Merge pull request #4879 from poettering/systemd

src/basic/cgroup-util.c

@@ -2361,6 +2361,7 @@ int cg_enable_everywhere(CGroupMask supported, CGroupMask mask, const char *p) {
 bool cg_is_unified_wanted(void) {
         static thread_local int wanted = -1;
         int r, unified;
+        bool b;
 
         /* If the hierarchy is already mounted, then follow whatever
          * was chosen for it. */
@@ -2374,20 +2375,11 @@ bool cg_is_unified_wanted(void) {
         if (wanted >= 0)
                 return wanted;
 
-        r = get_proc_cmdline_key("systemd.unified_cgroup_hierarchy", NULL);
-        if (r > 0)
-                return (wanted = true);
-        else {
-                _cleanup_free_ char *value = NULL;
+        r = proc_cmdline_get_bool("systemd.unified_cgroup_hierarchy", &b);
+        if (r < 0)
+                return false;
 
-                r = get_proc_cmdline_key("systemd.unified_cgroup_hierarchy=", &value);
-                if (r < 0)
-                        return false;
-                if (r == 0)
-                        return (wanted = false);
-
-                return (wanted = parse_boolean(value) > 0);
-        }
+        return (wanted = r > 0 ? b : false);
 }
 
 bool cg_is_legacy_wanted(void) {
@@ -2397,6 +2389,7 @@ bool cg_is_legacy_wanted(void) {
 bool cg_is_unified_systemd_controller_wanted(void) {
         static thread_local int wanted = -1;
         int r, unified;
+        bool b;
 
         /* If the unified hierarchy is requested in full, no need to
          * bother with this. */
@@ -2415,23 +2408,11 @@ bool cg_is_unified_systemd_controller_wanted(void) {
         if (wanted >= 0)
                 return wanted;
 
-        r = get_proc_cmdline_key("systemd.legacy_systemd_cgroup_controller", NULL);
-        if (r > 0)
-                wanted = false;
-        else {
-                _cleanup_free_ char *value = NULL;
+        r = proc_cmdline_get_bool("systemd.legacy_systemd_cgroup_controller", &b);
+        if (r < 0)
+                return false;
 
-                r = get_proc_cmdline_key("systemd.legacy_systemd_cgroup_controller=", &value);
-                if (r < 0)
-                        return false;
-
-                if (r == 0)
-                        wanted = false;
-                else
-                        wanted = parse_boolean(value) <= 0;
-        }
-
-        return wanted;
+        return (wanted = r > 0 ? b : false);
 }
 
 bool cg_is_legacy_systemd_controller_wanted(void) {

src/basic/fs-util.c

@@ -799,8 +799,10 @@ int chase_symlinks(const char *path, const char *original_root, unsigned flags,
                         return -ENOMEM;
         }
 
-        *ret = done;
-        done = NULL;
+        if (ret) {
+                *ret = done;
+                done = NULL;
+        }
 
         return exists;
 }

src/basic/hexdecoct.c

@@ -97,6 +97,9 @@ int unhexmem(const char *p, size_t l, void **mem, size_t *len) {
         assert(len);
         assert(p);
 
+        if (l % 2 != 0)
+                return -EINVAL;
+
         z = r = malloc((l + 1) / 2 + 1);
         if (!r)
                 return -ENOMEM;
@@ -107,12 +110,10 @@ int unhexmem(const char *p, size_t l, void **mem, size_t *len) {
                 a = unhexchar(x[0]);
                 if (a < 0)
                         return a;
-                else if (x+1 < p + l) {
-                        b = unhexchar(x[1]);
-                        if (b < 0)
-                                return b;
-                } else
-                        b = 0;
+
+                b = unhexchar(x[1]);
+                if (b < 0)
+                        return b;
 
                 *(z++) = (uint8_t) a << 4 | (uint8_t) b;
         }

src/basic/log.c

@@ -981,24 +981,30 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
         if (streq(key, "debug") && !value)
                 log_set_max_level(LOG_DEBUG);
 
-        else if (streq(key, "systemd.log_target") && value) {
+        else if (proc_cmdline_key_streq(key, "systemd.log_target")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (log_set_target_from_string(value) < 0)
                         log_warning("Failed to parse log target '%s'. Ignoring.", value);
 
-        } else if (streq(key, "systemd.log_level") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.log_level")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (log_set_max_level_from_string(value) < 0)
                         log_warning("Failed to parse log level '%s'. Ignoring.", value);
 
-        } else if (streq(key, "systemd.log_color") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.log_color")) {
 
-                if (log_show_color_from_string(value) < 0)
+                if (log_show_color_from_string(value ?: "1") < 0)
                         log_warning("Failed to parse log color setting '%s'. Ignoring.", value);
 
-        } else if (streq(key, "systemd.log_location") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.log_location")) {
 
-                if (log_show_location_from_string(value) < 0)
+                if (log_show_location_from_string(value ?: "1") < 0)
                         log_warning("Failed to parse log location setting '%s'. Ignoring.", value);
         }
 
@@ -1009,10 +1015,9 @@ void log_parse_environment(void) {
         const char *e;
 
         if (get_ctty_devnr(0, NULL) < 0)
-                /* Only try to read the command line in daemons.
-                   We assume that anything that has a controlling
-                   tty is user stuff. */
-                (void) parse_proc_cmdline(parse_proc_cmdline_item, NULL, true);
+                /* Only try to read the command line in daemons.  We assume that anything that has a controlling tty is
+                   user stuff. */
+                (void) proc_cmdline_parse(parse_proc_cmdline_item, NULL, PROC_CMDLINE_STRIP_RD_PREFIX);
 
         e = secure_getenv("SYSTEMD_LOG_TARGET");
         if (e && log_set_target_from_string(e) < 0)

src/basic/mount-util.c

@@ -673,6 +673,9 @@ int mount_verbose(
         else if ((flags & MS_BIND) && !type)
                 log_debug("Bind-mounting %s on %s (%s \"%s\")...",
                           what, where, strnull(fl), strempty(options));
+        else if (flags & MS_MOVE)
+                log_debug("Moving mount %s → %s (%s \"%s\")...",
+                          what, where, strnull(fl), strempty(options));
         else
                 log_debug("Mounting %s on %s (%s \"%s\")...",
                           strna(type), where, strnull(fl), strempty(options));

src/basic/proc-cmdline.c

@@ -34,17 +34,30 @@
 #include "virt.h"
 
 int proc_cmdline(char **ret) {
+        const char *e;
         assert(ret);
 
+        /* For testing purposes it is sometimes useful to be able to override what we consider /proc/cmdline to be */
+        e = secure_getenv("SYSTEMD_PROC_CMDLINE");
+        if (e) {
+                char *m;
+
+                m = strdup(e);
+                if (!m)
+                        return -ENOMEM;
+
+                *ret = m;
+                return 0;
+        }
+
         if (detect_container() > 0)
                 return get_process_cmdline(1, 0, false, ret);
         else
                 return read_one_line_file("/proc/cmdline", ret);
 }
 
-int parse_proc_cmdline(int (*parse_item)(const char *key, const char *value, void *data),
-                       void *data,
-                       bool strip_prefix) {
+int proc_cmdline_parse(proc_cmdline_parse_t parse_item, void *data, unsigned flags) {
+
         _cleanup_free_ char *line = NULL;
         const char *p;
         int r;
@@ -58,7 +71,7 @@ int parse_proc_cmdline(int (*parse_item)(const char *key, const char *value, voi
         p = line;
         for (;;) {
                 _cleanup_free_ char *word = NULL;
-                char *value = NULL, *unprefixed;
+                char *value, *key, *q;
 
                 r = extract_first_word(&p, &word, NULL, EXTRACT_QUOTES|EXTRACT_RELAX);
                 if (r < 0)
@@ -66,17 +79,23 @@ int parse_proc_cmdline(int (*parse_item)(const char *key, const char *value, voi
                 if (r == 0)
                         break;
 
-                /* Filter out arguments that are intended only for the
-                 * initrd */
-                unprefixed = startswith(word, "rd.");
-                if (unprefixed && !in_initrd())
-                        continue;
+                key = word;
 
-                value = strchr(word, '=');
+                /* Filter out arguments that are intended only for the initrd */
+                q = startswith(word, "rd.");
+                if (q) {
+                        if (!in_initrd())
+                                continue;
+
+                        if (flags & PROC_CMDLINE_STRIP_RD_PREFIX)
+                                key = q;
+                }
+
+                value = strchr(key, '=');
                 if (value)
                         *(value++) = 0;
 
-                r = parse_item(strip_prefix && unprefixed ? unprefixed : word, value, data);
+                r = parse_item(key, value, data);
                 if (r < 0)
                         return r;
         }
@@ -84,13 +103,64 @@ int parse_proc_cmdline(int (*parse_item)(const char *key, const char *value, voi
         return 0;
 }
 
-int get_proc_cmdline_key(const char *key, char **value) {
+static bool relaxed_equal_char(char a, char b) {
+
+        return a == b ||
+                (a == '_' && b == '-') ||
+                (a == '-' && b == '_');
+}
+
+char *proc_cmdline_key_startswith(const char *s, const char *prefix) {
+
+        assert(s);
+        assert(prefix);
+
+        /* Much like startswith(), but considers "-" and "_" the same */
+
+        for (; *prefix != 0; s++, prefix++)
+                if (!relaxed_equal_char(*s, *prefix))
+                        return NULL;
+
+        return (char*) s;
+}
+
+bool proc_cmdline_key_streq(const char *x, const char *y) {
+        assert(x);
+        assert(y);
+
+        /* Much like streq(), but considers "-" and "_" the same */
+
+        for (; *x != 0 || *y != 0; x++, y++)
+                if (!relaxed_equal_char(*x, *y))
+                        return false;
+
+        return true;
+}
+
+int proc_cmdline_get_key(const char *key, unsigned flags, char **value) {
         _cleanup_free_ char *line = NULL, *ret = NULL;
         bool found = false;
         const char *p;
         int r;
 
-        assert(key);
+        /* Looks for a specific key on the kernel command line. Supports two modes:
+         *
+         * a) The "value" parameter is used. In this case a parameter beginning with the "key" string followed by "="
+         *    is searched, and the value following this is returned in "value".
+         *
+         * b) as above, but the PROC_CMDLINE_VALUE_OPTIONAL flag is set. In this case if the the key is found as a
+         *    separate word (i.e. not followed by "=" but instead by whitespace or the end of the command line), then
+         *    this is also accepted, and "value" is returned as NULL.
+         *
+         * c) The "value" parameter is NULL. In this case a search for the exact "key" parameter is performed.
+         *
+         * In all three cases, > 0 is returned if the key is found, 0 if not.*/
+
+        if (isempty(key))
+                return -EINVAL;
+
+        if ((flags & PROC_CMDLINE_VALUE_OPTIONAL) && !value)
+                return -EINVAL;
 
         r = proc_cmdline(&line);
         if (r < 0)
@@ -107,21 +177,26 @@ int get_proc_cmdline_key(const char *key, char **value) {
                 if (r == 0)
                         break;
 
-                /* Filter out arguments that are intended only for the
-                 * initrd */
+                /* Automatically filter out arguments that are intended only for the initrd, if we are not in the
+                 * initrd. */
                 if (!in_initrd() && startswith(word, "rd."))
                         continue;
 
                 if (value) {
-                        e = startswith(word, key);
+                        e = proc_cmdline_key_startswith(word, key);
                         if (!e)
                                 continue;
 
-                        r = free_and_strdup(&ret, e);
-                        if (r < 0)
-                                return r;
+                        if (*e == '=') {
+                                r = free_and_strdup(&ret, e+1);
+                                if (r < 0)
+                                        return r;
+
+                                found = true;
+
+                        } else if (*e == 0 && (flags & PROC_CMDLINE_VALUE_OPTIONAL))
+                                found = true;
 
-                        found = true;
                 } else {
                         if (streq(word, key))
                                 found = true;
@@ -134,20 +209,42 @@ int get_proc_cmdline_key(const char *key, char **value) {
         }
 
         return found;
+}
 
+int proc_cmdline_get_bool(const char *key, bool *ret) {
+        _cleanup_free_ char *v = NULL;
+        int r;
+
+        assert(ret);
+
+        r = proc_cmdline_get_key(key, PROC_CMDLINE_VALUE_OPTIONAL, &v);
+        if (r < 0)
+                return r;
+        if (r == 0) {
+                *ret = false;
+                return 0;
+        }
+
+        if (v) { /* parameter passed */
+                r = parse_boolean(v);
+                if (r < 0)
+                        return r;
+                *ret = r;
+        } else /* no parameter passed */
+                *ret = true;
+
+        return 1;
 }
 
 int shall_restore_state(void) {
-        _cleanup_free_ char *value = NULL;
+        bool ret;
         int r;
 
-        r = get_proc_cmdline_key("systemd.restore_state=", &value);
+        r = proc_cmdline_get_bool("systemd.restore_state", &ret);
         if (r < 0)
                 return r;
-        if (r == 0)
-                return true;
 
-        return parse_boolean(value);
+        return r > 0 ? ret : true;
 }
 
 static const char * const rlmap[] = {

src/basic/proc-cmdline.h

@@ -19,11 +19,36 @@
   along with systemd; If not, see <http://www.gnu.org/licenses/>.
 ***/
 
+#include <stdbool.h>
+
+#include "log.h"
+
+enum {
+        PROC_CMDLINE_STRIP_RD_PREFIX = 1,
+        PROC_CMDLINE_VALUE_OPTIONAL = 2,
+};
+
+typedef int (*proc_cmdline_parse_t)(const char *key, const char *value, void *data);
+
 int proc_cmdline(char **ret);
-int parse_proc_cmdline(int (*parse_item)(const char *key, const char *value, void *data),
-                       void *data,
-                       bool strip_prefix);
-int get_proc_cmdline_key(const char *parameter, char **value);
+
+int proc_cmdline_parse(const proc_cmdline_parse_t parse, void *userdata, unsigned flags);
+
+int proc_cmdline_get_key(const char *parameter, unsigned flags, char **value);
+int proc_cmdline_get_bool(const char *key, bool *ret);
+
+char *proc_cmdline_key_startswith(const char *s, const char *prefix);
+bool proc_cmdline_key_streq(const char *x, const char *y);
 
 int shall_restore_state(void);
 const char* runlevel_to_target(const char *rl);
+
+/* A little helper call, to be used in proc_cmdline_parse_t callbacks */
+static inline bool proc_cmdline_value_missing(const char *key, const char *value) {
+        if (!value) {
+                log_warning("Missing argument for %s= kernel command line switch, ignoring.", key);
+                return true;
+        }
+
+        return false;
+}

src/basic/stat-util.c

@@ -28,6 +28,7 @@
 
 #include "dirent-util.h"
 #include "fd-util.h"
+#include "fs-util.h"
 #include "macro.h"
 #include "missing.h"
 #include "stat-util.h"
@@ -143,22 +144,29 @@ int path_is_read_only_fs(const char *path) {
 }
 
 int path_is_os_tree(const char *path) {
-        char *p;
         int r;
 
         assert(path);
 
+        /* Does the path exist at all? If not, generate an error immediately. This is useful so that a missing root dir
+         * always results in -ENOENT, and we can properly distuingish the case where the whole root doesn't exist from
+         * the case where just the os-release file is missing. */
+        if (laccess(path, F_OK) < 0)
+                return -errno;
+
         /* We use /usr/lib/os-release as flag file if something is an OS */
-        p = strjoina(path, "/usr/lib/os-release");
-        r = access(p, F_OK);
-        if (r >= 0)
-                return 1;
+        r = chase_symlinks("/usr/lib/os-release", path, CHASE_PREFIX_ROOT, NULL);
+        if (r == -ENOENT) {
 
-        /* Also check for the old location in /etc, just in case. */
-        p = strjoina(path, "/etc/os-release");
-        r = access(p, F_OK);
+                /* Also check for the old location in /etc, just in case. */
+                r = chase_symlinks("/etc/os-release", path, CHASE_PREFIX_ROOT, NULL);
+                if (r == -ENOENT)
+                        return 0; /* We got nothing */
+        }
+        if (r < 0)
+                return r;
 
-        return r >= 0;
+        return 1;
 }
 
 int files_same(const char *filea, const char *fileb) {
@@ -196,7 +204,7 @@ int fd_check_fstype(int fd, statfs_f_type_t magic_value) {
 int path_check_fstype(const char *path, statfs_f_type_t magic_value) {
         _cleanup_close_ int fd = -1;
 
-        fd = open(path, O_RDONLY);
+        fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_PATH);
         if (fd < 0)
                 return -errno;
 
@@ -216,3 +224,13 @@ int fd_is_temporary_fs(int fd) {
 
         return is_temporary_fs(&s);
 }
+
+int path_is_temporary_fs(const char *path) {
+        _cleanup_close_ int fd = -1;
+
+        fd = open(path, O_RDONLY|O_CLOEXEC|O_NOCTTY|O_PATH);
+        if (fd < 0)
+                return -errno;
+
+        return fd_is_temporary_fs(fd);
+}

src/basic/stat-util.h

@@ -61,6 +61,7 @@ int path_check_fstype(const char *path, statfs_f_type_t magic_value);
 
 bool is_temporary_fs(const struct statfs *s) _pure_;
 int fd_is_temporary_fs(int fd);
+int path_is_temporary_fs(const char *path);
 
 /* Because statfs.t_type can be int on some architectures, we have to cast
  * the const magic to the type, otherwise the compiler warns about

src/core/dbus-manager.c

@@ -35,6 +35,7 @@
 #include "fd-util.h"
 #include "fileio.h"
 #include "format-util.h"
+#include "fs-util.h"
 #include "install.h"
 #include "log.h"
 #include "path-util.h"
@@ -1484,25 +1485,36 @@ static int method_switch_root(sd_bus_message *message, void *userdata, sd_bus_er
         if (r < 0)
                 return r;
 
-        if (path_equal(root, "/") || !path_is_absolute(root))
-                return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid switch root path %s", root);
+        if (isempty(root))
+                return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "New root directory may not be the empty string.");
+        if (!path_is_absolute(root))
+                return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "New root path '%s' is not absolute.", root);
+        if (path_equal(root, "/"))
+                return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "New root directory cannot be the old root directory.");
 
         /* Safety check */
         if (isempty(init)) {
-                if (!path_is_os_tree(root))
-                        return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Specified switch root path %s does not seem to be an OS tree. os-release file is missing.", root);
+                r = path_is_os_tree(root);
+                if (r < 0)
+                        return sd_bus_error_set_errnof(error, r, "Failed to determine whether root path '%s' contains an OS tree: %m", root);
+                if (r == 0)
+                        return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Specified switch root path '%s' does not seem to be an OS tree. os-release file is missing.", root);
         } else {
-                _cleanup_free_ char *p = NULL;
+                _cleanup_free_ char *chased = NULL;
 
                 if (!path_is_absolute(init))
-                        return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Invalid init path %s", init);
+                        return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Path to init binary '%s' not absolute.", init);
 
-                p = strappend(root, init);
-                if (!p)
-                        return -ENOMEM;
+                r = chase_symlinks(init, root, CHASE_PREFIX_ROOT, &chased);
+                if (r < 0)
+                        return sd_bus_error_set_errnof(error, r, "Could not resolve init executable %s: %m", init);
 
-                if (access(p, X_OK) < 0)
-                        return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Specified init binary %s does not exist.", p);
+                if (laccess(chased, X_OK) < 0) {
+                        if (errno == EACCES)
+                                return sd_bus_error_setf(error, SD_BUS_ERROR_INVALID_ARGS, "Init binary %s is not executable.", init);
+
+                        return sd_bus_error_set_errnof(error, r, "Could not check whether init binary %s is executable: %m", init);
+                }
         }
 
         rt = strdup(root);

src/core/job.c

@@ -645,7 +645,7 @@ _pure_ static const char *job_get_status_message_format(Unit *u, JobType t, JobR
                 [JOB_DEPENDENCY]  = "Dependency failed for %s.",
                 [JOB_ASSERT]      = "Assertion failed for %s.",
                 [JOB_UNSUPPORTED] = "Starting of %s not supported.",
-                [JOB_COLLECTED]   = "Unecessary job for %s was removed.",
+                [JOB_COLLECTED]   = "Unnecessary job for %s was removed.",
         };
         static const char *const generic_finished_stop_job[_JOB_RESULT_MAX] = {
                 [JOB_DONE]        = "Stopped %s.",

src/core/main.c

@@ -49,6 +49,7 @@
 #include "cpu-set-util.h"
 #include "dbus-manager.h"
 #include "def.h"
+#include "emergency-action.h"
 #include "env-util.h"
 #include "fd-util.h"
 #include "fdset.h"
@@ -90,7 +91,6 @@
 #include "user-util.h"
 #include "virt.h"
 #include "watchdog.h"
-#include "emergency-action.h"
 
 static enum {
         ACTION_RUN,
@@ -337,60 +337,73 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
 
         assert(key);
 
-        if (streq(key, "systemd.unit") && value) {
+        if (STR_IN_SET(key, "systemd.unit", "rd.systemd.unit")) {
 
-                if (!in_initrd())
-                        return free_and_strdup(&arg_default_unit, value);
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
-        } else if (streq(key, "rd.systemd.unit") && value) {
+                if (!unit_name_is_valid(value, UNIT_NAME_PLAIN|UNIT_NAME_INSTANCE))
+                        log_warning("Unit name specified on %s= is not valid, ignoring: %s", key, value);
+                else if (in_initrd() == !!startswith(key, "rd.")) {
+                        if (free_and_strdup(&arg_default_unit, value) < 0)
+                                return log_oom();
+                }
 
-                if (in_initrd())
-                        return free_and_strdup(&arg_default_unit, value);
+        } else if (proc_cmdline_key_streq(key, "systemd.dump_core")) {
 
-        } else if (streq(key, "systemd.dump_core") && value) {
-
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse dump core switch %s. Ignoring.", value);
                 else
                         arg_dump_core = r;
 
-        } else if (streq(key, "systemd.crash_chvt") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.crash_chvt")) {
 
-                if (parse_crash_chvt(value) < 0)
+                if (!value)
+                        arg_crash_chvt = 0; /* turn on */
+                else if (parse_crash_chvt(value) < 0)
                         log_warning("Failed to parse crash chvt switch %s. Ignoring.", value);
 
-        } else if (streq(key, "systemd.crash_shell") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.crash_shell")) {
 
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse crash shell switch %s. Ignoring.", value);
                 else
                         arg_crash_shell = r;
 
-        } else if (streq(key, "systemd.crash_reboot") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.crash_reboot")) {
 
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse crash reboot switch %s. Ignoring.", value);
                 else
                         arg_crash_reboot = r;
 
-        } else if (streq(key, "systemd.confirm_spawn") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.confirm_spawn")) {
+                char *s;
 
-                arg_confirm_spawn = mfree(arg_confirm_spawn);
-
-                r = parse_confirm_spawn(value, &arg_confirm_spawn);
+                r = parse_confirm_spawn(value, &s);
                 if (r < 0)
                         log_warning_errno(r, "Failed to parse confirm_spawn switch %s. Ignoring.", value);
+                else {
+                        free(arg_confirm_spawn);
+                        arg_confirm_spawn = s;
+                }
 
-        } else if (streq(key, "systemd.show_status") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.show_status")) {
 
-                r = parse_show_status(value, &arg_show_status);
-                if (r < 0)
-                        log_warning("Failed to parse show status switch %s. Ignoring.", value);
+                if (value) {
+                        r = parse_show_status(value, &arg_show_status);
+                        if (r < 0)
+                                log_warning("Failed to parse show status switch %s. Ignoring.", value);
+                } else
+                        arg_show_status = SHOW_STATUS_YES;
 
-        } else if (streq(key, "systemd.default_standard_output") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.default_standard_output")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 r = exec_output_from_string(value);
                 if (r < 0)
@@ -398,7 +411,10 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                 else
                         arg_default_std_output = r;
 
-        } else if (streq(key, "systemd.default_standard_error") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.default_standard_error")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 r = exec_output_from_string(value);
                 if (r < 0)
@@ -406,24 +422,42 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                 else
                         arg_default_std_error = r;
 
-        } else if (streq(key, "systemd.setenv") && value) {
+        } else if (streq(key, "systemd.setenv")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (env_assignment_is_valid(value)) {
                         char **env;
 
                         env = strv_env_set(arg_default_environment, value);
-                        if (env)
-                                arg_default_environment = env;
-                        else
-                                log_warning_errno(ENOMEM, "Setting environment variable '%s' failed, ignoring: %m", value);
+                        if (!env)
+                                return log_oom();
+
+                        arg_default_environment = env;
                 } else
                         log_warning("Environment variable name '%s' is not valid. Ignoring.", value);
 
-        } else if (streq(key, "systemd.machine_id") && value) {
+        } else if (proc_cmdline_key_streq(key, "systemd.machine_id")) {
 
-               r = set_machine_id(value);
-               if (r < 0)
-                       log_warning("MachineID '%s' is not valid. Ignoring.", value);
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = set_machine_id(value);
+                if (r < 0)
+                        log_warning("MachineID '%s' is not valid. Ignoring.", value);
+
+        } else if (proc_cmdline_key_streq(key, "systemd.default_timeout_start_sec")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = parse_sec(value, &arg_default_timeout_start_usec);
+                if (r < 0)
+                        log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
+
+                if (arg_default_timeout_start_usec <= 0)
+                        arg_default_timeout_start_usec = USEC_INFINITY;
 
         } else if (streq(key, "quiet") && !value) {
 
@@ -445,15 +479,6 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                 target = runlevel_to_target(key);
                 if (target)
                         return free_and_strdup(&arg_default_unit, target);
-
-        } else if (streq(key, "systemd.default_timeout_start_sec") && value) {
-
-                r = parse_sec(value, &arg_default_timeout_start_usec);
-                if (r < 0)
-                        log_warning_errno(r, "Failed to parse default start timeout: %s, ignoring.", value);
-
-                if (arg_default_timeout_start_usec <= 0)
-                        arg_default_timeout_start_usec = USEC_INFINITY;
         }
 
         return 0;
@@ -1341,10 +1366,9 @@ static int fixup_environment(void) {
          * However if TERM was configured through the kernel
          * command line then leave it alone. */
 
-        r = get_proc_cmdline_key("TERM=", &term);
+        r = proc_cmdline_get_key("TERM", 0, &term);
         if (r < 0)
                 return r;
-
         if (r == 0) {
                 term = strdup(default_term_for_tty("/dev/console"));
                 if (!term)
@@ -1411,7 +1435,7 @@ int main(int argc, char *argv[]) {
            called 'systemd'. That is confusing, hence let's call us
            systemd right-away. */
         program_invocation_short_name = systemd;
-        prctl(PR_SET_NAME, systemd);
+        (void) prctl(PR_SET_NAME, systemd);
 
         saved_argv = argv;
         saved_argc = argc;
@@ -1435,9 +1459,10 @@ int main(int argc, char *argv[]) {
                 if (!skip_setup) {
                         r = mount_setup_early();
                         if (r < 0) {
-                                error_message = "Failed to early mount API filesystems";
+                                error_message = "Failed to mount early API filesystems";
                                 goto finish;
                         }
+
                         dual_timestamp_get(&security_start_timestamp);
                         if (mac_selinux_setup(&loaded_policy) < 0) {
                                 error_message = "Failed to load SELinux policy";
@@ -1513,7 +1538,7 @@ int main(int argc, char *argv[]) {
                 log_close_console(); /* force reopen of /dev/console */
                 log_open();
 
-                /* For the later on, see above... */
+                /* For later on, see above... */
                 log_set_target(LOG_TARGET_JOURNAL);
 
                 /* clear the kernel timestamp,
@@ -1590,7 +1615,7 @@ int main(int argc, char *argv[]) {
         }
 
         if (arg_system) {
-                r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
+                r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
                 if (r < 0)
                         log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
         }

src/core/mount-setup.c

@@ -360,7 +360,6 @@ int mount_setup(bool loaded_policy) {
         int r = 0;
 
         r = mount_points_setup(ELEMENTSOF(mount_table), loaded_policy);
-
         if (r < 0)
                 return r;
 
@@ -391,25 +390,24 @@ int mount_setup(bool loaded_policy) {
          * udevd. */
         dev_setup(NULL, UID_INVALID, GID_INVALID);
 
-        /* Mark the root directory as shared in regards to mount
-         * propagation. The kernel defaults to "private", but we think
-         * it makes more sense to have a default of "shared" so that
-         * nspawn and the container tools work out of the box. If
-         * specific setups need other settings they can reset the
-         * propagation mode to private if needed. */
+        /* Mark the root directory as shared in regards to mount propagation. The kernel defaults to "private", but we
+         * think it makes more sense to have a default of "shared" so that nspawn and the container tools work out of
+         * the box. If specific setups need other settings they can reset the propagation mode to private if
+         * needed. Note that we set this only when we are invoked directly by the kernel. If we are invoked by a
+         * container manager we assume the container manager knows what it is doing (for example, because it set up
+         * some directories with different propagation modes). */
         if (detect_container() <= 0)
                 if (mount(NULL, "/", NULL, MS_REC|MS_SHARED, NULL) < 0)
                         log_warning_errno(errno, "Failed to set up the root directory for shared mount propagation: %m");
 
-        /* Create a few directories we always want around, Note that
-         * sd_booted() checks for /run/systemd/system, so this mkdir
-         * really needs to stay for good, otherwise software that
-         * copied sd-daemon.c into their sources will misdetect
-         * systemd. */
+        /* Create a few directories we always want around, Note that sd_booted() checks for /run/systemd/system, so
+         * this mkdir really needs to stay for good, otherwise software that copied sd-daemon.c into their sources will
+         * misdetect systemd. */
         (void) mkdir_label("/run/systemd", 0755);
         (void) mkdir_label("/run/systemd/system", 0755);
-        (void) mkdir_label("/run/systemd/inaccessible", 0000);
+
         /* Set up inaccessible items */
+        (void) mkdir_label("/run/systemd/inaccessible", 0000);
         (void) mknod("/run/systemd/inaccessible/reg", S_IFREG | 0000, 0);
         (void) mkdir_label("/run/systemd/inaccessible/dir", 0000);
         (void) mknod("/run/systemd/inaccessible/chr", S_IFCHR | 0000, makedev(0, 0));

src/cryptsetup/cryptsetup-generator.c

@@ -102,18 +102,17 @@ static int create_disk(
         if (!f)
                 return log_error_errno(errno, "Failed to create unit file %s: %m", p);
 
-        fputs(
-                "# Automatically generated by systemd-cryptsetup-generator\n\n"
-                "[Unit]\n"
-                "Description=Cryptography Setup for %I\n"
-                "Documentation=man:crypttab(5) man:systemd-cryptsetup-generator(8) man:systemd-cryptsetup@.service(8)\n"
-                "SourcePath=/etc/crypttab\n"
-                "DefaultDependencies=no\n"
-                "Conflicts=umount.target\n"
-                "BindsTo=dev-mapper-%i.device\n"
-                "IgnoreOnIsolate=true\n"
-                "After=cryptsetup-pre.target\n",
-                f);
+        fputs("# Automatically generated by systemd-cryptsetup-generator\n\n"
+              "[Unit]\n"
+              "Description=Cryptography Setup for %I\n"
+              "Documentation=man:crypttab(5) man:systemd-cryptsetup-generator(8) man:systemd-cryptsetup@.service(8)\n"
+              "SourcePath=/etc/crypttab\n"
+              "DefaultDependencies=no\n"
+              "Conflicts=umount.target\n"
+              "BindsTo=dev-mapper-%i.device\n"
+              "IgnoreOnIsolate=true\n"
+              "After=cryptsetup-pre.target\n",
+              f);
 
         if (!nofail)
                 fprintf(f,
@@ -278,27 +277,30 @@ static crypto_device *get_crypto_device(const char *uuid) {
 }
 
 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
-        int r;
-        crypto_device *d;
         _cleanup_free_ char *uuid = NULL, *uuid_value = NULL;
+        crypto_device *d;
+        int r;
 
-        if (streq(key, "luks") && value) {
+        if (streq(key, "luks")) {
 
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : 1;
                 if (r < 0)
-                        log_warning("Failed to parse luks switch %s. Ignoring.", value);
+                        log_warning("Failed to parse luks= kernel command line switch %s. Ignoring.", value);
                 else
                         arg_enabled = r;
 
-        } else if (streq(key, "luks.crypttab") && value) {
+        } else if (streq(key, "luks.crypttab")) {
 
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : 1;
                 if (r < 0)
-                        log_warning("Failed to parse luks crypttab switch %s. Ignoring.", value);
+                        log_warning("Failed to parse luks.crypttab= kernel command line switch %s. Ignoring.", value);
                 else
                         arg_read_crypttab = r;
 
-        } else if (streq(key, "luks.uuid") && value) {
+        } else if (streq(key, "luks.uuid")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 d = get_crypto_device(startswith(value, "luks-") ? value+5 : value);
                 if (!d)
@@ -306,7 +308,10 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
 
                 d->create = arg_whitelist = true;
 
-        } else if (streq(key, "luks.options") && value) {
+        } else if (streq(key, "luks.options")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 r = sscanf(value, "%m[0-9a-fA-F-]=%ms", &uuid, &uuid_value);
                 if (r == 2) {
@@ -314,13 +319,14 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                         if (!d)
                                 return log_oom();
 
-                        free(d->options);
-                        d->options = uuid_value;
-                        uuid_value = NULL;
+                        free_and_replace(d->options, uuid_value);
                 } else if (free_and_strdup(&arg_default_options, value) < 0)
                         return log_oom();
 
-        } else if (streq(key, "luks.key") && value) {
+        } else if (streq(key, "luks.key")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 r = sscanf(value, "%m[0-9a-fA-F-]=%ms", &uuid, &uuid_value);
                 if (r == 2) {
@@ -328,13 +334,14 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                         if (!d)
                                 return log_oom();
 
-                        free(d->keyfile);
-                        d->keyfile = uuid_value;
-                        uuid_value = NULL;
+                        free_and_replace(d->keyfile, uuid_value);
                 } else if (free_and_strdup(&arg_default_keyfile, value) < 0)
                         return log_oom();
 
-        } else if (streq(key, "luks.name") && value) {
+        } else if (streq(key, "luks.name")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 r = sscanf(value, "%m[0-9a-fA-F-]=%ms", &uuid, &uuid_value);
                 if (r == 2) {
@@ -349,7 +356,6 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                         uuid_value = NULL;
                 } else
                         log_warning("Failed to parse luks name switch %s. Ignoring.", value);
-
         }
 
         return 0;
@@ -458,7 +464,7 @@ static int add_proc_cmdline_devices(void) {
 }
 
 int main(int argc, char *argv[]) {
-        int r = EXIT_FAILURE;
+        int r;
 
         if (argc > 1 && argc != 4) {
                 log_error("This program takes three or no arguments.");
@@ -475,32 +481,36 @@ int main(int argc, char *argv[]) {
         umask(0022);
 
         arg_disks = hashmap_new(&string_hash_ops);
-        if (!arg_disks)
-                goto cleanup;
+        if (!arg_disks) {
+                r = log_oom();
+                goto finish;
+        }
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, true);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, PROC_CMDLINE_STRIP_RD_PREFIX);
         if (r < 0) {
-                log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
-                r = EXIT_FAILURE;
+                log_warning_errno(r, "Failed to parse kernel command line: %m");
+                goto finish;
         }
 
         if (!arg_enabled) {
-                r = EXIT_SUCCESS;
-                goto cleanup;
+                r = 0;
+                goto finish;
         }
 
-        if (add_crypttab_devices() < 0)
-                goto cleanup;
+        r = add_crypttab_devices();
+        if (r < 0)
+                goto finish;
 
-        if (add_proc_cmdline_devices() < 0)
-                goto cleanup;
+        r = add_proc_cmdline_devices();
+        if (r < 0)
+                goto finish;
 
-        r = EXIT_SUCCESS;
+        r = 0;
 
-cleanup:
+finish:
         free_arg_disks();
         free(arg_default_options);
         free(arg_default_keyfile);
 
-        return r;
+        return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
 }

src/cryptsetup/cryptsetup.c

@@ -310,7 +310,7 @@ static char *disk_mount_point(const char *label) {
         if (asprintf(&device, "/dev/mapper/%s", label) < 0)
                 return NULL;
 
-        f = setmntent("/etc/fstab", "r");
+        f = setmntent("/etc/fstab", "re");
         if (!f)
                 return NULL;
 
@@ -593,17 +593,18 @@ static int help(void) {
 }
 
 int main(int argc, char *argv[]) {
-        int r = EXIT_FAILURE;
         struct crypt_device *cd = NULL;
+        int r;
 
         if (argc <= 1) {
-                help();
-                return EXIT_SUCCESS;
+                r = help();
+                goto finish;
         }
 
         if (argc < 3) {
                 log_error("This program requires at least two arguments.");
-                return EXIT_FAILURE;
+                r = -EINVAL;
+                goto finish;
         }
 
         log_set_target(LOG_TARGET_AUTO);
@@ -614,7 +615,6 @@ int main(int argc, char *argv[]) {
 
         if (streq(argv[1], "attach")) {
                 uint32_t flags = 0;
-                int k;
                 unsigned tries;
                 usec_t until;
                 crypt_status_info status;
@@ -648,11 +648,11 @@ int main(int argc, char *argv[]) {
 
                 if (arg_header) {
                         log_debug("LUKS header: %s", arg_header);
-                        k = crypt_init(&cd, arg_header);
+                        r = crypt_init(&cd, arg_header);
                 } else
-                        k = crypt_init(&cd, argv[3]);
-                if (k != 0) {
-                        log_error_errno(k, "crypt_init() failed: %m");
+                        r = crypt_init(&cd, argv[3]);
+                if (r < 0) {
+                        log_error_errno(r, "crypt_init() failed: %m");
                         goto finish;
                 }
 
@@ -661,7 +661,7 @@ int main(int argc, char *argv[]) {
                 status = crypt_status(cd, argv[2]);
                 if (status == CRYPT_ACTIVE || status == CRYPT_BUSY) {
                         log_info("Volume %s already active.", argv[2]);
-                        r = EXIT_SUCCESS;
+                        r = 0;
                         goto finish;
                 }
 
@@ -691,29 +691,30 @@ int main(int argc, char *argv[]) {
                         _cleanup_strv_free_erase_ char **passwords = NULL;
 
                         if (!key_file) {
-                                k = get_password(argv[2], argv[3], until, tries == 0 && !arg_verify, &passwords);
-                                if (k == -EAGAIN)
+                                r = get_password(argv[2], argv[3], until, tries == 0 && !arg_verify, &passwords);
+                                if (r == -EAGAIN)
                                         continue;
-                                else if (k < 0)
+                                if (r < 0)
                                         goto finish;
                         }
 
                         if (streq_ptr(arg_type, CRYPT_TCRYPT))
-                                k = attach_tcrypt(cd, argv[2], key_file, passwords, flags);
+                                r = attach_tcrypt(cd, argv[2], key_file, passwords, flags);
                         else
-                                k = attach_luks_or_plain(cd,
+                                r = attach_luks_or_plain(cd,
                                                          argv[2],
                                                          key_file,
                                                          arg_header ? argv[3] : NULL,
                                                          passwords,
                                                          flags);
-                        if (k >= 0)
+                        if (r >= 0)
                                 break;
-                        else if (k == -EAGAIN) {
+                        if (r == -EAGAIN) {
                                 key_file = NULL;
                                 continue;
-                        } else if (k != -EPERM) {
-                                log_error_errno(k, "Failed to activate: %m");
+                        }
+                        if (r != -EPERM) {
+                                log_error_errno(r, "Failed to activate: %m");
                                 goto finish;
                         }
 
@@ -722,28 +723,28 @@ int main(int argc, char *argv[]) {
 
                 if (arg_tries != 0 && tries >= arg_tries) {
                         log_error("Too many attempts; giving up.");
-                        r = EXIT_FAILURE;
+                        r = -EPERM;
                         goto finish;
                 }
 
         } else if (streq(argv[1], "detach")) {
-                int k;
 
-                k = crypt_init_by_name(&cd, argv[2]);
-                if (k == -ENODEV) {
+                r = crypt_init_by_name(&cd, argv[2]);
+                if (r == -ENODEV) {
                         log_info("Volume %s already inactive.", argv[2]);
-                        r = EXIT_SUCCESS;
+                        r = 0;
                         goto finish;
-                } else if (k) {
-                        log_error_errno(k, "crypt_init_by_name() failed: %m");
+                }
+                if (r < 0) {
+                        log_error_errno(r, "crypt_init_by_name() failed: %m");
                         goto finish;
                 }
 
                 crypt_set_log_callback(cd, log_glue, NULL);
 
-                k = crypt_deactivate(cd, argv[2]);
-                if (k < 0) {
-                        log_error_errno(k, "Failed to deactivate: %m");
+                r = crypt_deactivate(cd, argv[2]);
+                if (r < 0) {
+                        log_error_errno(r, "Failed to deactivate: %m");
                         goto finish;
                 }
 
@@ -752,10 +753,9 @@ int main(int argc, char *argv[]) {
                 goto finish;
         }
 
-        r = EXIT_SUCCESS;
+        r = 0;
 
 finish:
-
         if (cd)
                 crypt_free(cd);
 
@@ -764,5 +764,5 @@ finish:
         free(arg_header);
         strv_free(arg_tcrypt_keyfiles);
 
-        return r;
+        return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
 }

src/debug-generator/debug-generator.c

@@ -39,56 +39,53 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
         assert(key);
 
         if (streq(key, "systemd.mask")) {
+                char *n;
 
-                if (!value)
-                        log_error("Missing argument for systemd.mask= kernel command line parameter.");
-                else {
-                        char *n;
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
-                        r = unit_name_mangle(value, UNIT_NAME_NOGLOB, &n);
-                        if (r < 0)
-                                return log_error_errno(r, "Failed to glob unit name: %m");
+                r = unit_name_mangle(value, UNIT_NAME_NOGLOB, &n);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to glob unit name: %m");
 
-                        r = strv_consume(&arg_mask, n);
-                        if (r < 0)
-                                return log_oom();
-                }
+                r = strv_consume(&arg_mask, n);
+                if (r < 0)
+                        return log_oom();
 
         } else if (streq(key, "systemd.wants")) {
+                char *n;
 
-                if (!value)
-                        log_error("Missing argument for systemd.want= kernel command line parameter.");
-                else {
-                        char *n;
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
-                        r = unit_name_mangle(value, UNIT_NAME_NOGLOB, &n);
-                        if (r < 0)
-                                return log_error_errno(r, "Failed to glob unit name: %m");
+                r = unit_name_mangle(value, UNIT_NAME_NOGLOB, &n);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to glob unit name: %m");
 
-                        r = strv_consume(&arg_wants, n);
-                        if (r < 0)
-                                return log_oom();
-                }
+                r = strv_consume(&arg_wants, n);
+                if (r < 0)
+                        return log_oom();
 
-        } else if (streq(key, "systemd.debug-shell")) {
+        } else if (proc_cmdline_key_streq(key, "systemd.debug_shell")) {
 
                 if (value) {
                         r = parse_boolean(value);
                         if (r < 0)
-                                log_error("Failed to parse systemd.debug-shell= argument '%s', ignoring.", value);
+                                log_error("Failed to parse systemd.debug_shell= argument '%s', ignoring.", value);
                         else
                                 arg_debug_shell = r;
                 } else
                         arg_debug_shell = true;
+
         } else if (streq(key, "systemd.unit")) {
 
-                if (!value)
-                        log_error("Missing argument for systemd.unit= kernel command line parameter.");
-                else {
-                        r = free_and_strdup(&arg_default_unit, value);
-                        if (r < 0)
-                                return log_error_errno(r, "Failed to set default unit %s: %m", value);
-                }
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = free_and_strdup(&arg_default_unit, value);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to set default unit %s: %m", value);
+
         } else if (!value) {
                 const char *target;
 
@@ -173,7 +170,7 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 

src/dissect/dissect.c

@@ -35,7 +35,7 @@ static enum {
 } arg_action = ACTION_DISSECT;
 static const char *arg_image = NULL;
 static const char *arg_path = NULL;
-static DissectImageFlags arg_flags = DISSECT_IMAGE_DISCARD_ON_LOOP;
+static DissectImageFlags arg_flags = DISSECT_IMAGE_REQUIRE_ROOT|DISSECT_IMAGE_DISCARD_ON_LOOP;
 static void *arg_root_hash = NULL;
 static size_t arg_root_hash_size = 0;
 
@@ -191,7 +191,7 @@ int main(int argc, char *argv[]) {
                 goto finish;
         }
 
-        r = dissect_image(d->fd, arg_root_hash, arg_root_hash_size, &m);
+        r = dissect_image(d->fd, arg_root_hash, arg_root_hash_size, arg_flags, &m);
         if (r == -ENOPKG) {
                 log_error_errno(r, "Couldn't identify a suitable partition table or file system in %s.", arg_image);
                 goto finish;
@@ -221,6 +221,9 @@ int main(int argc, char *argv[]) {
                                p->rw ? "writable" : "read-only",
                                partition_designator_to_string(i));
 
+                        if (!sd_id128_is_null(p->uuid))
+                                printf(" (UUID " SD_ID128_FORMAT_STR ")", SD_ID128_FORMAT_VAL(p->uuid));
+
                         if (p->fstype)
                                 printf(" of type %s", p->fstype);
 

src/firstboot/firstboot.c

@@ -33,6 +33,7 @@
 #include "mkdir.h"
 #include "parse-util.h"
 #include "path-util.h"
+#include "proc-cmdline.h"
 #include "random-util.h"
 #include "string-util.h"
 #include "strv.h"
@@ -825,6 +826,7 @@ static int parse_argv(int argc, char *argv[]) {
 }
 
 int main(int argc, char *argv[]) {
+        bool enabled;
         int r;
 
         r = parse_argv(argc, argv);
@@ -837,6 +839,16 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
+        r = proc_cmdline_get_bool("systemd.firstboot", &enabled);
+        if (r < 0) {
+                log_error_errno(r, "Failed to parse systemd.firstboot= kernel command line argument, ignoring.");
+                goto finish;
+        }
+        if (r > 0 && !enabled) {
+                r = 0; /* disabled */
+                goto finish;
+        }
+
         r = process_locale();
         if (r < 0)
                 goto finish;

src/fsck/fsck.c

@@ -99,7 +99,10 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
 
         assert(key);
 
-        if (streq(key, "fsck.mode") && value) {
+        if (streq(key, "fsck.mode")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (streq(value, "auto"))
                         arg_force = arg_skip = false;
@@ -110,7 +113,10 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                 else
                         log_warning("Invalid fsck.mode= parameter '%s'. Ignoring.", value);
 
-        } else if (streq(key, "fsck.repair") && value) {
+        } else if (streq(key, "fsck.repair")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (streq(value, "preen"))
                         arg_repair = "-a";
@@ -293,7 +299,7 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, true);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, PROC_CMDLINE_STRIP_RD_PREFIX);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 

src/fstab-generator/fstab-generator.c

@@ -42,16 +42,20 @@
 #include "unit-name.h"
 #include "util.h"
 #include "virt.h"
+#include "volatile-util.h"
 
 static const char *arg_dest = "/tmp";
+static const char *arg_dest_late = "/tmp";
 static bool arg_fstab_enabled = true;
 static char *arg_root_what = NULL;
 static char *arg_root_fstype = NULL;
 static char *arg_root_options = NULL;
+static char *arg_root_hash = NULL;
 static int arg_root_rw = -1;
 static char *arg_usr_what = NULL;
 static char *arg_usr_fstype = NULL;
 static char *arg_usr_options = NULL;
+static VolatileMode arg_volatile_mode = _VOLATILE_MODE_INVALID;
 
 static int add_swap(
                 const char *what,
@@ -235,6 +239,7 @@ static int write_requires_mounts_for(FILE *f, const char *opts) {
 }
 
 static int add_mount(
+                const char *dest,
                 const char *what,
                 const char *where,
                 const char *fstype,
@@ -286,7 +291,7 @@ static int add_mount(
         if (r < 0)
                 return log_error_errno(r, "Failed to generate unit name: %m");
 
-        unit = strjoin(arg_dest, "/", name);
+        unit = strjoin(dest, "/", name);
         if (!unit)
                 return log_oom();
 
@@ -318,7 +323,7 @@ static int add_mount(
         }
 
         if (passno != 0) {
-                r = generator_write_fsck_deps(f, arg_dest, what, where, fstype);
+                r = generator_write_fsck_deps(f, dest, what, where, fstype);
                 if (r < 0)
                         return r;
         }
@@ -334,7 +339,7 @@ static int add_mount(
         if (!isempty(fstype) && !streq(fstype, "auto"))
                 fprintf(f, "Type=%s\n", fstype);
 
-        r = generator_write_timeouts(arg_dest, what, where, opts, &filtered);
+        r = generator_write_timeouts(dest, what, where, opts, &filtered);
         if (r < 0)
                 return r;
 
@@ -350,7 +355,7 @@ static int add_mount(
                 return log_error_errno(r, "Failed to write unit file %s: %m", unit);
 
         if (!noauto && !automount) {
-                lnk = strjoin(arg_dest, "/", post, nofail ? ".wants/" : ".requires/", name);
+                lnk = strjoin(dest, "/", post, nofail ? ".wants/" : ".requires/", name);
                 if (!lnk)
                         return log_oom();
 
@@ -364,7 +369,7 @@ static int add_mount(
                 if (r < 0)
                         return log_error_errno(r, "Failed to generate unit name: %m");
 
-                automount_unit = strjoin(arg_dest, "/", automount_name);
+                automount_unit = strjoin(dest, "/", automount_name);
                 if (!automount_unit)
                         return log_oom();
 
@@ -406,7 +411,7 @@ static int add_mount(
                         return log_error_errno(r, "Failed to write unit file %s: %m", automount_unit);
 
                 free(lnk);
-                lnk = strjoin(arg_dest, "/", post, nofail ? ".wants/" : ".requires/", automount_name);
+                lnk = strjoin(dest, "/", post, nofail ? ".wants/" : ".requires/", automount_name);
                 if (!lnk)
                         return log_oom();
 
@@ -479,7 +484,8 @@ static int parse_fstab(bool initrd) {
                         else
                                 post = SPECIAL_LOCAL_FS_TARGET;
 
-                        k = add_mount(what,
+                        k = add_mount(arg_dest,
+                                      what,
                                       where,
                                       me->mnt_type,
                                       me->mnt_opts,
@@ -540,7 +546,8 @@ static int add_sysroot_mount(void) {
                         return r;
         }
 
-        return add_mount(what,
+        return add_mount(arg_dest,
+                         what,
                          "/sysroot",
                          arg_root_fstype,
                          opts,
@@ -593,7 +600,8 @@ static int add_sysroot_usr_mount(void) {
                 opts = arg_usr_options;
 
         log_debug("Found entry what=%s where=/sysroot/usr type=%s", what, strna(arg_usr_fstype));
-        return add_mount(what,
+        return add_mount(arg_dest,
+                         what,
                          "/sysroot/usr",
                          arg_usr_fstype,
                          opts,
@@ -605,6 +613,46 @@ static int add_sysroot_usr_mount(void) {
                          "/proc/cmdline");
 }
 
+static int add_volatile_root(void) {
+        const char *from, *to;
+
+        if (arg_volatile_mode != VOLATILE_YES)
+                return 0;
+
+        /* Let's add in systemd-remount-volatile.service which will remount the root device to tmpfs if this is
+         * requested, leaving only /usr from the root mount inside. */
+
+        from = strjoina(SYSTEM_DATA_UNIT_PATH "/systemd-volatile-root.service");
+        to = strjoina(arg_dest, "/" SPECIAL_INITRD_ROOT_FS_TARGET, ".requires/systemd-volatile-root.service");
+
+        (void) mkdir_parents(to, 0755);
+
+        if (symlink(from, to) < 0)
+                return log_error_errno(errno, "Failed to hook in volatile remount service: %m");
+
+        return 0;
+}
+
+static int add_volatile_var(void) {
+
+        if (arg_volatile_mode != VOLATILE_STATE)
+                return 0;
+
+        /* If requested, mount /var as tmpfs, but do so only if there's nothing else defined for this. */
+
+        return add_mount(arg_dest_late,
+                         "tmpfs",
+                         "/var",
+                         "tmpfs",
+                         "mode=0755",
+                         0,
+                         false,
+                         false,
+                         false,
+                         SPECIAL_LOCAL_FS_TARGET,
+                         "/proc/cmdline");
+}
+
 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
         int r;
 
@@ -612,27 +660,36 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
          * instance should take precedence.  In the case of multiple rootflags=
          * or usrflags= the arguments should be concatenated */
 
-        if (STR_IN_SET(key, "fstab", "rd.fstab") && value) {
+        if (STR_IN_SET(key, "fstab", "rd.fstab")) {
 
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : 1;
                 if (r < 0)
                         log_warning("Failed to parse fstab switch %s. Ignoring.", value);
                 else
                         arg_fstab_enabled = r;
 
-        } else if (streq(key, "root") && value) {
+        } else if (streq(key, "root")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (free_and_strdup(&arg_root_what, value) < 0)
                         return log_oom();
 
-        } else if (streq(key, "rootfstype") && value) {
+        } else if (streq(key, "rootfstype")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (free_and_strdup(&arg_root_fstype, value) < 0)
                         return log_oom();
 
-        } else if (streq(key, "rootflags") && value) {
+        } else if (streq(key, "rootflags")) {
                 char *o;
 
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 o = arg_root_options ?
                         strjoin(arg_root_options, ",", value) :
                         strdup(value);
@@ -641,20 +698,36 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
 
                 free(arg_root_options);
                 arg_root_options = o;
+        } else if (streq(key, "roothash")) {
 
-        } else if (streq(key, "mount.usr") && value) {
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                if (free_and_strdup(&arg_root_hash, value) < 0)
+                        return log_oom();
+
+        } else if (streq(key, "mount.usr")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (free_and_strdup(&arg_usr_what, value) < 0)
                         return log_oom();
 
-        } else if (streq(key, "mount.usrfstype") && value) {
+        } else if (streq(key, "mount.usrfstype")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (free_and_strdup(&arg_usr_fstype, value) < 0)
                         return log_oom();
 
-        } else if (streq(key, "mount.usrflags") && value) {
+        } else if (streq(key, "mount.usrflags")) {
                 char *o;
 
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 o = arg_usr_options ?
                         strjoin(arg_usr_options, ",", value) :
                         strdup(value);
@@ -668,10 +741,40 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
                 arg_root_rw = true;
         else if (streq(key, "ro") && !value)
                 arg_root_rw = false;
+        else if (streq(key, "systemd.volatile")) {
+                VolatileMode m;
+
+                if (value) {
+                        m = volatile_mode_from_string(value);
+                        if (m < 0)
+                                log_warning("Failed to parse systemd.volatile= argument: %s", value);
+                        else
+                                arg_volatile_mode = m;
+                } else
+                        arg_volatile_mode = VOLATILE_YES;
+        }
 
         return 0;
 }
 
+static int determine_root(void) {
+        /* If we have a root hash but no root device then Verity is used, and we use the "root" DM device as root. */
+
+        if (arg_root_what)
+                return 0;
+
+        if (!arg_root_hash)
+                return 0;
+
+        arg_root_what = strdup("/dev/mapper/root");
+        if (!arg_root_what)
+                return log_oom();
+
+        log_info("Using verity root device %s.", arg_root_what);
+
+        return 1;
+}
+
 int main(int argc, char *argv[]) {
         int r = 0;
 
@@ -682,6 +785,8 @@ int main(int argc, char *argv[]) {
 
         if (argc > 1)
                 arg_dest = argv[1];
+        if (argc > 3)
+                arg_dest_late = argv[3];
 
         log_set_target(LOG_TARGET_SAFE);
         log_parse_environment();
@@ -689,10 +794,12 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 
+        (void) determine_root();
+
         /* Always honour root= and usr= in the kernel command line if we are in an initrd */
         if (in_initrd()) {
                 int k;
@@ -702,8 +809,12 @@ int main(int argc, char *argv[]) {
                 k = add_sysroot_usr_mount();
                 if (k < 0)
                         r = k;
+
+                k = add_volatile_root();
+                if (k < 0)
+                        r = k;
         } else
-                r = 0;
+                r = add_volatile_var();
 
         /* Honour /etc/fstab only when that's enabled */
         if (arg_fstab_enabled) {
@@ -729,6 +840,7 @@ int main(int argc, char *argv[]) {
         free(arg_root_what);
         free(arg_root_fstype);
         free(arg_root_options);
+        free(arg_root_hash);
 
         free(arg_usr_what);
         free(arg_usr_fstype);

src/gpt-auto-generator/gpt-auto-generator.c

@@ -29,6 +29,7 @@
 #include "blkid-util.h"
 #include "btrfs-util.h"
 #include "dirent-util.h"
+#include "dissect-image.h"
 #include "efivars.h"
 #include "fd-util.h"
 #include "fileio.h"
@@ -54,7 +55,7 @@ static bool arg_enabled = true;
 static bool arg_root_enabled = true;
 static bool arg_root_rw = false;
 
-static int add_cryptsetup(const char *id, const char *what, bool rw, char **device) {
+static int add_cryptsetup(const char *id, const char *what, bool rw, bool require, char **device) {
         _cleanup_free_ char *e = NULL, *n = NULL, *p = NULL, *d = NULL, *to = NULL;
         _cleanup_fclose_ FILE *f = NULL;
         char *from, *ret;
@@ -62,7 +63,6 @@ static int add_cryptsetup(const char *id, const char *what, bool rw, char **devi
 
         assert(id);
         assert(what);
-        assert(device);
 
         r = unit_name_from_path(what, ".device", &d);
         if (r < 0)
@@ -119,23 +119,26 @@ static int add_cryptsetup(const char *id, const char *what, bool rw, char **devi
         if (symlink(from, to) < 0)
                 return log_error_errno(errno, "Failed to create symlink %s: %m", to);
 
-        free(to);
-        to = strjoin(arg_dest, "/cryptsetup.target.requires/", n);
-        if (!to)
-                return log_oom();
+        if (require) {
+                free(to);
 
-        mkdir_parents_label(to, 0755);
-        if (symlink(from, to) < 0)
-                return log_error_errno(errno, "Failed to create symlink %s: %m", to);
+                to = strjoin(arg_dest, "/cryptsetup.target.requires/", n);
+                if (!to)
+                        return log_oom();
 
-        free(to);
-        to = strjoin(arg_dest, "/dev-mapper-", e, ".device.requires/", n);
-        if (!to)
-                return log_oom();
+                mkdir_parents_label(to, 0755);
+                if (symlink(from, to) < 0)
+                        return log_error_errno(errno, "Failed to create symlink %s: %m", to);
 
-        mkdir_parents_label(to, 0755);
-        if (symlink(from, to) < 0)
-                return log_error_errno(errno, "Failed to create symlink %s: %m", to);
+                free(to);
+                to = strjoin(arg_dest, "/dev-mapper-", e, ".device.requires/", n);
+                if (!to)
+                        return log_oom();
+
+                mkdir_parents_label(to, 0755);
+                if (symlink(from, to) < 0)
+                        return log_error_errno(errno, "Failed to create symlink %s: %m", to);
+        }
 
         free(p);
         p = strjoin(arg_dest, "/dev-mapper-", e, ".device.d/50-job-timeout-sec-0.conf");
@@ -155,7 +158,8 @@ static int add_cryptsetup(const char *id, const char *what, bool rw, char **devi
         if (!ret)
                 return log_oom();
 
-        *device = ret;
+        if (device)
+                *device = ret;
         return 0;
 }
 
@@ -182,7 +186,7 @@ static int add_mount(
 
         if (streq_ptr(fstype, "crypto_LUKS")) {
 
-                r = add_cryptsetup(id, what, rw, &crypto_what);
+                r = add_cryptsetup(id, what, rw, true, &crypto_what);
                 if (r < 0)
                         return r;
 
@@ -270,61 +274,28 @@ static bool path_is_busy(const char *where) {
         return false;
 }
 
-static int probe_and_add_mount(
+static int add_partition_mount(
+                DissectedPartition *p,
                 const char *id,
-                const char *what,
                 const char *where,
-                bool rw,
-                const char *description,
-                const char *post) {
+                const char *description) {
 
-        _cleanup_blkid_free_probe_ blkid_probe b = NULL;
-        const char *fstype = NULL;
-        int r;
-
-        assert(id);
-        assert(what);
-        assert(where);
-        assert(description);
+        assert(p);
 
         if (path_is_busy(where)) {
                 log_debug("%s already populated, ignoring.", where);
                 return 0;
         }
 
-        /* Let's check the partition type here, so that we know
-         * whether to do LUKS magic. */
-
-        errno = 0;
-        b = blkid_new_probe_from_filename(what);
-        if (!b) {
-                if (errno == 0)
-                        return log_oom();
-                return log_error_errno(errno, "Failed to allocate prober: %m");
-        }
-
-        blkid_probe_enable_superblocks(b, 1);
-        blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE);
-
-        errno = 0;
-        r = blkid_do_safeprobe(b);
-        if (r == -2 || r == 1) /* no result or uncertain */
-                return 0;
-        else if (r != 0)
-                return log_error_errno(errno ?: EIO, "Failed to probe %s: %m", what);
-
-        /* add_mount is OK with fstype being NULL. */
-        (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL);
-
         return add_mount(
                         id,
-                        what,
+                        p->node,
                         where,
-                        fstype,
-                        rw,
+                        p->fstype,
+                        p->rw,
                         NULL,
                         description,
-                        post);
+                        SPECIAL_LOCAL_FS_TARGET);
 }
 
 static int add_swap(const char *path) {
@@ -449,22 +420,17 @@ static int add_automount(
         return 0;
 }
 
-static int add_boot(const char *what) {
+static int add_esp(DissectedPartition *p) {
         const char *esp;
         int r;
 
-        assert(what);
+        assert(p);
 
         if (in_initrd()) {
                 log_debug("In initrd, ignoring the ESP.");
                 return 0;
         }
 
-        if (detect_container() > 0) {
-                log_debug("In a container, ignoring the ESP.");
-                return 0;
-        }
-
         /* If /efi exists we'll use that. Otherwise we'll use /boot, as that's usually the better choice */
         esp = access("/efi/", F_OK) >= 0 ? "/efi" : "/boot";
 
@@ -480,9 +446,7 @@ static int add_boot(const char *what) {
         }
 
         if (is_efi_boot()) {
-                _cleanup_blkid_free_probe_ blkid_probe b = NULL;
-                const char *fstype = NULL, *uuid_string = NULL;
-                sd_id128_t loader_uuid, part_uuid;
+                sd_id128_t loader_uuid;
 
                 /* If this is an EFI boot, be extra careful, and only mount the ESP if it was the ESP used for booting. */
 
@@ -494,43 +458,7 @@ static int add_boot(const char *what) {
                 if (r < 0)
                         return log_error_errno(r, "Failed to read ESP partition UUID: %m");
 
-                errno = 0;
-                b = blkid_new_probe_from_filename(what);
-                if (!b) {
-                        if (errno == 0)
-                                return log_oom();
-                        return log_error_errno(errno, "Failed to allocate prober: %m");
-                }
-
-                blkid_probe_enable_partitions(b, 1);
-                blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
-
-                errno = 0;
-                r = blkid_do_safeprobe(b);
-                if (r == -2 || r == 1) /* no result or uncertain */
-                        return 0;
-                else if (r != 0)
-                        return log_error_errno(errno ?: EIO, "Failed to probe %s: %m", what);
-
-                (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL);
-                if (!streq_ptr(fstype, "vfat")) {
-                        log_debug("Partition for %s is not a FAT filesystem, ignoring.", esp);
-                        return 0;
-                }
-
-                errno = 0;
-                r = blkid_probe_lookup_value(b, "PART_ENTRY_UUID", &uuid_string, NULL);
-                if (r != 0) {
-                        log_debug_errno(errno, "Partition for %s does not have a UUID, ignoring.", esp);
-                        return 0;
-                }
-
-                if (sd_id128_from_string(uuid_string, &part_uuid) < 0) {
-                        log_debug("Partition for %s does not have a valid UUID, ignoring.", esp);
-                        return 0;
-                }
-
-                if (!sd_id128_equal(part_uuid, loader_uuid)) {
+                if (!sd_id128_equal(p->uuid, loader_uuid)) {
                         log_debug("Partition for %s does not appear to be the partition we are booted from.", esp);
                         return 0;
                 }
@@ -538,35 +466,29 @@ static int add_boot(const char *what) {
                 log_debug("Not an EFI boot, skipping ESP check.");
 
         return add_automount("boot",
-                          what,
-                          esp,
-                          "vfat",
-                          true,
-                          "umask=0077",
-                          "EFI System Partition Automount",
-                          120 * USEC_PER_SEC);
+                             p->node,
+                             esp,
+                             p->fstype,
+                             true,
+                             "umask=0077",
+                             "EFI System Partition Automount",
+                             120 * USEC_PER_SEC);
 }
 #else
-static int add_boot(const char *what) {
+static int add_esp(const char *what) {
         return 0;
 }
 #endif
 
-static int enumerate_partitions(dev_t devnum) {
-
-        _cleanup_udev_enumerate_unref_ struct udev_enumerate *e = NULL;
+static int open_parent(dev_t devnum, int *ret) {
         _cleanup_udev_device_unref_ struct udev_device *d = NULL;
-        _cleanup_blkid_free_probe_ blkid_probe b = NULL;
         _cleanup_udev_unref_ struct udev *udev = NULL;
-        _cleanup_free_ char *boot = NULL, *home = NULL, *srv = NULL;
-        struct udev_list_entry *first, *item;
-        struct udev_device *parent = NULL;
-        const char *name, *node, *pttype, *devtype;
-        int boot_nr = -1, home_nr = -1, srv_nr = -1;
-        bool home_rw = true, srv_rw = true;
-        blkid_partlist pl;
-        int r, k;
+        const char *name, *devtype, *node;
+        struct udev_device *parent;
         dev_t pn;
+        int fd;
+
+        assert(ret);
 
         udev = udev_new();
         if (!udev)
@@ -580,228 +502,94 @@ static int enumerate_partitions(dev_t devnum) {
         if (!name)
                 name = udev_device_get_syspath(d);
         if (!name) {
-                log_debug("Device %u:%u does not have a name, ignoring.",
-                          major(devnum), minor(devnum));
-                return 0;
+                log_debug("Device %u:%u does not have a name, ignoring.", major(devnum), minor(devnum));
+                goto not_found;
         }
 
         parent = udev_device_get_parent(d);
         if (!parent) {
                 log_debug("%s: not a partitioned device, ignoring.", name);
-                return 0;
+                goto not_found;
         }
 
         /* Does it have a devtype? */
         devtype = udev_device_get_devtype(parent);
         if (!devtype) {
                 log_debug("%s: parent doesn't have a device type, ignoring.", name);
-                return 0;
+                goto not_found;
         }
 
         /* Is this a disk or a partition? We only care for disks... */
         if (!streq(devtype, "disk")) {
                 log_debug("%s: parent isn't a raw disk, ignoring.", name);
-                return 0;
+                goto not_found;
         }
 
         /* Does it have a device node? */
         node = udev_device_get_devnode(parent);
         if (!node) {
                 log_debug("%s: parent device does not have device node, ignoring.", name);
-                return 0;
+                goto not_found;
         }
 
         log_debug("%s: root device %s.", name, node);
 
         pn = udev_device_get_devnum(parent);
-        if (major(pn) == 0)
-                return 0;
-
-        errno = 0;
-        b = blkid_new_probe_from_filename(node);
-        if (!b) {
-                if (errno == 0)
-                        return log_oom();
-
-                return log_error_errno(errno, "%s: failed to allocate prober: %m", node);
+        if (major(pn) == 0) {
+                log_debug("%s: parent device is not a proper block device, ignoring.", name);
+                goto not_found;
         }
 
-        blkid_probe_enable_partitions(b, 1);
-        blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
+        fd = open(node, O_RDONLY|O_CLOEXEC|O_NOCTTY);
+        if (fd < 0)
+                return log_error_errno(errno, "Failed to open %s: %m", node);
 
-        errno = 0;
-        r = blkid_do_safeprobe(b);
-        if (r == 1)
-                return 0; /* no results */
-        else if (r == -2) {
-                log_warning("%s: probe gave ambiguous results, ignoring.", node);
-                return 0;
-        } else if (r != 0)
-                return log_error_errno(errno ?: EIO, "%s: failed to probe: %m", node);
+        *ret = fd;
+        return 1;
 
-        errno = 0;
-        r = blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
-        if (r != 0) {
-                if (errno == 0)
-                        return 0; /* No partition table found. */
+not_found:
+        *ret = -1;
+        return 0;
+}
 
-                return log_error_errno(errno, "%s: failed to determine partition table type: %m", node);
-        }
+static int enumerate_partitions(dev_t devnum) {
 
-        /* We only do this all for GPT... */
-        if (!streq_ptr(pttype, "gpt")) {
-                log_debug("%s: not a GPT partition table, ignoring.", node);
+        _cleanup_close_ int fd = -1;
+        _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL;
+        int r, k;
+
+        r = open_parent(devnum, &fd);
+        if (r <= 0)
+                return r;
+
+        r = dissect_image(fd, NULL, 0, DISSECT_IMAGE_GPT_ONLY, &m);
+        if (r == -ENOPKG) {
+                log_debug_errno(r, "No suitable partition table found, ignoring.");
                 return 0;
         }
-
-        errno = 0;
-        pl = blkid_probe_get_partitions(b);
-        if (!pl) {
-                if (errno == 0)
-                        return log_oom();
-
-                return log_error_errno(errno, "%s: failed to list partitions: %m", node);
-        }
-
-        e = udev_enumerate_new(udev);
-        if (!e)
-                return log_oom();
-
-        r = udev_enumerate_add_match_parent(e, parent);
         if (r < 0)
-                return log_oom();
+                return log_error_errno(r, "Failed to dissect: %m");
 
-        r = udev_enumerate_add_match_subsystem(e, "block");
-        if (r < 0)
-                return log_oom();
-
-        r = udev_enumerate_scan_devices(e);
-        if (r < 0)
-                return log_error_errno(r, "%s: failed to enumerate partitions: %m", node);
-
-        first = udev_enumerate_get_list_entry(e);
-        udev_list_entry_foreach(item, first) {
-                _cleanup_udev_device_unref_ struct udev_device *q;
-                unsigned long long flags;
-                const char *stype, *subnode;
-                sd_id128_t type_id;
-                blkid_partition pp;
-                dev_t qn;
-                int nr;
-
-                q = udev_device_new_from_syspath(udev, udev_list_entry_get_name(item));
-                if (!q)
-                        continue;
-
-                qn = udev_device_get_devnum(q);
-                if (major(qn) == 0)
-                        continue;
-
-                if (qn == devnum)
-                        continue;
-
-                if (qn == pn)
-                        continue;
-
-                subnode = udev_device_get_devnode(q);
-                if (!subnode)
-                        continue;
-
-                pp = blkid_partlist_devno_to_partition(pl, qn);
-                if (!pp)
-                        continue;
-
-                nr = blkid_partition_get_partno(pp);
-                if (nr < 0)
-                        continue;
-
-                stype = blkid_partition_get_type_string(pp);
-                if (!stype)
-                        continue;
-
-                if (sd_id128_from_string(stype, &type_id) < 0)
-                        continue;
-
-                flags = blkid_partition_get_flags(pp);
-
-                if (sd_id128_equal(type_id, GPT_SWAP)) {
-
-                        if (flags & GPT_FLAG_NO_AUTO)
-                                continue;
-
-                        if (flags & GPT_FLAG_READ_ONLY) {
-                                log_debug("%s marked as read-only swap partition, which is bogus. Ignoring.", subnode);
-                                continue;
-                        }
-
-                        k = add_swap(subnode);
-                        if (k < 0)
-                                r = k;
-
-                } else if (sd_id128_equal(type_id, GPT_ESP)) {
-
-                        /* We only care for the first /boot partition */
-                        if (boot && nr >= boot_nr)
-                                continue;
-
-                        /* Note that we do not honour the "no-auto"
-                         * flag for the ESP, as it is often unset, to
-                         * hide it from Windows. */
-
-                        boot_nr = nr;
-
-                        r = free_and_strdup(&boot, subnode);
-                        if (r < 0)
-                                return log_oom();
-
-                } else if (sd_id128_equal(type_id, GPT_HOME)) {
-
-                        if (flags & GPT_FLAG_NO_AUTO)
-                                continue;
-
-                        /* We only care for the first /home partition */
-                        if (home && nr >= home_nr)
-                                continue;
-
-                        home_nr = nr;
-                        home_rw = !(flags & GPT_FLAG_READ_ONLY),
-
-                        r = free_and_strdup(&home, subnode);
-                        if (r < 0)
-                                return log_oom();
-
-                } else if (sd_id128_equal(type_id, GPT_SRV)) {
-
-                        if (flags & GPT_FLAG_NO_AUTO)
-                                continue;
-
-                        /* We only care for the first /srv partition */
-                        if (srv && nr >= srv_nr)
-                                continue;
-
-                        srv_nr = nr;
-                        srv_rw = !(flags & GPT_FLAG_READ_ONLY),
-
-                        r = free_and_strdup(&srv, subnode);
-                        if (r < 0)
-                                return log_oom();
-                }
-        }
-
-        if (boot) {
-                k = add_boot(boot);
+        if (m->partitions[PARTITION_SWAP].found) {
+                k = add_swap(m->partitions[PARTITION_SWAP].node);
                 if (k < 0)
                         r = k;
         }
 
-        if (home) {
-                k = probe_and_add_mount("home", home, "/home", home_rw, "Home Partition", SPECIAL_LOCAL_FS_TARGET);
+        if (m->partitions[PARTITION_ESP].found) {
+                k = add_esp(m->partitions + PARTITION_ESP);
                 if (k < 0)
                         r = k;
         }
 
-        if (srv) {
-                k = probe_and_add_mount("srv", srv, "/srv", srv_rw, "Server Data Partition", SPECIAL_LOCAL_FS_TARGET);
+        if (m->partitions[PARTITION_HOME].found) {
+                k = add_partition_mount(m->partitions + PARTITION_HOME, "home", "/home", "Home Partition");
+                if (k < 0)
+                        r = k;
+        }
+
+        if (m->partitions[PARTITION_SRV].found) {
+                k = add_partition_mount(m->partitions + PARTITION_SRV, "srv", "/srv", "Server Data Partition");
                 if (k < 0)
                         r = k;
         }
@@ -876,8 +664,40 @@ static int get_block_device_harder(const char *path, dev_t *dev) {
                 if (!IN_SET(de->d_type, DT_LNK, DT_UNKNOWN))
                         continue;
 
-                if (found) /* Don't try to support multiple backing block devices */
-                        goto fallback;
+                if (found) {
+                        _cleanup_free_ char *u = NULL, *v = NULL, *a = NULL, *b = NULL;
+
+                        /* We found a device backed by multiple other devices. We don't really support automatic
+                         * discovery on such setups, with the exception of dm-verity partitions. In this case there are
+                         * two backing devices: the data partition and the hash partition. We are fine with such
+                         * setups, however, only if both partitions are on the same physical device. Hence, let's
+                         * verify this. */
+
+                        u = strjoin(p, "/", de->d_name, "/../dev");
+                        if (!u)
+                                return -ENOMEM;
+
+                        v = strjoin(p, "/", found->d_name, "/../dev");
+                        if (!v)
+                                return -ENOMEM;
+
+                        r = read_one_line_file(u, &a);
+                        if (r < 0) {
+                                log_debug_errno(r, "Failed to read %s: %m", u);
+                                goto fallback;
+                        }
+
+                        r = read_one_line_file(v, &b);
+                        if (r < 0) {
+                                log_debug_errno(r, "Failed to read %s: %m", v);
+                                goto fallback;
+                        }
+
+                        /* Check if the parent device is the same. If not, then the two backing devices are on
+                         * different physical devices, and we don't support that. */
+                        if (!streq(a, b))
+                                goto fallback;
+                }
 
                 found = de;
         }
@@ -912,21 +732,33 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
 
         assert(key);
 
-        if (STR_IN_SET(key, "systemd.gpt_auto", "rd.systemd.gpt_auto") && value) {
+        if (STR_IN_SET(key, "systemd.gpt_auto", "rd.systemd.gpt_auto")) {
 
-                r = parse_boolean(value);
+                r = value ? parse_boolean(value) : 1;
                 if (r < 0)
                         log_warning("Failed to parse gpt-auto switch \"%s\". Ignoring.", value);
                 else
                         arg_enabled = r;
 
-        } else if (streq(key, "root") && value) {
+        } else if (streq(key, "root")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 /* Disable root disk logic if there's a root= value
                  * specified (unless it happens to be "gpt-auto") */
 
                 arg_root_enabled = streq(value, "gpt-auto");
 
+        } else if (streq(key, "roothash")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                /* Disable root disk logic if there's roothash= defined (i.e. verity enabled) */
+
+                arg_root_enabled = false;
+
         } else if (streq(key, "rw") && !value)
                 arg_root_rw = true;
         else if (streq(key, "ro") && !value)
@@ -935,6 +767,16 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
         return 0;
 }
 
+#ifdef ENABLE_EFI
+static int add_root_cryptsetup(void) {
+
+        /* If a device /dev/gpt-auto-root-luks appears, then make it pull in systemd-cryptsetup-root.service, which
+         * sets it up, and causes /dev/gpt-auto-root to appear which is all we are looking for. */
+
+        return add_cryptsetup("root", "/dev/gpt-auto-root-luks", true, false, NULL);
+}
+#endif
+
 static int add_root_mount(void) {
 
 #ifdef ENABLE_EFI
@@ -960,6 +802,10 @@ static int add_root_mount(void) {
                 r = generator_write_initrd_root_device_deps(arg_dest, "/dev/gpt-auto-root");
                 if (r < 0)
                         return 0;
+
+                r = add_root_cryptsetup();
+                if (r < 0)
+                        return r;
         }
 
         return add_mount(
@@ -983,11 +829,11 @@ static int add_mounts(void) {
         r = get_block_device_harder("/", &devno);
         if (r < 0)
                 return log_error_errno(r, "Failed to determine block device of root file system: %m");
-        else if (r == 0) {
+        if (r == 0) {
                 r = get_block_device_harder("/usr", &devno);
                 if (r < 0)
                         return log_error_errno(r, "Failed to determine block device of /usr file system: %m");
-                else if (r == 0) {
+                if (r == 0) {
                         log_debug("Neither root nor /usr file system are on a (single) block device.");
                         return 0;
                 }
@@ -997,7 +843,7 @@ static int add_mounts(void) {
 }
 
 int main(int argc, char *argv[]) {
-        int r = 0;
+        int r = 0, k;
 
         if (argc > 1 && argc != 4) {
                 log_error("This program takes three or no arguments.");
@@ -1018,7 +864,7 @@ int main(int argc, char *argv[]) {
                 return EXIT_SUCCESS;
         }
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 
@@ -1031,8 +877,6 @@ int main(int argc, char *argv[]) {
                 r = add_root_mount();
 
         if (!in_initrd()) {
-                int k;
-
                 k = add_mounts();
                 if (k < 0)
                         r = k;

src/hibernate-resume/hibernate-resume-generator.c

@@ -31,15 +31,22 @@
 #include "util.h"
 
 static const char *arg_dest = "/tmp";
-static char *arg_resume_dev = NULL;
+static char *arg_resume_device = NULL;
 
 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
 
-        if (streq(key, "resume") && value) {
-                free(arg_resume_dev);
-                arg_resume_dev = fstab_node_to_udev_node(value);
-                if (!arg_resume_dev)
+        if (streq(key, "resume")) {
+                char *s;
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                s = fstab_node_to_udev_node(value);
+                if (!s)
                         return log_oom();
+
+                free(arg_resume_device);
+                arg_resume_device = s;
         }
 
         return 0;
@@ -49,10 +56,10 @@ static int process_resume(void) {
         _cleanup_free_ char *name = NULL, *lnk = NULL;
         int r;
 
-        if (!arg_resume_dev)
+        if (!arg_resume_device)
                 return 0;
 
-        r = unit_name_from_path_instance("systemd-hibernate-resume", arg_resume_dev, ".service", &name);
+        r = unit_name_from_path_instance("systemd-hibernate-resume", arg_resume_device, ".service", &name);
         if (r < 0)
                 return log_error_errno(r, "Failed to generate unit name: %m");
 
@@ -88,12 +95,12 @@ int main(int argc, char *argv[]) {
         if (!in_initrd())
                 return EXIT_SUCCESS;
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 
         r = process_resume();
-        free(arg_resume_dev);
+        free(arg_resume_device);
 
         return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
 }

src/import/pull-common.c

@@ -218,37 +218,40 @@ int pull_make_path(const char *url, const char *etag, const char *image_root, co
         return 0;
 }
 
-int pull_make_settings_job(
+int pull_make_auxiliary_job(
                 PullJob **ret,
                 const char *url,
+                int (*strip_suffixes)(const char *name, char **ret),
+                const char *suffix,
                 CurlGlue *glue,
                 PullJobFinished on_finished,
                 void *userdata) {
 
-        _cleanup_free_ char *last_component = NULL, *ll = NULL, *settings_url = NULL;
+        _cleanup_free_ char *last_component = NULL, *ll = NULL, *auxiliary_url = NULL;
         _cleanup_(pull_job_unrefp) PullJob *job = NULL;
         const char *q;
         int r;
 
         assert(ret);
         assert(url);
+        assert(strip_suffixes);
         assert(glue);
 
         r = import_url_last_component(url, &last_component);
         if (r < 0)
                 return r;
 
-        r = tar_strip_suffixes(last_component, &ll);
+        r = strip_suffixes(last_component, &ll);
         if (r < 0)
                 return r;
 
-        q = strjoina(ll, ".nspawn");
+        q = strjoina(ll, suffix);
 
-        r = import_url_change_last_component(url, q, &settings_url);
+        r = import_url_change_last_component(url, q, &auxiliary_url);
         if (r < 0)
                 return r;
 
-        r = pull_job_new(&job, settings_url, glue, userdata);
+        r = pull_job_new(&job, auxiliary_url, glue, userdata);
         if (r < 0)
                 return r;
 
@@ -320,7 +323,56 @@ int pull_make_verification_jobs(
         return 0;
 }
 
+static int verify_one(PullJob *checksum_job, PullJob *job) {
+        _cleanup_free_ char *fn = NULL;
+        const char *line, *p;
+        int r;
+
+        assert(checksum_job);
+
+        if (!job)
+                return 0;
+
+        assert(IN_SET(job->state, PULL_JOB_DONE, PULL_JOB_FAILED));
+
+        /* Don't verify the checksum if we didn't actually successfully download something new */
+        if (job->state != PULL_JOB_DONE)
+                return 0;
+        if (job->error != 0)
+                return 0;
+        if (job->etag_exists)
+                return 0;
+
+        assert(job->calc_checksum);
+        assert(job->checksum);
+
+        r = import_url_last_component(job->url, &fn);
+        if (r < 0)
+                return log_oom();
+
+        if (!filename_is_valid(fn)) {
+                log_error("Cannot verify checksum, could not determine server-side file name.");
+                return -EBADMSG;
+        }
+
+        line = strjoina(job->checksum, " *", fn, "\n");
+
+        p = memmem(checksum_job->payload,
+                   checksum_job->payload_size,
+                   line,
+                   strlen(line));
+
+        if (!p || (p != (char*) checksum_job->payload && p[-1] != '\n')) {
+                log_error("DOWNLOAD INVALID: Checksum of %s file did not checkout, file has been tampered with.", fn);
+                return -EBADMSG;
+        }
+
+        log_info("SHA256 checksum of %s is valid.", job->url);
+        return 1;
+}
+
 int pull_verify(PullJob *main_job,
+                PullJob *roothash_job,
                 PullJob *settings_job,
                 PullJob *checksum_job,
                 PullJob *signature_job) {
@@ -328,7 +380,6 @@ int pull_verify(PullJob *main_job,
         _cleanup_close_pair_ int gpg_pipe[2] = { -1, -1 };
         _cleanup_free_ char *fn = NULL;
         _cleanup_close_ int sig_file = -1;
-        const char *p, *line;
         char sig_file_path[] = "/tmp/sigXXXXXX", gpg_home[] = "/tmp/gpghomeXXXXXX";
         _cleanup_(sigkill_waitp) pid_t pid = 0;
         bool gpg_home_created = false;
@@ -342,6 +393,7 @@ int pull_verify(PullJob *main_job,
 
         assert(main_job->calc_checksum);
         assert(main_job->checksum);
+
         assert(checksum_job->state == PULL_JOB_DONE);
 
         if (!checksum_job->payload || checksum_job->payload_size <= 0) {
@@ -349,64 +401,17 @@ int pull_verify(PullJob *main_job,
                 return -EBADMSG;
         }
 
-        r = import_url_last_component(main_job->url, &fn);
+        r = verify_one(checksum_job, main_job);
         if (r < 0)
-                return log_oom();
+                return r;
 
-        if (!filename_is_valid(fn)) {
-                log_error("Cannot verify checksum, could not determine valid server-side file name.");
-                return -EBADMSG;
-        }
+        r = verify_one(checksum_job, roothash_job);
+        if (r < 0)
+                return r;
 
-        line = strjoina(main_job->checksum, " *", fn, "\n");
-
-        p = memmem(checksum_job->payload,
-                   checksum_job->payload_size,
-                   line,
-                   strlen(line));
-
-        if (!p || (p != (char*) checksum_job->payload && p[-1] != '\n')) {
-                log_error("DOWNLOAD INVALID: Checksum did not check out, payload has been tampered with.");
-                return -EBADMSG;
-        }
-
-        log_info("SHA256 checksum of %s is valid.", main_job->url);
-
-        assert(!settings_job || IN_SET(settings_job->state, PULL_JOB_DONE, PULL_JOB_FAILED));
-
-        if (settings_job &&
-            settings_job->state == PULL_JOB_DONE &&
-            settings_job->error == 0 &&
-            !settings_job->etag_exists) {
-
-                _cleanup_free_ char *settings_fn = NULL;
-
-                assert(settings_job->calc_checksum);
-                assert(settings_job->checksum);
-
-                r = import_url_last_component(settings_job->url, &settings_fn);
-                if (r < 0)
-                        return log_oom();
-
-                if (!filename_is_valid(settings_fn)) {
-                        log_error("Cannot verify checksum, could not determine server-side settings file name.");
-                        return -EBADMSG;
-                }
-
-                line = strjoina(settings_job->checksum, " *", settings_fn, "\n");
-
-                p = memmem(checksum_job->payload,
-                           checksum_job->payload_size,
-                           line,
-                           strlen(line));
-
-                if (!p || (p != (char*) checksum_job->payload && p[-1] != '\n')) {
-                        log_error("DOWNLOAD INVALID: Checksum of settings file did not checkout, settings file has been tampered with.");
-                        return -EBADMSG;
-                }
-
-                log_info("SHA256 checksum of %s is valid.", settings_job->url);
-        }
+        r = verify_one(checksum_job, settings_job);
+        if (r < 0)
+                return r;
 
         if (!signature_job)
                 return 0;

src/import/pull-common.h

@@ -30,7 +30,7 @@ int pull_find_old_etags(const char *url, const char *root, int dt, const char *p
 
 int pull_make_path(const char *url, const char *etag, const char *image_root, const char *prefix, const char *suffix, char **ret);
 
-int pull_make_settings_job(PullJob **ret, const char *url, CurlGlue *glue, PullJobFinished on_finished, void *userdata);
+int pull_make_auxiliary_job(PullJob **ret, const char *url, int (*strip_suffixes)(const char *name, char **ret), const char *suffix, CurlGlue *glue, PullJobFinished on_finished, void *userdata);
 int pull_make_verification_jobs(PullJob **ret_checksum_job, PullJob **ret_signature_job, ImportVerify verify, const char *url, CurlGlue *glue, PullJobFinished on_finished, void *userdata);
 
-int pull_verify(PullJob *main_job, PullJob *settings_job, PullJob *checksum_job, PullJob *signature_job);
+int pull_verify(PullJob *main_job, PullJob *roothash_job, PullJob *settings_job, PullJob *checksum_job, PullJob *signature_job);

src/import/pull-raw.c

@@ -63,6 +63,7 @@ struct RawPull {
         char *image_root;
 
         PullJob *raw_job;
+        PullJob *roothash_job;
         PullJob *settings_job;
         PullJob *checksum_job;
         PullJob *signature_job;
@@ -74,6 +75,7 @@ struct RawPull {
         bool force_local;
         bool grow_machine_directory;
         bool settings;
+        bool roothash;
 
         char *final_path;
         char *temp_path;
@@ -81,6 +83,9 @@ struct RawPull {
         char *settings_path;
         char *settings_temp_path;
 
+        char *roothash_path;
+        char *roothash_temp_path;
+
         ImportVerify verify;
 };
 
@@ -90,6 +95,7 @@ RawPull* raw_pull_unref(RawPull *i) {
 
         pull_job_unref(i->raw_job);
         pull_job_unref(i->settings_job);
+        pull_job_unref(i->roothash_job);
         pull_job_unref(i->checksum_job);
         pull_job_unref(i->signature_job);
 
@@ -101,12 +107,18 @@ RawPull* raw_pull_unref(RawPull *i) {
                 free(i->temp_path);
         }
 
+        if (i->roothash_temp_path) {
+                (void) unlink(i->roothash_temp_path);
+                free(i->roothash_temp_path);
+        }
+
         if (i->settings_temp_path) {
                 (void) unlink(i->settings_temp_path);
                 free(i->settings_temp_path);
         }
 
         free(i->final_path);
+        free(i->roothash_path);
         free(i->settings_path);
         free(i->image_root);
         free(i->local);
@@ -176,6 +188,11 @@ static void raw_pull_report_progress(RawPull *i, RawProgress p) {
                         remain -= 5;
                 }
 
+                if (i->roothash_job) {
+                        percent += i->roothash_job->progress_percent * 5 / 100;
+                        remain -= 5;
+                }
+
                 if (i->checksum_job) {
                         percent += i->checksum_job->progress_percent * 5 / 100;
                         remain -= 5;
@@ -262,6 +279,55 @@ static int raw_pull_maybe_convert_qcow2(RawPull *i) {
         return 1;
 }
 
+static int raw_pull_determine_path(RawPull *i, const char *suffix, char **field) {
+        int r;
+
+        assert(i);
+        assert(field);
+
+        if (*field)
+                return 0;
+
+        assert(i->raw_job);
+
+        r = pull_make_path(i->raw_job->url, i->raw_job->etag, i->image_root, ".raw-", suffix, field);
+        if (r < 0)
+                return log_oom();
+
+        return 1;
+}
+
+static int raw_pull_copy_auxiliary_file(
+                RawPull *i,
+                const char *suffix,
+                char **path) {
+
+        const char *local;
+        int r;
+
+        assert(i);
+        assert(suffix);
+        assert(path);
+
+        r = raw_pull_determine_path(i, suffix, path);
+        if (r < 0)
+                return r;
+
+        local = strjoina(i->image_root, "/", i->local, suffix);
+
+        r = copy_file_atomic(*path, local, 0644, i->force_local, 0);
+        if (r == -EEXIST)
+                log_warning_errno(r, "File %s already exists, not replacing.", local);
+        else if (r == -ENOENT)
+                log_debug_errno(r, "Skipping creation of auxiliary file, since none was found.");
+        else if (r < 0)
+                log_warning_errno(r, "Failed to copy file %s, ignoring: %m", local);
+        else
+                log_info("Created new file %s.", local);
+
+        return 0;
+}
+
 static int raw_pull_make_local_copy(RawPull *i) {
         _cleanup_free_ char *tp = NULL;
         _cleanup_close_ int dfd = -1;
@@ -274,12 +340,6 @@ static int raw_pull_make_local_copy(RawPull *i) {
         if (!i->local)
                 return 0;
 
-        if (!i->final_path) {
-                r = pull_make_path(i->raw_job->url, i->raw_job->etag, i->image_root, ".raw-", ".raw", &i->final_path);
-                if (r < 0)
-                        return log_oom();
-        }
-
         if (i->raw_job->etag_exists) {
                 /* We have downloaded this one previously, reopen it */
 
@@ -338,27 +398,16 @@ static int raw_pull_make_local_copy(RawPull *i) {
 
         log_info("Created new local image '%s'.", i->local);
 
+        if (i->roothash) {
+                r = raw_pull_copy_auxiliary_file(i, ".roothash", &i->roothash_path);
+                if (r < 0)
+                        return r;
+        }
+
         if (i->settings) {
-                const char *local_settings;
-                assert(i->settings_job);
-
-                if (!i->settings_path) {
-                        r = pull_make_path(i->settings_job->url, i->settings_job->etag, i->image_root, ".settings-", NULL, &i->settings_path);
-                        if (r < 0)
-                                return log_oom();
-                }
-
-                local_settings = strjoina(i->image_root, "/", i->local, ".nspawn");
-
-                r = copy_file_atomic(i->settings_path, local_settings, 0644, i->force_local, 0);
-                if (r == -EEXIST)
-                        log_warning_errno(r, "Settings file %s already exists, not replacing.", local_settings);
-                else if (r == -ENOENT)
-                        log_debug_errno(r, "Skipping creation of settings file, since none was found.");
-                else if (r < 0)
-                        log_warning_errno(r, "Failed to copy settings files %s, ignoring: %m", local_settings);
-                else
-                        log_info("Created new settings file %s.", local_settings);
+                r = raw_pull_copy_auxiliary_file(i, ".nspawn", &i->settings_path);
+                if (r < 0)
+                        return r;
         }
 
         return 0;
@@ -370,6 +419,8 @@ static bool raw_pull_is_done(RawPull *i) {
 
         if (!PULL_JOB_IS_COMPLETE(i->raw_job))
                 return false;
+        if (i->roothash_job && !PULL_JOB_IS_COMPLETE(i->roothash_job))
+                return false;
         if (i->settings_job && !PULL_JOB_IS_COMPLETE(i->settings_job))
                 return false;
         if (i->checksum_job && !PULL_JOB_IS_COMPLETE(i->checksum_job))
@@ -380,6 +431,39 @@ static bool raw_pull_is_done(RawPull *i) {
         return true;
 }
 
+static int raw_pull_rename_auxiliary_file(
+                RawPull *i,
+                const char *suffix,
+                char **temp_path,
+                char **path) {
+
+        int r;
+
+        assert(i);
+        assert(temp_path);
+        assert(suffix);
+        assert(path);
+
+        /* Regenerate final name for this auxiliary file, we might know the etag of the raw file now, and we shoud
+         * incorporate it in the file name if we can */
+        *path = mfree(*path);
+        r = raw_pull_determine_path(i, suffix, path);
+        if (r < 0)
+                return r;
+
+        r = import_make_read_only(*temp_path);
+        if (r < 0)
+                return r;
+
+        r = rename_noreplace(AT_FDCWD, *temp_path, AT_FDCWD, *path);
+        if (r < 0)
+                return log_error_errno(r, "Failed to rename file %s to %s: %m", *temp_path, *path);
+
+        *temp_path = mfree(*temp_path);
+
+        return 1;
+}
+
 static void raw_pull_job_on_finished(PullJob *j) {
         RawPull *i;
         int r;
@@ -388,7 +472,10 @@ static void raw_pull_job_on_finished(PullJob *j) {
         assert(j->userdata);
 
         i = j->userdata;
-        if (j == i->settings_job) {
+        if (j == i->roothash_job) {
+                if (j->error != 0)
+                        log_info_errno(j->error, "Root hash file could not be retrieved, proceeding without.");
+        } else if (j == i->settings_job) {
                 if (j->error != 0)
                         log_info_errno(j->error, "Settings file could not be retrieved, proceeding without.");
         } else if (j->error != 0) {
@@ -413,16 +500,22 @@ static void raw_pull_job_on_finished(PullJob *j) {
         if (!raw_pull_is_done(i))
                 return;
 
+        if (i->roothash_job)
+                i->roothash_job->disk_fd = safe_close(i->roothash_job->disk_fd);
         if (i->settings_job)
                 i->settings_job->disk_fd = safe_close(i->settings_job->disk_fd);
 
+        r = raw_pull_determine_path(i, ".raw", &i->final_path);
+        if (r < 0)
+                goto finish;
+
         if (!i->raw_job->etag_exists) {
                 /* This is a new download, verify it, and move it into place */
                 assert(i->raw_job->disk_fd >= 0);
 
                 raw_pull_report_progress(i, RAW_VERIFYING);
 
-                r = pull_verify(i->raw_job, i->settings_job, i->checksum_job, i->signature_job);
+                r = pull_verify(i->raw_job, i->roothash_job, i->settings_job, i->checksum_job, i->signature_job);
                 if (r < 0)
                         goto finish;
 
@@ -446,24 +539,18 @@ static void raw_pull_job_on_finished(PullJob *j) {
 
                 i->temp_path = mfree(i->temp_path);
 
-                if (i->settings_job &&
-                    i->settings_job->error == 0 &&
-                    !i->settings_job->etag_exists) {
-
-                        assert(i->settings_temp_path);
-                        assert(i->settings_path);
-
-                        r = import_make_read_only(i->settings_temp_path);
+                if (i->roothash_job &&
+                    i->roothash_job->error == 0) {
+                        r = raw_pull_rename_auxiliary_file(i, ".roothash", &i->roothash_temp_path, &i->roothash_path);
                         if (r < 0)
                                 goto finish;
+                }
 
-                        r = rename_noreplace(AT_FDCWD, i->settings_temp_path, AT_FDCWD, i->settings_path);
-                        if (r < 0) {
-                                log_error_errno(r, "Failed to rename settings file: %m");
+                if (i->settings_job &&
+                    i->settings_job->error == 0) {
+                        r = raw_pull_rename_auxiliary_file(i, ".nspawn", &i->settings_temp_path, &i->settings_path);
+                        if (r < 0)
                                 goto finish;
-                        }
-
-                        i->settings_temp_path = mfree(i->settings_temp_path);
                 }
         }
 
@@ -482,6 +569,35 @@ finish:
                 sd_event_exit(i->event, r);
 }
 
+static int raw_pull_job_on_open_disk_generic(
+                RawPull *i,
+                PullJob *j,
+                const char *extra,
+                char **temp_path) {
+
+        _cleanup_free_ char *p = NULL;
+        int r;
+
+        assert(i);
+        assert(j);
+        assert(extra);
+        assert(temp_path);
+
+        if (!*temp_path) {
+                r = tempfn_random_child(i->image_root, extra, temp_path);
+                if (r < 0)
+                        return log_oom();
+        }
+
+        (void) mkdir_parents_label(*temp_path, 0700);
+
+        j->disk_fd = open(*temp_path, O_RDWR|O_CREAT|O_EXCL|O_NOCTTY|O_CLOEXEC, 0664);
+        if (j->disk_fd < 0)
+                return log_error_errno(errno, "Failed to create %s: %m", *temp_path);
+
+        return 0;
+}
+
 static int raw_pull_job_on_open_disk_raw(PullJob *j) {
         RawPull *i;
         int r;
@@ -491,57 +607,40 @@ static int raw_pull_job_on_open_disk_raw(PullJob *j) {
 
         i = j->userdata;
         assert(i->raw_job == j);
-        assert(!i->final_path);
-        assert(!i->temp_path);
 
-        r = pull_make_path(j->url, j->etag, i->image_root, ".raw-", ".raw", &i->final_path);
+        r = raw_pull_job_on_open_disk_generic(i, j, "raw", &i->temp_path);
         if (r < 0)
-                return log_oom();
-
-        r = tempfn_random(i->final_path, NULL, &i->temp_path);
-        if (r < 0)
-                return log_oom();
-
-        (void) mkdir_parents_label(i->temp_path, 0700);
-
-        j->disk_fd = open(i->temp_path, O_RDWR|O_CREAT|O_EXCL|O_NOCTTY|O_CLOEXEC, 0664);
-        if (j->disk_fd < 0)
-                return log_error_errno(errno, "Failed to create %s: %m", i->temp_path);
+                return r;
 
         r = chattr_fd(j->disk_fd, FS_NOCOW_FL, FS_NOCOW_FL);
         if (r < 0)
-                log_warning_errno(r, "Failed to set file attributes on %s: %m", i->temp_path);
+                log_warning_errno(r, "Failed to set file attributes on %s, ignoring: %m", i->temp_path);
 
         return 0;
 }
 
+static int raw_pull_job_on_open_disk_roothash(PullJob *j) {
+        RawPull *i;
+
+        assert(j);
+        assert(j->userdata);
+
+        i = j->userdata;
+        assert(i->roothash_job == j);
+
+        return raw_pull_job_on_open_disk_generic(i, j, "roothash", &i->roothash_temp_path);
+}
+
 static int raw_pull_job_on_open_disk_settings(PullJob *j) {
         RawPull *i;
-        int r;
 
         assert(j);
         assert(j->userdata);
 
         i = j->userdata;
         assert(i->settings_job == j);
-        assert(!i->settings_path);
-        assert(!i->settings_temp_path);
 
-        r = pull_make_path(j->url, j->etag, i->image_root, ".settings-", NULL, &i->settings_path);
-        if (r < 0)
-                return log_oom();
-
-        r = tempfn_random(i->settings_path, NULL, &i->settings_temp_path);
-        if (r < 0)
-                return log_oom();
-
-        mkdir_parents_label(i->settings_temp_path, 0700);
-
-        j->disk_fd = open(i->settings_temp_path, O_RDWR|O_CREAT|O_EXCL|O_NOCTTY|O_CLOEXEC, 0664);
-        if (j->disk_fd < 0)
-                return log_error_errno(errno, "Failed to create %s: %m", i->settings_temp_path);
-
-        return 0;
+        return raw_pull_job_on_open_disk_generic(i, j, "settings", &i->settings_temp_path);
 }
 
 static void raw_pull_job_on_progress(PullJob *j) {
@@ -561,7 +660,8 @@ int raw_pull_start(
                 const char *local,
                 bool force_local,
                 ImportVerify verify,
-                bool settings) {
+                bool settings,
+                bool roothash) {
 
         int r;
 
@@ -585,6 +685,7 @@ int raw_pull_start(
         i->force_local = force_local;
         i->verify = verify;
         i->settings = settings;
+        i->roothash = roothash;
 
         /* Queue job for the image itself */
         r = pull_job_new(&i->raw_job, url, i->glue, i);
@@ -601,18 +702,24 @@ int raw_pull_start(
         if (r < 0)
                 return r;
 
+        if (roothash) {
+                r = pull_make_auxiliary_job(&i->roothash_job, url, raw_strip_suffixes, ".roothash", i->glue, raw_pull_job_on_finished, i);
+                if (r < 0)
+                        return r;
+
+                i->roothash_job->on_open_disk = raw_pull_job_on_open_disk_roothash;
+                i->roothash_job->on_progress = raw_pull_job_on_progress;
+                i->roothash_job->calc_checksum = verify != IMPORT_VERIFY_NO;
+        }
+
         if (settings) {
-                r = pull_make_settings_job(&i->settings_job, url, i->glue, raw_pull_job_on_finished, i);
+                r = pull_make_auxiliary_job(&i->settings_job, url, raw_strip_suffixes, ".nspawn", i->glue, raw_pull_job_on_finished, i);
                 if (r < 0)
                         return r;
 
                 i->settings_job->on_open_disk = raw_pull_job_on_open_disk_settings;
                 i->settings_job->on_progress = raw_pull_job_on_progress;
                 i->settings_job->calc_checksum = verify != IMPORT_VERIFY_NO;
-
-                r = pull_find_old_etags(i->settings_job->url, i->image_root, DT_REG, ".settings-", NULL, &i->settings_job->old_etags);
-                if (r < 0)
-                        return r;
         }
 
         r = pull_make_verification_jobs(&i->checksum_job, &i->signature_job, verify, url, i->glue, raw_pull_job_on_finished, i);
@@ -623,6 +730,12 @@ int raw_pull_start(
         if (r < 0)
                 return r;
 
+        if (i->roothash_job) {
+                r = pull_job_begin(i->roothash_job);
+                if (r < 0)
+                        return r;
+        }
+
         if (i->settings_job) {
                 r = pull_job_begin(i->settings_job);
                 if (r < 0)

src/import/pull-raw.h

@@ -33,4 +33,4 @@ RawPull* raw_pull_unref(RawPull *pull);
 
 DEFINE_TRIVIAL_CLEANUP_FUNC(RawPull*, raw_pull_unref);
 
-int raw_pull_start(RawPull *pull, const char *url, const char *local, bool force_local, ImportVerify verify, bool settings);
+int raw_pull_start(RawPull *pull, const char *url, const char *local, bool force_local, ImportVerify verify, bool settings, bool roothash);

src/import/pull-tar.c

@@ -215,6 +215,24 @@ static void tar_pull_report_progress(TarPull *i, TarProgress p) {
         log_debug("Combined progress %u%%", percent);
 }
 
+static int tar_pull_determine_path(TarPull *i, const char *suffix, char **field) {
+        int r;
+
+        assert(i);
+        assert(field);
+
+        if (*field)
+                return 0;
+
+        assert(i->tar_job);
+
+        r = pull_make_path(i->tar_job->url, i->tar_job->etag, i->image_root, ".tar-", suffix, field);
+        if (r < 0)
+                return log_oom();
+
+        return 1;
+}
+
 static int tar_pull_make_local_copy(TarPull *i) {
         int r;
 
@@ -224,12 +242,6 @@ static int tar_pull_make_local_copy(TarPull *i) {
         if (!i->local)
                 return 0;
 
-        if (!i->final_path) {
-                r = pull_make_path(i->tar_job->url, i->tar_job->etag, i->image_root, ".tar-", NULL, &i->final_path);
-                if (r < 0)
-                        return log_oom();
-        }
-
         r = pull_make_local_copy(i->final_path, i->image_root, i->local, i->force_local);
         if (r < 0)
                 return r;
@@ -238,11 +250,9 @@ static int tar_pull_make_local_copy(TarPull *i) {
                 const char *local_settings;
                 assert(i->settings_job);
 
-                if (!i->settings_path) {
-                        r = pull_make_path(i->settings_job->url, i->settings_job->etag, i->image_root, ".settings-", NULL, &i->settings_path);
-                        if (r < 0)
-                                return log_oom();
-                }
+                r = tar_pull_determine_path(i, ".nspawn", &i->settings_path);
+                if (r < 0)
+                        return r;
 
                 local_settings = strjoina(i->image_root, "/", i->local, ".nspawn");
 
@@ -311,6 +321,10 @@ static void tar_pull_job_on_finished(PullJob *j) {
         if (i->settings_job)
                 i->settings_job->disk_fd = safe_close(i->settings_job->disk_fd);
 
+        r = tar_pull_determine_path(i, NULL, &i->final_path);
+        if (r < 0)
+                goto finish;
+
         if (i->tar_pid > 0) {
                 r = wait_for_terminate_and_warn("tar", i->tar_pid, true);
                 i->tar_pid = 0;
@@ -327,7 +341,7 @@ static void tar_pull_job_on_finished(PullJob *j) {
 
                 tar_pull_report_progress(i, TAR_VERIFYING);
 
-                r = pull_verify(i->tar_job, i->settings_job, i->checksum_job, i->signature_job);
+                r = pull_verify(i->tar_job, NULL, i->settings_job, i->checksum_job, i->signature_job);
                 if (r < 0)
                         goto finish;
 
@@ -346,16 +360,18 @@ static void tar_pull_job_on_finished(PullJob *j) {
                 i->temp_path = mfree(i->temp_path);
 
                 if (i->settings_job &&
-                    i->settings_job->error == 0 &&
-                    !i->settings_job->etag_exists) {
+                    i->settings_job->error == 0) {
 
                         assert(i->settings_temp_path);
                         assert(i->settings_path);
 
-                        /* Also move the settings file into place, if
-                         * it exist. Note that we do so only if we
-                         * also moved the tar file in place, to keep
-                         * things strictly in sync. */
+                        /* Also move the settings file into place, if it exist. Note that we do so only if we also
+                         * moved the tar file in place, to keep things strictly in sync. */
+
+                        i->settings_path = mfree(i->settings_path);
+                        r = tar_pull_determine_path(i, ".nspawn", &i->settings_path);
+                        if (r < 0)
+                                goto finish;
 
                         r = import_make_read_only(i->settings_temp_path);
                         if (r < 0)
@@ -395,17 +411,13 @@ static int tar_pull_job_on_open_disk_tar(PullJob *j) {
 
         i = j->userdata;
         assert(i->tar_job == j);
-        assert(!i->final_path);
-        assert(!i->temp_path);
         assert(i->tar_pid <= 0);
 
-        r = pull_make_path(j->url, j->etag, i->image_root, ".tar-", NULL, &i->final_path);
-        if (r < 0)
-                return log_oom();
-
-        r = tempfn_random(i->final_path, NULL, &i->temp_path);
-        if (r < 0)
-                return log_oom();
+        if (!i->temp_path) {
+                r = tempfn_random_child(i->image_root, "tar", &i->temp_path);
+                if (r < 0)
+                        return log_oom();
+        }
 
         mkdir_parents_label(i->temp_path, 0700);
 
@@ -434,16 +446,12 @@ static int tar_pull_job_on_open_disk_settings(PullJob *j) {
 
         i = j->userdata;
         assert(i->settings_job == j);
-        assert(!i->settings_path);
-        assert(!i->settings_temp_path);
 
-        r = pull_make_path(j->url, j->etag, i->image_root, ".settings-", NULL, &i->settings_path);
-        if (r < 0)
-                return log_oom();
-
-        r = tempfn_random(i->settings_path, NULL, &i->settings_temp_path);
-        if (r < 0)
-                return log_oom();
+        if (!i->settings_temp_path) {
+                r = tempfn_random_child(i->image_root, "settings", &i->settings_temp_path);
+                if (r < 0)
+                        return log_oom();
+        }
 
         mkdir_parents_label(i->settings_temp_path, 0700);
 
@@ -513,17 +521,13 @@ int tar_pull_start(
 
         /* Set up download job for the settings file (.nspawn) */
         if (settings) {
-                r = pull_make_settings_job(&i->settings_job, url, i->glue, tar_pull_job_on_finished, i);
+                r = pull_make_auxiliary_job(&i->settings_job, url, tar_strip_suffixes, ".nspawn", i->glue, tar_pull_job_on_finished, i);
                 if (r < 0)
                         return r;
 
                 i->settings_job->on_open_disk = tar_pull_job_on_open_disk_settings;
                 i->settings_job->on_progress = tar_pull_job_on_progress;
                 i->settings_job->calc_checksum = verify != IMPORT_VERIFY_NO;
-
-                r = pull_find_old_etags(i->settings_job->url, i->image_root, DT_REG, ".settings-", NULL, &i->settings_job->old_etags);
-                if (r < 0)
-                        return r;
         }
 
         /* Set up download of checksum/signature files */

src/import/pull.c

@@ -37,6 +37,7 @@ static bool arg_force = false;
 static const char *arg_image_root = "/var/lib/machines";
 static ImportVerify arg_verify = IMPORT_VERIFY_SIGNATURE;
 static bool arg_settings = true;
+static bool arg_roothash = true;
 
 static int interrupt_signal_handler(sd_event_source *s, const struct signalfd_siginfo *si, void *userdata) {
         log_notice("Transfer aborted.");
@@ -204,7 +205,7 @@ static int pull_raw(int argc, char *argv[], void *userdata) {
         if (r < 0)
                 return log_error_errno(r, "Failed to allocate puller: %m");
 
-        r = raw_pull_start(pull, url, local, arg_force, arg_verify, arg_settings);
+        r = raw_pull_start(pull, url, local, arg_force, arg_verify, arg_settings, arg_roothash);
         if (r < 0)
                 return log_error_errno(r, "Failed to pull image: %m");
 
@@ -226,6 +227,7 @@ static int help(int argc, char *argv[], void *userdata) {
                "     --verify=MODE            Verify downloaded image, one of: 'no',\n"
                "                              'checksum', 'signature'\n"
                "     --settings=BOOL          Download settings file with image\n"
+               "     --roothash=BOOL          Download root hash file with image\n"
                "     --image-root=PATH        Image root directory\n\n"
                "Commands:\n"
                "  tar URL [NAME]              Download a TAR image\n"
@@ -243,6 +245,7 @@ static int parse_argv(int argc, char *argv[]) {
                 ARG_IMAGE_ROOT,
                 ARG_VERIFY,
                 ARG_SETTINGS,
+                ARG_ROOTHASH,
         };
 
         static const struct option options[] = {
@@ -252,6 +255,7 @@ static int parse_argv(int argc, char *argv[]) {
                 { "image-root",      required_argument, NULL, ARG_IMAGE_ROOT      },
                 { "verify",          required_argument, NULL, ARG_VERIFY          },
                 { "settings",        required_argument, NULL, ARG_SETTINGS        },
+                { "roothash",        required_argument, NULL, ARG_ROOTHASH        },
                 {}
         };
 
@@ -295,6 +299,14 @@ static int parse_argv(int argc, char *argv[]) {
                         arg_settings = r;
                         break;
 
+                case ARG_ROOTHASH:
+                        r = parse_boolean(optarg);
+                        if (r < 0)
+                                return log_error_errno(r, "Failed to parse --roothash= parameter '%s'", optarg);
+
+                        arg_roothash = r;
+                        break;
+
                 case '?':
                         return -EINVAL;
 

src/journal/journald-server.c

@@ -283,17 +283,16 @@ static int open_journal(
 }
 
 static bool flushed_flag_is_set(void) {
-        return (access("/run/systemd/journal/flushed", F_OK) >= 0);
+        return access("/run/systemd/journal/flushed", F_OK) >= 0;
 }
 
 static int system_journal_open(Server *s, bool flush_requested) {
-        bool flushed = false;
         const char *fn;
         int r = 0;
 
         if (!s->system_journal &&
-            (s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) &&
-            (flush_requested || (flushed = flushed_flag_is_set()))) {
+            IN_SET(s->storage, STORAGE_PERSISTENT, STORAGE_AUTO) &&
+            (flush_requested || flushed_flag_is_set())) {
 
                 /* If in auto mode: first try to create the machine
                  * path, but not the prefix.
@@ -326,8 +325,8 @@ static int system_journal_open(Server *s, bool flush_requested) {
                  * Perform an implicit flush to var, leaving the runtime
                  * journal closed, now that the system journal is back.
                  */
-                if (s->runtime_journal && flushed)
-                        (void) server_flush_to_var(s);
+                if (!flush_requested)
+                        (void) server_flush_to_var(s, true);
         }
 
         if (!s->runtime_journal &&
@@ -1183,7 +1182,7 @@ finish:
         dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id, priority, object_pid);
 }
 
-int server_flush_to_var(Server *s) {
+int server_flush_to_var(Server *s, bool require_flag_file) {
         sd_id128_t machine;
         sd_journal *j = NULL;
         char ts[FORMAT_TIMESPAN_MAX];
@@ -1193,13 +1192,15 @@ int server_flush_to_var(Server *s) {
 
         assert(s);
 
-        if (s->storage != STORAGE_AUTO &&
-            s->storage != STORAGE_PERSISTENT)
+        if (!IN_SET(s->storage, STORAGE_AUTO, STORAGE_PERSISTENT))
                 return 0;
 
         if (!s->runtime_journal)
                 return 0;
 
+        if (require_flag_file && !flushed_flag_is_set())
+                return 0;
+
         (void) system_journal_open(s, true);
 
         if (!s->system_journal)
@@ -1411,7 +1412,7 @@ static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *
 
         log_info("Received request to flush runtime journal from PID " PID_FMT, si->ssi_pid);
 
-        (void) server_flush_to_var(s);
+        (void) server_flush_to_var(s, false);
         server_sync(s);
         server_vacuum(s, false);
 
@@ -1532,60 +1533,93 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
 
         assert(s);
 
-        if (streq(key, "systemd.journald.forward_to_syslog")) {
+        if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_syslog")) {
+
                 r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse forward to syslog switch \"%s\". Ignoring.", value);
                 else
                         s->forward_to_syslog = r;
-        } else if (streq(key, "systemd.journald.forward_to_kmsg")) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_kmsg")) {
+
                 r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse forward to kmsg switch \"%s\". Ignoring.", value);
                 else
                         s->forward_to_kmsg = r;
-        } else if (streq(key, "systemd.journald.forward_to_console")) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_console")) {
+
                 r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse forward to console switch \"%s\". Ignoring.", value);
                 else
                         s->forward_to_console = r;
-        } else if (streq(key, "systemd.journald.forward_to_wall")) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.forward_to_wall")) {
+
                 r = value ? parse_boolean(value) : true;
                 if (r < 0)
                         log_warning("Failed to parse forward to wall switch \"%s\". Ignoring.", value);
                 else
                         s->forward_to_wall = r;
-        } else if (streq(key, "systemd.journald.max_level_console") && value) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_console")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = log_level_from_string(value);
                 if (r < 0)
                         log_warning("Failed to parse max level console value \"%s\". Ignoring.", value);
                 else
                         s->max_level_console = r;
-        } else if (streq(key, "systemd.journald.max_level_store") && value) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_store")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = log_level_from_string(value);
                 if (r < 0)
                         log_warning("Failed to parse max level store value \"%s\". Ignoring.", value);
                 else
                         s->max_level_store = r;
-        } else if (streq(key, "systemd.journald.max_level_syslog") && value) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_syslog")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = log_level_from_string(value);
                 if (r < 0)
                         log_warning("Failed to parse max level syslog value \"%s\". Ignoring.", value);
                 else
                         s->max_level_syslog = r;
-        } else if (streq(key, "systemd.journald.max_level_kmsg") && value) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_kmsg")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = log_level_from_string(value);
                 if (r < 0)
                         log_warning("Failed to parse max level kmsg value \"%s\". Ignoring.", value);
                 else
                         s->max_level_kmsg = r;
-        } else if (streq(key, "systemd.journald.max_level_wall") && value) {
+
+        } else if (proc_cmdline_key_streq(key, "systemd.journald.max_level_wall")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = log_level_from_string(value);
                 if (r < 0)
                         log_warning("Failed to parse max level wall value \"%s\". Ignoring.", value);
                 else
                         s->max_level_wall = r;
+
         } else if (startswith(key, "systemd.journald"))
                 log_warning("Unknown journald kernel command line option \"%s\". Ignoring.", key);
 
@@ -1898,7 +1932,10 @@ int server_init(Server *s) {
         journal_reset_metrics(&s->runtime_storage.metrics);
 
         server_parse_config_file(s);
-        parse_proc_cmdline(parse_proc_cmdline_item, s, true);
+
+        r = proc_cmdline_parse(parse_proc_cmdline_item, s, PROC_CMDLINE_STRIP_RD_PREFIX);
+        if (r < 0)
+                log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 
         if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) {
                 log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0",

src/journal/journald-server.h

@@ -197,7 +197,7 @@ void server_sync(Server *s);
 int server_vacuum(Server *s, bool verbose);
 void server_rotate(Server *s);
 int server_schedule_sync(Server *s, int priority);
-int server_flush_to_var(Server *s);
+int server_flush_to_var(Server *s, bool require_flag_file);
 void server_maybe_append_tags(Server *s);
 int server_process_datagram(sd_event_source *es, int fd, uint32_t revents, void *userdata);
 void server_space_usage_message(Server *s, JournalStorage *storage);

src/journal/journald.c

@@ -52,7 +52,7 @@ int main(int argc, char *argv[]) {
                 goto finish;
 
         server_vacuum(&server, false);
-        server_flush_to_var(&server);
+        server_flush_to_var(&server, true);
         server_flush_dev_kmsg(&server);
 
         log_debug("systemd-journald running as pid "PID_FMT, getpid());

src/machine/image-dbus.c

@@ -335,7 +335,7 @@ static int raw_image_get_os_release(Image *image, char ***ret, sd_bus_error *err
         if (r < 0)
                 return sd_bus_error_set_errnof(error, r, "Failed to set up loop block device for %s: %m", image->path);
 
-        r = dissect_image(d->fd, NULL, 0, &m);
+        r = dissect_image(d->fd, NULL, 0, DISSECT_IMAGE_REQUIRE_ROOT, &m);
         if (r == -ENOPKG)
                 return sd_bus_error_set_errnof(error, r, "Disk image %s not understood: %m", image->path);
         if (r < 0)

src/machine/machinectl.c

@@ -2666,9 +2666,9 @@ static int help(int argc, char *argv[], void *userdata) {
                "  -o --output=STRING          Change journal output mode (short,\n"
                "                              short-monotonic, verbose, export, json,\n"
                "                              json-pretty, json-sse, cat)\n"
-               "      --verify=MODE           Verification mode for downloaded images (no,\n"
+               "     --verify=MODE            Verification mode for downloaded images (no,\n"
                "                              checksum, signature)\n"
-               "      --force                 Download image even if already exists\n\n"
+               "     --force                  Download image even if already exists\n\n"
                "Machine Commands:\n"
                "  list                        List running VMs and containers\n"
                "  status NAME...              Show VM/container details\n"

src/modules-load/modules-load.c

@@ -62,7 +62,11 @@ static int add_modules(const char *p) {
 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
         int r;
 
-        if (streq(key, "modules-load") && value) {
+        if (proc_cmdline_key_streq(key, "modules_load")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = add_modules(value);
                 if (r < 0)
                         return r;
@@ -226,7 +230,7 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, true);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, PROC_CMDLINE_STRIP_RD_PREFIX);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 

src/nspawn/nspawn-mount.c

@@ -1349,21 +1349,3 @@ fail:
         (void) rmdir(template);
         return r;
 }
-
-VolatileMode volatile_mode_from_string(const char *s) {
-        int b;
-
-        if (isempty(s))
-                return _VOLATILE_MODE_INVALID;
-
-        b = parse_boolean(s);
-        if (b > 0)
-                return VOLATILE_YES;
-        if (b == 0)
-                return VOLATILE_NO;
-
-        if (streq(s, "state"))
-                return VOLATILE_STATE;
-
-        return _VOLATILE_MODE_INVALID;
-}

src/nspawn/nspawn-mount.h

@@ -22,6 +22,7 @@
 #include <stdbool.h>
 
 #include "cgroup-util.h"
+#include "volatile-util.h"
 
 typedef enum MountSettingsMask {
         MOUNT_FATAL              = 1 << 0, /* if set, a mount error is considered fatal */
@@ -32,14 +33,6 @@ typedef enum MountSettingsMask {
                                               Works only if MOUNT_APPLY_APIVFS_RO is also set. */
 } MountSettingsMask;
 
-typedef enum VolatileMode {
-        VOLATILE_NO,
-        VOLATILE_YES,
-        VOLATILE_STATE,
-        _VOLATILE_MODE_MAX,
-        _VOLATILE_MODE_INVALID = -1
-} VolatileMode;
-
 typedef enum CustomMountType {
         CUSTOM_MOUNT_BIND,
         CUSTOM_MOUNT_TMPFS,
@@ -77,5 +70,3 @@ int mount_custom(const char *dest, CustomMount *mounts, unsigned n, bool userns,
 
 int setup_volatile(const char *directory, VolatileMode mode, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context);
 int setup_volatile_state(const char *directory, VolatileMode mode, bool userns, uid_t uid_shift, uid_t uid_range, const char *selinux_apifs_context);
-
-VolatileMode volatile_mode_from_string(const char *s);

src/nspawn/nspawn.c

@@ -1288,15 +1288,18 @@ static int setup_timezone(const char *dest) {
                 return 0;
         }
 
-        r = unlink(where);
-        if (r < 0 && errno != ENOENT) {
-                log_error_errno(errno, "Failed to remove existing timezone info %s in container: %m", where);
+        if (unlink(where) < 0 && errno != ENOENT) {
+                log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING, /* Don't complain on read-only images */
+                               errno,
+                               "Failed to remove existing timezone info %s in container, ignoring: %m", where);
                 return 0;
         }
 
         what = strjoina("../usr/share/zoneinfo/", z);
         if (symlink(what, where) < 0) {
-                log_error_errno(errno, "Failed to correct timezone of container: %m");
+                log_full_errno(IN_SET(errno, EROFS, EACCES, EPERM) ? LOG_DEBUG : LOG_WARNING,
+                               errno,
+                               "Failed to correct timezone of container, ignoring: %m");
                 return 0;
         }
 
@@ -1308,31 +1311,43 @@ static int setup_timezone(const char *dest) {
 }
 
 static int setup_resolv_conf(const char *dest) {
-        const char *where = NULL;
-        int r;
+        _cleanup_free_ char *resolved = NULL, *etc = NULL;
+        const char *where;
+        int r, found;
 
         assert(dest);
 
         if (arg_private_network)
                 return 0;
 
-        /* Fix resolv.conf, if possible */
-        where = prefix_roota(dest, "/etc/resolv.conf");
+        r = chase_symlinks("/etc", dest, CHASE_PREFIX_ROOT, &etc);
+        if (r < 0) {
+                log_warning_errno(r, "Failed to resolve /etc path in container, ignoring: %m");
+                return 0;
+        }
+
+        where = strjoina(etc, "/resolv.conf");
+        found = chase_symlinks(where, dest, CHASE_NONEXISTENT, &resolved);
+        if (found < 0) {
+                log_warning_errno(found, "Failed to resolve /etc/resolv.conf path in container, ignoring: %m");
+                return 0;
+        }
 
         if (access("/run/systemd/resolve/resolv.conf", F_OK) >= 0 &&
-                        access("/usr/lib/systemd/resolv.conf", F_OK) >= 0) {
+            access("/usr/lib/systemd/resolv.conf", F_OK) >= 0) {
+
                 /* resolved is enabled on the host. In this, case bind mount its static resolv.conf file into the
                  * container, so that the container can use the host's resolver. Given that network namespacing is
                  * disabled it's only natural of the container also uses the host's resolver. It also has the big
                  * advantage that the container will be able to follow the host's DNS server configuration changes
                  * transparently. */
 
-                (void) touch(where);
+                if (found == 0) /* missing? */
+                        (void) touch(resolved);
 
-                r = mount_verbose(LOG_WARNING, "/usr/lib/systemd/resolv.conf", where, NULL, MS_BIND, NULL);
+                r = mount_verbose(LOG_DEBUG, "/usr/lib/systemd/resolv.conf", resolved, NULL, MS_BIND, NULL);
                 if (r >= 0)
-                        return mount_verbose(LOG_ERR, NULL, where, NULL,
-                                             MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
+                        return mount_verbose(LOG_ERR, NULL, resolved, NULL, MS_BIND|MS_REMOUNT|MS_RDONLY|MS_NOSUID|MS_NODEV, NULL);
         }
 
         /* If that didn't work, let's copy the file */
@@ -1343,7 +1358,7 @@ static int setup_resolv_conf(const char *dest) {
                  *
                  * If the disk image is read-only, there's also no point in complaining.
                  */
-                log_full_errno(IN_SET(r, -ELOOP, -EROFS) ? LOG_DEBUG : LOG_WARNING, r,
+                log_full_errno(IN_SET(r, -ELOOP, -EROFS, -EACCES, -EPERM) ? LOG_DEBUG : LOG_WARNING, r,
                                "Failed to copy /etc/resolv.conf to %s, ignoring: %m", where);
                 return 0;
         }
@@ -2467,20 +2482,6 @@ static int outer_child(
         if (r < 0)
                 return r;
 
-        /* Mark everything as shared so our mounts get propagated down. This is
-         * required to make new bind mounts available in systemd services
-         * inside the containter that create a new mount namespace.
-         * See https://github.com/systemd/systemd/issues/3860
-         * Further submounts (such as /dev) done after this will inherit the
-         * shared propagation mode.*/
-        r = mount_verbose(LOG_ERR, NULL, directory, NULL, MS_SHARED|MS_REC, NULL);
-        if (r < 0)
-                return r;
-
-        r = recursive_chown(directory, arg_uid_shift, arg_uid_range);
-        if (r < 0)
-                return r;
-
         r = setup_volatile(
                         directory,
                         arg_volatile_mode,
@@ -2501,6 +2502,20 @@ static int outer_child(
         if (r < 0)
                 return r;
 
+        /* Mark everything as shared so our mounts get propagated down. This is
+         * required to make new bind mounts available in systemd services
+         * inside the containter that create a new mount namespace.
+         * See https://github.com/systemd/systemd/issues/3860
+         * Further submounts (such as /dev) done after this will inherit the
+         * shared propagation mode.*/
+        r = mount_verbose(LOG_ERR, NULL, directory, NULL, MS_SHARED|MS_REC, NULL);
+        if (r < 0)
+                return r;
+
+        r = recursive_chown(directory, arg_uid_shift, arg_uid_range);
+        if (r < 0)
+                return r;
+
         r = base_filesystem_create(directory, arg_uid_shift, (gid_t) arg_uid_shift);
         if (r < 0)
                 return r;
@@ -3740,7 +3755,11 @@ int main(int argc, char *argv[]) {
                         goto finish;
                 }
 
-                r = dissect_image(loop->fd, arg_root_hash, arg_root_hash_size, &dissected_image);
+                r = dissect_image(
+                                loop->fd,
+                                arg_root_hash, arg_root_hash_size,
+                                DISSECT_IMAGE_REQUIRE_ROOT,
+                                &dissected_image);
                 if (r == -ENOPKG) {
                         log_error_errno(r, "Could not find a suitable file system or partition table in image: %s", arg_image);
 

src/quotacheck/quotacheck.c

@@ -34,7 +34,10 @@ static bool arg_force = false;
 
 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
 
-        if (streq(key, "quotacheck.mode") && value) {
+        if (streq(key, "quotacheck.mode")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
 
                 if (streq(value, "auto"))
                         arg_force = arg_skip = false;
@@ -88,7 +91,7 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, false);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, 0);
         if (r < 0)
                 log_warning_errno(r, "Failed to parse kernel command line, ignoring: %m");
 
@@ -104,9 +107,10 @@ int main(int argc, char *argv[]) {
 
         pid = fork();
         if (pid < 0) {
-                log_error_errno(errno, "fork(): %m");
-                return EXIT_FAILURE;
-        } else if (pid == 0) {
+                r = log_error_errno(errno, "fork(): %m");
+                goto finish;
+        }
+        if (pid == 0) {
 
                 /* Child */
 
@@ -120,5 +124,6 @@ int main(int argc, char *argv[]) {
 
         r = wait_for_terminate_and_warn("quotacheck", pid, true);
 
+finish:
         return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
 }

src/remount-fs/remount-fs.c

@@ -56,7 +56,7 @@ int main(int argc, char *argv[]) {
 
         umask(0022);
 
-        f = setmntent("/etc/fstab", "r");
+        f = setmntent("/etc/fstab", "re");
         if (!f) {
                 if (errno == ENOENT) {
                         r = 0;

src/shared/base-filesystem.c

@@ -101,7 +101,7 @@ int base_filesystem_create(const char *root, uid_t uid, gid_t gid) {
                         if (r < 0 && errno != EEXIST)
                                 return log_error_errno(errno, "Failed to create symlink at %s/%s: %m", root, table[i].dir);
 
-                        if (uid != UID_INVALID || gid != UID_INVALID) {
+                        if (uid_is_valid(uid) || gid_is_valid(gid)) {
                                 if (fchownat(fd, table[i].dir, uid, gid, AT_SYMLINK_NOFOLLOW) < 0)
                                         return log_error_errno(errno, "Failed to chown symlink at %s/%s: %m", root, table[i].dir);
                         }

src/shared/dissect-image.c

@@ -84,7 +84,7 @@ not_found:
 #endif
 }
 
-int dissect_image(int fd, const void *root_hash, size_t root_hash_size, DissectedImage **ret) {
+int dissect_image(int fd, const void *root_hash, size_t root_hash_size, DissectImageFlags flags, DissectedImage **ret) {
 
 #ifdef HAVE_BLKID
         sd_id128_t root_uuid = SD_ID128_NULL, verity_uuid = SD_ID128_NULL;
@@ -95,7 +95,8 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
         _cleanup_blkid_free_probe_ blkid_probe b = NULL;
         _cleanup_udev_unref_ struct udev *udev = NULL;
         _cleanup_free_ char *generic_node = NULL;
-        const char *pttype = NULL, *usage = NULL;
+        sd_id128_t generic_uuid = SD_ID128_NULL;
+        const char *pttype = NULL;
         struct udev_list_entry *first, *item;
         blkid_partlist pl;
         int r, generic_nr;
@@ -147,8 +148,12 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                 return -errno;
         }
 
-        blkid_probe_enable_superblocks(b, 1);
-        blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE|BLKID_SUBLKS_USAGE);
+        if ((flags & DISSECT_IMAGE_GPT_ONLY) == 0) {
+                /* Look for file system superblocks, unless we only shall look for GPT partition tables */
+                blkid_probe_enable_superblocks(b, 1);
+                blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE|BLKID_SUBLKS_USAGE);
+        }
+
         blkid_probe_enable_partitions(b, 1);
         blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS);
 
@@ -169,40 +174,45 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
         if (!m)
                 return -ENOMEM;
 
-        (void) blkid_probe_lookup_value(b, "USAGE", &usage, NULL);
-        if (STRPTR_IN_SET(usage, "filesystem", "crypto")) {
-                _cleanup_free_ char *t = NULL, *n = NULL;
-                const char *fstype = NULL;
+        if (!(flags & DISSECT_IMAGE_GPT_ONLY) &&
+            (flags & DISSECT_IMAGE_REQUIRE_ROOT)) {
+                const char *usage = NULL;
 
-                /* OK, we have found a file system, that's our root partition then. */
-                (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL);
+                (void) blkid_probe_lookup_value(b, "USAGE", &usage, NULL);
+                if (STRPTR_IN_SET(usage, "filesystem", "crypto")) {
+                        _cleanup_free_ char *t = NULL, *n = NULL;
+                        const char *fstype = NULL;
 
-                if (fstype) {
-                        t = strdup(fstype);
-                        if (!t)
+                        /* OK, we have found a file system, that's our root partition then. */
+                        (void) blkid_probe_lookup_value(b, "TYPE", &fstype, NULL);
+
+                        if (fstype) {
+                                t = strdup(fstype);
+                                if (!t)
+                                        return -ENOMEM;
+                        }
+
+                        if (asprintf(&n, "/dev/block/%u:%u", major(st.st_rdev), minor(st.st_rdev)) < 0)
                                 return -ENOMEM;
+
+                        m->partitions[PARTITION_ROOT] = (DissectedPartition) {
+                                .found = true,
+                                .rw = true,
+                                .partno = -1,
+                                .architecture = _ARCHITECTURE_INVALID,
+                                .fstype = t,
+                                .node = n,
+                        };
+
+                        t = n = NULL;
+
+                        m->encrypted = streq(fstype, "crypto_LUKS");
+
+                        *ret = m;
+                        m = NULL;
+
+                        return 0;
                 }
-
-                if (asprintf(&n, "/dev/block/%u:%u", major(st.st_rdev), minor(st.st_rdev)) < 0)
-                        return -ENOMEM;
-
-                m->partitions[PARTITION_ROOT] = (DissectedPartition) {
-                        .found = true,
-                        .rw = true,
-                        .partno = -1,
-                        .architecture = _ARCHITECTURE_INVALID,
-                        .fstype = t,
-                        .node = n,
-                };
-
-                t = n = NULL;
-
-                m->encrypted = streq(fstype, "crypto_LUKS");
-
-                *ret = m;
-                m = NULL;
-
-                return 0;
         }
 
         (void) blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL);
@@ -212,7 +222,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
         is_gpt = streq_ptr(pttype, "gpt");
         is_mbr = streq_ptr(pttype, "dos");
 
-        if (!is_gpt && !is_mbr)
+        if (!is_gpt && ((flags & DISSECT_IMAGE_GPT_ONLY) || !is_mbr))
                 return -ENOPKG;
 
         errno = 0;
@@ -300,7 +310,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
         first = udev_enumerate_get_list_entry(e);
         udev_list_entry_foreach(item, first) {
                 _cleanup_udev_device_unref_ struct udev_device *q;
-                unsigned long long flags;
+                unsigned long long pflags;
                 blkid_partition pp;
                 const char *node;
                 dev_t qn;
@@ -325,7 +335,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                 if (!pp)
                         continue;
 
-                flags = blkid_partition_get_flags(pp);
+                pflags = blkid_partition_get_flags(pp);
 
                 nr = blkid_partition_get_partno(pp);
                 if (nr < 0)
@@ -337,7 +347,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                         sd_id128_t type_id, id;
                         bool rw = true;
 
-                        if (flags & GPT_FLAG_NO_AUTO)
+                        if (pflags & GPT_FLAG_NO_AUTO)
                                 continue;
 
                         sid = blkid_partition_get_uuid(pp);
@@ -354,10 +364,10 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
 
                         if (sd_id128_equal(type_id, GPT_HOME)) {
                                 designator = PARTITION_HOME;
-                                rw = !(flags & GPT_FLAG_READ_ONLY);
+                                rw = !(pflags & GPT_FLAG_READ_ONLY);
                         } else if (sd_id128_equal(type_id, GPT_SRV)) {
                                 designator = PARTITION_SRV;
-                                rw = !(flags & GPT_FLAG_READ_ONLY);
+                                rw = !(pflags & GPT_FLAG_READ_ONLY);
                         } else if (sd_id128_equal(type_id, GPT_ESP)) {
                                 designator = PARTITION_ESP;
                                 fstype = "vfat";
@@ -371,7 +381,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
 
                                 designator = PARTITION_ROOT;
                                 architecture = native_architecture();
-                                rw = !(flags & GPT_FLAG_READ_ONLY);
+                                rw = !(pflags & GPT_FLAG_READ_ONLY);
                         } else if (sd_id128_equal(type_id, GPT_ROOT_NATIVE_VERITY)) {
 
                                 m->can_verity = true;
@@ -395,9 +405,8 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
 
                                 designator = PARTITION_ROOT_SECONDARY;
                                 architecture = SECONDARY_ARCHITECTURE;
-                                rw = !(flags & GPT_FLAG_READ_ONLY);
+                                rw = !(pflags & GPT_FLAG_READ_ONLY);
                         } else if (sd_id128_equal(type_id, GPT_ROOT_SECONDARY_VERITY)) {
-
                                 m->can_verity = true;
 
                                 /* Ignore verity unless root has is specified */
@@ -419,7 +428,8 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                                         multiple_generic = true;
                                 else {
                                         generic_nr = nr;
-                                        generic_rw = !(flags & GPT_FLAG_READ_ONLY);
+                                        generic_rw = !(pflags & GPT_FLAG_READ_ONLY);
+                                        generic_uuid = id;
                                         generic_node = strdup(node);
                                         if (!generic_node)
                                                 return -ENOMEM;
@@ -450,6 +460,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                                         .architecture = architecture,
                                         .node = n,
                                         .fstype = t,
+                                        .uuid = id,
                                 };
 
                                 n = t = NULL;
@@ -457,7 +468,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
 
                 } else if (is_mbr) {
 
-                        if (flags != 0x80) /* Bootable flag */
+                        if (pflags != 0x80) /* Bootable flag */
                                 continue;
 
                         if (blkid_partition_get_type(pp) != 0x83) /* Linux partition */
@@ -480,7 +491,7 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                  * either, then check if there's a single generic one, and use that. */
 
                 if (m->partitions[PARTITION_ROOT_VERITY].found)
-                        return -ENXIO;
+                        return -EADDRNOTAVAIL;
 
                 if (m->partitions[PARTITION_ROOT_SECONDARY].found) {
                         m->partitions[PARTITION_ROOT] = m->partitions[PARTITION_ROOT_SECONDARY];
@@ -489,8 +500,19 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                         m->partitions[PARTITION_ROOT_VERITY] = m->partitions[PARTITION_ROOT_SECONDARY_VERITY];
                         zero(m->partitions[PARTITION_ROOT_SECONDARY_VERITY]);
 
-                } else if (generic_node && !root_hash) {
+                } else if (flags & DISSECT_IMAGE_REQUIRE_ROOT) {
 
+                        /* If the root has was set, then we won't fallback to a generic node, because the root hash
+                         * decides */
+                        if (root_hash)
+                                return -EADDRNOTAVAIL;
+
+                        /* If we didn't find a generic node, then we can't fix this up either */
+                        if (!generic_node)
+                                return -ENXIO;
+
+                        /* If we didn't find a properly marked root partition, but we did find a single suitable
+                         * generic Linux partition, then use this as root partition, if the caller asked for it. */
                         if (multiple_generic)
                                 return -ENOTUNIQ;
 
@@ -500,17 +522,15 @@ int dissect_image(int fd, const void *root_hash, size_t root_hash_size, Dissecte
                                 .partno = generic_nr,
                                 .architecture = _ARCHITECTURE_INVALID,
                                 .node = generic_node,
+                                .uuid = generic_uuid,
                         };
 
                         generic_node = NULL;
-                } else
-                        return -ENXIO;
+                }
         }
 
-        assert(m->partitions[PARTITION_ROOT].found);
-
         if (root_hash) {
-                if (!m->partitions[PARTITION_ROOT_VERITY].found)
+                if (!m->partitions[PARTITION_ROOT_VERITY].found || !m->partitions[PARTITION_ROOT].found)
                         return -EADDRNOTAVAIL;
 
                 /* If we found the primary root with the hash, then we definitely want to suppress any secondary root

src/shared/dissect-image.h

@@ -32,6 +32,7 @@ struct DissectedPartition {
         bool rw:1;
         int partno;        /* -1 if there was no partition and the images contains a file system directly */
         int architecture;  /* Intended architecture: either native, secondary or unset (-1). */
+        sd_id128_t uuid;   /* Partition entry UUID as reported by the GPT */
         char *fstype;
         char *node;
         char *decrypted_node;
@@ -67,6 +68,8 @@ typedef enum DissectImageFlags {
         DISSECT_IMAGE_DISCARD_ANY = DISSECT_IMAGE_DISCARD_ON_LOOP |
                                     DISSECT_IMAGE_DISCARD |
                                     DISSECT_IMAGE_DISCARD_ON_CRYPTO,
+        DISSECT_IMAGE_GPT_ONLY = 16,         /* Only recognize images with GPT partition tables */
+        DISSECT_IMAGE_REQUIRE_ROOT = 32,     /* Don't accept disks without root partition */
 } DissectImageFlags;
 
 struct DissectedImage {
@@ -76,7 +79,7 @@ struct DissectedImage {
         DissectedPartition partitions[_PARTITION_DESIGNATOR_MAX];
 };
 
-int dissect_image(int fd, const void *root_hash, size_t root_hash_size, DissectedImage **ret);
+int dissect_image(int fd, const void *root_hash, size_t root_hash_size, DissectImageFlags flags, DissectedImage **ret);
 
 DissectedImage* dissected_image_unref(DissectedImage *m);
 DEFINE_TRIVIAL_CLEANUP_FUNC(DissectedImage*, dissected_image_unref);

src/shared/fstab-util.c

@@ -38,7 +38,7 @@ bool fstab_is_mount_point(const char *mount) {
         _cleanup_endmntent_ FILE *f = NULL;
         struct mntent *m;
 
-        f = setmntent("/etc/fstab", "r");
+        f = setmntent("/etc/fstab", "re");
         if (!f)
                 return false;
 

src/shared/machine-image.c

@@ -99,6 +99,16 @@ static char **image_settings_path(Image *image) {
         return ret;
 }
 
+static char *image_roothash_path(Image *image) {
+        const char *fn;
+
+        assert(image);
+
+        fn = strjoina(image->name, ".roothash");
+
+        return file_in_same_dir(image->path, fn);
+}
+
 static int image_new(
                 ImageType t,
                 const char *pretty,
@@ -397,6 +407,7 @@ void image_hashmap_free(Hashmap *map) {
 int image_remove(Image *i) {
         _cleanup_release_lock_file_ LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT;
         _cleanup_strv_free_ char **settings = NULL;
+        _cleanup_free_ char *roothash = NULL;
         char **j;
         int r;
 
@@ -409,6 +420,10 @@ int image_remove(Image *i) {
         if (!settings)
                 return -ENOMEM;
 
+        roothash = image_roothash_path(i);
+        if (!roothash)
+                return -ENOMEM;
+
         /* Make sure we don't interfere with a running nspawn */
         r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock);
         if (r < 0)
@@ -445,14 +460,17 @@ int image_remove(Image *i) {
                         log_debug_errno(errno, "Failed to unlink %s, ignoring: %m", *j);
         }
 
+        if (unlink(roothash) < 0 && errno != ENOENT)
+                log_debug_errno(errno, "Failed to unlink %s, ignoring: %m", roothash);
+
         return 0;
 }
 
-static int rename_settings_file(const char *path, const char *new_name) {
+static int rename_auxiliary_file(const char *path, const char *new_name, const char *suffix) {
         _cleanup_free_ char *rs = NULL;
         const char *fn;
 
-        fn = strjoina(new_name, ".nspawn");
+        fn = strjoina(new_name, suffix);
 
         rs = file_in_same_dir(path, fn);
         if (!rs)
@@ -463,7 +481,7 @@ static int rename_settings_file(const char *path, const char *new_name) {
 
 int image_rename(Image *i, const char *new_name) {
         _cleanup_release_lock_file_ LockFile global_lock = LOCK_FILE_INIT, local_lock = LOCK_FILE_INIT, name_lock = LOCK_FILE_INIT;
-        _cleanup_free_ char *new_path = NULL, *nn = NULL;
+        _cleanup_free_ char *new_path = NULL, *nn = NULL, *roothash = NULL;
         _cleanup_strv_free_ char **settings = NULL;
         unsigned file_attr = 0;
         char **j;
@@ -481,6 +499,10 @@ int image_rename(Image *i, const char *new_name) {
         if (!settings)
                 return -ENOMEM;
 
+        roothash = image_roothash_path(i);
+        if (!roothash)
+                return -ENOMEM;
+
         /* Make sure we don't interfere with a running nspawn */
         r = image_path_lock(i->path, LOCK_EX|LOCK_NB, &global_lock, &local_lock);
         if (r < 0)
@@ -550,19 +572,23 @@ int image_rename(Image *i, const char *new_name) {
         nn = NULL;
 
         STRV_FOREACH(j, settings) {
-                r = rename_settings_file(*j, new_name);
+                r = rename_auxiliary_file(*j, new_name, ".nspawn");
                 if (r < 0 && r != -ENOENT)
                         log_debug_errno(r, "Failed to rename settings file %s, ignoring: %m", *j);
         }
 
+        r = rename_auxiliary_file(roothash, new_name, ".roothash");
+        if (r < 0 && r != -ENOENT)
+                log_debug_errno(r, "Failed to rename roothash file %s, ignoring: %m", roothash);
+
         return 0;
 }
 
-static int clone_settings_file(const char *path, const char *new_name) {
+static int clone_auxiliary_file(const char *path, const char *new_name, const char *suffix) {
         _cleanup_free_ char *rs = NULL;
         const char *fn;
 
-        fn = strjoina(new_name, ".nspawn");
+        fn = strjoina(new_name, suffix);
 
         rs = file_in_same_dir(path, fn);
         if (!rs)
@@ -574,6 +600,7 @@ static int clone_settings_file(const char *path, const char *new_name) {
 int image_clone(Image *i, const char *new_name, bool read_only) {
         _cleanup_release_lock_file_ LockFile name_lock = LOCK_FILE_INIT;
         _cleanup_strv_free_ char **settings = NULL;
+        _cleanup_free_ char *roothash = NULL;
         const char *new_path;
         char **j;
         int r;
@@ -587,6 +614,10 @@ int image_clone(Image *i, const char *new_name, bool read_only) {
         if (!settings)
                 return -ENOMEM;
 
+        roothash = image_roothash_path(i);
+        if (!roothash)
+                return -ENOMEM;
+
         /* Make sure nobody takes the new name, between the time we
          * checked it is currently unused in all search paths, and the
          * time we take possession of it */
@@ -636,11 +667,15 @@ int image_clone(Image *i, const char *new_name, bool read_only) {
                 return r;
 
         STRV_FOREACH(j, settings) {
-                r = clone_settings_file(*j, new_name);
+                r = clone_auxiliary_file(*j, new_name, ".nspawn");
                 if (r < 0 && r != -ENOENT)
                         log_debug_errno(r, "Failed to clone settings %s, ignoring: %m", *j);
         }
 
+        r = clone_auxiliary_file(roothash, new_name, ".roothash");
+        if (r < 0 && r != -ENOENT)
+                log_debug_errno(r, "Failed to clone root hash file %s, ignoring: %m", roothash);
+
         return 0;
 }
 

src/shared/switch-root.c

@@ -28,123 +28,102 @@
 
 #include "base-filesystem.h"
 #include "fd-util.h"
+#include "fs-util.h"
 #include "log.h"
 #include "missing.h"
 #include "mkdir.h"
+#include "mount-util.h"
 #include "path-util.h"
 #include "rm-rf.h"
 #include "stdio-util.h"
 #include "string-util.h"
+#include "strv.h"
 #include "switch-root.h"
 #include "user-util.h"
 #include "util.h"
 
-int switch_root(const char *new_root, const char *oldroot, bool detach_oldroot,  unsigned long mountflags) {
-
-        /*  Don't try to unmount/move the old "/", there's no way to do it. */
-        static const char move_mounts[] =
-                "/dev\0"
-                "/proc\0"
-                "/sys\0"
-                "/run\0";
+int switch_root(const char *new_root,
+                const char *old_root_after, /* path below the new root, where to place the old root after the transition */
+                bool unmount_old_root,
+                unsigned long mount_flags) {  /* MS_MOVE or MS_BIND */
 
+        _cleanup_free_ char *resolved_old_root_after = NULL;
         _cleanup_close_ int old_root_fd = -1;
-        struct stat new_root_stat;
         bool old_root_remove;
-        const char *i, *temporary_old_root;
+        const char *i;
+        int r;
+
+        assert(new_root);
+        assert(old_root_after);
 
         if (path_equal(new_root, "/"))
                 return 0;
 
-        temporary_old_root = strjoina(new_root, oldroot);
-        mkdir_p_label(temporary_old_root, 0755);
-
+        /* Check if we shall remove the contents of the old root */
         old_root_remove = in_initrd();
+        if (old_root_remove) {
+                old_root_fd = open("/", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY|O_DIRECTORY);
+                if (old_root_fd < 0)
+                        return log_error_errno(errno, "Failed to open root directory: %m");
+        }
 
-        if (stat(new_root, &new_root_stat) < 0)
-                return log_error_errno(errno, "Failed to stat directory %s: %m", new_root);
+        /* Determine where we shall place the old root after the transition */
+        r = chase_symlinks(old_root_after, new_root, CHASE_PREFIX_ROOT|CHASE_NONEXISTENT, &resolved_old_root_after);
+        if (r < 0)
+                return log_error_errno(r, "Failed to resolve %s/%s: %m", new_root, old_root_after);
+        if (r == 0) /* Doesn't exist yet. Let's create it */
+                (void) mkdir_p_label(resolved_old_root_after, 0755);
 
-        /* Work-around for kernel design: the kernel refuses switching
-         * root if any file systems are mounted MS_SHARED. Hence
+        /* Work-around for kernel design: the kernel refuses MS_MOVE if any file systems are mounted MS_SHARED. Hence
          * remount them MS_PRIVATE here as a work-around.
          *
          * https://bugzilla.redhat.com/show_bug.cgi?id=847418 */
         if (mount(NULL, "/", NULL, MS_REC|MS_PRIVATE, NULL) < 0)
-                log_warning_errno(errno, "Failed to make \"/\" private mount: %m");
+                return log_error_errno(errno, "Failed to set \"/\" mount propagation to private: %m");
 
-        NULSTR_FOREACH(i, move_mounts) {
-                char new_mount[PATH_MAX];
-                struct stat sb;
-                size_t n;
+        FOREACH_STRING(i, "/sys", "/dev", "/run", "/proc") {
+                _cleanup_free_ char *chased = NULL;
 
-                n = snprintf(new_mount, sizeof new_mount, "%s%s", new_root, i);
-                if (n >= sizeof new_mount) {
-                        bool move = mountflags & MS_MOVE;
+                r = chase_symlinks(i, new_root, CHASE_PREFIX_ROOT|CHASE_NONEXISTENT, &chased);
+                if (r < 0)
+                        return log_error_errno(r, "Failed to resolve %s/%s: %m", new_root, i);
+                if (r > 0) {
+                        /* Already exists. Let's see if it is a mount point already. */
+                        r = path_is_mount_point(chased, NULL, 0);
+                        if (r < 0)
+                                return log_error_errno(r, "Failed to determine whether %s is a mount point: %m", chased);
+                        if (r > 0) /* If it is already mounted, then do nothing */
+                                continue;
+                } else
+                         /* Doesn't exist yet? */
+                        (void) mkdir_p_label(chased, 0755);
 
-                        log_warning("New path is too long, %s: %s%s",
-                                    move ? "forcing unmount instead" : "ignoring",
-                                    new_root, i);
-
-                        if (move)
-                                if (umount2(i, MNT_FORCE) < 0)
-                                        log_warning_errno(errno, "Failed to unmount %s: %m", i);
-                        continue;
-                }
-
-                mkdir_p_label(new_mount, 0755);
-
-                if (stat(new_mount, &sb) < 0 ||
-                    sb.st_dev != new_root_stat.st_dev) {
-
-                        /* Mount point seems to be mounted already or
-                         * stat failed. Unmount the old mount point. */
-                        if (umount2(i, MNT_DETACH) < 0)
-                                log_warning_errno(errno, "Failed to unmount %s: %m", i);
-                        continue;
-                }
-
-                if (mount(i, new_mount, NULL, mountflags, NULL) < 0) {
-                        if (mountflags & MS_MOVE) {
-                                log_error_errno(errno, "Failed to move mount %s to %s, forcing unmount: %m", i, new_mount);
-
-                                if (umount2(i, MNT_FORCE) < 0)
-                                        log_warning_errno(errno, "Failed to unmount %s: %m", i);
-
-                        } else if (mountflags & MS_BIND)
-                                log_error_errno(errno, "Failed to bind mount %s to %s: %m", i, new_mount);
-                }
+                if (mount(i, chased, NULL, mount_flags, NULL) < 0)
+                        return log_error_errno(r, "Failed to mount %s to %s: %m", i, chased);
         }
 
-        /* Do not fail, if base_filesystem_create() fails. Not all
-         * switch roots are like base_filesystem_create() wants them
-         * to look like. They might even boot, if they are RO and
-         * don't have the FS layout. Just ignore the error and
-         * switch_root() nevertheless. */
+        /* Do not fail if base_filesystem_create() fails. Not all switch roots are like base_filesystem_create() wants
+         * them to look like. They might even boot, if they are RO and don't have the FS layout. Just ignore the error
+         * and switch_root() nevertheless. */
         (void) base_filesystem_create(new_root, UID_INVALID, GID_INVALID);
 
         if (chdir(new_root) < 0)
                 return log_error_errno(errno, "Failed to change directory to %s: %m", new_root);
 
-        if (old_root_remove) {
-                old_root_fd = open("/", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY|O_DIRECTORY);
-                if (old_root_fd < 0)
-                        log_warning_errno(errno, "Failed to open root directory: %m");
-        }
-
-        /* We first try a pivot_root() so that we can umount the old
-         * root dir. In many cases (i.e. where rootfs is /), that's
-         * not possible however, and hence we simply overmount root */
-        if (pivot_root(new_root, temporary_old_root) >= 0) {
+        /* We first try a pivot_root() so that we can umount the old root dir. In many cases (i.e. where rootfs is /),
+         * that's not possible however, and hence we simply overmount root */
+        if (pivot_root(new_root, resolved_old_root_after) >= 0) {
 
                 /* Immediately get rid of the old root, if detach_oldroot is set.
                  * Since we are running off it we need to do this lazily. */
-                if (detach_oldroot && umount2(oldroot, MNT_DETACH) < 0)
-                        log_error_errno(errno, "Failed to lazily umount old root dir %s, %s: %m",
-                                  oldroot,
-                                  errno == ENOENT ? "ignoring" : "leaving it around");
+                if (unmount_old_root) {
+                        r = umount_recursive(old_root_after, MNT_DETACH);
+                        if (r < 0)
+                                log_warning_errno(r, "Failed to unmount old root directory tree, ignoring: %m");
+                }
 
         } else if (mount(new_root, "/", NULL, MS_MOVE, NULL) < 0)
-                return log_error_errno(errno, "Failed to mount moving %s to /: %m", new_root);
+                return log_error_errno(errno, "Failed to move %s to /: %m", new_root);
 
         if (chroot(".") < 0)
                 return log_error_errno(errno, "Failed to change root: %m");

src/shared/volatile-util.c

@@ -0,0 +1,68 @@
+/***
+  This file is part of systemd.
+
+  Copyright 2015 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
+  (at your option) any later version.
+
+  systemd is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+#include "alloc-util.h"
+#include "macro.h"
+#include "parse-util.h"
+#include "proc-cmdline.h"
+#include "string-util.h"
+#include "volatile-util.h"
+
+VolatileMode volatile_mode_from_string(const char *s) {
+        int b;
+
+        if (isempty(s))
+                return _VOLATILE_MODE_INVALID;
+
+        b = parse_boolean(s);
+        if (b > 0)
+                return VOLATILE_YES;
+        if (b == 0)
+                return VOLATILE_NO;
+
+        if (streq(s, "state"))
+                return VOLATILE_STATE;
+
+        return _VOLATILE_MODE_INVALID;
+}
+
+int query_volatile_mode(VolatileMode *ret) {
+        _cleanup_free_ char *mode = NULL;
+        VolatileMode m = VOLATILE_NO;
+        int r;
+
+        r = proc_cmdline_get_key("systemd.volatile", PROC_CMDLINE_VALUE_OPTIONAL, &mode);
+        if (r < 0)
+                return r;
+        if (r == 0)
+                goto finish;
+
+        if (mode) {
+                m = volatile_mode_from_string(mode);
+                if (m < 0)
+                        return -EINVAL;
+        } else
+                m = VOLATILE_YES;
+
+        r = 1;
+
+finish:
+        *ret = m;
+        return r;
+}

src/shared/volatile-util.h

@@ -0,0 +1,32 @@
+#pragma once
+
+/***
+  This file is part of systemd.
+
+  Copyright 2016 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
+  (at your option) any later version.
+
+  systemd is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+typedef enum VolatileMode {
+        VOLATILE_NO,
+        VOLATILE_YES,
+        VOLATILE_STATE,
+        _VOLATILE_MODE_MAX,
+        _VOLATILE_MODE_INVALID = -1
+} VolatileMode;
+
+VolatileMode volatile_mode_from_string(const char *s);
+
+int query_volatile_mode(VolatileMode *ret);

src/test/test-dissect-image.c

@@ -43,7 +43,7 @@ int main(int argc, char *argv[]) {
                 return EXIT_FAILURE;
         }
 
-        r = dissect_image(d->fd, NULL, 0, &m);
+        r = dissect_image(d->fd, NULL, 0, DISSECT_IMAGE_REQUIRE_ROOT, &m);
         if (r < 0) {
                 log_error_errno(r, "Failed to dissect image: %m");
                 return EXIT_FAILURE;

src/test/test-execute.c

@@ -33,6 +33,7 @@
 #ifdef HAVE_SECCOMP
 #include "seccomp-util.h"
 #endif
+#include "stat-util.h"
 #include "test-helper.h"
 #include "unit.h"
 #include "util.h"
@@ -188,15 +189,27 @@ static void test_exec_protectkernelmodules(Manager *m) {
 }
 
 static void test_exec_readonlypaths(Manager *m) {
+
+        if (path_is_read_only_fs("/var") > 0)
+                return;
+
         test(m, "exec-readonlypaths.service", 0, CLD_EXITED);
         test(m, "exec-readonlypaths-mount-propagation.service", 0, CLD_EXITED);
 }
 
 static void test_exec_readwritepaths(Manager *m) {
+
+        if (path_is_read_only_fs("/") > 0)
+                return;
+
         test(m, "exec-readwritepaths-mount-propagation.service", 0, CLD_EXITED);
 }
 
 static void test_exec_inaccessiblepaths(Manager *m) {
+
+        if (path_is_read_only_fs("/") > 0)
+                return;
+
         test(m, "exec-inaccessiblepaths-mount-propagation.service", 0, CLD_EXITED);
 }
 

src/test/test-hexdecoct.c

@@ -87,27 +87,19 @@ static void test_undecchar(void) {
 }
 
 static void test_unhexmem(void) {
-        const char *hex = "efa214921";
+        const char *hex = "efa2149213";
         const char *hex_invalid = "efa214921o";
         _cleanup_free_ char *hex2 = NULL;
         _cleanup_free_ void *mem = NULL;
         size_t len;
 
-        assert_se(unhexmem(hex, strlen(hex), &mem, &len) == 0);
-        assert_se(unhexmem(hex, strlen(hex) + 1, &mem, &len) == -EINVAL);
         assert_se(unhexmem(hex_invalid, strlen(hex_invalid), &mem, &len) == -EINVAL);
+        assert_se(unhexmem(hex, strlen(hex) + 1, &mem, &len) == -EINVAL);
+        assert_se(unhexmem(hex, strlen(hex) - 1, &mem, &len) == -EINVAL);
+        assert_se(unhexmem(hex, strlen(hex), &mem, &len) == 0);
 
         assert_se((hex2 = hexmem(mem, len)));
-
-        free(mem);
-
-        assert_se(memcmp(hex, hex2, strlen(hex)) == 0);
-
-        free(hex2);
-
-        assert_se(unhexmem(hex, strlen(hex) - 1, &mem, &len) == 0);
-        assert_se((hex2 = hexmem(mem, len)));
-        assert_se(memcmp(hex, hex2, strlen(hex) - 1) == 0);
+        assert_se(streq(hex, hex2));
 }
 
 /* https://tools.ietf.org/html/rfc4648#section-10 */

src/test/test-proc-cmdline.c

@@ -35,8 +35,8 @@ static int parse_item(const char *key, const char *value, void *data) {
         return 0;
 }
 
-static void test_parse_proc_cmdline(void) {
-        assert_se(parse_proc_cmdline(parse_item, &obj, true) >= 0);
+static void test_proc_cmdline_parse(void) {
+        assert_se(proc_cmdline_parse(parse_item, &obj, true) >= 0);
 }
 
 static void test_runlevel_to_target(void) {
@@ -55,11 +55,101 @@ static void test_runlevel_to_target(void) {
         assert_se(streq_ptr(runlevel_to_target("rd.rescue"), SPECIAL_RESCUE_TARGET));
 }
 
+static void test_proc_cmdline_get_key(void) {
+        _cleanup_free_ char *value = NULL;
+
+        putenv((char*) "SYSTEMD_PROC_CMDLINE=foo_bar=quux wuff-piep=tuet zumm");
+
+        assert_se(proc_cmdline_get_key("", 0, &value) == -EINVAL);
+        assert_se(proc_cmdline_get_key("abc", 0, NULL) == 0);
+        assert_se(proc_cmdline_get_key("abc", 0, &value) == 0 && value == NULL);
+        assert_se(proc_cmdline_get_key("abc", PROC_CMDLINE_VALUE_OPTIONAL, &value) == 0 && value == NULL);
+
+        assert_se(proc_cmdline_get_key("foo_bar", 0, &value) > 0 && streq_ptr(value, "quux"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("foo_bar", PROC_CMDLINE_VALUE_OPTIONAL, &value) > 0 && streq_ptr(value, "quux"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("foo-bar", 0, &value) > 0 && streq_ptr(value, "quux"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("foo-bar", PROC_CMDLINE_VALUE_OPTIONAL, &value) > 0 && streq_ptr(value, "quux"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("foo-bar", 0, NULL) == 0);
+        assert_se(proc_cmdline_get_key("foo-bar", PROC_CMDLINE_VALUE_OPTIONAL, NULL) == -EINVAL);
+
+        assert_se(proc_cmdline_get_key("wuff-piep", 0, &value) > 0 && streq_ptr(value, "tuet"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("wuff-piep", PROC_CMDLINE_VALUE_OPTIONAL, &value) > 0 && streq_ptr(value, "tuet"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("wuff_piep", 0, &value) > 0 && streq_ptr(value, "tuet"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("wuff_piep", PROC_CMDLINE_VALUE_OPTIONAL, &value) > 0 && streq_ptr(value, "tuet"));
+        value = mfree(value);
+        assert_se(proc_cmdline_get_key("wuff_piep", 0, NULL) == 0);
+        assert_se(proc_cmdline_get_key("wuff_piep", PROC_CMDLINE_VALUE_OPTIONAL, NULL) == -EINVAL);
+
+        assert_se(proc_cmdline_get_key("zumm", 0, &value) == 0 && value == NULL);
+        assert_se(proc_cmdline_get_key("zumm", PROC_CMDLINE_VALUE_OPTIONAL, &value) > 0 && value == NULL);
+        assert_se(proc_cmdline_get_key("zumm", 0, NULL) > 0);
+}
+
+static void test_proc_cmdline_get_bool(void) {
+        bool value = false;
+
+        putenv((char*) "SYSTEMD_PROC_CMDLINE=foo_bar bar-waldo=1 x_y-z=0 quux=miep");
+
+        assert_se(proc_cmdline_get_bool("", &value) == -EINVAL);
+        assert_se(proc_cmdline_get_bool("abc", &value) == 0 && value == false);
+        assert_se(proc_cmdline_get_bool("foo_bar", &value) > 0 && value == true);
+        assert_se(proc_cmdline_get_bool("foo-bar", &value) > 0 && value == true);
+        assert_se(proc_cmdline_get_bool("bar-waldo", &value) > 0 && value == true);
+        assert_se(proc_cmdline_get_bool("bar_waldo", &value) > 0 && value == true);
+        assert_se(proc_cmdline_get_bool("x_y-z", &value) > 0 && value == false);
+        assert_se(proc_cmdline_get_bool("x-y-z", &value) > 0 && value == false);
+        assert_se(proc_cmdline_get_bool("x-y_z", &value) > 0 && value == false);
+        assert_se(proc_cmdline_get_bool("x_y_z", &value) > 0 && value == false);
+        assert_se(proc_cmdline_get_bool("quux", &value) == -EINVAL && value == false);
+}
+
+static void test_proc_cmdline_key_streq(void) {
+
+        assert_se(proc_cmdline_key_streq("", ""));
+        assert_se(proc_cmdline_key_streq("a", "a"));
+        assert_se(!proc_cmdline_key_streq("", "a"));
+        assert_se(!proc_cmdline_key_streq("a", ""));
+        assert_se(proc_cmdline_key_streq("a", "a"));
+        assert_se(!proc_cmdline_key_streq("a", "b"));
+        assert_se(proc_cmdline_key_streq("x-y-z", "x-y-z"));
+        assert_se(proc_cmdline_key_streq("x-y-z", "x_y_z"));
+        assert_se(proc_cmdline_key_streq("x-y-z", "x-y_z"));
+        assert_se(proc_cmdline_key_streq("x-y-z", "x_y-z"));
+        assert_se(proc_cmdline_key_streq("x_y-z", "x-y_z"));
+        assert_se(!proc_cmdline_key_streq("x_y-z", "x-z_z"));
+}
+
+static void test_proc_cmdline_key_startswith(void) {
+
+        assert_se(proc_cmdline_key_startswith("", ""));
+        assert_se(proc_cmdline_key_startswith("x", ""));
+        assert_se(!proc_cmdline_key_startswith("", "x"));
+        assert_se(proc_cmdline_key_startswith("x", "x"));
+        assert_se(!proc_cmdline_key_startswith("x", "y"));
+        assert_se(!proc_cmdline_key_startswith("foo-bar", "quux"));
+        assert_se(proc_cmdline_key_startswith("foo-bar", "foo"));
+        assert_se(proc_cmdline_key_startswith("foo-bar", "foo-bar"));
+        assert_se(proc_cmdline_key_startswith("foo-bar", "foo_bar"));
+        assert_se(proc_cmdline_key_startswith("foo-bar", "foo_"));
+        assert_se(!proc_cmdline_key_startswith("foo-bar", "foo_xx"));
+}
+
 int main(void) {
         log_parse_environment();
         log_open();
 
-        test_parse_proc_cmdline();
+        test_proc_cmdline_parse();
+        test_proc_cmdline_key_streq();
+        test_proc_cmdline_key_startswith();
+        test_proc_cmdline_get_key();
+        test_proc_cmdline_get_bool();
         test_runlevel_to_target();
 
         return 0;

src/test/test-stat-util.c

@@ -18,12 +18,14 @@
 ***/
 
 #include <fcntl.h>
+#include <linux/magic.h>
 #include <unistd.h>
 
 #include "alloc-util.h"
 #include "fd-util.h"
 #include "fileio.h"
 #include "macro.h"
+#include "missing.h"
 #include "stat-util.h"
 
 static void test_files_same(void) {
@@ -60,9 +62,33 @@ static void test_is_symlink(void) {
         unlink(name_link);
 }
 
+static void test_path_is_os_tree(void) {
+        assert_se(path_is_os_tree("/") > 0);
+        assert_se(path_is_os_tree("/etc") == 0);
+        assert_se(path_is_os_tree("/idontexist") == -ENOENT);
+}
+
+static void test_path_check_fstype(void) {
+        assert_se(path_check_fstype("/run", TMPFS_MAGIC) > 0);
+        assert_se(path_check_fstype("/run", BTRFS_SUPER_MAGIC) == 0);
+        assert_se(path_check_fstype("/proc", PROC_SUPER_MAGIC) > 0);
+        assert_se(path_check_fstype("/proc", BTRFS_SUPER_MAGIC) == 0);
+        assert_se(path_check_fstype("/proc", BTRFS_SUPER_MAGIC) == 0);
+        assert_se(path_check_fstype("/i-dont-exist", BTRFS_SUPER_MAGIC) == -ENOENT);
+}
+
+static void test_path_is_temporary_fs(void) {
+        assert_se(path_is_temporary_fs("/run") > 0);
+        assert_se(path_is_temporary_fs("/proc") == 0);
+        assert_se(path_is_temporary_fs("/i-dont-exist") == -ENOENT);
+}
+
 int main(int argc, char *argv[]) {
         test_files_same();
         test_is_symlink();
+        test_path_is_os_tree();
+        test_path_check_fstype();
+        test_path_is_temporary_fs();
 
         return 0;
 }

src/udev/net/link-config.c

@@ -192,14 +192,9 @@ static int load_link(link_config_ctx *ctx, const char *filename) {
 }
 
 static bool enable_name_policy(void) {
-        _cleanup_free_ char *value = NULL;
-        int r;
+        bool b;
 
-        r = get_proc_cmdline_key("net.ifnames=", &value);
-        if (r > 0 && streq(value, "0"))
-            return false;
-
-        return true;
+        return proc_cmdline_get_bool("net.ifnames", &b) <= 0 || b;
 }
 
 int link_config_load(link_config_ctx *ctx) {

src/udev/udevd.c

@@ -1357,10 +1357,10 @@ static int listen_fds(int *rctrl, int *rnetlink) {
 
 /*
  * read the kernel command line, in case we need to get into debug mode
- *   udev.log-priority=<level>                 syslog priority
- *   udev.children-max=<number of workers>     events are fully serialized if set to 1
- *   udev.exec-delay=<number of seconds>       delay execution of every executed program
- *   udev.event-timeout=<number of seconds>    seconds to wait before terminating an event
+ *   udev.log_priority=<level>                 syslog priority
+ *   udev.children_max=<number of workers>     events are fully serialized if set to 1
+ *   udev.exec_delay=<number of seconds>       delay execution of every executed program
+ *   udev.event_timeout=<number of seconds>    seconds to wait before terminating an event
  */
 static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
         int r = 0;
@@ -1370,25 +1370,46 @@ static int parse_proc_cmdline_item(const char *key, const char *value, void *dat
         if (!value)
                 return 0;
 
-        if (streq(key, "udev.log-priority") && value) {
+        if (proc_cmdline_key_streq(key, "udev.log_priority")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = util_log_priority(value);
                 if (r >= 0)
                         log_set_max_level(r);
-        } else if (streq(key, "udev.event-timeout") && value) {
+
+        } else if (proc_cmdline_key_streq(key, "udev.event_timeout")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = safe_atou64(value, &arg_event_timeout_usec);
                 if (r >= 0) {
                         arg_event_timeout_usec *= USEC_PER_SEC;
                         arg_event_timeout_warn_usec = (arg_event_timeout_usec / 3) ? : 1;
                 }
-        } else if (streq(key, "udev.children-max") && value)
+
+        } else if (proc_cmdline_key_streq(key, "udev.children_max")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = safe_atou(value, &arg_children_max);
-        else if (streq(key, "udev.exec-delay") && value)
+
+        } else if (proc_cmdline_key_streq(key, "udev.exec_delay")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
                 r = safe_atoi(value, &arg_exec_delay);
-        else if (startswith(key, "udev."))
+
+        } else if (startswith(key, "udev."))
                 log_warning("Unknown udev kernel command line option \"%s\"", key);
 
         if (r < 0)
                 log_warning_errno(r, "Failed to parse \"%s=%s\", ignoring: %m", key, value);
+
         return 0;
 }
 
@@ -1649,7 +1670,7 @@ int main(int argc, char *argv[]) {
         if (r <= 0)
                 goto exit;
 
-        r = parse_proc_cmdline(parse_proc_cmdline_item, NULL, true);
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, PROC_CMDLINE_STRIP_RD_PREFIX);
         if (r < 0)
                 log_warning_errno(r, "failed to parse kernel command line, ignoring: %m");
 

src/veritysetup/Makefile

@@ -0,0 +1 @@
+../Makefile

src/veritysetup/veritysetup-generator.c

@@ -0,0 +1,253 @@
+/***
+  This file is part of systemd.
+
+  Copyright 2016 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
+  (at your option) any later version.
+
+  systemd is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+#include <stdbool.h>
+#include <stdlib.h>
+#include <sys/stat.h>
+#include <unistd.h>
+
+#include "alloc-util.h"
+#include "fd-util.h"
+#include "fileio.h"
+#include "fstab-util.h"
+#include "hexdecoct.h"
+#include "id128-util.h"
+#include "mkdir.h"
+#include "parse-util.h"
+#include "proc-cmdline.h"
+#include "string-util.h"
+#include "unit-name.h"
+
+static char *arg_dest = NULL;
+static bool arg_enabled = true;
+static char *arg_root_hash = NULL;
+static char *arg_data_what = NULL;
+static char *arg_hash_what = NULL;
+
+static int create_device(void) {
+        _cleanup_free_ char *u = NULL, *v = NULL, *d = NULL, *e = NULL;
+        _cleanup_fclose_ FILE *f = NULL;
+        const char *p, *to;
+        int r;
+
+        /* If all three pieces of information are missing, then verity is turned off */
+        if (!arg_root_hash && !arg_data_what && !arg_hash_what)
+                return 0;
+
+        /* if one of them is missing however, the data is simply incomplete and this is an error */
+        if (!arg_root_hash)
+                log_error("Verity information incomplete, root hash unspecified.");
+        if (!arg_data_what)
+                log_error("Verity information incomplete, root data device unspecified.");
+        if (!arg_hash_what)
+                log_error("Verity information incomplete, root hash device unspecified.");
+
+        if (!arg_root_hash || !arg_data_what || !arg_hash_what)
+                return -EINVAL;
+
+        log_debug("Using root verity data device %s,\n"
+                  "                  hash device %s,\n"
+                  "                and root hash %s.", arg_data_what, arg_hash_what, arg_root_hash);
+
+        p = strjoina(arg_dest, "/systemd-veritysetup@root.service");
+
+        u = fstab_node_to_udev_node(arg_data_what);
+        if (!u)
+                return log_oom();
+        v = fstab_node_to_udev_node(arg_hash_what);
+        if (!v)
+                return log_oom();
+
+        r = unit_name_from_path(u, ".device", &d);
+        if (r < 0)
+                return log_error_errno(r, "Failed to to generate unit name: %m");
+        r = unit_name_from_path(v, ".device", &e);
+        if (r < 0)
+                return log_error_errno(r, "Failed to to generate unit name: %m");
+
+        f = fopen(p, "wxe");
+        if (!f)
+                return log_error_errno(errno, "Failed to create unit file %s: %m", p);
+
+        fprintf(f,
+                "# Automatically generated by systemd-veritysetup-generator\n\n"
+                "[Unit]\n"
+                "Description=Integrity Protection Setup for %%I\n"
+                "Documentation=man:systemd-veritysetup-generator(8) man:systemd-veritysetup@.service(8)\n"
+                "SourcePath=/proc/cmdline\n"
+                "DefaultDependencies=no\n"
+                "Conflicts=umount.target\n"
+                "BindsTo=%s %s\n"
+                "IgnoreOnIsolate=true\n"
+                "After=cryptsetup-pre.target %s %s\n"
+                "Before=cryptsetup.target umount.target\n"
+                "\n[Service]\n"
+                "Type=oneshot\n"
+                "RemainAfterExit=yes\n"
+                "ExecStart=" ROOTLIBEXECDIR "/systemd-veritysetup attach root '%s' '%s' '%s'\n"
+                "ExecStop=" ROOTLIBEXECDIR "/systemd-veritysetup detach root\n",
+                d, e,
+                d, e,
+                u, v, arg_root_hash);
+
+        r = fflush_and_check(f);
+        if (r < 0)
+                return log_error_errno(r, "Failed to write file %s: %m", p);
+
+        to = strjoina(arg_dest, "/cryptsetup.target.requires/systemd-veritysetup@root.service");
+
+        (void) mkdir_parents(to, 0755);
+        if (symlink("../systemd-veritysetup@root.service", to) < 0)
+                return log_error_errno(errno, "Failed to create symlink %s: %m", to);
+
+        return 0;
+}
+
+static int parse_proc_cmdline_item(const char *key, const char *value, void *data) {
+        int r;
+
+        if (streq(key, "systemd.verity")) {
+
+                r = value ? parse_boolean(value) : 1;
+                if (r < 0)
+                        log_warning("Failed to parse verity= kernel command line switch %s. Ignoring.", value);
+                else
+                        arg_enabled = r;
+
+        } else if (streq(key, "roothash")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = free_and_strdup(&arg_root_hash, value);
+                if (r < 0)
+                        return log_oom();
+
+        } else if (streq(key, "systemd.verity_root_data")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = free_and_strdup(&arg_data_what, value);
+                if (r < 0)
+                        return log_oom();
+
+        } else if (streq(key, "systemd.verity_root_hash")) {
+
+                if (proc_cmdline_value_missing(key, value))
+                        return 0;
+
+                r = free_and_strdup(&arg_hash_what, value);
+                if (r < 0)
+                        return log_oom();
+        }
+
+        return 0;
+}
+
+static int determine_devices(void) {
+        _cleanup_free_ void *m = NULL;
+        sd_id128_t root_uuid, verity_uuid;
+        char ids[37];
+        size_t l;
+        int r;
+
+        /* Try to automatically derive the root data and hash device paths from the root hash */
+
+        if (!arg_root_hash)
+                return 0;
+
+        if (arg_data_what && arg_hash_what)
+                return 0;
+
+        r = unhexmem(arg_root_hash, strlen(arg_root_hash), &m, &l);
+        if (r < 0)
+                return log_error_errno(r, "Failed to parse root hash: %s", arg_root_hash);
+        if (l < sizeof(sd_id128_t)) {
+                log_debug("Root hash is shorter than 128 bits (32 characters), ignoring for discovering verity partition.");
+                return 0;
+        }
+
+        if (!arg_data_what) {
+                memcpy(&root_uuid, m, sizeof(root_uuid));
+
+                arg_data_what = strjoin("/dev/disk/by-partuuid/", id128_to_uuid_string(root_uuid, ids));
+                if (!arg_data_what)
+                        return log_oom();
+        }
+
+        if (!arg_hash_what) {
+                memcpy(&verity_uuid, (uint8_t*) m + l - sizeof(verity_uuid), sizeof(verity_uuid));
+
+                arg_hash_what = strjoin("/dev/disk/by-partuuid/", id128_to_uuid_string(verity_uuid, ids));
+                if (!arg_hash_what)
+                        return log_oom();
+        }
+
+        return 1;
+}
+
+int main(int argc, char *argv[]) {
+        int r;
+
+        if (argc > 1 && argc != 4) {
+                log_error("This program takes three or no arguments.");
+                return EXIT_FAILURE;
+        }
+
+        if (argc > 1)
+                arg_dest = argv[1];
+
+        log_set_target(LOG_TARGET_SAFE);
+        log_parse_environment();
+        log_open();
+
+        umask(0022);
+
+        r = proc_cmdline_parse(parse_proc_cmdline_item, NULL, PROC_CMDLINE_STRIP_RD_PREFIX);
+        if (r < 0) {
+                log_warning_errno(r, "Failed to parse kernel command line: %m");
+                goto finish;
+        }
+
+        /* For now we only support the root device on verity. Later on we might want to add support for /etc/veritytab
+         * or similar to define additional mappings */
+
+        if (!arg_enabled) {
+                r = 0;
+                goto finish;
+        }
+
+        r = determine_devices();
+        if (r < 0)
+                goto finish;
+
+        r = create_device();
+        if (r < 0)
+                goto finish;
+
+        r = 0;
+
+finish:
+        free(arg_root_hash);
+        free(arg_data_what);
+        free(arg_hash_what);
+
+        return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
+}

src/veritysetup/veritysetup.c

@@ -0,0 +1,154 @@
+/***
+  This file is part of systemd.
+
+  Copyright 2016 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
+  (at your option) any later version.
+
+  systemd is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+#include <libcryptsetup.h>
+#include <stdio.h>
+#include <sys/stat.h>
+
+#include "log.h"
+#include "hexdecoct.h"
+#include "string-util.h"
+#include "alloc-util.h"
+
+static char *arg_root_hash = NULL;
+static char *arg_data_what = NULL;
+static char *arg_hash_what = NULL;
+
+static int help(void) {
+        printf("%s attach VOLUME DATADEVICE HASHDEVICE ROOTHASH\n"
+               "%s detach VOLUME\n\n"
+               "Attaches or detaches an integrity protected block device.\n",
+               program_invocation_short_name,
+               program_invocation_short_name);
+
+        return 0;
+}
+
+static void log_glue(int level, const char *msg, void *usrptr) {
+        log_debug("%s", msg);
+}
+
+int main(int argc, char *argv[]) {
+        struct crypt_device *cd = NULL;
+        int r;
+
+        if (argc <= 1) {
+                r = help();
+                goto finish;
+        }
+
+        if (argc < 3) {
+                log_error("This program requires at least two arguments.");
+                r = -EINVAL;
+                goto finish;
+        }
+
+        log_set_target(LOG_TARGET_AUTO);
+        log_parse_environment();
+        log_open();
+
+        umask(0022);
+
+        if (streq(argv[1], "attach")) {
+                _cleanup_free_ void *m = NULL;
+                crypt_status_info status;
+                size_t l;
+
+                if (argc < 6) {
+                        log_error("attach requires at least two arguments.");
+                        r = -EINVAL;
+                        goto finish;
+                }
+
+                r = unhexmem(argv[5], strlen(argv[5]), &m, &l);
+                if (r < 0) {
+                        log_error("Failed to parse root hash.");
+                        goto finish;
+                }
+
+                r = crypt_init(&cd, argv[4]);
+                if (r < 0) {
+                        log_error_errno(r, "Failed to open verity device %s: %m", argv[4]);
+                        goto finish;
+                }
+
+                crypt_set_log_callback(cd, log_glue, NULL);
+
+                status = crypt_status(cd, argv[2]);
+                if (status == CRYPT_ACTIVE || status == CRYPT_BUSY) {
+                        log_info("Volume %s already active.", argv[2]);
+                        r = 0;
+                        goto finish;
+                }
+
+                r = crypt_load(cd, CRYPT_VERITY, NULL);
+                if (r < 0) {
+                        log_error_errno(r, "Failed to load verity superblock: %m");
+                        goto finish;
+                }
+
+                r = crypt_set_data_device(cd, argv[3]);
+                if (r < 0) {
+                        log_error_errno(r, "Failed to configure data device: %m");
+                        goto finish;
+                }
+
+                r = crypt_activate_by_volume_key(cd, argv[2], m, l, CRYPT_ACTIVATE_READONLY);
+                if (r < 0) {
+                        log_error_errno(r, "Failed to set up verity device: %m");
+                        goto finish;
+                }
+
+        } else if (streq(argv[1], "detach")) {
+
+                r = crypt_init_by_name(&cd, argv[2]);
+                if (r == -ENODEV) {
+                        log_info("Volume %s already inactive.", argv[2]);
+                        goto finish;
+                } else if (r < 0) {
+                        log_error_errno(r, "crypt_init_by_name() failed: %m");
+                        goto finish;
+                }
+
+                crypt_set_log_callback(cd, log_glue, NULL);
+
+                r = crypt_deactivate(cd, argv[2]);
+                if (r < 0) {
+                        log_error_errno(r, "Failed to deactivate: %m");
+                        goto finish;
+                }
+
+        } else {
+                log_error("Unknown verb %s.", argv[1]);
+                r = -EINVAL;
+                goto finish;
+        }
+
+        r = 0;
+
+finish:
+        if (cd)
+                crypt_free(cd);
+
+        free(arg_root_hash);
+        free(arg_data_what);
+        free(arg_hash_what);
+
+        return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
+}

src/volatile-root/Makefile

@@ -0,0 +1 @@
+../Makefile

src/volatile-root/volatile-root.c

@@ -0,0 +1,157 @@
+/***
+  This file is part of systemd.
+
+  Copyright 2016 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 of the License, or
+  (at your option) any later version.
+
+  systemd is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+#include <sys/mount.h>
+
+#include "alloc-util.h"
+#include "fs-util.h"
+#include "mkdir.h"
+#include "mount-util.h"
+#include "stat-util.h"
+#include "volatile-util.h"
+#include "string-util.h"
+#include "path-util.h"
+
+static int make_volatile(const char *path) {
+        _cleanup_free_ char *old_usr = NULL;
+        int r;
+
+        r = path_is_mount_point(path, NULL, AT_SYMLINK_FOLLOW);
+        if (r < 0)
+                return log_error_errno(r, "Couldn't determine whether %s is a mount point: %m", path);
+        if (r == 0) {
+                log_error("%s is not a mount point.", path);
+                return -EINVAL;
+        }
+
+        r = path_is_temporary_fs(path);
+        if (r < 0)
+                return log_error_errno(r, "Couldn't determine whether %s is a temporary file system: %m", path);
+        if (r > 0) {
+                log_info("%s already is a temporary file system.", path);
+                return 0;
+        }
+
+        r = chase_symlinks("/usr", path, CHASE_PREFIX_ROOT, &old_usr);
+        if (r < 0)
+                return log_error_errno(r, "/usr not available in old root: %m");
+
+        r = mkdir_p("/run/systemd/volatile-sysroot", 0700);
+        if (r < 0)
+                return log_error_errno(r, "Couldn't generate volatile sysroot directory: %m");
+
+        r = mount_verbose(LOG_ERR, "tmpfs", "/run/systemd/volatile-sysroot", "tmpfs", MS_STRICTATIME, "mode=755");
+        if (r < 0)
+                goto finish_rmdir;
+
+        if (mkdir("/run/systemd/volatile-sysroot/usr", 0755) < 0) {
+                r = -errno;
+                goto finish_umount;
+        }
+
+        r = mount_verbose(LOG_ERR, old_usr, "/run/systemd/volatile-sysroot/usr", NULL, MS_BIND|MS_REC, NULL);
+        if (r < 0)
+                goto finish_umount;
+
+        r = bind_remount_recursive("/run/systemd/volatile-sysroot/usr", true, NULL);
+        if (r < 0)
+                goto finish_umount;
+
+        r = umount_recursive(path, 0);
+        if (r < 0) {
+                log_error_errno(r, "Failed to unmount %s: %m", path);
+                goto finish_umount;
+        }
+
+        if (mount(NULL, "/", NULL, MS_SLAVE|MS_REC, NULL) < 0)
+                log_warning_errno(errno, "Failed to remount %s MS_SLAVE|MS_REC: %m", path);
+
+        r = mount_verbose(LOG_ERR, "/run/systemd/volatile-sysroot", path, NULL, MS_MOVE, NULL);
+
+finish_umount:
+        (void) umount_recursive("/run/systemd/volatile-sysroot", 0);
+
+finish_rmdir:
+        (void) rmdir("/run/systemd/volatile-sysroot");
+
+        return r;
+}
+
+int main(int argc, char *argv[]) {
+        VolatileMode m = _VOLATILE_MODE_INVALID;
+        const char *path;
+        int r;
+
+        log_set_target(LOG_TARGET_AUTO);
+        log_parse_environment();
+        log_open();
+
+        if (argc > 3) {
+                log_error("Too many arguments. Expected directory and mode.");
+                r = -EINVAL;
+                goto finish;
+        }
+
+        r = query_volatile_mode(&m);
+        if (r < 0) {
+                log_error_errno(r, "Failed to determine volatile mode from kernel command line.");
+                goto finish;
+        }
+        if (r == 0 && argc >= 2) {
+                /* The kernel command line always wins. However if nothing was set there, the argument passed here wins instead. */
+                m = volatile_mode_from_string(argv[1]);
+                if (m < 0) {
+                        log_error("Couldn't parse volatile mode: %s", argv[1]);
+                        r = -EINVAL;
+                        goto finish;
+                }
+        }
+
+        if (argc < 3)
+                path = "/sysroot";
+        else {
+                path = argv[2];
+
+                if (isempty(path)) {
+                        log_error("Directory name cannot be empty.");
+                        r = -EINVAL;
+                        goto finish;
+                }
+                if (!path_is_absolute(path)) {
+                        log_error("Directory must be specified as absolute path.");
+                        r = -EINVAL;
+                        goto finish;
+                }
+                if (path_equal(path, "/")) {
+                        log_error("Directory cannot be the root directory.");
+                        r = -EINVAL;
+                        goto finish;
+                }
+        }
+
+        if (m != VOLATILE_YES) {
+                r = 0;
+                goto finish;
+        }
+
+        r = make_volatile(path);
+
+finish:
+        return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
+}