tree.c 132.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 * 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
15 16
 * along with this program; if not, you can access it online at
 * http://www.gnu.org/licenses/gpl-2.0.html.
17 18 19 20 21 22 23 24 25 26 27
 *
 * 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
 */
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/smp.h>
35
#include <linux/rcupdate_wait.h>
36 37
#include <linux/interrupt.h>
#include <linux/sched.h>
38
#include <linux/sched/debug.h>
39
#include <linux/nmi.h>
40
#include <linux/atomic.h>
41
#include <linux/bitops.h>
42
#include <linux/export.h>
43 44 45 46 47 48 49
#include <linux/completion.h>
#include <linux/moduleparam.h>
#include <linux/percpu.h>
#include <linux/notifier.h>
#include <linux/cpu.h>
#include <linux/mutex.h>
#include <linux/time.h>
50
#include <linux/kernel_stat.h>
51 52
#include <linux/wait.h>
#include <linux/kthread.h>
53
#include <uapi/linux/sched/types.h>
54
#include <linux/prefetch.h>
55 56
#include <linux/delay.h>
#include <linux/stop_machine.h>
57
#include <linux/random.h>
58
#include <linux/trace_events.h>
59
#include <linux/suspend.h>
P
Paul E. McKenney 已提交
60
#include <linux/ftrace.h>
61

62
#include "tree.h"
63
#include "rcu.h"
64

65 66 67 68 69
#ifdef MODULE_PARAM_PREFIX
#undef MODULE_PARAM_PREFIX
#endif
#define MODULE_PARAM_PREFIX "rcutree."

70 71
/* Data structures. */

72 73 74 75 76 77 78 79
/*
 * In order to export the rcu_state name to the tracing tools, it
 * needs to be added in the __tracepoint_string section.
 * This requires defining a separate variable tp_<sname>_varname
 * that points to the string being used, and this will allow
 * the tracing userspace tools to be able to decipher the string
 * address to the matching string.
 */
80 81
#ifdef CONFIG_TRACING
# define DEFINE_RCU_TPS(sname) \
82
static char sname##_varname[] = #sname; \
83 84 85 86 87 88 89 90 91
static const char *tp_##sname##_varname __used __tracepoint_string = sname##_varname;
# define RCU_STATE_NAME(sname) sname##_varname
#else
# define DEFINE_RCU_TPS(sname)
# define RCU_STATE_NAME(sname) __stringify(sname)
#endif

#define RCU_STATE_INITIALIZER(sname, sabbr, cr) \
DEFINE_RCU_TPS(sname) \
92
static DEFINE_PER_CPU_SHARED_ALIGNED(struct rcu_data, sname##_data); \
93
struct rcu_state sname##_state = { \
94
	.level = { &sname##_state.node[0] }, \
95
	.rda = &sname##_data, \
96
	.call = cr, \
97
	.gp_state = RCU_GP_IDLE, \
P
Paul E. McKenney 已提交
98 99
	.gpnum = 0UL - 300UL, \
	.completed = 0UL - 300UL, \
100
	.orphan_lock = __RAW_SPIN_LOCK_UNLOCKED(&sname##_state.orphan_lock), \
101 102
	.orphan_pend = RCU_CBLIST_INITIALIZER(sname##_state.orphan_pend), \
	.orphan_done = RCU_CBLIST_INITIALIZER(sname##_state.orphan_done), \
103
	.barrier_mutex = __MUTEX_INITIALIZER(sname##_state.barrier_mutex), \
104
	.name = RCU_STATE_NAME(sname), \
105
	.abbr = sabbr, \
106
	.exp_mutex = __MUTEX_INITIALIZER(sname##_state.exp_mutex), \
107
	.exp_wake_mutex = __MUTEX_INITIALIZER(sname##_state.exp_wake_mutex), \
108
}
109

110 111
RCU_STATE_INITIALIZER(rcu_sched, 's', call_rcu_sched);
RCU_STATE_INITIALIZER(rcu_bh, 'b', call_rcu_bh);
112

113
static struct rcu_state *const rcu_state_p;
114
LIST_HEAD(rcu_struct_flavors);
115

116 117 118
/* Dump rcu_node combining tree at boot to verify correct setup. */
static bool dump_tree;
module_param(dump_tree, bool, 0444);
119 120 121
/* Control rcu_node-tree auto-balancing at boot time. */
static bool rcu_fanout_exact;
module_param(rcu_fanout_exact, bool, 0444);
122 123
/* Increase (but not decrease) the RCU_FANOUT_LEAF at boot time. */
static int rcu_fanout_leaf = RCU_FANOUT_LEAF;
124
module_param(rcu_fanout_leaf, int, 0444);
125
int rcu_num_lvls __read_mostly = RCU_NUM_LVLS;
126
/* Number of rcu_nodes at specified level. */
127
int num_rcu_lvl[] = NUM_RCU_LVL_INIT;
128
int rcu_num_nodes __read_mostly = NUM_RCU_NODES; /* Total # rcu_nodes in use. */
129 130
/* panic() on RCU Stall sysctl. */
int sysctl_panic_on_rcu_stall __read_mostly;
131

132
/*
133 134 135 136
 * The rcu_scheduler_active variable is initialized to the value
 * RCU_SCHEDULER_INACTIVE and transitions RCU_SCHEDULER_INIT just before the
 * first task is spawned.  So when this variable is RCU_SCHEDULER_INACTIVE,
 * RCU can assume that there is but one task, allowing RCU to (for example)
137
 * optimize synchronize_rcu() to a simple barrier().  When this variable
138 139 140 141 142
 * is RCU_SCHEDULER_INIT, 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.  Finally, it
 * transitions from RCU_SCHEDULER_INIT to RCU_SCHEDULER_RUNNING after RCU
 * is fully initialized, including all of its kthreads having been spawned.
143
 */
144 145 146
int rcu_scheduler_active __read_mostly;
EXPORT_SYMBOL_GPL(rcu_scheduler_active);

147 148 149 150 151 152 153 154 155 156 157 158 159 160
/*
 * 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;

161 162
static void rcu_init_new_rnp(struct rcu_node *rnp_leaf);
static void rcu_cleanup_dead_rnp(struct rcu_node *rnp_leaf);
T
Thomas Gleixner 已提交
163
static void rcu_boost_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu);
164 165
static void invoke_rcu_core(void);
static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp);
166 167
static void rcu_report_exp_rdp(struct rcu_state *rsp,
			       struct rcu_data *rdp, bool wake);
168
static void sync_sched_exp_online_cleanup(int cpu);
169

170
/* rcuc/rcub kthread realtime priority */
171
static int kthread_prio = IS_ENABLED(CONFIG_RCU_BOOST) ? 1 : 0;
172 173
module_param(kthread_prio, int, 0644);

174
/* Delay in jiffies for grace-period initialization delays, debug only. */
175

176 177 178 179 180 181
static int gp_preinit_delay;
module_param(gp_preinit_delay, int, 0444);
static int gp_init_delay;
module_param(gp_init_delay, int, 0444);
static int gp_cleanup_delay;
module_param(gp_cleanup_delay, int, 0444);
182

183 184
/*
 * Number of grace periods between delays, normalized by the duration of
185
 * the delay.  The longer the delay, the more the grace periods between
186 187 188 189 190 191 192
 * each delay.  The reason for this normalization is that it means that,
 * for non-zero delays, the overall slowdown of grace periods is constant
 * regardless of the duration of the delay.  This arrangement balances
 * the need for long delays to increase some race probabilities with the
 * need for fast grace periods to increase other race probabilities.
 */
#define PER_RCU_NODE_PERIOD 3	/* Number of grace periods between delays. */
193

194 195 196 197 198 199 200 201 202 203 204 205
/*
 * 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;

206 207 208 209 210 211 212 213
/*
 * Compute the mask of online CPUs for the specified rcu_node structure.
 * This will not be stable unless the rcu_node structure's ->lock is
 * held, but the bit corresponding to the current CPU will be stable
 * in most contexts.
 */
unsigned long rcu_rnp_online_cpus(struct rcu_node *rnp)
{
214
	return READ_ONCE(rnp->qsmaskinitnext);
215 216
}

217
/*
218
 * Return true if an RCU grace period is in progress.  The READ_ONCE()s
219 220 221 222 223
 * 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)
{
224
	return READ_ONCE(rsp->completed) != READ_ONCE(rsp->gpnum);
225 226
}

227
/*
228
 * Note a quiescent state.  Because we do not need to know
229
 * how many quiescent states passed, just if there was at least
230
 * one since the start of the grace period, this just sets a flag.
231
 * The caller must have disabled preemption.
232
 */
233
void rcu_sched_qs(void)
234
{
235
	RCU_LOCKDEP_WARN(preemptible(), "rcu_sched_qs() invoked with preemption enabled!!!");
236 237 238 239 240 241 242 243
	if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.s))
		return;
	trace_rcu_grace_period(TPS("rcu_sched"),
			       __this_cpu_read(rcu_sched_data.gpnum),
			       TPS("cpuqs"));
	__this_cpu_write(rcu_sched_data.cpu_no_qs.b.norm, false);
	if (!__this_cpu_read(rcu_sched_data.cpu_no_qs.b.exp))
		return;
244 245 246
	__this_cpu_write(rcu_sched_data.cpu_no_qs.b.exp, false);
	rcu_report_exp_rdp(&rcu_sched_state,
			   this_cpu_ptr(&rcu_sched_data), true);
247 248
}

249
void rcu_bh_qs(void)
250
{
251
	RCU_LOCKDEP_WARN(preemptible(), "rcu_bh_qs() invoked with preemption enabled!!!");
252
	if (__this_cpu_read(rcu_bh_data.cpu_no_qs.s)) {
253 254 255
		trace_rcu_grace_period(TPS("rcu_bh"),
				       __this_cpu_read(rcu_bh_data.gpnum),
				       TPS("cpuqs"));
256
		__this_cpu_write(rcu_bh_data.cpu_no_qs.b.norm, false);
257
	}
258
}
259

260 261 262 263 264 265 266 267 268
/*
 * Steal a bit from the bottom of ->dynticks for idle entry/exit
 * control.  Initially this is for TLB flushing.
 */
#define RCU_DYNTICK_CTRL_MASK 0x1
#define RCU_DYNTICK_CTRL_CTR  (RCU_DYNTICK_CTRL_MASK + 1)
#ifndef rcu_eqs_special_exit
#define rcu_eqs_special_exit() do { } while (0)
#endif
269 270 271

static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
	.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,
272
	.dynticks = ATOMIC_INIT(RCU_DYNTICK_CTRL_CTR),
273 274
};

275 276 277 278 279 280 281 282 283 284 285 286 287 288
/*
 * There's a few places, currently just in the tracing infrastructure,
 * that uses rcu_irq_enter() to make sure RCU is watching. But there's
 * a small location where that will not even work. In those cases
 * rcu_irq_enter_disabled() needs to be checked to make sure rcu_irq_enter()
 * can be called.
 */
static DEFINE_PER_CPU(bool, disable_rcu_irq_enter);

bool rcu_irq_enter_disabled(void)
{
	return this_cpu_read(disable_rcu_irq_enter);
}

289 290 291 292 293 294 295
/*
 * Record entry into an extended quiescent state.  This is only to be
 * called when not already in an extended quiescent state.
 */
static void rcu_dynticks_eqs_enter(void)
{
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
296
	int seq;
297 298

	/*
299
	 * CPUs seeing atomic_add_return() must see prior RCU read-side
300 301 302
	 * critical sections, and we also must force ordering with the
	 * next idle sojourn.
	 */
303 304 305 306 307 308 309
	seq = atomic_add_return(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
	/* Better be in an extended quiescent state! */
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
		     (seq & RCU_DYNTICK_CTRL_CTR));
	/* Better not have special action (TLB flush) pending! */
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
		     (seq & RCU_DYNTICK_CTRL_MASK));
310 311 312 313 314 315 316 317 318
}

/*
 * Record exit from an extended quiescent state.  This is only to be
 * called from an extended quiescent state.
 */
static void rcu_dynticks_eqs_exit(void)
{
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
319
	int seq;
320 321

	/*
322
	 * CPUs seeing atomic_add_return() must see prior idle sojourns,
323 324 325
	 * and we also must force ordering with the next RCU read-side
	 * critical section.
	 */
326 327 328 329 330 331 332 333 334
	seq = atomic_add_return(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
		     !(seq & RCU_DYNTICK_CTRL_CTR));
	if (seq & RCU_DYNTICK_CTRL_MASK) {
		atomic_andnot(RCU_DYNTICK_CTRL_MASK, &rdtp->dynticks);
		smp_mb__after_atomic(); /* _exit after clearing mask. */
		/* Prefer duplicate flushes to losing a flush. */
		rcu_eqs_special_exit();
	}
335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350
}

/*
 * Reset the current CPU's ->dynticks counter to indicate that the
 * newly onlined CPU is no longer in an extended quiescent state.
 * This will either leave the counter unchanged, or increment it
 * to the next non-quiescent value.
 *
 * The non-atomic test/increment sequence works because the upper bits
 * of the ->dynticks counter are manipulated only by the corresponding CPU,
 * or when the corresponding CPU is offline.
 */
static void rcu_dynticks_eqs_online(void)
{
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);

351
	if (atomic_read(&rdtp->dynticks) & RCU_DYNTICK_CTRL_CTR)
352
		return;
353
	atomic_add(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
354 355
}

356 357 358 359 360 361 362 363 364
/*
 * Is the current CPU in an extended quiescent state?
 *
 * No ordering, as we are sampling CPU-local information.
 */
bool rcu_dynticks_curr_cpu_in_eqs(void)
{
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);

365
	return !(atomic_read(&rdtp->dynticks) & RCU_DYNTICK_CTRL_CTR);
366 367
}

368 369 370 371
/*
 * Snapshot the ->dynticks counter with full ordering so as to allow
 * stable comparison of this counter with past and future snapshots.
 */
372
int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
373 374 375
{
	int snap = atomic_add_return(0, &rdtp->dynticks);

376
	return snap & ~RCU_DYNTICK_CTRL_MASK;
377 378
}

379 380 381 382 383 384
/*
 * Return true if the snapshot returned from rcu_dynticks_snap()
 * indicates that RCU is in an extended quiescent state.
 */
static bool rcu_dynticks_in_eqs(int snap)
{
385
	return !(snap & RCU_DYNTICK_CTRL_CTR);
386 387 388 389 390 391 392 393 394 395 396 397
}

/*
 * Return true if the CPU corresponding to the specified rcu_dynticks
 * structure has spent some time in an extended quiescent state since
 * rcu_dynticks_snap() returned the specified snapshot.
 */
static bool rcu_dynticks_in_eqs_since(struct rcu_dynticks *rdtp, int snap)
{
	return snap != rcu_dynticks_snap(rdtp);
}

398 399 400 401 402 403 404
/*
 * Do a double-increment of the ->dynticks counter to emulate a
 * momentary idle-CPU quiescent state.
 */
static void rcu_dynticks_momentary_idle(void)
{
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
405 406
	int special = atomic_add_return(2 * RCU_DYNTICK_CTRL_CTR,
					&rdtp->dynticks);
407 408

	/* It is illegal to call this from idle state. */
409
	WARN_ON_ONCE(!(special & RCU_DYNTICK_CTRL_CTR));
410 411
}

412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
/*
 * Set the special (bottom) bit of the specified CPU so that it
 * will take special action (such as flushing its TLB) on the
 * next exit from an extended quiescent state.  Returns true if
 * the bit was successfully set, or false if the CPU was not in
 * an extended quiescent state.
 */
bool rcu_eqs_special_set(int cpu)
{
	int old;
	int new;
	struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);

	do {
		old = atomic_read(&rdtp->dynticks);
		if (old & RCU_DYNTICK_CTRL_CTR)
			return false;
		new = old | RCU_DYNTICK_CTRL_MASK;
	} while (atomic_cmpxchg(&rdtp->dynticks, old, new) != old);
	return true;
432
}
433

434 435 436 437 438 439 440
/*
 * Let the RCU core know that this CPU has gone through the scheduler,
 * which is a quiescent state.  This is called when the need for a
 * quiescent state is urgent, so we burn an atomic operation and full
 * memory barriers to let the RCU core know about it, regardless of what
 * this CPU might (or might not) do in the near future.
 *
441
 * We inform the RCU core by emulating a zero-duration dyntick-idle period.
442 443
 *
 * The caller must have disabled interrupts.
444 445 446
 */
static void rcu_momentary_dyntick_idle(void)
{
447 448
	raw_cpu_write(rcu_dynticks.rcu_need_heavy_qs, false);
	rcu_dynticks_momentary_idle();
449 450
}

451 452 453
/*
 * Note a context switch.  This is a quiescent state for RCU-sched,
 * and requires special handling for preemptible RCU.
454
 * The caller must have disabled interrupts.
455
 */
456
void rcu_note_context_switch(bool preempt)
457
{
458
	barrier(); /* Avoid RCU read-side critical sections leaking down. */
459
	trace_rcu_utilization(TPS("Start context switch"));
460
	rcu_sched_qs();
461
	rcu_preempt_note_context_switch(preempt);
462 463 464 465
	/* Load rcu_urgent_qs before other flags. */
	if (!smp_load_acquire(this_cpu_ptr(&rcu_dynticks.rcu_urgent_qs)))
		goto out;
	this_cpu_write(rcu_dynticks.rcu_urgent_qs, false);
466
	if (unlikely(raw_cpu_read(rcu_dynticks.rcu_need_heavy_qs)))
467
		rcu_momentary_dyntick_idle();
468
	this_cpu_inc(rcu_dynticks.rcu_qs_ctr);
469 470
	if (!preempt)
		rcu_note_voluntary_context_switch_lite(current);
471
out:
472
	trace_rcu_utilization(TPS("End context switch"));
473
	barrier(); /* Avoid RCU read-side critical sections leaking up. */
474
}
475
EXPORT_SYMBOL_GPL(rcu_note_context_switch);
476

477
/*
478
 * Register a quiescent state for all RCU flavors.  If there is an
479 480
 * emergency, invoke rcu_momentary_dyntick_idle() to do a heavy-weight
 * dyntick-idle quiescent state visible to other CPUs (but only for those
481
 * RCU flavors in desperate need of a quiescent state, which will normally
482 483
 * be none of them).  Either way, do a lightweight quiescent state for
 * all RCU flavors.
484 485 486 487 488
 *
 * The barrier() calls are redundant in the common case when this is
 * called externally, but just in case this is called from within this
 * file.
 *
489 490 491
 */
void rcu_all_qs(void)
{
492 493
	unsigned long flags;

494 495 496 497 498 499 500 501 502
	if (!raw_cpu_read(rcu_dynticks.rcu_urgent_qs))
		return;
	preempt_disable();
	/* Load rcu_urgent_qs before other flags. */
	if (!smp_load_acquire(this_cpu_ptr(&rcu_dynticks.rcu_urgent_qs))) {
		preempt_enable();
		return;
	}
	this_cpu_write(rcu_dynticks.rcu_urgent_qs, false);
503
	barrier(); /* Avoid RCU read-side critical sections leaking down. */
504
	if (unlikely(raw_cpu_read(rcu_dynticks.rcu_need_heavy_qs))) {
505
		local_irq_save(flags);
506
		rcu_momentary_dyntick_idle();
507 508
		local_irq_restore(flags);
	}
509
	if (unlikely(raw_cpu_read(rcu_sched_data.cpu_no_qs.b.exp)))
510
		rcu_sched_qs();
511
	this_cpu_inc(rcu_dynticks.rcu_qs_ctr);
512
	barrier(); /* Avoid RCU read-side critical sections leaking up. */
513
	preempt_enable();
514 515 516
}
EXPORT_SYMBOL_GPL(rcu_all_qs);

517 518 519 520 521 522
#define DEFAULT_RCU_BLIMIT 10     /* Maximum callbacks per rcu_do_batch. */
static long blimit = DEFAULT_RCU_BLIMIT;
#define DEFAULT_RCU_QHIMARK 10000 /* If this many pending, ignore blimit. */
static long qhimark = DEFAULT_RCU_QHIMARK;
#define DEFAULT_RCU_QLOMARK 100   /* Once only this many pending, use blimit. */
static long qlowmark = DEFAULT_RCU_QLOMARK;
523

E
Eric Dumazet 已提交
524 525 526
module_param(blimit, long, 0444);
module_param(qhimark, long, 0444);
module_param(qlowmark, long, 0444);
527

528 529
static ulong jiffies_till_first_fqs = ULONG_MAX;
static ulong jiffies_till_next_fqs = ULONG_MAX;
530
static bool rcu_kick_kthreads;
531 532 533

module_param(jiffies_till_first_fqs, ulong, 0644);
module_param(jiffies_till_next_fqs, ulong, 0644);
534
module_param(rcu_kick_kthreads, bool, 0644);
535

536 537 538 539 540 541 542
/*
 * How long the grace period must be before we start recruiting
 * quiescent-state help from rcu_note_context_switch().
 */
static ulong jiffies_till_sched_qs = HZ / 20;
module_param(jiffies_till_sched_qs, ulong, 0644);

543
static bool rcu_start_gp_advanced(struct rcu_state *rsp, struct rcu_node *rnp,
544
				  struct rcu_data *rdp);
545
static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *rsp));
546
static void force_quiescent_state(struct rcu_state *rsp);
547
static int rcu_pending(void);
548 549

/*
550
 * Return the number of RCU batches started thus far for debug & stats.
551
 */
552 553 554 555 556 557 558 559
unsigned long rcu_batches_started(void)
{
	return rcu_state_p->gpnum;
}
EXPORT_SYMBOL_GPL(rcu_batches_started);

/*
 * Return the number of RCU-sched batches started thus far for debug & stats.
560
 */
561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586
unsigned long rcu_batches_started_sched(void)
{
	return rcu_sched_state.gpnum;
}
EXPORT_SYMBOL_GPL(rcu_batches_started_sched);

/*
 * Return the number of RCU BH batches started thus far for debug & stats.
 */
unsigned long rcu_batches_started_bh(void)
{
	return rcu_bh_state.gpnum;
}
EXPORT_SYMBOL_GPL(rcu_batches_started_bh);

/*
 * Return the number of RCU batches completed thus far for debug & stats.
 */
unsigned long rcu_batches_completed(void)
{
	return rcu_state_p->completed;
}
EXPORT_SYMBOL_GPL(rcu_batches_completed);

/*
 * Return the number of RCU-sched batches completed thus far for debug & stats.
587
 */
588
unsigned long rcu_batches_completed_sched(void)
589
{
590
	return rcu_sched_state.completed;
591
}
592
EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);
593 594

/*
595
 * Return the number of RCU BH batches completed thus far for debug & stats.
596
 */
597
unsigned long rcu_batches_completed_bh(void)
598 599 600 601 602
{
	return rcu_bh_state.completed;
}
EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);

603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624
/*
 * Return the number of RCU expedited batches completed thus far for
 * debug & stats.  Odd numbers mean that a batch is in progress, even
 * numbers mean idle.  The value returned will thus be roughly double
 * the cumulative batches since boot.
 */
unsigned long rcu_exp_batches_completed(void)
{
	return rcu_state_p->expedited_sequence;
}
EXPORT_SYMBOL_GPL(rcu_exp_batches_completed);

/*
 * Return the number of RCU-sched expedited batches completed thus far
 * for debug & stats.  Similar to rcu_exp_batches_completed().
 */
unsigned long rcu_exp_batches_completed_sched(void)
{
	return rcu_sched_state.expedited_sequence;
}
EXPORT_SYMBOL_GPL(rcu_exp_batches_completed_sched);

625 626 627 628 629
/*
 * Force a quiescent state.
 */
void rcu_force_quiescent_state(void)
{
630
	force_quiescent_state(rcu_state_p);
631 632 633
}
EXPORT_SYMBOL_GPL(rcu_force_quiescent_state);

634 635 636 637 638
/*
 * Force a quiescent state for RCU BH.
 */
void rcu_bh_force_quiescent_state(void)
{
639
	force_quiescent_state(&rcu_bh_state);
640 641 642
}
EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

643 644 645 646 647 648 649 650 651
/*
 * Force a quiescent state for RCU-sched.
 */
void rcu_sched_force_quiescent_state(void)
{
	force_quiescent_state(&rcu_sched_state);
}
EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
/*
 * Show the state of the grace-period kthreads.
 */
void show_rcu_gp_kthreads(void)
{
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp) {
		pr_info("%s: wait state: %d ->state: %#lx\n",
			rsp->name, rsp->gp_state, rsp->gp_kthread->state);
		/* sched_show_task(rsp->gp_kthread); */
	}
}
EXPORT_SYMBOL_GPL(show_rcu_gp_kthreads);

667 668 669 670 671 672 673 674 675 676 677 678 679 680
/*
 * 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);

681 682 683 684 685 686 687 688 689 690
/*
 * Send along grace-period-related data for rcutorture diagnostics.
 */
void rcutorture_get_gp_data(enum rcutorture_type test_type, int *flags,
			    unsigned long *gpnum, unsigned long *completed)
{
	struct rcu_state *rsp = NULL;

	switch (test_type) {
	case RCU_FLAVOR:
691
		rsp = rcu_state_p;
692 693 694 695 696 697 698 699 700 701
		break;
	case RCU_BH_FLAVOR:
		rsp = &rcu_bh_state;
		break;
	case RCU_SCHED_FLAVOR:
		rsp = &rcu_sched_state;
		break;
	default:
		break;
	}
702
	if (rsp == NULL)
703
		return;
704 705 706
	*flags = READ_ONCE(rsp->gp_flags);
	*gpnum = READ_ONCE(rsp->gpnum);
	*completed = READ_ONCE(rsp->completed);
707 708 709
}
EXPORT_SYMBOL_GPL(rcutorture_get_gp_data);

710 711 712 713 714 715 716 717 718 719 720
/*
 * 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);

721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736
/*
 * 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];
}

/*
 * Is there any need for future grace periods?
 * Interrupts must be disabled.  If the caller does not hold the root
 * rnp_node structure's ->lock, the results are advisory only.
 */
static int rcu_future_needs_gp(struct rcu_state *rsp)
{
	struct rcu_node *rnp = rcu_get_root(rsp);
737
	int idx = (READ_ONCE(rnp->completed) + 1) & 0x1;
738 739
	int *fp = &rnp->need_future_gp[idx];

740
	RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_future_needs_gp() invoked with irqs enabled!!!");
741
	return READ_ONCE(*fp);
742 743
}

744
/*
745 746 747
 * Does the current CPU require a not-yet-started grace period?
 * The caller must have disabled interrupts to prevent races with
 * normal callback registry.
748
 */
749
static bool
750 751
cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)
{
752
	RCU_LOCKDEP_WARN(!irqs_disabled(), "cpu_needs_another_gp() invoked with irqs enabled!!!");
753
	if (rcu_gp_in_progress(rsp))
754
		return false;  /* No, a grace period is already in progress. */
755
	if (rcu_future_needs_gp(rsp))
756
		return true;  /* Yes, a no-CBs CPU needs one. */
757
	if (!rcu_segcblist_is_enabled(&rdp->cblist))
758
		return false;  /* No, this is a no-CBs (or offline) CPU. */
759
	if (!rcu_segcblist_restempty(&rdp->cblist, RCU_NEXT_READY_TAIL))
760
		return true;  /* Yes, CPU has newly registered callbacks. */
761 762 763
	if (rcu_segcblist_future_gp_needed(&rdp->cblist,
					   READ_ONCE(rsp->completed)))
		return true;  /* Yes, CBs for future grace period. */
764
	return false; /* No grace period needed. */
765 766
}

767
/*
P
Paul E. McKenney 已提交
768
 * rcu_eqs_enter_common - current CPU is entering an extended quiescent state
769
 *
P
Paul E. McKenney 已提交
770 771
 * Enter idle, doing appropriate accounting.  The caller must have
 * disabled interrupts.
772
 */
P
Paul E. McKenney 已提交
773
static void rcu_eqs_enter_common(bool user)
774
{
775 776
	struct rcu_state *rsp;
	struct rcu_data *rdp;
P
Paul E. McKenney 已提交
777
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
778

779
	RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_eqs_enter_common() invoked with irqs enabled!!!");
P
Paul E. McKenney 已提交
780
	trace_rcu_dyntick(TPS("Start"), rdtp->dynticks_nesting, 0);
781 782
	if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
	    !user && !is_idle_task(current)) {
783 784
		struct task_struct *idle __maybe_unused =
			idle_task(smp_processor_id());
785

P
Paul E. McKenney 已提交
786
		trace_rcu_dyntick(TPS("Error on entry: not idle task"), rdtp->dynticks_nesting, 0);
787
		rcu_ftrace_dump(DUMP_ORIG);
788 789 790
		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! */
791
	}
792 793 794 795
	for_each_rcu_flavor(rsp) {
		rdp = this_cpu_ptr(rsp->rda);
		do_nocb_deferred_wakeup(rdp);
	}
796
	rcu_prepare_for_idle();
797
	__this_cpu_inc(disable_rcu_irq_enter);
P
Paul E. McKenney 已提交
798 799
	rdtp->dynticks_nesting = 0; /* Breaks tracing momentarily. */
	rcu_dynticks_eqs_enter(); /* After this, tracing works again. */
800
	__this_cpu_dec(disable_rcu_irq_enter);
801
	rcu_dynticks_task_enter();
802 803

	/*
804
	 * It is illegal to enter an extended quiescent state while
805 806
	 * in an RCU read-side critical section.
	 */
807 808 809 810 811 812
	RCU_LOCKDEP_WARN(lock_is_held(&rcu_lock_map),
			 "Illegal idle entry in RCU read-side critical section.");
	RCU_LOCKDEP_WARN(lock_is_held(&rcu_bh_lock_map),
			 "Illegal idle entry in RCU-bh read-side critical section.");
	RCU_LOCKDEP_WARN(lock_is_held(&rcu_sched_lock_map),
			 "Illegal idle entry in RCU-sched read-side critical section.");
813
}
814

815 816 817
/*
 * Enter an RCU extended quiescent state, which can be either the
 * idle loop or adaptive-tickless usermode execution.
818
 */
819
static void rcu_eqs_enter(bool user)
820 821 822
{
	struct rcu_dynticks *rdtp;

823
	rdtp = this_cpu_ptr(&rcu_dynticks);
824
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
P
Paul E. McKenney 已提交
825 826 827 828
		     (rdtp->dynticks_nesting & DYNTICK_TASK_NEST_MASK) == 0);
	if ((rdtp->dynticks_nesting & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE)
		rcu_eqs_enter_common(user);
	else
829
		rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE;
830
}
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845

/**
 * 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)
{
846 847 848
	unsigned long flags;

	local_irq_save(flags);
849
	rcu_eqs_enter(false);
850
	local_irq_restore(flags);
851
}
852
EXPORT_SYMBOL_GPL(rcu_idle_enter);
853

854
#ifdef CONFIG_NO_HZ_FULL
855 856 857 858 859 860 861 862 863 864
/**
 * 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)
{
865
	rcu_eqs_enter(1);
866
}
867
#endif /* CONFIG_NO_HZ_FULL */
868

869 870 871 872 873
/**
 * 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
874
 * sections can occur.  The caller must have disabled interrupts.
875
 *
876 877 878 879 880 881 882 883
 * 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.
884
 */
885
void rcu_irq_exit(void)
886 887 888
{
	struct rcu_dynticks *rdtp;

889
	RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_irq_exit() invoked with irqs enabled!!!");
890
	rdtp = this_cpu_ptr(&rcu_dynticks);
891
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
P
Paul E. McKenney 已提交
892 893 894 895 896 897 898
		     rdtp->dynticks_nesting < 1);
	if (rdtp->dynticks_nesting <= 1) {
		rcu_eqs_enter_common(true);
	} else {
		trace_rcu_dyntick(TPS("--="), rdtp->dynticks_nesting, rdtp->dynticks_nesting - 1);
		rdtp->dynticks_nesting--;
	}
899 900 901 902 903 904 905 906 907 908 909
}

/*
 * Wrapper for rcu_irq_exit() where interrupts are enabled.
 */
void rcu_irq_exit_irqson(void)
{
	unsigned long flags;

	local_irq_save(flags);
	rcu_irq_exit();
910 911 912 913
	local_irq_restore(flags);
}

/*
914
 * rcu_eqs_exit_common - current CPU moving away from extended quiescent state
915 916 917 918 919
 *
 * 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.
 */
920
static void rcu_eqs_exit_common(long long oldval, int user)
921
{
922
	RCU_TRACE(struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);)
923

924
	rcu_dynticks_task_exit();
925
	rcu_dynticks_eqs_exit();
926
	rcu_cleanup_after_idle();
927
	trace_rcu_dyntick(TPS("End"), oldval, rdtp->dynticks_nesting);
928 929
	if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
	    !user && !is_idle_task(current)) {
930 931
		struct task_struct *idle __maybe_unused =
			idle_task(smp_processor_id());
932

933
		trace_rcu_dyntick(TPS("Error on exit: not idle task"),
934
				  oldval, rdtp->dynticks_nesting);
935
		rcu_ftrace_dump(DUMP_ORIG);
936 937 938
		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! */
939 940 941
	}
}

942 943 944
/*
 * Exit an RCU extended quiescent state, which can be either the
 * idle loop or adaptive-tickless usermode execution.
945
 */
946
static void rcu_eqs_exit(bool user)
947 948 949 950
{
	struct rcu_dynticks *rdtp;
	long long oldval;

951
	RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_eqs_exit() invoked with irqs enabled!!!");
952
	rdtp = this_cpu_ptr(&rcu_dynticks);
953
	oldval = rdtp->dynticks_nesting;
954
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && oldval < 0);
955
	if (oldval & DYNTICK_TASK_NEST_MASK) {
956
		rdtp->dynticks_nesting += DYNTICK_TASK_NEST_VALUE;
957
	} else {
958
		rdtp->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
959
		rcu_eqs_exit_common(oldval, user);
960
	}
961
}
962 963 964 965 966 967 968 969 970 971 972 973 974 975

/**
 * 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)
{
976 977 978
	unsigned long flags;

	local_irq_save(flags);
979
	rcu_eqs_exit(false);
980
	local_irq_restore(flags);
981
}
982
EXPORT_SYMBOL_GPL(rcu_idle_exit);
983

984
#ifdef CONFIG_NO_HZ_FULL
985 986 987 988 989 990 991 992
/**
 * 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)
{
993
	rcu_eqs_exit(1);
994
}
995
#endif /* CONFIG_NO_HZ_FULL */
996

997 998 999 1000 1001
/**
 * 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
1002
 * sections can occur.  The caller must have disabled interrupts.
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
 *
 * 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)
{
	struct rcu_dynticks *rdtp;
	long long oldval;

1021
	RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_irq_enter() invoked with irqs enabled!!!");
1022
	rdtp = this_cpu_ptr(&rcu_dynticks);
1023 1024
	oldval = rdtp->dynticks_nesting;
	rdtp->dynticks_nesting++;
1025 1026
	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
		     rdtp->dynticks_nesting == 0);
1027
	if (oldval)
1028
		trace_rcu_dyntick(TPS("++="), oldval, rdtp->dynticks_nesting);
1029
	else
1030
		rcu_eqs_exit_common(oldval, true);
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
}

/*
 * Wrapper for rcu_irq_enter() where interrupts are enabled.
 */
void rcu_irq_enter_irqson(void)
{
	unsigned long flags;

	local_irq_save(flags);
	rcu_irq_enter();
1042 1043 1044 1045 1046 1047
	local_irq_restore(flags);
}

/**
 * rcu_nmi_enter - inform RCU of entry to NMI context
 *
1048 1049 1050 1051 1052
 * If the CPU was idle from RCU's viewpoint, update rdtp->dynticks and
 * rdtp->dynticks_nmi_nesting to let the RCU grace-period handling know
 * that the CPU is active.  This implementation permits nested NMIs, as
 * long as the nesting level does not overflow an int.  (You will probably
 * run out of stack space first.)
1053 1054 1055
 */
void rcu_nmi_enter(void)
{
1056
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1057
	int incby = 2;
1058

1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
	/* Complain about underflow. */
	WARN_ON_ONCE(rdtp->dynticks_nmi_nesting < 0);

	/*
	 * If idle from RCU viewpoint, atomically increment ->dynticks
	 * to mark non-idle and increment ->dynticks_nmi_nesting by one.
	 * Otherwise, increment ->dynticks_nmi_nesting by two.  This means
	 * if ->dynticks_nmi_nesting is equal to one, we are guaranteed
	 * to be in the outermost NMI handler that interrupted an RCU-idle
	 * period (observation due to Andy Lutomirski).
	 */
1070
	if (rcu_dynticks_curr_cpu_in_eqs()) {
1071
		rcu_dynticks_eqs_exit();
1072 1073 1074 1075
		incby = 1;
	}
	rdtp->dynticks_nmi_nesting += incby;
	barrier();
1076 1077 1078 1079 1080
}

/**
 * rcu_nmi_exit - inform RCU of exit from NMI context
 *
1081 1082 1083 1084
 * If we are returning from the outermost NMI handler that interrupted an
 * RCU-idle period, update rdtp->dynticks and rdtp->dynticks_nmi_nesting
 * to let the RCU grace-period handling know that the CPU is back to
 * being RCU-idle.
1085 1086 1087
 */
void rcu_nmi_exit(void)
{
1088
	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
1089

1090 1091 1092 1093 1094 1095
	/*
	 * Check for ->dynticks_nmi_nesting underflow and bad ->dynticks.
	 * (We are exiting an NMI handler, so RCU better be paying attention
	 * to us!)
	 */
	WARN_ON_ONCE(rdtp->dynticks_nmi_nesting <= 0);
1096
	WARN_ON_ONCE(rcu_dynticks_curr_cpu_in_eqs());
1097 1098 1099 1100 1101 1102 1103

	/*
	 * If the nesting level is not 1, the CPU wasn't RCU-idle, so
	 * leave it in non-RCU-idle state.
	 */
	if (rdtp->dynticks_nmi_nesting != 1) {
		rdtp->dynticks_nmi_nesting -= 2;
1104
		return;
1105 1106 1107 1108
	}

	/* This NMI interrupted an RCU-idle CPU, restore RCU-idleness. */
	rdtp->dynticks_nmi_nesting = 0;
1109
	rcu_dynticks_eqs_enter();
1110 1111
}

1112 1113
/**
 * rcu_is_watching - see if RCU thinks that the current CPU is idle
1114
 *
1115 1116 1117
 * Return true if RCU is watching the running CPU, which means that this
 * CPU can safely enter RCU read-side critical sections.  In other words,
 * if the current CPU is in its idle loop and is neither in an interrupt
1118
 * or NMI handler, return true.
1119
 */
1120
bool notrace rcu_is_watching(void)
1121
{
1122
	bool ret;
1123

1124
	preempt_disable_notrace();
1125
	ret = !rcu_dynticks_curr_cpu_in_eqs();
1126
	preempt_enable_notrace();
1127
	return ret;
1128
}
1129
EXPORT_SYMBOL_GPL(rcu_is_watching);
1130

1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
/*
 * If a holdout task is actually running, request an urgent quiescent
 * state from its CPU.  This is unsynchronized, so migrations can cause
 * the request to go to the wrong CPU.  Which is OK, all that will happen
 * is that the CPU's next context switch will be a bit slower and next
 * time around this task will generate another request.
 */
void rcu_request_urgent_qs_task(struct task_struct *t)
{
	int cpu;

	barrier();
	cpu = task_cpu(t);
	if (!task_curr(t))
		return; /* This task is not running on that CPU. */
	smp_store_release(per_cpu_ptr(&rcu_dynticks.rcu_urgent_qs, cpu), true);
}

1149
#if defined(CONFIG_PROVE_RCU) && defined(CONFIG_HOTPLUG_CPU)
1150 1151 1152 1153 1154 1155 1156

/*
 * 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
1157 1158 1159 1160 1161 1162
 * 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,
1163 1164
 * the fact that a CPU enters the scheduler after completing the teardown
 * of the CPU.
1165
 *
1166 1167
 * This is also why RCU internally marks CPUs online during in the
 * preparation phase and offline after the CPU has been taken down.
1168 1169 1170 1171 1172 1173
 *
 * Disable checking if in an NMI handler because we cannot safely report
 * errors from NMI handlers anyway.
 */
bool rcu_lockdep_current_cpu_online(void)
{
1174 1175
	struct rcu_data *rdp;
	struct rcu_node *rnp;
1176 1177 1178
	bool ret;

	if (in_nmi())
F
Fengguang Wu 已提交
1179
		return true;
1180
	preempt_disable();
1181
	rdp = this_cpu_ptr(&rcu_sched_data);
1182
	rnp = rdp->mynode;
1183
	ret = (rdp->grpmask & rcu_rnp_online_cpus(rnp)) ||
1184 1185 1186 1187 1188 1189
	      !rcu_scheduler_fully_active;
	preempt_enable();
	return ret;
}
EXPORT_SYMBOL_GPL(rcu_lockdep_current_cpu_online);

1190
#endif /* #if defined(CONFIG_PROVE_RCU) && defined(CONFIG_HOTPLUG_CPU) */
1191

1192
/**
1193
 * rcu_is_cpu_rrupt_from_idle - see if idle or immediately interrupted from idle
1194
 *
1195 1196 1197
 * 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.
1198
 */
1199
static int rcu_is_cpu_rrupt_from_idle(void)
1200
{
1201
	return __this_cpu_read(rcu_dynticks.dynticks_nesting) <= 1;
1202 1203 1204 1205 1206
}

/*
 * Snapshot the specified CPU's dynticks counter so that we can later
 * credit them with an implicit quiescent state.  Return 1 if this CPU
1207
 * is in dynticks idle mode, which is an extended quiescent state.
1208
 */
1209
static int dyntick_save_progress_counter(struct rcu_data *rdp)
1210
{
1211
	rdp->dynticks_snap = rcu_dynticks_snap(rdp->dynticks);
1212
	if (rcu_dynticks_in_eqs(rdp->dynticks_snap)) {
1213
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("dti"));
1214
		if (ULONG_CMP_LT(READ_ONCE(rdp->gpnum) + ULONG_MAX / 4,
1215
				 rdp->mynode->gpnum))
1216
			WRITE_ONCE(rdp->gpwrap, true);
1217
		return 1;
1218
	}
1219
	return 0;
1220 1221 1222 1223 1224 1225
}

/*
 * 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()
1226
 * for this same CPU, or by virtue of having been offline.
1227
 */
1228
static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
1229
{
1230
	unsigned long jtsq;
1231
	bool *rnhqp;
1232
	bool *ruqp;
1233 1234
	unsigned long rjtsc;
	struct rcu_node *rnp;
1235 1236 1237 1238 1239 1240 1241 1242 1243

	/*
	 * 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.
	 */
1244
	if (rcu_dynticks_in_eqs_since(rdp->dynticks, rdp->dynticks_snap)) {
1245
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("dti"));
1246 1247 1248 1249
		rdp->dynticks_fqs++;
		return 1;
	}

1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
	/* Compute and saturate jiffies_till_sched_qs. */
	jtsq = jiffies_till_sched_qs;
	rjtsc = rcu_jiffies_till_stall_check();
	if (jtsq > rjtsc / 2) {
		WRITE_ONCE(jiffies_till_sched_qs, rjtsc);
		jtsq = rjtsc / 2;
	} else if (jtsq < 1) {
		WRITE_ONCE(jiffies_till_sched_qs, 1);
		jtsq = 1;
	}

1261
	/*
1262 1263 1264 1265
	 * Has this CPU encountered a cond_resched_rcu_qs() since the
	 * beginning of the grace period?  For this to be the case,
	 * the CPU has to have noticed the current grace period.  This
	 * might not be the case for nohz_full CPUs looping in the kernel.
1266
	 */
1267
	rnp = rdp->mynode;
1268
	ruqp = per_cpu_ptr(&rcu_dynticks.rcu_urgent_qs, rdp->cpu);
1269
	if (time_after(jiffies, rdp->rsp->gp_start + jtsq) &&
1270
	    READ_ONCE(rdp->rcu_qs_ctr_snap) != per_cpu(rcu_dynticks.rcu_qs_ctr, rdp->cpu) &&
1271 1272 1273
	    READ_ONCE(rdp->gpnum) == rnp->gpnum && !rdp->gpwrap) {
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("rqc"));
		return 1;
1274 1275 1276
	} else {
		/* Load rcu_qs_ctr before store to rcu_urgent_qs. */
		smp_store_release(ruqp, true);
1277 1278
	}

1279 1280
	/* Check for the CPU being offline. */
	if (!(rdp->grpmask & rcu_rnp_online_cpus(rnp))) {
1281
		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("ofl"));
1282 1283 1284
		rdp->offline_fqs++;
		return 1;
	}
1285 1286

	/*
1287 1288 1289 1290 1291 1292
	 * A CPU running for an extended time within the kernel can
	 * delay RCU grace periods.  When the CPU is in NO_HZ_FULL mode,
	 * even context-switching back and forth between a pair of
	 * in-kernel CPU-bound tasks cannot advance grace periods.
	 * So if the grace period is old enough, make the CPU pay attention.
	 * Note that the unsynchronized assignments to the per-CPU
1293
	 * rcu_need_heavy_qs variable are safe.  Yes, setting of
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
	 * bits can be lost, but they will be set again on the next
	 * force-quiescent-state pass.  So lost bit sets do not result
	 * in incorrect behavior, merely in a grace period lasting
	 * a few jiffies longer than it might otherwise.  Because
	 * there are at most four threads involved, and because the
	 * updates are only once every few jiffies, the probability of
	 * lossage (and thus of slight grace-period extension) is
	 * quite low.
	 *
	 * Note that if the jiffies_till_sched_qs boot/sysfs parameter
	 * is set too high, we override with half of the RCU CPU stall
	 * warning delay.
1306
	 */
1307 1308 1309 1310 1311
	rnhqp = &per_cpu(rcu_dynticks.rcu_need_heavy_qs, rdp->cpu);
	if (!READ_ONCE(*rnhqp) &&
	    (time_after(jiffies, rdp->rsp->gp_start + jtsq) ||
	     time_after(jiffies, rdp->rsp->jiffies_resched))) {
		WRITE_ONCE(*rnhqp, true);
1312 1313
		/* Store rcu_need_heavy_qs before rcu_urgent_qs. */
		smp_store_release(ruqp, true);
1314
		rdp->rsp->jiffies_resched += 5; /* Re-enable beating. */
1315 1316
	}

1317 1318 1319 1320 1321 1322
	/*
	 * If more than halfway to RCU CPU stall-warning time, do
	 * a resched_cpu() to try to loosen things up a bit.
	 */
	if (jiffies - rdp->rsp->gp_start > rcu_jiffies_till_stall_check() / 2)
		resched_cpu(rdp->cpu);
1323

1324
	return 0;
1325 1326 1327 1328
}

static void record_gp_stall_check_time(struct rcu_state *rsp)
{
1329
	unsigned long j = jiffies;
1330
	unsigned long j1;
1331 1332 1333

	rsp->gp_start = j;
	smp_wmb(); /* Record start time before stall time. */
1334
	j1 = rcu_jiffies_till_stall_check();
1335
	WRITE_ONCE(rsp->jiffies_stall, j + j1);
1336
	rsp->jiffies_resched = j + j1 / 2;
1337
	rsp->n_force_qs_gpstart = READ_ONCE(rsp->n_force_qs);
1338 1339
}

1340 1341 1342 1343 1344 1345 1346 1347 1348 1349
/*
 * Convert a ->gp_state value to a character string.
 */
static const char *gp_state_getname(short gs)
{
	if (gs < 0 || gs >= ARRAY_SIZE(gp_state_names))
		return "???";
	return gp_state_names[gs];
}

1350 1351 1352 1353 1354 1355 1356 1357 1358
/*
 * Complain about starvation of grace-period kthread.
 */
static void rcu_check_gp_kthread_starvation(struct rcu_state *rsp)
{
	unsigned long gpa;
	unsigned long j;

	j = jiffies;
1359
	gpa = READ_ONCE(rsp->gp_activity);
1360
	if (j - gpa > 2 * HZ) {
1361
		pr_err("%s kthread starved for %ld jiffies! g%lu c%lu f%#x %s(%d) ->state=%#lx\n",
1362
		       rsp->name, j - gpa,
1363
		       rsp->gpnum, rsp->completed,
1364 1365
		       rsp->gp_flags,
		       gp_state_getname(rsp->gp_state), rsp->gp_state,
1366
		       rsp->gp_kthread ? rsp->gp_kthread->state : ~0);
1367
		if (rsp->gp_kthread) {
1368
			sched_show_task(rsp->gp_kthread);
1369 1370
			wake_up_process(rsp->gp_kthread);
		}
1371
	}
1372 1373
}

1374
/*
1375 1376 1377 1378
 * Dump stacks of all tasks running on stalled CPUs.  First try using
 * NMIs, but fall back to manual remote stack tracing on architectures
 * that don't support NMI-based stack dumps.  The NMI-triggered stack
 * traces are more accurate because they are printed by the target CPU.
1379 1380 1381 1382 1383 1384 1385 1386
 */
static void rcu_dump_cpu_stacks(struct rcu_state *rsp)
{
	int cpu;
	unsigned long flags;
	struct rcu_node *rnp;

	rcu_for_each_leaf_node(rsp, rnp) {
1387
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
1388 1389 1390
		for_each_leaf_node_possible_cpu(rnp, cpu)
			if (rnp->qsmask & leaf_node_cpu_bit(rnp, cpu))
				if (!trigger_single_cpu_backtrace(cpu))
1391
					dump_cpu_task(cpu);
B
Boqun Feng 已提交
1392
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
1393 1394 1395
	}
}

1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406
/*
 * If too much time has passed in the current grace period, and if
 * so configured, go kick the relevant kthreads.
 */
static void rcu_stall_kick_kthreads(struct rcu_state *rsp)
{
	unsigned long j;

	if (!rcu_kick_kthreads)
		return;
	j = READ_ONCE(rsp->jiffies_kick_kthreads);
1407 1408
	if (time_after(jiffies, j) && rsp->gp_kthread &&
	    (rcu_gp_in_progress(rsp) || READ_ONCE(rsp->gp_flags))) {
1409
		WARN_ONCE(1, "Kicking %s grace-period kthread\n", rsp->name);
1410
		rcu_ftrace_dump(DUMP_ALL);
1411 1412 1413 1414 1415
		wake_up_process(rsp->gp_kthread);
		WRITE_ONCE(rsp->jiffies_kick_kthreads, j + HZ);
	}
}

1416 1417 1418 1419 1420 1421
static inline void panic_on_rcu_stall(void)
{
	if (sysctl_panic_on_rcu_stall)
		panic("RCU Stall\n");
}

1422
static void print_other_cpu_stall(struct rcu_state *rsp, unsigned long gpnum)
1423 1424 1425 1426
{
	int cpu;
	long delta;
	unsigned long flags;
1427 1428
	unsigned long gpa;
	unsigned long j;
1429
	int ndetected = 0;
1430
	struct rcu_node *rnp = rcu_get_root(rsp);
1431
	long totqlen = 0;
1432

1433 1434 1435 1436 1437
	/* Kick and suppress, if so configured. */
	rcu_stall_kick_kthreads(rsp);
	if (rcu_cpu_stall_suppress)
		return;

1438 1439
	/* Only let one CPU complain about others per time interval. */

1440
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
1441
	delta = jiffies - READ_ONCE(rsp->jiffies_stall);
1442
	if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {
B
Boqun Feng 已提交
1443
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
1444 1445
		return;
	}
1446 1447
	WRITE_ONCE(rsp->jiffies_stall,
		   jiffies + 3 * rcu_jiffies_till_stall_check() + 3);
B
Boqun Feng 已提交
1448
	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
1449

1450 1451 1452 1453 1454
	/*
	 * OK, time to rat on our buddy...
	 * See Documentation/RCU/stallwarn.txt for info on how to debug
	 * RCU CPU stall warnings.
	 */
1455
	pr_err("INFO: %s detected stalls on CPUs/tasks:",
1456
	       rsp->name);
1457
	print_cpu_stall_info_begin();
1458
	rcu_for_each_leaf_node(rsp, rnp) {
1459
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
1460
		ndetected += rcu_print_task_stall(rnp);
1461
		if (rnp->qsmask != 0) {
1462 1463 1464
			for_each_leaf_node_possible_cpu(rnp, cpu)
				if (rnp->qsmask & leaf_node_cpu_bit(rnp, cpu)) {
					print_cpu_stall_info(rsp, cpu);
1465 1466 1467
					ndetected++;
				}
		}
B
Boqun Feng 已提交
1468
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
1469
	}
1470 1471

	print_cpu_stall_info_end();
1472
	for_each_possible_cpu(cpu)
1473 1474
		totqlen += rcu_segcblist_n_cbs(&per_cpu_ptr(rsp->rda,
							    cpu)->cblist);
1475
	pr_cont("(detected by %d, t=%ld jiffies, g=%ld, c=%ld, q=%lu)\n",
1476
	       smp_processor_id(), (long)(jiffies - rsp->gp_start),
1477
	       (long)rsp->gpnum, (long)rsp->completed, totqlen);
1478
	if (ndetected) {
1479
		rcu_dump_cpu_stacks(rsp);
1480 1481 1482

		/* Complain about tasks blocking the grace period. */
		rcu_print_detail_task_stall(rsp);
1483
	} else {
1484 1485
		if (READ_ONCE(rsp->gpnum) != gpnum ||
		    READ_ONCE(rsp->completed) == gpnum) {
1486 1487 1488
			pr_err("INFO: Stall ended before state dump start\n");
		} else {
			j = jiffies;
1489
			gpa = READ_ONCE(rsp->gp_activity);
1490
			pr_err("All QSes seen, last %s kthread activity %ld (%ld-%ld), jiffies_till_next_fqs=%ld, root ->qsmask %#lx\n",
1491
			       rsp->name, j - gpa, j, gpa,
1492 1493
			       jiffies_till_next_fqs,
			       rcu_get_root(rsp)->qsmask);
1494 1495 1496 1497
			/* In this case, the current CPU might be at fault. */
			sched_show_task(current);
		}
	}
1498

1499 1500
	rcu_check_gp_kthread_starvation(rsp);

1501 1502
	panic_on_rcu_stall();

1503
	force_quiescent_state(rsp);  /* Kick them all. */
1504 1505 1506 1507
}

static void print_cpu_stall(struct rcu_state *rsp)
{
1508
	int cpu;
1509 1510
	unsigned long flags;
	struct rcu_node *rnp = rcu_get_root(rsp);
1511
	long totqlen = 0;
1512

1513 1514 1515 1516 1517
	/* Kick and suppress, if so configured. */
	rcu_stall_kick_kthreads(rsp);
	if (rcu_cpu_stall_suppress)
		return;

1518 1519 1520 1521 1522
	/*
	 * OK, time to rat on ourselves...
	 * See Documentation/RCU/stallwarn.txt for info on how to debug
	 * RCU CPU stall warnings.
	 */
1523
	pr_err("INFO: %s self-detected stall on CPU", rsp->name);
1524 1525 1526
	print_cpu_stall_info_begin();
	print_cpu_stall_info(rsp, smp_processor_id());
	print_cpu_stall_info_end();
1527
	for_each_possible_cpu(cpu)
1528 1529
		totqlen += rcu_segcblist_n_cbs(&per_cpu_ptr(rsp->rda,
							    cpu)->cblist);
1530 1531 1532
	pr_cont(" (t=%lu jiffies g=%ld c=%ld q=%lu)\n",
		jiffies - rsp->gp_start,
		(long)rsp->gpnum, (long)rsp->completed, totqlen);
1533 1534 1535

	rcu_check_gp_kthread_starvation(rsp);

1536
	rcu_dump_cpu_stacks(rsp);
1537

1538
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
1539 1540 1541
	if (ULONG_CMP_GE(jiffies, READ_ONCE(rsp->jiffies_stall)))
		WRITE_ONCE(rsp->jiffies_stall,
			   jiffies + 3 * rcu_jiffies_till_stall_check() + 3);
B
Boqun Feng 已提交
1542
	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
1543

1544 1545
	panic_on_rcu_stall();

1546 1547 1548 1549 1550 1551 1552 1553
	/*
	 * Attempt to revive the RCU machinery by forcing a context switch.
	 *
	 * A context switch would normally allow the RCU state machine to make
	 * progress and it could be we're stuck in kernel space without context
	 * switches for an entirely unreasonable amount of time.
	 */
	resched_cpu(smp_processor_id());
1554 1555 1556 1557
}

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
{
1558 1559 1560
	unsigned long completed;
	unsigned long gpnum;
	unsigned long gps;
1561 1562
	unsigned long j;
	unsigned long js;
1563 1564
	struct rcu_node *rnp;

1565 1566
	if ((rcu_cpu_stall_suppress && !rcu_kick_kthreads) ||
	    !rcu_gp_in_progress(rsp))
1567
		return;
1568
	rcu_stall_kick_kthreads(rsp);
1569
	j = jiffies;
1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587

	/*
	 * Lots of memory barriers to reject false positives.
	 *
	 * The idea is to pick up rsp->gpnum, then rsp->jiffies_stall,
	 * then rsp->gp_start, and finally rsp->completed.  These values
	 * are updated in the opposite order with memory barriers (or
	 * equivalent) during grace-period initialization and cleanup.
	 * Now, a false positive can occur if we get an new value of
	 * rsp->gp_start and a old value of rsp->jiffies_stall.  But given
	 * the memory barriers, the only way that this can happen is if one
	 * grace period ends and another starts between these two fetches.
	 * Detect this by comparing rsp->completed with the previous fetch
	 * from rsp->gpnum.
	 *
	 * Given this check, comparisons of jiffies, rsp->jiffies_stall,
	 * and rsp->gp_start suffice to forestall false positives.
	 */
1588
	gpnum = READ_ONCE(rsp->gpnum);
1589
	smp_rmb(); /* Pick up ->gpnum first... */
1590
	js = READ_ONCE(rsp->jiffies_stall);
1591
	smp_rmb(); /* ...then ->jiffies_stall before the rest... */
1592
	gps = READ_ONCE(rsp->gp_start);
1593
	smp_rmb(); /* ...and finally ->gp_start before ->completed. */
1594
	completed = READ_ONCE(rsp->completed);
1595 1596 1597 1598
	if (ULONG_CMP_GE(completed, gpnum) ||
	    ULONG_CMP_LT(j, js) ||
	    ULONG_CMP_GE(gps, js))
		return; /* No stall or GP completed since entering function. */
1599
	rnp = rdp->mynode;
1600
	if (rcu_gp_in_progress(rsp) &&
1601
	    (READ_ONCE(rnp->qsmask) & rdp->grpmask)) {
1602 1603 1604 1605

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

1606 1607
	} else if (rcu_gp_in_progress(rsp) &&
		   ULONG_CMP_GE(j, js + RCU_STALL_RAT_DELAY)) {
1608

1609
		/* They had a few time units to dump stack, so complain. */
1610
		print_other_cpu_stall(rsp, gpnum);
1611 1612 1613
	}
}

1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
/**
 * 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)
{
1625 1626 1627
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
1628
		WRITE_ONCE(rsp->jiffies_stall, jiffies + ULONG_MAX / 2);
1629 1630
}

1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642
/*
 * Determine the value that ->completed will have at the end of the
 * next subsequent grace period.  This is used to tag callbacks so that
 * a CPU can invoke callbacks in a timely fashion even if that CPU has
 * been dyntick-idle for an extended period with callbacks under the
 * influence of RCU_FAST_NO_HZ.
 *
 * The caller must hold rnp->lock with interrupts disabled.
 */
static unsigned long rcu_cbs_completed(struct rcu_state *rsp,
				       struct rcu_node *rnp)
{
1643 1644
	lockdep_assert_held(&rnp->lock);

1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661
	/*
	 * If RCU is idle, we just wait for the next grace period.
	 * But we can only be sure that RCU is idle if we are looking
	 * at the root rcu_node structure -- otherwise, a new grace
	 * period might have started, but just not yet gotten around
	 * to initializing the current non-root rcu_node structure.
	 */
	if (rcu_get_root(rsp) == rnp && rnp->gpnum == rnp->completed)
		return rnp->completed + 1;

	/*
	 * Otherwise, wait for a possible partial grace period and
	 * then the subsequent full grace period.
	 */
	return rnp->completed + 2;
}

1662 1663 1664 1665 1666
/*
 * Trace-event helper function for rcu_start_future_gp() and
 * rcu_nocb_wait_gp().
 */
static void trace_rcu_future_gp(struct rcu_node *rnp, struct rcu_data *rdp,
1667
				unsigned long c, const char *s)
1668 1669 1670 1671 1672 1673 1674 1675 1676
{
	trace_rcu_future_grace_period(rdp->rsp->name, rnp->gpnum,
				      rnp->completed, c, rnp->level,
				      rnp->grplo, rnp->grphi, s);
}

/*
 * Start some future grace period, as needed to handle newly arrived
 * callbacks.  The required future grace periods are recorded in each
1677 1678
 * rcu_node structure's ->need_future_gp field.  Returns true if there
 * is reason to awaken the grace-period kthread.
1679 1680 1681
 *
 * The caller must hold the specified rcu_node structure's ->lock.
 */
1682 1683 1684
static bool __maybe_unused
rcu_start_future_gp(struct rcu_node *rnp, struct rcu_data *rdp,
		    unsigned long *c_out)
1685 1686
{
	unsigned long c;
1687
	bool ret = false;
1688 1689
	struct rcu_node *rnp_root = rcu_get_root(rdp->rsp);

1690 1691
	lockdep_assert_held(&rnp->lock);

1692 1693 1694 1695 1696
	/*
	 * Pick up grace-period number for new callbacks.  If this
	 * grace period is already marked as needed, return to the caller.
	 */
	c = rcu_cbs_completed(rdp->rsp, rnp);
1697
	trace_rcu_future_gp(rnp, rdp, c, TPS("Startleaf"));
1698
	if (rnp->need_future_gp[c & 0x1]) {
1699
		trace_rcu_future_gp(rnp, rdp, c, TPS("Prestartleaf"));
1700
		goto out;
1701 1702 1703 1704 1705 1706 1707
	}

	/*
	 * If either this rcu_node structure or the root rcu_node structure
	 * believe that a grace period is in progress, then we must wait
	 * for the one following, which is in "c".  Because our request
	 * will be noticed at the end of the current grace period, we don't
1708 1709 1710 1711 1712 1713 1714
	 * need to explicitly start one.  We only do the lockless check
	 * of rnp_root's fields if the current rcu_node structure thinks
	 * there is no grace period in flight, and because we hold rnp->lock,
	 * the only possible change is when rnp_root's two fields are
	 * equal, in which case rnp_root->gpnum might be concurrently
	 * incremented.  But that is OK, as it will just result in our
	 * doing some extra useless work.
1715 1716
	 */
	if (rnp->gpnum != rnp->completed ||
1717
	    READ_ONCE(rnp_root->gpnum) != READ_ONCE(rnp_root->completed)) {
1718
		rnp->need_future_gp[c & 0x1]++;
1719
		trace_rcu_future_gp(rnp, rdp, c, TPS("Startedleaf"));
1720
		goto out;
1721 1722 1723 1724 1725 1726 1727
	}

	/*
	 * There might be no grace period in progress.  If we don't already
	 * hold it, acquire the root rcu_node structure's lock in order to
	 * start one (if needed).
	 */
1728 1729
	if (rnp != rnp_root)
		raw_spin_lock_rcu_node(rnp_root);
1730 1731 1732 1733

	/*
	 * Get a new grace-period number.  If there really is no grace
	 * period in progress, it will be smaller than the one we obtained
1734
	 * earlier.  Adjust callbacks as needed.
1735 1736
	 */
	c = rcu_cbs_completed(rdp->rsp, rnp_root);
1737 1738
	if (!rcu_is_nocb_cpu(rdp->cpu))
		(void)rcu_segcblist_accelerate(&rdp->cblist, c);
1739 1740 1741 1742 1743 1744

	/*
	 * If the needed for the required grace period is already
	 * recorded, trace and leave.
	 */
	if (rnp_root->need_future_gp[c & 0x1]) {
1745
		trace_rcu_future_gp(rnp, rdp, c, TPS("Prestartedroot"));
1746 1747 1748 1749 1750 1751 1752 1753
		goto unlock_out;
	}

	/* Record the need for the future grace period. */
	rnp_root->need_future_gp[c & 0x1]++;

	/* If a grace period is not already in progress, start one. */
	if (rnp_root->gpnum != rnp_root->completed) {
1754
		trace_rcu_future_gp(rnp, rdp, c, TPS("Startedleafroot"));
1755
	} else {
1756
		trace_rcu_future_gp(rnp, rdp, c, TPS("Startedroot"));
1757
		ret = rcu_start_gp_advanced(rdp->rsp, rnp_root, rdp);
1758 1759 1760
	}
unlock_out:
	if (rnp != rnp_root)
B
Boqun Feng 已提交
1761
		raw_spin_unlock_rcu_node(rnp_root);
1762 1763 1764 1765
out:
	if (c_out != NULL)
		*c_out = c;
	return ret;
1766 1767 1768 1769
}

/*
 * Clean up any old requests for the just-ended grace period.  Also return
1770
 * whether any additional grace periods have been requested.
1771 1772 1773 1774 1775 1776 1777 1778 1779
 */
static int rcu_future_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp)
{
	int c = rnp->completed;
	int needmore;
	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

	rnp->need_future_gp[c & 0x1] = 0;
	needmore = rnp->need_future_gp[(c + 1) & 0x1];
1780 1781
	trace_rcu_future_gp(rnp, rdp, c,
			    needmore ? TPS("CleanupMore") : TPS("Cleanup"));
1782 1783 1784
	return needmore;
}

1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
/*
 * Awaken the grace-period kthread for the specified flavor of RCU.
 * Don't do a self-awaken, and don't bother awakening when there is
 * nothing for the grace-period kthread to do (as in several CPUs
 * raced to awaken, and we lost), and finally don't try to awaken
 * a kthread that has not yet been created.
 */
static void rcu_gp_kthread_wake(struct rcu_state *rsp)
{
	if (current == rsp->gp_kthread ||
1795
	    !READ_ONCE(rsp->gp_flags) ||
1796 1797
	    !rsp->gp_kthread)
		return;
1798
	swake_up(&rsp->gp_wq);
1799 1800
}

1801 1802 1803 1804 1805 1806 1807
/*
 * If there is room, assign a ->completed number to any callbacks on
 * this CPU that have not already been assigned.  Also accelerate any
 * callbacks that were previously assigned a ->completed number that has
 * since proven to be too conservative, which can happen if callbacks get
 * assigned a ->completed number while RCU is idle, but with reference to
 * a non-root rcu_node structure.  This function is idempotent, so it does
1808 1809
 * not hurt to call it repeatedly.  Returns an flag saying that we should
 * awaken the RCU grace-period kthread.
1810 1811 1812
 *
 * The caller must hold rnp->lock with interrupts disabled.
 */
1813
static bool rcu_accelerate_cbs(struct rcu_state *rsp, struct rcu_node *rnp,
1814 1815
			       struct rcu_data *rdp)
{
1816
	bool ret = false;
1817

1818 1819
	lockdep_assert_held(&rnp->lock);

1820 1821
	/* If no pending (not yet ready to invoke) callbacks, nothing to do. */
	if (!rcu_segcblist_pend_cbs(&rdp->cblist))
1822
		return false;
1823 1824

	/*
1825 1826 1827 1828 1829 1830 1831 1832
	 * Callbacks are often registered with incomplete grace-period
	 * information.  Something about the fact that getting exact
	 * information requires acquiring a global lock...  RCU therefore
	 * makes a conservative estimate of the grace period number at which
	 * a given callback will become ready to invoke.	The following
	 * code checks this estimate and improves it when possible, thus
	 * accelerating callback invocation to an earlier grace-period
	 * number.
1833
	 */
1834 1835
	if (rcu_segcblist_accelerate(&rdp->cblist, rcu_cbs_completed(rsp, rnp)))
		ret = rcu_start_future_gp(rnp, rdp, NULL);
1836 1837

	/* Trace depending on how much we were able to accelerate. */
1838
	if (rcu_segcblist_restempty(&rdp->cblist, RCU_WAIT_TAIL))
1839
		trace_rcu_grace_period(rsp->name, rdp->gpnum, TPS("AccWaitCB"));
1840
	else
1841
		trace_rcu_grace_period(rsp->name, rdp->gpnum, TPS("AccReadyCB"));
1842
	return ret;
1843 1844 1845 1846 1847 1848 1849 1850
}

/*
 * Move any callbacks whose grace period has completed to the
 * RCU_DONE_TAIL sublist, then compact the remaining sublists and
 * assign ->completed numbers to any callbacks in the RCU_NEXT_TAIL
 * sublist.  This function is idempotent, so it does not hurt to
 * invoke it repeatedly.  As long as it is not invoked -too- often...
1851
 * Returns true if the RCU grace-period kthread needs to be awakened.
1852 1853 1854
 *
 * The caller must hold rnp->lock with interrupts disabled.
 */
1855
static bool rcu_advance_cbs(struct rcu_state *rsp, struct rcu_node *rnp,
1856 1857
			    struct rcu_data *rdp)
{
1858 1859
	lockdep_assert_held(&rnp->lock);

1860 1861
	/* If no pending (not yet ready to invoke) callbacks, nothing to do. */
	if (!rcu_segcblist_pend_cbs(&rdp->cblist))
1862
		return false;
1863 1864 1865 1866 1867

	/*
	 * Find all callbacks whose ->completed numbers indicate that they
	 * are ready to invoke, and put them into the RCU_DONE_TAIL sublist.
	 */
1868
	rcu_segcblist_advance(&rdp->cblist, rnp->completed);
1869 1870

	/* Classify any remaining callbacks. */
1871
	return rcu_accelerate_cbs(rsp, rnp, rdp);
1872 1873
}

1874
/*
1875 1876 1877
 * Update CPU-local rcu_data state to record the beginnings and ends of
 * grace periods.  The caller must hold the ->lock of the leaf rcu_node
 * structure corresponding to the current CPU, and must have irqs disabled.
1878
 * Returns true if the grace-period kthread needs to be awakened.
1879
 */
1880 1881
static bool __note_gp_changes(struct rcu_state *rsp, struct rcu_node *rnp,
			      struct rcu_data *rdp)
1882
{
1883
	bool ret;
1884
	bool need_gp;
1885

1886 1887
	lockdep_assert_held(&rnp->lock);

1888
	/* Handle the ends of any preceding grace periods first. */
1889
	if (rdp->completed == rnp->completed &&
1890
	    !unlikely(READ_ONCE(rdp->gpwrap))) {
1891

1892
		/* No grace period end, so just accelerate recent callbacks. */
1893
		ret = rcu_accelerate_cbs(rsp, rnp, rdp);
1894

1895 1896 1897
	} else {

		/* Advance callbacks. */
1898
		ret = rcu_advance_cbs(rsp, rnp, rdp);
1899 1900 1901

		/* Remember that we saw this grace-period completion. */
		rdp->completed = rnp->completed;
1902
		trace_rcu_grace_period(rsp->name, rdp->gpnum, TPS("cpuend"));
1903
	}
1904

1905
	if (rdp->gpnum != rnp->gpnum || unlikely(READ_ONCE(rdp->gpwrap))) {
1906 1907 1908 1909 1910 1911
		/*
		 * If the current grace period is waiting for this CPU,
		 * set up to detect a quiescent state, otherwise don't
		 * go looking for one.
		 */
		rdp->gpnum = rnp->gpnum;
1912
		trace_rcu_grace_period(rsp->name, rdp->gpnum, TPS("cpustart"));
1913 1914
		need_gp = !!(rnp->qsmask & rdp->grpmask);
		rdp->cpu_no_qs.b.norm = need_gp;
1915
		rdp->rcu_qs_ctr_snap = __this_cpu_read(rcu_dynticks.rcu_qs_ctr);
1916
		rdp->core_needs_qs = need_gp;
1917
		zero_cpu_stall_ticks(rdp);
1918
		WRITE_ONCE(rdp->gpwrap, false);
1919
	}
1920
	return ret;
1921 1922
}

1923
static void note_gp_changes(struct rcu_state *rsp, struct rcu_data *rdp)
1924 1925
{
	unsigned long flags;
1926
	bool needwake;
1927 1928 1929 1930
	struct rcu_node *rnp;

	local_irq_save(flags);
	rnp = rdp->mynode;
1931 1932 1933
	if ((rdp->gpnum == READ_ONCE(rnp->gpnum) &&
	     rdp->completed == READ_ONCE(rnp->completed) &&
	     !unlikely(READ_ONCE(rdp->gpwrap))) || /* w/out lock. */
1934
	    !raw_spin_trylock_rcu_node(rnp)) { /* irqs already off, so later. */
1935 1936 1937
		local_irq_restore(flags);
		return;
	}
1938
	needwake = __note_gp_changes(rsp, rnp, rdp);
B
Boqun Feng 已提交
1939
	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
1940 1941
	if (needwake)
		rcu_gp_kthread_wake(rsp);
1942 1943
}

1944 1945 1946 1947 1948 1949 1950
static void rcu_gp_slow(struct rcu_state *rsp, int delay)
{
	if (delay > 0 &&
	    !(rsp->gpnum % (rcu_num_nodes * PER_RCU_NODE_PERIOD * delay)))
		schedule_timeout_uninterruptible(delay);
}

1951
/*
1952
 * Initialize a new grace period.  Return false if no grace period required.
1953
 */
1954
static bool rcu_gp_init(struct rcu_state *rsp)
1955
{
1956
	unsigned long oldmask;
1957
	struct rcu_data *rdp;
1958
	struct rcu_node *rnp = rcu_get_root(rsp);
1959

1960
	WRITE_ONCE(rsp->gp_activity, jiffies);
1961
	raw_spin_lock_irq_rcu_node(rnp);
1962
	if (!READ_ONCE(rsp->gp_flags)) {
1963
		/* Spurious wakeup, tell caller to go back to sleep.  */
B
Boqun Feng 已提交
1964
		raw_spin_unlock_irq_rcu_node(rnp);
1965
		return false;
1966
	}
1967
	WRITE_ONCE(rsp->gp_flags, 0); /* Clear all flags: New grace period. */
1968

1969 1970 1971 1972 1973
	if (WARN_ON_ONCE(rcu_gp_in_progress(rsp))) {
		/*
		 * Grace period already in progress, don't start another.
		 * Not supposed to be able to happen.
		 */
B
Boqun Feng 已提交
1974
		raw_spin_unlock_irq_rcu_node(rnp);
1975
		return false;
1976 1977 1978
	}

	/* Advance to a new grace period and initialize state. */
1979
	record_gp_stall_check_time(rsp);
1980 1981
	/* Record GP times before starting GP, hence smp_store_release(). */
	smp_store_release(&rsp->gpnum, rsp->gpnum + 1);
1982
	trace_rcu_grace_period(rsp->name, rsp->gpnum, TPS("start"));
B
Boqun Feng 已提交
1983
	raw_spin_unlock_irq_rcu_node(rnp);
1984

1985 1986 1987 1988 1989 1990 1991
	/*
	 * Apply per-leaf buffered online and offline operations to the
	 * rcu_node tree.  Note that this new grace period need not wait
	 * for subsequent online CPUs, and that quiescent-state forcing
	 * will handle subsequent offline CPUs.
	 */
	rcu_for_each_leaf_node(rsp, rnp) {
1992
		rcu_gp_slow(rsp, gp_preinit_delay);
1993
		raw_spin_lock_irq_rcu_node(rnp);
1994 1995 1996
		if (rnp->qsmaskinit == rnp->qsmaskinitnext &&
		    !rnp->wait_blkd_tasks) {
			/* Nothing to do on this leaf rcu_node structure. */
B
Boqun Feng 已提交
1997
			raw_spin_unlock_irq_rcu_node(rnp);
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
			continue;
		}

		/* Record old state, apply changes to ->qsmaskinit field. */
		oldmask = rnp->qsmaskinit;
		rnp->qsmaskinit = rnp->qsmaskinitnext;

		/* If zero-ness of ->qsmaskinit changed, propagate up tree. */
		if (!oldmask != !rnp->qsmaskinit) {
			if (!oldmask) /* First online CPU for this rcu_node. */
				rcu_init_new_rnp(rnp);
			else if (rcu_preempt_has_tasks(rnp)) /* blocked tasks */
				rnp->wait_blkd_tasks = true;
			else /* Last offline CPU and can propagate. */
				rcu_cleanup_dead_rnp(rnp);
		}

		/*
		 * If all waited-on tasks from prior grace period are
		 * done, and if all this rcu_node structure's CPUs are
		 * still offline, propagate up the rcu_node tree and
		 * clear ->wait_blkd_tasks.  Otherwise, if one of this
		 * rcu_node structure's CPUs has since come back online,
		 * simply clear ->wait_blkd_tasks (but rcu_cleanup_dead_rnp()
		 * checks for this, so just call it unconditionally).
		 */
		if (rnp->wait_blkd_tasks &&
		    (!rcu_preempt_has_tasks(rnp) ||
		     rnp->qsmaskinit)) {
			rnp->wait_blkd_tasks = false;
			rcu_cleanup_dead_rnp(rnp);
		}

B
Boqun Feng 已提交
2031
		raw_spin_unlock_irq_rcu_node(rnp);
2032
	}
2033 2034 2035 2036 2037 2038 2039 2040

	/*
	 * 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
2041
	 * leaf node has been initialized.
2042 2043 2044 2045 2046
	 *
	 * The grace period cannot complete until the initialization
	 * process finishes, because this kthread handles both.
	 */
	rcu_for_each_node_breadth_first(rsp, rnp) {
2047
		rcu_gp_slow(rsp, gp_init_delay);
2048
		raw_spin_lock_irq_rcu_node(rnp);
2049
		rdp = this_cpu_ptr(rsp->rda);
2050 2051
		rcu_preempt_check_blocked_tasks(rnp);
		rnp->qsmask = rnp->qsmaskinit;
2052
		WRITE_ONCE(rnp->gpnum, rsp->gpnum);
2053
		if (WARN_ON_ONCE(rnp->completed != rsp->completed))
2054
			WRITE_ONCE(rnp->completed, rsp->completed);
2055
		if (rnp == rdp->mynode)
2056
			(void)__note_gp_changes(rsp, rnp, rdp);
2057 2058 2059 2060
		rcu_preempt_boost_start_gp(rnp);
		trace_rcu_grace_period_init(rsp->name, rnp->gpnum,
					    rnp->level, rnp->grplo,
					    rnp->grphi, rnp->qsmask);
B
Boqun Feng 已提交
2061
		raw_spin_unlock_irq_rcu_node(rnp);
2062
		cond_resched_rcu_qs();
2063
		WRITE_ONCE(rsp->gp_activity, jiffies);
2064
	}
2065

2066
	return true;
2067
}
2068

2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088
/*
 * Helper function for wait_event_interruptible_timeout() wakeup
 * at force-quiescent-state time.
 */
static bool rcu_gp_fqs_check_wake(struct rcu_state *rsp, int *gfp)
{
	struct rcu_node *rnp = rcu_get_root(rsp);

	/* Someone like call_rcu() requested a force-quiescent-state scan. */
	*gfp = READ_ONCE(rsp->gp_flags);
	if (*gfp & RCU_GP_FLAG_FQS)
		return true;

	/* The current grace period has completed. */
	if (!READ_ONCE(rnp->qsmask) && !rcu_preempt_blocked_readers_cgp(rnp))
		return true;

	return false;
}

2089 2090 2091
/*
 * Do one round of quiescent-state forcing.
 */
2092
static void rcu_gp_fqs(struct rcu_state *rsp, bool first_time)
2093 2094 2095
{
	struct rcu_node *rnp = rcu_get_root(rsp);

2096
	WRITE_ONCE(rsp->gp_activity, jiffies);
2097
	rsp->n_force_qs++;
2098
	if (first_time) {
2099
		/* Collect dyntick-idle snapshots. */
2100
		force_qs_rnp(rsp, dyntick_save_progress_counter);
2101 2102
	} else {
		/* Handle dyntick-idle and offline CPUs. */
2103
		force_qs_rnp(rsp, rcu_implicit_dynticks_qs);
2104 2105
	}
	/* Clear flag to prevent immediate re-entry. */
2106
	if (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) {
2107
		raw_spin_lock_irq_rcu_node(rnp);
2108 2109
		WRITE_ONCE(rsp->gp_flags,
			   READ_ONCE(rsp->gp_flags) & ~RCU_GP_FLAG_FQS);
B
Boqun Feng 已提交
2110
		raw_spin_unlock_irq_rcu_node(rnp);
2111 2112 2113
	}
}

2114 2115 2116
/*
 * Clean up after the old grace period.
 */
2117
static void rcu_gp_cleanup(struct rcu_state *rsp)
2118 2119
{
	unsigned long gp_duration;
2120
	bool needgp = false;
2121
	int nocb = 0;
2122 2123
	struct rcu_data *rdp;
	struct rcu_node *rnp = rcu_get_root(rsp);
2124
	struct swait_queue_head *sq;
2125

2126
	WRITE_ONCE(rsp->gp_activity, jiffies);
2127
	raw_spin_lock_irq_rcu_node(rnp);
2128 2129 2130
	gp_duration = jiffies - rsp->gp_start;
	if (gp_duration > rsp->gp_max)
		rsp->gp_max = gp_duration;
2131

2132 2133 2134 2135 2136 2137 2138 2139
	/*
	 * 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.
	 */
B
Boqun Feng 已提交
2140
	raw_spin_unlock_irq_rcu_node(rnp);
2141

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151
	/*
	 * 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) {
2152
		raw_spin_lock_irq_rcu_node(rnp);
2153 2154
		WARN_ON_ONCE(rcu_preempt_blocked_readers_cgp(rnp));
		WARN_ON_ONCE(rnp->qsmask);
2155
		WRITE_ONCE(rnp->completed, rsp->gpnum);
2156 2157
		rdp = this_cpu_ptr(rsp->rda);
		if (rnp == rdp->mynode)
2158
			needgp = __note_gp_changes(rsp, rnp, rdp) || needgp;
2159
		/* smp_mb() provided by prior unlock-lock pair. */
2160
		nocb += rcu_future_gp_cleanup(rsp, rnp);
2161
		sq = rcu_nocb_gp_get(rnp);
B
Boqun Feng 已提交
2162
		raw_spin_unlock_irq_rcu_node(rnp);
2163
		rcu_nocb_gp_cleanup(sq);
2164
		cond_resched_rcu_qs();
2165
		WRITE_ONCE(rsp->gp_activity, jiffies);
2166
		rcu_gp_slow(rsp, gp_cleanup_delay);
2167
	}
2168
	rnp = rcu_get_root(rsp);
2169
	raw_spin_lock_irq_rcu_node(rnp); /* Order GP before ->completed update. */
2170
	rcu_nocb_gp_set(rnp, nocb);
2171

2172
	/* Declare grace period done. */
2173
	WRITE_ONCE(rsp->completed, rsp->gpnum);
2174
	trace_rcu_grace_period(rsp->name, rsp->completed, TPS("end"));
2175
	rsp->gp_state = RCU_GP_IDLE;
2176
	rdp = this_cpu_ptr(rsp->rda);
2177 2178 2179
	/* Advance CBs to reduce false positives below. */
	needgp = rcu_advance_cbs(rsp, rnp, rdp) || needgp;
	if (needgp || cpu_needs_another_gp(rsp, rdp)) {
2180
		WRITE_ONCE(rsp->gp_flags, RCU_GP_FLAG_INIT);
2181
		trace_rcu_grace_period(rsp->name,
2182
				       READ_ONCE(rsp->gpnum),
2183 2184
				       TPS("newreq"));
	}
B
Boqun Feng 已提交
2185
	raw_spin_unlock_irq_rcu_node(rnp);
2186 2187 2188 2189 2190 2191 2192
}

/*
 * Body of kthread that handles grace periods.
 */
static int __noreturn rcu_gp_kthread(void *arg)
{
2193
	bool first_gp_fqs;
2194
	int gf;
2195
	unsigned long j;
2196
	int ret;
2197 2198 2199
	struct rcu_state *rsp = arg;
	struct rcu_node *rnp = rcu_get_root(rsp);

2200
	rcu_bind_gp_kthread();
2201 2202 2203 2204
	for (;;) {

		/* Handle grace-period start. */
		for (;;) {
2205
			trace_rcu_grace_period(rsp->name,
2206
					       READ_ONCE(rsp->gpnum),
2207
					       TPS("reqwait"));
2208
			rsp->gp_state = RCU_GP_WAIT_GPS;
2209
			swait_event_interruptible(rsp->gp_wq,
2210
						 READ_ONCE(rsp->gp_flags) &
2211
						 RCU_GP_FLAG_INIT);
2212
			rsp->gp_state = RCU_GP_DONE_GPS;
2213
			/* Locking provides needed memory barrier. */
2214
			if (rcu_gp_init(rsp))
2215
				break;
2216
			cond_resched_rcu_qs();
2217
			WRITE_ONCE(rsp->gp_activity, jiffies);
2218
			WARN_ON(signal_pending(current));
2219
			trace_rcu_grace_period(rsp->name,
2220
					       READ_ONCE(rsp->gpnum),
2221
					       TPS("reqwaitsig"));
2222
		}
2223

2224
		/* Handle quiescent-state forcing. */
2225
		first_gp_fqs = true;
2226 2227 2228 2229 2230
		j = jiffies_till_first_fqs;
		if (j > HZ) {
			j = HZ;
			jiffies_till_first_fqs = HZ;
		}
2231
		ret = 0;
2232
		for (;;) {
2233
			if (!ret) {
2234
				rsp->jiffies_force_qs = jiffies + j;
2235 2236 2237
				WRITE_ONCE(rsp->jiffies_kick_kthreads,
					   jiffies + 3 * j);
			}
2238
			trace_rcu_grace_period(rsp->name,
2239
					       READ_ONCE(rsp->gpnum),
2240
					       TPS("fqswait"));
2241
			rsp->gp_state = RCU_GP_WAIT_FQS;
2242
			ret = swait_event_interruptible_timeout(rsp->gp_wq,
2243
					rcu_gp_fqs_check_wake(rsp, &gf), j);
2244
			rsp->gp_state = RCU_GP_DOING_FQS;
2245
			/* Locking provides needed memory barriers. */
2246
			/* If grace period done, leave loop. */
2247
			if (!READ_ONCE(rnp->qsmask) &&
2248
			    !rcu_preempt_blocked_readers_cgp(rnp))
2249
				break;
2250
			/* If time for quiescent-state forcing, do it. */
2251 2252
			if (ULONG_CMP_GE(jiffies, rsp->jiffies_force_qs) ||
			    (gf & RCU_GP_FLAG_FQS)) {
2253
				trace_rcu_grace_period(rsp->name,
2254
						       READ_ONCE(rsp->gpnum),
2255
						       TPS("fqsstart"));
2256 2257
				rcu_gp_fqs(rsp, first_gp_fqs);
				first_gp_fqs = false;
2258
				trace_rcu_grace_period(rsp->name,
2259
						       READ_ONCE(rsp->gpnum),
2260
						       TPS("fqsend"));
2261
				cond_resched_rcu_qs();
2262
				WRITE_ONCE(rsp->gp_activity, jiffies);
2263 2264 2265 2266 2267 2268 2269 2270 2271
				ret = 0; /* Force full wait till next FQS. */
				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;
				}
2272 2273
			} else {
				/* Deal with stray signal. */
2274
				cond_resched_rcu_qs();
2275
				WRITE_ONCE(rsp->gp_activity, jiffies);
2276
				WARN_ON(signal_pending(current));
2277
				trace_rcu_grace_period(rsp->name,
2278
						       READ_ONCE(rsp->gpnum),
2279
						       TPS("fqswaitsig"));
2280 2281 2282 2283 2284 2285
				ret = 1; /* Keep old FQS timing. */
				j = jiffies;
				if (time_after(jiffies, rsp->jiffies_force_qs))
					j = 1;
				else
					j = rsp->jiffies_force_qs - j;
2286
			}
2287
		}
2288 2289

		/* Handle grace-period end. */
2290
		rsp->gp_state = RCU_GP_CLEANUP;
2291
		rcu_gp_cleanup(rsp);
2292
		rsp->gp_state = RCU_GP_CLEANED;
2293 2294 2295
	}
}

2296 2297 2298
/*
 * Start a new RCU grace period if warranted, re-initializing the hierarchy
 * in preparation for detecting the next grace period.  The caller must hold
2299
 * the root node's ->lock and hard irqs must be disabled.
2300 2301 2302 2303
 *
 * 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.
2304 2305
 *
 * Returns true if the grace-period kthread must be awakened.
2306
 */
2307
static bool
2308 2309
rcu_start_gp_advanced(struct rcu_state *rsp, struct rcu_node *rnp,
		      struct rcu_data *rdp)
2310
{
2311
	lockdep_assert_held(&rnp->lock);
2312
	if (!rsp->gp_kthread || !cpu_needs_another_gp(rsp, rdp)) {
2313
		/*
2314
		 * Either we have not yet spawned the grace-period
2315 2316
		 * task, this CPU does not need another grace period,
		 * or a grace period is already in progress.
2317
		 * Either way, don't start a new grace period.
2318
		 */
2319
		return false;
2320
	}
2321 2322
	WRITE_ONCE(rsp->gp_flags, RCU_GP_FLAG_INIT);
	trace_rcu_grace_period(rsp->name, READ_ONCE(rsp->gpnum),
2323
			       TPS("newreq"));
2324

2325 2326
	/*
	 * We can't do wakeups while holding the rnp->lock, as that
2327
	 * could cause possible deadlocks with the rq->lock. Defer
2328
	 * the wakeup to our caller.
2329
	 */
2330
	return true;
2331 2332
}

2333 2334 2335 2336 2337 2338
/*
 * Similar to rcu_start_gp_advanced(), but also advance the calling CPU's
 * callbacks.  Note that rcu_start_gp_advanced() cannot do this because it
 * is invoked indirectly from rcu_advance_cbs(), which would result in
 * endless recursion -- or would do so if it wasn't for the self-deadlock
 * that is encountered beforehand.
2339 2340
 *
 * Returns true if the grace-period kthread needs to be awakened.
2341
 */
2342
static bool rcu_start_gp(struct rcu_state *rsp)
2343 2344 2345
{
	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
	struct rcu_node *rnp = rcu_get_root(rsp);
2346
	bool ret = false;
2347 2348 2349 2350 2351 2352 2353 2354 2355

	/*
	 * If there is no grace period in progress right now, any
	 * callbacks we have up to this point will be satisfied by the
	 * next grace period.  Also, advancing the callbacks reduces the
	 * probability of false positives from cpu_needs_another_gp()
	 * resulting in pointless grace periods.  So, advance callbacks
	 * then start the grace period!
	 */
2356 2357 2358
	ret = rcu_advance_cbs(rsp, rnp, rdp) || ret;
	ret = rcu_start_gp_advanced(rsp, rnp, rdp) || ret;
	return ret;
2359 2360
}

2361
/*
2362 2363 2364 2365 2366 2367 2368
 * Report a full set of quiescent states to the specified rcu_state data
 * structure.  Invoke rcu_gp_kthread_wake() to awaken the grace-period
 * kthread if another grace period is required.  Whether we wake
 * the grace-period kthread or it awakens itself for the next round
 * of quiescent-state forcing, that kthread will clean up after the
 * just-completed grace period.  Note that the caller must hold rnp->lock,
 * which is released before return.
2369
 */
P
Paul E. McKenney 已提交
2370
static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)
2371
	__releases(rcu_get_root(rsp)->lock)
2372
{
2373
	lockdep_assert_held(&rcu_get_root(rsp)->lock);
2374
	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
2375
	WRITE_ONCE(rsp->gp_flags, READ_ONCE(rsp->gp_flags) | RCU_GP_FLAG_FQS);
B
Boqun Feng 已提交
2376
	raw_spin_unlock_irqrestore_rcu_node(rcu_get_root(rsp), flags);
2377
	rcu_gp_kthread_wake(rsp);
2378 2379
}

2380
/*
P
Paul E. McKenney 已提交
2381 2382 2383
 * 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
2384 2385 2386 2387 2388
 * must be represented by the same rcu_node structure (which need not be a
 * leaf rcu_node structure, though it often will be).  The gps parameter
 * is the grace-period snapshot, which means that the quiescent states
 * are valid only if rnp->gpnum is equal to gps.  That structure's lock
 * must be held upon entry, and it is released before return.
2389 2390
 */
static void
P
Paul E. McKenney 已提交
2391
rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,
2392
		  struct rcu_node *rnp, unsigned long gps, unsigned long flags)
2393 2394
	__releases(rnp->lock)
{
2395
	unsigned long oldmask = 0;
2396 2397
	struct rcu_node *rnp_c;

2398 2399
	lockdep_assert_held(&rnp->lock);

2400 2401
	/* Walk up the rcu_node hierarchy. */
	for (;;) {
2402
		if (!(rnp->qsmask & mask) || rnp->gpnum != gps) {
2403

2404 2405 2406 2407
			/*
			 * Our bit has already been cleared, or the
			 * relevant grace period is already over, so done.
			 */
B
Boqun Feng 已提交
2408
			raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2409 2410
			return;
		}
2411
		WARN_ON_ONCE(oldmask); /* Any child must be all zeroed! */
2412
		rnp->qsmask &= ~mask;
2413 2414 2415 2416
		trace_rcu_quiescent_state_report(rsp->name, rnp->gpnum,
						 mask, rnp->qsmask, rnp->level,
						 rnp->grplo, rnp->grphi,
						 !!rnp->gp_tasks);
2417
		if (rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {
2418 2419

			/* Other bits still set at this level, so done. */
B
Boqun Feng 已提交
2420
			raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2421 2422 2423 2424 2425 2426 2427 2428 2429
			return;
		}
		mask = rnp->grpmask;
		if (rnp->parent == NULL) {

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

			break;
		}
B
Boqun Feng 已提交
2430
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2431
		rnp_c = rnp;
2432
		rnp = rnp->parent;
2433
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
2434
		oldmask = rnp_c->qsmask;
2435 2436 2437 2438
	}

	/*
	 * Get here if we are the last CPU to pass through a quiescent
P
Paul E. McKenney 已提交
2439
	 * state for this grace period.  Invoke rcu_report_qs_rsp()
2440
	 * to clean up and start the next grace period if one is needed.
2441
	 */
P
Paul E. McKenney 已提交
2442
	rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */
2443 2444
}

2445 2446 2447 2448 2449 2450 2451
/*
 * Record a quiescent state for all tasks that were previously queued
 * on the specified rcu_node structure and that were blocking the current
 * RCU grace period.  The caller must hold the specified rnp->lock with
 * irqs disabled, and this lock is released upon return, but irqs remain
 * disabled.
 */
2452
static void rcu_report_unblock_qs_rnp(struct rcu_state *rsp,
2453 2454 2455
				      struct rcu_node *rnp, unsigned long flags)
	__releases(rnp->lock)
{
2456
	unsigned long gps;
2457 2458 2459
	unsigned long mask;
	struct rcu_node *rnp_p;

2460
	lockdep_assert_held(&rnp->lock);
2461 2462
	if (rcu_state_p == &rcu_sched_state || rsp != rcu_state_p ||
	    rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {
B
Boqun Feng 已提交
2463
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2464 2465 2466 2467 2468 2469
		return;  /* Still need more quiescent states! */
	}

	rnp_p = rnp->parent;
	if (rnp_p == NULL) {
		/*
2470 2471
		 * Only one rcu_node structure in the tree, so don't
		 * try to report up to its nonexistent parent!
2472 2473 2474 2475 2476
		 */
		rcu_report_qs_rsp(rsp, flags);
		return;
	}

2477 2478
	/* Report up the rest of the hierarchy, tracking current ->gpnum. */
	gps = rnp->gpnum;
2479
	mask = rnp->grpmask;
B
Boqun Feng 已提交
2480
	raw_spin_unlock_rcu_node(rnp);	/* irqs remain disabled. */
2481
	raw_spin_lock_rcu_node(rnp_p);	/* irqs already disabled. */
2482
	rcu_report_qs_rnp(mask, rsp, rnp_p, gps, flags);
2483 2484
}

2485
/*
P
Paul E. McKenney 已提交
2486
 * Record a quiescent state for the specified CPU to that CPU's rcu_data
2487
 * structure.  This must be called from the specified CPU.
2488 2489
 */
static void
2490
rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp)
2491 2492 2493
{
	unsigned long flags;
	unsigned long mask;
2494
	bool needwake;
2495 2496 2497
	struct rcu_node *rnp;

	rnp = rdp->mynode;
2498
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
2499 2500
	if (rdp->cpu_no_qs.b.norm || rdp->gpnum != rnp->gpnum ||
	    rnp->completed == rnp->gpnum || rdp->gpwrap) {
2501 2502

		/*
2503 2504 2505 2506
		 * 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.
2507
		 */
2508
		rdp->cpu_no_qs.b.norm = true;	/* need qs for new gp. */
2509
		rdp->rcu_qs_ctr_snap = __this_cpu_read(rcu_dynticks.rcu_qs_ctr);
B
Boqun Feng 已提交
2510
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2511 2512 2513 2514
		return;
	}
	mask = rdp->grpmask;
	if ((rnp->qsmask & mask) == 0) {
B
Boqun Feng 已提交
2515
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2516
	} else {
2517
		rdp->core_needs_qs = false;
2518 2519 2520 2521 2522

		/*
		 * This GP can't end until cpu checks in, so all of our
		 * callbacks can be processed during the next GP.
		 */
2523
		needwake = rcu_accelerate_cbs(rsp, rnp, rdp);
2524

2525 2526
		rcu_report_qs_rnp(mask, rsp, rnp, rnp->gpnum, flags);
		/* ^^^ Released rnp->lock */
2527 2528
		if (needwake)
			rcu_gp_kthread_wake(rsp);
2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540
	}
}

/*
 * 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)
{
2541 2542
	/* Check for grace-period ends and beginnings. */
	note_gp_changes(rsp, rdp);
2543 2544 2545 2546 2547

	/*
	 * 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.
	 */
2548
	if (!rdp->core_needs_qs)
2549 2550 2551 2552 2553 2554
		return;

	/*
	 * Was there a quiescent state since the beginning of the grace
	 * period? If no, then exit and wait for the next call.
	 */
2555
	if (rdp->cpu_no_qs.b.norm)
2556 2557
		return;

P
Paul E. McKenney 已提交
2558 2559 2560 2561
	/*
	 * Tell RCU we are done (but rcu_report_qs_rdp() will be the
	 * judge of that).
	 */
2562
	rcu_report_qs_rdp(rdp->cpu, rsp, rdp);
2563 2564
}

2565 2566 2567 2568 2569
/*
 * Trace the fact that this CPU is going offline.
 */
static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
{
2570 2571 2572
	RCU_TRACE(unsigned long mask;)
	RCU_TRACE(struct rcu_data *rdp = this_cpu_ptr(rsp->rda);)
	RCU_TRACE(struct rcu_node *rnp = rdp->mynode;)
2573

2574 2575 2576
	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU))
		return;

2577
	RCU_TRACE(mask = rdp->grpmask;)
2578 2579
	trace_rcu_grace_period(rsp->name,
			       rnp->gpnum + 1 - !!(rnp->qsmask & mask),
2580
			       TPS("cpuofl"));
2581 2582
}

2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604
/*
 * All CPUs for the specified rcu_node structure have gone offline,
 * and all tasks that were preempted within an RCU read-side critical
 * section while running on one of those CPUs have since exited their RCU
 * read-side critical section.  Some other CPU is reporting this fact with
 * the specified rcu_node structure's ->lock held and interrupts disabled.
 * This function therefore goes up the tree of rcu_node structures,
 * clearing the corresponding bits in the ->qsmaskinit fields.  Note that
 * the leaf rcu_node structure's ->qsmaskinit field has already been
 * updated
 *
 * This function does check that the specified rcu_node structure has
 * all CPUs offline and no blocked tasks, so it is OK to invoke it
 * prematurely.  That said, invoking it after the fact will cost you
 * a needless lock acquisition.  So once it has done its work, don't
 * invoke it again.
 */
static void rcu_cleanup_dead_rnp(struct rcu_node *rnp_leaf)
{
	long mask;
	struct rcu_node *rnp = rnp_leaf;

2605
	lockdep_assert_held(&rnp->lock);
2606 2607
	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU) ||
	    rnp->qsmaskinit || rcu_preempt_has_tasks(rnp))
2608 2609 2610 2611 2612 2613
		return;
	for (;;) {
		mask = rnp->grpmask;
		rnp = rnp->parent;
		if (!rnp)
			break;
2614
		raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */
2615
		rnp->qsmaskinit &= ~mask;
2616
		rnp->qsmask &= ~mask;
2617
		if (rnp->qsmaskinit) {
B
Boqun Feng 已提交
2618 2619
			raw_spin_unlock_rcu_node(rnp);
			/* irqs remain disabled. */
2620 2621
			return;
		}
B
Boqun Feng 已提交
2622
		raw_spin_unlock_rcu_node(rnp); /* irqs remain disabled. */
2623 2624 2625
	}
}

2626
/*
2627
 * The CPU has been completely removed, and some other CPU is reporting
2628 2629 2630
 * this fact from process context.  Do the remainder of the cleanup.
 * There can only be one CPU hotplug operation at a time, so no need for
 * explicit locking.
2631
 */
2632
static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
2633
{
2634
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
2635
	struct rcu_node *rnp = rdp->mynode;  /* Outgoing CPU's rdp & rnp. */
2636

2637 2638 2639
	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU))
		return;

2640
	/* Adjust any no-longer-needed kthreads. */
T
Thomas Gleixner 已提交
2641
	rcu_boost_kthread_setaffinity(rnp, -1);
2642 2643 2644 2645 2646 2647
}

/*
 * Invoke any RCU callbacks that have made it to the end of their grace
 * period.  Thottle as specified by rdp->blimit.
 */
2648
static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)
2649 2650
{
	unsigned long flags;
2651 2652 2653
	struct rcu_head *rhp;
	struct rcu_cblist rcl = RCU_CBLIST_INITIALIZER(rcl);
	long bl, count;
2654

2655
	/* If no callbacks are ready, just return. */
2656 2657 2658 2659 2660 2661
	if (!rcu_segcblist_ready_cbs(&rdp->cblist)) {
		trace_rcu_batch_start(rsp->name,
				      rcu_segcblist_n_lazy_cbs(&rdp->cblist),
				      rcu_segcblist_n_cbs(&rdp->cblist), 0);
		trace_rcu_batch_end(rsp->name, 0,
				    !rcu_segcblist_empty(&rdp->cblist),
2662 2663
				    need_resched(), is_idle_task(current),
				    rcu_is_callbacks_kthread());
2664
		return;
2665
	}
2666 2667 2668

	/*
	 * Extract the list of ready callbacks, disabling to prevent
2669 2670
	 * races with call_rcu() from interrupt handlers.  Leave the
	 * callback counts, as rcu_barrier() needs to be conservative.
2671 2672
	 */
	local_irq_save(flags);
2673
	WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));
2674
	bl = rdp->blimit;
2675 2676 2677
	trace_rcu_batch_start(rsp->name, rcu_segcblist_n_lazy_cbs(&rdp->cblist),
			      rcu_segcblist_n_cbs(&rdp->cblist), bl);
	rcu_segcblist_extract_done_cbs(&rdp->cblist, &rcl);
2678 2679 2680
	local_irq_restore(flags);

	/* Invoke callbacks. */
2681 2682 2683 2684 2685 2686 2687 2688 2689
	rhp = rcu_cblist_dequeue(&rcl);
	for (; rhp; rhp = rcu_cblist_dequeue(&rcl)) {
		debug_rcu_head_unqueue(rhp);
		if (__rcu_reclaim(rsp->name, rhp))
			rcu_cblist_dequeued_lazy(&rcl);
		/*
		 * Stop only if limit reached and CPU has something to do.
		 * Note: The rcl structure counts down from zero.
		 */
2690
		if (-rcl.len >= bl &&
2691 2692
		    (need_resched() ||
		     (!is_idle_task(current) && !rcu_is_callbacks_kthread())))
2693 2694 2695 2696
			break;
	}

	local_irq_save(flags);
2697
	count = -rcl.len;
2698 2699
	trace_rcu_batch_end(rsp->name, count, !!rcl.head, need_resched(),
			    is_idle_task(current), rcu_is_callbacks_kthread());
2700

2701 2702
	/* Update counts and requeue any remaining callbacks. */
	rcu_segcblist_insert_done_cbs(&rdp->cblist, &rcl);
2703 2704
	smp_mb(); /* List handling before counting for rcu_barrier(). */
	rdp->n_cbs_invoked += count;
2705
	rcu_segcblist_insert_count(&rdp->cblist, &rcl);
2706 2707

	/* Reinstate batch limit if we have worked down the excess. */
2708 2709
	count = rcu_segcblist_n_cbs(&rdp->cblist);
	if (rdp->blimit == LONG_MAX && count <= qlowmark)
2710 2711
		rdp->blimit = blimit;

2712
	/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */
2713
	if (count == 0 && rdp->qlen_last_fqs_check != 0) {
2714 2715
		rdp->qlen_last_fqs_check = 0;
		rdp->n_force_qs_snap = rsp->n_force_qs;
2716 2717 2718
	} else if (count < rdp->qlen_last_fqs_check - qhimark)
		rdp->qlen_last_fqs_check = count;
	WARN_ON_ONCE(rcu_segcblist_empty(&rdp->cblist) != (count == 0));
2719

2720 2721
	local_irq_restore(flags);

2722
	/* Re-invoke RCU core processing if there are callbacks remaining. */
2723
	if (rcu_segcblist_ready_cbs(&rdp->cblist))
2724
		invoke_rcu_core();
2725 2726 2727 2728 2729
}

/*
 * 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).
2730
 * Also schedule RCU core processing.
2731
 *
2732
 * This function must be called from hardirq context.  It is normally
2733
 * invoked from the scheduling-clock interrupt.
2734
 */
2735
void rcu_check_callbacks(int user)
2736
{
2737
	trace_rcu_utilization(TPS("Start scheduler-tick"));
2738
	increment_cpu_stall_ticks();
2739
	if (user || rcu_is_cpu_rrupt_from_idle()) {
2740 2741 2742 2743 2744

		/*
		 * 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
2745
		 * a quiescent state, so note it.
2746 2747
		 *
		 * No memory barrier is required here because both
2748 2749 2750
		 * 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.
2751 2752
		 */

2753 2754
		rcu_sched_qs();
		rcu_bh_qs();
2755 2756 2757 2758 2759 2760 2761

	} 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
2762
		 * critical section, so note it.
2763 2764
		 */

2765
		rcu_bh_qs();
2766
	}
2767
	rcu_preempt_check_callbacks();
2768
	if (rcu_pending())
2769
		invoke_rcu_core();
P
Paul E. McKenney 已提交
2770 2771
	if (user)
		rcu_note_voluntary_context_switch(current);
2772
	trace_rcu_utilization(TPS("End scheduler-tick"));
2773 2774 2775 2776 2777
}

/*
 * Scan the leaf rcu_node structures, processing dyntick state for any that
 * have not yet encountered a quiescent state, using the function specified.
2778 2779
 * Also initiate boosting for any threads blocked on the root rcu_node.
 *
2780
 * The caller must have suppressed start of new grace periods.
2781
 */
2782
static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *rsp))
2783 2784 2785 2786
{
	int cpu;
	unsigned long flags;
	unsigned long mask;
2787
	struct rcu_node *rnp;
2788

2789
	rcu_for_each_leaf_node(rsp, rnp) {
2790
		cond_resched_rcu_qs();
2791
		mask = 0;
2792
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
2793
		if (rnp->qsmask == 0) {
2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816
			if (rcu_state_p == &rcu_sched_state ||
			    rsp != rcu_state_p ||
			    rcu_preempt_blocked_readers_cgp(rnp)) {
				/*
				 * No point in scanning bits because they
				 * are all zero.  But we might need to
				 * priority-boost blocked readers.
				 */
				rcu_initiate_boost(rnp, flags);
				/* rcu_initiate_boost() releases rnp->lock */
				continue;
			}
			if (rnp->parent &&
			    (rnp->parent->qsmask & rnp->grpmask)) {
				/*
				 * Race between grace-period
				 * initialization and task exiting RCU
				 * read-side critical section: Report.
				 */
				rcu_report_unblock_qs_rnp(rsp, rnp, flags);
				/* rcu_report_unblock_qs_rnp() rlses ->lock */
				continue;
			}
2817
		}
2818 2819
		for_each_leaf_node_possible_cpu(rnp, cpu) {
			unsigned long bit = leaf_node_cpu_bit(rnp, cpu);
2820
			if ((rnp->qsmask & bit) != 0) {
2821
				if (f(per_cpu_ptr(rsp->rda, cpu)))
2822 2823
					mask |= bit;
			}
2824
		}
2825
		if (mask != 0) {
2826 2827
			/* Idle/offline CPUs, report (releases rnp->lock. */
			rcu_report_qs_rnp(mask, rsp, rnp, rnp->gpnum, flags);
2828 2829
		} else {
			/* Nothing to do here, so just drop the lock. */
B
Boqun Feng 已提交
2830
			raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
2831 2832 2833 2834 2835 2836 2837 2838
		}
	}
}

/*
 * Force quiescent states on reluctant CPUs, and also detect which
 * CPUs are in dyntick-idle mode.
 */
2839
static void force_quiescent_state(struct rcu_state *rsp)
2840 2841
{
	unsigned long flags;
2842 2843 2844 2845 2846
	bool ret;
	struct rcu_node *rnp;
	struct rcu_node *rnp_old = NULL;

	/* Funnel through hierarchy to reduce memory contention. */
2847
	rnp = __this_cpu_read(rsp->rda->mynode);
2848
	for (; rnp != NULL; rnp = rnp->parent) {
2849
		ret = (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) ||
2850 2851 2852 2853
		      !raw_spin_trylock(&rnp->fqslock);
		if (rnp_old != NULL)
			raw_spin_unlock(&rnp_old->fqslock);
		if (ret) {
2854
			rsp->n_force_qs_lh++;
2855 2856 2857 2858 2859
			return;
		}
		rnp_old = rnp;
	}
	/* rnp_old == rcu_get_root(rsp), rnp == NULL. */
2860

2861
	/* Reached the root of the rcu_node tree, acquire lock. */
2862
	raw_spin_lock_irqsave_rcu_node(rnp_old, flags);
2863
	raw_spin_unlock(&rnp_old->fqslock);
2864
	if (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) {
2865
		rsp->n_force_qs_lh++;
B
Boqun Feng 已提交
2866
		raw_spin_unlock_irqrestore_rcu_node(rnp_old, flags);
2867
		return;  /* Someone beat us to it. */
2868
	}
2869
	WRITE_ONCE(rsp->gp_flags, READ_ONCE(rsp->gp_flags) | RCU_GP_FLAG_FQS);
B
Boqun Feng 已提交
2870
	raw_spin_unlock_irqrestore_rcu_node(rnp_old, flags);
2871
	rcu_gp_kthread_wake(rsp);
2872 2873 2874
}

/*
2875 2876 2877
 * 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.
2878 2879
 */
static void
2880
__rcu_process_callbacks(struct rcu_state *rsp)
2881 2882
{
	unsigned long flags;
2883
	bool needwake;
2884
	struct rcu_data *rdp = raw_cpu_ptr(rsp->rda);
2885

N
Nicholas Mc Guire 已提交
2886
	WARN_ON_ONCE(!rdp->beenonline);
2887

2888 2889 2890 2891
	/* 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? */
2892
	local_irq_save(flags);
2893
	if (cpu_needs_another_gp(rsp, rdp)) {
2894
		raw_spin_lock_rcu_node(rcu_get_root(rsp)); /* irqs disabled. */
2895
		needwake = rcu_start_gp(rsp);
B
Boqun Feng 已提交
2896
		raw_spin_unlock_irqrestore_rcu_node(rcu_get_root(rsp), flags);
2897 2898
		if (needwake)
			rcu_gp_kthread_wake(rsp);
2899 2900
	} else {
		local_irq_restore(flags);
2901 2902 2903
	}

	/* If there are callbacks ready, invoke them. */
2904
	if (rcu_segcblist_ready_cbs(&rdp->cblist))
2905
		invoke_rcu_callbacks(rsp, rdp);
2906 2907 2908

	/* Do any needed deferred wakeups of rcuo kthreads. */
	do_nocb_deferred_wakeup(rdp);
2909 2910
}

2911
/*
2912
 * Do RCU core processing for the current CPU.
2913
 */
2914
static __latent_entropy void rcu_process_callbacks(struct softirq_action *unused)
2915
{
2916 2917
	struct rcu_state *rsp;

2918 2919
	if (cpu_is_offline(smp_processor_id()))
		return;
2920
	trace_rcu_utilization(TPS("Start RCU core"));
2921 2922
	for_each_rcu_flavor(rsp)
		__rcu_process_callbacks(rsp);
2923
	trace_rcu_utilization(TPS("End RCU core"));
2924 2925
}

2926
/*
2927 2928 2929
 * 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
2930
 * are running on the current CPU with softirqs disabled, the
2931
 * rcu_cpu_kthread_task cannot disappear out from under us.
2932
 */
2933
static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)
2934
{
2935
	if (unlikely(!READ_ONCE(rcu_scheduler_fully_active)))
2936
		return;
2937 2938
	if (likely(!rsp->boost)) {
		rcu_do_batch(rsp, rdp);
2939 2940
		return;
	}
2941
	invoke_rcu_callbacks_kthread();
2942 2943
}

2944
static void invoke_rcu_core(void)
2945
{
2946 2947
	if (cpu_online(smp_processor_id()))
		raise_softirq(RCU_SOFTIRQ);
2948 2949
}

2950 2951 2952 2953 2954
/*
 * 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)
2955
{
2956 2957
	bool needwake;

2958 2959 2960 2961
	/*
	 * If called from an extended quiescent state, invoke the RCU
	 * core in order to force a re-evaluation of RCU's idleness.
	 */
2962
	if (!rcu_is_watching())
2963 2964
		invoke_rcu_core();

2965
	/* If interrupts were disabled or CPU offline, don't invoke RCU core. */
2966
	if (irqs_disabled_flags(flags) || cpu_is_offline(smp_processor_id()))
2967
		return;
2968

2969 2970 2971 2972 2973 2974 2975
	/*
	 * 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.
	 */
2976 2977
	if (unlikely(rcu_segcblist_n_cbs(&rdp->cblist) >
		     rdp->qlen_last_fqs_check + qhimark)) {
2978 2979

		/* Are we ignoring a completed grace period? */
2980
		note_gp_changes(rsp, rdp);
2981 2982 2983 2984 2985

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

2986
			raw_spin_lock_rcu_node(rnp_root);
2987
			needwake = rcu_start_gp(rsp);
B
Boqun Feng 已提交
2988
			raw_spin_unlock_rcu_node(rnp_root);
2989 2990
			if (needwake)
				rcu_gp_kthread_wake(rsp);
2991 2992 2993 2994
		} else {
			/* Give the grace period a kick. */
			rdp->blimit = LONG_MAX;
			if (rsp->n_force_qs == rdp->n_force_qs_snap &&
2995
			    rcu_segcblist_first_pend_cb(&rdp->cblist) != head)
2996
				force_quiescent_state(rsp);
2997
			rdp->n_force_qs_snap = rsp->n_force_qs;
2998
			rdp->qlen_last_fqs_check = rcu_segcblist_n_cbs(&rdp->cblist);
2999
		}
3000
	}
3001 3002
}

3003 3004 3005 3006 3007 3008 3009
/*
 * RCU callback function to leak a callback.
 */
static void rcu_leak_callback(struct rcu_head *rhp)
{
}

P
Paul E. McKenney 已提交
3010 3011 3012 3013 3014 3015
/*
 * Helper function for call_rcu() and friends.  The cpu argument will
 * normally be -1, indicating "currently running CPU".  It may specify
 * a CPU only if that CPU is a no-CBs CPU.  Currently, only _rcu_barrier()
 * is expected to specify a CPU.
 */
3016
static void
3017
__call_rcu(struct rcu_head *head, rcu_callback_t func,
P
Paul E. McKenney 已提交
3018
	   struct rcu_state *rsp, int cpu, bool lazy)
3019 3020 3021 3022
{
	unsigned long flags;
	struct rcu_data *rdp;

3023 3024 3025
	/* Misaligned rcu_head! */
	WARN_ON_ONCE((unsigned long)head & (sizeof(void *) - 1));

3026
	if (debug_rcu_head_queue(head)) {
3027 3028 3029 3030 3031 3032 3033
		/*
		 * Probable double call_rcu(), so leak the callback.
		 * Use rcu:rcu_callback trace event to find the previous
		 * time callback was passed to __call_rcu().
		 */
		WARN_ONCE(1, "__call_rcu(): Double-freed CB %p->%pF()!!!\n",
			  head, head->func);
3034
		WRITE_ONCE(head->func, rcu_leak_callback);
3035 3036
		return;
	}
3037 3038 3039
	head->func = func;
	head->next = NULL;
	local_irq_save(flags);
3040
	rdp = this_cpu_ptr(rsp->rda);
3041 3042

	/* Add the callback to our list. */
3043
	if (unlikely(!rcu_segcblist_is_enabled(&rdp->cblist)) || cpu != -1) {
P
Paul E. McKenney 已提交
3044 3045 3046 3047
		int offline;

		if (cpu != -1)
			rdp = per_cpu_ptr(rsp->rda, cpu);
3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060
		if (likely(rdp->mynode)) {
			/* Post-boot, so this should be for a no-CBs CPU. */
			offline = !__call_rcu_nocb(rdp, head, lazy, flags);
			WARN_ON_ONCE(offline);
			/* Offline CPU, _call_rcu() illegal, leak callback.  */
			local_irq_restore(flags);
			return;
		}
		/*
		 * Very early boot, before rcu_init().  Initialize if needed
		 * and then drop through to queue the callback.
		 */
		BUG_ON(cpu != -1);
3061
		WARN_ON_ONCE(!rcu_is_watching());
3062 3063
		if (rcu_segcblist_empty(&rdp->cblist))
			rcu_segcblist_init(&rdp->cblist);
3064
	}
3065 3066
	rcu_segcblist_enqueue(&rdp->cblist, head, lazy);
	if (!lazy)
3067
		rcu_idle_count_callbacks_posted();
3068

3069 3070
	if (__is_kfree_rcu_offset((unsigned long)func))
		trace_rcu_kfree_callback(rsp->name, head, (unsigned long)func,
3071 3072
					 rcu_segcblist_n_lazy_cbs(&rdp->cblist),
					 rcu_segcblist_n_cbs(&rdp->cblist));
3073
	else
3074 3075 3076
		trace_rcu_callback(rsp->name, head,
				   rcu_segcblist_n_lazy_cbs(&rdp->cblist),
				   rcu_segcblist_n_cbs(&rdp->cblist));
3077

3078 3079
	/* Go handle any RCU core processing required. */
	__call_rcu_core(rsp, rdp, head, flags);
3080 3081 3082
	local_irq_restore(flags);
}

3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100
/**
 * call_rcu_sched() - Queue an RCU for invocation after sched grace period.
 * @head: structure to be used for queueing the RCU updates.
 * @func: actual callback function to be invoked after the grace period
 *
 * The callback function will be invoked some time after a full grace
 * period elapses, in other words after all currently executing RCU
 * read-side critical sections have completed. call_rcu_sched() assumes
 * that the read-side critical sections end on enabling of preemption
 * or on voluntary preemption.
 * RCU read-side critical sections are delimited by :
 *  - rcu_read_lock_sched() and rcu_read_unlock_sched(), OR
 *  - anything that disables preemption.
 *
 *  These may be nested.
 *
 * See the description of call_rcu() for more detailed information on
 * memory ordering guarantees.
3101
 */
3102
void call_rcu_sched(struct rcu_head *head, rcu_callback_t func)
3103
{
P
Paul E. McKenney 已提交
3104
	__call_rcu(head, func, &rcu_sched_state, -1, 0);
3105
}
3106
EXPORT_SYMBOL_GPL(call_rcu_sched);
3107

3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127
/**
 * call_rcu_bh() - Queue an RCU for invocation after a quicker grace period.
 * @head: structure to be used for queueing the RCU updates.
 * @func: actual callback function to be invoked after the grace period
 *
 * The callback function will be invoked some time after a full grace
 * period elapses, in other words after all currently executing RCU
 * read-side critical sections have completed. call_rcu_bh() assumes
 * that the read-side critical sections end on completion of a softirq
 * handler. This means that read-side critical sections in process
 * context must not be interrupted by softirqs. This interface is to be
 * used when most of the read-side critical sections are in softirq context.
 * RCU read-side critical sections are delimited by :
 *  - rcu_read_lock() and  rcu_read_unlock(), if in interrupt context.
 *  OR
 *  - rcu_read_lock_bh() and rcu_read_unlock_bh(), if in process context.
 *  These may be nested.
 *
 * See the description of call_rcu() for more detailed information on
 * memory ordering guarantees.
3128
 */
3129
void call_rcu_bh(struct rcu_head *head, rcu_callback_t func)
3130
{
P
Paul E. McKenney 已提交
3131
	__call_rcu(head, func, &rcu_bh_state, -1, 0);
3132 3133 3134
}
EXPORT_SYMBOL_GPL(call_rcu_bh);

3135 3136 3137 3138 3139 3140 3141 3142
/*
 * Queue an RCU callback for lazy invocation after a grace period.
 * This will likely be later named something like "call_rcu_lazy()",
 * but this change will require some way of tagging the lazy RCU
 * callbacks in the list of pending callbacks. Until then, this
 * function may only be called from __kfree_rcu().
 */
void kfree_call_rcu(struct rcu_head *head,
3143
		    rcu_callback_t func)
3144
{
3145
	__call_rcu(head, func, rcu_state_p, -1, 1);
3146 3147 3148
}
EXPORT_SYMBOL_GPL(kfree_call_rcu);

3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159
/*
 * 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)
{
3160 3161
	int ret;

3162
	might_sleep();  /* Check for RCU read-side critical section. */
3163 3164 3165 3166
	preempt_disable();
	ret = num_online_cpus() <= 1;
	preempt_enable();
	return ret;
3167 3168
}

3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180
/**
 * 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
3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202
 * non-threaded 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.
 *
 * Note that this guarantee implies further memory-ordering guarantees.
 * On systems with more than one CPU, when synchronize_sched() returns,
 * each CPU is guaranteed to have executed a full memory barrier since the
 * end of its last RCU-sched read-side critical section whose beginning
 * preceded the call to synchronize_sched().  In addition, each CPU having
 * an RCU read-side critical section that extends beyond the return from
 * synchronize_sched() is guaranteed to have executed a full memory barrier
 * after the beginning of synchronize_sched() and before the beginning of
 * that RCU read-side critical section.  Note that these guarantees include
 * CPUs that are offline, idle, or executing in user mode, as well as CPUs
 * that are executing in the kernel.
 *
 * Furthermore, if CPU A invoked synchronize_sched(), which returned
 * to its caller on CPU B, then both CPU A and CPU B are guaranteed
 * to have executed a full memory barrier during the execution of
 * synchronize_sched() -- even if CPU A and CPU B are the same CPU (but
 * again only if the system has more than one CPU).
3203 3204 3205
 */
void synchronize_sched(void)
{
3206 3207 3208 3209
	RCU_LOCKDEP_WARN(lock_is_held(&rcu_bh_lock_map) ||
			 lock_is_held(&rcu_lock_map) ||
			 lock_is_held(&rcu_sched_lock_map),
			 "Illegal synchronize_sched() in RCU-sched read-side critical section");
3210 3211
	if (rcu_blocking_is_gp())
		return;
3212
	if (rcu_gp_is_expedited())
3213 3214 3215
		synchronize_sched_expedited();
	else
		wait_rcu_gp(call_rcu_sched);
3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226
}
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.
3227 3228 3229
 *
 * See the description of synchronize_sched() for more detailed information
 * on memory ordering guarantees.
3230 3231 3232
 */
void synchronize_rcu_bh(void)
{
3233 3234 3235 3236
	RCU_LOCKDEP_WARN(lock_is_held(&rcu_bh_lock_map) ||
			 lock_is_held(&rcu_lock_map) ||
			 lock_is_held(&rcu_sched_lock_map),
			 "Illegal synchronize_rcu_bh() in RCU-bh read-side critical section");
3237 3238
	if (rcu_blocking_is_gp())
		return;
3239
	if (rcu_gp_is_expedited())
3240 3241 3242
		synchronize_rcu_bh_expedited();
	else
		wait_rcu_gp(call_rcu_bh);
3243 3244 3245
}
EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265
/**
 * get_state_synchronize_rcu - Snapshot current RCU state
 *
 * Returns a cookie that is used by a later call to cond_synchronize_rcu()
 * to determine whether or not a full grace period has elapsed in the
 * meantime.
 */
unsigned long get_state_synchronize_rcu(void)
{
	/*
	 * Any prior manipulation of RCU-protected data must happen
	 * before the load from ->gpnum.
	 */
	smp_mb();  /* ^^^ */

	/*
	 * Make sure this load happens before the purportedly
	 * time-consuming work between get_state_synchronize_rcu()
	 * and cond_synchronize_rcu().
	 */
3266
	return smp_load_acquire(&rcu_state_p->gpnum);
3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291
}
EXPORT_SYMBOL_GPL(get_state_synchronize_rcu);

/**
 * cond_synchronize_rcu - Conditionally wait for an RCU grace period
 *
 * @oldstate: return value from earlier call to get_state_synchronize_rcu()
 *
 * If a full RCU grace period has elapsed since the earlier call to
 * get_state_synchronize_rcu(), just return.  Otherwise, invoke
 * synchronize_rcu() to wait for a full grace period.
 *
 * Yes, this function does not take counter wrap into account.  But
 * counter wrap is harmless.  If the counter wraps, we have waited for
 * more than 2 billion grace periods (and way more on a 64-bit system!),
 * so waiting for one additional grace period should be just fine.
 */
void cond_synchronize_rcu(unsigned long oldstate)
{
	unsigned long newstate;

	/*
	 * Ensure that this load happens before any RCU-destructive
	 * actions the caller might carry out after we return.
	 */
3292
	newstate = smp_load_acquire(&rcu_state_p->completed);
3293 3294 3295 3296 3297
	if (ULONG_CMP_GE(oldstate, newstate))
		synchronize_rcu();
}
EXPORT_SYMBOL_GPL(cond_synchronize_rcu);

3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349
/**
 * get_state_synchronize_sched - Snapshot current RCU-sched state
 *
 * Returns a cookie that is used by a later call to cond_synchronize_sched()
 * to determine whether or not a full grace period has elapsed in the
 * meantime.
 */
unsigned long get_state_synchronize_sched(void)
{
	/*
	 * Any prior manipulation of RCU-protected data must happen
	 * before the load from ->gpnum.
	 */
	smp_mb();  /* ^^^ */

	/*
	 * Make sure this load happens before the purportedly
	 * time-consuming work between get_state_synchronize_sched()
	 * and cond_synchronize_sched().
	 */
	return smp_load_acquire(&rcu_sched_state.gpnum);
}
EXPORT_SYMBOL_GPL(get_state_synchronize_sched);

/**
 * cond_synchronize_sched - Conditionally wait for an RCU-sched grace period
 *
 * @oldstate: return value from earlier call to get_state_synchronize_sched()
 *
 * If a full RCU-sched grace period has elapsed since the earlier call to
 * get_state_synchronize_sched(), just return.  Otherwise, invoke
 * synchronize_sched() to wait for a full grace period.
 *
 * Yes, this function does not take counter wrap into account.  But
 * counter wrap is harmless.  If the counter wraps, we have waited for
 * more than 2 billion grace periods (and way more on a 64-bit system!),
 * so waiting for one additional grace period should be just fine.
 */
void cond_synchronize_sched(unsigned long oldstate)
{
	unsigned long newstate;

	/*
	 * Ensure that this load happens before any RCU-destructive
	 * actions the caller might carry out after we return.
	 */
	newstate = smp_load_acquire(&rcu_sched_state.completed);
	if (ULONG_CMP_GE(oldstate, newstate))
		synchronize_sched();
}
EXPORT_SYMBOL_GPL(cond_synchronize_sched);

3350 3351 3352 3353 3354 3355 3356 3357 3358
/*
 * 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)
{
3359 3360
	struct rcu_node *rnp = rdp->mynode;

3361 3362 3363 3364 3365
	rdp->n_rcu_pending++;

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

3366 3367 3368 3369
	/* Is this CPU a NO_HZ_FULL CPU that should ignore RCU? */
	if (rcu_nohz_full_cpu(rsp))
		return 0;

3370
	/* Is the RCU core waiting for a quiescent state from this CPU? */
3371
	if (rcu_scheduler_fully_active &&
3372
	    rdp->core_needs_qs && rdp->cpu_no_qs.b.norm &&
3373
	    rdp->rcu_qs_ctr_snap == __this_cpu_read(rcu_dynticks.rcu_qs_ctr)) {
3374
		rdp->n_rp_core_needs_qs++;
3375
	} else if (rdp->core_needs_qs && !rdp->cpu_no_qs.b.norm) {
3376
		rdp->n_rp_report_qs++;
3377
		return 1;
3378
	}
3379 3380

	/* Does this CPU have callbacks ready to invoke? */
3381
	if (rcu_segcblist_ready_cbs(&rdp->cblist)) {
3382
		rdp->n_rp_cb_ready++;
3383
		return 1;
3384
	}
3385 3386

	/* Has RCU gone idle with this CPU needing another grace period? */
3387 3388
	if (cpu_needs_another_gp(rsp, rdp)) {
		rdp->n_rp_cpu_needs_gp++;
3389
		return 1;
3390
	}
3391 3392

	/* Has another RCU grace period completed?  */
3393
	if (READ_ONCE(rnp->completed) != rdp->completed) { /* outside lock */
3394
		rdp->n_rp_gp_completed++;
3395
		return 1;
3396
	}
3397 3398

	/* Has a new RCU grace period started? */
3399 3400
	if (READ_ONCE(rnp->gpnum) != rdp->gpnum ||
	    unlikely(READ_ONCE(rdp->gpwrap))) { /* outside lock */
3401
		rdp->n_rp_gp_started++;
3402
		return 1;
3403
	}
3404

3405 3406 3407 3408 3409 3410
	/* Does this CPU need a deferred NOCB wakeup? */
	if (rcu_nocb_need_deferred_wakeup(rdp)) {
		rdp->n_rp_nocb_defer_wakeup++;
		return 1;
	}

3411
	/* nothing to do */
3412
	rdp->n_rp_need_nothing++;
3413 3414 3415 3416 3417 3418 3419 3420
	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.
 */
3421
static int rcu_pending(void)
3422
{
3423 3424 3425
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
3426
		if (__rcu_pending(rsp, this_cpu_ptr(rsp->rda)))
3427 3428
			return 1;
	return 0;
3429 3430 3431
}

/*
3432 3433 3434
 * Return true if the specified CPU has any callback.  If all_lazy is
 * non-NULL, store an indication of whether all callbacks are lazy.
 * (If there are no callbacks, all of them are deemed to be lazy.)
3435
 */
3436
static bool __maybe_unused rcu_cpu_has_callbacks(bool *all_lazy)
3437
{
3438 3439 3440
	bool al = true;
	bool hc = false;
	struct rcu_data *rdp;
3441 3442
	struct rcu_state *rsp;

3443
	for_each_rcu_flavor(rsp) {
3444
		rdp = this_cpu_ptr(rsp->rda);
3445
		if (rcu_segcblist_empty(&rdp->cblist))
3446 3447
			continue;
		hc = true;
3448
		if (rcu_segcblist_n_nonlazy_cbs(&rdp->cblist) || !all_lazy) {
3449
			al = false;
3450 3451
			break;
		}
3452 3453 3454 3455
	}
	if (all_lazy)
		*all_lazy = al;
	return hc;
3456 3457
}

3458 3459 3460 3461
/*
 * Helper function for _rcu_barrier() tracing.  If tracing is disabled,
 * the compiler is expected to optimize this away.
 */
3462
static void _rcu_barrier_trace(struct rcu_state *rsp, const char *s,
3463 3464 3465 3466 3467 3468
			       int cpu, unsigned long done)
{
	trace_rcu_barrier(rsp->name, s, cpu,
			  atomic_read(&rsp->barrier_cpu_count), done);
}

3469 3470 3471 3472
/*
 * RCU callback function for _rcu_barrier().  If we are last, wake
 * up the task executing _rcu_barrier().
 */
3473
static void rcu_barrier_callback(struct rcu_head *rhp)
3474
{
3475 3476 3477
	struct rcu_data *rdp = container_of(rhp, struct rcu_data, barrier_head);
	struct rcu_state *rsp = rdp->rsp;

3478
	if (atomic_dec_and_test(&rsp->barrier_cpu_count)) {
3479
		_rcu_barrier_trace(rsp, "LastCB", -1, rsp->barrier_sequence);
3480
		complete(&rsp->barrier_completion);
3481
	} else {
3482
		_rcu_barrier_trace(rsp, "CB", -1, rsp->barrier_sequence);
3483
	}
3484 3485 3486 3487 3488 3489 3490
}

/*
 * Called with preemption disabled, and from cross-cpu IRQ context.
 */
static void rcu_barrier_func(void *type)
{
3491
	struct rcu_state *rsp = type;
3492
	struct rcu_data *rdp = raw_cpu_ptr(rsp->rda);
3493

3494
	_rcu_barrier_trace(rsp, "IRQ", -1, rsp->barrier_sequence);
3495 3496 3497 3498 3499 3500 3501 3502
	rdp->barrier_head.func = rcu_barrier_callback;
	debug_rcu_head_queue(&rdp->barrier_head);
	if (rcu_segcblist_entrain(&rdp->cblist, &rdp->barrier_head, 0)) {
		atomic_inc(&rsp->barrier_cpu_count);
	} else {
		debug_rcu_head_unqueue(&rdp->barrier_head);
		_rcu_barrier_trace(rsp, "IRQNQ", -1, rsp->barrier_sequence);
	}
3503 3504 3505 3506 3507 3508
}

/*
 * Orchestrate the specified type of RCU barrier, waiting for all
 * RCU callbacks of the specified type to complete.
 */
3509
static void _rcu_barrier(struct rcu_state *rsp)
3510
{
3511 3512
	int cpu;
	struct rcu_data *rdp;
3513
	unsigned long s = rcu_seq_snap(&rsp->barrier_sequence);
3514

3515
	_rcu_barrier_trace(rsp, "Begin", -1, s);
3516

3517
	/* Take mutex to serialize concurrent rcu_barrier() requests. */
3518
	mutex_lock(&rsp->barrier_mutex);
3519

3520 3521 3522
	/* Did someone else do our work for us? */
	if (rcu_seq_done(&rsp->barrier_sequence, s)) {
		_rcu_barrier_trace(rsp, "EarlyExit", -1, rsp->barrier_sequence);
3523 3524 3525 3526 3527
		smp_mb(); /* caller's subsequent code after above check. */
		mutex_unlock(&rsp->barrier_mutex);
		return;
	}

3528 3529 3530
	/* Mark the start of the barrier operation. */
	rcu_seq_start(&rsp->barrier_sequence);
	_rcu_barrier_trace(rsp, "Inc1", -1, rsp->barrier_sequence);
3531

3532
	/*
3533 3534
	 * 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
3535 3536
	 * (or preemption of this task).  Exclude CPU-hotplug operations
	 * to ensure that no offline CPU has callbacks queued.
3537
	 */
3538
	init_completion(&rsp->barrier_completion);
3539
	atomic_set(&rsp->barrier_cpu_count, 1);
3540
	get_online_cpus();
3541 3542

	/*
3543 3544 3545
	 * 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.
3546
	 */
P
Paul E. McKenney 已提交
3547
	for_each_possible_cpu(cpu) {
3548
		if (!cpu_online(cpu) && !rcu_is_nocb_cpu(cpu))
P
Paul E. McKenney 已提交
3549
			continue;
3550
		rdp = per_cpu_ptr(rsp->rda, cpu);
3551
		if (rcu_is_nocb_cpu(cpu)) {
3552 3553
			if (!rcu_nocb_cpu_needs_barrier(rsp, cpu)) {
				_rcu_barrier_trace(rsp, "OfflineNoCB", cpu,
3554
						   rsp->barrier_sequence);
3555 3556
			} else {
				_rcu_barrier_trace(rsp, "OnlineNoCB", cpu,
3557
						   rsp->barrier_sequence);
3558
				smp_mb__before_atomic();
3559 3560 3561 3562
				atomic_inc(&rsp->barrier_cpu_count);
				__call_rcu(&rdp->barrier_head,
					   rcu_barrier_callback, rsp, cpu, 0);
			}
3563
		} else if (rcu_segcblist_n_cbs(&rdp->cblist)) {
3564
			_rcu_barrier_trace(rsp, "OnlineQ", cpu,
3565
					   rsp->barrier_sequence);
3566
			smp_call_function_single(cpu, rcu_barrier_func, rsp, 1);
3567
		} else {
3568
			_rcu_barrier_trace(rsp, "OnlineNQ", cpu,
3569
					   rsp->barrier_sequence);
3570 3571
		}
	}
3572
	put_online_cpus();
3573 3574 3575 3576 3577

	/*
	 * Now that we have an rcu_barrier_callback() callback on each
	 * CPU, and thus each counted, remove the initial count.
	 */
3578
	if (atomic_dec_and_test(&rsp->barrier_cpu_count))
3579
		complete(&rsp->barrier_completion);
3580 3581

	/* Wait for all rcu_barrier_callback() callbacks to be invoked. */
3582
	wait_for_completion(&rsp->barrier_completion);
3583

3584 3585 3586 3587
	/* Mark the end of the barrier operation. */
	_rcu_barrier_trace(rsp, "Inc2", -1, rsp->barrier_sequence);
	rcu_seq_end(&rsp->barrier_sequence);

3588
	/* Other rcu_barrier() invocations can now safely proceed. */
3589
	mutex_unlock(&rsp->barrier_mutex);
3590 3591 3592 3593 3594 3595 3596
}

/**
 * rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.
 */
void rcu_barrier_bh(void)
{
3597
	_rcu_barrier(&rcu_bh_state);
3598 3599 3600 3601 3602 3603 3604 3605
}
EXPORT_SYMBOL_GPL(rcu_barrier_bh);

/**
 * rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.
 */
void rcu_barrier_sched(void)
{
3606
	_rcu_barrier(&rcu_sched_state);
3607 3608 3609
}
EXPORT_SYMBOL_GPL(rcu_barrier_sched);

3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620
/*
 * Propagate ->qsinitmask bits up the rcu_node tree to account for the
 * first CPU in a given leaf rcu_node structure coming online.  The caller
 * must hold the corresponding leaf rcu_node ->lock with interrrupts
 * disabled.
 */
static void rcu_init_new_rnp(struct rcu_node *rnp_leaf)
{
	long mask;
	struct rcu_node *rnp = rnp_leaf;

3621
	lockdep_assert_held(&rnp->lock);
3622 3623 3624 3625 3626
	for (;;) {
		mask = rnp->grpmask;
		rnp = rnp->parent;
		if (rnp == NULL)
			return;
3627
		raw_spin_lock_rcu_node(rnp); /* Interrupts already disabled. */
3628
		rnp->qsmaskinit |= mask;
B
Boqun Feng 已提交
3629
		raw_spin_unlock_rcu_node(rnp); /* Interrupts remain disabled. */
3630 3631 3632
	}
}

3633
/*
3634
 * Do boot-time initialization of a CPU's per-CPU RCU data.
3635
 */
3636 3637
static void __init
rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)
3638 3639
{
	unsigned long flags;
3640
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
3641 3642 3643
	struct rcu_node *rnp = rcu_get_root(rsp);

	/* Set up local state, ensuring consistent view of global state. */
3644
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
3645
	rdp->grpmask = leaf_node_cpu_bit(rdp->mynode, cpu);
3646
	rdp->dynticks = &per_cpu(rcu_dynticks, cpu);
3647
	WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE);
3648
	WARN_ON_ONCE(rcu_dynticks_in_eqs(rcu_dynticks_snap(rdp->dynticks)));
3649
	rdp->cpu = cpu;
3650
	rdp->rsp = rsp;
P
Paul E. McKenney 已提交
3651
	rcu_boot_init_nocb_percpu_data(rdp);
B
Boqun Feng 已提交
3652
	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
3653 3654 3655 3656 3657 3658 3659
}

/*
 * 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.
3660
 */
3661
static void
3662
rcu_init_percpu_data(int cpu, struct rcu_state *rsp)
3663 3664
{
	unsigned long flags;
3665
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
3666 3667 3668
	struct rcu_node *rnp = rcu_get_root(rsp);

	/* Set up local state, ensuring consistent view of global state. */
3669
	raw_spin_lock_irqsave_rcu_node(rnp, flags);
3670 3671
	rdp->qlen_last_fqs_check = 0;
	rdp->n_force_qs_snap = rsp->n_force_qs;
3672
	rdp->blimit = blimit;
3673 3674 3675
	if (rcu_segcblist_empty(&rdp->cblist) && /* No early-boot CBs? */
	    !init_nocb_callback_list(rdp))
		rcu_segcblist_init(&rdp->cblist);  /* Re-enable callbacks. */
3676
	rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
3677
	rcu_dynticks_eqs_online();
B
Boqun Feng 已提交
3678
	raw_spin_unlock_rcu_node(rnp);		/* irqs remain disabled. */
3679

3680 3681 3682 3683 3684
	/*
	 * Add CPU to leaf rcu_node pending-online bitmask.  Any needed
	 * propagation up the rcu_node tree will happen at the beginning
	 * of the next grace period.
	 */
3685
	rnp = rdp->mynode;
3686
	raw_spin_lock_rcu_node(rnp);		/* irqs already disabled. */
3687 3688 3689
	if (!rdp->beenonline)
		WRITE_ONCE(rsp->ncpus, READ_ONCE(rsp->ncpus) + 1);
	rdp->beenonline = true;	 /* We have now been online. */
3690 3691
	rdp->gpnum = rnp->completed; /* Make CPU later note any new GP. */
	rdp->completed = rnp->completed;
3692
	rdp->cpu_no_qs.b.norm = true;
3693
	rdp->rcu_qs_ctr_snap = per_cpu(rcu_dynticks.rcu_qs_ctr, cpu);
3694
	rdp->core_needs_qs = false;
3695
	trace_rcu_grace_period(rsp->name, rdp->gpnum, TPS("cpuonl"));
B
Boqun Feng 已提交
3696
	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
3697 3698
}

3699 3700 3701 3702
/*
 * Invoked early in the CPU-online process, when pretty much all
 * services are available.  The incoming CPU is not present.
 */
3703
int rcutree_prepare_cpu(unsigned int cpu)
3704
{
3705 3706 3707
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
3708
		rcu_init_percpu_data(cpu, rsp);
3709 3710 3711 3712 3713 3714 3715

	rcu_prepare_kthreads(cpu);
	rcu_spawn_all_nocb_kthreads(cpu);

	return 0;
}

3716 3717 3718
/*
 * Update RCU priority boot kthread affinity for CPU-hotplug changes.
 */
3719 3720 3721 3722 3723 3724 3725
static void rcutree_affinity_setting(unsigned int cpu, int outgoing)
{
	struct rcu_data *rdp = per_cpu_ptr(rcu_state_p->rda, cpu);

	rcu_boost_kthread_setaffinity(rdp->mynode, outgoing);
}

3726 3727 3728 3729
/*
 * Near the end of the CPU-online process.  Pretty much all services
 * enabled, and the CPU is now very much alive.
 */
3730 3731 3732 3733
int rcutree_online_cpu(unsigned int cpu)
{
	sync_sched_exp_online_cleanup(cpu);
	rcutree_affinity_setting(cpu, -1);
3734 3735
	if (IS_ENABLED(CONFIG_TREE_SRCU))
		srcu_online_cpu(cpu);
3736 3737 3738
	return 0;
}

3739 3740 3741 3742
/*
 * Near the beginning of the process.  The CPU is still very much alive
 * with pretty much all services enabled.
 */
3743 3744 3745
int rcutree_offline_cpu(unsigned int cpu)
{
	rcutree_affinity_setting(cpu, cpu);
3746 3747
	if (IS_ENABLED(CONFIG_TREE_SRCU))
		srcu_offline_cpu(cpu);
3748 3749 3750
	return 0;
}

3751 3752 3753
/*
 * Near the end of the offline process.  We do only tracing here.
 */
3754 3755 3756 3757 3758 3759 3760 3761 3762
int rcutree_dying_cpu(unsigned int cpu)
{
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
		rcu_cleanup_dying_cpu(rsp);
	return 0;
}

3763 3764 3765
/*
 * The outgoing CPU is gone and we are running elsewhere.
 */
3766 3767 3768 3769 3770 3771 3772 3773 3774
int rcutree_dead_cpu(unsigned int cpu)
{
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp) {
		rcu_cleanup_dead_cpu(cpu, rsp);
		do_nocb_deferred_wakeup(per_cpu_ptr(rsp->rda, cpu));
	}
	return 0;
3775 3776
}

3777 3778 3779 3780 3781 3782
/*
 * Mark the specified CPU as being online so that subsequent grace periods
 * (both expedited and normal) will wait on it.  Note that this means that
 * incoming CPUs are not allowed to use RCU read-side critical sections
 * until this function is called.  Failing to observe this restriction
 * will result in lockdep splats.
3783 3784 3785 3786
 *
 * Note that this function is special in that it is invoked directly
 * from the incoming CPU rather than from the cpuhp_step mechanism.
 * This is because this function must be invoked at a precise location.
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796
 */
void rcu_cpu_starting(unsigned int cpu)
{
	unsigned long flags;
	unsigned long mask;
	struct rcu_data *rdp;
	struct rcu_node *rnp;
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp) {
3797
		rdp = per_cpu_ptr(rsp->rda, cpu);
3798 3799 3800 3801 3802 3803 3804 3805 3806
		rnp = rdp->mynode;
		mask = rdp->grpmask;
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
		rnp->qsmaskinitnext |= mask;
		rnp->expmaskinitnext |= mask;
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
	}
}

3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823
#ifdef CONFIG_HOTPLUG_CPU
/*
 * The CPU is exiting the idle loop into the arch_cpu_idle_dead()
 * function.  We now remove it from the rcu_node tree's ->qsmaskinit
 * bit masks.
 */
static void rcu_cleanup_dying_idle_cpu(int cpu, struct rcu_state *rsp)
{
	unsigned long flags;
	unsigned long mask;
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
	struct rcu_node *rnp = rdp->mynode;  /* Outgoing CPU's rdp & rnp. */

	/* Remove outgoing CPU from mask in the leaf rcu_node structure. */
	mask = rdp->grpmask;
	raw_spin_lock_irqsave_rcu_node(rnp, flags); /* Enforce GP memory-order guarantee. */
	rnp->qsmaskinitnext &= ~mask;
3824
	raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
3825 3826
}

3827 3828 3829 3830 3831 3832 3833 3834
/*
 * The outgoing function has no further need of RCU, so remove it from
 * the list of CPUs that RCU must track.
 *
 * Note that this function is special in that it is invoked directly
 * from the outgoing CPU rather than from the cpuhp_step mechanism.
 * This is because this function must be invoked at a precise location.
 */
3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846
void rcu_report_dead(unsigned int cpu)
{
	struct rcu_state *rsp;

	/* QS for any half-done expedited RCU-sched GP. */
	preempt_disable();
	rcu_report_exp_rdp(&rcu_sched_state,
			   this_cpu_ptr(rcu_sched_state.rda), true);
	preempt_enable();
	for_each_rcu_flavor(rsp)
		rcu_cleanup_dying_idle_cpu(cpu, rsp);
}
3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956

/*
 * Send the specified CPU's RCU callbacks to the orphanage.  The
 * specified CPU must be offline, and the caller must hold the
 * ->orphan_lock.
 */
static void
rcu_send_cbs_to_orphanage(int cpu, struct rcu_state *rsp,
			  struct rcu_node *rnp, struct rcu_data *rdp)
{
	lockdep_assert_held(&rsp->orphan_lock);

	/* No-CBs CPUs do not have orphanable callbacks. */
	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU) || rcu_is_nocb_cpu(rdp->cpu))
		return;

	/*
	 * Orphan the callbacks.  First adjust the counts.  This is safe
	 * because _rcu_barrier() excludes CPU-hotplug operations, so it
	 * cannot be running now.  Thus no memory barrier is required.
	 */
	rdp->n_cbs_orphaned += rcu_segcblist_n_cbs(&rdp->cblist);
	rcu_segcblist_extract_count(&rdp->cblist, &rsp->orphan_done);

	/*
	 * 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.
	 */
	rcu_segcblist_extract_pend_cbs(&rdp->cblist, &rsp->orphan_pend);

	/*
	 * 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.
	 */
	rcu_segcblist_extract_done_cbs(&rdp->cblist, &rsp->orphan_done);

	/* Finally, disallow further callbacks on this CPU.  */
	rcu_segcblist_disable(&rdp->cblist);
}

/*
 * Adopt the RCU callbacks from the specified rcu_state structure's
 * orphanage.  The caller must hold the ->orphan_lock.
 */
static void rcu_adopt_orphan_cbs(struct rcu_state *rsp, unsigned long flags)
{
	struct rcu_data *rdp = raw_cpu_ptr(rsp->rda);

	lockdep_assert_held(&rsp->orphan_lock);

	/* No-CBs CPUs are handled specially. */
	if (!IS_ENABLED(CONFIG_HOTPLUG_CPU) ||
	    rcu_nocb_adopt_orphan_cbs(rsp, rdp, flags))
		return;

	/* Do the accounting first. */
	rdp->n_cbs_adopted += rsp->orphan_done.len;
	if (rsp->orphan_done.len_lazy != rsp->orphan_done.len)
		rcu_idle_count_callbacks_posted();
	rcu_segcblist_insert_count(&rdp->cblist, &rsp->orphan_done);

	/*
	 * 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, then the done ones. */
	rcu_segcblist_insert_done_cbs(&rdp->cblist, &rsp->orphan_done);
	WARN_ON_ONCE(rsp->orphan_done.head);
	rcu_segcblist_insert_pend_cbs(&rdp->cblist, &rsp->orphan_pend);
	WARN_ON_ONCE(rsp->orphan_pend.head);
	WARN_ON_ONCE(rcu_segcblist_empty(&rdp->cblist) !=
		     !rcu_segcblist_n_cbs(&rdp->cblist));
}

/* Orphan the dead CPU's callbacks, and then adopt them. */
static void rcu_migrate_callbacks(int cpu, struct rcu_state *rsp)
{
	unsigned long flags;
	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
	struct rcu_node *rnp = rdp->mynode;  /* Outgoing CPU's rdp & rnp. */

	raw_spin_lock_irqsave(&rsp->orphan_lock, flags);
	rcu_send_cbs_to_orphanage(cpu, rsp, rnp, rdp);
	rcu_adopt_orphan_cbs(rsp, flags);
	raw_spin_unlock_irqrestore(&rsp->orphan_lock, flags);
	WARN_ONCE(rcu_segcblist_n_cbs(&rdp->cblist) != 0 ||
		  !rcu_segcblist_empty(&rdp->cblist),
		  "rcu_cleanup_dead_cpu: Callbacks on offline CPU %d: qlen=%lu, 1stCB=%p\n",
		  cpu, rcu_segcblist_n_cbs(&rdp->cblist),
		  rcu_segcblist_first_cb(&rdp->cblist));
}

/*
 * The outgoing CPU has just passed through the dying-idle state,
 * and we are being invoked from the CPU that was IPIed to continue the
 * offline operation.  We need to migrate the outgoing CPU's callbacks.
 */
void rcutree_migrate_callbacks(int cpu)
{
	struct rcu_state *rsp;

	for_each_rcu_flavor(rsp)
		rcu_migrate_callbacks(cpu, rsp);
}
3957 3958
#endif

3959 3960 3961 3962
/*
 * On non-huge systems, use expedited RCU grace periods to make suspend
 * and hibernation run faster.
 */
3963 3964 3965 3966 3967 3968 3969
static int rcu_pm_notify(struct notifier_block *self,
			 unsigned long action, void *hcpu)
{
	switch (action) {
	case PM_HIBERNATION_PREPARE:
	case PM_SUSPEND_PREPARE:
		if (nr_cpu_ids <= 256) /* Expediting bad for large systems. */
3970
			rcu_expedite_gp();
3971 3972 3973
		break;
	case PM_POST_HIBERNATION:
	case PM_POST_SUSPEND:
3974 3975
		if (nr_cpu_ids <= 256) /* Expediting bad for large systems. */
			rcu_unexpedite_gp();
3976 3977 3978 3979 3980 3981 3982
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}

3983
/*
3984
 * Spawn the kthreads that handle each RCU flavor's grace periods.
3985 3986 3987 3988
 */
static int __init rcu_spawn_gp_kthread(void)
{
	unsigned long flags;
3989
	int kthread_prio_in = kthread_prio;
3990 3991
	struct rcu_node *rnp;
	struct rcu_state *rsp;
3992
	struct sched_param sp;
3993 3994
	struct task_struct *t;

3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005
	/* Force priority into range. */
	if (IS_ENABLED(CONFIG_RCU_BOOST) && kthread_prio < 1)
		kthread_prio = 1;
	else if (kthread_prio < 0)
		kthread_prio = 0;
	else if (kthread_prio > 99)
		kthread_prio = 99;
	if (kthread_prio != kthread_prio_in)
		pr_alert("rcu_spawn_gp_kthread(): Limited prio to %d from %d\n",
			 kthread_prio, kthread_prio_in);

4006
	rcu_scheduler_fully_active = 1;
4007
	for_each_rcu_flavor(rsp) {
4008
		t = kthread_create(rcu_gp_kthread, rsp, "%s", rsp->name);
4009 4010
		BUG_ON(IS_ERR(t));
		rnp = rcu_get_root(rsp);
4011
		raw_spin_lock_irqsave_rcu_node(rnp, flags);
4012
		rsp->gp_kthread = t;
4013 4014 4015 4016
		if (kthread_prio) {
			sp.sched_priority = kthread_prio;
			sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
		}
B
Boqun Feng 已提交
4017
		raw_spin_unlock_irqrestore_rcu_node(rnp, flags);
4018
		wake_up_process(t);
4019
	}
4020
	rcu_spawn_nocb_kthreads();
4021
	rcu_spawn_boost_kthreads();
4022 4023 4024 4025
	return 0;
}
early_initcall(rcu_spawn_gp_kthread);

4026
/*
4027 4028 4029 4030 4031 4032
 * This function is invoked towards the end of the scheduler's
 * initialization process.  Before this is called, the idle task might
 * contain synchronous grace-period primitives (during which time, this idle
 * task is booting the system, and such primitives are no-ops).  After this
 * function is called, any synchronous grace-period primitives are run as
 * expedited, with the requesting task driving the grace period forward.
4033
 * A later core_initcall() rcu_set_runtime_mode() will switch to full
4034
 * runtime RCU functionality.
4035 4036 4037 4038 4039
 */
void rcu_scheduler_starting(void)
{
	WARN_ON(num_online_cpus() != 1);
	WARN_ON(nr_context_switches() > 0);
4040 4041 4042
	rcu_test_sync_prims();
	rcu_scheduler_active = RCU_SCHEDULER_INIT;
	rcu_test_sync_prims();
4043 4044
}

4045 4046 4047
/*
 * Helper function for rcu_init() that initializes one rcu_state structure.
 */
4048
static void __init rcu_init_one(struct rcu_state *rsp)
4049
{
4050 4051
	static const char * const buf[] = RCU_NODE_NAME_INIT;
	static const char * const fqs[] = RCU_FQS_NAME_INIT;
4052 4053
	static struct lock_class_key rcu_node_class[RCU_NUM_LVLS];
	static struct lock_class_key rcu_fqs_class[RCU_NUM_LVLS];
4054 4055

	int levelspread[RCU_NUM_LVLS];		/* kids/node in each level. */
4056 4057 4058 4059 4060
	int cpustride = 1;
	int i;
	int j;
	struct rcu_node *rnp;

4061
	BUILD_BUG_ON(RCU_NUM_LVLS > ARRAY_SIZE(buf));  /* Fix buf[] init! */
4062

4063 4064 4065
	/* Silence gcc 4.8 false positive about array index out of range. */
	if (rcu_num_lvls <= 0 || rcu_num_lvls > RCU_NUM_LVLS)
		panic("rcu_init_one: rcu_num_lvls out of range");
4066

4067 4068
	/* Initialize the level-tracking arrays. */

4069
	for (i = 1; i < rcu_num_lvls; i++)
4070 4071
		rsp->level[i] = rsp->level[i - 1] + num_rcu_lvl[i - 1];
	rcu_init_levelspread(levelspread, num_rcu_lvl);
4072 4073 4074

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

4075
	for (i = rcu_num_lvls - 1; i >= 0; i--) {
4076
		cpustride *= levelspread[i];
4077
		rnp = rsp->level[i];
4078
		for (j = 0; j < num_rcu_lvl[i]; j++, rnp++) {
B
Boqun Feng 已提交
4079 4080
			raw_spin_lock_init(&ACCESS_PRIVATE(rnp, lock));
			lockdep_set_class_and_name(&ACCESS_PRIVATE(rnp, lock),
4081
						   &rcu_node_class[i], buf[i]);
4082 4083 4084
			raw_spin_lock_init(&rnp->fqslock);
			lockdep_set_class_and_name(&rnp->fqslock,
						   &rcu_fqs_class[i], fqs[i]);
4085 4086
			rnp->gpnum = rsp->gpnum;
			rnp->completed = rsp->completed;
4087 4088 4089 4090
			rnp->qsmask = 0;
			rnp->qsmaskinit = 0;
			rnp->grplo = j * cpustride;
			rnp->grphi = (j + 1) * cpustride - 1;
4091 4092
			if (rnp->grphi >= nr_cpu_ids)
				rnp->grphi = nr_cpu_ids - 1;
4093 4094 4095 4096 4097
			if (i == 0) {
				rnp->grpnum = 0;
				rnp->grpmask = 0;
				rnp->parent = NULL;
			} else {
4098
				rnp->grpnum = j % levelspread[i - 1];
4099 4100
				rnp->grpmask = 1UL << rnp->grpnum;
				rnp->parent = rsp->level[i - 1] +
4101
					      j / levelspread[i - 1];
4102 4103
			}
			rnp->level = i;
4104
			INIT_LIST_HEAD(&rnp->blkd_tasks);
4105
			rcu_init_one_nocb(rnp);
4106 4107
			init_waitqueue_head(&rnp->exp_wq[0]);
			init_waitqueue_head(&rnp->exp_wq[1]);
4108 4109
			init_waitqueue_head(&rnp->exp_wq[2]);
			init_waitqueue_head(&rnp->exp_wq[3]);
4110
			spin_lock_init(&rnp->exp_lock);
4111 4112
		}
	}
4113

4114 4115
	init_swait_queue_head(&rsp->gp_wq);
	init_swait_queue_head(&rsp->expedited_wq);
4116
	rnp = rsp->level[rcu_num_lvls - 1];
4117
	for_each_possible_cpu(i) {
4118
		while (i > rnp->grphi)
4119
			rnp++;
4120
		per_cpu_ptr(rsp->rda, i)->mynode = rnp;
4121 4122
		rcu_boot_init_percpu_data(i, rsp);
	}
4123
	list_add(&rsp->flavors, &rcu_struct_flavors);
4124 4125
}

4126 4127
/*
 * Compute the rcu_node tree geometry from kernel parameters.  This cannot
4128
 * replace the definitions in tree.h because those are needed to size
4129 4130 4131 4132
 * the ->node array in the rcu_state structure.
 */
static void __init rcu_init_geometry(void)
{
4133
	ulong d;
4134
	int i;
4135
	int rcu_capacity[RCU_NUM_LVLS];
4136

4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149
	/*
	 * Initialize any unspecified boot parameters.
	 * The default values of jiffies_till_first_fqs and
	 * jiffies_till_next_fqs are set to the RCU_JIFFIES_TILL_FORCE_QS
	 * value, which is a function of HZ, then adding one for each
	 * RCU_JIFFIES_FQS_DIV CPUs that might be on the system.
	 */
	d = RCU_JIFFIES_TILL_FORCE_QS + nr_cpu_ids / RCU_JIFFIES_FQS_DIV;
	if (jiffies_till_first_fqs == ULONG_MAX)
		jiffies_till_first_fqs = d;
	if (jiffies_till_next_fqs == ULONG_MAX)
		jiffies_till_next_fqs = d;

4150
	/* If the compile-time values are accurate, just leave. */
4151
	if (rcu_fanout_leaf == RCU_FANOUT_LEAF &&
4152
	    nr_cpu_ids == NR_CPUS)
4153
		return;
4154 4155
	pr_info("RCU: Adjusting geometry for rcu_fanout_leaf=%d, nr_cpu_ids=%d\n",
		rcu_fanout_leaf, nr_cpu_ids);
4156 4157

	/*
4158 4159 4160 4161
	 * The boot-time rcu_fanout_leaf parameter must be at least two
	 * and cannot exceed the number of bits in the rcu_node masks.
	 * Complain and fall back to the compile-time values if this
	 * limit is exceeded.
4162
	 */
4163
	if (rcu_fanout_leaf < 2 ||
4164
	    rcu_fanout_leaf > sizeof(unsigned long) * 8) {
4165
		rcu_fanout_leaf = RCU_FANOUT_LEAF;
4166 4167 4168 4169 4170 4171
		WARN_ON(1);
		return;
	}

	/*
	 * Compute number of nodes that can be handled an rcu_node tree
4172
	 * with the given number of levels.
4173
	 */
4174
	rcu_capacity[0] = rcu_fanout_leaf;
4175
	for (i = 1; i < RCU_NUM_LVLS; i++)
4176
		rcu_capacity[i] = rcu_capacity[i - 1] * RCU_FANOUT;
4177 4178

	/*
4179
	 * The tree must be able to accommodate the configured number of CPUs.
4180
	 * If this limit is exceeded, fall back to the compile-time values.
4181
	 */
4182 4183 4184 4185 4186
	if (nr_cpu_ids > rcu_capacity[RCU_NUM_LVLS - 1]) {
		rcu_fanout_leaf = RCU_FANOUT_LEAF;
		WARN_ON(1);
		return;
	}
4187

4188
	/* Calculate the number of levels in the tree. */
4189
	for (i = 0; nr_cpu_ids > rcu_capacity[i]; i++) {
4190
	}
4191
	rcu_num_lvls = i + 1;
4192

4193
	/* Calculate the number of rcu_nodes at each level of the tree. */
4194
	for (i = 0; i < rcu_num_lvls; i++) {
4195
		int cap = rcu_capacity[(rcu_num_lvls - 1) - i];
4196 4197
		num_rcu_lvl[i] = DIV_ROUND_UP(nr_cpu_ids, cap);
	}
4198 4199 4200

	/* Calculate the total number of rcu_node structures. */
	rcu_num_nodes = 0;
4201
	for (i = 0; i < rcu_num_lvls; i++)
4202 4203 4204
		rcu_num_nodes += num_rcu_lvl[i];
}

4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226
/*
 * Dump out the structure of the rcu_node combining tree associated
 * with the rcu_state structure referenced by rsp.
 */
static void __init rcu_dump_rcu_node_tree(struct rcu_state *rsp)
{
	int level = 0;
	struct rcu_node *rnp;

	pr_info("rcu_node tree layout dump\n");
	pr_info(" ");
	rcu_for_each_node_breadth_first(rsp, rnp) {
		if (rnp->level != level) {
			pr_cont("\n");
			pr_info(" ");
			level = rnp->level;
		}
		pr_cont("%d:%d ^%d  ", rnp->grplo, rnp->grphi, rnp->grpnum);
	}
	pr_cont("\n");
}

4227
void __init rcu_init(void)
4228
{
P
Paul E. McKenney 已提交
4229
	int cpu;
4230

4231 4232
	rcu_early_boot_tests();

4233
	rcu_bootup_announce();
4234
	rcu_init_geometry();
4235 4236
	rcu_init_one(&rcu_bh_state);
	rcu_init_one(&rcu_sched_state);
4237 4238
	if (dump_tree)
		rcu_dump_rcu_node_tree(&rcu_sched_state);
4239
	__rcu_init_preempt();
J
Jiang Fang 已提交
4240
	open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
4241 4242 4243 4244 4245 4246

	/*
	 * We don't need protection against CPU-hotplug here because
	 * this is called early in boot, before either interrupts
	 * or the scheduler are operational.
	 */
4247
	pm_notifier(rcu_pm_notify, 0);
4248
	for_each_online_cpu(cpu) {
4249
		rcutree_prepare_cpu(cpu);
4250
		rcu_cpu_starting(cpu);
4251 4252
		if (IS_ENABLED(CONFIG_TREE_SRCU))
			srcu_online_cpu(cpu);
4253
	}
4254 4255
}

4256
#include "tree_exp.h"
4257
#include "tree_plugin.h"