tree_plugin.h 89.9 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
/*
 * Check the RCU kernel configuration parameters and print informative
66
 * messages about anything out of the ordinary.
67 68 69
 */
static void __init rcu_bootup_announce_oddness(void)
{
70 71
	if (IS_ENABLED(CONFIG_RCU_TRACE))
		pr_info("\tRCU debugfs-based tracing is enabled.\n");
72 73
	if ((IS_ENABLED(CONFIG_64BIT) && RCU_FANOUT != 64) ||
	    (!IS_ENABLED(CONFIG_64BIT) && RCU_FANOUT != 32))
74
		pr_info("\tCONFIG_RCU_FANOUT set to non-default value of %d\n",
75
		       RCU_FANOUT);
76
	if (rcu_fanout_exact)
77 78 79 80 81 82 83
		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");
84 85
	if (RCU_NUM_LVLS >= 4)
		pr_info("\tFour(or more)-level hierarchy is enabled.\n");
86
	if (RCU_FANOUT_LEAF != 16)
87
		pr_info("\tBuild-time adjustment of leaf fanout to %d.\n",
88 89
			RCU_FANOUT_LEAF);
	if (rcu_fanout_leaf != RCU_FANOUT_LEAF)
90
		pr_info("\tBoot-time adjustment of leaf fanout to %d.\n", rcu_fanout_leaf);
91
	if (nr_cpu_ids != NR_CPUS)
92
		pr_info("\tRCU restricting CPUs from NR_CPUS=%d to nr_cpu_ids=%d.\n", NR_CPUS, nr_cpu_ids);
93 94
	if (IS_ENABLED(CONFIG_RCU_BOOST))
		pr_info("\tRCU kthread priority: %d.\n", kthread_prio);
95 96
}

97
#ifdef CONFIG_PREEMPT_RCU
98

99
RCU_STATE_INITIALIZER(rcu_preempt, 'p', call_rcu);
100
static struct rcu_state *const rcu_state_p = &rcu_preempt_state;
101
static struct rcu_data __percpu *const rcu_data_p = &rcu_preempt_data;
102

103 104
static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp,
			       bool wake);
105

106 107 108
/*
 * Tell them what RCU they are running.
 */
109
static void __init rcu_bootup_announce(void)
110
{
111
	pr_info("Preemptible hierarchical RCU implementation.\n");
112
	rcu_bootup_announce_oddness();
113 114
}

115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
/* Flags for rcu_preempt_ctxt_queue() decision table. */
#define RCU_GP_TASKS	0x8
#define RCU_EXP_TASKS	0x4
#define RCU_GP_BLKD	0x2
#define RCU_EXP_BLKD	0x1

/*
 * Queues a task preempted within an RCU-preempt read-side critical
 * section into the appropriate location within the ->blkd_tasks list,
 * depending on the states of any ongoing normal and expedited grace
 * periods.  The ->gp_tasks pointer indicates which element the normal
 * grace period is waiting on (NULL if none), and the ->exp_tasks pointer
 * indicates which element the expedited grace period is waiting on (again,
 * NULL if none).  If a grace period is waiting on a given element in the
 * ->blkd_tasks list, it also waits on all subsequent elements.  Thus,
 * adding a task to the tail of the list blocks any grace period that is
 * already waiting on one of the elements.  In contrast, adding a task
 * to the head of the list won't block any grace period that is already
 * waiting on one of the elements.
 *
 * This queuing is imprecise, and can sometimes make an ongoing grace
 * period wait for a task that is not strictly speaking blocking it.
 * Given the choice, we needlessly block a normal grace period rather than
 * blocking an expedited grace period.
 *
 * Note that an endless sequence of expedited grace periods still cannot
 * indefinitely postpone a normal grace period.  Eventually, all of the
 * fixed number of preempted tasks blocking the normal grace period that are
 * not also blocking the expedited grace period will resume and complete
 * their RCU read-side critical sections.  At that point, the ->gp_tasks
 * pointer will equal the ->exp_tasks pointer, at which point the end of
 * the corresponding expedited grace period will also be the end of the
 * normal grace period.
 */
static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp,
				   unsigned long flags) __releases(rnp->lock)
{
	int blkd_state = (rnp->gp_tasks ? RCU_GP_TASKS : 0) +
			 (rnp->exp_tasks ? RCU_EXP_TASKS : 0) +
			 (rnp->qsmask & rdp->grpmask ? RCU_GP_BLKD : 0) +
			 (rnp->expmask & rdp->grpmask ? RCU_EXP_BLKD : 0);
	struct task_struct *t = current;

	/*
	 * Decide where to queue the newly blocked task.  In theory,
	 * this could be an if-statement.  In practice, when I tried
	 * that, it was quite messy.
	 */
	switch (blkd_state) {
	case 0:
	case                RCU_EXP_TASKS:
	case                RCU_EXP_TASKS + RCU_GP_BLKD:
	case RCU_GP_TASKS:
	case RCU_GP_TASKS + RCU_EXP_TASKS:

		/*
		 * Blocking neither GP, or first task blocking the normal
		 * GP but not blocking the already-waiting expedited GP.
		 * Queue at the head of the list to avoid unnecessarily
		 * blocking the already-waiting GPs.
		 */
		list_add(&t->rcu_node_entry, &rnp->blkd_tasks);
		break;

	case                                              RCU_EXP_BLKD:
	case                                RCU_GP_BLKD:
	case                                RCU_GP_BLKD + RCU_EXP_BLKD:
	case RCU_GP_TASKS +                               RCU_EXP_BLKD:
	case RCU_GP_TASKS +                 RCU_GP_BLKD + RCU_EXP_BLKD:
	case RCU_GP_TASKS + RCU_EXP_TASKS + RCU_GP_BLKD + RCU_EXP_BLKD:

		/*
		 * First task arriving that blocks either GP, or first task
		 * arriving that blocks the expedited GP (with the normal
		 * GP already waiting), or a task arriving that blocks
		 * both GPs with both GPs already waiting.  Queue at the
		 * tail of the list to avoid any GP waiting on any of the
		 * already queued tasks that are not blocking it.
		 */
		list_add_tail(&t->rcu_node_entry, &rnp->blkd_tasks);
		break;

	case                RCU_EXP_TASKS +               RCU_EXP_BLKD:
	case                RCU_EXP_TASKS + RCU_GP_BLKD + RCU_EXP_BLKD:
	case RCU_GP_TASKS + RCU_EXP_TASKS +               RCU_EXP_BLKD:

		/*
		 * Second or subsequent task blocking the expedited GP.
		 * The task either does not block the normal GP, or is the
		 * first task blocking the normal GP.  Queue just after
		 * the first task blocking the expedited GP.
		 */
		list_add(&t->rcu_node_entry, rnp->exp_tasks);
		break;

	case RCU_GP_TASKS +                 RCU_GP_BLKD:
	case RCU_GP_TASKS + RCU_EXP_TASKS + RCU_GP_BLKD:

		/*
		 * Second or subsequent task blocking the normal GP.
		 * The task does not block the expedited GP. Queue just
		 * after the first task blocking the normal GP.
		 */
		list_add(&t->rcu_node_entry, rnp->gp_tasks);
		break;

	default:

		/* Yet another exercise in excessive paranoia. */
		WARN_ON_ONCE(1);
		break;
	}

	/*
	 * We have now queued the task.  If it was the first one to
	 * block either grace period, update the ->gp_tasks and/or
	 * ->exp_tasks pointers, respectively, to reference the newly
	 * blocked tasks.
	 */
	if (!rnp->gp_tasks && (blkd_state & RCU_GP_BLKD))
		rnp->gp_tasks = &t->rcu_node_entry;
	if (!rnp->exp_tasks && (blkd_state & RCU_EXP_BLKD))
		rnp->exp_tasks = &t->rcu_node_entry;
	raw_spin_unlock(&rnp->lock);

	/*
	 * Report the quiescent state for the expedited GP.  This expedited
	 * GP should not be able to end until we report, so there should be
	 * no need to check for a subsequent expedited GP.  (Though we are
	 * still in a quiescent state in any case.)
	 */
	if (blkd_state & RCU_EXP_BLKD &&
	    t->rcu_read_unlock_special.b.exp_need_qs) {
		t->rcu_read_unlock_special.b.exp_need_qs = false;
		rcu_report_exp_rdp(rdp->rsp, rdp, true);
	} else {
		WARN_ON_ONCE(t->rcu_read_unlock_special.b.exp_need_qs);
	}
	local_irq_restore(flags);
}

256
/*
P
Paul E. McKenney 已提交
257
 * Record a preemptible-RCU quiescent state for the specified CPU.  Note
258 259 260
 * 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.
261
 *
262 263
 * As with the other rcu_*_qs() functions, callers to this function
 * must disable preemption.
264
 */
265
static void rcu_preempt_qs(void)
266
{
267
	if (__this_cpu_read(rcu_data_p->cpu_no_qs.s)) {
268
		trace_rcu_grace_period(TPS("rcu_preempt"),
269
				       __this_cpu_read(rcu_data_p->gpnum),
270
				       TPS("cpuqs"));
271
		__this_cpu_write(rcu_data_p->cpu_no_qs.b.norm, false);
272 273 274
		barrier(); /* Coordinate with rcu_preempt_check_callbacks(). */
		current->rcu_read_unlock_special.b.need_qs = false;
	}
275 276 277
}

/*
278 279 280
 * 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
281 282 283 284 285 286
 * 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.
287 288
 *
 * Caller must disable preemption.
289
 */
290
static void rcu_preempt_note_context_switch(void)
291 292
{
	struct task_struct *t = current;
293
	unsigned long flags;
294 295 296
	struct rcu_data *rdp;
	struct rcu_node *rnp;

297
	if (t->rcu_read_lock_nesting > 0 &&
298
	    !t->rcu_read_unlock_special.b.blocked) {
299 300

		/* Possibly blocking in an RCU read-side critical section. */
301
		rdp = this_cpu_ptr(rcu_state_p->rda);
302
		rnp = rdp->mynode;
303
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
304
		t->rcu_read_unlock_special.b.blocked = true;
305
		t->rcu_blocked_node = rnp;
306 307

		/*
308 309 310
		 * Verify the CPU's sanity, trace the preemption, and
		 * then queue the task as required based on the states
		 * of any ongoing and expedited grace periods.
311
		 */
312
		WARN_ON_ONCE((rdp->grpmask & rcu_rnp_online_cpus(rnp)) == 0);
313
		WARN_ON_ONCE(!list_empty(&t->rcu_node_entry));
314 315 316 317 318
		trace_rcu_preempt_task(rdp->rsp->name,
				       t->pid,
				       (rnp->qsmask & rdp->grpmask)
				       ? rnp->gpnum
				       : rnp->gpnum + 1);
319
		rcu_preempt_ctxt_queue(rnp, rdp, flags);
320
	} else if (t->rcu_read_lock_nesting < 0 &&
321
		   t->rcu_read_unlock_special.s) {
322 323 324 325 326 327

		/*
		 * Complete exit from RCU read-side critical section on
		 * behalf of preempted instance of __rcu_read_unlock().
		 */
		rcu_read_unlock_special(t);
328 329 330 331 332 333 334 335 336 337 338
	}

	/*
	 * 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.
	 */
339
	rcu_preempt_qs();
340 341
}

342 343 344 345 346
/*
 * 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.
 */
347
static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp)
348
{
349
	return rnp->gp_tasks != NULL;
350 351
}

352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
/*
 * 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;
}

367 368 369 370 371 372 373 374 375
/*
 * 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);
}

376 377 378 379 380
/*
 * 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.
 */
381
void rcu_read_unlock_special(struct task_struct *t)
382
{
383 384 385
	bool empty_exp;
	bool empty_norm;
	bool empty_exp_now;
386
	unsigned long flags;
387
	struct list_head *np;
388
	bool drop_boost_mutex = false;
389
	struct rcu_data *rdp;
390
	struct rcu_node *rnp;
391
	union rcu_special special;
392 393 394 395 396 397 398 399

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

	local_irq_save(flags);

	/*
400 401
	 * If RCU core is waiting for this CPU to exit its critical section,
	 * report the fact that it has exited.  Because irqs are disabled,
402
	 * t->rcu_read_unlock_special cannot change.
403 404
	 */
	special = t->rcu_read_unlock_special;
405
	if (special.b.need_qs) {
406
		rcu_preempt_qs();
407
		t->rcu_read_unlock_special.b.need_qs = false;
408
		if (!t->rcu_read_unlock_special.s) {
409 410 411
			local_irq_restore(flags);
			return;
		}
412 413
	}

414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
	/*
	 * Respond to a request for an expedited grace period, but only if
	 * we were not preempted, meaning that we were running on the same
	 * CPU throughout.  If we were preempted, the exp_need_qs flag
	 * would have been cleared at the time of the first preemption,
	 * and the quiescent state would be reported when we were dequeued.
	 */
	if (special.b.exp_need_qs) {
		WARN_ON_ONCE(special.b.blocked);
		t->rcu_read_unlock_special.b.exp_need_qs = false;
		rdp = this_cpu_ptr(rcu_state_p->rda);
		rcu_report_exp_rdp(rcu_state_p, rdp, true);
		if (!t->rcu_read_unlock_special.s) {
			local_irq_restore(flags);
			return;
		}
	}

432
	/* Hardware IRQ handlers cannot block, complain if they get here. */
433 434 435
	if (in_irq() || in_serving_softirq()) {
		lockdep_rcu_suspicious(__FILE__, __LINE__,
				       "rcu_read_unlock() from irq or softirq with blocking in critical section!!!\n");
436
		pr_alert("->rcu_read_unlock_special: %#x (b: %d, enq: %d nq: %d)\n",
437 438
			 t->rcu_read_unlock_special.s,
			 t->rcu_read_unlock_special.b.blocked,
439
			 t->rcu_read_unlock_special.b.exp_need_qs,
440
			 t->rcu_read_unlock_special.b.need_qs);
441 442 443 444 445
		local_irq_restore(flags);
		return;
	}

	/* Clean up if blocked during RCU read-side critical section. */
446 447
	if (special.b.blocked) {
		t->rcu_read_unlock_special.b.blocked = false;
448

449
		/*
450
		 * Remove this task from the list it blocked on.  The task
451 452 453
		 * now remains queued on the rcu_node corresponding to the
		 * CPU it first blocked on, so there is no longer any need
		 * to loop.  Retain a WARN_ON_ONCE() out of sheer paranoia.
454
		 */
455 456 457
		rnp = t->rcu_blocked_node;
		raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */
		WARN_ON_ONCE(rnp != t->rcu_blocked_node);
458
		empty_norm = !rcu_preempt_blocked_readers_cgp(rnp);
459
		empty_exp = sync_rcu_preempt_exp_done(rnp);
460
		smp_mb(); /* ensure expedited fastpath sees end of RCU c-s. */
461
		np = rcu_next_node_entry(t, rnp);
462
		list_del_init(&t->rcu_node_entry);
463
		t->rcu_blocked_node = NULL;
464
		trace_rcu_unlock_preempted_task(TPS("rcu_preempt"),
465
						rnp->gpnum, t->pid);
466 467 468 469
		if (&t->rcu_node_entry == rnp->gp_tasks)
			rnp->gp_tasks = np;
		if (&t->rcu_node_entry == rnp->exp_tasks)
			rnp->exp_tasks = np;
470 471 472 473 474 475
		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;
		}
476 477 478 479

		/*
		 * If this was the last task on the current list, and if
		 * we aren't waiting on any CPUs, report the quiescent state.
480 481
		 * Note that rcu_report_unblock_qs_rnp() releases rnp->lock,
		 * so we must take a snapshot of the expedited state.
482
		 */
483
		empty_exp_now = sync_rcu_preempt_exp_done(rnp);
484
		if (!empty_norm && !rcu_preempt_blocked_readers_cgp(rnp)) {
485
			trace_rcu_quiescent_state_report(TPS("preempt_rcu"),
486 487 488 489 490 491
							 rnp->gpnum,
							 0, rnp->qsmask,
							 rnp->level,
							 rnp->grplo,
							 rnp->grphi,
							 !!rnp->gp_tasks);
492
			rcu_report_unblock_qs_rnp(rcu_state_p, rnp, flags);
493
		} else {
494
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
495
		}
496

497
		/* Unboost if we were boosted. */
498
		if (IS_ENABLED(CONFIG_RCU_BOOST) && drop_boost_mutex)
499
			rt_mutex_unlock(&rnp->boost_mtx);
500

501 502 503 504
		/*
		 * If this was the last task on the expedited lists,
		 * then we need to report up the rcu_node hierarchy.
		 */
505
		if (!empty_exp && empty_exp_now)
506
			rcu_report_exp_rnp(rcu_state_p, rnp, true);
507 508
	} else {
		local_irq_restore(flags);
509 510 511
	}
}

512 513 514 515 516 517 518 519 520
/*
 * 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;

521
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
522 523 524 525
	if (!rcu_preempt_blocked_readers_cgp(rnp)) {
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
		return;
	}
526
	t = list_entry(rnp->gp_tasks->prev,
527 528 529 530
		       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);
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
}

/*
 * 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);
}

546 547
static void rcu_print_task_stall_begin(struct rcu_node *rnp)
{
548
	pr_err("\tTasks blocked on level-%d rcu_node (CPUs %d-%d):",
549 550 551 552 553
	       rnp->level, rnp->grplo, rnp->grphi);
}

static void rcu_print_task_stall_end(void)
{
554
	pr_cont("\n");
555 556
}

557 558 559 560
/*
 * Scan the current list of tasks blocked within RCU read-side critical
 * sections, printing out the tid of each.
 */
561
static int rcu_print_task_stall(struct rcu_node *rnp)
562 563
{
	struct task_struct *t;
564
	int ndetected = 0;
565

566
	if (!rcu_preempt_blocked_readers_cgp(rnp))
567
		return 0;
568
	rcu_print_task_stall_begin(rnp);
569
	t = list_entry(rnp->gp_tasks->prev,
570
		       struct task_struct, rcu_node_entry);
571
	list_for_each_entry_continue(t, &rnp->blkd_tasks, rcu_node_entry) {
572
		pr_cont(" P%d", t->pid);
573 574
		ndetected++;
	}
575
	rcu_print_task_stall_end();
576
	return ndetected;
577 578
}

579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
/*
 * Scan the current list of tasks blocked within RCU read-side critical
 * sections, printing out the tid of each that is blocking the current
 * expedited grace period.
 */
static int rcu_print_task_exp_stall(struct rcu_node *rnp)
{
	struct task_struct *t;
	int ndetected = 0;

	if (!rnp->exp_tasks)
		return 0;
	t = list_entry(rnp->exp_tasks->prev,
		       struct task_struct, rcu_node_entry);
	list_for_each_entry_continue(t, &rnp->blkd_tasks, rcu_node_entry) {
		pr_cont(" P%d", t->pid);
		ndetected++;
	}
	return ndetected;
}

600 601 602 603 604 605
/*
 * 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.
606 607 608
 *
 * Also, if there are blocked tasks on the list, they automatically
 * block the newly created grace period, so set up ->gp_tasks accordingly.
609 610 611
 */
static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
{
612
	WARN_ON_ONCE(rcu_preempt_blocked_readers_cgp(rnp));
613
	if (rcu_preempt_has_tasks(rnp))
614
		rnp->gp_tasks = rnp->blkd_tasks.next;
615
	WARN_ON_ONCE(rnp->qsmask);
616 617
}

618 619 620 621 622 623 624
/*
 * 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.
 */
625
static void rcu_preempt_check_callbacks(void)
626 627 628 629
{
	struct task_struct *t = current;

	if (t->rcu_read_lock_nesting == 0) {
630
		rcu_preempt_qs();
631 632
		return;
	}
633
	if (t->rcu_read_lock_nesting > 0 &&
634
	    __this_cpu_read(rcu_data_p->core_needs_qs) &&
635
	    __this_cpu_read(rcu_data_p->cpu_no_qs.b.norm))
636
		t->rcu_read_unlock_special.b.need_qs = true;
637 638
}

639 640
#ifdef CONFIG_RCU_BOOST

641 642
static void rcu_preempt_do_callbacks(void)
{
643
	rcu_do_batch(rcu_state_p, this_cpu_ptr(rcu_data_p));
644 645
}

646 647
#endif /* #ifdef CONFIG_RCU_BOOST */

648
/*
P
Paul E. McKenney 已提交
649
 * Queue a preemptible-RCU callback for invocation after a grace period.
650
 */
651
void call_rcu(struct rcu_head *head, rcu_callback_t func)
652
{
653
	__call_rcu(head, func, rcu_state_p, -1, 0);
654 655 656
}
EXPORT_SYMBOL_GPL(call_rcu);

657 658 659 660 661
/**
 * 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
662 663 664 665 666
 * 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.
667 668 669
 *
 * See the description of synchronize_sched() for more detailed information
 * on memory ordering guarantees.
670 671 672
 */
void synchronize_rcu(void)
{
673 674 675 676
	RCU_LOCKDEP_WARN(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");
677 678
	if (!rcu_scheduler_active)
		return;
679
	if (rcu_gp_is_expedited())
680 681 682
		synchronize_rcu_expedited();
	else
		wait_rcu_gp(call_rcu);
683 684 685
}
EXPORT_SYMBOL_GPL(synchronize_rcu);

686
/*
687 688 689 690 691
 * Remote handler for smp_call_function_single().  If there is an
 * RCU read-side critical section in effect, request that the
 * next rcu_read_unlock() record the quiescent state up the
 * ->expmask fields in the rcu_node tree.  Otherwise, immediately
 * report the quiescent state.
692
 */
693
static void sync_rcu_exp_handler(void *info)
694
{
695 696 697
	struct rcu_data *rdp;
	struct rcu_state *rsp = info;
	struct task_struct *t = current;
698 699

	/*
700 701 702 703
	 * Within an RCU read-side critical section, request that the next
	 * rcu_read_unlock() report.  Unless this RCU read-side critical
	 * section has already blocked, in which case it is already set
	 * up for the expedited grace period to wait on it.
704
	 */
705 706 707
	if (t->rcu_read_lock_nesting > 0 &&
	    !t->rcu_read_unlock_special.b.blocked) {
		t->rcu_read_unlock_special.b.exp_need_qs = true;
708
		return;
709
	}
710

711 712 713 714 715 716 717 718 719 720
	/*
	 * We are either exiting an RCU read-side critical section (negative
	 * values of t->rcu_read_lock_nesting) or are not in one at all
	 * (zero value of t->rcu_read_lock_nesting).  Or we are in an RCU
	 * read-side critical section that blocked before this expedited
	 * grace period started.  Either way, we can immediately report
	 * the quiescent state.
	 */
	rdp = this_cpu_ptr(rsp->rda);
	rcu_report_exp_rdp(rsp, rdp, true);
721 722
}

723 724 725 726 727 728 729 730 731 732 733
/**
 * 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.
734 735 736
 */
void synchronize_rcu_expedited(void)
{
737
	struct rcu_node *rnp;
738
	struct rcu_node *rnp_unlock;
739
	struct rcu_state *rsp = rcu_state_p;
740
	unsigned long s;
741

742
	s = rcu_exp_gp_seq_snap(rsp);
743

744 745 746
	rnp_unlock = exp_funnel_lock(rsp, s);
	if (rnp_unlock == NULL)
		return;  /* Someone else did our work for us. */
747

748
	rcu_exp_gp_seq_start(rsp);
749

750
	/* Initialize the rcu_node tree in preparation for the wait. */
751
	sync_rcu_exp_select_cpus(rsp, sync_rcu_exp_handler);
752

753
	/* Wait for snapshotted ->blkd_tasks lists to drain. */
754
	rnp = rcu_get_root(rsp);
755
	synchronize_sched_expedited_wait(rsp);
756 757

	/* Clean up and exit. */
758
	rcu_exp_gp_seq_end(rsp);
759
	mutex_unlock(&rnp_unlock->exp_funnel_mutex);
760 761 762
}
EXPORT_SYMBOL_GPL(synchronize_rcu_expedited);

763 764
/**
 * rcu_barrier - Wait until all in-flight call_rcu() callbacks complete.
765 766 767 768 769
 *
 * 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.
770 771 772
 */
void rcu_barrier(void)
{
773
	_rcu_barrier(rcu_state_p);
774 775 776
}
EXPORT_SYMBOL_GPL(rcu_barrier);

777
/*
P
Paul E. McKenney 已提交
778
 * Initialize preemptible RCU's state structures.
779 780 781
 */
static void __init __rcu_init_preempt(void)
{
782
	rcu_init_one(rcu_state_p, rcu_data_p);
783 784
}

785 786 787 788 789 790 791 792 793 794 795 796 797 798
/*
 * 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();
799
	t->rcu_read_unlock_special.b.blocked = true;
800 801 802
	__rcu_read_unlock();
}

803
#else /* #ifdef CONFIG_PREEMPT_RCU */
804

805
static struct rcu_state *const rcu_state_p = &rcu_sched_state;
806
static struct rcu_data __percpu *const rcu_data_p = &rcu_sched_data;
807

808 809 810
/*
 * Tell them what RCU they are running.
 */
811
static void __init rcu_bootup_announce(void)
812
{
813
	pr_info("Hierarchical RCU implementation.\n");
814
	rcu_bootup_announce_oddness();
815 816
}

817 818 819 820
/*
 * Because preemptible RCU does not exist, we never have to check for
 * CPUs being in quiescent states.
 */
821
static void rcu_preempt_note_context_switch(void)
822 823 824
{
}

825
/*
P
Paul E. McKenney 已提交
826
 * Because preemptible RCU does not exist, there are never any preempted
827 828
 * RCU readers.
 */
829
static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp)
830 831 832 833
{
	return 0;
}

834 835 836 837
/*
 * Because there is no preemptible RCU, there can be no readers blocked.
 */
static bool rcu_preempt_has_tasks(struct rcu_node *rnp)
838
{
839
	return false;
840 841
}

842
/*
P
Paul E. McKenney 已提交
843
 * Because preemptible RCU does not exist, we never have to check for
844 845 846 847 848 849
 * tasks blocked within RCU read-side critical sections.
 */
static void rcu_print_detail_task_stall(struct rcu_state *rsp)
{
}

850
/*
P
Paul E. McKenney 已提交
851
 * Because preemptible RCU does not exist, we never have to check for
852 853
 * tasks blocked within RCU read-side critical sections.
 */
854
static int rcu_print_task_stall(struct rcu_node *rnp)
855
{
856
	return 0;
857 858
}

859 860 861 862 863 864 865 866 867 868
/*
 * Because preemptible RCU does not exist, we never have to check for
 * tasks blocked within RCU read-side critical sections that are
 * blocking the current expedited grace period.
 */
static int rcu_print_task_exp_stall(struct rcu_node *rnp)
{
	return 0;
}

869
/*
P
Paul E. McKenney 已提交
870
 * Because there is no preemptible RCU, there can be no readers blocked,
871 872
 * so there is no need to check for blocked tasks.  So check only for
 * bogus qsmask values.
873 874 875
 */
static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp)
{
876
	WARN_ON_ONCE(rnp->qsmask);
877 878
}

879
/*
P
Paul E. McKenney 已提交
880
 * Because preemptible RCU does not exist, it never has any callbacks
881 882
 * to check.
 */
883
static void rcu_preempt_check_callbacks(void)
884 885 886
{
}

887 888
/*
 * Wait for an rcu-preempt grace period, but make it happen quickly.
P
Paul E. McKenney 已提交
889
 * But because preemptible RCU does not exist, map to rcu-sched.
890 891 892 893 894 895 896
 */
void synchronize_rcu_expedited(void)
{
	synchronize_sched_expedited();
}
EXPORT_SYMBOL_GPL(synchronize_rcu_expedited);

897
/*
P
Paul E. McKenney 已提交
898
 * Because preemptible RCU does not exist, rcu_barrier() is just
899 900 901 902 903 904 905 906
 * another name for rcu_barrier_sched().
 */
void rcu_barrier(void)
{
	rcu_barrier_sched();
}
EXPORT_SYMBOL_GPL(rcu_barrier);

907
/*
P
Paul E. McKenney 已提交
908
 * Because preemptible RCU does not exist, it need not be initialized.
909 910 911 912 913
 */
static void __init __rcu_init_preempt(void)
{
}

914 915 916 917 918 919 920 921
/*
 * Because preemptible RCU does not exist, tasks cannot possibly exit
 * while in preemptible RCU read-side critical sections.
 */
void exit_rcu(void)
{
}

922
#endif /* #else #ifdef CONFIG_PREEMPT_RCU */
923

924 925
#ifdef CONFIG_RCU_BOOST

926
#include "../locking/rtmutex_common.h"
927

928 929 930 931
#ifdef CONFIG_RCU_TRACE

static void rcu_initiate_boost_trace(struct rcu_node *rnp)
{
932
	if (!rcu_preempt_has_tasks(rnp))
933 934 935 936 937 938 939 940
		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 &&
941
		 ULONG_CMP_LT(jiffies, rnp->boost_time))
942 943 944 945 946 947 948 949 950 951 952 953 954
		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
955 956 957 958 959 960 961 962 963 964
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);
}

965 966 967 968 969 970 971 972 973 974 975 976 977 978
/*
 * 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;

979 980
	if (READ_ONCE(rnp->exp_tasks) == NULL &&
	    READ_ONCE(rnp->boost_tasks) == NULL)
981 982
		return 0;  /* Nothing left to boost. */

983
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999

	/*
	 * 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.
	 */
1000
	if (rnp->exp_tasks != NULL) {
1001
		tb = rnp->exp_tasks;
1002 1003
		rnp->n_exp_boosts++;
	} else {
1004
		tb = rnp->boost_tasks;
1005 1006 1007
		rnp->n_normal_boosts++;
	}
	rnp->n_tasks_boosted++;
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025

	/*
	 * 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);
1026
	rt_mutex_init_proxy_locked(&rnp->boost_mtx, t);
1027
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1028 1029 1030
	/* 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. */
1031

1032 1033
	return READ_ONCE(rnp->exp_tasks) != NULL ||
	       READ_ONCE(rnp->boost_tasks) != NULL;
1034 1035 1036
}

/*
1037
 * Priority-boosting kthread, one per leaf rcu_node.
1038 1039 1040 1041 1042 1043 1044
 */
static int rcu_boost_kthread(void *arg)
{
	struct rcu_node *rnp = (struct rcu_node *)arg;
	int spincnt = 0;
	int more2boost;

1045
	trace_rcu_utilization(TPS("Start boost kthread@init"));
1046
	for (;;) {
1047
		rnp->boost_kthread_status = RCU_KTHREAD_WAITING;
1048
		trace_rcu_utilization(TPS("End boost kthread@rcu_wait"));
1049
		rcu_wait(rnp->boost_tasks || rnp->exp_tasks);
1050
		trace_rcu_utilization(TPS("Start boost kthread@rcu_wait"));
1051
		rnp->boost_kthread_status = RCU_KTHREAD_RUNNING;
1052 1053 1054 1055 1056 1057
		more2boost = rcu_boost(rnp);
		if (more2boost)
			spincnt++;
		else
			spincnt = 0;
		if (spincnt > 10) {
T
Thomas Gleixner 已提交
1058
			rnp->boost_kthread_status = RCU_KTHREAD_YIELDING;
1059
			trace_rcu_utilization(TPS("End boost kthread@rcu_yield"));
T
Thomas Gleixner 已提交
1060
			schedule_timeout_interruptible(2);
1061
			trace_rcu_utilization(TPS("Start boost kthread@rcu_yield"));
1062 1063 1064
			spincnt = 0;
		}
	}
1065
	/* NOTREACHED */
1066
	trace_rcu_utilization(TPS("End boost kthread@notreached"));
1067 1068 1069 1070 1071 1072 1073 1074 1075
	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.
 *
1076 1077 1078
 * 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.
1079
 */
1080
static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags)
1081
	__releases(rnp->lock)
1082 1083 1084
{
	struct task_struct *t;

1085 1086
	if (!rcu_preempt_blocked_readers_cgp(rnp) && rnp->exp_tasks == NULL) {
		rnp->n_balk_exp_gp_tasks++;
1087
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1088
		return;
1089
	}
1090 1091 1092 1093 1094 1095 1096
	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;
1097
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1098
		t = rnp->boost_kthread_task;
T
Thomas Gleixner 已提交
1099 1100
		if (t)
			rcu_wake_cond(t, rnp->boost_kthread_status);
1101
	} else {
1102
		rcu_initiate_boost_trace(rnp);
1103 1104
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
	}
1105 1106
}

1107 1108 1109 1110 1111 1112 1113 1114 1115
/*
 * 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);
1116
	if (__this_cpu_read(rcu_cpu_kthread_task) != NULL &&
T
Thomas Gleixner 已提交
1117 1118 1119 1120
	    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));
	}
1121 1122 1123
	local_irq_restore(flags);
}

1124 1125 1126 1127 1128 1129
/*
 * Is the current CPU running the RCU-callbacks kthread?
 * Caller must have preemption disabled.
 */
static bool rcu_is_callbacks_kthread(void)
{
1130
	return __this_cpu_read(rcu_cpu_kthread_task) == current;
1131 1132
}

1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
#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.
 */
1148
static int rcu_spawn_one_boost_kthread(struct rcu_state *rsp,
1149
				       struct rcu_node *rnp)
1150
{
T
Thomas Gleixner 已提交
1151
	int rnp_index = rnp - &rsp->node[0];
1152 1153 1154 1155
	unsigned long flags;
	struct sched_param sp;
	struct task_struct *t;

1156
	if (rcu_state_p != rsp)
1157
		return 0;
T
Thomas Gleixner 已提交
1158

1159
	if (!rcu_scheduler_fully_active || rcu_rnp_online_cpus(rnp) == 0)
T
Thomas Gleixner 已提交
1160 1161
		return 0;

1162
	rsp->boost = 1;
1163 1164 1165
	if (rnp->boost_kthread_task != NULL)
		return 0;
	t = kthread_create(rcu_boost_kthread, (void *)rnp,
1166
			   "rcub/%d", rnp_index);
1167 1168
	if (IS_ERR(t))
		return PTR_ERR(t);
1169
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
1170 1171
	rnp->boost_kthread_task = t;
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1172
	sp.sched_priority = kthread_prio;
1173
	sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
1174
	wake_up_process(t); /* get to TASK_INTERRUPTIBLE quickly. */
1175 1176 1177
	return 0;
}

1178 1179
static void rcu_kthread_do_work(void)
{
1180 1181
	rcu_do_batch(&rcu_sched_state, this_cpu_ptr(&rcu_sched_data));
	rcu_do_batch(&rcu_bh_state, this_cpu_ptr(&rcu_bh_data));
1182 1183 1184
	rcu_preempt_do_callbacks();
}

1185
static void rcu_cpu_kthread_setup(unsigned int cpu)
1186 1187 1188
{
	struct sched_param sp;

1189
	sp.sched_priority = kthread_prio;
1190
	sched_setscheduler_nocheck(current, SCHED_FIFO, &sp);
1191 1192
}

1193
static void rcu_cpu_kthread_park(unsigned int cpu)
1194
{
1195
	per_cpu(rcu_cpu_kthread_status, cpu) = RCU_KTHREAD_OFFCPU;
1196 1197
}

1198
static int rcu_cpu_kthread_should_run(unsigned int cpu)
1199
{
1200
	return __this_cpu_read(rcu_cpu_has_work);
1201 1202 1203 1204
}

/*
 * Per-CPU kernel thread that invokes RCU callbacks.  This replaces the
1205 1206
 * RCU softirq used in flavors and configurations of RCU that do not
 * support RCU priority boosting.
1207
 */
1208
static void rcu_cpu_kthread(unsigned int cpu)
1209
{
1210 1211
	unsigned int *statusp = this_cpu_ptr(&rcu_cpu_kthread_status);
	char work, *workp = this_cpu_ptr(&rcu_cpu_has_work);
1212
	int spincnt;
1213

1214
	for (spincnt = 0; spincnt < 10; spincnt++) {
1215
		trace_rcu_utilization(TPS("Start CPU kthread@rcu_wait"));
1216 1217
		local_bh_disable();
		*statusp = RCU_KTHREAD_RUNNING;
1218 1219
		this_cpu_inc(rcu_cpu_kthread_loops);
		local_irq_disable();
1220 1221
		work = *workp;
		*workp = 0;
1222
		local_irq_enable();
1223 1224 1225
		if (work)
			rcu_kthread_do_work();
		local_bh_enable();
1226
		if (*workp == 0) {
1227
			trace_rcu_utilization(TPS("End CPU kthread@rcu_wait"));
1228 1229
			*statusp = RCU_KTHREAD_WAITING;
			return;
1230 1231
		}
	}
1232
	*statusp = RCU_KTHREAD_YIELDING;
1233
	trace_rcu_utilization(TPS("Start CPU kthread@rcu_yield"));
1234
	schedule_timeout_interruptible(2);
1235
	trace_rcu_utilization(TPS("End CPU kthread@rcu_yield"));
1236
	*statusp = RCU_KTHREAD_WAITING;
1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247
}

/*
 * 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
1248
static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu)
1249
{
T
Thomas Gleixner 已提交
1250
	struct task_struct *t = rnp->boost_kthread_task;
1251
	unsigned long mask = rcu_rnp_online_cpus(rnp);
1252 1253 1254
	cpumask_var_t cm;
	int cpu;

T
Thomas Gleixner 已提交
1255
	if (!t)
1256
		return;
T
Thomas Gleixner 已提交
1257
	if (!zalloc_cpumask_var(&cm, GFP_KERNEL))
1258 1259 1260 1261
		return;
	for (cpu = rnp->grplo; cpu <= rnp->grphi; cpu++, mask >>= 1)
		if ((mask & 0x1) && cpu != outgoingcpu)
			cpumask_set_cpu(cpu, cm);
1262
	if (cpumask_weight(cm) == 0)
1263
		cpumask_setall(cm);
T
Thomas Gleixner 已提交
1264
	set_cpus_allowed_ptr(t, cm);
1265 1266 1267
	free_cpumask_var(cm);
}

1268 1269 1270 1271 1272 1273 1274 1275
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,
};
1276 1277

/*
1278
 * Spawn boost kthreads -- called as soon as the scheduler is running.
1279
 */
1280
static void __init rcu_spawn_boost_kthreads(void)
1281 1282
{
	struct rcu_node *rnp;
T
Thomas Gleixner 已提交
1283
	int cpu;
1284

1285
	for_each_possible_cpu(cpu)
1286
		per_cpu(rcu_cpu_has_work, cpu) = 0;
1287
	BUG_ON(smpboot_register_percpu_thread(&rcu_cpu_thread_spec));
1288 1289
	rcu_for_each_leaf_node(rcu_state_p, rnp)
		(void)rcu_spawn_one_boost_kthread(rcu_state_p, rnp);
1290 1291
}

1292
static void rcu_prepare_kthreads(int cpu)
1293
{
1294
	struct rcu_data *rdp = per_cpu_ptr(rcu_state_p->rda, cpu);
1295 1296 1297
	struct rcu_node *rnp = rdp->mynode;

	/* Fire up the incoming CPU's kthread and leaf rcu_node kthread. */
1298
	if (rcu_scheduler_fully_active)
1299
		(void)rcu_spawn_one_boost_kthread(rcu_state_p, rnp);
1300 1301
}

1302 1303
#else /* #ifdef CONFIG_RCU_BOOST */

1304
static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags)
1305
	__releases(rnp->lock)
1306
{
1307
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1308 1309
}

1310
static void invoke_rcu_callbacks_kthread(void)
1311
{
1312
	WARN_ON_ONCE(1);
1313 1314
}

1315 1316 1317 1318 1319
static bool rcu_is_callbacks_kthread(void)
{
	return false;
}

1320 1321 1322 1323
static void rcu_preempt_boost_start_gp(struct rcu_node *rnp)
{
}

T
Thomas Gleixner 已提交
1324
static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu)
1325 1326 1327
{
}

1328
static void __init rcu_spawn_boost_kthreads(void)
1329 1330 1331
{
}

1332
static void rcu_prepare_kthreads(int cpu)
1333 1334 1335
{
}

1336 1337
#endif /* #else #ifdef CONFIG_RCU_BOOST */

1338 1339 1340 1341 1342 1343 1344 1345
#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.
 *
1346 1347
 * Because we not have RCU_FAST_NO_HZ, just check whether this CPU needs
 * any flavor of RCU.
1348
 */
1349
int rcu_needs_cpu(u64 basemono, u64 *nextevt)
1350
{
1351
	*nextevt = KTIME_MAX;
1352 1353
	return IS_ENABLED(CONFIG_RCU_NOCB_CPU_ALL)
	       ? 0 : rcu_cpu_has_callbacks(NULL);
1354 1355 1356 1357 1358 1359
}

/*
 * Because we do not have RCU_FAST_NO_HZ, don't bother cleaning up
 * after it.
 */
1360
static void rcu_cleanup_after_idle(void)
1361 1362 1363
{
}

1364
/*
1365
 * Do the idle-entry grace-period work, which, because CONFIG_RCU_FAST_NO_HZ=n,
1366 1367
 * is nothing.
 */
1368
static void rcu_prepare_for_idle(void)
1369 1370 1371
{
}

1372 1373 1374 1375 1376 1377 1378 1379
/*
 * 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)
{
}

1380 1381
#else /* #if !defined(CONFIG_RCU_FAST_NO_HZ) */

1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396
/*
 * 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!
1397 1398 1399
 * 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.
1400 1401 1402 1403 1404
 *
 * 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.
 */
1405
#define RCU_IDLE_GP_DELAY 4		/* Roughly one grace period. */
1406
#define RCU_IDLE_LAZY_GP_DELAY (6 * HZ)	/* Roughly six seconds. */
1407

1408 1409 1410 1411
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);
1412 1413

/*
1414 1415 1416
 * 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.
1417
 */
1418
static bool __maybe_unused rcu_try_advance_all_cbs(void)
1419
{
1420 1421
	bool cbs_ready = false;
	struct rcu_data *rdp;
1422
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1423 1424
	struct rcu_node *rnp;
	struct rcu_state *rsp;
1425

1426 1427
	/* Exit early if we advanced recently. */
	if (jiffies == rdtp->last_advance_all)
1428
		return false;
1429 1430
	rdtp->last_advance_all = jiffies;

1431 1432 1433
	for_each_rcu_flavor(rsp) {
		rdp = this_cpu_ptr(rsp->rda);
		rnp = rdp->mynode;
1434

1435 1436 1437 1438 1439
		/*
		 * Don't bother checking unless a grace period has
		 * completed since we last checked and there are
		 * callbacks not yet ready to invoke.
		 */
1440
		if ((rdp->completed != rnp->completed ||
1441
		     unlikely(READ_ONCE(rdp->gpwrap))) &&
1442
		    rdp->nxttail[RCU_DONE_TAIL] != rdp->nxttail[RCU_NEXT_TAIL])
1443
			note_gp_changes(rsp, rdp);
1444

1445 1446 1447 1448
		if (cpu_has_callbacks_ready_to_invoke(rdp))
			cbs_ready = true;
	}
	return cbs_ready;
1449 1450
}

1451
/*
1452 1453 1454 1455
 * 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.
1456
 *
1457
 * The caller must have disabled interrupts.
1458
 */
1459
int rcu_needs_cpu(u64 basemono, u64 *nextevt)
1460
{
1461
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1462
	unsigned long dj;
1463

1464
	if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_ALL)) {
1465
		*nextevt = KTIME_MAX;
1466 1467 1468
		return 0;
	}

1469 1470 1471
	/* Snapshot to detect later posting of non-lazy callback. */
	rdtp->nonlazy_posted_snap = rdtp->nonlazy_posted;

1472
	/* If no callbacks, RCU doesn't need the CPU. */
1473
	if (!rcu_cpu_has_callbacks(&rdtp->all_lazy)) {
1474
		*nextevt = KTIME_MAX;
1475 1476
		return 0;
	}
1477 1478 1479 1480 1481

	/* Attempt to advance callbacks. */
	if (rcu_try_advance_all_cbs()) {
		/* Some ready to invoke, so initiate later invocation. */
		invoke_rcu_core();
1482 1483
		return 1;
	}
1484 1485 1486
	rdtp->last_accelerate = jiffies;

	/* Request timer delay depending on laziness, and round. */
1487
	if (!rdtp->all_lazy) {
1488
		dj = round_up(rcu_idle_gp_delay + jiffies,
1489
			       rcu_idle_gp_delay) - jiffies;
1490
	} else {
1491
		dj = round_jiffies(rcu_idle_lazy_gp_delay + jiffies) - jiffies;
1492
	}
1493
	*nextevt = basemono + dj * TICK_NSEC;
1494 1495 1496
	return 0;
}

1497
/*
1498 1499 1500 1501 1502 1503
 * 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.
1504 1505
 *
 * The caller must have disabled interrupts.
1506
 */
1507
static void rcu_prepare_for_idle(void)
1508
{
1509
	bool needwake;
1510
	struct rcu_data *rdp;
1511
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1512 1513
	struct rcu_node *rnp;
	struct rcu_state *rsp;
1514 1515
	int tne;

1516 1517 1518
	if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_ALL))
		return;

1519
	/* Handle nohz enablement switches conservatively. */
1520
	tne = READ_ONCE(tick_nohz_active);
1521
	if (tne != rdtp->tick_nohz_enabled_snap) {
1522
		if (rcu_cpu_has_callbacks(NULL))
1523 1524 1525 1526 1527 1528
			invoke_rcu_core(); /* force nohz to see update. */
		rdtp->tick_nohz_enabled_snap = tne;
		return;
	}
	if (!tne)
		return;
1529

1530
	/* If this is a no-CBs CPU, no callbacks, just return. */
1531
	if (rcu_is_nocb_cpu(smp_processor_id()))
1532 1533
		return;

1534
	/*
1535 1536 1537
	 * 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.
1538
	 */
1539 1540
	if (rdtp->all_lazy &&
	    rdtp->nonlazy_posted != rdtp->nonlazy_posted_snap) {
1541 1542
		rdtp->all_lazy = false;
		rdtp->nonlazy_posted_snap = rdtp->nonlazy_posted;
1543
		invoke_rcu_core();
1544 1545 1546
		return;
	}

1547
	/*
1548 1549
	 * If we have not yet accelerated this jiffy, accelerate all
	 * callbacks on this CPU.
1550
	 */
1551
	if (rdtp->last_accelerate == jiffies)
1552
		return;
1553 1554
	rdtp->last_accelerate = jiffies;
	for_each_rcu_flavor(rsp) {
1555
		rdp = this_cpu_ptr(rsp->rda);
1556 1557 1558
		if (!*rdp->nxttail[RCU_DONE_TAIL])
			continue;
		rnp = rdp->mynode;
1559
		raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */
1560
		needwake = rcu_accelerate_cbs(rsp, rnp, rdp);
1561
		raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
1562 1563
		if (needwake)
			rcu_gp_kthread_wake(rsp);
1564
	}
1565
}
1566

1567 1568 1569 1570 1571
/*
 * 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.
 */
1572
static void rcu_cleanup_after_idle(void)
1573
{
1574 1575
	if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_ALL) ||
	    rcu_is_nocb_cpu(smp_processor_id()))
1576
		return;
1577 1578
	if (rcu_try_advance_all_cbs())
		invoke_rcu_core();
1579 1580
}

1581
/*
1582 1583 1584 1585 1586 1587
 * 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().
1588 1589 1590
 */
static void rcu_idle_count_callbacks_posted(void)
{
1591
	__this_cpu_add(rcu_dynticks.nonlazy_posted, 1);
1592 1593
}

1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
/*
 * 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) {
1623
		rdp = raw_cpu_ptr(rsp->rda);
1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644
		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);
1645
	smp_mb(); /* Ensure callback reuse happens after callback invocation. */
1646 1647 1648 1649 1650 1651 1652 1653 1654

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

	for_each_online_cpu(cpu) {
		smp_call_function_single(cpu, rcu_oom_notify_cpu, NULL, 1);
1655
		cond_resched_rcu_qs();
1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674
	}

	/* 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);

1675
#endif /* #else #if !defined(CONFIG_RCU_FAST_NO_HZ) */
1676 1677 1678 1679 1680

#ifdef CONFIG_RCU_FAST_NO_HZ

static void print_cpu_stall_fast_no_hz(char *cp, int cpu)
{
1681
	struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);
1682
	unsigned long nlpd = rdtp->nonlazy_posted - rdtp->nonlazy_posted_snap;
1683

1684 1685 1686 1687 1688
	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');
1689 1690 1691 1692 1693 1694
}

#else /* #ifdef CONFIG_RCU_FAST_NO_HZ */

static void print_cpu_stall_fast_no_hz(char *cp, int cpu)
{
1695
	*cp = '\0';
1696 1697 1698 1699 1700 1701 1702
}

#endif /* #else #ifdef CONFIG_RCU_FAST_NO_HZ */

/* Initiate the stall-info list. */
static void print_cpu_stall_info_begin(void)
{
1703
	pr_cont("\n");
1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733
}

/*
 * 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);
1734 1735 1736 1737 1738 1739
	pr_err("\t%d-%c%c%c: (%lu %s) idle=%03x/%llx/%d softirq=%u/%u fqs=%ld %s\n",
	       cpu,
	       "O."[!!cpu_online(cpu)],
	       "o."[!!(rdp->grpmask & rdp->mynode->qsmaskinit)],
	       "N."[!!(rdp->grpmask & rdp->mynode->qsmaskinitnext)],
	       ticks_value, ticks_title,
1740 1741
	       atomic_read(&rdtp->dynticks) & 0xfff,
	       rdtp->dynticks_nesting, rdtp->dynticks_nmi_nesting,
1742
	       rdp->softirq_snap, kstat_softirqs_cpu(RCU_SOFTIRQ, cpu),
1743
	       READ_ONCE(rsp->n_force_qs) - rsp->n_force_qs_gpstart,
1744 1745 1746 1747 1748 1749
	       fast_no_hz);
}

/* Terminate the stall-info list. */
static void print_cpu_stall_info_end(void)
{
1750
	pr_err("\t");
1751 1752 1753 1754 1755 1756
}

/* Zero ->ticks_this_gp for all flavors of RCU. */
static void zero_cpu_stall_ticks(struct rcu_data *rdp)
{
	rdp->ticks_this_gp = 0;
1757
	rdp->softirq_snap = kstat_softirqs_cpu(RCU_SOFTIRQ, smp_processor_id());
1758 1759 1760 1761 1762
}

/* Increment ->ticks_this_gp for all flavors of RCU. */
static void increment_cpu_stall_ticks(void)
{
1763 1764 1765
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
1766
		raw_cpu_inc(rsp->rda->ticks_this_gp);
1767 1768
}

P
Paul E. McKenney 已提交
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
#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);

1802 1803 1804 1805 1806 1807 1808
static int __init parse_rcu_nocb_poll(char *arg)
{
	rcu_nocb_poll = 1;
	return 0;
}
early_param("rcu_nocb_poll", parse_rcu_nocb_poll);

1809
/*
1810 1811
 * Wake up any no-CBs CPUs' kthreads that were waiting on the just-ended
 * grace period.
1812
 */
1813
static void rcu_nocb_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
1814
{
1815
	wake_up_all(&rnp->nocb_gp_wq[rnp->completed & 0x1]);
1816 1817 1818
}

/*
1819
 * Set the root rcu_node structure's ->need_future_gp field
1820 1821 1822 1823 1824
 * 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.
1825
 */
1826 1827
static void rcu_nocb_gp_set(struct rcu_node *rnp, int nrq)
{
1828
	rnp->need_future_gp[(rnp->completed + 1) & 0x1] += nrq;
1829 1830 1831
}

static void rcu_init_one_nocb(struct rcu_node *rnp)
1832
{
1833 1834
	init_waitqueue_head(&rnp->nocb_gp_wq[0]);
	init_waitqueue_head(&rnp->nocb_gp_wq[1]);
1835 1836
}

1837
#ifndef CONFIG_RCU_NOCB_CPU_ALL
L
Liu Ping Fan 已提交
1838
/* Is the specified CPU a no-CBs CPU? */
1839
bool rcu_is_nocb_cpu(int cpu)
P
Paul E. McKenney 已提交
1840 1841 1842 1843 1844
{
	if (have_rcu_nocb_mask)
		return cpumask_test_cpu(cpu, rcu_nocb_mask);
	return false;
}
1845
#endif /* #ifndef CONFIG_RCU_NOCB_CPU_ALL */
P
Paul E. McKenney 已提交
1846

1847 1848 1849 1850 1851 1852 1853
/*
 * 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;

1854
	if (!READ_ONCE(rdp_leader->nocb_kthread))
1855
		return;
1856
	if (READ_ONCE(rdp_leader->nocb_leader_sleep) || force) {
1857
		/* Prior smp_mb__after_atomic() orders against prior enqueue. */
1858
		WRITE_ONCE(rdp_leader->nocb_leader_sleep, false);
1859 1860 1861 1862
		wake_up(&rdp_leader->nocb_wq);
	}
}

1863 1864 1865 1866 1867 1868 1869
/*
 * 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);
1870 1871
	unsigned long ret;
#ifdef CONFIG_PROVE_RCU
1872
	struct rcu_head *rhp;
1873
#endif /* #ifdef CONFIG_PROVE_RCU */
1874

1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887
	/*
	 * 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);
1888

1889
#ifdef CONFIG_PROVE_RCU
1890
	rhp = READ_ONCE(rdp->nocb_head);
1891
	if (!rhp)
1892
		rhp = READ_ONCE(rdp->nocb_gp_head);
1893
	if (!rhp)
1894
		rhp = READ_ONCE(rdp->nocb_follower_head);
1895 1896

	/* Having no rcuo kthread but CBs after scheduler starts is bad! */
1897
	if (!READ_ONCE(rdp->nocb_kthread) && rhp &&
1898
	    rcu_scheduler_fully_active) {
1899 1900 1901 1902 1903
		/* 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);
	}
1904
#endif /* #ifdef CONFIG_PROVE_RCU */
1905

1906
	return !!ret;
1907 1908
}

P
Paul E. McKenney 已提交
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
/*
 * 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,
1920 1921
				    int rhcount, int rhcount_lazy,
				    unsigned long flags)
P
Paul E. McKenney 已提交
1922 1923 1924 1925 1926 1927
{
	int len;
	struct rcu_head **old_rhpp;
	struct task_struct *t;

	/* Enqueue the callback on the nocb list and update counts. */
1928 1929
	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
1930
	old_rhpp = xchg(&rdp->nocb_tail, rhtp);
1931
	WRITE_ONCE(*old_rhpp, rhp);
P
Paul E. McKenney 已提交
1932
	atomic_long_add(rhcount_lazy, &rdp->nocb_q_count_lazy);
1933
	smp_mb__after_atomic(); /* Store *old_rhpp before _wake test. */
P
Paul E. McKenney 已提交
1934 1935

	/* If we are not being polled and there is a kthread, awaken it ... */
1936
	t = READ_ONCE(rdp->nocb_kthread);
1937
	if (rcu_nocb_poll || !t) {
1938 1939
		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
				    TPS("WakeNotPoll"));
P
Paul E. McKenney 已提交
1940
		return;
1941
	}
P
Paul E. McKenney 已提交
1942 1943
	len = atomic_long_read(&rdp->nocb_q_count);
	if (old_rhpp == &rdp->nocb_head) {
1944
		if (!irqs_disabled_flags(flags)) {
1945 1946
			/* ... if queue was empty ... */
			wake_nocb_leader(rdp, false);
1947 1948 1949
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    TPS("WakeEmpty"));
		} else {
1950
			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE;
1951 1952 1953
			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
					    TPS("WakeEmptyIsDeferred"));
		}
P
Paul E. McKenney 已提交
1954 1955
		rdp->qlen_last_fqs_check = 0;
	} else if (len > rdp->qlen_last_fqs_check + qhimark) {
1956
		/* ... or if many callbacks queued. */
1957 1958 1959 1960 1961 1962 1963 1964 1965
		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
1966
		rdp->qlen_last_fqs_check = LONG_MAX / 2;
1967 1968
	} else {
		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeNot"));
P
Paul E. McKenney 已提交
1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
	}
	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,
1983
			    bool lazy, unsigned long flags)
P
Paul E. McKenney 已提交
1984 1985
{

1986
	if (!rcu_is_nocb_cpu(rdp->cpu))
1987
		return false;
1988
	__call_rcu_nocb_enqueue(rdp, rhp, &rhp->next, 1, lazy, flags);
1989 1990 1991
	if (__is_kfree_rcu_offset((unsigned long)rhp->func))
		trace_rcu_kfree_callback(rdp->rsp->name, rhp,
					 (unsigned long)rhp->func,
1992 1993
					 -atomic_long_read(&rdp->nocb_q_count_lazy),
					 -atomic_long_read(&rdp->nocb_q_count));
1994 1995
	else
		trace_rcu_callback(rdp->rsp->name, rhp,
1996 1997
				   -atomic_long_read(&rdp->nocb_q_count_lazy),
				   -atomic_long_read(&rdp->nocb_q_count));
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

	/*
	 * 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();

2009
	return true;
P
Paul E. McKenney 已提交
2010 2011 2012 2013 2014 2015 2016
}

/*
 * 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,
2017 2018
						     struct rcu_data *rdp,
						     unsigned long flags)
P
Paul E. McKenney 已提交
2019 2020 2021 2022 2023
{
	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. */
2024
	if (!rcu_is_nocb_cpu(smp_processor_id()))
2025
		return false;
P
Paul E. McKenney 已提交
2026 2027 2028 2029 2030 2031
	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,
2032
					rsp->orphan_donetail, ql, qll, flags);
P
Paul E. McKenney 已提交
2033 2034 2035 2036 2037 2038
		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,
2039
					rsp->orphan_nxttail, ql, qll, flags);
P
Paul E. McKenney 已提交
2040 2041 2042 2043
		ql = qll = 0;
		rsp->orphan_nxtlist = NULL;
		rsp->orphan_nxttail = &rsp->orphan_nxtlist;
	}
2044
	return true;
P
Paul E. McKenney 已提交
2045 2046 2047
}

/*
2048 2049
 * 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
2050
 */
2051
static void rcu_nocb_wait_gp(struct rcu_data *rdp)
P
Paul E. McKenney 已提交
2052
{
2053
	unsigned long c;
2054
	bool d;
2055
	unsigned long flags;
2056
	bool needwake;
2057 2058
	struct rcu_node *rnp = rdp->mynode;

2059
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
2060
	needwake = rcu_start_future_gp(rnp, rdp, &c);
2061
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
2062 2063
	if (needwake)
		rcu_gp_kthread_wake(rdp->rsp);
P
Paul E. McKenney 已提交
2064 2065

	/*
2066 2067
	 * Wait for the grace period.  Do so interruptibly to avoid messing
	 * up the load average.
P
Paul E. McKenney 已提交
2068
	 */
2069
	trace_rcu_future_gp(rnp, rdp, c, TPS("StartWait"));
2070
	for (;;) {
2071 2072
		wait_event_interruptible(
			rnp->nocb_gp_wq[c & 0x1],
2073
			(d = ULONG_CMP_GE(READ_ONCE(rnp->completed), c)));
2074
		if (likely(d))
2075
			break;
2076
		WARN_ON(signal_pending(current));
2077
		trace_rcu_future_gp(rnp, rdp, c, TPS("ResumeWait"));
2078
	}
2079
	trace_rcu_future_gp(rnp, rdp, c, TPS("EndWait"));
2080
	smp_mb(); /* Ensure that CB invocation happens after GP end. */
P
Paul E. McKenney 已提交
2081 2082
}

2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
/*
 * 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,
2100
				!READ_ONCE(my_rdp->nocb_leader_sleep));
2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113
		/* 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) {
2114
		rdp->nocb_gp_head = READ_ONCE(rdp->nocb_head);
2115 2116 2117 2118
		if (!rdp->nocb_gp_head)
			continue;  /* No CBs here, try next follower. */

		/* Move callbacks to wait-for-GP list, which is empty. */
2119
		WRITE_ONCE(rdp->nocb_head, NULL);
2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131
		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");
2132
		WARN_ON(signal_pending(current));
2133 2134 2135
		schedule_timeout_interruptible(1);

		/* Rescan in case we were a victim of memory ordering. */
2136 2137
		my_rdp->nocb_leader_sleep = true;
		smp_mb();  /* Ensure _sleep true before scan. */
2138
		for (rdp = my_rdp; rdp; rdp = rdp->nocb_next_follower)
2139
			if (READ_ONCE(rdp->nocb_head)) {
2140
				/* Found CB, so short-circuit next wait. */
2141
				my_rdp->nocb_leader_sleep = false;
2142 2143 2144 2145 2146 2147 2148 2149 2150
				break;
			}
		goto wait_again;
	}

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

	/*
2151 2152
	 * We left ->nocb_leader_sleep unset to reduce cache thrashing.
	 * We set it now, but recheck for new callbacks while
2153 2154
	 * traversing our follower list.
	 */
2155 2156
	my_rdp->nocb_leader_sleep = true;
	smp_mb(); /* Ensure _sleep true before scan of ->nocb_head. */
2157 2158 2159

	/* Each pass through the following loop wakes a follower, if needed. */
	for (rdp = my_rdp; rdp; rdp = rdp->nocb_next_follower) {
2160
		if (READ_ONCE(rdp->nocb_head))
2161
			my_rdp->nocb_leader_sleep = false;/* No need to sleep.*/
2162 2163 2164 2165 2166 2167
		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;
2168
		smp_mb__after_atomic(); /* Store *tail before wakeup. */
2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195
		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,
2196
						 READ_ONCE(rdp->nocb_follower_head));
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208
		} 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");
2209
		WARN_ON(signal_pending(current));
2210 2211 2212 2213
		schedule_timeout_interruptible(1);
	}
}

P
Paul E. McKenney 已提交
2214 2215
/*
 * Per-rcu_data kthread, but only for no-CBs CPUs.  Each kthread invokes
2216 2217 2218
 * 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
2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
 */
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 (;;) {
2230 2231 2232 2233 2234 2235 2236
		/* 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. */
2237
		list = READ_ONCE(rdp->nocb_follower_head);
2238 2239
		BUG_ON(!list);
		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, "WokeNonEmpty");
2240
		WRITE_ONCE(rdp->nocb_follower_head, NULL);
2241
		tail = xchg(&rdp->nocb_follower_tail, &rdp->nocb_follower_head);
P
Paul E. McKenney 已提交
2242 2243

		/* Each pass through the following loop invokes a callback. */
2244 2245 2246
		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
2247 2248 2249 2250 2251
		c = cl = 0;
		while (list) {
			next = list->next;
			/* Wait for enqueuing to complete, if needed. */
			while (next == NULL && &list->next != tail) {
2252 2253
				trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
						    TPS("WaitQueue"));
P
Paul E. McKenney 已提交
2254
				schedule_timeout_interruptible(1);
2255 2256
				trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
						    TPS("WokeQueue"));
P
Paul E. McKenney 已提交
2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267
				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);
2268 2269 2270
		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);
2271
		rdp->n_nocbs_invoked += c;
P
Paul E. McKenney 已提交
2272 2273 2274 2275
	}
	return 0;
}

2276
/* Is a deferred wakeup of rcu_nocb_kthread() required? */
2277
static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
2278
{
2279
	return READ_ONCE(rdp->nocb_defer_wakeup);
2280 2281 2282 2283 2284
}

/* Do a deferred wakeup of rcu_nocb_kthread(). */
static void do_nocb_deferred_wakeup(struct rcu_data *rdp)
{
2285 2286
	int ndw;

2287 2288
	if (!rcu_nocb_need_deferred_wakeup(rdp))
		return;
2289 2290
	ndw = READ_ONCE(rdp->nocb_defer_wakeup);
	WRITE_ONCE(rdp->nocb_defer_wakeup, RCU_NOGP_WAKE_NOT);
2291 2292
	wake_nocb_leader(rdp, ndw == RCU_NOGP_WAKE_FORCE);
	trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("DeferredWake"));
2293 2294
}

2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310
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) {
2311 2312 2313 2314
		if (!zalloc_cpumask_var(&rcu_nocb_mask, GFP_KERNEL)) {
			pr_info("rcu_nocb_mask allocation failed, callback offloading disabled.\n");
			return;
		}
2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337
		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);
	}
2338 2339
	pr_info("\tOffload RCU callbacks from CPUs: %*pbl.\n",
		cpumask_pr_args(rcu_nocb_mask));
2340 2341 2342 2343
	if (rcu_nocb_poll)
		pr_info("\tPoll for callbacks from no-CBs CPUs.\n");

	for_each_rcu_flavor(rsp) {
2344 2345
		for_each_cpu(cpu, rcu_nocb_mask)
			init_nocb_callback_list(per_cpu_ptr(rsp->rda, cpu));
2346
		rcu_organize_nocb_kthreads(rsp);
2347
	}
2348 2349
}

P
Paul E. McKenney 已提交
2350 2351 2352 2353 2354
/* 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);
2355
	rdp->nocb_follower_tail = &rdp->nocb_follower_head;
P
Paul E. McKenney 已提交
2356 2357
}

2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
/*
 * 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;
2388 2389 2390 2391 2392 2393 2394
			if (rdp == rdp_spawn) {
				rdp = rdp->nocb_next_follower;
			} else {
				rdp_last = rdp;
				rdp = rdp->nocb_next_follower;
				rdp_last->nocb_next_follower = NULL;
			}
2395 2396 2397 2398 2399 2400 2401 2402
		} 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));
2403
	WRITE_ONCE(rdp_spawn->nocb_kthread, t);
2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432
}

/*
 * 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);
}

2433 2434 2435 2436 2437
/* 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);

/*
2438
 * Initialize leader-follower relationships for all no-CBs CPU.
2439
 */
2440
static void __init rcu_organize_nocb_kthreads(struct rcu_state *rsp)
P
Paul E. McKenney 已提交
2441 2442
{
	int cpu;
2443 2444
	int ls = rcu_nocb_leader_stride;
	int nl = 0;  /* Next leader. */
P
Paul E. McKenney 已提交
2445
	struct rcu_data *rdp;
2446 2447
	struct rcu_data *rdp_leader = NULL;  /* Suppress misguided gcc warn. */
	struct rcu_data *rdp_prev = NULL;
P
Paul E. McKenney 已提交
2448

2449
	if (!have_rcu_nocb_mask)
P
Paul E. McKenney 已提交
2450
		return;
2451 2452 2453 2454 2455 2456 2457 2458 2459
	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
2460 2461
	for_each_cpu(cpu, rcu_nocb_mask) {
		rdp = per_cpu_ptr(rsp->rda, cpu);
2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
		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
2473 2474 2475 2476
	}
}

/* Prevent __call_rcu() from enqueuing callbacks on no-CBs CPUs */
2477
static bool init_nocb_callback_list(struct rcu_data *rdp)
P
Paul E. McKenney 已提交
2478
{
2479
	if (!rcu_is_nocb_cpu(rdp->cpu))
2480
		return false;
2481

2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
	/* 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
2492
	rdp->nxttail[RCU_NEXT_TAIL] = NULL;
2493
	return true;
P
Paul E. McKenney 已提交
2494 2495
}

2496 2497
#else /* #ifdef CONFIG_RCU_NOCB_CPU */

2498 2499 2500 2501 2502 2503
static bool rcu_nocb_cpu_needs_barrier(struct rcu_state *rsp, int cpu)
{
	WARN_ON_ONCE(1); /* Should be dead code. */
	return false;
}

2504
static void rcu_nocb_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
P
Paul E. McKenney 已提交
2505 2506 2507
{
}

2508 2509 2510 2511 2512 2513 2514
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
2515 2516

static bool __call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *rhp,
2517
			    bool lazy, unsigned long flags)
P
Paul E. McKenney 已提交
2518
{
2519
	return false;
P
Paul E. McKenney 已提交
2520 2521 2522
}

static bool __maybe_unused rcu_nocb_adopt_orphan_cbs(struct rcu_state *rsp,
2523 2524
						     struct rcu_data *rdp,
						     unsigned long flags)
P
Paul E. McKenney 已提交
2525
{
2526
	return false;
P
Paul E. McKenney 已提交
2527 2528 2529 2530 2531 2532
}

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

2533
static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
2534 2535 2536 2537 2538 2539 2540 2541
{
	return false;
}

static void do_nocb_deferred_wakeup(struct rcu_data *rdp)
{
}

2542 2543 2544 2545 2546
static void rcu_spawn_all_nocb_kthreads(int cpu)
{
}

static void __init rcu_spawn_nocb_kthreads(void)
P
Paul E. McKenney 已提交
2547 2548 2549
{
}

2550
static bool init_nocb_callback_list(struct rcu_data *rdp)
P
Paul E. McKenney 已提交
2551
{
2552
	return false;
P
Paul E. McKenney 已提交
2553 2554 2555
}

#endif /* #else #ifdef CONFIG_RCU_NOCB_CPU */
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565

/*
 * 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.
 */
2566
static void __maybe_unused rcu_kick_nohz_cpu(int cpu)
2567 2568 2569 2570 2571 2572
{
#ifdef CONFIG_NO_HZ_FULL
	if (tick_nohz_full_cpu(cpu))
		smp_send_reschedule(cpu);
#endif /* #ifdef CONFIG_NO_HZ_FULL */
}
2573 2574 2575 2576


#ifdef CONFIG_NO_HZ_FULL_SYSIDLE

2577
static int full_sysidle_state;		/* Current system-idle state. */
2578 2579 2580 2581 2582 2583
#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. */

2584 2585 2586 2587 2588 2589
/*
 * 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.
 */
2590
static void rcu_sysidle_enter(int irq)
2591 2592
{
	unsigned long j;
2593
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
2594

2595 2596 2597 2598
	/* If there are no nohz_full= CPUs, no need to track this. */
	if (!tick_nohz_full_enabled())
		return;

2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
	/* 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;
2618
	WRITE_ONCE(rdtp->dynticks_idle_jiffies, j);
2619
	smp_mb__before_atomic();
2620
	atomic_inc(&rdtp->dynticks_idle);
2621
	smp_mb__after_atomic();
2622 2623 2624
	WARN_ON_ONCE(atomic_read(&rdtp->dynticks_idle) & 0x1);
}

2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635
/*
 * 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)
{
2636
	int oldstate = READ_ONCE(full_sysidle_state);
2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656
	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. */
}

2657 2658 2659 2660 2661
/*
 * 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.
 */
2662
static void rcu_sysidle_exit(int irq)
2663
{
2664 2665
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);

2666 2667 2668 2669
	/* If there are no nohz_full= CPUs, no need to track this. */
	if (!tick_nohz_full_enabled())
		return;

2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691
	/* 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. */
2692
	smp_mb__before_atomic();
2693
	atomic_inc(&rdtp->dynticks_idle);
2694
	smp_mb__after_atomic();
2695
	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks_idle) & 0x1));
2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714

	/*
	 * 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
2715 2716
 * does not count as idle.  The caller must have disabled interrupts,
 * and must be running on tick_do_timer_cpu.
2717 2718 2719 2720 2721 2722 2723 2724
 */
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;

2725 2726 2727 2728
	/* If there are no nohz_full= CPUs, don't check system-wide idleness. */
	if (!tick_nohz_full_enabled())
		return;

2729 2730 2731 2732 2733
	/*
	 * 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.
	 */
2734
	if (!*isidle || rdp->rsp != rcu_state_p ||
2735 2736
	    cpu_is_offline(rdp->cpu) || rdp->cpu == tick_do_timer_cpu)
		return;
2737 2738
	/* Verify affinity of current kthread. */
	WARN_ON_ONCE(smp_processor_id() != tick_do_timer_cpu);
2739 2740 2741 2742 2743 2744 2745 2746 2747 2748

	/* 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. */
2749
	j = READ_ONCE(rdtp->dynticks_idle_jiffies);
2750 2751 2752 2753 2754 2755 2756 2757 2758 2759
	/* 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)
{
2760
	return rsp == rcu_state_p;
2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785
}

/*
 * 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. */
2786
	switch (READ_ONCE(full_sysidle_state)) {
2787 2788 2789
	case RCU_SYSIDLE_NOT:

		/* First time all are idle, so note a short idle period. */
2790
		WRITE_ONCE(full_sysidle_state, RCU_SYSIDLE_SHORT);
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826
		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();
2827
	if (full_sysidle_state > RCU_SYSIDLE_SHORT)
2828
		WRITE_ONCE(full_sysidle_state, RCU_SYSIDLE_NOT);
2829 2830 2831 2832 2833 2834 2835 2836 2837
}

/*
 * 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)
{
2838
	if (rsp != rcu_state_p)
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854
		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)
{
2855 2856 2857 2858
	/* If there are no nohz_full= CPUs, no need to track this. */
	if (!tick_nohz_full_enabled())
		return;

2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
	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);
2880
	WRITE_ONCE(rshp->inuse, 0);
2881 2882 2883 2884
}

/*
 * Check to see if the system is fully idle, other than the timekeeping CPU.
2885 2886
 * The caller must have disabled interrupts.  This is not intended to be
 * called unless tick_nohz_full_enabled().
2887 2888 2889 2890
 */
bool rcu_sys_is_idle(void)
{
	static struct rcu_sysidle_head rsh;
2891
	int rss = READ_ONCE(full_sysidle_state);
2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911

	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) {
2912
				rdp = per_cpu_ptr(rcu_state_p->rda, cpu);
2913 2914 2915 2916
				rcu_sysidle_check_cpu(rdp, &isidle, &maxj);
				if (!isidle)
					break;
			}
2917
			rcu_sysidle_report(rcu_state_p, isidle, maxj, false);
2918
			oldrss = rss;
2919
			rss = READ_ONCE(full_sysidle_state);
2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943
		}
	}

	/* 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 &&
2944
	    !rcu_gp_in_progress(rcu_state_p) &&
2945 2946 2947
	    !rsh.inuse && xchg(&rsh.inuse, 1) == 0)
		call_rcu(&rsh.rh, rcu_sysidle_cb);
	return false;
2948 2949
}

2950 2951 2952 2953 2954 2955 2956 2957 2958 2959
/*
 * 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 */

2960
static void rcu_sysidle_enter(int irq)
2961 2962 2963
{
}

2964
static void rcu_sysidle_exit(int irq)
2965 2966 2967
{
}

2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
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)
{
}

2983 2984 2985 2986 2987
static void rcu_sysidle_init_percpu_data(struct rcu_dynticks *rdtp)
{
}

#endif /* #else #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
2988 2989 2990 2991 2992 2993 2994 2995

/*
 * 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
2996
 * CONFIG_RCU_NOCB_CPU CPUs.
2997 2998 2999 3000 3001 3002
 */
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) ||
3003
	     ULONG_CMP_LT(jiffies, READ_ONCE(rsp->gp_start) + HZ)))
3004
		return true;
3005
#endif /* #ifdef CONFIG_NO_HZ_FULL */
3006
	return false;
3007
}
3008 3009 3010 3011 3012 3013 3014

/*
 * Bind the grace-period kthread for the sysidle flavor of RCU to the
 * timekeeping CPU.
 */
static void rcu_bind_gp_kthread(void)
{
3015
	int __maybe_unused cpu;
3016

3017
	if (!tick_nohz_full_enabled())
3018
		return;
3019 3020
#ifdef CONFIG_NO_HZ_FULL_SYSIDLE
	cpu = tick_do_timer_cpu;
3021
	if (cpu >= 0 && cpu < nr_cpu_ids)
3022
		set_cpus_allowed_ptr(current, cpumask_of(cpu));
3023
#else /* #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
3024
	housekeeping_affine(current);
3025
#endif /* #else #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
3026
}
3027 3028 3029 3030 3031

/* 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)
3032
	WRITE_ONCE(current->rcu_tasks_idle_cpu, smp_processor_id());
3033 3034 3035 3036 3037 3038 3039
#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)
3040
	WRITE_ONCE(current->rcu_tasks_idle_cpu, -1);
3041 3042
#endif /* #if defined(CONFIG_TASKS_RCU) && defined(CONFIG_NO_HZ_FULL) */
}