Linux 3.13-rc2
[linux-2.6-block.git] / kernel / rcu / tree.h
CommitLineData
9f77da9f
PM
1/*
2 * Read-Copy Update mechanism for mutual exclusion (tree-based version)
3 * Internal non-public definitions.
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 *
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
18 *
19 * Copyright IBM Corporation, 2008
20 *
21 * Author: Ingo Molnar <mingo@elte.hu>
22 * Paul E. McKenney <paulmck@linux.vnet.ibm.com>
23 */
24
25#include <linux/cache.h>
26#include <linux/spinlock.h>
27#include <linux/threads.h>
28#include <linux/cpumask.h>
29#include <linux/seqlock.h>
016a8d5b 30#include <linux/irq_work.h>
9f77da9f
PM
31
32/*
8932a63d
PM
33 * Define shape of hierarchy based on NR_CPUS, CONFIG_RCU_FANOUT, and
34 * CONFIG_RCU_FANOUT_LEAF.
9f77da9f 35 * In theory, it should be possible to add more levels straightforwardly.
0209f649
PM
36 * In practice, this did work well going from three levels to four.
37 * Of course, your mileage may vary.
9f77da9f 38 */
cf244dc0 39#define MAX_RCU_LVLS 4
8932a63d 40#define RCU_FANOUT_1 (CONFIG_RCU_FANOUT_LEAF)
0209f649
PM
41#define RCU_FANOUT_2 (RCU_FANOUT_1 * CONFIG_RCU_FANOUT)
42#define RCU_FANOUT_3 (RCU_FANOUT_2 * CONFIG_RCU_FANOUT)
43#define RCU_FANOUT_4 (RCU_FANOUT_3 * CONFIG_RCU_FANOUT)
9f77da9f 44
0209f649 45#if NR_CPUS <= RCU_FANOUT_1
f885b7f2 46# define RCU_NUM_LVLS 1
9f77da9f
PM
47# define NUM_RCU_LVL_0 1
48# define NUM_RCU_LVL_1 (NR_CPUS)
49# define NUM_RCU_LVL_2 0
50# define NUM_RCU_LVL_3 0
cf244dc0 51# define NUM_RCU_LVL_4 0
0209f649 52#elif NR_CPUS <= RCU_FANOUT_2
f885b7f2 53# define RCU_NUM_LVLS 2
9f77da9f 54# define NUM_RCU_LVL_0 1
0209f649 55# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
9f77da9f
PM
56# define NUM_RCU_LVL_2 (NR_CPUS)
57# define NUM_RCU_LVL_3 0
cf244dc0 58# define NUM_RCU_LVL_4 0
0209f649 59#elif NR_CPUS <= RCU_FANOUT_3
f885b7f2 60# define RCU_NUM_LVLS 3
9f77da9f 61# define NUM_RCU_LVL_0 1
0209f649
PM
62# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2)
63# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
64# define NUM_RCU_LVL_3 (NR_CPUS)
cf244dc0 65# define NUM_RCU_LVL_4 0
0209f649 66#elif NR_CPUS <= RCU_FANOUT_4
f885b7f2 67# define RCU_NUM_LVLS 4
cf244dc0 68# define NUM_RCU_LVL_0 1
0209f649
PM
69# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_3)
70# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2)
71# define NUM_RCU_LVL_3 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
72# define NUM_RCU_LVL_4 (NR_CPUS)
9f77da9f
PM
73#else
74# error "CONFIG_RCU_FANOUT insufficient for NR_CPUS"
0209f649 75#endif /* #if (NR_CPUS) <= RCU_FANOUT_1 */
9f77da9f 76
cf244dc0 77#define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3 + NUM_RCU_LVL_4)
9f77da9f
PM
78#define NUM_RCU_NODES (RCU_SUM - NR_CPUS)
79
f885b7f2
PM
80extern int rcu_num_lvls;
81extern int rcu_num_nodes;
82
9f77da9f
PM
83/*
84 * Dynticks per-CPU state.
85 */
86struct rcu_dynticks {
9b2e4f18
PM
87 long long dynticks_nesting; /* Track irq/process nesting level. */
88 /* Process level is worth LLONG_MAX/2. */
89 int dynticks_nmi_nesting; /* Track NMI nesting level. */
90 atomic_t dynticks; /* Even value for idle, else odd. */
2333210b
PM
91#ifdef CONFIG_NO_HZ_FULL_SYSIDLE
92 long long dynticks_idle_nesting;
93 /* irq/process nesting level from idle. */
94 atomic_t dynticks_idle; /* Even value for idle, else odd. */
95 /* "Idle" excludes userspace execution. */
96 unsigned long dynticks_idle_jiffies;
97 /* End of last non-NMI non-idle period. */
98#endif /* #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
5955f7ee 99#ifdef CONFIG_RCU_FAST_NO_HZ
c0f4dfd4 100 bool all_lazy; /* Are all CPU's CBs lazy? */
5955f7ee
PM
101 unsigned long nonlazy_posted;
102 /* # times non-lazy CBs posted to CPU. */
103 unsigned long nonlazy_posted_snap;
104 /* idle-period nonlazy_posted snapshot. */
c0f4dfd4
PM
105 unsigned long last_accelerate;
106 /* Last jiffy CBs were accelerated. */
c229828c
PM
107 unsigned long last_advance_all;
108 /* Last jiffy CBs were all advanced. */
9d2ad243 109 int tick_nohz_enabled_snap; /* Previously seen value from sysfs. */
5955f7ee 110#endif /* #ifdef CONFIG_RCU_FAST_NO_HZ */
9f77da9f
PM
111};
112
d71df90e
PM
113/* RCU's kthread states for tracing. */
114#define RCU_KTHREAD_STOPPED 0
115#define RCU_KTHREAD_RUNNING 1
116#define RCU_KTHREAD_WAITING 2
15ba0ba8
PM
117#define RCU_KTHREAD_OFFCPU 3
118#define RCU_KTHREAD_YIELDING 4
119#define RCU_KTHREAD_MAX 4
d71df90e 120
9f77da9f
PM
121/*
122 * Definition for node within the RCU grace-period-detection hierarchy.
123 */
124struct rcu_node {
1304afb2 125 raw_spinlock_t lock; /* Root rcu_node's lock protects some */
1eba8f84 126 /* rcu_state fields as well as following. */
20133cfc 127 unsigned long gpnum; /* Current grace period for this node. */
86848966
PM
128 /* This will either be equal to or one */
129 /* behind the root rcu_node's gpnum. */
20133cfc 130 unsigned long completed; /* Last GP completed for this node. */
d09b62df
PM
131 /* This will either be equal to or one */
132 /* behind the root rcu_node's gpnum. */
9f77da9f
PM
133 unsigned long qsmask; /* CPUs or groups that need to switch in */
134 /* order for current grace period to proceed.*/
1eba8f84
PM
135 /* In leaf rcu_node, each bit corresponds to */
136 /* an rcu_data structure, otherwise, each */
137 /* bit corresponds to a child rcu_node */
138 /* structure. */
12f5f524 139 unsigned long expmask; /* Groups that have ->blkd_tasks */
d9a3da06
PM
140 /* elements that need to drain to allow the */
141 /* current expedited grace period to */
142 /* complete (only for TREE_PREEMPT_RCU). */
9f77da9f 143 unsigned long qsmaskinit;
d9a3da06 144 /* Per-GP initial value for qsmask & expmask. */
9f77da9f 145 unsigned long grpmask; /* Mask to apply to parent qsmask. */
1eba8f84 146 /* Only one bit will be set in this mask. */
9f77da9f
PM
147 int grplo; /* lowest-numbered CPU or group here. */
148 int grphi; /* highest-numbered CPU or group here. */
149 u8 grpnum; /* CPU/group number for next level up. */
150 u8 level; /* root is at level 0. */
151 struct rcu_node *parent;
12f5f524
PM
152 struct list_head blkd_tasks;
153 /* Tasks blocked in RCU read-side critical */
154 /* section. Tasks are placed at the head */
155 /* of this list and age towards the tail. */
156 struct list_head *gp_tasks;
157 /* Pointer to the first task blocking the */
158 /* current grace period, or NULL if there */
159 /* is no such task. */
160 struct list_head *exp_tasks;
161 /* Pointer to the first task blocking the */
162 /* current expedited grace period, or NULL */
163 /* if there is no such task. If there */
164 /* is no current expedited grace period, */
165 /* then there can cannot be any such task. */
27f4d280
PM
166#ifdef CONFIG_RCU_BOOST
167 struct list_head *boost_tasks;
168 /* Pointer to first task that needs to be */
169 /* priority boosted, or NULL if no priority */
170 /* boosting is needed for this rcu_node */
171 /* structure. If there are no tasks */
172 /* queued on this rcu_node structure that */
173 /* are blocking the current grace period, */
174 /* there can be no such task. */
175 unsigned long boost_time;
176 /* When to start boosting (jiffies). */
177 struct task_struct *boost_kthread_task;
178 /* kthread that takes care of priority */
179 /* boosting for this rcu_node structure. */
d71df90e
PM
180 unsigned int boost_kthread_status;
181 /* State of boost_kthread_task for tracing. */
0ea1f2eb
PM
182 unsigned long n_tasks_boosted;
183 /* Total number of tasks boosted. */
184 unsigned long n_exp_boosts;
185 /* Number of tasks boosted for expedited GP. */
186 unsigned long n_normal_boosts;
187 /* Number of tasks boosted for normal GP. */
188 unsigned long n_balk_blkd_tasks;
189 /* Refused to boost: no blocked tasks. */
190 unsigned long n_balk_exp_gp_tasks;
191 /* Refused to boost: nothing blocking GP. */
192 unsigned long n_balk_boost_tasks;
193 /* Refused to boost: already boosting. */
194 unsigned long n_balk_notblocked;
195 /* Refused to boost: RCU RS CS still running. */
196 unsigned long n_balk_notyet;
197 /* Refused to boost: not yet time. */
198 unsigned long n_balk_nos;
199 /* Refused to boost: not sure why, though. */
200 /* This can happen due to race conditions. */
27f4d280 201#endif /* #ifdef CONFIG_RCU_BOOST */
dae6e64d
PM
202#ifdef CONFIG_RCU_NOCB_CPU
203 wait_queue_head_t nocb_gp_wq[2];
204 /* Place for rcu_nocb_kthread() to wait GP. */
dae6e64d 205#endif /* #ifdef CONFIG_RCU_NOCB_CPU */
8b425aa8
PM
206 int need_future_gp[2];
207 /* Counts of upcoming no-CB GP requests. */
394f2769 208 raw_spinlock_t fqslock ____cacheline_internodealigned_in_smp;
9f77da9f
PM
209} ____cacheline_internodealigned_in_smp;
210
a0b6c9a7
PM
211/*
212 * Do a full breadth-first scan of the rcu_node structures for the
213 * specified rcu_state structure.
214 */
215#define rcu_for_each_node_breadth_first(rsp, rnp) \
216 for ((rnp) = &(rsp)->node[0]; \
f885b7f2 217 (rnp) < &(rsp)->node[rcu_num_nodes]; (rnp)++)
a0b6c9a7 218
d9a3da06
PM
219/*
220 * Do a breadth-first scan of the non-leaf rcu_node structures for the
221 * specified rcu_state structure. Note that if there is a singleton
222 * rcu_node tree with but one rcu_node structure, this loop is a no-op.
223 */
224#define rcu_for_each_nonleaf_node_breadth_first(rsp, rnp) \
225 for ((rnp) = &(rsp)->node[0]; \
f885b7f2 226 (rnp) < (rsp)->level[rcu_num_lvls - 1]; (rnp)++)
d9a3da06
PM
227
228/*
229 * Scan the leaves of the rcu_node hierarchy for the specified rcu_state
230 * structure. Note that if there is a singleton rcu_node tree with but
231 * one rcu_node structure, this loop -will- visit the rcu_node structure.
232 * It is still a leaf node, even if it is also the root node.
233 */
a0b6c9a7 234#define rcu_for_each_leaf_node(rsp, rnp) \
f885b7f2
PM
235 for ((rnp) = (rsp)->level[rcu_num_lvls - 1]; \
236 (rnp) < &(rsp)->node[rcu_num_nodes]; (rnp)++)
a0b6c9a7 237
9f77da9f
PM
238/* Index values for nxttail array in struct rcu_data. */
239#define RCU_DONE_TAIL 0 /* Also RCU_WAIT head. */
240#define RCU_WAIT_TAIL 1 /* Also RCU_NEXT_READY head. */
241#define RCU_NEXT_READY_TAIL 2 /* Also RCU_NEXT head. */
242#define RCU_NEXT_TAIL 3
243#define RCU_NEXT_SIZE 4
244
245/* Per-CPU data for read-copy update. */
246struct rcu_data {
247 /* 1) quiescent-state and grace-period handling : */
20133cfc 248 unsigned long completed; /* Track rsp->completed gp number */
9f77da9f 249 /* in order to detect GP end. */
20133cfc 250 unsigned long gpnum; /* Highest gp number that this CPU */
9f77da9f 251 /* is aware of having started. */
e4cc1f22 252 bool passed_quiesce; /* User-mode/idle loop etc. */
9f77da9f
PM
253 bool qs_pending; /* Core waits for quiesc state. */
254 bool beenonline; /* CPU online at least once. */
6cc68793 255 bool preemptible; /* Preemptible RCU? */
9f77da9f
PM
256 struct rcu_node *mynode; /* This CPU's leaf of hierarchy */
257 unsigned long grpmask; /* Mask to apply to leaf qsmask. */
a858af28
PM
258#ifdef CONFIG_RCU_CPU_STALL_INFO
259 unsigned long ticks_this_gp; /* The number of scheduling-clock */
260 /* ticks this CPU has handled */
261 /* during and after the last grace */
262 /* period it is aware of. */
263#endif /* #ifdef CONFIG_RCU_CPU_STALL_INFO */
9f77da9f
PM
264
265 /* 2) batch handling */
266 /*
267 * If nxtlist is not NULL, it is partitioned as follows.
268 * Any of the partitions might be empty, in which case the
269 * pointer to that partition will be equal to the pointer for
270 * the following partition. When the list is empty, all of
1eba8f84
PM
271 * the nxttail elements point to the ->nxtlist pointer itself,
272 * which in that case is NULL.
9f77da9f 273 *
9f77da9f
PM
274 * [nxtlist, *nxttail[RCU_DONE_TAIL]):
275 * Entries that batch # <= ->completed
276 * The grace period for these entries has completed, and
277 * the other grace-period-completed entries may be moved
278 * here temporarily in rcu_process_callbacks().
1eba8f84
PM
279 * [*nxttail[RCU_DONE_TAIL], *nxttail[RCU_WAIT_TAIL]):
280 * Entries that batch # <= ->completed - 1: waiting for current GP
281 * [*nxttail[RCU_WAIT_TAIL], *nxttail[RCU_NEXT_READY_TAIL]):
282 * Entries known to have arrived before current GP ended
283 * [*nxttail[RCU_NEXT_READY_TAIL], *nxttail[RCU_NEXT_TAIL]):
284 * Entries that might have arrived after current GP ended
285 * Note that the value of *nxttail[RCU_NEXT_TAIL] will
286 * always be NULL, as this is the end of the list.
9f77da9f
PM
287 */
288 struct rcu_head *nxtlist;
289 struct rcu_head **nxttail[RCU_NEXT_SIZE];
dc35c893
PM
290 unsigned long nxtcompleted[RCU_NEXT_SIZE];
291 /* grace periods for sublists. */
486e2593
PM
292 long qlen_lazy; /* # of lazy queued callbacks */
293 long qlen; /* # of queued callbacks, incl lazy */
37c72e56
PM
294 long qlen_last_fqs_check;
295 /* qlen at last check for QS forcing */
269dcc1c 296 unsigned long n_cbs_invoked; /* count of RCU cbs invoked. */
c635a4e1 297 unsigned long n_nocbs_invoked; /* count of no-CBs RCU cbs invoked. */
29494be7
LJ
298 unsigned long n_cbs_orphaned; /* RCU cbs orphaned by dying CPU */
299 unsigned long n_cbs_adopted; /* RCU cbs adopted from dying CPU */
37c72e56
PM
300 unsigned long n_force_qs_snap;
301 /* did other CPU force QS recently? */
9f77da9f
PM
302 long blimit; /* Upper limit on a processed batch */
303
9f77da9f
PM
304 /* 3) dynticks interface. */
305 struct rcu_dynticks *dynticks; /* Shared per-CPU dynticks state. */
306 int dynticks_snap; /* Per-GP tracking for dynticks. */
9f77da9f
PM
307
308 /* 4) reasons this CPU needed to be kicked by force_quiescent_state */
9f77da9f 309 unsigned long dynticks_fqs; /* Kicked due to dynticks idle. */
9f77da9f 310 unsigned long offline_fqs; /* Kicked due to being offline. */
9f77da9f
PM
311
312 /* 5) __rcu_pending() statistics. */
20133cfc
PM
313 unsigned long n_rcu_pending; /* rcu_pending() calls since boot. */
314 unsigned long n_rp_qs_pending;
d21670ac 315 unsigned long n_rp_report_qs;
20133cfc
PM
316 unsigned long n_rp_cb_ready;
317 unsigned long n_rp_cpu_needs_gp;
318 unsigned long n_rp_gp_completed;
319 unsigned long n_rp_gp_started;
20133cfc 320 unsigned long n_rp_need_nothing;
9f77da9f 321
b626c1b6 322 /* 6) _rcu_barrier() and OOM callbacks. */
06668efa 323 struct rcu_head barrier_head;
b626c1b6
PM
324#ifdef CONFIG_RCU_FAST_NO_HZ
325 struct rcu_head oom_head;
326#endif /* #ifdef CONFIG_RCU_FAST_NO_HZ */
06668efa 327
3fbfbf7a
PM
328 /* 7) Callback offloading. */
329#ifdef CONFIG_RCU_NOCB_CPU
330 struct rcu_head *nocb_head; /* CBs waiting for kthread. */
331 struct rcu_head **nocb_tail;
332 atomic_long_t nocb_q_count; /* # CBs waiting for kthread */
333 atomic_long_t nocb_q_count_lazy; /* (approximate). */
334 int nocb_p_count; /* # CBs being invoked by kthread */
335 int nocb_p_count_lazy; /* (approximate). */
336 wait_queue_head_t nocb_wq; /* For nocb kthreads to sleep on. */
337 struct task_struct *nocb_kthread;
338#endif /* #ifdef CONFIG_RCU_NOCB_CPU */
339
6231069b
PM
340 /* 8) RCU CPU stall data. */
341#ifdef CONFIG_RCU_CPU_STALL_INFO
342 unsigned int softirq_snap; /* Snapshot of softirq activity. */
343#endif /* #ifdef CONFIG_RCU_CPU_STALL_INFO */
344
9f77da9f 345 int cpu;
d4c08f2a 346 struct rcu_state *rsp;
9f77da9f
PM
347};
348
af446b70 349/* Values for fqs_state field in struct rcu_state. */
83f5b01f
PM
350#define RCU_GP_IDLE 0 /* No grace period in progress. */
351#define RCU_GP_INIT 1 /* Grace period being initialized. */
352#define RCU_SAVE_DYNTICK 2 /* Need to scan dyntick state. */
ee47eb9f 353#define RCU_FORCE_QS 3 /* Need to force quiescent state. */
9f77da9f 354#define RCU_SIGNAL_INIT RCU_SAVE_DYNTICK
9f77da9f 355
026ad283
PM
356#define RCU_JIFFIES_TILL_FORCE_QS (1 + (HZ > 250) + (HZ > 500))
357 /* For jiffies_till_first_fqs and */
358 /* and jiffies_till_next_fqs. */
007b0924 359
026ad283
PM
360#define RCU_JIFFIES_FQS_DIV 256 /* Very large systems need more */
361 /* delay between bouts of */
362 /* quiescent-state forcing. */
363
364#define RCU_STALL_RAT_DELAY 2 /* Allow other CPUs time to take */
365 /* at least one scheduling clock */
366 /* irq before ratting on them. */
9f77da9f 367
08bca60a
PZ
368#define rcu_wait(cond) \
369do { \
370 for (;;) { \
371 set_current_state(TASK_INTERRUPTIBLE); \
372 if (cond) \
373 break; \
374 schedule(); \
375 } \
376 __set_current_state(TASK_RUNNING); \
377} while (0)
9f77da9f
PM
378
379/*
380 * RCU global state, including node hierarchy. This hierarchy is
381 * represented in "heap" form in a dense array. The root (first level)
382 * of the hierarchy is in ->node[0] (referenced by ->level[0]), the second
383 * level in ->node[1] through ->node[m] (->node[1] referenced by ->level[1]),
384 * and the third level in ->node[m+1] and following (->node[m+1] referenced
385 * by ->level[2]). The number of levels is determined by the number of
386 * CPUs and by CONFIG_RCU_FANOUT. Small systems will have a "hierarchy"
387 * consisting of a single rcu_node.
388 */
389struct rcu_state {
390 struct rcu_node node[NUM_RCU_NODES]; /* Hierarchy. */
f885b7f2 391 struct rcu_node *level[RCU_NUM_LVLS]; /* Hierarchy levels. */
9f77da9f 392 u32 levelcnt[MAX_RCU_LVLS + 1]; /* # nodes in each level. */
f885b7f2 393 u8 levelspread[RCU_NUM_LVLS]; /* kids/node in each level. */
394f99a9 394 struct rcu_data __percpu *rda; /* pointer of percu rcu_data. */
037b64ed
PM
395 void (*call)(struct rcu_head *head, /* call_rcu() flavor. */
396 void (*func)(struct rcu_head *head));
9f77da9f
PM
397
398 /* The following fields are guarded by the root rcu_node's lock. */
399
af446b70 400 u8 fqs_state ____cacheline_internodealigned_in_smp;
9f77da9f 401 /* Force QS state. */
a46e0899 402 u8 boost; /* Subject to priority boost. */
20133cfc
PM
403 unsigned long gpnum; /* Current gp number. */
404 unsigned long completed; /* # of last completed gp. */
b3dbec76
PM
405 struct task_struct *gp_kthread; /* Task for grace periods. */
406 wait_queue_head_t gp_wq; /* Where GP task waits. */
407 int gp_flags; /* Commands for GP task. */
1eba8f84 408
d9a3da06 409 /* End of fields guarded by root rcu_node's lock. */
1eba8f84 410
7b2e6011
PM
411 raw_spinlock_t orphan_lock ____cacheline_internodealigned_in_smp;
412 /* Protect following fields. */
b1420f1c
PM
413 struct rcu_head *orphan_nxtlist; /* Orphaned callbacks that */
414 /* need a grace period. */
415 struct rcu_head **orphan_nxttail; /* Tail of above. */
416 struct rcu_head *orphan_donelist; /* Orphaned callbacks that */
417 /* are ready to invoke. */
418 struct rcu_head **orphan_donetail; /* Tail of above. */
419 long qlen_lazy; /* Number of lazy callbacks. */
420 long qlen; /* Total number of callbacks. */
7b2e6011 421 /* End of fields guarded by orphan_lock. */
a4fbe35a
PM
422
423 struct mutex onoff_mutex; /* Coordinate hotplug & GPs. */
424
7be7f0be 425 struct mutex barrier_mutex; /* Guards barrier fields. */
24ebbca8 426 atomic_t barrier_cpu_count; /* # CPUs waiting on. */
7db74df8 427 struct completion barrier_completion; /* Wake at barrier end. */
cf3a9c48
PM
428 unsigned long n_barrier_done; /* ++ at start and end of */
429 /* _rcu_barrier(). */
a4fbe35a
PM
430 /* End of fields guarded by barrier_mutex. */
431
40694d66
PM
432 atomic_long_t expedited_start; /* Starting ticket. */
433 atomic_long_t expedited_done; /* Done ticket. */
a30489c5
PM
434 atomic_long_t expedited_wrap; /* # near-wrap incidents. */
435 atomic_long_t expedited_tryfail; /* # acquisition failures. */
436 atomic_long_t expedited_workdone1; /* # done by others #1. */
437 atomic_long_t expedited_workdone2; /* # done by others #2. */
438 atomic_long_t expedited_normal; /* # fallbacks to normal. */
439 atomic_long_t expedited_stoppedcpus; /* # successful stop_cpus. */
440 atomic_long_t expedited_done_tries; /* # tries to update _done. */
441 atomic_long_t expedited_done_lost; /* # times beaten to _done. */
442 atomic_long_t expedited_done_exit; /* # times exited _done loop. */
40694d66 443
9f77da9f
PM
444 unsigned long jiffies_force_qs; /* Time at which to invoke */
445 /* force_quiescent_state(). */
446 unsigned long n_force_qs; /* Number of calls to */
447 /* force_quiescent_state(). */
448 unsigned long n_force_qs_lh; /* ~Number of calls leaving */
449 /* due to lock unavailable. */
450 unsigned long n_force_qs_ngp; /* Number of calls leaving */
451 /* due to no GP active. */
9f77da9f
PM
452 unsigned long gp_start; /* Time at which GP started, */
453 /* but in jiffies. */
454 unsigned long jiffies_stall; /* Time at which to check */
455 /* for CPU stalls. */
15ba0ba8
PM
456 unsigned long gp_max; /* Maximum GP duration in */
457 /* jiffies. */
e66c33d5 458 const char *name; /* Name of structure. */
a4889858 459 char abbr; /* Abbreviated name. */
6ce75a23 460 struct list_head flavors; /* List of RCU flavors. */
016a8d5b 461 struct irq_work wakeup_work; /* Postponed wakeups */
9f77da9f
PM
462};
463
4cdfc175
PM
464/* Values for rcu_state structure's gp_flags field. */
465#define RCU_GP_FLAG_INIT 0x1 /* Need grace-period initialization. */
466#define RCU_GP_FLAG_FQS 0x2 /* Need grace-period quiescent-state forcing. */
467
6ce75a23 468extern struct list_head rcu_struct_flavors;
3fbfbf7a
PM
469
470/* Sequence through rcu_state structures for each RCU flavor. */
6ce75a23
PM
471#define for_each_rcu_flavor(rsp) \
472 list_for_each_entry((rsp), &rcu_struct_flavors, flavors)
473
d9a3da06
PM
474/* Return values for rcu_preempt_offline_tasks(). */
475
476#define RCU_OFL_TASKS_NORM_GP 0x1 /* Tasks blocking normal */
477 /* GP were moved to root. */
478#define RCU_OFL_TASKS_EXP_GP 0x2 /* Tasks blocking expedited */
479 /* GP were moved to root. */
480
6258c4fb
IM
481/*
482 * RCU implementation internal declarations:
483 */
d6714c22
PM
484extern struct rcu_state rcu_sched_state;
485DECLARE_PER_CPU(struct rcu_data, rcu_sched_data);
6258c4fb
IM
486
487extern struct rcu_state rcu_bh_state;
488DECLARE_PER_CPU(struct rcu_data, rcu_bh_data);
489
f41d911f
PM
490#ifdef CONFIG_TREE_PREEMPT_RCU
491extern struct rcu_state rcu_preempt_state;
492DECLARE_PER_CPU(struct rcu_data, rcu_preempt_data);
493#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
494
eab0993c
PM
495#ifdef CONFIG_RCU_BOOST
496DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
497DECLARE_PER_CPU(int, rcu_cpu_kthread_cpu);
498DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);
499DECLARE_PER_CPU(char, rcu_cpu_has_work);
500#endif /* #ifdef CONFIG_RCU_BOOST */
501
017c4261 502#ifndef RCU_TREE_NONCORE
9f77da9f 503
9b2619af 504/* Forward declarations for rcutree_plugin.h */
dbe01350 505static void rcu_bootup_announce(void);
9b2619af 506long rcu_batches_completed(void);
cba6d0d6 507static void rcu_preempt_note_context_switch(int cpu);
27f4d280 508static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp);
b668c9cf 509#ifdef CONFIG_HOTPLUG_CPU
d3f6bad3
PM
510static void rcu_report_unblock_qs_rnp(struct rcu_node *rnp,
511 unsigned long flags);
b668c9cf 512#endif /* #ifdef CONFIG_HOTPLUG_CPU */
1ed509a2 513static void rcu_print_detail_task_stall(struct rcu_state *rsp);
9bc8b558 514static int rcu_print_task_stall(struct rcu_node *rnp);
9b2619af
PM
515static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp);
516#ifdef CONFIG_HOTPLUG_CPU
237c80c5
PM
517static int rcu_preempt_offline_tasks(struct rcu_state *rsp,
518 struct rcu_node *rnp,
519 struct rcu_data *rdp);
9b2619af
PM
520#endif /* #ifdef CONFIG_HOTPLUG_CPU */
521static void rcu_preempt_check_callbacks(int cpu);
9b2619af 522void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu));
d9a3da06 523#if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_TREE_PREEMPT_RCU)
b40d293e
TG
524static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp,
525 bool wake);
d9a3da06 526#endif /* #if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_TREE_PREEMPT_RCU) */
9b2619af 527static void __init __rcu_init_preempt(void);
1217ed1b 528static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags);
a46e0899
PM
529static void rcu_preempt_boost_start_gp(struct rcu_node *rnp);
530static void invoke_rcu_callbacks_kthread(void);
dff1672d 531static bool rcu_is_callbacks_kthread(void);
a46e0899
PM
532#ifdef CONFIG_RCU_BOOST
533static void rcu_preempt_do_callbacks(void);
49fb4c62 534static int rcu_spawn_one_boost_kthread(struct rcu_state *rsp,
5d01bbd1 535 struct rcu_node *rnp);
a46e0899 536#endif /* #ifdef CONFIG_RCU_BOOST */
49fb4c62 537static void rcu_prepare_kthreads(int cpu);
7cb92499 538static void rcu_cleanup_after_idle(int cpu);
aea1b35e 539static void rcu_prepare_for_idle(int cpu);
c57afe80 540static void rcu_idle_count_callbacks_posted(void);
a858af28
PM
541static void print_cpu_stall_info_begin(void);
542static void print_cpu_stall_info(struct rcu_state *rsp, int cpu);
543static void print_cpu_stall_info_end(void);
544static void zero_cpu_stall_ticks(struct rcu_data *rdp);
545static void increment_cpu_stall_ticks(void);
dae6e64d
PM
546static int rcu_nocb_needs_gp(struct rcu_state *rsp);
547static void rcu_nocb_gp_set(struct rcu_node *rnp, int nrq);
0446be48 548static void rcu_nocb_gp_cleanup(struct rcu_state *rsp, struct rcu_node *rnp);
dae6e64d 549static void rcu_init_one_nocb(struct rcu_node *rnp);
3fbfbf7a
PM
550static bool __call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *rhp,
551 bool lazy);
552static bool rcu_nocb_adopt_orphan_cbs(struct rcu_state *rsp,
553 struct rcu_data *rdp);
3fbfbf7a
PM
554static void rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp);
555static void rcu_spawn_nocb_kthreads(struct rcu_state *rsp);
65d798f0 556static void rcu_kick_nohz_cpu(int cpu);
34ed6246 557static bool init_nocb_callback_list(struct rcu_data *rdp);
eb348b89
PM
558static void rcu_sysidle_enter(struct rcu_dynticks *rdtp, int irq);
559static void rcu_sysidle_exit(struct rcu_dynticks *rdtp, int irq);
0edd1b17
PM
560static void rcu_sysidle_check_cpu(struct rcu_data *rdp, bool *isidle,
561 unsigned long *maxj);
562static bool is_sysidle_rcu_state(struct rcu_state *rsp);
563static void rcu_sysidle_report_gp(struct rcu_state *rsp, int isidle,
564 unsigned long maxj);
eb75767b 565static void rcu_bind_gp_kthread(void);
2333210b 566static void rcu_sysidle_init_percpu_data(struct rcu_dynticks *rdtp);
9b2619af 567
017c4261 568#endif /* #ifndef RCU_TREE_NONCORE */
3fbfbf7a
PM
569
570#ifdef CONFIG_RCU_TRACE
571#ifdef CONFIG_RCU_NOCB_CPU
572/* Sum up queue lengths for tracing. */
573static inline void rcu_nocb_q_lengths(struct rcu_data *rdp, long *ql, long *qll)
574{
575 *ql = atomic_long_read(&rdp->nocb_q_count) + rdp->nocb_p_count;
576 *qll = atomic_long_read(&rdp->nocb_q_count_lazy) + rdp->nocb_p_count_lazy;
577}
578#else /* #ifdef CONFIG_RCU_NOCB_CPU */
579static inline void rcu_nocb_q_lengths(struct rcu_data *rdp, long *ql, long *qll)
580{
581 *ql = 0;
582 *qll = 0;
583}
584#endif /* #else #ifdef CONFIG_RCU_NOCB_CPU */
585#endif /* #ifdef CONFIG_RCU_TRACE */