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  * Use the gcc __atomic_load() rather than C11/C++11 atomic load
  65  * explicit because the pointer used as input argument is a pointer,
  66  * not an _Atomic type as required by C11/C++11.
  67  *
  68  * By defining URCU_DEREFERENCE_USE_VOLATILE, the user requires use of
  69  * volatile access to implement rcu_dereference rather than
  70  * memory_order_consume load from the C11/C++11 standards.
  71  *
  72  * This may improve performance on weakly-ordered architectures where
  73  * the compiler implements memory_order_consume as a
  74  * memory_order_acquire, which is stricter than required by the
  75  * standard.
  76  *
  77  * Note that using volatile accesses for rcu_dereference may cause
  78  * LTO to generate incorrectly ordered code starting from C11/C++11.
  79  *
  80  * Should match rcu_assign_pointer() or rcu_xchg_pointer().
  81  *
  82  * This macro is less than 10 lines long.  The intent is that this macro
  83  * meets the 10-line criterion in LGPL, allowing this function to be
  84  * expanded directly in non-LGPL code.
  85  */
  86
  87 #if !defined (URCU_DEREFERENCE_USE_VOLATILE) &&         \
  88         ((defined (__cplusplus) && __cplusplus >= 201103L) ||   \
  89         (defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L))
  90 # define __URCU_DEREFERENCE_USE_ATOMIC_CONSUME
  91 #endif
  92
  93 #ifdef __URCU_DEREFERENCE_USE_ATOMIC_CONSUME
  94 # define _rcu_dereference(p) __extension__ ({                                           \
  95                                 __typeof__(p) _________p1;                              \
  96                                 __atomic_load(&(p), &_________p1, __ATOMIC_CONSUME);    \
  97                                 (_________p1);                                          \
  98                         })
  99 #else
 100 # define _rcu_dereference(p) __extension__ ({                                           \
 101                                 __typeof__(p) _________p1 = CMM_LOAD_SHARED(p);         \
 102                                 cmm_smp_read_barrier_depends();                         \
 103                                 (_________p1);                                          \
 104                         })
 105 #endif
 106
 107 /**
 108  * _rcu_cmpxchg_pointer - same as rcu_assign_pointer, but tests if the pointer
 109  * is as expected by "old". If succeeds, returns the previous pointer to the
 110  * data structure, which can be safely freed after waiting for a quiescent state
 111  * using synchronize_rcu(). If fails (unexpected value), returns old (which
 112  * should not be freed !).
 113  *
 114  * uatomic_cmpxchg() acts as both release and acquire barriers.
 115  *
 116  * This macro is less than 10 lines long.  The intent is that this macro
 117  * meets the 10-line criterion in LGPL, allowing this function to be
 118  * expanded directly in non-LGPL code.
 119  */
 120 #define _rcu_cmpxchg_pointer(p, old, _new)                              \
 121         __extension__                                                   \
 122         ({                                                              \
 123                 __typeof__(*p) _________pold = (old);                   \
 124                 __typeof__(*p) _________pnew = (_new);                  \
 125                 uatomic_cmpxchg(p, _________pold, _________pnew);       \
 126         })
 127
 128 /**
 129  * _rcu_xchg_pointer - same as rcu_assign_pointer, but returns the previous
 130  * pointer to the data structure, which can be safely freed after waiting for a
 131  * quiescent state using synchronize_rcu().
 132  *
 133  * uatomic_xchg() acts as both release and acquire barriers.
 134  *
 135  * This macro is less than 10 lines long.  The intent is that this macro
 136  * meets the 10-line criterion in LGPL, allowing this function to be
 137  * expanded directly in non-LGPL code.
 138  */
 139 #define _rcu_xchg_pointer(p, v)                         \
 140         __extension__                                   \
 141         ({                                              \
 142                 __typeof__(*p) _________pv = (v);       \
 143                 uatomic_xchg(p, _________pv);           \
 144         })
 145
 146
 147 #define _rcu_set_pointer(p, v)                          \
 148         do {                                            \
 149                 __typeof__(*p) _________pv = (v);       \
 150                 if (!__builtin_constant_p(v) ||         \
 151                     ((v) != NULL))                      \
 152                         cmm_wmb();                              \
 153                 uatomic_set(p, _________pv);            \
 154         } while (0)
 155
 156 /**
 157  * _rcu_assign_pointer - assign (publicize) a pointer to a new data structure
 158  * meant to be read by RCU read-side critical sections. Returns the assigned
 159  * value.
 160  *
 161  * Documents which pointers will be dereferenced by RCU read-side critical
 162  * sections and adds the required memory barriers on architectures requiring
 163  * them. It also makes sure the compiler does not reorder code initializing the
 164  * data structure before its publication.
 165  *
 166  * Should match rcu_dereference().
 167  *
 168  * This macro is less than 10 lines long.  The intent is that this macro
 169  * meets the 10-line criterion in LGPL, allowing this function to be
 170  * expanded directly in non-LGPL code.
 171  */
 172 #define _rcu_assign_pointer(p, v)       _rcu_set_pointer(&(p), v)
 173
 174 #ifdef __cplusplus
 175 }
 176 #endif
 177
 178 #endif /* _URCU_POINTER_STATIC_H */