Add connection close proactively for Walk() http/rpc (#7645)

errs was passed to many goroutines but they are all allowed
to update errs if any error happens during deletion, which
can cause a data race.

This commit will avoid issuing bulk delete operations in parallel
to avoid the warning race.

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,46 @@
+---
+name: Bug report
+about: Create a report to help us improve
+title: ''
+labels: community, triage
+assignees: ''
+
+---
+
+<!--- Provide a general summary of the issue in the Title above -->
+
+## Expected Behavior
+<!--- If you're describing a bug, tell us what should happen -->
+<!--- If you're suggesting a change/improvement, tell us how it should work -->
+
+## Current Behavior
+<!--- If describing a bug, tell us what happens instead of the expected behavior -->
+<!--- If suggesting a change/improvement, explain the difference from current behavior -->
+
+## Possible Solution
+<!--- Not obligatory, but suggest a fix/reason for the bug, -->
+<!--- or ideas how to implement the addition or change -->
+
+## Steps to Reproduce (for bugs)
+<!--- Provide a link to a live example, or an unambiguous set of steps to -->
+<!--- reproduce this bug. Include code to reproduce, if relevant -->
+1.
+2.
+3.
+4.
+
+## Context
+<!--- How has this issue affected you? What are you trying to accomplish? -->
+<!--- Providing context helps us come up with a solution that is most useful in the real world -->
+
+## Regression
+<!-- Is this issue a regression? (Yes / No) -->
+<!-- If Yes, optionally please include minio version or commit id or PR# that caused this regression, if you have these details. -->
+
+## Your Environment
+<!--- Include as many relevant details about the environment you experienced the bug in -->
+* Version used (`minio version`):
+* Environment name and version (e.g. nginx 1.9.1):
+* Server type and version:
+* Operating System and version (`uname -a`):
+* Link to your project:

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,20 @@
+---
+name: Feature request
+about: Suggest an idea for this project
+title: ''
+labels: community, triage
+assignees: ''
+
+---
+
+**Is your feature request related to a problem? Please describe.**
+A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
+
+**Describe the solution you'd like**
+A clear and concise description of what you want to happen.
+
+**Describe alternatives you've considered**
+A clear and concise description of any alternative solutions or features you've considered.
+
+**Additional context**
+Add any other context or screenshots about the feature request here.

Makefile

@@ -74,7 +74,7 @@ docker: build
 # Builds minio and installs it to $GOPATH/bin.
 install: build
 	@echo "Installing minio binary to '$(GOPATH)/bin/minio'"
-	@mkdir -p $(GOPATH)/bin && cp -uf $(PWD)/minio $(GOPATH)/bin/minio
+	@mkdir -p $(GOPATH)/bin && cp -f $(PWD)/minio $(GOPATH)/bin/minio
 	@echo "Installation successful. To learn more, try \"minio --help\"."
 
 clean:

cmd/admin-handlers.go

@@ -35,7 +35,6 @@ import (
 	"github.com/tidwall/sjson"
 
 	"github.com/minio/minio/cmd/logger"
-	"github.com/minio/minio/pkg/auth"
 	"github.com/minio/minio/pkg/cpu"
 	"github.com/minio/minio/pkg/disk"
 	"github.com/minio/minio/pkg/handlers"
@@ -1416,70 +1415,3 @@ func (a adminAPIHandlers) SetConfigKeysHandler(w http.ResponseWriter, r *http.Re
 	// Send success response
 	writeSuccessResponseHeadersOnly(w)
 }
-
-// UpdateAdminCredsHandler - POST /minio/admin/v1/config/credential
-// ----------
-// Update admin credentials in a minio server
-func (a adminAPIHandlers) UpdateAdminCredentialsHandler(w http.ResponseWriter,
-	r *http.Request) {
-
-	ctx := newContext(r, w, "UpdateCredentialsHandler")
-
-	objectAPI := validateAdminReq(ctx, w, r)
-	if objectAPI == nil {
-		return
-	}
-
-	// Avoid setting new credentials when they are already passed
-	// by the environment. Deny if WORM is enabled.
-	if globalIsEnvCreds || globalWORMEnabled {
-		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrMethodNotAllowed), r.URL)
-		return
-	}
-
-	if r.ContentLength > maxEConfigJSONSize || r.ContentLength == -1 {
-		// More than maxConfigSize bytes were available
-		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAdminConfigTooLarge), r.URL)
-		return
-	}
-
-	password := globalServerConfig.GetCredential().SecretKey
-	configBytes, err := madmin.DecryptData(password, io.LimitReader(r.Body, r.ContentLength))
-	if err != nil {
-		logger.LogIf(ctx, err)
-		writeCustomErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrAdminConfigBadJSON), err.Error(), r.URL)
-		return
-	}
-
-	// Decode request body
-	var req madmin.SetCredsReq
-	if err = json.Unmarshal(configBytes, &req); err != nil {
-		logger.LogIf(ctx, err)
-		writeErrorResponseJSON(ctx, w, errorCodes.ToAPIErr(ErrRequestBodyParse), r.URL)
-		return
-	}
-
-	creds, err := auth.CreateCredentials(req.AccessKey, req.SecretKey)
-	if err != nil {
-		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
-		return
-	}
-
-	// Acquire lock before updating global configuration.
-	globalServerConfigMu.Lock()
-	defer globalServerConfigMu.Unlock()
-
-	// Update local credentials in memory.
-	globalServerConfig.SetCredential(creds)
-
-	// Set active creds.
-	globalActiveCred = creds
-
-	if err = saveServerConfig(ctx, objectAPI, globalServerConfig); err != nil {
-		writeErrorResponseJSON(ctx, w, toAdminAPIErr(ctx, err), r.URL)
-		return
-	}
-
-	// Reply to the client before restarting minio server.
-	writeSuccessResponseHeadersOnly(w)
-}

cmd/admin-handlers_test.go

@@ -20,7 +20,6 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
-	"fmt"
 	"io"
 	"io/ioutil"
 	"net/http"
@@ -298,60 +297,6 @@ func (atb *adminXLTestBed) TearDown() {
 	resetTestGlobals()
 }
 
-func (atb *adminXLTestBed) GenerateHealTestData(t *testing.T) {
-	// Create an object myobject under bucket mybucket.
-	bucketName := "mybucket"
-	err := atb.objLayer.MakeBucketWithLocation(context.Background(), bucketName, "")
-	if err != nil {
-		t.Fatalf("Failed to make bucket %s - %v", bucketName,
-			err)
-	}
-
-	// create some objects
-	{
-		objName := "myobject"
-		for i := 0; i < 10; i++ {
-			objectName := fmt.Sprintf("%s-%d", objName, i)
-			_, err = atb.objLayer.PutObject(context.Background(), bucketName, objectName,
-				mustGetPutObjReader(t, bytes.NewReader([]byte("hello")),
-					int64(len("hello")), "", ""), ObjectOptions{})
-			if err != nil {
-				t.Fatalf("Failed to create %s - %v", objectName,
-					err)
-			}
-		}
-	}
-
-	// create a multipart upload (incomplete)
-	{
-		objName := "mpObject"
-		uploadID, err := atb.objLayer.NewMultipartUpload(context.Background(), bucketName,
-			objName, ObjectOptions{})
-		if err != nil {
-			t.Fatalf("mp new error: %v", err)
-		}
-
-		_, err = atb.objLayer.PutObjectPart(context.Background(), bucketName, objName,
-			uploadID, 3, mustGetPutObjReader(t, bytes.NewReader(
-				[]byte("hello")), int64(len("hello")), "", ""), ObjectOptions{})
-		if err != nil {
-			t.Fatalf("mp put error: %v", err)
-		}
-
-	}
-}
-
-func (atb *adminXLTestBed) CleanupHealTestData(t *testing.T) {
-	bucketName := "mybucket"
-	objName := "myobject"
-	for i := 0; i < 10; i++ {
-		atb.objLayer.DeleteObject(context.Background(), bucketName,
-			fmt.Sprintf("%s-%d", objName, i))
-	}
-
-	atb.objLayer.DeleteBucket(context.Background(), bucketName)
-}
-
 // initTestObjLayer - Helper function to initialize an XL-based object
 // layer and set globalObjectAPI.
 func initTestXLObjLayer() (ObjectLayer, []string, error) {
@@ -599,86 +544,6 @@ func TestServiceRestartHandler(t *testing.T) {
 	testServicesCmdHandler(restartCmd, t)
 }
 
-// Test for service set creds management REST API.
-func TestServiceSetCreds(t *testing.T) {
-	adminTestBed, err := prepareAdminXLTestBed()
-	if err != nil {
-		t.Fatal("Failed to initialize a single node XL backend for admin handler tests.")
-	}
-	defer adminTestBed.TearDown()
-
-	// Initialize admin peers to make admin RPC calls. Note: In a
-	// single node setup, this degenerates to a simple function
-	// call under the hood.
-	globalMinioAddr = "127.0.0.1:9000"
-
-	credentials := globalServerConfig.GetCredential()
-
-	testCases := []struct {
-		AccessKey          string
-		SecretKey          string
-		EnvKeysSet         bool
-		ExpectedStatusCode int
-	}{
-		// Bad secret key
-		{"minio", "minio", false, http.StatusBadRequest},
-		// Bad  secret key set from the env
-		{"minio", "minio", true, http.StatusMethodNotAllowed},
-		// Good keys set from the env
-		{"minio", "minio123", true, http.StatusMethodNotAllowed},
-		// Successful operation should be the last one to
-		// not change server credentials during tests.
-		{"minio", "minio123", false, http.StatusOK},
-	}
-	for i, testCase := range testCases {
-		// Set or unset environement keys
-		globalIsEnvCreds = testCase.EnvKeysSet
-
-		// Construct setCreds request body
-		body, err := json.Marshal(madmin.SetCredsReq{
-			AccessKey: testCase.AccessKey,
-			SecretKey: testCase.SecretKey})
-		if err != nil {
-			t.Fatalf("JSONify err: %v", err)
-		}
-
-		ebody, err := madmin.EncryptData(credentials.SecretKey, body)
-		if err != nil {
-			t.Fatal(err)
-		}
-
-		// Construct setCreds request
-		req, err := getServiceCmdRequest(setCreds, credentials, ebody)
-		if err != nil {
-			t.Fatalf("Failed to build service status request %v", err)
-		}
-
-		rec := httptest.NewRecorder()
-
-		// Execute request
-		adminTestBed.router.ServeHTTP(rec, req)
-
-		// Check if the http code response is expected
-		if rec.Code != testCase.ExpectedStatusCode {
-			t.Errorf("Test %d: Wrong status code, expected = %d, found = %d", i+1, testCase.ExpectedStatusCode, rec.Code)
-			resp, _ := ioutil.ReadAll(rec.Body)
-			t.Errorf("Expected to receive %d status code but received %d. Body (%s)",
-				http.StatusOK, rec.Code, string(resp))
-		}
-
-		// If we got 200 OK, check if new credentials are really set
-		if rec.Code == http.StatusOK {
-			cred := globalServerConfig.GetCredential()
-			if cred.AccessKey != testCase.AccessKey {
-				t.Errorf("Test %d: Wrong access key, expected = %s, found = %s", i+1, testCase.AccessKey, cred.AccessKey)
-			}
-			if cred.SecretKey != testCase.SecretKey {
-				t.Errorf("Test %d: Wrong secret key, expected = %s, found = %s", i+1, testCase.SecretKey, cred.SecretKey)
-			}
-		}
-	}
-}
-
 // buildAdminRequest - helper function to build an admin API request.
 func buildAdminRequest(queryVal url.Values, method, path string,
 	contentLength int64, bodySeeker io.ReadSeeker) (*http.Request, error) {

cmd/admin-router.go

@@ -74,8 +74,6 @@ func registerAdminRouter(router *mux.Router, enableConfigOps, enableIAMOps bool)
 
 	/// Config operations
 	if enableConfigOps {
-		// Update credentials
-		adminV1Router.Methods(http.MethodPut).Path("/config/credential").HandlerFunc(httpTraceHdrs(adminAPI.UpdateAdminCredentialsHandler))
 		// Get config
 		adminV1Router.Methods(http.MethodGet).Path("/config").HandlerFunc(httpTraceHdrs(adminAPI.GetConfigHandler))
 		// Set config

cmd/api-errors.go

@@ -65,6 +65,7 @@ const (
 	ErrBadDigest
 	ErrEntityTooSmall
 	ErrEntityTooLarge
+	ErrPolicyTooLarge
 	ErrIncompleteBody
 	ErrInternalError
 	ErrInvalidAccessKeyID
@@ -398,6 +399,11 @@ var errorCodes = errorCodeMap{
 		Description:    "Your proposed upload exceeds the maximum allowed object size.",
 		HTTPStatusCode: http.StatusBadRequest,
 	},
+	ErrPolicyTooLarge: {
+		Code:           "PolicyTooLarge",
+		Description:    "Policy exceeds the maximum allowed document size.",
+		HTTPStatusCode: http.StatusBadRequest,
+	},
 	ErrIncompleteBody: {
 		Code:           "IncompleteBody",
 		Description:    "You did not provide the number of bytes specified by the Content-Length HTTP header.",
@@ -1479,7 +1485,6 @@ func toAPIErrorCode(ctx context.Context, err error) (apiErr APIErrorCode) {
 	if err == nil {
 		return ErrNone
 	}
-
 	// Verify if the underlying error is signature mismatch.
 	switch err {
 	case errInvalidArgument:
@@ -1529,6 +1534,8 @@ func toAPIErrorCode(ctx context.Context, err error) (apiErr APIErrorCode) {
 		apiErr = ErrKMSAuthFailure
 	case errOperationTimedOut, context.Canceled, context.DeadlineExceeded:
 		apiErr = ErrOperationTimedOut
+	case errNetworkConnReset:
+		apiErr = ErrSlowDown
 	}
 
 	// Compression errors

cmd/api-router.go

@@ -46,6 +46,7 @@ func registerAPIRouter(router *mux.Router, encryptionEnabled bool) {
 	var routers []*mux.Router
 	for _, domainName := range globalDomainNames {
 		routers = append(routers, apiRouter.Host("{bucket:.+}."+domainName).Subrouter())
+		routers = append(routers, apiRouter.Host("{bucket:.+}."+domainName+":{port:.*}").Subrouter())
 	}
 	routers = append(routers, apiRouter.PathPrefix("/{bucket}").Subrouter())
 

cmd/auth-handler.go

@@ -362,7 +362,7 @@ func isReqAuthenticated(ctx context.Context, r *http.Request, region string, sty
 
 	// Verify 'Content-Md5' and/or 'X-Amz-Content-Sha256' if present.
 	// The verification happens implicit during reading.
-	reader, err := hash.NewReader(r.Body, -1, hex.EncodeToString(contentMD5), hex.EncodeToString(contentSHA256), -1)
+	reader, err := hash.NewReader(r.Body, -1, hex.EncodeToString(contentMD5), hex.EncodeToString(contentSHA256), -1, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return toAPIErrorCode(ctx, err)
 	}

cmd/bitrot-streaming.go

@@ -32,33 +32,20 @@ type streamingBitrotWriter struct {
 	h         hash.Hash
 	shardSize int64
 	canClose  chan struct{} // Needed to avoid race explained in Close() call.
-
-	// Following two fields are used only to make sure that Write(p) is called such that
-	// len(p) is always the block size except the last block, i.e prevent programmer errors.
-	currentBlockIdx int
-	verifyTillIdx   int
 }
 
 func (b *streamingBitrotWriter) Write(p []byte) (int, error) {
-	if b.currentBlockIdx < b.verifyTillIdx && int64(len(p)) != b.shardSize {
-		// All blocks except last should be of the length b.shardSize
-		logger.LogIf(context.Background(), errUnexpected)
-		return 0, errUnexpected
-	}
 	if len(p) == 0 {
 		return 0, nil
 	}
 	b.h.Reset()
 	b.h.Write(p)
 	hashBytes := b.h.Sum(nil)
-	n, err := b.iow.Write(hashBytes)
-	if n != len(hashBytes) {
-		logger.LogIf(context.Background(), err)
+	_, err := b.iow.Write(hashBytes)
+	if err != nil {
 		return 0, err
 	}
-	n, err = b.iow.Write(p)
-	b.currentBlockIdx++
-	return n, err
+	return b.iow.Write(p)
 }
 
 func (b *streamingBitrotWriter) Close() error {
@@ -78,16 +65,14 @@ func (b *streamingBitrotWriter) Close() error {
 func newStreamingBitrotWriter(disk StorageAPI, volume, filePath string, length int64, algo BitrotAlgorithm, shardSize int64) io.WriteCloser {
 	r, w := io.Pipe()
 	h := algo.New()
-	bw := &streamingBitrotWriter{w, h, shardSize, make(chan struct{}), 0, int(length / shardSize)}
+	bw := &streamingBitrotWriter{w, h, shardSize, make(chan struct{})}
 	go func() {
-		bitrotSumsTotalSize := ceilFrac(length, shardSize) * int64(h.Size()) // Size used for storing bitrot checksums.
-		totalFileSize := bitrotSumsTotalSize + length
-		err := disk.CreateFile(volume, filePath, totalFileSize, r)
-		if err != nil {
-			reqInfo := (&logger.ReqInfo{}).AppendTags("storageDisk", disk.String())
-			ctx := logger.SetReqInfo(context.Background(), reqInfo)
-			logger.LogIf(ctx, err)
+		totalFileSize := int64(-1) // For compressed objects length will be unknown (represented by length=-1)
+		if length != -1 {
+			bitrotSumsTotalSize := ceilFrac(length, shardSize) * int64(h.Size()) // Size used for storing bitrot checksums.
+			totalFileSize = bitrotSumsTotalSize + length
 		}
+		err := disk.CreateFile(volume, filePath, totalFileSize, r)
 		r.CloseWithError(err)
 		close(bw.canClose)
 	}()
@@ -118,7 +103,7 @@ func (b *streamingBitrotReader) ReadAt(buf []byte, offset int64) (int, error) {
 	var err error
 	if offset%b.shardSize != 0 {
 		// Offset should always be aligned to b.shardSize
-		logger.LogIf(context.Background(), errUnexpected)
+		// Can never happen unless there are programmer bugs
 		return 0, errUnexpected
 	}
 	if b.rc == nil {
@@ -127,25 +112,20 @@ func (b *streamingBitrotReader) ReadAt(buf []byte, offset int64) (int, error) {
 		streamOffset := (offset/b.shardSize)*int64(b.h.Size()) + offset
 		b.rc, err = b.disk.ReadFileStream(b.volume, b.filePath, streamOffset, b.tillOffset-streamOffset)
 		if err != nil {
-			reqInfo := (&logger.ReqInfo{}).AppendTags("storageDisk", b.disk.String())
-			ctx := logger.SetReqInfo(context.Background(), reqInfo)
-			logger.LogIf(ctx, err)
 			return 0, err
 		}
 	}
 	if offset != b.currOffset {
-		logger.LogIf(context.Background(), errUnexpected)
+		// Can never happen unless there are programmer bugs
 		return 0, errUnexpected
 	}
 	b.h.Reset()
 	_, err = io.ReadFull(b.rc, b.hashBytes)
 	if err != nil {
-		logger.LogIf(context.Background(), err)
 		return 0, err
 	}
 	_, err = io.ReadFull(b.rc, buf)
 	if err != nil {
-		logger.LogIf(context.Background(), err)
 		return 0, err
 	}
 	b.h.Write(buf)

cmd/bitrot-whole.go

@@ -31,19 +31,9 @@ type wholeBitrotWriter struct {
 	filePath  string
 	shardSize int64 // This is the shard size of the erasure logic
 	hash.Hash       // For bitrot hash
-
-	// Following two fields are used only to make sure that Write(p) is called such that
-	// len(p) is always the block size except the last block and prevent programmer errors.
-	currentBlockIdx int
-	lastBlockIdx    int
 }
 
 func (b *wholeBitrotWriter) Write(p []byte) (int, error) {
-	if b.currentBlockIdx < b.lastBlockIdx && int64(len(p)) != b.shardSize {
-		// All blocks except last should be of the length b.shardSize
-		logger.LogIf(context.Background(), errUnexpected)
-		return 0, errUnexpected
-	}
 	err := b.disk.AppendFile(b.volume, b.filePath, p)
 	if err != nil {
 		logger.LogIf(context.Background(), err)
@@ -54,7 +44,6 @@ func (b *wholeBitrotWriter) Write(p []byte) (int, error) {
 		logger.LogIf(context.Background(), err)
 		return 0, err
 	}
-	b.currentBlockIdx++
 	return len(p), nil
 }
 
@@ -63,8 +52,8 @@ func (b *wholeBitrotWriter) Close() error {
 }
 
 // Returns whole-file bitrot writer.
-func newWholeBitrotWriter(disk StorageAPI, volume, filePath string, length int64, algo BitrotAlgorithm, shardSize int64) io.WriteCloser {
-	return &wholeBitrotWriter{disk, volume, filePath, shardSize, algo.New(), 0, int(length / shardSize)}
+func newWholeBitrotWriter(disk StorageAPI, volume, filePath string, algo BitrotAlgorithm, shardSize int64) io.WriteCloser {
+	return &wholeBitrotWriter{disk, volume, filePath, shardSize, algo.New()}
 }
 
 // Implementation to verify bitrot for the whole file.

cmd/bitrot.go

@@ -120,7 +120,7 @@ func newBitrotWriter(disk StorageAPI, volume, filePath string, length int64, alg
 	if algo == HighwayHash256S {
 		return newStreamingBitrotWriter(disk, volume, filePath, length, algo, shardSize)
 	}
-	return newWholeBitrotWriter(disk, volume, filePath, length, algo, shardSize)
+	return newWholeBitrotWriter(disk, volume, filePath, algo, shardSize)
 }
 
 func newBitrotReader(disk StorageAPI, bucket string, filePath string, tillOffset int64, algo BitrotAlgorithm, sum []byte, shardSize int64) io.ReaderAt {

cmd/bucket-handlers.go

@@ -263,16 +263,6 @@ func (api objectAPIHandlers) DeleteMultipleObjectsHandler(w http.ResponseWriter,
 		return
 	}
 
-	var s3Error APIErrorCode
-	if s3Error = checkRequestAuthType(ctx, r, policy.DeleteObjectAction, bucket, ""); s3Error != ErrNone {
-		// In the event access is denied, a 200 response should still be returned
-		// http://docs.aws.amazon.com/AmazonS3/latest/API/multiobjectdeleteapi.html
-		if s3Error != ErrAccessDenied {
-			writeErrorResponse(ctx, w, errorCodes.ToAPIErr(s3Error), r.URL, guessIsBrowserReq(r))
-			return
-		}
-	}
-
 	// Content-Length is required and should be non-zero
 	// http://docs.aws.amazon.com/AmazonS3/latest/API/multiobjectdeleteapi.html
 	if r.ContentLength <= 0 {
@@ -319,42 +309,60 @@ func (api objectAPIHandlers) DeleteMultipleObjectsHandler(w http.ResponseWriter,
 		return
 	}
 
-	deleteObject := objectAPI.DeleteObject
+	deleteObjectsFn := objectAPI.DeleteObjects
 	if api.CacheAPI() != nil {
-		deleteObject = api.CacheAPI().DeleteObject
+		deleteObjectsFn = api.CacheAPI().DeleteObjects
 	}
 
-	var dErrs = make([]error, len(deleteObjects.Objects))
+	type delObj struct {
+		origIndex int
+		name      string
+	}
+
+	var objectsToDelete []delObj
+	var dErrs = make([]APIErrorCode, len(deleteObjects.Objects))
+
 	for index, object := range deleteObjects.Objects {
-		// If the request is denied access, each item
-		// should be marked as 'AccessDenied'
-		if s3Error == ErrAccessDenied {
-			dErrs[index] = PrefixAccessDenied{
-				Bucket: bucket,
-				Object: object.ObjectName,
+		if dErrs[index] = checkRequestAuthType(ctx, r, policy.DeleteObjectAction, bucket, object.ObjectName); dErrs[index] != ErrNone {
+			if dErrs[index] == ErrSignatureDoesNotMatch || dErrs[index] == ErrInvalidAccessKeyID {
+				writeErrorResponse(ctx, w, errorCodes.ToAPIErr(dErrs[index]), r.URL, guessIsBrowserReq(r))
+				return
 			}
 			continue
 		}
-		dErrs[index] = deleteObject(ctx, bucket, object.ObjectName)
+
+		objectsToDelete = append(objectsToDelete, delObj{index, object.ObjectName})
+	}
+
+	toNames := func(input []delObj) (output []string) {
+		output = make([]string, len(input))
+		for i := range input {
+			output[i] = input[i].name
+		}
+		return
+	}
+
+	errs, err := deleteObjectsFn(ctx, bucket, toNames(objectsToDelete))
+	if err != nil {
+		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
+		return
+	}
+
+	for i, obj := range objectsToDelete {
+		dErrs[obj.origIndex] = toAPIErrorCode(ctx, errs[i])
 	}
 
 	// Collect deleted objects and errors if any.
 	var deletedObjects []ObjectIdentifier
 	var deleteErrors []DeleteError
-	for index, err := range dErrs {
+	for index, errCode := range dErrs {
 		object := deleteObjects.Objects[index]
 		// Success deleted objects are collected separately.
-		if err == nil {
+		if errCode == ErrNone || errCode == ErrNoSuchKey {
 			deletedObjects = append(deletedObjects, object)
 			continue
 		}
-		if _, ok := err.(ObjectNotFound); ok {
-			// If the object is not found it should be
-			// accounted as deleted as per S3 spec.
-			deletedObjects = append(deletedObjects, object)
-			continue
-		}
-		apiErr := toAPIError(ctx, err)
+		apiErr := getAPIError(errCode)
 		// Error during delete should be collected separately.
 		deleteErrors = append(deleteErrors, DeleteError{
 			Code:    apiErr.Code,
@@ -615,7 +623,7 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 		return
 	}
 
-	hashReader, err := hash.NewReader(fileBody, fileSize, "", "", fileSize)
+	hashReader, err := hash.NewReader(fileBody, fileSize, "", "", fileSize, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		logger.LogIf(ctx, err)
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
@@ -653,7 +661,8 @@ func (api objectAPIHandlers) PostPolicyBucketHandler(w http.ResponseWriter, r *h
 				return
 			}
 			info := ObjectInfo{Size: fileSize}
-			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", fileSize) // do not try to verify encrypted content
+			// do not try to verify encrypted content
+			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", fileSize, globalCLIContext.StrictS3Compat)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 				return

cmd/bucket-handlers_test.go

@@ -762,7 +762,7 @@ func testAPIDeleteMultipleObjectsHandler(obj ObjectLayer, instanceType, bucketNa
 		apiRouter.ServeHTTP(rec, req)
 		// Assert the response code with the expected status.
 		if rec.Code != testCase.expectedRespStatus {
-			t.Errorf("Case %d: MinIO %s: Expected the response status to be `%d`, but instead found `%d`", i+1, instanceType, testCase.expectedRespStatus, rec.Code)
+			t.Errorf("Test %d: MinIO %s: Expected the response status to be `%d`, but instead found `%d`", i+1, instanceType, testCase.expectedRespStatus, rec.Code)
 		}
 
 		// read the response body.

cmd/bucket-policy-handlers.go

@@ -71,7 +71,7 @@ func (api objectAPIHandlers) PutBucketPolicyHandler(w http.ResponseWriter, r *ht
 
 	// Error out if Content-Length is beyond allowed size.
 	if r.ContentLength > maxBucketPolicySize {
-		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrEntityTooLarge), r.URL, guessIsBrowserReq(r))
+		writeErrorResponse(ctx, w, errorCodes.ToAPIErr(ErrPolicyTooLarge), r.URL, guessIsBrowserReq(r))
 		return
 	}
 

cmd/common-main.go

@@ -161,6 +161,9 @@ func handleCommonCmdArgs(ctx *cli.Context) {
 	globalCertsCADir = &ConfigDir{path: filepath.Join(globalCertsDir.Get(), certsCADir)}
 
 	logger.FatalIf(mkdirAllIgnorePerm(globalCertsCADir.Get()), "Unable to create certs CA directory at %s", globalCertsCADir.Get())
+
+	// Check "compat" flag from command line argument.
+	globalCLIContext.StrictS3Compat = ctx.IsSet("compat") || ctx.GlobalIsSet("compat")
 }
 
 // Parses the given compression exclude list `extensions` or `content-types`.

cmd/config-common.go

@@ -21,6 +21,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"time"
 
 	etcd "github.com/coreos/etcd/clientv3"
 	"github.com/minio/minio/cmd/logger"
@@ -73,7 +74,7 @@ func saveConfigEtcd(ctx context.Context, client *etcd.Client, configFile string,
 }
 
 func saveConfig(ctx context.Context, objAPI ObjectLayer, configFile string, data []byte) error {
-	hashReader, err := hash.NewReader(bytes.NewReader(data), int64(len(data)), "", getSHA256Hash(data), int64(len(data)))
+	hashReader, err := hash.NewReader(bytes.NewReader(data), int64(len(data)), "", getSHA256Hash(data), int64(len(data)), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return err
 	}
@@ -105,12 +106,26 @@ func readConfigEtcd(ctx context.Context, client *etcd.Client, configFile string)
 
 // watchConfigEtcd - watches for changes on `configFile` on etcd and loads them.
 func watchConfigEtcd(objAPI ObjectLayer, configFile string, loadCfgFn func(ObjectLayer) error) {
-	ctx, cancel := context.WithTimeout(context.Background(), defaultContextTimeout)
-	defer cancel()
-	for watchResp := range globalEtcdClient.Watch(ctx, configFile) {
-		for _, event := range watchResp.Events {
-			if event.IsModify() || event.IsCreate() {
-				loadCfgFn(objAPI)
+	for {
+		watchCh := globalEtcdClient.Watch(context.Background(), iamConfigPrefix)
+		select {
+		case <-GlobalServiceDoneCh:
+			return
+		case watchResp, ok := <-watchCh:
+			if !ok {
+				time.Sleep(1 * time.Second)
+				continue
+			}
+			if err := watchResp.Err(); err != nil {
+				logger.LogIf(context.Background(), err)
+				// log and retry.
+				time.Sleep(1 * time.Second)
+				continue
+			}
+			for _, event := range watchResp.Events {
+				if event.IsModify() || event.IsCreate() {
+					loadCfgFn(objAPI)
+				}
 			}
 		}
 	}

cmd/disk-cache-fs.go

@@ -18,7 +18,6 @@ package cmd
 
 import (
 	"context"
-	"encoding/hex"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -399,7 +398,7 @@ func (cfs *cacheFSObjects) PutObject(ctx context.Context, bucket string, object
 		return ObjectInfo{}, toObjectErr(err, bucket, object)
 	}
 	if fsMeta.Meta["etag"] == "" {
-		fsMeta.Meta["etag"] = hex.EncodeToString(data.MD5Current())
+		fsMeta.Meta["etag"] = r.MD5CurrentHexString()
 	}
 	// Should return IncompleteBody{} error when reader has fewer
 	// bytes than specified in request header.

cmd/disk-cache.go

@@ -62,6 +62,7 @@ type cacheObjects struct {
 	GetObjectInfoFn           func(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
 	PutObjectFn               func(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error)
 	DeleteObjectFn            func(ctx context.Context, bucket, object string) error
+	DeleteObjectsFn           func(ctx context.Context, bucket string, objects []string) ([]error, error)
 	ListObjectsFn             func(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (result ListObjectsInfo, err error)
 	ListObjectsV2Fn           func(ctx context.Context, bucket, prefix, continuationToken, delimiter string, maxKeys int, fetchOwner bool, startAfter string) (result ListObjectsV2Info, err error)
 	ListBucketsFn             func(ctx context.Context) (buckets []BucketInfo, err error)
@@ -94,6 +95,7 @@ type CacheObjectLayer interface {
 	GetObjectInfo(ctx context.Context, bucket, object string, opts ObjectOptions) (objInfo ObjectInfo, err error)
 	PutObject(ctx context.Context, bucket, object string, data *PutObjReader, opts ObjectOptions) (objInfo ObjectInfo, err error)
 	DeleteObject(ctx context.Context, bucket, object string) error
+	DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error)
 
 	// Multipart operations.
 	NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (uploadID string, err error)
@@ -243,7 +245,7 @@ func (c cacheObjects) GetObjectNInfo(ctx context.Context, bucket, object string,
 	// Initialize pipe.
 	pipeReader, pipeWriter := io.Pipe()
 	teeReader := io.TeeReader(bkReader, pipeWriter)
-	hashReader, herr := hash.NewReader(pipeReader, bkReader.ObjInfo.Size, "", "", bkReader.ObjInfo.Size)
+	hashReader, herr := hash.NewReader(pipeReader, bkReader.ObjInfo.Size, "", "", bkReader.ObjInfo.Size, globalCLIContext.StrictS3Compat)
 	if herr != nil {
 		bkReader.Close()
 		return nil, herr
@@ -314,7 +316,7 @@ func (c cacheObjects) GetObject(ctx context.Context, bucket, object string, star
 	}
 	// Initialize pipe.
 	pipeReader, pipeWriter := io.Pipe()
-	hashReader, err := hash.NewReader(pipeReader, objInfo.Size, "", "", objInfo.Size)
+	hashReader, err := hash.NewReader(pipeReader, objInfo.Size, "", "", objInfo.Size, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return err
 	}
@@ -375,7 +377,7 @@ func (c cacheObjects) GetObjectInfo(ctx context.Context, bucket, object string,
 // Returns function "listDir" of the type listDirFunc.
 // isLeaf - is used by listDir function to check if an entry is a leaf or non-leaf entry.
 // disks - list of fsObjects
-func listDirCacheFactory(isLeaf IsLeafFunc, disks []*cacheFSObjects) ListDirFunc {
+func listDirCacheFactory(isLeaf func(string, string) bool, disks []*cacheFSObjects) ListDirFunc {
 	listCacheDirs := func(bucket, prefixDir, prefixEntry string) (dirs []string) {
 		var entries []string
 		for _, disk := range disks {
@@ -391,6 +393,12 @@ func listDirCacheFactory(isLeaf IsLeafFunc, disks []*cacheFSObjects) ListDirFunc
 				continue
 			}
 
+			for i := range entries {
+				if isLeaf(bucket, entries[i]) {
+					entries[i] = strings.TrimSuffix(entries[i], slashSeparator)
+				}
+			}
+
 			// Filter entries that have the prefix prefixEntry.
 			entries = filterMatchingPrefix(entries, prefixEntry)
 			dirs = append(dirs, entries...)
@@ -399,7 +407,7 @@ func listDirCacheFactory(isLeaf IsLeafFunc, disks []*cacheFSObjects) ListDirFunc
 	}
 
 	// listDir - lists all the entries at a given prefix and given entry in the prefix.
-	listDir := func(bucket, prefixDir, prefixEntry string) (mergedEntries []string, delayIsLeaf bool) {
+	listDir := func(bucket, prefixDir, prefixEntry string) (mergedEntries []string) {
 		cacheEntries := listCacheDirs(bucket, prefixDir, prefixEntry)
 		for _, entry := range cacheEntries {
 			// Find elements in entries which are not in mergedEntries
@@ -411,7 +419,7 @@ func listDirCacheFactory(isLeaf IsLeafFunc, disks []*cacheFSObjects) ListDirFunc
 			mergedEntries = append(mergedEntries, entry)
 			sort.Strings(mergedEntries)
 		}
-		return mergedEntries, false
+		return mergedEntries
 	}
 	return listDir
 }
@@ -430,25 +438,16 @@ func (c cacheObjects) listCacheObjects(ctx context.Context, bucket, prefix, mark
 	walkResultCh, endWalkCh := c.listPool.Release(listParams{bucket, recursive, marker, prefix})
 	if walkResultCh == nil {
 		endWalkCh = make(chan struct{})
-		isLeaf := func(bucket, object string) bool {
+
+		listDir := listDirCacheFactory(func(bucket, object string) bool {
 			fs, err := c.cache.getCacheFS(ctx, bucket, object)
 			if err != nil {
 				return false
 			}
 			_, err = fs.getObjectInfo(ctx, bucket, object)
 			return err == nil
-		}
-
-		isLeafDir := func(bucket, object string) bool {
-			fs, err := c.cache.getCacheFS(ctx, bucket, object)
-			if err != nil {
-				return false
-			}
-			return fs.isObjectDir(bucket, object)
-		}
-
-		listDir := listDirCacheFactory(isLeaf, c.cache.cfs)
-		walkResultCh = startTreeWalk(ctx, bucket, prefix, marker, recursive, listDir, isLeaf, isLeafDir, endWalkCh)
+		}, c.cache.cfs)
+		walkResultCh = startTreeWalk(ctx, bucket, prefix, marker, recursive, listDir, endWalkCh)
 	}
 
 	for i := 0; i < maxKeys; {
@@ -458,10 +457,6 @@ func (c cacheObjects) listCacheObjects(ctx context.Context, bucket, prefix, mark
 			eof = true
 			break
 		}
-		// For any walk error return right away.
-		if walkResult.err != nil {
-			return result, toObjectErr(walkResult.err, bucket, prefix)
-		}
 
 		entry := walkResult.entry
 		var objInfo ObjectInfo
@@ -636,6 +631,14 @@ func (c cacheObjects) DeleteObject(ctx context.Context, bucket, object string) (
 	return
 }
 
+func (c cacheObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = c.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // Returns true if object should be excluded from cache
 func (c cacheObjects) isCacheExclude(bucket, object string) bool {
 	for _, pattern := range c.exclude {
@@ -671,13 +674,13 @@ func (c cacheObjects) PutObject(ctx context.Context, bucket, object string, r *P
 	objInfo = ObjectInfo{}
 	// Initialize pipe to stream data to backend
 	pipeReader, pipeWriter := io.Pipe()
-	hashReader, err := hash.NewReader(pipeReader, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize())
+	hashReader, err := hash.NewReader(pipeReader, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize(), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return ObjectInfo{}, err
 	}
 	// Initialize pipe to stream data to cache
 	rPipe, wPipe := io.Pipe()
-	cHashReader, err := hash.NewReader(rPipe, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize())
+	cHashReader, err := hash.NewReader(rPipe, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize(), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return ObjectInfo{}, err
 	}
@@ -765,13 +768,13 @@ func (c cacheObjects) PutObjectPart(ctx context.Context, bucket, object, uploadI
 	info = PartInfo{}
 	// Initialize pipe to stream data to backend
 	pipeReader, pipeWriter := io.Pipe()
-	hashReader, err := hash.NewReader(pipeReader, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize())
+	hashReader, err := hash.NewReader(pipeReader, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize(), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return
 	}
 	// Initialize pipe to stream data to cache
 	rPipe, wPipe := io.Pipe()
-	cHashReader, err := hash.NewReader(rPipe, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize())
+	cHashReader, err := hash.NewReader(rPipe, size, data.MD5HexString(), data.SHA256HexString(), data.ActualSize(), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		return
 	}
@@ -981,6 +984,14 @@ func newServerCacheObjects(config CacheConfig) (CacheObjectLayer, error) {
 		DeleteObjectFn: func(ctx context.Context, bucket, object string) error {
 			return newObjectLayerFn().DeleteObject(ctx, bucket, object)
 		},
+		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
+		},
+
 		ListObjectsFn: func(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (result ListObjectsInfo, err error) {
 			return newObjectLayerFn().ListObjects(ctx, bucket, prefix, marker, delimiter, maxKeys)
 		},

cmd/disk-cache_test.go

@@ -194,7 +194,7 @@ func TestDiskCache(t *testing.T) {
 	objInfo.UserDefined = httpMeta
 	var opts ObjectOptions
 	byteReader := bytes.NewReader([]byte(content))
-	hashReader, err := hash.NewReader(byteReader, int64(size), "", "", int64(size))
+	hashReader, err := hash.NewReader(byteReader, int64(size), "", "", int64(size), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -269,7 +269,7 @@ func TestDiskCacheMaxUse(t *testing.T) {
 	opts := ObjectOptions{}
 
 	byteReader := bytes.NewReader([]byte(content))
-	hashReader, err := hash.NewReader(byteReader, int64(size), "", "", int64(size))
+	hashReader, err := hash.NewReader(byteReader, int64(size), "", "", int64(size), globalCLIContext.StrictS3Compat)
 	if err != nil {
 		t.Fatal(err)
 	}

cmd/endpoint-ellipses_test.go

@@ -44,7 +44,6 @@ func TestCreateServerEndpoints(t *testing.T) {
 		{":9000", []string{"/export1{1...32}", "/export1{1...32}"}, false},
 		// Same host cannot export same disk on two ports - special case localhost.
 		{":9001", []string{"http://localhost:900{1...2}/export{1...64}"}, false},
-
 		// Valid inputs.
 		{":9000", []string{"/export1"}, true},
 		{":9000", []string{"/export1", "/export2", "/export3", "/export4"}, true},

cmd/endpoint.go

@@ -17,6 +17,7 @@
 package cmd
 
 import (
+	"context"
 	"fmt"
 	"net"
 	"net/url"
@@ -26,7 +27,10 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
+	"time"
 
+	humanize "github.com/dustin/go-humanize"
+	"github.com/minio/cli"
 	"github.com/minio/minio-go/pkg/set"
 	"github.com/minio/minio/cmd/logger"
 	"github.com/minio/minio/pkg/cpu"
@@ -44,6 +48,8 @@ const (
 
 	// URLEndpointType - URL style endpoint type enum.
 	URLEndpointType
+
+	retryInterval = 5 // In Seconds.
 )
 
 // Endpoint - any type of endpoint.
@@ -51,6 +57,7 @@ type Endpoint struct {
 	*url.URL
 	IsLocal  bool
 	SetIndex int
+	HostName string
 }
 
 func (endpoint Endpoint) String() string {
@@ -75,6 +82,19 @@ func (endpoint Endpoint) IsHTTPS() bool {
 	return endpoint.Scheme == "https"
 }
 
+// UpdateIsLocal - resolves the host and updates if it is local or not.
+func (endpoint *Endpoint) UpdateIsLocal() error {
+	if !endpoint.IsLocal {
+		isLocal, err := isLocalHost(endpoint.HostName)
+		if err != nil {
+			return err
+		}
+		endpoint.IsLocal = isLocal
+
+	}
+	return nil
+}
+
 // NewEndpoint - returns new endpoint based on given arguments.
 func NewEndpoint(arg string) (ep Endpoint, e error) {
 	// isEmptyPath - check whether given path is not empty.
@@ -87,6 +107,7 @@ func NewEndpoint(arg string) (ep Endpoint, e error) {
 	}
 
 	var isLocal bool
+	var host string
 	u, err := url.Parse(arg)
 	if err == nil && u.Host != "" {
 		// URL style of endpoint.
@@ -98,7 +119,7 @@ func NewEndpoint(arg string) (ep Endpoint, e error) {
 			return ep, fmt.Errorf("invalid URL endpoint format")
 		}
 
-		var host, port string
+		var port string
 		host, port, err = net.SplitHostPort(u.Host)
 		if err != nil {
 			if !strings.Contains(err.Error(), "missing port in address") {
@@ -150,10 +171,6 @@ func NewEndpoint(arg string) (ep Endpoint, e error) {
 			}
 		}
 
-		isLocal, err = isLocalHost(host)
-		if err != nil {
-			return ep, err
-		}
 	} else {
 		// Only check if the arg is an ip address and ask for scheme since its absent.
 		// localhost, example.com, any FQDN cannot be disambiguated from a regular file path such as
@@ -166,8 +183,9 @@ func NewEndpoint(arg string) (ep Endpoint, e error) {
 	}
 
 	return Endpoint{
-		URL:     u,
-		IsLocal: isLocal,
+		URL:      u,
+		IsLocal:  isLocal,
+		HostName: host,
 	}, nil
 }
 
@@ -200,6 +218,74 @@ func (endpoints EndpointList) GetString(i int) string {
 	return endpoints[i].String()
 }
 
+// UpdateIsLocal - resolves the host and discovers the local host.
+func (endpoints EndpointList) UpdateIsLocal() error {
+	var epsResolved int
+	var foundLocal bool
+	resolvedList := make([]bool, len(endpoints))
+	// Mark the starting time
+	startTime := time.Now()
+	keepAliveTicker := time.NewTicker(retryInterval * time.Second)
+	defer keepAliveTicker.Stop()
+	for {
+		// Break if the local endpoint is found already. Or all the endpoints are resolved.
+		if foundLocal || (epsResolved == len(endpoints)) {
+			break
+		}
+		// Retry infinitely on Kubernetes and Docker swarm.
+		// This is needed as the remote hosts are sometime
+		// not available immediately.
+		select {
+		case <-globalOSSignalCh:
+			return fmt.Errorf("The endpoint resolution got interrupted")
+		default:
+			for i, resolved := range resolvedList {
+				if resolved {
+					continue
+				}
+
+				// return err if not Docker or Kubernetes
+				// We use IsDocker() method to check for Docker Swarm environment
+				// as there is no reliable way to clearly identify Swarm from
+				// Docker environment.
+				isLocal, err := isLocalHost(endpoints[i].HostName)
+				if err != nil {
+					if !IsDocker() && !IsKubernetes() {
+						return err
+					}
+					// time elapsed
+					timeElapsed := time.Since(startTime)
+					// log error only if more than 1s elapsed
+					if timeElapsed > time.Second {
+						// log the message to console about the host not being
+						// resolveable.
+						reqInfo := (&logger.ReqInfo{}).AppendTags("host", endpoints[i].HostName)
+						reqInfo.AppendTags("elapsedTime", humanize.RelTime(startTime, startTime.Add(timeElapsed), "elapsed", ""))
+						ctx := logger.SetReqInfo(context.Background(), reqInfo)
+						logger.LogIf(ctx, err)
+					}
+				} else {
+					resolvedList[i] = true
+					endpoints[i].IsLocal = isLocal
+					epsResolved++
+					if !foundLocal {
+						foundLocal = isLocal
+					}
+				}
+			}
+
+			// Wait for the tick, if the there exist a local endpoint in discovery.
+			// Non docker/kubernetes environment does not need to wait.
+			if !foundLocal && (IsDocker() && IsKubernetes()) {
+				<-keepAliveTicker.C
+			}
+		}
+	}
+
+	return nil
+
+}
+
 // localEndpointsMemUsage - returns ServerMemUsageInfo for only the
 // local endpoints from given list of endpoints
 func localEndpointsMemUsage(endpoints EndpointList) ServerMemUsageInfo {
@@ -302,9 +388,30 @@ func NewEndpointList(args ...string) (endpoints EndpointList, err error) {
 		uniqueArgs.Add(arg)
 		endpoints = append(endpoints, endpoint)
 	}
+
 	return endpoints, nil
 }
 
+func checkEndpointsSubOptimal(ctx *cli.Context, setupType SetupType, endpoints EndpointList) (err error) {
+	// Validate sub optimal ordering only for distributed setup.
+	if setupType != DistXLSetupType {
+		return nil
+	}
+	var endpointOrder int
+	err = fmt.Errorf("Too many disk args are local, input is in sub-optimal order. Please review input args: %s", ctx.Args())
+	for _, endpoint := range endpoints {
+		if endpoint.IsLocal {
+			endpointOrder++
+		} else {
+			endpointOrder--
+		}
+		if endpointOrder >= 2 {
+			return err
+		}
+	}
+	return nil
+}
+
 // Checks if there are any cross device mounts.
 func checkCrossDeviceMounts(endpoints EndpointList) (err error) {
 	var absPaths []string
@@ -341,6 +448,9 @@ func CreateEndpoints(serverAddr string, args ...[]string) (string, EndpointList,
 		if err != nil {
 			return serverAddr, endpoints, setupType, err
 		}
+		if err := endpoint.UpdateIsLocal(); err != nil {
+			return serverAddr, endpoints, setupType, err
+		}
 		if endpoint.Type() != PathEndpointType {
 			return serverAddr, endpoints, setupType, uiErrInvalidFSEndpoint(nil).Msg("use path style endpoint for FS setup")
 		}
@@ -381,6 +491,10 @@ func CreateEndpoints(serverAddr string, args ...[]string) (string, EndpointList,
 		return serverAddr, endpoints, setupType, nil
 	}
 
+	if err := endpoints.UpdateIsLocal(); err != nil {
+		return serverAddr, endpoints, setupType, uiErrInvalidErasureEndpoints(nil).Msg(err.Error())
+	}
+
 	// Here all endpoints are URL style.
 	endpointPathSet := set.NewStringSet()
 	localEndpointCount := 0

cmd/endpoint_test.go

@@ -17,13 +17,53 @@
 package cmd
 
 import (
+	"flag"
 	"fmt"
 	"net/url"
 	"reflect"
 	"strings"
 	"testing"
+
+	"github.com/minio/cli"
 )
 
+func TestSubOptimalEndpointInput(t *testing.T) {
+	args1 := []string{"http://localhost/d1", "http://localhost/d2", "http://localhost/d3", "http://localhost/d4"}
+	args2 := []string{"http://example.org/d1", "http://example.com/d1", "http://example.net/d1", "http://example.edu/d1"}
+
+	tests := []struct {
+		setupType SetupType
+		ctx       *cli.Context
+		endpoints EndpointList
+		isErr     bool
+	}{
+		{
+			setupType: DistXLSetupType,
+			ctx:       cli.NewContext(cli.NewApp(), flag.NewFlagSet("", flag.ContinueOnError), nil),
+			endpoints: mustGetNewEndpointList(args1...),
+			isErr:     false,
+		},
+		{
+			setupType: DistXLSetupType,
+			ctx:       cli.NewContext(cli.NewApp(), flag.NewFlagSet("", flag.ContinueOnError), nil),
+			endpoints: mustGetNewEndpointList(args2...),
+			isErr:     false,
+		},
+	}
+	for i, test := range tests {
+		test := test
+		t.Run(fmt.Sprintf("Test%d", i+1), func(t *testing.T) {
+			err := checkEndpointsSubOptimal(test.ctx, test.setupType, test.endpoints)
+			if test.isErr && err == nil {
+				t.Error("expected err but found nil")
+			}
+			if !test.isErr && err != nil {
+				t.Errorf("expected err nil but found an err %s", err)
+			}
+		})
+	}
+}
+
 func TestNewEndpoint(t *testing.T) {
 	u1, _ := url.Parse("http://localhost/path")
 	u2, _ := url.Parse("https://example.org/path")
@@ -49,11 +89,11 @@ func TestNewEndpoint(t *testing.T) {
 		{"http:path", Endpoint{URL: &url.URL{Path: "http:path"}, IsLocal: true}, PathEndpointType, nil},
 		{"http:/path", Endpoint{URL: &url.URL{Path: "http:/path"}, IsLocal: true}, PathEndpointType, nil},
 		{"http:///path", Endpoint{URL: &url.URL{Path: "http:/path"}, IsLocal: true}, PathEndpointType, nil},
-		{"http://localhost/path", Endpoint{URL: u1, IsLocal: true}, URLEndpointType, nil},
-		{"http://localhost/path//", Endpoint{URL: u1, IsLocal: true}, URLEndpointType, nil},
-		{"https://example.org/path", Endpoint{URL: u2}, URLEndpointType, nil},
-		{"http://127.0.0.1:8080/path", Endpoint{URL: u3, IsLocal: true}, URLEndpointType, nil},
-		{"http://192.168.253.200/path", Endpoint{URL: u4}, URLEndpointType, nil},
+		{"http://localhost/path", Endpoint{URL: u1, IsLocal: true, HostName: "localhost"}, URLEndpointType, nil},
+		{"http://localhost/path//", Endpoint{URL: u1, IsLocal: true, HostName: "localhost"}, URLEndpointType, nil},
+		{"https://example.org/path", Endpoint{URL: u2, IsLocal: false, HostName: "example.org"}, URLEndpointType, nil},
+		{"http://127.0.0.1:8080/path", Endpoint{URL: u3, IsLocal: true, HostName: "127.0.0.1"}, URLEndpointType, nil},
+		{"http://192.168.253.200/path", Endpoint{URL: u4, IsLocal: false, HostName: "192.168.253.200"}, URLEndpointType, nil},
 		{"", Endpoint{}, -1, fmt.Errorf("empty or root endpoint is not supported")},
 		{"/", Endpoint{}, -1, fmt.Errorf("empty or root endpoint is not supported")},
 		{`\`, Endpoint{}, -1, fmt.Errorf("empty or root endpoint is not supported")},
@@ -71,6 +111,10 @@ func TestNewEndpoint(t *testing.T) {
 
 	for _, testCase := range testCases {
 		endpoint, err := NewEndpoint(testCase.arg)
+		if err == nil {
+			err = endpoint.UpdateIsLocal()
+		}
+
 		if testCase.expectedErr == nil {
 			if err != nil {
 				t.Fatalf("error: expected = <nil>, got = %v", err)
@@ -261,46 +305,46 @@ func TestCreateEndpoints(t *testing.T) {
 
 		// DistXL type
 		{"127.0.0.1:10000", [][]string{{case1Endpoint1, case1Endpoint2, "http://example.org/d3", "http://example.com/d4"}}, "127.0.0.1:10000", EndpointList{
-			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]},
+			Endpoint{URL: case1URLs[0], IsLocal: case1LocalFlags[0], HostName: nonLoopBackIP},
+			Endpoint{URL: case1URLs[1], IsLocal: case1LocalFlags[1], HostName: nonLoopBackIP},
+			Endpoint{URL: case1URLs[2], IsLocal: case1LocalFlags[2], HostName: "example.org"},
+			Endpoint{URL: case1URLs[3], IsLocal: case1LocalFlags[3], HostName: "example.com"},
 		}, DistXLSetupType, nil},
 
 		{"127.0.0.1:10000", [][]string{{case2Endpoint1, case2Endpoint2, "http://example.org/d3", "http://example.com/d4"}}, "127.0.0.1:10000", EndpointList{
-			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]},
+			Endpoint{URL: case2URLs[0], IsLocal: case2LocalFlags[0], HostName: nonLoopBackIP},
+			Endpoint{URL: case2URLs[1], IsLocal: case2LocalFlags[1], HostName: nonLoopBackIP},
+			Endpoint{URL: case2URLs[2], IsLocal: case2LocalFlags[2], HostName: "example.org"},
+			Endpoint{URL: case2URLs[3], IsLocal: case2LocalFlags[3], HostName: "example.com"},
 		}, DistXLSetupType, nil},
 
 		{":80", [][]string{{case3Endpoint1, "http://example.org:9000/d2", "http://example.com/d3", "http://example.net/d4"}}, ":80", EndpointList{
-			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]},
+			Endpoint{URL: case3URLs[0], IsLocal: case3LocalFlags[0], HostName: nonLoopBackIP},
+			Endpoint{URL: case3URLs[1], IsLocal: case3LocalFlags[1], HostName: "example.org"},
+			Endpoint{URL: case3URLs[2], IsLocal: case3LocalFlags[2], HostName: "example.com"},
+			Endpoint{URL: case3URLs[3], IsLocal: case3LocalFlags[3], HostName: "example.net"},
 		}, DistXLSetupType, nil},
 
 		{":9000", [][]string{{case4Endpoint1, "http://example.org/d2", "http://example.com/d3", "http://example.net/d4"}}, ":9000", EndpointList{
-			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]},
+			Endpoint{URL: case4URLs[0], IsLocal: case4LocalFlags[0], HostName: nonLoopBackIP},
+			Endpoint{URL: case4URLs[1], IsLocal: case4LocalFlags[1], HostName: "example.org"},
+			Endpoint{URL: case4URLs[2], IsLocal: case4LocalFlags[2], HostName: "example.com"},
+			Endpoint{URL: case4URLs[3], IsLocal: case4LocalFlags[3], HostName: "example.net"},
 		}, DistXLSetupType, nil},
 
 		{":9000", [][]string{{case5Endpoint1, case5Endpoint2, case5Endpoint3, case5Endpoint4}}, ":9000", EndpointList{
-			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]},
+			Endpoint{URL: case5URLs[0], IsLocal: case5LocalFlags[0], HostName: nonLoopBackIP},
+			Endpoint{URL: case5URLs[1], IsLocal: case5LocalFlags[1], HostName: nonLoopBackIP},
+			Endpoint{URL: case5URLs[2], IsLocal: case5LocalFlags[2], HostName: nonLoopBackIP},
+			Endpoint{URL: case5URLs[3], IsLocal: case5LocalFlags[3], HostName: nonLoopBackIP},
 		}, DistXLSetupType, nil},
 
 		// DistXL Setup using only local host.
 		{":9003", [][]string{{"http://localhost:9000/d1", "http://localhost:9001/d2", "http://127.0.0.1:9002/d3", case6Endpoint}}, ":9003", EndpointList{
-			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]},
+			Endpoint{URL: case6URLs[0], IsLocal: case6LocalFlags[0], HostName: "localhost"},
+			Endpoint{URL: case6URLs[1], IsLocal: case6LocalFlags[1], HostName: "localhost"},
+			Endpoint{URL: case6URLs[2], IsLocal: case6LocalFlags[2], HostName: "127.0.0.1"},
+			Endpoint{URL: case6URLs[3], IsLocal: case6LocalFlags[3], HostName: nonLoopBackIP},
 		}, DistXLSetupType, nil},
 	}
 
@@ -357,6 +401,11 @@ func TestGetLocalPeer(t *testing.T) {
 
 	for i, testCase := range testCases {
 		endpoints, _ := NewEndpointList(testCase.endpointArgs...)
+		if !endpoints[0].IsLocal {
+			if err := endpoints.UpdateIsLocal(); err != nil {
+				t.Fatalf("error: expected = <nil>, got = %v", err)
+			}
+		}
 		remotePeer := GetLocalPeer(endpoints)
 		if remotePeer != testCase.expectedResult {
 			t.Fatalf("Test %d: expected: %v, got: %v", i+1, testCase.expectedResult, remotePeer)
@@ -384,6 +433,11 @@ func TestGetRemotePeers(t *testing.T) {
 
 	for _, testCase := range testCases {
 		endpoints, _ := NewEndpointList(testCase.endpointArgs...)
+		if !endpoints[0].IsLocal {
+			if err := endpoints.UpdateIsLocal(); err != nil {
+				t.Fatalf("error: expected = <nil>, got = %v", err)
+			}
+		}
 		remotePeers := GetRemotePeers(endpoints)
 		if !reflect.DeepEqual(remotePeers, testCase.expectedResult) {
 			t.Fatalf("expected: %v, got: %v", testCase.expectedResult, remotePeers)

cmd/erasure.go

@@ -103,6 +103,9 @@ 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))

cmd/fs-v1-helpers.go

@@ -206,6 +206,9 @@ func osErrToFSFileErr(err error) error {
 	if isSysErrPathNotFound(err) {
 		return errFileNotFound
 	}
+	if isSysErrTooManyFiles(err) {
+		return errTooManyOpenFiles
+	}
 	return err
 }
 

cmd/fs-v1-multipart.go

@@ -504,10 +504,7 @@ func (fs *FSObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 	}
 
 	// Calculate s3 compatible md5sum for complete multipart.
-	s3MD5, err := getCompleteMultipartMD5(ctx, parts)
-	if err != nil {
-		return oi, err
-	}
+	s3MD5 := getCompleteMultipartMD5(parts)
 
 	partSize := int64(-1) // Used later to ensure that all parts sizes are same.
 

cmd/fs-v1.go

@@ -450,6 +450,11 @@ func (fs *FSObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBu
 		// Return the new object info.
 		return fsMeta.ToObjectInfo(srcBucket, srcObject, fi), nil
 	}
+
+	if err := checkPutObjectArgs(ctx, dstBucket, dstObject, fs, srcInfo.PutObjReader.Size()); err != nil {
+		return ObjectInfo{}, err
+	}
+
 	objInfo, err := fs.putObject(ctx, dstBucket, dstObject, srcInfo.PutObjReader, ObjectOptions{ServerSideEncryption: dstOpts.ServerSideEncryption, UserDefined: srcInfo.UserDefined})
 	if err != nil {
 		return oi, toObjectErr(err, dstBucket, dstObject)
@@ -813,6 +818,7 @@ func (fs *FSObjects) PutObject(ctx context.Context, bucket string, object string
 		return objInfo, err
 	}
 	defer objectLock.Unlock()
+
 	return fs.putObject(ctx, bucket, object, r, opts)
 }
 
@@ -854,10 +860,6 @@ func (fs *FSObjects) putObject(ctx context.Context, bucket string, object string
 		return fsMeta.ToObjectInfo(bucket, object, fi), nil
 	}
 
-	if err = checkPutObjectArgs(ctx, bucket, object, fs, data.Size()); err != nil {
-		return ObjectInfo{}, err
-	}
-
 	// Check if an object is present as one of the parent dir.
 	if fs.parentDirIsObject(ctx, bucket, path.Dir(object)) {
 		return ObjectInfo{}, toObjectErr(errFileParentIsFile, bucket, object)
@@ -951,6 +953,16 @@ func (fs *FSObjects) putObject(ctx context.Context, bucket string, object string
 	return fsMeta.ToObjectInfo(bucket, object, fi), nil
 }
 
+// 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) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = fs.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // 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 {
@@ -1001,17 +1013,17 @@ func (fs *FSObjects) DeleteObject(ctx context.Context, bucket, object string) er
 // Returns function "listDir" of the type listDirFunc.
 // isLeaf - is used by listDir function to check if an entry
 // is a leaf or non-leaf entry.
-func (fs *FSObjects) listDirFactory(isLeaf IsLeafFunc) ListDirFunc {
+func (fs *FSObjects) listDirFactory() ListDirFunc {
 	// listDir - lists all the entries at a given prefix and given entry in the prefix.
-	listDir := func(bucket, prefixDir, prefixEntry string) (entries []string, delayIsLeaf bool) {
+	listDir := func(bucket, prefixDir, prefixEntry string) (entries []string) {
 		var err error
 		entries, err = readDir(pathJoin(fs.fsPath, bucket, prefixDir))
 		if err != nil && err != errFileNotFound {
 			logger.LogIf(context.Background(), err)
 			return
 		}
-		entries, delayIsLeaf = filterListEntries(bucket, prefixDir, entries, prefixEntry, isLeaf)
-		return entries, delayIsLeaf
+		sort.Strings(entries)
+		return filterMatchingPrefix(entries, prefixEntry)
 	}
 
 	// Return list factory instance.
@@ -1097,22 +1109,8 @@ func (fs *FSObjects) getObjectETag(ctx context.Context, bucket, entry string, lo
 // 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) {
-	isLeaf := func(bucket, object string) bool {
-		// bucket argument is unused as we don't need to StatFile
-		// to figure if it's a file, just need to check that the
-		// object string does not end with "/".
-		return !hasSuffix(object, slashSeparator)
-	}
-	// Return true if the specified object is an empty directory
-	isLeafDir := func(bucket, object string) bool {
-		if !hasSuffix(object, slashSeparator) {
-			return false
-		}
-		return fs.isObjectDir(bucket, object)
-	}
-	listDir := fs.listDirFactory(isLeaf)
-
-	return listObjects(ctx, fs, bucket, prefix, marker, delimiter, maxKeys, fs.listPool, isLeaf, isLeafDir, listDir, fs.getObjectInfo, fs.getObjectInfo)
+	return listObjects(ctx, fs, bucket, prefix, marker, delimiter, maxKeys, fs.listPool,
+		fs.listDirFactory(), fs.getObjectInfo, fs.getObjectInfo)
 }
 
 // ReloadFormat - no-op for fs, Valid only for XL.

cmd/gateway-common.go

@@ -17,7 +17,6 @@
 package cmd
 
 import (
-	"context"
 	"net/http"
 	"os"
 	"strings"
@@ -44,14 +43,11 @@ var (
 	// ListObjects function alias.
 	ListObjects = listObjects
 
-	// FilterListEntries function alias.
-	FilterListEntries = filterListEntries
+	// FilterMatchingPrefix function alias.
+	FilterMatchingPrefix = filterMatchingPrefix
 
 	// IsStringEqual is string equal.
 	IsStringEqual = isStringEqual
-
-	// GetCompleteMultipartMD5 returns multipart MD5
-	GetCompleteMultipartMD5 = getCompleteMultipartMD5
 )
 
 // StatInfo -  alias for statInfo
@@ -351,8 +347,8 @@ func ErrorRespToObjectError(err error, params ...string) error {
 }
 
 // ComputeCompleteMultipartMD5 calculates MD5 ETag for complete multipart responses
-func ComputeCompleteMultipartMD5(parts []CompletePart) (string, error) {
-	return getCompleteMultipartMD5(context.Background(), parts)
+func ComputeCompleteMultipartMD5(parts []CompletePart) string {
+	return getCompleteMultipartMD5(parts)
 }
 
 // parse gateway sse env variable

cmd/gateway/azure/gateway-azure.go

@@ -835,7 +835,7 @@ func (a *azureObjects) PutObject(ctx context.Context, bucket, object string, r *
 	}
 
 	// Save md5sum for future processing on the object.
-	opts.UserDefined["x-amz-meta-md5sum"] = hex.EncodeToString(data.MD5Current())
+	opts.UserDefined["x-amz-meta-md5sum"] = r.MD5CurrentHexString()
 	objBlob.Metadata, objBlob.Properties, err = s3MetaToAzureProperties(ctx, opts.UserDefined)
 	if err != nil {
 		return objInfo, azureToObjectError(err, bucket, object)
@@ -896,6 +896,14 @@ func (a *azureObjects) DeleteObject(ctx context.Context, bucket, object string)
 	return nil
 }
 
+func (a *azureObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = a.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // ListMultipartUploads - It's decided not to support List Multipart Uploads, hence returning empty result.
 func (a *azureObjects) ListMultipartUploads(ctx context.Context, bucket, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result minio.ListMultipartsInfo, err error) {
 	// It's decided not to support List Multipart Uploads, hence returning empty result.
@@ -1207,10 +1215,7 @@ func (a *azureObjects) CompleteMultipartUpload(ctx context.Context, bucket, obje
 	if err != nil {
 		return objInfo, azureToObjectError(err, bucket, object)
 	}
-	objBlob.Metadata["md5sum"], err = cmd.ComputeCompleteMultipartMD5(uploadedParts)
-	if err != nil {
-		return objInfo, err
-	}
+	objBlob.Metadata["md5sum"] = cmd.ComputeCompleteMultipartMD5(uploadedParts)
 	err = objBlob.SetProperties(nil)
 	if err != nil {
 		return objInfo, azureToObjectError(err, bucket, object)

cmd/gateway/b2/gateway-b2.go

@@ -595,6 +595,14 @@ func (l *b2Objects) DeleteObject(ctx context.Context, bucket string, object stri
 	return b2ToObjectError(err, bucket, object)
 }
 
+func (l *b2Objects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = l.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // ListMultipartUploads lists all multipart uploads.
 func (l *b2Objects) ListMultipartUploads(ctx context.Context, bucket string, prefix string, keyMarker string, uploadIDMarker string,
 	delimiter string, maxUploads int) (lmi minio.ListMultipartsInfo, err error) {

cmd/gateway/gcs/gateway-gcs.go

@@ -915,15 +915,12 @@ func (l *gcsGateway) PutObject(ctx context.Context, bucket string, key string, r
 	}
 
 	// Close the object writer upon success.
-	w.Close()
-
-	attrs, err := object.Attrs(ctx)
-	if err != nil {
+	if err := w.Close(); err != nil {
 		logger.LogIf(ctx, err)
 		return minio.ObjectInfo{}, gcsToObjectError(err, bucket, key)
 	}
 
-	return fromGCSAttrsToObjectInfo(attrs), nil
+	return fromGCSAttrsToObjectInfo(w.Attrs()), nil
 }
 
 // CopyObject - Copies a blob from source container to destination container.
@@ -958,6 +955,14 @@ func (l *gcsGateway) DeleteObject(ctx context.Context, bucket string, object str
 	return nil
 }
 
+func (l *gcsGateway) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = l.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // NewMultipartUpload - upload object in multiple parts
 func (l *gcsGateway) NewMultipartUpload(ctx context.Context, bucket string, key string, o minio.ObjectOptions) (uploadID string, err error) {
 	// generate new uploadid
@@ -1093,7 +1098,10 @@ func (l *gcsGateway) PutObjectPart(ctx context.Context, bucket string, key strin
 		return minio.PartInfo{}, gcsToObjectError(err, bucket, key)
 	}
 	// Make sure to close the object writer upon success.
-	w.Close()
+	if err := w.Close(); err != nil {
+		logger.LogIf(ctx, err)
+		return minio.PartInfo{}, gcsToObjectError(err, bucket, key)
+	}
 	return minio.PartInfo{
 		PartNumber:   partNumber,
 		ETag:         etag,

cmd/gateway/hdfs/gateway-hdfs.go

@@ -29,8 +29,8 @@ import (
 	"syscall"
 	"time"
 
-	"github.com/colinmarc/hdfs/v2"
 	"github.com/minio/cli"
+	"github.com/minio/hdfs/v3"
 	"github.com/minio/minio-go/pkg/s3utils"
 	minio "github.com/minio/minio/cmd"
 	"github.com/minio/minio/cmd/logger"
@@ -287,24 +287,9 @@ func (n *hdfsObjects) ListBuckets(ctx context.Context) (buckets []minio.BucketIn
 	return buckets, nil
 }
 
-func (n *hdfsObjects) isObjectDir(bucket, object string) bool {
-	f, err := n.clnt.Open(minio.PathJoin(hdfsSeparator, bucket, object))
-	if err != nil {
-		return false
-	}
-	defer f.Close()
-
-	entries, err := f.Readdir(1)
-	if err != nil {
-		return false
-	}
-
-	return len(entries) == 0
-}
-
-func (n *hdfsObjects) listDirFactory(isLeaf minio.IsLeafFunc) minio.ListDirFunc {
+func (n *hdfsObjects) listDirFactory() minio.ListDirFunc {
 	// listDir - lists all the entries at a given prefix and given entry in the prefix.
-	listDir := func(bucket, prefixDir, prefixEntry string) (entries []string, delayIsLeaf bool) {
+	listDir := func(bucket, prefixDir, prefixEntry string) (entries []string) {
 		f, err := n.clnt.Open(minio.PathJoin(hdfsSeparator, bucket, prefixDir))
 		if err != nil {
 			if os.IsNotExist(err) {
@@ -327,8 +312,7 @@ func (n *hdfsObjects) listDirFactory(isLeaf minio.IsLeafFunc) minio.ListDirFunc
 			}
 		}
 		fis = nil
-		entries, delayIsLeaf = minio.FilterListEntries(bucket, prefixDir, entries, prefixEntry, isLeaf)
-		return entries, delayIsLeaf
+		return minio.FilterMatchingPrefix(entries, prefixEntry)
 	}
 
 	// Return list factory instance.
@@ -337,27 +321,11 @@ func (n *hdfsObjects) listDirFactory(isLeaf minio.IsLeafFunc) minio.ListDirFunc
 
 // ListObjects lists all blobs in HDFS bucket filtered by prefix.
 func (n *hdfsObjects) ListObjects(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (loi minio.ListObjectsInfo, err error) {
-	isLeaf := func(bucket, object string) bool {
-		// bucket argument is unused as we don't need to StatFile
-		// to figure if it's a file, just need to check that the
-		// object string does not end with "/".
-		return !strings.HasSuffix(object, hdfsSeparator)
-	}
-
-	// Return true if the specified object is an empty directory
-	isLeafDir := func(bucket, object string) bool {
-		if !strings.HasSuffix(object, hdfsSeparator) {
-			return false
-		}
-		return n.isObjectDir(bucket, object)
-	}
-	listDir := n.listDirFactory(isLeaf)
-
 	getObjectInfo := func(ctx context.Context, bucket, entry string) (minio.ObjectInfo, error) {
 		return n.GetObjectInfo(ctx, bucket, entry, minio.ObjectOptions{})
 	}
 
-	return minio.ListObjects(ctx, n, bucket, prefix, marker, delimiter, maxKeys, n.listPool, isLeaf, isLeafDir, listDir, getObjectInfo, getObjectInfo)
+	return minio.ListObjects(ctx, n, bucket, prefix, marker, delimiter, maxKeys, n.listPool, n.listDirFactory(), getObjectInfo, getObjectInfo)
 }
 
 // Check if the given error corresponds to ENOTEMPTY for unix
@@ -432,6 +400,14 @@ func (n *hdfsObjects) DeleteObject(ctx context.Context, bucket, object string) e
 	return hdfsToObjectErr(ctx, n.deleteObject(minio.PathJoin(hdfsSeparator, bucket), minio.PathJoin(hdfsSeparator, bucket, object)), bucket, object)
 }
 
+func (n *hdfsObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = n.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 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) {
 	objInfo, err := n.GetObjectInfo(ctx, bucket, object, opts)
 	if err != nil {
@@ -477,13 +453,15 @@ func (n *hdfsObjects) GetObject(ctx context.Context, bucket, key string, startOf
 	if err != nil {
 		return hdfsToObjectErr(ctx, err, bucket, key)
 	}
-	if _, err = rd.Seek(startOffset, io.SeekStart); err != nil {
-		return hdfsToObjectErr(ctx, err, bucket, key)
+	defer rd.Close()
+	_, err = io.Copy(writer, io.NewSectionReader(rd, startOffset, length))
+	if err == io.ErrClosedPipe {
+		// hdfs library doesn't send EOF correctly, so io.Copy attempts
+		// to write which returns io.ErrClosedPipe - just ignore
+		// this for now.
+		err = nil
 	}
-	if _, err = io.Copy(writer, io.LimitReader(rd, length)); err != nil {
-		return hdfsToObjectErr(ctx, err, bucket, key)
-	}
-	return nil
+	return hdfsToObjectErr(ctx, err, bucket, key)
 }
 
 // GetObjectInfo reads object info and replies back ObjectInfo.
@@ -528,17 +506,19 @@ func (n *hdfsObjects) PutObject(ctx context.Context, bucket string, object strin
 			return objInfo, hdfsToObjectErr(ctx, err, bucket, object)
 		}
 		defer n.deleteObject(minio.PathJoin(hdfsSeparator, minioMetaTmpBucket), tmpname)
-		defer w.Close()
 		if _, err = io.Copy(w, r); err != nil {
+			w.Close()
 			return objInfo, hdfsToObjectErr(ctx, err, bucket, object)
 		}
 		dir := path.Dir(name)
 		if dir != "" {
 			if err = n.clnt.MkdirAll(dir, os.FileMode(0755)); err != nil {
+				w.Close()
 				n.deleteObject(minio.PathJoin(hdfsSeparator, bucket), dir)
 				return objInfo, hdfsToObjectErr(ctx, err, bucket, object)
 			}
 		}
+		w.Close()
 		if err = n.clnt.Rename(tmpname, name); err != nil {
 			return objInfo, hdfsToObjectErr(ctx, err, bucket, object)
 		}
@@ -675,10 +655,7 @@ func (n *hdfsObjects) CompleteMultipartUpload(ctx context.Context, bucket, objec
 	}
 
 	// Calculate s3 compatible md5sum for complete multipart.
-	s3MD5, err := minio.GetCompleteMultipartMD5(ctx, parts)
-	if err != nil {
-		return objInfo, err
-	}
+	s3MD5 := minio.ComputeCompleteMultipartMD5(parts)
 
 	return minio.ObjectInfo{
 		Bucket:  bucket,

cmd/gateway/oss/gateway-oss.go

@@ -710,6 +710,14 @@ func (l *ossObjects) DeleteObject(ctx context.Context, bucket, object string) er
 	return nil
 }
 
+func (l *ossObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = l.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // fromOSSClientListMultipartsInfo converts oss ListMultipartUploadResult to ListMultipartsInfo
 func fromOSSClientListMultipartsInfo(lmur oss.ListMultipartUploadResult) minio.ListMultipartsInfo {
 	uploads := make([]minio.MultipartInfo, len(lmur.Uploads))

cmd/gateway/s3/gateway-s3-metadata.go

@@ -241,7 +241,7 @@ func getGWMetadata(ctx context.Context, bucket, prefix string, gwMeta gwMetaV1)
 		logger.LogIf(ctx, err)
 		return nil, err
 	}
-	hashReader, err := hash.NewReader(bytes.NewReader(metadataBytes), int64(len(metadataBytes)), "", "", int64(len(metadataBytes)))
+	hashReader, err := hash.NewReader(bytes.NewReader(metadataBytes), int64(len(metadataBytes)), "", "", int64(len(metadataBytes)), false)
 	if err != nil {
 		return nil, err
 	}

cmd/gateway/s3/gateway-s3.go

@@ -500,6 +500,14 @@ func (l *s3Objects) DeleteObject(ctx context.Context, bucket string, object stri
 	return nil
 }
 
+func (l *s3Objects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	for idx, object := range objects {
+		errs[idx] = l.DeleteObject(ctx, bucket, object)
+	}
+	return errs, nil
+}
+
 // ListMultipartUploads lists all multipart uploads.
 func (l *s3Objects) ListMultipartUploads(ctx context.Context, bucket string, prefix string, keyMarker string, uploadIDMarker string, delimiter string, maxUploads int) (lmi minio.ListMultipartsInfo, e error) {
 	result, err := l.Client.ListMultipartUploads(bucket, prefix, keyMarker, uploadIDMarker, delimiter, maxUploads)

cmd/globals.go

@@ -92,9 +92,10 @@ const (
 )
 
 var globalCLIContext = struct {
-	JSON, Quiet bool
-	Anonymous   bool
-	Addr        string
+	JSON, Quiet    bool
+	Anonymous      bool
+	Addr           string
+	StrictS3Compat bool
 }{}
 
 var (

cmd/http/conn_bug_21133.go

@@ -1,48 +0,0 @@
-/*
- * 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 http
-
-import (
-	"net"
-	"sync/atomic"
-	"time"
-)
-
-// QuirkConn - similar to golang net.Conn struct, but contains a workaround of the
-// following the go bug reported here https://github.com/golang/go/issues/21133.
-// Once the bug will be fixed, we can remove this structure and replaces it with
-// the standard net.Conn
-type QuirkConn struct {
-	net.Conn
-	hadReadDeadlineInPast int32 // atomic
-}
-
-// SetReadDeadline - implements a workaround of SetReadDeadline go bug
-func (q *QuirkConn) SetReadDeadline(t time.Time) error {
-	inPast := int32(0)
-	if t.Before(time.Now()) {
-		inPast = 1
-	}
-	atomic.StoreInt32(&q.hadReadDeadlineInPast, inPast)
-	return q.Conn.SetReadDeadline(t)
-}
-
-// canSetReadDeadline - returns if it is safe to set a new
-// read deadline without triggering golang/go#21133 issue.
-func (q *QuirkConn) canSetReadDeadline() bool {
-	return atomic.LoadInt32(&q.hadReadDeadlineInPast) != 1
-}

cmd/http/deadlineconn.go

@@ -23,7 +23,7 @@ import (
 
 // DeadlineConn - is a generic stream-oriented network connection supporting buffered reader and read/write timeout.
 type DeadlineConn struct {
-	QuirkConn
+	net.Conn
 	readTimeout            time.Duration // sets the read timeout in the connection.
 	writeTimeout           time.Duration // sets the write timeout in the connection.
 	updateBytesReadFunc    func(int)     // function to be called to update bytes read.
@@ -32,7 +32,7 @@ type DeadlineConn struct {
 
 // Sets read timeout
 func (c *DeadlineConn) setReadTimeout() {
-	if c.readTimeout != 0 && c.canSetReadDeadline() {
+	if c.readTimeout != 0 {
 		c.SetReadDeadline(time.Now().UTC().Add(c.readTimeout))
 	}
 }
@@ -68,7 +68,7 @@ func (c *DeadlineConn) Write(b []byte) (n int, err error) {
 // newDeadlineConn - creates a new connection object wrapping net.Conn with deadlines.
 func newDeadlineConn(c net.Conn, readTimeout, writeTimeout time.Duration, updateBytesReadFunc, updateBytesWrittenFunc func(int)) *DeadlineConn {
 	return &DeadlineConn{
-		QuirkConn:              QuirkConn{Conn: c},
+		Conn:                   c,
 		readTimeout:            readTimeout,
 		writeTimeout:           writeTimeout,
 		updateBytesReadFunc:    updateBytesReadFunc,

cmd/http/timeoutconn.go

@@ -1,62 +0,0 @@
-/*
- * MinIO Cloud Storage, (C) 2018 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 http
-
-import (
-	"net"
-	"time"
-)
-
-// TimeoutConn - is wrapped net.Conn with read/write timeouts.
-type TimeoutConn struct {
-	QuirkConn
-	readTimeout  time.Duration
-	writeTimeout time.Duration
-}
-
-func (c *TimeoutConn) setReadTimeout() {
-	if c.readTimeout != 0 && c.canSetReadDeadline() {
-		c.SetReadDeadline(time.Now().UTC().Add(c.readTimeout))
-	}
-}
-
-func (c *TimeoutConn) setWriteTimeout() {
-	if c.writeTimeout != 0 {
-		c.SetWriteDeadline(time.Now().UTC().Add(c.writeTimeout))
-	}
-}
-
-// Read - reads data from the connection with timeout.
-func (c *TimeoutConn) Read(b []byte) (n int, err error) {
-	c.setReadTimeout()
-	return c.Conn.Read(b)
-}
-
-// Write - writes data to the connection with timeout.
-func (c *TimeoutConn) Write(b []byte) (n int, err error) {
-	c.setWriteTimeout()
-	return c.Conn.Write(b)
-}
-
-// NewTimeoutConn - creates a new timeout connection.
-func NewTimeoutConn(c net.Conn, readTimeout, writeTimeout time.Duration) *TimeoutConn {
-	return &TimeoutConn{
-		QuirkConn:    QuirkConn{Conn: c},
-		readTimeout:  readTimeout,
-		writeTimeout: writeTimeout,
-	}
-}

cmd/iam.go

@@ -62,6 +62,9 @@ type IAMSys struct {
 
 // Load - loads iam subsystem
 func (sys *IAMSys) Load(objAPI ObjectLayer) error {
+	if globalEtcdClient != nil {
+		return sys.refreshEtcd()
+	}
 	return sys.refresh(objAPI)
 }
 
@@ -71,21 +74,52 @@ func (sys *IAMSys) Init(objAPI ObjectLayer) error {
 		return errInvalidArgument
 	}
 
-	defer func() {
-		// Refresh IAMSys in background.
-		go func() {
-			ticker := time.NewTicker(globalRefreshIAMInterval)
-			defer ticker.Stop()
-			for {
-				select {
-				case <-GlobalServiceDoneCh:
-					return
-				case <-ticker.C:
-					sys.refresh(objAPI)
+	if globalEtcdClient != nil {
+		defer func() {
+			go func() {
+				// Refresh IAMSys with etcd watch.
+				for {
+					watchCh := globalEtcdClient.Watch(context.Background(), iamConfigPrefix)
+					select {
+					case <-GlobalServiceDoneCh:
+						return
+					case watchResp, ok := <-watchCh:
+						if !ok {
+							time.Sleep(1 * time.Second)
+							continue
+						}
+						if err := watchResp.Err(); err != nil {
+							logger.LogIf(context.Background(), err)
+							// log and retry.
+							time.Sleep(1 * time.Second)
+							continue
+						}
+						for _, event := range watchResp.Events {
+							if event.IsModify() || event.IsCreate() || event.Type == etcd.EventTypeDelete {
+								sys.refreshEtcd()
+							}
+						}
+					}
 				}
-			}
+			}()
 		}()
-	}()
+	} else {
+		defer func() {
+			// Refresh IAMSys in background.
+			go func() {
+				ticker := time.NewTicker(globalRefreshIAMInterval)
+				defer ticker.Stop()
+				for {
+					select {
+					case <-GlobalServiceDoneCh:
+						return
+					case <-ticker.C:
+						sys.refresh(objAPI)
+					}
+				}
+			}()
+		}()
+	}
 
 	doneCh := make(chan struct{})
 	defer close(doneCh)
@@ -95,6 +129,9 @@ func (sys *IAMSys) Init(objAPI ObjectLayer) error {
 	//  - Read quorum is lost just after the initialization
 	//    of the object layer.
 	for range newRetryTimerSimple(doneCh) {
+		if globalEtcdClient != nil {
+			return sys.refreshEtcd()
+		}
 		// Load IAMSys once during boot.
 		if err := sys.refresh(objAPI); err != nil {
 			if err == errDiskNotFound ||
@@ -697,35 +734,51 @@ func setDefaultCannedPolicies(policies map[string]iampolicy.Policy) {
 	}
 }
 
+func (sys *IAMSys) refreshEtcd() error {
+	iamUsersMap := make(map[string]auth.Credentials)
+	iamPolicyMap := make(map[string]string)
+	iamCannedPolicyMap := make(map[string]iampolicy.Policy)
+
+	if err := reloadEtcdPolicies(iamConfigPoliciesPrefix, iamCannedPolicyMap); err != nil {
+		return err
+	}
+	if err := reloadEtcdUsers(iamConfigUsersPrefix, iamUsersMap, iamPolicyMap); err != nil {
+		return err
+	}
+	if err := reloadEtcdUsers(iamConfigSTSPrefix, iamUsersMap, iamPolicyMap); err != nil {
+		return err
+	}
+
+	// Sets default canned policies, if none are set.
+	setDefaultCannedPolicies(iamCannedPolicyMap)
+
+	sys.Lock()
+	defer sys.Unlock()
+
+	sys.iamUsersMap = iamUsersMap
+	sys.iamPolicyMap = iamPolicyMap
+	sys.iamCannedPolicyMap = iamCannedPolicyMap
+
+	return nil
+}
+
 // Refresh IAMSys.
 func (sys *IAMSys) refresh(objAPI ObjectLayer) error {
 	iamUsersMap := make(map[string]auth.Credentials)
 	iamPolicyMap := make(map[string]string)
 	iamCannedPolicyMap := make(map[string]iampolicy.Policy)
 
-	if globalEtcdClient != nil {
-		if err := reloadEtcdPolicies(iamConfigPoliciesPrefix, iamCannedPolicyMap); err != nil {
-			return err
-		}
-		if err := reloadEtcdUsers(iamConfigUsersPrefix, iamUsersMap, iamPolicyMap); err != nil {
-			return err
-		}
-		if err := reloadEtcdUsers(iamConfigSTSPrefix, iamUsersMap, iamPolicyMap); err != nil {
-			return err
-		}
-	} else {
-		if err := reloadPolicies(objAPI, iamConfigPoliciesPrefix, iamCannedPolicyMap); err != nil {
-			return err
-		}
-		if err := reloadUsers(objAPI, iamConfigUsersPrefix, iamUsersMap, iamPolicyMap); err != nil {
-			return err
-		}
-		if err := reloadUsers(objAPI, iamConfigSTSPrefix, iamUsersMap, iamPolicyMap); err != nil {
-			return err
-		}
+	if err := reloadPolicies(objAPI, iamConfigPoliciesPrefix, iamCannedPolicyMap); err != nil {
+		return err
+	}
+	if err := reloadUsers(objAPI, iamConfigUsersPrefix, iamUsersMap, iamPolicyMap); err != nil {
+		return err
+	}
+	if err := reloadUsers(objAPI, iamConfigSTSPrefix, iamUsersMap, iamPolicyMap); err != nil {
+		return err
 	}
 
-	// Sets default canned policies, if none set.
+	// Sets default canned policies, if none are set.
 	setDefaultCannedPolicies(iamCannedPolicyMap)
 
 	sys.Lock()

cmd/logger/target/console/console.go

@@ -45,12 +45,13 @@ func (c *Target) Send(e interface{}) error {
 		return nil
 	}
 
-	trace := make([]string, len(entry.Trace.Source))
+	traceLength := len(entry.Trace.Source)
+	trace := make([]string, traceLength)
 
 	// Add a sequence number and formatting for each stack trace
 	// No formatting is required for the first entry
 	for i, element := range entry.Trace.Source {
-		trace[i] = fmt.Sprintf("%8v: %s", i+1, element)
+		trace[i] = fmt.Sprintf("%8v: %s", traceLength-i, element)
 	}
 
 	tagString := ""

cmd/main.go

@@ -32,24 +32,28 @@ var globalFlags = []cli.Flag{
 	cli.StringFlag{
 		Name:  "config-dir, C",
 		Value: defaultConfigDir.Get(),
-		Usage: "[DEPRECATED] Path to legacy configuration directory.",
+		Usage: "[DEPRECATED] path to legacy configuration directory",
 	},
 	cli.StringFlag{
 		Name:  "certs-dir, S",
 		Value: defaultCertsDir.Get(),
-		Usage: "Path to certs directory.",
+		Usage: "path to certs directory",
 	},
 	cli.BoolFlag{
 		Name:  "quiet",
-		Usage: "Disable startup information.",
+		Usage: "disable startup information",
 	},
 	cli.BoolFlag{
 		Name:  "anonymous",
-		Usage: "Hide sensitive information from logging.",
+		Usage: "hide sensitive information from logging",
 	},
 	cli.BoolFlag{
 		Name:  "json",
-		Usage: "Output server logs and startup information in json format.",
+		Usage: "output server logs and startup information in json format",
+	},
+	cli.BoolFlag{
+		Name:  "compat",
+		Usage: "trade off performance for S3 compatibility",
 	},
 }
 
@@ -118,7 +122,7 @@ func newApp(name string) *cli.App {
 	// Set up app.
 	cli.HelpFlag = cli.BoolFlag{
 		Name:  "help, h",
-		Usage: "Show help.",
+		Usage: "show help",
 	}
 
 	app := cli.NewApp()

cmd/merge-walk-pool.go

@@ -0,0 +1,143 @@
+/*
+ * MinIO Cloud Storage, (C) 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 (
+	"reflect"
+	"sync"
+	"time"
+)
+
+const (
+	globalMergeLookupTimeout = time.Minute * 1 // 1 minutes.
+)
+
+// mergeWalk - represents the go routine that does the merge walk.
+type mergeWalk struct {
+	entryChs   []FileInfoCh
+	endWalkCh  chan struct{}   // To signal when mergeWalk go-routine should end.
+	endTimerCh chan<- struct{} // To signal when timer go-routine should end.
+}
+
+// MergeWalkPool - pool of mergeWalk go routines.
+// A mergeWalk is added to the pool by Set() and removed either by
+// doing a Release() or if the concerned timer goes off.
+// mergeWalkPool's purpose is to maintain active mergeWalk go-routines in a map so that
+// it can be looked up across related list calls.
+type MergeWalkPool struct {
+	pool    map[listParams][]mergeWalk
+	timeOut time.Duration
+	lock    *sync.Mutex
+}
+
+// NewMergeWalkPool - initialize new tree walk pool.
+func NewMergeWalkPool(timeout time.Duration) *MergeWalkPool {
+	tPool := &MergeWalkPool{
+		pool:    make(map[listParams][]mergeWalk),
+		timeOut: timeout,
+		lock:    &sync.Mutex{},
+	}
+	return tPool
+}
+
+// Release - selects a mergeWalk from the pool based on the input
+// listParams, removes it from the pool, and returns the MergeWalkResult
+// channel.
+// Returns nil if listParams does not have an asccociated mergeWalk.
+func (t MergeWalkPool) Release(params listParams) ([]FileInfoCh, chan struct{}) {
+	t.lock.Lock()
+	defer t.lock.Unlock()
+	walks, ok := t.pool[params] // Pick the valid walks.
+	if ok {
+		if len(walks) > 0 {
+			// Pop out the first valid walk entry.
+			walk := walks[0]
+			walks = walks[1:]
+			if len(walks) > 0 {
+				t.pool[params] = walks
+			} else {
+				delete(t.pool, params)
+			}
+			walk.endTimerCh <- struct{}{}
+			return walk.entryChs, walk.endWalkCh
+		}
+	}
+	// Release return nil if params not found.
+	return nil, nil
+}
+
+// Set - adds a mergeWalk to the mergeWalkPool.
+// Also starts a timer go-routine that ends when:
+// 1) time.After() expires after t.timeOut seconds.
+//    The expiration is needed so that the mergeWalk go-routine resources are freed after a timeout
+//    if the S3 client does only partial listing of objects.
+// 2) Relase() signals the timer go-routine to end on endTimerCh.
+//    During listing the timer should not timeout and end the mergeWalk go-routine, hence the
+//    timer go-routine should be ended.
+func (t MergeWalkPool) Set(params listParams, resultChs []FileInfoCh, endWalkCh chan struct{}) {
+	t.lock.Lock()
+	defer t.lock.Unlock()
+
+	// Should be a buffered channel so that Release() never blocks.
+	endTimerCh := make(chan struct{}, 1)
+
+	walkInfo := mergeWalk{
+		entryChs:   resultChs,
+		endWalkCh:  endWalkCh,
+		endTimerCh: endTimerCh,
+	}
+
+	// Append new walk info.
+	t.pool[params] = append(t.pool[params], walkInfo)
+
+	// Timer go-routine which times out after t.timeOut seconds.
+	go func(endTimerCh <-chan struct{}, walkInfo mergeWalk) {
+		select {
+		// Wait until timeOut
+		case <-time.After(t.timeOut):
+			// Timeout has expired. Remove the mergeWalk from mergeWalkPool and
+			// end the mergeWalk go-routine.
+			t.lock.Lock()
+			walks, ok := t.pool[params]
+			if ok {
+				// Trick of filtering without allocating
+				// https://github.com/golang/go/wiki/SliceTricks#filtering-without-allocating
+				nwalks := walks[:0]
+				// Look for walkInfo, remove it from the walks list.
+				for _, walk := range walks {
+					if !reflect.DeepEqual(walk, walkInfo) {
+						nwalks = append(nwalks, walk)
+					}
+				}
+				if len(nwalks) == 0 {
+					// No more mergeWalk go-routines associated with listParams
+					// hence remove map entry.
+					delete(t.pool, params)
+				} else {
+					// There are more mergeWalk go-routines associated with listParams
+					// hence save the list in the map.
+					t.pool[params] = nwalks
+				}
+			}
+			// Signal the mergeWalk go-routine to die.
+			close(endWalkCh)
+			t.lock.Unlock()
+		case <-endTimerCh:
+			return
+		}
+	}(endTimerCh, walkInfo)
+}

cmd/merge-walk-pool_test.go

@@ -0,0 +1,103 @@
+/*
+ * MinIO Cloud Storage, (C) 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 (
+	"testing"
+	"time"
+)
+
+// Test if tree walker go-routine is removed from the pool after timeout
+// and that is available in the pool before the timeout.
+func TestMergeWalkPoolBasic(t *testing.T) {
+	// Create a treeWalkPool
+	tw := NewMergeWalkPool(1 * time.Second)
+
+	// Create sample params
+	params := listParams{
+		bucket: "test-bucket",
+	}
+
+	endWalkCh := make(chan struct{})
+	// Add a treeWalk to the pool
+	tw.Set(params, []FileInfoCh{}, endWalkCh)
+
+	// Wait for treeWalkPool timeout to happen
+	<-time.After(2 * time.Second)
+	if c1, _ := tw.Release(params); c1 != nil {
+		t.Error("treeWalk go-routine must have been freed")
+	}
+
+	// Add the treeWalk back to the pool
+	endWalkCh = make(chan struct{})
+	tw.Set(params, []FileInfoCh{}, endWalkCh)
+
+	// Release the treeWalk before timeout
+	select {
+	case <-time.After(1 * time.Second):
+		break
+	default:
+		if c1, _ := tw.Release(params); c1 == nil {
+			t.Error("treeWalk go-routine got freed before timeout")
+		}
+	}
+}
+
+// Test if multiple merge walkers for the same listParams are managed as expected by the pool.
+func TestManyMergeWalksSameParam(t *testing.T) {
+	// Create a treeWalkPool.
+	tw := NewMergeWalkPool(5 * time.Second)
+
+	// Create sample params.
+	params := listParams{
+		bucket: "test-bucket",
+	}
+
+	select {
+	// This timeout is an upper-bound. This is started
+	// before the first treeWalk go-routine's timeout period starts.
+	case <-time.After(5 * time.Second):
+		break
+	default:
+		// Create many treeWalk go-routines for the same params.
+		for i := 0; i < 10; i++ {
+			endWalkCh := make(chan struct{})
+			walkChs := make([]FileInfoCh, 0)
+			tw.Set(params, walkChs, endWalkCh)
+		}
+
+		tw.lock.Lock()
+		if walks, ok := tw.pool[params]; ok {
+			if len(walks) != 10 {
+				t.Error("There aren't as many walks as were Set")
+			}
+		}
+		tw.lock.Unlock()
+		for i := 0; i < 10; i++ {
+			tw.lock.Lock()
+			if walks, ok := tw.pool[params]; ok {
+				// Before ith Release we should have 10-i treeWalk go-routines.
+				if 10-i != len(walks) {
+					t.Error("There aren't as many walks as were Set")
+				}
+			}
+			tw.lock.Unlock()
+			tw.Release(params)
+		}
+	}
+
+}

cmd/naughty-disk_test.go

@@ -111,11 +111,18 @@ func (d *naughtyDisk) DeleteVol(volume string) (err error) {
 	return d.disk.DeleteVol(volume)
 }
 
-func (d *naughtyDisk) ListDir(volume, path string, count int) (entries []string, err error) {
+func (d *naughtyDisk) Walk(volume, path, marker string, recursive bool, leafFile string, readMetadataFn readMetadataFunc, endWalkCh chan struct{}) (chan FileInfo, error) {
+	if err := d.calcError(); err != nil {
+		return nil, err
+	}
+	return d.disk.Walk(volume, path, marker, recursive, leafFile, readMetadataFn, endWalkCh)
+}
+
+func (d *naughtyDisk) ListDir(volume, path string, count int, leafFile string) (entries []string, err error) {
 	if err := d.calcError(); err != nil {
 		return []string{}, err
 	}
-	return d.disk.ListDir(volume, path, count)
+	return d.disk.ListDir(volume, path, count, leafFile)
 }
 
 func (d *naughtyDisk) ReadFile(volume string, path string, offset int64, buf []byte, verifier *BitrotVerifier) (n int64, err error) {
@@ -167,6 +174,14 @@ func (d *naughtyDisk) DeleteFile(volume string, path string) (err error) {
 	return d.disk.DeleteFile(volume, path)
 }
 
+func (d *naughtyDisk) DeleteFileBulk(volume string, paths []string) ([]error, error) {
+	errs := make([]error, len(paths))
+	for idx, path := range paths {
+		errs[idx] = d.disk.DeleteFile(volume, path)
+	}
+	return errs, nil
+}
+
 func (d *naughtyDisk) WriteAll(volume string, path string, buf []byte) (err error) {
 	if err := d.calcError(); err != nil {
 		return err

cmd/net.go

@@ -17,7 +17,6 @@
 package cmd
 
 import (
-	"context"
 	"errors"
 	"fmt"
 	"net"
@@ -27,9 +26,7 @@ import (
 	"strconv"
 	"strings"
 	"syscall"
-	"time"
 
-	humanize "github.com/dustin/go-humanize"
 	"github.com/minio/minio-go/pkg/set"
 	"github.com/minio/minio/cmd/logger"
 )
@@ -102,40 +99,7 @@ func getHostIP(host string) (ipList set.StringSet, err error) {
 	var ips []net.IP
 
 	if ips, err = net.LookupIP(host); err != nil {
-		// return err if not Docker or Kubernetes
-		// We use IsDocker() method to check for Docker Swarm environment
-		// as there is no reliable way to clearly identify Swarm from
-		// Docker environment.
-		if !IsDocker() && !IsKubernetes() {
-			return ipList, err
-		}
-
-		// channel to indicate completion of host resolution
-		doneCh := make(chan struct{})
-		// Indicate retry routine to exit cleanly, upon this function return.
-		defer close(doneCh)
-		// Mark the starting time
-		startTime := time.Now()
-		// wait for hosts to resolve in exponentialbackoff manner
-		for range newRetryTimerSimple(doneCh) {
-			// Retry infinitely on Kubernetes and Docker swarm.
-			// This is needed as the remote hosts are sometime
-			// not available immediately.
-			if ips, err = net.LookupIP(host); err == nil {
-				break
-			}
-			// time elapsed
-			timeElapsed := time.Since(startTime)
-			// log error only if more than 1s elapsed
-			if timeElapsed > time.Second {
-				// log the message to console about the host not being
-				// resolveable.
-				reqInfo := (&logger.ReqInfo{}).AppendTags("host", host)
-				reqInfo.AppendTags("elapsedTime", humanize.RelTime(startTime, startTime.Add(timeElapsed), "elapsed", ""))
-				ctx := logger.SetReqInfo(context.Background(), reqInfo)
-				logger.LogIf(ctx, err)
-			}
-		}
+		return ipList, err
 	}
 
 	ipList = set.NewStringSet()

cmd/object-api-common.go

@@ -116,7 +116,7 @@ func cleanupDir(ctx context.Context, storage StorageAPI, volume, dirPath string)
 		}
 
 		// If it's a directory, list and call delFunc() for each entry.
-		entries, err := storage.ListDir(volume, entryPath, -1)
+		entries, err := storage.ListDir(volume, entryPath, -1, "")
 		// If entryPath prefix never existed, safe to ignore.
 		if err == errFileNotFound {
 			return nil
@@ -144,6 +144,78 @@ func cleanupDir(ctx context.Context, storage StorageAPI, volume, dirPath string)
 	return err
 }
 
+// Cleanup objects in bulk and recursively: each object will have a list of sub-files to delete in the backend
+func cleanupObjectsBulk(ctx context.Context, storage StorageAPI, volume string, objsPaths []string, errs []error) ([]error, error) {
+	// The list of files in disk to delete
+	var filesToDelete []string
+	// Map files to delete to the passed objsPaths
+	var filesToDeleteObjsIndexes []int
+
+	// Traverse and return the list of sub entries
+	var traverse func(string) ([]string, error)
+	traverse = func(entryPath string) ([]string, error) {
+		var output = make([]string, 0)
+		if !hasSuffix(entryPath, slashSeparator) {
+			output = append(output, entryPath)
+			return output, nil
+		}
+		entries, err := storage.ListDir(volume, entryPath, -1, "")
+		if err != nil {
+			if err == errFileNotFound {
+				return nil, nil
+			}
+			return nil, err
+		}
+
+		for _, entry := range entries {
+			subEntries, err := traverse(pathJoin(entryPath, entry))
+			if err != nil {
+				return nil, err
+			}
+			output = append(output, subEntries...)
+		}
+		return output, nil
+	}
+
+	// Find and collect the list of files to remove associated
+	// to the passed objects paths
+	for idx, objPath := range objsPaths {
+		if errs[idx] != nil {
+			continue
+		}
+		output, err := traverse(objPath)
+		if err != nil {
+			errs[idx] = err
+			continue
+		} else {
+			errs[idx] = nil
+		}
+		filesToDelete = append(filesToDelete, output...)
+		for i := 0; i < len(output); i++ {
+			filesToDeleteObjsIndexes = append(filesToDeleteObjsIndexes, idx)
+		}
+	}
+
+	// Reverse the list so remove can succeed
+	reverseStringSlice(filesToDelete)
+
+	dErrs, err := storage.DeleteFileBulk(volume, filesToDelete)
+	if err != nil {
+		return nil, err
+	}
+
+	// Map files deletion errors to the correspondent objects
+	for i := range dErrs {
+		if dErrs[i] != nil {
+			if errs[filesToDeleteObjsIndexes[i]] != nil {
+				errs[filesToDeleteObjsIndexes[i]] = dErrs[i]
+			}
+		}
+	}
+
+	return errs, nil
+}
+
 // Removes notification.xml for a given bucket, only used during DeleteBucket.
 func removeNotificationConfig(ctx context.Context, objAPI ObjectLayer, bucket string) error {
 	// Verify bucket is valid.
@@ -162,7 +234,7 @@ func removeListenerConfig(ctx context.Context, objAPI ObjectLayer, bucket string
 	return objAPI.DeleteObject(ctx, minioMetaBucket, lcPath)
 }
 
-func listObjects(ctx context.Context, obj ObjectLayer, bucket, prefix, marker, delimiter string, maxKeys int, tpool *TreeWalkPool, isLeaf IsLeafFunc, isLeafDir IsLeafDirFunc, listDir ListDirFunc, getObjInfo func(context.Context, string, string) (ObjectInfo, error), getObjectInfoDirs ...func(context.Context, string, string) (ObjectInfo, error)) (loi ListObjectsInfo, err error) {
+func listObjects(ctx context.Context, obj ObjectLayer, bucket, prefix, marker, delimiter string, maxKeys int, tpool *TreeWalkPool, listDir ListDirFunc, getObjInfo func(context.Context, string, string) (ObjectInfo, error), getObjectInfoDirs ...func(context.Context, string, string) (ObjectInfo, error)) (loi ListObjectsInfo, err error) {
 	if err := checkListObjsArgs(ctx, bucket, prefix, marker, delimiter, obj); err != nil {
 		return loi, err
 	}
@@ -203,7 +275,7 @@ func listObjects(ctx context.Context, obj ObjectLayer, bucket, prefix, marker, d
 	walkResultCh, endWalkCh := tpool.Release(listParams{bucket, recursive, marker, prefix})
 	if walkResultCh == nil {
 		endWalkCh = make(chan struct{})
-		walkResultCh = startTreeWalk(ctx, bucket, prefix, marker, recursive, listDir, isLeaf, isLeafDir, endWalkCh)
+		walkResultCh = startTreeWalk(ctx, bucket, prefix, marker, recursive, listDir, endWalkCh)
 	}
 
 	var objInfos []ObjectInfo
@@ -218,14 +290,6 @@ func listObjects(ctx context.Context, obj ObjectLayer, bucket, prefix, marker, d
 			eof = true
 			break
 		}
-		// For any walk error return right away.
-		if walkResult.err != nil {
-			// File not found is a valid case.
-			if walkResult.err == errFileNotFound {
-				continue
-			}
-			return loi, toObjectErr(walkResult.err, bucket, prefix)
-		}
 
 		var objInfo ObjectInfo
 		var err error
@@ -235,6 +299,14 @@ func listObjects(ctx context.Context, obj ObjectLayer, bucket, prefix, marker, d
 				if err == nil {
 					break
 				}
+				if err == errFileNotFound {
+					err = nil
+					objInfo = ObjectInfo{
+						Bucket: bucket,
+						Name:   walkResult.entry,
+						IsDir:  true,
+					}
+				}
 			}
 		} else {
 			objInfo, err = getObjInfo(ctx, bucket, walkResult.entry)

cmd/object-api-errors.go

@@ -27,6 +27,22 @@ import (
 // underlying storage layer.
 func toObjectErr(err error, params ...string) error {
 	switch err {
+	case errDiskNotFound:
+		switch len(params) {
+		case 1:
+			err = BucketNotFound{Bucket: params[0]}
+		case 2:
+			err = ObjectNotFound{
+				Bucket: params[0],
+				Object: params[1],
+			}
+		case 3:
+			err = InvalidUploadID{
+				Bucket:   params[0],
+				Object:   params[1],
+				UploadID: params[2],
+			}
+		}
 	case errVolumeNotFound:
 		if len(params) >= 1 {
 			err = BucketNotFound{Bucket: params[0]}
@@ -41,6 +57,8 @@ func toObjectErr(err error, params ...string) error {
 		}
 	case errDiskFull:
 		err = StorageFull{}
+	case errTooManyOpenFiles:
+		err = SlowDown{}
 	case errFileAccessDenied:
 		if len(params) >= 2 {
 			err = PrefixAccessDenied{
@@ -63,11 +81,18 @@ func toObjectErr(err error, params ...string) error {
 			}
 		}
 	case errFileNotFound:
-		if len(params) >= 2 {
+		switch len(params) {
+		case 2:
 			err = ObjectNotFound{
 				Bucket: params[0],
 				Object: params[1],
 			}
+		case 3:
+			err = InvalidUploadID{
+				Bucket:   params[0],
+				Object:   params[1],
+				UploadID: params[2],
+			}
 		}
 	case errFileNameTooLong:
 		if len(params) >= 2 {
@@ -114,6 +139,13 @@ func (e StorageFull) Error() string {
 	return "Storage reached its minimum free disk threshold."
 }
 
+// SlowDown  too many file descriptors open or backend busy .
+type SlowDown struct{}
+
+func (e SlowDown) Error() string {
+	return "Please reduce your request rate"
+}
+
 // InsufficientReadQuorum storage cannot satisfy quorum for read operation.
 type InsufficientReadQuorum struct{}
 
@@ -321,6 +353,8 @@ func (e MalformedUploadID) Error() string {
 
 // InvalidUploadID invalid upload id.
 type InvalidUploadID struct {
+	Bucket   string
+	Object   string
 	UploadID string
 }
 

cmd/object-api-interface.go

@@ -73,6 +73,7 @@ type ObjectLayer interface {
 	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)
 	DeleteObject(ctx context.Context, bucket, object string) error
+	DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error)
 
 	// Multipart operations.
 	ListMultipartUploads(ctx context.Context, bucket, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error)

cmd/object-api-listobjects_test.go

@@ -35,11 +35,13 @@ func TestListObjects(t *testing.T) {
 }
 
 // Unit test for ListObjects in general.
-func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
-
+func testListObjects(obj ObjectLayer, instanceType string, t1 TestErrHandler) {
+	t, _ := t1.(*testing.T)
 	testBuckets := []string{
 		// This bucket is used for testing ListObject operations.
 		"test-bucket-list-object",
+		// This bucket will be tested with empty directories
+		"test-bucket-empty-dir",
 		// Will not store any objects in this bucket,
 		// Its to test ListObjects on an empty bucket.
 		"empty-bucket",
@@ -53,24 +55,27 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 
 	var err error
 	testObjects := []struct {
-		name    string
-		content string
-		meta    map[string]string
+		parentBucket string
+		name         string
+		content      string
+		meta         map[string]string
 	}{
-		{"Asia-maps.png", "asis-maps", map[string]string{"content-type": "image/png"}},
-		{"Asia/India/India-summer-photos-1", "contentstring", nil},
-		{"Asia/India/Karnataka/Bangalore/Koramangala/pics", "contentstring", nil},
-		{"newPrefix0", "newPrefix0", nil},
-		{"newPrefix1", "newPrefix1", nil},
-		{"newzen/zen/recurse/again/again/again/pics", "recurse", nil},
-		{"obj0", "obj0", nil},
-		{"obj1", "obj1", nil},
-		{"obj2", "obj2", nil},
-		{"z-empty-dir/", "", nil},
+		{testBuckets[0], "Asia-maps.png", "asis-maps", map[string]string{"content-type": "image/png"}},
+		{testBuckets[0], "Asia/India/India-summer-photos-1", "contentstring", nil},
+		{testBuckets[0], "Asia/India/Karnataka/Bangalore/Koramangala/pics", "contentstring", nil},
+		{testBuckets[0], "newPrefix0", "newPrefix0", nil},
+		{testBuckets[0], "newPrefix1", "newPrefix1", nil},
+		{testBuckets[0], "newzen/zen/recurse/again/again/again/pics", "recurse", nil},
+		{testBuckets[0], "obj0", "obj0", nil},
+		{testBuckets[0], "obj1", "obj1", nil},
+		{testBuckets[0], "obj2", "obj2", nil},
+		{testBuckets[1], "obj1", "obj1", nil},
+		{testBuckets[1], "obj2", "obj2", nil},
+		{testBuckets[1], "temporary/0/", "", nil},
 	}
 	for _, object := range testObjects {
 		md5Bytes := md5.Sum([]byte(object.content))
-		_, err = obj.PutObject(context.Background(), testBuckets[0], object.name, mustGetPutObjReader(t, bytes.NewBufferString(object.content),
+		_, err = obj.PutObject(context.Background(), object.parentBucket, object.name, mustGetPutObjReader(t, bytes.NewBufferString(object.content),
 			int64(len(object.content)), hex.EncodeToString(md5Bytes[:]), ""), ObjectOptions{UserDefined: object.meta})
 		if err != nil {
 			t.Fatalf("%s : %s", instanceType, err.Error())
@@ -96,7 +101,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-1.
@@ -193,7 +197,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-10.
@@ -205,7 +208,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-11.
@@ -215,7 +217,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 			Objects: []ObjectInfo{
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-12.
@@ -224,7 +225,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 			IsTruncated: false,
 			Objects: []ObjectInfo{
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-13.
@@ -238,7 +238,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-14.
@@ -255,7 +254,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-15.
@@ -270,7 +268,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-16.
@@ -284,7 +281,6 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj0"},
 				{Name: "obj1"},
 				{Name: "obj2"},
-				{Name: "z-empty-dir/"},
 			},
 		},
 		// ListObjectsResult-17.
@@ -371,6 +367,7 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 			Objects: []ObjectInfo{
 				{Name: "Asia-maps.png"},
 			},
+			Prefixes: []string{"Asia/"},
 		},
 		// ListObjectsResult-26.
 		// prefix = "new" and delimiter is set in the testCase.(testCase 58).
@@ -380,6 +377,7 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "newPrefix0"},
 				{Name: "newPrefix1"},
 			},
+			Prefixes: []string{"newzen/"},
 		},
 		// ListObjectsResult-27.
 		// Prefix is set to "Asia/India/" in the testCase, and delimiter is set to forward slash '/' (testCase 59).
@@ -388,6 +386,7 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 			Objects: []ObjectInfo{
 				{Name: "Asia/India/India-summer-photos-1"},
 			},
+			Prefixes: []string{"Asia/India/Karnataka/"},
 		},
 		// ListObjectsResult-28.
 		// Marker is set to "Asia/India/India-summer-photos-1" and delimiter set in the testCase, (testCase 60).
@@ -400,6 +399,7 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj1"},
 				{Name: "obj2"},
 			},
+			Prefixes: []string{"newzen/"},
 		},
 		// ListObjectsResult-29.
 		// Marker is set to "Asia/India/Karnataka/Bangalore/Koramangala/pics" in the testCase and delimeter set, (testCase 61).
@@ -412,6 +412,7 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 				{Name: "obj1"},
 				{Name: "obj2"},
 			},
+			Prefixes: []string{"newzen/"},
 		},
 		// ListObjectsResult-30.
 		// Prefix and Delimiter is set to '/', (testCase 62).
@@ -419,6 +420,31 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 			IsTruncated: false,
 			Objects:     []ObjectInfo{},
 		},
+		// ListObjectsResult-31 Empty directory, recursive listing
+		{
+			IsTruncated: false,
+			Objects: []ObjectInfo{
+				{Name: "obj1"},
+				{Name: "obj2"},
+				{Name: "temporary/0/"},
+			},
+		},
+		// ListObjectsResult-32 Empty directory, non recursive listing
+		{
+			IsTruncated: false,
+			Objects: []ObjectInfo{
+				{Name: "obj1"},
+				{Name: "obj2"},
+			},
+			Prefixes: []string{"temporary/"},
+		},
+		// ListObjectsResult-33 Listing empty directory only
+		{
+			IsTruncated: false,
+			Objects: []ObjectInfo{
+				{Name: "temporary/0/"},
+			},
+		},
 	}
 
 	testCases := []struct {
@@ -529,67 +555,87 @@ func testListObjects(obj ObjectLayer, instanceType string, t TestErrHandler) {
 		{"test-bucket-list-object", "", "Asia/India/Karnataka/Bangalore/Koramangala/pics", "/", 10, resultCases[29], nil, true},
 		// Test with prefix and delimiter set to '/'. (60)
 		{"test-bucket-list-object", "/", "", "/", 10, resultCases[30], nil, true},
-
 		// Test with invalid prefix (61)
-		{"test-bucket-list-object", "\\", "", "/", 10, resultCases[30], ObjectNameInvalid{Bucket: "test-bucket-list-object", Object: "\\"}, false},
+		{"test-bucket-list-object", "\\", "", "/", 10, ListObjectsInfo{}, ObjectNameInvalid{Bucket: "test-bucket-list-object", Object: "\\"}, false},
+		// Test listing an empty directory in recursive mode (62)
+		{"test-bucket-empty-dir", "", "", "", 10, resultCases[31], nil, true},
+		// Test listing an empty directory in a non recursive mode (63)
+		{"test-bucket-empty-dir", "", "", "/", 10, resultCases[32], nil, true},
+		// Test listing a directory which contains an empty directory (64)
+		{"test-bucket-empty-dir", "", "temporary/", "", 10, resultCases[33], nil, true},
 	}
 
 	for i, testCase := range testCases {
-		result, err := obj.ListObjects(context.Background(), testCase.bucketName, testCase.prefix, testCase.marker, testCase.delimeter, int(testCase.maxKeys))
-		if err != nil && testCase.shouldPass {
-			t.Errorf("Test %d: %s:  Expected to pass, but failed with: <ERROR> %s", i+1, instanceType, err.Error())
-		}
-		if err == nil && !testCase.shouldPass {
-			t.Errorf("Test %d: %s: Expected to fail with <ERROR> \"%s\", but passed instead", i+1, instanceType, testCase.err.Error())
-		}
-		// Failed as expected, but does it fail for the expected reason.
-		if err != nil && !testCase.shouldPass {
-			if !strings.Contains(err.Error(), testCase.err.Error()) {
-				t.Errorf("Test %d: %s: Expected to fail with error \"%s\", but instead failed with error \"%s\" instead", i+1, instanceType, testCase.err.Error(), err.Error())
+		testCase := testCase
+		t.Run(fmt.Sprintf("Test%d-%s", i+1, instanceType), func(t *testing.T) {
+			result, err := obj.ListObjects(context.Background(), testCase.bucketName,
+				testCase.prefix, testCase.marker, testCase.delimeter, int(testCase.maxKeys))
+			if err != nil && testCase.shouldPass {
+				t.Errorf("Test %d: %s:  Expected to pass, but failed with: <ERROR> %s", i+1, instanceType, err.Error())
 			}
-		}
-		// Since there are cases for which ListObjects fails, this is
-		// necessary. Test passes as expected, but the output values
-		// are verified for correctness here.
-		if err == nil && testCase.shouldPass {
-			// The length of the expected ListObjectsResult.Objects
-			// should match in both expected result from test cases
-			// and in the output. On failure calling t.Fatalf,
-			// otherwise it may lead to index out of range error in
-			// assertion following this.
-			if len(testCase.result.Objects) != len(result.Objects) {
-				t.Fatalf("Test %d: %s: Expected number of object in the result to be '%d', but found '%d' objects instead", i+1, instanceType, len(testCase.result.Objects), len(result.Objects))
+			if err == nil && !testCase.shouldPass {
+				t.Errorf("Test %d: %s: Expected to fail with <ERROR> \"%s\", but passed instead", i+1, instanceType, testCase.err.Error())
 			}
-			for j := 0; j < len(testCase.result.Objects); j++ {
-				if testCase.result.Objects[j].Name != result.Objects[j].Name {
-					t.Errorf("Test %d: %s: Expected object name to be \"%s\", but found \"%s\" instead", i+1, instanceType, testCase.result.Objects[j].Name, result.Objects[j].Name)
+			// Failed as expected, but does it fail for the expected reason.
+			if err != nil && !testCase.shouldPass {
+				if !strings.Contains(err.Error(), testCase.err.Error()) {
+					t.Errorf("Test %d: %s: Expected to fail with error \"%s\", but instead failed with error \"%s\" instead", i+1, instanceType, testCase.err.Error(), err.Error())
 				}
-				//FIXME: we should always check for ETag
-				if result.Objects[j].ETag == "" && !strings.HasSuffix(result.Objects[j].Name, slashSeparator) {
-					t.Errorf("Test %d: %s: Expected ETag to be not empty, but found empty instead (%v)", i+1, instanceType, result.Objects[j].Name)
+			}
+			// Since there are cases for which ListObjects fails, this is
+			// necessary. Test passes as expected, but the output values
+			// are verified for correctness here.
+			if err == nil && testCase.shouldPass {
+				// The length of the expected ListObjectsResult.Objects
+				// should match in both expected result from test cases
+				// and in the output. On failure calling t.Fatalf,
+				// otherwise it may lead to index out of range error in
+				// assertion following this.
+				if len(testCase.result.Objects) != len(result.Objects) {
+					t.Fatalf("Test %d: %s: Expected number of object in the result to be '%d', but found '%d' objects instead", i+1, instanceType, len(testCase.result.Objects), len(result.Objects))
+				}
+				for j := 0; j < len(testCase.result.Objects); j++ {
+					if testCase.result.Objects[j].Name != result.Objects[j].Name {
+						t.Errorf("Test %d: %s: Expected object name to be \"%s\", but found \"%s\" instead", i+1, instanceType, testCase.result.Objects[j].Name, result.Objects[j].Name)
+					}
+					// FIXME: we should always check for ETag
+					if result.Objects[j].ETag == "" && !strings.HasSuffix(result.Objects[j].Name, slashSeparator) {
+						t.Errorf("Test %d: %s: Expected ETag to be not empty, but found empty instead (%v)", i+1, instanceType, result.Objects[j].Name)
+					}
+
+				}
+
+				if len(testCase.result.Prefixes) != len(result.Prefixes) {
+					t.Fatalf("Test %d: %s: Expected number of prefixes in the result to be '%d', but found '%d' prefixes instead", i+1, instanceType, len(testCase.result.Prefixes), len(result.Prefixes))
+				}
+				for j := 0; j < len(testCase.result.Prefixes); j++ {
+					if testCase.result.Prefixes[j] != result.Prefixes[j] {
+						t.Errorf("Test %d: %s: Expected prefix name to be \"%s\", but found \"%s\" instead", i+1, instanceType, testCase.result.Prefixes[j], result.Prefixes[j])
+					}
+				}
+
+				if testCase.result.IsTruncated != result.IsTruncated {
+					t.Errorf("Test %d: %s: Expected IsTruncated flag to be %v, but instead found it to be %v", i+1, instanceType, testCase.result.IsTruncated, result.IsTruncated)
+				}
+
+				if testCase.result.IsTruncated && result.NextMarker == "" {
+					t.Errorf("Test %d: %s: Expected NextContinuationToken to contain a string since listing is truncated, but instead found it to be empty", i+1, instanceType)
+				}
+
+				if !testCase.result.IsTruncated && result.NextMarker != "" {
+					t.Errorf("Test %d: %s: Expected NextContinuationToken to be empty since listing is not truncated, but instead found `%v`", i+1, instanceType, result.NextMarker)
 				}
 
 			}
-			if testCase.result.IsTruncated != result.IsTruncated {
-				t.Errorf("Test %d: %s: Expected IsTruncated flag to be %v, but instead found it to be %v", i+1, instanceType, testCase.result.IsTruncated, result.IsTruncated)
+			// Take ListObject treeWalk go-routine to completion, if available in the treewalk pool.
+			if result.IsTruncated {
+				_, err = obj.ListObjects(context.Background(), testCase.bucketName,
+					testCase.prefix, result.NextMarker, testCase.delimeter, 1000)
+				if err != nil {
+					t.Fatal(err)
+				}
 			}
-
-			if testCase.result.IsTruncated && result.NextMarker == "" {
-				t.Errorf("Test %d: %s: Expected NextContinuationToken to contain a string since listing is truncated, but instead found it to be empty", i+1, instanceType)
-			}
-
-			if !testCase.result.IsTruncated && result.NextMarker != "" {
-				t.Errorf("Test %d: %s: Expected NextContinuationToken to be empty since listing is not truncated, but instead found `%v`", i+1, instanceType, result.NextMarker)
-			}
-
-		}
-		// Take ListObject treeWalk go-routine to completion, if available in the treewalk pool.
-		if result.IsTruncated {
-			_, err = obj.ListObjects(context.Background(), testCase.bucketName, testCase.prefix, result.NextMarker, testCase.delimeter, 1000)
-			if err != nil {
-				t.Fatal(err)
-			}
-		}
+		})
 	}
 }
 

cmd/object-api-multipart_test.go

@@ -19,7 +19,6 @@ package cmd
 import (
 	"bytes"
 	"context"
-	"encoding/hex"
 	"fmt"
 	"os"
 	"reflect"
@@ -228,8 +227,7 @@ func testPutObjectPartDiskNotFound(obj ObjectLayer, instanceType string, disks [
 	}
 
 	// This causes quorum failure verify.
-	disks = disks[len(disks)-3:]
-	for _, disk := range disks {
+	for _, disk := range disks[len(disks)-3:] {
 		os.RemoveAll(disk)
 	}
 
@@ -240,9 +238,26 @@ func testPutObjectPartDiskNotFound(obj ObjectLayer, instanceType string, disks [
 		t.Fatalf("Test %s: expected to fail but passed instead", instanceType)
 	}
 	// as majority of xl.json are not available, we expect uploadID to be not available.
-	expectedErr := InvalidUploadID{UploadID: testCase.uploadID}
-	if err.Error() != expectedErr.Error() {
-		t.Fatalf("Test %s: expected error %s, got %s instead.", instanceType, expectedErr, err)
+	expectedErr1 := InsufficientReadQuorum{}
+	if err.Error() != expectedErr1.Error() {
+		t.Fatalf("Test %s: expected error %s, got %s instead.", instanceType, expectedErr1, err)
+	}
+
+	// This causes invalid upload id.
+	for _, disk := range disks {
+		os.RemoveAll(disk)
+	}
+
+	// Object part upload should fail with bucket not found.
+	_, err = obj.PutObjectPart(context.Background(), testCase.bucketName, testCase.objName, testCase.uploadID, testCase.PartID, mustGetPutObjReader(t, bytes.NewBufferString(testCase.inputReaderData), testCase.intputDataSize, testCase.inputMd5, sha256sum), ObjectOptions{})
+	if err == nil {
+		t.Fatalf("Test %s: expected to fail but passed instead", instanceType)
+	}
+
+	// As all disks at not available, bucket not found.
+	expectedErr2 := BucketNotFound{Bucket: testCase.bucketName}
+	if err.Error() != expectedErr2.Error() {
+		t.Fatalf("Test %s: expected error %s, got %s instead.", instanceType, expectedErr2, err)
 	}
 }
 
@@ -1850,10 +1865,7 @@ func testObjectCompleteMultipartUpload(obj ObjectLayer, instanceType string, t T
 			},
 		},
 	}
-	s3MD5, err := getCompleteMultipartMD5(context.Background(), inputParts[3].parts)
-	if err != nil {
-		t.Fatalf("Obtaining S3MD5 failed")
-	}
+	s3MD5 := getCompleteMultipartMD5(inputParts[3].parts)
 
 	// Test cases with sample input values for CompleteMultipartUpload.
 	testCases := []struct {
@@ -1882,8 +1894,8 @@ func testObjectCompleteMultipartUpload(obj ObjectLayer, instanceType string, t T
 		// Asserting for Invalid UploadID (Test number 9).
 		{bucketNames[0], objectNames[0], "abc", []CompletePart{}, "", InvalidUploadID{UploadID: "abc"}, false},
 		// Test case with invalid Part Etag (Test number 10-11).
-		{bucketNames[0], objectNames[0], uploadIDs[0], []CompletePart{{ETag: "abc"}}, "", hex.ErrLength, false},
-		{bucketNames[0], objectNames[0], uploadIDs[0], []CompletePart{{ETag: "abcz"}}, "", hex.InvalidByteError(00), false},
+		{bucketNames[0], objectNames[0], uploadIDs[0], []CompletePart{{ETag: "abc"}}, "", InvalidPart{}, false},
+		{bucketNames[0], objectNames[0], uploadIDs[0], []CompletePart{{ETag: "abcz"}}, "", InvalidPart{}, false},
 		// Part number 0 doesn't exist, expecting InvalidPart error (Test number 12).
 		{bucketNames[0], objectNames[0], uploadIDs[0], []CompletePart{{ETag: "abcd", PartNumber: 0}}, "", InvalidPart{}, false},
 		// // Upload and PartNumber exists, But a deliberate ETag mismatch is introduced (Test number 13).

cmd/object-api-utils.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2015, 2016 MinIO, Inc.
+ * MinIO Cloud Storage, (C) 2015-2019 MinIO, Inc.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -190,18 +190,18 @@ func mustGetUUID() string {
 }
 
 // Create an s3 compatible MD5sum for complete multipart transaction.
-func getCompleteMultipartMD5(ctx context.Context, parts []CompletePart) (string, error) {
+func getCompleteMultipartMD5(parts []CompletePart) string {
 	var finalMD5Bytes []byte
 	for _, part := range parts {
 		md5Bytes, err := hex.DecodeString(canonicalizeETag(part.ETag))
 		if err != nil {
-			logger.LogIf(ctx, err)
-			return "", err
+			finalMD5Bytes = append(finalMD5Bytes, []byte(part.ETag)...)
+		} else {
+			finalMD5Bytes = append(finalMD5Bytes, md5Bytes...)
 		}
-		finalMD5Bytes = append(finalMD5Bytes, md5Bytes...)
 	}
 	s3MD5 := fmt.Sprintf("%s-%d", getMD5Hash(finalMD5Bytes), len(parts))
-	return s3MD5, nil
+	return s3MD5
 }
 
 // Clean unwanted fields from metadata
@@ -675,9 +675,26 @@ func (p *PutObjReader) Size() int64 {
 // as a hex encoded string
 func (p *PutObjReader) MD5CurrentHexString() string {
 	md5sumCurr := p.rawReader.MD5Current()
+	var appendHyphen bool
+	// md5sumcurr is not empty in two scenarios
+	// - server is running in strict compatibility mode
+	// - client set Content-Md5 during PUT operation
+	if len(md5sumCurr) == 0 {
+		// md5sumCurr is only empty when we are running
+		// in non-compatibility mode.
+		md5sumCurr = make([]byte, 16)
+		rand.Read(md5sumCurr)
+		appendHyphen = true
+	}
 	if p.sealMD5Fn != nil {
 		md5sumCurr = p.sealMD5Fn(md5sumCurr)
 	}
+	if appendHyphen {
+		// Make sure to return etag string upto 32 length, for SSE
+		// requests ETag might be longer and the code decrypting the
+		// ETag ignores ETag in multipart ETag form i.e <hex>-N
+		return hex.EncodeToString(md5sumCurr)[:32] + "-1"
+	}
 	return hex.EncodeToString(md5sumCurr)
 }
 

cmd/object-api-utils_test.go

@@ -18,7 +18,6 @@ package cmd
 
 import (
 	"bytes"
-	"context"
 	"io"
 	"net/http"
 	"reflect"
@@ -143,8 +142,8 @@ func TestGetCompleteMultipartMD5(t *testing.T) {
 		expectedResult string
 		expectedErr    string
 	}{
-		// Wrong MD5 hash string
-		{[]CompletePart{{ETag: "wrong-md5-hash-string"}}, "", "encoding/hex: invalid byte: U+0077 'w'"},
+		// Wrong MD5 hash string, returns md5um of hash
+		{[]CompletePart{{ETag: "wrong-md5-hash-string"}}, "0deb8cb07527b4b2669c861cb9653607-1", ""},
 
 		// Single CompletePart with valid MD5 hash string.
 		{[]CompletePart{{ETag: "cf1f738a5924e645913c984e0fe3d708"}}, "10dc1617fbcf0bd0858048cb96e6bd77-1", ""},
@@ -154,17 +153,10 @@ func TestGetCompleteMultipartMD5(t *testing.T) {
 	}
 
 	for i, test := range testCases {
-		result, err := getCompleteMultipartMD5(context.Background(), test.parts)
+		result := getCompleteMultipartMD5(test.parts)
 		if result != test.expectedResult {
 			t.Fatalf("test %d failed: expected: result=%v, got=%v", i+1, test.expectedResult, result)
 		}
-		errString := ""
-		if err != nil {
-			errString = err.Error()
-		}
-		if errString != test.expectedErr {
-			t.Fatalf("test %d failed: expected: err=%v, got=%v", i+1, test.expectedErr, err)
-		}
 	}
 }
 

cmd/object-handlers.go

@@ -827,7 +827,7 @@ func (api objectAPIHandlers) CopyObjectHandler(w http.ResponseWriter, r *http.Re
 		reader = gr
 	}
 
-	srcInfo.Reader, err = hash.NewReader(reader, length, "", "", actualSize)
+	srcInfo.Reader, err = hash.NewReader(reader, length, "", "", actualSize, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
@@ -934,7 +934,8 @@ func (api objectAPIHandlers) CopyObjectHandler(w http.ResponseWriter, r *http.Re
 				crypto.RemoveInternalEntries(srcInfo.UserDefined)
 			}
 
-			srcInfo.Reader, err = hash.NewReader(reader, targetSize, "", "", targetSize) // do not try to verify encrypted content
+			// do not try to verify encrypted content
+			srcInfo.Reader, err = hash.NewReader(reader, targetSize, "", "", targetSize, globalCLIContext.StrictS3Compat)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 				return
@@ -1176,6 +1177,11 @@ func (api objectAPIHandlers) PutObjectHandler(w http.ResponseWriter, r *http.Req
 		}
 	}
 
+	// This request header needs to be set prior to setting ObjectOptions
+	if globalAutoEncryption && !crypto.SSEC.IsRequested(r.Header) {
+		r.Header.Add(crypto.SSEHeader, crypto.SSEAlgorithmAES256)
+	}
+
 	actualSize := size
 
 	if objectAPI.IsCompressionSupported() && isCompressible(r.Header, object) && size > 0 {
@@ -1183,7 +1189,7 @@ func (api objectAPIHandlers) PutObjectHandler(w http.ResponseWriter, r *http.Req
 		metadata[ReservedMetadataPrefix+"compression"] = compressionAlgorithmV1
 		metadata[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(size, 10)
 
-		actualReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize)
+		actualReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize, globalCLIContext.StrictS3Compat)
 		if err != nil {
 			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 			return
@@ -1196,7 +1202,7 @@ func (api objectAPIHandlers) PutObjectHandler(w http.ResponseWriter, r *http.Req
 		sha256hex = ""
 	}
 
-	hashReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize)
+	hashReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
@@ -1205,10 +1211,6 @@ func (api objectAPIHandlers) PutObjectHandler(w http.ResponseWriter, r *http.Req
 	rawReader := hashReader
 	pReader := NewPutObjReader(rawReader, nil, nil)
 
-	// This request header needs to be set prior to setting ObjectOptions
-	if globalAutoEncryption && !crypto.SSEC.IsRequested(r.Header) {
-		r.Header.Add(crypto.SSEHeader, crypto.SSEAlgorithmAES256)
-	}
 	// get gateway encryption options
 	var opts ObjectOptions
 	opts, err = putOpts(ctx, r, bucket, object, metadata)
@@ -1234,7 +1236,8 @@ func (api objectAPIHandlers) PutObjectHandler(w http.ResponseWriter, r *http.Req
 				return
 			}
 			info := ObjectInfo{Size: size}
-			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", size) // do not try to verify encrypted content
+			// do not try to verify encrypted content
+			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", size, globalCLIContext.StrictS3Compat)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 				return
@@ -1259,8 +1262,9 @@ func (api objectAPIHandlers) PutObjectHandler(w http.ResponseWriter, r *http.Req
 
 	etag := objInfo.ETag
 	if objInfo.IsCompressed() {
-		// Ignore compressed ETag.
-		etag = objInfo.ETag + "-1"
+		if !strings.HasSuffix(objInfo.ETag, "-1") {
+			etag = objInfo.ETag + "-1"
+		}
 	} else if hasServerSideEncryptionHeader(r.Header) {
 		etag = getDecryptedETag(r.Header, objInfo, false)
 	}
@@ -1612,7 +1616,7 @@ func (api objectAPIHandlers) CopyObjectPartHandler(w http.ResponseWriter, r *htt
 		reader = gr
 	}
 
-	srcInfo.Reader, err = hash.NewReader(reader, length, "", "", actualPartSize)
+	srcInfo.Reader, err = hash.NewReader(reader, length, "", "", actualPartSize, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
@@ -1672,7 +1676,7 @@ func (api objectAPIHandlers) CopyObjectPartHandler(w http.ResponseWriter, r *htt
 			}
 
 			info := ObjectInfo{Size: length}
-			srcInfo.Reader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", length)
+			srcInfo.Reader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", length, globalCLIContext.StrictS3Compat)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 				return
@@ -1839,7 +1843,7 @@ func (api objectAPIHandlers) PutObjectPartHandler(w http.ResponseWriter, r *http
 
 	isCompressed := false
 	if objectAPI.IsCompressionSupported() && compressPart {
-		actualReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize)
+		actualReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize, globalCLIContext.StrictS3Compat)
 		if err != nil {
 			writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 			return
@@ -1853,7 +1857,7 @@ func (api objectAPIHandlers) PutObjectPartHandler(w http.ResponseWriter, r *http
 		isCompressed = true
 	}
 
-	hashReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize)
+	hashReader, err := hash.NewReader(reader, size, md5hex, sha256hex, actualSize, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 		return
@@ -1920,7 +1924,8 @@ func (api objectAPIHandlers) PutObjectPartHandler(w http.ResponseWriter, r *http
 				return
 			}
 			info := ObjectInfo{Size: size}
-			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", size) // do not try to verify encrypted content
+			// do not try to verify encrypted content
+			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", size, globalCLIContext.StrictS3Compat)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 				return
@@ -1941,10 +1946,7 @@ func (api objectAPIHandlers) PutObjectPartHandler(w http.ResponseWriter, r *http
 	}
 
 	etag := partInfo.ETag
-	if isCompressed {
-		// Suppress compressed ETag.
-		etag = partInfo.ETag + "-1"
-	} else if isEncrypted {
+	if isEncrypted {
 		etag = tryDecryptETag(objectEncryptionKey, partInfo.ETag, crypto.SSEC.IsRequested(r.Header))
 	}
 	w.Header()["ETag"] = []string{"\"" + etag + "\""}
@@ -2234,11 +2236,6 @@ func (api objectAPIHandlers) CompleteMultipartUploadHandler(w http.ResponseWrite
 	// Complete parts.
 	var completeParts []CompletePart
 	for _, part := range complMultipartUpload.Parts {
-		// Avoiding for gateway parts.
-		// `strings.TrimPrefix` does not work here as intended. So `Replace` is used instead.
-		if objectAPI.IsCompressionSupported() {
-			part.ETag = strings.Replace(part.ETag, "-1", "", -1) // For compressed multiparts, We append '-1' for part.ETag.
-		}
 		part.ETag = canonicalizeETag(part.ETag)
 		if isEncrypted {
 			// ETag is stored in the backend in encrypted form. Validate client sent ETag with

cmd/object-handlers_test.go

@@ -1916,7 +1916,7 @@ func testAPICopyObjectHandler(obj ObjectLayer, instanceType, bucketName string,
 	for i, input := range putObjectInputs {
 		// uploading the object.
 		var objInfo ObjectInfo
-		objInfo, err = obj.PutObject(context.Background(), input.bucketName, input.objectName, mustGetPutObjReader(t, bytes.NewBuffer(input.textData), input.contentLength, input.metaData[""], ""), ObjectOptions{UserDefined: input.metaData})
+		objInfo, err = obj.PutObject(context.Background(), input.bucketName, input.objectName, mustGetPutObjReader(t, bytes.NewBuffer(input.textData), input.contentLength, input.md5sum, ""), ObjectOptions{UserDefined: input.metaData})
 		// if object upload fails stop the test.
 		if err != nil {
 			t.Fatalf("Put Object case %d:  Error uploading object: <ERROR> %v", i+1, err)
@@ -2233,11 +2233,6 @@ func testAPICopyObjectHandler(obj ObjectLayer, instanceType, bucketName string,
 				t.Fatalf("Test %d: %s: Failed to parse the CopyObjectResult response: <ERROR> %s", i+1, instanceType, err)
 			}
 
-			retag := canonicalizeETag(cpObjResp.ETag)
-			if retag != bytesData[0].md5sum {
-				t.Fatalf("Test %d: %s: Failed to CopyObject: <ERROR> got %s, expected %s", i+1, instanceType, retag, bytesData[0].md5sum)
-			}
-
 			// See if the new object is formed.
 			// testing whether the copy was successful.
 			err = obj.GetObject(context.Background(), testCase.bucketName, testCase.newObjectName, 0, int64(len(bytesData[0].byteData)), buffers[0], "", opts)
@@ -2660,10 +2655,7 @@ func testAPICompleteMultipartHandler(obj ObjectLayer, instanceType, bucketName s
 	}
 
 	// on successful complete multipart operation the s3MD5 for the parts uploaded will be returned.
-	s3MD5, err := getCompleteMultipartMD5(context.Background(), inputParts[3].parts)
-	if err != nil {
-		t.Fatalf("Obtaining S3MD5 failed")
-	}
+	s3MD5 := getCompleteMultipartMD5(inputParts[3].parts)
 
 	// generating the response body content for the success case.
 	successResponse := generateCompleteMultpartUploadResponse(bucketName, objectName, getGetObjectURL("", bucketName, objectName), s3MD5)

cmd/posix-errors.go

@@ -136,3 +136,12 @@ func isSysErrCrossDevice(err error) bool {
 	e, ok := err.(*os.LinkError)
 	return ok && e.Err == syscall.EXDEV
 }
+
+// Check if given error corresponds to too many open files
+func isSysErrTooManyFiles(err error) bool {
+	if err == syscall.ENFILE || err == syscall.EMFILE {
+		return true
+	}
+	pathErr, ok := err.(*os.PathError)
+	return ok && (pathErr.Err == syscall.ENFILE || pathErr.Err == syscall.EMFILE)
+}

cmd/posix-list-dir_unix.go

@@ -20,7 +20,6 @@ package cmd
 
 import (
 	"os"
-	"path"
 	"runtime"
 	"sync"
 	"syscall"
@@ -79,7 +78,7 @@ func parseDirents(dirPath string, buf []byte) (entries []string, err error) {
 			// On Linux XFS does not implement d_type for on disk
 			// format << v5. Fall back to OsStat().
 			var fi os.FileInfo
-			fi, err = os.Stat(path.Join(dirPath, name))
+			fi, err = os.Stat(pathJoin(dirPath, name))
 			if err != nil {
 				// If file does not exist, we continue and skip it.
 				// Could happen if it was deleted in the middle while

cmd/posix-list-dir_windows.go

@@ -20,7 +20,6 @@ package cmd
 
 import (
 	"os"
-	"path"
 	"strings"
 	"syscall"
 )
@@ -82,7 +81,7 @@ func readDirN(dirPath string, count int) (entries []string, err error) {
 		case data.FileAttributes&syscall.FILE_ATTRIBUTE_REPARSE_POINT != 0:
 			// If its symbolic link, follow the link using os.Stat()
 			var fi os.FileInfo
-			fi, err = os.Stat(path.Join(dirPath, name))
+			fi, err = os.Stat(pathJoin(dirPath, name))
 			if err != nil {
 				// If file does not exist, we continue and skip it.
 				// Could happen if it was deleted in the middle while

cmd/posix.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017, 2018 MinIO, Inc.
+ * 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.
@@ -25,6 +25,7 @@ import (
 	slashpath "path"
 	"path/filepath"
 	"runtime"
+	"sort"
 	"strings"
 	"sync"
 	"sync/atomic"
@@ -37,12 +38,17 @@ import (
 	"github.com/minio/minio/cmd/logger"
 	"github.com/minio/minio/pkg/disk"
 	"github.com/minio/minio/pkg/mountinfo"
+	"github.com/ncw/directio"
 )
 
 const (
-	diskMinFreeSpace  = 900 * humanize.MiByte // Min 900MiB free space.
-	diskMinTotalSpace = diskMinFreeSpace      // Min 900MiB total space.
-	maxAllowedIOError = 5
+	diskMinFreeSpace    = 900 * humanize.MiByte // Min 900MiB free space.
+	diskMinTotalSpace   = diskMinFreeSpace      // Min 900MiB total space.
+	maxAllowedIOError   = 5
+	posixWriteBlockSize = 4 * humanize.MiByte
+	// DirectIO alignment needs to be 4K. Defined here as
+	// directio.AlignSize is defined as 0 in MacOS causing divide by 0 error.
+	directioAlignSize = 4096
 )
 
 // isValidVolname verifies a volname name in accordance with object
@@ -71,7 +77,6 @@ type posix struct {
 	connected bool
 
 	diskMount bool // indicates if the path is an actual mount.
-	driveSync bool // indicates if the backend is synchronous.
 
 	diskFileInfo os.FileInfo
 	// Disk usage metrics
@@ -185,10 +190,10 @@ func newPosix(path string) (*posix, error) {
 	p := &posix{
 		connected: true,
 		diskPath:  path,
-		// 1MiB buffer pool for posix internal operations.
+		// 4MiB buffer pool for posix internal operations.
 		pool: sync.Pool{
 			New: func() interface{} {
-				b := make([]byte, readSizeV1)
+				b := directio.AlignedBlock(posixWriteBlockSize)
 				return &b
 			},
 		},
@@ -197,15 +202,6 @@ func newPosix(path string) (*posix, error) {
 		diskMount:    mountinfo.IsLikelyMountPoint(path),
 	}
 
-	var pf BoolFlag
-	if driveSync := os.Getenv("MINIO_DRIVE_SYNC"); driveSync != "" {
-		pf, err = ParseBoolFlag(driveSync)
-		if err != nil {
-			return nil, err
-		}
-		p.driveSync = bool(pf)
-	}
-
 	if !p.diskMount {
 		go p.diskUsage(GlobalServiceDoneCh)
 	}
@@ -649,9 +645,88 @@ func (s *posix) DeleteVol(volume string) (err error) {
 	return nil
 }
 
+// Walk - is a sorted walker which returns file entries in lexically
+// sorted order, additionally along with metadata about each of those entries.
+func (s *posix) Walk(volume, dirPath, marker string, recursive bool, leafFile string,
+	readMetadataFn readMetadataFunc, endWalkCh chan struct{}) (ch chan FileInfo, err error) {
+
+	defer func() {
+		if err == errFaultyDisk {
+			atomic.AddInt32(&s.ioErrCount, 1)
+		}
+	}()
+
+	if atomic.LoadInt32(&s.ioErrCount) > maxAllowedIOError {
+		return nil, errFaultyDisk
+	}
+
+	if err = s.checkDiskFound(); err != nil {
+		return nil, err
+	}
+
+	// Verify if volume is valid and it exists.
+	volumeDir, err := s.getVolDir(volume)
+	if err != nil {
+		return nil, err
+	}
+
+	// Stat a volume entry.
+	_, err = os.Stat(volumeDir)
+	if err != nil {
+		if os.IsNotExist(err) {
+			return nil, errVolumeNotFound
+		} else if isSysErrIO(err) {
+			return nil, errFaultyDisk
+		}
+		return nil, err
+	}
+
+	ch = make(chan FileInfo)
+	go func() {
+		defer close(ch)
+		listDir := func(volume, dirPath, dirEntry string) (entries []string) {
+			entries, err := s.ListDir(volume, dirPath, -1, leafFile)
+			if err != nil {
+				return
+			}
+			sort.Strings(entries)
+			return filterMatchingPrefix(entries, dirEntry)
+		}
+
+		walkResultCh := startTreeWalk(context.Background(), volume, dirPath, marker, recursive, listDir, endWalkCh)
+		for {
+			walkResult, ok := <-walkResultCh
+			if !ok {
+				return
+			}
+			var fi FileInfo
+			if hasSuffix(walkResult.entry, slashSeparator) {
+				fi = FileInfo{
+					Volume: volume,
+					Name:   walkResult.entry,
+					Mode:   os.ModeDir,
+				}
+			} else {
+				buf, err := s.ReadAll(volume, pathJoin(walkResult.entry, leafFile))
+				if err != nil {
+					continue
+				}
+				fi = readMetadataFn(buf, volume, walkResult.entry)
+			}
+			select {
+			case ch <- fi:
+			case <-endWalkCh:
+				return
+			}
+		}
+	}()
+
+	return ch, nil
+}
+
 // ListDir - return all the entries at the given directory path.
 // If an entry is a directory it will be returned with a trailing "/".
-func (s *posix) ListDir(volume, dirPath string, count int) (entries []string, err error) {
+func (s *posix) ListDir(volume, dirPath string, count int, leafFile string) (entries []string, err error) {
 	defer func() {
 		if err == errFaultyDisk {
 			atomic.AddInt32(&s.ioErrCount, 1)
@@ -684,9 +759,21 @@ func (s *posix) ListDir(volume, dirPath string, count int) (entries []string, er
 
 	dirPath = pathJoin(volumeDir, dirPath)
 	if count > 0 {
-		return readDirN(dirPath, count)
+		entries, err = readDirN(dirPath, count)
+	} else {
+		entries, err = readDir(dirPath)
 	}
-	return readDir(dirPath)
+
+	// If leaf file is specified, filter out the entries.
+	if leafFile != "" {
+		for i, entry := range entries {
+			if _, serr := os.Stat(pathJoin(dirPath, entry, leafFile)); serr == nil {
+				entries[i] = strings.TrimSuffix(entry, slashSeparator)
+			}
+		}
+	}
+
+	return entries, err
 }
 
 // ReadAll reads from r until an error or EOF and returns the data it read.
@@ -721,6 +808,8 @@ func (s *posix) ReadAll(volume, path string) (buf []byte, err error) {
 			return nil, errVolumeNotFound
 		} else if isSysErrIO(err) {
 			return nil, errFaultyDisk
+		} else if isSysErrTooManyFiles(err) {
+			return nil, errTooManyOpenFiles
 		}
 		return nil, err
 	}
@@ -824,6 +913,8 @@ func (s *posix) ReadFile(volume, path string, offset int64, buffer []byte, verif
 			return 0, errFileAccessDenied
 		case isSysErrIO(err):
 			return 0, errFaultyDisk
+		case isSysErrTooManyFiles(err):
+			return 0, errTooManyOpenFiles
 		default:
 			return 0, err
 		}
@@ -934,6 +1025,8 @@ func (s *posix) openFile(volume, path string, mode int) (f *os.File, err error)
 			return nil, errFileAccessDenied
 		case isSysErrIO(err):
 			return nil, errFaultyDisk
+		case isSysErrTooManyFiles(err):
+			return nil, errTooManyOpenFiles
 		default:
 			return nil, err
 		}
@@ -996,6 +1089,8 @@ func (s *posix) ReadFileStream(volume, path string, offset, length int64) (io.Re
 			return nil, errFileAccessDenied
 		case isSysErrIO(err):
 			return nil, errFaultyDisk
+		case isSysErrTooManyFiles(err):
+			return nil, errTooManyOpenFiles
 		default:
 			return nil, err
 		}
@@ -1023,10 +1118,9 @@ func (s *posix) ReadFileStream(volume, path string, offset, length int64) (io.Re
 
 // CreateFile - creates the file.
 func (s *posix) CreateFile(volume, path string, fileSize int64, r io.Reader) (err error) {
-	if fileSize < 0 {
+	if fileSize < -1 {
 		return errInvalidArgument
 	}
-
 	defer func() {
 		if err == errFaultyDisk {
 			atomic.AddInt32(&s.ioErrCount, 1)
@@ -1045,13 +1139,50 @@ func (s *posix) CreateFile(volume, path string, fileSize int64, r io.Reader) (er
 		return err
 	}
 
-	// Create file if not found. Note that it is created with os.O_EXCL flag as the file
-	// always is supposed to be created in the tmp directory with a unique file name.
-	w, err := s.openFile(volume, path, os.O_CREATE|os.O_APPEND|os.O_WRONLY|os.O_EXCL)
-	if err != nil {
+	if err = s.checkDiskFound(); err != nil {
 		return err
 	}
 
+	volumeDir, err := s.getVolDir(volume)
+	if err != nil {
+		return err
+	}
+	// Stat a volume entry.
+	_, err = os.Stat((volumeDir))
+	if err != nil {
+		if os.IsNotExist(err) {
+			return errVolumeNotFound
+		} else if isSysErrIO(err) {
+			return errFaultyDisk
+		}
+		return err
+	}
+
+	filePath := pathJoin(volumeDir, path)
+	if err = checkPathLength((filePath)); err != nil {
+		return err
+	}
+
+	// Create top level directories if they don't exist.
+	// with mode 0777 mkdir honors system umask.
+	if err = mkdirAll(slashpath.Dir(filePath), 0777); err != nil {
+		return err
+	}
+
+	w, err := disk.OpenFileDirectIO(filePath, os.O_CREATE|os.O_WRONLY|os.O_EXCL, 0666)
+	if err != nil {
+		switch {
+		case os.IsPermission(err):
+			return errFileAccessDenied
+		case os.IsExist(err):
+			return errFileAccessDenied
+		case isSysErrIO(err):
+			return errFaultyDisk
+		default:
+			return err
+		}
+	}
+	defer w.Sync() // Sync before close.
 	defer w.Close()
 
 	var e error
@@ -1078,17 +1209,70 @@ func (s *posix) CreateFile(volume, path string, fileSize int64, r io.Reader) (er
 	bufp := s.pool.Get().(*[]byte)
 	defer s.pool.Put(bufp)
 
-	n, err := io.CopyBuffer(w, r, *bufp)
-	if err != nil {
-		return err
+	buf := *bufp
+
+	// Writes remaining bytes in the buffer.
+	writeRemaining := func(w *os.File, buf []byte) (remainingWritten int, err error) {
+		var n int
+		remaining := len(buf)
+		// The following logic writes the remainging data such that it writes whatever best is possible (aligned buffer)
+		// in O_DIRECT mode and remaining (unaligned buffer) in non-O_DIRECT mode.
+		remainingAligned := (remaining / directioAlignSize) * directioAlignSize
+		remainingAlignedBuf := buf[:remainingAligned]
+		remainingUnalignedBuf := buf[remainingAligned:]
+		if len(remainingAlignedBuf) > 0 {
+			n, err = w.Write(remainingAlignedBuf)
+			if err != nil {
+				return 0, err
+			}
+			remainingWritten += n
+		}
+		if len(remainingUnalignedBuf) > 0 {
+			// Write on O_DIRECT fds fail if buffer is not 4K aligned, hence disable O_DIRECT.
+			if err = disk.DisableDirectIO(w); err != nil {
+				return 0, err
+			}
+			n, err = w.Write(remainingUnalignedBuf)
+			if err != nil {
+				return 0, err
+			}
+			remainingWritten += n
+		}
+		return remainingWritten, nil
 	}
-	if n < fileSize {
-		return errLessData
+
+	var written int
+	for {
+		var n int
+		n, err = io.ReadFull(r, buf)
+		switch err {
+		case nil:
+			n, err = w.Write(buf)
+			if err != nil {
+				return err
+			}
+			written += n
+		case io.ErrUnexpectedEOF:
+			n, err = writeRemaining(w, buf[:n])
+			if err != nil {
+				return err
+			}
+			written += n
+			fallthrough
+		case io.EOF:
+			if fileSize != -1 {
+				if written < int(fileSize) {
+					return errLessData
+				}
+				if written > int(fileSize) {
+					return errMoreData
+				}
+			}
+			return nil
+		default:
+			return err
+		}
 	}
-	if n > fileSize {
-		return errMoreData
-	}
-	return nil
 }
 
 func (s *posix) WriteAll(volume, path string, buf []byte) (err error) {
@@ -1130,13 +1314,9 @@ func (s *posix) AppendFile(volume, path string, buf []byte) (err error) {
 	}
 
 	var w *os.File
-	// Create file if not found, additionally also enables synchronous
-	// operation if asked by the user.
-	if s.driveSync {
-		w, err = s.openFile(volume, path, os.O_CREATE|os.O_SYNC|os.O_APPEND|os.O_WRONLY)
-	} else {
-		w, err = s.openFile(volume, path, os.O_CREATE|os.O_APPEND|os.O_WRONLY)
-	}
+	// Create file if not found. Not doing O_DIRECT here to avoid the code that does buffer aligned writes.
+	// AppendFile() is only used by healing code to heal objects written in old format.
+	w, err = s.openFile(volume, path, os.O_CREATE|os.O_SYNC|os.O_APPEND|os.O_WRONLY)
 	if err != nil {
 		return err
 	}
@@ -1290,6 +1470,14 @@ func (s *posix) DeleteFile(volume, path string) (err error) {
 	return deleteFile(volumeDir, filePath)
 }
 
+func (s *posix) DeleteFileBulk(volume string, paths []string) (errs []error, err error) {
+	errs = make([]error, len(paths))
+	for idx, path := range paths {
+		errs[idx] = s.DeleteFile(volume, path)
+	}
+	return
+}
+
 // RenameFile - rename source path to destination path atomically.
 func (s *posix) RenameFile(srcVolume, srcPath, dstVolume, dstPath string) (err error) {
 	defer func() {

cmd/posix_test.go

@@ -839,7 +839,7 @@ func TestPosixPosixListDir(t *testing.T) {
 		} else {
 			t.Errorf("Expected the StorageAPI to be of type *posix")
 		}
-		dirList, err = posixStorage.ListDir(testCase.srcVol, testCase.srcPath, -1)
+		dirList, err = posixStorage.ListDir(testCase.srcVol, testCase.srcPath, -1, "")
 		if err != testCase.expectedErr {
 			t.Fatalf("TestPosix case %d: Expected: \"%s\", got: \"%s\"", i+1, testCase.expectedErr, err)
 		}

cmd/rest/client.go

@@ -89,12 +89,7 @@ func newCustomDialContext(timeout time.Duration) func(ctx context.Context, netwo
 			DualStack: true,
 		}
 
-		conn, err := dialer.DialContext(ctx, network, addr)
-		if err != nil {
-			return nil, err
-		}
-
-		return xhttp.NewTimeoutConn(conn, timeout, timeout), nil
+		return dialer.DialContext(ctx, network, addr)
 	}
 }
 

cmd/server-main.go

@@ -167,6 +167,8 @@ func serverHandleCmdArgs(ctx *cli.Context) {
 	}
 	logger.FatalIf(err, "Invalid command line arguments")
 
+	logger.LogIf(context.Background(), checkEndpointsSubOptimal(ctx, setupType, globalEndpoints))
+
 	globalMinioHost, globalMinioPort = mustSplitHostPort(globalMinioAddr)
 
 	// On macOS, if a process already listens on LOCALIPADDR:PORT, net.Listen() falls back
@@ -200,6 +202,8 @@ func serverMain(ctx *cli.Context) {
 		cli.ShowCommandHelpAndExit(ctx, "server", 1)
 	}
 
+	signal.Notify(globalOSSignalCh, os.Interrupt, syscall.SIGTERM)
+
 	// Disable logging until server initialization is complete, any
 	// error during initialization will be shown as a fatal message
 	logger.Disable = true
@@ -305,8 +309,6 @@ func serverMain(ctx *cli.Context) {
 		globalHTTPServerErrorCh <- globalHTTPServer.Start()
 	}()
 
-	signal.Notify(globalOSSignalCh, os.Interrupt, syscall.SIGTERM)
-
 	newObject, err := newObjectLayer(globalEndpoints)
 	if err != nil {
 		// Stop watching for any certificate changes.

cmd/server_test.go

@@ -18,7 +18,6 @@ package cmd
 
 import (
 	"bytes"
-	"context"
 	"crypto/tls"
 	"crypto/x509"
 	"encoding/xml"
@@ -2725,12 +2724,9 @@ func (s *TestSuiteCommon) TestObjectMultipart(c *check) {
 	c.Assert(response.StatusCode, http.StatusOK)
 	var parts []CompletePart
 	for _, part := range completeUploads.Parts {
-		// For compressed objects, we dont treat E-Tag as checksum.
-		part.ETag = strings.Replace(part.ETag, "-1", "", -1)
 		part.ETag = canonicalizeETag(part.ETag)
 		parts = append(parts, part)
 	}
-	etag, err := getCompleteMultipartMD5(context.Background(), parts)
-	c.Assert(err, nil)
+	etag := getCompleteMultipartMD5(parts)
 	c.Assert(canonicalizeETag(response.Header.Get("Etag")), etag)
 }

cmd/storage-datatypes.go

@@ -30,6 +30,13 @@ type VolInfo struct {
 	Created time.Time
 }
 
+// FilesInfo represent a list of files, additionally
+// indicates if the list is last.
+type FilesInfo struct {
+	Files       []FileInfo
+	IsTruncated bool
+}
+
 // FileInfo - represents file stat information.
 type FileInfo struct {
 	// Name of the volume.
@@ -46,4 +53,12 @@ type FileInfo struct {
 
 	// File mode bits.
 	Mode os.FileMode
+
+	// File metadata
+	Metadata map[string]string
+
+	// All the parts per object.
+	Parts []ObjectPartInfo
+
+	Quorum int
 }

cmd/storage-errors.go

@@ -48,6 +48,9 @@ var errDiskAccessDenied = errors.New("disk access denied")
 // errFileNotFound - cannot find the file.
 var errFileNotFound = errors.New("file not found")
 
+// errTooManyOpenFiles - too many open files.
+var errTooManyOpenFiles = errors.New("too many open files")
+
 // errFileNameTooLong - given file name is too long than supported length.
 var errFileNameTooLong = errors.New("file name too long")
 

cmd/storage-interface.go

@@ -38,8 +38,12 @@ type StorageAPI interface {
 	StatVol(volume string) (vol VolInfo, err error)
 	DeleteVol(volume string) (err error)
 
+	// Walk in sorted order directly on disk.
+	Walk(volume, dirPath string, marker string, recursive bool, leafFile string,
+		readMetadataFn readMetadataFunc, endWalkCh chan struct{}) (chan FileInfo, error)
+
 	// File operations.
-	ListDir(volume, dirPath string, count int) ([]string, error)
+	ListDir(volume, dirPath string, count int, leafFile string) ([]string, error)
 	ReadFile(volume string, path string, offset int64, buf []byte, verifier *BitrotVerifier) (n int64, err error)
 	AppendFile(volume string, path string, buf []byte) (err error)
 	CreateFile(volume, path string, size int64, reader io.Reader) error
@@ -47,6 +51,7 @@ type StorageAPI interface {
 	RenameFile(srcVolume, srcPath, dstVolume, dstPath string) error
 	StatFile(volume string, path string) (file FileInfo, err error)
 	DeleteFile(volume string, path string) (err error)
+	DeleteFileBulk(volume string, paths []string) (errs []error, err error)
 
 	// Write all data, syncs the data to disk.
 	WriteAll(volume string, path string, buf []byte) (err error)

cmd/storage-rest-client.go

@@ -44,6 +44,9 @@ func isNetworkError(err error) bool {
 	if err.Error() == errConnectionStale.Error() {
 		return true
 	}
+	if strings.Contains(err.Error(), "connection reset by peer") {
+		return true
+	}
 	if uerr, isURLError := err.(*url.Error); isURLError {
 		if uerr.Timeout() {
 			return true
@@ -56,6 +59,19 @@ func isNetworkError(err error) bool {
 	return isNetOpError
 }
 
+// Attempt to approximate network error with a
+// typed network error, otherwise default to
+// errDiskNotFound
+func toNetworkError(err error) error {
+	if err == nil {
+		return err
+	}
+	if strings.Contains(err.Error(), "connection reset by peer") {
+		return errNetworkConnReset
+	}
+	return errDiskNotFound
+}
+
 // Converts rpc.ServerError to underlying error. This function is
 // written so that the storageAPI errors are consistent across network
 // disks as well.
@@ -65,7 +81,7 @@ func toStorageErr(err error) error {
 	}
 
 	if isNetworkError(err) {
-		return errDiskNotFound
+		return toNetworkError(err)
 	}
 
 	switch err.Error() {
@@ -234,7 +250,7 @@ func (client *storageRESTClient) CreateFile(volume, path string, length int64, r
 	values.Set(storageRESTVolume, volume)
 	values.Set(storageRESTFilePath, path)
 	values.Set(storageRESTLength, strconv.Itoa(int(length)))
-	respBody, err := client.call(storageRESTMethodCreateFile, values, r, length)
+	respBody, err := client.call(storageRESTMethodCreateFile, values, ioutil.NopCloser(r), length)
 	defer http.DrainBody(respBody)
 	return err
 }
@@ -314,12 +330,50 @@ func (client *storageRESTClient) ReadFile(volume, path string, offset int64, buf
 	return int64(n), err
 }
 
+func (client *storageRESTClient) Walk(volume, dirPath, marker string, recursive bool, leafFile string,
+	readMetadataFn readMetadataFunc, endWalkCh chan struct{}) (chan FileInfo, error) {
+	values := make(url.Values)
+	values.Set(storageRESTVolume, volume)
+	values.Set(storageRESTDirPath, dirPath)
+	values.Set(storageRESTMarkerPath, marker)
+	values.Set(storageRESTRecursive, strconv.FormatBool(recursive))
+	values.Set(storageRESTLeafFile, leafFile)
+	respBody, err := client.call(storageRESTMethodWalk, values, nil, -1)
+	if err != nil {
+		return nil, err
+	}
+
+	ch := make(chan FileInfo)
+	go func() {
+		defer close(ch)
+		defer http.DrainBody(respBody)
+
+		decoder := gob.NewDecoder(respBody)
+		for {
+			var fi FileInfo
+			if gerr := decoder.Decode(&fi); gerr != nil {
+				// Upon error return
+				return
+			}
+			select {
+			case ch <- fi:
+			case <-endWalkCh:
+				return
+			}
+
+		}
+	}()
+
+	return ch, nil
+}
+
 // ListDir - lists a directory.
-func (client *storageRESTClient) ListDir(volume, dirPath string, count int) (entries []string, err error) {
+func (client *storageRESTClient) ListDir(volume, dirPath string, count int, leafFile string) (entries []string, err error) {
 	values := make(url.Values)
 	values.Set(storageRESTVolume, volume)
 	values.Set(storageRESTDirPath, dirPath)
 	values.Set(storageRESTCount, strconv.Itoa(count))
+	values.Set(storageRESTLeafFile, leafFile)
 	respBody, err := client.call(storageRESTMethodListDir, values, nil, -1)
 	if err != nil {
 		return nil, err
@@ -339,6 +393,34 @@ func (client *storageRESTClient) DeleteFile(volume, path string) error {
 	return err
 }
 
+// DeleteFileBulk - deletes files in bulk.
+func (client *storageRESTClient) DeleteFileBulk(volume string, paths []string) (errs []error, err error) {
+	errs = make([]error, len(paths))
+	values := make(url.Values)
+	values.Set(storageRESTVolume, volume)
+	for _, path := range paths {
+		values.Add(storageRESTFilePath, path)
+	}
+	respBody, err := client.call(storageRESTMethodDeleteFileBulk, values, nil, -1)
+	defer http.DrainBody(respBody)
+
+	if err != nil {
+		return nil, err
+	}
+
+	bulkErrs := bulkErrorsResponse{}
+	gob.NewDecoder(respBody).Decode(&bulkErrs)
+	if err != nil {
+		return nil, err
+	}
+
+	for i, dErr := range bulkErrs.Errs {
+		errs[i] = toStorageErr(dErr)
+	}
+
+	return errs, nil
+}
+
 // RenameFile - renames a file.
 func (client *storageRESTClient) RenameFile(srcVolume, srcPath, dstVolume, dstPath string) (err error) {
 	values := make(url.Values)

cmd/storage-rest-common.go

@@ -16,7 +16,7 @@
 
 package cmd
 
-const storageRESTVersion = "v4"
+const storageRESTVersion = "v6"
 const storageRESTPath = minioReservedBucketPath + "/storage/" + storageRESTVersion + "/"
 
 const (
@@ -34,7 +34,9 @@ const (
 	storageRESTMethodReadFile       = "readfile"
 	storageRESTMethodReadFileStream = "readfilestream"
 	storageRESTMethodListDir        = "listdir"
+	storageRESTMethodWalk           = "walk"
 	storageRESTMethodDeleteFile     = "deletefile"
+	storageRESTMethodDeleteFileBulk = "deletefilebulk"
 	storageRESTMethodRenameFile     = "renamefile"
 	storageRESTMethodGetInstanceID  = "getinstanceid"
 )
@@ -50,6 +52,9 @@ const (
 	storageRESTOffset     = "offset"
 	storageRESTLength     = "length"
 	storageRESTCount      = "count"
+	storageRESTMarkerPath = "marker"
+	storageRESTLeafFile   = "leaf-file"
+	storageRESTRecursive  = "recursive"
 	storageRESTBitrotAlgo = "bitrot-algo"
 	storageRESTBitrotHash = "bitrot-hash"
 	storageRESTInstanceID = "instance-id"

cmd/storage-rest-server.go

@@ -17,6 +17,7 @@
 package cmd
 
 import (
+	"context"
 	"encoding/gob"
 	"encoding/hex"
 	"errors"
@@ -47,6 +48,22 @@ func (s *storageRESTServer) writeErrorResponse(w http.ResponseWriter, err error)
 	w.(http.Flusher).Flush()
 }
 
+type bulkErrorsResponse struct {
+	Errs []error `json:"errors"`
+}
+
+func (s *storageRESTServer) writeErrorsResponse(w http.ResponseWriter, errs []error) {
+	resp := bulkErrorsResponse{Errs: make([]error, len(errs))}
+	for idx, err := range errs {
+		if err == nil {
+			continue
+		}
+		resp.Errs[idx] = err
+	}
+	gob.NewEncoder(w).Encode(resp)
+	w.(http.Flusher).Flush()
+}
+
 // DefaultSkewTime - skew time is 15 minutes between minio peers.
 const DefaultSkewTime = 15 * time.Minute
 
@@ -353,6 +370,58 @@ func (s *storageRESTServer) ReadFileStreamHandler(w http.ResponseWriter, r *http
 	w.(http.Flusher).Flush()
 }
 
+// readMetadata func provides the function types for reading leaf metadata.
+type readMetadataFunc func(buf []byte, volume, entry string) FileInfo
+
+func readMetadata(buf []byte, volume, entry string) FileInfo {
+	m, err := xlMetaV1UnmarshalJSON(context.Background(), buf)
+	if err != nil {
+		return FileInfo{}
+	}
+	return FileInfo{
+		Volume:   volume,
+		Name:     entry,
+		ModTime:  m.Stat.ModTime,
+		Size:     m.Stat.Size,
+		Metadata: m.Meta,
+		Parts:    m.Parts,
+		Quorum:   m.Erasure.DataBlocks,
+	}
+}
+
+// WalkHandler - remote caller to start walking at a requested directory path.
+func (s *storageRESTServer) WalkHandler(w http.ResponseWriter, r *http.Request) {
+	if !s.IsValid(w, r) {
+		return
+	}
+	vars := mux.Vars(r)
+	volume := vars[storageRESTVolume]
+	dirPath := vars[storageRESTDirPath]
+	markerPath := vars[storageRESTMarkerPath]
+	recursive, err := strconv.ParseBool(vars[storageRESTRecursive])
+	if err != nil {
+		s.writeErrorResponse(w, err)
+		return
+	}
+	leafFile := vars[storageRESTLeafFile]
+
+	endWalkCh := make(chan struct{})
+	defer close(endWalkCh)
+
+	fch, err := s.storage.Walk(volume, dirPath, markerPath, recursive, leafFile, readMetadata, endWalkCh)
+	if err != nil {
+		s.writeErrorResponse(w, err)
+		return
+	}
+	defer w.(http.Flusher).Flush()
+
+	w.Header().Set("Connection", "close") // Pro-actively ask client to close this connection.
+	encoder := gob.NewEncoder(w)
+	for fi := range fch {
+		encoder.Encode(&fi)
+	}
+}
+
 // ListDirHandler - list a directory.
 func (s *storageRESTServer) ListDirHandler(w http.ResponseWriter, r *http.Request) {
 	if !s.IsValid(w, r) {
@@ -361,12 +430,13 @@ func (s *storageRESTServer) ListDirHandler(w http.ResponseWriter, r *http.Reques
 	vars := mux.Vars(r)
 	volume := vars[storageRESTVolume]
 	dirPath := vars[storageRESTDirPath]
+	leafFile := vars[storageRESTLeafFile]
 	count, err := strconv.Atoi(vars[storageRESTCount])
 	if err != nil {
 		s.writeErrorResponse(w, err)
 		return
 	}
-	entries, err := s.storage.ListDir(volume, dirPath, count)
+	entries, err := s.storage.ListDir(volume, dirPath, count, leafFile)
 	if err != nil {
 		s.writeErrorResponse(w, err)
 		return
@@ -390,6 +460,24 @@ func (s *storageRESTServer) DeleteFileHandler(w http.ResponseWriter, r *http.Req
 	}
 }
 
+// DeleteFileBulkHandler - delete a file.
+func (s *storageRESTServer) DeleteFileBulkHandler(w http.ResponseWriter, r *http.Request) {
+	if !s.IsValid(w, r) {
+		return
+	}
+	vars := r.URL.Query()
+	volume := vars.Get(storageRESTVolume)
+	filePaths := vars[storageRESTFilePath]
+
+	errs, err := s.storage.DeleteFileBulk(volume, filePaths)
+	if err != nil {
+		s.writeErrorResponse(w, err)
+		return
+	}
+
+	s.writeErrorsResponse(w, errs)
+}
+
 // RenameFileHandler - rename a file.
 func (s *storageRESTServer) RenameFileHandler(w http.ResponseWriter, r *http.Request) {
 	if !s.IsValid(w, r) {
@@ -443,9 +531,14 @@ func registerStorageRESTHandlers(router *mux.Router, endpoints EndpointList) {
 		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodReadFileStream).HandlerFunc(httpTraceHdrs(server.ReadFileStreamHandler)).
 			Queries(restQueries(storageRESTVolume, storageRESTFilePath, storageRESTOffset, storageRESTLength)...)
 		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodListDir).HandlerFunc(httpTraceHdrs(server.ListDirHandler)).
-			Queries(restQueries(storageRESTVolume, storageRESTDirPath, storageRESTCount)...)
+			Queries(restQueries(storageRESTVolume, storageRESTDirPath, storageRESTCount, storageRESTLeafFile)...)
+		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodWalk).HandlerFunc(httpTraceHdrs(server.WalkHandler)).
+			Queries(restQueries(storageRESTVolume, storageRESTDirPath, storageRESTMarkerPath, storageRESTRecursive, storageRESTLeafFile)...)
 		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodDeleteFile).HandlerFunc(httpTraceHdrs(server.DeleteFileHandler)).
 			Queries(restQueries(storageRESTVolume, storageRESTFilePath)...)
+		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodDeleteFileBulk).HandlerFunc(httpTraceHdrs(server.DeleteFileBulkHandler)).
+			Queries(restQueries(storageRESTVolume, storageRESTFilePath)...)
+
 		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodRenameFile).HandlerFunc(httpTraceHdrs(server.RenameFileHandler)).
 			Queries(restQueries(storageRESTSrcVolume, storageRESTSrcPath, storageRESTDstVolume, storageRESTDstPath)...)
 		subrouter.Methods(http.MethodPost).Path("/" + storageRESTMethodGetInstanceID).HandlerFunc(httpTraceAll(server.GetInstanceID))

cmd/storage-rest_test.go

@@ -260,7 +260,7 @@ func testStorageAPIListDir(t *testing.T, storage StorageAPI) {
 	}
 
 	for i, testCase := range testCases {
-		result, err := storage.ListDir(testCase.volumeName, testCase.prefix, -1)
+		result, err := storage.ListDir(testCase.volumeName, testCase.prefix, -1, "")
 		expectErr := (err != nil)
 
 		if expectErr != testCase.expectErr {
@@ -508,6 +508,10 @@ func newStorageRESTHTTPServerClient(t *testing.T) (*httptest.Server, *storageRES
 		t.Fatalf("NewEndpoint failed %v", endpoint)
 	}
 
+	if err := endpoint.UpdateIsLocal(); err != nil {
+		t.Fatalf("UpdateIsLocal failed %v", err)
+	}
+
 	registerStorageRESTHandlers(router, EndpointList{endpoint})
 	restClient, err := newStorageRESTClient(endpoint)
 	if err != nil {

cmd/sts-handlers.go

@@ -25,6 +25,7 @@ import (
 	"github.com/minio/minio/cmd/logger"
 	"github.com/minio/minio/pkg/auth"
 	"github.com/minio/minio/pkg/iam/validator"
+	"github.com/minio/minio/pkg/wildcard"
 )
 
 const (
@@ -49,13 +50,19 @@ func registerSTSRouter(router *mux.Router) {
 	stsRouter := router.NewRoute().PathPrefix("/").Subrouter()
 
 	// Assume roles with no JWT, handles AssumeRole.
-	stsRouter.Methods("POST").HeadersRegexp("Content-Type", "application/x-www-form-urlencoded*").
-		HeadersRegexp("Authorization", "AWS4-HMAC-SHA256*").
-		HandlerFunc(httpTraceAll(sts.AssumeRole))
+	stsRouter.Methods("POST").MatcherFunc(func(r *http.Request, rm *mux.RouteMatch) bool {
+		ctypeOk := wildcard.MatchSimple("application/x-www-form-urlencoded*", r.Header.Get("Content-Type"))
+		authOk := wildcard.MatchSimple("AWS4-HMAC-SHA256*", r.Header.Get("Authorization"))
+		noQueries := len(r.URL.Query()) == 0
+		return ctypeOk && authOk && noQueries
+	}).HandlerFunc(httpTraceAll(sts.AssumeRole))
 
 	// Assume roles with JWT handler, handles both ClientGrants and WebIdentity.
-	stsRouter.Methods("POST").HeadersRegexp("Content-Type", "application/x-www-form-urlencoded*").
-		HandlerFunc(httpTraceAll(sts.AssumeRoleWithJWT))
+	stsRouter.Methods("POST").MatcherFunc(func(r *http.Request, rm *mux.RouteMatch) bool {
+		ctypeOk := wildcard.MatchSimple("application/x-www-form-urlencoded*", r.Header.Get("Content-Type"))
+		noQueries := len(r.URL.Query()) == 0
+		return ctypeOk && noQueries
+	}).HandlerFunc(httpTraceAll(sts.AssumeRoleWithJWT))
 
 	// AssumeRoleWithClientGrants
 	stsRouter.Methods("POST").HandlerFunc(httpTraceAll(sts.AssumeRoleWithClientGrants)).

cmd/test-utils_test.go

@@ -138,7 +138,7 @@ func calculateSignedChunkLength(chunkDataSize int64) int64 {
 }
 
 func mustGetPutObjReader(t TestErrHandler, data io.Reader, size int64, md5hex, sha256hex string) *PutObjReader {
-	hr, err := hash.NewReader(data, size, md5hex, sha256hex, size)
+	hr, err := hash.NewReader(data, size, md5hex, sha256hex, size, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -1927,6 +1927,10 @@ func ExecObjectLayerAPITest(t *testing.T, objAPITest objAPITestType, endpoints [
 
 	globalIAMSys = NewIAMSys()
 	globalIAMSys.Init(objLayer)
+
+	globalPolicySys = NewPolicySys()
+	globalPolicySys.Init(objLayer)
+
 	// initialize the server and obtain the credentials and root.
 	// credentials are necessary to sign the HTTP request.
 	if err = newTestConfig(globalMinioDefaultRegion, objLayer); err != nil {

cmd/tree-walk-pool.go

@@ -18,6 +18,7 @@ package cmd
 
 import (
 	"errors"
+	"reflect"
 	"sync"
 	"time"
 )
@@ -127,20 +128,23 @@ func (t TreeWalkPool) Set(params listParams, resultCh chan TreeWalkResult, endWa
 			t.lock.Lock()
 			walks, ok := t.pool[params]
 			if ok {
+				// Trick of filtering without allocating
+				// https://github.com/golang/go/wiki/SliceTricks#filtering-without-allocating
+				nwalks := walks[:0]
 				// Look for walkInfo, remove it from the walks list.
-				for i, walk := range walks {
-					if walk == walkInfo {
-						walks = append(walks[:i], walks[i+1:]...)
+				for _, walk := range walks {
+					if !reflect.DeepEqual(walk, walkInfo) {
+						nwalks = append(nwalks, walk)
 					}
 				}
-				if len(walks) == 0 {
+				if len(nwalks) == 0 {
 					// No more treeWalk go-routines associated with listParams
 					// hence remove map entry.
 					delete(t.pool, params)
 				} else {
 					// There are more treeWalk go-routines associated with listParams
 					// hence save the list in the map.
-					t.pool[params] = walks
+					t.pool[params] = nwalks
 				}
 			}
 			// Signal the treeWalk go-routine to die.

cmd/tree-walk.go

@@ -25,40 +25,9 @@ import (
 // TreeWalkResult - Tree walk result carries results of tree walking.
 type TreeWalkResult struct {
 	entry string
-	err   error
 	end   bool
 }
 
-// posix.ListDir returns entries with trailing "/" for directories. At the object layer
-// we need to remove this trailing "/" for objects and retain "/" for prefixes before
-// sorting because the trailing "/" can affect the sorting results for certain cases.
-// Ex. lets say entries = ["a-b/", "a/"] and both are objects.
-//     sorting with out trailing "/" = ["a", "a-b"]
-//     sorting with trailing "/"     = ["a-b/", "a/"]
-// Hence if entries[] does not have a case like the above example then isLeaf() check
-// can be delayed till the entry is pushed into the TreeWalkResult channel.
-// delayIsLeafCheck() returns true if isLeaf can be delayed or false if
-// isLeaf should be done in listDir()
-func delayIsLeafCheck(entries []string) bool {
-	for i, entry := range entries {
-		if i == len(entries)-1 {
-			break
-		}
-		// If any byte in the "entry" string is less than '/' then the
-		// next "entry" should not contain '/' at the same same byte position.
-		for j := 0; j < len(entry); j++ {
-			if entry[j] < '/' {
-				if len(entries[i+1]) > j {
-					if entries[i+1][j] == '/' {
-						return false
-					}
-				}
-			}
-		}
-	}
-	return true
-}
-
 // Return entries that have prefix prefixEntry.
 // Note: input entries are expected to be sorted.
 func filterMatchingPrefix(entries []string, prefixEntry string) []string {
@@ -82,51 +51,15 @@ func filterMatchingPrefix(entries []string, prefixEntry string) []string {
 		}
 		end--
 	}
+	sort.Strings(entries[start:end])
 	return entries[start:end]
 }
 
 // ListDirFunc - "listDir" function of type listDirFunc returned by listDirFactory() - explained below.
-type ListDirFunc func(bucket, prefixDir, prefixEntry string) (entries []string, delayIsLeaf bool)
-
-// IsLeafFunc - A function isLeaf of type isLeafFunc is used to detect if an entry is a leaf entry. There are four scenarios
-// where isLeaf should behave differently:
-// 1. FS backend object listing - isLeaf is true if the entry has a trailing "/"
-// 2. FS backend multipart listing - isLeaf is true if the entry is a directory and contains uploads.json
-// 3. XL backend object listing - isLeaf is true if the entry is a directory and contains xl.json
-// 4. XL backend multipart listing - isLeaf is true if the entry is a directory and contains uploads.json
-type IsLeafFunc func(string, string) bool
-
-// IsLeafDirFunc - A function isLeafDir of type isLeafDirFunc is used to detect if an entry represents an empty directory.
-type IsLeafDirFunc func(string, string) bool
-
-func filterListEntries(bucket, prefixDir string, entries []string, prefixEntry string, isLeaf IsLeafFunc) ([]string, bool) {
-	// Listing needs to be sorted.
-	sort.Strings(entries)
-
-	// Filter entries that have the prefix prefixEntry.
-	entries = filterMatchingPrefix(entries, prefixEntry)
-
-	// Can isLeaf() check be delayed till when it has to be sent down the
-	// TreeWalkResult channel?
-	delayIsLeaf := delayIsLeafCheck(entries)
-	if delayIsLeaf {
-		return entries, true
-	}
-
-	// isLeaf() check has to happen here so that trailing "/" for objects can be removed.
-	for i, entry := range entries {
-		if isLeaf(bucket, pathJoin(prefixDir, entry)) {
-			entries[i] = strings.TrimSuffix(entry, slashSeparator)
-		}
-	}
-	// Sort again after removing trailing "/" for objects as the previous sort
-	// does not hold good anymore.
-	sort.Strings(entries)
-	return entries, false
-}
+type ListDirFunc func(bucket, prefixDir, prefixEntry string) (entries []string)
 
 // treeWalk walks directory tree recursively pushing TreeWalkResult into the channel as and when it encounters files.
-func doTreeWalk(ctx context.Context, bucket, prefixDir, entryPrefixMatch, marker string, recursive bool, listDir ListDirFunc, isLeaf IsLeafFunc, isLeafDir IsLeafDirFunc, resultCh chan TreeWalkResult, endWalkCh chan struct{}, isEnd bool) error {
+func doTreeWalk(ctx context.Context, bucket, prefixDir, entryPrefixMatch, marker string, recursive bool, listDir ListDirFunc, resultCh chan TreeWalkResult, endWalkCh chan struct{}, isEnd bool) (totalNum int, treeErr error) {
 	// Example:
 	// if prefixDir="one/two/three/" and marker="four/five.txt" treeWalk is recursively
 	// called with prefixDir="one/two/three/four/" and marker="five.txt"
@@ -142,15 +75,10 @@ func doTreeWalk(ctx context.Context, bucket, prefixDir, entryPrefixMatch, marker
 		}
 	}
 
-	entries, delayIsLeaf := listDir(bucket, prefixDir, entryPrefixMatch)
-	// When isleaf check is delayed, make sure that it is set correctly here.
-	if delayIsLeaf && isLeaf == nil {
-		return errInvalidArgument
-	}
-
+	entries := listDir(bucket, prefixDir, entryPrefixMatch)
 	// For an empty list return right here.
 	if len(entries) == 0 {
-		return nil
+		return 0, nil
 	}
 
 	// example:
@@ -162,27 +90,13 @@ func doTreeWalk(ctx context.Context, bucket, prefixDir, entryPrefixMatch, marker
 	entries = entries[idx:]
 	// For an empty list after search through the entries, return right here.
 	if len(entries) == 0 {
-		return nil
+		return 0, nil
 	}
 
 	for i, entry := range entries {
-		var leaf, leafDir bool
+		pentry := pathJoin(prefixDir, entry)
+		isDir := hasSuffix(pentry, slashSeparator)
 
-		// Decision to do isLeaf check was pushed from listDir() to here.
-		if delayIsLeaf {
-			leaf = isLeaf(bucket, pathJoin(prefixDir, entry))
-			if leaf {
-				entry = strings.TrimSuffix(entry, slashSeparator)
-			}
-		} else {
-			leaf = !strings.HasSuffix(entry, slashSeparator)
-		}
-
-		if strings.HasSuffix(entry, slashSeparator) {
-			leafDir = isLeafDir(bucket, pathJoin(prefixDir, entry))
-		}
-
-		isDir := !leafDir && !leaf
 		if i == 0 && markerDir == entry {
 			if !recursive {
 				// Skip as the marker would already be listed in the previous listing.
@@ -209,26 +123,35 @@ func doTreeWalk(ctx context.Context, bucket, prefixDir, entryPrefixMatch, marker
 			// markIsEnd is passed to this entry's treeWalk() so that treeWalker.end can be marked
 			// true at the end of the treeWalk stream.
 			markIsEnd := i == len(entries)-1 && isEnd
-			if tErr := doTreeWalk(ctx, bucket, pathJoin(prefixDir, entry), prefixMatch, markerArg, recursive, listDir, isLeaf, isLeafDir, resultCh, endWalkCh, markIsEnd); tErr != nil {
-				return tErr
+			totalFound, err := doTreeWalk(ctx, bucket, pentry, prefixMatch, markerArg, recursive,
+				listDir, resultCh, endWalkCh, markIsEnd)
+			if err != nil {
+				return 0, err
+			}
+
+			// A nil totalFound means this is an empty directory that
+			// needs to be sent to the result channel, otherwise continue
+			// to the next entry.
+			if totalFound > 0 {
+				continue
 			}
-			continue
 		}
+
 		// EOF is set if we are at last entry and the caller indicated we at the end.
 		isEOF := ((i == len(entries)-1) && isEnd)
 		select {
 		case <-endWalkCh:
-			return errWalkAbort
-		case resultCh <- TreeWalkResult{entry: pathJoin(prefixDir, entry), end: isEOF}:
+			return 0, errWalkAbort
+		case resultCh <- TreeWalkResult{entry: pentry, end: isEOF}:
 		}
 	}
 
 	// Everything is listed.
-	return nil
+	return len(entries), nil
 }
 
 // Initiate a new treeWalk in a goroutine.
-func startTreeWalk(ctx context.Context, bucket, prefix, marker string, recursive bool, listDir ListDirFunc, isLeaf IsLeafFunc, isLeafDir IsLeafDirFunc, endWalkCh chan struct{}) chan TreeWalkResult {
+func startTreeWalk(ctx context.Context, bucket, prefix, marker string, recursive bool, listDir ListDirFunc, endWalkCh chan struct{}) chan TreeWalkResult {
 	// Example 1
 	// If prefix is "one/two/three/" and marker is "one/two/three/four/five.txt"
 	// treeWalk is called with prefixDir="one/two/three/" and marker="four/five.txt"
@@ -250,7 +173,7 @@ func startTreeWalk(ctx context.Context, bucket, prefix, marker string, recursive
 	marker = strings.TrimPrefix(marker, prefixDir)
 	go func() {
 		isEnd := true // Indication to start walking the tree with end as true.
-		doTreeWalk(ctx, bucket, prefixDir, entryPrefixMatch, marker, recursive, listDir, isLeaf, isLeafDir, resultCh, endWalkCh, isEnd)
+		doTreeWalk(ctx, bucket, prefixDir, entryPrefixMatch, marker, recursive, listDir, resultCh, endWalkCh, isEnd)
 		close(resultCh)
 	}()
 	return resultCh

cmd/tree-walk_test.go

@@ -30,49 +30,6 @@ import (
 // Fixed volume name that could be used across tests
 const volume = "testvolume"
 
-// Test for delayIsLeafCheck.
-func TestDelayIsLeafCheck(t *testing.T) {
-	testCases := []struct {
-		entries []string
-		delay   bool
-	}{
-		// Test cases where isLeaf check can't be delayed.
-		{
-			[]string{"a-b/", "a/"},
-			false,
-		},
-		{
-			[]string{"a%b/", "a/"},
-			false,
-		},
-		{
-			[]string{"a-b-c", "a-b/"},
-			false,
-		},
-
-		// Test cases where isLeaf check can be delayed.
-		{
-			[]string{"a-b/", "aa/"},
-			true,
-		},
-		{
-			[]string{"a", "a-b"},
-			true,
-		},
-		{
-			[]string{"aaa", "bbb"},
-			true,
-		},
-	}
-	for i, testCase := range testCases {
-		expected := testCase.delay
-		got := delayIsLeafCheck(testCase.entries)
-		if expected != got {
-			t.Errorf("Test %d : Expected %t got %t", i+1, expected, got)
-		}
-	}
-}
-
 // Test for filterMatchingPrefix.
 func TestFilterMatchingPrefix(t *testing.T) {
 	entries := []string{"a", "aab", "ab", "abbbb", "zzz"}
@@ -128,11 +85,11 @@ func createNamespace(disk StorageAPI, volume string, files []string) error {
 
 // Test if tree walker returns entries matching prefix alone are received
 // when a non empty prefix is supplied.
-func testTreeWalkPrefix(t *testing.T, listDir ListDirFunc, isLeaf IsLeafFunc, isLeafDir IsLeafDirFunc) {
+func testTreeWalkPrefix(t *testing.T, listDir ListDirFunc) {
 	// Start the tree walk go-routine.
 	prefix := "d/"
 	endWalkCh := make(chan struct{})
-	twResultCh := startTreeWalk(context.Background(), volume, prefix, "", true, listDir, isLeaf, isLeafDir, endWalkCh)
+	twResultCh := startTreeWalk(context.Background(), volume, prefix, "", true, listDir, endWalkCh)
 
 	// Check if all entries received on the channel match the prefix.
 	for res := range twResultCh {
@@ -143,11 +100,11 @@ func testTreeWalkPrefix(t *testing.T, listDir ListDirFunc, isLeaf IsLeafFunc, is
 }
 
 // Test if entries received on tree walk's channel appear after the supplied marker.
-func testTreeWalkMarker(t *testing.T, listDir ListDirFunc, isLeaf IsLeafFunc, isLeafDir IsLeafDirFunc) {
+func testTreeWalkMarker(t *testing.T, listDir ListDirFunc) {
 	// Start the tree walk go-routine.
 	prefix := ""
 	endWalkCh := make(chan struct{})
-	twResultCh := startTreeWalk(context.Background(), volume, prefix, "d/g", true, listDir, isLeaf, isLeafDir, endWalkCh)
+	twResultCh := startTreeWalk(context.Background(), volume, prefix, "d/g", true, listDir, endWalkCh)
 
 	// Check if only 3 entries, namely d/g/h, i/j/k, lmn are received on the channel.
 	expectedCount := 3
@@ -184,23 +141,13 @@ func TestTreeWalk(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	isLeaf := func(volume, prefix string) bool {
-		return !hasSuffix(prefix, slashSeparator)
-	}
+	listDir := listDirFactory(context.Background(), disk)
 
-	isLeafDir := func(volume, prefix string) bool {
-		entries, listErr := disk.ListDir(volume, prefix, 1)
-		if listErr != nil {
-			return false
-		}
-		return len(entries) == 0
-	}
-
-	listDir := listDirFactory(context.Background(), isLeaf, disk)
 	// Simple test for prefix based walk.
-	testTreeWalkPrefix(t, listDir, isLeaf, isLeafDir)
+	testTreeWalkPrefix(t, listDir)
 	// Simple test when marker is set.
-	testTreeWalkMarker(t, listDir, isLeaf, isLeafDir)
+	testTreeWalkMarker(t, listDir)
+
 	err = os.RemoveAll(fsDir)
 	if err != nil {
 		t.Fatal(err)
@@ -228,19 +175,7 @@ func TestTreeWalkTimeout(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	isLeaf := func(volume, prefix string) bool {
-		return !hasSuffix(prefix, slashSeparator)
-	}
-
-	isLeafDir := func(volume, prefix string) bool {
-		entries, listErr := disk.ListDir(volume, prefix, 1)
-		if listErr != nil {
-			return false
-		}
-		return len(entries) == 0
-	}
-
-	listDir := listDirFactory(context.Background(), isLeaf, disk)
+	listDir := listDirFactory(context.Background(), disk)
 
 	// TreeWalk pool with 2 seconds timeout for tree-walk go routines.
 	pool := NewTreeWalkPool(2 * time.Second)
@@ -249,7 +184,7 @@ func TestTreeWalkTimeout(t *testing.T) {
 	prefix := ""
 	marker := ""
 	recursive := true
-	resultCh := startTreeWalk(context.Background(), volume, prefix, marker, recursive, listDir, isLeaf, isLeafDir, endWalkCh)
+	resultCh := startTreeWalk(context.Background(), volume, prefix, marker, recursive, listDir, endWalkCh)
 
 	params := listParams{
 		bucket:    volume,
@@ -313,9 +248,7 @@ func TestListDir(t *testing.T) {
 	}
 
 	// create listDir function.
-	listDir := listDirFactory(context.Background(), func(volume, prefix string) bool {
-		return !hasSuffix(prefix, slashSeparator)
-	}, disk1, disk2)
+	listDir := listDirFactory(context.Background(), disk1, disk2)
 
 	// Create file1 in fsDir1 and file2 in fsDir2.
 	disks := []StorageAPI{disk1, disk2}
@@ -327,7 +260,7 @@ func TestListDir(t *testing.T) {
 	}
 
 	// Should list "file1" from fsDir1.
-	entries, _ := listDir(volume, "", "")
+	entries := listDir(volume, "", "")
 	if len(entries) != 2 {
 		t.Fatal("Expected the number of entries to be 2")
 	}
@@ -345,7 +278,7 @@ func TestListDir(t *testing.T) {
 	}
 
 	// Should list "file2" from fsDir2.
-	entries, _ = listDir(volume, "", "")
+	entries = listDir(volume, "", "")
 	if len(entries) != 1 {
 		t.Fatal("Expected the number of entries to be 1")
 	}
@@ -373,21 +306,8 @@ func TestRecursiveTreeWalk(t *testing.T) {
 		t.Fatalf("Unable to create StorageAPI: %s", err)
 	}
 
-	// Simple isLeaf check, returns true if there is no trailing "/"
-	isLeaf := func(volume, prefix string) bool {
-		return !hasSuffix(prefix, slashSeparator)
-	}
-
-	isLeafDir := func(volume, prefix string) bool {
-		entries, listErr := disk1.ListDir(volume, prefix, 1)
-		if listErr != nil {
-			return false
-		}
-		return len(entries) == 0
-	}
-
 	// Create listDir function.
-	listDir := listDirFactory(context.Background(), isLeaf, disk1)
+	listDir := listDirFactory(context.Background(), disk1)
 
 	// Create the namespace.
 	var files = []string{
@@ -461,13 +381,16 @@ func TestRecursiveTreeWalk(t *testing.T) {
 		}},
 	}
 	for i, testCase := range testCases {
-		for entry := range startTreeWalk(context.Background(), volume,
-			testCase.prefix, testCase.marker, testCase.recursive,
-			listDir, isLeaf, isLeafDir, endWalkCh) {
-			if _, found := testCase.expected[entry.entry]; !found {
-				t.Errorf("Test %d: Expected %s, but couldn't find", i+1, entry.entry)
+		testCase := testCase
+		t.Run(fmt.Sprintf("Test%d", i+1), func(t *testing.T) {
+			for entry := range startTreeWalk(context.Background(), volume,
+				testCase.prefix, testCase.marker, testCase.recursive,
+				listDir, endWalkCh) {
+				if _, found := testCase.expected[entry.entry]; !found {
+					t.Errorf("Expected %s, but couldn't find", entry.entry)
+				}
 			}
-		}
+		})
 	}
 	err = os.RemoveAll(fsDir1)
 	if err != nil {
@@ -488,21 +411,8 @@ func TestSortedness(t *testing.T) {
 		t.Fatalf("Unable to create StorageAPI: %s", err)
 	}
 
-	// Simple isLeaf check, returns true if there is no trailing "/"
-	isLeaf := func(volume, prefix string) bool {
-		return !hasSuffix(prefix, slashSeparator)
-	}
-
-	isLeafDir := func(volume, prefix string) bool {
-		entries, listErr := disk1.ListDir(volume, prefix, 1)
-		if listErr != nil {
-			return false
-		}
-		return len(entries) == 0
-	}
-
 	// Create listDir function.
-	listDir := listDirFactory(context.Background(), isLeaf, disk1)
+	listDir := listDirFactory(context.Background(), disk1)
 
 	// Create the namespace.
 	var files = []string{
@@ -544,7 +454,7 @@ func TestSortedness(t *testing.T) {
 		var actualEntries []string
 		for entry := range startTreeWalk(context.Background(), volume,
 			test.prefix, test.marker, test.recursive,
-			listDir, isLeaf, isLeafDir, endWalkCh) {
+			listDir, endWalkCh) {
 			actualEntries = append(actualEntries, entry.entry)
 		}
 		if !sort.IsSorted(sort.StringSlice(actualEntries)) {
@@ -572,20 +482,8 @@ func TestTreeWalkIsEnd(t *testing.T) {
 		t.Fatalf("Unable to create StorageAPI: %s", err)
 	}
 
-	isLeaf := func(volume, prefix string) bool {
-		return !hasSuffix(prefix, slashSeparator)
-	}
-
-	isLeafDir := func(volume, prefix string) bool {
-		entries, listErr := disk1.ListDir(volume, prefix, 1)
-		if listErr != nil {
-			return false
-		}
-		return len(entries) == 0
-	}
-
 	// Create listDir function.
-	listDir := listDirFactory(context.Background(), isLeaf, disk1)
+	listDir := listDirFactory(context.Background(), disk1)
 
 	// Create the namespace.
 	var files = []string{
@@ -626,7 +524,8 @@ func TestTreeWalkIsEnd(t *testing.T) {
 	}
 	for i, test := range testCases {
 		var entry TreeWalkResult
-		for entry = range startTreeWalk(context.Background(), volume, test.prefix, test.marker, test.recursive, listDir, isLeaf, isLeafDir, endWalkCh) {
+		for entry = range startTreeWalk(context.Background(), volume, test.prefix,
+			test.marker, test.recursive, listDir, endWalkCh) {
 		}
 		if entry.entry != test.expectedEntry {
 			t.Errorf("Test %d: Expected entry %s, but received %s with the EOF marker", i, test.expectedEntry, entry.entry)

cmd/typed-errors.go

@@ -85,3 +85,6 @@ var errNoSuchPolicy = errors.New("Specified canned policy does not exist")
 
 // error returned when access is denied.
 var errAccessDenied = errors.New("Do not have enough permissions to access this resource")
+
+// errNetworkConnReset - connection reset by peer
+var errNetworkConnReset = errors.New("connection reset by peer")

cmd/utils.go

@@ -474,3 +474,10 @@ func restQueries(keys ...string) []string {
 	}
 	return accumulator
 }
+
+// Reverse the input order of a slice of string
+func reverseStringSlice(input []string) {
+	for left, right := 0, len(input)-1; left < right; left, right = left+1, right-1 {
+		input[left], input[right] = input[right], input[left]
+	}
+}

cmd/web-handlers.go

@@ -583,7 +583,22 @@ func (web *webAPIHandlers) RemoveObject(r *http.Request, args *RemoveObjectArgs,
 
 	claims, owner, authErr := webRequestAuthenticate(r)
 	if authErr != nil {
-		return toJSONError(authErr)
+		if authErr == errNoAuthToken {
+			// Check if all objects are allowed to be deleted anonymously
+			for _, object := range args.Objects {
+				if !globalPolicySys.IsAllowed(policy.Args{
+					Action:          policy.DeleteObjectAction,
+					BucketName:      args.BucketName,
+					ConditionValues: getConditionValues(r, "", ""),
+					IsOwner:         false,
+					ObjectName:      object,
+				}) {
+					return toJSONError(errAuthentication)
+				}
+			}
+		} else {
+			return toJSONError(authErr)
+		}
 	}
 
 	if args.BucketName == "" || len(args.Objects) == 0 {
@@ -640,16 +655,20 @@ next:
 					return toJSONError(errMethodNotAllowed)
 				}
 			}
-
-			if !globalIAMSys.IsAllowed(iampolicy.Args{
-				AccountName:     claims.Subject,
-				Action:          iampolicy.DeleteObjectAction,
-				BucketName:      args.BucketName,
-				ConditionValues: getConditionValues(r, "", claims.Subject),
-				IsOwner:         owner,
-				ObjectName:      objectName,
-			}) {
-				return toJSONError(errAccessDenied)
+			// Check for permissions only in the case of
+			// non-anonymous login. For anonymous login, policy has already
+			// been checked.
+			if authErr != errNoAuthToken {
+				if !globalIAMSys.IsAllowed(iampolicy.Args{
+					AccountName:     claims.Subject,
+					Action:          iampolicy.DeleteObjectAction,
+					BucketName:      args.BucketName,
+					ConditionValues: getConditionValues(r, "", claims.Subject),
+					IsOwner:         owner,
+					ObjectName:      objectName,
+				}) {
+					return toJSONError(errAccessDenied)
+				}
 			}
 
 			if err = deleteObject(context.Background(), objectAPI, web.CacheAPI(), args.BucketName, objectName, r); err != nil {
@@ -934,7 +953,7 @@ func (web *webAPIHandlers) Upload(w http.ResponseWriter, r *http.Request) {
 	var reader io.Reader = r.Body
 	actualSize := size
 
-	hashReader, err := hash.NewReader(reader, size, "", "", actualSize)
+	hashReader, err := hash.NewReader(reader, size, "", "", actualSize, globalCLIContext.StrictS3Compat)
 	if err != nil {
 		writeWebErrorResponse(w, err)
 		return
@@ -944,7 +963,7 @@ func (web *webAPIHandlers) Upload(w http.ResponseWriter, r *http.Request) {
 		metadata[ReservedMetadataPrefix+"compression"] = compressionAlgorithmV1
 		metadata[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(size, 10)
 
-		actualReader, err := hash.NewReader(reader, size, "", "", actualSize)
+		actualReader, err := hash.NewReader(reader, size, "", "", actualSize, globalCLIContext.StrictS3Compat)
 		if err != nil {
 			writeWebErrorResponse(w, err)
 			return
@@ -953,7 +972,7 @@ func (web *webAPIHandlers) Upload(w http.ResponseWriter, r *http.Request) {
 		// Set compression metrics.
 		size = -1 // Since compressed size is un-predictable.
 		reader = newSnappyCompressReader(actualReader)
-		hashReader, err = hash.NewReader(reader, size, "", "", actualSize)
+		hashReader, err = hash.NewReader(reader, size, "", "", actualSize, globalCLIContext.StrictS3Compat)
 		if err != nil {
 			writeWebErrorResponse(w, err)
 			return
@@ -977,7 +996,8 @@ func (web *webAPIHandlers) Upload(w http.ResponseWriter, r *http.Request) {
 				return
 			}
 			info := ObjectInfo{Size: size}
-			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", size) // do not try to verify encrypted content
+			// do not try to verify encrypted content
+			hashReader, err = hash.NewReader(reader, info.EncryptedSize(), "", "", size, globalCLIContext.StrictS3Compat)
 			if err != nil {
 				writeErrorResponse(ctx, w, toAPIError(ctx, err), r.URL, guessIsBrowserReq(r))
 				return

cmd/xl-sets.go

@@ -23,7 +23,6 @@ import (
 	"io"
 	"net/http"
 	"sort"
-	"strings"
 	"sync"
 	"time"
 
@@ -76,8 +75,8 @@ type xlSets struct {
 	// Distribution algorithm of choice.
 	distributionAlgo string
 
-	// Pack level listObjects pool management.
-	listPool *TreeWalkPool
+	// Merge tree walk
+	pool *MergeWalkPool
 }
 
 // isConnected - checks if the endpoint is connected or not.
@@ -270,7 +269,7 @@ func newXLSets(endpoints EndpointList, format *formatXLV3, setCount int, drivesP
 		format:             format,
 		disksConnectDoneCh: make(chan struct{}),
 		distributionAlgo:   format.XL.DistributionAlgo,
-		listPool:           NewTreeWalkPool(globalLookupTimeout),
+		pool:               NewMergeWalkPool(globalMergeLookupTimeout),
 	}
 
 	mutex := newNSLock(globalIsDistXL)
@@ -466,9 +465,14 @@ func hashKey(algo string, key string, cardinality int) int {
 	}
 }
 
+// Returns always a same erasure coded set for a given input.
+func (s *xlSets) getHashedSetIndex(input string) int {
+	return hashKey(s.distributionAlgo, input, len(s.sets))
+}
+
 // Returns always a same erasure coded set for a given input.
 func (s *xlSets) getHashedSet(input string) (set *xlObjects) {
-	return s.sets[hashKey(s.distributionAlgo, input, len(s.sets))]
+	return s.sets[s.getHashedSetIndex(input)]
 }
 
 // GetBucketInfo - returns bucket info from one of the erasure coded set.
@@ -618,6 +622,58 @@ func (s *xlSets) DeleteObject(ctx context.Context, bucket string, object string)
 	return s.getHashedSet(object).DeleteObject(ctx, bucket, object)
 }
 
+// DeleteObjects - bulk delete of objects
+// Bulk delete is only possible within one set. For that purpose
+// objects are group by set first, and then bulk delete is invoked
+// for each set, the error response of each delete will be returned
+func (s *xlSets) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+
+	type delObj struct {
+		// Set index associated to this object
+		setIndex int
+		// Original index from the list of arguments
+		// where this object is passed
+		origIndex int
+		// Object name
+		name string
+	}
+
+	// Transform []delObj to the list of object names
+	toNames := func(delObjs []delObj) []string {
+		names := make([]string, len(delObjs))
+		for i, obj := range delObjs {
+			names[i] = obj.name
+		}
+		return names
+	}
+
+	// The result of delete operation on all passed objects
+	var delErrs = make([]error, len(objects))
+
+	// A map between a set and its associated objects
+	var objSetMap = make(map[int][]delObj)
+
+	// Group objects by set index
+	for i, object := range objects {
+		index := s.getHashedSetIndex(object)
+		objSetMap[index] = append(objSetMap[index], delObj{setIndex: index, origIndex: i, name: object})
+	}
+
+	// Invoke bulk delete on objects per set and save
+	// the result of the delete operation
+	for _, objsGroup := range objSetMap {
+		errs, err := s.getHashedSet(objsGroup[0].name).DeleteObjects(ctx, bucket, toNames(objsGroup))
+		if err != nil {
+			return nil, err
+		}
+		for i, obj := range objsGroup {
+			delErrs[obj.origIndex] = errs[i]
+		}
+	}
+
+	return delErrs, nil
+}
+
 // CopyObject - copies objects from one hashedSet to another hashedSet, on server side.
 func (s *xlSets) CopyObject(ctx context.Context, srcBucket, srcObject, destBucket, destObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (objInfo ObjectInfo, err error) {
 	srcSet := s.getHashedSet(srcObject)
@@ -641,9 +697,8 @@ func (s *xlSets) CopyObject(ctx context.Context, srcBucket, srcObject, destBucke
 }
 
 // Returns function "listDir" of the type listDirFunc.
-// isLeaf - is used by listDir function to check if an entry is a leaf or non-leaf entry.
 // disks - used for doing disk.ListDir(). Sets passes set of disks.
-func listDirSetsFactory(ctx context.Context, isLeaf IsLeafFunc, isLeafDir IsLeafDirFunc, sets ...*xlObjects) ListDirFunc {
+func listDirSetsFactory(ctx context.Context, sets ...*xlObjects) ListDirFunc {
 	listDirInternal := func(bucket, prefixDir, prefixEntry string, disks []StorageAPI) (mergedEntries []string) {
 		var diskEntries = make([][]string, len(disks))
 		var wg sync.WaitGroup
@@ -654,7 +709,7 @@ func listDirSetsFactory(ctx context.Context, isLeaf IsLeafFunc, isLeafDir IsLeaf
 			wg.Add(1)
 			go func(index int, disk StorageAPI) {
 				defer wg.Done()
-				diskEntries[index], _ = disk.ListDir(bucket, prefixDir, -1)
+				diskEntries[index], _ = disk.ListDir(bucket, prefixDir, -1, xlMetaJSONFile)
 			}(index, disk)
 		}
 
@@ -684,7 +739,7 @@ func listDirSetsFactory(ctx context.Context, isLeaf IsLeafFunc, isLeafDir IsLeaf
 	}
 
 	// listDir - lists all the entries at a given prefix and given entry in the prefix.
-	listDir := func(bucket, prefixDir, prefixEntry string) (mergedEntries []string, delayIsLeaf bool) {
+	listDir := func(bucket, prefixDir, prefixEntry string) (mergedEntries []string) {
 		for _, set := range sets {
 			var newEntries []string
 			// Find elements in entries which are not in mergedEntries
@@ -703,47 +758,245 @@ func listDirSetsFactory(ctx context.Context, isLeaf IsLeafFunc, isLeafDir IsLeaf
 				sort.Strings(mergedEntries)
 			}
 		}
-		return filterListEntries(bucket, prefixDir, mergedEntries, prefixEntry, isLeaf)
+		return filterMatchingPrefix(mergedEntries, prefixEntry)
 	}
 	return listDir
 }
 
-// ListObjects - implements listing of objects across sets, each set is independently
-// listed and subsequently merge lexically sorted inside listDirSetsFactory(). Resulting
-// value through the walk channel receives the data properly lexically sorted.
-func (s *xlSets) ListObjects(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (ListObjectsInfo, error) {
-	isLeaf := func(bucket, entry string) bool {
-		entry = strings.TrimSuffix(entry, slashSeparator)
-		// Verify if we are at the leaf, a leaf is where we
-		// see `xl.json` inside a directory.
-		return s.getHashedSet(entry).isObject(bucket, entry)
+// FileInfoCh - file info channel
+type FileInfoCh struct {
+	Ch    chan FileInfo
+	Prev  FileInfo
+	Valid bool
+}
+
+// Pop - pops a cached entry if any, or from the cached channel.
+func (f *FileInfoCh) Pop() (fi FileInfo, ok bool) {
+	if f.Valid {
+		f.Valid = false
+		return f.Prev, true
+	} // No cached entries found, read from channel
+	f.Prev, ok = <-f.Ch
+	return f.Prev, ok
+}
+
+// Push - cache an entry, for Pop() later.
+func (f *FileInfoCh) Push(fi FileInfo) {
+	f.Prev = fi
+	f.Valid = true
+}
+
+// Calculate least entry across multiple FileInfo channels, additionally
+// returns a boolean to indicate if the caller needs to call again.
+func leastEntry(entriesCh []FileInfoCh, readQuorum int) (FileInfo, bool) {
+	var entriesValid = make([]bool, len(entriesCh))
+	var entries = make([]FileInfo, len(entriesCh))
+	for i := range entriesCh {
+		entries[i], entriesValid[i] = entriesCh[i].Pop()
 	}
 
-	isLeafDir := func(bucket, entry string) bool {
-		// Verify prefixes in all sets.
-		var ok bool
-		for _, set := range s.sets {
-			ok = set.isObjectDir(bucket, entry)
-			if ok {
-				return true
+	var isTruncated = false
+	for _, valid := range entriesValid {
+		if !valid {
+			continue
+		}
+		isTruncated = true
+		break
+	}
+
+	var lentry FileInfo
+	var found bool
+	for i, valid := range entriesValid {
+		if !valid {
+			continue
+		}
+		if !found {
+			lentry = entries[i]
+			found = true
+			continue
+		}
+		if entries[i].Name < lentry.Name {
+			lentry = entries[i]
+		}
+	}
+
+	// We haven't been able to find any least entry,
+	// this would mean that we don't have valid.
+	if !found {
+		return lentry, isTruncated
+	}
+
+	leastEntryCount := 0
+	for i, valid := range entriesValid {
+		if !valid {
+			continue
+		}
+
+		// Entries are duplicated across disks,
+		// we should simply skip such entries.
+		if lentry.Name == entries[i].Name && lentry.ModTime.Equal(entries[i].ModTime) {
+			leastEntryCount++
+			continue
+		}
+
+		// Push all entries which are lexically higher
+		// and will be returned later in Pop()
+		entriesCh[i].Push(entries[i])
+	}
+
+	quorum := lentry.Quorum
+	if quorum == 0 {
+		quorum = readQuorum
+	}
+
+	if leastEntryCount >= quorum {
+		return lentry, isTruncated
+	}
+
+	return leastEntry(entriesCh, readQuorum)
+}
+
+// mergeEntriesCh - merges FileInfo channel to entries upto maxKeys.
+func mergeEntriesCh(entriesCh []FileInfoCh, maxKeys int, readQuorum int) (entries FilesInfo) {
+	for i := 0; i < maxKeys; {
+		var fi FileInfo
+		fi, entries.IsTruncated = leastEntry(entriesCh, readQuorum)
+		if !entries.IsTruncated {
+			break
+		}
+		entries.Files = append(entries.Files, fi)
+		i++
+	}
+	return entries
+}
+
+// Starts a walk channel across all disks and returns a slice.
+func (s *xlSets) startMergeWalks(ctx context.Context, bucket, prefix, marker string, recursive bool, endWalkCh chan struct{}) []FileInfoCh {
+	var entryChs []FileInfoCh
+	for _, set := range s.sets {
+		for _, disk := range set.getDisks() {
+			if disk == nil {
+				// Disk can be offline
+				continue
+			}
+			entryCh, err := disk.Walk(bucket, prefix, marker, recursive, xlMetaJSONFile, readMetadata, endWalkCh)
+			if err != nil {
+				// Disk walk returned error, ignore it.
+				continue
+			}
+			entryChs = append(entryChs, FileInfoCh{
+				Ch: entryCh,
+			})
+		}
+	}
+	return entryChs
+}
+
+// ListObjects - implements listing of objects across disks, each disk is indepenently
+// walked and merged at this layer. Resulting value through the merge process sends
+// the data in lexically sorted order.
+func (s *xlSets) ListObjects(ctx context.Context, bucket, prefix, marker, delimiter string, maxKeys int) (loi ListObjectsInfo, err error) {
+	if err = checkListObjsArgs(ctx, bucket, prefix, marker, delimiter, s); err != nil {
+		return loi, err
+	}
+
+	// Marker is set validate pre-condition.
+	if marker != "" {
+		// Marker not common with prefix is not implemented. Send an empty response
+		if !hasPrefix(marker, prefix) {
+			return loi, nil
+		}
+	}
+
+	// With max keys of zero we have reached eof, return right here.
+	if maxKeys == 0 {
+		return loi, nil
+	}
+
+	// For delimiter and prefix as '/' we do not list anything at all
+	// since according to s3 spec we stop at the 'delimiter'
+	// along // with the prefix. On a flat namespace with 'prefix'
+	// as '/' we don't have any entries, since all the keys are
+	// of form 'keyName/...'
+	if delimiter == slashSeparator && prefix == slashSeparator {
+		return loi, nil
+	}
+
+	// Over flowing count - reset to maxObjectList.
+	if maxKeys < 0 || maxKeys > maxObjectList {
+		maxKeys = maxObjectList
+	}
+
+	// Default is recursive, if delimiter is set then list non recursive.
+	recursive := true
+	if delimiter == slashSeparator {
+		recursive = false
+	}
+
+	entryChs, endWalkCh := s.pool.Release(listParams{bucket, recursive, marker, prefix})
+	if entryChs == nil {
+		endWalkCh = make(chan struct{})
+		entryChs = s.startMergeWalks(context.Background(), bucket, prefix, marker, recursive, endWalkCh)
+	}
+
+	entries := mergeEntriesCh(entryChs, maxKeys, s.drivesPerSet/2)
+	if len(entries.Files) == 0 {
+		return loi, nil
+	}
+
+	loi.Objects = make([]ObjectInfo, len(entries.Files))
+	loi.IsTruncated = entries.IsTruncated
+	if loi.IsTruncated {
+		loi.NextMarker = entries.Files[len(entries.Files)-1].Name
+	}
+
+	for _, entry := range entries.Files {
+		var objInfo ObjectInfo
+		if hasSuffix(entry.Name, slashSeparator) {
+			if !recursive {
+				loi.Prefixes = append(loi.Prefixes, entry.Name)
+				continue
+			}
+			objInfo = ObjectInfo{
+				Bucket: bucket,
+				Name:   entry.Name,
+				IsDir:  true,
+			}
+		} else {
+			objInfo = ObjectInfo{
+				IsDir:           false,
+				Bucket:          bucket,
+				Name:            entry.Name,
+				ModTime:         entry.ModTime,
+				Size:            entry.Size,
+				ContentType:     entry.Metadata["content-type"],
+				ContentEncoding: entry.Metadata["content-encoding"],
+			}
+
+			// Extract etag from metadata.
+			objInfo.ETag = extractETag(entry.Metadata)
+
+			// All the parts per object.
+			objInfo.Parts = entry.Parts
+
+			// 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-*.
+			objInfo.UserDefined = cleanMetadata(entry.Metadata)
+
+			// Update storage class
+			if sc, ok := entry.Metadata[amzStorageClass]; ok {
+				objInfo.StorageClass = sc
+			} else {
+				objInfo.StorageClass = globalMinioDefaultStorageClass
 			}
 		}
-		return false
+		loi.Objects = append(loi.Objects, objInfo)
 	}
-
-	listDir := listDirSetsFactory(ctx, isLeaf, isLeafDir, s.sets...)
-
-	var getObjectInfoDirs []func(context.Context, string, string) (ObjectInfo, error)
-	// Verify prefixes in all sets.
-	for _, set := range s.sets {
-		getObjectInfoDirs = append(getObjectInfoDirs, set.getObjectInfoDir)
+	if loi.IsTruncated {
+		s.pool.Set(listParams{bucket, recursive, loi.NextMarker, prefix}, entryChs, endWalkCh)
 	}
-
-	var getObjectInfo = func(ctx context.Context, bucket string, entry string) (ObjectInfo, error) {
-		return s.getHashedSet(entry).getObjectInfo(ctx, bucket, entry)
-	}
-
-	return listObjects(ctx, s, bucket, prefix, marker, delimiter, maxKeys, s.listPool, isLeaf, isLeafDir, listDir, getObjectInfo, getObjectInfoDirs...)
+	return loi, nil
 }
 
 func (s *xlSets) ListMultipartUploads(ctx context.Context, bucket, prefix, keyMarker, uploadIDMarker, delimiter string, maxUploads int) (result ListMultipartsInfo, err error) {
@@ -1256,26 +1509,14 @@ func (s *xlSets) HealObjects(ctx context.Context, bucket, prefix string, healObj
 	recursive := true
 
 	endWalkCh := make(chan struct{})
-	isLeaf := func(bucket, entry string) bool {
-		return hasSuffix(entry, xlMetaJSONFile)
-	}
-
-	isLeafDir := func(bucket, entry string) bool {
-		return false
-	}
-
-	listDir := listDirSetsFactory(ctx, isLeaf, isLeafDir, s.sets...)
-	walkResultCh := startTreeWalk(ctx, bucket, prefix, "", recursive, listDir, isLeaf, isLeafDir, endWalkCh)
+	listDir := listDirSetsFactory(ctx, s.sets...)
+	walkResultCh := startTreeWalk(ctx, bucket, prefix, "", recursive, listDir, endWalkCh)
 	for {
 		walkResult, ok := <-walkResultCh
 		if !ok {
 			break
 		}
-		// For any walk error return right away.
-		if walkResult.err != nil {
-			return toObjectErr(walkResult.err, bucket, prefix)
-		}
-		if err := healObjectFn(bucket, strings.TrimSuffix(walkResult.entry, slashSeparator+xlMetaJSONFile)); err != nil {
+		if err := healObjectFn(bucket, walkResult.entry); err != nil {
 			return toObjectErr(err, bucket, walkResult.entry)
 		}
 		if walkResult.end {

cmd/xl-v1-common.go

@@ -85,34 +85,3 @@ func (xl xlObjects) isObject(bucket, prefix string) (ok bool) {
 
 	return reduceReadQuorumErrs(context.Background(), errs, objectOpIgnoredErrs, readQuorum) == nil
 }
-
-// isObjectDir returns if the specified path represents an empty directory.
-func (xl xlObjects) isObjectDir(bucket, prefix string) (ok bool) {
-	var errs = make([]error, len(xl.getDisks()))
-	var wg sync.WaitGroup
-	for index, disk := range xl.getDisks() {
-		if disk == nil {
-			continue
-		}
-		wg.Add(1)
-		go func(index int, disk StorageAPI) {
-			defer wg.Done()
-			// Check if 'prefix' is an object on this 'disk', else continue the check the next disk
-			entries, err := disk.ListDir(bucket, prefix, 1)
-			if err != nil {
-				errs[index] = err
-				return
-			}
-			if len(entries) > 0 {
-				errs[index] = errVolumeNotEmpty
-				return
-			}
-		}(index, disk)
-	}
-
-	wg.Wait()
-
-	readQuorum := len(xl.getDisks()) / 2
-
-	return reduceReadQuorumErrs(context.Background(), errs, objectOpIgnoredErrs, readQuorum) == nil
-}

cmd/xl-v1-healing-common_test.go

@@ -311,9 +311,9 @@ func TestDisksWithAllParts(t *testing.T) {
 
 	diskFailures := make(map[int]string)
 	// key = disk index, value = part name with hash mismatch
-	diskFailures[0] = "part.3"
+	diskFailures[0] = "part.1"
 	diskFailures[3] = "part.1"
-	diskFailures[15] = "part.2"
+	diskFailures[15] = "part.1"
 
 	for diskIndex, partName := range diskFailures {
 		for _, info := range partsMetadata[diskIndex].Erasure.Checksums {
@@ -354,5 +354,4 @@ func TestDisksWithAllParts(t *testing.T) {
 
 		}
 	}
-
 }

cmd/xl-v1-healing.go

@@ -341,7 +341,7 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 		}
 
 		// List and delete the object directory,
-		files, derr := disk.ListDir(bucket, object, -1)
+		files, derr := disk.ListDir(bucket, object, -1, "")
 		if derr == nil {
 			for _, entry := range files {
 				_ = disk.DeleteFile(bucket,
@@ -425,6 +425,10 @@ func (xl xlObjects) healObject(ctx context.Context, bucket string, object string
 		}
 	}
 
+	// Cleanup in case of xl.json writing failure
+	writeQuorum := latestMeta.Erasure.DataBlocks + 1
+	defer xl.deleteObject(ctx, minioMetaTmpBucket, tmpID, writeQuorum, false)
+
 	// Generate and write `xl.json` generated from other disks.
 	outDatedDisks, aErr := writeUniqueXLMetadata(ctx, outDatedDisks, minioMetaTmpBucket, tmpID,
 		partsMetadata, diskCount(outDatedDisks))
@@ -478,52 +482,51 @@ func (xl xlObjects) healObjectDir(ctx context.Context, bucket, object string, dr
 	hr.Before.Drives = make([]madmin.HealDriveInfo, len(storageDisks))
 	hr.After.Drives = make([]madmin.HealDriveInfo, len(storageDisks))
 
-	var wg sync.WaitGroup
-
-	// Prepare object creation in all disks
-	for i, d := range storageDisks {
-		wg.Add(1)
-		go func(idx int, disk StorageAPI) {
-			defer wg.Done()
-			if disk == nil {
-				hr.Before.Drives[idx] = madmin.HealDriveInfo{State: madmin.DriveStateOffline}
-				hr.After.Drives[idx] = madmin.HealDriveInfo{State: madmin.DriveStateOffline}
-				return
+	errs := statAllDirs(ctx, storageDisks, bucket, object)
+	if isObjectDirDangling(errs) {
+		for i, err := range errs {
+			if err == nil {
+				storageDisks[i].DeleteFile(bucket, object)
 			}
-
-			drive := disk.String()
-			hr.Before.Drives[idx] = madmin.HealDriveInfo{UUID: "", Endpoint: drive, State: madmin.DriveStateOffline}
-			hr.After.Drives[idx] = madmin.HealDriveInfo{UUID: "", Endpoint: drive, State: madmin.DriveStateOffline}
-
-			_, statErr := disk.StatVol(pathJoin(bucket, object))
-			switch statErr {
-			case nil:
-				hr.Before.Drives[idx].State = madmin.DriveStateOk
-				hr.After.Drives[idx].State = madmin.DriveStateOk
-				// Object is fine in this disk, nothing to be done anymore, exiting
-				return
-			case errVolumeNotFound:
-				hr.Before.Drives[idx].State = madmin.DriveStateMissing
-				hr.After.Drives[idx].State = madmin.DriveStateMissing
-			default:
-				logger.LogIf(ctx, err)
-				return
-			}
-
-			if dryRun {
-				return
-			}
-
-			if err := disk.MakeVol(pathJoin(bucket, object)); err == nil || err == errVolumeExists {
-				hr.After.Drives[idx].State = madmin.DriveStateOk
-			} else {
-				logger.LogIf(ctx, err)
-				hr.After.Drives[idx].State = madmin.DriveStateOffline
-			}
-		}(i, d)
+		}
 	}
 
-	wg.Wait()
+	// Prepare object creation in all disks
+	for i, err := range errs {
+		var drive string
+		if storageDisks[i] != nil {
+			drive = storageDisks[i].String()
+		}
+		switch err {
+		case errDiskNotFound:
+			hr.Before.Drives[i] = madmin.HealDriveInfo{State: madmin.DriveStateOffline}
+			hr.After.Drives[i] = madmin.HealDriveInfo{State: madmin.DriveStateOffline}
+		case errVolumeNotFound:
+			hr.Before.Drives[i] = madmin.HealDriveInfo{Endpoint: drive, State: madmin.DriveStateMissing}
+			hr.After.Drives[i] = madmin.HealDriveInfo{Endpoint: drive, State: madmin.DriveStateMissing}
+		default:
+			hr.Before.Drives[i] = madmin.HealDriveInfo{Endpoint: drive, State: madmin.DriveStateCorrupt}
+			hr.After.Drives[i] = madmin.HealDriveInfo{Endpoint: drive, State: madmin.DriveStateCorrupt}
+		}
+	}
+	if dryRun {
+		return hr, nil
+	}
+	for i, err := range errs {
+		switch err {
+		case errVolumeNotFound:
+			merr := storageDisks[i].MakeVol(pathJoin(bucket, object))
+			switch merr {
+			case nil, errVolumeExists:
+				hr.After.Drives[i].State = madmin.DriveStateOk
+			case errDiskNotFound:
+				hr.After.Drives[i].State = madmin.DriveStateOffline
+			default:
+				logger.LogIf(ctx, merr)
+				hr.After.Drives[i].State = madmin.DriveStateCorrupt
+			}
+		}
+	}
 	return hr, nil
 }
 
@@ -587,6 +590,46 @@ func defaultHealResult(latestXLMeta xlMetaV1, storageDisks []StorageAPI, errs []
 	return result
 }
 
+// Stat all directories.
+func statAllDirs(ctx context.Context, storageDisks []StorageAPI, bucket, prefix string) []error {
+	var errs = make([]error, len(storageDisks))
+	var wg sync.WaitGroup
+	for index, disk := range storageDisks {
+		if disk == nil {
+			continue
+		}
+		wg.Add(1)
+		go func(index int, disk StorageAPI) {
+			defer wg.Done()
+			entries, err := disk.ListDir(bucket, prefix, 1, "")
+			if err != nil {
+				errs[index] = err
+				return
+			}
+			if len(entries) > 0 {
+				errs[index] = errVolumeNotEmpty
+				return
+			}
+		}(index, disk)
+	}
+
+	wg.Wait()
+	return errs
+}
+
+// ObjectDir is considered dangling/corrupted if any only
+// if total disks - a combination of corrupted and missing
+// files is lesser than N/2+1 number of disks.
+func isObjectDirDangling(errs []error) (ok bool) {
+	var notFoundDir int
+	for _, readErr := range errs {
+		if readErr == errFileNotFound {
+			notFoundDir++
+		}
+	}
+	return notFoundDir > len(errs)/2
+}
+
 // 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.

cmd/xl-v1-list-objects.go

@@ -22,11 +22,10 @@ import (
 )
 
 // Returns function "listDir" of the type listDirFunc.
-// isLeaf - is used by listDir function to check if an entry is a leaf or non-leaf entry.
 // disks - used for doing disk.ListDir()
-func listDirFactory(ctx context.Context, isLeaf IsLeafFunc, disks ...StorageAPI) ListDirFunc {
+func listDirFactory(ctx context.Context, disks ...StorageAPI) ListDirFunc {
 	// Returns sorted merged entries from all the disks.
-	listDir := func(bucket, prefixDir, prefixEntry string) (mergedEntries []string, delayIsLeaf bool) {
+	listDir := func(bucket, prefixDir, prefixEntry string) (mergedEntries []string) {
 		for _, disk := range disks {
 			if disk == nil {
 				continue
@@ -34,7 +33,7 @@ func listDirFactory(ctx context.Context, isLeaf IsLeafFunc, disks ...StorageAPI)
 			var entries []string
 			var newEntries []string
 			var err error
-			entries, err = disk.ListDir(bucket, prefixDir, -1)
+			entries, err = disk.ListDir(bucket, prefixDir, -1, xlMetaJSONFile)
 			if err != nil {
 				continue
 			}
@@ -55,8 +54,7 @@ func listDirFactory(ctx context.Context, isLeaf IsLeafFunc, disks ...StorageAPI)
 				sort.Strings(mergedEntries)
 			}
 		}
-		mergedEntries, delayIsLeaf = filterListEntries(bucket, prefixDir, mergedEntries, prefixEntry, isLeaf)
-		return mergedEntries, delayIsLeaf
+		return filterMatchingPrefix(mergedEntries, prefixEntry)
 	}
 	return listDir
 }
@@ -72,10 +70,8 @@ func (xl xlObjects) listObjects(ctx context.Context, bucket, prefix, marker, del
 	walkResultCh, endWalkCh := xl.listPool.Release(listParams{bucket, recursive, marker, prefix})
 	if walkResultCh == nil {
 		endWalkCh = make(chan struct{})
-		isLeaf := xl.isObject
-		isLeafDir := xl.isObjectDir
-		listDir := listDirFactory(ctx, isLeaf, xl.getLoadBalancedDisks()...)
-		walkResultCh = startTreeWalk(ctx, bucket, prefix, marker, recursive, listDir, isLeaf, isLeafDir, endWalkCh)
+		listDir := listDirFactory(ctx, xl.getLoadBalancedDisks()...)
+		walkResultCh = startTreeWalk(ctx, bucket, prefix, marker, recursive, listDir, endWalkCh)
 	}
 
 	var objInfos []ObjectInfo
@@ -89,10 +85,6 @@ func (xl xlObjects) listObjects(ctx context.Context, bucket, prefix, marker, del
 			eof = true
 			break
 		}
-		// For any walk error return right away.
-		if walkResult.err != nil {
-			return loi, toObjectErr(walkResult.err, bucket, prefix)
-		}
 		entry := walkResult.entry
 		var objInfo ObjectInfo
 		if hasSuffix(entry, slashSeparator) {
@@ -108,8 +100,10 @@ func (xl xlObjects) listObjects(ctx context.Context, bucket, prefix, marker, del
 				// Ignore errFileNotFound as the object might have got
 				// deleted in the interim period of listing and getObjectInfo(),
 				// ignore quorum error as it might be an entry from an outdated disk.
-				switch err {
-				case errFileNotFound, errXLReadQuorum:
+				if IsErrIgnored(err, []error{
+					errFileNotFound,
+					errXLReadQuorum,
+				}...) {
 					continue
 				}
 				return loi, toObjectErr(err, bucket, prefix)

cmd/xl-v1-metadata.go

@@ -1,5 +1,5 @@
 /*
- * MinIO Cloud Storage, (C) 2016, 2017, 2017 MinIO, Inc.
+ * 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.
@@ -27,6 +27,7 @@ import (
 	"sync"
 	"time"
 
+	jsoniter "github.com/json-iterator/go"
 	"github.com/minio/minio/cmd/logger"
 )
 
@@ -74,8 +75,8 @@ func (c ChecksumInfo) MarshalJSON() ([]byte, error) {
 
 // UnmarshalJSON - should never be called, instead xlMetaV1UnmarshalJSON() should be used.
 func (c *ChecksumInfo) UnmarshalJSON(data []byte) error {
-	logger.LogIf(context.Background(), errUnexpected)
 	var info checksumInfoJSON
+	var json = jsoniter.ConfigCompatibleWithStandardLibrary
 	if err := json.Unmarshal(data, &info); err != nil {
 		return err
 	}
@@ -424,14 +425,6 @@ func (xl xlObjects) readXLMetaStat(ctx context.Context, bucket, object string) (
 	return statInfo{}, nil, reduceReadQuorumErrs(ctx, ignoredErrs, nil, readQuorum)
 }
 
-// deleteXLMetadata - deletes `xl.json` on a single disk.
-func deleteXLMetdata(ctx context.Context, disk StorageAPI, bucket, prefix string) error {
-	jsonFile := path.Join(prefix, xlMetaJSONFile)
-	err := disk.DeleteFile(bucket, jsonFile)
-	logger.LogIf(ctx, err)
-	return err
-}
-
 // writeXLMetadata - writes `xl.json` to a single disk.
 func writeXLMetadata(ctx context.Context, disk StorageAPI, bucket, prefix string, xlMeta xlMetaV1) error {
 	jsonFile := path.Join(prefix, xlMetaJSONFile)
@@ -449,27 +442,6 @@ func writeXLMetadata(ctx context.Context, disk StorageAPI, bucket, prefix string
 	return err
 }
 
-// deleteAllXLMetadata - deletes all partially written `xl.json` depending on errs.
-func deleteAllXLMetadata(ctx context.Context, disks []StorageAPI, bucket, prefix string, errs []error) {
-	var wg = &sync.WaitGroup{}
-	// Delete all the `xl.json` left over.
-	for index, disk := range disks {
-		if disk == nil {
-			continue
-		}
-		// Undo rename object in parallel.
-		wg.Add(1)
-		go func(index int, disk StorageAPI) {
-			defer wg.Done()
-			if errs[index] != nil {
-				return
-			}
-			_ = deleteXLMetdata(ctx, disk, bucket, prefix)
-		}(index, disk)
-	}
-	wg.Wait()
-}
-
 // Rename `xl.json` content to destination location for each disk in order.
 func renameXLMetadata(ctx context.Context, disks []StorageAPI, srcBucket, srcEntry, dstBucket, dstEntry string, quorum int) ([]StorageAPI, error) {
 	isDir := false
@@ -509,10 +481,6 @@ func writeUniqueXLMetadata(ctx context.Context, disks []StorageAPI, bucket, pref
 	wg.Wait()
 
 	err := reduceWriteQuorumErrs(ctx, mErrs, objectOpIgnoredErrs, quorum)
-	if err == errXLWriteQuorum {
-		// Delete all `xl.json` successfully renamed.
-		deleteAllXLMetadata(ctx, disks, bucket, prefix, mErrs)
-	}
 	return evalDisks(disks, mErrs), err
 }
 
@@ -547,9 +515,5 @@ func writeSameXLMetadata(ctx context.Context, disks []StorageAPI, bucket, prefix
 	wg.Wait()
 
 	err := reduceWriteQuorumErrs(ctx, mErrs, objectOpIgnoredErrs, writeQuorum)
-	if err == errXLWriteQuorum {
-		// Delete all `xl.json` successfully renamed.
-		deleteAllXLMetadata(ctx, disks, bucket, prefix, mErrs)
-	}
 	return evalDisks(disks, mErrs), err
 }

cmd/xl-v1-multipart.go

@@ -47,9 +47,10 @@ func (xl xlObjects) getMultipartSHADir(bucket, object string) string {
 	return getSHA256Hash([]byte(pathJoin(bucket, object)))
 }
 
-// isUploadIDExists - verify if a given uploadID exists and is valid.
-func (xl xlObjects) isUploadIDExists(ctx context.Context, bucket, object, uploadID string) bool {
-	return xl.isObject(minioMetaMultipartBucket, xl.getUploadIDDir(bucket, object, uploadID))
+// checkUploadIDExists - verify if a given uploadID exists and is valid.
+func (xl xlObjects) checkUploadIDExists(ctx context.Context, bucket, object, uploadID string) error {
+	_, err := xl.getObjectInfo(ctx, minioMetaMultipartBucket, xl.getUploadIDDir(bucket, object, uploadID))
+	return err
 }
 
 // Removes part given by partName belonging to a mulitpart upload from minioMetaBucket
@@ -135,10 +136,6 @@ func commitXLMetadata(ctx context.Context, disks []StorageAPI, srcBucket, srcPre
 	wg.Wait()
 
 	err := reduceWriteQuorumErrs(ctx, mErrs, objectOpIgnoredErrs, quorum)
-	if err == errXLWriteQuorum {
-		// Delete all `xl.json` successfully renamed.
-		deleteAllXLMetadata(ctx, disks, dstBucket, dstPrefix, mErrs)
-	}
 	return evalDisks(disks, mErrs), err
 }
 
@@ -163,7 +160,7 @@ func (xl xlObjects) ListMultipartUploads(ctx context.Context, bucket, object, ke
 		if disk == nil {
 			continue
 		}
-		uploadIDs, err := disk.ListDir(minioMetaMultipartBucket, xl.getMultipartSHADir(bucket, object), -1)
+		uploadIDs, err := disk.ListDir(minioMetaMultipartBucket, xl.getMultipartSHADir(bucket, object), -1, "")
 		if err != nil {
 			if err == errFileNotFound {
 				return result, nil
@@ -216,15 +213,16 @@ func (xl xlObjects) newMultipartUpload(ctx context.Context, bucket string, objec
 	uploadIDPath := xl.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)
+
 	// Write updated `xl.json` to all disks.
 	disks, err := writeSameXLMetadata(ctx, xl.getDisks(), minioMetaTmpBucket, tempUploadIDPath, xlMeta, writeQuorum)
 	if err != nil {
 		return "", toObjectErr(err, minioMetaTmpBucket, tempUploadIDPath)
 	}
-	// delete the tmp path later in case we fail to rename (ignore
-	// returned errors) - this will be a no-op in case of a rename
-	// success.
-	defer xl.deleteObject(ctx, minioMetaTmpBucket, tempUploadIDPath, writeQuorum, false)
 
 	// Attempt to rename temp upload object to actual upload path object
 	_, rErr := rename(ctx, disks, minioMetaTmpBucket, tempUploadIDPath, minioMetaMultipartBucket, uploadIDPath, true, writeQuorum, nil)
@@ -308,9 +306,9 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	}
 
 	// Validates if upload ID exists.
-	if !xl.isUploadIDExists(ctx, bucket, object, uploadID) {
+	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
 		preUploadIDLock.RUnlock()
-		return pi, InvalidUploadID{UploadID: uploadID}
+		return pi, toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Read metadata associated with the object from all disks.
@@ -406,9 +404,9 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	}
 	defer postUploadIDLock.Unlock()
 
-	// Validate again if upload ID still exists.
-	if !xl.isUploadIDExists(ctx, bucket, object, uploadID) {
-		return pi, InvalidUploadID{UploadID: uploadID}
+	// Validates if upload ID exists.
+	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+		return pi, toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Rename temporary part file to its final location.
@@ -452,8 +450,10 @@ func (xl xlObjects) PutObjectPart(ctx context.Context, bucket, object, uploadID
 	}
 
 	// Write all the checksum metadata.
-	newUUID := mustGetUUID()
-	tempXLMetaPath := newUUID
+	tempXLMetaPath := mustGetUUID()
+
+	// 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.
 	if onlineDisks, err = writeUniqueXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempXLMetaPath, partsMetadata, writeQuorum); err != nil {
@@ -499,8 +499,8 @@ func (xl xlObjects) ListObjectParts(ctx context.Context, bucket, object, uploadI
 	}
 	defer uploadIDLock.Unlock()
 
-	if !xl.isUploadIDExists(ctx, bucket, object, uploadID) {
-		return result, InvalidUploadID{UploadID: uploadID}
+	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+		return result, toObjectErr(err, bucket, object, uploadID)
 	}
 
 	uploadIDPath := xl.getUploadIDDir(bucket, object, uploadID)
@@ -617,8 +617,8 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 	}
 	defer uploadIDLock.Unlock()
 
-	if !xl.isUploadIDExists(ctx, bucket, object, uploadID) {
-		return oi, InvalidUploadID{UploadID: uploadID}
+	if err := xl.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.
@@ -628,10 +628,7 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 	}
 
 	// Calculate s3 compatible md5sum for complete multipart.
-	s3MD5, err := getCompleteMultipartMD5(ctx, parts)
-	if err != nil {
-		return oi, err
-	}
+	s3MD5 := getCompleteMultipartMD5(parts)
 
 	// Read metadata associated with the object from all disks.
 	partsMetadata, errs := readAllXLMetadata(ctx, xl.getDisks(), minioMetaMultipartBucket, uploadIDPath)
@@ -731,8 +728,6 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 	// Save the consolidated actual size.
 	xlMeta.Meta[ReservedMetadataPrefix+"actual-size"] = strconv.FormatInt(objectActualSize, 10)
 
-	tempUploadIDPath := uploadID
-
 	// Update all xl metadata, make sure to not modify fields like
 	// checksum which are different on each disks.
 	for index := range partsMetadata {
@@ -741,13 +736,18 @@ func (xl xlObjects) CompleteMultipartUpload(ctx context.Context, bucket string,
 		partsMetadata[index].Parts = xlMeta.Parts
 	}
 
+	tempXLMetaPath := mustGetUUID()
+
+	// Cleanup in case of failure
+	defer xl.deleteObject(ctx, minioMetaTmpBucket, tempXLMetaPath, writeQuorum, false)
+
 	// Write unique `xl.json` for each disk.
-	if onlineDisks, err = writeUniqueXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempUploadIDPath, partsMetadata, writeQuorum); err != nil {
-		return oi, toObjectErr(err, minioMetaTmpBucket, tempUploadIDPath)
+	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, tempUploadIDPath, minioMetaMultipartBucket, uploadIDPath, writeQuorum)
+	onlineDisks, rErr = commitXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempXLMetaPath, minioMetaMultipartBucket, uploadIDPath, writeQuorum)
 	if rErr != nil {
 		return oi, toObjectErr(rErr, minioMetaMultipartBucket, uploadIDPath)
 	}
@@ -821,8 +821,9 @@ func (xl xlObjects) AbortMultipartUpload(ctx context.Context, bucket, object, up
 	}
 	defer uploadIDLock.Unlock()
 
-	if !xl.isUploadIDExists(ctx, bucket, object, uploadID) {
-		return InvalidUploadID{UploadID: uploadID}
+	// Validates if upload ID exists.
+	if err := xl.checkUploadIDExists(ctx, bucket, object, uploadID); err != nil {
+		return toObjectErr(err, bucket, object, uploadID)
 	}
 
 	// Read metadata associated with the object from all disks.
@@ -871,12 +872,12 @@ func (xl xlObjects) cleanupStaleMultipartUploads(ctx context.Context, cleanupInt
 // 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)
+	shaDirs, err := disk.ListDir(minioMetaMultipartBucket, "", -1, "")
 	if err != nil {
 		return
 	}
 	for _, shaDir := range shaDirs {
-		uploadIDDirs, err := disk.ListDir(minioMetaMultipartBucket, shaDir, -1)
+		uploadIDDirs, err := disk.ListDir(minioMetaMultipartBucket, shaDir, -1, "")
 		if err != nil {
 			continue
 		}

cmd/xl-v1-object.go

@@ -22,11 +22,9 @@ import (
 	"io"
 	"net/http"
 	"path"
-	"strconv"
 	"sync"
 
 	"github.com/minio/minio/cmd/logger"
-	"github.com/minio/minio/pkg/hash"
 	"github.com/minio/minio/pkg/mimedb"
 )
 
@@ -64,35 +62,32 @@ func (xl xlObjects) putObjectDir(ctx context.Context, bucket, object string, wri
 func (xl xlObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBucket, dstObject string, srcInfo ObjectInfo, srcOpts, dstOpts ObjectOptions) (oi ObjectInfo, e error) {
 	cpSrcDstSame := isStringEqual(pathJoin(srcBucket, srcObject), pathJoin(dstBucket, dstObject))
 
-	// Read metadata associated with the object from all disks.
-	storageDisks := xl.getDisks()
-
-	metaArr, errs := readAllXLMetadata(ctx, storageDisks, srcBucket, srcObject)
-
-	// get Quorum for this object
-	readQuorum, writeQuorum, err := objectQuorumFromMeta(ctx, xl, metaArr, errs)
-	if err != nil {
-		return oi, toObjectErr(err, srcBucket, srcObject)
-	}
-
-	if reducedErr := reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum); reducedErr != nil {
-		return oi, toObjectErr(reducedErr, srcBucket, srcObject)
-	}
-
-	// List all online disks.
-	_, modTime := listOnlineDisks(storageDisks, metaArr, errs)
-
-	// Pick latest valid metadata.
-	xlMeta, err := pickValidXLMeta(ctx, metaArr, modTime, readQuorum)
-	if err != nil {
-		return oi, toObjectErr(err, srcBucket, srcObject)
-	}
-
-	// Length of the file to read.
-	length := xlMeta.Stat.Size
-
 	// Check if this request is only metadata update.
 	if cpSrcDstSame {
+		// Read metadata associated with the object from all disks.
+		storageDisks := xl.getDisks()
+
+		metaArr, errs := readAllXLMetadata(ctx, storageDisks, srcBucket, srcObject)
+
+		// get Quorum for this object
+		readQuorum, writeQuorum, err := objectQuorumFromMeta(ctx, xl, metaArr, errs)
+		if err != nil {
+			return oi, toObjectErr(err, srcBucket, srcObject)
+		}
+
+		if reducedErr := reduceReadQuorumErrs(ctx, errs, objectOpIgnoredErrs, readQuorum); reducedErr != nil {
+			return oi, toObjectErr(reducedErr, srcBucket, srcObject)
+		}
+
+		// List all online disks.
+		_, modTime := listOnlineDisks(storageDisks, metaArr, errs)
+
+		// Pick latest valid metadata.
+		xlMeta, err := pickValidXLMeta(ctx, metaArr, modTime, readQuorum)
+		if err != nil {
+			return oi, toObjectErr(err, srcBucket, srcObject)
+		}
+
 		// Update `xl.json` content on each disks.
 		for index := range metaArr {
 			metaArr[index].Meta = srcInfo.UserDefined
@@ -103,44 +98,24 @@ func (xl xlObjects) CopyObject(ctx context.Context, srcBucket, srcObject, dstBuc
 
 		tempObj := mustGetUUID()
 
+		// Cleanup in case of xl.json writing failure
+		defer xl.deleteObject(ctx, minioMetaTmpBucket, tempObj, writeQuorum, false)
+
 		// Write unique `xl.json` for each disk.
 		if onlineDisks, err = writeUniqueXLMetadata(ctx, storageDisks, minioMetaTmpBucket, tempObj, metaArr, writeQuorum); err != nil {
 			return oi, toObjectErr(err, srcBucket, srcObject)
 		}
+
 		// Rename atomically `xl.json` from tmp location to destination for each disk.
 		if _, err = renameXLMetadata(ctx, onlineDisks, minioMetaTmpBucket, tempObj, srcBucket, srcObject, writeQuorum); err != nil {
 			return oi, toObjectErr(err, srcBucket, srcObject)
 		}
+
 		return xlMeta.ToObjectInfo(srcBucket, srcObject), nil
 	}
 
-	// Initialize pipe.
-	pipeReader, pipeWriter := io.Pipe()
-
-	go func() {
-		var startOffset int64 // Read the whole file.
-		if gerr := xl.getObject(ctx, srcBucket, srcObject, startOffset, length, pipeWriter, srcInfo.ETag, srcOpts); gerr != nil {
-			pipeWriter.CloseWithError(toObjectErr(gerr, srcBucket, srcObject))
-			return
-		}
-		pipeWriter.Close() // Close writer explicitly signaling we wrote all data.
-	}()
-
-	hashReader, err := hash.NewReader(pipeReader, length, "", "", length)
-	if err != nil {
-		logger.LogIf(ctx, err)
-		return oi, toObjectErr(err, dstBucket, dstObject)
-	}
-	putOpts := ObjectOptions{UserDefined: srcInfo.UserDefined, ServerSideEncryption: dstOpts.ServerSideEncryption}
-	objInfo, err := xl.putObject(ctx, dstBucket, dstObject, NewPutObjReader(hashReader, nil, nil), putOpts)
-	if err != nil {
-		return oi, toObjectErr(err, dstBucket, dstObject)
-	}
-
-	// Explicitly close the reader.
-	pipeReader.Close()
-
-	return objInfo, nil
+	putOpts := ObjectOptions{ServerSideEncryption: dstOpts.ServerSideEncryption, UserDefined: srcInfo.UserDefined}
+	return xl.PutObject(ctx, dstBucket, dstObject, srcInfo.PutObjReader, putOpts)
 }
 
 // GetObjectNInfo - returns object info and an object
@@ -173,10 +148,6 @@ func (xl xlObjects) GetObjectNInfo(ctx context.Context, bucket, object string, r
 	// Handler directory request by returning a reader that
 	// returns no bytes.
 	if hasSuffix(object, slashSeparator) {
-		if !xl.isObjectDir(bucket, object) {
-			nsUnlocker()
-			return nil, toObjectErr(errFileNotFound, bucket, object)
-		}
 		var objInfo ObjectInfo
 		if objInfo, err = xl.getObjectInfoDir(ctx, bucket, object); err != nil {
 			nsUnlocker()
@@ -375,19 +346,16 @@ func (xl xlObjects) getObjectInfoDir(ctx context.Context, bucket, object string)
 		wg.Add(1)
 		go func(index int, disk StorageAPI) {
 			defer wg.Done()
-			if _, err := disk.StatVol(pathJoin(bucket, object)); err != nil {
-				// Since we are re-purposing StatVol, an object which
-				// is a directory if it doesn't exist should be
-				// returned as errFileNotFound instead, convert
-				// the error right here accordingly.
-				if err == errVolumeNotFound {
-					err = errFileNotFound
-				} else if err == errVolumeAccessDenied {
-					err = errFileAccessDenied
-				}
-
-				// Save error to reduce it later
+			// Check if 'prefix' is an object on this 'disk'.
+			entries, err := disk.ListDir(bucket, object, 1, "")
+			if err != nil {
 				errs[index] = err
+				return
+			}
+			if len(entries) > 0 {
+				// Not a directory if not empty.
+				errs[index] = errFileNotFound
+				return
 			}
 		}(index, disk)
 	}
@@ -412,13 +380,11 @@ func (xl xlObjects) GetObjectInfo(ctx context.Context, bucket, object string, op
 	}
 
 	if hasSuffix(object, slashSeparator) {
-		if !xl.isObjectDir(bucket, object) {
-			return oi, toObjectErr(errFileNotFound, bucket, object)
+		info, err := xl.getObjectInfoDir(ctx, bucket, object)
+		if err != nil {
+			return oi, toObjectErr(err, bucket, object)
 		}
-		if oi, e = xl.getObjectInfoDir(ctx, bucket, object); e != nil {
-			return oi, toObjectErr(e, bucket, object)
-		}
-		return oi, nil
+		return info, nil
 	}
 
 	info, err := xl.getObjectInfo(ctx, bucket, object)
@@ -543,6 +509,7 @@ func (xl xlObjects) PutObject(ctx context.Context, bucket string, object string,
 		return objInfo, err
 	}
 	defer objectLock.Unlock()
+
 	return xl.putObject(ctx, bucket, object, data, opts)
 }
 
@@ -593,11 +560,6 @@ func (xl xlObjects) putObject(ctx context.Context, bucket string, object string,
 		return dirObjectInfo(bucket, object, data.Size(), opts.UserDefined), nil
 	}
 
-	// Validate put object input args.
-	if err = checkPutObjectArgs(ctx, bucket, object, xl, data.Size()); err != nil {
-		return ObjectInfo{}, err
-	}
-
 	// Validate input data size and it can never be less than zero.
 	if data.Size() < -1 {
 		logger.LogIf(ctx, errInvalidArgument)
@@ -611,9 +573,6 @@ func (xl xlObjects) putObject(ctx context.Context, bucket string, object string,
 		return ObjectInfo{}, toObjectErr(errFileParentIsFile, bucket, object)
 	}
 
-	// Limit the reader to its provided size if specified.
-	var reader io.Reader = data
-
 	// Initialize parts metadata
 	partsMetadata := make([]xlMetaV1, len(xl.getDisks()))
 
@@ -625,10 +584,7 @@ func (xl xlObjects) putObject(ctx context.Context, bucket string, object string,
 	}
 
 	// Order disks according to erasure distribution
-	onlineDisks := shuffleDisks(xl.getDisks(), partsMetadata[0].Erasure.Distribution)
-
-	// Total size of the written object
-	var sizeWritten int64
+	onlineDisks := shuffleDisks(xl.getDisks(), xlMeta.Erasure.Distribution)
 
 	erasure, err := NewErasure(ctx, xlMeta.Erasure.DataBlocks, xlMeta.Erasure.ParityBlocks, xlMeta.Erasure.BlockSize)
 	if err != nil {
@@ -652,82 +608,37 @@ func (xl xlObjects) putObject(ctx context.Context, bucket string, object string,
 		buffer = buffer[:xlMeta.Erasure.BlockSize]
 	}
 
-	// Read data and split into parts - similar to multipart mechanism
-	for partIdx := 1; ; partIdx++ {
-		// Compute part name
-		partName := "part." + strconv.Itoa(partIdx)
-		// Compute the path of current part
-		tempErasureObj := pathJoin(uniqueID, partName)
+	partName := "part.1"
+	tempErasureObj := pathJoin(uniqueID, partName)
 
-		// Calculate the size of the current part.
-		var curPartSize int64
-		curPartSize, err = calculatePartSizeFromIdx(ctx, data.Size(), globalPutPartSize, partIdx)
-		if err != nil {
-			logger.LogIf(ctx, err)
-			return ObjectInfo{}, toObjectErr(err, bucket, object)
+	writers := make([]io.Writer, len(onlineDisks))
+	for i, disk := range onlineDisks {
+		if disk == nil {
+			continue
 		}
+		writers[i] = newBitrotWriter(disk, minioMetaTmpBucket, tempErasureObj, erasure.ShardFileSize(data.Size()), DefaultBitrotAlgorithm, erasure.ShardSize())
+	}
 
-		// Hint the filesystem to pre-allocate one continuous large block.
-		// This is only an optimization.
-		var curPartReader io.Reader
+	n, erasureErr := erasure.Encode(ctx, data, writers, buffer, erasure.dataBlocks+1)
+	closeBitrotWriters(writers)
+	if erasureErr != nil {
+		return ObjectInfo{}, toObjectErr(erasureErr, minioMetaTmpBucket, tempErasureObj)
+	}
 
-		if curPartSize < data.Size() {
-			curPartReader = io.LimitReader(reader, curPartSize)
-		} else {
-			curPartReader = reader
-		}
-
-		writers := make([]io.Writer, len(onlineDisks))
-		for i, disk := range onlineDisks {
-			if disk == nil {
-				continue
-			}
-			writers[i] = newBitrotWriter(disk, minioMetaTmpBucket, tempErasureObj, erasure.ShardFileSize(curPartSize), DefaultBitrotAlgorithm, erasure.ShardSize())
-		}
-
-		n, erasureErr := erasure.Encode(ctx, curPartReader, writers, buffer, erasure.dataBlocks+1)
-		// Note: we should not be defer'ing the following closeBitrotWriters() call as we are inside a for loop i.e if we use defer, we would accumulate a lot of open files by the time
-		// we return from this function.
-		closeBitrotWriters(writers)
-		if erasureErr != nil {
-			return ObjectInfo{}, toObjectErr(erasureErr, minioMetaTmpBucket, tempErasureObj)
-		}
-
-		// Should return IncompleteBody{} error when reader has fewer bytes
-		// than specified in request header.
-
-		if n < curPartSize && data.Size() > 0 {
-			logger.LogIf(ctx, IncompleteBody{})
-			return ObjectInfo{}, IncompleteBody{}
-		}
-
-		if n == 0 && data.Size() == -1 {
-			// The last part of a compressed object will always be empty
-			// Since the compressed size is unpredictable.
-			// Hence removing the last (empty) part from all `xl.disks`.
-			dErr := xl.deleteObject(ctx, minioMetaTmpBucket, tempErasureObj, writeQuorum, true)
-			if dErr != nil {
-				return ObjectInfo{}, toObjectErr(dErr, minioMetaTmpBucket, tempErasureObj)
-			}
-			break
-		}
-
-		// Update the total written size
-		sizeWritten += n
-
-		for i, w := range writers {
-			if w == nil {
-				onlineDisks[i] = nil
-				continue
-			}
-			partsMetadata[i].AddObjectPart(partIdx, partName, "", n, data.ActualSize())
-			partsMetadata[i].Erasure.AddChecksumInfo(ChecksumInfo{partName, DefaultBitrotAlgorithm, bitrotWriterSum(w)})
-		}
-
-		// We wrote everything, break out.
-		if sizeWritten == data.Size() {
-			break
+	// Should return IncompleteBody{} error when reader has fewer bytes
+	// than specified in request header.
+	if n < data.Size() {
+		logger.LogIf(ctx, IncompleteBody{})
+		return ObjectInfo{}, IncompleteBody{}
+	}
+
+	for i, w := range writers {
+		if w == nil {
+			onlineDisks[i] = nil
+			continue
 		}
+		partsMetadata[i].AddObjectPart(1, partName, "", n, data.ActualSize())
+		partsMetadata[i].Erasure.AddChecksumInfo(ChecksumInfo{partName, DefaultBitrotAlgorithm, bitrotWriterSum(w)})
 	}
 
 	// Save additional erasureMetadata.
@@ -765,7 +676,7 @@ func (xl xlObjects) putObject(ctx context.Context, bucket string, object string,
 	// Update `xl.json` content on each disks.
 	for index := range partsMetadata {
 		partsMetadata[index].Meta = opts.UserDefined
-		partsMetadata[index].Stat.Size = sizeWritten
+		partsMetadata[index].Stat.Size = n
 		partsMetadata[index].Stat.ModTime = modTime
 	}
 
@@ -795,7 +706,6 @@ func (xl xlObjects) putObject(ctx context.Context, bucket string, object string,
 		UserDefined:     xlMeta.Meta,
 	}
 
-	// Success, return object info.
 	return objInfo, nil
 }
 
@@ -861,6 +771,191 @@ func (xl xlObjects) deleteObject(ctx context.Context, bucket, object string, wri
 	return reduceWriteQuorumErrs(ctx, dErrs, objectOpIgnoredErrs, writeQuorum)
 }
 
+// deleteObject - wrapper for delete object, deletes an object from
+// all the disks in parallel, including `xl.json` associated with the
+// object.
+func (xl xlObjects) doDeleteObjects(ctx context.Context, bucket string, objects []string, errs []error, writeQuorums []int, isDirs []bool) ([]error, error) {
+	var tmpObjs = make([]string, len(objects))
+	var disks = xl.getDisks()
+
+	if bucket == minioMetaTmpBucket {
+		copy(tmpObjs, objects)
+	} else {
+		for i, object := range objects {
+			if errs[i] != nil {
+				continue
+			}
+			var err error
+			tmpObjs[i] = mustGetUUID()
+			// Rename the current object while requiring write quorum, but also consider
+			// that a non found object in a given disk as a success since it already
+			// confirms that the object doesn't have a part in that disk (already removed)
+			if isDirs[i] {
+				disks, err = rename(ctx, xl.getDisks(), bucket, object, minioMetaTmpBucket, tmpObjs[i], true, writeQuorums[i],
+					[]error{errFileNotFound, errFileAccessDenied})
+			} else {
+				disks, err = rename(ctx, xl.getDisks(), bucket, object, minioMetaTmpBucket, tmpObjs[i], true, writeQuorums[i],
+					[]error{errFileNotFound})
+			}
+			if err != nil {
+				errs[i] = err
+			}
+		}
+	}
+
+	// Initialize list of errors.
+	var opErrs = make([]error, len(disks))
+	var delObjErrs = make([][]error, len(disks))
+
+	// Remove objects in bulk for each disk
+	for index, disk := range disks {
+		if disk == nil {
+			opErrs[index] = errDiskNotFound
+			continue
+		}
+		delObjErrs[index], opErrs[index] = cleanupObjectsBulk(ctx, disk, minioMetaTmpBucket, tmpObjs, errs)
+		if opErrs[index] == errVolumeNotFound {
+			opErrs[index] = nil
+		}
+	}
+
+	// Return errors if any during deletion
+	if err := reduceWriteQuorumErrs(ctx, opErrs, objectOpIgnoredErrs, len(xl.getDisks())/2+1); err != nil {
+		return nil, err
+	}
+
+	// Reduce errors for each object
+	for objIndex := range objects {
+		if errs[objIndex] != nil {
+			continue
+		}
+		listErrs := make([]error, len(xl.getDisks()))
+		for i := range delObjErrs {
+			if delObjErrs[i] != nil {
+				listErrs[i] = delObjErrs[i][objIndex]
+			}
+		}
+		errs[objIndex] = reduceWriteQuorumErrs(ctx, listErrs, objectOpIgnoredErrs, writeQuorums[objIndex])
+	}
+
+	return errs, nil
+}
+
+func (xl xlObjects) deleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+	errs := make([]error, len(objects))
+	writeQuorums := make([]int, len(objects))
+	isObjectDirs := make([]bool, len(objects))
+
+	for i, object := range objects {
+		errs[i] = checkDelObjArgs(ctx, bucket, object)
+	}
+
+	var objectLocks = make([]RWLocker, len(objects))
+
+	for i, object := range objects {
+		if errs[i] != nil {
+			continue
+		}
+		// Acquire a write lock before deleting the object.
+		objectLocks[i] = xl.nsMutex.NewNSLock(bucket, object)
+		if errs[i] = objectLocks[i].GetLock(globalOperationTimeout); errs[i] != nil {
+			continue
+		}
+		defer objectLocks[i].Unlock()
+	}
+
+	for i, object := range objects {
+		isObjectDirs[i] = hasSuffix(object, slashSeparator)
+	}
+
+	for i, object := range objects {
+		if isObjectDirs[i] {
+			_, err := xl.getObjectInfoDir(ctx, bucket, object)
+			if err == errXLReadQuorum {
+				if isObjectDirDangling(statAllDirs(ctx, xl.getDisks(), bucket, object)) {
+					// If object is indeed dangling, purge it.
+					errs[i] = nil
+				}
+			}
+			if err != nil {
+				errs[i] = toObjectErr(err, bucket, object)
+				continue
+			}
+		}
+	}
+
+	for i, object := range objects {
+		if errs[i] != nil {
+			continue
+		}
+		if isObjectDirs[i] {
+			writeQuorums[i] = len(xl.getDisks())/2 + 1
+		} else {
+			var err error
+			// Read metadata associated with the object from all disks.
+			partsMetadata, readXLErrs := readAllXLMetadata(ctx, xl.getDisks(), bucket, object)
+			// get Quorum for this object
+			_, writeQuorums[i], err = objectQuorumFromMeta(ctx, xl, partsMetadata, readXLErrs)
+			if err != nil {
+				errs[i] = toObjectErr(err, bucket, object)
+				continue
+			}
+		}
+	}
+
+	return xl.doDeleteObjects(ctx, bucket, objects, errs, writeQuorums, isObjectDirs)
+}
+
+// DeleteObjects deletes objects in bulk, this function will still automatically split objects list
+// into smaller bulks if some object names are found to be duplicated in the delete list, splitting
+// into smaller bulks will avoid holding twice the write lock of the duplicated object names.
+func (xl xlObjects) DeleteObjects(ctx context.Context, bucket string, objects []string) ([]error, error) {
+
+	var (
+		i, start, end int
+		// Deletion result for all objects
+		deleteErrs []error
+		// Object names store will be used to check for object name duplication
+		objectNamesStore = make(map[string]interface{})
+	)
+
+	for {
+		if i >= len(objects) {
+			break
+		}
+
+		object := objects[i]
+
+		_, duplicationFound := objectNamesStore[object]
+		if duplicationFound {
+			end = i - 1
+		} else {
+			objectNamesStore[object] = true
+			end = i
+		}
+
+		if duplicationFound || i == len(objects)-1 {
+			errs, err := xl.deleteObjects(ctx, bucket, objects[start:end+1])
+			if err != nil {
+				return nil, err
+			}
+			deleteErrs = append(deleteErrs, errs...)
+			objectNamesStore = make(map[string]interface{})
+		}
+
+		if duplicationFound {
+			// Avoid to increase the index if object
+			// name is found to be duplicated.
+			start = i
+			end = i
+		} else {
+			i++
+		}
+	}
+
+	return deleteErrs, nil
+}
+
 // DeleteObject - deletes an object, this call doesn't necessary reply
 // any error as it is not necessary for the handler to reply back a
 // response to the client request.
@@ -879,8 +974,17 @@ func (xl xlObjects) DeleteObject(ctx context.Context, bucket, object string) (er
 	var writeQuorum int
 	var isObjectDir = hasSuffix(object, slashSeparator)
 
-	if isObjectDir && !xl.isObjectDir(bucket, object) {
-		return toObjectErr(errFileNotFound, bucket, object)
+	if isObjectDir {
+		_, err = xl.getObjectInfoDir(ctx, bucket, object)
+		if err == errXLReadQuorum {
+			if isObjectDirDangling(statAllDirs(ctx, xl.getDisks(), bucket, object)) {
+				// If object is indeed dangling, purge it.
+				err = nil
+			}
+		}
+		if err != nil {
+			return toObjectErr(err, bucket, object)
+		}
 	}
 
 	if isObjectDir {

cmd/xl-v1-object_test.go

@@ -111,6 +111,80 @@ func TestXLDeleteObjectBasic(t *testing.T) {
 	// Cleanup backend directories
 	removeRoots(fsDirs)
 }
+
+func TestXLDeleteObjectsXLSet(t *testing.T) {
+
+	var objs []*xlObjects
+	for i := 0; i < 32; i++ {
+		obj, fsDirs, err := prepareXL(16)
+		if err != nil {
+			t.Fatal("Unable to initialize 'XL' object layer.", err)
+		}
+		// Remove all dirs.
+		for _, dir := range fsDirs {
+			defer os.RemoveAll(dir)
+		}
+		objs = append(objs, obj.(*xlObjects))
+	}
+
+	xlSets := &xlSets{sets: objs, distributionAlgo: "CRCMOD"}
+
+	type testCaseType struct {
+		bucket string
+		object string
+	}
+
+	bucketName := "bucket"
+	testCases := []testCaseType{
+		{bucketName, "dir/obj1"},
+		{bucketName, "dir/obj2"},
+		{bucketName, "obj3"},
+		{bucketName, "obj_4"},
+	}
+
+	err := xlSets.MakeBucketWithLocation(context.Background(), bucketName, "")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	for _, testCase := range testCases {
+		_, err = xlSets.PutObject(context.Background(), 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)
+		}
+	}
+
+	toObjectNames := func(testCases []testCaseType) []string {
+		names := make([]string, len(testCases))
+		for i := range testCases {
+			names[i] = testCases[i].object
+		}
+		return names
+	}
+
+	objectNames := toObjectNames(testCases)
+	delErrs, err := xlSets.DeleteObjects(context.Background(), bucketName, objectNames)
+	if err != nil {
+		t.Errorf("Failed to call DeleteObjects with the error: `%v`", err)
+	}
+
+	for i := range delErrs {
+		if delErrs[i] != nil {
+			t.Errorf("Failed to remove object `%v` with the error: `%v`", objectNames[i], delErrs[i])
+		}
+	}
+
+	for _, test := range testCases {
+		_, statErr := xlSets.GetObjectInfo(context.Background(), test.bucket, test.object, ObjectOptions{})
+		switch statErr.(type) {
+		case ObjectNotFound:
+		default:
+			t.Fatalf("Object %s is not removed", test.bucket+"/"+test.object)
+		}
+	}
+}
+
 func TestXLDeleteObjectDiskNotFound(t *testing.T) {
 	// Reset global storage class flags
 	resetGlobalStorageEnvs()

cmd/xl-v1-utils.go

@@ -18,13 +18,13 @@ package cmd
 
 import (
 	"context"
-	"encoding/hex"
 	"errors"
 	"hash/crc32"
 	"path"
 	"sync"
 	"time"
 
+	jsoniter "github.com/json-iterator/go"
 	"github.com/minio/minio/cmd/logger"
 	"github.com/tidwall/gjson"
 )
@@ -139,59 +139,6 @@ func parseXLFormat(xlMetaBuf []byte) string {
 	return gjson.GetBytes(xlMetaBuf, "format").String()
 }
 
-func parseXLRelease(xlMetaBuf []byte) string {
-	return gjson.GetBytes(xlMetaBuf, "minio.release").String()
-}
-
-func parseXLErasureInfo(ctx context.Context, xlMetaBuf []byte) (ErasureInfo, error) {
-	erasure := ErasureInfo{}
-	erasureResult := gjson.GetBytes(xlMetaBuf, "erasure")
-	// parse the xlV1Meta.Erasure.Distribution.
-	disResult := erasureResult.Get("distribution").Array()
-
-	distribution := make([]int, len(disResult))
-	for i, dis := range disResult {
-		distribution[i] = int(dis.Int())
-	}
-	erasure.Distribution = distribution
-
-	erasure.Algorithm = erasureResult.Get("algorithm").String()
-	erasure.DataBlocks = int(erasureResult.Get("data").Int())
-	erasure.ParityBlocks = int(erasureResult.Get("parity").Int())
-	erasure.BlockSize = erasureResult.Get("blockSize").Int()
-	erasure.Index = int(erasureResult.Get("index").Int())
-
-	checkSumsResult := erasureResult.Get("checksum").Array()
-
-	// Check for scenario where checksum information missing for some parts.
-	partsResult := gjson.GetBytes(xlMetaBuf, "parts").Array()
-	if len(checkSumsResult) != len(partsResult) {
-		return erasure, errCorruptedFormat
-	}
-
-	// Parse xlMetaV1.Erasure.Checksum array.
-	checkSums := make([]ChecksumInfo, len(checkSumsResult))
-	for i, v := range checkSumsResult {
-		algorithm := BitrotAlgorithmFromString(v.Get("algorithm").String())
-		if !algorithm.Available() {
-			logger.LogIf(ctx, errBitrotHashAlgoInvalid)
-			return erasure, errBitrotHashAlgoInvalid
-		}
-		hash, err := hex.DecodeString(v.Get("hash").String())
-		if err != nil {
-			logger.LogIf(ctx, err)
-			return erasure, err
-		}
-		name := v.Get("name").String()
-		if name == "" {
-			return erasure, errCorruptedFormat
-		}
-		checkSums[i] = ChecksumInfo{Name: name, Algorithm: algorithm, Hash: hash}
-	}
-	erasure.Checksums = checkSums
-	return erasure, nil
-}
-
 func parseXLParts(xlMetaBuf []byte) []ObjectPartInfo {
 	// Parse the XL Parts.
 	partsResult := gjson.GetBytes(xlMetaBuf, "parts").Array()
@@ -220,32 +167,9 @@ func parseXLMetaMap(xlMetaBuf []byte) map[string]string {
 
 // Constructs XLMetaV1 using `gjson` lib to retrieve each field.
 func xlMetaV1UnmarshalJSON(ctx context.Context, xlMetaBuf []byte) (xlMeta xlMetaV1, e error) {
-	// obtain version.
-	xlMeta.Version = parseXLVersion(xlMetaBuf)
-	// obtain format.
-	xlMeta.Format = parseXLFormat(xlMetaBuf)
-	// Parse xlMetaV1.Stat .
-	stat, err := parseXLStat(xlMetaBuf)
-	if err != nil {
-		logger.LogIf(ctx, err)
-		return xlMeta, err
-	}
-
-	xlMeta.Stat = stat
-	// parse the xlV1Meta.Erasure fields.
-	xlMeta.Erasure, err = parseXLErasureInfo(ctx, xlMetaBuf)
-	if err != nil {
-		return xlMeta, err
-	}
-
-	// Parse the XL Parts.
-	xlMeta.Parts = parseXLParts(xlMetaBuf)
-	// Get the xlMetaV1.Realse field.
-	xlMeta.Minio.Release = parseXLRelease(xlMetaBuf)
-	// parse xlMetaV1.
-	xlMeta.Meta = parseXLMetaMap(xlMetaBuf)
-
-	return xlMeta, nil
+	var json = jsoniter.ConfigCompatibleWithStandardLibrary
+	e = json.Unmarshal(xlMetaBuf, &xlMeta)
+	return xlMeta, e
 }
 
 // read xl.json from the given disk, parse and return xlV1MetaV1.Parts.
@@ -426,6 +350,9 @@ func calculatePartSizeFromIdx(ctx context.Context, totalSize int64, partSize int
 		logger.LogIf(ctx, errPartSizeIndex)
 		return 0, errPartSizeIndex
 	}
+	if totalSize == -1 {
+		return -1, nil
+	}
 	if totalSize > 0 {
 		// Compute the total count of parts
 		partsCount := totalSize/partSize + 1

docs/bucket/notifications/README.md

@@ -2,29 +2,30 @@
 
 Events occurring on objects in a bucket can be monitored using bucket event notifications. Event types supported by MinIO server are
 
-| Supported Event Types | | |
-|:---------------------------|--------------------------------------------|-------------------------|
-| `s3:ObjectCreated:Put`     | `s3:ObjectCreated:CompleteMultipartUpload` | `s3:ObjectAccessed:Head`|
-| `s3:ObjectCreated:Post`    | `s3:ObjectRemoved:Delete`                  |
-| `s3:ObjectCreated:Copy`    | `s3:ObjectAccessed:Get`                    |
+| Supported Event Types   |                                            |                          |
+| :---------------------- | ------------------------------------------ | ------------------------ |
+| `s3:ObjectCreated:Put`  | `s3:ObjectCreated:CompleteMultipartUpload` | `s3:ObjectAccessed:Head` |
+| `s3:ObjectCreated:Post` | `s3:ObjectRemoved:Delete`                  |
+| `s3:ObjectCreated:Copy` | `s3:ObjectAccessed:Get`                    |
 
 Use client tools like `mc` to set and listen for event notifications using the [`event` sub-command](https://docs.min.io/docs/minio-client-complete-guide#events). MinIO SDK's [`BucketNotification` APIs](https://docs.min.io/docs/golang-client-api-reference#SetBucketNotification) can also be used. The notification message MinIO sends to publish an event is a JSON message with the following [structure](https://docs.aws.amazon.com/AmazonS3/latest/dev/notification-content-structure.html).
 
 Bucket events can be published to the following targets:
 
-| Supported Notification Targets | | |
-|:---------------------------|--------------------------------------------|-------------------------|
-| [`AMQP`](#AMQP) | [`Redis`](#Redis) | [`MySQL`](#MySQL) |
-| [`MQTT`](#MQTT) | [`NATS`](#NATS) | [`Apache Kafka`](#apache-kafka) |
-| [`Elasticsearch`](#Elasticsearch) | [`PostgreSQL`](#PostgreSQL) | [`Webhooks`](#webhooks) |
-| [`NSQ`](#NSQ) | | |
+| Supported Notification Targets    |                             |                                 |
+| :-------------------------------- | --------------------------- | ------------------------------- |
+| [`AMQP`](#AMQP)                   | [`Redis`](#Redis)           | [`MySQL`](#MySQL)               |
+| [`MQTT`](#MQTT)                   | [`NATS`](#NATS)             | [`Apache Kafka`](#apache-kafka) |
+| [`Elasticsearch`](#Elasticsearch) | [`PostgreSQL`](#PostgreSQL) | [`Webhooks`](#webhooks)         |
+| [`NSQ`](#NSQ)                     |                             |                                 |
 
 ## Prerequisites
 
-* Install and configure MinIO Server from [here](https://docs.min.io/docs/minio-quickstart-guide).
-* Install and configure MinIO Client from [here](https://docs.min.io/docs/minio-client-quickstart-guide).
+- Install and configure MinIO Server from [here](https://docs.min.io/docs/minio-quickstart-guide).
+- Install and configure MinIO Client from [here](https://docs.min.io/docs/minio-client-quickstart-guide).
 
 <a name="AMQP"></a>
+
 ## Publish MinIO events via AMQP
 
 Install RabbitMQ from [here](https://www.rabbitmq.com/).
@@ -33,20 +34,20 @@ Install RabbitMQ from [here](https://www.rabbitmq.com/).
 
 The MinIO server configuration file is stored on the backend in json format. The AMQP configuration is located in the `amqp` key under the `notify` top-level key. Create a configuration key-value pair here for your AMQP instance. The key is a name for your AMQP endpoint, and the value is a collection of key-value parameters described in the table below.
 
-| Parameter | Type | Description |
-|:---|:---|:---|
-| `enable` | _bool_ | (Required) Is this server endpoint configuration active/enabled? |
-| `url` | _string_ | (Required) AMQP server endpoint, e.g. `amqp://myuser:mypassword@localhost:5672` |
-| `exchange` | _string_ | Name of the exchange. |
-| `routingKey` | _string_ | Routing key for publishing. |
-| `exchangeType` | _string_ | Kind of exchange. |
-| `deliveryMode` | _uint8_ | Delivery mode for publishing. 0 or 1 - transient; 2 - persistent. |
-| `mandatory` | _bool_ | Publishing related bool. |
-| `immediate` | _bool_ | Publishing related bool. |
-| `durable` | _bool_ | Exchange declaration related bool. |
-| `internal` | _bool_ | Exchange declaration related bool. |
-| `noWait` | _bool_ | Exchange declaration related bool. |
-| `autoDeleted` | _bool_ | Exchange declaration related bool. |
+| Parameter      | Type     | Description                                                                     |
+| :------------- | :------- | :------------------------------------------------------------------------------ |
+| `enable`       | _bool_   | (Required) Is this server endpoint configuration active/enabled?                |
+| `url`          | _string_ | (Required) AMQP server endpoint, e.g. `amqp://myuser:mypassword@localhost:5672` |
+| `exchange`     | _string_ | Name of the exchange.                                                           |
+| `routingKey`   | _string_ | Routing key for publishing.                                                     |
+| `exchangeType` | _string_ | Kind of exchange.                                                               |
+| `deliveryMode` | _uint8_  | Delivery mode for publishing. 0 or 1 - transient; 2 - persistent.               |
+| `mandatory`    | _bool_   | Publishing related bool.                                                        |
+| `immediate`    | _bool_   | Publishing related bool.                                                        |
+| `durable`      | _bool_   | Exchange declaration related bool.                                              |
+| `internal`     | _bool_   | Exchange declaration related bool.                                              |
+| `noWait`       | _bool_   | Exchange declaration related bool.                                              |
+| `autoDeleted`  | _bool_   | Exchange declaration related bool.                                              |
 
 An example configuration for RabbitMQ is shown below:
 
@@ -68,22 +69,26 @@ An example configuration for RabbitMQ is shown below:
     }
 }
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the AMQP configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs:  arn:minio:sqs::1:amqp` at start-up if there were no errors.
+
+After updating the AMQP configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:amqp` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
-MinIO supports all the exchanges available in [RabbitMQ](https://www.rabbitmq.com/). For this setup, we are using ``fanout`` exchange.
+
+MinIO supports all the exchanges available in [RabbitMQ](https://www.rabbitmq.com/). For this setup, we are using `fanout` exchange.
 
 Note that, you can add as many AMQP server endpoint configurations as needed by providing an identifier (like "1" in the example above) for the AMQP instance and an object of per-server configuration parameters.
 
-
 ### Step 2: Enable bucket notification using MinIO client
 
-We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted ``images`` bucket on ``myminio`` server. Here ARN value is ``arn:minio:sqs::1:amqp``. To understand more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
+We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted `images` bucket on `myminio` server. Here ARN value is `arn:minio:sqs::1:amqp`. To understand more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
 
 ```
 mc mb myminio/images
@@ -94,7 +99,7 @@ arn:minio:sqs::1:amqp s3:ObjectCreated:*,s3:ObjectRemoved:* Filter: suffix=”.j
 
 ### Step 3: Test on RabbitMQ
 
-The python program below waits on the queue exchange ``bucketevents`` and prints event notifications on the console. We use [Pika Python Client](https://www.rabbitmq.com/tutorials/tutorial-three-python.html) library to do this.
+The python program below waits on the queue exchange `bucketevents` and prints event notifications on the console. We use [Pika Python Client](https://www.rabbitmq.com/tutorials/tutorial-three-python.html) library to do this.
 
 ```py
 #!/usr/bin/env python
@@ -131,7 +136,7 @@ Execute this example python program to watch for RabbitMQ events on the console.
 python rabbit.py
 ```
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
@@ -145,6 +150,7 @@ python rabbit.py
 ```
 
 <a name="MQTT"></a>
+
 ## Publish MinIO events MQTT
 
 Install an MQTT Broker from [here](https://mosquitto.org/).
@@ -153,17 +159,16 @@ Install an MQTT Broker from [here](https://mosquitto.org/).
 
 The MinIO server configuration file is stored on the backend in json format. The MQTT configuration is located in the `mqtt` key under the `notify` top-level key. Create a configuration key-value pair here for your MQTT instance. The key is a name for your MQTT endpoint, and the value is a collection of key-value parameters described in the table below.
 
-
-| Parameter | Type | Description |
-|:---|:---|:---|
-| `enable` | _bool_ | (Required) Is this server endpoint configuration active/enabled? |
-| `broker` | _string_ | (Required) MQTT server endpoint, e.g. `tcp://localhost:1883` |
-| `topic` | _string_ | (Required) Name of the MQTT topic to publish on, e.g. `minio` |
-| `qos` | _int_ | Set the Quality of Service Level |
-| `username` | _string_ | Username to connect to the MQTT server (if required) |
-| `password` | _string_ | Password to connect to the MQTT server (if required) |
-| `queueDir` | _string_ | Persistent store for events when MQTT broker is offline |
-| `queueLimit` | _int_ | Set the maximum event limit for the persistent store. The default limit is 10000 |
+| Parameter    | Type     | Description                                                                      |
+| :----------- | :------- | :------------------------------------------------------------------------------- |
+| `enable`     | _bool_   | (Required) Is this server endpoint configuration active/enabled?                 |
+| `broker`     | _string_ | (Required) MQTT server endpoint, e.g. `tcp://localhost:1883`                     |
+| `topic`      | _string_ | (Required) Name of the MQTT topic to publish on, e.g. `minio`                    |
+| `qos`        | _int_    | Set the Quality of Service Level                                                 |
+| `username`   | _string_ | Username to connect to the MQTT server (if required)                             |
+| `password`   | _string_ | Password to connect to the MQTT server (if required)                             |
+| `queueDir`   | _string_ | Persistent store for events when MQTT broker is offline                          |
+| `queueLimit` | _int_    | Set the maximum event limit for the persistent store. The default limit is 10000 |
 
 An example configuration for MQTT is shown below:
 
@@ -181,25 +186,28 @@ An example configuration for MQTT is shown below:
     }
 }
 ```
+
 MinIO supports persistent event store. The persistent store will backup events when the MQTT broker goes offline and replays it when the broker comes back online. The event store can be configured by setting the directory path in `queueDir` field and the maximum limit of events in the queueDir in `queueLimit` field. For eg, the `queueDir` can be `/home/events` and `queueLimit` can be `1000`. By default, the `queueLimit` is set to 10000.
 
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
+
 After updating the MQTT configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:mqtt` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
 
-MinIO supports any MQTT server that supports MQTT 3.1 or 3.1.1 and can connect to them over TCP, TLS, or a Websocket connection using ``tcp://``, ``tls://``, or ``ws://`` respectively as the scheme for the broker url. See the [Go Client](http://www.eclipse.org/paho/clients/golang/) documentation for more information.
+MinIO supports any MQTT server that supports MQTT 3.1 or 3.1.1 and can connect to them over TCP, TLS, or a Websocket connection using `tcp://`, `tls://`, or `ws://` respectively as the scheme for the broker url. See the [Go Client](http://www.eclipse.org/paho/clients/golang/) documentation for more information.
 
 Note that, you can add as many MQTT server endpoint configurations as needed by providing an identifier (like "1" in the example above) for the MQTT instance and an object of per-server configuration parameters.
 
-
 ### Step 2: Enable bucket notification using MinIO client
 
-We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted ``images`` bucket on ``myminio`` server. Here ARN value is ``arn:minio:sqs::1:mqtt``.
+We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted `images` bucket on `myminio` server. Here ARN value is `arn:minio:sqs::1:mqtt`.
 
 ```
 mc mb myminio/images
@@ -210,7 +218,7 @@ arn:minio:sqs::1:amqp s3:ObjectCreated:*,s3:ObjectRemoved:* Filter: suffix=”.j
 
 ### Step 3: Test on MQTT
 
-The python program below waits on mqtt topic ``/minio`` and prints event notifications on the console. We use [paho-mqtt](https://pypi.python.org/pypi/paho-mqtt/) library to do this.
+The python program below waits on mqtt topic `/minio` and prints event notifications on the console. We use [paho-mqtt](https://pypi.python.org/pypi/paho-mqtt/) library to do this.
 
 ```py
 #!/usr/bin/env python3
@@ -243,7 +251,7 @@ Execute this example python program to watch for MQTT events on the console.
 python mqtt.py
 ```
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
@@ -257,6 +265,7 @@ python mqtt.py
 ```
 
 <a name="Elasticsearch"></a>
+
 ## Publish MinIO events via Elasticsearch
 
 Install [Elasticsearch](https://www.elastic.co/downloads/elasticsearch) server.
@@ -277,12 +286,12 @@ MinIO requires a 5.x series version of Elasticsearch. This is the latest major r
 
 The MinIO server configuration file is stored on the backend in json format. The Elasticsearch configuration is located in the `elasticsearch` key under the `notify` top-level key. Create a configuration key-value pair here for your Elasticsearch instance. The key is a name for your Elasticsearch endpoint, and the value is a collection of key-value parameters described in the table below.
 
-| Parameter | Type | Description |
-|:---|:---|:---|
-| `enable` | _bool_ | (Required) Is this server endpoint configuration active/enabled? |
-| `format` | _string_ | (Required) Either `namespace` or `access`. |
-| `url` | _string_ | (Required) The Elasticsearch server's address, with optional authentication info. For example: `http://localhost:9200` or with authentication info `http://elastic:MagicWord@127.0.0.1:9200`. |
-| `index` | _string_ | (Required) The name of an Elasticsearch index in which MinIO will store documents. |
+| Parameter | Type     | Description                                                                                                                                                                                   |
+| :-------- | :------- | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `enable`  | _bool_   | (Required) Is this server endpoint configuration active/enabled?                                                                                                                              |
+| `format`  | _string_ | (Required) Either `namespace` or `access`.                                                                                                                                                    |
+| `url`     | _string_ | (Required) The Elasticsearch server's address, with optional authentication info. For example: `http://localhost:9200` or with authentication info `http://elastic:MagicWord@127.0.0.1:9200`. |
+| `index`   | _string_ | (Required) The name of an Elasticsearch index in which MinIO will store documents.                                                                                                            |
 
 An example of Elasticsearch configuration is as follows:
 
@@ -300,10 +309,13 @@ An example of Elasticsearch configuration is as follows:
 If Elasticsearch has authentication enabled, the credentials can be supplied to MinIO via the `url` parameter formatted as `PROTO://USERNAME:PASSWORD@ELASTICSEARCH_HOST:PORT`.
 
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the Elasticsearch configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs:  arn:minio:sqs::1:elasticsearch` at start-up if there were no errors.
+
+After updating the Elasticsearch configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:elasticsearch` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
@@ -327,13 +339,13 @@ arn:minio:sqs::1:elasticsearch s3:ObjectCreated:*,s3:ObjectRemoved:* Filter: suf
 
 ### Step 4: Test on Elasticsearch
 
-Upload a JPEG image into ``images`` bucket.
+Upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
 ```
 
-Use curl to view contents of ``minio_events`` index.
+Use curl to view contents of `minio_events` index.
 
 ```
 $ curl  "http://localhost:9200/minio_events/_search?pretty=true"
@@ -408,6 +420,7 @@ This output shows that a document has been created for the event in Elasticsearc
 Here we see that the document ID is the bucket and object name. In case `access` format was used, the document ID would be automatically generated by Elasticsearch.
 
 <a name="Redis"></a>
+
 ## Publish MinIO events via Redis
 
 Install [Redis](http://redis.io/download) server. For illustrative purposes, we have set the database password as "yoursecret".
@@ -424,13 +437,13 @@ The steps below show how to use this notification target in `namespace` and `acc
 
 The MinIO server configuration file is stored on the backend in json format.The Redis configuration is located in the `redis` key under the `notify` top-level key. Create a configuration key-value pair here for your Redis instance. The key is a name for your Redis endpoint, and the value is a collection of key-value parameters described in the table below.
 
-| Parameter | Type | Description |
-|:---|:---|:---|
-| `enable` | _bool_ | (Required) Is this server endpoint configuration active/enabled? |
-| `format` | _string_ | (Required) Either `namespace` or `access`. |
-| `address` | _string_ | (Required) The Redis server's address. For example: `localhost:6379`. |
-| `password` | _string_ | (Optional) The Redis server's password. |
-| `key` | _string_ | (Required) The name of the redis key under which events are stored. A hash is used in case of `namespace` format and a list in case of `access` format.|
+| Parameter  | Type     | Description                                                                                                                                             |
+| :--------- | :------- | :------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| `enable`   | _bool_   | (Required) Is this server endpoint configuration active/enabled?                                                                                        |
+| `format`   | _string_ | (Required) Either `namespace` or `access`.                                                                                                              |
+| `address`  | _string_ | (Required) The Redis server's address. For example: `localhost:6379`.                                                                                   |
+| `password` | _string_ | (Optional) The Redis server's password.                                                                                                                 |
+| `key`      | _string_ | (Required) The name of the redis key under which events are stored. A hash is used in case of `namespace` format and a list in case of `access` format. |
 
 An example of Redis configuration is as follows:
 
@@ -445,11 +458,15 @@ An example of Redis configuration is as follows:
     }
 }
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the Redis configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs:  arn:minio:sqs::1:redis` at start-up if there were no errors.
+
+After updating the Redis configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:redis` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
@@ -481,7 +498,7 @@ redis-cli -a yoursecret
 OK
 ```
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
@@ -501,13 +518,14 @@ Here we see that MinIO performed `HSET` on `minio_events` key.
 In case, `access` format was used, then `minio_events` would be a list, and the MinIO server would have performed an `RPUSH` to append to the list. A consumer of this list would ideally use `BLPOP` to remove list items from the left-end of the list.
 
 <a name="NATS"></a>
+
 ## Publish MinIO events via NATS
 
 Install NATS from [here](http://nats.io/).
 
 ### Step 1: Add NATS endpoint to MinIO
 
-The NATS configuration block in ``config.json`` is as follows:
+The NATS configuration block in `config.json` is as follows:
 
 ```
 "nats": {
@@ -530,14 +548,19 @@ The NATS configuration block in ``config.json`` is as follows:
     }
 },
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the NATS configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart MinIO server to reflect config changes. ``bucketevents`` is the subject used by NATS in this example.
+
+After updating the NATS configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart MinIO server to reflect config changes. `bucketevents` is the subject used by NATS in this example.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
+
 MinIO server also supports [NATS Streaming mode](http://nats.io/documentation/streaming/nats-streaming-intro/) that offers additional functionality like `Message/event persistence`, `At-least-once-delivery`, and `Publisher rate limiting`. To configure MinIO server to send notifications to NATS Streaming server, update the MinIO server configuration file as follows:
 
 ```
@@ -561,11 +584,12 @@ MinIO server also supports [NATS Streaming mode](http://nats.io/documentation/st
     }
 },
 ```
+
 Read more about sections `clusterID`, `clientID` on [NATS documentation](https://github.com/nats-io/nats-streaming-server/blob/master/README.md). Section `maxPubAcksInflight` is explained [here](https://github.com/nats-io/go-nats-streaming#publisher-rate-limiting).
 
 ### Step 2: Enable bucket notification using MinIO client
 
-We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted from ``images`` bucket on ``myminio`` server. Here ARN value is ``arn:minio:sqs::1:nats``. To understand more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
+We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted from `images` bucket on `myminio` server. Here ARN value is `arn:minio:sqs::1:nats`. To understand more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
 
 ```
 mc mb myminio/images
@@ -614,13 +638,13 @@ go run nats.go
 2016/10/12 06:39:18 Subscribing to subject 'bucketevents'
 ```
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
 ```
 
-The example ``nats.go`` program prints event notification to console.
+The example `nats.go` program prints event notification to console.
 
 ```
 go run nats.go
@@ -664,19 +688,21 @@ go run nats.go
 2017/07/07 11:47:40 Connected
 2017/07/07 11:47:40 Subscribing to subject 'bucketevents'
 ```
-Open another terminal and upload a JPEG image into ``images`` bucket.
+
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
 ```
 
-The example ``nats.go`` program prints event notification to console.
+The example `nats.go` program prints event notification to console.
 
 ```
 Received a message: {"EventType":"s3:ObjectCreated:Put","Key":"images/myphoto.jpg","Records":[{"eventVersion":"2.0","eventSource":"minio:s3","awsRegion":"","eventTime":"2017-07-07T18:46:37Z","eventName":"s3:ObjectCreated:Put","userIdentity":{"principalId":"minio"},"requestParameters":{"sourceIPAddress":"192.168.1.80:55328"},"responseElements":{"x-amz-request-id":"14CF20BD1EFD5B93","x-minio-origin-endpoint":"http://127.0.0.1:9000"},"s3":{"s3SchemaVersion":"1.0","configurationId":"Config","bucket":{"name":"images","ownerIdentity":{"principalId":"minio"},"arn":"arn:aws:s3:::images"},"object":{"key":"myphoto.jpg","size":248682,"eTag":"f1671feacb8bbf7b0397c6e9364e8c92","contentType":"image/jpeg","userDefined":{"content-type":"image/jpeg"},"versionId":"1","sequencer":"14CF20BD1EFD5B93"}},"source":{"host":"192.168.1.80","port":"55328","userAgent":"MinIO (linux; amd64) minio-go/2.0.4 mc/DEVELOPMENT.GOGET"}}],"level":"info","msg":"","time":"2017-07-07T11:46:37-07:00"}
 ```
 
 <a name="PostgreSQL"></a>
+
 ## Publish MinIO events via PostgreSQL
 
 Install [PostgreSQL](https://www.postgresql.org/) database server. For illustrative purposes, we have set the "postgres" user password as `password` and created a database called `minio_events` to store the events.
@@ -697,17 +723,17 @@ MinIO requires PostgreSQL version 9.5 or above. MinIO uses the [`INSERT ON CONFL
 
 The MinIO server configuration file is stored on the backend in json format. The PostgreSQL configuration is located in the `postgresql` key under the `notify` top-level key. Create a configuration key-value pair here for your PostgreSQL instance. The key is a name for your PostgreSQL endpoint, and the value is a collection of key-value parameters described in the table below.
 
-| Parameter | Type | Description |
-|:---|:---|:---|
-| `enable` | _bool_ | (Required) Is this server endpoint configuration active/enabled? |
-| `format` | _string_ | (Required) Either `namespace` or `access`. |
+| Parameter          | Type     | Description                                                                                                                                                                          |
+| :----------------- | :------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `enable`           | _bool_   | (Required) Is this server endpoint configuration active/enabled?                                                                                                                     |
+| `format`           | _string_ | (Required) Either `namespace` or `access`.                                                                                                                                           |
 | `connectionString` | _string_ | (Optional) [Connection string parameters](https://godoc.org/github.com/lib/pq#hdr-Connection_String_Parameters) for the PostgreSQL server. Can be used to set `sslmode` for example. |
-| `table` | _string_ | (Required) Table name in which events will be stored/updated. If the table does not exist, the MinIO server creates it at start-up.|
-| `host` | _string_ | (Optional) Host name of the PostgreSQL server. Defaults to `localhost`|
-| `port` | _string_ | (Optional) Port on which to connect to PostgreSQL server. Defaults to `5432`. |
-| `user` | _string_ | (Optional) Database user name. Defaults to user running the server process. |
-| `password` | _string_ | (Optional) Database password. |
-| `database` | _string_ | (Optional) Database name. |
+| `table`            | _string_ | (Required) Table name in which events will be stored/updated. If the table does not exist, the MinIO server creates it at start-up.                                                  |
+| `host`             | _string_ | (Optional) Host name of the PostgreSQL server. Defaults to `localhost`. IPv6 host should be enclosed with `[` and `]`                                                                |
+| `port`             | _string_ | (Optional) Port on which to connect to PostgreSQL server. Defaults to `5432`.                                                                                                        |
+| `user`             | _string_ | (Optional) Database user name. Defaults to user running the server process.                                                                                                          |
+| `password`         | _string_ | (Optional) Database password.                                                                                                                                                        |
+| `database`         | _string_ | (Optional) Database name.                                                                                                                                                            |
 
 An example of PostgreSQL configuration is as follows:
 
@@ -729,15 +755,18 @@ An example of PostgreSQL configuration is as follows:
 
 Note that for illustration here, we have disabled SSL. In the interest of security, for production this is not recommended.
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the Postgres configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs:  arn:minio:sqs::1:postgresql` at start-up if there were no errors.
+
+After updating the Postgres configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:postgresql` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
-Note that, you can add as many PostgreSQL server endpoint configurations as needed by providing an identifier (like "1" in the example above) for the PostgreSQL instance and an object of per-server configuration parameters.
 
+Note that, you can add as many PostgreSQL server endpoint configurations as needed by providing an identifier (like "1" in the example above) for the PostgreSQL instance and an object of per-server configuration parameters.
 
 ### Step 3: Enable bucket notification using MinIO client
 
@@ -760,7 +789,7 @@ arn:minio:sqs::1:postgresql s3:ObjectCreated:*,s3:ObjectRemoved:* Filter: suffix
 
 ### Step 4: Test on PostgreSQL
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
@@ -779,6 +808,7 @@ key                 |                      value
 ```
 
 <a name="MySQL"></a>
+
 ## Publish MinIO events via MySQL
 
 Install MySQL from [here](https://dev.mysql.com/downloads/mysql/). For illustrative purposes, we have set the root password as `password` and created a database called `miniodb` to store the events.
@@ -799,17 +829,17 @@ MinIO requires MySQL version 5.7.8 or above. MinIO uses the [JSON](https://dev.m
 
 The MinIO server configuration file is stored on the backend in json format. The MySQL configuration is located in the `mysql` key under the `notify` top-level key. Create a configuration key-value pair here for your MySQL instance. The key is a name for your MySQL endpoint, and the value is a collection of key-value parameters described in the table below.
 
-| Parameter | Type | Description |
-|:---|:---|:---|
-| `enable` | _bool_ | (Required) Is this server endpoint configuration active/enabled? |
-| `format` | _string_ | (Required) Either `namespace` or `access`. |
+| Parameter   | Type     | Description                                                                                                                                                                                                                                                             |
+| :---------- | :------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `enable`    | _bool_   | (Required) Is this server endpoint configuration active/enabled?                                                                                                                                                                                                        |
+| `format`    | _string_ | (Required) Either `namespace` or `access`.                                                                                                                                                                                                                              |
 | `dsnString` | _string_ | (Optional) [Data-Source-Name connection string](https://github.com/go-sql-driver/mysql#dsn-data-source-name) for the MySQL server. If not specified, the connection information specified by the `host`, `port`, `user`, `password` and `database` parameters are used. |
-| `table` | _string_ | (Required) Table name in which events will be stored/updated. If the table does not exist, the MinIO server creates it at start-up.|
-| `host` | _string_ | Host name of the MySQL server (used only if `dsnString` is empty). |
-| `port` | _string_ | Port on which to connect to the MySQL server (used only if `dsnString` is empty). |
-| `user` | _string_ | Database user-name (used only if `dsnString` is empty). |
-| `password` | _string_ | Database password (used only if `dsnString` is empty). |
-| `database` | _string_ | Database name (used only if `dsnString` is empty). |
+| `table`     | _string_ | (Required) Table name in which events will be stored/updated. If the table does not exist, the MinIO server creates it at start-up.                                                                                                                                     |
+| `host`      | _string_ | Host name of the MySQL server (used only if `dsnString` is empty).                                                                                                                                                                                                      |
+| `port`      | _string_ | Port on which to connect to the MySQL server (used only if `dsnString` is empty).                                                                                                                                                                                       |
+| `user`      | _string_ | Database user-name (used only if `dsnString` is empty).                                                                                                                                                                                                                 |
+| `password`  | _string_ | Database password (used only if `dsnString` is empty).                                                                                                                                                                                                                  |
+| `database`  | _string_ | Database name (used only if `dsnString` is empty).                                                                                                                                                                                                                      |
 
 An example of MySQL configuration is as follows:
 
@@ -827,16 +857,20 @@ An example of MySQL configuration is as follows:
         }
 }
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the MySQL configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs:  arn:minio:sqs::1:mysql` at start-up if there were no errors.
+
+After updating the MySQL configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:mysql` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
-Note that, you can add as many MySQL server endpoint configurations as needed by providing an identifier (like "1" in the example above) for the MySQL instance and an object of per-server configuration parameters.
 
+Note that, you can add as many MySQL server endpoint configurations as needed by providing an identifier (like "1" in the example above) for the MySQL instance and an object of per-server configuration parameters.
 
 ### Step 3: Enable bucket notification using MinIO client
 
@@ -858,7 +892,7 @@ arn:minio:sqs::1:postgresql s3:ObjectCreated:*,s3:ObjectRemoved:* Filter: suffix
 
 ### Step 4: Test on MySQL
 
-Open another terminal and upload a JPEG image into ``images`` bucket:
+Open another terminal and upload a JPEG image into `images` bucket:
 
 ```
 mc cp myphoto.jpg myminio/images
@@ -879,6 +913,7 @@ mysql> select * from minio_images;
 ```
 
 <a name="apache-kafka"></a>
+
 ## Publish MinIO events via Kafka
 
 Install Apache Kafka from [here](http://kafka.apache.org/).
@@ -889,7 +924,7 @@ MinIO requires Kafka version 0.10 or 0.9. Internally MinIO uses the [Shopify/sar
 
 ### Step 2: Add Kafka endpoint to MinIO
 
-The MinIO server configuration file is stored on the backend in json format. Update the kafka configuration block in ``config.json`` as follows:
+The MinIO server configuration file is stored on the backend in json format. Update the kafka configuration block in `config.json` as follows:
 
 ```
 "kafka": {
@@ -900,17 +935,22 @@ The MinIO server configuration file is stored on the backend in json format. Upd
     }
 }
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
-After updating the Kafka configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs:  arn:minio:sqs::1:kafka` at start-up if there were no errors.``bucketevents`` is the topic used by kafka in this example.
+
+After updating the Kafka configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:kafka` at start-up if there were no errors.`bucketevents` is the topic used by kafka in this example.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
+
 ### Step 3: Enable bucket notification using MinIO client
 
-We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted from ``images`` bucket on ``myminio`` server. Here ARN value is ``arn:minio:sqs::1:kafka``. To understand more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
+We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted from `images` bucket on `myminio` server. Here ARN value is `arn:minio:sqs::1:kafka`. To understand more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
 
 ```
 mc mb myminio/images
@@ -927,13 +967,13 @@ We used [kafkacat](https://github.com/edenhill/kafkacat) to print all notificati
 kafkacat -C -b localhost:9092 -t bucketevents
 ```
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp myphoto.jpg myminio/images
 ```
 
-``kafkacat`` prints the event notification to the console.
+`kafkacat` prints the event notification to the console.
 
 ```
 kafkacat -b localhost:9092 -t bucketevents
@@ -941,13 +981,14 @@ kafkacat -b localhost:9092 -t bucketevents
 ```
 
 <a name="webhooks"></a>
+
 ## Publish MinIO events via Webhooks
 
 [Webhooks](https://en.wikipedia.org/wiki/Webhook) are a way to receive information when it happens, rather than continually polling for that data.
 
 ### Step 1: Add Webhook endpoint to MinIO
 
-The MinIO server configuration file is stored on the backend in json format. Update the Webhook configuration block in ``config.json`` as follows
+The MinIO server configuration file is stored on the backend in json format. Update the Webhook configuration block in `config.json` as follows
 
 ```
 "webhook": {
@@ -956,17 +997,22 @@ The MinIO server configuration file is stored on the backend in json format. Upd
     "endpoint": "http://localhost:3000/"
 }
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
+
 After updating the webhook configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Here the endpoint is the server listening for webhook notifications. Save the file and restart the MinIO server for changes to take effect. Note that the endpoint needs to be live and reachable when you restart your MinIO server.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
+
 ### Step 2: Enable bucket notification using MinIO client
 
-We will enable bucket event notification to trigger whenever a JPEG image is uploaded to ``images`` bucket on ``myminio`` server. Here ARN value is ``arn:minio:sqs::1:webhook``. To learn more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
+We will enable bucket event notification to trigger whenever a JPEG image is uploaded to `images` bucket on `myminio` server. Here ARN value is `arn:minio:sqs::1:webhook`. To learn more about ARN please follow [AWS ARN](http://docs.aws.amazon.com/general/latest/gr/aws-arns-and-namespaces.html) documentation.
 
 ```
 mc mb myminio/images
@@ -995,18 +1041,19 @@ git clone https://github.com/minio/thumbnailer/
 npm install
 ```
 
-Then open the Thumbnailer config file at ``config/webhook.json`` and add the configuration for your MinIO server and then start Thumbnailer by
+Then open the Thumbnailer config file at `config/webhook.json` and add the configuration for your MinIO server and then start Thumbnailer by
 
 ```
 NODE_ENV=webhook node thumbnail-webhook.js
 ```
 
-Thumbnailer starts running at ``http://localhost:3000/``. Next, configure the MinIO server to send notifications to this URL (as mentioned in step 1) and use ``mc`` to set up bucket notifications (as mentioned in step 2). Then upload a JPEG image to MinIO server by
+Thumbnailer starts running at `http://localhost:3000/`. Next, configure the MinIO server to send notifications to this URL (as mentioned in step 1) and use `mc` to set up bucket notifications (as mentioned in step 2). Then upload a JPEG image to MinIO server by
 
 ```
 mc cp ~/images.jpg myminio/images
 .../images.jpg:  8.31 KB / 8.31 KB ┃▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓┃ 100.00% 59.42 KB/s 0s
 ```
+
 Wait a few moments, then check the bucket’s contents with mc ls — you will see a thumbnail appear.
 
 ```
@@ -1015,6 +1062,7 @@ mc ls myminio/images-thumbnail
 ```
 
 <a name="NSQ"></a>
+
 ## Publish MinIO events to NSQ
 
 Install an NSQ Daemon from [here](https://nsq.io/). Or use the following Docker
@@ -1043,21 +1091,24 @@ An example configuration for NSQ is shown below:
     }
 }
 ```
+
 To update the configuration, use `mc admin config get` command to get the current configuration file for the minio deployment in json format, and save it locally.
+
 ```sh
 $ mc admin config get myminio/ > /tmp/myconfig
 ```
+
 After updating the NSQ configuration in /tmp/myconfig , use `mc admin config set` command to update the configuration for the deployment.Restart the MinIO server to put the changes into effect. The server will print a line like `SQS ARNs: arn:minio:sqs::1:nsq` at start-up if there were no errors.
+
 ```sh
 $ mc admin config set myminio < /tmp/myconfig
 ```
 
 Note that, you can add as many NSQ daemon endpoint configurations as needed by providing an identifier (like "1" in the example above) for the NSQ instance and an object of per-server configuration parameters.
 
-
 ### Step 2: Enable bucket notification using MinIO client
 
-We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted ``images`` bucket on ``myminio`` server. Here ARN value is ``arn:minio:sqs::1:nsq``.
+We will enable bucket event notification to trigger whenever a JPEG image is uploaded or deleted `images` bucket on `myminio` server. Here ARN value is `arn:minio:sqs::1:nsq`.
 
 ```
 mc mb myminio/images
@@ -1074,7 +1125,7 @@ The simplest test is to download `nsq_tail` from [nsq github](https://github.com
 ./nsq_tail -nsqd-tcp-address 127.0.0.1:4150 -topic minio
 ```
 
-Open another terminal and upload a JPEG image into ``images`` bucket.
+Open another terminal and upload a JPEG image into `images` bucket.
 
 ```
 mc cp gopher.jpg myminio/images
@@ -1086,5 +1137,4 @@ You should receive the following event notification via NSQ once the upload comp
 {"EventName":"s3:ObjectCreated:Put","Key":"images/gopher.jpg","Records":[{"eventVersion":"2.0","eventSource":"minio:s3","awsRegion":"","eventTime":"2018-10-31T09:31:11Z","eventName":"s3:ObjectCreated:Put","userIdentity":{"principalId":"21EJ9HYV110O8NVX2VMS"},"requestParameters":{"sourceIPAddress":"10.1.1.1"},"responseElements":{"x-amz-request-id":"1562A792DAA53426","x-minio-origin-endpoint":"http://10.0.3.1:9000"},"s3":{"s3SchemaVersion":"1.0","configurationId":"Config","bucket":{"name":"images","ownerIdentity":{"principalId":"21EJ9HYV110O8NVX2VMS"},"arn":"arn:aws:s3:::images"},"object":{"key":"gopher.jpg","size":162023,"eTag":"5337769ffa594e742408ad3f30713cd7","contentType":"image/jpeg","userMetadata":{"content-type":"image/jpeg"},"versionId":"1","sequencer":"1562A792DAA53426"}},"source":{"host":"","port":"","userAgent":"MinIO (linux; amd64) minio-go/v6.0.8 mc/DEVELOPMENT.GOGET"}}]}
 ```
 
-
-*NOTE* If you are running [distributed MinIO](https://docs.min.io/docs/distributed-minio-quickstart-guide), modify ``~/.minio/config.json`` on all the nodes with your bucket event notification backend configuration.
+_NOTE_ If you are running [distributed MinIO](https://docs.min.io/docs/distributed-minio-quickstart-guide), modify `~/.minio/config.json` on all the nodes with your bucket event notification backend configuration.

docs/minio-limits.md

@@ -4,7 +4,8 @@
 
 |Item|Specification|
 |:---|:---|
-|Maximum number of servers| 32|
+|Maximum number of servers per cluster| 32|
+|Maximum number of Federated clusters | Unlimited|
 |Minimum number of servers| 02|
 |Maximum number of drives per server| Unlimited|
 |Read quorum| N/2|

docs/orchestration/docker-compose/docker-compose.yaml

@@ -5,7 +5,7 @@ version: '2'
 # 9001 through 9004.
 services:
  minio1:
-  image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+  image: minio/minio:RELEASE.2019-04-23T23-50-36Z
   volumes:
    - data1:/data
   ports:
@@ -15,7 +15,7 @@ services:
    MINIO_SECRET_KEY: minio123
   command: server http://minio1/data http://minio2/data http://minio3/data http://minio4/data 
  minio2:
-  image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+  image: minio/minio:RELEASE.2019-04-23T23-50-36Z
   volumes:
    - data2:/data
   ports:
@@ -25,7 +25,7 @@ services:
    MINIO_SECRET_KEY: minio123
   command: server http://minio1/data http://minio2/data http://minio3/data http://minio4/data 
  minio3:
-  image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+  image: minio/minio:RELEASE.2019-04-23T23-50-36Z
   volumes:
    - data3:/data
   ports:
@@ -35,7 +35,7 @@ services:
    MINIO_SECRET_KEY: minio123
   command: server http://minio1/data http://minio2/data http://minio3/data http://minio4/data 
  minio4:
-  image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+  image: minio/minio:RELEASE.2019-04-23T23-50-36Z
   volumes:
    - data4:/data
   ports:

docs/orchestration/docker-swarm/docker-compose-secrets.yaml

@@ -2,7 +2,7 @@ version: '3.1'
 
 services:
   minio1:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio1
     volumes:
       - minio1-data:/export
@@ -24,7 +24,7 @@ services:
       - access_key
 
   minio2:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio2
     volumes:
       - minio2-data:/export
@@ -46,7 +46,7 @@ services:
       - access_key
 
   minio3:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio3
     volumes:
       - minio3-data:/export
@@ -68,7 +68,7 @@ services:
       - access_key
 
   minio4:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio4
     volumes:
       - minio4-data:/export

docs/orchestration/docker-swarm/docker-compose.yaml

@@ -2,7 +2,7 @@ version: '3'
 
 services:
   minio1:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio1
     volumes:
       - minio1-data:/export
@@ -24,7 +24,7 @@ services:
     command: server http://minio1/export http://minio2/export http://minio3/export http://minio4/export
 
   minio2:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio2
     volumes:
       - minio2-data:/export
@@ -46,7 +46,7 @@ services:
     command: server http://minio1/export http://minio2/export http://minio3/export http://minio4/export
 
   minio3:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio3
     volumes:
       - minio3-data:/export
@@ -68,7 +68,7 @@ services:
     command: server http://minio1/export http://minio2/export http://minio3/export http://minio4/export
 
   minio4:
-    image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+    image: minio/minio:RELEASE.2019-04-23T23-50-36Z
     hostname: minio4
     volumes:
       - minio4-data:/export

docs/orchestration/kubernetes/minio-distributed-daemonset.yaml

@@ -30,7 +30,7 @@ spec:
           value: "minio"
         - name: MINIO_SECRET_KEY
           value: "minio123"
-        image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+        image: minio/minio:RELEASE.2019-04-23T23-50-36Z
         # Unfortunately you must manually define each server. Perhaps autodiscovery via DNS can be implemented in the future.
         args:
         - server

docs/orchestration/kubernetes/minio-distributed-statefulset.yaml

@@ -21,7 +21,7 @@ spec:
           value: "minio"
         - name: MINIO_SECRET_KEY
           value: "minio123"
-        image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+        image: minio/minio:RELEASE.2019-04-23T23-50-36Z
         args:
         - server
         - http://minio-0.minio.default.svc.cluster.local/data

docs/orchestration/kubernetes/minio-gcs-gateway-deployment.yaml

@@ -21,7 +21,7 @@ spec:
       containers:
       - name: minio
         # Pulls the default Minio image from Docker Hub
-        image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+        image: minio/minio:RELEASE.2019-04-23T23-50-36Z
         args:
         - gateway
         - gcs

docs/orchestration/kubernetes/minio-standalone-deployment.yaml

@@ -29,7 +29,7 @@ spec:
         - name: data 
           mountPath: "/data"
         # Pulls the lastest Minio image from Docker Hub
-        image: minio/minio:RELEASE.2019-04-18T01-15-57Z
+        image: minio/minio:RELEASE.2019-04-23T23-50-36Z
         args:
         - server
         - /data

docs/tls/README.md

@@ -22,7 +22,7 @@ Copy the existing private key and public certificate to the `certs` directory. T
 > NOTE: Location of custom certs directory can be specified using `--certs-dir` command line option.
 
 **Note:** 
-* The key and certificate files must be appended with `.key` and `.crt`, respectively.
+* Inside the `certs` directory, the private key must by named `private.key` and the public key must be named `public.crt`.
 * A certificate signed by a CA contains information about the issued identity (e.g. name, expiry, public key) and any intermediate certificates. The root CA is not included.
 
 ## <a name="generate-use-self-signed-keys-certificates"></a>3. Generate and use Self-signed Keys and Certificates with MinIO

go.mod

@@ -10,9 +10,9 @@ require (
 	github.com/DataDog/zstd v1.3.5 // indirect
 	github.com/alecthomas/participle v0.2.1
 	github.com/aliyun/aliyun-oss-go-sdk v0.0.0-20190307165228-86c17b95fcd5
+	github.com/armon/go-metrics v0.0.0-20190430140413-ec5e00d3c878 // indirect
 	github.com/bcicen/jstream v0.0.0-20190220045926-16c1f8af81c2
 	github.com/cheggaaa/pb v1.0.28
-	github.com/colinmarc/hdfs/v2 v2.0.0
 	github.com/coredns/coredns v1.4.0
 	github.com/coreos/bbolt v1.3.2 // indirect
 	github.com/coreos/etcd v3.3.12+incompatible
@@ -37,12 +37,15 @@ require (
 	github.com/grpc-ecosystem/go-grpc-prometheus v1.2.0 // indirect
 	github.com/grpc-ecosystem/grpc-gateway v1.8.5 // indirect
 	github.com/hashicorp/go-cleanhttp v0.5.1 // indirect
-	github.com/hashicorp/go-retryablehttp v0.5.3 // indirect
+	github.com/hashicorp/go-msgpack v0.5.4 // indirect
 	github.com/hashicorp/go-rootcerts v1.0.0 // indirect
+	github.com/hashicorp/raft v1.0.1 // indirect
 	github.com/hashicorp/vault v1.1.0
 	github.com/howeyc/gopass v0.0.0-20170109162249-bf9dde6d0d2c // indirect
 	github.com/inconshreveable/go-update v0.0.0-20160112193335-8152e7eb6ccf
+	github.com/jcmturner/gofork v0.0.0-20190328161633-dc7c13fece03 // indirect
 	github.com/jonboulle/clockwork v0.1.0 // indirect
+	github.com/json-iterator/go v1.1.6
 	github.com/klauspost/compress v1.4.1 // indirect
 	github.com/klauspost/cpuid v1.2.0 // indirect
 	github.com/klauspost/pgzip v1.2.1
@@ -53,7 +56,8 @@ require (
 	github.com/miekg/dns v1.1.8
 	github.com/minio/blazer v0.0.0-20171126203752-2081f5bf0465
 	github.com/minio/cli v1.3.0
-	github.com/minio/dsync v0.0.0-20190131060523-fb604afd87b2
+	github.com/minio/dsync v1.0.0
+	github.com/minio/hdfs/v3 v3.0.0
 	github.com/minio/highwayhash v1.0.0
 	github.com/minio/lsync v0.0.0-20190207022115-a4e43e3d0887
 	github.com/minio/mc v0.0.0-20190401030144-a1355e50e2e8
@@ -66,11 +70,11 @@ require (
 	github.com/nats-io/go-nats v1.7.2 // indirect
 	github.com/nats-io/go-nats-streaming v0.4.2
 	github.com/nats-io/nats v1.7.2
+	github.com/nats-io/nats-streaming-server v0.14.1 // indirect
 	github.com/nats-io/nkeys v0.0.2 // indirect
 	github.com/nats-io/nuid v1.0.1 // indirect
+	github.com/ncw/directio v1.0.5
 	github.com/nsqio/go-nsq v1.0.7
-	github.com/pascaldekloe/goe v0.1.0 // indirect
-	github.com/pkg/errors v0.8.1 // indirect
 	github.com/pkg/profile v1.3.0
 	github.com/prometheus/client_golang v0.9.3-0.20190127221311-3c4408c8b829
 	github.com/prometheus/client_model v0.0.0-20190129233127-fd36f4220a90 // indirect
@@ -93,12 +97,12 @@ require (
 	go.uber.org/atomic v1.3.2
 	go.uber.org/multierr v1.1.0 // indirect
 	go.uber.org/zap v1.9.1 // indirect
-	golang.org/x/crypto v0.0.0-20190417174047-f416ebab96af
-	golang.org/x/lint v0.0.0-20190409202823-959b441ac422 // indirect
-	golang.org/x/net v0.0.0-20190415214537-1da14a5a36f2
-	golang.org/x/sys v0.0.0-20190416152802-12500544f89f
-	golang.org/x/time v0.0.0-20190308202827-9d24e82272b4
-	golang.org/x/tools v0.0.0-20190417223002-a5870b403859 // indirect
+	golang.org/x/crypto v0.0.0-20190426145343-a29dc8fdc734
+	golang.org/x/net v0.0.0-20190424112056-4829fb13d2c6
+	golang.org/x/sync v0.0.0-20190423024810-112230192c58 // indirect
+	golang.org/x/sys v0.0.0-20190429190828-d89cdac9e872
+	golang.org/x/text v0.3.2 // indirect
+	golang.org/x/time v0.0.0-20190308202827-9d24e82272b4 // indirect
 	google.golang.org/api v0.3.0
 	gopkg.in/Shopify/sarama.v1 v1.20.0
 	gopkg.in/cheggaaa/pb.v1 v1.0.28 // indirect

go.sum

@@ -19,6 +19,7 @@ github.com/Azure/go-autorest v11.5.2+incompatible/go.mod h1:r+4oMnoxhatjLLJ6zxSW
 github.com/Azure/go-autorest v11.7.0+incompatible h1:gzma19dc9ejB75D90E5S+/wXouzpZyA+CV+/MJPSD/k=
 github.com/Azure/go-autorest v11.7.0+incompatible/go.mod h1:r+4oMnoxhatjLLJ6zxSWATqVooLgysK6ZNox3g/xq24=
 github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
+github.com/DataDog/datadog-go v2.2.0+incompatible/go.mod h1:LButxg5PwREeZtORoXG3tL4fMGNddJ+vMq1mwgfaqoQ=
 github.com/DataDog/zstd v1.3.5 h1:DtpNbljikUepEPD16hD4LvIcmhnhdLTiW/5pHgbmp14=
 github.com/DataDog/zstd v1.3.5/go.mod h1:1jcaCB/ufaK+sKp1NBhlGmpz41jOoPQ35bpF36t7BBo=
 github.com/Jeffail/gabs v1.1.1/go.mod h1:6xMvQMK4k33lb7GUUpaAPh6nKMmemQeg5d4gn7/bOXc=
@@ -45,6 +46,8 @@ github.com/araddon/gou v0.0.0-20190110011759-c797efecbb61/go.mod h1:ikc1XA58M+Rx
 github.com/armon/circbuf v0.0.0-20150827004946-bbbad097214e/go.mod h1:3U/XgcO3hCbHZ8TKRvWD2dDTCfh9M9ya+I9JpbB7O8o=
 github.com/armon/go-metrics v0.0.0-20180917152333-f0300d1749da h1:8GUt8eRujhVEGZFFEjBj46YV4rDjvGrNxb0KMWYkL2I=
 github.com/armon/go-metrics v0.0.0-20180917152333-f0300d1749da/go.mod h1:Q73ZrmVTwzkszR9V5SSuryQ31EELlFMUz1kKyl939pY=
+github.com/armon/go-metrics v0.0.0-20190430140413-ec5e00d3c878 h1:EFSB7Zo9Eg91v7MJPVsifUysc/wPdN+NOnVe6bWbdBM=
+github.com/armon/go-metrics v0.0.0-20190430140413-ec5e00d3c878/go.mod h1:3AMJUQhVx52RsWOnlkpikZr01T/yAVN2gn0861vByNg=
 github.com/armon/go-radix v0.0.0-20180808171621-7fddfc383310/go.mod h1:ufUuZ+zHj4x4TnLV4JWEpy2hxWSpsRywHrMgIH9cCH8=
 github.com/asaskevich/govalidator v0.0.0-20180720115003-f9ffefc3facf/go.mod h1:lB+ZfQJz7igIIfQNfa7Ml4HSf2uFQQRzpGGRXenZAgY=
 github.com/aws/aws-sdk-go v1.15.31/go.mod h1:mFuSZ37Z9YOHbQEwBWztmVzqXrEkub65tZoCYDt7FT0=
@@ -69,9 +72,9 @@ github.com/cheggaaa/pb v0.0.0-20160713104425-73ae1d68fe0b/go.mod h1:pQciLPpbU0ox
 github.com/cheggaaa/pb v1.0.28 h1:kWGpdAcSp3MxMU9CCHOwz/8V0kCHN4+9yQm2MzWuI98=
 github.com/cheggaaa/pb v1.0.28/go.mod h1:pQciLPpbU0oxA0h+VJYYLxO+XeDQb5pZijXscXHm81s=
 github.com/chrismalek/oktasdk-go v0.0.0-20181212195951-3430665dfaa0/go.mod h1:5d8DqS60xkj9k3aXfL3+mXBH0DPYO0FQjcKosxl+b/Q=
+github.com/circonus-labs/circonus-gometrics v2.3.1+incompatible/go.mod h1:nmEj6Dob7S7YxXgwXpfOuvO54S+tGdZdw9fuRZt25Ag=
+github.com/circonus-labs/circonusllhist v0.1.3/go.mod h1:kMXHVDlOchFAehlya5ePtbp5jckzBHf4XRpQvBOLI+I=
 github.com/client9/misspell v0.3.4/go.mod h1:qj6jICC3Q7zFZvVWo7KLAzC3yx5G7kyvSDkc90ppPyw=
-github.com/colinmarc/hdfs/v2 v2.0.0 h1:zXrYPcGbdTS8li50eel904btC8XI35dwDYlTvtFKQe4=
-github.com/colinmarc/hdfs/v2 v2.0.0/go.mod h1:uU5kspcuAB3y3XVcBew3IEXsYDTeyOHBbiBFSqCQHnk=
 github.com/containerd/continuity v0.0.0-20181203112020-004b46473808/go.mod h1:GL3xCUCBDV3CZiTSEKksMWbLE66hEyuu9qyDOOqM47Y=
 github.com/coredns/coredns v0.0.0-20180121192821-d4bf076ccf4e/go.mod h1:zASH/MVDgR6XZTbxvOnsZfffS+31vg6Ackf/wo1+AM0=
 github.com/coredns/coredns v1.4.0 h1:RubBkYmkByUqZWWkjRHvNLnUHgkRVqAWgSMmRFvpE1A=
@@ -244,6 +247,8 @@ github.com/hashicorp/go-immutable-radix v1.0.0/go.mod h1:0y9vanUI8NX6FsYoO3zeMjh
 github.com/hashicorp/go-memdb v0.0.0-20190306140544-eea0b16292ad/go.mod h1:kbfItVoBJwCfKXDXN4YoAXjxcFVZ7MRrJzyTX6H4giE=
 github.com/hashicorp/go-msgpack v0.5.3 h1:zKjpN5BK/P5lMYrLmBHdBULWbJ0XpYR+7NGzqkZzoD4=
 github.com/hashicorp/go-msgpack v0.5.3/go.mod h1:ahLV/dePpqEmjfWmKiqvPkv/twdG7iPBM1vqhUKIvfM=
+github.com/hashicorp/go-msgpack v0.5.4 h1:SFT72YqIkOcLdWJUYcriVX7hbrZpwc/f7h8aW2NUqrA=
+github.com/hashicorp/go-msgpack v0.5.4/go.mod h1:ahLV/dePpqEmjfWmKiqvPkv/twdG7iPBM1vqhUKIvfM=
 github.com/hashicorp/go-multierror v1.0.0 h1:iVjPR7a6H0tWELX5NxNe7bYopibicUzc7uPribsnS6o=
 github.com/hashicorp/go-multierror v1.0.0/go.mod h1:dHtQlpGsu+cZNNAkkCN/P3hoUDHhCYQXV3UM06sGGrk=
 github.com/hashicorp/go-plugin v0.0.0-20190220160451-3f118e8ee104/go.mod h1:++UyYGoz3o5w9ZzAdZxtQKrWWP+iqPBn3cQptSMzBuY=
@@ -275,6 +280,8 @@ github.com/hashicorp/memberlist v0.1.3/go.mod h1:ajVTdAv/9Im8oMAAj5G31PhhMCZJV2p
 github.com/hashicorp/nomad v0.8.7/go.mod h1:WRaKjdO1G2iqi86TvTjIYtKTyxg4pl7NLr9InxtWaI0=
 github.com/hashicorp/raft v1.0.0 h1:htBVktAOtGs4Le5Z7K8SF5H2+oWsQFYVmOgH5loro7Y=
 github.com/hashicorp/raft v1.0.0/go.mod h1:DVSAWItjLjTOkVbSpWQ0j0kUADIvDaCtBxIcbNAQLkI=
+github.com/hashicorp/raft v1.0.1 h1:94uRdS11oEneUkxmXq6Vg9shNhBILh2UTb9crQjJWl0=
+github.com/hashicorp/raft v1.0.1/go.mod h1:DVSAWItjLjTOkVbSpWQ0j0kUADIvDaCtBxIcbNAQLkI=
 github.com/hashicorp/serf v0.8.2/go.mod h1:6hOLApaqBFA1NXqRQAsxw9QxuDEvNxSQRwA/JwenrHc=
 github.com/hashicorp/vault v0.0.0-20181121181053-d4367e581fe1/go.mod h1:KfSyffbKxoVyspOdlaGVjIuwLobi07qD1bAbosPMpP0=
 github.com/hashicorp/vault v1.1.0 h1:v79NUgO5xCZnXVzUkIqFOXtP8YhpnHAi1fk3eo9cuOE=
@@ -299,12 +306,15 @@ github.com/inconshreveable/go-update v0.0.0-20160112193335-8152e7eb6ccf h1:WfD7V
 github.com/inconshreveable/go-update v0.0.0-20160112193335-8152e7eb6ccf/go.mod h1:hyb9oH7vZsitZCiBt0ZvifOrB+qc8PS5IiilCIb87rg=
 github.com/jcmturner/gofork v0.0.0-20180107083740-2aebee971930 h1:v4CYlQ+HeysPHsr2QFiEO60gKqnvn1xwvuKhhAhuEkk=
 github.com/jcmturner/gofork v0.0.0-20180107083740-2aebee971930/go.mod h1:MK8+TM0La+2rjBD4jE12Kj1pCCxK7d2LK/UM3ncEo0o=
+github.com/jcmturner/gofork v0.0.0-20190328161633-dc7c13fece03 h1:FUwcHNlEqkqLjLBdCp5PRlCFijNjvcYANOZXzCfXwCM=
+github.com/jcmturner/gofork v0.0.0-20190328161633-dc7c13fece03/go.mod h1:MK8+TM0La+2rjBD4jE12Kj1pCCxK7d2LK/UM3ncEo0o=
 github.com/jeffchao/backoff v0.0.0-20140404060208-9d7fd7aa17f2/go.mod h1:xkfESuHriIekR+4RoV+fu91j/CfnYM29Zi2tMFw5iD4=
 github.com/jefferai/jsonx v1.0.0/go.mod h1:OGmqmi2tTeI/PS+qQfBDToLHHJIy/RMp24fPo8vFvoQ=
 github.com/jellevandenhooff/dkim v0.0.0-20150330215556-f50fe3d243e1/go.mod h1:E0B/fFc00Y+Rasa88328GlI/XbtyysCtTHZS8h7IrBU=
 github.com/jmespath/go-jmespath v0.0.0-20160202185014-0b12d6b521d8/go.mod h1:Nht3zPeWKUH0NzdCt2Blrr5ys8VGpn0CEB0cQHVjt7k=
 github.com/jonboulle/clockwork v0.1.0 h1:VKV+ZcuP6l3yW9doeqz6ziZGgcynBVQO+obU0+0hcPo=
 github.com/jonboulle/clockwork v0.1.0/go.mod h1:Ii8DK3G1RaLaWxj9trq07+26W01tbo22gdxWY5EU2bo=
+github.com/json-iterator/go v1.1.6 h1:MrUvLMLTMxbqFJ9kzlvat/rYZqZnW3u4wkLzWTaFwKs=
 github.com/json-iterator/go v1.1.6/go.mod h1:+SdeFBvtyEkXs7REEP0seUULqWtbJapLOCVDaaPEHmU=
 github.com/jstemmer/go-junit-report v0.0.0-20190106144839-af01ea7f8024/go.mod h1:6v2b51hI/fHJwM22ozAgKL4VKDeJcHhJFhtBdhmNjmU=
 github.com/jtolds/gls v4.2.1+incompatible/go.mod h1:QJZ7F/aHp+rZTRtaJ1ow/lLfFfVYBRgL+9YlvaHOwJU=
@@ -367,8 +377,10 @@ github.com/minio/cli v0.0.0-20170227073228-b8ae5507c0ce/go.mod h1:hLsWNQy2wIf3FK
 github.com/minio/cli v1.3.0 h1:vB0iUpmyaH54+1jJJj62Aa0qFF3xO3i0J3IcKiM6bHM=
 github.com/minio/cli v1.3.0/go.mod h1:hLsWNQy2wIf3FKFnMlH69f4RdEyn8nbRA2shaulTjGY=
 github.com/minio/dsync v0.0.0-20190104003057-61c41ffdeea2/go.mod h1:eLQe3mXL0h02kNpPtBJiLr1fIEIJftgXRAjncjQbxJo=
-github.com/minio/dsync v0.0.0-20190131060523-fb604afd87b2 h1:5Aq4Aro/PSNVgoWWTLPX+zcfDY87VhtKPv+7x1ERJ1w=
-github.com/minio/dsync v0.0.0-20190131060523-fb604afd87b2/go.mod h1:eLQe3mXL0h02kNpPtBJiLr1fIEIJftgXRAjncjQbxJo=
+github.com/minio/dsync v1.0.0 h1:l6pQgUPBM41idlR0UOcpAP+EYim9MCwIAUh6sQQI1gk=
+github.com/minio/dsync v1.0.0/go.mod h1:eLQe3mXL0h02kNpPtBJiLr1fIEIJftgXRAjncjQbxJo=
+github.com/minio/hdfs/v3 v3.0.0 h1:yHa9ugB2UeazkiphO6Q2dmbqiKR5lZssN/1vda4gwoY=
+github.com/minio/hdfs/v3 v3.0.0/go.mod h1:k04lEYpgeojX3o1vQep6rQs4MCTD+qlh2xHFEa/BH6A=
 github.com/minio/highwayhash v0.0.0-20181220011308-93ed73d64169/go.mod h1:NL8wme5P5MoscwAkXfGroz3VgpCdhBw3KYOu5mEsvpU=
 github.com/minio/highwayhash v1.0.0 h1:iMSDhgUILCr0TNm8LWlSjF8N0ZIj2qbO8WHp6Q/J2BA=
 github.com/minio/highwayhash v1.0.0/go.mod h1:xQboMTeM9nY9v/LlAOxFctujiv5+Aq2hR5dxBpaMbdc=
@@ -407,7 +419,9 @@ github.com/mitchellh/mapstructure v1.1.2 h1:fmNYVwqnSfB9mZU6OS2O6GsXM+wcskZDuKQz
 github.com/mitchellh/mapstructure v1.1.2/go.mod h1:FVVH3fgwuzCH5S8UJGiWEs2h04kUh9fWfEaFds41c1Y=
 github.com/mitchellh/pointerstructure v0.0.0-20170205204203-f2329fcfa9e2/go.mod h1:KMNPMpc0BU/kZEgyDhBplsDn/mjnJMhyMjq4MWboN20=
 github.com/mitchellh/reflectwalk v1.0.0/go.mod h1:mSTlrgnPZtwu0c4WaC2kGObEpuNDbx0jmZXqmk4esnw=
+github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd h1:TRLaZ9cD/w8PVh93nsPXa1VrQ6jlwL5oN8l14QlcNfg=
 github.com/modern-go/concurrent v0.0.0-20180306012644-bacd9c7ef1dd/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
+github.com/modern-go/reflect2 v1.0.1 h1:9f412s+6RmYXLWZSEzVVgPGK7C2PphHj5RJrvfx9AWI=
 github.com/modern-go/reflect2 v1.0.1/go.mod h1:bx2lNnkwVCuqBIxFjflWJWanXIb3RllmbCylyMrvgv0=
 github.com/mwitkow/go-conntrack v0.0.0-20161129095857-cc309e4a2223/go.mod h1:qRWi+5nqEBWmkhHvq77mSJWrCKwh8bxhgT7d/eI7P4U=
 github.com/nats-io/gnatsd v1.4.1 h1:RconcfDeWpKCD6QIIwiVFcvForlXpWeJP7i5/lDLy44=
@@ -423,12 +437,16 @@ github.com/nats-io/nats v1.7.2 h1:8hasuHgNFiDGc7MTkGpPFc0v1Jz1nEoQwIHKAH2hFuo=
 github.com/nats-io/nats v1.7.2/go.mod h1:PpmYZwlgTfBI56QypJLfIMOfLnMRuVs+VL6r8mQ2SoQ=
 github.com/nats-io/nats-streaming-server v0.12.2 h1:EpyLfUBZgwu5c0mdSSytQsapm615AyitPssq7jgafdw=
 github.com/nats-io/nats-streaming-server v0.12.2/go.mod h1:RyqtDJZvMZO66YmyjIYdIvS69zu/wDAkyNWa8PIUa5c=
+github.com/nats-io/nats-streaming-server v0.14.1 h1:NLZNqluNLmsijog4x0VXofHGaSdAIH16Ge55maG5ItM=
+github.com/nats-io/nats-streaming-server v0.14.1/go.mod h1:RyqtDJZvMZO66YmyjIYdIvS69zu/wDAkyNWa8PIUa5c=
 github.com/nats-io/nkeys v0.0.2 h1:+qM7QpgXnvDDixitZtQUBDY9w/s9mu1ghS+JIbsrx6M=
 github.com/nats-io/nkeys v0.0.2/go.mod h1:dab7URMsZm6Z/jp9Z5UGa87Uutgc2mVpXLC4B7TDb/4=
 github.com/nats-io/nuid v1.0.0 h1:44QGdhbiANq8ZCbUkdn6W5bqtg+mHuDE4wOUuxxndFs=
 github.com/nats-io/nuid v1.0.0/go.mod h1:19wcPz3Ph3q0Jbyiqsd0kePYG7A95tJPxeL+1OSON2c=
 github.com/nats-io/nuid v1.0.1 h1:5iA8DT8V7q8WK2EScv2padNa/rTESc1KdnPw4TC2paw=
 github.com/nats-io/nuid v1.0.1/go.mod h1:19wcPz3Ph3q0Jbyiqsd0kePYG7A95tJPxeL+1OSON2c=
+github.com/ncw/directio v1.0.5 h1:JSUBhdjEvVaJvOoyPAbcW0fnd0tvRXD76wEfZ1KcQz4=
+github.com/ncw/directio v1.0.5/go.mod h1:rX/pKEYkOXBGOggmcyJeJGloCkleSvphPx2eV3t6ROk=
 github.com/nsqio/go-nsq v0.0.0-20181028195256-0527e80f3ba5/go.mod h1:XP5zaUs3pqf+Q71EqUJs3HYfBIqfK6G83WQMdNN+Ito=
 github.com/nsqio/go-nsq v1.0.7 h1:O0pIZJYTf+x7cZBA0UMY8WxFG79lYTURmWzAAh48ljY=
 github.com/nsqio/go-nsq v1.0.7/go.mod h1:XP5zaUs3pqf+Q71EqUJs3HYfBIqfK6G83WQMdNN+Ito=
@@ -553,6 +571,7 @@ github.com/tidwall/sjson v1.0.4 h1:UcdIRXff12Lpnu3OLtZvnc03g4vH2suXDXhBwBqmzYg=
 github.com/tidwall/sjson v1.0.4/go.mod h1:bURseu1nuBkFpIES5cz6zBtjmYeOQmEESshn7VpF15Y=
 github.com/tmc/grpc-websocket-proxy v0.0.0-20190109142713-0ad062ec5ee5 h1:LnC5Kc/wtumK+WB441p7ynQJzVuNRJiqddSIE3IlSEQ=
 github.com/tmc/grpc-websocket-proxy v0.0.0-20190109142713-0ad062ec5ee5/go.mod h1:ncp9v5uamzpCO7NfCPTXjqaC+bZgJeR0sMTm6dMHP7U=
+github.com/tv42/httpunix v0.0.0-20150427012821-b75d8614f926/go.mod h1:9ESjWnEqriFuLhtthL60Sar/7RFoluCcXsuvEwTV5KM=
 github.com/ugorji/go v0.0.0-20180628102755-7d51bbe6161d/go.mod h1:hnLbHMwcvSihnDhEfx2/BzKp2xb0Y+ErdfYcrs9tkJQ=
 github.com/ugorji/go v1.1.2 h1:JON3E2/GPW2iDNGoSAusl1KDf5TRQ8k8q7Tp097pZGs=
 github.com/ugorji/go v1.1.2/go.mod h1:hnLbHMwcvSihnDhEfx2/BzKp2xb0Y+ErdfYcrs9tkJQ=
@@ -592,10 +611,8 @@ golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACk
 golang.org/x/crypto v0.0.0-20190313024323-a1f597ede03a/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20190325154230-a5d413f7728c h1:Vj5n4GlwjmQteupaxJ9+0FNOmBrHfq7vN4btdGoDZgI=
 golang.org/x/crypto v0.0.0-20190325154230-a5d413f7728c/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
-golang.org/x/crypto v0.0.0-20190404164418-38d8ce5564a5 h1:bselrhR0Or1vomJZC8ZIjWtbDmn9OYFLX5Ik9alpJpE=
-golang.org/x/crypto v0.0.0-20190404164418-38d8ce5564a5/go.mod h1:WFFai1msRO1wXaEeE5yQxYXgSfI8pQAWXbQop6sCtWE=
-golang.org/x/crypto v0.0.0-20190417174047-f416ebab96af h1:6qGQw30u837TXZbCmLFR9AVA+RjJU1LIbvk0oIkDZGY=
-golang.org/x/crypto v0.0.0-20190417174047-f416ebab96af/go.mod h1:WFFai1msRO1wXaEeE5yQxYXgSfI8pQAWXbQop6sCtWE=
+golang.org/x/crypto v0.0.0-20190426145343-a29dc8fdc734 h1:p/H982KKEjUnLJkM3tt/LemDnOc1GiZL5FCVlORJ5zo=
+golang.org/x/crypto v0.0.0-20190426145343-a29dc8fdc734/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/exp v0.0.0-20190121172915-509febef88a4/go.mod h1:CJ0aWSM057203Lf6IL+f9T1iT9GByDxfZKAQTCR3kQA=
 golang.org/x/lint v0.0.0-20180702182130-06c8688daad7/go.mod h1:UVdnD1Gm6xHRNCYTkRU2/jEulfH38KcIWyp/GAMgvoE=
 golang.org/x/lint v0.0.0-20181026193005-c67002cb31c3/go.mod h1:UVdnD1Gm6xHRNCYTkRU2/jEulfH38KcIWyp/GAMgvoE=
@@ -604,8 +621,6 @@ golang.org/x/lint v0.0.0-20190227174305-5b3e6a55c961/go.mod h1:wehouNa3lNwaWXcvx
 golang.org/x/lint v0.0.0-20190301231843-5614ed5bae6f/go.mod h1:UVdnD1Gm6xHRNCYTkRU2/jEulfH38KcIWyp/GAMgvoE=
 golang.org/x/lint v0.0.0-20190313153728-d0100b6bd8b3 h1:XQyxROzUlZH+WIQwySDgnISgOivlhjIEwaQaJEJrrN0=
 golang.org/x/lint v0.0.0-20190313153728-d0100b6bd8b3/go.mod h1:6SW0HCj/g11FgYtHlgUYUwCkIfeOF89ocIRzGO/8vkc=
-golang.org/x/lint v0.0.0-20190409202823-959b441ac422 h1:QzoH/1pFpZguR8NrRHLcO6jKqfv2zpuSqZLgdm7ZmjI=
-golang.org/x/lint v0.0.0-20190409202823-959b441ac422/go.mod h1:6SW0HCj/g11FgYtHlgUYUwCkIfeOF89ocIRzGO/8vkc=
 golang.org/x/net v0.0.0-20180530234432-1e491301e022/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20180826012351-8a410e7b638d/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
@@ -625,8 +640,8 @@ golang.org/x/net v0.0.0-20190318221613-d196dffd7c2b/go.mod h1:t9HGtf8HONx5eT2rtn
 golang.org/x/net v0.0.0-20190324223953-e3b2ff56ed87/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
 golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3 h1:0GoQqolDA55aaLxZyTzK/Y2ePZzZTUrRacwib7cNsYQ=
 golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
-golang.org/x/net v0.0.0-20190415214537-1da14a5a36f2 h1:iC0Y6EDq+rhnAePxGvJs2kzUAYcwESqdcGRPzEUfzTU=
-golang.org/x/net v0.0.0-20190415214537-1da14a5a36f2/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
+golang.org/x/net v0.0.0-20190424112056-4829fb13d2c6 h1:FP8hkuE6yUEaJnK7O2eTuejKWwW+Rhfj80dQ2JcKxCU=
+golang.org/x/net v0.0.0-20190424112056-4829fb13d2c6/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
 golang.org/x/oauth2 v0.0.0-20180603041954-1e0a3fa8ba9a/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
 golang.org/x/oauth2 v0.0.0-20180821212333-d2e6202438be/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
 golang.org/x/oauth2 v0.0.0-20181017192945-9dcd33a902f4/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
@@ -639,6 +654,8 @@ golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20190227155943-e225da77a7e6 h1:bjcUS9ztw9kFmmIxJInhon/0Is3p+EHBKNgquIzo1OI=
 golang.org/x/sync v0.0.0-20190227155943-e225da77a7e6/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
+golang.org/x/sync v0.0.0-20190423024810-112230192c58 h1:8gQV6CLnAEikrhgkHFbMAEhagSSnXWGV915qUMm9mrU=
+golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sys v0.0.0-20180823144017-11551d06cbcc/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180828065106-d99a578cf41b/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180830151530-49385e6e1522/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -661,21 +678,21 @@ golang.org/x/sys v0.0.0-20190222072716-a9d3bda3a223/go.mod h1:STP8DvDyc/dI5b8T5h
 golang.org/x/sys v0.0.0-20190304154630-e844e0132e93/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190318195719-6c81ef8f67ca/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20190322080309-f49334f85ddc/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20190403152447-81d4e9dc473e h1:nFYrTHrdrAOpShe27kaFHjsqYSEQ0KWqdWLu3xuZJts=
-golang.org/x/sys v0.0.0-20190403152447-81d4e9dc473e/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20190405154228-4b34438f7a67 h1:1Fzlr8kkDLQwqMP8GxrhptBLqZG/EDpiATneiZHY998=
-golang.org/x/sys v0.0.0-20190405154228-4b34438f7a67/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20190416152802-12500544f89f h1:1ZH9RnjNgLzh6YrsRp/c6ddZ8Lq0fq9xztNOoWJ2sz4=
-golang.org/x/sys v0.0.0-20190416152802-12500544f89f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20190429190828-d89cdac9e872 h1:cGjJzUd8RgBw428LXP65YXni0aiGNA4Bl+ls8SmLOm8=
+golang.org/x/sys v0.0.0-20190429190828-d89cdac9e872/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.1-0.20180807135948-17ff2d5776d2 h1:z99zHgr7hKfrUcX/KsoJk5FJfjTceCKIp96+biqP4To=
 golang.org/x/text v0.3.1-0.20180807135948-17ff2d5776d2/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
+golang.org/x/text v0.3.2 h1:tW2bmiBqwgJj/UpqtC8EpXEZVYOwU0yG4iWbprSVAcs=
+golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/time v0.0.0-20180412165947-fbb02b2291d2/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/time v0.0.0-20181108054448-85acf8d2951c/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/time v0.0.0-20190308202827-9d24e82272b4 h1:SvFZT6jyqRaOeXpc5h/JSfZenJ2O330aBsf7JfSUXmQ=
 golang.org/x/time v0.0.0-20190308202827-9d24e82272b4/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/tools v0.0.0-20180221164845-07fd8470d635/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20180828015842-6cd1fcedba52/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
+golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20181219222714-6e267b5cc78e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190114222345-bf090417da8b/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190226205152-f727befe758c/go.mod h1:9Yl7xja0Znq3iFh3HoIrodX9oNMXvdceNzlUR8zjMvY=
@@ -683,8 +700,6 @@ golang.org/x/tools v0.0.0-20190311212946-11955173bddd/go.mod h1:LCzVGOaR6xXOjkQ3
 golang.org/x/tools v0.0.0-20190312170243-e65039ee4138/go.mod h1:LCzVGOaR6xXOjkQ3onu1FJEFr0SW1gC7cKk1uF8kGRs=
 golang.org/x/tools v0.0.0-20190318200714-bb1270c20edf/go.mod h1:LCzVGOaR6xXOjkQ3onu1FJEFr0SW1gC7cKk1uF8kGRs=
 golang.org/x/tools v0.0.0-20190328211700-ab21143f2384/go.mod h1:LCzVGOaR6xXOjkQ3onu1FJEFr0SW1gC7cKk1uF8kGRs=
-golang.org/x/tools v0.0.0-20190417223002-a5870b403859 h1:9IwAPkJAVzY3ES7jydtY2zc5YltyS6lQuXUJoqoWQYg=
-golang.org/x/tools v0.0.0-20190417223002-a5870b403859/go.mod h1:LCzVGOaR6xXOjkQ3onu1FJEFr0SW1gC7cKk1uF8kGRs=
 google.golang.org/api v0.0.0-20180603000442-8e296ef26005/go.mod h1:4mhQ8q/RsB7i+udVvVy5NUi08OU8ZlA0gRVgrF7VFY0=
 google.golang.org/api v0.0.0-20180910000450-7ca32eb868bf/go.mod h1:4mhQ8q/RsB7i+udVvVy5NUi08OU8ZlA0gRVgrF7VFY0=
 google.golang.org/api v0.0.0-20180916000451-19ff8768a5c0/go.mod h1:4mhQ8q/RsB7i+udVvVy5NUi08OU8ZlA0gRVgrF7VFY0=