workqueue.c 159 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
/*
Tejun Heo's avatar
Tejun Heo committed
2
 * kernel/workqueue.c - generic async execution with shared worker pool
Linus Torvalds's avatar
Linus Torvalds committed
3
 *
Tejun Heo's avatar
Tejun Heo committed
4
 * Copyright (C) 2002		Ingo Molnar
Linus Torvalds's avatar
Linus Torvalds committed
5
 *
Tejun Heo's avatar
Tejun Heo committed
6 7 8 9 10
 *   Derived from the taskqueue/keventd code by:
 *     David Woodhouse <dwmw2@infradead.org>
 *     Andrew Morton
 *     Kai Petzke <wpp@marie.physik.tu-berlin.de>
 *     Theodore Ts'o <tytso@mit.edu>
Linus Torvalds's avatar
Linus Torvalds committed
11
 *
Tejun Heo's avatar
Tejun Heo committed
12
 * Made to use alloc_percpu by Christoph Lameter.
Linus Torvalds's avatar
Linus Torvalds committed
13
 *
Tejun Heo's avatar
Tejun Heo committed
14 15
 * Copyright (C) 2010		SUSE Linux Products GmbH
 * Copyright (C) 2010		Tejun Heo <tj@kernel.org>
16
 *
Tejun Heo's avatar
Tejun Heo committed
17 18
 * This is the generic async execution mechanism.  Work items as are
 * executed in process context.  The worker pool is shared and
19 20 21 22
 * automatically managed.  There are two worker pools for each CPU (one for
 * normal work items and the other for high priority ones) and some extra
 * pools for workqueues which are not bound to any specific CPU - the
 * number of these backing pools is dynamic.
Tejun Heo's avatar
Tejun Heo committed
23
 *
24
 * Please read Documentation/core-api/workqueue.rst for details.
Linus Torvalds's avatar
Linus Torvalds committed
25 26
 */

27
#include <linux/export.h>
Linus Torvalds's avatar
Linus Torvalds committed
28 29 30 31 32 33 34 35 36 37
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/signal.h>
#include <linux/completion.h>
#include <linux/workqueue.h>
#include <linux/slab.h>
#include <linux/cpu.h>
#include <linux/notifier.h>
#include <linux/kthread.h>
38
#include <linux/hardirq.h>
39
#include <linux/mempolicy.h>
40
#include <linux/freezer.h>
41
#include <linux/debug_locks.h>
42
#include <linux/lockdep.h>
Tejun Heo's avatar
Tejun Heo committed
43
#include <linux/idr.h>
44
#include <linux/jhash.h>
45
#include <linux/hashtable.h>
46
#include <linux/rculist.h>
47
#include <linux/nodemask.h>
48
#include <linux/moduleparam.h>
49
#include <linux/uaccess.h>
50
#include <linux/sched/isolation.h>
51
#include <linux/nmi.h>
52

53
#include "workqueue_internal.h"
Linus Torvalds's avatar
Linus Torvalds committed
54

55
enum {
56 57
	/*
	 * worker_pool flags
58
	 *
59
	 * A bound pool is either associated or disassociated with its CPU.
60 61 62 63 64 65
	 * 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
66
	 * be executing on any CPU.  The pool behaves as an unbound one.
67
	 *
68
	 * Note that DISASSOCIATED should be flipped only while holding
69
	 * wq_pool_attach_mutex to avoid changing binding state while
70
	 * worker_attach_to_pool() is in progress.
71
	 */
72
	POOL_MANAGER_ACTIVE	= 1 << 0,	/* being managed */
73
	POOL_DISASSOCIATED	= 1 << 2,	/* cpu can't serve workers */
74

75 76 77
	/* worker flags */
	WORKER_DIE		= 1 << 1,	/* die die die */
	WORKER_IDLE		= 1 << 2,	/* is idle */
78
	WORKER_PREP		= 1 << 3,	/* preparing to run works */
79
	WORKER_CPU_INTENSIVE	= 1 << 6,	/* cpu intensive */
80
	WORKER_UNBOUND		= 1 << 7,	/* worker is unbound */
81
	WORKER_REBOUND		= 1 << 8,	/* worker was rebound */
82

83 84
	WORKER_NOT_RUNNING	= WORKER_PREP | WORKER_CPU_INTENSIVE |
				  WORKER_UNBOUND | WORKER_REBOUND,
85

86
	NR_STD_WORKER_POOLS	= 2,		/* # standard pools per cpu */
87

88
	UNBOUND_POOL_HASH_ORDER	= 6,		/* hashed by pool->attrs */
89
	BUSY_WORKER_HASH_ORDER	= 6,		/* 64 pointers */
90

91 92 93
	MAX_IDLE_WORKERS_RATIO	= 4,		/* 1/4 of busy can be idle */
	IDLE_WORKER_TIMEOUT	= 300 * HZ,	/* keep idle ones for 5 mins */

94 95 96
	MAYDAY_INITIAL_TIMEOUT  = HZ / 100 >= 2 ? HZ / 100 : 2,
						/* call for help after 10ms
						   (min two ticks) */
97 98 99 100 101
	MAYDAY_INTERVAL		= HZ / 10,	/* and then every 100ms */
	CREATE_COOLDOWN		= HZ,		/* time to breath after fail */

	/*
	 * Rescue workers are used only on emergencies and shared by
102
	 * all cpus.  Give MIN_NICE.
103
	 */
104 105
	RESCUER_NICE_LEVEL	= MIN_NICE,
	HIGHPRI_NICE_LEVEL	= MIN_NICE,
106 107

	WQ_NAME_LEN		= 24,
108
};
Linus Torvalds's avatar
Linus Torvalds committed
109 110

/*
111 112
 * Structure fields follow one of the following exclusion rules.
 *
113 114
 * I: Modifiable by initialization/destruction paths and read-only for
 *    everyone else.
115
 *
116 117 118
 * P: Preemption protected.  Disabling preemption is enough and should
 *    only be modified and accessed from the local cpu.
 *
119
 * L: pool->lock protected.  Access with pool->lock held.
120
 *
121 122 123 124
 * X: During normal operation, modification requires pool->lock and should
 *    be done only from local cpu.  Either disabling preemption on local
 *    cpu or grabbing pool->lock is enough for read access.  If
 *    POOL_DISASSOCIATED is set, it's identical to L.
125
 *
126
 * A: wq_pool_attach_mutex protected.
127
 *
128
 * PL: wq_pool_mutex protected.
129
 *
130
 * PR: wq_pool_mutex protected for writes.  Sched-RCU protected for reads.
131
 *
132 133 134 135 136
 * PW: wq_pool_mutex and wq->mutex protected for writes.  Either for reads.
 *
 * PWR: wq_pool_mutex and wq->mutex protected for writes.  Either or
 *      sched-RCU for reads.
 *
137 138
 * WQ: wq->mutex protected.
 *
139
 * WR: wq->mutex protected for writes.  Sched-RCU protected for reads.
140 141
 *
 * MD: wq_mayday_lock protected.
Linus Torvalds's avatar
Linus Torvalds committed
142 143
 */

144
/* struct worker is defined in workqueue_internal.h */
Tejun Heo's avatar
Tejun Heo committed
145

146
struct worker_pool {
147
	spinlock_t		lock;		/* the pool lock */
148
	int			cpu;		/* I: the associated cpu */
149
	int			node;		/* I: the associated node ID */
150
	int			id;		/* I: pool ID */
151
	unsigned int		flags;		/* X: flags */
152

153 154
	unsigned long		watchdog_ts;	/* L: watchdog timestamp */

155
	struct list_head	worklist;	/* L: list of pending works */
156

157 158
	int			nr_workers;	/* L: total number of workers */
	int			nr_idle;	/* L: currently idle workers */
159 160 161 162 163

	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 */

164
	/* a workers is either on busy_hash or idle_list, or the manager */
165 166 167
	DECLARE_HASHTABLE(busy_hash, BUSY_WORKER_HASH_ORDER);
						/* L: hash of busy workers */

168
	struct worker		*manager;	/* L: purely informational */
169
	struct list_head	workers;	/* A: attached workers */
170
	struct completion	*detach_completion; /* all workers detached */
171

172
	struct ida		worker_ida;	/* worker IDs for task name */
173

174
	struct workqueue_attrs	*attrs;		/* I: worker attributes */
175 176
	struct hlist_node	hash_node;	/* PL: unbound_pool_hash node */
	int			refcnt;		/* PL: refcnt for unbound pools */
177

178 179 180 181 182 183
	/*
	 * The current concurrency level.  As it's likely to be accessed
	 * from other CPUs during try_to_wake_up(), put it in a separate
	 * cacheline.
	 */
	atomic_t		nr_running ____cacheline_aligned_in_smp;
184 185 186 187 188 189

	/*
	 * Destruction of pool is sched-RCU protected to allow dereferences
	 * from get_work_pool().
	 */
	struct rcu_head		rcu;
190 191
} ____cacheline_aligned_in_smp;

Linus Torvalds's avatar
Linus Torvalds committed
192
/*
193 194 195 196
 * The per-pool workqueue.  While queued, the lower WORK_STRUCT_FLAG_BITS
 * of work_struct->data are used for flags and the remaining high bits
 * point to the pwq; thus, pwqs need to be aligned at two's power of the
 * number of flag bits.
Linus Torvalds's avatar
Linus Torvalds committed
197
 */
198
struct pool_workqueue {
199
	struct worker_pool	*pool;		/* I: the associated pool */
200
	struct workqueue_struct *wq;		/* I: the owning workqueue */
201 202
	int			work_color;	/* L: current color */
	int			flush_color;	/* L: flushing color */
203
	int			refcnt;		/* L: reference count */
204 205
	int			nr_in_flight[WORK_NR_COLORS];
						/* L: nr of in_flight works */
206
	int			nr_active;	/* L: nr of active works */
207
	int			max_active;	/* L: max active works */
208
	struct list_head	delayed_works;	/* L: delayed works */
209
	struct list_head	pwqs_node;	/* WR: node on wq->pwqs */
210
	struct list_head	mayday_node;	/* MD: node on wq->maydays */
211 212 213 214 215

	/*
	 * Release of unbound pwq is punted to system_wq.  See put_pwq()
	 * and pwq_unbound_release_workfn() for details.  pool_workqueue
	 * itself is also sched-RCU protected so that the first pwq can be
216
	 * determined without grabbing wq->mutex.
217 218 219
	 */
	struct work_struct	unbound_release_work;
	struct rcu_head		rcu;
220
} __aligned(1 << WORK_STRUCT_FLAG_BITS);
Linus Torvalds's avatar
Linus Torvalds committed
221

222 223 224 225
/*
 * Structure used to wait for workqueue flush.
 */
struct wq_flusher {
226 227
	struct list_head	list;		/* WQ: list of flushers */
	int			flush_color;	/* WQ: flush color waiting for */
228 229 230
	struct completion	done;		/* flush completion */
};

231 232
struct wq_device;

Linus Torvalds's avatar
Linus Torvalds committed
233
/*
234 235
 * The externally visible workqueue.  It relays the issued work items to
 * the appropriate worker_pool through its pool_workqueues.
Linus Torvalds's avatar
Linus Torvalds committed
236 237
 */
struct workqueue_struct {
238
	struct list_head	pwqs;		/* WR: all pwqs of this wq */
239
	struct list_head	list;		/* PR: list of all workqueues */
240

241 242 243
	struct mutex		mutex;		/* protects this wq */
	int			work_color;	/* WQ: current work color */
	int			flush_color;	/* WQ: current flush color */
244
	atomic_t		nr_pwqs_to_flush; /* flush in progress */
245 246 247
	struct wq_flusher	*first_flusher;	/* WQ: first flusher */
	struct list_head	flusher_queue;	/* WQ: flush waiters */
	struct list_head	flusher_overflow; /* WQ: flush overflow list */
248

249
	struct list_head	maydays;	/* MD: pwqs requesting rescue */
250 251
	struct worker		*rescuer;	/* I: rescue worker */

252
	int			nr_drainers;	/* WQ: drain in progress */
253
	int			saved_max_active; /* WQ: saved pwq max_active */
254

255 256
	struct workqueue_attrs	*unbound_attrs;	/* PW: only for unbound wqs */
	struct pool_workqueue	*dfl_pwq;	/* PW: only for unbound wqs */
257

258 259 260
#ifdef CONFIG_SYSFS
	struct wq_device	*wq_dev;	/* I: for sysfs interface */
#endif
261
#ifdef CONFIG_LOCKDEP
262
	struct lockdep_map	lockdep_map;
263
#endif
264
	char			name[WQ_NAME_LEN]; /* I: workqueue name */
265

266 267 268 269 270 271 272
	/*
	 * Destruction of workqueue_struct is sched-RCU protected to allow
	 * walking the workqueues list without grabbing wq_pool_mutex.
	 * This is used to dump all workqueues from sysrq.
	 */
	struct rcu_head		rcu;

273 274 275
	/* hot fields used during command issue, aligned to cacheline */
	unsigned int		flags ____cacheline_aligned; /* WQ: WQ_* flags */
	struct pool_workqueue __percpu *cpu_pwqs; /* I: per-cpu pwqs */
276
	struct pool_workqueue __rcu *numa_pwq_tbl[]; /* PWR: unbound pwqs indexed by node */
Linus Torvalds's avatar
Linus Torvalds committed
277 278
};

279 280
static struct kmem_cache *pwq_cache;

281 282 283
static cpumask_var_t *wq_numa_possible_cpumask;
					/* possible CPUs of each node */

284 285 286
static bool wq_disable_numa;
module_param_named(disable_numa, wq_disable_numa, bool, 0444);

287
/* see the comment above the definition of WQ_POWER_EFFICIENT */
288
static bool wq_power_efficient = IS_ENABLED(CONFIG_WQ_POWER_EFFICIENT_DEFAULT);
289 290
module_param_named(power_efficient, wq_power_efficient, bool, 0444);

291
static bool wq_online;			/* can kworkers be created yet? */
292

293 294
static bool wq_numa_enabled;		/* unbound NUMA affinity enabled */

295 296 297
/* buf for wq_update_unbound_numa_attrs(), protected by CPU hotplug exclusion */
static struct workqueue_attrs *wq_update_unbound_numa_attrs_buf;

298
static DEFINE_MUTEX(wq_pool_mutex);	/* protects pools and workqueues list */
299
static DEFINE_MUTEX(wq_pool_attach_mutex); /* protects worker attach/detach */
300
static DEFINE_SPINLOCK(wq_mayday_lock);	/* protects wq->maydays list */
301
static DECLARE_WAIT_QUEUE_HEAD(wq_manager_wait); /* wait for manager to go away */
302

303
static LIST_HEAD(workqueues);		/* PR: list of all workqueues */
304
static bool workqueue_freezing;		/* PL: have wqs started freezing? */
305

306 307 308 309 310
/* PL: allowable cpus for unbound wqs and work items */
static cpumask_var_t wq_unbound_cpumask;

/* CPU where unbound work was last round robin scheduled from this CPU */
static DEFINE_PER_CPU(int, wq_rr_cpu_last);
311

312 313 314 315 316 317 318 319 320 321 322 323
/*
 * Local execution of unbound work items is no longer guaranteed.  The
 * following always forces round-robin CPU selection on unbound work items
 * to uncover usages which depend on it.
 */
#ifdef CONFIG_DEBUG_WQ_FORCE_RR_CPU
static bool wq_debug_force_rr_cpu = true;
#else
static bool wq_debug_force_rr_cpu = false;
#endif
module_param_named(debug_force_rr_cpu, wq_debug_force_rr_cpu, bool, 0644);

324
/* the per-cpu worker pools */
325
static DEFINE_PER_CPU_SHARED_ALIGNED(struct worker_pool [NR_STD_WORKER_POOLS], cpu_worker_pools);
326

327
static DEFINE_IDR(worker_pool_idr);	/* PR: idr of all pools */
328

329
/* PL: hash of all unbound pools keyed by pool->attrs */
330 331
static DEFINE_HASHTABLE(unbound_pool_hash, UNBOUND_POOL_HASH_ORDER);

332
/* I: attributes used when instantiating standard unbound pools on demand */
333 334
static struct workqueue_attrs *unbound_std_wq_attrs[NR_STD_WORKER_POOLS];

335 336 337
/* I: attributes used when instantiating ordered pools on demand */
static struct workqueue_attrs *ordered_wq_attrs[NR_STD_WORKER_POOLS];

338
struct workqueue_struct *system_wq __read_mostly;
339
EXPORT_SYMBOL(system_wq);
340
struct workqueue_struct *system_highpri_wq __read_mostly;
341
EXPORT_SYMBOL_GPL(system_highpri_wq);
342
struct workqueue_struct *system_long_wq __read_mostly;
343
EXPORT_SYMBOL_GPL(system_long_wq);
344
struct workqueue_struct *system_unbound_wq __read_mostly;
345
EXPORT_SYMBOL_GPL(system_unbound_wq);
346
struct workqueue_struct *system_freezable_wq __read_mostly;
347
EXPORT_SYMBOL_GPL(system_freezable_wq);
348 349 350 351
struct workqueue_struct *system_power_efficient_wq __read_mostly;
EXPORT_SYMBOL_GPL(system_power_efficient_wq);
struct workqueue_struct *system_freezable_power_efficient_wq __read_mostly;
EXPORT_SYMBOL_GPL(system_freezable_power_efficient_wq);
352

353
static int worker_thread(void *__worker);
354
static void workqueue_sysfs_unregister(struct workqueue_struct *wq);
355

356 357 358
#define CREATE_TRACE_POINTS
#include <trace/events/workqueue.h>

359
#define assert_rcu_or_pool_mutex()					\
360 361 362
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held() &&			\
			 !lockdep_is_held(&wq_pool_mutex),		\
			 "sched RCU or wq_pool_mutex should be held")
363

364
#define assert_rcu_or_wq_mutex(wq)					\
365 366 367
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held() &&			\
			 !lockdep_is_held(&wq->mutex),			\
			 "sched RCU or wq->mutex should be held")
368

369
#define assert_rcu_or_wq_mutex_or_pool_mutex(wq)			\
370 371 372 373
	RCU_LOCKDEP_WARN(!rcu_read_lock_sched_held() &&			\
			 !lockdep_is_held(&wq->mutex) &&		\
			 !lockdep_is_held(&wq_pool_mutex),		\
			 "sched RCU, wq->mutex or wq_pool_mutex should be held")
374

375 376 377
#define for_each_cpu_worker_pool(pool, cpu)				\
	for ((pool) = &per_cpu(cpu_worker_pools, cpu)[0];		\
	     (pool) < &per_cpu(cpu_worker_pools, cpu)[NR_STD_WORKER_POOLS]; \
378
	     (pool)++)
379

380 381 382
/**
 * for_each_pool - iterate through all worker_pools in the system
 * @pool: iteration cursor
383
 * @pi: integer used for iteration
384
 *
385 386 387
 * This must be called either with wq_pool_mutex held or sched RCU read
 * locked.  If the pool needs to be used beyond the locking in effect, the
 * caller is responsible for guaranteeing that the pool stays online.
388 389 390
 *
 * The if/else clause exists only for the lockdep assertion and can be
 * ignored.
391
 */
392 393
#define for_each_pool(pool, pi)						\
	idr_for_each_entry(&worker_pool_idr, pool, pi)			\
394
		if (({ assert_rcu_or_pool_mutex(); false; })) { }	\
395
		else
396

397 398 399 400 401
/**
 * for_each_pool_worker - iterate through all workers of a worker_pool
 * @worker: iteration cursor
 * @pool: worker_pool to iterate workers of
 *
402
 * This must be called with wq_pool_attach_mutex.
403 404 405 406
 *
 * The if/else clause exists only for the lockdep assertion and can be
 * ignored.
 */
407 408
#define for_each_pool_worker(worker, pool)				\
	list_for_each_entry((worker), &(pool)->workers, node)		\
409
		if (({ lockdep_assert_held(&wq_pool_attach_mutex); false; })) { } \
410 411
		else

412 413 414 415
/**
 * for_each_pwq - iterate through all pool_workqueues of the specified workqueue
 * @pwq: iteration cursor
 * @wq: the target workqueue
416
 *
417
 * This must be called either with wq->mutex held or sched RCU read locked.
418 419
 * If the pwq needs to be used beyond the locking in effect, the caller is
 * responsible for guaranteeing that the pwq stays online.
420 421 422
 *
 * The if/else clause exists only for the lockdep assertion and can be
 * ignored.
423 424
 */
#define for_each_pwq(pwq, wq)						\
425
	list_for_each_entry_rcu((pwq), &(wq)->pwqs, pwqs_node)		\
426
		if (({ assert_rcu_or_wq_mutex(wq); false; })) { }	\
427
		else
428

429 430 431 432
#ifdef CONFIG_DEBUG_OBJECTS_WORK

static struct debug_obj_descr work_debug_descr;

433 434 435 436 437
static void *work_debug_hint(void *addr)
{
	return ((struct work_struct *) addr)->func;
}

438 439 440 441 442 443 444
static bool work_is_static_object(void *addr)
{
	struct work_struct *work = addr;

	return test_bit(WORK_STRUCT_STATIC_BIT, work_data_bits(work));
}

445 446 447 448
/*
 * fixup_init is called when:
 * - an active object is initialized
 */
449
static bool work_fixup_init(void *addr, enum debug_obj_state state)
450 451 452 453 454 455 456
{
	struct work_struct *work = addr;

	switch (state) {
	case ODEBUG_STATE_ACTIVE:
		cancel_work_sync(work);
		debug_object_init(work, &work_debug_descr);
457
		return true;
458
	default:
459
		return false;
460 461 462 463 464 465 466
	}
}

/*
 * fixup_free is called when:
 * - an active object is freed
 */
467
static bool work_fixup_free(void *addr, enum debug_obj_state state)
468 469 470 471 472 473 474
{
	struct work_struct *work = addr;

	switch (state) {
	case ODEBUG_STATE_ACTIVE:
		cancel_work_sync(work);
		debug_object_free(work, &work_debug_descr);
475
		return true;
476
	default:
477
		return false;
478 479 480 481 482
	}
}

static struct debug_obj_descr work_debug_descr = {
	.name		= "work_struct",
483
	.debug_hint	= work_debug_hint,
484
	.is_static_object = work_is_static_object,
485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513
	.fixup_init	= work_fixup_init,
	.fixup_free	= work_fixup_free,
};

static inline void debug_work_activate(struct work_struct *work)
{
	debug_object_activate(work, &work_debug_descr);
}

static inline void debug_work_deactivate(struct work_struct *work)
{
	debug_object_deactivate(work, &work_debug_descr);
}

void __init_work(struct work_struct *work, int onstack)
{
	if (onstack)
		debug_object_init_on_stack(work, &work_debug_descr);
	else
		debug_object_init(work, &work_debug_descr);
}
EXPORT_SYMBOL_GPL(__init_work);

void destroy_work_on_stack(struct work_struct *work)
{
	debug_object_free(work, &work_debug_descr);
}
EXPORT_SYMBOL_GPL(destroy_work_on_stack);

514 515 516 517 518 519 520
void destroy_delayed_work_on_stack(struct delayed_work *work)
{
	destroy_timer_on_stack(&work->timer);
	debug_object_free(&work->work, &work_debug_descr);
}
EXPORT_SYMBOL_GPL(destroy_delayed_work_on_stack);

521 522 523 524 525
#else
static inline void debug_work_activate(struct work_struct *work) { }
static inline void debug_work_deactivate(struct work_struct *work) { }
#endif

526 527 528 529 530 531 532
/**
 * worker_pool_assign_id - allocate ID and assing it to @pool
 * @pool: the pool pointer of interest
 *
 * Returns 0 if ID in [0, WORK_OFFQ_POOL_NONE) is allocated and assigned
 * successfully, -errno on failure.
 */
533 534 535 536
static int worker_pool_assign_id(struct worker_pool *pool)
{
	int ret;

537
	lockdep_assert_held(&wq_pool_mutex);
538

539 540
	ret = idr_alloc(&worker_pool_idr, pool, 0, WORK_OFFQ_POOL_NONE,
			GFP_KERNEL);
541
	if (ret >= 0) {
542
		pool->id = ret;
543 544
		return 0;
	}
545
	return ret;
546 547
}

548 549 550 551 552
/**
 * unbound_pwq_by_node - return the unbound pool_workqueue for the given node
 * @wq: the target workqueue
 * @node: the node ID
 *
553 554
 * This must be called with any of wq_pool_mutex, wq->mutex or sched RCU
 * read locked.
555 556
 * If the pwq needs to be used beyond the locking in effect, the caller is
 * responsible for guaranteeing that the pwq stays online.
557 558
 *
 * Return: The unbound pool_workqueue for @node.
559 560 561 562
 */
static struct pool_workqueue *unbound_pwq_by_node(struct workqueue_struct *wq,
						  int node)
{
563
	assert_rcu_or_wq_mutex_or_pool_mutex(wq);
564 565 566 567 568 569 570 571 572 573

	/*
	 * XXX: @node can be NUMA_NO_NODE if CPU goes offline while a
	 * delayed item is pending.  The plan is to keep CPU -> NODE
	 * mapping valid and stable across CPU on/offlines.  Once that
	 * happens, this workaround can be removed.
	 */
	if (unlikely(node == NUMA_NO_NODE))
		return wq->dfl_pwq;

574 575 576
	return rcu_dereference_raw(wq->numa_pwq_tbl[node]);
}

577 578 579 580 581 582 583 584 585 586 587 588 589 590 591
static unsigned int work_color_to_flags(int color)
{
	return color << WORK_STRUCT_COLOR_SHIFT;
}

static int get_work_color(struct work_struct *work)
{
	return (*work_data_bits(work) >> WORK_STRUCT_COLOR_SHIFT) &
		((1 << WORK_STRUCT_COLOR_BITS) - 1);
}

static int work_next_color(int color)
{
	return (color + 1) % WORK_NR_COLORS;
}
Linus Torvalds's avatar
Linus Torvalds committed
592

593
/*
594 595
 * While queued, %WORK_STRUCT_PWQ is set and non flag bits of a work's data
 * contain the pointer to the queued pwq.  Once execution starts, the flag
596
 * is cleared and the high bits contain OFFQ flags and pool ID.
597
 *
598 599
 * set_work_pwq(), set_work_pool_and_clear_pending(), mark_work_canceling()
 * and clear_work_data() can be used to set the pwq, pool or clear
600 601
 * work->data.  These functions should only be called while the work is
 * owned - ie. while the PENDING bit is set.
602
 *
603
 * get_work_pool() and get_work_pwq() can be used to obtain the pool or pwq
604
 * corresponding to a work.  Pool is available once the work has been
605
 * queued anywhere after initialization until it is sync canceled.  pwq is
606
 * available only while the work item is queued.
607
 *
608 609 610 611
 * %WORK_OFFQ_CANCELING is used to mark a work item which is being
 * canceled.  While being canceled, a work item may have its PENDING set
 * but stay off timer and worklist for arbitrarily long and nobody should
 * try to steal the PENDING bit.
612
 */
613 614
static inline void set_work_data(struct work_struct *work, unsigned long data,
				 unsigned long flags)
615
{
616
	WARN_ON_ONCE(!work_pending(work));
617 618
	atomic_long_set(&work->data, data | flags | work_static(work));
}
619

620
static void set_work_pwq(struct work_struct *work, struct pool_workqueue *pwq,
621 622
			 unsigned long extra_flags)
{
623 624
	set_work_data(work, (unsigned long)pwq,
		      WORK_STRUCT_PENDING | WORK_STRUCT_PWQ | extra_flags);
625 626
}

627 628 629 630 631 632 633
static void set_work_pool_and_keep_pending(struct work_struct *work,
					   int pool_id)
{
	set_work_data(work, (unsigned long)pool_id << WORK_OFFQ_POOL_SHIFT,
		      WORK_STRUCT_PENDING);
}

634 635
static void set_work_pool_and_clear_pending(struct work_struct *work,
					    int pool_id)
636
{
637 638 639 640 641 642 643
	/*
	 * The following wmb is paired with the implied mb in
	 * test_and_set_bit(PENDING) and ensures all updates to @work made
	 * here are visible to and precede any updates by the next PENDING
	 * owner.
	 */
	smp_wmb();
644
	set_work_data(work, (unsigned long)pool_id << WORK_OFFQ_POOL_SHIFT, 0);
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
	/*
	 * The following mb guarantees that previous clear of a PENDING bit
	 * will not be reordered with any speculative LOADS or STORES from
	 * work->current_func, which is executed afterwards.  This possible
	 * reordering can lead to a missed execution on attempt to qeueue
	 * the same @work.  E.g. consider this case:
	 *
	 *   CPU#0                         CPU#1
	 *   ----------------------------  --------------------------------
	 *
	 * 1  STORE event_indicated
	 * 2  queue_work_on() {
	 * 3    test_and_set_bit(PENDING)
	 * 4 }                             set_..._and_clear_pending() {
	 * 5                                 set_work_data() # clear bit
	 * 6                                 smp_mb()
	 * 7                               work->current_func() {
	 * 8				      LOAD event_indicated
	 *				   }
	 *
	 * Without an explicit full barrier speculative LOAD on line 8 can
	 * be executed before CPU#0 does STORE on line 1.  If that happens,
	 * CPU#0 observes the PENDING bit is still set and new execution of
	 * a @work is not queued in a hope, that CPU#1 will eventually
	 * finish the queued @work.  Meanwhile CPU#1 does not see
	 * event_indicated is set, because speculative LOAD was executed
	 * before actual STORE.
	 */
	smp_mb();
674
}
675

676
static void clear_work_data(struct work_struct *work)
Linus Torvalds's avatar
Linus Torvalds committed
677
{
678 679
	smp_wmb();	/* see set_work_pool_and_clear_pending() */
	set_work_data(work, WORK_STRUCT_NO_POOL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
680 681
}

682
static struct pool_workqueue *get_work_pwq(struct work_struct *work)
683
{
684
	unsigned long data = atomic_long_read(&work->data);
685

686
	if (data & WORK_STRUCT_PWQ)
687 688 689
		return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
	else
		return NULL;
690 691
}

692 693 694 695
/**
 * get_work_pool - return the worker_pool a given work was associated with
 * @work: the work item of interest
 *
696 697 698
 * Pools are created and destroyed under wq_pool_mutex, and allows read
 * access under sched-RCU read lock.  As such, this function should be
 * called under wq_pool_mutex or with preemption disabled.
699 700 701 702 703
 *
 * All fields of the returned pool are accessible as long as the above
 * mentioned locking is in effect.  If the returned pool needs to be used
 * beyond the critical section, the caller is responsible for ensuring the
 * returned pool is and stays online.
704 705
 *
 * Return: The worker_pool @work was last associated with.  %NULL if none.
706 707
 */
static struct worker_pool *get_work_pool(struct work_struct *work)
708
{
709
	unsigned long data = atomic_long_read(&work->data);
710
	int pool_id;
711

712
	assert_rcu_or_pool_mutex();
713

714 715
	if (data & WORK_STRUCT_PWQ)
		return ((struct pool_workqueue *)
716
			(data & WORK_STRUCT_WQ_DATA_MASK))->pool;
717

718 719
	pool_id = data >> WORK_OFFQ_POOL_SHIFT;
	if (pool_id == WORK_OFFQ_POOL_NONE)
720 721
		return NULL;

722
	return idr_find(&worker_pool_idr, pool_id);
723 724 725 726 727 728
}

/**
 * get_work_pool_id - return the worker pool ID a given work is associated with
 * @work: the work item of interest
 *
729
 * Return: The worker_pool ID @work was last associated with.
730 731 732 733
 * %WORK_OFFQ_POOL_NONE if none.
 */
static int get_work_pool_id(struct work_struct *work)
{
734 735
	unsigned long data = atomic_long_read(&work->data);

736 737
	if (data & WORK_STRUCT_PWQ)
		return ((struct pool_workqueue *)
738
			(data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
739

740
	return data >> WORK_OFFQ_POOL_SHIFT;
741 742
}

743 744
static void mark_work_canceling(struct work_struct *work)
{
745
	unsigned long pool_id = get_work_pool_id(work);
746

747 748
	pool_id <<= WORK_OFFQ_POOL_SHIFT;
	set_work_data(work, pool_id | WORK_OFFQ_CANCELING, WORK_STRUCT_PENDING);
749 750 751 752 753 754
}

static bool work_is_canceling(struct work_struct *work)
{
	unsigned long data = atomic_long_read(&work->data);

755
	return !(data & WORK_STRUCT_PWQ) && (data & WORK_OFFQ_CANCELING);
756 757
}

758
/*
759 760
 * Policy functions.  These define the policies on how the global worker
 * pools are managed.  Unless noted otherwise, these functions assume that
761
 * they're being called with pool->lock held.
762 763
 */

764
static bool __need_more_worker(struct worker_pool *pool)
765
{
766
	return !atomic_read(&pool->nr_running);
767 768
}

769
/*
770 771
 * Need to wake up a worker?  Called from anything but currently
 * running workers.
772 773
 *
 * Note that, because unbound workers never contribute to nr_running, this
774
 * function will always return %true for unbound pools as long as the
775
 * worklist isn't empty.
776
 */
777
static bool need_more_worker(struct worker_pool *pool)
778
{
779
	return !list_empty(&pool->worklist) && __need_more_worker(pool);
780
}
781

782
/* Can I start working?  Called from busy but !running workers. */
783
static bool may_start_working(struct worker_pool *pool)
784
{
785
	return pool->nr_idle;
786 787 788
}

/* Do I need to keep working?  Called from currently running workers. */
789
static bool keep_working(struct worker_pool *pool)
790
{
791 792
	return !list_empty(&pool->worklist) &&
		atomic_read(&pool->nr_running) <= 1;
793 794 795
}

/* Do we need a new worker?  Called from manager. */
796
static bool need_to_create_worker(struct worker_pool *pool)
797
{
798
	return need_more_worker(pool) && !may_start_working(pool);
799
}
800

801
/* Do we have too many workers and should some go away? */
802
static bool too_many_workers(struct worker_pool *pool)
803
{
804
	bool managing = pool->flags & POOL_MANAGER_ACTIVE;
805 806
	int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
	int nr_busy = pool->nr_workers - nr_idle;
807 808

	return nr_idle > 2 && (nr_idle - 2) * MAX_IDLE_WORKERS_RATIO >= nr_busy;
809 810
}

811
/*
812 813 814
 * Wake up functions.
 */

815 816
/* Return the first idle worker.  Safe with preemption disabled */
static struct worker *first_idle_worker(struct worker_pool *pool)
817
{
818
	if (unlikely(list_empty(&pool->idle_list)))
819 820
		return NULL;

821
	return list_first_entry(&pool->idle_list, struct worker, entry);
822 823 824 825
}

/**
 * wake_up_worker - wake up an idle worker
826
 * @pool: worker pool to wake worker from
827
 *
828
 * Wake up the first idle worker of @pool.
829 830
 *
 * CONTEXT:
831
 * spin_lock_irq(pool->lock).
832
 */
833
static void wake_up_worker(struct worker_pool *pool)
834
{
835
	struct worker *worker = first_idle_worker(pool);
836 837 838 839 840

	if (likely(worker))
		wake_up_process(worker->task);
}

841
/**
842 843 844 845 846 847 848 849 850 851
 * wq_worker_waking_up - a worker is waking up
 * @task: task waking up
 * @cpu: CPU @task is waking up to
 *
 * This function is called during try_to_wake_up() when a worker is
 * being awoken.
 *
 * CONTEXT:
 * spin_lock_irq(rq->lock)
 */
852
void wq_worker_waking_up(struct task_struct *task, int cpu)
853 854 855
{
	struct worker *worker = kthread_data(task);

856
	if (!(worker->flags & WORKER_NOT_RUNNING)) {
857
		WARN_ON_ONCE(worker->pool->cpu != cpu);
858
		atomic_inc(&worker->pool->nr_running);
859
	}
860 861 862 863 864 865 866 867 868 869 870 871 872
}

/**
 * wq_worker_sleeping - a worker is going to sleep
 * @task: task going to sleep
 *
 * This function is called during schedule() when a busy worker is
 * going to sleep.  Worker on the same cpu can be woken up by
 * returning pointer to its task.
 *
 * CONTEXT:
 * spin_lock_irq(rq->lock)
 *
873
 * Return:
874 875
 * Worker task on @cpu to wake up, %NULL if none.
 */
876
struct task_struct *wq_worker_sleeping(struct task_struct *task)
877 878
{
	struct worker *worker = kthread_data(task), *to_wakeup = NULL;
879
	struct worker_pool *pool;
880

881 882 883 884 885
	/*
	 * Rescuers, which may not have all the fields set up like normal
	 * workers, also reach here, let's not access anything before
	 * checking NOT_RUNNING.
	 */
886
	if (worker->flags & WORKER_NOT_RUNNING)
887 888
		return NULL;

889 890
	pool = worker->pool;

891
	/* this can only happen on the local cpu */
892
	if (WARN_ON_ONCE(pool->cpu != raw_smp_processor_id()))
893
		return NULL;
894 895 896 897 898 899

	/*
	 * The counterpart of the following dec_and_test, implied mb,
	 * worklist not empty test sequence is in insert_work().
	 * Please read comment there.
	 *
900 901 902
	 * 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
903
	 * manipulating idle_list, so dereferencing idle_list without pool
904
	 * lock is safe.
905
	 */
906 907
	if (atomic_dec_and_test(&pool->nr_running) &&
	    !list_empty(&pool->worklist))
908
		to_wakeup = first_idle_worker(pool);
909 910 911 912 913
	return to_wakeup ? to_wakeup->task : NULL;
}

/**
 * worker_set_flags - set worker flags and adjust nr_running accordingly
914
 * @worker: self
915 916
 * @flags: flags to set
 *
917
 * Set @flags in @worker->flags and adjust nr_running accordingly.
918
 *
919
 * CONTEXT:
920
 * spin_lock_irq(pool->lock)
921
 */
922
static inline void worker_set_flags(struct worker *worker, unsigned int flags)
923
{
924
	struct worker_pool *pool = worker->pool;
925

926 927
	WARN_ON_ONCE(worker->task != current);

928
	/* If transitioning into NOT_RUNNING, adjust nr_running. */
929 930
	if ((flags & WORKER_NOT_RUNNING) &&
	    !(worker->flags & WORKER_NOT_RUNNING)) {
931
		atomic_dec(&pool->nr_running);
932 933
	}

934 935 936 937
	worker->flags |= flags;
}

/**
938
 * worker_clr_flags - clear worker flags and adjust nr_running accordingly
939
 * @worker: self
940 941
 * @flags: flags to clear
 *
942
 * Clear @flags in @worker->flags and adjust nr_running accordingly.
943
 *
944
 * CONTEXT:
945
 * spin_lock_irq(pool->lock)
946 947 948
 */
static inline void worker_clr_flags(struct worker *worker, unsigned int flags)
{
949
	struct worker_pool *pool = worker->pool;
950 951
	unsigned int oflags = worker->flags;

952 953
	WARN_ON_ONCE(worker->task != current);

954
	worker->flags &= ~flags;
955

956 957 958 959 960
	/*
	 * If transitioning out of NOT_RUNNING, increment nr_running.  Note
	 * that the nested NOT_RUNNING is not a noop.  NOT_RUNNING is mask
	 * of multiple flags, not a single flag.
	 */
961 962
	if ((flags & WORKER_NOT_RUNNING) && (oflags & WORKER_NOT_RUNNING))
		if (!(worker->flags & WORKER_NOT_RUNNING))
963
			atomic_inc(&pool->nr_running);
964 965
}

966 967
/**
 * find_worker_executing_work - find worker which is executing a work
968
 * @pool: pool of interest
969 970
 * @work: work to find worker for
 *
971 972
 * Find a worker which is executing @work on @pool by searching
 * @pool->busy_hash which is keyed by the address of @work.  For a worker
973 974 975 976 977 978 979 980 981 982 983 984
 * to match, its current execution should match the address of @work and
 * its work function.  This is to avoid unwanted dependency between
 * unrelated work executions through a work item being recycled while still
 * being executed.
 *
 * This is a bit tricky.  A work item may be freed once its execution
 * starts and nothing prevents the freed area from being recycled for
 * another work item.  If the same work item address ends up being reused
 * before the original execution finishes, workqueue will identify the
 * recycled work item as currently executing and make it wait until the
 * current execution finishes, introducing an unwanted dependency.
 *
985 986 987 988 989 990
 * This function checks the work item address and work function to avoid
 * false positives.  Note that this isn't complete as one may construct a
 * work function which can introduce dependency onto itself through a
 * recycled work item.  Well, if somebody wants to shoot oneself in the
 * foot that badly, there's only so much we can do, and if such deadlock
 * actually occurs, it should be easy to locate the culprit work function.
991 992
 *
 * CONTEXT:
993
 * spin_lock_irq(pool->lock).
994
 *
995 996
 * Return:
 * Pointer to worker which is executing @work if found, %NULL
997
 * otherwise.
998
 */
999
static struct worker *find_worker_executing_work(struct worker_pool *pool,
1000
						 struct work_struct *work)
1001
{
1002 1003
	struct worker *worker;

1004
	hash_for_each_possible(pool->busy_hash, worker, hentry,
1005 1006 1007
			       (unsigned long)work)
		if (worker->current_work == work &&
		    worker->current_func == work->func)
1008 1009 1010
			return worker;

	return NULL;
1011 1012
}

1013 1014 1015 1016
/**
 * move_linked_works - move linked works to a list
 * @work: start of series of works to be scheduled
 * @head: target list to append @work to
1017
 * @nextp: out parameter for nested worklist walking
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
 *
 * Schedule linked works starting from @work to @head.  Work series to
 * be scheduled starts at @work and includes any consecutive work with
 * WORK_STRUCT_LINKED set in its predecessor.
 *
 * If @nextp is not NULL, it's updated to point to the next work of
 * the last scheduled work.  This allows move_linked_works() to be
 * nested inside outer list_for_each_entry_safe().
 *
 * CONTEXT:
1028
 * spin_lock_irq(pool->lock).
1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
 */
static void move_linked_works(struct work_struct *work, struct list_head *head,
			      struct work_struct **nextp)
{
	struct work_struct *n;

	/*
	 * Linked worklist will always end before the end of the list,
	 * use NULL for list head.
	 */
	list_for_each_entry_safe_from(work, n, NULL, entry) {
		list_move_tail(&work->entry, head);
		if (!(*work_data_bits(work) & WORK_STRUCT_LINKED))
			break;
	}

	/*
	 * If we're already inside safe list traversal and have moved
	 * multiple works to the scheduled queue, the next position
	 * needs to be updated.
	 */
	if (nextp)
		*nextp = n;
}

1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
/**
 * get_pwq - get an extra reference on the specified pool_workqueue
 * @pwq: pool_workqueue to get
 *
 * Obtain an extra reference on @pwq.  The caller should guarantee that
 * @pwq has positive refcnt and be holding the matching pool->lock.
 */
static void get_pwq(struct pool_workqueue *pwq)
{
	lockdep_assert_held(&pwq->pool->lock);
	WARN_ON_ONCE(pwq->refcnt <= 0);
	pwq->refcnt++;
}

/**
 * put_pwq - put a pool_workqueue reference
 * @pwq: pool_workqueue to put
 *
 * Drop a reference of @pwq.  If its refcnt reaches zero, schedule its
 * destruction.  The caller should be holding the matching pool->lock.
 */
static void put_pwq(struct pool_workqueue *pwq)
{
	lockdep_assert_held(&pwq->pool->lock);
	if (likely(--pwq->refcnt))
		return;
	if (WARN_ON_ONCE(!(pwq->wq->flags & WQ_UNBOUND)))
		return;
	/*
	 * @pwq can't be released under pool->lock, bounce to
	 * pwq_unbound_release_workfn().  This never recurses on the same
	 * pool->lock as this path is taken only for unbound workqueues and
	 * the release work item is scheduled on a per-cpu workqueue.  To
	 * avoid lockdep warning, unbound pool->locks are given lockdep
	 * subclass of 1 in get_unbound_pool().
	 */
	schedule_work(&pwq->unbound_release_work);
}

1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111
/**
 * put_pwq_unlocked - put_pwq() with surrounding pool lock/unlock
 * @pwq: pool_workqueue to put (can be %NULL)
 *
 * put_pwq() with locking.  This function also allows %NULL @pwq.
 */
static void put_pwq_unlocked(struct pool_workqueue *pwq)
{
	if (pwq) {
		/*
		 * As both pwqs and pools are sched-RCU protected, the
		 * following lock operations are safe.
		 */
		spin_lock_irq(&pwq->pool->lock);
		put_pwq(pwq);
		spin_unlock_irq(&pwq->pool->lock);
	}
}

1112
static void pwq_activate_delayed_work(struct work_struct *work)
1113
{
1114
	struct pool_workqueue *pwq = get_work_pwq(work);
1115 1116

	trace_workqueue_activate_work(work);
1117 1118
	if (list_empty(&pwq->pool->worklist))
		pwq->pool->watchdog_ts = jiffies;
1119
	move_linked_works(work, &pwq->pool->worklist, NULL);
1120
	__clear_bit(WORK_STRUCT_DELAYED_BIT, work_data_bits(work));
1121
	pwq->nr_active++;
1122 1123
}

1124
static void pwq_activate_first_delayed(struct pool_workqueue *pwq)
1125
{
1126
	struct work_struct *work = list_first_entry(&pwq->delayed_works,
1127 1128
						    struct work_struct, entry);

1129
	pwq_activate_delayed_work(work);
1130 1131
}

1132
/**
1133 1134
 * pwq_dec_nr_in_flight - decrement pwq's nr_in_flight
 * @pwq: pwq of interest
1135 1136 1137
 * @color: color of work which left the queue
 *
 * A work either has completed or is removed from pending queue,
1138
 * decrement nr_in_flight of its pwq and handle workqueue flushing.
1139 1140
 *
 * CONTEXT:
1141
 * spin_lock_irq(pool->lock).
1142
 */
1143
static void pwq_dec_nr_in_flight(struct pool_workqueue *pwq, int color)
1144
{
1145
	/* uncolored work items don't participate in flushing or nr_active */
1146
	if (color == WORK_NO_COLOR)
1147
		goto out_put;
1148

1149
	pwq->nr_in_flight[color]--;
1150

1151 1152
	pwq->nr_active--;
	if (!list_empty(&pwq->delayed_works)) {
1153
		/* one down, submit a delayed one */
1154 1155
		if (pwq->nr_active < pwq->max_active)
			pwq_activate_first_delayed(pwq);
1156 1157 1158
	}

	/* is flush in progress and are we at the flushing tip? */
1159
	if (likely(pwq->flush_color != color))
1160
		goto out_put;
1161 1162

	/* are there still in-flight works? */
1163
	if (pwq->nr_in_flight[color])
1164
		goto out_put;
1165

1166 1167
	/* this pwq is done, clear flush_color */
	pwq->flush_color = -1;
1168 1169

	/*
1170
	 * If this was the last pwq, wake up the first flusher.  It
1171 1172
	 * will handle the rest.
	 */
1173 1174
	if (atomic_dec_and_test(&pwq->wq->nr_pwqs_to_flush))
		complete(&pwq->wq->first_flusher->done);
1175 1176
out_put:
	put_pwq(pwq);
1177 1178
}

1179
/**
1180
 * try_to_grab_pending - steal work item from worklist and disable irq
1181 1182
 * @work: work item to steal
 * @is_dwork: @work is a delayed_work
1183
 * @flags: place to store irq state
1184 1185
 *
 * Try to grab PENDING bit of @work.  This function can handle @work in any
1186
 * stable state - idle, on timer or on worklist.
1187
 *
1188
 * Return:
1189 1190 1191
 *  1		if @work was pending and we successfully stole PENDING
 *  0		if @work was idle and we claimed PENDING
 *  -EAGAIN	if PENDING couldn't be grabbed at the moment, safe to busy-retry
1192 1193
 *  -ENOENT	if someone else is canceling @work, this state may persist
 *		for arbitrarily long
1194
 *
1195
 * Note:
1196
 * On >= 0 return, the caller owns @work's PENDING bit.  To avoid getting
1197 1198 1199
 * interrupted while holding PENDING and @work off queue, irq must be
 * disabled on entry.  This, combined with delayed_work->timer being
 * irqsafe, ensures that we return -EAGAIN for finite short period of time.
1200 1201 1202 1203
 *
 * On successful return, >= 0, irq is disabled and the caller is
 * responsible for releasing it using local_irq_restore(*@flags).
 *
1204
 * This function is safe to call from any context including IRQ handler.
1205
 */
1206 1207
static int try_to_grab_pending(struct work_struct *work, bool is_dwork,
			       unsigned long *flags)