1 //===-- atomic.c - Implement support functions for atomic operations.------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // atomic.c defines a set of functions for performing atomic accesses on
10 // arbitrary-sized memory locations. This design uses locks that should
11 // be fast in the uncontended case, for two reasons:
12 //
13 // 1) This code must work with C programs that do not link to anything
14 // (including pthreads) and so it should not depend on any pthread
15 // functions.
16 // 2) Atomic operations, rather than explicit mutexes, are most commonly used
17 // on code where contended operations are rate.
18 //
19 // To avoid needing a per-object lock, this code allocates an array of
20 // locks and hashes the object pointers to find the one that it should use.
21 // For operations that must be atomic on two locations, the lower lock is
22 // always acquired first, to avoid deadlock.
23 //
24 //===----------------------------------------------------------------------===//
25
26 #include <stdbool.h>
27 #include <stddef.h>
28 #include <stdint.h>
29
30 #include "assembly.h"
31
32 // We use __builtin_mem* here to avoid dependencies on libc-provided headers.
33 #define memcpy __builtin_memcpy
34 #define memcmp __builtin_memcmp
35
36 // Clang objects if you redefine a builtin. This little hack allows us to
37 // define a function with the same name as an intrinsic.
38 #pragma redefine_extname __atomic_load_c SYMBOL_NAME(__atomic_load)
39 #pragma redefine_extname __atomic_store_c SYMBOL_NAME(__atomic_store)
40 #pragma redefine_extname __atomic_exchange_c SYMBOL_NAME(__atomic_exchange)
41 #pragma redefine_extname __atomic_compare_exchange_c SYMBOL_NAME( \
42 __atomic_compare_exchange)
43 #pragma redefine_extname __atomic_is_lock_free_c SYMBOL_NAME( \
44 __atomic_is_lock_free)
45
46 /// Number of locks. This allocates one page on 32-bit platforms, two on
47 /// 64-bit. This can be specified externally if a different trade between
48 /// memory usage and contention probability is required for a given platform.
49 #ifndef SPINLOCK_COUNT
50 #define SPINLOCK_COUNT (1 << 10)
51 #endif
52 static const long SPINLOCK_MASK = SPINLOCK_COUNT - 1;
53
54 ////////////////////////////////////////////////////////////////////////////////
55 // Platform-specific lock implementation. Falls back to spinlocks if none is
56 // defined. Each platform should define the Lock type, and corresponding
57 // lock() and unlock() functions.
58 ////////////////////////////////////////////////////////////////////////////////
59 #if defined(__FreeBSD__) || defined(__DragonFly__)
60 #include <errno.h>
61 // clang-format off
62 #include <sys/types.h>
63 #include <machine/atomic.h>
64 #include <sys/umtx.h>
65 // clang-format on
66 typedef struct _usem Lock;
unlock(Lock * l)67 __inline static void unlock(Lock *l) {
68 __c11_atomic_store((_Atomic(uint32_t) *)&l->_count, 1, __ATOMIC_RELEASE);
69 __c11_atomic_thread_fence(__ATOMIC_SEQ_CST);
70 if (l->_has_waiters)
71 _umtx_op(l, UMTX_OP_SEM_WAKE, 1, 0, 0);
72 }
lock(Lock * l)73 __inline static void lock(Lock *l) {
74 uint32_t old = 1;
75 while (!__c11_atomic_compare_exchange_weak((_Atomic(uint32_t) *)&l->_count,
76 &old, 0, __ATOMIC_ACQUIRE,
77 __ATOMIC_RELAXED)) {
78 _umtx_op(l, UMTX_OP_SEM_WAIT, 0, 0, 0);
79 old = 1;
80 }
81 }
82 /// locks for atomic operations
83 static Lock locks[SPINLOCK_COUNT] = {[0 ... SPINLOCK_COUNT - 1] = {0, 1, 0}};
84
85 #elif defined(__APPLE__)
86 #include <libkern/OSAtomic.h>
87 typedef OSSpinLock Lock;
unlock(Lock * l)88 __inline static void unlock(Lock *l) { OSSpinLockUnlock(l); }
89 /// Locks a lock. In the current implementation, this is potentially
90 /// unbounded in the contended case.
lock(Lock * l)91 __inline static void lock(Lock *l) { OSSpinLockLock(l); }
92 static Lock locks[SPINLOCK_COUNT]; // initialized to OS_SPINLOCK_INIT which is 0
93
94 #else
95 typedef _Atomic(uintptr_t) Lock;
96 /// Unlock a lock. This is a release operation.
unlock(Lock * l)97 __inline static void unlock(Lock *l) {
98 __c11_atomic_store(l, 0, __ATOMIC_RELEASE);
99 }
100 /// Locks a lock. In the current implementation, this is potentially
101 /// unbounded in the contended case.
lock(Lock * l)102 __inline static void lock(Lock *l) {
103 uintptr_t old = 0;
104 while (!__c11_atomic_compare_exchange_weak(l, &old, 1, __ATOMIC_ACQUIRE,
105 __ATOMIC_RELAXED))
106 old = 0;
107 }
108 /// locks for atomic operations
109 static Lock locks[SPINLOCK_COUNT];
110 #endif
111
112 /// Returns a lock to use for a given pointer.
lock_for_pointer(void * ptr)113 static __inline Lock *lock_for_pointer(void *ptr) {
114 intptr_t hash = (intptr_t)ptr;
115 // Disregard the lowest 4 bits. We want all values that may be part of the
116 // same memory operation to hash to the same value and therefore use the same
117 // lock.
118 hash >>= 4;
119 // Use the next bits as the basis for the hash
120 intptr_t low = hash & SPINLOCK_MASK;
121 // Now use the high(er) set of bits to perturb the hash, so that we don't
122 // get collisions from atomic fields in a single object
123 hash >>= 16;
124 hash ^= low;
125 // Return a pointer to the word to use
126 return locks + (hash & SPINLOCK_MASK);
127 }
128
129 /// Macros for determining whether a size is lock free.
130 #define ATOMIC_ALWAYS_LOCK_FREE_OR_ALIGNED_LOCK_FREE(size, p) \
131 (__atomic_always_lock_free(size, p) || \
132 (__atomic_always_lock_free(size, 0) && ((uintptr_t)p % size) == 0))
133 #define IS_LOCK_FREE_1(p) ATOMIC_ALWAYS_LOCK_FREE_OR_ALIGNED_LOCK_FREE(1, p)
134 #define IS_LOCK_FREE_2(p) ATOMIC_ALWAYS_LOCK_FREE_OR_ALIGNED_LOCK_FREE(2, p)
135 #define IS_LOCK_FREE_4(p) ATOMIC_ALWAYS_LOCK_FREE_OR_ALIGNED_LOCK_FREE(4, p)
136 #define IS_LOCK_FREE_8(p) ATOMIC_ALWAYS_LOCK_FREE_OR_ALIGNED_LOCK_FREE(8, p)
137 #define IS_LOCK_FREE_16(p) ATOMIC_ALWAYS_LOCK_FREE_OR_ALIGNED_LOCK_FREE(16, p)
138
139 /// Macro that calls the compiler-generated lock-free versions of functions
140 /// when they exist.
141 #define TRY_LOCK_FREE_CASE(n, type, ptr) \
142 case n: \
143 if (IS_LOCK_FREE_##n(ptr)) { \
144 LOCK_FREE_ACTION(type); \
145 } \
146 break;
147 #ifdef __SIZEOF_INT128__
148 #define TRY_LOCK_FREE_CASE_16(p) TRY_LOCK_FREE_CASE(16, __uint128_t, p)
149 #else
150 #define TRY_LOCK_FREE_CASE_16(p) /* __uint128_t not available */
151 #endif
152
153 #define LOCK_FREE_CASES(ptr) \
154 do { \
155 switch (size) { \
156 TRY_LOCK_FREE_CASE(1, uint8_t, ptr) \
157 TRY_LOCK_FREE_CASE(2, uint16_t, ptr) \
158 TRY_LOCK_FREE_CASE(4, uint32_t, ptr) \
159 TRY_LOCK_FREE_CASE(8, uint64_t, ptr) \
160 TRY_LOCK_FREE_CASE_16(ptr) /* __uint128_t may not be supported */ \
161 default: \
162 break; \
163 } \
164 } while (0)
165
166 /// Whether atomic operations for the given size (and alignment) are lock-free.
__atomic_is_lock_free_c(size_t size,void * ptr)167 bool __atomic_is_lock_free_c(size_t size, void *ptr) {
168 #define LOCK_FREE_ACTION(type) return true;
169 LOCK_FREE_CASES(ptr);
170 #undef LOCK_FREE_ACTION
171 return false;
172 }
173
174 /// An atomic load operation. This is atomic with respect to the source
175 /// pointer only.
__atomic_load_c(int size,void * src,void * dest,int model)176 void __atomic_load_c(int size, void *src, void *dest, int model) {
177 #define LOCK_FREE_ACTION(type) \
178 *((type *)dest) = __c11_atomic_load((_Atomic(type) *)src, model); \
179 return;
180 LOCK_FREE_CASES(src);
181 #undef LOCK_FREE_ACTION
182 Lock *l = lock_for_pointer(src);
183 lock(l);
184 memcpy(dest, src, size);
185 unlock(l);
186 }
187
188 /// An atomic store operation. This is atomic with respect to the destination
189 /// pointer only.
__atomic_store_c(int size,void * dest,void * src,int model)190 void __atomic_store_c(int size, void *dest, void *src, int model) {
191 #define LOCK_FREE_ACTION(type) \
192 __c11_atomic_store((_Atomic(type) *)dest, *(type *)src, model); \
193 return;
194 LOCK_FREE_CASES(dest);
195 #undef LOCK_FREE_ACTION
196 Lock *l = lock_for_pointer(dest);
197 lock(l);
198 memcpy(dest, src, size);
199 unlock(l);
200 }
201
202 /// Atomic compare and exchange operation. If the value at *ptr is identical
203 /// to the value at *expected, then this copies value at *desired to *ptr. If
204 /// they are not, then this stores the current value from *ptr in *expected.
205 ///
206 /// This function returns 1 if the exchange takes place or 0 if it fails.
__atomic_compare_exchange_c(int size,void * ptr,void * expected,void * desired,int success,int failure)207 int __atomic_compare_exchange_c(int size, void *ptr, void *expected,
208 void *desired, int success, int failure) {
209 #define LOCK_FREE_ACTION(type) \
210 return __c11_atomic_compare_exchange_strong( \
211 (_Atomic(type) *)ptr, (type *)expected, *(type *)desired, success, \
212 failure)
213 LOCK_FREE_CASES(ptr);
214 #undef LOCK_FREE_ACTION
215 Lock *l = lock_for_pointer(ptr);
216 lock(l);
217 if (memcmp(ptr, expected, size) == 0) {
218 memcpy(ptr, desired, size);
219 unlock(l);
220 return 1;
221 }
222 memcpy(expected, ptr, size);
223 unlock(l);
224 return 0;
225 }
226
227 /// Performs an atomic exchange operation between two pointers. This is atomic
228 /// with respect to the target address.
__atomic_exchange_c(int size,void * ptr,void * val,void * old,int model)229 void __atomic_exchange_c(int size, void *ptr, void *val, void *old, int model) {
230 #define LOCK_FREE_ACTION(type) \
231 *(type *)old = \
232 __c11_atomic_exchange((_Atomic(type) *)ptr, *(type *)val, model); \
233 return;
234 LOCK_FREE_CASES(ptr);
235 #undef LOCK_FREE_ACTION
236 Lock *l = lock_for_pointer(ptr);
237 lock(l);
238 memcpy(old, ptr, size);
239 memcpy(ptr, val, size);
240 unlock(l);
241 }
242
243 ////////////////////////////////////////////////////////////////////////////////
244 // Where the size is known at compile time, the compiler may emit calls to
245 // specialised versions of the above functions.
246 ////////////////////////////////////////////////////////////////////////////////
247 #ifdef __SIZEOF_INT128__
248 #define OPTIMISED_CASES \
249 OPTIMISED_CASE(1, IS_LOCK_FREE_1, uint8_t) \
250 OPTIMISED_CASE(2, IS_LOCK_FREE_2, uint16_t) \
251 OPTIMISED_CASE(4, IS_LOCK_FREE_4, uint32_t) \
252 OPTIMISED_CASE(8, IS_LOCK_FREE_8, uint64_t) \
253 OPTIMISED_CASE(16, IS_LOCK_FREE_16, __uint128_t)
254 #else
255 #define OPTIMISED_CASES \
256 OPTIMISED_CASE(1, IS_LOCK_FREE_1, uint8_t) \
257 OPTIMISED_CASE(2, IS_LOCK_FREE_2, uint16_t) \
258 OPTIMISED_CASE(4, IS_LOCK_FREE_4, uint32_t) \
259 OPTIMISED_CASE(8, IS_LOCK_FREE_8, uint64_t)
260 #endif
261
262 #define OPTIMISED_CASE(n, lockfree, type) \
263 type __atomic_load_##n(type *src, int model) { \
264 if (lockfree(src)) \
265 return __c11_atomic_load((_Atomic(type) *)src, model); \
266 Lock *l = lock_for_pointer(src); \
267 lock(l); \
268 type val = *src; \
269 unlock(l); \
270 return val; \
271 }
272 OPTIMISED_CASES
273 #undef OPTIMISED_CASE
274
275 #define OPTIMISED_CASE(n, lockfree, type) \
276 void __atomic_store_##n(type *dest, type val, int model) { \
277 if (lockfree(dest)) { \
278 __c11_atomic_store((_Atomic(type) *)dest, val, model); \
279 return; \
280 } \
281 Lock *l = lock_for_pointer(dest); \
282 lock(l); \
283 *dest = val; \
284 unlock(l); \
285 return; \
286 }
287 OPTIMISED_CASES
288 #undef OPTIMISED_CASE
289
290 #define OPTIMISED_CASE(n, lockfree, type) \
291 type __atomic_exchange_##n(type *dest, type val, int model) { \
292 if (lockfree(dest)) \
293 return __c11_atomic_exchange((_Atomic(type) *)dest, val, model); \
294 Lock *l = lock_for_pointer(dest); \
295 lock(l); \
296 type tmp = *dest; \
297 *dest = val; \
298 unlock(l); \
299 return tmp; \
300 }
301 OPTIMISED_CASES
302 #undef OPTIMISED_CASE
303
304 #define OPTIMISED_CASE(n, lockfree, type) \
305 bool __atomic_compare_exchange_##n(type *ptr, type *expected, type desired, \
306 int success, int failure) { \
307 if (lockfree(ptr)) \
308 return __c11_atomic_compare_exchange_strong( \
309 (_Atomic(type) *)ptr, expected, desired, success, failure); \
310 Lock *l = lock_for_pointer(ptr); \
311 lock(l); \
312 if (*ptr == *expected) { \
313 *ptr = desired; \
314 unlock(l); \
315 return true; \
316 } \
317 *expected = *ptr; \
318 unlock(l); \
319 return false; \
320 }
321 OPTIMISED_CASES
322 #undef OPTIMISED_CASE
323
324 ////////////////////////////////////////////////////////////////////////////////
325 // Atomic read-modify-write operations for integers of various sizes.
326 ////////////////////////////////////////////////////////////////////////////////
327 #define ATOMIC_RMW(n, lockfree, type, opname, op) \
328 type __atomic_fetch_##opname##_##n(type *ptr, type val, int model) { \
329 if (lockfree(ptr)) \
330 return __c11_atomic_fetch_##opname((_Atomic(type) *)ptr, val, model); \
331 Lock *l = lock_for_pointer(ptr); \
332 lock(l); \
333 type tmp = *ptr; \
334 *ptr = tmp op val; \
335 unlock(l); \
336 return tmp; \
337 }
338
339 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, add, +)
340 OPTIMISED_CASES
341 #undef OPTIMISED_CASE
342 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, sub, -)
343 OPTIMISED_CASES
344 #undef OPTIMISED_CASE
345 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, and, &)
346 OPTIMISED_CASES
347 #undef OPTIMISED_CASE
348 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, or, |)
349 OPTIMISED_CASES
350 #undef OPTIMISED_CASE
351 #define OPTIMISED_CASE(n, lockfree, type) ATOMIC_RMW(n, lockfree, type, xor, ^)
352 OPTIMISED_CASES
353 #undef OPTIMISED_CASE
354