uatomic/x86: Remove redundant memory barriers
[urcu.git] / include / urcu / static / pointer.h
1 #ifndef _URCU_POINTER_STATIC_H
2 #define _URCU_POINTER_STATIC_H
3
4 /*
5 * urcu/static/pointer.h
6 *
7 * Userspace RCU header. Operations on pointers.
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 <urcu/compiler.h>
33 #include <urcu/arch.h>
34 #include <urcu/system.h>
35 #include <urcu/uatomic.h>
36
37 #ifdef __cplusplus
38 extern "C" {
39 #endif
40
41 /**
42 * _rcu_dereference - reads (copy) a RCU-protected pointer to a local variable
43 * into a RCU read-side critical section. The pointer can later be safely
44 * dereferenced within the critical section.
45 *
46 * This ensures that the pointer copy is invariant thorough the whole critical
47 * section.
48 *
49 * Inserts memory barriers on architectures that require them (currently only
50 * Alpha) and documents which pointers are protected by RCU.
51 *
52 * With C standards prior to C11/C++11, the compiler memory barrier in
53 * CMM_LOAD_SHARED() ensures that value-speculative optimizations (e.g.
54 * VSS: Value Speculation Scheduling) does not perform the data read
55 * before the pointer read by speculating the value of the pointer.
56 * Correct ordering is ensured because the pointer is read as a volatile
57 * access. This acts as a global side-effect operation, which forbids
58 * reordering of dependent memory operations.
59 *
60 * With C standards C11/C++11, concerns about dependency-breaking
61 * optimizations are taken care of by the "memory_order_consume" atomic
62 * load.
63 *
64 * By defining URCU_DEREFERENCE_USE_VOLATILE, the user requires use of
65 * volatile access to implement rcu_dereference rather than
66 * memory_order_consume load from the C11/C++11 standards.
67 *
68 * This may improve performance on weakly-ordered architectures where
69 * the compiler implements memory_order_consume as a
70 * memory_order_acquire, which is stricter than required by the
71 * standard.
72 *
73 * Note that using volatile accesses for rcu_dereference may cause
74 * LTO to generate incorrectly ordered code starting from C11/C++11.
75 *
76 * Should match rcu_assign_pointer() or rcu_xchg_pointer().
77 *
78 * This macro is less than 10 lines long. The intent is that this macro
79 * meets the 10-line criterion in LGPL, allowing this function to be
80 * expanded directly in non-LGPL code.
81 */
82
83 #ifdef URCU_DEREFERENCE_USE_VOLATILE
84 # define __rcu_dereference(p) CMM_LOAD_SHARED(p)
85 #else
86 # if defined (__cplusplus)
87 # if __cplusplus >= 201103L
88 # include <atomic>
89 # define __rcu_dereference(p) ((std::atomic<__typeof__(p)>)(p)).load(std::memory_order_consume)
90 # else
91 # define __rcu_dereference(p) CMM_LOAD_SHARED(x)
92 # endif
93 # else
94 # if (defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L)
95 # include <stdatomic.h>
96 # define __rcu_dereference(p) atomic_load_explicit(&(p), memory_order_consume)
97 # else
98 # define __rcu_dereference(p) CMM_LOAD_SHARED(p)
99 # endif
100 # endif
101 #endif
102
103 #define _rcu_dereference(p) \
104 __extension__ \
105 ({ \
106 __typeof__(p) _________p1 = __rcu_dereference(p); \
107 cmm_smp_read_barrier_depends(); \
108 (_________p1); \
109 })
110
111 /**
112 * _rcu_cmpxchg_pointer - same as rcu_assign_pointer, but tests if the pointer
113 * is as expected by "old". If succeeds, returns the previous pointer to the
114 * data structure, which can be safely freed after waiting for a quiescent state
115 * using synchronize_rcu(). If fails (unexpected value), returns old (which
116 * should not be freed !).
117 *
118 * uatomic_cmpxchg() acts as both release and acquire barriers.
119 *
120 * This macro is less than 10 lines long. The intent is that this macro
121 * meets the 10-line criterion in LGPL, allowing this function to be
122 * expanded directly in non-LGPL code.
123 */
124 #define _rcu_cmpxchg_pointer(p, old, _new) \
125 __extension__ \
126 ({ \
127 __typeof__(*p) _________pold = (old); \
128 __typeof__(*p) _________pnew = (_new); \
129 uatomic_cmpxchg(p, _________pold, _________pnew); \
130 })
131
132 /**
133 * _rcu_xchg_pointer - same as rcu_assign_pointer, but returns the previous
134 * pointer to the data structure, which can be safely freed after waiting for a
135 * quiescent state using synchronize_rcu().
136 *
137 * uatomic_xchg() acts as both release and acquire barriers.
138 *
139 * This macro is less than 10 lines long. The intent is that this macro
140 * meets the 10-line criterion in LGPL, allowing this function to be
141 * expanded directly in non-LGPL code.
142 */
143 #define _rcu_xchg_pointer(p, v) \
144 __extension__ \
145 ({ \
146 __typeof__(*p) _________pv = (v); \
147 uatomic_xchg(p, _________pv); \
148 })
149
150
151 #define _rcu_set_pointer(p, v) \
152 do { \
153 __typeof__(*p) _________pv = (v); \
154 if (!__builtin_constant_p(v) || \
155 ((v) != NULL)) \
156 cmm_wmb(); \
157 uatomic_set(p, _________pv); \
158 } while (0)
159
160 /**
161 * _rcu_assign_pointer - assign (publicize) a pointer to a new data structure
162 * meant to be read by RCU read-side critical sections. Returns the assigned
163 * value.
164 *
165 * Documents which pointers will be dereferenced by RCU read-side critical
166 * sections and adds the required memory barriers on architectures requiring
167 * them. It also makes sure the compiler does not reorder code initializing the
168 * data structure before its publication.
169 *
170 * Should match rcu_dereference().
171 *
172 * This macro is less than 10 lines long. The intent is that this macro
173 * meets the 10-line criterion in LGPL, allowing this function to be
174 * expanded directly in non-LGPL code.
175 */
176 #define _rcu_assign_pointer(p, v) _rcu_set_pointer(&(p), v)
177
178 #ifdef __cplusplus
179 }
180 #endif
181
182 #endif /* _URCU_POINTER_STATIC_H */
This page took 0.032016 seconds and 4 git commands to generate.