rcu: Clean up code based on review feedback from Josh Triplett, part 2
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / kernel / rcutree.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>
30
31/*
32 * Define shape of hierarchy based on NR_CPUS and CONFIG_RCU_FANOUT.
33 * In theory, it should be possible to add more levels straightforwardly.
34 * In practice, this has not been tested, so there is probably some
35 * bug somewhere.
36 */
37#define MAX_RCU_LVLS 3
38#define RCU_FANOUT (CONFIG_RCU_FANOUT)
39#define RCU_FANOUT_SQ (RCU_FANOUT * RCU_FANOUT)
40#define RCU_FANOUT_CUBE (RCU_FANOUT_SQ * RCU_FANOUT)
41
42#if NR_CPUS <= RCU_FANOUT
43# define NUM_RCU_LVLS 1
44# define NUM_RCU_LVL_0 1
45# define NUM_RCU_LVL_1 (NR_CPUS)
46# define NUM_RCU_LVL_2 0
47# define NUM_RCU_LVL_3 0
48#elif NR_CPUS <= RCU_FANOUT_SQ
49# define NUM_RCU_LVLS 2
50# define NUM_RCU_LVL_0 1
fc2219d4 51# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
9f77da9f
PM
52# define NUM_RCU_LVL_2 (NR_CPUS)
53# define NUM_RCU_LVL_3 0
54#elif NR_CPUS <= RCU_FANOUT_CUBE
55# define NUM_RCU_LVLS 3
56# define NUM_RCU_LVL_0 1
fc2219d4
PM
57# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ)
58# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
9f77da9f
PM
59# define NUM_RCU_LVL_3 NR_CPUS
60#else
61# error "CONFIG_RCU_FANOUT insufficient for NR_CPUS"
62#endif /* #if (NR_CPUS) <= RCU_FANOUT */
63
64#define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3)
65#define NUM_RCU_NODES (RCU_SUM - NR_CPUS)
66
67/*
68 * Dynticks per-CPU state.
69 */
70struct rcu_dynticks {
71 int dynticks_nesting; /* Track nesting level, sort of. */
72 int dynticks; /* Even value for dynticks-idle, else odd. */
73 int dynticks_nmi; /* Even value for either dynticks-idle or */
74 /* not in nmi handler, else odd. So this */
75 /* remains even for nmi from irq handler. */
76};
77
78/*
79 * Definition for node within the RCU grace-period-detection hierarchy.
80 */
81struct rcu_node {
1eba8f84
PM
82 spinlock_t lock; /* Root rcu_node's lock protects some */
83 /* rcu_state fields as well as following. */
f41d911f 84 long gpnum; /* Current grace period for this node. */
86848966
PM
85 /* This will either be equal to or one */
86 /* behind the root rcu_node's gpnum. */
9f77da9f
PM
87 unsigned long qsmask; /* CPUs or groups that need to switch in */
88 /* order for current grace period to proceed.*/
1eba8f84
PM
89 /* In leaf rcu_node, each bit corresponds to */
90 /* an rcu_data structure, otherwise, each */
91 /* bit corresponds to a child rcu_node */
92 /* structure. */
9f77da9f
PM
93 unsigned long qsmaskinit;
94 /* Per-GP initialization for qsmask. */
95 unsigned long grpmask; /* Mask to apply to parent qsmask. */
1eba8f84 96 /* Only one bit will be set in this mask. */
9f77da9f
PM
97 int grplo; /* lowest-numbered CPU or group here. */
98 int grphi; /* highest-numbered CPU or group here. */
99 u8 grpnum; /* CPU/group number for next level up. */
100 u8 level; /* root is at level 0. */
101 struct rcu_node *parent;
f41d911f
PM
102 struct list_head blocked_tasks[2];
103 /* Tasks blocked in RCU read-side critsect. */
1eba8f84
PM
104 /* Grace period number (->gpnum) x blocked */
105 /* by tasks on the (x & 0x1) element of the */
106 /* blocked_tasks[] array. */
9f77da9f
PM
107} ____cacheline_internodealigned_in_smp;
108
109/* Index values for nxttail array in struct rcu_data. */
110#define RCU_DONE_TAIL 0 /* Also RCU_WAIT head. */
111#define RCU_WAIT_TAIL 1 /* Also RCU_NEXT_READY head. */
112#define RCU_NEXT_READY_TAIL 2 /* Also RCU_NEXT head. */
113#define RCU_NEXT_TAIL 3
114#define RCU_NEXT_SIZE 4
115
116/* Per-CPU data for read-copy update. */
117struct rcu_data {
118 /* 1) quiescent-state and grace-period handling : */
119 long completed; /* Track rsp->completed gp number */
120 /* in order to detect GP end. */
121 long gpnum; /* Highest gp number that this CPU */
122 /* is aware of having started. */
123 long passed_quiesc_completed;
124 /* Value of completed at time of qs. */
125 bool passed_quiesc; /* User-mode/idle loop etc. */
126 bool qs_pending; /* Core waits for quiesc state. */
127 bool beenonline; /* CPU online at least once. */
f41d911f 128 bool preemptable; /* Preemptable RCU? */
9f77da9f
PM
129 struct rcu_node *mynode; /* This CPU's leaf of hierarchy */
130 unsigned long grpmask; /* Mask to apply to leaf qsmask. */
131
132 /* 2) batch handling */
133 /*
134 * If nxtlist is not NULL, it is partitioned as follows.
135 * Any of the partitions might be empty, in which case the
136 * pointer to that partition will be equal to the pointer for
137 * the following partition. When the list is empty, all of
1eba8f84
PM
138 * the nxttail elements point to the ->nxtlist pointer itself,
139 * which in that case is NULL.
9f77da9f 140 *
9f77da9f
PM
141 * [nxtlist, *nxttail[RCU_DONE_TAIL]):
142 * Entries that batch # <= ->completed
143 * The grace period for these entries has completed, and
144 * the other grace-period-completed entries may be moved
145 * here temporarily in rcu_process_callbacks().
1eba8f84
PM
146 * [*nxttail[RCU_DONE_TAIL], *nxttail[RCU_WAIT_TAIL]):
147 * Entries that batch # <= ->completed - 1: waiting for current GP
148 * [*nxttail[RCU_WAIT_TAIL], *nxttail[RCU_NEXT_READY_TAIL]):
149 * Entries known to have arrived before current GP ended
150 * [*nxttail[RCU_NEXT_READY_TAIL], *nxttail[RCU_NEXT_TAIL]):
151 * Entries that might have arrived after current GP ended
152 * Note that the value of *nxttail[RCU_NEXT_TAIL] will
153 * always be NULL, as this is the end of the list.
9f77da9f
PM
154 */
155 struct rcu_head *nxtlist;
156 struct rcu_head **nxttail[RCU_NEXT_SIZE];
a71fca58 157 long qlen; /* # of queued callbacks */
9f77da9f
PM
158 long blimit; /* Upper limit on a processed batch */
159
160#ifdef CONFIG_NO_HZ
161 /* 3) dynticks interface. */
162 struct rcu_dynticks *dynticks; /* Shared per-CPU dynticks state. */
163 int dynticks_snap; /* Per-GP tracking for dynticks. */
164 int dynticks_nmi_snap; /* Per-GP tracking for dynticks_nmi. */
165#endif /* #ifdef CONFIG_NO_HZ */
166
167 /* 4) reasons this CPU needed to be kicked by force_quiescent_state */
168#ifdef CONFIG_NO_HZ
169 unsigned long dynticks_fqs; /* Kicked due to dynticks idle. */
170#endif /* #ifdef CONFIG_NO_HZ */
171 unsigned long offline_fqs; /* Kicked due to being offline. */
172 unsigned long resched_ipi; /* Sent a resched IPI. */
173
174 /* 5) __rcu_pending() statistics. */
175 long n_rcu_pending; /* rcu_pending() calls since boot. */
176 long n_rp_qs_pending;
177 long n_rp_cb_ready;
178 long n_rp_cpu_needs_gp;
179 long n_rp_gp_completed;
180 long n_rp_gp_started;
181 long n_rp_need_fqs;
182 long n_rp_need_nothing;
183
184 int cpu;
185};
186
187/* Values for signaled field in struct rcu_state. */
188#define RCU_GP_INIT 0 /* Grace period being initialized. */
189#define RCU_SAVE_DYNTICK 1 /* Need to scan dyntick state. */
190#define RCU_FORCE_QS 2 /* Need to force quiescent state. */
191#ifdef CONFIG_NO_HZ
192#define RCU_SIGNAL_INIT RCU_SAVE_DYNTICK
193#else /* #ifdef CONFIG_NO_HZ */
194#define RCU_SIGNAL_INIT RCU_FORCE_QS
195#endif /* #else #ifdef CONFIG_NO_HZ */
196
197#define RCU_JIFFIES_TILL_FORCE_QS 3 /* for rsp->jiffies_force_qs */
198#ifdef CONFIG_RCU_CPU_STALL_DETECTOR
199#define RCU_SECONDS_TILL_STALL_CHECK (10 * HZ) /* for rsp->jiffies_stall */
200#define RCU_SECONDS_TILL_STALL_RECHECK (30 * HZ) /* for rsp->jiffies_stall */
201#define RCU_STALL_RAT_DELAY 2 /* Allow other CPUs time */
202 /* to take at least one */
203 /* scheduling clock irq */
204 /* before ratting on them. */
205
206#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
207
208/*
209 * RCU global state, including node hierarchy. This hierarchy is
210 * represented in "heap" form in a dense array. The root (first level)
211 * of the hierarchy is in ->node[0] (referenced by ->level[0]), the second
212 * level in ->node[1] through ->node[m] (->node[1] referenced by ->level[1]),
213 * and the third level in ->node[m+1] and following (->node[m+1] referenced
214 * by ->level[2]). The number of levels is determined by the number of
215 * CPUs and by CONFIG_RCU_FANOUT. Small systems will have a "hierarchy"
216 * consisting of a single rcu_node.
217 */
218struct rcu_state {
219 struct rcu_node node[NUM_RCU_NODES]; /* Hierarchy. */
220 struct rcu_node *level[NUM_RCU_LVLS]; /* Hierarchy levels. */
221 u32 levelcnt[MAX_RCU_LVLS + 1]; /* # nodes in each level. */
222 u8 levelspread[NUM_RCU_LVLS]; /* kids/node in each level. */
223 struct rcu_data *rda[NR_CPUS]; /* array of rdp pointers. */
224
225 /* The following fields are guarded by the root rcu_node's lock. */
226
227 u8 signaled ____cacheline_internodealigned_in_smp;
228 /* Force QS state. */
229 long gpnum; /* Current gp number. */
230 long completed; /* # of last completed gp. */
1eba8f84
PM
231
232 /* End of fields guarded by root rcu_node's lock. */
233
9f77da9f
PM
234 spinlock_t onofflock; /* exclude on/offline and */
235 /* starting new GP. */
236 spinlock_t fqslock; /* Only one task forcing */
237 /* quiescent states. */
238 unsigned long jiffies_force_qs; /* Time at which to invoke */
239 /* force_quiescent_state(). */
240 unsigned long n_force_qs; /* Number of calls to */
241 /* force_quiescent_state(). */
242 unsigned long n_force_qs_lh; /* ~Number of calls leaving */
243 /* due to lock unavailable. */
244 unsigned long n_force_qs_ngp; /* Number of calls leaving */
245 /* due to no GP active. */
246#ifdef CONFIG_RCU_CPU_STALL_DETECTOR
247 unsigned long gp_start; /* Time at which GP started, */
248 /* but in jiffies. */
249 unsigned long jiffies_stall; /* Time at which to check */
250 /* for CPU stalls. */
251#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
252#ifdef CONFIG_NO_HZ
253 long dynticks_completed; /* Value of completed @ snap. */
254#endif /* #ifdef CONFIG_NO_HZ */
255};
256
257#ifdef RCU_TREE_NONCORE
6258c4fb
IM
258
259/*
260 * RCU implementation internal declarations:
261 */
d6714c22
PM
262extern struct rcu_state rcu_sched_state;
263DECLARE_PER_CPU(struct rcu_data, rcu_sched_data);
6258c4fb
IM
264
265extern struct rcu_state rcu_bh_state;
266DECLARE_PER_CPU(struct rcu_data, rcu_bh_data);
267
f41d911f
PM
268#ifdef CONFIG_TREE_PREEMPT_RCU
269extern struct rcu_state rcu_preempt_state;
270DECLARE_PER_CPU(struct rcu_data, rcu_preempt_data);
271#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
272
9f77da9f
PM
273#endif /* #ifdef RCU_TREE_NONCORE */
274