Allow forcing the use of sys membarrier
[urcu.git] / include / urcu / static / urcu-bp.h
1 #ifndef _URCU_BP_STATIC_H
2 #define _URCU_BP_STATIC_H
3
4 /*
5 * urcu-bp-static.h
6 *
7 * Userspace RCU header.
8 *
9 * TO BE INCLUDED ONLY IN CODE THAT IS TO BE RECOMPILED ON EACH LIBURCU
10 * RELEASE. See urcu.h for linking dynamically with the userspace rcu library.
11 *
12 * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
13 * Copyright (c) 2009 Paul E. McKenney, IBM Corporation.
14 *
15 * This library is free software; you can redistribute it and/or
16 * modify it under the terms of the GNU Lesser General Public
17 * License as published by the Free Software Foundation; either
18 * version 2.1 of the License, or (at your option) any later version.
19 *
20 * This library is distributed in the hope that it will be useful,
21 * but WITHOUT ANY WARRANTY; without even the implied warranty of
22 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
23 * Lesser General Public License for more details.
24 *
25 * You should have received a copy of the GNU Lesser General Public
26 * License along with this library; if not, write to the Free Software
27 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
28 *
29 * IBM's contributions to this file may be relicensed under LGPLv2 or later.
30 */
31
32 #include <stdlib.h>
33 #include <pthread.h>
34 #include <unistd.h>
35
36 #include <urcu/config.h>
37 #include <urcu/compiler.h>
38 #include <urcu/arch.h>
39 #include <urcu/system.h>
40 #include <urcu/uatomic.h>
41 #include <urcu/list.h>
42 #include <urcu/tls-compat.h>
43 #include <urcu/debug.h>
44
45 /*
46 * This code section can only be included in LGPL 2.1 compatible source code.
47 * See below for the function call wrappers which can be used in code meant to
48 * be only linked with the Userspace RCU library. This comes with a small
49 * performance degradation on the read-side due to the added function calls.
50 * This is required to permit relinking with newer versions of the library.
51 */
52
53 #ifdef __cplusplus
54 extern "C" {
55 #endif
56
57 enum rcu_state {
58 RCU_READER_ACTIVE_CURRENT,
59 RCU_READER_ACTIVE_OLD,
60 RCU_READER_INACTIVE,
61 };
62
63 /*
64 * The trick here is that RCU_GP_CTR_PHASE must be a multiple of 8 so we can use a
65 * full 8-bits, 16-bits or 32-bits bitmask for the lower order bits.
66 */
67 #define RCU_GP_COUNT (1UL << 0)
68 /* Use the amount of bits equal to half of the architecture long size */
69 #define RCU_GP_CTR_PHASE (1UL << (sizeof(long) << 2))
70 #define RCU_GP_CTR_NEST_MASK (RCU_GP_CTR_PHASE - 1)
71
72 /*
73 * Used internally by _rcu_read_lock.
74 */
75 extern void rcu_bp_register(void);
76
77 struct rcu_gp {
78 /*
79 * Global grace period counter.
80 * Contains the current RCU_GP_CTR_PHASE.
81 * Also has a RCU_GP_COUNT of 1, to accelerate the reader fast path.
82 * Written to only by writer with mutex taken.
83 * Read by both writer and readers.
84 */
85 unsigned long ctr;
86 } __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
87
88 extern struct rcu_gp rcu_gp;
89
90 struct rcu_reader {
91 /* Data used by both reader and synchronize_rcu() */
92 unsigned long ctr;
93 /* Data used for registry */
94 struct cds_list_head node __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
95 pthread_t tid;
96 int alloc; /* registry entry allocated */
97 };
98
99 /*
100 * Bulletproof version keeps a pointer to a registry not part of the TLS.
101 * Adds a pointer dereference on the read-side, but won't require to unregister
102 * the reader thread.
103 */
104 extern DECLARE_URCU_TLS(struct rcu_reader *, rcu_reader);
105
106 #ifdef CONFIG_RCU_FORCE_SYS_MEMBARRIER
107 #define urcu_bp_has_sys_membarrier 1
108 #else
109 extern int urcu_bp_has_sys_membarrier;
110 #endif
111
112 static inline void urcu_bp_smp_mb_slave(void)
113 {
114 if (caa_likely(urcu_bp_has_sys_membarrier))
115 cmm_barrier();
116 else
117 cmm_smp_mb();
118 }
119
120 static inline enum rcu_state rcu_reader_state(unsigned long *ctr)
121 {
122 unsigned long v;
123
124 if (ctr == NULL)
125 return RCU_READER_INACTIVE;
126 /*
127 * Make sure both tests below are done on the same version of *value
128 * to insure consistency.
129 */
130 v = CMM_LOAD_SHARED(*ctr);
131 if (!(v & RCU_GP_CTR_NEST_MASK))
132 return RCU_READER_INACTIVE;
133 if (!((v ^ rcu_gp.ctr) & RCU_GP_CTR_PHASE))
134 return RCU_READER_ACTIVE_CURRENT;
135 return RCU_READER_ACTIVE_OLD;
136 }
137
138 /*
139 * Helper for _rcu_read_lock(). The format of rcu_gp.ctr (as well as
140 * the per-thread rcu_reader.ctr) has the upper bits containing a count of
141 * _rcu_read_lock() nesting, and a lower-order bit that contains either zero
142 * or RCU_GP_CTR_PHASE. The smp_mb_slave() ensures that the accesses in
143 * _rcu_read_lock() happen before the subsequent read-side critical section.
144 */
145 static inline void _rcu_read_lock_update(unsigned long tmp)
146 {
147 if (caa_likely(!(tmp & RCU_GP_CTR_NEST_MASK))) {
148 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, _CMM_LOAD_SHARED(rcu_gp.ctr));
149 urcu_bp_smp_mb_slave();
150 } else
151 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, tmp + RCU_GP_COUNT);
152 }
153
154 /*
155 * Enter an RCU read-side critical section.
156 *
157 * The first cmm_barrier() call ensures that the compiler does not reorder
158 * the body of _rcu_read_lock() with a mutex.
159 *
160 * This function and its helper are both less than 10 lines long. The
161 * intent is that this function meets the 10-line criterion in LGPL,
162 * allowing this function to be invoked directly from non-LGPL code.
163 */
164 static inline void _rcu_read_lock(void)
165 {
166 unsigned long tmp;
167
168 if (caa_unlikely(!URCU_TLS(rcu_reader)))
169 rcu_bp_register(); /* If not yet registered. */
170 cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */
171 tmp = URCU_TLS(rcu_reader)->ctr;
172 urcu_assert((tmp & RCU_GP_CTR_NEST_MASK) != RCU_GP_CTR_NEST_MASK);
173 _rcu_read_lock_update(tmp);
174 }
175
176 /*
177 * Exit an RCU read-side critical section. This function is less than
178 * 10 lines of code, and is intended to be usable by non-LGPL code, as
179 * called out in LGPL.
180 */
181 static inline void _rcu_read_unlock(void)
182 {
183 unsigned long tmp;
184
185 tmp = URCU_TLS(rcu_reader)->ctr;
186 urcu_assert(tmp & RCU_GP_CTR_NEST_MASK);
187 /* Finish using rcu before decrementing the pointer. */
188 urcu_bp_smp_mb_slave();
189 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, tmp - RCU_GP_COUNT);
190 cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */
191 }
192
193 /*
194 * Returns whether within a RCU read-side critical section.
195 *
196 * This function is less than 10 lines long. The intent is that this
197 * function meets the 10-line criterion for LGPL, allowing this function
198 * to be invoked directly from non-LGPL code.
199 */
200 static inline int _rcu_read_ongoing(void)
201 {
202 if (caa_unlikely(!URCU_TLS(rcu_reader)))
203 rcu_bp_register(); /* If not yet registered. */
204 return URCU_TLS(rcu_reader)->ctr & RCU_GP_CTR_NEST_MASK;
205 }
206
207 #ifdef __cplusplus
208 }
209 #endif
210
211 #endif /* _URCU_BP_STATIC_H */
This page took 0.040072 seconds and 5 git commands to generate.