rcutree.c 90.5 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
#include <linux/atomic.h>
40
#include <linux/bitops.h>
41
#include <linux/export.h>
42 43 44 45 46 47 48
#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>
49
#include <linux/kernel_stat.h>
50 51
#include <linux/wait.h>
#include <linux/kthread.h>
52
#include <linux/prefetch.h>
53 54
#include <linux/delay.h>
#include <linux/stop_machine.h>
55
#include <linux/random.h>
56

57
#include "rcutree.h"
58 59 60
#include <trace/events/rcu.h>

#include "rcu.h"
61

62 63
/* Data structures. */

64
static struct lock_class_key rcu_node_class[RCU_NUM_LVLS];
65
static struct lock_class_key rcu_fqs_class[RCU_NUM_LVLS];
66

67
#define RCU_STATE_INITIALIZER(sname, cr) { \
68
	.level = { &sname##_state.node[0] }, \
69
	.call = cr, \
70
	.fqs_state = RCU_GP_IDLE, \
71 72
	.gpnum = -300, \
	.completed = -300, \
73 74 75
	.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&sname##_state.onofflock), \
	.orphan_nxttail = &sname##_state.orphan_nxtlist, \
	.orphan_donetail = &sname##_state.orphan_donelist, \
76
	.barrier_mutex = __MUTEX_INITIALIZER(sname##_state.barrier_mutex), \
77
	.name = #sname, \
78 79
}

80 81
struct rcu_state rcu_sched_state =
	RCU_STATE_INITIALIZER(rcu_sched, call_rcu_sched);
82
DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
83

84
struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh, call_rcu_bh);
85
DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
86

87
static struct rcu_state *rcu_state;
88
LIST_HEAD(rcu_struct_flavors);
89

90 91
/* Increase (but not decrease) the CONFIG_RCU_FANOUT_LEAF at boot time. */
static int rcu_fanout_leaf = CONFIG_RCU_FANOUT_LEAF;
92
module_param(rcu_fanout_leaf, int, 0444);
93 94 95 96 97 98 99 100 101 102
int rcu_num_lvls __read_mostly = RCU_NUM_LVLS;
static int num_rcu_lvl[] = {  /* Number of rcu_nodes at specified level. */
	NUM_RCU_LVL_0,
	NUM_RCU_LVL_1,
	NUM_RCU_LVL_2,
	NUM_RCU_LVL_3,
	NUM_RCU_LVL_4,
};
int rcu_num_nodes __read_mostly = NUM_RCU_NODES; /* Total # rcu_nodes in use. */

103 104 105 106 107 108 109 110 111
/*
 * The rcu_scheduler_active variable transitions from zero to one just
 * before the first task is spawned.  So when this variable is zero, RCU
 * can assume that there is but one task, allowing RCU to (for example)
 * optimized synchronize_sched() to a simple barrier().  When this variable
 * is one, RCU must actually do all the hard work required to detect real
 * grace periods.  This variable is also used to suppress boot-time false
 * positives from lockdep-RCU error checking.
 */
112 113 114
int rcu_scheduler_active __read_mostly;
EXPORT_SYMBOL_GPL(rcu_scheduler_active);

115 116 117 118 119 120 121 122 123 124 125 126 127 128
/*
 * The rcu_scheduler_fully_active variable transitions from zero to one
 * during the early_initcall() processing, which is after the scheduler
 * is capable of creating new tasks.  So RCU processing (for example,
 * creating tasks for RCU priority boosting) must be delayed until after
 * rcu_scheduler_fully_active transitions from zero to one.  We also
 * currently delay invocation of any RCU callbacks until after this point.
 *
 * It might later prove better for people registering RCU callbacks during
 * early boot to take responsibility for these callbacks, but one step at
 * a time.
 */
static int rcu_scheduler_fully_active __read_mostly;

129 130
#ifdef CONFIG_RCU_BOOST

131 132 133 134 135
/*
 * 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);
136
DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
137
DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);
138
DEFINE_PER_CPU(char, rcu_cpu_has_work);
139

140 141
#endif /* #ifdef CONFIG_RCU_BOOST */

T
Thomas Gleixner 已提交
142
static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu);
143 144
static void invoke_rcu_core(void);
static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp);
145

146 147 148 149 150 151 152 153 154 155 156 157
/*
 * Track the rcutorture test sequence number and the update version
 * number within a given test.  The rcutorture_testseq is incremented
 * on every rcutorture module load and unload, so has an odd value
 * when a test is running.  The rcutorture_vernum is set to zero
 * when rcutorture starts and is incremented on each rcutorture update.
 * These variables enable correlating rcutorture output with the
 * RCU tracing information.
 */
unsigned long rcutorture_testseq;
unsigned long rcutorture_vernum;

158 159 160 161 162 163 164 165 166 167
/*
 * 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);
}

168
/*
169
 * Note a quiescent state.  Because we do not need to know
170
 * how many quiescent states passed, just if there was at least
171
 * one since the start of the grace period, this just sets a flag.
172
 * The caller must have disabled preemption.
173
 */
174
void rcu_sched_qs(int cpu)
175
{
176
	struct rcu_data *rdp = &per_cpu(rcu_sched_data, cpu);
177

178
	if (rdp->passed_quiesce == 0)
179
		trace_rcu_grace_period("rcu_sched", rdp->gpnum, "cpuqs");
180
	rdp->passed_quiesce = 1;
181 182
}

183
void rcu_bh_qs(int cpu)
184
{
185
	struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);
186

187
	if (rdp->passed_quiesce == 0)
188
		trace_rcu_grace_period("rcu_bh", rdp->gpnum, "cpuqs");
189
	rdp->passed_quiesce = 1;
190
}
191

192 193 194
/*
 * Note a context switch.  This is a quiescent state for RCU-sched,
 * and requires special handling for preemptible RCU.
195
 * The caller must have disabled preemption.
196 197 198
 */
void rcu_note_context_switch(int cpu)
{
199
	trace_rcu_utilization("Start context switch");
200
	rcu_sched_qs(cpu);
201
	rcu_preempt_note_context_switch(cpu);
202
	trace_rcu_utilization("End context switch");
203
}
204
EXPORT_SYMBOL_GPL(rcu_note_context_switch);
205

206
DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
207
	.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,
208
	.dynticks = ATOMIC_INIT(1),
209 210 211
#ifdef CONFIG_RCU_USER_QS
	.ignore_user_qs = true,
#endif
212
};
213

214
static int blimit = 10;		/* Maximum callbacks per rcu_do_batch. */
215 216 217
static int qhimark = 10000;	/* If this many pending, ignore blimit. */
static int qlowmark = 100;	/* Once only this many pending, use blimit. */

218 219 220
module_param(blimit, int, 0444);
module_param(qhimark, int, 0444);
module_param(qlowmark, int, 0444);
221

222 223 224
int rcu_cpu_stall_suppress __read_mostly; /* 1 = suppress stall warnings. */
int rcu_cpu_stall_timeout __read_mostly = CONFIG_RCU_CPU_STALL_TIMEOUT;

225
module_param(rcu_cpu_stall_suppress, int, 0644);
226
module_param(rcu_cpu_stall_timeout, int, 0644);
227

228 229 230 231 232 233
static ulong jiffies_till_first_fqs = RCU_JIFFIES_TILL_FORCE_QS;
static ulong jiffies_till_next_fqs = RCU_JIFFIES_TILL_FORCE_QS;

module_param(jiffies_till_first_fqs, ulong, 0644);
module_param(jiffies_till_next_fqs, ulong, 0644);

234 235
static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *));
static void force_quiescent_state(struct rcu_state *rsp);
236
static int rcu_pending(int cpu);
237 238

/*
239
 * Return the number of RCU-sched batches processed thus far for debug & stats.
240
 */
241
long rcu_batches_completed_sched(void)
242
{
243
	return rcu_sched_state.completed;
244
}
245
EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);
246 247 248 249 250 251 252 253 254 255

/*
 * 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);

256 257 258 259 260
/*
 * Force a quiescent state for RCU BH.
 */
void rcu_bh_force_quiescent_state(void)
{
261
	force_quiescent_state(&rcu_bh_state);
262 263 264
}
EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289
/*
 * Record the number of times rcutorture tests have been initiated and
 * terminated.  This information allows the debugfs tracing stats to be
 * correlated to the rcutorture messages, even when the rcutorture module
 * is being repeatedly loaded and unloaded.  In other words, we cannot
 * store this state in rcutorture itself.
 */
void rcutorture_record_test_transition(void)
{
	rcutorture_testseq++;
	rcutorture_vernum = 0;
}
EXPORT_SYMBOL_GPL(rcutorture_record_test_transition);

/*
 * Record the number of writer passes through the current rcutorture test.
 * This is also used to correlate debugfs tracing stats with the rcutorture
 * messages.
 */
void rcutorture_record_progress(unsigned long vernum)
{
	rcutorture_vernum++;
}
EXPORT_SYMBOL_GPL(rcutorture_record_progress);

290 291 292 293 294
/*
 * Force a quiescent state for RCU-sched.
 */
void rcu_sched_force_quiescent_state(void)
{
295
	force_quiescent_state(&rcu_sched_state);
296 297 298
}
EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

299 300 301 302 303 304 305 306 307 308 309 310 311 312 313
/*
 * 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)
{
314 315 316
	return *rdp->nxttail[RCU_DONE_TAIL +
			     ACCESS_ONCE(rsp->completed) != rdp->completed] &&
	       !rcu_gp_in_progress(rsp);
317 318 319 320 321 322 323 324 325 326
}

/*
 * 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];
}

327
/*
328
 * rcu_eqs_enter_common - current CPU is moving towards extended quiescent state
329 330 331 332 333
 *
 * If the new value of the ->dynticks_nesting counter now is zero,
 * we really have entered idle, and must do the appropriate accounting.
 * The caller must have disabled interrupts.
 */
334 335
static void rcu_eqs_enter_common(struct rcu_dynticks *rdtp, long long oldval,
				bool user)
336
{
337
	trace_rcu_dyntick("Start", oldval, 0);
338
	if (!is_idle_task(current) && !user) {
339 340
		struct task_struct *idle = idle_task(smp_processor_id());

341
		trace_rcu_dyntick("Error on entry: not idle task", oldval, 0);
342
		ftrace_dump(DUMP_ORIG);
343 344 345
		WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",
			  current->pid, current->comm,
			  idle->pid, idle->comm); /* must be idle task! */
346
	}
347
	rcu_prepare_for_idle(smp_processor_id());
348 349 350 351 352
	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
	smp_mb__before_atomic_inc();  /* See above. */
	atomic_inc(&rdtp->dynticks);
	smp_mb__after_atomic_inc();  /* Force ordering with next sojourn. */
	WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);
353 354

	/*
355
	 * It is illegal to enter an extended quiescent state while
356 357 358 359 360 361 362 363
	 * in an RCU read-side critical section.
	 */
	rcu_lockdep_assert(!lock_is_held(&rcu_lock_map),
			   "Illegal idle entry in RCU read-side critical section.");
	rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map),
			   "Illegal idle entry in RCU-bh read-side critical section.");
	rcu_lockdep_assert(!lock_is_held(&rcu_sched_lock_map),
			   "Illegal idle entry in RCU-sched read-side critical section.");
364
}
365

366 367 368
/*
 * Enter an RCU extended quiescent state, which can be either the
 * idle loop or adaptive-tickless usermode execution.
369
 */
370
static void rcu_eqs_enter(bool user)
371
{
372
	long long oldval;
373 374 375
	struct rcu_dynticks *rdtp;

	rdtp = &__get_cpu_var(rcu_dynticks);
376
	oldval = rdtp->dynticks_nesting;
377 378 379 380 381
	WARN_ON_ONCE((oldval & DYNTICK_TASK_NEST_MASK) == 0);
	if ((oldval & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE)
		rdtp->dynticks_nesting = 0;
	else
		rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE;
382
	rcu_eqs_enter_common(rdtp, oldval, user);
383
}
384 385 386 387 388 389 390 391 392 393 394 395 396 397 398

/**
 * rcu_idle_enter - inform RCU that current CPU is entering idle
 *
 * Enter idle 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 idle, a possibility
 * handled by irq_enter() and irq_exit().)
 *
 * We crowbar the ->dynticks_nesting field to zero to allow for
 * the possibility of usermode upcalls having messed up our count
 * of interrupt nesting level during the prior busy period.
 */
void rcu_idle_enter(void)
{
399 400 401
	unsigned long flags;

	local_irq_save(flags);
402
	rcu_eqs_enter(0);
403
	local_irq_restore(flags);
404
}
405
EXPORT_SYMBOL_GPL(rcu_idle_enter);
406

407
#ifdef CONFIG_RCU_USER_QS
408 409 410 411 412 413 414 415 416 417
/**
 * rcu_user_enter - inform RCU that we are resuming userspace.
 *
 * Enter RCU idle mode right before resuming userspace.  No use of RCU
 * is permitted between this call and rcu_user_exit(). This way the
 * CPU doesn't need to maintain the tick for RCU maintenance purposes
 * when the CPU runs in userspace.
 */
void rcu_user_enter(void)
{
418 419 420
	unsigned long flags;
	struct rcu_dynticks *rdtp;

421 422 423 424 425 426 427 428 429 430 431
	/*
	 * Some contexts may involve an exception occuring in an irq,
	 * leading to that nesting:
	 * rcu_irq_enter() rcu_user_exit() rcu_user_exit() rcu_irq_exit()
	 * This would mess up the dyntick_nesting count though. And rcu_irq_*()
	 * helpers are enough to protect RCU uses inside the exception. So
	 * just return immediately if we detect we are in an IRQ.
	 */
	if (in_interrupt())
		return;

432 433 434 435
	WARN_ON_ONCE(!current->mm);

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
436
	if (!rdtp->ignore_user_qs && !rdtp->in_user) {
437 438 439 440
		rdtp->in_user = true;
		rcu_eqs_enter(1);
	}
	local_irq_restore(flags);
441 442
}

443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
/**
 * rcu_user_enter_after_irq - inform RCU that we are going to resume userspace
 * after the current irq returns.
 *
 * This is similar to rcu_user_enter() but in the context of a non-nesting
 * irq. After this call, RCU enters into idle mode when the interrupt
 * returns.
 */
void rcu_user_enter_after_irq(void)
{
	unsigned long flags;
	struct rcu_dynticks *rdtp;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
	/* Ensure this irq is interrupting a non-idle RCU state.  */
	WARN_ON_ONCE(!(rdtp->dynticks_nesting & DYNTICK_TASK_MASK));
	rdtp->dynticks_nesting = 1;
	local_irq_restore(flags);
}
463
#endif /* CONFIG_RCU_USER_QS */
464

465 466 467 468 469 470
/**
 * rcu_irq_exit - inform RCU that current CPU is exiting irq towards idle
 *
 * Exit from an interrupt handler, which might possibly result in entering
 * idle mode, in other words, leaving the mode in which read-side critical
 * sections can occur.
471
 *
472 473 474 475 476 477 478 479
 * This code assumes that the idle loop never does anything that might
 * result in unbalanced calls to irq_enter() and irq_exit().  If your
 * architecture violates this assumption, RCU will give you what you
 * deserve, good and hard.  But very infrequently and irreproducibly.
 *
 * Use things like work queues to work around this limitation.
 *
 * You have been warned.
480
 */
481
void rcu_irq_exit(void)
482 483
{
	unsigned long flags;
484
	long long oldval;
485 486 487 488
	struct rcu_dynticks *rdtp;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
489
	oldval = rdtp->dynticks_nesting;
490 491
	rdtp->dynticks_nesting--;
	WARN_ON_ONCE(rdtp->dynticks_nesting < 0);
492 493 494
	if (rdtp->dynticks_nesting)
		trace_rcu_dyntick("--=", oldval, rdtp->dynticks_nesting);
	else
495
		rcu_eqs_enter_common(rdtp, oldval, 1);
496 497 498 499
	local_irq_restore(flags);
}

/*
500
 * rcu_eqs_exit_common - current CPU moving away from extended quiescent state
501 502 503 504 505
 *
 * If the new value of the ->dynticks_nesting counter was previously zero,
 * we really have exited idle, and must do the appropriate accounting.
 * The caller must have disabled interrupts.
 */
506 507
static void rcu_eqs_exit_common(struct rcu_dynticks *rdtp, long long oldval,
			       int user)
508
{
509 510 511 512 513
	smp_mb__before_atomic_inc();  /* Force ordering w/previous sojourn. */
	atomic_inc(&rdtp->dynticks);
	/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */
	smp_mb__after_atomic_inc();  /* See above. */
	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
514
	rcu_cleanup_after_idle(smp_processor_id());
515
	trace_rcu_dyntick("End", oldval, rdtp->dynticks_nesting);
516
	if (!is_idle_task(current) && !user) {
517 518
		struct task_struct *idle = idle_task(smp_processor_id());

519 520
		trace_rcu_dyntick("Error on exit: not idle task",
				  oldval, rdtp->dynticks_nesting);
521
		ftrace_dump(DUMP_ORIG);
522 523 524
		WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",
			  current->pid, current->comm,
			  idle->pid, idle->comm); /* must be idle task! */
525 526 527
	}
}

528 529 530
/*
 * Exit an RCU extended quiescent state, which can be either the
 * idle loop or adaptive-tickless usermode execution.
531
 */
532
static void rcu_eqs_exit(bool user)
533 534 535 536 537 538
{
	struct rcu_dynticks *rdtp;
	long long oldval;

	rdtp = &__get_cpu_var(rcu_dynticks);
	oldval = rdtp->dynticks_nesting;
539 540 541 542 543
	WARN_ON_ONCE(oldval < 0);
	if (oldval & DYNTICK_TASK_NEST_MASK)
		rdtp->dynticks_nesting += DYNTICK_TASK_NEST_VALUE;
	else
		rdtp->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
544
	rcu_eqs_exit_common(rdtp, oldval, user);
545
}
546 547 548 549 550 551 552 553 554 555 556 557 558 559

/**
 * rcu_idle_exit - inform RCU that current CPU is leaving idle
 *
 * Exit idle mode, in other words, -enter- the mode in which RCU
 * read-side critical sections can occur.
 *
 * We crowbar the ->dynticks_nesting field to DYNTICK_TASK_NEST to
 * allow for the possibility of usermode upcalls messing up our count
 * of interrupt nesting level during the busy period that is just
 * now starting.
 */
void rcu_idle_exit(void)
{
560 561 562
	unsigned long flags;

	local_irq_save(flags);
563
	rcu_eqs_exit(0);
564
	local_irq_restore(flags);
565
}
566
EXPORT_SYMBOL_GPL(rcu_idle_exit);
567

568
#ifdef CONFIG_RCU_USER_QS
569 570 571 572 573 574 575 576
/**
 * rcu_user_exit - inform RCU that we are exiting userspace.
 *
 * Exit RCU idle mode while entering the kernel because it can
 * run a RCU read side critical section anytime.
 */
void rcu_user_exit(void)
{
577 578 579
	unsigned long flags;
	struct rcu_dynticks *rdtp;

580 581 582 583 584 585 586 587 588 589 590
	/*
	 * Some contexts may involve an exception occuring in an irq,
	 * leading to that nesting:
	 * rcu_irq_enter() rcu_user_exit() rcu_user_exit() rcu_irq_exit()
	 * This would mess up the dyntick_nesting count though. And rcu_irq_*()
	 * helpers are enough to protect RCU uses inside the exception. So
	 * just return immediately if we detect we are in an IRQ.
	 */
	if (in_interrupt())
		return;

591 592 593 594 595 596 597
	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
	if (rdtp->in_user) {
		rdtp->in_user = false;
		rcu_eqs_exit(1);
	}
	local_irq_restore(flags);
598 599
}

600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
/**
 * rcu_user_exit_after_irq - inform RCU that we won't resume to userspace
 * idle mode after the current non-nesting irq returns.
 *
 * This is similar to rcu_user_exit() but in the context of an irq.
 * This is called when the irq has interrupted a userspace RCU idle mode
 * context. When the current non-nesting interrupt returns after this call,
 * the CPU won't restore the RCU idle mode.
 */
void rcu_user_exit_after_irq(void)
{
	unsigned long flags;
	struct rcu_dynticks *rdtp;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
	/* Ensure we are interrupting an RCU idle mode. */
	WARN_ON_ONCE(rdtp->dynticks_nesting & DYNTICK_TASK_NEST_MASK);
	rdtp->dynticks_nesting += DYNTICK_TASK_EXIT_IDLE;
	local_irq_restore(flags);
}
621
#endif /* CONFIG_RCU_USER_QS */
622

623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652
/**
 * rcu_irq_enter - inform RCU that current CPU is entering irq away from idle
 *
 * Enter an interrupt handler, which might possibly result in exiting
 * idle mode, in other words, entering the mode in which read-side critical
 * sections can occur.
 *
 * Note that the Linux kernel is fully capable of entering an interrupt
 * handler that it never exits, for example when doing upcalls to
 * user mode!  This code assumes that the idle loop never does upcalls to
 * user mode.  If your architecture does do upcalls from the idle loop (or
 * does anything else that results in unbalanced calls to the irq_enter()
 * and irq_exit() functions), RCU will give you what you deserve, good
 * and hard.  But very infrequently and irreproducibly.
 *
 * Use things like work queues to work around this limitation.
 *
 * You have been warned.
 */
void rcu_irq_enter(void)
{
	unsigned long flags;
	struct rcu_dynticks *rdtp;
	long long oldval;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
	oldval = rdtp->dynticks_nesting;
	rdtp->dynticks_nesting++;
	WARN_ON_ONCE(rdtp->dynticks_nesting == 0);
653 654 655
	if (oldval)
		trace_rcu_dyntick("++=", oldval, rdtp->dynticks_nesting);
	else
656
		rcu_eqs_exit_common(rdtp, oldval, 1);
657 658 659 660 661 662 663 664 665 666 667 668 669 670
	local_irq_restore(flags);
}

/**
 * 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);

671 672
	if (rdtp->dynticks_nmi_nesting == 0 &&
	    (atomic_read(&rdtp->dynticks) & 0x1))
673
		return;
674 675 676 677 678 679
	rdtp->dynticks_nmi_nesting++;
	smp_mb__before_atomic_inc();  /* Force delay from prior write. */
	atomic_inc(&rdtp->dynticks);
	/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */
	smp_mb__after_atomic_inc();  /* See above. */
	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
680 681 682 683 684 685 686 687 688 689 690 691 692
}

/**
 * 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);

693 694
	if (rdtp->dynticks_nmi_nesting == 0 ||
	    --rdtp->dynticks_nmi_nesting != 0)
695
		return;
696 697 698 699 700
	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
	smp_mb__before_atomic_inc();  /* See above. */
	atomic_inc(&rdtp->dynticks);
	smp_mb__after_atomic_inc();  /* Force delay to next write. */
	WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);
701 702 703
}

/**
704
 * rcu_is_cpu_idle - see if RCU thinks that the current CPU is idle
705
 *
706
 * If the current CPU is in its idle loop and is neither in an interrupt
707
 * or NMI handler, return true.
708
 */
709
int rcu_is_cpu_idle(void)
710
{
711 712 713 714 715 716
	int ret;

	preempt_disable();
	ret = (atomic_read(&__get_cpu_var(rcu_dynticks).dynticks) & 0x1) == 0;
	preempt_enable();
	return ret;
717
}
718
EXPORT_SYMBOL(rcu_is_cpu_idle);
719

720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
#ifdef CONFIG_RCU_USER_QS
void rcu_user_hooks_switch(struct task_struct *prev,
			   struct task_struct *next)
{
	struct rcu_dynticks *rdtp;

	/* Interrupts are disabled in context switch */
	rdtp = &__get_cpu_var(rcu_dynticks);
	if (!rdtp->ignore_user_qs) {
		clear_tsk_thread_flag(prev, TIF_NOHZ);
		set_tsk_thread_flag(next, TIF_NOHZ);
	}
}
#endif /* #ifdef CONFIG_RCU_USER_QS */

735
#if defined(CONFIG_PROVE_RCU) && defined(CONFIG_HOTPLUG_CPU)
736 737 738 739 740 741 742

/*
 * Is the current CPU online?  Disable preemption to avoid false positives
 * that could otherwise happen due to the current CPU number being sampled,
 * this task being preempted, its old CPU being taken offline, resuming
 * on some other CPU, then determining that its old CPU is now offline.
 * It is OK to use RCU on an offline processor during initial boot, hence
743 744 745 746 747 748 749 750 751 752 753
 * the check for rcu_scheduler_fully_active.  Note also that it is OK
 * for a CPU coming online to use RCU for one jiffy prior to marking itself
 * online in the cpu_online_mask.  Similarly, it is OK for a CPU going
 * offline to continue to use RCU for one jiffy after marking itself
 * offline in the cpu_online_mask.  This leniency is necessary given the
 * non-atomic nature of the online and offline processing, for example,
 * the fact that a CPU enters the scheduler after completing the CPU_DYING
 * notifiers.
 *
 * This is also why RCU internally marks CPUs online during the
 * CPU_UP_PREPARE phase and offline during the CPU_DEAD phase.
754 755 756 757 758 759
 *
 * Disable checking if in an NMI handler because we cannot safely report
 * errors from NMI handlers anyway.
 */
bool rcu_lockdep_current_cpu_online(void)
{
760 761
	struct rcu_data *rdp;
	struct rcu_node *rnp;
762 763 764 765 766
	bool ret;

	if (in_nmi())
		return 1;
	preempt_disable();
767 768 769
	rdp = &__get_cpu_var(rcu_sched_data);
	rnp = rdp->mynode;
	ret = (rdp->grpmask & rnp->qsmaskinit) ||
770 771 772 773 774 775
	      !rcu_scheduler_fully_active;
	preempt_enable();
	return ret;
}
EXPORT_SYMBOL_GPL(rcu_lockdep_current_cpu_online);

776
#endif /* #if defined(CONFIG_PROVE_RCU) && defined(CONFIG_HOTPLUG_CPU) */
777

778
/**
779
 * rcu_is_cpu_rrupt_from_idle - see if idle or immediately interrupted from idle
780
 *
781 782 783
 * If the current CPU is idle or running at a first-level (not nested)
 * interrupt from idle, return true.  The caller must have at least
 * disabled preemption.
784
 */
785
int rcu_is_cpu_rrupt_from_idle(void)
786
{
787
	return __get_cpu_var(rcu_dynticks).dynticks_nesting <= 1;
788 789 790 791 792
}

/*
 * Snapshot the specified CPU's dynticks counter so that we can later
 * credit them with an implicit quiescent state.  Return 1 if this CPU
793
 * is in dynticks idle mode, which is an extended quiescent state.
794 795 796
 */
static int dyntick_save_progress_counter(struct rcu_data *rdp)
{
797
	rdp->dynticks_snap = atomic_add_return(0, &rdp->dynticks->dynticks);
798
	return (rdp->dynticks_snap & 0x1) == 0;
799 800 801 802 803 804
}

/*
 * 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()
805
 * for this same CPU, or by virtue of having been offline.
806 807 808
 */
static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
{
809 810
	unsigned int curr;
	unsigned int snap;
811

812 813
	curr = (unsigned int)atomic_add_return(0, &rdp->dynticks->dynticks);
	snap = (unsigned int)rdp->dynticks_snap;
814 815 816 817 818 819 820 821 822

	/*
	 * 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.
	 */
823
	if ((curr & 0x1) == 0 || UINT_CMP_GE(curr, snap + 2)) {
824
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "dti");
825 826 827 828
		rdp->dynticks_fqs++;
		return 1;
	}

829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
	/*
	 * Check for the CPU being offline, but only if the grace period
	 * is old enough.  We don't need to worry about the CPU changing
	 * state: If we see it offline even once, it has been through a
	 * quiescent state.
	 *
	 * The reason for insisting that the grace period be at least
	 * one jiffy old is that CPUs that are not quite online and that
	 * have just gone offline can still execute RCU read-side critical
	 * sections.
	 */
	if (ULONG_CMP_GE(rdp->rsp->gp_start + 2, jiffies))
		return 0;  /* Grace period is not old enough. */
	barrier();
	if (cpu_is_offline(rdp->cpu)) {
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "ofl");
		rdp->offline_fqs++;
		return 1;
	}
	return 0;
849 850
}

851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868
static int jiffies_till_stall_check(void)
{
	int till_stall_check = ACCESS_ONCE(rcu_cpu_stall_timeout);

	/*
	 * Limit check must be consistent with the Kconfig limits
	 * for CONFIG_RCU_CPU_STALL_TIMEOUT.
	 */
	if (till_stall_check < 3) {
		ACCESS_ONCE(rcu_cpu_stall_timeout) = 3;
		till_stall_check = 3;
	} else if (till_stall_check > 300) {
		ACCESS_ONCE(rcu_cpu_stall_timeout) = 300;
		till_stall_check = 300;
	}
	return till_stall_check * HZ + RCU_STALL_DELAY_DELTA;
}

869 870 871
static void record_gp_stall_check_time(struct rcu_state *rsp)
{
	rsp->gp_start = jiffies;
872
	rsp->jiffies_stall = jiffies + jiffies_till_stall_check();
873 874 875 876 877 878 879
}

static void print_other_cpu_stall(struct rcu_state *rsp)
{
	int cpu;
	long delta;
	unsigned long flags;
880
	int ndetected = 0;
881 882 883 884
	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
885
	raw_spin_lock_irqsave(&rnp->lock, flags);
886
	delta = jiffies - rsp->jiffies_stall;
887
	if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {
P
Paul E. McKenney 已提交
888
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
889 890
		return;
	}
891
	rsp->jiffies_stall = jiffies + 3 * jiffies_till_stall_check() + 3;
P
Paul E. McKenney 已提交
892
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
893

894 895 896 897 898
	/*
	 * OK, time to rat on our buddy...
	 * See Documentation/RCU/stallwarn.txt for info on how to debug
	 * RCU CPU stall warnings.
	 */
899
	printk(KERN_ERR "INFO: %s detected stalls on CPUs/tasks:",
900
	       rsp->name);
901
	print_cpu_stall_info_begin();
902
	rcu_for_each_leaf_node(rsp, rnp) {
903
		raw_spin_lock_irqsave(&rnp->lock, flags);
904
		ndetected += rcu_print_task_stall(rnp);
905 906 907 908 909 910 911 912
		if (rnp->qsmask != 0) {
			for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)
				if (rnp->qsmask & (1UL << cpu)) {
					print_cpu_stall_info(rsp,
							     rnp->grplo + cpu);
					ndetected++;
				}
		}
913
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
914
	}
915 916 917 918 919 920 921

	/*
	 * Now rat on any tasks that got kicked up to the root rcu_node
	 * due to CPU offlining.
	 */
	rnp = rcu_get_root(rsp);
	raw_spin_lock_irqsave(&rnp->lock, flags);
922
	ndetected += rcu_print_task_stall(rnp);
923 924 925 926
	raw_spin_unlock_irqrestore(&rnp->lock, flags);

	print_cpu_stall_info_end();
	printk(KERN_CONT "(detected by %d, t=%ld jiffies)\n",
927
	       smp_processor_id(), (long)(jiffies - rsp->gp_start));
928 929 930
	if (ndetected == 0)
		printk(KERN_ERR "INFO: Stall ended before state dump start\n");
	else if (!trigger_all_cpu_backtrace())
931
		dump_stack();
932

933
	/* Complain about tasks blocking the grace period. */
934 935 936

	rcu_print_detail_task_stall(rsp);

937
	force_quiescent_state(rsp);  /* Kick them all. */
938 939 940 941 942 943 944
}

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

945 946 947 948 949
	/*
	 * OK, time to rat on ourselves...
	 * See Documentation/RCU/stallwarn.txt for info on how to debug
	 * RCU CPU stall warnings.
	 */
950 951 952 953 954
	printk(KERN_ERR "INFO: %s self-detected stall on CPU", rsp->name);
	print_cpu_stall_info_begin();
	print_cpu_stall_info(rsp, smp_processor_id());
	print_cpu_stall_info_end();
	printk(KERN_CONT " (t=%lu jiffies)\n", jiffies - rsp->gp_start);
955 956
	if (!trigger_all_cpu_backtrace())
		dump_stack();
957

P
Paul E. McKenney 已提交
958
	raw_spin_lock_irqsave(&rnp->lock, flags);
959
	if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))
960 961
		rsp->jiffies_stall = jiffies +
				     3 * jiffies_till_stall_check() + 3;
P
Paul E. McKenney 已提交
962
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
963

964 965 966 967 968
	set_need_resched();  /* kick ourselves to get things going. */
}

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
{
969 970
	unsigned long j;
	unsigned long js;
971 972
	struct rcu_node *rnp;

973
	if (rcu_cpu_stall_suppress)
974
		return;
975 976
	j = ACCESS_ONCE(jiffies);
	js = ACCESS_ONCE(rsp->jiffies_stall);
977
	rnp = rdp->mynode;
978 979
	if (rcu_gp_in_progress(rsp) &&
	    (ACCESS_ONCE(rnp->qsmask) & rdp->grpmask) && ULONG_CMP_GE(j, js)) {
980 981 982 983

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

984 985
	} else if (rcu_gp_in_progress(rsp) &&
		   ULONG_CMP_GE(j, js + RCU_STALL_RAT_DELAY)) {
986

987
		/* They had a few time units to dump stack, so complain. */
988 989 990 991
		print_other_cpu_stall(rsp);
	}
}

992 993
static int rcu_panic(struct notifier_block *this, unsigned long ev, void *ptr)
{
994
	rcu_cpu_stall_suppress = 1;
995 996 997
	return NOTIFY_DONE;
}

998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
/**
 * rcu_cpu_stall_reset - prevent further stall warnings in current grace period
 *
 * Set the stall-warning timeout way off into the future, thus preventing
 * any RCU CPU stall-warning messages from appearing in the current set of
 * RCU grace periods.
 *
 * The caller must disable hard irqs.
 */
void rcu_cpu_stall_reset(void)
{
1009 1010 1011 1012
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
		rsp->jiffies_stall = jiffies + ULONG_MAX / 2;
1013 1014
}

1015 1016 1017 1018 1019 1020 1021 1022 1023
static struct notifier_block rcu_panic_block = {
	.notifier_call = rcu_panic,
};

static void __init check_cpu_stall_init(void)
{
	atomic_notifier_chain_register(&panic_notifier_list, &rcu_panic_block);
}

1024 1025 1026
/*
 * 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
1027 1028 1029
 * 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.
1030
 */
1031 1032 1033
static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
{
	if (rdp->gpnum != rnp->gpnum) {
1034 1035 1036 1037 1038
		/*
		 * If the current grace period is waiting for this CPU,
		 * set up to detect a quiescent state, otherwise don't
		 * go looking for one.
		 */
1039
		rdp->gpnum = rnp->gpnum;
1040
		trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpustart");
1041 1042
		rdp->passed_quiesce = 0;
		rdp->qs_pending = !!(rnp->qsmask & rdp->grpmask);
1043
		zero_cpu_stall_ticks(rdp);
1044 1045 1046
	}
}

1047 1048
static void note_new_gpnum(struct rcu_state *rsp, struct rcu_data *rdp)
{
1049 1050 1051 1052 1053 1054
	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
1055
	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
1056 1057 1058 1059
		local_irq_restore(flags);
		return;
	}
	__note_new_gpnum(rsp, rnp, rdp);
P
Paul E. McKenney 已提交
1060
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
}

/*
 * 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;
}

1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
/*
 * Initialize the specified rcu_data structure's callback list to empty.
 */
static void init_callback_list(struct rcu_data *rdp)
{
	int i;

	rdp->nxtlist = NULL;
	for (i = 0; i < RCU_NEXT_SIZE; i++)
		rdp->nxttail[i] = &rdp->nxtlist;
}

1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113
/*
 * 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;
1114
		trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuend");
1115

1116 1117
		/*
		 * If we were in an extended quiescent state, we may have
1118
		 * missed some grace periods that others CPUs handled on
1119
		 * our behalf. Catch up with this state to avoid noting
1120 1121
		 * spurious new grace periods.  If another grace period
		 * has started, then rnp->gpnum will have advanced, so
1122 1123
		 * we will detect this later on.  Of course, any quiescent
		 * states we found for the old GP are now invalid.
1124
		 */
1125
		if (ULONG_CMP_LT(rdp->gpnum, rdp->completed)) {
1126
			rdp->gpnum = rdp->completed;
1127 1128
			rdp->passed_quiesce = 0;
		}
1129

1130
		/*
1131 1132
		 * If RCU does not need a quiescent state from this CPU,
		 * then make sure that this CPU doesn't go looking for one.
1133
		 */
1134
		if ((rnp->qsmask & rdp->grpmask) == 0)
1135
			rdp->qs_pending = 0;
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
	}
}

/*
 * 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
1153
	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
1154 1155 1156 1157
		local_irq_restore(flags);
		return;
	}
	__rcu_process_gp_end(rsp, rnp, rdp);
P
Paul E. McKenney 已提交
1158
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171
}

/*
 * 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);

1172 1173
	/* Set state so that this CPU will detect the next quiescent state. */
	__note_new_gpnum(rsp, rnp, rdp);
1174 1175
}

1176
/*
1177
 * Initialize a new grace period.
1178
 */
1179
static int rcu_gp_init(struct rcu_state *rsp)
1180 1181
{
	struct rcu_data *rdp;
1182
	struct rcu_node *rnp = rcu_get_root(rsp);
1183

1184
	raw_spin_lock_irq(&rnp->lock);
1185
	rsp->gp_flags = 0; /* Clear all flags: New grace period. */
1186

1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
	if (rcu_gp_in_progress(rsp)) {
		/* Grace period already in progress, don't start another.  */
		raw_spin_unlock_irq(&rnp->lock);
		return 0;
	}

	/* Advance to a new grace period and initialize state. */
	rsp->gpnum++;
	trace_rcu_grace_period(rsp->name, rsp->gpnum, "start");
	record_gp_stall_check_time(rsp);
	raw_spin_unlock_irq(&rnp->lock);

	/* Exclude any concurrent CPU-hotplug operations. */
	get_online_cpus();

	/*
	 * 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, relying on the layout
	 * of the tree within the rsp->node[] array.  Note that other CPUs
	 * will access only the leaves of the hierarchy, thus seeing 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.
	 *
	 * The grace period cannot complete until the initialization
	 * process finishes, because this kthread handles both.
	 */
	rcu_for_each_node_breadth_first(rsp, rnp) {
1216 1217
		raw_spin_lock_irq(&rnp->lock);
		rdp = this_cpu_ptr(rsp->rda);
1218 1219 1220
		rcu_preempt_check_blocked_tasks(rnp);
		rnp->qsmask = rnp->qsmaskinit;
		rnp->gpnum = rsp->gpnum;
1221
		WARN_ON_ONCE(rnp->completed != rsp->completed);
1222 1223 1224 1225 1226 1227 1228 1229
		rnp->completed = rsp->completed;
		if (rnp == rdp->mynode)
			rcu_start_gp_per_cpu(rsp, rnp, rdp);
		rcu_preempt_boost_start_gp(rnp);
		trace_rcu_grace_period_init(rsp->name, rnp->gpnum,
					    rnp->level, rnp->grplo,
					    rnp->grphi, rnp->qsmask);
		raw_spin_unlock_irq(&rnp->lock);
1230 1231 1232 1233
#ifdef CONFIG_PROVE_RCU_DELAY
		if ((random32() % (rcu_num_nodes * 8)) == 0)
			schedule_timeout_uninterruptible(2);
#endif /* #ifdef CONFIG_PROVE_RCU_DELAY */
1234 1235
		cond_resched();
	}
1236

1237 1238 1239
	put_online_cpus();
	return 1;
}
1240

1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
/*
 * Do one round of quiescent-state forcing.
 */
int rcu_gp_fqs(struct rcu_state *rsp, int fqs_state_in)
{
	int fqs_state = fqs_state_in;
	struct rcu_node *rnp = rcu_get_root(rsp);

	rsp->n_force_qs++;
	if (fqs_state == RCU_SAVE_DYNTICK) {
		/* Collect dyntick-idle snapshots. */
		force_qs_rnp(rsp, dyntick_save_progress_counter);
		fqs_state = RCU_FORCE_QS;
	} else {
		/* Handle dyntick-idle and offline CPUs. */
		force_qs_rnp(rsp, rcu_implicit_dynticks_qs);
	}
	/* Clear flag to prevent immediate re-entry. */
	if (ACCESS_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) {
		raw_spin_lock_irq(&rnp->lock);
		rsp->gp_flags &= ~RCU_GP_FLAG_FQS;
		raw_spin_unlock_irq(&rnp->lock);
	}
	return fqs_state;
}

1267 1268 1269
/*
 * Clean up after the old grace period.
 */
1270
static void rcu_gp_cleanup(struct rcu_state *rsp)
1271 1272 1273 1274
{
	unsigned long gp_duration;
	struct rcu_data *rdp;
	struct rcu_node *rnp = rcu_get_root(rsp);
1275

1276 1277 1278 1279
	raw_spin_lock_irq(&rnp->lock);
	gp_duration = jiffies - rsp->gp_start;
	if (gp_duration > rsp->gp_max)
		rsp->gp_max = gp_duration;
1280

1281 1282 1283 1284 1285 1286 1287 1288
	/*
	 * We know the grace period is complete, but to everyone else
	 * it appears to still be ongoing.  But it is also the case
	 * that to everyone else it looks like there is nothing that
	 * they can do to advance the grace period.  It is therefore
	 * safe for us to drop the lock in order to mark the grace
	 * period as completed in all of the rcu_node structures.
	 */
1289
	raw_spin_unlock_irq(&rnp->lock);
1290

1291 1292 1293 1294 1295 1296 1297 1298 1299 1300
	/*
	 * 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.  This also avoids
	 * some nasty RCU grace-period initialization races by forcing
	 * the end of the current grace period to be completely recorded in
	 * all of the rcu_node structures before the beginning of the next
	 * grace period is recorded in any of the rcu_node structures.
	 */
	rcu_for_each_node_breadth_first(rsp, rnp) {
1301
		raw_spin_lock_irq(&rnp->lock);
1302 1303 1304
		rnp->completed = rsp->gpnum;
		raw_spin_unlock_irq(&rnp->lock);
		cond_resched();
1305
	}
1306 1307
	rnp = rcu_get_root(rsp);
	raw_spin_lock_irq(&rnp->lock);
1308 1309 1310 1311

	rsp->completed = rsp->gpnum; /* Declare grace period done. */
	trace_rcu_grace_period(rsp->name, rsp->completed, "end");
	rsp->fqs_state = RCU_GP_IDLE;
1312
	rdp = this_cpu_ptr(rsp->rda);
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
	if (cpu_needs_another_gp(rsp, rdp))
		rsp->gp_flags = 1;
	raw_spin_unlock_irq(&rnp->lock);
}

/*
 * Body of kthread that handles grace periods.
 */
static int __noreturn rcu_gp_kthread(void *arg)
{
1323
	int fqs_state;
1324
	unsigned long j;
1325
	int ret;
1326 1327 1328 1329 1330 1331 1332
	struct rcu_state *rsp = arg;
	struct rcu_node *rnp = rcu_get_root(rsp);

	for (;;) {

		/* Handle grace-period start. */
		for (;;) {
1333 1334 1335 1336 1337
			wait_event_interruptible(rsp->gp_wq,
						 rsp->gp_flags &
						 RCU_GP_FLAG_INIT);
			if ((rsp->gp_flags & RCU_GP_FLAG_INIT) &&
			    rcu_gp_init(rsp))
1338 1339 1340 1341
				break;
			cond_resched();
			flush_signals(current);
		}
1342

1343 1344
		/* Handle quiescent-state forcing. */
		fqs_state = RCU_SAVE_DYNTICK;
1345 1346 1347 1348 1349
		j = jiffies_till_first_fqs;
		if (j > HZ) {
			j = HZ;
			jiffies_till_first_fqs = HZ;
		}
1350
		for (;;) {
1351
			rsp->jiffies_force_qs = jiffies + j;
1352 1353 1354 1355
			ret = wait_event_interruptible_timeout(rsp->gp_wq,
					(rsp->gp_flags & RCU_GP_FLAG_FQS) ||
					(!ACCESS_ONCE(rnp->qsmask) &&
					 !rcu_preempt_blocked_readers_cgp(rnp)),
1356
					j);
1357
			/* If grace period done, leave loop. */
1358
			if (!ACCESS_ONCE(rnp->qsmask) &&
1359
			    !rcu_preempt_blocked_readers_cgp(rnp))
1360
				break;
1361 1362 1363 1364 1365 1366 1367 1368 1369
			/* If time for quiescent-state forcing, do it. */
			if (ret == 0 || (rsp->gp_flags & RCU_GP_FLAG_FQS)) {
				fqs_state = rcu_gp_fqs(rsp, fqs_state);
				cond_resched();
			} else {
				/* Deal with stray signal. */
				cond_resched();
				flush_signals(current);
			}
1370 1371 1372 1373 1374 1375 1376 1377
			j = jiffies_till_next_fqs;
			if (j > HZ) {
				j = HZ;
				jiffies_till_next_fqs = HZ;
			} else if (j < 1) {
				j = 1;
				jiffies_till_next_fqs = 1;
			}
1378
		}
1379 1380 1381

		/* Handle grace-period end. */
		rcu_gp_cleanup(rsp);
1382 1383 1384
	}
}

1385 1386 1387 1388 1389
/*
 * 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.
1390 1391 1392 1393
 *
 * Note that it is legal for a dying CPU (which is marked as offline) to
 * invoke this function.  This can happen when the dying CPU reports its
 * quiescent state.
1394 1395 1396 1397 1398
 */
static void
rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
	__releases(rcu_get_root(rsp)->lock)
{
1399
	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
1400 1401
	struct rcu_node *rnp = rcu_get_root(rsp);

1402
	if (!rsp->gp_kthread ||
1403 1404
	    !cpu_needs_another_gp(rsp, rdp)) {
		/*
1405 1406 1407
		 * Either we have not yet spawned the grace-period
		 * task or this CPU does not need another grace period.
		 * Either way, don't start a new grace period.
1408 1409 1410 1411
		 */
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
		return;
	}
1412

1413
	rsp->gp_flags = RCU_GP_FLAG_INIT;
1414 1415
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
	wake_up(&rsp->gp_wq);
1416 1417
}

1418
/*
P
Paul E. McKenney 已提交
1419 1420 1421 1422 1423
 * 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.
1424
 */
P
Paul E. McKenney 已提交
1425
static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)
1426
	__releases(rcu_get_root(rsp)->lock)
1427
{
1428
	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
1429 1430
	raw_spin_unlock_irqrestore(&rcu_get_root(rsp)->lock, flags);
	wake_up(&rsp->gp_wq);  /* Memory barrier implied by wake_up() path. */
1431 1432
}

1433
/*
P
Paul E. McKenney 已提交
1434 1435 1436 1437 1438 1439
 * 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.
1440 1441
 */
static void
P
Paul E. McKenney 已提交
1442 1443
rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,
		  struct rcu_node *rnp, unsigned long flags)
1444 1445
	__releases(rnp->lock)
{
1446 1447
	struct rcu_node *rnp_c;

1448 1449 1450 1451 1452
	/* 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
1453
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1454 1455 1456
			return;
		}
		rnp->qsmask &= ~mask;
1457 1458 1459 1460
		trace_rcu_quiescent_state_report(rsp->name, rnp->gpnum,
						 mask, rnp->qsmask, rnp->level,
						 rnp->grplo, rnp->grphi,
						 !!rnp->gp_tasks);
1461
		if (rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {
1462 1463

			/* Other bits still set at this level, so done. */
P
Paul E. McKenney 已提交
1464
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1465 1466 1467 1468 1469 1470 1471 1472 1473
			return;
		}
		mask = rnp->grpmask;
		if (rnp->parent == NULL) {

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

			break;
		}
P
Paul E. McKenney 已提交
1474
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1475
		rnp_c = rnp;
1476
		rnp = rnp->parent;
P
Paul E. McKenney 已提交
1477
		raw_spin_lock_irqsave(&rnp->lock, flags);
1478
		WARN_ON_ONCE(rnp_c->qsmask);
1479 1480 1481 1482
	}

	/*
	 * Get here if we are the last CPU to pass through a quiescent
P
Paul E. McKenney 已提交
1483
	 * state for this grace period.  Invoke rcu_report_qs_rsp()
1484
	 * to clean up and start the next grace period if one is needed.
1485
	 */
P
Paul E. McKenney 已提交
1486
	rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */
1487 1488 1489
}

/*
P
Paul E. McKenney 已提交
1490 1491 1492 1493 1494 1495 1496
 * 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!
1497 1498
 */
static void
1499
rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp)
1500 1501 1502 1503 1504 1505
{
	unsigned long flags;
	unsigned long mask;
	struct rcu_node *rnp;

	rnp = rdp->mynode;
P
Paul E. McKenney 已提交
1506
	raw_spin_lock_irqsave(&rnp->lock, flags);
1507 1508
	if (rdp->passed_quiesce == 0 || rdp->gpnum != rnp->gpnum ||
	    rnp->completed == rnp->gpnum) {
1509 1510

		/*
1511 1512 1513 1514
		 * The grace period in which this quiescent state was
		 * recorded has ended, so don't report it upwards.
		 * We will instead need a new quiescent state that lies
		 * within the current grace period.
1515
		 */
1516
		rdp->passed_quiesce = 0;	/* need qs for new gp. */
P
Paul E. McKenney 已提交
1517
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1518 1519 1520 1521
		return;
	}
	mask = rdp->grpmask;
	if ((rnp->qsmask & mask) == 0) {
P
Paul E. McKenney 已提交
1522
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1523 1524 1525 1526 1527 1528 1529 1530 1531
	} 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
1532
		rcu_report_qs_rnp(mask, rsp, rnp, flags); /* rlses rnp->lock */
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559
	}
}

/*
 * 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.
	 */
1560
	if (!rdp->passed_quiesce)
1561 1562
		return;

P
Paul E. McKenney 已提交
1563 1564 1565 1566
	/*
	 * Tell RCU we are done (but rcu_report_qs_rdp() will be the
	 * judge of that).
	 */
1567
	rcu_report_qs_rdp(rdp->cpu, rsp, rdp);
1568 1569 1570 1571
}

#ifdef CONFIG_HOTPLUG_CPU

1572
/*
1573 1574 1575
 * Send the specified CPU's RCU callbacks to the orphanage.  The
 * specified CPU must be offline, and the caller must hold the
 * ->onofflock.
1576
 */
1577 1578 1579
static void
rcu_send_cbs_to_orphanage(int cpu, struct rcu_state *rsp,
			  struct rcu_node *rnp, struct rcu_data *rdp)
1580
{
1581 1582 1583 1584 1585
	/*
	 * Orphan the callbacks.  First adjust the counts.  This is safe
	 * because ->onofflock excludes _rcu_barrier()'s adoption of
	 * the callbacks, thus no memory barrier is required.
	 */
1586
	if (rdp->nxtlist != NULL) {
1587 1588 1589
		rsp->qlen_lazy += rdp->qlen_lazy;
		rsp->qlen += rdp->qlen;
		rdp->n_cbs_orphaned += rdp->qlen;
1590
		rdp->qlen_lazy = 0;
1591
		ACCESS_ONCE(rdp->qlen) = 0;
1592 1593 1594
	}

	/*
1595 1596 1597 1598 1599 1600 1601
	 * Next, move those callbacks still needing a grace period to
	 * the orphanage, where some other CPU will pick them up.
	 * Some of the callbacks might have gone partway through a grace
	 * period, but that is too bad.  They get to start over because we
	 * cannot assume that grace periods are synchronized across CPUs.
	 * We don't bother updating the ->nxttail[] array yet, instead
	 * we just reset the whole thing later on.
1602
	 */
1603 1604 1605 1606
	if (*rdp->nxttail[RCU_DONE_TAIL] != NULL) {
		*rsp->orphan_nxttail = *rdp->nxttail[RCU_DONE_TAIL];
		rsp->orphan_nxttail = rdp->nxttail[RCU_NEXT_TAIL];
		*rdp->nxttail[RCU_DONE_TAIL] = NULL;
1607 1608 1609
	}

	/*
1610 1611 1612
	 * Then move the ready-to-invoke callbacks to the orphanage,
	 * where some other CPU will pick them up.  These will not be
	 * required to pass though another grace period: They are done.
1613
	 */
1614
	if (rdp->nxtlist != NULL) {
1615 1616
		*rsp->orphan_donetail = rdp->nxtlist;
		rsp->orphan_donetail = rdp->nxttail[RCU_DONE_TAIL];
1617
	}
1618

1619
	/* Finally, initialize the rcu_data structure's list to empty.  */
1620
	init_callback_list(rdp);
1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
}

/*
 * Adopt the RCU callbacks from the specified rcu_state structure's
 * orphanage.  The caller must hold the ->onofflock.
 */
static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
{
	int i;
	struct rcu_data *rdp = __this_cpu_ptr(rsp->rda);

	/* Do the accounting first. */
	rdp->qlen_lazy += rsp->qlen_lazy;
	rdp->qlen += rsp->qlen;
	rdp->n_cbs_adopted += rsp->qlen;
1636 1637
	if (rsp->qlen_lazy != rsp->qlen)
		rcu_idle_count_callbacks_posted();
1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
	rsp->qlen_lazy = 0;
	rsp->qlen = 0;

	/*
	 * We do not need a memory barrier here because the only way we
	 * can get here if there is an rcu_barrier() in flight is if
	 * we are the task doing the rcu_barrier().
	 */

	/* First adopt the ready-to-invoke callbacks. */
	if (rsp->orphan_donelist != NULL) {
		*rsp->orphan_donetail = *rdp->nxttail[RCU_DONE_TAIL];
		*rdp->nxttail[RCU_DONE_TAIL] = rsp->orphan_donelist;
		for (i = RCU_NEXT_SIZE - 1; i >= RCU_DONE_TAIL; i--)
			if (rdp->nxttail[i] == rdp->nxttail[RCU_DONE_TAIL])
				rdp->nxttail[i] = rsp->orphan_donetail;
		rsp->orphan_donelist = NULL;
		rsp->orphan_donetail = &rsp->orphan_donelist;
	}

	/* And then adopt the callbacks that still need a grace period. */
	if (rsp->orphan_nxtlist != NULL) {
		*rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_nxtlist;
		rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_nxttail;
		rsp->orphan_nxtlist = NULL;
		rsp->orphan_nxttail = &rsp->orphan_nxtlist;
	}
}

/*
 * Trace the fact that this CPU is going offline.
 */
static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
{
	RCU_TRACE(unsigned long mask);
	RCU_TRACE(struct rcu_data *rdp = this_cpu_ptr(rsp->rda));
	RCU_TRACE(struct rcu_node *rnp = rdp->mynode);

	RCU_TRACE(mask = rdp->grpmask);
1677 1678 1679
	trace_rcu_grace_period(rsp->name,
			       rnp->gpnum + 1 - !!(rnp->qsmask & mask),
			       "cpuofl");
1680 1681 1682
}

/*
1683
 * The CPU has been completely removed, and some other CPU is reporting
1684 1685
 * this fact from process context.  Do the remainder of the cleanup,
 * including orphaning the outgoing CPU's RCU callbacks, and also
1686 1687
 * adopting them.  There can only be one CPU hotplug operation at a time,
 * so no other CPU can be attempting to update rcu_cpu_kthread_task.
1688
 */
1689
static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
1690
{
1691 1692 1693
	unsigned long flags;
	unsigned long mask;
	int need_report = 0;
1694
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
1695
	struct rcu_node *rnp = rdp->mynode;  /* Outgoing CPU's rdp & rnp. */
1696

1697
	/* Adjust any no-longer-needed kthreads. */
T
Thomas Gleixner 已提交
1698
	rcu_boost_kthread_setaffinity(rnp, -1);
1699

1700
	/* Remove the dead CPU from the bitmasks in the rcu_node hierarchy. */
1701 1702 1703 1704

	/* Exclude any attempts to start a new grace period. */
	raw_spin_lock_irqsave(&rsp->onofflock, flags);

1705 1706 1707 1708
	/* Orphan the dead CPU's callbacks, and adopt them if appropriate. */
	rcu_send_cbs_to_orphanage(cpu, rsp, rnp, rdp);
	rcu_adopt_orphan_cbs(rsp);

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 1734 1735 1736 1737 1738 1739 1740
	/* Remove the outgoing CPU from the masks in the rcu_node hierarchy. */
	mask = rdp->grpmask;	/* rnp->grplo is constant. */
	do {
		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
		rnp->qsmaskinit &= ~mask;
		if (rnp->qsmaskinit != 0) {
			if (rnp != rdp->mynode)
				raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
			break;
		}
		if (rnp == rdp->mynode)
			need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);
		else
			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
		mask = rnp->grpmask;
		rnp = rnp->parent;
	} while (rnp != NULL);

	/*
	 * We still hold the leaf rcu_node structure lock here, and
	 * irqs are still disabled.  The reason for this subterfuge is
	 * because invoking rcu_report_unblock_qs_rnp() with ->onofflock
	 * held leads to deadlock.
	 */
	raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */
	rnp = rdp->mynode;
	if (need_report & RCU_OFL_TASKS_NORM_GP)
		rcu_report_unblock_qs_rnp(rnp, flags);
	else
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
	if (need_report & RCU_OFL_TASKS_EXP_GP)
		rcu_report_exp_rnp(rsp, rnp, true);
1741 1742 1743
	WARN_ONCE(rdp->qlen != 0 || rdp->nxtlist != NULL,
		  "rcu_cleanup_dead_cpu: Callbacks on offline CPU %d: qlen=%lu, nxtlist=%p\n",
		  cpu, rdp->qlen, rdp->nxtlist);
1744 1745 1746
	init_callback_list(rdp);
	/* Disallow further callbacks on this CPU. */
	rdp->nxttail[RCU_NEXT_TAIL] = NULL;
1747 1748 1749 1750
}

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1751
static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
1752 1753 1754
{
}

1755
static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
1756 1757 1758 1759 1760 1761 1762 1763 1764
{
}

#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.
 */
1765
static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)
1766 1767 1768
{
	unsigned long flags;
	struct rcu_head *next, *list, **tail;
1769
	int bl, count, count_lazy, i;
1770 1771

	/* If no callbacks are ready, just return.*/
1772
	if (!cpu_has_callbacks_ready_to_invoke(rdp)) {
1773
		trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, 0);
1774 1775 1776
		trace_rcu_batch_end(rsp->name, 0, !!ACCESS_ONCE(rdp->nxtlist),
				    need_resched(), is_idle_task(current),
				    rcu_is_callbacks_kthread());
1777
		return;
1778
	}
1779 1780 1781 1782 1783 1784

	/*
	 * Extract the list of ready callbacks, disabling to prevent
	 * races with call_rcu() from interrupt handlers.
	 */
	local_irq_save(flags);
1785
	WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));
1786
	bl = rdp->blimit;
1787
	trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, bl);
1788 1789 1790 1791
	list = rdp->nxtlist;
	rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];
	*rdp->nxttail[RCU_DONE_TAIL] = NULL;
	tail = rdp->nxttail[RCU_DONE_TAIL];
1792 1793 1794
	for (i = RCU_NEXT_SIZE - 1; i >= 0; i--)
		if (rdp->nxttail[i] == rdp->nxttail[RCU_DONE_TAIL])
			rdp->nxttail[i] = &rdp->nxtlist;
1795 1796 1797
	local_irq_restore(flags);

	/* Invoke callbacks. */
1798
	count = count_lazy = 0;
1799 1800 1801
	while (list) {
		next = list->next;
		prefetch(next);
1802
		debug_rcu_head_unqueue(list);
1803 1804
		if (__rcu_reclaim(rsp->name, list))
			count_lazy++;
1805
		list = next;
1806 1807 1808 1809
		/* Stop only if limit reached and CPU has something to do. */
		if (++count >= bl &&
		    (need_resched() ||
		     (!is_idle_task(current) && !rcu_is_callbacks_kthread())))
1810 1811 1812 1813
			break;
	}

	local_irq_save(flags);
1814 1815 1816
	trace_rcu_batch_end(rsp->name, count, !!list, need_resched(),
			    is_idle_task(current),
			    rcu_is_callbacks_kthread());
1817 1818 1819 1820 1821

	/* Update count, and requeue any remaining callbacks. */
	if (list != NULL) {
		*tail = rdp->nxtlist;
		rdp->nxtlist = list;
1822 1823 1824
		for (i = 0; i < RCU_NEXT_SIZE; i++)
			if (&rdp->nxtlist == rdp->nxttail[i])
				rdp->nxttail[i] = tail;
1825 1826 1827
			else
				break;
	}
1828 1829
	smp_mb(); /* List handling before counting for rcu_barrier(). */
	rdp->qlen_lazy -= count_lazy;
1830
	ACCESS_ONCE(rdp->qlen) -= count;
1831
	rdp->n_cbs_invoked += count;
1832 1833 1834 1835 1836

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

1837 1838 1839 1840 1841 1842
	/* 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;
1843
	WARN_ON_ONCE((rdp->nxtlist == NULL) != (rdp->qlen == 0));
1844

1845 1846
	local_irq_restore(flags);

1847
	/* Re-invoke RCU core processing if there are callbacks remaining. */
1848
	if (cpu_has_callbacks_ready_to_invoke(rdp))
1849
		invoke_rcu_core();
1850 1851 1852 1853 1854
}

/*
 * 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).
1855
 * Also schedule RCU core processing.
1856
 *
1857
 * This function must be called from hardirq context.  It is normally
1858 1859 1860 1861 1862
 * 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)
{
1863
	trace_rcu_utilization("Start scheduler-tick");
1864
	increment_cpu_stall_ticks();
1865
	if (user || rcu_is_cpu_rrupt_from_idle()) {
1866 1867 1868 1869 1870

		/*
		 * 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
1871
		 * a quiescent state, so note it.
1872 1873
		 *
		 * No memory barrier is required here because both
1874 1875 1876
		 * 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.
1877 1878
		 */

1879 1880
		rcu_sched_qs(cpu);
		rcu_bh_qs(cpu);
1881 1882 1883 1884 1885 1886 1887

	} 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
1888
		 * critical section, so note it.
1889 1890
		 */

1891
		rcu_bh_qs(cpu);
1892
	}
1893
	rcu_preempt_check_callbacks(cpu);
1894
	if (rcu_pending(cpu))
1895
		invoke_rcu_core();
1896
	trace_rcu_utilization("End scheduler-tick");
1897 1898 1899 1900 1901
}

/*
 * Scan the leaf rcu_node structures, processing dyntick state for any that
 * have not yet encountered a quiescent state, using the function specified.
1902 1903
 * Also initiate boosting for any threads blocked on the root rcu_node.
 *
1904
 * The caller must have suppressed start of new grace periods.
1905
 */
1906
static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))
1907 1908 1909 1910 1911
{
	unsigned long bit;
	int cpu;
	unsigned long flags;
	unsigned long mask;
1912
	struct rcu_node *rnp;
1913

1914
	rcu_for_each_leaf_node(rsp, rnp) {
1915
		cond_resched();
1916
		mask = 0;
P
Paul E. McKenney 已提交
1917
		raw_spin_lock_irqsave(&rnp->lock, flags);
1918
		if (!rcu_gp_in_progress(rsp)) {
P
Paul E. McKenney 已提交
1919
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1920
			return;
1921
		}
1922
		if (rnp->qsmask == 0) {
1923
			rcu_initiate_boost(rnp, flags); /* releases rnp->lock */
1924 1925
			continue;
		}
1926
		cpu = rnp->grplo;
1927
		bit = 1;
1928
		for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {
1929 1930
			if ((rnp->qsmask & bit) != 0 &&
			    f(per_cpu_ptr(rsp->rda, cpu)))
1931 1932
				mask |= bit;
		}
1933
		if (mask != 0) {
1934

P
Paul E. McKenney 已提交
1935 1936
			/* rcu_report_qs_rnp() releases rnp->lock. */
			rcu_report_qs_rnp(mask, rsp, rnp, flags);
1937 1938
			continue;
		}
P
Paul E. McKenney 已提交
1939
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1940
	}
1941
	rnp = rcu_get_root(rsp);
1942 1943 1944 1945
	if (rnp->qsmask == 0) {
		raw_spin_lock_irqsave(&rnp->lock, flags);
		rcu_initiate_boost(rnp, flags); /* releases rnp->lock. */
	}
1946 1947 1948 1949 1950 1951
}

/*
 * Force quiescent states on reluctant CPUs, and also detect which
 * CPUs are in dyntick-idle mode.
 */
1952
static void force_quiescent_state(struct rcu_state *rsp)
1953 1954
{
	unsigned long flags;
1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
	bool ret;
	struct rcu_node *rnp;
	struct rcu_node *rnp_old = NULL;

	/* Funnel through hierarchy to reduce memory contention. */
	rnp = per_cpu_ptr(rsp->rda, raw_smp_processor_id())->mynode;
	for (; rnp != NULL; rnp = rnp->parent) {
		ret = (ACCESS_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) ||
		      !raw_spin_trylock(&rnp->fqslock);
		if (rnp_old != NULL)
			raw_spin_unlock(&rnp_old->fqslock);
		if (ret) {
			rsp->n_force_qs_lh++;
			return;
		}
		rnp_old = rnp;
	}
	/* rnp_old == rcu_get_root(rsp), rnp == NULL. */
1973

1974 1975 1976 1977 1978 1979
	/* Reached the root of the rcu_node tree, acquire lock. */
	raw_spin_lock_irqsave(&rnp_old->lock, flags);
	raw_spin_unlock(&rnp_old->fqslock);
	if (ACCESS_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) {
		rsp->n_force_qs_lh++;
		raw_spin_unlock_irqrestore(&rnp_old->lock, flags);
1980
		return;  /* Someone beat us to it. */
1981
	}
1982
	rsp->gp_flags |= RCU_GP_FLAG_FQS;
1983
	raw_spin_unlock_irqrestore(&rnp_old->lock, flags);
1984
	wake_up(&rsp->gp_wq);  /* Memory barrier implied by wake_up() path. */
1985 1986 1987
}

/*
1988 1989 1990
 * This does the RCU core processing work for the specified rcu_state
 * and rcu_data structures.  This may be called only from the CPU to
 * whom the rdp belongs.
1991 1992
 */
static void
1993
__rcu_process_callbacks(struct rcu_state *rsp)
1994 1995
{
	unsigned long flags;
1996
	struct rcu_data *rdp = __this_cpu_ptr(rsp->rda);
1997

1998 1999
	WARN_ON_ONCE(rdp->beenonline == 0);

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
	/*
	 * 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
2011
		raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);
2012 2013 2014 2015
		rcu_start_gp(rsp, flags);  /* releases above lock */
	}

	/* If there are callbacks ready, invoke them. */
2016
	if (cpu_has_callbacks_ready_to_invoke(rdp))
2017
		invoke_rcu_callbacks(rsp, rdp);
2018 2019
}

2020
/*
2021
 * Do RCU core processing for the current CPU.
2022
 */
2023
static void rcu_process_callbacks(struct softirq_action *unused)
2024
{
2025 2026
	struct rcu_state *rsp;

2027 2028
	if (cpu_is_offline(smp_processor_id()))
		return;
2029
	trace_rcu_utilization("Start RCU core");
2030 2031
	for_each_rcu_flavor(rsp)
		__rcu_process_callbacks(rsp);
2032
	trace_rcu_utilization("End RCU core");
2033 2034
}

2035
/*
2036 2037 2038 2039 2040
 * Schedule RCU callback invocation.  If the specified type of RCU
 * does not support RCU priority boosting, just do a direct call,
 * otherwise wake up the per-CPU kernel kthread.  Note that because we
 * are running on the current CPU with interrupts disabled, the
 * rcu_cpu_kthread_task cannot disappear out from under us.
2041
 */
2042
static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)
2043
{
2044 2045
	if (unlikely(!ACCESS_ONCE(rcu_scheduler_fully_active)))
		return;
2046 2047
	if (likely(!rsp->boost)) {
		rcu_do_batch(rsp, rdp);
2048 2049
		return;
	}
2050
	invoke_rcu_callbacks_kthread();
2051 2052
}

2053
static void invoke_rcu_core(void)
2054 2055 2056 2057
{
	raise_softirq(RCU_SOFTIRQ);
}

2058 2059 2060 2061 2062
/*
 * Handle any core-RCU processing required by a call_rcu() invocation.
 */
static void __call_rcu_core(struct rcu_state *rsp, struct rcu_data *rdp,
			    struct rcu_head *head, unsigned long flags)
2063
{
2064 2065 2066 2067
	/*
	 * If called from an extended quiescent state, invoke the RCU
	 * core in order to force a re-evaluation of RCU's idleness.
	 */
2068
	if (rcu_is_cpu_idle() && cpu_online(smp_processor_id()))
2069 2070
		invoke_rcu_core();

2071
	/* If interrupts were disabled or CPU offline, don't invoke RCU core. */
2072
	if (irqs_disabled_flags(flags) || cpu_is_offline(smp_processor_id()))
2073
		return;
2074

2075 2076 2077 2078 2079 2080 2081
	/*
	 * 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.
	 */
2082
	if (unlikely(rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {
2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099

		/* Are we ignoring a completed grace period? */
		rcu_process_gp_end(rsp, rdp);
		check_for_new_grace_period(rsp, rdp);

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

			raw_spin_lock_irqsave(&rnp_root->lock, nestflag);
			rcu_start_gp(rsp, nestflag);  /* rlses rnp_root->lock */
		} else {
			/* Give the grace period a kick. */
			rdp->blimit = LONG_MAX;
			if (rsp->n_force_qs == rdp->n_force_qs_snap &&
			    *rdp->nxttail[RCU_DONE_TAIL] != head)
2100
				force_quiescent_state(rsp);
2101 2102 2103
			rdp->n_force_qs_snap = rsp->n_force_qs;
			rdp->qlen_last_fqs_check = rdp->qlen;
		}
2104
	}
2105 2106
}

2107 2108
static void
__call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),
2109
	   struct rcu_state *rsp, bool lazy)
2110 2111 2112 2113
{
	unsigned long flags;
	struct rcu_data *rdp;

2114
	WARN_ON_ONCE((unsigned long)head & 0x3); /* Misaligned rcu_head! */
2115
	debug_rcu_head_queue(head);
2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
	head->func = func;
	head->next = NULL;

	/*
	 * 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);
2126
	rdp = this_cpu_ptr(rsp->rda);
2127 2128

	/* Add the callback to our list. */
2129 2130 2131 2132 2133 2134
	if (unlikely(rdp->nxttail[RCU_NEXT_TAIL] == NULL)) {
		/* _call_rcu() is illegal on offline CPU; leak the callback. */
		WARN_ON_ONCE(1);
		local_irq_restore(flags);
		return;
	}
2135
	ACCESS_ONCE(rdp->qlen)++;
2136 2137
	if (lazy)
		rdp->qlen_lazy++;
2138 2139
	else
		rcu_idle_count_callbacks_posted();
2140 2141 2142
	smp_mb();  /* Count before adding callback for rcu_barrier(). */
	*rdp->nxttail[RCU_NEXT_TAIL] = head;
	rdp->nxttail[RCU_NEXT_TAIL] = &head->next;
2143

2144 2145
	if (__is_kfree_rcu_offset((unsigned long)func))
		trace_rcu_kfree_callback(rsp->name, head, (unsigned long)func,
2146
					 rdp->qlen_lazy, rdp->qlen);
2147
	else
2148
		trace_rcu_callback(rsp->name, head, rdp->qlen_lazy, rdp->qlen);
2149

2150 2151
	/* Go handle any RCU core processing required. */
	__call_rcu_core(rsp, rdp, head, flags);
2152 2153 2154 2155
	local_irq_restore(flags);
}

/*
2156
 * Queue an RCU-sched callback for invocation after a grace period.
2157
 */
2158
void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
2159
{
2160
	__call_rcu(head, func, &rcu_sched_state, 0);
2161
}
2162
EXPORT_SYMBOL_GPL(call_rcu_sched);
2163 2164

/*
2165
 * Queue an RCU callback for invocation after a quicker grace period.
2166 2167 2168
 */
void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
{
2169
	__call_rcu(head, func, &rcu_bh_state, 0);
2170 2171 2172
}
EXPORT_SYMBOL_GPL(call_rcu_bh);

2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183
/*
 * Because a context switch is a grace period for RCU-sched and RCU-bh,
 * any blocking grace-period wait automatically implies a grace period
 * if there is only one CPU online at any point time during execution
 * of either synchronize_sched() or synchronize_rcu_bh().  It is OK to
 * occasionally incorrectly indicate that there are multiple CPUs online
 * when there was in fact only one the whole time, as this just adds
 * some overhead: RCU still operates correctly.
 */
static inline int rcu_blocking_is_gp(void)
{
2184 2185
	int ret;

2186
	might_sleep();  /* Check for RCU read-side critical section. */
2187 2188 2189 2190
	preempt_disable();
	ret = num_online_cpus() <= 1;
	preempt_enable();
	return ret;
2191 2192
}

2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217
/**
 * 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)
{
2218 2219 2220 2221
	rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&
			   !lock_is_held(&rcu_lock_map) &&
			   !lock_is_held(&rcu_sched_lock_map),
			   "Illegal synchronize_sched() in RCU-sched read-side critical section");
2222 2223
	if (rcu_blocking_is_gp())
		return;
2224
	wait_rcu_gp(call_rcu_sched);
2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238
}
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)
{
2239 2240 2241 2242
	rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&
			   !lock_is_held(&rcu_lock_map) &&
			   !lock_is_held(&rcu_sched_lock_map),
			   "Illegal synchronize_rcu_bh() in RCU-bh read-side critical section");
2243 2244
	if (rcu_blocking_is_gp())
		return;
2245
	wait_rcu_gp(call_rcu_bh);
2246 2247 2248
}
EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268
static atomic_t sync_sched_expedited_started = ATOMIC_INIT(0);
static atomic_t sync_sched_expedited_done = ATOMIC_INIT(0);

static int synchronize_sched_expedited_cpu_stop(void *data)
{
	/*
	 * There must be a full memory barrier on each affected CPU
	 * between the time that try_stop_cpus() is called and the
	 * time that it returns.
	 *
	 * In the current initial implementation of cpu_stop, the
	 * above condition is already met when the control reaches
	 * this point and the following smp_mb() is not strictly
	 * necessary.  Do smp_mb() anyway for documentation and
	 * robustness against future implementation changes.
	 */
	smp_mb(); /* See above comment block. */
	return 0;
}

2269 2270 2271 2272 2273 2274 2275 2276 2277 2278
/**
 * synchronize_sched_expedited - Brute-force RCU-sched grace period
 *
 * Wait for an RCU-sched grace period to elapse, but use a "big hammer"
 * approach to force the grace period to end quickly.  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_sched_expedited() in a loop, please
 * restructure your code to batch your updates, and then use a single
 * synchronize_sched() instead.
2279
 *
2280 2281 2282 2283
 * Note that it is illegal to call this function while holding any lock
 * that is acquired by a CPU-hotplug notifier.  And yes, it is also illegal
 * to call this function from a CPU-hotplug notifier.  Failing to observe
 * these restriction will result in deadlock.
2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312
 *
 * This implementation can be thought of as an application of ticket
 * locking to RCU, with sync_sched_expedited_started and
 * sync_sched_expedited_done taking on the roles of the halves
 * of the ticket-lock word.  Each task atomically increments
 * sync_sched_expedited_started upon entry, snapshotting the old value,
 * then attempts to stop all the CPUs.  If this succeeds, then each
 * CPU will have executed a context switch, resulting in an RCU-sched
 * grace period.  We are then done, so we use atomic_cmpxchg() to
 * update sync_sched_expedited_done to match our snapshot -- but
 * only if someone else has not already advanced past our snapshot.
 *
 * On the other hand, if try_stop_cpus() fails, we check the value
 * of sync_sched_expedited_done.  If it has advanced past our
 * initial snapshot, then someone else must have forced a grace period
 * some time after we took our snapshot.  In this case, our work is
 * done for us, and we can simply return.  Otherwise, we try again,
 * but keep our initial snapshot for purposes of checking for someone
 * doing our work for us.
 *
 * If we fail too many times in a row, we fall back to synchronize_sched().
 */
void synchronize_sched_expedited(void)
{
	int firstsnap, s, snap, trycount = 0;

	/* Note that atomic_inc_return() implies full memory barrier. */
	firstsnap = snap = atomic_inc_return(&sync_sched_expedited_started);
	get_online_cpus();
2313
	WARN_ON_ONCE(cpu_is_offline(raw_smp_processor_id()));
2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324

	/*
	 * Each pass through the following loop attempts to force a
	 * context switch on each CPU.
	 */
	while (try_stop_cpus(cpu_online_mask,
			     synchronize_sched_expedited_cpu_stop,
			     NULL) == -EAGAIN) {
		put_online_cpus();

		/* No joy, try again later.  Or just synchronize_sched(). */
2325
		if (trycount++ < 10) {
2326
			udelay(trycount * num_online_cpus());
2327
		} else {
2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369
			synchronize_sched();
			return;
		}

		/* Check to see if someone else did our work for us. */
		s = atomic_read(&sync_sched_expedited_done);
		if (UINT_CMP_GE((unsigned)s, (unsigned)firstsnap)) {
			smp_mb(); /* ensure test happens before caller kfree */
			return;
		}

		/*
		 * Refetching sync_sched_expedited_started allows later
		 * callers to piggyback on our grace period.  We subtract
		 * 1 to get the same token that the last incrementer got.
		 * We retry after they started, so our grace period works
		 * for them, and they started after our first try, so their
		 * grace period works for us.
		 */
		get_online_cpus();
		snap = atomic_read(&sync_sched_expedited_started);
		smp_mb(); /* ensure read is before try_stop_cpus(). */
	}

	/*
	 * Everyone up to our most recent fetch is covered by our grace
	 * period.  Update the counter, but only if our work is still
	 * relevant -- which it won't be if someone who started later
	 * than we did beat us to the punch.
	 */
	do {
		s = atomic_read(&sync_sched_expedited_done);
		if (UINT_CMP_GE((unsigned)s, (unsigned)snap)) {
			smp_mb(); /* ensure test happens before caller kfree */
			break;
		}
	} while (atomic_cmpxchg(&sync_sched_expedited_done, s, snap) != s);

	put_online_cpus();
}
EXPORT_SYMBOL_GPL(synchronize_sched_expedited);

2370 2371 2372 2373 2374 2375 2376 2377 2378
/*
 * 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)
{
2379 2380
	struct rcu_node *rnp = rdp->mynode;

2381 2382 2383 2384 2385 2386
	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? */
2387 2388
	if (rcu_scheduler_fully_active &&
	    rdp->qs_pending && !rdp->passed_quiesce) {
2389
		rdp->n_rp_qs_pending++;
2390
	} else if (rdp->qs_pending && rdp->passed_quiesce) {
2391
		rdp->n_rp_report_qs++;
2392
		return 1;
2393
	}
2394 2395

	/* Does this CPU have callbacks ready to invoke? */
2396 2397
	if (cpu_has_callbacks_ready_to_invoke(rdp)) {
		rdp->n_rp_cb_ready++;
2398
		return 1;
2399
	}
2400 2401

	/* Has RCU gone idle with this CPU needing another grace period? */
2402 2403
	if (cpu_needs_another_gp(rsp, rdp)) {
		rdp->n_rp_cpu_needs_gp++;
2404
		return 1;
2405
	}
2406 2407

	/* Has another RCU grace period completed?  */
2408
	if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */
2409
		rdp->n_rp_gp_completed++;
2410
		return 1;
2411
	}
2412 2413

	/* Has a new RCU grace period started? */
2414
	if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */
2415
		rdp->n_rp_gp_started++;
2416
		return 1;
2417
	}
2418 2419

	/* nothing to do */
2420
	rdp->n_rp_need_nothing++;
2421 2422 2423 2424 2425 2426 2427 2428
	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.
 */
2429
static int rcu_pending(int cpu)
2430
{
2431 2432 2433 2434 2435 2436
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
		if (__rcu_pending(rsp, per_cpu_ptr(rsp->rda, cpu)))
			return 1;
	return 0;
2437 2438 2439 2440 2441
}

/*
 * 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
2442
 * 1 if so.
2443
 */
2444
static int rcu_cpu_has_callbacks(int cpu)
2445
{
2446 2447
	struct rcu_state *rsp;

2448
	/* RCU callbacks either ready or pending? */
2449 2450 2451 2452
	for_each_rcu_flavor(rsp)
		if (per_cpu_ptr(rsp->rda, cpu)->nxtlist)
			return 1;
	return 0;
2453 2454
}

2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
/*
 * Helper function for _rcu_barrier() tracing.  If tracing is disabled,
 * the compiler is expected to optimize this away.
 */
static void _rcu_barrier_trace(struct rcu_state *rsp, char *s,
			       int cpu, unsigned long done)
{
	trace_rcu_barrier(rsp->name, s, cpu,
			  atomic_read(&rsp->barrier_cpu_count), done);
}

2466 2467 2468 2469
/*
 * RCU callback function for _rcu_barrier().  If we are last, wake
 * up the task executing _rcu_barrier().
 */
2470
static void rcu_barrier_callback(struct rcu_head *rhp)
2471
{
2472 2473 2474
	struct rcu_data *rdp = container_of(rhp, struct rcu_data, barrier_head);
	struct rcu_state *rsp = rdp->rsp;

2475 2476
	if (atomic_dec_and_test(&rsp->barrier_cpu_count)) {
		_rcu_barrier_trace(rsp, "LastCB", -1, rsp->n_barrier_done);
2477
		complete(&rsp->barrier_completion);
2478 2479 2480
	} else {
		_rcu_barrier_trace(rsp, "CB", -1, rsp->n_barrier_done);
	}
2481 2482 2483 2484 2485 2486 2487
}

/*
 * Called with preemption disabled, and from cross-cpu IRQ context.
 */
static void rcu_barrier_func(void *type)
{
2488
	struct rcu_state *rsp = type;
2489
	struct rcu_data *rdp = __this_cpu_ptr(rsp->rda);
2490

2491
	_rcu_barrier_trace(rsp, "IRQ", -1, rsp->n_barrier_done);
2492
	atomic_inc(&rsp->barrier_cpu_count);
2493
	rsp->call(&rdp->barrier_head, rcu_barrier_callback);
2494 2495 2496 2497 2498 2499
}

/*
 * Orchestrate the specified type of RCU barrier, waiting for all
 * RCU callbacks of the specified type to complete.
 */
2500
static void _rcu_barrier(struct rcu_state *rsp)
2501
{
2502 2503
	int cpu;
	struct rcu_data *rdp;
2504 2505
	unsigned long snap = ACCESS_ONCE(rsp->n_barrier_done);
	unsigned long snap_done;
2506

2507
	_rcu_barrier_trace(rsp, "Begin", -1, snap);
2508

2509
	/* Take mutex to serialize concurrent rcu_barrier() requests. */
2510
	mutex_lock(&rsp->barrier_mutex);
2511

2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524
	/*
	 * Ensure that all prior references, including to ->n_barrier_done,
	 * are ordered before the _rcu_barrier() machinery.
	 */
	smp_mb();  /* See above block comment. */

	/*
	 * Recheck ->n_barrier_done to see if others did our work for us.
	 * This means checking ->n_barrier_done for an even-to-odd-to-even
	 * transition.  The "if" expression below therefore rounds the old
	 * value up to the next even number and adds two before comparing.
	 */
	snap_done = ACCESS_ONCE(rsp->n_barrier_done);
2525
	_rcu_barrier_trace(rsp, "Check", -1, snap_done);
2526
	if (ULONG_CMP_GE(snap_done, ((snap + 1) & ~0x1) + 2)) {
2527
		_rcu_barrier_trace(rsp, "EarlyExit", -1, snap_done);
2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
		smp_mb(); /* caller's subsequent code after above check. */
		mutex_unlock(&rsp->barrier_mutex);
		return;
	}

	/*
	 * Increment ->n_barrier_done to avoid duplicate work.  Use
	 * ACCESS_ONCE() to prevent the compiler from speculating
	 * the increment to precede the early-exit check.
	 */
	ACCESS_ONCE(rsp->n_barrier_done)++;
	WARN_ON_ONCE((rsp->n_barrier_done & 0x1) != 1);
2540
	_rcu_barrier_trace(rsp, "Inc1", -1, rsp->n_barrier_done);
2541
	smp_mb(); /* Order ->n_barrier_done increment with below mechanism. */
2542

2543
	/*
2544 2545
	 * Initialize the count to one rather than to zero in order to
	 * avoid a too-soon return to zero in case of a short grace period
2546 2547
	 * (or preemption of this task).  Exclude CPU-hotplug operations
	 * to ensure that no offline CPU has callbacks queued.
2548
	 */
2549
	init_completion(&rsp->barrier_completion);
2550
	atomic_set(&rsp->barrier_cpu_count, 1);
2551
	get_online_cpus();
2552 2553

	/*
2554 2555 2556
	 * Force each CPU with callbacks to register a new callback.
	 * When that callback is invoked, we will know that all of the
	 * corresponding CPU's preceding callbacks have been invoked.
2557
	 */
2558
	for_each_online_cpu(cpu) {
2559
		rdp = per_cpu_ptr(rsp->rda, cpu);
2560
		if (ACCESS_ONCE(rdp->qlen)) {
2561 2562
			_rcu_barrier_trace(rsp, "OnlineQ", cpu,
					   rsp->n_barrier_done);
2563
			smp_call_function_single(cpu, rcu_barrier_func, rsp, 1);
2564
		} else {
2565 2566
			_rcu_barrier_trace(rsp, "OnlineNQ", cpu,
					   rsp->n_barrier_done);
2567 2568
		}
	}
2569
	put_online_cpus();
2570 2571 2572 2573 2574

	/*
	 * Now that we have an rcu_barrier_callback() callback on each
	 * CPU, and thus each counted, remove the initial count.
	 */
2575
	if (atomic_dec_and_test(&rsp->barrier_cpu_count))
2576
		complete(&rsp->barrier_completion);
2577

2578 2579 2580 2581
	/* Increment ->n_barrier_done to prevent duplicate work. */
	smp_mb(); /* Keep increment after above mechanism. */
	ACCESS_ONCE(rsp->n_barrier_done)++;
	WARN_ON_ONCE((rsp->n_barrier_done & 0x1) != 0);
2582
	_rcu_barrier_trace(rsp, "Inc2", -1, rsp->n_barrier_done);
2583 2584
	smp_mb(); /* Keep increment before caller's subsequent code. */

2585
	/* Wait for all rcu_barrier_callback() callbacks to be invoked. */
2586
	wait_for_completion(&rsp->barrier_completion);
2587 2588

	/* Other rcu_barrier() invocations can now safely proceed. */
2589
	mutex_unlock(&rsp->barrier_mutex);
2590 2591 2592 2593 2594 2595 2596
}

/**
 * rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.
 */
void rcu_barrier_bh(void)
{
2597
	_rcu_barrier(&rcu_bh_state);
2598 2599 2600 2601 2602 2603 2604 2605
}
EXPORT_SYMBOL_GPL(rcu_barrier_bh);

/**
 * rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.
 */
void rcu_barrier_sched(void)
{
2606
	_rcu_barrier(&rcu_sched_state);
2607 2608 2609
}
EXPORT_SYMBOL_GPL(rcu_barrier_sched);

2610
/*
2611
 * Do boot-time initialization of a CPU's per-CPU RCU data.
2612
 */
2613 2614
static void __init
rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)
2615 2616
{
	unsigned long flags;
2617
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
2618 2619 2620
	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
2621
	raw_spin_lock_irqsave(&rnp->lock, flags);
2622
	rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);
2623
	init_callback_list(rdp);
2624
	rdp->qlen_lazy = 0;
2625
	ACCESS_ONCE(rdp->qlen) = 0;
2626
	rdp->dynticks = &per_cpu(rcu_dynticks, cpu);
2627
	WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE);
2628
	WARN_ON_ONCE(atomic_read(&rdp->dynticks->dynticks) != 1);
2629 2630 2631
#ifdef CONFIG_RCU_USER_QS
	WARN_ON_ONCE(rdp->dynticks->in_user);
#endif
2632
	rdp->cpu = cpu;
2633
	rdp->rsp = rsp;
P
Paul E. McKenney 已提交
2634
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
2635 2636 2637 2638 2639 2640 2641
}

/*
 * 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.
2642
 */
2643
static void __cpuinit
P
Paul E. McKenney 已提交
2644
rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptible)
2645 2646 2647
{
	unsigned long flags;
	unsigned long mask;
2648
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
2649 2650 2651
	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
2652
	raw_spin_lock_irqsave(&rnp->lock, flags);
2653
	rdp->beenonline = 1;	 /* We have now been online. */
P
Paul E. McKenney 已提交
2654
	rdp->preemptible = preemptible;
2655 2656
	rdp->qlen_last_fqs_check = 0;
	rdp->n_force_qs_snap = rsp->n_force_qs;
2657
	rdp->blimit = blimit;
2658
	init_callback_list(rdp);  /* Re-enable callbacks on this CPU. */
2659
	rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
2660 2661
	atomic_set(&rdp->dynticks->dynticks,
		   (atomic_read(&rdp->dynticks->dynticks) & ~0x1) + 1);
2662
	rcu_prepare_for_idle_init(cpu);
P
Paul E. McKenney 已提交
2663
	raw_spin_unlock(&rnp->lock);		/* irqs remain disabled. */
2664 2665 2666 2667 2668 2669 2670

	/*
	 * 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
2671
	raw_spin_lock(&rsp->onofflock);		/* irqs already disabled. */
2672 2673 2674 2675 2676 2677

	/* 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
2678
		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
2679 2680
		rnp->qsmaskinit |= mask;
		mask = rnp->grpmask;
2681
		if (rnp == rdp->mynode) {
2682 2683 2684 2685 2686 2687
			/*
			 * If there is a grace period in progress, we will
			 * set up to wait for it next time we run the
			 * RCU core code.
			 */
			rdp->gpnum = rnp->completed;
2688
			rdp->completed = rnp->completed;
2689 2690
			rdp->passed_quiesce = 0;
			rdp->qs_pending = 0;
2691
			trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuonl");
2692
		}
P
Paul E. McKenney 已提交
2693
		raw_spin_unlock(&rnp->lock); /* irqs already disabled. */
2694 2695 2696
		rnp = rnp->parent;
	} while (rnp != NULL && !(rnp->qsmaskinit & mask));

P
Paul E. McKenney 已提交
2697
	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
2698 2699
}

P
Peter Zijlstra 已提交
2700
static void __cpuinit rcu_prepare_cpu(int cpu)
2701
{
2702 2703 2704 2705 2706
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
		rcu_init_percpu_data(cpu, rsp,
				     strcmp(rsp->name, "rcu_preempt") == 0);
2707 2708 2709
}

/*
2710
 * Handle CPU online/offline notification events.
2711
 */
2712 2713
static int __cpuinit rcu_cpu_notify(struct notifier_block *self,
				    unsigned long action, void *hcpu)
2714 2715
{
	long cpu = (long)hcpu;
2716
	struct rcu_data *rdp = per_cpu_ptr(rcu_state->rda, cpu);
2717
	struct rcu_node *rnp = rdp->mynode;
2718
	struct rcu_state *rsp;
2719

2720
	trace_rcu_utilization("Start CPU hotplug");
2721 2722 2723
	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
P
Peter Zijlstra 已提交
2724 2725
		rcu_prepare_cpu(cpu);
		rcu_prepare_kthreads(cpu);
2726 2727
		break;
	case CPU_ONLINE:
2728
	case CPU_DOWN_FAILED:
T
Thomas Gleixner 已提交
2729
		rcu_boost_kthread_setaffinity(rnp, -1);
2730 2731
		break;
	case CPU_DOWN_PREPARE:
T
Thomas Gleixner 已提交
2732
		rcu_boost_kthread_setaffinity(rnp, cpu);
2733
		break;
2734 2735 2736
	case CPU_DYING:
	case CPU_DYING_FROZEN:
		/*
2737 2738 2739
		 * The whole machine is "stopped" except this CPU, so we can
		 * touch any data without introducing corruption. We send the
		 * dying CPU's callbacks to an arbitrarily chosen online CPU.
2740
		 */
2741 2742
		for_each_rcu_flavor(rsp)
			rcu_cleanup_dying_cpu(rsp);
2743
		rcu_cleanup_after_idle(cpu);
2744
		break;
2745 2746 2747 2748
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
	case CPU_UP_CANCELED:
	case CPU_UP_CANCELED_FROZEN:
2749 2750
		for_each_rcu_flavor(rsp)
			rcu_cleanup_dead_cpu(cpu, rsp);
2751 2752 2753 2754
		break;
	default:
		break;
	}
2755
	trace_rcu_utilization("End CPU hotplug");
2756 2757 2758
	return NOTIFY_OK;
}

2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
/*
 * Spawn the kthread that handles this RCU flavor's grace periods.
 */
static int __init rcu_spawn_gp_kthread(void)
{
	unsigned long flags;
	struct rcu_node *rnp;
	struct rcu_state *rsp;
	struct task_struct *t;

	for_each_rcu_flavor(rsp) {
		t = kthread_run(rcu_gp_kthread, rsp, rsp->name);
		BUG_ON(IS_ERR(t));
		rnp = rcu_get_root(rsp);
		raw_spin_lock_irqsave(&rnp->lock, flags);
		rsp->gp_kthread = t;
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
	}
	return 0;
}
early_initcall(rcu_spawn_gp_kthread);

2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795
/*
 * This function is invoked towards the end of the scheduler's initialization
 * process.  Before this is called, the idle task might contain
 * RCU read-side critical sections (during which time, this idle
 * task is booting the system).  After this function is called, the
 * idle tasks are prohibited from containing RCU read-side critical
 * sections.  This function also enables RCU lockdep checking.
 */
void rcu_scheduler_starting(void)
{
	WARN_ON(num_online_cpus() != 1);
	WARN_ON(nr_context_switches() > 0);
	rcu_scheduler_active = 1;
}

2796 2797 2798 2799 2800 2801 2802 2803 2804
/*
 * 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;

2805
	for (i = rcu_num_lvls - 1; i > 0; i--)
2806
		rsp->levelspread[i] = CONFIG_RCU_FANOUT;
2807
	rsp->levelspread[0] = rcu_fanout_leaf;
2808 2809 2810 2811 2812 2813 2814 2815
}
#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */
static void __init rcu_init_levelspread(struct rcu_state *rsp)
{
	int ccur;
	int cprv;
	int i;

2816
	cprv = nr_cpu_ids;
2817
	for (i = rcu_num_lvls - 1; i >= 0; i--) {
2818 2819 2820 2821 2822 2823 2824 2825 2826 2827
		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.
 */
2828 2829
static void __init rcu_init_one(struct rcu_state *rsp,
		struct rcu_data __percpu *rda)
2830
{
2831 2832 2833 2834 2835 2836 2837 2838
	static char *buf[] = { "rcu_node_0",
			       "rcu_node_1",
			       "rcu_node_2",
			       "rcu_node_3" };  /* Match MAX_RCU_LVLS */
	static char *fqs[] = { "rcu_node_fqs_0",
			       "rcu_node_fqs_1",
			       "rcu_node_fqs_2",
			       "rcu_node_fqs_3" };  /* Match MAX_RCU_LVLS */
2839 2840 2841 2842 2843
	int cpustride = 1;
	int i;
	int j;
	struct rcu_node *rnp;

2844 2845
	BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf));  /* Fix buf[] init! */

2846 2847
	/* Initialize the level-tracking arrays. */

2848 2849 2850
	for (i = 0; i < rcu_num_lvls; i++)
		rsp->levelcnt[i] = num_rcu_lvl[i];
	for (i = 1; i < rcu_num_lvls; i++)
2851 2852 2853 2854 2855
		rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];
	rcu_init_levelspread(rsp);

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

2856
	for (i = rcu_num_lvls - 1; i >= 0; i--) {
2857 2858 2859
		cpustride *= rsp->levelspread[i];
		rnp = rsp->level[i];
		for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {
P
Paul E. McKenney 已提交
2860
			raw_spin_lock_init(&rnp->lock);
2861 2862
			lockdep_set_class_and_name(&rnp->lock,
						   &rcu_node_class[i], buf[i]);
2863 2864 2865
			raw_spin_lock_init(&rnp->fqslock);
			lockdep_set_class_and_name(&rnp->fqslock,
						   &rcu_fqs_class[i], fqs[i]);
2866 2867
			rnp->gpnum = rsp->gpnum;
			rnp->completed = rsp->completed;
2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884
			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;
2885
			INIT_LIST_HEAD(&rnp->blkd_tasks);
2886 2887
		}
	}
2888

2889
	rsp->rda = rda;
2890
	init_waitqueue_head(&rsp->gp_wq);
2891
	rnp = rsp->level[rcu_num_lvls - 1];
2892
	for_each_possible_cpu(i) {
2893
		while (i > rnp->grphi)
2894
			rnp++;
2895
		per_cpu_ptr(rsp->rda, i)->mynode = rnp;
2896 2897
		rcu_boot_init_percpu_data(i, rsp);
	}
2898
	list_add(&rsp->flavors, &rcu_struct_flavors);
2899 2900
}

2901 2902 2903 2904 2905 2906 2907 2908 2909
/*
 * Compute the rcu_node tree geometry from kernel parameters.  This cannot
 * replace the definitions in rcutree.h because those are needed to size
 * the ->node array in the rcu_state structure.
 */
static void __init rcu_init_geometry(void)
{
	int i;
	int j;
2910
	int n = nr_cpu_ids;
2911 2912 2913
	int rcu_capacity[MAX_RCU_LVLS + 1];

	/* If the compile-time values are accurate, just leave. */
2914 2915
	if (rcu_fanout_leaf == CONFIG_RCU_FANOUT_LEAF &&
	    nr_cpu_ids == NR_CPUS)
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961
		return;

	/*
	 * Compute number of nodes that can be handled an rcu_node tree
	 * with the given number of levels.  Setting rcu_capacity[0] makes
	 * some of the arithmetic easier.
	 */
	rcu_capacity[0] = 1;
	rcu_capacity[1] = rcu_fanout_leaf;
	for (i = 2; i <= MAX_RCU_LVLS; i++)
		rcu_capacity[i] = rcu_capacity[i - 1] * CONFIG_RCU_FANOUT;

	/*
	 * The boot-time rcu_fanout_leaf parameter is only permitted
	 * to increase the leaf-level fanout, not decrease it.  Of course,
	 * the leaf-level fanout cannot exceed the number of bits in
	 * the rcu_node masks.  Finally, the tree must be able to accommodate
	 * the configured number of CPUs.  Complain and fall back to the
	 * compile-time values if these limits are exceeded.
	 */
	if (rcu_fanout_leaf < CONFIG_RCU_FANOUT_LEAF ||
	    rcu_fanout_leaf > sizeof(unsigned long) * 8 ||
	    n > rcu_capacity[MAX_RCU_LVLS]) {
		WARN_ON(1);
		return;
	}

	/* Calculate the number of rcu_nodes at each level of the tree. */
	for (i = 1; i <= MAX_RCU_LVLS; i++)
		if (n <= rcu_capacity[i]) {
			for (j = 0; j <= i; j++)
				num_rcu_lvl[j] =
					DIV_ROUND_UP(n, rcu_capacity[i - j]);
			rcu_num_lvls = i;
			for (j = i + 1; j <= MAX_RCU_LVLS; j++)
				num_rcu_lvl[j] = 0;
			break;
		}

	/* Calculate the total number of rcu_node structures. */
	rcu_num_nodes = 0;
	for (i = 0; i <= MAX_RCU_LVLS; i++)
		rcu_num_nodes += num_rcu_lvl[i];
	rcu_num_nodes -= n;
}

2962
void __init rcu_init(void)
2963
{
P
Paul E. McKenney 已提交
2964
	int cpu;
2965

2966
	rcu_bootup_announce();
2967
	rcu_init_geometry();
2968 2969
	rcu_init_one(&rcu_sched_state, &rcu_sched_data);
	rcu_init_one(&rcu_bh_state, &rcu_bh_data);
2970
	__rcu_init_preempt();
2971
	 open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
2972 2973 2974 2975 2976 2977 2978

	/*
	 * 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
2979 2980
	for_each_online_cpu(cpu)
		rcu_cpu_notify(NULL, CPU_UP_PREPARE, (void *)(long)cpu);
2981
	check_cpu_stall_init();
2982 2983
}

2984
#include "rcutree_plugin.h"