update Dockerfile, with Giuseppe's newuidmap

hack/test.sh

@@ -3,6 +3,21 @@ set -eu -o pipefail -x
 
 cd $(dirname $0)/..
 
+ref=rootless-containers/rootlesskit:test
 # build -q but keep printing something (....) so as to avoid Travis timeout
-(while sleep 60; do echo -n .; done) & docker build -q -t rootless-containers/rootlesskit:test -f ./hack/test/Dockerfile . ; kill %1
-docker run -it --rm --privileged rootless-containers/rootlesskit:test
+(while sleep 60; do echo -n .; done) & docker build -q -t ${ref} -f ./hack/test/Dockerfile . ; kill %1
+
+# TODO: add `--security-opt unconfined-paths=/sys` when https://github.com/docker/cli/pull/1347 gets merged.
+# See https://github.com/rootless-containers/rootlesskit/pull/23 for the context.
+#
+# NOTE: bind-mounting /dev/net/tun is needed because we cannot mknod the tun device, at least with kernel < 4.18
+#
+# FIXME: --privilege is needed because the test suite compares rootlesskit with rootful `ip netns`.
+# TODO: split the rootful test to separate container and remove --privilege
+docker run \
+       --rm \
+       --security-opt seccomp=unconfined \
+       --security-opt apparmor=unconfined \
+       -v /dev/net/tun:/dev/net/tun \
+       --privileged \
+       ${ref}

hack/test/Dockerfile

@@ -1,19 +1,29 @@
-FROM golang:1.10 AS rootlesskit
+FROM golang:1.11 AS rootlesskit
 ADD . /go/src/github.com/rootless-containers/rootlesskit
 RUN go build -o /rootlesskit github.com/rootless-containers/rootlesskit/cmd/rootlesskit
 
-FROM ubuntu as build-c
+FROM ubuntu:18.04 as build-c
 RUN apt update && apt install -y git make gcc automake autotools-dev libtool
 
+# As of October 2018, the upstream shadow newuidmap/newgidmap requires `docker run --cap-add sys_admin`.
+# We can use https://github.com/shadow-maint/shadow/pull/132 for eliminating unneeded dependency on CAP_SYS_ADMIN.
+FROM build-c as idmap
+RUN apt update && apt install -y byacc gettext autopoint libcap-dev libxslt1-dev
+RUN ( git clone -b no-cap-sys-admin https://github.com/giuseppe/shadow.git /shadow && cd /shadow )
+WORKDIR /shadow
+RUN ./autogen.sh --disable-nls --disable-man --without-audit --without-selinux --without-acl --without-attr --without-tcb --without-nscd \
+&& make \
+&& cp src/newuidmap src/newgidmap /usr/bin
+
 FROM build-c AS slirp4netns
-ARG SLIRP4NETNS_COMMIT=0037042e2facc8818d0a254e320a739648283f2e
+ARG SLIRP4NETNS_COMMIT=39e87c0e66345edf7fd6e0bd1f61aa842617e757
 RUN git clone https://github.com/rootless-containers/slirp4netns.git /slirp4netns && \
   cd /slirp4netns && git checkout ${SLIRP4NETNS_COMMIT} && \
   ./autogen.sh && ./configure && make
 
-FROM ubuntu AS vpnkit
+FROM ubuntu:18.04 AS vpnkit
 RUN apt update && apt install -y ocaml opam
-ARG VPNKIT_COMMIT=bbedcf42d9e6ba0afba65aca18dd5fc087192b13
+ARG VPNKIT_COMMIT=7d18ec48982340f49461379d718a4538f3090e13
 RUN git clone https://github.com/moby/vpnkit.git /vpnkit && \
   cd /vpnkit && git checkout ${VPNKIT_COMMIT} && \
   make
@@ -36,24 +46,33 @@ RUN git clone https://github.com/rd235/vdeplug_slirp.git /vdeplug_slirp && \
   cd /vdeplug_slirp && git checkout ${VDEPLUGSLIRP_COMMIT} && \
   autoreconf -if && ./configure && make && make install
 
-FROM ubuntu AS test
-# busybox is only for debugging purpose
-# sudo is only for rootful veth benchmark (for comparison)
-RUN apt update && apt install -y iperf3 uidmap iproute2 busybox sudo
-RUN useradd --create-home --home-dir /home/user --uid 1000 user \
+FROM ubuntu:18.04 AS test-base
+# iproute2: for `ip` command that rootlesskit needs to exec
+# iperf3: only for benchmark purpose
+# busybox: only for debugging purpose
+# sudo: only for rootful veth benchmark (for comparison)
+RUN apt update && apt install -y iproute2 iperf3 busybox sudo
+COPY --from=idmap /usr/bin/newuidmap /usr/bin/newuidmap
+COPY --from=idmap /usr/bin/newgidmap /usr/bin/newgidmap
+COPY --from=rootlesskit /rootlesskit /home/user/bin/
+RUN chmod u+s /usr/bin/newuidmap /usr/bin/newgidmap \
+  && useradd --create-home --home-dir /home/user --uid 1000 user \
   && mkdir -p /run/user/1000 \
+  && chown -R user:user /run/user/1000 /home/user \
   && echo "user ALL=(ALL) NOPASSWD:ALL" > /etc/sudoers.d/user
-COPY --from=rootlesskit /rootlesskit /home/user/bin/
-COPY --from=slirp4netns /slirp4netns/slirp4netns /home/user/bin/
-COPY --from=vpnkit /vpnkit/vpnkit.exe /home/user/bin/vpnkit
-COPY --from=vdeplug_slirp /usr/local/bin/* /home/user/bin/
-COPY --from=vdeplug_slirp /usr/local/lib/* /home/user/lib/
-ADD ./hack/test/docker-entrypoint.sh /home/user/bin/
-RUN chown -R user:user /run/user/1000 /home/user
 USER user
 ENV HOME /home/user
 ENV USER user
 ENV XDG_RUNTIME_DIR=/run/user/1000
 ENV PATH /home/user/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
 ENV LD_LIBRARY_PATH=/home/user/lib
+
+FROM test-base AS test-light
+COPY --from=slirp4netns /slirp4netns/slirp4netns /home/user/bin/
+
+FROM test-light AS test-full
+COPY --from=vpnkit /vpnkit/vpnkit.exe /home/user/bin/vpnkit
+COPY --from=vdeplug_slirp /usr/local/bin/* /home/user/bin/
+COPY --from=vdeplug_slirp /usr/local/lib/* /home/user/lib/
+ADD ./hack/test/docker-entrypoint.sh /home/user/bin/
 ENTRYPOINT ["/home/user/bin/docker-entrypoint.sh"]

pkg/child/child.go

@@ -83,11 +83,25 @@ func mountSysfs() error {
 		return errors.Wrap(err, "creating a directory under /tmp")
 	}
 	defer os.RemoveAll(tmp)
-	cmds := [][]string{
-		{"mount", "--rbind", "/sys/fs/cgroup", tmp},
-		{"mount", "-t", "sysfs", "none", "/sys"},
-		{"mount", "-n", "--move", tmp, "/sys/fs/cgroup"},
+	cmds := [][]string{{"mount", "--rbind", "/sys/fs/cgroup", tmp}}
+	if err := common.Execs(os.Stderr, os.Environ(), cmds); err != nil {
+		return errors.Wrapf(err, "executing %v", cmds)
+	}
+	cmds = [][]string{{"mount", "-t", "sysfs", "none", "/sys"}}
+	if err := common.Execs(os.Stderr, os.Environ(), cmds); err != nil {
+		// when the sysfs in the parent namespace is RO,
+		// we can't mount RW sysfs even in the child namespace.
+		// https://github.com/rootless-containers/rootlesskit/pull/23#issuecomment-429292632
+		// https://github.com/torvalds/linux/blob/9f203e2f2f065cd74553e6474f0ae3675f39fb0f/fs/namespace.c#L3326-L3328
+		cmdsRo := [][]string{{"mount", "-t", "sysfs", "-o", "ro", "none", "/sys"}}
+		logrus.Warnf("failed to mount sysfs (%v), falling back to read-only mount (%v): %v",
+			cmds, cmdsRo, err)
+		if err := common.Execs(os.Stderr, os.Environ(), cmdsRo); err != nil {
+			// when /sys/firmware is masked, even RO sysfs can't be mounted
+			logrus.Warnf("failed to mount sysfs (%v): %v", cmdsRo, err)
+		}
 	}
+	cmds = [][]string{{"mount", "-n", "--move", tmp, "/sys/fs/cgroup"}}
 	if err := common.Execs(os.Stderr, os.Environ(), cmds); err != nil {
 		return errors.Wrapf(err, "executing %v", cmds)
 	}