Commit | Line | Data |
---|---|---|
fdee2e6d MD |
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 | * | |
a5a9f428 PM |
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. | |
fdee2e6d | 11 | * |
6982d6d7 | 12 | * Copyright (c) 2009 Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
fdee2e6d MD |
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> | |
fdee2e6d MD |
34 | #include <unistd.h> |
35 | ||
01477510 | 36 | #include <urcu/debug.h> |
d8d9a340 | 37 | #include <urcu/config.h> |
fdee2e6d MD |
38 | #include <urcu/compiler.h> |
39 | #include <urcu/arch.h> | |
40 | #include <urcu/system.h> | |
a2e7bf9c | 41 | #include <urcu/uatomic.h> |
fdee2e6d | 42 | #include <urcu/list.h> |
bd252a04 | 43 | #include <urcu/tls-compat.h> |
fdee2e6d MD |
44 | |
45 | /* | |
46 | * This code section can only be included in LGPL 2.1 compatible source code. | |
47 | * See below for the function call wrappers which can be used in code meant to | |
48 | * be only linked with the Userspace RCU library. This comes with a small | |
49 | * performance degradation on the read-side due to the added function calls. | |
50 | * This is required to permit relinking with newer versions of the library. | |
51 | */ | |
52 | ||
36bc70a8 MD |
53 | #ifdef __cplusplus |
54 | extern "C" { | |
55 | #endif | |
56 | ||
4477a870 MD |
57 | enum urcu_bp_state { |
58 | URCU_BP_READER_ACTIVE_CURRENT, | |
59 | URCU_BP_READER_ACTIVE_OLD, | |
60 | URCU_BP_READER_INACTIVE, | |
52c75091 MD |
61 | }; |
62 | ||
fdee2e6d | 63 | /* |
4477a870 | 64 | * The trick here is that URCU_BP_GP_CTR_PHASE must be a multiple of 8 so we can use a |
fdee2e6d MD |
65 | * full 8-bits, 16-bits or 32-bits bitmask for the lower order bits. |
66 | */ | |
4477a870 | 67 | #define URCU_BP_GP_COUNT (1UL << 0) |
fdee2e6d | 68 | /* Use the amount of bits equal to half of the architecture long size */ |
4477a870 MD |
69 | #define URCU_BP_GP_CTR_PHASE (1UL << (sizeof(long) << 2)) |
70 | #define URCU_BP_GP_CTR_NEST_MASK (URCU_BP_GP_CTR_PHASE - 1) | |
fdee2e6d MD |
71 | |
72 | /* | |
4477a870 | 73 | * Used internally by _urcu_bp_read_lock. |
fdee2e6d | 74 | */ |
4477a870 | 75 | extern void urcu_bp_register(void); |
fdee2e6d | 76 | |
4477a870 | 77 | struct urcu_bp_gp { |
c13c2e55 MD |
78 | /* |
79 | * Global grace period counter. | |
4477a870 MD |
80 | * Contains the current URCU_BP_GP_CTR_PHASE. |
81 | * Also has a URCU_BP_GP_COUNT of 1, to accelerate the reader fast path. | |
c13c2e55 MD |
82 | * Written to only by writer with mutex taken. |
83 | * Read by both writer and readers. | |
84 | */ | |
85 | unsigned long ctr; | |
86 | } __attribute__((aligned(CAA_CACHE_LINE_SIZE))); | |
87 | ||
4477a870 | 88 | extern struct urcu_bp_gp urcu_bp_gp; |
fdee2e6d | 89 | |
4477a870 MD |
90 | struct urcu_bp_reader { |
91 | /* Data used by both reader and urcu_bp_synchronize_rcu() */ | |
52c75091 | 92 | unsigned long ctr; |
fdee2e6d | 93 | /* Data used for registry */ |
16aa9ee8 | 94 | struct cds_list_head node __attribute__((aligned(CAA_CACHE_LINE_SIZE))); |
fdee2e6d MD |
95 | pthread_t tid; |
96 | int alloc; /* registry entry allocated */ | |
97 | }; | |
98 | ||
99 | /* | |
100 | * Bulletproof version keeps a pointer to a registry not part of the TLS. | |
101 | * Adds a pointer dereference on the read-side, but won't require to unregister | |
102 | * the reader thread. | |
103 | */ | |
4477a870 | 104 | extern DECLARE_URCU_TLS(struct urcu_bp_reader *, urcu_bp_reader); |
fdee2e6d | 105 | |
d8d9a340 MD |
106 | #ifdef CONFIG_RCU_FORCE_SYS_MEMBARRIER |
107 | #define urcu_bp_has_sys_membarrier 1 | |
108 | #else | |
f541831e | 109 | extern int urcu_bp_has_sys_membarrier; |
d8d9a340 | 110 | #endif |
f541831e MD |
111 | |
112 | static inline void urcu_bp_smp_mb_slave(void) | |
113 | { | |
114 | if (caa_likely(urcu_bp_has_sys_membarrier)) | |
115 | cmm_barrier(); | |
116 | else | |
117 | cmm_smp_mb(); | |
118 | } | |
119 | ||
4477a870 | 120 | static inline enum urcu_bp_state urcu_bp_reader_state(unsigned long *ctr) |
fdee2e6d | 121 | { |
52c75091 | 122 | unsigned long v; |
fdee2e6d | 123 | |
52c75091 | 124 | if (ctr == NULL) |
4477a870 | 125 | return URCU_BP_READER_INACTIVE; |
fdee2e6d MD |
126 | /* |
127 | * Make sure both tests below are done on the same version of *value | |
128 | * to insure consistency. | |
129 | */ | |
52c75091 | 130 | v = CMM_LOAD_SHARED(*ctr); |
4477a870 MD |
131 | if (!(v & URCU_BP_GP_CTR_NEST_MASK)) |
132 | return URCU_BP_READER_INACTIVE; | |
133 | if (!((v ^ urcu_bp_gp.ctr) & URCU_BP_GP_CTR_PHASE)) | |
134 | return URCU_BP_READER_ACTIVE_CURRENT; | |
135 | return URCU_BP_READER_ACTIVE_OLD; | |
fdee2e6d MD |
136 | } |
137 | ||
a5a9f428 | 138 | /* |
4477a870 | 139 | * Helper for _urcu_bp_read_lock(). The format of urcu_bp_gp.ctr (as well as |
be152bdb LKO |
140 | * the per-thread rcu_reader.ctr) has the lower-order bits containing a count of |
141 | * _urcu_bp_read_lock() nesting, and a single high-order URCU_BP_GP_CTR_PHASE bit | |
142 | * that contains either zero or one. The smp_mb_slave() ensures that the accesses in | |
4477a870 | 143 | * _urcu_bp_read_lock() happen before the subsequent read-side critical section. |
a5a9f428 | 144 | */ |
4477a870 | 145 | static inline void _urcu_bp_read_lock_update(unsigned long tmp) |
a5a9f428 | 146 | { |
4477a870 MD |
147 | if (caa_likely(!(tmp & URCU_BP_GP_CTR_NEST_MASK))) { |
148 | _CMM_STORE_SHARED(URCU_TLS(urcu_bp_reader)->ctr, _CMM_LOAD_SHARED(urcu_bp_gp.ctr)); | |
f541831e | 149 | urcu_bp_smp_mb_slave(); |
a5a9f428 | 150 | } else |
4477a870 | 151 | _CMM_STORE_SHARED(URCU_TLS(urcu_bp_reader)->ctr, tmp + URCU_BP_GP_COUNT); |
a5a9f428 PM |
152 | } |
153 | ||
154 | /* | |
155 | * Enter an RCU read-side critical section. | |
156 | * | |
157 | * The first cmm_barrier() call ensures that the compiler does not reorder | |
4477a870 | 158 | * the body of _urcu_bp_read_lock() with a mutex. |
a5a9f428 PM |
159 | * |
160 | * This function and its helper are both less than 10 lines long. The | |
161 | * intent is that this function meets the 10-line criterion in LGPL, | |
162 | * allowing this function to be invoked directly from non-LGPL code. | |
163 | */ | |
4477a870 | 164 | static inline void _urcu_bp_read_lock(void) |
fdee2e6d | 165 | { |
52c75091 | 166 | unsigned long tmp; |
fdee2e6d | 167 | |
4477a870 MD |
168 | if (caa_unlikely(!URCU_TLS(urcu_bp_reader))) |
169 | urcu_bp_register(); /* If not yet registered. */ | |
5481ddb3 | 170 | cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */ |
4477a870 | 171 | tmp = URCU_TLS(urcu_bp_reader)->ctr; |
2a27e931 | 172 | urcu_assert_debug((tmp & URCU_BP_GP_CTR_NEST_MASK) != URCU_BP_GP_CTR_NEST_MASK); |
4477a870 | 173 | _urcu_bp_read_lock_update(tmp); |
fdee2e6d MD |
174 | } |
175 | ||
a5a9f428 PM |
176 | /* |
177 | * Exit an RCU read-side critical section. This function is less than | |
178 | * 10 lines of code, and is intended to be usable by non-LGPL code, as | |
179 | * called out in LGPL. | |
180 | */ | |
4477a870 | 181 | static inline void _urcu_bp_read_unlock(void) |
fdee2e6d | 182 | { |
343c8b13 MD |
183 | unsigned long tmp; |
184 | ||
4477a870 | 185 | tmp = URCU_TLS(urcu_bp_reader)->ctr; |
2a27e931 | 186 | urcu_assert_debug(tmp & URCU_BP_GP_CTR_NEST_MASK); |
343c8b13 | 187 | /* Finish using rcu before decrementing the pointer. */ |
f541831e | 188 | urcu_bp_smp_mb_slave(); |
4477a870 | 189 | _CMM_STORE_SHARED(URCU_TLS(urcu_bp_reader)->ctr, tmp - URCU_BP_GP_COUNT); |
5481ddb3 | 190 | cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */ |
fdee2e6d MD |
191 | } |
192 | ||
882f3357 MD |
193 | /* |
194 | * Returns whether within a RCU read-side critical section. | |
195 | * | |
196 | * This function is less than 10 lines long. The intent is that this | |
197 | * function meets the 10-line criterion for LGPL, allowing this function | |
198 | * to be invoked directly from non-LGPL code. | |
199 | */ | |
4477a870 | 200 | static inline int _urcu_bp_read_ongoing(void) |
882f3357 | 201 | { |
4477a870 MD |
202 | if (caa_unlikely(!URCU_TLS(urcu_bp_reader))) |
203 | urcu_bp_register(); /* If not yet registered. */ | |
204 | return URCU_TLS(urcu_bp_reader)->ctr & URCU_BP_GP_CTR_NEST_MASK; | |
882f3357 MD |
205 | } |
206 | ||
67ecffc0 | 207 | #ifdef __cplusplus |
36bc70a8 MD |
208 | } |
209 | #endif | |
210 | ||
fdee2e6d | 211 | #endif /* _URCU_BP_STATIC_H */ |