rcutree.c 78.9 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

56
#include "rcutree.h"
57 58 59
#include <trace/events/rcu.h>

#include "rcu.h"
60

61 62
/* Data structures. */

63
static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];
64

65
#define RCU_STATE_INITIALIZER(structname) { \
66
	.level = { &structname##_state.node[0] }, \
67 68 69 70
	.levelcnt = { \
		NUM_RCU_LVL_0,  /* root of hierarchy. */ \
		NUM_RCU_LVL_1, \
		NUM_RCU_LVL_2, \
71 72
		NUM_RCU_LVL_3, \
		NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \
73
	}, \
74
	.fqs_state = RCU_GP_IDLE, \
75 76
	.gpnum = -300, \
	.completed = -300, \
77 78
	.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.onofflock), \
	.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.fqslock), \
79 80
	.n_force_qs = 0, \
	.n_force_qs_ngp = 0, \
81
	.name = #structname, \
82 83
}

84
struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched);
85
DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
86

87
struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh);
88
DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
89

90 91
static struct rcu_state *rcu_state;

92 93 94 95 96 97 98 99 100
/*
 * 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.
 */
101 102 103
int rcu_scheduler_active __read_mostly;
EXPORT_SYMBOL_GPL(rcu_scheduler_active);

104 105 106 107 108 109 110 111 112 113 114 115 116 117
/*
 * 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;

118 119
#ifdef CONFIG_RCU_BOOST

120 121 122 123 124
/*
 * 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);
125
DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
126
DEFINE_PER_CPU(int, rcu_cpu_kthread_cpu);
127
DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);
128
DEFINE_PER_CPU(char, rcu_cpu_has_work);
129

130 131
#endif /* #ifdef CONFIG_RCU_BOOST */

132
static void rcu_node_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu);
133 134
static void invoke_rcu_core(void);
static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp);
135

136 137 138 139 140 141 142 143 144 145 146 147
/*
 * 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;

148 149 150 151 152 153 154 155 156 157
/*
 * 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);
}

158
/*
159
 * Note a quiescent state.  Because we do not need to know
160
 * how many quiescent states passed, just if there was at least
161
 * one since the start of the grace period, this just sets a flag.
162
 * The caller must have disabled preemption.
163
 */
164
void rcu_sched_qs(int cpu)
165
{
166
	struct rcu_data *rdp = &per_cpu(rcu_sched_data, cpu);
167

168
	rdp->passed_quiesce_gpnum = rdp->gpnum;
169
	barrier();
170
	if (rdp->passed_quiesce == 0)
171
		trace_rcu_grace_period("rcu_sched", rdp->gpnum, "cpuqs");
172
	rdp->passed_quiesce = 1;
173 174
}

175
void rcu_bh_qs(int cpu)
176
{
177
	struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);
178

179
	rdp->passed_quiesce_gpnum = rdp->gpnum;
180
	barrier();
181
	if (rdp->passed_quiesce == 0)
182
		trace_rcu_grace_period("rcu_bh", rdp->gpnum, "cpuqs");
183
	rdp->passed_quiesce = 1;
184
}
185

186 187 188
/*
 * Note a context switch.  This is a quiescent state for RCU-sched,
 * and requires special handling for preemptible RCU.
189
 * The caller must have disabled preemption.
190 191 192
 */
void rcu_note_context_switch(int cpu)
{
193
	trace_rcu_utilization("Start context switch");
194 195
	rcu_sched_qs(cpu);
	rcu_preempt_note_context_switch(cpu);
196
	trace_rcu_utilization("End context switch");
197
}
198
EXPORT_SYMBOL_GPL(rcu_note_context_switch);
199

200
DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
201
	.dynticks_nesting = DYNTICK_TASK_NESTING,
202
	.dynticks = ATOMIC_INIT(1),
203
};
204

205
static int blimit = 10;		/* Maximum callbacks per rcu_do_batch. */
206 207 208
static int qhimark = 10000;	/* If this many pending, ignore blimit. */
static int qlowmark = 100;	/* Once only this many pending, use blimit. */

209 210 211 212
module_param(blimit, int, 0);
module_param(qhimark, int, 0);
module_param(qlowmark, int, 0);

213 214 215
int rcu_cpu_stall_suppress __read_mostly; /* 1 = suppress stall warnings. */
int rcu_cpu_stall_timeout __read_mostly = CONFIG_RCU_CPU_STALL_TIMEOUT;

216
module_param(rcu_cpu_stall_suppress, int, 0644);
217
module_param(rcu_cpu_stall_timeout, int, 0644);
218

219
static void force_quiescent_state(struct rcu_state *rsp, int relaxed);
220
static int rcu_pending(int cpu);
221 222

/*
223
 * Return the number of RCU-sched batches processed thus far for debug & stats.
224
 */
225
long rcu_batches_completed_sched(void)
226
{
227
	return rcu_sched_state.completed;
228
}
229
EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);
230 231 232 233 234 235 236 237 238 239

/*
 * 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);

240 241 242 243 244 245 246 247 248
/*
 * Force a quiescent state for RCU BH.
 */
void rcu_bh_force_quiescent_state(void)
{
	force_quiescent_state(&rcu_bh_state, 0);
}
EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
/*
 * 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);

274 275 276 277 278 279 280 281 282
/*
 * Force a quiescent state for RCU-sched.
 */
void rcu_sched_force_quiescent_state(void)
{
	force_quiescent_state(&rcu_sched_state, 0);
}
EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
/*
 * 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)
{
298
	return *rdp->nxttail[RCU_DONE_TAIL] && !rcu_gp_in_progress(rsp);
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322
}

/*
 * 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];
}

/*
 * If the specified CPU is offline, tell the caller that it is in
 * a quiescent state.  Otherwise, whack it with a reschedule IPI.
 * Grace periods can end up waiting on an offline CPU when that
 * CPU is in the process of coming online -- it will be added to the
 * rcu_node bitmasks before it actually makes it online.  The same thing
 * can happen while a CPU is in the process of coming online.  Because this
 * race is quite rare, we check for it after detecting that the grace
 * period has been delayed rather than checking each and every CPU
 * each and every time we start a new grace period.
 */
static int rcu_implicit_offline_qs(struct rcu_data *rdp)
{
	/*
323 324 325 326 327
	 * If the CPU is offline for more than a jiffy, it is in a quiescent
	 * state.  We can trust its state not to change because interrupts
	 * are disabled.  The reason for the jiffy's worth of slack is to
	 * handle CPUs initializing on the way up and finding their way
	 * to the idle loop on the way down.
328
	 */
329 330
	if (cpu_is_offline(rdp->cpu) &&
	    ULONG_CMP_LT(rdp->rsp->gp_start + 2, jiffies)) {
331
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "ofl");
332 333 334 335 336 337
		rdp->offline_fqs++;
		return 1;
	}
	return 0;
}

338 339 340 341 342 343 344
/*
 * rcu_idle_enter_common - inform RCU that current CPU is moving towards idle
 *
 * 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.
 */
345
static void rcu_idle_enter_common(struct rcu_dynticks *rdtp, long long oldval)
346
{
347
	trace_rcu_dyntick("Start", oldval, 0);
348
	if (!is_idle_task(current)) {
349 350
		struct task_struct *idle = idle_task(smp_processor_id());

351
		trace_rcu_dyntick("Error on entry: not idle task", oldval, 0);
352
		ftrace_dump(DUMP_ALL);
353 354 355
		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! */
356
	}
357
	rcu_prepare_for_idle(smp_processor_id());
358 359 360 361 362
	/* 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);
363 364 365 366 367 368 369 370 371 372 373

	/*
	 * The idle task is not permitted to enter the idle loop while
	 * 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.");
374
}
375 376

/**
377
 * rcu_idle_enter - inform RCU that current CPU is entering idle
378
 *
379
 * Enter idle mode, in other words, -leave- the mode in which RCU
380
 * read-side critical sections can occur.  (Though RCU read-side
381 382 383 384 385 386
 * 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.
387
 */
388
void rcu_idle_enter(void)
389 390
{
	unsigned long flags;
391
	long long oldval;
392 393 394 395
	struct rcu_dynticks *rdtp;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
396
	oldval = rdtp->dynticks_nesting;
397
	rdtp->dynticks_nesting = 0;
398
	rcu_idle_enter_common(rdtp, oldval);
399 400 401
	local_irq_restore(flags);
}

402 403 404 405 406 407
/**
 * 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.
408
 *
409 410 411 412 413 414 415 416
 * 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.
417
 */
418
void rcu_irq_exit(void)
419 420
{
	unsigned long flags;
421
	long long oldval;
422 423 424 425
	struct rcu_dynticks *rdtp;

	local_irq_save(flags);
	rdtp = &__get_cpu_var(rcu_dynticks);
426
	oldval = rdtp->dynticks_nesting;
427 428
	rdtp->dynticks_nesting--;
	WARN_ON_ONCE(rdtp->dynticks_nesting < 0);
429 430 431 432
	if (rdtp->dynticks_nesting)
		trace_rcu_dyntick("--=", oldval, rdtp->dynticks_nesting);
	else
		rcu_idle_enter_common(rdtp, oldval);
433 434 435 436 437 438 439 440 441 442 443 444
	local_irq_restore(flags);
}

/*
 * rcu_idle_exit_common - inform RCU that current CPU is moving away from idle
 *
 * 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.
 */
static void rcu_idle_exit_common(struct rcu_dynticks *rdtp, long long oldval)
{
445 446 447 448 449
	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));
450
	rcu_cleanup_after_idle(smp_processor_id());
451
	trace_rcu_dyntick("End", oldval, rdtp->dynticks_nesting);
452
	if (!is_idle_task(current)) {
453 454
		struct task_struct *idle = idle_task(smp_processor_id());

455 456
		trace_rcu_dyntick("Error on exit: not idle task",
				  oldval, rdtp->dynticks_nesting);
457
		ftrace_dump(DUMP_ALL);
458 459 460
		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! */
461 462 463 464 465 466 467 468 469
	}
}

/**
 * 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.
 *
470 471
 * We crowbar the ->dynticks_nesting field to DYNTICK_TASK_NESTING to
 * allow for the possibility of usermode upcalls messing up our count
472 473 474 475 476 477 478 479 480 481 482 483 484
 * of interrupt nesting level during the busy period that is just
 * now starting.
 */
void rcu_idle_exit(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;
	WARN_ON_ONCE(oldval != 0);
485
	rdtp->dynticks_nesting = DYNTICK_TASK_NESTING;
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519
	rcu_idle_exit_common(rdtp, oldval);
	local_irq_restore(flags);
}

/**
 * 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);
520 521 522 523
	if (oldval)
		trace_rcu_dyntick("++=", oldval, rdtp->dynticks_nesting);
	else
		rcu_idle_exit_common(rdtp, oldval);
524 525 526 527 528 529 530 531 532 533 534 535 536 537
	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);

538 539
	if (rdtp->dynticks_nmi_nesting == 0 &&
	    (atomic_read(&rdtp->dynticks) & 0x1))
540
		return;
541 542 543 544 545 546
	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));
547 548 549 550 551 552 553 554 555 556 557 558 559
}

/**
 * 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);

560 561
	if (rdtp->dynticks_nmi_nesting == 0 ||
	    --rdtp->dynticks_nmi_nesting != 0)
562
		return;
563 564 565 566 567
	/* 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);
568 569
}

570 571
#ifdef CONFIG_PROVE_RCU

572
/**
573
 * rcu_is_cpu_idle - see if RCU thinks that the current CPU is idle
574
 *
575
 * If the current CPU is in its idle loop and is neither in an interrupt
576
 * or NMI handler, return true.
577
 */
578
int rcu_is_cpu_idle(void)
579
{
580 581 582 583 584 585
	int ret;

	preempt_disable();
	ret = (atomic_read(&__get_cpu_var(rcu_dynticks).dynticks) & 0x1) == 0;
	preempt_enable();
	return ret;
586
}
587
EXPORT_SYMBOL(rcu_is_cpu_idle);
588

589 590 591 592 593 594 595 596
#ifdef CONFIG_HOTPLUG_CPU

/*
 * 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
597 598 599 600 601 602 603 604 605 606 607
 * 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.
608 609 610 611 612 613
 *
 * Disable checking if in an NMI handler because we cannot safely report
 * errors from NMI handlers anyway.
 */
bool rcu_lockdep_current_cpu_online(void)
{
614 615
	struct rcu_data *rdp;
	struct rcu_node *rnp;
616 617 618 619 620
	bool ret;

	if (in_nmi())
		return 1;
	preempt_disable();
621 622 623
	rdp = &__get_cpu_var(rcu_sched_data);
	rnp = rdp->mynode;
	ret = (rdp->grpmask & rnp->qsmaskinit) ||
624 625 626 627 628 629 630 631
	      !rcu_scheduler_fully_active;
	preempt_enable();
	return ret;
}
EXPORT_SYMBOL_GPL(rcu_lockdep_current_cpu_online);

#endif /* #ifdef CONFIG_HOTPLUG_CPU */

632 633
#endif /* #ifdef CONFIG_PROVE_RCU */

634
/**
635
 * rcu_is_cpu_rrupt_from_idle - see if idle or immediately interrupted from idle
636
 *
637 638 639
 * 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.
640
 */
641
int rcu_is_cpu_rrupt_from_idle(void)
642
{
643
	return __get_cpu_var(rcu_dynticks).dynticks_nesting <= 1;
644 645 646 647 648
}

/*
 * Snapshot the specified CPU's dynticks counter so that we can later
 * credit them with an implicit quiescent state.  Return 1 if this CPU
649
 * is in dynticks idle mode, which is an extended quiescent state.
650 651 652
 */
static int dyntick_save_progress_counter(struct rcu_data *rdp)
{
653
	rdp->dynticks_snap = atomic_add_return(0, &rdp->dynticks->dynticks);
654
	return (rdp->dynticks_snap & 0x1) == 0;
655 656 657 658 659 660 661 662 663 664
}

/*
 * Return true if the specified CPU has passed through a quiescent
 * state by virtue of being in or having passed through an dynticks
 * idle state since the last call to dyntick_save_progress_counter()
 * for this same CPU.
 */
static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
{
665 666
	unsigned int curr;
	unsigned int snap;
667

668 669
	curr = (unsigned int)atomic_add_return(0, &rdp->dynticks->dynticks);
	snap = (unsigned int)rdp->dynticks_snap;
670 671 672 673 674 675 676 677 678

	/*
	 * 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.
	 */
679
	if ((curr & 0x1) == 0 || UINT_CMP_GE(curr, snap + 2)) {
680
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "dti");
681 682 683 684 685 686 687 688
		rdp->dynticks_fqs++;
		return 1;
	}

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

689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706
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;
}

707 708 709
static void record_gp_stall_check_time(struct rcu_state *rsp)
{
	rsp->gp_start = jiffies;
710
	rsp->jiffies_stall = jiffies + jiffies_till_stall_check();
711 712 713 714 715 716 717
}

static void print_other_cpu_stall(struct rcu_state *rsp)
{
	int cpu;
	long delta;
	unsigned long flags;
718
	int ndetected;
719 720 721 722
	struct rcu_node *rnp = rcu_get_root(rsp);

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

723
	raw_spin_lock_irqsave(&rnp->lock, flags);
724
	delta = jiffies - rsp->jiffies_stall;
725
	if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {
726
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
727 728
		return;
	}
729
	rsp->jiffies_stall = jiffies + 3 * jiffies_till_stall_check() + 3;
730
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
731

732 733 734 735 736
	/*
	 * OK, time to rat on our buddy...
	 * See Documentation/RCU/stallwarn.txt for info on how to debug
	 * RCU CPU stall warnings.
	 */
737
	printk(KERN_ERR "INFO: %s detected stalls on CPUs/tasks:",
738
	       rsp->name);
739
	print_cpu_stall_info_begin();
740
	rcu_for_each_leaf_node(rsp, rnp) {
741
		raw_spin_lock_irqsave(&rnp->lock, flags);
742
		ndetected += rcu_print_task_stall(rnp);
743
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
744
		if (rnp->qsmask == 0)
745
			continue;
746
		for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)
747
			if (rnp->qsmask & (1UL << cpu)) {
748
				print_cpu_stall_info(rsp, rnp->grplo + cpu);
749 750
				ndetected++;
			}
751
	}
752 753 754 755 756 757 758 759 760 761 762 763

	/*
	 * 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);
	ndetected = rcu_print_task_stall(rnp);
	raw_spin_unlock_irqrestore(&rnp->lock, flags);

	print_cpu_stall_info_end();
	printk(KERN_CONT "(detected by %d, t=%ld jiffies)\n",
764
	       smp_processor_id(), (long)(jiffies - rsp->gp_start));
765 766 767
	if (ndetected == 0)
		printk(KERN_ERR "INFO: Stall ended before state dump start\n");
	else if (!trigger_all_cpu_backtrace())
768
		dump_stack();
769

770 771 772 773
	/* If so configured, complain about tasks blocking the grace period. */

	rcu_print_detail_task_stall(rsp);

774 775 776 777 778 779 780 781
	force_quiescent_state(rsp, 0);  /* Kick them all. */
}

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

782 783 784 785 786
	/*
	 * OK, time to rat on ourselves...
	 * See Documentation/RCU/stallwarn.txt for info on how to debug
	 * RCU CPU stall warnings.
	 */
787 788 789 790 791
	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);
792 793
	if (!trigger_all_cpu_backtrace())
		dump_stack();
794

795
	raw_spin_lock_irqsave(&rnp->lock, flags);
796
	if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))
797 798
		rsp->jiffies_stall = jiffies +
				     3 * jiffies_till_stall_check() + 3;
799
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
800

801 802 803 804 805
	set_need_resched();  /* kick ourselves to get things going. */
}

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
{
806 807
	unsigned long j;
	unsigned long js;
808 809
	struct rcu_node *rnp;

810
	if (rcu_cpu_stall_suppress)
811
		return;
812 813
	j = ACCESS_ONCE(jiffies);
	js = ACCESS_ONCE(rsp->jiffies_stall);
814
	rnp = rdp->mynode;
815
	if ((ACCESS_ONCE(rnp->qsmask) & rdp->grpmask) && ULONG_CMP_GE(j, js)) {
816 817 818 819

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

820 821
	} else if (rcu_gp_in_progress(rsp) &&
		   ULONG_CMP_GE(j, js + RCU_STALL_RAT_DELAY)) {
822

823
		/* They had a few time units to dump stack, so complain. */
824 825 826 827
		print_other_cpu_stall(rsp);
	}
}

828 829
static int rcu_panic(struct notifier_block *this, unsigned long ev, void *ptr)
{
830
	rcu_cpu_stall_suppress = 1;
831 832 833
	return NOTIFY_DONE;
}

834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849
/**
 * 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)
{
	rcu_sched_state.jiffies_stall = jiffies + ULONG_MAX / 2;
	rcu_bh_state.jiffies_stall = jiffies + ULONG_MAX / 2;
	rcu_preempt_stall_reset();
}

850 851 852 853 854 855 856 857 858
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);
}

859 860 861
/*
 * 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
862 863 864
 * 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.
865
 */
866 867 868
static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
{
	if (rdp->gpnum != rnp->gpnum) {
869 870 871 872 873
		/*
		 * If the current grace period is waiting for this CPU,
		 * set up to detect a quiescent state, otherwise don't
		 * go looking for one.
		 */
874
		rdp->gpnum = rnp->gpnum;
875
		trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpustart");
876 877
		if (rnp->qsmask & rdp->grpmask) {
			rdp->qs_pending = 1;
878
			rdp->passed_quiesce = 0;
879 880
		} else
			rdp->qs_pending = 0;
881
		zero_cpu_stall_ticks(rdp);
882 883 884
	}
}

885 886
static void note_new_gpnum(struct rcu_state *rsp, struct rcu_data *rdp)
{
887 888 889 890 891 892
	unsigned long flags;
	struct rcu_node *rnp;

	local_irq_save(flags);
	rnp = rdp->mynode;
	if (rdp->gpnum == ACCESS_ONCE(rnp->gpnum) || /* outside lock. */
893
	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
894 895 896 897
		local_irq_restore(flags);
		return;
	}
	__note_new_gpnum(rsp, rnp, rdp);
898
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920
}

/*
 * 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;
}

921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939
/*
 * 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;
940
		trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuend");
941

942 943
		/*
		 * If we were in an extended quiescent state, we may have
944
		 * missed some grace periods that others CPUs handled on
945
		 * our behalf. Catch up with this state to avoid noting
946 947 948
		 * spurious new grace periods.  If another grace period
		 * has started, then rnp->gpnum will have advanced, so
		 * we will detect this later on.
949
		 */
950
		if (ULONG_CMP_LT(rdp->gpnum, rdp->completed))
951 952
			rdp->gpnum = rdp->completed;

953
		/*
954 955
		 * If RCU does not need a quiescent state from this CPU,
		 * then make sure that this CPU doesn't go looking for one.
956
		 */
957
		if ((rnp->qsmask & rdp->grpmask) == 0)
958
			rdp->qs_pending = 0;
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
	}
}

/*
 * 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. */
976
	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
977 978 979 980
		local_irq_restore(flags);
		return;
	}
	__rcu_process_gp_end(rsp, rnp, rdp);
981
	raw_spin_unlock_irqrestore(&rnp->lock, flags);
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007
}

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

	/*
	 * Because this CPU just now started the new grace period, we know
	 * that all of its callbacks will be covered by this upcoming grace
	 * period, even the ones that were registered arbitrarily recently.
	 * Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.
	 *
	 * Other CPUs cannot be sure exactly when the grace period started.
	 * Therefore, their recently registered callbacks must pass through
	 * an additional RCU_NEXT_READY stage, so that they will be handled
	 * by the next RCU grace period.
	 */
	rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
	rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
1008 1009 1010

	/* Set state so that this CPU will detect the next quiescent state. */
	__note_new_gpnum(rsp, rnp, rdp);
1011 1012
}

1013 1014 1015 1016 1017
/*
 * 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.
1018 1019 1020 1021
 *
 * 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.
1022 1023 1024 1025 1026
 */
static void
rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
	__releases(rcu_get_root(rsp)->lock)
{
1027
	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
1028 1029
	struct rcu_node *rnp = rcu_get_root(rsp);

1030
	if (!rcu_scheduler_fully_active ||
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
	    !cpu_needs_another_gp(rsp, rdp)) {
		/*
		 * Either the scheduler hasn't yet spawned the first
		 * non-idle task or this CPU does not need another
		 * grace period.  Either way, don't start a new grace
		 * period.
		 */
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
		return;
	}
1041

1042
	if (rsp->fqs_active) {
1043
		/*
1044 1045
		 * This CPU needs a grace period, but force_quiescent_state()
		 * is running.  Tell it to start one on this CPU's behalf.
1046
		 */
1047 1048
		rsp->fqs_need_gp = 1;
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1049 1050 1051 1052 1053
		return;
	}

	/* Advance to a new grace period and initialize state. */
	rsp->gpnum++;
1054
	trace_rcu_grace_period(rsp->name, rsp->gpnum, "start");
1055 1056
	WARN_ON_ONCE(rsp->fqs_state == RCU_GP_INIT);
	rsp->fqs_state = RCU_GP_INIT; /* Hold off force_quiescent_state. */
1057 1058
	rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
	record_gp_stall_check_time(rsp);
1059
	raw_spin_unlock(&rnp->lock);  /* leave irqs disabled. */
1060 1061

	/* Exclude any concurrent CPU-hotplug operations. */
1062
	raw_spin_lock(&rsp->onofflock);  /* irqs already disabled. */
1063 1064

	/*
1065 1066 1067 1068 1069 1070 1071 1072 1073
	 * Set the quiescent-state-needed bits in all the rcu_node
	 * structures for all currently online CPUs in breadth-first
	 * order, starting from the root rcu_node structure.  This
	 * operation relies on the layout of the hierarchy within the
	 * rsp->node[] array.  Note that other CPUs will access only
	 * the leaves of the hierarchy, which still indicate that no
	 * grace period is in progress, at least until the corresponding
	 * leaf node has been initialized.  In addition, we have excluded
	 * CPU-hotplug operations.
1074 1075 1076 1077
	 *
	 * Note that the grace period cannot complete until we finish
	 * the initialization process, as there will be at least one
	 * qsmask bit set in the root node until that time, namely the
1078 1079
	 * one corresponding to this CPU, due to the fact that we have
	 * irqs disabled.
1080
	 */
1081
	rcu_for_each_node_breadth_first(rsp, rnp) {
1082
		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
1083
		rcu_preempt_check_blocked_tasks(rnp);
1084
		rnp->qsmask = rnp->qsmaskinit;
1085
		rnp->gpnum = rsp->gpnum;
1086 1087 1088
		rnp->completed = rsp->completed;
		if (rnp == rdp->mynode)
			rcu_start_gp_per_cpu(rsp, rnp, rdp);
1089
		rcu_preempt_boost_start_gp(rnp);
1090 1091 1092
		trace_rcu_grace_period_init(rsp->name, rnp->gpnum,
					    rnp->level, rnp->grplo,
					    rnp->grphi, rnp->qsmask);
1093
		raw_spin_unlock(&rnp->lock);	/* irqs remain disabled. */
1094 1095
	}

1096
	rnp = rcu_get_root(rsp);
1097
	raw_spin_lock(&rnp->lock);		/* irqs already disabled. */
1098
	rsp->fqs_state = RCU_SIGNAL_INIT; /* force_quiescent_state now OK. */
1099 1100
	raw_spin_unlock(&rnp->lock);		/* irqs remain disabled. */
	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
1101 1102
}

1103
/*
1104 1105 1106 1107 1108
 * 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.
1109
 */
1110
static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)
1111
	__releases(rcu_get_root(rsp)->lock)
1112
{
1113
	unsigned long gp_duration;
1114 1115
	struct rcu_node *rnp = rcu_get_root(rsp);
	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
1116

1117
	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
1118 1119 1120 1121 1122 1123

	/*
	 * Ensure that all grace-period and pre-grace-period activity
	 * is seen before the assignment to rsp->completed.
	 */
	smp_mb(); /* See above block comment. */
1124 1125 1126
	gp_duration = jiffies - rsp->gp_start;
	if (gp_duration > rsp->gp_max)
		rsp->gp_max = gp_duration;
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160

	/*
	 * 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.
	 *
	 * But if this CPU needs another grace period, it will take
	 * care of this while initializing the next grace period.
	 * We use RCU_WAIT_TAIL instead of the usual RCU_DONE_TAIL
	 * because the callbacks have not yet been advanced: Those
	 * callbacks are waiting on the grace period that just now
	 * completed.
	 */
	if (*rdp->nxttail[RCU_WAIT_TAIL] == NULL) {
		raw_spin_unlock(&rnp->lock);	 /* irqs remain disabled. */

		/*
		 * Propagate new ->completed value to rcu_node structures
		 * so that other CPUs don't have to wait until the start
		 * of the next grace period to process their callbacks.
		 */
		rcu_for_each_node_breadth_first(rsp, rnp) {
			raw_spin_lock(&rnp->lock); /* irqs already disabled. */
			rnp->completed = rsp->gpnum;
			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
		}
		rnp = rcu_get_root(rsp);
		raw_spin_lock(&rnp->lock); /* irqs already disabled. */
	}

	rsp->completed = rsp->gpnum;  /* Declare the grace period complete. */
1161
	trace_rcu_grace_period(rsp->name, rsp->completed, "end");
1162
	rsp->fqs_state = RCU_GP_IDLE;
1163 1164 1165
	rcu_start_gp(rsp, flags);  /* releases root node's rnp->lock. */
}

1166
/*
1167 1168 1169 1170 1171 1172
 * 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.
1173 1174
 */
static void
1175 1176
rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,
		  struct rcu_node *rnp, unsigned long flags)
1177 1178
	__releases(rnp->lock)
{
1179 1180
	struct rcu_node *rnp_c;

1181 1182 1183 1184 1185
	/* Walk up the rcu_node hierarchy. */
	for (;;) {
		if (!(rnp->qsmask & mask)) {

			/* Our bit has already been cleared, so done. */
1186
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1187 1188 1189
			return;
		}
		rnp->qsmask &= ~mask;
1190 1191 1192 1193
		trace_rcu_quiescent_state_report(rsp->name, rnp->gpnum,
						 mask, rnp->qsmask, rnp->level,
						 rnp->grplo, rnp->grphi,
						 !!rnp->gp_tasks);
1194
		if (rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {
1195 1196

			/* Other bits still set at this level, so done. */
1197
			raw_spin_unlock_irqrestore(&rnp->lock, flags);
1198 1199 1200 1201 1202 1203 1204 1205 1206
			return;
		}
		mask = rnp->grpmask;
		if (rnp->parent == NULL) {

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

			break;
		}
1207
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1208
		rnp_c = rnp;
1209
		rnp = rnp->parent;
1210
		raw_spin_lock_irqsave(&rnp->lock, flags);
1211
		WARN_ON_ONCE(rnp_c->qsmask);
1212 1213 1214 1215
	}

	/*
	 * Get here if we are the last CPU to pass through a quiescent
1216
	 * state for this grace period.  Invoke rcu_report_qs_rsp()
1217
	 * to clean up and start the next grace period if one is needed.
1218
	 */
1219
	rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */
1220 1221 1222
}

/*
1223 1224 1225 1226 1227 1228 1229
 * 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!
1230 1231
 */
static void
1232
rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp, long lastgp)
1233 1234 1235 1236 1237 1238
{
	unsigned long flags;
	unsigned long mask;
	struct rcu_node *rnp;

	rnp = rdp->mynode;
1239
	raw_spin_lock_irqsave(&rnp->lock, flags);
1240
	if (lastgp != rnp->gpnum || rnp->completed == rnp->gpnum) {
1241 1242

		/*
1243 1244 1245 1246
		 * 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.
1247
		 */
1248
		rdp->passed_quiesce = 0;	/* need qs for new gp. */
1249
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1250 1251 1252 1253
		return;
	}
	mask = rdp->grpmask;
	if ((rnp->qsmask & mask) == 0) {
1254
		raw_spin_unlock_irqrestore(&rnp->lock, flags);
1255 1256 1257 1258 1259 1260 1261 1262 1263
	} 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];

1264
		rcu_report_qs_rnp(mask, rsp, rnp, flags); /* rlses rnp->lock */
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
	}
}

/*
 * 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.
	 */
1292
	if (!rdp->passed_quiesce)
1293 1294
		return;

1295 1296 1297 1298
	/*
	 * Tell RCU we are done (but rcu_report_qs_rdp() will be the
	 * judge of that).
	 */
1299
	rcu_report_qs_rdp(rdp->cpu, rsp, rdp, rdp->passed_quiesce_gpnum);
1300 1301 1302 1303
}

#ifdef CONFIG_HOTPLUG_CPU

1304
/*
1305
 * Move a dying CPU's RCU callbacks to online CPU's callback list.
1306 1307 1308 1309 1310 1311 1312 1313 1314
 * Also record a quiescent state for this CPU for the current grace period.
 * Synchronization and interrupt disabling are not required because
 * this function executes in stop_machine() context.  Therefore, cleanup
 * operations that might block must be done later from the CPU_DEAD
 * notifier.
 *
 * Note that the outgoing CPU's bit has already been cleared in the
 * cpu_online_mask.  This allows us to randomly pick a callback
 * destination from the bits set in that mask.
1315
 */
1316
static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
1317 1318
{
	int i;
1319
	unsigned long mask;
1320
	int receive_cpu = cpumask_any(cpu_online_mask);
1321
	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
1322
	struct rcu_data *receive_rdp = per_cpu_ptr(rsp->rda, receive_cpu);
1323
	RCU_TRACE(struct rcu_node *rnp = rdp->mynode); /* For dying CPU. */
1324

1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364
	/* First, adjust the counts. */
	if (rdp->nxtlist != NULL) {
		receive_rdp->qlen_lazy += rdp->qlen_lazy;
		receive_rdp->qlen += rdp->qlen;
		rdp->qlen_lazy = 0;
		rdp->qlen = 0;
	}

	/*
	 * Next, move ready-to-invoke callbacks to be invoked on some
	 * other CPU.  These will not be required to pass through another
	 * grace period:  They are done, regardless of CPU.
	 */
	if (rdp->nxtlist != NULL &&
	    rdp->nxttail[RCU_DONE_TAIL] != &rdp->nxtlist) {
		struct rcu_head *oldhead;
		struct rcu_head **oldtail;
		struct rcu_head **newtail;

		oldhead = rdp->nxtlist;
		oldtail = receive_rdp->nxttail[RCU_DONE_TAIL];
		rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];
		*rdp->nxttail[RCU_DONE_TAIL] = *oldtail;
		*receive_rdp->nxttail[RCU_DONE_TAIL] = oldhead;
		newtail = rdp->nxttail[RCU_DONE_TAIL];
		for (i = RCU_DONE_TAIL; i < RCU_NEXT_SIZE; i++) {
			if (receive_rdp->nxttail[i] == oldtail)
				receive_rdp->nxttail[i] = newtail;
			if (rdp->nxttail[i] == newtail)
				rdp->nxttail[i] = &rdp->nxtlist;
		}
	}

	/*
	 * Finally, put the rest of the callbacks at the end of the list.
	 * The ones that made it partway through get to start over:  We
	 * cannot assume that grace periods are synchronized across CPUs.
	 * (We could splice RCU_WAIT_TAIL into RCU_NEXT_READY_TAIL, but
	 * this does not seem compelling.  Not yet, anyway.)
	 */
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375
	if (rdp->nxtlist != NULL) {
		*receive_rdp->nxttail[RCU_NEXT_TAIL] = rdp->nxtlist;
		receive_rdp->nxttail[RCU_NEXT_TAIL] =
				rdp->nxttail[RCU_NEXT_TAIL];
		receive_rdp->n_cbs_adopted += rdp->qlen;
		rdp->n_cbs_orphaned += rdp->qlen;

		rdp->nxtlist = NULL;
		for (i = 0; i < RCU_NEXT_SIZE; i++)
			rdp->nxttail[i] = &rdp->nxtlist;
	}
1376

1377 1378 1379 1380 1381 1382
	/*
	 * Record a quiescent state for the dying CPU.  This is safe
	 * only because we have already cleared out the callbacks.
	 * (Otherwise, the RCU core might try to schedule the invocation
	 * of callbacks on this now-offline CPU, which would be bad.)
	 */
1383
	mask = rdp->grpmask;	/* rnp->grplo is constant. */
1384 1385 1386 1387 1388
	trace_rcu_grace_period(rsp->name,
			       rnp->gpnum + 1 - !!(rnp->qsmask & mask),
			       "cpuofl");
	rcu_report_qs_rdp(smp_processor_id(), rsp, rdp, rsp->gpnum);
	/* Note that rcu_report_qs_rdp() might call trace_rcu_grace_period(). */
1389 1390 1391
}

/*
1392 1393 1394 1395
 * The CPU has been completely removed, and some other CPU is reporting
 * this fact from process context.  Do the remainder of the cleanup.
 * There can only be one CPU hotplug operation at a time, so no other
 * CPU can be attempting to update rcu_cpu_kthread_task.
1396
 */
1397
static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
1398
{
1399 1400 1401
	unsigned long flags;
	unsigned long mask;
	int need_report = 0;
1402
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
1403
	struct rcu_node *rnp = rdp->mynode;  /* Outgoing CPU's rnp. */
1404

1405
	/* Adjust any no-longer-needed kthreads. */
1406 1407
	rcu_stop_cpu_kthread(cpu);
	rcu_node_kthread_setaffinity(rnp, -1);
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445

	/* Remove the dying CPU from the bitmasks in the rcu_node hierarchy. */

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

	/* 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);
1446 1447 1448 1449
}

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1450
static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
1451 1452 1453
{
}

1454
static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
1455 1456 1457 1458 1459 1460 1461 1462 1463
{
}

#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.
 */
1464
static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)
1465 1466 1467
{
	unsigned long flags;
	struct rcu_head *next, *list, **tail;
1468
	int bl, count, count_lazy;
Paul E. McKenney's avatar