Merge pull request #2903 from keszybz/cgroup2-v3

core: cgroup2 support

8 files changed, 53 insertions, 19 deletions

diff --git a/src/basic/cgroup-util.c b/src/basic/cgroup-util.c
index 56c1fcaab9..5043180747 100644
--- a/src/basic/cgroup-util.c
+++ b/src/basic/cgroup-util.c
@@ -101,6 +101,39 @@ int cg_read_pid(FILE *f, pid_t *_pid) {
         return 1;
 }
 
+int cg_read_event(const char *controller, const char *path, const char *event,
+                  char **val)
+{
+        _cleanup_free_ char *events = NULL, *content = NULL;
+        char *p, *line;
+        int r;
+
+        r = cg_get_path(controller, path, "cgroup.events", &events);
+        if (r < 0)
+                return r;
+
+        r = read_full_file(events, &content, NULL);
+        if (r < 0)
+                return r;
+
+        p = content;
+        while ((line = strsep(&p, "\n"))) {
+                char *key;
+
+                key = strsep(&line, " ");
+                if (!key || !line)
+                        return -EINVAL;
+
+                if (strcmp(key, event))
+                        continue;
+
+                *val = strdup(line);
+                return 0;
+        }
+
+        return -ENOENT;
+}
+
 int cg_enumerate_subgroups(const char *controller, const char *path, DIR **_d) {
         _cleanup_free_ char *fs = NULL;
         int r;
@@ -1007,18 +1040,12 @@ int cg_is_empty_recursive(const char *controller, const char *path) {
                 return unified;
 
         if (unified > 0) {
-                _cleanup_free_ char *populated = NULL, *t = NULL;
+                _cleanup_free_ char *t = NULL;
 
                 /* On the unified hierarchy we can check empty state
-                 * via the "cgroup.populated" attribute. */
+                 * via the "populated" attribute of "cgroup.events". */
 
-                r = cg_get_path(controller, path, "cgroup.populated", &populated);
-                if (r < 0)
-                        return r;
-
-                r = read_one_line_file(populated, &t);
-                if (r == -ENOENT)
-                        return 1;
+                r = cg_read_event(controller, path, "populated", &t);
                 if (r < 0)
                         return r;
 
@@ -2129,7 +2156,7 @@ int cg_unified(void) {
         if (statfs("/sys/fs/cgroup/", &fs) < 0)
                 return -errno;
 
-        if (F_TYPE_EQUAL(fs.f_type, CGROUP_SUPER_MAGIC))
+        if (F_TYPE_EQUAL(fs.f_type, CGROUP2_SUPER_MAGIC))
                 unified_cache = true;
         else if (F_TYPE_EQUAL(fs.f_type, TMPFS_MAGIC))
                 unified_cache = false;
diff --git a/src/basic/cgroup-util.h b/src/basic/cgroup-util.h
index ad1edd9cdb..4254e51e5d 100644
--- a/src/basic/cgroup-util.h
+++ b/src/basic/cgroup-util.h
@@ -96,6 +96,8 @@ static inline bool CGROUP_BLKIO_WEIGHT_IS_OK(uint64_t x) {
 
 int cg_enumerate_processes(const char *controller, const char *path, FILE **_f);
 int cg_read_pid(FILE *f, pid_t *_pid);
+int cg_read_event(const char *controller, const char *path, const char *event,
+                  char **val);
 
 int cg_enumerate_subgroups(const char *controller, const char *path, DIR **_d);
 int cg_read_subgroup(DIR *d, char **fn);
diff --git a/src/basic/missing.h b/src/basic/missing.h
index 034e334e66..66cd5921ad 100644
--- a/src/basic/missing.h
+++ b/src/basic/missing.h
@@ -437,6 +437,10 @@ struct btrfs_ioctl_quota_ctl_args {
 #define CGROUP_SUPER_MAGIC 0x27e0eb
 #endif
 
+#ifndef CGROUP2_SUPER_MAGIC
+#define CGROUP2_SUPER_MAGIC 0x63677270
+#endif
+
 #ifndef TMPFS_MAGIC
 #define TMPFS_MAGIC 0x01021994
 #endif
diff --git a/src/core/cgroup.c b/src/core/cgroup.c
index 39235a95f6..9c34928052 100644
--- a/src/core/cgroup.c
+++ b/src/core/cgroup.c
@@ -765,7 +765,7 @@ int unit_set_cgroup_path(Unit *u, const char *path) {
 }
 
 int unit_watch_cgroup(Unit *u) {
-        _cleanup_free_ char *populated = NULL;
+        _cleanup_free_ char *events = NULL;
         int r;
 
         assert(u);
@@ -791,11 +791,11 @@ int unit_watch_cgroup(Unit *u) {
         if (r < 0)
                 return log_oom();
 
-        r = cg_get_path(SYSTEMD_CGROUP_CONTROLLER, u->cgroup_path, "cgroup.populated", &populated);
+        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, populated, IN_MODIFY);
+        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
diff --git a/src/core/manager.c b/src/core/manager.c
index 26cb588e30..14ee68a3d2 100644
--- a/src/core/manager.c
+++ b/src/core/manager.c
@@ -1631,7 +1631,9 @@ static void invoke_sigchld_event(Manager *m, Unit *u, const siginfo_t *si) {
         log_unit_debug(u, "Child "PID_FMT" belongs to %s", si->si_pid, u->id);
 
         unit_unwatch_pid(u, si->si_pid);
-        UNIT_VTABLE(u)->sigchld_event(u, si->si_pid, si->si_code, si->si_status);
+
+        if (UNIT_VTABLE(u)->sigchld_event)
+                UNIT_VTABLE(u)->sigchld_event(u, si->si_pid, si->si_code, si->si_status);
 }
 
 static int manager_dispatch_sigchld(Manager *m) {
diff --git a/src/core/mount-setup.c b/src/core/mount-setup.c
index de1a361cc4..32fe51c67e 100644
--- a/src/core/mount-setup.c
+++ b/src/core/mount-setup.c
@@ -94,7 +94,7 @@ static const MountPoint mount_table[] = {
 #endif
         { "tmpfs",       "/run",                      "tmpfs",      "mode=755",                MS_NOSUID|MS_NODEV|MS_STRICTATIME,
           NULL,          MNT_FATAL|MNT_IN_CONTAINER },
-        { "cgroup",      "/sys/fs/cgroup",            "cgroup",     "__DEVEL__sane_behavior",  MS_NOSUID|MS_NOEXEC|MS_NODEV,
+        { "cgroup",      "/sys/fs/cgroup",            "cgroup2",    NULL,                      MS_NOSUID|MS_NOEXEC|MS_NODEV,
           cg_is_unified_wanted, MNT_FATAL|MNT_IN_CONTAINER },
         { "tmpfs",       "/sys/fs/cgroup",            "tmpfs",      "mode=755",                MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
           cg_is_legacy_wanted, MNT_FATAL|MNT_IN_CONTAINER },
diff --git a/src/nspawn/nspawn-cgroup.c b/src/nspawn/nspawn-cgroup.c
index 9f9a4759d1..f50f1ad6c2 100644
--- a/src/nspawn/nspawn-cgroup.c
+++ b/src/nspawn/nspawn-cgroup.c
@@ -55,8 +55,7 @@ int chown_cgroup(pid_t pid, uid_t uid_shift) {
                        "cgroup.events",
                        "cgroup.clone_children",
                        "cgroup.controllers",
-                       "cgroup.subtree_control",
-                       "cgroup.populated")
+                       "cgroup.subtree_control")
                 if (fchownat(fd, fn, uid_shift, uid_shift, 0) < 0)
                         log_full_errno(errno == ENOENT ? LOG_DEBUG :  LOG_WARNING, errno,
                                        "Failed to chown() cgroup file %s, ignoring: %m", fn);
@@ -94,7 +93,7 @@ int sync_cgroup(pid_t pid, bool unified_requested) {
         if (unified)
                 r = mount("cgroup", tree, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, "none,name=systemd,xattr");
         else
-                r = mount("cgroup", tree, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, "__DEVEL__sane_behavior");
+                r = mount("cgroup", tree, "cgroup2", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL);
         if (r < 0) {
                 r = log_error_errno(errno, "Failed to mount unified hierarchy: %m");
                 goto finish;
diff --git a/src/nspawn/nspawn-mount.c b/src/nspawn/nspawn-mount.c
index 70cca15278..64cb6b3ce3 100644
--- a/src/nspawn/nspawn-mount.c
+++ b/src/nspawn/nspawn-mount.c
@@ -750,7 +750,7 @@ static int mount_unified_cgroups(const char *dest) {
                 return -EINVAL;
         }
 
-        if (mount("cgroup", p, "cgroup", MS_NOSUID|MS_NOEXEC|MS_NODEV, "__DEVEL__sane_behavior") < 0)
+        if (mount("cgroup", p, "cgroup2", MS_NOSUID|MS_NOEXEC|MS_NODEV, NULL) < 0)
                 return log_error_errno(errno, "Failed to mount unified cgroup hierarchy to %s: %m", p);
 
         return 0;