Merge pull request #3098 from creack/bump_v0.7.1

Bump v0.7.1

.gitignore

@@ -1,3 +1,6 @@
+# Docker project generated files to ignore
+#  if you want to ignore files created by your editor/tools,
+#  please consider a global .gitignore https://help.github.com/articles/ignoring-files
 .vagrant*
 bin
 docker/docker

CHANGELOG.md

@@ -1,5 +1,73 @@
 # Changelog
 
+## 0.7.1 (2013-12-05)
+
+#### Documentation
+
++ Add @SvenDowideit as documentation maintainer
++ Add links example
++ Add documentation regarding ambassador pattern
++ Add Google Cloud Platform docs
++ Add dockerfile best practices
+* Update doc for RHEL
+* Update doc for registry
+* Update Postgres examples
+* Update doc for Ubuntu install
+* Improve remote api doc
+
+#### Runtime
+
++ Add hostconfig to docker inspect
++ Implement `docker log -f` to stream logs
++ Add env variable to disable kernel version warning
++ Add -format to `docker inspect`
++ Support bind-mount for files
+- Fix bridge creation on RHEL
+- Fix image size calculation
+- Make sure iptables are called even if the bridge already exists
+- Fix issue with stderr only attach
+- Remove init layer when destroying a container
+- Fix same port binding on different interfaces
+- `docker build` now returns the correct exit code
+- Fix `docker port` to display correct port
+- `docker build` now check that the dockerfile exists client side
+- `docker attach` now returns the correct exit code
+- Remove the name entry when the container does not exist
+
+#### Registry
+
+* Improve progress bars, add ETA for downloads
+* Simultaneous pulls now waits for the first to finish instead of failing
+- Tag only the top-layer image when pushing to registry
+- Fix issue with offline image transfer
+- Fix issue preventing using ':' in password for registry
+
+#### Other
+
++ Add pprof handler for debug
++ Create a Makefile
+* Use stdlib tar that now includes fix
+* Improve make.sh test script
+* Handle SIGQUIT on the daemon
+* Disable verbose during tests
+* Upgrade to go1.2 for official build
+* Improve unit tests
+* The test suite now runs all tests even if one fails
+* Refactor C in Go (Devmapper)
+- Fix OSX compilation
+
+## 0.7.0 (2013-11-25)
+
+#### Notable features since 0.6.0
+
+* Storage drivers: choose from aufs, device-mapper, or vfs.
+* Standard Linux support: docker now runs on unmodified Linux kernels and all major distributions.
+* Links: compose complex software stacks by connecting containers to each other.
+* Container naming: organize your containers by giving them memorable names.
+* Advanced port redirects: specify port redirects per interface, or keep sensitive ports private.
+* Offline transfer: push and pull images to the filesystem without losing information.
+* Quality: numerous bugfixes and small usability improvements. Significant increase in test coverage.
+
 ## 0.6.7 (2013-11-21)
 
 #### Runtime

CONTRIBUTING.md

@@ -4,6 +4,13 @@ Want to hack on Docker? Awesome! Here are instructions to get you
 started. They are probably not perfect, please let us know if anything
 feels wrong or incomplete.
 
+## Reporting Issues
+
+When reporting [issues](https://github.com/dotcloud/docker/issues) 
+on Github please include your host OS ( Ubuntu 12.04, Fedora 19, etc... )
+and the output of `docker version` along with the output of `docker info` if possible.  
+This information will help us review and fix your issue faster.
+
 ## Build Environment
 
 For instructions on setting up your development environment, please
@@ -64,7 +71,7 @@ your branch before submitting a pull request.
 
 Update the documentation when creating or modifying features. Test
 your documentation changes for clarity, concision, and correctness, as
-well as a clean docmuent build. See ``docs/README.md`` for more
+well as a clean documentation build. See ``docs/README.md`` for more
 information on building the docs and how docs get released.
 
 Write clean code. Universally formatted code promotes ease of writing, reading,
@@ -115,6 +122,7 @@ For more details see [MAINTAINERS.md](hack/MAINTAINERS.md)
 * Step 1: learn the component inside out
 * Step 2: make yourself useful by contributing code, bugfixes, support etc.
 * Step 3: volunteer on the irc channel (#docker@freenode)
+* Step 4: propose yourself at a scheduled #docker-meeting
 
 Don't forget: being a maintainer is a time investment. Make sure you will have time to make yourself available.
 You don't have to be a maintainer to make a difference on the project!

Dockerfile

@@ -23,44 +23,53 @@
 # the case. Therefore, you don't have to disable it anymore.
 #
 
-docker-version 0.6.1
-from	ubuntu:12.04
-maintainer	Solomon Hykes <solomon@dotcloud.com>
+docker-version	0.6.1
+FROM	ubuntu:12.04
+MAINTAINER	Solomon Hykes <solomon@dotcloud.com>
 
 # Build dependencies
-run	echo 'deb http://archive.ubuntu.com/ubuntu precise main universe' > /etc/apt/sources.list
-run	apt-get update
-run	apt-get install -y -q curl
-run	apt-get install -y -q git
-run	apt-get install -y -q mercurial
-run apt-get install -y -q build-essential libsqlite3-dev
+RUN	echo 'deb http://archive.ubuntu.com/ubuntu precise main universe' > /etc/apt/sources.list
+RUN	apt-get update
+RUN	apt-get install -y -q curl
+RUN	apt-get install -y -q git
+RUN	apt-get install -y -q mercurial
+RUN	apt-get install -y -q build-essential libsqlite3-dev
 
 # Install Go
-run	curl -s https://go.googlecode.com/files/go1.2rc5.src.tar.gz | tar -v -C /usr/local -xz
-env	PATH	/usr/local/go/bin:/usr/local/bin:/usr/local/sbin:/usr/bin:/usr/sbin:/bin:/sbin
-env	GOPATH	/go:/go/src/github.com/dotcloud/docker/vendor
-run cd /usr/local/go/src && ./make.bash && go install -ldflags '-w -linkmode external -extldflags "-static -Wl,--unresolved-symbols=ignore-in-shared-libs"' -tags netgo -a std
+RUN	curl -s https://go.googlecode.com/files/go1.2.src.tar.gz | tar -v -C /usr/local -xz
+ENV	PATH	/usr/local/go/bin:/usr/local/bin:/usr/local/sbin:/usr/bin:/usr/sbin:/bin:/sbin
+ENV	GOPATH	/go:/go/src/github.com/dotcloud/docker/vendor
+RUN	cd /usr/local/go/src && ./make.bash && go install -ldflags '-w -linkmode external -extldflags "-static -Wl,--unresolved-symbols=ignore-in-shared-libs"' -tags netgo -a std
 
 # Ubuntu stuff
-run	apt-get install -y -q ruby1.9.3 rubygems libffi-dev
-run	gem install --no-rdoc --no-ri fpm
-run	apt-get install -y -q reprepro dpkg-sig
+RUN	apt-get install -y -q ruby1.9.3 rubygems libffi-dev
+RUN	gem install --no-rdoc --no-ri fpm
+RUN	apt-get install -y -q reprepro dpkg-sig
 
-run	apt-get install -y -q python-pip
-run	pip install s3cmd==1.1.0-beta3
-run	pip install python-magic==0.4.6
-run	/bin/echo -e '[default]\naccess_key=$AWS_ACCESS_KEY\nsecret_key=$AWS_SECRET_KEY\n' > /.s3cfg
+RUN	apt-get install -y -q python-pip
+RUN	pip install s3cmd==1.1.0-beta3
+RUN	pip install python-magic==0.4.6
+RUN	/bin/echo -e '[default]\naccess_key=$AWS_ACCESS_KEY\nsecret_key=$AWS_SECRET_KEY\n' > /.s3cfg
 
 # Runtime dependencies
-run	apt-get install -y -q iptables
-run	apt-get install -y -q lxc
-run	apt-get install -y -q aufs-tools
+RUN	apt-get install -y -q iptables
+RUN	apt-get install -y -q lxc
+RUN	apt-get install -y -q aufs-tools
 
-volume	/var/lib/docker
-workdir	/go/src/github.com/dotcloud/docker
+# Get lvm2 source for compiling statically
+RUN	git clone https://git.fedorahosted.org/git/lvm2.git /usr/local/lvm2 && cd /usr/local/lvm2 && git checkout v2_02_103
+# see https://git.fedorahosted.org/cgit/lvm2.git/refs/tags for release tags
+# note: we can't use "git clone -b" above because it requires at least git 1.7.10 to be able to use that on a tag instead of a branch and we only have 1.7.9.5
+
+# Compile and install lvm2
+RUN	cd /usr/local/lvm2 && ./configure --enable-static_link && make device-mapper && make install_device-mapper
+# see https://git.fedorahosted.org/cgit/lvm2.git/tree/INSTALL
+
+VOLUME	/var/lib/docker
+WORKDIR	/go/src/github.com/dotcloud/docker
 
 # Wrap all commands in the "docker-in-docker" script to allow nested containers
-entrypoint ["hack/dind"]
+ENTRYPOINT	["hack/dind"]
 
 # Upload docker source
-add	.       /go/src/github.com/dotcloud/docker
+ADD	.	/go/src/github.com/dotcloud/docker

Makefile

@@ -0,0 +1,26 @@
+.PHONY: all binary build default doc shell test
+
+DOCKER_RUN_DOCKER := docker run -rm -i -t -privileged -e TESTFLAGS -v $(CURDIR)/bundles:/go/src/github.com/dotcloud/docker/bundles docker
+
+default: binary
+
+all: build
+	$(DOCKER_RUN_DOCKER) hack/make.sh
+
+binary: build
+	$(DOCKER_RUN_DOCKER) hack/make.sh binary
+
+doc:
+	docker build -t docker-docs docs && docker run -p 8000:8000 docker-docs
+
+test: build
+	$(DOCKER_RUN_DOCKER) hack/make.sh test
+
+shell: build
+	$(DOCKER_RUN_DOCKER) bash
+
+build: bundles
+	docker build -t docker .
+
+bundles:
+	mkdir bundles

VERSION

@@ -1 +1 @@
-0.6.7
+0.7.1

api.go

@@ -1,12 +1,16 @@
 package docker
 
 import (
+	"bufio"
+	"bytes"
 	"code.google.com/p/go.net/websocket"
 	"encoding/base64"
 	"encoding/json"
+	"expvar"
 	"fmt"
 	"github.com/dotcloud/docker/archive"
 	"github.com/dotcloud/docker/auth"
+	"github.com/dotcloud/docker/systemd"
 	"github.com/dotcloud/docker/utils"
 	"github.com/gorilla/mux"
 	"io"
@@ -15,6 +19,7 @@ import (
 	"mime"
 	"net"
 	"net/http"
+	"net/http/pprof"
 	"os"
 	"os/exec"
 	"regexp"
@@ -23,7 +28,7 @@ import (
 )
 
 const (
-	APIVERSION        = 1.7
+	APIVERSION        = 1.8
 	DEFAULTHTTPHOST   = "127.0.0.1"
 	DEFAULTHTTPPORT   = 4243
 	DEFAULTUNIXSOCKET = "/var/run/docker.sock"
@@ -254,7 +259,7 @@ func getEvents(srv *Server, version float64, w http.ResponseWriter, r *http.Requ
 	wf.Flush()
 	if since != 0 {
 		// If since, send previous events that happened after the timestamp
-		for _, event := range srv.events {
+		for _, event := range srv.GetEvents() {
 			if event.Time >= since {
 				err := sendEvent(wf, &event)
 				if err != nil && err.Error() == "JSON error" {
@@ -564,12 +569,18 @@ func postContainersCreate(srv *Server, version float64, w http.ResponseWriter, r
 		job.SetenvList("Dns", defaultDns)
 	}
 	// Read container ID from the first line of stdout
-	job.StdoutParseString(&out.ID)
+	job.Stdout.AddString(&out.ID)
 	// Read warnings from stderr
-	job.StderrParseLines(&out.Warnings, 0)
+	warnings := &bytes.Buffer{}
+	job.Stderr.Add(warnings)
 	if err := job.Run(); err != nil {
 		return err
 	}
+	// Parse warnings from stderr
+	scanner := bufio.NewScanner(warnings)
+	for scanner.Scan() {
+		out.Warnings = append(out.Warnings, scanner.Text())
+	}
 	if job.GetenvInt("Memory") > 0 && !srv.runtime.capabilities.MemoryLimit {
 		log.Println("WARNING: Your kernel does not support memory limit capabilities. Limitation discarded.")
 		out.Warnings = append(out.Warnings, "Your kernel does not support memory limit capabilities. Limitation discarded.")
@@ -865,7 +876,10 @@ func getContainersByName(srv *Server, version float64, w http.ResponseWriter, r
 		return fmt.Errorf("Conflict between containers and images")
 	}
 
-	return writeJSON(w, http.StatusOK, container)
+	container.readHostConfig()
+	c := APIContainer{container, container.hostConfig}
+
+	return writeJSON(w, http.StatusOK, c)
 }
 
 func getImagesByName(srv *Server, version float64, w http.ResponseWriter, r *http.Request, vars map[string]string) error {
@@ -922,7 +936,7 @@ func postBuild(srv *Server, version float64, w http.ResponseWriter, r *http.Requ
 		}
 		context = c
 	} else if utils.IsURL(remoteURL) {
-		f, err := utils.Download(remoteURL, ioutil.Discard)
+		f, err := utils.Download(remoteURL)
 		if err != nil {
 			return err
 		}
@@ -951,9 +965,26 @@ func postBuild(srv *Server, version float64, w http.ResponseWriter, r *http.Requ
 		return err
 	}
 
-	b := NewBuildFile(srv, utils.NewWriteFlusher(w), !suppressOutput, !noCache, rm)
+	if version >= 1.8 {
+		w.Header().Set("Content-Type", "application/json")
+	}
+	sf := utils.NewStreamFormatter(version >= 1.8)
+	b := NewBuildFile(srv,
+		&StdoutFormater{
+			Writer:          utils.NewWriteFlusher(w),
+			StreamFormatter: sf,
+		},
+		&StderrFormater{
+			Writer:          utils.NewWriteFlusher(w),
+			StreamFormatter: sf,
+		},
+		!suppressOutput, !noCache, rm, utils.NewWriteFlusher(w), sf)
 	id, err := b.Build(context)
 	if err != nil {
+		if sf.Used() {
+			w.Write(sf.FormatError(err))
+			return nil
+		}
 		return fmt.Errorf("Error build: %s", err)
 	}
 	if repoName != "" {
@@ -1037,9 +1068,37 @@ func makeHttpHandler(srv *Server, logging bool, localMethod string, localRoute s
 	}
 }
 
+// Replicated from expvar.go as not public.
+func expvarHandler(w http.ResponseWriter, r *http.Request) {
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
+	fmt.Fprintf(w, "{\n")
+	first := true
+	expvar.Do(func(kv expvar.KeyValue) {
+		if !first {
+			fmt.Fprintf(w, ",\n")
+		}
+		first = false
+		fmt.Fprintf(w, "%q: %s", kv.Key, kv.Value)
+	})
+	fmt.Fprintf(w, "\n}\n")
+}
+
+func AttachProfiler(router *mux.Router) {
+	router.HandleFunc("/debug/vars", expvarHandler)
+	router.HandleFunc("/debug/pprof/", pprof.Index)
+	router.HandleFunc("/debug/pprof/cmdline", pprof.Cmdline)
+	router.HandleFunc("/debug/pprof/profile", pprof.Profile)
+	router.HandleFunc("/debug/pprof/symbol", pprof.Symbol)
+	router.HandleFunc("/debug/pprof/heap", pprof.Handler("heap").ServeHTTP)
+	router.HandleFunc("/debug/pprof/goroutine", pprof.Handler("goroutine").ServeHTTP)
+	router.HandleFunc("/debug/pprof/threadcreate", pprof.Handler("threadcreate").ServeHTTP)
+}
+
 func createRouter(srv *Server, logging bool) (*mux.Router, error) {
 	r := mux.NewRouter()
-
+	if os.Getenv("DEBUG") != "" {
+		AttachProfiler(r)
+	}
 	m := map[string]map[string]HttpApiFunc{
 		"GET": {
 			"/events":                         getEvents,
@@ -1126,8 +1185,6 @@ func ServeRequest(srv *Server, apiversion float64, w http.ResponseWriter, req *h
 }
 
 func ListenAndServe(proto, addr string, srv *Server, logging bool) error {
-	log.Printf("Listening for HTTP on %s (%s)\n", addr, proto)
-
 	r, err := createRouter(srv, logging)
 	if err != nil {
 		return err
@@ -1158,5 +1215,9 @@ func ListenAndServe(proto, addr string, srv *Server, logging bool) error {
 		}
 	}
 	httpSrv := http.Server{Addr: addr, Handler: r}
+
+	log.Printf("Listening for HTTP on %s (%s)\n", addr, proto)
+	// Tell the init daemon we are accepting requests
+	go systemd.SdNotify("READY=1")
 	return httpSrv.Serve(l)
 }

api_params.go

@@ -33,15 +33,17 @@ type (
 		Debug              bool
 		Containers         int
 		Images             int
-		NFd                int    `json:",omitempty"`
-		NGoroutines        int    `json:",omitempty"`
-		MemoryLimit        bool   `json:",omitempty"`
-		SwapLimit          bool   `json:",omitempty"`
-		IPv4Forwarding     bool   `json:",omitempty"`
-		LXCVersion         string `json:",omitempty"`
-		NEventsListener    int    `json:",omitempty"`
-		KernelVersion      string `json:",omitempty"`
-		IndexServerAddress string `json:",omitempty"`
+		Driver             string      `json:",omitempty"`
+		DriverStatus       [][2]string `json:",omitempty"`
+		NFd                int         `json:",omitempty"`
+		NGoroutines        int         `json:",omitempty"`
+		MemoryLimit        bool        `json:",omitempty"`
+		SwapLimit          bool        `json:",omitempty"`
+		IPv4Forwarding     bool        `json:",omitempty"`
+		LXCVersion         string      `json:",omitempty"`
+		NEventsListener    int         `json:",omitempty"`
+		KernelVersion      string      `json:",omitempty"`
+		IndexServerAddress string      `json:",omitempty"`
 	}
 
 	APITop struct {
@@ -116,6 +118,10 @@ type (
 		Resource string
 		HostPath string
 	}
+	APIContainer struct {
+		*Container
+		HostConfig *HostConfig
+	}
 )
 
 func (api APIImages) ToLegacy() []APIImagesOld {

archive/archive.go

@@ -15,7 +15,15 @@ import (
 
 type Archive io.Reader
 
-type Compression uint32
+type Compression int
+
+type TarOptions struct {
+	Includes    []string
+	Excludes    []string
+	Recursive   bool
+	Compression Compression
+	CreateFiles []string
+}
 
 const (
 	Uncompressed Compression = iota
@@ -80,20 +88,78 @@ func (compression *Compression) Extension() string {
 // Tar creates an archive from the directory at `path`, and returns it as a
 // stream of bytes.
 func Tar(path string, compression Compression) (io.Reader, error) {
-	return TarFilter(path, compression, nil)
+	return TarFilter(path, &TarOptions{Recursive: true, Compression: compression})
+}
+
+func escapeName(name string) string {
+	escaped := make([]byte, 0)
+	for i, c := range []byte(name) {
+		if i == 0 && c == '/' {
+			continue
+		}
+		// all printable chars except "-" which is 0x2d
+		if (0x20 <= c && c <= 0x7E) && c != 0x2d {
+			escaped = append(escaped, c)
+		} else {
+			escaped = append(escaped, fmt.Sprintf("\\%03o", c)...)
+		}
+	}
+	return string(escaped)
 }
 
 // Tar creates an archive from the directory at `path`, only including files whose relative
 // paths are included in `filter`. If `filter` is nil, then all files are included.
-func TarFilter(path string, compression Compression, filter []string) (io.Reader, error) {
-	args := []string{"tar", "--numeric-owner", "-f", "-", "-C", path}
-	if filter == nil {
-		filter = []string{"."}
+func TarFilter(path string, options *TarOptions) (io.Reader, error) {
+	args := []string{"tar", "--numeric-owner", "-f", "-", "-C", path, "-T", "-"}
+	if options.Includes == nil {
+		options.Includes = []string{"."}
 	}
-	for _, f := range filter {
-		args = append(args, "-c"+compression.Flag(), f)
+	args = append(args, "-c"+options.Compression.Flag())
+
+	for _, exclude := range options.Excludes {
+		args = append(args, fmt.Sprintf("--exclude=%s", exclude))
 	}
-	return CmdStream(exec.Command(args[0], args[1:]...))
+
+	if !options.Recursive {
+		args = append(args, "--no-recursion")
+	}
+
+	files := ""
+	for _, f := range options.Includes {
+		files = files + escapeName(f) + "\n"
+	}
+
+	tmpDir := ""
+
+	if options.CreateFiles != nil {
+		var err error // Can't use := here or we override the outer tmpDir
+		tmpDir, err = ioutil.TempDir("", "docker-tar")
+		if err != nil {
+			return nil, err
+		}
+
+		files = files + "-C" + tmpDir + "\n"
+		for _, f := range options.CreateFiles {
+			path := filepath.Join(tmpDir, f)
+			err := os.MkdirAll(filepath.Dir(path), 0600)
+			if err != nil {
+				return nil, err
+			}
+
+			if file, err := os.OpenFile(path, os.O_CREATE, 0600); err != nil {
+				return nil, err
+			} else {
+				file.Close()
+			}
+			files = files + escapeName(f) + "\n"
+		}
+	}
+
+	return CmdStream(exec.Command(args[0], args[1:]...), &files, func() {
+		if tmpDir != "" {
+			_ = os.RemoveAll(tmpDir)
+		}
+	})
 }
 
 // Untar reads a stream of bytes from `archive`, parses it as a tar archive,
@@ -101,7 +167,7 @@ func TarFilter(path string, compression Compression, filter []string) (io.Reader
 // The archive may be compressed with one of the following algorithms:
 //  identity (uncompressed), gzip, bzip2, xz.
 // FIXME: specify behavior when target path exists vs. doesn't exist.
-func Untar(archive io.Reader, path string) error {
+func Untar(archive io.Reader, path string, options *TarOptions) error {
 	if archive == nil {
 		return fmt.Errorf("Empty archive")
 	}
@@ -123,8 +189,15 @@ func Untar(archive io.Reader, path string) error {
 	compression := DetectCompression(buf)
 
 	utils.Debugf("Archive compression detected: %s", compression.Extension())
+	args := []string{"--numeric-owner", "-f", "-", "-C", path, "-x" + compression.Flag()}
 
-	cmd := exec.Command("tar", "--numeric-owner", "-f", "-", "-C", path, "-x"+compression.Flag())
+	if options != nil {
+		for _, exclude := range options.Excludes {
+			args = append(args, fmt.Sprintf("--exclude=%s", exclude))
+		}
+	}
+
+	cmd := exec.Command("tar", args...)
 	cmd.Stdin = io.MultiReader(bytes.NewReader(buf), archive)
 	// Hardcode locale environment for predictable outcome regardless of host configuration.
 	//   (see https://github.com/dotcloud/docker/issues/355)
@@ -141,11 +214,11 @@ func Untar(archive io.Reader, path string) error {
 // TarUntar aborts and returns the error.
 func TarUntar(src string, filter []string, dst string) error {
 	utils.Debugf("TarUntar(%s %s %s)", src, filter, dst)
-	archive, err := TarFilter(src, Uncompressed, filter)
+	archive, err := TarFilter(src, &TarOptions{Compression: Uncompressed, Includes: filter, Recursive: true})
 	if err != nil {
 		return err
 	}
-	return Untar(archive, dst)
+	return Untar(archive, dst, nil)
 }
 
 // UntarPath is a convenience function which looks for an archive
@@ -153,7 +226,7 @@ func TarUntar(src string, filter []string, dst string) error {
 func UntarPath(src, dst string) error {
 	if archive, err := os.Open(src); err != nil {
 		return err
-	} else if err := Untar(archive, dst); err != nil {
+	} else if err := Untar(archive, dst, nil); err != nil {
 		return err
 	}
 	return nil
@@ -222,19 +295,39 @@ func CopyFileWithTar(src, dst string) error {
 		return err
 	}
 	tw.Close()
-	return Untar(buf, filepath.Dir(dst))
+	return Untar(buf, filepath.Dir(dst), nil)
 }
 
 // CmdStream executes a command, and returns its stdout as a stream.
 // If the command fails to run or doesn't complete successfully, an error
 // will be returned, including anything written on stderr.
-func CmdStream(cmd *exec.Cmd) (io.Reader, error) {
+func CmdStream(cmd *exec.Cmd, input *string, atEnd func()) (io.Reader, error) {
+	if input != nil {
+		stdin, err := cmd.StdinPipe()
+		if err != nil {
+			if atEnd != nil {
+				atEnd()
+			}
+			return nil, err
+		}
+		// Write stdin if any
+		go func() {
+			_, _ = stdin.Write([]byte(*input))
+			stdin.Close()
+		}()
+	}
 	stdout, err := cmd.StdoutPipe()
 	if err != nil {
+		if atEnd != nil {
+			atEnd()
+		}
 		return nil, err
 	}
 	stderr, err := cmd.StderrPipe()
 	if err != nil {
+		if atEnd != nil {
+			atEnd()
+		}
 		return nil, err
 	}
 	pipeR, pipeW := io.Pipe()
@@ -259,6 +352,9 @@ func CmdStream(cmd *exec.Cmd) (io.Reader, error) {
 		} else {
 			pipeW.Close()
 		}
+		if atEnd != nil {
+			atEnd()
+		}
 	}()
 	// Run the command and return the pipe
 	if err := cmd.Start(); err != nil {

archive/archive_test.go

@@ -14,7 +14,7 @@ import (
 
 func TestCmdStreamLargeStderr(t *testing.T) {
 	cmd := exec.Command("/bin/sh", "-c", "dd if=/dev/zero bs=1k count=1000 of=/dev/stderr; echo hello")
-	out, err := CmdStream(cmd)
+	out, err := CmdStream(cmd, nil, nil)
 	if err != nil {
 		t.Fatalf("Failed to start command: %s", err)
 	}
@@ -35,7 +35,7 @@ func TestCmdStreamLargeStderr(t *testing.T) {
 
 func TestCmdStreamBad(t *testing.T) {
 	badCmd := exec.Command("/bin/sh", "-c", "echo hello; echo >&2 error couldn\\'t reverse the phase pulser; exit 1")
-	out, err := CmdStream(badCmd)
+	out, err := CmdStream(badCmd, nil, nil)
 	if err != nil {
 		t.Fatalf("Failed to start command: %s", err)
 	}
@@ -50,7 +50,7 @@ func TestCmdStreamBad(t *testing.T) {
 
 func TestCmdStreamGood(t *testing.T) {
 	cmd := exec.Command("/bin/sh", "-c", "echo hello; exit 0")
-	out, err := CmdStream(cmd)
+	out, err := CmdStream(cmd, nil, nil)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -83,7 +83,7 @@ func tarUntar(t *testing.T, origin string, compression Compression) error {
 		return err
 	}
 	defer os.RemoveAll(tmp)
-	if err := Untar(archive, tmp); err != nil {
+	if err := Untar(archive, tmp, nil); err != nil {
 		return err
 	}
 	if _, err := os.Stat(tmp); err != nil {

archive/changes.go

@@ -0,0 +1,317 @@
+package archive
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"syscall"
+)
+
+type ChangeType int
+
+const (
+	ChangeModify = iota
+	ChangeAdd
+	ChangeDelete
+)
+
+type Change struct {
+	Path string
+	Kind ChangeType
+}
+
+func (change *Change) String() string {
+	var kind string
+	switch change.Kind {
+	case ChangeModify:
+		kind = "C"
+	case ChangeAdd:
+		kind = "A"
+	case ChangeDelete:
+		kind = "D"
+	}
+	return fmt.Sprintf("%s %s", kind, change.Path)
+}
+
+func Changes(layers []string, rw string) ([]Change, error) {
+	var changes []Change
+	err := filepath.Walk(rw, func(path string, f os.FileInfo, err error) error {
+		if err != nil {
+			return err
+		}
+
+		// Rebase path
+		path, err = filepath.Rel(rw, path)
+		if err != nil {
+			return err
+		}
+		path = filepath.Join("/", path)
+
+		// Skip root
+		if path == "/" {
+			return nil
+		}
+
+		// Skip AUFS metadata
+		if matched, err := filepath.Match("/.wh..wh.*", path); err != nil || matched {
+			return err
+		}
+
+		change := Change{
+			Path: path,
+		}
+
+		// Find out what kind of modification happened
+		file := filepath.Base(path)
+		// If there is a whiteout, then the file was removed
+		if strings.HasPrefix(file, ".wh.") {
+			originalFile := file[len(".wh."):]
+			change.Path = filepath.Join(filepath.Dir(path), originalFile)
+			change.Kind = ChangeDelete
+		} else {
+			// Otherwise, the file was added
+			change.Kind = ChangeAdd
+
+			// ...Unless it already existed in a top layer, in which case, it's a modification
+			for _, layer := range layers {
+				stat, err := os.Stat(filepath.Join(layer, path))
+				if err != nil && !os.IsNotExist(err) {
+					return err
+				}
+				if err == nil {
+					// The file existed in the top layer, so that's a modification
+
+					// However, if it's a directory, maybe it wasn't actually modified.
+					// If you modify /foo/bar/baz, then /foo will be part of the changed files only because it's the parent of bar
+					if stat.IsDir() && f.IsDir() {
+						if f.Size() == stat.Size() && f.Mode() == stat.Mode() && f.ModTime() == stat.ModTime() {
+							// Both directories are the same, don't record the change
+							return nil
+						}
+					}
+					change.Kind = ChangeModify
+					break
+				}
+			}
+		}
+
+		// Record change
+		changes = append(changes, change)
+		return nil
+	})
+	if err != nil && !os.IsNotExist(err) {
+		return nil, err
+	}
+	return changes, nil
+}
+
+type FileInfo struct {
+	parent   *FileInfo
+	name     string
+	stat     syscall.Stat_t
+	children map[string]*FileInfo
+}
+
+func (root *FileInfo) LookUp(path string) *FileInfo {
+	parent := root
+	if path == "/" {
+		return root
+	}
+
+	pathElements := strings.Split(path, "/")
+	for _, elem := range pathElements {
+		if elem != "" {
+			child := parent.children[elem]
+			if child == nil {
+				return nil
+			}
+			parent = child
+		}
+	}
+	return parent
+}
+
+func (info *FileInfo) path() string {
+	if info.parent == nil {
+		return "/"
+	}
+	return filepath.Join(info.parent.path(), info.name)
+}
+
+func (info *FileInfo) isDir() bool {
+	return info.parent == nil || info.stat.Mode&syscall.S_IFDIR == syscall.S_IFDIR
+}
+
+func (info *FileInfo) addChanges(oldInfo *FileInfo, changes *[]Change) {
+	if oldInfo == nil {
+		// add
+		change := Change{
+			Path: info.path(),
+			Kind: ChangeAdd,
+		}
+		*changes = append(*changes, change)
+	}
+
+	// We make a copy so we can modify it to detect additions
+	// also, we only recurse on the old dir if the new info is a directory
+	// otherwise any previous delete/change is considered recursive
+	oldChildren := make(map[string]*FileInfo)
+	if oldInfo != nil && info.isDir() {
+		for k, v := range oldInfo.children {
+			oldChildren[k] = v
+		}
+	}
+
+	for name, newChild := range info.children {
+		oldChild, _ := oldChildren[name]
+		if oldChild != nil {
+			// change?
+			oldStat := &oldChild.stat
+			newStat := &newChild.stat
+			// Note: We can't compare inode or ctime or blocksize here, because these change
+			// when copying a file into a container. However, that is not generally a problem
+			// because any content change will change mtime, and any status change should
+			// be visible when actually comparing the stat fields. The only time this
+			// breaks down is if some code intentionally hides a change by setting
+			// back mtime
+			if oldStat.Mode != newStat.Mode ||
+				oldStat.Uid != newStat.Uid ||
+				oldStat.Gid != newStat.Gid ||
+				oldStat.Rdev != newStat.Rdev ||
+				// Don't look at size for dirs, its not a good measure of change
+				(oldStat.Size != newStat.Size && oldStat.Mode&syscall.S_IFDIR != syscall.S_IFDIR) ||
+				getLastModification(oldStat) != getLastModification(newStat) {
+				change := Change{
+					Path: newChild.path(),
+					Kind: ChangeModify,
+				}
+				*changes = append(*changes, change)
+			}
+
+			// Remove from copy so we can detect deletions
+			delete(oldChildren, name)
+		}
+
+		newChild.addChanges(oldChild, changes)
+	}
+	for _, oldChild := range oldChildren {
+		// delete
+		change := Change{
+			Path: oldChild.path(),
+			Kind: ChangeDelete,
+		}
+		*changes = append(*changes, change)
+	}
+
+}
+
+func (info *FileInfo) Changes(oldInfo *FileInfo) []Change {
+	var changes []Change
+
+	info.addChanges(oldInfo, &changes)
+
+	return changes
+}
+
+func newRootFileInfo() *FileInfo {
+	root := &FileInfo{
+		name:     "/",
+		children: make(map[string]*FileInfo),
+	}
+	return root
+}
+
+func collectFileInfo(sourceDir string) (*FileInfo, error) {
+	root := newRootFileInfo()
+
+	err := filepath.Walk(sourceDir, func(path string, f os.FileInfo, err error) error {
+		if err != nil {
+			return err
+		}
+
+		// Rebase path
+		relPath, err := filepath.Rel(sourceDir, path)
+		if err != nil {
+			return err
+		}
+		relPath = filepath.Join("/", relPath)
+
+		if relPath == "/" {
+			return nil
+		}
+
+		parent := root.LookUp(filepath.Dir(relPath))
+		if parent == nil {
+			return fmt.Errorf("collectFileInfo: Unexpectedly no parent for %s", relPath)
+		}
+
+		info := &FileInfo{
+			name:     filepath.Base(relPath),
+			children: make(map[string]*FileInfo),
+			parent:   parent,
+		}
+
+		if err := syscall.Lstat(path, &info.stat); err != nil {
+			return err
+		}
+
+		parent.children[info.name] = info
+
+		return nil
+	})
+	if err != nil {
+		return nil, err
+	}
+	return root, nil
+}
+
+// Compare two directories and generate an array of Change objects describing the changes
+func ChangesDirs(newDir, oldDir string) ([]Change, error) {
+	oldRoot, err := collectFileInfo(oldDir)
+	if err != nil {
+		return nil, err
+	}
+	newRoot, err := collectFileInfo(newDir)
+	if err != nil {
+		return nil, err
+	}
+
+	return newRoot.Changes(oldRoot), nil
+}
+
+func ChangesSize(newDir string, changes []Change) int64 {
+	var size int64
+	for _, change := range changes {
+		if change.Kind == ChangeModify || change.Kind == ChangeAdd {
+			file := filepath.Join(newDir, change.Path)
+			fileInfo, _ := os.Lstat(file)
+			if fileInfo != nil && !fileInfo.IsDir() {
+				size += fileInfo.Size()
+			}
+		}
+	}
+	return size
+}
+
+func ExportChanges(dir string, changes []Change) (Archive, error) {
+	files := make([]string, 0)
+	deletions := make([]string, 0)
+	for _, change := range changes {
+		if change.Kind == ChangeModify || change.Kind == ChangeAdd {
+			files = append(files, change.Path)
+		}
+		if change.Kind == ChangeDelete {
+			base := filepath.Base(change.Path)
+			dir := filepath.Dir(change.Path)
+			deletions = append(deletions, filepath.Join(dir, ".wh."+base))
+		}
+	}
+	// FIXME: Why do we create whiteout files inside Tar code ?
+	return TarFilter(dir, &TarOptions{
+		Compression: Uncompressed,
+		Includes:    files,
+		Recursive:   false,
+		CreateFiles: deletions,
+	})
+}

archive/changes_test.go

@@ -0,0 +1,305 @@
+package archive
+
+import (
+	"io/ioutil"
+	"os"
+	"os/exec"
+	"path"
+	"sort"
+	"testing"
+	"time"
+)
+
+func max(x, y int) int {
+	if x >= y {
+		return x
+	}
+	return y
+}
+
+func copyDir(src, dst string) error {
+	cmd := exec.Command("cp", "-a", src, dst)
+	if err := cmd.Run(); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Helper to sort []Change by path
+type byPath struct{ changes []Change }
+
+func (b byPath) Less(i, j int) bool { return b.changes[i].Path < b.changes[j].Path }
+func (b byPath) Len() int           { return len(b.changes) }
+func (b byPath) Swap(i, j int)      { b.changes[i], b.changes[j] = b.changes[j], b.changes[i] }
+
+type FileType uint32
+
+const (
+	Regular FileType = iota
+	Dir
+	Symlink
+)
+
+type FileData struct {
+	filetype    FileType
+	path        string
+	contents    string
+	permissions os.FileMode
+}
+
+func createSampleDir(t *testing.T, root string) {
+	files := []FileData{
+		{Regular, "file1", "file1\n", 0600},
+		{Regular, "file2", "file2\n", 0666},
+		{Regular, "file3", "file3\n", 0404},
+		{Regular, "file4", "file4\n", 0600},
+		{Regular, "file5", "file5\n", 0600},
+		{Regular, "file6", "file6\n", 0600},
+		{Regular, "file7", "file7\n", 0600},
+		{Dir, "dir1", "", 0740},
+		{Regular, "dir1/file1-1", "file1-1\n", 01444},
+		{Regular, "dir1/file1-2", "file1-2\n", 0666},
+		{Dir, "dir2", "", 0700},
+		{Regular, "dir2/file2-1", "file2-1\n", 0666},
+		{Regular, "dir2/file2-2", "file2-2\n", 0666},
+		{Dir, "dir3", "", 0700},
+		{Regular, "dir3/file3-1", "file3-1\n", 0666},
+		{Regular, "dir3/file3-2", "file3-2\n", 0666},
+		{Dir, "dir4", "", 0700},
+		{Regular, "dir4/file3-1", "file4-1\n", 0666},
+		{Regular, "dir4/file3-2", "file4-2\n", 0666},
+		{Symlink, "symlink1", "target1", 0666},
+		{Symlink, "symlink2", "target2", 0666},
+	}
+
+	now := time.Now()
+	for _, info := range files {
+		p := path.Join(root, info.path)
+		if info.filetype == Dir {
+			if err := os.MkdirAll(p, info.permissions); err != nil {
+				t.Fatal(err)
+			}
+		} else if info.filetype == Regular {
+			if err := ioutil.WriteFile(p, []byte(info.contents), info.permissions); err != nil {
+				t.Fatal(err)
+			}
+		} else if info.filetype == Symlink {
+			if err := os.Symlink(info.contents, p); err != nil {
+				t.Fatal(err)
+			}
+		}
+
+		if info.filetype != Symlink {
+			// Set a consistent ctime, atime for all files and dirs
+			if err := os.Chtimes(p, now, now); err != nil {
+				t.Fatal(err)
+			}
+		}
+	}
+}
+
+// Create an directory, copy it, make sure we report no changes between the two
+func TestChangesDirsEmpty(t *testing.T) {
+	src, err := ioutil.TempDir("", "docker-changes-test")
+	if err != nil {
+		t.Fatal(err)
+	}
+	createSampleDir(t, src)
+	dst := src + "-copy"
+	if err := copyDir(src, dst); err != nil {
+		t.Fatal(err)
+	}
+	changes, err := ChangesDirs(dst, src)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if len(changes) != 0 {
+		t.Fatalf("Reported changes for identical dirs: %v", changes)
+	}
+	os.RemoveAll(src)
+	os.RemoveAll(dst)
+}
+
+func mutateSampleDir(t *testing.T, root string) {
+	// Remove a regular file
+	if err := os.RemoveAll(path.Join(root, "file1")); err != nil {
+		t.Fatal(err)
+	}
+
+	// Remove a directory
+	if err := os.RemoveAll(path.Join(root, "dir1")); err != nil {
+		t.Fatal(err)
+	}
+
+	// Remove a symlink
+	if err := os.RemoveAll(path.Join(root, "symlink1")); err != nil {
+		t.Fatal(err)
+	}
+
+	// Rewrite a file
+	if err := ioutil.WriteFile(path.Join(root, "file2"), []byte("fileN\n"), 0777); err != nil {
+		t.Fatal(err)
+	}
+
+	// Replace a file
+	if err := os.RemoveAll(path.Join(root, "file3")); err != nil {
+		t.Fatal(err)
+	}
+	if err := ioutil.WriteFile(path.Join(root, "file3"), []byte("fileM\n"), 0404); err != nil {
+		t.Fatal(err)
+	}
+
+	// Touch file
+	if err := os.Chtimes(path.Join(root, "file4"), time.Now(), time.Now()); err != nil {
+		t.Fatal(err)
+	}
+
+	// Replace file with dir
+	if err := os.RemoveAll(path.Join(root, "file5")); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.MkdirAll(path.Join(root, "file5"), 0666); err != nil {
+		t.Fatal(err)
+	}
+
+	// Create new file
+	if err := ioutil.WriteFile(path.Join(root, "filenew"), []byte("filenew\n"), 0777); err != nil {
+		t.Fatal(err)
+	}
+
+	// Create new dir
+	if err := os.MkdirAll(path.Join(root, "dirnew"), 0766); err != nil {
+		t.Fatal(err)
+	}
+
+	// Create a new symlink
+	if err := os.Symlink("targetnew", path.Join(root, "symlinknew")); err != nil {
+		t.Fatal(err)
+	}
+
+	// Change a symlink
+	if err := os.RemoveAll(path.Join(root, "symlink2")); err != nil {
+		t.Fatal(err)
+	}
+	if err := os.Symlink("target2change", path.Join(root, "symlink2")); err != nil {
+		t.Fatal(err)
+	}
+
+	// Replace dir with file
+	if err := os.RemoveAll(path.Join(root, "dir2")); err != nil {
+		t.Fatal(err)
+	}
+	if err := ioutil.WriteFile(path.Join(root, "dir2"), []byte("dir2\n"), 0777); err != nil {
+		t.Fatal(err)
+	}
+
+	// Touch dir
+	if err := os.Chtimes(path.Join(root, "dir3"), time.Now(), time.Now()); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestChangesDirsMutated(t *testing.T) {
+	src, err := ioutil.TempDir("", "docker-changes-test")
+	if err != nil {
+		t.Fatal(err)
+	}
+	createSampleDir(t, src)
+	dst := src + "-copy"
+	if err := copyDir(src, dst); err != nil {
+		t.Fatal(err)
+	}
+	defer os.RemoveAll(src)
+	defer os.RemoveAll(dst)
+
+	mutateSampleDir(t, dst)
+
+	changes, err := ChangesDirs(dst, src)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	sort.Sort(byPath{changes})
+
+	expectedChanges := []Change{
+		{"/dir1", ChangeDelete},
+		{"/dir2", ChangeModify},
+		{"/dir3", ChangeModify},
+		{"/dirnew", ChangeAdd},
+		{"/file1", ChangeDelete},
+		{"/file2", ChangeModify},
+		{"/file3", ChangeModify},
+		{"/file4", ChangeModify},
+		{"/file5", ChangeModify},
+		{"/filenew", ChangeAdd},
+		{"/symlink1", ChangeDelete},
+		{"/symlink2", ChangeModify},
+		{"/symlinknew", ChangeAdd},
+	}
+
+	for i := 0; i < max(len(changes), len(expectedChanges)); i++ {
+		if i >= len(expectedChanges) {
+			t.Fatalf("unexpected change %s\n", changes[i].String())
+		}
+		if i >= len(changes) {
+			t.Fatalf("no change for expected change %s\n", expectedChanges[i].String())
+		}
+		if changes[i].Path == expectedChanges[i].Path {
+			if changes[i] != expectedChanges[i] {
+				t.Fatalf("Wrong change for %s, expected %s, got %s\n", changes[i].Path, changes[i].String(), expectedChanges[i].String())
+			}
+		} else if changes[i].Path < expectedChanges[i].Path {
+			t.Fatalf("unexpected change %s\n", changes[i].String())
+		} else {
+			t.Fatalf("no change for expected change %s != %s\n", expectedChanges[i].String(), changes[i].String())
+		}
+	}
+}
+
+func TestApplyLayer(t *testing.T) {
+	t.Skip("Skipping TestApplyLayer due to known failures") // Disable this for now as it is broken
+	return
+
+	// src, err := ioutil.TempDir("", "docker-changes-test")
+	// if err != nil {
+	// 	t.Fatal(err)
+	// }
+	// createSampleDir(t, src)
+	// dst := src + "-copy"
+	// if err := copyDir(src, dst); err != nil {
+	// 	t.Fatal(err)
+	// }
+	// mutateSampleDir(t, dst)
+
+	// changes, err := ChangesDirs(dst, src)
+	// if err != nil {
+	// 	t.Fatal(err)
+	// }
+
+	// layer, err := ExportChanges(dst, changes)
+	// if err != nil {
+	// 	t.Fatal(err)
+	// }
+
+	// layerCopy, err := NewTempArchive(layer, "")
+	// if err != nil {
+	// 	t.Fatal(err)
+	// }
+
+	// if err := ApplyLayer(src, layerCopy); err != nil {
+	// 	t.Fatal(err)
+	// }
+
+	// changes2, err := ChangesDirs(src, dst)
+	// if err != nil {
+	// 	t.Fatal(err)
+	// }
+
+	// if len(changes2) != 0 {
+	// 	t.Fatalf("Unexpected differences after re applying mutation: %v", changes)
+	// }
+
+	// os.RemoveAll(src)
+	// os.RemoveAll(dst)
+}

archive/diff.go

@@ -0,0 +1,97 @@
+package archive
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"syscall"
+	"time"
+)
+
+// ApplyLayer parses a diff in the standard layer format from `layer`, and
+// applies it to the directory `dest`.
+func ApplyLayer(dest string, layer Archive) error {
+	// Poor man's diff applyer in 2 steps:
+
+	// Step 1: untar everything in place
+	if err := Untar(layer, dest, nil); err != nil {
+		return err
+	}
+
+	modifiedDirs := make(map[string]*syscall.Stat_t)
+	addDir := func(file string) {
+		d := filepath.Dir(file)
+		if _, exists := modifiedDirs[d]; !exists {
+			if s, err := os.Lstat(d); err == nil {
+				if sys := s.Sys(); sys != nil {
+					if stat, ok := sys.(*syscall.Stat_t); ok {
+						modifiedDirs[d] = stat
+					}
+				}
+			}
+		}
+	}
+
+	// Step 2: walk for whiteouts and apply them, removing them in the process
+	err := filepath.Walk(dest, func(fullPath string, f os.FileInfo, err error) error {
+		if err != nil {
+			if os.IsNotExist(err) {
+				// This happens in the case of whiteouts in parent dir removing a directory
+				// We just ignore it
+				return filepath.SkipDir
+			}
+			return err
+		}
+
+		// Rebase path
+		path, err := filepath.Rel(dest, fullPath)
+		if err != nil {
+			return err
+		}
+		path = filepath.Join("/", path)
+
+		// Skip AUFS metadata
+		if matched, err := filepath.Match("/.wh..wh.*", path); err != nil {
+			return err
+		} else if matched {
+			addDir(fullPath)
+			if err := os.RemoveAll(fullPath); err != nil {
+				return err
+			}
+		}
+
+		filename := filepath.Base(path)
+		if strings.HasPrefix(filename, ".wh.") {
+			rmTargetName := filename[len(".wh."):]
+			rmTargetPath := filepath.Join(filepath.Dir(fullPath), rmTargetName)
+
+			// Remove the file targeted by the whiteout
+			addDir(rmTargetPath)
+			if err := os.RemoveAll(rmTargetPath); err != nil {
+				return err
+			}
+			// Remove the whiteout itself
+			addDir(fullPath)
+			if err := os.RemoveAll(fullPath); err != nil {
+				return err
+			}
+		}
+		return nil
+	})
+	if err != nil {
+		return err
+	}
+
+	for k, v := range modifiedDirs {
+		lastAccess := getLastAccess(v)
+		lastModification := getLastModification(v)
+		aTime := time.Unix(lastAccess.Unix())
+		mTime := time.Unix(lastModification.Unix())
+
+		if err := os.Chtimes(k, aTime, mTime); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}

archive/stat_darwin.go

@@ -0,0 +1,11 @@
+package archive
+
+import "syscall"
+
+func getLastAccess(stat *syscall.Stat_t) syscall.Timespec {
+	return stat.Atimespec
+}
+
+func getLastModification(stat *syscall.Stat_t) syscall.Timespec {
+	return stat.Mtimespec
+}

archive/stat_linux.go

@@ -0,0 +1,11 @@
+package archive
+
+import "syscall"
+
+func getLastAccess(stat *syscall.Stat_t) syscall.Timespec {
+	return stat.Atim
+}
+
+func getLastModification(stat *syscall.Stat_t) syscall.Timespec {
+	return stat.Mtim
+}

auth/auth.go

@@ -63,7 +63,7 @@ func decodeAuth(authStr string) (string, string, error) {
 	if n > decLen {
 		return "", "", fmt.Errorf("Something went wrong decoding auth config")
 	}
-	arr := strings.Split(string(decoded), ":")
+	arr := strings.SplitN(string(decoded), ":", 2)
 	if len(arr) != 2 {
 		return "", "", fmt.Errorf("Invalid auth configuration file")
 	}
@@ -223,6 +223,28 @@ func Login(authConfig *AuthConfig, factory *utils.HTTPRequestFactory) (string, e
 		} else {
 			return "", fmt.Errorf("Registration: %s", reqBody)
 		}
+	} else if reqStatusCode == 401 {
+		// This case would happen with private registries where /v1/users is
+		// protected, so people can use `docker login` as an auth check.
+		req, err := factory.NewRequest("GET", serverAddress+"users/", nil)
+		req.SetBasicAuth(authConfig.Username, authConfig.Password)
+		resp, err := client.Do(req)
+		if err != nil {
+			return "", err
+		}
+		defer resp.Body.Close()
+		body, err := ioutil.ReadAll(resp.Body)
+		if err != nil {
+			return "", err
+    	}
+		if resp.StatusCode == 200 {
+			status = "Login Succeeded"
+		} else if resp.StatusCode == 401 {
+			return "", fmt.Errorf("Wrong login/password, please try again")
+		} else {
+			return "", fmt.Errorf("Login: %s (Code: %d; Headers: %s)", body,
+				resp.StatusCode, resp.Header)
+		}
 	} else {
 		return "", fmt.Errorf("Unexpected status code [%d] : %s", reqStatusCode, reqBody)
 	}

buildfile.go

@@ -36,14 +36,19 @@ type buildFile struct {
 	tmpContainers map[string]struct{}
 	tmpImages     map[string]struct{}
 
-	out io.Writer
+	outStream io.Writer
+	errStream io.Writer
+
+	// Deprecated, original writer used for ImagePull. To be removed.
+	outOld io.Writer
+	sf     *utils.StreamFormatter
 }
 
 func (b *buildFile) clearTmp(containers map[string]struct{}) {
 	for c := range containers {
 		tmp := b.runtime.Get(c)
 		b.runtime.Destroy(tmp)
-		fmt.Fprintf(b.out, "Removing intermediate container %s\n", utils.TruncateID(c))
+		fmt.Fprintf(b.outStream, "Removing intermediate container %s\n", utils.TruncateID(c))
 	}
 }
 
@@ -52,7 +57,7 @@ func (b *buildFile) CmdFrom(name string) error {
 	if err != nil {
 		if b.runtime.graph.IsNotExist(err) {
 			remote, tag := utils.ParseRepositoryTag(name)
-			if err := b.srv.ImagePull(remote, tag, b.out, utils.NewStreamFormatter(false), nil, nil, true); err != nil {
+			if err := b.srv.ImagePull(remote, tag, b.outOld, b.sf, nil, nil, true); err != nil {
 				return err
 			}
 			image, err = b.runtime.repositories.LookupImage(name)
@@ -100,7 +105,7 @@ func (b *buildFile) CmdRun(args string) error {
 		if cache, err := b.srv.ImageGetCached(b.image, b.config); err != nil {
 			return err
 		} else if cache != nil {
-			fmt.Fprintf(b.out, " ---> Using cache\n")
+			fmt.Fprintf(b.outStream, " ---> Using cache\n")
 			utils.Debugf("[BUILDER] Use cached version")
 			b.image = cache.ID
 			return nil
@@ -241,7 +246,7 @@ func (b *buildFile) CmdVolume(args string) error {
 		volume = []string{args}
 	}
 	if b.config.Volumes == nil {
-		b.config.Volumes = NewPathOpts()
+		b.config.Volumes = map[string]struct{}{}
 	}
 	for _, v := range volume {
 		b.config.Volumes[v] = struct{}{}
@@ -253,7 +258,7 @@ func (b *buildFile) CmdVolume(args string) error {
 }
 
 func (b *buildFile) addRemote(container *Container, orig, dest string) error {
-	file, err := utils.Download(orig, ioutil.Discard)
+	file, err := utils.Download(orig)
 	if err != nil {
 		return err
 	}
@@ -288,7 +293,7 @@ func (b *buildFile) addContext(container *Container, orig, dest string) error {
 		destPath = destPath + "/"
 	}
 	if !strings.HasPrefix(origPath, b.context) {
-		return fmt.Errorf("Forbidden path: %s", origPath)
+		return fmt.Errorf("Forbidden path outside the build context: %s (%s)", orig, origPath)
 	}
 	fi, err := os.Stat(origPath)
 	if err != nil {
@@ -364,6 +369,34 @@ func (b *buildFile) CmdAdd(args string) error {
 	return nil
 }
 
+type StdoutFormater struct {
+	io.Writer
+	*utils.StreamFormatter
+}
+
+func (sf *StdoutFormater) Write(buf []byte) (int, error) {
+	formattedBuf := sf.StreamFormatter.FormatStream(string(buf))
+	n, err := sf.Writer.Write(formattedBuf)
+	if n != len(formattedBuf) {
+		return n, io.ErrShortWrite
+	}
+	return len(buf), err
+}
+
+type StderrFormater struct {
+	io.Writer
+	*utils.StreamFormatter
+}
+
+func (sf *StderrFormater) Write(buf []byte) (int, error) {
+	formattedBuf := sf.StreamFormatter.FormatStream("\033[91m" + string(buf) + "\033[0m")
+	n, err := sf.Writer.Write(formattedBuf)
+	if n != len(formattedBuf) {
+		return n, io.ErrShortWrite
+	}
+	return len(buf), err
+}
+
 func (b *buildFile) run() (string, error) {
 	if b.image == "" {
 		return "", fmt.Errorf("Please provide a source image with `from` prior to run")
@@ -376,7 +409,7 @@ func (b *buildFile) run() (string, error) {
 		return "", err
 	}
 	b.tmpContainers[c.ID] = struct{}{}
-	fmt.Fprintf(b.out, " ---> Running in %s\n", utils.TruncateID(c.ID))
+	fmt.Fprintf(b.outStream, " ---> Running in %s\n", utils.TruncateID(c.ID))
 
 	// override the entry point that may have been picked up from the base image
 	c.Path = b.config.Cmd[0]
@@ -386,7 +419,7 @@ func (b *buildFile) run() (string, error) {
 
 	if b.verbose {
 		errCh = utils.Go(func() error {
-			return <-c.Attach(nil, nil, b.out, b.out)
+			return <-c.Attach(nil, nil, b.outStream, b.errStream)
 		})
 	}
 
@@ -403,7 +436,11 @@ func (b *buildFile) run() (string, error) {
 
 	// Wait for it to finish
 	if ret := c.Wait(); ret != 0 {
-		return "", fmt.Errorf("The command %v returned a non-zero code: %d", b.config.Cmd, ret)
+		err := &utils.JSONError{
+			Message: fmt.Sprintf("The command %v returned a non-zero code: %d", b.config.Cmd, ret),
+			Code:    ret,
+		}
+		return "", err
 	}
 
 	return c.ID, nil
@@ -424,7 +461,7 @@ func (b *buildFile) commit(id string, autoCmd []string, comment string) error {
 			if cache, err := b.srv.ImageGetCached(b.image, b.config); err != nil {
 				return err
 			} else if cache != nil {
-				fmt.Fprintf(b.out, " ---> Using cache\n")
+				fmt.Fprintf(b.outStream, " ---> Using cache\n")
 				utils.Debugf("[BUILDER] Use cached version")
 				b.image = cache.ID
 				return nil
@@ -438,10 +475,10 @@ func (b *buildFile) commit(id string, autoCmd []string, comment string) error {
 			return err
 		}
 		for _, warning := range warnings {
-			fmt.Fprintf(b.out, " ---> [Warning] %s\n", warning)
+			fmt.Fprintf(b.outStream, " ---> [Warning] %s\n", warning)
 		}
 		b.tmpContainers[container.ID] = struct{}{}
-		fmt.Fprintf(b.out, " ---> Running in %s\n", utils.TruncateID(container.ID))
+		fmt.Fprintf(b.outStream, " ---> Running in %s\n", utils.TruncateID(container.ID))
 		id = container.ID
 		if err := container.EnsureMounted(); err != nil {
 			return err
@@ -476,7 +513,7 @@ func (b *buildFile) Build(context io.Reader) (string, error) {
 	if err != nil {
 		return "", err
 	}
-	if err := archive.Untar(context, name); err != nil {
+	if err := archive.Untar(context, name, nil); err != nil {
 		return "", err
 	}
 	defer os.RemoveAll(name)
@@ -507,22 +544,22 @@ func (b *buildFile) Build(context io.Reader) (string, error) {
 
 		method, exists := reflect.TypeOf(b).MethodByName("Cmd" + strings.ToUpper(instruction[:1]) + strings.ToLower(instruction[1:]))
 		if !exists {
-			fmt.Fprintf(b.out, "# Skipping unknown instruction %s\n", strings.ToUpper(instruction))
+			fmt.Fprintf(b.errStream, "# Skipping unknown instruction %s\n", strings.ToUpper(instruction))
 			continue
 		}
 
 		stepN += 1
-		fmt.Fprintf(b.out, "Step %d : %s %s\n", stepN, strings.ToUpper(instruction), arguments)
+		fmt.Fprintf(b.outStream, "Step %d : %s %s\n", stepN, strings.ToUpper(instruction), arguments)
 
 		ret := method.Func.Call([]reflect.Value{reflect.ValueOf(b), reflect.ValueOf(arguments)})[0].Interface()
 		if ret != nil {
 			return "", ret.(error)
 		}
 
-		fmt.Fprintf(b.out, " ---> %v\n", utils.TruncateID(b.image))
+		fmt.Fprintf(b.outStream, " ---> %s\n", utils.TruncateID(b.image))
 	}
 	if b.image != "" {
-		fmt.Fprintf(b.out, "Successfully built %s\n", utils.TruncateID(b.image))
+		fmt.Fprintf(b.outStream, "Successfully built %s\n", utils.TruncateID(b.image))
 		if b.rm {
 			b.clearTmp(b.tmpContainers)
 		}
@@ -531,16 +568,19 @@ func (b *buildFile) Build(context io.Reader) (string, error) {
 	return "", fmt.Errorf("An error occurred during the build\n")
 }
 
-func NewBuildFile(srv *Server, out io.Writer, verbose, utilizeCache, rm bool) BuildFile {
+func NewBuildFile(srv *Server, outStream, errStream io.Writer, verbose, utilizeCache, rm bool, outOld io.Writer, sf *utils.StreamFormatter) BuildFile {
 	return &buildFile{
 		runtime:       srv.runtime,
 		srv:           srv,
 		config:        &Config{},
-		out:           out,
+		outStream:     outStream,
+		errStream:     errStream,
 		tmpContainers: make(map[string]struct{}),
 		tmpImages:     make(map[string]struct{}),
 		verbose:       verbose,
 		utilizeCache:  utilizeCache,
 		rm:            rm,
+		sf:            sf,
+		outOld:        outOld,
 	}
 }

changes.go

@@ -1,106 +0,0 @@
-package docker
-
-import (
-	"fmt"
-	"os"
-	"path/filepath"
-	"strings"
-)
-
-type ChangeType int
-
-const (
-	ChangeModify = iota
-	ChangeAdd
-	ChangeDelete
-)
-
-type Change struct {
-	Path string
-	Kind ChangeType
-}
-
-func (change *Change) String() string {
-	var kind string
-	switch change.Kind {
-	case ChangeModify:
-		kind = "C"
-	case ChangeAdd:
-		kind = "A"
-	case ChangeDelete:
-		kind = "D"
-	}
-	return fmt.Sprintf("%s %s", kind, change.Path)
-}
-
-func Changes(layers []string, rw string) ([]Change, error) {
-	var changes []Change
-	err := filepath.Walk(rw, func(path string, f os.FileInfo, err error) error {
-		if err != nil {
-			return err
-		}
-
-		// Rebase path
-		path, err = filepath.Rel(rw, path)
-		if err != nil {
-			return err
-		}
-		path = filepath.Join("/", path)
-
-		// Skip root
-		if path == "/" {
-			return nil
-		}
-
-		// Skip AUFS metadata
-		if matched, err := filepath.Match("/.wh..wh.*", path); err != nil || matched {
-			return err
-		}
-
-		change := Change{
-			Path: path,
-		}
-
-		// Find out what kind of modification happened
-		file := filepath.Base(path)
-		// If there is a whiteout, then the file was removed
-		if strings.HasPrefix(file, ".wh.") {
-			originalFile := file[len(".wh."):]
-			change.Path = filepath.Join(filepath.Dir(path), originalFile)
-			change.Kind = ChangeDelete
-		} else {
-			// Otherwise, the file was added
-			change.Kind = ChangeAdd
-
-			// ...Unless it already existed in a top layer, in which case, it's a modification
-			for _, layer := range layers {
-				stat, err := os.Stat(filepath.Join(layer, path))
-				if err != nil && !os.IsNotExist(err) {
-					return err
-				}
-				if err == nil {
-					// The file existed in the top layer, so that's a modification
-
-					// However, if it's a directory, maybe it wasn't actually modified.
-					// If you modify /foo/bar/baz, then /foo will be part of the changed files only because it's the parent of bar
-					if stat.IsDir() && f.IsDir() {
-						if f.Size() == stat.Size() && f.Mode() == stat.Mode() && f.ModTime() == stat.ModTime() {
-							// Both directories are the same, don't record the change
-							return nil
-						}
-					}
-					change.Kind = ChangeModify
-					break
-				}
-			}
-		}
-
-		// Record change
-		changes = append(changes, change)
-		return nil
-	})
-	if err != nil && !os.IsNotExist(err) {
-		return nil, err
-	}
-	return changes, nil
-}

commands.go

@@ -23,7 +23,6 @@ import (
 	"os"
 	"os/signal"
 	"path"
-	"path/filepath"
 	"reflect"
 	"regexp"
 	"runtime"
@@ -32,6 +31,7 @@ import (
 	"strings"
 	"syscall"
 	"text/tabwriter"
+	"text/template"
 	"time"
 )
 
@@ -195,6 +195,10 @@ func (cli *DockerCli) CmdBuild(args ...string) error {
 		if _, err := os.Stat(cmd.Arg(0)); err != nil {
 			return err
 		}
+		filename := path.Join(cmd.Arg(0), "Dockerfile")
+		if _, err = os.Stat(filename); os.IsNotExist(err) {
+			return fmt.Errorf("no Dockerfile found in %s", cmd.Arg(0))
+		}
 		context, err = archive.Tar(cmd.Arg(0), archive.Uncompressed)
 	}
 	var body io.Reader
@@ -202,7 +206,7 @@ func (cli *DockerCli) CmdBuild(args ...string) error {
 	// FIXME: ProgressReader shouldn't be this annoying to use
 	if context != nil {
 		sf := utils.NewStreamFormatter(false)
-		body = utils.ProgressReader(ioutil.NopCloser(context), 0, cli.err, sf.FormatProgress("", "Uploading context", "%v bytes%0.0s%0.0s"), sf, true)
+		body = utils.ProgressReader(ioutil.NopCloser(context), 0, cli.err, sf, true, "", "Uploading context")
 	}
 	// Upload the build context
 	v := &url.Values{}
@@ -220,42 +224,16 @@ func (cli *DockerCli) CmdBuild(args ...string) error {
 	if *rm {
 		v.Set("rm", "1")
 	}
-	req, err := http.NewRequest("POST", fmt.Sprintf("/v%g/build?%s", APIVERSION, v.Encode()), body)
-	if err != nil {
-		return err
-	}
+
+	headers := http.Header(make(map[string][]string))
 	if context != nil {
-		req.Header.Set("Content-Type", "application/tar")
+		headers.Set("Content-Type", "application/tar")
 	}
-	dial, err := net.Dial(cli.proto, cli.addr)
-	if err != nil {
-		return err
+	err = cli.stream("POST", fmt.Sprintf("/build?%s", v.Encode()), body, cli.out, headers)
+	if jerr, ok := err.(*utils.JSONError); ok {
+		return &utils.StatusError{Status: jerr.Message, StatusCode: jerr.Code}
 	}
-	clientconn := httputil.NewClientConn(dial, nil)
-	resp, err := clientconn.Do(req)
-	defer clientconn.Close()
-	if err != nil {
-		return err
-	}
-	defer resp.Body.Close()
-	// Check for errors
-	if resp.StatusCode < 200 || resp.StatusCode >= 400 {
-		body, err := ioutil.ReadAll(resp.Body)
-		if err != nil {
-			return err
-		}
-		if len(body) == 0 {
-			return fmt.Errorf("Error: %s", http.StatusText(resp.StatusCode))
-		}
-		return fmt.Errorf("Error: %s", body)
-	}
-
-	// Output the result
-	if _, err := io.Copy(cli.out, resp.Body); err != nil {
-		return err
-	}
-
-	return nil
+	return err
 }
 
 // 'docker login': login / register a user to registry service.
@@ -463,6 +441,10 @@ func (cli *DockerCli) CmdInfo(args ...string) error {
 
 	fmt.Fprintf(cli.out, "Containers: %d\n", out.Containers)
 	fmt.Fprintf(cli.out, "Images: %d\n", out.Images)
+	fmt.Fprintf(cli.out, "Driver: %s\n", out.Driver)
+	for _, pair := range out.DriverStatus {
+		fmt.Fprintf(cli.out, " %s: %s\n", pair[0], pair[1])
+	}
 	if out.Debug || os.Getenv("DEBUG") != "" {
 		fmt.Fprintf(cli.out, "Debug mode (server): %v\n", out.Debug)
 		fmt.Fprintf(cli.out, "Debug mode (client): %v\n", os.Getenv("DEBUG") != "")
@@ -651,6 +633,7 @@ func (cli *DockerCli) CmdStart(args ...string) error {
 
 func (cli *DockerCli) CmdInspect(args ...string) error {
 	cmd := cli.Subcmd("inspect", "CONTAINER|IMAGE [CONTAINER|IMAGE...]", "Return low-level information on a container/image")
+	tmplStr := cmd.String("format", "", "Format the output using the given go template.")
 	if err := cmd.Parse(args); err != nil {
 		return nil
 	}
@@ -659,10 +642,21 @@ func (cli *DockerCli) CmdInspect(args ...string) error {
 		return nil
 	}
 
+	var tmpl *template.Template
+	if *tmplStr != "" {
+		var err error
+		if tmpl, err = template.New("").Parse(*tmplStr); err != nil {
+			fmt.Fprintf(cli.err, "Template parsing error: %v\n", err)
+			return &utils.StatusError{StatusCode: 64,
+				Status: "Template parsing error: " + err.Error()}
+		}
+	}
+
 	indented := new(bytes.Buffer)
+	indented.WriteByte('[')
 	status := 0
 
-	for _, name := range args {
+	for _, name := range cmd.Args() {
 		obj, _, err := cli.call("GET", "/containers/"+name+"/json", nil)
 		if err != nil {
 			obj, _, err = cli.call("GET", "/images/"+name+"/json", nil)
@@ -677,25 +671,42 @@ func (cli *DockerCli) CmdInspect(args ...string) error {
 			}
 		}
 
-		if err = json.Indent(indented, obj, "", "    "); err != nil {
-			fmt.Fprintf(cli.err, "%s\n", err)
-			status = 1
-			continue
+		if tmpl == nil {
+			if err = json.Indent(indented, obj, "", "    "); err != nil {
+				fmt.Fprintf(cli.err, "%s\n", err)
+				status = 1
+				continue
+			}
+		} else {
+			// Has template, will render
+			var value interface{}
+			if err := json.Unmarshal(obj, &value); err != nil {
+				fmt.Fprintf(cli.err, "%s\n", err)
+				status = 1
+				continue
+			}
+			if err := tmpl.Execute(cli.out, value); err != nil {
+				return err
+			}
+			cli.out.Write([]byte{'\n'})
 		}
 		indented.WriteString(",")
 	}
 
-	if indented.Len() > 0 {
+	if indented.Len() > 1 {
 		// Remove trailing ','
 		indented.Truncate(indented.Len() - 1)
 	}
-	fmt.Fprintf(cli.out, "[")
-	if _, err := io.Copy(cli.out, indented); err != nil {
-		return err
+	indented.WriteByte(']')
+
+	if tmpl == nil {
+		if _, err := io.Copy(cli.out, indented); err != nil {
+			return err
+		}
 	}
-	fmt.Fprintf(cli.out, "]")
+
 	if status != 0 {
-		return &utils.StatusError{Status: status}
+		return &utils.StatusError{StatusCode: status}
 	}
 	return nil
 }
@@ -759,16 +770,12 @@ func (cli *DockerCli) CmdPort(args ...string) error {
 		return err
 	}
 
-	if frontends, exists := out.NetworkSettings.Ports[Port(port+"/"+proto)]; exists {
-		if frontends == nil {
-			fmt.Fprintf(cli.out, "%s\n", port)
-		} else {
-			for _, frontend := range frontends {
-				fmt.Fprintf(cli.out, "%s:%s\n", frontend.HostIp, frontend.HostPort)
-			}
+	if frontends, exists := out.NetworkSettings.Ports[Port(port+"/"+proto)]; exists && frontends != nil {
+		for _, frontend := range frontends {
+			fmt.Fprintf(cli.out, "%s:%s\n", frontend.HostIp, frontend.HostPort)
 		}
 	} else {
-		return fmt.Errorf("Error: No private port '%s' allocated on %s", cmd.Arg(1), cmd.Arg(0))
+		return fmt.Errorf("Error: No public port '%s' published for %s", cmd.Arg(1), cmd.Arg(0))
 	}
 	return nil
 }
@@ -847,7 +854,7 @@ func (cli *DockerCli) CmdHistory(args ...string) error {
 				fmt.Fprintf(w, "%s\t", utils.TruncateID(out.ID))
 			}
 
-			fmt.Fprintf(w, "%s ago\t", utils.HumanDuration(time.Now().Sub(time.Unix(out.Created, 0))))
+			fmt.Fprintf(w, "%s ago\t", utils.HumanDuration(time.Now().UTC().Sub(time.Unix(out.Created, 0))))
 
 			if *noTrunc {
 				fmt.Fprintf(w, "%s\t", out.CreatedBy)
@@ -924,7 +931,7 @@ func (cli *DockerCli) CmdKill(args ...string) error {
 }
 
 func (cli *DockerCli) CmdImport(args ...string) error {
-	cmd := cli.Subcmd("import", "URL|- [REPOSITORY[:TAG]]", "Create a new filesystem image from the contents of a tarball(.tar, .tar.gz, .tgz, .bzip, .tar.xz, .txz).")
+	cmd := cli.Subcmd("import", "URL|- [REPOSITORY[:TAG]]", "Create an empty filesystem image and import the contents of the tarball (.tar, .tar.gz, .tgz, .bzip, .tar.xz, .txz) into it, then optionally tag it.")
 
 	if err := cmd.Parse(args); err != nil {
 		return nil
@@ -1128,16 +1135,18 @@ func (cli *DockerCli) CmdImages(args ...string) error {
 		}
 
 		var outs []APIImages
-		err = json.Unmarshal(body, &outs)
-		if err != nil {
+		if err := json.Unmarshal(body, &outs); err != nil {
 			return err
 		}
 
-		var startImageArg = cmd.Arg(0)
-		var startImage APIImages
+		var (
+			startImageArg = cmd.Arg(0)
+			startImage    APIImages
+
+			roots    []APIImages
+			byParent = make(map[string][]APIImages)
+		)
 
-		var roots []APIImages
-		var byParent = make(map[string][]APIImages)
 		for _, image := range outs {
 			if image.ParentId == "" {
 				roots = append(roots, image)
@@ -1189,7 +1198,7 @@ func (cli *DockerCli) CmdImages(args ...string) error {
 
 		w := tabwriter.NewWriter(cli.out, 20, 1, 3, ' ', 0)
 		if !*quiet {
-			fmt.Fprintln(w, "REPOSITORY\tTAG\tIMAGE ID\tCREATED\tSIZE")
+			fmt.Fprintln(w, "REPOSITORY\tTAG\tIMAGE ID\tCREATED\tVIRTUAL SIZE")
 		}
 
 		for _, out := range outs {
@@ -1202,12 +1211,7 @@ func (cli *DockerCli) CmdImages(args ...string) error {
 				}
 
 				if !*quiet {
-					fmt.Fprintf(w, "%s\t%s\t%s\t%s ago\t", repo, tag, out.ID, utils.HumanDuration(time.Now().Sub(time.Unix(out.Created, 0))))
-					if out.VirtualSize > 0 {
-						fmt.Fprintf(w, "%s (virtual %s)\n", utils.HumanSize(out.Size), utils.HumanSize(out.VirtualSize))
-					} else {
-						fmt.Fprintf(w, "%s\n", utils.HumanSize(out.Size))
-					}
+					fmt.Fprintf(w, "%s\t%s\t%s\t%s ago\t%s\n", repo, tag, out.ID, utils.HumanDuration(time.Now().UTC().Sub(time.Unix(out.Created, 0))), utils.HumanSize(out.VirtualSize))
 				} else {
 					fmt.Fprintln(w, out.ID)
 				}
@@ -1344,7 +1348,7 @@ func (cli *DockerCli) CmdPs(args ...string) error {
 			if !*noTrunc {
 				out.Command = utils.Trunc(out.Command, 20)
 			}
-			fmt.Fprintf(w, "%s\t%s\t%s\t%s ago\t%s\t%s\t%s\t", out.ID, out.Image, out.Command, utils.HumanDuration(time.Now().Sub(time.Unix(out.Created, 0))), out.Status, displayablePorts(out.Ports), strings.Join(out.Names, ","))
+			fmt.Fprintf(w, "%s\t%s\t%s\t%s ago\t%s\t%s\t%s\t", out.ID, out.Image, out.Command, utils.HumanDuration(time.Now().UTC().Sub(time.Unix(out.Created, 0))), out.Status, displayablePorts(out.Ports), strings.Join(out.Names, ","))
 			if *size {
 				if out.SizeRootFs > 0 {
 					fmt.Fprintf(w, "%s (virtual %s)\n", utils.HumanSize(out.SizeRw), utils.HumanSize(out.SizeRootFs))
@@ -1451,7 +1455,7 @@ func (cli *DockerCli) CmdEvents(args ...string) error {
 }
 
 func (cli *DockerCli) CmdExport(args ...string) error {
-	cmd := cli.Subcmd("export", "CONTAINER", "Export the contents of a filesystem as a tar archive")
+	cmd := cli.Subcmd("export", "CONTAINER", "Export the contents of a filesystem as a tar archive to STDOUT")
 	if err := cmd.Parse(args); err != nil {
 		return nil
 	}
@@ -1495,6 +1499,7 @@ func (cli *DockerCli) CmdDiff(args ...string) error {
 
 func (cli *DockerCli) CmdLogs(args ...string) error {
 	cmd := cli.Subcmd("logs", "CONTAINER", "Fetch the logs of a container")
+	follow := cmd.Bool("f", false, "Follow log output")
 	if err := cmd.Parse(args); err != nil {
 		return nil
 	}
@@ -1514,7 +1519,15 @@ func (cli *DockerCli) CmdLogs(args ...string) error {
 		return err
 	}
 
-	if err := cli.hijack("POST", "/containers/"+name+"/attach?logs=1&stdout=1&stderr=1", container.Config.Tty, nil, cli.out, cli.err, nil); err != nil {
+	v := url.Values{}
+	v.Set("logs", "1")
+	v.Set("stdout", "1")
+	v.Set("stderr", "1")
+	if *follow && container.State.Running {
+		v.Set("stream", "1")
+	}
+
+	if err := cli.hijack("POST", "/containers/"+name+"/attach?"+v.Encode(), container.Config.Tty, nil, cli.out, cli.err, nil); err != nil {
 		return err
 	}
 	return nil
@@ -1572,6 +1585,15 @@ func (cli *DockerCli) CmdAttach(args ...string) error {
 	if err := cli.hijack("POST", "/containers/"+cmd.Arg(0)+"/attach?"+v.Encode(), container.Config.Tty, in, cli.out, cli.err, nil); err != nil {
 		return err
 	}
+
+	_, status, err := getExitCode(cli, cmd.Arg(0))
+	if err != nil {
+		return err
+	}
+	if status != 0 {
+		return &utils.StatusError{StatusCode: status}
+	}
+
 	return nil
 }
 
@@ -1629,64 +1651,6 @@ func (cli *DockerCli) CmdSearch(args ...string) error {
 // Ports type - Used to parse multiple -p flags
 type ports []int
 
-// AttachOpts stores arguments to 'docker run -a', eg. which streams to attach to
-type AttachOpts map[string]bool
-
-func NewAttachOpts() AttachOpts {
-	return make(AttachOpts)
-}
-
-func (opts AttachOpts) String() string {
-	// Cast to underlying map type to avoid infinite recursion
-	return fmt.Sprintf("%v", map[string]bool(opts))
-}
-
-func (opts AttachOpts) Set(val string) error {
-	if val != "stdin" && val != "stdout" && val != "stderr" {
-		return fmt.Errorf("Unsupported stream name: %s", val)
-	}
-	opts[val] = true
-	return nil
-}
-
-func (opts AttachOpts) Get(val string) bool {
-	if res, exists := opts[val]; exists {
-		return res
-	}
-	return false
-}
-
-// PathOpts stores a unique set of absolute paths
-type PathOpts map[string]struct{}
-
-func NewPathOpts() PathOpts {
-	return make(PathOpts)
-}
-
-func (opts PathOpts) String() string {
-	return fmt.Sprintf("%v", map[string]struct{}(opts))
-}
-
-func (opts PathOpts) Set(val string) error {
-	var containerPath string
-
-	splited := strings.SplitN(val, ":", 2)
-	if len(splited) == 1 {
-		containerPath = splited[0]
-		val = filepath.Clean(splited[0])
-	} else {
-		containerPath = splited[1]
-		val = fmt.Sprintf("%s:%s", splited[0], filepath.Clean(splited[1]))
-	}
-
-	if !filepath.IsAbs(containerPath) {
-		utils.Debugf("%s is not an absolute path", containerPath)
-		return fmt.Errorf("%s is not an absolute path", containerPath)
-	}
-	opts[val] = struct{}{}
-	return nil
-}
-
 func (cli *DockerCli) CmdTag(args ...string) error {
 	cmd := cli.Subcmd("tag", "[OPTIONS] IMAGE REPOSITORY[:TAG]", "Tag an image into a repository")
 	force := cmd.Bool("f", false, "Force")
@@ -1730,61 +1694,61 @@ func ParseRun(args []string, capabilities *Capabilities) (*Config, *HostConfig,
 }
 
 func parseRun(cmd *flag.FlagSet, args []string, capabilities *Capabilities) (*Config, *HostConfig, *flag.FlagSet, error) {
+	var (
+		// FIXME: use utils.ListOpts for attach and volumes?
+		flAttach  = NewListOpts(ValidateAttach)
+		flVolumes = NewListOpts(ValidatePath)
+		flLinks   = NewListOpts(ValidateLink)
+		flEnv     = NewListOpts(ValidateEnv)
 
-	flHostname := cmd.String("h", "", "Container host name")
-	flWorkingDir := cmd.String("w", "", "Working directory inside the container")
-	flUser := cmd.String("u", "", "Username or UID")
-	flDetach := cmd.Bool("d", false, "Detached mode: Run container in the background, print new container id")
-	flAttach := NewAttachOpts()
-	cmd.Var(flAttach, "a", "Attach to stdin, stdout or stderr.")
-	flStdin := cmd.Bool("i", false, "Keep stdin open even if not attached")
-	flTty := cmd.Bool("t", false, "Allocate a pseudo-tty")
-	flMemoryString := cmd.String("m", "", "Memory limit (format: <number><optional unit>, where unit = b, k, m or g)")
-	flContainerIDFile := cmd.String("cidfile", "", "Write the container ID to the file")
-	flNetwork := cmd.Bool("n", true, "Enable networking for this container")
-	flPrivileged := cmd.Bool("privileged", false, "Give extended privileges to this container")
-	flAutoRemove := cmd.Bool("rm", false, "Automatically remove the container when it exits (incompatible with -d)")
-	cmd.Bool("sig-proxy", true, "Proxify all received signal to the process (even in non-tty mode)")
-	cmd.String("name", "", "Assign a name to the container")
-	flPublishAll := cmd.Bool("P", false, "Publish all exposed ports to the host interfaces")
+		flPublish     ListOpts
+		flExpose      ListOpts
+		flDns         ListOpts
+		flVolumesFrom ListOpts
+		flLxcOpts     ListOpts
 
-	if capabilities != nil && *flMemoryString != "" && !capabilities.MemoryLimit {
-		//fmt.Fprintf(stdout, "WARNING: Your kernel does not support memory limit capabilities. Limitation discarded.\n")
-		*flMemoryString = ""
-	}
+		flAutoRemove      = cmd.Bool("rm", false, "Automatically remove the container when it exits (incompatible with -d)")
+		flDetach          = cmd.Bool("d", false, "Detached mode: Run container in the background, print new container id")
+		flNetwork         = cmd.Bool("n", true, "Enable networking for this container")
+		flPrivileged      = cmd.Bool("privileged", false, "Give extended privileges to this container")
+		flPublishAll      = cmd.Bool("P", false, "Publish all exposed ports to the host interfaces")
+		flStdin           = cmd.Bool("i", false, "Keep stdin open even if not attached")
+		flTty             = cmd.Bool("t", false, "Allocate a pseudo-tty")
+		flContainerIDFile = cmd.String("cidfile", "", "Write the container ID to the file")
+		flEntrypoint      = cmd.String("entrypoint", "", "Overwrite the default entrypoint of the image")
+		flHostname        = cmd.String("h", "", "Container host name")
+		flMemoryString    = cmd.String("m", "", "Memory limit (format: <number><optional unit>, where unit = b, k, m or g)")
+		flUser            = cmd.String("u", "", "Username or UID")
+		flWorkingDir      = cmd.String("w", "", "Working directory inside the container")
+		flCpuShares       = cmd.Int64("c", 0, "CPU shares (relative weight)")
 
-	flCpuShares := cmd.Int64("c", 0, "CPU shares (relative weight)")
+		// For documentation purpose
+		_ = cmd.Bool("sig-proxy", true, "Proxify all received signal to the process (even in non-tty mode)")
+		_ = cmd.String("name", "", "Assign a name to the container")
+	)
 
-	var flPublish utils.ListOpts
-	cmd.Var(&flPublish, "p", "Publish a container's port to the host (use 'docker port' to see the actual mapping)")
-
-	var flExpose utils.ListOpts
-	cmd.Var(&flExpose, "expose", "Expose a port from the container without publishing it to your host")
-
-	var flEnv utils.ListOpts
+	cmd.Var(&flAttach, "a", "Attach to stdin, stdout or stderr.")
+	cmd.Var(&flVolumes, "v", "Bind mount a volume (e.g. from the host: -v /host:/container, from docker: -v /container)")
+	cmd.Var(&flLinks, "link", "Add link to another container (name:alias)")
 	cmd.Var(&flEnv, "e", "Set environment variables")
 
-	var flDns utils.ListOpts
+	cmd.Var(&flPublish, "p", fmt.Sprintf("Publish a container's port to the host (format: %s) (use 'docker port' to see the actual mapping)", PortSpecTemplateFormat))
+	cmd.Var(&flExpose, "expose", "Expose a port from the container without publishing it to your host")
 	cmd.Var(&flDns, "dns", "Set custom dns servers")
-
-	flVolumes := NewPathOpts()
-	cmd.Var(flVolumes, "v", "Bind mount a volume (e.g. from the host: -v /host:/container, from docker: -v /container)")
-
-	var flVolumesFrom utils.ListOpts
 	cmd.Var(&flVolumesFrom, "volumes-from", "Mount volumes from the specified container(s)")
-
-	flEntrypoint := cmd.String("entrypoint", "", "Overwrite the default entrypoint of the image")
-
-	var flLxcOpts utils.ListOpts
 	cmd.Var(&flLxcOpts, "lxc-conf", "Add custom lxc options -lxc-conf=\"lxc.cgroup.cpuset.cpus = 0,1\"")
 
-	var flLinks utils.ListOpts
-	cmd.Var(&flLinks, "link", "Add link to another container (name:alias)")
-
 	if err := cmd.Parse(args); err != nil {
 		return nil, nil, cmd, err
 	}
-	if *flDetach && len(flAttach) > 0 {
+
+	// Check if the kernel supports memory limit cgroup.
+	if capabilities != nil && *flMemoryString != "" && !capabilities.MemoryLimit {
+		*flMemoryString = ""
+	}
+
+	// Validate input params
+	if *flDetach && flAttach.Len() > 0 {
 		return nil, nil, cmd, ErrConflictAttachDetach
 	}
 	if *flWorkingDir != "" && !path.IsAbs(*flWorkingDir) {
@@ -1795,7 +1759,7 @@ func parseRun(cmd *flag.FlagSet, args []string, capabilities *Capabilities) (*Co
 	}
 
 	// If neither -d or -a are set, attach to everything by default
-	if len(flAttach) == 0 && !*flDetach {
+	if flAttach.Len() == 0 && !*flDetach {
 		if !*flDetach {
 			flAttach.Set("stdout")
 			flAttach.Set("stderr")
@@ -1805,50 +1769,35 @@ func parseRun(cmd *flag.FlagSet, args []string, capabilities *Capabilities) (*Co
 		}
 	}
 
-	envs := []string{}
-
-	for _, env := range flEnv {
-		arr := strings.Split(env, "=")
-		if len(arr) > 1 {
-			envs = append(envs, env)
-		} else {
-			v := os.Getenv(env)
-			envs = append(envs, env+"="+v)
-		}
-	}
-
 	var flMemory int64
-
 	if *flMemoryString != "" {
 		parsedMemory, err := utils.RAMInBytes(*flMemoryString)
-
 		if err != nil {
 			return nil, nil, cmd, err
 		}
-
 		flMemory = parsedMemory
 	}
 
 	var binds []string
-
 	// add any bind targets to the list of container volumes
-	for bind := range flVolumes {
-		arr := strings.Split(bind, ":")
-		if len(arr) > 1 {
+	for bind := range flVolumes.GetMap() {
+		if arr := strings.Split(bind, ":"); len(arr) > 1 {
 			if arr[0] == "/" {
 				return nil, nil, cmd, fmt.Errorf("Invalid bind mount: source can't be '/'")
 			}
 			dstDir := arr[1]
-			flVolumes[dstDir] = struct{}{}
+			flVolumes.Set(dstDir)
 			binds = append(binds, bind)
-			delete(flVolumes, bind)
+			flVolumes.Delete(bind)
 		}
 	}
 
-	parsedArgs := cmd.Args()
-	runCmd := []string{}
-	entrypoint := []string{}
-	image := ""
+	var (
+		parsedArgs = cmd.Args()
+		runCmd     []string
+		entrypoint []string
+		image      string
+	)
 	if len(parsedArgs) >= 1 {
 		image = cmd.Arg(0)
 	}
@@ -1859,28 +1808,28 @@ func parseRun(cmd *flag.FlagSet, args []string, capabilities *Capabilities) (*Co
 		entrypoint = []string{*flEntrypoint}
 	}
 
-	var lxcConf []KeyValuePair
 	lxcConf, err := parseLxcConfOpts(flLxcOpts)
 	if err != nil {
 		return nil, nil, cmd, err
 	}
 
-	hostname := *flHostname
-	domainname := ""
-
-	parts := strings.SplitN(hostname, ".", 2)
+	var (
+		domainname string
+		hostname   = *flHostname
+		parts      = strings.SplitN(hostname, ".", 2)
+	)
 	if len(parts) > 1 {
 		hostname = parts[0]
 		domainname = parts[1]
 	}
 
-	ports, portBindings, err := parsePortSpecs(flPublish)
+	ports, portBindings, err := parsePortSpecs(flPublish.GetAll())
 	if err != nil {
 		return nil, nil, cmd, err
 	}
 
 	// Merge in exposed ports to the map of published ports
-	for _, e := range flExpose {
+	for _, e := range flExpose.GetAll() {
 		if strings.Contains(e, ":") {
 			return nil, nil, cmd, fmt.Errorf("Invalid port format for -expose: %s", e)
 		}
@@ -1904,12 +1853,12 @@ func parseRun(cmd *flag.FlagSet, args []string, capabilities *Capabilities) (*Co
 		AttachStdin:     flAttach.Get("stdin"),
 		AttachStdout:    flAttach.Get("stdout"),
 		AttachStderr:    flAttach.Get("stderr"),
-		Env:             envs,
+		Env:             flEnv.GetAll(),
 		Cmd:             runCmd,
-		Dns:             flDns,
+		Dns:             flDns.GetAll(),
 		Image:           image,
-		Volumes:         flVolumes,
-		VolumesFrom:     strings.Join(flVolumesFrom, ","),
+		Volumes:         flVolumes.GetMap(),
+		VolumesFrom:     strings.Join(flVolumesFrom.GetAll(), ","),
 		Entrypoint:      entrypoint,
 		WorkingDir:      *flWorkingDir,
 	}
@@ -1920,7 +1869,7 @@ func parseRun(cmd *flag.FlagSet, args []string, capabilities *Capabilities) (*Co
 		LxcConf:         lxcConf,
 		Privileged:      *flPrivileged,
 		PortBindings:    portBindings,
-		Links:           flLinks,
+		Links:           flLinks.GetAll(),
 		PublishAllPorts: *flPublishAll,
 	}
 
@@ -1946,30 +1895,33 @@ func (cli *DockerCli) CmdRun(args ...string) error {
 		return nil
 	}
 
-	flRm := cmd.Lookup("rm")
-	autoRemove, _ := strconv.ParseBool(flRm.Value.String())
+	// Retrieve relevant client-side config
+	var (
+		flName        = cmd.Lookup("name")
+		flRm          = cmd.Lookup("rm")
+		flSigProxy    = cmd.Lookup("sig-proxy")
+		autoRemove, _ = strconv.ParseBool(flRm.Value.String())
+		sigProxy, _   = strconv.ParseBool(flSigProxy.Value.String())
+	)
 
-	flSigProxy := cmd.Lookup("sig-proxy")
-	sigProxy, _ := strconv.ParseBool(flSigProxy.Value.String())
-	flName := cmd.Lookup("name")
+	// Disable sigProxy in case on TTY
 	if config.Tty {
 		sigProxy = false
 	}
 
-	var containerIDFile *os.File
+	var containerIDFile io.WriteCloser
 	if len(hostConfig.ContainerIDFile) > 0 {
-		if _, err := ioutil.ReadFile(hostConfig.ContainerIDFile); err == nil {
+		if _, err := os.Stat(hostConfig.ContainerIDFile); err == nil {
 			return fmt.Errorf("cid file found, make sure the other container isn't running or delete %s", hostConfig.ContainerIDFile)
 		}
-		containerIDFile, err = os.Create(hostConfig.ContainerIDFile)
-		if err != nil {
+		if containerIDFile, err = os.Create(hostConfig.ContainerIDFile); err != nil {
 			return fmt.Errorf("failed to create the container ID file: %s", err)
 		}
 		defer containerIDFile.Close()
 	}
+
 	containerValues := url.Values{}
-	name := flName.Value.String()
-	if name != "" {
+	if name := flName.Value.String(); name != "" {
 		containerValues.Set("name", name)
 	}
 
@@ -1990,8 +1942,7 @@ func (cli *DockerCli) CmdRun(args ...string) error {
 		v.Set("tag", tag)
 
 		// Resolve the Repository name from fqn to endpoint + name
-		var endpoint string
-		endpoint, _, err = registry.ResolveRepositoryName(repos)
+		endpoint, _, err := registry.ResolveRepositoryName(repos)
 		if err != nil {
 			return err
 		}
@@ -2009,32 +1960,27 @@ func (cli *DockerCli) CmdRun(args ...string) error {
 		registryAuthHeader := []string{
 			base64.URLEncoding.EncodeToString(buf),
 		}
-		err = cli.stream("POST", "/images/create?"+v.Encode(), nil, cli.err, map[string][]string{
-			"X-Registry-Auth": registryAuthHeader,
-		})
-		if err != nil {
+		if err = cli.stream("POST", "/images/create?"+v.Encode(), nil, cli.err, map[string][]string{"X-Registry-Auth": registryAuthHeader}); err != nil {
 			return err
 		}
-		body, _, err = cli.call("POST", "/containers/create?"+containerValues.Encode(), config)
-		if err != nil {
+		if body, _, err = cli.call("POST", "/containers/create?"+containerValues.Encode(), config); err != nil {
 			return err
 		}
-	}
-	if err != nil {
+	} else if err != nil {
 		return err
 	}
 
-	runResult := &APIRun{}
-	err = json.Unmarshal(body, runResult)
-	if err != nil {
+	var runResult APIRun
+	if err := json.Unmarshal(body, &runResult); err != nil {
 		return err
 	}
 
 	for _, warning := range runResult.Warnings {
 		fmt.Fprintf(cli.err, "WARNING: %s\n", warning)
 	}
+
 	if len(hostConfig.ContainerIDFile) > 0 {
-		if _, err = containerIDFile.WriteString(runResult.ID); err != nil {
+		if _, err = containerIDFile.Write([]byte(runResult.ID)); err != nil {
 			return fmt.Errorf("failed to write the container ID to the file: %s", err)
 		}
 	}
@@ -2045,27 +1991,38 @@ func (cli *DockerCli) CmdRun(args ...string) error {
 	}
 
 	var (
-		wait  chan struct{}
-		errCh chan error
+		waitDisplayId chan struct{}
+		errCh         chan error
 	)
 
 	if !config.AttachStdout && !config.AttachStderr {
 		// Make this asynchrone in order to let the client write to stdin before having to read the ID
-		wait = make(chan struct{})
+		waitDisplayId = make(chan struct{})
 		go func() {
-			defer close(wait)
+			defer close(waitDisplayId)
 			fmt.Fprintf(cli.out, "%s\n", runResult.ID)
 		}()
 	}
 
-	hijacked := make(chan bool)
+	// We need to instanciate the chan because the select needs it. It can
+	// be closed but can't be uninitialized.
+	hijacked := make(chan io.Closer)
+
+	// Block the return until the chan gets closed
+	defer func() {
+		utils.Debugf("End of CmdRun(), Waiting for hijack to finish.")
+		if _, ok := <-hijacked; ok {
+			utils.Errorf("Hijack did not finish (chan still open)")
+		}
+	}()
 
 	if config.AttachStdin || config.AttachStdout || config.AttachStderr {
-
-		v := url.Values{}
+		var (
+			out, stderr io.Writer
+			in          io.ReadCloser
+			v           = url.Values{}
+		)
 		v.Set("stream", "1")
-		var out, stderr io.Writer
-		var in io.ReadCloser
 
 		if config.AttachStdin {
 			v.Set("stdin", "1")
@@ -2093,7 +2050,12 @@ func (cli *DockerCli) CmdRun(args ...string) error {
 
 	// Acknowledge the hijack before starting
 	select {
-	case <-hijacked:
+	case closer := <-hijacked:
+		// Make sure that hijack gets closed when returning. (result
+		// in closing hijack chan and freeing server's goroutines.
+		if closer != nil {
+			defer closer.Close()
+		}
 	case err := <-errCh:
 		if err != nil {
 			utils.Debugf("Error hijack: %s", err)
@@ -2119,31 +2081,37 @@ func (cli *DockerCli) CmdRun(args ...string) error {
 		}
 	}
 
+	// Detached mode: wait for the id to be displayed and return.
 	if !config.AttachStdout && !config.AttachStderr {
 		// Detached mode
-		<-wait
-	} else {
-		running, status, err := getExitCode(cli, runResult.ID)
-		if err != nil {
-			return err
-		}
-		if autoRemove {
-			if running {
-				return fmt.Errorf("Impossible to auto-remove a detached container")
-			}
-			// Wait for the process to
-			if _, _, err := cli.call("POST", "/containers/"+runResult.ID+"/wait", nil); err != nil {
-				return err
-			}
-			if _, _, err := cli.call("DELETE", "/containers/"+runResult.ID, nil); err != nil {
-				return err
-			}
-		}
-		if status != 0 {
-			return &utils.StatusError{Status: status}
-		}
+		<-waitDisplayId
+		return nil
 	}
 
+	var status int
+
+	// Attached mode
+	if autoRemove {
+		// Autoremove: wait for the container to finish, retrieve
+		// the exit code and remove the container
+		if _, _, err := cli.call("POST", "/containers/"+runResult.ID+"/wait", nil); err != nil {
+			return err
+		}
+		if _, status, err = getExitCode(cli, runResult.ID); err != nil {
+			return err
+		}
+		if _, _, err := cli.call("DELETE", "/containers/"+runResult.ID, nil); err != nil {
+			return err
+		}
+	} else {
+		// No Autoremove: Simply retrieve the exit code
+		if _, status, err = getExitCode(cli, runResult.ID); err != nil {
+			return err
+		}
+	}
+	if status != 0 {
+		return &utils.StatusError{StatusCode: status}
+	}
 	return nil
 }
 
@@ -2175,7 +2143,7 @@ func (cli *DockerCli) CmdCp(args ...string) error {
 
 	if statusCode == 200 {
 		r := bytes.NewReader(data)
-		if err := archive.Untar(r, copyData.HostPath); err != nil {
+		if err := archive.Untar(r, copyData.HostPath, nil); err != nil {
 			return err
 		}
 	}
@@ -2183,7 +2151,7 @@ func (cli *DockerCli) CmdCp(args ...string) error {
 }
 
 func (cli *DockerCli) CmdSave(args ...string) error {
-	cmd := cli.Subcmd("save", "IMAGE DESTINATION", "Save an image to a tar archive")
+	cmd := cli.Subcmd("save", "IMAGE", "Save an image to a tar archive (streamed to stdout)")
 	if err := cmd.Parse(args); err != nil {
 		return err
 	}
@@ -2328,7 +2296,7 @@ func (cli *DockerCli) stream(method, path string, in io.Reader, out io.Writer, h
 	}
 
 	if matchesContentType(resp.Header.Get("Content-Type"), "application/json") {
-		return utils.DisplayJSONMessagesStream(resp.Body, out)
+		return utils.DisplayJSONMessagesStream(resp.Body, out, cli.terminalFd, cli.isTerminal)
 	}
 	if _, err := io.Copy(out, resp.Body); err != nil {
 		return err
@@ -2336,7 +2304,12 @@ func (cli *DockerCli) stream(method, path string, in io.Reader, out io.Writer, h
 	return nil
 }
 
-func (cli *DockerCli) hijack(method, path string, setRawTerminal bool, in io.ReadCloser, stdout, stderr io.Writer, started chan bool) error {
+func (cli *DockerCli) hijack(method, path string, setRawTerminal bool, in io.ReadCloser, stdout, stderr io.Writer, started chan io.Closer) error {
+	defer func() {
+		if started != nil {
+			close(started)
+		}
+	}()
 	// fixme: refactor client to support redirect
 	re := regexp.MustCompile("/+")
 	path = re.ReplaceAllString(path, "/")
@@ -2366,13 +2339,32 @@ func (cli *DockerCli) hijack(method, path string, setRawTerminal bool, in io.Rea
 	defer rwc.Close()
 
 	if started != nil {
-		started <- true
+		started <- rwc
 	}
 
 	var receiveStdout chan error
 
-	if stdout != nil {
+	var oldState *term.State
+
+	if in != nil && setRawTerminal && cli.isTerminal && os.Getenv("NORAW") == "" {
+		oldState, err = term.SetRawTerminal(cli.terminalFd)
+		if err != nil {
+			return err
+		}
+		defer term.RestoreTerminal(cli.terminalFd, oldState)
+	}
+
+	if stdout != nil || stderr != nil {
 		receiveStdout = utils.Go(func() (err error) {
+			defer func() {
+				if in != nil {
+					if setRawTerminal && cli.isTerminal {
+						term.RestoreTerminal(cli.terminalFd, oldState)
+					}
+					in.Close()
+				}
+			}()
+
 			// When TTY is ON, use regular copy
 			if setRawTerminal {
 				_, err = io.Copy(stdout, br)
@@ -2384,14 +2376,6 @@ func (cli *DockerCli) hijack(method, path string, setRawTerminal bool, in io.Rea
 		})
 	}
 
-	if in != nil && setRawTerminal && cli.isTerminal && os.Getenv("NORAW") == "" {
-		oldState, err := term.SetRawTerminal(cli.terminalFd)
-		if err != nil {
-			return err
-		}
-		defer term.RestoreTerminal(cli.terminalFd, oldState)
-	}
-
 	sendStdin := utils.Go(func() error {
 		if in != nil {
 			io.Copy(rwc, in)
@@ -2410,7 +2394,7 @@ func (cli *DockerCli) hijack(method, path string, setRawTerminal bool, in io.Rea
 		return nil
 	})
 
-	if stdout != nil {
+	if stdout != nil || stderr != nil {
 		if err := <-receiveStdout; err != nil {
 			utils.Errorf("Error receiveStdout: %s", err)
 			return err

commands_unit_test.go

@@ -0,0 +1,157 @@
+package docker
+
+import (
+	"strings"
+	"testing"
+)
+
+func parse(t *testing.T, args string) (*Config, *HostConfig, error) {
+	config, hostConfig, _, err := ParseRun(strings.Split(args+" ubuntu bash", " "), nil)
+	return config, hostConfig, err
+}
+
+func mustParse(t *testing.T, args string) (*Config, *HostConfig) {
+	config, hostConfig, err := parse(t, args)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return config, hostConfig
+}
+
+func TestParseRunLinks(t *testing.T) {
+	if _, hostConfig := mustParse(t, "-link a:b"); len(hostConfig.Links) == 0 || hostConfig.Links[0] != "a:b" {
+		t.Fatalf("Error parsing links. Expected []string{\"a:b\"}, received: %v", hostConfig.Links)
+	}
+	if _, hostConfig := mustParse(t, "-link a:b -link c:d"); len(hostConfig.Links) < 2 || hostConfig.Links[0] != "a:b" || hostConfig.Links[1] != "c:d" {
+		t.Fatalf("Error parsing links. Expected []string{\"a:b\", \"c:d\"}, received: %v", hostConfig.Links)
+	}
+	if _, hostConfig := mustParse(t, ""); len(hostConfig.Links) != 0 {
+		t.Fatalf("Error parsing links. No link expected, received: %v", hostConfig.Links)
+	}
+
+	if _, _, err := parse(t, "-link a"); err == nil {
+		t.Fatalf("Error parsing links. `-link a` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-link"); err == nil {
+		t.Fatalf("Error parsing links. `-link` should be an error but is not")
+	}
+}
+
+func TestParseRunAttach(t *testing.T) {
+	if config, _ := mustParse(t, "-a stdin"); !config.AttachStdin || config.AttachStdout || config.AttachStderr {
+		t.Fatalf("Error parsing attach flags. Expect only Stdin enabled. Received: in: %v, out: %v, err: %v", config.AttachStdin, config.AttachStdout, config.AttachStderr)
+	}
+	if config, _ := mustParse(t, "-a stdin -a stdout"); !config.AttachStdin || !config.AttachStdout || config.AttachStderr {
+		t.Fatalf("Error parsing attach flags. Expect only Stdin and Stdout enabled. Received: in: %v, out: %v, err: %v", config.AttachStdin, config.AttachStdout, config.AttachStderr)
+	}
+	if config, _ := mustParse(t, "-a stdin -a stdout -a stderr"); !config.AttachStdin || !config.AttachStdout || !config.AttachStderr {
+		t.Fatalf("Error parsing attach flags. Expect all attach enabled. Received: in: %v, out: %v, err: %v", config.AttachStdin, config.AttachStdout, config.AttachStderr)
+	}
+	if config, _ := mustParse(t, ""); config.AttachStdin || !config.AttachStdout || !config.AttachStderr {
+		t.Fatalf("Error parsing attach flags. Expect Stdin disabled. Received: in: %v, out: %v, err: %v", config.AttachStdin, config.AttachStdout, config.AttachStderr)
+	}
+
+	if _, _, err := parse(t, "-a"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-a invalid"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a invalid` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-a invalid -a stdout"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a stdout -a invalid` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-a stdout -a stderr -d"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a stdout -a stderr -d` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-a stdin -d"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a stdin -d` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-a stdout -d"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a stdout -d` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-a stderr -d"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-a stderr -d` should be an error but is not")
+	}
+	if _, _, err := parse(t, "-d -rm"); err == nil {
+		t.Fatalf("Error parsing attach flags, `-d -rm` should be an error but is not")
+	}
+}
+
+func TestParseRunVolumes(t *testing.T) {
+	if config, hostConfig := mustParse(t, "-v /tmp"); hostConfig.Binds != nil {
+		t.Fatalf("Error parsing volume flags, `-v /tmp` should not mount-bind anything. Received %v", hostConfig.Binds)
+	} else if _, exists := config.Volumes["/tmp"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /tmp` is missing from volumes. Received %v", config.Volumes)
+	}
+
+	if config, hostConfig := mustParse(t, "-v /tmp -v /var"); hostConfig.Binds != nil {
+		t.Fatalf("Error parsing volume flags, `-v /tmp -v /var` should not mount-bind anything. Received %v", hostConfig.Binds)
+	} else if _, exists := config.Volumes["/tmp"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /tmp` is missing from volumes. Recevied %v", config.Volumes)
+	} else if _, exists := config.Volumes["/var"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /var` is missing from volumes. Received %v", config.Volumes)
+	}
+
+	if config, hostConfig := mustParse(t, "-v /hostTmp:/containerTmp"); hostConfig.Binds == nil || hostConfig.Binds[0] != "/hostTmp:/containerTmp" {
+		t.Fatalf("Error parsing volume flags, `-v /hostTmp:/containerTmp` should mount-bind /hostTmp into /containeTmp. Received %v", hostConfig.Binds)
+	} else if _, exists := config.Volumes["/containerTmp"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /tmp` is missing from volumes. Received %v", config.Volumes)
+	}
+
+	if config, hostConfig := mustParse(t, "-v /hostTmp:/containerTmp -v /hostVar:/containerVar"); hostConfig.Binds == nil || hostConfig.Binds[0] != "/hostTmp:/containerTmp" || hostConfig.Binds[1] != "/hostVar:/containerVar" {
+		t.Fatalf("Error parsing volume flags, `-v /hostTmp:/containerTmp -v /hostVar:/containerVar` should mount-bind /hostTmp into /containeTmp and /hostVar into /hostContainer. Received %v", hostConfig.Binds)
+	} else if _, exists := config.Volumes["/containerTmp"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /containerTmp` is missing from volumes. Received %v", config.Volumes)
+	} else if _, exists := config.Volumes["/containerVar"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /containerVar` is missing from volumes. Received %v", config.Volumes)
+	}
+
+	if config, hostConfig := mustParse(t, "-v /hostTmp:/containerTmp:ro -v /hostVar:/containerVar:rw"); hostConfig.Binds == nil || hostConfig.Binds[0] != "/hostTmp:/containerTmp:ro" || hostConfig.Binds[1] != "/hostVar:/containerVar:rw" {
+		t.Fatalf("Error parsing volume flags, `-v /hostTmp:/containerTmp:ro -v /hostVar:/containerVar:rw` should mount-bind /hostTmp into /containeTmp and /hostVar into /hostContainer. Received %v", hostConfig.Binds)
+	} else if _, exists := config.Volumes["/containerTmp"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /containerTmp` is missing from volumes. Received %v", config.Volumes)
+	} else if _, exists := config.Volumes["/containerVar"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /containerVar` is missing from volumes. Received %v", config.Volumes)
+	}
+
+	if config, hostConfig := mustParse(t, "-v /hostTmp:/containerTmp -v /containerVar"); hostConfig.Binds == nil || len(hostConfig.Binds) > 1 || hostConfig.Binds[0] != "/hostTmp:/containerTmp" {
+		t.Fatalf("Error parsing volume flags, `-v /hostTmp:/containerTmp -v /containerVar` should mount-bind only /hostTmp into /containeTmp. Received %v", hostConfig.Binds)
+	} else if _, exists := config.Volumes["/containerTmp"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /containerTmp` is missing from volumes. Received %v", config.Volumes)
+	} else if _, exists := config.Volumes["/containerVar"]; !exists {
+		t.Fatalf("Error parsing volume flags, `-v /containerVar` is missing from volumes. Received %v", config.Volumes)
+	}
+
+	if config, hostConfig := mustParse(t, ""); hostConfig.Binds != nil {
+		t.Fatalf("Error parsing volume flags, without volume, nothing should be mount-binded. Received %v", hostConfig.Binds)
+	} else if len(config.Volumes) != 0 {
+		t.Fatalf("Error parsing volume flags, without volume, no volume should be present. Received %v", config.Volumes)
+	}
+
+	mustParse(t, "-v /")
+
+	if _, _, err := parse(t, "-v /:/"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v /:/` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v /tmp:"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v /tmp:` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v /tmp:ro"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v /tmp:ro` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v /tmp::"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v /tmp::` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v :"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v :` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v ::"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v ::` should fail but didn't")
+	}
+	if _, _, err := parse(t, "-v /tmp:/tmp:/tmp:/tmp"); err == nil {
+		t.Fatalf("Error parsing volume flags, `-v /tmp:/tmp:/tmp:/tmp` should fail but didn't")
+	}
+}

config.go

@@ -16,6 +16,7 @@ type DaemonConfig struct {
 	BridgeIface                 string
 	DefaultIp                   net.IP
 	InterContainerCommunication bool
+	GraphDriver                 string
 }
 
 // ConfigFromJob creates and returns a new DaemonConfig object
@@ -26,8 +27,8 @@ func ConfigFromJob(job *engine.Job) *DaemonConfig {
 	config.Root = job.Getenv("Root")
 	config.AutoRestart = job.GetenvBool("AutoRestart")
 	config.EnableCors = job.GetenvBool("EnableCors")
-	if dns := job.Getenv("Dns"); dns != "" {
-		config.Dns = []string{dns}
+	if dns := job.GetenvList("Dns"); dns != nil {
+		config.Dns = dns
 	}
 	config.EnableIptables = job.GetenvBool("EnableIptables")
 	if br := job.Getenv("BridgeIface"); br != "" {
@@ -37,5 +38,6 @@ func ConfigFromJob(job *engine.Job) *DaemonConfig {
 	}
 	config.DefaultIp = net.ParseIP(job.Getenv("DefaultIp"))
 	config.InterContainerCommunication = job.GetenvBool("InterContainerCommunication")
+	config.GraphDriver = job.Getenv("GraphDriver")
 	return &config
 }

container.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"fmt"
 	"github.com/dotcloud/docker/archive"
+	"github.com/dotcloud/docker/graphdriver"
 	"github.com/dotcloud/docker/term"
 	"github.com/dotcloud/docker/utils"
 	"github.com/kr/pty"
@@ -16,7 +17,6 @@ import (
 	"os"
 	"os/exec"
 	"path"
-	"path/filepath"
 	"strconv"
 	"strings"
 	"sync"
@@ -24,10 +24,15 @@ import (
 	"time"
 )
 
+var (
+	ErrNotATTY = errors.New("The PTY is not a file")
+	ErrNoTTY   = errors.New("No PTY found")
+)
+
 type Container struct {
 	sync.Mutex
-
-	root string
+	root   string // Path to the "home" of the container, including metadata.
+	rootfs string // Path to the root filesystem of the container.
 
 	ID string
 
@@ -48,6 +53,7 @@ type Container struct {
 	HostnamePath   string
 	HostsPath      string
 	Name           string
+	Driver         string
 
 	cmd       *exec.Cmd
 	stdout    *utils.WriteBroadcaster
@@ -196,8 +202,13 @@ func (settings *NetworkSettings) PortMappingAPI() []APIPort {
 
 // Inject the io.Reader at the given path. Note: do not close the reader
 func (container *Container) Inject(file io.Reader, pth string) error {
+	if err := container.EnsureMounted(); err != nil {
+		return fmt.Errorf("inject: error mounting container %s: %s", container.ID, err)
+	}
+
 	// Return error if path exists
-	if _, err := os.Stat(path.Join(container.rwPath(), pth)); err == nil {
+	destPath := path.Join(container.RootfsPath(), pth)
+	if _, err := os.Stat(destPath); err == nil {
 		// Since err is nil, the path could be stat'd and it exists
 		return fmt.Errorf("%s exists", pth)
 	} else if !os.IsNotExist(err) {
@@ -208,14 +219,16 @@ func (container *Container) Inject(file io.Reader, pth string) error {
 	}
 
 	// Make sure the directory exists
-	if err := os.MkdirAll(path.Join(container.rwPath(), path.Dir(pth)), 0755); err != nil {
+	if err := os.MkdirAll(path.Join(container.RootfsPath(), path.Dir(pth)), 0755); err != nil {
 		return err
 	}
 
-	dest, err := os.Create(path.Join(container.rwPath(), pth))
+	dest, err := os.Create(destPath)
 	if err != nil {
 		return err
 	}
+	defer dest.Close()
+
 	if _, err := io.Copy(dest, file); err != nil {
 		return err
 	}
@@ -528,157 +541,18 @@ func (container *Container) Start() (err error) {
 		log.Printf("WARNING: IPv4 forwarding is disabled. Networking will not work")
 	}
 
-	// Create the requested bind mounts
-	binds := make(map[string]BindMap)
-	// Define illegal container destinations
-	illegalDsts := []string{"/", "."}
-
-	for _, bind := range container.hostConfig.Binds {
-		// FIXME: factorize bind parsing in parseBind
-		var src, dst, mode string
-		arr := strings.Split(bind, ":")
-		if len(arr) == 2 {
-			src = arr[0]
-			dst = arr[1]
-			mode = "rw"
-		} else if len(arr) == 3 {
-			src = arr[0]
-			dst = arr[1]
-			mode = arr[2]
-		} else {
-			return fmt.Errorf("Invalid bind specification: %s", bind)
-		}
-
-		// Bail if trying to mount to an illegal destination
-		for _, illegal := range illegalDsts {
-			if dst == illegal {
-				return fmt.Errorf("Illegal bind destination: %s", dst)
-			}
-		}
-
-		bindMap := BindMap{
-			SrcPath: src,
-			DstPath: dst,
-			Mode:    mode,
-		}
-		binds[path.Clean(dst)] = bindMap
-	}
-
 	if container.Volumes == nil || len(container.Volumes) == 0 {
 		container.Volumes = make(map[string]string)
 		container.VolumesRW = make(map[string]bool)
 	}
 
 	// Apply volumes from another container if requested
-	if container.Config.VolumesFrom != "" {
-		containerSpecs := strings.Split(container.Config.VolumesFrom, ",")
-		for _, containerSpec := range containerSpecs {
-			mountRW := true
-			specParts := strings.SplitN(containerSpec, ":", 2)
-			switch len(specParts) {
-			case 0:
-				return fmt.Errorf("Malformed volumes-from specification: %s", container.Config.VolumesFrom)
-			case 2:
-				switch specParts[1] {
-				case "ro":
-					mountRW = false
-				case "rw": // mountRW is already true
-				default:
-					return fmt.Errorf("Malformed volumes-from speficication: %s", containerSpec)
-				}
-			}
-			c := container.runtime.Get(specParts[0])
-			if c == nil {
-				return fmt.Errorf("Container %s not found. Impossible to mount its volumes", container.ID)
-			}
-			for volPath, id := range c.Volumes {
-				if _, exists := container.Volumes[volPath]; exists {
-					continue
-				}
-				if err := os.MkdirAll(path.Join(container.RootfsPath(), volPath), 0755); err != nil {
-					return err
-				}
-				container.Volumes[volPath] = id
-				if isRW, exists := c.VolumesRW[volPath]; exists {
-					container.VolumesRW[volPath] = isRW && mountRW
-				}
-			}
-
-		}
+	if err := container.applyExternalVolumes(); err != nil {
+		return err
 	}
 
-	// Create the requested volumes if they don't exist
-	for volPath := range container.Config.Volumes {
-		volPath = path.Clean(volPath)
-		// Skip existing volumes
-		if _, exists := container.Volumes[volPath]; exists {
-			continue
-		}
-		var srcPath string
-		var isBindMount bool
-		srcRW := false
-		// If an external bind is defined for this volume, use that as a source
-		if bindMap, exists := binds[volPath]; exists {
-			isBindMount = true
-			srcPath = bindMap.SrcPath
-			if strings.ToLower(bindMap.Mode) == "rw" {
-				srcRW = true
-			}
-			// Otherwise create an directory in $ROOT/volumes/ and use that
-		} else {
-			c, err := container.runtime.volumes.Create(nil, container, "", "", nil)
-			if err != nil {
-				return err
-			}
-			srcPath, err = c.layer()
-			if err != nil {
-				return err
-			}
-			srcRW = true // RW by default
-		}
-		container.Volumes[volPath] = srcPath
-		container.VolumesRW[volPath] = srcRW
-		// Create the mountpoint
-		rootVolPath := path.Join(container.RootfsPath(), volPath)
-		if err := os.MkdirAll(rootVolPath, 0755); err != nil {
-			return err
-		}
-
-		// Do not copy or change permissions if we are mounting from the host
-		if srcRW && !isBindMount {
-			volList, err := ioutil.ReadDir(rootVolPath)
-			if err != nil {
-				return err
-			}
-			if len(volList) > 0 {
-				srcList, err := ioutil.ReadDir(srcPath)
-				if err != nil {
-					return err
-				}
-				if len(srcList) == 0 {
-					// If the source volume is empty copy files from the root into the volume
-					if err := archive.CopyWithTar(rootVolPath, srcPath); err != nil {
-						return err
-					}
-
-					var stat syscall.Stat_t
-					if err := syscall.Stat(rootVolPath, &stat); err != nil {
-						return err
-					}
-					var srcStat syscall.Stat_t
-					if err := syscall.Stat(srcPath, &srcStat); err != nil {
-						return err
-					}
-					// Change the source volume's ownership if it differs from the root
-					// files that where just copied
-					if stat.Uid != srcStat.Uid || stat.Gid != srcStat.Gid {
-						if err := os.Chown(srcPath, int(stat.Uid), int(stat.Gid)); err != nil {
-							return err
-						}
-					}
-				}
-			}
-		}
+	if err := container.createVolumes(); err != nil {
+		return err
 	}
 
 	if err := container.generateLXCConfig(); err != nil {
@@ -864,6 +738,204 @@ func (container *Container) Start() (err error) {
 	return ErrContainerStart
 }
 
+func (container *Container) getBindMap() (map[string]BindMap, error) {
+	// Create the requested bind mounts
+	binds := make(map[string]BindMap)
+	// Define illegal container destinations
+	illegalDsts := []string{"/", "."}
+
+	for _, bind := range container.hostConfig.Binds {
+		// FIXME: factorize bind parsing in parseBind
+		var src, dst, mode string
+		arr := strings.Split(bind, ":")
+		if len(arr) == 2 {
+			src = arr[0]
+			dst = arr[1]
+			mode = "rw"
+		} else if len(arr) == 3 {
+			src = arr[0]
+			dst = arr[1]
+			mode = arr[2]
+		} else {
+			return nil, fmt.Errorf("Invalid bind specification: %s", bind)
+		}
+
+		// Bail if trying to mount to an illegal destination
+		for _, illegal := range illegalDsts {
+			if dst == illegal {
+				return nil, fmt.Errorf("Illegal bind destination: %s", dst)
+			}
+		}
+
+		bindMap := BindMap{
+			SrcPath: src,
+			DstPath: dst,
+			Mode:    mode,
+		}
+		binds[path.Clean(dst)] = bindMap
+	}
+  return binds, nil
+}
+
+func (container *Container) createVolumes() error {
+  binds, err := container.getBindMap()
+  if err != nil {
+    return err
+  }
+	volumesDriver := container.runtime.volumes.driver
+	// Create the requested volumes if they don't exist
+	for volPath := range container.Config.Volumes {
+		volPath = path.Clean(volPath)
+		volIsDir := true
+		// Skip existing volumes
+		if _, exists := container.Volumes[volPath]; exists {
+			continue
+		}
+		var srcPath string
+		var isBindMount bool
+		srcRW := false
+		// If an external bind is defined for this volume, use that as a source
+		if bindMap, exists := binds[volPath]; exists {
+			isBindMount = true
+			srcPath = bindMap.SrcPath
+			if strings.ToLower(bindMap.Mode) == "rw" {
+				srcRW = true
+			}
+			if file, err := os.Open(bindMap.SrcPath); err != nil {
+				return err
+			} else {
+				defer file.Close()
+				if stat, err := file.Stat(); err != nil {
+					return err
+				} else {
+					volIsDir = stat.IsDir()
+				}
+			}
+			// Otherwise create an directory in $ROOT/volumes/ and use that
+		} else {
+
+			// Do not pass a container as the parameter for the volume creation.
+			// The graph driver using the container's information ( Image ) to
+			// create the parent.
+			c, err := container.runtime.volumes.Create(nil, nil, "", "", nil)
+			if err != nil {
+				return err
+			}
+			srcPath, err = volumesDriver.Get(c.ID)
+			if err != nil {
+				return fmt.Errorf("Driver %s failed to get volume rootfs %s: %s", volumesDriver, c.ID, err)
+			}
+			srcRW = true // RW by default
+		}
+		container.Volumes[volPath] = srcPath
+		container.VolumesRW[volPath] = srcRW
+		// Create the mountpoint
+		rootVolPath := path.Join(container.RootfsPath(), volPath)
+		if volIsDir {
+			if err := os.MkdirAll(rootVolPath, 0755); err != nil {
+				return err
+			}
+		}
+
+		volPath = path.Join(container.RootfsPath(), volPath)
+		if _, err := os.Stat(volPath); err != nil {
+			if os.IsNotExist(err) {
+				if volIsDir {
+					if err := os.MkdirAll(volPath, 0755); err != nil {
+						return err
+					}
+				} else {
+					if err := os.MkdirAll(path.Dir(volPath), 0755); err != nil {
+						return err
+					}
+					if f, err := os.OpenFile(volPath, os.O_CREATE, 0755); err != nil {
+						return err
+					} else {
+						f.Close()
+					}
+				}
+			}
+		}
+
+		// Do not copy or change permissions if we are mounting from the host
+		if srcRW && !isBindMount {
+			volList, err := ioutil.ReadDir(rootVolPath)
+			if err != nil {
+				return err
+			}
+			if len(volList) > 0 {
+				srcList, err := ioutil.ReadDir(srcPath)
+				if err != nil {
+					return err
+				}
+				if len(srcList) == 0 {
+					// If the source volume is empty copy files from the root into the volume
+					if err := archive.CopyWithTar(rootVolPath, srcPath); err != nil {
+						return err
+					}
+
+					var stat syscall.Stat_t
+					if err := syscall.Stat(rootVolPath, &stat); err != nil {
+						return err
+					}
+					var srcStat syscall.Stat_t
+					if err := syscall.Stat(srcPath, &srcStat); err != nil {
+						return err
+					}
+					// Change the source volume's ownership if it differs from the root
+					// files that where just copied
+					if stat.Uid != srcStat.Uid || stat.Gid != srcStat.Gid {
+						if err := os.Chown(srcPath, int(stat.Uid), int(stat.Gid)); err != nil {
+							return err
+						}
+					}
+				}
+			}
+		}
+	}
+	return nil
+}
+
+func (container *Container) applyExternalVolumes() error {
+	if container.Config.VolumesFrom != "" {
+		containerSpecs := strings.Split(container.Config.VolumesFrom, ",")
+		for _, containerSpec := range containerSpecs {
+			mountRW := true
+			specParts := strings.SplitN(containerSpec, ":", 2)
+			switch len(specParts) {
+			case 0:
+				return fmt.Errorf("Malformed volumes-from specification: %s", container.Config.VolumesFrom)
+			case 2:
+				switch specParts[1] {
+				case "ro":
+					mountRW = false
+				case "rw": // mountRW is already true
+				default:
+					return fmt.Errorf("Malformed volumes-from speficication: %s", containerSpec)
+				}
+			}
+			c := container.runtime.Get(specParts[0])
+			if c == nil {
+				return fmt.Errorf("Container %s not found. Impossible to mount its volumes", container.ID)
+			}
+			for volPath, id := range c.Volumes {
+				if _, exists := container.Volumes[volPath]; exists {
+					continue
+				}
+				if err := os.MkdirAll(path.Join(container.RootfsPath(), volPath), 0755); err != nil {
+					return err
+				}
+				container.Volumes[volPath] = id
+				if isRW, exists := c.VolumesRW[volPath]; exists {
+					container.VolumesRW[volPath] = isRW && mountRW
+				}
+			}
+
+		}
+	}
+	return nil
+}
+
 func (container *Container) Run() error {
 	if err := container.Start(); err != nil {
 		return err
@@ -1231,15 +1303,14 @@ func (container *Container) Resize(h, w int) error {
 }
 
 func (container *Container) ExportRw() (archive.Archive, error) {
-	return archive.Tar(container.rwPath(), archive.Uncompressed)
-}
-
-func (container *Container) RwChecksum() (string, error) {
-	rwData, err := archive.Tar(container.rwPath(), archive.Xz)
-	if err != nil {
-		return "", err
+	if err := container.EnsureMounted(); err != nil {
+		return nil, err
 	}
-	return utils.HashData(rwData)
+	if container.runtime == nil {
+		return nil, fmt.Errorf("Can't load storage driver for unregistered container %s", container.ID)
+	}
+
+	return container.runtime.Diff(container)
 }
 
 func (container *Container) Export() (archive.Archive, error) {
@@ -1265,28 +1336,17 @@ func (container *Container) WaitTimeout(timeout time.Duration) error {
 }
 
 func (container *Container) EnsureMounted() error {
-	if mounted, err := container.Mounted(); err != nil {
-		return err
-	} else if mounted {
-		return nil
-	}
+	// FIXME: EnsureMounted is deprecated because drivers are now responsible
+	// for re-entrant mounting in their Get() method.
 	return container.Mount()
 }
 
 func (container *Container) Mount() error {
-	image, err := container.GetImage()
-	if err != nil {
-		return err
-	}
-	return image.Mount(container.RootfsPath(), container.rwPath())
+	return container.runtime.Mount(container)
 }
 
-func (container *Container) Changes() ([]Change, error) {
-	image, err := container.GetImage()
-	if err != nil {
-		return nil, err
-	}
-	return image.Changes(container.rwPath())
+func (container *Container) Changes() ([]archive.Change, error) {
+	return container.runtime.Changes(container)
 }
 
 func (container *Container) GetImage() (*Image, error) {
@@ -1296,18 +1356,8 @@ func (container *Container) GetImage() (*Image, error) {
 	return container.runtime.graph.Get(container.Image)
 }
 
-func (container *Container) Mounted() (bool, error) {
-	return Mounted(container.RootfsPath())
-}
-
 func (container *Container) Unmount() error {
-	if _, err := os.Stat(container.RootfsPath()); err != nil {
-		if os.IsNotExist(err) {
-			return nil
-		}
-		return err
-	}
-	return Unmount(container.RootfsPath())
+	return container.runtime.Unmount(container)
 }
 
 func (container *Container) logPath(name string) string {
@@ -1336,11 +1386,7 @@ func (container *Container) lxcConfigPath() string {
 
 // This method must be exported to be used from the lxc template
 func (container *Container) RootfsPath() string {
-	return path.Join(container.root, "rootfs")
-}
-
-func (container *Container) rwPath() string {
-	return path.Join(container.root, "rw")
+	return container.rootfs
 }
 
 func validateID(id string) error {
@@ -1352,49 +1398,38 @@ func validateID(id string) error {
 
 // GetSize, return real size, virtual size
 func (container *Container) GetSize() (int64, int64) {
-	var sizeRw, sizeRootfs int64
-	data := make(map[uint64]bool)
+	var (
+		sizeRw, sizeRootfs int64
+		err                error
+		driver             = container.runtime.driver
+	)
 
-	filepath.Walk(container.rwPath(), func(path string, fileInfo os.FileInfo, err error) error {
-		if fileInfo == nil {
-			return nil
+	if err := container.EnsureMounted(); err != nil {
+		utils.Errorf("Warning: failed to compute size of container rootfs %s: %s", container.ID, err)
+		return sizeRw, sizeRootfs
+	}
+
+	if differ, ok := container.runtime.driver.(graphdriver.Differ); ok {
+		sizeRw, err = differ.DiffSize(container.ID)
+		if err != nil {
+			utils.Errorf("Warning: driver %s couldn't return diff size of container %s: %s", driver, container.ID, err)
+			// FIXME: GetSize should return an error. Not changing it now in case
+			// there is a side-effect.
+			sizeRw = -1
 		}
-		size := fileInfo.Size()
-		if size == 0 {
-			return nil
+	} else {
+		changes, _ := container.Changes()
+		if changes != nil {
+			sizeRw = archive.ChangesSize(container.RootfsPath(), changes)
+		} else {
+			sizeRw = -1
 		}
+	}
 
-		inode := fileInfo.Sys().(*syscall.Stat_t).Ino
-		if _, entryExists := data[inode]; entryExists {
-			return nil
+	if _, err = os.Stat(container.RootfsPath()); err != nil {
+		if sizeRootfs, err = utils.TreeSize(container.RootfsPath()); err != nil {
+			sizeRootfs = -1
 		}
-		data[inode] = false
-
-		sizeRw += size
-		return nil
-	})
-
-	data = make(map[uint64]bool)
-	_, err := os.Stat(container.RootfsPath())
-	if err == nil {
-		filepath.Walk(container.RootfsPath(), func(path string, fileInfo os.FileInfo, err error) error {
-			if fileInfo == nil {
-				return nil
-			}
-			size := fileInfo.Size()
-			if size == 0 {
-				return nil
-			}
-
-			inode := fileInfo.Sys().(*syscall.Stat_t).Ino
-			if _, entryExists := data[inode]; entryExists {
-				return nil
-			}
-			data[inode] = false
-
-			sizeRootfs += size
-			return nil
-		})
 	}
 	return sizeRw, sizeRootfs
 }
@@ -1417,7 +1452,11 @@ func (container *Container) Copy(resource string) (archive.Archive, error) {
 		filter = []string{path.Base(basePath)}
 		basePath = path.Dir(basePath)
 	}
-	return archive.TarFilter(basePath, archive.Uncompressed, filter)
+	return archive.TarFilter(basePath, &archive.TarOptions{
+		Compression: archive.Uncompressed,
+		Includes:    filter,
+		Recursive:   true,
+	})
 }
 
 // Returns true if the container exposes a certain port
@@ -1425,3 +1464,13 @@ func (container *Container) Exposes(p Port) bool {
 	_, exists := container.Config.ExposedPorts[p]
 	return exists
 }
+
+func (container *Container) GetPtyMaster() (*os.File, error) {
+	if container.ptyMaster == nil {
+		return nil, ErrNoTTY
+	}
+	if pty, ok := container.ptyMaster.(*os.File); ok {
+		return pty, nil
+	}
+	return nil, ErrNotATTY
+}

contrib/docker-device-tool/device_tool.go

@@ -0,0 +1,170 @@
+package main
+
+import (
+	"flag"
+	"fmt"
+	"github.com/dotcloud/docker/graphdriver/devmapper"
+	"os"
+	"path"
+	"sort"
+	"strconv"
+	"strings"
+)
+
+func usage() {
+	fmt.Fprintf(os.Stderr, "Usage: %s <flags>  [status] | [list] | [device id]  | [resize new-pool-size] | [snap new-id base-id] | [remove id] | [mount id mountpoint]\n", os.Args[0])
+	flag.PrintDefaults()
+	os.Exit(1)
+}
+
+func byteSizeFromString(arg string) (int64, error) {
+	digits := ""
+	rest := ""
+	last := strings.LastIndexAny(arg, "0123456789")
+	if last >= 0 {
+		digits = arg[:last+1]
+		rest = arg[last+1:]
+	}
+
+	val, err := strconv.ParseInt(digits, 10, 64)
+	if err != nil {
+		return val, err
+	}
+
+	rest = strings.ToLower(strings.TrimSpace(rest))
+
+	var multiplier int64 = 1
+	switch rest {
+	case "":
+		multiplier = 1
+	case "k", "kb":
+		multiplier = 1024
+	case "m", "mb":
+		multiplier = 1024 * 1024
+	case "g", "gb":
+		multiplier = 1024 * 1024 * 1024
+	case "t", "tb":
+		multiplier = 1024 * 1024 * 1024 * 1024
+	default:
+		return 0, fmt.Errorf("Unknown size unit: %s", rest)
+	}
+
+	return val * multiplier, nil
+}
+
+func main() {
+	root := flag.String("r", "/var/lib/docker", "Docker root dir")
+	flDebug := flag.Bool("D", false, "Debug mode")
+
+	flag.Parse()
+
+	if *flDebug {
+		os.Setenv("DEBUG", "1")
+	}
+
+	if flag.NArg() < 1 {
+		usage()
+	}
+
+	args := flag.Args()
+
+	home := path.Join(*root, "devicemapper")
+	devices, err := devmapper.NewDeviceSet(home, false)
+	if err != nil {
+		fmt.Println("Can't initialize device mapper: ", err)
+		os.Exit(1)
+	}
+
+	switch args[0] {
+	case "status":
+		status := devices.Status()
+		fmt.Printf("Pool name: %s\n", status.PoolName)
+		fmt.Printf("Data Loopback file: %s\n", status.DataLoopback)
+		fmt.Printf("Metadata Loopback file: %s\n", status.MetadataLoopback)
+		fmt.Printf("Sector size: %d\n", status.SectorSize)
+		fmt.Printf("Data use: %d of %d (%.1f %%)\n", status.Data.Used, status.Data.Total, 100.0*float64(status.Data.Used)/float64(status.Data.Total))
+		fmt.Printf("Metadata use: %d of %d (%.1f %%)\n", status.Metadata.Used, status.Metadata.Total, 100.0*float64(status.Metadata.Used)/float64(status.Metadata.Total))
+		break
+	case "list":
+		ids := devices.List()
+		sort.Strings(ids)
+		for _, id := range ids {
+			fmt.Println(id)
+		}
+		break
+	case "device":
+		if flag.NArg() < 2 {
+			usage()
+		}
+		status, err := devices.GetDeviceStatus(args[1])
+		if err != nil {
+			fmt.Println("Can't get device info: ", err)
+			os.Exit(1)
+		}
+		fmt.Printf("Id: %d\n", status.DeviceId)
+		fmt.Printf("Size: %d\n", status.Size)
+		fmt.Printf("Transaction Id: %d\n", status.TransactionId)
+		fmt.Printf("Size in Sectors: %d\n", status.SizeInSectors)
+		fmt.Printf("Mapped Sectors: %d\n", status.MappedSectors)
+		fmt.Printf("Highest Mapped Sector: %d\n", status.HighestMappedSector)
+		break
+	case "resize":
+		if flag.NArg() < 2 {
+			usage()
+		}
+
+		size, err := byteSizeFromString(args[1])
+		if err != nil {
+			fmt.Println("Invalid size: ", err)
+			os.Exit(1)
+		}
+
+		err = devices.ResizePool(size)
+		if err != nil {
+			fmt.Println("Error resizeing pool: ", err)
+			os.Exit(1)
+		}
+
+		break
+	case "snap":
+		if flag.NArg() < 3 {
+			usage()
+		}
+
+		err := devices.AddDevice(args[1], args[2])
+		if err != nil {
+			fmt.Println("Can't create snap device: ", err)
+			os.Exit(1)
+		}
+		break
+	case "remove":
+		if flag.NArg() < 2 {
+			usage()
+		}
+
+		err := devices.RemoveDevice(args[1])
+		if err != nil {
+			fmt.Println("Can't remove device: ", err)
+			os.Exit(1)
+		}
+		break
+	case "mount":
+		if flag.NArg() < 3 {
+			usage()
+		}
+
+		err := devices.MountDevice(args[1], args[2], false)
+		if err != nil {
+			fmt.Println("Can't create snap device: ", err)
+			os.Exit(1)
+		}
+		break
+	default:
+		fmt.Printf("Unknown command %s\n", args[0])
+		usage()
+
+		os.Exit(1)
+	}
+
+	return
+}

contrib/init/systemd/docker.service

@@ -1,11 +1,9 @@
 [Unit]
 Description=Docker Application Container Engine 
 Documentation=http://docs.docker.io
-Requires=network.target
-After=multi-user.target
+After=network.target
 
 [Service]
-Type=simple
 ExecStartPre=/bin/mount --make-rprivate /
 ExecStart=/usr/bin/docker -d
 

contrib/mkimage-centos.sh

@@ -1,15 +0,0 @@
-#!/bin/bash
-# Create a CentOS base image for Docker
-# From unclejack https://github.com/dotcloud/docker/issues/290
-set -e
-
-MIRROR_URL="http://centos.netnitco.net/6.4/os/x86_64/"
-MIRROR_URL_UPDATES="http://centos.netnitco.net/6.4/updates/x86_64/"
-
-yum install -y febootstrap xz
-
-febootstrap -i bash -i coreutils -i tar -i bzip2 -i gzip -i vim-minimal -i wget -i patch -i diffutils -i iproute -i yum centos centos64  $MIRROR_URL -u $MIRROR_URL_UPDATES
-touch centos64/etc/resolv.conf
-touch centos64/sbin/init
-
-tar --numeric-owner -Jcpf centos-64.tar.xz -C centos64 .

contrib/mkimage-rinse.sh

@@ -0,0 +1,112 @@
+#!/bin/bash
+set -e
+
+repo="$1"
+distro="$2"
+mirror="$3"
+
+if [ ! "$repo" ] || [ ! "$distro" ]; then
+	self="$(basename $0)"
+	echo >&2 "usage: $self repo distro [mirror]"
+	echo >&2
+	echo >&2 "   ie: $self username/centos centos-5"
+	echo >&2 "       $self username/centos centos-6"
+	echo >&2
+	echo >&2 "   ie: $self username/slc slc-5"
+	echo >&2 "       $self username/slc slc-6"
+	echo >&2
+	echo >&2 "   ie: $self username/centos centos-5 http://vault.centos.org/5.8/os/x86_64/CentOS/"
+	echo >&2 "       $self username/centos centos-6 http://vault.centos.org/6.3/os/x86_64/Packages/"
+	echo >&2
+	echo >&2 'See /etc/rinse for supported values of "distro" and for examples of'
+	echo >&2 '  expected values of "mirror".'
+	echo >&2
+	echo >&2 'This script is tested to work with the original upstream version of rinse,'
+	echo >&2 '  found at http://www.steve.org.uk/Software/rinse/ and also in Debian at'
+	echo >&2 '  http://packages.debian.org/wheezy/rinse -- as always, YMMV.'
+	echo >&2
+	exit 1
+fi
+
+target="/tmp/docker-rootfs-rinse-$distro-$$-$RANDOM"
+
+cd "$(dirname "$(readlink -f "$BASH_SOURCE")")"
+returnTo="$(pwd -P)"
+
+rinseArgs=( --arch amd64 --distribution "$distro" --directory "$target" )
+if [ "$mirror" ]; then
+	rinseArgs+=( --mirror "$mirror" )
+fi
+
+set -x
+
+mkdir -p "$target"
+
+sudo rinse "${rinseArgs[@]}"
+
+cd "$target"
+
+# rinse fails a little at setting up /dev, so we'll just wipe it out and create our own
+sudo rm -rf dev
+sudo mkdir -m 755 dev
+(
+	cd dev
+	sudo ln -sf /proc/self/fd ./
+	sudo mkdir -m 755 pts
+	sudo mkdir -m 1777 shm
+	sudo mknod -m 600 console c 5 1
+	sudo mknod -m 600 initctl p
+	sudo mknod -m 666 full c 1 7
+	sudo mknod -m 666 null c 1 3
+	sudo mknod -m 666 ptmx c 5 2
+	sudo mknod -m 666 random c 1 8
+	sudo mknod -m 666 tty c 5 0
+	sudo mknod -m 666 tty0 c 4 0
+	sudo mknod -m 666 urandom c 1 9
+	sudo mknod -m 666 zero c 1 5
+)
+
+# effectively: febootstrap-minimize --keep-zoneinfo --keep-rpmdb --keep-services "$target"
+#  locales
+sudo rm -rf usr/{{lib,share}/locale,{lib,lib64}/gconv,bin/localedef,sbin/build-locale-archive}
+#  docs
+sudo rm -rf usr/share/{man,doc,info,gnome/help}
+#  cracklib
+sudo rm -rf usr/share/cracklib
+#  i18n
+sudo rm -rf usr/share/i18n
+#  yum cache
+sudo rm -rf var/cache/yum
+sudo mkdir -p --mode=0755 var/cache/yum
+#  sln
+sudo rm -rf sbin/sln
+#  ldconfig
+#sudo rm -rf sbin/ldconfig
+sudo rm -rf etc/ld.so.cache var/cache/ldconfig
+sudo mkdir -p --mode=0755 var/cache/ldconfig
+
+# allow networking init scripts inside the container to work without extra steps
+echo 'NETWORKING=yes' | sudo tee etc/sysconfig/network > /dev/null
+
+# to restore locales later:
+#  yum reinstall glibc-common
+
+version=
+if [ -r etc/redhat-release ]; then
+	version="$(sed -E 's/^[^0-9.]*([0-9.]+).*$/\1/' etc/redhat-release)"
+elif [ -r etc/SuSE-release ]; then
+	version="$(awk '/^VERSION/ { print $3 }' etc/SuSE-release)"
+fi
+
+if [ -z "$version" ]; then
+	echo >&2 "warning: cannot autodetect OS version, using $distro as tag"
+	sleep 20
+	version="$distro"
+fi
+
+sudo tar --numeric-owner -c . | docker import - $repo:$version
+
+docker run -i -t $repo:$version echo success
+
+cd "$returnTo"
+sudo rm -rf "$target"

contrib/mkseccomp.pl

@@ -0,0 +1,77 @@
+#!/usr/bin/perl
+#
+# A simple helper script to help people build seccomp profiles for
+# Docker/LXC.  The goal is mostly to reduce the attack surface to the
+# kernel, by restricting access to rarely used, recently added or not used
+# syscalls.
+#
+# This script processes one or more files which contain the list of system
+# calls to be allowed.  See mkseccomp.sample for more information how you
+# can configure the list of syscalls.  When run, this script produces output
+# which, when stored in a file, can be passed to docker as follows:
+#
+# docker run -lxc-conf="lxc.seccomp=$file" <rest of arguments>
+#
+# The included sample file shows how to cut about a quarter of all syscalls,
+# which affecting most applications.
+#
+# For specific situations it is possible to reduce the list further. By
+# reducing the list to just those syscalls required by a certain application
+# you can make it difficult for unknown/unexpected code to run.
+#
+# Run this script as follows:
+#
+# ./mkseccomp.pl < mkseccomp.sample >syscalls.list
+# or
+# ./mkseccomp.pl mkseccomp.sample >syscalls.list
+#
+# Multiple files can be specified, in which case the lists of syscalls are
+# combined.
+#
+# By Martijn van Oosterhout <kleptog@svana.org> Nov 2013
+
+# How it works:
+#
+# This program basically spawns two processes to form a chain like:
+#
+# <process data section to prefix __NR_> | cpp | <add header and filter unknown syscalls>
+
+use strict;
+use warnings;
+
+if( -t ) {
+    print STDERR "Helper script to make seccomp filters for Docker/LXC.\n";
+    print STDERR "Usage: mkseccomp.pl [files...]\n";
+    exit 1;
+}
+
+my $pid = open(my $in, "-|") // die "Couldn't fork1 ($!)\n";
+
+if($pid == 0) {  # Child
+    $pid = open(my $out, "|-") // die "Couldn't fork2 ($!)\n";
+
+    if($pid == 0) { # Child, which execs cpp
+        exec "cpp" or die "Couldn't exec cpp ($!)\n";
+        exit 1;
+    }
+
+    # Process the DATA section and output to cpp
+    print $out "#include <sys/syscall.h>\n";
+    while(<>) {
+        if(/^\w/) {
+            print $out "__NR_$_";
+        }
+    }
+    close $out;
+    exit 0;
+
+}
+
+# Print header and then process output from cpp.
+print "1\n";
+print "whitelist\n";
+
+while(<$in>) {
+    print if( /^[0-9]/ );
+}
+

contrib/mkseccomp.sample

@@ -0,0 +1,444 @@
+/* This sample file is an example for mkseccomp.pl to produce a seccomp file
+ * which restricts syscalls that are only useful for an admin but allows the
+ * vast majority of normal userspace programs to run normally.
+ *
+ * The format of this file is one line per syscall.  This is then processed
+ * and passed to 'cpp' to convert the names to numbers using whatever is
+ * correct for your platform.  As such C-style comments are permitted.  Note
+ * this also means that C preprocessor macros are also allowed.  So it is
+ * possible to create groups surrounded by #ifdef/#endif and control their
+ * inclusion via #define (not #include).
+ *
+ * Syscalls that don't exist on your architecture are silently filtered out.
+ * Syscalls marked with (*) are required for a container to spawn a bash
+ * shell successfully (not necessarily full featured).  Listing the same
+ * syscall multiple times is no problem.
+ *
+ * If you want to make a list specifically for one application the easiest
+ * way is to run the application under strace, like so:
+ *
+ * $ strace -f -q -c -o strace.out application args...
+ *
+ * Once you have a reasonable sample of the execution of the program, exit
+ * it.  The file strace.out will have a summary of the syscalls used.  Copy
+ * that list into this file, comment out everything else except the starred
+ * syscalls (which you need for the container to start) and you're done.
+ *
+ * To get the list of syscalls from the strace output this works well for
+ * me
+ *
+ * $ cut -c52 < strace.out
+ *
+ * This sample list was compiled as a combination of all the syscalls
+ * available on i386 and amd64 on Ubuntu Precise, as such it may not contain
+ * everything and not everything may be relevent for your system.  This
+ * shouldn't be a problem.
+ */
+
+// Filesystem/File descriptor related
+access                 // (*)
+chdir                  // (*)
+chmod
+chown
+chown32
+close                  // (*)
+creat
+dup                    // (*)
+dup2                   // (*)
+dup3
+epoll_create
+epoll_create1
+epoll_ctl
+epoll_ctl_old
+epoll_pwait
+epoll_wait
+epoll_wait_old
+eventfd
+eventfd2
+faccessat              // (*)
+fadvise64
+fadvise64_64
+fallocate
+fanotify_init
+fanotify_mark
+ioctl                  // (*)
+fchdir
+fchmod
+fchmodat
+fchown
+fchown32
+fchownat
+fcntl                  // (*)
+fcntl64
+fdatasync
+fgetxattr
+flistxattr
+flock
+fremovexattr
+fsetxattr
+fstat                  // (*)
+fstat64
+fstatat64
+fstatfs
+fstatfs64
+fsync
+ftruncate
+ftruncate64
+getcwd                 // (*)
+getdents               // (*)
+getdents64
+getxattr
+inotify_add_watch
+inotify_init
+inotify_init1
+inotify_rm_watch
+io_cancel
+io_destroy
+io_getevents
+io_setup
+io_submit
+lchown
+lchown32
+lgetxattr
+link
+linkat
+listxattr
+llistxattr
+llseek
+_llseek
+lremovexattr
+lseek                  // (*)
+lsetxattr
+lstat
+lstat64
+mkdir
+mkdirat
+mknod
+mknodat
+newfstatat
+_newselect
+oldfstat
+oldlstat
+oldolduname
+oldstat
+olduname
+oldwait4
+open                   // (*)
+openat                 // (*)
+pipe                   // (*)
+pipe2
+poll
+ppoll
+pread64
+preadv
+futimesat
+pselect6
+pwrite64
+pwritev
+read                   // (*)
+readahead
+readdir
+readlink
+readlinkat
+readv
+removexattr
+rename
+renameat
+rmdir
+select
+sendfile
+sendfile64
+setxattr
+splice
+stat                   // (*)
+stat64
+statfs                 // (*)
+statfs64
+symlink
+symlinkat
+sync
+sync_file_range
+sync_file_range2
+syncfs
+tee
+truncate
+truncate64
+umask
+unlink
+unlinkat
+ustat
+utime
+utimensat
+utimes
+write                  // (*)
+writev
+
+// Network related
+accept
+accept4
+bind                   // (*)
+connect                // (*)
+getpeername
+getsockname            // (*)
+getsockopt
+listen
+recv
+recvfrom               // (*)
+recvmmsg
+recvmsg
+send
+sendmmsg
+sendmsg
+sendto                 // (*)
+setsockopt
+shutdown
+socket                 // (*)
+socketcall
+socketpair
+
+// Signal related
+pause
+rt_sigaction           // (*)
+rt_sigpending
+rt_sigprocmask         // (*)
+rt_sigqueueinfo
+rt_sigreturn           // (*)
+rt_sigsuspend
+rt_sigtimedwait
+rt_tgsigqueueinfo
+sigaction
+sigaltstack            // (*)
+signal
+signalfd
+signalfd4
+sigpending
+sigprocmask
+sigreturn
+sigsuspend
+
+// Other needed POSIX
+alarm
+brk                    // (*)
+clock_adjtime
+clock_getres
+clock_gettime
+clock_nanosleep
+//clock_settime
+gettimeofday
+nanosleep
+nice
+sysinfo
+syslog
+time
+timer_create
+timer_delete
+timerfd_create
+timerfd_gettime
+timerfd_settime
+timer_getoverrun
+timer_gettime
+timer_settime
+times
+uname                  // (*)
+
+// Memory control
+madvise
+mbind
+mincore
+mlock
+mlockall
+mmap                   // (*)
+mmap2
+mprotect               // (*)
+mremap
+msync
+munlock
+munlockall
+munmap                 // (*)
+remap_file_pages
+set_mempolicy
+vmsplice
+
+// Process control
+capget
+//capset
+clone                  // (*)
+execve                 // (*)
+exit                   // (*)
+exit_group             // (*)
+fork
+getcpu
+getpgid
+getpgrp                // (*)
+getpid                 // (*)
+getppid                // (*)
+getpriority
+getresgid
+getresgid32
+getresuid
+getresuid32
+getrlimit              // (*)
+getrusage
+getsid
+getuid                 // (*)
+getuid32
+getegid                // (*)
+getegid32
+geteuid                // (*)
+geteuid32
+getgid                 // (*)
+getgid32
+getgroups
+getgroups32
+getitimer
+get_mempolicy
+kill
+//personality
+prctl
+prlimit64
+sched_getaffinity
+sched_getparam
+sched_get_priority_max
+sched_get_priority_min
+sched_getscheduler
+sched_rr_get_interval
+//sched_setaffinity
+//sched_setparam
+//sched_setscheduler
+sched_yield
+setfsgid
+setfsgid32
+setfsuid
+setfsuid32
+setgid
+setgid32
+setgroups
+setgroups32
+setitimer
+setpgid                // (*)
+setpriority
+setregid
+setregid32
+setresgid
+setresgid32
+setresuid
+setresuid32
+setreuid
+setreuid32
+setrlimit
+setsid
+setuid
+setuid32
+ugetrlimit
+vfork
+wait4                  // (*)
+waitid
+waitpid
+
+// IPC
+ipc
+mq_getsetattr
+mq_notify
+mq_open
+mq_timedreceive
+mq_timedsend
+mq_unlink
+msgctl
+msgget
+msgrcv
+msgsnd
+semctl
+semget
+semop
+semtimedop
+shmat
+shmctl
+shmdt
+shmget
+
+// Linux specific, mostly needed for thread-related stuff
+arch_prctl             // (*)
+get_robust_list
+get_thread_area
+gettid
+futex                  // (*)
+restart_syscall        // (*)
+set_robust_list        // (*)
+set_thread_area
+set_tid_address        // (*)
+tgkill
+tkill
+
+// Admin syscalls, these are blocked
+//acct
+//adjtimex
+//bdflush
+//chroot
+//create_module
+//delete_module
+//get_kernel_syms      // Obsolete
+//idle                 // Obsolete
+//init_module
+//ioperm
+//iopl
+//ioprio_get
+//ioprio_set
+//kexec_load
+//lookup_dcookie       // oprofile only?
+//migrate_pages        // NUMA
+//modify_ldt
+//mount
+//move_pages           // NUMA
+//name_to_handle_at    // NFS server
+//nfsservctl           // NFS server
+//open_by_handle_at    // NFS server
+//perf_event_open
+//pivot_root
+//process_vm_readv     // For debugger
+//process_vm_writev    // For debugger
+//ptrace               // For debugger
+//query_module
+//quotactl
+//reboot
+//setdomainname
+//sethostname
+//setns
+//settimeofday
+//sgetmask             // Obsolete
+//ssetmask             // Obsolete
+//stime
+//swapoff
+//swapon
+//_sysctl
+//sysfs
+//sys_setaltroot
+//umount
+//umount2
+//unshare
+//uselib
+//vhangup
+//vm86
+//vm86old
+
+// Kernel key management
+//add_key
+//keyctl
+//request_key
+
+// Unimplemented
+//afs_syscall
+//break
+//ftime
+//getpmsg
+//gtty
+//lock
+//madvise1
+//mpx
+//prof
+//profil
+//putpmsg
+//security
+//stty
+//tuxcall
+//ulimit
+//vserver

contrib/udev/80-docker.rules

@@ -0,0 +1,3 @@
+# hide docker's loopback devices from udisks, and thus from user desktops
+SUBSYSTEM=="block", ENV{DM_NAME}=="docker-*", ENV{UDISKS_PRESENTATION_HIDE}="1", ENV{UDISKS_IGNORE}="1"
+SUBSYSTEM=="block", DEVPATH=="/devices/virtual/block/loop*", ATTR{loop/backing_file}=="/var/lib/docker/*", ENV{UDISKS_PRESENTATION_HIDE}="1", ENV{UDISKS_IGNORE}="1"

contrib/vagrant-docker/README.md

@@ -17,3 +17,34 @@ meaning you can use Vagrant to control Docker containers.
 
 * [docker-provider](https://github.com/fgrehm/docker-provider)
 * [vagrant-shell](https://github.com/destructuring/vagrant-shell)
+
+## Setting up Vagrant-docker with the Remote API
+
+The initial Docker upstart script will not work because it runs on `127.0.0.1`, which is not accessible to the host machine. Instead, we need to change the script to connect to `0.0.0.0`. To do this, modify `/etc/init/docker.conf` to look like this:
+
+```
+description     "Docker daemon"
+
+start on filesystem and started lxc-net
+stop on runlevel [!2345]
+
+respawn
+
+script
+    /usr/bin/docker -d -H=tcp://0.0.0.0:4243/
+end script
+```
+
+Once that's done, you need to set up a SSH tunnel between your host machine and the vagrant machine that's running Docker. This can be done by running the following command in a host terminal:
+
+```
+ssh -L 4243:localhost:4243 -p 2222 vagrant@localhost
+```
+
+(The first 4243 is what your host can connect to, the second 4243 is what port Docker is running on in the vagrant machine, and the 2222 is the port Vagrant is providing for SSH. If VirtualBox is the VM you're using, you can see what value "2222" should be by going to: Network > Adapter 1 > Advanced > Port Forwarding in the VirtualBox GUI.)
+
+Note that because the port has been changed, to run docker commands from within the command line you must run them like this:
+
+```
+sudo docker -H 0.0.0.0:4243 < commands for docker >
+```

contrib/zfs/MAINTAINERS

@@ -0,0 +1 @@
+Gurjeet Singh <gurjeet@singh.im> (gurjeet.singh.im)

contrib/zfs/README.md

@@ -0,0 +1,22 @@
+# ZFS Storage Driver
+
+This is a placeholder to declare the presence and status of ZFS storage driver
+for containers.
+
+The current development is done in Gurjeet Singh's fork of Docker, under the
+branch named [zfs_driver].
+
+[zfs_driver]: https://github.com/gurjeet/docker/tree/zfs_driver
+
+
+# Status
+
+Pre-alpha
+
+The code is under development. Contributions in the form of suggestions,
+code-reviews, and patches are welcome.
+
+Please send the communication to gurjeet@singh.im and CC at least one Docker
+mailing list.
+
+

docker/docker.go

@@ -23,21 +23,25 @@ func main() {
 		sysinit.SysInit()
 		return
 	}
-	// FIXME: Switch d and D ? (to be more sshd like)
-	flVersion := flag.Bool("v", false, "Print version information and quit")
-	flDaemon := flag.Bool("d", false, "Daemon mode")
-	flDebug := flag.Bool("D", false, "Debug mode")
-	flAutoRestart := flag.Bool("r", true, "Restart previously running containers")
-	bridgeName := flag.String("b", "", "Attach containers to a pre-existing network bridge. Use 'none' to disable container networking")
-	pidfile := flag.String("p", "/var/run/docker.pid", "File containing process PID")
-	flRoot := flag.String("g", "/var/lib/docker", "Path to use as the root of the docker runtime.")
-	flEnableCors := flag.Bool("api-enable-cors", false, "Enable CORS requests in the remote api.")
-	flDns := flag.String("dns", "", "Set custom dns servers")
-	flHosts := utils.ListOpts{fmt.Sprintf("unix://%s", docker.DEFAULTUNIXSOCKET)}
-	flag.Var(&flHosts, "H", "tcp://host:port to bind/connect to or unix://path/to/socket to use")
-	flEnableIptables := flag.Bool("iptables", true, "Disable iptables within docker")
-	flDefaultIp := flag.String("ip", "0.0.0.0", "Default ip address to use when binding a containers ports")
-	flInterContainerComm := flag.Bool("icc", true, "Enable inter-container communication")
+
+	var (
+		flVersion            = flag.Bool("v", false, "Print version information and quit")
+		flDaemon             = flag.Bool("d", false, "Enable daemon mode")
+		flDebug              = flag.Bool("D", false, "Enable debug mode")
+		flAutoRestart        = flag.Bool("r", true, "Restart previously running containers")
+		bridgeName           = flag.String("b", "", "Attach containers to a pre-existing network bridge; use 'none' to disable container networking")
+		pidfile              = flag.String("p", "/var/run/docker.pid", "Path to use for daemon PID file")
+		flRoot               = flag.String("g", "/var/lib/docker", "Path to use as the root of the docker runtime")
+		flEnableCors         = flag.Bool("api-enable-cors", false, "Enable CORS headers in the remote API")
+		flDns                = docker.NewListOpts(docker.ValidateIp4Address)
+		flEnableIptables     = flag.Bool("iptables", true, "Disable docker's addition of iptables rules")
+		flDefaultIp          = flag.String("ip", "0.0.0.0", "Default IP address to use when binding container ports")
+		flInterContainerComm = flag.Bool("icc", true, "Enable inter-container communication")
+		flGraphDriver        = flag.String("s", "", "Force the docker runtime to use a specific storage driver")
+		flHosts              = docker.NewListOpts(docker.ValidateHost)
+	)
+	flag.Var(&flDns, "dns", "Force docker to use specific DNS servers")
+	flag.Var(&flHosts, "H", "Multiple tcp://host:port or unix://path/to/socket to bind in daemon mode, single connection otherwise")
 
 	flag.Parse()
 
@@ -45,16 +49,9 @@ func main() {
 		showVersion()
 		return
 	}
-	if len(flHosts) > 1 {
-		flHosts = flHosts[1:] //trick to display a nice default value in the usage
-	}
-	for i, flHost := range flHosts {
-		host, err := utils.ParseHost(docker.DEFAULTHTTPHOST, docker.DEFAULTHTTPPORT, flHost)
-		if err == nil {
-			flHosts[i] = host
-		} else {
-			log.Fatal(err)
-		}
+	if flHosts.Len() == 0 {
+		// If we do not have a host, default to unix socket
+		flHosts.Set(fmt.Sprintf("unix://%s", docker.DEFAULTUNIXSOCKET))
 	}
 
 	if *flDebug {
@@ -77,28 +74,32 @@ func main() {
 		job.Setenv("Root", *flRoot)
 		job.SetenvBool("AutoRestart", *flAutoRestart)
 		job.SetenvBool("EnableCors", *flEnableCors)
-		job.Setenv("Dns", *flDns)
+		job.SetenvList("Dns", flDns.GetAll())
 		job.SetenvBool("EnableIptables", *flEnableIptables)
 		job.Setenv("BridgeIface", *bridgeName)
 		job.Setenv("DefaultIp", *flDefaultIp)
 		job.SetenvBool("InterContainerCommunication", *flInterContainerComm)
+		job.Setenv("GraphDriver", *flGraphDriver)
 		if err := job.Run(); err != nil {
 			log.Fatal(err)
 		}
 		// Serve api
-		job = eng.Job("serveapi", flHosts...)
+		job = eng.Job("serveapi", flHosts.GetAll()...)
 		job.SetenvBool("Logging", true)
 		if err := job.Run(); err != nil {
 			log.Fatal(err)
 		}
 	} else {
-		if len(flHosts) > 1 {
+		if flHosts.Len() > 1 {
 			log.Fatal("Please specify only one -H")
 		}
-		protoAddrParts := strings.SplitN(flHosts[0], "://", 2)
+		protoAddrParts := strings.SplitN(flHosts.GetAll()[0], "://", 2)
 		if err := docker.ParseCommands(protoAddrParts[0], protoAddrParts[1], flag.Args()...); err != nil {
 			if sterr, ok := err.(*utils.StatusError); ok {
-				os.Exit(sterr.Status)
+				if sterr.Status != "" {
+					log.Println(sterr.Status)
+				}
+				os.Exit(sterr.StatusCode)
 			}
 			log.Fatal(err)
 		}

docs/Dockerfile

@@ -9,7 +9,7 @@ run apt-get install -y python-setuptools make
 run easy_install pip
 #from docs/requirements.txt, but here to increase cacheability
 run pip install Sphinx==1.1.3
-run pip install sphinxcontrib-httpdomain==1.1.8
+run pip install sphinxcontrib-httpdomain==1.1.9
 add . /docs
 run cd /docs; make docs
 

docs/MAINTAINERS

@@ -1,2 +1,4 @@
 Andy Rothfusz <andy@dotcloud.com> (@metalivedev)
 Ken Cochrane <ken@dotcloud.com> (@kencochrane)
+James Turnbull <james@lovedthanlost.net> (@jamesturnbull)
+Sven Dowideit <SvenDowideit@fosiki.com> (@SvenDowideit)

docs/README.md

@@ -41,11 +41,12 @@ its dependencies. There are two main ways to install this tool:
 
 ###Native Installation
 
-* Install sphinx: `pip install sphinx`
-    * Mac OS X: `[sudo] pip-2.7 install sphinx`
-* Install sphinx httpdomain contrib package: `pip install sphinxcontrib-httpdomain`
-    * Mac OS X: `[sudo] pip-2.7 install sphinxcontrib-httpdomain`
-* If pip is not available you can probably install it using your favorite package manager as **python-pip**
+Install dependencies from `requirements.txt` file in your `docker/docs`
+directory:
+
+* Linux: `pip install -r docs/requirements.txt`
+
+* Mac OS X: `[sudo] pip-2.7 -r docs/requirements.txt`
 
 ###Alternative Installation: Docker Container
 
@@ -136,7 +137,7 @@ Manpages
 --------
 
 * To make the manpages, run ``make man``. Please note there is a bug
-  in spinx 1.1.3 which makes this fail.  Upgrade to the latest version
+  in Sphinx 1.1.3 which makes this fail.  Upgrade to the latest version
   of Sphinx.
 * Then preview the manpage by running ``man _build/man/docker.1``,
   where ``_build/man/docker.1`` is the path to the generated manfile

docs/sources/api/docker_remote_api.rst

@@ -34,6 +34,28 @@ Calling /images/<name>/insert is the same as calling
 You can still call an old version of the api using
 /v1.0/images/<name>/insert
 
+
+v1.8
+****
+
+Full Documentation
+------------------
+
+:doc:`docker_remote_api_v1.8`
+
+What's new
+----------
+
+.. http:post:: /build
+
+   **New!** This endpoint now returns build status as json stream. In case
+   of a build error, it returns the exit status of the failed command.
+
+.. http:get:: /containers/(id)/json
+
+    **New!** This endpoint now returns the host config for the container.
+
+
 v1.7
 ****
 

docs/sources/api/docker_remote_api_v1.7.rst

@@ -132,7 +132,9 @@ Create a container
 		],
 		"Dns":null,
 		"Image":"base",
-		"Volumes":{},
+		"Volumes":{
+			"/tmp": {}
+		},
 		"VolumesFrom":"",
 		"WorkingDir":""
 
@@ -361,8 +363,12 @@ Start a container
 
            {
                 "Binds":["/tmp:/tmp"],
-                "LxcConf":{"lxc.utsname":"docker"}
+                "LxcConf":{"lxc.utsname":"docker"},
+                "PortBindings":null
+                "PublishAllPorts":false
            }
+           
+        Binds need to reference Volumes that were defined during container creation.
 
         **Example response**:
 
@@ -990,10 +996,10 @@ Build an image from Dockerfile via stdin
    .. sourcecode:: http
 
       HTTP/1.1 200 OK
+      Content-Type: application/json
 
       {{ STREAM }}
 
-
    The stream must be a tar archive compressed with one of the
    following algorithms: identity (no compression), gzip, bzip2,
    xz. 

docs/sources/api/registry_index_spec.rst

@@ -37,7 +37,7 @@ We expect that there will be only one instance of the index, run and managed by
 - It delegates authentication and authorization to the Index Auth service using tokens
 - It supports different storage backends (S3, cloud files, local FS)
 - It doesn’t have a local database
-- It will be open-sourced at some point
+- `Source Code <https://github.com/dotcloud/docker-registry>`_
 
 We expect that there will be multiple registries out there. To help to grasp the context, here are some examples of registries:
 
@@ -46,10 +46,6 @@ We expect that there will be multiple registries out there. To help to grasp the
 - **vendor registry**: such a registry is provided by a software vendor, who wants to distribute docker images. It would be operated and managed by the vendor. Only users authorized by the vendor would be able to get write access. Some images would be public (accessible for anyone), others private (accessible only for authorized users). Authentication and authorization would be delegated to the Index. The goal of vendor registries is to let someone do “docker pull basho/riak1.3” and automatically push from the vendor registry (instead of a sponsor registry); i.e. get all the convenience of a sponsor registry, while retaining control on the asset distribution.
 - **private registry**: such a registry is located behind a firewall, or protected by an additional security layer (HTTP authorization, SSL client-side certificates, IP address authorization...). The registry is operated by a private entity, outside of dotCloud’s control. It can optionally delegate additional authorization to the Index, but it is not mandatory.
 
-.. note::
-
-    Mirror registries and private registries which do not use the Index don’t even need to run the registry code. They can be implemented by any kind of transport implementing HTTP GET and PUT. Read-only registries can be powered by a simple static HTTP server.
-
 .. note::
 
     The latter implies that while HTTP is the protocol of choice for a registry, multiple schemes are possible (and in some cases, trivial):

docs/sources/commandline/cli.rst

@@ -18,6 +18,38 @@ To list available commands, either run ``docker`` with no parameters or execute
 
     ...
 
+.. _cli_daemon:
+
+``daemon``
+----------
+
+::
+
+    Usage of docker:
+      -D=false: Enable debug mode
+      -H=[unix:///var/run/docker.sock]: Multiple tcp://host:port or unix://path/to/socket to bind in daemon mode, single connection otherwise
+      -api-enable-cors=false: Enable CORS headers in the remote API
+      -b="": Attach containers to a pre-existing network bridge; use 'none' to disable container networking
+      -d=false: Enable daemon mode
+      -dns="": Force docker to use specific DNS servers
+      -g="/var/lib/docker": Path to use as the root of the docker runtime
+      -icc=true: Enable inter-container communication
+      -ip="0.0.0.0": Default IP address to use when binding container ports
+      -iptables=true: Disable docker's addition of iptables rules
+      -p="/var/run/docker.pid": Path to use for daemon PID file
+      -r=true: Restart previously running containers
+      -s="": Force the docker runtime to use a specific storage driver
+      -v=false: Print version information and quit
+
+The docker daemon is the persistent process that manages containers.  Docker uses the same binary for both the 
+daemon and client.  To run the daemon you provide the ``-d`` flag.
+
+To force docker to use devicemapper as the storage driver, use ``docker -d -s devicemapper``
+
+To set the dns server for all docker containers, use ``docker -d -dns 8.8.8.8``
+
+To run the daemon with debug output, use ``docker -d -D``
+
 .. _cli_attach:
 
 ``attach``
@@ -34,7 +66,8 @@ To list available commands, either run ``docker`` with no parameters or execute
 
 You can detach from the container again (and leave it running) with
 ``CTRL-c`` (for a quiet exit) or ``CTRL-\`` to get a stacktrace of
-the Docker client when it quits.
+the Docker client when it quits.  When you detach from the container's 
+process the exit code will be retuned to the client.
 
 To stop a container, use ``docker stop``
 
@@ -110,7 +143,7 @@ Examples:
 
 .. code-block:: bash
 
-    sudo docker build .
+    $ sudo docker build .
     Uploading context 10240 bytes
     Step 1 : FROM busybox
     Pulling repository busybox
@@ -150,7 +183,7 @@ message.
 
 .. code-block:: bash
 
-   sudo docker build -t vieux/apache:2.0 .
+   $ sudo docker build -t vieux/apache:2.0 .
 
 This will build like the previous example, but it will then tag the
 resulting image. The repository name will be ``vieux/apache`` and the
@@ -159,7 +192,7 @@ tag will be ``2.0``
 
 .. code-block:: bash
 
-    sudo docker build - < Dockerfile
+    $ sudo docker build - < Dockerfile
 
 This will read a ``Dockerfile`` from *stdin* without context. Due to
 the lack of a context, no contents of any local directory will be sent
@@ -168,7 +201,7 @@ to the ``docker`` daemon.  Since there is no context, a Dockerfile
 
 .. code-block:: bash
 
-    sudo docker build github.com/creack/docker-firefox
+    $ sudo docker build github.com/creack/docker-firefox
 
 This will clone the Github repository and use the cloned repository as
 context. The ``Dockerfile`` at the root of the repository is used as
@@ -197,15 +230,15 @@ Simple commit of an existing container
 
 .. code-block:: bash
 
-	$ docker ps
+	$ sudo docker ps
 	ID                  IMAGE               COMMAND             CREATED             STATUS              PORTS
 	c3f279d17e0a        ubuntu:12.04        /bin/bash           7 days ago          Up 25 hours                             
 	197387f1b436        ubuntu:12.04        /bin/bash           7 days ago          Up 25 hours                             
 	$ docker commit c3f279d17e0a  SvenDowideit/testimage:version3
 	f5283438590d
 	$ docker images | head
-	REPOSITORY                        TAG                 ID                  CREATED             SIZE
-	SvenDowideit/testimage            version3            f5283438590d        16 seconds ago      204.2 MB (virtual 335.7 MB)
+	REPOSITORY                        TAG                 ID                  CREATED             VIRTUAL SIZE
+	SvenDowideit/testimage            version3            f5283438590d        16 seconds ago      335.7 MB
 	
 
 Full -run example
@@ -369,7 +402,13 @@ Show events in the past from a specified time
 
     Usage: docker export CONTAINER
 
-    Export the contents of a filesystem as a tar archive
+    Export the contents of a filesystem as a tar archive to STDOUT
+    
+for example:
+
+.. code-block:: bash
+
+    $ sudo docker export red_panda > latest.tar
 
 .. _cli_history:
 
@@ -442,16 +481,16 @@ Listing the most recently created images
 .. code-block:: bash
 
 	$ sudo docker images | head
-	REPOSITORY                    TAG                 IMAGE ID            CREATED             SIZE
-	<none>                        <none>              77af4d6b9913        19 hours ago        30.53 MB (virtual 1.089 GB)
-	committest                    latest              b6fa739cedf5        19 hours ago        30.53 MB (virtual 1.089 GB)
-	<none>                        <none>              78a85c484f71        19 hours ago        30.53 MB (virtual 1.089 GB)
-	docker                        latest              30557a29d5ab        20 hours ago        30.53 MB (virtual 1.089 GB)
-	<none>                        <none>              0124422dd9f9        20 hours ago        30.53 MB (virtual 1.089 GB)
-	<none>                        <none>              18ad6fad3402        22 hours ago        23.68 MB (virtual 1.082 GB)
-	<none>                        <none>              f9f1e26352f0        23 hours ago        30.46 MB (virtual 1.089 GB)
-	tryout                        latest              2629d1fa0b81        23 hours ago        16.4 kB (virtual 131.5 MB)
-	<none>                        <none>              5ed6274db6ce        24 hours ago        30.44 MB (virtual 1.089 GB)
+	REPOSITORY                    TAG                 IMAGE ID            CREATED             VIRTUAL SIZE
+	<none>                        <none>              77af4d6b9913        19 hours ago        1.089 GB
+	committest                    latest              b6fa739cedf5        19 hours ago        1.089 GB
+	<none>                        <none>              78a85c484f71        19 hours ago        1.089 GB
+	docker                        latest              30557a29d5ab        20 hours ago        1.089 GB
+	<none>                        <none>              0124422dd9f9        20 hours ago        1.089 GB
+	<none>                        <none>              18ad6fad3402        22 hours ago        1.082 GB
+	<none>                        <none>              f9f1e26352f0        23 hours ago        1.089 GB
+	tryout                        latest              2629d1fa0b81        23 hours ago        131.5 MB
+	<none>                        <none>              5ed6274db6ce        24 hours ago        1.089 GB
 
 Listing the full length image IDs
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -459,16 +498,16 @@ Listing the full length image IDs
 .. code-block:: bash
 
 	$ sudo docker images -notrunc | head
-	REPOSITORY                    TAG                 IMAGE ID                                                           CREATED             SIZE
-	<none>                        <none>              77af4d6b9913e693e8d0b4b294fa62ade6054e6b2f1ffb617ac955dd63fb0182   19 hours ago        30.53 MB (virtual 1.089 GB)
-	committest                    latest              b6fa739cedf5ea12a620a439402b6004d057da800f91c7524b5086a5e4749c9f   19 hours ago        30.53 MB (virtual 1.089 GB)
-	<none>                        <none>              78a85c484f71509adeaace20e72e941f6bdd2b25b4c75da8693efd9f61a37921   19 hours ago        30.53 MB (virtual 1.089 GB)
-	docker                        latest              30557a29d5abc51e5f1d5b472e79b7e296f595abcf19fe6b9199dbbc809c6ff4   20 hours ago        30.53 MB (virtual 1.089 GB)
-	<none>                        <none>              0124422dd9f9cf7ef15c0617cda3931ee68346455441d66ab8bdc5b05e9fdce5   20 hours ago        30.53 MB (virtual 1.089 GB)
-	<none>                        <none>              18ad6fad340262ac2a636efd98a6d1f0ea775ae3d45240d3418466495a19a81b   22 hours ago        23.68 MB (virtual 1.082 GB)
-	<none>                        <none>              f9f1e26352f0a3ba6a0ff68167559f64f3e21ff7ada60366e2d44a04befd1d3a   23 hours ago        30.46 MB (virtual 1.089 GB)
-	tryout                        latest              2629d1fa0b81b222fca63371ca16cbf6a0772d07759ff80e8d1369b926940074   23 hours ago        16.4 kB (virtual 131.5 MB)
-	<none>                        <none>              5ed6274db6ceb2397844896966ea239290555e74ef307030ebb01ff91b1914df   24 hours ago        30.44 MB (virtual 1.089 GB)
+	REPOSITORY                    TAG                 IMAGE ID                                                           CREATED             VIRTUAL SIZE
+	<none>                        <none>              77af4d6b9913e693e8d0b4b294fa62ade6054e6b2f1ffb617ac955dd63fb0182   19 hours ago        1.089 GB
+	committest                    latest              b6fa739cedf5ea12a620a439402b6004d057da800f91c7524b5086a5e4749c9f   19 hours ago        1.089 GB
+	<none>                        <none>              78a85c484f71509adeaace20e72e941f6bdd2b25b4c75da8693efd9f61a37921   19 hours ago        1.089 GB
+	docker                        latest              30557a29d5abc51e5f1d5b472e79b7e296f595abcf19fe6b9199dbbc809c6ff4   20 hours ago        1.089 GB
+	<none>                        <none>              0124422dd9f9cf7ef15c0617cda3931ee68346455441d66ab8bdc5b05e9fdce5   20 hours ago        1.089 GB
+	<none>                        <none>              18ad6fad340262ac2a636efd98a6d1f0ea775ae3d45240d3418466495a19a81b   22 hours ago        1.082 GB
+	<none>                        <none>              f9f1e26352f0a3ba6a0ff68167559f64f3e21ff7ada60366e2d44a04befd1d3a   23 hours ago        1.089 GB
+	tryout                        latest              2629d1fa0b81b222fca63371ca16cbf6a0772d07759ff80e8d1369b926940074   23 hours ago        131.5 MB
+	<none>                        <none>              5ed6274db6ceb2397844896966ea239290555e74ef307030ebb01ff91b1914df   24 hours ago        1.089 GB
 
 Displaying images visually
 ~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -519,7 +558,8 @@ Displaying image hierarchy
 
     Usage: docker import URL|- [REPOSITORY[:TAG]]
 
-    Create a new filesystem image from the contents of a tarball
+    Create an empty filesystem image and import the contents of the tarball 
+    (.tar, .tar.gz, .tgz, .bzip, .tar.xz, .txz) into it, then optionally tag it.
 
 At this time, the URL must start with ``http`` and point to a single
 file archive (.tar, .tar.gz, .tgz, .bzip, .tar.xz, .txz) containing a
@@ -591,6 +631,12 @@ might not get preserved.
 
     Insert a file from URL in the IMAGE at PATH
 
+Use the specified IMAGE as the parent for a new image which adds a
+:ref:`layer <layer_def>` containing the new file. ``insert`` does not modify 
+the original image, and the new image has the contents of the parent image, 
+plus the new file.
+
+
 Examples
 ~~~~~~~~
 
@@ -600,6 +646,7 @@ Insert file from github
 .. code-block:: bash
 
     $ sudo docker insert 8283e18b24bc https://raw.github.com/metalivedev/django/master/postinstall /tmp/postinstall.sh
+    06fd35556d7b
 
 .. _cli_inspect:
 
@@ -612,6 +659,52 @@ Insert file from github
 
     Return low-level information on a container
 
+      -format="": template to output results
+
+By default, this will render all results in a JSON array.  If a format
+is specified, the given template will be executed for each result.
+
+Go's `text/template <http://golang.org/pkg/text/template/>` package
+describes all the details of the format.
+
+Examples
+~~~~~~~~
+
+Get an instance's IP Address
+............................
+
+For the most part, you can pick out any field from the JSON in a
+fairly straightforward manner.
+
+.. code-block:: bash
+
+    $ sudo docker inspect -format='{{.NetworkSettings.IPAddress}}' $INSTANCE_ID
+
+List All Port Bindings
+......................
+
+One can loop over arrays and maps in the results to produce simple
+text output:
+
+.. code-block:: bash
+
+    $ sudo docker inspect -format='{{range $p, $conf := .NetworkSettings.Ports}} {{$p}} -> {{(index $conf 0).HostPort}} {{end}}' $INSTANCE_ID
+
+Find a Specific Port Mapping
+............................
+
+The ``.Field`` syntax doesn't work when the field name begins with a
+number, but the template language's ``index`` function does.  The
+``.NetworkSettings.Ports`` section contains a map of the internal port
+mappings to a list of external address/port objects, so to grab just
+the numeric public port, you use ``index`` to find the specific port
+map, and then ``index`` 0 contains first object inside of that.  Then
+we ask for the ``HostPort`` field to get the public address.
+
+.. code-block:: bash
+
+    $ sudo docker inspect -format='{{(index (index .NetworkSettings.Ports "8787/tcp") 0).HostPort}}' $INSTANCE_ID
+
 .. _cli_kill:
 
 ``kill``
@@ -704,6 +797,15 @@ Known Issues (kill)
       -notrunc=false: Don't truncate output
       -q=false: Only display numeric IDs
 
+Running ``docker ps`` showing 2 linked containers.
+
+.. code-block:: bash
+
+    $ docker ps
+    CONTAINER ID        IMAGE                        COMMAND                CREATED              STATUS              PORTS               NAMES
+    4c01db0b339c        ubuntu:12.04                 bash                   17 seconds ago       Up 16 seconds                           webapp              
+    d7886598dbe2        crosbymichael/redis:latest   /redis-server --dir    33 minutes ago       Up 33 minutes       6379/tcp            redis,webapp/db     
+
 .. _cli_pull:
 
 ``pull``
@@ -752,7 +854,7 @@ Known Issues (kill)
         -link="": Remove the link instead of the actual container
 
 Known Issues (rm)
-~~~~~~~~~~~~~~~~~~~
+~~~~~~~~~~~~~~~~~
 
 * :issue:`197` indicates that ``docker kill`` may leave directories
   behind and make it difficult to remove the container.
@@ -763,7 +865,7 @@ Examples:
 
 .. code-block:: bash
 
-    $ docker rm /redis
+    $ sudo docker rm /redis
     /redis
 
 
@@ -772,7 +874,7 @@ This will remove the container referenced under the link ``/redis``.
 
 .. code-block:: bash
 
-    $ docker rm -link /webapp/redis
+    $ sudo docker rm -link /webapp/redis
     /webapp/redis
 
 
@@ -781,7 +883,7 @@ network communication.
 
 .. code-block:: bash
 
-    $ docker rm `docker ps -a -q`
+    $ sudo docker rm `docker ps -a -q`
 
 
 This command will delete all stopped containers. The command ``docker ps -a -q`` will return all
@@ -836,12 +938,19 @@ containers will not be deleted.
       -name="": Assign the specified name to the container. If no name is specific docker will generate a random name
       -P=false: Publish all exposed ports to the host interfaces
 
-Examples
---------
+Known Issues (run -volumes-from)
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+* :issue:`2702`: "lxc-start: Permission denied - failed to mount"
+  could indicate a permissions problem with AppArmor. Please see the
+  issue for a workaround.
+
+Examples:
+~~~~~~~~~
 
 .. code-block:: bash
 
-    sudo docker run -cidfile /tmp/docker_test.cid ubuntu echo "test"
+    $ sudo docker run -cidfile /tmp/docker_test.cid ubuntu echo "test"
 
 This will create a container and print "test" to the console. The
 ``cidfile`` flag makes docker attempt to create a new file and write the
@@ -850,7 +959,10 @@ error. Docker will close this file when docker run exits.
 
 .. code-block:: bash
 
-   docker run mount -t tmpfs none /var/spool/squid
+   $ sudo docker run -t -i -rm ubuntu bash
+   root@bc338942ef20:/# mount -t tmpfs none /mnt
+   mount: permission denied
+
 
 This will *not* work, because by default, most potentially dangerous
 kernel capabilities are dropped; including ``cap_sys_admin`` (which is
@@ -859,7 +971,12 @@ allow it to run:
 
 .. code-block:: bash
 
-   docker run -privileged mount -t tmpfs none /var/spool/squid
+   $ sudo docker run -privileged ubuntu bash
+   root@50e3f57e16e6:/# mount -t tmpfs none /mnt
+   root@50e3f57e16e6:/# df -h
+   Filesystem      Size  Used Avail Use% Mounted on
+   none            1.9G     0  1.9G   0% /mnt
+
 
 The ``-privileged`` flag gives *all* capabilities to the container,
 and it also lifts all the limitations enforced by the ``device``
@@ -869,7 +986,7 @@ use-cases, like running Docker within Docker.
 
 .. code-block:: bash
 
-   docker  run -w /path/to/dir/ -i -t  ubuntu pwd
+   $ sudo docker  run -w /path/to/dir/ -i -t  ubuntu pwd
 
 The ``-w`` lets the command being executed inside directory given,
 here /path/to/dir/. If the path does not exists it is created inside the
@@ -877,7 +994,7 @@ container.
 
 .. code-block:: bash
 
-   docker  run  -v `pwd`:`pwd` -w `pwd` -i -t  ubuntu pwd
+   $ sudo docker  run  -v `pwd`:`pwd` -w `pwd` -i -t  ubuntu pwd
 
 The ``-v`` flag mounts the current working directory into the container.
 The ``-w`` lets the command being executed inside the current
@@ -887,7 +1004,7 @@ using the container, but inside the current working directory.
 
 .. code-block:: bash
 
-    docker run -p 127.0.0.1:80:8080 ubuntu bash
+   $ sudo docker run -p 127.0.0.1:80:8080 ubuntu bash
 
 This binds port ``8080`` of the container to port ``80`` on 127.0.0.1 of the
 host machine. :ref:`port_redirection` explains in detail how to manipulate ports
@@ -895,7 +1012,7 @@ in Docker.
 
 .. code-block:: bash
 
-    docker run -expose 80 ubuntu bash
+    $ sudo docker run -expose 80 ubuntu bash
 
 This exposes port ``80`` of the container for use within a link without
 publishing the port to the host system's interfaces. :ref:`port_redirection`
@@ -903,14 +1020,14 @@ explains in detail how to manipulate ports in Docker.
 
 .. code-block:: bash
 
-    docker run -name console -t -i ubuntu bash
+    $ sudo docker run -name console -t -i ubuntu bash
 
 This will create and run a new container with the container name
 being ``console``.
 
 .. code-block:: bash
 
-    docker run -link /redis:redis -name console ubuntu bash
+    $ sudo docker run -link /redis:redis -name console ubuntu bash
 
 The ``-link`` flag will link the container named ``/redis`` into the
 newly created container with the alias ``redis``.  The new container
@@ -920,7 +1037,7 @@ to the newly created container.
 
 .. code-block:: bash
 
-   docker run -volumes-from 777f7dc92da7,ba8c0c54f0f2:ro -i -t ubuntu pwd
+   $ sudo docker run -volumes-from 777f7dc92da7,ba8c0c54f0f2:ro -i -t ubuntu pwd
 
 The ``-volumes-from`` flag mounts all the defined volumes from the
 refrence containers. Containers can be specified by a comma seperated
@@ -929,16 +1046,10 @@ id may be optionally suffixed with ``:ro`` or ``:rw`` to mount the volumes in
 read-only or read-write mode, respectively. By default, the volumes are mounted
 in the same mode (rw or ro) as the reference container.
 
-Known Issues (run -volumes-from)
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-* :issue:`2702`: "lxc-start: Permission denied - failed to mount"
-  could indicate a permissions problem with AppArmor. Please see the
-  issue for a workaround.
-
 .. _cli_save:
 
 ``save``
+---------
 
 ::
 

docs/sources/contributing/devenvironment.rst

@@ -42,9 +42,7 @@ This following command will build a development environment using the Dockerfile
 
 .. code-block:: bash
 
-    sudo docker build -t docker .
-
-
+    sudo make build
 
 If the build is successful, congratulations! You have produced a clean build of docker, neatly encapsulated in a standard build environment. 
 
@@ -56,7 +54,7 @@ To create the Docker binary, run this command:
 
 .. code-block:: bash
 
-	sudo docker run -privileged -v `pwd`:/go/src/github.com/dotcloud/docker docker hack/make.sh binary
+	sudo make binary
 
 This will create the Docker binary in ``./bundles/<version>-dev/binary/``
 
@@ -68,10 +66,15 @@ To execute the test cases, run this command:
 
 .. code-block:: bash
 
-	sudo docker run -privileged -v `pwd`:/go/src/github.com/dotcloud/docker docker hack/make.sh test
+	sudo make test
 
 
-Note: if you're running the tests in vagrant, you need to specify a dns entry in the command: `-dns 8.8.8.8`
+Note: if you're running the tests in vagrant, you need to specify a dns entry in 
+the command (either edit the Makefile, or run the step manually): 
+
+.. code-block:: bash
+
+	sudo docker run -dns 8.8.8.8 -privileged -v `pwd`:/go/src/github.com/dotcloud/docker docker hack/make.sh test
 
 If the test are successful then the tail of the output should look something like this
 
@@ -113,11 +116,23 @@ You can run an interactive session in the newly built container:
 
 .. code-block:: bash
 
-	sudo docker run -privileged -i -t docker bash
+	sudo make shell
 
-	# type 'exit' to exit
+	# type 'exit' or Ctrl-D to exit
 
 
+Extra Step: Build and view the Documenation
+-------------------------------------------
+
+If you want to read the documentation from a local website, or are making changes
+to it, you can build the documentation and then serve it by:
+
+.. code-block:: bash
+
+	sudo make doc
+    # when its done, you can point your browser to http://yourdockerhost:8000
+	# type Ctrl-C to exit
+
 
 .. note:: The binary is available outside the container in the directory  ``./bundles/<version>-dev/binary/``. You can swap your host docker executable with this binary for live testing - for example, on ubuntu: ``sudo service docker stop ; sudo cp $(which docker) $(which docker)_ ; sudo cp ./bundles/<version>-dev/binary/docker-<version>-dev $(which docker);sudo service docker start``.
 

docs/sources/examples/cfengine_process_management.rst

@@ -0,0 +1,137 @@
+:title: Process Management with CFEngine
+:description: Managing containerized processes with CFEngine
+:keywords: cfengine, process, management, usage, docker, documentation
+
+Process Management with CFEngine
+================================
+
+Create Docker containers with managed processes.
+
+Docker monitors one process in each running container and the container lives or dies with that process.
+By introducing CFEngine inside Docker containers, we can alleviate a few of the issues that may arise:
+
+* It is possible to easily start multiple processes within a container, all of which will be managed automatically, with the normal ``docker run`` command.
+* If a managed process dies or crashes, CFEngine will start it again within 1 minute.
+* The container itself will live as long as the CFEngine scheduling daemon (cf-execd) lives. With CFEngine, we are able to decouple the life of the container from the uptime of the service it provides.
+
+
+How it works
+------------
+
+CFEngine, together with the cfe-docker integration policies, are installed as part of the Dockerfile. This builds CFEngine into our Docker image.
+
+The Dockerfile's ``ENTRYPOINT`` takes an arbitrary amount of commands (with any desired arguments) as parameters.
+When we run the Docker container these parameters get written to CFEngine policies and CFEngine takes over to ensure that the desired processes are running in the container.
+
+CFEngine scans the process table for the ``basename`` of the commands given to the ``ENTRYPOINT`` and runs the command to start the process if the ``basename`` is not found.
+For example, if we start the container with ``docker run "/path/to/my/application parameters"``, CFEngine will look for a process named ``application`` and run the command.
+If an entry for ``application`` is not found in the process table at any point in time, CFEngine will execute ``/path/to/my/application parameters`` to start the application once again.
+The check on the process table happens every minute.
+
+Note that it is therefore important that the command to start your application leaves a process with the basename of the command.
+This can be made more flexible by making some minor adjustments to the CFEngine policies, if desired.
+
+
+Usage
+-----
+
+This example assumes you have Docker installed and working.
+We will install and manage ``apache2`` and ``sshd`` in a single container.
+
+There are three steps:
+
+1. Install CFEngine into the container.
+2. Copy the CFEngine Docker process management policy into the containerized CFEngine installation.
+3. Start your application processes as part of the ``docker run`` command.
+
+
+Building the container image
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+The first two steps can be done as part of a Dockerfile, as follows.
+
+.. code-block:: bash
+
+    FROM ubuntu
+    MAINTAINER Eystein Måløy Stenberg <eytein.stenberg@gmail.com>
+
+    RUN apt-get -y install wget lsb-release unzip
+
+    # install latest CFEngine
+    RUN wget -qO- http://cfengine.com/pub/gpg.key | apt-key add -
+    RUN echo "deb http://cfengine.com/pub/apt $(lsb_release -cs) main" > /etc/apt/sources.list.d/cfengine-community.list
+    RUN apt-get update
+    RUN apt-get install cfengine-community
+
+    # install cfe-docker process management policy
+    RUN wget --no-check-certificate https://github.com/estenberg/cfe-docker/archive/master.zip -P /tmp/ && unzip /tmp/master.zip -d /tmp/
+    RUN cp /tmp/cfe-docker-master/cfengine/bin/* /var/cfengine/bin/
+    RUN cp /tmp/cfe-docker-master/cfengine/inputs/* /var/cfengine/inputs/
+    RUN rm -rf /tmp/cfe-docker-master /tmp/master.zip
+
+    # apache2 and openssh are just for testing purposes, install your own apps here
+    RUN apt-get -y install openssh-server apache2
+    RUN mkdir -p /var/run/sshd
+    RUN echo "root:password" | chpasswd  # need a password for ssh
+
+    ENTRYPOINT ["/var/cfengine/bin/docker_processes_run.sh"]
+
+
+By saving this file as ``Dockerfile`` to a working directory, you can then build your container with the docker build command,
+e.g. ``docker build -t managed_image``.
+
+Testing the container
+~~~~~~~~~~~~~~~~~~~~~
+
+Start the container with ``apache2`` and ``sshd`` running and managed, forwarding a port to our SSH instance:
+
+.. code-block:: bash
+
+    docker run -p 127.0.0.1:222:22 -d managed_image "/usr/sbin/sshd" "/etc/init.d/apache2 start"
+
+We now clearly see one of the benefits of the cfe-docker integration: it allows to start several processes
+as part of a normal ``docker run`` command.
+
+We can now log in to our new container and see that both ``apache2`` and ``sshd`` are running. We have set the root password to
+"password" in the Dockerfile above and can use that to log in with ssh:
+
+.. code-block:: bash
+
+    ssh -p222 root@127.0.0.1
+
+    ps -ef
+    UID        PID  PPID  C STIME TTY          TIME CMD
+    root         1     0  0 07:48 ?        00:00:00 /bin/bash /var/cfengine/bin/docker_processes_run.sh /usr/sbin/sshd /etc/init.d/apache2 start
+    root        18     1  0 07:48 ?        00:00:00 /var/cfengine/bin/cf-execd -F
+    root        20     1  0 07:48 ?        00:00:00 /usr/sbin/sshd
+    root        32     1  0 07:48 ?        00:00:00 /usr/sbin/apache2 -k start
+    www-data    34    32  0 07:48 ?        00:00:00 /usr/sbin/apache2 -k start
+    www-data    35    32  0 07:48 ?        00:00:00 /usr/sbin/apache2 -k start
+    www-data    36    32  0 07:48 ?        00:00:00 /usr/sbin/apache2 -k start
+    root        93    20  0 07:48 ?        00:00:00 sshd: root@pts/0 
+    root       105    93  0 07:48 pts/0    00:00:00 -bash
+    root       112   105  0 07:49 pts/0    00:00:00 ps -ef
+
+
+If we stop apache2, it will be started again within a minute by CFEngine.
+
+.. code-block:: bash
+
+    service apache2 status
+     Apache2 is running (pid 32).
+    service apache2 stop
+             * Stopping web server apache2 ... waiting    [ OK ]
+    service apache2 status
+     Apache2 is NOT running.
+    # ... wait up to 1 minute...
+    service apache2 status
+     Apache2 is running (pid 173).
+
+
+Adapting to your applications
+-----------------------------
+
+To make sure your applications get managed in the same manner, there are just two things you need to adjust from the above example:
+
+* In the Dockerfile used above, install your applications instead of ``apache2`` and ``sshd``.
+* When you start the container with ``docker run``, specify the command line arguments to your applications rather than ``apache2`` and ``sshd``.

docs/sources/examples/hello_world.rst

@@ -131,8 +131,6 @@ Attach to the container to see the results in real-time.
 
 - **"docker attach**" This will allow us to attach to a background
   process to see what is going on.
-- **"-sig-proxy=true"** Proxify all received signal to the process
-  (even in non-tty mode)
 - **$CONTAINER_ID** The Id of the container we want to attach too.
 
 Exit from the container attachment by pressing Control-C.

docs/sources/examples/index.rst

@@ -24,3 +24,5 @@ to more substantial services like those which you might find in production.
    postgresql_service
    mongodb
    running_riak_service
+   using_supervisord
+   cfengine_process_management

docs/sources/examples/postgresql_service.rst

@@ -7,26 +7,18 @@
 PostgreSQL Service
 ==================
 
+.. include:: example_header.inc
+
 .. note::
 
     A shorter version of `this blog post`_.
 
-.. note::
-
-    As of version 0.5.2, Docker requires root privileges to run.
-    You have to either manually adjust your system configuration (permissions on
-    /var/run/docker.sock or sudo config), or prefix `docker` with `sudo`. Check
-    `this thread`_ for details.
-
 .. _this blog post: http://zaiste.net/2013/08/docker_postgresql_how_to/
-.. _this thread: https://groups.google.com/forum/?fromgroups#!topic/docker-club/P3xDLqmLp0E
 
 Installing PostgreSQL on Docker
 -------------------------------
 
-For clarity I won't be showing command output.
-
-Run an interactive shell in Docker container.
+Run an interactive shell in a Docker container.
 
 .. code-block:: bash
 
@@ -38,26 +30,26 @@ Update its dependencies.
 
     apt-get update
 
-Install ``python-software-properties``.
+Install ``python-software-properties``, ``software-properties-common``, ``wget`` and ``vim``.
 
 .. code-block:: bash
 
-    apt-get -y install python-software-properties
-    apt-get -y install software-properties-common
+    apt-get -y install python-software-properties software-properties-common wget vim
 
-Add Pitti's PostgreSQL repository. It contains the most recent stable release
-of PostgreSQL i.e. ``9.2``.
+Add PostgreSQL's repository. It contains the most recent stable release
+of PostgreSQL, ``9.3``.
 
 .. code-block:: bash
 
-    add-apt-repository ppa:pitti/postgresql
+    wget --quiet -O - https://www.postgresql.org/media/keys/ACCC4CF8.asc | apt-key add -
+    echo "deb http://apt.postgresql.org/pub/repos/apt/ precise-pgdg main" > /etc/apt/sources.list.d/pgdg.list
     apt-get update
 
-Finally, install PostgreSQL 9.2
+Finally, install PostgreSQL 9.3
 
 .. code-block:: bash
 
-    apt-get -y install postgresql-9.2 postgresql-client-9.2 postgresql-contrib-9.2
+    apt-get -y install postgresql-9.3 postgresql-client-9.3 postgresql-contrib-9.3
 
 Now, create a PostgreSQL superuser role that can create databases and
 other roles.  Following Vagrant's convention the role will be named
@@ -76,15 +68,14 @@ role.
 
 Adjust PostgreSQL configuration so that remote connections to the
 database are possible. Make sure that inside
-``/etc/postgresql/9.2/main/pg_hba.conf`` you have following line (you will need
-to install an editor, e.g. ``apt-get install vim``):
+``/etc/postgresql/9.3/main/pg_hba.conf`` you have following line:
 
 .. code-block:: bash
 
     host    all             all             0.0.0.0/0               md5
 
-Additionaly, inside ``/etc/postgresql/9.2/main/postgresql.conf``
-uncomment ``listen_addresses`` so it is as follows:
+Additionaly, inside ``/etc/postgresql/9.3/main/postgresql.conf``
+uncomment ``listen_addresses`` like so:
 
 .. code-block:: bash
 
@@ -94,7 +85,7 @@ uncomment ``listen_addresses`` so it is as follows:
 
     This PostgreSQL setup is for development only purposes. Refer
     to PostgreSQL documentation how to fine-tune these settings so that it
-    is enough secure.
+    is secure enough.
 
 Exit.
 
@@ -102,43 +93,43 @@ Exit.
 
     exit
 
-Create an image and assign it a name. ``<container_id>`` is in the
-Bash prompt; you can also locate it using ``docker ps -a``.
+Create an image from our container and assign it a name. The ``<container_id>``
+is in the Bash prompt; you can also locate it using ``docker ps -a``.
 
 .. code-block:: bash
 
     sudo docker commit <container_id> <your username>/postgresql
 
-Finally, run PostgreSQL server via ``docker``.
+Finally, run the PostgreSQL server via ``docker``.
 
 .. code-block:: bash
 
     CONTAINER=$(sudo docker run -d -p 5432 \
       -t <your username>/postgresql \
-      /bin/su postgres -c '/usr/lib/postgresql/9.2/bin/postgres \
-        -D /var/lib/postgresql/9.2/main \
-        -c config_file=/etc/postgresql/9.2/main/postgresql.conf')
+      /bin/su postgres -c '/usr/lib/postgresql/9.3/bin/postgres \
+        -D /var/lib/postgresql/9.3/main \
+        -c config_file=/etc/postgresql/9.3/main/postgresql.conf')
 
-Connect the PostgreSQL server using ``psql`` (You will need postgres installed
-on the machine.  For ubuntu, use something like
-``sudo apt-get install postgresql``).
+Connect the PostgreSQL server using ``psql`` (You will need the
+postgresql client installed on the machine.  For ubuntu, use something
+like ``sudo apt-get install postgresql-client``).
 
 .. code-block:: bash
 
-    CONTAINER_IP=$(sudo docker inspect $CONTAINER | grep IPAddress | awk '{ print $2 }' | tr -d ',"')
+    CONTAINER_IP=$(sudo docker inspect -format='{{.NetworkSettings.IPAddress}}' $CONTAINER)
     psql -h $CONTAINER_IP -p 5432 -d docker -U docker -W
 
 As before, create roles or databases if needed.
 
 .. code-block:: bash
 
-    psql (9.2.4)
+    psql (9.3.1)
     Type "help" for help.
 
     docker=# CREATE DATABASE foo OWNER=docker;
     CREATE DATABASE
 
-Additionally, publish your newly created image on Docker Index.
+Additionally, publish your newly created image on the Docker Index.
 
 .. code-block:: bash
 
@@ -160,9 +151,9 @@ container starts.
 .. code-block:: bash
 
     sudo docker commit -run='{"Cmd": \
-      ["/bin/su", "postgres", "-c", "/usr/lib/postgresql/9.2/bin/postgres -D \
-      /var/lib/postgresql/9.2/main -c \
-      config_file=/etc/postgresql/9.2/main/postgresql.conf"], "PortSpecs": ["5432"]}' \
+      ["/bin/su", "postgres", "-c", "/usr/lib/postgresql/9.3/bin/postgres -D \
+      /var/lib/postgresql/9.3/main -c \
+      config_file=/etc/postgresql/9.3/main/postgresql.conf"], "PortSpecs": ["5432"]}' \
       <container_id> <your username>/postgresql
 
 From now on, just type ``docker run <your username>/postgresql`` and

docs/sources/examples/using_supervisord.rst

@@ -0,0 +1,128 @@
+:title: Using Supervisor with Docker
+:description: How to use Supervisor process management with Docker
+:keywords: docker, supervisor, process management
+
+.. _using_supervisord:
+
+Using Supervisor with Docker
+============================
+
+.. include:: example_header.inc
+
+Traditionally a Docker container runs a single process when it is launched, for
+example an Apache daemon or a SSH server daemon. Often though you want to run
+more than one process in a container. There are a number of ways you can
+achieve this ranging from using a simple Bash script as the value of your
+container's ``CMD`` instruction to installing a process management tool.
+
+In this example we're going to make use of the process management tool,
+`Supervisor <http://supervisord.org/>`_, to manage multiple processes in our
+container. Using Supervisor allows us to better control, manage, and restart the
+processes we want to run. To demonstrate this we're going to install and manage both an
+SSH daemon and an Apache daemon.
+
+Creating a Dockerfile
+---------------------
+
+Let's start by creating a basic ``Dockerfile`` for our new image.
+
+.. code-block:: bash
+
+    FROM ubuntu:latest
+    MAINTAINER examples@docker.io
+    RUN echo "deb http://archive.ubuntu.com/ubuntu precise main universe" > /etc/apt/sources.list
+    RUN apt-get update
+    RUN apt-get upgrade -y
+
+Installing Supervisor
+---------------------
+
+We can now install our SSH and Apache daemons as well as Supervisor in our container.
+
+.. code-block:: bash
+
+    RUN apt-get install -y openssh-server apache2 supervisor
+    RUN mkdir -p /var/run/sshd
+    RUN mkdir -p /var/log/supervisor
+
+Here we're installing the ``openssh-server``, ``apache2`` and ``supervisor``
+(which provides the Supervisor daemon) packages. We're also creating two new
+directories that are needed to run our SSH daemon and Supervisor.
+
+Adding Supervisor's configuration file
+--------------------------------------
+
+Now let's add a configuration file for Supervisor. The default file is called
+``supervisord.conf`` and is located in ``/etc/supervisor/conf.d/``.
+
+.. code-block:: bash
+
+    ADD supervisord.conf /etc/supervisor/conf.d/supervisord.conf
+
+Let's see what is inside our ``supervisord.conf`` file.
+
+.. code-block:: bash
+
+    [supervisord]
+    nodaemon=true
+
+    [program:sshd]
+    command=/usr/sbin/sshd -D
+
+    [program:apache2]
+    command=/bin/bash -c "source /etc/apache2/envvars && /usr/sbin/apache2 -DFOREGROUND"
+
+The ``supervisord.conf`` configuration file contains directives that configure
+Supervisor and the processes it manages. The first block ``[supervisord]``
+provides configuration for Supervisor itself. We're using one directive,
+``nodaemon`` which tells Supervisor to run interactively rather than daemonize.
+
+The next two blocks manage the services we wish to control. Each block controls
+a separate process. The blocks contain a single directive, ``command``, which
+specifies what command to run to start each process.
+
+Exposing ports and running Supervisor
+-------------------------------------
+
+Now let's finish our ``Dockerfile`` by exposing some required ports and
+specifying the ``CMD`` instruction to start Supervisor when our container
+launches.
+
+.. code-block:: bash
+
+    EXPOSE 22 80
+    CMD ["/usr/bin/supervisord"]
+
+Here we've exposed ports 22 and 80 on the container and we're running the
+``/usr/bin/supervisord`` binary when the container launches.
+
+Building our container
+----------------------
+
+We can now build our new container.
+
+.. code-block:: bash
+
+    sudo docker build -t <yourname>/supervisord .
+
+Running our Supervisor container
+--------------------------------
+
+Once we've got a built image we can launch a container from it.
+
+.. code-block:: bash
+
+    sudo docker run -p 22 -p 80 -t -i <yourname>/supervisor
+    2013-11-25 18:53:22,312 CRIT Supervisor running as root (no user in config file)
+    2013-11-25 18:53:22,312 WARN Included extra file "/etc/supervisor/conf.d/supervisord.conf" during parsing
+    2013-11-25 18:53:22,342 INFO supervisord started with pid 1
+    2013-11-25 18:53:23,346 INFO spawned: 'sshd' with pid 6
+    2013-11-25 18:53:23,349 INFO spawned: 'apache2' with pid 7
+    . . .
+
+We've launched a new container interactively using the ``docker run`` command.
+That container has run Supervisor and launched the SSH and Apache daemons with
+it. We've specified the ``-p`` flag to expose ports 22 and 80. From here we can
+now identify the exposed ports and connect to one or both of the SSH and Apache
+daemons.
+

docs/sources/installation/amazon.rst

@@ -22,20 +22,37 @@ Amazon QuickStart
 
 1. **Choose an image:**
 
-   * Launch the `Create Instance Wizard` <https://console.aws.amazon.com/ec2/v2/home?#LaunchInstanceWizard:> menu on your AWS Console
-   * Select "Community AMIs" option and serch for ``amd64 precise`` (click enter to search)
-   * If you choose a EBS enabled AMI you will be able to launch a `t1.micro` instance (more info on `pricing` <http://aws.amazon.com/en/ec2/pricing/> )
-   * When you click select you'll be taken to the instance setup, and you're one click away from having your Ubuntu VM up and running.
+   * Launch the `Create Instance Wizard
+     <https://console.aws.amazon.com/ec2/v2/home?#LaunchInstanceWizard:>`_ menu
+     on your AWS Console.
+
+   * When picking the source AMI for your instance type, select "Community
+     AMIs".
+
+   * Search for ``amd64 precise``. Pick one of the amd64 Ubuntu images.
+
+   * If you choose a EBS enabled AMI, you'll also be able to launch a
+     ``t1.micro`` instance (more info on `pricing
+     <http://aws.amazon.com/en/ec2/pricing/>`_).  ``t1.micro`` instances are
+     eligible for Amazon's Free Usage Tier.
+
+   * When you click select you'll be taken to the instance setup, and you're one
+     click away from having your Ubuntu VM up and running.
 
 2. **Tell CloudInit to install Docker:**
 
-   * Enter ``#include https://get.docker.io`` into the instance *User
-     Data*. `CloudInit <https://help.ubuntu.com/community/CloudInit>`_
-     is part of the Ubuntu image you chose and it bootstraps from this
-     *User Data*.
+   * When you're on the "Configure Instance Details" step, expand the "Advanced
+     Details" section.
 
-3. After a few more standard choices where defaults are probably ok, your
-   AWS Ubuntu instance with Docker should be running!
+   * Under "User data", select "As text".
+
+   * Enter ``#include https://get.docker.io`` into the instance *User Data*.
+     `CloudInit <https://help.ubuntu.com/community/CloudInit>`_ is part of the
+     Ubuntu image you chose; it will bootstrap Docker by running the shell
+     script located at this URL.
+
+3. After a few more standard choices where defaults are probably ok, your AWS
+   Ubuntu instance with Docker should be running!
 
 **If this is your first AWS instance, you may need to set up your
 Security Group to allow SSH.** By default all incoming ports to your
@@ -152,7 +169,7 @@ Docker that way too. Vagrant 1.1 or higher is required.
    includes rights to SSH (port 22) to your container.
 
    If you have an advanced AWS setup, you might want to have a look at
-   https://github.com/mitchellh/vagrant-aws
+   `vagrant-aws <https://github.com/mitchellh/vagrant-aws>`_.
 
 7. Connect to your machine
 

docs/sources/installation/archlinux.rst

@@ -1,5 +1,5 @@
 :title: Installation on Arch Linux
-:description: Docker installation on Arch Linux. 
+:description: Docker installation on Arch Linux.
 :keywords: arch linux, virtualization, docker, documentation, installation
 
 .. _arch_linux:
@@ -7,54 +7,49 @@
 Arch Linux
 ==========
 
+.. include:: install_header.inc
+
+.. include:: install_unofficial.inc
+
 Installing on Arch Linux is not officially supported but can be handled via 
-either of the following AUR packages:
+one of the following AUR packages:
 
 * `lxc-docker <https://aur.archlinux.org/packages/lxc-docker/>`_
 * `lxc-docker-git <https://aur.archlinux.org/packages/lxc-docker-git/>`_
+* `lxc-docker-nightly <https://aur.archlinux.org/packages/lxc-docker-nightly/>`_
 
 The lxc-docker package will install the latest tagged version of docker. 
 The lxc-docker-git package will build from the current master branch.
+The lxc-docker-nightly package will install the latest build.
 
 Dependencies
 ------------
 
 Docker depends on several packages which are specified as dependencies in
-either AUR package.
+the AUR packages. The core dependencies are:
 
-* aufs3
 * bridge-utils
-* go
+* device-mapper
 * iproute2
-* linux-aufs_friendly
 * lxc
 
+
 Installation
 ------------
 
-.. include:: install_header.inc
-
-.. include:: install_unofficial.inc
-
 The instructions here assume **yaourt** is installed.  See 
 `Arch User Repository <https://wiki.archlinux.org/index.php/Arch_User_Repository#Installing_packages>`_
 for information on building and installing packages from the AUR if you have not
 done so before.
 
-Keep in mind that if **linux-aufs_friendly** is not already installed that a
-new kernel will be compiled and this can take quite a while.
-
 ::
 
-    yaourt -S lxc-docker-git
+    yaourt -S lxc-docker
 
 
 Starting Docker
 ---------------
 
-Prior to starting docker modify your bootloader to use the 
-**linux-aufs_friendly** kernel and reboot your system.
-
 There is a systemd service unit created for docker.  To start the docker service:
 
 ::

docs/sources/installation/binaries.rst

@@ -12,17 +12,9 @@ Binaries
 **This instruction set is meant for hackers who want to try out Docker
 on a variety of environments.**
 
-Right now, the officially supported distributions are:
-
-- :ref:`ubuntu_precise`
-- :ref:`ubuntu_raring`
-
-
-But we know people have had success running it under
-
-- Debian
-- Suse
-- :ref:`arch_linux`
+Before following these directions, you should really check if a packaged version
+of Docker is already available for your distribution.  We have packages for many
+distributions, and more keep showing up all the time!
 
 Check Your Kernel
 -----------------
@@ -34,7 +26,7 @@ Get the docker binary:
 
 .. code-block:: bash
 
-    wget --output-document=docker https://get.docker.io/builds/Linux/x86_64/docker-latest
+    wget https://get.docker.io/builds/Linux/x86_64/docker-latest -O docker
     chmod +x docker
 
 

docs/sources/installation/fedora.rst

@@ -0,0 +1,52 @@
+:title: Requirements and Installation on Fedora
+:description: Please note this project is currently under heavy development. It should not be used in production.
+:keywords: Docker, Docker documentation, requirements, virtualbox, vagrant, git, ssh, putty, cygwin, linux
+
+.. _fedora:
+
+Fedora
+======
+
+.. include:: install_header.inc
+
+.. include:: install_unofficial.inc
+
+Docker is available in **Fedora 19 and later**. Please note that due to the
+current Docker limitations Docker is able to run only on the **64 bit**
+architecture.
+
+Installation
+------------
+
+Firstly, let's make sure our Fedora host is up-to-date.
+
+.. code-block:: bash
+
+    sudo yum -y upgrade
+
+Next let's install the ``docker-io`` package which will install Docker on our host.
+
+.. code-block:: bash
+
+   sudo yum -y install docker-io
+
+Now it's installed lets start the Docker daemon.
+
+.. code-block:: bash
+
+    sudo systemctl start docker
+
+If we want Docker to start at boot we should also:
+
+.. code-block:: bash
+
+   sudo systemctl enable docker
+
+Now let's verify that Docker is working.
+
+.. code-block:: bash
+
+   sudo docker run -i -t ubuntu /bin/bash
+
+**Done!**, now continue with the :ref:`hello_world` example.
+

docs/sources/installation/gentoolinux.rst

@@ -4,8 +4,8 @@
 
 .. _gentoo_linux:
 
-Gentoo Linux
-============
+Gentoo
+======
 
 .. include:: install_header.inc
 
@@ -22,17 +22,19 @@ provided at https://github.com/tianon/docker-overlay which can be added using
 properly installing and using the overlay can be found in `the overlay README
 <https://github.com/tianon/docker-overlay/blob/master/README.md#using-this-overlay>`_.
 
+Note that sometimes there is a disparity between the latest version and what's
+in the overlay, and between the latest version in the overlay and what's in the
+portage tree.  Please be patient, and the latest version should propagate
+shortly.
+
 Installation
 ^^^^^^^^^^^^
 
 The package should properly pull in all the necessary dependencies and prompt
-for all necessary kernel options.  For the most straightforward installation
-experience, use ``sys-kernel/aufs-sources`` as your kernel sources.  If you
-prefer not to use ``sys-kernel/aufs-sources``, the portage tree also contains
-``sys-fs/aufs3``, which includes the patches necessary for adding AUFS support
-to other kernel source packages such as ``sys-kernel/gentoo-sources`` (and a
-``kernel-patch`` USE flag to perform the patching to ``/usr/src/linux``
-automatically).
+for all necessary kernel options.  The ebuilds for 0.7+ include use flags to
+pull in the proper dependencies of the major storage drivers, with the
+"device-mapper" use flag being enabled by default, since that is the simplest
+installation path.
 
 .. code-block:: bash
 
@@ -47,9 +49,9 @@ the #docker IRC channel on the freenode network.
 Starting Docker
 ^^^^^^^^^^^^^^^
 
-Ensure that you are running a kernel that includes the necessary AUFS
-patches/support and includes all the necessary modules and/or configuration for
-LXC.
+Ensure that you are running a kernel that includes all the necessary modules
+and/or configuration for LXC (and optionally for device-mapper and/or AUFS,
+depending on the storage driver you've decided to use).
 
 OpenRC
 ------

docs/sources/installation/google.rst

@@ -0,0 +1,65 @@
+:title: Installation on Google Cloud Platform
+:description: Please note this project is currently under heavy development. It should not be used in production.
+:keywords: Docker, Docker documentation, installation, google, Google Compute Engine, Google Cloud Platform
+
+`Google Cloud Platform <https://cloud.google.com/>`_
+====================================================
+
+.. include:: install_header.inc
+
+.. _googlequickstart:
+
+`Compute Engine <https://developers.google.com/compute>`_ QuickStart for `Debian <https://www.debian.org>`_
+-----------------------------------------------------------------------------------------------------------
+
+1. Go to `Google Cloud Console <https://cloud.google.com/console>`_ and create a new Cloud Project with billing enabled.
+
+2. Download and configure the `Google Cloud SDK <https://developers.google.com/cloud/sdk/>`_ to use your project with the following commands:
+
+.. code-block:: bash
+
+    $ curl https://dl.google.com/dl/cloudsdk/release/install_google_cloud_sdk.bash | bash
+    $ gcloud auth login
+    Enter a cloud project id (or leave blank to not set): <google-cloud-project-id>
+
+3. Start a new instance, select a zone close to you and the desired instance size:
+
+.. code-block:: bash
+
+    $ gcutil addinstance docker-playground --image=backports-debian-7
+    1: europe-west1-a
+    ...
+    4: us-central1-b
+    >>> <zone-index>
+    1: machineTypes/n1-standard-1
+    ...
+    12: machineTypes/g1-small
+    >>> <machine-type-index>
+
+4. Connect to the instance using SSH:
+
+.. code-block:: bash
+
+    $ gcutil ssh docker-playground
+    docker-playground:~$ 
+
+5. Enable IP forwarding:
+
+.. code-block:: bash
+
+    docker-playground:~$ echo net.ipv4.ip_forward=1 | sudo tee /etc/sysctl.d/99-docker.conf 
+    docker-playground:~$ sudo sysctl --system
+
+6. Install the latest Docker release and configure it to start when the instance boots:
+
+.. code-block:: bash
+
+    docker-playground:~$ curl get.docker.io | bash
+    docker-playground:~$ sudo update-rc.d docker defaults
+
+7. Start a new container:
+
+.. code-block:: bash
+
+    docker-playground:~$ sudo docker run busybox echo 'docker on GCE \o/'
+    docker on GCE \o/

docs/sources/installation/index.rst

@@ -9,7 +9,7 @@ Installation
 
 There are a number of ways to install Docker, depending on where you
 want to run the daemon. The :ref:`ubuntu_linux` installation is the
-officially-tested version, and the community adds more techniques for
+officially-tested version. The community adds more techniques for
 installing Docker all the time.
 
 Contents:
@@ -18,13 +18,16 @@ Contents:
    :maxdepth: 1
 
    ubuntulinux
-   binaries
-   security
-   upgrading
-   kernel
+   rhel
+   fedora
+   archlinux
+   gentoolinux
    vagrant
    windows
    amazon
    rackspace
-   archlinux
-   gentoolinux
+   google
+   kernel
+   binaries
+   security
+   upgrading

docs/sources/installation/kernel.rst

@@ -11,10 +11,10 @@ In short, Docker has the following kernel requirements:
 
 - Linux version 3.8 or above.
 
-- `AUFS support <http://aufs.sourceforge.net/>`_.
-
 - Cgroups and namespaces must be enabled.
 
+*Note: as of 0.7 docker no longer requires aufs. AUFS support is still available as an optional driver.*
+
 The officially supported kernel is the one recommended by the
 :ref:`ubuntu_linux` installation path. It is the one that most developers
 will use, and the one that receives the most attention from the core
@@ -58,17 +58,6 @@ detects something older than 3.8.
 See issue `#407 <https://github.com/dotcloud/docker/issues/407>`_ for details.
 
 
-AUFS support
-------------
-
-Docker currently relies on AUFS, an unioning filesystem.
-While AUFS is included in the kernels built by the Debian and Ubuntu
-distributions, is not part of the standard kernel. This means that if
-you decide to roll your own kernel, you will have to patch your
-kernel tree to add AUFS. The process is documented on
-`AUFS webpage <http://aufs.sourceforge.net/>`_.
-
-
 Cgroups and namespaces
 ----------------------
 

docs/sources/installation/rackspace.rst

@@ -2,7 +2,6 @@
 :description: Installing Docker on Ubuntu proviced by Rackspace
 :keywords: Rackspace Cloud, installation, docker, linux, ubuntu
 
-===============
 Rackspace Cloud
 ===============
 
@@ -14,14 +13,14 @@ straightforward, and you should mostly be able to follow the
 
 **However, there is one caveat:**
 
-If you are using any linux not already shipping with the 3.8 kernel
+If you are using any Linux not already shipping with the 3.8 kernel
 you will need to install it. And this is a little more difficult on
 Rackspace.
 
 Rackspace boots their servers using grub's ``menu.lst`` and does not
-like non 'virtual' packages (e.g. xen compatible) kernels there,
-although they do work. This makes ``update-grub`` to not have the
-expected result, and you need to set the kernel manually.
+like non 'virtual' packages (e.g. Xen compatible) kernels there,
+although they do work. This results in ``update-grub`` not having the
+expected result, and you will need to set the kernel manually.
 
 **Do not attempt this on a production machine!**
 
@@ -34,7 +33,7 @@ expected result, and you need to set the kernel manually.
     apt-get install linux-generic-lts-raring
 
 
-Great, now you have kernel installed in ``/boot/``, next is to make it
+Great, now you have the kernel installed in ``/boot/``, next you need to make it
 boot next time.
 
 .. code-block:: bash
@@ -48,9 +47,9 @@ boot next time.
 Now you need to manually edit ``/boot/grub/menu.lst``, you will find a
 section at the bottom with the existing options.  Copy the top one and
 substitute the new kernel into that. Make sure the new kernel is on
-top, and double check kernel and initrd point to the right files.
+top, and double check the kernel and initrd lines point to the right files.
 
-Make special care to double check the kernel and initrd entries.
+Take special care to double check the kernel and initrd entries.
 
 .. code-block:: bash
 
@@ -79,7 +78,7 @@ It will probably look something like this:
      initrd		/boot/initrd.img-3.2.0-38-virtual
 
 
-Reboot server (either via command line or console)
+Reboot the server (either via command line or console)
 
 .. code-block:: bash
 

docs/sources/installation/rhel.rst

@@ -0,0 +1,65 @@
+:title: Requirements and Installation on Red Hat Enterprise Linux / CentOS
+:description: Please note this project is currently under heavy development. It should not be used in production.
+:keywords: Docker, Docker documentation, requirements, linux, rhel, centos
+
+.. _rhel:
+
+Red Hat Enterprise Linux / CentOS
+=================================
+
+.. include:: install_header.inc
+
+.. include:: install_unofficial.inc
+
+Docker is available for **RHEL/CentOS 6**.
+
+Please note that this package is part of a `Extra Packages for Enterprise Linux (EPEL)`_, a community effort to create and maintain additional packages for RHEL distribution.
+
+Please note that due to the current Docker limitations Docker is able to run only on the **64 bit** architecture.
+
+Installation
+------------
+
+1. Firstly, let's make sure our RHEL host is up-to-date.
+
+.. code-block:: bash
+
+    sudo yum -y upgrade
+
+2. Next you need to install the EPEL repository. Please follow the `EPEL installation instructions`_.
+
+3. Next let's install the ``docker-io`` package which will install Docker on our host.
+
+.. code-block:: bash
+
+   sudo yum -y install docker-io
+
+4. Now it's installed lets start the Docker daemon.
+
+.. code-block:: bash
+
+    sudo service docker start
+
+If we want Docker to start at boot we should also:
+
+.. code-block:: bash
+
+   sudo chkconfig docker on
+
+5. Now let's verify that Docker is working.
+
+.. code-block:: bash
+
+   sudo docker run -i -t ubuntu /bin/bash
+
+**Done!**, now continue with the :ref:`hello_world` example.
+
+Issues?
+-------
+
+If you have any issues - please report them directly in the `Red Hat Bugzilla for docker-io component`_.
+
+.. _Extra Packages for Enterprise Linux (EPEL): https://fedoraproject.org/wiki/EPEL
+.. _EPEL installation instructions: https://fedoraproject.org/wiki/EPEL#How_can_I_use_these_extra_packages.3F
+.. _Red Hat Bugzilla for docker-io component : https://bugzilla.redhat.com/enter_bug.cgi?product=Fedora%20EPEL&component=docker-io
+

docs/sources/installation/ubuntulinux.rst

@@ -4,8 +4,8 @@
 
 .. _ubuntu_linux:
 
-Ubuntu Linux
-============
+Ubuntu
+======
 
 .. warning::
 
@@ -14,16 +14,11 @@ Ubuntu Linux
 
 .. include:: install_header.inc
 
-Right now, the officially supported distribution are:
+Docker is supported on the following versions of Ubuntu:
 
 - :ref:`ubuntu_precise`
 - :ref:`ubuntu_raring`
 
-Docker has the following dependencies
-
-* Linux kernel 3.8 (read more about :ref:`kernel`)
-* AUFS file system support (we are working on BTRFS support as an alternative)
-
 Please read :ref:`ufw`, if you plan to use `UFW (Uncomplicated
 Firewall) <https://help.ubuntu.com/community/UFW>`_
 
@@ -70,32 +65,43 @@ Installation
 
 Docker is available as a Debian package, which makes installation easy.
 
+First add the Docker repository key to your local keychain. You can use the
+``apt-key`` command to check the fingerprint matches: ``36A1 D786 9245 C895 0F96
+6E92 D857 6A8B A88D 21E9``
 
 .. code-block:: bash
 
-   # Add the Docker repository key to your local keychain
-   # using apt-key finger you can check the fingerprint matches 36A1 D786 9245 C895 0F96 6E92 D857 6A8B A88D 21E9
    sudo sh -c "wget -qO- https://get.docker.io/gpg | apt-key add -"
 
-   # Add the Docker repository to your apt sources list.
-   sudo sh -c "echo deb http://get.docker.io/ubuntu docker main\
-   > /etc/apt/sources.list.d/docker.list"
+Add the Docker repository to your apt sources list, update and install the
+``lxc-docker`` package. 
 
-   # Update your sources
-   sudo apt-get update
-
-   # Install, you will see another warning that the package cannot be authenticated. Confirm install.
-   sudo apt-get install lxc-docker
-
-Verify it worked
+*You may receive a warning that the package isn't trusted. Answer yes to
+continue installation.*
+
+.. code-block:: bash
+
+   sudo sh -c "echo deb http://get.docker.io/ubuntu docker main\
+   > /etc/apt/sources.list.d/docker.list"
+   sudo apt-get update
+   sudo apt-get install lxc-docker
+
+.. note::
+
+    There is also a simple ``curl`` script available to help with this process.
+
+    .. code-block:: bash
+
+        curl -s http://get.docker.io/ubuntu/ | sudo sh
+
+Now verify that the installation has worked by downloading the ``ubuntu`` image
+and launching a container.
 
 .. code-block:: bash
 
-   # download the base 'ubuntu' container and run bash inside it while setting up an interactive shell
    sudo docker run -i -t ubuntu /bin/bash
 
-   # type 'exit' to exit
-
+Type ``exit`` to exit
 
 **Done!**, now continue with the :ref:`hello_world` example.
 
@@ -107,10 +113,13 @@ Ubuntu Raring 13.04 (64 bit)
 Dependencies
 ------------
 
-**AUFS filesystem support**
+**Optional AUFS filesystem support**
 
 Ubuntu Raring already comes with the 3.8 kernel, so we don't need to install it. However, not all systems
-have AUFS filesystem support enabled, so we need to install it.
+have AUFS filesystem support enabled. AUFS support is optional as of version 0.7, but it's still available as
+a driver and we recommend using it if you can.
+
+To make sure AUFS is installed, run the following commands:
 
 .. code-block:: bash
 
@@ -123,36 +132,37 @@ Installation
 
 Docker is available as a Debian package, which makes installation easy.
 
-*Please note that these instructions have changed for 0.6. If you are upgrading from an earlier version, you will need
-to follow them again.*
+.. warning::
+
+    Please note that these instructions have changed for 0.6. If you are upgrading from an earlier version, you will need
+    to follow them again.
+
+First add the Docker repository key to your local keychain. You can use the
+``apt-key`` command to check the fingerprint matches: ``36A1 D786 9245 C895 0F96
+6E92 D857 6A8B A88D 21E9``
 
 .. code-block:: bash
 
-   # Add the Docker repository key to your local keychain
-   # using apt-key finger you can check the fingerprint matches 36A1 D786 9245 C895 0F96 6E92 D857 6A8B A88D 21E9
    sudo sh -c "wget -qO- https://get.docker.io/gpg | apt-key add -"
 
-   # Add the Docker repository to your apt sources list.
-   sudo sh -c "echo deb http://get.docker.io/ubuntu docker main\
-   > /etc/apt/sources.list.d/docker.list"
-
-   # update
-   sudo apt-get update
-
-   # install
-   sudo apt-get install lxc-docker
-
-
-Verify it worked
+Add the Docker repository to your apt sources list, update and install the
+``lxc-docker`` package.
+
+.. code-block:: bash
+
+   sudo sh -c "echo deb http://get.docker.io/ubuntu docker main\
+   > /etc/apt/sources.list.d/docker.list"
+   sudo apt-get update
+   sudo apt-get install lxc-docker
+
+Now verify that the installation has worked by downloading the ``ubuntu`` image
+and launching a container.
 
 .. code-block:: bash
 
-   # download the base 'ubuntu' container
-   # and run bash inside it while setting up an interactive shell
    sudo docker run -i -t ubuntu /bin/bash
 
-   # type exit to exit
-
+Type ``exit`` to exit
 
 **Done!**, now continue with the :ref:`hello_world` example.
 
@@ -162,8 +172,8 @@ Verify it worked
 Docker and UFW
 ^^^^^^^^^^^^^^
 
-Docker uses a bridge to manage container networking. By default, UFW
-drops all `forwarding`, thus a first step is to enable UFW forwarding:
+Docker uses a bridge to manage container networking. By default, UFW drops all
+`forwarding` traffic. As a result will you need to enable UFW forwarding:
 
 .. code-block:: bash
 
@@ -181,9 +191,9 @@ Then reload UFW:
    sudo ufw reload
 
 
-UFW's default set of rules denied all `incoming`, so if you want to be
-able to reach your containers from another host, you should allow
-incoming connections on the docker port (default 4243):
+UFW's default set of rules denies all `incoming` traffic. If you want to be
+able to reach your containers from another host then you should allow
+incoming connections on the Docker port (default 4243):
 
 .. code-block:: bash
 

docs/sources/use/ambassador_pattern_linking.rst

@@ -0,0 +1,182 @@
+:title: Ambassador pattern linking
+:description: Using the Ambassador pattern to abstract (network) services
+:keywords: Examples, Usage, links, docker, documentation, examples, names, name, container naming
+
+.. _ambassador_pattern_linking:
+
+Ambassador pattern linking
+==========================
+
+Rather than hardcoding network links between a service consumer and provider, Docker
+encourages service portability.
+
+eg, instead of
+
+.. code-block:: bash
+
+	(consumer) --> (redis)
+
+requiring you to restart the ``consumer`` to attach it to a different ``redis`` service, 
+you can add ambassadors
+
+.. code-block:: bash
+
+	(consumer) --> (redis-ambassador) --> (redis)
+
+	or
+
+	(consumer) --> (redis-ambassador) ---network---> (redis-ambassador) --> (redis)
+
+When you need to rewire your consumer to talk to a different resdis server, you 
+can just restart the ``redis-ambassador`` container that the consumer is connected to.
+
+This pattern also allows you to transparently move the redis server to a different
+docker host from the consumer.
+
+Using the ``svendowideit/ambassador`` container, the link wiring is controlled entirely 
+from the ``docker run`` parameters.
+
+Two host Example
+----------------
+
+Start actual redis server on one Docker host
+
+.. code-block:: bash
+
+	big-server $ docker run -d -name redis crosbymichael/redis
+
+Then add an ambassador linked to the redis server, mapping a port to the outside world
+
+.. code-block:: bash
+
+	big-server $ docker run -d -link redis:redis -name redis_ambassador -p 6379:6379 svendowideit/ambassador
+
+On the other host, you can set up another ambassador setting environment variables for each remote port we want to proxy to the ``big-server``
+
+.. code-block:: bash
+
+	client-server $ docker run -d -name redis_ambassador -expose 6379 -e REDIS_PORT_6379_TCP=tcp://192.168.1.52:6379 svendowideit/ambassador
+
+Then on the ``client-server`` host, you can use a redis client container to talk 
+to the remote redis server, just by linking to the local redis ambassador.
+
+.. code-block:: bash
+
+	client-server $ docker run -i -t -rm -link redis_ambassador:redis relateiq/redis-cli
+	redis 172.17.0.160:6379> ping
+	PONG
+
+
+
+How it works
+------------
+
+The following example shows what the ``svendowideit/ambassador`` container does 
+automatically (with a tiny amount of ``sed``)
+
+On the docker host (192.168.1.52) that redis will run on:
+
+.. code-block:: bash
+
+	# start actual redis server
+	$ docker run -d -name redis crosbymichael/redis
+
+	# get a redis-cli container for connection testing	
+	$ docker pull relateiq/redis-cli
+
+	# test the redis server by talking to it directly
+	$ docker run -t -i -rm -link redis:redis relateiq/redis-cli
+	redis 172.17.0.136:6379> ping
+	PONG
+	^D
+	
+	# add redis ambassador
+	$ docker run -t -i -link redis:redis -name redis_ambassador -p 6379:6379 busybox sh
+	
+in the redis_ambassador container, you can see the linked redis containers's env
+
+.. code-block:: bash
+
+	$ env
+	REDIS_PORT=tcp://172.17.0.136:6379
+	REDIS_PORT_6379_TCP_ADDR=172.17.0.136
+	REDIS_NAME=/redis_ambassador/redis
+	HOSTNAME=19d7adf4705e
+	REDIS_PORT_6379_TCP_PORT=6379
+	HOME=/
+	REDIS_PORT_6379_TCP_PROTO=tcp
+	container=lxc
+	REDIS_PORT_6379_TCP=tcp://172.17.0.136:6379
+	TERM=xterm
+	PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
+	PWD=/
+	
+	
+This environment is used by the ambassador socat script to expose redis to the world 
+(via the -p 6379:6379 port mapping)
+
+.. code-block:: bash
+
+	$ docker rm redis_ambassador
+	$ sudo ./contrib/mkimage-unittest.sh
+	$ docker run -t -i -link redis:redis -name redis_ambassador -p 6379:6379 docker-ut sh
+	
+	$ socat TCP4-LISTEN:6379,fork,reuseaddr TCP4:172.17.0.136:6379
+	
+then ping the redis server via the ambassador
+
+.. code-block::bash
+
+	$ docker run -i -t -rm -link redis_ambassador:redis relateiq/redis-cli
+	redis 172.17.0.160:6379> ping
+	PONG
+
+Now goto a different server
+
+.. code-block:: bash
+
+	$ sudo ./contrib/mkimage-unittest.sh
+	$ docker run -t -i  -expose 6379 -name redis_ambassador docker-ut sh
+	
+	$ socat TCP4-LISTEN:6379,fork,reuseaddr TCP4:192.168.1.52:6379
+
+and get the redis-cli image so we can talk over the ambassador bridge
+
+.. code-block:: bash
+
+	$ docker pull relateiq/redis-cli
+	$ docker run -i -t -rm -link redis_ambassador:redis relateiq/redis-cli
+	redis 172.17.0.160:6379> ping
+	PONG
+
+The svendowideit/ambassador Dockerfile
+--------------------------------------
+
+The ``svendowideit/ambassador`` image is a small busybox image with ``socat`` built in.
+When you start the container, it uses a small ``sed`` script to parse out the (possibly multiple)
+link environment variables to set up the port forwarding. On the remote host, you need to set the 
+variable using the ``-e`` command line option.
+
+``-expose 1234 -e REDIS_PORT_1234_TCP=tcp://192.168.1.52:6379`` will forward the 
+local ``1234`` port to the remote IP and port - in this case ``192.168.1.52:6379``.
+
+
+.. code-block:: Dockerfile
+
+	#
+	#
+	# first you need to build the docker-ut image using ./contrib/mkimage-unittest.sh
+	# then 
+	#   docker build -t SvenDowideit/ambassador .
+	#   docker tag SvenDowideit/ambassador ambassador
+	# then to run it (on the host that has the real backend on it)
+	#   docker run -t -i -link redis:redis -name redis_ambassador -p 6379:6379 ambassador
+	# on the remote host, you can set up another ambassador
+	#    docker run -t -i -name redis_ambassador -expose 6379 sh
+
+	FROM	docker-ut
+	MAINTAINER	SvenDowideit@home.org.au
+
+
+	CMD	env | grep _TCP= | sed 's/.*_PORT_\([0-9]*\)_TCP=tcp:\/\/\(.*\):\(.*\)/socat TCP4-LISTEN:\1,fork,reuseaddr TCP4:\2:\3 \&/'  | sh && top
+

docs/sources/use/baseimages.rst

@@ -37,7 +37,7 @@ There are more example scripts for creating base images in the
 Docker Github Repo:
 
 * `BusyBox <https://github.com/dotcloud/docker/blob/master/contrib/mkimage-busybox.sh>`_
-* `CentOS
-  <https://github.com/dotcloud/docker/blob/master/contrib/mkimage-centos.sh>`_
-* `Debian/Ubuntu
+* `CentOS / Scientific Linux CERN (SLC)
+  <https://github.com/dotcloud/docker/blob/master/contrib/mkimage-rinse.sh>`_
+* `Debian / Ubuntu
   <https://github.com/dotcloud/docker/blob/master/contrib/mkimage-debootstrap.sh>`_

docs/sources/use/basics.rst

@@ -76,11 +76,11 @@ client commands.
   # Add the docker group if it doesn't already exist.
   sudo groupadd docker
 
-  # Add the user "ubuntu" to the docker group.
+  # Add the connected user "${USERNAME}" to the docker group.
   # Change the user name to match your preferred user.
   # You may have to logout and log back in again for
   # this to take effect.
-  sudo gpasswd -a ubuntu docker
+  sudo gpasswd -a ${USERNAME} docker
 
   # Restart the docker daemon.
   sudo service docker restart

docs/sources/use/index.rst

@@ -21,3 +21,4 @@ Contents:
    host_integration
    working_with_volumes
    working_with_links_names
+   ambassador_pattern_linking

docs/sources/use/working_with_links_names.rst

@@ -54,9 +54,9 @@ inter-container communication is set to false.
 
 .. code-block:: bash
 
-    # Example: there is an image called redis-2.6 that exposes the port 6379 and starts redis-server.
+    # Example: there is an image called crosbymichael/redis that exposes the port 6379 and starts redis-server.
     # Let's name the container as "redis" based on that image and run it as daemon.
-    $ sudo docker run -d -name redis redis-2.6
+    $ sudo docker run -d -name redis crosbymichael/redis
 
 We can issue all the commands that you would expect using the name "redis"; start, stop,
 attach, using the name for our container. The name also allows us to link other containers
@@ -102,3 +102,12 @@ about the child container.
 
 Accessing the network information along with the environment of the child container allows
 us to easily connect to the Redis service on the specific IP and port in the environment.
+
+Running ``docker ps`` shows the 2 containers, and the webapp/db alias name for the redis container.
+
+.. code-block:: bash
+
+    $ docker ps
+    CONTAINER ID        IMAGE                        COMMAND                CREATED              STATUS              PORTS               NAMES
+    4c01db0b339c        ubuntu:12.04                 bash                   17 seconds ago       Up 16 seconds                           webapp              
+    d7886598dbe2        crosbymichael/redis:latest   /redis-server --dir    33 minutes ago       Up 33 minutes       6379/tcp            redis,webapp/db     

docs/theme/docker/layout.html

@@ -35,7 +35,7 @@
     %}
 
     {#
-        This part is hopefully complex because things like |cut '/index/' are not available in spinx jinja
+        This part is hopefully complex because things like |cut '/index/' are not available in Sphinx jinja
         and will make it crash. (and we need index/ out.
     #}
     <link rel="canonical" href="http://docs.docker.io/en/latest/

engine/engine.go

@@ -9,7 +9,7 @@ import (
 	"strings"
 )
 
-type Handler func(*Job) string
+type Handler func(*Job) Status
 
 var globalHandlers map[string]Handler
 
@@ -70,7 +70,9 @@ func New(root string) (*Engine, error) {
 		log.Printf("WARNING: %s\n", err)
 	} else {
 		if utils.CompareKernelVersion(k, &utils.KernelVersionInfo{Kernel: 3, Major: 8, Minor: 0}) < 0 {
-			log.Printf("WARNING: You are running linux kernel version %s, which might be unstable running docker. Please upgrade your kernel to 3.8.0.", k.String())
+			if os.Getenv("DOCKER_NOWARN_KERNEL_VERSION") == "" {
+				log.Printf("WARNING: You are running linux kernel version %s, which might be unstable running docker. Please upgrade your kernel to 3.8.0.", k.String())
+			}
 		}
 	}
 	if err := os.MkdirAll(root, 0700); err != nil && !os.IsExist(err) {
@@ -99,10 +101,12 @@ func (eng *Engine) Job(name string, args ...string) *Job {
 		Eng:    eng,
 		Name:   name,
 		Args:   args,
-		Stdin:  os.Stdin,
-		Stdout: os.Stdout,
-		Stderr: os.Stderr,
+		Stdin:  NewInput(),
+		Stdout: NewOutput(),
+		Stderr: NewOutput(),
 	}
+	job.Stdout.Add(utils.NopWriteCloser(os.Stdout))
+	job.Stderr.Add(utils.NopWriteCloser(os.Stderr))
 	handler, exists := eng.handlers[name]
 	if exists {
 		job.handler = handler

engine/engine_test.go

@@ -1,6 +1,9 @@
 package engine
 
 import (
+	"io/ioutil"
+	"os"
+	"path"
 	"testing"
 )
 
@@ -38,8 +41,9 @@ func TestJob(t *testing.T) {
 		t.Fatalf("job1.handler should be empty")
 	}
 
-	h := func(j *Job) string {
-		return j.Name
+	h := func(j *Job) Status {
+		j.Printf("%s\n", j.Name)
+		return 42
 	}
 
 	eng.Register("dummy2", h)
@@ -49,7 +53,51 @@ func TestJob(t *testing.T) {
 		t.Fatalf("job2.handler shouldn't be nil")
 	}
 
-	if job2.handler(job2) != job2.Name {
+	if job2.handler(job2) != 42 {
 		t.Fatalf("handler dummy2 was not found in job2")
 	}
 }
+
+func TestEngineRoot(t *testing.T) {
+	tmp, err := ioutil.TempDir("", "docker-test-TestEngineCreateDir")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer os.RemoveAll(tmp)
+	dir := path.Join(tmp, "dir")
+	eng, err := New(dir)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if st, err := os.Stat(dir); err != nil {
+		t.Fatal(err)
+	} else if !st.IsDir() {
+		t.Fatalf("engine.New() created something other than a directory at %s", dir)
+	}
+	if r := eng.Root(); r != dir {
+		t.Fatalf("Expected: %v\nReceived: %v", dir, r)
+	}
+}
+
+func TestEngineString(t *testing.T) {
+	eng1 := newTestEngine(t)
+	defer os.RemoveAll(eng1.Root())
+	eng2 := newTestEngine(t)
+	defer os.RemoveAll(eng2.Root())
+	s1 := eng1.String()
+	s2 := eng2.String()
+	if eng1 == eng2 {
+		t.Fatalf("Different engines should have different names (%v == %v)", s1, s2)
+	}
+}
+
+func TestEngineLogf(t *testing.T) {
+	eng := newTestEngine(t)
+	defer os.RemoveAll(eng.Root())
+	input := "Test log line"
+	if n, err := eng.Logf("%s\n", input); err != nil {
+		t.Fatal(err)
+	} else if n < len(input) {
+		t.Fatalf("Test: Logf() should print at least as much as the input\ninput=%d\nprinted=%d", len(input), n)
+	}
+}

engine/env_test.go

@@ -37,16 +37,16 @@ func TestSetenvBool(t *testing.T) {
 	job := mkJob(t, "dummy")
 	job.SetenvBool("foo", true)
 	if val := job.GetenvBool("foo"); !val {
-		t.Fatalf("GetenvBool returns incorrect value: %b", val)
+		t.Fatalf("GetenvBool returns incorrect value: %t", val)
 	}
 
 	job.SetenvBool("bar", false)
 	if val := job.GetenvBool("bar"); val {
-		t.Fatalf("GetenvBool returns incorrect value: %b", val)
+		t.Fatalf("GetenvBool returns incorrect value: %t", val)
 	}
 
 	if val := job.GetenvBool("nonexistent"); val {
-		t.Fatalf("GetenvBool returns incorrect value: %b", val)
+		t.Fatalf("GetenvBool returns incorrect value: %t", val)
 	}
 }
 

engine/helpers_test.go

@@ -1,32 +1,18 @@
 package engine
 
 import (
-	"fmt"
 	"github.com/dotcloud/docker/utils"
-	"io/ioutil"
-	"runtime"
-	"strings"
 	"testing"
 )
 
 var globalTestID string
 
 func newTestEngine(t *testing.T) *Engine {
-	// Use the caller function name as a prefix.
-	// This helps trace temp directories back to their test.
-	pc, _, _, _ := runtime.Caller(1)
-	callerLongName := runtime.FuncForPC(pc).Name()
-	parts := strings.Split(callerLongName, ".")
-	callerShortName := parts[len(parts)-1]
-	if globalTestID == "" {
-		globalTestID = utils.RandomString()[:4]
-	}
-	prefix := fmt.Sprintf("docker-test%s-%s-", globalTestID, callerShortName)
-	root, err := ioutil.TempDir("", prefix)
+	tmp, err := utils.TestDirectory("")
 	if err != nil {
 		t.Fatal(err)
 	}
-	eng, err := New(root)
+	eng, err := New(tmp)
 	if err != nil {
 		t.Fatal(err)
 	}

engine/job.go

@@ -1,16 +1,13 @@
 package engine
 
 import (
-	"bufio"
 	"bytes"
 	"encoding/json"
 	"fmt"
 	"io"
-	"io/ioutil"
-	"os"
 	"strconv"
 	"strings"
-	"sync"
+	"time"
 )
 
 // A job is the fundamental unit of work in the docker engine.
@@ -31,126 +28,75 @@ type Job struct {
 	Name    string
 	Args    []string
 	env     []string
-	Stdin   io.Reader
-	Stdout  io.Writer
-	Stderr  io.Writer
-	handler func(*Job) string
-	status  string
+	Stdout  *Output
+	Stderr  *Output
+	Stdin   *Input
+	handler Handler
+	status  Status
+	end     time.Time
 	onExit  []func()
 }
 
+type Status int
+
+const (
+	StatusOK       Status = 0
+	StatusErr      Status = 1
+	StatusNotFound Status = 127
+)
+
 // Run executes the job and blocks until the job completes.
 // If the job returns a failure status, an error is returned
 // which includes the status.
 func (job *Job) Run() error {
-	defer func() {
-		var wg sync.WaitGroup
-		for _, f := range job.onExit {
-			wg.Add(1)
-			go func(f func()) {
-				f()
-				wg.Done()
-			}(f)
-		}
-		wg.Wait()
-	}()
-	if job.Stdout != nil && job.Stdout != os.Stdout {
-		job.Stdout = io.MultiWriter(job.Stdout, os.Stdout)
-	}
-	if job.Stderr != nil && job.Stderr != os.Stderr {
-		job.Stderr = io.MultiWriter(job.Stderr, os.Stderr)
+	// FIXME: make this thread-safe
+	// FIXME: implement wait
+	if !job.end.IsZero() {
+		return fmt.Errorf("%s: job has already completed", job.Name)
 	}
+	// Log beginning and end of the job
 	job.Eng.Logf("+job %s", job.CallString())
 	defer func() {
 		job.Eng.Logf("-job %s%s", job.CallString(), job.StatusString())
 	}()
+	var errorMessage string
+	job.Stderr.AddString(&errorMessage)
 	if job.handler == nil {
-		job.status = "command not found"
+		job.Errorf("%s: command not found", job.Name)
+		job.status = 127
 	} else {
 		job.status = job.handler(job)
+		job.end = time.Now()
 	}
-	if job.status != "0" {
-		return fmt.Errorf("%s: %s", job.Name, job.status)
+	// Wait for all background tasks to complete
+	if err := job.Stdout.Close(); err != nil {
+		return err
+	}
+	if err := job.Stderr.Close(); err != nil {
+		return err
+	}
+	if job.status != 0 {
+		return fmt.Errorf("%s: %s", job.Name, errorMessage)
 	}
 	return nil
 }
 
-func (job *Job) StdoutParseLines(dst *[]string, limit int) {
-	job.parseLines(job.StdoutPipe(), dst, limit)
-}
-
-func (job *Job) StderrParseLines(dst *[]string, limit int) {
-	job.parseLines(job.StderrPipe(), dst, limit)
-}
-
-func (job *Job) parseLines(src io.Reader, dst *[]string, limit int) {
-	var wg sync.WaitGroup
-	wg.Add(1)
-	go func() {
-		defer wg.Done()
-		scanner := bufio.NewScanner(src)
-		for scanner.Scan() {
-			// If the limit is reached, flush the rest of the source and return
-			if limit > 0 && len(*dst) >= limit {
-				io.Copy(ioutil.Discard, src)
-				return
-			}
-			line := scanner.Text()
-			// Append the line (with delimitor removed)
-			*dst = append(*dst, line)
-		}
-	}()
-	job.onExit = append(job.onExit, wg.Wait)
-}
-
-func (job *Job) StdoutParseString(dst *string) {
-	lines := make([]string, 0, 1)
-	job.StdoutParseLines(&lines, 1)
-	job.onExit = append(job.onExit, func() {
-		if len(lines) >= 1 {
-			*dst = lines[0]
-		}
-	})
-}
-
-func (job *Job) StderrParseString(dst *string) {
-	lines := make([]string, 0, 1)
-	job.StderrParseLines(&lines, 1)
-	job.onExit = append(job.onExit, func() { *dst = lines[0] })
-}
-
-func (job *Job) StdoutPipe() io.ReadCloser {
-	r, w := io.Pipe()
-	job.Stdout = w
-	job.onExit = append(job.onExit, func() { w.Close() })
-	return r
-}
-
-func (job *Job) StderrPipe() io.ReadCloser {
-	r, w := io.Pipe()
-	job.Stderr = w
-	job.onExit = append(job.onExit, func() { w.Close() })
-	return r
-}
-
 func (job *Job) CallString() string {
 	return fmt.Sprintf("%s(%s)", job.Name, strings.Join(job.Args, ", "))
 }
 
 func (job *Job) StatusString() string {
-	// FIXME: if a job returns the empty string, it will be printed
-	// as not having returned.
-	// (this only affects String which is a convenience function).
-	if job.status != "" {
-		var okerr string
-		if job.status == "0" {
-			okerr = "OK"
-		} else {
-			okerr = "ERR"
-		}
-		return fmt.Sprintf(" = %s (%s)", okerr, job.status)
+	// If the job hasn't completed, status string is empty
+	if job.end.IsZero() {
+		return ""
 	}
-	return ""
+	var okerr string
+	if job.status == StatusOK {
+		okerr = "OK"
+	} else {
+		okerr = "ERR"
+	}
+	return fmt.Sprintf(" = %s (%d)", okerr, job.status)
 }
 
 // String returns a human-readable description of `job`
@@ -338,5 +284,8 @@ func (job *Job) Printf(format string, args ...interface{}) (n int, err error) {
 
 func (job *Job) Errorf(format string, args ...interface{}) (n int, err error) {
 	return fmt.Fprintf(job.Stderr, format, args...)
-
+}
+
+func (job *Job) Error(err error) (int, error) {
+	return fmt.Fprintf(job.Stderr, "%s", err)
 }

engine/job_test.go

@@ -0,0 +1,80 @@
+package engine
+
+import (
+	"os"
+	"testing"
+)
+
+func TestJobStatusOK(t *testing.T) {
+	eng := newTestEngine(t)
+	defer os.RemoveAll(eng.Root())
+	eng.Register("return_ok", func(job *Job) Status { return StatusOK })
+	err := eng.Job("return_ok").Run()
+	if err != nil {
+		t.Fatalf("Expected: err=%v\nReceived: err=%v", nil, err)
+	}
+}
+
+func TestJobStatusErr(t *testing.T) {
+	eng := newTestEngine(t)
+	defer os.RemoveAll(eng.Root())
+	eng.Register("return_err", func(job *Job) Status { return StatusErr })
+	err := eng.Job("return_err").Run()
+	if err == nil {
+		t.Fatalf("When a job returns StatusErr, Run() should return an error")
+	}
+}
+
+func TestJobStatusNotFound(t *testing.T) {
+	eng := newTestEngine(t)
+	defer os.RemoveAll(eng.Root())
+	eng.Register("return_not_found", func(job *Job) Status { return StatusNotFound })
+	err := eng.Job("return_not_found").Run()
+	if err == nil {
+		t.Fatalf("When a job returns StatusNotFound, Run() should return an error")
+	}
+}
+
+func TestJobStdoutString(t *testing.T) {
+	eng := newTestEngine(t)
+	defer os.RemoveAll(eng.Root())
+	// FIXME: test multiple combinations of output and status
+	eng.Register("say_something_in_stdout", func(job *Job) Status {
+		job.Printf("Hello world\n")
+		return StatusOK
+	})
+
+	job := eng.Job("say_something_in_stdout")
+	var output string
+	if err := job.Stdout.AddString(&output); err != nil {
+		t.Fatal(err)
+	}
+	if err := job.Run(); err != nil {
+		t.Fatal(err)
+	}
+	if expectedOutput := "Hello world"; output != expectedOutput {
+		t.Fatalf("Stdout last line:\nExpected: %v\nReceived: %v", expectedOutput, output)
+	}
+}
+
+func TestJobStderrString(t *testing.T) {
+	eng := newTestEngine(t)
+	defer os.RemoveAll(eng.Root())
+	// FIXME: test multiple combinations of output and status
+	eng.Register("say_something_in_stderr", func(job *Job) Status {
+		job.Errorf("Warning, something might happen\nHere it comes!\nOh no...\nSomething happened\n")
+		return StatusOK
+	})
+
+	job := eng.Job("say_something_in_stderr")
+	var output string
+	if err := job.Stderr.AddString(&output); err != nil {
+		t.Fatal(err)
+	}
+	if err := job.Run(); err != nil {
+		t.Fatal(err)
+	}
+	if expectedOutput := "Something happened"; output != expectedOutput {
+		t.Fatalf("Stderr last line:\nExpected: %v\nReceived: %v", expectedOutput, output)
+	}
+}

engine/streams.go

@@ -0,0 +1,166 @@
+package engine
+
+import (
+	"bufio"
+	"container/ring"
+	"fmt"
+	"io"
+	"sync"
+)
+
+type Output struct {
+	sync.Mutex
+	dests []io.Writer
+	tasks sync.WaitGroup
+}
+
+// NewOutput returns a new Output object with no destinations attached.
+// Writing to an empty Output will cause the written data to be discarded.
+func NewOutput() *Output {
+	return &Output{}
+}
+
+// Add attaches a new destination to the Output. Any data subsequently written
+// to the output will be written to the new destination in addition to all the others.
+// This method is thread-safe.
+// FIXME: Add cannot fail
+func (o *Output) Add(dst io.Writer) error {
+	o.Mutex.Lock()
+	defer o.Mutex.Unlock()
+	o.dests = append(o.dests, dst)
+	return nil
+}
+
+// AddPipe creates an in-memory pipe with io.Pipe(), adds its writing end as a destination,
+// and returns its reading end for consumption by the caller.
+// This is a rough equivalent similar to Cmd.StdoutPipe() in the standard os/exec package.
+// This method is thread-safe.
+func (o *Output) AddPipe() (io.Reader, error) {
+	r, w := io.Pipe()
+	o.Add(w)
+	return r, nil
+}
+
+// AddTail starts a new goroutine which will read all subsequent data written to the output,
+// line by line, and append the last `n` lines to `dst`.
+func (o *Output) AddTail(dst *[]string, n int) error {
+	src, err := o.AddPipe()
+	if err != nil {
+		return err
+	}
+	o.tasks.Add(1)
+	go func() {
+		defer o.tasks.Done()
+		Tail(src, n, dst)
+	}()
+	return nil
+}
+
+// AddString starts a new goroutine which will read all subsequent data written to the output,
+// line by line, and store the last line into `dst`.
+func (o *Output) AddString(dst *string) error {
+	src, err := o.AddPipe()
+	if err != nil {
+		return err
+	}
+	o.tasks.Add(1)
+	go func() {
+		defer o.tasks.Done()
+		lines := make([]string, 0, 1)
+		Tail(src, 1, &lines)
+		if len(lines) == 0 {
+			*dst = ""
+		} else {
+			*dst = lines[0]
+		}
+	}()
+	return nil
+}
+
+// Write writes the same data to all registered destinations.
+// This method is thread-safe.
+func (o *Output) Write(p []byte) (n int, err error) {
+	o.Mutex.Lock()
+	defer o.Mutex.Unlock()
+	var firstErr error
+	for _, dst := range o.dests {
+		_, err := dst.Write(p)
+		if err != nil && firstErr == nil {
+			firstErr = err
+		}
+	}
+	return len(p), firstErr
+}
+
+// Close unregisters all destinations and waits for all background
+// AddTail and AddString tasks to complete.
+// The Close method of each destination is called if it exists.
+func (o *Output) Close() error {
+	o.Mutex.Lock()
+	defer o.Mutex.Unlock()
+	var firstErr error
+	for _, dst := range o.dests {
+		if closer, ok := dst.(io.WriteCloser); ok {
+			err := closer.Close()
+			if err != nil && firstErr == nil {
+				firstErr = err
+			}
+		}
+	}
+	o.tasks.Wait()
+	return firstErr
+}
+
+type Input struct {
+	src io.Reader
+	sync.Mutex
+}
+
+// NewInput returns a new Input object with no source attached.
+// Reading to an empty Input will return io.EOF.
+func NewInput() *Input {
+	return &Input{}
+}
+
+// Read reads from the input in a thread-safe way.
+func (i *Input) Read(p []byte) (n int, err error) {
+	i.Mutex.Lock()
+	defer i.Mutex.Unlock()
+	if i.src == nil {
+		return 0, io.EOF
+	}
+	return i.src.Read(p)
+}
+
+// Add attaches a new source to the input.
+// Add can only be called once per input. Subsequent calls will
+// return an error.
+func (i *Input) Add(src io.Reader) error {
+	i.Mutex.Lock()
+	defer i.Mutex.Unlock()
+	if i.src != nil {
+		return fmt.Errorf("Maximum number of sources reached: 1")
+	}
+	i.src = src
+	return nil
+}
+
+// Tail reads from `src` line per line, and returns the last `n` lines as an array.
+// A ring buffer is used to only store `n` lines at any time.
+func Tail(src io.Reader, n int, dst *[]string) {
+	scanner := bufio.NewScanner(src)
+	r := ring.New(n)
+	for scanner.Scan() {
+		if n == 0 {
+			continue
+		}
+		r.Value = scanner.Text()
+		r = r.Next()
+	}
+	r.Do(func(v interface{}) {
+		if v == nil {
+			return
+		}
+		*dst = append(*dst, v.(string))
+	})
+}

engine/streams_test.go

@@ -0,0 +1,274 @@
+package engine
+
+import (
+	"bufio"
+	"bytes"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"strings"
+	"testing"
+)
+
+func TestOutputAddString(t *testing.T) {
+	var testInputs = [][2]string{
+		{
+			"hello, world!",
+			"hello, world!",
+		},
+
+		{
+			"One\nTwo\nThree",
+			"Three",
+		},
+
+		{
+			"",
+			"",
+		},
+
+		{
+			"A line\nThen another nl-terminated line\n",
+			"Then another nl-terminated line",
+		},
+
+		{
+			"A line followed by an empty line\n\n",
+			"",
+		},
+	}
+	for _, testData := range testInputs {
+		input := testData[0]
+		expectedOutput := testData[1]
+		o := NewOutput()
+		var output string
+		if err := o.AddString(&output); err != nil {
+			t.Error(err)
+		}
+		if n, err := o.Write([]byte(input)); err != nil {
+			t.Error(err)
+		} else if n != len(input) {
+			t.Errorf("Expected %d, got %d", len(input), n)
+		}
+		o.Close()
+		if output != expectedOutput {
+			t.Errorf("Last line is not stored as return string.\nInput:   '%s'\nExpected: '%s'\nGot:       '%s'", input, expectedOutput, output)
+		}
+	}
+}
+
+type sentinelWriteCloser struct {
+	calledWrite bool
+	calledClose bool
+}
+
+func (w *sentinelWriteCloser) Write(p []byte) (int, error) {
+	w.calledWrite = true
+	return len(p), nil
+}
+
+func (w *sentinelWriteCloser) Close() error {
+	w.calledClose = true
+	return nil
+}
+
+func TestOutputAddClose(t *testing.T) {
+	o := NewOutput()
+	var s sentinelWriteCloser
+	if err := o.Add(&s); err != nil {
+		t.Fatal(err)
+	}
+	if err := o.Close(); err != nil {
+		t.Fatal(err)
+	}
+	// Write data after the output is closed.
+	// Write should succeed, but no destination should receive it.
+	if _, err := o.Write([]byte("foo bar")); err != nil {
+		t.Fatal(err)
+	}
+	if !s.calledClose {
+		t.Fatal("Output.Close() didn't close the destination")
+	}
+}
+
+func TestOutputAddPipe(t *testing.T) {
+	var testInputs = []string{
+		"hello, world!",
+		"One\nTwo\nThree",
+		"",
+		"A line\nThen another nl-terminated line\n",
+		"A line followed by an empty line\n\n",
+	}
+	for _, input := range testInputs {
+		expectedOutput := input
+		o := NewOutput()
+		r, err := o.AddPipe()
+		if err != nil {
+			t.Fatal(err)
+		}
+		go func(o *Output) {
+			if n, err := o.Write([]byte(input)); err != nil {
+				t.Error(err)
+			} else if n != len(input) {
+				t.Errorf("Expected %d, got %d", len(input), n)
+			}
+			if err := o.Close(); err != nil {
+				t.Error(err)
+			}
+		}(o)
+		output, err := ioutil.ReadAll(r)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if string(output) != expectedOutput {
+			t.Errorf("Last line is not stored as return string.\nExpected: '%s'\nGot:       '%s'", expectedOutput, output)
+		}
+	}
+}
+
+func TestTail(t *testing.T) {
+	var tests = make(map[string][][]string)
+	tests["hello, world!"] = [][]string{
+		{},
+		{"hello, world!"},
+		{"hello, world!"},
+		{"hello, world!"},
+	}
+	tests["One\nTwo\nThree"] = [][]string{
+		{},
+		{"Three"},
+		{"Two", "Three"},
+		{"One", "Two", "Three"},
+	}
+	for input, outputs := range tests {
+		for n, expectedOutput := range outputs {
+			var output []string
+			Tail(strings.NewReader(input), n, &output)
+			if fmt.Sprintf("%v", output) != fmt.Sprintf("%v", expectedOutput) {
+				t.Errorf("Tail n=%d returned wrong result.\nExpected: '%s'\nGot     : '%s'", expectedOutput, output)
+			}
+		}
+	}
+}
+
+func TestOutputAddTail(t *testing.T) {
+	var tests = make(map[string][][]string)
+	tests["hello, world!"] = [][]string{
+		{},
+		{"hello, world!"},
+		{"hello, world!"},
+		{"hello, world!"},
+	}
+	tests["One\nTwo\nThree"] = [][]string{
+		{},
+		{"Three"},
+		{"Two", "Three"},
+		{"One", "Two", "Three"},
+	}
+	for input, outputs := range tests {
+		for n, expectedOutput := range outputs {
+			o := NewOutput()
+			var output []string
+			if err := o.AddTail(&output, n); err != nil {
+				t.Error(err)
+			}
+			if n, err := o.Write([]byte(input)); err != nil {
+				t.Error(err)
+			} else if n != len(input) {
+				t.Errorf("Expected %d, got %d", len(input), n)
+			}
+			o.Close()
+			if fmt.Sprintf("%v", output) != fmt.Sprintf("%v", expectedOutput) {
+				t.Errorf("Tail(%d) returned wrong result.\nExpected: %v\nGot:      %v", n, expectedOutput, output)
+			}
+		}
+	}
+}
+
+func lastLine(txt string) string {
+	scanner := bufio.NewScanner(strings.NewReader(txt))
+	var lastLine string
+	for scanner.Scan() {
+		lastLine = scanner.Text()
+	}
+	return lastLine
+}
+
+func TestOutputAdd(t *testing.T) {
+	o := NewOutput()
+	b := &bytes.Buffer{}
+	o.Add(b)
+	input := "hello, world!"
+	if n, err := o.Write([]byte(input)); err != nil {
+		t.Fatal(err)
+	} else if n != len(input) {
+		t.Fatalf("Expected %d, got %d", len(input), n)
+	}
+	if output := b.String(); output != input {
+		t.Fatal("Received wrong data from Add.\nExpected: '%s'\nGot:     '%s'", input, output)
+	}
+}
+
+func TestOutputWriteError(t *testing.T) {
+	o := NewOutput()
+	buf := &bytes.Buffer{}
+	o.Add(buf)
+	r, w := io.Pipe()
+	input := "Hello there"
+	expectedErr := fmt.Errorf("This is an error")
+	r.CloseWithError(expectedErr)
+	o.Add(w)
+	n, err := o.Write([]byte(input))
+	if err != expectedErr {
+		t.Fatalf("Output.Write() should return the first error encountered, if any")
+	}
+	if buf.String() != input {
+		t.Fatalf("Output.Write() should attempt write on all destinations, even after encountering an error")
+	}
+	if n != len(input) {
+		t.Fatalf("Output.Write() should return the size of the input if it successfully writes to at least one destination")
+	}
+}
+
+func TestInputAddEmpty(t *testing.T) {
+	i := NewInput()
+	var b bytes.Buffer
+	if err := i.Add(&b); err != nil {
+		t.Fatal(err)
+	}
+	data, err := ioutil.ReadAll(i)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(data) > 0 {
+		t.Fatalf("Read from empty input shoul yield no data")
+	}
+}
+
+func TestInputAddTwo(t *testing.T) {
+	i := NewInput()
+	var b1 bytes.Buffer
+	// First add should succeed
+	if err := i.Add(&b1); err != nil {
+		t.Fatal(err)
+	}
+	var b2 bytes.Buffer
+	// Second add should fail
+	if err := i.Add(&b2); err == nil {
+		t.Fatalf("Adding a second source should return an error")
+	}
+}
+
+func TestInputAddNotEmpty(t *testing.T) {
+	i := NewInput()
+	b := bytes.NewBufferString("hello world\nabc")
+	expectedResult := b.String()
+	i.Add(b)
+	result, err := ioutil.ReadAll(i)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if string(result) != expectedResult {
+		t.Fatalf("Expected: %v\nReceived: %v", expectedResult, result)
+	}
+}

graph.go

@@ -3,6 +3,7 @@ package docker
 import (
 	"fmt"
 	"github.com/dotcloud/docker/archive"
+	"github.com/dotcloud/docker/graphdriver"
 	"github.com/dotcloud/docker/utils"
 	"io"
 	"io/ioutil"
@@ -10,6 +11,7 @@ import (
 	"path"
 	"path/filepath"
 	"strings"
+	"syscall"
 	"time"
 )
 
@@ -17,11 +19,12 @@ import (
 type Graph struct {
 	Root    string
 	idIndex *utils.TruncIndex
+	driver  graphdriver.Driver
 }
 
 // NewGraph instantiates a new graph at the given root path in the filesystem.
 // `root` will be created if it doesn't exist.
-func NewGraph(root string) (*Graph, error) {
+func NewGraph(root string, driver graphdriver.Driver) (*Graph, error) {
 	abspath, err := filepath.Abs(root)
 	if err != nil {
 		return nil, err
@@ -30,9 +33,11 @@ func NewGraph(root string) (*Graph, error) {
 	if err := os.MkdirAll(root, 0700); err != nil && !os.IsExist(err) {
 		return nil, err
 	}
+
 	graph := &Graph{
 		Root:    abspath,
 		idIndex: utils.NewTruncIndex(),
+		driver:  driver,
 	}
 	if err := graph.restore(); err != nil {
 		return nil, err
@@ -47,7 +52,9 @@ func (graph *Graph) restore() error {
 	}
 	for _, v := range dir {
 		id := v.Name()
-		graph.idIndex.Add(id)
+		if graph.driver.Exists(id) {
+			graph.idIndex.Add(id)
+		}
 	}
 	return nil
 }
@@ -78,16 +85,36 @@ func (graph *Graph) Get(name string) (*Image, error) {
 	if err != nil {
 		return nil, err
 	}
+	// Check that the filesystem layer exists
+	rootfs, err := graph.driver.Get(img.ID)
+	if err != nil {
+		return nil, fmt.Errorf("Driver %s failed to get image rootfs %s: %s", graph.driver, img.ID, err)
+	}
 	if img.ID != id {
 		return nil, fmt.Errorf("Image stored at '%s' has wrong id '%s'", id, img.ID)
 	}
 	img.graph = graph
-	if img.Size == 0 {
-		root, err := img.root()
-		if err != nil {
-			return nil, err
+
+	if img.Size < 0 {
+		var size int64
+		if img.Parent == "" {
+			if size, err = utils.TreeSize(rootfs); err != nil {
+				return nil, err
+			}
+		} else {
+			parentFs, err := graph.driver.Get(img.Parent)
+			if err != nil {
+				return nil, err
+			}
+			changes, err := archive.ChangesDirs(rootfs, parentFs)
+			if err != nil {
+				return nil, err
+			}
+			size = archive.ChangesSize(rootfs, changes)
 		}
-		if err := StoreSize(img, root); err != nil {
+
+		img.Size = size
+		if err := img.SaveSize(graph.imageRoot(id)); err != nil {
 			return nil, err
 		}
 	}
@@ -99,7 +126,7 @@ func (graph *Graph) Create(layerData archive.Archive, container *Container, comm
 	img := &Image{
 		ID:            GenerateID(),
 		Comment:       comment,
-		Created:       time.Now(),
+		Created:       time.Now().UTC(),
 		DockerVersion: VERSION,
 		Author:        author,
 		Config:        config,
@@ -118,7 +145,15 @@ func (graph *Graph) Create(layerData archive.Archive, container *Container, comm
 
 // Register imports a pre-existing image into the graph.
 // FIXME: pass img as first argument
-func (graph *Graph) Register(jsonData []byte, layerData archive.Archive, img *Image) error {
+func (graph *Graph) Register(jsonData []byte, layerData archive.Archive, img *Image) (err error) {
+	defer func() {
+		// If any error occurs, remove the new dir from the driver.
+		// Don't check for errors since the dir might not have been created.
+		// FIXME: this leaves a possible race condition.
+		if err != nil {
+			graph.driver.Remove(img.ID)
+		}
+	}()
 	if err := ValidateID(img.ID); err != nil {
 		return err
 	}
@@ -126,19 +161,43 @@ func (graph *Graph) Register(jsonData []byte, layerData archive.Archive, img *Im
 	if graph.Exists(img.ID) {
 		return fmt.Errorf("Image %s already exists", img.ID)
 	}
+
+	// Ensure that the image root does not exist on the filesystem
+	// when it is not registered in the graph.
+	// This is common when you switch from one graph driver to another
+	if err := os.RemoveAll(graph.imageRoot(img.ID)); err != nil && !os.IsNotExist(err) {
+		return err
+	}
+
+	// If the driver has this ID but the graph doesn't, remove it from the driver to start fresh.
+	// (the graph is the source of truth).
+	// Ignore errors, since we don't know if the driver correctly returns ErrNotExist.
+	// (FIXME: make that mandatory for drivers).
+	graph.driver.Remove(img.ID)
+
 	tmp, err := graph.Mktemp("")
 	defer os.RemoveAll(tmp)
 	if err != nil {
 		return fmt.Errorf("Mktemp failed: %s", err)
 	}
-	if err := StoreImage(img, jsonData, layerData, tmp); err != nil {
+
+	// Create root filesystem in the driver
+	if err := graph.driver.Create(img.ID, img.Parent); err != nil {
+		return fmt.Errorf("Driver %s failed to create image rootfs %s: %s", graph.driver, img.ID, err)
+	}
+	// Mount the root filesystem so we can apply the diff/layer
+	rootfs, err := graph.driver.Get(img.ID)
+	if err != nil {
+		return fmt.Errorf("Driver %s failed to get image rootfs %s: %s", graph.driver, img.ID, err)
+	}
+	img.graph = graph
+	if err := StoreImage(img, jsonData, layerData, tmp, rootfs); err != nil {
 		return err
 	}
 	// Commit
 	if err := os.Rename(tmp, graph.imageRoot(img.ID)); err != nil {
 		return err
 	}
-	img.graph = graph
 	graph.idIndex.Add(img.ID)
 	return nil
 }
@@ -152,50 +211,33 @@ func (graph *Graph) TempLayerArchive(id string, compression archive.Compression,
 	if err != nil {
 		return nil, err
 	}
-	tmp, err := graph.tmp()
+	tmp, err := graph.Mktemp("")
 	if err != nil {
 		return nil, err
 	}
-	a, err := image.TarLayer(compression)
+	a, err := image.TarLayer()
 	if err != nil {
 		return nil, err
 	}
-	return archive.NewTempArchive(utils.ProgressReader(ioutil.NopCloser(a), 0, output, sf.FormatProgress("", "Buffering to disk", "%v/%v (%v)"), sf, true), tmp.Root)
+	return archive.NewTempArchive(utils.ProgressReader(ioutil.NopCloser(a), 0, output, sf, true, "", "Buffering to disk"), tmp)
 }
 
 // Mktemp creates a temporary sub-directory inside the graph's filesystem.
 func (graph *Graph) Mktemp(id string) (string, error) {
-	if id == "" {
-		id = GenerateID()
+	dir := path.Join(graph.Root, "_tmp", GenerateID())
+	if err := os.MkdirAll(dir, 0700); err != nil {
+		return "", err
 	}
-	tmp, err := graph.tmp()
-	if err != nil {
-		return "", fmt.Errorf("Couldn't create temp: %s", err)
-	}
-	if tmp.Exists(id) {
-		return "", fmt.Errorf("Image %s already exists", id)
-	}
-	return tmp.imageRoot(id), nil
+	return dir, nil
 }
 
-// getDockerInitLayer returns the path of a layer containing a mountpoint suitable
+// setupInitLayer populates a directory with mountpoints suitable
 // for bind-mounting dockerinit into the container. The mountpoint is simply an
 // empty file at /.dockerinit
 //
 // This extra layer is used by all containers as the top-most ro layer. It protects
 // the container from unwanted side-effects on the rw layer.
-func (graph *Graph) getDockerInitLayer() (string, error) {
-	tmp, err := graph.tmp()
-	if err != nil {
-		return "", err
-	}
-	initLayer := tmp.imageRoot("_dockerinit")
-	if err := os.Mkdir(initLayer, 0755); err != nil && !os.IsExist(err) {
-		// If directory already existed, keep going.
-		// For all other errors, abort.
-		return "", err
-	}
-
+func setupInitLayer(initLayer string) error {
 	for pth, typ := range map[string]string{
 		"/dev/pts":         "dir",
 		"/dev/shm":         "dir",
@@ -209,36 +251,38 @@ func (graph *Graph) getDockerInitLayer() (string, error) {
 		// "var/run": "dir",
 		// "var/lock": "dir",
 	} {
+		parts := strings.Split(pth, "/")
+		prev := "/"
+		for _, p := range parts[1:] {
+			prev = path.Join(prev, p)
+			syscall.Unlink(path.Join(initLayer, prev))
+		}
+
 		if _, err := os.Stat(path.Join(initLayer, pth)); err != nil {
 			if os.IsNotExist(err) {
 				switch typ {
 				case "dir":
 					if err := os.MkdirAll(path.Join(initLayer, pth), 0755); err != nil {
-						return "", err
+						return err
 					}
 				case "file":
 					if err := os.MkdirAll(path.Join(initLayer, path.Dir(pth)), 0755); err != nil {
-						return "", err
+						return err
 					}
 					f, err := os.OpenFile(path.Join(initLayer, pth), os.O_CREATE, 0755)
 					if err != nil {
-						return "", err
+						return err
 					}
 					f.Close()
 				}
 			} else {
-				return "", err
+				return err
 			}
 		}
 	}
 
 	// Layer is ready to use, if it wasn't before.
-	return initLayer, nil
-}
-
-func (graph *Graph) tmp() (*Graph, error) {
-	// Changed to _tmp from :tmp:, because it messed with ":" separators in aufs branch syntax...
-	return NewGraph(path.Join(graph.Root, "_tmp"))
+	return nil
 }
 
 // Check if given error is "not empty".
@@ -270,6 +314,9 @@ func (graph *Graph) Delete(name string) error {
 	if err != nil {
 		return err
 	}
+	// Remove rootfs data from the driver
+	graph.driver.Remove(id)
+	// Remove the trashed image directory
 	return os.RemoveAll(tmp)
 }
 
@@ -344,3 +391,7 @@ func (graph *Graph) Heads() (map[string]*Image, error) {
 func (graph *Graph) imageRoot(id string) string {
 	return path.Join(graph.Root, id)
 }
+
+func (graph *Graph) Driver() graphdriver.Driver {
+	return graph.driver
+}

graph_test.go

@@ -1,292 +0,0 @@
-package docker
-
-import (
-	"archive/tar"
-	"bytes"
-	"errors"
-	"github.com/dotcloud/docker/archive"
-	"github.com/dotcloud/docker/utils"
-	"io"
-	"io/ioutil"
-	"os"
-	"testing"
-	"time"
-)
-
-func TestInit(t *testing.T) {
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	// Root should exist
-	if _, err := os.Stat(graph.Root); err != nil {
-		t.Fatal(err)
-	}
-	// Map() should be empty
-	if l, err := graph.Map(); err != nil {
-		t.Fatal(err)
-	} else if len(l) != 0 {
-		t.Fatalf("len(Map()) should return %d, not %d", 0, len(l))
-	}
-}
-
-// Test that Register can be interrupted cleanly without side effects
-func TestInterruptedRegister(t *testing.T) {
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	badArchive, w := io.Pipe() // Use a pipe reader as a fake archive which never yields data
-	image := &Image{
-		ID:      GenerateID(),
-		Comment: "testing",
-		Created: time.Now(),
-	}
-	go graph.Register(nil, badArchive, image)
-	time.Sleep(200 * time.Millisecond)
-	w.CloseWithError(errors.New("But I'm not a tarball!")) // (Nobody's perfect, darling)
-	if _, err := graph.Get(image.ID); err == nil {
-		t.Fatal("Image should not exist after Register is interrupted")
-	}
-	// Registering the same image again should succeed if the first register was interrupted
-	goodArchive, err := fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	if err := graph.Register(nil, goodArchive, image); err != nil {
-		t.Fatal(err)
-	}
-}
-
-// FIXME: Do more extensive tests (ex: create multiple, delete, recreate;
-//       create multiple, check the amount of images and paths, etc..)
-func TestGraphCreate(t *testing.T) {
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	archive, err := fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	image, err := graph.Create(archive, nil, "Testing", "", nil)
-	if err != nil {
-		t.Fatal(err)
-	}
-	if err := ValidateID(image.ID); err != nil {
-		t.Fatal(err)
-	}
-	if image.Comment != "Testing" {
-		t.Fatalf("Wrong comment: should be '%s', not '%s'", "Testing", image.Comment)
-	}
-	if image.DockerVersion != VERSION {
-		t.Fatalf("Wrong docker_version: should be '%s', not '%s'", VERSION, image.DockerVersion)
-	}
-	images, err := graph.Map()
-	if err != nil {
-		t.Fatal(err)
-	} else if l := len(images); l != 1 {
-		t.Fatalf("Wrong number of images. Should be %d, not %d", 1, l)
-	}
-	if images[image.ID] == nil {
-		t.Fatalf("Could not find image with id %s", image.ID)
-	}
-}
-
-func TestRegister(t *testing.T) {
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	archive, err := fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	image := &Image{
-		ID:      GenerateID(),
-		Comment: "testing",
-		Created: time.Now(),
-	}
-	err = graph.Register(nil, archive, image)
-	if err != nil {
-		t.Fatal(err)
-	}
-	if images, err := graph.Map(); err != nil {
-		t.Fatal(err)
-	} else if l := len(images); l != 1 {
-		t.Fatalf("Wrong number of images. Should be %d, not %d", 1, l)
-	}
-	if resultImg, err := graph.Get(image.ID); err != nil {
-		t.Fatal(err)
-	} else {
-		if resultImg.ID != image.ID {
-			t.Fatalf("Wrong image ID. Should be '%s', not '%s'", image.ID, resultImg.ID)
-		}
-		if resultImg.Comment != image.Comment {
-			t.Fatalf("Wrong image comment. Should be '%s', not '%s'", image.Comment, resultImg.Comment)
-		}
-	}
-}
-
-// Test that an image can be deleted by its shorthand prefix
-func TestDeletePrefix(t *testing.T) {
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	img := createTestImage(graph, t)
-	if err := graph.Delete(utils.TruncateID(img.ID)); err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 0)
-}
-
-func createTestImage(graph *Graph, t *testing.T) *Image {
-	archive, err := fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	img, err := graph.Create(archive, nil, "Test image", "", nil)
-	if err != nil {
-		t.Fatal(err)
-	}
-	return img
-}
-
-func TestDelete(t *testing.T) {
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	archive, err := fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 0)
-	img, err := graph.Create(archive, nil, "Bla bla", "", nil)
-	if err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 1)
-	if err := graph.Delete(img.ID); err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 0)
-
-	archive, err = fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	// Test 2 create (same name) / 1 delete
-	img1, err := graph.Create(archive, nil, "Testing", "", nil)
-	if err != nil {
-		t.Fatal(err)
-	}
-	archive, err = fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	if _, err = graph.Create(archive, nil, "Testing", "", nil); err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 2)
-	if err := graph.Delete(img1.ID); err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 1)
-
-	// Test delete wrong name
-	if err := graph.Delete("Not_foo"); err == nil {
-		t.Fatalf("Deleting wrong ID should return an error")
-	}
-	assertNImages(graph, t, 1)
-
-	archive, err = fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	// Test delete twice (pull -> rm -> pull -> rm)
-	if err := graph.Register(nil, archive, img1); err != nil {
-		t.Fatal(err)
-	}
-	if err := graph.Delete(img1.ID); err != nil {
-		t.Fatal(err)
-	}
-	assertNImages(graph, t, 1)
-}
-
-func TestByParent(t *testing.T) {
-	archive1, _ := fakeTar()
-	archive2, _ := fakeTar()
-	archive3, _ := fakeTar()
-
-	graph := tempGraph(t)
-	defer os.RemoveAll(graph.Root)
-	parentImage := &Image{
-		ID:      GenerateID(),
-		Comment: "parent",
-		Created: time.Now(),
-		Parent:  "",
-	}
-	childImage1 := &Image{
-		ID:      GenerateID(),
-		Comment: "child1",
-		Created: time.Now(),
-		Parent:  parentImage.ID,
-	}
-	childImage2 := &Image{
-		ID:      GenerateID(),
-		Comment: "child2",
-		Created: time.Now(),
-		Parent:  parentImage.ID,
-	}
-	_ = graph.Register(nil, archive1, parentImage)
-	_ = graph.Register(nil, archive2, childImage1)
-	_ = graph.Register(nil, archive3, childImage2)
-
-	byParent, err := graph.ByParent()
-	if err != nil {
-		t.Fatal(err)
-	}
-	numChildren := len(byParent[parentImage.ID])
-	if numChildren != 2 {
-		t.Fatalf("Expected 2 children, found %d", numChildren)
-	}
-}
-
-func assertNImages(graph *Graph, t *testing.T, n int) {
-	if images, err := graph.Map(); err != nil {
-		t.Fatal(err)
-	} else if actualN := len(images); actualN != n {
-		t.Fatalf("Expected %d images, found %d", n, actualN)
-	}
-}
-
-/*
- * HELPER FUNCTIONS
- */
-
-func tempGraph(t *testing.T) *Graph {
-	tmp, err := ioutil.TempDir("", "docker-graph-")
-	if err != nil {
-		t.Fatal(err)
-	}
-	graph, err := NewGraph(tmp)
-	if err != nil {
-		t.Fatal(err)
-	}
-	return graph
-}
-
-func testArchive(t *testing.T) archive.Archive {
-	archive, err := fakeTar()
-	if err != nil {
-		t.Fatal(err)
-	}
-	return archive
-}
-
-func fakeTar() (io.Reader, error) {
-	content := []byte("Hello world!\n")
-	buf := new(bytes.Buffer)
-	tw := tar.NewWriter(buf)
-	for _, name := range []string{"/etc/postgres/postgres.conf", "/etc/passwd", "/var/log/postgres/postgres.conf"} {
-		hdr := new(tar.Header)
-		hdr.Size = int64(len(content))
-		hdr.Name = name
-		if err := tw.WriteHeader(hdr); err != nil {
-			return nil, err
-		}
-		tw.Write([]byte(content))
-	}
-	tw.Close()
-	return buf, nil
-}

graphdb/graphdb.go

@@ -1,4 +1,4 @@
-package gograph
+package graphdb
 
 import (
 	"database/sql"

graphdb/graphdb_test.go

@@ -1,4 +1,4 @@
-package gograph
+package graphdb
 
 import (
 	_ "code.google.com/p/gosqlite/sqlite3"

graphdb/sort.go

@@ -1,4 +1,4 @@
-package gograph
+package graphdb
 
 import "sort"
 

graphdb/sort_test.go

@@ -1,4 +1,4 @@
-package gograph
+package graphdb
 
 import (
 	"testing"

graphdb/utils.go

@@ -1,4 +1,4 @@
-package gograph
+package graphdb
 
 import (
 	"path"

graphdriver/aufs/aufs.go

@@ -0,0 +1,336 @@
+/*
+
+aufs driver directory structure
+
+.
+├── layers // Metadata of layers
+│   ├── 1
+│   ├── 2
+│   └── 3
+├── diffs  // Content of the layer
+│   ├── 1  // Contains layers that need to be mounted for the id
+│   ├── 2
+│   └── 3
+└── mnt    // Mount points for the rw layers to be mounted
+    ├── 1
+    ├── 2
+    └── 3
+
+*/
+
+package aufs
+
+import (
+	"bufio"
+	"fmt"
+	"github.com/dotcloud/docker/archive"
+	"github.com/dotcloud/docker/graphdriver"
+	"github.com/dotcloud/docker/utils"
+	"log"
+	"os"
+	"os/exec"
+	"path"
+	"strings"
+)
+
+func init() {
+	graphdriver.Register("aufs", Init)
+}
+
+type Driver struct {
+	root string
+}
+
+// New returns a new AUFS driver.
+// An error is returned if AUFS is not supported.
+func Init(root string) (graphdriver.Driver, error) {
+	// Try to load the aufs kernel module
+	if err := supportsAufs(); err != nil {
+		return nil, err
+	}
+	paths := []string{
+		"mnt",
+		"diff",
+		"layers",
+	}
+
+	// Create the root aufs driver dir and return
+	// if it already exists
+	// If not populate the dir structure
+	if err := os.MkdirAll(root, 0755); err != nil {
+		if os.IsExist(err) {
+			return &Driver{root}, nil
+		}
+		return nil, err
+	}
+
+	for _, p := range paths {
+		if err := os.MkdirAll(path.Join(root, p), 0755); err != nil {
+			return nil, err
+		}
+	}
+	return &Driver{root}, nil
+}
+
+// Return a nil error if the kernel supports aufs
+// We cannot modprobe because inside dind modprobe fails
+// to run
+func supportsAufs() error {
+	// We can try to modprobe aufs first before looking at
+	// proc/filesystems for when aufs is supported
+	exec.Command("modprobe", "aufs").Run()
+
+	f, err := os.Open("/proc/filesystems")
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+
+	s := bufio.NewScanner(f)
+	for s.Scan() {
+		if strings.Contains(s.Text(), "aufs") {
+			return nil
+		}
+	}
+	return fmt.Errorf("AUFS was not found in /proc/filesystems")
+}
+
+func (a Driver) rootPath() string {
+	return a.root
+}
+
+func (Driver) String() string {
+	return "aufs"
+}
+
+func (a Driver) Status() [][2]string {
+	ids, _ := loadIds(path.Join(a.rootPath(), "layers"))
+	return [][2]string{
+		{"Root Dir", a.rootPath()},
+		{"Dirs", fmt.Sprintf("%d", len(ids))},
+	}
+}
+
+// Exists returns true if the given id is registered with
+// this driver
+func (a Driver) Exists(id string) bool {
+	if _, err := os.Lstat(path.Join(a.rootPath(), "layers", id)); err != nil {
+		return false
+	}
+	return true
+}
+
+// Three folders are created for each id
+// mnt, layers, and diff
+func (a *Driver) Create(id, parent string) error {
+	if err := a.createDirsFor(id); err != nil {
+		return err
+	}
+	// Write the layers metadata
+	f, err := os.Create(path.Join(a.rootPath(), "layers", id))
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+
+	if parent != "" {
+		ids, err := getParentIds(a.rootPath(), parent)
+		if err != nil {
+			return err
+		}
+
+		if _, err := fmt.Fprintln(f, parent); err != nil {
+			return err
+		}
+		for _, i := range ids {
+			if _, err := fmt.Fprintln(f, i); err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
+func (a *Driver) createDirsFor(id string) error {
+	paths := []string{
+		"mnt",
+		"diff",
+	}
+
+	for _, p := range paths {
+		if err := os.MkdirAll(path.Join(a.rootPath(), p, id), 0755); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// Unmount and remove the dir information
+func (a *Driver) Remove(id string) error {
+	// Make sure the dir is umounted first
+	if err := a.unmount(id); err != nil {
+		return err
+	}
+	tmpDirs := []string{
+		"mnt",
+		"diff",
+	}
+
+	// Remove the dirs atomically
+	for _, p := range tmpDirs {
+		// We need to use a temp dir in the same dir as the driver so Rename
+		// does not fall back to the slow copy if /tmp and the driver dir
+		// are on different devices
+		tmp := path.Join(a.rootPath(), "tmp", p, id)
+		if err := os.MkdirAll(tmp, 0755); err != nil {
+			return err
+		}
+		realPath := path.Join(a.rootPath(), p, id)
+		if err := os.Rename(realPath, tmp); err != nil && !os.IsNotExist(err) {
+			return err
+		}
+		defer os.RemoveAll(tmp)
+	}
+
+	// Remove the layers file for the id
+	if err := os.Remove(path.Join(a.rootPath(), "layers", id)); err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	return nil
+}
+
+// Return the rootfs path for the id
+// This will mount the dir at it's given path
+func (a *Driver) Get(id string) (string, error) {
+	ids, err := getParentIds(a.rootPath(), id)
+	if err != nil {
+		if !os.IsNotExist(err) {
+			return "", err
+		}
+		ids = []string{}
+	}
+
+	// If a dir does not have a parent ( no layers )do not try to mount
+	// just return the diff path to the data
+	out := path.Join(a.rootPath(), "diff", id)
+	if len(ids) > 0 {
+		out = path.Join(a.rootPath(), "mnt", id)
+		if err := a.mount(id); err != nil {
+			return "", err
+		}
+	}
+	return out, nil
+}
+
+// Returns an archive of the contents for the id
+func (a *Driver) Diff(id string) (archive.Archive, error) {
+	return archive.TarFilter(path.Join(a.rootPath(), "diff", id), &archive.TarOptions{
+		Recursive:   true,
+		Compression: archive.Uncompressed,
+	})
+}
+
+func (a *Driver) ApplyDiff(id string, diff archive.Archive) error {
+	return archive.Untar(diff, path.Join(a.rootPath(), "diff", id), nil)
+}
+
+// Returns the size of the contents for the id
+func (a *Driver) DiffSize(id string) (int64, error) {
+	return utils.TreeSize(path.Join(a.rootPath(), "diff", id))
+}
+
+func (a *Driver) Changes(id string) ([]archive.Change, error) {
+	layers, err := a.getParentLayerPaths(id)
+	if err != nil {
+		return nil, err
+	}
+	return archive.Changes(layers, path.Join(a.rootPath(), "diff", id))
+}
+
+func (a *Driver) getParentLayerPaths(id string) ([]string, error) {
+	parentIds, err := getParentIds(a.rootPath(), id)
+	if err != nil {
+		return nil, err
+	}
+	if len(parentIds) == 0 {
+		return nil, fmt.Errorf("Dir %s does not have any parent layers", id)
+	}
+	layers := make([]string, len(parentIds))
+
+	// Get the diff paths for all the parent ids
+	for i, p := range parentIds {
+		layers[i] = path.Join(a.rootPath(), "diff", p)
+	}
+	return layers, nil
+}
+
+func (a *Driver) mount(id string) error {
+	// If the id is mounted or we get an error return
+	if mounted, err := a.mounted(id); err != nil || mounted {
+		return err
+	}
+
+	var (
+		target = path.Join(a.rootPath(), "mnt", id)
+		rw     = path.Join(a.rootPath(), "diff", id)
+	)
+
+	layers, err := a.getParentLayerPaths(id)
+	if err != nil {
+		return err
+	}
+
+	if err := a.aufsMount(layers, rw, target); err != nil {
+		return err
+	}
+	return nil
+}
+
+func (a *Driver) unmount(id string) error {
+	if mounted, err := a.mounted(id); err != nil || !mounted {
+		return err
+	}
+	target := path.Join(a.rootPath(), "mnt", id)
+	return Unmount(target)
+}
+
+func (a *Driver) mounted(id string) (bool, error) {
+	target := path.Join(a.rootPath(), "mnt", id)
+	return Mounted(target)
+}
+
+// During cleanup aufs needs to unmount all mountpoints
+func (a *Driver) Cleanup() error {
+	ids, err := loadIds(path.Join(a.rootPath(), "layers"))
+	if err != nil {
+		return err
+	}
+	for _, id := range ids {
+		if err := a.unmount(id); err != nil {
+			utils.Errorf("Unmounting %s: %s", utils.TruncateID(id), err)
+		}
+	}
+	return nil
+}
+
+func (a *Driver) aufsMount(ro []string, rw, target string) error {
+	rwBranch := fmt.Sprintf("%v=rw", rw)
+	roBranches := ""
+	for _, layer := range ro {
+		roBranches += fmt.Sprintf("%v=ro+wh:", layer)
+	}
+	branches := fmt.Sprintf("br:%v:%v,xino=/dev/shm/aufs.xino", rwBranch, roBranches)
+
+	//if error, try to load aufs kernel module
+	if err := mount("none", target, "aufs", 0, branches); err != nil {
+		log.Printf("Kernel does not support AUFS, trying to load the AUFS module with modprobe...")
+		if err := exec.Command("modprobe", "aufs").Run(); err != nil {
+			return fmt.Errorf("Unable to load the AUFS module")
+		}
+		log.Printf("...module loaded.")
+		if err := mount("none", target, "aufs", 0, branches); err != nil {
+			return fmt.Errorf("Unable to mount using aufs %s", err)
+		}
+	}
+	return nil
+}

graphdriver/aufs/aufs_test.go

@@ -0,0 +1,623 @@
+package aufs
+
+import (
+	"github.com/dotcloud/docker/archive"
+	"os"
+	"path"
+	"testing"
+)
+
+var (
+	tmp = path.Join(os.TempDir(), "aufs-tests", "aufs")
+)
+
+func newDriver(t *testing.T) *Driver {
+	if err := os.MkdirAll(tmp, 0755); err != nil {
+		t.Fatal(err)
+	}
+
+	d, err := Init(tmp)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return d.(*Driver)
+}
+
+func TestNewDriver(t *testing.T) {
+	if err := os.MkdirAll(tmp, 0755); err != nil {
+		t.Fatal(err)
+	}
+
+	d, err := Init(tmp)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer os.RemoveAll(tmp)
+	if d == nil {
+		t.Fatalf("Driver should not be nil")
+	}
+}
+
+func TestAufsString(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if d.String() != "aufs" {
+		t.Fatalf("Expected aufs got %s", d.String())
+	}
+}
+
+func TestCreateDirStructure(t *testing.T) {
+	newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	paths := []string{
+		"mnt",
+		"layers",
+		"diff",
+	}
+
+	for _, p := range paths {
+		if _, err := os.Stat(path.Join(tmp, p)); err != nil {
+			t.Fatal(err)
+		}
+	}
+}
+
+// We should be able to create two drivers with the same dir structure
+func TestNewDriverFromExistingDir(t *testing.T) {
+	if err := os.MkdirAll(tmp, 0755); err != nil {
+		t.Fatal(err)
+	}
+
+	if _, err := Init(tmp); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := Init(tmp); err != nil {
+		t.Fatal(err)
+	}
+	os.RemoveAll(tmp)
+}
+
+func TestCreateNewDir(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestCreateNewDirStructure(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	paths := []string{
+		"mnt",
+		"diff",
+		"layers",
+	}
+
+	for _, p := range paths {
+		if _, err := os.Stat(path.Join(tmp, p, "1")); err != nil {
+			t.Fatal(err)
+		}
+	}
+}
+
+func TestRemoveImage(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := d.Remove("1"); err != nil {
+		t.Fatal(err)
+	}
+
+	paths := []string{
+		"mnt",
+		"diff",
+		"layers",
+	}
+
+	for _, p := range paths {
+		if _, err := os.Stat(path.Join(tmp, p, "1")); err == nil {
+			t.Fatalf("Error should not be nil because dirs with id 1 should be delted: %s", p)
+		}
+	}
+}
+
+func TestGetWithoutParent(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	diffPath, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+	expected := path.Join(tmp, "diff", "1")
+	if diffPath != expected {
+		t.Fatalf("Expected path %s got %s", expected, diffPath)
+	}
+}
+
+func TestCleanupWithNoDirs(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Cleanup(); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestCleanupWithDir(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := d.Cleanup(); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestMountedFalseResponse(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	response, err := d.mounted("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if response != false {
+		t.Fatalf("Response if dir id 1 is mounted should be false")
+	}
+}
+
+func TestMountedTrueReponse(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+	defer d.Cleanup()
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	if err := d.Create("2", "1"); err != nil {
+		t.Fatal(err)
+	}
+
+	_, err := d.Get("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	response, err := d.mounted("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if response != true {
+		t.Fatalf("Response if dir id 2 is mounted should be true")
+	}
+}
+
+func TestMountWithParent(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	if err := d.Create("2", "1"); err != nil {
+		t.Fatal(err)
+	}
+
+	defer func() {
+		if err := d.Cleanup(); err != nil {
+			t.Fatal(err)
+		}
+	}()
+
+	mntPath, err := d.Get("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if mntPath == "" {
+		t.Fatal("mntPath should not be empty string")
+	}
+
+	expected := path.Join(tmp, "mnt", "2")
+	if mntPath != expected {
+		t.Fatalf("Expected %s got %s", expected, mntPath)
+	}
+}
+
+func TestRemoveMountedDir(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	if err := d.Create("2", "1"); err != nil {
+		t.Fatal(err)
+	}
+
+	defer func() {
+		if err := d.Cleanup(); err != nil {
+			t.Fatal(err)
+		}
+	}()
+
+	mntPath, err := d.Get("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if mntPath == "" {
+		t.Fatal("mntPath should not be empty string")
+	}
+
+	mounted, err := d.mounted("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if !mounted {
+		t.Fatalf("Dir id 2 should be mounted")
+	}
+
+	if err := d.Remove("2"); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestCreateWithInvalidParent(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", "docker"); err == nil {
+		t.Fatalf("Error should not be nil with parent does not exist")
+	}
+}
+
+func TestGetDiff(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	diffPath, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Add a file to the diff path with a fixed size
+	size := int64(1024)
+
+	f, err := os.Create(path.Join(diffPath, "test_file"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Truncate(size); err != nil {
+		t.Fatal(err)
+	}
+	f.Close()
+
+	a, err := d.Diff("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if a == nil {
+		t.Fatalf("Archive should not be nil")
+	}
+}
+
+func TestChanges(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	if err := d.Create("2", "1"); err != nil {
+		t.Fatal(err)
+	}
+
+	defer func() {
+		if err := d.Cleanup(); err != nil {
+			t.Fatal(err)
+		}
+	}()
+
+	mntPoint, err := d.Get("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Create a file to save in the mountpoint
+	f, err := os.Create(path.Join(mntPoint, "test.txt"))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if _, err := f.WriteString("testline"); err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Close(); err != nil {
+		t.Fatal(err)
+	}
+
+	changes, err := d.Changes("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(changes) != 1 {
+		t.Fatalf("Dir 2 should have one change from parent got %d", len(changes))
+	}
+	change := changes[0]
+
+	expectedPath := "/test.txt"
+	if change.Path != expectedPath {
+		t.Fatalf("Expected path %s got %s", expectedPath, change.Path)
+	}
+
+	if change.Kind != archive.ChangeAdd {
+		t.Fatalf("Change kind should be ChangeAdd got %s", change.Kind)
+	}
+
+	if err := d.Create("3", "2"); err != nil {
+		t.Fatal(err)
+	}
+	mntPoint, err = d.Get("3")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Create a file to save in the mountpoint
+	f, err = os.Create(path.Join(mntPoint, "test2.txt"))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if _, err := f.WriteString("testline"); err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Close(); err != nil {
+		t.Fatal(err)
+	}
+
+	changes, err = d.Changes("3")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if len(changes) != 1 {
+		t.Fatalf("Dir 2 should have one change from parent got %d", len(changes))
+	}
+	change = changes[0]
+
+	expectedPath = "/test2.txt"
+	if change.Path != expectedPath {
+		t.Fatalf("Expected path %s got %s", expectedPath, change.Path)
+	}
+
+	if change.Kind != archive.ChangeAdd {
+		t.Fatalf("Change kind should be ChangeAdd got %s", change.Kind)
+	}
+}
+
+func TestDiffSize(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	diffPath, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Add a file to the diff path with a fixed size
+	size := int64(1024)
+
+	f, err := os.Create(path.Join(diffPath, "test_file"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Truncate(size); err != nil {
+		t.Fatal(err)
+	}
+	s, err := f.Stat()
+	if err != nil {
+		t.Fatal(err)
+	}
+	size = s.Size()
+	if err := f.Close(); err != nil {
+		t.Fatal(err)
+	}
+
+	diffSize, err := d.DiffSize("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if diffSize != size {
+		t.Fatalf("Expected size to be %d got %d", size, diffSize)
+	}
+}
+
+func TestChildDiffSize(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+	defer d.Cleanup()
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	diffPath, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Add a file to the diff path with a fixed size
+	size := int64(1024)
+
+	f, err := os.Create(path.Join(diffPath, "test_file"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Truncate(size); err != nil {
+		t.Fatal(err)
+	}
+	s, err := f.Stat()
+	if err != nil {
+		t.Fatal(err)
+	}
+	size = s.Size()
+	if err := f.Close(); err != nil {
+		t.Fatal(err)
+	}
+
+	diffSize, err := d.DiffSize("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if diffSize != size {
+		t.Fatalf("Expected size to be %d got %d", size, diffSize)
+	}
+
+	if err := d.Create("2", "1"); err != nil {
+		t.Fatal(err)
+	}
+
+	diffSize, err = d.DiffSize("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+	// The diff size for the child should be zero
+	if diffSize != 0 {
+		t.Fatalf("Expected size to be %d got %d", 0, diffSize)
+	}
+}
+
+func TestExists(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+	defer d.Cleanup()
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	if d.Exists("none") {
+		t.Fatal("id name should not exist in the driver")
+	}
+
+	if !d.Exists("1") {
+		t.Fatal("id 1 should exist in the driver")
+	}
+}
+
+func TestStatus(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+	defer d.Cleanup()
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	status := d.Status()
+	if status == nil || len(status) == 0 {
+		t.Fatal("Status should not be nil or empty")
+	}
+	rootDir := status[0]
+	dirs := status[1]
+	if rootDir[0] != "Root Dir" {
+		t.Fatalf("Expected Root Dir got %s", rootDir[0])
+	}
+	if rootDir[1] != d.rootPath() {
+		t.Fatalf("Expected %s got %s", d.rootPath(), rootDir[1])
+	}
+	if dirs[0] != "Dirs" {
+		t.Fatalf("Expected Dirs got %s", dirs[0])
+	}
+	if dirs[1] != "1" {
+		t.Fatalf("Expected 1 got %s", dirs[1])
+	}
+}
+
+func TestApplyDiff(t *testing.T) {
+	d := newDriver(t)
+	defer os.RemoveAll(tmp)
+	defer d.Cleanup()
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	diffPath, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Add a file to the diff path with a fixed size
+	size := int64(1024)
+
+	f, err := os.Create(path.Join(diffPath, "test_file"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Truncate(size); err != nil {
+		t.Fatal(err)
+	}
+	f.Close()
+
+	diff, err := d.Diff("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if err := d.Create("2", ""); err != nil {
+		t.Fatal(err)
+	}
+	if err := d.Create("3", "2"); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := d.ApplyDiff("3", diff); err != nil {
+		t.Fatal(err)
+	}
+
+	// Ensure that the file is in the mount point for id 3
+
+	mountPoint, err := d.Get("3")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if _, err := os.Stat(path.Join(mountPoint, "test_file")); err != nil {
+		t.Fatal(err)
+	}
+}

graphdriver/aufs/dirs.go

@@ -0,0 +1,46 @@
+package aufs
+
+import (
+	"bufio"
+	"io/ioutil"
+	"os"
+	"path"
+)
+
+// Return all the directories
+func loadIds(root string) ([]string, error) {
+	dirs, err := ioutil.ReadDir(root)
+	if err != nil {
+		return nil, err
+	}
+	out := []string{}
+	for _, d := range dirs {
+		if !d.IsDir() {
+			out = append(out, d.Name())
+		}
+	}
+	return out, nil
+}
+
+// Read the layers file for the current id and return all the
+// layers represented by new lines in the file
+//
+// If there are no lines in the file then the id has no parent
+// and an empty slice is returned.
+func getParentIds(root, id string) ([]string, error) {
+	f, err := os.Open(path.Join(root, "layers", id))
+	if err != nil {
+		return nil, err
+	}
+	defer f.Close()
+
+	out := []string{}
+	s := bufio.NewScanner(f)
+
+	for s.Scan() {
+		if t := s.Text(); t != "" {
+			out = append(out, s.Text())
+		}
+	}
+	return out, s.Err()
+}

graphdriver/aufs/migrate.go

@@ -0,0 +1,194 @@
+package aufs
+
+import (
+	"encoding/json"
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path"
+)
+
+type metadata struct {
+	ID       string `json:"id"`
+	ParentID string `json:"parent,omitempty"`
+	Image    string `json:"Image,omitempty"`
+
+	parent *metadata
+}
+
+func pathExists(pth string) bool {
+	if _, err := os.Stat(pth); err != nil {
+		return false
+	}
+	return true
+}
+
+// Migrate existing images and containers from docker < 0.7.x
+//
+// The format pre 0.7 is for docker to store the metadata and filesystem
+// content in the same directory.  For the migration to work we need to move Image layer
+// data from /var/lib/docker/graph/<id>/layers to the diff of the registered id.
+//
+// Next we need to migrate the container's rw layer to diff of the driver.  After the
+// contents are migrated we need to register the image and container ids with the
+// driver.
+//
+// For the migration we try to move the folder containing the layer files, if that
+// fails because the data is currently mounted we will fallback to creating a
+// symlink.
+func (a *Driver) Migrate(pth string, setupInit func(p string) error) error {
+	if pathExists(path.Join(pth, "graph")) {
+		if err := a.migrateRepositories(pth); err != nil {
+			return err
+		}
+		if err := a.migrateImages(path.Join(pth, "graph")); err != nil {
+			return err
+		}
+		return a.migrateContainers(path.Join(pth, "containers"), setupInit)
+	}
+	return nil
+}
+
+func (a *Driver) migrateRepositories(pth string) error {
+	name := path.Join(pth, "repositories")
+	if err := os.Rename(name, name+"-aufs"); err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	return nil
+}
+
+func (a *Driver) migrateContainers(pth string, setupInit func(p string) error) error {
+	fis, err := ioutil.ReadDir(pth)
+	if err != nil {
+		return err
+	}
+
+	for _, fi := range fis {
+		if id := fi.Name(); fi.IsDir() && pathExists(path.Join(pth, id, "rw")) {
+			if err := tryRelocate(path.Join(pth, id, "rw"), path.Join(a.rootPath(), "diff", id)); err != nil {
+				return err
+			}
+
+			if !a.Exists(id) {
+
+				metadata, err := loadMetadata(path.Join(pth, id, "config.json"))
+				if err != nil {
+					return err
+				}
+
+				initID := fmt.Sprintf("%s-init", id)
+				if err := a.Create(initID, metadata.Image); err != nil {
+					return err
+				}
+
+				initPath, err := a.Get(initID)
+				if err != nil {
+					return err
+				}
+				// setup init layer
+				if err := setupInit(initPath); err != nil {
+					return err
+				}
+
+				if err := a.Create(id, initID); err != nil {
+					return err
+				}
+			}
+		}
+	}
+	return nil
+}
+
+func (a *Driver) migrateImages(pth string) error {
+	fis, err := ioutil.ReadDir(pth)
+	if err != nil {
+		return err
+	}
+	var (
+		m       = make(map[string]*metadata)
+		current *metadata
+		exists  bool
+	)
+
+	for _, fi := range fis {
+		if id := fi.Name(); fi.IsDir() && pathExists(path.Join(pth, id, "layer")) {
+			if current, exists = m[id]; !exists {
+				current, err = loadMetadata(path.Join(pth, id, "json"))
+				if err != nil {
+					return err
+				}
+				m[id] = current
+			}
+		}
+	}
+
+	for _, v := range m {
+		v.parent = m[v.ParentID]
+	}
+
+	migrated := make(map[string]bool)
+	for _, v := range m {
+		if err := a.migrateImage(v, pth, migrated); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func (a *Driver) migrateImage(m *metadata, pth string, migrated map[string]bool) error {
+	if !migrated[m.ID] {
+		if m.parent != nil {
+			a.migrateImage(m.parent, pth, migrated)
+		}
+		if err := tryRelocate(path.Join(pth, m.ID, "layer"), path.Join(a.rootPath(), "diff", m.ID)); err != nil {
+			return err
+		}
+		if !a.Exists(m.ID) {
+			if err := a.Create(m.ID, m.ParentID); err != nil {
+				return err
+			}
+		}
+		migrated[m.ID] = true
+	}
+	return nil
+}
+
+// tryRelocate will try to rename the old path to the new pack and if
+// the operation fails, it will fallback to a symlink
+func tryRelocate(oldPath, newPath string) error {
+	s, err := os.Lstat(newPath)
+	if err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	// If the destination is a symlink then we already tried to relocate once before
+	// and it failed so we delete it and try to remove
+	if s != nil && s.Mode()&os.ModeSymlink == os.ModeSymlink {
+		if err := os.RemoveAll(newPath); err != nil {
+			return err
+		}
+	}
+	if err := os.Rename(oldPath, newPath); err != nil {
+		if sErr := os.Symlink(oldPath, newPath); sErr != nil {
+			return fmt.Errorf("Unable to relocate %s to %s: Rename err %s Symlink err %s", oldPath, newPath, err, sErr)
+		}
+	}
+	return nil
+}
+
+func loadMetadata(pth string) (*metadata, error) {
+	f, err := os.Open(pth)
+	if err != nil {
+		return nil, err
+	}
+	defer f.Close()
+
+	var (
+		out = &metadata{}
+		dec = json.NewDecoder(f)
+	)
+
+	if err := dec.Decode(out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}

graphdriver/aufs/mount.go

@@ -1,13 +1,11 @@
-package docker
+package aufs
 
 import (
-	"fmt"
 	"github.com/dotcloud/docker/utils"
 	"os"
 	"os/exec"
 	"path/filepath"
 	"syscall"
-	"time"
 )
 
 func Unmount(target string) error {
@@ -17,22 +15,7 @@ func Unmount(target string) error {
 	if err := syscall.Unmount(target, 0); err != nil {
 		return err
 	}
-	// Even though we just unmounted the filesystem, AUFS will prevent deleting the mntpoint
-	// for some time. We'll just keep retrying until it succeeds.
-	for retries := 0; retries < 1000; retries++ {
-		err := os.Remove(target)
-		if err == nil {
-			// rm mntpoint succeeded
-			return nil
-		}
-		if os.IsNotExist(err) {
-			// mntpoint doesn't exist anymore. Success.
-			return nil
-		}
-		// fmt.Printf("(%v) Remove %v returned: %v\n", retries, target, err)
-		time.Sleep(10 * time.Millisecond)
-	}
-	return fmt.Errorf("Umount: Failed to umount %v", target)
+	return nil
 }
 
 func Mounted(mountpoint string) (bool, error) {
@@ -49,5 +32,6 @@ func Mounted(mountpoint string) (bool, error) {
 	}
 	mntpointSt := mntpoint.Sys().(*syscall.Stat_t)
 	parentSt := parent.Sys().(*syscall.Stat_t)
+
 	return mntpointSt.Dev != parentSt.Dev, nil
 }

graphdriver/aufs/mount_darwin.go

@@ -1,4 +1,4 @@
-package docker
+package aufs
 
 import "errors"
 

graphdriver/aufs/mount_linux.go

@@ -1,4 +1,4 @@
-package docker
+package aufs
 
 import "syscall"
 

graphdriver/devmapper/attach_loopback.go

@@ -0,0 +1,126 @@
+// +build linux
+
+package devmapper
+
+import (
+	"fmt"
+	"github.com/dotcloud/docker/utils"
+)
+
+func stringToLoopName(src string) [LoNameSize]uint8 {
+	var dst [LoNameSize]uint8
+	copy(dst[:], src[:])
+	return dst
+}
+
+func getNextFreeLoopbackIndex() (int, error) {
+	f, err := osOpenFile("/dev/loop-control", osORdOnly, 0644)
+	if err != nil {
+		return 0, err
+	}
+	defer f.Close()
+
+	index, err := ioctlLoopCtlGetFree(f.Fd())
+	if index < 0 {
+		index = 0
+	}
+	return index, err
+}
+
+func openNextAvailableLoopback(index int, sparseFile *osFile) (loopFile *osFile, err error) {
+	// Start looking for a free /dev/loop
+	for {
+		target := fmt.Sprintf("/dev/loop%d", index)
+		index++
+
+		fi, err := osStat(target)
+		if err != nil {
+			if osIsNotExist(err) {
+				utils.Errorf("There are no more loopback device available.")
+			}
+			return nil, ErrAttachLoopbackDevice
+		}
+
+		if fi.Mode()&osModeDevice != osModeDevice {
+			utils.Errorf("Loopback device %s is not a block device.", target)
+			continue
+		}
+
+		// OpenFile adds O_CLOEXEC
+		loopFile, err = osOpenFile(target, osORdWr, 0644)
+		if err != nil {
+			utils.Errorf("Error openning loopback device: %s", err)
+			return nil, ErrAttachLoopbackDevice
+		}
+
+		// Try to attach to the loop file
+		if err := ioctlLoopSetFd(loopFile.Fd(), sparseFile.Fd()); err != nil {
+			loopFile.Close()
+
+			// If the error is EBUSY, then try the next loopback
+			if err != sysEBusy {
+				utils.Errorf("Cannot set up loopback device %s: %s", target, err)
+				return nil, ErrAttachLoopbackDevice
+			}
+
+			// Otherwise, we keep going with the loop
+			continue
+		}
+		// In case of success, we finished. Break the loop.
+		break
+	}
+
+	// This can't happen, but let's be sure
+	if loopFile == nil {
+		utils.Errorf("Unreachable code reached! Error attaching %s to a loopback device.", sparseFile.Name())
+		return nil, ErrAttachLoopbackDevice
+	}
+
+	return loopFile, nil
+}
+
+// attachLoopDevice attaches the given sparse file to the next
+// available loopback device. It returns an opened *osFile.
+func attachLoopDevice(sparseName string) (loop *osFile, err error) {
+
+	// Try to retrieve the next available loopback device via syscall.
+	// If it fails, we discard error and start loopking for a
+	// loopback from index 0.
+	startIndex, err := getNextFreeLoopbackIndex()
+	if err != nil {
+		utils.Debugf("Error retrieving the next available loopback: %s", err)
+	}
+
+	// OpenFile adds O_CLOEXEC
+	sparseFile, err := osOpenFile(sparseName, osORdWr, 0644)
+	if err != nil {
+		utils.Errorf("Error openning sparse file %s: %s", sparseName, err)
+		return nil, ErrAttachLoopbackDevice
+	}
+	defer sparseFile.Close()
+
+	loopFile, err := openNextAvailableLoopback(startIndex, sparseFile)
+	if err != nil {
+		return nil, err
+	}
+
+	// Set the status of the loopback device
+	loopInfo := &LoopInfo64{
+		loFileName: stringToLoopName(loopFile.Name()),
+		loOffset:   0,
+		loFlags:    LoFlagsAutoClear,
+	}
+
+	if err := ioctlLoopSetStatus64(loopFile.Fd(), loopInfo); err != nil {
+		utils.Errorf("Cannot set up loopback device info: %s", err)
+
+		// If the call failed, then free the loopback device
+		if err := ioctlLoopClrFd(loopFile.Fd()); err != nil {
+			utils.Errorf("Error while cleaning up the loopback device")
+		}
+		loopFile.Close()
+		return nil, ErrAttachLoopbackDevice
+	}
+
+	return loopFile, nil
+}

graphdriver/devmapper/deviceset.go

@@ -0,0 +1,959 @@
+// +build linux
+
+package devmapper
+
+import (
+	"encoding/json"
+	"errors"
+	"fmt"
+	"github.com/dotcloud/docker/utils"
+	"io"
+	"io/ioutil"
+	"path"
+	"path/filepath"
+	"strconv"
+	"sync"
+	"time"
+)
+
+var (
+	DefaultDataLoopbackSize     int64  = 100 * 1024 * 1024 * 1024
+	DefaultMetaDataLoopbackSize int64  = 2 * 1024 * 1024 * 1024
+	DefaultBaseFsSize           uint64 = 10 * 1024 * 1024 * 1024
+)
+
+type DevInfo struct {
+	Hash          string     `json:"-"`
+	DeviceId      int        `json:"device_id"`
+	Size          uint64     `json:"size"`
+	TransactionId uint64     `json:"transaction_id"`
+	Initialized   bool       `json:"initialized"`
+	devices       *DeviceSet `json:"-"`
+}
+
+type MetaData struct {
+	Devices map[string]*DevInfo `json:devices`
+}
+
+type DeviceSet struct {
+	MetaData
+	sync.Mutex
+	root             string
+	devicePrefix     string
+	TransactionId    uint64
+	NewTransactionId uint64
+	nextFreeDevice   int
+	activeMounts     map[string]int
+}
+
+type DiskUsage struct {
+	Used  uint64
+	Total uint64
+}
+
+type Status struct {
+	PoolName         string
+	DataLoopback     string
+	MetadataLoopback string
+	Data             DiskUsage
+	Metadata         DiskUsage
+	SectorSize       uint64
+}
+
+type DevStatus struct {
+	DeviceId            int
+	Size                uint64
+	TransactionId       uint64
+	SizeInSectors       uint64
+	MappedSectors       uint64
+	HighestMappedSector uint64
+}
+
+func getDevName(name string) string {
+	return "/dev/mapper/" + name
+}
+
+func (info *DevInfo) Name() string {
+	hash := info.Hash
+	if hash == "" {
+		hash = "base"
+	}
+	return fmt.Sprintf("%s-%s", info.devices.devicePrefix, hash)
+}
+
+func (info *DevInfo) DevName() string {
+	return getDevName(info.Name())
+}
+
+func (devices *DeviceSet) loopbackDir() string {
+	return path.Join(devices.root, "devicemapper")
+}
+
+func (devices *DeviceSet) jsonFile() string {
+	return path.Join(devices.loopbackDir(), "json")
+}
+
+func (devices *DeviceSet) getPoolName() string {
+	return devices.devicePrefix + "-pool"
+}
+
+func (devices *DeviceSet) getPoolDevName() string {
+	return getDevName(devices.getPoolName())
+}
+
+func (devices *DeviceSet) hasImage(name string) bool {
+	dirname := devices.loopbackDir()
+	filename := path.Join(dirname, name)
+
+	_, err := osStat(filename)
+	return err == nil
+}
+
+// ensureImage creates a sparse file of <size> bytes at the path
+// <root>/devicemapper/<name>.
+// If the file already exists, it does nothing.
+// Either way it returns the full path.
+func (devices *DeviceSet) ensureImage(name string, size int64) (string, error) {
+	dirname := devices.loopbackDir()
+	filename := path.Join(dirname, name)
+
+	if err := osMkdirAll(dirname, 0700); err != nil && !osIsExist(err) {
+		return "", err
+	}
+
+	if _, err := osStat(filename); err != nil {
+		if !osIsNotExist(err) {
+			return "", err
+		}
+		utils.Debugf("Creating loopback file %s for device-manage use", filename)
+		file, err := osOpenFile(filename, osORdWr|osOCreate, 0600)
+		if err != nil {
+			return "", err
+		}
+		defer file.Close()
+
+		if err = file.Truncate(size); err != nil {
+			return "", err
+		}
+	}
+	return filename, nil
+}
+
+func (devices *DeviceSet) allocateDeviceId() int {
+	// TODO: Add smarter reuse of deleted devices
+	id := devices.nextFreeDevice
+	devices.nextFreeDevice = devices.nextFreeDevice + 1
+	return id
+}
+
+func (devices *DeviceSet) allocateTransactionId() uint64 {
+	devices.NewTransactionId = devices.NewTransactionId + 1
+	return devices.NewTransactionId
+}
+
+func (devices *DeviceSet) saveMetadata() error {
+	jsonData, err := json.Marshal(devices.MetaData)
+	if err != nil {
+		return fmt.Errorf("Error encoding metaadata to json: %s", err)
+	}
+	tmpFile, err := ioutil.TempFile(filepath.Dir(devices.jsonFile()), ".json")
+	if err != nil {
+		return fmt.Errorf("Error creating metadata file: %s", err)
+	}
+
+	n, err := tmpFile.Write(jsonData)
+	if err != nil {
+		return fmt.Errorf("Error writing metadata to %s: %s", tmpFile.Name(), err)
+	}
+	if n < len(jsonData) {
+		return io.ErrShortWrite
+	}
+	if err := tmpFile.Sync(); err != nil {
+		return fmt.Errorf("Error syncing metadata file %s: %s", tmpFile.Name(), err)
+	}
+	if err := tmpFile.Close(); err != nil {
+		return fmt.Errorf("Error closing metadata file %s: %s", tmpFile.Name(), err)
+	}
+	if err := osRename(tmpFile.Name(), devices.jsonFile()); err != nil {
+		return fmt.Errorf("Error committing metadata file %s: %s", tmpFile.Name(), err)
+	}
+
+	if devices.NewTransactionId != devices.TransactionId {
+		if err = setTransactionId(devices.getPoolDevName(), devices.TransactionId, devices.NewTransactionId); err != nil {
+			return fmt.Errorf("Error setting devmapper transition ID: %s", err)
+		}
+		devices.TransactionId = devices.NewTransactionId
+	}
+	return nil
+}
+
+func (devices *DeviceSet) registerDevice(id int, hash string, size uint64) (*DevInfo, error) {
+	utils.Debugf("registerDevice(%v, %v)", id, hash)
+	info := &DevInfo{
+		Hash:          hash,
+		DeviceId:      id,
+		Size:          size,
+		TransactionId: devices.allocateTransactionId(),
+		Initialized:   false,
+		devices:       devices,
+	}
+
+	devices.Devices[hash] = info
+	if err := devices.saveMetadata(); err != nil {
+		// Try to remove unused device
+		delete(devices.Devices, hash)
+		return nil, err
+	}
+
+	return info, nil
+}
+
+func (devices *DeviceSet) activateDeviceIfNeeded(hash string) error {
+	utils.Debugf("activateDeviceIfNeeded(%v)", hash)
+	info := devices.Devices[hash]
+	if info == nil {
+		return fmt.Errorf("Unknown device %s", hash)
+	}
+
+	if devinfo, _ := getInfo(info.Name()); devinfo != nil && devinfo.Exists != 0 {
+		return nil
+	}
+
+	return activateDevice(devices.getPoolDevName(), info.Name(), info.DeviceId, info.Size)
+}
+
+func (devices *DeviceSet) createFilesystem(info *DevInfo) error {
+	devname := info.DevName()
+
+	err := execRun("mkfs.ext4", "-E", "discard,lazy_itable_init=0,lazy_journal_init=0", devname)
+	if err != nil {
+		err = execRun("mkfs.ext4", "-E", "discard,lazy_itable_init=0", devname)
+	}
+	if err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+	return nil
+}
+
+func (devices *DeviceSet) loadMetaData() error {
+	utils.Debugf("loadMetadata()")
+	defer utils.Debugf("loadMetadata END")
+	_, _, _, params, err := getStatus(devices.getPoolName())
+	if err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	if _, err := fmt.Sscanf(params, "%d", &devices.TransactionId); err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+	devices.NewTransactionId = devices.TransactionId
+
+	jsonData, err := ioutil.ReadFile(devices.jsonFile())
+	if err != nil && !osIsNotExist(err) {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	devices.MetaData.Devices = make(map[string]*DevInfo)
+	if jsonData != nil {
+		if err := json.Unmarshal(jsonData, &devices.MetaData); err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+	}
+
+	for hash, d := range devices.Devices {
+		d.Hash = hash
+		d.devices = devices
+
+		if d.DeviceId >= devices.nextFreeDevice {
+			devices.nextFreeDevice = d.DeviceId + 1
+		}
+
+		// If the transaction id is larger than the actual one we lost the device due to some crash
+		if d.TransactionId > devices.TransactionId {
+			utils.Debugf("Removing lost device %s with id %d", hash, d.TransactionId)
+			delete(devices.Devices, hash)
+		}
+	}
+	return nil
+}
+
+func (devices *DeviceSet) setupBaseImage() error {
+	oldInfo := devices.Devices[""]
+	if oldInfo != nil && oldInfo.Initialized {
+		return nil
+	}
+
+	if oldInfo != nil && !oldInfo.Initialized {
+		utils.Debugf("Removing uninitialized base image")
+		if err := devices.removeDevice(""); err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+	}
+
+	utils.Debugf("Initializing base device-manager snapshot")
+
+	id := devices.allocateDeviceId()
+
+	// Create initial device
+	if err := createDevice(devices.getPoolDevName(), id); err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	utils.Debugf("Registering base device (id %v) with FS size %v", id, DefaultBaseFsSize)
+	info, err := devices.registerDevice(id, "", DefaultBaseFsSize)
+	if err != nil {
+		_ = deleteDevice(devices.getPoolDevName(), id)
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	utils.Debugf("Creating filesystem on base device-manager snapshot")
+
+	if err = devices.activateDeviceIfNeeded(""); err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	if err := devices.createFilesystem(info); err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	info.Initialized = true
+	if err = devices.saveMetadata(); err != nil {
+		info.Initialized = false
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	return nil
+}
+
+func setCloseOnExec(name string) {
+	if fileInfos, _ := ioutil.ReadDir("/proc/self/fd"); fileInfos != nil {
+		for _, i := range fileInfos {
+			link, _ := osReadlink(filepath.Join("/proc/self/fd", i.Name()))
+			if link == name {
+				fd, err := strconv.Atoi(i.Name())
+				if err == nil {
+					sysCloseOnExec(fd)
+				}
+			}
+		}
+	}
+}
+
+func (devices *DeviceSet) log(level int, file string, line int, dmError int, message string) {
+	if level >= 7 {
+		return // Ignore _LOG_DEBUG
+	}
+
+	utils.Debugf("libdevmapper(%d): %s:%d (%d) %s", level, file, line, dmError, message)
+}
+
+func major(device uint64) uint64 {
+	return (device >> 8) & 0xfff
+}
+
+func minor(device uint64) uint64 {
+	return (device & 0xff) | ((device >> 12) & 0xfff00)
+}
+
+func (devices *DeviceSet) ResizePool(size int64) error {
+	dirname := devices.loopbackDir()
+	datafilename := path.Join(dirname, "data")
+	metadatafilename := path.Join(dirname, "metadata")
+
+	datafile, err := osOpenFile(datafilename, osORdWr, 0)
+	if datafile == nil {
+		return err
+	}
+	defer datafile.Close()
+
+	fi, err := datafile.Stat()
+	if fi == nil {
+		return err
+	}
+
+	if fi.Size() > size {
+		return fmt.Errorf("Can't shrink file")
+	}
+
+	dataloopback := FindLoopDeviceFor(datafile)
+	if dataloopback == nil {
+		return fmt.Errorf("Unable to find loopback mount for: %s", datafilename)
+	}
+	defer dataloopback.Close()
+
+	metadatafile, err := osOpenFile(metadatafilename, osORdWr, 0)
+	if metadatafile == nil {
+		return err
+	}
+	defer metadatafile.Close()
+
+	metadataloopback := FindLoopDeviceFor(metadatafile)
+	if metadataloopback == nil {
+		return fmt.Errorf("Unable to find loopback mount for: %s", metadatafilename)
+	}
+	defer metadataloopback.Close()
+
+	// Grow loopback file
+	if err := datafile.Truncate(size); err != nil {
+		return fmt.Errorf("Unable to grow loopback file: %s", err)
+	}
+
+	// Reload size for loopback device
+	if err := LoopbackSetCapacity(dataloopback); err != nil {
+		return fmt.Errorf("Unable to update loopback capacity: %s", err)
+	}
+
+	// Suspend the pool
+	if err := suspendDevice(devices.getPoolName()); err != nil {
+		return fmt.Errorf("Unable to suspend pool: %s", err)
+	}
+
+	// Reload with the new block sizes
+	if err := reloadPool(devices.getPoolName(), dataloopback, metadataloopback); err != nil {
+		return fmt.Errorf("Unable to reload pool: %s", err)
+	}
+
+	// Resume the pool
+	if err := resumeDevice(devices.getPoolName()); err != nil {
+		return fmt.Errorf("Unable to resume pool: %s", err)
+	}
+
+	return nil
+}
+
+func (devices *DeviceSet) initDevmapper(doInit bool) error {
+	logInit(devices)
+
+	// Make sure the sparse images exist in <root>/devicemapper/data and
+	// <root>/devicemapper/metadata
+
+	hasData := devices.hasImage("data")
+	hasMetadata := devices.hasImage("metadata")
+
+	if !doInit && !hasData {
+		return errors.New("Loopback data file not found")
+	}
+
+	if !doInit && !hasMetadata {
+		return errors.New("Loopback metadata file not found")
+	}
+
+	createdLoopback := !hasData || !hasMetadata
+	data, err := devices.ensureImage("data", DefaultDataLoopbackSize)
+	if err != nil {
+		utils.Debugf("Error device ensureImage (data): %s\n", err)
+		return err
+	}
+	metadata, err := devices.ensureImage("metadata", DefaultMetaDataLoopbackSize)
+	if err != nil {
+		utils.Debugf("Error device ensureImage (metadata): %s\n", err)
+		return err
+	}
+
+	// Set the device prefix from the device id and inode of the docker root dir
+
+	st, err := osStat(devices.root)
+	if err != nil {
+		return fmt.Errorf("Error looking up dir %s: %s", devices.root, err)
+	}
+	sysSt := toSysStatT(st.Sys())
+	// "reg-" stands for "regular file".
+	// In the future we might use "dev-" for "device file", etc.
+	// docker-maj,min[-inode] stands for:
+	//	- Managed by docker
+	//	- The target of this device is at major <maj> and minor <min>
+	//	- If <inode> is defined, use that file inside the device as a loopback image. Otherwise use the device itself.
+	devices.devicePrefix = fmt.Sprintf("docker-%d:%d-%d", major(sysSt.Dev), minor(sysSt.Dev), sysSt.Ino)
+	utils.Debugf("Generated prefix: %s", devices.devicePrefix)
+
+	// Check for the existence of the device <prefix>-pool
+	utils.Debugf("Checking for existence of the pool '%s'", devices.getPoolName())
+	info, err := getInfo(devices.getPoolName())
+	if info == nil {
+		utils.Debugf("Error device getInfo: %s", err)
+		return err
+	}
+
+	// It seems libdevmapper opens this without O_CLOEXEC, and go exec will not close files
+	// that are not Close-on-exec, and lxc-start will die if it inherits any unexpected files,
+	// so we add this badhack to make sure it closes itself
+	setCloseOnExec("/dev/mapper/control")
+
+	// If the pool doesn't exist, create it
+	if info.Exists == 0 {
+		utils.Debugf("Pool doesn't exist. Creating it.")
+
+		dataFile, err := attachLoopDevice(data)
+		if err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+		defer dataFile.Close()
+
+		metadataFile, err := attachLoopDevice(metadata)
+		if err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+		defer metadataFile.Close()
+
+		if err := createPool(devices.getPoolName(), dataFile, metadataFile); err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+	}
+
+	// If we didn't just create the data or metadata image, we need to
+	// load the metadata from the existing file.
+	if !createdLoopback {
+		if err = devices.loadMetaData(); err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+	}
+
+	// Setup the base image
+	if doInit {
+		if err := devices.setupBaseImage(); err != nil {
+			utils.Debugf("Error device setupBaseImage: %s\n", err)
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (devices *DeviceSet) AddDevice(hash, baseHash string) error {
+	devices.Lock()
+	defer devices.Unlock()
+
+	if devices.Devices[hash] != nil {
+		return fmt.Errorf("hash %s already exists", hash)
+	}
+
+	baseInfo := devices.Devices[baseHash]
+	if baseInfo == nil {
+		return fmt.Errorf("Error adding device for '%s': can't find device for parent '%s'", hash, baseHash)
+	}
+
+	deviceId := devices.allocateDeviceId()
+
+	if err := devices.createSnapDevice(devices.getPoolDevName(), deviceId, baseInfo.Name(), baseInfo.DeviceId); err != nil {
+		utils.Debugf("Error creating snap device: %s\n", err)
+		return err
+	}
+
+	if _, err := devices.registerDevice(deviceId, hash, baseInfo.Size); err != nil {
+		deleteDevice(devices.getPoolDevName(), deviceId)
+		utils.Debugf("Error registering device: %s\n", err)
+		return err
+	}
+	return nil
+}
+
+func (devices *DeviceSet) removeDevice(hash string) error {
+	info := devices.Devices[hash]
+	if info == nil {
+		return fmt.Errorf("hash %s doesn't exists", hash)
+	}
+
+	devinfo, _ := getInfo(info.Name())
+	if devinfo != nil && devinfo.Exists != 0 {
+		if err := removeDevice(info.Name()); err != nil {
+			utils.Debugf("Error removing device: %s\n", err)
+			return err
+		}
+	}
+
+	if info.Initialized {
+		info.Initialized = false
+		if err := devices.saveMetadata(); err != nil {
+			utils.Debugf("Error saving meta data: %s\n", err)
+			return err
+		}
+	}
+
+	if err := deleteDevice(devices.getPoolDevName(), info.DeviceId); err != nil {
+		utils.Debugf("Error deleting device: %s\n", err)
+		return err
+	}
+
+	devices.allocateTransactionId()
+	delete(devices.Devices, info.Hash)
+
+	if err := devices.saveMetadata(); err != nil {
+		devices.Devices[info.Hash] = info
+		utils.Debugf("Error saving meta data: %s\n", err)
+		return err
+	}
+
+	return nil
+}
+
+func (devices *DeviceSet) RemoveDevice(hash string) error {
+	devices.Lock()
+	defer devices.Unlock()
+
+	return devices.removeDevice(hash)
+}
+
+func (devices *DeviceSet) deactivateDevice(hash string) error {
+	utils.Debugf("[devmapper] deactivateDevice(%s)", hash)
+	defer utils.Debugf("[devmapper] deactivateDevice END")
+	var devname string
+	// FIXME: shouldn't we just register the pool into devices?
+	devname, err := devices.byHash(hash)
+	if err != nil {
+		return err
+	}
+	devinfo, err := getInfo(devname)
+	if err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+	if devinfo.Exists != 0 {
+		if err := removeDevice(devname); err != nil {
+			utils.Debugf("\n--->Err: %s\n", err)
+			return err
+		}
+		if err := devices.waitRemove(hash); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+// waitRemove blocks until either:
+// a) the device registered at <device_set_prefix>-<hash> is removed,
+// or b) the 1 second timeout expires.
+func (devices *DeviceSet) waitRemove(hash string) error {
+	utils.Debugf("[deviceset %s] waitRemove(%s)", devices.devicePrefix, hash)
+	defer utils.Debugf("[deviceset %s] waitRemove(%) END", devices.devicePrefix, hash)
+	devname, err := devices.byHash(hash)
+	if err != nil {
+		return err
+	}
+	i := 0
+	for ; i < 1000; i += 1 {
+		devinfo, err := getInfo(devname)
+		if err != nil {
+			// If there is an error we assume the device doesn't exist.
+			// The error might actually be something else, but we can't differentiate.
+			return nil
+		}
+		if i%100 == 0 {
+			utils.Debugf("Waiting for removal of %s: exists=%d", devname, devinfo.Exists)
+		}
+		if devinfo.Exists == 0 {
+			break
+		}
+
+		time.Sleep(1 * time.Millisecond)
+	}
+	if i == 1000 {
+		return fmt.Errorf("Timeout while waiting for device %s to be removed", devname)
+	}
+	return nil
+}
+
+// waitClose blocks until either:
+// a) the device registered at <device_set_prefix>-<hash> is closed,
+// or b) the 1 second timeout expires.
+func (devices *DeviceSet) waitClose(hash string) error {
+	devname, err := devices.byHash(hash)
+	if err != nil {
+		return err
+	}
+	i := 0
+	for ; i < 1000; i += 1 {
+		devinfo, err := getInfo(devname)
+		if err != nil {
+			return err
+		}
+		if i%100 == 0 {
+			utils.Debugf("Waiting for unmount of %s: opencount=%d", devname, devinfo.OpenCount)
+		}
+		if devinfo.OpenCount == 0 {
+			break
+		}
+		time.Sleep(1 * time.Millisecond)
+	}
+	if i == 1000 {
+		return fmt.Errorf("Timeout while waiting for device %s to close", devname)
+	}
+	return nil
+}
+
+// byHash is a hack to allow looking up the deviceset's pool by the hash "pool".
+// FIXME: it seems probably cleaner to register the pool in devices.Devices,
+// but I am afraid of arcane implications deep in the devicemapper code,
+// so this will do.
+func (devices *DeviceSet) byHash(hash string) (devname string, err error) {
+	if hash == "pool" {
+		return devices.getPoolDevName(), nil
+	}
+	info := devices.Devices[hash]
+	if info == nil {
+		return "", fmt.Errorf("hash %s doesn't exists", hash)
+	}
+	return info.Name(), nil
+}
+
+func (devices *DeviceSet) Shutdown() error {
+	devices.Lock()
+	defer devices.Unlock()
+
+	utils.Debugf("[deviceset %s] shutdown()", devices.devicePrefix)
+	utils.Debugf("[devmapper] Shutting down DeviceSet: %s", devices.root)
+	defer utils.Debugf("[deviceset %s] shutdown END", devices.devicePrefix)
+
+	for path, count := range devices.activeMounts {
+		for i := count; i > 0; i-- {
+			if err := sysUnmount(path, 0); err != nil {
+				utils.Debugf("Shutdown unmounting %s, error: %s\n", path, err)
+			}
+		}
+		delete(devices.activeMounts, path)
+	}
+
+	for _, d := range devices.Devices {
+		if err := devices.waitClose(d.Hash); err != nil {
+			utils.Errorf("Warning: error waiting for device %s to unmount: %s\n", d.Hash, err)
+		}
+		if err := devices.deactivateDevice(d.Hash); err != nil {
+			utils.Debugf("Shutdown deactivate %s , error: %s\n", d.Hash, err)
+		}
+	}
+
+	pool := devices.getPoolDevName()
+	if devinfo, err := getInfo(pool); err == nil && devinfo.Exists != 0 {
+		if err := devices.deactivateDevice("pool"); err != nil {
+			utils.Debugf("Shutdown deactivate %s , error: %s\n", pool, err)
+		}
+	}
+
+	return nil
+}
+
+func (devices *DeviceSet) MountDevice(hash, path string, readOnly bool) error {
+	devices.Lock()
+	defer devices.Unlock()
+
+	if err := devices.activateDeviceIfNeeded(hash); err != nil {
+		return fmt.Errorf("Error activating devmapper device for '%s': %s", hash, err)
+	}
+
+	info := devices.Devices[hash]
+
+	var flags uintptr = sysMsMgcVal
+
+	if readOnly {
+		flags = flags | sysMsRdOnly
+	}
+
+	err := sysMount(info.DevName(), path, "ext4", flags, "discard")
+	if err != nil && err == sysEInval {
+		err = sysMount(info.DevName(), path, "ext4", flags, "")
+	}
+	if err != nil {
+		return fmt.Errorf("Error mounting '%s' on '%s': %s", info.DevName(), path, err)
+	}
+
+	count := devices.activeMounts[path]
+	devices.activeMounts[path] = count + 1
+
+	return devices.setInitialized(hash)
+}
+
+func (devices *DeviceSet) UnmountDevice(hash, path string, deactivate bool) error {
+	utils.Debugf("[devmapper] UnmountDevice(hash=%s path=%s)", hash, path)
+	defer utils.Debugf("[devmapper] UnmountDevice END")
+	devices.Lock()
+	defer devices.Unlock()
+
+	utils.Debugf("[devmapper] Unmount(%s)", path)
+	if err := sysUnmount(path, 0); err != nil {
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+	utils.Debugf("[devmapper] Unmount done")
+	// Wait for the unmount to be effective,
+	// by watching the value of Info.OpenCount for the device
+	if err := devices.waitClose(hash); err != nil {
+		return err
+	}
+
+	if count := devices.activeMounts[path]; count > 1 {
+		devices.activeMounts[path] = count - 1
+	} else {
+		delete(devices.activeMounts, path)
+	}
+
+	if deactivate {
+		devices.deactivateDevice(hash)
+	}
+
+	return nil
+}
+
+func (devices *DeviceSet) HasDevice(hash string) bool {
+	devices.Lock()
+	defer devices.Unlock()
+
+	return devices.Devices[hash] != nil
+}
+
+func (devices *DeviceSet) HasInitializedDevice(hash string) bool {
+	devices.Lock()
+	defer devices.Unlock()
+
+	info := devices.Devices[hash]
+	return info != nil && info.Initialized
+}
+
+func (devices *DeviceSet) HasActivatedDevice(hash string) bool {
+	devices.Lock()
+	defer devices.Unlock()
+
+	info := devices.Devices[hash]
+	if info == nil {
+		return false
+	}
+	devinfo, _ := getInfo(info.Name())
+	return devinfo != nil && devinfo.Exists != 0
+}
+
+func (devices *DeviceSet) setInitialized(hash string) error {
+	info := devices.Devices[hash]
+	if info == nil {
+		return fmt.Errorf("Unknown device %s", hash)
+	}
+
+	info.Initialized = true
+	if err := devices.saveMetadata(); err != nil {
+		info.Initialized = false
+		utils.Debugf("\n--->Err: %s\n", err)
+		return err
+	}
+
+	return nil
+}
+
+func (devices *DeviceSet) List() []string {
+	devices.Lock()
+	defer devices.Unlock()
+
+	ids := make([]string, len(devices.Devices))
+	i := 0
+	for k := range devices.Devices {
+		ids[i] = k
+		i++
+	}
+	return ids
+}
+
+func (devices *DeviceSet) deviceStatus(devName string) (sizeInSectors, mappedSectors, highestMappedSector uint64, err error) {
+	var params string
+	_, sizeInSectors, _, params, err = getStatus(devName)
+	if err != nil {
+		return
+	}
+	if _, err = fmt.Sscanf(params, "%d %d", &mappedSectors, &highestMappedSector); err == nil {
+		return
+	}
+	return
+}
+
+func (devices *DeviceSet) GetDeviceStatus(hash string) (*DevStatus, error) {
+	devices.Lock()
+	defer devices.Unlock()
+
+	info := devices.Devices[hash]
+	if info == nil {
+		return nil, fmt.Errorf("No device %s", hash)
+	}
+
+	status := &DevStatus{
+		DeviceId:      info.DeviceId,
+		Size:          info.Size,
+		TransactionId: info.TransactionId,
+	}
+
+	if err := devices.activateDeviceIfNeeded(hash); err != nil {
+		return nil, fmt.Errorf("Error activating devmapper device for '%s': %s", hash, err)
+	}
+
+	if sizeInSectors, mappedSectors, highestMappedSector, err := devices.deviceStatus(info.DevName()); err != nil {
+		return nil, err
+	} else {
+		status.SizeInSectors = sizeInSectors
+		status.MappedSectors = mappedSectors
+		status.HighestMappedSector = highestMappedSector
+	}
+
+	return status, nil
+}
+
+func (devices *DeviceSet) poolStatus() (totalSizeInSectors, transactionId, dataUsed, dataTotal, metadataUsed, metadataTotal uint64, err error) {
+	var params string
+	if _, totalSizeInSectors, _, params, err = getStatus(devices.getPoolName()); err == nil {
+		_, err = fmt.Sscanf(params, "%d %d/%d %d/%d", &transactionId, &metadataUsed, &metadataTotal, &dataUsed, &dataTotal)
+	}
+	return
+}
+
+func (devices *DeviceSet) Status() *Status {
+	devices.Lock()
+	defer devices.Unlock()
+
+	status := &Status{}
+
+	status.PoolName = devices.getPoolName()
+	status.DataLoopback = path.Join(devices.loopbackDir(), "data")
+	status.MetadataLoopback = path.Join(devices.loopbackDir(), "metadata")
+
+	totalSizeInSectors, _, dataUsed, dataTotal, metadataUsed, metadataTotal, err := devices.poolStatus()
+	if err == nil {
+		// Convert from blocks to bytes
+		blockSizeInSectors := totalSizeInSectors / dataTotal
+
+		status.Data.Used = dataUsed * blockSizeInSectors * 512
+		status.Data.Total = dataTotal * blockSizeInSectors * 512
+
+		// metadata blocks are always 4k
+		status.Metadata.Used = metadataUsed * 4096
+		status.Metadata.Total = metadataTotal * 4096
+
+		status.SectorSize = blockSizeInSectors * 512
+	}
+
+	return status
+}
+
+func NewDeviceSet(root string, doInit bool) (*DeviceSet, error) {
+	SetDevDir("/dev")
+
+	devices := &DeviceSet{
+		root:         root,
+		MetaData:     MetaData{Devices: make(map[string]*DevInfo)},
+		activeMounts: make(map[string]int),
+	}
+
+	if err := devices.initDevmapper(doInit); err != nil {
+		return nil, err
+	}
+
+	return devices, nil
+}

graphdriver/devmapper/devmapper.go

@@ -0,0 +1,571 @@
+// +build linux
+
+package devmapper
+
+import (
+	"errors"
+	"fmt"
+	"github.com/dotcloud/docker/utils"
+	"runtime"
+)
+
+type DevmapperLogger interface {
+	log(level int, file string, line int, dmError int, message string)
+}
+
+const (
+	DeviceCreate TaskType = iota
+	DeviceReload
+	DeviceRemove
+	DeviceRemoveAll
+	DeviceSuspend
+	DeviceResume
+	DeviceInfo
+	DeviceDeps
+	DeviceRename
+	DeviceVersion
+	DeviceStatus
+	DeviceTable
+	DeviceWaitevent
+	DeviceList
+	DeviceClear
+	DeviceMknodes
+	DeviceListVersions
+	DeviceTargetMsg
+	DeviceSetGeometry
+)
+
+const (
+	AddNodeOnResume AddNodeType = iota
+	AddNodeOnCreate
+)
+
+var (
+	ErrTaskRun                = errors.New("dm_task_run failed")
+	ErrTaskSetName            = errors.New("dm_task_set_name failed")
+	ErrTaskSetMessage         = errors.New("dm_task_set_message failed")
+	ErrTaskSetAddNode         = errors.New("dm_task_set_add_node failed")
+	ErrTaskSetRo              = errors.New("dm_task_set_ro failed")
+	ErrTaskAddTarget          = errors.New("dm_task_add_target failed")
+	ErrTaskSetSector          = errors.New("dm_task_set_sector failed")
+	ErrTaskGetInfo            = errors.New("dm_task_get_info failed")
+	ErrTaskSetCookie          = errors.New("dm_task_set_cookie failed")
+	ErrNilCookie              = errors.New("cookie ptr can't be nil")
+	ErrAttachLoopbackDevice   = errors.New("loopback mounting failed")
+	ErrGetBlockSize           = errors.New("Can't get block size")
+	ErrUdevWait               = errors.New("wait on udev cookie failed")
+	ErrSetDevDir              = errors.New("dm_set_dev_dir failed")
+	ErrGetLibraryVersion      = errors.New("dm_get_library_version failed")
+	ErrCreateRemoveTask       = errors.New("Can't create task of type DeviceRemove")
+	ErrRunRemoveDevice        = errors.New("running removeDevice failed")
+	ErrInvalidAddNode         = errors.New("Invalide AddNoce type")
+	ErrGetLoopbackBackingFile = errors.New("Unable to get loopback backing file")
+	ErrLoopbackSetCapacity    = errors.New("Unable set loopback capacity")
+)
+
+type (
+	Task struct {
+		unmanaged *CDmTask
+	}
+	Info struct {
+		Exists        int
+		Suspended     int
+		LiveTable     int
+		InactiveTable int
+		OpenCount     int32
+		EventNr       uint32
+		Major         uint32
+		Minor         uint32
+		ReadOnly      int
+		TargetCount   int32
+	}
+	TaskType    int
+	AddNodeType int
+)
+
+func (t *Task) destroy() {
+	if t != nil {
+		DmTaskDestroy(t.unmanaged)
+		runtime.SetFinalizer(t, nil)
+	}
+}
+
+func TaskCreate(tasktype TaskType) *Task {
+	Ctask := DmTaskCreate(int(tasktype))
+	if Ctask == nil {
+		return nil
+	}
+	task := &Task{unmanaged: Ctask}
+	runtime.SetFinalizer(task, (*Task).destroy)
+	return task
+}
+
+func (t *Task) Run() error {
+	if res := DmTaskRun(t.unmanaged); res != 1 {
+		return ErrTaskRun
+	}
+	return nil
+}
+
+func (t *Task) SetName(name string) error {
+	if res := DmTaskSetName(t.unmanaged, name); res != 1 {
+		return ErrTaskSetName
+	}
+	return nil
+}
+
+func (t *Task) SetMessage(message string) error {
+	if res := DmTaskSetMessage(t.unmanaged, message); res != 1 {
+		return ErrTaskSetMessage
+	}
+	return nil
+}
+
+func (t *Task) SetSector(sector uint64) error {
+	if res := DmTaskSetSector(t.unmanaged, sector); res != 1 {
+		return ErrTaskSetSector
+	}
+	return nil
+}
+
+func (t *Task) SetCookie(cookie *uint, flags uint16) error {
+	if cookie == nil {
+		return ErrNilCookie
+	}
+	if res := DmTaskSetCookie(t.unmanaged, cookie, flags); res != 1 {
+		return ErrTaskSetCookie
+	}
+	return nil
+}
+
+func (t *Task) SetAddNode(addNode AddNodeType) error {
+	if addNode != AddNodeOnResume && addNode != AddNodeOnCreate {
+		return ErrInvalidAddNode
+	}
+	if res := DmTaskSetAddNode(t.unmanaged, addNode); res != 1 {
+		return ErrTaskSetAddNode
+	}
+	return nil
+}
+
+func (t *Task) SetRo() error {
+	if res := DmTaskSetRo(t.unmanaged); res != 1 {
+		return ErrTaskSetRo
+	}
+	return nil
+}
+
+func (t *Task) AddTarget(start, size uint64, ttype, params string) error {
+	if res := DmTaskAddTarget(t.unmanaged, start, size,
+		ttype, params); res != 1 {
+		return ErrTaskAddTarget
+	}
+	return nil
+}
+
+func (t *Task) GetInfo() (*Info, error) {
+	info := &Info{}
+	if res := DmTaskGetInfo(t.unmanaged, info); res != 1 {
+		return nil, ErrTaskGetInfo
+	}
+	return info, nil
+}
+
+func (t *Task) GetNextTarget(next uintptr) (nextPtr uintptr, start uint64,
+	length uint64, targetType string, params string) {
+
+	return DmGetNextTarget(t.unmanaged, next, &start, &length,
+			&targetType, &params),
+		start, length, targetType, params
+}
+
+func getLoopbackBackingFile(file *osFile) (uint64, uint64, error) {
+	loopInfo, err := ioctlLoopGetStatus64(file.Fd())
+	if err != nil {
+		utils.Errorf("Error get loopback backing file: %s\n", err)
+		return 0, 0, ErrGetLoopbackBackingFile
+	}
+	return loopInfo.loDevice, loopInfo.loInode, nil
+}
+
+func LoopbackSetCapacity(file *osFile) error {
+	if err := ioctlLoopSetCapacity(file.Fd(), 0); err != nil {
+		utils.Errorf("Error loopbackSetCapacity: %s", err)
+		return ErrLoopbackSetCapacity
+	}
+	return nil
+}
+
+func FindLoopDeviceFor(file *osFile) *osFile {
+	stat, err := file.Stat()
+	if err != nil {
+		return nil
+	}
+	targetInode := stat.Sys().(*sysStatT).Ino
+	targetDevice := stat.Sys().(*sysStatT).Dev
+
+	for i := 0; true; i++ {
+		path := fmt.Sprintf("/dev/loop%d", i)
+
+		file, err := osOpenFile(path, osORdWr, 0)
+		if err != nil {
+			if osIsNotExist(err) {
+				return nil
+			}
+
+			// Ignore all errors until the first not-exist
+			// we want to continue looking for the file
+			continue
+		}
+
+		dev, inode, err := getLoopbackBackingFile(file)
+		if err == nil && dev == targetDevice && inode == targetInode {
+			return file
+		}
+		file.Close()
+	}
+
+	return nil
+}
+
+func UdevWait(cookie uint) error {
+	if res := DmUdevWait(cookie); res != 1 {
+		utils.Debugf("Failed to wait on udev cookie %d", cookie)
+		return ErrUdevWait
+	}
+	return nil
+}
+
+func LogInitVerbose(level int) {
+	DmLogInitVerbose(level)
+}
+
+var dmLogger DevmapperLogger = nil
+
+func logInit(logger DevmapperLogger) {
+	dmLogger = logger
+	LogWithErrnoInit()
+}
+
+func SetDevDir(dir string) error {
+	if res := DmSetDevDir(dir); res != 1 {
+		utils.Debugf("Error dm_set_dev_dir")
+		return ErrSetDevDir
+	}
+	return nil
+}
+
+func GetLibraryVersion() (string, error) {
+	var version string
+	if res := DmGetLibraryVersion(&version); res != 1 {
+		return "", ErrGetLibraryVersion
+	}
+	return version, nil
+}
+
+// Useful helper for cleanup
+func RemoveDevice(name string) error {
+	task := TaskCreate(DeviceRemove)
+	if task == nil {
+		return ErrCreateRemoveTask
+	}
+	if err := task.SetName(name); err != nil {
+		utils.Debugf("Can't set task name %s", name)
+		return err
+	}
+	if err := task.Run(); err != nil {
+		return ErrRunRemoveDevice
+	}
+	return nil
+}
+
+func GetBlockDeviceSize(file *osFile) (uint64, error) {
+	size, err := ioctlBlkGetSize64(file.Fd())
+	if err != nil {
+		utils.Errorf("Error getblockdevicesize: %s", err)
+		return 0, ErrGetBlockSize
+	}
+	return uint64(size), nil
+}
+
+// This is the programmatic example of "dmsetup create"
+func createPool(poolName string, dataFile, metadataFile *osFile) error {
+	task, err := createTask(DeviceCreate, poolName)
+	if task == nil {
+		return err
+	}
+
+	size, err := GetBlockDeviceSize(dataFile)
+	if err != nil {
+		return fmt.Errorf("Can't get data size")
+	}
+
+	params := metadataFile.Name() + " " + dataFile.Name() + " 128 32768"
+	if err := task.AddTarget(0, size/512, "thin-pool", params); err != nil {
+		return fmt.Errorf("Can't add target")
+	}
+
+	var cookie uint = 0
+	if err := task.SetCookie(&cookie, 0); err != nil {
+		return fmt.Errorf("Can't set cookie")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running DeviceCreate (createPool)")
+	}
+
+	UdevWait(cookie)
+
+	return nil
+}
+
+func reloadPool(poolName string, dataFile, metadataFile *osFile) error {
+	task, err := createTask(DeviceReload, poolName)
+	if task == nil {
+		return err
+	}
+
+	size, err := GetBlockDeviceSize(dataFile)
+	if err != nil {
+		return fmt.Errorf("Can't get data size")
+	}
+
+	params := metadataFile.Name() + " " + dataFile.Name() + " 128 32768"
+	if err := task.AddTarget(0, size/512, "thin-pool", params); err != nil {
+		return fmt.Errorf("Can't add target")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running DeviceCreate")
+	}
+
+	return nil
+}
+
+func createTask(t TaskType, name string) (*Task, error) {
+	task := TaskCreate(t)
+	if task == nil {
+		return nil, fmt.Errorf("Can't create task of type %d", int(t))
+	}
+	if err := task.SetName(name); err != nil {
+		return nil, fmt.Errorf("Can't set task name %s", name)
+	}
+	return task, nil
+}
+
+func getInfo(name string) (*Info, error) {
+	task, err := createTask(DeviceInfo, name)
+	if task == nil {
+		return nil, err
+	}
+	if err := task.Run(); err != nil {
+		return nil, err
+	}
+	return task.GetInfo()
+}
+
+func getStatus(name string) (uint64, uint64, string, string, error) {
+	task, err := createTask(DeviceStatus, name)
+	if task == nil {
+		utils.Debugf("getStatus: Error createTask: %s", err)
+		return 0, 0, "", "", err
+	}
+	if err := task.Run(); err != nil {
+		utils.Debugf("getStatus: Error Run: %s", err)
+		return 0, 0, "", "", err
+	}
+
+	devinfo, err := task.GetInfo()
+	if err != nil {
+		utils.Debugf("getStatus: Error GetInfo: %s", err)
+		return 0, 0, "", "", err
+	}
+	if devinfo.Exists == 0 {
+		utils.Debugf("getStatus: Non existing device %s", name)
+		return 0, 0, "", "", fmt.Errorf("Non existing device %s", name)
+	}
+
+	_, start, length, targetType, params := task.GetNextTarget(0)
+	return start, length, targetType, params, nil
+}
+
+func setTransactionId(poolName string, oldId uint64, newId uint64) error {
+	task, err := createTask(DeviceTargetMsg, poolName)
+	if task == nil {
+		return err
+	}
+
+	if err := task.SetSector(0); err != nil {
+		return fmt.Errorf("Can't set sector")
+	}
+
+	if err := task.SetMessage(fmt.Sprintf("set_transaction_id %d %d", oldId, newId)); err != nil {
+		return fmt.Errorf("Can't set message")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running setTransactionId")
+	}
+	return nil
+}
+
+func suspendDevice(name string) error {
+	task, err := createTask(DeviceSuspend, name)
+	if task == nil {
+		return err
+	}
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running DeviceSuspend: %s", err)
+	}
+	return nil
+}
+
+func resumeDevice(name string) error {
+	task, err := createTask(DeviceResume, name)
+	if task == nil {
+		return err
+	}
+
+	var cookie uint = 0
+	if err := task.SetCookie(&cookie, 0); err != nil {
+		return fmt.Errorf("Can't set cookie")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running DeviceResume")
+	}
+
+	UdevWait(cookie)
+
+	return nil
+}
+
+func createDevice(poolName string, deviceId int) error {
+	utils.Debugf("[devmapper] createDevice(poolName=%v, deviceId=%v)", poolName, deviceId)
+	task, err := createTask(DeviceTargetMsg, poolName)
+	if task == nil {
+		return err
+	}
+
+	if err := task.SetSector(0); err != nil {
+		return fmt.Errorf("Can't set sector")
+	}
+
+	if err := task.SetMessage(fmt.Sprintf("create_thin %d", deviceId)); err != nil {
+		return fmt.Errorf("Can't set message")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running createDevice")
+	}
+	return nil
+}
+
+func deleteDevice(poolName string, deviceId int) error {
+	task, err := createTask(DeviceTargetMsg, poolName)
+	if task == nil {
+		return err
+	}
+
+	if err := task.SetSector(0); err != nil {
+		return fmt.Errorf("Can't set sector")
+	}
+
+	if err := task.SetMessage(fmt.Sprintf("delete %d", deviceId)); err != nil {
+		return fmt.Errorf("Can't set message")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running deleteDevice")
+	}
+	return nil
+}
+
+func removeDevice(name string) error {
+	utils.Debugf("[devmapper] removeDevice START")
+	defer utils.Debugf("[devmapper] removeDevice END")
+	task, err := createTask(DeviceRemove, name)
+	if task == nil {
+		return err
+	}
+	if err = task.Run(); err != nil {
+		return fmt.Errorf("Error running removeDevice")
+	}
+	return nil
+}
+
+func activateDevice(poolName string, name string, deviceId int, size uint64) error {
+	task, err := createTask(DeviceCreate, name)
+	if task == nil {
+		return err
+	}
+
+	params := fmt.Sprintf("%s %d", poolName, deviceId)
+	if err := task.AddTarget(0, size/512, "thin", params); err != nil {
+		return fmt.Errorf("Can't add target")
+	}
+	if err := task.SetAddNode(AddNodeOnCreate); err != nil {
+		return fmt.Errorf("Can't add node")
+	}
+
+	var cookie uint = 0
+	if err := task.SetCookie(&cookie, 0); err != nil {
+		return fmt.Errorf("Can't set cookie")
+	}
+
+	if err := task.Run(); err != nil {
+		return fmt.Errorf("Error running DeviceCreate (activateDevice)")
+	}
+
+	UdevWait(cookie)
+
+	return nil
+}
+
+func (devices *DeviceSet) createSnapDevice(poolName string, deviceId int, baseName string, baseDeviceId int) error {
+	devinfo, _ := getInfo(baseName)
+	doSuspend := devinfo != nil && devinfo.Exists != 0
+
+	if doSuspend {
+		if err := suspendDevice(baseName); err != nil {
+			return err
+		}
+	}
+
+	task, err := createTask(DeviceTargetMsg, poolName)
+	if task == nil {
+		if doSuspend {
+			resumeDevice(baseName)
+		}
+		return err
+	}
+
+	if err := task.SetSector(0); err != nil {
+		if doSuspend {
+			resumeDevice(baseName)
+		}
+		return fmt.Errorf("Can't set sector")
+	}
+
+	if err := task.SetMessage(fmt.Sprintf("create_snap %d %d", deviceId, baseDeviceId)); err != nil {
+		if doSuspend {
+			resumeDevice(baseName)
+		}
+		return fmt.Errorf("Can't set message")
+	}
+
+	if err := task.Run(); err != nil {
+		if doSuspend {
+			resumeDevice(baseName)
+		}
+		return fmt.Errorf("Error running DeviceCreate (createSnapDevice)")
+	}
+
+	if doSuspend {
+		if err := resumeDevice(baseName); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}

graphdriver/devmapper/devmapper_doc.go

@@ -0,0 +1,106 @@
+package devmapper
+
+// Definition of struct dm_task and sub structures (from lvm2)
+//
+// struct dm_ioctl {
+// 	/*
+// 	 * The version number is made up of three parts:
+// 	 * major - no backward or forward compatibility,
+// 	 * minor - only backwards compatible,
+// 	 * patch - both backwards and forwards compatible.
+// 	 *
+// 	 * All clients of the ioctl interface should fill in the
+// 	 * version number of the interface that they were
+// 	 * compiled with.
+// 	 *
+// 	 * All recognised ioctl commands (ie. those that don't
+// 	 * return -ENOTTY) fill out this field, even if the
+// 	 * command failed.
+// 	 */
+// 	uint32_t version[3];	/* in/out */
+// 	uint32_t data_size;	/* total size of data passed in
+// 				 * including this struct */
+
+// 	uint32_t data_start;	/* offset to start of data
+// 				 * relative to start of this struct */
+
+// 	uint32_t target_count;	/* in/out */
+// 	int32_t open_count;	/* out */
+// 	uint32_t flags;		/* in/out */
+
+// 	/*
+// 	 * event_nr holds either the event number (input and output) or the
+// 	 * udev cookie value (input only).
+// 	 * The DM_DEV_WAIT ioctl takes an event number as input.
+// 	 * The DM_SUSPEND, DM_DEV_REMOVE and DM_DEV_RENAME ioctls
+// 	 * use the field as a cookie to return in the DM_COOKIE
+// 	 * variable with the uevents they issue.
+// 	 * For output, the ioctls return the event number, not the cookie.
+// 	 */
+// 	uint32_t event_nr;      	/* in/out */
+// 	uint32_t padding;
+
+// 	uint64_t dev;		/* in/out */
+
+// 	char name[DM_NAME_LEN];	/* device name */
+// 	char uuid[DM_UUID_LEN];	/* unique identifier for
+// 				 * the block device */
+// 	char data[7];		/* padding or data */
+// };
+
+// struct target {
+// 	uint64_t start;
+// 	uint64_t length;
+// 	char *type;
+// 	char *params;
+
+// 	struct target *next;
+// };
+
+// typedef enum {
+// 	DM_ADD_NODE_ON_RESUME, /* add /dev/mapper node with dmsetup resume */
+// 	DM_ADD_NODE_ON_CREATE  /* add /dev/mapper node with dmsetup create */
+// } dm_add_node_t;
+
+// struct dm_task {
+// 	int type;
+// 	char *dev_name;
+// 	char *mangled_dev_name;
+
+// 	struct target *head, *tail;
+
+// 	int read_only;
+// 	uint32_t event_nr;
+// 	int major;
+// 	int minor;
+// 	int allow_default_major_fallback;
+// 	uid_t uid;
+// 	gid_t gid;
+// 	mode_t mode;
+// 	uint32_t read_ahead;
+// 	uint32_t read_ahead_flags;
+// 	union {
+// 		struct dm_ioctl *v4;
+// 	} dmi;
+// 	char *newname;
+// 	char *message;
+// 	char *geometry;
+// 	uint64_t sector;
+// 	int no_flush;
+// 	int no_open_count;
+// 	int skip_lockfs;
+// 	int query_inactive_table;
+// 	int suppress_identical_reload;
+// 	dm_add_node_t add_node;
+// 	uint64_t existing_table_size;
+// 	int cookie_set;
+// 	int new_uuid;
+// 	int secure_data;
+// 	int retry_remove;
+// 	int enable_checks;
+// 	int expected_errno;
+
+// 	char *uuid;
+// 	char *mangled_uuid;
+// };
+//

graphdriver/devmapper/devmapper_log.go

@@ -0,0 +1,15 @@
+// +build linux
+
+package devmapper
+
+import "C"
+
+// Due to the way cgo works this has to be in a separate file, as devmapper.go has
+// definitions in the cgo block, which is incompatible with using "//export"
+
+//export DevmapperLogCallback
+func DevmapperLogCallback(level C.int, file *C.char, line C.int, dm_errno_or_class C.int, message *C.char) {
+	if dmLogger != nil {
+		dmLogger.log(int(level), C.GoString(file), int(line), int(dm_errno_or_class), C.GoString(message))
+	}
+}

graphdriver/devmapper/devmapper_test.go

@@ -0,0 +1,287 @@
+// +build linux
+
+package devmapper
+
+import (
+	"testing"
+)
+
+func TestTaskCreate(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	// Test success
+	taskCreate(t, DeviceInfo)
+
+	// Test Failure
+	DmTaskCreate = dmTaskCreateFail
+	defer func() { DmTaskCreate = dmTaskCreateFct }()
+	if task := TaskCreate(-1); task != nil {
+		t.Fatalf("An error should have occured while creating an invalid task.")
+	}
+}
+
+func TestTaskRun(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	// Perform the RUN
+	if err := task.Run(); err != nil {
+		t.Fatal(err)
+	}
+	// Make sure we don't have error with GetInfo
+	if _, err := task.GetInfo(); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	DmTaskRun = dmTaskRunFail
+	defer func() { DmTaskRun = dmTaskRunFct }()
+
+	task = taskCreate(t, DeviceInfo)
+	// Perform the RUN
+	if err := task.Run(); err != ErrTaskRun {
+		t.Fatalf("An error should have occured while running task.")
+	}
+	// Make sure GetInfo also fails
+	if _, err := task.GetInfo(); err != ErrTaskGetInfo {
+		t.Fatalf("GetInfo should fail if task.Run() failed.")
+	}
+}
+
+func TestTaskSetName(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	if err := task.SetName("test"); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	DmTaskSetName = dmTaskSetNameFail
+	defer func() { DmTaskSetName = dmTaskSetNameFct }()
+
+	if err := task.SetName("test"); err != ErrTaskSetName {
+		t.Fatalf("An error should have occured while runnign SetName.")
+	}
+}
+
+func TestTaskSetMessage(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	if err := task.SetMessage("test"); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	DmTaskSetMessage = dmTaskSetMessageFail
+	defer func() { DmTaskSetMessage = dmTaskSetMessageFct }()
+
+	if err := task.SetMessage("test"); err != ErrTaskSetMessage {
+		t.Fatalf("An error should have occured while runnign SetMessage.")
+	}
+}
+
+func TestTaskSetSector(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	if err := task.SetSector(128); err != nil {
+		t.Fatal(err)
+	}
+
+	DmTaskSetSector = dmTaskSetSectorFail
+	defer func() { DmTaskSetSector = dmTaskSetSectorFct }()
+
+	// Test failure
+	if err := task.SetSector(0); err != ErrTaskSetSector {
+		t.Fatalf("An error should have occured while running SetSector.")
+	}
+}
+
+func TestTaskSetCookie(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	var (
+		cookie uint = 0
+		task        = taskCreate(t, DeviceInfo)
+	)
+
+	// Test success
+	if err := task.SetCookie(&cookie, 0); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	if err := task.SetCookie(nil, 0); err != ErrNilCookie {
+		t.Fatalf("An error should have occured while running SetCookie with nil cookie.")
+	}
+
+	DmTaskSetCookie = dmTaskSetCookieFail
+	defer func() { DmTaskSetCookie = dmTaskSetCookieFct }()
+
+	if err := task.SetCookie(&cookie, 0); err != ErrTaskSetCookie {
+		t.Fatalf("An error should have occured while running SetCookie.")
+	}
+}
+
+func TestTaskSetAddNode(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	if err := task.SetAddNode(0); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	if err := task.SetAddNode(-1); err != ErrInvalidAddNode {
+		t.Fatalf("An error should have occured running SetAddNode with wrong node.")
+	}
+
+	DmTaskSetAddNode = dmTaskSetAddNodeFail
+	defer func() { DmTaskSetAddNode = dmTaskSetAddNodeFct }()
+
+	if err := task.SetAddNode(0); err != ErrTaskSetAddNode {
+		t.Fatalf("An error should have occured running SetAddNode.")
+	}
+}
+
+func TestTaskSetRo(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	if err := task.SetRo(); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	DmTaskSetRo = dmTaskSetRoFail
+	defer func() { DmTaskSetRo = dmTaskSetRoFct }()
+
+	if err := task.SetRo(); err != ErrTaskSetRo {
+		t.Fatalf("An error should have occured running SetRo.")
+	}
+}
+
+func TestTaskAddTarget(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	task := taskCreate(t, DeviceInfo)
+
+	// Test success
+	if err := task.AddTarget(0, 128, "thinp", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	// Test failure
+	DmTaskAddTarget = dmTaskAddTargetFail
+	defer func() { DmTaskAddTarget = dmTaskAddTargetFct }()
+
+	if err := task.AddTarget(0, 128, "thinp", ""); err != ErrTaskAddTarget {
+		t.Fatalf("An error should have occured running AddTarget.")
+	}
+}
+
+// func TestTaskGetInfo(t *testing.T) {
+// 	task := taskCreate(t, DeviceInfo)
+
+// 	// Test success
+// 	if _, err := task.GetInfo(); err != nil {
+// 		t.Fatal(err)
+// 	}
+
+// 	// Test failure
+// 	DmTaskGetInfo = dmTaskGetInfoFail
+// 	defer func() { DmTaskGetInfo = dmTaskGetInfoFct }()
+
+// 	if _, err := task.GetInfo(); err != ErrTaskGetInfo {
+// 		t.Fatalf("An error should have occured running GetInfo.")
+// 	}
+// }
+
+// func TestTaskGetNextTarget(t *testing.T) {
+// 	task := taskCreate(t, DeviceInfo)
+
+// 	if next, _, _, _, _ := task.GetNextTarget(0); next == 0 {
+// 		t.Fatalf("The next target should not be 0.")
+// 	}
+// }
+
+/// Utils
+func taskCreate(t *testing.T, taskType TaskType) *Task {
+	task := TaskCreate(taskType)
+	if task == nil {
+		t.Fatalf("Error creating task")
+	}
+	return task
+}
+
+/// Failure function replacement
+func dmTaskCreateFail(t int) *CDmTask {
+	return nil
+}
+
+func dmTaskRunFail(task *CDmTask) int {
+	return -1
+}
+
+func dmTaskSetNameFail(task *CDmTask, name string) int {
+	return -1
+}
+
+func dmTaskSetMessageFail(task *CDmTask, message string) int {
+	return -1
+}
+
+func dmTaskSetSectorFail(task *CDmTask, sector uint64) int {
+	return -1
+}
+
+func dmTaskSetCookieFail(task *CDmTask, cookie *uint, flags uint16) int {
+	return -1
+}
+
+func dmTaskSetAddNodeFail(task *CDmTask, addNode AddNodeType) int {
+	return -1
+}
+
+func dmTaskSetRoFail(task *CDmTask) int {
+	return -1
+}
+
+func dmTaskAddTargetFail(task *CDmTask,
+	start, size uint64, ttype, params string) int {
+	return -1
+}
+
+func dmTaskGetInfoFail(task *CDmTask, info *Info) int {
+	return -1
+}
+
+func dmGetNextTargetFail(task *CDmTask, next uintptr, start, length *uint64,
+	target, params *string) uintptr {
+	return 0
+}
+
+func dmAttachLoopDeviceFail(filename string, fd *int) string {
+	return ""
+}
+
+func sysGetBlockSizeFail(fd uintptr, size *uint64) sysErrno {
+	return 1
+}
+
+func dmUdevWaitFail(cookie uint) int {
+	return -1
+}
+
+func dmSetDevDirFail(dir string) int {
+	return -1
+}
+
+func dmGetLibraryVersionFail(version *string) int {
+	return -1
+}

graphdriver/devmapper/devmapper_wrapper.go

@@ -0,0 +1,221 @@
+// +build linux
+
+package devmapper
+
+/*
+#cgo LDFLAGS: -L. -ldevmapper
+#include <libdevmapper.h>
+#include <linux/loop.h> // FIXME: present only for defines, maybe we can remove it?
+#include <linux/fs.h>   // FIXME: present only for BLKGETSIZE64, maybe we can remove it?
+
+// FIXME: Can't we find a way to do the logging in pure Go?
+extern void DevmapperLogCallback(int level, char *file, int line, int dm_errno_or_class, char *str);
+
+static void	log_cb(int level, const char *file, int line, int dm_errno_or_class, const char *f, ...)
+{
+  char buffer[256];
+  va_list ap;
+
+  va_start(ap, f);
+  vsnprintf(buffer, 256, f, ap);
+  va_end(ap);
+
+  DevmapperLogCallback(level, (char *)file, line, dm_errno_or_class, buffer);
+}
+
+static void	log_with_errno_init()
+{
+  dm_log_with_errno_init(log_cb);
+}
+*/
+import "C"
+
+import (
+	"unsafe"
+)
+
+type (
+	CDmTask C.struct_dm_task
+
+	CLoopInfo64 C.struct_loop_info64
+	LoopInfo64  struct {
+		loDevice           uint64 /* ioctl r/o */
+		loInode            uint64 /* ioctl r/o */
+		loRdevice          uint64 /* ioctl r/o */
+		loOffset           uint64
+		loSizelimit        uint64 /* bytes, 0 == max available */
+		loNumber           uint32 /* ioctl r/o */
+		loEncrypt_type     uint32
+		loEncrypt_key_size uint32 /* ioctl w/o */
+		loFlags            uint32 /* ioctl r/o */
+		loFileName         [LoNameSize]uint8
+		loCryptName        [LoNameSize]uint8
+		loEncryptKey       [LoKeySize]uint8 /* ioctl w/o */
+		loInit             [2]uint64
+	}
+)
+
+// FIXME: Make sure the values are defined in C
+// IOCTL consts
+const (
+	BlkGetSize64 = C.BLKGETSIZE64
+
+	LoopSetFd       = C.LOOP_SET_FD
+	LoopCtlGetFree  = C.LOOP_CTL_GET_FREE
+	LoopGetStatus64 = C.LOOP_GET_STATUS64
+	LoopSetStatus64 = C.LOOP_SET_STATUS64
+	LoopClrFd       = C.LOOP_CLR_FD
+	LoopSetCapacity = C.LOOP_SET_CAPACITY
+)
+
+const (
+	LoFlagsAutoClear = C.LO_FLAGS_AUTOCLEAR
+	LoFlagsReadOnly  = C.LO_FLAGS_READ_ONLY
+	LoFlagsPartScan  = C.LO_FLAGS_PARTSCAN
+	LoKeySize        = C.LO_KEY_SIZE
+	LoNameSize       = C.LO_NAME_SIZE
+)
+
+var (
+	DmGetLibraryVersion = dmGetLibraryVersionFct
+	DmGetNextTarget     = dmGetNextTargetFct
+	DmLogInitVerbose    = dmLogInitVerboseFct
+	DmSetDevDir         = dmSetDevDirFct
+	DmTaskAddTarget     = dmTaskAddTargetFct
+	DmTaskCreate        = dmTaskCreateFct
+	DmTaskDestroy       = dmTaskDestroyFct
+	DmTaskGetInfo       = dmTaskGetInfoFct
+	DmTaskRun           = dmTaskRunFct
+	DmTaskSetAddNode    = dmTaskSetAddNodeFct
+	DmTaskSetCookie     = dmTaskSetCookieFct
+	DmTaskSetMessage    = dmTaskSetMessageFct
+	DmTaskSetName       = dmTaskSetNameFct
+	DmTaskSetRo         = dmTaskSetRoFct
+	DmTaskSetSector     = dmTaskSetSectorFct
+	DmUdevWait          = dmUdevWaitFct
+	LogWithErrnoInit    = logWithErrnoInitFct
+)
+
+func free(p *C.char) {
+	C.free(unsafe.Pointer(p))
+}
+
+func dmTaskDestroyFct(task *CDmTask) {
+	C.dm_task_destroy((*C.struct_dm_task)(task))
+}
+
+func dmTaskCreateFct(taskType int) *CDmTask {
+	return (*CDmTask)(C.dm_task_create(C.int(taskType)))
+}
+
+func dmTaskRunFct(task *CDmTask) int {
+	ret, _ := C.dm_task_run((*C.struct_dm_task)(task))
+	return int(ret)
+}
+
+func dmTaskSetNameFct(task *CDmTask, name string) int {
+	Cname := C.CString(name)
+	defer free(Cname)
+
+	return int(C.dm_task_set_name((*C.struct_dm_task)(task), Cname))
+}
+
+func dmTaskSetMessageFct(task *CDmTask, message string) int {
+	Cmessage := C.CString(message)
+	defer free(Cmessage)
+
+	return int(C.dm_task_set_message((*C.struct_dm_task)(task), Cmessage))
+}
+
+func dmTaskSetSectorFct(task *CDmTask, sector uint64) int {
+	return int(C.dm_task_set_sector((*C.struct_dm_task)(task), C.uint64_t(sector)))
+}
+
+func dmTaskSetCookieFct(task *CDmTask, cookie *uint, flags uint16) int {
+	cCookie := C.uint32_t(*cookie)
+	defer func() {
+		*cookie = uint(cCookie)
+	}()
+	return int(C.dm_task_set_cookie((*C.struct_dm_task)(task), &cCookie, C.uint16_t(flags)))
+}
+
+func dmTaskSetAddNodeFct(task *CDmTask, addNode AddNodeType) int {
+	return int(C.dm_task_set_add_node((*C.struct_dm_task)(task), C.dm_add_node_t(addNode)))
+}
+
+func dmTaskSetRoFct(task *CDmTask) int {
+	return int(C.dm_task_set_ro((*C.struct_dm_task)(task)))
+}
+
+func dmTaskAddTargetFct(task *CDmTask,
+	start, size uint64, ttype, params string) int {
+
+	Cttype := C.CString(ttype)
+	defer free(Cttype)
+
+	Cparams := C.CString(params)
+	defer free(Cparams)
+
+	return int(C.dm_task_add_target((*C.struct_dm_task)(task), C.uint64_t(start), C.uint64_t(size), Cttype, Cparams))
+}
+
+func dmTaskGetInfoFct(task *CDmTask, info *Info) int {
+	Cinfo := C.struct_dm_info{}
+	defer func() {
+		info.Exists = int(Cinfo.exists)
+		info.Suspended = int(Cinfo.suspended)
+		info.LiveTable = int(Cinfo.live_table)
+		info.InactiveTable = int(Cinfo.inactive_table)
+		info.OpenCount = int32(Cinfo.open_count)
+		info.EventNr = uint32(Cinfo.event_nr)
+		info.Major = uint32(Cinfo.major)
+		info.Minor = uint32(Cinfo.minor)
+		info.ReadOnly = int(Cinfo.read_only)
+		info.TargetCount = int32(Cinfo.target_count)
+	}()
+	return int(C.dm_task_get_info((*C.struct_dm_task)(task), &Cinfo))
+}
+
+func dmGetNextTargetFct(task *CDmTask, next uintptr, start, length *uint64, target, params *string) uintptr {
+	var (
+		Cstart, Clength      C.uint64_t
+		CtargetType, Cparams *C.char
+	)
+	defer func() {
+		*start = uint64(Cstart)
+		*length = uint64(Clength)
+		*target = C.GoString(CtargetType)
+		*params = C.GoString(Cparams)
+	}()
+
+	nextp := C.dm_get_next_target((*C.struct_dm_task)(task), unsafe.Pointer(next), &Cstart, &Clength, &CtargetType, &Cparams)
+	return uintptr(nextp)
+}
+
+func dmUdevWaitFct(cookie uint) int {
+	return int(C.dm_udev_wait(C.uint32_t(cookie)))
+}
+
+func dmLogInitVerboseFct(level int) {
+	C.dm_log_init_verbose(C.int(level))
+}
+
+func logWithErrnoInitFct() {
+	C.log_with_errno_init()
+}
+
+func dmSetDevDirFct(dir string) int {
+	Cdir := C.CString(dir)
+	defer free(Cdir)
+
+	return int(C.dm_set_dev_dir(Cdir))
+}
+
+func dmGetLibraryVersionFct(version *string) int {
+	buffer := C.CString(string(make([]byte, 128)))
+	defer free(buffer)
+	defer func() {
+		*version = C.GoString(buffer)
+	}()
+	return int(C.dm_get_library_version(buffer, 128))
+}

graphdriver/devmapper/driver.go

@@ -0,0 +1,128 @@
+// +build linux
+
+package devmapper
+
+import (
+	"fmt"
+	"github.com/dotcloud/docker/graphdriver"
+	"io/ioutil"
+	"path"
+)
+
+func init() {
+	graphdriver.Register("devicemapper", Init)
+}
+
+// Placeholder interfaces, to be replaced
+// at integration.
+
+// End of placeholder interfaces.
+
+type Driver struct {
+	*DeviceSet
+	home string
+}
+
+var Init = func(home string) (graphdriver.Driver, error) {
+	deviceSet, err := NewDeviceSet(home, true)
+	if err != nil {
+		return nil, err
+	}
+	d := &Driver{
+		DeviceSet: deviceSet,
+		home:      home,
+	}
+	return d, nil
+}
+
+func (d *Driver) String() string {
+	return "devicemapper"
+}
+
+func (d *Driver) Status() [][2]string {
+	s := d.DeviceSet.Status()
+
+	status := [][2]string{
+		{"Pool Name", s.PoolName},
+		{"Data file", s.DataLoopback},
+		{"Metadata file", s.MetadataLoopback},
+		{"Data Space Used", fmt.Sprintf("%.1f Mb", float64(s.Data.Used)/(1024*1024))},
+		{"Data Space Total", fmt.Sprintf("%.1f Mb", float64(s.Data.Total)/(1024*1024))},
+		{"Metadata Space Used", fmt.Sprintf("%.1f Mb", float64(s.Metadata.Used)/(1024*1024))},
+		{"Metadata Space Total", fmt.Sprintf("%.1f Mb", float64(s.Metadata.Total)/(1024*1024))},
+	}
+	return status
+}
+
+func (d *Driver) Cleanup() error {
+	return d.DeviceSet.Shutdown()
+}
+
+func (d *Driver) Create(id, parent string) error {
+	if err := d.DeviceSet.AddDevice(id, parent); err != nil {
+		return err
+	}
+
+	mp := path.Join(d.home, "mnt", id)
+	if err := d.mount(id, mp); err != nil {
+		return err
+	}
+
+	if err := osMkdirAll(path.Join(mp, "rootfs"), 0755); err != nil && !osIsExist(err) {
+		return err
+	}
+
+	// Create an "id" file with the container/image id in it to help reconscruct this in case
+	// of later problems
+	if err := ioutil.WriteFile(path.Join(mp, "id"), []byte(id), 0600); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func (d *Driver) Remove(id string) error {
+	mp := path.Join(d.home, "mnt", id)
+	if err := d.unmount(id, mp); err != nil {
+		return err
+	}
+	return d.DeviceSet.RemoveDevice(id)
+}
+
+func (d *Driver) Get(id string) (string, error) {
+	mp := path.Join(d.home, "mnt", id)
+	if err := d.mount(id, mp); err != nil {
+		return "", err
+	}
+	return path.Join(mp, "rootfs"), nil
+}
+
+func (d *Driver) mount(id, mountPoint string) error {
+	// Create the target directories if they don't exist
+	if err := osMkdirAll(mountPoint, 0755); err != nil && !osIsExist(err) {
+		return err
+	}
+	// If mountpoint is already mounted, do nothing
+	if mounted, err := Mounted(mountPoint); err != nil {
+		return fmt.Errorf("Error checking mountpoint: %s", err)
+	} else if mounted {
+		return nil
+	}
+	// Mount the device
+	return d.DeviceSet.MountDevice(id, mountPoint, false)
+}
+
+func (d *Driver) unmount(id, mountPoint string) error {
+	// If mountpoint is not mounted, do nothing
+	if mounted, err := Mounted(mountPoint); err != nil {
+		return fmt.Errorf("Error checking mountpoint: %s", err)
+	} else if !mounted {
+		return nil
+	}
+	// Unmount the device
+	return d.DeviceSet.UnmountDevice(id, mountPoint, true)
+}
+
+func (d *Driver) Exists(id string) bool {
+	return d.Devices[id] != nil
+}

graphdriver/devmapper/driver_test.go

@@ -0,0 +1,878 @@
+// +build linux
+
+package devmapper
+
+import (
+	"fmt"
+	"github.com/dotcloud/docker/graphdriver"
+	"io/ioutil"
+	"path"
+	"runtime"
+	"strings"
+	"syscall"
+	"testing"
+)
+
+func init() {
+	// Reduce the size the the base fs and loopback for the tests
+	DefaultDataLoopbackSize = 300 * 1024 * 1024
+	DefaultMetaDataLoopbackSize = 200 * 1024 * 1024
+	DefaultBaseFsSize = 300 * 1024 * 1024
+}
+
+// denyAllDevmapper mocks all calls to libdevmapper in the unit tests, and denies them by default
+func denyAllDevmapper() {
+	// Hijack all calls to libdevmapper with default panics.
+	// Authorized calls are selectively hijacked in each tests.
+	DmTaskCreate = func(t int) *CDmTask {
+		panic("DmTaskCreate: this method should not be called here")
+	}
+	DmTaskRun = func(task *CDmTask) int {
+		panic("DmTaskRun: this method should not be called here")
+	}
+	DmTaskSetName = func(task *CDmTask, name string) int {
+		panic("DmTaskSetName: this method should not be called here")
+	}
+	DmTaskSetMessage = func(task *CDmTask, message string) int {
+		panic("DmTaskSetMessage: this method should not be called here")
+	}
+	DmTaskSetSector = func(task *CDmTask, sector uint64) int {
+		panic("DmTaskSetSector: this method should not be called here")
+	}
+	DmTaskSetCookie = func(task *CDmTask, cookie *uint, flags uint16) int {
+		panic("DmTaskSetCookie: this method should not be called here")
+	}
+	DmTaskSetAddNode = func(task *CDmTask, addNode AddNodeType) int {
+		panic("DmTaskSetAddNode: this method should not be called here")
+	}
+	DmTaskSetRo = func(task *CDmTask) int {
+		panic("DmTaskSetRo: this method should not be called here")
+	}
+	DmTaskAddTarget = func(task *CDmTask, start, size uint64, ttype, params string) int {
+		panic("DmTaskAddTarget: this method should not be called here")
+	}
+	DmTaskGetInfo = func(task *CDmTask, info *Info) int {
+		panic("DmTaskGetInfo: this method should not be called here")
+	}
+	DmGetNextTarget = func(task *CDmTask, next uintptr, start, length *uint64, target, params *string) uintptr {
+		panic("DmGetNextTarget: this method should not be called here")
+	}
+	DmUdevWait = func(cookie uint) int {
+		panic("DmUdevWait: this method should not be called here")
+	}
+	DmSetDevDir = func(dir string) int {
+		panic("DmSetDevDir: this method should not be called here")
+	}
+	DmGetLibraryVersion = func(version *string) int {
+		panic("DmGetLibraryVersion: this method should not be called here")
+	}
+	DmLogInitVerbose = func(level int) {
+		panic("DmLogInitVerbose: this method should not be called here")
+	}
+	DmTaskDestroy = func(task *CDmTask) {
+		panic("DmTaskDestroy: this method should not be called here")
+	}
+	LogWithErrnoInit = func() {
+		panic("LogWithErrnoInit: this method should not be called here")
+	}
+}
+
+func denyAllSyscall() {
+	sysMount = func(source, target, fstype string, flags uintptr, data string) (err error) {
+		panic("sysMount: this method should not be called here")
+	}
+	sysUnmount = func(target string, flags int) (err error) {
+		panic("sysUnmount: this method should not be called here")
+	}
+	sysCloseOnExec = func(fd int) {
+		panic("sysCloseOnExec: this method should not be called here")
+	}
+	sysSyscall = func(trap, a1, a2, a3 uintptr) (r1, r2 uintptr, err syscall.Errno) {
+		panic("sysSyscall: this method should not be called here")
+	}
+	// Not a syscall, but forbidding it here anyway
+	Mounted = func(mnt string) (bool, error) {
+		panic("devmapper.Mounted: this method should not be called here")
+	}
+	// osOpenFile = os.OpenFile
+	// osNewFile = os.NewFile
+	// osCreate = os.Create
+	// osStat = os.Stat
+	// osIsNotExist = os.IsNotExist
+	// osIsExist = os.IsExist
+	// osMkdirAll = os.MkdirAll
+	// osRemoveAll = os.RemoveAll
+	// osRename = os.Rename
+	// osReadlink = os.Readlink
+
+	// execRun = func(name string, args ...string) error {
+	// 	return exec.Command(name, args...).Run()
+	// }
+}
+
+func mkTestDirectory(t *testing.T) string {
+	dir, err := ioutil.TempDir("", "docker-test-devmapper-")
+	if err != nil {
+		t.Fatal(err)
+	}
+	return dir
+}
+
+func newDriver(t *testing.T) *Driver {
+	home := mkTestDirectory(t)
+	d, err := Init(home)
+	if err != nil {
+		t.Fatal(err)
+	}
+	return d.(*Driver)
+}
+
+func cleanup(d *Driver) {
+	d.Cleanup()
+	osRemoveAll(d.home)
+}
+
+type Set map[string]bool
+
+func (r Set) Assert(t *testing.T, names ...string) {
+	for _, key := range names {
+		if _, exists := r[key]; !exists {
+			t.Fatalf("Key not set: %s", key)
+		}
+		delete(r, key)
+	}
+	if len(r) != 0 {
+		t.Fatalf("Unexpected keys: %v", r)
+	}
+}
+
+func TestInit(t *testing.T) {
+	var (
+		calls        = make(Set)
+		taskMessages = make(Set)
+		taskTypes    = make(Set)
+		home         = mkTestDirectory(t)
+	)
+	defer osRemoveAll(home)
+
+	func() {
+		denyAllDevmapper()
+		DmSetDevDir = func(dir string) int {
+			calls["DmSetDevDir"] = true
+			expectedDir := "/dev"
+			if dir != expectedDir {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmSetDevDir(%v)\nReceived: DmSetDevDir(%v)\n", expectedDir, dir)
+			}
+			return 0
+		}
+		LogWithErrnoInit = func() {
+			calls["DmLogWithErrnoInit"] = true
+		}
+		var task1 CDmTask
+		DmTaskCreate = func(taskType int) *CDmTask {
+			calls["DmTaskCreate"] = true
+			taskTypes[fmt.Sprintf("%d", taskType)] = true
+			return &task1
+		}
+		DmTaskSetName = func(task *CDmTask, name string) int {
+			calls["DmTaskSetName"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskSetName(%v)\nReceived: DmTaskSetName(%v)\n", expectedTask, task)
+			}
+			// FIXME: use Set.AssertRegexp()
+			if !strings.HasPrefix(name, "docker-") && !strings.HasPrefix(name, "/dev/mapper/docker-") ||
+				!strings.HasSuffix(name, "-pool") && !strings.HasSuffix(name, "-base") {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskSetName(%v)\nReceived: DmTaskSetName(%v)\n", "docker-...-pool", name)
+			}
+			return 1
+		}
+		DmTaskRun = func(task *CDmTask) int {
+			calls["DmTaskRun"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskRun(%v)\nReceived: DmTaskRun(%v)\n", expectedTask, task)
+			}
+			return 1
+		}
+		DmTaskGetInfo = func(task *CDmTask, info *Info) int {
+			calls["DmTaskGetInfo"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskGetInfo(%v)\nReceived: DmTaskGetInfo(%v)\n", expectedTask, task)
+			}
+			// This will crash if info is not dereferenceable
+			info.Exists = 0
+			return 1
+		}
+		DmTaskSetSector = func(task *CDmTask, sector uint64) int {
+			calls["DmTaskSetSector"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskSetSector(%v)\nReceived: DmTaskSetSector(%v)\n", expectedTask, task)
+			}
+			if expectedSector := uint64(0); sector != expectedSector {
+				t.Fatalf("Wrong libdevmapper call to DmTaskSetSector\nExpected: %v\nReceived: %v\n", expectedSector, sector)
+			}
+			return 1
+		}
+		DmTaskSetMessage = func(task *CDmTask, message string) int {
+			calls["DmTaskSetMessage"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskSetSector(%v)\nReceived: DmTaskSetSector(%v)\n", expectedTask, task)
+			}
+			taskMessages[message] = true
+			return 1
+		}
+		DmTaskDestroy = func(task *CDmTask) {
+			calls["DmTaskDestroy"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskDestroy(%v)\nReceived: DmTaskDestroy(%v)\n", expectedTask, task)
+			}
+		}
+		DmTaskAddTarget = func(task *CDmTask, start, size uint64, ttype, params string) int {
+			calls["DmTaskSetTarget"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskDestroy(%v)\nReceived: DmTaskDestroy(%v)\n", expectedTask, task)
+			}
+			if start != 0 {
+				t.Fatalf("Wrong start: %d != %d", start, 0)
+			}
+			if ttype != "thin" && ttype != "thin-pool" {
+				t.Fatalf("Wrong ttype: %s", ttype)
+			}
+			// Quick smoke test
+			if params == "" {
+				t.Fatalf("Params should not be empty")
+			}
+			return 1
+		}
+		fakeCookie := uint(4321)
+		DmTaskSetCookie = func(task *CDmTask, cookie *uint, flags uint16) int {
+			calls["DmTaskSetCookie"] = true
+			expectedTask := &task1
+			if task != expectedTask {
+				t.Fatalf("Wrong libdevmapper call\nExpected: DmTaskDestroy(%v)\nReceived: DmTaskDestroy(%v)\n", expectedTask, task)
+			}
+			if flags != 0 {
+				t.Fatalf("Cookie flags should be 0 (not %x)", flags)
+			}
+			*cookie = fakeCookie
+			return 1
+		}
+		DmUdevWait = func(cookie uint) int {
+			calls["DmUdevWait"] = true
+			if cookie != fakeCookie {
+				t.Fatalf("Wrong cookie: %d != %d", cookie, fakeCookie)
+			}
+			return 1
+		}
+		DmTaskSetAddNode = func(task *CDmTask, addNode AddNodeType) int {
+			if addNode != AddNodeOnCreate {
+				t.Fatalf("Wrong AddNoteType: %v (expected %v)", addNode, AddNodeOnCreate)
+			}
+			calls["DmTaskSetAddNode"] = true
+			return 1
+		}
+		execRun = func(name string, args ...string) error {
+			calls["execRun"] = true
+			if name != "mkfs.ext4" {
+				t.Fatalf("Expected %s to be executed, not %s", "mkfs.ext4", name)
+			}
+			return nil
+		}
+		driver, err := Init(home)
+		if err != nil {
+			t.Fatal(err)
+		}
+		defer func() {
+			if err := driver.Cleanup(); err != nil {
+				t.Fatal(err)
+			}
+		}()
+	}()
+	// Put all tests in a funciton to make sure the garbage collection will
+	// occur.
+
+	// Call GC to cleanup runtime.Finalizers
+	runtime.GC()
+
+	calls.Assert(t,
+		"DmSetDevDir",
+		"DmLogWithErrnoInit",
+		"DmTaskSetName",
+		"DmTaskRun",
+		"DmTaskGetInfo",
+		"DmTaskDestroy",
+		"execRun",
+		"DmTaskCreate",
+		"DmTaskSetTarget",
+		"DmTaskSetCookie",
+		"DmUdevWait",
+		"DmTaskSetSector",
+		"DmTaskSetMessage",
+		"DmTaskSetAddNode",
+	)
+	taskTypes.Assert(t, "0", "6", "17")
+	taskMessages.Assert(t, "create_thin 0", "set_transaction_id 0 1")
+}
+
+func fakeInit() func(home string) (graphdriver.Driver, error) {
+	oldInit := Init
+	Init = func(home string) (graphdriver.Driver, error) {
+		return &Driver{
+			home: home,
+		}, nil
+	}
+	return oldInit
+}
+
+func restoreInit(init func(home string) (graphdriver.Driver, error)) {
+	Init = init
+}
+
+func mockAllDevmapper(calls Set) {
+	DmSetDevDir = func(dir string) int {
+		calls["DmSetDevDir"] = true
+		return 0
+	}
+	LogWithErrnoInit = func() {
+		calls["DmLogWithErrnoInit"] = true
+	}
+	DmTaskCreate = func(taskType int) *CDmTask {
+		calls["DmTaskCreate"] = true
+		return &CDmTask{}
+	}
+	DmTaskSetName = func(task *CDmTask, name string) int {
+		calls["DmTaskSetName"] = true
+		return 1
+	}
+	DmTaskRun = func(task *CDmTask) int {
+		calls["DmTaskRun"] = true
+		return 1
+	}
+	DmTaskGetInfo = func(task *CDmTask, info *Info) int {
+		calls["DmTaskGetInfo"] = true
+		return 1
+	}
+	DmTaskSetSector = func(task *CDmTask, sector uint64) int {
+		calls["DmTaskSetSector"] = true
+		return 1
+	}
+	DmTaskSetMessage = func(task *CDmTask, message string) int {
+		calls["DmTaskSetMessage"] = true
+		return 1
+	}
+	DmTaskDestroy = func(task *CDmTask) {
+		calls["DmTaskDestroy"] = true
+	}
+	DmTaskAddTarget = func(task *CDmTask, start, size uint64, ttype, params string) int {
+		calls["DmTaskSetTarget"] = true
+		return 1
+	}
+	DmTaskSetCookie = func(task *CDmTask, cookie *uint, flags uint16) int {
+		calls["DmTaskSetCookie"] = true
+		return 1
+	}
+	DmUdevWait = func(cookie uint) int {
+		calls["DmUdevWait"] = true
+		return 1
+	}
+	DmTaskSetAddNode = func(task *CDmTask, addNode AddNodeType) int {
+		calls["DmTaskSetAddNode"] = true
+		return 1
+	}
+	execRun = func(name string, args ...string) error {
+		calls["execRun"] = true
+		return nil
+	}
+}
+
+func TestDriverName(t *testing.T) {
+	denyAllDevmapper()
+	defer denyAllDevmapper()
+
+	oldInit := fakeInit()
+	defer restoreInit(oldInit)
+
+	d := newDriver(t)
+	if d.String() != "devicemapper" {
+		t.Fatalf("Expected driver name to be devicemapper got %s", d.String())
+	}
+}
+
+func TestDriverCreate(t *testing.T) {
+	denyAllDevmapper()
+	denyAllSyscall()
+	defer denyAllSyscall()
+	defer denyAllDevmapper()
+
+	calls := make(Set)
+	mockAllDevmapper(calls)
+
+	sysMount = func(source, target, fstype string, flags uintptr, data string) (err error) {
+		calls["sysMount"] = true
+		// FIXME: compare the exact source and target strings (inodes + devname)
+		if expectedSource := "/dev/mapper/docker-"; !strings.HasPrefix(source, expectedSource) {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedSource, source)
+		}
+		if expectedTarget := "/tmp/docker-test-devmapper-"; !strings.HasPrefix(target, expectedTarget) {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedTarget, target)
+		}
+		if expectedFstype := "ext4"; fstype != expectedFstype {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedFstype, fstype)
+		}
+		if expectedFlags := uintptr(3236757504); flags != expectedFlags {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedFlags, flags)
+		}
+		return nil
+	}
+
+	Mounted = func(mnt string) (bool, error) {
+		calls["Mounted"] = true
+		if !strings.HasPrefix(mnt, "/tmp/docker-test-devmapper-") || !strings.HasSuffix(mnt, "/mnt/1") {
+			t.Fatalf("Wrong mounted call\nExpected: Mounted(%v)\nReceived: Mounted(%v)\n", "/tmp/docker-test-devmapper-.../mnt/1", mnt)
+		}
+		return false, nil
+	}
+
+	sysSyscall = func(trap, a1, a2, a3 uintptr) (r1, r2 uintptr, err syscall.Errno) {
+		calls["sysSyscall"] = true
+		if trap != sysSysIoctl {
+			t.Fatalf("Unexpected syscall. Expecting SYS_IOCTL, received: %d", trap)
+		}
+		switch a2 {
+		case LoopSetFd:
+			calls["ioctl.loopsetfd"] = true
+		case LoopCtlGetFree:
+			calls["ioctl.loopctlgetfree"] = true
+		case LoopGetStatus64:
+			calls["ioctl.loopgetstatus"] = true
+		case LoopSetStatus64:
+			calls["ioctl.loopsetstatus"] = true
+		case LoopClrFd:
+			calls["ioctl.loopclrfd"] = true
+		case LoopSetCapacity:
+			calls["ioctl.loopsetcapacity"] = true
+		case BlkGetSize64:
+			calls["ioctl.blkgetsize"] = true
+		default:
+			t.Fatalf("Unexpected IOCTL. Received %d", a2)
+		}
+		return 0, 0, 0
+	}
+
+	func() {
+		d := newDriver(t)
+
+		calls.Assert(t,
+			"DmSetDevDir",
+			"DmLogWithErrnoInit",
+			"DmTaskSetName",
+			"DmTaskRun",
+			"DmTaskGetInfo",
+			"execRun",
+			"DmTaskCreate",
+			"DmTaskSetTarget",
+			"DmTaskSetCookie",
+			"DmUdevWait",
+			"DmTaskSetSector",
+			"DmTaskSetMessage",
+			"DmTaskSetAddNode",
+			"sysSyscall",
+			"ioctl.blkgetsize",
+			"ioctl.loopsetfd",
+			"ioctl.loopsetstatus",
+		)
+
+		if err := d.Create("1", ""); err != nil {
+			t.Fatal(err)
+		}
+		calls.Assert(t,
+			"DmTaskCreate",
+			"DmTaskGetInfo",
+			"sysMount",
+			"Mounted",
+			"DmTaskRun",
+			"DmTaskSetTarget",
+			"DmTaskSetSector",
+			"DmTaskSetCookie",
+			"DmUdevWait",
+			"DmTaskSetName",
+			"DmTaskSetMessage",
+			"DmTaskSetAddNode",
+		)
+
+	}()
+
+	runtime.GC()
+
+	calls.Assert(t,
+		"DmTaskDestroy",
+	)
+}
+
+func TestDriverRemove(t *testing.T) {
+	denyAllDevmapper()
+	denyAllSyscall()
+	defer denyAllSyscall()
+	defer denyAllDevmapper()
+
+	calls := make(Set)
+	mockAllDevmapper(calls)
+
+	sysMount = func(source, target, fstype string, flags uintptr, data string) (err error) {
+		calls["sysMount"] = true
+		// FIXME: compare the exact source and target strings (inodes + devname)
+		if expectedSource := "/dev/mapper/docker-"; !strings.HasPrefix(source, expectedSource) {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedSource, source)
+		}
+		if expectedTarget := "/tmp/docker-test-devmapper-"; !strings.HasPrefix(target, expectedTarget) {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedTarget, target)
+		}
+		if expectedFstype := "ext4"; fstype != expectedFstype {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedFstype, fstype)
+		}
+		if expectedFlags := uintptr(3236757504); flags != expectedFlags {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedFlags, flags)
+		}
+		return nil
+	}
+	sysUnmount = func(target string, flags int) (err error) {
+		calls["sysUnmount"] = true
+		// FIXME: compare the exact source and target strings (inodes + devname)
+		if expectedTarget := "/tmp/docker-test-devmapper-"; !strings.HasPrefix(target, expectedTarget) {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedTarget, target)
+		}
+		if expectedFlags := 0; flags != expectedFlags {
+			t.Fatalf("Wrong syscall call\nExpected: Mount(%v)\nReceived: Mount(%v)\n", expectedFlags, flags)
+		}
+		return nil
+	}
+	Mounted = func(mnt string) (bool, error) {
+		calls["Mounted"] = true
+		return false, nil
+	}
+
+	sysSyscall = func(trap, a1, a2, a3 uintptr) (r1, r2 uintptr, err syscall.Errno) {
+		calls["sysSyscall"] = true
+		if trap != sysSysIoctl {
+			t.Fatalf("Unexpected syscall. Expecting SYS_IOCTL, received: %d", trap)
+		}
+		switch a2 {
+		case LoopSetFd:
+			calls["ioctl.loopsetfd"] = true
+		case LoopCtlGetFree:
+			calls["ioctl.loopctlgetfree"] = true
+		case LoopGetStatus64:
+			calls["ioctl.loopgetstatus"] = true
+		case LoopSetStatus64:
+			calls["ioctl.loopsetstatus"] = true
+		case LoopClrFd:
+			calls["ioctl.loopclrfd"] = true
+		case LoopSetCapacity:
+			calls["ioctl.loopsetcapacity"] = true
+		case BlkGetSize64:
+			calls["ioctl.blkgetsize"] = true
+		default:
+			t.Fatalf("Unexpected IOCTL. Received %d", a2)
+		}
+		return 0, 0, 0
+	}
+
+	func() {
+		d := newDriver(t)
+
+		calls.Assert(t,
+			"DmSetDevDir",
+			"DmLogWithErrnoInit",
+			"DmTaskSetName",
+			"DmTaskRun",
+			"DmTaskGetInfo",
+			"execRun",
+			"DmTaskCreate",
+			"DmTaskSetTarget",
+			"DmTaskSetCookie",
+			"DmUdevWait",
+			"DmTaskSetSector",
+			"DmTaskSetMessage",
+			"DmTaskSetAddNode",
+			"sysSyscall",
+			"ioctl.blkgetsize",
+			"ioctl.loopsetfd",
+			"ioctl.loopsetstatus",
+		)
+
+		if err := d.Create("1", ""); err != nil {
+			t.Fatal(err)
+		}
+
+		calls.Assert(t,
+			"DmTaskCreate",
+			"DmTaskGetInfo",
+			"sysMount",
+			"Mounted",
+			"DmTaskRun",
+			"DmTaskSetTarget",
+			"DmTaskSetSector",
+			"DmTaskSetCookie",
+			"DmUdevWait",
+			"DmTaskSetName",
+			"DmTaskSetMessage",
+			"DmTaskSetAddNode",
+		)
+
+		Mounted = func(mnt string) (bool, error) {
+			calls["Mounted"] = true
+			return true, nil
+		}
+
+		if err := d.Remove("1"); err != nil {
+			t.Fatal(err)
+		}
+
+		calls.Assert(t,
+			"DmTaskRun",
+			"DmTaskSetSector",
+			"DmTaskSetName",
+			"DmTaskSetMessage",
+			"DmTaskCreate",
+			"DmTaskGetInfo",
+			"Mounted",
+			"sysUnmount",
+		)
+	}()
+	runtime.GC()
+
+	calls.Assert(t,
+		"DmTaskDestroy",
+	)
+}
+
+func TestCleanup(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	t.Skip("Unimplemented")
+	d := newDriver(t)
+	defer osRemoveAll(d.home)
+
+	mountPoints := make([]string, 2)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	// Mount the id
+	p, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+	mountPoints[0] = p
+
+	if err := d.Create("2", "1"); err != nil {
+		t.Fatal(err)
+	}
+
+	p, err = d.Get("2")
+	if err != nil {
+		t.Fatal(err)
+	}
+	mountPoints[1] = p
+
+	// Ensure that all the mount points are currently mounted
+	for _, p := range mountPoints {
+		if mounted, err := Mounted(p); err != nil {
+			t.Fatal(err)
+		} else if !mounted {
+			t.Fatalf("Expected %s to be mounted", p)
+		}
+	}
+
+	// Ensure that devices are active
+	for _, p := range []string{"1", "2"} {
+		if !d.HasActivatedDevice(p) {
+			t.Fatalf("Expected %s to have an active device", p)
+		}
+	}
+
+	if err := d.Cleanup(); err != nil {
+		t.Fatal(err)
+	}
+
+	// Ensure that all the mount points are no longer mounted
+	for _, p := range mountPoints {
+		if mounted, err := Mounted(p); err != nil {
+			t.Fatal(err)
+		} else if mounted {
+			t.Fatalf("Expected %s to not be mounted", p)
+		}
+	}
+
+	// Ensure that devices are no longer activated
+	for _, p := range []string{"1", "2"} {
+		if d.HasActivatedDevice(p) {
+			t.Fatalf("Expected %s not be an active device", p)
+		}
+	}
+}
+
+func TestNotMounted(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	t.Skip("Not implemented")
+	d := newDriver(t)
+	defer cleanup(d)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	mounted, err := Mounted(path.Join(d.home, "mnt", "1"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if mounted {
+		t.Fatal("Id 1 should not be mounted")
+	}
+}
+
+func TestMounted(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	d := newDriver(t)
+	defer cleanup(d)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := d.Get("1"); err != nil {
+		t.Fatal(err)
+	}
+
+	mounted, err := Mounted(path.Join(d.home, "mnt", "1"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !mounted {
+		t.Fatal("Id 1 should be mounted")
+	}
+}
+
+func TestInitCleanedDriver(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	d := newDriver(t)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := d.Get("1"); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := d.Cleanup(); err != nil {
+		t.Fatal(err)
+	}
+
+	driver, err := Init(d.home)
+	if err != nil {
+		t.Fatal(err)
+	}
+	d = driver.(*Driver)
+	defer cleanup(d)
+
+	if _, err := d.Get("1"); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestMountMountedDriver(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	d := newDriver(t)
+	defer cleanup(d)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	// Perform get on same id to ensure that it will
+	// not be mounted twice
+	if _, err := d.Get("1"); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := d.Get("1"); err != nil {
+		t.Fatal(err)
+	}
+}
+
+func TestGetReturnsValidDevice(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	d := newDriver(t)
+	defer cleanup(d)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	if !d.HasDevice("1") {
+		t.Fatalf("Expected id 1 to be in device set")
+	}
+
+	if _, err := d.Get("1"); err != nil {
+		t.Fatal(err)
+	}
+
+	if !d.HasActivatedDevice("1") {
+		t.Fatalf("Expected id 1 to be activated")
+	}
+
+	if !d.HasInitializedDevice("1") {
+		t.Fatalf("Expected id 1 to be initialized")
+	}
+}
+
+func TestDriverGetSize(t *testing.T) {
+	t.Skip("FIXME: not a unit test")
+	t.Skipf("Size is currently not implemented")
+
+	d := newDriver(t)
+	defer cleanup(d)
+
+	if err := d.Create("1", ""); err != nil {
+		t.Fatal(err)
+	}
+
+	mountPoint, err := d.Get("1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	size := int64(1024)
+
+	f, err := osCreate(path.Join(mountPoint, "test_file"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := f.Truncate(size); err != nil {
+		t.Fatal(err)
+	}
+	f.Close()
+
+	// diffSize, err := d.DiffSize("1")
+	// if err != nil {
+	// 	t.Fatal(err)
+	// }
+	// if diffSize != size {
+	// 	t.Fatalf("Expected size %d got %d", size, diffSize)
+	// }
+}
+
+func assertMap(t *testing.T, m map[string]bool, keys ...string) {
+	for _, key := range keys {
+		if _, exists := m[key]; !exists {
+			t.Fatalf("Key not set: %s", key)
+		}
+		delete(m, key)
+	}
+	if len(m) != 0 {
+		t.Fatalf("Unexpected keys: %v", m)
+	}
+}

graphdriver/devmapper/ioctl.go

@@ -0,0 +1,60 @@
+// +build linux
+
+package devmapper
+
+import (
+	"unsafe"
+)
+
+func ioctlLoopCtlGetFree(fd uintptr) (int, error) {
+	index, _, err := sysSyscall(sysSysIoctl, fd, LoopCtlGetFree, 0)
+	if err != 0 {
+		return 0, err
+	}
+	return int(index), nil
+}
+
+func ioctlLoopSetFd(loopFd, sparseFd uintptr) error {
+	if _, _, err := sysSyscall(sysSysIoctl, loopFd, LoopSetFd, sparseFd); err != 0 {
+		return err
+	}
+	return nil
+}
+
+func ioctlLoopSetStatus64(loopFd uintptr, loopInfo *LoopInfo64) error {
+	if _, _, err := sysSyscall(sysSysIoctl, loopFd, LoopSetStatus64, uintptr(unsafe.Pointer(loopInfo))); err != 0 {
+		return err
+	}
+	return nil
+}
+
+func ioctlLoopClrFd(loopFd uintptr) error {
+	if _, _, err := sysSyscall(sysSysIoctl, loopFd, LoopClrFd, 0); err != 0 {
+		return err
+	}
+	return nil
+}
+
+func ioctlLoopGetStatus64(loopFd uintptr) (*LoopInfo64, error) {
+	loopInfo := &LoopInfo64{}
+
+	if _, _, err := sysSyscall(sysSysIoctl, loopFd, LoopGetStatus64, uintptr(unsafe.Pointer(loopInfo))); err != 0 {
+		return nil, err
+	}
+	return loopInfo, nil
+}
+
+func ioctlLoopSetCapacity(loopFd uintptr, value int) error {
+	if _, _, err := sysSyscall(sysSysIoctl, loopFd, LoopSetCapacity, uintptr(value)); err != 0 {
+		return err
+	}
+	return nil
+}
+
+func ioctlBlkGetSize64(fd uintptr) (int64, error) {
+	var size int64
+	if _, _, err := sysSyscall(sysSysIoctl, fd, BlkGetSize64, uintptr(unsafe.Pointer(&size))); err != 0 {
+		return 0, err
+	}
+	return size, nil
+}

graphdriver/devmapper/mount.go

@@ -0,0 +1,27 @@
+// +build linux
+
+package devmapper
+
+import (
+	"path/filepath"
+)
+
+// FIXME: this is copy-pasted from the aufs driver.
+// It should be moved into the core.
+
+var Mounted = func(mountpoint string) (bool, error) {
+	mntpoint, err := osStat(mountpoint)
+	if err != nil {
+		if osIsNotExist(err) {
+			return false, nil
+		}
+		return false, err
+	}
+	parent, err := osStat(filepath.Join(mountpoint, ".."))
+	if err != nil {
+		return false, err
+	}
+	mntpointSt := toSysStatT(mntpoint.Sys())
+	parentSt := toSysStatT(parent.Sys())
+	return mntpointSt.Dev != parentSt.Dev, nil
+}

graphdriver/devmapper/sys.go

@@ -0,0 +1,57 @@
+// +build linux
+
+package devmapper
+
+import (
+	"os"
+	"os/exec"
+	"syscall"
+)
+
+type (
+	sysStatT syscall.Stat_t
+	sysErrno syscall.Errno
+
+	osFile struct{ *os.File }
+)
+
+var (
+	sysMount       = syscall.Mount
+	sysUnmount     = syscall.Unmount
+	sysCloseOnExec = syscall.CloseOnExec
+	sysSyscall     = syscall.Syscall
+
+	osOpenFile = func(name string, flag int, perm os.FileMode) (*osFile, error) {
+		f, err := os.OpenFile(name, flag, perm)
+		return &osFile{File: f}, err
+	}
+	osOpen       = func(name string) (*osFile, error) { f, err := os.Open(name); return &osFile{File: f}, err }
+	osNewFile    = os.NewFile
+	osCreate     = os.Create
+	osStat       = os.Stat
+	osIsNotExist = os.IsNotExist
+	osIsExist    = os.IsExist
+	osMkdirAll   = os.MkdirAll
+	osRemoveAll  = os.RemoveAll
+	osRename     = os.Rename
+	osReadlink   = os.Readlink
+
+	execRun = func(name string, args ...string) error { return exec.Command(name, args...).Run() }
+)
+
+const (
+	sysMsMgcVal = syscall.MS_MGC_VAL
+	sysMsRdOnly = syscall.MS_RDONLY
+	sysEInval   = syscall.EINVAL
+	sysSysIoctl = syscall.SYS_IOCTL
+	sysEBusy    = syscall.EBUSY
+
+	osORdOnly    = os.O_RDONLY
+	osORdWr      = os.O_RDWR
+	osOCreate    = os.O_CREATE
+	osModeDevice = os.ModeDevice
+)
+
+func toSysStatT(i interface{}) *sysStatT {
+	return (*sysStatT)(i.(*syscall.Stat_t))
+}

graphdriver/driver.go

@@ -0,0 +1,90 @@
+package graphdriver
+
+import (
+	"fmt"
+	"github.com/dotcloud/docker/archive"
+	"github.com/dotcloud/docker/utils"
+	"os"
+	"path"
+)
+
+type InitFunc func(root string) (Driver, error)
+
+type Driver interface {
+	String() string
+
+	Create(id, parent string) error
+	Remove(id string) error
+
+	Get(id string) (dir string, err error)
+	Exists(id string) bool
+
+	Status() [][2]string
+
+	Cleanup() error
+}
+
+type Differ interface {
+	Diff(id string) (archive.Archive, error)
+	Changes(id string) ([]archive.Change, error)
+	ApplyDiff(id string, diff archive.Archive) error
+	DiffSize(id string) (bytes int64, err error)
+}
+
+var (
+	DefaultDriver string
+	// All registred drivers
+	drivers map[string]InitFunc
+	// Slice of drivers that should be used in an order
+	priority = []string{
+		"aufs",
+		"devicemapper",
+		"vfs",
+	}
+)
+
+func init() {
+	drivers = make(map[string]InitFunc)
+}
+
+func Register(name string, initFunc InitFunc) error {
+	if _, exists := drivers[name]; exists {
+		return fmt.Errorf("Name already registered %s", name)
+	}
+	drivers[name] = initFunc
+
+	return nil
+}
+
+func GetDriver(name, home string) (Driver, error) {
+	if initFunc, exists := drivers[name]; exists {
+		return initFunc(path.Join(home, name))
+	}
+	return nil, fmt.Errorf("No such driver: %s", name)
+}
+
+func New(root string) (driver Driver, err error) {
+	for _, name := range []string{os.Getenv("DOCKER_DRIVER"), DefaultDriver} {
+		if name != "" {
+			return GetDriver(name, root)
+		}
+	}
+
+	// Check for priority drivers first
+	for _, name := range priority {
+		if driver, err = GetDriver(name, root); err != nil {
+			utils.Debugf("Error loading driver %s: %s", name, err)
+			continue
+		}
+		return driver, nil
+	}
+
+	// Check all registered drivers if no priority driver is found
+	for _, initFunc := range drivers {
+		if driver, err = initFunc(root); err != nil {
+			continue
+		}
+		return driver, nil
+	}
+	return nil, err
+}