Advertise RDT L3 num_closid

Signed-off-by: PiotrProkop <pprokop@nvidia.com>

docs/usage/customization-guide.md

@@ -601,8 +601,9 @@ The following features are available for matching:
 |                  |              | **`status`** | string | Status of the driver, possible values are 'active' and 'passive'
 |                  |              | **`turbo`**  | bool   | 'true' if turbo frequencies are enabled, otherwise 'false'
 |                  |              | **`scaling`** | string | Active scaling_governor, possible values are 'powersave' or 'performance'.
-| **`cpu.rdt`**    | flag         |          |            | Intel RDT capabilities supported by the system
+| **`cpu.rdt`**    | attribute    |          |            | Intel RDT capabilities supported by the system
 |                  |              | **`<rdt-flag>`** |    | RDT capability is supported, see [RDT flags](#intel-rdt-flags) for details
+|                  |              | **`RDTL3CA_NUM_CLOSID`** | int  | The number or available CLOSID (Class of service ID) for Intel L3 Cache Allocation Technology
 | **`cpu.security`** | attribute  |          |            | Features related to security and trusted execution environments
 |                  |              | **`sgx.enabled`** | bool | `true` if Intel SGX (Software Guard Extensions) has been enabled, otherwise does not exist
 |                  |              | **`se.enabled`** | bool  | `true` if IBM Secure Execution for Linux is available and has been enabled, otherwise does not exist

source/cpu/cpu.go

@@ -164,8 +164,12 @@ func (s *cpuSource) GetLabels() (source.FeatureLabels, error) {
 	}
 
 	// RDT
-	for k := range features.Flags[RdtFeature].Elements {
-		labels["rdt."+k] = true
+	for k, v := range features.Attributes[RdtFeature].Elements {
+		if k == "RDTL3CA_NUM_CLOSID" {
+			continue
+		}
+
+		labels["rdt."+k] = v
 	}
 
 	// Security
@@ -227,7 +231,7 @@ func (s *cpuSource) Discover() error {
 	s.features.Attributes[PstateFeature] = nfdv1alpha1.NewAttributeFeatures(pstate)
 
 	// Detect RDT features
-	s.features.Flags[RdtFeature] = nfdv1alpha1.NewFlagFeatures(discoverRDT()...)
+	s.features.Attributes[RdtFeature] = nfdv1alpha1.NewAttributeFeatures(discoverRDT())
 
 	// Detect available guest protection(SGX,TDX,SEV) features
 	s.features.Attributes[SecurityFeature] = nfdv1alpha1.NewAttributeFeatures(discoverSecurity())

source/cpu/rdt_amd64.go

@@ -20,6 +20,14 @@ limitations under the License.
 package cpu
 
 import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strconv"
+
+	"github.com/opencontainers/runc/libcontainer/intelrdt"
+	"k8s.io/klog/v2"
+
 	"sigs.k8s.io/node-feature-discovery/pkg/cpuid"
 )
 
@@ -44,8 +52,8 @@ const (
 	RDT_ALLOCATION_EBX_MEMORY_BANDWIDTH_ALLOCATION              = 1 << 3
 )
 
-func discoverRDT() []string {
-	features := []string{}
+func discoverRDT() map[string]string {
+	attributes := map[string]string{}
 
 	// Read cpuid information
 	extFeatures := cpuid.Cpuid(LEAF_EXT_FEATURE_FLAGS, 0)
@@ -57,16 +65,16 @@ func discoverRDT() []string {
 	if extFeatures.EBX&EXT_FEATURE_FLAGS_EBX_RDT_M != 0 {
 		if rdtMonitoring.EDX&RDT_MONITORING_EDX_L3_MONITORING != 0 {
 			// Monitoring is supported
-			features = append(features, "RDTMON")
+			attributes["RDTMON"] = "true"
 
 			// Cache Monitoring Technology (L3 occupancy monitoring)
 			if rdtL3Monitoring.EDX&RDT_MONITORING_SUBLEAF_L3_EDX_L3_OCCUPANCY_MONITORING != 0 {
-				features = append(features, "RDTCMT")
+				attributes["RDTCMT"] = "true"
 			}
 			// Memore Bandwidth Monitoring (L3 local&total bandwidth monitoring)
 			if rdtL3Monitoring.EDX&RDT_MONITORING_SUBLEAF_L3_EDX_L3_TOTAL_BANDWIDTH_MONITORING != 0 &&
 				rdtL3Monitoring.EDX&RDT_MONITORING_SUBLEAF_L3_EDX_L3_LOCAL_BANDWIDTH_MONITORING != 0 {
-				features = append(features, "RDTMBM")
+				attributes["RDTMBM"] = "true"
 			}
 		}
 	}
@@ -75,17 +83,51 @@ func discoverRDT() []string {
 	if extFeatures.EBX&EXT_FEATURE_FLAGS_EBX_RDT_A != 0 {
 		// L3 Cache Allocation
 		if rdtAllocation.EBX&RDT_ALLOCATION_EBX_L3_CACHE_ALLOCATION != 0 {
-			features = append(features, "RDTL3CA")
+			attributes["RDTL3CA"] = "true"
+			numClosID := getNumClosID("L3")
+			if numClosID > -1 {
+				attributes["RDTL3CA_NUM_CLOSID"] = strconv.FormatInt(int64(numClosID), 10)
+			}
 		}
 		// L2 Cache Allocation
 		if rdtAllocation.EBX&RDT_ALLOCATION_EBX_L2_CACHE_ALLOCATION != 0 {
-			features = append(features, "RDTL2CA")
+			attributes["RDTL2CA"] = "true"
 		}
 		// Memory Bandwidth Allocation
 		if rdtAllocation.EBX&RDT_ALLOCATION_EBX_MEMORY_BANDWIDTH_ALLOCATION != 0 {
-			features = append(features, "RDTMBA")
+			attributes["RDTMBA"] = "true"
 		}
 	}
 
-	return features
+	return attributes
+}
+
+func getNumClosID(level string) int64 {
+	resctrlRootDir, err := intelrdt.Root()
+	if err != nil {
+		klog.V(4).ErrorS(err, "can't find resctrl filesystem")
+		return -1
+	}
+
+	closidFile := filepath.Join(resctrlRootDir, "info", level, "num_closids")
+
+	if _, err := os.Stat(closidFile); err != nil {
+		klog.V(4).ErrorS(err, "failed to stat file", "fileName", closidFile)
+		return -1
+	}
+
+	closidsBytes, err := os.ReadFile(filepath.Join(resctrlRootDir, "info", level, "num_closids"))
+	if err != nil {
+		klog.V(4).ErrorS(err, "failed to read file", "fileName", closidFile)
+		return -1
+	}
+
+	numClosIDs, err := strconv.ParseInt(string(bytes.TrimSpace(closidsBytes)), 10, 64)
+	if err != nil {
+		klog.V(4).ErrorS(err, "failed to ParseInt", "num_closids", string(bytes.TrimSpace(closidsBytes)))
+		return -1
+	}
+
+	// subtract 1 for default control group
+	return numClosIDs - 1
 }

source/cpu/rdt_stub.go

@@ -19,6 +19,6 @@ limitations under the License.
 
 package cpu
 
-func discoverRDT() []string {
-	return []string{}
+func discoverRDT() map[string]string {
+	return map[string]string{}
 }