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/compiler.h>
  37 #include <urcu/arch.h>
  38 #include <urcu/system.h>
  39 #include <urcu/uatomic.h>
  40 #include <urcu/list.h>
  41 #include <urcu/tls-compat.h>
  42 #include <urcu/debug.h>
  43
  44 /*
  45  * This code section can only be included in LGPL 2.1 compatible source code.
  46  * See below for the function call wrappers which can be used in code meant to
  47  * be only linked with the Userspace RCU library. This comes with a small
  48  * performance degradation on the read-side due to the added function calls.
  49  * This is required to permit relinking with newer versions of the library.
  50  */
  51
  52 #ifdef __cplusplus
  53 extern "C" {
  54 #endif
  55
  56 enum rcu_state {
  57         RCU_READER_ACTIVE_CURRENT,
  58         RCU_READER_ACTIVE_OLD,
  59         RCU_READER_INACTIVE,
  60 };
  61
  62 /*
  63  * The trick here is that RCU_GP_CTR_PHASE must be a multiple of 8 so we can use a
  64  * full 8-bits, 16-bits or 32-bits bitmask for the lower order bits.
  65  */
  66 #define RCU_GP_COUNT            (1UL << 0)
  67 /* Use the amount of bits equal to half of the architecture long size */
  68 #define RCU_GP_CTR_PHASE                (1UL << (sizeof(long) << 2))
  69 #define RCU_GP_CTR_NEST_MASK    (RCU_GP_CTR_PHASE - 1)
  70
  71 /*
  72  * Used internally by _rcu_read_lock.
  73  */
  74 extern void rcu_bp_register(void);
  75
  76 struct rcu_gp {
  77         /*
  78          * Global grace period counter.
  79          * Contains the current RCU_GP_CTR_PHASE.
  80          * Also has a RCU_GP_COUNT of 1, to accelerate the reader fast path.
  81          * Written to only by writer with mutex taken.
  82          * Read by both writer and readers.
  83          */
  84         unsigned long ctr;
  85 } __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
  86
  87 extern struct rcu_gp rcu_gp;
  88
  89 struct rcu_reader {
  90         /* Data used by both reader and synchronize_rcu() */
  91         unsigned long ctr;
  92         /* Data used for registry */
  93         struct cds_list_head node __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
  94         pthread_t tid;
  95         int alloc;      /* registry entry allocated */
  96 };
  97
  98 /*
  99  * Bulletproof version keeps a pointer to a registry not part of the TLS.
 100  * Adds a pointer dereference on the read-side, but won't require to unregister
 101  * the reader thread.
 102  */
 103 extern DECLARE_URCU_TLS(struct rcu_reader *, rcu_reader);
 104
 105 extern int urcu_bp_has_sys_membarrier;
 106
 107 static inline void urcu_bp_smp_mb_slave(void)
 108 {
 109         if (caa_likely(urcu_bp_has_sys_membarrier))
 110                 cmm_barrier();
 111         else
 112                 cmm_smp_mb();
 113 }
 114
 115 static inline enum rcu_state rcu_reader_state(unsigned long *ctr)
 116 {
 117         unsigned long v;
 118
 119         if (ctr == NULL)
 120                 return RCU_READER_INACTIVE;
 121         /*
 122          * Make sure both tests below are done on the same version of *value
 123          * to insure consistency.
 124          */
 125         v = CMM_LOAD_SHARED(*ctr);
 126         if (!(v & RCU_GP_CTR_NEST_MASK))
 127                 return RCU_READER_INACTIVE;
 128         if (!((v ^ rcu_gp.ctr) & RCU_GP_CTR_PHASE))
 129                 return RCU_READER_ACTIVE_CURRENT;
 130         return RCU_READER_ACTIVE_OLD;
 131 }
 132
 133 /*
 134  * Helper for _rcu_read_lock().  The format of rcu_gp.ctr (as well as
 135  * the per-thread rcu_reader.ctr) has the upper bits containing a count of
 136  * _rcu_read_lock() nesting, and a lower-order bit that contains either zero
 137  * or RCU_GP_CTR_PHASE.  The smp_mb_slave() ensures that the accesses in
 138  * _rcu_read_lock() happen before the subsequent read-side critical section.
 139  */
 140 static inline void _rcu_read_lock_update(unsigned long tmp)
 141 {
 142         if (caa_likely(!(tmp & RCU_GP_CTR_NEST_MASK))) {
 143                 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, _CMM_LOAD_SHARED(rcu_gp.ctr));
 144                 urcu_bp_smp_mb_slave();
 145         } else
 146                 _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, tmp + RCU_GP_COUNT);
 147 }
 148
 149 /*
 150  * Enter an RCU read-side critical section.
 151  *
 152  * The first cmm_barrier() call ensures that the compiler does not reorder
 153  * the body of _rcu_read_lock() with a mutex.
 154  *
 155  * This function and its helper are both less than 10 lines long.  The
 156  * intent is that this function meets the 10-line criterion in LGPL,
 157  * allowing this function to be invoked directly from non-LGPL code.
 158  */
 159 static inline void _rcu_read_lock(void)
 160 {
 161         unsigned long tmp;
 162
 163         if (caa_unlikely(!URCU_TLS(rcu_reader)))
 164                 rcu_bp_register(); /* If not yet registered. */
 165         cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
 166         tmp = URCU_TLS(rcu_reader)->ctr;
 167         urcu_assert((tmp & RCU_GP_CTR_NEST_MASK) != RCU_GP_CTR_NEST_MASK);
 168         _rcu_read_lock_update(tmp);
 169 }
 170
 171 /*
 172  * Exit an RCU read-side critical section.  This function is less than
 173  * 10 lines of code, and is intended to be usable by non-LGPL code, as
 174  * called out in LGPL.
 175  */
 176 static inline void _rcu_read_unlock(void)
 177 {
 178         unsigned long tmp;
 179
 180         tmp = URCU_TLS(rcu_reader)->ctr;
 181         urcu_assert(tmp & RCU_GP_CTR_NEST_MASK);
 182         /* Finish using rcu before decrementing the pointer. */
 183         urcu_bp_smp_mb_slave();
 184         _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, tmp - RCU_GP_COUNT);
 185         cmm_barrier();  /* Ensure the compiler does not reorder us with mutex */
 186 }
 187
 188 /*
 189  * Returns whether within a RCU read-side critical section.
 190  *
 191  * This function is less than 10 lines long.  The intent is that this
 192  * function meets the 10-line criterion for LGPL, allowing this function
 193  * to be invoked directly from non-LGPL code.
 194  */
 195 static inline int _rcu_read_ongoing(void)
 196 {
 197         if (caa_unlikely(!URCU_TLS(rcu_reader)))
 198                 rcu_bp_register(); /* If not yet registered. */
 199         return URCU_TLS(rcu_reader)->ctr & RCU_GP_CTR_NEST_MASK;
 200 }
 201
 202 #ifdef __cplusplus
 203 }
 204 #endif
 205
 206 #endif /* _URCU_BP_STATIC_H */