Merge pull request #19187 from estesp/lets-do-this

User namespaces: graduate from experimental

daemon/config_experimental.go

@@ -2,118 +2,7 @@
 
 package daemon
 
-import (
+import flag "github.com/docker/docker/pkg/mflag"
-	"fmt"
-	"strconv"
-	"strings"
-
-	"github.com/docker/docker/pkg/idtools"
-	flag "github.com/docker/docker/pkg/mflag"
-	"github.com/opencontainers/runc/libcontainer/user"
-)
 
 func (config *Config) attachExperimentalFlags(cmd *flag.FlagSet, usageFn func(string) string) {
-	cmd.StringVar(&config.RemappedRoot, []string{"-userns-remap"}, "", usageFn("User/Group setting for user namespaces"))
-}
-
-const (
-	defaultIDSpecifier string = "default"
-	defaultRemappedID  string = "dockremap"
-)
-
-// Parse the remapped root (user namespace) option, which can be one of:
-//   username            - valid username from /etc/passwd
-//   username:groupname  - valid username; valid groupname from /etc/group
-//   uid                 - 32-bit unsigned int valid Linux UID value
-//   uid:gid             - uid value; 32-bit unsigned int Linux GID value
-//
-//  If no groupname is specified, and a username is specified, an attempt
-//  will be made to lookup a gid for that username as a groupname
-//
-//  If names are used, they are verified to exist in passwd/group
-func parseRemappedRoot(usergrp string) (string, string, error) {
-
-	var (
-		userID, groupID     int
-		username, groupname string
-	)
-
-	idparts := strings.Split(usergrp, ":")
-	if len(idparts) > 2 {
-		return "", "", fmt.Errorf("Invalid user/group specification in --userns-remap: %q", usergrp)
-	}
-
-	if uid, err := strconv.ParseInt(idparts[0], 10, 32); err == nil {
-		// must be a uid; take it as valid
-		userID = int(uid)
-		luser, err := user.LookupUid(userID)
-		if err != nil {
-			return "", "", fmt.Errorf("Uid %d has no entry in /etc/passwd: %v", userID, err)
-		}
-		username = luser.Name
-		if len(idparts) == 1 {
-			// if the uid was numeric and no gid was specified, take the uid as the gid
-			groupID = userID
-			lgrp, err := user.LookupGid(groupID)
-			if err != nil {
-				return "", "", fmt.Errorf("Gid %d has no entry in /etc/group: %v", groupID, err)
-			}
-			groupname = lgrp.Name
-		}
-	} else {
-		lookupName := idparts[0]
-		// special case: if the user specified "default", they want Docker to create or
-		// use (after creation) the "dockremap" user/group for root remapping
-		if lookupName == defaultIDSpecifier {
-			lookupName = defaultRemappedID
-		}
-		luser, err := user.LookupUser(lookupName)
-		if err != nil && idparts[0] != defaultIDSpecifier {
-			// error if the name requested isn't the special "dockremap" ID
-			return "", "", fmt.Errorf("Error during uid lookup for %q: %v", lookupName, err)
-		} else if err != nil {
-			// special case-- if the username == "default", then we have been asked
-			// to create a new entry pair in /etc/{passwd,group} for which the /etc/sub{uid,gid}
-			// ranges will be used for the user and group mappings in user namespaced containers
-			_, _, err := idtools.AddNamespaceRangesUser(defaultRemappedID)
-			if err == nil {
-				return defaultRemappedID, defaultRemappedID, nil
-			}
-			return "", "", fmt.Errorf("Error during %q user creation: %v", defaultRemappedID, err)
-		}
-		userID = luser.Uid
-		username = luser.Name
-		if len(idparts) == 1 {
-			// we only have a string username, and no group specified; look up gid from username as group
-			group, err := user.LookupGroup(lookupName)
-			if err != nil {
-				return "", "", fmt.Errorf("Error during gid lookup for %q: %v", lookupName, err)
-			}
-			groupID = group.Gid
-			groupname = group.Name
-		}
-	}
-
-	if len(idparts) == 2 {
-		// groupname or gid is separately specified and must be resolved
-		// to a unsigned 32-bit gid
-		if gid, err := strconv.ParseInt(idparts[1], 10, 32); err == nil {
-			// must be a gid, take it as valid
-			groupID = int(gid)
-			lgrp, err := user.LookupGid(groupID)
-			if err != nil {
-				return "", "", fmt.Errorf("Gid %d has no entry in /etc/passwd: %v", groupID, err)
-			}
-			groupname = lgrp.Name
-		} else {
-			// not a number; attempt a lookup
-			group, err := user.LookupGroup(idparts[1])
-			if err != nil {
-				return "", "", fmt.Errorf("Error during gid lookup for %q: %v", idparts[1], err)
-			}
-			groupID = group.Gid
-			groupname = idparts[1]
-		}
-	}
-	return username, groupname, nil
 }

daemon/config_unix.go

@@ -79,6 +79,7 @@ func (config *Config) InstallFlags(cmd *flag.FlagSet, usageFn func(string) strin
 	cmd.BoolVar(&config.EnableCors, []string{"#api-enable-cors", "#-api-enable-cors"}, false, usageFn("Enable CORS headers in the remote API, this is deprecated by --api-cors-header"))
 	cmd.StringVar(&config.CorsHeaders, []string{"-api-cors-header"}, "", usageFn("Set CORS headers in the remote API"))
 	cmd.StringVar(&config.CgroupParent, []string{"-cgroup-parent"}, "", usageFn("Set parent cgroup for all containers"))
+	cmd.StringVar(&config.RemappedRoot, []string{"-userns-remap"}, "", usageFn("User/Group setting for user namespaces"))
 
 	config.attachExperimentalFlags(cmd, usageFn)
 }

daemon/daemon_experimental.go

@@ -2,88 +2,8 @@
 
 package daemon
 
-import (
+import "github.com/docker/engine-api/types/container"
-	"fmt"
-	"os"
-	"path/filepath"
-	"runtime"
-
-	"github.com/Sirupsen/logrus"
-	"github.com/docker/docker/pkg/idtools"
-	"github.com/docker/engine-api/types/container"
-)
-
-func setupRemappedRoot(config *Config) ([]idtools.IDMap, []idtools.IDMap, error) {
-	if runtime.GOOS != "linux" && config.RemappedRoot != "" {
-		return nil, nil, fmt.Errorf("User namespaces are only supported on Linux")
-	}
-
-	// if the daemon was started with remapped root option, parse
-	// the config option to the int uid,gid values
-	var (
-		uidMaps, gidMaps []idtools.IDMap
-	)
-	if config.RemappedRoot != "" {
-		username, groupname, err := parseRemappedRoot(config.RemappedRoot)
-		if err != nil {
-			return nil, nil, err
-		}
-		if username == "root" {
-			// Cannot setup user namespaces with a 1-to-1 mapping; "--root=0:0" is a no-op
-			// effectively
-			logrus.Warnf("User namespaces: root cannot be remapped with itself; user namespaces are OFF")
-			return uidMaps, gidMaps, nil
-		}
-		logrus.Infof("User namespaces: ID ranges will be mapped to subuid/subgid ranges of: %s:%s", username, groupname)
-		// update remapped root setting now that we have resolved them to actual names
-		config.RemappedRoot = fmt.Sprintf("%s:%s", username, groupname)
-
-		uidMaps, gidMaps, err = idtools.CreateIDMappings(username, groupname)
-		if err != nil {
-			return nil, nil, fmt.Errorf("Can't create ID mappings: %v", err)
-		}
-	}
-	return uidMaps, gidMaps, nil
-}
-
-func setupDaemonRoot(config *Config, rootDir string, rootUID, rootGID int) error {
-	config.Root = rootDir
-	// the docker root metadata directory needs to have execute permissions for all users (o+x)
-	// so that syscalls executing as non-root, operating on subdirectories of the graph root
-	// (e.g. mounted layers of a container) can traverse this path.
-	// The user namespace support will create subdirectories for the remapped root host uid:gid
-	// pair owned by that same uid:gid pair for proper write access to those needed metadata and
-	// layer content subtrees.
-	if _, err := os.Stat(rootDir); err == nil {
-		// root current exists; verify the access bits are correct by setting them
-		if err = os.Chmod(rootDir, 0701); err != nil {
-			return err
-		}
-	} else if os.IsNotExist(err) {
-		// no root exists yet, create it 0701 with root:root ownership
-		if err := os.MkdirAll(rootDir, 0701); err != nil {
-			return err
-		}
-	}
-
-	// if user namespaces are enabled we will create a subtree underneath the specified root
-	// with any/all specified remapped root uid/gid options on the daemon creating
-	// a new subdirectory with ownership set to the remapped uid/gid (so as to allow
-	// `chdir()` to work for containers namespaced to that uid/gid)
-	if config.RemappedRoot != "" {
-		config.Root = filepath.Join(rootDir, fmt.Sprintf("%d.%d", rootUID, rootGID))
-		logrus.Debugf("Creating user namespaced daemon root: %s", config.Root)
-		// Create the root directory if it doesn't exists
-		if err := idtools.MkdirAllAs(config.Root, 0700, rootUID, rootGID); err != nil {
-			return fmt.Errorf("Cannot create daemon root: %s: %v", config.Root, err)
-		}
-	}
-	return nil
-}
 
 func (daemon *Daemon) verifyExperimentalContainerSettings(hostConfig *container.HostConfig, config *container.Config) ([]string, error) {
-	if hostConfig.Privileged && daemon.configStore.RemappedRoot != "" {
-		return nil, fmt.Errorf("Privileged mode is incompatible with user namespace mappings")
-	}
 	return nil, nil
 }

daemon/daemon_stub.go

@@ -2,26 +2,7 @@
 
 package daemon
 
-import (
+import "github.com/docker/engine-api/types/container"
-	"os"
-
-	"github.com/docker/docker/pkg/idtools"
-	"github.com/docker/docker/pkg/system"
-	"github.com/docker/engine-api/types/container"
-)
-
-func setupRemappedRoot(config *Config) ([]idtools.IDMap, []idtools.IDMap, error) {
-	return nil, nil, nil
-}
-
-func setupDaemonRoot(config *Config, rootDir string, rootUID, rootGID int) error {
-	config.Root = rootDir
-	// Create the root directory if it doesn't exists
-	if err := system.MkdirAll(config.Root, 0700); err != nil && !os.IsExist(err) {
-		return err
-	}
-	return nil
-}
 
 func (daemon *Daemon) verifyExperimentalContainerSettings(hostConfig *container.HostConfig, config *container.Config) ([]string, error) {
 	return nil, nil

daemon/daemon_unix.go

@@ -7,6 +7,7 @@ import (
 	"net"
 	"os"
 	"path/filepath"
+	"runtime"
 	"strconv"
 	"strings"
 	"syscall"
@@ -33,6 +34,7 @@ import (
 	"github.com/docker/libnetwork/types"
 	blkiodev "github.com/opencontainers/runc/libcontainer/configs"
 	"github.com/opencontainers/runc/libcontainer/label"
+	"github.com/opencontainers/runc/libcontainer/user"
 )
 
 const (
@@ -42,6 +44,9 @@ const (
 	platformSupported = true
 	// It's not kernel limit, we want this 4M limit to supply a reasonable functional container
 	linuxMinMemory = 4194304
+	// constants for remapped root settings
+	defaultIDSpecifier string = "default"
+	defaultRemappedID  string = "dockremap"
 )
 
 func getBlkioWeightDevices(config *containertypes.HostConfig) ([]*blkiodev.WeightDevice, error) {
@@ -375,6 +380,24 @@ func verifyPlatformContainerSettings(daemon *Daemon, hostConfig *containertypes.
 		warnings = append(warnings, "IPv4 forwarding is disabled. Networking will not work.")
 		logrus.Warnf("IPv4 forwarding is disabled. Networking will not work")
 	}
+	// check for various conflicting options with user namespaces
+	if daemon.configStore.RemappedRoot != "" {
+		if hostConfig.Privileged {
+			return warnings, fmt.Errorf("Privileged mode is incompatible with user namespaces.")
+		}
+		if hostConfig.NetworkMode.IsHost() || hostConfig.NetworkMode.IsContainer() {
+			return warnings, fmt.Errorf("Cannot share the host or a container's network namespace when user namespaces are enabled.")
+		}
+		if hostConfig.PidMode.IsHost() {
+			return warnings, fmt.Errorf("Cannot share the host PID namespace when user namespaces are enabled.")
+		}
+		if hostConfig.IpcMode.IsContainer() {
+			return warnings, fmt.Errorf("Cannot share a container's IPC namespace when user namespaces are enabled.")
+		}
+		if hostConfig.ReadonlyRootfs {
+			return warnings, fmt.Errorf("Cannot use the --read-only option when user namespaces are enabled.")
+		}
+	}
 	return warnings, nil
 }
 
@@ -674,6 +697,171 @@ func setupInitLayer(initLayer string, rootUID, rootGID int) error {
 	return nil
 }
 
+// Parse the remapped root (user namespace) option, which can be one of:
+//   username            - valid username from /etc/passwd
+//   username:groupname  - valid username; valid groupname from /etc/group
+//   uid                 - 32-bit unsigned int valid Linux UID value
+//   uid:gid             - uid value; 32-bit unsigned int Linux GID value
+//
+//  If no groupname is specified, and a username is specified, an attempt
+//  will be made to lookup a gid for that username as a groupname
+//
+//  If names are used, they are verified to exist in passwd/group
+func parseRemappedRoot(usergrp string) (string, string, error) {
+
+	var (
+		userID, groupID     int
+		username, groupname string
+	)
+
+	idparts := strings.Split(usergrp, ":")
+	if len(idparts) > 2 {
+		return "", "", fmt.Errorf("Invalid user/group specification in --userns-remap: %q", usergrp)
+	}
+
+	if uid, err := strconv.ParseInt(idparts[0], 10, 32); err == nil {
+		// must be a uid; take it as valid
+		userID = int(uid)
+		luser, err := user.LookupUid(userID)
+		if err != nil {
+			return "", "", fmt.Errorf("Uid %d has no entry in /etc/passwd: %v", userID, err)
+		}
+		username = luser.Name
+		if len(idparts) == 1 {
+			// if the uid was numeric and no gid was specified, take the uid as the gid
+			groupID = userID
+			lgrp, err := user.LookupGid(groupID)
+			if err != nil {
+				return "", "", fmt.Errorf("Gid %d has no entry in /etc/group: %v", groupID, err)
+			}
+			groupname = lgrp.Name
+		}
+	} else {
+		lookupName := idparts[0]
+		// special case: if the user specified "default", they want Docker to create or
+		// use (after creation) the "dockremap" user/group for root remapping
+		if lookupName == defaultIDSpecifier {
+			lookupName = defaultRemappedID
+		}
+		luser, err := user.LookupUser(lookupName)
+		if err != nil && idparts[0] != defaultIDSpecifier {
+			// error if the name requested isn't the special "dockremap" ID
+			return "", "", fmt.Errorf("Error during uid lookup for %q: %v", lookupName, err)
+		} else if err != nil {
+			// special case-- if the username == "default", then we have been asked
+			// to create a new entry pair in /etc/{passwd,group} for which the /etc/sub{uid,gid}
+			// ranges will be used for the user and group mappings in user namespaced containers
+			_, _, err := idtools.AddNamespaceRangesUser(defaultRemappedID)
+			if err == nil {
+				return defaultRemappedID, defaultRemappedID, nil
+			}
+			return "", "", fmt.Errorf("Error during %q user creation: %v", defaultRemappedID, err)
+		}
+		userID = luser.Uid
+		username = luser.Name
+		if len(idparts) == 1 {
+			// we only have a string username, and no group specified; look up gid from username as group
+			group, err := user.LookupGroup(lookupName)
+			if err != nil {
+				return "", "", fmt.Errorf("Error during gid lookup for %q: %v", lookupName, err)
+			}
+			groupID = group.Gid
+			groupname = group.Name
+		}
+	}
+
+	if len(idparts) == 2 {
+		// groupname or gid is separately specified and must be resolved
+		// to a unsigned 32-bit gid
+		if gid, err := strconv.ParseInt(idparts[1], 10, 32); err == nil {
+			// must be a gid, take it as valid
+			groupID = int(gid)
+			lgrp, err := user.LookupGid(groupID)
+			if err != nil {
+				return "", "", fmt.Errorf("Gid %d has no entry in /etc/passwd: %v", groupID, err)
+			}
+			groupname = lgrp.Name
+		} else {
+			// not a number; attempt a lookup
+			group, err := user.LookupGroup(idparts[1])
+			if err != nil {
+				return "", "", fmt.Errorf("Error during gid lookup for %q: %v", idparts[1], err)
+			}
+			groupID = group.Gid
+			groupname = idparts[1]
+		}
+	}
+	return username, groupname, nil
+}
+
+func setupRemappedRoot(config *Config) ([]idtools.IDMap, []idtools.IDMap, error) {
+	if runtime.GOOS != "linux" && config.RemappedRoot != "" {
+		return nil, nil, fmt.Errorf("User namespaces are only supported on Linux")
+	}
+
+	// if the daemon was started with remapped root option, parse
+	// the config option to the int uid,gid values
+	var (
+		uidMaps, gidMaps []idtools.IDMap
+	)
+	if config.RemappedRoot != "" {
+		username, groupname, err := parseRemappedRoot(config.RemappedRoot)
+		if err != nil {
+			return nil, nil, err
+		}
+		if username == "root" {
+			// Cannot setup user namespaces with a 1-to-1 mapping; "--root=0:0" is a no-op
+			// effectively
+			logrus.Warnf("User namespaces: root cannot be remapped with itself; user namespaces are OFF")
+			return uidMaps, gidMaps, nil
+		}
+		logrus.Infof("User namespaces: ID ranges will be mapped to subuid/subgid ranges of: %s:%s", username, groupname)
+		// update remapped root setting now that we have resolved them to actual names
+		config.RemappedRoot = fmt.Sprintf("%s:%s", username, groupname)
+
+		uidMaps, gidMaps, err = idtools.CreateIDMappings(username, groupname)
+		if err != nil {
+			return nil, nil, fmt.Errorf("Can't create ID mappings: %v", err)
+		}
+	}
+	return uidMaps, gidMaps, nil
+}
+
+func setupDaemonRoot(config *Config, rootDir string, rootUID, rootGID int) error {
+	config.Root = rootDir
+	// the docker root metadata directory needs to have execute permissions for all users (o+x)
+	// so that syscalls executing as non-root, operating on subdirectories of the graph root
+	// (e.g. mounted layers of a container) can traverse this path.
+	// The user namespace support will create subdirectories for the remapped root host uid:gid
+	// pair owned by that same uid:gid pair for proper write access to those needed metadata and
+	// layer content subtrees.
+	if _, err := os.Stat(rootDir); err == nil {
+		// root current exists; verify the access bits are correct by setting them
+		if err = os.Chmod(rootDir, 0701); err != nil {
+			return err
+		}
+	} else if os.IsNotExist(err) {
+		// no root exists yet, create it 0701 with root:root ownership
+		if err := os.MkdirAll(rootDir, 0701); err != nil {
+			return err
+		}
+	}
+
+	// if user namespaces are enabled we will create a subtree underneath the specified root
+	// with any/all specified remapped root uid/gid options on the daemon creating
+	// a new subdirectory with ownership set to the remapped uid/gid (so as to allow
+	// `chdir()` to work for containers namespaced to that uid/gid)
+	if config.RemappedRoot != "" {
+		config.Root = filepath.Join(rootDir, fmt.Sprintf("%d.%d", rootUID, rootGID))
+		logrus.Debugf("Creating user namespaced daemon root: %s", config.Root)
+		// Create the root directory if it doesn't exists
+		if err := idtools.MkdirAllAs(config.Root, 0700, rootUID, rootGID); err != nil {
+			return fmt.Errorf("Cannot create daemon root: %s: %v", config.Root, err)
+		}
+	}
+	return nil
+}
+
 // registerLinks writes the links to a file.
 func (daemon *Daemon) registerLinks(container *container.Container, hostConfig *containertypes.HostConfig) error {
 	if hostConfig == nil {

daemon/daemon_windows.go

@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"os"
 	"path/filepath"
 	"runtime"
 	"strings"
@@ -18,6 +19,7 @@ import (
 	containertypes "github.com/docker/engine-api/types/container"
 	// register the windows graph driver
 	"github.com/docker/docker/daemon/graphdriver/windows"
+	"github.com/docker/docker/pkg/idtools"
 	"github.com/docker/docker/pkg/system"
 	"github.com/docker/libnetwork"
 	blkiodev "github.com/opencontainers/runc/libcontainer/configs"
@@ -135,6 +137,19 @@ func (daemon *Daemon) cleanupMounts() error {
 	return nil
 }
 
+func setupRemappedRoot(config *Config) ([]idtools.IDMap, []idtools.IDMap, error) {
+	return nil, nil, nil
+}
+
+func setupDaemonRoot(config *Config, rootDir string, rootUID, rootGID int) error {
+	config.Root = rootDir
+	// Create the root directory if it doesn't exists
+	if err := system.MkdirAll(config.Root, 0700); err != nil && !os.IsExist(err) {
+		return err
+	}
+	return nil
+}
+
 // conditionalMountOnStart is a platform specific helper function during the
 // container start to call mount.
 func (daemon *Daemon) conditionalMountOnStart(container *container.Container) error {

docs/reference/commandline/daemon.md

@@ -62,6 +62,7 @@ weight = -1
       --tlscert="~/.docker/cert.pem"         Path to TLS certificate file
       --tlskey="~/.docker/key.pem"           Path to TLS key file
       --tlsverify                            Use TLS and verify the remote
+      --userns-remap="default"               Enable user namespace remapping
       --userland-proxy=true                  Use userland proxy for loopback traffic
 
 Options with [] may be specified multiple times.
@@ -632,6 +633,133 @@ For information about how to create an authorization plugin, see [authorization
 plugin](../../extend/authorization.md) section in the Docker extend section of this documentation.
 
 
+## Daemon user namespace options
+
+The Linux kernel [user namespace support](http://man7.org/linux/man-pages/man7/user_namespaces.7.html) provides additional security by enabling
+a process, and therefore a container, to have a unique range of user and
+group IDs which are outside the traditional user and group range utilized by
+the host system. Potentially the most important security improvement is that,
+by default, container processes running as the `root` user will have expected
+administrative privilege (with some restrictions) inside the container but will
+effectively be mapped to an unprivileged `uid` on the host.
+
+When user namespace support is enabled, Docker creates a single daemon-wide mapping
+for all containers running on the same engine instance. The mappings will
+utilize the existing subordinate user and group ID feature available on all modern
+Linux distributions.
+The [`/etc/subuid`](http://man7.org/linux/man-pages/man5/subuid.5.html) and
+[`/etc/subgid`](http://man7.org/linux/man-pages/man5/subgid.5.html) files will be
+read for the user, and optional group, specified to the `--userns-remap`
+parameter.  If you do not wish to specify your own user and/or group, you can
+provide `default` as the value to this flag, and a user will be created on your behalf
+and provided subordinate uid and gid ranges. This default user will be named
+`dockremap`, and entries will be created for it in `/etc/passwd` and
+`/etc/group` using your distro's standard user and group creation tools.
+
+> **Note**: The single mapping per-daemon restriction is in place for now
+> because Docker shares image layers from its local cache across all
+> containers running on the engine instance.  Since file ownership must be
+> the same for all containers sharing the same layer content, the decision
+> was made to map the file ownership on `docker pull` to the daemon's user and
+> group mappings so that there is no delay for running containers once the
+> content is downloaded. This design preserves the same performance for `docker
+> pull`, `docker push`, and container startup as users expect with
+> user namespaces disabled.
+
+### Starting the daemon with user namespaces enabled
+
+To enable user namespace support, start the daemon with the
+`--userns-remap` flag, which accepts values in the following formats:
+
+ - uid
+ - uid:gid
+ - username
+ - username:groupname
+
+If numeric IDs are provided, translation back to valid user or group names
+will occur so that the subordinate uid and gid information can be read, given
+these resources are name-based, not id-based.  If the numeric ID information
+provided does not exist as entries in `/etc/passwd` or `/etc/group`, daemon
+startup will fail with an error message.
+
+*Example: starting with default Docker user management:*
+
+```
+     $ docker daemon --userns-remap=default
+```    
+When `default` is provided, Docker will create - or find the existing - user and group
+named `dockremap`. If the user is created, and the Linux distribution has
+appropriate support, the `/etc/subuid` and `/etc/subgid` files will be populated
+with a contiguous 65536 length range of subordinate user and group IDs, starting
+at an offset based on prior entries in those files.  For example, Ubuntu will
+create the following range, based on an existing user named `user1` already owning
+the first 65536 range:
+
+```
+     $ cat /etc/subuid
+     user1:100000:65536
+     dockremap:165536:65536
+```
+
+> **Note:** On a fresh Fedora install, we had to `touch` the
+> `/etc/subuid` and `/etc/subgid` files to have ranges assigned when users
+> were created.  Once these files existed, range assignment on user creation
+> worked properly.
+
+If you have a preferred/self-managed user with subordinate ID mappings already
+configured, you can provide that username or uid to the `--userns-remap` flag.
+If you have a group that doesn't match the username, you may provide the `gid`
+or group name as well; otherwise the username will be used as the group name
+when querying the system for the subordinate group ID range.
+
+### Detailed information on `subuid`/`subgid` ranges
+
+Given potential advanced use of the subordinate ID ranges by power users, the 
+following paragraphs define how the Docker daemon currently uses the range entries
+found within the subordinate range files.
+
+The simplest case is that only one contiguous range is defined for the
+provided user or group. In this case, Docker will use that entire contiguous
+range for the mapping of host uids and gids to the container process.  This
+means that the first ID in the range will be the remapped root user, and the
+IDs above that initial ID will map host ID 1 through the end of the range.
+
+From the example `/etc/subid` content shown above, the remapped root
+user would be uid 165536.
+
+If the system administrator has set up multiple ranges for a single user or
+group, the Docker daemon will read all the available ranges and use the
+following algorithm to create the mapping ranges:
+
+1. The range segments found for the particular user will be sorted by *start ID* ascending.
+2. Map segments will be created from each range in increasing value with a length matching the length of each segment. Therefore the range segment with the lowest numeric starting value will be equal to the remapped root, and continue up through host uid/gid equal to the range segment length. As an example, if the lowest segment starts at ID 1000 and has a length of 100, then a map of 1000 -> 0 (the remapped root) up through 1100 -> 100 will be created from this segment. If the next segment starts at ID 10000, then the next map will start with mapping 10000 -> 101 up to the length of this second segment. This will continue until no more segments are found in the subordinate files for this user.
+3. If more than five range segments exist for a single user, only the first five will be utilized, matching the kernel's limitation of only five entries in `/proc/self/uid_map` and `proc/self/gid_map`.
+
+### User namespace known restrictions
+
+The following standard Docker features are currently incompatible when
+running a Docker daemon with user namespaces enabled:
+
+ - sharing PID or NET namespaces with the host (`--pid=host` or `--net=host`)
+ - sharing a network namespace with an existing container (`--net=container:*other*`)
+ - sharing an IPC namespace with an existing container (`--ipc=container:*other*`)
+ - A `--readonly` container filesystem (this is a Linux kernel restriction against remounting with modified flags of a currently mounted filesystem when inside a user namespace)
+ - external (volume or graph) drivers which are unaware/incapable of using daemon user mappings
+ - Using `--privileged` mode flag on `docker run`
+
+In general, user namespaces are an advanced feature and will require
+coordination with other capabilities. For example, if volumes are mounted from
+the host, file ownership will have to be pre-arranged if the user or
+administrator wishes the containers to have expected access to the volume
+contents.
+
+Finally, while the `root` user inside a user namespaced container process has
+many of the expected admin privileges that go along with being the superuser, the
+Linux kernel has restrictions based on internal knowledge that this is a user namespaced
+process. The most notable restriction that we are aware of at this time is the
+inability to use `mknod`. Permission will be denied for device creation even as
+container `root` inside a user namespace.
+
 ## Miscellaneous options
 
 IP masquerading uses address translation to allow containers without a public

experimental/README.md

@@ -72,7 +72,7 @@ to build a Docker binary with the experimental features enabled:
 ## Current experimental features
 
  * [External graphdriver plugins](plugins_graphdriver.md)
- * [User namespaces](userns.md)
+ * The user namespaces feature has graduated from experimental.
 
 ## How to comment on an experimental feature
 

experimental/userns.md

@@ -1,119 +0,0 @@
-# Experimental: User namespace support
-
-Linux kernel [user namespace support](http://man7.org/linux/man-pages/man7/user_namespaces.7.html) provides additional security by enabling
-a process--and therefore a container--to have a unique range of user and
-group IDs which are outside the traditional user and group range utilized by
-the host system. Potentially the most important security improvement is that,
-by default, container processes running as the `root` user will have expected
-administrative privilege (with some restrictions) inside the container but will
-effectively be mapped to an unprivileged `uid` on the host.
-
-In this experimental phase, the Docker daemon creates a single daemon-wide mapping
-for all containers running on the same engine instance. The mappings will
-utilize the existing subordinate user and group ID feature available on all modern
-Linux distributions.
-The [`/etc/subuid`](http://man7.org/linux/man-pages/man5/subuid.5.html) and
-[`/etc/subgid`](http://man7.org/linux/man-pages/man5/subgid.5.html) files will be
-read for the user, and optional group, specified to the `--userns-remap`
-parameter.  If you do not wish to specify your own user and/or group, you can
-provide `default` as the value to this flag, and a user will be created on your behalf
-and provided subordinate uid and gid ranges. This default user will be named
-`dockremap`, and entries will be created for it in `/etc/passwd` and
-`/etc/group` using your distro's standard user and group creation tools.
-
-> **Note**: The single mapping per-daemon restriction exists for this experimental
-> phase because Docker shares image layers from its local cache across all
-> containers running on the engine instance.  Since file ownership must be
-> the same for all containers sharing the same layer content, the decision
-> was made to map the file ownership on `docker pull` to the daemon's user and
-> group mappings so that there is no delay for running containers once the
-> content is downloaded--exactly the same performance characteristics as with
-> user namespaces disabled.
-
-## Starting the daemon with user namespaces enabled
-To enable this experimental user namespace support for a Docker daemon instance,
-start the daemon with the aforementioned `--userns-remap` flag, which accepts
-values in the following formats:
-
- - uid
- - uid:gid
- - username
- - username:groupname
-
-If numeric IDs are provided, translation back to valid user or group names
-will occur so that the subordinate uid and gid information can be read, given
-these resources are name-based, not id-based.  If the numeric ID information
-provided does not exist as entries in `/etc/passwd` or `/etc/group`, daemon
-startup will fail with an error message.
-
-*An example: starting with default Docker user management:*
-
-```
-     $ docker daemon --userns-remap=default
-```    
-In this case, Docker will create--or find the existing--user and group
-named `dockremap`. If the user is created, and the Linux distribution has
-appropriate support, the `/etc/subuid` and `/etc/subgid` files will be populated
-with a contiguous 65536 length range of subordinate user and group IDs, starting
-at an offset based on prior entries in those files.  For example, Ubuntu will
-create the following range, based on an existing user already having the first
-65536 range:
-
-```
-     $ cat /etc/subuid
-     user1:100000:65536
-     dockremap:165536:65536
-```
-
-> **Note:** On a fresh Fedora install, we found that we had to `touch` the
-> `/etc/subuid` and `/etc/subgid` files to have ranges assigned when users
-> were created.  Once these files existed, range assignment on user creation
-> worked properly.
-
-If you have a preferred/self-managed user with subordinate ID mappings already
-configured, you can provide that username or uid to the `--userns-remap` flag.
-If you have a group that doesn't match the username, you may provide the `gid`
-or group name as well; otherwise the username will be used as the group name
-when querying the system for the subordinate group ID range.
-
-## Detailed information on `subuid`/`subgid` ranges
-
-Given there may be advanced use of the subordinate ID ranges by power users, we will
-describe how the Docker daemon uses the range entries within these files under the
-current experimental user namespace support.
-
-The simplest case exists where only one contiguous range is defined for the
-provided user or group. In this case, Docker will use that entire contiguous
-range for the mapping of host uids and gids to the container process.  This
-means that the first ID in the range will be the remapped root user, and the
-IDs above that initial ID will map host ID 1 through the end of the range.
-
-From the example `/etc/subid` content shown above, that means the remapped root
-user would be uid 165536.
-
-If the system administrator has set up multiple ranges for a single user or
-group, the Docker daemon will read all the available ranges and use the
-following algorithm to create the mapping ranges:
-
-1. The ranges will be sorted by *start ID* ascending
-2. Maps will be created from each range with where the host ID will increment starting at 0 for the first range, 0+*range1* length for the second, and so on.  This means that the lowest range start ID will be the remapped root, and all further ranges will map IDs from 1 through the uid or gid that equals the sum of all range lengths.
-3. Ranges segments above five will be ignored as the kernel ignores any ID maps after five (in `/proc/self/{u,g}id_map`)
-
-## User namespace known restrictions
-
-The following standard Docker features are currently incompatible when
-running a Docker daemon with experimental user namespaces enabled:
-
- - sharing namespaces with the host (--pid=host, --net=host, etc.)
- - sharing namespaces with other containers (--net=container:*other*)
- - A `--readonly` container filesystem (a Linux kernel restriction on remount with new flags of a currently mounted filesystem when inside a user namespace)
- - external (volume/graph) drivers which are unaware/incapable of using daemon user mappings
- - Using `--privileged` mode containers
- - volume use without pre-arranging proper file ownership in mounted volumes
-
-Additionally, while the `root` user inside a user namespaced container
-process has many of the privileges of the administrative root user, the
-following operations will fail:
-
- - Use of `mknod` - permission is denied for device creation by the container root
- - others will be listed here when fully tested

hack/make.sh

@@ -99,7 +99,7 @@ if [ ! "$GOPATH" ]; then
 	exit 1
 fi
 
-if [ "$DOCKER_EXPERIMENTAL" ] || [ "$DOCKER_REMAP_ROOT" ]; then
+if [ "$DOCKER_EXPERIMENTAL" ]; then
 	echo >&2 '# WARNING! DOCKER_EXPERIMENTAL is set: building experimental features'
 	echo >&2
 	DOCKER_BUILDTAGS+=" experimental pkcs11"

integration-cli/docker_api_containers_test.go

@@ -652,10 +652,14 @@ func (s *DockerSuite) TestContainerApiCreateWithDomainName(c *check.C) {
 	c.Assert(containerJSON.Config.Domainname, checker.Equals, domainName, check.Commentf("Mismatched Domainname"))
 }
 
-func (s *DockerSuite) TestContainerApiCreateNetworkMode(c *check.C) {
+func (s *DockerSuite) TestContainerApiCreateBridgeNetworkMode(c *check.C) {
 	testRequires(c, DaemonIsLinux)
-	UtilCreateNetworkMode(c, "host")
 	UtilCreateNetworkMode(c, "bridge")
+}
+
+func (s *DockerSuite) TestContainerApiCreateOtherNetworkModes(c *check.C) {
+	testRequires(c, DaemonIsLinux, NotUserNamespace)
+	UtilCreateNetworkMode(c, "host")
 	UtilCreateNetworkMode(c, "container:web1")
 }
 

man/docker-daemon.8.md

@@ -53,6 +53,7 @@ docker-daemon - Enable daemon mode
 [**--tlskey**[=*~/.docker/key.pem*]]
 [**--tlsverify**]
 [**--userland-proxy**[=*true*]]
+[**--userns-remap**[=*default*]]
 
 # DESCRIPTION
 **docker** has two distinct functions. It is used for starting the Docker
@@ -223,6 +224,9 @@ unix://[/path/to/socket] to use.
 **--userland-proxy**=*true*|*false*
     Rely on a userland proxy implementation for inter-container and outside-to-container loopback communications. Default is true.
 
+**--userns-remap**=*default*|*uid:gid*|*user:group*|*user*|*uid*
+    Enable user namespaces for containers on the daemon. Specifying "default" will cause a new user and group to be created to handle UID and GID range remapping for the user namespace mappings used for contained processes. Specifying a user (or uid) and optionally a group (or gid) will cause the daemon to lookup the user and group's subordinate ID ranges for use as the user namespace mappings for contained processes.
+
 # STORAGE DRIVER OPTIONS
 
 Docker uses storage backends (known as "graphdrivers" in the Docker