core: Private*/Protect* options with RootDirectory

When a service is chrooted with the option RootDirectory=/opt/..., then
the options PrivateDevices, PrivateTmp, ProtectHome, ProtectSystem must
mount the directories under $RootDirectory/{dev,tmp,home,usr,boot}.

The test-ns tool can test setup_namespace() with and without chroot:
 $ sudo TEST_NS_PROJECTS=/home/lennart/projects ./test-ns
 $ sudo TEST_NS_CHROOT=/home/alban/debian-tree TEST_NS_PROJECTS=/home/alban/debian-tree/home/alban/Documents ./test-ns

4 files changed, 100 insertions, 15 deletions

diff --git a/src/core/execute.c b/src/core/execute.c
index 0cca4813a8..97498b23d7 100644
--- a/src/core/execute.c
+++ b/src/core/execute.c
@@ -1307,6 +1307,7 @@ static int exec_child(
         uid_t uid = UID_INVALID;
         gid_t gid = GID_INVALID;
         int i, r;
+        bool needs_mount_namespace;
 
         assert(unit);
         assert(command);
@@ -1585,7 +1586,9 @@ static int exec_child(
                 }
         }
 
-        if (exec_needs_mount_namespace(context, params, runtime)) {
+        needs_mount_namespace = exec_needs_mount_namespace(context, params, runtime);
+
+        if (needs_mount_namespace) {
                 char *tmp = NULL, *var = NULL;
 
                 /* The runtime struct only contains the parent
@@ -1602,6 +1605,7 @@ static int exec_child(
                 }
 
                 r = setup_namespace(
+                                params->apply_chroot ? context->root_directory : NULL,
                                 context->read_write_dirs,
                                 context->read_only_dirs,
                                 context->inaccessible_dirs,
@@ -1627,7 +1631,7 @@ static int exec_child(
         }
 
         if (params->apply_chroot) {
-                if (context->root_directory)
+                if (!needs_mount_namespace && context->root_directory)
                         if (chroot(context->root_directory) < 0) {
                                 *exit_status = EXIT_CHROOT;
                                 return -errno;
diff --git a/src/core/namespace.c b/src/core/namespace.c
index bee2839f54..5b782798c8 100644
--- a/src/core/namespace.c
+++ b/src/core/namespace.c
@@ -36,6 +36,7 @@
 #include "dev-setup.h"
 #include "selinux-util.h"
 #include "namespace.h"
+#include "mkdir.h"
 
 typedef enum MountMode {
         /* This is ordered by priority! */
@@ -124,6 +125,22 @@ static void drop_duplicates(BindMount *m, unsigned *n) {
         *n = t - m;
 }
 
+static int mount_move_root(const char *path) {
+        if (chdir(path) < 0)
+                return -errno;
+
+        if (mount(path, "/", NULL, MS_MOVE, NULL) < 0)
+                return -errno;
+
+        if (chroot(".") < 0)
+                return -errno;
+
+        if (chdir("/") < 0)
+                return -errno;
+
+        return 0;
+}
+
 static int mount_dev(BindMount *m) {
         static const char devnodes[] =
                 "/dev/null\0"
@@ -225,7 +242,13 @@ static int mount_dev(BindMount *m) {
 
         dev_setup(temporary_mount);
 
-        if (mount(dev, "/dev/", NULL, MS_MOVE, NULL) < 0) {
+        /* Create the /dev directory if missing. It is more likely to be
+         * missing when the service is started with RootDirectory. This is
+         * consistent with mount units creating the mount points when missing.
+         */
+        (void) mkdir_p_label(m->path, 0755);
+
+        if (mount(dev, m->path, NULL, MS_MOVE, NULL) < 0) {
                 r = -errno;
                 goto fail;
         }
@@ -404,6 +427,7 @@ static int make_read_only(BindMount *m) {
 }
 
 int setup_namespace(
+                const char* root_directory,
                 char** read_write_dirs,
                 char** read_only_dirs,
                 char** inaccessible_dirs,
@@ -449,37 +473,56 @@ int setup_namespace(
                         return r;
 
                 if (tmp_dir) {
-                        m->path = "/tmp";
+                        m->path = prefix_roota(root_directory, "/tmp");
                         m->mode = PRIVATE_TMP;
                         m++;
                 }
 
                 if (var_tmp_dir) {
-                        m->path = "/var/tmp";
+                        m->path = prefix_roota(root_directory, "/var/tmp");
                         m->mode = PRIVATE_VAR_TMP;
                         m++;
                 }
 
                 if (private_dev) {
-                        m->path = "/dev";
+                        m->path = prefix_roota(root_directory, "/dev");
                         m->mode = PRIVATE_DEV;
                         m++;
                 }
 
                 if (bus_endpoint_path) {
-                        m->path = bus_endpoint_path;
+                        m->path = prefix_roota(root_directory, bus_endpoint_path);
                         m->mode = PRIVATE_BUS_ENDPOINT;
                         m++;
                 }
 
                 if (protect_home != PROTECT_HOME_NO) {
-                        r = append_mounts(&m, STRV_MAKE("-/home", "-/run/user", "-/root"), protect_home == PROTECT_HOME_READ_ONLY ? READONLY : INACCESSIBLE);
+                        const char *home_dir, *run_user_dir, *root_dir;
+
+                        home_dir = prefix_roota(root_directory, "/home");
+                        home_dir = strjoina("-", home_dir);
+                        run_user_dir = prefix_roota(root_directory, "/run/user");
+                        run_user_dir = strjoina("-", run_user_dir);
+                        root_dir = prefix_roota(root_directory, "/root");
+                        root_dir = strjoina("-", root_dir);
+
+                        r = append_mounts(&m, STRV_MAKE(home_dir, run_user_dir, root_dir),
+                                protect_home == PROTECT_HOME_READ_ONLY ? READONLY : INACCESSIBLE);
                         if (r < 0)
                                 return r;
                 }
 
                 if (protect_system != PROTECT_SYSTEM_NO) {
-                        r = append_mounts(&m, protect_system == PROTECT_SYSTEM_FULL ? STRV_MAKE("/usr", "-/boot", "/etc") : STRV_MAKE("/usr", "-/boot"), READONLY);
+                        const char *usr_dir, *boot_dir, *etc_dir;
+
+                        usr_dir = prefix_roota(root_directory, "/home");
+                        boot_dir = prefix_roota(root_directory, "/boot");
+                        boot_dir = strjoina("-", boot_dir);
+                        etc_dir = prefix_roota(root_directory, "/etc");
+
+                        r = append_mounts(&m, protect_system == PROTECT_SYSTEM_FULL
+                                ? STRV_MAKE(usr_dir, boot_dir, etc_dir)
+                                : STRV_MAKE(usr_dir, boot_dir), READONLY);
                         if (r < 0)
                                 return r;
                 }
@@ -490,12 +533,20 @@ int setup_namespace(
                 drop_duplicates(mounts, &n);
         }
 
-        if (n > 0) {
+        if (n > 0 || root_directory) {
                 /* Remount / as SLAVE so that nothing now mounted in the namespace
                    shows up in the parent */
                 if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0)
                         return -errno;
+        }
+
+        if (root_directory) {
+                /* Turn directory into bind mount */
+                if (mount(root_directory, root_directory, NULL, MS_BIND|MS_REC, NULL) < 0)
+                        return -errno;
+        }
 
+        if (n > 0) {
                 for (m = mounts; m < mounts + n; ++m) {
                         r = apply_mount(m, tmp_dir, var_tmp_dir);
                         if (r < 0)
@@ -509,12 +560,21 @@ int setup_namespace(
                 }
         }
 
+        if (root_directory) {
+                /* MS_MOVE does not work on MS_SHARED so the remount MS_SHARED will be done later */
+                r = mount_move_root(root_directory);
+
+                /* at this point, we cannot rollback */
+                if (r < 0)
+                        return r;
+        }
+
         /* Remount / as the desired mode. Not that this will not
          * reestablish propagation from our side to the host, since
          * what's disconnected is disconnected. */
         if (mount(NULL, "/", NULL, mount_flags | MS_REC, NULL) < 0) {
-                r = -errno;
-                goto fail;
+                /* at this point, we cannot rollback */
+                return -errno;
         }
 
         return 0;
diff --git a/src/core/namespace.h b/src/core/namespace.h
index 42b92e7803..00ab22bf2e 100644
--- a/src/core/namespace.h
+++ b/src/core/namespace.h
@@ -41,7 +41,8 @@ typedef enum ProtectSystem {
         _PROTECT_SYSTEM_INVALID = -1
 } ProtectSystem;
 
-int setup_namespace(char **read_write_dirs,
+int setup_namespace(const char *chroot,
+                    char **read_write_dirs,
                     char **read_only_dirs,
                     char **inaccessible_dirs,
                     const char *tmp_dir,
diff --git a/src/test/test-ns.c b/src/test/test-ns.c
index 76b131c284..3050be9e9d 100644
--- a/src/test/test-ns.c
+++ b/src/test/test-ns.c
@@ -38,10 +38,12 @@ int main(int argc, char *argv[]) {
                 NULL
         };
 
-        const char * const inaccessible[] = {
+        const char *inaccessible[] = {
                 "/home/lennart/projects",
                 NULL
         };
+        char *root_directory;
+        char *projects_directory;
 
         int r;
         char tmp_dir[] = "/tmp/systemd-private-XXXXXX",
@@ -50,7 +52,20 @@ int main(int argc, char *argv[]) {
         assert_se(mkdtemp(tmp_dir));
         assert_se(mkdtemp(var_tmp_dir));
 
-        r = setup_namespace((char **) writable,
+        root_directory = getenv("TEST_NS_CHROOT");
+        projects_directory = getenv("TEST_NS_PROJECTS");
+
+        if (projects_directory)
+                inaccessible[0] = projects_directory;
+
+        log_info("Inaccessible directory: '%s'", inaccessible[0]);
+        if (root_directory)
+                log_info("Chroot: '%s'", root_directory);
+        else
+                log_info("Not chrooted");
+
+        r = setup_namespace(root_directory,
+                            (char **) writable,
                             (char **) readonly,
                             (char **) inaccessible,
                             tmp_dir,
@@ -62,6 +77,11 @@ int main(int argc, char *argv[]) {
                             0);
         if (r < 0) {
                 log_error_errno(r, "Failed to setup namespace: %m");
+
+                log_info("Usage:\n"
+                         "  sudo TEST_NS_PROJECTS=/home/lennart/projects ./test-ns\n"
+                         "  sudo TEST_NS_CHROOT=/home/alban/debian-tree TEST_NS_PROJECTS=/home/alban/debian-tree/home/alban/Documents ./test-ns");
+
                 return 1;
         }