allow running without securityContext.procMount=Unmasked (BuildKit v0.4)

Dockerfile

@@ -1,5 +1,4 @@
-ARG RUNC_VERSION=9f9c96235cc97674e935002fc3d78361b696a69e
-FROM golang:1.10-alpine AS gobuild-base
+FROM golang:1.11-alpine AS gobuild-base
 RUN apk add --no-cache \
 	bash \
 	build-base \
@@ -9,13 +8,6 @@ RUN apk add --no-cache \
 	linux-headers \
 	make
 
-FROM gobuild-base AS runc
-ARG RUNC_VERSION
-RUN git clone https://github.com/opencontainers/runc.git "$GOPATH/src/github.com/opencontainers/runc" \
-	&& cd "$GOPATH/src/github.com/opencontainers/runc" \
-	&& make static BUILDTAGS="seccomp" EXTRA_FLAGS="-buildmode pie" EXTRA_LDFLAGS="-extldflags \\\"-fno-PIC -static\\\"" \
-	&& mv runc /usr/bin/runc
-
 FROM gobuild-base AS img
 WORKDIR /go/src/github.com/genuinetools/img
 COPY . .
@@ -43,18 +35,19 @@ FROM alpine:3.8 AS base
 MAINTAINER Jessica Frazelle <jess@linux.com>
 RUN apk add --no-cache git
 COPY --from=img /usr/bin/img /usr/bin/img
-COPY --from=runc /usr/bin/runc /usr/bin/runc
 COPY --from=idmap /usr/bin/newuidmap /usr/bin/newuidmap
 COPY --from=idmap /usr/bin/newgidmap /usr/bin/newgidmap
 RUN chmod u+s /usr/bin/newuidmap /usr/bin/newgidmap \
   && adduser -D -u 1000 user \
   && mkdir -p /run/user/1000 \
   && chown -R user /run/user/1000 /home/user \
   && echo user:100000:65536 | tee /etc/subuid | tee /etc/subgid
-# As of v3.8.1, Alpine does not set SUID bit on the busybox version of /bin/su.
-# However, future version may set SUID bit on /bin/su.
-# We lock the root account so as to disable su completely.
-RUN passwd -l root
+# In previous version of `alpine:3.8`, the root was not locked and su-able
+# without any password when SUID bit is set on `/bin/su`.
+#
+# As of 3/15/2019, the root is locked by default, but we expliciltly lock the
+# root just in case.
+RUN passwd -l root || true
 
 FROM base AS debug
 RUN apk add --no-cache bash strace

Makefile

@@ -13,8 +13,10 @@ include basic.mk
 prebuild: runc
 
 RUNCBUILDDIR=$(BUILDDIR)/src/github.com/opencontainers/runc
+RUNCCOMMIT=7cb3cde1f49eae53fb8fff5012c0750a64eb928b
 $(RUNCBUILDDIR):
-	git clone --depth 1 https://github.com/opencontainers/runc.git "$@"
+	git clone https://github.com/opencontainers/runc.git "$@"
+	( cd "$@" ; git checkout $(RUNCCOMMIT) )
 
 $(RUNCBUILDDIR)/runc: $(RUNCBUILDDIR)
 	GOPATH=$(BUILDDIR) $(MAKE) -C "$(RUNCBUILDDIR)" static BUILDTAGS="seccomp apparmor"

README.md

@@ -148,18 +148,32 @@ $ sudo emerge -a app-emulation/img
 
 #### Running with Docker
 
-This currently **does not work** without `--privileged`, to track the progress of making this work in a container without the `--privileged` flag see [upstream patches](#upstream-patches).
+Docker image `r.j3ss.co/img` is configured to be executed as an unprivileged user with UID 1000 and it does not need `--privileged` since `img` v0.6.0.
 
 ```console
 $ docker run --rm -it \
     --name img \
     --volume $(pwd):/home/user/src:ro \ # for the build context and dockerfile, can be read-only since we won't modify it
     --workdir /home/user/src \ # set the builder working directory
     --volume "${HOME}/.docker:/root/.docker:ro" \ # for credentials to push to docker hub or a registry
-    --privileged \
+    --security-opt seccomp=unconfined --security-opt apparmor=unconfined \ # required by runc
     r.j3ss.co/img build -t user/myimage .
 ```
 
+To enable PID namespace isolation (which disallows build containers to `kill(2)` the `img` process), you need to specify
+`--privileged` so that build containers can mount `/proc` with unshared PID namespaces.
+Note that even with `--privileged`, `img` works as an unprivileged user with UID 1000.
+
+See [docker/cli patch](#upstream-patches) for how to allow mounting `/proc` without `--privileged`.
+
+### Running with Kubernetes
+
+Since `img` v0.6.0, you don't need to specify any `securityContext` for running `img` as a Kubernetes container.
+
+However, to enable PID namespace isolation, you need to set `securityContext.procMount` to `Unmasked` (or simply set
+`securityContext.privileged` to `true`).
+`securityContext.procMount` is available since Kubernetes 1.12 with Docker 18.06/containerd 1.2/CRI-O 1.12.
+
 ## Usage
 
 Make sure you have user namespace support enabled. On some distros (Debian and
@@ -608,4 +622,4 @@ be unprivileged.
 
 A lot of this is based on the work of [moby/buildkit](https://github.com/moby/buildkit). 
 Thanks [@tonistiigi](https://github.com/tonistiigi) and
-[@AkihiroSuda](https://github.com/AkihiroSuda)!
+[@AkihiroSuda](https://github.com/AkihiroSuda)!

client/pull.go

@@ -15,6 +15,10 @@ import (
 
 // Pull retrieves an image from a remote registry.
 func (c *Client) Pull(ctx context.Context, image string) (*ListedImage, error) {
+	sm, err := c.getSessionManager()
+	if err != nil {
+		return nil, err
+	}
 	// Parse the image name and tag.
 	named, err := reference.ParseNormalizedNamed(image)
 	if err != nil {
@@ -46,18 +50,17 @@ func (c *Client) Pull(ctx context.Context, image string) (*ListedImage, error) {
 
 	// Create the source for the pull.
 	srcOpt := containerimage.SourceOpt{
-		SessionManager: opt.SessionManager,
-		Snapshotter:    opt.Snapshotter,
-		ContentStore:   opt.ContentStore,
-		Applier:        opt.Applier,
-		CacheAccessor:  cm,
-		ImageStore:     opt.ImageStore,
+		Snapshotter:   opt.Snapshotter,
+		ContentStore:  opt.ContentStore,
+		Applier:       opt.Applier,
+		CacheAccessor: cm,
+		ImageStore:    opt.ImageStore,
 	}
 	src, err := containerimage.NewSource(srcOpt)
 	if err != nil {
 		return nil, err
 	}
-	s, err := src.Resolve(ctx, identifier)
+	s, err := src.Resolve(ctx, identifier, sm)
 	if err != nil {
 		return nil, err
 	}
@@ -76,9 +79,8 @@ func (c *Client) Pull(ctx context.Context, image string) (*ListedImage, error) {
 		return nil, err
 	}
 	expOpt := imageexporter.Opt{
-		SessionManager: opt.SessionManager,
-		Images:         opt.ImageStore,
-		ImageWriter:    iw,
+		Images:      opt.ImageStore,
+		ImageWriter: iw,
 	}
 	exp, err := imageexporter.New(expOpt)
 	if err != nil {

client/push.go

@@ -30,5 +30,9 @@ func (c *Client) Push(ctx context.Context, image string, insecure bool) error {
 		return fmt.Errorf("getting image %q failed: %v", image, err)
 	}
 
-	return push.Push(ctx, opt.SessionManager, opt.ContentStore, imgObj.Target.Digest, image, insecure, opt.ResolveOptionsFunc)
+	sm, err := c.getSessionManager()
+	if err != nil {
+		return err
+	}
+	return push.Push(ctx, sm, opt.ContentStore, imgObj.Target.Digest, image, insecure, opt.ResolveOptionsFunc)
 }

client/workeropt.go

@@ -3,7 +3,9 @@ package client
 import (
 	"context"
 	"fmt"
+	"os/exec"
 	"path/filepath"
+	"syscall"
 	"time"
 
 	"github.com/containerd/containerd/content/local"
@@ -17,6 +19,7 @@ import (
 	"github.com/genuinetools/img/types"
 	"github.com/moby/buildkit/cache/metadata"
 	"github.com/moby/buildkit/executor"
+	executoroci "github.com/moby/buildkit/executor/oci"
 	"github.com/moby/buildkit/executor/runcexecutor"
 	containerdsnapshot "github.com/moby/buildkit/snapshot/containerd"
 	"github.com/moby/buildkit/util/network"
@@ -31,10 +34,6 @@ import (
 
 // createWorkerOpt creates a base.WorkerOpt to be used for a new worker.
 func (c *Client) createWorkerOpt(withExecutor bool) (opt base.WorkerOpt, err error) {
-	sm, err := c.getSessionManager()
-	if err != nil {
-		return opt, err
-	}
 	// Create the metadata store.
 	md, err := metadata.NewStore(filepath.Join(c.root, "metadata.db"))
 	if err != nil {
@@ -65,8 +64,9 @@ func (c *Client) createWorkerOpt(withExecutor bool) (opt base.WorkerOpt, err err
 	var exe executor.Executor
 	if withExecutor {
 		exeOpt := runcexecutor.Opt{
-			Root:     filepath.Join(c.root, "executor"),
-			Rootless: unprivileged,
+			Root:        filepath.Join(c.root, "executor"),
+			Rootless:    unprivileged,
+			ProcessMode: processMode(),
 		}
 		exe, err = runcexecutor.New(exeOpt, network.Default())
 		if err != nil {
@@ -121,7 +121,6 @@ func (c *Client) createWorkerOpt(withExecutor bool) (opt base.WorkerOpt, err err
 	opt = base.WorkerOpt{
 		ID:                 id,
 		Labels:             xlabels,
-		SessionManager:     sm,
 		MetadataStore:      md,
 		Executor:           exe,
 		Snapshotter:        containerdsnapshot.NewSnapshotter(mdb.Snapshotter(c.backend), contentStore, md, "buildkit", gc),
@@ -135,3 +134,18 @@ func (c *Client) createWorkerOpt(withExecutor bool) (opt base.WorkerOpt, err err
 
 	return opt, err
 }
+
+func processMode() executoroci.ProcessMode {
+	mountArgs := []string{"-t", "proc", "none", "/proc"}
+	cmd := exec.Command("mount", mountArgs...)
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		Pdeathsig:    syscall.SIGKILL,
+		Cloneflags:   syscall.CLONE_NEWPID,
+		Unshareflags: syscall.CLONE_NEWNS,
+	}
+	if b, err := cmd.CombinedOutput(); err != nil {
+		logrus.Warnf("Process sandbox is not available, consider unmasking procfs: %v", string(b))
+		return executoroci.NoProcessSandbox
+	}
+	return executoroci.ProcessSandbox
+}

contrib/e2e-dockerfiles-build-test.sh

@@ -24,7 +24,17 @@ build_and_push(){
 
 	echo "[In Container] Building ${REPO_URL}/${base}:${suite} for context ${build_dir}"
 	# Do the same but in a docker container.
-	docker run --rm --name img --volume $(pwd):/home/user/src:ro --workdir /home/user/src --privileged r.j3ss.co/img build -t ${REPO_URL}/${base}:${suite} ${build_dir}
+    name=${REPO_URL}-${base}-${suite}
+    set +e
+	docker run --name $name --volume $(pwd):/home/user/src:ro --workdir /home/user/src --privileged r.j3ss.co/img build --no-console -t ${REPO_URL}/${base}:${suite} ${build_dir}  > /dev/null 2>&1
+    status=$?
+    set -e
+    if [[ $status != 0 ]]; then
+        docker logs $name
+        docker rm -f $name
+        exit $status
+    fi
+    docker rm -f $name
 
 	# on successful build, push the image
 	echo "                       ---                                   "

go.mod

@@ -3,51 +3,41 @@ module github.com/genuinetools/img
 replace github.com/hashicorp/go-immutable-radix => github.com/tonistiigi/go-immutable-radix v0.0.0-20170803185627-826af9ccf0fe
 
 require (
-	github.com/BurntSushi/toml v0.3.1 // indirect
-	github.com/Microsoft/hcsshim v0.7.4 // indirect
-	github.com/containerd/cgroups v0.0.0-20180917172123-5017d4e9a9cf // indirect
+	github.com/checkpoint-restore/go-criu v0.0.0-20190109184317-bdb7599cd87b
 	github.com/containerd/console v0.0.0-20180822173158-c12b1e7919c1
-	github.com/containerd/containerd v0.0.0-20180925133938-3bc4ba271e1c
-	github.com/containerd/cri v1.11.1 // indirect
-	github.com/containerd/fifo v0.0.0-20180307165137-3d5202aec260 // indirect
+	github.com/containerd/containerd v1.3.0-0.20190212172151-f5b0fa220df8
 	github.com/containerd/go-runc v0.0.0-20180907222934-5a6d9f37cfa3
-	github.com/containerd/typeurl v0.0.0-20180627222232-a93fcdb778cd // indirect
-	github.com/coreos/go-systemd v0.0.0-20180828140353-eee3db372b31 // indirect
+	github.com/coreos/go-systemd v0.0.0-20181031085051-9002847aa142
+	github.com/coreos/pkg v0.0.0-20180928190104-399ea9e2e55f // indirect
 	github.com/cyphar/filepath-securejoin v0.2.2
-	github.com/docker/cli v0.0.0-20180920165730-54c19e67f69c
-	github.com/docker/distribution v0.0.0-20180920194744-16128bbac47f
-	github.com/docker/docker v0.0.0-20180924202107-a9c061deec0f
-	github.com/docker/docker-ce v17.12.1-ce-rc2+incompatible // indirect
+	github.com/docker/cli v0.0.0-20190131223713-234462756460
+	github.com/docker/distribution v2.7.1-0.20190205005809-0d3efadf0154+incompatible
+	github.com/docker/docker v0.7.3-0.20190315021241-05e7d000f2fc
 	github.com/docker/go-connections v0.4.0 // indirect
-	github.com/docker/go-events v0.0.0-20170721190031-9461782956ad // indirect
 	github.com/docker/go-units v0.3.3
-	github.com/docker/libnetwork v0.0.0-20171208192558-822e5b59d346 // indirect
 	github.com/genuinetools/pkg v0.0.0-20180910213200-1c141f661797
 	github.com/genuinetools/reg v0.16.0
-	github.com/godbus/dbus v4.1.0+incompatible // indirect
+	github.com/godbus/dbus v4.1.0+incompatible
 	github.com/gogo/googleapis v1.1.0 // indirect
-	github.com/google/shlex v0.0.0-20150127133951-6f45313302b9 // indirect
-	github.com/grpc-ecosystem/go-grpc-prometheus v1.2.0 // indirect
-	github.com/grpc-ecosystem/grpc-opentracing v0.0.0-20180507213350-8e809c8a8645 // indirect
-	github.com/hashicorp/go-immutable-radix v1.0.0 // indirect
+	github.com/golang/protobuf v1.2.0
 	github.com/hashicorp/golang-lru v0.5.0 // indirect
-	github.com/hashicorp/uuid v0.0.0-20160311170451-ebb0a03e909c // indirect
 	github.com/mitchellh/hashstructure v1.0.0 // indirect
-	github.com/moby/buildkit v0.3.2
-	github.com/morikuni/aec v0.0.0-20170113033406-39771216ff4c // indirect
+	github.com/moby/buildkit v0.4.0
+	github.com/mrunalp/fileutils v0.0.0-20171103030105-7d4729fb3618
 	github.com/opencontainers/image-spec v1.0.1
 	github.com/opencontainers/runc v1.0.0-rc2.0.20181113215238-10d38b660a77
-	github.com/opencontainers/runtime-spec v1.0.1 // indirect
+	github.com/opencontainers/runtime-spec v1.0.1
+	github.com/opencontainers/selinux v1.0.0
 	github.com/opentracing-contrib/go-stdlib v0.0.0-20180702182724-07a764486eb1 // indirect
 	github.com/opentracing/opentracing-go v1.0.2 // indirect
-	github.com/pkg/errors v0.8.0
+	github.com/pkg/errors v0.8.1
+	github.com/seccomp/libseccomp-golang v0.9.0
 	github.com/sirupsen/logrus v1.0.6
-	github.com/syndtr/gocapability v0.0.0-20180916011248-d98352740cb2 // indirect
-	github.com/tonistiigi/fsutil v0.0.0-20181011223333-2862f6bc5ac9 // indirect
-	github.com/tonistiigi/units v0.0.0-20180711220420-6950e57a87ea // indirect
-	github.com/vishvananda/netlink v1.0.0 // indirect
-	github.com/vishvananda/netns v0.0.0-20180720170159-13995c7128cc // indirect
+	github.com/syndtr/gocapability v0.0.0-20180916011248-d98352740cb2
+	github.com/urfave/cli v0.0.0-20171014202726-7bc6a0acffa5
+	github.com/vishvananda/netlink v1.0.0
 	go.etcd.io/bbolt v1.3.1-etcd.8
 	golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f
+	golang.org/x/sys v0.0.0-20180925112736-b09afc3d579e
 	google.golang.org/grpc v1.15.0
 )