Merge tag 'pci-v6.16-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci
[linux-2.6-block.git] / kernel / rcu / tree.h
CommitLineData
22e40925 1/* SPDX-License-Identifier: GPL-2.0+ */
9f77da9f
PM
2/*
3 * Read-Copy Update mechanism for mutual exclusion (tree-based version)
4 * Internal non-public definitions.
5 *
9f77da9f
PM
6 * Copyright IBM Corporation, 2008
7 *
8 * Author: Ingo Molnar <mingo@elte.hu>
22e40925 9 * Paul E. McKenney <paulmck@linux.ibm.com>
9f77da9f
PM
10 */
11
12#include <linux/cache.h>
9621fbee 13#include <linux/kthread.h>
9f77da9f 14#include <linux/spinlock.h>
037741a6 15#include <linux/rtmutex.h>
9f77da9f
PM
16#include <linux/threads.h>
17#include <linux/cpumask.h>
18#include <linux/seqlock.h>
abedf8e2 19#include <linux/swait.h>
f2425b4e 20#include <linux/rcu_node_tree.h>
f885b7f2 21
45753c5f
IM
22#include "rcu_segcblist.h"
23
23da2ad6 24/* Communicate arguments to a kthread worker handler. */
25f3d7ef 25struct rcu_exp_work {
25f3d7ef 26 unsigned long rew_s;
9621fbee 27 struct kthread_work rew_work;
25f3d7ef
PM
28};
29
d71df90e
PM
30/* RCU's kthread states for tracing. */
31#define RCU_KTHREAD_STOPPED 0
32#define RCU_KTHREAD_RUNNING 1
33#define RCU_KTHREAD_WAITING 2
15ba0ba8
PM
34#define RCU_KTHREAD_OFFCPU 3
35#define RCU_KTHREAD_YIELDING 4
36#define RCU_KTHREAD_MAX 4
d71df90e 37
9f77da9f
PM
38/*
39 * Definition for node within the RCU grace-period-detection hierarchy.
40 */
41struct rcu_node {
67c583a7
BF
42 raw_spinlock_t __private lock; /* Root rcu_node's lock protects */
43 /* some rcu_state fields as well as */
44 /* following. */
360fbbb4 45 unsigned long gp_seq; /* Track rsp->gp_seq. */
adbccddb 46 unsigned long gp_seq_needed; /* Track furthest future GP request. */
4bc8d555 47 unsigned long completedqs; /* All QSes done for this node. */
9f77da9f
PM
48 unsigned long qsmask; /* CPUs or groups that need to switch in */
49 /* order for current grace period to proceed.*/
1eba8f84
PM
50 /* In leaf rcu_node, each bit corresponds to */
51 /* an rcu_data structure, otherwise, each */
52 /* bit corresponds to a child rcu_node */
53 /* structure. */
f2e2df59 54 unsigned long rcu_gp_init_mask; /* Mask of offline CPUs at GP init. */
9f77da9f 55 unsigned long qsmaskinit;
b9585e94 56 /* Per-GP initial value for qsmask. */
0aa04b05
PM
57 /* Initialized from ->qsmaskinitnext at the */
58 /* beginning of each grace period. */
59 unsigned long qsmaskinitnext;
b9585e94
PM
60 unsigned long expmask; /* CPUs or groups that need to check in */
61 /* to allow the current expedited GP */
62 /* to complete. */
63 unsigned long expmaskinit;
64 /* Per-GP initial values for expmask. */
65 /* Initialized from ->expmaskinitnext at the */
66 /* beginning of each expedited GP. */
67 unsigned long expmaskinitnext;
68 /* Online CPUs for next expedited GP. */
1de6e56d
PM
69 /* Any CPU that has ever been online will */
70 /* have its bit set. */
8e5e6215
FW
71 struct kthread_worker *exp_kworker;
72 /* Workers performing per node expedited GP */
73 /* initialization. */
b2b00ddf
PM
74 unsigned long cbovldmask;
75 /* CPUs experiencing callback overload. */
9b9500da 76 unsigned long ffmask; /* Fully functional CPUs. */
9f77da9f 77 unsigned long grpmask; /* Mask to apply to parent qsmask. */
1eba8f84 78 /* Only one bit will be set in this mask. */
a2dae430
WY
79 int grplo; /* lowest-numbered CPU here. */
80 int grphi; /* highest-numbered CPU here. */
7a0c2b09 81 u8 grpnum; /* group number for next level up. */
9f77da9f 82 u8 level; /* root is at level 0. */
0aa04b05
PM
83 bool wait_blkd_tasks;/* Necessary to wait for blocked tasks to */
84 /* exit RCU read-side critical sections */
85 /* before propagating offline up the */
86 /* rcu_node tree? */
9f77da9f 87 struct rcu_node *parent;
12f5f524
PM
88 struct list_head blkd_tasks;
89 /* Tasks blocked in RCU read-side critical */
90 /* section. Tasks are placed at the head */
91 /* of this list and age towards the tail. */
92 struct list_head *gp_tasks;
93 /* Pointer to the first task blocking the */
94 /* current grace period, or NULL if there */
95 /* is no such task. */
96 struct list_head *exp_tasks;
97 /* Pointer to the first task blocking the */
98 /* current expedited grace period, or NULL */
99 /* if there is no such task. If there */
100 /* is no current expedited grace period, */
101 /* then there can cannot be any such task. */
27f4d280
PM
102 struct list_head *boost_tasks;
103 /* Pointer to first task that needs to be */
104 /* priority boosted, or NULL if no priority */
105 /* boosting is needed for this rcu_node */
106 /* structure. If there are no tasks */
107 /* queued on this rcu_node structure that */
108 /* are blocking the current grace period, */
109 /* there can be no such task. */
abaa93d9
PM
110 struct rt_mutex boost_mtx;
111 /* Used only for the priority-boosting */
112 /* side effect, not as a lock. */
27f4d280
PM
113 unsigned long boost_time;
114 /* When to start boosting (jiffies). */
7836b270 115 struct mutex kthread_mutex;
218b957a
DW
116 /* Exclusion for thread spawning and affinity */
117 /* manipulation. */
27f4d280
PM
118 struct task_struct *boost_kthread_task;
119 /* kthread that takes care of priority */
120 /* boosting for this rcu_node structure. */
d71df90e
PM
121 unsigned int boost_kthread_status;
122 /* State of boost_kthread_task for tracing. */
396eba65 123 unsigned long n_boosts; /* Number of boosts for this rcu_node structure. */
dae6e64d 124#ifdef CONFIG_RCU_NOCB_CPU
abedf8e2 125 struct swait_queue_head nocb_gp_wq[2];
dae6e64d 126 /* Place for rcu_nocb_kthread() to wait GP. */
dae6e64d 127#endif /* #ifdef CONFIG_RCU_NOCB_CPU */
394f2769 128 raw_spinlock_t fqslock ____cacheline_internodealigned_in_smp;
385b73c0 129
f6a12f34
PM
130 spinlock_t exp_lock ____cacheline_internodealigned_in_smp;
131 unsigned long exp_seq_rq;
3b5f668e 132 wait_queue_head_t exp_wq[4];
25f3d7ef
PM
133 struct rcu_exp_work rew;
134 bool exp_need_flush; /* Need to flush workitem? */
d96c52fe
PM
135 raw_spinlock_t exp_poll_lock;
136 /* Lock and data for polled expedited grace periods. */
137 unsigned long exp_seq_poll_rq;
138 struct work_struct exp_poll_wq;
9f77da9f
PM
139} ____cacheline_internodealigned_in_smp;
140
bc75e999
MR
141/*
142 * Bitmasks in an rcu_node cover the interval [grplo, grphi] of CPU IDs, and
143 * are indexed relative to this interval rather than the global CPU ID space.
144 * This generates the bit for a CPU in node-local masks.
145 */
df63fa5b 146#define leaf_node_cpu_bit(rnp, cpu) (BIT((cpu) - (rnp)->grplo))
bc75e999 147
5b74c458
PM
148/*
149 * Union to allow "aggregate OR" operation on the need for a quiescent
150 * state by the normal and expedited grace periods.
151 */
152union rcu_noqs {
153 struct {
154 u8 norm;
155 u8 exp;
156 } b; /* Bits. */
157 u16 s; /* Set of bits, aggregate OR here. */
158};
159
be42f00b
ZL
160/*
161 * Record the snapshot of the core stats at half of the first RCU stall timeout.
162 * The member gp_seq is used to ensure that all members are updated only once
163 * during the sampling period. The snapshot is taken only if this gp_seq is not
164 * equal to rdp->gp_seq.
165 */
166struct rcu_snap_record {
167 unsigned long gp_seq; /* Track rdp->gp_seq counter */
168 u64 cputime_irq; /* Accumulated cputime of hard irqs */
169 u64 cputime_softirq;/* Accumulated cputime of soft irqs */
170 u64 cputime_system; /* Accumulated cputime of kernel tasks */
da6b8559 171 u64 nr_hardirqs; /* Accumulated number of hard irqs */
be42f00b
ZL
172 unsigned int nr_softirqs; /* Accumulated number of soft irqs */
173 unsigned long long nr_csw; /* Accumulated number of task switches */
174 unsigned long jiffies; /* Track jiffies value */
175};
176
9f77da9f
PM
177/* Per-CPU data for read-copy update. */
178struct rcu_data {
179 /* 1) quiescent-state and grace-period handling : */
360fbbb4 180 unsigned long gp_seq; /* Track rsp->gp_seq counter. */
adbccddb 181 unsigned long gp_seq_needed; /* Track furthest future GP request. */
5b74c458 182 union rcu_noqs cpu_no_qs; /* No QSes yet for this CPU. */
a616aec9 183 bool core_needs_qs; /* Core waits for quiescent state. */
9f77da9f 184 bool beenonline; /* CPU online at least once. */
ff3bb6f4 185 bool gpwrap; /* Possible ->gp_seq wrap. */
aafe12f9 186 unsigned int gpwrap_count; /* Count of GP sequence wrap. */
c0f97f20 187 bool cpu_started; /* RCU watching this onlining CPU. */
9f77da9f
PM
188 struct rcu_node *mynode; /* This CPU's leaf of hierarchy */
189 unsigned long grpmask; /* Mask to apply to leaf qsmask. */
a858af28
PM
190 unsigned long ticks_this_gp; /* The number of scheduling-clock */
191 /* ticks this CPU has handled */
192 /* during and after the last grace */
193 /* period it is aware of. */
0864f057
PM
194 struct irq_work defer_qs_iw; /* Obtain later scheduler attention. */
195 bool defer_qs_iw_pending; /* Scheduler attention pending? */
a657f261 196 struct work_struct strict_work; /* Schedule readers for strict GPs. */
9f77da9f
PM
197
198 /* 2) batch handling */
15fecf89
PM
199 struct rcu_segcblist cblist; /* Segmented callback list, with */
200 /* different callbacks waiting for */
201 /* different grace periods. */
37c72e56
PM
202 long qlen_last_fqs_check;
203 /* qlen at last check for QS forcing */
e816d56f 204 unsigned long n_cbs_invoked; /* # callbacks invoked since boot. */
37c72e56
PM
205 unsigned long n_force_qs_snap;
206 /* did other CPU force QS recently? */
9f77da9f
PM
207 long blimit; /* Upper limit on a processed batch */
208
9f77da9f 209 /* 3) dynticks interface. */
2dba2230 210 int watching_snap; /* Per-GP tracking for dynticks. */
dc5a4f29
PM
211 bool rcu_need_heavy_qs; /* GP old, so heavy quiescent state! */
212 bool rcu_urgent_qs; /* GP old need light quiescent state. */
66e4c33b 213 bool rcu_forced_tick; /* Forced tick to provide QS. */
df1e849a 214 bool rcu_forced_tick_exp; /* ... provide QS to expedited GP. */
9f77da9f 215
8d8a9d0e 216 /* 4) rcu_barrier(), OOM callbacks, and expediting. */
a16578dd 217 unsigned long barrier_seq_snap; /* Snap of rcu_state.barrier_sequence. */
06668efa 218 struct rcu_head barrier_head;
76ce2b3d 219 int exp_watching_snap; /* Double-check need for IPI. */
06668efa 220
8d8a9d0e 221 /* 5) Callback offloading. */
3fbfbf7a 222#ifdef CONFIG_RCU_NOCB_CPU
12f54c3a 223 struct swait_queue_head nocb_cb_wq; /* For nocb kthreads to sleep on. */
d97b0781 224 struct swait_queue_head nocb_state_wq; /* For offloading state changes */
12f54c3a 225 struct task_struct *nocb_gp_kthread;
8be6e1b1 226 raw_spinlock_t nocb_lock; /* Guard following pair of fields. */
9fdd3bc9 227 int nocb_defer_wakeup; /* Defer wakeup of nocb_kthread. */
8be6e1b1 228 struct timer_list nocb_timer; /* Enforce finite deferral. */
d1b222c6 229 unsigned long nocb_gp_adv_time; /* Last call_rcu() CB adv (jiffies). */
02e30241
NU
230 struct mutex nocb_gp_kthread_mutex; /* Exclusion for nocb gp kthread */
231 /* spawning */
d1b222c6
PM
232
233 /* The following fields are used by call_rcu, hence own cacheline. */
234 raw_spinlock_t nocb_bypass_lock ____cacheline_internodealigned_in_smp;
235 struct rcu_cblist nocb_bypass; /* Lock-contention-bypass CB list. */
236 unsigned long nocb_bypass_first; /* Time (jiffies) of first enqueue. */
237 unsigned long nocb_nobypass_last; /* Last ->cblist enqueue (jiffies). */
238 int nocb_nobypass_count; /* # ->cblist enqueues at ^^^ time. */
fbce7497 239
6484fe54 240 /* The following fields are used by GP kthread, hence own cacheline. */
4fd8c5f1 241 raw_spinlock_t nocb_gp_lock ____cacheline_internodealigned_in_smp;
f7a81b12
PM
242 u8 nocb_gp_sleep; /* Is the nocb GP thread asleep? */
243 u8 nocb_gp_bypass; /* Found a bypass on last scan? */
244 u8 nocb_gp_gp; /* GP to wait for on last scan? */
245 unsigned long nocb_gp_seq; /* If so, ->gp_seq to wait for. */
246 unsigned long nocb_gp_loops; /* # passes through wait code. */
12f54c3a 247 struct swait_queue_head nocb_gp_wq; /* For nocb kthreads to sleep on. */
5d6742b3 248 bool nocb_cb_sleep; /* Is the nocb CB thread asleep? */
12f54c3a 249 struct task_struct *nocb_cb_kthread;
2ebc45c4
FW
250 struct list_head nocb_head_rdp; /*
251 * Head of rcu_data list in wakeup chain,
252 * if rdp_gp.
253 */
254 struct list_head nocb_entry_rdp; /* rcu_data node in wakeup chain. */
1598f4a4 255 struct rcu_data *nocb_toggling_rdp; /* rdp queued for (de-)offloading */
fbce7497 256
d1b222c6 257 /* The following fields are used by CB kthread, hence new cacheline. */
58bf6f77 258 struct rcu_data *nocb_gp_rdp ____cacheline_internodealigned_in_smp;
6484fe54 259 /* GP rdp takes GP-end wakeups. */
3fbfbf7a
PM
260#endif /* #ifdef CONFIG_RCU_NOCB_CPU */
261
37f62d7c
PM
262 /* 6) RCU priority boosting. */
263 struct task_struct *rcu_cpu_kthread_task;
264 /* rcuc per-CPU kthread or NULL. */
6ffdde28 265 unsigned int rcu_cpu_kthread_status;
f7e972ee 266 char rcu_cpu_has_work;
c9515875 267 unsigned long rcuc_activity;
37f62d7c
PM
268
269 /* 7) Diagnostic data, including RCU CPU stall warnings. */
6231069b 270 unsigned int softirq_snap; /* Snapshot of softirq activity. */
9b9500da
PM
271 /* ->rcu_iw* fields protected by leaf rcu_node ->lock. */
272 struct irq_work rcu_iw; /* Check for non-irq activity. */
273 bool rcu_iw_pending; /* Is ->rcu_iw pending? */
8aa670cd 274 unsigned long rcu_iw_gp_seq; /* ->gp_seq associated with ->rcu_iw. */
57738942 275 unsigned long rcu_ofl_gp_seq; /* ->gp_seq at last offline. */
ae2b217a 276 short rcu_ofl_gp_state; /* ->gp_state at last offline. */
57738942 277 unsigned long rcu_onl_gp_seq; /* ->gp_seq at last online. */
ae2b217a 278 short rcu_onl_gp_state; /* ->gp_state at last online. */
d3052109 279 unsigned long last_fqs_resched; /* Time of last rcu_resched(). */
c708b08c 280 unsigned long last_sched_clock; /* Jiffies of last rcu_sched_clock_irq(). */
be42f00b
ZL
281 struct rcu_snap_record snap_record; /* Snapshot of core stats at half of */
282 /* the first RCU stall timeout */
6231069b 283
3cb278e7 284 long lazy_len; /* Length of buffered lazy callbacks. */
9f77da9f
PM
285 int cpu;
286};
287
9fdd3bc9 288/* Values for nocb_defer_wakeup field in struct rcu_data. */
511324e4 289#define RCU_NOCB_WAKE_NOT 0
e75bcd48 290#define RCU_NOCB_WAKE_BYPASS 1
3cb278e7
JFG
291#define RCU_NOCB_WAKE_LAZY 2
292#define RCU_NOCB_WAKE 3
293#define RCU_NOCB_WAKE_FORCE 4
9fdd3bc9 294
026ad283
PM
295#define RCU_JIFFIES_TILL_FORCE_QS (1 + (HZ > 250) + (HZ > 500))
296 /* For jiffies_till_first_fqs and */
297 /* and jiffies_till_next_fqs. */
007b0924 298
026ad283
PM
299#define RCU_JIFFIES_FQS_DIV 256 /* Very large systems need more */
300 /* delay between bouts of */
301 /* quiescent-state forcing. */
302
303#define RCU_STALL_RAT_DELAY 2 /* Allow other CPUs time to take */
304 /* at least one scheduling clock */
305 /* irq before ratting on them. */
9f77da9f 306
08bca60a
PZ
307#define rcu_wait(cond) \
308do { \
309 for (;;) { \
310 set_current_state(TASK_INTERRUPTIBLE); \
311 if (cond) \
312 break; \
313 schedule(); \
314 } \
315 __set_current_state(TASK_RUNNING); \
316} while (0)
9f77da9f 317
462df2f5
URS
318/*
319 * A max threshold for synchronize_rcu() users which are
320 * awaken directly by the rcu_gp_kthread(). Left part is
321 * deferred to the main worker.
322 */
323#define SR_MAX_USERS_WAKE_FROM_GP 5
dfd458a9
URS
324#define SR_NORMAL_GP_WAIT_HEAD_MAX 5
325
326struct sr_wait_node {
327 atomic_t inuse;
328 struct llist_node node;
329};
330
9f77da9f
PM
331/*
332 * RCU global state, including node hierarchy. This hierarchy is
333 * represented in "heap" form in a dense array. The root (first level)
334 * of the hierarchy is in ->node[0] (referenced by ->level[0]), the second
335 * level in ->node[1] through ->node[m] (->node[1] referenced by ->level[1]),
336 * and the third level in ->node[m+1] and following (->node[m+1] referenced
337 * by ->level[2]). The number of levels is determined by the number of
338 * CPUs and by CONFIG_RCU_FANOUT. Small systems will have a "hierarchy"
339 * consisting of a single rcu_node.
340 */
341struct rcu_state {
342 struct rcu_node node[NUM_RCU_NODES]; /* Hierarchy. */
032dfc87
AG
343 struct rcu_node *level[RCU_NUM_LVLS + 1];
344 /* Hierarchy levels (+1 to */
345 /* shut bogus gcc warning) */
b9585e94 346 int ncpus; /* # CPUs seen so far. */
ed73860c 347 int n_online_cpus; /* # CPUs online for RCU. */
9f77da9f
PM
348
349 /* The following fields are guarded by the root rcu_node's lock. */
350
eae9f147
NU
351 unsigned long gp_seq ____cacheline_internodealigned_in_smp;
352 /* Grace-period sequence #. */
00943a60
WY
353 unsigned long gp_max; /* Maximum GP duration in */
354 /* jiffies. */
b3dbec76 355 struct task_struct *gp_kthread; /* Task for grace periods. */
abedf8e2 356 struct swait_queue_head gp_wq; /* Where GP task waits. */
afea227f
PM
357 short gp_flags; /* Commands for GP task. */
358 short gp_state; /* GP kthread sleep state. */
fd897573
PM
359 unsigned long gp_wake_time; /* Last GP kthread wake. */
360 unsigned long gp_wake_seq; /* ->gp_seq at ^^^. */
bf95b2bc
PM
361 unsigned long gp_seq_polled; /* GP seq for polled API. */
362 unsigned long gp_seq_polled_snap; /* ->gp_seq_polled at normal GP start. */
dd041405 363 unsigned long gp_seq_polled_exp_snap; /* ->gp_seq_polled at expedited GP start. */
1eba8f84 364
d9a3da06 365 /* End of fields guarded by root rcu_node's lock. */
1eba8f84 366
7be7f0be 367 struct mutex barrier_mutex; /* Guards barrier fields. */
24ebbca8 368 atomic_t barrier_cpu_count; /* # CPUs waiting on. */
7db74df8 369 struct completion barrier_completion; /* Wake at barrier end. */
4f525a52 370 unsigned long barrier_sequence; /* ++ at start and end of */
dd46a788 371 /* rcu_barrier(). */
a4fbe35a
PM
372 /* End of fields guarded by barrier_mutex. */
373
80b3fd47
PM
374 raw_spinlock_t barrier_lock; /* Protects ->barrier_seq_snap. */
375
f6a12f34 376 struct mutex exp_mutex; /* Serialize expedited GP. */
3b5f668e 377 struct mutex exp_wake_mutex; /* Serialize wakeup. */
d6ada2cf 378 unsigned long expedited_sequence; /* Take a ticket. */
3a6d7c64 379 atomic_t expedited_need_qs; /* # CPUs left to check in. */
abedf8e2 380 struct swait_queue_head expedited_wq; /* Wait for check-ins. */
b9585e94 381 int ncpus_snap; /* # CPUs seen last time. */
b2b00ddf
PM
382 u8 cbovld; /* Callback overload now? */
383 u8 cbovldnext; /* ^ ^ next time? */
40694d66 384
9f77da9f
PM
385 unsigned long jiffies_force_qs; /* Time at which to invoke */
386 /* force_quiescent_state(). */
8c7c4829
PM
387 unsigned long jiffies_kick_kthreads; /* Time at which to kick */
388 /* kthreads, if configured. */
9f77da9f
PM
389 unsigned long n_force_qs; /* Number of calls to */
390 /* force_quiescent_state(). */
9f77da9f
PM
391 unsigned long gp_start; /* Time at which GP started, */
392 /* but in jiffies. */
c51d7b5e
PM
393 unsigned long gp_end; /* Time last GP ended, again */
394 /* in jiffies. */
6ccd2ecd
PM
395 unsigned long gp_activity; /* Time of last GP kthread */
396 /* activity in jiffies. */
26d950a9
PM
397 unsigned long gp_req_activity; /* Time of last GP request */
398 /* in jiffies. */
9f77da9f
PM
399 unsigned long jiffies_stall; /* Time at which to check */
400 /* for CPU stalls. */
b96e7a5f
JFG
401 int nr_fqs_jiffies_stall; /* Number of fqs loops after
402 * which read jiffies and set
403 * jiffies_stall. Stall
404 * warnings disabled if !0. */
6193c76a
PM
405 unsigned long jiffies_resched; /* Time at which to resched */
406 /* a reluctant CPU. */
fc908ed3
PM
407 unsigned long n_force_qs_gpstart; /* Snapshot of n_force_qs at */
408 /* GP start. */
e66c33d5 409 const char *name; /* Name of structure. */
a4889858 410 char abbr; /* Abbreviated name. */
1e64b15a 411
82980b16 412 arch_spinlock_t ofl_lock ____cacheline_internodealigned_in_smp;
1e64b15a
PM
413 /* Synchronize offline with */
414 /* GP pre-initialization. */
dfd458a9
URS
415
416 /* synchronize_rcu() part. */
417 struct llist_head srs_next; /* request a GP users. */
418 struct llist_node *srs_wait_tail; /* wait for GP users. */
419 struct llist_node *srs_done_tail; /* ready for GP users. */
420 struct sr_wait_node srs_wait_nodes[SR_NORMAL_GP_WAIT_HEAD_MAX];
421 struct work_struct srs_cleanup_work;
6f948568 422 atomic_t srs_cleanups_pending; /* srs inflight worker cleanups. */
ff81428e
FW
423
424#ifdef CONFIG_RCU_NOCB_CPU
7121dd91 425 struct mutex nocb_mutex; /* Guards (de-)offloading */
ff81428e
FW
426 int nocb_is_setup; /* nocb is setup from boot */
427#endif
9f77da9f
PM
428};
429
4cdfc175
PM
430/* Values for rcu_state structure's gp_flags field. */
431#define RCU_GP_FLAG_INIT 0x1 /* Need grace-period initialization. */
432#define RCU_GP_FLAG_FQS 0x2 /* Need grace-period quiescent-state forcing. */
1fca4d12 433#define RCU_GP_FLAG_OVLD 0x4 /* Experiencing callback overload. */
4cdfc175 434
c34d2f41 435/* Values for rcu_state structure's gp_state field. */
77f81fe0 436#define RCU_GP_IDLE 0 /* Initial state and no GP in progress. */
afea227f 437#define RCU_GP_WAIT_GPS 1 /* Wait for grace-period start. */
319362c9 438#define RCU_GP_DONE_GPS 2 /* Wait done for grace-period start. */
fea3f222
PM
439#define RCU_GP_ONOFF 3 /* Grace-period initialization hotplug. */
440#define RCU_GP_INIT 4 /* Grace-period initialization. */
441#define RCU_GP_WAIT_FQS 5 /* Wait for force-quiescent-state time. */
442#define RCU_GP_DOING_FQS 6 /* Wait done for force-quiescent-state time. */
443#define RCU_GP_CLEANUP 7 /* Grace-period cleanup started. */
444#define RCU_GP_CLEANED 8 /* Grace-period cleanup complete. */
afea227f 445
358be2d3
PM
446/*
447 * In order to export the rcu_state name to the tracing tools, it
448 * needs to be added in the __tracepoint_string section.
449 * This requires defining a separate variable tp_<sname>_varname
450 * that points to the string being used, and this will allow
451 * the tracing userspace tools to be able to decipher the string
452 * address to the matching string.
453 */
454#ifdef CONFIG_PREEMPT_RCU
455#define RCU_ABBR 'p'
456#define RCU_NAME_RAW "rcu_preempt"
457#else /* #ifdef CONFIG_PREEMPT_RCU */
458#define RCU_ABBR 's'
459#define RCU_NAME_RAW "rcu_sched"
460#endif /* #else #ifdef CONFIG_PREEMPT_RCU */
461#ifndef CONFIG_TRACING
462#define RCU_NAME RCU_NAME_RAW
463#else /* #ifdef CONFIG_TRACING */
464static char rcu_name[] = RCU_NAME_RAW;
465static const char *tp_rcu_varname __used __tracepoint_string = rcu_name;
466#define RCU_NAME rcu_name
467#endif /* #else #ifdef CONFIG_TRACING */
6b50e119 468
32255d51 469/* Forward declarations for tree_plugin.h */
dbe01350 470static void rcu_bootup_announce(void);
45975c7d 471static void rcu_qs(void);
27f4d280 472static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp);
b668c9cf 473#ifdef CONFIG_HOTPLUG_CPU
8af3a5e7 474static bool rcu_preempt_has_tasks(struct rcu_node *rnp);
b668c9cf 475#endif /* #ifdef CONFIG_HOTPLUG_CPU */
74611ecb 476static int rcu_print_task_exp_stall(struct rcu_node *rnp);
81ab59a3 477static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp);
c98cac60 478static void rcu_flavor_sched_clock_irq(int user);
81ab59a3 479static void dump_blkd_tasks(struct rcu_node *rnp, int ncheck);
1217ed1b 480static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags);
a46e0899 481static void rcu_preempt_boost_start_gp(struct rcu_node *rnp);
51038506 482static bool rcu_is_callbacks_kthread(struct rcu_data *rdp);
48d07c04 483static void rcu_cpu_kthread_setup(unsigned int cpu);
3ef5a1c3 484static void rcu_spawn_one_boost_kthread(struct rcu_node *rnp);
0aa04b05 485static bool rcu_preempt_has_tasks(struct rcu_node *rnp);
3e310098 486static bool rcu_preempt_need_deferred_qs(struct task_struct *t);
a858af28 487static void zero_cpu_stall_ticks(struct rcu_data *rdp);
abedf8e2
PG
488static struct swait_queue_head *rcu_nocb_gp_get(struct rcu_node *rnp);
489static void rcu_nocb_gp_cleanup(struct swait_queue_head *sq);
dae6e64d 490static void rcu_init_one_nocb(struct rcu_node *rnp);
b8f7aca3 491static bool wake_nocb_gp(struct rcu_data *rdp, bool force);
d1b222c6 492static bool rcu_nocb_flush_bypass(struct rcu_data *rdp, struct rcu_head *rhp,
3cb278e7 493 unsigned long j, bool lazy);
afd4e696
FW
494static void call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *head,
495 rcu_callback_t func, unsigned long flags, bool lazy);
496static void __maybe_unused __call_rcu_nocb_wake(struct rcu_data *rdp, bool was_empty,
497 unsigned long flags);
87090516 498static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp, int level);
f8bb5cae 499static bool do_nocb_deferred_wakeup(struct rcu_data *rdp);
3fbfbf7a 500static void rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp);
ad368d15 501static void rcu_spawn_cpu_nocb_kthread(int cpu);
f7a81b12 502static void show_rcu_nocb_state(struct rcu_data *rdp);
5d6742b3
PM
503static void rcu_nocb_lock(struct rcu_data *rdp);
504static void rcu_nocb_unlock(struct rcu_data *rdp);
505static void rcu_nocb_unlock_irqrestore(struct rcu_data *rdp,
506 unsigned long flags);
d1b222c6 507static void rcu_lockdep_assert_cblist_protected(struct rcu_data *rdp);
35ce7f29 508#ifdef CONFIG_RCU_NOCB_CPU
4580b054 509static void __init rcu_organize_nocb_kthreads(void);
118e0d4a
FW
510
511/*
512 * Disable IRQs before checking offloaded state so that local
513 * locking is safe against concurrent de-offloading.
514 */
515#define rcu_nocb_lock_irqsave(rdp, flags) \
516do { \
517 local_irq_save(flags); \
518 if (rcu_segcblist_is_offloaded(&(rdp)->cblist)) \
519 raw_spin_lock(&(rdp)->nocb_lock); \
81c0b3d7
PM
520} while (0)
521#else /* #ifdef CONFIG_RCU_NOCB_CPU */
522#define rcu_nocb_lock_irqsave(rdp, flags) local_irq_save(flags)
523#endif /* #else #ifdef CONFIG_RCU_NOCB_CPU */
524
eb75767b 525static void rcu_bind_gp_kthread(void);
4580b054 526static bool rcu_nohz_full_cpu(void);
32255d51
PM
527
528/* Forward declarations for tree_stall.h */
32255d51 529static void record_gp_stall_check_time(void);
7ac1907c 530static void rcu_iw_handler(struct irq_work *iwp);
32255d51 531static void check_cpu_stall(struct rcu_data *rdp);
b51bcbbf
PM
532static void rcu_check_gp_start_stall(struct rcu_node *rnp, struct rcu_data *rdp,
533 const unsigned long gpssdelay);
d96c52fe
PM
534
535/* Forward declarations for tree_exp.h. */
536static void sync_rcu_do_polled_gp(struct work_struct *wp);