 	&sensor_dev_attr_temp2_crit.dev_attr.attr,
 	&sensor_dev_attr_temp2_crit_hyst.dev_attr.attr,
+	&sensor_dev_attr_temp3_input.dev_attr.attr,
 	&sensor_dev_attr_temp3_crit.dev_attr.attr,
 	&sensor_dev_attr_temp3_crit_hyst.dev_attr.attr,
 	&sensor_dev_attr_temp1_emergency.dev_attr.attr,
@@ -2272,7 +2299,9 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj,
 	     attr == &sensor_dev_attr_temp3_crit_hyst.dev_attr.attr ||
 	     attr == &sensor_dev_attr_temp1_emergency.dev_attr.attr ||
 	     attr == &sensor_dev_attr_temp2_emergency.dev_attr.attr ||
-	     attr == &sensor_dev_attr_temp3_emergency.dev_attr.attr))
+	     attr == &sensor_dev_attr_temp3_emergency.dev_attr.attr ||
+	     attr == &sensor_dev_attr_temp2_input.dev_attr.attr ||
+	     attr == &sensor_dev_attr_temp3_input.dev_attr.attr))
 		return 0;
 
 	return effective_mode;

drivers/gpu/drm/amd/include/kgd_pp_interface.h

@@ -111,6 +111,9 @@ enum amd_pp_sensors {
 	AMDGPU_PP_SENSOR_GPU_LOAD,
 	AMDGPU_PP_SENSOR_GFX_MCLK,
 	AMDGPU_PP_SENSOR_GPU_TEMP,
+	AMDGPU_PP_SENSOR_EDGE_TEMP = AMDGPU_PP_SENSOR_GPU_TEMP,
+	AMDGPU_PP_SENSOR_HOTSPOT_TEMP,
+	AMDGPU_PP_SENSOR_MEM_TEMP,
 	AMDGPU_PP_SENSOR_VCE_POWER,
 	AMDGPU_PP_SENSOR_UVD_POWER,
 	AMDGPU_PP_SENSOR_GPU_POWER,

drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c

@@ -3785,6 +3785,18 @@ static int vega10_read_sensor(struct pp_hwmgr *hwmgr, int idx,
 		*((uint32_t *)value) = vega10_thermal_get_temperature(hwmgr);
 		*size = 4;
 		break;
+	case AMDGPU_PP_SENSOR_HOTSPOT_TEMP:
+		smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetTemperatureHotspot);
+		*((uint32_t *)value) = smum_get_argument(hwmgr) *
+			PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
+		*size = 4;
+		break;
+	case AMDGPU_PP_SENSOR_MEM_TEMP:
+		smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetTemperatureHBM);
+		*((uint32_t *)value) = smum_get_argument(hwmgr) *
+			PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
+		*size = 4;
+		break;
 	case AMDGPU_PP_SENSOR_UVD_POWER:
 		*((uint32_t *)value) = data->uvd_power_gated ? 0 : 1;
 		*size = 4;

drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c

@@ -1338,6 +1338,7 @@ static int vega12_read_sensor(struct pp_hwmgr *hwmgr, int idx,
 			      void *value, int *size)
 {
 	struct vega12_hwmgr *data = (struct vega12_hwmgr *)(hwmgr->backend);
+	SmuMetrics_t metrics_table;
 	int ret = 0;
 
 	switch (idx) {
@@ -1360,6 +1361,24 @@ static int vega12_read_sensor(struct pp_hwmgr *hwmgr, int idx,
 		*((uint32_t *)value) = vega12_thermal_get_temperature(hwmgr);
 		*size = 4;
 		break;
+	case AMDGPU_PP_SENSOR_HOTSPOT_TEMP:
+		ret = vega12_get_metrics_table(hwmgr, &metrics_table);
+		if (ret)
+			return ret;
+
+		*((uint32_t *)value) = metrics_table.TemperatureHotspot *
+			PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
+		*size = 4;
+		break;
+	case AMDGPU_PP_SENSOR_MEM_TEMP:
+		ret = vega12_get_metrics_table(hwmgr, &metrics_table);
+		if (ret)
+			return ret;
+
+		*((uint32_t *)value) = metrics_table.TemperatureHBM *
+			PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
+		*size = 4;
+		break;
 	case AMDGPU_PP_SENSOR_UVD_POWER:
 		*((uint32_t *)value) = data->uvd_power_gated ? 0 : 1;
 		*size = 4;

drivers/gpu/drm/amd/powerplay/hwmgr/vega20_hwmgr.c

@@ -2138,10 +2138,28 @@ static int vega20_read_sensor(struct pp_hwmgr *hwmgr, int idx,
 		if (!ret)
 			*size = 4;
 		break;
-	case AMDGPU_PP_SENSOR_GPU_TEMP:
+	case AMDGPU_PP_SENSOR_HOTSPOT_TEMP:
 		*((uint32_t *)value) = vega20_thermal_get_temperature(hwmgr);
 		*size = 4;
 		break;
+	case AMDGPU_PP_SENSOR_EDGE_TEMP:
+		ret = vega20_get_metrics_table(hwmgr, &metrics_table);
+		if (ret)
+			return ret;
+
+		*((uint32_t *)value) = metrics_table.TemperatureEdge *
+			PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
+		*size = 4;
+		break;
+	case AMDGPU_PP_SENSOR_MEM_TEMP:
+		ret = vega20_get_metrics_table(hwmgr, &metrics_table);
+		if (ret)
+			return ret;
+
+		*((uint32_t *)value) = metrics_table.TemperatureHBM *
+			PP_TEMPERATURE_UNITS_PER_CENTIGRADES;
+		*size = 4;
+		break;
 	case AMDGPU_PP_SENSOR_UVD_POWER:
 		*((uint32_t *)value) = data->uvd_power_gated ? 0 : 1;
 		*size = 4;