Merge pull request #4962 from poettering/root-directory-2

Add new MountAPIVFS= boolean unit file setting + RootImage=

8 files changed, 243 insertions, 41 deletions

diff --git a/src/core/dbus-execute.c b/src/core/dbus-execute.c
index cc10e2d8e7..7df4cab3f6 100644
--- a/src/core/dbus-execute.c
+++ b/src/core/dbus-execute.c
@@ -758,6 +758,7 @@ const sd_bus_vtable bus_exec_vtable[] = {
         SD_BUS_PROPERTY("LimitRTTIMESoft", "t", bus_property_get_rlimit, offsetof(ExecContext, rlimit[RLIMIT_RTTIME]), SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("WorkingDirectory", "s", property_get_working_directory, 0, SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("RootDirectory", "s", NULL, offsetof(ExecContext, root_directory), SD_BUS_VTABLE_PROPERTY_CONST),
+        SD_BUS_PROPERTY("RootImage", "s", NULL, offsetof(ExecContext, root_image), SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("OOMScoreAdjust", "i", property_get_oom_score_adjust, 0, SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("Nice", "i", property_get_nice, 0, SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("IOScheduling", "i", property_get_ioprio, 0, SD_BUS_VTABLE_PROPERTY_CONST),
@@ -828,6 +829,7 @@ const sd_bus_vtable bus_exec_vtable[] = {
         SD_BUS_PROPERTY("RestrictNamespaces", "t", bus_property_get_ulong, offsetof(ExecContext, restrict_namespaces), SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("BindPaths", "a(ssbt)", property_get_bind_paths, 0, SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_PROPERTY("BindReadOnlyPaths", "a(ssbt)", property_get_bind_paths, 0, SD_BUS_VTABLE_PROPERTY_CONST),
+        SD_BUS_PROPERTY("MountAPIVFS", "b", bus_property_get_bool, offsetof(ExecContext, mount_apivfs), SD_BUS_VTABLE_PROPERTY_CONST),
         SD_BUS_VTABLE_END
 };
 
@@ -1047,7 +1049,7 @@ int bus_exec_context_set_transient_property(
 
                 return 1;
 
-        } else if (STR_IN_SET(name, "TTYPath", "RootDirectory")) {
+        } else if (STR_IN_SET(name, "TTYPath", "RootDirectory", "RootImage")) {
                 const char *s;
 
                 r = sd_bus_message_read(message, "s", &s);
@@ -1060,6 +1062,8 @@ int bus_exec_context_set_transient_property(
                 if (mode != UNIT_CHECK) {
                         if (streq(name, "TTYPath"))
                                 r = free_and_strdup(&c->tty_path, s);
+                        else if (streq(name, "RootImage"))
+                                r = free_and_strdup(&c->root_image, s);
                         else {
                                 assert(streq(name, "RootDirectory"));
                                 r = free_and_strdup(&c->root_directory, s);
@@ -1207,7 +1211,7 @@ int bus_exec_context_set_transient_property(
                               "PrivateTmp", "PrivateDevices", "PrivateNetwork", "PrivateUsers",
                               "NoNewPrivileges", "SyslogLevelPrefix", "MemoryDenyWriteExecute",
                               "RestrictRealtime", "DynamicUser", "RemoveIPC", "ProtectKernelTunables",
-                              "ProtectKernelModules", "ProtectControlGroups")) {
+                              "ProtectKernelModules", "ProtectControlGroups", "MountAPIVFS")) {
                 int b;
 
                 r = sd_bus_message_read(message, "b", &b);
@@ -1247,6 +1251,8 @@ int bus_exec_context_set_transient_property(
                                 c->protect_kernel_modules = b;
                         else if (streq(name, "ProtectControlGroups"))
                                 c->protect_control_groups = b;
+                        else if (streq(name, "MountAPIVFS"))
+                                c->mount_apivfs = b;
 
                         unit_write_drop_in_private_format(u, mode, name, "%s=%s", name, yes_no(b));
                 }
@@ -1495,12 +1501,15 @@ int bus_exec_context_set_transient_property(
                         return r;
 
                 STRV_FOREACH(p, l) {
-                        int offset;
-                        if (!utf8_is_valid(*p))
+                        const char *i = *p;
+                        size_t offset;
+
+                        if (!utf8_is_valid(i))
                                 return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid %s", name);
 
-                        offset = **p == '-';
-                        if (!path_is_absolute(*p + offset))
+                        offset = i[0] == '-';
+                        offset += i[offset] == '+';
+                        if (!path_is_absolute(i + offset))
                                 return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid %s", name);
                 }
 
@@ -1519,7 +1528,6 @@ int bus_exec_context_set_transient_property(
                                 unit_write_drop_in_private_format(u, mode, name, "%s=", name);
                         } else {
                                 r = strv_extend_strv(dirs, l, true);
-
                                 if (r < 0)
                                         return -ENOMEM;
 
diff --git a/src/core/execute.c b/src/core/execute.c
index aa0ddb564e..f57eb26388 100644
--- a/src/core/execute.c
+++ b/src/core/execute.c
@@ -1640,6 +1640,9 @@ static bool exec_needs_mount_namespace(
         assert(context);
         assert(params);
 
+        if (context->root_image)
+                return true;
+
         if (!strv_isempty(context->read_write_paths) ||
             !strv_isempty(context->read_only_paths) ||
             !strv_isempty(context->inaccessible_paths))
@@ -1662,6 +1665,9 @@ static bool exec_needs_mount_namespace(
             context->protect_control_groups)
                 return true;
 
+        if (context->mount_apivfs)
+                return true;
+
         return false;
 }
 
@@ -1935,13 +1941,14 @@ static int apply_mount_namespace(Unit *u, const ExecContext *context,
         int r;
         _cleanup_strv_free_ char **rw = NULL;
         char *tmp = NULL, *var = NULL;
-        const char *root_dir = NULL;
+        const char *root_dir = NULL, *root_image = NULL;
         NameSpaceInfo ns_info = {
                 .ignore_protect_paths = false,
                 .private_dev = context->private_devices,
                 .protect_control_groups = context->protect_control_groups,
                 .protect_kernel_tunables = context->protect_kernel_tunables,
                 .protect_kernel_modules = context->protect_kernel_modules,
+                .mount_apivfs = context->mount_apivfs,
         };
 
         assert(context);
@@ -1961,8 +1968,12 @@ static int apply_mount_namespace(Unit *u, const ExecContext *context,
         if (r < 0)
                 return r;
 
-        if (params->flags & EXEC_APPLY_CHROOT)
-                root_dir = context->root_directory;
+        if (params->flags & EXEC_APPLY_CHROOT) {
+                root_image = context->root_image;
+
+                if (!root_image)
+                        root_dir = context->root_directory;
+        }
 
         /*
          * If DynamicUser=no and RootDirectory= is set then lets pass a relaxed
@@ -1972,7 +1983,8 @@ static int apply_mount_namespace(Unit *u, const ExecContext *context,
         if (!context->dynamic_user && root_dir)
                 ns_info.ignore_protect_paths = true;
 
-        r = setup_namespace(root_dir, &ns_info, rw,
+        r = setup_namespace(root_dir, root_image,
+                            &ns_info, rw,
                             context->read_only_paths,
                             context->inaccessible_paths,
                             context->bind_mounts,
@@ -1981,7 +1993,8 @@ static int apply_mount_namespace(Unit *u, const ExecContext *context,
                             var,
                             context->protect_home,
                             context->protect_system,
-                            context->mount_flags);
+                            context->mount_flags,
+                            DISSECT_IMAGE_DISCARD_ON_LOOP);
 
         /* If we couldn't set up the namespace this is probably due to a
          * missing capability. In this case, silently proceeed. */
@@ -1995,10 +2008,12 @@ static int apply_mount_namespace(Unit *u, const ExecContext *context,
         return r;
 }
 
-static int apply_working_directory(const ExecContext *context,
-                                   const ExecParameters *params,
-                                   const char *home,
-                                   const bool needs_mount_ns) {
+static int apply_working_directory(
+                const ExecContext *context,
+                const ExecParameters *params,
+                const char *home,
+                const bool needs_mount_ns) {
+
         const char *d;
         const char *wd;
 
@@ -2979,6 +2994,7 @@ void exec_context_done(ExecContext *c) {
 
         c->working_directory = mfree(c->working_directory);
         c->root_directory = mfree(c->root_directory);
+        c->root_image = mfree(c->root_image);
         c->tty_path = mfree(c->tty_path);
         c->syslog_identifier = mfree(c->syslog_identifier);
         c->user = mfree(c->user);
@@ -3294,6 +3310,7 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
                 "%sPrivateUsers: %s\n"
                 "%sProtectHome: %s\n"
                 "%sProtectSystem: %s\n"
+                "%sMountAPIVFS: %s\n"
                 "%sIgnoreSIGPIPE: %s\n"
                 "%sMemoryDenyWriteExecute: %s\n"
                 "%sRestrictRealtime: %s\n",
@@ -3310,10 +3327,14 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
                 prefix, yes_no(c->private_users),
                 prefix, protect_home_to_string(c->protect_home),
                 prefix, protect_system_to_string(c->protect_system),
+                prefix, yes_no(c->mount_apivfs),
                 prefix, yes_no(c->ignore_sigpipe),
                 prefix, yes_no(c->memory_deny_write_execute),
                 prefix, yes_no(c->restrict_realtime));
 
+        if (c->root_image)
+                fprintf(f, "%sRootImage: %s\n", prefix, c->root_image);
+
         STRV_FOREACH(e, c->environment)
                 fprintf(f, "%sEnvironment: %s\n", prefix, *e);
 
diff --git a/src/core/execute.h b/src/core/execute.h
index f8694ef520..9f2b6fd39e 100644
--- a/src/core/execute.h
+++ b/src/core/execute.h
@@ -106,7 +106,7 @@ struct ExecContext {
         char **pass_environment;
 
         struct rlimit *rlimit[_RLIMIT_MAX];
-        char *working_directory, *root_directory;
+        char *working_directory, *root_directory, *root_image;
         bool working_directory_missing_ok;
         bool working_directory_home;
 
@@ -183,6 +183,7 @@ struct ExecContext {
         bool protect_kernel_tunables;
         bool protect_kernel_modules;
         bool protect_control_groups;
+        bool mount_apivfs;
 
         bool no_new_privileges;
 
diff --git a/src/core/load-fragment-gperf.gperf.m4 b/src/core/load-fragment-gperf.gperf.m4
index 15f22a2681..cb9e6fea27 100644
--- a/src/core/load-fragment-gperf.gperf.m4
+++ b/src/core/load-fragment-gperf.gperf.m4
@@ -19,6 +19,7 @@ m4_dnl Define the context options only once
 m4_define(`EXEC_CONTEXT_CONFIG_ITEMS',
 `$1.WorkingDirectory,            config_parse_working_directory,     0,                             offsetof($1, exec_context)
 $1.RootDirectory,                config_parse_unit_path_printf,      0,                             offsetof($1, exec_context.root_directory)
+$1.RootImage,                    config_parse_unit_path_printf,      0,                             offsetof($1, exec_context.root_image)
 $1.User,                         config_parse_user_group,            0,                             offsetof($1, exec_context.user)
 $1.Group,                        config_parse_user_group,            0,                             offsetof($1, exec_context.group)
 $1.SupplementaryGroups,          config_parse_user_group_strv,       0,                             offsetof($1, exec_context.supplementary_groups)
@@ -101,6 +102,7 @@ $1.PrivateUsers,                 config_parse_bool,                  0,
 $1.ProtectSystem,                config_parse_protect_system,        0,                             offsetof($1, exec_context)
 $1.ProtectHome,                  config_parse_protect_home,          0,                             offsetof($1, exec_context)
 $1.MountFlags,                   config_parse_exec_mount_flags,      0,                             offsetof($1, exec_context)
+$1.MountAPIVFS,                  config_parse_bool,                  0,                             offsetof($1, exec_context.mount_apivfs)
 $1.Personality,                  config_parse_personality,           0,                             offsetof($1, exec_context.personality)
 $1.RuntimeDirectoryMode,         config_parse_mode,                  0,                             offsetof($1, exec_context.runtime_directory_mode)
 $1.RuntimeDirectory,             config_parse_runtime_directory,     0,                             offsetof($1, exec_context.runtime_directory)
diff --git a/src/core/load-fragment.c b/src/core/load-fragment.c
index 243c288885..5b7471c0d0 100644
--- a/src/core/load-fragment.c
+++ b/src/core/load-fragment.c
@@ -3839,7 +3839,8 @@ int config_parse_namespace_path_strv(
         cur = rvalue;
         for (;;) {
                 _cleanup_free_ char *word = NULL, *resolved = NULL, *joined = NULL;
-                bool ignore_enoent;
+                const char *w;
+                bool ignore_enoent = false, shall_prefix = false;
 
                 r = extract_first_word(&cur, &word, NULL, EXTRACT_QUOTES);
                 if (r == 0)
@@ -3856,9 +3857,17 @@ int config_parse_namespace_path_strv(
                         continue;
                 }
 
-                ignore_enoent = word[0] == '-';
+                w = word;
+                if (startswith(w, "-")) {
+                        ignore_enoent = true;
+                        w++;
+                }
+                if (startswith(w, "+")) {
+                        shall_prefix = true;
+                        w++;
+                }
 
-                r = unit_full_printf(u, word + ignore_enoent, &resolved);
+                r = unit_full_printf(u, w, &resolved);
                 if (r < 0) {
                         log_syntax(unit, LOG_ERR, filename, line, r, "Failed to resolve specifiers in %s: %m", word);
                         continue;
@@ -3871,7 +3880,9 @@ int config_parse_namespace_path_strv(
 
                 path_kill_slashes(resolved);
 
-                joined = strjoin(ignore_enoent ? "-" : "", resolved);
+                joined = strjoin(ignore_enoent ? "-" : "",
+                                 shall_prefix ? "+" : "",
+                                 resolved);
 
                 r = strv_push(sv, joined);
                 if (r < 0)
diff --git a/src/core/namespace.c b/src/core/namespace.c
index 834883267c..75dca5b791 100644
--- a/src/core/namespace.c
+++ b/src/core/namespace.c
@@ -30,6 +30,7 @@
 #include "dev-setup.h"
 #include "fd-util.h"
 #include "fs-util.h"
+#include "loop-util.h"
 #include "loopback-setup.h"
 #include "missing.h"
 #include "mkdir.h"
@@ -52,10 +53,13 @@ typedef enum MountMode {
         INACCESSIBLE,
         BIND_MOUNT,
         BIND_MOUNT_RECURSIVE,
-        READONLY,
         PRIVATE_TMP,
         PRIVATE_VAR_TMP,
         PRIVATE_DEV,
+        BIND_DEV,
+        SYSFS,
+        PROCFS,
+        READONLY,
         READWRITE,
 } MountMode;
 
@@ -70,13 +74,13 @@ typedef struct MountEntry {
         char *source_malloc;
 } MountEntry;
 
-/*
- * The following Protect tables are to protect paths and mark some of them
- * READONLY, in case a path is covered by an option from another table, then
- * it is marked READWRITE in the current one, and the more restrictive mode is
- * applied from that other table. This way all options can be combined in a
- * safe and comprehensible way for users.
- */
+/* If MountAPIVFS= is used, let's mount /sys and /proc into the it, but only as a fallback if the user hasn't mounted
+ * something there already. These mounts are hence overriden by any other explicitly configured mounts. */
+static const MountEntry apivfs_table[] = {
+        { "/proc",               PROCFS,       false },
+        { "/dev",                BIND_DEV,     false },
+        { "/sys",                SYSFS,        false },
+};
 
 /* ProtectKernelTunables= option and the related filesystem APIs */
 static const MountEntry protect_kernel_tunables_table[] = {
@@ -176,6 +180,13 @@ static const char *mount_entry_source(const MountEntry *p) {
         return p->source_malloc ?: p->source_const;
 }
 
+static void mount_entry_done(MountEntry *p) {
+        assert(p);
+
+        p->path_malloc = mfree(p->path_malloc);
+        p->source_malloc = mfree(p->source_malloc);
+}
+
 static int append_access_mounts(MountEntry **p, char **strv, MountMode mode) {
         char **i;
 
@@ -351,7 +362,7 @@ static void drop_duplicates(MountEntry *m, unsigned *n) {
                 if (previous && path_equal(mount_entry_path(f), mount_entry_path(previous))) {
                         log_debug("%s is duplicate.", mount_entry_path(f));
                         previous->read_only = previous->read_only || mount_entry_read_only(f); /* Propagate the read-only flag to the remaining entry */
-                        f->path_malloc = mfree(f->path_malloc);
+                        mount_entry_done(f);
                         continue;
                 }
 
@@ -379,7 +390,7 @@ static void drop_inaccessible(MountEntry *m, unsigned *n) {
                  * it, as inaccessible paths really should drop the entire subtree. */
                 if (clear && path_startswith(mount_entry_path(f), clear)) {
                         log_debug("%s is masked by %s.", mount_entry_path(f), clear);
-                        f->path_malloc = mfree(f->path_malloc);
+                        mount_entry_done(f);
                         continue;
                 }
 
@@ -419,7 +430,7 @@ static void drop_nop(MountEntry *m, unsigned *n) {
                         /* We found it, let's see if it's the same mode, if so, we can drop this entry */
                         if (found && p->mode == f->mode) {
                                 log_debug("%s is redundant by %s", mount_entry_path(f), mount_entry_path(p));
-                                f->path_malloc = mfree(f->path_malloc);
+                                mount_entry_done(f);
                                 continue;
                         }
                 }
@@ -447,7 +458,7 @@ static void drop_outside_root(const char *root_directory, MountEntry *m, unsigne
 
                 if (!path_startswith(mount_entry_path(f), root_directory)) {
                         log_debug("%s is outside of root directory.", mount_entry_path(f));
-                        f->path_malloc = mfree(f->path_malloc);
+                        mount_entry_done(f);
                         continue;
                 }
 
@@ -458,7 +469,7 @@ static void drop_outside_root(const char *root_directory, MountEntry *m, unsigne
         *n = t - m;
 }
 
-static int mount_dev(MountEntry *m) {
+static int mount_private_dev(MountEntry *m) {
         static const char devnodes[] =
                 "/dev/null\0"
                 "/dev/zero\0"
@@ -597,6 +608,62 @@ fail:
         return r;
 }
 
+static int mount_bind_dev(MountEntry *m) {
+        int r;
+
+        assert(m);
+
+        /* Implements the little brother of mount_private_dev(): simply bind mounts the host's /dev into the service's
+         * /dev. This is only used when RootDirectory= is set. */
+
+        r = path_is_mount_point(mount_entry_path(m), NULL, 0);
+        if (r < 0)
+                return log_debug_errno(r, "Unable to determine whether /dev is already mounted: %m");
+        if (r > 0) /* make this a NOP if /dev is already a mount point */
+                return 0;
+
+        if (mount("/dev", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
+                return log_debug_errno(errno, "Failed to bind mount %s: %m", mount_entry_path(m));
+
+        return 1;
+}
+
+static int mount_sysfs(MountEntry *m) {
+        int r;
+
+        assert(m);
+
+        r = path_is_mount_point(mount_entry_path(m), NULL, 0);
+        if (r < 0)
+                return log_debug_errno(r, "Unable to determine whether /sys is already mounted: %m");
+        if (r > 0) /* make this a NOP if /sys is already a mount point */
+                return 0;
+
+        /* Bind mount the host's version so that we get all child mounts of it, too. */
+        if (mount("/sys", mount_entry_path(m), NULL, MS_BIND|MS_REC, NULL) < 0)
+                return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
+
+        return 1;
+}
+
+static int mount_procfs(MountEntry *m) {
+        int r;
+
+        assert(m);
+
+        r = path_is_mount_point(mount_entry_path(m), NULL, 0);
+        if (r < 0)
+                return log_debug_errno(r, "Unable to determine whether /proc is already mounted: %m");
+        if (r > 0) /* make this a NOP if /proc is already a mount point */
+                return 0;
+
+        /* Mount a new instance, so that we get the one that matches our user namespace, if we are running in one */
+        if (mount("proc", mount_entry_path(m), "proc", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0)
+                return log_debug_errno(errno, "Failed to mount %s: %m", mount_entry_path(m));
+
+        return 1;
+}
+
 static int mount_entry_chase(
                 const char *root_directory,
                 MountEntry *m,
@@ -684,6 +751,7 @@ static int apply_mount(
 
         case BIND_MOUNT_RECURSIVE:
                 /* Also chase the source mount */
+
                 r = mount_entry_chase(root_directory, m, mount_entry_source(m), &m->source_malloc);
                 if (r <= 0)
                         return r;
@@ -700,7 +768,16 @@ static int apply_mount(
                 break;
 
         case PRIVATE_DEV:
-                return mount_dev(m);
+                return mount_private_dev(m);
+
+        case BIND_DEV:
+                return mount_bind_dev(m);
+
+        case SYSFS:
+                return mount_sysfs(m);
+
+        case PROCFS:
+                return mount_procfs(m);
 
         default:
                 assert_not_reached("Unknown mode");
@@ -722,7 +799,7 @@ static int make_read_only(MountEntry *m, char **blacklist) {
 
         if (mount_entry_read_only(m))
                 r = bind_remount_recursive(mount_entry_path(m), true, blacklist);
-        else if (m->mode == PRIVATE_DEV) { /* Can be readonly but the submounts can't*/
+        else if (m->mode == PRIVATE_DEV) { /* Superblock can be readonly but the submounts can't*/
                 if (mount(NULL, mount_entry_path(m), NULL, MS_REMOUNT|DEV_MOUNT_OPTIONS|MS_RDONLY, NULL) < 0)
                         r = -errno;
         } else
@@ -738,6 +815,17 @@ static int make_read_only(MountEntry *m, char **blacklist) {
         return r;
 }
 
+static bool namespace_info_mount_apivfs(const NameSpaceInfo *ns_info) {
+        assert(ns_info);
+
+        /* ProtectControlGroups= and ProtectKernelTunables= imply MountAPIVFS=, since to protect the API VFS mounts,
+         * they need to be around in the first place... */
+
+        return ns_info->mount_apivfs ||
+                ns_info->protect_control_groups ||
+                ns_info->protect_kernel_tunables;
+}
+
 static unsigned namespace_calculate_mounts(
                 const NameSpaceInfo *ns_info,
                 char** read_write_paths,
@@ -774,11 +862,13 @@ static unsigned namespace_calculate_mounts(
                 (ns_info->protect_kernel_tunables ? ELEMENTSOF(protect_kernel_tunables_table) : 0) +
                 (ns_info->protect_control_groups ? 1 : 0) +
                 (ns_info->protect_kernel_modules ? ELEMENTSOF(protect_kernel_modules_table) : 0) +
-                protect_home_cnt + protect_system_cnt;
+                protect_home_cnt + protect_system_cnt +
+                (namespace_info_mount_apivfs(ns_info) ? ELEMENTSOF(apivfs_table) : 0);
 }
 
 int setup_namespace(
                 const char* root_directory,
+                const char* root_image,
                 const NameSpaceInfo *ns_info,
                 char** read_write_paths,
                 char** read_only_paths,
@@ -789,16 +879,57 @@ int setup_namespace(
                 const char* var_tmp_dir,
                 ProtectHome protect_home,
                 ProtectSystem protect_system,
-                unsigned long mount_flags) {
+                unsigned long mount_flags,
+                DissectImageFlags dissect_image_flags) {
 
+        _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL;
+        _cleanup_(decrypted_image_unrefp) DecryptedImage *decrypted_image = NULL;
+        _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL;
+        _cleanup_free_ void *root_hash = NULL;
         MountEntry *m, *mounts = NULL;
+        size_t root_hash_size = 0;
         bool make_slave = false;
         unsigned n_mounts;
         int r = 0;
 
+        assert(ns_info);
+
         if (mount_flags == 0)
                 mount_flags = MS_SHARED;
 
+        if (root_image) {
+                dissect_image_flags |= DISSECT_IMAGE_REQUIRE_ROOT;
+
+                if (protect_system == PROTECT_SYSTEM_STRICT && strv_isempty(read_write_paths))
+                        dissect_image_flags |= DISSECT_IMAGE_READ_ONLY;
+
+                r = loop_device_make_by_path(root_image,
+                                             dissect_image_flags & DISSECT_IMAGE_READ_ONLY ? O_RDONLY : O_RDWR,
+                                             &loop_device);
+                if (r < 0)
+                        return r;
+
+                r = root_hash_load(root_image, &root_hash, &root_hash_size);
+                if (r < 0)
+                        return r;
+
+                r = dissect_image(loop_device->fd, root_hash, root_hash_size, dissect_image_flags, &dissected_image);
+                if (r < 0)
+                        return r;
+
+                r = dissected_image_decrypt(dissected_image, NULL, root_hash, root_hash_size, dissect_image_flags, &decrypted_image);
+                if (r < 0)
+                        return r;
+
+                if (!root_directory) {
+                        /* Create a mount point for the image, if it's still missing. We use the same mount point for
+                         * all images, which is safe, since they all live in their own namespaces after all, and hence
+                         * won't see each other. */
+                        root_directory = "/run/systemd/unit-root";
+                        (void) mkdir(root_directory, 0700);
+                }
+        }
+
         n_mounts = namespace_calculate_mounts(
                         ns_info,
                         read_write_paths,
@@ -878,6 +1009,12 @@ int setup_namespace(
                 if (r < 0)
                         goto finish;
 
+                if (namespace_info_mount_apivfs(ns_info)) {
+                        r = append_static_mounts(&m, apivfs_table, ELEMENTSOF(apivfs_table), ns_info->ignore_protect_paths);
+                        if (r < 0)
+                                goto finish;
+                }
+
                 assert(mounts + n_mounts == m);
 
                 /* Prepend the root directory where that's necessary */
@@ -907,7 +1044,19 @@ int setup_namespace(
                 }
         }
 
-        if (root_directory) {
+        if (root_image) {
+                r = dissected_image_mount(dissected_image, root_directory, dissect_image_flags);
+                if (r < 0)
+                        goto finish;
+
+                r = decrypted_image_relinquish(decrypted_image);
+                if (r < 0)
+                        goto finish;
+
+                loop_device_relinquish(loop_device);
+
+        } else if (root_directory) {
+
                 /* Turn directory into bind mount, if it isn't one yet */
                 r = path_is_mount_point(root_directory, NULL, AT_SYMLINK_FOLLOW);
                 if (r < 0)
@@ -964,7 +1113,7 @@ int setup_namespace(
 
 finish:
         for (m = mounts; m < mounts + n_mounts; m++)
-                free(m->path_malloc);
+                mount_entry_done(m);
 
         return r;
 }
diff --git a/src/core/namespace.h b/src/core/namespace.h
index de3edc419c..f54954bd86 100644
--- a/src/core/namespace.h
+++ b/src/core/namespace.h
@@ -25,6 +25,7 @@ typedef struct BindMount BindMount;
 
 #include <stdbool.h>
 
+#include "dissect-image.h"
 #include "macro.h"
 
 typedef enum ProtectHome {
@@ -50,6 +51,7 @@ struct NameSpaceInfo {
         bool protect_control_groups:1;
         bool protect_kernel_tunables:1;
         bool protect_kernel_modules:1;
+        bool mount_apivfs:1;
 };
 
 struct BindMount {
@@ -62,6 +64,7 @@ struct BindMount {
 
 int setup_namespace(
                 const char *root_directory,
+                const char *root_image,
                 const NameSpaceInfo *ns_info,
                 char **read_write_paths,
                 char **read_only_paths,
@@ -72,7 +75,8 @@ int setup_namespace(
                 const char *var_tmp_dir,
                 ProtectHome protect_home,
                 ProtectSystem protect_system,
-                unsigned long mount_flags);
+                unsigned long mount_flags,
+                DissectImageFlags dissected_image_flags);
 
 int setup_tmp_dirs(
                 const char *id,
diff --git a/src/core/unit.c b/src/core/unit.c
index 44f1d5e206..90d7eea956 100644
--- a/src/core/unit.c
+++ b/src/core/unit.c
@@ -862,6 +862,12 @@ int unit_add_exec_dependencies(Unit *u, ExecContext *c) {
                         return r;
         }
 
+        if (c->root_image) {
+                r = unit_require_mounts_for(u, c->root_image);
+                if (r < 0)
+                        return r;
+        }
+
         if (!MANAGER_IS_SYSTEM(u->manager))
                 return 0;