From: jsg <jsg@openbsd.org>
Date: Wed, 26 Jul 2023 06:17:53 +0000 (+0000)
Subject: drm/amdgpu: add the fan abnormal detection feature
X-Git-Url: http://artulab.com/gitweb/?a=commitdiff_plain;h=ff290ac45dd9bb0e33c99c93b809136c23bcb5df;p=openbsd

drm/amdgpu: add the fan abnormal detection feature

From lyndonli
d7d53c669da90181827e42583582192646a4d933 in linux-6.1.y/6.1.40
ef5fca9f7294509ee5013af9e879edc5837c1d6c in mainline linux
---

diff --git a/sys/dev/pci/drm/amd/pm/swsmu/inc/amdgpu_smu.h b/sys/dev/pci/drm/amd/pm/swsmu/inc/amdgpu_smu.h
index adcf99148c7..1bb953c9796 100644
--- a/sys/dev/pci/drm/amd/pm/swsmu/inc/amdgpu_smu.h
+++ b/sys/dev/pci/drm/amd/pm/swsmu/inc/amdgpu_smu.h
@@ -168,6 +168,7 @@ struct smu_temperature_range {
 	int mem_crit_max;
 	int mem_emergency_max;
 	int software_shutdown_temp;
+	int software_shutdown_temp_offset;
 };
 
 struct smu_state_validation_block {
diff --git a/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0.c b/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0.c
index d829f952560..a72fa30ea5e 100644
--- a/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0.c
+++ b/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0.c
@@ -1381,6 +1381,7 @@ static int smu_v13_0_irq_process(struct amdgpu_device *adev,
 	 */
 	uint32_t ctxid = entry->src_data[0];
 	uint32_t data;
+	uint32_t high;
 
 	if (client_id == SOC15_IH_CLIENTID_THM) {
 		switch (src_id) {
@@ -1437,6 +1438,33 @@ static int smu_v13_0_irq_process(struct amdgpu_device *adev,
 					schedule_work(&smu->throttling_logging_work);
 
 				break;
+			case 0x8:
+				high = smu->thermal_range.software_shutdown_temp +
+					smu->thermal_range.software_shutdown_temp_offset;
+				high = min(SMU_THERMAL_MAXIMUM_ALERT_TEMP, high);
+				dev_emerg(adev->dev, "Reduce soft CTF limit to %d (by an offset %d)\n",
+							high,
+							smu->thermal_range.software_shutdown_temp_offset);
+
+				data = RREG32_SOC15(THM, 0, regTHM_THERMAL_INT_CTRL);
+				data = REG_SET_FIELD(data, THM_THERMAL_INT_CTRL,
+							DIG_THERM_INTH,
+							(high & 0xff));
+				data = data & (~THM_THERMAL_INT_CTRL__THERM_TRIGGER_MASK_MASK);
+				WREG32_SOC15(THM, 0, regTHM_THERMAL_INT_CTRL, data);
+				break;
+			case 0x9:
+				high = min(SMU_THERMAL_MAXIMUM_ALERT_TEMP,
+					smu->thermal_range.software_shutdown_temp);
+				dev_emerg(adev->dev, "Recover soft CTF limit to %d\n", high);
+
+				data = RREG32_SOC15(THM, 0, regTHM_THERMAL_INT_CTRL);
+				data = REG_SET_FIELD(data, THM_THERMAL_INT_CTRL,
+							DIG_THERM_INTH,
+							(high & 0xff));
+				data = data & (~THM_THERMAL_INT_CTRL__THERM_TRIGGER_MASK_MASK);
+				WREG32_SOC15(THM, 0, regTHM_THERMAL_INT_CTRL, data);
+				break;
 			}
 		}
 	}
diff --git a/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c b/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c
index 57579f5e00e..67cbd2efe2b 100644
--- a/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c
+++ b/sys/dev/pci/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c
@@ -1288,6 +1288,7 @@ static int smu_v13_0_7_get_thermal_temperature_range(struct smu_context *smu,
 	range->mem_emergency_max = (pptable->SkuTable.TemperatureLimit[TEMP_MEM] + CTF_OFFSET_MEM)*
 		SMU_TEMPERATURE_UNITS_PER_CENTIGRADES;
 	range->software_shutdown_temp = powerplay_table->software_shutdown_temp;
+	range->software_shutdown_temp_offset = pptable->SkuTable.FanAbnormalTempLimitOffset;
 
 	return 0;
 }