Merge pull request #1258 from Random-Liu/windows-support-1

Build CRI Plugin on Windows and Add Presubmit

.appveyor.yml

@@ -0,0 +1,28 @@
+version: "{build}"
+
+image: Visual Studio 2019
+
+clone_folder: c:\gopath\src\github.com\containerd\cri
+
+environment:
+  GOPATH: C:\gopath
+  PATH: C:\go\bin;C:\tools\mingw64\bin;$(PATH)
+  CGO_ENABLED: 1
+  matrix:
+    - GO_VERSION: 1.12.9
+
+install:
+  # Install Mingw
+  - choco install -y mingw --version 5.3.0
+  # Install Go
+  - appveyor DownloadFile https://storage.googleapis.com/golang/go%GO_VERSION%.windows-amd64.zip
+  - 7z x go%GO_VERSION%.windows-amd64.zip -oC:\ >nul
+  - go version
+  # Print powershell version
+  - ps: $psversiontable
+
+build_script:
+  - bash.exe -elc "mingw32-make"
+
+test_script:
+  - bash.exe -elc "mingw32-make test"

Makefile

@@ -15,8 +15,12 @@
 GO := go
 GOOS := $(shell $(GO) env GOOS)
 GOARCH := $(shell $(GO) env GOARCH)
-WHALE = "🇩"
+WHALE := "🇩"
-ONI = "👹"
+ONI := "👹"
+ifeq ($(GOOS),windows)
+	WHALE = "+"
+	ONI = "-"
+endif
 EPOCH_TEST_COMMIT := f9e02affccd51702191e5312665a16045ffef8ab
 PROJECT := github.com/containerd/cri
 BINDIR := ${DESTDIR}/usr/local/bin
@@ -26,13 +30,19 @@ BUILD_DIR := _output
 VERSION := $(shell git rev-parse --short HEAD)
 TARBALL_PREFIX := cri-containerd
 TARBALL := $(TARBALL_PREFIX)-$(VERSION).$(GOOS)-$(GOARCH).tar.gz
+ifneq ($(GOOS),windows)
 	BUILD_TAGS := seccomp apparmor
+endif
 # Add `-TEST` suffix to indicate that all binaries built from this repo are for test.
 GO_LDFLAGS := -X $(PROJECT)/vendor/github.com/containerd/containerd/version.Version=$(VERSION)-TEST
 SOURCES := $(shell find cmd/ pkg/ vendor/ -name '*.go')
 PLUGIN_SOURCES := $(shell ls *.go)
 INTEGRATION_SOURCES := $(shell find integration/ -name '*.go')
 
+ifeq ($(GOOS),windows)
+	BIN_EXT := .exe
+endif
+
 all: binaries
 
 help: ## this help
@@ -74,7 +84,7 @@ update-vendor: sync-vendor sort-vendor ## Syncs containerd/vendor.conf -> vendor
 
 $(BUILD_DIR)/containerd: $(SOURCES) $(PLUGIN_SOURCES)
 	@echo "$(WHALE) $@"
-	$(GO) build -o $@ \
+	$(GO) build -o $@$(BIN_EXT) \
 		-tags '$(BUILD_TAGS)' \
 		-ldflags '$(GO_LDFLAGS)' \
 		-gcflags '$(GO_GCFLAGS)' \

cmd/containerd/builtins_unix.go

@@ -0,0 +1,25 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package main
+
+import (
+	_ "github.com/containerd/containerd/metrics/cgroups"
+	_ "github.com/containerd/containerd/runtime/v1/linux"
+	_ "github.com/containerd/containerd/snapshots/overlay"
+)

cmd/containerd/builtins_windows.go

@@ -0,0 +1,25 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package main
+
+import (
+	_ "github.com/containerd/containerd/diff/lcow"
+	_ "github.com/containerd/containerd/diff/windows"
+	_ "github.com/containerd/containerd/snapshots/windows"
+)

cmd/containerd/containerd.go

@@ -24,8 +24,6 @@ import (
 
 	_ "github.com/containerd/containerd/diff/walking/plugin"
 	_ "github.com/containerd/containerd/gc/scheduler"
-	_ "github.com/containerd/containerd/metrics/cgroups"
-	_ "github.com/containerd/containerd/runtime/v1/linux"
 	_ "github.com/containerd/containerd/runtime/v2"
 	_ "github.com/containerd/containerd/services/containers"
 	_ "github.com/containerd/containerd/services/content"
@@ -39,7 +37,6 @@ import (
 	_ "github.com/containerd/containerd/services/snapshots"
 	_ "github.com/containerd/containerd/services/tasks"
 	_ "github.com/containerd/containerd/services/version"
-	_ "github.com/containerd/containerd/snapshots/overlay"
 	_ "github.com/containerd/cri"
 
 	"github.com/containerd/containerd/cmd/containerd/command"

pkg/config/config.go

@@ -21,11 +21,9 @@ import (
 	"time"
 
 	"github.com/BurntSushi/toml"
-	"github.com/containerd/containerd"
 	"github.com/containerd/containerd/log"
 	"github.com/containerd/containerd/plugin"
 	"github.com/pkg/errors"
-	"k8s.io/kubernetes/pkg/kubelet/server/streaming"
 )
 
 // Runtime struct to contain the type(ID), engine, and root variables for a default runtime
@@ -227,51 +225,6 @@ type Config struct {
 	StateDir string `json:"stateDir"`
 }
 
-// DefaultConfig returns default configurations of cri plugin.
-func DefaultConfig() PluginConfig {
-	return PluginConfig{
-		CniConfig: CniConfig{
-			NetworkPluginBinDir:       "/opt/cni/bin",
-			NetworkPluginConfDir:      "/etc/cni/net.d",
-			NetworkPluginMaxConfNum:   1, // only one CNI plugin config file will be loaded
-			NetworkPluginConfTemplate: "",
-		},
-		ContainerdConfig: ContainerdConfig{
-			Snapshotter:        containerd.DefaultSnapshotter,
-			DefaultRuntimeName: "runc",
-			NoPivot:            false,
-			Runtimes: map[string]Runtime{
-				"runc": {
-					Type: "io.containerd.runc.v1",
-				},
-			},
-		},
-		DisableTCPService:   true,
-		StreamServerAddress: "127.0.0.1",
-		StreamServerPort:    "0",
-		StreamIdleTimeout:   streaming.DefaultConfig.StreamIdleTimeout.String(), // 4 hour
-		EnableSelinux:       false,
-		EnableTLSStreaming:  false,
-		X509KeyPairStreaming: X509KeyPairStreaming{
-			TLSKeyFile:  "",
-			TLSCertFile: "",
-		},
-		SandboxImage:            "k8s.gcr.io/pause:3.1",
-		StatsCollectPeriod:      10,
-		SystemdCgroup:           false,
-		MaxContainerLogLineSize: 16 * 1024,
-		Registry: Registry{
-			Mirrors: map[string]Mirror{
-				"docker.io": {
-					Endpoints: []string{"https://registry-1.docker.io"},
-				},
-			},
-		},
-		MaxConcurrentDownloads: 3,
-		DisableProcMount:       false,
-	}
-}
-
 const (
 	// RuntimeUntrusted is the implicit runtime defined for ContainerdConfig.UntrustedWorkloadRuntime
 	RuntimeUntrusted = "untrusted"

pkg/config/config_unix.go

@@ -0,0 +1,69 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package config
+
+import (
+	"github.com/containerd/containerd"
+	"k8s.io/kubernetes/pkg/kubelet/server/streaming"
+)
+
+// DefaultConfig returns default configurations of cri plugin.
+func DefaultConfig() PluginConfig {
+	return PluginConfig{
+		CniConfig: CniConfig{
+			NetworkPluginBinDir:       "/opt/cni/bin",
+			NetworkPluginConfDir:      "/etc/cni/net.d",
+			NetworkPluginMaxConfNum:   1, // only one CNI plugin config file will be loaded
+			NetworkPluginConfTemplate: "",
+		},
+		ContainerdConfig: ContainerdConfig{
+			Snapshotter:        containerd.DefaultSnapshotter,
+			DefaultRuntimeName: "runc",
+			NoPivot:            false,
+			Runtimes: map[string]Runtime{
+				"runc": {
+					Type: "io.containerd.runc.v1",
+				},
+			},
+		},
+		DisableTCPService:   true,
+		StreamServerAddress: "127.0.0.1",
+		StreamServerPort:    "0",
+		StreamIdleTimeout:   streaming.DefaultConfig.StreamIdleTimeout.String(), // 4 hour
+		EnableSelinux:       false,
+		EnableTLSStreaming:  false,
+		X509KeyPairStreaming: X509KeyPairStreaming{
+			TLSKeyFile:  "",
+			TLSCertFile: "",
+		},
+		SandboxImage:            "k8s.gcr.io/pause:3.1",
+		StatsCollectPeriod:      10,
+		SystemdCgroup:           false,
+		MaxContainerLogLineSize: 16 * 1024,
+		Registry: Registry{
+			Mirrors: map[string]Mirror{
+				"docker.io": {
+					Endpoints: []string{"https://registry-1.docker.io"},
+				},
+			},
+		},
+		MaxConcurrentDownloads: 3,
+		DisableProcMount:       false,
+	}
+}

pkg/config/config_windows.go

@@ -0,0 +1,24 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package config
+
+// DefaultConfig returns default configurations of cri plugin.
+func DefaultConfig() PluginConfig {
+	return PluginConfig{}
+}

pkg/containerd/opts/spec.go

@@ -18,100 +18,19 @@ package opts
 
 import (
 	"context"
-	"fmt"
-	"io/ioutil"
 	"os"
 	"path/filepath"
-	"sort"
-	"strconv"
 	"strings"
 
 	"github.com/containerd/containerd/containers"
-	"github.com/containerd/containerd/log"
-	"github.com/containerd/containerd/mount"
 	"github.com/containerd/containerd/oci"
 
 	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
-	"github.com/opencontainers/runc/libcontainer/devices"
 	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
-	"github.com/opencontainers/selinux/go-selinux/label"
 	"github.com/pkg/errors"
-	"golang.org/x/sys/unix"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
-
-	osinterface "github.com/containerd/cri/pkg/os"
-	"github.com/containerd/cri/pkg/util"
 )
 
-const (
-	// DefaultSandboxCPUshares is default cpu shares for sandbox container.
-	DefaultSandboxCPUshares = 2
-)
-
-// WithAdditionalGIDs adds any additional groups listed for a particular user in the
-// /etc/groups file of the image's root filesystem to the OCI spec's additionalGids array.
-func WithAdditionalGIDs(userstr string) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
-		if s.Process == nil {
-			s.Process = &runtimespec.Process{}
-		}
-		gids := s.Process.User.AdditionalGids
-		if err := oci.WithAdditionalGIDs(userstr)(ctx, client, c, s); err != nil {
-			return err
-		}
-		// Merge existing gids and new gids.
-		s.Process.User.AdditionalGids = mergeGids(s.Process.User.AdditionalGids, gids)
-		return nil
-	}
-}
-
-func mergeGids(gids1, gids2 []uint32) []uint32 {
-	gidsMap := make(map[uint32]struct{})
-	for _, gid1 := range gids1 {
-		gidsMap[gid1] = struct{}{}
-	}
-	for _, gid2 := range gids2 {
-		gidsMap[gid2] = struct{}{}
-	}
-	var gids []uint32
-	for gid := range gidsMap {
-		gids = append(gids, gid)
-	}
-	sort.Slice(gids, func(i, j int) bool { return gids[i] < gids[j] })
-	return gids
-}
-
-// WithoutRunMount removes the `/run` inside the spec
-func WithoutRunMount(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-	var (
-		mounts  []runtimespec.Mount
-		current = s.Mounts
-	)
-	for _, m := range current {
-		if filepath.Clean(m.Destination) == "/run" {
-			continue
-		}
-		mounts = append(mounts, m)
-	}
-	s.Mounts = mounts
-	return nil
-}
-
-// WithoutDefaultSecuritySettings removes the default security settings generated on a spec
-func WithoutDefaultSecuritySettings(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-	if s.Process == nil {
-		s.Process = &runtimespec.Process{}
-	}
-	// Make sure no default seccomp/apparmor is specified
-	s.Process.ApparmorProfile = ""
-	if s.Linux != nil {
-		s.Linux.Seccomp = nil
-	}
-	// Remove default rlimits (See issue #515)
-	s.Process.Rlimits = nil
-	return nil
-}
-
 // WithRelativeRoot sets the root for the container
 func WithRelativeRoot(root string) oci.SpecOpts {
 	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
@@ -145,141 +64,6 @@ func WithProcessArgs(config *runtime.ContainerConfig, image *imagespec.ImageConf
 	}
 }
 
-// WithMounts sorts and adds runtime and CRI mounts to the spec
-func WithMounts(osi osinterface.OS, config *runtime.ContainerConfig, extra []*runtime.Mount, mountLabel string) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, _ *containers.Container, s *runtimespec.Spec) (err error) {
-		// mergeMounts merge CRI mounts with extra mounts. If a mount destination
-		// is mounted by both a CRI mount and an extra mount, the CRI mount will
-		// be kept.
-		var (
-			criMounts = config.GetMounts()
-			mounts    = append([]*runtime.Mount{}, criMounts...)
-		)
-		// Copy all mounts from extra mounts, except for mounts overriden by CRI.
-		for _, e := range extra {
-			found := false
-			for _, c := range criMounts {
-				if filepath.Clean(e.ContainerPath) == filepath.Clean(c.ContainerPath) {
-					found = true
-					break
-				}
-			}
-			if !found {
-				mounts = append(mounts, e)
-			}
-		}
-		// ---
-
-		// Sort mounts in number of parts. This ensures that high level mounts don't
-		// shadow other mounts.
-		sort.Sort(orderedMounts(mounts))
-
-		// Mount cgroup into the container as readonly, which inherits docker's behavior.
-		s.Mounts = append(s.Mounts, runtimespec.Mount{
-			Source:      "cgroup",
-			Destination: "/sys/fs/cgroup",
-			Type:        "cgroup",
-			Options:     []string{"nosuid", "noexec", "nodev", "relatime", "ro"},
-		})
-
-		// Copy all mounts from default mounts, except for
-		// - mounts overriden by supplied mount;
-		// - all mounts under /dev if a supplied /dev is present.
-		mountSet := make(map[string]struct{})
-		for _, m := range mounts {
-			mountSet[filepath.Clean(m.ContainerPath)] = struct{}{}
-		}
-
-		defaultMounts := s.Mounts
-		s.Mounts = nil
-
-		for _, m := range defaultMounts {
-			dst := filepath.Clean(m.Destination)
-			if _, ok := mountSet[dst]; ok {
-				// filter out mount overridden by a supplied mount
-				continue
-			}
-			if _, mountDev := mountSet["/dev"]; mountDev && strings.HasPrefix(dst, "/dev/") {
-				// filter out everything under /dev if /dev is a supplied mount
-				continue
-			}
-			s.Mounts = append(s.Mounts, m)
-		}
-
-		for _, mount := range mounts {
-			var (
-				dst = mount.GetContainerPath()
-				src = mount.GetHostPath()
-			)
-			// Create the host path if it doesn't exist.
-			// TODO(random-liu): Add CRI validation test for this case.
-			if _, err := osi.Stat(src); err != nil {
-				if !os.IsNotExist(err) {
-					return errors.Wrapf(err, "failed to stat %q", src)
-				}
-				if err := osi.MkdirAll(src, 0755); err != nil {
-					return errors.Wrapf(err, "failed to mkdir %q", src)
-				}
-			}
-			// TODO(random-liu): Add cri-containerd integration test or cri validation test
-			// for this.
-			src, err := osi.ResolveSymbolicLink(src)
-			if err != nil {
-				return errors.Wrapf(err, "failed to resolve symlink %q", src)
-			}
-			if s.Linux == nil {
-				s.Linux = &runtimespec.Linux{}
-			}
-			options := []string{"rbind"}
-			switch mount.GetPropagation() {
-			case runtime.MountPropagation_PROPAGATION_PRIVATE:
-				options = append(options, "rprivate")
-				// Since default root propogation in runc is rprivate ignore
-				// setting the root propagation
-			case runtime.MountPropagation_PROPAGATION_BIDIRECTIONAL:
-				if err := ensureShared(src, osi.LookupMount); err != nil {
-					return err
-				}
-				options = append(options, "rshared")
-				s.Linux.RootfsPropagation = "rshared"
-			case runtime.MountPropagation_PROPAGATION_HOST_TO_CONTAINER:
-				if err := ensureSharedOrSlave(src, osi.LookupMount); err != nil {
-					return err
-				}
-				options = append(options, "rslave")
-				if s.Linux.RootfsPropagation != "rshared" &&
-					s.Linux.RootfsPropagation != "rslave" {
-					s.Linux.RootfsPropagation = "rslave"
-				}
-			default:
-				log.G(ctx).Warnf("Unknown propagation mode for hostPath %q", mount.HostPath)
-				options = append(options, "rprivate")
-			}
-
-			// NOTE(random-liu): we don't change all mounts to `ro` when root filesystem
-			// is readonly. This is different from docker's behavior, but make more sense.
-			if mount.GetReadonly() {
-				options = append(options, "ro")
-			} else {
-				options = append(options, "rw")
-			}
-
-			if mount.GetSelinuxRelabel() {
-				if err := label.Relabel(src, mountLabel, true); err != nil && err != unix.ENOTSUP {
-					return errors.Wrapf(err, "relabel %q with %q failed", src, mountLabel)
-				}
-			}
-			s.Mounts = append(s.Mounts, runtimespec.Mount{
-				Source:      src,
-				Destination: dst,
-				Type:        "bind",
-				Options:     options,
-			})
-		}
-		return nil
-	}
-}
-
 // mounts defines how to sort runtime.Mount.
 // This is the same with the Docker implementation:
 //   https://github.com/moby/moby/blob/17.05.x/daemon/volumes.go#L26
@@ -307,325 +91,6 @@ func (m orderedMounts) parts(i int) int {
 	return strings.Count(filepath.Clean(m[i].ContainerPath), string(os.PathSeparator))
 }
 
-// Ensure mount point on which path is mounted, is shared.
-func ensureShared(path string, lookupMount func(string) (mount.Info, error)) error {
-	mountInfo, err := lookupMount(path)
-	if err != nil {
-		return err
-	}
-
-	// Make sure source mount point is shared.
-	optsSplit := strings.Split(mountInfo.Optional, " ")
-	for _, opt := range optsSplit {
-		if strings.HasPrefix(opt, "shared:") {
-			return nil
-		}
-	}
-
-	return errors.Errorf("path %q is mounted on %q but it is not a shared mount", path, mountInfo.Mountpoint)
-}
-
-// ensure mount point on which path is mounted, is either shared or slave.
-func ensureSharedOrSlave(path string, lookupMount func(string) (mount.Info, error)) error {
-	mountInfo, err := lookupMount(path)
-	if err != nil {
-		return err
-	}
-	// Make sure source mount point is shared.
-	optsSplit := strings.Split(mountInfo.Optional, " ")
-	for _, opt := range optsSplit {
-		if strings.HasPrefix(opt, "shared:") {
-			return nil
-		} else if strings.HasPrefix(opt, "master:") {
-			return nil
-		}
-	}
-	return errors.Errorf("path %q is mounted on %q but it is not a shared or slave mount", path, mountInfo.Mountpoint)
-}
-
-// WithPrivilegedDevices allows all host devices inside the container
-func WithPrivilegedDevices(_ context.Context, _ oci.Client, _ *containers.Container, s *runtimespec.Spec) error {
-	if s.Linux == nil {
-		s.Linux = &runtimespec.Linux{}
-	}
-	if s.Linux.Resources == nil {
-		s.Linux.Resources = &runtimespec.LinuxResources{}
-	}
-	hostDevices, err := devices.HostDevices()
-	if err != nil {
-		return err
-	}
-	for _, hostDevice := range hostDevices {
-		rd := runtimespec.LinuxDevice{
-			Path:  hostDevice.Path,
-			Type:  string(hostDevice.Type),
-			Major: hostDevice.Major,
-			Minor: hostDevice.Minor,
-			UID:   &hostDevice.Uid,
-			GID:   &hostDevice.Gid,
-		}
-		if hostDevice.Major == 0 && hostDevice.Minor == 0 {
-			// Invalid device, most likely a symbolic link, skip it.
-			continue
-		}
-		addDevice(s, rd)
-	}
-	s.Linux.Resources.Devices = []runtimespec.LinuxDeviceCgroup{
-		{
-			Allow:  true,
-			Access: "rwm",
-		},
-	}
-	return nil
-}
-
-func addDevice(s *runtimespec.Spec, rd runtimespec.LinuxDevice) {
-	for i, dev := range s.Linux.Devices {
-		if dev.Path == rd.Path {
-			s.Linux.Devices[i] = rd
-			return
-		}
-	}
-	s.Linux.Devices = append(s.Linux.Devices, rd)
-}
-
-// WithDevices sets the provided devices onto the container spec
-func WithDevices(osi osinterface.OS, config *runtime.ContainerConfig) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
-		if s.Linux == nil {
-			s.Linux = &runtimespec.Linux{}
-		}
-		if s.Linux.Resources == nil {
-			s.Linux.Resources = &runtimespec.LinuxResources{}
-		}
-		for _, device := range config.GetDevices() {
-			path, err := osi.ResolveSymbolicLink(device.HostPath)
-			if err != nil {
-				return err
-			}
-			dev, err := devices.DeviceFromPath(path, device.Permissions)
-			if err != nil {
-				return err
-			}
-			rd := runtimespec.LinuxDevice{
-				Path:  device.ContainerPath,
-				Type:  string(dev.Type),
-				Major: dev.Major,
-				Minor: dev.Minor,
-				UID:   &dev.Uid,
-				GID:   &dev.Gid,
-			}
-
-			addDevice(s, rd)
-
-			s.Linux.Resources.Devices = append(s.Linux.Resources.Devices, runtimespec.LinuxDeviceCgroup{
-				Allow:  true,
-				Type:   string(dev.Type),
-				Major:  &dev.Major,
-				Minor:  &dev.Minor,
-				Access: dev.Permissions,
-			})
-		}
-		return nil
-	}
-}
-
-// WithCapabilities sets the provided capabilties from the security context
-func WithCapabilities(sc *runtime.LinuxContainerSecurityContext) oci.SpecOpts {
-	capabilities := sc.GetCapabilities()
-	if capabilities == nil {
-		return nullOpt
-	}
-
-	var opts []oci.SpecOpts
-	// Add/drop all capabilities if "all" is specified, so that
-	// following individual add/drop could still work. E.g.
-	// AddCapabilities: []string{"ALL"}, DropCapabilities: []string{"CHOWN"}
-	// will be all capabilities without `CAP_CHOWN`.
-	if util.InStringSlice(capabilities.GetAddCapabilities(), "ALL") {
-		opts = append(opts, oci.WithAllCapabilities)
-	}
-	if util.InStringSlice(capabilities.GetDropCapabilities(), "ALL") {
-		opts = append(opts, oci.WithCapabilities(nil))
-	}
-
-	var caps []string
-	for _, c := range capabilities.GetAddCapabilities() {
-		if strings.ToUpper(c) == "ALL" {
-			continue
-		}
-		// Capabilities in CRI doesn't have `CAP_` prefix, so add it.
-		caps = append(caps, "CAP_"+strings.ToUpper(c))
-	}
-	opts = append(opts, oci.WithAddedCapabilities(caps))
-
-	caps = []string{}
-	for _, c := range capabilities.GetDropCapabilities() {
-		if strings.ToUpper(c) == "ALL" {
-			continue
-		}
-		caps = append(caps, "CAP_"+strings.ToUpper(c))
-	}
-	opts = append(opts, oci.WithDroppedCapabilities(caps))
-	return oci.Compose(opts...)
-}
-
-// WithoutAmbientCaps removes the ambient caps from the spec
-func WithoutAmbientCaps(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-	if s.Process == nil {
-		s.Process = &runtimespec.Process{}
-	}
-	if s.Process.Capabilities == nil {
-		s.Process.Capabilities = &runtimespec.LinuxCapabilities{}
-	}
-	s.Process.Capabilities.Ambient = nil
-	return nil
-}
-
-// WithDisabledCgroups clears the Cgroups Path from the spec
-func WithDisabledCgroups(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-	if s.Linux == nil {
-		s.Linux = &runtimespec.Linux{}
-	}
-	s.Linux.CgroupsPath = ""
-	return nil
-}
-
-// WithSelinuxLabels sets the mount and process labels
-func WithSelinuxLabels(process, mount string) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
-		if s.Linux == nil {
-			s.Linux = &runtimespec.Linux{}
-		}
-		if s.Process == nil {
-			s.Process = &runtimespec.Process{}
-		}
-		s.Linux.MountLabel = mount
-		s.Process.SelinuxLabel = process
-		return nil
-	}
-}
-
-// WithResources sets the provided resource restrictions
-func WithResources(resources *runtime.LinuxContainerResources) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
-		if resources == nil {
-			return nil
-		}
-		if s.Linux == nil {
-			s.Linux = &runtimespec.Linux{}
-		}
-		if s.Linux.Resources == nil {
-			s.Linux.Resources = &runtimespec.LinuxResources{}
-		}
-		if s.Linux.Resources.CPU == nil {
-			s.Linux.Resources.CPU = &runtimespec.LinuxCPU{}
-		}
-		if s.Linux.Resources.Memory == nil {
-			s.Linux.Resources.Memory = &runtimespec.LinuxMemory{}
-		}
-		var (
-			p      = uint64(resources.GetCpuPeriod())
-			q      = resources.GetCpuQuota()
-			shares = uint64(resources.GetCpuShares())
-			limit  = resources.GetMemoryLimitInBytes()
-		)
-
-		if p != 0 {
-			s.Linux.Resources.CPU.Period = &p
-		}
-		if q != 0 {
-			s.Linux.Resources.CPU.Quota = &q
-		}
-		if shares != 0 {
-			s.Linux.Resources.CPU.Shares = &shares
-		}
-		if cpus := resources.GetCpusetCpus(); cpus != "" {
-			s.Linux.Resources.CPU.Cpus = cpus
-		}
-		if mems := resources.GetCpusetMems(); mems != "" {
-			s.Linux.Resources.CPU.Mems = resources.GetCpusetMems()
-		}
-		if limit != 0 {
-			s.Linux.Resources.Memory.Limit = &limit
-		}
-		return nil
-	}
-}
-
-// WithOOMScoreAdj sets the oom score
-func WithOOMScoreAdj(config *runtime.ContainerConfig, restrict bool) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-		if s.Process == nil {
-			s.Process = &runtimespec.Process{}
-		}
-
-		resources := config.GetLinux().GetResources()
-		if resources == nil {
-			return nil
-		}
-		adj := int(resources.GetOomScoreAdj())
-		if restrict {
-			var err error
-			adj, err = restrictOOMScoreAdj(adj)
-			if err != nil {
-				return err
-			}
-		}
-		s.Process.OOMScoreAdj = &adj
-		return nil
-	}
-}
-
-// WithSysctls sets the provided sysctls onto the spec
-func WithSysctls(sysctls map[string]string) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-		if s.Linux == nil {
-			s.Linux = &runtimespec.Linux{}
-		}
-		if s.Linux.Sysctl == nil {
-			s.Linux.Sysctl = make(map[string]string)
-		}
-		for k, v := range sysctls {
-			s.Linux.Sysctl[k] = v
-		}
-		return nil
-	}
-}
-
-// WithPodOOMScoreAdj sets the oom score for the pod sandbox
-func WithPodOOMScoreAdj(adj int, restrict bool) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-		if s.Process == nil {
-			s.Process = &runtimespec.Process{}
-		}
-		if restrict {
-			var err error
-			adj, err = restrictOOMScoreAdj(adj)
-			if err != nil {
-				return err
-			}
-		}
-		s.Process.OOMScoreAdj = &adj
-		return nil
-	}
-}
-
-// WithSupplementalGroups sets the supplemental groups for the process
-func WithSupplementalGroups(groups []int64) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-		if s.Process == nil {
-			s.Process = &runtimespec.Process{}
-		}
-		var guids []uint32
-		for _, g := range groups {
-			guids = append(guids, uint32(g))
-		}
-		s.Process.User.AdditionalGids = mergeGids(s.Process.User.AdditionalGids, guids)
-		return nil
-	}
-}
-
 // WithAnnotation sets the provided annotation
 func WithAnnotation(k, v string) oci.SpecOpts {
 	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
@@ -636,110 +101,3 @@ func WithAnnotation(k, v string) oci.SpecOpts {
 		return nil
 	}
 }
-
-// WithPodNamespaces sets the pod namespaces for the container
-func WithPodNamespaces(config *runtime.LinuxContainerSecurityContext, pid uint32) oci.SpecOpts {
-	namespaces := config.GetNamespaceOptions()
-
-	opts := []oci.SpecOpts{
-		oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.NetworkNamespace, Path: GetNetworkNamespace(pid)}),
-		oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.IPCNamespace, Path: GetIPCNamespace(pid)}),
-		oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.UTSNamespace, Path: GetUTSNamespace(pid)}),
-	}
-	if namespaces.GetPid() != runtime.NamespaceMode_CONTAINER {
-		opts = append(opts, oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.PIDNamespace, Path: GetPIDNamespace(pid)}))
-	}
-	return oci.Compose(opts...)
-}
-
-// WithDefaultSandboxShares sets the default sandbox CPU shares
-func WithDefaultSandboxShares(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-	if s.Linux == nil {
-		s.Linux = &runtimespec.Linux{}
-	}
-	if s.Linux.Resources == nil {
-		s.Linux.Resources = &runtimespec.LinuxResources{}
-	}
-	if s.Linux.Resources.CPU == nil {
-		s.Linux.Resources.CPU = &runtimespec.LinuxCPU{}
-	}
-	i := uint64(DefaultSandboxCPUshares)
-	s.Linux.Resources.CPU.Shares = &i
-	return nil
-}
-
-// WithoutNamespace removes the provided namespace
-func WithoutNamespace(t runtimespec.LinuxNamespaceType) oci.SpecOpts {
-	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
-		if s.Linux == nil {
-			return nil
-		}
-		var namespaces []runtimespec.LinuxNamespace
-		for i, ns := range s.Linux.Namespaces {
-			if ns.Type != t {
-				namespaces = append(namespaces, s.Linux.Namespaces[i])
-			}
-		}
-		s.Linux.Namespaces = namespaces
-		return nil
-	}
-}
-
-func nullOpt(_ context.Context, _ oci.Client, _ *containers.Container, _ *runtimespec.Spec) error {
-	return nil
-}
-
-func getCurrentOOMScoreAdj() (int, error) {
-	b, err := ioutil.ReadFile("/proc/self/oom_score_adj")
-	if err != nil {
-		return 0, errors.Wrap(err, "could not get the daemon oom_score_adj")
-	}
-	s := strings.TrimSpace(string(b))
-	i, err := strconv.Atoi(s)
-	if err != nil {
-		return 0, errors.Wrap(err, "could not get the daemon oom_score_adj")
-	}
-	return i, nil
-}
-
-func restrictOOMScoreAdj(preferredOOMScoreAdj int) (int, error) {
-	currentOOMScoreAdj, err := getCurrentOOMScoreAdj()
-	if err != nil {
-		return preferredOOMScoreAdj, err
-	}
-	if preferredOOMScoreAdj < currentOOMScoreAdj {
-		return currentOOMScoreAdj, nil
-	}
-	return preferredOOMScoreAdj, nil
-}
-
-const (
-	// netNSFormat is the format of network namespace of a process.
-	netNSFormat = "/proc/%v/ns/net"
-	// ipcNSFormat is the format of ipc namespace of a process.
-	ipcNSFormat = "/proc/%v/ns/ipc"
-	// utsNSFormat is the format of uts namespace of a process.
-	utsNSFormat = "/proc/%v/ns/uts"
-	// pidNSFormat is the format of pid namespace of a process.
-	pidNSFormat = "/proc/%v/ns/pid"
-)
-
-// GetNetworkNamespace returns the network namespace of a process.
-func GetNetworkNamespace(pid uint32) string {
-	return fmt.Sprintf(netNSFormat, pid)
-}
-
-// GetIPCNamespace returns the ipc namespace of a process.
-func GetIPCNamespace(pid uint32) string {
-	return fmt.Sprintf(ipcNSFormat, pid)
-}
-
-// GetUTSNamespace returns the uts namespace of a process.
-func GetUTSNamespace(pid uint32) string {
-	return fmt.Sprintf(utsNSFormat, pid)
-}
-
-// GetPIDNamespace returns the pid namespace of a process.
-func GetPIDNamespace(pid uint32) string {
-	return fmt.Sprintf(pidNSFormat, pid)
-}

pkg/containerd/opts/spec_test.go

@@ -1,5 +1,5 @@
 /*
-Copyright 2018 The containerd Authors.
+Copyright The containerd Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -21,16 +21,9 @@ import (
 	"testing"
 
 	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
 )
 
-func TestMergeGids(t *testing.T) {
-	gids1 := []uint32{3, 2, 1}
-	gids2 := []uint32{2, 3, 4}
-	assert.Equal(t, []uint32{1, 2, 3, 4}, mergeGids(gids1, gids2))
-}
-
 func TestOrderedMounts(t *testing.T) {
 	mounts := []*runtime.Mount{
 		{ContainerPath: "/a/b/c"},
@@ -51,20 +44,3 @@ func TestOrderedMounts(t *testing.T) {
 	sort.Stable(orderedMounts(mounts))
 	assert.Equal(t, expected, mounts)
 }
-
-func TestRestrictOOMScoreAdj(t *testing.T) {
-	current, err := getCurrentOOMScoreAdj()
-	require.NoError(t, err)
-
-	got, err := restrictOOMScoreAdj(current - 1)
-	require.NoError(t, err)
-	assert.Equal(t, got, current)
-
-	got, err = restrictOOMScoreAdj(current)
-	require.NoError(t, err)
-	assert.Equal(t, got, current)
-
-	got, err = restrictOOMScoreAdj(current + 1)
-	require.NoError(t, err)
-	assert.Equal(t, got, current+1)
-}

pkg/containerd/opts/spec_unix.go

@@ -0,0 +1,676 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package opts
+
+import (
+	"context"
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"sort"
+	"strconv"
+	"strings"
+
+	"github.com/containerd/containerd/containers"
+	"github.com/containerd/containerd/log"
+	"github.com/containerd/containerd/mount"
+	"github.com/containerd/containerd/oci"
+	"github.com/opencontainers/runc/libcontainer/devices"
+	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/opencontainers/selinux/go-selinux/label"
+	"github.com/pkg/errors"
+	"golang.org/x/sys/unix"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	osinterface "github.com/containerd/cri/pkg/os"
+	"github.com/containerd/cri/pkg/util"
+)
+
+const (
+	// DefaultSandboxCPUshares is default cpu shares for sandbox container.
+	// TODO(windows): Evaluate whether this can be used for windows sandbox
+	// container cpu shares.
+	DefaultSandboxCPUshares = 2
+)
+
+// WithAdditionalGIDs adds any additional groups listed for a particular user in the
+// /etc/groups file of the image's root filesystem to the OCI spec's additionalGids array.
+func WithAdditionalGIDs(userstr string) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
+		if s.Process == nil {
+			s.Process = &runtimespec.Process{}
+		}
+		gids := s.Process.User.AdditionalGids
+		if err := oci.WithAdditionalGIDs(userstr)(ctx, client, c, s); err != nil {
+			return err
+		}
+		// Merge existing gids and new gids.
+		s.Process.User.AdditionalGids = mergeGids(s.Process.User.AdditionalGids, gids)
+		return nil
+	}
+}
+
+func mergeGids(gids1, gids2 []uint32) []uint32 {
+	gidsMap := make(map[uint32]struct{})
+	for _, gid1 := range gids1 {
+		gidsMap[gid1] = struct{}{}
+	}
+	for _, gid2 := range gids2 {
+		gidsMap[gid2] = struct{}{}
+	}
+	var gids []uint32
+	for gid := range gidsMap {
+		gids = append(gids, gid)
+	}
+	sort.Slice(gids, func(i, j int) bool { return gids[i] < gids[j] })
+	return gids
+}
+
+// WithoutRunMount removes the `/run` inside the spec
+func WithoutRunMount(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+	var (
+		mounts  []runtimespec.Mount
+		current = s.Mounts
+	)
+	for _, m := range current {
+		if filepath.Clean(m.Destination) == "/run" {
+			continue
+		}
+		mounts = append(mounts, m)
+	}
+	s.Mounts = mounts
+	return nil
+}
+
+// WithoutDefaultSecuritySettings removes the default security settings generated on a spec
+func WithoutDefaultSecuritySettings(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+	if s.Process == nil {
+		s.Process = &runtimespec.Process{}
+	}
+	// Make sure no default seccomp/apparmor is specified
+	s.Process.ApparmorProfile = ""
+	if s.Linux != nil {
+		s.Linux.Seccomp = nil
+	}
+	// Remove default rlimits (See issue #515)
+	s.Process.Rlimits = nil
+	return nil
+}
+
+// WithMounts sorts and adds runtime and CRI mounts to the spec
+func WithMounts(osi osinterface.OS, config *runtime.ContainerConfig, extra []*runtime.Mount, mountLabel string) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, _ *containers.Container, s *runtimespec.Spec) (err error) {
+		// mergeMounts merge CRI mounts with extra mounts. If a mount destination
+		// is mounted by both a CRI mount and an extra mount, the CRI mount will
+		// be kept.
+		var (
+			criMounts = config.GetMounts()
+			mounts    = append([]*runtime.Mount{}, criMounts...)
+		)
+		// Copy all mounts from extra mounts, except for mounts overriden by CRI.
+		for _, e := range extra {
+			found := false
+			for _, c := range criMounts {
+				if filepath.Clean(e.ContainerPath) == filepath.Clean(c.ContainerPath) {
+					found = true
+					break
+				}
+			}
+			if !found {
+				mounts = append(mounts, e)
+			}
+		}
+		// ---
+
+		// Sort mounts in number of parts. This ensures that high level mounts don't
+		// shadow other mounts.
+		sort.Sort(orderedMounts(mounts))
+
+		// Mount cgroup into the container as readonly, which inherits docker's behavior.
+		s.Mounts = append(s.Mounts, runtimespec.Mount{
+			Source:      "cgroup",
+			Destination: "/sys/fs/cgroup",
+			Type:        "cgroup",
+			Options:     []string{"nosuid", "noexec", "nodev", "relatime", "ro"},
+		})
+
+		// Copy all mounts from default mounts, except for
+		// - mounts overriden by supplied mount;
+		// - all mounts under /dev if a supplied /dev is present.
+		mountSet := make(map[string]struct{})
+		for _, m := range mounts {
+			mountSet[filepath.Clean(m.ContainerPath)] = struct{}{}
+		}
+
+		defaultMounts := s.Mounts
+		s.Mounts = nil
+
+		for _, m := range defaultMounts {
+			dst := filepath.Clean(m.Destination)
+			if _, ok := mountSet[dst]; ok {
+				// filter out mount overridden by a supplied mount
+				continue
+			}
+			if _, mountDev := mountSet["/dev"]; mountDev && strings.HasPrefix(dst, "/dev/") {
+				// filter out everything under /dev if /dev is a supplied mount
+				continue
+			}
+			s.Mounts = append(s.Mounts, m)
+		}
+
+		for _, mount := range mounts {
+			var (
+				dst = mount.GetContainerPath()
+				src = mount.GetHostPath()
+			)
+			// Create the host path if it doesn't exist.
+			// TODO(random-liu): Add CRI validation test for this case.
+			if _, err := osi.Stat(src); err != nil {
+				if !os.IsNotExist(err) {
+					return errors.Wrapf(err, "failed to stat %q", src)
+				}
+				if err := osi.MkdirAll(src, 0755); err != nil {
+					return errors.Wrapf(err, "failed to mkdir %q", src)
+				}
+			}
+			// TODO(random-liu): Add cri-containerd integration test or cri validation test
+			// for this.
+			src, err := osi.ResolveSymbolicLink(src)
+			if err != nil {
+				return errors.Wrapf(err, "failed to resolve symlink %q", src)
+			}
+			if s.Linux == nil {
+				s.Linux = &runtimespec.Linux{}
+			}
+			options := []string{"rbind"}
+			switch mount.GetPropagation() {
+			case runtime.MountPropagation_PROPAGATION_PRIVATE:
+				options = append(options, "rprivate")
+				// Since default root propogation in runc is rprivate ignore
+				// setting the root propagation
+			case runtime.MountPropagation_PROPAGATION_BIDIRECTIONAL:
+				if err := ensureShared(src, osi.(osinterface.UNIX).LookupMount); err != nil {
+					return err
+				}
+				options = append(options, "rshared")
+				s.Linux.RootfsPropagation = "rshared"
+			case runtime.MountPropagation_PROPAGATION_HOST_TO_CONTAINER:
+				if err := ensureSharedOrSlave(src, osi.(osinterface.UNIX).LookupMount); err != nil {
+					return err
+				}
+				options = append(options, "rslave")
+				if s.Linux.RootfsPropagation != "rshared" &&
+					s.Linux.RootfsPropagation != "rslave" {
+					s.Linux.RootfsPropagation = "rslave"
+				}
+			default:
+				log.G(ctx).Warnf("Unknown propagation mode for hostPath %q", mount.HostPath)
+				options = append(options, "rprivate")
+			}
+
+			// NOTE(random-liu): we don't change all mounts to `ro` when root filesystem
+			// is readonly. This is different from docker's behavior, but make more sense.
+			if mount.GetReadonly() {
+				options = append(options, "ro")
+			} else {
+				options = append(options, "rw")
+			}
+
+			if mount.GetSelinuxRelabel() {
+				if err := label.Relabel(src, mountLabel, true); err != nil && err != unix.ENOTSUP {
+					return errors.Wrapf(err, "relabel %q with %q failed", src, mountLabel)
+				}
+			}
+			s.Mounts = append(s.Mounts, runtimespec.Mount{
+				Source:      src,
+				Destination: dst,
+				Type:        "bind",
+				Options:     options,
+			})
+		}
+		return nil
+	}
+}
+
+// Ensure mount point on which path is mounted, is shared.
+func ensureShared(path string, lookupMount func(string) (mount.Info, error)) error {
+	mountInfo, err := lookupMount(path)
+	if err != nil {
+		return err
+	}
+
+	// Make sure source mount point is shared.
+	optsSplit := strings.Split(mountInfo.Optional, " ")
+	for _, opt := range optsSplit {
+		if strings.HasPrefix(opt, "shared:") {
+			return nil
+		}
+	}
+
+	return errors.Errorf("path %q is mounted on %q but it is not a shared mount", path, mountInfo.Mountpoint)
+}
+
+// ensure mount point on which path is mounted, is either shared or slave.
+func ensureSharedOrSlave(path string, lookupMount func(string) (mount.Info, error)) error {
+	mountInfo, err := lookupMount(path)
+	if err != nil {
+		return err
+	}
+	// Make sure source mount point is shared.
+	optsSplit := strings.Split(mountInfo.Optional, " ")
+	for _, opt := range optsSplit {
+		if strings.HasPrefix(opt, "shared:") {
+			return nil
+		} else if strings.HasPrefix(opt, "master:") {
+			return nil
+		}
+	}
+	return errors.Errorf("path %q is mounted on %q but it is not a shared or slave mount", path, mountInfo.Mountpoint)
+}
+
+// WithPrivilegedDevices allows all host devices inside the container
+func WithPrivilegedDevices(_ context.Context, _ oci.Client, _ *containers.Container, s *runtimespec.Spec) error {
+	if s.Linux == nil {
+		s.Linux = &runtimespec.Linux{}
+	}
+	if s.Linux.Resources == nil {
+		s.Linux.Resources = &runtimespec.LinuxResources{}
+	}
+	hostDevices, err := devices.HostDevices()
+	if err != nil {
+		return err
+	}
+	for _, hostDevice := range hostDevices {
+		rd := runtimespec.LinuxDevice{
+			Path:  hostDevice.Path,
+			Type:  string(hostDevice.Type),
+			Major: hostDevice.Major,
+			Minor: hostDevice.Minor,
+			UID:   &hostDevice.Uid,
+			GID:   &hostDevice.Gid,
+		}
+		if hostDevice.Major == 0 && hostDevice.Minor == 0 {
+			// Invalid device, most likely a symbolic link, skip it.
+			continue
+		}
+		addDevice(s, rd)
+	}
+	s.Linux.Resources.Devices = []runtimespec.LinuxDeviceCgroup{
+		{
+			Allow:  true,
+			Access: "rwm",
+		},
+	}
+	return nil
+}
+
+func addDevice(s *runtimespec.Spec, rd runtimespec.LinuxDevice) {
+	for i, dev := range s.Linux.Devices {
+		if dev.Path == rd.Path {
+			s.Linux.Devices[i] = rd
+			return
+		}
+	}
+	s.Linux.Devices = append(s.Linux.Devices, rd)
+}
+
+// WithDevices sets the provided devices onto the container spec
+func WithDevices(osi osinterface.OS, config *runtime.ContainerConfig) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
+		if s.Linux == nil {
+			s.Linux = &runtimespec.Linux{}
+		}
+		if s.Linux.Resources == nil {
+			s.Linux.Resources = &runtimespec.LinuxResources{}
+		}
+		for _, device := range config.GetDevices() {
+			path, err := osi.ResolveSymbolicLink(device.HostPath)
+			if err != nil {
+				return err
+			}
+			dev, err := devices.DeviceFromPath(path, device.Permissions)
+			if err != nil {
+				return err
+			}
+			rd := runtimespec.LinuxDevice{
+				Path:  device.ContainerPath,
+				Type:  string(dev.Type),
+				Major: dev.Major,
+				Minor: dev.Minor,
+				UID:   &dev.Uid,
+				GID:   &dev.Gid,
+			}
+
+			addDevice(s, rd)
+
+			s.Linux.Resources.Devices = append(s.Linux.Resources.Devices, runtimespec.LinuxDeviceCgroup{
+				Allow:  true,
+				Type:   string(dev.Type),
+				Major:  &dev.Major,
+				Minor:  &dev.Minor,
+				Access: dev.Permissions,
+			})
+		}
+		return nil
+	}
+}
+
+// WithCapabilities sets the provided capabilties from the security context
+func WithCapabilities(sc *runtime.LinuxContainerSecurityContext) oci.SpecOpts {
+	capabilities := sc.GetCapabilities()
+	if capabilities == nil {
+		return nullOpt
+	}
+
+	var opts []oci.SpecOpts
+	// Add/drop all capabilities if "all" is specified, so that
+	// following individual add/drop could still work. E.g.
+	// AddCapabilities: []string{"ALL"}, DropCapabilities: []string{"CHOWN"}
+	// will be all capabilities without `CAP_CHOWN`.
+	if util.InStringSlice(capabilities.GetAddCapabilities(), "ALL") {
+		opts = append(opts, oci.WithAllCapabilities)
+	}
+	if util.InStringSlice(capabilities.GetDropCapabilities(), "ALL") {
+		opts = append(opts, oci.WithCapabilities(nil))
+	}
+
+	var caps []string
+	for _, c := range capabilities.GetAddCapabilities() {
+		if strings.ToUpper(c) == "ALL" {
+			continue
+		}
+		// Capabilities in CRI doesn't have `CAP_` prefix, so add it.
+		caps = append(caps, "CAP_"+strings.ToUpper(c))
+	}
+	opts = append(opts, oci.WithAddedCapabilities(caps))
+
+	caps = []string{}
+	for _, c := range capabilities.GetDropCapabilities() {
+		if strings.ToUpper(c) == "ALL" {
+			continue
+		}
+		caps = append(caps, "CAP_"+strings.ToUpper(c))
+	}
+	opts = append(opts, oci.WithDroppedCapabilities(caps))
+	return oci.Compose(opts...)
+}
+
+// WithoutAmbientCaps removes the ambient caps from the spec
+func WithoutAmbientCaps(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+	if s.Process == nil {
+		s.Process = &runtimespec.Process{}
+	}
+	if s.Process.Capabilities == nil {
+		s.Process.Capabilities = &runtimespec.LinuxCapabilities{}
+	}
+	s.Process.Capabilities.Ambient = nil
+	return nil
+}
+
+// WithDisabledCgroups clears the Cgroups Path from the spec
+func WithDisabledCgroups(_ context.Context, _ oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+	if s.Linux == nil {
+		s.Linux = &runtimespec.Linux{}
+	}
+	s.Linux.CgroupsPath = ""
+	return nil
+}
+
+// WithSelinuxLabels sets the mount and process labels
+func WithSelinuxLabels(process, mount string) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
+		if s.Linux == nil {
+			s.Linux = &runtimespec.Linux{}
+		}
+		if s.Process == nil {
+			s.Process = &runtimespec.Process{}
+		}
+		s.Linux.MountLabel = mount
+		s.Process.SelinuxLabel = process
+		return nil
+	}
+}
+
+// WithResources sets the provided resource restrictions
+func WithResources(resources *runtime.LinuxContainerResources) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) (err error) {
+		if resources == nil {
+			return nil
+		}
+		if s.Linux == nil {
+			s.Linux = &runtimespec.Linux{}
+		}
+		if s.Linux.Resources == nil {
+			s.Linux.Resources = &runtimespec.LinuxResources{}
+		}
+		if s.Linux.Resources.CPU == nil {
+			s.Linux.Resources.CPU = &runtimespec.LinuxCPU{}
+		}
+		if s.Linux.Resources.Memory == nil {
+			s.Linux.Resources.Memory = &runtimespec.LinuxMemory{}
+		}
+		var (
+			p      = uint64(resources.GetCpuPeriod())
+			q      = resources.GetCpuQuota()
+			shares = uint64(resources.GetCpuShares())
+			limit  = resources.GetMemoryLimitInBytes()
+		)
+
+		if p != 0 {
+			s.Linux.Resources.CPU.Period = &p
+		}
+		if q != 0 {
+			s.Linux.Resources.CPU.Quota = &q
+		}
+		if shares != 0 {
+			s.Linux.Resources.CPU.Shares = &shares
+		}
+		if cpus := resources.GetCpusetCpus(); cpus != "" {
+			s.Linux.Resources.CPU.Cpus = cpus
+		}
+		if mems := resources.GetCpusetMems(); mems != "" {
+			s.Linux.Resources.CPU.Mems = resources.GetCpusetMems()
+		}
+		if limit != 0 {
+			s.Linux.Resources.Memory.Limit = &limit
+		}
+		return nil
+	}
+}
+
+// WithOOMScoreAdj sets the oom score
+func WithOOMScoreAdj(config *runtime.ContainerConfig, restrict bool) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+		if s.Process == nil {
+			s.Process = &runtimespec.Process{}
+		}
+
+		resources := config.GetLinux().GetResources()
+		if resources == nil {
+			return nil
+		}
+		adj := int(resources.GetOomScoreAdj())
+		if restrict {
+			var err error
+			adj, err = restrictOOMScoreAdj(adj)
+			if err != nil {
+				return err
+			}
+		}
+		s.Process.OOMScoreAdj = &adj
+		return nil
+	}
+}
+
+// WithSysctls sets the provided sysctls onto the spec
+func WithSysctls(sysctls map[string]string) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+		if s.Linux == nil {
+			s.Linux = &runtimespec.Linux{}
+		}
+		if s.Linux.Sysctl == nil {
+			s.Linux.Sysctl = make(map[string]string)
+		}
+		for k, v := range sysctls {
+			s.Linux.Sysctl[k] = v
+		}
+		return nil
+	}
+}
+
+// WithPodOOMScoreAdj sets the oom score for the pod sandbox
+func WithPodOOMScoreAdj(adj int, restrict bool) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+		if s.Process == nil {
+			s.Process = &runtimespec.Process{}
+		}
+		if restrict {
+			var err error
+			adj, err = restrictOOMScoreAdj(adj)
+			if err != nil {
+				return err
+			}
+		}
+		s.Process.OOMScoreAdj = &adj
+		return nil
+	}
+}
+
+// WithSupplementalGroups sets the supplemental groups for the process
+func WithSupplementalGroups(groups []int64) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+		if s.Process == nil {
+			s.Process = &runtimespec.Process{}
+		}
+		var guids []uint32
+		for _, g := range groups {
+			guids = append(guids, uint32(g))
+		}
+		s.Process.User.AdditionalGids = mergeGids(s.Process.User.AdditionalGids, guids)
+		return nil
+	}
+}
+
+// WithPodNamespaces sets the pod namespaces for the container
+func WithPodNamespaces(config *runtime.LinuxContainerSecurityContext, pid uint32) oci.SpecOpts {
+	namespaces := config.GetNamespaceOptions()
+
+	opts := []oci.SpecOpts{
+		oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.NetworkNamespace, Path: GetNetworkNamespace(pid)}),
+		oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.IPCNamespace, Path: GetIPCNamespace(pid)}),
+		oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.UTSNamespace, Path: GetUTSNamespace(pid)}),
+	}
+	if namespaces.GetPid() != runtime.NamespaceMode_CONTAINER {
+		opts = append(opts, oci.WithLinuxNamespace(runtimespec.LinuxNamespace{Type: runtimespec.PIDNamespace, Path: GetPIDNamespace(pid)}))
+	}
+	return oci.Compose(opts...)
+}
+
+// WithDefaultSandboxShares sets the default sandbox CPU shares
+func WithDefaultSandboxShares(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+	if s.Linux == nil {
+		s.Linux = &runtimespec.Linux{}
+	}
+	if s.Linux.Resources == nil {
+		s.Linux.Resources = &runtimespec.LinuxResources{}
+	}
+	if s.Linux.Resources.CPU == nil {
+		s.Linux.Resources.CPU = &runtimespec.LinuxCPU{}
+	}
+	i := uint64(DefaultSandboxCPUshares)
+	s.Linux.Resources.CPU.Shares = &i
+	return nil
+}
+
+// WithoutNamespace removes the provided namespace
+func WithoutNamespace(t runtimespec.LinuxNamespaceType) oci.SpecOpts {
+	return func(ctx context.Context, client oci.Client, c *containers.Container, s *runtimespec.Spec) error {
+		if s.Linux == nil {
+			return nil
+		}
+		var namespaces []runtimespec.LinuxNamespace
+		for i, ns := range s.Linux.Namespaces {
+			if ns.Type != t {
+				namespaces = append(namespaces, s.Linux.Namespaces[i])
+			}
+		}
+		s.Linux.Namespaces = namespaces
+		return nil
+	}
+}
+
+func nullOpt(_ context.Context, _ oci.Client, _ *containers.Container, _ *runtimespec.Spec) error {
+	return nil
+}
+
+func getCurrentOOMScoreAdj() (int, error) {
+	b, err := ioutil.ReadFile("/proc/self/oom_score_adj")
+	if err != nil {
+		return 0, errors.Wrap(err, "could not get the daemon oom_score_adj")
+	}
+	s := strings.TrimSpace(string(b))
+	i, err := strconv.Atoi(s)
+	if err != nil {
+		return 0, errors.Wrap(err, "could not get the daemon oom_score_adj")
+	}
+	return i, nil
+}
+
+func restrictOOMScoreAdj(preferredOOMScoreAdj int) (int, error) {
+	currentOOMScoreAdj, err := getCurrentOOMScoreAdj()
+	if err != nil {
+		return preferredOOMScoreAdj, err
+	}
+	if preferredOOMScoreAdj < currentOOMScoreAdj {
+		return currentOOMScoreAdj, nil
+	}
+	return preferredOOMScoreAdj, nil
+}
+
+const (
+	// netNSFormat is the format of network namespace of a process.
+	netNSFormat = "/proc/%v/ns/net"
+	// ipcNSFormat is the format of ipc namespace of a process.
+	ipcNSFormat = "/proc/%v/ns/ipc"
+	// utsNSFormat is the format of uts namespace of a process.
+	utsNSFormat = "/proc/%v/ns/uts"
+	// pidNSFormat is the format of pid namespace of a process.
+	pidNSFormat = "/proc/%v/ns/pid"
+)
+
+// GetNetworkNamespace returns the network namespace of a process.
+func GetNetworkNamespace(pid uint32) string {
+	return fmt.Sprintf(netNSFormat, pid)
+}
+
+// GetIPCNamespace returns the ipc namespace of a process.
+func GetIPCNamespace(pid uint32) string {
+	return fmt.Sprintf(ipcNSFormat, pid)
+}
+
+// GetUTSNamespace returns the uts namespace of a process.
+func GetUTSNamespace(pid uint32) string {
+	return fmt.Sprintf(utsNSFormat, pid)
+}
+
+// GetPIDNamespace returns the pid namespace of a process.
+func GetPIDNamespace(pid uint32) string {
+	return fmt.Sprintf(pidNSFormat, pid)
+}

pkg/containerd/opts/spec_unix_test.go

@@ -0,0 +1,49 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package opts
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestMergeGids(t *testing.T) {
+	gids1 := []uint32{3, 2, 1}
+	gids2 := []uint32{2, 3, 4}
+	assert.Equal(t, []uint32{1, 2, 3, 4}, mergeGids(gids1, gids2))
+}
+
+func TestRestrictOOMScoreAdj(t *testing.T) {
+	current, err := getCurrentOOMScoreAdj()
+	require.NoError(t, err)
+
+	got, err := restrictOOMScoreAdj(current - 1)
+	require.NoError(t, err)
+	assert.Equal(t, got, current)
+
+	got, err = restrictOOMScoreAdj(current)
+	require.NoError(t, err)
+	assert.Equal(t, got, current)
+
+	got, err = restrictOOMScoreAdj(current + 1)
+	require.NoError(t, err)
+	assert.Equal(t, got, current+1)
+}

pkg/ioutil/write_closer_test.go

@@ -69,7 +69,7 @@ func TestSerialWriteCloser(t *testing.T) {
 			testData[i] = []byte(repeatNumber(i, dataLen) + "\n")
 		}
 
-		f, err := ioutil.TempFile("/tmp", "serial-write-closer")
+		f, err := ioutil.TempFile("", "serial-write-closer")
 		require.NoError(t, err)
 		defer os.RemoveAll(f.Name())
 		defer f.Close()

pkg/netns/netns_unix.go

@@ -1,5 +1,7 @@
+// +build !windows
+
 /*
-Copyright 2018 The Containerd Authors.
+Copyright The containerd Authors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.

pkg/netns/netns_windows.go

@@ -0,0 +1,50 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package netns
+
+// TODO(windows): Implement netns for windows.
+// NetNS holds network namespace.
+type NetNS struct {
+}
+
+// NewNetNS creates a network namespace.
+func NewNetNS() (*NetNS, error) {
+	return nil, nil
+}
+
+// LoadNetNS loads existing network namespace.
+func LoadNetNS(path string) *NetNS {
+	return nil
+}
+
+// Remove removes network namepace. Remove is idempotent, meaning it might
+// be invoked multiple times and provides consistent result.
+func (n *NetNS) Remove() error {
+	return nil
+}
+
+// Closed checks whether the network namespace has been closed.
+func (n *NetNS) Closed() (bool, error) {
+	return false, nil
+}
+
+// GetPath returns network namespace path for sandbox container
+func (n *NetNS) GetPath() string {
+	return ""
+}

pkg/os/os.go

@@ -22,11 +22,7 @@ import (
 	"os"
 	"path/filepath"
 
-	"github.com/containerd/containerd/mount"
-	"github.com/containerd/fifo"
 	"github.com/docker/docker/pkg/symlink"
-	"golang.org/x/net/context"
-	"golang.org/x/sys/unix"
 )
 
 // OS collects system level operations that need to be mocked out
@@ -34,15 +30,11 @@ import (
 type OS interface {
 	MkdirAll(path string, perm os.FileMode) error
 	RemoveAll(path string) error
-	OpenFifo(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error)
 	Stat(name string) (os.FileInfo, error)
 	ResolveSymbolicLink(name string) (string, error)
 	FollowSymlinkInScope(path, scope string) (string, error)
 	CopyFile(src, dest string, perm os.FileMode) error
 	WriteFile(filename string, data []byte, perm os.FileMode) error
-	Mount(source string, target string, fstype string, flags uintptr, data string) error
-	Unmount(target string) error
-	LookupMount(path string) (mount.Info, error)
 	Hostname() (string, error)
 }
 
@@ -59,11 +51,6 @@ func (RealOS) RemoveAll(path string) error {
 	return os.RemoveAll(path)
 }
 
-// OpenFifo will call fifo.OpenFifo to open a fifo.
-func (RealOS) OpenFifo(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error) {
-	return fifo.OpenFifo(ctx, fn, flag, perm)
-}
-
 // Stat will call os.Stat to get the status of the given file.
 func (RealOS) Stat(name string) (os.FileInfo, error) {
 	return os.Stat(name)
@@ -109,33 +96,6 @@ func (RealOS) WriteFile(filename string, data []byte, perm os.FileMode) error {
 	return ioutil.WriteFile(filename, data, perm)
 }
 
-// Mount will call unix.Mount to mount the file.
-func (RealOS) Mount(source string, target string, fstype string, flags uintptr, data string) error {
-	return unix.Mount(source, target, fstype, flags, data)
-}
-
-// Unmount will call Unmount to unmount the file.
-func (RealOS) Unmount(target string) error {
-	return Unmount(target)
-}
-
-// LookupMount gets mount info of a given path.
-func (RealOS) LookupMount(path string) (mount.Info, error) {
-	return mount.Lookup(path)
-}
-
-// Unmount unmounts the target. It does not return an error in case the target is not mounted.
-// In case the target does not exist, the appropriate error is returned.
-func Unmount(target string) error {
-	err := unix.Unmount(target, unix.MNT_DETACH)
-	if err == unix.EINVAL {
-		// ignore "not mounted" error
-		err = nil
-	}
-
-	return err
-}
-
 // Hostname will call os.Hostname to get the hostname of the host.
 func (RealOS) Hostname() (string, error) {
 	return os.Hostname()

pkg/os/os_unix.go

@@ -0,0 +1,59 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package os
+
+import (
+	"github.com/containerd/containerd/mount"
+	"golang.org/x/sys/unix"
+)
+
+// UNIX collects unix system level operations that need to be
+// mocked out during tests.
+type UNIX interface {
+	Mount(source string, target string, fstype string, flags uintptr, data string) error
+	Unmount(target string) error
+	LookupMount(path string) (mount.Info, error)
+}
+
+// Mount will call unix.Mount to mount the file.
+func (RealOS) Mount(source string, target string, fstype string, flags uintptr, data string) error {
+	return unix.Mount(source, target, fstype, flags, data)
+}
+
+// Unmount will call Unmount to unmount the file.
+func (RealOS) Unmount(target string) error {
+	return Unmount(target)
+}
+
+// LookupMount gets mount info of a given path.
+func (RealOS) LookupMount(path string) (mount.Info, error) {
+	return mount.Lookup(path)
+}
+
+// Unmount unmounts the target. It does not return an error in case the target is not mounted.
+// In case the target does not exist, the appropriate error is returned.
+func Unmount(target string) error {
+	err := unix.Unmount(target, unix.MNT_DETACH)
+	if err == unix.EINVAL {
+		// ignore "not mounted" error
+		err = nil
+	}
+
+	return err
+}

pkg/os/testing/fake_os.go

@@ -17,12 +17,10 @@ limitations under the License.
 package testing
 
 import (
-	"io"
 	"os"
 	"sync"
 
 	containerdmount "github.com/containerd/containerd/mount"
-	"golang.org/x/net/context"
 
 	osInterface "github.com/containerd/cri/pkg/os"
 )
@@ -42,7 +40,6 @@ type FakeOS struct {
 	sync.Mutex
 	MkdirAllFn             func(string, os.FileMode) error
 	RemoveAllFn            func(string) error
-	OpenFifoFn             func(context.Context, string, int, os.FileMode) (io.ReadWriteCloser, error)
 	StatFn                 func(string) (os.FileInfo, error)
 	ResolveSymbolicLinkFn  func(string) (string, error)
 	FollowSymlinkInScopeFn func(string, string) (string, error)
@@ -139,19 +136,6 @@ func (f *FakeOS) RemoveAll(path string) error {
 	return nil
 }
 
-// OpenFifo is a fake call that invokes OpenFifoFn or just returns nil.
-func (f *FakeOS) OpenFifo(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error) {
-	f.appendCalls("OpenFifo", ctx, fn, flag, perm)
-	if err := f.getError("OpenFifo"); err != nil {
-		return nil, err
-	}
-
-	if f.OpenFifoFn != nil {
-		return f.OpenFifoFn(ctx, fn, flag, perm)
-	}
-	return nil, nil
-}
-
 // Stat is a fake call that invokes StatFn or just return nil.
 func (f *FakeOS) Stat(name string) (os.FileInfo, error) {
 	f.appendCalls("Stat", name)

pkg/os/testing/fake_os_unix.go

@@ -0,0 +1,23 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package testing
+
+import osInterface "github.com/containerd/cri/pkg/os"
+
+var _ osInterface.UNIX = &FakeOS{}

pkg/server/container_create.go

@@ -18,23 +18,12 @@ package server
 
 import (
 	"path/filepath"
-	"strconv"
-	"strings"
 	"time"
 
 	"github.com/containerd/containerd"
 	"github.com/containerd/containerd/containers"
-	"github.com/containerd/containerd/contrib/apparmor"
-	"github.com/containerd/containerd/contrib/seccomp"
 	"github.com/containerd/containerd/log"
 	"github.com/containerd/containerd/oci"
-	"github.com/containerd/cri/pkg/annotations"
-	"github.com/containerd/cri/pkg/config"
-	customopts "github.com/containerd/cri/pkg/containerd/opts"
-	ctrdutil "github.com/containerd/cri/pkg/containerd/util"
-	cio "github.com/containerd/cri/pkg/server/io"
-	containerstore "github.com/containerd/cri/pkg/store/container"
-	"github.com/containerd/cri/pkg/util"
 	"github.com/containerd/typeurl"
 	"github.com/davecgh/go-spew/spew"
 	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
@@ -42,21 +31,12 @@ import (
 	"github.com/pkg/errors"
 	"golang.org/x/net/context"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
-)
 
-const (
+	customopts "github.com/containerd/cri/pkg/containerd/opts"
-	// profileNamePrefix is the prefix for loading profiles on a localhost. Eg. AppArmor localhost/profileName.
+	ctrdutil "github.com/containerd/cri/pkg/containerd/util"
-	profileNamePrefix = "localhost/" // TODO (mikebrow): get localhost/ & runtime/default from CRI kubernetes/kubernetes#51747
+	cio "github.com/containerd/cri/pkg/server/io"
-	// runtimeDefault indicates that we should use or create a runtime default profile.
+	containerstore "github.com/containerd/cri/pkg/store/container"
-	runtimeDefault = "runtime/default"
+	"github.com/containerd/cri/pkg/util"
-	// dockerDefault indicates that we should use or create a docker default profile.
-	dockerDefault = "docker/default"
-	// appArmorDefaultProfileName is name to use when creating a default apparmor profile.
-	appArmorDefaultProfileName = "cri-containerd.apparmor.d"
-	// unconfinedProfile is a string indicating one should run a pod/containerd without a security profile
-	unconfinedProfile = "unconfined"
-	// seccompDefaultProfile is the default seccomp profile.
-	seccompDefaultProfile = dockerDefault
 )
 
 func init() {
@@ -156,10 +136,10 @@ func (c *criService) CreateContainer(ctx context.Context, r *runtime.CreateConta
 	}()
 
 	// Create container volumes mounts.
-	volumeMounts := c.generateVolumeMounts(containerRootDir, config.GetMounts(), &image.ImageSpec.Config)
+	volumeMounts := c.volumeMounts(containerRootDir, config.GetMounts(), &image.ImageSpec.Config)
 
-	// Generate container runtime spec.
+	// Generate container mounts.
-	mounts := c.generateContainerMounts(sandboxID, config)
+	mounts := c.containerMounts(sandboxID, config)
 
 	ociRuntime, err := c.getSandboxRuntime(sandboxConfig, sandbox.Metadata.RuntimeHandler)
 	if err != nil {
@@ -167,7 +147,7 @@ func (c *criService) CreateContainer(ctx context.Context, r *runtime.CreateConta
 	}
 	log.G(ctx).Debugf("Use OCI runtime %+v for sandbox %q and container %q", ociRuntime, sandboxID, id)
 
-	spec, err := c.generateContainerSpec(id, sandboxID, sandboxPid, config, sandboxConfig,
+	spec, err := c.containerSpec(id, sandboxID, sandboxPid, sandbox.NetNSPath, config, sandboxConfig,
 		&image.ImageSpec.Config, append(mounts, volumeMounts...), ociRuntime)
 	if err != nil {
 		return nil, errors.Wrapf(err, "failed to generate container %q spec", id)
@@ -185,7 +165,6 @@ func (c *criService) CreateContainer(ctx context.Context, r *runtime.CreateConta
 		// rootfs readonly (requested by spec.Root.Readonly).
 		customopts.WithNewSnapshot(id, containerdImage),
 	}
-
 	if len(volumeMounts) > 0 {
 		mountMap := make(map[string]string)
 		for _, v := range volumeMounts {
@@ -219,58 +198,11 @@ func (c *criService) CreateContainer(ctx context.Context, r *runtime.CreateConta
 		}
 	}()
 
-	var specOpts []oci.SpecOpts
+	specOpts, err := c.containerSpecOpts(config, &image.ImageSpec.Config)
-	securityContext := config.GetLinux().GetSecurityContext()
-	// Set container username. This could only be done by containerd, because it needs
-	// access to the container rootfs. Pass user name to containerd, and let it overwrite
-	// the spec for us.
-	userstr, err := generateUserString(
-		securityContext.GetRunAsUsername(),
-		securityContext.GetRunAsUser(),
-		securityContext.GetRunAsGroup())
-
 	if err != nil {
-		return nil, errors.Wrap(err, "failed to generate user string")
+		return nil, errors.Wrap(err, "")
-	}
-	if userstr == "" {
-		// Lastly, since no user override was passed via CRI try to set via OCI
-		// Image
-		userstr = image.ImageSpec.Config.User
-	}
-	if userstr != "" {
-		specOpts = append(specOpts, oci.WithUser(userstr))
 	}
 
-	if securityContext.GetRunAsUsername() != "" {
-		userstr = securityContext.GetRunAsUsername()
-	} else {
-		// Even if RunAsUser is not set, we still call `GetValue` to get uid 0.
-		// Because it is still useful to get additional gids for uid 0.
-		userstr = strconv.FormatInt(securityContext.GetRunAsUser().GetValue(), 10)
-	}
-	specOpts = append(specOpts, customopts.WithAdditionalGIDs(userstr))
-
-	apparmorSpecOpts, err := generateApparmorSpecOpts(
-		securityContext.GetApparmorProfile(),
-		securityContext.GetPrivileged(),
-		c.apparmorEnabled)
-	if err != nil {
-		return nil, errors.Wrap(err, "failed to generate apparmor spec opts")
-	}
-	if apparmorSpecOpts != nil {
-		specOpts = append(specOpts, apparmorSpecOpts)
-	}
-
-	seccompSpecOpts, err := generateSeccompSpecOpts(
-		securityContext.GetSeccompProfilePath(),
-		securityContext.GetPrivileged(),
-		c.seccompEnabled)
-	if err != nil {
-		return nil, errors.Wrap(err, "failed to generate seccomp spec opts")
-	}
-	if seccompSpecOpts != nil {
-		specOpts = append(specOpts, seccompSpecOpts)
-	}
 	containerLabels := buildLabels(config.Labels, containerKindContainer)
 
 	runtimeOptions, err := getRuntimeOptions(sandboxInfo)
@@ -322,128 +254,10 @@ func (c *criService) CreateContainer(ctx context.Context, r *runtime.CreateConta
 	return &runtime.CreateContainerResponse{ContainerId: id}, nil
 }
 
-func (c *criService) generateContainerSpec(id string, sandboxID string, sandboxPid uint32, config *runtime.ContainerConfig,
+// volumeMounts sets up image volumes for container. Rely on the removal of container
-	sandboxConfig *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig, extraMounts []*runtime.Mount,
-	ociRuntime config.Runtime) (*runtimespec.Spec, error) {
-
-	specOpts := []oci.SpecOpts{
-		customopts.WithoutRunMount,
-		customopts.WithoutDefaultSecuritySettings,
-		customopts.WithRelativeRoot(relativeRootfsPath),
-		customopts.WithProcessArgs(config, imageConfig),
-		// this will be set based on the security context below
-		oci.WithNewPrivileges,
-	}
-	if config.GetWorkingDir() != "" {
-		specOpts = append(specOpts, oci.WithProcessCwd(config.GetWorkingDir()))
-	} else if imageConfig.WorkingDir != "" {
-		specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
-	}
-
-	if config.GetTty() {
-		specOpts = append(specOpts, oci.WithTTY)
-	}
-
-	// Add HOSTNAME env.
-	var (
-		err      error
-		hostname = sandboxConfig.GetHostname()
-	)
-	if hostname == "" {
-		if hostname, err = c.os.Hostname(); err != nil {
-			return nil, err
-		}
-	}
-	specOpts = append(specOpts, oci.WithEnv([]string{hostnameEnv + "=" + hostname}))
-
-	// Apply envs from image config first, so that envs from container config
-	// can override them.
-	env := imageConfig.Env
-	for _, e := range config.GetEnvs() {
-		env = append(env, e.GetKey()+"="+e.GetValue())
-	}
-	specOpts = append(specOpts, oci.WithEnv(env))
-
-	securityContext := config.GetLinux().GetSecurityContext()
-	selinuxOpt := securityContext.GetSelinuxOptions()
-	processLabel, mountLabel, err := initSelinuxOpts(selinuxOpt)
-	if err != nil {
-		return nil, errors.Wrapf(err, "failed to init selinux options %+v", securityContext.GetSelinuxOptions())
-	}
-	specOpts = append(specOpts, customopts.WithMounts(c.os, config, extraMounts, mountLabel))
-
-	if !c.config.DisableProcMount {
-		// Apply masked paths if specified.
-		// If the container is privileged, this will be cleared later on.
-		specOpts = append(specOpts, oci.WithMaskedPaths(securityContext.GetMaskedPaths()))
-
-		// Apply readonly paths if specified.
-		// If the container is privileged, this will be cleared later on.
-		specOpts = append(specOpts, oci.WithReadonlyPaths(securityContext.GetReadonlyPaths()))
-	}
-
-	if securityContext.GetPrivileged() {
-		if !sandboxConfig.GetLinux().GetSecurityContext().GetPrivileged() {
-			return nil, errors.New("no privileged container allowed in sandbox")
-		}
-		specOpts = append(specOpts, oci.WithPrivileged)
-		if !ociRuntime.PrivilegedWithoutHostDevices {
-			specOpts = append(specOpts, customopts.WithPrivilegedDevices)
-		}
-	} else { // not privileged
-		specOpts = append(specOpts, customopts.WithDevices(c.os, config), customopts.WithCapabilities(securityContext))
-	}
-
-	// Clear all ambient capabilities. The implication of non-root + caps
-	// is not clearly defined in Kubernetes.
-	// See https://github.com/kubernetes/kubernetes/issues/56374
-	// Keep docker's behavior for now.
-	specOpts = append(specOpts,
-		customopts.WithoutAmbientCaps,
-		customopts.WithSelinuxLabels(processLabel, mountLabel),
-	)
-
-	// TODO: Figure out whether we should set no new privilege for sandbox container by default
-	if securityContext.GetNoNewPrivs() {
-		specOpts = append(specOpts, oci.WithNoNewPrivileges)
-	}
-	// TODO(random-liu): [P1] Set selinux options (privileged or not).
-	if securityContext.GetReadonlyRootfs() {
-		specOpts = append(specOpts, oci.WithRootFSReadonly())
-	}
-
-	if c.config.DisableCgroup {
-		specOpts = append(specOpts, customopts.WithDisabledCgroups)
-	} else {
-		specOpts = append(specOpts, customopts.WithResources(config.GetLinux().GetResources()))
-		if sandboxConfig.GetLinux().GetCgroupParent() != "" {
-			cgroupsPath := getCgroupsPath(sandboxConfig.GetLinux().GetCgroupParent(), id)
-			specOpts = append(specOpts, oci.WithCgroup(cgroupsPath))
-		}
-	}
-
-	supplementalGroups := securityContext.GetSupplementalGroups()
-
-	for pKey, pValue := range getPassthroughAnnotations(sandboxConfig.Annotations,
-		ociRuntime.PodAnnotations) {
-		specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
-	}
-
-	specOpts = append(specOpts,
-		customopts.WithOOMScoreAdj(config, c.config.RestrictOOMScoreAdj),
-		customopts.WithPodNamespaces(securityContext, sandboxPid),
-		customopts.WithSupplementalGroups(supplementalGroups),
-		customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeContainer),
-		customopts.WithAnnotation(annotations.SandboxID, sandboxID),
-	)
-
-	return runtimeSpec(id, specOpts...)
-}
-
-// generateVolumeMounts sets up image volumes for container. Rely on the removal of container
 // root directory to do cleanup. Note that image volume will be skipped, if there is criMounts
 // specified with the same destination.
-func (c *criService) generateVolumeMounts(containerRootDir string, criMounts []*runtime.Mount, config *imagespec.ImageConfig) []*runtime.Mount {
+func (c *criService) volumeMounts(containerRootDir string, criMounts []*runtime.Mount, config *imagespec.ImageConfig) []*runtime.Mount {
 	if len(config.Volumes) == 0 {
 		return nil
 	}
@@ -469,61 +283,9 @@ func (c *criService) generateVolumeMounts(containerRootDir string, criMounts []*
 	return mounts
 }
 
-// generateContainerMounts sets up necessary container mounts including /dev/shm, /etc/hosts
-// and /etc/resolv.conf.
-func (c *criService) generateContainerMounts(sandboxID string, config *runtime.ContainerConfig) []*runtime.Mount {
-	var mounts []*runtime.Mount
-	securityContext := config.GetLinux().GetSecurityContext()
-	if !isInCRIMounts(etcHostname, config.GetMounts()) {
-		// /etc/hostname is added since 1.1.6, 1.2.4 and 1.3.
-		// For in-place upgrade, the old sandbox doesn't have the hostname file,
-		// do not mount this in that case.
-		// TODO(random-liu): Remove the check and always mount this when
-		// containerd 1.1 and 1.2 are deprecated.
-		hostpath := c.getSandboxHostname(sandboxID)
-		if _, err := c.os.Stat(hostpath); err == nil {
-			mounts = append(mounts, &runtime.Mount{
-				ContainerPath: etcHostname,
-				HostPath:      hostpath,
-				Readonly:      securityContext.GetReadonlyRootfs(),
-			})
-		}
-	}
-
-	if !isInCRIMounts(etcHosts, config.GetMounts()) {
-		mounts = append(mounts, &runtime.Mount{
-			ContainerPath: etcHosts,
-			HostPath:      c.getSandboxHosts(sandboxID),
-			Readonly:      securityContext.GetReadonlyRootfs(),
-		})
-	}
-
-	// Mount sandbox resolv.config.
-	// TODO: Need to figure out whether we should always mount it as read-only
-	if !isInCRIMounts(resolvConfPath, config.GetMounts()) {
-		mounts = append(mounts, &runtime.Mount{
-			ContainerPath: resolvConfPath,
-			HostPath:      c.getResolvPath(sandboxID),
-			Readonly:      securityContext.GetReadonlyRootfs(),
-		})
-	}
-
-	if !isInCRIMounts(devShm, config.GetMounts()) {
-		sandboxDevShm := c.getSandboxDevShm(sandboxID)
-		if securityContext.GetNamespaceOptions().GetIpc() == runtime.NamespaceMode_NODE {
-			sandboxDevShm = devShm
-		}
-		mounts = append(mounts, &runtime.Mount{
-			ContainerPath: devShm,
-			HostPath:      sandboxDevShm,
-			Readonly:      false,
-		})
-	}
-	return mounts
-}
-
 // runtimeSpec returns a default runtime spec used in cri-containerd.
-func runtimeSpec(id string, opts ...oci.SpecOpts) (*runtimespec.Spec, error) {
+// TODO(windows): Remove nolint after windows starts using this helper.
+func runtimeSpec(id string, opts ...oci.SpecOpts) (*runtimespec.Spec, error) { // nolint: deadcode, unused
 	// GenerateSpec needs namespace.
 	ctx := ctrdutil.NamespacedContext()
 	spec, err := oci.GenerateSpec(ctx, nil, &containers.Container{ID: id}, opts...)
@@ -532,105 +294,3 @@ func runtimeSpec(id string, opts ...oci.SpecOpts) (*runtimespec.Spec, error) {
 	}
 	return spec, nil
 }
-
-// generateSeccompSpecOpts generates containerd SpecOpts for seccomp.
-func generateSeccompSpecOpts(seccompProf string, privileged, seccompEnabled bool) (oci.SpecOpts, error) {
-	if privileged {
-		// Do not set seccomp profile when container is privileged
-		return nil, nil
-	}
-	// Set seccomp profile
-	if seccompProf == runtimeDefault || seccompProf == dockerDefault {
-		// use correct default profile (Eg. if not configured otherwise, the default is docker/default)
-		seccompProf = seccompDefaultProfile
-	}
-	if !seccompEnabled {
-		if seccompProf != "" && seccompProf != unconfinedProfile {
-			return nil, errors.New("seccomp is not supported")
-		}
-		return nil, nil
-	}
-	switch seccompProf {
-	case "", unconfinedProfile:
-		// Do not set seccomp profile.
-		return nil, nil
-	case dockerDefault:
-		// Note: WithDefaultProfile specOpts must be added after capabilities
-		return seccomp.WithDefaultProfile(), nil
-	default:
-		// Require and Trim default profile name prefix
-		if !strings.HasPrefix(seccompProf, profileNamePrefix) {
-			return nil, errors.Errorf("invalid seccomp profile %q", seccompProf)
-		}
-		return seccomp.WithProfile(strings.TrimPrefix(seccompProf, profileNamePrefix)), nil
-	}
-}
-
-// generateApparmorSpecOpts generates containerd SpecOpts for apparmor.
-func generateApparmorSpecOpts(apparmorProf string, privileged, apparmorEnabled bool) (oci.SpecOpts, error) {
-	if !apparmorEnabled {
-		// Should fail loudly if user try to specify apparmor profile
-		// but we don't support it.
-		if apparmorProf != "" && apparmorProf != unconfinedProfile {
-			return nil, errors.New("apparmor is not supported")
-		}
-		return nil, nil
-	}
-	switch apparmorProf {
-	// Based on kubernetes#51746, default apparmor profile should be applied
-	// for when apparmor is not specified.
-	case runtimeDefault, "":
-		if privileged {
-			// Do not set apparmor profile when container is privileged
-			return nil, nil
-		}
-		// TODO (mikebrow): delete created apparmor default profile
-		return apparmor.WithDefaultProfile(appArmorDefaultProfileName), nil
-	case unconfinedProfile:
-		return nil, nil
-	default:
-		// Require and Trim default profile name prefix
-		if !strings.HasPrefix(apparmorProf, profileNamePrefix) {
-			return nil, errors.Errorf("invalid apparmor profile %q", apparmorProf)
-		}
-		return apparmor.WithProfile(strings.TrimPrefix(apparmorProf, profileNamePrefix)), nil
-	}
-}
-
-// generateUserString generates valid user string based on OCI Image Spec
-// v1.0.0.
-//
-// CRI defines that the following combinations are valid:
-//
-// (none) -> ""
-// username -> username
-// username, uid -> username
-// username, uid, gid -> username:gid
-// username, gid -> username:gid
-// uid -> uid
-// uid, gid -> uid:gid
-// gid -> error
-//
-// TODO(random-liu): Add group name support in CRI.
-func generateUserString(username string, uid, gid *runtime.Int64Value) (string, error) {
-	var userstr, groupstr string
-	if uid != nil {
-		userstr = strconv.FormatInt(uid.GetValue(), 10)
-	}
-	if username != "" {
-		userstr = username
-	}
-	if gid != nil {
-		groupstr = strconv.FormatInt(gid.GetValue(), 10)
-	}
-	if userstr == "" {
-		if groupstr != "" {
-			return "", errors.Errorf("user group %q is specified without user", groupstr)
-		}
-		return "", nil
-	}
-	if groupstr != "" {
-		userstr = userstr + ":" + groupstr
-	}
-	return userstr, nil
-}

pkg/server/container_create_unix.go

@@ -0,0 +1,379 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"strconv"
+	"strings"
+
+	"github.com/containerd/containerd/contrib/apparmor"
+	"github.com/containerd/containerd/contrib/seccomp"
+	"github.com/containerd/containerd/oci"
+	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
+	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/pkg/errors"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	"github.com/containerd/cri/pkg/annotations"
+	"github.com/containerd/cri/pkg/config"
+	customopts "github.com/containerd/cri/pkg/containerd/opts"
+)
+
+const (
+	// profileNamePrefix is the prefix for loading profiles on a localhost. Eg. AppArmor localhost/profileName.
+	profileNamePrefix = "localhost/" // TODO (mikebrow): get localhost/ & runtime/default from CRI kubernetes/kubernetes#51747
+	// runtimeDefault indicates that we should use or create a runtime default profile.
+	runtimeDefault = "runtime/default"
+	// dockerDefault indicates that we should use or create a docker default profile.
+	dockerDefault = "docker/default"
+	// appArmorDefaultProfileName is name to use when creating a default apparmor profile.
+	appArmorDefaultProfileName = "cri-containerd.apparmor.d"
+	// unconfinedProfile is a string indicating one should run a pod/containerd without a security profile
+	unconfinedProfile = "unconfined"
+	// seccompDefaultProfile is the default seccomp profile.
+	seccompDefaultProfile = dockerDefault
+)
+
+// containerMounts sets up necessary container system file mounts
+// including /dev/shm, /etc/hosts and /etc/resolv.conf.
+func (c *criService) containerMounts(sandboxID string, config *runtime.ContainerConfig) []*runtime.Mount {
+	var mounts []*runtime.Mount
+	securityContext := config.GetLinux().GetSecurityContext()
+	if !isInCRIMounts(etcHostname, config.GetMounts()) {
+		// /etc/hostname is added since 1.1.6, 1.2.4 and 1.3.
+		// For in-place upgrade, the old sandbox doesn't have the hostname file,
+		// do not mount this in that case.
+		// TODO(random-liu): Remove the check and always mount this when
+		// containerd 1.1 and 1.2 are deprecated.
+		hostpath := c.getSandboxHostname(sandboxID)
+		if _, err := c.os.Stat(hostpath); err == nil {
+			mounts = append(mounts, &runtime.Mount{
+				ContainerPath: etcHostname,
+				HostPath:      hostpath,
+				Readonly:      securityContext.GetReadonlyRootfs(),
+			})
+		}
+	}
+
+	if !isInCRIMounts(etcHosts, config.GetMounts()) {
+		mounts = append(mounts, &runtime.Mount{
+			ContainerPath: etcHosts,
+			HostPath:      c.getSandboxHosts(sandboxID),
+			Readonly:      securityContext.GetReadonlyRootfs(),
+		})
+	}
+
+	// Mount sandbox resolv.config.
+	// TODO: Need to figure out whether we should always mount it as read-only
+	if !isInCRIMounts(resolvConfPath, config.GetMounts()) {
+		mounts = append(mounts, &runtime.Mount{
+			ContainerPath: resolvConfPath,
+			HostPath:      c.getResolvPath(sandboxID),
+			Readonly:      securityContext.GetReadonlyRootfs(),
+		})
+	}
+
+	if !isInCRIMounts(devShm, config.GetMounts()) {
+		sandboxDevShm := c.getSandboxDevShm(sandboxID)
+		if securityContext.GetNamespaceOptions().GetIpc() == runtime.NamespaceMode_NODE {
+			sandboxDevShm = devShm
+		}
+		mounts = append(mounts, &runtime.Mount{
+			ContainerPath: devShm,
+			HostPath:      sandboxDevShm,
+			Readonly:      false,
+		})
+	}
+	return mounts
+}
+
+func (c *criService) containerSpec(id string, sandboxID string, sandboxPid uint32, netNSPath string,
+	config *runtime.ContainerConfig, sandboxConfig *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig,
+	extraMounts []*runtime.Mount, ociRuntime config.Runtime) (*runtimespec.Spec, error) {
+
+	specOpts := []oci.SpecOpts{
+		customopts.WithoutRunMount,
+		customopts.WithoutDefaultSecuritySettings,
+		customopts.WithRelativeRoot(relativeRootfsPath),
+		customopts.WithProcessArgs(config, imageConfig),
+		// this will be set based on the security context below
+		oci.WithNewPrivileges,
+	}
+	if config.GetWorkingDir() != "" {
+		specOpts = append(specOpts, oci.WithProcessCwd(config.GetWorkingDir()))
+	} else if imageConfig.WorkingDir != "" {
+		specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
+	}
+
+	if config.GetTty() {
+		specOpts = append(specOpts, oci.WithTTY)
+	}
+
+	// Add HOSTNAME env.
+	var (
+		err      error
+		hostname = sandboxConfig.GetHostname()
+	)
+	if hostname == "" {
+		if hostname, err = c.os.Hostname(); err != nil {
+			return nil, err
+		}
+	}
+	specOpts = append(specOpts, oci.WithEnv([]string{hostnameEnv + "=" + hostname}))
+
+	// Apply envs from image config first, so that envs from container config
+	// can override them.
+	env := imageConfig.Env
+	for _, e := range config.GetEnvs() {
+		env = append(env, e.GetKey()+"="+e.GetValue())
+	}
+	specOpts = append(specOpts, oci.WithEnv(env))
+
+	securityContext := config.GetLinux().GetSecurityContext()
+	selinuxOpt := securityContext.GetSelinuxOptions()
+	processLabel, mountLabel, err := initSelinuxOpts(selinuxOpt)
+	if err != nil {
+		return nil, errors.Wrapf(err, "failed to init selinux options %+v", securityContext.GetSelinuxOptions())
+	}
+	specOpts = append(specOpts, customopts.WithMounts(c.os, config, extraMounts, mountLabel))
+
+	if !c.config.DisableProcMount {
+		// Apply masked paths if specified.
+		// If the container is privileged, this will be cleared later on.
+		specOpts = append(specOpts, oci.WithMaskedPaths(securityContext.GetMaskedPaths()))
+
+		// Apply readonly paths if specified.
+		// If the container is privileged, this will be cleared later on.
+		specOpts = append(specOpts, oci.WithReadonlyPaths(securityContext.GetReadonlyPaths()))
+	}
+
+	if securityContext.GetPrivileged() {
+		if !sandboxConfig.GetLinux().GetSecurityContext().GetPrivileged() {
+			return nil, errors.New("no privileged container allowed in sandbox")
+		}
+		specOpts = append(specOpts, oci.WithPrivileged)
+		if !ociRuntime.PrivilegedWithoutHostDevices {
+			specOpts = append(specOpts, customopts.WithPrivilegedDevices)
+		}
+	} else { // not privileged
+		specOpts = append(specOpts, customopts.WithDevices(c.os, config), customopts.WithCapabilities(securityContext))
+	}
+
+	// Clear all ambient capabilities. The implication of non-root + caps
+	// is not clearly defined in Kubernetes.
+	// See https://github.com/kubernetes/kubernetes/issues/56374
+	// Keep docker's behavior for now.
+	specOpts = append(specOpts,
+		customopts.WithoutAmbientCaps,
+		customopts.WithSelinuxLabels(processLabel, mountLabel),
+	)
+
+	// TODO: Figure out whether we should set no new privilege for sandbox container by default
+	if securityContext.GetNoNewPrivs() {
+		specOpts = append(specOpts, oci.WithNoNewPrivileges)
+	}
+	// TODO(random-liu): [P1] Set selinux options (privileged or not).
+	if securityContext.GetReadonlyRootfs() {
+		specOpts = append(specOpts, oci.WithRootFSReadonly())
+	}
+
+	if c.config.DisableCgroup {
+		specOpts = append(specOpts, customopts.WithDisabledCgroups)
+	} else {
+		specOpts = append(specOpts, customopts.WithResources(config.GetLinux().GetResources()))
+		if sandboxConfig.GetLinux().GetCgroupParent() != "" {
+			cgroupsPath := getCgroupsPath(sandboxConfig.GetLinux().GetCgroupParent(), id)
+			specOpts = append(specOpts, oci.WithCgroup(cgroupsPath))
+		}
+	}
+
+	supplementalGroups := securityContext.GetSupplementalGroups()
+
+	for pKey, pValue := range getPassthroughAnnotations(sandboxConfig.Annotations,
+		ociRuntime.PodAnnotations) {
+		specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
+	}
+
+	specOpts = append(specOpts,
+		customopts.WithOOMScoreAdj(config, c.config.RestrictOOMScoreAdj),
+		customopts.WithPodNamespaces(securityContext, sandboxPid),
+		customopts.WithSupplementalGroups(supplementalGroups),
+		customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeContainer),
+		customopts.WithAnnotation(annotations.SandboxID, sandboxID),
+	)
+
+	return runtimeSpec(id, specOpts...)
+}
+
+func (c *criService) containerSpecOpts(config *runtime.ContainerConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
+	var specOpts []oci.SpecOpts
+	securityContext := config.GetLinux().GetSecurityContext()
+	// Set container username. This could only be done by containerd, because it needs
+	// access to the container rootfs. Pass user name to containerd, and let it overwrite
+	// the spec for us.
+	userstr, err := generateUserString(
+		securityContext.GetRunAsUsername(),
+		securityContext.GetRunAsUser(),
+		securityContext.GetRunAsGroup())
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to generate user string")
+	}
+	if userstr == "" {
+		// Lastly, since no user override was passed via CRI try to set via OCI
+		// Image
+		userstr = imageConfig.User
+	}
+	if userstr != "" {
+		specOpts = append(specOpts, oci.WithUser(userstr))
+	}
+
+	if securityContext.GetRunAsUsername() != "" {
+		userstr = securityContext.GetRunAsUsername()
+	} else {
+		// Even if RunAsUser is not set, we still call `GetValue` to get uid 0.
+		// Because it is still useful to get additional gids for uid 0.
+		userstr = strconv.FormatInt(securityContext.GetRunAsUser().GetValue(), 10)
+	}
+	specOpts = append(specOpts, customopts.WithAdditionalGIDs(userstr))
+
+	apparmorSpecOpts, err := generateApparmorSpecOpts(
+		securityContext.GetApparmorProfile(),
+		securityContext.GetPrivileged(),
+		c.apparmorEnabled())
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to generate apparmor spec opts")
+	}
+	if apparmorSpecOpts != nil {
+		specOpts = append(specOpts, apparmorSpecOpts)
+	}
+
+	seccompSpecOpts, err := generateSeccompSpecOpts(
+		securityContext.GetSeccompProfilePath(),
+		securityContext.GetPrivileged(),
+		c.seccompEnabled())
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to generate seccomp spec opts")
+	}
+	if seccompSpecOpts != nil {
+		specOpts = append(specOpts, seccompSpecOpts)
+	}
+	return specOpts, nil
+}
+
+// generateSeccompSpecOpts generates containerd SpecOpts for seccomp.
+func generateSeccompSpecOpts(seccompProf string, privileged, seccompEnabled bool) (oci.SpecOpts, error) {
+	if privileged {
+		// Do not set seccomp profile when container is privileged
+		return nil, nil
+	}
+	// Set seccomp profile
+	if seccompProf == runtimeDefault || seccompProf == dockerDefault {
+		// use correct default profile (Eg. if not configured otherwise, the default is docker/default)
+		seccompProf = seccompDefaultProfile
+	}
+	if !seccompEnabled {
+		if seccompProf != "" && seccompProf != unconfinedProfile {
+			return nil, errors.New("seccomp is not supported")
+		}
+		return nil, nil
+	}
+	switch seccompProf {
+	case "", unconfinedProfile:
+		// Do not set seccomp profile.
+		return nil, nil
+	case dockerDefault:
+		// Note: WithDefaultProfile specOpts must be added after capabilities
+		return seccomp.WithDefaultProfile(), nil
+	default:
+		// Require and Trim default profile name prefix
+		if !strings.HasPrefix(seccompProf, profileNamePrefix) {
+			return nil, errors.Errorf("invalid seccomp profile %q", seccompProf)
+		}
+		return seccomp.WithProfile(strings.TrimPrefix(seccompProf, profileNamePrefix)), nil
+	}
+}
+
+// generateApparmorSpecOpts generates containerd SpecOpts for apparmor.
+func generateApparmorSpecOpts(apparmorProf string, privileged, apparmorEnabled bool) (oci.SpecOpts, error) {
+	if !apparmorEnabled {
+		// Should fail loudly if user try to specify apparmor profile
+		// but we don't support it.
+		if apparmorProf != "" && apparmorProf != unconfinedProfile {
+			return nil, errors.New("apparmor is not supported")
+		}
+		return nil, nil
+	}
+	switch apparmorProf {
+	// Based on kubernetes#51746, default apparmor profile should be applied
+	// for when apparmor is not specified.
+	case runtimeDefault, "":
+		if privileged {
+			// Do not set apparmor profile when container is privileged
+			return nil, nil
+		}
+		// TODO (mikebrow): delete created apparmor default profile
+		return apparmor.WithDefaultProfile(appArmorDefaultProfileName), nil
+	case unconfinedProfile:
+		return nil, nil
+	default:
+		// Require and Trim default profile name prefix
+		if !strings.HasPrefix(apparmorProf, profileNamePrefix) {
+			return nil, errors.Errorf("invalid apparmor profile %q", apparmorProf)
+		}
+		return apparmor.WithProfile(strings.TrimPrefix(apparmorProf, profileNamePrefix)), nil
+	}
+}
+
+// generateUserString generates valid user string based on OCI Image Spec
+// v1.0.0.
+//
+// CRI defines that the following combinations are valid:
+//
+// (none) -> ""
+// username -> username
+// username, uid -> username
+// username, uid, gid -> username:gid
+// username, gid -> username:gid
+// uid -> uid
+// uid, gid -> uid:gid
+// gid -> error
+//
+// TODO(random-liu): Add group name support in CRI.
+func generateUserString(username string, uid, gid *runtime.Int64Value) (string, error) {
+	var userstr, groupstr string
+	if uid != nil {
+		userstr = strconv.FormatInt(uid.GetValue(), 10)
+	}
+	if username != "" {
+		userstr = username
+	}
+	if gid != nil {
+		groupstr = strconv.FormatInt(gid.GetValue(), 10)
+	}
+	if userstr == "" {
+		if groupstr != "" {
+			return "", errors.Errorf("user group %q is specified without user", groupstr)
+		}
+		return "", nil
+	}
+	if groupstr != "" {
+		userstr = userstr + ":" + groupstr
+	}
+	return userstr, nil
+}

pkg/server/container_create_windows.go

@@ -0,0 +1,46 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"github.com/containerd/containerd/errdefs"
+	"github.com/containerd/containerd/oci"
+	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
+	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	"github.com/containerd/cri/pkg/config"
+)
+
+// No container mounts for windows.
+func (c *criService) containerMounts(sandboxID string, config *runtime.ContainerConfig) []*runtime.Mount {
+	return nil
+}
+
+// TODO(windows): Add windows container spec.
+func (c *criService) containerSpec(id string, sandboxID string, sandboxPid uint32, netNSPath string,
+	config *runtime.ContainerConfig, sandboxConfig *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig,
+	extraMounts []*runtime.Mount, ociRuntime config.Runtime) (*runtimespec.Spec, error) {
+	return nil, errdefs.ErrNotImplemented
+}
+
+// No extra spec options needed for windows.
+func (c *criService) containerSpecOpts(config *runtime.ContainerConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
+	return nil, nil
+}

pkg/server/container_start.go

@@ -25,7 +25,6 @@ import (
 	containerdio "github.com/containerd/containerd/cio"
 	"github.com/containerd/containerd/errdefs"
 	"github.com/containerd/containerd/log"
-	"github.com/containerd/containerd/plugin"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 	"golang.org/x/net/context"
@@ -99,11 +98,7 @@ func (c *criService) StartContainer(ctx context.Context, r *runtime.StartContain
 		return nil, errors.Wrap(err, "failed to get container info")
 	}
 
-	var taskOpts []containerd.NewTaskOpts
+	taskOpts := c.taskOpts(ctrInfo.Runtime.Name)
-	// TODO(random-liu): Remove this after shim v1 is deprecated.
-	if c.config.NoPivot && ctrInfo.Runtime.Name == plugin.RuntimeLinuxV1 {
-		taskOpts = append(taskOpts, containerd.WithNoPivotRoot)
-	}
 	task, err := container.NewTask(ctx, ioCreation, taskOpts...)
 	if err != nil {
 		return nil, errors.Wrap(err, "failed to create containerd task")

pkg/server/container_stats.go

@@ -25,6 +25,8 @@ import (
 
 // ContainerStats returns stats of the container. If the container does not
 // exist, the call returns an error.
+// TODO(windows): hcsshim Stats is not implemented, add windows support after
+// that is implemented.
 func (c *criService) ContainerStats(ctx context.Context, in *runtime.ContainerStatsRequest) (*runtime.ContainerStatsResponse, error) {
 	cntr, err := c.containerStore.Get(in.GetContainerId())
 	if err != nil {
@@ -39,7 +41,7 @@ func (c *criService) ContainerStats(ctx context.Context, in *runtime.ContainerSt
 		return nil, errors.Errorf("unexpected metrics response: %+v", resp.Metrics)
 	}
 
-	cs, err := c.getContainerMetrics(cntr.Metadata, resp.Metrics[0])
+	cs, err := c.containerMetrics(cntr.Metadata, resp.Metrics[0])
 	if err != nil {
 		return nil, errors.Wrap(err, "failed to decode container metrics")
 	}

pkg/server/container_stats_list.go

@@ -17,10 +17,8 @@ limitations under the License.
 package server
 
 import (
-	"github.com/containerd/cgroups"
 	tasks "github.com/containerd/containerd/api/services/tasks/v1"
 	"github.com/containerd/containerd/api/types"
-	"github.com/containerd/typeurl"
 	"github.com/pkg/errors"
 	"golang.org/x/net/context"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
@@ -58,7 +56,7 @@ func (c *criService) toCRIContainerStats(
 	}
 	containerStats := new(runtime.ListContainerStatsResponse)
 	for _, cntr := range containers {
-		cs, err := c.getContainerMetrics(cntr.Metadata, statsMap[cntr.ID])
+		cs, err := c.containerMetrics(cntr.Metadata, statsMap[cntr.ID])
 		if err != nil {
 			return nil, errors.Wrapf(err, "failed to decode container metrics for %q", cntr.ID)
 		}
@@ -67,59 +65,6 @@ func (c *criService) toCRIContainerStats(
 	return containerStats, nil
 }
 
-func (c *criService) getContainerMetrics(
-	meta containerstore.Metadata,
-	stats *types.Metric,
-) (*runtime.ContainerStats, error) {
-	var cs runtime.ContainerStats
-	var usedBytes, inodesUsed uint64
-	sn, err := c.snapshotStore.Get(meta.ID)
-	// If snapshotstore doesn't have cached snapshot information
-	// set WritableLayer usage to zero
-	if err == nil {
-		usedBytes = sn.Size
-		inodesUsed = sn.Inodes
-	}
-	cs.WritableLayer = &runtime.FilesystemUsage{
-		Timestamp: sn.Timestamp,
-		FsId: &runtime.FilesystemIdentifier{
-			Mountpoint: c.imageFSPath,
-		},
-		UsedBytes:  &runtime.UInt64Value{Value: usedBytes},
-		InodesUsed: &runtime.UInt64Value{Value: inodesUsed},
-	}
-	cs.Attributes = &runtime.ContainerAttributes{
-		Id:          meta.ID,
-		Metadata:    meta.Config.GetMetadata(),
-		Labels:      meta.Config.GetLabels(),
-		Annotations: meta.Config.GetAnnotations(),
-	}
-
-	if stats != nil {
-		s, err := typeurl.UnmarshalAny(stats.Data)
-		if err != nil {
-			return nil, errors.Wrap(err, "failed to extract container metrics")
-		}
-		metrics := s.(*cgroups.Metrics)
-		if metrics.CPU != nil && metrics.CPU.Usage != nil {
-			cs.Cpu = &runtime.CpuUsage{
-				Timestamp:            stats.Timestamp.UnixNano(),
-				UsageCoreNanoSeconds: &runtime.UInt64Value{Value: metrics.CPU.Usage.Total},
-			}
-		}
-		if metrics.Memory != nil && metrics.Memory.Usage != nil {
-			cs.Memory = &runtime.MemoryUsage{
-				Timestamp: stats.Timestamp.UnixNano(),
-				WorkingSetBytes: &runtime.UInt64Value{
-					Value: getWorkingSet(metrics.Memory),
-				},
-			}
-		}
-	}
-
-	return &cs, nil
-}
-
 func (c *criService) normalizeContainerStatsFilter(filter *runtime.ContainerStatsFilter) {
 	if cntr, err := c.containerStore.Get(filter.GetId()); err == nil {
 		filter.Id = cntr.ID
@@ -169,17 +114,3 @@ func matchLabelSelector(selector, labels map[string]string) bool {
 	}
 	return true
 }
-
-// getWorkingSet calculates workingset memory from cgroup memory stats.
-// The caller should make sure memory is not nil.
-// workingset = usage - total_inactive_file
-func getWorkingSet(memory *cgroups.MemoryStat) uint64 {
-	if memory.Usage == nil {
-		return 0
-	}
-	var workingSet uint64
-	if memory.TotalInactiveFile < memory.Usage.Usage {
-		workingSet = memory.Usage.Usage - memory.TotalInactiveFile
-	}
-	return workingSet
-}

pkg/server/container_stats_list_unix.go

@@ -0,0 +1,96 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"github.com/containerd/cgroups"
+	"github.com/containerd/containerd/api/types"
+	"github.com/containerd/typeurl"
+	"github.com/pkg/errors"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	containerstore "github.com/containerd/cri/pkg/store/container"
+)
+
+func (c *criService) containerMetrics(
+	meta containerstore.Metadata,
+	stats *types.Metric,
+) (*runtime.ContainerStats, error) {
+	var cs runtime.ContainerStats
+	var usedBytes, inodesUsed uint64
+	sn, err := c.snapshotStore.Get(meta.ID)
+	// If snapshotstore doesn't have cached snapshot information
+	// set WritableLayer usage to zero
+	if err == nil {
+		usedBytes = sn.Size
+		inodesUsed = sn.Inodes
+	}
+	cs.WritableLayer = &runtime.FilesystemUsage{
+		Timestamp: sn.Timestamp,
+		FsId: &runtime.FilesystemIdentifier{
+			Mountpoint: c.imageFSPath,
+		},
+		UsedBytes:  &runtime.UInt64Value{Value: usedBytes},
+		InodesUsed: &runtime.UInt64Value{Value: inodesUsed},
+	}
+	cs.Attributes = &runtime.ContainerAttributes{
+		Id:          meta.ID,
+		Metadata:    meta.Config.GetMetadata(),
+		Labels:      meta.Config.GetLabels(),
+		Annotations: meta.Config.GetAnnotations(),
+	}
+
+	if stats != nil {
+		s, err := typeurl.UnmarshalAny(stats.Data)
+		if err != nil {
+			return nil, errors.Wrap(err, "failed to extract container metrics")
+		}
+		metrics := s.(*cgroups.Metrics)
+		if metrics.CPU != nil && metrics.CPU.Usage != nil {
+			cs.Cpu = &runtime.CpuUsage{
+				Timestamp:            stats.Timestamp.UnixNano(),
+				UsageCoreNanoSeconds: &runtime.UInt64Value{Value: metrics.CPU.Usage.Total},
+			}
+		}
+		if metrics.Memory != nil && metrics.Memory.Usage != nil {
+			cs.Memory = &runtime.MemoryUsage{
+				Timestamp: stats.Timestamp.UnixNano(),
+				WorkingSetBytes: &runtime.UInt64Value{
+					Value: getWorkingSet(metrics.Memory),
+				},
+			}
+		}
+	}
+
+	return &cs, nil
+}
+
+// getWorkingSet calculates workingset memory from cgroup memory stats.
+// The caller should make sure memory is not nil.
+// workingset = usage - total_inactive_file
+func getWorkingSet(memory *cgroups.MemoryStat) uint64 {
+	if memory.Usage == nil {
+		return 0
+	}
+	var workingSet uint64
+	if memory.TotalInactiveFile < memory.Usage.Usage {
+		workingSet = memory.Usage.Usage - memory.TotalInactiveFile
+	}
+	return workingSet
+}

pkg/server/container_stats_list_unix_test.go

@@ -1,3 +1,5 @@
+// +build !windows
+
 /*
 Copyright 2018 The containerd Authors.
 

pkg/server/container_stats_list_windows.go

@@ -0,0 +1,35 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"github.com/containerd/containerd/api/types"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	containerstore "github.com/containerd/cri/pkg/store/container"
+)
+
+// TODO(windows): Implement a dummy version of this, and actually support this
+// when stats is supported by the hcs containerd shim.
+func (c *criService) containerMetrics(
+	meta containerstore.Metadata,
+	stats *types.Metric,
+) (*runtime.ContainerStats, error) {
+	return nil, nil
+}

pkg/server/container_update_resources_unix.go

@@ -1,3 +1,5 @@
+// +build !windows
+
 /*
 Copyright 2017 The Kubernetes Authors.
 

pkg/server/container_update_resources_unix_test.go

@@ -1,3 +1,5 @@
+// +build !windows
+
 /*
 Copyright 2017 The Kubernetes Authors.
 

pkg/server/container_update_resources_windows.go

@@ -0,0 +1,31 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"github.com/containerd/containerd/errdefs"
+	"golang.org/x/net/context"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+)
+
+// UpdateContainerResources updates ContainerConfig of the container.
+// TODO(windows): Figure out whether windows support this.
+func (c *criService) UpdateContainerResources(ctx context.Context, r *runtime.UpdateContainerResourcesRequest) (*runtime.UpdateContainerResourcesResponse, error) {
+	return nil, errdefs.ErrNotImplemented
+}

pkg/server/helpers.go

@@ -20,7 +20,6 @@ import (
 	"fmt"
 	"path"
 	"path/filepath"
-	"regexp"
 	"strconv"
 	"strings"
 
@@ -33,7 +32,6 @@ import (
 	"github.com/containerd/typeurl"
 	"github.com/docker/distribution/reference"
 	imagedigest "github.com/opencontainers/go-digest"
-	"github.com/opencontainers/selinux/go-selinux/label"
 	"github.com/pkg/errors"
 	"golang.org/x/net/context"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
@@ -51,6 +49,7 @@ const (
 	errorStartReason = "StartError"
 	// errorStartExitCode is the exit code when fails to start container.
 	// 128 is the same with Docker's behavior.
+	// TODO(windows): Figure out what should be used for windows.
 	errorStartExitCode = 128
 	// completeExitReason is the exit reason when container exits with code 0.
 	completeExitReason = "Completed"
@@ -58,39 +57,16 @@ const (
 	errorExitReason = "Error"
 	// oomExitReason is the exit reason when process in container is oom killed.
 	oomExitReason = "OOMKilled"
-)
 
-const (
-	// defaultSandboxOOMAdj is default omm adj for sandbox container. (kubernetes#47938).
-	defaultSandboxOOMAdj = -998
-	// defaultShmSize is the default size of the sandbox shm.
-	defaultShmSize = int64(1024 * 1024 * 64)
-	// relativeRootfsPath is the rootfs path relative to bundle path.
-	relativeRootfsPath = "rootfs"
 	// sandboxesDir contains all sandbox root. A sandbox root is the running
 	// directory of the sandbox, all files created for the sandbox will be
 	// placed under this directory.
 	sandboxesDir = "sandboxes"
 	// containersDir contains all container root.
 	containersDir = "containers"
-	// According to http://man7.org/linux/man-pages/man5/resolv.conf.5.html:
-	// "The search list is currently limited to six domains with a total of 256 characters."
-	maxDNSSearches = 6
 	// Delimiter used to construct container/sandbox names.
 	nameDelimiter = "_"
-	// devShm is the default path of /dev/shm.
-	devShm = "/dev/shm"
-	// etcHosts is the default path of /etc/hosts file.
-	etcHosts = "/etc/hosts"
-	// etcHostname is the default path of /etc/hostname file.
-	etcHostname = "/etc/hostname"
-	// resolvConfPath is the abs path of resolv.conf on host or container.
-	resolvConfPath = "/etc/resolv.conf"
-	// hostnameEnv is the key for HOSTNAME env.
-	hostnameEnv = "HOSTNAME"
-)
 
-const (
 	// criContainerdPrefix is common prefix for cri-containerd
 	criContainerdPrefix = "io.cri-containerd"
 	// containerKindLabel is a label key indicating container is sandbox container or application container
@@ -107,14 +83,9 @@ const (
 	sandboxMetadataExtension = criContainerdPrefix + ".sandbox.metadata"
 	// containerMetadataExtension is an extension name that identify metadata of container in CreateContainerRequest
 	containerMetadataExtension = criContainerdPrefix + ".container.metadata"
-)
 
-const (
 	// defaultIfName is the default network interface for the pods
 	defaultIfName = "eth0"
-	// networkAttachCount is the minimum number of networks the PodSandbox
-	// attaches to
-	networkAttachCount = 2
 )
 
 // makeSandboxName generates sandbox name from sandbox metadata. The name
@@ -141,17 +112,6 @@ func makeContainerName(c *runtime.ContainerMetadata, s *runtime.PodSandboxMetada
 	}, nameDelimiter)
 }
 
-// getCgroupsPath generates container cgroups path.
-func getCgroupsPath(cgroupsParent, id string) string {
-	base := path.Base(cgroupsParent)
-	if strings.HasSuffix(base, ".slice") {
-		// For a.slice/b.slice/c.slice, base is c.slice.
-		// runc systemd cgroup path format is "slice:prefix:name".
-		return strings.Join([]string{base, "cri-containerd", id}, ":")
-	}
-	return filepath.Join(cgroupsParent, id)
-}
-
 // getSandboxRootDir returns the root directory for managing sandbox files,
 // e.g. hosts files.
 func (c *criService) getSandboxRootDir(id string) string {
@@ -176,26 +136,6 @@ func (c *criService) getVolatileContainerRootDir(id string) string {
 	return filepath.Join(c.config.StateDir, containersDir, id)
 }
 
-// getSandboxHostname returns the hostname file path inside the sandbox root directory.
-func (c *criService) getSandboxHostname(id string) string {
-	return filepath.Join(c.getSandboxRootDir(id), "hostname")
-}
-
-// getSandboxHosts returns the hosts file path inside the sandbox root directory.
-func (c *criService) getSandboxHosts(id string) string {
-	return filepath.Join(c.getSandboxRootDir(id), "hosts")
-}
-
-// getResolvPath returns resolv.conf filepath for specified sandbox.
-func (c *criService) getResolvPath(id string) string {
-	return filepath.Join(c.getSandboxRootDir(id), "resolv.conf")
-}
-
-// getSandboxDevShm returns the shm file path inside the sandbox root directory.
-func (c *criService) getSandboxDevShm(id string) string {
-	return filepath.Join(c.getVolatileSandboxRootDir(id), "shm")
-}
-
 // criContainerStateToString formats CRI container state to string.
 func criContainerStateToString(state runtime.ContainerState) string {
 	return runtime.ContainerState_name[int32(state)]
@@ -298,49 +238,6 @@ func (c *criService) ensureImageExists(ctx context.Context, ref string, config *
 	return &newImage, nil
 }
 
-func initSelinuxOpts(selinuxOpt *runtime.SELinuxOption) (string, string, error) {
-	if selinuxOpt == nil {
-		return "", "", nil
-	}
-
-	// Should ignored selinuxOpts if they are incomplete.
-	if selinuxOpt.GetUser() == "" ||
-		selinuxOpt.GetRole() == "" ||
-		selinuxOpt.GetType() == "" {
-		return "", "", nil
-	}
-
-	// make sure the format of "level" is correct.
-	ok, err := checkSelinuxLevel(selinuxOpt.GetLevel())
-	if err != nil || !ok {
-		return "", "", err
-	}
-
-	labelOpts := fmt.Sprintf("%s:%s:%s:%s",
-		selinuxOpt.GetUser(),
-		selinuxOpt.GetRole(),
-		selinuxOpt.GetType(),
-		selinuxOpt.GetLevel())
-
-	options, err := label.DupSecOpt(labelOpts)
-	if err != nil {
-		return "", "", err
-	}
-	return label.InitLabels(options)
-}
-
-func checkSelinuxLevel(level string) (bool, error) {
-	if len(level) == 0 {
-		return true, nil
-	}
-
-	matched, err := regexp.MatchString(`^s\d(-s\d)??(:c\d{1,4}((.c\d{1,4})?,c\d{1,4})*(.c\d{1,4})?(,c\d{1,4}(.c\d{1,4})?)*)?$`, level)
-	if err != nil || !matched {
-		return false, errors.Wrapf(err, "the format of 'level' %q is not correct", level)
-	}
-	return true, nil
-}
-
 // isInCRIMounts checks whether a destination is in CRI mount list.
 func isInCRIMounts(dst string, mounts []*runtime.Mount) bool {
 	for _, m := range mounts {

pkg/server/helpers_test.go

@@ -114,39 +114,6 @@ func TestGetRepoDigestAndTag(t *testing.T) {
 	}
 }
 
-func TestGetCgroupsPath(t *testing.T) {
-	testID := "test-id"
-	for desc, test := range map[string]struct {
-		cgroupsParent string
-		expected      string
-	}{
-		"should support regular cgroup path": {
-			cgroupsParent: "/a/b",
-			expected:      "/a/b/test-id",
-		},
-		"should support systemd cgroup path": {
-			cgroupsParent: "/a.slice/b.slice",
-			expected:      "b.slice:cri-containerd:test-id",
-		},
-		"should support tailing slash for regular cgroup path": {
-			cgroupsParent: "/a/b/",
-			expected:      "/a/b/test-id",
-		},
-		"should support tailing slash for systemd cgroup path": {
-			cgroupsParent: "/a.slice/b.slice/",
-			expected:      "b.slice:cri-containerd:test-id",
-		},
-		"should treat root cgroup as regular cgroup path": {
-			cgroupsParent: "/",
-			expected:      "/test-id",
-		},
-	} {
-		t.Logf("TestCase %q", desc)
-		got := getCgroupsPath(test.cgroupsParent, testID)
-		assert.Equal(t, test.expected, got)
-	}
-}
-
 func TestBuildLabels(t *testing.T) {
 	configLabels := map[string]string{
 		"a": "b",

pkg/server/helpers_unix.go

@@ -0,0 +1,137 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"fmt"
+	"path"
+	"path/filepath"
+	"regexp"
+	"strings"
+
+	runcapparmor "github.com/opencontainers/runc/libcontainer/apparmor"
+	runcseccomp "github.com/opencontainers/runc/libcontainer/seccomp"
+	"github.com/opencontainers/selinux/go-selinux/label"
+	"github.com/pkg/errors"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+)
+
+const (
+	// defaultSandboxOOMAdj is default omm adj for sandbox container. (kubernetes#47938).
+	defaultSandboxOOMAdj = -998
+	// defaultShmSize is the default size of the sandbox shm.
+	defaultShmSize = int64(1024 * 1024 * 64)
+	// relativeRootfsPath is the rootfs path relative to bundle path.
+	relativeRootfsPath = "rootfs"
+	// According to http://man7.org/linux/man-pages/man5/resolv.conf.5.html:
+	// "The search list is currently limited to six domains with a total of 256 characters."
+	maxDNSSearches = 6
+	// devShm is the default path of /dev/shm.
+	devShm = "/dev/shm"
+	// etcHosts is the default path of /etc/hosts file.
+	etcHosts = "/etc/hosts"
+	// etcHostname is the default path of /etc/hostname file.
+	etcHostname = "/etc/hostname"
+	// resolvConfPath is the abs path of resolv.conf on host or container.
+	resolvConfPath = "/etc/resolv.conf"
+	// hostnameEnv is the key for HOSTNAME env.
+	hostnameEnv = "HOSTNAME"
+)
+
+// getCgroupsPath generates container cgroups path.
+func getCgroupsPath(cgroupsParent, id string) string {
+	base := path.Base(cgroupsParent)
+	if strings.HasSuffix(base, ".slice") {
+		// For a.slice/b.slice/c.slice, base is c.slice.
+		// runc systemd cgroup path format is "slice:prefix:name".
+		return strings.Join([]string{base, "cri-containerd", id}, ":")
+	}
+	return filepath.Join(cgroupsParent, id)
+}
+
+// getSandboxHostname returns the hostname file path inside the sandbox root directory.
+func (c *criService) getSandboxHostname(id string) string {
+	return filepath.Join(c.getSandboxRootDir(id), "hostname")
+}
+
+// getSandboxHosts returns the hosts file path inside the sandbox root directory.
+func (c *criService) getSandboxHosts(id string) string {
+	return filepath.Join(c.getSandboxRootDir(id), "hosts")
+}
+
+// getResolvPath returns resolv.conf filepath for specified sandbox.
+func (c *criService) getResolvPath(id string) string {
+	return filepath.Join(c.getSandboxRootDir(id), "resolv.conf")
+}
+
+// getSandboxDevShm returns the shm file path inside the sandbox root directory.
+func (c *criService) getSandboxDevShm(id string) string {
+	return filepath.Join(c.getVolatileSandboxRootDir(id), "shm")
+}
+
+func initSelinuxOpts(selinuxOpt *runtime.SELinuxOption) (string, string, error) {
+	if selinuxOpt == nil {
+		return "", "", nil
+	}
+
+	// Should ignored selinuxOpts if they are incomplete.
+	if selinuxOpt.GetUser() == "" ||
+		selinuxOpt.GetRole() == "" ||
+		selinuxOpt.GetType() == "" {
+		return "", "", nil
+	}
+
+	// make sure the format of "level" is correct.
+	ok, err := checkSelinuxLevel(selinuxOpt.GetLevel())
+	if err != nil || !ok {
+		return "", "", err
+	}
+
+	labelOpts := fmt.Sprintf("%s:%s:%s:%s",
+		selinuxOpt.GetUser(),
+		selinuxOpt.GetRole(),
+		selinuxOpt.GetType(),
+		selinuxOpt.GetLevel())
+
+	options, err := label.DupSecOpt(labelOpts)
+	if err != nil {
+		return "", "", err
+	}
+	return label.InitLabels(options)
+}
+
+func checkSelinuxLevel(level string) (bool, error) {
+	if len(level) == 0 {
+		return true, nil
+	}
+
+	matched, err := regexp.MatchString(`^s\d(-s\d)??(:c\d{1,4}((.c\d{1,4})?,c\d{1,4})*(.c\d{1,4})?(,c\d{1,4}(.c\d{1,4})?)*)?$`, level)
+	if err != nil || !matched {
+		return false, errors.Wrapf(err, "the format of 'level' %q is not correct", level)
+	}
+	return true, nil
+}
+
+func (c *criService) apparmorEnabled() bool {
+	return runcapparmor.IsEnabled() && !c.config.DisableApparmor
+}
+
+func (c *criService) seccompEnabled() bool {
+	return runcseccomp.IsEnabled()
+}

pkg/server/helpers_unix_test.go

@@ -0,0 +1,58 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestGetCgroupsPath(t *testing.T) {
+	testID := "test-id"
+	for desc, test := range map[string]struct {
+		cgroupsParent string
+		expected      string
+	}{
+		"should support regular cgroup path": {
+			cgroupsParent: "/a/b",
+			expected:      "/a/b/test-id",
+		},
+		"should support systemd cgroup path": {
+			cgroupsParent: "/a.slice/b.slice",
+			expected:      "b.slice:cri-containerd:test-id",
+		},
+		"should support tailing slash for regular cgroup path": {
+			cgroupsParent: "/a/b/",
+			expected:      "/a/b/test-id",
+		},
+		"should support tailing slash for systemd cgroup path": {
+			cgroupsParent: "/a.slice/b.slice/",
+			expected:      "b.slice:cri-containerd:test-id",
+		},
+		"should treat root cgroup as regular cgroup path": {
+			cgroupsParent: "/",
+			expected:      "/test-id",
+		},
+	} {
+		t.Logf("TestCase %q", desc)
+		got := getCgroupsPath(test.cgroupsParent, testID)
+		assert.Equal(t, test.expected, got)
+	}
+}

pkg/server/imagefs_info.go

@@ -25,6 +25,7 @@ import (
 )
 
 // ImageFsInfo returns information of the filesystem that is used to store images.
+// TODO(windows): Usage for windows is always 0 right now. Support this for windows.
 func (c *criService) ImageFsInfo(ctx context.Context, r *runtime.ImageFsInfoRequest) (*runtime.ImageFsInfoResponse, error) {
 	snapshots := c.snapshotStore.List()
 	timestamp := time.Now().UnixNano()

pkg/server/io/helpers.go

@@ -24,7 +24,6 @@ import (
 	"syscall"
 
 	"github.com/containerd/containerd/cio"
-	"github.com/containerd/fifo"
 	"golang.org/x/net/context"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
 )
@@ -113,7 +112,7 @@ func newStdioPipes(fifos *cio.FIFOSet) (_ *stdioPipes, _ *wgCloser, err error) {
 	}()
 
 	if fifos.Stdin != "" {
-		if f, err = fifo.OpenFifo(ctx, fifos.Stdin, syscall.O_WRONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
+		if f, err = openFifo(ctx, fifos.Stdin, syscall.O_WRONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
 			return nil, nil, err
 		}
 		p.stdin = f
@@ -121,7 +120,7 @@ func newStdioPipes(fifos *cio.FIFOSet) (_ *stdioPipes, _ *wgCloser, err error) {
 	}
 
 	if fifos.Stdout != "" {
-		if f, err = fifo.OpenFifo(ctx, fifos.Stdout, syscall.O_RDONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
+		if f, err = openFifo(ctx, fifos.Stdout, syscall.O_RDONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
 			return nil, nil, err
 		}
 		p.stdout = f
@@ -129,7 +128,7 @@ func newStdioPipes(fifos *cio.FIFOSet) (_ *stdioPipes, _ *wgCloser, err error) {
 	}
 
 	if fifos.Stderr != "" {
-		if f, err = fifo.OpenFifo(ctx, fifos.Stderr, syscall.O_RDONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
+		if f, err = openFifo(ctx, fifos.Stderr, syscall.O_RDONLY|syscall.O_CREAT|syscall.O_NONBLOCK, 0700); err != nil {
 			return nil, nil, err
 		}
 		p.stderr = f

pkg/server/io/helpers_unix.go

@@ -0,0 +1,31 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package io
+
+import (
+	"io"
+	"os"
+
+	"github.com/containerd/fifo"
+	"golang.org/x/net/context"
+)
+
+func openFifo(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error) {
+	return fifo.OpenFifo(ctx, fn, flag, perm)
+}

pkg/server/io/helpers_windows.go

@@ -0,0 +1,31 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package io
+
+import (
+	"io"
+	"os"
+
+	"golang.org/x/net/context"
+)
+
+// TODO(windows): Add windows FIFO support.
+func openFifo(ctx context.Context, fn string, flag int, perm os.FileMode) (io.ReadWriteCloser, error) {
+	return nil, nil
+}

pkg/server/restart.go

@@ -20,6 +20,7 @@ import (
 	"io/ioutil"
 	"os"
 	"path/filepath"
+	goruntime "runtime"
 	"time"
 
 	"github.com/containerd/containerd"
@@ -407,7 +408,8 @@ func (c *criService) loadSandbox(ctx context.Context, cntr containerd.Container)
 	sandbox.Container = cntr
 
 	// Load network namespace.
-	if meta.Config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
+	if goruntime.GOOS != "windows" &&
+		meta.Config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
 		// Don't need to load netns for host network sandbox.
 		return sandbox, nil
 	}

pkg/server/sandbox_portforward_unix.go

@@ -1,3 +1,5 @@
+// +build !windows
+
 /*
 Copyright 2017 The Kubernetes Authors.
 

pkg/server/sandbox_portforward_windows.go

@@ -0,0 +1,37 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"io"
+
+	"github.com/containerd/containerd/errdefs"
+	"golang.org/x/net/context"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+)
+
+// PortForward prepares a streaming endpoint to forward ports from a PodSandbox, and returns the address.
+// TODO(windows): Implement this for windows.
+func (c *criService) PortForward(ctx context.Context, r *runtime.PortForwardRequest) (*runtime.PortForwardResponse, error) {
+	return nil, errdefs.ErrNotImplemented
+}
+
+func (c *criService) portForward(ctx context.Context, id string, port int32, stream io.ReadWriter) error {
+	return errdefs.ErrNotImplemented
+}

pkg/server/sandbox_run.go

@@ -18,26 +18,20 @@ package server
 
 import (
 	"encoding/json"
-	"fmt"
 	"math"
-	"os"
+	goruntime "runtime"
 	"strings"
 
 	"github.com/containerd/containerd"
 	containerdio "github.com/containerd/containerd/cio"
 	"github.com/containerd/containerd/errdefs"
 	"github.com/containerd/containerd/log"
-	"github.com/containerd/containerd/oci"
-	"github.com/containerd/containerd/plugin"
 	cni "github.com/containerd/go-cni"
 	"github.com/containerd/typeurl"
 	"github.com/davecgh/go-spew/spew"
-	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
-	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 	"golang.org/x/net/context"
-	"golang.org/x/sys/unix"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
 	"k8s.io/kubernetes/pkg/util/bandwidth"
 
@@ -110,10 +104,14 @@ func (c *criService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandbox
 	}
 	log.G(ctx).Debugf("Use OCI %+v for sandbox %q", ociRuntime, id)
 
-	securityContext := config.GetLinux().GetSecurityContext()
+	podNetwork := true
-	//Create Network Namespace if it is not in host network
+	// Pod network is always needed on windows.
-	hostNet := securityContext.GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE
+	if goruntime.GOOS != "windows" &&
-	if !hostNet {
+		config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
+		// Pod network is not needed on linux with host network.
+		podNetwork = false
+	}
+	if podNetwork {
 		// If it is not in host network namespace then create a namespace and set the sandbox
 		// handle. NetNSPath in sandbox metadata and NetNS is non empty only for non host network
 		// namespaces. If the pod is in host network namespace then both are empty and should not
@@ -154,39 +152,19 @@ func (c *criService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandbox
 	}
 
 	// Create sandbox container.
-	spec, err := c.generateSandboxContainerSpec(id, config, &image.ImageSpec.Config, sandbox.NetNSPath, ociRuntime.PodAnnotations)
+	// NOTE: sandboxContainerSpec SHOULD NOT have side
+	// effect, e.g. accessing/creating files, so that we can test
+	// it safely.
+	spec, err := c.sandboxContainerSpec(id, config, &image.ImageSpec.Config, sandbox.NetNSPath, ociRuntime.PodAnnotations)
 	if err != nil {
 		return nil, errors.Wrap(err, "failed to generate sandbox container spec")
 	}
 	log.G(ctx).Debugf("Sandbox container %q spec: %#+v", id, spew.NewFormatter(spec))
 
-	var specOpts []oci.SpecOpts
+	// Generate spec options that will be applied to the spec later.
-	userstr, err := generateUserString(
+	specOpts, err := c.sandboxContainerSpecOpts(config, &image.ImageSpec.Config)
-		"",
-		securityContext.GetRunAsUser(),
-		securityContext.GetRunAsGroup(),
-	)
 	if err != nil {
-		return nil, errors.Wrap(err, "failed to generate user string")
+		return nil, errors.Wrap(err, "failed to generate sanbdox container spec options")
-	}
-	if userstr == "" {
-		// Lastly, since no user override was passed via CRI try to set via OCI
-		// Image
-		userstr = image.ImageSpec.Config.User
-	}
-	if userstr != "" {
-		specOpts = append(specOpts, oci.WithUser(userstr))
-	}
-
-	seccompSpecOpts, err := generateSeccompSpecOpts(
-		securityContext.GetSeccompProfilePath(),
-		securityContext.GetPrivileged(),
-		c.seccompEnabled)
-	if err != nil {
-		return nil, errors.Wrap(err, "failed to generate seccomp spec opts")
-	}
-	if seccompSpecOpts != nil {
-		specOpts = append(specOpts, seccompSpecOpts)
 	}
 
 	sandboxLabels := buildLabels(config.Labels, containerKindSandbox)
@@ -247,14 +225,14 @@ func (c *criService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandbox
 		}
 	}()
 
-	// Setup sandbox /dev/shm, /etc/hosts, /etc/resolv.conf and /etc/hostname.
+	// Setup files required for the sandbox.
 	if err = c.setupSandboxFiles(id, config); err != nil {
 		return nil, errors.Wrapf(err, "failed to setup sandbox files")
 	}
 	defer func() {
 		if retErr != nil {
-			if err = c.unmountSandboxFiles(id, config); err != nil {
+			if err = c.cleanupSandboxFiles(id, config); err != nil {
-				log.G(ctx).WithError(err).Errorf("Failed to unmount sandbox files in %q",
+				log.G(ctx).WithError(err).Errorf("Failed to cleanup sandbox files in %q",
 					sandboxRootDir)
 			}
 		}
@@ -270,11 +248,7 @@ func (c *criService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandbox
 	log.G(ctx).Tracef("Create sandbox container (id=%q, name=%q).",
 		id, name)
 
-	var taskOpts []containerd.NewTaskOpts
+	taskOpts := c.taskOpts(ociRuntime.Type)
-	// TODO(random-liu): Remove this after shim v1 is deprecated.
-	if c.config.NoPivot && ociRuntime.Type == plugin.RuntimeRuncV1 {
-		taskOpts = append(taskOpts, containerd.WithNoPivotRoot)
-	}
 	// We don't need stdio for sandbox container.
 	task, err := container.NewTask(ctx, containerdio.NullIO, taskOpts...)
 	if err != nil {
@@ -328,222 +302,6 @@ func (c *criService) RunPodSandbox(ctx context.Context, r *runtime.RunPodSandbox
 	return &runtime.RunPodSandboxResponse{PodSandboxId: id}, nil
 }
 
-func (c *criService) generateSandboxContainerSpec(id string, config *runtime.PodSandboxConfig,
-	imageConfig *imagespec.ImageConfig, nsPath string, runtimePodAnnotations []string) (*runtimespec.Spec, error) {
-	// Creates a spec Generator with the default spec.
-	// TODO(random-liu): [P1] Compare the default settings with docker and containerd default.
-	specOpts := []oci.SpecOpts{
-		customopts.WithoutRunMount,
-		customopts.WithoutDefaultSecuritySettings,
-		customopts.WithRelativeRoot(relativeRootfsPath),
-		oci.WithEnv(imageConfig.Env),
-		oci.WithRootFSReadonly(),
-		oci.WithHostname(config.GetHostname()),
-	}
-	if imageConfig.WorkingDir != "" {
-		specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
-	}
-
-	if len(imageConfig.Entrypoint) == 0 && len(imageConfig.Cmd) == 0 {
-		// Pause image must have entrypoint or cmd.
-		return nil, errors.Errorf("invalid empty entrypoint and cmd in image config %+v", imageConfig)
-	}
-	specOpts = append(specOpts, oci.WithProcessArgs(append(imageConfig.Entrypoint, imageConfig.Cmd...)...))
-
-	// TODO(random-liu): [P2] Consider whether to add labels and annotations to the container.
-
-	// Set cgroups parent.
-	if c.config.DisableCgroup {
-		specOpts = append(specOpts, customopts.WithDisabledCgroups)
-	} else {
-		if config.GetLinux().GetCgroupParent() != "" {
-			cgroupsPath := getCgroupsPath(config.GetLinux().GetCgroupParent(), id)
-			specOpts = append(specOpts, oci.WithCgroup(cgroupsPath))
-		}
-	}
-
-	// When cgroup parent is not set, containerd-shim will create container in a child cgroup
-	// of the cgroup itself is in.
-	// TODO(random-liu): [P2] Set default cgroup path if cgroup parent is not specified.
-
-	// Set namespace options.
-	var (
-		securityContext = config.GetLinux().GetSecurityContext()
-		nsOptions       = securityContext.GetNamespaceOptions()
-	)
-	if nsOptions.GetNetwork() == runtime.NamespaceMode_NODE {
-		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.NetworkNamespace))
-		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.UTSNamespace))
-	} else {
-		//TODO(Abhi): May be move this to containerd spec opts (WithLinuxSpaceOption)
-		specOpts = append(specOpts, oci.WithLinuxNamespace(
-			runtimespec.LinuxNamespace{
-				Type: runtimespec.NetworkNamespace,
-				Path: nsPath,
-			}))
-	}
-	if nsOptions.GetPid() == runtime.NamespaceMode_NODE {
-		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.PIDNamespace))
-	}
-	if nsOptions.GetIpc() == runtime.NamespaceMode_NODE {
-		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.IPCNamespace))
-	}
-
-	// It's fine to generate the spec before the sandbox /dev/shm
-	// is actually created.
-	sandboxDevShm := c.getSandboxDevShm(id)
-	if nsOptions.GetIpc() == runtime.NamespaceMode_NODE {
-		sandboxDevShm = devShm
-	}
-	specOpts = append(specOpts, oci.WithMounts([]runtimespec.Mount{
-		{
-			Source:      sandboxDevShm,
-			Destination: devShm,
-			Type:        "bind",
-			Options:     []string{"rbind", "ro"},
-		},
-	}))
-
-	selinuxOpt := securityContext.GetSelinuxOptions()
-	processLabel, mountLabel, err := initSelinuxOpts(selinuxOpt)
-	if err != nil {
-		return nil, errors.Wrapf(err, "failed to init selinux options %+v", securityContext.GetSelinuxOptions())
-	}
-
-	supplementalGroups := securityContext.GetSupplementalGroups()
-	specOpts = append(specOpts,
-		customopts.WithSelinuxLabels(processLabel, mountLabel),
-		customopts.WithSupplementalGroups(supplementalGroups),
-	)
-
-	// Add sysctls
-	sysctls := config.GetLinux().GetSysctls()
-	specOpts = append(specOpts, customopts.WithSysctls(sysctls))
-
-	// Note: LinuxSandboxSecurityContext does not currently provide an apparmor profile
-
-	if !c.config.DisableCgroup {
-		specOpts = append(specOpts, customopts.WithDefaultSandboxShares)
-	}
-	specOpts = append(specOpts, customopts.WithPodOOMScoreAdj(int(defaultSandboxOOMAdj), c.config.RestrictOOMScoreAdj))
-
-	for pKey, pValue := range getPassthroughAnnotations(config.Annotations,
-		runtimePodAnnotations) {
-		specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
-	}
-
-	specOpts = append(specOpts,
-		customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeSandbox),
-		customopts.WithAnnotation(annotations.SandboxID, id),
-		customopts.WithAnnotation(annotations.SandboxLogDir, config.GetLogDirectory()),
-	)
-
-	return runtimeSpec(id, specOpts...)
-}
-
-// setupSandboxFiles sets up necessary sandbox files including /dev/shm, /etc/hosts,
-// /etc/resolv.conf and /etc/hostname.
-func (c *criService) setupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
-	sandboxEtcHostname := c.getSandboxHostname(id)
-	hostname := config.GetHostname()
-	if hostname == "" {
-		var err error
-		hostname, err = c.os.Hostname()
-		if err != nil {
-			return errors.Wrap(err, "failed to get hostname")
-		}
-	}
-	if err := c.os.WriteFile(sandboxEtcHostname, []byte(hostname+"\n"), 0644); err != nil {
-		return errors.Wrapf(err, "failed to write hostname to %q", sandboxEtcHostname)
-	}
-
-	// TODO(random-liu): Consider whether we should maintain /etc/hosts and /etc/resolv.conf in kubelet.
-	sandboxEtcHosts := c.getSandboxHosts(id)
-	if err := c.os.CopyFile(etcHosts, sandboxEtcHosts, 0644); err != nil {
-		return errors.Wrapf(err, "failed to generate sandbox hosts file %q", sandboxEtcHosts)
-	}
-
-	// Set DNS options. Maintain a resolv.conf for the sandbox.
-	var err error
-	resolvContent := ""
-	if dnsConfig := config.GetDnsConfig(); dnsConfig != nil {
-		resolvContent, err = parseDNSOptions(dnsConfig.Servers, dnsConfig.Searches, dnsConfig.Options)
-		if err != nil {
-			return errors.Wrapf(err, "failed to parse sandbox DNSConfig %+v", dnsConfig)
-		}
-	}
-	resolvPath := c.getResolvPath(id)
-	if resolvContent == "" {
-		// copy host's resolv.conf to resolvPath
-		err = c.os.CopyFile(resolvConfPath, resolvPath, 0644)
-		if err != nil {
-			return errors.Wrapf(err, "failed to copy host's resolv.conf to %q", resolvPath)
-		}
-	} else {
-		err = c.os.WriteFile(resolvPath, []byte(resolvContent), 0644)
-		if err != nil {
-			return errors.Wrapf(err, "failed to write resolv content to %q", resolvPath)
-		}
-	}
-
-	// Setup sandbox /dev/shm.
-	if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetIpc() == runtime.NamespaceMode_NODE {
-		if _, err := c.os.Stat(devShm); err != nil {
-			return errors.Wrapf(err, "host %q is not available for host ipc", devShm)
-		}
-	} else {
-		sandboxDevShm := c.getSandboxDevShm(id)
-		if err := c.os.MkdirAll(sandboxDevShm, 0700); err != nil {
-			return errors.Wrap(err, "failed to create sandbox shm")
-		}
-		shmproperty := fmt.Sprintf("mode=1777,size=%d", defaultShmSize)
-		if err := c.os.Mount("shm", sandboxDevShm, "tmpfs", uintptr(unix.MS_NOEXEC|unix.MS_NOSUID|unix.MS_NODEV), shmproperty); err != nil {
-			return errors.Wrap(err, "failed to mount sandbox shm")
-		}
-	}
-
-	return nil
-}
-
-// parseDNSOptions parse DNS options into resolv.conf format content,
-// if none option is specified, will return empty with no error.
-func parseDNSOptions(servers, searches, options []string) (string, error) {
-	resolvContent := ""
-
-	if len(searches) > maxDNSSearches {
-		return "", errors.Errorf("DNSOption.Searches has more than %d domains", maxDNSSearches)
-	}
-
-	if len(searches) > 0 {
-		resolvContent += fmt.Sprintf("search %s\n", strings.Join(searches, " "))
-	}
-
-	if len(servers) > 0 {
-		resolvContent += fmt.Sprintf("nameserver %s\n", strings.Join(servers, "\nnameserver "))
-	}
-
-	if len(options) > 0 {
-		resolvContent += fmt.Sprintf("options %s\n", strings.Join(options, " "))
-	}
-
-	return resolvContent, nil
-}
-
-// unmountSandboxFiles unmount some sandbox files, we rely on the removal of sandbox root directory to
-// remove these files. Unmount should *NOT* return error if the mount point is already unmounted.
-func (c *criService) unmountSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
-	if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetIpc() != runtime.NamespaceMode_NODE {
-		path, err := c.os.FollowSymlinkInScope(c.getSandboxDevShm(id), "/")
-		if err != nil {
-			return errors.Wrap(err, "failed to follow symlink")
-		}
-		if err := c.os.Unmount(path); err != nil && !os.IsNotExist(err) {
-			return errors.Wrapf(err, "failed to unmount %q", path)
-		}
-	}
-	return nil
-}
-
 // setupPod setups up the network for a pod
 func (c *criService) setupPod(ctx context.Context, id string, path string, config *runtime.PodSandboxConfig) (string, *cni.CNIResult, error) {
 	if c.netPlugin == nil {
@@ -713,6 +471,7 @@ func (c *criService) getSandboxRuntime(config *runtime.PodSandboxConfig, runtime
 		//  Note: If the workload is marked untrusted but requests privileged, this can be granted, as the
 		// runtime may support this.  For example, in a virtual-machine isolated runtime, privileged
 		// is a supported option, granting the workload to access the entire guest VM instead of host.
+		// TODO(windows): Deprecate this so that we don't need to handle it for windows.
 		if hostAccessingSandbox(config) {
 			return criconfig.Runtime{}, errors.New("untrusted workload with host access is not allowed")
 		}

pkg/server/sandbox_run_test.go

@@ -18,449 +18,16 @@ package server
 
 import (
 	"net"
-	"os"
-	"path/filepath"
 	"testing"
 
 	cni "github.com/containerd/go-cni"
-	"github.com/containerd/typeurl"
-	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
-	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
 
 	"github.com/containerd/cri/pkg/annotations"
 	criconfig "github.com/containerd/cri/pkg/config"
-	"github.com/containerd/cri/pkg/containerd/opts"
-	ostesting "github.com/containerd/cri/pkg/os/testing"
-	sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
 )
 
-func getRunPodSandboxTestData() (*runtime.PodSandboxConfig, *imagespec.ImageConfig, func(*testing.T, string, *runtimespec.Spec)) {
-	config := &runtime.PodSandboxConfig{
-		Metadata: &runtime.PodSandboxMetadata{
-			Name:      "test-name",
-			Uid:       "test-uid",
-			Namespace: "test-ns",
-			Attempt:   1,
-		},
-		Hostname:     "test-hostname",
-		LogDirectory: "test-log-directory",
-		Labels:       map[string]string{"a": "b"},
-		Annotations:  map[string]string{"c": "d"},
-		Linux: &runtime.LinuxPodSandboxConfig{
-			CgroupParent: "/test/cgroup/parent",
-		},
-	}
-	imageConfig := &imagespec.ImageConfig{
-		Env:        []string{"a=b", "c=d"},
-		Entrypoint: []string{"/pause"},
-		Cmd:        []string{"forever"},
-		WorkingDir: "/workspace",
-	}
-	specCheck := func(t *testing.T, id string, spec *runtimespec.Spec) {
-		assert.Equal(t, "test-hostname", spec.Hostname)
-		assert.Equal(t, getCgroupsPath("/test/cgroup/parent", id), spec.Linux.CgroupsPath)
-		assert.Equal(t, relativeRootfsPath, spec.Root.Path)
-		assert.Equal(t, true, spec.Root.Readonly)
-		assert.Contains(t, spec.Process.Env, "a=b", "c=d")
-		assert.Equal(t, []string{"/pause", "forever"}, spec.Process.Args)
-		assert.Equal(t, "/workspace", spec.Process.Cwd)
-		assert.EqualValues(t, *spec.Linux.Resources.CPU.Shares, opts.DefaultSandboxCPUshares)
-		assert.EqualValues(t, *spec.Process.OOMScoreAdj, defaultSandboxOOMAdj)
-
-		t.Logf("Check PodSandbox annotations")
-		assert.Contains(t, spec.Annotations, annotations.SandboxID)
-		assert.EqualValues(t, spec.Annotations[annotations.SandboxID], id)
-
-		assert.Contains(t, spec.Annotations, annotations.ContainerType)
-		assert.EqualValues(t, spec.Annotations[annotations.ContainerType], annotations.ContainerTypeSandbox)
-
-		assert.Contains(t, spec.Annotations, annotations.SandboxLogDir)
-		assert.EqualValues(t, spec.Annotations[annotations.SandboxLogDir], "test-log-directory")
-	}
-	return config, imageConfig, specCheck
-}
-
-func TestGenerateSandboxContainerSpec(t *testing.T) {
-	testID := "test-id"
-	nsPath := "test-cni"
-	for desc, test := range map[string]struct {
-		configChange      func(*runtime.PodSandboxConfig)
-		podAnnotations    []string
-		imageConfigChange func(*imagespec.ImageConfig)
-		specCheck         func(*testing.T, *runtimespec.Spec)
-		expectErr         bool
-	}{
-		"spec should reflect original config": {
-			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
-				// runtime spec should have expected namespaces enabled by default.
-				require.NotNil(t, spec.Linux)
-				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.NetworkNamespace,
-					Path: nsPath,
-				})
-				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.UTSNamespace,
-				})
-				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.PIDNamespace,
-				})
-				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.IPCNamespace,
-				})
-			},
-		},
-		"host namespace": {
-			configChange: func(c *runtime.PodSandboxConfig) {
-				c.Linux.SecurityContext = &runtime.LinuxSandboxSecurityContext{
-					NamespaceOptions: &runtime.NamespaceOption{
-						Network: runtime.NamespaceMode_NODE,
-						Pid:     runtime.NamespaceMode_NODE,
-						Ipc:     runtime.NamespaceMode_NODE,
-					},
-				}
-			},
-			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
-				// runtime spec should disable expected namespaces in host mode.
-				require.NotNil(t, spec.Linux)
-				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.NetworkNamespace,
-				})
-				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.UTSNamespace,
-				})
-				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.PIDNamespace,
-				})
-				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
-					Type: runtimespec.IPCNamespace,
-				})
-			},
-		},
-		"should return error when entrypoint and cmd are empty": {
-			imageConfigChange: func(c *imagespec.ImageConfig) {
-				c.Entrypoint = nil
-				c.Cmd = nil
-			},
-			expectErr: true,
-		},
-		"should set supplemental groups correctly": {
-			configChange: func(c *runtime.PodSandboxConfig) {
-				c.Linux.SecurityContext = &runtime.LinuxSandboxSecurityContext{
-					SupplementalGroups: []int64{1111, 2222},
-				}
-			},
-			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
-				require.NotNil(t, spec.Process)
-				assert.Contains(t, spec.Process.User.AdditionalGids, uint32(1111))
-				assert.Contains(t, spec.Process.User.AdditionalGids, uint32(2222))
-			},
-		},
-		"a passthrough annotation should be passed as an OCI annotation": {
-			podAnnotations: []string{"c"},
-			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
-				assert.Equal(t, spec.Annotations["c"], "d")
-			},
-		},
-		"a non-passthrough annotation should not be passed as an OCI annotation": {
-			configChange: func(c *runtime.PodSandboxConfig) {
-				c.Annotations["d"] = "e"
-			},
-			podAnnotations: []string{"c"},
-			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
-				assert.Equal(t, spec.Annotations["c"], "d")
-				_, ok := spec.Annotations["d"]
-				assert.False(t, ok)
-			},
-		},
-		"passthrough annotations should support wildcard match": {
-			configChange: func(c *runtime.PodSandboxConfig) {
-				c.Annotations["t.f"] = "j"
-				c.Annotations["z.g"] = "o"
-				c.Annotations["z"] = "o"
-				c.Annotations["y.ca"] = "b"
-				c.Annotations["y"] = "b"
-			},
-			podAnnotations: []string{"t*", "z.*", "y.c*"},
-			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
-				assert.Equal(t, spec.Annotations["t.f"], "j")
-				assert.Equal(t, spec.Annotations["z.g"], "o")
-				assert.Equal(t, spec.Annotations["y.ca"], "b")
-				_, ok := spec.Annotations["y"]
-				assert.False(t, ok)
-				_, ok = spec.Annotations["z"]
-				assert.False(t, ok)
-			},
-		},
-	} {
-		t.Logf("TestCase %q", desc)
-		c := newTestCRIService()
-		config, imageConfig, specCheck := getRunPodSandboxTestData()
-		if test.configChange != nil {
-			test.configChange(config)
-		}
-
-		if test.imageConfigChange != nil {
-			test.imageConfigChange(imageConfig)
-		}
-		spec, err := c.generateSandboxContainerSpec(testID, config, imageConfig, nsPath,
-			test.podAnnotations)
-		if test.expectErr {
-			assert.Error(t, err)
-			assert.Nil(t, spec)
-			continue
-		}
-		assert.NoError(t, err)
-		assert.NotNil(t, spec)
-		specCheck(t, testID, spec)
-		if test.specCheck != nil {
-			test.specCheck(t, spec)
-		}
-	}
-}
-
-func TestSetupSandboxFiles(t *testing.T) {
-	const (
-		testID       = "test-id"
-		realhostname = "test-real-hostname"
-	)
-	for desc, test := range map[string]struct {
-		dnsConfig     *runtime.DNSConfig
-		hostname      string
-		ipcMode       runtime.NamespaceMode
-		expectedCalls []ostesting.CalledDetail
-	}{
-		"should check host /dev/shm existence when ipc mode is NODE": {
-			ipcMode: runtime.NamespaceMode_NODE,
-			expectedCalls: []ostesting.CalledDetail{
-				{
-					Name: "Hostname",
-				},
-				{
-					Name: "WriteFile",
-					Arguments: []interface{}{
-						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
-						[]byte(realhostname + "\n"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/hosts",
-						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/resolv.conf",
-						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name:      "Stat",
-					Arguments: []interface{}{"/dev/shm"},
-				},
-			},
-		},
-		"should create new /etc/resolv.conf if DNSOptions is set": {
-			dnsConfig: &runtime.DNSConfig{
-				Servers:  []string{"8.8.8.8"},
-				Searches: []string{"114.114.114.114"},
-				Options:  []string{"timeout:1"},
-			},
-			ipcMode: runtime.NamespaceMode_NODE,
-			expectedCalls: []ostesting.CalledDetail{
-				{
-					Name: "Hostname",
-				},
-				{
-					Name: "WriteFile",
-					Arguments: []interface{}{
-						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
-						[]byte(realhostname + "\n"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/hosts",
-						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "WriteFile",
-					Arguments: []interface{}{
-						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
-						[]byte(`search 114.114.114.114
-nameserver 8.8.8.8
-options timeout:1
-`), os.FileMode(0644),
-					},
-				},
-				{
-					Name:      "Stat",
-					Arguments: []interface{}{"/dev/shm"},
-				},
-			},
-		},
-		"should create sandbox shm when ipc namespace mode is not NODE": {
-			ipcMode: runtime.NamespaceMode_POD,
-			expectedCalls: []ostesting.CalledDetail{
-				{
-					Name: "Hostname",
-				},
-				{
-					Name: "WriteFile",
-					Arguments: []interface{}{
-						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
-						[]byte(realhostname + "\n"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/hosts",
-						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/resolv.conf",
-						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "MkdirAll",
-					Arguments: []interface{}{
-						filepath.Join(testStateDir, sandboxesDir, testID, "shm"),
-						os.FileMode(0700),
-					},
-				},
-				{
-					Name: "Mount",
-					// Ignore arguments which are too complex to check.
-				},
-			},
-		},
-		"should create /etc/hostname when hostname is set": {
-			hostname: "test-hostname",
-			ipcMode:  runtime.NamespaceMode_NODE,
-			expectedCalls: []ostesting.CalledDetail{
-				{
-					Name: "WriteFile",
-					Arguments: []interface{}{
-						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
-						[]byte("test-hostname\n"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/hosts",
-						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name: "CopyFile",
-					Arguments: []interface{}{
-						"/etc/resolv.conf",
-						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
-						os.FileMode(0644),
-					},
-				},
-				{
-					Name:      "Stat",
-					Arguments: []interface{}{"/dev/shm"},
-				},
-			},
-		},
-	} {
-		t.Logf("TestCase %q", desc)
-		c := newTestCRIService()
-		c.os.(*ostesting.FakeOS).HostnameFn = func() (string, error) {
-			return realhostname, nil
-		}
-		cfg := &runtime.PodSandboxConfig{
-			Hostname:  test.hostname,
-			DnsConfig: test.dnsConfig,
-			Linux: &runtime.LinuxPodSandboxConfig{
-				SecurityContext: &runtime.LinuxSandboxSecurityContext{
-					NamespaceOptions: &runtime.NamespaceOption{
-						Ipc: test.ipcMode,
-					},
-				},
-			},
-		}
-		c.setupSandboxFiles(testID, cfg)
-		calls := c.os.(*ostesting.FakeOS).GetCalls()
-		assert.Len(t, calls, len(test.expectedCalls))
-		for i, expected := range test.expectedCalls {
-			if expected.Arguments == nil {
-				// Ignore arguments.
-				expected.Arguments = calls[i].Arguments
-			}
-			assert.Equal(t, expected, calls[i])
-		}
-	}
-}
-
-func TestParseDNSOption(t *testing.T) {
-	for desc, test := range map[string]struct {
-		servers         []string
-		searches        []string
-		options         []string
-		expectedContent string
-		expectErr       bool
-	}{
-		"empty dns options should return empty content": {},
-		"non-empty dns options should return correct content": {
-			servers:  []string{"8.8.8.8", "server.google.com"},
-			searches: []string{"114.114.114.114"},
-			options:  []string{"timeout:1"},
-			expectedContent: `search 114.114.114.114
-nameserver 8.8.8.8
-nameserver server.google.com
-options timeout:1
-`,
-		},
-		"should return error if dns search exceeds limit(6)": {
-			searches: []string{
-				"server0.google.com",
-				"server1.google.com",
-				"server2.google.com",
-				"server3.google.com",
-				"server4.google.com",
-				"server5.google.com",
-				"server6.google.com",
-			},
-			expectErr: true,
-		},
-	} {
-		t.Logf("TestCase %q", desc)
-		resolvContent, err := parseDNSOptions(test.servers, test.searches, test.options)
-		if test.expectErr {
-			assert.Error(t, err)
-			continue
-		}
-		assert.NoError(t, err)
-		assert.Equal(t, resolvContent, test.expectedContent)
-	}
-}
-
 func TestToCNIPortMappings(t *testing.T) {
 	for desc, test := range map[string]struct {
 		criPortMappings []*runtime.PortMapping
@@ -575,46 +142,6 @@ func TestSelectPodIP(t *testing.T) {
 	}
 }
 
-func TestTypeurlMarshalUnmarshalSandboxMeta(t *testing.T) {
-	for desc, test := range map[string]struct {
-		configChange func(*runtime.PodSandboxConfig)
-	}{
-		"should marshal original config": {},
-		"should marshal Linux": {
-			configChange: func(c *runtime.PodSandboxConfig) {
-				c.Linux.SecurityContext = &runtime.LinuxSandboxSecurityContext{
-					NamespaceOptions: &runtime.NamespaceOption{
-						Network: runtime.NamespaceMode_NODE,
-						Pid:     runtime.NamespaceMode_NODE,
-						Ipc:     runtime.NamespaceMode_NODE,
-					},
-					SupplementalGroups: []int64{1111, 2222},
-				}
-			},
-		},
-	} {
-		t.Logf("TestCase %q", desc)
-		meta := &sandboxstore.Metadata{
-			ID:        "1",
-			Name:      "sandbox_1",
-			NetNSPath: "/home/cloud",
-		}
-		meta.Config, _, _ = getRunPodSandboxTestData()
-		if test.configChange != nil {
-			test.configChange(meta.Config)
-		}
-
-		any, err := typeurl.MarshalAny(meta)
-		assert.NoError(t, err)
-		data, err := typeurl.UnmarshalAny(any)
-		assert.NoError(t, err)
-		assert.IsType(t, &sandboxstore.Metadata{}, data)
-		curMeta, ok := data.(*sandboxstore.Metadata)
-		assert.True(t, ok)
-		assert.Equal(t, meta, curMeta)
-	}
-}
-
 func TestHostAccessingSandbox(t *testing.T) {
 	privilegedContext := &runtime.PodSandboxConfig{
 		Linux: &runtime.LinuxPodSandboxConfig{
@@ -823,21 +350,3 @@ func TestGetSandboxRuntime(t *testing.T) {
 		})
 	}
 }
-
-func TestSandboxDisableCgroup(t *testing.T) {
-	config, imageConfig, _ := getRunPodSandboxTestData()
-	c := newTestCRIService()
-	c.config.DisableCgroup = true
-	spec, err := c.generateSandboxContainerSpec("test-id", config, imageConfig, "test-cni", []string{})
-	require.NoError(t, err)
-
-	t.Log("resource limit should not be set")
-	assert.Nil(t, spec.Linux.Resources.Memory)
-	assert.Nil(t, spec.Linux.Resources.CPU)
-
-	t.Log("cgroup path should be empty")
-	assert.Empty(t, spec.Linux.CgroupsPath)
-}
-
-// TODO(random-liu): [P1] Add unit test for different error cases to make sure
-// the function cleans up on error properly.

pkg/server/sandbox_run_unix.go

@@ -0,0 +1,298 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/containerd/containerd"
+	"github.com/containerd/containerd/oci"
+	"github.com/containerd/containerd/plugin"
+	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
+	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/pkg/errors"
+	"golang.org/x/sys/unix"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	"github.com/containerd/cri/pkg/annotations"
+	customopts "github.com/containerd/cri/pkg/containerd/opts"
+	osinterface "github.com/containerd/cri/pkg/os"
+)
+
+func (c *criService) sandboxContainerSpec(id string, config *runtime.PodSandboxConfig,
+	imageConfig *imagespec.ImageConfig, nsPath string, runtimePodAnnotations []string) (*runtimespec.Spec, error) {
+	// Creates a spec Generator with the default spec.
+	// TODO(random-liu): [P1] Compare the default settings with docker and containerd default.
+	specOpts := []oci.SpecOpts{
+		customopts.WithoutRunMount,
+		customopts.WithoutDefaultSecuritySettings,
+		customopts.WithRelativeRoot(relativeRootfsPath),
+		oci.WithEnv(imageConfig.Env),
+		oci.WithRootFSReadonly(),
+		oci.WithHostname(config.GetHostname()),
+	}
+	if imageConfig.WorkingDir != "" {
+		specOpts = append(specOpts, oci.WithProcessCwd(imageConfig.WorkingDir))
+	}
+
+	if len(imageConfig.Entrypoint) == 0 && len(imageConfig.Cmd) == 0 {
+		// Pause image must have entrypoint or cmd.
+		return nil, errors.Errorf("invalid empty entrypoint and cmd in image config %+v", imageConfig)
+	}
+	specOpts = append(specOpts, oci.WithProcessArgs(append(imageConfig.Entrypoint, imageConfig.Cmd...)...))
+
+	// Set cgroups parent.
+	if c.config.DisableCgroup {
+		specOpts = append(specOpts, customopts.WithDisabledCgroups)
+	} else {
+		if config.GetLinux().GetCgroupParent() != "" {
+			cgroupsPath := getCgroupsPath(config.GetLinux().GetCgroupParent(), id)
+			specOpts = append(specOpts, oci.WithCgroup(cgroupsPath))
+		}
+	}
+
+	// When cgroup parent is not set, containerd-shim will create container in a child cgroup
+	// of the cgroup itself is in.
+	// TODO(random-liu): [P2] Set default cgroup path if cgroup parent is not specified.
+
+	// Set namespace options.
+	var (
+		securityContext = config.GetLinux().GetSecurityContext()
+		nsOptions       = securityContext.GetNamespaceOptions()
+	)
+	if nsOptions.GetNetwork() == runtime.NamespaceMode_NODE {
+		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.NetworkNamespace))
+		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.UTSNamespace))
+	} else {
+		specOpts = append(specOpts, oci.WithLinuxNamespace(
+			runtimespec.LinuxNamespace{
+				Type: runtimespec.NetworkNamespace,
+				Path: nsPath,
+			}))
+	}
+	if nsOptions.GetPid() == runtime.NamespaceMode_NODE {
+		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.PIDNamespace))
+	}
+	if nsOptions.GetIpc() == runtime.NamespaceMode_NODE {
+		specOpts = append(specOpts, customopts.WithoutNamespace(runtimespec.IPCNamespace))
+	}
+
+	// It's fine to generate the spec before the sandbox /dev/shm
+	// is actually created.
+	sandboxDevShm := c.getSandboxDevShm(id)
+	if nsOptions.GetIpc() == runtime.NamespaceMode_NODE {
+		sandboxDevShm = devShm
+	}
+	specOpts = append(specOpts, oci.WithMounts([]runtimespec.Mount{
+		{
+			Source:      sandboxDevShm,
+			Destination: devShm,
+			Type:        "bind",
+			Options:     []string{"rbind", "ro"},
+		},
+	}))
+
+	selinuxOpt := securityContext.GetSelinuxOptions()
+	processLabel, mountLabel, err := initSelinuxOpts(selinuxOpt)
+	if err != nil {
+		return nil, errors.Wrapf(err, "failed to init selinux options %+v", securityContext.GetSelinuxOptions())
+	}
+
+	supplementalGroups := securityContext.GetSupplementalGroups()
+	specOpts = append(specOpts,
+		customopts.WithSelinuxLabels(processLabel, mountLabel),
+		customopts.WithSupplementalGroups(supplementalGroups),
+	)
+
+	// Add sysctls
+	sysctls := config.GetLinux().GetSysctls()
+	specOpts = append(specOpts, customopts.WithSysctls(sysctls))
+
+	// Note: LinuxSandboxSecurityContext does not currently provide an apparmor profile
+
+	if !c.config.DisableCgroup {
+		specOpts = append(specOpts, customopts.WithDefaultSandboxShares)
+	}
+	specOpts = append(specOpts, customopts.WithPodOOMScoreAdj(int(defaultSandboxOOMAdj), c.config.RestrictOOMScoreAdj))
+
+	for pKey, pValue := range getPassthroughAnnotations(config.Annotations,
+		runtimePodAnnotations) {
+		specOpts = append(specOpts, customopts.WithAnnotation(pKey, pValue))
+	}
+
+	specOpts = append(specOpts,
+		customopts.WithAnnotation(annotations.ContainerType, annotations.ContainerTypeSandbox),
+		customopts.WithAnnotation(annotations.SandboxID, id),
+		customopts.WithAnnotation(annotations.SandboxLogDir, config.GetLogDirectory()),
+	)
+
+	return runtimeSpec(id, specOpts...)
+}
+
+// sandboxContainerSpecOpts generates OCI spec options for
+// the sandbox container.
+func (c *criService) sandboxContainerSpecOpts(config *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
+	var (
+		securityContext = config.GetLinux().GetSecurityContext()
+		specOpts        []oci.SpecOpts
+	)
+	seccompSpecOpts, err := generateSeccompSpecOpts(
+		securityContext.GetSeccompProfilePath(),
+		securityContext.GetPrivileged(),
+		c.seccompEnabled())
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to generate seccomp spec opts")
+	}
+	if seccompSpecOpts != nil {
+		specOpts = append(specOpts, seccompSpecOpts)
+	}
+
+	userstr, err := generateUserString(
+		"",
+		securityContext.GetRunAsUser(),
+		securityContext.GetRunAsGroup(),
+	)
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to generate user string")
+	}
+	if userstr == "" {
+		// Lastly, since no user override was passed via CRI try to set via OCI
+		// Image
+		userstr = imageConfig.User
+	}
+	if userstr != "" {
+		specOpts = append(specOpts, oci.WithUser(userstr))
+	}
+	return specOpts, nil
+}
+
+// setupSandboxFiles sets up necessary sandbox files including /dev/shm, /etc/hosts,
+// /etc/resolv.conf and /etc/hostname.
+func (c *criService) setupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
+	sandboxEtcHostname := c.getSandboxHostname(id)
+	hostname := config.GetHostname()
+	if hostname == "" {
+		var err error
+		hostname, err = c.os.Hostname()
+		if err != nil {
+			return errors.Wrap(err, "failed to get hostname")
+		}
+	}
+	if err := c.os.WriteFile(sandboxEtcHostname, []byte(hostname+"\n"), 0644); err != nil {
+		return errors.Wrapf(err, "failed to write hostname to %q", sandboxEtcHostname)
+	}
+
+	// TODO(random-liu): Consider whether we should maintain /etc/hosts and /etc/resolv.conf in kubelet.
+	sandboxEtcHosts := c.getSandboxHosts(id)
+	if err := c.os.CopyFile(etcHosts, sandboxEtcHosts, 0644); err != nil {
+		return errors.Wrapf(err, "failed to generate sandbox hosts file %q", sandboxEtcHosts)
+	}
+
+	// Set DNS options. Maintain a resolv.conf for the sandbox.
+	var err error
+	resolvContent := ""
+	if dnsConfig := config.GetDnsConfig(); dnsConfig != nil {
+		resolvContent, err = parseDNSOptions(dnsConfig.Servers, dnsConfig.Searches, dnsConfig.Options)
+		if err != nil {
+			return errors.Wrapf(err, "failed to parse sandbox DNSConfig %+v", dnsConfig)
+		}
+	}
+	resolvPath := c.getResolvPath(id)
+	if resolvContent == "" {
+		// copy host's resolv.conf to resolvPath
+		err = c.os.CopyFile(resolvConfPath, resolvPath, 0644)
+		if err != nil {
+			return errors.Wrapf(err, "failed to copy host's resolv.conf to %q", resolvPath)
+		}
+	} else {
+		err = c.os.WriteFile(resolvPath, []byte(resolvContent), 0644)
+		if err != nil {
+			return errors.Wrapf(err, "failed to write resolv content to %q", resolvPath)
+		}
+	}
+
+	// Setup sandbox /dev/shm.
+	if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetIpc() == runtime.NamespaceMode_NODE {
+		if _, err := c.os.Stat(devShm); err != nil {
+			return errors.Wrapf(err, "host %q is not available for host ipc", devShm)
+		}
+	} else {
+		sandboxDevShm := c.getSandboxDevShm(id)
+		if err := c.os.MkdirAll(sandboxDevShm, 0700); err != nil {
+			return errors.Wrap(err, "failed to create sandbox shm")
+		}
+		shmproperty := fmt.Sprintf("mode=1777,size=%d", defaultShmSize)
+		if err := c.os.(osinterface.UNIX).Mount("shm", sandboxDevShm, "tmpfs", uintptr(unix.MS_NOEXEC|unix.MS_NOSUID|unix.MS_NODEV), shmproperty); err != nil {
+			return errors.Wrap(err, "failed to mount sandbox shm")
+		}
+	}
+
+	return nil
+}
+
+// parseDNSOptions parse DNS options into resolv.conf format content,
+// if none option is specified, will return empty with no error.
+func parseDNSOptions(servers, searches, options []string) (string, error) {
+	resolvContent := ""
+
+	if len(searches) > maxDNSSearches {
+		return "", errors.Errorf("DNSOption.Searches has more than %d domains", maxDNSSearches)
+	}
+
+	if len(searches) > 0 {
+		resolvContent += fmt.Sprintf("search %s\n", strings.Join(searches, " "))
+	}
+
+	if len(servers) > 0 {
+		resolvContent += fmt.Sprintf("nameserver %s\n", strings.Join(servers, "\nnameserver "))
+	}
+
+	if len(options) > 0 {
+		resolvContent += fmt.Sprintf("options %s\n", strings.Join(options, " "))
+	}
+
+	return resolvContent, nil
+}
+
+// cleanupSandboxFiles unmount some sandbox files, we rely on the removal of sandbox root directory to
+// remove these files. Unmount should *NOT* return error if the mount point is already unmounted.
+func (c *criService) cleanupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
+	if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetIpc() != runtime.NamespaceMode_NODE {
+		path, err := c.os.FollowSymlinkInScope(c.getSandboxDevShm(id), "/")
+		if err != nil {
+			return errors.Wrap(err, "failed to follow symlink")
+		}
+		if err := c.os.(osinterface.UNIX).Unmount(path); err != nil && !os.IsNotExist(err) {
+			return errors.Wrapf(err, "failed to unmount %q", path)
+		}
+	}
+	return nil
+}
+
+// taskOpts generates task options for a (sandbox) container.
+func (c *criService) taskOpts(runtimeType string) []containerd.NewTaskOpts {
+	// TODO(random-liu): Remove this after shim v1 is deprecated.
+	var taskOpts []containerd.NewTaskOpts
+	if c.config.NoPivot && runtimeType == plugin.RuntimeRuncV1 {
+		taskOpts = append(taskOpts, containerd.WithNoPivotRoot)
+	}
+	return taskOpts
+}

pkg/server/sandbox_run_unix_test.go

@@ -0,0 +1,520 @@
+// +build !windows
+
+/*
+Copyright 2017 The Kubernetes Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/containerd/typeurl"
+	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
+	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+
+	"github.com/containerd/cri/pkg/annotations"
+	"github.com/containerd/cri/pkg/containerd/opts"
+	ostesting "github.com/containerd/cri/pkg/os/testing"
+	sandboxstore "github.com/containerd/cri/pkg/store/sandbox"
+)
+
+func getRunPodSandboxTestData() (*runtime.PodSandboxConfig, *imagespec.ImageConfig, func(*testing.T, string, *runtimespec.Spec)) {
+	config := &runtime.PodSandboxConfig{
+		Metadata: &runtime.PodSandboxMetadata{
+			Name:      "test-name",
+			Uid:       "test-uid",
+			Namespace: "test-ns",
+			Attempt:   1,
+		},
+		Hostname:     "test-hostname",
+		LogDirectory: "test-log-directory",
+		Labels:       map[string]string{"a": "b"},
+		Annotations:  map[string]string{"c": "d"},
+		Linux: &runtime.LinuxPodSandboxConfig{
+			CgroupParent: "/test/cgroup/parent",
+		},
+	}
+	imageConfig := &imagespec.ImageConfig{
+		Env:        []string{"a=b", "c=d"},
+		Entrypoint: []string{"/pause"},
+		Cmd:        []string{"forever"},
+		WorkingDir: "/workspace",
+	}
+	specCheck := func(t *testing.T, id string, spec *runtimespec.Spec) {
+		assert.Equal(t, "test-hostname", spec.Hostname)
+		assert.Equal(t, getCgroupsPath("/test/cgroup/parent", id), spec.Linux.CgroupsPath)
+		assert.Equal(t, relativeRootfsPath, spec.Root.Path)
+		assert.Equal(t, true, spec.Root.Readonly)
+		assert.Contains(t, spec.Process.Env, "a=b", "c=d")
+		assert.Equal(t, []string{"/pause", "forever"}, spec.Process.Args)
+		assert.Equal(t, "/workspace", spec.Process.Cwd)
+		assert.EqualValues(t, *spec.Linux.Resources.CPU.Shares, opts.DefaultSandboxCPUshares)
+		assert.EqualValues(t, *spec.Process.OOMScoreAdj, defaultSandboxOOMAdj)
+
+		t.Logf("Check PodSandbox annotations")
+		assert.Contains(t, spec.Annotations, annotations.SandboxID)
+		assert.EqualValues(t, spec.Annotations[annotations.SandboxID], id)
+
+		assert.Contains(t, spec.Annotations, annotations.ContainerType)
+		assert.EqualValues(t, spec.Annotations[annotations.ContainerType], annotations.ContainerTypeSandbox)
+
+		assert.Contains(t, spec.Annotations, annotations.SandboxLogDir)
+		assert.EqualValues(t, spec.Annotations[annotations.SandboxLogDir], "test-log-directory")
+	}
+	return config, imageConfig, specCheck
+}
+
+func TestSandboxContainerSpec(t *testing.T) {
+	testID := "test-id"
+	nsPath := "test-cni"
+	for desc, test := range map[string]struct {
+		configChange      func(*runtime.PodSandboxConfig)
+		podAnnotations    []string
+		imageConfigChange func(*imagespec.ImageConfig)
+		specCheck         func(*testing.T, *runtimespec.Spec)
+		expectErr         bool
+	}{
+		"spec should reflect original config": {
+			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
+				// runtime spec should have expected namespaces enabled by default.
+				require.NotNil(t, spec.Linux)
+				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.NetworkNamespace,
+					Path: nsPath,
+				})
+				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.UTSNamespace,
+				})
+				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.PIDNamespace,
+				})
+				assert.Contains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.IPCNamespace,
+				})
+			},
+		},
+		"host namespace": {
+			configChange: func(c *runtime.PodSandboxConfig) {
+				c.Linux.SecurityContext = &runtime.LinuxSandboxSecurityContext{
+					NamespaceOptions: &runtime.NamespaceOption{
+						Network: runtime.NamespaceMode_NODE,
+						Pid:     runtime.NamespaceMode_NODE,
+						Ipc:     runtime.NamespaceMode_NODE,
+					},
+				}
+			},
+			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
+				// runtime spec should disable expected namespaces in host mode.
+				require.NotNil(t, spec.Linux)
+				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.NetworkNamespace,
+				})
+				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.UTSNamespace,
+				})
+				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.PIDNamespace,
+				})
+				assert.NotContains(t, spec.Linux.Namespaces, runtimespec.LinuxNamespace{
+					Type: runtimespec.IPCNamespace,
+				})
+			},
+		},
+		"should return error when entrypoint and cmd are empty": {
+			imageConfigChange: func(c *imagespec.ImageConfig) {
+				c.Entrypoint = nil
+				c.Cmd = nil
+			},
+			expectErr: true,
+		},
+		"should set supplemental groups correctly": {
+			configChange: func(c *runtime.PodSandboxConfig) {
+				c.Linux.SecurityContext = &runtime.LinuxSandboxSecurityContext{
+					SupplementalGroups: []int64{1111, 2222},
+				}
+			},
+			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
+				require.NotNil(t, spec.Process)
+				assert.Contains(t, spec.Process.User.AdditionalGids, uint32(1111))
+				assert.Contains(t, spec.Process.User.AdditionalGids, uint32(2222))
+			},
+		},
+		"a passthrough annotation should be passed as an OCI annotation": {
+			podAnnotations: []string{"c"},
+			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
+				assert.Equal(t, spec.Annotations["c"], "d")
+			},
+		},
+		"a non-passthrough annotation should not be passed as an OCI annotation": {
+			configChange: func(c *runtime.PodSandboxConfig) {
+				c.Annotations["d"] = "e"
+			},
+			podAnnotations: []string{"c"},
+			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
+				assert.Equal(t, spec.Annotations["c"], "d")
+				_, ok := spec.Annotations["d"]
+				assert.False(t, ok)
+			},
+		},
+		"passthrough annotations should support wildcard match": {
+			configChange: func(c *runtime.PodSandboxConfig) {
+				c.Annotations["t.f"] = "j"
+				c.Annotations["z.g"] = "o"
+				c.Annotations["z"] = "o"
+				c.Annotations["y.ca"] = "b"
+				c.Annotations["y"] = "b"
+			},
+			podAnnotations: []string{"t*", "z.*", "y.c*"},
+			specCheck: func(t *testing.T, spec *runtimespec.Spec) {
+				assert.Equal(t, spec.Annotations["t.f"], "j")
+				assert.Equal(t, spec.Annotations["z.g"], "o")
+				assert.Equal(t, spec.Annotations["y.ca"], "b")
+				_, ok := spec.Annotations["y"]
+				assert.False(t, ok)
+				_, ok = spec.Annotations["z"]
+				assert.False(t, ok)
+			},
+		},
+	} {
+		t.Logf("TestCase %q", desc)
+		c := newTestCRIService()
+		config, imageConfig, specCheck := getRunPodSandboxTestData()
+		if test.configChange != nil {
+			test.configChange(config)
+		}
+
+		if test.imageConfigChange != nil {
+			test.imageConfigChange(imageConfig)
+		}
+		spec, err := c.sandboxContainerSpec(testID, config, imageConfig, nsPath,
+			test.podAnnotations)
+		if test.expectErr {
+			assert.Error(t, err)
+			assert.Nil(t, spec)
+			continue
+		}
+		assert.NoError(t, err)
+		assert.NotNil(t, spec)
+		specCheck(t, testID, spec)
+		if test.specCheck != nil {
+			test.specCheck(t, spec)
+		}
+	}
+}
+
+func TestSetupSandboxFiles(t *testing.T) {
+	const (
+		testID       = "test-id"
+		realhostname = "test-real-hostname"
+	)
+	for desc, test := range map[string]struct {
+		dnsConfig     *runtime.DNSConfig
+		hostname      string
+		ipcMode       runtime.NamespaceMode
+		expectedCalls []ostesting.CalledDetail
+	}{
+		"should check host /dev/shm existence when ipc mode is NODE": {
+			ipcMode: runtime.NamespaceMode_NODE,
+			expectedCalls: []ostesting.CalledDetail{
+				{
+					Name: "Hostname",
+				},
+				{
+					Name: "WriteFile",
+					Arguments: []interface{}{
+						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
+						[]byte(realhostname + "\n"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/hosts",
+						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/resolv.conf",
+						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name:      "Stat",
+					Arguments: []interface{}{"/dev/shm"},
+				},
+			},
+		},
+		"should create new /etc/resolv.conf if DNSOptions is set": {
+			dnsConfig: &runtime.DNSConfig{
+				Servers:  []string{"8.8.8.8"},
+				Searches: []string{"114.114.114.114"},
+				Options:  []string{"timeout:1"},
+			},
+			ipcMode: runtime.NamespaceMode_NODE,
+			expectedCalls: []ostesting.CalledDetail{
+				{
+					Name: "Hostname",
+				},
+				{
+					Name: "WriteFile",
+					Arguments: []interface{}{
+						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
+						[]byte(realhostname + "\n"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/hosts",
+						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "WriteFile",
+					Arguments: []interface{}{
+						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
+						[]byte(`search 114.114.114.114
+nameserver 8.8.8.8
+options timeout:1
+`), os.FileMode(0644),
+					},
+				},
+				{
+					Name:      "Stat",
+					Arguments: []interface{}{"/dev/shm"},
+				},
+			},
+		},
+		"should create sandbox shm when ipc namespace mode is not NODE": {
+			ipcMode: runtime.NamespaceMode_POD,
+			expectedCalls: []ostesting.CalledDetail{
+				{
+					Name: "Hostname",
+				},
+				{
+					Name: "WriteFile",
+					Arguments: []interface{}{
+						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
+						[]byte(realhostname + "\n"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/hosts",
+						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/resolv.conf",
+						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "MkdirAll",
+					Arguments: []interface{}{
+						filepath.Join(testStateDir, sandboxesDir, testID, "shm"),
+						os.FileMode(0700),
+					},
+				},
+				{
+					Name: "Mount",
+					// Ignore arguments which are too complex to check.
+				},
+			},
+		},
+		"should create /etc/hostname when hostname is set": {
+			hostname: "test-hostname",
+			ipcMode:  runtime.NamespaceMode_NODE,
+			expectedCalls: []ostesting.CalledDetail{
+				{
+					Name: "WriteFile",
+					Arguments: []interface{}{
+						filepath.Join(testRootDir, sandboxesDir, testID, "hostname"),
+						[]byte("test-hostname\n"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/hosts",
+						filepath.Join(testRootDir, sandboxesDir, testID, "hosts"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name: "CopyFile",
+					Arguments: []interface{}{
+						"/etc/resolv.conf",
+						filepath.Join(testRootDir, sandboxesDir, testID, "resolv.conf"),
+						os.FileMode(0644),
+					},
+				},
+				{
+					Name:      "Stat",
+					Arguments: []interface{}{"/dev/shm"},
+				},
+			},
+		},
+	} {
+		t.Logf("TestCase %q", desc)
+		c := newTestCRIService()
+		c.os.(*ostesting.FakeOS).HostnameFn = func() (string, error) {
+			return realhostname, nil
+		}
+		cfg := &runtime.PodSandboxConfig{
+			Hostname:  test.hostname,
+			DnsConfig: test.dnsConfig,
+			Linux: &runtime.LinuxPodSandboxConfig{
+				SecurityContext: &runtime.LinuxSandboxSecurityContext{
+					NamespaceOptions: &runtime.NamespaceOption{
+						Ipc: test.ipcMode,
+					},
+				},
+			},
+		}
+		c.setupSandboxFiles(testID, cfg)
+		calls := c.os.(*ostesting.FakeOS).GetCalls()
+		assert.Len(t, calls, len(test.expectedCalls))
+		for i, expected := range test.expectedCalls {
+			if expected.Arguments == nil {
+				// Ignore arguments.
+				expected.Arguments = calls[i].Arguments
+			}
+			assert.Equal(t, expected, calls[i])
+		}
+	}
+}
+
+func TestParseDNSOption(t *testing.T) {
+	for desc, test := range map[string]struct {
+		servers         []string
+		searches        []string
+		options         []string
+		expectedContent string
+		expectErr       bool
+	}{
+		"empty dns options should return empty content": {},
+		"non-empty dns options should return correct content": {
+			servers:  []string{"8.8.8.8", "server.google.com"},
+			searches: []string{"114.114.114.114"},
+			options:  []string{"timeout:1"},
+			expectedContent: `search 114.114.114.114
+nameserver 8.8.8.8
+nameserver server.google.com
+options timeout:1
+`,
+		},
+		"should return error if dns search exceeds limit(6)": {
+			searches: []string{
+				"server0.google.com",
+				"server1.google.com",
+				"server2.google.com",
+				"server3.google.com",
+				"server4.google.com",
+				"server5.google.com",
+				"server6.google.com",
+			},
+			expectErr: true,
+		},
+	} {
+		t.Logf("TestCase %q", desc)
+		resolvContent, err := parseDNSOptions(test.servers, test.searches, test.options)
+		if test.expectErr {
+			assert.Error(t, err)
+			continue
+		}
+		assert.NoError(t, err)
+		assert.Equal(t, resolvContent, test.expectedContent)
+	}
+}
+
+// TODO(windows): Move this to sandbox_run_test.go
+func TestTypeurlMarshalUnmarshalSandboxMeta(t *testing.T) {
+	for desc, test := range map[string]struct {
+		configChange func(*runtime.PodSandboxConfig)
+	}{
+		"should marshal original config": {},
+		"should marshal Linux": {
+			configChange: func(c *runtime.PodSandboxConfig) {
+				c.Linux.SecurityContext = &runtime.LinuxSandboxSecurityContext{
+					NamespaceOptions: &runtime.NamespaceOption{
+						Network: runtime.NamespaceMode_NODE,
+						Pid:     runtime.NamespaceMode_NODE,
+						Ipc:     runtime.NamespaceMode_NODE,
+					},
+					SupplementalGroups: []int64{1111, 2222},
+				}
+			},
+		},
+	} {
+		t.Logf("TestCase %q", desc)
+		meta := &sandboxstore.Metadata{
+			ID:        "1",
+			Name:      "sandbox_1",
+			NetNSPath: "/home/cloud",
+		}
+		meta.Config, _, _ = getRunPodSandboxTestData()
+		if test.configChange != nil {
+			test.configChange(meta.Config)
+		}
+
+		any, err := typeurl.MarshalAny(meta)
+		assert.NoError(t, err)
+		data, err := typeurl.UnmarshalAny(any)
+		assert.NoError(t, err)
+		assert.IsType(t, &sandboxstore.Metadata{}, data)
+		curMeta, ok := data.(*sandboxstore.Metadata)
+		assert.True(t, ok)
+		assert.Equal(t, meta, curMeta)
+	}
+}
+
+func TestSandboxDisableCgroup(t *testing.T) {
+	config, imageConfig, _ := getRunPodSandboxTestData()
+	c := newTestCRIService()
+	c.config.DisableCgroup = true
+	spec, err := c.sandboxContainerSpec("test-id", config, imageConfig, "test-cni", []string{})
+	require.NoError(t, err)
+
+	t.Log("resource limit should not be set")
+	assert.Nil(t, spec.Linux.Resources.Memory)
+	assert.Nil(t, spec.Linux.Resources.CPU)
+
+	t.Log("cgroup path should be empty")
+	assert.Empty(t, spec.Linux.CgroupsPath)
+}
+
+// TODO(random-liu): [P1] Add unit test for different error cases to make sure
+// the function cleans up on error properly.

pkg/server/sandbox_run_windows.go

@@ -0,0 +1,55 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"github.com/containerd/containerd"
+	"github.com/containerd/containerd/errdefs"
+	"github.com/containerd/containerd/oci"
+	imagespec "github.com/opencontainers/image-spec/specs-go/v1"
+	runtimespec "github.com/opencontainers/runtime-spec/specs-go"
+	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
+)
+
+// TODO(windows): Add windows support.
+// TODO(windows): Configure windows sandbox shares
+func (c *criService) sandboxContainerSpec(id string, config *runtime.PodSandboxConfig,
+	imageConfig *imagespec.ImageConfig, nsPath string, runtimePodAnnotations []string) (*runtimespec.Spec, error) {
+	return nil, errdefs.ErrNotImplemented
+}
+
+// No sandbox container spec options for windows yet.
+func (c *criService) sandboxContainerSpecOpts(config *runtime.PodSandboxConfig, imageConfig *imagespec.ImageConfig) ([]oci.SpecOpts, error) {
+	return nil, nil
+}
+
+// No sandbox files needed for windows.
+func (c *criService) setupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
+	return nil
+}
+
+// No sandbox files needed for windows.
+func (c *criService) cleanupSandboxFiles(id string, config *runtime.PodSandboxConfig) error {
+	return nil
+}
+
+// No task options needed for windows.
+func (c *criService) taskOpts(runtimeType string) []containerd.NewTaskOpts {
+	return nil
+}

pkg/server/sandbox_status.go

@@ -18,6 +18,7 @@ package server
 
 import (
 	"encoding/json"
+	goruntime "runtime"
 
 	"github.com/containerd/containerd"
 	"github.com/containerd/containerd/errdefs"
@@ -69,7 +70,8 @@ func (c *criService) PodSandboxStatus(ctx context.Context, r *runtime.PodSandbox
 func (c *criService) getIP(sandbox sandboxstore.Sandbox) (string, error) {
 	config := sandbox.Config
 
-	if config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
+	if goruntime.GOOS != "windows" &&
+		config.GetLinux().GetSecurityContext().GetNamespaceOptions().GetNetwork() == runtime.NamespaceMode_NODE {
 		// For sandboxes using the node network we are not
 		// responsible for reporting the IP.
 		return "", nil

pkg/server/sandbox_stop.go

@@ -57,8 +57,8 @@ func (c *criService) StopPodSandbox(ctx context.Context, r *runtime.StopPodSandb
 		}
 	}
 
-	if err := c.unmountSandboxFiles(id, sandbox.Config); err != nil {
+	if err := c.cleanupSandboxFiles(id, sandbox.Config); err != nil {
-		return nil, errors.Wrap(err, "failed to unmount sandbox files")
+		return nil, errors.Wrap(err, "failed to cleanup sandbox files")
 	}
 
 	// Only stop sandbox container when it's running or unknown.

pkg/server/service.go

@@ -26,10 +26,6 @@ import (
 	"github.com/containerd/containerd"
 	"github.com/containerd/containerd/plugin"
 	cni "github.com/containerd/go-cni"
-	runcapparmor "github.com/opencontainers/runc/libcontainer/apparmor"
-	runcseccomp "github.com/opencontainers/runc/libcontainer/seccomp"
-	runcsystem "github.com/opencontainers/runc/libcontainer/system"
-	"github.com/opencontainers/selinux/go-selinux"
 	"github.com/pkg/errors"
 	"github.com/sirupsen/logrus"
 	"google.golang.org/grpc"
@@ -68,10 +64,6 @@ type criService struct {
 	config criconfig.Config
 	// imageFSPath is the path to image filesystem.
 	imageFSPath string
-	// apparmorEnabled indicates whether apparmor is enabled.
-	apparmorEnabled bool
-	// seccompEnabled indicates whether seccomp is enabled.
-	seccompEnabled bool
 	// os is an interface for all required os operations.
 	os osinterface.OS
 	// sandboxStore stores all resources associated with sandboxes.
@@ -107,8 +99,6 @@ func NewCRIService(config criconfig.Config, client *containerd.Client) (CRIServi
 	c := &criService{
 		config:             config,
 		client:             client,
-		apparmorEnabled:    runcapparmor.IsEnabled() && !config.DisableApparmor,
-		seccompEnabled:     runcseccomp.IsEnabled(),
 		os:                 osinterface.RealOS{},
 		sandboxStore:       sandboxstore.NewStore(),
 		containerStore:     containerstore.NewStore(),
@@ -119,20 +109,6 @@ func NewCRIService(config criconfig.Config, client *containerd.Client) (CRIServi
 		initialized:        atomic.NewBool(false),
 	}
 
-	if runcsystem.RunningInUserNS() {
-		if !(config.DisableCgroup && !c.apparmorEnabled && config.RestrictOOMScoreAdj) {
-			logrus.Warn("Running containerd in a user namespace typically requires disable_cgroup, disable_apparmor, restrict_oom_score_adj set to be true")
-		}
-	}
-
-	if c.config.EnableSelinux {
-		if !selinux.GetEnabled() {
-			logrus.Warn("Selinux is not supported")
-		}
-	} else {
-		selinux.SetDisabled()
-	}
-
 	if client.SnapshotService(c.config.ContainerdConfig.Snapshotter) == nil {
 		return nil, errors.Errorf("failed to find snapshotter %q", c.config.ContainerdConfig.Snapshotter)
 	}
@@ -140,23 +116,10 @@ func NewCRIService(config criconfig.Config, client *containerd.Client) (CRIServi
 	c.imageFSPath = imageFSPath(config.ContainerdRootDir, config.ContainerdConfig.Snapshotter)
 	logrus.Infof("Get image filesystem path %q", c.imageFSPath)
 
-	// Pod needs to attach to atleast loopback network and a non host network,
+	if err := c.initPlatform(); err != nil {
-	// hence networkAttachCount is 2. If there are more network configs the
+		return nil, errors.Wrap(err, "initialize platform")
-	// pod will be attached to all the networks but we will only use the ip
-	// of the default network interface as the pod IP.
-	c.netPlugin, err = cni.New(cni.WithMinNetworkCount(networkAttachCount),
-		cni.WithPluginConfDir(config.NetworkPluginConfDir),
-		cni.WithPluginMaxConfNum(config.NetworkPluginMaxConfNum),
-		cni.WithPluginDir([]string{config.NetworkPluginBinDir}))
-	if err != nil {
-		return nil, errors.Wrap(err, "failed to initialize cni")
 	}
 
-	// Try to load the config if it exists. Just log the error if load fails
-	// This is not disruptive for containerd to panic
-	if err := c.netPlugin.Load(cni.WithLoNetwork, cni.WithDefaultConf); err != nil {
-		logrus.WithError(err).Error("Failed to load cni during init, please check CRI plugin status before setting up network for pods")
-	}
 	// prepare streaming server
 	c.streamServer, err = newStreamServer(c, config.StreamServerAddress, config.StreamServerPort, config.StreamIdleTimeout)
 	if err != nil {

pkg/server/service_unix.go

@@ -0,0 +1,74 @@
+// +build !windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	cni "github.com/containerd/go-cni"
+	runcsystem "github.com/opencontainers/runc/libcontainer/system"
+	"github.com/opencontainers/selinux/go-selinux"
+	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
+)
+
+// networkAttachCount is the minimum number of networks the PodSandbox
+// attaches to
+const networkAttachCount = 2
+
+// initPlatform handles linux specific initialization for the CRI service.
+func (c *criService) initPlatform() error {
+	var err error
+
+	if runcsystem.RunningInUserNS() {
+		if !(c.config.DisableCgroup && !c.apparmorEnabled() && c.config.RestrictOOMScoreAdj) {
+			logrus.Warn("Running containerd in a user namespace typically requires disable_cgroup, disable_apparmor, restrict_oom_score_adj set to be true")
+		}
+	}
+
+	if c.config.EnableSelinux {
+		if !selinux.GetEnabled() {
+			logrus.Warn("Selinux is not supported")
+		}
+	} else {
+		selinux.SetDisabled()
+	}
+
+	// Pod needs to attach to at least loopback network and a non host network,
+	// hence networkAttachCount is 2. If there are more network configs the
+	// pod will be attached to all the networks but we will only use the ip
+	// of the default network interface as the pod IP.
+	c.netPlugin, err = cni.New(cni.WithMinNetworkCount(networkAttachCount),
+		cni.WithPluginConfDir(c.config.NetworkPluginConfDir),
+		cni.WithPluginMaxConfNum(c.config.NetworkPluginMaxConfNum),
+		cni.WithPluginDir([]string{c.config.NetworkPluginBinDir}))
+	if err != nil {
+		return errors.Wrap(err, "failed to initialize cni")
+	}
+
+	// Try to load the config if it exists. Just log the error if load fails
+	// This is not disruptive for containerd to panic
+	if err := c.netPlugin.Load(c.cniLoadOptions()...); err != nil {
+		logrus.WithError(err).Error("Failed to load cni during init, please check CRI plugin status before setting up network for pods")
+	}
+	return nil
+}
+
+// cniLoadOptions returns cni load options for the linux.
+func (c *criService) cniLoadOptions() []cni.CNIOpt {
+	return []cni.CNIOpt{cni.WithLoNetwork, cni.WithDefaultConf}
+}

pkg/server/service_windows.go

@@ -0,0 +1,35 @@
+// +build windows
+
+/*
+Copyright The containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	cni "github.com/containerd/go-cni"
+)
+
+// initPlatform handles linux specific initialization for the CRI service.
+// TODO(windows): Initialize CRI plugin for windows
+func (c *criService) initPlatform() error {
+	return nil
+}
+
+// cniLoadOptions returns cni load options for the windows.
+// TODO(windows): Implement CNI options for windows.
+func (c *criService) cniLoadOptions() []cni.CNIOpt {
+	return nil
+}

pkg/server/status.go

@@ -22,7 +22,6 @@ import (
 	goruntime "runtime"
 
 	"github.com/containerd/containerd/log"
-	cni "github.com/containerd/go-cni"
 	"golang.org/x/net/context"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
 )
@@ -42,9 +41,8 @@ func (c *criService) Status(ctx context.Context, r *runtime.StatusRequest) (*run
 		Type:   runtime.NetworkReady,
 		Status: true,
 	}
-
 	// Load the latest cni configuration to be in sync with the latest network configuration
-	if err := c.netPlugin.Load(cni.WithLoNetwork, cni.WithDefaultConf); err != nil {
+	if err := c.netPlugin.Load(c.cniLoadOptions()...); err != nil {
 		log.G(ctx).WithError(err).Errorf("Failed to load cni configuration")
 	}
 	// Check the status of the cni initialization

pkg/server/update_runtime_config.go

@@ -22,7 +22,6 @@ import (
 	"text/template"
 
 	"github.com/containerd/containerd/log"
-	cni "github.com/containerd/go-cni"
 	"github.com/pkg/errors"
 	"golang.org/x/net/context"
 	runtime "k8s.io/cri-api/pkg/apis/runtime/v1alpha2"
@@ -52,7 +51,7 @@ func (c *criService) UpdateRuntimeConfig(ctx context.Context, r *runtime.UpdateR
 	if err := c.netPlugin.Status(); err == nil {
 		log.G(ctx).Infof("Network plugin is ready, skip generating cni config from template %q", confTemplate)
 		return &runtime.UpdateRuntimeConfigResponse{}, nil
-	} else if err := c.netPlugin.Load(cni.WithLoNetwork, cni.WithDefaultConf); err == nil {
+	} else if err := c.netPlugin.Load(c.cniLoadOptions()...); err == nil {
 		log.G(ctx).Infof("CNI config is successfully loaded, skip generating cni config from template %q", confTemplate)
 		return &runtime.UpdateRuntimeConfigResponse{}, nil
 	}

test/verify.sh

@@ -26,3 +26,4 @@ cd "${ROOT}"
 
 make install.tools
 make verify
+GOOS=windows make verify

vendor.conf

@@ -25,6 +25,7 @@ github.com/opencontainers/runc f4982d86f7fde0b6f953cc62ccc4022c519a10a9 # v1.0.0
 github.com/opencontainers/image-spec v1.0.1
 github.com/opencontainers/go-digest c9281466c8b2f606084ac71339773efd177436e7
 github.com/matttproud/golang_protobuf_extensions v1.0.1
+github.com/konsorten/go-windows-terminal-sequences v1.0.1
 github.com/grpc-ecosystem/go-grpc-prometheus v1.1
 github.com/google/uuid v1.1.1
 github.com/golang/protobuf v1.2.0

vendor/github.com/Microsoft/go-winio/pkg/fs/fs_windows.go

@@ -0,0 +1,47 @@
+package fs
+
+import (
+	"errors"
+	"path/filepath"
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var (
+	// ErrInvalidPath is returned when the location of a file path doesn't begin with a driver letter.
+	ErrInvalidPath = errors.New("the path provided to GetFileSystemType must start with a drive letter")
+)
+
+// GetFileSystemType obtains the type of a file system through GetVolumeInformation.
+// https://msdn.microsoft.com/en-us/library/windows/desktop/aa364993(v=vs.85).aspx
+func GetFileSystemType(path string) (fsType string, hr error) {
+	drive := filepath.VolumeName(path)
+	if len(drive) != 2 {
+		return "", ErrInvalidPath
+	}
+
+	var (
+		modkernel32              = windows.NewLazySystemDLL("kernel32.dll")
+		procGetVolumeInformation = modkernel32.NewProc("GetVolumeInformationW")
+		buf                      = make([]uint16, 255)
+		size                     = windows.MAX_PATH + 1
+	)
+	drive += `\`
+	n := uintptr(unsafe.Pointer(nil))
+	r0, _, _ := syscall.Syscall9(procGetVolumeInformation.Addr(), 8, uintptr(unsafe.Pointer(windows.StringToUTF16Ptr(drive))), n, n, n, n, n, uintptr(unsafe.Pointer(&buf[0])), uintptr(size), 0)
+	if int32(r0) < 0 {
+		hr = syscall.Errno(win32FromHresult(r0))
+	}
+	fsType = windows.UTF16ToString(buf)
+	return
+}
+
+// win32FromHresult is a helper function to get the win32 error code from an HRESULT.
+func win32FromHresult(hr uintptr) uintptr {
+	if hr&0x1fff0000 == 0x00070000 {
+		return hr & 0xffff
+	}
+	return hr
+}

vendor/github.com/Microsoft/go-winio/pkg/security/grantvmgroupaccess.go

@@ -0,0 +1,159 @@
+package security
+
+import (
+	"os"
+	"syscall"
+	"unsafe"
+
+	"github.com/pkg/errors"
+)
+
+type (
+	accessMask          uint32
+	accessMode          uint32
+	desiredAccess       uint32
+	inheritMode         uint32
+	objectType          uint32
+	shareMode           uint32
+	securityInformation uint32
+	trusteeForm         uint32
+	trusteeType         uint32
+
+	explicitAccess struct {
+		accessPermissions accessMask
+		accessMode        accessMode
+		inheritance       inheritMode
+		trustee           trustee
+	}
+
+	trustee struct {
+		multipleTrustee          *trustee
+		multipleTrusteeOperation int32
+		trusteeForm              trusteeForm
+		trusteeType              trusteeType
+		name                     uintptr
+	}
+)
+
+const (
+	accessMaskDesiredPermission accessMask = 1 << 31 // GENERIC_READ
+
+	accessModeGrant accessMode = 1
+
+	desiredAccessReadControl desiredAccess = 0x20000
+	desiredAccessWriteDac    desiredAccess = 0x40000
+
+	gvmga = "GrantVmGroupAccess:"
+
+	inheritModeNoInheritance                  inheritMode = 0x0
+	inheritModeSubContainersAndObjectsInherit inheritMode = 0x3
+
+	objectTypeFileObject objectType = 0x1
+
+	securityInformationDACL securityInformation = 0x4
+
+	shareModeRead  shareMode = 0x1
+	shareModeWrite shareMode = 0x2
+
+	sidVmGroup = "S-1-5-83-0"
+
+	trusteeFormIsSid trusteeForm = 0
+
+	trusteeTypeWellKnownGroup trusteeType = 5
+)
+
+// GrantVMGroupAccess sets the DACL for a specified file or directory to
+// include Grant ACE entries for the VM Group SID. This is a golang re-
+// implementation of the same function in vmcompute, just not exported in
+// RS5. Which kind of sucks. Sucks a lot :/
+func GrantVmGroupAccess(name string) error {
+	// Stat (to determine if `name` is a directory).
+	s, err := os.Stat(name)
+	if err != nil {
+		return errors.Wrapf(err, "%s os.Stat %s", gvmga, name)
+	}
+
+	// Get a handle to the file/directory. Must defer Close on success.
+	fd, err := createFile(name, s.IsDir())
+	if err != nil {
+		return err // Already wrapped
+	}
+	defer syscall.CloseHandle(fd)
+
+	// Get the current DACL and Security Descriptor. Must defer LocalFree on success.
+	ot := objectTypeFileObject
+	si := securityInformationDACL
+	sd := uintptr(0)
+	origDACL := uintptr(0)
+	if err := getSecurityInfo(fd, uint32(ot), uint32(si), nil, nil, &origDACL, nil, &sd); err != nil {
+		return errors.Wrapf(err, "%s GetSecurityInfo %s", gvmga, name)
+	}
+	defer syscall.LocalFree((syscall.Handle)(unsafe.Pointer(sd)))
+
+	// Generate a new DACL which is the current DACL with the required ACEs added.
+	// Must defer LocalFree on success.
+	newDACL, err := generateDACLWithAcesAdded(name, s.IsDir(), origDACL)
+	if err != nil {
+		return err // Already wrapped
+	}
+	defer syscall.LocalFree((syscall.Handle)(unsafe.Pointer(newDACL)))
+
+	// And finally use SetSecurityInfo to apply the updated DACL.
+	if err := setSecurityInfo(fd, uint32(ot), uint32(si), uintptr(0), uintptr(0), newDACL, uintptr(0)); err != nil {
+		return errors.Wrapf(err, "%s SetSecurityInfo %s", gvmga, name)
+	}
+
+	return nil
+}
+
+// createFile is a helper function to call [Nt]CreateFile to get a handle to
+// the file or directory.
+func createFile(name string, isDir bool) (syscall.Handle, error) {
+	namep := syscall.StringToUTF16(name)
+	da := uint32(desiredAccessReadControl | desiredAccessWriteDac)
+	sm := uint32(shareModeRead | shareModeWrite)
+	fa := uint32(syscall.FILE_ATTRIBUTE_NORMAL)
+	if isDir {
+		fa = uint32(fa | syscall.FILE_FLAG_BACKUP_SEMANTICS)
+	}
+	fd, err := syscall.CreateFile(&namep[0], da, sm, nil, syscall.OPEN_EXISTING, fa, 0)
+	if err != nil {
+		return 0, errors.Wrapf(err, "%s syscall.CreateFile %s", gvmga, name)
+	}
+	return fd, nil
+}
+
+// generateDACLWithAcesAdded generates a new DACL with the two needed ACEs added.
+// The caller is responsible for LocalFree of the returned DACL on success.
+func generateDACLWithAcesAdded(name string, isDir bool, origDACL uintptr) (uintptr, error) {
+	// Generate pointers to the SIDs based on the string SIDs
+	sid, err := syscall.StringToSid(sidVmGroup)
+	if err != nil {
+		return 0, errors.Wrapf(err, "%s syscall.StringToSid %s %s", gvmga, name, sidVmGroup)
+	}
+
+	inheritance := inheritModeNoInheritance
+	if isDir {
+		inheritance = inheritModeSubContainersAndObjectsInherit
+	}
+
+	eaArray := []explicitAccess{
+		explicitAccess{
+			accessPermissions: accessMaskDesiredPermission,
+			accessMode:        accessModeGrant,
+			inheritance:       inheritance,
+			trustee: trustee{
+				trusteeForm: trusteeFormIsSid,
+				trusteeType: trusteeTypeWellKnownGroup,
+				name:        uintptr(unsafe.Pointer(sid)),
+			},
+		},
+	}
+
+	modifiedDACL := uintptr(0)
+	if err := setEntriesInAcl(uintptr(uint32(1)), uintptr(unsafe.Pointer(&eaArray[0])), origDACL, &modifiedDACL); err != nil {
+		return 0, errors.Wrapf(err, "%s SetEntriesInAcl %s", gvmga, name)
+	}
+
+	return modifiedDACL, nil
+}

vendor/github.com/Microsoft/go-winio/pkg/security/syscall_windows.go

@@ -0,0 +1,7 @@
+package security
+
+//go:generate go run mksyscall_windows.go -output zsyscall_windows.go syscall_windows.go
+
+//sys getSecurityInfo(handle syscall.Handle, objectType uint32, si uint32, ppsidOwner **uintptr, ppsidGroup **uintptr, ppDacl *uintptr, ppSacl *uintptr, ppSecurityDescriptor *uintptr) (err error) [failretval!=0] = advapi32.GetSecurityInfo
+//sys setSecurityInfo(handle syscall.Handle, objectType uint32, si uint32, psidOwner uintptr, psidGroup uintptr, pDacl uintptr, pSacl uintptr) (err error) [failretval!=0] = advapi32.SetSecurityInfo
+//sys setEntriesInAcl(count uintptr, pListOfEEs uintptr, oldAcl uintptr, newAcl *uintptr) (err error) [failretval!=0] = advapi32.SetEntriesInAclW

vendor/github.com/Microsoft/go-winio/pkg/security/zsyscall_windows.go

@@ -0,0 +1,81 @@
+// Code generated mksyscall_windows.exe DO NOT EDIT
+
+package security
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return nil
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	// TODO: add more here, after collecting data on the common
+	// error values see on Windows. (perhaps when running
+	// all.bat?)
+	return e
+}
+
+var (
+	modadvapi32 = windows.NewLazySystemDLL("advapi32.dll")
+
+	procGetSecurityInfo  = modadvapi32.NewProc("GetSecurityInfo")
+	procSetSecurityInfo  = modadvapi32.NewProc("SetSecurityInfo")
+	procSetEntriesInAclW = modadvapi32.NewProc("SetEntriesInAclW")
+)
+
+func getSecurityInfo(handle syscall.Handle, objectType uint32, si uint32, ppsidOwner **uintptr, ppsidGroup **uintptr, ppDacl *uintptr, ppSacl *uintptr, ppSecurityDescriptor *uintptr) (err error) {
+	r1, _, e1 := syscall.Syscall9(procGetSecurityInfo.Addr(), 8, uintptr(handle), uintptr(objectType), uintptr(si), uintptr(unsafe.Pointer(ppsidOwner)), uintptr(unsafe.Pointer(ppsidGroup)), uintptr(unsafe.Pointer(ppDacl)), uintptr(unsafe.Pointer(ppSacl)), uintptr(unsafe.Pointer(ppSecurityDescriptor)), 0)
+	if r1 != 0 {
+		if e1 != 0 {
+			err = errnoErr(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return
+}
+
+func setSecurityInfo(handle syscall.Handle, objectType uint32, si uint32, psidOwner uintptr, psidGroup uintptr, pDacl uintptr, pSacl uintptr) (err error) {
+	r1, _, e1 := syscall.Syscall9(procSetSecurityInfo.Addr(), 7, uintptr(handle), uintptr(objectType), uintptr(si), uintptr(psidOwner), uintptr(psidGroup), uintptr(pDacl), uintptr(pSacl), 0, 0)
+	if r1 != 0 {
+		if e1 != 0 {
+			err = errnoErr(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return
+}
+
+func setEntriesInAcl(count uintptr, pListOfEEs uintptr, oldAcl uintptr, newAcl *uintptr) (err error) {
+	r1, _, e1 := syscall.Syscall6(procSetEntriesInAclW.Addr(), 4, uintptr(count), uintptr(pListOfEEs), uintptr(oldAcl), uintptr(unsafe.Pointer(newAcl)), 0, 0)
+	if r1 != 0 {
+		if e1 != 0 {
+			err = errnoErr(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return
+}

vendor/github.com/Microsoft/go-winio/vhd/vhd.go

@@ -0,0 +1,151 @@
+// +build windows
+
+package vhd
+
+import "syscall"
+
+//go:generate go run mksyscall_windows.go -output zvhd.go vhd.go
+
+//sys createVirtualDisk(virtualStorageType *virtualStorageType, path string, virtualDiskAccessMask uint32, securityDescriptor *uintptr, flags uint32, providerSpecificFlags uint32, parameters *createVirtualDiskParameters, o *syscall.Overlapped, handle *syscall.Handle) (err error) [failretval != 0] = VirtDisk.CreateVirtualDisk
+//sys openVirtualDisk(virtualStorageType *virtualStorageType, path string, virtualDiskAccessMask uint32, flags uint32, parameters *openVirtualDiskParameters, handle *syscall.Handle) (err error) [failretval != 0] = VirtDisk.OpenVirtualDisk
+//sys detachVirtualDisk(handle syscall.Handle, flags uint32, providerSpecificFlags uint32) (err error) [failretval != 0] = VirtDisk.DetachVirtualDisk
+
+type virtualStorageType struct {
+	DeviceID uint32
+	VendorID [16]byte
+}
+
+type (
+	createVirtualDiskFlag uint32
+	VirtualDiskAccessMask uint32
+	VirtualDiskFlag       uint32
+)
+
+const (
+	// Flags for creating a VHD (not exported)
+	createVirtualDiskFlagNone                        createVirtualDiskFlag = 0
+	createVirtualDiskFlagFullPhysicalAllocation      createVirtualDiskFlag = 1
+	createVirtualDiskFlagPreventWritesToSourceDisk   createVirtualDiskFlag = 2
+	createVirtualDiskFlagDoNotCopyMetadataFromParent createVirtualDiskFlag = 4
+
+	// Access Mask for opening a VHD
+	VirtualDiskAccessNone     VirtualDiskAccessMask = 0
+	VirtualDiskAccessAttachRO VirtualDiskAccessMask = 65536
+	VirtualDiskAccessAttachRW VirtualDiskAccessMask = 131072
+	VirtualDiskAccessDetach   VirtualDiskAccessMask = 262144
+	VirtualDiskAccessGetInfo  VirtualDiskAccessMask = 524288
+	VirtualDiskAccessCreate   VirtualDiskAccessMask = 1048576
+	VirtualDiskAccessMetaOps  VirtualDiskAccessMask = 2097152
+	VirtualDiskAccessRead     VirtualDiskAccessMask = 851968
+	VirtualDiskAccessAll      VirtualDiskAccessMask = 4128768
+	VirtualDiskAccessWritable VirtualDiskAccessMask = 3276800
+
+	// Flags for opening a VHD
+	OpenVirtualDiskFlagNone                        VirtualDiskFlag = 0
+	OpenVirtualDiskFlagNoParents                   VirtualDiskFlag = 0x1
+	OpenVirtualDiskFlagBlankFile                   VirtualDiskFlag = 0x2
+	OpenVirtualDiskFlagBootDrive                   VirtualDiskFlag = 0x4
+	OpenVirtualDiskFlagCachedIO                    VirtualDiskFlag = 0x8
+	OpenVirtualDiskFlagCustomDiffChain             VirtualDiskFlag = 0x10
+	OpenVirtualDiskFlagParentCachedIO              VirtualDiskFlag = 0x20
+	OpenVirtualDiskFlagVhdSetFileOnly              VirtualDiskFlag = 0x40
+	OpenVirtualDiskFlagIgnoreRelativeParentLocator VirtualDiskFlag = 0x80
+	OpenVirtualDiskFlagNoWriteHardening            VirtualDiskFlag = 0x100
+)
+
+type createVersion2 struct {
+	UniqueID                 [16]byte // GUID
+	MaximumSize              uint64
+	BlockSizeInBytes         uint32
+	SectorSizeInBytes        uint32
+	ParentPath               *uint16 // string
+	SourcePath               *uint16 // string
+	OpenFlags                uint32
+	ParentVirtualStorageType virtualStorageType
+	SourceVirtualStorageType virtualStorageType
+	ResiliencyGUID           [16]byte // GUID
+}
+
+type createVirtualDiskParameters struct {
+	Version  uint32 // Must always be set to 2
+	Version2 createVersion2
+}
+
+type openVersion2 struct {
+	GetInfoOnly    int32    // bool but 4-byte aligned
+	ReadOnly       int32    // bool but 4-byte aligned
+	ResiliencyGUID [16]byte // GUID
+}
+
+type openVirtualDiskParameters struct {
+	Version  uint32 // Must always be set to 2
+	Version2 openVersion2
+}
+
+// CreateVhdx will create a simple vhdx file at the given path using default values.
+func CreateVhdx(path string, maxSizeInGb, blockSizeInMb uint32) error {
+	var (
+		defaultType virtualStorageType
+		handle      syscall.Handle
+	)
+
+	parameters := createVirtualDiskParameters{
+		Version: 2,
+		Version2: createVersion2{
+			MaximumSize:      uint64(maxSizeInGb) * 1024 * 1024 * 1024,
+			BlockSizeInBytes: blockSizeInMb * 1024 * 1024,
+		},
+	}
+
+	if err := createVirtualDisk(
+		&defaultType,
+		path,
+		uint32(VirtualDiskAccessNone),
+		nil,
+		uint32(createVirtualDiskFlagNone),
+		0,
+		&parameters,
+		nil,
+		&handle); err != nil {
+		return err
+	}
+
+	if err := syscall.CloseHandle(handle); err != nil {
+		return err
+	}
+
+	return nil
+}
+
+// DetachVhd detaches a mounted container layer vhd found at `path`.
+func DetachVhd(path string) error {
+	handle, err := OpenVirtualDisk(
+		path,
+		VirtualDiskAccessNone,
+		OpenVirtualDiskFlagCachedIO|OpenVirtualDiskFlagIgnoreRelativeParentLocator)
+
+	if err != nil {
+		return err
+	}
+	defer syscall.CloseHandle(handle)
+	return detachVirtualDisk(handle, 0, 0)
+}
+
+// OpenVirtualDisk obtains a handle to a VHD opened with supplied access mask and flags.
+func OpenVirtualDisk(path string, accessMask VirtualDiskAccessMask, flag VirtualDiskFlag) (syscall.Handle, error) {
+	var (
+		defaultType virtualStorageType
+		handle      syscall.Handle
+	)
+	parameters := openVirtualDiskParameters{Version: 2}
+	if err := openVirtualDisk(
+		&defaultType,
+		path,
+		uint32(accessMask),
+		uint32(flag),
+		&parameters,
+		&handle); err != nil {
+		return 0, err
+	}
+	return handle, nil
+}

vendor/github.com/Microsoft/go-winio/vhd/zvhd.go

@@ -0,0 +1,99 @@
+// MACHINE GENERATED BY 'go generate' COMMAND; DO NOT EDIT
+
+package vhd
+
+import (
+	"syscall"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+var _ unsafe.Pointer
+
+// Do the interface allocations only once for common
+// Errno values.
+const (
+	errnoERROR_IO_PENDING = 997
+)
+
+var (
+	errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING)
+)
+
+// errnoErr returns common boxed Errno values, to prevent
+// allocations at runtime.
+func errnoErr(e syscall.Errno) error {
+	switch e {
+	case 0:
+		return nil
+	case errnoERROR_IO_PENDING:
+		return errERROR_IO_PENDING
+	}
+	// TODO: add more here, after collecting data on the common
+	// error values see on Windows. (perhaps when running
+	// all.bat?)
+	return e
+}
+
+var (
+	modVirtDisk = windows.NewLazySystemDLL("VirtDisk.dll")
+
+	procCreateVirtualDisk = modVirtDisk.NewProc("CreateVirtualDisk")
+	procOpenVirtualDisk   = modVirtDisk.NewProc("OpenVirtualDisk")
+	procDetachVirtualDisk = modVirtDisk.NewProc("DetachVirtualDisk")
+)
+
+func createVirtualDisk(virtualStorageType *virtualStorageType, path string, virtualDiskAccessMask uint32, securityDescriptor *uintptr, flags uint32, providerSpecificFlags uint32, parameters *createVirtualDiskParameters, o *syscall.Overlapped, handle *syscall.Handle) (err error) {
+	var _p0 *uint16
+	_p0, err = syscall.UTF16PtrFromString(path)
+	if err != nil {
+		return
+	}
+	return _createVirtualDisk(virtualStorageType, _p0, virtualDiskAccessMask, securityDescriptor, flags, providerSpecificFlags, parameters, o, handle)
+}
+
+func _createVirtualDisk(virtualStorageType *virtualStorageType, path *uint16, virtualDiskAccessMask uint32, securityDescriptor *uintptr, flags uint32, providerSpecificFlags uint32, parameters *createVirtualDiskParameters, o *syscall.Overlapped, handle *syscall.Handle) (err error) {
+	r1, _, e1 := syscall.Syscall9(procCreateVirtualDisk.Addr(), 9, uintptr(unsafe.Pointer(virtualStorageType)), uintptr(unsafe.Pointer(path)), uintptr(virtualDiskAccessMask), uintptr(unsafe.Pointer(securityDescriptor)), uintptr(flags), uintptr(providerSpecificFlags), uintptr(unsafe.Pointer(parameters)), uintptr(unsafe.Pointer(o)), uintptr(unsafe.Pointer(handle)))
+	if r1 != 0 {
+		if e1 != 0 {
+			err = errnoErr(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return
+}
+
+func openVirtualDisk(virtualStorageType *virtualStorageType, path string, virtualDiskAccessMask uint32, flags uint32, parameters *openVirtualDiskParameters, handle *syscall.Handle) (err error) {
+	var _p0 *uint16
+	_p0, err = syscall.UTF16PtrFromString(path)
+	if err != nil {
+		return
+	}
+	return _openVirtualDisk(virtualStorageType, _p0, virtualDiskAccessMask, flags, parameters, handle)
+}
+
+func _openVirtualDisk(virtualStorageType *virtualStorageType, path *uint16, virtualDiskAccessMask uint32, flags uint32, parameters *openVirtualDiskParameters, handle *syscall.Handle) (err error) {
+	r1, _, e1 := syscall.Syscall6(procOpenVirtualDisk.Addr(), 6, uintptr(unsafe.Pointer(virtualStorageType)), uintptr(unsafe.Pointer(path)), uintptr(virtualDiskAccessMask), uintptr(flags), uintptr(unsafe.Pointer(parameters)), uintptr(unsafe.Pointer(handle)))
+	if r1 != 0 {
+		if e1 != 0 {
+			err = errnoErr(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return
+}
+
+func detachVirtualDisk(handle syscall.Handle, flags uint32, providerSpecificFlags uint32) (err error) {
+	r1, _, e1 := syscall.Syscall(procDetachVirtualDisk.Addr(), 3, uintptr(handle), uintptr(flags), uintptr(providerSpecificFlags))
+	if r1 != 0 {
+		if e1 != 0 {
+			err = errnoErr(e1)
+		} else {
+			err = syscall.EINVAL
+		}
+	}
+	return
+}

vendor/github.com/Microsoft/hcsshim/ext4/internal/format/format.go

@@ -0,0 +1,411 @@
+package format
+
+type SuperBlock struct {
+	InodesCount          uint32
+	BlocksCountLow       uint32
+	RootBlocksCountLow   uint32
+	FreeBlocksCountLow   uint32
+	FreeInodesCount      uint32
+	FirstDataBlock       uint32
+	LogBlockSize         uint32
+	LogClusterSize       uint32
+	BlocksPerGroup       uint32
+	ClustersPerGroup     uint32
+	InodesPerGroup       uint32
+	Mtime                uint32
+	Wtime                uint32
+	MountCount           uint16
+	MaxMountCount        uint16
+	Magic                uint16
+	State                uint16
+	Errors               uint16
+	MinorRevisionLevel   uint16
+	LastCheck            uint32
+	CheckInterval        uint32
+	CreatorOS            uint32
+	RevisionLevel        uint32
+	DefaultReservedUid   uint16
+	DefaultReservedGid   uint16
+	FirstInode           uint32
+	InodeSize            uint16
+	BlockGroupNr         uint16
+	FeatureCompat        CompatFeature
+	FeatureIncompat      IncompatFeature
+	FeatureRoCompat      RoCompatFeature
+	UUID                 [16]uint8
+	VolumeName           [16]byte
+	LastMounted          [64]byte
+	AlgorithmUsageBitmap uint32
+	PreallocBlocks       uint8
+	PreallocDirBlocks    uint8
+	ReservedGdtBlocks    uint16
+	JournalUUID          [16]uint8
+	JournalInum          uint32
+	JournalDev           uint32
+	LastOrphan           uint32
+	HashSeed             [4]uint32
+	DefHashVersion       uint8
+	JournalBackupType    uint8
+	DescSize             uint16
+	DefaultMountOpts     uint32
+	FirstMetaBg          uint32
+	MkfsTime             uint32
+	JournalBlocks        [17]uint32
+	BlocksCountHigh      uint32
+	RBlocksCountHigh     uint32
+	FreeBlocksCountHigh  uint32
+	MinExtraIsize        uint16
+	WantExtraIsize       uint16
+	Flags                uint32
+	RaidStride           uint16
+	MmpInterval          uint16
+	MmpBlock             uint64
+	RaidStripeWidth      uint32
+	LogGroupsPerFlex     uint8
+	ChecksumType         uint8
+	ReservedPad          uint16
+	KbytesWritten        uint64
+	SnapshotInum         uint32
+	SnapshotID           uint32
+	SnapshotRBlocksCount uint64
+	SnapshotList         uint32
+	ErrorCount           uint32
+	FirstErrorTime       uint32
+	FirstErrorInode      uint32
+	FirstErrorBlock      uint64
+	FirstErrorFunc       [32]uint8
+	FirstErrorLine       uint32
+	LastErrorTime        uint32
+	LastErrorInode       uint32
+	LastErrorLine        uint32
+	LastErrorBlock       uint64
+	LastErrorFunc        [32]uint8
+	MountOpts            [64]uint8
+	UserQuotaInum        uint32
+	GroupQuotaInum       uint32
+	OverheadBlocks       uint32
+	BackupBgs            [2]uint32
+	EncryptAlgos         [4]uint8
+	EncryptPwSalt        [16]uint8
+	LpfInode             uint32
+	ProjectQuotaInum     uint32
+	ChecksumSeed         uint32
+	WtimeHigh            uint8
+	MtimeHigh            uint8
+	MkfsTimeHigh         uint8
+	LastcheckHigh        uint8
+	FirstErrorTimeHigh   uint8
+	LastErrorTimeHigh    uint8
+	Pad                  [2]uint8
+	Reserved             [96]uint32
+	Checksum             uint32
+}
+
+const SuperBlockMagic uint16 = 0xef53
+
+type CompatFeature uint32
+type IncompatFeature uint32
+type RoCompatFeature uint32
+
+const (
+	CompatDirPrealloc   CompatFeature = 0x1
+	CompatImagicInodes  CompatFeature = 0x2
+	CompatHasJournal    CompatFeature = 0x4
+	CompatExtAttr       CompatFeature = 0x8
+	CompatResizeInode   CompatFeature = 0x10
+	CompatDirIndex      CompatFeature = 0x20
+	CompatLazyBg        CompatFeature = 0x40
+	CompatExcludeInode  CompatFeature = 0x80
+	CompatExcludeBitmap CompatFeature = 0x100
+	CompatSparseSuper2  CompatFeature = 0x200
+
+	IncompatCompression IncompatFeature = 0x1
+	IncompatFiletype    IncompatFeature = 0x2
+	IncompatRecover     IncompatFeature = 0x4
+	IncompatJournalDev  IncompatFeature = 0x8
+	IncompatMetaBg      IncompatFeature = 0x10
+	IncompatExtents     IncompatFeature = 0x40
+	Incompat_64Bit      IncompatFeature = 0x80
+	IncompatMmp         IncompatFeature = 0x100
+	IncompatFlexBg      IncompatFeature = 0x200
+	IncompatEaInode     IncompatFeature = 0x400
+	IncompatDirdata     IncompatFeature = 0x1000
+	IncompatCsumSeed    IncompatFeature = 0x2000
+	IncompatLargedir    IncompatFeature = 0x4000
+	IncompatInlineData  IncompatFeature = 0x8000
+	IncompatEncrypt     IncompatFeature = 0x10000
+
+	RoCompatSparseSuper  RoCompatFeature = 0x1
+	RoCompatLargeFile    RoCompatFeature = 0x2
+	RoCompatBtreeDir     RoCompatFeature = 0x4
+	RoCompatHugeFile     RoCompatFeature = 0x8
+	RoCompatGdtCsum      RoCompatFeature = 0x10
+	RoCompatDirNlink     RoCompatFeature = 0x20
+	RoCompatExtraIsize   RoCompatFeature = 0x40
+	RoCompatHasSnapshot  RoCompatFeature = 0x80
+	RoCompatQuota        RoCompatFeature = 0x100
+	RoCompatBigalloc     RoCompatFeature = 0x200
+	RoCompatMetadataCsum RoCompatFeature = 0x400
+	RoCompatReplica      RoCompatFeature = 0x800
+	RoCompatReadonly     RoCompatFeature = 0x1000
+	RoCompatProject      RoCompatFeature = 0x2000
+)
+
+type BlockGroupFlag uint16
+
+const (
+	BlockGroupInodeUninit BlockGroupFlag = 0x1
+	BlockGroupBlockUninit BlockGroupFlag = 0x2
+	BlockGroupInodeZeroed BlockGroupFlag = 0x4
+)
+
+type GroupDescriptor struct {
+	BlockBitmapLow     uint32
+	InodeBitmapLow     uint32
+	InodeTableLow      uint32
+	FreeBlocksCountLow uint16
+	FreeInodesCountLow uint16
+	UsedDirsCountLow   uint16
+	Flags              BlockGroupFlag
+	ExcludeBitmapLow   uint32
+	BlockBitmapCsumLow uint16
+	InodeBitmapCsumLow uint16
+	ItableUnusedLow    uint16
+	Checksum           uint16
+}
+
+type GroupDescriptor64 struct {
+	GroupDescriptor
+	BlockBitmapHigh     uint32
+	InodeBitmapHigh     uint32
+	InodeTableHigh      uint32
+	FreeBlocksCountHigh uint16
+	FreeInodesCountHigh uint16
+	UsedDirsCountHigh   uint16
+	ItableUnusedHigh    uint16
+	ExcludeBitmapHigh   uint32
+	BlockBitmapCsumHigh uint16
+	InodeBitmapCsumHigh uint16
+	Reserved            uint32
+}
+
+const (
+	S_IXOTH  = 0x1
+	S_IWOTH  = 0x2
+	S_IROTH  = 0x4
+	S_IXGRP  = 0x8
+	S_IWGRP  = 0x10
+	S_IRGRP  = 0x20
+	S_IXUSR  = 0x40
+	S_IWUSR  = 0x80
+	S_IRUSR  = 0x100
+	S_ISVTX  = 0x200
+	S_ISGID  = 0x400
+	S_ISUID  = 0x800
+	S_IFIFO  = 0x1000
+	S_IFCHR  = 0x2000
+	S_IFDIR  = 0x4000
+	S_IFBLK  = 0x6000
+	S_IFREG  = 0x8000
+	S_IFLNK  = 0xA000
+	S_IFSOCK = 0xC000
+
+	TypeMask uint16 = 0xF000
+)
+
+type InodeNumber uint32
+
+const (
+	InodeRoot = 2
+)
+
+type Inode struct {
+	Mode                 uint16
+	Uid                  uint16
+	SizeLow              uint32
+	Atime                uint32
+	Ctime                uint32
+	Mtime                uint32
+	Dtime                uint32
+	Gid                  uint16
+	LinksCount           uint16
+	BlocksLow            uint32
+	Flags                InodeFlag
+	Version              uint32
+	Block                [60]byte
+	Generation           uint32
+	XattrBlockLow        uint32
+	SizeHigh             uint32
+	ObsoleteFragmentAddr uint32
+	BlocksHigh           uint16
+	XattrBlockHigh       uint16
+	UidHigh              uint16
+	GidHigh              uint16
+	ChecksumLow          uint16
+	Reserved             uint16
+	ExtraIsize           uint16
+	ChecksumHigh         uint16
+	CtimeExtra           uint32
+	MtimeExtra           uint32
+	AtimeExtra           uint32
+	Crtime               uint32
+	CrtimeExtra          uint32
+	VersionHigh          uint32
+	Projid               uint32
+}
+
+type InodeFlag uint32
+
+const (
+	InodeFlagSecRm              InodeFlag = 0x1
+	InodeFlagUnRm               InodeFlag = 0x2
+	InodeFlagCompressed         InodeFlag = 0x4
+	InodeFlagSync               InodeFlag = 0x8
+	InodeFlagImmutable          InodeFlag = 0x10
+	InodeFlagAppend             InodeFlag = 0x20
+	InodeFlagNoDump             InodeFlag = 0x40
+	InodeFlagNoAtime            InodeFlag = 0x80
+	InodeFlagDirtyCompressed    InodeFlag = 0x100
+	InodeFlagCompressedClusters InodeFlag = 0x200
+	InodeFlagNoCompress         InodeFlag = 0x400
+	InodeFlagEncrypted          InodeFlag = 0x800
+	InodeFlagHashedIndex        InodeFlag = 0x1000
+	InodeFlagMagic              InodeFlag = 0x2000
+	InodeFlagJournalData        InodeFlag = 0x4000
+	InodeFlagNoTail             InodeFlag = 0x8000
+	InodeFlagDirSync            InodeFlag = 0x10000
+	InodeFlagTopDir             InodeFlag = 0x20000
+	InodeFlagHugeFile           InodeFlag = 0x40000
+	InodeFlagExtents            InodeFlag = 0x80000
+	InodeFlagEaInode            InodeFlag = 0x200000
+	InodeFlagEOFBlocks          InodeFlag = 0x400000
+	InodeFlagSnapfile           InodeFlag = 0x01000000
+	InodeFlagSnapfileDeleted    InodeFlag = 0x04000000
+	InodeFlagSnapfileShrunk     InodeFlag = 0x08000000
+	InodeFlagInlineData         InodeFlag = 0x10000000
+	InodeFlagProjectIDInherit   InodeFlag = 0x20000000
+	InodeFlagReserved           InodeFlag = 0x80000000
+)
+
+const (
+	MaxLinks = 65000
+)
+
+type ExtentHeader struct {
+	Magic      uint16
+	Entries    uint16
+	Max        uint16
+	Depth      uint16
+	Generation uint32
+}
+
+const ExtentHeaderMagic uint16 = 0xf30a
+
+type ExtentIndexNode struct {
+	Block    uint32
+	LeafLow  uint32
+	LeafHigh uint16
+	Unused   uint16
+}
+
+type ExtentLeafNode struct {
+	Block     uint32
+	Length    uint16
+	StartHigh uint16
+	StartLow  uint32
+}
+
+type ExtentTail struct {
+	Checksum uint32
+}
+
+type DirectoryEntry struct {
+	Inode        InodeNumber
+	RecordLength uint16
+	NameLength   uint8
+	FileType     FileType
+	//Name         []byte
+}
+
+type FileType uint8
+
+const (
+	FileTypeUnknown      FileType = 0x0
+	FileTypeRegular      FileType = 0x1
+	FileTypeDirectory    FileType = 0x2
+	FileTypeCharacter    FileType = 0x3
+	FileTypeBlock        FileType = 0x4
+	FileTypeFIFO         FileType = 0x5
+	FileTypeSocket       FileType = 0x6
+	FileTypeSymbolicLink FileType = 0x7
+)
+
+type DirectoryEntryTail struct {
+	ReservedZero1 uint32
+	RecordLength  uint16
+	ReservedZero2 uint8
+	FileType      uint8
+	Checksum      uint32
+}
+
+type DirectoryTreeRoot struct {
+	Dot            DirectoryEntry
+	DotName        [4]byte
+	DotDot         DirectoryEntry
+	DotDotName     [4]byte
+	ReservedZero   uint32
+	HashVersion    uint8
+	InfoLength     uint8
+	IndirectLevels uint8
+	UnusedFlags    uint8
+	Limit          uint16
+	Count          uint16
+	Block          uint32
+	//Entries        []DirectoryTreeEntry
+}
+
+type DirectoryTreeNode struct {
+	FakeInode        uint32
+	FakeRecordLength uint16
+	NameLength       uint8
+	FileType         uint8
+	Limit            uint16
+	Count            uint16
+	Block            uint32
+	//Entries          []DirectoryTreeEntry
+}
+
+type DirectoryTreeEntry struct {
+	Hash  uint32
+	Block uint32
+}
+
+type DirectoryTreeTail struct {
+	Reserved uint32
+	Checksum uint32
+}
+
+type XAttrInodeBodyHeader struct {
+	Magic uint32
+}
+
+type XAttrHeader struct {
+	Magic          uint32
+	ReferenceCount uint32
+	Blocks         uint32
+	Hash           uint32
+	Checksum       uint32
+	Reserved       [3]uint32
+}
+
+const XAttrHeaderMagic uint32 = 0xea020000
+
+type XAttrEntry struct {
+	NameLength  uint8
+	NameIndex   uint8
+	ValueOffset uint16
+	ValueInum   uint32
+	ValueSize   uint32
+	Hash        uint32
+	//Name        []byte
+}

vendor/github.com/Microsoft/hcsshim/ext4/tar2ext4/tar2ext4.go

@@ -0,0 +1,174 @@
+package tar2ext4
+
+import (
+	"archive/tar"
+	"bufio"
+	"encoding/binary"
+	"io"
+	"path"
+	"strings"
+
+	"github.com/Microsoft/hcsshim/ext4/internal/compactext4"
+)
+
+type params struct {
+	convertWhiteout bool
+	appendVhdFooter bool
+	ext4opts        []compactext4.Option
+}
+
+// Option is the type for optional parameters to Convert.
+type Option func(*params)
+
+// ConvertWhiteout instructs the converter to convert OCI-style whiteouts
+// (beginning with .wh.) to overlay-style whiteouts.
+func ConvertWhiteout(p *params) {
+	p.convertWhiteout = true
+}
+
+// AppendVhdFooter instructs the converter to add a fixed VHD footer to the
+// file.
+func AppendVhdFooter(p *params) {
+	p.appendVhdFooter = true
+}
+
+// InlineData instructs the converter to write small files into the inode
+// structures directly. This creates smaller images but currently is not
+// compatible with DAX.
+func InlineData(p *params) {
+	p.ext4opts = append(p.ext4opts, compactext4.InlineData)
+}
+
+// MaximumDiskSize instructs the writer to limit the disk size to the specified
+// value. This also reserves enough metadata space for the specified disk size.
+// If not provided, then 16GB is the default.
+func MaximumDiskSize(size int64) Option {
+	return func(p *params) {
+		p.ext4opts = append(p.ext4opts, compactext4.MaximumDiskSize(size))
+	}
+}
+
+const (
+	whiteoutPrefix = ".wh."
+	opaqueWhiteout = ".wh..wh..opq"
+)
+
+// Convert writes a compact ext4 file system image that contains the files in the
+// input tar stream.
+func Convert(r io.Reader, w io.ReadWriteSeeker, options ...Option) error {
+	var p params
+	for _, opt := range options {
+		opt(&p)
+	}
+	t := tar.NewReader(bufio.NewReader(r))
+	fs := compactext4.NewWriter(w, p.ext4opts...)
+	for {
+		hdr, err := t.Next()
+		if err == io.EOF {
+			break
+		}
+		if err != nil {
+			return err
+		}
+
+		if p.convertWhiteout {
+			dir, name := path.Split(hdr.Name)
+			if strings.HasPrefix(name, whiteoutPrefix) {
+				if name == opaqueWhiteout {
+					// Update the directory with the appropriate xattr.
+					f, err := fs.Stat(dir)
+					if err != nil {
+						return err
+					}
+					f.Xattrs["trusted.overlay.opaque"] = []byte("y")
+					err = fs.Create(dir, f)
+					if err != nil {
+						return err
+					}
+				} else {
+					// Create an overlay-style whiteout.
+					f := &compactext4.File{
+						Mode:     compactext4.S_IFCHR,
+						Devmajor: 0,
+						Devminor: 0,
+					}
+					err = fs.Create(path.Join(dir, name[len(whiteoutPrefix):]), f)
+					if err != nil {
+						return err
+					}
+				}
+
+				continue
+			}
+		}
+
+		if hdr.Typeflag == tar.TypeLink {
+			err = fs.Link(hdr.Linkname, hdr.Name)
+			if err != nil {
+				return err
+			}
+		} else {
+			f := &compactext4.File{
+				Mode:     uint16(hdr.Mode),
+				Atime:    hdr.AccessTime,
+				Mtime:    hdr.ModTime,
+				Ctime:    hdr.ChangeTime,
+				Crtime:   hdr.ModTime,
+				Size:     hdr.Size,
+				Uid:      uint32(hdr.Uid),
+				Gid:      uint32(hdr.Gid),
+				Linkname: hdr.Linkname,
+				Devmajor: uint32(hdr.Devmajor),
+				Devminor: uint32(hdr.Devminor),
+				Xattrs:   make(map[string][]byte),
+			}
+			for key, value := range hdr.PAXRecords {
+				const xattrPrefix = "SCHILY.xattr."
+				if strings.HasPrefix(key, xattrPrefix) {
+					f.Xattrs[key[len(xattrPrefix):]] = []byte(value)
+				}
+			}
+
+			var typ uint16
+			switch hdr.Typeflag {
+			case tar.TypeReg, tar.TypeRegA:
+				typ = compactext4.S_IFREG
+			case tar.TypeSymlink:
+				typ = compactext4.S_IFLNK
+			case tar.TypeChar:
+				typ = compactext4.S_IFCHR
+			case tar.TypeBlock:
+				typ = compactext4.S_IFBLK
+			case tar.TypeDir:
+				typ = compactext4.S_IFDIR
+			case tar.TypeFifo:
+				typ = compactext4.S_IFIFO
+			}
+			f.Mode &= ^compactext4.TypeMask
+			f.Mode |= typ
+			err = fs.Create(hdr.Name, f)
+			if err != nil {
+				return err
+			}
+			_, err = io.Copy(fs, t)
+			if err != nil {
+				return err
+			}
+		}
+	}
+	err := fs.Close()
+	if err != nil {
+		return err
+	}
+	if p.appendVhdFooter {
+		size, err := w.Seek(0, io.SeekEnd)
+		if err != nil {
+			return err
+		}
+		err = binary.Write(w, binary.BigEndian, makeFixedVHDFooter(size))
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}

vendor/github.com/Microsoft/hcsshim/ext4/tar2ext4/vhdfooter.go

@@ -0,0 +1,76 @@
+package tar2ext4
+
+import (
+	"bytes"
+	"crypto/rand"
+	"encoding/binary"
+)
+
+// Constants for the VHD footer
+const (
+	cookieMagic            = "conectix"
+	featureMask            = 0x2
+	fileFormatVersionMagic = 0x00010000
+	fixedDataOffset        = -1
+	creatorVersionMagic    = 0x000a0000
+	diskTypeFixed          = 2
+)
+
+type vhdFooter struct {
+	Cookie             [8]byte
+	Features           uint32
+	FileFormatVersion  uint32
+	DataOffset         int64
+	TimeStamp          uint32
+	CreatorApplication [4]byte
+	CreatorVersion     uint32
+	CreatorHostOS      [4]byte
+	OriginalSize       int64
+	CurrentSize        int64
+	DiskGeometry       uint32
+	DiskType           uint32
+	Checksum           uint32
+	UniqueID           [16]uint8
+	SavedState         uint8
+	Reserved           [427]uint8
+}
+
+func makeFixedVHDFooter(size int64) *vhdFooter {
+	footer := &vhdFooter{
+		Features:          featureMask,
+		FileFormatVersion: fileFormatVersionMagic,
+		DataOffset:        fixedDataOffset,
+		CreatorVersion:    creatorVersionMagic,
+		OriginalSize:      size,
+		CurrentSize:       size,
+		DiskType:          diskTypeFixed,
+		UniqueID:          generateUUID(),
+	}
+	copy(footer.Cookie[:], cookieMagic)
+	footer.Checksum = calculateCheckSum(footer)
+	return footer
+}
+
+func calculateCheckSum(footer *vhdFooter) uint32 {
+	oldchk := footer.Checksum
+	footer.Checksum = 0
+
+	buf := &bytes.Buffer{}
+	binary.Write(buf, binary.BigEndian, footer)
+
+	var chk uint32
+	bufBytes := buf.Bytes()
+	for i := 0; i < len(bufBytes); i++ {
+		chk += uint32(bufBytes[i])
+	}
+	footer.Checksum = oldchk
+	return uint32(^chk)
+}
+
+func generateUUID() [16]byte {
+	res := [16]byte{}
+	if _, err := rand.Read(res[:]); err != nil {
+		panic(err)
+	}
+	return res
+}

vendor/github.com/containerd/containerd/diff/lcow/lcow.go

@@ -0,0 +1,210 @@
+// +build windows
+
+/*
+   Copyright The containerd Authors.
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
+*/
+
+package lcow
+
+import (
+	"context"
+	"io"
+	"os"
+	"path"
+	"time"
+
+	"github.com/Microsoft/go-winio/pkg/security"
+	"github.com/Microsoft/hcsshim/ext4/tar2ext4"
+	"github.com/containerd/containerd/content"
+	"github.com/containerd/containerd/diff"
+	"github.com/containerd/containerd/errdefs"
+	"github.com/containerd/containerd/log"
+	"github.com/containerd/containerd/metadata"
+	"github.com/containerd/containerd/mount"
+	"github.com/containerd/containerd/plugin"
+	digest "github.com/opencontainers/go-digest"
+	ocispec "github.com/opencontainers/image-spec/specs-go/v1"
+	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
+)
+
+const (
+	// maxLcowVhdSizeGB is the max size in GB of any layer
+	maxLcowVhdSizeGB = 128 * 1024 * 1024 * 1024
+)
+
+func init() {
+	plugin.Register(&plugin.Registration{
+		Type: plugin.DiffPlugin,
+		ID:   "windows-lcow",
+		Requires: []plugin.Type{
+			plugin.MetadataPlugin,
+		},
+		InitFn: func(ic *plugin.InitContext) (interface{}, error) {
+			md, err := ic.Get(plugin.MetadataPlugin)
+			if err != nil {
+				return nil, err
+			}
+
+			ic.Meta.Platforms = append(ic.Meta.Platforms, ocispec.Platform{
+				OS:           "linux",
+				Architecture: "amd64",
+			})
+			return NewWindowsLcowDiff(md.(*metadata.DB).ContentStore())
+		},
+	})
+}
+
+// CompareApplier handles both comparison and
+// application of layer diffs.
+type CompareApplier interface {
+	diff.Applier
+	diff.Comparer
+}
+
+// windowsLcowDiff does filesystem comparison and application
+// for Windows specific Linux layer diffs.
+type windowsLcowDiff struct {
+	store content.Store
+}
+
+var emptyDesc = ocispec.Descriptor{}
+
+// NewWindowsLcowDiff is the Windows LCOW container layer implementation
+// for comparing and applying Linux filesystem layers on Windows
+func NewWindowsLcowDiff(store content.Store) (CompareApplier, error) {
+	return windowsLcowDiff{
+		store: store,
+	}, nil
+}
+
+// Apply applies the content associated with the provided digests onto the
+// provided mounts. Archive content will be extracted and decompressed if
+// necessary.
+func (s windowsLcowDiff) Apply(ctx context.Context, desc ocispec.Descriptor, mounts []mount.Mount, opts ...diff.ApplyOpt) (d ocispec.Descriptor, err error) {
+	t1 := time.Now()
+	defer func() {
+		if err == nil {
+			log.G(ctx).WithFields(logrus.Fields{
+				"d":     time.Since(t1),
+				"dgst":  desc.Digest,
+				"size":  desc.Size,
+				"media": desc.MediaType,
+			}).Debugf("diff applied")
+		}
+	}()
+
+	var config diff.ApplyConfig
+	for _, o := range opts {
+		if err := o(ctx, desc, &config); err != nil {
+			return emptyDesc, errors.Wrap(err, "failed to apply config opt")
+		}
+	}
+
+	layer, _, err := mountsToLayerAndParents(mounts)
+	if err != nil {
+		return emptyDesc, err
+	}
+
+	ra, err := s.store.ReaderAt(ctx, desc)
+	if err != nil {
+		return emptyDesc, errors.Wrap(err, "failed to get reader from content store")
+	}
+	defer ra.Close()
+
+	processor := diff.NewProcessorChain(desc.MediaType, content.NewReader(ra))
+	for {
+		if processor, err = diff.GetProcessor(ctx, processor, config.ProcessorPayloads); err != nil {
+			return emptyDesc, errors.Wrapf(err, "failed to get stream processor for %s", desc.MediaType)
+		}
+		if processor.MediaType() == ocispec.MediaTypeImageLayer {
+			break
+		}
+	}
+	defer processor.Close()
+
+	// Calculate the Digest as we go
+	digester := digest.Canonical.Digester()
+	rc := &readCounter{
+		r: io.TeeReader(processor, digester.Hash()),
+	}
+
+	layerPath := path.Join(layer, "layer.vhd")
+	outFile, err := os.Create(layerPath)
+	if err != nil {
+		return emptyDesc, err
+	}
+	defer func() {
+		if err != nil {
+			outFile.Close()
+			os.Remove(layerPath)
+		}
+	}()
+
+	err = tar2ext4.Convert(rc, outFile, tar2ext4.ConvertWhiteout, tar2ext4.AppendVhdFooter, tar2ext4.MaximumDiskSize(maxLcowVhdSizeGB))
+	if err != nil {
+		return emptyDesc, errors.Wrapf(err, "failed to convert tar2ext4 vhd")
+	}
+	err = outFile.Sync()
+	if err != nil {
+		return emptyDesc, errors.Wrapf(err, "failed to sync tar2ext4 vhd to disk")
+	}
+	outFile.Close()
+
+	err = security.GrantVmGroupAccess(layerPath)
+	if err != nil {
+		return emptyDesc, errors.Wrapf(err, "failed GrantVmGroupAccess on layer vhd: %v", layerPath)
+	}
+
+	return ocispec.Descriptor{
+		MediaType: ocispec.MediaTypeImageLayer,
+		Size:      rc.c,
+		Digest:    digester.Digest(),
+	}, nil
+}
+
+// Compare creates a diff between the given mounts and uploads the result
+// to the content store.
+func (s windowsLcowDiff) Compare(ctx context.Context, lower, upper []mount.Mount, opts ...diff.Opt) (d ocispec.Descriptor, err error) {
+	return emptyDesc, errdefs.ErrNotImplemented
+}
+
+type readCounter struct {
+	r io.Reader
+	c int64
+}
+
+func (rc *readCounter) Read(p []byte) (n int, err error) {
+	n, err = rc.r.Read(p)
+	rc.c += int64(n)
+	return
+}
+
+func mountsToLayerAndParents(mounts []mount.Mount) (string, []string, error) {
+	if len(mounts) != 1 {
+		return "", nil, errors.Wrap(errdefs.ErrInvalidArgument, "number of mounts should always be 1 for Windows lcow-layers")
+	}
+	mnt := mounts[0]
+	if mnt.Type != "lcow-layer" {
+		return "", nil, errors.Wrap(errdefs.ErrInvalidArgument, "mount layer type must be lcow-layer")
+	}
+
+	parentLayerPaths, err := mnt.GetParentPaths()
+	if err != nil {
+		return "", nil, err
+	}
+
+	return mnt.Source, parentLayerPaths, nil
+}

vendor/github.com/containerd/containerd/diff/windows/windows.go

@@ -0,0 +1,193 @@
+// +build windows
+
+/*
+   Copyright The containerd Authors.
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
+*/
+
+package windows
+
+import (
+	"context"
+	"io"
+	"io/ioutil"
+	"time"
+
+	winio "github.com/Microsoft/go-winio"
+	"github.com/containerd/containerd/archive"
+	"github.com/containerd/containerd/content"
+	"github.com/containerd/containerd/diff"
+	"github.com/containerd/containerd/errdefs"
+	"github.com/containerd/containerd/log"
+	"github.com/containerd/containerd/metadata"
+	"github.com/containerd/containerd/mount"
+	"github.com/containerd/containerd/platforms"
+	"github.com/containerd/containerd/plugin"
+	digest "github.com/opencontainers/go-digest"
+	ocispec "github.com/opencontainers/image-spec/specs-go/v1"
+	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
+)
+
+func init() {
+	plugin.Register(&plugin.Registration{
+		Type: plugin.DiffPlugin,
+		ID:   "windows",
+		Requires: []plugin.Type{
+			plugin.MetadataPlugin,
+		},
+		InitFn: func(ic *plugin.InitContext) (interface{}, error) {
+			md, err := ic.Get(plugin.MetadataPlugin)
+			if err != nil {
+				return nil, err
+			}
+
+			ic.Meta.Platforms = append(ic.Meta.Platforms, platforms.DefaultSpec())
+			return NewWindowsDiff(md.(*metadata.DB).ContentStore())
+		},
+	})
+}
+
+// CompareApplier handles both comparison and
+// application of layer diffs.
+type CompareApplier interface {
+	diff.Applier
+	diff.Comparer
+}
+
+// windowsDiff does filesystem comparison and application
+// for Windows specific layer diffs.
+type windowsDiff struct {
+	store content.Store
+}
+
+var emptyDesc = ocispec.Descriptor{}
+
+// NewWindowsDiff is the Windows container layer implementation
+// for comparing and applying filesystem layers
+func NewWindowsDiff(store content.Store) (CompareApplier, error) {
+	return windowsDiff{
+		store: store,
+	}, nil
+}
+
+// Apply applies the content associated with the provided digests onto the
+// provided mounts. Archive content will be extracted and decompressed if
+// necessary.
+func (s windowsDiff) Apply(ctx context.Context, desc ocispec.Descriptor, mounts []mount.Mount, opts ...diff.ApplyOpt) (d ocispec.Descriptor, err error) {
+	t1 := time.Now()
+	defer func() {
+		if err == nil {
+			log.G(ctx).WithFields(logrus.Fields{
+				"d":     time.Since(t1),
+				"dgst":  desc.Digest,
+				"size":  desc.Size,
+				"media": desc.MediaType,
+			}).Debugf("diff applied")
+		}
+	}()
+
+	var config diff.ApplyConfig
+	for _, o := range opts {
+		if err := o(ctx, desc, &config); err != nil {
+			return emptyDesc, errors.Wrap(err, "failed to apply config opt")
+		}
+	}
+
+	ra, err := s.store.ReaderAt(ctx, desc)
+	if err != nil {
+		return emptyDesc, errors.Wrap(err, "failed to get reader from content store")
+	}
+	defer ra.Close()
+
+	processor := diff.NewProcessorChain(desc.MediaType, content.NewReader(ra))
+	for {
+		if processor, err = diff.GetProcessor(ctx, processor, config.ProcessorPayloads); err != nil {
+			return emptyDesc, errors.Wrapf(err, "failed to get stream processor for %s", desc.MediaType)
+		}
+		if processor.MediaType() == ocispec.MediaTypeImageLayer {
+			break
+		}
+	}
+	defer processor.Close()
+
+	digester := digest.Canonical.Digester()
+	rc := &readCounter{
+		r: io.TeeReader(processor, digester.Hash()),
+	}
+
+	layer, parentLayerPaths, err := mountsToLayerAndParents(mounts)
+	if err != nil {
+		return emptyDesc, err
+	}
+
+	// TODO darrenstahlmsft: When this is done isolated, we should disable these.
+	// it currently cannot be disabled, unless we add ref counting. Since this is
+	// temporary, leaving it enabled is OK for now.
+	if err := winio.EnableProcessPrivileges([]string{winio.SeBackupPrivilege, winio.SeRestorePrivilege}); err != nil {
+		return emptyDesc, err
+	}
+
+	if _, err := archive.Apply(ctx, layer, rc, archive.WithParentLayers(parentLayerPaths), archive.AsWindowsContainerLayer()); err != nil {
+		return emptyDesc, err
+	}
+
+	// Read any trailing data
+	if _, err := io.Copy(ioutil.Discard, rc); err != nil {
+		return emptyDesc, err
+	}
+
+	return ocispec.Descriptor{
+		MediaType: ocispec.MediaTypeImageLayer,
+		Size:      rc.c,
+		Digest:    digester.Digest(),
+	}, nil
+}
+
+// Compare creates a diff between the given mounts and uploads the result
+// to the content store.
+func (s windowsDiff) Compare(ctx context.Context, lower, upper []mount.Mount, opts ...diff.Opt) (d ocispec.Descriptor, err error) {
+	return emptyDesc, errdefs.ErrNotImplemented
+}
+
+type readCounter struct {
+	r io.Reader
+	c int64
+}
+
+func (rc *readCounter) Read(p []byte) (n int, err error) {
+	n, err = rc.r.Read(p)
+	rc.c += int64(n)
+	return
+}
+
+func mountsToLayerAndParents(mounts []mount.Mount) (string, []string, error) {
+	if len(mounts) != 1 {
+		return "", nil, errors.Wrap(errdefs.ErrInvalidArgument, "number of mounts should always be 1 for Windows layers")
+	}
+	mnt := mounts[0]
+	if mnt.Type != "windows-layer" {
+		// This is a special case error. When this is received the diff service
+		// will attempt the next differ in the chain which for Windows is the
+		// lcow differ that we want.
+		return "", nil, errdefs.ErrNotImplemented
+	}
+
+	parentLayerPaths, err := mnt.GetParentPaths()
+	if err != nil {
+		return "", nil, err
+	}
+
+	return mnt.Source, parentLayerPaths, nil
+}

vendor/github.com/containerd/containerd/snapshots/windows/windows.go

@@ -0,0 +1,338 @@
+// +build windows
+
+/*
+   Copyright The containerd Authors.
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
+*/
+
+package windows
+
+import (
+	"context"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"strings"
+
+	winfs "github.com/Microsoft/go-winio/pkg/fs"
+	"github.com/Microsoft/go-winio/vhd"
+	"github.com/Microsoft/hcsshim"
+	"github.com/containerd/containerd/errdefs"
+	"github.com/containerd/containerd/log"
+	"github.com/containerd/containerd/mount"
+	"github.com/containerd/containerd/plugin"
+	"github.com/containerd/containerd/snapshots"
+	"github.com/containerd/containerd/snapshots/storage"
+	"github.com/containerd/continuity/fs"
+	"github.com/pkg/errors"
+)
+
+func init() {
+	plugin.Register(&plugin.Registration{
+		Type: plugin.SnapshotPlugin,
+		ID:   "windows",
+		InitFn: func(ic *plugin.InitContext) (interface{}, error) {
+			return NewSnapshotter(ic.Root)
+		},
+	})
+}
+
+type snapshotter struct {
+	root string
+	info hcsshim.DriverInfo
+	ms   *storage.MetaStore
+}
+
+// NewSnapshotter returns a new windows snapshotter
+func NewSnapshotter(root string) (snapshots.Snapshotter, error) {
+	fsType, err := winfs.GetFileSystemType(root)
+	if err != nil {
+		return nil, err
+	}
+	if strings.ToLower(fsType) != "ntfs" {
+		return nil, errors.Wrapf(errdefs.ErrInvalidArgument, "%s is not on an NTFS volume - only NTFS volumes are supported", root)
+	}
+
+	if err := os.MkdirAll(root, 0700); err != nil {
+		return nil, err
+	}
+	ms, err := storage.NewMetaStore(filepath.Join(root, "metadata.db"))
+	if err != nil {
+		return nil, err
+	}
+
+	if err := os.Mkdir(filepath.Join(root, "snapshots"), 0700); err != nil && !os.IsExist(err) {
+		return nil, err
+	}
+
+	return &snapshotter{
+		info: hcsshim.DriverInfo{
+			HomeDir: filepath.Join(root, "snapshots"),
+		},
+		root: root,
+		ms:   ms,
+	}, nil
+}
+
+// Stat returns the info for an active or committed snapshot by name or
+// key.
+//
+// Should be used for parent resolution, existence checks and to discern
+// the kind of snapshot.
+func (s *snapshotter) Stat(ctx context.Context, key string) (snapshots.Info, error) {
+	ctx, t, err := s.ms.TransactionContext(ctx, false)
+	if err != nil {
+		return snapshots.Info{}, err
+	}
+	defer t.Rollback()
+
+	_, info, _, err := storage.GetInfo(ctx, key)
+	return info, err
+}
+
+func (s *snapshotter) Update(ctx context.Context, info snapshots.Info, fieldpaths ...string) (snapshots.Info, error) {
+	ctx, t, err := s.ms.TransactionContext(ctx, true)
+	if err != nil {
+		return snapshots.Info{}, err
+	}
+	defer t.Rollback()
+
+	info, err = storage.UpdateInfo(ctx, info, fieldpaths...)
+	if err != nil {
+		return snapshots.Info{}, err
+	}
+
+	if err := t.Commit(); err != nil {
+		return snapshots.Info{}, err
+	}
+
+	return info, nil
+}
+
+func (s *snapshotter) Usage(ctx context.Context, key string) (snapshots.Usage, error) {
+	ctx, t, err := s.ms.TransactionContext(ctx, false)
+	if err != nil {
+		return snapshots.Usage{}, err
+	}
+	defer t.Rollback()
+
+	_, info, usage, err := storage.GetInfo(ctx, key)
+	if err != nil {
+		return snapshots.Usage{}, err
+	}
+
+	if info.Kind == snapshots.KindActive {
+		du := fs.Usage{
+			Size: 0,
+		}
+		usage = snapshots.Usage(du)
+	}
+
+	return usage, nil
+}
+
+func (s *snapshotter) Prepare(ctx context.Context, key, parent string, opts ...snapshots.Opt) ([]mount.Mount, error) {
+	return s.createSnapshot(ctx, snapshots.KindActive, key, parent, opts)
+}
+
+func (s *snapshotter) View(ctx context.Context, key, parent string, opts ...snapshots.Opt) ([]mount.Mount, error) {
+	return s.createSnapshot(ctx, snapshots.KindView, key, parent, opts)
+}
+
+// Mounts returns the mounts for the transaction identified by key. Can be
+// called on an read-write or readonly transaction.
+//
+// This can be used to recover mounts after calling View or Prepare.
+func (s *snapshotter) Mounts(ctx context.Context, key string) ([]mount.Mount, error) {
+	ctx, t, err := s.ms.TransactionContext(ctx, false)
+	if err != nil {
+		return nil, err
+	}
+	defer t.Rollback()
+
+	snapshot, err := storage.GetSnapshot(ctx, key)
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to get snapshot mount")
+	}
+	return s.mounts(snapshot), nil
+}
+
+func (s *snapshotter) Commit(ctx context.Context, name, key string, opts ...snapshots.Opt) error {
+	ctx, t, err := s.ms.TransactionContext(ctx, true)
+	if err != nil {
+		return err
+	}
+	defer t.Rollback()
+
+	usage := fs.Usage{
+		Size: 0,
+	}
+
+	if _, err = storage.CommitActive(ctx, key, name, snapshots.Usage(usage), opts...); err != nil {
+		return errors.Wrap(err, "failed to commit snapshot")
+	}
+
+	if err := t.Commit(); err != nil {
+		return err
+	}
+	return nil
+}
+
+// Remove abandons the transaction identified by key. All resources
+// associated with the key will be removed.
+func (s *snapshotter) Remove(ctx context.Context, key string) error {
+	ctx, t, err := s.ms.TransactionContext(ctx, true)
+	if err != nil {
+		return err
+	}
+	defer t.Rollback()
+
+	id, _, err := storage.Remove(ctx, key)
+	if err != nil {
+		return errors.Wrap(err, "failed to remove")
+	}
+
+	path := s.getSnapshotDir(id)
+	renamedID := "rm-" + id
+	renamed := s.getSnapshotDir(renamedID)
+	if err := os.Rename(path, renamed); err != nil && !os.IsNotExist(err) {
+		if !os.IsPermission(err) {
+			return err
+		}
+		// If permission denied, it's possible that the scratch is still mounted, an
+		// artifact after a hard daemon crash for example. Worth a shot to try detaching it
+		// before retrying the rename.
+		if detachErr := vhd.DetachVhd(filepath.Join(path, "sandbox.vhdx")); detachErr != nil {
+			return errors.Wrapf(err, "failed to detach VHD: %s", detachErr)
+		}
+		if renameErr := os.Rename(path, renamed); renameErr != nil && !os.IsNotExist(renameErr) {
+			return errors.Wrapf(err, "second rename attempt following detach failed: %s", renameErr)
+		}
+	}
+
+	if err := t.Commit(); err != nil {
+		if err1 := os.Rename(renamed, path); err1 != nil {
+			// May cause inconsistent data on disk
+			log.G(ctx).WithError(err1).WithField("path", renamed).Errorf("Failed to rename after failed commit")
+		}
+		return errors.Wrap(err, "failed to commit")
+	}
+
+	if err := hcsshim.DestroyLayer(s.info, renamedID); err != nil {
+		// Must be cleaned up, any "rm-*" could be removed if no active transactions
+		log.G(ctx).WithError(err).WithField("path", renamed).Warnf("Failed to remove root filesystem")
+	}
+
+	return nil
+}
+
+// Walk the committed snapshots.
+func (s *snapshotter) Walk(ctx context.Context, fn func(context.Context, snapshots.Info) error) error {
+	ctx, t, err := s.ms.TransactionContext(ctx, false)
+	if err != nil {
+		return err
+	}
+	defer t.Rollback()
+
+	return storage.WalkInfo(ctx, fn)
+}
+
+// Close closes the snapshotter
+func (s *snapshotter) Close() error {
+	return s.ms.Close()
+}
+
+func (s *snapshotter) mounts(sn storage.Snapshot) []mount.Mount {
+	var (
+		roFlag           string
+		source           string
+		parentLayerPaths []string
+	)
+
+	if sn.Kind == snapshots.KindView {
+		roFlag = "ro"
+	} else {
+		roFlag = "rw"
+	}
+
+	if len(sn.ParentIDs) == 0 || sn.Kind == snapshots.KindActive {
+		source = s.getSnapshotDir(sn.ID)
+		parentLayerPaths = s.parentIDsToParentPaths(sn.ParentIDs)
+	} else {
+		source = s.getSnapshotDir(sn.ParentIDs[0])
+		parentLayerPaths = s.parentIDsToParentPaths(sn.ParentIDs[1:])
+	}
+
+	// error is not checked here, as a string array will never fail to Marshal
+	parentLayersJSON, _ := json.Marshal(parentLayerPaths)
+	parentLayersOption := mount.ParentLayerPathsFlag + string(parentLayersJSON)
+
+	var mounts []mount.Mount
+	mounts = append(mounts, mount.Mount{
+		Source: source,
+		Type:   "windows-layer",
+		Options: []string{
+			roFlag,
+			parentLayersOption,
+		},
+	})
+
+	return mounts
+}
+
+func (s *snapshotter) getSnapshotDir(id string) string {
+	return filepath.Join(s.root, "snapshots", id)
+}
+
+func (s *snapshotter) createSnapshot(ctx context.Context, kind snapshots.Kind, key, parent string, opts []snapshots.Opt) ([]mount.Mount, error) {
+	ctx, t, err := s.ms.TransactionContext(ctx, true)
+	if err != nil {
+		return nil, err
+	}
+	defer t.Rollback()
+
+	newSnapshot, err := storage.CreateSnapshot(ctx, kind, key, parent, opts...)
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to create snapshot")
+	}
+
+	if kind == snapshots.KindActive {
+		parentLayerPaths := s.parentIDsToParentPaths(newSnapshot.ParentIDs)
+
+		var parentPath string
+		if len(parentLayerPaths) != 0 {
+			parentPath = parentLayerPaths[0]
+		}
+
+		if err := hcsshim.CreateSandboxLayer(s.info, newSnapshot.ID, parentPath, parentLayerPaths); err != nil {
+			return nil, errors.Wrap(err, "failed to create sandbox layer")
+		}
+
+		// TODO(darrenstahlmsft): Allow changing sandbox size
+	}
+
+	if err := t.Commit(); err != nil {
+		return nil, errors.Wrap(err, "commit failed")
+	}
+
+	return s.mounts(newSnapshot), nil
+}
+
+func (s *snapshotter) parentIDsToParentPaths(parentIDs []string) []string {
+	var parentLayerPaths []string
+	for _, ID := range parentIDs {
+		parentLayerPaths = append(parentLayerPaths, s.getSnapshotDir(ID))
+	}
+	return parentLayerPaths
+}

vendor/github.com/konsorten/go-windows-terminal-sequences/LICENSE

@@ -0,0 +1,9 @@
+(The MIT License)
+
+Copyright (c) 2017 marvin + konsorten GmbH (open-source@konsorten.de)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the 'Software'), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

vendor/github.com/konsorten/go-windows-terminal-sequences/README.md

@@ -0,0 +1,40 @@
+# Windows Terminal Sequences
+
+This library allow for enabling Windows terminal color support for Go.
+
+See [Console Virtual Terminal Sequences](https://docs.microsoft.com/en-us/windows/console/console-virtual-terminal-sequences) for details.
+
+## Usage
+
+```go
+import (
+	"syscall"
+	
+	sequences "github.com/konsorten/go-windows-terminal-sequences"
+)
+
+func main() {
+	sequences.EnableVirtualTerminalProcessing(syscall.Stdout, true)
+}
+
+```
+
+## Authors
+
+The tool is sponsored by the [marvin + konsorten GmbH](http://www.konsorten.de).
+
+We thank all the authors who provided code to this library:
+
+* Felix Kollmann
+
+## License
+
+(The MIT License)
+
+Copyright (c) 2018 marvin + konsorten GmbH (open-source@konsorten.de)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the 'Software'), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

vendor/github.com/konsorten/go-windows-terminal-sequences/go.mod

@@ -0,0 +1 @@
+module github.com/konsorten/go-windows-terminal-sequences

vendor/github.com/konsorten/go-windows-terminal-sequences/sequences.go

@@ -0,0 +1,36 @@
+// +build windows
+
+package sequences
+
+import (
+	"syscall"
+	"unsafe"
+)
+
+var (
+	kernel32Dll    *syscall.LazyDLL  = syscall.NewLazyDLL("Kernel32.dll")
+	setConsoleMode *syscall.LazyProc = kernel32Dll.NewProc("SetConsoleMode")
+)
+
+func EnableVirtualTerminalProcessing(stream syscall.Handle, enable bool) error {
+	const ENABLE_VIRTUAL_TERMINAL_PROCESSING uint32 = 0x4
+
+	var mode uint32
+	err := syscall.GetConsoleMode(syscall.Stdout, &mode)
+	if err != nil {
+		return err
+	}
+
+	if enable {
+		mode |= ENABLE_VIRTUAL_TERMINAL_PROCESSING
+	} else {
+		mode &^= ENABLE_VIRTUAL_TERMINAL_PROCESSING
+	}
+
+	ret, _, err := setConsoleMode.Call(uintptr(unsafe.Pointer(stream)), uintptr(mode))
+	if ret == 0 {
+		return err
+	}
+
+	return nil
+}