Merge branch 'for-3.9-cpuset' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup

Pull cpuset changes from Tejun Heo:

 - Synchornization has seen a lot of changes with focus on decoupling
   cpuset synchronization from cgroup internal locking.

   After this change, there only remain a couple of mostly trivial
   dependencies on cgroup_lock outside cgroup core proper.  cgroup_lock
   is scheduled to be unexported in this devel cycle.

   This will finally remove the fragile locking order around cgroup
   (cgroup locking wants to / should be one of the outermost but yet has
   been acquired from deep inside individual controllers).

 - At this point, Li is most knowlegeable with cpuset and taking over
   the maintainership of cpuset.

* 'for-3.9-cpuset' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cpuset: drop spurious retval assignment in proc_cpuset_show()
  cpuset: fix RCU lockdep splat
  cpuset: update MAINTAINERS
  cpuset: remove cpuset->parent
  cpuset: replace cpuset->stack_list with cpuset_for_each_descendant_pre()
  cpuset: replace cgroup_mutex locking with cpuset internal locking
  cpuset: schedule hotplug propagation from cpuset_attach() if the cpuset is empty
  cpuset: pin down cpus and mems while a task is being attached
  cpuset: make CPU / memory hotplug propagation asynchronous
  cpuset: drop async_rebuild_sched_domains()
  cpuset: don't nest cgroup_mutex inside get_online_cpus()
  cpuset: reorganize CPU / memory hotplug handling
  cpuset: cleanup cpuset[_can]_attach()
  cpuset: introduce cpuset_for_each_child()
  cpuset: introduce CS_ONLINE
  cpuset: introduce ->css_on/offline()
  cpuset: remove fast exit path from remove_tasks_in_empty_cpuset()
  cpuset: remove unused cpuset_unlock()

1 files changed, 483 insertions, 389 deletions

diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index 5bb9bf18438..4f9dfe43ecb 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -61,14 +61,6 @@
 #include <linux/cgroup.h>
 
 /*
- * Workqueue for cpuset related tasks.
- *
- * Using kevent workqueue may cause deadlock when memory_migrate
- * is set. So we create a separate workqueue thread for cpuset.
- */
-static struct workqueue_struct *cpuset_wq;
-
-/*
  * Tracks how many cpusets are currently defined in system.
  * When there is only one cpuset (the root cpuset) we can
  * short circuit some hooks.
@@ -95,18 +87,21 @@ struct cpuset {
 	cpumask_var_t cpus_allowed;	/* CPUs allowed to tasks in cpuset */
 	nodemask_t mems_allowed;	/* Memory Nodes allowed to tasks */
 
-	struct cpuset *parent;		/* my parent */
-
 	struct fmeter fmeter;		/* memory_pressure filter */
 
+	/*
+	 * Tasks are being attached to this cpuset.  Used to prevent
+	 * zeroing cpus/mems_allowed between ->can_attach() and ->attach().
+	 */
+	int attach_in_progress;
+
 	/* partition number for rebuild_sched_domains() */
 	int pn;
 
 	/* for custom sched domain */
 	int relax_domain_level;
 
-	/* used for walking a cpuset hierarchy */
-	struct list_head stack_list;
+	struct work_struct hotplug_work;
 };
 
 /* Retrieve the cpuset for a cgroup */
@@ -123,6 +118,15 @@ static inline struct cpuset *task_cs(struct task_struct *task)
 			    struct cpuset, css);
 }
 
+static inline struct cpuset *parent_cs(const struct cpuset *cs)
+{
+	struct cgroup *pcgrp = cs->css.cgroup->parent;
+
+	if (pcgrp)
+		return cgroup_cs(pcgrp);
+	return NULL;
+}
+
 #ifdef CONFIG_NUMA
 static inline bool task_has_mempolicy(struct task_struct *task)
 {
@@ -138,6 +142,7 @@ static inline bool task_has_mempolicy(struct task_struct *task)
 
 /* bits in struct cpuset flags field */
 typedef enum {
+	CS_ONLINE,
 	CS_CPU_EXCLUSIVE,
 	CS_MEM_EXCLUSIVE,
 	CS_MEM_HARDWALL,
@@ -147,13 +152,12 @@ typedef enum {
 	CS_SPREAD_SLAB,
 } cpuset_flagbits_t;
 
-/* the type of hotplug event */
-enum hotplug_event {
-	CPUSET_CPU_OFFLINE,
-	CPUSET_MEM_OFFLINE,
-};
-
 /* convenient tests for these bits */
+static inline bool is_cpuset_online(const struct cpuset *cs)
+{
+	return test_bit(CS_ONLINE, &cs->flags);
+}
+
 static inline int is_cpu_exclusive(const struct cpuset *cs)
 {
 	return test_bit(CS_CPU_EXCLUSIVE, &cs->flags);
@@ -190,27 +194,52 @@ static inline int is_spread_slab(const struct cpuset *cs)
 }
 
 static struct cpuset top_cpuset = {
-	.flags = ((1 << CS_CPU_EXCLUSIVE) | (1 << CS_MEM_EXCLUSIVE)),
+	.flags = ((1 << CS_ONLINE) | (1 << CS_CPU_EXCLUSIVE) |
+		  (1 << CS_MEM_EXCLUSIVE)),
 };
 
+/**
+ * cpuset_for_each_child - traverse online children of a cpuset
+ * @child_cs: loop cursor pointing to the current child
+ * @pos_cgrp: used for iteration
+ * @parent_cs: target cpuset to walk children of
+ *
+ * Walk @child_cs through the online children of @parent_cs.  Must be used
+ * with RCU read locked.
+ */
+#define cpuset_for_each_child(child_cs, pos_cgrp, parent_cs)		\
+	cgroup_for_each_child((pos_cgrp), (parent_cs)->css.cgroup)	\
+		if (is_cpuset_online(((child_cs) = cgroup_cs((pos_cgrp)))))
+
+/**
+ * cpuset_for_each_descendant_pre - pre-order walk of a cpuset's descendants
+ * @des_cs: loop cursor pointing to the current descendant
+ * @pos_cgrp: used for iteration
+ * @root_cs: target cpuset to walk ancestor of
+ *
+ * Walk @des_cs through the online descendants of @root_cs.  Must be used
+ * with RCU read locked.  The caller may modify @pos_cgrp by calling
+ * cgroup_rightmost_descendant() to skip subtree.
+ */
+#define cpuset_for_each_descendant_pre(des_cs, pos_cgrp, root_cs)	\
+	cgroup_for_each_descendant_pre((pos_cgrp), (root_cs)->css.cgroup) \
+		if (is_cpuset_online(((des_cs) = cgroup_cs((pos_cgrp)))))
+
 /*
- * There are two global mutexes guarding cpuset structures.  The first
- * is the main control groups cgroup_mutex, accessed via
- * cgroup_lock()/cgroup_unlock().  The second is the cpuset-specific
- * callback_mutex, below. They can nest.  It is ok to first take
- * cgroup_mutex, then nest callback_mutex.  We also require taking
- * task_lock() when dereferencing a task's cpuset pointer.  See "The
- * task_lock() exception", at the end of this comment.
- *
- * A task must hold both mutexes to modify cpusets.  If a task
- * holds cgroup_mutex, then it blocks others wanting that mutex,
- * ensuring that it is the only task able to also acquire callback_mutex
- * and be able to modify cpusets.  It can perform various checks on
- * the cpuset structure first, knowing nothing will change.  It can
- * also allocate memory while just holding cgroup_mutex.  While it is
- * performing these checks, various callback routines can briefly
- * acquire callback_mutex to query cpusets.  Once it is ready to make
- * the changes, it takes callback_mutex, blocking everyone else.
+ * There are two global mutexes guarding cpuset structures - cpuset_mutex
+ * and callback_mutex.  The latter may nest inside the former.  We also
+ * require taking task_lock() when dereferencing a task's cpuset pointer.
+ * See "The task_lock() exception", at the end of this comment.
+ *
+ * A task must hold both mutexes to modify cpusets.  If a task holds
+ * cpuset_mutex, then it blocks others wanting that mutex, ensuring that it
+ * is the only task able to also acquire callback_mutex and be able to
+ * modify cpusets.  It can perform various checks on the cpuset structure
+ * first, knowing nothing will change.  It can also allocate memory while
+ * just holding cpuset_mutex.  While it is performing these checks, various
+ * callback routines can briefly acquire callback_mutex to query cpusets.
+ * Once it is ready to make the changes, it takes callback_mutex, blocking
+ * everyone else.
  *
  * Calls to the kernel memory allocator can not be made while holding
  * callback_mutex, as that would risk double tripping on callback_mutex
@@ -232,6 +261,7 @@ static struct cpuset top_cpuset = {
  * guidelines for accessing subsystem state in kernel/cgroup.c
  */
 
+static DEFINE_MUTEX(cpuset_mutex);
 static DEFINE_MUTEX(callback_mutex);
 
 /*
@@ -246,6 +276,17 @@ static char cpuset_nodelist[CPUSET_NODELIST_LEN];
 static DEFINE_SPINLOCK(cpuset_buffer_lock);
 
 /*
+ * CPU / memory hotplug is handled asynchronously.
+ */
+static struct workqueue_struct *cpuset_propagate_hotplug_wq;
+
+static void cpuset_hotplug_workfn(struct work_struct *work);
+static void cpuset_propagate_hotplug_workfn(struct work_struct *work);
+static void schedule_cpuset_propagate_hotplug(struct cpuset *cs);
+
+static DECLARE_WORK(cpuset_hotplug_work, cpuset_hotplug_workfn);
+
+/*
  * This is ugly, but preserves the userspace API for existing cpuset
  * users. If someone tries to mount the "cpuset" filesystem, we
  * silently switch it to mount "cgroup" instead
@@ -289,7 +330,7 @@ static void guarantee_online_cpus(const struct cpuset *cs,
 				  struct cpumask *pmask)
 {
 	while (cs && !cpumask_intersects(cs->cpus_allowed, cpu_online_mask))
-		cs = cs->parent;
+		cs = parent_cs(cs);
 	if (cs)
 		cpumask_and(pmask, cs->cpus_allowed, cpu_online_mask);
 	else
@@ -314,7 +355,7 @@ static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
 {
 	while (cs && !nodes_intersects(cs->mems_allowed,
 					node_states[N_MEMORY]))
-		cs = cs->parent;
+		cs = parent_cs(cs);
 	if (cs)
 		nodes_and(*pmask, cs->mems_allowed,
 					node_states[N_MEMORY]);
@@ -326,7 +367,7 @@ static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
 /*
  * update task's spread flag if cpuset's page/slab spread flag is set
  *
- * Called with callback_mutex/cgroup_mutex held
+ * Called with callback_mutex/cpuset_mutex held
  */
 static void cpuset_update_task_spread_flag(struct cpuset *cs,
 					struct task_struct *tsk)
@@ -346,7 +387,7 @@ static void cpuset_update_task_spread_flag(struct cpuset *cs,
  *
  * One cpuset is a subset of another if all its allowed CPUs and
  * Memory Nodes are a subset of the other, and its exclusive flags
- * are only set if the other's are set.  Call holding cgroup_mutex.
+ * are only set if the other's are set.  Call holding cpuset_mutex.
  */
 
 static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
@@ -395,7 +436,7 @@ static void free_trial_cpuset(struct cpuset *trial)
  * If we replaced the flag and mask values of the current cpuset
  * (cur) with those values in the trial cpuset (trial), would
  * our various subset and exclusive rules still be valid?  Presumes
- * cgroup_mutex held.
+ * cpuset_mutex held.
  *
  * 'cur' is the address of an actual, in-use cpuset.  Operations
  * such as list traversal that depend on the actual address of the
@@ -412,48 +453,58 @@ static int validate_change(const struct cpuset *cur, const struct cpuset *trial)
 {
 	struct cgroup *cont;
 	struct cpuset *c, *par;
+	int ret;
+
+	rcu_read_lock();
 
 	/* Each of our child cpusets must be a subset of us */
-	list_for_each_entry(cont, &cur->css.cgroup->children, sibling) {
-		if (!is_cpuset_subset(cgroup_cs(cont), trial))
-			return -EBUSY;
-	}
+	ret = -EBUSY;
+	cpuset_for_each_child(c, cont, cur)
+		if (!is_cpuset_subset(c, trial))
+			goto out;
 
 	/* Remaining checks don't apply to root cpuset */
+	ret = 0;
 	if (cur == &top_cpuset)
-		return 0;
+		goto out;
 
-	par = cur->parent;
+	par = parent_cs(cur);
 
 	/* We must be a subset of our parent cpuset */
+	ret = -EACCES;
 	if (!is_cpuset_subset(trial, par))
-		return -EACCES;
+		goto out;
 
 	/*
 	 * If either I or some sibling (!= me) is exclusive, we can't
 	 * overlap
 	 */
-	list_for_each_entry(cont, &par->css.cgroup->children, sibling) {
-		c = cgroup_cs(cont);
+	ret = -EINVAL;
+	cpuset_for_each_child(c, cont, par) {
 		if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) &&
 		    c != cur &&
 		    cpumask_intersects(trial->cpus_allowed, c->cpus_allowed))
-			return -EINVAL;
+			goto out;
 		if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) &&
 		    c != cur &&
 		    nodes_intersects(trial->mems_allowed, c->mems_allowed))
-			return -EINVAL;
+			goto out;
 	}
 
-	/* Cpusets with tasks can't have empty cpus_allowed or mems_allowed */
-	if (cgroup_task_count(cur->css.cgroup)) {
-		if (cpumask_empty(trial->cpus_allowed) ||
-		    nodes_empty(trial->mems_allowed)) {
-			return -ENOSPC;
-		}
-	}
+	/*
+	 * Cpusets with tasks - existing or newly being attached - can't
+	 * have empty cpus_allowed or mems_allowed.
+	 */
+	ret = -ENOSPC;
+	if ((cgroup_task_count(cur->css.cgroup) || cur->attach_in_progress) &&
+	    (cpumask_empty(trial->cpus_allowed) ||
+	     nodes_empty(trial->mems_allowed)))
+		goto out;
 
-	return 0;
+	ret = 0;
+out:
+	rcu_read_unlock();
+	return ret;
 }
 
 #ifdef CONFIG_SMP
@@ -474,31 +525,24 @@ update_domain_attr(struct sched_domain_attr *dattr, struct cpuset *c)
 	return;
 }
 
-static void
-update_domain_attr_tree(struct sched_domain_attr *dattr, struct cpuset *c)
+static void update_domain_attr_tree(struct sched_domain_attr *dattr,
+				    struct cpuset *root_cs)
 {
-	LIST_HEAD(q);
-
-	list_add(&c->stack_list, &q);
-	while (!list_empty(&q)) {
-		struct cpuset *cp;
-		struct cgroup *cont;
-		struct cpuset *child;
-
-		cp = list_first_entry(&q, struct cpuset, stack_list);
-		list_del(q.next);
+	struct cpuset *cp;
+	struct cgroup *pos_cgrp;
 
-		if (cpumask_empty(cp->cpus_allowed))
+	rcu_read_lock();
+	cpuset_for_each_descendant_pre(cp, pos_cgrp, root_cs) {
+		/* skip the whole subtree if @cp doesn't have any CPU */
+		if (cpumask_empty(cp->cpus_allowed)) {
+			pos_cgrp = cgroup_rightmost_descendant(pos_cgrp);
 			continue;
+		}
 
 		if (is_sched_load_balance(cp))
 			update_domain_attr(dattr, cp);
-
-		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
-			child = cgroup_cs(cont);
-			list_add_tail(&child->stack_list, &q);
-		}
 	}
+	rcu_read_unlock();
 }
 
 /*
@@ -520,7 +564,7 @@ update_domain_attr_tree(struct sched_domain_attr *dattr, struct cpuset *c)
  * domains when operating in the severe memory shortage situations
  * that could cause allocation failures below.
  *
- * Must be called with cgroup_lock held.
+ * Must be called with cpuset_mutex held.
  *
  * The three key local variables below are:
  *    q  - a linked-list queue of cpuset pointers, used to implement a
@@ -558,7 +602,6 @@ update_domain_attr_tree(struct sched_domain_attr *dattr, struct cpuset *c)
 static int generate_sched_domains(cpumask_var_t **domains,
 			struct sched_domain_attr **attributes)
 {
-	LIST_HEAD(q);		/* queue of cpusets to be scanned */
 	struct cpuset *cp;	/* scans q */
 	struct cpuset **csa;	/* array of all cpuset ptrs */
 	int csn;		/* how many cpuset ptrs in csa so far */
@@ -567,6 +610,7 @@ static int generate_sched_domains(cpumask_var_t **domains,
 	struct sched_domain_attr *dattr;  /* attributes for custom domains */
 	int ndoms = 0;		/* number of sched domains in result */
 	int nslot;		/* next empty doms[] struct cpumask slot */
+	struct cgroup *pos_cgrp;
 
 	doms = NULL;
 	dattr = NULL;
@@ -594,33 +638,27 @@ static int generate_sched_domains(cpumask_var_t **domains,
 		goto done;
 	csn = 0;
 
-	list_add(&top_cpuset.stack_list, &q);
-	while (!list_empty(&q)) {
-		struct cgroup *cont;
-		struct cpuset *child;   /* scans child cpusets of cp */
-
-		cp = list_first_entry(&q, struct cpuset, stack_list);
-		list_del(q.next);
-
-		if (cpumask_empty(cp->cpus_allowed))
-			continue;
-
+	rcu_read_lock();
+	cpuset_for_each_descendant_pre(cp, pos_cgrp, &top_cpuset) {
 		/*
-		 * All child cpusets contain a subset of the parent's cpus, so
-		 * just skip them, and then we call update_domain_attr_tree()
-		 * to calc relax_domain_level of the corresponding sched
-		 * domain.
+		 * Continue traversing beyond @cp iff @cp has some CPUs and
+		 * isn't load balancing.  The former is obvious.  The
+		 * latter: All child cpusets contain a subset of the
+		 * parent's cpus, so just skip them, and then we call
+		 * update_domain_attr_tree() to calc relax_domain_level of
+		 * the corresponding sched domain.
 		 */
-		if (is_sched_load_balance(cp)) {
-			csa[csn++] = cp;
+		if (!cpumask_empty(cp->cpus_allowed) &&
+		    !is_sched_load_balance(cp))
 			continue;
-		}
 
-		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
-			child = cgroup_cs(cont);
-			list_add_tail(&child->stack_list, &q);
-		}
-  	}
+		if (is_sched_load_balance(cp))
+			csa[csn++] = cp;
+
+		/* skip @cp's subtree */
+		pos_cgrp = cgroup_rightmost_descendant(pos_cgrp);
+	}
+	rcu_read_unlock();
 
 	for (i = 0; i < csn; i++)
 		csa[i]->pn = i;
@@ -725,25 +763,25 @@ done:
 /*
  * Rebuild scheduler domains.
  *
- * Call with neither cgroup_mutex held nor within get_online_cpus().
- * Takes both cgroup_mutex and get_online_cpus().
+ * If the flag 'sched_load_balance' of any cpuset with non-empty
+ * 'cpus' changes, or if the 'cpus' allowed changes in any cpuset
+ * which has that flag enabled, or if any cpuset with a non-empty
+ * 'cpus' is removed, then call this routine to rebuild the
+ * scheduler's dynamic sched domains.
  *
- * Cannot be directly called from cpuset code handling changes
- * to the cpuset pseudo-filesystem, because it cannot be called
- * from code that already holds cgroup_mutex.
+ * Call with cpuset_mutex held.  Takes get_online_cpus().
  */
-static void do_rebuild_sched_domains(struct work_struct *unused)
+static void rebuild_sched_domains_locked(void)
 {
 	struct sched_domain_attr *attr;
 	cpumask_var_t *doms;
 	int ndoms;
 
+	lockdep_assert_held(&cpuset_mutex);
 	get_online_cpus();
 
 	/* Generate domain masks and attrs */
-	cgroup_lock();
 	ndoms = generate_sched_domains(&doms, &attr);
-	cgroup_unlock();
 
 	/* Have scheduler rebuild the domains */
 	partition_sched_domains(ndoms, doms, attr);
@@ -751,7 +789,7 @@ static void do_rebuild_sched_domains(struct work_struct *unused)
 	put_online_cpus();
 }
 #else /* !CONFIG_SMP */
-static void do_rebuild_sched_domains(struct work_struct *unused)
+static void rebuild_sched_domains_locked(void)
 {
 }
 
@@ -763,44 +801,11 @@ static int generate_sched_domains(cpumask_var_t **domains,
 }
 #endif /* CONFIG_SMP */
 
-static DECLARE_WORK(rebuild_sched_domains_work, do_rebuild_sched_domains);
-
-/*
- * Rebuild scheduler domains, asynchronously via workqueue.
- *
- * If the flag 'sched_load_balance' of any cpuset with non-empty
- * 'cpus' changes, or if the 'cpus' allowed changes in any cpuset
- * which has that flag enabled, or if any cpuset with a non-empty
- * 'cpus' is removed, then call this routine to rebuild the
- * scheduler's dynamic sched domains.
- *
- * The rebuild_sched_domains() and partition_sched_domains()
- * routines must nest cgroup_lock() inside get_online_cpus(),
- * but such cpuset changes as these must nest that locking the
- * other way, holding cgroup_lock() for much of the code.
- *
- * So in order to avoid an ABBA deadlock, the cpuset code handling
- * these user changes delegates the actual sched domain rebuilding
- * to a separate workqueue thread, which ends up processing the
- * above do_rebuild_sched_domains() function.
- */
-static void async_rebuild_sched_domains(void)
-{
-	queue_work(cpuset_wq, &rebuild_sched_domains_work);
-}
-
-/*
- * Accomplishes the same scheduler domain rebuild as the above
- * async_rebuild_sched_domains(), however it directly calls the
- * rebuild routine synchronously rather than calling it via an
- * asynchronous work thread.
- *
- * This can only be called from code that is not holding
- * cgroup_mutex (not nested in a cgroup_lock() call.)
- */
 void rebuild_sched_domains(void)
 {
-	do_rebuild_sched_domains(NULL);
+	mutex_lock(&cpuset_mutex);
+	rebuild_sched_domains_locked();
+	mutex_unlock(&cpuset_mutex);
 }
 
 /**
@@ -808,7 +813,7 @@ void rebuild_sched_domains(void)
  * @tsk: task to test
  * @scan: struct cgroup_scanner contained in its struct cpuset_hotplug_scanner
  *
- * Call with cgroup_mutex held.  May take callback_mutex during call.
+ * Call with cpuset_mutex held.  May take callback_mutex during call.
  * Called for each task in a cgroup by cgroup_scan_tasks().
  * Return nonzero if this tasks's cpus_allowed mask should be changed (in other
  * words, if its mask is not equal to its cpuset's mask).
@@ -829,7 +834,7 @@ static int cpuset_test_cpumask(struct task_struct *tsk,
  * cpus_allowed mask needs to be changed.
  *
  * We don't need to re-check for the cgroup/cpuset membership, since we're
- * holding cgroup_lock() at this point.
+ * holding cpuset_mutex at this point.
  */
 static void cpuset_change_cpumask(struct task_struct *tsk,
 				  struct cgroup_scanner *scan)
@@ -842,7 +847,7 @@ static void cpuset_change_cpumask(struct task_struct *tsk,
  * @cs: the cpuset in which each task's cpus_allowed mask needs to be changed
  * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
  *
- * Called with cgroup_mutex held
+ * Called with cpuset_mutex held
  *
  * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
  * calling callback functions for each.
@@ -920,7 +925,7 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
 	heap_free(&heap);
 
 	if (is_load_balanced)
-		async_rebuild_sched_domains();
+		rebuild_sched_domains_locked();
 	return 0;
 }
 
@@ -932,7 +937,7 @@ static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
  *    Temporarilly set tasks mems_allowed to target nodes of migration,
  *    so that the migration code can allocate pages on these nodes.
  *
- *    Call holding cgroup_mutex, so current's cpuset won't change
+ *    Call holding cpuset_mutex, so current's cpuset won't change
  *    during this call, as manage_mutex holds off any cpuset_attach()
  *    calls.  Therefore we don't need to take task_lock around the
  *    call to guarantee_online_mems(), as we know no one is changing
@@ -1007,7 +1012,7 @@ static void cpuset_change_task_nodemask(struct task_struct *tsk,
 /*
  * Update task's mems_allowed and rebind its mempolicy and vmas' mempolicy
  * of it to cpuset's new mems_allowed, and migrate pages to new nodes if
- * memory_migrate flag is set. Called with cgroup_mutex held.
+ * memory_migrate flag is set. Called with cpuset_mutex held.
  */
 static void cpuset_change_nodemask(struct task_struct *p,
 				   struct cgroup_scanner *scan)
@@ -1016,7 +1021,7 @@ static void cpuset_change_nodemask(struct task_struct *p,
 	struct cpuset *cs;
 	int migrate;
 	const nodemask_t *oldmem = scan->data;
-	static nodemask_t newmems;	/* protected by cgroup_mutex */
+	static nodemask_t newmems;	/* protected by cpuset_mutex */
 
 	cs = cgroup_cs(scan->cg);
 	guarantee_online_mems(cs, &newmems);
@@ -1043,7 +1048,7 @@ static void *cpuset_being_rebound;
  * @oldmem: old mems_allowed of cpuset cs
  * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
  *
- * Called with cgroup_mutex held
+ * Called with cpuset_mutex held
  * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
  * if @heap != NULL.
  */
@@ -1065,7 +1070,7 @@ static void update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem,
 	 * take while holding tasklist_lock.  Forks can happen - the
 	 * mpol_dup() cpuset_being_rebound check will catch such forks,
 	 * and rebind their vma mempolicies too.  Because we still hold
-	 * the global cgroup_mutex, we know that no other rebind effort
+	 * the global cpuset_mutex, we know that no other rebind effort
 	 * will be contending for the global variable cpuset_being_rebound.
 	 * It's ok if we rebind the same mm twice; mpol_rebind_mm()
 	 * is idempotent.  Also migrate pages in each mm to new nodes.
@@ -1084,7 +1089,7 @@ static void update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem,
  * mempolicies and if the cpuset is marked 'memory_migrate',
  * migrate the tasks pages to the new memory.
  *
- * Call with cgroup_mutex held.  May take callback_mutex during call.
+ * Call with cpuset_mutex held.  May take callback_mutex during call.
  * Will take tasklist_lock, scan tasklist for tasks in cpuset cs,
  * lock each such tasks mm->mmap_sem, scan its vma's and rebind
  * their mempolicies to the cpusets new mems_allowed.
@@ -1168,7 +1173,7 @@ static int update_relax_domain_level(struct cpuset *cs, s64 val)
 		cs->relax_domain_level = val;
 		if (!cpumask_empty(cs->cpus_allowed) &&
 		    is_sched_load_balance(cs))
-			async_rebuild_sched_domains();
+			rebuild_sched_domains_locked();
 	}
 
 	return 0;
@@ -1182,7 +1187,7 @@ static int update_relax_domain_level(struct cpuset *cs, s64 val)
  * Called by cgroup_scan_tasks() for each task in a cgroup.
  *
  * We don't need to re-check for the cgroup/cpuset membership, since we're
- * holding cgroup_lock() at this point.
+ * holding cpuset_mutex at this point.
  */
 static void cpuset_change_flag(struct task_struct *tsk,
 				struct cgroup_scanner *scan)
@@ -1195,7 +1200,7 @@ static void cpuset_change_flag(struct task_struct *tsk,
  * @cs: the cpuset in which each task's spread flags needs to be changed
  * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
  *
- * Called with cgroup_mutex held
+ * Called with cpuset_mutex held
  *
  * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
  * calling callback functions for each.
@@ -1220,7 +1225,7 @@ static void update_tasks_flags(struct cpuset *cs, struct ptr_heap *heap)
  * cs:		the cpuset to update
  * turning_on: 	whether the flag is being set or cleared
  *
- * Call with cgroup_mutex held.
+ * Call with cpuset_mutex held.
  */
 
 static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
@@ -1260,7 +1265,7 @@ static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
 	mutex_unlock(&callback_mutex);
 
 	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed)
-		async_rebuild_sched_domains();
+		rebuild_sched_domains_locked();
 
 	if (spread_flag_changed)
 		update_tasks_flags(cs, &heap);
@@ -1368,24 +1373,18 @@ static int fmeter_getrate(struct fmeter *fmp)
 	return val;
 }
 
-/*
- * Protected by cgroup_lock. The nodemasks must be stored globally because
- * dynamically allocating them is not allowed in can_attach, and they must
- * persist until attach.
- */
-static cpumask_var_t cpus_attach;
-static nodemask_t cpuset_attach_nodemask_from;
-static nodemask_t cpuset_attach_nodemask_to;
-
-/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
+/* Called by cgroups to determine if a cpuset is usable; cpuset_mutex held */
 static int cpuset_can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 {
 	struct cpuset *cs = cgroup_cs(cgrp);
 	struct task_struct *task;
 	int ret;
 
+	mutex_lock(&cpuset_mutex);
+
+	ret = -ENOSPC;
 	if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
-		return -ENOSPC;
+		goto out_unlock;
 
 	cgroup_taskset_for_each(task, cgrp, tset) {
 		/*
@@ -1397,25 +1396,45 @@ static int cpuset_can_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 		 * set_cpus_allowed_ptr() on all attached tasks before
 		 * cpus_allowed may be changed.
 		 */
+		ret = -EINVAL;
 		if (task->flags & PF_THREAD_BOUND)
-			return -EINVAL;
-		if ((ret = security_task_setscheduler(task)))
-			return ret;
+			goto out_unlock;
+		ret = security_task_setscheduler(task);
+		if (ret)
+			goto out_unlock;
 	}
 
-	/* prepare for attach */
-	if (cs == &top_cpuset)
-		cpumask_copy(cpus_attach, cpu_possible_mask);
-	else
-		guarantee_online_cpus(cs, cpus_attach);
-
-	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
+	/*
+	 * Mark attach is in progress.  This makes validate_change() fail
+	 * changes which zero cpus/mems_allowed.
+	 */
+	cs->attach_in_progress++;
+	ret = 0;
+out_unlock:
+	mutex_unlock(&cpuset_mutex);
+	return ret;
+}
 
-	return 0;
+static void cpuset_cancel_attach(struct cgroup *cgrp,
+				 struct cgroup_taskset *tset)
+{
+	mutex_lock(&cpuset_mutex);
+	cgroup_cs(cgrp)->attach_in_progress--;
+	mutex_unlock(&cpuset_mutex);
 }
 
+/*
+ * Protected by cpuset_mutex.  cpus_attach is used only by cpuset_attach()
+ * but we can't allocate it dynamically there.  Define it global and
+ * allocate from cpuset_init().
+ */
+static cpumask_var_t cpus_attach;
+
 static void cpuset_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 {
+	/* static bufs protected by cpuset_mutex */
+	static nodemask_t cpuset_attach_nodemask_from;
+	static nodemask_t cpuset_attach_nodemask_to;
 	struct mm_struct *mm;
 	struct task_struct *task;
 	struct task_struct *leader = cgroup_taskset_first(tset);
@@ -1423,6 +1442,16 @@ static void cpuset_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 	struct cpuset *cs = cgroup_cs(cgrp);
 	struct cpuset *oldcs = cgroup_cs(oldcgrp);
 
+	mutex_lock(&cpuset_mutex);
+
+	/* prepare for attach */
+	if (cs == &top_cpuset)
+		cpumask_copy(cpus_attach, cpu_possible_mask);
+	else
+		guarantee_online_cpus(cs, cpus_attach);
+
+	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
+
 	cgroup_taskset_for_each(task, cgrp, tset) {
 		/*
 		 * can_attach beforehand should guarantee that this doesn't
@@ -1448,6 +1477,18 @@ static void cpuset_attach(struct cgroup *cgrp, struct cgroup_taskset *tset)
 					  &cpuset_attach_nodemask_to);
 		mmput(mm);
 	}
+
+	cs->attach_in_progress--;
+
+	/*
+	 * We may have raced with CPU/memory hotunplug.  Trigger hotplug
+	 * propagation if @cs doesn't have any CPU or memory.  It will move
+	 * the newly added tasks to the nearest parent which can execute.
+	 */
+	if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
+		schedule_cpuset_propagate_hotplug(cs);
+
+	mutex_unlock(&cpuset_mutex);
 }
 
 /* The various types of files and directories in a cpuset file system */
@@ -1469,12 +1510,13 @@ typedef enum {
 
 static int cpuset_write_u64(struct cgroup *cgrp, struct cftype *cft, u64 val)
 {
-	int retval = 0;
 	struct cpuset *cs = cgroup_cs(cgrp);
 	cpuset_filetype_t type = cft->private;
+	int retval = -ENODEV;
 
-	if (!cgroup_lock_live_group(cgrp))
-		return -ENODEV;
+	mutex_lock(&cpuset_mutex);
+	if (!is_cpuset_online(cs))
+		goto out_unlock;
 
 	switch (type) {
 	case FILE_CPU_EXCLUSIVE:
@@ -1508,18 +1550,20 @@ static int cpuset_write_u64(struct cgroup *cgrp, struct cftype *cft, u64 val)
 		retval = -EINVAL;
 		break;
 	}
-	cgroup_unlock();
+out_unlock:
+	mutex_unlock(&cpuset_mutex);
 	return retval;
 }
 
 static int cpuset_write_s64(struct cgroup *cgrp, struct cftype *cft, s64 val)
 {
-	int retval = 0;
 	struct cpuset *cs = cgroup_cs(cgrp);
 	cpuset_filetype_t type = cft->private;
+	int retval = -ENODEV;
 
-	if (!cgroup_lock_live_group(cgrp))
-		return -ENODEV;
+	mutex_lock(&cpuset_mutex);
+	if (!is_cpuset_online(cs))
+		goto out_unlock;
 
 	switch (type) {
 	case FILE_SCHED_RELAX_DOMAIN_LEVEL:
@@ -1529,7 +1573,8 @@ static int cpuset_write_s64(struct cgroup *cgrp, struct cftype *cft, s64 val)
 		retval = -EINVAL;
 		break;
 	}
-	cgroup_unlock();
+out_unlock:
+	mutex_unlock(&cpuset_mutex);
 	return retval;
 }
 
@@ -1539,17 +1584,36 @@ static int cpuset_write_s64(struct cgroup *cgrp, struct cftype *cft, s64 val)
 static int cpuset_write_resmask(struct cgroup *cgrp, struct cftype *cft,
 				const char *buf)
 {
-	int retval = 0;
 	struct cpuset *cs = cgroup_cs(cgrp);
 	struct cpuset *trialcs;
+	int retval = -ENODEV;
 
-	if (!cgroup_lock_live_group(cgrp))
-		return -ENODEV;
+	/*
+	 * CPU or memory hotunplug may leave @cs w/o any execution
+	 * resources, in which case the hotplug code asynchronously updates
+	 * configuration and transfers all tasks to the nearest ancestor
+	 * which can execute.
+	 *
+	 * As writes to "cpus" or "mems" may restore @cs's execution
+	 * resources, wait for the previously scheduled operations before
+	 * proceeding, so that we don't end up keep removing tasks added
+	 * after execution capability is restored.
+	 *
+	 * Flushing cpuset_hotplug_work is enough to synchronize against
+	 * hotplug hanlding; however, cpuset_attach() may schedule
+	 * propagation work directly.  Flush the workqueue too.
+	 */
+	flush_work(&cpuset_hotplug_work);
+	flush_workqueue(cpuset_propagate_hotplug_wq);
+
+	mutex_lock(&cpuset_mutex);
+	if (!is_cpuset_online(cs))
+		goto out_unlock;
 
 	trialcs = alloc_trial_cpuset(cs);
 	if (!trialcs) {
 		retval = -ENOMEM;
-		goto out;
+		goto out_unlock;
 	}
 
 	switch (cft->private) {
@@ -1565,8 +1629,8 @@ static int cpuset_write_resmask(struct cgroup *cgrp, struct cftype *cft,
 	}
 
 	free_trial_cpuset(trialcs);
-out:
-	cgroup_unlock();
+out_unlock:
+	mutex_unlock(&cpuset_mutex);
 	return retval;
 }
 
@@ -1790,15 +1854,12 @@ static struct cftype files[] = {
 
 static struct cgroup_subsys_state *cpuset_css_alloc(struct cgroup *cont)
 {
-	struct cgroup *parent_cg = cont->parent;
-	struct cgroup *tmp_cg;
-	struct cpuset *parent, *cs;
+	struct cpuset *cs;
 
-	if (!parent_cg)
+	if (!cont->parent)
 		return &top_cpuset.css;
-	parent = cgroup_cs(parent_cg);
 
-	cs = kmalloc(sizeof(*cs), GFP_KERNEL);
+	cs = kzalloc(sizeof(*cs), GFP_KERNEL);
 	if (!cs)
 		return ERR_PTR(-ENOMEM);
 	if (!alloc_cpumask_var(&cs->cpus_allowed, GFP_KERNEL)) {
@@ -1806,22 +1867,38 @@ static struct cgroup_subsys_state *cpuset_css_alloc(struct cgroup *cont)
 		return ERR_PTR(-ENOMEM);
 	}
 
-	cs->flags = 0;
-	if (is_spread_page(parent))
-		set_bit(CS_SPREAD_PAGE, &cs->flags);
-	if (is_spread_slab(parent))
-		set_bit(CS_SPREAD_SLAB, &cs->flags);
 	set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
 	cpumask_clear(cs->cpus_allowed);
 	nodes_clear(cs->mems_allowed);
 	fmeter_init(&cs->fmeter);
+	INIT_WORK(&cs->hotplug_work, cpuset_propagate_hotplug_workfn);
 	cs->relax_domain_level = -1;
 
-	cs->parent = parent;
+	return &cs->css;
+}
+
+static int cpuset_css_online(struct cgroup *cgrp)
+{
+	struct cpuset *cs = cgroup_cs(cgrp);
+	struct cpuset *parent = parent_cs(cs);
+	struct cpuset *tmp_cs;
+	struct cgroup *pos_cg;
+
+	if (!parent)
+		return 0;
+
+	mutex_lock(&cpuset_mutex);
+
+	set_bit(CS_ONLINE, &cs->flags);
+	if (is_spread_page(parent))
+		set_bit(CS_SPREAD_PAGE, &cs->flags);
+	if (is_spread_slab(parent))
+		set_bit(CS_SPREAD_SLAB, &cs->flags);
+
 	number_of_cpusets++;
 
-	if (!test_bit(CGRP_CPUSET_CLONE_CHILDREN, &cont->flags))
-		goto skip_clone;
+	if (!test_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags))
+		goto out_unlock;
 
 	/*
 	 * Clone @parent's configuration if CGRP_CPUSET_CLONE_CHILDREN is
@@ -1836,35 +1913,49 @@ static struct cgroup_subsys_state *cpuset_css_alloc(struct cgroup *cont)
 	 * changed to grant parent->cpus_allowed-sibling_cpus_exclusive
 	 * (and likewise for mems) to the new cgroup.
 	 */
-	list_for_each_entry(tmp_cg, &parent_cg->children, sibling) {
-		struct cpuset *tmp_cs = cgroup_cs(tmp_cg);
-
-		if (is_mem_exclusive(tmp_cs) || is_cpu_exclusive(tmp_cs))
-			goto skip_clone;
+	rcu_read_lock();
+	cpuset_for_each_child(tmp_cs, pos_cg, parent) {
+		if (is_mem_exclusive(tmp_cs) || is_cpu_exclusive(tmp_cs)) {
+			rcu_read_unlock();
+			goto out_unlock;
+		}
 	}
+	rcu_read_unlock();
 
 	mutex_lock(&callback_mutex);
 	cs->mems_allowed = parent->mems_allowed;
 	cpumask_copy(cs->cpus_allowed, parent->cpus_allowed);
 	mutex_unlock(&callback_mutex);
-skip_clone:
-	return &cs->css;
+out_unlock:
+	mutex_unlock(&cpuset_mutex);
+	return 0;
+}
+
+static void cpuset_css_offline(struct cgroup *cgrp)
+{
+	struct cpuset *cs = cgroup_cs(cgrp);
+
+	mutex_lock(&cpuset_mutex);
+
+	if (is_sched_load_balance(cs))
+		update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
+
+	number_of_cpusets--;
+	clear_bit(CS_ONLINE, &cs->flags);
+
+	mutex_unlock(&cpuset_mutex);
 }
 
 /*
  * If the cpuset being removed has its flag 'sched_load_balance'
  * enabled, then simulate turning sched_load_balance off, which
- * will call async_rebuild_sched_domains().
+ * will call rebuild_sched_domains_locked().
  */
 
 static void cpuset_css_free(struct cgroup *cont)
 {
 	struct cpuset *cs = cgroup_cs(cont);
 
-	if (is_sched_load_balance(cs))
-		update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
-
-	number_of_cpusets--;
 	free_cpumask_var(cs->cpus_allowed);
 	kfree(cs);
 }
@@ -1872,8 +1963,11 @@ static void cpuset_css_free(struct cgroup *cont)
 struct cgroup_subsys cpuset_subsys = {
 	.name = "cpuset",
 	.css_alloc = cpuset_css_alloc,
+	.css_online = cpuset_css_online,
+	.css_offline = cpu