drm/amd/pm: Use gpu metrics 1.9 for SMUv13.0.12

Fill and publish GPU metrics in v1.9 format for SMUv13.0.12 SOCs

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

3 files changed, 51 insertions, 67 deletions

diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c
index a0c844bf852c..f4bf0b558d86 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_12_ppt.c
@@ -58,7 +58,7 @@
 
 #define NUM_JPEG_RINGS_FW	10
 #define NUM_JPEG_RINGS_GPU_METRICS(gpu_metrics) \
-	(ARRAY_SIZE(gpu_metrics->xcp_stats[0].jpeg_busy) / 4)
+	(ARRAY_SIZE(gpu_metrics->jpeg_busy) / 4)
 
 const struct cmn2asic_mapping smu_v13_0_12_feature_mask_map[SMU_FEATURE_COUNT] = {
 	SMU_13_0_12_FEA_MAP(SMU_FEATURE_DATA_CALCULATIONS_BIT, 		FEATURE_DATA_CALCULATION),
@@ -772,22 +772,17 @@ ssize_t smu_v13_0_12_get_xcp_metrics(struct smu_context *smu, struct amdgpu_xcp
 	return sizeof(*xcp_metrics);
 }
 
-ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, void *smu_metrics)
+void smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table,
+				  void *smu_metrics,
+				  struct smu_v13_0_6_gpu_metrics *gpu_metrics)
 {
-	struct smu_table_context *smu_table = &smu->smu_table;
-	struct gpu_metrics_v1_8 *gpu_metrics =
-		(struct gpu_metrics_v1_8 *)smu_table->gpu_metrics_table;
-	int ret = 0, xcc_id, inst, i, j, k, idx;
 	struct amdgpu_device *adev = smu->adev;
+	int ret = 0, xcc_id, inst, i, j;
 	u8 num_jpeg_rings_gpu_metrics;
 	MetricsTable_t *metrics;
-	struct amdgpu_xcp *xcp;
-	u32 inst_mask;
 
 	metrics = (MetricsTable_t *)smu_metrics;
 
-	smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 8);
-
 	gpu_metrics->temperature_hotspot =
 		SMUQ10_ROUND(metrics->MaxSocketTemperature);
 	/* Individual HBM stack temperature is not reported */
@@ -877,57 +872,47 @@ ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, void
 			gpu_metrics->xgmi_link_status[j] = ret;
 	}
 
-	gpu_metrics->num_partition = adev->xcp_mgr->num_xcps;
-
 	num_jpeg_rings_gpu_metrics = NUM_JPEG_RINGS_GPU_METRICS(gpu_metrics);
-	for_each_xcp(adev->xcp_mgr, xcp, i) {
-		amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask);
-		idx = 0;
-		for_each_inst(k, inst_mask) {
-			/* Both JPEG and VCN has same instances */
-			inst = GET_INST(VCN, k);
-
-			for (j = 0; j < num_jpeg_rings_gpu_metrics; ++j) {
-				gpu_metrics->xcp_stats[i].jpeg_busy
-					[(idx * num_jpeg_rings_gpu_metrics) + j] =
-					SMUQ10_ROUND(metrics->JpegBusy
-							[(inst * NUM_JPEG_RINGS_FW) + j]);
-			}
-			gpu_metrics->xcp_stats[i].vcn_busy[idx] =
-			       SMUQ10_ROUND(metrics->VcnBusy[inst]);
-			idx++;
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		inst = GET_INST(VCN, i);
+
+		for (j = 0; j < num_jpeg_rings_gpu_metrics; ++j) {
+			gpu_metrics->jpeg_busy[(i * num_jpeg_rings_gpu_metrics) +
+					       j] =
+				SMUQ10_ROUND(
+					metrics->JpegBusy[(inst *
+							   NUM_JPEG_RINGS_FW) +
+							  j]);
 		}
+		gpu_metrics->vcn_busy[i] = SMUQ10_ROUND(metrics->VcnBusy[inst]);
+	}
 
-		amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask);
-		idx = 0;
-		for_each_inst(k, inst_mask) {
-			inst = GET_INST(GC, k);
-			gpu_metrics->xcp_stats[i].gfx_busy_inst[idx] =
-				SMUQ10_ROUND(metrics->GfxBusy[inst]);
-			gpu_metrics->xcp_stats[i].gfx_busy_acc[idx] =
-				SMUQ10_ROUND(metrics->GfxBusyAcc[inst]);
-			if (smu_v13_0_6_cap_supported(smu, SMU_CAP(HST_LIMIT_METRICS))) {
-				gpu_metrics->xcp_stats[i].gfx_below_host_limit_ppt_acc[idx] =
-					SMUQ10_ROUND(metrics->GfxclkBelowHostLimitPptAcc[inst]);
-				gpu_metrics->xcp_stats[i].gfx_below_host_limit_thm_acc[idx] =
-					SMUQ10_ROUND(metrics->GfxclkBelowHostLimitThmAcc[inst]);
-				gpu_metrics->xcp_stats[i].gfx_low_utilization_acc[idx] =
-					SMUQ10_ROUND(metrics->GfxclkLowUtilizationAcc[inst]);
-				gpu_metrics->xcp_stats[i].gfx_below_host_limit_total_acc[idx] =
-					SMUQ10_ROUND(metrics->GfxclkBelowHostLimitTotalAcc[inst]);
-			}
-			idx++;
-		}
+	for (i = 0; i < NUM_XCC(adev->gfx.xcc_mask); ++i) {
+		inst = GET_INST(GC, i);
+		gpu_metrics->gfx_busy_inst[i] =
+			SMUQ10_ROUND(metrics->GfxBusy[inst]);
+		gpu_metrics->gfx_busy_acc[i] =
+			SMUQ10_ROUND(metrics->GfxBusyAcc[inst]);
+		if (smu_v13_0_6_cap_supported(smu,
+					      SMU_CAP(HST_LIMIT_METRICS))) {
+			gpu_metrics
+				->gfx_below_host_limit_ppt_acc[i] = SMUQ10_ROUND(
+				metrics->GfxclkBelowHostLimitPptAcc[inst]);
+			gpu_metrics
+				->gfx_below_host_limit_thm_acc[i] = SMUQ10_ROUND(
+				metrics->GfxclkBelowHostLimitThmAcc[inst]);
+			gpu_metrics->gfx_low_utilization_acc[i] = SMUQ10_ROUND(
+				metrics->GfxclkLowUtilizationAcc[inst]);
+			gpu_metrics->gfx_below_host_limit_total_acc
+				[i] = SMUQ10_ROUND(
+				metrics->GfxclkBelowHostLimitTotalAcc[inst]);
+		};
 	}
 
 	gpu_metrics->xgmi_link_width = metrics->XgmiWidth;
 	gpu_metrics->xgmi_link_speed = metrics->XgmiBitrate;
 
 	gpu_metrics->firmware_timestamp = metrics->Timestamp;
-
-	*table = (void *)gpu_metrics;
-
-	return sizeof(*gpu_metrics);
 }
 
 const struct smu_temp_funcs smu_v13_0_12_temp_funcs = {
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
index 6d39b02a3257..556116712171 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
@@ -549,7 +549,6 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu)
 {
 	struct smu_table_context *smu_table = &smu->smu_table;
 	struct smu_table *tables = smu_table->tables;
-	void *gpu_metrics_table __free(kfree) = NULL;
 	struct smu_v13_0_6_gpu_metrics *gpu_metrics;
 	void *driver_pptable __free(kfree) = NULL;
 	void *metrics_table __free(kfree) = NULL;
@@ -580,12 +579,6 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu)
 		return -ENOMEM;
 	smu_table->metrics_time = 0;
 
-	smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_8);
-	gpu_metrics_table =
-		kzalloc(smu_table->gpu_metrics_table_size, GFP_KERNEL);
-	if (!gpu_metrics_table)
-		return -ENOMEM;
-
 	driver_pptable = kzalloc(sizeof(struct PPTable_t), GFP_KERNEL);
 	if (!driver_pptable)
 		return -ENOMEM;
@@ -608,7 +601,6 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu)
 		}
 	}
 
-	smu_table->gpu_metrics_table = no_free_ptr(gpu_metrics_table);
 	smu_table->metrics_table = no_free_ptr(metrics_table);
 	smu_table->driver_pptable = no_free_ptr(driver_pptable);
 
@@ -2792,16 +2784,20 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 	if (ret)
 		return ret;
 
-	if (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) ==
-		    IP_VERSION(13, 0, 12) &&
-	    smu_v13_0_6_cap_supported(smu, SMU_CAP(STATIC_METRICS)))
-		return smu_v13_0_12_get_gpu_metrics(smu, table, metrics_v0);
-
-	metrics_v1 = (MetricsTableV1_t *)metrics_v0;
 	metrics_v2 = (MetricsTableV2_t *)metrics_v0;
 	gpu_metrics = (struct smu_v13_0_6_gpu_metrics
 			       *)(tables[SMU_TABLE_SMU_METRICS].cache.buffer);
 
+	if (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 12) &&
+	    smu_v13_0_6_cap_supported(smu, SMU_CAP(STATIC_METRICS))) {
+		smu_v13_0_12_get_gpu_metrics(smu, table, metrics_v0,
+					     gpu_metrics);
+		goto fill;
+	}
+
+	metrics_v1 = (MetricsTableV1_t *)metrics_v0;
+	metrics_v2 = (MetricsTableV2_t *)metrics_v0;
+
 	gpu_metrics->temperature_hotspot =
 		SMUQ10_ROUND(GET_METRIC_FIELD(MaxSocketTemperature, version));
 	/* Individual HBM stack temperature is not reported */
@@ -2974,6 +2970,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 
 	gpu_metrics->firmware_timestamp = GET_METRIC_FIELD(Timestamp, version);
 
+fill:
 	*table = tables[SMU_TABLE_SMU_METRICS].cache.buffer;
 
 	return sizeof(*gpu_metrics);
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h
index 3f57e2a33fb4..86d82044a255 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.h
@@ -94,7 +94,6 @@ size_t smu_v13_0_12_get_system_metrics_size(void);
 int smu_v13_0_12_setup_driver_pptable(struct smu_context *smu);
 int smu_v13_0_12_get_smu_metrics_data(struct smu_context *smu,
 				      MetricsMember_t member, uint32_t *value);
-ssize_t smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table, void *smu_metrics);
 ssize_t smu_v13_0_12_get_xcp_metrics(struct smu_context *smu,
 				     struct amdgpu_xcp *xcp, void *table,
 				     void *smu_metrics);
@@ -216,6 +215,9 @@ extern const struct smu_temp_funcs smu_v13_0_12_temp_funcs;
 		  SMU_13_0_6_MAX_XCC);
 
 DECLARE_SMU_METRICS_CLASS(smu_v13_0_6_gpu_metrics, SMU_13_0_6_METRICS_FIELDS);
+void smu_v13_0_12_get_gpu_metrics(struct smu_context *smu, void **table,
+				  void *smu_metrics,
+				  struct smu_v13_0_6_gpu_metrics *gpu_metrics);
 
 #endif /* SWSMU_CODE_LAYER_L2 */