Merge pull request #13512 from msekletar/freezer

core: introduce support for cgroup freezer

NEWS

@@ -10,6 +10,10 @@ CHANGES WITH 246 in spe:
           their first output column with --no-legend. To hide the first column,
           use --plain.
 
+        * The service manager gained basic support for cgroup v2 freezer. Units
+          can now be suspended or resumed either using new systemctl verbs,
+          freeze and thaw respectively, or via D-Bus.
+
 CHANGES WITH 245:
 
         * A new tool "systemd-repart" has been added, that operates as an

man/systemctl.xml

@@ -303,6 +303,30 @@ Sun 2017-02-26 20:57:49 EST  2h 3min left  Sun 2017-02-26 11:56:36 EST  6h ago
             generally redundant and reproducible on the next invocation of the unit).</para>
           </listitem>
         </varlistentry>
+        <varlistentry>
+          <term><command>freeze <replaceable>PATTERN</replaceable>…</command></term>
+
+          <listitem>
+            <para>Freeze one or more units specified on the
+            command line using cgroup freezer</para>
+
+            <para>Freezing the unit will cause all processes contained within the cgroup corresponding to the unit
+            to be suspended. Being suspended means that unit's processes won't be scheduled to run on CPU until thawed.
+            Note that this command is supported only on systems that use unified cgroup hierarchy. Unit is automatically
+            thawed just before we execute a job against the unit, e.g. before the unit is stopped.</para>
+          </listitem>
+        </varlistentry>
+        <varlistentry>
+          <term><command>thaw <replaceable>PATTERN</replaceable>…</command></term>
+
+          <listitem>
+            <para>Thaw (unfreeze) one or more units specified on the
+            command line.</para>
+
+            <para>This is the inverse operation to the <command>freeze</command> command and resumes the execution of
+            processes in the unit's cgroup.</para>
+          </listitem>
+        </varlistentry>
         <varlistentry>
           <term><command>is-active <replaceable>PATTERN</replaceable>…</command></term>
 

src/basic/cgroup-util.c

@@ -149,6 +149,17 @@ bool cg_ns_supported(void) {
         return enabled;
 }
 
+bool cg_freezer_supported(void) {
+        static thread_local int supported = -1;
+
+        if (supported >= 0)
+                return supported;
+
+        supported = cg_all_unified() > 0 && access("/sys/fs/cgroup/init.scope/cgroup.freeze", F_OK) == 0;
+
+        return supported;
+}
+
 int cg_enumerate_subgroups(const char *controller, const char *path, DIR **_d) {
         _cleanup_free_ char *fs = NULL;
         int r;
@@ -1684,12 +1695,13 @@ int cg_get_attribute_as_uint64(const char *controller, const char *path, const c
         return 0;
 }
 
-int cg_get_keyed_attribute(
+int cg_get_keyed_attribute_full(
                 const char *controller,
                 const char *path,
                 const char *attribute,
                 char **keys,
-                char **ret_values) {
+                char **ret_values,
+                CGroupKeyMode mode) {
 
         _cleanup_free_ char *filename = NULL, *contents = NULL;
         const char *p;
@@ -1701,7 +1713,8 @@ int cg_get_keyed_attribute(
          * all keys to retrieve. The 'ret_values' parameter should be passed as string size with the same number of
          * entries as 'keys'. On success each entry will be set to the value of the matching key.
          *
-         * If the attribute file doesn't exist at all returns ENOENT, if any key is not found returns ENXIO. */
+         * If the attribute file doesn't exist at all returns ENOENT, if any key is not found returns ENXIO. If mode
+         * is set to GG_KEY_MODE_GRACEFUL we ignore missing keys and return those that were parsed successfully. */
 
         r = cg_get_path(controller, path, attribute, &filename);
         if (r < 0)
@@ -1749,6 +1762,9 @@ int cg_get_keyed_attribute(
                 p += strspn(p, NEWLINE);
         }
 
+        if (mode & CG_KEY_MODE_GRACEFUL)
+                goto done;
+
         r = -ENXIO;
 
 fail:
@@ -1759,6 +1775,9 @@ int cg_get_keyed_attribute(
 
 done:
         memcpy(ret_values, v, sizeof(char*) * n);
+        if (mode & CG_KEY_MODE_GRACEFUL)
+                return n_done;
+
         return 0;
 }
 

src/basic/cgroup-util.h

@@ -180,9 +180,31 @@ int cg_pid_get_path(const char *controller, pid_t pid, char **path);
 
 int cg_rmdir(const char *controller, const char *path);
 
+typedef enum  {
+        CG_KEY_MODE_GRACEFUL = 1 << 0,
+} CGroupKeyMode;
+
 int cg_set_attribute(const char *controller, const char *path, const char *attribute, const char *value);
 int cg_get_attribute(const char *controller, const char *path, const char *attribute, char **ret);
-int cg_get_keyed_attribute(const char *controller, const char *path, const char *attribute, char **keys, char **values);
+int cg_get_keyed_attribute_full(const char *controller, const char *path, const char *attribute, char **keys, char **values, CGroupKeyMode mode);
+
+static inline int cg_get_keyed_attribute(
+                const char *controller,
+                const char *path,
+                const char *attribute,
+                char **keys,
+                char **ret_values) {
+        return cg_get_keyed_attribute_full(controller, path, attribute, keys, ret_values, 0);
+}
+
+static inline int cg_get_keyed_attribute_graceful(
+                const char *controller,
+                const char *path,
+                const char *attribute,
+                char **keys,
+                char **ret_values) {
+        return cg_get_keyed_attribute_full(controller, path, attribute, keys, ret_values, CG_KEY_MODE_GRACEFUL);
+}
 
 int cg_get_attribute_as_uint64(const char *controller, const char *path, const char *attribute, uint64_t *ret);
 
@@ -238,6 +260,7 @@ int cg_mask_to_string(CGroupMask mask, char **ret);
 int cg_kernel_controllers(Set **controllers);
 
 bool cg_ns_supported(void);
+bool cg_freezer_supported(void);
 
 int cg_all_unified(void);
 int cg_hybrid_unified(void);

src/basic/unit-def.c

@@ -108,6 +108,15 @@ static const char* const unit_active_state_table[_UNIT_ACTIVE_STATE_MAX] = {
 
 DEFINE_STRING_TABLE_LOOKUP(unit_active_state, UnitActiveState);
 
+static const char* const freezer_state_table[_FREEZER_STATE_MAX] = {
+        [FREEZER_RUNNING] = "running",
+        [FREEZER_FREEZING] = "freezing",
+        [FREEZER_FROZEN] = "frozen",
+        [FREEZER_THAWING] = "thawing",
+};
+
+DEFINE_STRING_TABLE_LOOKUP(freezer_state, FreezerState);
+
 static const char* const automount_state_table[_AUTOMOUNT_STATE_MAX] = {
         [AUTOMOUNT_DEAD] = "dead",
         [AUTOMOUNT_WAITING] = "waiting",

src/basic/unit-def.h

@@ -48,6 +48,15 @@ typedef enum UnitActiveState {
         _UNIT_ACTIVE_STATE_INVALID = -1
 } UnitActiveState;
 
+typedef enum FreezerState {
+        FREEZER_RUNNING,
+        FREEZER_FREEZING,
+        FREEZER_FROZEN,
+        FREEZER_THAWING,
+        _FREEZER_STATE_MAX,
+        _FREEZER_STATE_INVALID = -1
+} FreezerState;
+
 typedef enum AutomountState {
         AUTOMOUNT_DEAD,
         AUTOMOUNT_WAITING,
@@ -253,6 +262,9 @@ UnitLoadState unit_load_state_from_string(const char *s) _pure_;
 const char *unit_active_state_to_string(UnitActiveState i) _const_;
 UnitActiveState unit_active_state_from_string(const char *s) _pure_;
 
+const char *freezer_state_to_string(FreezerState i) _const_;
+FreezerState freezer_state_from_string(const char *s) _pure_;
+
 const char* automount_state_to_string(AutomountState i) _const_;
 AutomountState automount_state_from_string(const char *s) _pure_;
 

src/core/cgroup.c

@@ -16,7 +16,9 @@
 #include "fd-util.h"
 #include "fileio.h"
 #include "fs-util.h"
+#include "io-util.h"
 #include "limits-util.h"
+#include "nulstr-util.h"
 #include "parse-util.h"
 #include "path-util.h"
 #include "process-util.h"
@@ -2661,6 +2663,16 @@ void unit_add_to_cgroup_empty_queue(Unit *u) {
                 log_debug_errno(r, "Failed to enable cgroup empty event source: %m");
 }
 
+static void unit_remove_from_cgroup_empty_queue(Unit *u) {
+        assert(u);
+
+        if (!u->in_cgroup_empty_queue)
+                return;
+
+        LIST_REMOVE(cgroup_empty_queue, u->manager->cgroup_empty_queue, u);
+        u->in_cgroup_empty_queue = false;
+}
+
 int unit_check_oom(Unit *u) {
         _cleanup_free_ char *oom_kill = NULL;
         bool increased;
@@ -2761,6 +2773,41 @@ static void unit_add_to_cgroup_oom_queue(Unit *u) {
                 log_error_errno(r, "Failed to enable cgroup oom event source: %m");
 }
 
+static int unit_check_cgroup_events(Unit *u) {
+        char *values[2] = {};
+        int r;
+
+        assert(u);
+
+        r = cg_get_keyed_attribute_graceful(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, "cgroup.events",
+                                            STRV_MAKE("populated", "frozen"), values);
+        if (r < 0)
+                return r;
+
+        /* The cgroup.events notifications can be merged together so act as we saw the given state for the
+         * first time. The functions we call to handle given state are idempotent, which makes them
+         * effectively remember the previous state. */
+        if (values[0]) {
+                if (streq(values[0], "1"))
+                        unit_remove_from_cgroup_empty_queue(u);
+                else
+                        unit_add_to_cgroup_empty_queue(u);
+        }
+
+        /* Disregard freezer state changes due to operations not initiated by us */
+        if (values[1] && IN_SET(u->freezer_state, FREEZER_FREEZING, FREEZER_THAWING)) {
+                if (streq(values[1], "0"))
+                        unit_thawed(u);
+                else
+                        unit_frozen(u);
+        }
+
+        free(values[0]);
+        free(values[1]);
+
+        return 0;
+}
+
 static int on_cgroup_inotify_event(sd_event_source *s, int fd, uint32_t revents, void *userdata) {
         Manager *m = userdata;
 
@@ -2797,7 +2844,7 @@ static int on_cgroup_inotify_event(sd_event_source *s, int fd, uint32_t revents,
 
                         u = hashmap_get(m->cgroup_control_inotify_wd_unit, INT_TO_PTR(e->wd));
                         if (u)
-                                unit_add_to_cgroup_empty_queue(u);
+                                unit_check_cgroup_events(u);
 
                         u = hashmap_get(m->cgroup_memory_inotify_wd_unit, INT_TO_PTR(e->wd));
                         if (u)
@@ -3550,6 +3597,46 @@ int compare_job_priority(const void *a, const void *b) {
         return strcmp(x->unit->id, y->unit->id);
 }
 
+int unit_cgroup_freezer_action(Unit *u, FreezerAction action) {
+        _cleanup_free_ char *path = NULL;
+        FreezerState target, kernel = _FREEZER_STATE_INVALID;
+        int r;
+
+        assert(u);
+        assert(IN_SET(action, FREEZER_FREEZE, FREEZER_THAW));
+
+        if (!u->cgroup_realized)
+                return -EBUSY;
+
+        target = action == FREEZER_FREEZE ? FREEZER_FROZEN : FREEZER_RUNNING;
+
+        r = unit_freezer_state_kernel(u, &kernel);
+        if (r < 0)
+                log_unit_debug_errno(u, r, "Failed to obtain cgroup freezer state: %m");
+
+        if (target == kernel) {
+                u->freezer_state = target;
+                return 0;
+        }
+
+        r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, "cgroup.freeze", &path);
+        if (r < 0)
+                return r;
+
+        log_unit_debug(u, "%s unit.", action == FREEZER_FREEZE ? "Freezing" : "Thawing");
+
+        if (action == FREEZER_FREEZE)
+                u->freezer_state = FREEZER_FREEZING;
+        else
+                u->freezer_state = FREEZER_THAWING;
+
+        r = write_string_file(path, one_zero(action == FREEZER_FREEZE), WRITE_STRING_FILE_DISABLE_BUFFER);
+        if (r < 0)
+                return r;
+
+        return 0;
+}
+
 static const char* const cgroup_device_policy_table[_CGROUP_DEVICE_POLICY_MAX] = {
         [CGROUP_DEVICE_POLICY_AUTO]   = "auto",
         [CGROUP_DEVICE_POLICY_CLOSED] = "closed",
@@ -3585,3 +3672,10 @@ int unit_get_cpuset(Unit *u, CPUSet *cpus, const char *name) {
 }
 
 DEFINE_STRING_TABLE_LOOKUP(cgroup_device_policy, CGroupDevicePolicy);
+
+static const char* const freezer_action_table[_FREEZER_ACTION_MAX] = {
+        [FREEZER_FREEZE] = "freeze",
+        [FREEZER_THAW] = "thaw",
+};
+
+DEFINE_STRING_TABLE_LOOKUP(freezer_action, FreezerAction);

src/core/cgroup.h

@@ -47,6 +47,14 @@ typedef enum CGroupDevicePolicy {
         _CGROUP_DEVICE_POLICY_INVALID = -1
 } CGroupDevicePolicy;
 
+typedef enum FreezerAction {
+        FREEZER_FREEZE,
+        FREEZER_THAW,
+
+        _FREEZER_ACTION_MAX,
+        _FREEZER_ACTION_INVALID = -1,
+} FreezerAction;
+
 struct CGroupDeviceAllow {
         LIST_FIELDS(CGroupDeviceAllow, device_allow);
         char *path;
@@ -274,3 +282,7 @@ bool unit_cgroup_delegate(Unit *u);
 int compare_job_priority(const void *a, const void *b);
 
 int unit_get_cpuset(Unit *u, CPUSet *cpus, const char *name);
+int unit_cgroup_freezer_action(Unit *u, FreezerAction action);
+
+const char* freezer_action_to_string(FreezerAction a) _const_;
+FreezerAction freezer_action_from_string(const char *s) _pure_;

src/core/dbus-manager.c

@@ -620,6 +620,14 @@ static int method_clean_unit(sd_bus_message *message, void *userdata, sd_bus_err
         return method_generic_unit_operation(message, userdata, error, bus_unit_method_clean, GENERIC_UNIT_LOAD|GENERIC_UNIT_VALIDATE_LOADED);
 }
 
+static int method_freeze_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) {
+        return method_generic_unit_operation(message, userdata, error, bus_unit_method_freeze, 0);
+}
+
+static int method_thaw_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) {
+        return method_generic_unit_operation(message, userdata, error, bus_unit_method_thaw, 0);
+}
+
 static int method_reset_failed_unit(sd_bus_message *message, void *userdata, sd_bus_error *error) {
         /* Don't load the unit (because unloaded units can't be in failed state), and don't insist on the
          * unit to be loaded properly (since a failed unit might have its unit file disappeared) */
@@ -2584,6 +2592,18 @@ const sd_bus_vtable bus_manager_vtable[] = {
                                  NULL,,
                                  method_clean_unit,
                                  SD_BUS_VTABLE_UNPRIVILEGED),
+        SD_BUS_METHOD_WITH_NAMES("FreezeUnit",
+                                 "s",
+                                 SD_BUS_PARAM(name),
+                                 NULL,,
+                                 method_freeze_unit,
+                                 SD_BUS_VTABLE_UNPRIVILEGED),
+        SD_BUS_METHOD_WITH_NAMES("ThawUnit",
+                                 "s",
+                                 SD_BUS_PARAM(name),
+                                 NULL,,
+                                 method_thaw_unit,
+                                 SD_BUS_VTABLE_UNPRIVILEGED),
         SD_BUS_METHOD_WITH_NAMES("ResetFailedUnit",
                                  "s",
                                  SD_BUS_PARAM(name),