Allow reporting cpu physical core count.

cpu.Info() reports a list of cpus and the list will be double in length when hyperthreading is enabled. This difference also scales cpu utilization. New config property: cpu.stats.physicalcore.enabled Default set to false to allow opt-in usage. When set to true domainmgr will: Take the last cpu.InfoStat's CoreID Add 1 (since CoreID starts at 0) Set HostMemory.Ncpus to that value. Signed-off-by: Andrew Durbin <[email protected]>
lf-edge · Oct 25, 2024 · 82fa369 · 82fa369
1 parent 8e7e500
commit 82fa369
Show file tree

Hide file tree

Showing 8 changed files with 32 additions and 10 deletions.
diff --git a/docs/CONFIG-PROPERTIES.md b/docs/CONFIG-PROPERTIES.md
@@ -4,6 +4,7 @@
 | ---- | ---- | ------- | ----------- |
 | app.allow.vnc | boolean | false (only local access) | allow access to EVE's VNC ports from external IPs |
 | app.fml.resolution | string | notset | Set system-wide value of forced resolution for applications running in FML mode, it can be one of [predefined](/pkg/pillar/types/global.go) FmlResolution* values. |
+| cpu.stats.physicalcore.enable | boolean  | false | Report Ncpus as physical cores instead of HyperThread/SMT cores |
 | timer.config.interval | integer in seconds | 60 | how frequently device gets config |
 | timer.cert.interval | integer in seconds | 1 day (24*3600) | how frequently device checks for new controller certificates |
 | timer.metric.interval  | integer in seconds | 60 | how frequently device reports metrics |

diff --git a/pkg/pillar/cmd/domainmgr/domainmgr.go b/pkg/pillar/cmd/domainmgr/domainmgr.go
@@ -107,6 +107,7 @@ type domainContext struct {
 	setInitialVgaAccess     bool
 	consoleAccess           bool
 	setInitialConsoleAccess bool
+	reportPhyCores          bool
 
 	GCInitialized       bool
 	domainBootRetryTime uint32 // In seconds
@@ -543,7 +544,7 @@ func Run(ps *pubsub.PubSub, loggerArg *logrus.Logger, logArg *base.LogObject, ar
 	var resources types.HostMemory
 	for i := 0; true; i++ {
 		delay := 10
-		resources, err = hyper.GetHostCPUMem()
+		resources, err = hyper.GetHostCPUMem(domainCtx.reportPhyCores)
 		if err == nil {
 			break
 		}
@@ -2639,6 +2640,7 @@ func handleGlobalConfigImpl(ctxArg interface{}, key string,
 			ctx.metricInterval = metricInterval
 		}
 		ctx.processCloudInitMultiPart = gcp.GlobalValueBool(types.ProcessCloudInitMultiPart)
+		ctx.reportPhyCores = gcp.GlobalValueBool(types.CpuStatsPhysicalCoreEnable)
 		ctx.GCInitialized = true
 	}
 	log.Functionf("handleGlobalConfigImpl done for %s. "+

diff --git a/pkg/pillar/cmd/domainmgr/metric.go b/pkg/pillar/cmd/domainmgr/metric.go
@@ -5,9 +5,10 @@ package domainmgr
 
 import (
 	"fmt"
-	"github.com/lf-edge/eve/pkg/pillar/hypervisor"
 	"time"
 
+	"github.com/lf-edge/eve/pkg/pillar/hypervisor"
+
 	"github.com/lf-edge/eve/pkg/pillar/flextimer"
 	"github.com/lf-edge/eve/pkg/pillar/types"
 	"github.com/shirou/gopsutil/cpu"
@@ -93,7 +94,7 @@ func logWatermarks(ctx *domainContext, status *types.DomainStatus, dm *types.Dom
 
 func getAndPublishMetrics(ctx *domainContext, hyper hypervisor.Hypervisor) {
 	dmList, _ := hyper.GetDomsCPUMem()
-	hm, err := hyper.GetHostCPUMem()
+	hm, err := hyper.GetHostCPUMem(ctx.reportPhyCores)
 	if err != nil {
 		log.Errorf("Cannot obtain HostCPUMem: %s", err)
 		return

diff --git a/pkg/pillar/hypervisor/containerd.go b/pkg/pillar/hypervisor/containerd.go
@@ -278,8 +278,8 @@ func (ctx ctrdContext) PCISameController(id1 string, id2 string) bool {
 	return types.PCISameController(id1, id2)
 }
 
-func (ctx ctrdContext) GetHostCPUMem() (types.HostMemory, error) {
-	return selfDomCPUMem()
+func (ctx ctrdContext) GetHostCPUMem(reportPhyCores bool) (types.HostMemory, error) {
+	return selfDomCPUMem(reportPhyCores)
 }
 
 const nanoSecToSec uint64 = 1000000000

diff --git a/pkg/pillar/hypervisor/hypervisor.go b/pkg/pillar/hypervisor/hypervisor.go
@@ -9,6 +9,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"strconv"
 	"strings"
 
 	"github.com/lf-edge/eve/pkg/pillar/base"
@@ -33,7 +34,7 @@ type Hypervisor interface {
 	PCIRelease(string) error
 	PCISameController(string, string) bool
 
-	GetHostCPUMem() (types.HostMemory, error)
+	GetHostCPUMem(reportPhyCores bool) (types.HostMemory, error)
 	GetDomsCPUMem() (map[string]types.DomainMetric, error)
 
 	GetCapabilities() (*types.Capabilities, error)
@@ -111,7 +112,7 @@ func GetAvailableHypervisors() (all []string, enabled []string) {
 	return
 }
 
-func selfDomCPUMem() (types.HostMemory, error) {
+func selfDomCPUMem(reportPhysCores bool) (types.HostMemory, error) {
 	hm := types.HostMemory{}
 	vm, err := mem.VirtualMemory()
 	if err != nil {
@@ -139,6 +140,20 @@ func selfDomCPUMem() (types.HostMemory, error) {
 		return hm, err
 	}
 	hm.Ncpus = uint32(len(info))
+	if reportPhysCores {
+		if len(info) < 1 {
+			return hm, nil
+		}
+		// The list should be ordered so that CoreIds for a 4 core / 8 thread
+		// CPU would be eg. 0,1,2,3,0,1,2,3.  Pull the last entry:
+		lastInfoStat := info[len(info)-1]
+		val, err := strconv.ParseInt(lastInfoStat.CoreID, 10, 32)
+		if err != nil {
+			return hm, err
+		}
+		// Account for coreid 0
+		hm.Ncpus = uint32(val) + 1
+	}
 	return hm, nil
 }
 

diff --git a/pkg/pillar/hypervisor/null.go b/pkg/pillar/hypervisor/null.go
@@ -166,8 +166,8 @@ func (ctx nullContext) PCISameController(id1 string, id2 string) bool {
 	return types.PCISameController(id1, id2)
 }
 
-func (ctx nullContext) GetHostCPUMem() (types.HostMemory, error) {
-	return selfDomCPUMem()
+func (ctx nullContext) GetHostCPUMem(reportPhyCores bool) (types.HostMemory, error) {
+	return selfDomCPUMem(reportPhyCores)
 }
 
 func (ctx nullContext) GetDomsCPUMem() (map[string]types.DomainMetric, error) {

diff --git a/pkg/pillar/hypervisor/xen.go b/pkg/pillar/hypervisor/xen.go
@@ -623,7 +623,7 @@ func (ctx xenContext) PCISameController(id1 string, id2 string) bool {
 	return false
 }
 
-func (ctx xenContext) GetHostCPUMem() (types.HostMemory, error) {
+func (ctx xenContext) GetHostCPUMem(reportPhyCores bool) (types.HostMemory, error) {
 	hm := types.HostMemory{}
 	ctrdSystemCtx, done := ctx.ctrdClient.CtrNewSystemServicesCtx()
 	defer done()

diff --git a/pkg/pillar/types/global.go b/pkg/pillar/types/global.go
@@ -256,6 +256,8 @@ const (
 	EnableARPSnoop GlobalSettingKey = "network.switch.enable.arpsnoop"
 	// WwanQueryVisibleProviders : periodically query visible cellular service providers
 	WwanQueryVisibleProviders GlobalSettingKey = "wwan.query.visible.providers"
+	// CpuStatsPhysicalCoreEnable: report Ncpus as Physical Cores instead of Hyperthread/SMT
+	CpuStatsPhysicalCoreEnable GlobalSettingKey = "cpu.stats.physicalcore.enable"
 
 	// TriState Items
 	// NetworkFallbackAnyEth global setting key
@@ -947,6 +949,7 @@ func NewConfigItemSpecMap() ConfigItemSpecMap {
 	configItemSpecMap.AddBoolItem(EnableARPSnoop, true)
 	configItemSpecMap.AddBoolItem(WwanQueryVisibleProviders, false)
 	configItemSpecMap.AddBoolItem(NetworkLocalLegacyMACAddress, false)
+	configItemSpecMap.AddBoolItem(CpuStatsPhysicalCoreEnable, false)
 
 	// Add TriState Items
 	configItemSpecMap.AddTriStateItem(NetworkFallbackAnyEth, TS_DISABLED)