Merge branch 'for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq

Pull workqueue changes from Tejun Heo:
 "There are three major changes.

   - WQ_HIGHPRI has been reimplemented so that high priority work items
     are served by worker threads with -20 nice value from dedicated
     highpri worker pools.

   - CPU hotplug support has been reimplemented such that idle workers
     are kept across CPU hotplug events.  This makes CPU hotplug cheaper
     (for PM) and makes the code simpler.

   - flush_kthread_work() has been reimplemented so that a work item can
     be freed while executing.  This removes an annoying behavior
     difference between kthread_worker and workqueue."

* 'for-3.6' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: fix spurious CPU locality WARN from process_one_work()
  kthread_worker: reimplement flush_kthread_work() to allow freeing the work item being executed
  kthread_worker: reorganize to prepare for flush_kthread_work() reimplementation
  workqueue: simplify CPU hotplug code
  workqueue: remove CPU offline trustee
  workqueue: don't butcher idle workers on an offline CPU
  workqueue: reimplement CPU online rebinding to handle idle workers
  workqueue: drop @bind from create_worker()
  workqueue: use mutex for global_cwq manager exclusion
  workqueue: ROGUE workers are UNBOUND workers
  workqueue: drop CPU_DYING notifier operation
  workqueue: perform cpu down operations from low priority cpu_notifier()
  workqueue: reimplement WQ_HIGHPRI using a separate worker_pool
  workqueue: introduce NR_WORKER_POOLS and for_each_worker_pool()
  workqueue: separate out worker_pool flags
  workqueue: use @pool instead of @gcwq or @cpu where applicable
  workqueue: factor out worker_pool from global_cwq
  workqueue: don't use WQ_HIGHPRI for unbound workqueues

Documentation/workqueue.txt

@@ -89,25 +89,28 @@ called thread-pools.
 
 The cmwq design differentiates between the user-facing workqueues that
 subsystems and drivers queue work items on and the backend mechanism
-which manages thread-pool and processes the queued work items.
+which manages thread-pools and processes the queued work items.
 
 The backend is called gcwq.  There is one gcwq for each possible CPU
-and one gcwq to serve work items queued on unbound workqueues.
+and one gcwq to serve work items queued on unbound workqueues.  Each
+gcwq has two thread-pools - one for normal work items and the other
+for high priority ones.
 
 Subsystems and drivers can create and queue work items through special
 workqueue API functions as they see fit. They can influence some
 aspects of the way the work items are executed by setting flags on the
 workqueue they are putting the work item on. These flags include
-things like CPU locality, reentrancy, concurrency limits and more. To
-get a detailed overview refer to the API description of
+things like CPU locality, reentrancy, concurrency limits, priority and
+more.  To get a detailed overview refer to the API description of
 alloc_workqueue() below.
 
-When a work item is queued to a workqueue, the target gcwq is
-determined according to the queue parameters and workqueue attributes
-and appended on the shared worklist of the gcwq.  For example, unless
-specifically overridden, a work item of a bound workqueue will be
-queued on the worklist of exactly that gcwq that is associated to the
-CPU the issuer is running on.
+When a work item is queued to a workqueue, the target gcwq and
+thread-pool is determined according to the queue parameters and
+workqueue attributes and appended on the shared worklist of the
+thread-pool.  For example, unless specifically overridden, a work item
+of a bound workqueue will be queued on the worklist of either normal
+or highpri thread-pool of the gcwq that is associated to the CPU the
+issuer is running on.
 
 For any worker pool implementation, managing the concurrency level
 (how many execution contexts are active) is an important issue.  cmwq
@@ -115,26 +118,26 @@ tries to keep the concurrency at a minimal but sufficient level.
 Minimal to save resources and sufficient in that the system is used at
 its full capacity.
 
-Each gcwq bound to an actual CPU implements concurrency management by
-hooking into the scheduler.  The gcwq is notified whenever an active
-worker wakes up or sleeps and keeps track of the number of the
-currently runnable workers.  Generally, work items are not expected to
-hog a CPU and consume many cycles.  That means maintaining just enough
-concurrency to prevent work processing from stalling should be
-optimal.  As long as there are one or more runnable workers on the
-CPU, the gcwq doesn't start execution of a new work, but, when the
-last running worker goes to sleep, it immediately schedules a new
-worker so that the CPU doesn't sit idle while there are pending work
-items.  This allows using a minimal number of workers without losing
-execution bandwidth.
+Each thread-pool bound to an actual CPU implements concurrency
+management by hooking into the scheduler.  The thread-pool is notified
+whenever an active worker wakes up or sleeps and keeps track of the
+number of the currently runnable workers.  Generally, work items are
+not expected to hog a CPU and consume many cycles.  That means
+maintaining just enough concurrency to prevent work processing from
+stalling should be optimal.  As long as there are one or more runnable
+workers on the CPU, the thread-pool doesn't start execution of a new
+work, but, when the last running worker goes to sleep, it immediately
+schedules a new worker so that the CPU doesn't sit idle while there
+are pending work items.  This allows using a minimal number of workers
+without losing execution bandwidth.
 
 Keeping idle workers around doesn't cost other than the memory space
 for kthreads, so cmwq holds onto idle ones for a while before killing
 them.
 
 For an unbound wq, the above concurrency management doesn't apply and
-the gcwq for the pseudo unbound CPU tries to start executing all work
-items as soon as possible.  The responsibility of regulating
+the thread-pools for the pseudo unbound CPU try to start executing all
+work items as soon as possible.  The responsibility of regulating
 concurrency level is on the users.  There is also a flag to mark a
 bound wq to ignore the concurrency management.  Please refer to the
 API section for details.
@@ -205,31 +208,22 @@ resources, scheduled and executed.
 
   WQ_HIGHPRI
 
-	Work items of a highpri wq are queued at the head of the
-	worklist of the target gcwq and start execution regardless of
-	the current concurrency level.  In other words, highpri work
-	items will always start execution as soon as execution
-	resource is available.
+	Work items of a highpri wq are queued to the highpri
+	thread-pool of the target gcwq.  Highpri thread-pools are
+	served by worker threads with elevated nice level.
 
-	Ordering among highpri work items is preserved - a highpri
-	work item queued after another highpri work item will start
-	execution after the earlier highpri work item starts.
-
-	Although highpri work items are not held back by other
-	runnable work items, they still contribute to the concurrency
-	level.  Highpri work items in runnable state will prevent
-	non-highpri work items from starting execution.
-
-	This flag is meaningless for unbound wq.
+	Note that normal and highpri thread-pools don't interact with
+	each other.  Each maintain its separate pool of workers and
+	implements concurrency management among its workers.
 
   WQ_CPU_INTENSIVE
 
 	Work items of a CPU intensive wq do not contribute to the
 	concurrency level.  In other words, runnable CPU intensive
-	work items will not prevent other work items from starting
-	execution.  This is useful for bound work items which are
-	expected to hog CPU cycles so that their execution is
-	regulated by the system scheduler.
+	work items will not prevent other work items in the same
+	thread-pool from starting execution.  This is useful for bound
+	work items which are expected to hog CPU cycles so that their
+	execution is regulated by the system scheduler.
 
 	Although CPU intensive work items don't contribute to the
 	concurrency level, start of their executions is still
@@ -239,14 +233,6 @@ resources, scheduled and executed.
 
 	This flag is meaningless for unbound wq.
 
-  WQ_HIGHPRI | WQ_CPU_INTENSIVE
-
-	This combination makes the wq avoid interaction with
-	concurrency management completely and behave as a simple
-	per-CPU execution context provider.  Work items queued on a
-	highpri CPU-intensive wq start execution as soon as resources
-	are available and don't affect execution of other work items.
-
 @max_active:
 
 @max_active determines the maximum number of execution contexts per
@@ -328,20 +314,7 @@ If @max_active == 2,
  35		w2 wakes up and finishes
 
 Now, let's assume w1 and w2 are queued to a different wq q1 which has
-WQ_HIGHPRI set,
-
- TIME IN MSECS	EVENT
- 0		w1 and w2 start and burn CPU
- 5		w1 sleeps
- 10		w2 sleeps
- 10		w0 starts and burns CPU
- 15		w0 sleeps
- 15		w1 wakes up and finishes
- 20		w2 wakes up and finishes
- 25		w0 wakes up and burns CPU
- 30		w0 finishes
-
-If q1 has WQ_CPU_INTENSIVE set,
+WQ_CPU_INTENSIVE set,
 
  TIME IN MSECS	EVENT
  0		w0 starts and burns CPU

include/linux/cpu.h

@@ -73,8 +73,9 @@ enum {
 	/* migration should happen before other stuff but after perf */
 	CPU_PRI_PERF		= 20,
 	CPU_PRI_MIGRATION	= 10,
-	/* prepare workqueues for other notifiers */
-	CPU_PRI_WORKQUEUE	= 5,
+	/* bring up workqueues before normal notifiers and down after */
+	CPU_PRI_WORKQUEUE_UP	= 5,
+	CPU_PRI_WORKQUEUE_DOWN	= -5,
 };
 
 #define CPU_ONLINE		0x0002 /* CPU (unsigned)v is up */

include/linux/kthread.h

@@ -49,8 +49,6 @@ extern int tsk_fork_get_node(struct task_struct *tsk);
  * can be queued and flushed using queue/flush_kthread_work()
  * respectively.  Queued kthread_works are processed by a kthread
  * running kthread_worker_fn().
- *
- * A kthread_work can't be freed while it is executing.
  */
 struct kthread_work;
 typedef void (*kthread_work_func_t)(struct kthread_work *work);
@@ -59,15 +57,14 @@ struct kthread_worker {
 	spinlock_t		lock;
 	struct list_head	work_list;
 	struct task_struct	*task;
+	struct kthread_work	*current_work;
 };
 
 struct kthread_work {
 	struct list_head	node;
 	kthread_work_func_t	func;
 	wait_queue_head_t	done;
-	atomic_t		flushing;
-	int			queue_seq;
-	int			done_seq;
+	struct kthread_worker	*worker;
 };
 
 #define KTHREAD_WORKER_INIT(worker)	{				\
@@ -79,7 +76,6 @@ struct kthread_work {
 	.node = LIST_HEAD_INIT((work).node),				\
 	.func = (fn),							\
 	.done = __WAIT_QUEUE_HEAD_INITIALIZER((work).done),		\
-	.flushing = ATOMIC_INIT(0),					\
 	}
 
 #define DEFINE_KTHREAD_WORKER(worker)					\

include/trace/events/workqueue.h

@@ -54,7 +54,7 @@ TRACE_EVENT(workqueue_queue_work,
 		__entry->function	= work->func;
 		__entry->workqueue	= cwq->wq;
 		__entry->req_cpu	= req_cpu;
-		__entry->cpu		= cwq->gcwq->cpu;
+		__entry->cpu		= cwq->pool->gcwq->cpu;
 	),
 
 	TP_printk("work struct=%p function=%pf workqueue=%p req_cpu=%u cpu=%u",

kernel/kthread.c

@@ -360,16 +360,12 @@ int kthread_worker_fn(void *worker_ptr)
 					struct kthread_work, node);
 		list_del_init(&work->node);
 	}
+	worker->current_work = work;
 	spin_unlock_irq(&worker->lock);
 
 	if (work) {
 		__set_current_state(TASK_RUNNING);
 		work->func(work);
-		smp_wmb();	/* wmb worker-b0 paired with flush-b1 */
-		work->done_seq = work->queue_seq;
-		smp_mb();	/* mb worker-b1 paired with flush-b0 */
-		if (atomic_read(&work->flushing))
-			wake_up_all(&work->done);
 	} else if (!freezing(current))
 		schedule();
 
@@ -378,6 +374,19 @@ int kthread_worker_fn(void *worker_ptr)
 }
 EXPORT_SYMBOL_GPL(kthread_worker_fn);
 
+/* insert @work before @pos in @worker */
+static void insert_kthread_work(struct kthread_worker *worker,
+			       struct kthread_work *work,
+			       struct list_head *pos)
+{
+	lockdep_assert_held(&worker->lock);
+
+	list_add_tail(&work->node, pos);
+	work->worker = worker;
+	if (likely(worker->task))
+		wake_up_process(worker->task);
+}
+
 /**
  * queue_kthread_work - queue a kthread_work
  * @worker: target kthread_worker
@@ -395,10 +404,7 @@ bool queue_kthread_work(struct kthread_worker *worker,
 
 	spin_lock_irqsave(&worker->lock, flags);
 	if (list_empty(&work->node)) {
-		list_add_tail(&work->node, &worker->work_list);
-		work->queue_seq++;
-		if (likely(worker->task))
-			wake_up_process(worker->task);
+		insert_kthread_work(worker, work, &worker->work_list);
 		ret = true;
 	}
 	spin_unlock_irqrestore(&worker->lock, flags);
@@ -406,6 +412,18 @@ bool queue_kthread_work(struct kthread_worker *worker,
 }
 EXPORT_SYMBOL_GPL(queue_kthread_work);
 
+struct kthread_flush_work {
+	struct kthread_work	work;
+	struct completion	done;
+};
+
+static void kthread_flush_work_fn(struct kthread_work *work)
+{
+	struct kthread_flush_work *fwork =
+		container_of(work, struct kthread_flush_work, work);
+	complete(&fwork->done);
+}
+
 /**
  * flush_kthread_work - flush a kthread_work
  * @work: work to flush
@@ -414,39 +432,37 @@ EXPORT_SYMBOL_GPL(queue_kthread_work);
  */
 void flush_kthread_work(struct kthread_work *work)
 {
-	int seq = work->queue_seq;
-
-	atomic_inc(&work->flushing);
+	struct kthread_flush_work fwork = {
+		KTHREAD_WORK_INIT(fwork.work, kthread_flush_work_fn),
+		COMPLETION_INITIALIZER_ONSTACK(fwork.done),
+	};
+	struct kthread_worker *worker;
+	bool noop = false;
 
-	/*
-	 * mb flush-b0 paired with worker-b1, to make sure either
-	 * worker sees the above increment or we see done_seq update.
-	 */
-	smp_mb__after_atomic_inc();
+retry:
+	worker = work->worker;
+	if (!worker)
+		return;
 
-	/* A - B <= 0 tests whether B is in front of A regardless of overflow */
-	wait_event(work->done, seq - work->done_seq <= 0);
-	atomic_dec(&work->flushing);
+	spin_lock_irq(&worker->lock);
+	if (work->worker != worker) {
+		spin_unlock_irq(&worker->lock);
+		goto retry;
+	}
 
-	/*
-	 * rmb flush-b1 paired with worker-b0, to make sure our caller
-	 * sees every change made by work->func().
-	 */
-	smp_mb__after_atomic_dec();
-}
-EXPORT_SYMBOL_GPL(flush_kthread_work);
+	if (!list_empty(&work->node))
+		insert_kthread_work(worker, &fwork.work, work->node.next);
+	else if (worker->current_work == work)
+		insert_kthread_work(worker, &fwork.work, worker->work_list.next);
+	else
+		noop = true;
 
-struct kthread_flush_work {
-	struct kthread_work	work;
-	struct completion	done;
-};
+	spin_unlock_irq(&worker->lock);
 
-static void kthread_flush_work_fn(struct kthread_work *work)
-{
-	struct kthread_flush_work *fwork =
-		container_of(work, struct kthread_flush_work, work);
-	complete(&fwork->done);
+	if (!noop)
+		wait_for_completion(&fwork.done);
 }
+EXPORT_SYMBOL_GPL(flush_kthread_work);
 
 /**
  * flush_kthread_worker - flush all current works on a kthread_worker

kernel/workqueue.c

@@ -45,32 +45,41 @@
 #include "workqueue_sched.h"
 
 enum {
-	/* global_cwq flags */
-	GCWQ_MANAGE_WORKERS	= 1 << 0,	/* need to manage workers */
-	GCWQ_MANAGING_WORKERS	= 1 << 1,	/* managing workers */
-	GCWQ_DISASSOCIATED	= 1 << 2,	/* cpu can't serve workers */
-	GCWQ_FREEZING		= 1 << 3,	/* freeze in progress */
-	GCWQ_HIGHPRI_PENDING	= 1 << 4,	/* highpri works on queue */
+	/*
+	 * global_cwq flags
+	 *
+	 * A bound gcwq is either associated or disassociated with its CPU.
+	 * While associated (!DISASSOCIATED), all workers are bound to the
+	 * CPU and none has %WORKER_UNBOUND set and concurrency management
+	 * is in effect.
+	 *
+	 * While DISASSOCIATED, the cpu may be offline and all workers have
+	 * %WORKER_UNBOUND set and concurrency management disabled, and may
+	 * be executing on any CPU.  The gcwq behaves as an unbound one.
+	 *
+	 * Note that DISASSOCIATED can be flipped only while holding
+	 * managership of all pools on the gcwq to avoid changing binding
+	 * state while create_worker() is in progress.
+	 */
+	GCWQ_DISASSOCIATED	= 1 << 0,	/* cpu can't serve workers */
+	GCWQ_FREEZING		= 1 << 1,	/* freeze in progress */
+
+	/* pool flags */
+	POOL_MANAGE_WORKERS	= 1 << 0,	/* need to manage workers */
 
 	/* worker flags */
 	WORKER_STARTED		= 1 << 0,	/* started */
 	WORKER_DIE		= 1 << 1,	/* die die die */
 	WORKER_IDLE		= 1 << 2,	/* is idle */
 	WORKER_PREP		= 1 << 3,	/* preparing to run works */
-	WORKER_ROGUE		= 1 << 4,	/* not bound to any cpu */
 	WORKER_REBIND		= 1 << 5,	/* mom is home, come back */
 	WORKER_CPU_INTENSIVE	= 1 << 6,	/* cpu intensive */
 	WORKER_UNBOUND		= 1 << 7,	/* worker is unbound */
 
-	WORKER_NOT_RUNNING	= WORKER_PREP | WORKER_ROGUE | WORKER_REBIND |
-				  WORKER_CPU_INTENSIVE | WORKER_UNBOUND,
+	WORKER_NOT_RUNNING	= WORKER_PREP | WORKER_REBIND | WORKER_UNBOUND |
+				  WORKER_CPU_INTENSIVE,
 
-	/* gcwq->trustee_state */
-	TRUSTEE_START		= 0,		/* start */
-	TRUSTEE_IN_CHARGE	= 1,		/* trustee in charge of gcwq */
-	TRUSTEE_BUTCHER		= 2,		/* butcher workers */
-	TRUSTEE_RELEASE		= 3,		/* release workers */
-	TRUSTEE_DONE		= 4,		/* trustee is done */
+	NR_WORKER_POOLS		= 2,		/* # worker pools per gcwq */
 
 	BUSY_WORKER_HASH_ORDER	= 6,		/* 64 pointers */
 	BUSY_WORKER_HASH_SIZE	= 1 << BUSY_WORKER_HASH_ORDER,
@@ -84,13 +93,13 @@ enum {
 						   (min two ticks) */
 	MAYDAY_INTERVAL		= HZ / 10,	/* and then every 100ms */
 	CREATE_COOLDOWN		= HZ,		/* time to breath after fail */
-	TRUSTEE_COOLDOWN	= HZ / 10,	/* for trustee draining */
 
 	/*
 	 * Rescue workers are used only on emergencies and shared by
 	 * all cpus.  Give -20.
 	 */
 	RESCUER_NICE_LEVEL	= -20,
+	HIGHPRI_NICE_LEVEL	= -20,
 };
 
 /*
@@ -115,6 +124,8 @@ enum {
  */
 
 struct global_cwq;
+struct worker_pool;
+struct idle_rebind;
 
 /*
  * The poor guys doing the actual heavy lifting.  All on-duty workers
@@ -131,12 +142,31 @@ struct worker {
 	struct cpu_workqueue_struct *current_cwq; /* L: current_work's cwq */
 	struct list_head	scheduled;	/* L: scheduled works */
 	struct task_struct	*task;		/* I: worker task */
-	struct global_cwq	*gcwq;		/* I: the associated gcwq */
+	struct worker_pool	*pool;		/* I: the associated pool */
 	/* 64 bytes boundary on 64bit, 32 on 32bit */
 	unsigned long		last_active;	/* L: last active timestamp */
 	unsigned int		flags;		/* X: flags */
 	int			id;		/* I: worker id */
-	struct work_struct	rebind_work;	/* L: rebind worker to cpu */
+
+	/* for rebinding worker to CPU */
+	struct idle_rebind	*idle_rebind;	/* L: for idle worker */
+	struct work_struct	rebind_work;	/* L: for busy worker */
+};
+
+struct worker_pool {
+	struct global_cwq	*gcwq;		/* I: the owning gcwq */
+	unsigned int		flags;		/* X: flags */
+
+	struct list_head	worklist;	/* L: list of pending works */
+	int			nr_workers;	/* L: total number of workers */
+	int			nr_idle;	/* L: currently idle ones */
+
+	struct list_head	idle_list;	/* X: list of idle workers */
+	struct timer_list	idle_timer;	/* L: worker idle timeout */
+	struct timer_list	mayday_timer;	/* L: SOS timer for workers */
+
+	struct mutex		manager_mutex;	/* mutex manager should hold */
+	struct ida		worker_ida;	/* L: for worker IDs */
 };
 
 /*
@@ -146,27 +176,16 @@ struct worker {
  */
 struct global_cwq {
 	spinlock_t		lock;		/* the gcwq lock */
-	struct list_head	worklist;	/* L: list of pending works */
 	unsigned int		cpu;		/* I: the associated cpu */
 	unsigned int		flags;		/* L: GCWQ_* flags */
 
-	int			nr_workers;	/* L: total number of workers */
-	int			nr_idle;	/* L: currently idle ones */
-
-	/* workers are chained either in the idle_list or busy_hash */
-	struct list_head	idle_list;	/* X: list of idle workers */
+	/* workers are chained either in busy_hash or pool idle_list */
 	struct hlist_head	busy_hash[BUSY_WORKER_HASH_SIZE];
 						/* L: hash of busy workers */
 
-	struct timer_list	idle_timer;	/* L: worker idle timeout */
-	struct timer_list	mayday_timer;	/* L: SOS timer for dworkers */
-
-	struct ida		worker_ida;	/* L: for worker IDs */
+	struct worker_pool	pools[2];	/* normal and highpri pools */
 
-	struct task_struct	*trustee;	/* L: for gcwq shutdown */
-	unsigned int		trustee_state;	/* L: trustee state */
-	wait_queue_head_t	trustee_wait;	/* trustee wait */
-	struct worker		*first_idle;	/* L: first idle worker */
+	wait_queue_head_t	rebind_hold;	/* rebind hold wait */
 } ____cacheline_aligned_in_smp;
 
 /*
@@ -175,7 +194,7 @@ struct global_cwq {
  * aligned at two's power of the number of flag bits.
  */
 struct cpu_workqueue_struct {
-	struct global_cwq	*gcwq;		/* I: the associated gcwq */
+	struct worker_pool	*pool;		/* I: the associated pool */
 	struct workqueue_struct *wq;		/* I: the owning workqueue */
 	int			work_color;	/* L: current color */
 	int			flush_color;	/* L: flushing color */
@@ -264,6 +283,10 @@ EXPORT_SYMBOL_GPL(system_nrt_freezable_wq);
 #define CREATE_TRACE_POINTS
 #include <trace/events/workqueue.h>
 
+#define for_each_worker_pool(pool, gcwq)				\
+	for ((pool) = &(gcwq)->pools[0];				\
+	     (pool) < &(gcwq)->pools[NR_WORKER_POOLS]; (pool)++)
+
 #define for_each_busy_worker(worker, i, pos, gcwq)			\
 	for (i = 0; i < BUSY_WORKER_HASH_SIZE; i++)			\
 		hlist_for_each_entry(worker, pos, &gcwq->busy_hash[i], hentry)
@@ -444,7 +467,7 @@ static bool workqueue_freezing;		/* W: have wqs started freezing? */
  * try_to_wake_up().  Put it in a separate cacheline.
  */
 static DEFINE_PER_CPU(struct global_cwq, global_cwq);
-static DEFINE_PER_CPU_SHARED_ALIGNED(atomic_t, gcwq_nr_running);
+static DEFINE_PER_CPU_SHARED_ALIGNED(atomic_t, pool_nr_running[NR_WORKER_POOLS]);
 
 /*
  * Global cpu workqueue and nr_running counter for unbound gcwq.  The
@@ -452,10 +475,17 @@ static DEFINE_PER_CPU_SHARED_ALIGNED(atomic_t, gcwq_nr_running);
  * workers have WORKER_UNBOUND set.
  */
 static struct global_cwq unbound_global_cwq;
-static atomic_t unbound_gcwq_nr_running = ATOMIC_INIT(0);	/* always 0 */
+static atomic_t unbound_pool_nr_running[NR_WORKER_POOLS] = {
+	[0 ... NR_WORKER_POOLS - 1]	= ATOMIC_INIT(0),	/* always 0 */
+};
 
 static int worker_thread(void *__worker);
 
+static int worker_pool_pri(struct worker_pool *pool)
+{
+	return pool - pool->gcwq->pools;
+}
+
 static struct global_cwq *get_gcwq(unsigned int cpu)
 {
 	if (cpu != WORK_CPU_UNBOUND)
@@ -464,12 +494,15 @@ static struct global_cwq *get_gcwq(unsigned int cpu)
 		return &unbound_global_cwq;
 }
 
-static atomic_t *get_gcwq_nr_running(unsigned int cpu)
+static atomic_t *get_pool_nr_running(struct worker_pool *pool)
 {
+	int cpu = pool->gcwq->cpu;
+	int idx = worker_pool_pri(pool);
+
 	if (cpu != WORK_CPU_UNBOUND)
-		return &per_cpu(gcwq_nr_running, cpu);
+		return &per_cpu(pool_nr_running, cpu)[idx];
 	else
-		return &unbound_gcwq_nr_running;
+		return &unbound_pool_nr_running[idx];
 }
 
 static struct cpu_workqueue_struct *get_cwq(unsigned int cpu,
@@ -555,7 +588,7 @@ static struct global_cwq *get_work_gcwq(struct work_struct *work)
 
 	if (data & WORK_STRUCT_CWQ)
 		return ((struct cpu_workqueue_struct *)
-			(data & WORK_STRUCT_WQ_DATA_MASK))->gcwq;
+			(data & WORK_STRUCT_WQ_DATA_MASK))->pool->gcwq;
 
 	cpu = data >> WORK_STRUCT_FLAG_BITS;
 	if (cpu == WORK_CPU_NONE)
@@ -566,60 +599,62 @@ static struct global_cwq *get_work_gcwq(struct work_struct *work)
 }
 
 /*
- * Policy functions.  These define the policies on how the global
- * worker pool is managed.  Unless noted otherwise, these functions
- * assume that they're being called with gcwq->lock held.
+ * Policy functions.  These define the policies on how the global worker
+ * pools are managed.  Unless noted otherwise, these functions assume that
+ * they're being called with gcwq->lock held.
  */
 
-static bool __need_more_worker(struct global_cwq *gcwq)
+static bool __need_more_worker(struct worker_pool *pool)
 {
-	return !atomic_read(get_gcwq_nr_running(gcwq->cpu)) ||
-		gcwq->flags & GCWQ_HIGHPRI_PENDING;
+	return !atomic_read(get_pool_nr_running(pool));
 }
 
 /*
  * Need to wake up a worker?  Called from anything but currently
  * running workers.
+ *
+ * Note that, because unbound workers never contribute to nr_running, this
+ * function will always return %true for unbound gcwq as long as the
+ * worklist isn't empty.
  */
-static bool need_more_worker(struct global_cwq *gcwq)
+static bool need_more_worker(struct worker_pool *pool)
 {
-	return !list_empty(&gcwq->worklist) && __need_more_worker(gcwq);
+	return !list_empty(&pool->worklist) && __need_more_worker(pool);
 }
 
 /* Can I start working?  Called from busy but !running workers. */
-static bool may_start_working(struct global_cwq *gcwq)
+static bool may_start_working(struct worker_pool *pool)
 {
-	return gcwq->nr_idle;
+	return pool->nr_idle;
 }
 
 /* Do I need to keep working?  Called from currently running workers. */
-static bool keep_working(struct global_cwq *gcwq)
+static bool keep_working(struct worker_pool *pool)
 {
-	atomic_t *nr_running = get_gcwq_nr_running(gcwq->cpu);
+	atomic_t *nr_running = get_pool_nr_running(pool);
 
-	return !list_empty(&gcwq->worklist) &&
-		(atomic_read(nr_running) <= 1 ||
-		 gcwq->flags & GCWQ_HIGHPRI_PENDING);
+	return !list_empty(&pool->worklist) && atomic_read(nr_running) <= 1;
 }
 
 /* Do we need a new worker?  Called from manager. */
-static bool need_to_create_worker(struct global_cwq *gcwq)
+static bool need_to_create_worker(struct worker_pool *pool)
 {
-	return need_more_worker(gcwq) && !may_start_working(gcwq);
+	return need_more_worker(pool) && !may_start_working(pool);
 }
 
 /* Do I need to be the manager? */
-static bool need_to_manage_workers(struct global_cwq *gcwq)
+static bool need_to_manage_workers(struct worker_pool *pool)
 {
-	return need_to_create_worker(gcwq) || gcwq->flags & GCWQ_MANAGE_WORKERS;
+	return need_to_create_worker(pool) ||
+		(pool->flags & POOL_MANAGE_WORKERS);
 }
 
 /* Do we have too many workers and should some go away? */
-static bool too_many_workers(struct global_cwq *gcwq)
+static bool too_many_workers(struct worker_pool *pool)
 {
-	bool managing = gcwq->flags & GCWQ_MANAGING_WORKERS;
-	int nr_idle = gcwq->nr_idle + managing; /* manager is considered idle */
-	int nr_busy = gcwq->nr_workers - nr_idle;
+	bool managing = mutex_is_locked(&pool->manager_mutex);
+	int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
+	int nr_busy = pool->nr_workers - nr_idle;
 
 	return nr_idle > 2 && (nr_idle - 2) * MAX_IDLE_WORKERS_RATIO >= nr_busy;
 }
@@ -629,26 +664,26 @@ static bool too_many_workers(struct global_cwq *gcwq)
  */
 
 /* Return the first worker.  Safe with preemption disabled */
-static struct worker *first_worker(struct global_cwq *gcwq)
+static struct worker *first_worker(struct worker_pool *pool)
 {
-	if (unlikely(list_empty(&gcwq->idle_list)))
+	if (unlikely(list_empty(&pool->idle_list)))
 		return NULL;
 
-	return list_first_entry(&gcwq->idle_list, struct worker, entry);
+	return list_first_entry(&pool->idle_list, struct worker, entry);
 }
 
 /**
  * wake_up_worker - wake up an idle worker
- * @gcwq: gcwq to wake worker for
+ * @pool: worker pool to wake worker from
  *
- * Wake up the first idle worker of @gcwq.
+ * Wake up the first idle worker of @pool.
  *
  * CONTEXT:
  * spin_lock_irq(gcwq->lock).
  */
-static void wake_up_worker(struct global_cwq *gcwq)
+static void wake_up_worker(struct worker_pool *pool)
 {
-	struct worker *worker = first_worker(gcwq);
+	struct worker *worker = first_worker(pool);
 
 	if (likely(worker))
 		wake_up_process(worker->task);
@@ -670,7 +705,7 @@ void wq_worker_waking_up(struct task_struct *task, unsigned int cpu)
 	struct worker *worker = kthread_data(task);
 
 	if (!(worker->flags & WORKER_NOT_RUNNING))
-		atomic_inc(get_gcwq_nr_running(cpu));
+		atomic_inc(get_pool_nr_running(worker->pool));
 }
 
 /**
@@ -692,8 +727,8 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task,
 				       unsigned int cpu)
 {
 	struct worker *worker = kthread_data(task), *to_wakeup = NULL;
-	struct global_cwq *gcwq = get_gcwq(cpu);
-	atomic_t *nr_running = get_gcwq_nr_running(cpu);
+	struct worker_pool *pool = worker->pool;
+	atomic_t *nr_running = get_pool_nr_running(pool);
 
 	if (worker->flags & WORKER_NOT_RUNNING)
 		return NULL;
@@ -706,14 +741,14 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task,
 	 * worklist not empty test sequence is in insert_work().
 	 * Please read comment there.
 	 *
-	 * NOT_RUNNING is clear.  This means that trustee is not in
-	 * charge and we're running on the local cpu w/ rq lock held
-	 * and preemption disabled, which in turn means that none else
-	 * could be manipulating idle_list, so dereferencing idle_list
-	 * without gcwq lock is safe.
+	 * NOT_RUNNING is clear.  This means that we're bound to and
+	 * running on the local cpu w/ rq lock held and preemption
+	 * disabled, which in turn means that none else could be
+	 * manipulating idle_list, so dereferencing idle_list without gcwq
+	 * lock is safe.
 	 */
-	if (atomic_dec_and_test(nr_running) && !list_empty(&gcwq->worklist))
-		to_wakeup = first_worker(gcwq);
+	if (atomic_dec_and_test(nr_running) && !list_empty(&pool->worklist))
+		to_wakeup = first_worker(pool);
 	return to_wakeup ? to_wakeup->task : NULL;
 }
 
@@ -733,7 +768,7 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task,
 static inline void worker_set_flags(struct worker *worker, unsigned int flags,
 				    bool wakeup)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
 
 	WARN_ON_ONCE(worker->task != current);
 
@@ -744,12 +779,12 @@ static inline void worker_set_flags(struct worker *worker, unsigned int flags,
 	 */
 	if ((flags & WORKER_NOT_RUNNING) &&
 	    !(worker->flags & WORKER_NOT_RUNNING)) {
-		atomic_t *nr_running = get_gcwq_nr_running(gcwq->cpu);
+		atomic_t *nr_running = get_pool_nr_running(pool);
 
 		if (wakeup) {
 			if (atomic_dec_and_test(nr_running) &&
-			    !list_empty(&gcwq->worklist))
-				wake_up_worker(gcwq);
+			    !list_empty(&pool->worklist))
+				wake_up_worker(pool);
 		} else
 			atomic_dec(nr_running);
 	}
@@ -769,7 +804,7 @@ static inline void worker_set_flags(struct worker *worker, unsigned int flags,
  */
 static inline void worker_clr_flags(struct worker *worker, unsigned int flags)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
 	unsigned int oflags = worker->flags;
 
 	WARN_ON_ONCE(worker->task != current);
@@ -783,7 +818,7 @@ static inline void worker_clr_flags(struct worker *worker, unsigned int flags)
 	 */
 	if ((flags & WORKER_NOT_RUNNING) && (oflags & WORKER_NOT_RUNNING))
 		if (!(worker->flags & WORKER_NOT_RUNNING))
-			atomic_inc(get_gcwq_nr_running(gcwq->cpu));
+			atomic_inc(get_pool_nr_running(pool));
 }
 
 /**
@@ -866,43 +901,6 @@ static struct worker *find_worker_executing_work(struct global_cwq *gcwq,
 					    work);
 }
 
-/**
- * gcwq_determine_ins_pos - find insertion position
- * @gcwq: gcwq of interest
- * @cwq: cwq a work is being queued for
- *
- * A work for @cwq is about to be queued on @gcwq, determine insertion
- * position for the work.  If @cwq is for HIGHPRI wq, the work is
- * queued at the head of the queue but in FIFO order with respect to
- * other HIGHPRI works; otherwise, at the end of the queue.  This
- * function also sets GCWQ_HIGHPRI_PENDING flag to hint @gcwq that
- * there are HIGHPRI works pending.
- *
- * CONTEXT:
- * spin_lock_irq(gcwq->lock).
- *
- * RETURNS:
- * Pointer to inserstion position.
- */
-static inline struct list_head *gcwq_determine_ins_pos(struct global_cwq *gcwq,
-					       struct cpu_workqueue_struct *cwq)
-{
-	struct work_struct *twork;
-
-	if (likely(!(cwq->wq->flags & WQ_HIGHPRI)))
-		return &gcwq->worklist;
-
-	list_for_each_entry(twork, &gcwq->worklist, entry) {
-		struct cpu_workqueue_struct *tcwq = get_work_cwq(twork);
-
-		if (!(tcwq->wq->flags & WQ_HIGHPRI))
-			break;
-	}
-
-	gcwq->flags |= GCWQ_HIGHPRI_PENDING;
-	return &twork->entry;
-}
-
 /**
  * insert_work - insert a work into gcwq
  * @cwq: cwq @work belongs to
@@ -920,7 +918,7 @@ static void insert_work(struct cpu_workqueue_struct *cwq,
 			struct work_struct *work, struct list_head *head,
 			unsigned int extra_flags)
 {
-	struct global_cwq *gcwq = cwq->gcwq;
+	struct worker_pool *pool = cwq->pool;
 
 	/* we own @work, set data and link */
 	set_work_cwq(work, cwq, extra_flags);
@@ -940,8 +938,8 @@ static void insert_work(struct cpu_workqueue_struct *cwq,
 	 */
 	smp_mb();
 
-	if (__need_more_worker(gcwq))
-		wake_up_worker(gcwq);
+	if (__need_more_worker(pool))
+		wake_up_worker(pool);
 }
 
 /*
@@ -1043,7 +1041,7 @@ static void __queue_work(unsigned int cpu, struct workqueue_struct *wq,
 	if (likely(cwq->nr_active < cwq->max_active)) {
 		trace_workqueue_activate_work(work);
 		cwq->nr_active++;
-		worklist = gcwq_determine_ins_pos(gcwq, cwq);
+		worklist = &cwq->pool->worklist;
 	} else {
 		work_flags |= WORK_STRUCT_DELAYED;
 		worklist = &cwq->delayed_works;
@@ -1192,7 +1190,8 @@ EXPORT_SYMBOL_GPL(queue_delayed_work_on);
  */
 static void worker_enter_idle(struct worker *worker)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
+	struct global_cwq *gcwq = pool->gcwq;
 
 	BUG_ON(worker->flags & WORKER_IDLE);
 	BUG_ON(!list_empty(&worker->entry) &&
@@ -1200,27 +1199,24 @@ static void worker_enter_idle(struct worker *worker)
 
 	/* can't use worker_set_flags(), also called from start_worker() */
 	worker->flags |= WORKER_IDLE;
-	gcwq->nr_idle++;
+	pool->nr_idle++;
 	worker->last_active = jiffies;
 
 	/* idle_list is LIFO */
-	list_add(&worker->entry, &gcwq->idle_list);
+	list_add(&worker->entry, &pool->idle_list);
 
-	if (likely(!(worker->flags & WORKER_ROGUE))) {
-		if (too_many_workers(gcwq) && !timer_pending(&gcwq->idle_timer))
-			mod_timer(&gcwq->idle_timer,
-				  jiffies + IDLE_WORKER_TIMEOUT);
-	} else
-		wake_up_all(&gcwq->trustee_wait);
+	if (too_many_workers(pool) && !timer_pending(&pool->idle_timer))
+		mod_timer(&pool->idle_timer, jiffies + IDLE_WORKER_TIMEOUT);
 
 	/*
-	 * Sanity check nr_running.  Because trustee releases gcwq->lock
-	 * between setting %WORKER_ROGUE and zapping nr_running, the
-	 * warning may trigger spuriously.  Check iff trustee is idle.
+	 * Sanity check nr_running.  Because gcwq_unbind_fn() releases
+	 * gcwq->lock between setting %WORKER_UNBOUND and zapping
+	 * nr_running, the warning may trigger spuriously.  Check iff
+	 * unbind is not in progress.
 	 */
-	WARN_ON_ONCE(gcwq->trustee_state == TRUSTEE_DONE &&
-		     gcwq->nr_workers == gcwq->nr_idle &&
-		     atomic_read(get_gcwq_nr_running(gcwq->cpu)));
+	WARN_ON_ONCE(!(gcwq->flags & GCWQ_DISASSOCIATED) &&
+		     pool->nr_workers == pool->nr_idle &&
+		     atomic_read(get_pool_nr_running(pool)));
 }
 
 /**
@@ -1234,11 +1230,11 @@ static void worker_enter_idle(struct worker *worker)
  */
 static void worker_leave_idle(struct worker *worker)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
 
 	BUG_ON(!(worker->flags & WORKER_IDLE));
 	worker_clr_flags(worker, WORKER_IDLE);
-	gcwq->nr_idle--;
+	pool->nr_idle--;
 	list_del_init(&worker->entry);
 }
 
@@ -1258,11 +1254,11 @@ static void worker_leave_idle(struct worker *worker)
  * verbatim as it's best effort and blocking and gcwq may be
  * [dis]associated in the meantime.
  *
- * This function tries set_cpus_allowed() and locks gcwq and verifies
- * the binding against GCWQ_DISASSOCIATED which is set during
- * CPU_DYING and cleared during CPU_ONLINE, so if the worker enters
- * idle state or fetches works without dropping lock, it can guarantee
- * the scheduling requirement described in the first paragraph.
+ * This function tries set_cpus_allowed() and locks gcwq and verifies the
+ * binding against %GCWQ_DISASSOCIATED which is set during
+ * %CPU_DOWN_PREPARE and cleared during %CPU_ONLINE, so if the worker
+ * enters idle state or fetches works without dropping lock, it can
+ * guarantee the scheduling requirement described in the first paragraph.
  *
  * CONTEXT:
  * Might sleep.  Called without any lock but returns with gcwq->lock
@@ -1275,7 +1271,7 @@ static void worker_leave_idle(struct worker *worker)
 static bool worker_maybe_bind_and_lock(struct worker *worker)
 __acquires(&gcwq->lock)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct global_cwq *gcwq = worker->pool->gcwq;
 	struct task_struct *task = worker->task;
 
 	while (true) {
@@ -1308,16 +1304,40 @@ __acquires(&gcwq->lock)
 	}
 }
 
+struct idle_rebind {
+	int			cnt;		/* # workers to be rebound */
+	struct completion	done;		/* all workers rebound */
+};
+
 /*
- * Function for worker->rebind_work used to rebind rogue busy workers
- * to the associated cpu which is coming back online.  This is
- * scheduled by cpu up but can race with other cpu hotplug operations
- * and may be executed twice without intervening cpu down.
+ * Rebind an idle @worker to its CPU.  During CPU onlining, this has to
+ * happen synchronously for idle workers.  worker_thread() will test
+ * %WORKER_REBIND before leaving idle and call this function.
  */
-static void worker_rebind_fn(struct work_struct *work)
+static void idle_worker_rebind(struct worker *worker)
+{
+	struct global_cwq *gcwq = worker->pool->gcwq;
+
+	/* CPU must be online at this point */
+	WARN_ON(!worker_maybe_bind_and_lock(worker));
+	if (!--worker->idle_rebind->cnt)
+		complete(&worker->idle_rebind->done);
+	spin_unlock_irq(&worker->pool->gcwq->lock);
+
+	/* we did our part, wait for rebind_workers() to finish up */
+	wait_event(gcwq->rebind_hold, !(worker->flags & WORKER_REBIND));
+}
+
+/*
+ * Function for @worker->rebind.work used to rebind unbound busy workers to
+ * the associated cpu which is coming back online.  This is scheduled by
+ * cpu up but can race with other cpu hotplug operations and may be
+ * executed twice without intervening cpu down.
+ */
+static void busy_worker_rebind_fn(struct work_struct *work)
 {
 	struct worker *worker = container_of(work, struct worker, rebind_work);
-	struct global_cwq *gcwq = worker->gcwq;
+	struct global_cwq *gcwq = worker->pool->gcwq;
 
 	if (worker_maybe_bind_and_lock(worker))
 		worker_clr_flags(worker, WORKER_REBIND);
@@ -1325,6 +1345,112 @@ static void worker_rebind_fn(struct work_struct *work)
 	spin_unlock_irq(&gcwq->lock);
 }
 
+/**
+ * rebind_workers - rebind all workers of a gcwq to the associated CPU
+ * @gcwq: gcwq of interest
+ *
+ * @gcwq->cpu is coming online.  Rebind all workers to the CPU.  Rebinding
+ * is different for idle and busy ones.
+ *
+ * The idle ones should be rebound synchronously and idle rebinding should
+ * be complete before any worker starts executing work items with
+ * concurrency management enabled; otherwise, scheduler may oops trying to
+ * wake up non-local idle worker from wq_worker_sleeping().
+ *
+ * This is achieved by repeatedly requesting rebinding until all idle
+ * workers are known to have been rebound under @gcwq->lock and holding all
+ * idle workers from becoming busy until idle rebinding is complete.
+ *
+ * Once idle workers are rebound, busy workers can be rebound as they
+ * finish executing their current work items.  Queueing the rebind work at
+ * the head of their scheduled lists is enough.  Note that nr_running will
+ * be properbly bumped as busy workers rebind.
+ *
+ * On return, all workers are guaranteed to either be bound or have rebind
+ * work item scheduled.
+ */
+static void rebind_workers(struct global_cwq *gcwq)
+	__releases(&gcwq->lock) __acquires(&gcwq->lock)
+{
+	struct idle_rebind idle_rebind;
+	struct worker_pool *pool;
+	struct worker *worker;
+	struct hlist_node *pos;
+	int i;
+
+	lockdep_assert_held(&gcwq->lock);
+
+	for_each_worker_pool(pool, gcwq)
+		lockdep_assert_held(&pool->manager_mutex);
+
+	/*
+	 * Rebind idle workers.  Interlocked both ways.  We wait for
+	 * workers to rebind via @idle_rebind.done.  Workers will wait for
+	 * us to finish up by watching %WORKER_REBIND.
+	 */
+	init_completion(&idle_rebind.done);
+retry:
+	idle_rebind.cnt = 1;
+	INIT_COMPLETION(idle_rebind.done);
+
+	/* set REBIND and kick idle ones, we'll wait for these later */
+	for_each_worker_pool(pool, gcwq) {
+		list_for_each_entry(worker, &pool->idle_list, entry) {
+			if (worker->flags & WORKER_REBIND)
+				continue;
+
+			/* morph UNBOUND to REBIND */
+			worker->flags &= ~WORKER_UNBOUND;
+			worker->flags |= WORKER_REBIND;
+
+			idle_rebind.cnt++;
+			worker->idle_rebind = &idle_rebind;
+
+			/* worker_thread() will call idle_worker_rebind() */
+			wake_up_process(worker->task);
+		}
+	}
+
+	if (--idle_rebind.cnt) {
+		spin_unlock_irq(&gcwq->lock);
+		wait_for_completion(&idle_rebind.done);
+		spin_lock_irq(&gcwq->lock);
+		/* busy ones might have become idle while waiting, retry */
+		goto retry;
+	}
+
+	/*
+	 * All idle workers are rebound and waiting for %WORKER_REBIND to
+	 * be cleared inside idle_worker_rebind().  Clear and release.
+	 * Clearing %WORKER_REBIND from this foreign context is safe
+	 * because these workers are still guaranteed to be idle.
+	 */
+	for_each_worker_pool(pool, gcwq)
+		list_for_each_entry(worker, &pool->idle_list, entry)
+			worker->flags &= ~WORKER_REBIND;
+
+	wake_up_all(&gcwq->rebind_hold);
+
+	/* rebind busy workers */
+	for_each_busy_worker(worker, i, pos, gcwq) {
+		struct work_struct *rebind_work = &worker->rebind_work;
+
+		/* morph UNBOUND to REBIND */
+		worker->flags &= ~WORKER_UNBOUND;
+		worker->flags |= WORKER_REBIND;
+
+		if (test_and_set_bit(WORK_STRUCT_PENDING_BIT,
+				     work_data_bits(rebind_work)))
+			continue;
+
+		/* wq doesn't matter, use the default one */
+		debug_work_activate(rebind_work);
+		insert_work(get_cwq(gcwq->cpu, system_wq), rebind_work,
+			    worker->scheduled.next,
+			    work_color_to_flags(WORK_NO_COLOR));
+	}
+}
+
 static struct worker *alloc_worker(void)
 {
 	struct worker *worker;
@@ -1333,7 +1459,7 @@ static struct worker *alloc_worker(void)
 	if (worker) {
 		INIT_LIST_HEAD(&worker->entry);
 		INIT_LIST_HEAD(&worker->scheduled);
-		INIT_WORK(&worker->rebind_work, worker_rebind_fn);
+		INIT_WORK(&worker->rebind_work, busy_worker_rebind_fn);
 		/* on creation a worker is in !idle && prep state */
 		worker->flags = WORKER_PREP;
 	}
@@ -1342,10 +1468,9 @@ static struct worker *alloc_worker(void)
 
 /**
  * create_worker - create a new workqueue worker
- * @gcwq: gcwq the new worker will belong to
- * @bind: whether to set affinity to @cpu or not
+ * @pool: pool the new worker will belong to
  *
- * Create a new worker which is bound to @gcwq.  The returned worker
+ * Create a new worker which is bound to @pool.  The returned worker
  * can be started by calling start_worker() or destroyed using
  * destroy_worker().
  *
@@ -1355,16 +1480,17 @@ static struct worker *alloc_worker(void)
  * RETURNS:
  * Pointer to the newly created worker.
  */
-static struct worker *create_worker(struct global_cwq *gcwq, bool bind)
+static struct worker *create_worker(struct worker_pool *pool)
 {
-	bool on_unbound_cpu = gcwq->cpu == WORK_CPU_UNBOUND;
+	struct global_cwq *gcwq = pool->gcwq;
+	const char *pri = worker_pool_pri(pool) ? "H" : "";
 	struct worker *worker = NULL;
 	int id = -1;
 
 	spin_lock_irq(&gcwq->lock);
-	while (ida_get_new(&gcwq->worker_ida, &id)) {
+	while (ida_get_new(&pool->worker_ida, &id)) {
 		spin_unlock_irq(&gcwq->lock);
-		if (!ida_pre_get(&gcwq->worker_ida, GFP_KERNEL))
+		if (!ida_pre_get(&pool->worker_ida, GFP_KERNEL))
 			goto fail;
 		spin_lock_irq(&gcwq->lock);
 	}
@@ -1374,30 +1500,35 @@ static struct worker *create_worker(struct global_cwq *gcwq, bool bind)
 	if (!worker)
 		goto fail;
 
-	worker->gcwq = gcwq;
+	worker->pool = pool;
 	worker->id = id;
 
-	if (!on_unbound_cpu)
+	if (gcwq->cpu != WORK_CPU_UNBOUND)
 		worker->task = kthread_create_on_node(worker_thread,
-						      worker,
-						      cpu_to_node(gcwq->cpu),
-						      "kworker/%u:%d", gcwq->cpu, id);
+					worker, cpu_to_node(gcwq->cpu),
+					"kworker/%u:%d%s", gcwq->cpu, id, pri);
 	else
 		worker->task = kthread_create(worker_thread, worker,
-					      "kworker/u:%d", id);
+					      "kworker/u:%d%s", id, pri);
 	if (IS_ERR(worker->task))
 		goto fail;
 
+	if (worker_pool_pri(pool))
+		set_user_nice(worker->task, HIGHPRI_NICE_LEVEL);
+
 	/*
-	 * A rogue worker will become a regular one if CPU comes
-	 * online later on.  Make sure every worker has
-	 * PF_THREAD_BOUND set.
+	 * Determine CPU binding of the new worker depending on
+	 * %GCWQ_DISASSOCIATED.  The caller is responsible for ensuring the
+	 * flag remains stable across this function.  See the comments
+	 * above the flag definition for details.
+	 *
+	 * As an unbound worker may later become a regular one if CPU comes
+	 * online, make sure every worker has %PF_THREAD_BOUND set.
 	 */
-	if (bind && !on_unbound_cpu)
+	if (!(gcwq->flags & GCWQ_DISASSOCIATED)) {
 		kthread_bind(worker->task, gcwq->cpu);
-	else {
+	} else {
 		worker->task->flags |= PF_THREAD_BOUND;
-		if (on_unbound_cpu)
 		worker->flags |= WORKER_UNBOUND;
 	}
 
@@ -1405,7 +1536,7 @@ static struct worker *create_worker(struct global_cwq *gcwq, bool bind)
 fail:
 	if (id >= 0) {
 		spin_lock_irq(&gcwq->lock);
-		ida_remove(&gcwq->worker_ida, id);
+		ida_remove(&pool->worker_ida, id);
 		spin_unlock_irq(&gcwq->lock);
 	}
 	kfree(worker);
@@ -1424,7 +1555,7 @@ static struct worker *create_worker(struct global_cwq *gcwq, bool bind)
 static void start_worker(struct worker *worker)
 {
 	worker->flags |= WORKER_STARTED;
-	worker->gcwq->nr_workers++;
+	worker->pool->nr_workers++;
 	worker_enter_idle(worker);
 	wake_up_process(worker->task);
 }
@@ -1440,7 +1571,8 @@ static void start_worker(struct worker *worker)
  */
 static void destroy_worker(struct worker *worker)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
+	struct global_cwq *gcwq = pool->gcwq;
 	int id = worker->id;
 
 	/* sanity check frenzy */
@@ -1448,9 +1580,9 @@ static void destroy_worker(struct worker *worker)
 	BUG_ON(!list_empty(&worker->scheduled));
 
 	if (worker->flags & WORKER_STARTED)
-		gcwq->nr_workers--;
+		pool->nr_workers--;
 	if (worker->flags & WORKER_IDLE)
-		gcwq->nr_idle--;
+		pool->nr_idle--;
 
 	list_del_init(&worker->entry);
 	worker->flags |= WORKER_DIE;
@@ -1461,29 +1593,30 @@ static void destroy_worker(struct worker *worker)
 	kfree(worker);
 
 	spin_lock_irq(&gcwq->lock);
-	ida_remove(&gcwq->worker_ida, id);
+	ida_remove(&pool->worker_ida, id);
 }
 
-static void idle_worker_timeout(unsigned long __gcwq)
+static void idle_worker_timeout(unsigned long __pool)
 {
-	struct global_cwq *gcwq = (void *)__gcwq;
+	struct worker_pool *pool = (void *)__pool;
+	struct global_cwq *gcwq = pool->gcwq;
 
 	spin_lock_irq(&gcwq->lock);
 
-	if (too_many_workers(gcwq)) {
+	if (too_many_workers(pool)) {
 		struct worker *worker;
 		unsigned long expires;
 
 		/* idle_list is kept in LIFO order, check the last one */
-		worker = list_entry(gcwq->idle_list.prev, struct worker, entry);
+		worker = list_entry(pool->idle_list.prev, struct worker, entry);
 		expires = worker->last_active + IDLE_WORKER_TIMEOUT;
 
 		if (time_before(jiffies, expires))
-			mod_timer(&gcwq->idle_timer, expires);
+			mod_timer(&pool->idle_timer, expires);
 		else {
 			/* it's been idle for too long, wake up manager */
-			gcwq->flags |= GCWQ_MANAGE_WORKERS;
-			wake_up_worker(gcwq);
+			pool->flags |= POOL_MANAGE_WORKERS;
+			wake_up_worker(pool);
 		}
 	}
 
@@ -1500,7 +1633,7 @@ static bool send_mayday(struct work_struct *work)
 		return false;
 
 	/* mayday mayday mayday */
-	cpu = cwq->gcwq->cpu;
+	cpu = cwq->pool->gcwq->cpu;
 	/* WORK_CPU_UNBOUND can't be set in cpumask, use cpu 0 instead */
 	if (cpu == WORK_CPU_UNBOUND)
 		cpu = 0;
@@ -1509,37 +1642,38 @@ static bool send_mayday(struct work_struct *work)
 	return true;
 }
 
-static void gcwq_mayday_timeout(unsigned long __gcwq)
+static void gcwq_mayday_timeout(unsigned long __pool)
 {
-	struct global_cwq *gcwq = (void *)__gcwq;
+	struct worker_pool *pool = (void *)__pool;
+	struct global_cwq *gcwq = pool->gcwq;
 	struct work_struct *work;
 
 	spin_lock_irq(&gcwq->lock);
 
-	if (need_to_create_worker(gcwq)) {
+	if (need_to_create_worker(pool)) {
 		/*
 		 * We've been trying to create a new worker but
 		 * haven't been successful.  We might be hitting an
 		 * allocation deadlock.  Send distress signals to
 		 * rescuers.
 		 */
-		list_for_each_entry(work, &gcwq->worklist, entry)
+		list_for_each_entry(work, &pool->worklist, entry)
 			send_mayday(work);
 	}
 
 	spin_unlock_irq(&gcwq->lock);
 
-	mod_timer(&gcwq->mayday_timer, jiffies + MAYDAY_INTERVAL);
+	mod_timer(&pool->mayday_timer, jiffies + MAYDAY_INTERVAL);
 }
 
 /**
  * maybe_create_worker - create a new worker if necessary
- * @gcwq: gcwq to create a new worker for
+ * @pool: pool to create a new worker for
  *
- * Create a new worker for @gcwq if necessary.  @gcwq is guaranteed to
+ * Create a new worker for @pool if necessary.  @pool is guaranteed to
  * have at least one idle worker on return from this function.  If
  * creating a new worker takes longer than MAYDAY_INTERVAL, mayday is
- * sent to all rescuers with works scheduled on @gcwq to resolve
+ * sent to all rescuers with works scheduled on @pool to resolve
  * possible allocation deadlock.
  *
  * On return, need_to_create_worker() is guaranteed to be false and
@@ -1554,52 +1688,54 @@ static void gcwq_mayday_timeout(unsigned long __gcwq)
  * false if no action was taken and gcwq->lock stayed locked, true
  * otherwise.
  */
-static bool maybe_create_worker(struct global_cwq *gcwq)
+static bool maybe_create_worker(struct worker_pool *pool)
 __releases(&gcwq->lock)
 __acquires(&gcwq->lock)
 {
-	if (!need_to_create_worker(gcwq))
+	struct global_cwq *gcwq = pool->gcwq;
+
+	if (!need_to_create_worker(pool))
 		return false;
 restart:
 	spin_unlock_irq(&gcwq->lock);
 
 	/* if we don't make progress in MAYDAY_INITIAL_TIMEOUT, call for help */
-	mod_timer(&gcwq->mayday_timer, jiffies + MAYDAY_INITIAL_TIMEOUT);
+	mod_timer(&pool->mayday_timer, jiffies + MAYDAY_INITIAL_TIMEOUT);
 
 	while (true) {
 		struct worker *worker;
 
-		worker = create_worker(gcwq, true);
+		worker = create_worker(pool);
 		if (worker) {
-			del_timer_sync(&gcwq->mayday_timer);
+			del_timer_sync(&pool->mayday_timer);
 			spin_lock_irq(&gcwq->lock);
 			start_worker(worker);
-			BUG_ON(need_to_create_worker(gcwq));
+			BUG_ON(need_to_create_worker(pool));
 			return true;
 		}
 
-		if (!need_to_create_worker(gcwq))
+		if (!need_to_create_worker(pool))
 			break;
 
 		__set_current_state(TASK_INTERRUPTIBLE);
 		schedule_timeout(CREATE_COOLDOWN);
 
-		if (!need_to_create_worker(gcwq))
+		if (!need_to_create_worker(pool))
 			break;
 	}
 
-	del_timer_sync(&gcwq->mayday_timer);
+	del_timer_sync(&pool->mayday_timer);
 	spin_lock_irq(&gcwq->lock);
-	if (need_to_create_worker(gcwq))
+	if (need_to_create_worker(pool))
 		goto restart;
 	return true;
 }
 
 /**
  * maybe_destroy_worker - destroy workers which have been idle for a while
- * @gcwq: gcwq to destroy workers for
+ * @pool: pool to destroy workers for
  *
- * Destroy @gcwq workers which have been idle for longer than
+ * Destroy @pool workers which have been idle for longer than
  * IDLE_WORKER_TIMEOUT.
  *
  * LOCKING:
@@ -1610,19 +1746,19 @@ __acquires(&gcwq->lock)
  * false if no action was taken and gcwq->lock stayed locked, true
  * otherwise.
  */
-static bool maybe_destroy_workers(struct global_cwq *gcwq)
+static bool maybe_destroy_workers(struct worker_pool *pool)
 {
 	bool ret = false;
 
-	while (too_many_workers(gcwq)) {
+	while (too_many_workers(pool)) {
 		struct worker *worker;
 		unsigned long expires;
 
-		worker = list_entry(gcwq->idle_list.prev, struct worker, entry);
+		worker = list_entry(pool->idle_list.prev, struct worker, entry);
 		expires = worker->last_active + IDLE_WORKER_TIMEOUT;
 
 		if (time_before(jiffies, expires)) {
-			mod_timer(&gcwq->idle_timer, expires);
+			mod_timer(&pool->idle_timer, expires);
 			break;
 		}
 
@@ -1655,31 +1791,22 @@ static bool maybe_destroy_workers(struct global_cwq *gcwq)
  */
 static bool manage_workers(struct worker *worker)
 {
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
 	bool ret = false;
 
-	if (gcwq->flags & GCWQ_MANAGING_WORKERS)
+	if (!mutex_trylock(&pool->manager_mutex))
 		return ret;
 
-	gcwq->flags &= ~GCWQ_MANAGE_WORKERS;
-	gcwq->flags |= GCWQ_MANAGING_WORKERS;
+	pool->flags &= ~POOL_MANAGE_WORKERS;
 
 	/*
 	 * Destroy and then create so that may_start_working() is true
 	 * on return.
 	 */
-	ret |= maybe_destroy_workers(gcwq);
-	ret |= maybe_create_worker(gcwq);
-
-	gcwq->flags &= ~GCWQ_MANAGING_WORKERS;
-
-	/*
-	 * The trustee might be waiting to take over the manager
-	 * position, tell it we're done.
-	 */
-	if (unlikely(gcwq->trustee))
-		wake_up_all(&gcwq->trustee_wait);
+	ret |= maybe_destroy_workers(pool);
+	ret |= maybe_create_worker(pool);
 
+	mutex_unlock(&pool->manager_mutex);
 	return ret;
 }
 
@@ -1728,10 +1855,9 @@ static void cwq_activate_first_delayed(struct cpu_workqueue_struct *cwq)
 {
 	struct work_struct *work = list_first_entry(&cwq->delayed_works,
 						    struct work_struct, entry);
-	struct list_head *pos = gcwq_determine_ins_pos(cwq->gcwq, cwq);
 
 	trace_workqueue_activate_work(work);
-	move_linked_works(work, pos, NULL);
+	move_linked_works(work, &cwq->pool->worklist, NULL);
 	__clear_bit(WORK_STRUCT_DELAYED_BIT, work_data_bits(work));
 	cwq->nr_active++;
 }
@@ -1804,7 +1930,8 @@ __releases(&gcwq->lock)
 __acquires(&gcwq->lock)
 {
 	struct cpu_workqueue_struct *cwq = get_work_cwq(work);
-	struct global_cwq *gcwq = cwq->gcwq;
+	struct worker_pool *pool = worker->pool;
+	struct global_cwq *gcwq = pool->gcwq;
 	struct hlist_head *bwh = busy_worker_head(gcwq, work);
 	bool cpu_intensive = cwq->wq->flags & WQ_CPU_INTENSIVE;
 	work_func_t f = work->func;
@@ -1822,6 +1949,15 @@ __acquires(&gcwq->lock)
 
 	lockdep_copy_map(&lockdep_map, &work->lockdep_map);
 #endif
+	/*
+	 * Ensure we're on the correct CPU.  DISASSOCIATED test is
+	 * necessary to avoid spurious warnings from rescuers servicing the
+	 * unbound or a disassociated gcwq.
+	 */
+	WARN_ON_ONCE(!(worker->flags & (WORKER_UNBOUND | WORKER_REBIND)) &&
+		     !(gcwq->flags & GCWQ_DISASSOCIATED) &&
+		     raw_smp_processor_id() != gcwq->cpu);
+
 	/*
 	 * A single work shouldn't be executed concurrently by
 	 * multiple workers on a single cpu.  Check whether anyone is
@@ -1845,21 +1981,6 @@ __acquires(&gcwq->lock)
 	set_work_cpu(work, gcwq->cpu);
 	list_del_init(&work->entry);
 
-	/*
-	 * If HIGHPRI_PENDING, check the next work, and, if HIGHPRI,
-	 * wake up another worker; otherwise, clear HIGHPRI_PENDING.
-	 */
-	if (unlikely(gcwq->flags & GCWQ_HIGHPRI_PENDING)) {
-		struct work_struct *nwork = list_first_entry(&gcwq->worklist,
-						struct work_struct, entry);
-
-		if (!list_empty(&gcwq->worklist) &&
-		    get_work_cwq(nwork)->wq->flags & WQ_HIGHPRI)
-			wake_up_worker(gcwq);
-		else
-			gcwq->flags &= ~GCWQ_HIGHPRI_PENDING;
-	}
-
 	/*
 	 * CPU intensive works don't participate in concurrency
 	 * management.  They're the scheduler's responsibility.
@@ -1867,6 +1988,13 @@ __acquires(&gcwq->lock)
 	if (unlikely(cpu_intensive))
 		worker_set_flags(worker, WORKER_CPU_INTENSIVE, true);
 
+	/*
+	 * Unbound gcwq isn't concurrency managed and work items should be
+	 * executed ASAP.  Wake up another worker if necessary.
+	 */
+	if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
+		wake_up_worker(pool);
+
 	spin_unlock_irq(&gcwq->lock);
 
 	work_clear_pending(work);
@@ -1939,28 +2067,38 @@ static void process_scheduled_works(struct worker *worker)
 static int worker_thread(void *__worker)
 {
 	struct worker *worker = __worker;
-	struct global_cwq *gcwq = worker->gcwq;
+	struct worker_pool *pool = worker->pool;
+	struct global_cwq *gcwq = pool->gcwq;
 
 	/* tell the scheduler that this is a workqueue worker */
 	worker->task->flags |= PF_WQ_WORKER;
 woke_up:
 	spin_lock_irq(&gcwq->lock);
 
-	/* DIE can be set only while we're idle, checking here is enough */
-	if (worker->flags & WORKER_DIE) {
+	/*
+	 * DIE can be set only while idle and REBIND set while busy has
+	 * @worker->rebind_work scheduled.  Checking here is enough.
+	 */
+	if (unlikely(worker->flags & (WORKER_REBIND | WORKER_DIE))) {
 		spin_unlock_irq(&gcwq->lock);
+
+		if (worker->flags & WORKER_DIE) {
 			worker->task->flags &= ~PF_WQ_WORKER;
 			return 0;
 		}
 
+		idle_worker_rebind(worker);
+		goto woke_up;
+	}
+
 	worker_leave_idle(worker);
 recheck:
 	/* no more worker necessary? */
-	if (!need_more_worker(gcwq))
+	if (!need_more_worker(pool))
 		goto sleep;
 
 	/* do we need to manage? */
-	if (unlikely(!may_start_working(gcwq)) && manage_workers(worker))
+	if (unlikely(!may_start_working(pool)) && manage_workers(worker))
 		goto recheck;
 
 	/*
@@ -1979,7 +2117,7 @@ static int worker_thread(void *__worker)
 
 	do {
 		struct work_struct *work =
-			list_first_entry(&gcwq->worklist,
+			list_first_entry(&pool->worklist,
 					 struct work_struct, entry);
 
 		if (likely(!(*work_data_bits(work) & WORK_STRUCT_LINKED))) {
@@ -1991,11 +2129,11 @@ static int worker_thread(void *__worker)
 			move_linked_works(work, &worker->scheduled, NULL);
 			process_scheduled_works(worker);
 		}
-	} while (keep_working(gcwq));
+	} while (keep_working(pool));
 
 	worker_set_flags(worker, WORKER_PREP, false);
 sleep:
-	if (unlikely(need_to_manage_workers(gcwq)) && manage_workers(worker))
+	if (unlikely(need_to_manage_workers(pool)) && manage_workers(worker))
 		goto recheck;
 
 	/*
@@ -2053,14 +2191,15 @@ static int rescuer_thread(void *__wq)
 	for_each_mayday_cpu(cpu, wq->mayday_mask) {
 		unsigned int tcpu = is_unbound ? WORK_CPU_UNBOUND : cpu;
 		struct cpu_workqueue_struct *cwq = get_cwq(tcpu, wq);
-		struct global_cwq *gcwq = cwq->gcwq;
+		struct worker_pool *pool = cwq->pool;
+		struct global_cwq *gcwq = pool->gcwq;
 		struct work_struct *work, *n;
 
 		__set_current_state(TASK_RUNNING);
 		mayday_clear_cpu(cpu, wq->mayday_mask);
 
 		/* migrate to the target cpu if possible */
-		rescuer->gcwq = gcwq;
+		rescuer->pool = pool;
 		worker_maybe_bind_and_lock(rescuer);
 
 		/*
@@ -2068,7 +2207,7 @@ static int rescuer_thread(void *__wq)
 		 * process'em.
 		 */
 		BUG_ON(!list_empty(&rescuer->scheduled));
-		list_for_each_entry_safe(work, n, &gcwq->worklist, entry)
+		list_for_each_entry_safe(work, n, &pool->worklist, entry)
 			if (get_work_cwq(work) == cwq)
 				move_linked_works(work, scheduled, &n);
 
@@ -2079,8 +2218,8 @@ static int rescuer_thread(void *__wq)
 		 * regular worker; otherwise, we end up with 0 concurrency
 		 * and stalling the execution.
 		 */
-		if (keep_working(gcwq))
-			wake_up_worker(gcwq);
+		if (keep_working(pool))
+			wake_up_worker(pool);
 
 		spin_unlock_irq(&gcwq->lock);
 	}
@@ -2205,7 +2344,7 @@ static bool flush_workqueue_prep_cwqs(struct workqueue_struct *wq,
 
 	for_each_cwq_cpu(cpu, wq) {
 		struct cpu_workqueue_struct *cwq = get_cwq(cpu, wq);
-		struct global_cwq *gcwq = cwq->gcwq;
+		struct global_cwq *gcwq = cwq->pool->gcwq;
 
 		spin_lock_irq(&gcwq->lock);
 
@@ -2421,9 +2560,9 @@ void drain_workqueue(struct workqueue_struct *wq)
 		struct cpu_workqueue_struct *cwq = get_cwq(cpu, wq);
 		bool drained;
 
-		spin_lock_irq(&cwq->gcwq->lock);
+		spin_lock_irq(&cwq->pool->gcwq->lock);
 		drained = !cwq->nr_active && list_empty(&cwq->delayed_works);
-		spin_unlock_irq(&cwq->gcwq->lock);
+		spin_unlock_irq(&cwq->pool->gcwq->lock);
 
 		if (drained)
 			continue;
@@ -2463,7 +2602,7 @@ static bool start_flush_work(struct work_struct *work, struct wq_barrier *barr,
 		 */
 		smp_rmb();
 		cwq = get_work_cwq(work);
-		if (unlikely(!cwq || gcwq != cwq->gcwq))
+		if (unlikely(!cwq || gcwq != cwq->pool->gcwq))
 			goto already_gone;
 	} else if (wait_executing) {
 		worker = find_worker_executing_work(gcwq, work);
@@ -2984,13 +3123,6 @@ struct workqueue_struct *__alloc_workqueue_key(const char *fmt,
 	if (flags & WQ_MEM_RECLAIM)
 		flags |= WQ_RESCUER;
 
-	/*
-	 * Unbound workqueues aren't concurrency managed and should be
-	 * dispatched to workers immediately.
-	 */
-	if (flags & WQ_UNBOUND)
-		flags |= WQ_HIGHPRI;
-
 	max_active = max_active ?: WQ_DFL_ACTIVE;
 	max_active = wq_clamp_max_active(max_active, flags, wq->name);
 
@@ -3011,9 +3143,10 @@ struct workqueue_struct *__alloc_workqueue_key(const char *fmt,
 	for_each_cwq_cpu(cpu, wq) {
 		struct cpu_workqueue_struct *cwq = get_cwq(cpu, wq);
 		struct global_cwq *gcwq = get_gcwq(cpu);
+		int pool_idx = (bool)(flags & WQ_HIGHPRI);
 
 		BUG_ON((unsigned long)cwq & WORK_STRUCT_FLAG_MASK);
-		cwq->gcwq = gcwq;
+		cwq->pool = &gcwq->pools[pool_idx];
 		cwq->wq = wq;
 		cwq->flush_color = -1;
 		cwq->max_active = max_active;
@@ -3225,369 +3358,143 @@ EXPORT_SYMBOL_GPL(work_busy);
  * gcwqs serve mix of short, long and very long running works making
  * blocked draining impractical.
  *
- * This is solved by allowing a gcwq to be detached from CPU, running
- * it with unbound (rogue) workers and allowing it to be reattached
- * later if the cpu comes back online.  A separate thread is created
- * to govern a gcwq in such state and is called the trustee of the
- * gcwq.
- *
- * Trustee states and their descriptions.
- *
- * START	Command state used on startup.  On CPU_DOWN_PREPARE, a
- *		new trustee is started with this state.
- *
- * IN_CHARGE	Once started, trustee will enter this state after
- *		assuming the manager role and making all existing
- *		workers rogue.  DOWN_PREPARE waits for trustee to
- *		enter this state.  After reaching IN_CHARGE, trustee
- *		tries to execute the pending worklist until it's empty
- *		and the state is set to BUTCHER, or the state is set
- *		to RELEASE.
- *
- * BUTCHER	Command state which is set by the cpu callback after
- *		the cpu has went down.  Once this state is set trustee
- *		knows that there will be no new works on the worklist
- *		and once the worklist is empty it can proceed to
- *		killing idle workers.
- *
- * RELEASE	Command state which is set by the cpu callback if the
- *		cpu down has been canceled or it has come online
- *		again.  After recognizing this state, trustee stops
- *		trying to drain or butcher and clears ROGUE, rebinds
- *		all remaining workers back to the cpu and releases
- *		manager role.
- *
- * DONE		Trustee will enter this state after BUTCHER or RELEASE
- *		is complete.
- *
- *          trustee                 CPU                draining
- *         took over                down               complete
- * START -----------> IN_CHARGE -----------> BUTCHER -----------> DONE
- *                        |                     |                  ^
- *                        | CPU is back online  v   return workers |
- *                         ----------------> RELEASE --------------
+ * This is solved by allowing a gcwq to be disassociated from the CPU
+ * running as an unbound one and allowing it to be reattached later if the
+ * cpu comes back online.
  */
 
-/**
- * trustee_wait_event_timeout - timed event wait for trustee
- * @cond: condition to wait for
- * @timeout: timeout in jiffies
- *
- * wait_event_timeout() for trustee to use.  Handles locking and
- * checks for RELEASE request.
- *
- * CONTEXT:
- * spin_lock_irq(gcwq->lock) which may be released and regrabbed
- * multiple times.  To be used by trustee.
- *
- * RETURNS:
- * Positive indicating left time if @cond is satisfied, 0 if timed
- * out, -1 if canceled.
- */
-#define trustee_wait_event_timeout(cond, timeout) ({			\
-	long __ret = (timeout);						\
-	while (!((cond) || (gcwq->trustee_state == TRUSTEE_RELEASE)) &&	\
-	       __ret) {							\
-		spin_unlock_irq(&gcwq->lock);				\
-		__wait_event_timeout(gcwq->trustee_wait, (cond) ||	\
-			(gcwq->trustee_state == TRUSTEE_RELEASE),	\
-			__ret);						\
-		spin_lock_irq(&gcwq->lock);				\
-	}								\
-	gcwq->trustee_state == TRUSTEE_RELEASE ? -1 : (__ret);		\
-})
+/* claim manager positions of all pools */
+static void gcwq_claim_management_and_lock(struct global_cwq *gcwq)
+{
+	struct worker_pool *pool;
 
-/**
- * trustee_wait_event - event wait for trustee
- * @cond: condition to wait for
- *
- * wait_event() for trustee to use.  Automatically handles locking and
- * checks for CANCEL request.
- *
- * CONTEXT:
- * spin_lock_irq(gcwq->lock) which may be released and regrabbed
- * multiple times.  To be used by trustee.
- *
- * RETURNS:
- * 0 if @cond is satisfied, -1 if canceled.
- */
-#define trustee_wait_event(cond) ({					\
-	long __ret1;							\
-	__ret1 = trustee_wait_event_timeout(cond, MAX_SCHEDULE_TIMEOUT);\
-	__ret1 < 0 ? -1 : 0;						\
-})
+	for_each_worker_pool(pool, gcwq)
+		mutex_lock_nested(&pool->manager_mutex, pool - gcwq->pools);
+	spin_lock_irq(&gcwq->lock);
+}
 
-static int __cpuinit trustee_thread(void *__gcwq)
+/* release manager positions */
+static void gcwq_release_management_and_unlock(struct global_cwq *gcwq)
 {
-	struct global_cwq *gcwq = __gcwq;
+	struct worker_pool *pool;
+
+	spin_unlock_irq(&gcwq->lock);
+	for_each_worker_pool(pool, gcwq)
+		mutex_unlock(&pool->manager_mutex);
+}
+
+static void gcwq_unbind_fn(struct work_struct *work)
+{
+	struct global_cwq *gcwq = get_gcwq(smp_processor_id());
+	struct worker_pool *pool;
 	struct worker *worker;
-	struct work_struct *work;
 	struct hlist_node *pos;
-	long rc;
 	int i;
 
 	BUG_ON(gcwq->cpu != smp_processor_id());
 
-	spin_lock_irq(&gcwq->lock);
-	/*
-	 * Claim the manager position and make all workers rogue.
-	 * Trustee must be bound to the target cpu and can't be
-	 * cancelled.
-	 */
-	BUG_ON(gcwq->cpu != smp_processor_id());
-	rc = trustee_wait_event(!(gcwq->flags & GCWQ_MANAGING_WORKERS));
-	BUG_ON(rc < 0);
-
-	gcwq->flags |= GCWQ_MANAGING_WORKERS;
-
-	list_for_each_entry(worker, &gcwq->idle_list, entry)
-		worker->flags |= WORKER_ROGUE;
-
-	for_each_busy_worker(worker, i, pos, gcwq)
-		worker->flags |= WORKER_ROGUE;
+	gcwq_claim_management_and_lock(gcwq);
 
 	/*
-	 * Call schedule() so that we cross rq->lock and thus can
-	 * guarantee sched callbacks see the rogue flag.  This is
-	 * necessary as scheduler callbacks may be invoked from other
-	 * cpus.
+	 * We've claimed all manager positions.  Make all workers unbound
+	 * and set DISASSOCIATED.  Before this, all workers except for the
+	 * ones which are still executing works from before the last CPU
+	 * down must be on the cpu.  After this, they may become diasporas.
 	 */
-	spin_unlock_irq(&gcwq->lock);
-	schedule();
-	spin_lock_irq(&gcwq->lock);
-
-	/*
-	 * Sched callbacks are disabled now.  Zap nr_running.  After
-	 * this, nr_running stays zero and need_more_worker() and
-	 * keep_working() are always true as long as the worklist is
-	 * not empty.
-	 */
-	atomic_set(get_gcwq_nr_running(gcwq->cpu), 0);
-
-	spin_unlock_irq(&gcwq->lock);
-	del_timer_sync(&gcwq->idle_timer);
-	spin_lock_irq(&gcwq->lock);
-
-	/*
-	 * We're now in charge.  Notify and proceed to drain.  We need
-	 * to keep the gcwq running during the whole CPU down
-	 * procedure as other cpu hotunplug callbacks may need to
-	 * flush currently running tasks.
-	 */
-	gcwq->trustee_state = TRUSTEE_IN_CHARGE;
-	wake_up_all(&gcwq->trustee_wait);
-
-	/*
-	 * The original cpu is in the process of dying and may go away
-	 * anytime now.  When that happens, we and all workers would
-	 * be migrated to other cpus.  Try draining any left work.  We
-	 * want to get it over with ASAP - spam rescuers, wake up as
-	 * many idlers as necessary and create new ones till the
-	 * worklist is empty.  Note that if the gcwq is frozen, there
-	 * may be frozen works in freezable cwqs.  Don't declare
-	 * completion while frozen.
-	 */
-	while (gcwq->nr_workers != gcwq->nr_idle ||
-	       gcwq->flags & GCWQ_FREEZING ||
-	       gcwq->trustee_state == TRUSTEE_IN_CHARGE) {
-		int nr_works = 0;
-
-		list_for_each_entry(work, &gcwq->worklist, entry) {
-			send_mayday(work);
-			nr_works++;
-		}
-
-		list_for_each_entry(worker, &gcwq->idle_list, entry) {
-			if (!nr_works--)
-				break;
-			wake_up_process(worker->task);
-		}
+	for_each_worker_pool(pool, gcwq)
+		list_for_each_entry(worker, &pool->idle_list, entry)
+			worker->flags |= WORKER_UNBOUND;
 
-		if (need_to_create_worker(gcwq)) {
-			spin_unlock_irq(&gcwq->lock);
-			worker = create_worker(gcwq, false);
-			spin_lock_irq(&gcwq->lock);
-			if (worker) {
-				worker->flags |= WORKER_ROGUE;
-				start_worker(worker);
-			}
-		}
+	for_each_busy_worker(worker, i, pos, gcwq)
+		worker->flags |= WORKER_UNBOUND;
 
-		/* give a breather */
-		if (trustee_wait_event_timeout(false, TRUSTEE_COOLDOWN) < 0)
-			break;
-	}
+	gcwq->flags |= GCWQ_DISASSOCIATED;
 
-	/*
-	 * Either all works have been scheduled and cpu is down, or
-	 * cpu down has already been canceled.  Wait for and butcher
-	 * all workers till we're canceled.
-	 */
-	do {
-		rc = trustee_wait_event(!list_empty(&gcwq->idle_list));
-		while (!list_empty(&gcwq->idle_list))
-			destroy_worker(list_first_entry(&gcwq->idle_list,
-							struct worker, entry));
-	} while (gcwq->nr_workers && rc >= 0);
+	gcwq_release_management_and_unlock(gcwq);
 
 	/*
-	 * At this point, either draining has completed and no worker
-	 * is left, or cpu down has been canceled or the cpu is being
-	 * brought back up.  There shouldn't be any idle one left.
-	 * Tell the remaining busy ones to rebind once it finishes the
-	 * currently scheduled works by scheduling the rebind_work.
+	 * Call schedule() so that we cross rq->lock and thus can guarantee
+	 * sched callbacks see the %WORKER_UNBOUND flag.  This is necessary
+	 * as scheduler callbacks may be invoked from other cpus.
 	 */
-	WARN_ON(!list_empty(&gcwq->idle_list));
-
-	for_each_busy_worker(worker, i, pos, gcwq) {
-		struct work_struct *rebind_work = &worker->rebind_work;
+	schedule();
 
 	/*
-		 * Rebind_work may race with future cpu hotplug
-		 * operations.  Use a separate flag to mark that
-		 * rebinding is scheduled.
-		 */
-		worker->flags |= WORKER_REBIND;
-		worker->flags &= ~WORKER_ROGUE;
-
-		/* queue rebind_work, wq doesn't matter, use the default one */
-		if (test_and_set_bit(WORK_STRUCT_PENDING_BIT,
-				     work_data_bits(rebind_work)))
-			continue;
-
-		debug_work_activate(rebind_work);
-		insert_work(get_cwq(gcwq->cpu, system_wq), rebind_work,
-			    worker->scheduled.next,
-			    work_color_to_flags(WORK_NO_COLOR));
-	}
-
-	/* relinquish manager role */
-	gcwq->flags &= ~GCWQ_MANAGING_WORKERS;
-
-	/* notify completion */
-	gcwq->trustee = NULL;
-	gcwq->trustee_state = TRUSTEE_DONE;
-	wake_up_all(&gcwq->trustee_wait);
-	spin_unlock_irq(&gcwq->lock);
-	return 0;
-}
-
-/**
- * wait_trustee_state - wait for trustee to enter the specified state
- * @gcwq: gcwq the trustee of interest belongs to
- * @state: target state to wait for
+	 * Sched callbacks are disabled now.  Zap nr_running.  After this,
+	 * nr_running stays zero and need_more_worker() and keep_working()
+	 * are always true as long as the worklist is not empty.  @gcwq now
+	 * behaves as unbound (in terms of concurrency management) gcwq
+	 * which is served by workers tied to the CPU.
 	 *
- * Wait for the trustee to reach @state.  DONE is already matched.
- *
- * CONTEXT:
- * spin_lock_irq(gcwq->lock) which may be released and regrabbed
- * multiple times.  To be used by cpu_callback.
+	 * On return from this function, the current worker would trigger
+	 * unbound chain execution of pending work items if other workers
+	 * didn't already.
 	 */
-static void __cpuinit wait_trustee_state(struct global_cwq *gcwq, int state)
-__releases(&gcwq->lock)
-__acquires(&gcwq->lock)
-{
-	if (!(gcwq->trustee_state == state ||
-	      gcwq->trustee_state == TRUSTEE_DONE)) {
-		spin_unlock_irq(&gcwq->lock);
-		__wait_event(gcwq->trustee_wait,
-			     gcwq->trustee_state == state ||
-			     gcwq->trustee_state == TRUSTEE_DONE);
-		spin_lock_irq(&gcwq->lock);
-	}
+	for_each_worker_pool(pool, gcwq)
+		atomic_set(get_pool_nr_running(pool), 0);
 }
 
-static int __devinit workqueue_cpu_callback(struct notifier_block *nfb,
+/*
+ * Workqueues should be brought up before normal priority CPU notifiers.
+ * This will be registered high priority CPU notifier.
+ */
+static int __devinit workqueue_cpu_up_callback(struct notifier_block *nfb,
 					       unsigned long action,
 					       void *hcpu)
 {
 	unsigned int cpu = (unsigned long)hcpu;
 	struct global_cwq *gcwq = get_gcwq(cpu);
-	struct task_struct *new_trustee = NULL;
-	struct worker *uninitialized_var(new_worker);
-	unsigned long flags;
-
-	action &= ~CPU_TASKS_FROZEN;
+	struct worker_pool *pool;
 
-	switch (action) {
-	case CPU_DOWN_PREPARE:
-		new_trustee = kthread_create(trustee_thread, gcwq,
-					     "workqueue_trustee/%d\n", cpu);
-		if (IS_ERR(new_trustee))
-			return notifier_from_errno(PTR_ERR(new_trustee));
-		kthread_bind(new_trustee, cpu);
-		/* fall through */
+	switch (action & ~CPU_TASKS_FROZEN) {
 	case CPU_UP_PREPARE:
-		BUG_ON(gcwq->first_idle);
-		new_worker = create_worker(gcwq, false);
-		if (!new_worker) {
-			if (new_trustee)
-				kthread_stop(new_trustee);
-			return NOTIFY_BAD;
-		}
-	}
-
-	/* some are called w/ irq disabled, don't disturb irq status */
-	spin_lock_irqsave(&gcwq->lock, flags);
+		for_each_worker_pool(pool, gcwq) {
+			struct worker *worker;
 
-	switch (action) {
-	case CPU_DOWN_PREPARE:
-		/* initialize trustee and tell it to acquire the gcwq */
-		BUG_ON(gcwq->trustee || gcwq->trustee_state != TRUSTEE_DONE);
-		gcwq->trustee = new_trustee;
-		gcwq->trustee_state = TRUSTEE_START;
-		wake_up_process(gcwq->trustee);
-		wait_trustee_state(gcwq, TRUSTEE_IN_CHARGE);
-		/* fall through */
-	case CPU_UP_PREPARE:
-		BUG_ON(gcwq->first_idle);
-		gcwq->first_idle = new_worker;
-		break;
+			if (pool->nr_workers)
+				continue;
 
-	case CPU_DYING:
-		/*
-		 * Before this, the trustee and all workers except for
-		 * the ones which are still executing works from
-		 * before the last CPU down must be on the cpu.  After
-		 * this, they'll all be diasporas.
-		 */
-		gcwq->flags |= GCWQ_DISASSOCIATED;
-		break;
+			worker = create_worker(pool);
+			if (!worker)
+				return NOTIFY_BAD;
 
-	case CPU_POST_DEAD:
-		gcwq->trustee_state = TRUSTEE_BUTCHER;
-		/* fall through */
-	case CPU_UP_CANCELED:
-		destroy_worker(gcwq->first_idle);
-		gcwq->first_idle = NULL;
+			spin_lock_irq(&gcwq->lock);
+			start_worker(worker);
+			spin_unlock_irq(&gcwq->lock);
+		}
 		break;
 
 	case CPU_DOWN_FAILED:
 	case CPU_ONLINE:
+		gcwq_claim_management_and_lock(gcwq);
 		gcwq->flags &= ~GCWQ_DISASSOCIATED;
-		if (gcwq->trustee_state != TRUSTEE_DONE) {
-			gcwq->trustee_state = TRUSTEE_RELEASE;
-			wake_up_process(gcwq->trustee);
-			wait_trustee_state(gcwq, TRUSTEE_DONE);
+		rebind_workers(gcwq);
+		gcwq_release_management_and_unlock(gcwq);
+		break;
 	}
+	return NOTIFY_OK;
+}
 
-		/*
-		 * Trustee is done and there might be no worker left.
-		 * Put the first_idle in and request a real manager to
-		 * take a look.
+/*
+ * Workqueues should be brought down after normal priority CPU notifiers.
+ * This will be registered as low priority CPU notifier.
  */
-		spin_unlock_irq(&gcwq->lock);
-		kthread_bind(gcwq->first_idle->task, cpu);
-		spin_lock_irq(&gcwq->lock);
-		gcwq->flags |= GCWQ_MANAGE_WORKERS;
-		start_worker(gcwq->first_idle);
-		gcwq->first_idle = NULL;
+static int __devinit workqueue_cpu_down_callback(struct notifier_block *nfb,
+						 unsigned long action,
+						 void *hcpu)
+{
+	unsigned int cpu = (unsigned long)hcpu;
+	struct work_struct unbind_work;
+
+	switch (action & ~CPU_TASKS_FROZEN) {
+	case CPU_DOWN_PREPARE:
+		/* unbinding should happen on the local CPU */
+		INIT_WORK_ONSTACK(&unbind_work, gcwq_unbind_fn);
+		schedule_work_on(cpu, &unbind_work);
+		flush_work(&unbind_work);
 		break;
 	}
-
-	spin_unlock_irqrestore(&gcwq->lock, flags);
-
-	return notifier_from_errno(0);
+	return NOTIFY_OK;
 }
 
 #ifdef CONFIG_SMP
@@ -3746,6 +3653,7 @@ void thaw_workqueues(void)
 
 	for_each_gcwq_cpu(cpu) {
 		struct global_cwq *gcwq = get_gcwq(cpu);
+		struct worker_pool *pool;
 		struct workqueue_struct *wq;
 
 		spin_lock_irq(&gcwq->lock);
@@ -3767,7 +3675,8 @@ void thaw_workqueues(void)
 				cwq_activate_first_delayed(cwq);
 		}
 
-		wake_up_worker(gcwq);
+		for_each_worker_pool(pool, gcwq)
+			wake_up_worker(pool);
 
 		spin_unlock_irq(&gcwq->lock);
 	}
@@ -3783,47 +3692,58 @@ static int __init init_workqueues(void)
 	unsigned int cpu;
 	int i;
 
-	cpu_notifier(workqueue_cpu_callback, CPU_PRI_WORKQUEUE);
+	cpu_notifier(workqueue_cpu_up_callback, CPU_PRI_WORKQUEUE_UP);
+	cpu_notifier(workqueue_cpu_down_callback, CPU_PRI_WORKQUEUE_DOWN);
 
 	/* initialize gcwqs */
 	for_each_gcwq_cpu(cpu) {
 		struct global_cwq *gcwq = get_gcwq(cpu);
+		struct worker_pool *pool;
 
 		spin_lock_init(&gcwq->lock);
-		INIT_LIST_HEAD(&gcwq->worklist);
 		gcwq->cpu = cpu;
 		gcwq->flags |= GCWQ_DISASSOCIATED;
 
-		INIT_LIST_HEAD(&gcwq->idle_list);
 		for (i = 0; i < BUSY_WORKER_HASH_SIZE; i++)
 			INIT_HLIST_HEAD(&gcwq->busy_hash[i]);
 
-		init_timer_deferrable(&gcwq->idle_timer);
-		gcwq->idle_timer.function = idle_worker_timeout;
-		gcwq->idle_timer.data = (unsigned long)gcwq;
+		for_each_worker_pool(pool, gcwq) {
+			pool->gcwq = gcwq;
+			INIT_LIST_HEAD(&pool->worklist);
+			INIT_LIST_HEAD(&pool->idle_list);
 
-		setup_timer(&gcwq->mayday_timer, gcwq_mayday_timeout,
-			    (unsigned long)gcwq);
+			init_timer_deferrable(&pool->idle_timer);
+			pool->idle_timer.function = idle_worker_timeout;
+			pool->idle_timer.data = (unsigned long)pool;
 
-		ida_init(&gcwq->worker_ida);
+			setup_timer(&pool->mayday_timer, gcwq_mayday_timeout,
+				    (unsigned long)pool);
 
-		gcwq->trustee_state = TRUSTEE_DONE;
-		init_waitqueue_head(&gcwq->trustee_wait);
+			mutex_init(&pool->manager_mutex);
+			ida_init(&pool->worker_ida);
+		}
+
+		init_waitqueue_head(&gcwq->rebind_hold);
 	}
 
 	/* create the initial worker */
 	for_each_online_gcwq_cpu(cpu) {
 		struct global_cwq *gcwq = get_gcwq(cpu);
-		struct worker *worker;
+		struct worker_pool *pool;
 
 		if (cpu != WORK_CPU_UNBOUND)
 			gcwq->flags &= ~GCWQ_DISASSOCIATED;
-		worker = create_worker(gcwq, true);
+
+		for_each_worker_pool(pool, gcwq) {
+			struct worker *worker;
+
+			worker = create_worker(pool);
 			BUG_ON(!worker);
 			spin_lock_irq(&gcwq->lock);
 			start_worker(worker);
 			spin_unlock_irq(&gcwq->lock);
 		}
+	}
 
 	system_wq = alloc_workqueue("events", 0, 0);
 	system_long_wq = alloc_workqueue("events_long", 0, 0);