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