From 099619957a0d9ae0e6e5e221493e40fb8fc2cd81 Mon Sep 17 00:00:00 2001 From: Alban Crequy Date: Mon, 7 Dec 2015 01:10:50 +0100 Subject: cgroup2: use new fstype for unified hierarchy Since Linux v4.4-rc1, __DEVEL__sane_behavior does not exist anymore and is replaced by a new fstype "cgroup2". With this patch, systemd no longer supports the old (unstable) way of doing unified hierarchy with __DEVEL__sane_behavior and systemd now requires Linux v4.4 for unified hierarchy. Non-unified hierarchy is still the default and is unchanged by this patch. https://github.com/torvalds/linux/commit/67e9c74b8a873408c27ac9a8e4c1d1c8d72c93ff --- src/core/mount-setup.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/core') 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 }, -- cgit v1.2.3-54-g00ecf From ab2c3861dcbece00d1c03357556faced6e048de0 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 25 Mar 2016 11:38:50 -0400 Subject: core: update populated event handling in unified hierarchy Earlier during the development of unified hierarchy, the populated event was reported through by the dedicated "cgroup.populated" file; however, the interface was updated so that it's reported through the "populated" field of "cgroup.events" file. Update populated event handling logic accordingly. --- src/basic/cgroup-util.c | 45 ++++++++++++++++++++++++++++++++++++--------- src/basic/cgroup-util.h | 2 ++ src/core/cgroup.c | 6 +++--- src/nspawn/nspawn-cgroup.c | 3 +-- 4 files changed, 42 insertions(+), 14 deletions(-) (limited to 'src/core') diff --git a/src/basic/cgroup-util.c b/src/basic/cgroup-util.c index 5124b5bf93..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; 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/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/nspawn/nspawn-cgroup.c b/src/nspawn/nspawn-cgroup.c index 53a7ee7134..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); -- cgit v1.2.3-54-g00ecf From e57051f542080a4ec1d9e4e1c428b39defe9f098 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Fri, 25 Mar 2016 11:38:50 -0400 Subject: core: update invoke_sigchld_event() to handle NULL ->sigchld_event() After receiving SIGCHLD, one of the ways manager_dispatch_sigchld() maps the now zombie $PID to its unit is through manager_get_unit_by_pid_cgroup() which reads /proc/$PID/cgroup and looks up the unit associated with the cgroup path. On non-unified cgroup hierarchies, a process is immediately migrated to the root cgroup on death and the cgroup lookup would always have returned the unit associated with it, making it rather pointless but safe. On unified hierarchy, a zombie remains associated with the cgroup that it was associated with at the time of death and thus manager_get_unit_by_pid_cgroup() will look up the unit properly. However, by the time manager_dispatch_sigchld() is running, the original cgroup may have become empty and it and its associated unit might already have been removed. If the cgroup path doesn't yield a match, manager_dispatch_sigchld() keeps pruning the leaf component. This means that the function may return a slice unit for a pid and as a slice doesn't have ->sigchld_event() handler, calling invoke_sigchld_event() on it causes a segfault. This patch updates invoke_sigchld_event() so that it skips calling if the handler is not set. --- src/core/manager.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src/core') diff --git a/src/core/manager.c b/src/core/manager.c index f13e933578..78f2a21061 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) { -- cgit v1.2.3-54-g00ecf