tree_plugin.h 90.5 KB
Newer Older
1 2 3
/*
 * Read-Copy Update mechanism for mutual exclusion (tree-based version)
 * Internal non-public definitions that provide either classic
P
Paul E. McKenney 已提交
4
 * or preemptible semantics.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
17 18
 * along with this program; if not, you can access it online at
 * http://www.gnu.org/licenses/gpl-2.0.html.
19 20 21 22 23 24 25 26
 *
 * Copyright Red Hat, 2009
 * Copyright IBM Corporation, 2009
 *
 * Author: Ingo Molnar <mingo@elte.hu>
 *	   Paul E. McKenney <paulmck@linux.vnet.ibm.com>
 */

27
#include <linux/delay.h>
P
Paul E. McKenney 已提交
28
#include <linux/gfp.h>
29
#include <linux/oom.h>
30
#include <linux/smpboot.h>
31
#include "../time/tick-internal.h"
32

33
#ifdef CONFIG_RCU_BOOST
34

35
#include "../locking/rtmutex_common.h"
36

37 38 39 40 41 42 43 44 45
/*
 * Control variables for per-CPU and per-rcu_node kthreads.  These
 * handle all flavors of RCU.
 */
static DEFINE_PER_CPU(struct task_struct *, rcu_cpu_kthread_task);
DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);
DEFINE_PER_CPU(char, rcu_cpu_has_work);

46 47 48 49 50 51 52 53 54 55 56
#else /* #ifdef CONFIG_RCU_BOOST */

/*
 * Some architectures do not define rt_mutexes, but if !CONFIG_RCU_BOOST,
 * all uses are in dead code.  Provide a definition to keep the compiler
 * happy, but add WARN_ON_ONCE() to complain if used in the wrong place.
 * This probably needs to be excluded from -rt builds.
 */
#define rt_mutex_owner(a) ({ WARN_ON_ONCE(1); NULL; })

#endif /* #else #ifdef CONFIG_RCU_BOOST */
57

P
Paul E. McKenney 已提交
58 59 60
#ifdef CONFIG_RCU_NOCB_CPU
static cpumask_var_t rcu_nocb_mask; /* CPUs to have callbacks offloaded. */
static bool have_rcu_nocb_mask;	    /* Was rcu_nocb_mask allocated? */
61
static bool __read_mostly rcu_nocb_poll;    /* Offload kthread are to poll. */
P
Paul E. McKenney 已提交
62 63
#endif /* #ifdef CONFIG_RCU_NOCB_CPU */

64 65 66 67 68 69 70
/*
 * Check the RCU kernel configuration parameters and print informative
 * messages about anything out of the ordinary.  If you like #ifdef, you
 * will love this function.
 */
static void __init rcu_bootup_announce_oddness(void)
{
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88
	if (IS_ENABLED(CONFIG_RCU_TRACE))
		pr_info("\tRCU debugfs-based tracing is enabled.\n");
	if ((IS_ENABLED(CONFIG_64BIT) && CONFIG_RCU_FANOUT != 64) ||
	    (!IS_ENABLED(CONFIG_64BIT) && CONFIG_RCU_FANOUT != 32))
		pr_info("\tCONFIG_RCU_FANOUT set to non-default value of %d\n",
		       CONFIG_RCU_FANOUT);
	if (IS_ENABLED(CONFIG_RCU_FANOUT_EXACT))
		pr_info("\tHierarchical RCU autobalancing is disabled.\n");
	if (IS_ENABLED(CONFIG_RCU_FAST_NO_HZ))
		pr_info("\tRCU dyntick-idle grace-period acceleration is enabled.\n");
	if (IS_ENABLED(CONFIG_PROVE_RCU))
		pr_info("\tRCU lockdep checking is enabled.\n");
	if (IS_ENABLED(CONFIG_RCU_TORTURE_TEST_RUNNABLE))
		pr_info("\tRCU torture testing starts during boot.\n");
	if (IS_ENABLED(CONFIG_RCU_CPU_STALL_INFO))
		pr_info("\tAdditional per-CPU info printed with stalls.\n");
	if (NUM_RCU_LVL_4 != 0)
		pr_info("\tFour-level hierarchy is enabled.\n");
89 90 91
	if (CONFIG_RCU_FANOUT_LEAF != 16)
		pr_info("\tBuild-time adjustment of leaf fanout to %d.\n",
			CONFIG_RCU_FANOUT_LEAF);
92
	if (rcu_fanout_leaf != CONFIG_RCU_FANOUT_LEAF)
93
		pr_info("\tBoot-time adjustment of leaf fanout to %d.\n", rcu_fanout_leaf);
94
	if (nr_cpu_ids != NR_CPUS)
95
		pr_info("\tRCU restricting CPUs from NR_CPUS=%d to nr_cpu_ids=%d.\n", NR_CPUS, nr_cpu_ids);
96 97
	if (IS_ENABLED(CONFIG_RCU_BOOST))
		pr_info("\tRCU kthread priority: %d.\n", kthread_prio);
98 99
}

100
#ifdef CONFIG_PREEMPT_RCU
101

102
RCU_STATE_INITIALIZER(rcu_preempt, 'p', call_rcu);
103
static struct rcu_state *const rcu_state_p = &rcu_preempt_state;
104
static struct rcu_data __percpu *const rcu_data_p = &rcu_preempt_data;
105

106
static int rcu_preempted_readers_exp(struct rcu_node *rnp);
107 108
static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp,
			       bool wake);
109

110 111 112
/*
 * Tell them what RCU they are running.
 */
113
static void __init rcu_bootup_announce(void)
114
{
115
	pr_info("Preemptible hierarchical RCU implementation.\n");
116
	rcu_bootup_announce_oddness();
117 118 119
}

/*
P
Paul E. McKenney 已提交
120
 * Record a preemptible-RCU quiescent state for the specified CPU.  Note
121 122 123
 * that this just means that the task currently running on the CPU is
 * not in a quiescent state.  There might be any number of tasks blocked
 * while in an RCU read-side critical section.
124
 *
125 126
 * As with the other rcu_*_qs() functions, callers to this function
 * must disable preemption.
127
 */
128
static void rcu_preempt_qs(void)
129
{
130
	if (!__this_cpu_read(rcu_data_p->passed_quiesce)) {
131
		trace_rcu_grace_period(TPS("rcu_preempt"),
132
				       __this_cpu_read(rcu_data_p->gpnum),
133
				       TPS("cpuqs"));
134
		__this_cpu_write(rcu_data_p->passed_quiesce, 1);
135 136 137
		barrier(); /* Coordinate with rcu_preempt_check_callbacks(). */
		current->rcu_read_unlock_special.b.need_qs = false;
	}
138 139 140
}

/*
141 142 143
 * We have entered the scheduler, and the current task might soon be
 * context-switched away from.  If this task is in an RCU read-side
 * critical section, we will no longer be able to rely on the CPU to
144 145 146 147 148 149
 * record that fact, so we enqueue the task on the blkd_tasks list.
 * The task will dequeue itself when it exits the outermost enclosing
 * RCU read-side critical section.  Therefore, the current grace period
 * cannot be permitted to complete until the blkd_tasks list entries
 * predating the current grace period drain, in other words, until
 * rnp->gp_tasks becomes NULL.
150 151
 *
 * Caller must disable preemption.
152
 */
153
static void rcu_preempt_note_context_switch(void)
154 155
{
	struct task_struct *t = current;
156
	unsigned long flags;
157 158 159
	struct rcu_data *rdp;
	struct rcu_node *rnp;

160
	if (t->rcu_read_lock_nesting > 0 &&
161
	    !t->rcu_read_unlock_special.b.blocked) {
162 163

		/* Possibly blocking in an RCU read-side critical section. */
164
		rdp = this_cpu_ptr(rcu_state_p->rda);
165
		rnp = rdp->mynode;
P
Paul E. McKenney 已提交
166
		raw_spin_lock_irqsave(&rnp->lock, flags);
167
		smp_mb__after_unlock_lock();
168
		t->rcu_read_unlock_special.b.blocked = true;
169
		t->rcu_blocked_node = rnp;
170 171 172 173 174 175 176 177 178

		/*
		 * If this CPU has already checked in, then this task
		 * will hold up the next grace period rather than the
		 * current grace period.  Queue the task accordingly.
		 * If the task is queued for the current grace period
		 * (i.e., this CPU has not yet passed through a quiescent
		 * state for the current grace period), then as long
		 * as that task remains queued, the current grace period
179 180 181 182 183 184
		 * cannot end.  Note that there is some uncertainty as
		 * to exactly when the current grace period started.
		 * We take a conservative approach, which can result
		 * in unnecessarily waiting on tasks that started very
		 * slightly after the current grace period began.  C'est
		 * la vie!!!
185 186 187
		 *
		 * But first, note that the current CPU must still be
		 * on line!
188
		 */
189
		WARN_ON_ONCE((rdp->grpmask & rcu_rnp_online_cpus(rnp)) == 0);
190
		WARN_ON_ONCE(!list_empty(&t->rcu_node_entry));
191 192 193
		if ((rnp->qsmask & rdp->grpmask) && rnp->gp_tasks != NULL) {
			list_add(&t->rcu_node_entry, rnp->gp_tasks->prev);
			rnp->gp_tasks = &t->rcu_node_entry;
194 195
			if (IS_ENABLED(CONFIG_RCU_BOOST) &&
			    rnp->boost_tasks != NULL)
196
				rnp->boost_tasks = rnp->gp_tasks;
197 198 199 200 201
		} else {
			list_add(&t->rcu_node_entry, &rnp->blkd_tasks);
			if (rnp->qsmask & rdp->grpmask)
				rnp->gp_tasks = &t->rcu_node_entry;
		}
202 203 204 205 206
		trace_rcu_preempt_task(rdp->rsp->name,
				       t->pid,
				       (rnp->qsmask & rdp->grpmask)
				       ? rnp->gpnum
				       : rnp->gpnum + 1);
P
Paul E. McKenney 已提交
207
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
208
	} else if (t->rcu_read_lock_nesting < 0 &&
209
		   t->rcu_read_unlock_special.s) {
210 211 212 213 214 215

		/*
		 * Complete exit from RCU read-side critical section on
		 * behalf of preempted instance of __rcu_read_unlock().
		 */
		rcu_read_unlock_special(t);
216 217 218 219 220 221 222 223 224 225 226
	}

	/*
	 * Either we were not in an RCU read-side critical section to
	 * begin with, or we have now recorded that critical section
	 * globally.  Either way, we can now note a quiescent state
	 * for this CPU.  Again, if we were in an RCU read-side critical
	 * section, and if that critical section was blocking the current
	 * grace period, then the fact that the task has been enqueued
	 * means that we continue to block the current grace period.
	 */
227
	rcu_preempt_qs();
228 229
}

230 231 232 233 234
/*
 * Check for preempted RCU readers blocking the current grace period
 * for the specified rcu_node structure.  If the caller needs a reliable
 * answer, it must hold the rcu_node's ->lock.
 */
235
static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp)
236
{
237
	return rnp->gp_tasks != NULL;
238 239
}

240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
/*
 * Advance a ->blkd_tasks-list pointer to the next entry, instead
 * returning NULL if at the end of the list.
 */
static struct list_head *rcu_next_node_entry(struct task_struct *t,
					     struct rcu_node *rnp)
{
	struct list_head *np;

	np = t->rcu_node_entry.next;
	if (np == &rnp->blkd_tasks)
		np = NULL;
	return np;
}

255 256 257 258 259 260 261 262 263
/*
 * Return true if the specified rcu_node structure has tasks that were
 * preempted within an RCU read-side critical section.
 */
static bool rcu_preempt_has_tasks(struct rcu_node *rnp)
{
	return !list_empty(&rnp->blkd_tasks);
}

264 265 266 267 268
/*
 * Handle special cases during rcu_read_unlock(), such as needing to
 * notify RCU core processing or task having blocked during the RCU
 * read-side critical section.
 */
269
void rcu_read_unlock_special(struct task_struct *t)
270
{
271 272 273
	bool empty_exp;
	bool empty_norm;
	bool empty_exp_now;
274
	unsigned long flags;
275
	struct list_head *np;
276
	bool drop_boost_mutex = false;
277
	struct rcu_node *rnp;
278
	union rcu_special special;
279 280 281 282 283 284 285 286 287

	/* NMI handlers cannot block and cannot safely manipulate state. */
	if (in_nmi())
		return;

	local_irq_save(flags);

	/*
	 * If RCU core is waiting for this CPU to exit critical section,
288 289
	 * let it know that we have done so.  Because irqs are disabled,
	 * t->rcu_read_unlock_special cannot change.
290 291
	 */
	special = t->rcu_read_unlock_special;
292
	if (special.b.need_qs) {
293
		rcu_preempt_qs();
294
		t->rcu_read_unlock_special.b.need_qs = false;
295
		if (!t->rcu_read_unlock_special.s) {
296 297 298
			local_irq_restore(flags);
			return;
		}
299 300
	}

301
	/* Hardware IRQ handlers cannot block, complain if they get here. */
302 303 304 305 306 307 308
	if (in_irq() || in_serving_softirq()) {
		lockdep_rcu_suspicious(__FILE__, __LINE__,
				       "rcu_read_unlock() from irq or softirq with blocking in critical section!!!\n");
		pr_alert("->rcu_read_unlock_special: %#x (b: %d, nq: %d)\n",
			 t->rcu_read_unlock_special.s,
			 t->rcu_read_unlock_special.b.blocked,
			 t->rcu_read_unlock_special.b.need_qs);
309 310 311 312 313
		local_irq_restore(flags);
		return;
	}

	/* Clean up if blocked during RCU read-side critical section. */
314 315
	if (special.b.blocked) {
		t->rcu_read_unlock_special.b.blocked = false;
316

317 318 319 320 321 322
		/*
		 * Remove this task from the list it blocked on.  The
		 * task can migrate while we acquire the lock, but at
		 * most one time.  So at most two passes through loop.
		 */
		for (;;) {
323
			rnp = t->rcu_blocked_node;
P
Paul E. McKenney 已提交
324
			raw_spin_lock(&rnp->lock);  /* irqs already disabled. */
325
			smp_mb__after_unlock_lock();
326
			if (rnp == t->rcu_blocked_node)
327
				break;
P
Paul E. McKenney 已提交
328
			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
329
		}
330
		empty_norm = !rcu_preempt_blocked_readers_cgp(rnp);
331 332
		empty_exp = !rcu_preempted_readers_exp(rnp);
		smp_mb(); /* ensure expedited fastpath sees end of RCU c-s. */
333
		np = rcu_next_node_entry(t, rnp);
334
		list_del_init(&t->rcu_node_entry);
335
		t->rcu_blocked_node = NULL;
336
		trace_rcu_unlock_preempted_task(TPS("rcu_preempt"),
337
						rnp->gpnum, t->pid);
338 339 340 341
		if (&t->rcu_node_entry == rnp->gp_tasks)
			rnp->gp_tasks = np;
		if (&t->rcu_node_entry == rnp->exp_tasks)
			rnp->exp_tasks = np;
342 343 344 345 346 347
		if (IS_ENABLED(CONFIG_RCU_BOOST)) {
			if (&t->rcu_node_entry == rnp->boost_tasks)
				rnp->boost_tasks = np;
			/* Snapshot ->boost_mtx ownership w/rnp->lock held. */
			drop_boost_mutex = rt_mutex_owner(&rnp->boost_mtx) == t;
		}
348 349 350 351

		/*
		 * If this was the last task on the current list, and if
		 * we aren't waiting on any CPUs, report the quiescent state.
352 353
		 * Note that rcu_report_unblock_qs_rnp() releases rnp->lock,
		 * so we must take a snapshot of the expedited state.
354
		 */
355
		empty_exp_now = !rcu_preempted_readers_exp(rnp);
356
		if (!empty_norm && !rcu_preempt_blocked_readers_cgp(rnp)) {
357
			trace_rcu_quiescent_state_report(TPS("preempt_rcu"),
358 359 360 361 362 363
							 rnp->gpnum,
							 0, rnp->qsmask,
							 rnp->level,
							 rnp->grplo,
							 rnp->grphi,
							 !!rnp->gp_tasks);
364
			rcu_report_unblock_qs_rnp(rcu_state_p, rnp, flags);
365
		} else {
366
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
367
		}
368

369
		/* Unboost if we were boosted. */
370
		if (IS_ENABLED(CONFIG_RCU_BOOST) && drop_boost_mutex)
371
			rt_mutex_unlock(&rnp->boost_mtx);
372

373 374 375 376
		/*
		 * If this was the last task on the expedited lists,
		 * then we need to report up the rcu_node hierarchy.
		 */
377
		if (!empty_exp && empty_exp_now)
378
			rcu_report_exp_rnp(rcu_state_p, rnp, true);
379 380
	} else {
		local_irq_restore(flags);
381 382 383
	}
}

384 385 386 387 388 389 390 391 392
/*
 * Dump detailed information for all tasks blocking the current RCU
 * grace period on the specified rcu_node structure.
 */
static void rcu_print_detail_task_stall_rnp(struct rcu_node *rnp)
{
	unsigned long flags;
	struct task_struct *t;

393
	raw_spin_lock_irqsave(&rnp->lock, flags);
394 395 396 397
	if (!rcu_preempt_blocked_readers_cgp(rnp)) {
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
		return;
	}
398 399 400 401 402
	t = list_entry(rnp->gp_tasks,
		       struct task_struct, rcu_node_entry);
	list_for_each_entry_continue(t, &rnp->blkd_tasks, rcu_node_entry)
		sched_show_task(t);
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417
}

/*
 * Dump detailed information for all tasks blocking the current RCU
 * grace period.
 */
static void rcu_print_detail_task_stall(struct rcu_state *rsp)
{
	struct rcu_node *rnp = rcu_get_root(rsp);

	rcu_print_detail_task_stall_rnp(rnp);
	rcu_for_each_leaf_node(rsp, rnp)
		rcu_print_detail_task_stall_rnp(rnp);
}

418 419 420 421
#ifdef CONFIG_RCU_CPU_STALL_INFO

static void rcu_print_task_stall_begin(struct rcu_node *rnp)
{
422
	pr_err("\tTasks blocked on level-%d rcu_node (CPUs %d-%d):",
423 424 425 426 427
	       rnp->level, rnp->grplo, rnp->grphi);
}

static void rcu_print_task_stall_end(void)
{
428
	pr_cont("\n");
429 430 431 432 433 434 435 436 437 438 439 440 441 442
}

#else /* #ifdef CONFIG_RCU_CPU_STALL_INFO */

static void rcu_print_task_stall_begin(struct rcu_node *rnp)
{
}

static void rcu_print_task_stall_end(void)
{
}

#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_INFO */

443 444 445 446
/*
 * Scan the current list of tasks blocked within RCU read-side critical
 * sections, printing out the tid of each.
 */
447
static int rcu_print_task_stall(struct rcu_node *rnp)
448 449
{
	struct task_struct *t;
450
	int ndetected = 0;
451

452
	if (!rcu_preempt_blocked_readers_cgp(rnp))
453
		return 0;
454
	rcu_print_task_stall_begin(rnp);
455 456
	t = list_entry(rnp->gp_tasks,
		       struct task_struct, rcu_node_entry);
457
	list_for_each_entry_continue(t, &rnp->blkd_tasks, rcu_node_entry) {
458
		pr_cont(" P%d", t->pid);
459 460
		ndetected++;
	}
461
	rcu_print_task_stall_end();
462
	return ndetected;
463 464
}

465 466 467 468 469 470
/*
 * Check that the list of blocked tasks for the newly completed grace
 * period is in fact empty.  It is a serious bug to complete a grace
 * period that still has RCU readers blocked!  This function must be
 * invoked -before- updating this rnp's ->gpnum, and the rnp's ->lock
 * must be held by the caller.
471 472 473
 *
 * Also, if there are blocked tasks on the list, they automatically
 * block the newly created grace period, so set up ->gp_tasks accordingly.
474 475 476
 */
static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
{
477
	WARN_ON_ONCE(rcu_preempt_blocked_readers_cgp(rnp));
478
	if (rcu_preempt_has_tasks(rnp))
479
		rnp->gp_tasks = rnp->blkd_tasks.next;
480
	WARN_ON_ONCE(rnp->qsmask);
481 482
}

483 484 485 486 487 488 489
/*
 * Check for a quiescent state from the current CPU.  When a task blocks,
 * the task is recorded in the corresponding CPU's rcu_node structure,
 * which is checked elsewhere.
 *
 * Caller must disable hard irqs.
 */
490
static void rcu_preempt_check_callbacks(void)
491 492 493 494
{
	struct task_struct *t = current;

	if (t->rcu_read_lock_nesting == 0) {
495
		rcu_preempt_qs();
496 497
		return;
	}
498
	if (t->rcu_read_lock_nesting > 0 &&
499 500
	    __this_cpu_read(rcu_data_p->qs_pending) &&
	    !__this_cpu_read(rcu_data_p->passed_quiesce))
501
		t->rcu_read_unlock_special.b.need_qs = true;
502 503
}

504 505
#ifdef CONFIG_RCU_BOOST

506 507
static void rcu_preempt_do_callbacks(void)
{
508
	rcu_do_batch(rcu_state_p, this_cpu_ptr(rcu_data_p));
509 510
}

511 512
#endif /* #ifdef CONFIG_RCU_BOOST */

513
/*
P
Paul E. McKenney 已提交
514
 * Queue a preemptible-RCU callback for invocation after a grace period.
515 516 517
 */
void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
{
518
	__call_rcu(head, func, rcu_state_p, -1, 0);
519 520 521
}
EXPORT_SYMBOL_GPL(call_rcu);

522 523 524 525 526
/**
 * synchronize_rcu - wait until a grace period has elapsed.
 *
 * Control will return to the caller some time after a full grace
 * period has elapsed, in other words after all currently executing RCU
527 528 529 530 531
 * read-side critical sections have completed.  Note, however, that
 * upon return from synchronize_rcu(), the caller might well be executing
 * concurrently with new RCU read-side critical sections that began while
 * synchronize_rcu() was waiting.  RCU read-side critical sections are
 * delimited by rcu_read_lock() and rcu_read_unlock(), and may be nested.
532 533 534
 *
 * See the description of synchronize_sched() for more detailed information
 * on memory ordering guarantees.
535 536 537
 */
void synchronize_rcu(void)
{
538 539 540 541
	rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&
			   !lock_is_held(&rcu_lock_map) &&
			   !lock_is_held(&rcu_sched_lock_map),
			   "Illegal synchronize_rcu() in RCU read-side critical section");
542 543
	if (!rcu_scheduler_active)
		return;
544
	if (rcu_gp_is_expedited())
545 546 547
		synchronize_rcu_expedited();
	else
		wait_rcu_gp(call_rcu);
548 549 550
}
EXPORT_SYMBOL_GPL(synchronize_rcu);

551
static DECLARE_WAIT_QUEUE_HEAD(sync_rcu_preempt_exp_wq);
552
static unsigned long sync_rcu_preempt_exp_count;
553 554 555 556 557 558 559 560 561 562
static DEFINE_MUTEX(sync_rcu_preempt_exp_mutex);

/*
 * Return non-zero if there are any tasks in RCU read-side critical
 * sections blocking the current preemptible-RCU expedited grace period.
 * If there is no preemptible-RCU expedited grace period currently in
 * progress, returns zero unconditionally.
 */
static int rcu_preempted_readers_exp(struct rcu_node *rnp)
{
563
	return rnp->exp_tasks != NULL;
564 565 566 567 568 569 570 571 572 573 574 575 576 577
}

/*
 * return non-zero if there is no RCU expedited grace period in progress
 * for the specified rcu_node structure, in other words, if all CPUs and
 * tasks covered by the specified rcu_node structure have done their bit
 * for the current expedited grace period.  Works only for preemptible
 * RCU -- other RCU implementation use other means.
 *
 * Caller must hold sync_rcu_preempt_exp_mutex.
 */
static int sync_rcu_preempt_exp_done(struct rcu_node *rnp)
{
	return !rcu_preempted_readers_exp(rnp) &&
578
	       READ_ONCE(rnp->expmask) == 0;
579 580 581 582 583 584 585 586 587 588 589 590
}

/*
 * Report the exit from RCU read-side critical section for the last task
 * that queued itself during or before the current expedited preemptible-RCU
 * grace period.  This event is reported either to the rcu_node structure on
 * which the task was queued or to one of that rcu_node structure's ancestors,
 * recursively up the tree.  (Calm down, calm down, we do the recursion
 * iteratively!)
 *
 * Caller must hold sync_rcu_preempt_exp_mutex.
 */
591 592
static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp,
			       bool wake)
593 594 595 596
{
	unsigned long flags;
	unsigned long mask;

P
Paul E. McKenney 已提交
597
	raw_spin_lock_irqsave(&rnp->lock, flags);
598
	smp_mb__after_unlock_lock();
599
	for (;;) {
600 601
		if (!sync_rcu_preempt_exp_done(rnp)) {
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
602
			break;
603
		}
604
		if (rnp->parent == NULL) {
605
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
606 607
			if (wake) {
				smp_mb(); /* EGP done before wake_up(). */
608
				wake_up(&sync_rcu_preempt_exp_wq);
609
			}
610 611 612
			break;
		}
		mask = rnp->grpmask;
P
Paul E. McKenney 已提交
613
		raw_spin_unlock(&rnp->lock); /* irqs remain disabled */
614
		rnp = rnp->parent;
P
Paul E. McKenney 已提交
615
		raw_spin_lock(&rnp->lock); /* irqs already disabled */
616
		smp_mb__after_unlock_lock();
617 618 619 620 621 622
		rnp->expmask &= ~mask;
	}
}

/*
 * Snapshot the tasks blocking the newly started preemptible-RCU expedited
623 624 625 626
 * grace period for the specified rcu_node structure, phase 1.  If there
 * are such tasks, set the ->expmask bits up the rcu_node tree and also
 * set the ->expmask bits on the leaf rcu_node structures to tell phase 2
 * that work is needed here.
627
 *
628
 * Caller must hold sync_rcu_preempt_exp_mutex.
629 630
 */
static void
631
sync_rcu_preempt_exp_init1(struct rcu_state *rsp, struct rcu_node *rnp)
632
{
633
	unsigned long flags;
634 635
	unsigned long mask;
	struct rcu_node *rnp_up;
636

637
	raw_spin_lock_irqsave(&rnp->lock, flags);
638
	smp_mb__after_unlock_lock();
639 640
	WARN_ON_ONCE(rnp->expmask);
	WARN_ON_ONCE(rnp->exp_tasks);
641
	if (!rcu_preempt_has_tasks(rnp)) {
642
		/* No blocked tasks, nothing to do. */
643
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
		return;
	}
	/* Call for Phase 2 and propagate ->expmask bits up the tree. */
	rnp->expmask = 1;
	rnp_up = rnp;
	while (rnp_up->parent) {
		mask = rnp_up->grpmask;
		rnp_up = rnp_up->parent;
		if (rnp_up->expmask & mask)
			break;
		raw_spin_lock(&rnp_up->lock); /* irqs already off */
		smp_mb__after_unlock_lock();
		rnp_up->expmask |= mask;
		raw_spin_unlock(&rnp_up->lock); /* irqs still off */
	}
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
}

/*
 * Snapshot the tasks blocking the newly started preemptible-RCU expedited
 * grace period for the specified rcu_node structure, phase 2.  If the
 * leaf rcu_node structure has its ->expmask field set, check for tasks.
 * If there are some, clear ->expmask and set ->exp_tasks accordingly,
 * then initiate RCU priority boosting.  Otherwise, clear ->expmask and
 * invoke rcu_report_exp_rnp() to clear out the upper-level ->expmask bits,
 * enabling rcu_read_unlock_special() to do the bit-clearing.
 *
 * Caller must hold sync_rcu_preempt_exp_mutex.
 */
static void
sync_rcu_preempt_exp_init2(struct rcu_state *rsp, struct rcu_node *rnp)
{
	unsigned long flags;

	raw_spin_lock_irqsave(&rnp->lock, flags);
	smp_mb__after_unlock_lock();
	if (!rnp->expmask) {
		/* Phase 1 didn't do anything, so Phase 2 doesn't either. */
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
		return;
	}

	/* Phase 1 is over. */
	rnp->expmask = 0;

	/*
	 * If there are still blocked tasks, set up ->exp_tasks so that
	 * rcu_read_unlock_special() will wake us and then boost them.
	 */
	if (rcu_preempt_has_tasks(rnp)) {
694
		rnp->exp_tasks = rnp->blkd_tasks.next;
695
		rcu_initiate_boost(rnp, flags);  /* releases rnp->lock */
696
		return;
697
	}
698 699 700 701

	/* No longer any blocked tasks, so undo bit setting. */
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
	rcu_report_exp_rnp(rsp, rnp, false);
702 703
}

704 705 706 707 708 709 710 711 712 713 714
/**
 * synchronize_rcu_expedited - Brute-force RCU grace period
 *
 * Wait for an RCU-preempt grace period, but expedite it.  The basic
 * idea is to invoke synchronize_sched_expedited() to push all the tasks to
 * the ->blkd_tasks lists and wait for this list to drain.  This consumes
 * significant time on all CPUs and is unfriendly to real-time workloads,
 * so is thus not recommended for any sort of common-case code.
 * In fact, if you are using synchronize_rcu_expedited() in a loop,
 * please restructure your code to batch your updates, and then Use a
 * single synchronize_rcu() instead.
715 716 717
 */
void synchronize_rcu_expedited(void)
{
718
	struct rcu_node *rnp;
719
	struct rcu_state *rsp = rcu_state_p;
720
	unsigned long snap;
721 722 723
	int trycount = 0;

	smp_mb(); /* Caller's modifications seen first by other CPUs. */
724
	snap = READ_ONCE(sync_rcu_preempt_exp_count) + 1;
725 726
	smp_mb(); /* Above access cannot bleed into critical section. */

727 728 729 730 731 732 733 734
	/*
	 * Block CPU-hotplug operations.  This means that any CPU-hotplug
	 * operation that finds an rcu_node structure with tasks in the
	 * process of being boosted will know that all tasks blocking
	 * this expedited grace period will already be in the process of
	 * being boosted.  This simplifies the process of moving tasks
	 * from leaf to root rcu_node structures.
	 */
735 736 737 738 739
	if (!try_get_online_cpus()) {
		/* CPU-hotplug operation in flight, fall back to normal GP. */
		wait_rcu_gp(call_rcu);
		return;
	}
740

741 742 743 744 745 746
	/*
	 * Acquire lock, falling back to synchronize_rcu() if too many
	 * lock-acquisition failures.  Of course, if someone does the
	 * expedited grace period for us, just leave.
	 */
	while (!mutex_trylock(&sync_rcu_preempt_exp_mutex)) {
747
		if (ULONG_CMP_LT(snap,
748
		    READ_ONCE(sync_rcu_preempt_exp_count))) {
749 750 751
			put_online_cpus();
			goto mb_ret; /* Others did our work for us. */
		}
752
		if (trycount++ < 10) {
753
			udelay(trycount * num_online_cpus());
754
		} else {
755
			put_online_cpus();
756
			wait_rcu_gp(call_rcu);
757 758 759
			return;
		}
	}
760
	if (ULONG_CMP_LT(snap, READ_ONCE(sync_rcu_preempt_exp_count))) {
761
		put_online_cpus();
762
		goto unlock_mb_ret; /* Others did our work for us. */
763
	}
764

765
	/* force all RCU readers onto ->blkd_tasks lists. */
766 767
	synchronize_sched_expedited();

768 769 770 771 772 773
	/*
	 * Snapshot current state of ->blkd_tasks lists into ->expmask.
	 * Phase 1 sets bits and phase 2 permits rcu_read_unlock_special()
	 * to start clearing them.  Doing this in one phase leads to
	 * strange races between setting and clearing bits, so just say "no"!
	 */
774
	rcu_for_each_leaf_node(rsp, rnp)
775
		sync_rcu_preempt_exp_init1(rsp, rnp);
776
	rcu_for_each_leaf_node(rsp, rnp)
777
		sync_rcu_preempt_exp_init2(rsp, rnp);
778

779
	put_online_cpus();
780

781
	/* Wait for snapshotted ->blkd_tasks lists to drain. */
782 783 784 785 786 787
	rnp = rcu_get_root(rsp);
	wait_event(sync_rcu_preempt_exp_wq,
		   sync_rcu_preempt_exp_done(rnp));

	/* Clean up and exit. */
	smp_mb(); /* ensure expedited GP seen before counter increment. */
788
	WRITE_ONCE(sync_rcu_preempt_exp_count, sync_rcu_preempt_exp_count + 1);
789 790 791 792
unlock_mb_ret:
	mutex_unlock(&sync_rcu_preempt_exp_mutex);
mb_ret:
	smp_mb(); /* ensure subsequent action seen after grace period. */
793 794 795
}
EXPORT_SYMBOL_GPL(synchronize_rcu_expedited);

796 797
/**
 * rcu_barrier - Wait until all in-flight call_rcu() callbacks complete.
798 799 800 801 802
 *
 * Note that this primitive does not necessarily wait for an RCU grace period
 * to complete.  For example, if there are no RCU callbacks queued anywhere
 * in the system, then rcu_barrier() is within its rights to return
 * immediately, without waiting for anything, much less an RCU grace period.
803 804 805
 */
void rcu_barrier(void)
{
806
	_rcu_barrier(rcu_state_p);
807 808 809
}
EXPORT_SYMBOL_GPL(rcu_barrier);

810
/*
P
Paul E. McKenney 已提交
811
 * Initialize preemptible RCU's state structures.
812 813 814
 */
static void __init __rcu_init_preempt(void)
{
815
	rcu_init_one(rcu_state_p, rcu_data_p);
816 817
}

818 819 820 821 822 823 824 825 826 827 828 829 830 831
/*
 * Check for a task exiting while in a preemptible-RCU read-side
 * critical section, clean up if so.  No need to issue warnings,
 * as debug_check_no_locks_held() already does this if lockdep
 * is enabled.
 */
void exit_rcu(void)
{
	struct task_struct *t = current;

	if (likely(list_empty(&current->rcu_node_entry)))
		return;
	t->rcu_read_lock_nesting = 1;
	barrier();
832
	t->rcu_read_unlock_special.b.blocked = true;
833 834 835
	__rcu_read_unlock();
}

836
#else /* #ifdef CONFIG_PREEMPT_RCU */
837

838
static struct rcu_state *const rcu_state_p = &rcu_sched_state;
839
static struct rcu_data __percpu *const rcu_data_p = &rcu_sched_data;
840

841 842 843
/*
 * Tell them what RCU they are running.
 */
844
static void __init rcu_bootup_announce(void)
845
{
846
	pr_info("Hierarchical RCU implementation.\n");
847
	rcu_bootup_announce_oddness();
848 849
}

850 851 852 853
/*
 * Because preemptible RCU does not exist, we never have to check for
 * CPUs being in quiescent states.
 */
854
static void rcu_preempt_note_context_switch(void)
855 856 857
{
}

858
/*
P
Paul E. McKenney 已提交
859
 * Because preemptible RCU does not exist, there are never any preempted
860 861
 * RCU readers.
 */
862
static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp)
863 864 865 866
{
	return 0;
}

867 868 869 870
/*
 * Because there is no preemptible RCU, there can be no readers blocked.
 */
static bool rcu_preempt_has_tasks(struct rcu_node *rnp)
871
{
872
	return false;
873 874
}

875
/*
P
Paul E. McKenney 已提交
876
 * Because preemptible RCU does not exist, we never have to check for
877 878 879 880 881 882
 * tasks blocked within RCU read-side critical sections.
 */
static void rcu_print_detail_task_stall(struct rcu_state *rsp)
{
}

883
/*
P
Paul E. McKenney 已提交
884
 * Because preemptible RCU does not exist, we never have to check for
885 886
 * tasks blocked within RCU read-side critical sections.
 */
887
static int rcu_print_task_stall(struct rcu_node *rnp)
888
{
889
	return 0;
890 891
}

892
/*
P
Paul E. McKenney 已提交
893
 * Because there is no preemptible RCU, there can be no readers blocked,
894 895
 * so there is no need to check for blocked tasks.  So check only for
 * bogus qsmask values.
896 897 898
 */
static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
{
899
	WARN_ON_ONCE(rnp->qsmask);
900 901
}

902
/*
P
Paul E. McKenney 已提交
903
 * Because preemptible RCU does not exist, it never has any callbacks
904 905
 * to check.
 */
906
static void rcu_preempt_check_callbacks(void)
907 908 909
{
}

910 911
/*
 * Wait for an rcu-preempt grace period, but make it happen quickly.
P
Paul E. McKenney 已提交
912
 * But because preemptible RCU does not exist, map to rcu-sched.
913 914 915 916 917 918 919
 */
void synchronize_rcu_expedited(void)
{
	synchronize_sched_expedited();
}
EXPORT_SYMBOL_GPL(synchronize_rcu_expedited);

920
/*
P
Paul E. McKenney 已提交
921
 * Because preemptible RCU does not exist, rcu_barrier() is just
922 923 924 925 926 927 928 929
 * another name for rcu_barrier_sched().
 */
void rcu_barrier(void)
{
	rcu_barrier_sched();
}
EXPORT_SYMBOL_GPL(rcu_barrier);

930
/*
P
Paul E. McKenney 已提交
931
 * Because preemptible RCU does not exist, it need not be initialized.
932 933 934 935 936
 */
static void __init __rcu_init_preempt(void)
{
}

937 938 939 940 941 942 943 944
/*
 * Because preemptible RCU does not exist, tasks cannot possibly exit
 * while in preemptible RCU read-side critical sections.
 */
void exit_rcu(void)
{
}

945
#endif /* #else #ifdef CONFIG_PREEMPT_RCU */
946

947 948
#ifdef CONFIG_RCU_BOOST

949
#include "../locking/rtmutex_common.h"
950

951 952 953 954
#ifdef CONFIG_RCU_TRACE

static void rcu_initiate_boost_trace(struct rcu_node *rnp)
{
955
	if (!rcu_preempt_has_tasks(rnp))
956 957 958 959 960 961 962 963
		rnp->n_balk_blkd_tasks++;
	else if (rnp->exp_tasks == NULL && rnp->gp_tasks == NULL)
		rnp->n_balk_exp_gp_tasks++;
	else if (rnp->gp_tasks != NULL && rnp->boost_tasks != NULL)
		rnp->n_balk_boost_tasks++;
	else if (rnp->gp_tasks != NULL && rnp->qsmask != 0)
		rnp->n_balk_notblocked++;
	else if (rnp->gp_tasks != NULL &&
964
		 ULONG_CMP_LT(jiffies, rnp->boost_time))
965 966 967 968 969 970 971 972 973 974 975 976 977
		rnp->n_balk_notyet++;
	else
		rnp->n_balk_nos++;
}

#else /* #ifdef CONFIG_RCU_TRACE */

static void rcu_initiate_boost_trace(struct rcu_node *rnp)
{
}

#endif /* #else #ifdef CONFIG_RCU_TRACE */

T
Thomas Gleixner 已提交
978 979 980 981 982 983 984 985 986 987
static void rcu_wake_cond(struct task_struct *t, int status)
{
	/*
	 * If the thread is yielding, only wake it when this
	 * is invoked from idle
	 */
	if (status != RCU_KTHREAD_YIELDING || is_idle_task(current))
		wake_up_process(t);
}

988 989 990 991 992 993 994 995 996 997 998 999 1000 1001
/*
 * Carry out RCU priority boosting on the task indicated by ->exp_tasks
 * or ->boost_tasks, advancing the pointer to the next task in the
 * ->blkd_tasks list.
 *
 * Note that irqs must be enabled: boosting the task can block.
 * Returns 1 if there are more tasks needing to be boosted.
 */
static int rcu_boost(struct rcu_node *rnp)
{
	unsigned long flags;
	struct task_struct *t;
	struct list_head *tb;

1002 1003
	if (READ_ONCE(rnp->exp_tasks) == NULL &&
	    READ_ONCE(rnp->boost_tasks) == NULL)
1004 1005 1006
		return 0;  /* Nothing left to boost. */

	raw_spin_lock_irqsave(&rnp->lock, flags);
1007
	smp_mb__after_unlock_lock();
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023

	/*
	 * Recheck under the lock: all tasks in need of boosting
	 * might exit their RCU read-side critical sections on their own.
	 */
	if (rnp->exp_tasks == NULL && rnp->boost_tasks == NULL) {
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
		return 0;
	}

	/*
	 * Preferentially boost tasks blocking expedited grace periods.
	 * This cannot starve the normal grace periods because a second
	 * expedited grace period must boost all blocked tasks, including
	 * those blocking the pre-existing normal grace period.
	 */
1024
	if (rnp->exp_tasks != NULL) {
1025
		tb = rnp->exp_tasks;
1026 1027
		rnp->n_exp_boosts++;
	} else {
1028
		tb = rnp->boost_tasks;
1029 1030 1031
		rnp->n_normal_boosts++;
	}
	rnp->n_tasks_boosted++;
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049

	/*
	 * We boost task t by manufacturing an rt_mutex that appears to
	 * be held by task t.  We leave a pointer to that rt_mutex where
	 * task t can find it, and task t will release the mutex when it
	 * exits its outermost RCU read-side critical section.  Then
	 * simply acquiring this artificial rt_mutex will boost task
	 * t's priority.  (Thanks to tglx for suggesting this approach!)
	 *
	 * Note that task t must acquire rnp->lock to remove itself from
	 * the ->blkd_tasks list, which it will do from exit() if from
	 * nowhere else.  We therefore are guaranteed that task t will
	 * stay around at least until we drop rnp->lock.  Note that
	 * rnp->lock also resolves races between our priority boosting
	 * and task t's exiting its outermost RCU read-side critical
	 * section.
	 */
	t = container_of(tb, struct task_struct, rcu_node_entry);
1050
	rt_mutex_init_proxy_locked(&rnp->boost_mtx, t);
1051
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1052 1053 1054
	/* Lock only for side effect: boosts task t's priority. */
	rt_mutex_lock(&rnp->boost_mtx);
	rt_mutex_unlock(&rnp->boost_mtx);  /* Then keep lockdep happy. */
1055

1056 1057
	return READ_ONCE(rnp->exp_tasks) != NULL ||
	       READ_ONCE(rnp->boost_tasks) != NULL;
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
}

/*
 * Priority-boosting kthread.  One per leaf rcu_node and one for the
 * root rcu_node.
 */
static int rcu_boost_kthread(void *arg)
{
	struct rcu_node *rnp = (struct rcu_node *)arg;
	int spincnt = 0;
	int more2boost;

1070
	trace_rcu_utilization(TPS("Start boost kthread@init"));
1071
	for (;;) {
1072
		rnp->boost_kthread_status = RCU_KTHREAD_WAITING;
1073
		trace_rcu_utilization(TPS("End boost kthread@rcu_wait"));
1074
		rcu_wait(rnp->boost_tasks || rnp->exp_tasks);
1075
		trace_rcu_utilization(TPS("Start boost kthread@rcu_wait"));
1076
		rnp->boost_kthread_status = RCU_KTHREAD_RUNNING;
1077 1078 1079 1080 1081 1082
		more2boost = rcu_boost(rnp);
		if (more2boost)
			spincnt++;
		else
			spincnt = 0;
		if (spincnt > 10) {
T
Thomas Gleixner 已提交
1083
			rnp->boost_kthread_status = RCU_KTHREAD_YIELDING;
1084
			trace_rcu_utilization(TPS("End boost kthread@rcu_yield"));
T
Thomas Gleixner 已提交
1085
			schedule_timeout_interruptible(2);
1086
			trace_rcu_utilization(TPS("Start boost kthread@rcu_yield"));
1087 1088 1089
			spincnt = 0;
		}
	}
1090
	/* NOTREACHED */
1091
	trace_rcu_utilization(TPS("End boost kthread@notreached"));
1092 1093 1094 1095 1096 1097 1098 1099 1100
	return 0;
}

/*
 * Check to see if it is time to start boosting RCU readers that are
 * blocking the current grace period, and, if so, tell the per-rcu_node
 * kthread to start boosting them.  If there is an expedited grace
 * period in progress, it is always time to boost.
 *
1101 1102 1103
 * The caller must hold rnp->lock, which this function releases.
 * The ->boost_kthread_task is immortal, so we don't need to worry
 * about it going away.
1104
 */
1105
static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags)
1106
	__releases(rnp->lock)
1107 1108 1109
{
	struct task_struct *t;

1110 1111
	if (!rcu_preempt_blocked_readers_cgp(rnp) && rnp->exp_tasks == NULL) {
		rnp->n_balk_exp_gp_tasks++;
1112
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1113
		return;
1114
	}
1115 1116 1117 1118 1119 1120 1121
	if (rnp->exp_tasks != NULL ||
	    (rnp->gp_tasks != NULL &&
	     rnp->boost_tasks == NULL &&
	     rnp->qsmask == 0 &&
	     ULONG_CMP_GE(jiffies, rnp->boost_time))) {
		if (rnp->exp_tasks == NULL)
			rnp->boost_tasks = rnp->gp_tasks;
1122
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1123
		t = rnp->boost_kthread_task;
T
Thomas Gleixner 已提交
1124 1125
		if (t)
			rcu_wake_cond(t, rnp->boost_kthread_status);
1126
	} else {
1127
		rcu_initiate_boost_trace(rnp);
1128 1129
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
	}
1130 1131
}

1132 1133 1134 1135 1136 1137 1138 1139 1140
/*
 * Wake up the per-CPU kthread to invoke RCU callbacks.
 */
static void invoke_rcu_callbacks_kthread(void)
{
	unsigned long flags;

	local_irq_save(flags);
	__this_cpu_write(rcu_cpu_has_work, 1);
1141
	if (__this_cpu_read(rcu_cpu_kthread_task) != NULL &&
T
Thomas Gleixner 已提交
1142 1143 1144 1145
	    current != __this_cpu_read(rcu_cpu_kthread_task)) {
		rcu_wake_cond(__this_cpu_read(rcu_cpu_kthread_task),
			      __this_cpu_read(rcu_cpu_kthread_status));
	}
1146 1147 1148
	local_irq_restore(flags);
}

1149 1150 1151 1152 1153 1154
/*
 * Is the current CPU running the RCU-callbacks kthread?
 * Caller must have preemption disabled.
 */
static bool rcu_is_callbacks_kthread(void)
{
1155
	return __this_cpu_read(rcu_cpu_kthread_task) == current;
1156 1157
}

1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172
#define RCU_BOOST_DELAY_JIFFIES DIV_ROUND_UP(CONFIG_RCU_BOOST_DELAY * HZ, 1000)

/*
 * Do priority-boost accounting for the start of a new grace period.
 */
static void rcu_preempt_boost_start_gp(struct rcu_node *rnp)
{
	rnp->boost_time = jiffies + RCU_BOOST_DELAY_JIFFIES;
}

/*
 * Create an RCU-boost kthread for the specified node if one does not
 * already exist.  We only create this kthread for preemptible RCU.
 * Returns zero if all is well, a negated errno otherwise.
 */
1173
static int rcu_spawn_one_boost_kthread(struct rcu_state *rsp,
1174
				       struct rcu_node *rnp)
1175
{
T
Thomas Gleixner 已提交
1176
	int rnp_index = rnp - &rsp->node[0];
1177 1178 1179 1180
	unsigned long flags;
	struct sched_param sp;
	struct task_struct *t;

1181
	if (rcu_state_p != rsp)
1182
		return 0;
T
Thomas Gleixner 已提交
1183

1184
	if (!rcu_scheduler_fully_active || rcu_rnp_online_cpus(rnp) == 0)
T
Thomas Gleixner 已提交
1185 1186
		return 0;

1187
	rsp->boost = 1;
1188 1189 1190
	if (rnp->boost_kthread_task != NULL)
		return 0;
	t = kthread_create(rcu_boost_kthread, (void *)rnp,
1191
			   "rcub/%d", rnp_index);
1192 1193 1194
	if (IS_ERR(t))
		return PTR_ERR(t);
	raw_spin_lock_irqsave(&rnp->lock, flags);
1195
	smp_mb__after_unlock_lock();
1196 1197
	rnp->boost_kthread_task = t;
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1198
	sp.sched_priority = kthread_prio;
1199
	sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
1200
	wake_up_process(t); /* get to TASK_INTERRUPTIBLE quickly. */
1201 1202 1203
	return 0;
}

1204 1205
static void rcu_kthread_do_work(void)
{
1206 1207
	rcu_do_batch(&rcu_sched_state, this_cpu_ptr(&rcu_sched_data));
	rcu_do_batch(&rcu_bh_state, this_cpu_ptr(&rcu_bh_data));
1208 1209 1210
	rcu_preempt_do_callbacks();
}

1211
static void rcu_cpu_kthread_setup(unsigned int cpu)
1212 1213 1214
{
	struct sched_param sp;

1215
	sp.sched_priority = kthread_prio;
1216
	sched_setscheduler_nocheck(current, SCHED_FIFO, &sp);
1217 1218
}

1219
static void rcu_cpu_kthread_park(unsigned int cpu)
1220
{
1221
	per_cpu(rcu_cpu_kthread_status, cpu) = RCU_KTHREAD_OFFCPU;
1222 1223
}

1224
static int rcu_cpu_kthread_should_run(unsigned int cpu)
1225
{
1226
	return __this_cpu_read(rcu_cpu_has_work);
1227 1228 1229 1230
}

/*
 * Per-CPU kernel thread that invokes RCU callbacks.  This replaces the
1231 1232
 * RCU softirq used in flavors and configurations of RCU that do not
 * support RCU priority boosting.
1233
 */
1234
static void rcu_cpu_kthread(unsigned int cpu)
1235
{
1236 1237
	unsigned int *statusp = this_cpu_ptr(&rcu_cpu_kthread_status);
	char work, *workp = this_cpu_ptr(&rcu_cpu_has_work);
1238
	int spincnt;
1239

1240
	for (spincnt = 0; spincnt < 10; spincnt++) {
1241
		trace_rcu_utilization(TPS("Start CPU kthread@rcu_wait"));
1242 1243
		local_bh_disable();
		*statusp = RCU_KTHREAD_RUNNING;
1244 1245
		this_cpu_inc(rcu_cpu_kthread_loops);
		local_irq_disable();
1246 1247
		work = *workp;
		*workp = 0;
1248
		local_irq_enable();
1249 1250 1251
		if (work)
			rcu_kthread_do_work();
		local_bh_enable();
1252
		if (*workp == 0) {
1253
			trace_rcu_utilization(TPS("End CPU kthread@rcu_wait"));
1254 1255
			*statusp = RCU_KTHREAD_WAITING;
			return;
1256 1257
		}
	}
1258
	*statusp = RCU_KTHREAD_YIELDING;
1259
	trace_rcu_utilization(TPS("Start CPU kthread@rcu_yield"));
1260
	schedule_timeout_interruptible(2);
1261
	trace_rcu_utilization(TPS("End CPU kthread@rcu_yield"));
1262
	*statusp = RCU_KTHREAD_WAITING;
1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273
}

/*
 * Set the per-rcu_node kthread's affinity to cover all CPUs that are
 * served by the rcu_node in question.  The CPU hotplug lock is still
 * held, so the value of rnp->qsmaskinit will be stable.
 *
 * We don't include outgoingcpu in the affinity set, use -1 if there is
 * no outgoing CPU.  If there are no CPUs left in the affinity set,
 * this function allows the kthread to execute on any CPU.
 */
T
Thomas Gleixner 已提交
1274
static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu)
1275
{
T
Thomas Gleixner 已提交
1276
	struct task_struct *t = rnp->boost_kthread_task;
1277
	unsigned long mask = rcu_rnp_online_cpus(rnp);
1278 1279 1280
	cpumask_var_t cm;
	int cpu;

T
Thomas Gleixner 已提交
1281
	if (!t)
1282
		return;
T
Thomas Gleixner 已提交
1283
	if (!zalloc_cpumask_var(&cm, GFP_KERNEL))
1284 1285 1286 1287
		return;
	for (cpu = rnp->grplo; cpu <= rnp->grphi; cpu++, mask >>= 1)
		if ((mask & 0x1) && cpu != outgoingcpu)
			cpumask_set_cpu(cpu, cm);
1288
	if (cpumask_weight(cm) == 0)
1289
		cpumask_setall(cm);
T
Thomas Gleixner 已提交
1290
	set_cpus_allowed_ptr(t, cm);
1291 1292 1293
	free_cpumask_var(cm);
}

1294 1295 1296 1297 1298 1299 1300 1301
static struct smp_hotplug_thread rcu_cpu_thread_spec = {
	.store			= &rcu_cpu_kthread_task,
	.thread_should_run	= rcu_cpu_kthread_should_run,
	.thread_fn		= rcu_cpu_kthread,
	.thread_comm		= "rcuc/%u",
	.setup			= rcu_cpu_kthread_setup,
	.park			= rcu_cpu_kthread_park,
};
1302 1303

/*
1304
 * Spawn boost kthreads -- called as soon as the scheduler is running.
1305
 */
1306
static void __init rcu_spawn_boost_kthreads(void)
1307 1308
{
	struct rcu_node *rnp;
T
Thomas Gleixner 已提交
1309
	int cpu;
1310

1311
	for_each_possible_cpu(cpu)
1312
		per_cpu(rcu_cpu_has_work, cpu) = 0;
1313
	BUG_ON(smpboot_register_percpu_thread(&rcu_cpu_thread_spec));
1314 1315
	rcu_for_each_leaf_node(rcu_state_p, rnp)
		(void)rcu_spawn_one_boost_kthread(rcu_state_p, rnp);
1316 1317
}

1318
static void rcu_prepare_kthreads(int cpu)
1319
{
1320
	struct rcu_data *rdp = per_cpu_ptr(rcu_state_p->rda, cpu);
1321 1322 1323
	struct rcu_node *rnp = rdp->mynode;

	/* Fire up the incoming CPU's kthread and leaf rcu_node kthread. */
1324
	if (rcu_scheduler_fully_active)
1325
		(void)rcu_spawn_one_boost_kthread(rcu_state_p, rnp);
1326 1327
}

1328 1329
#else /* #ifdef CONFIG_RCU_BOOST */

1330
static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags)
1331
	__releases(rnp->lock)
1332
{
1333
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1334 1335
}

1336
static void invoke_rcu_callbacks_kthread(void)
1337
{
1338
	WARN_ON_ONCE(1);
1339 1340
}

1341 1342 1343 1344 1345
static bool rcu_is_callbacks_kthread(void)
{
	return false;
}

1346 1347 1348 1349
static void rcu_preempt_boost_start_gp(struct rcu_node *rnp)
{
}

T
Thomas Gleixner 已提交
1350
static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu)
1351 1352 1353
{
}

1354
static void __init rcu_spawn_boost_kthreads(void)
1355 1356 1357
{
}

1358
static void rcu_prepare_kthreads(int cpu)
1359 1360 1361
{
}

1362 1363
#endif /* #else #ifdef CONFIG_RCU_BOOST */

1364 1365 1366 1367 1368 1369 1370 1371
#if !defined(CONFIG_RCU_FAST_NO_HZ)

/*
 * Check to see if any future RCU-related work will need to be done
 * by the current CPU, even if none need be done immediately, returning
 * 1 if so.  This function is part of the RCU implementation; it is -not-
 * an exported member of the RCU API.
 *
1372 1373
 * Because we not have RCU_FAST_NO_HZ, just check whether this CPU needs
 * any flavor of RCU.
1374
 */
1375
#ifndef CONFIG_RCU_NOCB_CPU_ALL
1376
int rcu_needs_cpu(unsigned long *delta_jiffies)
1377
{
1378
	*delta_jiffies = ULONG_MAX;
1379
	return rcu_cpu_has_callbacks(NULL);
1380
}
1381
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_ALL */
1382 1383 1384 1385 1386

/*
 * Because we do not have RCU_FAST_NO_HZ, don't bother cleaning up
 * after it.
 */
1387
static void rcu_cleanup_after_idle(void)
1388 1389 1390
{
}

1391
/*
1392
 * Do the idle-entry grace-period work, which, because CONFIG_RCU_FAST_NO_HZ=n,
1393 1394
 * is nothing.
 */
1395
static void rcu_prepare_for_idle(void)
1396 1397 1398
{
}

1399 1400 1401 1402 1403 1404 1405 1406
/*
 * Don't bother keeping a running count of the number of RCU callbacks
 * posted because CONFIG_RCU_FAST_NO_HZ=n.
 */
static void rcu_idle_count_callbacks_posted(void)
{
}

1407 1408
#else /* #if !defined(CONFIG_RCU_FAST_NO_HZ) */

1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423
/*
 * This code is invoked when a CPU goes idle, at which point we want
 * to have the CPU do everything required for RCU so that it can enter
 * the energy-efficient dyntick-idle mode.  This is handled by a
 * state machine implemented by rcu_prepare_for_idle() below.
 *
 * The following three proprocessor symbols control this state machine:
 *
 * RCU_IDLE_GP_DELAY gives the number of jiffies that a CPU is permitted
 *	to sleep in dyntick-idle mode with RCU callbacks pending.  This
 *	is sized to be roughly one RCU grace period.  Those energy-efficiency
 *	benchmarkers who might otherwise be tempted to set this to a large
 *	number, be warned: Setting RCU_IDLE_GP_DELAY too high can hang your
 *	system.  And if you are -that- concerned about energy efficiency,
 *	just power the system down and be done with it!
1424 1425 1426
 * RCU_IDLE_LAZY_GP_DELAY gives the number of jiffies that a CPU is
 *	permitted to sleep in dyntick-idle mode with only lazy RCU
 *	callbacks pending.  Setting this too high can OOM your system.
1427 1428 1429 1430 1431
 *
 * The values below work well in practice.  If future workloads require
 * adjustment, they can be converted into kernel config parameters, though
 * making the state machine smarter might be a better option.
 */
1432
#define RCU_IDLE_GP_DELAY 4		/* Roughly one grace period. */
1433
#define RCU_IDLE_LAZY_GP_DELAY (6 * HZ)	/* Roughly six seconds. */
1434

1435 1436 1437 1438
static int rcu_idle_gp_delay = RCU_IDLE_GP_DELAY;
module_param(rcu_idle_gp_delay, int, 0644);
static int rcu_idle_lazy_gp_delay = RCU_IDLE_LAZY_GP_DELAY;
module_param(rcu_idle_lazy_gp_delay, int, 0644);
1439

1440
extern int tick_nohz_active;
1441 1442

/*
1443 1444 1445
 * Try to advance callbacks for all flavors of RCU on the current CPU, but
 * only if it has been awhile since the last time we did so.  Afterwards,
 * if there are any callbacks ready for immediate invocation, return true.
1446
 */
1447
static bool __maybe_unused rcu_try_advance_all_cbs(void)
1448
{
1449 1450
	bool cbs_ready = false;
	struct rcu_data *rdp;
1451
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1452 1453
	struct rcu_node *rnp;
	struct rcu_state *rsp;
1454

1455 1456
	/* Exit early if we advanced recently. */
	if (jiffies == rdtp->last_advance_all)
1457
		return false;
1458 1459
	rdtp->last_advance_all = jiffies;

1460 1461 1462
	for_each_rcu_flavor(rsp) {
		rdp = this_cpu_ptr(rsp->rda);
		rnp = rdp->mynode;
1463

1464 1465 1466 1467 1468
		/*
		 * Don't bother checking unless a grace period has
		 * completed since we last checked and there are
		 * callbacks not yet ready to invoke.
		 */
1469
		if ((rdp->completed != rnp->completed ||
1470
		     unlikely(READ_ONCE(rdp->gpwrap))) &&
1471
		    rdp->nxttail[RCU_DONE_TAIL] != rdp->nxttail[RCU_NEXT_TAIL])
1472
			note_gp_changes(rsp, rdp);
1473

1474 1475 1476 1477
		if (cpu_has_callbacks_ready_to_invoke(rdp))
			cbs_ready = true;
	}
	return cbs_ready;
1478 1479
}

1480
/*
1481 1482 1483 1484
 * Allow the CPU to enter dyntick-idle mode unless it has callbacks ready
 * to invoke.  If the CPU has callbacks, try to advance them.  Tell the
 * caller to set the timeout based on whether or not there are non-lazy
 * callbacks.
1485
 *
1486
 * The caller must have disabled interrupts.
1487
 */
1488
#ifndef CONFIG_RCU_NOCB_CPU_ALL
1489
int rcu_needs_cpu(unsigned long *dj)
1490
{
1491
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1492

1493 1494 1495
	/* Snapshot to detect later posting of non-lazy callback. */
	rdtp->nonlazy_posted_snap = rdtp->nonlazy_posted;

1496
	/* If no callbacks, RCU doesn't need the CPU. */
1497
	if (!rcu_cpu_has_callbacks(&rdtp->all_lazy)) {
1498
		*dj = ULONG_MAX;
1499 1500
		return 0;
	}
1501 1502 1503 1504 1505

	/* Attempt to advance callbacks. */
	if (rcu_try_advance_all_cbs()) {
		/* Some ready to invoke, so initiate later invocation. */
		invoke_rcu_core();
1506 1507
		return 1;
	}
1508 1509 1510
	rdtp->last_accelerate = jiffies;

	/* Request timer delay depending on laziness, and round. */
1511
	if (!rdtp->all_lazy) {
1512 1513
		*dj = round_up(rcu_idle_gp_delay + jiffies,
			       rcu_idle_gp_delay) - jiffies;
1514
	} else {
1515
		*dj = round_jiffies(rcu_idle_lazy_gp_delay + jiffies) - jiffies;
1516
	}
1517 1518
	return 0;
}
1519
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_ALL */
1520

1521
/*
1522 1523 1524 1525 1526 1527
 * Prepare a CPU for idle from an RCU perspective.  The first major task
 * is to sense whether nohz mode has been enabled or disabled via sysfs.
 * The second major task is to check to see if a non-lazy callback has
 * arrived at a CPU that previously had only lazy callbacks.  The third
 * major task is to accelerate (that is, assign grace-period numbers to)
 * any recently arrived callbacks.
1528 1529
 *
 * The caller must have disabled interrupts.
1530
 */
1531
static void rcu_prepare_for_idle(void)
1532
{
1533
#ifndef CONFIG_RCU_NOCB_CPU_ALL
1534
	bool needwake;
1535
	struct rcu_data *rdp;
1536
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1537 1538
	struct rcu_node *rnp;
	struct rcu_state *rsp;
1539 1540 1541
	int tne;

	/* Handle nohz enablement switches conservatively. */
1542
	tne = READ_ONCE(tick_nohz_active);
1543
	if (tne != rdtp->tick_nohz_enabled_snap) {
1544
		if (rcu_cpu_has_callbacks(NULL))
1545 1546 1547 1548 1549 1550
			invoke_rcu_core(); /* force nohz to see update. */
		rdtp->tick_nohz_enabled_snap = tne;
		return;
	}
	if (!tne)
		return;
1551

1552
	/* If this is a no-CBs CPU, no callbacks, just return. */
1553
	if (rcu_is_nocb_cpu(smp_processor_id()))
1554 1555
		return;

1556
	/*
1557 1558 1559
	 * If a non-lazy callback arrived at a CPU having only lazy
	 * callbacks, invoke RCU core for the side-effect of recalculating
	 * idle duration on re-entry to idle.
1560
	 */
1561 1562
	if (rdtp->all_lazy &&
	    rdtp->nonlazy_posted != rdtp->nonlazy_posted_snap) {
1563 1564
		rdtp->all_lazy = false;
		rdtp->nonlazy_posted_snap = rdtp->nonlazy_posted;
1565
		invoke_rcu_core();
1566 1567 1568
		return;
	}

1569
	/*
1570 1571
	 * If we have not yet accelerated this jiffy, accelerate all
	 * callbacks on this CPU.
1572
	 */
1573
	if (rdtp->last_accelerate == jiffies)
1574
		return;
1575 1576
	rdtp->last_accelerate = jiffies;
	for_each_rcu_flavor(rsp) {
1577
		rdp = this_cpu_ptr(rsp->rda);
1578 1579 1580 1581
		if (!*rdp->nxttail[RCU_DONE_TAIL])
			continue;
		rnp = rdp->mynode;
		raw_spin_lock(&rnp->lock); /* irqs already disabled. */
1582
		smp_mb__after_unlock_lock();
1583
		needwake = rcu_accelerate_cbs(rsp, rnp, rdp);
1584
		raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
1585 1586
		if (needwake)
			rcu_gp_kthread_wake(rsp);
1587
	}
1588
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_ALL */
1589
}
1590

1591 1592 1593 1594 1595
/*
 * Clean up for exit from idle.  Attempt to advance callbacks based on
 * any grace periods that elapsed while the CPU was idle, and if any
 * callbacks are now ready to invoke, initiate invocation.
 */
1596
static void rcu_cleanup_after_idle(void)
1597
{
1598
#ifndef CONFIG_RCU_NOCB_CPU_ALL
1599
	if (rcu_is_nocb_cpu(smp_processor_id()))
1600
		return;
1601 1602
	if (rcu_try_advance_all_cbs())
		invoke_rcu_core();
1603
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_ALL */
1604 1605
}

1606
/*
1607 1608 1609 1610 1611 1612
 * Keep a running count of the number of non-lazy callbacks posted
 * on this CPU.  This running counter (which is never decremented) allows
 * rcu_prepare_for_idle() to detect when something out of the idle loop
 * posts a callback, even if an equal number of callbacks are invoked.
 * Of course, callbacks should only be posted from within a trace event
 * designed to be called from idle or from within RCU_NONIDLE().
1613 1614 1615
 */
static void rcu_idle_count_callbacks_posted(void)
{
1616
	__this_cpu_add(rcu_dynticks.nonlazy_posted, 1);
1617 1618
}

1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
/*
 * Data for flushing lazy RCU callbacks at OOM time.
 */
static atomic_t oom_callback_count;
static DECLARE_WAIT_QUEUE_HEAD(oom_callback_wq);

/*
 * RCU OOM callback -- decrement the outstanding count and deliver the
 * wake-up if we are the last one.
 */
static void rcu_oom_callback(struct rcu_head *rhp)
{
	if (atomic_dec_and_test(&oom_callback_count))
		wake_up(&oom_callback_wq);
}

/*
 * Post an rcu_oom_notify callback on the current CPU if it has at
 * least one lazy callback.  This will unnecessarily post callbacks
 * to CPUs that already have a non-lazy callback at the end of their
 * callback list, but this is an infrequent operation, so accept some
 * extra overhead to keep things simple.
 */
static void rcu_oom_notify_cpu(void *unused)
{
	struct rcu_state *rsp;
	struct rcu_data *rdp;

	for_each_rcu_flavor(rsp) {
1648
		rdp = raw_cpu_ptr(rsp->rda);
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
		if (rdp->qlen_lazy != 0) {
			atomic_inc(&oom_callback_count);
			rsp->call(&rdp->oom_head, rcu_oom_callback);
		}
	}
}

/*
 * If low on memory, ensure that each CPU has a non-lazy callback.
 * This will wake up CPUs that have only lazy callbacks, in turn
 * ensuring that they free up the corresponding memory in a timely manner.
 * Because an uncertain amount of memory will be freed in some uncertain
 * timeframe, we do not claim to have freed anything.
 */
static int rcu_oom_notify(struct notifier_block *self,
			  unsigned long notused, void *nfreed)
{
	int cpu;

	/* Wait for callbacks from earlier instance to complete. */
	wait_event(oom_callback_wq, atomic_read(&oom_callback_count) == 0);
1670
	smp_mb(); /* Ensure callback reuse happens after callback invocation. */
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680

	/*
	 * Prevent premature wakeup: ensure that all increments happen
	 * before there is a chance of the counter reaching zero.
	 */
	atomic_set(&oom_callback_count, 1);

	get_online_cpus();
	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu, rcu_oom_notify_cpu, NULL, 1);
1681
		cond_resched_rcu_qs();
1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
	}
	put_online_cpus();

	/* Unconditionally decrement: no need to wake ourselves up. */
	atomic_dec(&oom_callback_count);

	return NOTIFY_OK;
}

static struct notifier_block rcu_oom_nb = {
	.notifier_call = rcu_oom_notify
};

static int __init rcu_register_oom_notifier(void)
{
	register_oom_notifier(&rcu_oom_nb);
	return 0;
}
early_initcall(rcu_register_oom_notifier);

1702
#endif /* #else #if !defined(CONFIG_RCU_FAST_NO_HZ) */
1703 1704 1705 1706 1707 1708 1709

#ifdef CONFIG_RCU_CPU_STALL_INFO

#ifdef CONFIG_RCU_FAST_NO_HZ

static void print_cpu_stall_fast_no_hz(char *cp, int cpu)
{
1710
	struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);
1711
	unsigned long nlpd = rdtp->nonlazy_posted - rdtp->nonlazy_posted_snap;
1712

1713 1714 1715 1716 1717
	sprintf(cp, "last_accelerate: %04lx/%04lx, nonlazy_posted: %ld, %c%c",
		rdtp->last_accelerate & 0xffff, jiffies & 0xffff,
		ulong2long(nlpd),
		rdtp->all_lazy ? 'L' : '.',
		rdtp->tick_nohz_enabled_snap ? '.' : 'D');
1718 1719 1720 1721 1722 1723
}

#else /* #ifdef CONFIG_RCU_FAST_NO_HZ */

static void print_cpu_stall_fast_no_hz(char *cp, int cpu)
{
1724
	*cp = '\0';
1725 1726 1727 1728 1729 1730 1731
}

#endif /* #else #ifdef CONFIG_RCU_FAST_NO_HZ */

/* Initiate the stall-info list. */
static void print_cpu_stall_info_begin(void)
{
1732
	pr_cont("\n");
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
}

/*
 * Print out diagnostic information for the specified stalled CPU.
 *
 * If the specified CPU is aware of the current RCU grace period
 * (flavor specified by rsp), then print the number of scheduling
 * clock interrupts the CPU has taken during the time that it has
 * been aware.  Otherwise, print the number of RCU grace periods
 * that this CPU is ignorant of, for example, "1" if the CPU was
 * aware of the previous grace period.
 *
 * Also print out idle and (if CONFIG_RCU_FAST_NO_HZ) idle-entry info.
 */
static void print_cpu_stall_info(struct rcu_state *rsp, int cpu)
{
	char fast_no_hz[72];
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
	struct rcu_dynticks *rdtp = rdp->dynticks;
	char *ticks_title;
	unsigned long ticks_value;

	if (rsp->gpnum == rdp->gpnum) {
		ticks_title = "ticks this GP";
		ticks_value = rdp->ticks_this_gp;
	} else {
		ticks_title = "GPs behind";
		ticks_value = rsp->gpnum - rdp->gpnum;
	}
	print_cpu_stall_fast_no_hz(fast_no_hz, cpu);
1763
	pr_err("\t%d: (%lu %s) idle=%03x/%llx/%d softirq=%u/%u fqs=%ld %s\n",
1764 1765 1766
	       cpu, ticks_value, ticks_title,
	       atomic_read(&rdtp->dynticks) & 0xfff,
	       rdtp->dynticks_nesting, rdtp->dynticks_nmi_nesting,
1767
	       rdp->softirq_snap, kstat_softirqs_cpu(RCU_SOFTIRQ, cpu),
1768
	       READ_ONCE(rsp->n_force_qs) - rsp->n_force_qs_gpstart,
1769 1770 1771 1772 1773 1774
	       fast_no_hz);
}

/* Terminate the stall-info list. */
static void print_cpu_stall_info_end(void)
{
1775
	pr_err("\t");
1776 1777 1778 1779 1780 1781
}

/* Zero ->ticks_this_gp for all flavors of RCU. */
static void zero_cpu_stall_ticks(struct rcu_data *rdp)
{
	rdp->ticks_this_gp = 0;
1782
	rdp->softirq_snap = kstat_softirqs_cpu(RCU_SOFTIRQ, smp_processor_id());
1783 1784 1785 1786 1787
}

/* Increment ->ticks_this_gp for all flavors of RCU. */
static void increment_cpu_stall_ticks(void)
{
1788 1789 1790
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
1791
		raw_cpu_inc(rsp->rda->ticks_this_gp);
1792 1793 1794 1795 1796 1797
}

#else /* #ifdef CONFIG_RCU_CPU_STALL_INFO */

static void print_cpu_stall_info_begin(void)
{
1798
	pr_cont(" {");
1799 1800 1801 1802
}

static void print_cpu_stall_info(struct rcu_state *rsp, int cpu)
{
1803
	pr_cont(" %d", cpu);
1804 1805 1806 1807
}

static void print_cpu_stall_info_end(void)
{
1808
	pr_cont("} ");
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
}

static void zero_cpu_stall_ticks(struct rcu_data *rdp)
{
}

static void increment_cpu_stall_ticks(void)
{
}

#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_INFO */
P
Paul E. McKenney 已提交
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853

#ifdef CONFIG_RCU_NOCB_CPU

/*
 * Offload callback processing from the boot-time-specified set of CPUs
 * specified by rcu_nocb_mask.  For each CPU in the set, there is a
 * kthread created that pulls the callbacks from the corresponding CPU,
 * waits for a grace period to elapse, and invokes the callbacks.
 * The no-CBs CPUs do a wake_up() on their kthread when they insert
 * a callback into any empty list, unless the rcu_nocb_poll boot parameter
 * has been specified, in which case each kthread actively polls its
 * CPU.  (Which isn't so great for energy efficiency, but which does
 * reduce RCU's overhead on that CPU.)
 *
 * This is intended to be used in conjunction with Frederic Weisbecker's
 * adaptive-idle work, which would seriously reduce OS jitter on CPUs
 * running CPU-bound user-mode computations.
 *
 * Offloading of callback processing could also in theory be used as
 * an energy-efficiency measure because CPUs with no RCU callbacks
 * queued are more aggressive about entering dyntick-idle mode.
 */


/* Parse the boot-time rcu_nocb_mask CPU list from the kernel parameters. */
static int __init rcu_nocb_setup(char *str)
{
	alloc_bootmem_cpumask_var(&rcu_nocb_mask);
	have_rcu_nocb_mask = true;
	cpulist_parse(str, rcu_nocb_mask);
	return 1;
}
__setup("rcu_nocbs=", rcu_nocb_setup);

1854 1855 1856 1857 1858 1859 1860
static int __init parse_rcu_nocb_poll(char *arg)
{
	rcu_nocb_poll = 1;
	return 0;
}
early_param("rcu_nocb_poll", parse_rcu_nocb_poll);

1861
/*
1862 1863
 * Wake up any no-CBs CPUs' kthreads that were waiting on the just-ended
 * grace period.
1864
 */
1865
static void rcu_nocb_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
1866
{
1867
	wake_up_all(&rnp->nocb_gp_wq[rnp->completed & 0x1]);
1868 1869 1870
}

/*
1871
 * Set the root rcu_node structure's ->need_future_gp field
1872 1873 1874 1875 1876
 * based on the sum of those of all rcu_node structures.  This does
 * double-count the root rcu_node structure's requests, but this
 * is necessary to handle the possibility of a rcu_nocb_kthread()
 * having awakened during the time that the rcu_node structures
 * were being updated for the end of the previous grace period.
1877
 */
1878 1879
static void rcu_nocb_gp_set(struct rcu_node *rnp, int nrq)
{
1880
	rnp->need_future_gp[(rnp->completed + 1) & 0x1] += nrq;
1881 1882 1883
}

static void rcu_init_one_nocb(struct rcu_node *rnp)
1884
{
1885 1886
	init_waitqueue_head(&rnp->nocb_gp_wq[0]);
	init_waitqueue_head(&rnp->nocb_gp_wq[1]);
1887 1888
}

1889
#ifndef CONFIG_RCU_NOCB_CPU_ALL
L
Liu Ping Fan 已提交
1890
/* Is the specified CPU a no-CBs CPU? */
1891
bool rcu_is_nocb_cpu(int cpu)
P
Paul E. McKenney 已提交
1892 1893 1894 1895 1896
{
	if (have_rcu_nocb_mask)
		return cpumask_test_cpu(cpu, rcu_nocb_mask);
	return false;
}
1897
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_ALL */
P
Paul E. McKenney 已提交
1898

1899 1900 1901 1902 1903 1904 1905
/*
 * Kick the leader kthread for this NOCB group.
 */
static void wake_nocb_leader(struct rcu_data *rdp, bool force)
{
	struct rcu_data *rdp_leader = rdp->nocb_leader;

1906
	if (!READ_ONCE(rdp_leader->nocb_kthread))
1907
		return;
1908
	if (READ_ONCE(rdp_leader->nocb_leader_sleep) || force) {
1909
		/* Prior smp_mb__after_atomic() orders against prior enqueue. */
1910
		WRITE_ONCE(rdp_leader->nocb_leader_sleep, false);
1911 1912 1913 1914
		wake_up(&rdp_leader->nocb_wq);
	}
}

1915 1916 1917 1918 1919 1920 1921
/*
 * Does the specified CPU need an RCU callback for the specified flavor
 * of rcu_barrier()?
 */
static bool rcu_nocb_cpu_needs_barrier(struct rcu_state *rsp, int cpu)
{
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
1922 1923
	unsigned long ret;
#ifdef CONFIG_PROVE_RCU
1924
	struct rcu_head *rhp;
1925
#endif /* #ifdef CONFIG_PROVE_RCU */
1926

1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939
	/*
	 * Check count of all no-CBs callbacks awaiting invocation.
	 * There needs to be a barrier before this function is called,
	 * but associated with a prior determination that no more
	 * callbacks would be posted.  In the worst case, the first
	 * barrier in _rcu_barrier() suffices (but the caller cannot
	 * necessarily rely on this, not a substitute for the caller
	 * getting the concurrency design right!).  There must also be
	 * a barrier between the following load an posting of a callback
	 * (if a callback is in fact needed).  This is associated with an
	 * atomic_inc() in the caller.
	 */
	ret = atomic_long_read(&rdp->nocb_q_count);
1940

1941
#ifdef CONFIG_PROVE_RCU
1942
	rhp = READ_ONCE(rdp->nocb_head);
1943
	if (!rhp)
1944
		rhp = READ_ONCE(rdp->nocb_gp_head);
1945
	if (!rhp)
1946
		rhp = READ_ONCE(rdp->nocb_follower_head);
1947 1948

	/* Having no rcuo kthread but CBs after scheduler starts is bad! */
1949
	if (!READ_ONCE(rdp->nocb_kthread) && rhp &&
1950
	    rcu_scheduler_fully_active) {
1951 1952 1953 1954 1955
		/* RCU callback enqueued before CPU first came online??? */
		pr_err("RCU: Never-onlined no-CBs CPU %d has CB %p\n",
		       cpu, rhp->func);
		WARN_ON_ONCE(1);
	}
1956
#endif /* #ifdef CONFIG_PROVE_RCU */
1957

1958
	return !!ret;
1959 1960
}

P
Paul E. McKenney 已提交
1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971
/*
 * Enqueue the specified string of rcu_head structures onto the specified
 * CPU's no-CBs lists.  The CPU is specified by rdp, the head of the
 * string by rhp, and the tail of the string by rhtp.  The non-lazy/lazy
 * counts are supplied by rhcount and rhcount_lazy.
 *
 * If warranted, also wake up the kthread servicing this CPUs queues.
 */
static void __call_rcu_nocb_enqueue(struct rcu_data *rdp,
				    struct rcu_head *rhp,
				    struct rcu_head **rhtp,
1972 1973
				    int rhcount, int rhcount_lazy,
				    unsigned long flags)
P
Paul E. McKenney 已提交
1974 1975 1976 1977 1978 1979
{
	int len;
	struct rcu_head **old_rhpp;
	struct task_struct *t;

	/* Enqueue the callback on the nocb list and update counts. */
1980 1981
	atomic_long_add(rhcount, &rdp->nocb_q_count);
	/* rcu_barrier() relies on ->nocb_q_count add before xchg. */
P
Paul E. McKenney 已提交
1982
	old_rhpp = xchg(&rdp->nocb_tail, rhtp);
1983
	WRITE_ONCE(*old_rhpp, rhp);
P
Paul E. McKenney 已提交
1984
	atomic_long_add(rhcount_lazy, &rdp->nocb_q_count_lazy);
1985
	smp_mb__after_atomic(); /* Store *old_rhpp before _wake test. */
P
Paul E. McKenney 已提交
1986 1987

	/* If we are not being polled and there is a kthread, awaken it ... */
1988
	t = READ_ONCE(rdp->nocb_kthread);
1989
	if (rcu_nocb_poll || !t) {
1990 1991
		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
				    TPS("WakeNotPoll"));
P
Paul E. McKenney 已提交
1992
		return;
1993
	}
P
Paul E. McKenney 已提交
1994 1995
	len = atomic_long_read(&rdp->nocb_q_count);
	if (old_rhpp == &rdp->nocb_head) {
1996
		if (!irqs_disabled_flags(flags)) {
1997 1998
			/* ... if queue was empty ... */
			wake_nocb_leader(rdp, false);
1999 2000 2001
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    TPS("WakeEmpty"));
		} else {
2002
			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE;
2003 2004 2005
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    TPS("WakeEmptyIsDeferred"));
		}
P
Paul E. McKenney 已提交
2006 2007
		rdp->qlen_last_fqs_check = 0;
	} else if (len > rdp->qlen_last_fqs_check + qhimark) {
2008
		/* ... or if many callbacks queued. */
2009 2010 2011 2012 2013 2014 2015 2016 2017
		if (!irqs_disabled_flags(flags)) {
			wake_nocb_leader(rdp, true);
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    TPS("WakeOvf"));
		} else {
			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE_FORCE;
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    TPS("WakeOvfIsDeferred"));
		}
P
Paul E. McKenney 已提交
2018
		rdp->qlen_last_fqs_check = LONG_MAX / 2;
2019 2020
	} else {
		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeNot"));
P
Paul E. McKenney 已提交
2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034
	}
	return;
}

/*
 * This is a helper for __call_rcu(), which invokes this when the normal
 * callback queue is inoperable.  If this is not a no-CBs CPU, this
 * function returns failure back to __call_rcu(), which can complain
 * appropriately.
 *
 * Otherwise, this function queues the callback where the corresponding
 * "rcuo" kthread can find it.
 */
static bool __call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *rhp,
2035
			    bool lazy, unsigned long flags)
P
Paul E. McKenney 已提交
2036 2037
{

2038
	if (!rcu_is_nocb_cpu(rdp->cpu))
2039
		return false;
2040
	__call_rcu_nocb_enqueue(rdp, rhp, &rhp->next, 1, lazy, flags);
2041 2042 2043
	if (__is_kfree_rcu_offset((unsigned long)rhp->func))
		trace_rcu_kfree_callback(rdp->rsp->name, rhp,
					 (unsigned long)rhp->func,
2044 2045
					 -atomic_long_read(&rdp->nocb_q_count_lazy),
					 -atomic_long_read(&rdp->nocb_q_count));
2046 2047
	else
		trace_rcu_callback(rdp->rsp->name, rhp,
2048 2049
				   -atomic_long_read(&rdp->nocb_q_count_lazy),
				   -atomic_long_read(&rdp->nocb_q_count));
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060

	/*
	 * If called from an extended quiescent state with interrupts
	 * disabled, invoke the RCU core in order to allow the idle-entry
	 * deferred-wakeup check to function.
	 */
	if (irqs_disabled_flags(flags) &&
	    !rcu_is_watching() &&
	    cpu_online(smp_processor_id()))
		invoke_rcu_core();

2061
	return true;
P
Paul E. McKenney 已提交
2062 2063 2064 2065 2066 2067 2068
}

/*
 * Adopt orphaned callbacks on a no-CBs CPU, or return 0 if this is
 * not a no-CBs CPU.
 */
static bool __maybe_unused rcu_nocb_adopt_orphan_cbs(struct rcu_state *rsp,
2069 2070
						     struct rcu_data *rdp,
						     unsigned long flags)
P
Paul E. McKenney 已提交
2071 2072 2073 2074 2075
{
	long ql = rsp->qlen;
	long qll = rsp->qlen_lazy;

	/* If this is not a no-CBs CPU, tell the caller to do it the old way. */
2076
	if (!rcu_is_nocb_cpu(smp_processor_id()))
2077
		return false;
P
Paul E. McKenney 已提交
2078 2079 2080 2081 2082 2083
	rsp->qlen = 0;
	rsp->qlen_lazy = 0;

	/* First, enqueue the donelist, if any.  This preserves CB ordering. */
	if (rsp->orphan_donelist != NULL) {
		__call_rcu_nocb_enqueue(rdp, rsp->orphan_donelist,
2084
					rsp->orphan_donetail, ql, qll, flags);
P
Paul E. McKenney 已提交
2085 2086 2087 2088 2089 2090
		ql = qll = 0;
		rsp->orphan_donelist = NULL;
		rsp->orphan_donetail = &rsp->orphan_donelist;
	}
	if (rsp->orphan_nxtlist != NULL) {
		__call_rcu_nocb_enqueue(rdp, rsp->orphan_nxtlist,
2091
					rsp->orphan_nxttail, ql, qll, flags);
P
Paul E. McKenney 已提交
2092 2093 2094 2095
		ql = qll = 0;
		rsp->orphan_nxtlist = NULL;
		rsp->orphan_nxttail = &rsp->orphan_nxtlist;
	}
2096
	return true;
P
Paul E. McKenney 已提交
2097 2098 2099
}

/*
2100 2101
 * If necessary, kick off a new grace period, and either way wait
 * for a subsequent grace period to complete.
P
Paul E. McKenney 已提交
2102
 */
2103
static void rcu_nocb_wait_gp(struct rcu_data *rdp)
P
Paul E. McKenney 已提交
2104
{
2105
	unsigned long c;
2106
	bool d;
2107
	unsigned long flags;
2108
	bool needwake;
2109 2110 2111
	struct rcu_node *rnp = rdp->mynode;

	raw_spin_lock_irqsave(&rnp->lock, flags);
2112
	smp_mb__after_unlock_lock();
2113
	needwake = rcu_start_future_gp(rnp, rdp, &c);
2114
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
2115 2116
	if (needwake)
		rcu_gp_kthread_wake(rdp->rsp);
P
Paul E. McKenney 已提交
2117 2118

	/*
2119 2120
	 * Wait for the grace period.  Do so interruptibly to avoid messing
	 * up the load average.
P
Paul E. McKenney 已提交
2121
	 */
2122
	trace_rcu_future_gp(rnp, rdp, c, TPS("StartWait"));
2123
	for (;;) {
2124 2125
		wait_event_interruptible(
			rnp->nocb_gp_wq[c & 0x1],
2126
			(d = ULONG_CMP_GE(READ_ONCE(rnp->completed), c)));
2127
		if (likely(d))
2128
			break;
2129
		WARN_ON(signal_pending(current));
2130
		trace_rcu_future_gp(rnp, rdp, c, TPS("ResumeWait"));
2131
	}
2132
	trace_rcu_future_gp(rnp, rdp, c, TPS("EndWait"));
2133
	smp_mb(); /* Ensure that CB invocation happens after GP end. */
P
Paul E. McKenney 已提交
2134 2135
}

2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
/*
 * Leaders come here to wait for additional callbacks to show up.
 * This function does not return until callbacks appear.
 */
static void nocb_leader_wait(struct rcu_data *my_rdp)
{
	bool firsttime = true;
	bool gotcbs;
	struct rcu_data *rdp;
	struct rcu_head **tail;

wait_again:

	/* Wait for callbacks to appear. */
	if (!rcu_nocb_poll) {
		trace_rcu_nocb_wake(my_rdp->rsp->name, my_rdp->cpu, "Sleep");
		wait_event_interruptible(my_rdp->nocb_wq,
2153
				!READ_ONCE(my_rdp->nocb_leader_sleep));
2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166
		/* Memory barrier handled by smp_mb() calls below and repoll. */
	} else if (firsttime) {
		firsttime = false; /* Don't drown trace log with "Poll"! */
		trace_rcu_nocb_wake(my_rdp->rsp->name, my_rdp->cpu, "Poll");
	}

	/*
	 * Each pass through the following loop checks a follower for CBs.
	 * We are our own first follower.  Any CBs found are moved to
	 * nocb_gp_head, where they await a grace period.
	 */
	gotcbs = false;
	for (rdp = my_rdp; rdp; rdp = rdp->nocb_next_follower) {
2167
		rdp->nocb_gp_head = READ_ONCE(rdp->nocb_head);
2168 2169 2170 2171
		if (!rdp->nocb_gp_head)
			continue;  /* No CBs here, try next follower. */

		/* Move callbacks to wait-for-GP list, which is empty. */
2172
		WRITE_ONCE(rdp->nocb_head, NULL);
2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184
		rdp->nocb_gp_tail = xchg(&rdp->nocb_tail, &rdp->nocb_head);
		gotcbs = true;
	}

	/*
	 * If there were no callbacks, sleep a bit, rescan after a
	 * memory barrier, and go retry.
	 */
	if (unlikely(!gotcbs)) {
		if (!rcu_nocb_poll)
			trace_rcu_nocb_wake(my_rdp->rsp->name, my_rdp->cpu,
					    "WokeEmpty");
2185
		WARN_ON(signal_pending(current));
2186 2187 2188
		schedule_timeout_interruptible(1);

		/* Rescan in case we were a victim of memory ordering. */
2189 2190
		my_rdp->nocb_leader_sleep = true;
		smp_mb();  /* Ensure _sleep true before scan. */
2191
		for (rdp = my_rdp; rdp; rdp = rdp->nocb_next_follower)
2192
			if (READ_ONCE(rdp->nocb_head)) {
2193
				/* Found CB, so short-circuit next wait. */
2194
				my_rdp->nocb_leader_sleep = false;
2195 2196 2197 2198 2199 2200 2201 2202 2203
				break;
			}
		goto wait_again;
	}

	/* Wait for one grace period. */
	rcu_nocb_wait_gp(my_rdp);

	/*
2204 2205
	 * We left ->nocb_leader_sleep unset to reduce cache thrashing.
	 * We set it now, but recheck for new callbacks while
2206 2207
	 * traversing our follower list.
	 */
2208 2209
	my_rdp->nocb_leader_sleep = true;
	smp_mb(); /* Ensure _sleep true before scan of ->nocb_head. */
2210 2211 2212

	/* Each pass through the following loop wakes a follower, if needed. */
	for (rdp = my_rdp; rdp; rdp = rdp->nocb_next_follower) {
2213
		if (READ_ONCE(rdp->nocb_head))
2214
			my_rdp->nocb_leader_sleep = false;/* No need to sleep.*/
2215 2216 2217 2218 2219 2220
		if (!rdp->nocb_gp_head)
			continue; /* No CBs, so no need to wake follower. */

		/* Append callbacks to follower's "done" list. */
		tail = xchg(&rdp->nocb_follower_tail, rdp->nocb_gp_tail);
		*tail = rdp->nocb_gp_head;
2221
		smp_mb__after_atomic(); /* Store *tail before wakeup. */
2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
		if (rdp != my_rdp && tail == &rdp->nocb_follower_head) {
			/*
			 * List was empty, wake up the follower.
			 * Memory barriers supplied by atomic_long_add().
			 */
			wake_up(&rdp->nocb_wq);
		}
	}

	/* If we (the leader) don't have CBs, go wait some more. */
	if (!my_rdp->nocb_follower_head)
		goto wait_again;
}

/*
 * Followers come here to wait for additional callbacks to show up.
 * This function does not return until callbacks appear.
 */
static void nocb_follower_wait(struct rcu_data *rdp)
{
	bool firsttime = true;

	for (;;) {
		if (!rcu_nocb_poll) {
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    "FollowerSleep");
			wait_event_interruptible(rdp->nocb_wq,
2249
						 READ_ONCE(rdp->nocb_follower_head));
2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261
		} else if (firsttime) {
			/* Don't drown trace log with "Poll"! */
			firsttime = false;
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, "Poll");
		}
		if (smp_load_acquire(&rdp->nocb_follower_head)) {
			/* ^^^ Ensure CB invocation follows _head test. */
			return;
		}
		if (!rcu_nocb_poll)
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    "WokeEmpty");
2262
		WARN_ON(signal_pending(current));
2263 2264 2265 2266
		schedule_timeout_interruptible(1);
	}
}

P
Paul E. McKenney 已提交
2267 2268
/*
 * Per-rcu_data kthread, but only for no-CBs CPUs.  Each kthread invokes
2269 2270 2271
 * callbacks queued by the corresponding no-CBs CPU, however, there is
 * an optional leader-follower relationship so that the grace-period
 * kthreads don't have to do quite so many wakeups.
P
Paul E. McKenney 已提交
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282
 */
static int rcu_nocb_kthread(void *arg)
{
	int c, cl;
	struct rcu_head *list;
	struct rcu_head *next;
	struct rcu_head **tail;
	struct rcu_data *rdp = arg;

	/* Each pass through this loop invokes one batch of callbacks */
	for (;;) {
2283 2284 2285 2286 2287 2288 2289
		/* Wait for callbacks. */
		if (rdp->nocb_leader == rdp)
			nocb_leader_wait(rdp);
		else
			nocb_follower_wait(rdp);

		/* Pull the ready-to-invoke callbacks onto local list. */
2290
		list = READ_ONCE(rdp->nocb_follower_head);
2291 2292
		BUG_ON(!list);
		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, "WokeNonEmpty");
2293
		WRITE_ONCE(rdp->nocb_follower_head, NULL);
2294
		tail = xchg(&rdp->nocb_follower_tail, &rdp->nocb_follower_head);
P
Paul E. McKenney 已提交
2295 2296

		/* Each pass through the following loop invokes a callback. */
2297 2298 2299
		trace_rcu_batch_start(rdp->rsp->name,
				      atomic_long_read(&rdp->nocb_q_count_lazy),
				      atomic_long_read(&rdp->nocb_q_count), -1);
P
Paul E. McKenney 已提交
2300 2301 2302 2303 2304
		c = cl = 0;
		while (list) {
			next = list->next;
			/* Wait for enqueuing to complete, if needed. */
			while (next == NULL && &list->next != tail) {
2305 2306
				trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
						    TPS("WaitQueue"));
P
Paul E. McKenney 已提交
2307
				schedule_timeout_interruptible(1);
2308 2309
				trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
						    TPS("WokeQueue"));
P
Paul E. McKenney 已提交
2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
				next = list->next;
			}
			debug_rcu_head_unqueue(list);
			local_bh_disable();
			if (__rcu_reclaim(rdp->rsp->name, list))
				cl++;
			c++;
			local_bh_enable();
			list = next;
		}
		trace_rcu_batch_end(rdp->rsp->name, c, !!list, 0, 0, 1);
2321 2322 2323
		smp_mb__before_atomic();  /* _add after CB invocation. */
		atomic_long_add(-c, &rdp->nocb_q_count);
		atomic_long_add(-cl, &rdp->nocb_q_count_lazy);
2324
		rdp->n_nocbs_invoked += c;
P
Paul E. McKenney 已提交
2325 2326 2327 2328
	}
	return 0;
}

2329
/* Is a deferred wakeup of rcu_nocb_kthread() required? */
2330
static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
2331
{
2332
	return READ_ONCE(rdp->nocb_defer_wakeup);
2333 2334 2335 2336 2337
}

/* Do a deferred wakeup of rcu_nocb_kthread(). */
static void do_nocb_deferred_wakeup(struct rcu_data *rdp)
{
2338 2339
	int ndw;

2340 2341
	if (!rcu_nocb_need_deferred_wakeup(rdp))
		return;
2342 2343
	ndw = READ_ONCE(rdp->nocb_defer_wakeup);
	WRITE_ONCE(rdp->nocb_defer_wakeup, RCU_NOGP_WAKE_NOT);
2344 2345
	wake_nocb_leader(rdp, ndw == RCU_NOGP_WAKE_FORCE);
	trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("DeferredWake"));
2346 2347
}

2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363
void __init rcu_init_nohz(void)
{
	int cpu;
	bool need_rcu_nocb_mask = true;
	struct rcu_state *rsp;

#ifdef CONFIG_RCU_NOCB_CPU_NONE
	need_rcu_nocb_mask = false;
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_NONE */

#if defined(CONFIG_NO_HZ_FULL)
	if (tick_nohz_full_running && cpumask_weight(tick_nohz_full_mask))
		need_rcu_nocb_mask = true;
#endif /* #if defined(CONFIG_NO_HZ_FULL) */

	if (!have_rcu_nocb_mask && need_rcu_nocb_mask) {
2364 2365 2366 2367
		if (!zalloc_cpumask_var(&rcu_nocb_mask, GFP_KERNEL)) {
			pr_info("rcu_nocb_mask allocation failed, callback offloading disabled.\n");
			return;
		}
2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
		have_rcu_nocb_mask = true;
	}
	if (!have_rcu_nocb_mask)
		return;

#ifdef CONFIG_RCU_NOCB_CPU_ZERO
	pr_info("\tOffload RCU callbacks from CPU 0\n");
	cpumask_set_cpu(0, rcu_nocb_mask);
#endif /* #ifdef CONFIG_RCU_NOCB_CPU_ZERO */
#ifdef CONFIG_RCU_NOCB_CPU_ALL
	pr_info("\tOffload RCU callbacks from all CPUs\n");
	cpumask_copy(rcu_nocb_mask, cpu_possible_mask);
#endif /* #ifdef CONFIG_RCU_NOCB_CPU_ALL */
#if defined(CONFIG_NO_HZ_FULL)
	if (tick_nohz_full_running)
		cpumask_or(rcu_nocb_mask, rcu_nocb_mask, tick_nohz_full_mask);
#endif /* #if defined(CONFIG_NO_HZ_FULL) */

	if (!cpumask_subset(rcu_nocb_mask, cpu_possible_mask)) {
		pr_info("\tNote: kernel parameter 'rcu_nocbs=' contains nonexistent CPUs.\n");
		cpumask_and(rcu_nocb_mask, cpu_possible_mask,
			    rcu_nocb_mask);
	}
2391 2392
	pr_info("\tOffload RCU callbacks from CPUs: %*pbl.\n",
		cpumask_pr_args(rcu_nocb_mask));
2393 2394 2395 2396
	if (rcu_nocb_poll)
		pr_info("\tPoll for callbacks from no-CBs CPUs.\n");

	for_each_rcu_flavor(rsp) {
2397 2398
		for_each_cpu(cpu, rcu_nocb_mask)
			init_nocb_callback_list(per_cpu_ptr(rsp->rda, cpu));
2399
		rcu_organize_nocb_kthreads(rsp);
2400
	}
2401 2402
}

P
Paul E. McKenney 已提交
2403 2404 2405 2406 2407
/* Initialize per-rcu_data variables for no-CBs CPUs. */
static void __init rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp)
{
	rdp->nocb_tail = &rdp->nocb_head;
	init_waitqueue_head(&rdp->nocb_wq);
2408
	rdp->nocb_follower_tail = &rdp->nocb_follower_head;
P
Paul E. McKenney 已提交
2409 2410
}

2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440
/*
 * If the specified CPU is a no-CBs CPU that does not already have its
 * rcuo kthread for the specified RCU flavor, spawn it.  If the CPUs are
 * brought online out of order, this can require re-organizing the
 * leader-follower relationships.
 */
static void rcu_spawn_one_nocb_kthread(struct rcu_state *rsp, int cpu)
{
	struct rcu_data *rdp;
	struct rcu_data *rdp_last;
	struct rcu_data *rdp_old_leader;
	struct rcu_data *rdp_spawn = per_cpu_ptr(rsp->rda, cpu);
	struct task_struct *t;

	/*
	 * If this isn't a no-CBs CPU or if it already has an rcuo kthread,
	 * then nothing to do.
	 */
	if (!rcu_is_nocb_cpu(cpu) || rdp_spawn->nocb_kthread)
		return;

	/* If we didn't spawn the leader first, reorganize! */
	rdp_old_leader = rdp_spawn->nocb_leader;
	if (rdp_old_leader != rdp_spawn && !rdp_old_leader->nocb_kthread) {
		rdp_last = NULL;
		rdp = rdp_old_leader;
		do {
			rdp->nocb_leader = rdp_spawn;
			if (rdp_last && rdp != rdp_spawn)
				rdp_last->nocb_next_follower = rdp;
2441 2442 2443 2444 2445 2446 2447
			if (rdp == rdp_spawn) {
				rdp = rdp->nocb_next_follower;
			} else {
				rdp_last = rdp;
				rdp = rdp->nocb_next_follower;
				rdp_last->nocb_next_follower = NULL;
			}
2448 2449 2450 2451 2452 2453 2454 2455
		} while (rdp);
		rdp_spawn->nocb_next_follower = rdp_old_leader;
	}

	/* Spawn the kthread for this CPU and RCU flavor. */
	t = kthread_run(rcu_nocb_kthread, rdp_spawn,
			"rcuo%c/%d", rsp->abbr, cpu);
	BUG_ON(IS_ERR(t));
2456
	WRITE_ONCE(rdp_spawn->nocb_kthread, t);
2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
}

/*
 * If the specified CPU is a no-CBs CPU that does not already have its
 * rcuo kthreads, spawn them.
 */
static void rcu_spawn_all_nocb_kthreads(int cpu)
{
	struct rcu_state *rsp;

	if (rcu_scheduler_fully_active)
		for_each_rcu_flavor(rsp)
			rcu_spawn_one_nocb_kthread(rsp, cpu);
}

/*
 * Once the scheduler is running, spawn rcuo kthreads for all online
 * no-CBs CPUs.  This assumes that the early_initcall()s happen before
 * non-boot CPUs come online -- if this changes, we will need to add
 * some mutual exclusion.
 */
static void __init rcu_spawn_nocb_kthreads(void)
{
	int cpu;

	for_each_online_cpu(cpu)
		rcu_spawn_all_nocb_kthreads(cpu);
}

2486 2487 2488 2489 2490
/* How many follower CPU IDs per leader?  Default of -1 for sqrt(nr_cpu_ids). */
static int rcu_nocb_leader_stride = -1;
module_param(rcu_nocb_leader_stride, int, 0444);

/*
2491
 * Initialize leader-follower relationships for all no-CBs CPU.
2492
 */
2493
static void __init rcu_organize_nocb_kthreads(struct rcu_state *rsp)
P
Paul E. McKenney 已提交
2494 2495
{
	int cpu;
2496 2497
	int ls = rcu_nocb_leader_stride;
	int nl = 0;  /* Next leader. */
P
Paul E. McKenney 已提交
2498
	struct rcu_data *rdp;
2499 2500
	struct rcu_data *rdp_leader = NULL;  /* Suppress misguided gcc warn. */
	struct rcu_data *rdp_prev = NULL;
P
Paul E. McKenney 已提交
2501

2502
	if (!have_rcu_nocb_mask)
P
Paul E. McKenney 已提交
2503
		return;
2504 2505 2506 2507 2508 2509 2510 2511 2512
	if (ls == -1) {
		ls = int_sqrt(nr_cpu_ids);
		rcu_nocb_leader_stride = ls;
	}

	/*
	 * Each pass through this loop sets up one rcu_data structure and
	 * spawns one rcu_nocb_kthread().
	 */
P
Paul E. McKenney 已提交
2513 2514
	for_each_cpu(cpu, rcu_nocb_mask) {
		rdp = per_cpu_ptr(rsp->rda, cpu);
2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525
		if (rdp->cpu >= nl) {
			/* New leader, set up for followers & next leader. */
			nl = DIV_ROUND_UP(rdp->cpu + 1, ls) * ls;
			rdp->nocb_leader = rdp;
			rdp_leader = rdp;
		} else {
			/* Another follower, link to previous leader. */
			rdp->nocb_leader = rdp_leader;
			rdp_prev->nocb_next_follower = rdp;
		}
		rdp_prev = rdp;
P
Paul E. McKenney 已提交
2526 2527 2528 2529
	}
}

/* Prevent __call_rcu() from enqueuing callbacks on no-CBs CPUs */
2530
static bool init_nocb_callback_list(struct rcu_data *rdp)
P
Paul E. McKenney 已提交
2531
{
2532
	if (!rcu_is_nocb_cpu(rdp->cpu))
2533
		return false;
2534

2535 2536 2537 2538 2539 2540 2541 2542 2543 2544
	/* If there are early-boot callbacks, move them to nocb lists. */
	if (rdp->nxtlist) {
		rdp->nocb_head = rdp->nxtlist;
		rdp->nocb_tail = rdp->nxttail[RCU_NEXT_TAIL];
		atomic_long_set(&rdp->nocb_q_count, rdp->qlen);
		atomic_long_set(&rdp->nocb_q_count_lazy, rdp->qlen_lazy);
		rdp->nxtlist = NULL;
		rdp->qlen = 0;
		rdp->qlen_lazy = 0;
	}
P
Paul E. McKenney 已提交
2545
	rdp->nxttail[RCU_NEXT_TAIL] = NULL;
2546
	return true;
P
Paul E. McKenney 已提交
2547 2548
}

2549 2550
#else /* #ifdef CONFIG_RCU_NOCB_CPU */

2551 2552 2553 2554 2555 2556
static bool rcu_nocb_cpu_needs_barrier(struct rcu_state *rsp, int cpu)
{
	WARN_ON_ONCE(1); /* Should be dead code. */
	return false;
}

2557
static void rcu_nocb_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
P
Paul E. McKenney 已提交
2558 2559 2560
{
}

2561 2562 2563 2564 2565 2566 2567
static void rcu_nocb_gp_set(struct rcu_node *rnp, int nrq)
{
}

static void rcu_init_one_nocb(struct rcu_node *rnp)
{
}
P
Paul E. McKenney 已提交
2568 2569

static bool __call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *rhp,
2570
			    bool lazy, unsigned long flags)
P
Paul E. McKenney 已提交
2571
{
2572
	return false;
P
Paul E. McKenney 已提交
2573 2574 2575
}

static bool __maybe_unused rcu_nocb_adopt_orphan_cbs(struct rcu_state *rsp,
2576 2577
						     struct rcu_data *rdp,
						     unsigned long flags)
P
Paul E. McKenney 已提交
2578
{
2579
	return false;
P
Paul E. McKenney 已提交
2580 2581 2582 2583 2584 2585
}

static void __init rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp)
{
}

2586
static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
2587 2588 2589 2590 2591 2592 2593 2594
{
	return false;
}

static void do_nocb_deferred_wakeup(struct rcu_data *rdp)
{
}

2595 2596 2597 2598 2599
static void rcu_spawn_all_nocb_kthreads(int cpu)
{
}

static void __init rcu_spawn_nocb_kthreads(void)
P
Paul E. McKenney 已提交
2600 2601 2602
{
}

2603
static bool init_nocb_callback_list(struct rcu_data *rdp)
P
Paul E. McKenney 已提交
2604
{
2605
	return false;
P
Paul E. McKenney 已提交
2606 2607 2608
}

#endif /* #else #ifdef CONFIG_RCU_NOCB_CPU */
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618

/*
 * An adaptive-ticks CPU can potentially execute in kernel mode for an
 * arbitrarily long period of time with the scheduling-clock tick turned
 * off.  RCU will be paying attention to this CPU because it is in the
 * kernel, but the CPU cannot be guaranteed to be executing the RCU state
 * machine because the scheduling-clock tick has been disabled.  Therefore,
 * if an adaptive-ticks CPU is failing to respond to the current grace
 * period and has not be idle from an RCU perspective, kick it.
 */
2619
static void __maybe_unused rcu_kick_nohz_cpu(int cpu)
2620 2621 2622 2623 2624 2625
{
#ifdef CONFIG_NO_HZ_FULL
	if (tick_nohz_full_cpu(cpu))
		smp_send_reschedule(cpu);
#endif /* #ifdef CONFIG_NO_HZ_FULL */
}
2626 2627 2628 2629


#ifdef CONFIG_NO_HZ_FULL_SYSIDLE

2630
static int full_sysidle_state;		/* Current system-idle state. */
2631 2632 2633 2634 2635 2636
#define RCU_SYSIDLE_NOT		0	/* Some CPU is not idle. */
#define RCU_SYSIDLE_SHORT	1	/* All CPUs idle for brief period. */
#define RCU_SYSIDLE_LONG	2	/* All CPUs idle for long enough. */
#define RCU_SYSIDLE_FULL	3	/* All CPUs idle, ready for sysidle. */
#define RCU_SYSIDLE_FULL_NOTED	4	/* Actually entered sysidle state. */

2637 2638 2639 2640 2641 2642
/*
 * Invoked to note exit from irq or task transition to idle.  Note that
 * usermode execution does -not- count as idle here!  After all, we want
 * to detect full-system idle states, not RCU quiescent states and grace
 * periods.  The caller must have disabled interrupts.
 */
2643
static void rcu_sysidle_enter(int irq)
2644 2645
{
	unsigned long j;
2646
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
2647

2648 2649 2650 2651
	/* If there are no nohz_full= CPUs, no need to track this. */
	if (!tick_nohz_full_enabled())
		return;

2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670
	/* Adjust nesting, check for fully idle. */
	if (irq) {
		rdtp->dynticks_idle_nesting--;
		WARN_ON_ONCE(rdtp->dynticks_idle_nesting < 0);
		if (rdtp->dynticks_idle_nesting != 0)
			return;  /* Still not fully idle. */
	} else {
		if ((rdtp->dynticks_idle_nesting & DYNTICK_TASK_NEST_MASK) ==
		    DYNTICK_TASK_NEST_VALUE) {
			rdtp->dynticks_idle_nesting = 0;
		} else {
			rdtp->dynticks_idle_nesting -= DYNTICK_TASK_NEST_VALUE;
			WARN_ON_ONCE(rdtp->dynticks_idle_nesting < 0);
			return;  /* Still not fully idle. */
		}
	}

	/* Record start of fully idle period. */
	j = jiffies;
2671
	WRITE_ONCE(rdtp->dynticks_idle_jiffies, j);
2672
	smp_mb__before_atomic();
2673
	atomic_inc(&rdtp->dynticks_idle);
2674
	smp_mb__after_atomic();
2675 2676 2677
	WARN_ON_ONCE(atomic_read(&rdtp->dynticks_idle) & 0x1);
}

2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688
/*
 * Unconditionally force exit from full system-idle state.  This is
 * invoked when a normal CPU exits idle, but must be called separately
 * for the timekeeping CPU (tick_do_timer_cpu).  The reason for this
 * is that the timekeeping CPU is permitted to take scheduling-clock
 * interrupts while the system is in system-idle state, and of course
 * rcu_sysidle_exit() has no way of distinguishing a scheduling-clock
 * interrupt from any other type of interrupt.
 */
void rcu_sysidle_force_exit(void)
{
2689
	int oldstate = READ_ONCE(full_sysidle_state);
2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709
	int newoldstate;

	/*
	 * Each pass through the following loop attempts to exit full
	 * system-idle state.  If contention proves to be a problem,
	 * a trylock-based contention tree could be used here.
	 */
	while (oldstate > RCU_SYSIDLE_SHORT) {
		newoldstate = cmpxchg(&full_sysidle_state,
				      oldstate, RCU_SYSIDLE_NOT);
		if (oldstate == newoldstate &&
		    oldstate == RCU_SYSIDLE_FULL_NOTED) {
			rcu_kick_nohz_cpu(tick_do_timer_cpu);
			return; /* We cleared it, done! */
		}
		oldstate = newoldstate;
	}
	smp_mb(); /* Order initial oldstate fetch vs. later non-idle work. */
}

2710 2711 2712 2713 2714
/*
 * Invoked to note entry to irq or task transition from idle.  Note that
 * usermode execution does -not- count as idle here!  The caller must
 * have disabled interrupts.
 */
2715
static void rcu_sysidle_exit(int irq)
2716
{
2717 2718
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);

2719 2720 2721 2722
	/* If there are no nohz_full= CPUs, no need to track this. */
	if (!tick_nohz_full_enabled())
		return;

2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744
	/* Adjust nesting, check for already non-idle. */
	if (irq) {
		rdtp->dynticks_idle_nesting++;
		WARN_ON_ONCE(rdtp->dynticks_idle_nesting <= 0);
		if (rdtp->dynticks_idle_nesting != 1)
			return; /* Already non-idle. */
	} else {
		/*
		 * Allow for irq misnesting.  Yes, it really is possible
		 * to enter an irq handler then never leave it, and maybe
		 * also vice versa.  Handle both possibilities.
		 */
		if (rdtp->dynticks_idle_nesting & DYNTICK_TASK_NEST_MASK) {
			rdtp->dynticks_idle_nesting += DYNTICK_TASK_NEST_VALUE;
			WARN_ON_ONCE(rdtp->dynticks_idle_nesting <= 0);
			return; /* Already non-idle. */
		} else {
			rdtp->dynticks_idle_nesting = DYNTICK_TASK_EXIT_IDLE;
		}
	}

	/* Record end of idle period. */
2745
	smp_mb__before_atomic();
2746
	atomic_inc(&rdtp->dynticks_idle);
2747
	smp_mb__after_atomic();
2748
	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks_idle) & 0x1));
2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767

	/*
	 * If we are the timekeeping CPU, we are permitted to be non-idle
	 * during a system-idle state.  This must be the case, because
	 * the timekeeping CPU has to take scheduling-clock interrupts
	 * during the time that the system is transitioning to full
	 * system-idle state.  This means that the timekeeping CPU must
	 * invoke rcu_sysidle_force_exit() directly if it does anything
	 * more than take a scheduling-clock interrupt.
	 */
	if (smp_processor_id() == tick_do_timer_cpu)
		return;

	/* Update system-idle state: We are clearly no longer fully idle! */
	rcu_sysidle_force_exit();
}

/*
 * Check to see if the current CPU is idle.  Note that usermode execution
2768 2769
 * does not count as idle.  The caller must have disabled interrupts,
 * and must be running on tick_do_timer_cpu.
2770 2771 2772 2773 2774 2775 2776 2777
 */
static void rcu_sysidle_check_cpu(struct rcu_data *rdp, bool *isidle,
				  unsigned long *maxj)
{
	int cur;
	unsigned long j;
	struct rcu_dynticks *rdtp = rdp->dynticks;

2778 2779 2780 2781
	/* If there are no nohz_full= CPUs, don't check system-wide idleness. */
	if (!tick_nohz_full_enabled())
		return;

2782 2783 2784 2785 2786
	/*
	 * If some other CPU has already reported non-idle, if this is
	 * not the flavor of RCU that tracks sysidle state, or if this
	 * is an offline or the timekeeping CPU, nothing to do.
	 */
2787
	if (!*isidle || rdp->rsp != rcu_state_p ||
2788 2789
	    cpu_is_offline(rdp->cpu) || rdp->cpu == tick_do_timer_cpu)
		return;
2790 2791
	/* Verify affinity of current kthread. */
	WARN_ON_ONCE(smp_processor_id() != tick_do_timer_cpu);
2792 2793 2794 2795 2796 2797 2798 2799 2800 2801

	/* Pick up current idle and NMI-nesting counter and check. */
	cur = atomic_read(&rdtp->dynticks_idle);
	if (cur & 0x1) {
		*isidle = false; /* We are not idle! */
		return;
	}
	smp_mb(); /* Read counters before timestamps. */

	/* Pick up timestamps. */
2802
	j = READ_ONCE(rdtp->dynticks_idle_jiffies);
2803 2804 2805 2806 2807 2808 2809 2810 2811 2812
	/* If this CPU entered idle more recently, update maxj timestamp. */
	if (ULONG_CMP_LT(*maxj, j))
		*maxj = j;
}

/*
 * Is this the flavor of RCU that is handling full-system idle?
 */
static bool is_sysidle_rcu_state(struct rcu_state *rsp)
{
2813
	return rsp == rcu_state_p;
2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
}

/*
 * Return a delay in jiffies based on the number of CPUs, rcu_node
 * leaf fanout, and jiffies tick rate.  The idea is to allow larger
 * systems more time to transition to full-idle state in order to
 * avoid the cache thrashing that otherwise occur on the state variable.
 * Really small systems (less than a couple of tens of CPUs) should
 * instead use a single global atomically incremented counter, and later
 * versions of this will automatically reconfigure themselves accordingly.
 */
static unsigned long rcu_sysidle_delay(void)
{
	if (nr_cpu_ids <= CONFIG_NO_HZ_FULL_SYSIDLE_SMALL)
		return 0;
	return DIV_ROUND_UP(nr_cpu_ids * HZ, rcu_fanout_leaf * 1000);
}

/*
 * Advance the full-system-idle state.  This is invoked when all of
 * the non-timekeeping CPUs are idle.
 */
static void rcu_sysidle(unsigned long j)
{
	/* Check the current state. */
2839
	switch (READ_ONCE(full_sysidle_state)) {
2840 2841 2842
	case RCU_SYSIDLE_NOT:

		/* First time all are idle, so note a short idle period. */
2843
		WRITE_ONCE(full_sysidle_state, RCU_SYSIDLE_SHORT);
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
		break;

	case RCU_SYSIDLE_SHORT:

		/*
		 * Idle for a bit, time to advance to next state?
		 * cmpxchg failure means race with non-idle, let them win.
		 */
		if (ULONG_CMP_GE(jiffies, j + rcu_sysidle_delay()))
			(void)cmpxchg(&full_sysidle_state,
				      RCU_SYSIDLE_SHORT, RCU_SYSIDLE_LONG);
		break;

	case RCU_SYSIDLE_LONG:

		/*
		 * Do an additional check pass before advancing to full.
		 * cmpxchg failure means race with non-idle, let them win.
		 */
		if (ULONG_CMP_GE(jiffies, j + rcu_sysidle_delay()))
			(void)cmpxchg(&full_sysidle_state,
				      RCU_SYSIDLE_LONG, RCU_SYSIDLE_FULL);
		break;

	default:
		break;
	}
}

/*
 * Found a non-idle non-timekeeping CPU, so kick the system-idle state
 * back to the beginning.
 */
static void rcu_sysidle_cancel(void)
{
	smp_mb();
2880
	if (full_sysidle_state > RCU_SYSIDLE_SHORT)
2881
		WRITE_ONCE(full_sysidle_state, RCU_SYSIDLE_NOT);
2882 2883 2884 2885 2886 2887 2888 2889 2890
}

/*
 * Update the sysidle state based on the results of a force-quiescent-state
 * scan of the CPUs' dyntick-idle state.
 */
static void rcu_sysidle_report(struct rcu_state *rsp, int isidle,
			       unsigned long maxj, bool gpkt)
{
2891
	if (rsp != rcu_state_p)
2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907
		return;  /* Wrong flavor, ignore. */
	if (gpkt && nr_cpu_ids <= CONFIG_NO_HZ_FULL_SYSIDLE_SMALL)
		return;  /* Running state machine from timekeeping CPU. */
	if (isidle)
		rcu_sysidle(maxj);    /* More idle! */
	else
		rcu_sysidle_cancel(); /* Idle is over. */
}

/*
 * Wrapper for rcu_sysidle_report() when called from the grace-period
 * kthread's context.
 */
static void rcu_sysidle_report_gp(struct rcu_state *rsp, int isidle,
				  unsigned long maxj)
{
2908 2909 2910 2911
	/* If there are no nohz_full= CPUs, no need to track this. */
	if (!tick_nohz_full_enabled())
		return;

2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932
	rcu_sysidle_report(rsp, isidle, maxj, true);
}

/* Callback and function for forcing an RCU grace period. */
struct rcu_sysidle_head {
	struct rcu_head rh;
	int inuse;
};

static void rcu_sysidle_cb(struct rcu_head *rhp)
{
	struct rcu_sysidle_head *rshp;

	/*
	 * The following memory barrier is needed to replace the
	 * memory barriers that would normally be in the memory
	 * allocator.
	 */
	smp_mb();  /* grace period precedes setting inuse. */

	rshp = container_of(rhp, struct rcu_sysidle_head, rh);
2933
	WRITE_ONCE(rshp->inuse, 0);
2934 2935 2936 2937
}

/*
 * Check to see if the system is fully idle, other than the timekeeping CPU.
2938 2939
 * The caller must have disabled interrupts.  This is not intended to be
 * called unless tick_nohz_full_enabled().
2940 2941 2942 2943
 */
bool rcu_sys_is_idle(void)
{
	static struct rcu_sysidle_head rsh;
2944
	int rss = READ_ONCE(full_sysidle_state);
2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964

	if (WARN_ON_ONCE(smp_processor_id() != tick_do_timer_cpu))
		return false;

	/* Handle small-system case by doing a full scan of CPUs. */
	if (nr_cpu_ids <= CONFIG_NO_HZ_FULL_SYSIDLE_SMALL) {
		int oldrss = rss - 1;

		/*
		 * One pass to advance to each state up to _FULL.
		 * Give up if any pass fails to advance the state.
		 */
		while (rss < RCU_SYSIDLE_FULL && oldrss < rss) {
			int cpu;
			bool isidle = true;
			unsigned long maxj = jiffies - ULONG_MAX / 4;
			struct rcu_data *rdp;

			/* Scan all the CPUs looking for nonidle CPUs. */
			for_each_possible_cpu(cpu) {
2965
				rdp = per_cpu_ptr(rcu_state_p->rda, cpu);
2966 2967 2968 2969
				rcu_sysidle_check_cpu(rdp, &isidle, &maxj);
				if (!isidle)
					break;
			}
2970
			rcu_sysidle_report(rcu_state_p, isidle, maxj, false);
2971
			oldrss = rss;
2972
			rss = READ_ONCE(full_sysidle_state);
2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996
		}
	}

	/* If this is the first observation of an idle period, record it. */
	if (rss == RCU_SYSIDLE_FULL) {
		rss = cmpxchg(&full_sysidle_state,
			      RCU_SYSIDLE_FULL, RCU_SYSIDLE_FULL_NOTED);
		return rss == RCU_SYSIDLE_FULL;
	}

	smp_mb(); /* ensure rss load happens before later caller actions. */

	/* If already fully idle, tell the caller (in case of races). */
	if (rss == RCU_SYSIDLE_FULL_NOTED)
		return true;

	/*
	 * If we aren't there yet, and a grace period is not in flight,
	 * initiate a grace period.  Either way, tell the caller that
	 * we are not there yet.  We use an xchg() rather than an assignment
	 * to make up for the memory barriers that would otherwise be
	 * provided by the memory allocator.
	 */
	if (nr_cpu_ids > CONFIG_NO_HZ_FULL_SYSIDLE_SMALL &&
2997
	    !rcu_gp_in_progress(rcu_state_p) &&
2998 2999 3000
	    !rsh.inuse && xchg(&rsh.inuse, 1) == 0)
		call_rcu(&rsh.rh, rcu_sysidle_cb);
	return false;
3001 3002
}

3003 3004 3005 3006 3007 3008 3009 3010 3011 3012
/*
 * Initialize dynticks sysidle state for CPUs coming online.
 */
static void rcu_sysidle_init_percpu_data(struct rcu_dynticks *rdtp)
{
	rdtp->dynticks_idle_nesting = DYNTICK_TASK_NEST_VALUE;
}

#else /* #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */

3013
static void rcu_sysidle_enter(int irq)
3014 3015 3016
{
}

3017
static void rcu_sysidle_exit(int irq)
3018 3019 3020
{
}

3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035
static void rcu_sysidle_check_cpu(struct rcu_data *rdp, bool *isidle,
				  unsigned long *maxj)
{
}

static bool is_sysidle_rcu_state(struct rcu_state *rsp)
{
	return false;
}

static void rcu_sysidle_report_gp(struct rcu_state *rsp, int isidle,
				  unsigned long maxj)
{
}

3036 3037 3038 3039 3040
static void rcu_sysidle_init_percpu_data(struct rcu_dynticks *rdtp)
{
}

#endif /* #else #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
3041 3042 3043 3044 3045 3046 3047 3048

/*
 * Is this CPU a NO_HZ_FULL CPU that should ignore RCU so that the
 * grace-period kthread will do force_quiescent_state() processing?
 * The idea is to avoid waking up RCU core processing on such a
 * CPU unless the grace period has extended for too long.
 *
 * This code relies on the fact that all NO_HZ_FULL CPUs are also
3049
 * CONFIG_RCU_NOCB_CPU CPUs.
3050 3051 3052 3053 3054 3055
 */
static bool rcu_nohz_full_cpu(struct rcu_state *rsp)
{
#ifdef CONFIG_NO_HZ_FULL
	if (tick_nohz_full_cpu(smp_processor_id()) &&
	    (!rcu_gp_in_progress(rsp) ||
3056
	     ULONG_CMP_LT(jiffies, READ_ONCE(rsp->gp_start) + HZ)))
3057 3058 3059 3060
		return 1;
#endif /* #ifdef CONFIG_NO_HZ_FULL */
	return 0;
}
3061 3062 3063 3064 3065 3066 3067

/*
 * Bind the grace-period kthread for the sysidle flavor of RCU to the
 * timekeeping CPU.
 */
static void rcu_bind_gp_kthread(void)
{
3068
	int __maybe_unused cpu;
3069

3070
	if (!tick_nohz_full_enabled())
3071
		return;
3072 3073
#ifdef CONFIG_NO_HZ_FULL_SYSIDLE
	cpu = tick_do_timer_cpu;
3074
	if (cpu >= 0 && cpu < nr_cpu_ids)
3075
		set_cpus_allowed_ptr(current, cpumask_of(cpu));
3076
#else /* #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
3077
	housekeeping_affine(current);
3078
#endif /* #else #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
3079
}
3080 3081 3082 3083 3084

/* Record the current task on dyntick-idle entry. */
static void rcu_dynticks_task_enter(void)
{
#if defined(CONFIG_TASKS_RCU) && defined(CONFIG_NO_HZ_FULL)
3085
	WRITE_ONCE(current->rcu_tasks_idle_cpu, smp_processor_id());
3086 3087 3088 3089 3090 3091 3092
#endif /* #if defined(CONFIG_TASKS_RCU) && defined(CONFIG_NO_HZ_FULL) */
}

/* Record no current task on dyntick-idle exit. */
static void rcu_dynticks_task_exit(void)
{
#if defined(CONFIG_TASKS_RCU) && defined(CONFIG_NO_HZ_FULL)
3093
	WRITE_ONCE(current->rcu_tasks_idle_cpu, -1);
3094 3095
#endif /* #if defined(CONFIG_TASKS_RCU) && defined(CONFIG_NO_HZ_FULL) */
}