rcutree.c 55.8 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
/*
 * Read-Copy Update mechanism for mutual exclusion
 *
 * 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
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
 *
 * Copyright IBM Corporation, 2008
 *
 * Authors: Dipankar Sarma <dipankar@in.ibm.com>
 *	    Manfred Spraul <manfred@colorfullife.com>
 *	    Paul E. McKenney <paulmck@linux.vnet.ibm.com> Hierarchical version
 *
 * Based on the original work by Paul McKenney <paulmck@us.ibm.com>
 * and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.
 *
 * For detailed explanation of Read-Copy Update mechanism see -
28
 *	Documentation/RCU
29 30 31 32 33 34 35 36 37
 */
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/smp.h>
#include <linux/rcupdate.h>
#include <linux/interrupt.h>
#include <linux/sched.h>
38
#include <linux/nmi.h>
39 40 41 42 43 44 45 46 47 48 49
#include <asm/atomic.h>
#include <linux/bitops.h>
#include <linux/module.h>
#include <linux/completion.h>
#include <linux/moduleparam.h>
#include <linux/percpu.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/mutex.h>
#include <linux/time.h>

50 51
#include "rcutree.h"

52 53
/* Data structures. */

54
static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];
55

56 57 58 59 60 61
#define RCU_STATE_INITIALIZER(name) { \
	.level = { &name.node[0] }, \
	.levelcnt = { \
		NUM_RCU_LVL_0,  /* root of hierarchy. */ \
		NUM_RCU_LVL_1, \
		NUM_RCU_LVL_2, \
62 63
		NUM_RCU_LVL_3, \
		NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \
64
	}, \
65
	.signaled = RCU_GP_IDLE, \
66 67
	.gpnum = -300, \
	.completed = -300, \
P
Paul E. McKenney 已提交
68
	.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&name.onofflock), \
69 70 71
	.orphan_cbs_list = NULL, \
	.orphan_cbs_tail = &name.orphan_cbs_list, \
	.orphan_qlen = 0, \
P
Paul E. McKenney 已提交
72
	.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&name.fqslock), \
73 74 75 76
	.n_force_qs = 0, \
	.n_force_qs_ngp = 0, \
}

77 78
struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched_state);
DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
79

80 81
struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh_state);
DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
82

83 84 85 86 87 88 89 90 91 92
/*
 * Return true if an RCU grace period is in progress.  The ACCESS_ONCE()s
 * permit this function to be invoked without holding the root rcu_node
 * structure's ->lock, but of course results can be subject to change.
 */
static int rcu_gp_in_progress(struct rcu_state *rsp)
{
	return ACCESS_ONCE(rsp->completed) != ACCESS_ONCE(rsp->gpnum);
}

93
/*
94
 * Note a quiescent state.  Because we do not need to know
95
 * how many quiescent states passed, just if there was at least
96
 * one since the start of the grace period, this just sets a flag.
97
 */
98
void rcu_sched_qs(int cpu)
99
{
100
	struct rcu_data *rdp = &per_cpu(rcu_sched_data, cpu);
101

102
	rdp->passed_quiesc_completed = rdp->gpnum - 1;
103 104
	barrier();
	rdp->passed_quiesc = 1;
105 106
}

107
void rcu_bh_qs(int cpu)
108
{
109
	struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);
110

111
	rdp->passed_quiesc_completed = rdp->gpnum - 1;
112 113
	barrier();
	rdp->passed_quiesc = 1;
114
}
115

116 117 118 119 120 121 122 123 124 125
/*
 * Note a context switch.  This is a quiescent state for RCU-sched,
 * and requires special handling for preemptible RCU.
 */
void rcu_note_context_switch(int cpu)
{
	rcu_sched_qs(cpu);
	rcu_preempt_note_context_switch(cpu);
}

126
#ifdef CONFIG_NO_HZ
127 128 129 130
DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
	.dynticks_nesting = 1,
	.dynticks = 1,
};
131 132 133 134 135 136
#endif /* #ifdef CONFIG_NO_HZ */

static int blimit = 10;		/* Maximum callbacks per softirq. */
static int qhimark = 10000;	/* If this many pending, ignore blimit. */
static int qlowmark = 100;	/* Once only this many pending, use blimit. */

137 138 139 140
module_param(blimit, int, 0);
module_param(qhimark, int, 0);
module_param(qlowmark, int, 0);

141
static void force_quiescent_state(struct rcu_state *rsp, int relaxed);
142
static int rcu_pending(int cpu);
143 144

/*
145
 * Return the number of RCU-sched batches processed thus far for debug & stats.
146
 */
147
long rcu_batches_completed_sched(void)
148
{
149
	return rcu_sched_state.completed;
150
}
151
EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);
152 153 154 155 156 157 158 159 160 161

/*
 * Return the number of RCU BH batches processed thus far for debug & stats.
 */
long rcu_batches_completed_bh(void)
{
	return rcu_bh_state.completed;
}
EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);

162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
/*
 * Force a quiescent state for RCU BH.
 */
void rcu_bh_force_quiescent_state(void)
{
	force_quiescent_state(&rcu_bh_state, 0);
}
EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

/*
 * Force a quiescent state for RCU-sched.
 */
void rcu_sched_force_quiescent_state(void)
{
	force_quiescent_state(&rcu_sched_state, 0);
}
EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
/*
 * Does the CPU have callbacks ready to be invoked?
 */
static int
cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)
{
	return &rdp->nxtlist != rdp->nxttail[RCU_DONE_TAIL];
}

/*
 * Does the current CPU require a yet-as-unscheduled grace period?
 */
static int
cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)
{
195
	return *rdp->nxttail[RCU_DONE_TAIL] && !rcu_gp_in_progress(rsp);
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
}

/*
 * Return the root node of the specified rcu_state structure.
 */
static struct rcu_node *rcu_get_root(struct rcu_state *rsp)
{
	return &rsp->node[0];
}

#ifdef CONFIG_SMP

/*
 * If the specified CPU is offline, tell the caller that it is in
 * a quiescent state.  Otherwise, whack it with a reschedule IPI.
 * Grace periods can end up waiting on an offline CPU when that
 * CPU is in the process of coming online -- it will be added to the
 * rcu_node bitmasks before it actually makes it online.  The same thing
 * can happen while a CPU is in the process of coming online.  Because this
 * race is quite rare, we check for it after detecting that the grace
 * period has been delayed rather than checking each and every CPU
 * each and every time we start a new grace period.
 */
static int rcu_implicit_offline_qs(struct rcu_data *rdp)
{
	/*
	 * If the CPU is offline, it is in a quiescent state.  We can
	 * trust its state not to change because interrupts are disabled.
	 */
	if (cpu_is_offline(rdp->cpu)) {
		rdp->offline_fqs++;
		return 1;
	}

230 231 232 233
	/* If preemptable RCU, no point in sending reschedule IPI. */
	if (rdp->preemptable)
		return 0;

234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264
	/* The CPU is online, so send it a reschedule IPI. */
	if (rdp->cpu != smp_processor_id())
		smp_send_reschedule(rdp->cpu);
	else
		set_need_resched();
	rdp->resched_ipi++;
	return 0;
}

#endif /* #ifdef CONFIG_SMP */

#ifdef CONFIG_NO_HZ

/**
 * rcu_enter_nohz - inform RCU that current CPU is entering nohz
 *
 * Enter nohz mode, in other words, -leave- the mode in which RCU
 * read-side critical sections can occur.  (Though RCU read-side
 * critical sections can occur in irq handlers in nohz mode, a possibility
 * handled by rcu_irq_enter() and rcu_irq_exit()).
 */
void rcu_enter_nohz(void)
{
	unsigned long flags;
	struct rcu_dynticks *rdtp;

	smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
	rdtp->dynticks++;
	rdtp->dynticks_nesting--;
265
	WARN_ON_ONCE(rdtp->dynticks & 0x1);
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
	local_irq_restore(flags);
}

/*
 * rcu_exit_nohz - inform RCU that current CPU is leaving nohz
 *
 * Exit nohz mode, in other words, -enter- the mode in which RCU
 * read-side critical sections normally occur.
 */
void rcu_exit_nohz(void)
{
	unsigned long flags;
	struct rcu_dynticks *rdtp;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
	rdtp->dynticks++;
	rdtp->dynticks_nesting++;
284
	WARN_ON_ONCE(!(rdtp->dynticks & 0x1));
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302
	local_irq_restore(flags);
	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
}

/**
 * rcu_nmi_enter - inform RCU of entry to NMI context
 *
 * If the CPU was idle with dynamic ticks active, and there is no
 * irq handler running, this updates rdtp->dynticks_nmi to let the
 * RCU grace-period handling know that the CPU is active.
 */
void rcu_nmi_enter(void)
{
	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

	if (rdtp->dynticks & 0x1)
		return;
	rdtp->dynticks_nmi++;
303
	WARN_ON_ONCE(!(rdtp->dynticks_nmi & 0x1));
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321
	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
}

/**
 * rcu_nmi_exit - inform RCU of exit from NMI context
 *
 * If the CPU was idle with dynamic ticks active, and there is no
 * irq handler running, this updates rdtp->dynticks_nmi to let the
 * RCU grace-period handling know that the CPU is no longer active.
 */
void rcu_nmi_exit(void)
{
	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

	if (rdtp->dynticks & 0x1)
		return;
	smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
	rdtp->dynticks_nmi++;
322
	WARN_ON_ONCE(rdtp->dynticks_nmi & 0x1);
323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
}

/**
 * rcu_irq_enter - inform RCU of entry to hard irq context
 *
 * If the CPU was idle with dynamic ticks active, this updates the
 * rdtp->dynticks to let the RCU handling know that the CPU is active.
 */
void rcu_irq_enter(void)
{
	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

	if (rdtp->dynticks_nesting++)
		return;
	rdtp->dynticks++;
338
	WARN_ON_ONCE(!(rdtp->dynticks & 0x1));
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356
	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
}

/**
 * rcu_irq_exit - inform RCU of exit from hard irq context
 *
 * If the CPU was idle with dynamic ticks active, update the rdp->dynticks
 * to put let the RCU handling be aware that the CPU is going back to idle
 * with no ticks.
 */
void rcu_irq_exit(void)
{
	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

	if (--rdtp->dynticks_nesting)
		return;
	smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
	rdtp->dynticks++;
357
	WARN_ON_ONCE(rdtp->dynticks & 0x1);
358 359

	/* If the interrupt queued a callback, get out of dyntick mode. */
360
	if (__get_cpu_var(rcu_sched_data).nxtlist ||
361 362 363 364 365 366 367 368 369
	    __get_cpu_var(rcu_bh_data).nxtlist)
		set_need_resched();
}

#ifdef CONFIG_SMP

/*
 * Snapshot the specified CPU's dynticks counter so that we can later
 * credit them with an implicit quiescent state.  Return 1 if this CPU
370
 * is in dynticks idle mode, which is an extended quiescent state.
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
 */
static int dyntick_save_progress_counter(struct rcu_data *rdp)
{
	int ret;
	int snap;
	int snap_nmi;

	snap = rdp->dynticks->dynticks;
	snap_nmi = rdp->dynticks->dynticks_nmi;
	smp_mb();	/* Order sampling of snap with end of grace period. */
	rdp->dynticks_snap = snap;
	rdp->dynticks_nmi_snap = snap_nmi;
	ret = ((snap & 0x1) == 0) && ((snap_nmi & 0x1) == 0);
	if (ret)
		rdp->dynticks_fqs++;
	return ret;
}

/*
 * Return true if the specified CPU has passed through a quiescent
 * state by virtue of being in or having passed through an dynticks
 * idle state since the last call to dyntick_save_progress_counter()
 * for this same CPU.
 */
static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
{
	long curr;
	long curr_nmi;
	long snap;
	long snap_nmi;

	curr = rdp->dynticks->dynticks;
	snap = rdp->dynticks_snap;
	curr_nmi = rdp->dynticks->dynticks_nmi;
	snap_nmi = rdp->dynticks_nmi_snap;
	smp_mb(); /* force ordering with cpu entering/leaving dynticks. */

	/*
	 * If the CPU passed through or entered a dynticks idle phase with
	 * no active irq/NMI handlers, then we can safely pretend that the CPU
	 * already acknowledged the request to pass through a quiescent
	 * state.  Either way, that CPU cannot possibly be in an RCU
	 * read-side critical section that started before the beginning
	 * of the current RCU grace period.
	 */
	if ((curr != snap || (curr & 0x1) == 0) &&
	    (curr_nmi != snap_nmi || (curr_nmi & 0x1) == 0)) {
		rdp->dynticks_fqs++;
		return 1;
	}

	/* Go check for the CPU being offline. */
	return rcu_implicit_offline_qs(rdp);
}

#endif /* #ifdef CONFIG_SMP */

#else /* #ifdef CONFIG_NO_HZ */

#ifdef CONFIG_SMP

static int dyntick_save_progress_counter(struct rcu_data *rdp)
{
	return 0;
}

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
{
	return rcu_implicit_offline_qs(rdp);
}

#endif /* #ifdef CONFIG_SMP */

#endif /* #else #ifdef CONFIG_NO_HZ */

#ifdef CONFIG_RCU_CPU_STALL_DETECTOR

static void record_gp_stall_check_time(struct rcu_state *rsp)
{
	rsp->gp_start = jiffies;
	rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_CHECK;
}

static void print_other_cpu_stall(struct rcu_state *rsp)
{
	int cpu;
	long delta;
	unsigned long flags;
	struct rcu_node *rnp = rcu_get_root(rsp);

	/* Only let one CPU complain about others per time interval. */

P
Paul E. McKenney 已提交
463
	raw_spin_lock_irqsave(&rnp->lock, flags);
464
	delta = jiffies - rsp->jiffies_stall;
465
	if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {
P
Paul E. McKenney 已提交
466
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
467 468 469
		return;
	}
	rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_RECHECK;
470 471 472 473 474 475

	/*
	 * Now rat on any tasks that got kicked up to the root rcu_node
	 * due to CPU offlining.
	 */
	rcu_print_task_stall(rnp);
P
Paul E. McKenney 已提交
476
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
477 478 479 480

	/* OK, time to rat on our buddy... */

	printk(KERN_ERR "INFO: RCU detected CPU stalls:");
481
	rcu_for_each_leaf_node(rsp, rnp) {
482
		raw_spin_lock_irqsave(&rnp->lock, flags);
483
		rcu_print_task_stall(rnp);
484
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
485
		if (rnp->qsmask == 0)
486
			continue;
487 488 489
		for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)
			if (rnp->qsmask & (1UL << cpu))
				printk(" %d", rnp->grplo + cpu);
490 491 492
	}
	printk(" (detected by %d, t=%ld jiffies)\n",
	       smp_processor_id(), (long)(jiffies - rsp->gp_start));
493 494
	trigger_all_cpu_backtrace();

495 496 497 498
	/* If so configured, complain about tasks blocking the grace period. */

	rcu_print_detail_task_stall(rsp);

499 500 501 502 503 504 505 506 507 508
	force_quiescent_state(rsp, 0);  /* Kick them all. */
}

static void print_cpu_stall(struct rcu_state *rsp)
{
	unsigned long flags;
	struct rcu_node *rnp = rcu_get_root(rsp);

	printk(KERN_ERR "INFO: RCU detected CPU %d stall (t=%lu jiffies)\n",
			smp_processor_id(), jiffies - rsp->gp_start);
509 510
	trigger_all_cpu_backtrace();

P
Paul E. McKenney 已提交
511
	raw_spin_lock_irqsave(&rnp->lock, flags);
512
	if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))
513 514
		rsp->jiffies_stall =
			jiffies + RCU_SECONDS_TILL_STALL_RECHECK;
P
Paul E. McKenney 已提交
515
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
516

517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
	set_need_resched();  /* kick ourselves to get things going. */
}

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
{
	long delta;
	struct rcu_node *rnp;

	delta = jiffies - rsp->jiffies_stall;
	rnp = rdp->mynode;
	if ((rnp->qsmask & rdp->grpmask) && delta >= 0) {

		/* We haven't checked in, so go dump stack. */
		print_cpu_stall(rsp);

532
	} else if (rcu_gp_in_progress(rsp) && delta >= RCU_STALL_RAT_DELAY) {
533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553

		/* They had two time units to dump stack, so complain. */
		print_other_cpu_stall(rsp);
	}
}

#else /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

static void record_gp_stall_check_time(struct rcu_state *rsp)
{
}

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
{
}

#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

/*
 * Update CPU-local rcu_data state to record the newly noticed grace period.
 * This is used both when we started the grace period and when we notice
554 555 556
 * that someone else started the grace period.  The caller must hold the
 * ->lock of the leaf rcu_node structure corresponding to the current CPU,
 *  and must have irqs disabled.
557
 */
558 559 560 561 562 563 564 565 566
static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
{
	if (rdp->gpnum != rnp->gpnum) {
		rdp->qs_pending = 1;
		rdp->passed_quiesc = 0;
		rdp->gpnum = rnp->gpnum;
	}
}

567 568
static void note_new_gpnum(struct rcu_state *rsp, struct rcu_data *rdp)
{
569 570 571 572 573 574
	unsigned long flags;
	struct rcu_node *rnp;

	local_irq_save(flags);
	rnp = rdp->mynode;
	if (rdp->gpnum == ACCESS_ONCE(rnp->gpnum) || /* outside lock. */
P
Paul E. McKenney 已提交
575
	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
576 577 578 579
		local_irq_restore(flags);
		return;
	}
	__note_new_gpnum(rsp, rnp, rdp);
P
Paul E. McKenney 已提交
580
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
}

/*
 * Did someone else start a new RCU grace period start since we last
 * checked?  Update local state appropriately if so.  Must be called
 * on the CPU corresponding to rdp.
 */
static int
check_for_new_grace_period(struct rcu_state *rsp, struct rcu_data *rdp)
{
	unsigned long flags;
	int ret = 0;

	local_irq_save(flags);
	if (rdp->gpnum != rsp->gpnum) {
		note_new_gpnum(rsp, rdp);
		ret = 1;
	}
	local_irq_restore(flags);
	return ret;
}

603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638
/*
 * Advance this CPU's callbacks, but only if the current grace period
 * has ended.  This may be called only from the CPU to whom the rdp
 * belongs.  In addition, the corresponding leaf rcu_node structure's
 * ->lock must be held by the caller, with irqs disabled.
 */
static void
__rcu_process_gp_end(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
{
	/* Did another grace period end? */
	if (rdp->completed != rnp->completed) {

		/* Advance callbacks.  No harm if list empty. */
		rdp->nxttail[RCU_DONE_TAIL] = rdp->nxttail[RCU_WAIT_TAIL];
		rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_READY_TAIL];
		rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

		/* Remember that we saw this grace-period completion. */
		rdp->completed = rnp->completed;
	}
}

/*
 * Advance this CPU's callbacks, but only if the current grace period
 * has ended.  This may be called only from the CPU to whom the rdp
 * belongs.
 */
static void
rcu_process_gp_end(struct rcu_state *rsp, struct rcu_data *rdp)
{
	unsigned long flags;
	struct rcu_node *rnp;

	local_irq_save(flags);
	rnp = rdp->mynode;
	if (rdp->completed == ACCESS_ONCE(rnp->completed) || /* outside lock. */
P
Paul E. McKenney 已提交
639
	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
640 641 642 643
		local_irq_restore(flags);
		return;
	}
	__rcu_process_gp_end(rsp, rnp, rdp);
P
Paul E. McKenney 已提交
644
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670
}

/*
 * Do per-CPU grace-period initialization for running CPU.  The caller
 * must hold the lock of the leaf rcu_node structure corresponding to
 * this CPU.
 */
static void
rcu_start_gp_per_cpu(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
{
	/* Prior grace period ended, so advance callbacks for current CPU. */
	__rcu_process_gp_end(rsp, rnp, rdp);

	/*
	 * Because this CPU just now started the new grace period, we know
	 * that all of its callbacks will be covered by this upcoming grace
	 * period, even the ones that were registered arbitrarily recently.
	 * Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.
	 *
	 * Other CPUs cannot be sure exactly when the grace period started.
	 * Therefore, their recently registered callbacks must pass through
	 * an additional RCU_NEXT_READY stage, so that they will be handled
	 * by the next RCU grace period.
	 */
	rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
	rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
671 672 673

	/* Set state so that this CPU will detect the next quiescent state. */
	__note_new_gpnum(rsp, rnp, rdp);
674 675
}

676 677 678 679 680 681 682 683 684 685 686 687 688
/*
 * Start a new RCU grace period if warranted, re-initializing the hierarchy
 * in preparation for detecting the next grace period.  The caller must hold
 * the root node's ->lock, which is released before return.  Hard irqs must
 * be disabled.
 */
static void
rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
	__releases(rcu_get_root(rsp)->lock)
{
	struct rcu_data *rdp = rsp->rda[smp_processor_id()];
	struct rcu_node *rnp = rcu_get_root(rsp);

689
	if (!cpu_needs_another_gp(rsp, rdp) || rsp->fqs_active) {
690 691
		if (cpu_needs_another_gp(rsp, rdp))
			rsp->fqs_need_gp = 1;
692
		if (rnp->completed == rsp->completed) {
P
Paul E. McKenney 已提交
693
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
694 695
			return;
		}
P
Paul E. McKenney 已提交
696
		raw_spin_unlock(&rnp->lock);	 /* irqs remain disabled. */
697 698 699 700 701 702 703

		/*
		 * Propagate new ->completed value to rcu_node structures
		 * so that other CPUs don't have to wait until the start
		 * of the next grace period to process their callbacks.
		 */
		rcu_for_each_node_breadth_first(rsp, rnp) {
P
Paul E. McKenney 已提交
704
			raw_spin_lock(&rnp->lock); /* irqs already disabled. */
705
			rnp->completed = rsp->completed;
P
Paul E. McKenney 已提交
706
			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
707 708
		}
		local_irq_restore(flags);
709 710 711 712 713
		return;
	}

	/* Advance to a new grace period and initialize state. */
	rsp->gpnum++;
714
	WARN_ON_ONCE(rsp->signaled == RCU_GP_INIT);
715 716 717 718 719 720
	rsp->signaled = RCU_GP_INIT; /* Hold off force_quiescent_state. */
	rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
	record_gp_stall_check_time(rsp);

	/* Special-case the common single-level case. */
	if (NUM_RCU_NODES == 1) {
721
		rcu_preempt_check_blocked_tasks(rnp);
722
		rnp->qsmask = rnp->qsmaskinit;
723
		rnp->gpnum = rsp->gpnum;
724
		rnp->completed = rsp->completed;
725
		rsp->signaled = RCU_SIGNAL_INIT; /* force_quiescent_state OK. */
726
		rcu_start_gp_per_cpu(rsp, rnp, rdp);
P
Paul E. McKenney 已提交
727
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
728 729 730
		return;
	}

P
Paul E. McKenney 已提交
731
	raw_spin_unlock(&rnp->lock);  /* leave irqs disabled. */
732 733 734


	/* Exclude any concurrent CPU-hotplug operations. */
P
Paul E. McKenney 已提交
735
	raw_spin_lock(&rsp->onofflock);  /* irqs already disabled. */
736 737

	/*
738 739 740 741 742 743 744 745 746
	 * Set the quiescent-state-needed bits in all the rcu_node
	 * structures for all currently online CPUs in breadth-first
	 * order, starting from the root rcu_node structure.  This
	 * operation relies on the layout of the hierarchy within the
	 * rsp->node[] array.  Note that other CPUs will access only
	 * the leaves of the hierarchy, which still indicate that no
	 * grace period is in progress, at least until the corresponding
	 * leaf node has been initialized.  In addition, we have excluded
	 * CPU-hotplug operations.
747 748 749 750
	 *
	 * Note that the grace period cannot complete until we finish
	 * the initialization process, as there will be at least one
	 * qsmask bit set in the root node until that time, namely the
751 752
	 * one corresponding to this CPU, due to the fact that we have
	 * irqs disabled.
753
	 */
754
	rcu_for_each_node_breadth_first(rsp, rnp) {
P
Paul E. McKenney 已提交
755
		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
756
		rcu_preempt_check_blocked_tasks(rnp);
757
		rnp->qsmask = rnp->qsmaskinit;
758
		rnp->gpnum = rsp->gpnum;
759 760 761
		rnp->completed = rsp->completed;
		if (rnp == rdp->mynode)
			rcu_start_gp_per_cpu(rsp, rnp, rdp);
P
Paul E. McKenney 已提交
762
		raw_spin_unlock(&rnp->lock);	/* irqs remain disabled. */
763 764
	}

765
	rnp = rcu_get_root(rsp);
P
Paul E. McKenney 已提交
766
	raw_spin_lock(&rnp->lock);		/* irqs already disabled. */
767
	rsp->signaled = RCU_SIGNAL_INIT; /* force_quiescent_state now OK. */
P
Paul E. McKenney 已提交
768 769
	raw_spin_unlock(&rnp->lock);		/* irqs remain disabled. */
	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
770 771
}

772
/*
P
Paul E. McKenney 已提交
773 774 775 776 777
 * Report a full set of quiescent states to the specified rcu_state
 * data structure.  This involves cleaning up after the prior grace
 * period and letting rcu_start_gp() start up the next grace period
 * if one is needed.  Note that the caller must hold rnp->lock, as
 * required by rcu_start_gp(), which will release it.
778
 */
P
Paul E. McKenney 已提交
779
static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)
780
	__releases(rcu_get_root(rsp)->lock)
781
{
782
	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
783
	rsp->completed = rsp->gpnum;
784
	rsp->signaled = RCU_GP_IDLE;
785 786 787
	rcu_start_gp(rsp, flags);  /* releases root node's rnp->lock. */
}

788
/*
P
Paul E. McKenney 已提交
789 790 791 792 793 794
 * Similar to rcu_report_qs_rdp(), for which it is a helper function.
 * Allows quiescent states for a group of CPUs to be reported at one go
 * to the specified rcu_node structure, though all the CPUs in the group
 * must be represented by the same rcu_node structure (which need not be
 * a leaf rcu_node structure, though it often will be).  That structure's
 * lock must be held upon entry, and it is released before return.
795 796
 */
static void
P
Paul E. McKenney 已提交
797 798
rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,
		  struct rcu_node *rnp, unsigned long flags)
799 800
	__releases(rnp->lock)
{
801 802
	struct rcu_node *rnp_c;

803 804 805 806 807
	/* Walk up the rcu_node hierarchy. */
	for (;;) {
		if (!(rnp->qsmask & mask)) {

			/* Our bit has already been cleared, so done. */
P
Paul E. McKenney 已提交
808
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
809 810 811
			return;
		}
		rnp->qsmask &= ~mask;
812
		if (rnp->qsmask != 0 || rcu_preempted_readers(rnp)) {
813 814

			/* Other bits still set at this level, so done. */
P
Paul E. McKenney 已提交
815
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
816 817 818 819 820 821 822 823 824
			return;
		}
		mask = rnp->grpmask;
		if (rnp->parent == NULL) {

			/* No more levels.  Exit loop holding root lock. */

			break;
		}
P
Paul E. McKenney 已提交
825
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
826
		rnp_c = rnp;
827
		rnp = rnp->parent;
P
Paul E. McKenney 已提交
828
		raw_spin_lock_irqsave(&rnp->lock, flags);
829
		WARN_ON_ONCE(rnp_c->qsmask);
830 831 832 833
	}

	/*
	 * Get here if we are the last CPU to pass through a quiescent
P
Paul E. McKenney 已提交
834
	 * state for this grace period.  Invoke rcu_report_qs_rsp()
835
	 * to clean up and start the next grace period if one is needed.
836
	 */
P
Paul E. McKenney 已提交
837
	rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */
838 839 840
}

/*
P
Paul E. McKenney 已提交
841 842 843 844 845 846 847
 * Record a quiescent state for the specified CPU to that CPU's rcu_data
 * structure.  This must be either called from the specified CPU, or
 * called when the specified CPU is known to be offline (and when it is
 * also known that no other CPU is concurrently trying to help the offline
 * CPU).  The lastcomp argument is used to make sure we are still in the
 * grace period of interest.  We don't want to end the current grace period
 * based on quiescent states detected in an earlier grace period!
848 849
 */
static void
P
Paul E. McKenney 已提交
850
rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp, long lastcomp)
851 852 853 854 855 856
{
	unsigned long flags;
	unsigned long mask;
	struct rcu_node *rnp;

	rnp = rdp->mynode;
P
Paul E. McKenney 已提交
857
	raw_spin_lock_irqsave(&rnp->lock, flags);
858
	if (lastcomp != rnp->completed) {
859 860 861 862 863 864

		/*
		 * Someone beat us to it for this grace period, so leave.
		 * The race with GP start is resolved by the fact that we
		 * hold the leaf rcu_node lock, so that the per-CPU bits
		 * cannot yet be initialized -- so we would simply find our
P
Paul E. McKenney 已提交
865 866
		 * CPU's bit already cleared in rcu_report_qs_rnp() if this
		 * race occurred.
867 868
		 */
		rdp->passed_quiesc = 0;	/* try again later! */
P
Paul E. McKenney 已提交
869
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
870 871 872 873
		return;
	}
	mask = rdp->grpmask;
	if ((rnp->qsmask & mask) == 0) {
P
Paul E. McKenney 已提交
874
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
875 876 877 878 879 880 881 882 883
	} else {
		rdp->qs_pending = 0;

		/*
		 * This GP can't end until cpu checks in, so all of our
		 * callbacks can be processed during the next GP.
		 */
		rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

P
Paul E. McKenney 已提交
884
		rcu_report_qs_rnp(mask, rsp, rnp, flags); /* rlses rnp->lock */
885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914
	}
}

/*
 * Check to see if there is a new grace period of which this CPU
 * is not yet aware, and if so, set up local rcu_data state for it.
 * Otherwise, see if this CPU has just passed through its first
 * quiescent state for this grace period, and record that fact if so.
 */
static void
rcu_check_quiescent_state(struct rcu_state *rsp, struct rcu_data *rdp)
{
	/* If there is now a new grace period, record and return. */
	if (check_for_new_grace_period(rsp, rdp))
		return;

	/*
	 * Does this CPU still need to do its part for current grace period?
	 * If no, return and let the other CPUs do their part as well.
	 */
	if (!rdp->qs_pending)
		return;

	/*
	 * Was there a quiescent state since the beginning of the grace
	 * period? If no, then exit and wait for the next call.
	 */
	if (!rdp->passed_quiesc)
		return;

P
Paul E. McKenney 已提交
915 916 917 918 919
	/*
	 * Tell RCU we are done (but rcu_report_qs_rdp() will be the
	 * judge of that).
	 */
	rcu_report_qs_rdp(rdp->cpu, rsp, rdp, rdp->passed_quiesc_completed);
920 921 922 923
}

#ifdef CONFIG_HOTPLUG_CPU

924 925 926 927 928 929 930 931 932 933 934 935 936 937
/*
 * Move a dying CPU's RCU callbacks to the ->orphan_cbs_list for the
 * specified flavor of RCU.  The callbacks will be adopted by the next
 * _rcu_barrier() invocation or by the CPU_DEAD notifier, whichever
 * comes first.  Because this is invoked from the CPU_DYING notifier,
 * irqs are already disabled.
 */
static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)
{
	int i;
	struct rcu_data *rdp = rsp->rda[smp_processor_id()];

	if (rdp->nxtlist == NULL)
		return;  /* irqs disabled, so comparison is stable. */
P
Paul E. McKenney 已提交
938
	raw_spin_lock(&rsp->onofflock);  /* irqs already disabled. */
939 940 941 942 943 944 945
	*rsp->orphan_cbs_tail = rdp->nxtlist;
	rsp->orphan_cbs_tail = rdp->nxttail[RCU_NEXT_TAIL];
	rdp->nxtlist = NULL;
	for (i = 0; i < RCU_NEXT_SIZE; i++)
		rdp->nxttail[i] = &rdp->nxtlist;
	rsp->orphan_qlen += rdp->qlen;
	rdp->qlen = 0;
P
Paul E. McKenney 已提交
946
	raw_spin_unlock(&rsp->onofflock);  /* irqs remain disabled. */
947 948 949 950 951 952 953 954 955 956
}

/*
 * Adopt previously orphaned RCU callbacks.
 */
static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
{
	unsigned long flags;
	struct rcu_data *rdp;

P
Paul E. McKenney 已提交
957
	raw_spin_lock_irqsave(&rsp->onofflock, flags);
958 959
	rdp = rsp->rda[smp_processor_id()];
	if (rsp->orphan_cbs_list == NULL) {
P
Paul E. McKenney 已提交
960
		raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
961 962 963 964 965 966 967 968
		return;
	}
	*rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_list;
	rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_tail;
	rdp->qlen += rsp->orphan_qlen;
	rsp->orphan_cbs_list = NULL;
	rsp->orphan_cbs_tail = &rsp->orphan_cbs_list;
	rsp->orphan_qlen = 0;
P
Paul E. McKenney 已提交
969
	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
970 971
}

972 973 974 975 976 977 978 979
/*
 * Remove the outgoing CPU from the bitmasks in the rcu_node hierarchy
 * and move all callbacks from the outgoing CPU to the current one.
 */
static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp)
{
	unsigned long flags;
	unsigned long mask;
980
	int need_report = 0;
981 982 983 984
	struct rcu_data *rdp = rsp->rda[cpu];
	struct rcu_node *rnp;

	/* Exclude any attempts to start a new grace period. */
P
Paul E. McKenney 已提交
985
	raw_spin_lock_irqsave(&rsp->onofflock, flags);
986 987

	/* Remove the outgoing CPU from the masks in the rcu_node hierarchy. */
988
	rnp = rdp->mynode;	/* this is the outgoing CPU's rnp. */
989 990
	mask = rdp->grpmask;	/* rnp->grplo is constant. */
	do {
P
Paul E. McKenney 已提交
991
		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
992 993
		rnp->qsmaskinit &= ~mask;
		if (rnp->qsmaskinit != 0) {
994
			if (rnp != rdp->mynode)
P
Paul E. McKenney 已提交
995
				raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
996 997
			break;
		}
998
		if (rnp == rdp->mynode)
999
			need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);
1000
		else
P
Paul E. McKenney 已提交
1001
			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
1002 1003 1004 1005
		mask = rnp->grpmask;
		rnp = rnp->parent;
	} while (rnp != NULL);

1006 1007 1008
	/*
	 * We still hold the leaf rcu_node structure lock here, and
	 * irqs are still disabled.  The reason for this subterfuge is
P
Paul E. McKenney 已提交
1009 1010
	 * because invoking rcu_report_unblock_qs_rnp() with ->onofflock
	 * held leads to deadlock.
1011
	 */
P
Paul E. McKenney 已提交
1012
	raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */
1013
	rnp = rdp->mynode;
1014
	if (need_report & RCU_OFL_TASKS_NORM_GP)
P
Paul E. McKenney 已提交
1015
		rcu_report_unblock_qs_rnp(rnp, flags);
1016
	else
P
Paul E. McKenney 已提交
1017
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1018 1019
	if (need_report & RCU_OFL_TASKS_EXP_GP)
		rcu_report_exp_rnp(rsp, rnp);
1020

1021
	rcu_adopt_orphan_cbs(rsp);
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
}

/*
 * Remove the specified CPU from the RCU hierarchy and move any pending
 * callbacks that it might have to the current CPU.  This code assumes
 * that at least one CPU in the system will remain running at all times.
 * Any attempt to offline -all- CPUs is likely to strand RCU callbacks.
 */
static void rcu_offline_cpu(int cpu)
{
1032
	__rcu_offline_cpu(cpu, &rcu_sched_state);
1033
	__rcu_offline_cpu(cpu, &rcu_bh_state);
1034
	rcu_preempt_offline_cpu(cpu);
1035 1036 1037 1038
}

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1039 1040 1041 1042 1043 1044 1045 1046
static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)
{
}

static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
{
}

1047 1048 1049 1050 1051 1052 1053 1054 1055 1056
static void rcu_offline_cpu(int cpu)
{
}

#endif /* #else #ifdef CONFIG_HOTPLUG_CPU */

/*
 * Invoke any RCU callbacks that have made it to the end of their grace
 * period.  Thottle as specified by rdp->blimit.
 */
1057
static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
{
	unsigned long flags;
	struct rcu_head *next, *list, **tail;
	int count;

	/* If no callbacks are ready, just return.*/
	if (!cpu_has_callbacks_ready_to_invoke(rdp))
		return;

	/*
	 * Extract the list of ready callbacks, disabling to prevent
	 * races with call_rcu() from interrupt handlers.
	 */
	local_irq_save(flags);
	list = rdp->nxtlist;
	rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];
	*rdp->nxttail[RCU_DONE_TAIL] = NULL;
	tail = rdp->nxttail[RCU_DONE_TAIL];
	for (count = RCU_NEXT_SIZE - 1; count >= 0; count--)
		if (rdp->nxttail[count] == rdp->nxttail[RCU_DONE_TAIL])
			rdp->nxttail[count] = &rdp->nxtlist;
	local_irq_restore(flags);

	/* Invoke callbacks. */
	count = 0;
	while (list) {
		next = list->next;
		prefetch(next);
		list->func(list);
		list = next;
		if (++count >= rdp->blimit)
			break;
	}

	local_irq_save(flags);

	/* Update count, and requeue any remaining callbacks. */
	rdp->qlen -= count;
	if (list != NULL) {
		*tail = rdp->nxtlist;
		rdp->nxtlist = list;
		for (count = 0; count < RCU_NEXT_SIZE; count++)
			if (&rdp->nxtlist == rdp->nxttail[count])
				rdp->nxttail[count] = tail;
			else
				break;
	}

	/* Reinstate batch limit if we have worked down the excess. */
	if (rdp->blimit == LONG_MAX && rdp->qlen <= qlowmark)
		rdp->blimit = blimit;

1110 1111 1112 1113 1114 1115 1116
	/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */
	if (rdp->qlen == 0 && rdp->qlen_last_fqs_check != 0) {
		rdp->qlen_last_fqs_check = 0;
		rdp->n_force_qs_snap = rsp->n_force_qs;
	} else if (rdp->qlen < rdp->qlen_last_fqs_check - qhimark)
		rdp->qlen_last_fqs_check = rdp->qlen;

1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
	local_irq_restore(flags);

	/* Re-raise the RCU softirq if there are callbacks remaining. */
	if (cpu_has_callbacks_ready_to_invoke(rdp))
		raise_softirq(RCU_SOFTIRQ);
}

/*
 * Check to see if this CPU is in a non-context-switch quiescent state
 * (user mode or idle loop for rcu, non-softirq execution for rcu_bh).
 * Also schedule the RCU softirq handler.
 *
 * This function must be called with hardirqs disabled.  It is normally
 * invoked from the scheduling-clock interrupt.  If rcu_pending returns
 * false, there is no point in invoking rcu_check_callbacks().
 */
void rcu_check_callbacks(int cpu, int user)
{
1135 1136
	if (!rcu_pending(cpu))
		return; /* if nothing for RCU to do. */
1137
	if (user ||
1138 1139
	    (idle_cpu(cpu) && rcu_scheduler_active &&
	     !in_softirq() && hardirq_count() <= (1 << HARDIRQ_SHIFT))) {
1140 1141 1142 1143 1144

		/*
		 * Get here if this CPU took its interrupt from user
		 * mode or from the idle loop, and if this is not a
		 * nested interrupt.  In this case, the CPU is in
1145
		 * a quiescent state, so note it.
1146 1147
		 *
		 * No memory barrier is required here because both
1148 1149 1150
		 * rcu_sched_qs() and rcu_bh_qs() reference only CPU-local
		 * variables that other CPUs neither access nor modify,
		 * at least not while the corresponding CPU is online.
1151 1152
		 */

1153 1154
		rcu_sched_qs(cpu);
		rcu_bh_qs(cpu);
1155 1156 1157 1158 1159 1160 1161

	} else if (!in_softirq()) {

		/*
		 * Get here if this CPU did not take its interrupt from
		 * softirq, in other words, if it is not interrupting
		 * a rcu_bh read-side critical section.  This is an _bh
1162
		 * critical section, so note it.
1163 1164
		 */

1165
		rcu_bh_qs(cpu);
1166
	}
1167
	rcu_preempt_check_callbacks(cpu);
1168 1169 1170 1171 1172 1173 1174 1175
	raise_softirq(RCU_SOFTIRQ);
}

#ifdef CONFIG_SMP

/*
 * Scan the leaf rcu_node structures, processing dyntick state for any that
 * have not yet encountered a quiescent state, using the function specified.
1176
 * The caller must have suppressed start of new grace periods.
1177
 */
1178
static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))
1179 1180 1181 1182 1183
{
	unsigned long bit;
	int cpu;
	unsigned long flags;
	unsigned long mask;
1184
	struct rcu_node *rnp;
1185

1186
	rcu_for_each_leaf_node(rsp, rnp) {
1187
		mask = 0;
P
Paul E. McKenney 已提交
1188
		raw_spin_lock_irqsave(&rnp->lock, flags);
1189
		if (!rcu_gp_in_progress(rsp)) {
P
Paul E. McKenney 已提交
1190
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1191
			return;
1192
		}
1193
		if (rnp->qsmask == 0) {
P
Paul E. McKenney 已提交
1194
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1195 1196
			continue;
		}
1197
		cpu = rnp->grplo;
1198
		bit = 1;
1199 1200
		for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {
			if ((rnp->qsmask & bit) != 0 && f(rsp->rda[cpu]))
1201 1202
				mask |= bit;
		}
1203
		if (mask != 0) {
1204

P
Paul E. McKenney 已提交
1205 1206
			/* rcu_report_qs_rnp() releases rnp->lock. */
			rcu_report_qs_rnp(mask, rsp, rnp, flags);
1207 1208
			continue;
		}
P
Paul E. McKenney 已提交
1209
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
	}
}

/*
 * Force quiescent states on reluctant CPUs, and also detect which
 * CPUs are in dyntick-idle mode.
 */
static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
{
	unsigned long flags;
	struct rcu_node *rnp = rcu_get_root(rsp);

1222
	if (!rcu_gp_in_progress(rsp))
1223
		return;  /* No grace period in progress, nothing to force. */
P
Paul E. McKenney 已提交
1224
	if (!raw_spin_trylock_irqsave(&rsp->fqslock, flags)) {
1225 1226 1227
		rsp->n_force_qs_lh++; /* Inexact, can lose counts.  Tough! */
		return;	/* Someone else is already on the job. */
	}
1228
	if (relaxed && ULONG_CMP_GE(rsp->jiffies_force_qs, jiffies))
1229
		goto unlock_fqs_ret; /* no emergency and done recently. */
1230
	rsp->n_force_qs++;
P
Paul E. McKenney 已提交
1231
	raw_spin_lock(&rnp->lock);  /* irqs already disabled */
1232
	rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
1233
	if(!rcu_gp_in_progress(rsp)) {
1234
		rsp->n_force_qs_ngp++;
P
Paul E. McKenney 已提交
1235
		raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
1236
		goto unlock_fqs_ret;  /* no GP in progress, time updated. */
1237
	}
1238
	rsp->fqs_active = 1;
1239
	switch (rsp->signaled) {
1240
	case RCU_GP_IDLE:
1241 1242
	case RCU_GP_INIT:

1243
		break; /* grace period idle or initializing, ignore. */
1244 1245 1246 1247 1248

	case RCU_SAVE_DYNTICK:
		if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)
			break; /* So gcc recognizes the dead code. */

L
Lai Jiangshan 已提交
1249 1250
		raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */

1251
		/* Record dyntick-idle state. */
1252
		force_qs_rnp(rsp, dyntick_save_progress_counter);
P
Paul E. McKenney 已提交
1253
		raw_spin_lock(&rnp->lock);  /* irqs already disabled */
1254
		if (rcu_gp_in_progress(rsp))
1255
			rsp->signaled = RCU_FORCE_QS;
1256
		break;
1257 1258 1259 1260

	case RCU_FORCE_QS:

		/* Check dyntick-idle state, send IPI to laggarts. */
P
Paul E. McKenney 已提交
1261
		raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
1262
		force_qs_rnp(rsp, rcu_implicit_dynticks_qs);
1263 1264 1265

		/* Leave state in case more forcing is required. */

P
Paul E. McKenney 已提交
1266
		raw_spin_lock(&rnp->lock);  /* irqs already disabled */
1267
		break;
1268
	}
1269
	rsp->fqs_active = 0;
1270
	if (rsp->fqs_need_gp) {
P
Paul E. McKenney 已提交
1271
		raw_spin_unlock(&rsp->fqslock); /* irqs remain disabled */
1272 1273 1274 1275
		rsp->fqs_need_gp = 0;
		rcu_start_gp(rsp, flags); /* releases rnp->lock */
		return;
	}
P
Paul E. McKenney 已提交
1276
	raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
1277
unlock_fqs_ret:
P
Paul E. McKenney 已提交
1278
	raw_spin_unlock_irqrestore(&rsp->fqslock, flags);
1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
}

#else /* #ifdef CONFIG_SMP */

static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
{
	set_need_resched();
}

#endif /* #else #ifdef CONFIG_SMP */

/*
 * This does the RCU processing work from softirq context for the
 * specified rcu_state and rcu_data structures.  This may be called
 * only from the CPU to whom the rdp belongs.
 */
static void
__rcu_process_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)
{
	unsigned long flags;

1300 1301
	WARN_ON_ONCE(rdp->beenonline == 0);

1302 1303 1304 1305
	/*
	 * If an RCU GP has gone long enough, go check for dyntick
	 * idle CPUs and, if needed, send resched IPIs.
	 */
1306
	if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319
		force_quiescent_state(rsp, 1);

	/*
	 * Advance callbacks in response to end of earlier grace
	 * period that some other CPU ended.
	 */
	rcu_process_gp_end(rsp, rdp);

	/* Update RCU state based on any recent quiescent states. */
	rcu_check_quiescent_state(rsp, rdp);

	/* Does this CPU require a not-yet-started grace period? */
	if (cpu_needs_another_gp(rsp, rdp)) {
P
Paul E. McKenney 已提交
1320
		raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);
1321 1322 1323 1324
		rcu_start_gp(rsp, flags);  /* releases above lock */
	}

	/* If there are callbacks ready, invoke them. */
1325
	rcu_do_batch(rsp, rdp);
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339
}

/*
 * Do softirq processing for the current CPU.
 */
static void rcu_process_callbacks(struct softirq_action *unused)
{
	/*
	 * Memory references from any prior RCU read-side critical sections
	 * executed by the interrupted code must be seen before any RCU
	 * grace-period manipulations below.
	 */
	smp_mb(); /* See above block comment. */

1340 1341
	__rcu_process_callbacks(&rcu_sched_state,
				&__get_cpu_var(rcu_sched_data));
1342
	__rcu_process_callbacks(&rcu_bh_state, &__get_cpu_var(rcu_bh_data));
1343
	rcu_preempt_process_callbacks();
1344 1345 1346 1347 1348 1349 1350

	/*
	 * Memory references from any later RCU read-side critical sections
	 * executed by the interrupted code must be seen after any RCU
	 * grace-period manipulations above.
	 */
	smp_mb(); /* See above block comment. */
1351 1352 1353

	/* If we are last CPU on way to dyntick-idle mode, accelerate it. */
	rcu_needs_cpu_flush();
1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
}

static void
__call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),
	   struct rcu_state *rsp)
{
	unsigned long flags;
	struct rcu_data *rdp;

	head->func = func;
	head->next = NULL;

	smp_mb(); /* Ensure RCU update seen before callback registry. */

	/*
	 * Opportunistically note grace-period endings and beginnings.
	 * Note that we might see a beginning right after we see an
	 * end, but never vice versa, since this CPU has to pass through
	 * a quiescent state betweentimes.
	 */
	local_irq_save(flags);
	rdp = rsp->rda[smp_processor_id()];
	rcu_process_gp_end(rsp, rdp);
	check_for_new_grace_period(rsp, rdp);

	/* Add the callback to our list. */
	*rdp->nxttail[RCU_NEXT_TAIL] = head;
	rdp->nxttail[RCU_NEXT_TAIL] = &head->next;

	/* Start a new grace period if one not already started. */
1384
	if (!rcu_gp_in_progress(rsp)) {
1385 1386 1387
		unsigned long nestflag;
		struct rcu_node *rnp_root = rcu_get_root(rsp);

P
Paul E. McKenney 已提交
1388
		raw_spin_lock_irqsave(&rnp_root->lock, nestflag);
1389 1390 1391
		rcu_start_gp(rsp, nestflag);  /* releases rnp_root->lock. */
	}

1392 1393 1394 1395 1396 1397 1398 1399
	/*
	 * Force the grace period if too many callbacks or too long waiting.
	 * Enforce hysteresis, and don't invoke force_quiescent_state()
	 * if some other CPU has recently done so.  Also, don't bother
	 * invoking force_quiescent_state() if the newly enqueued callback
	 * is the only one waiting for a grace period to complete.
	 */
	if (unlikely(++rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {
1400
		rdp->blimit = LONG_MAX;
1401 1402 1403 1404 1405
		if (rsp->n_force_qs == rdp->n_force_qs_snap &&
		    *rdp->nxttail[RCU_DONE_TAIL] != head)
			force_quiescent_state(rsp, 0);
		rdp->n_force_qs_snap = rsp->n_force_qs;
		rdp->qlen_last_fqs_check = rdp->qlen;
1406
	} else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))
1407 1408 1409 1410 1411
		force_quiescent_state(rsp, 1);
	local_irq_restore(flags);
}

/*
1412
 * Queue an RCU-sched callback for invocation after a grace period.
1413
 */
1414
void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
1415
{
1416
	__call_rcu(head, func, &rcu_sched_state);
1417
}
1418
EXPORT_SYMBOL_GPL(call_rcu_sched);
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428

/*
 * Queue an RCU for invocation after a quicker grace period.
 */
void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
{
	__call_rcu(head, func, &rcu_bh_state);
}
EXPORT_SYMBOL_GPL(call_rcu_bh);

1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
/**
 * synchronize_sched - wait until an rcu-sched grace period has elapsed.
 *
 * Control will return to the caller some time after a full rcu-sched
 * grace period has elapsed, in other words after all currently executing
 * rcu-sched read-side critical sections have completed.   These read-side
 * critical sections are delimited by rcu_read_lock_sched() and
 * rcu_read_unlock_sched(), and may be nested.  Note that preempt_disable(),
 * local_irq_disable(), and so on may be used in place of
 * rcu_read_lock_sched().
 *
 * This means that all preempt_disable code sequences, including NMI and
 * hardware-interrupt handlers, in progress on entry will have completed
 * before this primitive returns.  However, this does not guarantee that
 * softirq handlers will have completed, since in some kernels, these
 * handlers can run in process context, and can block.
 *
 * This primitive provides the guarantees made by the (now removed)
 * synchronize_kernel() API.  In contrast, synchronize_rcu() only
 * guarantees that rcu_read_lock() sections will have completed.
 * In "classic RCU", these two guarantees happen to be one and
 * the same, but can differ in realtime RCU implementations.
 */
void synchronize_sched(void)
{
	struct rcu_synchronize rcu;

	if (rcu_blocking_is_gp())
		return;

	init_completion(&rcu.completion);
	/* Will wake me after RCU finished. */
	call_rcu_sched(&rcu.head, wakeme_after_rcu);
	/* Wait for it. */
	wait_for_completion(&rcu.completion);
}
EXPORT_SYMBOL_GPL(synchronize_sched);

/**
 * synchronize_rcu_bh - wait until an rcu_bh grace period has elapsed.
 *
 * Control will return to the caller some time after a full rcu_bh grace
 * period has elapsed, in other words after all currently executing rcu_bh
 * read-side critical sections have completed.  RCU read-side critical
 * sections are delimited by rcu_read_lock_bh() and rcu_read_unlock_bh(),
 * and may be nested.
 */
void synchronize_rcu_bh(void)
{
	struct rcu_synchronize rcu;

	if (rcu_blocking_is_gp())
		return;

	init_completion(&rcu.completion);
	/* Will wake me after RCU finished. */
	call_rcu_bh(&rcu.head, wakeme_after_rcu);
	/* Wait for it. */
	wait_for_completion(&rcu.completion);
}
EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

1491 1492 1493 1494 1495 1496 1497 1498 1499
/*
 * Check to see if there is any immediate RCU-related work to be done
 * by the current CPU, for the specified type of RCU, returning 1 if so.
 * The checks are in order of increasing expense: checks that can be
 * carried out against CPU-local state are performed first.  However,
 * we must check for CPU stalls first, else we might not get a chance.
 */
static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)
{
1500 1501
	struct rcu_node *rnp = rdp->mynode;

1502 1503 1504 1505 1506 1507
	rdp->n_rcu_pending++;

	/* Check for CPU stalls, if enabled. */
	check_cpu_stall(rsp, rdp);

	/* Is the RCU core waiting for a quiescent state from this CPU? */
1508
	if (rdp->qs_pending) {
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518

		/*
		 * If force_quiescent_state() coming soon and this CPU
		 * needs a quiescent state, and this is either RCU-sched
		 * or RCU-bh, force a local reschedule.
		 */
		if (!rdp->preemptable &&
		    ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,
				 jiffies))
			set_need_resched();
1519
		rdp->n_rp_qs_pending++;
1520
		return 1;
1521
	}
1522 1523

	/* Does this CPU have callbacks ready to invoke? */
1524 1525
	if (cpu_has_callbacks_ready_to_invoke(rdp)) {
		rdp->n_rp_cb_ready++;
1526
		return 1;
1527
	}
1528 1529

	/* Has RCU gone idle with this CPU needing another grace period? */
1530 1531
	if (cpu_needs_another_gp(rsp, rdp)) {
		rdp->n_rp_cpu_needs_gp++;
1532
		return 1;
1533
	}
1534 1535

	/* Has another RCU grace period completed?  */
1536
	if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */
1537
		rdp->n_rp_gp_completed++;
1538
		return 1;
1539
	}
1540 1541

	/* Has a new RCU grace period started? */
1542
	if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */
1543
		rdp->n_rp_gp_started++;
1544
		return 1;
1545
	}
1546 1547

	/* Has an RCU GP gone long enough to send resched IPIs &c? */
1548
	if (rcu_gp_in_progress(rsp) &&
1549
	    ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) {
1550
		rdp->n_rp_need_fqs++;
1551
		return 1;
1552
	}
1553 1554

	/* nothing to do */
1555
	rdp->n_rp_need_nothing++;
1556 1557 1558 1559 1560 1561 1562 1563
	return 0;
}

/*
 * Check to see if there is any immediate RCU-related work to be done
 * by the current CPU, returning 1 if so.  This function is part of the
 * RCU implementation; it is -not- an exported member of the RCU API.
 */
1564
static int rcu_pending(int cpu)
1565
{
1566
	return __rcu_pending(&rcu_sched_state, &per_cpu(rcu_sched_data, cpu)) ||
1567 1568
	       __rcu_pending(&rcu_bh_state, &per_cpu(rcu_bh_data, cpu)) ||
	       rcu_preempt_pending(cpu);
1569 1570 1571 1572 1573
}

/*
 * 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
1574
 * 1 if so.
1575
 */
1576
static int rcu_needs_cpu_quick_check(int cpu)
1577 1578
{
	/* RCU callbacks either ready or pending? */
1579
	return per_cpu(rcu_sched_data, cpu).nxtlist ||
1580 1581
	       per_cpu(rcu_bh_data, cpu).nxtlist ||
	       rcu_preempt_needs_cpu(cpu);
1582 1583
}

1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};
static atomic_t rcu_barrier_cpu_count;
static DEFINE_MUTEX(rcu_barrier_mutex);
static struct completion rcu_barrier_completion;

static void rcu_barrier_callback(struct rcu_head *notused)
{
	if (atomic_dec_and_test(&rcu_barrier_cpu_count))
		complete(&rcu_barrier_completion);
}

/*
 * Called with preemption disabled, and from cross-cpu IRQ context.
 */
static void rcu_barrier_func(void *type)
{
	int cpu = smp_processor_id();
	struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);
	void (*call_rcu_func)(struct rcu_head *head,
			      void (*func)(struct rcu_head *head));

	atomic_inc(&rcu_barrier_cpu_count);
	call_rcu_func = type;
	call_rcu_func(head, rcu_barrier_callback);
}

/*
 * Orchestrate the specified type of RCU barrier, waiting for all
 * RCU callbacks of the specified type to complete.
 */
1614 1615
static void _rcu_barrier(struct rcu_state *rsp,
			 void (*call_rcu_func)(struct rcu_head *head,
1616 1617 1618
					       void (*func)(struct rcu_head *head)))
{
	BUG_ON(in_interrupt());
1619
	/* Take mutex to serialize concurrent rcu_barrier() requests. */
1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
	mutex_lock(&rcu_barrier_mutex);
	init_completion(&rcu_barrier_completion);
	/*
	 * Initialize rcu_barrier_cpu_count to 1, then invoke
	 * rcu_barrier_func() on each CPU, so that each CPU also has
	 * incremented rcu_barrier_cpu_count.  Only then is it safe to
	 * decrement rcu_barrier_cpu_count -- otherwise the first CPU
	 * might complete its grace period before all of the other CPUs
	 * did their increment, causing this function to return too
	 * early.
	 */
	atomic_set(&rcu_barrier_cpu_count, 1);
1632 1633
	preempt_disable(); /* stop CPU_DYING from filling orphan_cbs_list */
	rcu_adopt_orphan_cbs(rsp);
1634
	on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);
1635
	preempt_enable(); /* CPU_DYING can again fill orphan_cbs_list */
1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
	if (atomic_dec_and_test(&rcu_barrier_cpu_count))
		complete(&rcu_barrier_completion);
	wait_for_completion(&rcu_barrier_completion);
	mutex_unlock(&rcu_barrier_mutex);
}

/**
 * rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.
 */
void rcu_barrier_bh(void)
{
1647
	_rcu_barrier(&rcu_bh_state, call_rcu_bh);
1648 1649 1650 1651 1652 1653 1654 1655
}
EXPORT_SYMBOL_GPL(rcu_barrier_bh);

/**
 * rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.
 */
void rcu_barrier_sched(void)
{
1656
	_rcu_barrier(&rcu_sched_state, call_rcu_sched);
1657 1658 1659
}
EXPORT_SYMBOL_GPL(rcu_barrier_sched);

1660
/*
1661
 * Do boot-time initialization of a CPU's per-CPU RCU data.
1662
 */
1663 1664
static void __init
rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)
1665 1666 1667
{
	unsigned long flags;
	int i;
1668 1669 1670 1671
	struct rcu_data *rdp = rsp->rda[cpu];
	struct rcu_node *rnp = rcu_get_root(rsp);

	/* Set up local state, ensuring consistent view of global state. */
P
Paul E. McKenney 已提交
1672
	raw_spin_lock_irqsave(&rnp->lock, flags);
1673 1674 1675 1676 1677 1678 1679 1680 1681
	rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);
	rdp->nxtlist = NULL;
	for (i = 0; i < RCU_NEXT_SIZE; i++)
		rdp->nxttail[i] = &rdp->nxtlist;
	rdp->qlen = 0;
#ifdef CONFIG_NO_HZ
	rdp->dynticks = &per_cpu(rcu_dynticks, cpu);
#endif /* #ifdef CONFIG_NO_HZ */
	rdp->cpu = cpu;
P
Paul E. McKenney 已提交
1682
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1683 1684 1685 1686 1687 1688 1689
}

/*
 * Initialize a CPU's per-CPU RCU data.  Note that only one online or
 * offline event can be happening at a given time.  Note also that we
 * can accept some slop in the rsp->completed access due to the fact
 * that this CPU cannot possibly have any RCU callbacks in flight yet.
1690
 */
1691
static void __cpuinit
1692
rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptable)
1693 1694 1695 1696 1697 1698 1699
{
	unsigned long flags;
	unsigned long mask;
	struct rcu_data *rdp = rsp->rda[cpu];
	struct rcu_node *rnp = rcu_get_root(rsp);

	/* Set up local state, ensuring consistent view of global state. */
P
Paul E. McKenney 已提交
1700
	raw_spin_lock_irqsave(&rnp->lock, flags);
1701 1702 1703
	rdp->passed_quiesc = 0;  /* We could be racing with new GP, */
	rdp->qs_pending = 1;	 /*  so set up to respond to current GP. */
	rdp->beenonline = 1;	 /* We have now been online. */
1704
	rdp->preemptable = preemptable;
1705 1706
	rdp->qlen_last_fqs_check = 0;
	rdp->n_force_qs_snap = rsp->n_force_qs;
1707
	rdp->blimit = blimit;
P
Paul E. McKenney 已提交
1708
	raw_spin_unlock(&rnp->lock);		/* irqs remain disabled. */
1709 1710 1711 1712 1713 1714 1715

	/*
	 * A new grace period might start here.  If so, we won't be part
	 * of it, but that is OK, as we are currently in a quiescent state.
	 */

	/* Exclude any attempts to start a new GP on large systems. */
P
Paul E. McKenney 已提交
1716
	raw_spin_lock(&rsp->onofflock);		/* irqs already disabled. */
1717 1718 1719 1720 1721 1722

	/* Add CPU to rcu_node bitmasks. */
	rnp = rdp->mynode;
	mask = rdp->grpmask;
	do {
		/* Exclude any attempts to start a new GP on small systems. */
P
Paul E. McKenney 已提交
1723
		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
1724 1725
		rnp->qsmaskinit |= mask;
		mask = rnp->grpmask;
1726 1727 1728 1729 1730
		if (rnp == rdp->mynode) {
			rdp->gpnum = rnp->completed; /* if GP in progress... */
			rdp->completed = rnp->completed;
			rdp->passed_quiesc_completed = rnp->completed - 1;
		}
P
Paul E. McKenney 已提交
1731
		raw_spin_unlock(&rnp->lock); /* irqs already disabled. */
1732 1733 1734
		rnp = rnp->parent;
	} while (rnp != NULL && !(rnp->qsmaskinit & mask));

P
Paul E. McKenney 已提交
1735
	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
1736 1737 1738 1739
}

static void __cpuinit rcu_online_cpu(int cpu)
{
1740 1741 1742
	rcu_init_percpu_data(cpu, &rcu_sched_state, 0);
	rcu_init_percpu_data(cpu, &rcu_bh_state, 0);
	rcu_preempt_init_percpu_data(cpu);
1743 1744 1745
}

/*
1746
 * Handle CPU online/offline notification events.
1747
 */
1748 1749
static int __cpuinit rcu_cpu_notify(struct notifier_block *self,
				    unsigned long action, void *hcpu)
1750 1751 1752 1753 1754 1755 1756 1757
{
	long cpu = (long)hcpu;

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
		rcu_online_cpu(cpu);
		break;
1758 1759 1760
	case CPU_DYING:
	case CPU_DYING_FROZEN:
		/*
1761
		 * preempt_disable() in _rcu_barrier() prevents stop_machine(),
1762
		 * so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);"
1763 1764 1765 1766 1767 1768
		 * returns, all online cpus have queued rcu_barrier_func().
		 * The dying CPU clears its cpu_online_mask bit and
		 * moves all of its RCU callbacks to ->orphan_cbs_list
		 * in the context of stop_machine(), so subsequent calls
		 * to _rcu_barrier() will adopt these callbacks and only
		 * then queue rcu_barrier_func() on all remaining CPUs.
1769
		 */
1770 1771 1772
		rcu_send_cbs_to_orphanage(&rcu_bh_state);
		rcu_send_cbs_to_orphanage(&rcu_sched_state);
		rcu_preempt_send_cbs_to_orphanage();
1773
		break;
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 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:
		rcu_offline_cpu(cpu);
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}

/*
 * Compute the per-level fanout, either using the exact fanout specified
 * or balancing the tree, depending on CONFIG_RCU_FANOUT_EXACT.
 */
#ifdef CONFIG_RCU_FANOUT_EXACT
static void __init rcu_init_levelspread(struct rcu_state *rsp)
{
	int i;

	for (i = NUM_RCU_LVLS - 1; i >= 0; i--)
		rsp->levelspread[i] = CONFIG_RCU_FANOUT;
}
#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */
static void __init rcu_init_levelspread(struct rcu_state *rsp)
{
	int ccur;
	int cprv;
	int i;

	cprv = NR_CPUS;
	for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {
		ccur = rsp->levelcnt[i];
		rsp->levelspread[i] = (cprv + ccur - 1) / ccur;
		cprv = ccur;
	}
}
#endif /* #else #ifdef CONFIG_RCU_FANOUT_EXACT */

/*
 * Helper function for rcu_init() that initializes one rcu_state structure.
 */
static void __init rcu_init_one(struct rcu_state *rsp)
{
1819 1820 1821 1822
	static char *buf[] = { "rcu_node_level_0",
			       "rcu_node_level_1",
			       "rcu_node_level_2",
			       "rcu_node_level_3" };  /* Match MAX_RCU_LVLS */
1823 1824 1825 1826 1827
	int cpustride = 1;
	int i;
	int j;
	struct rcu_node *rnp;

1828 1829
	BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf));  /* Fix buf[] init! */

1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
	/* Initialize the level-tracking arrays. */

	for (i = 1; i < NUM_RCU_LVLS; i++)
		rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];
	rcu_init_levelspread(rsp);

	/* Initialize the elements themselves, starting from the leaves. */

	for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {
		cpustride *= rsp->levelspread[i];
		rnp = rsp->level[i];
		for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {
P
Paul E. McKenney 已提交
1842
			raw_spin_lock_init(&rnp->lock);
1843 1844
			lockdep_set_class_and_name(&rnp->lock,
						   &rcu_node_class[i], buf[i]);
1845
			rnp->gpnum = 0;
1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
			rnp->qsmask = 0;
			rnp->qsmaskinit = 0;
			rnp->grplo = j * cpustride;
			rnp->grphi = (j + 1) * cpustride - 1;
			if (rnp->grphi >= NR_CPUS)
				rnp->grphi = NR_CPUS - 1;
			if (i == 0) {
				rnp->grpnum = 0;
				rnp->grpmask = 0;
				rnp->parent = NULL;
			} else {
				rnp->grpnum = j % rsp->levelspread[i - 1];
				rnp->grpmask = 1UL << rnp->grpnum;
				rnp->parent = rsp->level[i - 1] +
					      j / rsp->levelspread[i - 1];
			}
			rnp->level = i;
1863 1864
			INIT_LIST_HEAD(&rnp->blocked_tasks[0]);
			INIT_LIST_HEAD(&rnp->blocked_tasks[1]);
1865 1866
			INIT_LIST_HEAD(&rnp->blocked_tasks[2]);
			INIT_LIST_HEAD(&rnp->blocked_tasks[3]);
1867 1868
		}
	}
1869 1870 1871 1872 1873 1874 1875 1876

	rnp = rsp->level[NUM_RCU_LVLS - 1];
	for_each_possible_cpu(i) {
		if (i > rnp->grphi)
			rnp++;
		rsp->rda[i]->mynode = rnp;
		rcu_boot_init_percpu_data(i, rsp);
	}
1877 1878 1879
}

/*
1880 1881 1882
 * Helper macro for __rcu_init() and __rcu_init_preempt().  To be used
 * nowhere else!  Assigns leaf node pointers into each CPU's rcu_data
 * structure.
1883
 */
1884
#define RCU_INIT_FLAVOR(rsp, rcu_data) \
1885
do { \
1886 1887
	int i; \
	\
1888 1889 1890
	for_each_possible_cpu(i) { \
		(rsp)->rda[i] = &per_cpu(rcu_data, i); \
	} \
1891
	rcu_init_one(rsp); \
1892 1893
} while (0)

1894
void __init rcu_init(void)
1895
{
P
Paul E. McKenney 已提交
1896
	int cpu;
1897

1898
	rcu_bootup_announce();
1899 1900 1901
#ifdef CONFIG_RCU_CPU_STALL_DETECTOR
	printk(KERN_INFO "RCU-based detection of stalled CPUs is enabled.\n");
#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
1902 1903 1904
#if NUM_RCU_LVL_4 != 0
	printk(KERN_INFO "Experimental four-level hierarchy is enabled.\n");
#endif /* #if NUM_RCU_LVL_4 != 0 */
1905 1906
	RCU_INIT_FLAVOR(&rcu_sched_state, rcu_sched_data);
	RCU_INIT_FLAVOR(&rcu_bh_state, rcu_bh_data);
1907
	__rcu_init_preempt();
1908
	open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
1909 1910 1911 1912 1913 1914 1915

	/*
	 * We don't need protection against CPU-hotplug here because
	 * this is called early in boot, before either interrupts
	 * or the scheduler are operational.
	 */
	cpu_notifier(rcu_cpu_notify, 0);
P
Paul E. McKenney 已提交
1916 1917
	for_each_online_cpu(cpu)
		rcu_cpu_notify(NULL, CPU_UP_PREPARE, (void *)(long)cpu);
1918 1919
}

1920
#include "rcutree_plugin.h"