Turn off 'noexec' option by default for named volumes

We previously enforced this for security reasons, but as Dan has
explained on several occasions, it's not very valuable there
(it's trivially easy to bypass) and it does seriously annoy folks
trying to use named volumes. Flip the default from 'on' to 'off'.

This is a backport from the master branch to v1.9 branch.

Signed-off-by: Matthew Heon <[email protected]>

pkg/util/mountOpts_linux.go

@@ -7,7 +7,7 @@ import (
 )
 
 func getDefaultMountOptions(path string) (defaultMountOptions, error) {
-	opts := defaultMountOptions{true, true, true}
+	opts := defaultMountOptions{false, true, true}
 	if path == "" {
 		return opts, nil
 	}

test/e2e/create_test.go

@@ -205,7 +205,7 @@ var _ = Describe("Podman create", func() {
 		session = podmanTest.Podman([]string{"logs", "test_tmpfs"})
 		session.WaitWithDefaultTimeout()
 		Expect(session.ExitCode()).To(Equal(0))
-		Expect(session.OutputToString()).To(ContainSubstring("/create/test rw,nosuid,nodev,noexec,relatime - tmpfs"))
+		Expect(session.OutputToString()).To(ContainSubstring("/create/test rw,nosuid,nodev,relatime - tmpfs"))
 	})
 
 	It("podman create --pod automatically", func() {

test/e2e/run_volume_test.go

@@ -117,7 +117,7 @@ var _ = Describe("Podman run with volumes", func() {
 		session = podmanTest.Podman([]string{"run", "--rm", "--mount", "type=tmpfs,target=" + dest, ALPINE, "grep", dest, "/proc/self/mountinfo"})
 		session.WaitWithDefaultTimeout()
 		Expect(session.ExitCode()).To(Equal(0))
-		Expect(session.OutputToString()).To(ContainSubstring(dest + " rw,nosuid,nodev,noexec,relatime - tmpfs"))
+		Expect(session.OutputToString()).To(ContainSubstring(dest + " rw,nosuid,nodev,relatime - tmpfs"))
 
 		session = podmanTest.Podman([]string{"run", "--rm", "--mount", "type=tmpfs,target=/etc/ssl,tmpcopyup", ALPINE, "ls", "/etc/ssl"})
 		session.WaitWithDefaultTimeout()

test/system/160-volumes.bats

@@ -0,0 +1,164 @@
+#!/usr/bin/env bats   -*- bats -*-
+#
+# podman volume-related tests
+#
+
+load helpers
+
+function setup() {
+    basic_setup
+
+    run_podman '?' volume rm -a
+}
+
+function teardown() {
+    run_podman '?' rm -a --volumes
+    run_podman '?' volume rm -a -f
+
+    basic_teardown
+}
+
+
+# Simple volume tests: share files between host and container
+@test "podman run --volumes : basic" {
+    skip_if_remote "volumes cannot be shared across hosts"
+
+    # Create three temporary directories
+    vol1=${PODMAN_TMPDIR}/v1_$(random_string)
+    vol2=${PODMAN_TMPDIR}/v2_$(random_string)
+    vol3=${PODMAN_TMPDIR}/v3_$(random_string)
+    mkdir $vol1 $vol2 $vol3
+
+    # In each directory, write a random string to a file
+    echo $(random_string) >$vol1/file1_in
+    echo $(random_string) >$vol2/file2_in
+    echo $(random_string) >$vol3/file3_in
+
+    # Run 'cat' on each file, and compare against local files. Mix -v / --volume
+    # flags, and specify them out of order just for grins. The shell wildcard
+    # expansion must sort vol1/2/3 lexically regardless.
+    v_opts="-v $vol1:/vol1:z --volume $vol3:/vol3:z -v $vol2:/vol2:z"
+    run_podman run --rm $v_opts $IMAGE sh -c "cat /vol?/file?_in"
+
+    for i in 1 2 3; do
+        eval voldir=\$vol${i}
+        is "${lines[$(($i - 1))]}" "$(< $voldir/file${i}_in)" \
+           "contents of /vol${i}/file${i}_in"
+    done
+
+    # Confirm that container sees vol1 as a mount point
+    run_podman run --rm $v_opts $IMAGE mount
+    is "$output" ".* on /vol1 type .*" "'mount' in container lists vol1"
+
+    # Have the container do write operations, confirm them on host
+    out1=$(random_string)
+    run_podman run --rm $v_opts $IMAGE sh -c "echo $out1 >/vol1/file1_out;
+                                              cp /vol2/file2_in /vol3/file3_out"
+    is "$(<$vol1/file1_out)" "$out1"              "contents of /vol1/file1_out"
+    is "$(<$vol3/file3_out)" "$(<$vol2/file2_in)" "contents of /vol3/file3_out"
+
+    # Writing to read-only volumes: not allowed
+    run_podman 1 run --rm -v $vol1:/vol1ro:z,ro $IMAGE sh -c "touch /vol1ro/abc"
+    is "$output" ".*Read-only file system"  "touch on read-only volume"
+}
+
+
+# Running scripts (executables) from a volume
+@test "podman volume: exec/noexec" {
+    myvolume=myvol$(random_string)
+
+    run_podman volume create $myvolume
+    is "$output" "$myvolume" "output from volume create"
+
+    run_podman volume inspect --format '{{.Mountpoint}}' $myvolume
+    mountpoint="$output"
+
+    # Create a script, make it runnable
+    rand=$(random_string)
+    cat >$mountpoint/myscript <<EOF
+#!/bin/sh
+echo "got here -$rand-"
+EOF
+    chmod 755 $mountpoint/myscript
+
+    # By default, volumes are mounted exec, but we have manually added the
+    # noexec option. This should fail.
+    # ARGH. Unfortunately, runc (used for cgroups v1) produces a different error
+    local expect_rc=126
+    local expect_msg='.* OCI runtime permission denied.*'
+    run_podman info --format '{{ .Host.OCIRuntime.Path }}'
+    if expr "$output" : ".*/runc"; then
+        expect_rc=1
+        expect_msg='.* exec user process caused.*permission denied'
+    fi
+
+    run_podman ${expect_rc} run --rm --volume $myvolume:/vol:noexec,z $IMAGE /vol/myscript
+    is "$output" "$expect_msg" "run on volume, noexec"
+
+    # With the default, it should pass
+    run_podman run --rm -v $myvolume:/vol:z $IMAGE /vol/myscript
+    is "$output" "got here -$rand-" "script in volume is runnable with default (exec)"
+
+    # Clean up
+    run_podman volume rm $myvolume
+}
+
+
+# Confirm that container sees the correct id
+@test "podman volume with --userns=keep-id" {
+    is_rootless || skip "only meaningful when run rootless"
+
+    myvoldir=${PODMAN_TMPDIR}/volume_$(random_string)
+    mkdir $myvoldir
+    touch $myvoldir/myfile
+
+    # With keep-id
+    run_podman run --rm -v $myvoldir:/vol:z --userns=keep-id $IMAGE \
+               stat -c "%u:%s" /vol/myfile
+    is "$output" "$(id -u):0" "with keep-id: stat(file in container) == my uid"
+
+    # Without
+    run_podman run --rm -v $myvoldir:/vol:z $IMAGE \
+               stat -c "%u:%s" /vol/myfile
+    is "$output" "0:0" "w/o keep-id: stat(file in container) == root"
+}
+
+
+# 'volume prune' identifies and cleans up unused volumes
+@test "podman volume prune" {
+    # Create four named volumes
+    local -a v=()
+    for i in 1 2 3 4;do
+        vol=myvol${i}$(random_string)
+        v[$i]=$vol
+        run_podman volume create $vol
+    done
+
+    # Run two containers: one mounting v1, one mounting v2 & v3
+    run_podman run --name c1 --volume ${v[1]}:/vol1 $IMAGE date
+    run_podman run --name c2 --volume ${v[2]}:/vol2 -v ${v[3]}:/vol3 \
+               $IMAGE date
+
+    # prune should remove v4
+    run_podman volume prune --force
+    is "$output" "${v[4]}" "volume prune, with 1, 2, 3 in use, deletes only 4"
+
+    # Remove the container using v2 and v3. Prune should now remove those.
+    # The 'echo sort' is to get the output sorted and in one line.
+    run_podman rm c2
+    run_podman volume prune --force
+    is "$(echo $(sort <<<$output))" "${v[2]} ${v[3]}" \
+       "volume prune, after rm c2, deletes volumes 2 and 3"
+
+    # Remove the final container. Prune should now remove v1.
+    run_podman rm c1
+    run_podman volume prune --force
+    is "$output"  "${v[1]}" "volume prune, after rm c2 & c1, deletes volume 1"
+
+    # Further prunes are NOPs
+    run_podman volume prune --force
+    is "$output"  "" "no more volumes to prune"
+}
+
+
+# vim: filetype=sh