]>
Commit | Line | Data |
---|---|---|
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> | |
30 | ||
31 | /* | |
8932a63d PM |
32 | * Define shape of hierarchy based on NR_CPUS, CONFIG_RCU_FANOUT, and |
33 | * CONFIG_RCU_FANOUT_LEAF. | |
9f77da9f | 34 | * In theory, it should be possible to add more levels straightforwardly. |
0209f649 PM |
35 | * In practice, this did work well going from three levels to four. |
36 | * Of course, your mileage may vary. | |
9f77da9f | 37 | */ |
cf244dc0 | 38 | #define MAX_RCU_LVLS 4 |
8932a63d | 39 | #define RCU_FANOUT_1 (CONFIG_RCU_FANOUT_LEAF) |
0209f649 PM |
40 | #define RCU_FANOUT_2 (RCU_FANOUT_1 * CONFIG_RCU_FANOUT) |
41 | #define RCU_FANOUT_3 (RCU_FANOUT_2 * CONFIG_RCU_FANOUT) | |
42 | #define RCU_FANOUT_4 (RCU_FANOUT_3 * CONFIG_RCU_FANOUT) | |
9f77da9f | 43 | |
0209f649 | 44 | #if NR_CPUS <= RCU_FANOUT_1 |
f885b7f2 | 45 | # define RCU_NUM_LVLS 1 |
9f77da9f PM |
46 | # define NUM_RCU_LVL_0 1 |
47 | # define NUM_RCU_LVL_1 (NR_CPUS) | |
48 | # define NUM_RCU_LVL_2 0 | |
49 | # define NUM_RCU_LVL_3 0 | |
cf244dc0 | 50 | # define NUM_RCU_LVL_4 0 |
0209f649 | 51 | #elif NR_CPUS <= RCU_FANOUT_2 |
f885b7f2 | 52 | # define RCU_NUM_LVLS 2 |
9f77da9f | 53 | # define NUM_RCU_LVL_0 1 |
0209f649 | 54 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1) |
9f77da9f PM |
55 | # define NUM_RCU_LVL_2 (NR_CPUS) |
56 | # define NUM_RCU_LVL_3 0 | |
cf244dc0 | 57 | # define NUM_RCU_LVL_4 0 |
0209f649 | 58 | #elif NR_CPUS <= RCU_FANOUT_3 |
f885b7f2 | 59 | # define RCU_NUM_LVLS 3 |
9f77da9f | 60 | # define NUM_RCU_LVL_0 1 |
0209f649 PM |
61 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2) |
62 | # define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1) | |
63 | # define NUM_RCU_LVL_3 (NR_CPUS) | |
cf244dc0 | 64 | # define NUM_RCU_LVL_4 0 |
0209f649 | 65 | #elif NR_CPUS <= RCU_FANOUT_4 |
f885b7f2 | 66 | # define RCU_NUM_LVLS 4 |
cf244dc0 | 67 | # define NUM_RCU_LVL_0 1 |
0209f649 PM |
68 | # define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_3) |
69 | # define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2) | |
70 | # define NUM_RCU_LVL_3 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1) | |
71 | # define NUM_RCU_LVL_4 (NR_CPUS) | |
9f77da9f PM |
72 | #else |
73 | # error "CONFIG_RCU_FANOUT insufficient for NR_CPUS" | |
0209f649 | 74 | #endif /* #if (NR_CPUS) <= RCU_FANOUT_1 */ |
9f77da9f | 75 | |
cf244dc0 | 76 | #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 |
77 | #define NUM_RCU_NODES (RCU_SUM - NR_CPUS) |
78 | ||
f885b7f2 PM |
79 | extern int rcu_num_lvls; |
80 | extern int rcu_num_nodes; | |
81 | ||
9f77da9f PM |
82 | /* |
83 | * Dynticks per-CPU state. | |
84 | */ | |
85 | struct rcu_dynticks { | |
9b2e4f18 PM |
86 | long long dynticks_nesting; /* Track irq/process nesting level. */ |
87 | /* Process level is worth LLONG_MAX/2. */ | |
88 | int dynticks_nmi_nesting; /* Track NMI nesting level. */ | |
89 | atomic_t dynticks; /* Even value for idle, else odd. */ | |
5955f7ee PM |
90 | #ifdef CONFIG_RCU_FAST_NO_HZ |
91 | int dyntick_drain; /* Prepare-for-idle state variable. */ | |
92 | unsigned long dyntick_holdoff; | |
93 | /* No retries for the jiffy of failure. */ | |
94 | struct timer_list idle_gp_timer; | |
95 | /* Wake up CPU sleeping with callbacks. */ | |
96 | unsigned long idle_gp_timer_expires; | |
97 | /* When to wake up CPU (for repost). */ | |
98 | bool idle_first_pass; /* First pass of attempt to go idle? */ | |
99 | unsigned long nonlazy_posted; | |
100 | /* # times non-lazy CBs posted to CPU. */ | |
101 | unsigned long nonlazy_posted_snap; | |
102 | /* idle-period nonlazy_posted snapshot. */ | |
9d2ad243 | 103 | int tick_nohz_enabled_snap; /* Previously seen value from sysfs. */ |
5955f7ee | 104 | #endif /* #ifdef CONFIG_RCU_FAST_NO_HZ */ |
9f77da9f PM |
105 | }; |
106 | ||
d71df90e PM |
107 | /* RCU's kthread states for tracing. */ |
108 | #define RCU_KTHREAD_STOPPED 0 | |
109 | #define RCU_KTHREAD_RUNNING 1 | |
110 | #define RCU_KTHREAD_WAITING 2 | |
15ba0ba8 PM |
111 | #define RCU_KTHREAD_OFFCPU 3 |
112 | #define RCU_KTHREAD_YIELDING 4 | |
113 | #define RCU_KTHREAD_MAX 4 | |
d71df90e | 114 | |
9f77da9f PM |
115 | /* |
116 | * Definition for node within the RCU grace-period-detection hierarchy. | |
117 | */ | |
118 | struct rcu_node { | |
1304afb2 | 119 | raw_spinlock_t lock; /* Root rcu_node's lock protects some */ |
1eba8f84 | 120 | /* rcu_state fields as well as following. */ |
20133cfc | 121 | unsigned long gpnum; /* Current grace period for this node. */ |
86848966 PM |
122 | /* This will either be equal to or one */ |
123 | /* behind the root rcu_node's gpnum. */ | |
20133cfc | 124 | unsigned long completed; /* Last GP completed for this node. */ |
d09b62df PM |
125 | /* This will either be equal to or one */ |
126 | /* behind the root rcu_node's gpnum. */ | |
9f77da9f PM |
127 | unsigned long qsmask; /* CPUs or groups that need to switch in */ |
128 | /* order for current grace period to proceed.*/ | |
1eba8f84 PM |
129 | /* In leaf rcu_node, each bit corresponds to */ |
130 | /* an rcu_data structure, otherwise, each */ | |
131 | /* bit corresponds to a child rcu_node */ | |
132 | /* structure. */ | |
12f5f524 | 133 | unsigned long expmask; /* Groups that have ->blkd_tasks */ |
d9a3da06 PM |
134 | /* elements that need to drain to allow the */ |
135 | /* current expedited grace period to */ | |
136 | /* complete (only for TREE_PREEMPT_RCU). */ | |
9f77da9f | 137 | unsigned long qsmaskinit; |
d9a3da06 | 138 | /* Per-GP initial value for qsmask & expmask. */ |
9f77da9f | 139 | unsigned long grpmask; /* Mask to apply to parent qsmask. */ |
1eba8f84 | 140 | /* Only one bit will be set in this mask. */ |
9f77da9f PM |
141 | int grplo; /* lowest-numbered CPU or group here. */ |
142 | int grphi; /* highest-numbered CPU or group here. */ | |
143 | u8 grpnum; /* CPU/group number for next level up. */ | |
144 | u8 level; /* root is at level 0. */ | |
145 | struct rcu_node *parent; | |
12f5f524 PM |
146 | struct list_head blkd_tasks; |
147 | /* Tasks blocked in RCU read-side critical */ | |
148 | /* section. Tasks are placed at the head */ | |
149 | /* of this list and age towards the tail. */ | |
150 | struct list_head *gp_tasks; | |
151 | /* Pointer to the first task blocking the */ | |
152 | /* current grace period, or NULL if there */ | |
153 | /* is no such task. */ | |
154 | struct list_head *exp_tasks; | |
155 | /* Pointer to the first task blocking the */ | |
156 | /* current expedited grace period, or NULL */ | |
157 | /* if there is no such task. If there */ | |
158 | /* is no current expedited grace period, */ | |
159 | /* then there can cannot be any such task. */ | |
27f4d280 PM |
160 | #ifdef CONFIG_RCU_BOOST |
161 | struct list_head *boost_tasks; | |
162 | /* Pointer to first task that needs to be */ | |
163 | /* priority boosted, or NULL if no priority */ | |
164 | /* boosting is needed for this rcu_node */ | |
165 | /* structure. If there are no tasks */ | |
166 | /* queued on this rcu_node structure that */ | |
167 | /* are blocking the current grace period, */ | |
168 | /* there can be no such task. */ | |
169 | unsigned long boost_time; | |
170 | /* When to start boosting (jiffies). */ | |
171 | struct task_struct *boost_kthread_task; | |
172 | /* kthread that takes care of priority */ | |
173 | /* boosting for this rcu_node structure. */ | |
d71df90e PM |
174 | unsigned int boost_kthread_status; |
175 | /* State of boost_kthread_task for tracing. */ | |
0ea1f2eb PM |
176 | unsigned long n_tasks_boosted; |
177 | /* Total number of tasks boosted. */ | |
178 | unsigned long n_exp_boosts; | |
179 | /* Number of tasks boosted for expedited GP. */ | |
180 | unsigned long n_normal_boosts; | |
181 | /* Number of tasks boosted for normal GP. */ | |
182 | unsigned long n_balk_blkd_tasks; | |
183 | /* Refused to boost: no blocked tasks. */ | |
184 | unsigned long n_balk_exp_gp_tasks; | |
185 | /* Refused to boost: nothing blocking GP. */ | |
186 | unsigned long n_balk_boost_tasks; | |
187 | /* Refused to boost: already boosting. */ | |
188 | unsigned long n_balk_notblocked; | |
189 | /* Refused to boost: RCU RS CS still running. */ | |
190 | unsigned long n_balk_notyet; | |
191 | /* Refused to boost: not yet time. */ | |
192 | unsigned long n_balk_nos; | |
193 | /* Refused to boost: not sure why, though. */ | |
194 | /* This can happen due to race conditions. */ | |
27f4d280 | 195 | #endif /* #ifdef CONFIG_RCU_BOOST */ |
394f2769 | 196 | raw_spinlock_t fqslock ____cacheline_internodealigned_in_smp; |
9f77da9f PM |
197 | } ____cacheline_internodealigned_in_smp; |
198 | ||
a0b6c9a7 PM |
199 | /* |
200 | * Do a full breadth-first scan of the rcu_node structures for the | |
201 | * specified rcu_state structure. | |
202 | */ | |
203 | #define rcu_for_each_node_breadth_first(rsp, rnp) \ | |
204 | for ((rnp) = &(rsp)->node[0]; \ | |
f885b7f2 | 205 | (rnp) < &(rsp)->node[rcu_num_nodes]; (rnp)++) |
a0b6c9a7 | 206 | |
d9a3da06 PM |
207 | /* |
208 | * Do a breadth-first scan of the non-leaf rcu_node structures for the | |
209 | * specified rcu_state structure. Note that if there is a singleton | |
210 | * rcu_node tree with but one rcu_node structure, this loop is a no-op. | |
211 | */ | |
212 | #define rcu_for_each_nonleaf_node_breadth_first(rsp, rnp) \ | |
213 | for ((rnp) = &(rsp)->node[0]; \ | |
f885b7f2 | 214 | (rnp) < (rsp)->level[rcu_num_lvls - 1]; (rnp)++) |
d9a3da06 PM |
215 | |
216 | /* | |
217 | * Scan the leaves of the rcu_node hierarchy for the specified rcu_state | |
218 | * structure. Note that if there is a singleton rcu_node tree with but | |
219 | * one rcu_node structure, this loop -will- visit the rcu_node structure. | |
220 | * It is still a leaf node, even if it is also the root node. | |
221 | */ | |
a0b6c9a7 | 222 | #define rcu_for_each_leaf_node(rsp, rnp) \ |
f885b7f2 PM |
223 | for ((rnp) = (rsp)->level[rcu_num_lvls - 1]; \ |
224 | (rnp) < &(rsp)->node[rcu_num_nodes]; (rnp)++) | |
a0b6c9a7 | 225 | |
9f77da9f PM |
226 | /* Index values for nxttail array in struct rcu_data. */ |
227 | #define RCU_DONE_TAIL 0 /* Also RCU_WAIT head. */ | |
228 | #define RCU_WAIT_TAIL 1 /* Also RCU_NEXT_READY head. */ | |
229 | #define RCU_NEXT_READY_TAIL 2 /* Also RCU_NEXT head. */ | |
230 | #define RCU_NEXT_TAIL 3 | |
231 | #define RCU_NEXT_SIZE 4 | |
232 | ||
233 | /* Per-CPU data for read-copy update. */ | |
234 | struct rcu_data { | |
235 | /* 1) quiescent-state and grace-period handling : */ | |
20133cfc | 236 | unsigned long completed; /* Track rsp->completed gp number */ |
9f77da9f | 237 | /* in order to detect GP end. */ |
20133cfc | 238 | unsigned long gpnum; /* Highest gp number that this CPU */ |
9f77da9f | 239 | /* is aware of having started. */ |
e4cc1f22 | 240 | bool passed_quiesce; /* User-mode/idle loop etc. */ |
9f77da9f PM |
241 | bool qs_pending; /* Core waits for quiesc state. */ |
242 | bool beenonline; /* CPU online at least once. */ | |
6cc68793 | 243 | bool preemptible; /* Preemptible RCU? */ |
9f77da9f PM |
244 | struct rcu_node *mynode; /* This CPU's leaf of hierarchy */ |
245 | unsigned long grpmask; /* Mask to apply to leaf qsmask. */ | |
a858af28 PM |
246 | #ifdef CONFIG_RCU_CPU_STALL_INFO |
247 | unsigned long ticks_this_gp; /* The number of scheduling-clock */ | |
248 | /* ticks this CPU has handled */ | |
249 | /* during and after the last grace */ | |
250 | /* period it is aware of. */ | |
251 | #endif /* #ifdef CONFIG_RCU_CPU_STALL_INFO */ | |
9f77da9f PM |
252 | |
253 | /* 2) batch handling */ | |
254 | /* | |
255 | * If nxtlist is not NULL, it is partitioned as follows. | |
256 | * Any of the partitions might be empty, in which case the | |
257 | * pointer to that partition will be equal to the pointer for | |
258 | * the following partition. When the list is empty, all of | |
1eba8f84 PM |
259 | * the nxttail elements point to the ->nxtlist pointer itself, |
260 | * which in that case is NULL. | |
9f77da9f | 261 | * |
9f77da9f PM |
262 | * [nxtlist, *nxttail[RCU_DONE_TAIL]): |
263 | * Entries that batch # <= ->completed | |
264 | * The grace period for these entries has completed, and | |
265 | * the other grace-period-completed entries may be moved | |
266 | * here temporarily in rcu_process_callbacks(). | |
1eba8f84 PM |
267 | * [*nxttail[RCU_DONE_TAIL], *nxttail[RCU_WAIT_TAIL]): |
268 | * Entries that batch # <= ->completed - 1: waiting for current GP | |
269 | * [*nxttail[RCU_WAIT_TAIL], *nxttail[RCU_NEXT_READY_TAIL]): | |
270 | * Entries known to have arrived before current GP ended | |
271 | * [*nxttail[RCU_NEXT_READY_TAIL], *nxttail[RCU_NEXT_TAIL]): | |
272 | * Entries that might have arrived after current GP ended | |
273 | * Note that the value of *nxttail[RCU_NEXT_TAIL] will | |
274 | * always be NULL, as this is the end of the list. | |
9f77da9f PM |
275 | */ |
276 | struct rcu_head *nxtlist; | |
277 | struct rcu_head **nxttail[RCU_NEXT_SIZE]; | |
dc35c893 PM |
278 | unsigned long nxtcompleted[RCU_NEXT_SIZE]; |
279 | /* grace periods for sublists. */ | |
486e2593 PM |
280 | long qlen_lazy; /* # of lazy queued callbacks */ |
281 | long qlen; /* # of queued callbacks, incl lazy */ | |
37c72e56 PM |
282 | long qlen_last_fqs_check; |
283 | /* qlen at last check for QS forcing */ | |
269dcc1c | 284 | unsigned long n_cbs_invoked; /* count of RCU cbs invoked. */ |
c635a4e1 | 285 | unsigned long n_nocbs_invoked; /* count of no-CBs RCU cbs invoked. */ |
29494be7 LJ |
286 | unsigned long n_cbs_orphaned; /* RCU cbs orphaned by dying CPU */ |
287 | unsigned long n_cbs_adopted; /* RCU cbs adopted from dying CPU */ | |
37c72e56 PM |
288 | unsigned long n_force_qs_snap; |
289 | /* did other CPU force QS recently? */ | |
9f77da9f PM |
290 | long blimit; /* Upper limit on a processed batch */ |
291 | ||
9f77da9f PM |
292 | /* 3) dynticks interface. */ |
293 | struct rcu_dynticks *dynticks; /* Shared per-CPU dynticks state. */ | |
294 | int dynticks_snap; /* Per-GP tracking for dynticks. */ | |
9f77da9f PM |
295 | |
296 | /* 4) reasons this CPU needed to be kicked by force_quiescent_state */ | |
9f77da9f | 297 | unsigned long dynticks_fqs; /* Kicked due to dynticks idle. */ |
9f77da9f | 298 | unsigned long offline_fqs; /* Kicked due to being offline. */ |
9f77da9f PM |
299 | |
300 | /* 5) __rcu_pending() statistics. */ | |
20133cfc PM |
301 | unsigned long n_rcu_pending; /* rcu_pending() calls since boot. */ |
302 | unsigned long n_rp_qs_pending; | |
d21670ac | 303 | unsigned long n_rp_report_qs; |
20133cfc PM |
304 | unsigned long n_rp_cb_ready; |
305 | unsigned long n_rp_cpu_needs_gp; | |
306 | unsigned long n_rp_gp_completed; | |
307 | unsigned long n_rp_gp_started; | |
20133cfc | 308 | unsigned long n_rp_need_nothing; |
9f77da9f | 309 | |
b626c1b6 | 310 | /* 6) _rcu_barrier() and OOM callbacks. */ |
06668efa | 311 | struct rcu_head barrier_head; |
b626c1b6 PM |
312 | #ifdef CONFIG_RCU_FAST_NO_HZ |
313 | struct rcu_head oom_head; | |
314 | #endif /* #ifdef CONFIG_RCU_FAST_NO_HZ */ | |
06668efa | 315 | |
3fbfbf7a PM |
316 | /* 7) Callback offloading. */ |
317 | #ifdef CONFIG_RCU_NOCB_CPU | |
318 | struct rcu_head *nocb_head; /* CBs waiting for kthread. */ | |
319 | struct rcu_head **nocb_tail; | |
320 | atomic_long_t nocb_q_count; /* # CBs waiting for kthread */ | |
321 | atomic_long_t nocb_q_count_lazy; /* (approximate). */ | |
322 | int nocb_p_count; /* # CBs being invoked by kthread */ | |
323 | int nocb_p_count_lazy; /* (approximate). */ | |
324 | wait_queue_head_t nocb_wq; /* For nocb kthreads to sleep on. */ | |
325 | struct task_struct *nocb_kthread; | |
326 | #endif /* #ifdef CONFIG_RCU_NOCB_CPU */ | |
327 | ||
9f77da9f | 328 | int cpu; |
d4c08f2a | 329 | struct rcu_state *rsp; |
9f77da9f PM |
330 | }; |
331 | ||
af446b70 | 332 | /* Values for fqs_state field in struct rcu_state. */ |
83f5b01f PM |
333 | #define RCU_GP_IDLE 0 /* No grace period in progress. */ |
334 | #define RCU_GP_INIT 1 /* Grace period being initialized. */ | |
335 | #define RCU_SAVE_DYNTICK 2 /* Need to scan dyntick state. */ | |
ee47eb9f | 336 | #define RCU_FORCE_QS 3 /* Need to force quiescent state. */ |
9f77da9f | 337 | #define RCU_SIGNAL_INIT RCU_SAVE_DYNTICK |
9f77da9f PM |
338 | |
339 | #define RCU_JIFFIES_TILL_FORCE_QS 3 /* for rsp->jiffies_force_qs */ | |
007b0924 | 340 | |
007b0924 PM |
341 | #define RCU_STALL_RAT_DELAY 2 /* Allow other CPUs time */ |
342 | /* to take at least one */ | |
343 | /* scheduling clock irq */ | |
344 | /* before ratting on them. */ | |
9f77da9f | 345 | |
08bca60a PZ |
346 | #define rcu_wait(cond) \ |
347 | do { \ | |
348 | for (;;) { \ | |
349 | set_current_state(TASK_INTERRUPTIBLE); \ | |
350 | if (cond) \ | |
351 | break; \ | |
352 | schedule(); \ | |
353 | } \ | |
354 | __set_current_state(TASK_RUNNING); \ | |
355 | } while (0) | |
9f77da9f PM |
356 | |
357 | /* | |
358 | * RCU global state, including node hierarchy. This hierarchy is | |
359 | * represented in "heap" form in a dense array. The root (first level) | |
360 | * of the hierarchy is in ->node[0] (referenced by ->level[0]), the second | |
361 | * level in ->node[1] through ->node[m] (->node[1] referenced by ->level[1]), | |
362 | * and the third level in ->node[m+1] and following (->node[m+1] referenced | |
363 | * by ->level[2]). The number of levels is determined by the number of | |
364 | * CPUs and by CONFIG_RCU_FANOUT. Small systems will have a "hierarchy" | |
365 | * consisting of a single rcu_node. | |
366 | */ | |
367 | struct rcu_state { | |
368 | struct rcu_node node[NUM_RCU_NODES]; /* Hierarchy. */ | |
f885b7f2 | 369 | struct rcu_node *level[RCU_NUM_LVLS]; /* Hierarchy levels. */ |
9f77da9f | 370 | u32 levelcnt[MAX_RCU_LVLS + 1]; /* # nodes in each level. */ |
f885b7f2 | 371 | u8 levelspread[RCU_NUM_LVLS]; /* kids/node in each level. */ |
394f99a9 | 372 | struct rcu_data __percpu *rda; /* pointer of percu rcu_data. */ |
037b64ed PM |
373 | void (*call)(struct rcu_head *head, /* call_rcu() flavor. */ |
374 | void (*func)(struct rcu_head *head)); | |
3fbfbf7a PM |
375 | #ifdef CONFIG_RCU_NOCB_CPU |
376 | void (*call_remote)(struct rcu_head *head, | |
377 | void (*func)(struct rcu_head *head)); | |
378 | /* call_rcu() flavor, but for */ | |
379 | /* placing on remote CPU. */ | |
380 | #endif /* #ifdef CONFIG_RCU_NOCB_CPU */ | |
9f77da9f PM |
381 | |
382 | /* The following fields are guarded by the root rcu_node's lock. */ | |
383 | ||
af446b70 | 384 | u8 fqs_state ____cacheline_internodealigned_in_smp; |
9f77da9f | 385 | /* Force QS state. */ |
a46e0899 | 386 | u8 boost; /* Subject to priority boost. */ |
20133cfc PM |
387 | unsigned long gpnum; /* Current gp number. */ |
388 | unsigned long completed; /* # of last completed gp. */ | |
b3dbec76 PM |
389 | struct task_struct *gp_kthread; /* Task for grace periods. */ |
390 | wait_queue_head_t gp_wq; /* Where GP task waits. */ | |
391 | int gp_flags; /* Commands for GP task. */ | |
1eba8f84 | 392 | |
d9a3da06 | 393 | /* End of fields guarded by root rcu_node's lock. */ |
1eba8f84 | 394 | |
7b2e6011 PM |
395 | raw_spinlock_t orphan_lock ____cacheline_internodealigned_in_smp; |
396 | /* Protect following fields. */ | |
b1420f1c PM |
397 | struct rcu_head *orphan_nxtlist; /* Orphaned callbacks that */ |
398 | /* need a grace period. */ | |
399 | struct rcu_head **orphan_nxttail; /* Tail of above. */ | |
400 | struct rcu_head *orphan_donelist; /* Orphaned callbacks that */ | |
401 | /* are ready to invoke. */ | |
402 | struct rcu_head **orphan_donetail; /* Tail of above. */ | |
403 | long qlen_lazy; /* Number of lazy callbacks. */ | |
404 | long qlen; /* Total number of callbacks. */ | |
7b2e6011 | 405 | /* End of fields guarded by orphan_lock. */ |
a4fbe35a PM |
406 | |
407 | struct mutex onoff_mutex; /* Coordinate hotplug & GPs. */ | |
408 | ||
7be7f0be | 409 | struct mutex barrier_mutex; /* Guards barrier fields. */ |
24ebbca8 | 410 | atomic_t barrier_cpu_count; /* # CPUs waiting on. */ |
7db74df8 | 411 | struct completion barrier_completion; /* Wake at barrier end. */ |
cf3a9c48 PM |
412 | unsigned long n_barrier_done; /* ++ at start and end of */ |
413 | /* _rcu_barrier(). */ | |
a4fbe35a PM |
414 | /* End of fields guarded by barrier_mutex. */ |
415 | ||
40694d66 PM |
416 | atomic_long_t expedited_start; /* Starting ticket. */ |
417 | atomic_long_t expedited_done; /* Done ticket. */ | |
a30489c5 PM |
418 | atomic_long_t expedited_wrap; /* # near-wrap incidents. */ |
419 | atomic_long_t expedited_tryfail; /* # acquisition failures. */ | |
420 | atomic_long_t expedited_workdone1; /* # done by others #1. */ | |
421 | atomic_long_t expedited_workdone2; /* # done by others #2. */ | |
422 | atomic_long_t expedited_normal; /* # fallbacks to normal. */ | |
423 | atomic_long_t expedited_stoppedcpus; /* # successful stop_cpus. */ | |
424 | atomic_long_t expedited_done_tries; /* # tries to update _done. */ | |
425 | atomic_long_t expedited_done_lost; /* # times beaten to _done. */ | |
426 | atomic_long_t expedited_done_exit; /* # times exited _done loop. */ | |
40694d66 | 427 | |
9f77da9f PM |
428 | unsigned long jiffies_force_qs; /* Time at which to invoke */ |
429 | /* force_quiescent_state(). */ | |
430 | unsigned long n_force_qs; /* Number of calls to */ | |
431 | /* force_quiescent_state(). */ | |
432 | unsigned long n_force_qs_lh; /* ~Number of calls leaving */ | |
433 | /* due to lock unavailable. */ | |
434 | unsigned long n_force_qs_ngp; /* Number of calls leaving */ | |
435 | /* due to no GP active. */ | |
9f77da9f PM |
436 | unsigned long gp_start; /* Time at which GP started, */ |
437 | /* but in jiffies. */ | |
438 | unsigned long jiffies_stall; /* Time at which to check */ | |
439 | /* for CPU stalls. */ | |
15ba0ba8 PM |
440 | unsigned long gp_max; /* Maximum GP duration in */ |
441 | /* jiffies. */ | |
4300aa64 | 442 | char *name; /* Name of structure. */ |
6ce75a23 | 443 | struct list_head flavors; /* List of RCU flavors. */ |
9f77da9f PM |
444 | }; |
445 | ||
4cdfc175 PM |
446 | /* Values for rcu_state structure's gp_flags field. */ |
447 | #define RCU_GP_FLAG_INIT 0x1 /* Need grace-period initialization. */ | |
448 | #define RCU_GP_FLAG_FQS 0x2 /* Need grace-period quiescent-state forcing. */ | |
449 | ||
6ce75a23 | 450 | extern struct list_head rcu_struct_flavors; |
3fbfbf7a PM |
451 | |
452 | /* Sequence through rcu_state structures for each RCU flavor. */ | |
6ce75a23 PM |
453 | #define for_each_rcu_flavor(rsp) \ |
454 | list_for_each_entry((rsp), &rcu_struct_flavors, flavors) | |
455 | ||
d9a3da06 PM |
456 | /* Return values for rcu_preempt_offline_tasks(). */ |
457 | ||
458 | #define RCU_OFL_TASKS_NORM_GP 0x1 /* Tasks blocking normal */ | |
459 | /* GP were moved to root. */ | |
460 | #define RCU_OFL_TASKS_EXP_GP 0x2 /* Tasks blocking expedited */ | |
461 | /* GP were moved to root. */ | |
462 | ||
6258c4fb IM |
463 | /* |
464 | * RCU implementation internal declarations: | |
465 | */ | |
d6714c22 PM |
466 | extern struct rcu_state rcu_sched_state; |
467 | DECLARE_PER_CPU(struct rcu_data, rcu_sched_data); | |
6258c4fb IM |
468 | |
469 | extern struct rcu_state rcu_bh_state; | |
470 | DECLARE_PER_CPU(struct rcu_data, rcu_bh_data); | |
471 | ||
f41d911f PM |
472 | #ifdef CONFIG_TREE_PREEMPT_RCU |
473 | extern struct rcu_state rcu_preempt_state; | |
474 | DECLARE_PER_CPU(struct rcu_data, rcu_preempt_data); | |
475 | #endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */ | |
476 | ||
eab0993c PM |
477 | #ifdef CONFIG_RCU_BOOST |
478 | DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_status); | |
479 | DECLARE_PER_CPU(int, rcu_cpu_kthread_cpu); | |
480 | DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_loops); | |
481 | DECLARE_PER_CPU(char, rcu_cpu_has_work); | |
482 | #endif /* #ifdef CONFIG_RCU_BOOST */ | |
483 | ||
017c4261 | 484 | #ifndef RCU_TREE_NONCORE |
9f77da9f | 485 | |
9b2619af | 486 | /* Forward declarations for rcutree_plugin.h */ |
dbe01350 | 487 | static void rcu_bootup_announce(void); |
9b2619af | 488 | long rcu_batches_completed(void); |
cba6d0d6 | 489 | static void rcu_preempt_note_context_switch(int cpu); |
27f4d280 | 490 | static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp); |
b668c9cf | 491 | #ifdef CONFIG_HOTPLUG_CPU |
d3f6bad3 PM |
492 | static void rcu_report_unblock_qs_rnp(struct rcu_node *rnp, |
493 | unsigned long flags); | |
b668c9cf | 494 | #endif /* #ifdef CONFIG_HOTPLUG_CPU */ |
1ed509a2 | 495 | static void rcu_print_detail_task_stall(struct rcu_state *rsp); |
9bc8b558 | 496 | static int rcu_print_task_stall(struct rcu_node *rnp); |
9b2619af PM |
497 | static void rcu_preempt_check_blocked_tasks(struct rcu_node *rnp); |
498 | #ifdef CONFIG_HOTPLUG_CPU | |
237c80c5 PM |
499 | static int rcu_preempt_offline_tasks(struct rcu_state *rsp, |
500 | struct rcu_node *rnp, | |
501 | struct rcu_data *rdp); | |
9b2619af PM |
502 | #endif /* #ifdef CONFIG_HOTPLUG_CPU */ |
503 | static void rcu_preempt_check_callbacks(int cpu); | |
9b2619af | 504 | void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu)); |
d9a3da06 | 505 | #if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_TREE_PREEMPT_RCU) |
b40d293e TG |
506 | static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp, |
507 | bool wake); | |
d9a3da06 | 508 | #endif /* #if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_TREE_PREEMPT_RCU) */ |
9b2619af | 509 | static void __init __rcu_init_preempt(void); |
1217ed1b | 510 | static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags); |
a46e0899 PM |
511 | static void rcu_preempt_boost_start_gp(struct rcu_node *rnp); |
512 | static void invoke_rcu_callbacks_kthread(void); | |
dff1672d | 513 | static bool rcu_is_callbacks_kthread(void); |
a46e0899 PM |
514 | #ifdef CONFIG_RCU_BOOST |
515 | static void rcu_preempt_do_callbacks(void); | |
27f4d280 | 516 | static int __cpuinit rcu_spawn_one_boost_kthread(struct rcu_state *rsp, |
5d01bbd1 | 517 | struct rcu_node *rnp); |
a46e0899 | 518 | #endif /* #ifdef CONFIG_RCU_BOOST */ |
f8b7fc6b | 519 | static void __cpuinit rcu_prepare_kthreads(int cpu); |
7cb92499 PM |
520 | static void rcu_prepare_for_idle_init(int cpu); |
521 | static void rcu_cleanup_after_idle(int cpu); | |
aea1b35e | 522 | static void rcu_prepare_for_idle(int cpu); |
c57afe80 | 523 | static void rcu_idle_count_callbacks_posted(void); |
a858af28 PM |
524 | static void print_cpu_stall_info_begin(void); |
525 | static void print_cpu_stall_info(struct rcu_state *rsp, int cpu); | |
526 | static void print_cpu_stall_info_end(void); | |
527 | static void zero_cpu_stall_ticks(struct rcu_data *rdp); | |
528 | static void increment_cpu_stall_ticks(void); | |
3fbfbf7a PM |
529 | static bool is_nocb_cpu(int cpu); |
530 | static bool __call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *rhp, | |
531 | bool lazy); | |
532 | static bool rcu_nocb_adopt_orphan_cbs(struct rcu_state *rsp, | |
533 | struct rcu_data *rdp); | |
534 | static bool nocb_cpu_expendable(int cpu); | |
535 | static void rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp); | |
536 | static void rcu_spawn_nocb_kthreads(struct rcu_state *rsp); | |
537 | static void init_nocb_callback_list(struct rcu_data *rdp); | |
538 | static void __init rcu_init_nocb(void); | |
9b2619af | 539 | |
017c4261 | 540 | #endif /* #ifndef RCU_TREE_NONCORE */ |
3fbfbf7a PM |
541 | |
542 | #ifdef CONFIG_RCU_TRACE | |
543 | #ifdef CONFIG_RCU_NOCB_CPU | |
544 | /* Sum up queue lengths for tracing. */ | |
545 | static inline void rcu_nocb_q_lengths(struct rcu_data *rdp, long *ql, long *qll) | |
546 | { | |
547 | *ql = atomic_long_read(&rdp->nocb_q_count) + rdp->nocb_p_count; | |
548 | *qll = atomic_long_read(&rdp->nocb_q_count_lazy) + rdp->nocb_p_count_lazy; | |
549 | } | |
550 | #else /* #ifdef CONFIG_RCU_NOCB_CPU */ | |
551 | static inline void rcu_nocb_q_lengths(struct rcu_data *rdp, long *ql, long *qll) | |
552 | { | |
553 | *ql = 0; | |
554 | *qll = 0; | |
555 | } | |
556 | #endif /* #else #ifdef CONFIG_RCU_NOCB_CPU */ | |
557 | #endif /* #ifdef CONFIG_RCU_TRACE */ |