Allow forcing the use of sys membarrier
[urcu.git] / include / urcu / static / urcu-bp.h
CommitLineData
fdee2e6d
MD
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 *
a5a9f428
PM
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.
fdee2e6d 11 *
6982d6d7 12 * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
fdee2e6d
MD
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>
fdee2e6d
MD
34#include <unistd.h>
35
d8d9a340 36#include <urcu/config.h>
fdee2e6d
MD
37#include <urcu/compiler.h>
38#include <urcu/arch.h>
39#include <urcu/system.h>
a2e7bf9c 40#include <urcu/uatomic.h>
fdee2e6d 41#include <urcu/list.h>
bd252a04 42#include <urcu/tls-compat.h>
a77f7d82 43#include <urcu/debug.h>
fdee2e6d
MD
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
36bc70a8
MD
53#ifdef __cplusplus
54extern "C" {
55#endif
56
52c75091
MD
57enum rcu_state {
58 RCU_READER_ACTIVE_CURRENT,
59 RCU_READER_ACTIVE_OLD,
60 RCU_READER_INACTIVE,
61};
62
fdee2e6d 63/*
02be5561 64 * The trick here is that RCU_GP_CTR_PHASE must be a multiple of 8 so we can use a
fdee2e6d
MD
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 */
02be5561
MD
69#define RCU_GP_CTR_PHASE (1UL << (sizeof(long) << 2))
70#define RCU_GP_CTR_NEST_MASK (RCU_GP_CTR_PHASE - 1)
fdee2e6d
MD
71
72/*
73 * Used internally by _rcu_read_lock.
74 */
75extern void rcu_bp_register(void);
76
c13c2e55
MD
77struct 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
88extern struct rcu_gp rcu_gp;
fdee2e6d 89
02be5561 90struct rcu_reader {
fdee2e6d 91 /* Data used by both reader and synchronize_rcu() */
52c75091 92 unsigned long ctr;
fdee2e6d 93 /* Data used for registry */
16aa9ee8 94 struct cds_list_head node __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
fdee2e6d
MD
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 */
bd252a04 104extern DECLARE_URCU_TLS(struct rcu_reader *, rcu_reader);
fdee2e6d 105
d8d9a340
MD
106#ifdef CONFIG_RCU_FORCE_SYS_MEMBARRIER
107#define urcu_bp_has_sys_membarrier 1
108#else
f541831e 109extern int urcu_bp_has_sys_membarrier;
d8d9a340 110#endif
f541831e
MD
111
112static 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
52c75091 120static inline enum rcu_state rcu_reader_state(unsigned long *ctr)
fdee2e6d 121{
52c75091 122 unsigned long v;
fdee2e6d 123
52c75091
MD
124 if (ctr == NULL)
125 return RCU_READER_INACTIVE;
fdee2e6d
MD
126 /*
127 * Make sure both tests below are done on the same version of *value
128 * to insure consistency.
129 */
52c75091
MD
130 v = CMM_LOAD_SHARED(*ctr);
131 if (!(v & RCU_GP_CTR_NEST_MASK))
132 return RCU_READER_INACTIVE;
c13c2e55 133 if (!((v ^ rcu_gp.ctr) & RCU_GP_CTR_PHASE))
52c75091
MD
134 return RCU_READER_ACTIVE_CURRENT;
135 return RCU_READER_ACTIVE_OLD;
fdee2e6d
MD
136}
137
a5a9f428 138/*
c13c2e55 139 * Helper for _rcu_read_lock(). The format of rcu_gp.ctr (as well as
a5a9f428
PM
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 */
145static inline void _rcu_read_lock_update(unsigned long tmp)
146{
147 if (caa_likely(!(tmp & RCU_GP_CTR_NEST_MASK))) {
c13c2e55 148 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, _CMM_LOAD_SHARED(rcu_gp.ctr));
f541831e 149 urcu_bp_smp_mb_slave();
a5a9f428
PM
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 */
fdee2e6d
MD
164static inline void _rcu_read_lock(void)
165{
52c75091 166 unsigned long tmp;
fdee2e6d 167
bd252a04 168 if (caa_unlikely(!URCU_TLS(rcu_reader)))
a5a9f428 169 rcu_bp_register(); /* If not yet registered. */
5481ddb3 170 cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */
bd252a04 171 tmp = URCU_TLS(rcu_reader)->ctr;
6ef5ddc9 172 urcu_assert((tmp & RCU_GP_CTR_NEST_MASK) != RCU_GP_CTR_NEST_MASK);
a5a9f428 173 _rcu_read_lock_update(tmp);
fdee2e6d
MD
174}
175
a5a9f428
PM
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 */
fdee2e6d
MD
181static inline void _rcu_read_unlock(void)
182{
343c8b13
MD
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. */
f541831e 188 urcu_bp_smp_mb_slave();
343c8b13 189 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, tmp - RCU_GP_COUNT);
5481ddb3 190 cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */
fdee2e6d
MD
191}
192
882f3357
MD
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 */
200static 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
67ecffc0 207#ifdef __cplusplus
36bc70a8
MD
208}
209#endif
210
fdee2e6d 211#endif /* _URCU_BP_STATIC_H */
This page took 0.065426 seconds and 4 git commands to generate.