Task Control Groups: shared cgroup subsystem group arrays

Replace the struct css_set embedded in task_struct with a pointer; all tasks
that have the same set of memberships across all hierarchies will share a
css_set object, and will be linked via their css_sets field to the "tasks"
list_head in the css_set.

Assuming that many tasks share the same cgroup assignments, this reduces
overall space usage and keeps the size of the task_struct down (three pointers
added to task_struct compared to a non-cgroups kernel, no matter how many
subsystems are registered).

[akpm@linux-foundation.org: fix a printk]
[akpm@linux-foundation.org: build fix]
Signed-off-by: Paul Menage <menage@google.com>
Cc: Serge E. Hallyn <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Serge E. Hallyn <serue@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Kirill Korotaev <dev@openvz.org>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation/cgroups.txt

@@ -176,7 +176,9 @@ Control Groups extends the kernel as follows:
    subsystem state is something that's expected to happen frequently
    and in performance-critical code, whereas operations that require a
    task's actual cgroup assignments (in particular, moving between
-   cgroups) are less common.
+   cgroups) are less common. A linked list runs through the cg_list
+   field of each task_struct using the css_set, anchored at
+   css_set->tasks.
 
  - A cgroup hierarchy filesystem can be mounted  for browsing and
    manipulation from user space.
@@ -252,6 +254,16 @@ linear search to locate an appropriate existing css_set, so isn't
 very efficient. A future version will use a hash table for better
 performance.
 
+To allow access from a cgroup to the css_sets (and hence tasks)
+that comprise it, a set of cg_cgroup_link objects form a lattice;
+each cg_cgroup_link is linked into a list of cg_cgroup_links for
+a single cgroup on its cont_link_list field, and a list of
+cg_cgroup_links for a single css_set on its cg_link_list.
+
+Thus the set of tasks in a cgroup can be listed by iterating over
+each css_set that references the cgroup, and sub-iterating over
+each css_set's task set.
+
 The use of a Linux virtual file system (vfs) to represent the
 cgroup hierarchy provides for a familiar permission and name space
 for cgroups, with a minimum of additional kernel code.

include/linux/cgroup.h

@@ -27,10 +27,19 @@ extern void cgroup_lock(void);
 extern void cgroup_unlock(void);
 extern void cgroup_fork(struct task_struct *p);
 extern void cgroup_fork_callbacks(struct task_struct *p);
+extern void cgroup_post_fork(struct task_struct *p);
 extern void cgroup_exit(struct task_struct *p, int run_callbacks);
 
 extern struct file_operations proc_cgroup_operations;
 
+/* Define the enumeration of all cgroup subsystems */
+#define SUBSYS(_x) _x ## _subsys_id,
+enum cgroup_subsys_id {
+#include <linux/cgroup_subsys.h>
+	CGROUP_SUBSYS_COUNT
+};
+#undef SUBSYS
+
 /* Per-subsystem/per-cgroup state maintained by the system. */
 struct cgroup_subsys_state {
 	/* The cgroup that this subsystem is attached to. Useful
@@ -97,6 +106,52 @@ struct cgroup {
 
 	struct cgroupfs_root *root;
 	struct cgroup *top_cgroup;
+
+	/*
+	 * List of cg_cgroup_links pointing at css_sets with
+	 * tasks in this cgroup. Protected by css_set_lock
+	 */
+	struct list_head css_sets;
+};
+
+/* A css_set is a structure holding pointers to a set of
+ * cgroup_subsys_state objects. This saves space in the task struct
+ * object and speeds up fork()/exit(), since a single inc/dec and a
+ * list_add()/del() can bump the reference count on the entire
+ * cgroup set for a task.
+ */
+
+struct css_set {
+
+	/* Reference count */
+	struct kref ref;
+
+	/*
+	 * List running through all cgroup groups. Protected by
+	 * css_set_lock
+	 */
+	struct list_head list;
+
+	/*
+	 * List running through all tasks using this cgroup
+	 * group. Protected by css_set_lock
+	 */
+	struct list_head tasks;
+
+	/*
+	 * List of cg_cgroup_link objects on link chains from
+	 * cgroups referenced from this css_set. Protected by
+	 * css_set_lock
+	 */
+	struct list_head cg_links;
+
+	/*
+	 * Set of subsystem states, one for each subsystem. This array
+	 * is immutable after creation apart from the init_css_set
+	 * during subsystem registration (at boot time).
+	 */
+	struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
+
 };
 
 /* struct cftype:
@@ -157,15 +212,7 @@ int cgroup_is_removed(const struct cgroup *cont);
 
 int cgroup_path(const struct cgroup *cont, char *buf, int buflen);
 
-int __cgroup_task_count(const struct cgroup *cont);
-static inline int cgroup_task_count(const struct cgroup *cont)
-{
-	int task_count;
-	rcu_read_lock();
-	task_count = __cgroup_task_count(cont);
-	rcu_read_unlock();
-	return task_count;
-}
+int cgroup_task_count(const struct cgroup *cont);
 
 /* Return true if the cgroup is a descendant of the current cgroup */
 int cgroup_is_descendant(const struct cgroup *cont);
@@ -213,7 +260,7 @@ static inline struct cgroup_subsys_state *cgroup_subsys_state(
 static inline struct cgroup_subsys_state *task_subsys_state(
 	struct task_struct *task, int subsys_id)
 {
-	return rcu_dereference(task->cgroups.subsys[subsys_id]);
+	return rcu_dereference(task->cgroups->subsys[subsys_id]);
 }
 
 static inline struct cgroup* task_cgroup(struct task_struct *task,
@@ -226,6 +273,27 @@ int cgroup_path(const struct cgroup *cont, char *buf, int buflen);
 
 int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *ss);
 
+/* A cgroup_iter should be treated as an opaque object */
+struct cgroup_iter {
+	struct list_head *cg_link;
+	struct list_head *task;
+};
+
+/* To iterate across the tasks in a cgroup:
+ *
+ * 1) call cgroup_iter_start to intialize an iterator
+ *
+ * 2) call cgroup_iter_next() to retrieve member tasks until it
+ *    returns NULL or until you want to end the iteration
+ *
+ * 3) call cgroup_iter_end() to destroy the iterator.
+ */
+void cgroup_iter_start(struct cgroup *cont, struct cgroup_iter *it);
+struct task_struct *cgroup_iter_next(struct cgroup *cont,
+					struct cgroup_iter *it);
+void cgroup_iter_end(struct cgroup *cont, struct cgroup_iter *it);
+
+
 #else /* !CONFIG_CGROUPS */
 
 static inline int cgroup_init_early(void) { return 0; }
@@ -233,6 +301,7 @@ static inline int cgroup_init(void) { return 0; }
 static inline void cgroup_init_smp(void) {}
 static inline void cgroup_fork(struct task_struct *p) {}
 static inline void cgroup_fork_callbacks(struct task_struct *p) {}
+static inline void cgroup_post_fork(struct task_struct *p) {}
 static inline void cgroup_exit(struct task_struct *p, int callbacks) {}
 
 static inline void cgroup_lock(void) {}

include/linux/sched.h

@@ -894,34 +894,6 @@ struct sched_entity {
 #endif
 };
 
-#ifdef CONFIG_CGROUPS
-
-#define SUBSYS(_x) _x ## _subsys_id,
-enum cgroup_subsys_id {
-#include <linux/cgroup_subsys.h>
-	CGROUP_SUBSYS_COUNT
-};
-#undef SUBSYS
-
-/* A css_set is a structure holding pointers to a set of
- * cgroup_subsys_state objects.
- */
-
-struct css_set {
-
-	/* Set of subsystem states, one for each subsystem. NULL for
-	 * subsystems that aren't part of this hierarchy. These
-	 * pointers reduce the number of dereferences required to get
-	 * from a task to its state for a given cgroup, but result
-	 * in increased space usage if tasks are in wildly different
-	 * groupings across different hierarchies. This array is
-	 * immutable after creation */
-	struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT];
-
-};
-
-#endif /* CONFIG_CGROUPS */
-
 struct task_struct {
 	volatile long state;	/* -1 unrunnable, 0 runnable, >0 stopped */
 	void *stack;
@@ -1159,7 +1131,10 @@ struct task_struct {
 	int cpuset_mem_spread_rotor;
 #endif
 #ifdef CONFIG_CGROUPS
-	struct css_set cgroups;
+	/* Control Group info protected by css_set_lock */
+	struct css_set *cgroups;
+	/* cg_list protected by css_set_lock and tsk->alloc_lock */
+	struct list_head cg_list;
 #endif
 #ifdef CONFIG_FUTEX
 	struct robust_list_head __user *robust_list;

kernel/cgroup.c

@@ -36,6 +36,7 @@
 #include <linux/proc_fs.h>
 #include <linux/rcupdate.h>
 #include <linux/sched.h>
+#include <linux/backing-dev.h>
 #include <linux/seq_file.h>
 #include <linux/slab.h>
 #include <linux/magic.h>
@@ -95,6 +96,7 @@ static struct cgroupfs_root rootnode;
 /* The list of hierarchy roots */
 
 static LIST_HEAD(roots);
+static int root_count;
 
 /* dummytop is a shorthand for the dummy hierarchy's top cgroup */
 #define dummytop (&rootnode.top_cgroup)
@@ -133,12 +135,49 @@ list_for_each_entry(_ss, &_root->subsys_list, sibling)
 #define for_each_root(_root) \
 list_for_each_entry(_root, &roots, root_list)
 
-/* Each task_struct has an embedded css_set, so the get/put
- * operation simply takes a reference count on all the cgroups
- * referenced by subsystems in this css_set. This can end up
- * multiple-counting some cgroups, but that's OK - the ref-count is
- * just a busy/not-busy indicator; ensuring that we only count each
- * cgroup once would require taking a global lock to ensure that no
+/* Link structure for associating css_set objects with cgroups */
+struct cg_cgroup_link {
+	/*
+	 * List running through cg_cgroup_links associated with a
+	 * cgroup, anchored on cgroup->css_sets
+	 */
+	struct list_head cont_link_list;
+	/*
+	 * List running through cg_cgroup_links pointing at a
+	 * single css_set object, anchored on css_set->cg_links
+	 */
+	struct list_head cg_link_list;
+	struct css_set *cg;
+};
+
+/* The default css_set - used by init and its children prior to any
+ * hierarchies being mounted. It contains a pointer to the root state
+ * for each subsystem. Also used to anchor the list of css_sets. Not
+ * reference-counted, to improve performance when child cgroups
+ * haven't been created.
+ */
+
+static struct css_set init_css_set;
+static struct cg_cgroup_link init_css_set_link;
+
+/* css_set_lock protects the list of css_set objects, and the
+ * chain of tasks off each css_set.  Nests outside task->alloc_lock
+ * due to cgroup_iter_start() */
+static DEFINE_RWLOCK(css_set_lock);
+static int css_set_count;
+
+/* We don't maintain the lists running through each css_set to its
+ * task until after the first call to cgroup_iter_start(). This
+ * reduces the fork()/exit() overhead for people who have cgroups
+ * compiled into their kernel but not actually in use */
+static int use_task_css_set_links;
+
+/* When we create or destroy a css_set, the operation simply
+ * takes/releases a reference count on all the cgroups referenced
+ * by subsystems in this css_set. This can end up multiple-counting
+ * some cgroups, but that's OK - the ref-count is just a
+ * busy/not-busy indicator; ensuring that we only count each cgroup
+ * once would require taking a global lock to ensure that no
  * subsystems moved between hierarchies while we were doing so.
  *
  * Possible TODO: decide at boot time based on the number of
@@ -146,18 +185,230 @@ list_for_each_entry(_root, &roots, root_list)
  * it's better for performance to ref-count every subsystem, or to
  * take a global lock and only add one ref count to each hierarchy.
  */
-static void get_css_set(struct css_set *cg)
+
+/*
+ * unlink a css_set from the list and free it
+ */
+static void release_css_set(struct kref *k)
 {
+	struct css_set *cg = container_of(k, struct css_set, ref);
 	int i;
+
+	write_lock(&css_set_lock);
+	list_del(&cg->list);
+	css_set_count--;
+	while (!list_empty(&cg->cg_links)) {
+		struct cg_cgroup_link *link;
+		link = list_entry(cg->cg_links.next,
+				  struct cg_cgroup_link, cg_link_list);
+		list_del(&link->cg_link_list);
+		list_del(&link->cont_link_list);
+		kfree(link);
+	}
+	write_unlock(&css_set_lock);
 	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
-		atomic_inc(&cg->subsys[i]->cgroup->count);
+		atomic_dec(&cg->subsys[i]->cgroup->count);
+	kfree(cg);
 }
 
-static void put_css_set(struct css_set *cg)
+/*
+ * refcounted get/put for css_set objects
+ */
+static inline void get_css_set(struct css_set *cg)
+{
+	kref_get(&cg->ref);
+}
+
+static inline void put_css_set(struct css_set *cg)
+{
+	kref_put(&cg->ref, release_css_set);
+}
+
+/*
+ * find_existing_css_set() is a helper for
+ * find_css_set(), and checks to see whether an existing
+ * css_set is suitable. This currently walks a linked-list for
+ * simplicity; a later patch will use a hash table for better
+ * performance
+ *
+ * oldcg: the cgroup group that we're using before the cgroup
+ * transition
+ *
+ * cont: the cgroup that we're moving into
+ *
+ * template: location in which to build the desired set of subsystem
+ * state objects for the new cgroup group
+ */
+
+static struct css_set *find_existing_css_set(
+	struct css_set *oldcg,
+	struct cgroup *cont,
+	struct cgroup_subsys_state *template[])
 {
 	int i;
-	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
-		atomic_dec(&cg->subsys[i]->cgroup->count);
+	struct cgroupfs_root *root = cont->root;
+	struct list_head *l = &init_css_set.list;
+
+	/* Built the set of subsystem state objects that we want to
+	 * see in the new css_set */
+	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
+		if (root->subsys_bits & (1ull << i)) {
+			/* Subsystem is in this hierarchy. So we want
+			 * the subsystem state from the new
+			 * cgroup */
+			template[i] = cont->subsys[i];
+		} else {
+			/* Subsystem is not in this hierarchy, so we
+			 * don't want to change the subsystem state */
+			template[i] = oldcg->subsys[i];
+		}
+	}
+
+	/* Look through existing cgroup groups to find one to reuse */
+	do {
+		struct css_set *cg =
+			list_entry(l, struct css_set, list);
+
+		if (!memcmp(template, cg->subsys, sizeof(cg->subsys))) {
+			/* All subsystems matched */
+			return cg;
+		}
+		/* Try the next cgroup group */
+		l = l->next;
+	} while (l != &init_css_set.list);
+
+	/* No existing cgroup group matched */
+	return NULL;
+}
+
+/*
+ * allocate_cg_links() allocates "count" cg_cgroup_link structures
+ * and chains them on tmp through their cont_link_list fields. Returns 0 on
+ * success or a negative error
+ */
+
+static int allocate_cg_links(int count, struct list_head *tmp)
+{
+	struct cg_cgroup_link *link;
+	int i;
+	INIT_LIST_HEAD(tmp);
+	for (i = 0; i < count; i++) {
+		link = kmalloc(sizeof(*link), GFP_KERNEL);
+		if (!link) {
+			while (!list_empty(tmp)) {
+				link = list_entry(tmp->next,
+						  struct cg_cgroup_link,
+						  cont_link_list);
+				list_del(&link->cont_link_list);
+				kfree(link);
+			}
+			return -ENOMEM;
+		}
+		list_add(&link->cont_link_list, tmp);
+	}
+	return 0;
+}
+
+static void free_cg_links(struct list_head *tmp)
+{
+	while (!list_empty(tmp)) {
+		struct cg_cgroup_link *link;
+		link = list_entry(tmp->next,
+				  struct cg_cgroup_link,
+				  cont_link_list);
+		list_del(&link->cont_link_list);
+		kfree(link);
+	}
+}
+
+/*
+ * find_css_set() takes an existing cgroup group and a
+ * cgroup object, and returns a css_set object that's
+ * equivalent to the old group, but with the given cgroup
+ * substituted into the appropriate hierarchy. Must be called with
+ * cgroup_mutex held
+ */
+
+static struct css_set *find_css_set(
+	struct css_set *oldcg, struct cgroup *cont)
+{
+	struct css_set *res;
+	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT];
+	int i;
+
+	struct list_head tmp_cg_links;
+	struct cg_cgroup_link *link;
+
+	/* First see if we already have a cgroup group that matches
+	 * the desired set */
+	write_lock(&css_set_lock);
+	res = find_existing_css_set(oldcg, cont, template);
+	if (res)
+		get_css_set(res);
+	write_unlock(&css_set_lock);
+
+	if (res)
+		return res;
+
+	res = kmalloc(sizeof(*res), GFP_KERNEL);
+	if (!res)
+		return NULL;
+
+	/* Allocate all the cg_cgroup_link objects that we'll need */
+	if (allocate_cg_links(root_count, &tmp_cg_links) < 0) {
+		kfree(res);
+		return NULL;
+	}
+
+	kref_init(&res->ref);
+	INIT_LIST_HEAD(&res->cg_links);
+	INIT_LIST_HEAD(&res->tasks);
+
+	/* Copy the set of subsystem state objects generated in
+	 * find_existing_css_set() */
+	memcpy(res->subsys, template, sizeof(res->subsys));
+
+	write_lock(&css_set_lock);
+	/* Add reference counts and links from the new css_set. */
+	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
+		struct cgroup *cont = res->subsys[i]->cgroup;
+		struct cgroup_subsys *ss = subsys[i];
+		atomic_inc(&cont->count);
+		/*
+		 * We want to add a link once per cgroup, so we
+		 * only do it for the first subsystem in each
+		 * hierarchy
+		 */
+		if (ss->root->subsys_list.next == &ss->sibling) {
+			BUG_ON(list_empty(&tmp_cg_links));
+			link = list_entry(tmp_cg_links.next,
+					  struct cg_cgroup_link,
+					  cont_link_list);
+			list_del(&link->cont_link_list);
+			list_add(&link->cont_link_list, &cont->css_sets);
+			link->cg = res;
+			list_add(&link->cg_link_list, &res->cg_links);
+		}
+	}
+	if (list_empty(&rootnode.subsys_list)) {
+		link = list_entry(tmp_cg_links.next,
+				  struct cg_cgroup_link,
+				  cont_link_list);
+		list_del(&link->cont_link_list);
+		list_add(&link->cont_link_list, &dummytop->css_sets);
+		link->cg = res;
+		list_add(&link->cg_link_list, &res->cg_links);
+	}
+
+	BUG_ON(!list_empty(&tmp_cg_links));
+
+	/* Link this cgroup group into the list */
+	list_add(&res->list, &init_css_set.list);
+	css_set_count++;
+	INIT_LIST_HEAD(&res->tasks);
+	write_unlock(&css_set_lock);
+
+	return res;
 }
 
 /*
@@ -504,6 +755,7 @@ static void init_cgroup_root(struct cgroupfs_root *root)
 	cont->top_cgroup = cont;
 	INIT_LIST_HEAD(&cont->sibling);
 	INIT_LIST_HEAD(&cont->children);
+	INIT_LIST_HEAD(&cont->css_sets);
 }
 
 static int cgroup_test_super(struct super_block *sb, void *data)
@@ -573,6 +825,8 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
 	int ret = 0;
 	struct super_block *sb;
 	struct cgroupfs_root *root;
+	struct list_head tmp_cg_links, *l;
+	INIT_LIST_HEAD(&tmp_cg_links);
 
 	/* First find the desired set of subsystems */
 	ret = parse_cgroupfs_options(data, &opts);
@@ -602,18 +856,36 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
 	} else {
 		/* New superblock */
 		struct cgroup *cont = &root->top_cgroup;
+		struct inode *inode;
 
 		BUG_ON(sb->s_root != NULL);
 
 		ret = cgroup_get_rootdir(sb);
 		if (ret)
 			goto drop_new_super;
+		inode = sb->s_root->d_inode;
 
+		mutex_lock(&inode->i_mutex);
 		mutex_lock(&cgroup_mutex);
 
+		/*
+		 * We're accessing css_set_count without locking
+		 * css_set_lock here, but that's OK - it can only be
+		 * increased by someone holding cgroup_lock, and
+		 * that's us. The worst that can happen is that we
+		 * have some link structures left over
+		 */
+		ret = allocate_cg_links(css_set_count, &tmp_cg_links);
+		if (ret) {
+			mutex_unlock(&cgroup_mutex);
+			mutex_unlock(&inode->i_mutex);
+			goto drop_new_super;
+		}
+
 		ret = rebind_subsystems(root, root->subsys_bits);
 		if (ret == -EBUSY) {
 			mutex_unlock(&cgroup_mutex);
+			mutex_unlock(&inode->i_mutex);
 			goto drop_new_super;
 		}
 
@@ -621,24 +893,40 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
 		BUG_ON(ret);
 
 		list_add(&root->root_list, &roots);
+		root_count++;
 
 		sb->s_root->d_fsdata = &root->top_cgroup;
 		root->top_cgroup.dentry = sb->s_root;
 
+		/* Link the top cgroup in this hierarchy into all
+		 * the css_set objects */
+		write_lock(&css_set_lock);
+		l = &init_css_set.list;
+		do {
+			struct css_set *cg;
+			struct cg_cgroup_link *link;
+			cg = list_entry(l, struct css_set, list);
+			BUG_ON(list_empty(&tmp_cg_links));
+			link = list_entry(tmp_cg_links.next,
+					  struct cg_cgroup_link,
+					  cont_link_list);
+			list_del(&link->cont_link_list);
+			link->cg = cg;
+			list_add(&link->cont_link_list,
+				 &root->top_cgroup.css_sets);
+			list_add(&link->cg_link_list, &cg->cg_links);
+			l = l->next;
+		} while (l != &init_css_set.list);
+		write_unlock(&css_set_lock);
+
+		free_cg_links(&tmp_cg_links);
+
 		BUG_ON(!list_empty(&cont->sibling));
 		BUG_ON(!list_empty(&cont->children));
 		BUG_ON(root->number_of_cgroups != 1);
 
-		/*
-		 * I believe that it's safe to nest i_mutex inside
-		 * cgroup_mutex in this case, since no-one else can
-		 * be accessing this directory yet. But we still need
-		 * to teach lockdep that this is the case - currently
-		 * a cgroupfs remount triggers a lockdep warning
-		 */
-		mutex_lock(&cont->dentry->d_inode->i_mutex);
 		cgroup_populate_dir(cont);
-		mutex_unlock(&cont->dentry->d_inode->i_mutex);
+		mutex_unlock(&inode->i_mutex);
 		mutex_unlock(&cgroup_mutex);
 	}
 
@@ -647,6 +935,7 @@ static int cgroup_get_sb(struct file_system_type *fs_type,
  drop_new_super:
 	up_write(&sb->s_umount);
 	deactivate_super(sb);
+	free_cg_links(&tmp_cg_links);
 	return ret;
 }
 
@@ -668,8 +957,25 @@ static void cgroup_kill_sb(struct super_block *sb) {
 	/* Shouldn't be able to fail ... */
 	BUG_ON(ret);
 
-	if (!list_empty(&root->root_list))
+	/*
+	 * Release all the links from css_sets to this hierarchy's
+	 * root cgroup
+	 */
+	write_lock(&css_set_lock);
+	while (!list_empty(&cont->css_sets)) {
+		struct cg_cgroup_link *link;
+		link = list_entry(cont->css_sets.next,
+				  struct cg_cgroup_link, cont_link_list);
+		list_del(&link->cg_link_list);
+		list_del(&link->cont_link_list);
+		kfree(link);
+	}
+	write_unlock(&css_set_lock);
+
+	if (!list_empty(&root->root_list)) {
 		list_del(&root->root_list);
+		root_count--;
+	}
 	mutex_unlock(&cgroup_mutex);
 
 	kfree(root);
@@ -762,9 +1068,9 @@ static int attach_task(struct cgroup *cont, struct task_struct *tsk)
 	int retval = 0;
 	struct cgroup_subsys *ss;
 	struct cgroup *oldcont;
-	struct css_set *cg = &tsk->cgroups;
+	struct css_set *cg = tsk->cgroups;
+	struct css_set *newcg;
 	struct cgroupfs_root *root = cont->root;
-	int i;
 	int subsys_id;
 
 	get_first_subsys(cont, NULL, &subsys_id);
@@ -783,26 +1089,32 @@ static int attach_task(struct cgroup *cont, struct task_struct *tsk)
 		}
 	}
 
+	/*
+	 * Locate or allocate a new css_set for this task,
+	 * based on its final set of cgroups
+	 */
+	newcg = find_css_set(cg, cont);
+	if (!newcg) {
+		return -ENOMEM;
+	}
+
 	task_lock(tsk);
 	if (tsk->flags & PF_EXITING) {
 		task_unlock(tsk);
+		put_css_set(newcg);
 		return -ESRCH;
 	}
-	/* Update the css_set pointers for the subsystems in this
-	 * hierarchy */
-	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
-		if (root->subsys_bits & (1ull << i)) {
-			/* Subsystem is in this hierarchy. So we want
-			 * the subsystem state from the new
-			 * cgroup. Transfer the refcount from the
-			 * old to the new */
-			atomic_inc(&cont->count);
-			atomic_dec(&cg->subsys[i]->cgroup->count);
-			rcu_assign_pointer(cg->subsys[i], cont->subsys[i]);
-		}
-	}
+	rcu_assign_pointer(tsk->cgroups, newcg);
 	task_unlock(tsk);
 
+	/* Update the css_set linked lists if we're using them */
+	write_lock(&css_set_lock);
+	if (!list_empty(&tsk->cg_list)) {
+		list_del(&tsk->cg_list);
+		list_add(&tsk->cg_list, &newcg->tasks);
+	}
+	write_unlock(&css_set_lock);
+
 	for_each_subsys(root, ss) {
 		if (ss->attach) {
 			ss->attach(ss, cont, oldcont, tsk);
@@ -810,6 +1122,7 @@ static int attach_task(struct cgroup *cont, struct task_struct *tsk)
 	}
 
 	synchronize_rcu();
+	put_css_set(cg);
 	return 0;
 }
 
@@ -1069,7 +1382,7 @@ static int cgroup_create_file(struct dentry *dentry, int mode,
 
 		/* start with the directory inode held, so that we can
 		 * populate it without racing with another mkdir */
-		mutex_lock(&inode->i_mutex);
+		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
 	} else if (S_ISREG(mode)) {
 		inode->i_size = 0;
 		inode->i_fop = &cgroup_file_operations;
@@ -1148,27 +1461,101 @@ int cgroup_add_files(struct cgroup *cont,
 	return 0;
 }
 
-/* Count the number of tasks in a cgroup. Could be made more
- * time-efficient but less space-efficient with more linked lists
- * running through each cgroup and the css_set structures that
- * referenced it. Must be called with tasklist_lock held for read or
- * write or in an rcu critical section.
- */
-int __cgroup_task_count(const struct cgroup *cont)
+/* Count the number of tasks in a cgroup. */
+
+int cgroup_task_count(const struct cgroup *cont)
 {
 	int count = 0;
-	struct task_struct *g, *p;
-	struct cgroup_subsys_state *css;
-	int subsys_id;
-
-	get_first_subsys(cont, &css, &subsys_id);
-	do_each_thread(g, p) {
-		if (task_subsys_state(p, subsys_id) == css)
-			count ++;
-	} while_each_thread(g, p);
+	struct list_head *l;
+
+	read_lock(&css_set_lock);
+	l = cont->css_sets.next;
+	while (l != &cont->css_sets) {
+		struct cg_cgroup_link *link =
+			list_entry(l, struct cg_cgroup_link, cont_link_list);
+		count += atomic_read(&link->cg->ref.refcount);
+		l = l->next;
+	}
+	read_unlock(&css_set_lock);
 	return count;
 }
 
+/*
+ * Advance a list_head iterator.  The iterator should be positioned at
+ * the start of a css_set
+ */
+static void cgroup_advance_iter(struct cgroup *cont,
+					  struct cgroup_iter *it)
+{
+	struct list_head *l = it->cg_link;
+	struct cg_cgroup_link *link;
+	struct css_set *cg;
+
+	/* Advance to the next non-empty css_set */
+	do {
+		l = l->next;
+		if (l == &cont->css_sets) {
+			it->cg_link = NULL;
+			return;
+		}
+		link = list_entry(l, struct cg_cgroup_link, cont_link_list);
+		cg = link->cg;
+	} while (list_empty(&cg->tasks));
+	it->cg_link = l;
+	it->task = cg->tasks.next;
+}
+
+void cgroup_iter_start(struct cgroup *cont, struct cgroup_iter *it)
+{
+	/*
+	 * The first time anyone tries to iterate across a cgroup,
+	 * we need to enable the list linking each css_set to its
+	 * tasks, and fix up all existing tasks.
+	 */
+	if (!use_task_css_set_links) {
+		struct task_struct *p, *g;
+		write_lock(&css_set_lock);
+		use_task_css_set_links = 1;
+ 		do_each_thread(g, p) {
+			task_lock(p);
+			if (list_empty(&p->cg_list))
+				list_add(&p->cg_list, &p->cgroups->tasks);
+			task_unlock(p);
+ 		} while_each_thread(g, p);
+		write_unlock(&css_set_lock);
+	}
+	read_lock(&css_set_lock);
+	it->cg_link = &cont->css_sets;
+	cgroup_advance_iter(cont, it);
+}
+
+struct task_struct *cgroup_iter_next(struct cgroup *cont,
+					struct cgroup_iter *it)
+{
+	struct task_struct *res;
+	struct list_head *l = it->task;
+
+	/* If the iterator cg is NULL, we have no tasks */
+	if (!it->cg_link)
+		return NULL;
+	res = list_entry(l, struct task_struct, cg_list);
+	/* Advance iterator to find next entry */
+	l = l->next;
+	if (l == &res->cgroups->tasks) {
+		/* We reached the end of this task list - move on to
+		 * the next cg_cgroup_link */
+		cgroup_advance_iter(cont, it);
+	} else {
+		it->task = l;
+	}
+	return res;
+}
+
+void cgroup_iter_end(struct cgroup *cont, struct cgroup_iter *it)
+{
+	read_unlock(&css_set_lock);
+}
+
 /*
  * Stuff for reading the 'tasks' file.
  *
@@ -1198,22 +1585,15 @@ struct ctr_struct {
 static int pid_array_load(pid_t *pidarray, int npids, struct cgroup *cont)
 {
 	int n = 0;
-	struct task_struct *g, *p;
-	struct cgroup_subsys_state *css;
-	int subsys_id;
-
-	get_first_subsys(cont, &css, &subsys_id);
-	rcu_read_lock();
-	do_each_thread(g, p) {
-		if (task_subsys_state(p, subsys_id) == css) {
-			pidarray[n++] = pid_nr(task_pid(p));
-			if (unlikely(n == npids))
-				goto array_full;
-		}
-	} while_each_thread(g, p);
-
-array_full:
-	rcu_read_unlock();
+	struct cgroup_iter it;
+	struct task_struct *tsk;
+	cgroup_iter_start(cont, &it);
+	while ((tsk = cgroup_iter_next(cont, &it))) {
+		if (unlikely(n == npids))
+			break;
+		pidarray[n++] = pid_nr(task_pid(tsk));
+	}
+	cgroup_iter_end(cont, &it);
 	return n;
 }
 
@@ -1398,6 +1778,7 @@ static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
 	cont->flags = 0;
 	INIT_LIST_HEAD(&cont->sibling);
 	INIT_LIST_HEAD(&cont->children);
+	INIT_LIST_HEAD(&cont->css_sets);
 
 	cont->parent = parent;
 	cont->root = parent->root;
@@ -1529,8 +1910,8 @@ static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry)
 
 static void cgroup_init_subsys(struct cgroup_subsys *ss)
 {
-	struct task_struct *g, *p;
 	struct cgroup_subsys_state *css;
+	struct list_head *l;
 	printk(KERN_ERR "Initializing cgroup subsys %s\n", ss->name);
 
 	/* Create the top cgroup state for this subsystem */
@@ -1540,26 +1921,32 @@ static void cgroup_init_subsys(struct cgroup_subsys *ss)
 	BUG_ON(IS_ERR(css));
 	init_cgroup_css(css, ss, dummytop);
 
-	/* Update all tasks to contain a subsys pointer to this state
-	 * - since the subsystem is newly registered, all tasks are in
-	 * the subsystem's top cgroup. */
+	/* Update all cgroup groups to contain a subsys
+	 * pointer to this state - since the subsystem is
+	 * newly registered, all tasks and hence all cgroup
+	 * groups are in the subsystem's top cgroup. */
+	write_lock(&css_set_lock);
+	l = &init_css_set.list;
+	do {
+		struct css_set *cg =
+			list_entry(l, struct css_set, list);
+		cg->subsys[ss->subsys_id] = dummytop->subsys[ss->subsys_id];
+		l = l->next;
+	} while (l != &init_css_set.list);
+	write_unlock(&css_set_lock);
 
  	/* If this subsystem requested that it be notified with fork
  	 * events, we should send it one now for every process in the
  	 * system */
+ 	if (ss->fork) {
+ 		struct task_struct *g, *p;
 
-	read_lock(&tasklist_lock);
-	init_task.cgroups.subsys[ss->subsys_id] = css;
-	if (ss->fork)
-		ss->fork(ss, &init_task);
-
-	do_each_thread(g, p) {
-		printk(KERN_INFO "Setting task %p css to %p (%d)\n", css, p, p->pid);
-		p->cgroups.subsys[ss->subsys_id] = css;
-		if (ss->fork)
-			ss->fork(ss, p);
-	} while_each_thread(g, p);
-	read_unlock(&tasklist_lock);
+ 		read_lock(&tasklist_lock);
+ 		do_each_thread(g, p) {
+ 			ss->fork(ss, p);
+ 		} while_each_thread(g, p);
+ 		read_unlock(&tasklist_lock);
+ 	}
 
 	need_forkexit_callback |= ss->fork || ss->exit;
 
@@ -1573,8 +1960,22 @@ static void cgroup_init_subsys(struct cgroup_subsys *ss)
 int __init cgroup_init_early(void)
 {
 	int i;
+	kref_init(&init_css_set.ref);
+	kref_get(&init_css_set.ref);
+	INIT_LIST_HEAD(&init_css_set.list);
+	INIT_LIST_HEAD(&init_css_set.cg_links);
+	INIT_LIST_HEAD(&init_css_set.tasks);
+	css_set_count = 1;
 	init_cgroup_root(&rootnode);
 	list_add(&rootnode.root_list, &roots);
+	root_count = 1;
+	init_task.cgroups = &init_css_set;
+
+	init_css_set_link.cg = &init_css_set;
+	list_add(&init_css_set_link.cont_link_list,
+		 &rootnode.top_cgroup.css_sets);
+	list_add(&init_css_set_link.cg_link_list,
+		 &init_css_set.cg_links);
 
 	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
 		struct cgroup_subsys *ss = subsys[i];
@@ -1715,29 +2116,13 @@ static int proc_cgroupstats_show(struct seq_file *m, void *v)
 	int i;
 	struct cgroupfs_root *root;
 
+	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\n");
 	mutex_lock(&cgroup_mutex);
-	seq_puts(m, "Hierarchies:\n");
-	for_each_root(root) {
-		struct cgroup_subsys *ss;
-		int first = 1;
-		seq_printf(m, "%p: bits=%lx cgroups=%d (", root,
-			   root->subsys_bits, root->number_of_cgroups);
-		for_each_subsys(root, ss) {
-			seq_printf(m, "%s%s", first ? "" : ", ", ss->name);
-			first = false;
-		}
-		seq_putc(m, ')');
-		if (root->sb) {
-			seq_printf(m, " s_active=%d",
-				   atomic_read(&root->sb->s_active));
-		}
-		seq_putc(m, '\n');
-	}
-	seq_puts(m, "Subsystems:\n");
 	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
 		struct cgroup_subsys *ss = subsys[i];
-		seq_printf(m, "%d: name=%s hierarchy=%p\n",
-			   i, ss->name, ss->root);
+		seq_printf(m, "%s\t%lu\t%d\n",
+			   ss->name, ss->root->subsys_bits,
+			   ss->root->number_of_cgroups);
 	}
 	mutex_unlock(&cgroup_mutex);
 	return 0;
@@ -1765,18 +2150,19 @@ static struct file_operations proc_cgroupstats_operations = {
  * fork.c by dup_task_struct().  However, we ignore that copy, since
  * it was not made under the protection of RCU or cgroup_mutex, so
  * might no longer be a valid cgroup pointer.  attach_task() might
- * have already changed current->cgroup, allowing the previously
- * referenced cgroup to be removed and freed.
+ * have already changed current->cgroups, allowing the previously
+ * referenced cgroup group to be removed and freed.
  *
  * At the point that cgroup_fork() is called, 'current' is the parent
  * task, and the passed argument 'child' points to the child task.
  */
 void cgroup_fork(struct task_struct *child)
 {
-	rcu_read_lock();
-	child->cgroups = rcu_dereference(current->cgroups);
-	get_css_set(&child->cgroups);
-	rcu_read_unlock();
+	task_lock(current);
+	child->cgroups = current->cgroups;
+	get_css_set(child->cgroups);
+	task_unlock(current);
+	INIT_LIST_HEAD(&child->cg_list);
 }
 
 /**
@@ -1796,6 +2182,21 @@ void cgroup_fork_callbacks(struct task_struct *child)
 	}
 }
 
+/**
+ * cgroup_post_fork - called on a new task after adding it to the
+ * task list. Adds the task to the list running through its css_set
+ * if necessary. Has to be after the task is visible on the task list
+ * in case we race with the first call to cgroup_iter_start() - to
+ * guarantee that the new task ends up on its list. */
+void cgroup_post_fork(struct task_struct *child)
+{
+	if (use_task_css_set_links) {
+		write_lock(&css_set_lock);
+		if (list_empty(&child->cg_list))
+			list_add(&child->cg_list, &child->cgroups->tasks);
+		write_unlock(&css_set_lock);
+	}
+}
 /**
  * cgroup_exit - detach cgroup from exiting task
  * @tsk: pointer to task_struct of exiting process
@@ -1834,6 +2235,7 @@ void cgroup_fork_callbacks(struct task_struct *child)
 void cgroup_exit(struct task_struct *tsk, int run_callbacks)
 {
 	int i;
+	struct css_set *cg;
 
 	if (run_callbacks && need_forkexit_callback) {
 		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
@@ -1842,11 +2244,26 @@ void cgroup_exit(struct task_struct *tsk, int run_callbacks)
 				ss->exit(ss, tsk);
 		}
 	}
+
+	/*
+	 * Unlink from the css_set task list if necessary.
+	 * Optimistically check cg_list before taking
+	 * css_set_lock
+	 */
+	if (!list_empty(&tsk->cg_list)) {
+		write_lock(&css_set_lock);
+		if (!list_empty(&tsk->cg_list))
+			list_del(&tsk->cg_list);
+		write_unlock(&css_set_lock);
+	}
+
 	/* Reassign the task to the init_css_set. */
 	task_lock(tsk);
-	put_css_set(&tsk->cgroups);
-	tsk->cgroups = init_task.cgroups;
+	cg = tsk->cgroups;
+	tsk->cgroups = &init_css_set;
 	task_unlock(tsk);
+	if (cg)
+		put_css_set(cg);
 }
 
 /**
@@ -1880,7 +2297,7 @@ int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys)
 		mutex_unlock(&cgroup_mutex);
 		return 0;
 	}
-	cg = &tsk->cgroups;
+	cg = tsk->cgroups;
 	parent = task_cgroup(tsk, subsys->subsys_id);
 
 	snprintf(nodename, MAX_CGROUP_TYPE_NAMELEN, "node_%d", tsk->pid);
@@ -1888,6 +2305,8 @@ int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys)
 	/* Pin the hierarchy */
 	atomic_inc(&parent->root->sb->s_active);
 
+	/* Keep the cgroup alive */
+	get_css_set(cg);
 	mutex_unlock(&cgroup_mutex);
 
 	/* Now do the VFS work to create a cgroup */
@@ -1931,6 +2350,7 @@ int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys)
 	    (parent != task_cgroup(tsk, subsys->subsys_id))) {
 		/* Aargh, we raced ... */
 		mutex_unlock(&inode->i_mutex);
+		put_css_set(cg);
 
 		deactivate_super(parent->root->sb);
 		/* The cgroup is still accessible in the VFS, but
@@ -1954,6 +2374,7 @@ int cgroup_clone(struct task_struct *tsk, struct cgroup_subsys *subsys)
 
  out_release:
 	mutex_unlock(&inode->i_mutex);
+	put_css_set(cg);
 	deactivate_super(parent->root->sb);
 	return ret;
 }

kernel/fork.c

@@ -1301,6 +1301,7 @@ static struct task_struct *copy_process(unsigned long clone_flags,
 	spin_unlock(&current->sighand->siglock);
 	write_unlock_irq(&tasklist_lock);
 	proc_fork_connector(p);
+	cgroup_post_fork(p);
 	return p;
 
 bad_fork_cleanup_namespaces: