Support bucket versioning (#9377)

- Implement a new xl.json 2.0.0 format to support,
  this moves the entire marshaling logic to POSIX
  layer, top layer always consumes a common FileInfo
  construct which simplifies the metadata reads.
- Implement list object versions
- Migrate to siphash from crchash for new deployments
  for object placements.

Fixes #2111

Dockerfile.dev.browser

@@ -10,4 +10,3 @@ ENV PATH=$PATH:/root/go/bin
 
 RUN go get github.com/go-bindata/go-bindata/go-bindata && \
     go get github.com/elazarl/go-bindata-assetfs/go-bindata-assetfs
-

cmd/admin-handlers.go

@@ -631,7 +631,7 @@ func (a adminAPIHandlers) HealHandler(w http.ResponseWriter, r *http.Request) {
 	}
 
 	// Check if this setup has an erasure coded backend.
-	if !globalIsXL {
+	if !globalIsErasure {
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrHealNotImplemented), r.URL)
 		return
 	}
@@ -779,7 +779,7 @@ func (a adminAPIHandlers) BackgroundHealStatusHandler(w http.ResponseWriter, r *
 	}
 
 	// Check if this setup has an erasure coded backend.
-	if !globalIsXL {
+	if !globalIsErasure {
 		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrHealNotImplemented), r.URL)
 		return
 	}
@@ -789,7 +789,7 @@ func (a adminAPIHandlers) BackgroundHealStatusHandler(w http.ResponseWriter, r *
 	// Get local heal status first
 	bgHealStates = append(bgHealStates, getLocalBackgroundHealStatus())
 
-	if globalIsDistXL {
+	if globalIsDistErasure {
 		// Get heal status from other peers
 		peersHealStates := globalNotificationSys.BackgroundHealStatus()
 		bgHealStates = append(bgHealStates, peersHealStates...)
@@ -862,11 +862,11 @@ const (
 	AdminUpdateApplyFailure      = "XMinioAdminUpdateApplyFailure"
 )
 
-// toAdminAPIErrCode - converts errXLWriteQuorum error to admin API
+// toAdminAPIErrCode - converts errErasureWriteQuorum error to admin API
 // specific error.
 func toAdminAPIErrCode(ctx context.Context, err error) APIErrorCode {
 	switch err {
-	case errXLWriteQuorum:
+	case errErasureWriteQuorum:
 		return ErrAdminConfigNoQuorum
 	default:
 		return toAPIErrorCode(ctx, err)
@@ -1277,7 +1277,7 @@ func (a adminAPIHandlers) OBDInfoHandler(w http.ResponseWriter, r *http.Request)
 			partialWrite(obdInfo)
 		}
 
-		if net, ok := vars["perfnet"]; ok && net == "true" && globalIsDistXL {
+		if net, ok := vars["perfnet"]; ok && net == "true" && globalIsDistErasure {
 			obdInfo.Perf.Net = append(obdInfo.Perf.Net, globalNotificationSys.NetOBDInfo(deadlinedCtx))
 			partialWrite(obdInfo)
 
@@ -1384,7 +1384,7 @@ func (a adminAPIHandlers) ServerInfoHandler(w http.ResponseWriter, r *http.Reque
 			OffDisks += v
 		}
 
-		backend = madmin.XLBackend{
+		backend = madmin.ErasureBackend{
 			Type:             madmin.ErasureType,
 			OnlineDisks:      OnDisks,
 			OfflineDisks:     OffDisks,
@@ -1413,10 +1413,10 @@ func (a adminAPIHandlers) ServerInfoHandler(w http.ResponseWriter, r *http.Reque
 	for _, sp := range servers {
 		for i, di := range sp.Disks {
 			path := ""
-			if globalIsXL {
+			if globalIsErasure {
 				path = di.DrivePath
 			}
-			if globalIsDistXL {
+			if globalIsDistErasure {
 				path = sp.Endpoint + di.DrivePath
 			}
 			// For distributed
@@ -1424,13 +1424,13 @@ func (a adminAPIHandlers) ServerInfoHandler(w http.ResponseWriter, r *http.Reque
 				for b := range storageInfo.Backend.Sets[a] {
 					ep := storageInfo.Backend.Sets[a][b].Endpoint
 
-					if globalIsDistXL {
+					if globalIsDistErasure {
 						if strings.Replace(ep, "http://", "", -1) == path || strings.Replace(ep, "https://", "", -1) == path {
 							sp.Disks[i].State = storageInfo.Backend.Sets[a][b].State
 							sp.Disks[i].UUID = storageInfo.Backend.Sets[a][b].UUID
 						}
 					}
-					if globalIsXL {
+					if globalIsErasure {
 						if ep == path {
 							sp.Disks[i].State = storageInfo.Backend.Sets[a][b].State
 							sp.Disks[i].UUID = storageInfo.Backend.Sets[a][b].UUID

cmd/admin-handlers_test.go

@@ -33,27 +33,27 @@ import (
 	"github.com/minio/minio/pkg/madmin"
 )
 
-// adminXLTestBed - encapsulates subsystems that need to be setup for
+// adminErasureTestBed - encapsulates subsystems that need to be setup for
 // admin-handler unit tests.
-type adminXLTestBed struct {
+type adminErasureTestBed struct {
-	xlDirs   []string
+	erasureDirs []string
 	objLayer    ObjectLayer
 	router      *mux.Router
 }
 
-// prepareAdminXLTestBed - helper function that setups a single-node
+// prepareAdminErasureTestBed - helper function that setups a single-node
-// XL backend for admin-handler tests.
+// Erasure backend for admin-handler tests.
-func prepareAdminXLTestBed(ctx context.Context) (*adminXLTestBed, error) {
+func prepareAdminErasureTestBed(ctx context.Context) (*adminErasureTestBed, error) {
 
 	// reset global variables to start afresh.
 	resetTestGlobals()
 
-	// Set globalIsXL to indicate that the setup uses an erasure
+	// Set globalIsErasure to indicate that the setup uses an erasure
 	// code backend.
-	globalIsXL = true
+	globalIsErasure = true
 
 	// Initializing objectLayer for HealFormatHandler.
-	objLayer, xlDirs, xlErr := initTestXLObjLayer(ctx)
+	objLayer, erasureDirs, xlErr := initTestErasureObjLayer(ctx)
 	if xlErr != nil {
 		return nil, xlErr
 	}
@@ -66,7 +66,7 @@ func prepareAdminXLTestBed(ctx context.Context) (*adminXLTestBed, error) {
 	// Initialize boot time
 	globalBootTime = UTCNow()
 
-	globalEndpoints = mustGetZoneEndpoints(xlDirs...)
+	globalEndpoints = mustGetZoneEndpoints(erasureDirs...)
 
 	newAllSubsystems()
 
@@ -76,8 +76,8 @@ func prepareAdminXLTestBed(ctx context.Context) (*adminXLTestBed, error) {
 	adminRouter := mux.NewRouter()
 	registerAdminRouter(adminRouter, true, true)
 
-	return &adminXLTestBed{
+	return &adminErasureTestBed{
-		xlDirs:   xlDirs,
+		erasureDirs: erasureDirs,
 		objLayer:    objLayer,
 		router:      adminRouter,
 	}, nil
@@ -85,27 +85,28 @@ func prepareAdminXLTestBed(ctx context.Context) (*adminXLTestBed, error) {
 
 // TearDown - method that resets the test bed for subsequent unit
 // tests to start afresh.
-func (atb *adminXLTestBed) TearDown() {
+func (atb *adminErasureTestBed) TearDown() {
-	removeRoots(atb.xlDirs)
+	removeRoots(atb.erasureDirs)
 	resetTestGlobals()
 }
 
-// initTestObjLayer - Helper function to initialize an XL-based object
+// initTestObjLayer - Helper function to initialize an Erasure-based object
 // layer and set globalObjectAPI.
-func initTestXLObjLayer(ctx context.Context) (ObjectLayer, []string, error) {
+func initTestErasureObjLayer(ctx context.Context) (ObjectLayer, []string, error) {
-	xlDirs, err := getRandomDisks(16)
+	erasureDirs, err := getRandomDisks(16)
 	if err != nil {
 		return nil, nil, err
 	}
-	endpoints := mustGetNewEndpoints(xlDirs...)
+	endpoints := mustGetNewEndpoints(erasureDirs...)
-	storageDisks, format, err := waitForFormatXL(true, endpoints, 1, 1, 16, "")
+	storageDisks, format, err := waitForFormatErasure(true, endpoints, 1, 1, 16, "")
 	if err != nil {
-		removeRoots(xlDirs)
+		removeRoots(erasureDirs)
 		return nil, nil, err
 	}
 
 	globalPolicySys = NewPolicySys()
-	objLayer, err := newXLSets(ctx, endpoints, storageDisks, format)
+	objLayer := &erasureZones{zones: make([]*erasureSets, 1)}
+	objLayer.zones[0], err = newErasureSets(ctx, endpoints, storageDisks, format)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -114,7 +115,7 @@ func initTestXLObjLayer(ctx context.Context) (ObjectLayer, []string, error) {
 	globalObjLayerMutex.Lock()
 	globalObjectAPI = objLayer
 	globalObjLayerMutex.Unlock()
-	return objLayer, xlDirs, nil
+	return objLayer, erasureDirs, nil
 }
 
 // cmdType - Represents different service subcomands like status, stop
@@ -183,9 +184,9 @@ func testServicesCmdHandler(cmd cmdType, t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	adminTestBed, err := prepareAdminXLTestBed(ctx)
+	adminTestBed, err := prepareAdminErasureTestBed(ctx)
 	if err != nil {
-		t.Fatal("Failed to initialize a single node XL backend for admin handler tests.")
+		t.Fatal("Failed to initialize a single node Erasure backend for admin handler tests.")
 	}
 	defer adminTestBed.TearDown()
 
@@ -254,9 +255,9 @@ func TestAdminServerInfo(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	adminTestBed, err := prepareAdminXLTestBed(ctx)
+	adminTestBed, err := prepareAdminErasureTestBed(ctx)
 	if err != nil {
-		t.Fatal("Failed to initialize a single node XL backend for admin handler tests.")
+		t.Fatal("Failed to initialize a single node Erasure backend for admin handler tests.")
 	}
 
 	defer adminTestBed.TearDown()
@@ -298,7 +299,7 @@ func TestToAdminAPIErrCode(t *testing.T) {
 	}{
 		// 1. Server not in quorum.
 		{
-			err:            errXLWriteQuorum,
+			err:            errErasureWriteQuorum,
 			expectedAPIErr: ErrAdminConfigNoQuorum,
 		},
 		// 2. No error.

cmd/admin-heal-ops.go

@@ -21,7 +21,6 @@ import (
 	"encoding/json"
 	"fmt"
 	"net/http"
-	"strings"
 	"sync"
 	"time"
 
@@ -193,7 +192,7 @@ func (ahs *allHealState) LaunchNewHealSequence(h *healSequence) (
 	respBytes []byte, apiErr APIError, errMsg string) {
 
 	existsAndLive := false
-	he, exists := ahs.getHealSequence(h.path)
+	he, exists := ahs.getHealSequence(pathJoin(h.bucket, h.object))
 	if exists {
 		existsAndLive = !he.hasEnded()
 	}
@@ -220,8 +219,9 @@ func (ahs *allHealState) LaunchNewHealSequence(h *healSequence) (
 
 	// Check if new heal sequence to be started overlaps with any
 	// existing, running sequence
+	hpath := pathJoin(h.bucket, h.object)
 	for k, hSeq := range ahs.healSeqMap {
-		if !hSeq.hasEnded() && (HasPrefix(k, h.path) || HasPrefix(h.path, k)) {
+		if !hSeq.hasEnded() && (HasPrefix(k, hpath) || HasPrefix(hpath, k)) {
 
 			errMsg = "The provided heal sequence path overlaps with an existing " +
 				fmt.Sprintf("heal path: %s", k)
@@ -230,7 +230,7 @@ func (ahs *allHealState) LaunchNewHealSequence(h *healSequence) (
 	}
 
 	// Add heal state and start sequence
-	ahs.healSeqMap[h.path] = h
+	ahs.healSeqMap[hpath] = h
 
 	// Launch top-level background heal go-routine
 	go h.healSequenceStart()
@@ -251,11 +251,11 @@ func (ahs *allHealState) LaunchNewHealSequence(h *healSequence) (
 // status results from global state and returns its JSON
 // representation. The clientToken helps ensure there aren't
 // conflicting clients fetching status.
-func (ahs *allHealState) PopHealStatusJSON(path string,
+func (ahs *allHealState) PopHealStatusJSON(hpath string,
 	clientToken string) ([]byte, APIErrorCode) {
 
 	// fetch heal state for given path
-	h, exists := ahs.getHealSequence(path)
+	h, exists := ahs.getHealSequence(hpath)
 	if !exists {
 		// If there is no such heal sequence, return error.
 		return nil, ErrHealNoSuchProcess
@@ -296,18 +296,17 @@ func (ahs *allHealState) PopHealStatusJSON(path string,
 
 // healSource denotes single entity and heal option.
 type healSource struct {
-	path string           // entity path (format, buckets, objects) to heal
+	bucket    string
+	object    string
+	versionID string
 	opts      *madmin.HealOpts // optional heal option overrides default setting
 }
 
 // healSequence - state for each heal sequence initiated on the
 // server.
 type healSequence struct {
-	// bucket, and prefix on which heal seq. was initiated
+	// bucket, and object on which heal seq. was initiated
-	bucket, objPrefix string
+	bucket, object string
-
-	// path is just pathJoin(bucket, objPrefix)
-	path string
 
 	// A channel of entities (format, buckets, objects) to heal
 	sourceCh chan healSource
@@ -377,8 +376,7 @@ func newHealSequence(ctx context.Context, bucket, objPrefix, clientAddr string,
 	return &healSequence{
 		respCh:         make(chan healResult),
 		bucket:         bucket,
-		objPrefix:      objPrefix,
+		object:         objPrefix,
-		path:           pathJoin(bucket, objPrefix),
 		reportProgress: true,
 		startTime:      UTCNow(),
 		clientToken:    mustGetUUID(),
@@ -618,7 +616,9 @@ func (h *healSequence) healSequenceStart() {
 func (h *healSequence) queueHealTask(source healSource, healType madmin.HealItemType) error {
 	// Send heal request
 	task := healTask{
-		path:       source.path,
+		bucket:     source.bucket,
+		object:     source.object,
+		versionID:  source.versionID,
 		opts:       h.settings,
 		responseCh: h.respCh,
 	}
@@ -690,11 +690,11 @@ func (h *healSequence) healItemsFromSourceCh() error {
 			}
 			var itemType madmin.HealItemType
 			switch {
-			case source.path == nopHeal:
+			case source.bucket == nopHeal:
 				continue
-			case source.path == SlashSeparator:
+			case source.bucket == SlashSeparator:
 				itemType = madmin.HealItemMetadata
-			case !strings.Contains(source.path, SlashSeparator):
+			case source.bucket != "" && source.object == "":
 				itemType = madmin.HealItemBucket
 			default:
 				itemType = madmin.HealItemObject
@@ -762,12 +762,16 @@ func (h *healSequence) healMinioSysMeta(metaPrefix string) func() error {
 		// NOTE: Healing on meta is run regardless
 		// of any bucket being selected, this is to ensure that
 		// meta are always upto date and correct.
-		return objectAPI.HealObjects(h.ctx, minioMetaBucket, metaPrefix, h.settings, func(bucket string, object string) error {
+		return objectAPI.HealObjects(h.ctx, minioMetaBucket, metaPrefix, h.settings, func(bucket, object, versionID string) error {
 			if h.isQuitting() {
 				return errHealStopSignalled
 			}
 
-			herr := h.queueHealTask(healSource{path: pathJoin(bucket, object)}, madmin.HealItemBucketMetadata)
+			herr := h.queueHealTask(healSource{
+				bucket:    bucket,
+				object:    object,
+				versionID: versionID,
+			}, madmin.HealItemBucketMetadata)
 			// Object might have been deleted, by the time heal
 			// was attempted we ignore this object an move on.
 			if isErrObjectNotFound(herr) {
@@ -791,7 +795,7 @@ func (h *healSequence) healDiskFormat() error {
 		return errServerNotInitialized
 	}
 
-	return h.queueHealTask(healSource{path: SlashSeparator}, madmin.HealItemMetadata)
+	return h.queueHealTask(healSource{bucket: SlashSeparator}, madmin.HealItemMetadata)
 }
 
 // healBuckets - check for all buckets heal or just particular bucket.
@@ -833,7 +837,7 @@ func (h *healSequence) healBucket(bucket string, bucketsOnly bool) error {
 		return errServerNotInitialized
 	}
 
-	if err := h.queueHealTask(healSource{path: bucket}, madmin.HealItemBucket); err != nil {
+	if err := h.queueHealTask(healSource{bucket: bucket}, madmin.HealItemBucket); err != nil {
 		return err
 	}
 
@@ -842,12 +846,12 @@ func (h *healSequence) healBucket(bucket string, bucketsOnly bool) error {
 	}
 
 	if !h.settings.Recursive {
-		if h.objPrefix != "" {
+		if h.object != "" {
 			// Check if an object named as the objPrefix exists,
 			// and if so heal it.
-			_, err := objectAPI.GetObjectInfo(h.ctx, bucket, h.objPrefix, ObjectOptions{})
+			_, err := objectAPI.GetObjectInfo(h.ctx, bucket, h.object, ObjectOptions{})
 			if err == nil {
-				if err = h.healObject(bucket, h.objPrefix); err != nil {
+				if err = h.healObject(bucket, h.object, ""); err != nil {
 					return err
 				}
 			}
@@ -856,14 +860,14 @@ func (h *healSequence) healBucket(bucket string, bucketsOnly bool) error {
 		return nil
 	}
 
-	if err := objectAPI.HealObjects(h.ctx, bucket, h.objPrefix, h.settings, h.healObject); err != nil {
+	if err := objectAPI.HealObjects(h.ctx, bucket, h.object, h.settings, h.healObject); err != nil {
 		return errFnHealFromAPIErr(h.ctx, err)
 	}
 	return nil
 }
 
 // healObject - heal the given object and record result
-func (h *healSequence) healObject(bucket, object string) error {
+func (h *healSequence) healObject(bucket, object, versionID string) error {
 	// Get current object layer instance.
 	objectAPI := newObjectLayerWithoutSafeModeFn()
 	if objectAPI == nil {
@@ -874,5 +878,9 @@ func (h *healSequence) healObject(bucket, object string) error {
 		return errHealStopSignalled
 	}
 
-	return h.queueHealTask(healSource{path: pathJoin(bucket, object)}, madmin.HealItemObject)
+	return h.queueHealTask(healSource{
+		bucket:    bucket,
+		object:    object,
+		versionID: versionID,
+	}, madmin.HealItemObject)
 }

cmd/admin-router.go

@@ -64,7 +64,7 @@ func registerAdminRouter(router *mux.Router, enableConfigOps, enableIAMOps bool)
 		// DataUsageInfo operations
 		adminRouter.Methods(http.MethodGet).Path(adminVersion + "/datausageinfo").HandlerFunc(httpTraceAll(adminAPI.DataUsageInfoHandler))
 
-		if globalIsDistXL || globalIsXL {
+		if globalIsDistErasure || globalIsErasure {
 			/// Heal operations
 
 			// Heal processing endpoint.
@@ -172,7 +172,7 @@ func registerAdminRouter(router *mux.Router, enableConfigOps, enableIAMOps bool)
 		}
 
 		// Quota operations
-		if globalIsXL || globalIsDistXL {
+		if globalIsDistErasure || globalIsErasure {
 			if env.Get(envDataUsageCrawlConf, config.EnableOn) == config.EnableOn {
 				// GetBucketQuotaConfig
 				adminRouter.Methods(http.MethodGet).Path(adminVersion+"/get-bucket-quota").HandlerFunc(
@@ -185,7 +185,7 @@ func registerAdminRouter(router *mux.Router, enableConfigOps, enableIAMOps bool)
 
 		// -- Top APIs --
 		// Top locks
-		if globalIsDistXL {
+		if globalIsDistErasure {
 			adminRouter.Methods(http.MethodGet).Path(adminVersion + "/top/locks").HandlerFunc(httpTraceHdrs(adminAPI.TopLocksHandler))
 		}
 

cmd/admin-server-info.go

@@ -29,7 +29,7 @@ import (
 func getLocalServerProperty(endpointZones EndpointZones, r *http.Request) madmin.ServerProperties {
 	var disks []madmin.Disk
 	addr := r.Host
-	if globalIsDistXL {
+	if globalIsDistErasure {
 		addr = GetLocalPeer(endpointZones)
 	}
 	network := make(map[string]string)

cmd/api-datatypes.go

@@ -20,9 +20,18 @@ import (
 	"encoding/xml"
 )
 
-// ObjectIdentifier carries key name for the object to delete.
+// DeletedObject objects deleted
-type ObjectIdentifier struct {
+type DeletedObject struct {
+	DeleteMarker          bool   `xml:"DeleteMarker"`
+	DeleteMarkerVersionID string `xml:"DeleteMarkerVersionId,omitempty"`
+	ObjectName            string `xml:"Key,omitempty"`
+	VersionID             string `xml:"VersionId,omitempty"`
+}
+
+// ObjectToDelete carries key name for the object to delete.
+type ObjectToDelete struct {
 	ObjectName string `xml:"Key"`
+	VersionID  string `xml:"VersionId"`
 }
 
 // createBucketConfiguration container for bucket configuration request from client.
@@ -37,5 +46,5 @@ type DeleteObjectsRequest struct {
 	// Element to enable quiet mode for the request
 	Quiet bool
 	// List of objects to be deleted
-	Objects []ObjectIdentifier `xml:"Object"`
+	Objects []ObjectToDelete `xml:"Object"`
 }

cmd/api-errors.go

@@ -36,6 +36,7 @@ import (
 	"github.com/minio/minio/pkg/bucket/lifecycle"
 	objectlock "github.com/minio/minio/pkg/bucket/object/lock"
 	"github.com/minio/minio/pkg/bucket/policy"
+	"github.com/minio/minio/pkg/bucket/versioning"
 	"github.com/minio/minio/pkg/event"
 	"github.com/minio/minio/pkg/hash"
 )
@@ -538,9 +539,9 @@ var errorCodes = errorCodeMap{
 		HTTPStatusCode: http.StatusNotFound,
 	},
 	ErrNoSuchVersion: {
-		Code:           "NoSuchVersion",
+		Code:           "InvalidArgument",
-		Description:    "Indicates that the version ID specified in the request does not match an existing version.",
+		Description:    "Invalid version id specified",
-		HTTPStatusCode: http.StatusNotFound,
+		HTTPStatusCode: http.StatusBadRequest,
 	},
 	ErrNotImplemented: {
 		Code:           "NotImplemented",
@@ -1782,6 +1783,10 @@ func toAPIErrorCode(ctx context.Context, err error) (apiErr APIErrorCode) {
 		apiErr = ErrBucketAlreadyOwnedByYou
 	case ObjectNotFound:
 		apiErr = ErrNoSuchKey
+	case MethodNotAllowed:
+		apiErr = ErrMethodNotAllowed
+	case VersionNotFound:
+		apiErr = ErrNoSuchVersion
 	case ObjectAlreadyExists:
 		apiErr = ErrMethodNotAllowed
 	case ObjectNameInvalid:
@@ -1918,6 +1923,12 @@ func toAPIError(ctx context.Context, err error) APIError {
 					e.Error()),
 				HTTPStatusCode: http.StatusBadRequest,
 			}
+		case versioning.Error:
+			apiErr = APIError{
+				Code:           "IllegalVersioningConfigurationException",
+				Description:    fmt.Sprintf("Versioning configuration specified in the request is invalid. (%s)", e.Error()),
+				HTTPStatusCode: http.StatusBadRequest,
+			}
 		case lifecycle.Error:
 			apiErr = APIError{
 				Code:           "InvalidRequest",

cmd/api-headers.go

@@ -29,6 +29,7 @@ import (
 
 	"github.com/minio/minio/cmd/crypto"
 	xhttp "github.com/minio/minio/cmd/http"
+	"github.com/minio/minio/pkg/bucket/lifecycle"
 )
 
 // Returns a hexadecimal representation of time at the
@@ -152,5 +153,26 @@ func setObjectHeaders(w http.ResponseWriter, objInfo ObjectInfo, rs *HTTPRangeSp
 		w.Header().Set(xhttp.ContentRange, contentRange)
 	}
 
+	// Set the relevant version ID as part of the response header.
+	if objInfo.VersionID != "" {
+		w.Header()[xhttp.AmzVersionID] = []string{objInfo.VersionID}
+	}
+
+	if lc, err := globalLifecycleSys.Get(objInfo.Bucket); err == nil {
+		ruleID, expiryTime := lc.PredictExpiryTime(lifecycle.ObjectOpts{
+			Name:         objInfo.Name,
+			UserTags:     objInfo.UserTags,
+			VersionID:    objInfo.VersionID,
+			ModTime:      objInfo.ModTime,
+			IsLatest:     objInfo.IsLatest,
+			DeleteMarker: objInfo.DeleteMarker,
+		})
+		if !expiryTime.IsZero() {
+			w.Header()[xhttp.AmzExpiration] = []string{
+				fmt.Sprintf(`expiry-date="%s", rule-id="%s"`, expiryTime.Format(http.TimeFormat), ruleID),
+			}
+		}
+	}
+
 	return nil
 }

cmd/api-response.go

@@ -81,6 +81,7 @@ type ListVersionsResponse struct {
 
 	CommonPrefixes []CommonPrefix
 	Versions       []ObjectVersion
+	DeleteMarkers  []DeletedVersion
 
 	// Encoding type used to encode object keys in the response.
 	EncodingType string `xml:"EncodingType,omitempty"`
@@ -237,8 +238,22 @@ type Bucket struct {
 type ObjectVersion struct {
 	XMLName xml.Name `xml:"http://s3.amazonaws.com/doc/2006-03-01/ Version" json:"-"`
 	Object
-	VersionID string `xml:"VersionId"`
 	IsLatest  bool
+	VersionID string `xml:"VersionId"`
+}
+
+// DeletedVersion container for the delete object version metadata.
+type DeletedVersion struct {
+	XMLName xml.Name `xml:"http://s3.amazonaws.com/doc/2006-03-01/ DeleteMarker" json:"-"`
+
+	IsLatest     bool
+	Key          string
+	LastModified string // time string of format "2006-01-02T15:04:05.000Z"
+
+	// Owner of the object.
+	Owner Owner
+
+	VersionID string `xml:"VersionId"`
 }
 
 // StringMap is a map[string]string.
@@ -336,6 +351,7 @@ type DeleteError struct {
 	Code      string
 	Message   string
 	Key       string
+	VersionID string `xml:"VersionId"`
 }
 
 // DeleteObjectsResponse container for multiple object deletes.
@@ -343,7 +359,7 @@ type DeleteObjectsResponse struct {
 	XMLName xml.Name `xml:"http://s3.amazonaws.com/doc/2006-03-01/ DeleteResult" json:"-"`
 
 	// Collection of all deleted objects
-	DeletedObjects []ObjectIdentifier `xml:"Deleted,omitempty"`
+	DeletedObjects []DeletedObject `xml:"Deleted,omitempty"`
 
 	// Collection of errors deleting certain objects.
 	Errors []DeleteError `xml:"Error,omitempty"`
@@ -413,8 +429,9 @@ func generateListBucketsResponse(buckets []BucketInfo) ListBucketsResponse {
 }
 
 // generates an ListBucketVersions response for the said bucket with other enumerated options.
-func generateListVersionsResponse(bucket, prefix, marker, delimiter, encodingType string, maxKeys int, resp ListObjectsInfo) ListVersionsResponse {
+func generateListVersionsResponse(bucket, prefix, marker, versionIDMarker, delimiter, encodingType string, maxKeys int, resp ListObjectVersionsInfo) ListVersionsResponse {
 	var versions []ObjectVersion
+	var deletedVersions []DeletedVersion
 	var prefixes []CommonPrefix
 	var owner = Owner{}
 	var data = ListVersionsResponse{}
@@ -436,15 +453,29 @@ func generateListVersionsResponse(bucket, prefix, marker, delimiter, encodingTyp
 		} else {
 			content.StorageClass = globalMinioDefaultStorageClass
 		}
-
 		content.Owner = owner
-		content.VersionID = "null"
+		content.VersionID = object.VersionID
-		content.IsLatest = true
+		if content.VersionID == "" {
+			content.VersionID = nullVersionID
+		}
+		content.IsLatest = object.IsLatest
 		versions = append(versions, content)
 	}
 
+	for _, deleted := range resp.DeleteObjects {
+		var dv = DeletedVersion{
+			Key:          s3EncodeName(deleted.Name, encodingType),
+			Owner:        owner,
+			LastModified: deleted.ModTime.UTC().Format(iso8601TimeFormat),
+			VersionID:    deleted.VersionID,
+			IsLatest:     deleted.IsLatest,
+		}
+		deletedVersions = append(deletedVersions, dv)
+	}
+
 	data.Name = bucket
 	data.Versions = versions
+	data.DeleteMarkers = deletedVersions
 	data.EncodingType = encodingType
 	data.Prefix = s3EncodeName(prefix, encodingType)
 	data.KeyMarker = s3EncodeName(marker, encodingType)
@@ -452,6 +483,8 @@ func generateListVersionsResponse(bucket, prefix, marker, delimiter, encodingTyp
 	data.MaxKeys = maxKeys
 
 	data.NextKeyMarker = s3EncodeName(resp.NextMarker, encodingType)
+	data.NextVersionIDMarker = resp.NextVersionIDMarker
+	data.VersionIDMarker = versionIDMarker
 	data.IsTruncated = resp.IsTruncated
 
 	for _, prefix := range resp.Prefixes {
@@ -666,11 +699,14 @@ func generateListMultipartUploadsResponse(bucket string, multipartsInfo ListMult
 }
 
 // generate multi objects delete response.
-func generateMultiDeleteResponse(quiet bool, deletedObjects []ObjectIdentifier, errs []DeleteError) DeleteObjectsResponse {
+func generateMultiDeleteResponse(quiet bool, deletedObjects []DeletedObject, errs []DeleteError) DeleteObjectsResponse {
 	deleteResp := DeleteObjectsResponse{}
 	if !quiet {
 		deleteResp.DeletedObjects = deletedObjects
 	}
+	if len(errs) == len(deletedObjects) {
+		deleteResp.DeletedObjects = nil
+	}
 	deleteResp.Errors = errs
 	return deleteResp
 }

cmd/api-router.go

@@ -224,9 +224,9 @@ func registerAPIRouter(router *mux.Router, encryptionEnabled, allowSSEKMS bool)
 		// ListObjectsV2
 		bucket.Methods(http.MethodGet).HandlerFunc(
 			maxClients(collectAPIStats("listobjectsv2", httpTraceAll(api.ListObjectsV2Handler)))).Queries("list-type", "2")
-		// ListBucketVersions
+		// ListObjectVersions
 		bucket.Methods(http.MethodGet).HandlerFunc(
-			maxClients(collectAPIStats("listbucketversions", httpTraceAll(api.ListBucketObjectVersionsHandler)))).Queries("versions", "")
+			maxClients(collectAPIStats("listobjectversions", httpTraceAll(api.ListObjectVersionsHandler)))).Queries("versions", "")
 		// ListObjectsV1 (Legacy)
 		bucket.Methods(http.MethodGet).HandlerFunc(
 			maxClients(collectAPIStats("listobjectsv1", httpTraceAll(api.ListObjectsV1Handler))))

cmd/background-heal-ops.go

@@ -18,6 +18,7 @@ package cmd
 
 import (
 	"context"
+	"path"
 	"time"
 
 	"github.com/minio/minio/cmd/logger"
@@ -29,7 +30,9 @@ import (
 //   path: 'bucket/' or '/bucket/' => Heal bucket
 //   path: 'bucket/object' => Heal object
 type healTask struct {
-	path string
+	bucket    string
+	object    string
+	versionID string
 	opts      madmin.HealOpts
 	// Healing response will be sent here
 	responseCh chan healResult
@@ -79,17 +82,18 @@ func (h *healRoutine) run(ctx context.Context, objAPI ObjectLayer) {
 
 			var res madmin.HealResultItem
 			var err error
-			bucket, object := path2BucketObject(task.path)
 			switch {
-			case bucket == "" && object == "":
+			case task.bucket == nopHeal:
+				continue
+			case task.bucket == SlashSeparator:
 				res, err = healDiskFormat(ctx, objAPI, task.opts)
-			case bucket != "" && object == "":
+			case task.bucket != "" && task.object == "":
-				res, err = objAPI.HealBucket(ctx, bucket, task.opts.DryRun, task.opts.Remove)
+				res, err = objAPI.HealBucket(ctx, task.bucket, task.opts.DryRun, task.opts.Remove)
-			case bucket != "" && object != "":
+			case task.bucket != "" && task.object != "":
-				res, err = objAPI.HealObject(ctx, bucket, object, task.opts)
+				res, err = objAPI.HealObject(ctx, task.bucket, task.object, task.versionID, task.opts)
 			}
-			if task.path != slashSeparator && task.path != nopHeal {
+			if task.bucket != "" && task.object != "" {
-				ObjectPathUpdated(task.path)
+				ObjectPathUpdated(path.Join(task.bucket, task.object))
 			}
 			task.responseCh <- healResult{result: res, err: err}
 		case <-h.doneCh:

cmd/background-newdisks-heal-ops.go

@@ -33,7 +33,7 @@ func initLocalDisksAutoHeal(ctx context.Context, objAPI ObjectLayer) {
 //  1. Only the concerned erasure set will be listed and healed
 //  2. Only the node hosting the disk is responsible to perform the heal
 func monitorLocalDisksAndHeal(ctx context.Context, objAPI ObjectLayer) {
-	z, ok := objAPI.(*xlZones)
+	z, ok := objAPI.(*erasureZones)
 	if !ok {
 		return
 	}
@@ -84,10 +84,10 @@ func monitorLocalDisksAndHeal(ctx context.Context, objAPI ObjectLayer) {
 			}
 
 			// Reformat disks
-			bgSeq.sourceCh <- healSource{path: SlashSeparator}
+			bgSeq.sourceCh <- healSource{bucket: SlashSeparator}
 
 			// Ensure that reformatting disks is finished
-			bgSeq.sourceCh <- healSource{path: nopHeal}
+			bgSeq.sourceCh <- healSource{bucket: nopHeal}
 
 			var erasureSetInZoneToHeal = make([][]int, len(localDisksInZoneHeal))
 			// Compute the list of erasure set to heal

cmd/benchmark-utils_test.go

@@ -35,7 +35,7 @@ func runPutObjectBenchmark(b *testing.B, obj ObjectLayer, objSize int) {
 	// obtains random bucket name.
 	bucket := getRandomBucketName()
 	// create bucket.
-	err = obj.MakeBucketWithLocation(context.Background(), bucket, "", false)
+	err = obj.MakeBucketWithLocation(context.Background(), bucket, BucketOptions{})
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -76,7 +76,7 @@ func runPutObjectPartBenchmark(b *testing.B, obj ObjectLayer, partSize int) {
 	object := getRandomObjectName()
 
 	// create bucket.
-	err = obj.MakeBucketWithLocation(context.Background(), bucket, "", false)
+	err = obj.MakeBucketWithLocation(context.Background(), bucket, BucketOptions{})
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -127,9 +127,9 @@ func runPutObjectPartBenchmark(b *testing.B, obj ObjectLayer, partSize int) {
 	b.StopTimer()
 }
 
-// creates XL/FS backend setup, obtains the object layer and calls the runPutObjectPartBenchmark function.
+// creates Erasure/FS backend setup, obtains the object layer and calls the runPutObjectPartBenchmark function.
 func benchmarkPutObjectPart(b *testing.B, instanceType string, objSize int) {
-	// create a temp XL/FS backend.
+	// create a temp Erasure/FS backend.
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 	objLayer, disks, err := prepareTestBackend(ctx, instanceType)
@@ -143,9 +143,9 @@ func benchmarkPutObjectPart(b *testing.B, instanceType string, objSize int) {
 	runPutObjectPartBenchmark(b, objLayer, objSize)
 }
 
-// creates XL/FS backend setup, obtains the object layer and calls the runPutObjectBenchmark function.
+// creates Erasure/FS backend setup, obtains the object layer and calls the runPutObjectBenchmark function.
 func benchmarkPutObject(b *testing.B, instanceType string, objSize int) {
-	// create a temp XL/FS backend.
+	// create a temp Erasure/FS backend.
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 	objLayer, disks, err := prepareTestBackend(ctx, instanceType)
@@ -159,9 +159,9 @@ func benchmarkPutObject(b *testing.B, instanceType string, objSize int) {
 	runPutObjectBenchmark(b, objLayer, objSize)
 }
 
-// creates XL/FS backend setup, obtains the object layer and runs parallel benchmark for put object.
+// creates Erasure/FS backend setup, obtains the object layer and runs parallel benchmark for put object.
 func benchmarkPutObjectParallel(b *testing.B, instanceType string, objSize int) {
-	// create a temp XL/FS backend.
+	// create a temp Erasure/FS backend.
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 	objLayer, disks, err := prepareTestBackend(ctx, instanceType)
@@ -181,7 +181,7 @@ func runGetObjectBenchmark(b *testing.B, obj ObjectLayer, objSize int) {
 	// obtains random bucket name.
 	bucket := getRandomBucketName()
 	// create bucket.
-	err := obj.MakeBucketWithLocation(context.Background(), bucket, "", false)
+	err := obj.MakeBucketWithLocation(context.Background(), bucket, BucketOptions{})
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -190,7 +190,7 @@ func runGetObjectBenchmark(b *testing.B, obj ObjectLayer, objSize int) {
 
 	// generate etag for the generated data.
 	// etag of the data to written is required as input for PutObject.
-	// PutObject is the functions which writes the data onto the FS/XL backend.
+	// PutObject is the functions which writes the data onto the FS/Erasure backend.
 
 	// get text data generated for number of bytes equal to object size.
 	md5hex := getMD5Hash(textData)
@@ -240,9 +240,9 @@ func generateBytesData(size int) []byte {
 	return bytes.Repeat(getRandomByte(), size)
 }
 
-// creates XL/FS backend setup, obtains the object layer and calls the runGetObjectBenchmark function.
+// creates Erasure/FS backend setup, obtains the object layer and calls the runGetObjectBenchmark function.
 func benchmarkGetObject(b *testing.B, instanceType string, objSize int) {
-	// create a temp XL/FS backend.
+	// create a temp Erasure/FS backend.
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 	objLayer, disks, err := prepareTestBackend(ctx, instanceType)
@@ -256,9 +256,9 @@ func benchmarkGetObject(b *testing.B, instanceType string, objSize int) {
 	runGetObjectBenchmark(b, objLayer, objSize)
 }
 
-// creates XL/FS backend setup, obtains the object layer and runs parallel benchmark for ObjectLayer.GetObject() .
+// creates Erasure/FS backend setup, obtains the object layer and runs parallel benchmark for ObjectLayer.GetObject() .
 func benchmarkGetObjectParallel(b *testing.B, instanceType string, objSize int) {
-	// create a temp XL/FS backend.
+	// create a temp Erasure/FS backend.
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 	objLayer, disks, err := prepareTestBackend(ctx, instanceType)
@@ -278,7 +278,7 @@ func runPutObjectBenchmarkParallel(b *testing.B, obj ObjectLayer, objSize int) {
 	// obtains random bucket name.
 	bucket := getRandomBucketName()
 	// create bucket.
-	err := obj.MakeBucketWithLocation(context.Background(), bucket, "", false)
+	err := obj.MakeBucketWithLocation(context.Background(), bucket, BucketOptions{})
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -322,7 +322,7 @@ func runGetObjectBenchmarkParallel(b *testing.B, obj ObjectLayer, objSize int) {
 	// obtains random bucket name.
 	bucket := getRandomBucketName()
 	// create bucket.
-	err := obj.MakeBucketWithLocation(context.Background(), bucket, "", false)
+	err := obj.MakeBucketWithLocation(context.Background(), bucket, BucketOptions{})
 	if err != nil {
 		b.Fatal(err)
 	}
@@ -331,7 +331,7 @@ func runGetObjectBenchmarkParallel(b *testing.B, obj ObjectLayer, objSize int) {
 	textData := generateBytesData(objSize)
 	// generate md5sum for the generated data.
 	// md5sum of the data to written is required as input for PutObject.
-	// PutObject is the functions which writes the data onto the FS/XL backend.
+	// PutObject is the functions which writes the data onto the FS/Erasure backend.
 
 	md5hex := getMD5Hash([]byte(textData))
 	sha256hex := ""

cmd/bitrot.go

@@ -30,25 +30,6 @@ import (
 // magic HH-256 key as HH-256 hash of the first 100 decimals of π as utf-8 string with a zero key.
 var magicHighwayHash256Key = []byte("\x4b\xe7\x34\xfa\x8e\x23\x8a\xcd\x26\x3e\x83\xe6\xbb\x96\x85\x52\x04\x0f\x93\x5d\xa3\x9f\x44\x14\x97\xe0\x9d\x13\x22\xde\x36\xa0")
 
-// BitrotAlgorithm specifies a algorithm used for bitrot protection.
-type BitrotAlgorithm uint
-
-const (
-	// SHA256 represents the SHA-256 hash function
-	SHA256 BitrotAlgorithm = 1 + iota
-	// HighwayHash256 represents the HighwayHash-256 hash function
-	HighwayHash256
-	// HighwayHash256S represents the Streaming HighwayHash-256 hash function
-	HighwayHash256S
-	// BLAKE2b512 represents the BLAKE2b-512 hash function
-	BLAKE2b512
-)
-
-// DefaultBitrotAlgorithm is the default algorithm used for bitrot protection.
-const (
-	DefaultBitrotAlgorithm = HighwayHash256S
-)
-
 var bitrotAlgorithms = map[BitrotAlgorithm]string{
 	SHA256:          "sha256",
 	BLAKE2b512:      "blake2b",

cmd/bitrot_test.go

@@ -34,7 +34,7 @@ func testBitrotReaderWriterAlgo(t *testing.T, bitrotAlgo BitrotAlgorithm) {
 	volume := "testvol"
 	filePath := "testfile"
 
-	disk, err := newPosix(tmpDir, "")
+	disk, err := newXLStorage(tmpDir, "")
 	if err != nil {
 		t.Fatal(err)
 	}

cmd/bucket-encryption.go

@@ -55,5 +55,6 @@ func validateBucketSSEConfig(r io.Reader) (*bucketsse.BucketSSEConfig, error) {
 	if len(encConfig.Rules) == 1 && encConfig.Rules[0].DefaultEncryptionAction.Algorithm == bucketsse.AES256 {
 		return encConfig, nil
 	}
+
 	return nil, errors.New("Unsupported bucket encryption configuration")
 }

cmd/bucket-handlers.go

@@ -45,7 +45,6 @@ import (
 )
 
 const (
-	getBucketVersioningResponse = `<VersioningConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/"/>`
 	objectLockConfig        = "object-lock.xml"
 	bucketTaggingConfigFile = "tagging.xml"
 )
@@ -382,75 +381,86 @@ func (api objectAPIHandlers) DeleteMultipleObjectsHandler(w http.ResponseWriter,
 		deleteObjectsFn = api.CacheAPI().DeleteObjects
 	}
 
-	var objectsToDelete = map[string]int{}
+	var objectsToDelete = map[ObjectToDelete]int{}
 	getObjectInfoFn := objectAPI.GetObjectInfo
 	if api.CacheAPI() != nil {
 		getObjectInfoFn = api.CacheAPI().GetObjectInfo
 	}
 
-	var dErrs = make([]APIErrorCode, len(deleteObjects.Objects))
+	dErrs := make([]DeleteError, len(deleteObjects.Objects))
-
 	for index, object := range deleteObjects.Objects {
-		if dErrs[index] = checkRequestAuthType(ctx, r, policy.DeleteObjectAction, bucket, object.ObjectName); dErrs[index] != ErrNone {
+		if apiErrCode := checkRequestAuthType(ctx, r, policy.DeleteObjectAction, bucket, object.ObjectName); apiErrCode != ErrNone {
-			if dErrs[index] == ErrSignatureDoesNotMatch || dErrs[index] == ErrInvalidAccessKeyID {
+			if apiErrCode == ErrSignatureDoesNotMatch || apiErrCode == ErrInvalidAccessKeyID {
-				writeErrorResponse(ctx, w, errorCodes.ToAPIErr(dErrs[index]), r.URL, guessIsBrowserReq(r))
+				writeErrorResponse(ctx, w, errorCodes.ToAPIErr(apiErrCode), r.URL, guessIsBrowserReq(r))
 				return
 			}
+			apiErr := errorCodes.ToAPIErr(apiErrCode)
+			dErrs[index] = DeleteError{
+				Code:      apiErr.Code,
+				Message:   apiErr.Description,
+				Key:       object.ObjectName,
+				VersionID: object.VersionID,
+			}
 			continue
 		}
 
+		if object.VersionID != "" {
 			if rcfg, _ := globalBucketObjectLockSys.Get(bucket); rcfg.LockEnabled {
-			if apiErr := enforceRetentionBypassForDelete(ctx, r, bucket, object.ObjectName, getObjectInfoFn); apiErr != ErrNone {
+				if apiErrCode := enforceRetentionBypassForDelete(ctx, r, bucket, object, getObjectInfoFn); apiErrCode != ErrNone {
-				dErrs[index] = apiErr
+					apiErr := errorCodes.ToAPIErr(apiErrCode)
+					dErrs[index] = DeleteError{
+						Code:      apiErr.Code,
+						Message:   apiErr.Description,
+						Key:       object.ObjectName,
+						VersionID: object.VersionID,
+					}
 					continue
 				}
 			}
+		}
 
 		// Avoid duplicate objects, we use map to filter them out.
-		if _, ok := objectsToDelete[object.ObjectName]; !ok {
+		if _, ok := objectsToDelete[object]; !ok {
-			objectsToDelete[object.ObjectName] = index
+			objectsToDelete[object] = index
 		}
 	}
 
-	toNames := func(input map[string]int) (output []string) {
+	toNames := func(input map[ObjectToDelete]int) (output []ObjectToDelete) {
-		output = make([]string, len(input))
+		output = make([]ObjectToDelete, len(input))
 		idx := 0
-		for name := range input {
+		for obj := range input {
-			output[idx] = name
+			output[idx] = obj
 			idx++
 		}
 		return
 	}
 
 	deleteList := toNames(objectsToDelete)
-	errs, err := deleteObjectsFn(ctx, bucket, deleteList)
+	dObjects, errs := deleteObjectsFn(ctx, bucket, deleteList, ObjectOptions{
-	if err != nil {
+		Versioned: globalBucketVersioningSys.Enabled(bucket),
-		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+	})
-		return
-	}
 
-	for i, objName := range deleteList {
+	deletedObjects := make([]DeletedObject, len(deleteObjects.Objects))
-		dIdx := objectsToDelete[objName]
+	for i := range errs {
-		dErrs[dIdx] = toAPIErrorCode(ctx, errs[i])
+		dindex := objectsToDelete[deleteList[i]]
-	}
+		apiErr := toAPIError(ctx, errs[i])
-
+		if apiErr.Code == "" || apiErr.Code == "NoSuchKey" {
-	// Collect deleted objects and errors if any.
+			deletedObjects[dindex] = dObjects[i]
-	var deletedObjects []ObjectIdentifier
-	var deleteErrors []DeleteError
-	for index, errCode := range dErrs {
-		object := deleteObjects.Objects[index]
-		// Success deleted objects are collected separately.
-		if errCode == ErrNone || errCode == ErrNoSuchKey {
-			deletedObjects = append(deletedObjects, object)
 			continue
 		}
-		apiErr := getAPIError(errCode)
+		dErrs[dindex] = DeleteError{
-		// Error during delete should be collected separately.
-		deleteErrors = append(deleteErrors, DeleteError{
 			Code:      apiErr.Code,
 			Message:   apiErr.Description,
-			Key:     object.ObjectName,
+			Key:       deleteList[i].ObjectName,
-		})
+			VersionID: deleteList[i].VersionID,
+		}
+	}
+
+	var deleteErrors []DeleteError
+	for _, dErr := range dErrs {
+		if dErr.Code != "" {
+			deleteErrors = append(deleteErrors, dErr)
+		}
 	}
 
 	// Generate response
@@ -462,12 +472,21 @@ func (api objectAPIHandlers) DeleteMultipleObjectsHandler(w http.ResponseWriter,
 
 	// Notify deleted event for objects.
 	for _, dobj := range deletedObjects {
+		objInfo := ObjectInfo{
+			Name:      dobj.ObjectName,
+			VersionID: dobj.VersionID,
+		}
+		if dobj.DeleteMarker {
+			objInfo = ObjectInfo{
+				Name:         dobj.ObjectName,
+				DeleteMarker: dobj.DeleteMarker,
+				VersionID:    dobj.DeleteMarkerVersionID,
+			}
+		}
 		sendEvent(eventArgs{
 			EventName:    event.ObjectRemovedDelete,
 			BucketName:   bucket,
-			Object: ObjectInfo{
+			Object:       objInfo,
-				Name: dobj.ObjectName,
-			},
 			ReqParams:    extractReqParams(r),
 			RespElements: extractRespElements(w),
 			UserAgent:    r.UserAgent(),
@@ -522,12 +541,17 @@ func (api objectAPIHandlers) PutBucketHandler(w http.ResponseWriter, r *http.Req
 		return
 	}
 
+	opts := BucketOptions{
+		Location:    location,
+		LockEnabled: objectLockEnabled,
+	}
+
 	if globalDNSConfig != nil {
 		sr, err := globalDNSConfig.Get(bucket)
 		if err != nil {
 			if err == dns.ErrNoEntriesFound {
 				// Proceed to creating a bucket.
-				if err = objectAPI.MakeBucketWithLocation(ctx, bucket, location, objectLockEnabled); err != nil {
+				if err = objectAPI.MakeBucketWithLocation(ctx, bucket, opts); err != nil {
 					writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 					return
 				}
@@ -565,7 +589,7 @@ func (api objectAPIHandlers) PutBucketHandler(w http.ResponseWriter, r *http.Req
 	}
 
 	// Proceed to creating a bucket.
-	err := objectAPI.MakeBucketWithLocation(ctx, bucket, location, objectLockEnabled)
+	err := objectAPI.MakeBucketWithLocation(ctx, bucket, opts)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
@@ -797,9 +821,17 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		return
 	}
 
-	location := getObjectLocation(r, globalDomainNames, bucket, object)
+	// We must not use the http.Header().Set method here because some (broken)
+	// clients expect the ETag header key to be literally "ETag" - not "Etag" (case-sensitive).
+	// Therefore, we have to set the ETag directly as map entry.
 	w.Header()[xhttp.ETag] = []string{`"` + objInfo.ETag + `"`}
-	w.Header().Set(xhttp.Location, location)
+
+	// Set the relevant version ID as part of the response header.
+	if objInfo.VersionID != "" {
+		w.Header()[xhttp.AmzVersionID] = []string{objInfo.VersionID}
+	}
+
+	w.Header().Set(xhttp.Location, getObjectLocation(r, globalDomainNames, bucket, object))
 
 	// Notify object created event.
 	defer sendEvent(eventArgs{
@@ -826,9 +858,9 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 			Bucket:   objInfo.Bucket,
 			Key:      objInfo.Name,
 			ETag:     `"` + objInfo.ETag + `"`,
-			Location: location,
+			Location: w.Header().Get(xhttp.Location),
 		})
-		writeResponse(w, http.StatusCreated, resp, "application/xml")
+		writeResponse(w, http.StatusCreated, resp, mimeXML)
 	case "200":
 		writeSuccessResponseHeadersOnly(w)
 	default:
@@ -921,79 +953,30 @@ func (api objectAPIHandlers) DeleteBucketHandler(w http.ResponseWriter, r *http.
 
 	// Attempt to delete bucket.
 	if err := deleteBucket(ctx, bucket, forceDelete); err != nil {
+		if _, ok := err.(BucketNotEmpty); ok && (globalBucketVersioningSys.Enabled(bucket) || globalBucketVersioningSys.Suspended(bucket)) {
+			apiErr := toAPIError(ctx, err)
+			apiErr.Description = "The bucket you tried to delete is not empty. You must delete all versions in the bucket."
+			writeErrorResponse(ctx, w, apiErr, r.URL, guessIsBrowserReq(r))
+		} else {
 			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
-		return
 		}
-
-	if globalDNSConfig != nil {
-		if err := globalDNSConfig.Delete(bucket); err != nil {
-			// Deleting DNS entry failed, attempt to create the bucket again.
-			objectAPI.MakeBucketWithLocation(ctx, bucket, "", false)
-			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
 	}
-	}
 
 	globalNotificationSys.DeleteBucketMetadata(ctx, bucket)
 
+	if globalDNSConfig != nil {
+		if err := globalDNSConfig.Delete(bucket); err != nil {
+			logger.LogIf(ctx, fmt.Errorf("Unable to delete bucket DNS entry %w, please delete it manually using etcdctl", err))
+			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+			return
+		}
+	}
+
 	// Write success response.
 	writeSuccessNoContent(w)
 }
 
-// PutBucketVersioningHandler - PUT Bucket Versioning.
-// ----------
-// No-op. Available for API compatibility.
-func (api objectAPIHandlers) PutBucketVersioningHandler(w http.ResponseWriter, r *http.Request) {
-	ctx := newContext(r, w, "PutBucketVersioning")
-
-	defer logger.AuditLog(w, r, "PutBucketVersioning", mustGetClaimsFromToken(r))
-
-	vars := mux.Vars(r)
-	bucket := vars["bucket"]
-
-	objectAPI := api.ObjectAPI()
-	if objectAPI == nil {
-		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL, guessIsBrowserReq(r))
-		return
-	}
-
-	getBucketInfo := objectAPI.GetBucketInfo
-	if _, err := getBucketInfo(ctx, bucket); err != nil {
-		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
-		return
-	}
-
-	// Write success response.
-	writeSuccessResponseHeadersOnly(w)
-}
-
-// GetBucketVersioningHandler - GET Bucket Versioning.
-// ----------
-// No-op. Available for API compatibility.
-func (api objectAPIHandlers) GetBucketVersioningHandler(w http.ResponseWriter, r *http.Request) {
-	ctx := newContext(r, w, "GetBucketVersioning")
-
-	defer logger.AuditLog(w, r, "GetBucketVersioning", mustGetClaimsFromToken(r))
-
-	vars := mux.Vars(r)
-	bucket := vars["bucket"]
-
-	objectAPI := api.ObjectAPI()
-	if objectAPI == nil {
-		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL, guessIsBrowserReq(r))
-		return
-	}
-
-	getBucketInfo := objectAPI.GetBucketInfo
-	if _, err := getBucketInfo(ctx, bucket); err != nil {
-		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
-		return
-	}
-
-	// Write success response.
-	writeSuccessResponseXML(w, []byte(getBucketVersioningResponse))
-}
-
 // PutBucketObjectLockConfigHandler - PUT Bucket object lock configuration.
 // ----------
 // Places an Object Lock configuration on the specified bucket. The rule

cmd/bucket-handlers_test.go

@@ -19,6 +19,7 @@ package cmd
 import (
 	"bytes"
 	"encoding/xml"
+	"fmt"
 	"io/ioutil"
 	"net/http"
 	"net/http/httptest"
@@ -28,7 +29,7 @@ import (
 	"github.com/minio/minio/pkg/auth"
 )
 
-// Wrapper for calling RemoveBucket HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling RemoveBucket HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestRemoveBucketHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testRemoveBucketHandler, []string{"RemoveBucket"})
 }
@@ -73,7 +74,7 @@ func testRemoveBucketHandler(obj ObjectLayer, instanceType, bucketName string, a
 	}
 }
 
-// Wrapper for calling GetBucketPolicy HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling GetBucketPolicy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestGetBucketLocationHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testGetBucketLocationHandler, []string{"GetBucketLocation"})
 }
@@ -217,7 +218,7 @@ func testGetBucketLocationHandler(obj ObjectLayer, instanceType, bucketName stri
 	ExecObjectLayerAPINilTest(t, nilBucket, "", instanceType, apiRouter, nilReq)
 }
 
-// Wrapper for calling HeadBucket HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling HeadBucket HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestHeadBucketHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testHeadBucketHandler, []string{"HeadBucket"})
 }
@@ -322,7 +323,7 @@ func testHeadBucketHandler(obj ObjectLayer, instanceType, bucketName string, api
 	ExecObjectLayerAPINilTest(t, nilBucket, "", instanceType, apiRouter, nilReq)
 }
 
-// Wrapper for calling TestListMultipartUploadsHandler tests for both XL multiple disks and single node setup.
+// Wrapper for calling TestListMultipartUploadsHandler tests for both Erasure multiple disks and single node setup.
 func TestListMultipartUploadsHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testListMultipartUploadsHandler, []string{"ListMultipartUploads"})
 }
@@ -559,7 +560,7 @@ func testListMultipartUploadsHandler(obj ObjectLayer, instanceType, bucketName s
 	ExecObjectLayerAPINilTest(t, nilBucket, "", instanceType, apiRouter, nilReq)
 }
 
-// Wrapper for calling TestListBucketsHandler tests for both XL multiple disks and single node setup.
+// Wrapper for calling TestListBucketsHandler tests for both Erasure multiple disks and single node setup.
 func TestListBucketsHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testListBucketsHandler, []string{"ListBuckets"})
 }
@@ -653,7 +654,7 @@ func testListBucketsHandler(obj ObjectLayer, instanceType, bucketName string, ap
 	ExecObjectLayerAPINilTest(t, "", "", instanceType, apiRouter, nilReq)
 }
 
-// Wrapper for calling DeleteMultipleObjects HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling DeleteMultipleObjects HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestAPIDeleteMultipleObjectsHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testAPIDeleteMultipleObjectsHandler, []string{"DeleteMultipleObjects"})
 }
@@ -679,14 +680,17 @@ func testAPIDeleteMultipleObjectsHandler(obj ObjectLayer, instanceType, bucketNa
 		objectNames = append(objectNames, objectName)
 	}
 
-	getObjectIdentifierList := func(objectNames []string) (objectIdentifierList []ObjectIdentifier) {
+	getObjectToDeleteList := func(objectNames []string) (objectList []ObjectToDelete) {
 		for _, objectName := range objectNames {
-			objectIdentifierList = append(objectIdentifierList, ObjectIdentifier{objectName})
+			objectList = append(objectList, ObjectToDelete{
+				ObjectName: objectName,
+			})
 		}
 
-		return objectIdentifierList
+		return objectList
 	}
-	getDeleteErrorList := func(objects []ObjectIdentifier) (deleteErrorList []DeleteError) {
+
+	getDeleteErrorList := func(objects []ObjectToDelete) (deleteErrorList []DeleteError) {
 		for _, obj := range objects {
 			deleteErrorList = append(deleteErrorList, DeleteError{
 				Code:    errorCodes[ErrAccessDenied].Code,
@@ -699,22 +703,38 @@ func testAPIDeleteMultipleObjectsHandler(obj ObjectLayer, instanceType, bucketNa
 	}
 
 	requestList := []DeleteObjectsRequest{
-		{Quiet: false, Objects: getObjectIdentifierList(objectNames[:5])},
+		{Quiet: false, Objects: getObjectToDeleteList(objectNames[:5])},
-		{Quiet: true, Objects: getObjectIdentifierList(objectNames[5:])},
+		{Quiet: true, Objects: getObjectToDeleteList(objectNames[5:])},
 	}
 
 	// generate multi objects delete response.
 	successRequest0 := encodeResponse(requestList[0])
-	successResponse0 := generateMultiDeleteResponse(requestList[0].Quiet, requestList[0].Objects, nil)
+
+	deletedObjects := make([]DeletedObject, len(requestList[0].Objects))
+	for i := range requestList[0].Objects {
+		deletedObjects[i] = DeletedObject{
+			ObjectName: requestList[0].Objects[i].ObjectName,
+		}
+	}
+
+	successResponse0 := generateMultiDeleteResponse(requestList[0].Quiet, deletedObjects, nil)
 	encodedSuccessResponse0 := encodeResponse(successResponse0)
 
 	successRequest1 := encodeResponse(requestList[1])
-	successResponse1 := generateMultiDeleteResponse(requestList[1].Quiet, requestList[1].Objects, nil)
+
+	deletedObjects = make([]DeletedObject, len(requestList[1].Objects))
+	for i := range requestList[0].Objects {
+		deletedObjects[i] = DeletedObject{
+			ObjectName: requestList[1].Objects[i].ObjectName,
+		}
+	}
+
+	successResponse1 := generateMultiDeleteResponse(requestList[1].Quiet, deletedObjects, nil)
 	encodedSuccessResponse1 := encodeResponse(successResponse1)
 
 	// generate multi objects delete response for errors.
 	// errorRequest := encodeResponse(requestList[1])
-	errorResponse := generateMultiDeleteResponse(requestList[1].Quiet, requestList[1].Objects, nil)
+	errorResponse := generateMultiDeleteResponse(requestList[1].Quiet, deletedObjects, nil)
 	encodedErrorResponse := encodeResponse(errorResponse)
 
 	anonRequest := encodeResponse(requestList[0])
@@ -817,6 +837,7 @@ func testAPIDeleteMultipleObjectsHandler(obj ObjectLayer, instanceType, bucketNa
 
 		// Verify whether the bucket obtained object is same as the one created.
 		if testCase.expectedContent != nil && !bytes.Equal(testCase.expectedContent, actualContent) {
+			fmt.Println(string(testCase.expectedContent), string(actualContent))
 			t.Errorf("Test %d : MinIO %s: Object content differs from expected value.", i+1, instanceType)
 		}
 	}

cmd/bucket-lifecycle.go

@@ -21,7 +21,6 @@ import (
 )
 
 const (
-
 	// Disabled means the lifecycle rule is inactive
 	Disabled = "Disabled"
 )

cmd/bucket-listobjects-handlers.go

@@ -49,13 +49,13 @@ func validateListObjectsArgs(marker, delimiter, encodingType string, maxKeys int
 	return ErrNone
 }
 
-// ListBucketObjectVersions - GET Bucket Object versions
+// ListObjectVersions - GET Bucket Object versions
 // You can use the versions subresource to list metadata about all
 // of the versions of objects in a bucket.
-func (api objectAPIHandlers) ListBucketObjectVersionsHandler(w http.ResponseWriter, r *http.Request) {
+func (api objectAPIHandlers) ListObjectVersionsHandler(w http.ResponseWriter, r *http.Request) {
-	ctx := newContext(r, w, "ListBucketObjectVersions")
+	ctx := newContext(r, w, "ListObjectVersions")
 
-	defer logger.AuditLog(w, r, "ListBucketObjectVersions", mustGetClaimsFromToken(r))
+	defer logger.AuditLog(w, r, "ListObjectVersions", mustGetClaimsFromToken(r))
 
 	vars := mux.Vars(r)
 	bucket := vars["bucket"]
@@ -74,8 +74,7 @@ func (api objectAPIHandlers) ListBucketObjectVersionsHandler(w http.ResponseWrit
 	urlValues := r.URL.Query()
 
 	// Extract all the listBucketVersions query params to their native values.
-	// versionIDMarker is ignored here.
+	prefix, marker, delimiter, maxkeys, encodingType, versionIDMarker, errCode := getListBucketObjectVersionsArgs(urlValues)
-	prefix, marker, delimiter, maxkeys, encodingType, _, errCode := getListBucketObjectVersionsArgs(urlValues)
 	if errCode != ErrNone {
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(errCode), r.URL, guessIsBrowserReq(r))
 		return
@@ -87,29 +86,29 @@ func (api objectAPIHandlers) ListBucketObjectVersionsHandler(w http.ResponseWrit
 		return
 	}
 
-	listObjects := objectAPI.ListObjects
+	listObjectVersions := objectAPI.ListObjectVersions
 
-	// Inititate a list objects operation based on the input params.
+	// Inititate a list object versions operation based on the input params.
 	// On success would return back ListObjectsInfo object to be
 	// marshaled into S3 compatible XML header.
-	listObjectsInfo, err := listObjects(ctx, bucket, prefix, marker, delimiter, maxkeys)
+	listObjectVersionsInfo, err := listObjectVersions(ctx, bucket, prefix, marker, versionIDMarker, delimiter, maxkeys)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
 	}
 
-	for i := range listObjectsInfo.Objects {
+	for i := range listObjectVersionsInfo.Objects {
-		if crypto.IsEncrypted(listObjectsInfo.Objects[i].UserDefined) {
+		if crypto.IsEncrypted(listObjectVersionsInfo.Objects[i].UserDefined) {
-			listObjectsInfo.Objects[i].ETag = getDecryptedETag(r.Header, listObjectsInfo.Objects[i], false)
+			listObjectVersionsInfo.Objects[i].ETag = getDecryptedETag(r.Header, listObjectVersionsInfo.Objects[i], false)
 		}
-		listObjectsInfo.Objects[i].Size, err = listObjectsInfo.Objects[i].GetActualSize()
+		listObjectVersionsInfo.Objects[i].Size, err = listObjectVersionsInfo.Objects[i].GetActualSize()
 		if err != nil {
 			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 			return
 		}
 	}
 
-	response := generateListVersionsResponse(bucket, prefix, marker, delimiter, encodingType, maxkeys, listObjectsInfo)
+	response := generateListVersionsResponse(bucket, prefix, marker, versionIDMarker, delimiter, encodingType, maxkeys, listObjectVersionsInfo)
 
 	// Write success response.
 	writeSuccessResponseXML(w, encodeResponse(response))

cmd/bucket-metadata-sys.go

@@ -28,6 +28,7 @@ import (
 	"github.com/minio/minio/pkg/bucket/lifecycle"
 	objectlock "github.com/minio/minio/pkg/bucket/object/lock"
 	"github.com/minio/minio/pkg/bucket/policy"
+	"github.com/minio/minio/pkg/bucket/versioning"
 	"github.com/minio/minio/pkg/event"
 	"github.com/minio/minio/pkg/madmin"
 	"github.com/minio/minio/pkg/sync/errgroup"
@@ -111,6 +112,8 @@ func (sys *BucketMetadataSys) Update(bucket string, configFile string, configDat
 		meta.TaggingConfigXML = configData
 	case objectLockConfig:
 		meta.ObjectLockConfigXML = configData
+	case bucketVersioningConfig:
+		meta.VersioningConfigXML = configData
 	case bucketQuotaConfigFile:
 		meta.QuotaConfigJSON = configData
 	default:
@@ -147,6 +150,16 @@ func (sys *BucketMetadataSys) Get(bucket string) (BucketMetadata, error) {
 	return meta, nil
 }
 
+// GetVersioningConfig returns configured versioning config
+// The returned object may not be modified.
+func (sys *BucketMetadataSys) GetVersioningConfig(bucket string) (*versioning.Versioning, error) {
+	meta, err := sys.GetConfig(bucket)
+	if err != nil {
+		return nil, err
+	}
+	return meta.versioningConfig, nil
+}
+
 // GetTaggingConfig returns configured tagging config
 // The returned object may not be modified.
 func (sys *BucketMetadataSys) GetTaggingConfig(bucket string) (*tags.Tags, error) {

cmd/bucket-metadata.go

@@ -32,6 +32,7 @@ import (
 	"github.com/minio/minio/pkg/bucket/lifecycle"
 	objectlock "github.com/minio/minio/pkg/bucket/object/lock"
 	"github.com/minio/minio/pkg/bucket/policy"
+	"github.com/minio/minio/pkg/bucket/versioning"
 	"github.com/minio/minio/pkg/event"
 	"github.com/minio/minio/pkg/madmin"
 )
@@ -47,6 +48,7 @@ const (
 
 var (
 	enabledBucketObjectLockConfig = []byte(`<ObjectLockConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/"><ObjectLockEnabled>Enabled</ObjectLockEnabled></ObjectLockConfiguration>`)
+	enabledBucketVersioningConfig = []byte(`<VersioningConfiguration xmlns="http://s3.amazonaws.com/doc/2006-03-01/"><Status>Enabled</Status></VersioningConfiguration>`)
 )
 
 //go:generate msgp -file $GOFILE
@@ -64,6 +66,7 @@ type BucketMetadata struct {
 	NotificationConfigXML []byte
 	LifecycleConfigXML    []byte
 	ObjectLockConfigXML   []byte
+	VersioningConfigXML   []byte
 	EncryptionConfigXML   []byte
 	TaggingConfigXML      []byte
 	QuotaConfigJSON       []byte
@@ -73,6 +76,7 @@ type BucketMetadata struct {
 	notificationConfig *event.Config
 	lifecycleConfig    *lifecycle.Lifecycle
 	objectLockConfig   *objectlock.Config
+	versioningConfig   *versioning.Versioning
 	sseConfig          *bucketsse.BucketSSEConfig
 	taggingConfig      *tags.Tags
 	quotaConfig        *madmin.BucketQuota
@@ -87,6 +91,9 @@ func newBucketMetadata(name string) BucketMetadata {
 			XMLNS: "http://s3.amazonaws.com/doc/2006-03-01/",
 		},
 		quotaConfig: &madmin.BucketQuota{},
+		versioningConfig: &versioning.Versioning{
+			XMLNS: "http://s3.amazonaws.com/doc/2006-03-01/",
+		},
 	}
 }
 
@@ -188,6 +195,13 @@ func (b *BucketMetadata) parseAllConfigs(ctx context.Context, objectAPI ObjectLa
 		b.objectLockConfig = nil
 	}
 
+	if len(b.VersioningConfigXML) != 0 {
+		b.versioningConfig, err = versioning.ParseConfig(bytes.NewReader(b.VersioningConfigXML))
+		if err != nil {
+			return err
+		}
+	}
+
 	if len(b.QuotaConfigJSON) != 0 {
 		b.quotaConfig, err = parseBucketQuota(b.Name, b.QuotaConfigJSON)
 		if err != nil {
@@ -244,6 +258,7 @@ func (b *BucketMetadata) convertLegacyConfigs(ctx context.Context, objectAPI Obj
 		case legacyBucketObjectLockEnabledConfigFile:
 			if string(configData) == legacyBucketObjectLockEnabledConfig {
 				b.ObjectLockConfigXML = enabledBucketObjectLockConfig
+				b.VersioningConfigXML = enabledBucketVersioningConfig
 				b.LockEnabled = false // legacy value unset it
 				// we are only interested in b.ObjectLockConfigXML
 			}
@@ -259,6 +274,7 @@ func (b *BucketMetadata) convertLegacyConfigs(ctx context.Context, objectAPI Obj
 			b.TaggingConfigXML = configData
 		case objectLockConfig:
 			b.ObjectLockConfigXML = configData
+			b.VersioningConfigXML = enabledBucketVersioningConfig
 		case bucketQuotaConfigFile:
 			b.QuotaConfigJSON = configData
 		}

cmd/bucket-metadata_gen.go

@@ -66,6 +66,12 @@ func (z *BucketMetadata) DecodeMsg(dc *msgp.Reader) (err error) {
 				err = msgp.WrapError(err, "ObjectLockConfigXML")
 				return
 			}
+		case "VersioningConfigXML":
+			z.VersioningConfigXML, err = dc.ReadBytes(z.VersioningConfigXML)
+			if err != nil {
+				err = msgp.WrapError(err, "VersioningConfigXML")
+				return
+			}
 		case "EncryptionConfigXML":
 			z.EncryptionConfigXML, err = dc.ReadBytes(z.EncryptionConfigXML)
 			if err != nil {
@@ -97,9 +103,9 @@ func (z *BucketMetadata) DecodeMsg(dc *msgp.Reader) (err error) {
 
 // EncodeMsg implements msgp.Encodable
 func (z *BucketMetadata) EncodeMsg(en *msgp.Writer) (err error) {
-	// map header, size 10
+	// map header, size 11
 	// write "Name"
-	err = en.Append(0x8a, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
+	err = en.Append(0x8b, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
 	if err != nil {
 		return
 	}
@@ -168,6 +174,16 @@ func (z *BucketMetadata) EncodeMsg(en *msgp.Writer) (err error) {
 		err = msgp.WrapError(err, "ObjectLockConfigXML")
 		return
 	}
+	// write "VersioningConfigXML"
+	err = en.Append(0xb3, 0x56, 0x65, 0x72, 0x73, 0x69, 0x6f, 0x6e, 0x69, 0x6e, 0x67, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x58, 0x4d, 0x4c)
+	if err != nil {
+		return
+	}
+	err = en.WriteBytes(z.VersioningConfigXML)
+	if err != nil {
+		err = msgp.WrapError(err, "VersioningConfigXML")
+		return
+	}
 	// write "EncryptionConfigXML"
 	err = en.Append(0xb3, 0x45, 0x6e, 0x63, 0x72, 0x79, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x58, 0x4d, 0x4c)
 	if err != nil {
@@ -204,9 +220,9 @@ func (z *BucketMetadata) EncodeMsg(en *msgp.Writer) (err error) {
 // MarshalMsg implements msgp.Marshaler
 func (z *BucketMetadata) MarshalMsg(b []byte) (o []byte, err error) {
 	o = msgp.Require(b, z.Msgsize())
-	// map header, size 10
+	// map header, size 11
 	// string "Name"
-	o = append(o, 0x8a, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
+	o = append(o, 0x8b, 0xa4, 0x4e, 0x61, 0x6d, 0x65)
 	o = msgp.AppendString(o, z.Name)
 	// string "Created"
 	o = append(o, 0xa7, 0x43, 0x72, 0x65, 0x61, 0x74, 0x65, 0x64)
@@ -226,6 +242,9 @@ func (z *BucketMetadata) MarshalMsg(b []byte) (o []byte, err error) {
 	// string "ObjectLockConfigXML"
 	o = append(o, 0xb3, 0x4f, 0x62, 0x6a, 0x65, 0x63, 0x74, 0x4c, 0x6f, 0x63, 0x6b, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x58, 0x4d, 0x4c)
 	o = msgp.AppendBytes(o, z.ObjectLockConfigXML)
+	// string "VersioningConfigXML"
+	o = append(o, 0xb3, 0x56, 0x65, 0x72, 0x73, 0x69, 0x6f, 0x6e, 0x69, 0x6e, 0x67, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x58, 0x4d, 0x4c)
+	o = msgp.AppendBytes(o, z.VersioningConfigXML)
 	// string "EncryptionConfigXML"
 	o = append(o, 0xb3, 0x45, 0x6e, 0x63, 0x72, 0x79, 0x70, 0x74, 0x69, 0x6f, 0x6e, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x58, 0x4d, 0x4c)
 	o = msgp.AppendBytes(o, z.EncryptionConfigXML)
@@ -298,6 +317,12 @@ func (z *BucketMetadata) UnmarshalMsg(bts []byte) (o []byte, err error) {
 				err = msgp.WrapError(err, "ObjectLockConfigXML")
 				return
 			}
+		case "VersioningConfigXML":
+			z.VersioningConfigXML, bts, err = msgp.ReadBytesBytes(bts, z.VersioningConfigXML)
+			if err != nil {
+				err = msgp.WrapError(err, "VersioningConfigXML")
+				return
+			}
 		case "EncryptionConfigXML":
 			z.EncryptionConfigXML, bts, err = msgp.ReadBytesBytes(bts, z.EncryptionConfigXML)
 			if err != nil {
@@ -330,6 +355,6 @@ func (z *BucketMetadata) UnmarshalMsg(bts []byte) (o []byte, err error) {
 
 // Msgsize returns an upper bound estimate of the number of bytes occupied by the serialized message
 func (z *BucketMetadata) Msgsize() (s int) {
-	s = 1 + 5 + msgp.StringPrefixSize + len(z.Name) + 8 + msgp.TimeSize + 12 + msgp.BoolSize + 17 + msgp.BytesPrefixSize + len(z.PolicyConfigJSON) + 22 + msgp.BytesPrefixSize + len(z.NotificationConfigXML) + 19 + msgp.BytesPrefixSize + len(z.LifecycleConfigXML) + 20 + msgp.BytesPrefixSize + len(z.ObjectLockConfigXML) + 20 + msgp.BytesPrefixSize + len(z.EncryptionConfigXML) + 17 + msgp.BytesPrefixSize + len(z.TaggingConfigXML) + 16 + msgp.BytesPrefixSize + len(z.QuotaConfigJSON)
+	s = 1 + 5 + msgp.StringPrefixSize + len(z.Name) + 8 + msgp.TimeSize + 12 + msgp.BoolSize + 17 + msgp.BytesPrefixSize + len(z.PolicyConfigJSON) + 22 + msgp.BytesPrefixSize + len(z.NotificationConfigXML) + 19 + msgp.BytesPrefixSize + len(z.LifecycleConfigXML) + 20 + msgp.BytesPrefixSize + len(z.ObjectLockConfigXML) + 20 + msgp.BytesPrefixSize + len(z.VersioningConfigXML) + 20 + msgp.BytesPrefixSize + len(z.EncryptionConfigXML) + 17 + msgp.BytesPrefixSize + len(z.TaggingConfigXML) + 16 + msgp.BytesPrefixSize + len(z.QuotaConfigJSON)
 	return
 }

cmd/bucket-object-lock.go

@@ -52,79 +52,6 @@ func (sys *BucketObjectLockSys) Get(bucketName string) (r objectlock.Retention,
 	return config.ToRetention(), nil
 }
 
-// Similar to enforceRetentionBypassForDelete but for WebUI
-func enforceRetentionBypassForDeleteWeb(ctx context.Context, r *http.Request, bucket, object string, getObjectInfoFn GetObjectInfoFn, govBypassPerms bool) APIErrorCode {
-	opts, err := getOpts(ctx, r, bucket, object)
-	if err != nil {
-		return toAPIErrorCode(ctx, err)
-	}
-
-	oi, err := getObjectInfoFn(ctx, bucket, object, opts)
-	if err != nil {
-		return toAPIErrorCode(ctx, err)
-	}
-
-	lhold := objectlock.GetObjectLegalHoldMeta(oi.UserDefined)
-	if lhold.Status.Valid() && lhold.Status == objectlock.LegalHoldOn {
-		return ErrObjectLocked
-	}
-
-	ret := objectlock.GetObjectRetentionMeta(oi.UserDefined)
-	if ret.Mode.Valid() {
-		switch ret.Mode {
-		case objectlock.RetCompliance:
-			// In compliance mode, a protected object version can't be overwritten
-			// or deleted by any user, including the root user in your AWS account.
-			// When an object is locked in compliance mode, its retention mode can't
-			// be changed, and its retention period can't be shortened. Compliance mode
-			// ensures that an object version can't be overwritten or deleted for the
-			// duration of the retention period.
-			t, err := objectlock.UTCNowNTP()
-			if err != nil {
-				logger.LogIf(ctx, err)
-				return ErrObjectLocked
-			}
-
-			if !ret.RetainUntilDate.Before(t) {
-				return ErrObjectLocked
-			}
-			return ErrNone
-		case objectlock.RetGovernance:
-			// In governance mode, users can't overwrite or delete an object
-			// version or alter its lock settings unless they have special
-			// permissions. With governance mode, you protect objects against
-			// being deleted by most users, but you can still grant some users
-			// permission to alter the retention settings or delete the object
-			// if necessary. You can also use governance mode to test retention-period
-			// settings before creating a compliance-mode retention period.
-			// To override or remove governance-mode retention settings, a
-			// user must have the s3:BypassGovernanceRetention permission
-			// and must explicitly include x-amz-bypass-governance-retention:true
-			// as a request header with any request that requires overriding
-			// governance mode.
-			byPassSet := govBypassPerms && objectlock.IsObjectLockGovernanceBypassSet(r.Header)
-			if !byPassSet {
-				t, err := objectlock.UTCNowNTP()
-				if err != nil {
-					logger.LogIf(ctx, err)
-					return ErrObjectLocked
-				}
-
-				if !ret.RetainUntilDate.Before(t) {
-					return ErrObjectLocked
-				}
-
-				if !govBypassPerms {
-					return ErrObjectLocked
-				}
-
-				return ErrNone
-			}
-		}
-	}
-	return ErrNone
-}
-
 // enforceRetentionForDeletion checks if it is appropriate to remove an
 // object according to locking configuration when this is lifecycle/ bucket quota asking.
 func enforceRetentionForDeletion(ctx context.Context, objInfo ObjectInfo) (locked bool) {
@@ -153,14 +80,23 @@ func enforceRetentionForDeletion(ctx context.Context, objInfo ObjectInfo) (locke
 // For objects in "Governance" mode, overwrite is allowed if a) object retention date is past OR
 // governance bypass headers are set and user has governance bypass permissions.
 // Objects in "Compliance" mode can be overwritten only if retention date is past.
-func enforceRetentionBypassForDelete(ctx context.Context, r *http.Request, bucket, object string, getObjectInfoFn GetObjectInfoFn) APIErrorCode {
+func enforceRetentionBypassForDelete(ctx context.Context, r *http.Request, bucket string, object ObjectToDelete, getObjectInfoFn GetObjectInfoFn) APIErrorCode {
-	opts, err := getOpts(ctx, r, bucket, object)
+	opts, err := getOpts(ctx, r, bucket, object.ObjectName)
 	if err != nil {
 		return toAPIErrorCode(ctx, err)
 	}
 
-	oi, err := getObjectInfoFn(ctx, bucket, object, opts)
+	opts.VersionID = object.VersionID
+
+	oi, err := getObjectInfoFn(ctx, bucket, object.ObjectName, opts)
 	if err != nil {
+		switch err.(type) {
+		case MethodNotAllowed: // This happens usually for a delete marker
+			if oi.DeleteMarker {
+				// Delete marker should be present and valid.
+				return ErrNone
+			}
+		}
 		return toAPIErrorCode(ctx, err)
 	}
 
@@ -219,8 +155,8 @@ func enforceRetentionBypassForDelete(ctx context.Context, r *http.Request, bucke
 			// https://docs.aws.amazon.com/AmazonS3/latest/dev/object-lock-overview.html#object-lock-retention-modes
 			// If you try to delete objects protected by governance mode and have s3:BypassGovernanceRetention
 			// or s3:GetBucketObjectLockConfiguration permissions, the operation will succeed.
-			govBypassPerms1 := checkRequestAuthType(ctx, r, policy.BypassGovernanceRetentionAction, bucket, object)
+			govBypassPerms1 := checkRequestAuthType(ctx, r, policy.BypassGovernanceRetentionAction, bucket, object.ObjectName)
-			govBypassPerms2 := checkRequestAuthType(ctx, r, policy.GetBucketObjectLockConfigurationAction, bucket, object)
+			govBypassPerms2 := checkRequestAuthType(ctx, r, policy.GetBucketObjectLockConfigurationAction, bucket, object.ObjectName)
 			if govBypassPerms1 != ErrNone && govBypassPerms2 != ErrNone {
 				return ErrAccessDenied
 			}
@@ -331,24 +267,25 @@ func checkPutObjectLockAllowed(ctx context.Context, r *http.Request, bucket, obj
 		return mode, retainDate, legalHold, ErrNone
 	}
 
-	var objExists bool
 	opts, err := getOpts(ctx, r, bucket, object)
 	if err != nil {
 		return mode, retainDate, legalHold, toAPIErrorCode(ctx, err)
 	}
 
+	if opts.VersionID != "" {
+		if objInfo, err := getObjectInfoFn(ctx, bucket, object, opts); err == nil {
+			r := objectlock.GetObjectRetentionMeta(objInfo.UserDefined)
+
 			t, err := objectlock.UTCNowNTP()
 			if err != nil {
 				logger.LogIf(ctx, err)
 				return mode, retainDate, legalHold, ErrObjectLocked
 			}
 
-	if objInfo, err := getObjectInfoFn(ctx, bucket, object, opts); err == nil {
-		objExists = true
-		r := objectlock.GetObjectRetentionMeta(objInfo.UserDefined)
 			if r.Mode == objectlock.RetCompliance && r.RetainUntilDate.After(t) {
 				return mode, retainDate, legalHold, ErrObjectLocked
 			}
+
 			mode = r.Mode
 			retainDate = r.RetainUntilDate
 			legalHold = objectlock.GetObjectLegalHoldMeta(objInfo.UserDefined)
@@ -357,6 +294,7 @@ func checkPutObjectLockAllowed(ctx context.Context, r *http.Request, bucket, obj
 				return mode, retainDate, legalHold, ErrObjectLocked
 			}
 		}
+	}
 
 	if legalHoldRequested {
 		var lerr error
@@ -374,9 +312,6 @@ func checkPutObjectLockAllowed(ctx context.Context, r *http.Request, bucket, obj
 		if err != nil {
 			return mode, retainDate, legalHold, toAPIErrorCode(ctx, err)
 		}
-		if objExists && retainDate.After(t) {
-			return mode, retainDate, legalHold, ErrObjectLocked
-		}
 		if retentionPermErr != ErrNone {
 			return mode, retainDate, legalHold, retentionPermErr
 		}
@@ -387,16 +322,14 @@ func checkPutObjectLockAllowed(ctx context.Context, r *http.Request, bucket, obj
 		if retentionPermErr != ErrNone {
 			return mode, retainDate, legalHold, retentionPermErr
 		}
+
 		t, err := objectlock.UTCNowNTP()
 		if err != nil {
 			logger.LogIf(ctx, err)
 			return mode, retainDate, legalHold, ErrObjectLocked
 		}
-		// AWS S3 just creates a new version of object when an object is being overwritten.
+
-		if objExists && retainDate.After(t) {
+		if !legalHoldRequested && retentionCfg.LockEnabled {
-			return mode, retainDate, legalHold, ErrObjectLocked
-		}
-		if !legalHoldRequested {
 			// inherit retention from bucket configuration
 			return retentionCfg.Mode, objectlock.RetentionDate{Time: t.Add(retentionCfg.Validity)}, legalHold, ErrNone
 		}

cmd/bucket-policy-handlers.go

@@ -164,7 +164,7 @@ func (api objectAPIHandlers) GetBucketPolicyHandler(w http.ResponseWriter, r *ht
 	}
 
 	// Read bucket access policy.
-	config, err := globalBucketMetadataSys.GetPolicyConfig(bucket)
+	config, err := globalPolicySys.Get(bucket)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return

cmd/bucket-policy-handlers_test.go

@@ -92,7 +92,7 @@ func getAnonWriteOnlyObjectPolicy(bucketName, prefix string) *policy.Policy {
 	}
 }
 
-// Wrapper for calling Put Bucket Policy HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling Put Bucket Policy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestPutBucketPolicyHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testPutBucketPolicyHandler, []string{"PutBucketPolicy"})
 }
@@ -102,7 +102,7 @@ func testPutBucketPolicyHandler(obj ObjectLayer, instanceType, bucketName string
 	credentials auth.Credentials, t *testing.T) {
 
 	bucketName1 := fmt.Sprintf("%s-1", bucketName)
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName1, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName1, BucketOptions{}); err != nil {
 		t.Fatal(err)
 	}
 
@@ -314,7 +314,7 @@ func testPutBucketPolicyHandler(obj ObjectLayer, instanceType, bucketName string
 
 }
 
-// Wrapper for calling Get Bucket Policy HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling Get Bucket Policy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestGetBucketPolicyHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testGetBucketPolicyHandler, []string{"PutBucketPolicy", "GetBucketPolicy"})
 }
@@ -520,7 +520,7 @@ func testGetBucketPolicyHandler(obj ObjectLayer, instanceType, bucketName string
 	ExecObjectLayerAPINilTest(t, nilBucket, "", instanceType, apiRouter, nilReq)
 }
 
-// Wrapper for calling Delete Bucket Policy HTTP handler tests for both XL multiple disks and single node setup.
+// Wrapper for calling Delete Bucket Policy HTTP handler tests for both Erasure multiple disks and single node setup.
 func TestDeleteBucketPolicyHandler(t *testing.T) {
 	ExecObjectLayerAPITest(t, testDeleteBucketPolicyHandler, []string{"PutBucketPolicy", "DeleteBucketPolicy"})
 }

cmd/bucket-policy.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2018 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2018,2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -70,6 +70,16 @@ func getConditionValues(r *http.Request, lc string, username string, claims map[
 		principalType = "User"
 	}
 
+	vid := r.URL.Query().Get("versionId")
+	if vid == "" {
+		if u, err := url.Parse(r.Header.Get(xhttp.AmzCopySource)); err == nil {
+			vid = u.Query().Get("versionId")
+		}
+		if vid == "" {
+			vid = r.Header.Get(xhttp.AmzCopySourceVersionID)
+		}
+	}
+
 	args := map[string][]string{
 		"CurrentTime":     {currTime.Format(time.RFC3339)},
 		"EpochTime":       {strconv.FormatInt(currTime.Unix(), 10)},
@@ -80,6 +90,7 @@ func getConditionValues(r *http.Request, lc string, username string, claims map[
 		"principaltype":   {principalType},
 		"userid":          {username},
 		"username":        {username},
+		"versionid":       {vid},
 	}
 
 	if lc != "" {
@@ -142,7 +153,7 @@ func getConditionValues(r *http.Request, lc string, username string, claims map[
 	return args
 }
 
-// PolicyToBucketAccessPolicy - converts policy.Policy to minio-go/policy.BucketAccessPolicy.
+// PolicyToBucketAccessPolicy converts a MinIO policy into a minio-go policy data structure.
 func PolicyToBucketAccessPolicy(bucketPolicy *policy.Policy) (*miniogopolicy.BucketAccessPolicy, error) {
 	// Return empty BucketAccessPolicy for empty bucket policy.
 	if bucketPolicy == nil {

cmd/bucket-quota.go

@@ -138,7 +138,7 @@ func startBucketQuotaEnforcement(ctx context.Context, objAPI ObjectLayer) {
 		case <-ctx.Done():
 			return
 		case <-time.NewTimer(bgQuotaInterval).C:
-			logger.LogIf(ctx, enforceFIFOQuota(ctx, objAPI))
+			enforceFIFOQuota(ctx, objAPI)
 		}
 
 	}
@@ -146,20 +146,22 @@ func startBucketQuotaEnforcement(ctx context.Context, objAPI ObjectLayer) {
 
 // enforceFIFOQuota deletes objects in FIFO order until sufficient objects
 // have been deleted so as to bring bucket usage within quota
-func enforceFIFOQuota(ctx context.Context, objectAPI ObjectLayer) error {
+func enforceFIFOQuota(ctx context.Context, objectAPI ObjectLayer) {
 	// Turn off quota enforcement if data usage info is unavailable.
 	if env.Get(envDataUsageCrawlConf, config.EnableOn) == config.EnableOff {
-		return nil
+		return
 	}
 
 	buckets, err := objectAPI.ListBuckets(ctx)
 	if err != nil {
-		return err
+		logger.LogIf(ctx, err)
+		return
 	}
 
 	dataUsageInfo, err := loadDataUsageFromBackend(ctx, objectAPI)
 	if err != nil {
-		return err
+		logger.LogIf(ctx, err)
+		return
 	}
 
 	for _, binfo := range buckets {
@@ -196,7 +198,8 @@ func enforceFIFOQuota(ctx context.Context, objectAPI ObjectLayer) error {
 
 		// Walk through all objects
 		if err := objectAPI.Walk(ctx, bucket, "", objInfoCh); err != nil {
-			return err
+			logger.LogIf(ctx, err)
+			continue
 		}
 
 		// reuse the fileScorer used by disk cache to score entries by
@@ -205,47 +208,57 @@ func enforceFIFOQuota(ctx context.Context, objectAPI ObjectLayer) error {
 		// irrelevant.
 		scorer, err := newFileScorer(toFree, time.Now().Unix(), 1)
 		if err != nil {
-			return err
+			logger.LogIf(ctx, err)
+			continue
 		}
 
 		rcfg, _ := globalBucketObjectLockSys.Get(bucket)
-
 		for obj := range objInfoCh {
+			if obj.DeleteMarker {
+				// Delete markers are automatically added for FIFO purge.
+				scorer.addFileWithObjInfo(obj, 1)
+				continue
+			}
 			// skip objects currently under retention
 			if rcfg.LockEnabled && enforceRetentionForDeletion(ctx, obj) {
 				continue
 			}
-			scorer.addFile(obj.Name, obj.ModTime, obj.Size, 1)
+			scorer.addFileWithObjInfo(obj, 1)
 		}
-		var objects []string
+
-		numKeys := len(scorer.fileNames())
+		versioned := globalBucketVersioningSys.Enabled(bucket)
-		for i, key := range scorer.fileNames() {
+
-			objects = append(objects, key)
+		var objects []ObjectToDelete
+		numKeys := len(scorer.fileObjInfos())
+		for i, obj := range scorer.fileObjInfos() {
+			objects = append(objects, ObjectToDelete{
+				ObjectName: obj.Name,
+				VersionID:  obj.VersionID,
+			})
 			if len(objects) < maxDeleteList && (i < numKeys-1) {
-				// skip deletion until maxObjectList or end of slice
+				// skip deletion until maxDeleteList or end of slice
 				continue
 			}
 
 			if len(objects) == 0 {
 				break
 			}
+
 			// Deletes a list of objects.
-			deleteErrs, err := objectAPI.DeleteObjects(ctx, bucket, objects)
+			_, deleteErrs := objectAPI.DeleteObjects(ctx, bucket, objects, ObjectOptions{
-			if err != nil {
+				Versioned: versioned,
-				logger.LogIf(ctx, err)
+			})
-			} else {
 			for i := range deleteErrs {
 				if deleteErrs[i] != nil {
 					logger.LogIf(ctx, deleteErrs[i])
 					continue
 				}
+
 				// Notify object deleted event.
 				sendEvent(eventArgs{
 					EventName:  event.ObjectRemovedDelete,
 					BucketName: bucket,
-						Object: ObjectInfo{
+					Object:     obj,
-							Name: objects[i],
-						},
 					Host:       "Internal: [FIFO-QUOTA-EXPIRY]",
 				})
 			}
@@ -253,5 +266,3 @@ func enforceFIFOQuota(ctx context.Context, objectAPI ObjectLayer) error {
 		}
 	}
 }
-	return nil
-}

cmd/bucket-versioning-handler.go

@@ -0,0 +1,128 @@
+/*
+ * MinIO Cloud Storage, (C) 2020 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"encoding/xml"
+	"io"
+	"net/http"
+
+	humanize "github.com/dustin/go-humanize"
+	"github.com/gorilla/mux"
+	xhttp "github.com/minio/minio/cmd/http"
+	"github.com/minio/minio/cmd/logger"
+	"github.com/minio/minio/pkg/bucket/policy"
+	"github.com/minio/minio/pkg/bucket/versioning"
+)
+
+const (
+	bucketVersioningConfig = "versioning.xml"
+
+	// Maximum size of bucket versioning configuration payload sent to the PutBucketVersioningHandler.
+	maxBucketVersioningConfigSize = 1 * humanize.MiByte
+)
+
+// PutBucketVersioningHandler - PUT Bucket Versioning.
+// ----------
+func (api objectAPIHandlers) PutBucketVersioningHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := newContext(r, w, "PutBucketVersioning")
+
+	defer logger.AuditLog(w, r, "PutBucketVersioning", mustGetClaimsFromToken(r))
+
+	vars := mux.Vars(r)
+	bucket := vars["bucket"]
+
+	objectAPI := api.ObjectAPI()
+	if objectAPI == nil {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	// PutBucketVersioning API requires Content-Md5
+	if _, ok := r.Header[xhttp.ContentMD5]; !ok {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrMissingContentMD5), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	if s3Error := checkRequestAuthType(ctx, r, policy.PutBucketVersioningAction, bucket, ""); s3Error != ErrNone {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	v, err := versioning.ParseConfig(io.LimitReader(r.Body, maxBucketVersioningConfigSize))
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	configData, err := xml.Marshal(v)
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	if err = globalBucketMetadataSys.Update(bucket, bucketVersioningConfig, configData); err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	writeSuccessResponseHeadersOnly(w)
+}
+
+// GetBucketVersioningHandler - GET Bucket Versioning.
+// ----------
+func (api objectAPIHandlers) GetBucketVersioningHandler(w http.ResponseWriter, r *http.Request) {
+	ctx := newContext(r, w, "GetBucketVersioning")
+
+	defer logger.AuditLog(w, r, "GetBucketVersioning", mustGetClaimsFromToken(r))
+
+	vars := mux.Vars(r)
+	bucket := vars["bucket"]
+
+	objectAPI := api.ObjectAPI()
+	if objectAPI == nil {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrServerNotInitialized), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	if s3Error := checkRequestAuthType(ctx, r, policy.GetBucketVersioningAction, bucket, ""); s3Error != ErrNone {
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	// Check if bucket exists.
+	if _, err := objectAPI.GetBucketInfo(ctx, bucket); err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	config, err := globalBucketVersioningSys.Get(bucket)
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	configData, err := xml.Marshal(config)
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	// Write bucket versioning configuration to client
+	writeSuccessResponseXML(w, configData)
+
+}

cmd/bucket-versioning.go

@@ -0,0 +1,57 @@
+/*
+ * MinIO Cloud Storage, (C) 2020 MinIO, Inc.
+ *
+ * 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 cmd
+
+import "github.com/minio/minio/pkg/bucket/versioning"
+
+// BucketVersioningSys - policy subsystem.
+type BucketVersioningSys struct{}
+
+// Enabled enabled versioning?
+func (sys *BucketVersioningSys) Enabled(bucket string) bool {
+	vc, err := globalBucketMetadataSys.GetVersioningConfig(bucket)
+	if err != nil {
+		return false
+	}
+	return vc.Enabled()
+}
+
+// Suspended suspended versioning?
+func (sys *BucketVersioningSys) Suspended(bucket string) bool {
+	vc, err := globalBucketMetadataSys.GetVersioningConfig(bucket)
+	if err != nil {
+		return false
+	}
+	return vc.Suspended()
+}
+
+// Get returns stored bucket policy
+func (sys *BucketVersioningSys) Get(bucket string) (*versioning.Versioning, error) {
+	if globalIsGateway {
+		objAPI := newObjectLayerFn()
+		if objAPI == nil {
+			return nil, errServerNotInitialized
+		}
+		return nil, NotImplemented{}
+	}
+	return globalBucketMetadataSys.GetVersioningConfig(bucket)
+}
+
+// NewBucketVersioningSys - creates new versioning system.
+func NewBucketVersioningSys() *BucketVersioningSys {
+	return &BucketVersioningSys{}
+}

cmd/config-common.go

@@ -50,7 +50,7 @@ func readConfig(ctx context.Context, objAPI ObjectLayer, configFile string) ([]b
 }
 
 func deleteConfig(ctx context.Context, objAPI ObjectLayer, configFile string) error {
-	err := objAPI.DeleteObject(ctx, minioMetaBucket, configFile)
+	_, err := objAPI.DeleteObject(ctx, minioMetaBucket, configFile, ObjectOptions{})
 	if err != nil && isErrObjectNotFound(err) {
 		return errConfigNotFound
 	}

cmd/config-current.go

@@ -59,7 +59,7 @@ func initHelp() {
 	for k, v := range notify.DefaultNotificationKVS {
 		kvs[k] = v
 	}
-	if globalIsXL {
+	if globalIsErasure {
 		kvs[config.StorageClassSubSys] = storageclass.DefaultKVS
 	}
 	config.RegisterDefaultKVS(kvs)
@@ -168,7 +168,7 @@ func initHelp() {
 		},
 	}
 
-	if globalIsXL {
+	if globalIsErasure {
 		helpSubSys = append(helpSubSys, config.HelpKV{})
 		copy(helpSubSys[2:], helpSubSys[1:])
 		helpSubSys[1] = config.HelpKV{
@@ -232,9 +232,9 @@ func validateConfig(s config.Config) error {
 		return err
 	}
 
-	if globalIsXL {
+	if globalIsErasure {
 		if _, err := storageclass.LookupConfig(s[config.StorageClassSubSys][config.Default],
-			globalXLSetDriveCount); err != nil {
+			globalErasureSetDriveCount); err != nil {
 			return err
 		}
 	}
@@ -367,9 +367,9 @@ func lookupConfigs(s config.Config) {
 
 	globalAPIConfig.init(apiConfig)
 
-	if globalIsXL {
+	if globalIsErasure {
 		globalStorageClass, err = storageclass.LookupConfig(s[config.StorageClassSubSys][config.Default],
-			globalXLSetDriveCount)
+			globalErasureSetDriveCount)
 		if err != nil {
 			logger.LogIf(ctx, fmt.Errorf("Unable to initialize storage class config: %w", err))
 		}

cmd/config.go

@@ -92,7 +92,8 @@ func listServerConfigHistory(ctx context.Context, objAPI ObjectLayer, withData b
 
 func delServerConfigHistory(ctx context.Context, objAPI ObjectLayer, uuidKV string) error {
 	historyFile := pathJoin(minioConfigHistoryPrefix, uuidKV+kvPrefix)
-	return objAPI.DeleteObject(ctx, minioMetaBucket, historyFile)
+	_, err := objAPI.DeleteObject(ctx, minioMetaBucket, historyFile, ObjectOptions{})
+	return err
 }
 
 func readServerConfigHistory(ctx context.Context, objAPI ObjectLayer, uuidKV string) ([]byte, error) {

cmd/consolelogger.go

@@ -45,7 +45,7 @@ func mustGetNodeName(endpointZones EndpointZones) (nodeName string) {
 	if err != nil {
 		logger.FatalIf(err, "Unable to start console logging subsystem")
 	}
-	if globalIsDistXL {
+	if globalIsDistErasure {
 		nodeName = host.Name
 	}
 	return nodeName

cmd/copy-part-range.go

@@ -32,7 +32,9 @@ func writeCopyPartErr(ctx context.Context, w http.ResponseWriter, err error, url
 		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrInvalidCopyPartRangeSource), url, browser)
 		return
 	default:
-		writeErrorResponse(ctx, w, toAPIError(ctx, err), url, browser)
+		apiErr := errorCodes.ToAPIErr(ErrInvalidCopyPartRangeSource)
+		apiErr.Description = err.Error()
+		writeErrorResponse(ctx, w, apiErr, url, browser)
 		return
 	}
 }

cmd/data-crawler.go

@@ -28,7 +28,6 @@ import (
 	"time"
 
 	"github.com/minio/minio/cmd/config"
-	xhttp "github.com/minio/minio/cmd/http"
 	"github.com/minio/minio/cmd/logger"
 	"github.com/minio/minio/pkg/bucket/lifecycle"
 	"github.com/minio/minio/pkg/color"
@@ -512,7 +511,6 @@ func (i *crawlItem) transformMetaDir() {
 type actionMeta struct {
 	oi      ObjectInfo
 	trustOI bool // Set true if oi can be trusted and has been read with quorum.
-	meta    map[string]string
 }
 
 // applyActions will apply lifecycle checks on to a scanned item.
@@ -528,7 +526,16 @@ func (i *crawlItem) applyActions(ctx context.Context, o ObjectLayer, meta action
 		return size
 	}
 
-	action := i.lifeCycle.ComputeAction(i.objectPath(), meta.meta[xhttp.AmzObjectTagging], meta.oi.ModTime)
+	versionID := meta.oi.VersionID
+	action := i.lifeCycle.ComputeAction(
+		lifecycle.ObjectOpts{
+			Name:         i.objectPath(),
+			UserTags:     meta.oi.UserTags,
+			ModTime:      meta.oi.ModTime,
+			VersionID:    meta.oi.VersionID,
+			DeleteMarker: meta.oi.DeleteMarker,
+			IsLatest:     meta.oi.IsLatest,
+		})
 	if i.debug {
 		logger.Info(color.Green("applyActions:")+" lifecycle: %q, Initial scan: %v", i.objectPath(), action)
 	}
@@ -542,19 +549,42 @@ func (i *crawlItem) applyActions(ctx context.Context, o ObjectLayer, meta action
 	// These (expensive) operations should only run on items we are likely to delete.
 	// Load to ensure that we have the correct version and not an unsynced version.
 	if !meta.trustOI {
-		obj, err := o.GetObjectInfo(ctx, i.bucket, i.objectPath(), ObjectOptions{})
+		obj, err := o.GetObjectInfo(ctx, i.bucket, i.objectPath(), ObjectOptions{
+			VersionID: versionID,
+		})
 		if err != nil {
+			switch err.(type) {
+			case MethodNotAllowed: // This happens usually for a delete marker
+				if !obj.DeleteMarker { // if this is not a delete marker log and return
 					// Do nothing - heal in the future.
 					logger.LogIf(ctx, err)
 					return size
 				}
+			case ObjectNotFound:
+				// object not found return 0
+				return 0
+			default:
+				// All other errors proceed.
+				logger.LogIf(ctx, err)
+				return size
+			}
+		}
 		size = obj.Size
 
 		// Recalculate action.
-		action = i.lifeCycle.ComputeAction(i.objectPath(), obj.UserTags, obj.ModTime)
+		action = i.lifeCycle.ComputeAction(
+			lifecycle.ObjectOpts{
+				Name:         i.objectPath(),
+				UserTags:     obj.UserTags,
+				ModTime:      obj.ModTime,
+				VersionID:    obj.VersionID,
+				DeleteMarker: obj.DeleteMarker,
+				IsLatest:     obj.IsLatest,
+			})
 		if i.debug {
 			logger.Info(color.Green("applyActions:")+" lifecycle: Secondary scan: %v", action)
 		}
+		versionID = obj.VersionID
 		switch action {
 		case lifecycle.DeleteAction:
 		default:
@@ -563,7 +593,7 @@ func (i *crawlItem) applyActions(ctx context.Context, o ObjectLayer, meta action
 		}
 	}
 
-	err = o.DeleteObject(ctx, i.bucket, i.objectPath())
+	obj, err := o.DeleteObject(ctx, i.bucket, i.objectPath(), ObjectOptions{VersionID: versionID})
 	if err != nil {
 		// Assume it is still there.
 		logger.LogIf(ctx, err)
@@ -574,9 +604,7 @@ func (i *crawlItem) applyActions(ctx context.Context, o ObjectLayer, meta action
 	sendEvent(eventArgs{
 		EventName:  event.ObjectRemovedDelete,
 		BucketName: i.bucket,
-		Object: ObjectInfo{
+		Object:     obj,
-			Name: i.objectPath(),
-		},
 		Host:       "Internal: [ILM-EXPIRY]",
 	})
 	return 0

cmd/disk-cache-backend.go

@@ -60,7 +60,7 @@ type CacheChecksumInfoV1 struct {
 // Represents the cache metadata struct
 type cacheMeta struct {
 	Version string   `json:"version"`
-	Stat    statInfo `json:"stat"` // Stat of the current object `cache.json`.
+	Stat    StatInfo `json:"stat"` // Stat of the current object `cache.json`.
 
 	// checksums of blocks on disk.
 	Checksum CacheChecksumInfoV1 `json:"checksum,omitempty"`
@@ -553,7 +553,7 @@ func (c *diskCache) bitrotWriteToCache(cachePath, fileName string, reader io.Rea
 	}
 	f, err := os.Create(filePath)
 	if err != nil {
-		return 0, osErrToFSFileErr(err)
+		return 0, osErrToFileErr(err)
 	}
 	defer f.Close()
 

cmd/disk-cache-utils.go

@@ -187,12 +187,12 @@ func readCacheFileStream(filePath string, offset, length int64) (io.ReadCloser,
 
 	fr, err := os.Open(filePath)
 	if err != nil {
-		return nil, osErrToFSFileErr(err)
+		return nil, osErrToFileErr(err)
 	}
 	// Stat to get the size of the file at path.
 	st, err := fr.Stat()
 	if err != nil {
-		err = osErrToFSFileErr(err)
+		err = osErrToFileErr(err)
 		return nil, err
 	}
 
@@ -299,6 +299,7 @@ type fileScorer struct {
 
 type queuedFile struct {
 	name      string
+	versionID string
 	size      uint64
 	score     float64
 }
@@ -321,15 +322,33 @@ func newFileScorer(saveBytes uint64, now int64, maxHits int) (*fileScorer, error
 	return &f, nil
 }
 
-func (f *fileScorer) addFile(name string, lastAccess time.Time, size int64, hits int) {
+func (f *fileScorer) addFile(name string, accTime time.Time, size int64, hits int) {
+	f.addFileWithObjInfo(ObjectInfo{
+		Name:    name,
+		AccTime: accTime,
+		Size:    size,
+	}, hits)
+}
+
+func (f *fileScorer) addFileWithObjInfo(objInfo ObjectInfo, hits int) {
 	// Calculate how much we want to delete this object.
 	file := queuedFile{
-		name: name,
+		name:      objInfo.Name,
-		size: uint64(size),
+		versionID: objInfo.VersionID,
+		size:      uint64(objInfo.Size),
 	}
-	score := float64(f.now - lastAccess.Unix())
+
+	var score float64
+	if objInfo.ModTime.IsZero() {
+		// Mod time is not available with disk cache use atime.
+		score = float64(f.now - objInfo.AccTime.Unix())
+	} else {
+		// if not used mod time when mod time is available.
+		score = float64(f.now - objInfo.ModTime.Unix())
+	}
+
 	// Size as fraction of how much we want to save, 0->1.
-	szWeight := math.Max(0, (math.Min(1, float64(size)*f.sizeMult)))
+	szWeight := math.Max(0, (math.Min(1, float64(file.size)*f.sizeMult)))
 	// 0 at f.maxHits, 1 at 0.
 	hitsWeight := (1.0 - math.Max(0, math.Min(1.0, float64(hits)/float64(f.maxHits))))
 	file.score = score * (1 + 0.25*szWeight + 0.25*hitsWeight)
@@ -404,6 +423,22 @@ func (f *fileScorer) trimQueue() {
 	}
 }
 
+// fileObjInfos returns all queued file object infos
+func (f *fileScorer) fileObjInfos() []ObjectInfo {
+	res := make([]ObjectInfo, 0, f.queue.Len())
+	e := f.queue.Front()
+	for e != nil {
+		qfile := e.Value.(queuedFile)
+		res = append(res, ObjectInfo{
+			Name:      qfile.name,
+			Size:      int64(qfile.size),
+			VersionID: qfile.versionID,
+		})
+		e = e.Next()
+	}
+	return res
+}
+
 // fileNames returns all queued file names.
 func (f *fileScorer) fileNames() []string {
 	res := make([]string, 0, f.queue.Len())

cmd/disk-cache.go

@@ -51,8 +51,8 @@ type CacheObjectLayer interface {
 	// Object operations.
 	GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error)
 	GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
-	DeleteObject(ctx context.Context, bucket, object string) error
+	DeleteObject(ctx context.Context, bucket, object string, opts ObjectOptions) (ObjectInfo, error)
-	DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error)
+	DeleteObjects(ctx context.Context, bucket string, objects []ObjectToDelete, opts ObjectOptions) ([]DeletedObject, []error)
 	PutObject(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error)
 	CopyObject(ctx context.Context, srcBucket, srcObject, destBucket, destObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (objInfo ObjectInfo, err error)
 	// Storage operations.
@@ -78,8 +78,7 @@ type cacheObjects struct {
 
 	GetObjectNInfoFn func(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error)
 	GetObjectInfoFn  func(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
-	DeleteObjectFn   func(ctx context.Context, bucket, object string) error
+	DeleteObjectFn   func(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
-	DeleteObjectsFn  func(ctx context.Context, bucket string, objects []string) ([]error, error)
 	PutObjectFn      func(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error)
 	CopyObjectFn     func(ctx context.Context, srcBucket, srcObject, destBucket, destObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (objInfo ObjectInfo, err error)
 }
@@ -120,8 +119,8 @@ func (c *cacheObjects) updateMetadataIfChanged(ctx context.Context, dcache *disk
 }
 
 // DeleteObject clears cache entry if backend delete operation succeeds
-func (c *cacheObjects) DeleteObject(ctx context.Context, bucket, object string) (err error) {
+func (c *cacheObjects) DeleteObject(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error) {
-	if err = c.DeleteObjectFn(ctx, bucket, object); err != nil {
+	if objInfo, err = c.DeleteObjectFn(ctx, bucket, object, opts); err != nil {
 		return
 	}
 	if c.isCacheExclude(bucket, object) || c.skipCache() {
@@ -130,19 +129,38 @@ func (c *cacheObjects) DeleteObject(ctx context.Context, bucket, object string)
 
 	dcache, cerr := c.getCacheLoc(bucket, object)
 	if cerr != nil {
-		return
+		return objInfo, cerr
 	}
 	dcache.Delete(ctx, bucket, object)
 	return
 }
 
 // DeleteObjects batch deletes objects in slice, and clears any cached entries
-func (c *cacheObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+func (c *cacheObjects) DeleteObjects(ctx context.Context, bucket string, objects []ObjectToDelete, opts ObjectOptions) ([]DeletedObject, []error) {
 	errs := make([]error, len(objects))
+	objInfos := make([]ObjectInfo, len(objects))
 	for idx, object := range objects {
-		errs[idx] = c.DeleteObject(ctx, bucket, object)
+		opts.VersionID = object.VersionID
+		objInfos[idx], errs[idx] = c.DeleteObject(ctx, bucket, object.ObjectName, opts)
 	}
-	return errs, nil
+	deletedObjects := make([]DeletedObject, len(objInfos))
+	for idx := range errs {
+		if errs[idx] != nil {
+			continue
+		}
+		if objInfos[idx].DeleteMarker {
+			deletedObjects[idx] = DeletedObject{
+				DeleteMarker:          objInfos[idx].DeleteMarker,
+				DeleteMarkerVersionID: objInfos[idx].VersionID,
+			}
+			continue
+		}
+		deletedObjects[idx] = DeletedObject{
+			ObjectName: objInfos[idx].Name,
+			VersionID:  objInfos[idx].VersionID,
+		}
+	}
+	return deletedObjects, errs
 }
 
 // construct a metadata k-v map
@@ -649,15 +667,8 @@ func newServerCacheObjects(ctx context.Context, config cache.Config) (CacheObjec
 		GetObjectNInfoFn: func(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error) {
 			return newObjectLayerFn().GetObjectNInfo(ctx, bucket, object, rs, h, lockType, opts)
 		},
-		DeleteObjectFn: func(ctx context.Context, bucket, object string) error {
+		DeleteObjectFn: func(ctx context.Context, bucket, object string, opts ObjectOptions) (ObjectInfo, error) {
-			return newObjectLayerFn().DeleteObject(ctx, bucket, object)
+			return newObjectLayerFn().DeleteObject(ctx, bucket, object, opts)
-		},
-		DeleteObjectsFn: func(ctx context.Context, bucket string, objects []string) ([]error, error) {
-			errs := make([]error, len(objects))
-			for idx, object := range objects {
-				errs[idx] = newObjectLayerFn().DeleteObject(ctx, bucket, object)
-			}
-			return errs, nil
 		},
 		PutObjectFn: func(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error) {
 			return newObjectLayerFn().PutObject(ctx, bucket, object, data, opts)

cmd/encryption-v1.go

@@ -31,6 +31,7 @@ import (
 	"strconv"
 	"strings"
 
+	"github.com/google/uuid"
 	"github.com/minio/minio-go/v6/pkg/encrypt"
 	"github.com/minio/minio/cmd/crypto"
 	"github.com/minio/minio/cmd/logger"
@@ -82,7 +83,7 @@ func isEncryptedMultipart(objInfo ObjectInfo) bool {
 		}
 	}
 	// Further check if this object is uploaded using multipart mechanism
-	// by the user and it is not about XL internally splitting the
+	// by the user and it is not about Erasure internally splitting the
 	// object into parts in PutObject()
 	return !(objInfo.backendType == BackendErasure && len(objInfo.ETag) == 32)
 }
@@ -859,6 +860,7 @@ func getDefaultOpts(header http.Header, copySource bool, metadata map[string]str
 	var clientKey [32]byte
 	var sse encrypt.ServerSide
 
+	opts = ObjectOptions{UserDefined: metadata}
 	if copySource {
 		if crypto.SSECopy.IsRequested(header) {
 			clientKey, err = crypto.SSECopy.ParseHTTP(header)
@@ -868,7 +870,8 @@ func getDefaultOpts(header http.Header, copySource bool, metadata map[string]str
 			if sse, err = encrypt.NewSSEC(clientKey[:]); err != nil {
 				return
 			}
-			return ObjectOptions{ServerSideEncryption: encrypt.SSECopy(sse), UserDefined: metadata}, nil
+			opts.ServerSideEncryption = encrypt.SSECopy(sse)
+			return
 		}
 		return
 	}
@@ -881,12 +884,13 @@ func getDefaultOpts(header http.Header, copySource bool, metadata map[string]str
 		if sse, err = encrypt.NewSSEC(clientKey[:]); err != nil {
 			return
 		}
-		return ObjectOptions{ServerSideEncryption: sse, UserDefined: metadata}, nil
+		opts.ServerSideEncryption = sse
+		return
 	}
 	if crypto.S3.IsRequested(header) || (metadata != nil && crypto.S3.IsEncrypted(metadata)) {
-		return ObjectOptions{ServerSideEncryption: encrypt.NewSSE(), UserDefined: metadata}, nil
+		opts.ServerSideEncryption = encrypt.NewSSE()
 	}
-	return ObjectOptions{UserDefined: metadata}, nil
+	return
 }
 
 // get ObjectOptions for GET calls from encryption headers
@@ -908,6 +912,19 @@ func getOpts(ctx context.Context, r *http.Request, bucket, object string) (Objec
 		}
 	}
 
+	vid := strings.TrimSpace(r.URL.Query().Get("versionId"))
+	if vid != "" && vid != nullVersionID {
+		_, err := uuid.Parse(vid)
+		if err != nil {
+			logger.LogIf(ctx, err)
+			return opts, VersionNotFound{
+				Bucket:    bucket,
+				Object:    object,
+				VersionID: vid,
+			}
+		}
+	}
+
 	if GlobalGatewaySSE.SSEC() && crypto.SSEC.IsRequested(r.Header) {
 		key, err := crypto.SSEC.ParseHTTP(r.Header)
 		if err != nil {
@@ -916,7 +933,11 @@ func getOpts(ctx context.Context, r *http.Request, bucket, object string) (Objec
 		derivedKey := deriveClientKey(key, bucket, object)
 		encryption, err = encrypt.NewSSEC(derivedKey[:])
 		logger.CriticalIf(ctx, err)
-		return ObjectOptions{ServerSideEncryption: encryption, PartNumber: partNumber}, nil
+		return ObjectOptions{
+			ServerSideEncryption: encryption,
+			VersionID:            vid,
+			PartNumber:           partNumber,
+		}, nil
 	}
 
 	// default case of passing encryption headers to backend
@@ -925,18 +946,21 @@ func getOpts(ctx context.Context, r *http.Request, bucket, object string) (Objec
 		return opts, err
 	}
 	opts.PartNumber = partNumber
+	opts.VersionID = vid
 	return opts, nil
 }
 
 // get ObjectOptions for PUT calls from encryption headers and metadata
 func putOpts(ctx context.Context, r *http.Request, bucket, object string, metadata map[string]string) (opts ObjectOptions, err error) {
+	versioned := globalBucketVersioningSys.Enabled(bucket)
 	// In the case of multipart custom format, the metadata needs to be checked in addition to header to see if it
 	// is SSE-S3 encrypted, primarily because S3 protocol does not require SSE-S3 headers in PutObjectPart calls
 	if GlobalGatewaySSE.SSES3() && (crypto.S3.IsRequested(r.Header) || crypto.S3.IsEncrypted(metadata)) {
-		return ObjectOptions{ServerSideEncryption: encrypt.NewSSE(), UserDefined: metadata}, nil
+		return ObjectOptions{ServerSideEncryption: encrypt.NewSSE(), UserDefined: metadata, Versioned: versioned}, nil
 	}
 	if GlobalGatewaySSE.SSEC() && crypto.SSEC.IsRequested(r.Header) {
 		opts, err = getOpts(ctx, r, bucket, object)
+		opts.Versioned = versioned
 		opts.UserDefined = metadata
 		return
 	}
@@ -949,10 +973,15 @@ func putOpts(ctx context.Context, r *http.Request, bucket, object string, metada
 		if err != nil {
 			return ObjectOptions{}, err
 		}
-		return ObjectOptions{ServerSideEncryption: sseKms, UserDefined: metadata}, nil
+		return ObjectOptions{ServerSideEncryption: sseKms, UserDefined: metadata, Versioned: versioned}, nil
 	}
 	// default case of passing encryption headers and UserDefined metadata to backend
-	return getDefaultOpts(r.Header, false, metadata)
+	opts, err = getDefaultOpts(r.Header, false, metadata)
+	if err != nil {
+		return opts, err
+	}
+	opts.Versioned = versioned
+	return opts, nil
 }
 
 // get ObjectOptions for Copy calls with encryption headers provided on the target side and source side metadata
@@ -981,5 +1010,9 @@ func copySrcOpts(ctx context.Context, r *http.Request, bucket, object string) (O
 	}
 
 	// default case of passing encryption headers to backend
-	return getDefaultOpts(r.Header, true, nil)
+	opts, err := getDefaultOpts(r.Header, false, nil)
+	if err != nil {
+		return opts, err
+	}
+	return opts, nil
 }

cmd/endpoint.go

@@ -547,9 +547,9 @@ func CreateEndpoints(serverAddr string, foundLocal bool, args ...[]string) (Endp
 		return endpoints, setupType, config.ErrInvalidErasureEndpoints(nil).Msg("invalid number of endpoints")
 	}
 
-	// Return XL setup when all endpoints are path style.
+	// Return Erasure setup when all endpoints are path style.
 	if endpoints[0].Type() == PathEndpointType {
-		setupType = XLSetupType
+		setupType = ErasureSetupType
 		return endpoints, setupType, nil
 	}
 
@@ -614,18 +614,18 @@ func CreateEndpoints(serverAddr string, foundLocal bool, args ...[]string) (Endp
 
 	// All endpoints are pointing to local host
 	if len(endpoints) == localEndpointCount {
-		// If all endpoints have same port number, Just treat it as distXL setup
+		// If all endpoints have same port number, Just treat it as distErasure setup
 		// using URL style endpoints.
 		if len(localPortSet) == 1 {
 			if len(localServerHostSet) > 1 {
 				return endpoints, setupType,
 					config.ErrInvalidErasureEndpoints(nil).Msg("all local endpoints should not have different hostnames/ips")
 			}
-			return endpoints, DistXLSetupType, nil
+			return endpoints, DistErasureSetupType, nil
 		}
 
 		// Even though all endpoints are local, but those endpoints use different ports.
-		// This means it is DistXL setup.
+		// This means it is DistErasure setup.
 	}
 
 	// Add missing port in all endpoints.
@@ -645,7 +645,7 @@ func CreateEndpoints(serverAddr string, foundLocal bool, args ...[]string) (Endp
 	}
 
 	// Error out if we have less than 2 unique servers.
-	if len(uniqueArgs.ToSlice()) < 2 && setupType == DistXLSetupType {
+	if len(uniqueArgs.ToSlice()) < 2 && setupType == DistErasureSetupType {
 		err := fmt.Errorf("Unsupported number of endpoints (%s), minimum number of servers cannot be less than 2 in distributed setup", endpoints)
 		return endpoints, setupType, err
 	}
@@ -655,7 +655,7 @@ func CreateEndpoints(serverAddr string, foundLocal bool, args ...[]string) (Endp
 		updateDomainIPs(uniqueArgs)
 	}
 
-	setupType = DistXLSetupType
+	setupType = DistErasureSetupType
 	return endpoints, setupType, nil
 }
 

cmd/endpoint_test.go

@@ -232,71 +232,71 @@ func TestCreateEndpoints(t *testing.T) {
 		{"localhost:10000", [][]string{{"/d1"}}, "localhost:10000", Endpoints{Endpoint{URL: &url.URL{Path: mustAbs("/d1")}, IsLocal: true}}, FSSetupType, nil},
 		{"localhost:9000", [][]string{{"https://127.0.0.1:9000/d1", "https://localhost:9001/d1", "https://example.com/d1", "https://example.com/d2"}}, "", Endpoints{}, -1, fmt.Errorf("path '/d1' can not be served by different port on same address")},
 
-		// XL Setup with PathEndpointType
+		// Erasure Setup with PathEndpointType
 		{":1234", [][]string{{"/d1", "/d2", "/d3", "/d4"}}, ":1234",
 			Endpoints{
 				Endpoint{URL: &url.URL{Path: mustAbs("/d1")}, IsLocal: true},
 				Endpoint{URL: &url.URL{Path: mustAbs("/d2")}, IsLocal: true},
 				Endpoint{URL: &url.URL{Path: mustAbs("/d3")}, IsLocal: true},
 				Endpoint{URL: &url.URL{Path: mustAbs("/d4")}, IsLocal: true},
-			}, XLSetupType, nil},
+			}, ErasureSetupType, nil},
-		// DistXL Setup with URLEndpointType
+		// DistErasure Setup with URLEndpointType
 		{":9000", [][]string{{"http://localhost/d1", "http://localhost/d2", "http://localhost/d3", "http://localhost/d4"}}, ":9000", Endpoints{
 			Endpoint{URL: &url.URL{Scheme: "http", Host: "localhost", Path: "/d1"}, IsLocal: true},
 			Endpoint{URL: &url.URL{Scheme: "http", Host: "localhost", Path: "/d2"}, IsLocal: true},
 			Endpoint{URL: &url.URL{Scheme: "http", Host: "localhost", Path: "/d3"}, IsLocal: true},
 			Endpoint{URL: &url.URL{Scheme: "http", Host: "localhost", Path: "/d4"}, IsLocal: true},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
-		// DistXL Setup with URLEndpointType having mixed naming to local host.
+		// DistErasure Setup with URLEndpointType having mixed naming to local host.
 		{"127.0.0.1:10000", [][]string{{"http://localhost/d1", "http://localhost/d2", "http://127.0.0.1/d3", "http://127.0.0.1/d4"}}, "", Endpoints{}, -1, fmt.Errorf("all local endpoints should not have different hostnames/ips")},
 
 		{":9001", [][]string{{"http://10.0.0.1:9000/export", "http://10.0.0.2:9000/export", "http://" + nonLoopBackIP + ":9001/export", "http://10.0.0.2:9001/export"}}, "", Endpoints{}, -1, fmt.Errorf("path '/export' can not be served by different port on same address")},
 
 		{":9000", [][]string{{"http://127.0.0.1:9000/export", "http://" + nonLoopBackIP + ":9000/export", "http://10.0.0.1:9000/export", "http://10.0.0.2:9000/export"}}, "", Endpoints{}, -1, fmt.Errorf("path '/export' cannot be served by different address on same server")},
 
-		// DistXL type
+		// DistErasure type
 		{"127.0.0.1:10000", [][]string{{case1Endpoint1, case1Endpoint2, "http://example.org/d3", "http://example.com/d4"}}, "127.0.0.1:10000", Endpoints{
 			Endpoint{URL: case1URLs[0], IsLocal: case1LocalFlags[0]},
 			Endpoint{URL: case1URLs[1], IsLocal: case1LocalFlags[1]},
 			Endpoint{URL: case1URLs[2], IsLocal: case1LocalFlags[2]},
 			Endpoint{URL: case1URLs[3], IsLocal: case1LocalFlags[3]},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
 
 		{"127.0.0.1:10000", [][]string{{case2Endpoint1, case2Endpoint2, "http://example.org/d3", "http://example.com/d4"}}, "127.0.0.1:10000", Endpoints{
 			Endpoint{URL: case2URLs[0], IsLocal: case2LocalFlags[0]},
 			Endpoint{URL: case2URLs[1], IsLocal: case2LocalFlags[1]},
 			Endpoint{URL: case2URLs[2], IsLocal: case2LocalFlags[2]},
 			Endpoint{URL: case2URLs[3], IsLocal: case2LocalFlags[3]},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
 
 		{":80", [][]string{{case3Endpoint1, "http://example.org:9000/d2", "http://example.com/d3", "http://example.net/d4"}}, ":80", Endpoints{
 			Endpoint{URL: case3URLs[0], IsLocal: case3LocalFlags[0]},
 			Endpoint{URL: case3URLs[1], IsLocal: case3LocalFlags[1]},
 			Endpoint{URL: case3URLs[2], IsLocal: case3LocalFlags[2]},
 			Endpoint{URL: case3URLs[3], IsLocal: case3LocalFlags[3]},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
 
 		{":9000", [][]string{{case4Endpoint1, "http://example.org/d2", "http://example.com/d3", "http://example.net/d4"}}, ":9000", Endpoints{
 			Endpoint{URL: case4URLs[0], IsLocal: case4LocalFlags[0]},
 			Endpoint{URL: case4URLs[1], IsLocal: case4LocalFlags[1]},
 			Endpoint{URL: case4URLs[2], IsLocal: case4LocalFlags[2]},
 			Endpoint{URL: case4URLs[3], IsLocal: case4LocalFlags[3]},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
 
 		{":9000", [][]string{{case5Endpoint1, case5Endpoint2, case5Endpoint3, case5Endpoint4}}, ":9000", Endpoints{
 			Endpoint{URL: case5URLs[0], IsLocal: case5LocalFlags[0]},
 			Endpoint{URL: case5URLs[1], IsLocal: case5LocalFlags[1]},
 			Endpoint{URL: case5URLs[2], IsLocal: case5LocalFlags[2]},
 			Endpoint{URL: case5URLs[3], IsLocal: case5LocalFlags[3]},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
 
-		// DistXL Setup using only local host.
+		// DistErasure Setup using only local host.
 		{":9003", [][]string{{"http://localhost:9000/d1", "http://localhost:9001/d2", "http://127.0.0.1:9002/d3", case6Endpoint}}, ":9003", Endpoints{
 			Endpoint{URL: case6URLs[0], IsLocal: case6LocalFlags[0]},
 			Endpoint{URL: case6URLs[1], IsLocal: case6LocalFlags[1]},
 			Endpoint{URL: case6URLs[2], IsLocal: case6LocalFlags[2]},
 			Endpoint{URL: case6URLs[3], IsLocal: case6LocalFlags[3]},
-		}, DistXLSetupType, nil},
+		}, DistErasureSetupType, nil},
 	}
 
 	for _, testCase := range testCases {

cmd/erasure-bucket.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -22,7 +22,6 @@ import (
 
 	"github.com/minio/minio-go/v6/pkg/s3utils"
 	"github.com/minio/minio/cmd/logger"
-
 	"github.com/minio/minio/pkg/sync/errgroup"
 )
 
@@ -35,13 +34,13 @@ var bucketMetadataOpIgnoredErrs = append(bucketOpIgnoredErrs, errVolumeNotFound)
 /// Bucket operations
 
 // MakeBucket - make a bucket.
-func (xl xlObjects) MakeBucketWithLocation(ctx context.Context, bucket, location string, lockEnabled bool) error {
+func (er erasureObjects) MakeBucketWithLocation(ctx context.Context, bucket string, opts BucketOptions) error {
 	// Verify if bucket is valid.
 	if err := s3utils.CheckValidBucketNameStrict(bucket); err != nil {
 		return BucketNameInvalid{Bucket: bucket}
 	}
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
 
 	g := errgroup.WithNErrs(len(storageDisks))
 
@@ -86,9 +85,9 @@ func undoDeleteBucket(storageDisks []StorageAPI, bucket string) {
 }
 
 // getBucketInfo - returns the BucketInfo from one of the load balanced disks.
-func (xl xlObjects) getBucketInfo(ctx context.Context, bucketName string) (bucketInfo BucketInfo, err error) {
+func (er erasureObjects) getBucketInfo(ctx context.Context, bucketName string) (bucketInfo BucketInfo, err error) {
 	var bucketErrs []error
-	for _, disk := range xl.getLoadBalancedDisks() {
+	for _, disk := range er.getLoadBalancedDisks() {
 		if disk == nil {
 			bucketErrs = append(bucketErrs, errDiskNotFound)
 			continue
@@ -110,13 +109,13 @@ func (xl xlObjects) getBucketInfo(ctx context.Context, bucketName string) (bucke
 	// reduce to one error based on read quorum.
 	// `nil` is deliberately passed for ignoredErrs
 	// because these errors were already ignored.
-	readQuorum := getReadQuorum(len(xl.getDisks()))
+	readQuorum := getReadQuorum(len(er.getDisks()))
 	return BucketInfo{}, reduceReadQuorumErrs(ctx, bucketErrs, nil, readQuorum)
 }
 
 // GetBucketInfo - returns BucketInfo for a bucket.
-func (xl xlObjects) GetBucketInfo(ctx context.Context, bucket string) (bi BucketInfo, e error) {
+func (er erasureObjects) GetBucketInfo(ctx context.Context, bucket string) (bi BucketInfo, e error) {
-	bucketInfo, err := xl.getBucketInfo(ctx, bucket)
+	bucketInfo, err := er.getBucketInfo(ctx, bucket)
 	if err != nil {
 		return bi, toObjectErr(err, bucket)
 	}
@@ -124,8 +123,8 @@ func (xl xlObjects) GetBucketInfo(ctx context.Context, bucket string) (bi Bucket
 }
 
 // listBuckets - returns list of all buckets from a disk picked at random.
-func (xl xlObjects) listBuckets(ctx context.Context) (bucketsInfo []BucketInfo, err error) {
+func (er erasureObjects) listBuckets(ctx context.Context) (bucketsInfo []BucketInfo, err error) {
-	for _, disk := range xl.getLoadBalancedDisks() {
+	for _, disk := range er.getLoadBalancedDisks() {
 		if disk == nil {
 			continue
 		}
@@ -161,8 +160,8 @@ func (xl xlObjects) listBuckets(ctx context.Context) (bucketsInfo []BucketInfo,
 }
 
 // ListBuckets - lists all the buckets, sorted by its name.
-func (xl xlObjects) ListBuckets(ctx context.Context) ([]BucketInfo, error) {
+func (er erasureObjects) ListBuckets(ctx context.Context) ([]BucketInfo, error) {
-	bucketInfos, err := xl.listBuckets(ctx)
+	bucketInfos, err := er.listBuckets(ctx)
 	if err != nil {
 		return nil, toObjectErr(err)
 	}
@@ -196,9 +195,9 @@ func deleteDanglingBucket(ctx context.Context, storageDisks []StorageAPI, dErrs
 }
 
 // DeleteBucket - deletes a bucket.
-func (xl xlObjects) DeleteBucket(ctx context.Context, bucket string, forceDelete bool) error {
+func (er erasureObjects) DeleteBucket(ctx context.Context, bucket string, forceDelete bool) error {
 	// Collect if all disks report volume not found.
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
 
 	g := errgroup.WithNErrs(len(storageDisks))
 
@@ -235,7 +234,7 @@ func (xl xlObjects) DeleteBucket(ctx context.Context, bucket string, forceDelete
 
 	writeQuorum := getWriteQuorum(len(storageDisks))
 	err := reduceWriteQuorumErrs(ctx, dErrs, bucketOpIgnoredErrs, writeQuorum)
-	if err == errXLWriteQuorum {
+	if err == errErasureWriteQuorum {
 		undoDeleteBucket(storageDisks, bucket)
 	}
 	if err != nil {
@@ -251,25 +250,26 @@ func (xl xlObjects) DeleteBucket(ctx context.Context, bucket string, forceDelete
 }
 
 // IsNotificationSupported returns whether bucket notification is applicable for this layer.
-func (xl xlObjects) IsNotificationSupported() bool {
+func (er erasureObjects) IsNotificationSupported() bool {
 	return true
 }
 
 // IsListenBucketSupported returns whether listen bucket notification is applicable for this layer.
-func (xl xlObjects) IsListenBucketSupported() bool {
+func (er erasureObjects) IsListenBucketSupported() bool {
 	return true
 }
 
 // IsEncryptionSupported returns whether server side encryption is implemented for this layer.
-func (xl xlObjects) IsEncryptionSupported() bool {
+func (er erasureObjects) IsEncryptionSupported() bool {
 	return true
 }
 
 // IsCompressionSupported returns whether compression is applicable for this layer.
-func (xl xlObjects) IsCompressionSupported() bool {
+func (er erasureObjects) IsCompressionSupported() bool {
 	return true
 }
 
-func (xl xlObjects) IsTaggingSupported() bool {
+// IsTaggingSupported indicates whether erasureObjects implements tagging support.
+func (er erasureObjects) IsTaggingSupported() bool {
 	return true
 }

cmd/erasure-coding.go

@@ -0,0 +1,143 @@
+/*
+ * MinIO Cloud Storage, (C) 2017 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"context"
+	"sync"
+
+	"github.com/klauspost/reedsolomon"
+	"github.com/minio/minio/cmd/logger"
+)
+
+// Erasure - erasure encoding details.
+type Erasure struct {
+	encoder                  func() reedsolomon.Encoder
+	dataBlocks, parityBlocks int
+	blockSize                int64
+}
+
+// NewErasure creates a new ErasureStorage.
+func NewErasure(ctx context.Context, dataBlocks, parityBlocks int, blockSize int64) (e Erasure, err error) {
+	// Check the parameters for sanity now.
+	if dataBlocks <= 0 || parityBlocks <= 0 {
+		return e, reedsolomon.ErrInvShardNum
+	}
+
+	if dataBlocks+parityBlocks > 256 {
+		return e, reedsolomon.ErrMaxShardNum
+	}
+
+	e = Erasure{
+		dataBlocks:   dataBlocks,
+		parityBlocks: parityBlocks,
+		blockSize:    blockSize,
+	}
+
+	// Encoder when needed.
+	var enc reedsolomon.Encoder
+	var once sync.Once
+	e.encoder = func() reedsolomon.Encoder {
+		once.Do(func() {
+			e, err := reedsolomon.New(dataBlocks, parityBlocks, reedsolomon.WithAutoGoroutines(int(e.ShardSize())))
+			if err != nil {
+				// Error conditions should be checked above.
+				panic(err)
+			}
+			enc = e
+		})
+		return enc
+	}
+	return
+}
+
+// EncodeData encodes the given data and returns the erasure-coded data.
+// It returns an error if the erasure coding failed.
+func (e *Erasure) EncodeData(ctx context.Context, data []byte) ([][]byte, error) {
+	if len(data) == 0 {
+		return make([][]byte, e.dataBlocks+e.parityBlocks), nil
+	}
+	encoded, err := e.encoder().Split(data)
+	if err != nil {
+		logger.LogIf(ctx, err)
+		return nil, err
+	}
+	if err = e.encoder().Encode(encoded); err != nil {
+		logger.LogIf(ctx, err)
+		return nil, err
+	}
+	return encoded, nil
+}
+
+// DecodeDataBlocks decodes the given erasure-coded data.
+// It only decodes the data blocks but does not verify them.
+// It returns an error if the decoding failed.
+func (e *Erasure) DecodeDataBlocks(data [][]byte) error {
+	var isZero = 0
+	for _, b := range data[:] {
+		if len(b) == 0 {
+			isZero++
+			break
+		}
+	}
+	if isZero == 0 || isZero == len(data) {
+		// If all are zero, payload is 0 bytes.
+		return nil
+	}
+	return e.encoder().ReconstructData(data)
+}
+
+// DecodeDataAndParityBlocks decodes the given erasure-coded data and verifies it.
+// It returns an error if the decoding failed.
+func (e *Erasure) DecodeDataAndParityBlocks(ctx context.Context, data [][]byte) error {
+	if err := e.encoder().Reconstruct(data); err != nil {
+		logger.LogIf(ctx, err)
+		return err
+	}
+	return nil
+}
+
+// ShardSize - returns actual shared size from erasure blockSize.
+func (e *Erasure) ShardSize() int64 {
+	return ceilFrac(e.blockSize, int64(e.dataBlocks))
+}
+
+// ShardFileSize - returns final erasure size from original size.
+func (e *Erasure) ShardFileSize(totalLength int64) int64 {
+	if totalLength == 0 {
+		return 0
+	}
+	if totalLength == -1 {
+		return -1
+	}
+	numShards := totalLength / e.blockSize
+	lastBlockSize := totalLength % int64(e.blockSize)
+	lastShardSize := ceilFrac(lastBlockSize, int64(e.dataBlocks))
+	return numShards*e.ShardSize() + lastShardSize
+}
+
+// ShardFileOffset - returns the effective offset where erasure reading begins.
+func (e *Erasure) ShardFileOffset(startOffset, length, totalLength int64) int64 {
+	shardSize := e.ShardSize()
+	shardFileSize := e.ShardFileSize(totalLength)
+	endShard := (startOffset + int64(length)) / e.blockSize
+	tillOffset := endShard*shardSize + shardSize
+	if tillOffset > shardFileSize {
+		tillOffset = shardFileSize
+	}
+	return tillOffset
+}

cmd/erasure-common.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -24,8 +24,8 @@ import (
 )
 
 // getLoadBalancedDisks - fetches load balanced (sufficiently randomized) disk slice.
-func (xl xlObjects) getLoadBalancedDisks() (newDisks []StorageAPI) {
+func (er erasureObjects) getLoadBalancedDisks() (newDisks []StorageAPI) {
-	disks := xl.getDisks()
+	disks := er.getDisks()
 	// Based on the random shuffling return back randomized disks.
 	for _, i := range hashOrder(UTCNow().String(), len(disks)) {
 		newDisks = append(newDisks, disks[i-1])
@@ -36,13 +36,13 @@ func (xl xlObjects) getLoadBalancedDisks() (newDisks []StorageAPI) {
 // This function does the following check, suppose
 // object is "a/b/c/d", stat makes sure that objects ""a/b/c""
 // "a/b" and "a" do not exist.
-func (xl xlObjects) parentDirIsObject(ctx context.Context, bucket, parent string) bool {
+func (er erasureObjects) parentDirIsObject(ctx context.Context, bucket, parent string) bool {
 	var isParentDirObject func(string) bool
 	isParentDirObject = func(p string) bool {
 		if p == "." || p == SlashSeparator {
 			return false
 		}
-		if xl.isObject(bucket, p) {
+		if er.isObject(ctx, bucket, p) {
 			// If there is already a file at prefix "p", return true.
 			return true
 		}
@@ -53,9 +53,9 @@ func (xl xlObjects) parentDirIsObject(ctx context.Context, bucket, parent string
 }
 
 // isObject - returns `true` if the prefix is an object i.e if
-// `xl.json` exists at the leaf, false otherwise.
+// `xl.meta` exists at the leaf, false otherwise.
-func (xl xlObjects) isObject(bucket, prefix string) (ok bool) {
+func (er erasureObjects) isObject(ctx context.Context, bucket, prefix string) (ok bool) {
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
 
 	g := errgroup.WithNErrs(len(storageDisks))
 
@@ -66,22 +66,15 @@ func (xl xlObjects) isObject(bucket, prefix string) (ok bool) {
 				return errDiskNotFound
 			}
 			// Check if 'prefix' is an object on this 'disk', else continue the check the next disk
-			fi, err := storageDisks[index].StatFile(bucket, pathJoin(prefix, xlMetaJSONFile))
+			return storageDisks[index].CheckFile(bucket, prefix)
-			if err != nil {
-				return err
-			}
-			if fi.Size == 0 {
-				return errCorruptedFormat
-			}
-			return nil
 		}, index)
 	}
 
-	// NOTE: Observe we are not trying to read `xl.json` and figure out the actual
+	// NOTE: Observe we are not trying to read `xl.meta` and figure out the actual
 	// quorum intentionally, but rely on the default case scenario. Actual quorum
 	// verification will happen by top layer by using getObjectInfo() and will be
 	// ignored if necessary.
 	readQuorum := getReadQuorum(len(storageDisks))
 
-	return reduceReadQuorumErrs(GlobalContext, g.Wait(), objectOpIgnoredErrs, readQuorum) == nil
+	return reduceReadQuorumErrs(ctx, g.Wait(), objectOpIgnoredErrs, readQuorum) == nil
 }

cmd/erasure-common_test.go

@@ -24,13 +24,13 @@ import (
 )
 
 // Tests for if parent directory is object
-func TestXLParentDirIsObject(t *testing.T) {
+func TestErasureParentDirIsObject(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	obj, fsDisks, err := prepareXL16(ctx)
+	obj, fsDisks, err := prepareErasure16(ctx)
 	if err != nil {
-		t.Fatalf("Unable to initialize 'XL' object layer.")
+		t.Fatalf("Unable to initialize 'Erasure' object layer.")
 	}
 
 	// Remove all disks.
@@ -41,7 +41,7 @@ func TestXLParentDirIsObject(t *testing.T) {
 	bucketName := "testbucket"
 	objectName := "object"
 
-	if err = obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err = obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal(err)
 	}
 	objectContent := "12345"
@@ -54,7 +54,7 @@ func TestXLParentDirIsObject(t *testing.T) {
 		t.Fatalf("Unexpected object name returned got %s, expected %s", objInfo.Name, objectName)
 	}
 
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
 	xl := z.zones[0].sets[0]
 	testCases := []struct {
 		parentIsObject bool

cmd/erasure-decode.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -191,7 +191,7 @@ func (p *parallelReader) Read(dst [][]byte) ([][]byte, error) {
 		return newBuf, nil
 	}
 
-	return nil, errXLReadQuorum
+	return nil, errErasureReadQuorum
 }
 
 type errDecodeHealRequired struct {

cmd/erasure-decode_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -132,7 +132,7 @@ func TestErasureDecode(t *testing.T) {
 			if disk == OfflineDisk {
 				continue
 			}
-			tillOffset := erasure.ShardFileTillOffset(test.offset, test.length, test.data)
+			tillOffset := erasure.ShardFileOffset(test.offset, test.length, test.data)
 
 			bitrotReaders[index] = newBitrotReader(disk, "testbucket", "object", tillOffset, writeAlgorithm, bitrotWriterSum(writers[index]), erasure.ShardSize())
 		}
@@ -163,7 +163,7 @@ func TestErasureDecode(t *testing.T) {
 				if disk == OfflineDisk {
 					continue
 				}
-				tillOffset := erasure.ShardFileTillOffset(test.offset, test.length, test.data)
+				tillOffset := erasure.ShardFileOffset(test.offset, test.length, test.data)
 				bitrotReaders[index] = newBitrotReader(disk, "testbucket", "object", tillOffset, writeAlgorithm, bitrotWriterSum(writers[index]), erasure.ShardSize())
 			}
 			for j := range disks[:test.offDisks] {
@@ -268,7 +268,7 @@ func TestErasureDecodeRandomOffsetLength(t *testing.T) {
 			if disk == OfflineDisk {
 				continue
 			}
-			tillOffset := erasure.ShardFileTillOffset(offset, readLen, length)
+			tillOffset := erasure.ShardFileOffset(offset, readLen, length)
 			bitrotReaders[index] = newStreamingBitrotReader(disk, "testbucket", "object", tillOffset, DefaultBitrotAlgorithm, erasure.ShardSize())
 		}
 		err = erasure.Decode(context.Background(), buf, bitrotReaders, offset, readLen, length, nil)
@@ -330,7 +330,7 @@ func benchmarkErasureDecode(data, parity, dataDown, parityDown int, size int64,
 			if writers[index] == nil {
 				continue
 			}
-			tillOffset := erasure.ShardFileTillOffset(0, size, size)
+			tillOffset := erasure.ShardFileOffset(0, size, size)
 			bitrotReaders[index] = newStreamingBitrotReader(disk, "testbucket", "object", tillOffset, DefaultBitrotAlgorithm, erasure.ShardSize())
 		}
 		if err = erasure.Decode(context.Background(), bytes.NewBuffer(content[:0]), bitrotReaders, 0, size, size, nil); err != nil {

cmd/erasure-encode.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

cmd/erasure-encode_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

cmd/erasure-errors.go

@@ -18,11 +18,11 @@ package cmd
 
 import "errors"
 
-// errXLReadQuorum - did not meet read quorum.
+// errErasureReadQuorum - did not meet read quorum.
-var errXLReadQuorum = errors.New("Read failed. Insufficient number of disks online")
+var errErasureReadQuorum = errors.New("Read failed. Insufficient number of disks online")
 
-// errXLWriteQuorum - did not meet write quorum.
+// errErasureWriteQuorum - did not meet write quorum.
-var errXLWriteQuorum = errors.New("Write failed. Insufficient number of disks online")
+var errErasureWriteQuorum = errors.New("Write failed. Insufficient number of disks online")
 
 // errNoHealRequired - returned when healing is attempted on a previously healed disks.
 var errNoHealRequired = errors.New("No healing is required")

cmd/erasure-heal_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -71,7 +71,7 @@ func TestErasureHeal(t *testing.T) {
 		// create some test data
 		setup, err := newErasureTestSetup(test.dataBlocks, test.disks-test.dataBlocks, test.blocksize)
 		if err != nil {
-			t.Fatalf("Test %d: failed to setup XL environment: %v", i, err)
+			t.Fatalf("Test %d: failed to setup Erasure environment: %v", i, err)
 		}
 		disks := setup.disks
 		erasure, err := NewErasure(context.Background(), test.dataBlocks, test.disks-test.dataBlocks, test.blocksize)

cmd/erasure-healing-common.go

@@ -18,10 +18,8 @@ package cmd
 
 import (
 	"context"
-	"fmt"
 	"time"
 
-	"github.com/minio/minio/cmd/logger"
 	"github.com/minio/minio/pkg/madmin"
 )
 
@@ -31,7 +29,7 @@ func commonTime(modTimes []time.Time) (modTime time.Time, count int) {
 	timeOccurenceMap := make(map[time.Time]int)
 	// Ignore the uuid sentinel and count the rest.
 	for _, time := range modTimes {
-		if time == timeSentinel {
+		if time.Equal(timeSentinel) {
 			continue
 		}
 		timeOccurenceMap[time]++
@@ -61,45 +59,45 @@ func bootModtimes(diskCount int) []time.Time {
 	return modTimes
 }
 
-// Extracts list of times from xlMetaV1 slice and returns, skips
+// Extracts list of times from FileInfo slice and returns, skips
 // slice elements which have errors.
-func listObjectModtimes(partsMetadata []xlMetaV1, errs []error) (modTimes []time.Time) {
+func listObjectModtimes(partsMetadata []FileInfo, errs []error) (modTimes []time.Time) {
 	modTimes = bootModtimes(len(partsMetadata))
 	for index, metadata := range partsMetadata {
 		if errs[index] != nil {
 			continue
 		}
 		// Once the file is found, save the uuid saved on disk.
-		modTimes[index] = metadata.Stat.ModTime
+		modTimes[index] = metadata.ModTime
 	}
 	return modTimes
 }
 
 // Notes:
 // There are 5 possible states a disk could be in,
-// 1. __online__             - has the latest copy of xl.json - returned by listOnlineDisks
+// 1. __online__             - has the latest copy of xl.meta - returned by listOnlineDisks
 //
 // 2. __offline__            - err == errDiskNotFound
 //
-// 3. __availableWithParts__ - has the latest copy of xl.json and has all
+// 3. __availableWithParts__ - has the latest copy of xl.meta and has all
 //                             parts with checksums matching; returned by disksWithAllParts
 //
 // 4. __outdated__           - returned by outDatedDisk, provided []StorageAPI
 //                             returned by diskWithAllParts is passed for latestDisks.
-//    - has an old copy of xl.json
+//    - has an old copy of xl.meta
-//    - doesn't have xl.json (errFileNotFound)
+//    - doesn't have xl.meta (errFileNotFound)
-//    - has the latest xl.json but one or more parts are corrupt
+//    - has the latest xl.meta but one or more parts are corrupt
 //
-// 5. __missingParts__       - has the latest copy of xl.json but has some parts
+// 5. __missingParts__       - has the latest copy of xl.meta but has some parts
 // missing.  This is identified separately since this may need manual
 // inspection to understand the root cause. E.g, this could be due to
 // backend filesystem corruption.
 
 // listOnlineDisks - returns
-// - a slice of disks where disk having 'older' xl.json (or nothing)
+// - a slice of disks where disk having 'older' xl.meta (or nothing)
 // are set to nil.
 // - latest (in time) of the maximally occurring modTime(s).
-func listOnlineDisks(disks []StorageAPI, partsMetadata []xlMetaV1, errs []error) (onlineDisks []StorageAPI, modTime time.Time) {
+func listOnlineDisks(disks []StorageAPI, partsMetadata []FileInfo, errs []error) (onlineDisks []StorageAPI, modTime time.Time) {
 	onlineDisks = make([]StorageAPI, len(disks))
 
 	// List all the file commit ids from parts metadata.
@@ -110,7 +108,7 @@ func listOnlineDisks(disks []StorageAPI, partsMetadata []xlMetaV1, errs []error)
 
 	// Create a new online disks slice, which have common uuid.
 	for index, t := range modTimes {
-		if t == modTime {
+		if t.Equal(modTime) {
 			onlineDisks[index] = disks[index]
 		} else {
 			onlineDisks[index] = nil
@@ -119,89 +117,67 @@ func listOnlineDisks(disks []StorageAPI, partsMetadata []xlMetaV1, errs []error)
 	return onlineDisks, modTime
 }
 
-// Returns the latest updated xlMeta files and error in case of failure.
+// Returns the latest updated FileInfo files and error in case of failure.
-func getLatestXLMeta(ctx context.Context, partsMetadata []xlMetaV1, errs []error) (xlMetaV1, error) {
+func getLatestFileInfo(ctx context.Context, partsMetadata []FileInfo, errs []error) (FileInfo, error) {
-
 	// There should be atleast half correct entries, if not return failure
 	if reducedErr := reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, len(partsMetadata)/2); reducedErr != nil {
-		return xlMetaV1{}, reducedErr
+		return FileInfo{}, reducedErr
 	}
 
 	// List all the file commit ids from parts metadata.
 	modTimes := listObjectModtimes(partsMetadata, errs)
 
-	// Count all latest updated xlMeta values
+	// Count all latest updated FileInfo values
 	var count int
-	var latestXLMeta xlMetaV1
+	var latestFileInfo FileInfo
 
 	// Reduce list of UUIDs to a single common value - i.e. the last updated Time
 	modTime, _ := commonTime(modTimes)
 
-	// Interate through all the modTimes and count the xlMeta(s) with latest time.
+	// Interate through all the modTimes and count the FileInfo(s) with latest time.
 	for index, t := range modTimes {
-		if t == modTime && partsMetadata[index].IsValid() {
+		if t.Equal(modTime) && partsMetadata[index].IsValid() {
-			latestXLMeta = partsMetadata[index]
+			latestFileInfo = partsMetadata[index]
 			count++
 		}
 	}
 	if count < len(partsMetadata)/2 {
-		return xlMetaV1{}, errXLReadQuorum
+		return FileInfo{}, errErasureReadQuorum
 	}
 
-	return latestXLMeta, nil
+	return latestFileInfo, nil
 }
 
 // disksWithAllParts - This function needs to be called with
 // []StorageAPI returned by listOnlineDisks. Returns,
 //
-// - disks which have all parts specified in the latest xl.json.
+// - disks which have all parts specified in the latest xl.meta.
 //
 // - slice of errors about the state of data files on disk - can have
 //   a not-found error or a hash-mismatch error.
-func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetadata []xlMetaV1, errs []error, bucket,
+func disksWithAllParts(ctx context.Context, onlineDisks []StorageAPI, partsMetadata []FileInfo, errs []error, bucket,
 	object string, scanMode madmin.HealScanMode) ([]StorageAPI, []error) {
 	availableDisks := make([]StorageAPI, len(onlineDisks))
 	dataErrs := make([]error, len(onlineDisks))
 
 	for i, onlineDisk := range onlineDisks {
-		if onlineDisk == nil {
+		if errs[i] != nil {
 			dataErrs[i] = errs[i]
 			continue
 		}
+		if onlineDisk == nil {
+			dataErrs[i] = errDiskNotFound
+			continue
+		}
 
 		switch scanMode {
 		case madmin.HealDeepScan:
-			erasure := partsMetadata[i].Erasure
+			// disk has a valid xl.meta but may not have all the
-
-			// disk has a valid xl.json but may not have all the
 			// parts. This is considered an outdated disk, since
 			// it needs healing too.
-			for _, part := range partsMetadata[i].Parts {
+			dataErrs[i] = onlineDisk.VerifyFile(bucket, object, partsMetadata[i])
-				checksumInfo := erasure.GetChecksumInfo(part.Number)
-				partPath := pathJoin(object, fmt.Sprintf("part.%d", part.Number))
-				err := onlineDisk.VerifyFile(bucket, partPath, erasure.ShardFileSize(part.Size), checksumInfo.Algorithm, checksumInfo.Hash, erasure.ShardSize())
-				if err != nil {
-					if !IsErr(err, []error{
-						errFileNotFound,
-						errVolumeNotFound,
-						errFileCorrupt,
-					}...) {
-						logger.GetReqInfo(ctx).AppendTags("disk", onlineDisk.String())
-						logger.LogIf(ctx, err)
-					}
-					dataErrs[i] = err
-					break
-				}
-			}
 		case madmin.HealNormalScan:
-			for _, part := range partsMetadata[i].Parts {
+			dataErrs[i] = onlineDisk.CheckParts(bucket, object, partsMetadata[i])
-				partPath := pathJoin(object, fmt.Sprintf("part.%d", part.Number))
-				_, err := onlineDisk.StatFile(bucket, partPath)
-				if err != nil {
-					dataErrs[i] = err
-					break
-				}
-			}
 		}
 
 		if dataErrs[i] == nil {

cmd/erasure-healing-common_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -95,9 +95,9 @@ func TestListOnlineDisks(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
-	obj, disks, err := prepareXL16(ctx)
+	obj, disks, err := prepareErasure16(ctx)
 	if err != nil {
-		t.Fatalf("Prepare XL backend failed - %v", err)
+		t.Fatalf("Prepare Erasure backend failed - %v", err)
 	}
 	defer removeRoots(disks)
 
@@ -141,9 +141,9 @@ func TestListOnlineDisks(t *testing.T) {
 			modTimes:     modTimesThreeNone,
 			expectedTime: threeNanoSecs,
 			errs: []error{
-				// Disks that have a valid xl.json.
+				// Disks that have a valid xl.meta.
 				nil, nil, nil, nil, nil, nil, nil,
-				// Majority of disks don't have xl.json.
+				// Majority of disks don't have xl.meta.
 				errFileNotFound, errFileNotFound,
 				errFileNotFound, errFileNotFound,
 				errFileNotFound, errDiskAccessDenied,
@@ -156,9 +156,9 @@ func TestListOnlineDisks(t *testing.T) {
 			modTimes:     modTimesThreeNone,
 			expectedTime: threeNanoSecs,
 			errs: []error{
-				// Disks that have a valid xl.json.
+				// Disks that have a valid xl.meta.
 				nil, nil, nil, nil, nil, nil, nil,
-				// Majority of disks don't have xl.json.
+				// Majority of disks don't have xl.meta.
 				errFileNotFound, errFileNotFound,
 				errFileNotFound, errFileNotFound,
 				errFileNotFound, errDiskAccessDenied,
@@ -170,27 +170,34 @@ func TestListOnlineDisks(t *testing.T) {
 	}
 
 	bucket := "bucket"
-	object := "object"
+	err = obj.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
-	data := bytes.Repeat([]byte("a"), 1024)
-	z := obj.(*xlZones)
-	xlDisks := z.zones[0].sets[0].getDisks()
-	for i, test := range testCases {
-		// Prepare bucket/object backend for the tests below.
-
-		// Cleanup from previous test.
-		obj.DeleteObject(GlobalContext, bucket, object)
-		obj.DeleteBucket(GlobalContext, bucket, false)
-
-		err = obj.MakeBucketWithLocation(GlobalContext, "bucket", "", false)
 	if err != nil {
 		t.Fatalf("Failed to make a bucket %v", err)
 	}
 
-		_, err = obj.PutObject(GlobalContext, bucket, object, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{})
+	object := "object"
+	data := bytes.Repeat([]byte("a"), 1024)
+	z := obj.(*erasureZones)
+	erasureDisks := z.zones[0].sets[0].getDisks()
+	for i, test := range testCases {
+		_, err = obj.PutObject(ctx, bucket, object, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{})
 		if err != nil {
 			t.Fatalf("Failed to putObject %v", err)
 		}
 
+		partsMetadata, errs := readAllFileInfo(ctx, erasureDisks, bucket, object, "")
+		fi, err := getLatestFileInfo(ctx, partsMetadata, errs)
+		if err != nil {
+			t.Fatalf("Failed to getLatestFileInfo %v", err)
+		}
+
+		for j := range partsMetadata {
+			if errs[j] != nil {
+				t.Fatalf("Test %d: expected error to be nil: %s", i+1, errs[j])
+			}
+			partsMetadata[j].ModTime = test.modTimes[j]
+		}
+
 		tamperedIndex := -1
 		switch test._tamperBackend {
 		case deletePart:
@@ -199,11 +206,11 @@ func TestListOnlineDisks(t *testing.T) {
 					continue
 				}
 				// Remove a part from a disk
-				// which has a valid xl.json,
+				// which has a valid xl.meta,
 				// and check if that disk
 				// appears in outDatedDisks.
 				tamperedIndex = index
-				dErr := xlDisks[index].DeleteFile(bucket, filepath.Join(object, "part.1"))
+				dErr := erasureDisks[index].DeleteFile(bucket, pathJoin(object, fi.DataDir, "part.1"))
 				if dErr != nil {
 					t.Fatalf("Test %d: Failed to delete %s - %v", i+1,
 						filepath.Join(object, "part.1"), dErr)
@@ -216,11 +223,11 @@ func TestListOnlineDisks(t *testing.T) {
 					continue
 				}
 				// Corrupt a part from a disk
-				// which has a valid xl.json,
+				// which has a valid xl.meta,
 				// and check if that disk
 				// appears in outDatedDisks.
 				tamperedIndex = index
-				filePath := pathJoin(xlDisks[index].String(), bucket, object, "part.1")
+				filePath := pathJoin(erasureDisks[index].String(), bucket, object, fi.DataDir, "part.1")
 				f, err := os.OpenFile(filePath, os.O_WRONLY|os.O_SYNC, 0)
 				if err != nil {
 					t.Fatalf("Failed to open %s: %s\n", filePath, err)
@@ -232,27 +239,19 @@ func TestListOnlineDisks(t *testing.T) {
 
 		}
 
-		partsMetadata, errs := readAllXLMetadata(GlobalContext, xlDisks, bucket, object)
+		onlineDisks, modTime := listOnlineDisks(erasureDisks, partsMetadata, test.errs)
-		for i := range partsMetadata {
-			if errs[i] != nil {
-				t.Fatalf("Test %d: expected error to be nil: %s", i+1, errs[i].Error())
-			}
-			partsMetadata[i].Stat.ModTime = test.modTimes[i]
-		}
-
-		onlineDisks, modTime := listOnlineDisks(xlDisks, partsMetadata, test.errs)
 		if !modTime.Equal(test.expectedTime) {
 			t.Fatalf("Test %d: Expected modTime to be equal to %v but was found to be %v",
 				i+1, test.expectedTime, modTime)
 		}
 
-		availableDisks, newErrs := disksWithAllParts(GlobalContext, onlineDisks, partsMetadata, test.errs, bucket, object, madmin.HealDeepScan)
+		availableDisks, newErrs := disksWithAllParts(ctx, onlineDisks, partsMetadata, test.errs, bucket, object, madmin.HealDeepScan)
 		test.errs = newErrs
 
 		if test._tamperBackend != noTamper {
 			if tamperedIndex != -1 && availableDisks[tamperedIndex] != nil {
 				t.Fatalf("Test %d: disk (%v) with part.1 missing is not a disk with available data",
-					i+1, xlDisks[tamperedIndex])
+					i+1, erasureDisks[tamperedIndex])
 			}
 		}
 
@@ -262,9 +261,9 @@ func TestListOnlineDisks(t *testing.T) {
 func TestDisksWithAllParts(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
-	obj, disks, err := prepareXL16(ctx)
+	obj, disks, err := prepareErasure16(ctx)
 	if err != nil {
-		t.Fatalf("Prepare XL backend failed - %v", err)
+		t.Fatalf("Prepare Erasure backend failed - %v", err)
 	}
 	defer removeRoots(disks)
 
@@ -273,10 +272,10 @@ func TestDisksWithAllParts(t *testing.T) {
 	// make data with more than one part
 	partCount := 3
 	data := bytes.Repeat([]byte("a"), 6*1024*1024*partCount)
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
-	xl := z.zones[0].sets[0]
+	s := z.zones[0].sets[0]
-	xlDisks := xl.getDisks()
+	erasureDisks := s.getDisks()
-	err = obj.MakeBucketWithLocation(ctx, "bucket", "", false)
+	err = obj.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
 	if err != nil {
 		t.Fatalf("Failed to make a bucket %v", err)
 	}
@@ -286,22 +285,22 @@ func TestDisksWithAllParts(t *testing.T) {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	_, errs := readAllXLMetadata(ctx, xlDisks, bucket, object)
+	_, errs := readAllFileInfo(ctx, erasureDisks, bucket, object, "")
-	readQuorum := len(xlDisks) / 2
+	readQuorum := len(erasureDisks) / 2
 	if reducedErr := reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum); reducedErr != nil {
 		t.Fatalf("Failed to read xl meta data %v", reducedErr)
 	}
 
 	// Test that all disks are returned without any failures with
 	// unmodified meta data
-	partsMetadata, errs := readAllXLMetadata(ctx, xlDisks, bucket, object)
+	partsMetadata, errs := readAllFileInfo(ctx, erasureDisks, bucket, object, "")
 	if err != nil {
 		t.Fatalf("Failed to read xl meta data %v", err)
 	}
 
-	filteredDisks, errs := disksWithAllParts(ctx, xlDisks, partsMetadata, errs, bucket, object, madmin.HealDeepScan)
+	filteredDisks, errs := disksWithAllParts(ctx, erasureDisks, partsMetadata, errs, bucket, object, madmin.HealDeepScan)
 
-	if len(filteredDisks) != len(xlDisks) {
+	if len(filteredDisks) != len(erasureDisks) {
 		t.Errorf("Unexpected number of disks: %d", len(filteredDisks))
 	}
 
@@ -324,7 +323,7 @@ func TestDisksWithAllParts(t *testing.T) {
 	for diskIndex, partName := range diskFailures {
 		for i := range partsMetadata[diskIndex].Erasure.Checksums {
 			if fmt.Sprintf("part.%d", i+1) == partName {
-				filePath := pathJoin(xlDisks[diskIndex].String(), bucket, object, partName)
+				filePath := pathJoin(erasureDisks[diskIndex].String(), bucket, object, partsMetadata[diskIndex].DataDir, partName)
 				f, err := os.OpenFile(filePath, os.O_WRONLY|os.O_SYNC, 0)
 				if err != nil {
 					t.Fatalf("Failed to open %s: %s\n", filePath, err)
@@ -335,10 +334,10 @@ func TestDisksWithAllParts(t *testing.T) {
 		}
 	}
 
-	errs = make([]error, len(xlDisks))
+	errs = make([]error, len(erasureDisks))
-	filteredDisks, errs = disksWithAllParts(ctx, xlDisks, partsMetadata, errs, bucket, object, madmin.HealDeepScan)
+	filteredDisks, errs = disksWithAllParts(ctx, erasureDisks, partsMetadata, errs, bucket, object, madmin.HealDeepScan)
 
-	if len(filteredDisks) != len(xlDisks) {
+	if len(filteredDisks) != len(erasureDisks) {
 		t.Errorf("Unexpected number of disks: %d", len(filteredDisks))
 	}
 

cmd/erasure-healing.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017, 2018 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -20,6 +20,7 @@ import (
 	"context"
 	"fmt"
 	"io"
+	"sync"
 	"time"
 
 	"github.com/minio/minio/cmd/logger"
@@ -27,12 +28,12 @@ import (
 	"github.com/minio/minio/pkg/sync/errgroup"
 )
 
-func (xl xlObjects) ReloadFormat(ctx context.Context, dryRun bool) error {
+func (er erasureObjects) ReloadFormat(ctx context.Context, dryRun bool) error {
 	logger.LogIf(ctx, NotImplemented{})
 	return NotImplemented{}
 }
 
-func (xl xlObjects) HealFormat(ctx context.Context, dryRun bool) (madmin.HealResultItem, error) {
+func (er erasureObjects) HealFormat(ctx context.Context, dryRun bool) (madmin.HealResultItem, error) {
 	logger.LogIf(ctx, NotImplemented{})
 	return madmin.HealResultItem{}, NotImplemented{}
 }
@@ -40,14 +41,14 @@ func (xl xlObjects) HealFormat(ctx context.Context, dryRun bool) (madmin.HealRes
 // Heals a bucket if it doesn't exist on one of the disks, additionally
 // also heals the missing entries for bucket metadata files
 // `policy.json, notification.xml, listeners.json`.
-func (xl xlObjects) HealBucket(ctx context.Context, bucket string, dryRun, remove bool) (
+func (er erasureObjects) HealBucket(ctx context.Context, bucket string, dryRun, remove bool) (
 	result madmin.HealResultItem, err error) {
 	if !dryRun {
 		defer ObjectPathUpdated(bucket)
 	}
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
-	storageEndpoints := xl.getEndpoints()
+	storageEndpoints := er.getEndpoints()
 
 	// get write quorum for an object
 	writeQuorum := getWriteQuorum(len(storageDisks))
@@ -158,7 +159,6 @@ func healBucket(ctx context.Context, storageDisks []StorageAPI, storageEndpoints
 			State:    afterState[i],
 		})
 	}
-
 	return res, nil
 }
 
@@ -196,22 +196,22 @@ func listAllBuckets(storageDisks []StorageAPI, healBuckets map[string]VolInfo) (
 
 // Only heal on disks where we are sure that healing is needed. We can expand
 // this list as and when we figure out more errors can be added to this list safely.
-func shouldHealObjectOnDisk(xlErr, dataErr error, meta xlMetaV1, quorumModTime time.Time) bool {
+func shouldHealObjectOnDisk(erErr, dataErr error, meta FileInfo, quorumModTime time.Time) bool {
-	switch xlErr {
+	switch erErr {
 	case errFileNotFound:
 		return true
 	case errCorruptedFormat:
 		return true
 	}
-	if xlErr == nil {
+	if erErr == nil {
-		// If xl.json was read fine but there may be problem with the part.N files.
+		// If er.meta was read fine but there may be problem with the part.N files.
 		if IsErr(dataErr, []error{
 			errFileNotFound,
 			errFileCorrupt,
 		}...) {
 			return true
 		}
-		if !quorumModTime.Equal(meta.Stat.ModTime) {
+		if !quorumModTime.Equal(meta.ModTime) {
 			return true
 		}
 	}
@@ -219,20 +219,20 @@ func shouldHealObjectOnDisk(xlErr, dataErr error, meta xlMetaV1, quorumModTime t
 }
 
 // Heals an object by re-writing corrupt/missing erasure blocks.
-func (xl xlObjects) healObject(ctx context.Context, bucket string, object string,
+func (er erasureObjects) healObject(ctx context.Context, bucket string, object string,
-	partsMetadata []xlMetaV1, errs []error, latestXLMeta xlMetaV1,
+	partsMetadata []FileInfo, errs []error, latestFileInfo FileInfo,
 	dryRun bool, remove bool, scanMode madmin.HealScanMode) (result madmin.HealResultItem, err error) {
 
-	dataBlocks := latestXLMeta.Erasure.DataBlocks
+	dataBlocks := latestFileInfo.Erasure.DataBlocks
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
-	storageEndpoints := xl.getEndpoints()
+	storageEndpoints := er.getEndpoints()
 
-	// List of disks having latest version of the object xl.json
+	// List of disks having latest version of the object er.meta
 	// (by modtime).
 	latestDisks, modTime := listOnlineDisks(storageDisks, partsMetadata, errs)
 
-	// List of disks having all parts as per latest xl.json.
+	// List of disks having all parts as per latest er.meta.
 	availableDisks, dataErrs := disksWithAllParts(ctx, latestDisks, partsMetadata, errs, bucket, object, scanMode)
 
 	// Initialize heal result object
@@ -241,8 +241,8 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 		Bucket:       bucket,
 		Object:       object,
 		DiskCount:    len(storageDisks),
-		ParityBlocks: latestXLMeta.Erasure.ParityBlocks,
+		ParityBlocks: latestFileInfo.Erasure.ParityBlocks,
-		DataBlocks:   latestXLMeta.Erasure.DataBlocks,
+		DataBlocks:   latestFileInfo.Erasure.DataBlocks,
 
 		// Initialize object size to -1, so we can detect if we are
 		// unable to reliably find the object size.
@@ -263,7 +263,7 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 			numAvailableDisks++
 			// If data is sane on any one disk, we can
 			// extract the correct object size.
-			result.ObjectSize = partsMetadata[i].Stat.Size
+			result.ObjectSize = partsMetadata[i].Size
 			result.ParityBlocks = partsMetadata[i].Erasure.ParityBlocks
 			result.DataBlocks = partsMetadata[i].Erasure.DataBlocks
 		case errs[i] == errDiskNotFound, dataErrs[i] == errDiskNotFound:
@@ -307,18 +307,18 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 	// If less than read quorum number of disks have all the parts
 	// of the data, we can't reconstruct the erasure-coded data.
 	if numAvailableDisks < dataBlocks {
-		// Check if xl.json, and corresponding parts are also missing.
+		// Check if er.meta, and corresponding parts are also missing.
 		if m, ok := isObjectDangling(partsMetadata, errs, dataErrs); ok {
 			writeQuorum := m.Erasure.DataBlocks + 1
 			if m.Erasure.DataBlocks == 0 {
 				writeQuorum = getWriteQuorum(len(storageDisks))
 			}
 			if !dryRun && remove {
-				err = xl.deleteObject(ctx, bucket, object, writeQuorum, false)
+				err = er.deleteObject(ctx, bucket, object, writeQuorum)
 			}
-			return defaultHealResult(latestXLMeta, storageDisks, storageEndpoints, errs, bucket, object), err
+			return defaultHealResult(latestFileInfo, storageDisks, storageEndpoints, errs, bucket, object), err
 		}
-		return result, toObjectErr(errXLReadQuorum, bucket, object)
+		return result, toObjectErr(errErasureReadQuorum, bucket, object)
 	}
 
 	if disksToHealCount == 0 {
@@ -332,32 +332,19 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 		return result, nil
 	}
 
-	// Latest xlMetaV1 for reference. If a valid metadata is not
+	// Latest FileInfo for reference. If a valid metadata is not
 	// present, it is as good as object not found.
-	latestMeta, pErr := pickValidXLMeta(ctx, partsMetadata, modTime, dataBlocks)
+	latestMeta, pErr := pickValidFileInfo(ctx, partsMetadata, modTime, dataBlocks)
 	if pErr != nil {
 		return result, toObjectErr(pErr, bucket, object)
 	}
 
-	// Clear data files of the object on outdated disks
+	cleanFileInfo := func(fi FileInfo) FileInfo {
-	for _, disk := range outDatedDisks {
+		// Returns a copy of the 'fi' with checksums and parts nil'ed.
-		// Before healing outdated disks, we need to remove
+		nfi := fi
-		// xl.json and part files from "bucket/object/" so
+		nfi.Erasure.Checksums = nil
-		// that rename(minioMetaBucket, "tmp/tmpuuid/",
+		nfi.Parts = nil
-		// "bucket", "object/") succeeds.
+		return nfi
-		if disk == nil {
-			// Not an outdated disk.
-			continue
-		}
-
-		// List and delete the object directory,
-		files, derr := disk.ListDir(bucket, object, -1, "")
-		if derr == nil {
-			for _, entry := range files {
-				_ = disk.DeleteFile(bucket,
-					pathJoin(object, entry))
-			}
-		}
 	}
 
 	// Reorder so that we have data disks first and parity disks next.
@@ -368,7 +355,7 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 		if outDatedDisks[i] == nil {
 			continue
 		}
-		partsMetadata[i] = newXLMetaFromXLMeta(latestMeta)
+		partsMetadata[i] = cleanFileInfo(latestMeta)
 	}
 
 	// We write at temporary location and then rename to final location.
@@ -388,7 +375,7 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 		partSize := latestMeta.Parts[partIndex].Size
 		partActualSize := latestMeta.Parts[partIndex].ActualSize
 		partNumber := latestMeta.Parts[partIndex].Number
-		tillOffset := erasure.ShardFileTillOffset(0, partSize, partSize)
+		tillOffset := erasure.ShardFileOffset(0, partSize, partSize)
 		readers := make([]io.ReaderAt, len(latestDisks))
 		checksumAlgo := erasureInfo.GetChecksumInfo(partNumber).Algorithm
 		for i, disk := range latestDisks {
@@ -396,7 +383,7 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 				continue
 			}
 			checksumInfo := partsMetadata[i].Erasure.GetChecksumInfo(partNumber)
-			partPath := pathJoin(object, fmt.Sprintf("part.%d", partNumber))
+			partPath := pathJoin(object, latestMeta.DataDir, fmt.Sprintf("part.%d", partNumber))
 			readers[i] = newBitrotReader(disk, bucket, partPath, tillOffset, checksumAlgo, checksumInfo.Hash, erasure.ShardSize())
 		}
 		writers := make([]io.Writer, len(outDatedDisks))
@@ -404,21 +391,22 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 			if disk == OfflineDisk {
 				continue
 			}
-			partPath := pathJoin(tmpID, fmt.Sprintf("part.%d", partNumber))
+			partPath := pathJoin(tmpID, latestMeta.DataDir, fmt.Sprintf("part.%d", partNumber))
-			writers[i] = newBitrotWriter(disk, minioMetaTmpBucket, partPath, tillOffset, checksumAlgo, erasure.ShardSize())
+			writers[i] = newBitrotWriter(disk, minioMetaTmpBucket, partPath, tillOffset, DefaultBitrotAlgorithm, erasure.ShardSize())
 		}
-		hErr := erasure.Heal(ctx, readers, writers, partSize)
+		err = erasure.Heal(ctx, readers, writers, partSize)
 		closeBitrotReaders(readers)
 		closeBitrotWriters(writers)
-		if hErr != nil {
+		if err != nil {
-			return result, toObjectErr(hErr, bucket, object)
+			return result, toObjectErr(err, bucket, object)
 		}
 		// outDatedDisks that had write errors should not be
 		// written to for remaining parts, so we nil it out.
 		for i, disk := range outDatedDisks {
-			if disk == nil {
+			if disk == OfflineDisk {
 				continue
 			}
+
 			// A non-nil stale disk which did not receive
 			// a healed part checksum had a write error.
 			if writers[i] == nil {
@@ -426,6 +414,7 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 				disksToHealCount--
 				continue
 			}
+
 			partsMetadata[i].AddObjectPart(partNumber, "", partSize, partActualSize)
 			partsMetadata[i].Erasure.AddChecksumInfo(ChecksumInfo{
 				PartNumber: partNumber,
@@ -436,33 +425,31 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 
 		// If all disks are having errors, we give up.
 		if disksToHealCount == 0 {
-			return result, fmt.Errorf("all disks without up-to-date data had write errors")
+			return result, fmt.Errorf("all disks had write errors, unable to heal")
 		}
 	}
 
-	// Cleanup in case of xl.json writing failure
+	// Cleanup in case of er.meta writing failure
 	writeQuorum := latestMeta.Erasure.DataBlocks + 1
-	defer xl.deleteObject(ctx, minioMetaTmpBucket, tmpID, writeQuorum, false)
+	defer er.deleteObject(ctx, minioMetaTmpBucket, tmpID, writeQuorum)
 
-	// Generate and write `xl.json` generated from other disks.
+	// Generate and write `xl.meta` generated from other disks.
-	outDatedDisks, aErr := writeUniqueXLMetadata(ctx, outDatedDisks, minioMetaTmpBucket, tmpID,
+	outDatedDisks, err = writeUniqueFileInfo(ctx, outDatedDisks, minioMetaTmpBucket, tmpID,
 		partsMetadata, diskCount(outDatedDisks))
-	if aErr != nil {
+	if err != nil {
-		return result, toObjectErr(aErr, bucket, object)
+		return result, toObjectErr(err, bucket, object)
 	}
 
 	// Rename from tmp location to the actual location.
 	for _, disk := range outDatedDisks {
-		if disk == nil {
+		if disk == OfflineDisk {
 			continue
 		}
 
 		// Attempt a rename now from healed data to final location.
-		aErr = disk.RenameFile(minioMetaTmpBucket, retainSlash(tmpID), bucket,
+		if err = disk.RenameData(minioMetaTmpBucket, tmpID, latestMeta.DataDir, bucket, object); err != nil {
-			retainSlash(object))
+			logger.LogIf(ctx, err)
-		if aErr != nil {
+			return result, toObjectErr(err, bucket, object)
-			logger.LogIf(ctx, aErr)
-			return result, toObjectErr(aErr, bucket, object)
 		}
 
 		for i, v := range result.Before.Drives {
@@ -473,16 +460,16 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 	}
 
 	// Set the size of the object in the heal result
-	result.ObjectSize = latestMeta.Stat.Size
+	result.ObjectSize = latestMeta.Size
 
 	return result, nil
 }
 
 // healObjectDir - heals object directory specifically, this special call
 // is needed since we do not have a special backend format for directories.
-func (xl xlObjects) healObjectDir(ctx context.Context, bucket, object string, dryRun bool, remove bool) (hr madmin.HealResultItem, err error) {
+func (er erasureObjects) healObjectDir(ctx context.Context, bucket, object string, dryRun bool, remove bool) (hr madmin.HealResultItem, err error) {
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
-	storageEndpoints := xl.getEndpoints()
+	storageEndpoints := er.getEndpoints()
 
 	// Initialize heal result object
 	hr = madmin.HealResultItem{
@@ -502,7 +489,19 @@ func (xl xlObjects) healObjectDir(ctx context.Context, bucket, object string, dr
 	danglingObject := isObjectDirDangling(errs)
 	if danglingObject {
 		if !dryRun && remove {
-			xl.deleteObject(ctx, bucket, object, hr.DataBlocks+1, true)
+			var wg sync.WaitGroup
+			// Remove versions in bulk for each disk
+			for index, disk := range storageDisks {
+				if disk == nil {
+					continue
+				}
+				wg.Add(1)
+				go func(index int, disk StorageAPI) {
+					defer wg.Done()
+					_ = disk.DeleteFile(bucket, object)
+				}(index, disk)
+			}
+			wg.Wait()
 		}
 	}
 
@@ -548,7 +547,7 @@ func (xl xlObjects) healObjectDir(ctx context.Context, bucket, object string, dr
 
 // Populates default heal result item entries with possible values when we are returning prematurely.
 // This is to ensure that in any circumstance we are not returning empty arrays with wrong values.
-func defaultHealResult(latestXLMeta xlMetaV1, storageDisks []StorageAPI, storageEndpoints []string, errs []error, bucket, object string) madmin.HealResultItem {
+func defaultHealResult(latestFileInfo FileInfo, storageDisks []StorageAPI, storageEndpoints []string, errs []error, bucket, object string) madmin.HealResultItem {
 	// Initialize heal result object
 	result := madmin.HealResultItem{
 		Type:      madmin.HealItemObject,
@@ -560,8 +559,8 @@ func defaultHealResult(latestXLMeta xlMetaV1, storageDisks []StorageAPI, storage
 		// unable to reliably find the object size.
 		ObjectSize: -1,
 	}
-	if latestXLMeta.IsValid() {
+	if latestFileInfo.IsValid() {
-		result.ObjectSize = latestXLMeta.Stat.Size
+		result.ObjectSize = latestFileInfo.Size
 	}
 
 	for index, disk := range storageDisks {
@@ -595,13 +594,13 @@ func defaultHealResult(latestXLMeta xlMetaV1, storageDisks []StorageAPI, storage
 		})
 	}
 
-	if !latestXLMeta.IsValid() {
+	if !latestFileInfo.IsValid() {
 		// Default to most common configuration for erasure blocks.
 		result.ParityBlocks = getDefaultParityBlocks(len(storageDisks))
 		result.DataBlocks = getDefaultDataBlocks(len(storageDisks))
 	} else {
-		result.ParityBlocks = latestXLMeta.Erasure.ParityBlocks
+		result.ParityBlocks = latestFileInfo.Erasure.ParityBlocks
-		result.DataBlocks = latestXLMeta.Erasure.DataBlocks
+		result.DataBlocks = latestFileInfo.Erasure.DataBlocks
 	}
 
 	return result
@@ -616,7 +615,7 @@ func statAllDirs(ctx context.Context, storageDisks []StorageAPI, bucket, prefix
 		}
 		index := index
 		g.Go(func() error {
-			entries, err := storageDisks[index].ListDir(bucket, prefix, 1, "")
+			entries, err := storageDisks[index].ListDir(bucket, prefix, 1)
 			if err != nil {
 				return err
 			}
@@ -655,23 +654,23 @@ func isObjectDirDangling(errs []error) (ok bool) {
 // Object is considered dangling/corrupted if any only
 // if total disks - a combination of corrupted and missing
 // files is lesser than number of data blocks.
-func isObjectDangling(metaArr []xlMetaV1, errs []error, dataErrs []error) (validMeta xlMetaV1, ok bool) {
+func isObjectDangling(metaArr []FileInfo, errs []error, dataErrs []error) (validMeta FileInfo, ok bool) {
 	// We can consider an object data not reliable
-	// when xl.json is not found in read quorum disks.
+	// when er.meta is not found in read quorum disks.
-	// or when xl.json is not readable in read quorum disks.
+	// or when er.meta is not readable in read quorum disks.
-	var notFoundXLJSON, corruptedXLJSON int
+	var notFoundErasureJSON, corruptedErasureJSON int
 	for _, readErr := range errs {
 		if readErr == errFileNotFound {
-			notFoundXLJSON++
+			notFoundErasureJSON++
 		} else if readErr == errCorruptedFormat {
-			corruptedXLJSON++
+			corruptedErasureJSON++
 		}
 	}
 	var notFoundParts int
 	for i := range dataErrs {
 		// Only count part errors, if the error is not
-		// same as xl.json error. This is to avoid
+		// same as er.meta error. This is to avoid
-		// double counting when both parts and xl.json
+		// double counting when both parts and er.meta
 		// are not available.
 		if errs[i] != dataErrs[i] {
 			if dataErrs[i] == errFileNotFound {
@@ -694,11 +693,11 @@ func isObjectDangling(metaArr []xlMetaV1, errs []error, dataErrs []error) (valid
 	}
 
 	// We have valid meta, now verify if we have enough files with parity blocks.
-	return validMeta, corruptedXLJSON+notFoundXLJSON+notFoundParts > validMeta.Erasure.ParityBlocks
+	return validMeta, corruptedErasureJSON+notFoundErasureJSON+notFoundParts > validMeta.Erasure.ParityBlocks
 }
 
 // HealObject - heal the given object, automatically deletes the object if stale/corrupted if `remove` is true.
-func (xl xlObjects) HealObject(ctx context.Context, bucket, object string, opts madmin.HealOpts) (hr madmin.HealResultItem, err error) {
+func (er erasureObjects) HealObject(ctx context.Context, bucket, object, versionID string, opts madmin.HealOpts) (hr madmin.HealResultItem, err error) {
 	// Create context that also contains information about the object and bucket.
 	// The top level handler might not have this information.
 	reqInfo := logger.GetReqInfo(ctx)
@@ -712,14 +711,14 @@ func (xl xlObjects) HealObject(ctx context.Context, bucket, object string, opts
 
 	// Healing directories handle it separately.
 	if HasSuffix(object, SlashSeparator) {
-		return xl.healObjectDir(healCtx, bucket, object, opts.DryRun, opts.Remove)
+		return er.healObjectDir(healCtx, bucket, object, opts.DryRun, opts.Remove)
 	}
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
-	storageEndpoints := xl.getEndpoints()
+	storageEndpoints := er.getEndpoints()
 
 	// Read metadata files from all the disks
-	partsMetadata, errs := readAllXLMetadata(healCtx, storageDisks, bucket, object)
+	partsMetadata, errs := readAllFileInfo(healCtx, storageDisks, bucket, object, versionID)
 
 	// Check if the object is dangling, if yes and user requested
 	// remove we simply delete it from namespace.
@@ -729,15 +728,15 @@ func (xl xlObjects) HealObject(ctx context.Context, bucket, object string, opts
 			writeQuorum = getWriteQuorum(len(storageDisks))
 		}
 		if !opts.DryRun && opts.Remove {
-			xl.deleteObject(healCtx, bucket, object, writeQuorum, false)
+			er.deleteObject(healCtx, bucket, object, writeQuorum)
 		}
 		err = reduceReadQuorumErrs(ctx, errs, nil, writeQuorum-1)
-		return defaultHealResult(xlMetaV1{}, storageDisks, storageEndpoints, errs, bucket, object), toObjectErr(err, bucket, object)
+		return defaultHealResult(FileInfo{}, storageDisks, storageEndpoints, errs, bucket, object), toObjectErr(err, bucket, object)
 	}
 
-	latestXLMeta, err := getLatestXLMeta(healCtx, partsMetadata, errs)
+	latestFileInfo, err := getLatestFileInfo(healCtx, partsMetadata, errs)
 	if err != nil {
-		return defaultHealResult(xlMetaV1{}, storageDisks, storageEndpoints, errs, bucket, object), toObjectErr(err, bucket, object)
+		return defaultHealResult(FileInfo{}, storageDisks, storageEndpoints, errs, bucket, object), toObjectErr(err, bucket, object)
 	}
 
 	errCount := 0
@@ -751,20 +750,20 @@ func (xl xlObjects) HealObject(ctx context.Context, bucket, object string, opts
 		// Only if we get errors from all the disks we return error. Else we need to
 		// continue to return filled madmin.HealResultItem struct which includes info
 		// on what disks the file is available etc.
-		if err = reduceReadQuorumErrs(ctx, errs, nil, latestXLMeta.Erasure.DataBlocks); err != nil {
+		if err = reduceReadQuorumErrs(ctx, errs, nil, latestFileInfo.Erasure.DataBlocks); err != nil {
 			if m, ok := isObjectDangling(partsMetadata, errs, []error{}); ok {
 				writeQuorum := m.Erasure.DataBlocks + 1
 				if m.Erasure.DataBlocks == 0 {
 					writeQuorum = getWriteQuorum(len(storageDisks))
 				}
 				if !opts.DryRun && opts.Remove {
-					xl.deleteObject(ctx, bucket, object, writeQuorum, false)
+					er.deleteObject(ctx, bucket, object, writeQuorum)
 				}
 			}
-			return defaultHealResult(latestXLMeta, storageDisks, storageEndpoints, errs, bucket, object), toObjectErr(err, bucket, object)
+			return defaultHealResult(latestFileInfo, storageDisks, storageEndpoints, errs, bucket, object), toObjectErr(err, bucket, object)
 		}
 	}
 
 	// Heal the object.
-	return xl.healObject(healCtx, bucket, object, partsMetadata, errs, latestXLMeta, opts.DryRun, opts.Remove, opts.ScanMode)
+	return er.healObject(healCtx, bucket, object, partsMetadata, errs, latestFileInfo, opts.DryRun, opts.Remove, opts.ScanMode)
 }

cmd/erasure-healing_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -19,12 +19,127 @@ package cmd
 import (
 	"bytes"
 	"context"
-	"path/filepath"
+	"crypto/rand"
+	"os"
+	"path"
+	"reflect"
 	"testing"
+	"time"
 
+	"github.com/dustin/go-humanize"
 	"github.com/minio/minio/pkg/madmin"
 )
 
+// Tests both object and bucket healing.
+func TestHealing(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	obj, fsDirs, err := prepareErasure16(ctx)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer removeRoots(fsDirs)
+
+	z := obj.(*erasureZones)
+	er := z.zones[0].sets[0]
+
+	// Create "bucket"
+	err = obj.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	bucket := "bucket"
+	object := "object"
+
+	data := make([]byte, 1*humanize.MiByte)
+	length := int64(len(data))
+	_, err = rand.Read(data)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	_, err = obj.PutObject(ctx, bucket, object, mustGetPutObjReader(t, bytes.NewReader(data), length, "", ""), ObjectOptions{})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	disk := er.getDisks()[0]
+	fileInfoPreHeal, err := disk.ReadVersion(bucket, object, "")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Remove the object - to simulate the case where the disk was down when the object
+	// was created.
+	err = removeAll(pathJoin(disk.String(), bucket, object))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	_, err = er.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealNormalScan})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	fileInfoPostHeal, err := disk.ReadVersion(bucket, object, "")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// After heal the meta file should be as expected.
+	if !reflect.DeepEqual(fileInfoPreHeal, fileInfoPostHeal) {
+		t.Fatal("HealObject failed")
+	}
+
+	err = os.RemoveAll(path.Join(fsDirs[0], bucket, object, "er.meta"))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Write er.meta with different modtime to simulate the case where a disk had
+	// gone down when an object was replaced by a new object.
+	fileInfoOutDated := fileInfoPreHeal
+	fileInfoOutDated.ModTime = time.Now()
+	err = disk.WriteMetadata(bucket, object, fileInfoOutDated)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	_, err = er.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealDeepScan})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	fileInfoPostHeal, err = disk.ReadVersion(bucket, object, "")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// After heal the meta file should be as expected.
+	if !reflect.DeepEqual(fileInfoPreHeal, fileInfoPostHeal) {
+		t.Fatal("HealObject failed")
+	}
+
+	// Remove the bucket - to simulate the case where bucket was
+	// created when the disk was down.
+	err = os.RemoveAll(path.Join(fsDirs[0], bucket))
+	if err != nil {
+		t.Fatal(err)
+	}
+	// This would create the bucket.
+	_, err = er.HealBucket(ctx, bucket, false, false)
+	if err != nil {
+		t.Fatal(err)
+	}
+	// Stat the bucket to make sure that it was created.
+	_, err = er.getDisks()[0].StatVol(bucket)
+	if err != nil {
+		t.Fatal(err)
+	}
+}
+
 func TestHealObjectCorrupted(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -51,7 +166,7 @@ func TestHealObjectCorrupted(t *testing.T) {
 	data := bytes.Repeat([]byte("a"), 5*1024*1024)
 	var opts ObjectOptions
 
-	err = objLayer.MakeBucketWithLocation(ctx, bucket, "", false)
+	err = objLayer.MakeBucketWithLocation(ctx, bucket, BucketOptions{})
 	if err != nil {
 		t.Fatalf("Failed to make a bucket - %v", err)
 	}
@@ -81,91 +196,96 @@ func TestHealObjectCorrupted(t *testing.T) {
 	}
 
 	// Test 1: Remove the object backend files from the first disk.
-	z := objLayer.(*xlZones)
+	z := objLayer.(*erasureZones)
-	xl := z.zones[0].sets[0]
+	er := z.zones[0].sets[0]
-	firstDisk := xl.getDisks()[0]
+	erasureDisks := er.getDisks()
-	err = firstDisk.DeleteFile(bucket, filepath.Join(object, xlMetaJSONFile))
+	firstDisk := erasureDisks[0]
+	err = firstDisk.DeleteFile(bucket, pathJoin(object, xlStorageFormatFile))
 	if err != nil {
 		t.Fatalf("Failed to delete a file - %v", err)
 	}
 
-	_, err = objLayer.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealNormalScan})
+	_, err = objLayer.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealNormalScan})
 	if err != nil {
 		t.Fatalf("Failed to heal object - %v", err)
 	}
 
-	_, err = firstDisk.StatFile(bucket, filepath.Join(object, xlMetaJSONFile))
+	fileInfos, errs := readAllFileInfo(ctx, erasureDisks, bucket, object, "")
+	fi, err := getLatestFileInfo(ctx, fileInfos, errs)
 	if err != nil {
-		t.Errorf("Expected xl.json file to be present but stat failed - %v", err)
+		t.Fatalf("Failed to getLatestFileInfo - %v", err)
 	}
 
-	// Test 2: Heal when part.1 is empty
+	if err = firstDisk.CheckFile(bucket, object); err != nil {
-	partSt1, err := firstDisk.StatFile(bucket, filepath.Join(object, "part.1"))
+		t.Errorf("Expected er.meta file to be present but stat failed - %v", err)
-	if err != nil {
-		t.Errorf("Expected part.1 file to be present but stat failed - %v", err)
 	}
-	err = firstDisk.DeleteFile(bucket, filepath.Join(object, "part.1"))
+
+	err = firstDisk.DeleteFile(bucket, pathJoin(object, fi.DataDir, "part.1"))
 	if err != nil {
 		t.Errorf("Failure during deleting part.1 - %v", err)
 	}
-	err = firstDisk.WriteAll(bucket, filepath.Join(object, "part.1"), bytes.NewReader([]byte{}))
+
+	err = firstDisk.WriteAll(bucket, pathJoin(object, fi.DataDir, "part.1"), bytes.NewReader([]byte{}))
 	if err != nil {
 		t.Errorf("Failure during creating part.1 - %v", err)
 	}
-	_, err = objLayer.HealObject(ctx, bucket, object, madmin.HealOpts{DryRun: false, Remove: true, ScanMode: madmin.HealDeepScan})
+
+	_, err = objLayer.HealObject(ctx, bucket, object, "", madmin.HealOpts{DryRun: false, Remove: true, ScanMode: madmin.HealDeepScan})
 	if err != nil {
 		t.Errorf("Expected nil but received %v", err)
 	}
-	partSt2, err := firstDisk.StatFile(bucket, filepath.Join(object, "part.1"))
+
+	fileInfos, errs = readAllFileInfo(ctx, erasureDisks, bucket, object, "")
+	nfi, err := getLatestFileInfo(ctx, fileInfos, errs)
 	if err != nil {
-		t.Errorf("Expected from part.1 file to be present but stat failed - %v", err)
+		t.Fatalf("Failed to getLatestFileInfo - %v", err)
-	}
-	if partSt1.Size != partSt2.Size {
-		t.Errorf("part.1 file size is not the same before and after heal")
 	}
 
-	// Test 3: Heal when part.1 is correct in size but corrupted
+	if !reflect.DeepEqual(fi, nfi) {
-	partSt1, err = firstDisk.StatFile(bucket, filepath.Join(object, "part.1"))
+		t.Fatalf("FileInfo not equal after healing")
-	if err != nil {
-		t.Errorf("Expected part.1 file to be present but stat failed - %v", err)
 	}
-	err = firstDisk.DeleteFile(bucket, filepath.Join(object, "part.1"))
+
+	err = firstDisk.DeleteFile(bucket, pathJoin(object, fi.DataDir, "part.1"))
 	if err != nil {
 		t.Errorf("Failure during deleting part.1 - %v", err)
 	}
-	bdata := bytes.Repeat([]byte("b"), int(partSt1.Size))
+
-	err = firstDisk.WriteAll(bucket, filepath.Join(object, "part.1"), bytes.NewReader(bdata))
+	bdata := bytes.Repeat([]byte("b"), int(nfi.Size))
+	err = firstDisk.WriteAll(bucket, pathJoin(object, fi.DataDir, "part.1"), bytes.NewReader(bdata))
 	if err != nil {
 		t.Errorf("Failure during creating part.1 - %v", err)
 	}
-	_, err = objLayer.HealObject(ctx, bucket, object, madmin.HealOpts{DryRun: false, Remove: true, ScanMode: madmin.HealDeepScan})
+
+	_, err = objLayer.HealObject(ctx, bucket, object, "", madmin.HealOpts{DryRun: false, Remove: true, ScanMode: madmin.HealDeepScan})
 	if err != nil {
 		t.Errorf("Expected nil but received %v", err)
 	}
-	partSt2, err = firstDisk.StatFile(bucket, filepath.Join(object, "part.1"))
+
+	fileInfos, errs = readAllFileInfo(ctx, erasureDisks, bucket, object, "")
+	nfi, err = getLatestFileInfo(ctx, fileInfos, errs)
 	if err != nil {
-		t.Errorf("Expected from part.1 file to be present but stat failed - %v", err)
+		t.Fatalf("Failed to getLatestFileInfo - %v", err)
-	}
-	if partSt1.Size != partSt2.Size {
-		t.Errorf("part.1 file size is not the same before and after heal")
 	}
 
-	// Test 4: checks if HealObject returns an error when xl.json is not found
+	if !reflect.DeepEqual(fi, nfi) {
+		t.Fatalf("FileInfo not equal after healing")
+	}
+
+	// Test 4: checks if HealObject returns an error when xl.meta is not found
 	// in more than read quorum number of disks, to create a corrupted situation.
-
+	for i := 0; i <= len(er.getDisks())/2; i++ {
-	for i := 0; i <= len(xl.getDisks())/2; i++ {
+		er.getDisks()[i].DeleteFile(bucket, pathJoin(object, xlStorageFormatFile))
-		xl.getDisks()[i].DeleteFile(bucket, filepath.Join(object, xlMetaJSONFile))
 	}
 
 	// Try healing now, expect to receive errFileNotFound.
-	_, err = objLayer.HealObject(ctx, bucket, object, madmin.HealOpts{DryRun: false, Remove: true, ScanMode: madmin.HealDeepScan})
+	_, err = objLayer.HealObject(ctx, bucket, object, "", madmin.HealOpts{DryRun: false, Remove: true, ScanMode: madmin.HealDeepScan})
 	if err != nil {
 		if _, ok := err.(ObjectNotFound); !ok {
 			t.Errorf("Expect %v but received %v", ObjectNotFound{Bucket: bucket, Object: object}, err)
 		}
 	}
 
-	// since majority of xl.jsons are not available, object should be successfully deleted.
+	// since majority of xl.meta's are not available, object should be successfully deleted.
 	_, err = objLayer.GetObjectInfo(ctx, bucket, object, ObjectOptions{})
 	if _, ok := err.(ObjectNotFound); !ok {
 		t.Errorf("Expect %v but received %v", ObjectNotFound{Bucket: bucket, Object: object}, err)
@@ -173,7 +293,7 @@ func TestHealObjectCorrupted(t *testing.T) {
 }
 
 // Tests healing of object.
-func TestHealObjectXL(t *testing.T) {
+func TestHealObjectErasure(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
@@ -196,7 +316,7 @@ func TestHealObjectXL(t *testing.T) {
 	data := bytes.Repeat([]byte("a"), 5*1024*1024)
 	var opts ObjectOptions
 
-	err = obj.MakeBucketWithLocation(ctx, bucket, "", false)
+	err = obj.MakeBucketWithLocation(ctx, bucket, BucketOptions{})
 	if err != nil {
 		t.Fatalf("Failed to make a bucket - %v", err)
 	}
@@ -220,51 +340,51 @@ func TestHealObjectXL(t *testing.T) {
 		})
 	}
 
+	// Remove the object backend files from the first disk.
+	z := obj.(*erasureZones)
+	er := z.zones[0].sets[0]
+	firstDisk := er.getDisks()[0]
+
 	_, err = obj.CompleteMultipartUpload(ctx, bucket, object, uploadID, uploadedParts, ObjectOptions{})
 	if err != nil {
 		t.Fatalf("Failed to complete multipart upload - %v", err)
 	}
 
-	// Remove the object backend files from the first disk.
+	err = firstDisk.DeleteFile(bucket, pathJoin(object, xlStorageFormatFile))
-	z := obj.(*xlZones)
-	xl := z.zones[0].sets[0]
-	firstDisk := xl.getDisks()[0]
-	err = firstDisk.DeleteFile(bucket, filepath.Join(object, xlMetaJSONFile))
 	if err != nil {
 		t.Fatalf("Failed to delete a file - %v", err)
 	}
 
-	_, err = obj.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealNormalScan})
+	_, err = obj.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealNormalScan})
 	if err != nil {
 		t.Fatalf("Failed to heal object - %v", err)
 	}
 
-	_, err = firstDisk.StatFile(bucket, filepath.Join(object, xlMetaJSONFile))
+	if err = firstDisk.CheckFile(bucket, object); err != nil {
-	if err != nil {
+		t.Errorf("Expected er.meta file to be present but stat failed - %v", err)
-		t.Errorf("Expected xl.json file to be present but stat failed - %v", err)
 	}
 
-	xlDisks := xl.getDisks()
+	erasureDisks := er.getDisks()
-	z.zones[0].xlDisksMu.Lock()
+	z.zones[0].erasureDisksMu.Lock()
-	xl.getDisks = func() []StorageAPI {
+	er.getDisks = func() []StorageAPI {
 		// Nil more than half the disks, to remove write quorum.
-		for i := 0; i <= len(xlDisks)/2; i++ {
+		for i := 0; i <= len(erasureDisks)/2; i++ {
-			xlDisks[i] = nil
+			erasureDisks[i] = nil
 		}
-		return xlDisks
+		return erasureDisks
 	}
-	z.zones[0].xlDisksMu.Unlock()
+	z.zones[0].erasureDisksMu.Unlock()
 
 	// Try healing now, expect to receive errDiskNotFound.
-	_, err = obj.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealDeepScan})
+	_, err = obj.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealDeepScan})
-	// since majority of xl.jsons are not available, object quorum can't be read properly and error will be errXLReadQuorum
+	// since majority of er.meta's are not available, object quorum can't be read properly and error will be errErasureReadQuorum
 	if _, ok := err.(InsufficientReadQuorum); !ok {
 		t.Errorf("Expected %v but received %v", InsufficientReadQuorum{}, err)
 	}
 }
 
 // Tests healing of empty directories
-func TestHealEmptyDirectoryXL(t *testing.T) {
+func TestHealEmptyDirectoryErasure(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
@@ -285,7 +405,7 @@ func TestHealEmptyDirectoryXL(t *testing.T) {
 	object := "empty-dir/"
 	var opts ObjectOptions
 
-	err = obj.MakeBucketWithLocation(ctx, bucket, "", false)
+	err = obj.MakeBucketWithLocation(ctx, bucket, BucketOptions{})
 	if err != nil {
 		t.Fatalf("Failed to make a bucket - %v", err)
 	}
@@ -298,16 +418,16 @@ func TestHealEmptyDirectoryXL(t *testing.T) {
 	}
 
 	// Remove the object backend files from the first disk.
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
-	xl := z.zones[0].sets[0]
+	er := z.zones[0].sets[0]
-	firstDisk := xl.getDisks()[0]
+	firstDisk := er.getDisks()[0]
 	err = firstDisk.DeleteFile(bucket, object)
 	if err != nil {
 		t.Fatalf("Failed to delete a file - %v", err)
 	}
 
 	// Heal the object
-	hr, err := obj.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealNormalScan})
+	hr, err := obj.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealNormalScan})
 	if err != nil {
 		t.Fatalf("Failed to heal object - %v", err)
 	}
@@ -331,7 +451,7 @@ func TestHealEmptyDirectoryXL(t *testing.T) {
 	}
 
 	// Heal the same object again
-	hr, err = obj.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealNormalScan})
+	hr, err = obj.HealObject(ctx, bucket, object, "", madmin.HealOpts{ScanMode: madmin.HealNormalScan})
 	if err != nil {
 		t.Fatalf("Failed to heal object - %v", err)
 	}

cmd/erasure-list-objects.go

@@ -0,0 +1,58 @@
+/*
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"context"
+
+	"github.com/minio/minio/pkg/madmin"
+)
+
+// ListObjectVersions - This is not implemented, look for erasure-zones.ListObjectVersions()
+func (er erasureObjects) ListObjectVersions(ctx context.Context, bucket, prefix, marker, versionMarker, delimiter string, maxKeys int) (loi ListObjectVersionsInfo, e error) {
+	return loi, NotImplemented{}
+}
+
+// ListObjectsV2 - This is not implemented/needed anymore, look for erasure-zones.ListObjectsV2()
+func (er erasureObjects) ListObjectsV2(ctx context.Context, bucket, prefix, continuationToken, delimiter string, maxKeys int, fetchOwner bool, startAfter string) (loi ListObjectsV2Info, e error) {
+	return loi, NotImplemented{}
+}
+
+// ListObjects - This is not implemented/needed anymore, look for erasure-zones.ListObjects()
+func (er erasureObjects) ListObjects(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (loi ListObjectsInfo, e error) {
+	return loi, NotImplemented{}
+}
+
+// ListBucketsHeal - This is not implemented/needed anymore, look for erasure-zones.ListBucketHeal()
+func (er erasureObjects) ListBucketsHeal(ctx context.Context) ([]BucketInfo, error) {
+	return nil, NotImplemented{}
+}
+
+// ListObjectsHeal - This is not implemented, look for erasure-zones.ListObjectsHeal()
+func (er erasureObjects) ListObjectsHeal(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (loi ListObjectsInfo, err error) {
+	return ListObjectsInfo{}, NotImplemented{}
+}
+
+// HealObjects - This is not implemented/needed anymore, look for erasure-zones.HealObjects()
+func (er erasureObjects) HealObjects(ctx context.Context, bucket, prefix string, _ madmin.HealOpts, _ HealObjectFn) (e error) {
+	return NotImplemented{}
+}
+
+// Walk - This is not implemented/needed anymore, look for erasure-zones.Walk()
+func (er erasureObjects) Walk(ctx context.Context, bucket, prefix string, results chan<- ObjectInfo) error {
+	return NotImplemented{}
+}

cmd/erasure-lowlevel-heal.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

cmd/erasure-metadata-utils.go

@@ -20,9 +20,7 @@ import (
 	"context"
 	"errors"
 	"hash/crc32"
-	"path"
 
-	jsoniter "github.com/json-iterator/go"
 	"github.com/minio/minio/cmd/logger"
 	"github.com/minio/minio/pkg/sync/errgroup"
 )
@@ -72,13 +70,13 @@ func reduceQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, qu
 // reduceReadQuorumErrs behaves like reduceErrs but only for returning
 // values of maximally occurring errors validated against readQuorum.
 func reduceReadQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, readQuorum int) (maxErr error) {
-	return reduceQuorumErrs(ctx, errs, ignoredErrs, readQuorum, errXLReadQuorum)
+	return reduceQuorumErrs(ctx, errs, ignoredErrs, readQuorum, errErasureReadQuorum)
 }
 
 // reduceWriteQuorumErrs behaves like reduceErrs but only for returning
 // values of maximally occurring errors validated against writeQuorum.
 func reduceWriteQuorumErrs(ctx context.Context, errs []error, ignoredErrs []error, writeQuorum int) (maxErr error) {
-	return reduceQuorumErrs(ctx, errs, ignoredErrs, writeQuorum, errXLWriteQuorum)
+	return reduceQuorumErrs(ctx, errs, ignoredErrs, writeQuorum, errErasureWriteQuorum)
 }
 
 // Similar to 'len(slice)' but returns the actual elements count
@@ -115,44 +113,26 @@ func hashOrder(key string, cardinality int) []int {
 	return nums
 }
 
-// Constructs xlMetaV1 using `jsoniter` lib.
+// Reads all `xl.meta` metadata as a FileInfo slice.
-func xlMetaV1UnmarshalJSON(ctx context.Context, xlMetaBuf []byte) (xlMeta xlMetaV1, err error) {
-	var json = jsoniter.ConfigCompatibleWithStandardLibrary
-	err = json.Unmarshal(xlMetaBuf, &xlMeta)
-	return xlMeta, err
-}
-
-// readXLMeta reads `xl.json` and returns back XL metadata structure.
-func readXLMeta(ctx context.Context, disk StorageAPI, bucket string, object string) (xlMeta xlMetaV1, err error) {
-	// Reads entire `xl.json`.
-	xlMetaBuf, err := disk.ReadAll(bucket, path.Join(object, xlMetaJSONFile))
-	if err != nil {
-		if err != errFileNotFound && err != errVolumeNotFound {
-			logger.GetReqInfo(ctx).AppendTags("disk", disk.String())
-			logger.LogIf(ctx, err)
-		}
-		return xlMetaV1{}, err
-	}
-	if len(xlMetaBuf) == 0 {
-		return xlMetaV1{}, errFileNotFound
-	}
-	return xlMetaV1UnmarshalJSON(ctx, xlMetaBuf)
-}
-
-// Reads all `xl.json` metadata as a xlMetaV1 slice.
 // Returns error slice indicating the failed metadata reads.
-func readAllXLMetadata(ctx context.Context, disks []StorageAPI, bucket, object string) ([]xlMetaV1, []error) {
+func readAllFileInfo(ctx context.Context, disks []StorageAPI, bucket, object, versionID string) ([]FileInfo, []error) {
-	metadataArray := make([]xlMetaV1, len(disks))
+	metadataArray := make([]FileInfo, len(disks))
 
 	g := errgroup.WithNErrs(len(disks))
-	// Read `xl.json` parallelly across disks.
+	// Read `xl.meta` parallelly across disks.
 	for index := range disks {
 		index := index
 		g.Go(func() (err error) {
 			if disks[index] == nil {
 				return errDiskNotFound
 			}
-			metadataArray[index], err = readXLMeta(ctx, disks[index], bucket, object)
+			metadataArray[index], err = disks[index].ReadVersion(bucket, object, versionID)
+			if err != nil {
+				if err != errFileNotFound && err != errVolumeNotFound && err != errFileVersionNotFound {
+					logger.GetReqInfo(ctx).AppendTags("disk", disks[index].String())
+					logger.LogIf(ctx, err)
+				}
+			}
 			return err
 		}, index)
 	}
@@ -162,11 +142,11 @@ func readAllXLMetadata(ctx context.Context, disks []StorageAPI, bucket, object s
 }
 
 // Return shuffled partsMetadata depending on distribution.
-func shufflePartsMetadata(partsMetadata []xlMetaV1, distribution []int) (shuffledPartsMetadata []xlMetaV1) {
+func shufflePartsMetadata(partsMetadata []FileInfo, distribution []int) (shuffledPartsMetadata []FileInfo) {
 	if distribution == nil {
 		return partsMetadata
 	}
-	shuffledPartsMetadata = make([]xlMetaV1, len(partsMetadata))
+	shuffledPartsMetadata = make([]FileInfo, len(partsMetadata))
 	// Shuffle slice xl metadata for expected distribution.
 	for index := range partsMetadata {
 		blockIndex := distribution[index]

cmd/erasure-metadata-utils_test.go

@@ -0,0 +1,201 @@
+/*
+ * MinIO Cloud Storage, (C) 2015, 2016, 2017 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"context"
+	"reflect"
+	"testing"
+)
+
+// Tests caclculating disk count.
+func TestDiskCount(t *testing.T) {
+	testCases := []struct {
+		disks     []StorageAPI
+		diskCount int
+	}{
+		// Test case - 1
+		{
+			disks:     []StorageAPI{&xlStorage{}, &xlStorage{}, &xlStorage{}, &xlStorage{}},
+			diskCount: 4,
+		},
+		// Test case - 2
+		{
+			disks:     []StorageAPI{nil, &xlStorage{}, &xlStorage{}, &xlStorage{}},
+			diskCount: 3,
+		},
+	}
+	for i, testCase := range testCases {
+		cdiskCount := diskCount(testCase.disks)
+		if cdiskCount != testCase.diskCount {
+			t.Errorf("Test %d: Expected %d, got %d", i+1, testCase.diskCount, cdiskCount)
+		}
+	}
+}
+
+// Test for reduceErrs, reduceErr reduces collection
+// of errors into a single maximal error with in the list.
+func TestReduceErrs(t *testing.T) {
+	// List all of all test cases to validate various cases of reduce errors.
+	testCases := []struct {
+		errs        []error
+		ignoredErrs []error
+		err         error
+	}{
+		// Validate if have reduced properly.
+		{[]error{
+			errDiskNotFound,
+			errDiskNotFound,
+			errDiskFull,
+		}, []error{}, errErasureReadQuorum},
+		// Validate if have no consensus.
+		{[]error{
+			errDiskFull,
+			errDiskNotFound,
+			nil, nil,
+		}, []error{}, errErasureReadQuorum},
+		// Validate if have consensus and errors ignored.
+		{[]error{
+			errVolumeNotFound,
+			errVolumeNotFound,
+			errVolumeNotFound,
+			errVolumeNotFound,
+			errVolumeNotFound,
+			errDiskNotFound,
+			errDiskNotFound,
+		}, []error{errDiskNotFound}, errVolumeNotFound},
+		{[]error{}, []error{}, errErasureReadQuorum},
+		{[]error{errFileNotFound, errFileNotFound, errFileNotFound,
+			errFileNotFound, errFileNotFound, nil, nil, nil, nil, nil},
+			nil, nil},
+	}
+	// Validates list of all the testcases for returning valid errors.
+	for i, testCase := range testCases {
+		gotErr := reduceReadQuorumErrs(context.Background(), testCase.errs, testCase.ignoredErrs, 5)
+		if gotErr != testCase.err {
+			t.Errorf("Test %d : expected %s, got %s", i+1, testCase.err, gotErr)
+		}
+		gotNewErr := reduceWriteQuorumErrs(context.Background(), testCase.errs, testCase.ignoredErrs, 6)
+		if gotNewErr != errErasureWriteQuorum {
+			t.Errorf("Test %d : expected %s, got %s", i+1, errErasureWriteQuorum, gotErr)
+		}
+	}
+}
+
+// TestHashOrder - test order of ints in array
+func TestHashOrder(t *testing.T) {
+	testCases := []struct {
+		objectName  string
+		hashedOrder []int
+	}{
+		// cases which should pass the test.
+		// passing in valid object name.
+		{"object", []int{14, 15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13}},
+		{"The Shining Script <v1>.pdf", []int{16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15}},
+		{"Cost Benefit Analysis (2009-2010).pptx", []int{15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14}},
+		{"117Gn8rfHL2ACARPAhaFd0AGzic9pUbIA/5OCn5A", []int{3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2}},
+		{"SHØRT", []int{11, 12, 13, 14, 15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}},
+		{"There are far too many object names, and far too few bucket names!", []int{15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14}},
+		{"a/b/c/", []int{3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2}},
+		{"/a/b/c", []int{6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2, 3, 4, 5}},
+		{string([]byte{0xff, 0xfe, 0xfd}), []int{15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14}},
+	}
+
+	// Tests hashing order to be consistent.
+	for i, testCase := range testCases {
+		hashedOrder := hashOrder(testCase.objectName, 16)
+		if !reflect.DeepEqual(testCase.hashedOrder, hashedOrder) {
+			t.Errorf("Test case %d: Expected \"%v\" but failed \"%v\"", i+1, testCase.hashedOrder, hashedOrder)
+		}
+	}
+
+	// Tests hashing order to fail for when order is '-1'.
+	if hashedOrder := hashOrder("This will fail", -1); hashedOrder != nil {
+		t.Errorf("Test: Expect \"nil\" but failed \"%#v\"", hashedOrder)
+	}
+
+	if hashedOrder := hashOrder("This will fail", 0); hashedOrder != nil {
+		t.Errorf("Test: Expect \"nil\" but failed \"%#v\"", hashedOrder)
+	}
+}
+
+func TestShuffleDisks(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	nDisks := 16
+	disks, err := getRandomDisks(nDisks)
+	if err != nil {
+		t.Fatal(err)
+	}
+	objLayer, _, err := initObjectLayer(ctx, mustGetZoneEndpoints(disks...))
+	if err != nil {
+		removeRoots(disks)
+		t.Fatal(err)
+	}
+	defer removeRoots(disks)
+	z := objLayer.(*erasureZones)
+	testShuffleDisks(t, z)
+}
+
+// Test shuffleDisks which returns shuffled slice of disks for their actual distribution.
+func testShuffleDisks(t *testing.T, z *erasureZones) {
+	disks := z.zones[0].GetDisks(0)()
+	distribution := []int{16, 14, 12, 10, 8, 6, 4, 2, 1, 3, 5, 7, 9, 11, 13, 15}
+	shuffledDisks := shuffleDisks(disks, distribution)
+	// From the "distribution" above you can notice that:
+	// 1st data block is in the 9th disk (i.e distribution index 8)
+	// 2nd data block is in the 8th disk (i.e distribution index 7) and so on.
+	if shuffledDisks[0] != disks[8] ||
+		shuffledDisks[1] != disks[7] ||
+		shuffledDisks[2] != disks[9] ||
+		shuffledDisks[3] != disks[6] ||
+		shuffledDisks[4] != disks[10] ||
+		shuffledDisks[5] != disks[5] ||
+		shuffledDisks[6] != disks[11] ||
+		shuffledDisks[7] != disks[4] ||
+		shuffledDisks[8] != disks[12] ||
+		shuffledDisks[9] != disks[3] ||
+		shuffledDisks[10] != disks[13] ||
+		shuffledDisks[11] != disks[2] ||
+		shuffledDisks[12] != disks[14] ||
+		shuffledDisks[13] != disks[1] ||
+		shuffledDisks[14] != disks[15] ||
+		shuffledDisks[15] != disks[0] {
+		t.Errorf("shuffleDisks returned incorrect order.")
+	}
+}
+
+// TestEvalDisks tests the behavior of evalDisks
+func TestEvalDisks(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	nDisks := 16
+	disks, err := getRandomDisks(nDisks)
+	if err != nil {
+		t.Fatal(err)
+	}
+	objLayer, _, err := initObjectLayer(ctx, mustGetZoneEndpoints(disks...))
+	if err != nil {
+		removeRoots(disks)
+		t.Fatal(err)
+	}
+	defer removeRoots(disks)
+	z := objLayer.(*erasureZones)
+	testShuffleDisks(t, z)
+}

cmd/erasure-metadata.go

@@ -0,0 +1,326 @@
+/*
+ * MinIO Cloud Storage, (C) 2016-2019 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"context"
+	"encoding/hex"
+	"fmt"
+	"net/http"
+	"sort"
+	"time"
+
+	xhttp "github.com/minio/minio/cmd/http"
+	"github.com/minio/minio/cmd/logger"
+	"github.com/minio/minio/pkg/sync/errgroup"
+	"github.com/minio/sha256-simd"
+)
+
+const erasureAlgorithm = "rs-vandermonde"
+
+// byObjectPartNumber is a collection satisfying sort.Interface.
+type byObjectPartNumber []ObjectPartInfo
+
+func (t byObjectPartNumber) Len() int           { return len(t) }
+func (t byObjectPartNumber) Swap(i, j int)      { t[i], t[j] = t[j], t[i] }
+func (t byObjectPartNumber) Less(i, j int) bool { return t[i].Number < t[j].Number }
+
+// AddChecksumInfo adds a checksum of a part.
+func (e *ErasureInfo) AddChecksumInfo(ckSumInfo ChecksumInfo) {
+	for i, sum := range e.Checksums {
+		if sum.PartNumber == ckSumInfo.PartNumber {
+			e.Checksums[i] = ckSumInfo
+			return
+		}
+	}
+	e.Checksums = append(e.Checksums, ckSumInfo)
+}
+
+// GetChecksumInfo - get checksum of a part.
+func (e ErasureInfo) GetChecksumInfo(partNumber int) (ckSum ChecksumInfo) {
+	for _, sum := range e.Checksums {
+		if sum.PartNumber == partNumber {
+			// Return the checksum
+			return sum
+		}
+	}
+	return ChecksumInfo{}
+}
+
+// ShardFileSize - returns final erasure size from original size.
+func (e ErasureInfo) ShardFileSize(totalLength int64) int64 {
+	if totalLength == 0 {
+		return 0
+	}
+	if totalLength == -1 {
+		return -1
+	}
+	numShards := totalLength / e.BlockSize
+	lastBlockSize := totalLength % e.BlockSize
+	lastShardSize := ceilFrac(lastBlockSize, int64(e.DataBlocks))
+	return numShards*e.ShardSize() + lastShardSize
+}
+
+// ShardSize - returns actual shared size from erasure blockSize.
+func (e ErasureInfo) ShardSize() int64 {
+	return ceilFrac(e.BlockSize, int64(e.DataBlocks))
+}
+
+// IsValid - tells if erasure info fields are valid.
+func (fi FileInfo) IsValid() bool {
+	if fi.Deleted {
+		// Delete marker has no data, no need to check
+		// for erasure coding information
+		return true
+	}
+	data := fi.Erasure.DataBlocks
+	parity := fi.Erasure.ParityBlocks
+	return ((data >= parity) && (data != 0) && (parity != 0))
+}
+
+// ToObjectInfo - Converts metadata to object info.
+func (fi FileInfo) ToObjectInfo(bucket, object string) ObjectInfo {
+	if HasSuffix(object, SlashSeparator) {
+		return ObjectInfo{
+			Bucket: bucket,
+			Name:   object,
+			IsDir:  true,
+		}
+	}
+	objInfo := ObjectInfo{
+		IsDir:           false,
+		Bucket:          bucket,
+		Name:            object,
+		VersionID:       fi.VersionID,
+		IsLatest:        fi.IsLatest,
+		DeleteMarker:    fi.Deleted,
+		Size:            fi.Size,
+		ModTime:         fi.ModTime,
+		ContentType:     fi.Metadata["content-type"],
+		ContentEncoding: fi.Metadata["content-encoding"],
+	}
+	// Update expires
+	var (
+		t time.Time
+		e error
+	)
+	if exp, ok := fi.Metadata["expires"]; ok {
+		if t, e = time.Parse(http.TimeFormat, exp); e == nil {
+			objInfo.Expires = t.UTC()
+		}
+	}
+	objInfo.backendType = BackendErasure
+
+	// Extract etag from metadata.
+	objInfo.ETag = extractETag(fi.Metadata)
+
+	// Add user tags to the object info
+	objInfo.UserTags = fi.Metadata[xhttp.AmzObjectTagging]
+
+	// etag/md5Sum has already been extracted. We need to
+	// remove to avoid it from appearing as part of
+	// response headers. e.g, X-Minio-* or X-Amz-*.
+	// Tags have also been extracted, we remove that as well.
+	objInfo.UserDefined = cleanMetadata(fi.Metadata)
+
+	// All the parts per object.
+	objInfo.Parts = fi.Parts
+
+	// Update storage class
+	if sc, ok := fi.Metadata[xhttp.AmzStorageClass]; ok {
+		objInfo.StorageClass = sc
+	} else {
+		objInfo.StorageClass = globalMinioDefaultStorageClass
+	}
+
+	// Success.
+	return objInfo
+}
+
+// objectPartIndex - returns the index of matching object part number.
+func objectPartIndex(parts []ObjectPartInfo, partNumber int) int {
+	for i, part := range parts {
+		if partNumber == part.Number {
+			return i
+		}
+	}
+	return -1
+}
+
+// AddObjectPart - add a new object part in order.
+func (fi *FileInfo) AddObjectPart(partNumber int, partETag string, partSize int64, actualSize int64) {
+	partInfo := ObjectPartInfo{
+		Number:     partNumber,
+		ETag:       partETag,
+		Size:       partSize,
+		ActualSize: actualSize,
+	}
+
+	// Update part info if it already exists.
+	for i, part := range fi.Parts {
+		if partNumber == part.Number {
+			fi.Parts[i] = partInfo
+			return
+		}
+	}
+
+	// Proceed to include new part info.
+	fi.Parts = append(fi.Parts, partInfo)
+
+	// Parts in FileInfo should be in sorted order by part number.
+	sort.Sort(byObjectPartNumber(fi.Parts))
+}
+
+// ObjectToPartOffset - translate offset of an object to offset of its individual part.
+func (fi FileInfo) ObjectToPartOffset(ctx context.Context, offset int64) (partIndex int, partOffset int64, err error) {
+	if offset == 0 {
+		// Special case - if offset is 0, then partIndex and partOffset are always 0.
+		return 0, 0, nil
+	}
+	partOffset = offset
+	// Seek until object offset maps to a particular part offset.
+	for i, part := range fi.Parts {
+		partIndex = i
+		// Offset is smaller than size we have reached the proper part offset.
+		if partOffset < part.Size {
+			return partIndex, partOffset, nil
+		}
+		// Continue to towards the next part.
+		partOffset -= part.Size
+	}
+	logger.LogIf(ctx, InvalidRange{})
+	// Offset beyond the size of the object return InvalidRange.
+	return 0, 0, InvalidRange{}
+}
+
+func findFileInfoInQuorum(ctx context.Context, metaArr []FileInfo, modTime time.Time, quorum int) (xmv FileInfo, e error) {
+	metaHashes := make([]string, len(metaArr))
+	for i, meta := range metaArr {
+		if meta.IsValid() && meta.ModTime.Equal(modTime) {
+			h := sha256.New()
+			for _, part := range meta.Parts {
+				h.Write([]byte(fmt.Sprintf("part.%d", part.Number)))
+			}
+			metaHashes[i] = hex.EncodeToString(h.Sum(nil))
+		}
+	}
+
+	metaHashCountMap := make(map[string]int)
+	for _, hash := range metaHashes {
+		if hash == "" {
+			continue
+		}
+		metaHashCountMap[hash]++
+	}
+
+	maxHash := ""
+	maxCount := 0
+	for hash, count := range metaHashCountMap {
+		if count > maxCount {
+			maxCount = count
+			maxHash = hash
+		}
+	}
+
+	if maxCount < quorum {
+		return FileInfo{}, errErasureReadQuorum
+	}
+
+	for i, hash := range metaHashes {
+		if hash == maxHash {
+			return metaArr[i], nil
+		}
+	}
+
+	return FileInfo{}, errErasureReadQuorum
+}
+
+// pickValidFileInfo - picks one valid FileInfo content and returns from a
+// slice of FileInfo.
+func pickValidFileInfo(ctx context.Context, metaArr []FileInfo, modTime time.Time, quorum int) (xmv FileInfo, e error) {
+	return findFileInfoInQuorum(ctx, metaArr, modTime, quorum)
+}
+
+// Rename metadata content to destination location for each disk concurrently.
+func renameFileInfo(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, dstBucket, dstEntry string, quorum int) ([]StorageAPI, error) {
+	ignoredErr := []error{errFileNotFound}
+
+	g := errgroup.WithNErrs(len(disks))
+
+	// Rename file on all underlying storage disks.
+	for index := range disks {
+		index := index
+		g.Go(func() error {
+			if disks[index] == nil {
+				return errDiskNotFound
+			}
+			if err := disks[index].RenameData(srcBucket, srcEntry, "", dstBucket, dstEntry); err != nil {
+				if !IsErrIgnored(err, ignoredErr...) {
+					return err
+				}
+			}
+			return nil
+		}, index)
+	}
+
+	// Wait for all renames to finish.
+	errs := g.Wait()
+
+	// We can safely allow RenameData errors up to len(er.getDisks()) - writeQuorum
+	// otherwise return failure. Cleanup successful renames.
+	err := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, quorum)
+	return evalDisks(disks, errs), err
+}
+
+// writeUniqueFileInfo - writes unique `xl.meta` content for each disk concurrently.
+func writeUniqueFileInfo(ctx context.Context, disks []StorageAPI, bucket, prefix string, files []FileInfo, quorum int) ([]StorageAPI, error) {
+	g := errgroup.WithNErrs(len(disks))
+
+	// Start writing `xl.meta` to all disks in parallel.
+	for index := range disks {
+		index := index
+		g.Go(func() error {
+			if disks[index] == nil {
+				return errDiskNotFound
+			}
+			// Pick one FileInfo for a disk at index.
+			files[index].Erasure.Index = index + 1
+			return disks[index].WriteMetadata(bucket, prefix, files[index])
+		}, index)
+	}
+
+	// Wait for all the routines.
+	mErrs := g.Wait()
+
+	err := reduceWriteQuorumErrs(ctx, mErrs, objectOpIgnoredErrs, quorum)
+	return evalDisks(disks, mErrs), err
+}
+
+// Returns per object readQuorum and writeQuorum
+// readQuorum is the min required disks to read data.
+// writeQuorum is the min required disks to write data.
+func objectQuorumFromMeta(ctx context.Context, er erasureObjects, partsMetaData []FileInfo, errs []error) (objectReadQuorum, objectWriteQuorum int, err error) {
+	// get the latest updated Metadata and a count of all the latest updated FileInfo(s)
+	latestFileInfo, err := getLatestFileInfo(ctx, partsMetaData, errs)
+	if err != nil {
+		return 0, 0, err
+	}
+
+	// Since all the valid erasure code meta updated at the same time are equivalent, pass dataBlocks
+	// from latestFileInfo to get the quorum
+	return latestFileInfo.Erasure.DataBlocks, latestFileInfo.Erasure.DataBlocks + 1, nil
+}

cmd/erasure-metadata_test.go

@@ -0,0 +1,153 @@
+/*
+ * MinIO Cloud Storage, (C) 2015, 2016, 2017 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"context"
+	"strconv"
+	"testing"
+
+	humanize "github.com/dustin/go-humanize"
+)
+
+const ActualSize = 1000
+
+// Test FileInfo.AddObjectPart()
+func TestAddObjectPart(t *testing.T) {
+	testCases := []struct {
+		partNum       int
+		expectedIndex int
+	}{
+		{1, 0},
+		{2, 1},
+		{4, 2},
+		{5, 3},
+		{7, 4},
+		// Insert part.
+		{3, 2},
+		// Replace existing part.
+		{4, 3},
+		// Missing part.
+		{6, -1},
+	}
+
+	// Setup.
+	fi := newFileInfo("test-object", 8, 8)
+	if !fi.IsValid() {
+		t.Fatalf("unable to get xl meta")
+	}
+
+	// Test them.
+	for _, testCase := range testCases {
+		if testCase.expectedIndex > -1 {
+			partNumString := strconv.Itoa(testCase.partNum)
+			fi.AddObjectPart(testCase.partNum, "etag."+partNumString, int64(testCase.partNum+humanize.MiByte), ActualSize)
+		}
+
+		if index := objectPartIndex(fi.Parts, testCase.partNum); index != testCase.expectedIndex {
+			t.Fatalf("%+v: expected = %d, got: %d", testCase, testCase.expectedIndex, index)
+		}
+	}
+}
+
+// Test objectPartIndex(). generates a sample FileInfo data and asserts
+// the output of objectPartIndex() with the expected value.
+func TestObjectPartIndex(t *testing.T) {
+	testCases := []struct {
+		partNum       int
+		expectedIndex int
+	}{
+		{2, 1},
+		{1, 0},
+		{5, 3},
+		{4, 2},
+		{7, 4},
+	}
+
+	// Setup.
+	fi := newFileInfo("test-object", 8, 8)
+	if !fi.IsValid() {
+		t.Fatalf("unable to get xl meta")
+	}
+
+	// Add some parts for testing.
+	for _, testCase := range testCases {
+		partNumString := strconv.Itoa(testCase.partNum)
+		fi.AddObjectPart(testCase.partNum, "etag."+partNumString, int64(testCase.partNum+humanize.MiByte), ActualSize)
+	}
+
+	// Add failure test case.
+	testCases = append(testCases, struct {
+		partNum       int
+		expectedIndex int
+	}{6, -1})
+
+	// Test them.
+	for _, testCase := range testCases {
+		if index := objectPartIndex(fi.Parts, testCase.partNum); index != testCase.expectedIndex {
+			t.Fatalf("%+v: expected = %d, got: %d", testCase, testCase.expectedIndex, index)
+		}
+	}
+}
+
+// Test FileInfo.ObjectToPartOffset().
+func TestObjectToPartOffset(t *testing.T) {
+	// Setup.
+	fi := newFileInfo("test-object", 8, 8)
+	if !fi.IsValid() {
+		t.Fatalf("unable to get xl meta")
+	}
+
+	// Add some parts for testing.
+	// Total size of all parts is 5,242,899 bytes.
+	for _, partNum := range []int{1, 2, 4, 5, 7} {
+		partNumString := strconv.Itoa(partNum)
+		fi.AddObjectPart(partNum, "etag."+partNumString, int64(partNum+humanize.MiByte), ActualSize)
+	}
+
+	testCases := []struct {
+		offset         int64
+		expectedIndex  int
+		expectedOffset int64
+		expectedErr    error
+	}{
+		{0, 0, 0, nil},
+		{1 * humanize.MiByte, 0, 1 * humanize.MiByte, nil},
+		{1 + humanize.MiByte, 1, 0, nil},
+		{2 + humanize.MiByte, 1, 1, nil},
+		// Its valid for zero sized object.
+		{-1, 0, -1, nil},
+		// Max fffset is always (size - 1).
+		{(1 + 2 + 4 + 5 + 7) + (5 * humanize.MiByte) - 1, 4, 1048582, nil},
+		// Error if offset is size.
+		{(1 + 2 + 4 + 5 + 7) + (5 * humanize.MiByte), 0, 0, InvalidRange{}},
+	}
+
+	// Test them.
+	for _, testCase := range testCases {
+		index, offset, err := fi.ObjectToPartOffset(context.Background(), testCase.offset)
+		if err != testCase.expectedErr {
+			t.Fatalf("%+v: expected = %s, got: %s", testCase, testCase.expectedErr, err)
+		}
+		if index != testCase.expectedIndex {
+			t.Fatalf("%+v: index: expected = %d, got: %d", testCase, testCase.expectedIndex, index)
+		}
+		if offset != testCase.expectedOffset {
+			t.Fatalf("%+v: offset: expected = %d, got: %d", testCase, testCase.expectedOffset, offset)
+		}
+	}
+}

cmd/erasure-multipart.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017, 2018 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -24,7 +24,6 @@ import (
 	"sort"
 	"strconv"
 	"strings"
-	"time"
 
 	xhttp "github.com/minio/minio/cmd/http"
 	"github.com/minio/minio/cmd/logger"
@@ -32,24 +31,25 @@ import (
 	"github.com/minio/minio/pkg/sync/errgroup"
 )
 
-func (xl xlObjects) getUploadIDDir(bucket, object, uploadID string) string {
+func (er erasureObjects) getUploadIDDir(bucket, object, uploadID string) string {
-	return pathJoin(xl.getMultipartSHADir(bucket, object), uploadID)
+	return pathJoin(er.getMultipartSHADir(bucket, object), uploadID)
 }
 
-func (xl xlObjects) getMultipartSHADir(bucket, object string) string {
+func (er erasureObjects) getMultipartSHADir(bucket, object string) string {
 	return getSHA256Hash([]byte(pathJoin(bucket, object)))
 }
 
 // checkUploadIDExists - verify if a given uploadID exists and is valid.
-func (xl xlObjects) checkUploadIDExists(ctx context.Context, bucket, object, uploadID string) error {
+func (er erasureObjects) checkUploadIDExists(ctx context.Context, bucket, object, uploadID string) error {
-	_, err := xl.getObjectInfo(ctx, minioMetaMultipartBucket, xl.getUploadIDDir(bucket, object, uploadID), ObjectOptions{})
+	_, err := er.getObjectInfo(ctx, minioMetaMultipartBucket, er.getUploadIDDir(bucket, object, uploadID), ObjectOptions{})
 	return err
 }
 
 // Removes part given by partName belonging to a mulitpart upload from minioMetaBucket
-func (xl xlObjects) removeObjectPart(bucket, object, uploadID string, partNumber int) {
+func (er erasureObjects) removeObjectPart(bucket, object, uploadID, dataDir string, partNumber int) {
-	curpartPath := pathJoin(xl.getUploadIDDir(bucket, object, uploadID), fmt.Sprintf("part.%d", partNumber))
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
-	storageDisks := xl.getDisks()
+	curpartPath := pathJoin(uploadIDPath, dataDir, fmt.Sprintf("part.%d", partNumber))
+	storageDisks := er.getDisks()
 
 	g := errgroup.WithNErrs(len(storageDisks))
 	for index, disk := range storageDisks {
@@ -59,7 +59,7 @@ func (xl xlObjects) removeObjectPart(bucket, object, uploadID string, partNumber
 		index := index
 		g.Go(func() error {
 			// Ignoring failure to remove parts that weren't present in CompleteMultipartUpload
-			// requests. xl.json is the authoritative source of truth on which parts constitute
+			// requests. xl.meta is the authoritative source of truth on which parts constitute
 			// the object. The presence of parts that don't belong in the object doesn't affect correctness.
 			_ = storageDisks[index].DeleteFile(minioMetaMultipartBucket, curpartPath)
 			return nil
@@ -68,36 +68,6 @@ func (xl xlObjects) removeObjectPart(bucket, object, uploadID string, partNumber
 	g.Wait()
 }
 
-// commitXLMetadata - commit `xl.json` from source prefix to destination prefix in the given slice of disks.
-func commitXLMetadata(ctx context.Context, disks []StorageAPI, srcBucket, srcPrefix, dstBucket, dstPrefix string, quorum int) ([]StorageAPI, error) {
-	srcJSONFile := path.Join(srcPrefix, xlMetaJSONFile)
-	dstJSONFile := path.Join(dstPrefix, xlMetaJSONFile)
-
-	g := errgroup.WithNErrs(len(disks))
-
-	// Rename `xl.json` to all disks in parallel.
-	for index := range disks {
-		index := index
-		g.Go(func() error {
-			if disks[index] == nil {
-				return errDiskNotFound
-			}
-
-			// Delete any dangling directories.
-			defer disks[index].DeleteFile(srcBucket, srcPrefix)
-
-			// Renames `xl.json` from source prefix to destination prefix.
-			return disks[index].RenameFile(srcBucket, srcJSONFile, dstBucket, dstJSONFile)
-		}, index)
-	}
-
-	// Wait for all the routines.
-	mErrs := g.Wait()
-
-	err := reduceWriteQuorumErrs(ctx, mErrs, objectOpIgnoredErrs, quorum)
-	return evalDisks(disks, mErrs), err
-}
-
 // ListMultipartUploads - lists all the pending multipart
 // uploads for a particular object in a bucket.
 //
@@ -105,17 +75,17 @@ func commitXLMetadata(ctx context.Context, disks []StorageAPI, srcBucket, srcPre
 // not support prefix based listing, this is a deliberate attempt
 // towards simplification of multipart APIs.
 // The resulting ListMultipartsInfo structure is unmarshalled directly as XML.
-func (xl xlObjects) ListMultipartUploads(ctx context.Context, bucket, object, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, e error) {
+func (er erasureObjects) ListMultipartUploads(ctx context.Context, bucket, object, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, e error) {
 	result.MaxUploads = maxUploads
 	result.KeyMarker = keyMarker
 	result.Prefix = object
 	result.Delimiter = delimiter
 
-	for _, disk := range xl.getLoadBalancedDisks() {
+	for _, disk := range er.getLoadBalancedDisks() {
 		if disk == nil {
 			continue
 		}
-		uploadIDs, err := disk.ListDir(minioMetaMultipartBucket, xl.getMultipartSHADir(bucket, object), -1, "")
+		uploadIDs, err := disk.ListDir(minioMetaMultipartBucket, er.getMultipartSHADir(bucket, object), -1)
 		if err != nil {
 			if err == errFileNotFound {
 				return result, nil
@@ -147,16 +117,16 @@ func (xl xlObjects) ListMultipartUploads(ctx context.Context, bucket, object, ke
 // '.minio.sys/multipart/bucket/object/uploads.json' on all the
 // disks. `uploads.json` carries metadata regarding on-going multipart
 // operation(s) on the object.
-func (xl xlObjects) newMultipartUpload(ctx context.Context, bucket string, object string, meta map[string]string) (string, error) {
+func (er erasureObjects) newMultipartUpload(ctx context.Context, bucket string, object string, opts ObjectOptions) (string, error) {
 
-	onlineDisks := xl.getDisks()
+	onlineDisks := er.getDisks()
-	parityBlocks := globalStorageClass.GetParityForSC(meta[xhttp.AmzStorageClass])
+	parityBlocks := globalStorageClass.GetParityForSC(opts.UserDefined[xhttp.AmzStorageClass])
 	if parityBlocks == 0 {
 		parityBlocks = len(onlineDisks) / 2
 	}
 	dataBlocks := len(onlineDisks) - parityBlocks
 
-	xlMeta := newXLMetaV1(object, dataBlocks, parityBlocks)
+	fi := newFileInfo(object, dataBlocks, parityBlocks)
 
 	// we now know the number of blocks this object needs for data and parity.
 	// establish the writeQuorum using this data
@@ -165,30 +135,37 @@ func (xl xlObjects) newMultipartUpload(ctx context.Context, bucket string, objec
 		writeQuorum = dataBlocks + 1
 	}
 
-	if meta["content-type"] == "" {
+	if opts.UserDefined["content-type"] == "" {
 		contentType := mimedb.TypeByExtension(path.Ext(object))
-		meta["content-type"] = contentType
+		opts.UserDefined["content-type"] = contentType
 	}
-	xlMeta.Stat.ModTime = UTCNow()
+
-	xlMeta.Meta = meta
+	// Calculate the version to be saved.
+	if opts.Versioned {
+		fi.VersionID = mustGetUUID()
+	}
+
+	fi.DataDir = mustGetUUID()
+	fi.ModTime = UTCNow()
+	fi.Metadata = opts.UserDefined
 
 	uploadID := mustGetUUID()
-	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 	tempUploadIDPath := uploadID
 
 	// Delete the tmp path later in case we fail to commit (ignore
 	// returned errors) - this will be a no-op in case of a commit
 	// success.
-	defer xl.deleteObject(ctx, minioMetaTmpBucket, tempUploadIDPath, writeQuorum, false)
+	defer er.deleteObject(ctx, minioMetaTmpBucket, tempUploadIDPath, writeQuorum)
 
-	var partsMetadata = make([]xlMetaV1, len(onlineDisks))
+	var partsMetadata = make([]FileInfo, len(onlineDisks))
 	for i := range onlineDisks {
-		partsMetadata[i] = xlMeta
+		partsMetadata[i] = fi
 	}
 
 	var err error
-	// Write updated `xl.json` to all disks.
+	// Write updated `xl.meta` to all disks.
-	onlineDisks, err = writeUniqueXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempUploadIDPath, partsMetadata, writeQuorum)
+	onlineDisks, err = writeUniqueFileInfo(ctx, onlineDisks, minioMetaTmpBucket, tempUploadIDPath, partsMetadata, writeQuorum)
 	if err != nil {
 		return "", toObjectErr(err, minioMetaTmpBucket, tempUploadIDPath)
 	}
@@ -208,12 +185,12 @@ func (xl xlObjects) newMultipartUpload(ctx context.Context, bucket string, objec
 // subsequent request each UUID is unique.
 //
 // Implements S3 compatible initiate multipart API.
-func (xl xlObjects) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (string, error) {
+func (er erasureObjects) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (string, error) {
 	// No metadata is set, allocate a new one.
 	if opts.UserDefined == nil {
 		opts.UserDefined = make(map[string]string)
 	}
-	return xl.newMultipartUpload(ctx, bucket, object, opts.UserDefined)
+	return er.newMultipartUpload(ctx, bucket, object, opts)
 }
 
 // CopyObjectPart - reads incoming stream and internally erasure codes
@@ -221,8 +198,8 @@ func (xl xlObjects) NewMultipartUpload(ctx context.Context, bucket, object strin
 // data is read from an existing object.
 //
 // Implements S3 compatible Upload Part Copy API.
-func (xl xlObjects) CopyObjectPart(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject, uploadID string, partID int, startOffset int64, length int64, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (pi PartInfo, e error) {
+func (er erasureObjects) CopyObjectPart(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject, uploadID string, partID int, startOffset int64, length int64, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (pi PartInfo, e error) {
-	partInfo, err := xl.PutObjectPart(ctx, dstBucket, dstObject, uploadID, partID, NewPutObjReader(srcInfo.Reader, nil, nil), dstOpts)
+	partInfo, err := er.PutObjectPart(ctx, dstBucket, dstObject, uploadID, partID, NewPutObjReader(srcInfo.Reader, nil, nil), dstOpts)
 	if err != nil {
 		return pi, toObjectErr(err, dstBucket, dstObject)
 	}
@@ -236,64 +213,60 @@ func (xl xlObjects) CopyObjectPart(ctx context.Context, srcBucket, srcObject, ds
 // of the multipart transaction.
 //
 // Implements S3 compatible Upload Part API.
-func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID string, partID int, r *PutObjReader, opts ObjectOptions) (pi PartInfo, e error) {
+func (er erasureObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID string, partID int, r *PutObjReader, opts ObjectOptions) (pi PartInfo, e error) {
 	data := r.Reader
-	if err := checkPutObjectPartArgs(ctx, bucket, object, xl); err != nil {
-		return pi, err
-	}
-
 	// Validate input data size and it can never be less than zero.
 	if data.Size() < -1 {
 		logger.LogIf(ctx, errInvalidArgument, logger.Application)
 		return pi, toObjectErr(errInvalidArgument)
 	}
 
-	var partsMetadata []xlMetaV1
+	var partsMetadata []FileInfo
 	var errs []error
-	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 
 	// Validates if upload ID exists.
-	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+	if err := er.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		return pi, toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Read metadata associated with the object from all disks.
-	partsMetadata, errs = readAllXLMetadata(ctx, xl.getDisks(), minioMetaMultipartBucket,
+	partsMetadata, errs = readAllFileInfo(ctx, er.getDisks(), minioMetaMultipartBucket,
-		uploadIDPath)
+		uploadIDPath, "")
 
 	// get Quorum for this object
-	_, writeQuorum, err := objectQuorumFromMeta(ctx, xl, partsMetadata, errs)
+	_, writeQuorum, err := objectQuorumFromMeta(ctx, er, partsMetadata, errs)
 	if err != nil {
 		return pi, toObjectErr(err, bucket, object)
 	}
 
 	reducedErr := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-	if reducedErr == errXLWriteQuorum {
+	if reducedErr == errErasureWriteQuorum {
 		return pi, toObjectErr(reducedErr, bucket, object)
 	}
 
 	// List all online disks.
-	onlineDisks, modTime := listOnlineDisks(xl.getDisks(), partsMetadata, errs)
+	onlineDisks, modTime := listOnlineDisks(er.getDisks(), partsMetadata, errs)
 
 	// Pick one from the first valid metadata.
-	xlMeta, err := pickValidXLMeta(ctx, partsMetadata, modTime, writeQuorum)
+	fi, err := pickValidFileInfo(ctx, partsMetadata, modTime, writeQuorum)
 	if err != nil {
 		return pi, err
 	}
 
-	onlineDisks = shuffleDisks(onlineDisks, xlMeta.Erasure.Distribution)
+	onlineDisks = shuffleDisks(onlineDisks, fi.Erasure.Distribution)
 
 	// Need a unique name for the part being written in minioMetaBucket to
 	// accommodate concurrent PutObjectPart requests
 
 	partSuffix := fmt.Sprintf("part.%d", partID)
 	tmpPart := mustGetUUID()
-	tmpPartPath := path.Join(tmpPart, partSuffix)
+	tmpPartPath := pathJoin(tmpPart, partSuffix)
 
 	// Delete the temporary object part. If PutObjectPart succeeds there would be nothing to delete.
-	defer xl.deleteObject(ctx, minioMetaTmpBucket, tmpPart, writeQuorum, false)
+	defer er.deleteObject(ctx, minioMetaTmpBucket, tmpPart, writeQuorum)
 
-	erasure, err := NewErasure(ctx, xlMeta.Erasure.DataBlocks, xlMeta.Erasure.ParityBlocks, xlMeta.Erasure.BlockSize)
+	erasure, err := NewErasure(ctx, fi.Erasure.DataBlocks, fi.Erasure.ParityBlocks, fi.Erasure.BlockSize)
 	if err != nil {
 		return pi, toObjectErr(err, bucket, object)
 	}
@@ -303,16 +276,16 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	switch size := data.Size(); {
 	case size == 0:
 		buffer = make([]byte, 1) // Allocate atleast a byte to reach EOF
-	case size == -1 || size >= blockSizeV1:
+	case size == -1 || size >= fi.Erasure.BlockSize:
-		buffer = xl.bp.Get()
+		buffer = er.bp.Get()
-		defer xl.bp.Put(buffer)
+		defer er.bp.Put(buffer)
-	case size < blockSizeV1:
+	case size < fi.Erasure.BlockSize:
-		// No need to allocate fully blockSizeV1 buffer if the incoming data is smaller.
+		// No need to allocate fully fi.Erasure.BlockSize buffer if the incoming data is smaller.
-		buffer = make([]byte, size, 2*size+int64(erasure.parityBlocks+erasure.dataBlocks-1))
+		buffer = make([]byte, size, 2*size+int64(fi.Erasure.ParityBlocks+fi.Erasure.DataBlocks-1))
 	}
 
-	if len(buffer) > int(xlMeta.Erasure.BlockSize) {
+	if len(buffer) > int(fi.Erasure.BlockSize) {
-		buffer = buffer[:xlMeta.Erasure.BlockSize]
+		buffer = buffer[:fi.Erasure.BlockSize]
 	}
 	writers := make([]io.Writer, len(onlineDisks))
 	for i, disk := range onlineDisks {
@@ -322,7 +295,7 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 		writers[i] = newBitrotWriter(disk, minioMetaTmpBucket, tmpPartPath, erasure.ShardFileSize(data.Size()), DefaultBitrotAlgorithm, erasure.ShardSize())
 	}
 
-	n, err := erasure.Encode(ctx, data, writers, buffer, erasure.dataBlocks+1)
+	n, err := erasure.Encode(ctx, data, writers, buffer, fi.Erasure.DataBlocks+1)
 	closeBitrotWriters(writers)
 	if err != nil {
 		return pi, toObjectErr(err, bucket, object)
@@ -341,21 +314,21 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	}
 
 	// Validates if upload ID exists.
-	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+	if err := er.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		return pi, toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Rename temporary part file to its final location.
-	partPath := path.Join(uploadIDPath, partSuffix)
+	partPath := pathJoin(uploadIDPath, fi.DataDir, partSuffix)
 	onlineDisks, err = rename(ctx, onlineDisks, minioMetaTmpBucket, tmpPartPath, minioMetaMultipartBucket, partPath, false, writeQuorum, nil)
 	if err != nil {
 		return pi, toObjectErr(err, minioMetaMultipartBucket, partPath)
 	}
 
 	// Read metadata again because it might be updated with parallel upload of another part.
-	partsMetadata, errs = readAllXLMetadata(ctx, onlineDisks, minioMetaMultipartBucket, uploadIDPath)
+	partsMetadata, errs = readAllFileInfo(ctx, onlineDisks, minioMetaMultipartBucket, uploadIDPath, "")
 	reducedErr = reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-	if reducedErr == errXLWriteQuorum {
+	if reducedErr == errErasureWriteQuorum {
 		return pi, toObjectErr(reducedErr, bucket, object)
 	}
 
@@ -363,25 +336,26 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	onlineDisks, modTime = listOnlineDisks(onlineDisks, partsMetadata, errs)
 
 	// Pick one from the first valid metadata.
-	xlMeta, err = pickValidXLMeta(ctx, partsMetadata, modTime, writeQuorum)
+	fi, err = pickValidFileInfo(ctx, partsMetadata, modTime, writeQuorum)
 	if err != nil {
 		return pi, err
 	}
 
-	// Once part is successfully committed, proceed with updating XL metadata.
+	// Once part is successfully committed, proceed with updating erasure metadata.
-	xlMeta.Stat.ModTime = UTCNow()
+	fi.ModTime = UTCNow()
 
 	md5hex := r.MD5CurrentHexString()
 
 	// Add the current part.
-	xlMeta.AddObjectPart(partID, md5hex, n, data.ActualSize())
+	fi.AddObjectPart(partID, md5hex, n, data.ActualSize())
 
 	for i, disk := range onlineDisks {
 		if disk == OfflineDisk {
 			continue
 		}
-		partsMetadata[i].Stat = xlMeta.Stat
+		partsMetadata[i].Size = fi.Size
-		partsMetadata[i].Parts = xlMeta.Parts
+		partsMetadata[i].ModTime = fi.ModTime
+		partsMetadata[i].Parts = fi.Parts
 		partsMetadata[i].Erasure.AddChecksumInfo(ChecksumInfo{
 			PartNumber: partID,
 			Algorithm:  DefaultBitrotAlgorithm,
@@ -389,19 +363,8 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 		})
 	}
 
-	// Write all the checksum metadata.
+	// Writes update `xl.meta` format for each disk.
-	tempXLMetaPath := mustGetUUID()
+	if _, err = writeUniqueFileInfo(ctx, onlineDisks, minioMetaMultipartBucket, uploadIDPath, partsMetadata, writeQuorum); err != nil {
-
-	// Cleanup in case of xl.json writing failure
-	defer xl.deleteObject(ctx, minioMetaTmpBucket, tempXLMetaPath, writeQuorum, false)
-
-	// Writes a unique `xl.json` each disk carrying new checksum related information.
-	onlineDisks, err = writeUniqueXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempXLMetaPath, partsMetadata, writeQuorum)
-	if err != nil {
-		return pi, toObjectErr(err, minioMetaTmpBucket, tempXLMetaPath)
-	}
-
-	if _, err = commitXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempXLMetaPath, minioMetaMultipartBucket, uploadIDPath, writeQuorum); err != nil {
 		return pi, toObjectErr(err, minioMetaMultipartBucket, uploadIDPath)
 	}
 
@@ -409,8 +372,8 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	return PartInfo{
 		PartNumber:   partID,
 		ETag:         md5hex,
-		LastModified: xlMeta.Stat.ModTime,
+		LastModified: fi.ModTime,
-		Size:         xlMeta.Stat.Size,
+		Size:         fi.Size,
 		ActualSize:   data.ActualSize(),
 	}, nil
 }
@@ -419,44 +382,44 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 // by callers to verify object states
 // - encrypted
 // - compressed
-func (xl xlObjects) GetMultipartInfo(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (MultipartInfo, error) {
+func (er erasureObjects) GetMultipartInfo(ctx context.Context, bucket, object, uploadID string, opts ObjectOptions) (MultipartInfo, error) {
 	result := MultipartInfo{
 		Bucket:   bucket,
 		Object:   object,
 		UploadID: uploadID,
 	}
 
-	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+	if err := er.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		return result, toObjectErr(err, bucket, object, uploadID)
 	}
 
-	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
 
 	// Read metadata associated with the object from all disks.
-	partsMetadata, errs := readAllXLMetadata(ctx, storageDisks, minioMetaMultipartBucket, uploadIDPath)
+	partsMetadata, errs := readAllFileInfo(ctx, storageDisks, minioMetaMultipartBucket, uploadIDPath, opts.VersionID)
 
 	// get Quorum for this object
-	_, writeQuorum, err := objectQuorumFromMeta(ctx, xl, partsMetadata, errs)
+	readQuorum, _, err := objectQuorumFromMeta(ctx, er, partsMetadata, errs)
 	if err != nil {
 		return result, toObjectErr(err, minioMetaMultipartBucket, uploadIDPath)
 	}
 
-	reducedErr := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
+	reducedErr := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum)
-	if reducedErr == errXLWriteQuorum {
+	if reducedErr == errErasureReadQuorum {
 		return result, toObjectErr(reducedErr, minioMetaMultipartBucket, uploadIDPath)
 	}
 
 	_, modTime := listOnlineDisks(storageDisks, partsMetadata, errs)
 
 	// Pick one from the first valid metadata.
-	xlMeta, err := pickValidXLMeta(ctx, partsMetadata, modTime, writeQuorum)
+	fi, err := pickValidFileInfo(ctx, partsMetadata, modTime, readQuorum)
 	if err != nil {
 		return result, err
 	}
 
-	result.UserDefined = xlMeta.Meta
+	result.UserDefined = fi.Metadata
 	return result, nil
 }
 
@@ -467,51 +430,47 @@ func (xl xlObjects) GetMultipartInfo(ctx context.Context, bucket, object, upload
 // Implements S3 compatible ListObjectParts API. The resulting
 // ListPartsInfo structure is marshaled directly into XML and
 // replied back to the client.
-func (xl xlObjects) ListObjectParts(ctx context.Context, bucket, object, uploadID string, partNumberMarker, maxParts int, opts ObjectOptions) (result ListPartsInfo, e error) {
+func (er erasureObjects) ListObjectParts(ctx context.Context, bucket, object, uploadID string, partNumberMarker, maxParts int, opts ObjectOptions) (result ListPartsInfo, e error) {
-
+	if err := er.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
-	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		return result, toObjectErr(err, bucket, object, uploadID)
 	}
 
-	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
 
 	// Read metadata associated with the object from all disks.
-	partsMetadata, errs := readAllXLMetadata(ctx, storageDisks, minioMetaMultipartBucket, uploadIDPath)
+	partsMetadata, errs := readAllFileInfo(ctx, storageDisks, minioMetaMultipartBucket, uploadIDPath, "")
 
 	// get Quorum for this object
-	_, writeQuorum, err := objectQuorumFromMeta(ctx, xl, partsMetadata, errs)
+	_, writeQuorum, err := objectQuorumFromMeta(ctx, er, partsMetadata, errs)
 	if err != nil {
 		return result, toObjectErr(err, minioMetaMultipartBucket, uploadIDPath)
 	}
 
 	reducedErr := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-	if reducedErr == errXLWriteQuorum {
+	if reducedErr == errErasureWriteQuorum {
 		return result, toObjectErr(reducedErr, minioMetaMultipartBucket, uploadIDPath)
 	}
 
 	_, modTime := listOnlineDisks(storageDisks, partsMetadata, errs)
 
 	// Pick one from the first valid metadata.
-	xlValidMeta, err := pickValidXLMeta(ctx, partsMetadata, modTime, writeQuorum)
+	fi, err := pickValidFileInfo(ctx, partsMetadata, modTime, writeQuorum)
 	if err != nil {
 		return result, err
 	}
 
-	var xlMeta = xlValidMeta.Meta
-	var xlParts = xlValidMeta.Parts
-
 	// Populate the result stub.
 	result.Bucket = bucket
 	result.Object = object
 	result.UploadID = uploadID
 	result.MaxParts = maxParts
 	result.PartNumberMarker = partNumberMarker
-	result.UserDefined = xlMeta
+	result.UserDefined = fi.Metadata
 
 	// For empty number of parts or maxParts as zero, return right here.
-	if len(xlParts) == 0 || maxParts == 0 {
+	if len(fi.Parts) == 0 || maxParts == 0 {
 		return result, nil
 	}
 
@@ -521,17 +480,17 @@ func (xl xlObjects) ListObjectParts(ctx context.Context, bucket, object, uploadI
 	}
 
 	// Only parts with higher part numbers will be listed.
-	partIdx := objectPartIndex(xlParts, partNumberMarker)
+	partIdx := objectPartIndex(fi.Parts, partNumberMarker)
-	parts := xlParts
+	parts := fi.Parts
 	if partIdx != -1 {
-		parts = xlParts[partIdx+1:]
+		parts = fi.Parts[partIdx+1:]
 	}
 	count := maxParts
 	for _, part := range parts {
 		result.Parts = append(result.Parts, PartInfo{
 			PartNumber:   part.Number,
 			ETag:         part.ETag,
-			LastModified: xlValidMeta.Stat.ModTime,
+			LastModified: fi.ModTime,
 			Size:         part.Size,
 		})
 		count--
@@ -556,14 +515,14 @@ func (xl xlObjects) ListObjectParts(ctx context.Context, bucket, object, uploadI
 // md5sums of all the parts.
 //
 // Implements S3 compatible Complete multipart API.
-func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string, object string, uploadID string, parts []CompletePart, opts ObjectOptions) (oi ObjectInfo, e error) {
+func (er erasureObjects) CompleteMultipartUpload(ctx context.Context, bucket string, object string, uploadID string, parts []CompletePart, opts ObjectOptions) (oi ObjectInfo, e error) {
-	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+	if err := er.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		return oi, toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Check if an object is present as one of the parent dir.
 	// -- FIXME. (needs a new kind of lock).
-	if xl.parentDirIsObject(ctx, bucket, path.Dir(object)) {
+	if er.parentDirIsObject(ctx, bucket, path.Dir(object)) {
 		return oi, toObjectErr(errFileParentIsFile, bucket, object)
 	}
 
@@ -572,21 +531,21 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 	// Calculate s3 compatible md5sum for complete multipart.
 	s3MD5 := getCompleteMultipartMD5(parts)
 
-	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 
-	storageDisks := xl.getDisks()
+	storageDisks := er.getDisks()
 
 	// Read metadata associated with the object from all disks.
-	partsMetadata, errs := readAllXLMetadata(ctx, storageDisks, minioMetaMultipartBucket, uploadIDPath)
+	partsMetadata, errs := readAllFileInfo(ctx, storageDisks, minioMetaMultipartBucket, uploadIDPath, "")
 
 	// get Quorum for this object
-	_, writeQuorum, err := objectQuorumFromMeta(ctx, xl, partsMetadata, errs)
+	_, writeQuorum, err := objectQuorumFromMeta(ctx, er, partsMetadata, errs)
 	if err != nil {
 		return oi, toObjectErr(err, bucket, object)
 	}
 
 	reducedErr := reduceWriteQuorumErrs(ctx, errs, objectOpIgnoredErrs, writeQuorum)
-	if reducedErr == errXLWriteQuorum {
+	if reducedErr == errErasureWriteQuorum {
 		return oi, toObjectErr(reducedErr, bucket, object)
 	}
 
@@ -599,28 +558,26 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 	var objectActualSize int64
 
 	// Pick one from the first valid metadata.
-	xlMeta, err := pickValidXLMeta(ctx, partsMetadata, modTime, writeQuorum)
+	fi, err := pickValidFileInfo(ctx, partsMetadata, modTime, writeQuorum)
 	if err != nil {
 		return oi, err
 	}
 
 	// Order online disks in accordance with distribution order.
-	onlineDisks = shuffleDisks(onlineDisks, xlMeta.Erasure.Distribution)
+	onlineDisks = shuffleDisks(onlineDisks, fi.Erasure.Distribution)
 
 	// Order parts metadata in accordance with distribution order.
-	partsMetadata = shufflePartsMetadata(partsMetadata, xlMeta.Erasure.Distribution)
+	partsMetadata = shufflePartsMetadata(partsMetadata, fi.Erasure.Distribution)
 
-	// Save current xl meta for validation.
+	// Save current erasure metadata for validation.
-	var currentXLMeta = xlMeta
+	var currentFI = fi
 
 	// Allocate parts similar to incoming slice.
-	xlMeta.Parts = make([]ObjectPartInfo, len(parts))
+	fi.Parts = make([]ObjectPartInfo, len(parts))
 
 	// Validate each part and then commit to disk.
 	for i, part := range parts {
-		// ensure that part ETag is canonicalized to strip off extraneous quotes
+		partIdx := objectPartIndex(currentFI.Parts, part.PartNumber)
-		part.ETag = canonicalizeETag(part.ETag)
-		partIdx := objectPartIndex(currentXLMeta.Parts, part.PartNumber)
 		// All parts should have same part number.
 		if partIdx == -1 {
 			invp := InvalidPart{
@@ -630,116 +587,103 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 			return oi, invp
 		}
 
-		if currentXLMeta.Parts[partIdx].ETag != part.ETag {
+		// ensure that part ETag is canonicalized to strip off extraneous quotes
+		part.ETag = canonicalizeETag(part.ETag)
+		if currentFI.Parts[partIdx].ETag != part.ETag {
 			invp := InvalidPart{
 				PartNumber: part.PartNumber,
-				ExpETag:    currentXLMeta.Parts[partIdx].ETag,
+				ExpETag:    currentFI.Parts[partIdx].ETag,
 				GotETag:    part.ETag,
 			}
 			return oi, invp
 		}
 
 		// All parts except the last part has to be atleast 5MB.
-		if (i < len(parts)-1) && !isMinAllowedPartSize(currentXLMeta.Parts[partIdx].ActualSize) {
+		if (i < len(parts)-1) && !isMinAllowedPartSize(currentFI.Parts[partIdx].ActualSize) {
 			return oi, PartTooSmall{
 				PartNumber: part.PartNumber,
-				PartSize:   currentXLMeta.Parts[partIdx].ActualSize,
+				PartSize:   currentFI.Parts[partIdx].ActualSize,
 				PartETag:   part.ETag,
 			}
 		}
 
 		// Save for total object size.
-		objectSize += currentXLMeta.Parts[partIdx].Size
+		objectSize += currentFI.Parts[partIdx].Size
 
 		// Save the consolidated actual size.
-		objectActualSize += currentXLMeta.Parts[partIdx].ActualSize
+		objectActualSize += currentFI.Parts[partIdx].ActualSize
 
 		// Add incoming parts.
-		xlMeta.Parts[i] = ObjectPartInfo{
+		fi.Parts[i] = ObjectPartInfo{
 			Number:     part.PartNumber,
-			Size:       currentXLMeta.Parts[partIdx].Size,
+			Size:       currentFI.Parts[partIdx].Size,
-			ActualSize: currentXLMeta.Parts[partIdx].ActualSize,
+			ActualSize: currentFI.Parts[partIdx].ActualSize,
 		}
 	}
 
 	// Save the final object size and modtime.
-	xlMeta.Stat.Size = objectSize
+	fi.Size = objectSize
-	xlMeta.Stat.ModTime = UTCNow()
+	fi.ModTime = UTCNow()
 
 	// Save successfully calculated md5sum.
-	xlMeta.Meta["etag"] = s3MD5
+	fi.Metadata["etag"] = s3MD5
 
 	// Save the consolidated actual size.
-	xlMeta.Meta[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(objectActualSize, 10)
+	fi.Metadata[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(objectActualSize, 10)
 
-	// Update all xl metadata, make sure to not modify fields like
+	// Update all erasure metadata, make sure to not modify fields like
 	// checksum which are different on each disks.
 	for index := range partsMetadata {
-		partsMetadata[index].Stat = xlMeta.Stat
+		partsMetadata[index].Size = fi.Size
-		partsMetadata[index].Meta = xlMeta.Meta
+		partsMetadata[index].ModTime = fi.ModTime
-		partsMetadata[index].Parts = xlMeta.Parts
+		partsMetadata[index].Metadata = fi.Metadata
+		partsMetadata[index].Parts = fi.Parts
 	}
 
-	tempXLMetaPath := mustGetUUID()
+	// Write final `xl.meta` at uploadID location
-
+	if onlineDisks, err = writeUniqueFileInfo(ctx, onlineDisks, minioMetaMultipartBucket, uploadIDPath, partsMetadata, writeQuorum); err != nil {
-	// Cleanup in case of failure
+		return oi, toObjectErr(err, minioMetaMultipartBucket, uploadIDPath)
-	defer xl.deleteObject(ctx, minioMetaTmpBucket, tempXLMetaPath, writeQuorum, false)
-
-	// Write unique `xl.json` for each disk.
-	if onlineDisks, err = writeUniqueXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempXLMetaPath, partsMetadata, writeQuorum); err != nil {
-		return oi, toObjectErr(err, minioMetaTmpBucket, tempXLMetaPath)
-	}
-
-	var rErr error
-	onlineDisks, rErr = commitXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempXLMetaPath, minioMetaMultipartBucket, uploadIDPath, writeQuorum)
-	if rErr != nil {
-		return oi, toObjectErr(rErr, minioMetaMultipartBucket, uploadIDPath)
-	}
-
-	if xl.isObject(bucket, object) {
-		// Rename if an object already exists to temporary location.
-		newUniqueID := mustGetUUID()
-
-		// Delete success renamed object.
-		defer xl.deleteObject(ctx, minioMetaTmpBucket, newUniqueID, writeQuorum, false)
-
-		// NOTE: Do not use online disks slice here: the reason is that existing object should be purged
-		// regardless of `xl.json` status and rolled back in case of errors. Also allow renaming of the
-		// existing object if it is not present in quorum disks so users can overwrite stale objects.
-		_, err = rename(ctx, xl.getDisks(), bucket, object, minioMetaTmpBucket, newUniqueID, true, writeQuorum, []error{errFileNotFound})
-		if err != nil {
-			return oi, toObjectErr(err, bucket, object)
-		}
 	}
 
 	// Remove parts that weren't present in CompleteMultipartUpload request.
-	for _, curpart := range currentXLMeta.Parts {
+	for _, curpart := range currentFI.Parts {
-		if objectPartIndex(xlMeta.Parts, curpart.Number) == -1 {
+		if objectPartIndex(fi.Parts, curpart.Number) == -1 {
 			// Delete the missing part files. e.g,
 			// Request 1: NewMultipart
 			// Request 2: PutObjectPart 1
 			// Request 3: PutObjectPart 2
 			// Request 4: CompleteMultipartUpload --part 2
 			// N.B. 1st part is not present. This part should be removed from the storage.
-			xl.removeObjectPart(bucket, object, uploadID, curpart.Number)
+			er.removeObjectPart(bucket, object, uploadID, fi.DataDir, curpart.Number)
 		}
 	}
 
 	// Rename the multipart object to final location.
-	if onlineDisks, err = rename(ctx, onlineDisks, minioMetaMultipartBucket, uploadIDPath, bucket, object, true, writeQuorum, nil); err != nil {
+	if onlineDisks, err = renameData(ctx, onlineDisks, minioMetaMultipartBucket, uploadIDPath,
+		fi.DataDir, bucket, object, writeQuorum, nil); err != nil {
 		return oi, toObjectErr(err, bucket, object)
 	}
 
 	// Check if there is any offline disk and add it to the MRF list
 	for i := 0; i < len(onlineDisks); i++ {
 		if onlineDisks[i] == nil || storageDisks[i] == nil {
-			xl.addPartialUpload(bucket, object)
+			er.addPartialUpload(bucket, object)
 			break
 		}
 	}
 
+	for i := 0; i < len(onlineDisks); i++ {
+		if onlineDisks[i] == nil {
+			continue
+		}
+		// Object info is the same in all disks, so we can pick
+		// the first meta from online disk
+		fi = partsMetadata[i]
+		break
+	}
+
 	// Success, return object info.
-	return xlMeta.ToObjectInfo(bucket, object), nil
+	return fi.ToObjectInfo(bucket, object), nil
 }
 
 // AbortMultipartUpload - aborts an ongoing multipart operation
@@ -753,79 +697,28 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 // Implements S3 compatible Abort multipart API, slight difference is
 // that this is an atomic idempotent operation. Subsequent calls have
 // no affect and further requests to the same uploadID would not be honored.
-func (xl xlObjects) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string) error {
+func (er erasureObjects) AbortMultipartUpload(ctx context.Context, bucket, object, uploadID string) error {
 	// Validates if upload ID exists.
-	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+	if err := er.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		return toObjectErr(err, bucket, object, uploadID)
 	}
 
-	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
+	uploadIDPath := er.getUploadIDDir(bucket, object, uploadID)
 
 	// Read metadata associated with the object from all disks.
-	partsMetadata, errs := readAllXLMetadata(ctx, xl.getDisks(), minioMetaMultipartBucket, uploadIDPath)
+	partsMetadata, errs := readAllFileInfo(ctx, er.getDisks(), minioMetaMultipartBucket, uploadIDPath, "")
 
 	// get Quorum for this object
-	_, writeQuorum, err := objectQuorumFromMeta(ctx, xl, partsMetadata, errs)
+	_, writeQuorum, err := objectQuorumFromMeta(ctx, er, partsMetadata, errs)
 	if err != nil {
 		return toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Cleanup all uploaded parts.
-	if err = xl.deleteObject(ctx, minioMetaMultipartBucket, uploadIDPath, writeQuorum, false); err != nil {
+	if err = er.deleteObject(ctx, minioMetaMultipartBucket, uploadIDPath, writeQuorum); err != nil {
 		return toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Successfully purged.
 	return nil
 }
-
-// Clean-up the old multipart uploads. Should be run in a Go routine.
-func (xl xlObjects) cleanupStaleMultipartUploads(ctx context.Context, cleanupInterval, expiry time.Duration, doneCh <-chan struct{}) {
-	ticker := time.NewTicker(cleanupInterval)
-	defer ticker.Stop()
-
-	for {
-		select {
-		case <-doneCh:
-			return
-		case <-ticker.C:
-			var disk StorageAPI
-			for _, d := range xl.getLoadBalancedDisks() {
-				if d != nil {
-					disk = d
-					break
-				}
-			}
-			if disk == nil {
-				continue
-			}
-			xl.cleanupStaleMultipartUploadsOnDisk(ctx, disk, expiry)
-		}
-	}
-}
-
-// Remove the old multipart uploads on the given disk.
-func (xl xlObjects) cleanupStaleMultipartUploadsOnDisk(ctx context.Context, disk StorageAPI, expiry time.Duration) {
-	now := time.Now()
-	shaDirs, err := disk.ListDir(minioMetaMultipartBucket, "", -1, "")
-	if err != nil {
-		return
-	}
-	for _, shaDir := range shaDirs {
-		uploadIDDirs, err := disk.ListDir(minioMetaMultipartBucket, shaDir, -1, "")
-		if err != nil {
-			continue
-		}
-		for _, uploadIDDir := range uploadIDDirs {
-			uploadIDPath := pathJoin(shaDir, uploadIDDir)
-			fi, err := disk.StatFile(minioMetaMultipartBucket, pathJoin(uploadIDPath, xlMetaJSONFile))
-			if err != nil {
-				continue
-			}
-			if now.Sub(fi.ModTime) > expiry {
-				writeQuorum := getWriteQuorum(len(xl.getDisks()))
-				xl.deleteObject(ctx, minioMetaMultipartBucket, uploadIDPath, writeQuorum, false)
-			}
-		}
-	}
-}

cmd/erasure-object_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -20,16 +20,11 @@ import (
 	"bytes"
 	"context"
 	"io/ioutil"
-	"math/rand"
 	"os"
-	"path"
-	"reflect"
 	"testing"
-	"time"
 
 	humanize "github.com/dustin/go-humanize"
 	"github.com/minio/minio/cmd/config/storageclass"
-	"github.com/minio/minio/pkg/madmin"
 )
 
 func TestRepeatPutObjectPart(t *testing.T) {
@@ -41,7 +36,7 @@ func TestRepeatPutObjectPart(t *testing.T) {
 	var err error
 	var opts ObjectOptions
 
-	objLayer, disks, err = prepareXL16(ctx)
+	objLayer, disks, err = prepareErasure16(ctx)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -49,7 +44,7 @@ func TestRepeatPutObjectPart(t *testing.T) {
 	// cleaning up of temporary test directories
 	defer removeRoots(disks)
 
-	err = objLayer.MakeBucketWithLocation(ctx, "bucket1", "", false)
+	err = objLayer.MakeBucketWithLocation(ctx, "bucket1", BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -71,7 +66,7 @@ func TestRepeatPutObjectPart(t *testing.T) {
 	}
 }
 
-func TestXLDeleteObjectBasic(t *testing.T) {
+func TestErasureDeleteObjectBasic(t *testing.T) {
 	testCases := []struct {
 		bucket      string
 		object      string
@@ -91,12 +86,12 @@ func TestXLDeleteObjectBasic(t *testing.T) {
 	defer cancel()
 
 	// Create an instance of xl backend
-	xl, fsDirs, err := prepareXL16(ctx)
+	xl, fsDirs, err := prepareErasure16(ctx)
 	if err != nil {
 		t.Fatal(err)
 	}
 
-	err = xl.MakeBucketWithLocation(ctx, "bucket", "", false)
+	err = xl.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -104,40 +99,43 @@ func TestXLDeleteObjectBasic(t *testing.T) {
 	// Create object "dir/obj" under bucket "bucket" for Test 7 to pass
 	_, err = xl.PutObject(ctx, "bucket", "dir/obj", mustGetPutObjReader(t, bytes.NewReader([]byte("abcd")), int64(len("abcd")), "", ""), ObjectOptions{})
 	if err != nil {
-		t.Fatalf("XL Object upload failed: <ERROR> %s", err)
+		t.Fatalf("Erasure Object upload failed: <ERROR> %s", err)
 	}
-	for i, test := range testCases {
+	for _, test := range testCases {
-		actualErr := xl.DeleteObject(ctx, test.bucket, test.object)
+		test := test
+		t.Run("", func(t *testing.T) {
+			_, actualErr := xl.DeleteObject(ctx, test.bucket, test.object, ObjectOptions{})
 			if test.expectedErr != nil && actualErr != test.expectedErr {
-			t.Errorf("Test %d: Expected to fail with %s, but failed with %s", i+1, test.expectedErr, actualErr)
+				t.Errorf("Expected to fail with %s, but failed with %s", test.expectedErr, actualErr)
 			}
 			if test.expectedErr == nil && actualErr != nil {
-			t.Errorf("Test %d: Expected to pass, but failed with %s", i+1, actualErr)
+				t.Errorf("Expected to pass, but failed with %s", actualErr)
 			}
+		})
 	}
 	// Cleanup backend directories
 	removeRoots(fsDirs)
 }
 
-func TestXLDeleteObjectsXLSet(t *testing.T) {
+func TestErasureDeleteObjectsErasureSet(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
-	var objs []*xlObjects
+	var objs []*erasureObjects
 	for i := 0; i < 32; i++ {
-		obj, fsDirs, err := prepareXL(ctx, 16)
+		obj, fsDirs, err := prepareErasure(ctx, 16)
 		if err != nil {
-			t.Fatal("Unable to initialize 'XL' object layer.", err)
+			t.Fatal("Unable to initialize 'Erasure' object layer.", err)
 		}
 		// Remove all dirs.
 		for _, dir := range fsDirs {
 			defer os.RemoveAll(dir)
 		}
-		z := obj.(*xlZones)
+		z := obj.(*erasureZones)
 		xl := z.zones[0].sets[0]
 		objs = append(objs, xl)
 	}
 
-	xlSets := &xlSets{sets: objs, distributionAlgo: "CRCMOD"}
+	erasureSets := &erasureSets{sets: objs, distributionAlgo: "CRCMOD"}
 
 	type testCaseType struct {
 		bucket string
@@ -152,32 +150,29 @@ func TestXLDeleteObjectsXLSet(t *testing.T) {
 		{bucketName, "obj_4"},
 	}
 
-	err := xlSets.MakeBucketWithLocation(GlobalContext, bucketName, "", false)
+	err := erasureSets.MakeBucketWithLocation(ctx, bucketName, BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
 
 	for _, testCase := range testCases {
-		_, err = xlSets.PutObject(GlobalContext, testCase.bucket, testCase.object,
+		_, err = erasureSets.PutObject(ctx, testCase.bucket, testCase.object,
 			mustGetPutObjReader(t, bytes.NewReader([]byte("abcd")), int64(len("abcd")), "", ""), ObjectOptions{})
 		if err != nil {
-			t.Fatalf("XL Object upload failed: <ERROR> %s", err)
+			t.Fatalf("Erasure Object upload failed: <ERROR> %s", err)
 		}
 	}
 
-	toObjectNames := func(testCases []testCaseType) []string {
+	toObjectNames := func(testCases []testCaseType) []ObjectToDelete {
-		names := make([]string, len(testCases))
+		names := make([]ObjectToDelete, len(testCases))
 		for i := range testCases {
-			names[i] = testCases[i].object
+			names[i] = ObjectToDelete{ObjectName: testCases[i].object}
 		}
 		return names
 	}
 
 	objectNames := toObjectNames(testCases)
-	delErrs, err := xlSets.DeleteObjects(GlobalContext, bucketName, objectNames)
+	_, delErrs := erasureSets.DeleteObjects(ctx, bucketName, objectNames, ObjectOptions{})
-	if err != nil {
-		t.Errorf("Failed to call DeleteObjects with the error: `%v`", err)
-	}
 
 	for i := range delErrs {
 		if delErrs[i] != nil {
@@ -186,7 +181,7 @@ func TestXLDeleteObjectsXLSet(t *testing.T) {
 	}
 
 	for _, test := range testCases {
-		_, statErr := xlSets.GetObjectInfo(GlobalContext, test.bucket, test.object, ObjectOptions{})
+		_, statErr := erasureSets.GetObjectInfo(ctx, test.bucket, test.object, ObjectOptions{})
 		switch statErr.(type) {
 		case ObjectNotFound:
 		default:
@@ -195,23 +190,23 @@ func TestXLDeleteObjectsXLSet(t *testing.T) {
 	}
 }
 
-func TestXLDeleteObjectDiskNotFound(t *testing.T) {
+func TestErasureDeleteObjectDiskNotFound(t *testing.T) {
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
 
 	// Create an instance of xl backend.
-	obj, fsDirs, err := prepareXL16(ctx)
+	obj, fsDirs, err := prepareErasure16(ctx)
 	if err != nil {
 		t.Fatal(err)
 	}
 	// Cleanup backend directories
 	defer removeRoots(fsDirs)
 
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
 	xl := z.zones[0].sets[0]
 
 	// Create "bucket"
-	err = obj.MakeBucketWithLocation(ctx, "bucket", "", false)
+	err = obj.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -226,16 +221,17 @@ func TestXLDeleteObjectDiskNotFound(t *testing.T) {
 	}
 	// for a 16 disk setup, quorum is 9. To simulate disks not found yet
 	// quorum is available, we remove disks leaving quorum disks behind.
-	xlDisks := xl.getDisks()
+	erasureDisks := xl.getDisks()
-	z.zones[0].xlDisksMu.Lock()
+	z.zones[0].erasureDisksMu.Lock()
 	xl.getDisks = func() []StorageAPI {
-		for i := range xlDisks[:7] {
+		for i := range erasureDisks[:7] {
-			xlDisks[i] = newNaughtyDisk(xlDisks[i], nil, errFaultyDisk)
+			erasureDisks[i] = newNaughtyDisk(erasureDisks[i], nil, errFaultyDisk)
 		}
-		return xlDisks
+		return erasureDisks
 	}
-	z.zones[0].xlDisksMu.Unlock()
+
-	err = obj.DeleteObject(ctx, bucket, object)
+	z.zones[0].erasureDisksMu.Unlock()
+	_, err = obj.DeleteObject(ctx, bucket, object, ObjectOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -247,18 +243,19 @@ func TestXLDeleteObjectDiskNotFound(t *testing.T) {
 	}
 
 	// Remove one more disk to 'lose' quorum, by setting it to nil.
-	xlDisks = xl.getDisks()
+	erasureDisks = xl.getDisks()
-	z.zones[0].xlDisksMu.Lock()
+	z.zones[0].erasureDisksMu.Lock()
 	xl.getDisks = func() []StorageAPI {
-		xlDisks[7] = nil
+		erasureDisks[7] = nil
-		xlDisks[8] = nil
+		erasureDisks[8] = nil
-		return xlDisks
+		return erasureDisks
 	}
-	z.zones[0].xlDisksMu.Unlock()
+
-	err = obj.DeleteObject(ctx, bucket, object)
+	z.zones[0].erasureDisksMu.Unlock()
-	// since majority of disks are not available, metaquorum is not achieved and hence errXLReadQuorum error
+	_, err = obj.DeleteObject(ctx, bucket, object, ObjectOptions{})
-	if err != toObjectErr(errXLReadQuorum, bucket, object) {
+	// since majority of disks are not available, metaquorum is not achieved and hence errErasureWriteQuorum error
-		t.Errorf("Expected deleteObject to fail with %v, but failed with %v", toObjectErr(errXLReadQuorum, bucket, object), err)
+	if err != toObjectErr(errErasureWriteQuorum, bucket, object) {
+		t.Errorf("Expected deleteObject to fail with %v, but failed with %v", toObjectErr(errErasureWriteQuorum, bucket, object), err)
 	}
 }
 
@@ -267,18 +264,18 @@ func TestGetObjectNoQuorum(t *testing.T) {
 	defer cancel()
 
 	// Create an instance of xl backend.
-	obj, fsDirs, err := prepareXL16(ctx)
+	obj, fsDirs, err := prepareErasure16(ctx)
 	if err != nil {
 		t.Fatal(err)
 	}
 	// Cleanup backend directories.
 	defer removeRoots(fsDirs)
 
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
 	xl := z.zones[0].sets[0]
 
 	// Create "bucket"
-	err = obj.MakeBucketWithLocation(ctx, "bucket", "", false)
+	err = obj.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -293,7 +290,7 @@ func TestGetObjectNoQuorum(t *testing.T) {
 	}
 
 	// Make 9 disks offline, which leaves less than quorum number of disks
-	// in a 16 disk XL setup. The original disks are 'replaced' with
+	// in a 16 disk Erasure setup. The original disks are 'replaced' with
 	// naughtyDisks that fail after 'f' successful StorageAPI method
 	// invocations, where f - [0,2)
 	for f := 0; f < 2; f++ {
@@ -301,24 +298,24 @@ func TestGetObjectNoQuorum(t *testing.T) {
 		for i := 0; i <= f; i++ {
 			diskErrors[i] = nil
 		}
-		xlDisks := xl.getDisks()
+		erasureDisks := xl.getDisks()
-		for i := range xlDisks[:9] {
+		for i := range erasureDisks[:9] {
-			switch diskType := xlDisks[i].(type) {
+			switch diskType := erasureDisks[i].(type) {
 			case *naughtyDisk:
-				xlDisks[i] = newNaughtyDisk(diskType.disk, diskErrors, errFaultyDisk)
+				erasureDisks[i] = newNaughtyDisk(diskType.disk, diskErrors, errFaultyDisk)
 			default:
-				xlDisks[i] = newNaughtyDisk(xlDisks[i], diskErrors, errFaultyDisk)
+				erasureDisks[i] = newNaughtyDisk(erasureDisks[i], diskErrors, errFaultyDisk)
 			}
 		}
-		z.zones[0].xlDisksMu.Lock()
+		z.zones[0].erasureDisksMu.Lock()
 		xl.getDisks = func() []StorageAPI {
-			return xlDisks
+			return erasureDisks
 		}
-		z.zones[0].xlDisksMu.Unlock()
+		z.zones[0].erasureDisksMu.Unlock()
 		// Fetch object from store.
 		err = xl.GetObject(ctx, bucket, object, 0, int64(len("abcd")), ioutil.Discard, "", opts)
-		if err != toObjectErr(errXLReadQuorum, bucket, object) {
+		if err != toObjectErr(errErasureReadQuorum, bucket, object) {
-			t.Errorf("Expected putObject to fail with %v, but failed with %v", toObjectErr(errXLWriteQuorum, bucket, object), err)
+			t.Errorf("Expected putObject to fail with %v, but failed with %v", toObjectErr(errErasureWriteQuorum, bucket, object), err)
 		}
 	}
 }
@@ -328,7 +325,7 @@ func TestPutObjectNoQuorum(t *testing.T) {
 	defer cancel()
 
 	// Create an instance of xl backend.
-	obj, fsDirs, err := prepareXL16(ctx)
+	obj, fsDirs, err := prepareErasure16(ctx)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -336,11 +333,11 @@ func TestPutObjectNoQuorum(t *testing.T) {
 	// Cleanup backend directories.
 	defer removeRoots(fsDirs)
 
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
 	xl := z.zones[0].sets[0]
 
 	// Create "bucket"
-	err = obj.MakeBucketWithLocation(ctx, "bucket", "", false)
+	err = obj.MakeBucketWithLocation(ctx, "bucket", BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -355,7 +352,7 @@ func TestPutObjectNoQuorum(t *testing.T) {
 	}
 
 	// Make 9 disks offline, which leaves less than quorum number of disks
-	// in a 16 disk XL setup. The original disks are 'replaced' with
+	// in a 16 disk Erasure setup. The original disks are 'replaced' with
 	// naughtyDisks that fail after 'f' successful StorageAPI method
 	// invocations, where f - [0,3)
 	for f := 0; f < 3; f++ {
@@ -363,143 +360,38 @@ func TestPutObjectNoQuorum(t *testing.T) {
 		for i := 0; i <= f; i++ {
 			diskErrors[i] = nil
 		}
-		xlDisks := xl.getDisks()
+		erasureDisks := xl.getDisks()
-		for i := range xlDisks[:9] {
+		for i := range erasureDisks[:9] {
-			switch diskType := xlDisks[i].(type) {
+			switch diskType := erasureDisks[i].(type) {
 			case *naughtyDisk:
-				xlDisks[i] = newNaughtyDisk(diskType.disk, diskErrors, errFaultyDisk)
+				erasureDisks[i] = newNaughtyDisk(diskType.disk, diskErrors, errFaultyDisk)
 			default:
-				xlDisks[i] = newNaughtyDisk(xlDisks[i], diskErrors, errFaultyDisk)
+				erasureDisks[i] = newNaughtyDisk(erasureDisks[i], diskErrors, errFaultyDisk)
 			}
 		}
-		z.zones[0].xlDisksMu.Lock()
+		z.zones[0].erasureDisksMu.Lock()
 		xl.getDisks = func() []StorageAPI {
-			return xlDisks
+			return erasureDisks
 		}
-		z.zones[0].xlDisksMu.Unlock()
+		z.zones[0].erasureDisksMu.Unlock()
 		// Upload new content to same object "object"
 		_, err = obj.PutObject(ctx, bucket, object, mustGetPutObjReader(t, bytes.NewReader([]byte("abcd")), int64(len("abcd")), "", ""), opts)
-		if err != toObjectErr(errXLWriteQuorum, bucket, object) {
+		if err != toObjectErr(errErasureWriteQuorum, bucket, object) {
-			t.Errorf("Expected putObject to fail with %v, but failed with %v", toObjectErr(errXLWriteQuorum, bucket, object), err)
+			t.Errorf("Expected putObject to fail with %v, but failed with %v", toObjectErr(errErasureWriteQuorum, bucket, object), err)
 		}
 	}
 }
 
-// Tests both object and bucket healing.
-func TestHealing(t *testing.T) {
-	ctx, cancel := context.WithCancel(context.Background())
-	defer cancel()
-
-	obj, fsDirs, err := prepareXL16(ctx)
-	if err != nil {
-		t.Fatal(err)
-	}
-	defer removeRoots(fsDirs)
-
-	z := obj.(*xlZones)
-	xl := z.zones[0].sets[0]
-
-	// Create "bucket"
-	err = obj.MakeBucketWithLocation(ctx, "bucket", "", false)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	bucket := "bucket"
-	object := "object"
-
-	data := make([]byte, 1*humanize.MiByte)
-	length := int64(len(data))
-	_, err = rand.Read(data)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	_, err = obj.PutObject(ctx, bucket, object, mustGetPutObjReader(t, bytes.NewReader(data), length, "", ""), ObjectOptions{})
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	disk := xl.getDisks()[0]
-	xlMetaPreHeal, err := readXLMeta(ctx, disk, bucket, object)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// Remove the object - to simulate the case where the disk was down when the object
-	// was created.
-	err = os.RemoveAll(path.Join(fsDirs[0], bucket, object))
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	_, err = xl.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealNormalScan})
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	xlMetaPostHeal, err := readXLMeta(ctx, disk, bucket, object)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// After heal the meta file should be as expected.
-	if !reflect.DeepEqual(xlMetaPreHeal, xlMetaPostHeal) {
-		t.Fatal("HealObject failed")
-	}
-
-	err = os.RemoveAll(path.Join(fsDirs[0], bucket, object, "xl.json"))
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// Write xl.json with different modtime to simulate the case where a disk had
-	// gone down when an object was replaced by a new object.
-	xlMetaOutDated := xlMetaPreHeal
-	xlMetaOutDated.Stat.ModTime = time.Now()
-	err = writeXLMetadata(ctx, disk, bucket, object, xlMetaOutDated)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	_, err = xl.HealObject(ctx, bucket, object, madmin.HealOpts{ScanMode: madmin.HealDeepScan})
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	xlMetaPostHeal, err = readXLMeta(ctx, disk, bucket, object)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	// After heal the meta file should be as expected.
-	if !reflect.DeepEqual(xlMetaPreHeal, xlMetaPostHeal) {
-		t.Fatal("HealObject failed")
-	}
-
-	// Remove the bucket - to simulate the case where bucket was
-	// created when the disk was down.
-	err = os.RemoveAll(path.Join(fsDirs[0], bucket))
-	if err != nil {
-		t.Fatal(err)
-	}
-	// This would create the bucket.
-	_, err = xl.HealBucket(ctx, bucket, false, false)
-	if err != nil {
-		t.Fatal(err)
-	}
-	// Stat the bucket to make sure that it was created.
-	_, err = xl.getDisks()[0].StatVol(bucket)
-	if err != nil {
-		t.Fatal(err)
-	}
-}
-
 func TestObjectQuorumFromMeta(t *testing.T) {
 	ExecObjectLayerTestWithDirs(t, testObjectQuorumFromMeta)
 }
 
 func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []string, t TestErrHandler) {
+	restoreGlobalStorageClass := globalStorageClass
+	defer func() {
+		globalStorageClass = restoreGlobalStorageClass
+	}()
+
 	bucket := getRandomBucketName()
 
 	var opts ObjectOptions
@@ -507,45 +399,48 @@ func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []strin
 	partCount := 3
 	data := bytes.Repeat([]byte("a"), 6*1024*1024*partCount)
 
-	z := obj.(*xlZones)
+	z := obj.(*erasureZones)
 	xl := z.zones[0].sets[0]
-	xlDisks := xl.getDisks()
+	erasureDisks := xl.getDisks()
 
-	err := obj.MakeBucketWithLocation(GlobalContext, bucket, globalMinioDefaultRegion, false)
+	ctx, cancel := context.WithCancel(GlobalContext)
+	defer cancel()
+
+	err := obj.MakeBucketWithLocation(ctx, bucket, BucketOptions{})
 	if err != nil {
 		t.Fatalf("Failed to make a bucket %v", err)
 	}
 
 	// Object for test case 1 - No StorageClass defined, no MetaData in PutObject
 	object1 := "object1"
-	_, err = obj.PutObject(GlobalContext, bucket, object1, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), opts)
+	_, err = obj.PutObject(ctx, bucket, object1, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), opts)
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts1, errs1 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object1)
+	parts1, errs1 := readAllFileInfo(ctx, erasureDisks, bucket, object1, "")
 
 	// Object for test case 2 - No StorageClass defined, MetaData in PutObject requesting RRS Class
 	object2 := "object2"
 	metadata2 := make(map[string]string)
 	metadata2["x-amz-storage-class"] = storageclass.RRS
-	_, err = obj.PutObject(GlobalContext, bucket, object2, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata2})
+	_, err = obj.PutObject(ctx, bucket, object2, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata2})
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts2, errs2 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object2)
+	parts2, errs2 := readAllFileInfo(ctx, erasureDisks, bucket, object2, "")
 
 	// Object for test case 3 - No StorageClass defined, MetaData in PutObject requesting Standard Storage Class
 	object3 := "object3"
 	metadata3 := make(map[string]string)
 	metadata3["x-amz-storage-class"] = storageclass.STANDARD
-	_, err = obj.PutObject(GlobalContext, bucket, object3, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata3})
+	_, err = obj.PutObject(ctx, bucket, object3, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata3})
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts3, errs3 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object3)
+	parts3, errs3 := readAllFileInfo(ctx, erasureDisks, bucket, object3, "")
 
 	// Object for test case 4 - Standard StorageClass defined as Parity 6, MetaData in PutObject requesting Standard Storage Class
 	object4 := "object4"
@@ -557,12 +452,12 @@ func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []strin
 		},
 	}
 
-	_, err = obj.PutObject(GlobalContext, bucket, object4, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata4})
+	_, err = obj.PutObject(ctx, bucket, object4, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata4})
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts4, errs4 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object4)
+	parts4, errs4 := readAllFileInfo(ctx, erasureDisks, bucket, object4, "")
 
 	// Object for test case 5 - RRS StorageClass defined as Parity 2, MetaData in PutObject requesting RRS Class
 	// Reset global storage class flags
@@ -575,12 +470,12 @@ func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []strin
 		},
 	}
 
-	_, err = obj.PutObject(GlobalContext, bucket, object5, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata5})
+	_, err = obj.PutObject(ctx, bucket, object5, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata5})
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts5, errs5 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object5)
+	parts5, errs5 := readAllFileInfo(ctx, erasureDisks, bucket, object5, "")
 
 	// Object for test case 6 - RRS StorageClass defined as Parity 2, MetaData in PutObject requesting Standard Storage Class
 	object6 := "object6"
@@ -592,12 +487,12 @@ func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []strin
 		},
 	}
 
-	_, err = obj.PutObject(GlobalContext, bucket, object6, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata6})
+	_, err = obj.PutObject(ctx, bucket, object6, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata6})
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts6, errs6 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object6)
+	parts6, errs6 := readAllFileInfo(ctx, erasureDisks, bucket, object6, "")
 
 	// Object for test case 7 - Standard StorageClass defined as Parity 5, MetaData in PutObject requesting RRS Class
 	// Reset global storage class flags
@@ -610,15 +505,15 @@ func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []strin
 		},
 	}
 
-	_, err = obj.PutObject(GlobalContext, bucket, object7, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata7})
+	_, err = obj.PutObject(ctx, bucket, object7, mustGetPutObjReader(t, bytes.NewReader(data), int64(len(data)), "", ""), ObjectOptions{UserDefined: metadata7})
 	if err != nil {
 		t.Fatalf("Failed to putObject %v", err)
 	}
 
-	parts7, errs7 := readAllXLMetadata(GlobalContext, xlDisks, bucket, object7)
+	parts7, errs7 := readAllFileInfo(ctx, erasureDisks, bucket, object7, "")
 
 	tests := []struct {
-		parts               []xlMetaV1
+		parts               []FileInfo
 		errs                []error
 		expectedReadQuorum  int
 		expectedWriteQuorum int
@@ -632,23 +527,22 @@ func testObjectQuorumFromMeta(obj ObjectLayer, instanceType string, dirs []strin
 		{parts6, errs6, 8, 9, nil},
 		{parts7, errs7, 14, 15, nil},
 	}
-	for i, tt := range tests {
+	for _, tt := range tests {
-		actualReadQuorum, actualWriteQuorum, err := objectQuorumFromMeta(GlobalContext, *xl, tt.parts, tt.errs)
+		tt := tt
+		t.(*testing.T).Run("", func(t *testing.T) {
+			actualReadQuorum, actualWriteQuorum, err := objectQuorumFromMeta(ctx, *xl, tt.parts, tt.errs)
 			if tt.expectedError != nil && err == nil {
-			t.Errorf("Test %d, Expected %s, got %s", i+1, tt.expectedError, err)
+				t.Errorf("Expected %s, got %s", tt.expectedError, err)
-			return
 			}
 			if tt.expectedError == nil && err != nil {
-			t.Errorf("Test %d, Expected %s, got %s", i+1, tt.expectedError, err)
+				t.Errorf("Expected %s, got %s", tt.expectedError, err)
-			return
 			}
 			if tt.expectedReadQuorum != actualReadQuorum {
-			t.Errorf("Test %d, Expected Read Quorum %d, got %d", i+1, tt.expectedReadQuorum, actualReadQuorum)
+				t.Errorf("Expected Read Quorum %d, got %d", tt.expectedReadQuorum, actualReadQuorum)
-			return
 			}
 			if tt.expectedWriteQuorum != actualWriteQuorum {
-			t.Errorf("Test %d, Expected Write Quorum %d, got %d", i+1, tt.expectedWriteQuorum, actualWriteQuorum)
+				t.Errorf("Expected Write Quorum %d, got %d", tt.expectedWriteQuorum, actualWriteQuorum)
-			return
+			}
-		}
+		})
 	}
 }

cmd/erasure-sets_test.go

@@ -0,0 +1,245 @@
+/*
+ * MinIO Cloud Storage, (C) 2017 MinIO, Inc.
+ *
+ * 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 cmd
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/google/uuid"
+)
+
+var testUUID = uuid.MustParse("f5c58c61-7175-4018-ab5e-a94fe9c2de4e")
+
+func BenchmarkCrcHash(b *testing.B) {
+	cases := []struct {
+		key int
+	}{
+		{16},
+		{64},
+		{128},
+		{256},
+		{512},
+		{1024},
+	}
+	for _, testCase := range cases {
+		testCase := testCase
+		key := randString(testCase.key)
+		b.Run("", func(b *testing.B) {
+			b.SetBytes(1024)
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				crcHashMod(key, 16)
+			}
+		})
+	}
+}
+
+func BenchmarkSipHash(b *testing.B) {
+	cases := []struct {
+		key int
+	}{
+		{16},
+		{64},
+		{128},
+		{256},
+		{512},
+		{1024},
+	}
+	for _, testCase := range cases {
+		testCase := testCase
+		key := randString(testCase.key)
+		b.Run("", func(b *testing.B) {
+			b.SetBytes(1024)
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				sipHashMod(key, 16, testUUID)
+			}
+		})
+	}
+}
+
+// TestSipHashMod - test sip hash.
+func TestSipHashMod(t *testing.T) {
+	testCases := []struct {
+		objectName string
+		sipHash    int
+	}{
+		// cases which should pass the test.
+		// passing in valid object name.
+		{"object", 37},
+		{"The Shining Script <v1>.pdf", 38},
+		{"Cost Benefit Analysis (2009-2010).pptx", 59},
+		{"117Gn8rfHL2ACARPAhaFd0AGzic9pUbIA/5OCn5A", 35},
+		{"SHØRT", 49},
+		{"There are far too many object names, and far too few bucket names!", 8},
+		{"a/b/c/", 159},
+		{"/a/b/c", 96},
+		{string([]byte{0xff, 0xfe, 0xfd}), 147},
+	}
+
+	// Tests hashing order to be consistent.
+	for i, testCase := range testCases {
+		if sipHashElement := hashKey("SIPMOD", testCase.objectName, 200, testUUID); sipHashElement != testCase.sipHash {
+			t.Errorf("Test case %d: Expected \"%v\" but failed \"%v\"", i+1, testCase.sipHash, sipHashElement)
+		}
+	}
+
+	if sipHashElement := hashKey("SIPMOD", "This will fail", -1, testUUID); sipHashElement != -1 {
+		t.Errorf("Test: Expected \"-1\" but got \"%v\"", sipHashElement)
+	}
+
+	if sipHashElement := hashKey("SIPMOD", "This will fail", 0, testUUID); sipHashElement != -1 {
+		t.Errorf("Test: Expected \"-1\" but got \"%v\"", sipHashElement)
+	}
+
+	if sipHashElement := hashKey("UNKNOWN", "This will fail", 0, testUUID); sipHashElement != -1 {
+		t.Errorf("Test: Expected \"-1\" but got \"%v\"", sipHashElement)
+	}
+}
+
+// TestCrcHashMod - test crc hash.
+func TestCrcHashMod(t *testing.T) {
+	testCases := []struct {
+		objectName string
+		crcHash    int
+	}{
+		// cases which should pass the test.
+		// passing in valid object name.
+		{"object", 28},
+		{"The Shining Script <v1>.pdf", 142},
+		{"Cost Benefit Analysis (2009-2010).pptx", 133},
+		{"117Gn8rfHL2ACARPAhaFd0AGzic9pUbIA/5OCn5A", 185},
+		{"SHØRT", 97},
+		{"There are far too many object names, and far too few bucket names!", 101},
+		{"a/b/c/", 193},
+		{"/a/b/c", 116},
+		{string([]byte{0xff, 0xfe, 0xfd}), 61},
+	}
+
+	// Tests hashing order to be consistent.
+	for i, testCase := range testCases {
+		if crcHashElement := hashKey("CRCMOD", testCase.objectName, 200, testUUID); crcHashElement != testCase.crcHash {
+			t.Errorf("Test case %d: Expected \"%v\" but failed \"%v\"", i+1, testCase.crcHash, crcHashElement)
+		}
+	}
+
+	if crcHashElement := hashKey("CRCMOD", "This will fail", -1, testUUID); crcHashElement != -1 {
+		t.Errorf("Test: Expected \"-1\" but got \"%v\"", crcHashElement)
+	}
+
+	if crcHashElement := hashKey("CRCMOD", "This will fail", 0, testUUID); crcHashElement != -1 {
+		t.Errorf("Test: Expected \"-1\" but got \"%v\"", crcHashElement)
+	}
+
+	if crcHashElement := hashKey("UNKNOWN", "This will fail", 0, testUUID); crcHashElement != -1 {
+		t.Errorf("Test: Expected \"-1\" but got \"%v\"", crcHashElement)
+	}
+}
+
+// TestNewErasure - tests initialization of all input disks
+// and constructs a valid `Erasure` object
+func TestNewErasureSets(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	var nDisks = 16 // Maximum disks.
+	var erasureDisks []string
+	for i := 0; i < nDisks; i++ {
+		// Do not attempt to create this path, the test validates
+		// so that newErasureSets initializes non existing paths
+		// and successfully returns initialized object layer.
+		disk := filepath.Join(globalTestTmpDir, "minio-"+nextSuffix())
+		erasureDisks = append(erasureDisks, disk)
+		defer os.RemoveAll(disk)
+	}
+
+	endpoints := mustGetNewEndpoints(erasureDisks...)
+	_, _, err := waitForFormatErasure(true, endpoints, 1, 0, 16, "")
+	if err != errInvalidArgument {
+		t.Fatalf("Expecting error, got %s", err)
+	}
+
+	_, _, err = waitForFormatErasure(true, nil, 1, 1, 16, "")
+	if err != errInvalidArgument {
+		t.Fatalf("Expecting error, got %s", err)
+	}
+
+	// Initializes all erasure disks
+	storageDisks, format, err := waitForFormatErasure(true, endpoints, 1, 1, 16, "")
+	if err != nil {
+		t.Fatalf("Unable to format disks for erasure, %s", err)
+	}
+
+	if _, err := newErasureSets(ctx, endpoints, storageDisks, format); err != nil {
+		t.Fatalf("Unable to initialize erasure")
+	}
+}
+
+// TestHashedLayer - tests the hashed layer which will be returned
+// consistently for a given object name.
+func TestHashedLayer(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	var objs []*erasureObjects
+	for i := 0; i < 16; i++ {
+		obj, fsDirs, err := prepareErasure16(ctx)
+		if err != nil {
+			t.Fatal("Unable to initialize 'Erasure' object layer.", err)
+		}
+
+		// Remove all dirs.
+		for _, dir := range fsDirs {
+			defer os.RemoveAll(dir)
+		}
+
+		z := obj.(*erasureZones)
+		objs = append(objs, z.zones[0].sets[0])
+	}
+
+	sets := &erasureSets{sets: objs, distributionAlgo: "CRCMOD"}
+
+	testCases := []struct {
+		objectName  string
+		expectedObj *erasureObjects
+	}{
+		// cases which should pass the test.
+		// passing in valid object name.
+		{"object", objs[12]},
+		{"The Shining Script <v1>.pdf", objs[14]},
+		{"Cost Benefit Analysis (2009-2010).pptx", objs[13]},
+		{"117Gn8rfHL2ACARPAhaFd0AGzic9pUbIA/5OCn5A", objs[1]},
+		{"SHØRT", objs[9]},
+		{"There are far too many object names, and far too few bucket names!", objs[13]},
+		{"a/b/c/", objs[1]},
+		{"/a/b/c", objs[4]},
+		{string([]byte{0xff, 0xfe, 0xfd}), objs[13]},
+	}
+
+	// Tests hashing order to be consistent.
+	for i, testCase := range testCases {
+		gotObj := sets.getHashedSet(testCase.objectName)
+		if gotObj != testCase.expectedObj {
+			t.Errorf("Test case %d: Expected \"%#v\" but failed \"%#v\"", i+1, testCase.expectedObj, gotObj)
+		}
+	}
+}

cmd/erasure-utils.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

cmd/erasure.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2017 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -18,136 +18,373 @@ package cmd
 
 import (
 	"context"
+	"fmt"
+	"sort"
 	"sync"
+	"time"
 
-	"github.com/klauspost/reedsolomon"
 	"github.com/minio/minio/cmd/logger"
+	"github.com/minio/minio/pkg/bpool"
+	"github.com/minio/minio/pkg/dsync"
+	"github.com/minio/minio/pkg/madmin"
+	"github.com/minio/minio/pkg/sync/errgroup"
 )
 
-// Erasure - erasure encoding details.
+// OfflineDisk represents an unavailable disk.
-type Erasure struct {
+var OfflineDisk StorageAPI // zero value is nil
-	encoder                  func() reedsolomon.Encoder
+
-	dataBlocks, parityBlocks int
+// partialUpload is a successful upload of an object
-	blockSize                int64
+// but not written in all disks (having quorum)
+type partialUpload struct {
+	bucket    string
+	object    string
+	failedSet int
 }
 
-// NewErasure creates a new ErasureStorage.
+// erasureObjects - Implements ER object layer.
-func NewErasure(ctx context.Context, dataBlocks, parityBlocks int, blockSize int64) (e Erasure, err error) {
+type erasureObjects struct {
-	e = Erasure{
+	GatewayUnsupported
-		dataBlocks:   dataBlocks,
+
-		parityBlocks: parityBlocks,
+	// getDisks returns list of storageAPIs.
-		blockSize:    blockSize,
+	getDisks func() []StorageAPI
+
+	// getLockers returns list of remote and local lockers.
+	getLockers func() []dsync.NetLocker
+
+	// getEndpoints returns list of endpoint strings belonging this set.
+	// some may be local and some remote.
+	getEndpoints func() []string
+
+	// Locker mutex map.
+	nsMutex *nsLockMap
+
+	// Byte pools used for temporary i/o buffers.
+	bp *bpool.BytePoolCap
+
+	mrfUploadCh chan partialUpload
 }
 
-	// Check the parameters for sanity now.
+// NewNSLock - initialize a new namespace RWLocker instance.
-	if dataBlocks <= 0 || parityBlocks <= 0 {
+func (er erasureObjects) NewNSLock(ctx context.Context, bucket string, objects ...string) RWLocker {
-		return e, reedsolomon.ErrInvShardNum
+	return er.nsMutex.NewNSLock(ctx, er.getLockers, bucket, objects...)
 }
 
-	if dataBlocks+parityBlocks > 256 {
+// Shutdown function for object storage interface.
-		return e, reedsolomon.ErrMaxShardNum
+func (er erasureObjects) Shutdown(ctx context.Context) error {
-	}
+	// Add any object layer shutdown activities here.
-
+	closeStorageDisks(er.getDisks())
-	// Encoder when needed.
-	var enc reedsolomon.Encoder
-	var once sync.Once
-	e.encoder = func() reedsolomon.Encoder {
-		once.Do(func() {
-			e, err := reedsolomon.New(dataBlocks, parityBlocks, reedsolomon.WithAutoGoroutines(int(e.ShardSize())))
-			if err != nil {
-				// Error conditions should be checked above.
-				panic(err)
-			}
-			enc = e
-		})
-		return enc
-	}
-	return
-}
-
-// EncodeData encodes the given data and returns the erasure-coded data.
-// It returns an error if the erasure coding failed.
-func (e *Erasure) EncodeData(ctx context.Context, data []byte) ([][]byte, error) {
-	if len(data) == 0 {
-		return make([][]byte, e.dataBlocks+e.parityBlocks), nil
-	}
-	encoded, err := e.encoder().Split(data)
-	if err != nil {
-		logger.LogIf(ctx, err)
-		return nil, err
-	}
-	if err = e.encoder().Encode(encoded); err != nil {
-		logger.LogIf(ctx, err)
-		return nil, err
-	}
-	return encoded, nil
-}
-
-// DecodeDataBlocks decodes the given erasure-coded data.
-// It only decodes the data blocks but does not verify them.
-// It returns an error if the decoding failed.
-func (e *Erasure) DecodeDataBlocks(data [][]byte) error {
-	var isZero = 0
-	for _, b := range data[:] {
-		if len(b) == 0 {
-			isZero++
-			break
-		}
-	}
-	if isZero == 0 || isZero == len(data) {
-		// If all are zero, payload is 0 bytes.
 	return nil
 }
-	return e.encoder().ReconstructData(data)
+
+// byDiskTotal is a collection satisfying sort.Interface.
+type byDiskTotal []DiskInfo
+
+func (d byDiskTotal) Len() int      { return len(d) }
+func (d byDiskTotal) Swap(i, j int) { d[i], d[j] = d[j], d[i] }
+func (d byDiskTotal) Less(i, j int) bool {
+	return d[i].Total < d[j].Total
 }
 
-// DecodeDataAndParityBlocks decodes the given erasure-coded data and verifies it.
+// getDisksInfo - fetch disks info across all other storage API.
-// It returns an error if the decoding failed.
+func getDisksInfo(disks []StorageAPI, local bool) (disksInfo []DiskInfo, errs []error, onlineDisks, offlineDisks madmin.BackendDisks) {
-func (e *Erasure) DecodeDataAndParityBlocks(ctx context.Context, data [][]byte) error {
+	disksInfo = make([]DiskInfo, len(disks))
-	needsReconstruction := false
+	onlineDisks = make(madmin.BackendDisks)
-	for _, b := range data {
+	offlineDisks = make(madmin.BackendDisks)
-		if b == nil {
+
-			needsReconstruction = true
+	for _, disk := range disks {
-			break
+		if disk == OfflineDisk {
+			continue
+		}
+		peerAddr := disk.Hostname()
+		if _, ok := offlineDisks[peerAddr]; !ok {
+			offlineDisks[peerAddr] = 0
+		}
+		if _, ok := onlineDisks[peerAddr]; !ok {
+			onlineDisks[peerAddr] = 0
 		}
 	}
-	if !needsReconstruction {
+
-		return nil
+	g := errgroup.WithNErrs(len(disks))
+	for index := range disks {
+		index := index
+		g.Go(func() error {
+			if disks[index] == OfflineDisk {
+				// Storage disk is empty, perhaps ignored disk or not available.
+				return errDiskNotFound
 			}
-	if err := e.encoder().Reconstruct(data); err != nil {
+			info, err := disks[index].DiskInfo()
+			if err != nil {
+				if !IsErr(err, baseErrs...) {
+					reqInfo := (&logger.ReqInfo{}).AppendTags("disk", disks[index].String())
+					ctx := logger.SetReqInfo(GlobalContext, reqInfo)
 					logger.LogIf(ctx, err)
+				}
 				return err
 			}
+			disksInfo[index] = info
+			return nil
+		}, index)
+	}
+
+	errs = g.Wait()
+	// Wait for the routines.
+	for i, diskInfoErr := range errs {
+		if disks[i] == OfflineDisk {
+			continue
+		}
+		if diskInfoErr != nil {
+			offlineDisks[disks[i].Hostname()]++
+			continue
+		}
+		onlineDisks[disks[i].Hostname()]++
+	}
+
+	// Iterate over the passed endpoints arguments and check
+	// if there are still disks missing from the offline/online lists
+	// and update them accordingly.
+	missingOfflineDisks := make(map[string]int)
+	for _, zone := range globalEndpoints {
+		for _, endpoint := range zone.Endpoints {
+			// if local is set and endpoint is not local
+			// we are not interested in remote disks.
+			if local && !endpoint.IsLocal {
+				continue
+			}
+
+			if _, ok := offlineDisks[endpoint.Host]; !ok {
+				missingOfflineDisks[endpoint.Host]++
+			}
+		}
+	}
+	for missingDisk, n := range missingOfflineDisks {
+		onlineDisks[missingDisk] = 0
+		offlineDisks[missingDisk] = n
+	}
+
+	// Success.
+	return disksInfo, errs, onlineDisks, offlineDisks
+}
+
+// Get an aggregated storage info across all disks.
+func getStorageInfo(disks []StorageAPI, local bool) (StorageInfo, []error) {
+	disksInfo, errs, onlineDisks, offlineDisks := getDisksInfo(disks, local)
+
+	// Sort so that the first element is the smallest.
+	sort.Sort(byDiskTotal(disksInfo))
+
+	// Combine all disks to get total usage
+	usedList := make([]uint64, len(disksInfo))
+	totalList := make([]uint64, len(disksInfo))
+	availableList := make([]uint64, len(disksInfo))
+	mountPaths := make([]string, len(disksInfo))
+
+	for i, di := range disksInfo {
+		usedList[i] = di.Used
+		totalList[i] = di.Total
+		availableList[i] = di.Free
+		mountPaths[i] = di.MountPath
+	}
+
+	storageInfo := StorageInfo{
+		Used:       usedList,
+		Total:      totalList,
+		Available:  availableList,
+		MountPaths: mountPaths,
+	}
+
+	storageInfo.Backend.Type = BackendErasure
+	storageInfo.Backend.OnlineDisks = onlineDisks
+	storageInfo.Backend.OfflineDisks = offlineDisks
+
+	return storageInfo, errs
+}
+
+// StorageInfo - returns underlying storage statistics.
+func (er erasureObjects) StorageInfo(ctx context.Context, local bool) (StorageInfo, []error) {
+	disks := er.getDisks()
+	if local {
+		var localDisks []StorageAPI
+		for _, disk := range disks {
+			if disk != nil {
+				if disk.IsLocal() {
+					// Append this local disk since local flag is true
+					localDisks = append(localDisks, disk)
+				}
+			}
+		}
+		disks = localDisks
+	}
+	return getStorageInfo(disks, local)
+}
+
+// GetMetrics - is not implemented and shouldn't be called.
+func (er erasureObjects) GetMetrics(ctx context.Context) (*Metrics, error) {
+	logger.LogIf(ctx, NotImplemented{})
+	return &Metrics{}, NotImplemented{}
+}
+
+// CrawlAndGetDataUsage collects usage from all buckets.
+// updates are sent as different parts of the underlying
+// structure has been traversed.
+func (er erasureObjects) CrawlAndGetDataUsage(ctx context.Context, bf *bloomFilter, updates chan<- DataUsageInfo) error {
+	return NotImplemented{API: "CrawlAndGetDataUsage"}
+}
+
+// CrawlAndGetDataUsage will start crawling buckets and send updated totals as they are traversed.
+// Updates are sent on a regular basis and the caller *must* consume them.
+func (er erasureObjects) crawlAndGetDataUsage(ctx context.Context, buckets []BucketInfo, bf *bloomFilter, updates chan<- dataUsageCache) error {
+	var disks []StorageAPI
+
+	for _, d := range er.getLoadBalancedDisks() {
+		if d == nil || !d.IsOnline() {
+			continue
+		}
+		disks = append(disks, d)
+	}
+	if len(disks) == 0 || len(buckets) == 0 {
 		return nil
 	}
 
-// ShardSize - returns actual shared size from erasure blockSize.
+	// Load bucket totals
-func (e *Erasure) ShardSize() int64 {
+	oldCache := dataUsageCache{}
-	return ceilFrac(e.blockSize, int64(e.dataBlocks))
+	err := oldCache.load(ctx, er, dataUsageCacheName)
+	if err != nil {
+		return err
 	}
 
-// ShardFileSize - returns final erasure size from original size.
+	// New cache..
-func (e *Erasure) ShardFileSize(totalLength int64) int64 {
+	cache := dataUsageCache{
-	if totalLength == 0 {
+		Info: dataUsageCacheInfo{
-		return 0
+			Name:      dataUsageRoot,
-	}
+			NextCycle: oldCache.Info.NextCycle,
-	if totalLength == -1 {
+		},
-		return -1
+		Cache: make(map[string]dataUsageEntry, len(oldCache.Cache)),
-	}
-	numShards := totalLength / e.blockSize
-	lastBlockSize := totalLength % int64(e.blockSize)
-	lastShardSize := ceilFrac(lastBlockSize, int64(e.dataBlocks))
-	return numShards*e.ShardSize() + lastShardSize
 	}
 
-// ShardFileTillOffset - returns the effectiv eoffset where erasure reading begins.
+	// Put all buckets into channel.
-func (e *Erasure) ShardFileTillOffset(startOffset, length, totalLength int64) int64 {
+	bucketCh := make(chan BucketInfo, len(buckets))
-	shardSize := e.ShardSize()
+	// Add new buckets first
-	shardFileSize := e.ShardFileSize(totalLength)
+	for _, b := range buckets {
-	endShard := (startOffset + int64(length)) / e.blockSize
+		if oldCache.find(b.Name) == nil {
-	tillOffset := endShard*shardSize + shardSize
+			bucketCh <- b
-	if tillOffset > shardFileSize {
-		tillOffset = shardFileSize
 		}
-	return tillOffset
+	}
+	// Add existing buckets.
+	for _, b := range buckets {
+		e := oldCache.find(b.Name)
+		if e != nil {
+			bucketCh <- b
+			cache.replace(b.Name, dataUsageRoot, *e)
+		}
+	}
+
+	close(bucketCh)
+	bucketResults := make(chan dataUsageEntryInfo, len(disks))
+
+	// Start async collector/saver.
+	// This goroutine owns the cache.
+	var saverWg sync.WaitGroup
+	saverWg.Add(1)
+	go func() {
+		const updateTime = 30 * time.Second
+		t := time.NewTicker(updateTime)
+		defer t.Stop()
+		defer saverWg.Done()
+		var lastSave time.Time
+
+	saveLoop:
+		for {
+			select {
+			case <-ctx.Done():
+				// Return without saving.
+				return
+			case <-t.C:
+				if cache.Info.LastUpdate.Equal(lastSave) {
+					continue
+				}
+				logger.LogIf(ctx, cache.save(ctx, er, dataUsageCacheName))
+				updates <- cache.clone()
+				lastSave = cache.Info.LastUpdate
+			case v, ok := <-bucketResults:
+				if !ok {
+					break saveLoop
+				}
+				cache.replace(v.Name, v.Parent, v.Entry)
+				cache.Info.LastUpdate = time.Now()
+			}
+		}
+		// Save final state...
+		cache.Info.NextCycle++
+		cache.Info.LastUpdate = time.Now()
+		logger.LogIf(ctx, cache.save(ctx, er, dataUsageCacheName))
+		updates <- cache
+	}()
+
+	// Start one crawler per disk
+	var wg sync.WaitGroup
+	wg.Add(len(disks))
+	for i := range disks {
+		go func(i int) {
+			defer wg.Done()
+			disk := disks[i]
+
+			for bucket := range bucketCh {
+				select {
+				case <-ctx.Done():
+					return
+				default:
+				}
+
+				// Load cache for bucket
+				cacheName := pathJoin(bucket.Name, dataUsageCacheName)
+				cache := dataUsageCache{}
+				logger.LogIf(ctx, cache.load(ctx, er, cacheName))
+				if cache.Info.Name == "" {
+					cache.Info.Name = bucket.Name
+				}
+				if cache.Info.Name != bucket.Name {
+					logger.LogIf(ctx, fmt.Errorf("cache name mismatch: %s != %s", cache.Info.Name, bucket.Name))
+					cache.Info = dataUsageCacheInfo{
+						Name:       bucket.Name,
+						LastUpdate: time.Time{},
+						NextCycle:  0,
+					}
+				}
+
+				// Calc usage
+				before := cache.Info.LastUpdate
+				cache, err = disk.CrawlAndGetDataUsage(ctx, cache)
+				if err != nil {
+					logger.LogIf(ctx, err)
+					if cache.Info.LastUpdate.After(before) {
+						logger.LogIf(ctx, cache.save(ctx, er, cacheName))
+					}
+					continue
+				}
+
+				var root dataUsageEntry
+				if r := cache.root(); r != nil {
+					root = cache.flatten(*r)
+				}
+				bucketResults <- dataUsageEntryInfo{
+					Name:   cache.Info.Name,
+					Parent: dataUsageRoot,
+					Entry:  root,
+				}
+				// Save cache
+				logger.LogIf(ctx, cache.save(ctx, er, cacheName))
+			}
+		}(i)
+	}
+	wg.Wait()
+	close(bucketResults)
+	saverWg.Wait()
+
+	return nil
+}
+
+// IsReady - shouldn't be called will panic.
+func (er erasureObjects) IsReady(ctx context.Context) bool {
+	logger.CriticalIf(ctx, NotImplemented{})
+	return true
 }

cmd/erasure_test.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2016-2020 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -130,7 +130,7 @@ func newErasureTestSetup(dataBlocks int, parityBlocks int, blockSize int64) (*er
 	disks := make([]StorageAPI, len(diskPaths))
 	var err error
 	for i := range diskPaths {
-		disks[i], diskPaths[i], err = newPosixTestSetup()
+		disks[i], diskPaths[i], err = newXLStorageTestSetup()
 		if err != nil {
 			return nil, err
 		}

cmd/format-disk-cache.go

@@ -366,7 +366,7 @@ func loadAndValidateCacheFormat(ctx context.Context, drives []string) (formats [
 func migrateCacheData(ctx context.Context, c *diskCache, bucket, object, oldfile, destDir string, metadata map[string]string) error {
 	st, err := os.Stat(oldfile)
 	if err != nil {
-		err = osErrToFSFileErr(err)
+		err = osErrToFileErr(err)
 		return err
 	}
 	readCloser, err := readCacheFileStream(oldfile, 0, st.Size())

cmd/format-erasure.go

@@ -36,20 +36,23 @@ import (
 )
 
 const (
-	// Represents XL backend.
+	// Represents Erasure backend.
-	formatBackendXL = "xl"
+	formatBackendErasure = "xl"
 
-	// formatXLV1.XL.Version - version '1'.
+	// formatErasureV1.Erasure.Version - version '1'.
-	formatXLVersionV1 = "1"
+	formatErasureVersionV1 = "1"
 
-	// formatXLV2.XL.Version - version '2'.
+	// formatErasureV2.Erasure.Version - version '2'.
-	formatXLVersionV2 = "2"
+	formatErasureVersionV2 = "2"
 
-	// formatXLV3.XL.Version - version '3'.
+	// formatErasureV3.Erasure.Version - version '3'.
-	formatXLVersionV3 = "3"
+	formatErasureVersionV3 = "3"
 
-	// Distribution algorithm used.
+	// Distribution algorithm used, legacy
-	formatXLVersionV2DistributionAlgo = "CRCMOD"
+	formatErasureVersionV2DistributionAlgoLegacy = "CRCMOD"
+
+	// Distributed algorithm used, current
+	formatErasureVersionV3DistributionAlgo = "SIPMOD"
 )
 
 // Offline disk UUID represents an offline disk.
@@ -68,34 +71,34 @@ var formatCriticalErrors = map[error]struct{}{
 }
 
 // Used to detect the version of "xl" format.
-type formatXLVersionDetect struct {
+type formatErasureVersionDetect struct {
-	XL struct {
+	Erasure struct {
 		Version string `json:"version"`
 	} `json:"xl"`
 }
 
 // Represents the V1 backend disk structure version
 // under `.minio.sys` and actual data namespace.
-// formatXLV1 - structure holds format config version '1'.
+// formatErasureV1 - structure holds format config version '1'.
-type formatXLV1 struct {
+type formatErasureV1 struct {
 	formatMetaV1
-	XL struct {
+	Erasure struct {
 		Version string `json:"version"` // Version of 'xl' format.
 		Disk    string `json:"disk"`    // Disk field carries assigned disk uuid.
 		// JBOD field carries the input disk order generated the first
 		// time when fresh disks were supplied.
 		JBOD []string `json:"jbod"`
-	} `json:"xl"` // XL field holds xl format.
+	} `json:"xl"` // Erasure field holds xl format.
 }
 
 // Represents the V2 backend disk structure version
 // under `.minio.sys` and actual data namespace.
-// formatXLV2 - structure holds format config version '2'.
+// formatErasureV2 - structure holds format config version '2'.
 // The V2 format to support "large bucket" support where a bucket
 // can span multiple erasure sets.
-type formatXLV2 struct {
+type formatErasureV2 struct {
 	formatMetaV1
-	XL struct {
+	Erasure struct {
 		Version string `json:"version"` // Version of 'xl' format.
 		This    string `json:"this"`    // This field carries assigned disk uuid.
 		// Sets field carries the input disk order generated the first
@@ -108,13 +111,13 @@ type formatXLV2 struct {
 	} `json:"xl"`
 }
 
-// formatXLV3 struct is same as formatXLV2 struct except that formatXLV3.XL.Version is "3" indicating
+// formatErasureV3 struct is same as formatErasureV2 struct except that formatErasureV3.Erasure.Version is "3" indicating
 // the simplified multipart backend which is a flat hierarchy now.
 // In .minio.sys/multipart we have:
-// sha256(bucket/object)/uploadID/[xl.json, part.1, part.2 ....]
+// sha256(bucket/object)/uploadID/[xl.meta, part.1, part.2 ....]
-type formatXLV3 struct {
+type formatErasureV3 struct {
 	formatMetaV1
-	XL struct {
+	Erasure struct {
 		Version string `json:"version"` // Version of 'xl' format.
 		This    string `json:"this"`    // This field carries assigned disk uuid.
 		// Sets field carries the input disk order generated the first
@@ -127,40 +130,40 @@ type formatXLV3 struct {
 	} `json:"xl"`
 }
 
-func (f *formatXLV3) Clone() *formatXLV3 {
+func (f *formatErasureV3) Clone() *formatErasureV3 {
 	b, err := json.Marshal(f)
 	if err != nil {
 		panic(err)
 	}
-	var dst formatXLV3
+	var dst formatErasureV3
 	if err = json.Unmarshal(b, &dst); err != nil {
 		panic(err)
 	}
 	return &dst
 }
 
-// Returns formatXL.XL.Version
+// Returns formatErasure.Erasure.Version
-func newFormatXLV3(numSets int, setLen int) *formatXLV3 {
+func newFormatErasureV3(numSets int, setLen int) *formatErasureV3 {
-	format := &formatXLV3{}
+	format := &formatErasureV3{}
 	format.Version = formatMetaVersionV1
-	format.Format = formatBackendXL
+	format.Format = formatBackendErasure
 	format.ID = mustGetUUID()
-	format.XL.Version = formatXLVersionV3
+	format.Erasure.Version = formatErasureVersionV3
-	format.XL.DistributionAlgo = formatXLVersionV2DistributionAlgo
+	format.Erasure.DistributionAlgo = formatErasureVersionV3DistributionAlgo
-	format.XL.Sets = make([][]string, numSets)
+	format.Erasure.Sets = make([][]string, numSets)
 
 	for i := 0; i < numSets; i++ {
-		format.XL.Sets[i] = make([]string, setLen)
+		format.Erasure.Sets[i] = make([]string, setLen)
 		for j := 0; j < setLen; j++ {
-			format.XL.Sets[i][j] = mustGetUUID()
+			format.Erasure.Sets[i][j] = mustGetUUID()
 		}
 	}
 	return format
 }
 
-// Returns format XL version after reading `format.json`, returns
+// Returns format Erasure version after reading `format.json`, returns
-// successfully the version only if the backend is XL.
+// successfully the version only if the backend is Erasure.
-func formatGetBackendXLVersion(formatPath string) (string, error) {
+func formatGetBackendErasureVersion(formatPath string) (string, error) {
 	meta := &formatMetaV1{}
 	b, err := ioutil.ReadFile(formatPath)
 	if err != nil {
@@ -172,42 +175,42 @@ func formatGetBackendXLVersion(formatPath string) (string, error) {
 	if meta.Version != formatMetaVersionV1 {
 		return "", fmt.Errorf(`format.Version expected: %s, got: %s`, formatMetaVersionV1, meta.Version)
 	}
-	if meta.Format != formatBackendXL {
+	if meta.Format != formatBackendErasure {
-		return "", fmt.Errorf(`found backend %s, expected %s`, meta.Format, formatBackendXL)
+		return "", fmt.Errorf(`found backend %s, expected %s`, meta.Format, formatBackendErasure)
 	}
-	// XL backend found, proceed to detect version.
+	// Erasure backend found, proceed to detect version.
-	format := &formatXLVersionDetect{}
+	format := &formatErasureVersionDetect{}
 	if err = json.Unmarshal(b, format); err != nil {
 		return "", err
 	}
-	return format.XL.Version, nil
+	return format.Erasure.Version, nil
 }
 
 // Migrates all previous versions to latest version of `format.json`,
 // this code calls migration in sequence, such as V1 is migrated to V2
 // first before it V2 migrates to V3.
-func formatXLMigrate(export string) error {
+func formatErasureMigrate(export string) error {
 	formatPath := pathJoin(export, minioMetaBucket, formatConfigFile)
-	version, err := formatGetBackendXLVersion(formatPath)
+	version, err := formatGetBackendErasureVersion(formatPath)
 	if err != nil {
 		return err
 	}
 	switch version {
-	case formatXLVersionV1:
+	case formatErasureVersionV1:
-		if err = formatXLMigrateV1ToV2(export, version); err != nil {
+		if err = formatErasureMigrateV1ToV2(export, version); err != nil {
 			return err
 		}
 		// Migrate successful v1 => v2, proceed to v2 => v3
-		version = formatXLVersionV2
+		version = formatErasureVersionV2
 		fallthrough
-	case formatXLVersionV2:
+	case formatErasureVersionV2:
-		if err = formatXLMigrateV2ToV3(export, version); err != nil {
+		if err = formatErasureMigrateV2ToV3(export, version); err != nil {
 			return err
 		}
 		// Migrate successful v2 => v3, v3 is latest
 		// version = formatXLVersionV3
 		fallthrough
-	case formatXLVersionV3:
+	case formatErasureVersionV3:
 		// v3 is the latest version, return.
 		return nil
 	}
@@ -216,14 +219,14 @@ func formatXLMigrate(export string) error {
 
 // Migrates version V1 of format.json to version V2 of format.json,
 // migration fails upon any error.
-func formatXLMigrateV1ToV2(export, version string) error {
+func formatErasureMigrateV1ToV2(export, version string) error {
-	if version != formatXLVersionV1 {
+	if version != formatErasureVersionV1 {
-		return fmt.Errorf(`Disk %s: format version expected %s, found %s`, export, formatXLVersionV1, version)
+		return fmt.Errorf(`Disk %s: format version expected %s, found %s`, export, formatErasureVersionV1, version)
 	}
 
 	formatPath := pathJoin(export, minioMetaBucket, formatConfigFile)
 
-	formatV1 := &formatXLV1{}
+	formatV1 := &formatErasureV1{}
 	b, err := ioutil.ReadFile(formatPath)
 	if err != nil {
 		return err
@@ -232,15 +235,15 @@ func formatXLMigrateV1ToV2(export, version string) error {
 		return err
 	}
 
-	formatV2 := &formatXLV2{}
+	formatV2 := &formatErasureV2{}
 	formatV2.Version = formatMetaVersionV1
-	formatV2.Format = formatBackendXL
+	formatV2.Format = formatBackendErasure
-	formatV2.XL.Version = formatXLVersionV2
+	formatV2.Erasure.Version = formatErasureVersionV2
-	formatV2.XL.DistributionAlgo = formatXLVersionV2DistributionAlgo
+	formatV2.Erasure.DistributionAlgo = formatErasureVersionV2DistributionAlgoLegacy
-	formatV2.XL.This = formatV1.XL.Disk
+	formatV2.Erasure.This = formatV1.Erasure.Disk
-	formatV2.XL.Sets = make([][]string, 1)
+	formatV2.Erasure.Sets = make([][]string, 1)
-	formatV2.XL.Sets[0] = make([]string, len(formatV1.XL.JBOD))
+	formatV2.Erasure.Sets[0] = make([]string, len(formatV1.Erasure.JBOD))
-	copy(formatV2.XL.Sets[0], formatV1.XL.JBOD)
+	copy(formatV2.Erasure.Sets[0], formatV1.Erasure.JBOD)
 
 	b, err = json.Marshal(formatV2)
 	if err != nil {
@@ -250,13 +253,13 @@ func formatXLMigrateV1ToV2(export, version string) error {
 }
 
 // Migrates V2 for format.json to V3 (Flat hierarchy for multipart)
-func formatXLMigrateV2ToV3(export, version string) error {
+func formatErasureMigrateV2ToV3(export, version string) error {
-	if version != formatXLVersionV2 {
+	if version != formatErasureVersionV2 {
-		return fmt.Errorf(`Disk %s: format version expected %s, found %s`, export, formatXLVersionV2, version)
+		return fmt.Errorf(`Disk %s: format version expected %s, found %s`, export, formatErasureVersionV2, version)
 	}
 
 	formatPath := pathJoin(export, minioMetaBucket, formatConfigFile)
-	formatV2 := &formatXLV2{}
+	formatV2 := &formatErasureV2{}
 	b, err := ioutil.ReadFile(formatPath)
 	if err != nil {
 		return err
@@ -276,13 +279,13 @@ func formatXLMigrateV2ToV3(export, version string) error {
 
 	// format-V2 struct is exactly same as format-V1 except that version is "3"
 	// which indicates the simplified multipart backend.
-	formatV3 := formatXLV3{}
+	formatV3 := formatErasureV3{}
 
 	formatV3.Version = formatV2.Version
 	formatV3.Format = formatV2.Format
-	formatV3.XL = formatV2.XL
+	formatV3.Erasure = formatV2.Erasure
 
-	formatV3.XL.Version = formatXLVersionV3
+	formatV3.Erasure.Version = formatErasureVersionV3
 
 	b, err = json.Marshal(formatV3)
 	if err != nil {
@@ -303,7 +306,7 @@ func countErrs(errs []error, err error) int {
 }
 
 // Does all errors indicate we need to initialize all disks?.
-func shouldInitXLDisks(errs []error) bool {
+func shouldInitErasureDisks(errs []error) bool {
 	return countErrs(errs, errUnformattedDisk) == len(errs)
 }
 
@@ -312,13 +315,13 @@ func quorumUnformattedDisks(errs []error) bool {
 	return countErrs(errs, errUnformattedDisk) >= (len(errs)/2)+1
 }
 
-// loadFormatXLAll - load all format config from all input disks in parallel.
+// loadFormatErasureAll - load all format config from all input disks in parallel.
-func loadFormatXLAll(storageDisks []StorageAPI, heal bool) ([]*formatXLV3, []error) {
+func loadFormatErasureAll(storageDisks []StorageAPI, heal bool) ([]*formatErasureV3, []error) {
 	// Initialize list of errors.
 	g := errgroup.WithNErrs(len(storageDisks))
 
 	// Initialize format configs.
-	var formats = make([]*formatXLV3, len(storageDisks))
+	var formats = make([]*formatErasureV3, len(storageDisks))
 
 	// Load format from each disk in parallel
 	for index := range storageDisks {
@@ -327,7 +330,7 @@ func loadFormatXLAll(storageDisks []StorageAPI, heal bool) ([]*formatXLV3, []err
 			if storageDisks[index] == nil {
 				return errDiskNotFound
 			}
-			format, err := loadFormatXL(storageDisks[index])
+			format, err := loadFormatErasure(storageDisks[index])
 			if err != nil {
 				return err
 			}
@@ -335,7 +338,7 @@ func loadFormatXLAll(storageDisks []StorageAPI, heal bool) ([]*formatXLV3, []err
 			if !heal {
 				// If no healing required, make the disks valid and
 				// online.
-				storageDisks[index].SetDiskID(format.XL.This)
+				storageDisks[index].SetDiskID(format.Erasure.This)
 			}
 			return nil
 		}, index)
@@ -345,12 +348,12 @@ func loadFormatXLAll(storageDisks []StorageAPI, heal bool) ([]*formatXLV3, []err
 	return formats, g.Wait()
 }
 
-func saveFormatXL(disk StorageAPI, format interface{}, diskID string) error {
+func saveFormatErasure(disk StorageAPI, format interface{}, diskID string) error {
 	if format == nil || disk == nil {
 		return errDiskNotFound
 	}
 
-	if err := makeFormatXLMetaVolumes(disk); err != nil {
+	if err := makeFormatErasureMetaVolumes(disk); err != nil {
 		return err
 	}
 
@@ -398,8 +401,8 @@ func isHiddenDirectories(vols ...VolInfo) bool {
 	return true
 }
 
-// loadFormatXL - loads format.json from disk.
+// loadFormatErasure - loads format.json from disk.
-func loadFormatXL(disk StorageAPI) (format *formatXLV3, err error) {
+func loadFormatErasure(disk StorageAPI) (format *formatErasureV3, err error) {
 	buf, err := disk.ReadAll(minioMetaBucket, formatConfigFile)
 	if err != nil {
 		// 'file not found' and 'volume not found' as
@@ -421,7 +424,7 @@ func loadFormatXL(disk StorageAPI) (format *formatXLV3, err error) {
 	}
 
 	// Try to decode format json into formatConfigV1 struct.
-	format = &formatXLV3{}
+	format = &formatErasureV3{}
 	if err = json.Unmarshal(buf, format); err != nil {
 		return nil, err
 	}
@@ -430,56 +433,56 @@ func loadFormatXL(disk StorageAPI) (format *formatXLV3, err error) {
 	return format, nil
 }
 
-// Valid formatXL basic versions.
+// Valid formatErasure basic versions.
-func checkFormatXLValue(formatXL *formatXLV3) error {
+func checkFormatErasureValue(formatErasure *formatErasureV3) error {
 	// Validate format version and format type.
-	if formatXL.Version != formatMetaVersionV1 {
+	if formatErasure.Version != formatMetaVersionV1 {
-		return fmt.Errorf("Unsupported version of backend format [%s] found", formatXL.Version)
+		return fmt.Errorf("Unsupported version of backend format [%s] found", formatErasure.Version)
 	}
-	if formatXL.Format != formatBackendXL {
+	if formatErasure.Format != formatBackendErasure {
-		return fmt.Errorf("Unsupported backend format [%s] found", formatXL.Format)
+		return fmt.Errorf("Unsupported backend format [%s] found", formatErasure.Format)
 	}
-	if formatXL.XL.Version != formatXLVersionV3 {
+	if formatErasure.Erasure.Version != formatErasureVersionV3 {
-		return fmt.Errorf("Unsupported XL backend format found [%s]", formatXL.XL.Version)
+		return fmt.Errorf("Unsupported Erasure backend format found [%s]", formatErasure.Erasure.Version)
 	}
 	return nil
 }
 
 // Check all format values.
-func checkFormatXLValues(formats []*formatXLV3, drivesPerSet int) error {
+func checkFormatErasureValues(formats []*formatErasureV3, drivesPerSet int) error {
-	for i, formatXL := range formats {
+	for i, formatErasure := range formats {
-		if formatXL == nil {
+		if formatErasure == nil {
 			continue
 		}
-		if err := checkFormatXLValue(formatXL); err != nil {
+		if err := checkFormatErasureValue(formatErasure); err != nil {
 			return err
 		}
-		if len(formats) != len(formatXL.XL.Sets)*len(formatXL.XL.Sets[0]) {
+		if len(formats) != len(formatErasure.Erasure.Sets)*len(formatErasure.Erasure.Sets[0]) {
 			return fmt.Errorf("%s disk is already being used in another erasure deployment. (Number of disks specified: %d but the number of disks found in the %s disk's format.json: %d)",
-				humanize.Ordinal(i+1), len(formats), humanize.Ordinal(i+1), len(formatXL.XL.Sets)*len(formatXL.XL.Sets[0]))
+				humanize.Ordinal(i+1), len(formats), humanize.Ordinal(i+1), len(formatErasure.Erasure.Sets)*len(formatErasure.Erasure.Sets[0]))
 		}
 		// Only if custom erasure drive count is set,
 		// we should fail here other proceed to honor what
 		// is present on the disk.
-		if globalCustomErasureDriveCount && len(formatXL.XL.Sets[0]) != drivesPerSet {
+		if globalCustomErasureDriveCount && len(formatErasure.Erasure.Sets[0]) != drivesPerSet {
-			return fmt.Errorf("%s disk is already formatted with %d drives per erasure set. This cannot be changed to %d, please revert your MINIO_ERASURE_SET_DRIVE_COUNT setting", humanize.Ordinal(i+1), len(formatXL.XL.Sets[0]), drivesPerSet)
+			return fmt.Errorf("%s disk is already formatted with %d drives per erasure set. This cannot be changed to %d, please revert your MINIO_ERASURE_SET_DRIVE_COUNT setting", humanize.Ordinal(i+1), len(formatErasure.Erasure.Sets[0]), drivesPerSet)
 		}
 	}
 	return nil
 }
 
-// Get Deployment ID for the XL sets from format.json.
+// Get Deployment ID for the Erasure sets from format.json.
 // This need not be in quorum. Even if one of the format.json
 // file has this value, we assume it is valid.
 // If more than one format.json's have different id, it is considered a corrupt
 // backend format.
-func formatXLGetDeploymentID(refFormat *formatXLV3, formats []*formatXLV3) (string, error) {
+func formatErasureGetDeploymentID(refFormat *formatErasureV3, formats []*formatErasureV3) (string, error) {
 	var deploymentID string
 	for _, format := range formats {
 		if format == nil || format.ID == "" {
 			continue
 		}
-		if reflect.DeepEqual(format.XL.Sets, refFormat.XL.Sets) {
+		if reflect.DeepEqual(format.Erasure.Sets, refFormat.Erasure.Sets) {
 			// Found an ID in one of the format.json file
 			// Set deploymentID for the first time.
 			if deploymentID == "" {
@@ -494,11 +497,11 @@ func formatXLGetDeploymentID(refFormat *formatXLV3, formats []*formatXLV3) (stri
 	return deploymentID, nil
 }
 
-// formatXLFixDeploymentID - Add deployment id if it is not present.
+// formatErasureFixDeploymentID - Add deployment id if it is not present.
-func formatXLFixDeploymentID(endpoints Endpoints, storageDisks []StorageAPI, refFormat *formatXLV3) (err error) {
+func formatErasureFixDeploymentID(endpoints Endpoints, storageDisks []StorageAPI, refFormat *formatErasureV3) (err error) {
 	// Attempt to load all `format.json` from all disks.
 	var sErrs []error
-	formats, sErrs := loadFormatXLAll(storageDisks, false)
+	formats, sErrs := loadFormatErasureAll(storageDisks, false)
 	for i, sErr := range sErrs {
 		if _, ok := formatCriticalErrors[sErr]; ok {
 			return config.ErrCorruptedBackend(err).Hint(fmt.Sprintf("Clear any pre-existing content on %s", endpoints[i]))
@@ -506,13 +509,13 @@ func formatXLFixDeploymentID(endpoints Endpoints, storageDisks []StorageAPI, ref
 	}
 
 	for index := range formats {
-		// If the XL sets do not match, set those formats to nil,
+		// If the Erasure sets do not match, set those formats to nil,
 		// We do not have to update the ID on those format.json file.
-		if formats[index] != nil && !reflect.DeepEqual(formats[index].XL.Sets, refFormat.XL.Sets) {
+		if formats[index] != nil && !reflect.DeepEqual(formats[index].Erasure.Sets, refFormat.Erasure.Sets) {
 			formats[index] = nil
 		}
 	}
-	refFormat.ID, err = formatXLGetDeploymentID(refFormat, formats)
+	refFormat.ID, err = formatErasureGetDeploymentID(refFormat, formats)
 	if err != nil {
 		return err
 	}
@@ -534,12 +537,12 @@ func formatXLFixDeploymentID(endpoints Endpoints, storageDisks []StorageAPI, ref
 	}
 	// Deployment ID needs to be set on all the disks.
 	// Save `format.json` across all disks.
-	return saveFormatXLAll(GlobalContext, storageDisks, formats)
+	return saveFormatErasureAll(GlobalContext, storageDisks, formats)
 
 }
 
 // Update only the valid local disks which have not been updated before.
-func formatXLFixLocalDeploymentID(endpoints Endpoints, storageDisks []StorageAPI, refFormat *formatXLV3) error {
+func formatErasureFixLocalDeploymentID(endpoints Endpoints, storageDisks []StorageAPI, refFormat *formatErasureV3) error {
 	// If this server was down when the deploymentID was updated
 	// then we make sure that we update the local disks with the deploymentID.
 
@@ -550,7 +553,7 @@ func formatXLFixLocalDeploymentID(endpoints Endpoints, storageDisks []StorageAPI
 		index := index
 		g.Go(func() error {
 			if endpoints[index].IsLocal && storageDisks[index] != nil && storageDisks[index].IsOnline() {
-				format, err := loadFormatXL(storageDisks[index])
+				format, err := loadFormatErasure(storageDisks[index])
 				if err != nil {
 					// Disk can be offline etc.
 					// ignore the errors seen here.
@@ -559,11 +562,11 @@ func formatXLFixLocalDeploymentID(endpoints Endpoints, storageDisks []StorageAPI
 				if format.ID != "" {
 					return nil
 				}
-				if !reflect.DeepEqual(format.XL.Sets, refFormat.XL.Sets) {
+				if !reflect.DeepEqual(format.Erasure.Sets, refFormat.Erasure.Sets) {
 					return nil
 				}
 				format.ID = refFormat.ID
-				if err := saveFormatXL(storageDisks[index], format, format.XL.This); err != nil {
+				if err := saveFormatErasure(storageDisks[index], format, format.Erasure.This); err != nil {
 					logger.LogIf(GlobalContext, err)
 					return fmt.Errorf("Unable to save format.json, %w", err)
 				}
@@ -579,15 +582,15 @@ func formatXLFixLocalDeploymentID(endpoints Endpoints, storageDisks []StorageAPI
 	return nil
 }
 
-// Get backend XL format in quorum `format.json`.
+// Get backend Erasure format in quorum `format.json`.
-func getFormatXLInQuorum(formats []*formatXLV3) (*formatXLV3, error) {
+func getFormatErasureInQuorum(formats []*formatErasureV3) (*formatErasureV3, error) {
 	formatHashes := make([]string, len(formats))
 	for i, format := range formats {
 		if format == nil {
 			continue
 		}
 		h := sha256.New()
-		for _, set := range format.XL.Sets {
+		for _, set := range format.Erasure.Sets {
 			for _, diskID := range set {
 				h.Write([]byte(diskID))
 			}
@@ -613,55 +616,55 @@ func getFormatXLInQuorum(formats []*formatXLV3) (*formatXLV3, error) {
 	}
 
 	if maxCount < len(formats)/2 {
-		return nil, errXLReadQuorum
+		return nil, errErasureReadQuorum
 	}
 
 	for i, hash := range formatHashes {
 		if hash == maxHash {
 			format := formats[i].Clone()
-			format.XL.This = ""
+			format.Erasure.This = ""
 			return format, nil
 		}
 	}
 
-	return nil, errXLReadQuorum
+	return nil, errErasureReadQuorum
 }
 
-func formatXLV3Check(reference *formatXLV3, format *formatXLV3) error {
+func formatErasureV3Check(reference *formatErasureV3, format *formatErasureV3) error {
 	tmpFormat := format.Clone()
-	this := tmpFormat.XL.This
+	this := tmpFormat.Erasure.This
-	tmpFormat.XL.This = ""
+	tmpFormat.Erasure.This = ""
-	if len(reference.XL.Sets) != len(format.XL.Sets) {
+	if len(reference.Erasure.Sets) != len(format.Erasure.Sets) {
-		return fmt.Errorf("Expected number of sets %d, got %d", len(reference.XL.Sets), len(format.XL.Sets))
+		return fmt.Errorf("Expected number of sets %d, got %d", len(reference.Erasure.Sets), len(format.Erasure.Sets))
 	}
 
 	// Make sure that the sets match.
-	for i := range reference.XL.Sets {
+	for i := range reference.Erasure.Sets {
-		if len(reference.XL.Sets[i]) != len(format.XL.Sets[i]) {
+		if len(reference.Erasure.Sets[i]) != len(format.Erasure.Sets[i]) {
 			return fmt.Errorf("Each set should be of same size, expected %d got %d",
-				len(reference.XL.Sets[i]), len(format.XL.Sets[i]))
+				len(reference.Erasure.Sets[i]), len(format.Erasure.Sets[i]))
 		}
-		for j := range reference.XL.Sets[i] {
+		for j := range reference.Erasure.Sets[i] {
-			if reference.XL.Sets[i][j] != format.XL.Sets[i][j] {
+			if reference.Erasure.Sets[i][j] != format.Erasure.Sets[i][j] {
 				return fmt.Errorf("UUID on positions %d:%d do not match with, expected %s got %s",
-					i, j, reference.XL.Sets[i][j], format.XL.Sets[i][j])
+					i, j, reference.Erasure.Sets[i][j], format.Erasure.Sets[i][j])
 			}
 		}
 	}
 
 	// Make sure that the diskID is found in the set.
-	for i := 0; i < len(tmpFormat.XL.Sets); i++ {
+	for i := 0; i < len(tmpFormat.Erasure.Sets); i++ {
-		for j := 0; j < len(tmpFormat.XL.Sets[i]); j++ {
+		for j := 0; j < len(tmpFormat.Erasure.Sets[i]); j++ {
-			if this == tmpFormat.XL.Sets[i][j] {
+			if this == tmpFormat.Erasure.Sets[i][j] {
 				return nil
 			}
 		}
 	}
-	return fmt.Errorf("Disk ID %s not found in any disk sets %s", this, format.XL.Sets)
+	return fmt.Errorf("Disk ID %s not found in any disk sets %s", this, format.Erasure.Sets)
 }
 
 // Initializes meta volume only on local storage disks.
-func initXLMetaVolumesInLocalDisks(storageDisks []StorageAPI, formats []*formatXLV3) error {
+func initErasureMetaVolumesInLocalDisks(storageDisks []StorageAPI, formats []*formatErasureV3) error {
 
 	// Compute the local disks eligible for meta volumes (re)initialization
 	var disksToInit []StorageAPI
@@ -682,7 +685,7 @@ func initXLMetaVolumesInLocalDisks(storageDisks []StorageAPI, formats []*formatX
 		// goroutine will return its own instance of index variable.
 		index := index
 		g.Go(func() error {
-			return makeFormatXLMetaVolumes(disksToInit[index])
+			return makeFormatErasureMetaVolumes(disksToInit[index])
 		}, index)
 	}
 
@@ -698,15 +701,15 @@ func initXLMetaVolumesInLocalDisks(storageDisks []StorageAPI, formats []*formatX
 	return nil
 }
 
-// saveFormatXLAll - populates `format.json` on disks in its order.
+// saveFormatErasureAll - populates `format.json` on disks in its order.
-func saveFormatXLAll(ctx context.Context, storageDisks []StorageAPI, formats []*formatXLV3) error {
+func saveFormatErasureAll(ctx context.Context, storageDisks []StorageAPI, formats []*formatErasureV3) error {
 	g := errgroup.WithNErrs(len(storageDisks))
 
 	// Write `format.json` to all disks.
 	for index := range storageDisks {
 		index := index
 		g.Go(func() error {
-			return saveFormatXL(storageDisks[index], formats[index], formats[index].XL.This)
+			return saveFormatErasure(storageDisks[index], formats[index], formats[index].Erasure.This)
 		}, index)
 	}
 
@@ -745,9 +748,9 @@ func initStorageDisksWithErrors(endpoints Endpoints) ([]StorageAPI, []error) {
 	return storageDisks, g.Wait()
 }
 
-// formatXLV3ThisEmpty - find out if '.This' field is empty
+// formatErasureV3ThisEmpty - find out if '.This' field is empty
 // in any of the input `formats`, if yes return true.
-func formatXLV3ThisEmpty(formats []*formatXLV3) bool {
+func formatErasureV3ThisEmpty(formats []*formatErasureV3) bool {
 	for _, format := range formats {
 		if format == nil {
 			continue
@@ -756,18 +759,18 @@ func formatXLV3ThisEmpty(formats []*formatXLV3) bool {
 		// V1 to V2 to V3, in a scenario such as this we only need to handle
 		// single sets since we never used to support multiple sets in releases
 		// with V1 format version.
-		if len(format.XL.Sets) > 1 {
+		if len(format.Erasure.Sets) > 1 {
 			continue
 		}
-		if format.XL.This == "" {
+		if format.Erasure.This == "" {
 			return true
 		}
 	}
 	return false
 }
 
-// fixFormatXLV3 - fix format XL configuration on all disks.
+// fixFormatErasureV3 - fix format Erasure configuration on all disks.
-func fixFormatXLV3(storageDisks []StorageAPI, endpoints Endpoints, formats []*formatXLV3) error {
+func fixFormatErasureV3(storageDisks []StorageAPI, endpoints Endpoints, formats []*formatErasureV3) error {
 	g := errgroup.WithNErrs(len(formats))
 	for i := range formats {
 		i := i
@@ -779,12 +782,12 @@ func fixFormatXLV3(storageDisks []StorageAPI, endpoints Endpoints, formats []*fo
 			// V1 to V2 to V3, in a scenario such as this we only need to handle
 			// single sets since we never used to support multiple sets in releases
 			// with V1 format version.
-			if len(formats[i].XL.Sets) > 1 {
+			if len(formats[i].Erasure.Sets) > 1 {
 				return nil
 			}
-			if formats[i].XL.This == "" {
+			if formats[i].Erasure.This == "" {
-				formats[i].XL.This = formats[i].XL.Sets[0][i]
+				formats[i].Erasure.This = formats[i].Erasure.Sets[0][i]
-				if err := saveFormatXL(storageDisks[i], formats[i], formats[i].XL.This); err != nil {
+				if err := saveFormatErasure(storageDisks[i], formats[i], formats[i].Erasure.This); err != nil {
 					return err
 				}
 			}
@@ -800,10 +803,10 @@ func fixFormatXLV3(storageDisks []StorageAPI, endpoints Endpoints, formats []*fo
 
 }
 
-// initFormatXL - save XL format configuration on all disks.
+// initFormatErasure - save Erasure format configuration on all disks.
-func initFormatXL(ctx context.Context, storageDisks []StorageAPI, setCount, drivesPerSet int, deploymentID string) (*formatXLV3, error) {
+func initFormatErasure(ctx context.Context, storageDisks []StorageAPI, setCount, drivesPerSet int, deploymentID string) (*formatErasureV3, error) {
-	format := newFormatXLV3(setCount, drivesPerSet)
+	format := newFormatErasureV3(setCount, drivesPerSet)
-	formats := make([]*formatXLV3, len(storageDisks))
+	formats := make([]*formatErasureV3, len(storageDisks))
 	wantAtMost := ecDrivesNoConfig(drivesPerSet)
 
 	for i := 0; i < setCount; i++ {
@@ -811,7 +814,7 @@ func initFormatXL(ctx context.Context, storageDisks []StorageAPI, setCount, driv
 		for j := 0; j < drivesPerSet; j++ {
 			disk := storageDisks[i*drivesPerSet+j]
 			newFormat := format.Clone()
-			newFormat.XL.This = format.XL.Sets[i][j]
+			newFormat.Erasure.This = format.Erasure.Sets[i][j]
 			if deploymentID != "" {
 				newFormat.ID = deploymentID
 			}
@@ -843,11 +846,11 @@ func initFormatXL(ctx context.Context, storageDisks []StorageAPI, setCount, driv
 	}
 
 	// Save formats `format.json` across all disks.
-	if err := saveFormatXLAll(ctx, storageDisks, formats); err != nil {
+	if err := saveFormatErasureAll(ctx, storageDisks, formats); err != nil {
 		return nil, err
 	}
 
-	return getFormatXLInQuorum(formats)
+	return getFormatErasureInQuorum(formats)
 }
 
 // ecDrivesNoConfig returns the erasure coded drives in a set if no config has been set.
@@ -866,8 +869,8 @@ func ecDrivesNoConfig(drivesPerSet int) int {
 	return ecDrives
 }
 
-// Make XL backend meta volumes.
+// Make Erasure backend meta volumes.
-func makeFormatXLMetaVolumes(disk StorageAPI) error {
+func makeFormatErasureMetaVolumes(disk StorageAPI) error {
 	if disk == nil {
 		return errDiskNotFound
 	}
@@ -878,14 +881,14 @@ func makeFormatXLMetaVolumes(disk StorageAPI) error {
 // Get all UUIDs which are present in reference format should
 // be present in the list of formats provided, those are considered
 // as online UUIDs.
-func getOnlineUUIDs(refFormat *formatXLV3, formats []*formatXLV3) (onlineUUIDs []string) {
+func getOnlineUUIDs(refFormat *formatErasureV3, formats []*formatErasureV3) (onlineUUIDs []string) {
 	for _, format := range formats {
 		if format == nil {
 			continue
 		}
-		for _, set := range refFormat.XL.Sets {
+		for _, set := range refFormat.Erasure.Sets {
 			for _, uuid := range set {
-				if format.XL.This == uuid {
+				if format.Erasure.This == uuid {
 					onlineUUIDs = append(onlineUUIDs, uuid)
 				}
 			}
@@ -897,13 +900,13 @@ func getOnlineUUIDs(refFormat *formatXLV3, formats []*formatXLV3) (onlineUUIDs [
 // Look for all UUIDs which are not present in reference format
 // but are present in the onlineUUIDs list, construct of list such
 // offline UUIDs.
-func getOfflineUUIDs(refFormat *formatXLV3, formats []*formatXLV3) (offlineUUIDs []string) {
+func getOfflineUUIDs(refFormat *formatErasureV3, formats []*formatErasureV3) (offlineUUIDs []string) {
 	onlineUUIDs := getOnlineUUIDs(refFormat, formats)
-	for i, set := range refFormat.XL.Sets {
+	for i, set := range refFormat.Erasure.Sets {
 		for j, uuid := range set {
 			var found bool
 			for _, onlineUUID := range onlineUUIDs {
-				if refFormat.XL.Sets[i][j] == onlineUUID {
+				if refFormat.Erasure.Sets[i][j] == onlineUUID {
 					found = true
 				}
 			}
@@ -916,13 +919,13 @@ func getOfflineUUIDs(refFormat *formatXLV3, formats []*formatXLV3) (offlineUUIDs
 }
 
 // Mark all UUIDs that are offline.
-func markUUIDsOffline(refFormat *formatXLV3, formats []*formatXLV3) {
+func markUUIDsOffline(refFormat *formatErasureV3, formats []*formatErasureV3) {
 	offlineUUIDs := getOfflineUUIDs(refFormat, formats)
-	for i, set := range refFormat.XL.Sets {
+	for i, set := range refFormat.Erasure.Sets {
 		for j := range set {
 			for _, offlineUUID := range offlineUUIDs {
-				if refFormat.XL.Sets[i][j] == offlineUUID {
+				if refFormat.Erasure.Sets[i][j] == offlineUUID {
-					refFormat.XL.Sets[i][j] = offlineDiskUUID
+					refFormat.Erasure.Sets[i][j] = offlineDiskUUID
 				}
 			}
 		}
@@ -930,29 +933,29 @@ func markUUIDsOffline(refFormat *formatXLV3, formats []*formatXLV3) {
 }
 
 // Initialize a new set of set formats which will be written to all disks.
-func newHealFormatSets(refFormat *formatXLV3, setCount, drivesPerSet int, formats []*formatXLV3, errs []error) [][]*formatXLV3 {
+func newHealFormatSets(refFormat *formatErasureV3, setCount, drivesPerSet int, formats []*formatErasureV3, errs []error) [][]*formatErasureV3 {
-	newFormats := make([][]*formatXLV3, setCount)
+	newFormats := make([][]*formatErasureV3, setCount)
-	for i := range refFormat.XL.Sets {
+	for i := range refFormat.Erasure.Sets {
-		newFormats[i] = make([]*formatXLV3, drivesPerSet)
+		newFormats[i] = make([]*formatErasureV3, drivesPerSet)
 	}
-	for i := range refFormat.XL.Sets {
+	for i := range refFormat.Erasure.Sets {
-		for j := range refFormat.XL.Sets[i] {
+		for j := range refFormat.Erasure.Sets[i] {
 			if errs[i*drivesPerSet+j] == errUnformattedDisk || errs[i*drivesPerSet+j] == nil {
-				newFormats[i][j] = &formatXLV3{}
+				newFormats[i][j] = &formatErasureV3{}
 				newFormats[i][j].Version = refFormat.Version
 				newFormats[i][j].ID = refFormat.ID
 				newFormats[i][j].Format = refFormat.Format
-				newFormats[i][j].XL.Version = refFormat.XL.Version
+				newFormats[i][j].Erasure.Version = refFormat.Erasure.Version
-				newFormats[i][j].XL.DistributionAlgo = refFormat.XL.DistributionAlgo
+				newFormats[i][j].Erasure.DistributionAlgo = refFormat.Erasure.DistributionAlgo
 			}
 			if errs[i*drivesPerSet+j] == errUnformattedDisk {
-				newFormats[i][j].XL.This = ""
+				newFormats[i][j].Erasure.This = ""
-				newFormats[i][j].XL.Sets = nil
+				newFormats[i][j].Erasure.Sets = nil
 				continue
 			}
 			if errs[i*drivesPerSet+j] == nil {
-				newFormats[i][j].XL.This = formats[i*drivesPerSet+j].XL.This
+				newFormats[i][j].Erasure.This = formats[i*drivesPerSet+j].Erasure.This
-				newFormats[i][j].XL.Sets = nil
+				newFormats[i][j].Erasure.Sets = nil
 			}
 		}
 	}

cmd/format-erasure_test.go

@@ -26,13 +26,13 @@ import (
 
 // Test get offline/online uuids.
 func TestGetUUIDs(t *testing.T) {
-	fmtV2 := newFormatXLV3(4, 16)
+	fmtV2 := newFormatErasureV3(4, 16)
-	formats := make([]*formatXLV3, 64)
+	formats := make([]*formatErasureV3, 64)
 
 	for i := 0; i < 4; i++ {
 		for j := 0; j < 16; j++ {
 			newFormat := *fmtV2
-			newFormat.XL.This = fmtV2.XL.Sets[i][j]
+			newFormat.Erasure.This = fmtV2.Erasure.Sets[i][j]
 			formats[i*16+j] = &newFormat
 		}
 	}
@@ -62,9 +62,9 @@ func TestGetUUIDs(t *testing.T) {
 
 	markUUIDsOffline(fmtV2, formats)
 	gotCount = 0
-	for i := range fmtV2.XL.Sets {
+	for i := range fmtV2.Erasure.Sets {
-		for j := range fmtV2.XL.Sets[i] {
+		for j := range fmtV2.Erasure.Sets[i] {
-			if fmtV2.XL.Sets[i][j] == offlineDiskUUID {
+			if fmtV2.Erasure.Sets[i][j] == offlineDiskUUID {
 				gotCount++
 			}
 		}
@@ -74,16 +74,16 @@ func TestGetUUIDs(t *testing.T) {
 	}
 }
 
-// tests fixFormatXLV3 - fix format.json on all disks.
+// tests fixFormatErasureV3 - fix format.json on all disks.
 func TestFixFormatV3(t *testing.T) {
-	xlDirs, err := getRandomDisks(8)
+	erasureDirs, err := getRandomDisks(8)
 	if err != nil {
 		t.Fatal(err)
 	}
-	for _, xlDir := range xlDirs {
+	for _, erasureDir := range erasureDirs {
-		defer os.RemoveAll(xlDir)
+		defer os.RemoveAll(erasureDir)
 	}
-	endpoints := mustGetNewEndpoints(xlDirs...)
+	endpoints := mustGetNewEndpoints(erasureDirs...)
 
 	storageDisks, errs := initStorageDisksWithErrors(endpoints)
 	for _, err := range errs {
@@ -92,46 +92,46 @@ func TestFixFormatV3(t *testing.T) {
 		}
 	}
 
-	format := newFormatXLV3(1, 8)
+	format := newFormatErasureV3(1, 8)
-	formats := make([]*formatXLV3, 8)
+	formats := make([]*formatErasureV3, 8)
 
 	for j := 0; j < 8; j++ {
 		newFormat := format.Clone()
-		newFormat.XL.This = format.XL.Sets[0][j]
+		newFormat.Erasure.This = format.Erasure.Sets[0][j]
 		formats[j] = newFormat
 	}
 
-	if err = initXLMetaVolumesInLocalDisks(storageDisks, formats); err != nil {
+	if err = initErasureMetaVolumesInLocalDisks(storageDisks, formats); err != nil {
 		t.Fatal(err)
 	}
 
 	formats[1] = nil
-	expThis := formats[2].XL.This
+	expThis := formats[2].Erasure.This
-	formats[2].XL.This = ""
+	formats[2].Erasure.This = ""
-	if err := fixFormatXLV3(storageDisks, endpoints, formats); err != nil {
+	if err := fixFormatErasureV3(storageDisks, endpoints, formats); err != nil {
 		t.Fatal(err)
 	}
 
-	newFormats, errs := loadFormatXLAll(storageDisks, false)
+	newFormats, errs := loadFormatErasureAll(storageDisks, false)
 	for _, err := range errs {
 		if err != nil && err != errUnformattedDisk {
 			t.Fatal(err)
 		}
 	}
-	gotThis := newFormats[2].XL.This
+	gotThis := newFormats[2].Erasure.This
 	if expThis != gotThis {
 		t.Fatalf("expected uuid %s, got %s", expThis, gotThis)
 	}
 }
 
-// tests formatXLV3ThisEmpty conditions.
+// tests formatErasureV3ThisEmpty conditions.
-func TestFormatXLEmpty(t *testing.T) {
+func TestFormatErasureEmpty(t *testing.T) {
-	format := newFormatXLV3(1, 16)
+	format := newFormatErasureV3(1, 16)
-	formats := make([]*formatXLV3, 16)
+	formats := make([]*formatErasureV3, 16)
 
 	for j := 0; j < 16; j++ {
 		newFormat := format.Clone()
-		newFormat.XL.This = format.XL.Sets[0][j]
+		newFormat.Erasure.This = format.Erasure.Sets[0][j]
 		formats[j] = newFormat
 	}
 
@@ -139,18 +139,18 @@ func TestFormatXLEmpty(t *testing.T) {
 	// empty should return false.
 	formats[0] = nil
 
-	if ok := formatXLV3ThisEmpty(formats); ok {
+	if ok := formatErasureV3ThisEmpty(formats); ok {
 		t.Fatalf("expected value false, got %t", ok)
 	}
 
-	formats[2].XL.This = ""
+	formats[2].Erasure.This = ""
-	if ok := formatXLV3ThisEmpty(formats); !ok {
+	if ok := formatErasureV3ThisEmpty(formats); !ok {
 		t.Fatalf("expected value true, got %t", ok)
 	}
 }
 
 // Tests xl format migration.
-func TestFormatXLMigrate(t *testing.T) {
+func TestFormatErasureMigrate(t *testing.T) {
 	// Get test root.
 	rootPath, err := getTestRoot()
 	if err != nil {
@@ -158,12 +158,12 @@ func TestFormatXLMigrate(t *testing.T) {
 	}
 	defer os.RemoveAll(rootPath)
 
-	m := &formatXLV1{}
+	m := &formatErasureV1{}
-	m.Format = formatBackendXL
+	m.Format = formatBackendErasure
 	m.Version = formatMetaVersionV1
-	m.XL.Version = formatXLVersionV1
+	m.Erasure.Version = formatErasureVersionV1
-	m.XL.Disk = mustGetUUID()
+	m.Erasure.Disk = mustGetUUID()
-	m.XL.JBOD = []string{m.XL.Disk, mustGetUUID(), mustGetUUID(), mustGetUUID()}
+	m.Erasure.JBOD = []string{m.Erasure.Disk, mustGetUUID(), mustGetUUID(), mustGetUUID()}
 
 	b, err := json.Marshal(m)
 	if err != nil {
@@ -178,43 +178,43 @@ func TestFormatXLMigrate(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	if err = formatXLMigrate(rootPath); err != nil {
+	if err = formatErasureMigrate(rootPath); err != nil {
 		t.Fatal(err)
 	}
 
-	migratedVersion, err := formatGetBackendXLVersion(pathJoin(rootPath, minioMetaBucket, formatConfigFile))
+	migratedVersion, err := formatGetBackendErasureVersion(pathJoin(rootPath, minioMetaBucket, formatConfigFile))
 	if err != nil {
 		t.Fatal(err)
 	}
 
-	if migratedVersion != formatXLVersionV3 {
+	if migratedVersion != formatErasureVersionV3 {
-		t.Fatalf("expected version: %s, got: %s", formatXLVersionV3, migratedVersion)
+		t.Fatalf("expected version: %s, got: %s", formatErasureVersionV3, migratedVersion)
 	}
 
 	b, err = ioutil.ReadFile(pathJoin(rootPath, minioMetaBucket, formatConfigFile))
 	if err != nil {
 		t.Fatal(err)
 	}
-	formatV3 := &formatXLV3{}
+	formatV3 := &formatErasureV3{}
 	if err = json.Unmarshal(b, formatV3); err != nil {
 		t.Fatal(err)
 	}
-	if formatV3.XL.This != m.XL.Disk {
+	if formatV3.Erasure.This != m.Erasure.Disk {
-		t.Fatalf("expected disk uuid: %s, got: %s", m.XL.Disk, formatV3.XL.This)
+		t.Fatalf("expected disk uuid: %s, got: %s", m.Erasure.Disk, formatV3.Erasure.This)
 	}
-	if len(formatV3.XL.Sets) != 1 {
+	if len(formatV3.Erasure.Sets) != 1 {
-		t.Fatalf("expected single set after migrating from v1 to v3, but found %d", len(formatV3.XL.Sets))
+		t.Fatalf("expected single set after migrating from v1 to v3, but found %d", len(formatV3.Erasure.Sets))
 	}
-	if !reflect.DeepEqual(formatV3.XL.Sets[0], m.XL.JBOD) {
+	if !reflect.DeepEqual(formatV3.Erasure.Sets[0], m.Erasure.JBOD) {
-		t.Fatalf("expected disk uuid: %v, got: %v", m.XL.JBOD, formatV3.XL.Sets[0])
+		t.Fatalf("expected disk uuid: %v, got: %v", m.Erasure.JBOD, formatV3.Erasure.Sets[0])
 	}
 
-	m = &formatXLV1{}
+	m = &formatErasureV1{}
 	m.Format = "unknown"
 	m.Version = formatMetaVersionV1
-	m.XL.Version = formatXLVersionV1
+	m.Erasure.Version = formatErasureVersionV1
-	m.XL.Disk = mustGetUUID()
+	m.Erasure.Disk = mustGetUUID()
-	m.XL.JBOD = []string{m.XL.Disk, mustGetUUID(), mustGetUUID(), mustGetUUID()}
+	m.Erasure.JBOD = []string{m.Erasure.Disk, mustGetUUID(), mustGetUUID(), mustGetUUID()}
 
 	b, err = json.Marshal(m)
 	if err != nil {
@@ -225,16 +225,16 @@ func TestFormatXLMigrate(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	if err = formatXLMigrate(rootPath); err == nil {
+	if err = formatErasureMigrate(rootPath); err == nil {
 		t.Fatal("Expected to fail with unexpected backend format")
 	}
 
-	m = &formatXLV1{}
+	m = &formatErasureV1{}
-	m.Format = formatBackendXL
+	m.Format = formatBackendErasure
 	m.Version = formatMetaVersionV1
-	m.XL.Version = "30"
+	m.Erasure.Version = "30"
-	m.XL.Disk = mustGetUUID()
+	m.Erasure.Disk = mustGetUUID()
-	m.XL.JBOD = []string{m.XL.Disk, mustGetUUID(), mustGetUUID(), mustGetUUID()}
+	m.Erasure.JBOD = []string{m.Erasure.Disk, mustGetUUID(), mustGetUUID(), mustGetUUID()}
 
 	b, err = json.Marshal(m)
 	if err != nil {
@@ -245,25 +245,25 @@ func TestFormatXLMigrate(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	if err = formatXLMigrate(rootPath); err == nil {
+	if err = formatErasureMigrate(rootPath); err == nil {
 		t.Fatal("Expected to fail with unexpected backend format version number")
 	}
 }
 
 // Tests check format xl value.
-func TestCheckFormatXLValue(t *testing.T) {
+func TestCheckFormatErasureValue(t *testing.T) {
 	testCases := []struct {
-		format  *formatXLV3
+		format  *formatErasureV3
 		success bool
 	}{
-		// Invalid XL format version "2".
+		// Invalid Erasure format version "2".
 		{
-			&formatXLV3{
+			&formatErasureV3{
 				formatMetaV1: formatMetaV1{
 					Version: "2",
-					Format:  "XL",
+					Format:  "Erasure",
 				},
-				XL: struct {
+				Erasure: struct {
 					Version          string     `json:"version"`
 					This             string     `json:"this"`
 					Sets             [][]string `json:"sets"`
@@ -274,14 +274,14 @@ func TestCheckFormatXLValue(t *testing.T) {
 			},
 			false,
 		},
-		// Invalid XL format "Unknown".
+		// Invalid Erasure format "Unknown".
 		{
-			&formatXLV3{
+			&formatErasureV3{
 				formatMetaV1: formatMetaV1{
 					Version: "1",
 					Format:  "Unknown",
 				},
-				XL: struct {
+				Erasure: struct {
 					Version          string     `json:"version"`
 					This             string     `json:"this"`
 					Sets             [][]string `json:"sets"`
@@ -292,14 +292,14 @@ func TestCheckFormatXLValue(t *testing.T) {
 			},
 			false,
 		},
-		// Invalid XL format version "0".
+		// Invalid Erasure format version "0".
 		{
-			&formatXLV3{
+			&formatErasureV3{
 				formatMetaV1: formatMetaV1{
 					Version: "1",
-					Format:  "XL",
+					Format:  "Erasure",
 				},
-				XL: struct {
+				Erasure: struct {
 					Version          string     `json:"version"`
 					This             string     `json:"this"`
 					Sets             [][]string `json:"sets"`
@@ -314,65 +314,65 @@ func TestCheckFormatXLValue(t *testing.T) {
 
 	// Valid all test cases.
 	for i, testCase := range testCases {
-		if err := checkFormatXLValue(testCase.format); err != nil && testCase.success {
+		if err := checkFormatErasureValue(testCase.format); err != nil && testCase.success {
 			t.Errorf("Test %d: Expected failure %s", i+1, err)
 		}
 	}
 }
 
-// Tests getFormatXLInQuorum()
+// Tests getFormatErasureInQuorum()
-func TestGetFormatXLInQuorumCheck(t *testing.T) {
+func TestGetFormatErasureInQuorumCheck(t *testing.T) {
 	setCount := 2
 	drivesPerSet := 16
 
-	format := newFormatXLV3(setCount, drivesPerSet)
+	format := newFormatErasureV3(setCount, drivesPerSet)
-	formats := make([]*formatXLV3, 32)
+	formats := make([]*formatErasureV3, 32)
 
 	for i := 0; i < setCount; i++ {
 		for j := 0; j < drivesPerSet; j++ {
 			newFormat := format.Clone()
-			newFormat.XL.This = format.XL.Sets[i][j]
+			newFormat.Erasure.This = format.Erasure.Sets[i][j]
 			formats[i*drivesPerSet+j] = newFormat
 		}
 	}
 
 	// Return a format from list of formats in quorum.
-	quorumFormat, err := getFormatXLInQuorum(formats)
+	quorumFormat, err := getFormatErasureInQuorum(formats)
 	if err != nil {
 		t.Fatal(err)
 	}
 
 	// Check if the reference format and input formats are same.
-	if err = formatXLV3Check(quorumFormat, formats[0]); err != nil {
+	if err = formatErasureV3Check(quorumFormat, formats[0]); err != nil {
 		t.Fatal(err)
 	}
 
 	// QuorumFormat has .This field empty on purpose, expect a failure.
-	if err = formatXLV3Check(formats[0], quorumFormat); err == nil {
+	if err = formatErasureV3Check(formats[0], quorumFormat); err == nil {
 		t.Fatal("Unexpected success")
 	}
 
 	formats[0] = nil
-	quorumFormat, err = getFormatXLInQuorum(formats)
+	quorumFormat, err = getFormatErasureInQuorum(formats)
 	if err != nil {
 		t.Fatal(err)
 	}
 
 	badFormat := *quorumFormat
-	badFormat.XL.Sets = nil
+	badFormat.Erasure.Sets = nil
-	if err = formatXLV3Check(quorumFormat, &badFormat); err == nil {
+	if err = formatErasureV3Check(quorumFormat, &badFormat); err == nil {
 		t.Fatal("Unexpected success")
 	}
 
 	badFormatUUID := *quorumFormat
-	badFormatUUID.XL.Sets[0][0] = "bad-uuid"
+	badFormatUUID.Erasure.Sets[0][0] = "bad-uuid"
-	if err = formatXLV3Check(quorumFormat, &badFormatUUID); err == nil {
+	if err = formatErasureV3Check(quorumFormat, &badFormatUUID); err == nil {
 		t.Fatal("Unexpected success")
 	}
 
 	badFormatSetSize := *quorumFormat
-	badFormatSetSize.XL.Sets[0] = nil
+	badFormatSetSize.Erasure.Sets[0] = nil
-	if err = formatXLV3Check(quorumFormat, &badFormatSetSize); err == nil {
+	if err = formatErasureV3Check(quorumFormat, &badFormatSetSize); err == nil {
 		t.Fatal("Unexpected success")
 	}
 
@@ -381,36 +381,36 @@ func TestGetFormatXLInQuorumCheck(t *testing.T) {
 			formats[i] = nil
 		}
 	}
-	if _, err = getFormatXLInQuorum(formats); err == nil {
+	if _, err = getFormatErasureInQuorum(formats); err == nil {
 		t.Fatal("Unexpected success")
 	}
 }
 
-// Tests formatXLGetDeploymentID()
+// Tests formatErasureGetDeploymentID()
-func TestGetXLID(t *testing.T) {
+func TestGetErasureID(t *testing.T) {
 	setCount := 2
 	drivesPerSet := 8
 
-	format := newFormatXLV3(setCount, drivesPerSet)
+	format := newFormatErasureV3(setCount, drivesPerSet)
-	formats := make([]*formatXLV3, 16)
+	formats := make([]*formatErasureV3, 16)
 
 	for i := 0; i < setCount; i++ {
 		for j := 0; j < drivesPerSet; j++ {
 			newFormat := format.Clone()
-			newFormat.XL.This = format.XL.Sets[i][j]
+			newFormat.Erasure.This = format.Erasure.Sets[i][j]
 			formats[i*drivesPerSet+j] = newFormat
 		}
 	}
 
 	// Return a format from list of formats in quorum.
-	quorumFormat, err := getFormatXLInQuorum(formats)
+	quorumFormat, err := getFormatErasureInQuorum(formats)
 	if err != nil {
 		t.Fatal(err)
 	}
 
 	// Check if the reference format and input formats are same.
 	var id string
-	if id, err = formatXLGetDeploymentID(quorumFormat, formats); err != nil {
+	if id, err = formatErasureGetDeploymentID(quorumFormat, formats); err != nil {
 		t.Fatal(err)
 	}
 
@@ -419,15 +419,15 @@ func TestGetXLID(t *testing.T) {
 	}
 
 	formats[0] = nil
-	if id, err = formatXLGetDeploymentID(quorumFormat, formats); err != nil {
+	if id, err = formatErasureGetDeploymentID(quorumFormat, formats); err != nil {
 		t.Fatal(err)
 	}
 	if id == "" {
 		t.Fatal("ID cannot be empty.")
 	}
 
-	formats[1].XL.Sets[0][0] = "bad-uuid"
+	formats[1].Erasure.Sets[0][0] = "bad-uuid"
-	if id, err = formatXLGetDeploymentID(quorumFormat, formats); err != nil {
+	if id, err = formatErasureGetDeploymentID(quorumFormat, formats); err != nil {
 		t.Fatal(err)
 	}
 
@@ -436,7 +436,7 @@ func TestGetXLID(t *testing.T) {
 	}
 
 	formats[2].ID = "bad-id"
-	if _, err = formatXLGetDeploymentID(quorumFormat, formats); err != errCorruptedFormat {
+	if _, err = formatErasureGetDeploymentID(quorumFormat, formats); err != errCorruptedFormat {
 		t.Fatal("Unexpected Success")
 	}
 }
@@ -446,19 +446,19 @@ func TestNewFormatSets(t *testing.T) {
 	setCount := 2
 	drivesPerSet := 16
 
-	format := newFormatXLV3(setCount, drivesPerSet)
+	format := newFormatErasureV3(setCount, drivesPerSet)
-	formats := make([]*formatXLV3, 32)
+	formats := make([]*formatErasureV3, 32)
 	errs := make([]error, 32)
 
 	for i := 0; i < setCount; i++ {
 		for j := 0; j < drivesPerSet; j++ {
 			newFormat := format.Clone()
-			newFormat.XL.This = format.XL.Sets[i][j]
+			newFormat.Erasure.This = format.Erasure.Sets[i][j]
 			formats[i*drivesPerSet+j] = newFormat
 		}
 	}
 
-	quorumFormat, err := getFormatXLInQuorum(formats)
+	quorumFormat, err := getFormatErasureInQuorum(formats)
 	if err != nil {
 		t.Fatal(err)
 	}

cmd/format-fs.go

@@ -75,7 +75,7 @@ func newFormatFSV1() (format *formatFSV1) {
 }
 
 // Returns the field formatMetaV1.Format i.e the string "fs" which is never likely to change.
-// We do not use this function in XL to get the format as the file is not fcntl-locked on XL.
+// We do not use this function in Erasure to get the format as the file is not fcntl-locked on Erasure.
 func formatMetaGetFormatBackendFS(r io.ReadSeeker) (string, error) {
 	format := &formatMetaV1{}
 	if err := jsonLoad(r, format); err != nil {

cmd/fs-v1-helpers.go

@@ -42,7 +42,7 @@ func fsRemoveFile(ctx context.Context, filePath string) (err error) {
 	}
 
 	if err = os.Remove((filePath)); err != nil {
-		err = osErrToFSFileErr(err)
+		err = osErrToFileErr(err)
 		if err != errFileNotFound {
 			logger.LogIf(ctx, err)
 		}
@@ -186,37 +186,11 @@ func fsStatVolume(ctx context.Context, volume string) (os.FileInfo, error) {
 	return fi, nil
 }
 
-// Is a one place function which converts all os.PathError
-// into a more FS object layer friendly form, converts
-// known errors into their typed form for top level
-// interpretation.
-func osErrToFSFileErr(err error) error {
-	if err == nil {
-		return nil
-	}
-	if os.IsNotExist(err) {
-		return errFileNotFound
-	}
-	if os.IsPermission(err) {
-		return errFileAccessDenied
-	}
-	if isSysErrNotDir(err) {
-		return errFileNotFound
-	}
-	if isSysErrPathNotFound(err) {
-		return errFileNotFound
-	}
-	if isSysErrTooManyFiles(err) {
-		return errTooManyOpenFiles
-	}
-	return err
-}
-
 // Lookup if directory exists, returns directory attributes upon success.
 func fsStatDir(ctx context.Context, statDir string) (os.FileInfo, error) {
 	fi, err := fsStat(ctx, statDir)
 	if err != nil {
-		err = osErrToFSFileErr(err)
+		err = osErrToFileErr(err)
 		if err != errFileNotFound {
 			logger.LogIf(ctx, err)
 		}
@@ -232,7 +206,7 @@ func fsStatDir(ctx context.Context, statDir string) (os.FileInfo, error) {
 func fsStatFile(ctx context.Context, statFile string) (os.FileInfo, error) {
 	fi, err := fsStat(ctx, statFile)
 	if err != nil {
-		err = osErrToFSFileErr(err)
+		err = osErrToFileErr(err)
 		if err != errFileNotFound {
 			logger.LogIf(ctx, err)
 		}
@@ -267,13 +241,13 @@ func fsOpenFile(ctx context.Context, readPath string, offset int64) (io.ReadClos
 
 	fr, err := os.Open(readPath)
 	if err != nil {
-		return nil, 0, osErrToFSFileErr(err)
+		return nil, 0, osErrToFileErr(err)
 	}
 
 	// Stat to get the size of the file at path.
 	st, err := fr.Stat()
 	if err != nil {
-		err = osErrToFSFileErr(err)
+		err = osErrToFileErr(err)
 		if err != errFileNotFound {
 			logger.LogIf(ctx, err)
 		}
@@ -327,7 +301,7 @@ func fsCreateFile(ctx context.Context, filePath string, reader io.Reader, buf []
 	}
 	writer, err := lock.Open(filePath, flags, 0666)
 	if err != nil {
-		return 0, osErrToFSFileErr(err)
+		return 0, osErrToFileErr(err)
 	}
 	defer writer.Close()
 
@@ -399,7 +373,7 @@ func fsSimpleRenameFile(ctx context.Context, sourcePath, destPath string) error
 
 	if err := os.Rename(sourcePath, destPath); err != nil {
 		logger.LogIf(ctx, err)
-		return osErrToFSFileErr(err)
+		return osErrToFileErr(err)
 	}
 
 	return nil

cmd/fs-v1-helpers_test.go

@@ -28,10 +28,10 @@ import (
 )
 
 func TestFSRenameFile(t *testing.T) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 
@@ -53,10 +53,10 @@ func TestFSRenameFile(t *testing.T) {
 }
 
 func TestFSStats(t *testing.T) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 
@@ -170,11 +170,11 @@ func TestFSStats(t *testing.T) {
 		if testCase.srcPath != "" {
 			if _, err := fsStatFile(GlobalContext, pathJoin(testCase.srcFSPath, testCase.srcVol,
 				testCase.srcPath)); err != testCase.expectedErr {
-				t.Fatalf("TestPosix case %d: Expected: \"%s\", got: \"%s\"", i+1, testCase.expectedErr, err)
+				t.Fatalf("TestErasureStorage case %d: Expected: \"%s\", got: \"%s\"", i+1, testCase.expectedErr, err)
 			}
 		} else {
 			if _, err := fsStatVolume(GlobalContext, pathJoin(testCase.srcFSPath, testCase.srcVol)); err != testCase.expectedErr {
-				t.Fatalf("TestPosix case %d: Expected: \"%s\", got: \"%s\"", i+1, testCase.expectedErr, err)
+				t.Fatalf("TestFS case %d: Expected: \"%s\", got: \"%s\"", i+1, testCase.expectedErr, err)
 			}
 		}
 	}
@@ -182,9 +182,9 @@ func TestFSStats(t *testing.T) {
 
 func TestFSCreateAndOpen(t *testing.T) {
 	// Setup test environment.
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 
@@ -246,10 +246,10 @@ func TestFSCreateAndOpen(t *testing.T) {
 }
 
 func TestFSDeletes(t *testing.T) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 
@@ -349,10 +349,10 @@ func TestFSDeletes(t *testing.T) {
 }
 
 func BenchmarkFSDeleteFile(b *testing.B) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		b.Fatalf("Unable to create posix test setup, %s", err)
+		b.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 
@@ -383,10 +383,10 @@ func BenchmarkFSDeleteFile(b *testing.B) {
 
 // Tests fs removes.
 func TestFSRemoves(t *testing.T) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 
@@ -500,10 +500,10 @@ func TestFSRemoves(t *testing.T) {
 }
 
 func TestFSRemoveMeta(t *testing.T) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, fsPath, err := newPosixTestSetup()
+	_, fsPath, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(fsPath)
 

cmd/fs-v1-metadata_test.go

@@ -31,7 +31,7 @@ func TestFSV1MetadataObjInfo(t *testing.T) {
 	if objInfo.Size != 0 {
 		t.Fatal("Unexpected object info value for Size", objInfo.Size)
 	}
-	if objInfo.ModTime != timeSentinel {
+	if !objInfo.ModTime.Equal(timeSentinel) {
 		t.Fatal("Unexpected object info value for ModTime ", objInfo.ModTime)
 	}
 	if objInfo.IsDir {
@@ -53,7 +53,7 @@ func TestReadFSMetadata(t *testing.T) {
 	bucketName := "bucket"
 	objectName := "object"
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Unexpected err: ", err)
 	}
 	if _, err := obj.PutObject(GlobalContext, bucketName, objectName, mustGetPutObjReader(t, bytes.NewReader([]byte("abcd")), int64(len("abcd")), "", ""), ObjectOptions{}); err != nil {
@@ -88,7 +88,7 @@ func TestWriteFSMetadata(t *testing.T) {
 	bucketName := "bucket"
 	objectName := "object"
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Unexpected err: ", err)
 	}
 	if _, err := obj.PutObject(GlobalContext, bucketName, objectName, mustGetPutObjReader(t, bytes.NewReader([]byte("abcd")), int64(len("abcd")), "", ""), ObjectOptions{}); err != nil {

cmd/fs-v1-multipart.go

@@ -252,6 +252,14 @@ func (fs *FSObjects) NewMultipartUpload(ctx context.Context, bucket, object stri
 func (fs *FSObjects) CopyObjectPart(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject, uploadID string, partID int,
 	startOffset int64, length int64, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (pi PartInfo, e error) {
 
+	if srcOpts.VersionID != "" && srcOpts.VersionID != nullVersionID {
+		return pi, VersionNotFound{
+			Bucket:    srcBucket,
+			Object:    srcObject,
+			VersionID: srcOpts.VersionID,
+		}
+	}
+
 	if err := checkNewMultipartArgs(ctx, srcBucket, srcObject, fs); err != nil {
 		return pi, toObjectErr(err)
 	}
@@ -269,6 +277,14 @@ func (fs *FSObjects) CopyObjectPart(ctx context.Context, srcBucket, srcObject, d
 // written to '.minio.sys/tmp' location and safely renamed to
 // '.minio.sys/multipart' for reach parts.
 func (fs *FSObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID string, partID int, r *PutObjReader, opts ObjectOptions) (pi PartInfo, e error) {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return pi, VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
+
 	data := r.Reader
 	if err := checkPutObjectPartArgs(ctx, bucket, object, fs); err != nil {
 		return pi, toObjectErr(err, bucket)

cmd/fs-v1-multipart_test.go

@@ -40,7 +40,7 @@ func TestFSCleanupMultipartUploadsInRoutine(t *testing.T) {
 
 	// Create a context we can cancel.
 	ctx, cancel := context.WithCancel(GlobalContext)
-	obj.MakeBucketWithLocation(ctx, bucketName, "", false)
+	obj.MakeBucketWithLocation(ctx, bucketName, BucketOptions{})
 
 	uploadID, err := obj.NewMultipartUpload(ctx, bucketName, objectName, ObjectOptions{})
 	if err != nil {
@@ -81,7 +81,7 @@ func TestNewMultipartUploadFaultyDisk(t *testing.T) {
 	bucketName := "bucket"
 	objectName := "object"
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Cannot create bucket, err: ", err)
 	}
 
@@ -106,7 +106,7 @@ func TestPutObjectPartFaultyDisk(t *testing.T) {
 	data := []byte("12345")
 	dataLen := int64(len(data))
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Cannot create bucket, err: ", err)
 	}
 
@@ -139,7 +139,7 @@ func TestCompleteMultipartUploadFaultyDisk(t *testing.T) {
 	objectName := "object"
 	data := []byte("12345")
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Cannot create bucket, err: ", err)
 	}
 
@@ -172,7 +172,7 @@ func TestCompleteMultipartUpload(t *testing.T) {
 	objectName := "object"
 	data := []byte("12345")
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Cannot create bucket, err: ", err)
 	}
 
@@ -204,7 +204,7 @@ func TestAbortMultipartUpload(t *testing.T) {
 	objectName := "object"
 	data := []byte("12345")
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Cannot create bucket, err: ", err)
 	}
 
@@ -235,7 +235,7 @@ func TestListMultipartUploadsFaultyDisk(t *testing.T) {
 	bucketName := "bucket"
 	objectName := "object"
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Cannot create bucket, err: ", err)
 	}
 

cmd/fs-v1-rwpool_test.go

@@ -46,10 +46,10 @@ func TestRWPoolLongPath(t *testing.T) {
 
 // Tests all RWPool methods.
 func TestRWPool(t *testing.T) {
-	// create posix test setup
+	// create xlStorage test setup
-	_, path, err := newPosixTestSetup()
+	_, path, err := newXLStorageTestSetup()
 	if err != nil {
-		t.Fatalf("Unable to create posix test setup, %s", err)
+		t.Fatalf("Unable to create xlStorage test setup, %s", err)
 	}
 	defer os.RemoveAll(path)
 

cmd/fs-v1.go

@@ -346,7 +346,7 @@ func (fs *FSObjects) crawlBucket(ctx context.Context, bucket string, cache dataU
 		}
 
 		oi := fsMeta.ToObjectInfo(bucket, object, fi)
-		sz := item.applyActions(ctx, fs, actionMeta{oi: oi, meta: fsMeta.Meta})
+		sz := item.applyActions(ctx, fs, actionMeta{oi: oi})
 		if sz >= 0 {
 			return sz, nil
 		}
@@ -382,10 +382,9 @@ func (fs *FSObjects) statBucketDir(ctx context.Context, bucket string) (os.FileI
 	return st, nil
 }
 
-// MakeBucketWithLocation - create a new bucket, returns if it
+// MakeBucketWithLocation - create a new bucket, returns if it already exists.
-// already exists.
+func (fs *FSObjects) MakeBucketWithLocation(ctx context.Context, bucket string, opts BucketOptions) error {
-func (fs *FSObjects) MakeBucketWithLocation(ctx context.Context, bucket, location string, lockEnabled bool) error {
+	if opts.LockEnabled || opts.VersioningEnabled {
-	if lockEnabled {
 		return NotImplemented{}
 	}
 
@@ -581,6 +580,14 @@ func (fs *FSObjects) DeleteBucket(ctx context.Context, bucket string, forceDelet
 // if source object and destination object are same we only
 // update metadata.
 func (fs *FSObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (oi ObjectInfo, e error) {
+	if srcOpts.VersionID != "" && srcOpts.VersionID != nullVersionID {
+		return oi, VersionNotFound{
+			Bucket:    srcBucket,
+			Object:    srcObject,
+			VersionID: srcOpts.VersionID,
+		}
+	}
+
 	cpSrcDstSame := isStringEqual(pathJoin(srcBucket, srcObject), pathJoin(dstBucket, dstObject))
 	defer ObjectPathUpdated(path.Join(dstBucket, dstObject))
 
@@ -649,6 +656,13 @@ func (fs *FSObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBu
 // GetObjectNInfo - returns object info and a reader for object
 // content.
 func (fs *FSObjects) GetObjectNInfo(ctx context.Context, bucket, object string, rs *HTTPRangeSpec, h http.Header, lockType LockType, opts ObjectOptions) (gr *GetObjectReader, err error) {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return nil, VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
 	if err = checkGetObjArgs(ctx, bucket, object); err != nil {
 		return nil, err
 	}
@@ -746,6 +760,14 @@ func (fs *FSObjects) GetObjectNInfo(ctx context.Context, bucket, object string,
 // startOffset indicates the starting read location of the object.
 // length indicates the total length of the object.
 func (fs *FSObjects) GetObject(ctx context.Context, bucket, object string, offset int64, length int64, writer io.Writer, etag string, opts ObjectOptions) (err error) {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
+
 	if err = checkGetObjArgs(ctx, bucket, object); err != nil {
 		return err
 	}
@@ -948,6 +970,13 @@ func (fs *FSObjects) getObjectInfoWithLock(ctx context.Context, bucket, object s
 
 // GetObjectInfo - reads object metadata and replies back ObjectInfo.
 func (fs *FSObjects) GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (oi ObjectInfo, e error) {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return oi, VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
 
 	atomic.AddInt64(&fs.activeIOCount, 1)
 	defer func() {
@@ -998,6 +1027,10 @@ func (fs *FSObjects) parentDirIsObject(ctx context.Context, bucket, parent strin
 // Additionally writes `fs.json` which carries the necessary metadata
 // for future object operations.
 func (fs *FSObjects) PutObject(ctx context.Context, bucket string, object string, r *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, retErr error) {
+	if opts.Versioned {
+		return objInfo, NotImplemented{}
+	}
+
 	if err := checkPutObjectArgs(ctx, bucket, object, fs, r.Size()); err != nil {
 		return ObjectInfo{}, err
 	}
@@ -1146,26 +1179,45 @@ func (fs *FSObjects) putObject(ctx context.Context, bucket string, object string
 
 // DeleteObjects - deletes an object from a bucket, this operation is destructive
 // and there are no rollbacks supported.
-func (fs *FSObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+func (fs *FSObjects) DeleteObjects(ctx context.Context, bucket string, objects []ObjectToDelete, opts ObjectOptions) ([]DeletedObject, []error) {
 	errs := make([]error, len(objects))
+	dobjects := make([]DeletedObject, len(objects))
 	for idx, object := range objects {
-		errs[idx] = fs.DeleteObject(ctx, bucket, object)
+		if object.VersionID != "" {
+			errs[idx] = NotImplemented{}
+			continue
 		}
-	return errs, nil
+		_, errs[idx] = fs.DeleteObject(ctx, bucket, object.ObjectName, opts)
+		if errs[idx] == nil || isErrObjectNotFound(errs[idx]) {
+			dobjects[idx] = DeletedObject{
+				ObjectName: object.ObjectName,
+			}
+			errs[idx] = nil
+		}
+	}
+	return dobjects, errs
 }
 
 // DeleteObject - deletes an object from a bucket, this operation is destructive
 // and there are no rollbacks supported.
-func (fs *FSObjects) DeleteObject(ctx context.Context, bucket, object string) error {
+func (fs *FSObjects) DeleteObject(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error) {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return objInfo, VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
+
 	// Acquire a write lock before deleting the object.
 	lk := fs.NewNSLock(ctx, bucket, object)
-	if err := lk.GetLock(globalOperationTimeout); err != nil {
+	if err = lk.GetLock(globalOperationTimeout); err != nil {
-		return err
+		return objInfo, err
 	}
 	defer lk.Unlock()
 
-	if err := checkDelObjArgs(ctx, bucket, object); err != nil {
+	if err = checkDelObjArgs(ctx, bucket, object); err != nil {
-		return err
+		return objInfo, err
 	}
 
 	defer ObjectPathUpdated(path.Join(bucket, object))
@@ -1175,8 +1227,8 @@ func (fs *FSObjects) DeleteObject(ctx context.Context, bucket, object string) er
 		atomic.AddInt64(&fs.activeIOCount, -1)
 	}()
 
-	if _, err := fs.statBucketDir(ctx, bucket); err != nil {
+	if _, err = fs.statBucketDir(ctx, bucket); err != nil {
-		return toObjectErr(err, bucket)
+		return objInfo, toObjectErr(err, bucket)
 	}
 
 	minioMetaBucketDir := pathJoin(fs.fsPath, minioMetaBucket)
@@ -1189,23 +1241,23 @@ func (fs *FSObjects) DeleteObject(ctx context.Context, bucket, object string) er
 		}
 		if lerr != nil && lerr != errFileNotFound {
 			logger.LogIf(ctx, lerr)
-			return toObjectErr(lerr, bucket, object)
+			return objInfo, toObjectErr(lerr, bucket, object)
 		}
 	}
 
 	// Delete the object.
-	if err := fsDeleteFile(ctx, pathJoin(fs.fsPath, bucket), pathJoin(fs.fsPath, bucket, object)); err != nil {
+	if err = fsDeleteFile(ctx, pathJoin(fs.fsPath, bucket), pathJoin(fs.fsPath, bucket, object)); err != nil {
-		return toObjectErr(err, bucket, object)
+		return objInfo, toObjectErr(err, bucket, object)
 	}
 
 	if bucket != minioMetaBucket {
 		// Delete the metadata object.
-		err := fsDeleteFile(ctx, minioMetaBucketDir, fsMetaPath)
+		err = fsDeleteFile(ctx, minioMetaBucketDir, fsMetaPath)
 		if err != nil && err != errFileNotFound {
-			return toObjectErr(err, bucket, object)
+			return objInfo, toObjectErr(err, bucket, object)
 		}
 	}
-	return nil
+	return ObjectInfo{Bucket: bucket, Name: object}, nil
 }
 
 // Returns function "listDir" of the type listDirFunc.
@@ -1313,6 +1365,11 @@ func (fs *FSObjects) getObjectETag(ctx context.Context, bucket, entry string, lo
 	return extractETag(fsMeta.Meta), nil
 }
 
+// ListObjectVersions not implemented for FS mode.
+func (fs *FSObjects) ListObjectVersions(ctx context.Context, bucket, prefix, marker, versionMarker, delimiter string, maxKeys int) (loi ListObjectVersionsInfo, e error) {
+	return loi, NotImplemented{}
+}
+
 // ListObjects - list all objects at prefix upto maxKeys., optionally delimited by '/'. Maintains the list pool
 // state for future re-entrant list requests.
 func (fs *FSObjects) ListObjects(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (loi ListObjectsInfo, e error) {
@@ -1327,7 +1384,14 @@ func (fs *FSObjects) ListObjects(ctx context.Context, bucket, prefix, marker, de
 }
 
 // GetObjectTags - get object tags from an existing object
-func (fs *FSObjects) GetObjectTags(ctx context.Context, bucket, object string) (*tags.Tags, error) {
+func (fs *FSObjects) GetObjectTags(ctx context.Context, bucket, object string, opts ObjectOptions) (*tags.Tags, error) {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return nil, VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
 	oi, err := fs.GetObjectInfo(ctx, bucket, object, ObjectOptions{})
 	if err != nil {
 		return nil, err
@@ -1337,7 +1401,15 @@ func (fs *FSObjects) GetObjectTags(ctx context.Context, bucket, object string) (
 }
 
 // PutObjectTags - replace or add tags to an existing object
-func (fs *FSObjects) PutObjectTags(ctx context.Context, bucket, object string, tags string) error {
+func (fs *FSObjects) PutObjectTags(ctx context.Context, bucket, object string, tags string, opts ObjectOptions) error {
+	if opts.VersionID != "" && opts.VersionID != nullVersionID {
+		return VersionNotFound{
+			Bucket:    bucket,
+			Object:    object,
+			VersionID: opts.VersionID,
+		}
+	}
+
 	fsMetaPath := pathJoin(fs.fsPath, minioMetaBucket, bucketMetaPrefix, bucket, object, fs.metaJSONFile)
 	fsMeta := fsMetaV1{}
 	wlk, err := fs.rwPool.Write(fsMetaPath)
@@ -1369,30 +1441,30 @@ func (fs *FSObjects) PutObjectTags(ctx context.Context, bucket, object string, t
 }
 
 // DeleteObjectTags - delete object tags from an existing object
-func (fs *FSObjects) DeleteObjectTags(ctx context.Context, bucket, object string) error {
+func (fs *FSObjects) DeleteObjectTags(ctx context.Context, bucket, object string, opts ObjectOptions) error {
-	return fs.PutObjectTags(ctx, bucket, object, "")
+	return fs.PutObjectTags(ctx, bucket, object, "", opts)
 }
 
-// ReloadFormat - no-op for fs, Valid only for XL.
+// ReloadFormat - no-op for fs, Valid only for Erasure.
 func (fs *FSObjects) ReloadFormat(ctx context.Context, dryRun bool) error {
 	logger.LogIf(ctx, NotImplemented{})
 	return NotImplemented{}
 }
 
-// HealFormat - no-op for fs, Valid only for XL.
+// HealFormat - no-op for fs, Valid only for Erasure.
 func (fs *FSObjects) HealFormat(ctx context.Context, dryRun bool) (madmin.HealResultItem, error) {
 	logger.LogIf(ctx, NotImplemented{})
 	return madmin.HealResultItem{}, NotImplemented{}
 }
 
-// HealObject - no-op for fs. Valid only for XL.
+// HealObject - no-op for fs. Valid only for Erasure.
-func (fs *FSObjects) HealObject(ctx context.Context, bucket, object string, opts madmin.HealOpts) (
+func (fs *FSObjects) HealObject(ctx context.Context, bucket, object, versionID string, opts madmin.HealOpts) (
 	res madmin.HealResultItem, err error) {
 	logger.LogIf(ctx, NotImplemented{})
 	return res, NotImplemented{}
 }
 
-// HealBucket - no-op for fs, Valid only for XL.
+// HealBucket - no-op for fs, Valid only for Erasure.
 func (fs *FSObjects) HealBucket(ctx context.Context, bucket string, dryRun, remove bool) (madmin.HealResultItem,
 	error) {
 	logger.LogIf(ctx, NotImplemented{})
@@ -1408,13 +1480,13 @@ func (fs *FSObjects) Walk(ctx context.Context, bucket, prefix string, results ch
 	return fsWalk(ctx, fs, bucket, prefix, fs.listDirFactory(), results, fs.getObjectInfo, fs.getObjectInfo)
 }
 
-// HealObjects - no-op for fs. Valid only for XL.
+// HealObjects - no-op for fs. Valid only for Erasure.
-func (fs *FSObjects) HealObjects(ctx context.Context, bucket, prefix string, opts madmin.HealOpts, fn healObjectFn) (e error) {
+func (fs *FSObjects) HealObjects(ctx context.Context, bucket, prefix string, opts madmin.HealOpts, fn HealObjectFn) (e error) {
 	logger.LogIf(ctx, NotImplemented{})
 	return NotImplemented{}
 }
 
-// ListBucketsHeal - list all buckets to be healed. Valid only for XL
+// ListBucketsHeal - list all buckets to be healed. Valid only for Erasure
 func (fs *FSObjects) ListBucketsHeal(ctx context.Context) ([]BucketInfo, error) {
 	logger.LogIf(ctx, NotImplemented{})
 	return []BucketInfo{}, NotImplemented{}

cmd/fs-v1_test.go

@@ -36,7 +36,7 @@ func TestFSParentDirIsObject(t *testing.T) {
 	bucketName := "testbucket"
 	objectName := "object"
 
-	if err = obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err = obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal(err)
 	}
 	objectContent := "12345"
@@ -124,7 +124,7 @@ func TestFSShutdown(t *testing.T) {
 		fs := obj.(*FSObjects)
 
 		objectContent := "12345"
-		obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false)
+		obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{})
 		obj.PutObject(GlobalContext, bucketName, objectName, mustGetPutObjReader(t, bytes.NewReader([]byte(objectContent)), int64(len(objectContent)), "", ""), ObjectOptions{})
 		return fs, disk
 	}
@@ -138,7 +138,7 @@ func TestFSShutdown(t *testing.T) {
 
 	// Test Shutdown with faulty disk
 	fs, disk = prepareTest()
-	fs.DeleteObject(GlobalContext, bucketName, objectName)
+	fs.DeleteObject(GlobalContext, bucketName, objectName, ObjectOptions{})
 	os.RemoveAll(disk)
 	if err := fs.Shutdown(GlobalContext); err != nil {
 		t.Fatal("Got unexpected fs shutdown error: ", err)
@@ -155,12 +155,12 @@ func TestFSGetBucketInfo(t *testing.T) {
 	fs := obj.(*FSObjects)
 	bucketName := "bucket"
 
-	err := obj.MakeBucketWithLocation(GlobalContext, "a", "", false)
+	err := obj.MakeBucketWithLocation(GlobalContext, "a", BucketOptions{})
 	if !isSameType(err, BucketNameInvalid{}) {
 		t.Fatal("BucketNameInvalid error not returned")
 	}
 
-	err = obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false)
+	err = obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{})
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -199,7 +199,7 @@ func TestFSPutObject(t *testing.T) {
 	bucketName := "bucket"
 	objectName := "1/2/3/4/object"
 
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal(err)
 	}
 
@@ -267,33 +267,33 @@ func TestFSDeleteObject(t *testing.T) {
 	bucketName := "bucket"
 	objectName := "object"
 
-	obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false)
+	obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{})
 	obj.PutObject(GlobalContext, bucketName, objectName, mustGetPutObjReader(t, bytes.NewReader([]byte("abcd")), int64(len("abcd")), "", ""), ObjectOptions{})
 
 	// Test with invalid bucket name
-	if err := fs.DeleteObject(GlobalContext, "fo", objectName); !isSameType(err, BucketNameInvalid{}) {
+	if _, err := fs.DeleteObject(GlobalContext, "fo", objectName, ObjectOptions{}); !isSameType(err, BucketNameInvalid{}) {
 		t.Fatal("Unexpected error: ", err)
 	}
 	// Test with bucket does not exist
-	if err := fs.DeleteObject(GlobalContext, "foobucket", "fooobject"); !isSameType(err, BucketNotFound{}) {
+	if _, err := fs.DeleteObject(GlobalContext, "foobucket", "fooobject", ObjectOptions{}); !isSameType(err, BucketNotFound{}) {
 		t.Fatal("Unexpected error: ", err)
 	}
 	// Test with invalid object name
-	if err := fs.DeleteObject(GlobalContext, bucketName, "\\"); !(isSameType(err, ObjectNotFound{}) || isSameType(err, ObjectNameInvalid{})) {
+	if _, err := fs.DeleteObject(GlobalContext, bucketName, "\\", ObjectOptions{}); !(isSameType(err, ObjectNotFound{}) || isSameType(err, ObjectNameInvalid{})) {
 		t.Fatal("Unexpected error: ", err)
 	}
 	// Test with object does not exist.
-	if err := fs.DeleteObject(GlobalContext, bucketName, "foooobject"); !isSameType(err, ObjectNotFound{}) {
+	if _, err := fs.DeleteObject(GlobalContext, bucketName, "foooobject", ObjectOptions{}); !isSameType(err, ObjectNotFound{}) {
 		t.Fatal("Unexpected error: ", err)
 	}
 	// Test with valid condition
-	if err := fs.DeleteObject(GlobalContext, bucketName, objectName); err != nil {
+	if _, err := fs.DeleteObject(GlobalContext, bucketName, objectName, ObjectOptions{}); err != nil {
 		t.Fatal("Unexpected error: ", err)
 	}
 
 	// Delete object should err disk not found.
 	os.RemoveAll(disk)
-	if err := fs.DeleteObject(GlobalContext, bucketName, objectName); err != nil {
+	if _, err := fs.DeleteObject(GlobalContext, bucketName, objectName, ObjectOptions{}); err != nil {
 		if !isSameType(err, BucketNotFound{}) {
 			t.Fatal("Unexpected error: ", err)
 		}
@@ -311,7 +311,7 @@ func TestFSDeleteBucket(t *testing.T) {
 	fs := obj.(*FSObjects)
 	bucketName := "bucket"
 
-	err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false)
+	err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{})
 	if err != nil {
 		t.Fatal("Unexpected error: ", err)
 	}
@@ -330,7 +330,7 @@ func TestFSDeleteBucket(t *testing.T) {
 		t.Fatal("Unexpected error: ", err)
 	}
 
-	obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false)
+	obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{})
 
 	// Delete bucket should get error disk not found.
 	os.RemoveAll(disk)
@@ -351,7 +351,7 @@ func TestFSListBuckets(t *testing.T) {
 	fs := obj.(*FSObjects)
 
 	bucketName := "bucket"
-	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, "", false); err != nil {
+	if err := obj.MakeBucketWithLocation(GlobalContext, bucketName, BucketOptions{}); err != nil {
 		t.Fatal("Unexpected error: ", err)
 	}
 
@@ -389,7 +389,7 @@ func TestFSHealObject(t *testing.T) {
 	defer os.RemoveAll(disk)
 
 	obj := initFSObjects(disk, t)
-	_, err := obj.HealObject(GlobalContext, "bucket", "object", madmin.HealOpts{})
+	_, err := obj.HealObject(GlobalContext, "bucket", "object", "", madmin.HealOpts{})
 	if err == nil || !isSameType(err, NotImplemented{}) {
 		t.Fatalf("Heal Object should return NotImplemented error ")
 	}

cmd/gateway-common.go

@@ -55,42 +55,6 @@ var (
 	IsStringEqual = isStringEqual
 )
 
-// StatInfo -  alias for statInfo
-type StatInfo struct {
-	statInfo
-}
-
-// AnonErrToObjectErr - converts standard http codes into meaningful object layer errors.
-func AnonErrToObjectErr(statusCode int, params ...string) error {
-	bucket := ""
-	object := ""
-	if len(params) >= 1 {
-		bucket = params[0]
-	}
-	if len(params) == 2 {
-		object = params[1]
-	}
-
-	switch statusCode {
-	case http.StatusNotFound:
-		if object != "" {
-			return ObjectNotFound{bucket, object}
-		}
-		return BucketNotFound{Bucket: bucket}
-	case http.StatusBadRequest:
-		if object != "" {
-			return ObjectNameInvalid{bucket, object}
-		}
-		return BucketNameInvalid{Bucket: bucket}
-	case http.StatusForbidden:
-		fallthrough
-	case http.StatusUnauthorized:
-		return AllAccessDisabled{bucket, object}
-	}
-
-	return errUnexpected
-}
-
 // FromMinioClientMetadata converts minio metadata to map[string]string
 func FromMinioClientMetadata(metadata map[string][]string) map[string]string {
 	mm := map[string]string{}

cmd/gateway-unsupported.go

@@ -26,6 +26,7 @@ import (
 	bucketsse "github.com/minio/minio/pkg/bucket/encryption"
 	"github.com/minio/minio/pkg/bucket/lifecycle"
 	"github.com/minio/minio/pkg/bucket/policy"
+	"github.com/minio/minio/pkg/bucket/versioning"
 
 	"github.com/minio/minio/pkg/madmin"
 )
@@ -88,6 +89,12 @@ func (a GatewayUnsupported) GetMultipartInfo(ctx context.Context, bucket string,
 	return MultipartInfo{}, NotImplemented{}
 }
 
+// ListObjectVersions returns all object parts for specified object in specified bucket
+func (a GatewayUnsupported) ListObjectVersions(ctx context.Context, bucket, prefix, marker, versionMarker, delimiter string, maxKeys int) (ListObjectVersionsInfo, error) {
+	logger.LogIf(ctx, NotImplemented{})
+	return ListObjectVersionsInfo{}, NotImplemented{}
+}
+
 // ListObjectParts returns all object parts for specified object in specified bucket
 func (a GatewayUnsupported) ListObjectParts(ctx context.Context, bucket string, object string, uploadID string, partNumberMarker int, maxParts int, opts ObjectOptions) (lpi ListPartsInfo, err error) {
 	logger.LogIf(ctx, NotImplemented{})
@@ -121,33 +128,45 @@ func (a GatewayUnsupported) DeleteBucketPolicy(ctx context.Context, bucket strin
 	return NotImplemented{}
 }
 
-// SetBucketLifecycle sets lifecycle on bucket
+// SetBucketVersioning enables versioning on a bucket.
+func (a GatewayUnsupported) SetBucketVersioning(ctx context.Context, bucket string, v *versioning.Versioning) error {
+	logger.LogIf(ctx, NotImplemented{})
+	return NotImplemented{}
+}
+
+// GetBucketVersioning retrieves versioning configuration of a bucket.
+func (a GatewayUnsupported) GetBucketVersioning(ctx context.Context, bucket string) (*versioning.Versioning, error) {
+	logger.LogIf(ctx, NotImplemented{})
+	return nil, NotImplemented{}
+}
+
+// SetBucketLifecycle enables lifecycle policies on a bucket.
 func (a GatewayUnsupported) SetBucketLifecycle(ctx context.Context, bucket string, lifecycle *lifecycle.Lifecycle) error {
 	logger.LogIf(ctx, NotImplemented{})
 	return NotImplemented{}
 }
 
-// GetBucketLifecycle will get lifecycle on bucket
+// GetBucketLifecycle retrieves lifecycle configuration of a bucket.
 func (a GatewayUnsupported) GetBucketLifecycle(ctx context.Context, bucket string) (*lifecycle.Lifecycle, error) {
 	return nil, NotImplemented{}
 }
 
-// DeleteBucketLifecycle deletes all lifecycle on bucket
+// DeleteBucketLifecycle deletes all lifecycle policies on a bucket
 func (a GatewayUnsupported) DeleteBucketLifecycle(ctx context.Context, bucket string) error {
 	return NotImplemented{}
 }
 
-// GetBucketSSEConfig returns bucket encryption config on given bucket
+// GetBucketSSEConfig returns bucket encryption config on a bucket
 func (a GatewayUnsupported) GetBucketSSEConfig(ctx context.Context, bucket string) (*bucketsse.BucketSSEConfig, error) {
 	return nil, NotImplemented{}
 }
 
-// SetBucketSSEConfig sets bucket encryption config on given bucket
+// SetBucketSSEConfig sets bucket encryption config on a bucket
 func (a GatewayUnsupported) SetBucketSSEConfig(ctx context.Context, bucket string, config *bucketsse.BucketSSEConfig) error {
 	return NotImplemented{}
 }
 
-// DeleteBucketSSEConfig deletes bucket encryption config on given bucket
+// DeleteBucketSSEConfig deletes bucket encryption config on a bucket
 func (a GatewayUnsupported) DeleteBucketSSEConfig(ctx context.Context, bucket string) error {
 	return NotImplemented{}
 }
@@ -173,7 +192,7 @@ func (a GatewayUnsupported) ListBucketsHeal(ctx context.Context) (buckets []Buck
 }
 
 // HealObject - Not implemented stub
-func (a GatewayUnsupported) HealObject(ctx context.Context, bucket, object string, opts madmin.HealOpts) (h madmin.HealResultItem, e error) {
+func (a GatewayUnsupported) HealObject(ctx context.Context, bucket, object, versionID string, opts madmin.HealOpts) (h madmin.HealResultItem, e error) {
 	return h, NotImplemented{}
 }
 
@@ -188,7 +207,7 @@ func (a GatewayUnsupported) Walk(ctx context.Context, bucket, prefix string, res
 }
 
 // HealObjects - Not implemented stub
-func (a GatewayUnsupported) HealObjects(ctx context.Context, bucket, prefix string, opts madmin.HealOpts, fn healObjectFn) (e error) {
+func (a GatewayUnsupported) HealObjects(ctx context.Context, bucket, prefix string, opts madmin.HealOpts, fn HealObjectFn) (e error) {
 	return NotImplemented{}
 }
 
@@ -205,19 +224,19 @@ func (a GatewayUnsupported) GetMetrics(ctx context.Context) (*Metrics, error) {
 }
 
 // PutObjectTags - not implemented.
-func (a GatewayUnsupported) PutObjectTags(ctx context.Context, bucket, object string, tags string) error {
+func (a GatewayUnsupported) PutObjectTags(ctx context.Context, bucket, object string, tags string, opts ObjectOptions) error {
 	logger.LogIf(ctx, NotImplemented{})
 	return NotImplemented{}
 }
 
 // GetObjectTags - not implemented.
-func (a GatewayUnsupported) GetObjectTags(ctx context.Context, bucket, object string) (*tags.Tags, error) {
+func (a GatewayUnsupported) GetObjectTags(ctx context.Context, bucket, object string, opts ObjectOptions) (*tags.Tags, error) {
 	logger.LogIf(ctx, NotImplemented{})
 	return nil, NotImplemented{}
 }
 
 // DeleteObjectTags - not implemented.
-func (a GatewayUnsupported) DeleteObjectTags(ctx context.Context, bucket, object string) error {
+func (a GatewayUnsupported) DeleteObjectTags(ctx context.Context, bucket, object string, opts ObjectOptions) error {
 	logger.LogIf(ctx, NotImplemented{})
 	return NotImplemented{}
 }

cmd/gateway/azure/gateway-azure.go

@@ -553,8 +553,8 @@ func (a *azureObjects) StorageInfo(ctx context.Context, _ bool) (si minio.Storag
 }
 
 // MakeBucketWithLocation - Create a new container on azure backend.
-func (a *azureObjects) MakeBucketWithLocation(ctx context.Context, bucket, location string, lockEnabled bool) error {
+func (a *azureObjects) MakeBucketWithLocation(ctx context.Context, bucket string, opts minio.BucketOptions) error {
-	if lockEnabled {
+	if opts.LockEnabled || opts.VersioningEnabled {
 		return minio.NotImplemented{}
 	}
 
@@ -966,21 +966,30 @@ func (a *azureObjects) CopyObject(ctx context.Context, srcBucket, srcObject, des
 
 // DeleteObject - Deletes a blob on azure container, uses Azure
 // equivalent `BlobURL.Delete`.
-func (a *azureObjects) DeleteObject(ctx context.Context, bucket, object string) error {
+func (a *azureObjects) DeleteObject(ctx context.Context, bucket, object string, opts minio.ObjectOptions) (minio.ObjectInfo, error) {
 	blob := a.client.NewContainerURL(bucket).NewBlobURL(object)
 	_, err := blob.Delete(ctx, azblob.DeleteSnapshotsOptionNone, azblob.BlobAccessConditions{})
 	if err != nil {
-		return azureToObjectError(err, bucket, object)
+		return minio.ObjectInfo{}, azureToObjectError(err, bucket, object)
 	}
-	return nil
+	return minio.ObjectInfo{
+		Bucket: bucket,
+		Name:   object,
+	}, nil
 }
 
-func (a *azureObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+func (a *azureObjects) DeleteObjects(ctx context.Context, bucket string, objects []minio.ObjectToDelete, opts minio.ObjectOptions) ([]minio.DeletedObject, []error) {
 	errs := make([]error, len(objects))
+	dobjects := make([]minio.DeletedObject, len(objects))
 	for idx, object := range objects {
-		errs[idx] = a.DeleteObject(ctx, bucket, object)
+		_, errs[idx] = a.DeleteObject(ctx, bucket, object.ObjectName, opts)
+		if errs[idx] == nil {
+			dobjects[idx] = minio.DeletedObject{
+				ObjectName: object.ObjectName,
 			}
-	return errs, nil
+		}
+	}
+	return dobjects, errs
 }
 
 // ListMultipartUploads - It's decided not to support List Multipart Uploads, hence returning empty result.

cmd/gateway/azure/gateway-azure_test.go

@@ -243,43 +243,6 @@ func TestAzureCodesToObjectError(t *testing.T) {
 	}
 }
 
-func TestAnonErrToObjectErr(t *testing.T) {
-	testCases := []struct {
-		name       string
-		statusCode int
-		params     []string
-		wantErr    error
-	}{
-		{"ObjectNotFound",
-			http.StatusNotFound,
-			[]string{"testBucket", "testObject"},
-			minio.ObjectNotFound{Bucket: "testBucket", Object: "testObject"},
-		},
-		{"BucketNotFound",
-			http.StatusNotFound,
-			[]string{"testBucket", ""},
-			minio.BucketNotFound{Bucket: "testBucket"},
-		},
-		{"ObjectNameInvalid",
-			http.StatusBadRequest,
-			[]string{"testBucket", "testObject"},
-			minio.ObjectNameInvalid{Bucket: "testBucket", Object: "testObject"},
-		},
-		{"BucketNameInvalid",
-			http.StatusBadRequest,
-			[]string{"testBucket", ""},
-			minio.BucketNameInvalid{Bucket: "testBucket"},
-		},
-	}
-	for _, test := range testCases {
-		t.Run(test.name, func(t *testing.T) {
-			if err := minio.AnonErrToObjectErr(test.statusCode, test.params...); !reflect.DeepEqual(err, test.wantErr) {
-				t.Errorf("anonErrToObjectErr() error = %v, wantErr %v", err, test.wantErr)
-			}
-		})
-	}
-}
-
 func TestCheckAzureUploadID(t *testing.T) {
 	invalidUploadIDs := []string{
 		"123456789abcdefg",

cmd/gateway/gcs/gateway-gcs.go

@@ -421,14 +421,15 @@ func (l *gcsGateway) StorageInfo(ctx context.Context, _ bool) (si minio.StorageI
 }
 
 // MakeBucketWithLocation - Create a new container on GCS backend.
-func (l *gcsGateway) MakeBucketWithLocation(ctx context.Context, bucket, location string, lockEnabled bool) error {
+func (l *gcsGateway) MakeBucketWithLocation(ctx context.Context, bucket string, opts minio.BucketOptions) error {
-	if lockEnabled {
+	if opts.LockEnabled || opts.VersioningEnabled {
 		return minio.NotImplemented{}
 	}
 
 	bkt := l.client.Bucket(bucket)
 
 	// we'll default to the us multi-region in case of us-east-1
+	location := opts.Location
 	if location == "us-east-1" {
 		location = "us"
 	}
@@ -958,22 +959,31 @@ func (l *gcsGateway) CopyObject(ctx context.Context, srcBucket string, srcObject
 }
 
 // DeleteObject - Deletes a blob in bucket
-func (l *gcsGateway) DeleteObject(ctx context.Context, bucket string, object string) error {
+func (l *gcsGateway) DeleteObject(ctx context.Context, bucket string, object string, opts minio.ObjectOptions) (minio.ObjectInfo, error) {
 	err := l.client.Bucket(bucket).Object(object).Delete(ctx)
 	if err != nil {
 		logger.LogIf(ctx, err)
-		return gcsToObjectError(err, bucket, object)
+		return minio.ObjectInfo{}, gcsToObjectError(err, bucket, object)
 	}
 
-	return nil
+	return minio.ObjectInfo{
+		Bucket: bucket,
+		Name:   object,
+	}, nil
 }
 
-func (l *gcsGateway) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+func (l *gcsGateway) DeleteObjects(ctx context.Context, bucket string, objects []minio.ObjectToDelete, opts minio.ObjectOptions) ([]minio.DeletedObject, []error) {
 	errs := make([]error, len(objects))
+	dobjects := make([]minio.DeletedObject, len(objects))
 	for idx, object := range objects {
-		errs[idx] = l.DeleteObject(ctx, bucket, object)
+		_, errs[idx] = l.DeleteObject(ctx, bucket, object.ObjectName, opts)
+		if errs[idx] == nil {
+			dobjects[idx] = minio.DeletedObject{
+				ObjectName: object.ObjectName,
 			}
-	return errs, nil
+		}
+	}
+	return dobjects, errs
 }
 
 // NewMultipartUpload - upload object in multiple parts

cmd/gateway/hdfs/gateway-hdfs.go

@@ -283,8 +283,8 @@ func (n *hdfsObjects) DeleteBucket(ctx context.Context, bucket string, forceDele
 	return hdfsToObjectErr(ctx, n.clnt.Remove(minio.PathJoin(hdfsSeparator, bucket)), bucket)
 }
 
-func (n *hdfsObjects) MakeBucketWithLocation(ctx context.Context, bucket, location string, lockEnabled bool) error {
+func (n *hdfsObjects) MakeBucketWithLocation(ctx context.Context, bucket string, opts minio.BucketOptions) error {
-	if lockEnabled {
+	if opts.LockEnabled || opts.VersioningEnabled {
 		return minio.NotImplemented{}
 	}
 
@@ -439,16 +439,26 @@ func (n *hdfsObjects) ListObjectsV2(ctx context.Context, bucket, prefix, continu
 	}, nil
 }
 
-func (n *hdfsObjects) DeleteObject(ctx context.Context, bucket, object string) error {
+func (n *hdfsObjects) DeleteObject(ctx context.Context, bucket, object string, opts minio.ObjectOptions) (minio.ObjectInfo, error) {
-	return hdfsToObjectErr(ctx, n.deleteObject(minio.PathJoin(hdfsSeparator, bucket), minio.PathJoin(hdfsSeparator, bucket, object)), bucket, object)
+	err := hdfsToObjectErr(ctx, n.deleteObject(minio.PathJoin(hdfsSeparator, bucket), minio.PathJoin(hdfsSeparator, bucket, object)), bucket, object)
+	return minio.ObjectInfo{
+		Bucket: bucket,
+		Name:   object,
+	}, err
 }
 
-func (n *hdfsObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+func (n *hdfsObjects) DeleteObjects(ctx context.Context, bucket string, objects []minio.ObjectToDelete, opts minio.ObjectOptions) ([]minio.DeletedObject, []error) {
 	errs := make([]error, len(objects))
+	dobjects := make([]minio.DeletedObject, len(objects))
 	for idx, object := range objects {
-		errs[idx] = n.DeleteObject(ctx, bucket, object)
+		_, errs[idx] = n.DeleteObject(ctx, bucket, object.ObjectName, opts)
+		if errs[idx] == nil {
+			dobjects[idx] = minio.DeletedObject{
+				ObjectName: object.ObjectName,
 			}
-	return errs, nil
+		}
+	}
+	return dobjects, errs
 }
 
 func (n *hdfsObjects) GetObjectNInfo(ctx context.Context, bucket, object string, rs *minio.HTTPRangeSpec, h http.Header, lockType minio.LockType, opts minio.ObjectOptions) (gr *minio.GetObjectReader, err error) {

cmd/gateway/s3/gateway-s3-sse.go

@@ -258,8 +258,8 @@ func getPartMetaPath(object, uploadID string, partID int) string {
 }
 
 // deletes the custom dare metadata file saved at the backend
-func (l *s3EncObjects) deleteGWMetadata(ctx context.Context, bucket, metaFileName string) error {
+func (l *s3EncObjects) deleteGWMetadata(ctx context.Context, bucket, metaFileName string) (minio.ObjectInfo, error) {
-	return l.s3Objects.DeleteObject(ctx, bucket, metaFileName)
+	return l.s3Objects.DeleteObject(ctx, bucket, metaFileName, minio.ObjectOptions{})
 }
 
 func (l *s3EncObjects) getObject(ctx context.Context, bucket string, key string, startOffset int64, length int64, writer io.Writer, etag string, opts minio.ObjectOptions) error {
@@ -381,14 +381,14 @@ func (l *s3EncObjects) CopyObject(ctx context.Context, srcBucket string, srcObje
 // DeleteObject deletes a blob in bucket
 // For custom gateway encrypted large objects, cleans up encrypted content and metadata files
 // from the backend.
-func (l *s3EncObjects) DeleteObject(ctx context.Context, bucket string, object string) error {
+func (l *s3EncObjects) DeleteObject(ctx context.Context, bucket string, object string, opts minio.ObjectOptions) (minio.ObjectInfo, error) {
-
 	// Get dare meta json
 	if _, err := l.getGWMetadata(ctx, bucket, getDareMetaPath(object)); err != nil {
-		return l.s3Objects.DeleteObject(ctx, bucket, object)
+		logger.LogIf(minio.GlobalContext, err)
+		return l.s3Objects.DeleteObject(ctx, bucket, object, opts)
 	}
 	// delete encrypted object
-	l.s3Objects.DeleteObject(ctx, bucket, getGWContentPath(object))
+	l.s3Objects.DeleteObject(ctx, bucket, getGWContentPath(object), opts)
 	return l.deleteGWMetadata(ctx, bucket, getDareMetaPath(object))
 }
 
@@ -446,7 +446,7 @@ func (l *s3EncObjects) PutObject(ctx context.Context, bucket string, object stri
 	}
 	if opts.ServerSideEncryption == nil {
 		defer l.deleteGWMetadata(ctx, bucket, getDareMetaPath(object))
-		defer l.DeleteObject(ctx, bucket, getGWContentPath(object))
+		defer l.DeleteObject(ctx, bucket, getGWContentPath(object), opts)
 		return l.s3Objects.PutObject(ctx, bucket, object, data, minio.ObjectOptions{UserDefined: opts.UserDefined})
 	}
 
@@ -470,7 +470,7 @@ func (l *s3EncObjects) PutObject(ctx context.Context, bucket string, object stri
 	}
 	objInfo = gwMeta.ToObjectInfo(bucket, object)
 	// delete any unencrypted content of the same name created previously
-	l.s3Objects.DeleteObject(ctx, bucket, object)
+	l.s3Objects.DeleteObject(ctx, bucket, object, opts)
 	return objInfo, nil
 }
 
@@ -586,7 +586,7 @@ func (l *s3EncObjects) AbortMultipartUpload(ctx context.Context, bucket string,
 			return minio.InvalidUploadID{UploadID: uploadID}
 		}
 		for _, obj := range loi.Objects {
-			if err := l.s3Objects.DeleteObject(ctx, bucket, obj.Name); err != nil {
+			if _, err := l.s3Objects.DeleteObject(ctx, bucket, obj.Name, minio.ObjectOptions{}); err != nil {
 				return minio.ErrorRespToObjectError(err)
 			}
 			startAfter = obj.Name
@@ -608,7 +608,7 @@ func (l *s3EncObjects) CompleteMultipartUpload(ctx context.Context, bucket, obje
 		if e == nil {
 			// delete any encrypted version of object that might exist
 			defer l.deleteGWMetadata(ctx, bucket, getDareMetaPath(object))
-			defer l.DeleteObject(ctx, bucket, getGWContentPath(object))
+			defer l.DeleteObject(ctx, bucket, getGWContentPath(object), opts)
 		}
 		return oi, e
 	}
@@ -640,7 +640,7 @@ func (l *s3EncObjects) CompleteMultipartUpload(ctx context.Context, bucket, obje
 	}
 
 	//delete any unencrypted version of object that might be on the backend
-	defer l.s3Objects.DeleteObject(ctx, bucket, object)
+	defer l.s3Objects.DeleteObject(ctx, bucket, object, opts)
 
 	// Save the final object size and modtime.
 	gwMeta.Stat.Size = objectSize
@@ -665,7 +665,7 @@ func (l *s3EncObjects) CompleteMultipartUpload(ctx context.Context, bucket, obje
 				break
 			}
 			startAfter = obj.Name
-			l.s3Objects.DeleteObject(ctx, bucket, obj.Name)
+			l.s3Objects.DeleteObject(ctx, bucket, obj.Name, opts)
 		}
 		continuationToken = loi.NextContinuationToken
 		if !loi.IsTruncated || done {
@@ -716,7 +716,7 @@ func (l *s3EncObjects) cleanupStaleEncMultipartUploadsOnGW(ctx context.Context,
 		for _, b := range buckets {
 			expParts := l.getStalePartsForBucket(ctx, b.Name, expiry)
 			for k := range expParts {
-				l.s3Objects.DeleteObject(ctx, b.Name, k)
+				l.s3Objects.DeleteObject(ctx, b.Name, k, minio.ObjectOptions{})
 			}
 		}
 	}
@@ -783,7 +783,7 @@ func (l *s3EncObjects) DeleteBucket(ctx context.Context, bucket string, forceDel
 		}
 	}
 	for k := range expParts {
-		l.s3Objects.DeleteObject(ctx, bucket, k)
+		l.s3Objects.DeleteObject(ctx, bucket, k, minio.ObjectOptions{})
 	}
 	err := l.Client.RemoveBucket(bucket)
 	if err != nil {

cmd/gateway/s3/gateway-s3.go

@@ -287,8 +287,8 @@ func (l *s3Objects) StorageInfo(ctx context.Context, _ bool) (si minio.StorageIn
 }
 
 // MakeBucket creates a new container on S3 backend.
-func (l *s3Objects) MakeBucketWithLocation(ctx context.Context, bucket, location string, lockEnabled bool) error {
+func (l *s3Objects) MakeBucketWithLocation(ctx context.Context, bucket string, opts minio.BucketOptions) error {
-	if lockEnabled {
+	if opts.LockEnabled || opts.VersioningEnabled {
 		return minio.NotImplemented{}
 	}
 
@@ -302,7 +302,7 @@ func (l *s3Objects) MakeBucketWithLocation(ctx context.Context, bucket, location
 	if s3utils.CheckValidBucketName(bucket) != nil {
 		return minio.BucketNameInvalid{Bucket: bucket}
 	}
-	err := l.Client.MakeBucket(bucket, location)
+	err := l.Client.MakeBucket(bucket, opts.Location)
 	if err != nil {
 		return minio.ErrorRespToObjectError(err, bucket)
 	}
@@ -518,21 +518,30 @@ func (l *s3Objects) CopyObject(ctx context.Context, srcBucket string, srcObject
 }
 
 // DeleteObject deletes a blob in bucket
-func (l *s3Objects) DeleteObject(ctx context.Context, bucket string, object string) error {
+func (l *s3Objects) DeleteObject(ctx context.Context, bucket string, object string, opts minio.ObjectOptions) (minio.ObjectInfo, error) {
 	err := l.Client.RemoveObject(bucket, object)
 	if err != nil {
-		return minio.ErrorRespToObjectError(err, bucket, object)
+		return minio.ObjectInfo{}, minio.ErrorRespToObjectError(err, bucket, object)
 	}
 
-	return nil
+	return minio.ObjectInfo{
+		Bucket: bucket,
+		Name:   object,
+	}, nil
 }
 
-func (l *s3Objects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+func (l *s3Objects) DeleteObjects(ctx context.Context, bucket string, objects []minio.ObjectToDelete, opts minio.ObjectOptions) ([]minio.DeletedObject, []error) {
 	errs := make([]error, len(objects))
+	dobjects := make([]minio.DeletedObject, len(objects))
 	for idx, object := range objects {
-		errs[idx] = l.DeleteObject(ctx, bucket, object)
+		_, errs[idx] = l.DeleteObject(ctx, bucket, object.ObjectName, opts)
+		if errs[idx] == nil {
+			dobjects[idx] = minio.DeletedObject{
+				ObjectName: object.ObjectName,
 			}
-	return errs, nil
+		}
+	}
+	return dobjects, errs
 }
 
 // ListMultipartUploads lists all multipart uploads.
@@ -700,11 +709,10 @@ func (l *s3Objects) DeleteBucketPolicy(ctx context.Context, bucket string) error
 }
 
 // GetObjectTags gets the tags set on the object
-func (l *s3Objects) GetObjectTags(ctx context.Context, bucket string, object string) (*tags.Tags, error) {
+func (l *s3Objects) GetObjectTags(ctx context.Context, bucket string, object string, opts minio.ObjectOptions) (*tags.Tags, error) {
 	var err error
 	var tagObj *tags.Tags
 	var tagStr string
-	var opts minio.ObjectOptions
 
 	if _, err = l.GetObjectInfo(ctx, bucket, object, opts); err != nil {
 		return nil, minio.ErrorRespToObjectError(err, bucket, object)
@@ -721,7 +729,7 @@ func (l *s3Objects) GetObjectTags(ctx context.Context, bucket string, object str
 }
 
 // PutObjectTags attaches the tags to the object
-func (l *s3Objects) PutObjectTags(ctx context.Context, bucket, object string, tagStr string) error {
+func (l *s3Objects) PutObjectTags(ctx context.Context, bucket, object string, tagStr string, opts minio.ObjectOptions) error {
 	tagObj, err := tags.Parse(tagStr, true)
 	if err != nil {
 		return minio.ErrorRespToObjectError(err, bucket, object)
@@ -733,7 +741,7 @@ func (l *s3Objects) PutObjectTags(ctx context.Context, bucket, object string, ta
 }
 
 // DeleteObjectTags removes the tags attached to the object
-func (l *s3Objects) DeleteObjectTags(ctx context.Context, bucket, object string) error {
+func (l *s3Objects) DeleteObjectTags(ctx context.Context, bucket, object string, opts minio.ObjectOptions) error {
 	if err := l.Client.RemoveObjectTagging(bucket, object); err != nil {
 		return minio.ErrorRespToObjectError(err, bucket, object)
 	}

cmd/generic-handlers.go

@@ -103,7 +103,7 @@ func isHTTPHeaderSizeTooLarge(header http.Header) bool {
 		length := len(key) + len(header.Get(key))
 		size += length
 		for _, prefix := range userMetadataKeyPrefixes {
-			if HasPrefix(key, prefix) {
+			if strings.HasPrefix(strings.ToLower(key), prefix) {
 				usersize += length
 				break
 			}
@@ -444,74 +444,75 @@ func setIgnoreResourcesHandler(h http.Handler) http.Handler {
 	return resourceHandler{h}
 }
 
+var supportedDummyBucketAPIs = map[string][]string{
+	"acl":            {http.MethodPut, http.MethodGet},
+	"cors":           {http.MethodGet},
+	"website":        {http.MethodGet, http.MethodDelete},
+	"logging":        {http.MethodGet},
+	"accelerate":     {http.MethodGet},
+	"replication":    {http.MethodGet},
+	"requestPayment": {http.MethodGet},
+}
+
+// List of not implemented bucket queries
+var notImplementedBucketResourceNames = map[string]struct{}{
+	"cors":           {},
+	"metrics":        {},
+	"website":        {},
+	"logging":        {},
+	"inventory":      {},
+	"accelerate":     {},
+	"replication":    {},
+	"requestPayment": {},
+}
+
 // Checks requests for not implemented Bucket resources
 func ignoreNotImplementedBucketResources(req *http.Request) bool {
 	for name := range req.URL.Query() {
-		// Enable PutBucketACL, GetBucketACL, GetBucketCors,
+		methods, ok := supportedDummyBucketAPIs[name]
-		// GetBucketWebsite, GetBucketAcccelerate,
+		if ok {
-		// GetBucketRequestPayment, GetBucketLogging,
+			for _, method := range methods {
-		// GetBucketLifecycle, GetBucketReplication,
+				if method == req.Method {
-		// GetBucketTagging, GetBucketVersioning,
-		// DeleteBucketTagging, and DeleteBucketWebsite
-		// dummy calls specifically.
-		if name == "acl" && req.Method == http.MethodPut {
 					return false
 				}
-		if ((name == "acl" ||
+			}
-			name == "cors" ||
-			name == "website" ||
-			name == "accelerate" ||
-			name == "requestPayment" ||
-			name == "logging" ||
-			name == "lifecycle" ||
-			name == "replication" ||
-			name == "tagging" ||
-			name == "versioning") && req.Method == http.MethodGet) ||
-			((name == "tagging" ||
-				name == "website") && req.Method == http.MethodDelete) {
-			return false
 		}
 
-		if notImplementedBucketResourceNames[name] {
+		if _, ok := notImplementedBucketResourceNames[name]; ok {
 			return true
 		}
 	}
 	return false
 }
 
+var supportedDummyObjectAPIs = map[string][]string{
+	"acl": {http.MethodGet, http.MethodPut},
+}
+
+// List of not implemented object APIs
+var notImplementedObjectResourceNames = map[string]struct{}{
+	"restore": {},
+	"torrent": {},
+}
+
 // Checks requests for not implemented Object resources
 func ignoreNotImplementedObjectResources(req *http.Request) bool {
 	for name := range req.URL.Query() {
-		// Enable Get/PutObjectACL dummy call specifically.
+		methods, ok := supportedDummyObjectAPIs[name]
-		if name == "acl" && (req.Method == http.MethodGet || req.Method == http.MethodPut) {
+		if ok {
+			for _, method := range methods {
+				if method == req.Method {
 					return false
 				}
-		if notImplementedObjectResourceNames[name] {
+			}
+		}
+		if _, ok := notImplementedObjectResourceNames[name]; ok {
 			return true
 		}
 	}
 	return false
 }
 
-// List of not implemented bucket queries
-var notImplementedBucketResourceNames = map[string]bool{
-	"accelerate":     true,
-	"cors":           true,
-	"inventory":      true,
-	"logging":        true,
-	"metrics":        true,
-	"replication":    true,
-	"requestPayment": true,
-	"versioning":     true,
-	"website":        true,
-}
-
-// List of not implemented object queries
-var notImplementedObjectResourceNames = map[string]bool{
-	"restore": true,
-	"torrent": true,
-}
-
 // Resource handler ServeHTTP() wrapper
 func (h resourceHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 	bucketName, objectName := request2BucketObjectName(r)

cmd/generic-handlers_test.go

@@ -199,12 +199,16 @@ var containsReservedMetadataTests = []struct {
 }
 
 func TestContainsReservedMetadata(t *testing.T) {
-	for i, test := range containsReservedMetadataTests {
+	for _, test := range containsReservedMetadataTests {
-		if contains := containsReservedMetadata(test.header); contains && !test.shouldFail {
+		test := test
-			t.Errorf("Test %d: contains reserved header but should not fail", i)
+		t.Run("", func(t *testing.T) {
+			contains := containsReservedMetadata(test.header)
+			if contains && !test.shouldFail {
+				t.Errorf("contains reserved header but should not fail")
 			} else if !contains && test.shouldFail {
-			t.Errorf("Test %d: does not contain reserved header but failed", i)
+				t.Errorf("does not contain reserved header but failed")
 			}
+		})
 	}
 }
 

cmd/global-heal.go

@@ -79,7 +79,7 @@ func getLocalBackgroundHealStatus() madmin.BgHealState {
 }
 
 // healErasureSet lists and heals all objects in a specific erasure set
-func healErasureSet(ctx context.Context, setIndex int, xlObj *xlObjects, drivesPerSet int) error {
+func healErasureSet(ctx context.Context, setIndex int, xlObj *erasureObjects, drivesPerSet int) error {
 	buckets, err := xlObj.ListBuckets(ctx)
 	if err != nil {
 		return err
@@ -105,32 +105,34 @@ func healErasureSet(ctx context.Context, setIndex int, xlObj *xlObjects, drivesP
 	for _, bucket := range buckets {
 		// Heal current bucket
 		bgSeq.sourceCh <- healSource{
-			path: bucket.Name,
+			bucket: bucket.Name,
 		}
 
-		var entryChs []FileInfoCh
+		var entryChs []FileInfoVersionsCh
 		for _, disk := range xlObj.getLoadBalancedDisks() {
 			if disk == nil {
 				// Disk can be offline
 				continue
 			}
-			entryCh, err := disk.Walk(bucket.Name, "", "", true, xlMetaJSONFile, readMetadata, ctx.Done())
+
+			entryCh, err := disk.WalkVersions(bucket.Name, "", "", true, ctx.Done())
 			if err != nil {
 				// Disk walk returned error, ignore it.
 				continue
 			}
-			entryChs = append(entryChs, FileInfoCh{
+
+			entryChs = append(entryChs, FileInfoVersionsCh{
 				Ch: entryCh,
 			})
 		}
 
 		entriesValid := make([]bool, len(entryChs))
-		entries := make([]FileInfo, len(entryChs))
+		entries := make([]FileInfoVersions, len(entryChs))
 
 		for {
-			entry, quorumCount, ok := lexicallySortedEntry(entryChs, entries, entriesValid)
+			entry, quorumCount, ok := lexicallySortedEntryVersions(entryChs, entries, entriesValid)
 			if !ok {
-				return nil
+				break
 			}
 
 			if quorumCount == drivesPerSet {
@@ -138,8 +140,12 @@ func healErasureSet(ctx context.Context, setIndex int, xlObj *xlObjects, drivesP
 				continue
 			}
 
+			for _, version := range entry.Versions {
 				bgSeq.sourceCh <- healSource{
-				path: pathJoin(bucket.Name, entry.Name),
+					bucket:    bucket.Name,
+					object:    version.Name,
+					versionID: version.VersionID,
+				}
 			}
 		}
 	}
@@ -148,12 +154,14 @@ func healErasureSet(ctx context.Context, setIndex int, xlObj *xlObjects, drivesP
 }
 
 // deepHealObject heals given object path in deep to fix bitrot.
-func deepHealObject(objectPath string) {
+func deepHealObject(bucket, object, versionID string) {
 	// Get background heal sequence to send elements to heal
 	bgSeq, _ := globalBackgroundHealState.getHealSequenceByToken(bgHealingUUID)
 
 	bgSeq.sourceCh <- healSource{
-		path: objectPath,
+		bucket:    bucket,
+		object:    object,
+		versionID: versionID,
 		opts:      &madmin.HealOpts{ScanMode: madmin.HealDeepScan},
 	}
 }
@@ -172,7 +180,7 @@ func durationToNextHealRound(lastHeal time.Time) time.Duration {
 }
 
 // Healing leader will take the charge of healing all erasure sets
-func execLeaderTasks(ctx context.Context, z *xlZones) {
+func execLeaderTasks(ctx context.Context, z *erasureZones) {
 	// So that we don't heal immediately, but after one month.
 	lastScanTime := UTCNow()
 	// Get background heal sequence to send elements to heal
@@ -211,7 +219,7 @@ func execLeaderTasks(ctx context.Context, z *xlZones) {
 }
 
 func startGlobalHeal(ctx context.Context, objAPI ObjectLayer) {
-	zones, ok := objAPI.(*xlZones)
+	zones, ok := objAPI.(*erasureZones)
 	if !ok {
 		return
 	}

cmd/globals.go

@@ -61,8 +61,8 @@ const (
 	globalNetBSDOSName             = "netbsd"
 	globalMacOSName                = "darwin"
 	globalMinioModeFS              = "mode-server-fs"
-	globalMinioModeXL              = "mode-server-xl"
+	globalMinioModeErasure         = "mode-server-xl"
-	globalMinioModeDistXL          = "mode-server-distributed-xl"
+	globalMinioModeDistErasure     = "mode-server-distributed-xl"
 	globalMinioModeGatewayPrefix   = "mode-gateway-"
 
 	// Add new global values here.
@@ -107,13 +107,13 @@ var globalCLIContext = struct {
 
 var (
 	// Indicates set drive count.
-	globalXLSetDriveCount int
+	globalErasureSetDriveCount int
 
 	// Indicates if the running minio server is distributed setup.
-	globalIsDistXL = false
+	globalIsDistErasure = false
 
 	// Indicates if the running minio server is an erasure-code backend.
-	globalIsXL = false
+	globalIsErasure = false
 
 	// Indicates if the running minio is in gateway mode.
 	globalIsGateway = false
@@ -215,6 +215,7 @@ var (
 
 	globalBucketObjectLockSys *BucketObjectLockSys
 	globalBucketQuotaSys      *BucketQuotaSys
+	globalBucketVersioningSys *BucketVersioningSys
 
 	// Disk cache drives
 	globalCacheConfig cache.Config

cmd/handler-utils.go

@@ -445,7 +445,7 @@ func errorResponseHandler(w http.ResponseWriter, r *http.Request) {
 
 // gets host name for current node
 func getHostName(r *http.Request) (hostName string) {
-	if globalIsDistXL {
+	if globalIsDistErasure {
 		hostName = GetLocalPeer(globalEndpoints)
 	} else {
 		hostName = r.Host

cmd/http-tracer.go

@@ -114,7 +114,7 @@ func Trace(f http.HandlerFunc, logBody bool, w http.ResponseWriter, r *http.Requ
 	reqBodyRecorder = &recordRequest{Reader: r.Body, logBody: logBody, headers: reqHeaders}
 	r.Body = ioutil.NopCloser(reqBodyRecorder)
 	t.NodeName = r.Host
-	if globalIsDistXL {
+	if globalIsDistErasure {
 		t.NodeName = GetLocalPeer(globalEndpoints)
 	}
 	// strip port from the host address

cmd/http/headers.go

@@ -56,6 +56,10 @@ const (
 	// S3 storage class
 	AmzStorageClass = "x-amz-storage-class"
 
+	// S3 object version ID
+	AmzVersionID    = "x-amz-version-id"
+	AmzDeleteMarker = "x-amz-delete-marker"
+
 	// S3 object tagging
 	AmzObjectTagging = "X-Amz-Tagging"
 	AmzTagCount      = "x-amz-tagging-count"

cmd/iam.go

@@ -469,7 +469,7 @@ func (sys *IAMSys) Init(ctx context.Context, objAPI ObjectLayer) {
 		}
 
 		// These messages only meant primarily for distributed setup, so only log during distributed setup.
-		if globalIsDistXL {
+		if globalIsDistErasure {
 			logger.Info("Waiting for all MinIO IAM sub-system to be initialized.. lock acquired")
 		}