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 LGPL-COMPATIBLE CODE. See urcu.h for linking
  10  * 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/compiler.h>
  37 #include <urcu/arch.h>
  38 #include <urcu/system.h>
  39 #include <urcu/uatomic.h>
  40 #include <urcu/list.h>
  41
  42 /*
  43  * This code section can only be included in LGPL 2.1 compatible source code.
  44  * See below for the function call wrappers which can be used in code meant to
  45  * be only linked with the Userspace RCU library. This comes with a small
  46  * performance degradation on the read-side due to the added function calls.
  47  * This is required to permit relinking with newer versions of the library.
  48  */
  49
  50 #ifdef __cplusplus
  51 extern "C" {
  52 #endif
  53
  54 #ifdef DEBUG_RCU
  55 #define rcu_assert(args...)     assert(args)
  56 #else
  57 #define rcu_assert(args...)
  58 #endif
  59
  60 #ifdef DEBUG_YIELD
  61 #include <sched.h>
  62 #include <time.h>
  63 #include <pthread.h>
  64 #include <unistd.h>
  65
  66 #define YIELD_READ      (1 << 0)
  67 #define YIELD_WRITE     (1 << 1)
  68
  69 /*
  70  * Updates without RCU_MB are much slower. Account this in
  71  * the delay.
  72  */
  73 /* maximum sleep delay, in us */
  74 #define MAX_SLEEP 50
  75
  76 extern unsigned int yield_active;
  77 extern unsigned int __thread rand_yield;
  78
  79 static inline void debug_yield_read(void)
  80 {
  81         if (yield_active & YIELD_READ)
  82                 if (rand_r(&rand_yield) & 0x1)
  83                         usleep(rand_r(&rand_yield) % MAX_SLEEP);
  84 }
  85
  86 static inline void debug_yield_write(void)
  87 {
  88         if (yield_active & YIELD_WRITE)
  89                 if (rand_r(&rand_yield) & 0x1)
  90                         usleep(rand_r(&rand_yield) % MAX_SLEEP);
  91 }
  92
  93 static inline void debug_yield_init(void)
  94 {
  95         rand_yield = time(NULL) ^ pthread_self();
  96 }
  97 #else
  98 static inline void debug_yield_read(void)
  99 {
 100 }
 101
 102 static inline void debug_yield_write(void)
 103 {
 104 }
 105
 106 static inline void debug_yield_init(void)
 107 {
 108
 109 }
 110 #endif
 111
 112 /*
 113  * The trick here is that RCU_GP_CTR_PHASE must be a multiple of 8 so we can use a
 114  * full 8-bits, 16-bits or 32-bits bitmask for the lower order bits.
 115  */
 116 #define RCU_GP_COUNT            (1UL << 0)
 117 /* Use the amount of bits equal to half of the architecture long size */
 118 #define RCU_GP_CTR_PHASE                (1UL << (sizeof(long) << 2))
 119 #define RCU_GP_CTR_NEST_MASK    (RCU_GP_CTR_PHASE - 1)
 120
 121 /*
 122  * Used internally by _rcu_read_lock.
 123  */
 124 extern void rcu_bp_register(void);
 125
 126 /*
 127  * Global quiescent period counter with low-order bits unused.
 128  * Using a int rather than a char to eliminate false register dependencies
 129  * causing stalls on some architectures.
 130  */
 131 extern long rcu_gp_ctr;
 132
 133 struct rcu_reader {
 134         /* Data used by both reader and synchronize_rcu() */
 135         long ctr;
 136         /* Data used for registry */
 137         struct cds_list_head node __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
 138         pthread_t tid;
 139         int alloc;      /* registry entry allocated */
 140 };
 141
 142 /*
 143  * Bulletproof version keeps a pointer to a registry not part of the TLS.
 144  * Adds a pointer dereference on the read-side, but won't require to unregister
 145  * the reader thread.
 146  */
 147 extern struct rcu_reader __thread *rcu_reader;
 148
 149 static inline int rcu_old_gp_ongoing(long *value)
 150 {
 151         long v;
 152
 153         if (value == NULL)
 154                 return 0;
 155         /*
 156          * Make sure both tests below are done on the same version of *value
 157          * to insure consistency.
 158          */
 159         v = CMM_LOAD_SHARED(*value);
 160         return (v & RCU_GP_CTR_NEST_MASK) &&
 161                  ((v ^ rcu_gp_ctr) & RCU_GP_CTR_PHASE);
 162 }
 163
 164 static inline void _rcu_read_lock(void)
 165 {
 166         long tmp;
 167
 168         /* Check if registered */
 169         if (caa_unlikely(!rcu_reader))
 170                 rcu_bp_register();
 171
 172         cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
 173         tmp = rcu_reader->ctr;
 174         /*
 175          * rcu_gp_ctr is
 176          *   RCU_GP_COUNT | (~RCU_GP_CTR_PHASE or RCU_GP_CTR_PHASE)
 177          */
 178         if (caa_likely(!(tmp & RCU_GP_CTR_NEST_MASK))) {
 179                 _CMM_STORE_SHARED(rcu_reader->ctr, _CMM_LOAD_SHARED(rcu_gp_ctr));
 180                 /*
 181                  * Set active readers count for outermost nesting level before
 182                  * accessing the pointer.
 183                  */
 184                 cmm_smp_mb();
 185         } else {
 186                 _CMM_STORE_SHARED(rcu_reader->ctr, tmp + RCU_GP_COUNT);
 187         }
 188 }
 189
 190 static inline void _rcu_read_unlock(void)
 191 {
 192         /*
 193          * Finish using rcu before decrementing the pointer.
 194          */
 195         cmm_smp_mb();
 196         _CMM_STORE_SHARED(rcu_reader->ctr, rcu_reader->ctr - RCU_GP_COUNT);
 197         cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
 198 }
 199
 200 #ifdef __cplusplus
 201 }
 202 #endif
 203
 204 #endif /* _URCU_BP_STATIC_H */