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