feat: wipe devices before using if enabled

Signed-off-by: Suleyman Akbas <[email protected]>

.mockery.yaml

@@ -10,3 +10,9 @@ packages:
   github.com/openshift/lvm-operator/pkg/lvm:
       interfaces:
         LVM:
+  github.com/openshift/lvm-operator/pkg/wipefs:
+      interfaces:
+        Wipefs:
+  github.com/openshift/lvm-operator/pkg/dmsetup:
+    interfaces:
+      Dmsetup:

api/v1alpha1/lvmcluster_types.go

@@ -21,9 +21,6 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 )
 
-// EDIT THIS FILE!  THIS IS SCAFFOLDING FOR YOU TO OWN!
-// NOTE: json tags are required.  Any new fields you add must have json tags for the fields to be serialized.
-
 // LVMClusterSpec defines the desired state of LVMCluster
 type LVMClusterSpec struct {
 	// Important: Run "make" to regenerate code after modifying this file
@@ -114,15 +111,13 @@ type DeviceSelector struct {
 	// We discourage using the device names as they can change over node restarts.
 	// +optional
 	OptionalPaths []string `json:"optionalPaths,omitempty"`
-}
-
-// type DeviceClassConfig struct {
-// 	LVMConfig *LVMConfig `json:"lvmConfig,omitempty"`
-// }
 
-// type LVMConfig struct {
-// 	thinProvision bool `json:"thinProvision,omitempty"`
-// }
+	// ForceWipeDevicesAndDestroyAllData runs wipefs to wipe the devices.
+	// This can lead to data lose. Enable this only when you know that the disk
+	// does not contain any important data.
+	// +optional
+	ForceWipeDevicesAndDestroyAllData bool `json:"forceWipeDevicesAndDestroyAllData,omitempty"`
+}
 
 type LVMStateType string
 

bundle/manifests/lvm.topolvm.io_lvmclusters.yaml

@@ -56,6 +56,12 @@ spec:
                           description: DeviceSelector is a set of rules that should
                             match for a device to be included in the LVMCluster
                           properties:
+                            forceWipeDevicesAndDestroyAllData:
+                              description: ForceWipeDevicesAndDestroyAllData runs
+                                wipefs to wipe the devices. This can lead to data
+                                lose. Enable this only when you know that the disk
+                                does not contain any important data.
+                              type: boolean
                             optionalPaths:
                               description: A list of device paths which could be chosen
                                 for creating Volume Group. For example "/dev/disk/by-path/pci-0000:04:00.0-nvme-1"

bundle/manifests/lvm.topolvm.io_lvmvolumegroups.yaml

@@ -42,6 +42,11 @@ spec:
                 description: DeviceSelector is a set of rules that should match for
                   a device to be included in this TopoLVMCluster
                 properties:
+                  forceWipeDevicesAndDestroyAllData:
+                    description: ForceWipeDevicesAndDestroyAllData runs wipefs to
+                      wipe the devices. This can lead to data lose. Enable this only
+                      when you know that the disk does not contain any important data.
+                    type: boolean
                   optionalPaths:
                     description: A list of device paths which could be chosen for
                       creating Volume Group. For example "/dev/disk/by-path/pci-0000:04:00.0-nvme-1"

cmd/vgmanager/vgmanager.go

@@ -21,11 +21,13 @@ import (
 	"os"
 
 	"github.com/go-logr/logr"
+	"github.com/openshift/lvm-operator/pkg/dmsetup"
 	"github.com/openshift/lvm-operator/pkg/filter"
 	"github.com/openshift/lvm-operator/pkg/lsblk"
 	"github.com/openshift/lvm-operator/pkg/lvm"
 	"github.com/openshift/lvm-operator/pkg/lvmd"
 	"github.com/openshift/lvm-operator/pkg/vgmanager"
+	"github.com/openshift/lvm-operator/pkg/wipefs"
 	"github.com/spf13/cobra"
 	"sigs.k8s.io/controller-runtime/pkg/cache"
 	metricsserver "sigs.k8s.io/controller-runtime/pkg/metrics/server"
@@ -106,6 +108,8 @@ func run(_ *cobra.Command, _ []string, opts *Options) error {
 		LVMD:          lvmd.DefaultConfigurator(),
 		Scheme:        mgr.GetScheme(),
 		LSBLK:         lsblk.NewDefaultHostLSBLK(),
+		Wipefs:        wipefs.NewDefaultHostWipefs(),
+		Dmsetup:       dmsetup.NewDefaultHostDmsetup(),
 		LVM:           lvm.NewDefaultHostLVM(),
 		NodeName:      os.Getenv("NODE_NAME"),
 		Namespace:     os.Getenv("POD_NAMESPACE"),

config/crd/bases/lvm.topolvm.io_lvmclusters.yaml

@@ -52,6 +52,12 @@ spec:
                           description: DeviceSelector is a set of rules that should
                             match for a device to be included in the LVMCluster
                           properties:
+                            forceWipeDevicesAndDestroyAllData:
+                              description: ForceWipeDevicesAndDestroyAllData runs
+                                wipefs to wipe the devices. This can lead to data
+                                lose. Enable this only when you know that the disk
+                                does not contain any important data.
+                              type: boolean
                             optionalPaths:
                               description: A list of device paths which could be chosen
                                 for creating Volume Group. For example "/dev/disk/by-path/pci-0000:04:00.0-nvme-1"

config/crd/bases/lvm.topolvm.io_lvmvolumegroups.yaml

@@ -42,6 +42,11 @@ spec:
                 description: DeviceSelector is a set of rules that should match for
                   a device to be included in this TopoLVMCluster
                 properties:
+                  forceWipeDevicesAndDestroyAllData:
+                    description: ForceWipeDevicesAndDestroyAllData runs wipefs to
+                      wipe the devices. This can lead to data lose. Enable this only
+                      when you know that the disk does not contain any important data.
+                    type: boolean
                   optionalPaths:
                     description: A list of device paths which could be chosen for
                       creating Volume Group. For example "/dev/disk/by-path/pci-0000:04:00.0-nvme-1"

pkg/dmsetup/dmsetup.go

@@ -0,0 +1,52 @@
+package dmsetup
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/openshift/lvm-operator/pkg/internal/exec"
+)
+
+var (
+	DefaultDMSetup       = "/usr/sbin/dmsetup"
+	ErrReferenceNotFound = fmt.Errorf("device-mapper reference not found")
+)
+
+type Dmsetup interface {
+	Remove(deviceName string) error
+}
+
+type HostDmsetup struct {
+	exec.Executor
+	dmsetup string
+}
+
+func NewDefaultHostDmsetup() *HostDmsetup {
+	return NewHostDmsetup(&exec.CommandExecutor{}, DefaultDMSetup)
+}
+
+func NewHostDmsetup(executor exec.Executor, dmsetup string) *HostDmsetup {
+	return &HostDmsetup{
+		Executor: executor,
+		dmsetup:  dmsetup,
+	}
+}
+
+// Remove removes the device's reference from the device-mapper
+func (dmsetup *HostDmsetup) Remove(deviceName string) error {
+	if len(deviceName) == 0 {
+		return fmt.Errorf("failed to remove device-mapper reference. Device name is empty")
+	}
+
+	args := []string{"remove"}
+	args = append(args, deviceName)
+	output, err := dmsetup.ExecuteCommandWithOutputAsHost(dmsetup.dmsetup, args...)
+	if err != nil {
+		if strings.Contains(output, "not found") {
+			return ErrReferenceNotFound
+		}
+		return fmt.Errorf("failed to remove the reference from device-mapper %q: %v", deviceName, err)
+	}
+
+	return nil
+}

pkg/internal/exec/exec.go

@@ -26,7 +26,7 @@ var (
 	nsenterPath = "/usr/bin/nsenter"
 )
 
-// Executor is the  interface for running exec commands
+// Executor is the interface for running exec commands
 type Executor interface {
 	ExecuteCommandWithOutput(command string, arg ...string) (string, error)
 	ExecuteCommandWithOutputAsHost(command string, arg ...string) (string, error)

pkg/vgmanager/devices.go

@@ -138,7 +138,7 @@ func (r *VGReconciler) filterDevices(ctx context.Context, devices []lsblk.BlockD
 }
 
 // getNewDevicesToBeAdded gets all devices that should be added to the volume group
-func (r *VGReconciler) getNewDevicesToBeAdded(ctx context.Context, blockDevices []lsblk.BlockDevice, vgs []lvm.VolumeGroup, volumeGroup *lvmv1alpha1.LVMVolumeGroup) ([]lsblk.BlockDevice, error) {
+func (r *VGReconciler) getNewDevicesToBeAdded(ctx context.Context, blockDevices []lsblk.BlockDevice, nodeStatus *lvmv1alpha1.LVMVolumeGroupNodeStatus, volumeGroup *lvmv1alpha1.LVMVolumeGroup) ([]lsblk.BlockDevice, error) {
 	logger := log.FromContext(ctx)
 
 	var validBlockDevices []lsblk.BlockDevice
@@ -151,7 +151,7 @@ func (r *VGReconciler) getNewDevicesToBeAdded(ctx context.Context, blockDevices
 
 	// If Paths is specified, treat it as required paths
 	for _, path := range volumeGroup.Spec.DeviceSelector.Paths {
-		blockDevice, err := getValidDevice(path, blockDevices, vgs, volumeGroup)
+		blockDevice, err := getValidDevice(path, blockDevices, nodeStatus, volumeGroup)
 		if err != nil {
 			// An error for required devices is critical
 			return nil, fmt.Errorf("unable to validate device %s: %v", path, err)
@@ -168,7 +168,7 @@ func (r *VGReconciler) getNewDevicesToBeAdded(ctx context.Context, blockDevices
 	}
 
 	for _, path := range volumeGroup.Spec.DeviceSelector.OptionalPaths {
-		blockDevice, err := getValidDevice(path, blockDevices, vgs, volumeGroup)
+		blockDevice, err := getValidDevice(path, blockDevices, nodeStatus, volumeGroup)
 
 		// Check if we should skip this device
 		if err != nil {
@@ -200,11 +200,11 @@ func (r *VGReconciler) getNewDevicesToBeAdded(ctx context.Context, blockDevices
 	return validBlockDevices, nil
 }
 
-func isDeviceAlreadyPartOfVG(vgs []lvm.VolumeGroup, diskName string, volumeGroup *lvmv1alpha1.LVMVolumeGroup) bool {
-	for _, vg := range vgs {
-		if vg.Name == volumeGroup.Name {
-			for _, pv := range vg.PVs {
-				if pv.PvName == diskName {
+func isDeviceAlreadyPartOfVG(nodeStatus *lvmv1alpha1.LVMVolumeGroupNodeStatus, diskName string, volumeGroup *lvmv1alpha1.LVMVolumeGroup) bool {
+	for _, vgStatus := range nodeStatus.Spec.LVMVGStatus {
+		if vgStatus.Name == volumeGroup.Name {
+			for _, pv := range vgStatus.Devices {
+				if pv == diskName {
 					return true
 				}
 			}
@@ -232,15 +232,15 @@ func hasExactDisk(blockDevices []lsblk.BlockDevice, deviceName string) (lsblk.Bl
 //
 //	An error will be returned if the device is invalid
 //	No error and an empty BlockDevice object will be returned if this device should be skipped (ex: duplicate device)
-func getValidDevice(devicePath string, blockDevices []lsblk.BlockDevice, vgs []lvm.VolumeGroup, volumeGroup *lvmv1alpha1.LVMVolumeGroup) (lsblk.BlockDevice, error) {
+func getValidDevice(devicePath string, blockDevices []lsblk.BlockDevice, nodeStatus *lvmv1alpha1.LVMVolumeGroupNodeStatus, volumeGroup *lvmv1alpha1.LVMVolumeGroup) (lsblk.BlockDevice, error) {
 	// Make sure the symlink exists
 	diskName, err := filepath.EvalSymlinks(devicePath)
 	if err != nil {
 		return lsblk.BlockDevice{}, fmt.Errorf("unable to find symlink for disk path %s: %v", devicePath, err)
 	}
 
 	// Make sure this isn't a duplicate in the VG
-	if isDeviceAlreadyPartOfVG(vgs, diskName, volumeGroup) {
+	if isDeviceAlreadyPartOfVG(nodeStatus, diskName, volumeGroup) {
 		return lsblk.BlockDevice{}, nil // No error, we just don't want a duplicate
 	}
 

pkg/vgmanager/devices_test.go

@@ -570,7 +570,7 @@ func Test_getNewDevicesToBeAdded(t *testing.T) {
 	for _, tc := range testCases {
 		t.Run(tc.description, func(t *testing.T) {
 			ctx := log.IntoContext(context.Background(), testr.New(t))
-			availableDevices, err := r.getNewDevicesToBeAdded(ctx, tc.existingBlockDevices, tc.existingVGs, &tc.volumeGroup)
+			availableDevices, err := r.getNewDevicesToBeAdded(ctx, tc.existingBlockDevices, &v1alpha1.LVMVolumeGroupNodeStatus{}, &tc.volumeGroup)
 			if !tc.expectError {
 				assert.NoError(t, err)
 			} else {

pkg/vgmanager/vgmanager_controller.go

@@ -25,10 +25,13 @@ import (
 	"github.com/google/go-cmp/cmp"
 	lvmv1alpha1 "github.com/openshift/lvm-operator/api/v1alpha1"
 	"github.com/openshift/lvm-operator/controllers"
+	"github.com/openshift/lvm-operator/pkg/dmsetup"
 	"github.com/openshift/lvm-operator/pkg/filter"
 	"github.com/openshift/lvm-operator/pkg/lsblk"
 	"github.com/openshift/lvm-operator/pkg/lvm"
 	"github.com/openshift/lvm-operator/pkg/lvmd"
+	"github.com/openshift/lvm-operator/pkg/wipefs"
+
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/errors"
 	v1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -87,6 +90,8 @@ type VGReconciler struct {
 	LVMD lvmd.Configurator
 	lvm.LVM
 	lsblk.LSBLK
+	wipefs.Wipefs
+	dmsetup.Dmsetup
 	NodeName  string
 	Namespace string
 	Filters   func(*lvmv1alpha1.LVMVolumeGroup, lvm.LVM, lsblk.LSBLK) filter.Filters
@@ -128,12 +133,13 @@ func (r *VGReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Re
 		}
 	}
 
-	return r.reconcile(ctx, volumeGroup)
+	return r.reconcile(ctx, volumeGroup, nodeStatus)
 }
 
 func (r *VGReconciler) reconcile(
 	ctx context.Context,
 	volumeGroup *lvmv1alpha1.LVMVolumeGroup,
+	nodeStatus *lvmv1alpha1.LVMVolumeGroupNodeStatus,
 ) (ctrl.Result, error) {
 	logger := log.FromContext(ctx)
 	// Check if the LVMVolumeGroup resource is deleted
@@ -184,10 +190,21 @@ func (r *VGReconciler) reconcile(
 		return ctrl.Result{}, fmt.Errorf("failed to list block devices: %w", err)
 	}
 
+	wiped, err := r.wipeDevicesIfNecessary(ctx, volumeGroup, nodeStatus, blockDevices)
+	if err != nil {
+		return ctrl.Result{}, fmt.Errorf("failed to wipe devices: %w", err)
+	}
+	if wiped {
+		blockDevices, err = r.LSBLK.ListBlockDevices()
+		if err != nil {
+			return ctrl.Result{}, fmt.Errorf("failed to list block devices: %w", err)
+		}
+	}
+
 	// Get the available block devices that can be used for this volume group
 	// valid means that it can be used to create or extend the volume group
 	// devices that are already part of the volume group will not be returned
-	newDevices, err := r.getNewDevicesToBeAdded(ctx, blockDevices, vgs, volumeGroup)
+	newDevices, err := r.getNewDevicesToBeAdded(ctx, blockDevices, nodeStatus, volumeGroup)
 	if err != nil {
 		return ctrl.Result{}, fmt.Errorf("failed to get matching available block devices for volumegroup %s: %w", volumeGroup.GetName(), err)
 	}