Optimize CPU Model Feature Validation for Node Labeller

**Issue:**
Libvirt validation fails when a CPU model is usable
by QEMU but lacks features listed in
`/usr/share/libvirt/cpu_map/[CPU Model].xml` on a node.

**Resolution:**
Nodes are now labeled with a CPU model only if all
features listed in the corresponding `cpu_map` XML file
are present and the model is usable by QEMU.
This prevents validation issues caused by missing features.

**Libvirt's Recommendation:**
Libvirt advises against using its internal data files for
feature validation.
As recommended, we're exploring the use only Libvirt's API
to identify usable CPU models, avoiding direct `cpu_map`
file dependency.

**Workaround:**
To avoid the validation error mentioned above we will disable
some of the features in the `/usr/share/libvirt/cpu_map/[CPU Model].xml` files:

- disable svm for Opteron_G2 to avoid:
https://bugzilla.redhat.com/show_bug.cgi?id=2122283
- disable mpx for Skylake, Cascadelake, and Icelake to avoid:
https://gitlab.com/libvirt/libvirt/-/issues/304

These features are disabled only if not explicitly required by the user,
allowing broader CPU model support when svm or mpx features are absent.

In addition, for future issues we allow setting
a config map called "cpu-model-to-features-to-disable"
in the installation namespace for quick problematic
features disablement.

Signed-off-by: bmordeha <bmordeha@redhat.com>

cmd/virt-handler/virt-handler.go

@@ -350,6 +350,7 @@ func (app *virtHandlerApp) Run() {
 		app.KubeletPodsDir,
 		vmiSourceInformer,
 		vmiTargetInformer,
+		factory.CpuModelToFeaturesToDisableConfigMap(),
 		domainSharedInformer,
 		int(app.WatchdogTimeoutDuration.Seconds()),
 		app.MaxDevices,

pkg/controller/virtinformers.go

@@ -165,6 +165,9 @@ type KubeInformerFactory interface {
 	// Watches for the export route config map
 	ExportRouteConfigMap() cache.SharedIndexInformer
 
+	// Watches for cpu model config map
+	CpuModelToFeaturesToDisableConfigMap() cache.SharedIndexInformer
+
 	// Watches for the kubevirt export service
 	ExportService() cache.SharedIndexInformer
 
@@ -892,6 +895,15 @@ func (f *kubeInformerFactory) ExportRouteConfigMap() cache.SharedIndexInformer {
 	})
 }
 
+func (f *kubeInformerFactory) CpuModelToFeaturesToDisableConfigMap() cache.SharedIndexInformer {
+	return f.getInformer("cpuModelToFeaturesToDisable", func() cache.SharedIndexInformer {
+		restClient := f.clientSet.CoreV1().RESTClient()
+		fieldSelector := fields.OneTermEqualSelector("metadata.name", "cpu-model-to-features-to-disable")
+		lw := cache.NewListWatchFromClient(restClient, "configmaps", f.kubevirtNamespace, fieldSelector)
+		return cache.NewSharedIndexInformer(lw, &k8sv1.ConfigMap{}, f.defaultResync, cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc})
+	})
+}
+
 func (f *kubeInformerFactory) ExportService() cache.SharedIndexInformer {
 	return f.getInformer("exportService", func() cache.SharedIndexInformer {
 		// Watch all service with the kubevirt app label

pkg/handler-launcher-com/cmd/v1/cmd.proto

@@ -104,6 +104,10 @@ message InterfaceBindingMigration{
   string Method = 1;
 }
 
+message FeaturesToDisable{
+  repeated string features = 1;
+}
+
 message VirtualMachineOptions {
   SMBios VirtualMachineSMBios = 1;
   uint32 MemBalloonStatsPeriod = 2;
@@ -115,6 +119,7 @@ message VirtualMachineOptions {
   ClusterConfig clusterConfig = 7;
   map<string, string> interfaceDomainAttachment = 8;
   map<string, InterfaceBindingMigration> interfaceMigration = 9;
+  map<string, FeaturesToDisable> cpuModelToFeaturesToDisable = 10;
 }
 
 message VMIRequest {

pkg/virt-config/BUILD.bazel

@@ -10,6 +10,7 @@ go_library(
     importpath = "kubevirt.io/kubevirt/pkg/virt-config",
     visibility = ["//visibility:public"],
     deps = [
+        "//pkg/handler-launcher-com/cmd/v1:go_default_library",
         "//pkg/pointer:go_default_library",
         "//pkg/virt-config/deprecation:go_default_library",
         "//staging/src/kubevirt.io/api/core/v1:go_default_library",

pkg/virt-config/virt-config.go

@@ -25,13 +25,16 @@ package virtconfig
 
 import (
 	"fmt"
-
-	"kubevirt.io/client-go/log"
+	"strings"
 
 	k8sv1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
+	"k8s.io/client-go/tools/cache"
 
 	v1 "kubevirt.io/api/core/v1"
+	"kubevirt.io/client-go/log"
+
+	v12 "kubevirt.io/kubevirt/pkg/handler-launcher-com/cmd/v1"
 )
 
 const (
@@ -101,6 +104,58 @@ func IsPPC64(arch string) bool {
 	return arch == "ppc64le"
 }
 
+func (c *ClusterConfig) GetCpuModelToFeaturesToDisableMap(cpuModelConfigMapInformer cache.SharedIndexInformer) map[string]*v12.FeaturesToDisable {
+	/*
+					Libvirt validation fails when a CPU model is usable
+					by QEMU but lacks features listed in
+					`/usr/share/libvirt/cpu_map/[CPU Model].xml` on a node
+
+					To avoid the validation error mentioned above we will disable
+					some of the features in the `/usr/share/libvirt/cpu_map/[CPU Model].xml` files
+					for some of the cpu models.
+					Examples of validation error:
+		    		https://bugzilla.redhat.com/show_bug.cgi?id=2122283
+					https://gitlab.com/libvirt/libvirt/-/issues/304
+
+					Additionally, to handle future issues, we allow setting a config map named
+					“cpu-model-to-features-to-disable” in the installation namespace.
+					This enables quick and targeted disabling of problematic features.
+	*/
+
+	cpuModelToFeaturesToDisable := map[string]*v12.FeaturesToDisable{
+		"Opteron_G2":                {Features: []string{"svm"}},
+		"Cascadelake-Server-noTSX":  {Features: []string{"mpx"}},
+		"Cascadelake-Server":        {Features: []string{"mpx"}},
+		"Icelake-Client-noTSX":      {Features: []string{"mpx"}},
+		"Icelake-Client":            {Features: []string{"mpx"}},
+		"Icelake-Server-noTSX":      {Features: []string{"mpx"}},
+		"Icelake-Server":            {Features: []string{"mpx"}},
+		"Skylake-Client-IBRS":       {Features: []string{"mpx"}},
+		"Skylake-Client-noTSX-IBRS": {Features: []string{"mpx"}},
+		"Skylake-Client":            {Features: []string{"mpx"}},
+		"Skylake-Server-IBRS":       {Features: []string{"mpx"}},
+		"Skylake-Server-noTSX-IBRS": {Features: []string{"mpx"}},
+		"Skylake-Server":            {Features: []string{"mpx"}},
+	}
+
+	cachedObjs := cpuModelConfigMapInformer.GetIndexer().List()
+
+	if len(cachedObjs) != 1 {
+		return cpuModelToFeaturesToDisable
+	}
+	cpuModelToFeaturesToDisableCM := cachedObjs[0].(*k8sv1.ConfigMap)
+
+	for model, featuresString := range cpuModelToFeaturesToDisableCM.Data {
+		features := strings.Split(featuresString, ",")
+		knowFeatures := []string{}
+		if _, ok := cpuModelToFeaturesToDisable[model]; ok {
+			knowFeatures = cpuModelToFeaturesToDisable[model].Features
+		}
+		cpuModelToFeaturesToDisable[model] = &v12.FeaturesToDisable{Features: mergeStringSlices(knowFeatures, features)}
+	}
+	return cpuModelToFeaturesToDisable
+}
+
 func (c *ClusterConfig) GetMemBalloonStatsPeriod() uint32 {
 	return *c.GetConfig().MemBalloonStatsPeriod
 }
@@ -476,3 +531,20 @@ func (c *ClusterConfig) GetNetworkBindings() map[string]v1.InterfaceBindingPlugi
 	}
 	return nil
 }
+
+func mergeStringSlices(slice1, slice2 []string) []string {
+	mergedMap := make(map[string]bool)
+	for _, s := range slice1 {
+		mergedMap[s] = true
+	}
+
+	for _, s := range slice2 {
+		mergedMap[s] = true
+	}
+
+	mergedSlice := make([]string, 0, len(mergedMap))
+	for s := range mergedMap {
+		mergedSlice = append(mergedSlice, s)
+	}
+	return mergedSlice
+}

pkg/virt-handler/vm.go

@@ -199,6 +199,7 @@ func NewController(
 	kubeletPodsDir string,
 	vmiSourceInformer cache.SharedIndexInformer,
 	vmiTargetInformer cache.SharedIndexInformer,
+	cpuModelConfigMapInformer cache.SharedIndexInformer,
 	domainInformer cache.SharedInformer,
 	watchdogTimeoutSeconds int,
 	maxDevices int,
@@ -222,6 +223,7 @@ func NewController(
 		vmiSourceInformer:           vmiSourceInformer,
 		vmiTargetInformer:           vmiTargetInformer,
 		domainInformer:              domainInformer,
+		cpuModelConfigMapInformer:   cpuModelConfigMapInformer,
 		heartBeatInterval:           1 * time.Minute,
 		watchdogTimeoutSeconds:      watchdogTimeoutSeconds,
 		migrationProxy:              migrationProxy,
@@ -298,27 +300,28 @@ func NewController(
 }
 
 type VirtualMachineController struct {
-	recorder                 record.EventRecorder
-	clientset                kubecli.KubevirtClient
-	host                     string
-	migrationIpAddress       string
-	virtShareDir             string
-	virtPrivateDir           string
-	Queue                    workqueue.RateLimitingInterface
-	vmiSourceInformer        cache.SharedIndexInformer
-	vmiTargetInformer        cache.SharedIndexInformer
-	domainInformer           cache.SharedInformer
-	launcherClients          virtcache.LauncherClientInfoByVMI
-	heartBeatInterval        time.Duration
-	watchdogTimeoutSeconds   int
-	deviceManagerController  *device_manager.DeviceController
-	migrationProxy           migrationproxy.ProxyManager
-	podIsolationDetector     isolation.PodIsolationDetector
-	containerDiskMounter     container_disk.Mounter
-	hotplugVolumeMounter     hotplug_volume.VolumeMounter
-	clusterConfig            *virtconfig.ClusterConfig
-	sriovHotplugExecutorPool *executor.RateLimitedExecutorPool
-	downwardMetricsManager   downwardMetricsManager
+	recorder                  record.EventRecorder
+	clientset                 kubecli.KubevirtClient
+	host                      string
+	migrationIpAddress        string
+	virtShareDir              string
+	virtPrivateDir            string
+	Queue                     workqueue.RateLimitingInterface
+	vmiSourceInformer         cache.SharedIndexInformer
+	vmiTargetInformer         cache.SharedIndexInformer
+	domainInformer            cache.SharedInformer
+	cpuModelConfigMapInformer cache.SharedIndexInformer
+	launcherClients           virtcache.LauncherClientInfoByVMI
+	heartBeatInterval         time.Duration
+	watchdogTimeoutSeconds    int
+	deviceManagerController   *device_manager.DeviceController
+	migrationProxy            migrationproxy.ProxyManager
+	podIsolationDetector      isolation.PodIsolationDetector
+	containerDiskMounter      container_disk.Mounter
+	hotplugVolumeMounter      hotplug_volume.VolumeMounter
+	clusterConfig             *virtconfig.ClusterConfig
+	sriovHotplugExecutorPool  *executor.RateLimitedExecutorPool
+	downwardMetricsManager    downwardMetricsManager
 
 	netConf netconf
 	netStat netstat
@@ -1621,7 +1624,9 @@ func (c *VirtualMachineController) Run(threadiness int, stopCh chan struct{}) {
 
 	go c.downwardMetricsManager.Run(stopCh)
 
-	cache.WaitForCacheSync(stopCh, c.domainInformer.HasSynced, c.vmiSourceInformer.HasSynced, c.vmiTargetInformer.HasSynced)
+	go c.cpuModelConfigMapInformer.Run(stopCh)
+
+	cache.WaitForCacheSync(stopCh, c.domainInformer.HasSynced, c.vmiSourceInformer.HasSynced, c.vmiTargetInformer.HasSynced, c.cpuModelConfigMapInformer.HasSynced)
 
 	// Queue keys for previous Domains on the host that no longer exist
 	// in the cache. This ensures we perform local cleanup of deleted VMs.
@@ -3095,8 +3100,8 @@ func (d *VirtualMachineController) vmUpdateHelperDefault(origVMI *v1.VirtualMach
 
 	smbios := d.clusterConfig.GetSMBIOS()
 	period := d.clusterConfig.GetMemBalloonStatsPeriod()
-
 	options := virtualMachineOptions(smbios, period, preallocatedVolumes, d.capabilities, disksInfo, d.clusterConfig)
+	options.CpuModelToFeaturesToDisable = d.clusterConfig.GetCpuModelToFeaturesToDisableMap(d.cpuModelConfigMapInformer)
 	options.InterfaceDomainAttachment = domainspec.DomainAttachmentByInterfaceName(vmi.Spec.Domain.Devices.Interfaces, d.clusterConfig.GetNetworkBindings())
 
 	err = client.SyncVirtualMachine(vmi, options)

pkg/virt-handler/vm_test.go

@@ -95,6 +95,7 @@ var _ = Describe("VirtualMachineInstance", func() {
 	var vmiSource *framework.FakeControllerSource
 	var vmiSourceInformer cache.SharedIndexInformer
 	var vmiTargetInformer cache.SharedIndexInformer
+	var cpuModelCMInformer cache.SharedIndexInformer
 	var domainSource *framework.FakeControllerSource
 	var domainInformer cache.SharedIndexInformer
 	var mockQueue *testutils.MockWorkQueue
@@ -177,6 +178,7 @@ var _ = Describe("VirtualMachineInstance", func() {
 
 		vmiSourceInformer, vmiSource = testutils.NewFakeInformerFor(&v1.VirtualMachineInstance{})
 		vmiTargetInformer, _ = testutils.NewFakeInformerFor(&v1.VirtualMachineInstance{})
+		cpuModelCMInformer, _ = testutils.NewFakeInformerFor(&k8sv1.ConfigMap{})
 		domainInformer, domainSource = testutils.NewFakeInformerFor(&api.Domain{})
 		recorder = record.NewFakeRecorder(100)
 		recorder.IncludeObject = true
@@ -225,6 +227,7 @@ var _ = Describe("VirtualMachineInstance", func() {
 			podsDir,
 			vmiSourceInformer,
 			vmiTargetInformer,
+			cpuModelCMInformer,
 			domainInformer,
 			1,
 			10,

pkg/virt-launcher/virtwrap/converter/converter.go

@@ -134,6 +134,7 @@ type ConverterContext struct {
 	BochsForEFIGuests               bool
 	SerialConsoleLog                bool
 	DomainAttachmentByInterfaceName map[string]string
+	CpuModelToFeaturesToDisable     map[string]*cmdv1.FeaturesToDisable
 }
 
 func contains(volumes []string, name string) bool {
@@ -1827,15 +1828,28 @@ func Convert_v1_VirtualMachineInstance_To_api_Domain(vmi *v1.VirtualMachineInsta
 		}
 
 		// Set VM CPU features
+		existingFeatures := make(map[string]struct{})
 		if vmi.Spec.Domain.CPU.Features != nil {
 			for _, feature := range vmi.Spec.Domain.CPU.Features {
+				existingFeatures[feature.Name] = struct{}{}
 				domain.Spec.CPU.Features = append(domain.Spec.CPU.Features, api.CPUFeature{
 					Name:   feature.Name,
 					Policy: feature.Policy,
 				})
 			}
 		}
 
+		if featuresToDisable, ok := c.CpuModelToFeaturesToDisable[vmi.Spec.Domain.CPU.Model]; ok {
+			for _, feature := range featuresToDisable.Features {
+				if _, exists := existingFeatures[feature]; !exists {
+					domain.Spec.CPU.Features = append(domain.Spec.CPU.Features, api.CPUFeature{
+						Name:   feature,
+						Policy: "disable",
+					})
+				}
+			}
+		}
+
 		// Adjust guest vcpu config. Currently will handle vCPUs to pCPUs pinning
 		if vmi.IsCPUDedicated() {
 			err = vcpu.AdjustDomainForTopologyAndCPUSet(domain, vmi, c.Topology, c.CPUSet, useIOThreads)

pkg/virt-launcher/virtwrap/manager.go

@@ -983,6 +983,7 @@ func (l *LibvirtDomainManager) generateConverterContext(vmi *v1.VirtualMachineIn
 			c.SerialConsoleLog = isSerialConsoleLogEnabled(options.GetClusterConfig().GetSerialConsoleLogDisabled(), vmi)
 		}
 
+		c.CpuModelToFeaturesToDisable = options.CpuModelToFeaturesToDisable
 		c.DomainAttachmentByInterfaceName = options.GetInterfaceDomainAttachment()
 	}
 	c.DisksInfo = l.disksInfo