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 | * | |
9 | * TO BE INCLUDED ONLY IN LGPL-COMPATIBLE CODE. See urcu.h for linking | |
10 | * dynamically with the userspace rcu library. | |
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 | ||
36 | #include <urcu/compiler.h> | |
37 | #include <urcu/arch.h> | |
38 | #include <urcu/system.h> | |
a2e7bf9c | 39 | #include <urcu/uatomic.h> |
fdee2e6d | 40 | #include <urcu/list.h> |
bd252a04 | 41 | #include <urcu/tls-compat.h> |
fdee2e6d MD |
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 | ||
36bc70a8 MD |
51 | #ifdef __cplusplus |
52 | extern "C" { | |
53 | #endif | |
54 | ||
fdee2e6d MD |
55 | #ifdef DEBUG_RCU |
56 | #define rcu_assert(args...) assert(args) | |
57 | #else | |
58 | #define rcu_assert(args...) | |
59 | #endif | |
60 | ||
61 | #ifdef DEBUG_YIELD | |
62 | #include <sched.h> | |
63 | #include <time.h> | |
64 | #include <pthread.h> | |
65 | #include <unistd.h> | |
66 | ||
67 | #define YIELD_READ (1 << 0) | |
68 | #define YIELD_WRITE (1 << 1) | |
69 | ||
70 | /* | |
02be5561 | 71 | * Updates without RCU_MB are much slower. Account this in |
fdee2e6d MD |
72 | * the delay. |
73 | */ | |
74 | /* maximum sleep delay, in us */ | |
75 | #define MAX_SLEEP 50 | |
76 | ||
77 | extern unsigned int yield_active; | |
bd252a04 | 78 | extern DECLARE_URCU_TLS(unsigned int, rand_yield); |
fdee2e6d MD |
79 | |
80 | static inline void debug_yield_read(void) | |
81 | { | |
82 | if (yield_active & YIELD_READ) | |
bd252a04 MD |
83 | if (rand_r(&URCU_TLS(rand_yield)) & 0x1) |
84 | usleep(rand_r(&URCU_TLS(rand_yield)) % MAX_SLEEP); | |
fdee2e6d MD |
85 | } |
86 | ||
87 | static inline void debug_yield_write(void) | |
88 | { | |
89 | if (yield_active & YIELD_WRITE) | |
bd252a04 MD |
90 | if (rand_r(&URCU_TLS(rand_yield)) & 0x1) |
91 | usleep(rand_r(&URCU_TLS(rand_yield)) % MAX_SLEEP); | |
fdee2e6d MD |
92 | } |
93 | ||
94 | static inline void debug_yield_init(void) | |
95 | { | |
bd252a04 | 96 | URCU_TLS(rand_yield) = time(NULL) ^ pthread_self(); |
fdee2e6d MD |
97 | } |
98 | #else | |
99 | static inline void debug_yield_read(void) | |
100 | { | |
101 | } | |
102 | ||
103 | static inline void debug_yield_write(void) | |
104 | { | |
105 | } | |
106 | ||
107 | static inline void debug_yield_init(void) | |
108 | { | |
109 | ||
110 | } | |
111 | #endif | |
112 | ||
113 | /* | |
02be5561 | 114 | * The trick here is that RCU_GP_CTR_PHASE must be a multiple of 8 so we can use a |
fdee2e6d MD |
115 | * full 8-bits, 16-bits or 32-bits bitmask for the lower order bits. |
116 | */ | |
117 | #define RCU_GP_COUNT (1UL << 0) | |
118 | /* Use the amount of bits equal to half of the architecture long size */ | |
02be5561 MD |
119 | #define RCU_GP_CTR_PHASE (1UL << (sizeof(long) << 2)) |
120 | #define RCU_GP_CTR_NEST_MASK (RCU_GP_CTR_PHASE - 1) | |
fdee2e6d MD |
121 | |
122 | /* | |
123 | * Used internally by _rcu_read_lock. | |
124 | */ | |
125 | extern void rcu_bp_register(void); | |
126 | ||
127 | /* | |
128 | * Global quiescent period counter with low-order bits unused. | |
129 | * Using a int rather than a char to eliminate false register dependencies | |
130 | * causing stalls on some architectures. | |
131 | */ | |
02be5561 | 132 | extern long rcu_gp_ctr; |
fdee2e6d | 133 | |
02be5561 | 134 | struct rcu_reader { |
fdee2e6d MD |
135 | /* Data used by both reader and synchronize_rcu() */ |
136 | long ctr; | |
137 | /* Data used for registry */ | |
16aa9ee8 | 138 | struct cds_list_head node __attribute__((aligned(CAA_CACHE_LINE_SIZE))); |
fdee2e6d MD |
139 | pthread_t tid; |
140 | int alloc; /* registry entry allocated */ | |
141 | }; | |
142 | ||
143 | /* | |
144 | * Bulletproof version keeps a pointer to a registry not part of the TLS. | |
145 | * Adds a pointer dereference on the read-side, but won't require to unregister | |
146 | * the reader thread. | |
147 | */ | |
bd252a04 | 148 | extern DECLARE_URCU_TLS(struct rcu_reader *, rcu_reader); |
fdee2e6d MD |
149 | |
150 | static inline int rcu_old_gp_ongoing(long *value) | |
151 | { | |
152 | long v; | |
153 | ||
154 | if (value == NULL) | |
155 | return 0; | |
156 | /* | |
157 | * Make sure both tests below are done on the same version of *value | |
158 | * to insure consistency. | |
159 | */ | |
6cf3827c | 160 | v = CMM_LOAD_SHARED(*value); |
fdee2e6d | 161 | return (v & RCU_GP_CTR_NEST_MASK) && |
02be5561 | 162 | ((v ^ rcu_gp_ctr) & RCU_GP_CTR_PHASE); |
fdee2e6d MD |
163 | } |
164 | ||
165 | static inline void _rcu_read_lock(void) | |
166 | { | |
167 | long tmp; | |
168 | ||
169 | /* Check if registered */ | |
bd252a04 | 170 | if (caa_unlikely(!URCU_TLS(rcu_reader))) |
fdee2e6d MD |
171 | rcu_bp_register(); |
172 | ||
5481ddb3 | 173 | cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */ |
bd252a04 | 174 | tmp = URCU_TLS(rcu_reader)->ctr; |
02be5561 MD |
175 | /* |
176 | * rcu_gp_ctr is | |
177 | * RCU_GP_COUNT | (~RCU_GP_CTR_PHASE or RCU_GP_CTR_PHASE) | |
178 | */ | |
a0b7f7ea | 179 | if (caa_likely(!(tmp & RCU_GP_CTR_NEST_MASK))) { |
bd252a04 | 180 | _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, _CMM_LOAD_SHARED(rcu_gp_ctr)); |
fdee2e6d MD |
181 | /* |
182 | * Set active readers count for outermost nesting level before | |
183 | * accessing the pointer. | |
184 | */ | |
5481ddb3 | 185 | cmm_smp_mb(); |
fdee2e6d | 186 | } else { |
bd252a04 | 187 | _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, tmp + RCU_GP_COUNT); |
fdee2e6d MD |
188 | } |
189 | } | |
190 | ||
191 | static inline void _rcu_read_unlock(void) | |
192 | { | |
193 | /* | |
194 | * Finish using rcu before decrementing the pointer. | |
195 | */ | |
5481ddb3 | 196 | cmm_smp_mb(); |
bd252a04 | 197 | _CMM_STORE_SHARED(URCU_TLS(rcu_reader)->ctr, URCU_TLS(rcu_reader)->ctr - RCU_GP_COUNT); |
5481ddb3 | 198 | cmm_barrier(); /* Ensure the compiler does not reorder us with mutex */ |
fdee2e6d MD |
199 | } |
200 | ||
36bc70a8 MD |
201 | #ifdef __cplusplus |
202 | } | |
203 | #endif | |
204 | ||
fdee2e6d | 205 | #endif /* _URCU_BP_STATIC_H */ |