+ u->cgroup_members_mask = 0;
+
+ if (u->type == UNIT_SLICE) {
+ Unit *member;
+ Iterator i;
+
+ SET_FOREACH(member, u->dependencies[UNIT_BEFORE], i) {
+
+ if (member == u)
+ continue;
+
+ if (UNIT_DEREF(member->slice) != u)
+ continue;
+
+ u->cgroup_members_mask |=
+ unit_get_own_mask(member) |
+ unit_get_members_mask(member);
+ }
+ }
+
+ u->cgroup_members_mask_valid = true;
+ return u->cgroup_members_mask;
+}
+
+CGroupMask unit_get_siblings_mask(Unit *u) {
+ assert(u);
+
+ /* Returns the mask of controllers all of the unit's siblings
+ * require, i.e. the members mask of the unit's parent slice
+ * if there is one. */
+
+ if (UNIT_ISSET(u->slice))
+ return unit_get_members_mask(UNIT_DEREF(u->slice));
+
+ return unit_get_own_mask(u) | unit_get_members_mask(u);
+}
+
+CGroupMask unit_get_subtree_mask(Unit *u) {
+
+ /* Returns the mask of this subtree, meaning of the group
+ * itself and its children. */
+
+ return unit_get_own_mask(u) | unit_get_members_mask(u);
+}
+
+CGroupMask unit_get_target_mask(Unit *u) {
+ CGroupMask mask;
+
+ /* This returns the cgroup mask of all controllers to enable
+ * for a specific cgroup, i.e. everything it needs itself,
+ * plus all that its children need, plus all that its siblings
+ * need. This is primarily useful on the legacy cgroup
+ * hierarchy, where we need to duplicate each cgroup in each
+ * hierarchy that shall be enabled for it. */
+
+ mask = unit_get_own_mask(u) | unit_get_members_mask(u) | unit_get_siblings_mask(u);
+ mask &= u->manager->cgroup_supported;
+
+ return mask;
+}
+
+CGroupMask unit_get_enable_mask(Unit *u) {
+ CGroupMask mask;
+
+ /* This returns the cgroup mask of all controllers to enable
+ * for the children of a specific cgroup. This is primarily
+ * useful for the unified cgroup hierarchy, where each cgroup
+ * controls which controllers are enabled for its children. */
+
+ mask = unit_get_members_mask(u);
+ mask &= u->manager->cgroup_supported;
+
+ return mask;
+}
+
+/* Recurse from a unit up through its containing slices, propagating
+ * mask bits upward. A unit is also member of itself. */
+void unit_update_cgroup_members_masks(Unit *u) {
+ CGroupMask m;
+ bool more;
+
+ assert(u);
+
+ /* Calculate subtree mask */
+ m = unit_get_subtree_mask(u);
+
+ /* See if anything changed from the previous invocation. If
+ * not, we're done. */
+ if (u->cgroup_subtree_mask_valid && m == u->cgroup_subtree_mask)
+ return;
+
+ more =
+ u->cgroup_subtree_mask_valid &&
+ ((m & ~u->cgroup_subtree_mask) != 0) &&
+ ((~m & u->cgroup_subtree_mask) == 0);
+
+ u->cgroup_subtree_mask = m;
+ u->cgroup_subtree_mask_valid = true;
+
+ if (UNIT_ISSET(u->slice)) {
+ Unit *s = UNIT_DEREF(u->slice);
+
+ if (more)
+ /* There's more set now than before. We
+ * propagate the new mask to the parent's mask
+ * (not caring if it actually was valid or
+ * not). */
+
+ s->cgroup_members_mask |= m;
+
+ else
+ /* There's less set now than before (or we
+ * don't know), we need to recalculate
+ * everything, so let's invalidate the
+ * parent's members mask */
+
+ s->cgroup_members_mask_valid = false;
+
+ /* And now make sure that this change also hits our
+ * grandparents */
+ unit_update_cgroup_members_masks(s);
+ }
+}
+
+static const char *migrate_callback(CGroupMask mask, void *userdata) {
+ Unit *u = userdata;
+
+ assert(mask != 0);
+ assert(u);
+
+ while (u) {
+ if (u->cgroup_path &&
+ u->cgroup_realized &&
+ (u->cgroup_realized_mask & mask) == mask)
+ return u->cgroup_path;
+
+ u = UNIT_DEREF(u->slice);
+ }
+
+ return NULL;
+}
+
+char *unit_default_cgroup_path(Unit *u) {
+ _cleanup_free_ char *escaped = NULL, *slice = NULL;
+ int r;
+
+ assert(u);
+
+ if (unit_has_name(u, SPECIAL_ROOT_SLICE))
+ return strdup(u->manager->cgroup_root);
+
+ if (UNIT_ISSET(u->slice) && !unit_has_name(UNIT_DEREF(u->slice), SPECIAL_ROOT_SLICE)) {
+ r = cg_slice_to_path(UNIT_DEREF(u->slice)->id, &slice);
+ if (r < 0)
+ return NULL;
+ }
+
+ escaped = cg_escape(u->id);
+ if (!escaped)
+ return NULL;
+
+ if (slice)
+ return strjoin(u->manager->cgroup_root, "/", slice, "/",
+ escaped);
+ else
+ return strjoin(u->manager->cgroup_root, "/", escaped);
+}
+
+int unit_set_cgroup_path(Unit *u, const char *path) {
+ _cleanup_free_ char *p = NULL;
+ int r;
+
+ assert(u);
+
+ if (path) {
+ p = strdup(path);
+ if (!p)
+ return -ENOMEM;
+ } else
+ p = NULL;
+
+ if (streq_ptr(u->cgroup_path, p))
+ return 0;
+
+ if (p) {
+ r = hashmap_put(u->manager->cgroup_unit, p, u);
+ if (r < 0)
+ return r;
+ }
+
+ unit_release_cgroup(u);
+
+ u->cgroup_path = p;
+ p = NULL;
+
+ return 1;
+}
+
+int unit_watch_cgroup(Unit *u) {
+ _cleanup_free_ char *events = NULL;
+ int r;
+
+ assert(u);
+
+ if (!u->cgroup_path)
+ return 0;
+
+ if (u->cgroup_inotify_wd >= 0)
+ return 0;
+
+ /* Only applies to the unified hierarchy */
+ if (!cg_unified(SYSTEMD_CGROUP_CONTROLLER))
+ return 0;
+
+ /* Don't watch the root slice, it's pointless. */
+ if (unit_has_name(u, SPECIAL_ROOT_SLICE))
+ return 0;
+
+ r = hashmap_ensure_allocated(&u->manager->cgroup_inotify_wd_unit, &trivial_hash_ops);
+ if (r < 0)
+ return log_oom();
+
+ r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, "cgroup.events", &events);
+ if (r < 0)
+ return log_oom();
+
+ u->cgroup_inotify_wd = inotify_add_watch(u->manager->cgroup_inotify_fd, events, IN_MODIFY);
+ if (u->cgroup_inotify_wd < 0) {
+
+ if (errno == ENOENT) /* If the directory is already
+ * gone we don't need to track
+ * it, so this is not an error */
+ return 0;
+
+ return log_unit_error_errno(u, errno, "Failed to add inotify watch descriptor for control group %s: %m", u->cgroup_path);
+ }
+
+ r = hashmap_put(u->manager->cgroup_inotify_wd_unit, INT_TO_PTR(u->cgroup_inotify_wd), u);
+ if (r < 0)
+ return log_unit_error_errno(u, r, "Failed to add inotify watch descriptor to hash map: %m");
+
+ return 0;
+}
+
+static int unit_create_cgroup(
+ Unit *u,
+ CGroupMask target_mask,
+ CGroupMask enable_mask) {
+
+ CGroupContext *c;
+ int r;
+
+ assert(u);
+
+ c = unit_get_cgroup_context(u);
+ if (!c)
+ return 0;
+
+ if (!u->cgroup_path) {
+ _cleanup_free_ char *path = NULL;
+
+ path = unit_default_cgroup_path(u);
+ if (!path)
+ return log_oom();
+
+ r = unit_set_cgroup_path(u, path);
+ if (r == -EEXIST)
+ return log_unit_error_errno(u, r, "Control group %s exists already.", path);
+ if (r < 0)
+ return log_unit_error_errno(u, r, "Failed to set unit's control group path to %s: %m", path);
+ }
+
+ /* First, create our own group */
+ r = cg_create_everywhere(u->manager->cgroup_supported, target_mask, u->cgroup_path);
+ if (r < 0)
+ return log_unit_error_errno(u, r, "Failed to create cgroup %s: %m", u->cgroup_path);
+
+ /* Start watching it */
+ (void) unit_watch_cgroup(u);
+
+ /* Enable all controllers we need */
+ r = cg_enable_everywhere(u->manager->cgroup_supported, enable_mask, u->cgroup_path);
+ if (r < 0)
+ log_unit_warning_errno(u, r, "Failed to enable controllers on cgroup %s, ignoring: %m", u->cgroup_path);
+
+ /* Keep track that this is now realized */
+ u->cgroup_realized = true;
+ u->cgroup_realized_mask = target_mask;
+ u->cgroup_enabled_mask = enable_mask;
+
+ if (u->type != UNIT_SLICE && !c->delegate) {
+
+ /* Then, possibly move things over, but not if
+ * subgroups may contain processes, which is the case
+ * for slice and delegation units. */
+ r = cg_migrate_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->cgroup_path, migrate_callback, u);
+ if (r < 0)
+ log_unit_warning_errno(u, r, "Failed to migrate cgroup from to %s, ignoring: %m", u->cgroup_path);
+ }
+
+ return 0;
+}
+
+int unit_attach_pids_to_cgroup(Unit *u) {
+ int r;
+ assert(u);
+
+ r = unit_realize_cgroup(u);
+ if (r < 0)
+ return r;
+
+ r = cg_attach_many_everywhere(u->manager->cgroup_supported, u->cgroup_path, u->pids, migrate_callback, u);
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
+static void cgroup_xattr_apply(Unit *u) {
+ char ids[SD_ID128_STRING_MAX];
+ int r;
+
+ assert(u);
+
+ if (!MANAGER_IS_SYSTEM(u->manager))
+ return;
+
+ if (sd_id128_is_null(u->invocation_id))
+ return;
+
+ r = cg_set_xattr(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path,
+ "trusted.invocation_id",
+ sd_id128_to_string(u->invocation_id, ids), 32,
+ 0);
+ if (r < 0)
+ log_unit_warning_errno(u, r, "Failed to set invocation ID on control group %s, ignoring: %m", u->cgroup_path);
+}
+
+static bool unit_has_mask_realized(Unit *u, CGroupMask target_mask, CGroupMask enable_mask) {
+ assert(u);
+
+ return u->cgroup_realized && u->cgroup_realized_mask == target_mask && u->cgroup_enabled_mask == enable_mask;
+}
+
+/* Check if necessary controllers and attributes for a unit are in place.
+ *
+ * If so, do nothing.
+ * If not, create paths, move processes over, and set attributes.
+ *
+ * Returns 0 on success and < 0 on failure. */
+static int unit_realize_cgroup_now(Unit *u, ManagerState state) {
+ CGroupMask target_mask, enable_mask;
+ int r;
+
+ assert(u);
+
+ if (u->in_cgroup_queue) {
+ LIST_REMOVE(cgroup_queue, u->manager->cgroup_queue, u);
+ u->in_cgroup_queue = false;
+ }
+
+ target_mask = unit_get_target_mask(u);
+ enable_mask = unit_get_enable_mask(u);
+
+ if (unit_has_mask_realized(u, target_mask, enable_mask))
+ return 0;
+
+ /* First, realize parents */
+ if (UNIT_ISSET(u->slice)) {
+ r = unit_realize_cgroup_now(UNIT_DEREF(u->slice), state);
+ if (r < 0)
+ return r;
+ }
+
+ /* And then do the real work */
+ r = unit_create_cgroup(u, target_mask, enable_mask);
+ if (r < 0)
+ return r;
+
+ /* Finally, apply the necessary attributes. */
+ cgroup_context_apply(u, target_mask, state);
+ cgroup_xattr_apply(u);
+
+ return 0;
+}
+
+static void unit_add_to_cgroup_queue(Unit *u) {
+
+ if (u->in_cgroup_queue)
+ return;
+
+ LIST_PREPEND(cgroup_queue, u->manager->cgroup_queue, u);
+ u->in_cgroup_queue = true;
+}
+
+unsigned manager_dispatch_cgroup_queue(Manager *m) {
+ ManagerState state;
+ unsigned n = 0;
+ Unit *i;
+ int r;
+
+ state = manager_state(m);
+
+ while ((i = m->cgroup_queue)) {
+ assert(i->in_cgroup_queue);
+
+ r = unit_realize_cgroup_now(i, state);
+ if (r < 0)
+ log_warning_errno(r, "Failed to realize cgroups for queued unit %s, ignoring: %m", i->id);
+
+ n++;
+ }
+
+ return n;
+}
+
+static void unit_queue_siblings(Unit *u) {
+ Unit *slice;
+
+ /* This adds the siblings of the specified unit and the
+ * siblings of all parent units to the cgroup queue. (But
+ * neither the specified unit itself nor the parents.) */
+
+ while ((slice = UNIT_DEREF(u->slice))) {
+ Iterator i;
+ Unit *m;
+
+ SET_FOREACH(m, slice->dependencies[UNIT_BEFORE], i) {
+ if (m == u)
+ continue;
+
+ /* Skip units that have a dependency on the slice
+ * but aren't actually in it. */
+ if (UNIT_DEREF(m->slice) != slice)
+ continue;
+
+ /* No point in doing cgroup application for units
+ * without active processes. */
+ if (UNIT_IS_INACTIVE_OR_FAILED(unit_active_state(m)))
+ continue;
+
+ /* If the unit doesn't need any new controllers
+ * and has current ones realized, it doesn't need
+ * any changes. */
+ if (unit_has_mask_realized(m, unit_get_target_mask(m), unit_get_enable_mask(m)))
+ continue;
+
+ unit_add_to_cgroup_queue(m);
+ }
+
+ u = slice;
+ }
+}
+
+int unit_realize_cgroup(Unit *u) {
+ assert(u);
+
+ if (!UNIT_HAS_CGROUP_CONTEXT(u))
+ return 0;
+
+ /* So, here's the deal: when realizing the cgroups for this
+ * unit, we need to first create all parents, but there's more
+ * actually: for the weight-based controllers we also need to
+ * make sure that all our siblings (i.e. units that are in the
+ * same slice as we are) have cgroups, too. Otherwise, things
+ * would become very uneven as each of their processes would
+ * get as much resources as all our group together. This call
+ * will synchronously create the parent cgroups, but will
+ * defer work on the siblings to the next event loop
+ * iteration. */
+
+ /* Add all sibling slices to the cgroup queue. */
+ unit_queue_siblings(u);
+
+ /* And realize this one now (and apply the values) */
+ return unit_realize_cgroup_now(u, manager_state(u->manager));
+}
+
+void unit_release_cgroup(Unit *u) {
+ assert(u);
+
+ /* Forgets all cgroup details for this cgroup */
+
+ if (u->cgroup_path) {
+ (void) hashmap_remove(u->manager->cgroup_unit, u->cgroup_path);
+ u->cgroup_path = mfree(u->cgroup_path);
+ }
+
+ if (u->cgroup_inotify_wd >= 0) {
+ if (inotify_rm_watch(u->manager->cgroup_inotify_fd, u->cgroup_inotify_wd) < 0)
+ log_unit_debug_errno(u, errno, "Failed to remove cgroup inotify watch %i for %s, ignoring", u->cgroup_inotify_wd, u->id);
+
+ (void) hashmap_remove(u->manager->cgroup_inotify_wd_unit, INT_TO_PTR(u->cgroup_inotify_wd));
+ u->cgroup_inotify_wd = -1;
+ }
+}
+
+void unit_prune_cgroup(Unit *u) {
+ int r;
+ bool is_root_slice;
+
+ assert(u);
+
+ /* Removes the cgroup, if empty and possible, and stops watching it. */
+
+ if (!u->cgroup_path)
+ return;
+
+ (void) unit_get_cpu_usage(u, NULL); /* Cache the last CPU usage value before we destroy the cgroup */
+
+ is_root_slice = unit_has_name(u, SPECIAL_ROOT_SLICE);
+
+ r = cg_trim_everywhere(u->manager->cgroup_supported, u->cgroup_path, !is_root_slice);
+ if (r < 0) {
+ log_unit_debug_errno(u, r, "Failed to destroy cgroup %s, ignoring: %m", u->cgroup_path);
+ return;
+ }
+
+ if (is_root_slice)
+ return;
+
+ unit_release_cgroup(u);
+
+ u->cgroup_realized = false;
+ u->cgroup_realized_mask = 0;
+ u->cgroup_enabled_mask = 0;
+}
+
+int unit_search_main_pid(Unit *u, pid_t *ret) {
+ _cleanup_fclose_ FILE *f = NULL;
+ pid_t pid = 0, npid, mypid;
+ int r;
+
+ assert(u);
+ assert(ret);
+
+ if (!u->cgroup_path)
+ return -ENXIO;
+
+ r = cg_enumerate_processes(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, &f);
+ if (r < 0)
+ return r;
+
+ mypid = getpid();
+ while (cg_read_pid(f, &npid) > 0) {
+ pid_t ppid;
+
+ if (npid == pid)
+ continue;
+
+ /* Ignore processes that aren't our kids */
+ if (get_process_ppid(npid, &ppid) >= 0 && ppid != mypid)
+ continue;
+
+ if (pid != 0)