Merge branch 'master' of github.com:Project-HAMi/HAMi

cmd/device-plugin/nvidia/vgpucfg.go

@@ -3,7 +3,6 @@ package main
 import (
 	"encoding/json"
 	"fmt"
-	"io/ioutil"
 	"os"
 	"strings"
 
@@ -80,7 +79,7 @@ func updateFromCLIFlag[T any](pflag **T, c *cli.Context, flagName string) {
 }
 
 func readFromConfigFile() error {
-	jsonbyte, err := ioutil.ReadFile("/config/config.json")
+	jsonbyte, err := os.ReadFile("/config/config.json")
 	if err != nil {
 		return err
 	}

cmd/vGPUmonitor/feedback.go

@@ -3,7 +3,6 @@ package main
 import (
 	"errors"
 	"fmt"
-	"io/ioutil"
 	"os"
 	"sort"
 	"strconv"
@@ -34,7 +33,7 @@ func init() {
 
 func setcGgroupDriver() int {
 	// 1 for cgroupfs 2 for systemd
-	kubeletconfig, err := ioutil.ReadFile("/hostvar/lib/kubelet/config.yaml")
+	kubeletconfig, err := os.ReadFile("/hostvar/lib/kubelet/config.yaml")
 	if err != nil {
 		return 0
 	}

cmd/vGPUmonitor/pathmonitor.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"io/ioutil"
 	"net"
 	"os"
 	"path/filepath"
@@ -38,7 +37,7 @@ func init() {
 
 func checkfiles(fpath string) (*sharedRegionT, error) {
 	klog.Infof("Checking path %s", fpath)
-	files, err := ioutil.ReadDir(fpath)
+	files, err := os.ReadDir(fpath)
 	if err != nil {
 		return nil, err
 	}
@@ -83,7 +82,7 @@ func isVaildPod(name string, pods *v1.PodList) bool {
 func monitorpath(podmap map[string]podusage) error {
 	lock.Lock()
 	defer lock.Unlock()
-	files, err := ioutil.ReadDir(containerPath)
+	files, err := os.ReadDir(containerPath)
 	if err != nil {
 		return err
 	}

pkg/device-plugin/hygon/dcu/amdgpu/amdgpu.go

@@ -13,7 +13,6 @@ import (
 	"bufio"
 	"errors"
 	"fmt"
-	"io/ioutil"
 	"os"
 	"path/filepath"
 	"regexp"
@@ -100,7 +99,7 @@ func GetAMDGPUs() map[string]map[string]int {
 // AMDGPU check if a particular card is an AMD GPU by checking the device's vendor ID
 func AMDGPU(cardName string) bool {
 	sysfsVendorPath := "/sys/class/drm/" + cardName + "/device/vendor"
-	b, err := ioutil.ReadFile(sysfsVendorPath)
+	b, err := os.ReadFile(sysfsVendorPath)
 	if err == nil {
 		vid := strings.TrimSpace(string(b))
 

pkg/device-plugin/hygon/dcu/server.go

@@ -452,8 +452,14 @@ func (p *Plugin) createvdevFile(current *v1.Pod, ctr *v1.Container, req util.Con
 		s = s + fmt.Sprintln("enable: 1")
 	}
 	cacheFileHostDirectory := "/usr/local/vgpu/dcu/" + string(current.UID) + "_" + ctr.Name + "_" + fmt.Sprint(devidx) + "_" + fmt.Sprint(pipeid) + "_" + fmt.Sprint(vdevidx) + "_" + coremsk
-	os.MkdirAll(cacheFileHostDirectory, 0777)
-	os.Chmod(cacheFileHostDirectory, 0777)
+	err := os.MkdirAll(cacheFileHostDirectory, 0777)
+	if err != nil {
+		return "", err
+	}
+	err = os.Chmod(cacheFileHostDirectory, 0777)
+	if err != nil {
+		return "", err
+	}
 	os.WriteFile(cacheFileHostDirectory+"/vdev0.conf", []byte(s), os.ModePerm)
 	return cacheFileHostDirectory, nil
 }

pkg/device-plugin/mlu/cndev/cndev.go

@@ -18,8 +18,8 @@ import (
 	"bytes"
 	"errors"
 	"fmt"
-	"io/ioutil"
 	"log"
+	"os"
 	"os/exec"
 	"strconv"
 	"strings"
@@ -132,7 +132,7 @@ func (d *Device) ValidateSriovNum(num int) error {
 }
 
 func getNumFromFile(path string) (int, error) {
-	output, err := ioutil.ReadFile(path)
+	output, err := os.ReadFile(path)
 	if err != nil {
 		return 0, err
 	}

pkg/device-plugin/mlu/cntopo/cntopo.go

@@ -16,7 +16,7 @@ package cntopo
 
 import (
 	"encoding/json"
-	"io/ioutil"
+	"os"
 	"os/exec"
 	"sync"
 )
@@ -70,15 +70,15 @@ func (c *cntopo) GetRings(available []uint, size int) ([]Ring, error) {
 	}
 	c.Lock()
 	defer c.Unlock()
-	err = ioutil.WriteFile("/tmp/cntopo_input.json", b, 0666)
+	err = os.WriteFile("/tmp/cntopo_input.json", b, 0666)
 	if err != nil {
 		return nil, err
 	}
 	err = exec.Command("sh", "-c", "cntopo find -I /tmp/cntopo_input.json -O /tmp/cntopo_output.json -R 1000000 -C").Run()
 	if err != nil {
 		return nil, err
 	}
-	j, err := ioutil.ReadFile("/tmp/cntopo_output.json")
+	j, err := os.ReadFile("/tmp/cntopo_output.json")
 	if err != nil {
 		return nil, err
 	}

pkg/scheduler/nodes.go

@@ -49,7 +49,7 @@ type NodeUsage struct {
 
 type nodeManager struct {
 	nodes map[string]*NodeInfo
-	mutex sync.Mutex
+	mutex sync.RWMutex
 }
 
 func (m *nodeManager) init() {
@@ -101,14 +101,16 @@ func (m *nodeManager) rmNodeDevice(nodeID string, nodeInfo *NodeInfo) {
 }
 
 func (m *nodeManager) GetNode(nodeID string) (*NodeInfo, error) {
-	m.mutex.Lock()
-	defer m.mutex.Unlock()
+	m.mutex.RLock()
+	defer m.mutex.RUnlock()
 	if n, ok := m.nodes[nodeID]; ok {
 		return n, nil
 	}
 	return &NodeInfo{}, fmt.Errorf("node %v not found", nodeID)
 }
 
 func (m *nodeManager) ListNodes() (map[string]*NodeInfo, error) {
+	m.mutex.RLock()
+	defer m.mutex.RUnlock()
 	return m.nodes, nil
 }

pkg/scheduler/pods.go

@@ -36,7 +36,7 @@ type podInfo struct {
 
 type podManager struct {
 	pods  map[k8stypes.UID]*podInfo
-	mutex sync.Mutex
+	mutex sync.RWMutex
 }
 
 func (m *podManager) init() {
@@ -65,6 +65,8 @@ func (m *podManager) delPod(pod *corev1.Pod) {
 }
 
 func (m *podManager) GetScheduledPods() (map[k8stypes.UID]*podInfo, error) {
+	m.mutex.RLock()
+	defer m.mutex.RUnlock()
 	klog.Infof("Getting all scheduled pods with %d nums", len(m.pods))
 	return m.pods, nil
 }

pkg/scheduler/score.go

@@ -88,7 +88,7 @@ func fitInCertainDevice(node *NodeUsage, request util.ContainerDeviceRequest, an
 	originReq := k.Nums
 	prevnuma := -1
 	klog.InfoS("Allocating device for container request", "pod", klog.KObj(pod), "card request", k)
-	tmpDevs := []util.ContainerDevice{}
+	var tmpDevs []util.ContainerDevice
 	for i := len(node.Devices) - 1; i >= 0; i-- {
 		klog.InfoS("scoring pod", "pod", klog.KObj(pod), "Memreq", k.Memreq, "MemPercentagereq", k.MemPercentagereq, "Coresreq", k.Coresreq, "Nums", k.Nums, "device index", i, "device", node.Devices[i].Id)
 		found, numa := checkType(annos, *node.Devices[i], k)