Azure disk: dealing with missing disk probe

Signed-off-by: Huamin Chen <hchen@redhat.com>

pkg/volume/azure_dd/attacher.go

@@ -171,6 +171,10 @@ func (attacher *azureDiskAttacher) WaitForAttach(spec *volume.Spec, lunStr strin
 	err = wait.Poll(checkSleepDuration, timeout, func() (bool, error) {
 		glog.V(4).Infof("Checking Azure disk %q(lun %s) is attached.", volumeSource.DiskName, lunStr)
 		if devicePath, err = findDiskByLun(lun, &osIOHandler{}, exe); err == nil {
+			if len(devicePath) == 0 {
+				glog.Warningf("cannot find attached Azure disk %q(lun %s) locally.", volumeSource.DiskName, lunStr)
+				return false, fmt.Errorf("cannot find attached Azure disk %q(lun %s) locally.", volumeSource.DiskName, lunStr)
+			}
 			glog.V(4).Infof("Successfully found attached Azure disk %q(lun %s, device path %s).", volumeSource.DiskName, lunStr, devicePath)
 			return true, nil
 		} else {

pkg/volume/azure_dd/vhd_util.go

@@ -31,6 +31,7 @@ import (
 type ioHandler interface {
 	ReadDir(dirname string) ([]os.FileInfo, error)
 	WriteFile(filename string, data []byte, perm os.FileMode) error
+	Readlink(name string) (string, error)
 }
 
 type osIOHandler struct{}
@@ -41,29 +42,50 @@ func (handler *osIOHandler) ReadDir(dirname string) ([]os.FileInfo, error) {
 func (handler *osIOHandler) WriteFile(filename string, data []byte, perm os.FileMode) error {
 	return ioutil.WriteFile(filename, data, perm)
 }
+func (handler *osIOHandler) Readlink(name string) (string, error) {
+	return os.Readlink(name)
+}
 
-// given a LUN find the VHD device path like /dev/sdb
-// VHD disks under sysfs are like /sys/bus/scsi/devices/3:0:1:0
-// return empty string if no disk is found
+// exclude those used by azure as resource and OS root in /dev/disk/azure
+func listAzureDiskPath(io ioHandler) []string {
+	azureDiskPath := "/dev/disk/azure/"
+	var azureDiskList []string
+	if dirs, err := io.ReadDir(azureDiskPath); err == nil {
+		for _, f := range dirs {
+			name := f.Name()
+			diskPath := azureDiskPath + name
+			if link, linkErr := io.Readlink(diskPath); linkErr == nil {
+				sd := link[(strings.LastIndex(link, "/") + 1):]
+				azureDiskList = append(azureDiskList, sd)
+			}
+		}
+	}
+	glog.V(12).Infof("Azure sys disks paths: %v", azureDiskList)
+	return azureDiskList
+}
+
+// given a LUN find the VHD device path like /dev/sdd
+// exclude those disks used by Azure resources and OS root
 func findDiskByLun(lun int, io ioHandler, exe exec.Interface) (string, error) {
+	azureDisks := listAzureDiskPath(io)
+	return findDiskByLunWithConstraint(lun, io, exe, azureDisks)
+}
+
+// look for device /dev/sdX and validate it is a VHD
+// return empty string if no disk is found
+func findDiskByLunWithConstraint(lun int, io ioHandler, exe exec.Interface, azureDisks []string) (string, error) {
 	var err error
 	sys_path := "/sys/bus/scsi/devices"
 	if dirs, err := io.ReadDir(sys_path); err == nil {
 		for _, f := range dirs {
 			name := f.Name()
-			// look for path like /sys/bus/scsi/devices/3:0:1:0
+			// look for path like /sys/bus/scsi/devices/3:0:0:1
 			arr := strings.Split(name, ":")
 			if len(arr) < 4 {
 				continue
 			}
-			target, err := strconv.Atoi(arr[0])
-			if err != nil {
-				glog.Errorf("failed to parse target from %v (%v), err %v", arr[0], name, err)
-				continue
-			}
-
-			// as observed, targets 0-3 are used by OS disks. Skip them
-			if target > 3 {
+			// extract LUN from the path.
+			// LUN is the last index of the array, i.e. 1 in /sys/bus/scsi/devices/3:0:0:1
 			l, err := strconv.Atoi(arr[3])
 			if err != nil {
 				// unknown path format, continue to read the next one
@@ -85,12 +107,18 @@ func findDiskByLun(lun int, io ioHandler, exe exec.Interface) (string, error) {
 					glog.V(4).Infof("doesn't match VHD, output %v, error %v", string(out), err)
 					continue
 				}
-					// find it!
+				// find a disk, validate name
 				dir := path.Join(sys_path, name, "block")
-					dev, err := io.ReadDir(dir)
-					if err != nil {
-						glog.Errorf("failed to read %s", dir)
-					} else {
+				if dev, err := io.ReadDir(dir); err == nil {
+					found := false
+					for _, diskName := range azureDisks {
+						glog.V(12).Infof("validating disk %q with sys disk %q", dev[0].Name(), diskName)
+						if string(dev[0].Name()) == diskName {
+							found = true
+							break
+						}
+					}
+					if !found {
 						return "/dev/" + dev[0].Name(), nil
 					}
 				}

pkg/volume/azure_dd/vhd_util_test.go

@@ -56,7 +56,11 @@ var (
 	lun       = 1
 	lunStr    = "1"
 	diskPath  = "4:0:0:" + lunStr
-	devName  = "sda"
+	devName   = "sdd"
+	lun1      = 2
+	lunStr1   = "2"
+	diskPath1 = "3:0:0:" + lunStr1
+	devName1  = "sde"
 )
 
 type fakeIOHandler struct{}
@@ -85,6 +89,11 @@ func (handler *fakeIOHandler) ReadDir(dirname string) ([]os.FileInfo, error) {
 			name: devName,
 		}
 		return []os.FileInfo{n}, nil
+	case "/sys/bus/scsi/devices/" + diskPath1 + "/block":
+		n := &fakeFileInfo{
+			name: devName1,
+		}
+		return []os.FileInfo{n}, nil
 	}
 	return nil, fmt.Errorf("bad dir")
 }
@@ -93,6 +102,10 @@ func (handler *fakeIOHandler) WriteFile(filename string, data []byte, perm os.Fi
 	return nil
 }
 
+func (handler *fakeIOHandler) Readlink(name string) (string, error) {
+	return "/dev/azure/disk/sda", nil
+}
+
 func TestIoHandler(t *testing.T) {
 	var fcmd exec.FakeCmd
 	fcmd = exec.FakeCmd{
@@ -101,16 +114,21 @@ func TestIoHandler(t *testing.T) {
 			func() ([]byte, error) {
 				return []byte("Msft    \nVirtual Disk \n"), nil
 			},
+			// cat
+			func() ([]byte, error) {
+				return []byte("Msft    \nVirtual Disk \n"), nil
+			},
 		},
 	}
 	fake := exec.FakeExec{
 		CommandScript: []exec.FakeCommandAction{
 			func(cmd string, args ...string) exec.Cmd { return exec.InitFakeCmd(&fcmd, cmd, args...) },
+			func(cmd string, args ...string) exec.Cmd { return exec.InitFakeCmd(&fcmd, cmd, args...) },
 		},
 	}
 	disk, err := findDiskByLun(lun, &fakeIOHandler{}, &fake)
 	// if no disk matches lun, exit
 	if disk != "/dev/"+devName || err != nil {
-		t.Errorf("no data disk disk found: disk %v err %v", disk, err)
+		t.Errorf("no data disk found: disk %v err %v", disk, err)
 	}
 }