endpoint: Receiver.Receive: save a peek of recv stream to a temporary dataset if placeholder + encryption recv -F error is detected

fixes #277

Makefile

@@ -23,7 +23,8 @@ GOHOSTARCH ?= $(shell bash -c 'source <($(GO) env) && echo "$$GOHOSTARCH"')
 GO_ENV_VARS := GO111MODULE=on
 GO_LDFLAGS := "-X github.com/zrepl/zrepl/version.zreplVersion=$(_ZREPL_VERSION)"
 GO_MOD_READONLY := -mod=readonly
-GO_BUILDFLAGS := $(GO_MOD_READONLY)
+GO_EXTRA_BUILDFLAGS :=
+GO_BUILDFLAGS := $(GO_MOD_READONLY) $(GO_EXTRA_BUILDFLAGS)
 GO_BUILD := $(GO_ENV_VARS) $(GO) build $(GO_BUILDFLAGS) -ldflags $(GO_LDFLAGS)
 GOLANGCI_LINT := golangci-lint
 ifneq ($(GOARM),)
@@ -65,6 +66,7 @@ wrapup-and-checksum:
 		-acf $(NOARCH_TARBALL) \
 		$(ARTIFACTDIR)/docs/html \
 		$(ARTIFACTDIR)/bash_completion \
+		$(ARTIFACTDIR)/_zrepl.zsh_completion \
 		$(ARTIFACTDIR)/go_env.txt \
 		dist \
 		config/samples
@@ -159,7 +161,7 @@ platformtest: # do not track dependency on platformtest-bin to allow build of pl
 		$(ZREPL_PLATFORMTEST_ARGS)
 
 ##################### NOARCH #####################
-.PHONY: noarch $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/go_env.txt docs docs-clean
+.PHONY: noarch $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/_zrepl.zsh_completion $(ARTIFACTDIR)/go_env.txt docs docs-clean
 
 
 $(ARTIFACTDIR):
@@ -167,12 +169,16 @@ $(ARTIFACTDIR):
 $(ARTIFACTDIR)/docs: $(ARTIFACTDIR)
 	mkdir -p "$@"
 
-noarch: $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/go_env.txt docs
+noarch: $(ARTIFACTDIR)/bash_completion $(ARTIFACTDIR)/_zrepl.zsh_completion $(ARTIFACTDIR)/go_env.txt docs
 	# pass
 
 $(ARTIFACTDIR)/bash_completion:
 	$(MAKE) zrepl-bin GOOS=$(GOHOSTOS) GOARCH=$(GOHOSTARCH)
-	artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) bashcomp "$@"
+	artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) gencompletion bash "$@"
+
+$(ARTIFACTDIR)/_zrepl.zsh_completion:
+	$(MAKE) zrepl-bin GOOS=$(GOHOSTOS) GOARCH=$(GOHOSTARCH)
+	artifacts/zrepl-$(GOHOSTOS)-$(GOHOSTARCH) gencompletion zsh "$@"
 
 $(ARTIFACTDIR)/go_env.txt:
 	$(GO_ENV_VARS) $(GO) env > $@

cli/cli.go

@@ -19,29 +19,55 @@ var rootCmd = &cobra.Command{
 	Short: "One-stop ZFS replication solution",
 }
 
-var bashcompCmd = &cobra.Command{
-	Use:   "bashcomp path/to/out/file",
-	Short: "generate bash completions",
-	Run: func(cmd *cobra.Command, args []string) {
-		if len(args) != 1 {
-			fmt.Fprintf(os.Stderr, "specify exactly one positional agument\n")
-			err := cmd.Usage()
-			if err != nil {
-				panic(err)
-			}
-			os.Exit(1)
-		}
-		if err := rootCmd.GenBashCompletionFile(args[0]); err != nil {
-			fmt.Fprintf(os.Stderr, "error generating bash completion: %s", err)
-			os.Exit(1)
-		}
+func init() {
+	rootCmd.PersistentFlags().StringVar(&rootArgs.configPath, "config", "", "config file path")
+}
+
+var genCompletionCmd = &cobra.Command{
+	Use:   "gencompletion",
+	Short: "generate shell auto-completions",
+}
+
+type completionCmdInfo struct {
+	genFunc func(outpath string) error
+	help    string
+}
+
+var completionCmdMap = map[string]completionCmdInfo{
+	"zsh": {
+		rootCmd.GenZshCompletionFile,
+		"  save to file `_zrepl` in your zsh's $fpath",
+	},
+	"bash": {
+		rootCmd.GenBashCompletionFile,
+		"  save to a path and source that path in your .bashrc",
 	},
-	Hidden: true,
 }
 
 func init() {
-	rootCmd.PersistentFlags().StringVar(&rootArgs.configPath, "config", "", "config file path")
-	rootCmd.AddCommand(bashcompCmd)
+	for sh, info := range completionCmdMap {
+		sh, info := sh, info
+		genCompletionCmd.AddCommand(&cobra.Command{
+			Use:     fmt.Sprintf("%s path/to/out/file", sh),
+			Short:   fmt.Sprintf("generate %s completions", sh),
+			Example: info.help,
+			Run: func(cmd *cobra.Command, args []string) {
+				if len(args) != 1 {
+					fmt.Fprintf(os.Stderr, "specify exactly one positional agument\n")
+					err := cmd.Usage()
+					if err != nil {
+						panic(err)
+					}
+					os.Exit(1)
+				}
+				if err := info.genFunc(args[0]); err != nil {
+					fmt.Fprintf(os.Stderr, "error generating %s completion: %s", sh, err)
+					os.Exit(1)
+				}
+			},
+		})
+	}
+	rootCmd.AddCommand(genCompletionCmd)
 }
 
 type Subcommand struct {

client/migrate.go

@@ -211,7 +211,7 @@ func doMigrateReplicationCursorFS(ctx context.Context, v1CursorJobs []job.Job, f
 	}
 	fmt.Printf("identified owning job %q\n", owningJob.Name())
 
-	bookmarks, err := zfs.ZFSListFilesystemVersions(fs, zfs.ListFilesystemVersionsOptions{
+	bookmarks, err := zfs.ZFSListFilesystemVersions(ctx, fs, zfs.ListFilesystemVersionsOptions{
 		Types: zfs.Bookmarks,
 	})
 	if err != nil {

daemon/snapper/snapper.go

@@ -485,7 +485,7 @@ var findSyncPointFSNoFilesystemVersionsErr = fmt.Errorf("no filesystem versions"
 
 func findSyncPointFSNextOptimalSnapshotTime(l Logger, now time.Time, interval time.Duration, prefix string, d *zfs.DatasetPath) (time.Time, error) {
 
-	fsvs, err := zfs.ZFSListFilesystemVersions(d, zfs.ListFilesystemVersionsOptions{
+	fsvs, err := zfs.ZFSListFilesystemVersions(context.TODO(), d, zfs.ListFilesystemVersionsOptions{
 		Types:           zfs.Snapshots,
 		ShortnamePrefix: prefix,
 	})

docs/changelog.rst

@@ -62,7 +62,9 @@ Actual changelog:
 * |bugfix| |docs| snapshotting: clarify sync-up behavior and warn about filesystems
   that will not be snapshotted until the sync-up phase is over
 * |docs| Document new replication features in the :ref:`config overview <overview-how-replication-works>` and :repomasterlink:`replication/design.md`.
+* |feature| documented subcommand to generate ``bash`` and ``zsh`` completions
 * **[MAINTAINER NOTICE]** New platform tests in this version, please make sure you run them for your distro!
+* **[MAINTAINER NOTICE]** Please add the shell completions to the zrepl packages.
 
 0.2.1
 -----

endpoint/endpoint.go

@@ -2,14 +2,17 @@
 package endpoint
 
 import (
+	"bytes"
 	"context"
 	"fmt"
+	"io"
 	"path"
 	"sync"
 
 	"github.com/pkg/errors"
 
 	"github.com/zrepl/zrepl/replication/logic/pdu"
+	"github.com/zrepl/zrepl/util/chainedio"
 	"github.com/zrepl/zrepl/util/chainlock"
 	"github.com/zrepl/zrepl/util/envconst"
 	"github.com/zrepl/zrepl/util/semaphore"
@@ -96,7 +99,7 @@ func (s *Sender) ListFilesystemVersions(ctx context.Context, r *pdu.ListFilesyst
 	if err != nil {
 		return nil, err
 	}
-	fsvs, err := zfs.ZFSListFilesystemVersions(lp, zfs.ListFilesystemVersionsOptions{})
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, lp, zfs.ListFilesystemVersionsOptions{})
 	if err != nil {
 		return nil, err
 	}
@@ -241,7 +244,7 @@ func sendArgsFromPDUAndValidateExistsAndGetVersion(ctx context.Context, fs strin
 	return version, nil
 }
 
-func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
 
 	_, err := s.filterCheckFS(r.Filesystem)
 	if err != nil {
@@ -339,11 +342,11 @@ func (s *Sender) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.St
 
 	// step holds & replication cursor released / moved forward in s.SendCompleted => s.moveCursorAndReleaseSendHolds
 
-	streamCopier, err := zfs.ZFSSend(ctx, sendArgs)
+	sendStream, err := zfs.ZFSSend(ctx, sendArgs)
 	if err != nil {
 		return nil, nil, errors.Wrap(err, "zfs send failed")
 	}
-	return res, streamCopier, nil
+	return res, sendStream, nil
 }
 
 func (p *Sender) SendCompleted(ctx context.Context, r *pdu.SendCompletedReq) (*pdu.SendCompletedRes, error) {
@@ -476,7 +479,7 @@ func (p *Sender) ReplicationCursor(ctx context.Context, req *pdu.ReplicationCurs
 	return &pdu.ReplicationCursorRes{Result: &pdu.ReplicationCursorRes_Guid{Guid: cursor.Guid}}, nil
 }
 
-func (p *Sender) Receive(ctx context.Context, r *pdu.ReceiveReq, receive zfs.StreamCopier) (*pdu.ReceiveRes, error) {
+func (p *Sender) Receive(ctx context.Context, r *pdu.ReceiveReq, _ io.ReadCloser) (*pdu.ReceiveRes, error) {
 	return nil, fmt.Errorf("sender does not implement Receive()")
 }
 
@@ -648,7 +651,7 @@ func (s *Receiver) ListFilesystemVersions(ctx context.Context, req *pdu.ListFile
 	}
 	// TODO share following code with sender
 
-	fsvs, err := zfs.ZFSListFilesystemVersions(lp, zfs.ListFilesystemVersionsOptions{})
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, lp, zfs.ListFilesystemVersionsOptions{})
 	if err != nil {
 		return nil, err
 	}
@@ -680,13 +683,13 @@ func (s *Receiver) ReplicationCursor(context.Context, *pdu.ReplicationCursorReq)
 	return nil, fmt.Errorf("ReplicationCursor not implemented for Receiver")
 }
 
-func (s *Receiver) Send(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (s *Receiver) Send(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
 	return nil, nil, fmt.Errorf("receiver does not implement Send()")
 }
 
 var maxConcurrentZFSRecvSemaphore = semaphore.New(envconst.Int64("ZREPL_ENDPOINT_MAX_CONCURRENT_RECV", 10))
 
-func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs.StreamCopier) (*pdu.ReceiveRes, error) {
+func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive io.ReadCloser) (*pdu.ReceiveRes, error) {
 	getLogger(ctx).Debug("incoming Receive")
 	defer receive.Close()
 
@@ -765,21 +768,30 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 		return nil, visitErr
 	}
 
+	log := getLogger(ctx).WithField("proto_fs", req.GetFilesystem()).WithField("local_fs", lp.ToString())
+
 	// determine whether we need to rollback the filesystem / change its placeholder state
 	var clearPlaceholderProperty bool
 	var recvOpts zfs.RecvOptions
 	ph, err := zfs.ZFSGetFilesystemPlaceholderState(ctx, lp)
-	if err == nil && ph.FSExists && ph.IsPlaceholder {
+	if err != nil {
+		return nil, errors.Wrap(err, "cannot get placeholder state")
+	}
+	log.WithField("placeholder_state", fmt.Sprintf("%#v", ph)).Debug("placeholder state")
+	if ph.FSExists && ph.IsPlaceholder {
 		recvOpts.RollbackAndForceRecv = true
 		clearPlaceholderProperty = true
 	}
+
 	if clearPlaceholderProperty {
+		log.Info("clearing placeholder property")
 		if err := zfs.ZFSSetPlaceholder(ctx, lp, false); err != nil {
 			return nil, fmt.Errorf("cannot clear placeholder property for forced receive: %s", err)
 		}
 	}
 
 	if req.ClearResumeToken && ph.FSExists {
+		log.Info("clearing resume token")
 		if err := zfs.ZFSRecvClearResumeToken(ctx, lp.ToString()); err != nil {
 			return nil, errors.Wrap(err, "cannot clear resume token")
 		}
@@ -790,7 +802,7 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 		return nil, errors.Wrap(err, "cannot determine whether we can use resumable send & recv")
 	}
 
-	getLogger(ctx).Debug("acquire concurrent recv semaphore")
+	log.Debug("acquire concurrent recv semaphore")
 	// TODO use try-acquire and fail with resource-exhaustion rpc status
 	// => would require handling on the client-side
 	// => this is a dataconn endpoint, doesn't have the status code semantics of gRPC
@@ -800,14 +812,89 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 	}
 	defer guard.Release()
 
-	getLogger(ctx).WithField("opts", fmt.Sprintf("%#v", recvOpts)).Debug("start receive command")
+	log.Info("peeking 1M ahead")
+	var peek bytes.Buffer
+	var MaxPeek = envconst.Int64("ZREPL_ENDPOINT_RECV_PEEK_SIZE", 1<<20)
+	if _, err := io.Copy(&peek, io.LimitReader(receive, MaxPeek)); err != nil {
+		log.WithError(err).Error("cannot read peek-buffer from send stream")
+	}
+	var peekCopy bytes.Buffer
+	if n, err := peekCopy.Write(peek.Bytes()); err != nil || n != peek.Len() {
+		panic(peek.Len())
+	}
+
+	log.WithField("opts", fmt.Sprintf("%#v", recvOpts)).Debug("start receive command")
 
 	snapFullPath := to.FullPath(lp.ToString())
-	if err := zfs.ZFSRecv(ctx, lp.ToString(), to, receive, recvOpts); err != nil {
-		getLogger(ctx).
+	if err := zfs.ZFSRecv(ctx, lp.ToString(), to, chainedio.NewChainedReader(&peek, receive), recvOpts); err != nil {
+
+		// best-effort rollback of placeholder state if the recv didn't start
+		_, resumableStatePresent := err.(*zfs.RecvFailedWithResumeTokenErr)
+		disablePlaceholderRestoration := envconst.Bool("ZREPL_ENDPOINT_DISABLE_PLACEHOLDER_RESTORATION", false)
+		placeholderRestored := !ph.IsPlaceholder
+		if !disablePlaceholderRestoration && !resumableStatePresent && recvOpts.RollbackAndForceRecv && ph.FSExists && ph.IsPlaceholder && clearPlaceholderProperty {
+			log.Info("restoring placeholder property")
+			if phErr := zfs.ZFSSetPlaceholder(ctx, lp, true); phErr != nil {
+				log.WithError(phErr).Error("cannot restore placeholder property after failed receive, subsequent replications will likely fail with a different error")
+				// fallthrough
+			} else {
+				placeholderRestored = true
+			}
+			// fallthrough
+		}
+
+		// deal with failing initial encrypted send & recv
+		if _, ok := err.(*zfs.RecvDestroyOrOverwriteEncryptedErr); ok && ph.IsPlaceholder && placeholderRestored {
+			msg := `cannot automatically replace placeholder filesystem with incoming send stream - please see receive-side log for details`
+			err := errors.New(msg)
+			log.Error(msg)
+
+			log.Error(`zrepl creates placeholder filesystems on the receiving side of a replication to match the sending side's dataset hierarchy`)
+			log.Error(`zrepl uses zfs receive -F to replace those placeholders with incoming full sends`)
+			log.Error(`OpenZFS native encryption prohibits zfs receive -F for encrypted filesystems`)
+			log.Error(`the current zrepl placeholder filesystem concept is thus incompatible with OpenZFS native encryption`)
+
+			tempStartFullRecvFS := lp.Copy().ToString() + ".zrepl.initial-recv"
+			tempStartFullRecvFSDP, dpErr := zfs.NewDatasetPath(tempStartFullRecvFS)
+			if dpErr != nil {
+				log.WithError(dpErr).Error("cannot determine temporary filesystem name for initial encrypted recv workaround")
+				return nil, err // yes, err, not dpErr
+			}
+
+			log := log.WithField("temp_recv_fs", tempStartFullRecvFS)
+			log.Error(`as a workaround, zrepl will now attempt to re-receive the beginning of the stream into a temporary filesystem temp_recv_fs`)
+			log.Error(`if that step succeeds: shut down zrepl and use 'zfs rename' to swap temp_recv_fs with local_fs, then restart zrepl`)
+			log.Error(`replication will then resume using resumable send+recv`)
+
+			tempPH, phErr := zfs.ZFSGetFilesystemPlaceholderState(ctx, tempStartFullRecvFSDP)
+			if phErr != nil {
+				log.WithError(phErr).Error("cannot determine placeholder state of temp_recv_fs")
+				return nil, err // yes, err, not dpErr
+			}
+			if tempPH.FSExists {
+				log.Error("temp_recv_fs already exists, assuming a (partial) initial recv to that filesystem has already been done")
+				return nil, err
+			}
+
+			recvOpts.RollbackAndForceRecv = false
+			recvOpts.SavePartialRecvState = true
+			rerecvErr := zfs.ZFSRecv(ctx, tempStartFullRecvFS, to, chainedio.NewChainedReader(&peekCopy), recvOpts)
+			if _, isResumable := rerecvErr.(*zfs.RecvFailedWithResumeTokenErr); rerecvErr == nil || isResumable {
+				log.Error("completed re-receive into temporary filesystem temp_recv_fs, now shut down zrepl and use zfs rename to swap temp_recv_fs with local_fs")
+			} else {
+				log.WithError(rerecvErr).Error("failed to receive the beginning of the stream into temporary filesystem temp_recv_fs")
+				log.Error("we advise you to collect the error log and current configuration, open an issue on GitHub, and revert to your previous configuration in the meantime")
+			}
+
+			log.Error(`if you would like to see improvements to this situation, please open an issue on GitHub`)
+			return nil, err
+		}
+
+		log.
 			WithError(err).
-			WithField("opts", recvOpts).
+			WithField("opts", fmt.Sprintf("%#v", recvOpts)).
 			Error("zfs receive failed")
+
 		return nil, err
 	}
 
@@ -815,13 +902,13 @@ func (s *Receiver) Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs
 	toRecvd, err := to.ValidateExistsAndGetVersion(ctx, lp.ToString())
 	if err != nil {
 		msg := "receive request's `To` version does not match what we received in the stream"
-		getLogger(ctx).WithError(err).WithField("snap", snapFullPath).Error(msg)
-		getLogger(ctx).Error("aborting recv request, but keeping received snapshot for inspection")
+		log.WithError(err).WithField("snap", snapFullPath).Error(msg)
+		log.Error("aborting recv request, but keeping received snapshot for inspection")
 		return nil, errors.Wrap(err, msg)
 	}
 
 	if s.conf.UpdateLastReceivedHold {
-		getLogger(ctx).Debug("move last-received-hold")
+		log.Debug("move last-received-hold")
 		if err := MoveLastReceivedHold(ctx, lp.ToString(), toRecvd, s.conf.JobID); err != nil {
 			return nil, errors.Wrap(err, "cannot move last-received-hold")
 		}

endpoint/endpoint_zfs_abstraction.go

@@ -573,7 +573,7 @@ func listAbstractionsImplFS(ctx context.Context, fs string, query *ListZFSHoldsA
 			whatTypes[zfs.Snapshot] = true
 		}
 	}
-	fsvs, err := zfs.ZFSListFilesystemVersions(fsp, zfs.ListFilesystemVersionsOptions{
+	fsvs, err := zfs.ZFSListFilesystemVersions(ctx, fsp, zfs.ListFilesystemVersionsOptions{
 		Types: whatTypes,
 	})
 	if err != nil {

platformtest/tests/helpers.go

@@ -136,7 +136,7 @@ func makeResumeSituation(ctx *platformtest.Context, src dummySnapshotSituation,
 		return situation
 	}
 
-	limitedCopier := zfs.NewReadCloserCopier(limitio.ReadCloser(copier, src.dummyDataLen/2))
+	limitedCopier := limitio.ReadCloser(copier, src.dummyDataLen/2)
 	defer limitedCopier.Close()
 
 	require.NotNil(ctx, sendArgs.To)

platformtest/tests/listFilesystemVersions.go

@@ -43,7 +43,7 @@ func ListFilesystemVersionsTypeFilteringAndPrefix(t *platformtest.Context) {
 	fs := fmt.Sprintf("%s/foo bar", t.RootDataset)
 
 	// no options := all types
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
 	require.NoError(t, err)
 	require.Equal(t, []string{
 		"#blup 1", "#bookfoo 1", "#bookfoo 2", "#foo 1", "#foo 2",
@@ -51,21 +51,21 @@ func ListFilesystemVersionsTypeFilteringAndPrefix(t *platformtest.Context) {
 	}, versionRelnamesSorted(vs))
 
 	// just snapshots
-	vs, err = zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
+	vs, err = zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
 		Types: zfs.Snapshots,
 	})
 	require.NoError(t, err)
 	require.Equal(t, []string{"@ foo with leading whitespace", "@blup 1", "@foo 1", "@foo 2"}, versionRelnamesSorted(vs))
 
 	// just bookmarks
-	vs, err = zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
+	vs, err = zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
 		Types: zfs.Bookmarks,
 	})
 	require.NoError(t, err)
 	require.Equal(t, []string{"#blup 1", "#bookfoo 1", "#bookfoo 2", "#foo 1", "#foo 2"}, versionRelnamesSorted(vs))
 
 	// just with prefix foo
-	vs, err = zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
+	vs, err = zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{
 		ShortnamePrefix: "foo",
 	})
 	require.NoError(t, err)
@@ -82,7 +82,7 @@ func ListFilesystemVersionsZeroExistIsNotAnError(t *platformtest.Context) {
 
 	fs := fmt.Sprintf("%s/foo bar", t.RootDataset)
 
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
 	require.Empty(t, vs)
 	require.NoError(t, err)
 	dsne, ok := err.(*zfs.DatasetDoesNotExist)
@@ -98,7 +98,7 @@ func ListFilesystemVersionsFilesystemNotExist(t *platformtest.Context) {
 
 	nonexistentFS := fmt.Sprintf("%s/not existent", t.RootDataset)
 
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(nonexistentFS), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(nonexistentFS), zfs.ListFilesystemVersionsOptions{})
 	require.Empty(t, vs)
 	require.Error(t, err)
 	t.Logf("err = %T\n%s", err, err)
@@ -141,7 +141,7 @@ func ListFilesystemVersionsUserrefs(t *platformtest.Context) {
 
 	fs := fmt.Sprintf("%s/foo bar", t.RootDataset)
 
-	vs, err := zfs.ZFSListFilesystemVersions(mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
+	vs, err := zfs.ZFSListFilesystemVersions(t, mustDatasetPath(fs), zfs.ListFilesystemVersionsOptions{})
 	require.NoError(t, err)
 
 	type expectation struct {

platformtest/tests/sendArgsValidation.go

@@ -35,7 +35,7 @@ func SendArgsValidationEncryptedSendOfUnencryptedDatasetForbidden(ctx *platformt
 		ResumeToken: "",
 	}.Validate(ctx)
 
-	var stream *zfs.ReadCloserCopier
+	var stream *zfs.SendStream
 	if err == nil {
 		stream, err = zfs.ZFSSend(ctx, sendArgs) // no shadow
 		if err == nil {

replication/driver/replication_driver.go

@@ -146,6 +146,12 @@ type fs struct {
 
 	l *chainlock.L
 
+	// ordering relationship that must be maintained for initial replication
+	initialRepOrd struct {
+		parents, children []*fs
+		parentDidUpdate   chan struct{}
+	}
+
 	planning struct {
 		done bool
 		err  *timedError
@@ -296,6 +302,7 @@ func (a *attempt) do(ctx context.Context, prev *attempt) {
 			fs: pfs,
 			l:  a.l,
 		}
+		fs.initialRepOrd.parentDidUpdate = make(chan struct{}, 1)
 		a.fss = append(a.fss, fs)
 	}
 
@@ -354,8 +361,20 @@ func (a *attempt) do(ctx context.Context, prev *attempt) {
 	}
 	// invariant: prevs contains an entry for each unambiguous correspondence
 
+	// build up parent-child relationship (FIXME (O(n^2), but who's going to have that many filesystems...))
+	for _, f1 := range a.fss {
+		fs1 := f1.fs.ReportInfo().Name
+		for _, f2 := range a.fss {
+			fs2 := f2.fs.ReportInfo().Name
+			if strings.HasPrefix(fs1, fs2) && fs1 != fs2 {
+				f1.initialRepOrd.parents = append(f1.initialRepOrd.parents, f2)
+				f2.initialRepOrd.children = append(f2.initialRepOrd.children, f1)
+			}
+		}
+	}
+
 	stepQueue := newStepQueue()
-	defer stepQueue.Start(1)() // TODO parallel replication
+	defer stepQueue.Start(envconst.Int("ZREPL_REPLICATION_EXPERIMENTAL_REPLICATION_CONCURRENCY", 1))() // TODO parallel replication
 	var fssesDone sync.WaitGroup
 	for _, f := range a.fss {
 		fssesDone.Add(1)
@@ -370,54 +389,76 @@ func (a *attempt) do(ctx context.Context, prev *attempt) {
 	a.finishedAt = time.Now()
 }
 
-func (fs *fs) do(ctx context.Context, pq *stepQueue, prev *fs) {
+func (f *fs) debug(format string, args ...interface{}) {
+	debugPrefix("fs=%s", f.fs.ReportInfo().Name)(format, args...)
+}
 
-	defer fs.l.Lock().Unlock()
+// wake up children that watch for f.{planning.{err,done},planned.{step,stepErr}}
+func (f *fs) initialRepOrdWakeupChildren() {
+	var children []string
+	for _, c := range f.initialRepOrd.children {
+		// no locking required, c.fs does not change
+		children = append(children, c.fs.ReportInfo().Name)
+	}
+	f.debug("wakeup children %s", children)
+	for _, child := range f.initialRepOrd.children {
+		select {
+		// no locking required, child.initialRepOrd does not change
+		case child.initialRepOrd.parentDidUpdate <- struct{}{}:
+		default:
+		}
+	}
+}
+
+func (f *fs) do(ctx context.Context, pq *stepQueue, prev *fs) {
+
+	defer f.l.Lock().Unlock()
+	defer f.initialRepOrdWakeupChildren()
 
 	// get planned steps from replication logic
 	var psteps []Step
 	var errTime time.Time
 	var err error
-	fs.l.DropWhile(func() {
+	f.l.DropWhile(func() {
 		// TODO hacky
 		// choose target time that is earlier than any snapshot, so fs planning is always prioritized
 		targetDate := time.Unix(0, 0)
-		defer pq.WaitReady(fs, targetDate)()
-		psteps, err = fs.fs.PlanFS(ctx) // no shadow
-		errTime = time.Now()            // no shadow
+		defer pq.WaitReady(f, targetDate)()
+		psteps, err = f.fs.PlanFS(ctx) // no shadow
+		errTime = time.Now()           // no shadow
 	})
-	debug := debugPrefix("fs=%s", fs.fs.ReportInfo().Name)
-	fs.planning.done = true
 	if err != nil {
-		fs.planning.err = newTimedError(err, errTime)
+		f.planning.err = newTimedError(err, errTime)
 		return
 	}
 	for _, pstep := range psteps {
 		step := &step{
-			l:    fs.l,
+			l:    f.l,
 			step: pstep,
 		}
-		fs.planned.steps = append(fs.planned.steps, step)
+		f.planned.steps = append(f.planned.steps, step)
 	}
-	debug("initial len(fs.planned.steps) = %d", len(fs.planned.steps))
+	// we're not done planning yet, f.planned.steps might still be changed by next block
+	// => don't set f.planning.done just yet
+	f.debug("initial len(fs.planned.steps) = %d", len(f.planned.steps))
 
 	// for not-first attempts, only allow fs.planned.steps
 	// up to including the originally planned target snapshot
 	if prev != nil && prev.planning.done && prev.planning.err == nil {
 		prevUncompleted := prev.planned.steps[prev.planned.step:]
 		if len(prevUncompleted) == 0 {
-			debug("prevUncompleted is empty")
+			f.debug("prevUncompleted is empty")
 			return
 		}
-		if len(fs.planned.steps) == 0 {
-			debug("fs.planned.steps is empty")
+		if len(f.planned.steps) == 0 {
+			f.debug("fs.planned.steps is empty")
 			return
 		}
 		prevFailed := prevUncompleted[0]
-		curFirst := fs.planned.steps[0]
+		curFirst := f.planned.steps[0]
 		// we assume that PlanFS retries prevFailed (using curFirst)
 		if !prevFailed.step.TargetEquals(curFirst.step) {
-			debug("Targets don't match")
+			f.debug("Targets don't match")
 			// Two options:
 			// A: planning algorithm is broken
 			// B: manual user intervention inbetween
@@ -433,44 +474,130 @@ func (fs *fs) do(ctx context.Context, pq *stepQueue, prev *fs) {
 			}
 			msg := fmt.Sprintf("last attempt's uncompleted step %s does not correspond to this attempt's first planned step %s",
 				stepFmt(prevFailed), stepFmt(curFirst))
-			fs.planned.stepErr = newTimedError(errors.New(msg), time.Now())
+			f.planned.stepErr = newTimedError(errors.New(msg), time.Now())
 			return
 		}
 		// only allow until step targets diverge
 		min := len(prevUncompleted)
-		if min > len(fs.planned.steps) {
-			min = len(fs.planned.steps)
+		if min > len(f.planned.steps) {
+			min = len(f.planned.steps)
 		}
 		diverge := 0
 		for ; diverge < min; diverge++ {
-			debug("diverge compare iteration %d", diverge)
-			if !fs.planned.steps[diverge].step.TargetEquals(prevUncompleted[diverge].step) {
+			f.debug("diverge compare iteration %d", diverge)
+			if !f.planned.steps[diverge].step.TargetEquals(prevUncompleted[diverge].step) {
 				break
 			}
 		}
-		debug("diverge is %d", diverge)
-		fs.planned.steps = fs.planned.steps[0:diverge]
+		f.debug("diverge is %d", diverge)
+		f.planned.steps = f.planned.steps[0:diverge]
 	}
-	debug("post-prev-merge len(fs.planned.steps) = %d", len(fs.planned.steps))
+	f.debug("post-prev-merge len(fs.planned.steps) = %d", len(f.planned.steps))
 
-	for i, s := range fs.planned.steps {
-		var (
-			err     error
-			errTime time.Time
-		)
-		// lock must not be held while executing step in order for reporting to work
-		fs.l.DropWhile(func() {
-			targetDate := s.step.TargetDate()
-			defer pq.WaitReady(fs, targetDate)()
-			err = s.step.Step(ctx) // no shadow
-			errTime = time.Now()   // no shadow
+	// now we are done planning (f.planned.steps won't change from now on)
+	f.planning.done = true
+
+	// wait for parents' initial replication
+	var parents []string
+	for _, p := range f.initialRepOrd.parents {
+		parents = append(parents, p.fs.ReportInfo().Name)
+	}
+	f.debug("wait for parents %s", parents)
+	for {
+		var initialReplicatingParentsWithErrors []string
+		allParentsPresentOnReceiver := true
+		f.l.DropWhile(func() {
+			for _, p := range f.initialRepOrd.parents {
+				p.l.HoldWhile(func() {
+					// (get the preconditions that allow us to inspect p.planned)
+					parentHasPlanningDone := p.planning.done && p.planning.err == nil
+					if !parentHasPlanningDone {
+						// if the parent couldn't be planned, we cannot know whether it needs initial replication
+						// or incremental replication => be conservative and assume it was initial replication
+						allParentsPresentOnReceiver = false
+						if p.planning.err != nil {
+							initialReplicatingParentsWithErrors = append(initialReplicatingParentsWithErrors, p.fs.ReportInfo().Name)
+						}
+						return
+					}
+					// now allowed to inspect p.planned
+
+					// if there are no steps to be done, the filesystem must exist on the receiving side
+					// (otherwise we'd replicate it, and there would be a step for that)
+					// (FIXME hardcoded initial replication policy, assuming the policy will always do _some_ initial replication)
+					parentHasNoSteps := len(p.planned.steps) == 0
+
+					// OR if it has completed at least one step
+					// (remember that .step points to the next step to be done)
+					// (TODO technically, we could make this step ready in the moment the recv-side
+					//  dataset exists, i.e. after the first few megabytes of transferred data, but we'd have to ask the receiver for that -> poll ListFilesystems RPC)
+					parentHasTakenAtLeastOneSuccessfulStep := !parentHasNoSteps && p.planned.step >= 1
+
+					parentFirstStepIsIncremental := // no need to lock for .report() because step.l == it's fs.l
+						len(p.planned.steps) > 0 && p.planned.steps[0].report().IsIncremental()
+
+					f.debug("parentHasNoSteps=%v parentFirstStepIsIncremental=%v parentHasTakenAtLeastOneSuccessfulStep=%v",
+						parentHasNoSteps, parentFirstStepIsIncremental, parentHasTakenAtLeastOneSuccessfulStep)
+
+					parentPresentOnReceiver := parentHasNoSteps || parentFirstStepIsIncremental || parentHasTakenAtLeastOneSuccessfulStep
+
+					allParentsPresentOnReceiver = allParentsPresentOnReceiver && parentPresentOnReceiver // no shadow
+
+					if !parentPresentOnReceiver && p.planned.stepErr != nil {
+						initialReplicatingParentsWithErrors = append(initialReplicatingParentsWithErrors, p.fs.ReportInfo().Name)
+					}
+
+				})
+			}
 		})
+
+		if len(initialReplicatingParentsWithErrors) > 0 {
+			f.planned.stepErr = newTimedError(fmt.Errorf("parent(s) failed during initial replication: %s", initialReplicatingParentsWithErrors), time.Now())
+			return
+		}
+
+		if allParentsPresentOnReceiver {
+			break // good to go
+		}
+
+		// wait for wakeups from parents, then check again
+		// lock must not be held while waiting in order for reporting to work
+		f.l.DropWhile(func() {
+			select {
+			case <-ctx.Done():
+				f.planned.stepErr = newTimedError(ctx.Err(), time.Now())
+				return
+			case <-f.initialRepOrd.parentDidUpdate:
+				// loop
+			}
+		})
+		if f.planned.stepErr != nil {
+			return
+		}
+	}
+
+	f.debug("all parents ready, start replication %s", parents)
+
+	// do our steps
+	for i, s := range f.planned.steps {
+		// lock must not be held while executing step in order for reporting to work
+		f.l.DropWhile(func() {
+			// wait for parallel replication
+			targetDate := s.step.TargetDate()
+			defer pq.WaitReady(f, targetDate)()
+			// do the step
+			err, errTime = s.step.Step(ctx), time.Now() // no shadow
+		})
+
 		if err != nil {
-			fs.planned.stepErr = newTimedError(err, errTime)
+			f.planned.stepErr = newTimedError(err, errTime)
 			break
 		}
-		fs.planned.step = i + 1 // fs.planned.step must be == len(fs.planned.steps) if all went OK
+		f.planned.step = i + 1 // fs.planned.step must be == len(fs.planned.steps) if all went OK
+
+		f.initialRepOrdWakeupChildren()
 	}
+
 }
 
 // caller must hold lock l

replication/driver/replication_driver_debug.go

@@ -26,6 +26,6 @@ type debugFunc func(format string, args ...interface{})
 func debugPrefix(prefixFormat string, prefixFormatArgs ...interface{}) debugFunc {
 	prefix := fmt.Sprintf(prefixFormat, prefixFormatArgs...)
 	return func(format string, args ...interface{}) {
-		debug("%s: %s", prefix, fmt.Sprintf(format, args))
+		debug("%s: %s", prefix, fmt.Sprintf(format, args...))
 	}
 }

replication/logic/replication_logic.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"io"
 	"sync"
 	"time"
 
@@ -38,7 +39,7 @@ type Sender interface {
 	// If a non-nil io.ReadCloser is returned, it is guaranteed to be closed before
 	// any next call to the parent github.com/zrepl/zrepl/replication.Endpoint.
 	// If the send request is for dry run the io.ReadCloser will be nil
-	Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error)
+	Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error)
 	SendCompleted(ctx context.Context, r *pdu.SendCompletedReq) (*pdu.SendCompletedRes, error)
 	ReplicationCursor(ctx context.Context, req *pdu.ReplicationCursorReq) (*pdu.ReplicationCursorRes, error)
 }
@@ -47,7 +48,7 @@ type Receiver interface {
 	Endpoint
 	// Receive sends r and sendStream (the latter containing a ZFS send stream)
 	// to the parent github.com/zrepl/zrepl/replication.Endpoint.
-	Receive(ctx context.Context, req *pdu.ReceiveReq, receive zfs.StreamCopier) (*pdu.ReceiveRes, error)
+	Receive(ctx context.Context, req *pdu.ReceiveReq, receive io.ReadCloser) (*pdu.ReceiveRes, error)
 }
 
 type PlannerPolicy struct {
@@ -162,7 +163,7 @@ type Step struct {
 
 	// byteCounter is nil initially, and set later in Step.doReplication
 	// => concurrent read of that pointer from Step.ReportInfo must be protected
-	byteCounter    bytecounter.StreamCopier
+	byteCounter    bytecounter.ReadCloser
 	byteCounterMtx chainlock.L
 }
 
@@ -606,19 +607,19 @@ func (s *Step) doReplication(ctx context.Context) error {
 	sr := s.buildSendRequest(false)
 
 	log.Debug("initiate send request")
-	sres, sstreamCopier, err := s.sender.Send(ctx, sr)
+	sres, stream, err := s.sender.Send(ctx, sr)
 	if err != nil {
 		log.WithError(err).Error("send request failed")
 		return err
 	}
-	if sstreamCopier == nil {
+	if stream == nil {
 		err := errors.New("send request did not return a stream, broken endpoint implementation")
 		return err
 	}
-	defer sstreamCopier.Close()
+	defer stream.Close()
 
 	// Install a byte counter to track progress + for status report
-	byteCountingStream := bytecounter.NewStreamCopier(sstreamCopier)
+	byteCountingStream := bytecounter.NewReadCloser(stream)
 	s.byteCounterMtx.Lock()
 	s.byteCounter = byteCountingStream
 	s.byteCounterMtx.Unlock()

rpc/dataconn/dataconn_client.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"context"
 	"fmt"
+	"io"
 	"strings"
 
 	"github.com/golang/protobuf/proto"
@@ -11,7 +12,6 @@ import (
 	"github.com/zrepl/zrepl/replication/logic/pdu"
 	"github.com/zrepl/zrepl/rpc/dataconn/stream"
 	"github.com/zrepl/zrepl/transport"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 type Client struct {
@@ -26,7 +26,7 @@ func NewClient(connecter transport.Connecter, log Logger) *Client {
 	}
 }
 
-func (c *Client) send(ctx context.Context, conn *stream.Conn, endpoint string, req proto.Message, streamCopier zfs.StreamCopier) error {
+func (c *Client) send(ctx context.Context, conn *stream.Conn, endpoint string, req proto.Message, stream io.ReadCloser) error {
 
 	var buf bytes.Buffer
 	_, memErr := buf.WriteString(endpoint)
@@ -46,8 +46,8 @@ func (c *Client) send(ctx context.Context, conn *stream.Conn, endpoint string, r
 		return err
 	}
 
-	if streamCopier != nil {
-		return conn.SendStream(ctx, streamCopier, ZFSStream)
+	if stream != nil {
+		return conn.SendStream(ctx, stream, ZFSStream)
 	} else {
 		return nil
 	}
@@ -109,7 +109,7 @@ func (c *Client) putWire(conn *stream.Conn) {
 	}
 }
 
-func (c *Client) ReqSend(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (c *Client) ReqSend(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
 	conn, err := c.getWire(ctx)
 	if err != nil {
 		return nil, nil, err
@@ -130,16 +130,19 @@ func (c *Client) ReqSend(ctx context.Context, req *pdu.SendReq) (*pdu.SendRes, z
 		return nil, nil, err
 	}
 
-	var copier zfs.StreamCopier = nil
+	var stream io.ReadCloser
 	if !req.DryRun {
 		putWireOnReturn = false
-		copier = &streamCopier{streamConn: conn, closeStreamOnClose: true}
+		stream, err = conn.ReadStream(ZFSStream, true) // no shadow
+		if err != nil {
+			return nil, nil, err
+		}
 	}
 
-	return &res, copier, nil
+	return &res, stream, nil
 }
 
-func (c *Client) ReqRecv(ctx context.Context, req *pdu.ReceiveReq, streamCopier zfs.StreamCopier) (*pdu.ReceiveRes, error) {
+func (c *Client) ReqRecv(ctx context.Context, req *pdu.ReceiveReq, stream io.ReadCloser) (*pdu.ReceiveRes, error) {
 
 	defer c.log.Debug("ReqRecv returns")
 	conn, err := c.getWire(ctx)
@@ -166,7 +169,7 @@ func (c *Client) ReqRecv(ctx context.Context, req *pdu.ReceiveReq, streamCopier
 
 	sendErrChan := make(chan error)
 	go func() {
-		if err := c.send(ctx, conn, EndpointRecv, req, streamCopier); err != nil {
+		if err := c.send(ctx, conn, EndpointRecv, req, stream); err != nil {
 			sendErrChan <- err
 		} else {
 			sendErrChan <- nil

rpc/dataconn/dataconn_server.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"context"
 	"fmt"
+	"io"
 
 	"github.com/golang/protobuf/proto"
 
@@ -11,7 +12,6 @@ import (
 	"github.com/zrepl/zrepl/replication/logic/pdu"
 	"github.com/zrepl/zrepl/rpc/dataconn/stream"
 	"github.com/zrepl/zrepl/transport"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 // WireInterceptor has a chance to exchange the context and connection on each client connection.
@@ -21,11 +21,11 @@ type WireInterceptor func(ctx context.Context, rawConn *transport.AuthConn) (con
 type Handler interface {
 	// Send handles a SendRequest.
 	// The returned io.ReadCloser is allowed to be nil, for example if the requested Send is a dry-run.
-	Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error)
+	Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error)
 	// Receive handles a ReceiveRequest.
 	// It is guaranteed that Server calls Receive with a stream that holds the IdleConnTimeout
 	// configured in ServerConfig.Shared.IdleConnTimeout.
-	Receive(ctx context.Context, r *pdu.ReceiveReq, receive zfs.StreamCopier) (*pdu.ReceiveRes, error)
+	Receive(ctx context.Context, r *pdu.ReceiveReq, receive io.ReadCloser) (*pdu.ReceiveRes, error)
 	// PingDataconn handles a PingReq
 	PingDataconn(ctx context.Context, r *pdu.PingReq) (*pdu.PingRes, error)
 }
@@ -111,7 +111,7 @@ func (s *Server) serveConn(nc *transport.AuthConn) {
 	s.log.WithField("endpoint", endpoint).Debug("calling handler")
 
 	var res proto.Message
-	var sendStream zfs.StreamCopier
+	var sendStream io.ReadCloser
 	var handlerErr error
 	switch endpoint {
 	case EndpointSend:
@@ -127,7 +127,12 @@ func (s *Server) serveConn(nc *transport.AuthConn) {
 			s.log.WithError(err).Error("cannot unmarshal receive request")
 			return
 		}
-		res, handlerErr = s.h.Receive(ctx, &req, &streamCopier{streamConn: c, closeStreamOnClose: false}) // SHADOWING
+		stream, err := c.ReadStream(ZFSStream, false)
+		if err != nil {
+			s.log.WithError(err).Error("cannot open stream in receive request")
+			return
+		}
+		res, handlerErr = s.h.Receive(ctx, &req, stream) // SHADOWING
 	case EndpointPing:
 		var req pdu.PingReq
 		if err := proto.Unmarshal(reqStructured, &req); err != nil {

rpc/dataconn/dataconn_shared.go

@@ -1,12 +1,7 @@
 package dataconn
 
 import (
-	"io"
-	"sync"
 	"time"
-
-	"github.com/zrepl/zrepl/rpc/dataconn/stream"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 const (
@@ -39,33 +34,3 @@ const (
 	responseHeaderHandlerOk          = "HANDLER OK\n"
 	responseHeaderHandlerErrorPrefix = "HANDLER ERROR:\n"
 )
-
-type streamCopier struct {
-	mtx                sync.Mutex
-	used               bool
-	streamConn         *stream.Conn
-	closeStreamOnClose bool
-}
-
-// WriteStreamTo implements zfs.StreamCopier
-func (s *streamCopier) WriteStreamTo(w io.Writer) zfs.StreamCopierError {
-	s.mtx.Lock()
-	defer s.mtx.Unlock()
-	if s.used {
-		panic("streamCopier used multiple times")
-	}
-	s.used = true
-	return s.streamConn.ReadStreamInto(w, ZFSStream)
-}
-
-// Close implements zfs.StreamCopier
-func (s *streamCopier) Close() error {
-	// only record the close here, what we do actually depends on whether
-	// the streamCopier is instantiated server-side or client-side
-	s.mtx.Lock()
-	defer s.mtx.Unlock()
-	if s.closeStreamOnClose {
-		return s.streamConn.Close()
-	}
-	return nil
-}

rpc/dataconn/microbenchmark/microbenchmark.go

@@ -29,7 +29,6 @@ import (
 	"github.com/zrepl/zrepl/rpc/dataconn/timeoutconn"
 	"github.com/zrepl/zrepl/transport"
 	"github.com/zrepl/zrepl/util/devnoop"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 func orDie(err error) {
@@ -42,23 +41,9 @@ type readerStreamCopier struct{ io.Reader }
 
 func (readerStreamCopier) Close() error { return nil }
 
-type readerStreamCopierErr struct {
-	error
-}
-
-func (readerStreamCopierErr) IsReadError() bool  { return false }
-func (readerStreamCopierErr) IsWriteError() bool { return true }
-
-func (c readerStreamCopier) WriteStreamTo(w io.Writer) zfs.StreamCopierError {
-	var buf [1 << 21]byte
-	_, err := io.CopyBuffer(w, c.Reader, buf[:])
-	// always assume write error
-	return readerStreamCopierErr{err}
-}
-
 type devNullHandler struct{}
 
-func (devNullHandler) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (devNullHandler) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
 	var res pdu.SendRes
 	if args.devnoopReader {
 		return &res, readerStreamCopier{devnoop.Get()}, nil
@@ -67,12 +52,12 @@ func (devNullHandler) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, z
 	}
 }
 
-func (devNullHandler) Receive(ctx context.Context, r *pdu.ReceiveReq, stream zfs.StreamCopier) (*pdu.ReceiveRes, error) {
+func (devNullHandler) Receive(ctx context.Context, r *pdu.ReceiveReq, stream io.ReadCloser) (*pdu.ReceiveRes, error) {
 	var out io.Writer = os.Stdout
 	if args.devnoopWriter {
 		out = devnoop.Get()
 	}
-	err := stream.WriteStreamTo(out)
+	_, err := io.Copy(out, stream)
 	var res pdu.ReceiveRes
 	return &res, err
 }
@@ -172,7 +157,7 @@ func client() {
 		req := pdu.SendReq{}
 		_, stream, err := client.ReqSend(ctx, &req)
 		orDie(err)
-		err = stream.WriteStreamTo(os.Stdout)
+		_, err = io.Copy(os.Stdout, stream)
 		orDie(err)
 	case "recv":
 		var r io.Reader = os.Stdin

rpc/dataconn/stream/stream.go

@@ -14,7 +14,6 @@ import (
 	"github.com/zrepl/zrepl/rpc/dataconn/base2bufpool"
 	"github.com/zrepl/zrepl/rpc/dataconn/frameconn"
 	"github.com/zrepl/zrepl/rpc/dataconn/heartbeatconn"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 type Logger = logger.Logger
@@ -198,7 +197,7 @@ func (e ReadStreamError) Temporary() bool {
 	return false
 }
 
-var _ zfs.StreamCopierError = &ReadStreamError{}
+var _ net.Error = &ReadStreamError{}
 
 func (e ReadStreamError) IsReadError() bool {
 	return e.Kind != ReadStreamErrorKindWrite

rpc/dataconn/stream/stream_conn.go

@@ -14,7 +14,6 @@ import (
 
 	"github.com/zrepl/zrepl/rpc/dataconn/heartbeatconn"
 	"github.com/zrepl/zrepl/rpc/dataconn/timeoutconn"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 type Conn struct {
@@ -40,15 +39,7 @@ type Conn struct {
 
 var readMessageSentinel = fmt.Errorf("read stream complete")
 
-type writeStreamToErrorUnknownState struct{}
-
-func (e writeStreamToErrorUnknownState) Error() string {
-	return "dataconn read stream: connection is in unknown state"
-}
-
-func (e writeStreamToErrorUnknownState) IsReadError() bool { return true }
-
-func (e writeStreamToErrorUnknownState) IsWriteError() bool { return false }
+var errWriteStreamToErrorUnknownState = fmt.Errorf("dataconn read stream: connection is in unknown state")
 
 func Wrap(nc timeoutconn.Wire, sendHeartbeatInterval, peerTimeout time.Duration) *Conn {
 	hc := heartbeatconn.Wrap(nc, sendHeartbeatInterval, peerTimeout)
@@ -123,14 +114,28 @@ func (c *Conn) ReadStreamedMessage(ctx context.Context, maxSize uint32, frameTyp
 	}
 }
 
+type StreamReader struct {
+	*io.PipeReader
+	conn             *Conn
+	closeConnOnClose bool
+}
+
+func (r *StreamReader) Close() error {
+	err := r.PipeReader.Close()
+	if r.closeConnOnClose {
+		r.conn.Close() // TODO error logging
+	}
+	return err
+}
+
 // WriteStreamTo reads a stream from Conn and writes it to w.
-func (c *Conn) ReadStreamInto(w io.Writer, frameType uint32) (err zfs.StreamCopierError) {
+func (c *Conn) ReadStream(frameType uint32, closeConnOnClose bool) (_ *StreamReader, err error) {
 
 	// if we are closed while writing, return that as an error
 	if closeGuard, cse := c.closeState.RWEntry(); cse != nil {
-		return cse
+		return nil, cse
 	} else {
-		defer func(err *zfs.StreamCopierError) {
+		defer func(err *error) {
 			if closed := closeGuard.RWExit(); closed != nil {
 				*err = closed
 			}
@@ -138,18 +143,23 @@ func (c *Conn) ReadStreamInto(w io.Writer, frameType uint32) (err zfs.StreamCopi
 	}
 
 	c.readMtx.Lock()
-	defer c.readMtx.Unlock()
 	if !c.readClean {
-		return writeStreamToErrorUnknownState{}
+		return nil, errWriteStreamToErrorUnknownState
 	}
-	var rse *ReadStreamError = readStream(c.frameReads, c.hc, w, frameType)
-	c.readClean = isConnCleanAfterRead(rse)
 
-	// https://golang.org/doc/faq#nil_error
-	if rse == nil {
-		return nil
-	}
-	return rse
+	r, w := io.Pipe()
+	go func() {
+		defer c.readMtx.Unlock()
+		var err *ReadStreamError = readStream(c.frameReads, c.hc, w, frameType)
+		if err != nil {
+			_ = w.CloseWithError(err) // doc guarantees that error will always be nil
+		} else {
+			w.Close()
+		}
+		c.readClean = isConnCleanAfterRead(err)
+	}()
+
+	return &StreamReader{PipeReader: r, conn: c, closeConnOnClose: closeConnOnClose}, nil
 }
 
 func (c *Conn) WriteStreamedMessage(ctx context.Context, buf io.Reader, frameType uint32) (err error) {
@@ -178,7 +188,7 @@ func (c *Conn) WriteStreamedMessage(ctx context.Context, buf io.Reader, frameTyp
 	return errConn
 }
 
-func (c *Conn) SendStream(ctx context.Context, src zfs.StreamCopier, frameType uint32) (err error) {
+func (c *Conn) SendStream(ctx context.Context, stream io.ReadCloser, frameType uint32) (err error) {
 
 	// if we are closed while reading, return that as an error
 	if closeGuard, cse := c.closeState.RWEntry(); cse != nil {
@@ -197,49 +207,17 @@ func (c *Conn) SendStream(ctx context.Context, src zfs.StreamCopier, frameType u
 		return fmt.Errorf("dataconn send stream: connection is in unknown state")
 	}
 
-	// avoid io.Pipe if zfs.StreamCopier is an io.Reader
-	var r io.Reader
-	var w *io.PipeWriter
-	streamCopierErrChan := make(chan zfs.StreamCopierError, 1)
-	if reader, ok := src.(io.Reader); ok {
-		r = reader
-		streamCopierErrChan <- nil
-		close(streamCopierErrChan)
-	} else {
-		r, w = io.Pipe()
-		go func() {
-			streamCopierErrChan <- src.WriteStreamTo(w)
-			w.Close()
-		}()
-	}
+	errStream, errConn := writeStream(ctx, c.hc, stream, frameType)
 
-	type writeStreamRes struct {
-		errStream, errConn error
-	}
-	writeStreamErrChan := make(chan writeStreamRes, 1)
-	go func() {
-		var res writeStreamRes
-		res.errStream, res.errConn = writeStream(ctx, c.hc, r, frameType)
-		if w != nil {
-			_ = w.CloseWithError(res.errStream) // always returns nil
-		}
-		writeStreamErrChan <- res
-	}()
+	c.writeClean = isConnCleanAfterWrite(errConn) // TODO correct?
 
-	writeRes := <-writeStreamErrChan
-	streamCopierErr := <-streamCopierErrChan
-	c.writeClean = isConnCleanAfterWrite(writeRes.errConn) // TODO correct?
-	if streamCopierErr != nil && streamCopierErr.IsReadError() {
-		return streamCopierErr // something on our side is bad
-	} else {
-		if writeRes.errStream != nil {
-			return writeRes.errStream
-		} else if writeRes.errConn != nil {
-			return writeRes.errConn
-		}
-		// TODO combined error?
-		return streamCopierErr
+	if errStream != nil {
+		return errStream
+	} else if errConn != nil {
+		return errConn
 	}
+	// TODO combined error?
+	return nil
 }
 
 type closeState struct {
@@ -248,17 +226,13 @@ type closeState struct {
 
 type closeStateErrConnectionClosed struct{}
 
-var _ zfs.StreamCopierError = (*closeStateErrConnectionClosed)(nil)
-var _ error = (*closeStateErrConnectionClosed)(nil)
 var _ net.Error = (*closeStateErrConnectionClosed)(nil)
 
 func (e *closeStateErrConnectionClosed) Error() string {
 	return "connection closed"
 }
-func (e *closeStateErrConnectionClosed) IsReadError() bool  { return true }
-func (e *closeStateErrConnectionClosed) IsWriteError() bool { return true }
-func (e *closeStateErrConnectionClosed) Timeout() bool      { return false }
-func (e *closeStateErrConnectionClosed) Temporary() bool    { return false }
+func (e *closeStateErrConnectionClosed) Timeout() bool   { return false }
+func (e *closeStateErrConnectionClosed) Temporary() bool { return false }
 
 func (s *closeState) CloseEntry() error {
 	firstCloser := atomic.AddUint32(&s.closeCount, 1) == 1
@@ -273,7 +247,7 @@ type closeStateEntry struct {
 	entryCount uint32
 }
 
-func (s *closeState) RWEntry() (e *closeStateEntry, err zfs.StreamCopierError) {
+func (s *closeState) RWEntry() (e *closeStateEntry, err net.Error) {
 	entry := &closeStateEntry{s, atomic.LoadUint32(&s.closeCount)}
 	if entry.entryCount > 0 {
 		return nil, &closeStateErrConnectionClosed{}
@@ -281,7 +255,7 @@ func (s *closeState) RWEntry() (e *closeStateEntry, err zfs.StreamCopierError) {
 	return entry, nil
 }
 
-func (e *closeStateEntry) RWExit() zfs.StreamCopierError {
+func (e *closeStateEntry) RWExit() net.Error {
 	if atomic.LoadUint32(&e.entryCount) == e.entryCount {
 		// no calls to Close() while running rw operation
 		return nil

rpc/rpc_client.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"io"
 	"net"
 	"sync"
 	"sync/atomic"
@@ -20,7 +21,6 @@ import (
 	"github.com/zrepl/zrepl/rpc/versionhandshake"
 	"github.com/zrepl/zrepl/transport"
 	"github.com/zrepl/zrepl/util/envconst"
-	"github.com/zrepl/zrepl/zfs"
 )
 
 // Client implements the active side of a replication setup.
@@ -82,22 +82,22 @@ func (c *Client) Close() {
 
 // callers must ensure that the returned io.ReadCloser is closed
 // TODO expose dataClient interface to the outside world
-func (c *Client) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, zfs.StreamCopier, error) {
+func (c *Client) Send(ctx context.Context, r *pdu.SendReq) (*pdu.SendRes, io.ReadCloser, error) {
 	// TODO the returned sendStream may return a read error created by the remote side
-	res, streamCopier, err := c.dataClient.ReqSend(ctx, r)
+	res, stream, err := c.dataClient.ReqSend(ctx, r)
 	if err != nil {
 		return nil, nil, err
 	}
-	if streamCopier == nil {
+	if stream == nil {
 		return res, nil, nil
 	}
 
-	return res, streamCopier, nil
+	return res, stream, nil
 
 }
 
-func (c *Client) Receive(ctx context.Context, req *pdu.ReceiveReq, streamCopier zfs.StreamCopier) (*pdu.ReceiveRes, error) {
-	return c.dataClient.ReqRecv(ctx, req, streamCopier)
+func (c *Client) Receive(ctx context.Context, req *pdu.ReceiveReq, stream io.ReadCloser) (*pdu.ReceiveRes, error) {
+	return c.dataClient.ReqRecv(ctx, req, stream)
 }
 
 func (c *Client) ListFilesystems(ctx context.Context, in *pdu.ListFilesystemReq) (*pdu.ListFilesystemRes, error) {

util/bytecounter/bytecounter_readcloser.go

@@ -0,0 +1,39 @@
+package bytecounter
+
+import (
+	"io"
+	"sync/atomic"
+)
+
+// ReadCloser wraps an io.ReadCloser, reimplementing
+// its interface and counting the bytes written to during copying.
+type ReadCloser interface {
+	io.ReadCloser
+	Count() int64
+}
+
+// NewReadCloser wraps rc.
+func NewReadCloser(rc io.ReadCloser) ReadCloser {
+	return &readCloser{rc, 0}
+}
+
+type readCloser struct {
+	rc    io.ReadCloser
+	count int64
+}
+
+func (r *readCloser) Count() int64 {
+	return atomic.LoadInt64(&r.count)
+}
+
+var _ io.ReadCloser = &readCloser{}
+
+func (r *readCloser) Close() error {
+	return r.rc.Close()
+}
+
+func (r *readCloser) Read(p []byte) (int, error) {
+	n, err := r.rc.Read(p)
+	atomic.AddInt64(&r.count, int64(n))
+	return n, err
+}

util/bytecounter/bytecounter_reader.go

@@ -1,49 +0,0 @@
-package bytecounter
-
-import (
-	"io"
-	"sync/atomic"
-	"time"
-)
-
-type ByteCounterReader struct {
-	reader io.ReadCloser
-
-	// called & accessed synchronously during Read, no external access
-	cb       func(full int64)
-	cbEvery  time.Duration
-	lastCbAt time.Time
-
-	// set atomically because it may be read by multiple threads
-	bytes int64
-}
-
-func NewByteCounterReader(reader io.ReadCloser) *ByteCounterReader {
-	return &ByteCounterReader{
-		reader: reader,
-	}
-}
-
-func (b *ByteCounterReader) SetCallback(every time.Duration, cb func(full int64)) {
-	b.cbEvery = every
-	b.cb = cb
-}
-
-func (b *ByteCounterReader) Close() error {
-	return b.reader.Close()
-}
-
-func (b *ByteCounterReader) Read(p []byte) (n int, err error) {
-	n, err = b.reader.Read(p)
-	full := atomic.AddInt64(&b.bytes, int64(n))
-	now := time.Now()
-	if b.cb != nil && now.Sub(b.lastCbAt) > b.cbEvery {
-		b.cb(full)
-		b.lastCbAt = now
-	}
-	return n, err
-}
-
-func (b *ByteCounterReader) Bytes() int64 {
-	return atomic.LoadInt64(&b.bytes)
-}

util/bytecounter/bytecounter_streamcopier.go

@@ -1,71 +0,0 @@
-package bytecounter
-
-import (
-	"io"
-	"sync/atomic"
-
-	"github.com/zrepl/zrepl/zfs"
-)
-
-// StreamCopier wraps a zfs.StreamCopier, reimplementing
-// its interface and counting the bytes written to during copying.
-type StreamCopier interface {
-	zfs.StreamCopier
-	Count() int64
-}
-
-// NewStreamCopier wraps sc into a StreamCopier.
-// If sc is io.Reader, it is guaranteed that the returned StreamCopier
-// implements that interface, too.
-func NewStreamCopier(sc zfs.StreamCopier) StreamCopier {
-	bsc := &streamCopier{sc, 0}
-	if scr, ok := sc.(io.Reader); ok {
-		return streamCopierAndReader{bsc, scr}
-	} else {
-		return bsc
-	}
-}
-
-type streamCopier struct {
-	sc    zfs.StreamCopier
-	count int64
-}
-
-// proxy writer used by streamCopier
-type streamCopierWriter struct {
-	parent *streamCopier
-	w      io.Writer
-}
-
-func (w streamCopierWriter) Write(p []byte) (n int, err error) {
-	n, err = w.w.Write(p)
-	atomic.AddInt64(&w.parent.count, int64(n))
-	return
-}
-
-func (s *streamCopier) Count() int64 {
-	return atomic.LoadInt64(&s.count)
-}
-
-var _ zfs.StreamCopier = &streamCopier{}
-
-func (s streamCopier) Close() error {
-	return s.sc.Close()
-}
-
-func (s *streamCopier) WriteStreamTo(w io.Writer) zfs.StreamCopierError {
-	ww := streamCopierWriter{s, w}
-	return s.sc.WriteStreamTo(ww)
-}
-
-// a streamCopier whose underlying sc is an io.Reader
-type streamCopierAndReader struct {
-	*streamCopier
-	asReader io.Reader
-}
-
-func (scr streamCopierAndReader) Read(p []byte) (int, error) {
-	n, err := scr.asReader.Read(p)
-	atomic.AddInt64(&scr.streamCopier.count, int64(n))
-	return n, err
-}

util/bytecounter/bytecounter_streamcopier_test.go

@@ -1,39 +0,0 @@
-package bytecounter
-
-import (
-	"io"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-
-	"github.com/zrepl/zrepl/zfs"
-)
-
-type mockStreamCopierAndReader struct {
-	zfs.StreamCopier // to satisfy interface
-	reads            int
-}
-
-func (r *mockStreamCopierAndReader) Read(p []byte) (int, error) {
-	r.reads++
-	return len(p), nil
-}
-
-var _ io.Reader = &mockStreamCopierAndReader{}
-
-func TestNewStreamCopierReexportsReader(t *testing.T) {
-	mock := &mockStreamCopierAndReader{}
-	x := NewStreamCopier(mock)
-
-	r, ok := x.(io.Reader)
-	if !ok {
-		t.Fatalf("%T does not implement io.Reader, hence reader cannout have been wrapped", x)
-	}
-
-	var buf [23]byte
-	n, err := r.Read(buf[:])
-	assert.True(t, mock.reads == 1)
-	assert.True(t, n == len(buf))
-	assert.NoError(t, err)
-	assert.True(t, x.Count() == 23)
-}

util/chainedio/chainedio_readcloser.go

@@ -0,0 +1,43 @@
+package chainedio
+
+import "io"
+
+type ChainedReadCloser struct {
+	readers   []io.Reader
+	curReader int
+}
+
+func NewChainedReader(reader ...io.Reader) *ChainedReadCloser {
+	return &ChainedReadCloser{
+		readers:   reader,
+		curReader: 0,
+	}
+}
+
+func (c *ChainedReadCloser) Read(buf []byte) (n int, err error) {
+
+	n = 0
+
+	for c.curReader < len(c.readers) {
+		n, err = c.readers[c.curReader].Read(buf)
+		if err == io.EOF {
+			c.curReader++
+			continue
+		}
+		break
+	}
+	if c.curReader == len(c.readers) {
+		err = io.EOF // actually, there was no gap
+	}
+
+	return
+}
+
+func (c *ChainedReadCloser) Close() error {
+	for _, r := range c.readers {
+		if c, ok := r.(io.Closer); ok {
+			c.Close() // TODO debug log error?
+		}
+	}
+	return nil
+}

util/chainedio/chainedio_reader.go

@@ -1,34 +0,0 @@
-package chainedio
-
-import "io"
-
-type ChainedReader struct {
-	Readers   []io.Reader
-	curReader int
-}
-
-func NewChainedReader(reader ...io.Reader) *ChainedReader {
-	return &ChainedReader{
-		Readers:   reader,
-		curReader: 0,
-	}
-}
-
-func (c *ChainedReader) Read(buf []byte) (n int, err error) {
-
-	n = 0
-
-	for c.curReader < len(c.Readers) {
-		n, err = c.Readers[c.curReader].Read(buf)
-		if err == io.EOF {
-			c.curReader++
-			continue
-		}
-		break
-	}
-	if c.curReader == len(c.Readers) {
-		err = io.EOF // actually, there was no gap
-	}
-
-	return
-}

util/chainlock/chainlock.go

@@ -40,3 +40,8 @@ func (l *L) DropWhile(f func()) {
 	defer l.Unlock().Lock()
 	f()
 }
+
+func (l *L) HoldWhile(f func()) {
+	defer l.Lock().Unlock()
+	f()
+}

zfs/versions.go

@@ -206,13 +206,13 @@ func (o *ListFilesystemVersionsOptions) matches(v FilesystemVersion) bool {
 }
 
 // returned versions are sorted by createtxg FIXME drop sort by createtxg requirement
-func ZFSListFilesystemVersions(fs *DatasetPath, options ListFilesystemVersionsOptions) (res []FilesystemVersion, err error) {
+func ZFSListFilesystemVersions(ctx context.Context, fs *DatasetPath, options ListFilesystemVersionsOptions) (res []FilesystemVersion, err error) {
 	listResults := make(chan ZFSListResult)
 
 	promTimer := prometheus.NewTimer(prom.ZFSListFilesystemVersionDuration.WithLabelValues(fs.ToString()))
 	defer promTimer.ObserveDuration()
 
-	ctx, cancel := context.WithCancel(context.Background())
+	ctx, cancel := context.WithCancel(ctx)
 	defer cancel()
 	go ZFSListChan(ctx, listResults,
 		[]string{"name", "guid", "createtxg", "creation", "userrefs"},

zfs/zfs.go

@@ -354,63 +354,6 @@ func (a ZFSSendArgsUnvalidated) buildCommonSendArgs() ([]string, error) {
 	return args, nil
 }
 
-type ReadCloserCopier struct {
-	recorder readErrRecorder
-}
-
-type readErrRecorder struct {
-	io.ReadCloser
-	readErr error
-}
-
-type sendStreamCopierError struct {
-	isReadErr bool // if false, it's a write error
-	err       error
-}
-
-func (e sendStreamCopierError) Error() string {
-	if e.isReadErr {
-		return fmt.Sprintf("stream: read error: %s", e.err)
-	} else {
-		return fmt.Sprintf("stream: writer error: %s", e.err)
-	}
-}
-
-func (e sendStreamCopierError) IsReadError() bool  { return e.isReadErr }
-func (e sendStreamCopierError) IsWriteError() bool { return !e.isReadErr }
-
-func (r *readErrRecorder) Read(p []byte) (n int, err error) {
-	n, err = r.ReadCloser.Read(p)
-	r.readErr = err
-	return n, err
-}
-
-func NewReadCloserCopier(stream io.ReadCloser) *ReadCloserCopier {
-	return &ReadCloserCopier{recorder: readErrRecorder{stream, nil}}
-}
-
-func (c *ReadCloserCopier) WriteStreamTo(w io.Writer) StreamCopierError {
-	debug("sendStreamCopier.WriteStreamTo: begin")
-	_, err := io.Copy(w, &c.recorder)
-	debug("sendStreamCopier.WriteStreamTo: copy done")
-	if err != nil {
-		if c.recorder.readErr != nil {
-			return sendStreamCopierError{isReadErr: true, err: c.recorder.readErr}
-		} else {
-			return sendStreamCopierError{isReadErr: false, err: err}
-		}
-	}
-	return nil
-}
-
-func (c *ReadCloserCopier) Read(p []byte) (n int, err error) {
-	return c.recorder.Read(p)
-}
-
-func (c *ReadCloserCopier) Close() error {
-	return c.recorder.ReadCloser.Close()
-}
-
 func pipeWithCapacityHint(capacity int) (r, w *os.File, err error) {
 	if capacity <= 0 {
 		panic(fmt.Sprintf("capacity must be positive %v", capacity))
@@ -423,7 +366,7 @@ func pipeWithCapacityHint(capacity int) (r, w *os.File, err error) {
 	return stdoutReader, stdoutWriter, nil
 }
 
-type sendStream struct {
+type SendStream struct {
 	cmd  *zfscmd.Cmd
 	kill context.CancelFunc
 
@@ -433,7 +376,7 @@ type sendStream struct {
 	opErr        error
 }
 
-func (s *sendStream) Read(p []byte) (n int, err error) {
+func (s *SendStream) Read(p []byte) (n int, err error) {
 	s.closeMtx.Lock()
 	opErr := s.opErr
 	s.closeMtx.Unlock()
@@ -454,12 +397,12 @@ func (s *sendStream) Read(p []byte) (n int, err error) {
 	return n, err
 }
 
-func (s *sendStream) Close() error {
+func (s *SendStream) Close() error {
 	debug("sendStream: close called")
 	return s.killAndWait(nil)
 }
 
-func (s *sendStream) killAndWait(precedingReadErr error) error {
+func (s *SendStream) killAndWait(precedingReadErr error) error {
 
 	debug("sendStream: killAndWait enter")
 	defer debug("sendStream: killAndWait leave")
@@ -830,7 +773,7 @@ var ErrEncryptedSendNotSupported = fmt.Errorf("raw sends which are required for
 // (if from is "" a full ZFS send is done)
 //
 // Returns ErrEncryptedSendNotSupported if encrypted send is requested but not supported by CLI
-func ZFSSend(ctx context.Context, sendArgs ZFSSendArgsValidated) (*ReadCloserCopier, error) {
+func ZFSSend(ctx context.Context, sendArgs ZFSSendArgsValidated) (*SendStream, error) {
 
 	args := make([]string, 0)
 	args = append(args, "send")
@@ -879,14 +822,14 @@ func ZFSSend(ctx context.Context, sendArgs ZFSSendArgsValidated) (*ReadCloserCop
 	// close our writing-end of the pipe so that we don't wait for ourselves when reading from the reading  end
 	stdoutWriter.Close()
 
-	stream := &sendStream{
+	stream := &SendStream{
 		cmd:          cmd,
 		kill:         cancel,
 		stdoutReader: stdoutReader,
 		stderrBuf:    stderrBuf,
 	}
 
-	return NewReadCloserCopier(stream), nil
+	return stream, nil
 }
 
 type DrySendType string
@@ -1025,24 +968,6 @@ func ZFSSendDry(ctx context.Context, sendArgs ZFSSendArgsValidated) (_ *DrySendI
 	return &si, nil
 }
 
-type StreamCopierError interface {
-	error
-	IsReadError() bool
-	IsWriteError() bool
-}
-
-type StreamCopier interface {
-	// WriteStreamTo writes the stream represented by this StreamCopier
-	// to the given io.Writer.
-	WriteStreamTo(w io.Writer) StreamCopierError
-	// Close must be called as soon as it is clear that no more data will
-	// be read from the StreamCopier.
-	// If StreamCopier gets its data from a connection, it might hold
-	// a lock on the connection until Close is called. Only closing ensures
-	// that the connection can be used afterwards.
-	Close() error
-}
-
 type RecvOptions struct {
 	// Rollback to the oldest snapshot, destroy it, then perform `recv -F`.
 	// Note that this doesn't change property values, i.e. an existing local property value will be kept.
@@ -1067,7 +992,9 @@ func (e *ErrRecvResumeNotSupported) Error() string {
 	return buf.String()
 }
 
-func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, streamCopier StreamCopier, opts RecvOptions) (err error) {
+const RecvStderrBufSiz = 1 << 15
+
+func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, stream io.ReadCloser, opts RecvOptions) (err error) {
 
 	if err := v.ValidateInMemory(fs); err != nil {
 		return errors.Wrap(err, "invalid version")
@@ -1084,7 +1011,7 @@ func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, streamCopier
 	if opts.RollbackAndForceRecv {
 		// destroy all snapshots before `recv -F` because `recv -F`
 		// does not perform a rollback unless `send -R` was used (which we assume hasn't been the case)
-		snaps, err := ZFSListFilesystemVersions(fsdp, ListFilesystemVersionsOptions{
+		snaps, err := ZFSListFilesystemVersions(ctx, fsdp, ListFilesystemVersionsOptions{
 			Types: Snapshots,
 		})
 		if err != nil {
@@ -1134,7 +1061,7 @@ func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, streamCopier
 	//   cannot receive new filesystem stream: invalid backup stream
 	stdout := bytes.NewBuffer(make([]byte, 0, 1024))
 
-	stderr := bytes.NewBuffer(make([]byte, 0, 1024))
+	stderr := bytes.NewBuffer(make([]byte, 0, RecvStderrBufSiz))
 
 	stdin, stdinWriter, err := pipeWithCapacityHint(ZFSRecvPipeCapacityHint)
 	if err != nil {
@@ -1162,9 +1089,10 @@ func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, streamCopier
 
 	debug("started")
 
-	copierErrChan := make(chan StreamCopierError)
+	copierErrChan := make(chan error)
 	go func() {
-		copierErrChan <- streamCopier.WriteStreamTo(stdinWriter)
+		_, err := io.Copy(stdinWriter, stream)
+		copierErrChan <- err
 		stdinWriter.Close()
 	}()
 	waitErrChan := make(chan error)
@@ -1173,6 +1101,10 @@ func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, streamCopier
 		if err = cmd.Wait(); err != nil {
 			if rtErr := tryRecvErrorWithResumeToken(ctx, stderr.String()); rtErr != nil {
 				waitErrChan <- rtErr
+			} else if owErr := tryRecvDestroyOrOverwriteEncryptedErr(stderr.Bytes()); owErr != nil {
+				waitErrChan <- owErr
+			} else if readErr := tryRecvCannotReadFromStreamErr(stderr.Bytes()); readErr != nil {
+				waitErrChan <- readErr
 			} else {
 				waitErrChan <- &ZFSError{
 					Stderr:  stderr.Bytes(),
@@ -1183,22 +1115,23 @@ func ZFSRecv(ctx context.Context, fs string, v *ZFSSendArgVersion, streamCopier
 		}
 	}()
 
-	// streamCopier always fails before or simultaneously with Wait
-	// thus receive from it first
 	copierErr := <-copierErrChan
 	debug("copierErr: %T %s", copierErr, copierErr)
 	if copierErr != nil {
+		debug("killing zfs recv command after copierErr")
 		cancelCmd()
 	}
 
 	waitErr := <-waitErrChan
 	debug("waitErr: %T %s", waitErr, waitErr)
+
 	if copierErr == nil && waitErr == nil {
 		return nil
-	} else if waitErr != nil && (copierErr == nil || copierErr.IsWriteError()) {
-		return waitErr // has more interesting info in that case
+	} else if _, isReadErr := waitErr.(*RecvCannotReadFromStreamErr); isReadErr {
+		return copierErr // likely network error reading from stream
+	} else {
+		return waitErr // almost always more interesting info. NOTE: do not wrap!
 	}
-	return copierErr // if it's not a write error, the copier error is more interesting
 }
 
 type RecvFailedWithResumeTokenErr struct {
@@ -1228,6 +1161,43 @@ func (e *RecvFailedWithResumeTokenErr) Error() string {
 	return fmt.Sprintf("receive failed, resume token available: %s\n%#v", e.ResumeTokenRaw, e.ResumeTokenParsed)
 }
 
+type RecvDestroyOrOverwriteEncryptedErr struct {
+	Msg string
+}
+
+func (e *RecvDestroyOrOverwriteEncryptedErr) Error() string {
+	return e.Msg
+}
+
+var recvDestroyOrOverwriteEncryptedErrRe = regexp.MustCompile(`^(cannot receive new filesystem stream: zfs receive -F cannot be used to destroy an encrypted filesystem or overwrite an unencrypted one with an encrypted one)`)
+
+func tryRecvDestroyOrOverwriteEncryptedErr(stderr []byte) *RecvDestroyOrOverwriteEncryptedErr {
+	debug("tryRecvDestroyOrOverwriteEncryptedErr: %v", stderr)
+	m := recvDestroyOrOverwriteEncryptedErrRe.FindSubmatch(stderr)
+	if m == nil {
+		return nil
+	}
+	return &RecvDestroyOrOverwriteEncryptedErr{Msg: string(m[1])}
+}
+
+type RecvCannotReadFromStreamErr struct {
+	Msg string
+}
+
+func (e *RecvCannotReadFromStreamErr) Error() string {
+	return e.Msg
+}
+
+var reRecvCannotReadFromStreamErr = regexp.MustCompile(`^(cannot receive: failed to read from stream)$`)
+
+func tryRecvCannotReadFromStreamErr(stderr []byte) *RecvCannotReadFromStreamErr {
+	m := reRecvCannotReadFromStreamErr.FindSubmatch(stderr)
+	if m == nil {
+		return nil
+	}
+	return &RecvCannotReadFromStreamErr{Msg: string(m[1])}
+}
+
 type ClearResumeTokenError struct {
 	ZFSOutput []byte
 	CmdError  error

zfs/zfs_test.go

@@ -2,9 +2,11 @@ package zfs
 
 import (
 	"context"
+	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestZFSListHandlesProducesZFSErrorOnNonZeroExit(t *testing.T) {
@@ -259,3 +261,12 @@ size	10518512
 		})
 	}
 }
+
+func TestTryRecvDestroyOrOverwriteEncryptedErr(t *testing.T) {
+	msg := "cannot receive new filesystem stream: zfs receive -F cannot be used to destroy an encrypted filesystem or overwrite an unencrypted one with an encrypted one\n"
+	assert.GreaterOrEqual(t, RecvStderrBufSiz, len(msg))
+
+	err := tryRecvDestroyOrOverwriteEncryptedErr([]byte(msg))
+	require.NotNil(t, err)
+	assert.EqualError(t, err, strings.TrimSpace(msg))
+}

zfs/zfscmd/zfscmd.go

@@ -18,10 +18,10 @@ import (
 )
 
 type Cmd struct {
-	cmd                       *exec.Cmd
-	ctx                       context.Context
-	mtx                       sync.RWMutex
-	startedAt, waitReturnedAt time.Time
+	cmd                                      *exec.Cmd
+	ctx                                      context.Context
+	mtx                                      sync.RWMutex
+	startedAt, waitStartedAt, waitReturnedAt time.Time
 }
 
 func CommandContext(ctx context.Context, name string, arg ...string) *Cmd {
@@ -119,19 +119,65 @@ func (c *Cmd) startPost(err error) {
 }
 
 func (c *Cmd) waitPre() {
-	waitPreLogging(c, time.Now())
+	now := time.Now()
+
+	// ignore duplicate waits
+	c.mtx.Lock()
+	// ignore duplicate waits
+	if !c.waitStartedAt.IsZero() {
+		c.mtx.Unlock()
+		return
+	}
+	c.waitStartedAt = now
+	c.mtx.Unlock()
+
+	waitPreLogging(c, now)
+}
+
+type usage struct {
+	total_secs, system_secs, user_secs float64
 }
 
 func (c *Cmd) waitPost(err error) {
 	now := time.Now()
 
 	c.mtx.Lock()
+	// ignore duplicate waits
+	if !c.waitReturnedAt.IsZero() {
+		c.mtx.Unlock()
+		return
+	}
 	c.waitReturnedAt = now
 	c.mtx.Unlock()
 
-	waitPostReport(c, now)
-	waitPostLogging(c, err, now)
-	waitPostPrometheus(c, err, now)
+	// build usage
+	var u usage
+	{
+		var s *os.ProcessState
+		if err == nil {
+			s = c.cmd.ProcessState
+		} else if ee, ok := err.(*exec.ExitError); ok {
+			s = ee.ProcessState
+		}
+
+		if s == nil {
+			u = usage{
+				total_secs:  c.Runtime().Seconds(),
+				system_secs: -1,
+				user_secs:   -1,
+			}
+		} else {
+			u = usage{
+				total_secs:  c.Runtime().Seconds(),
+				system_secs: s.SystemTime().Seconds(),
+				user_secs:   s.UserTime().Seconds(),
+			}
+		}
+	}
+
+	waitPostReport(c, u, now)
+	waitPostLogging(c, u, err, now)
+	waitPostPrometheus(c, u, err, now)
 }
 
 // returns 0 if the command did not yet finish

zfs/zfscmd/zfscmd_logging.go

@@ -1,7 +1,6 @@
 package zfscmd
 
 import (
-	"os/exec"
 	"time"
 )
 
@@ -28,23 +27,12 @@ func waitPreLogging(c *Cmd, now time.Time) {
 	c.log().Debug("start waiting")
 }
 
-func waitPostLogging(c *Cmd, err error, now time.Time) {
-
-	var total, system, user float64
-
-	total = c.Runtime().Seconds()
-	if ee, ok := err.(*exec.ExitError); ok {
-		system = ee.ProcessState.SystemTime().Seconds()
-		user = ee.ProcessState.UserTime().Seconds()
-	} else {
-		system = -1
-		user = -1
-	}
+func waitPostLogging(c *Cmd, u usage, err error, now time.Time) {
 
 	log := c.log().
-		WithField("total_time_s", total).
-		WithField("systemtime_s", system).
-		WithField("usertime_s", user)
+		WithField("total_time_s", u.total_secs).
+		WithField("systemtime_s", u.system_secs).
+		WithField("usertime_s", u.user_secs)
 
 	if err == nil {
 		log.Info("command exited without error")

zfs/zfscmd/zfscmd_platform_test.go

@@ -5,11 +5,14 @@ import (
 	"bytes"
 	"context"
 	"io"
+	"os"
 	"os/exec"
+	"strings"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"github.com/zrepl/zrepl/util/circlog"
 )
 
 const testBin = "./zfscmd_platform_test.bash"
@@ -85,5 +88,37 @@ func TestCmdProcessState(t *testing.T) {
 	require.True(t, ok)
 	require.NotNil(t, ee.ProcessState)
 	require.Contains(t, ee.Error(), "killed")
+}
+
+func TestSigpipe(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	cmd := CommandContext(ctx, "bash", "-c", "sleep 5; echo invalid input; exit 23")
+	r, w, err := os.Pipe()
+	require.NoError(t, err)
+	output := circlog.MustNewCircularLog(1 << 20)
+	cmd.SetStdio(Stdio{
+		Stdin:  r,
+		Stdout: output,
+		Stderr: output,
+	})
+	err = cmd.Start()
+	require.NoError(t, err)
+	err = r.Close()
+	require.NoError(t, err)
+
+	// the script doesn't read stdin, but this input is almost certainly smaller than the pipe buffer
+	const LargerThanPipeBuffer = 1 << 21
+	_, err = io.Copy(w, bytes.NewBuffer(bytes.Repeat([]byte("i"), LargerThanPipeBuffer)))
+	// => io.Copy is going to block because the pipe buffer is full and the
+	//    script is not reading from it
+	// => the script is going to exit after 5s
+	// => we should expect a broken pipe error from the copier's perspective
+	t.Logf("copy err = %T: %s", err, err)
+	require.NotNil(t, err)
+	require.True(t, strings.Contains(err.Error(), "broken pipe"))
+
+	err = cmd.Wait()
+	require.EqualError(t, err, "exit status 23")
 
 }

zfs/zfscmd/zfscmd_prometheus.go

@@ -46,7 +46,7 @@ func RegisterMetrics(r prometheus.Registerer) {
 	r.MustRegister(metrics.usertime)
 }
 
-func waitPostPrometheus(c *Cmd, err error, now time.Time) {
+func waitPostPrometheus(c *Cmd, u usage, err error, now time.Time) {
 
 	if len(c.cmd.Args) < 2 {
 		getLogger(c.ctx).WithField("args", c.cmd.Args).
@@ -64,10 +64,10 @@ func waitPostPrometheus(c *Cmd, err error, now time.Time) {
 
 	metrics.totaltime.
 		WithLabelValues(labelValues...).
-		Observe(c.Runtime().Seconds())
+		Observe(u.total_secs)
 	metrics.systemtime.WithLabelValues(labelValues...).
-		Observe(c.cmd.ProcessState.SystemTime().Seconds())
+		Observe(u.system_secs)
 	metrics.usertime.WithLabelValues(labelValues...).
-		Observe(c.cmd.ProcessState.UserTime().Seconds())
+		Observe(u.user_secs)
 
 }

zfs/zfscmd/zfscmd_report.go

@@ -57,7 +57,7 @@ func startPostReport(c *Cmd, err error, now time.Time) {
 	active.mtx.Unlock()
 }
 
-func waitPostReport(c *Cmd, now time.Time) {
+func waitPostReport(c *Cmd, _ usage, now time.Time) {
 	active.mtx.Lock()
 	defer active.mtx.Unlock()
 	prev := active.cmds[c]