drm/amd/powerplay: implement power_dpm_force_performance_level for SMU11

add get_performance_level and force_performance_level
to implement the sys interface for SMU11.
Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -264,7 +264,9 @@ static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return snprintf(buf, PAGE_SIZE, "off\n");
 
-	if (adev->powerplay.pp_funcs->get_performance_level)
+	if (is_support_sw_smu(adev))
+		level = smu_get_performance_level(&adev->smu);
+	else if (adev->powerplay.pp_funcs->get_performance_level)
 		level = amdgpu_dpm_get_performance_level(adev);
 	else
 		level = adev->pm.dpm.forced_level;
@@ -297,7 +299,9 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	     (ddev->switch_power_state != DRM_SWITCH_POWER_ON))
 		return -EINVAL;
 
-	if (adev->powerplay.pp_funcs->get_performance_level)
+	if (is_support_sw_smu(adev))
+		current_level = smu_get_performance_level(&adev->smu);
+	else if (adev->powerplay.pp_funcs->get_performance_level)
 		current_level = amdgpu_dpm_get_performance_level(adev);
 
 	if (strncmp("low", buf, strlen("low")) == 0) {
@@ -326,7 +330,20 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev,
 	if (current_level == level)
 		return count;
 
-	if (adev->powerplay.pp_funcs->force_performance_level) {
+	if (is_support_sw_smu(adev)) {
+		mutex_lock(&adev->pm.mutex);
+		if (adev->pm.dpm.thermal_active) {
+			count = -EINVAL;
+			mutex_unlock(&adev->pm.mutex);
+			goto fail;
+		}
+		ret = smu_force_performance_level(&adev->smu, level);
+		if (ret)
+			count = -EINVAL;
+		else
+			adev->pm.dpm.forced_level = level;
+		mutex_unlock(&adev->pm.mutex);
+	} else if (adev->powerplay.pp_funcs->force_performance_level) {
 		mutex_lock(&adev->pm.mutex);
 		if (adev->pm.dpm.thermal_active) {
 			count = -EINVAL;

drivers/gpu/drm/amd/powerplay/amdgpu_smu.c

@@ -351,6 +351,8 @@ static int smu_sw_init(void *handle)
 	smu->workload_setting[6] = PP_SMC_POWER_PROFILE_CUSTOM;
 	smu->display_config = &adev->pm.pm_display_cfg;
 
+	smu->smu_dpm.dpm_level = AMD_DPM_FORCED_LEVEL_AUTO;
+	smu->smu_dpm.requested_dpm_level = AMD_DPM_FORCED_LEVEL_AUTO;
 	ret = smu_init_microcode(smu);
 	if (ret) {
 		pr_err("Failed to load smu firmware!\n");

drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h

@@ -423,6 +423,8 @@ struct pptable_funcs {
 					      *clocks);
 	int (*get_power_profile_mode)(struct smu_context *smu, char *buf);
 	int (*set_power_profile_mode)(struct smu_context *smu, long *input, uint32_t size);
+	enum amd_dpm_forced_level (*get_performance_level)(struct smu_context *smu);
+	int (*force_performance_level)(struct smu_context *smu, enum amd_dpm_forced_level level);
 };
 
 struct smu_funcs
@@ -594,6 +596,10 @@ struct smu_funcs
 	((smu)->funcs->get_power_profile_mode ? (smu)->funcs->get_power_profile_mode((smu), buf) : 0)
 #define smu_set_power_profile_mode(smu, param, param_size) \
 	((smu)->funcs->set_power_profile_mode ? (smu)->funcs->set_power_profile_mode((smu), (param), (param_size)) : 0)
+#define smu_get_performance_level(smu) \
+	((smu)->ppt_funcs->get_performance_level ? (smu)->ppt_funcs->get_performance_level((smu)) : 0)
+#define smu_force_performance_level(smu, level) \
+	((smu)->ppt_funcs->force_performance_level ? (smu)->ppt_funcs->force_performance_level((smu), (level)) : 0)
 
 #define smu_msg_get_index(smu, msg) \
 	((smu)->ppt_funcs? ((smu)->ppt_funcs->get_smu_msg_index? (smu)->ppt_funcs->get_smu_msg_index((smu), (msg)) : -EINVAL) : -EINVAL)

drivers/gpu/drm/amd/powerplay/smu_v11_0.c

@@ -1100,7 +1100,6 @@ smu_v11_0_display_clock_voltage_request(struct smu_context *smu,
 	PPCLK_e clk_select = 0;
 	uint32_t clk_freq = clock_req->clock_freq_in_khz / 1000;
 
-	mutex_lock(&smu->mutex);
 	if (smu_feature_is_enabled(smu, FEATURE_DPM_DCEFCLK_BIT)) {
 		switch (clk_type) {
 		case amd_pp_dcef_clock:
@@ -1129,7 +1128,6 @@ smu_v11_0_display_clock_voltage_request(struct smu_context *smu,
 	}
 
 failed:
-	mutex_unlock(&smu->mutex);
 	return ret;
 }
 

drivers/gpu/drm/amd/powerplay/vega20_ppt.c

@@ -1744,6 +1744,103 @@ static int vega20_unforce_dpm_levels(struct smu_context *smu)
 	return ret;
 }
 
+static enum amd_dpm_forced_level vega20_get_performance_level(struct smu_context *smu)
+{
+	struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm);
+	if (!smu_dpm_ctx->dpm_context)
+		return -EINVAL;
+
+	if (smu_dpm_ctx->dpm_level != smu_dpm_ctx->saved_dpm_level) {
+		mutex_lock(&(smu->mutex));
+		smu_dpm_ctx->saved_dpm_level = smu_dpm_ctx->dpm_level;
+		mutex_unlock(&(smu->mutex));
+	}
+	return smu_dpm_ctx->dpm_level;
+}
+
+static int
+vega20_force_performance_level(struct smu_context *smu, enum amd_dpm_forced_level level)
+{
+	int ret = 0;
+	int index = 0;
+	int i = 0;
+	uint32_t sclk_mask, mclk_mask, soc_mask;
+	long workload;
+	struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm);
+	if (!smu_dpm_ctx->dpm_context)
+		return -EINVAL;
+
+	for (i = 0; i < smu->adev->num_ip_blocks; i++) {
+		if (smu->adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC)
+			break;
+	}
+	mutex_lock(&smu->mutex);
+	smu->adev->ip_blocks[i].version->funcs->enable_umd_pstate(smu, &level);
+	ret = vega20_display_config_changed(smu);
+	if (ret) {
+		pr_err("Failed to change display config!");
+		goto failed;
+	}
+	ret = vega20_apply_clocks_adjust_rules(smu);
+	if (ret) {
+		pr_err("Failed to apply clocks adjust rules!");
+		goto failed;
+	}
+	ret = vega20_notify_smc_dispaly_config(smu);
+	if (ret) {
+		pr_err("Failed to notify smc display config!");
+		goto failed;
+	}
+	switch (level) {
+	case AMD_DPM_FORCED_LEVEL_HIGH:
+		ret = vega20_force_dpm_highest(smu);
+		break;
+
+	case AMD_DPM_FORCED_LEVEL_LOW:
+		ret = vega20_force_dpm_lowest(smu);
+		break;
+
+	case AMD_DPM_FORCED_LEVEL_AUTO:
+		ret = vega20_unforce_dpm_levels(smu);
+		break;
+
+	case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD:
+	case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK:
+	case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK:
+	case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK:
+		ret = vega20_get_profiling_clk_mask(smu, level,
+						    &sclk_mask,
+						    &mclk_mask,
+						    &soc_mask);
+		if (ret)
+			goto failed;
+		vega20_force_clk_levels(smu, PP_SCLK, 1 << sclk_mask);
+		vega20_force_clk_levels(smu, PP_MCLK, 1 << mclk_mask);
+		break;
+
+	case AMD_DPM_FORCED_LEVEL_MANUAL:
+	case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT:
+	default:
+		break;
+	}
+
+	if (!ret)
+		smu_dpm_ctx->dpm_level = level;
+
+	if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL) {
+		index = fls(smu->workload_mask);
+		index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0;
+		workload = smu->workload_setting[index];
+
+		if (smu->power_profile_mode != workload)
+			smu->funcs->set_power_profile_mode(smu, &workload, 0);
+	}
+
+failed:
+	mutex_unlock(&smu->mutex);
+	return ret;
+}
+
 static const struct pptable_funcs vega20_ppt_funcs = {
 	.alloc_dpm_context = vega20_allocate_dpm_context,
 	.store_powerplay_table = vega20_store_powerplay_table,
@@ -1761,6 +1858,8 @@ static const struct pptable_funcs vega20_ppt_funcs = {
 	.get_clock_by_type_with_latency = vega20_get_clock_by_type_with_latency,
 	.set_default_od8_settings = vega20_set_default_od8_setttings,
 	.get_od_percentage = vega20_get_od_percentage,
+	.get_performance_level = vega20_get_performance_level,
+	.force_performance_level = vega20_force_performance_level,
 };
 
 void vega20_set_ppt_funcs(struct smu_context *smu)