[fix]: add test case with darwin OS (sustainable-computing-io#1856)

* [fix]: add test case with darwin OS

Signed-off-by: Sam Yuan <[email protected]>

* [fix]: use same function naming conventions and behavior

Signed-off-by: Sam Yuan <[email protected]>

* [fix]: update with review comments

Signed-off-by: Sam Yuan <[email protected]>

* [fix]: update contributing.md as PR review comments

Signed-off-by: Sam Yuan <[email protected]>

* [fix]: remove unused package

Signed-off-by: Sam Yuan <[email protected]>

---------

Signed-off-by: Sam Yuan <[email protected]>

CONTRIBUTING.md

@@ -135,3 +135,12 @@ Doc: update Developer.md
 
 We enabled [stale bot](https://github.com/probot/stale) for house keeping. An
 Issue or Pull Request becomes stale if no any inactivity for 60 days.
+
+## For Mac and Windows user
+
+kepler currently focus on linux platform.
+for other platforms, to make kepler is easy for anyone contributes from
+any platform, we are welcome any benefits(PRs) for kepler including
+parts as compilable on other platform.
+before the specific platform is supported, we just running CI on linux
+as PR merge standard and official support.

cmd/exporter/exporter.go

@@ -152,7 +152,7 @@ func main() {
 	platform.InitPowerImpl()
 	defer platform.StopPower()
 
-	if config.EnabledGPU() {
+	if config.IsGPUEnabled() {
 		r := accelerator.GetRegistry()
 		if a, err := accelerator.New(config.GPU, true); err == nil {
 			r.MustRegister(a) // Register the accelerator with the registry

pkg/bpf/exporter.go

@@ -1,3 +1,6 @@
+//go:build !darwin
+// +build !darwin
+
 /*
 Copyright 2021.
 

pkg/bpf/fake_mac.go

@@ -0,0 +1,293 @@
+//go:build darwin
+// +build darwin
+
+/*
+Copyright 2021.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+	http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package bpf
+
+import (
+	"errors"
+	"fmt"
+	"os"
+	"runtime"
+	"time"
+
+	"github.com/cilium/ebpf"
+	"github.com/cilium/ebpf/link"
+	"github.com/cilium/ebpf/rlimit"
+	"github.com/jaypipes/ghw"
+	"github.com/sustainable-computing-io/kepler/pkg/config"
+	"golang.org/x/sys/unix"
+	"k8s.io/apimachinery/pkg/util/sets"
+	"k8s.io/klog/v2"
+)
+
+type exporter struct {
+	bpfObjects keplerObjects
+
+	schedSwitchLink link.Link
+	irqLink         link.Link
+	pageWriteLink   link.Link
+	pageReadLink    link.Link
+
+	perfEvents *hardwarePerfEvents
+
+	enabledHardwareCounters sets.Set[string]
+	enabledSoftwareCounters sets.Set[string]
+}
+
+func NewExporter() (Exporter, error) {
+	e := &exporter{
+		enabledHardwareCounters: sets.New[string](config.BPFHwCounters()...),
+		enabledSoftwareCounters: sets.New[string](config.BPFSwCounters()...),
+	}
+	err := e.attach()
+	if err != nil {
+		e.Detach()
+	}
+	return e, err
+}
+
+func (e *exporter) SupportedMetrics() SupportedMetrics {
+	return SupportedMetrics{
+		HardwareCounters: e.enabledHardwareCounters.Clone(),
+		SoftwareCounters: e.enabledSoftwareCounters.Clone(),
+	}
+}
+
+func (e *exporter) attach() error {
+	// Remove resource limits for kernels <5.11.
+	if err := rlimit.RemoveMemlock(); err != nil {
+		return fmt.Errorf("error removing memlock: %v", err)
+	}
+
+	// Load eBPF Specs
+	specs, err := loadKepler()
+	if err != nil {
+		return fmt.Errorf("error loading eBPF specs: %v", err)
+	}
+
+	// Adjust map sizes to the number of available CPUs
+	numCPU := getCPUCores()
+	klog.Infof("Number of CPUs: %d", numCPU)
+	for _, m := range specs.Maps {
+		// Only resize maps that have a MaxEntries of NUM_CPUS constant
+		if m.MaxEntries == 128 {
+			m.MaxEntries = uint32(numCPU)
+		}
+	}
+
+	// Set program global variables
+	err = specs.RewriteConstants(map[string]interface{}{
+		"SAMPLE_RATE": int32(config.GetBPFSampleRate()),
+	})
+	if err != nil {
+		return fmt.Errorf("error rewriting program constants: %v", err)
+	}
+
+	// Load the eBPF program(s)
+	if err := specs.LoadAndAssign(&e.bpfObjects, nil); err != nil {
+		return fmt.Errorf("error loading eBPF objects: %v", err)
+	}
+
+	// Attach the eBPF program(s)
+	e.schedSwitchLink, err = link.AttachTracing(link.TracingOptions{
+		Program:    e.bpfObjects.KeplerSchedSwitchTrace,
+		AttachType: ebpf.AttachTraceRawTp,
+	})
+	if err != nil {
+		return fmt.Errorf("error attaching sched_switch tracepoint: %v", err)
+	}
+
+	if config.ExposeIRQCounterMetrics() {
+		e.irqLink, err = link.AttachTracing(link.TracingOptions{
+			Program:    e.bpfObjects.KeplerIrqTrace,
+			AttachType: ebpf.AttachTraceRawTp,
+		})
+		if err != nil {
+			return fmt.Errorf("could not attach irq/softirq_entry: %w", err)
+		}
+	}
+
+	group := "writeback"
+	name := "writeback_dirty_page"
+	if _, err := os.Stat("/sys/kernel/debug/tracing/events/writeback/writeback_dirty_folio"); err == nil {
+		name = "writeback_dirty_folio"
+	}
+	e.pageWriteLink, err = link.Tracepoint(group, name, e.bpfObjects.KeplerWritePageTrace, nil)
+	if err != nil {
+		klog.Warningf("failed to attach tp/%s/%s: %v. Kepler will not collect page cache write events. This will affect the DRAM power model estimation on VMs.", group, name, err)
+	} else {
+		e.enabledSoftwareCounters[config.PageCacheHit] = struct{}{}
+	}
+
+	e.pageReadLink, err = link.AttachTracing(link.TracingOptions{
+		Program:    e.bpfObjects.KeplerReadPageTrace,
+		AttachType: ebpf.AttachTraceFEntry,
+	})
+	if err != nil {
+		klog.Warningf("failed to attach fentry/mark_page_accessed: %v. Kepler will not collect page cache read events. This will affect the DRAM power model estimation on VMs.", err)
+	}
+
+	// Return early if hardware counters are not enabled
+	if !config.ExposeHardwareCounterMetrics() {
+		klog.Infof("Hardware counter metrics are disabled")
+		return nil
+	}
+
+	e.perfEvents, err = createHardwarePerfEvents(
+		e.bpfObjects.CpuInstructionsEventReader,
+		e.bpfObjects.CpuCyclesEventReader,
+		e.bpfObjects.CacheMissEventReader,
+		numCPU,
+	)
+	if err != nil {
+		return nil
+	}
+
+	return nil
+}
+
+func (e *exporter) Detach() {
+	// Links
+	if e.schedSwitchLink != nil {
+		e.schedSwitchLink.Close()
+		e.schedSwitchLink = nil
+	}
+
+	if e.irqLink != nil {
+		e.irqLink.Close()
+		e.irqLink = nil
+	}
+
+	if e.pageWriteLink != nil {
+		e.pageWriteLink.Close()
+		e.pageWriteLink = nil
+	}
+
+	if e.pageReadLink != nil {
+		e.pageReadLink.Close()
+		e.pageReadLink = nil
+	}
+
+	// Perf events
+	e.perfEvents.close()
+	e.perfEvents = nil
+
+	// Objects
+	e.bpfObjects.Close()
+}
+
+func (e *exporter) CollectProcesses() ([]ProcessMetrics, error) {
+	start := time.Now()
+	// Get the max number of entries in the map
+	maxEntries := e.bpfObjects.Processes.MaxEntries()
+	total := 0
+	deleteKeys := make([]uint32, maxEntries)
+	deleteValues := make([]ProcessMetrics, maxEntries)
+	var cursor ebpf.MapBatchCursor
+	for {
+		count, err := e.bpfObjects.Processes.BatchLookupAndDelete(
+			&cursor,
+			deleteKeys,
+			deleteValues,
+			&ebpf.BatchOptions{},
+		)
+		total += count
+		if errors.Is(err, ebpf.ErrKeyNotExist) {
+			break
+		}
+		if err != nil {
+			return nil, fmt.Errorf("failed to batch lookup and delete: %v", err)
+		}
+	}
+	klog.V(5).Infof("collected %d process samples in %v", total, time.Since(start))
+	return deleteValues[:total], nil
+}
+
+///////////////////////////////////////////////////////////////////////////
+// utility functions
+
+func unixOpenPerfEvent(typ, conf, cpuCores int) ([]int, error) {
+	return []int{}, nil
+}
+
+func unixClosePerfEvents(fds []int) {
+	for _, fd := range fds {
+		_ = unix.SetNonblock(fd, true)
+		unix.Close(fd)
+	}
+}
+
+func getCPUCores() int {
+	cores := runtime.NumCPU()
+	if cpu, err := ghw.CPU(); err == nil {
+		// we need to get the number of all CPUs,
+		// so if /proc/cpuinfo is available, we can get the number of all CPUs
+		cores = int(cpu.TotalThreads)
+	}
+	return cores
+}
+
+type hardwarePerfEvents struct {
+	cpuCyclesPerfEvents       []int
+	cpuInstructionsPerfEvents []int
+	cacheMissPerfEvents       []int
+}
+
+func (h *hardwarePerfEvents) close() {
+	unixClosePerfEvents(h.cpuCyclesPerfEvents)
+	unixClosePerfEvents(h.cpuInstructionsPerfEvents)
+	unixClosePerfEvents(h.cacheMissPerfEvents)
+}
+
+// CreateHardwarePerfEvents creates perf events for CPU cycles, CPU instructions, and cache misses
+// and updates the corresponding eBPF maps.
+func createHardwarePerfEvents(cpuInstructionsMap, cpuCyclesMap, cacheMissMap *ebpf.Map, numCPU int) (*hardwarePerfEvents, error) {
+	var err error
+	events := &hardwarePerfEvents{
+		cpuCyclesPerfEvents:       []int{},
+		cpuInstructionsPerfEvents: []int{},
+		cacheMissPerfEvents:       []int{},
+	}
+	defer func() {
+		if err != nil {
+			unixClosePerfEvents(events.cpuCyclesPerfEvents)
+			unixClosePerfEvents(events.cpuInstructionsPerfEvents)
+			unixClosePerfEvents(events.cacheMissPerfEvents)
+		}
+	}()
+	for i, fd := range events.cpuCyclesPerfEvents {
+		if err = cpuCyclesMap.Update(uint32(i), uint32(fd), ebpf.UpdateAny); err != nil {
+			klog.Warningf("Failed to update cpu_cycles_event_reader map: %v", err)
+			return nil, err
+		}
+	}
+	for i, fd := range events.cpuInstructionsPerfEvents {
+		if err = cpuInstructionsMap.Update(uint32(i), uint32(fd), ebpf.UpdateAny); err != nil {
+			klog.Warningf("Failed to update cpu_instructions_event_reader map: %v", err)
+			return nil, err
+		}
+	}
+	for i, fd := range events.cacheMissPerfEvents {
+		if err = cacheMissMap.Update(uint32(i), uint32(fd), ebpf.UpdateAny); err != nil {
+			klog.Warningf("Failed to update cache_miss_event_reader map: %v", err)
+			return nil, err
+		}
+	}
+	return events, nil
+}

pkg/bpftest/bpf_suite_test.go

@@ -1,3 +1,6 @@
+//go:build !darwin
+// +build !darwin
+
 package bpftest
 
 import (

pkg/collector/energy/node_energy_collector.go

@@ -66,7 +66,7 @@ func UpdateNodeComponentsEnergy(nodeStats *stats.NodeStats, wg *sync.WaitGroup)
 // UpdateNodeGPUEnergy updates each GPU power consumption. Right now we don't support other types of accelerators
 func UpdateNodeGPUEnergy(nodeStats *stats.NodeStats, wg *sync.WaitGroup) {
 	defer wg.Done()
-	if config.EnabledGPU() {
+	if config.IsGPUEnabled() {
 		if gpu := acc.GetActiveAcceleratorByType(config.GPU); gpu != nil {
 			gpuEnergy := gpu.Device().AbsEnergyFromDevice()
 			for gpu, energy := range gpuEnergy {

pkg/collector/metric_collector.go

@@ -158,7 +158,7 @@ func (c *Collector) updateProcessResourceUtilizationMetrics(wg *sync.WaitGroup)
 	// update process metrics regarding the resource utilization to be used to calculate the energy consumption
 	// we first updates the bpf which is responsible to include new processes in the ProcessStats collection
 	resourceBpf.UpdateProcessBPFMetrics(c.bpfExporter, c.ProcessStats)
-	if config.EnabledGPU() {
+	if config.IsGPUEnabled() {
 		if acc.GetActiveAcceleratorByType(config.GPU) != nil {
 			accelerator.UpdateProcessGPUUtilizationMetrics(c.ProcessStats)
 		}

pkg/collector/resourceutilization/bpf/fake_mac.go

@@ -0,0 +1,13 @@
+//go:build darwin
+// +build darwin
+
+package bpf
+
+import (
+	"github.com/sustainable-computing-io/kepler/pkg/bpf"
+	"github.com/sustainable-computing-io/kepler/pkg/collector/stats"
+)
+
+func UpdateProcessBPFMetrics(bpfExporter bpf.Exporter, processStats map[uint64]*stats.ProcessStats) {
+
+}

pkg/collector/resourceutilization/bpf/process_bpf_collector.go

@@ -1,3 +1,6 @@
+//go:build !darwin
+// +build !darwin
+
 /*
 Copyright 2021.
 

pkg/collector/resourceutilization/bpf/process_bpf_collector_test.go

@@ -1,3 +1,6 @@
+//go:build !darwin
+// +build !darwin
+
 package bpf
 
 import (

pkg/collector/stats/node_stats.go

@@ -50,7 +50,7 @@ func (ne *NodeStats) ResetDeltaValues() {
 
 func (ne *NodeStats) UpdateIdleEnergyWithMinValue(isComponentsSystemCollectionSupported bool) {
 	// gpu metric
-	if config.EnabledGPU() {
+	if config.IsGPUEnabled() {
 		if acc.GetActiveAcceleratorByType(config.GPU) != nil {
 			ne.CalcIdleEnergy(config.AbsEnergyInGPU, config.IdleEnergyInGPU, config.GPUComputeUtilization)
 		}