1 /*- 2 * Copyright (c) 2013 Andrew Turner <andrew@freebsd.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD$ 27 */ 28 29 #ifndef _MACHINE_ATOMIC_H_ 30 #define _MACHINE_ATOMIC_H_ 31 32 #define isb() __asm __volatile("isb" : : : "memory") 33 34 /* 35 * Options for DMB and DSB: 36 * oshld Outer Shareable, load 37 * oshst Outer Shareable, store 38 * osh Outer Shareable, all 39 * nshld Non-shareable, load 40 * nshst Non-shareable, store 41 * nsh Non-shareable, all 42 * ishld Inner Shareable, load 43 * ishst Inner Shareable, store 44 * ish Inner Shareable, all 45 * ld Full system, load 46 * st Full system, store 47 * sy Full system, all 48 */ 49 #define dsb(opt) __asm __volatile("dsb " __STRING(opt) : : : "memory") 50 #define dmb(opt) __asm __volatile("dmb " __STRING(opt) : : : "memory") 51 52 #define mb() dmb(sy) /* Full system memory barrier all */ 53 #define wmb() dmb(st) /* Full system memory barrier store */ 54 #define rmb() dmb(ld) /* Full system memory barrier load */ 55 56 #define ATOMIC_OP(op, asm_op, bar, a, l) \ 57 static __inline void \ 58 atomic_##op##_##bar##32(volatile uint32_t *p, uint32_t val) \ 59 { \ 60 uint32_t tmp; \ 61 int res; \ 62 \ 63 __asm __volatile( \ 64 "1: ld"#a"xr %w0, [%2] \n" \ 65 " "#asm_op" %w0, %w0, %w3 \n" \ 66 " st"#l"xr %w1, %w0, [%2] \n" \ 67 " cbnz %w1, 1b \n" \ 68 : "=&r"(tmp), "=&r"(res) \ 69 : "r" (p), "r" (val) \ 70 : "memory" \ 71 ); \ 72 } \ 73 \ 74 static __inline void \ 75 atomic_##op##_##bar##64(volatile uint64_t *p, uint64_t val) \ 76 { \ 77 uint64_t tmp; \ 78 int res; \ 79 \ 80 __asm __volatile( \ 81 "1: ld"#a"xr %0, [%2] \n" \ 82 " "#asm_op" %0, %0, %3 \n" \ 83 " st"#l"xr %w1, %0, [%2] \n" \ 84 " cbnz %w1, 1b \n" \ 85 : "=&r"(tmp), "=&r"(res) \ 86 : "r" (p), "r" (val) \ 87 : "memory" \ 88 ); \ 89 } 90 91 #define ATOMIC(op, asm_op) \ 92 ATOMIC_OP(op, asm_op, , , ) \ 93 ATOMIC_OP(op, asm_op, acq_, a, ) \ 94 ATOMIC_OP(op, asm_op, rel_, , l) \ 95 96 ATOMIC(add, add) 97 ATOMIC(clear, bic) 98 ATOMIC(set, orr) 99 ATOMIC(subtract, sub) 100 101 #define ATOMIC_FCMPSET(bar, a, l) \ 102 static __inline int \ 103 atomic_fcmpset_##bar##32(volatile uint32_t *p, uint32_t *cmpval, \ 104 uint32_t newval) \ 105 { \ 106 uint32_t tmp; \ 107 uint32_t _cmpval = *cmpval; \ 108 int res; \ 109 \ 110 __asm __volatile( \ 111 "1: mov %w1, #1 \n" \ 112 " ld"#a"xr %w0, [%2] \n" \ 113 " cmp %w0, %w3 \n" \ 114 " b.ne 2f \n" \ 115 " st"#l"xr %w1, %w4, [%2] \n" \ 116 "2:" \ 117 : "=&r"(tmp), "=&r"(res) \ 118 : "r" (p), "r" (_cmpval), "r" (newval) \ 119 : "cc", "memory" \ 120 ); \ 121 *cmpval = tmp; \ 122 \ 123 return (!res); \ 124 } \ 125 \ 126 static __inline int \ 127 atomic_fcmpset_##bar##64(volatile uint64_t *p, uint64_t *cmpval, \ 128 uint64_t newval) \ 129 { \ 130 uint64_t tmp; \ 131 uint64_t _cmpval = *cmpval; \ 132 int res; \ 133 \ 134 __asm __volatile( \ 135 "1: mov %w1, #1 \n" \ 136 " ld"#a"xr %0, [%2] \n" \ 137 " cmp %0, %3 \n" \ 138 " b.ne 2f \n" \ 139 " st"#l"xr %w1, %4, [%2] \n" \ 140 "2:" \ 141 : "=&r"(tmp), "=&r"(res) \ 142 : "r" (p), "r" (_cmpval), "r" (newval) \ 143 : "cc", "memory" \ 144 ); \ 145 *cmpval = tmp; \ 146 \ 147 return (!res); \ 148 } 149 150 ATOMIC_FCMPSET( , , ) 151 ATOMIC_FCMPSET(acq_, a, ) 152 ATOMIC_FCMPSET(rel_, ,l) 153 154 #undef ATOMIC_FCMPSET 155 156 #define ATOMIC_CMPSET(bar, a, l) \ 157 static __inline int \ 158 atomic_cmpset_##bar##32(volatile uint32_t *p, uint32_t cmpval, \ 159 uint32_t newval) \ 160 { \ 161 uint32_t tmp; \ 162 int res; \ 163 \ 164 __asm __volatile( \ 165 "1: mov %w1, #1 \n" \ 166 " ld"#a"xr %w0, [%2] \n" \ 167 " cmp %w0, %w3 \n" \ 168 " b.ne 2f \n" \ 169 " st"#l"xr %w1, %w4, [%2] \n" \ 170 " cbnz %w1, 1b \n" \ 171 "2:" \ 172 : "=&r"(tmp), "=&r"(res) \ 173 : "r" (p), "r" (cmpval), "r" (newval) \ 174 : "cc", "memory" \ 175 ); \ 176 \ 177 return (!res); \ 178 } \ 179 \ 180 static __inline int \ 181 atomic_cmpset_##bar##64(volatile uint64_t *p, uint64_t cmpval, \ 182 uint64_t newval) \ 183 { \ 184 uint64_t tmp; \ 185 int res; \ 186 \ 187 __asm __volatile( \ 188 "1: mov %w1, #1 \n" \ 189 " ld"#a"xr %0, [%2] \n" \ 190 " cmp %0, %3 \n" \ 191 " b.ne 2f \n" \ 192 " st"#l"xr %w1, %4, [%2] \n" \ 193 " cbnz %w1, 1b \n" \ 194 "2:" \ 195 : "=&r"(tmp), "=&r"(res) \ 196 : "r" (p), "r" (cmpval), "r" (newval) \ 197 : "cc", "memory" \ 198 ); \ 199 \ 200 return (!res); \ 201 } 202 203 ATOMIC_CMPSET( , , ) 204 ATOMIC_CMPSET(acq_, a, ) 205 ATOMIC_CMPSET(rel_, ,l) 206 207 static __inline uint32_t 208 atomic_fetchadd_32(volatile uint32_t *p, uint32_t val) 209 { 210 uint32_t tmp, ret; 211 int res; 212 213 __asm __volatile( 214 "1: ldxr %w2, [%3] \n" 215 " add %w0, %w2, %w4 \n" 216 " stxr %w1, %w0, [%3] \n" 217 " cbnz %w1, 1b \n" 218 : "=&r"(tmp), "=&r"(res), "=&r"(ret) 219 : "r" (p), "r" (val) 220 : "memory" 221 ); 222 223 return (ret); 224 } 225 226 static __inline uint64_t 227 atomic_fetchadd_64(volatile uint64_t *p, uint64_t val) 228 { 229 uint64_t tmp, ret; 230 int res; 231 232 __asm __volatile( 233 "1: ldxr %2, [%3] \n" 234 " add %0, %2, %4 \n" 235 " stxr %w1, %0, [%3] \n" 236 " cbnz %w1, 1b \n" 237 : "=&r"(tmp), "=&r"(res), "=&r"(ret) 238 : "r" (p), "r" (val) 239 : "memory" 240 ); 241 242 return (ret); 243 } 244 245 static __inline uint32_t 246 atomic_readandclear_32(volatile uint32_t *p) 247 { 248 uint32_t ret; 249 int res; 250 251 __asm __volatile( 252 "1: ldxr %w1, [%2] \n" 253 " stxr %w0, wzr, [%2] \n" 254 " cbnz %w0, 1b \n" 255 : "=&r"(res), "=&r"(ret) 256 : "r" (p) 257 : "memory" 258 ); 259 260 return (ret); 261 } 262 263 static __inline uint64_t 264 atomic_readandclear_64(volatile uint64_t *p) 265 { 266 uint64_t ret; 267 int res; 268 269 __asm __volatile( 270 "1: ldxr %1, [%2] \n" 271 " stxr %w0, xzr, [%2] \n" 272 " cbnz %w0, 1b \n" 273 : "=&r"(res), "=&r"(ret) 274 : "r" (p) 275 : "memory" 276 ); 277 278 return (ret); 279 } 280 281 static __inline uint32_t 282 atomic_swap_32(volatile uint32_t *p, uint32_t val) 283 { 284 uint32_t ret; 285 int res; 286 287 __asm __volatile( 288 "1: ldxr %w0, [%2] \n" 289 " stxr %w1, %w3, [%2] \n" 290 " cbnz %w1, 1b \n" 291 : "=&r"(ret), "=&r"(res) 292 : "r" (p), "r" (val) 293 : "memory" 294 ); 295 296 return (ret); 297 } 298 299 static __inline uint64_t 300 atomic_swap_64(volatile uint64_t *p, uint64_t val) 301 { 302 uint64_t ret; 303 int res; 304 305 __asm __volatile( 306 "1: ldxr %0, [%2] \n" 307 " stxr %w1, %3, [%2] \n" 308 " cbnz %w1, 1b \n" 309 : "=&r"(ret), "=&r"(res) 310 : "r" (p), "r" (val) 311 : "memory" 312 ); 313 314 return (ret); 315 } 316 317 static __inline uint32_t 318 atomic_load_acq_32(volatile uint32_t *p) 319 { 320 uint32_t ret; 321 322 __asm __volatile( 323 "ldar %w0, [%1] \n" 324 : "=&r" (ret) 325 : "r" (p) 326 : "memory"); 327 328 return (ret); 329 } 330 331 static __inline uint64_t 332 atomic_load_acq_64(volatile uint64_t *p) 333 { 334 uint64_t ret; 335 336 __asm __volatile( 337 "ldar %0, [%1] \n" 338 : "=&r" (ret) 339 : "r" (p) 340 : "memory"); 341 342 return (ret); 343 } 344 345 static __inline void 346 atomic_store_rel_32(volatile uint32_t *p, uint32_t val) 347 { 348 349 __asm __volatile( 350 "stlr %w0, [%1] \n" 351 : 352 : "r" (val), "r" (p) 353 : "memory"); 354 } 355 356 static __inline void 357 atomic_store_rel_64(volatile uint64_t *p, uint64_t val) 358 { 359 360 __asm __volatile( 361 "stlr %0, [%1] \n" 362 : 363 : "r" (val), "r" (p) 364 : "memory"); 365 } 366 367 368 #define atomic_add_int atomic_add_32 369 #define atomic_fcmpset_int atomic_fcmpset_32 370 #define atomic_clear_int atomic_clear_32 371 #define atomic_cmpset_int atomic_cmpset_32 372 #define atomic_fetchadd_int atomic_fetchadd_32 373 #define atomic_readandclear_int atomic_readandclear_32 374 #define atomic_set_int atomic_set_32 375 #define atomic_swap_int atomic_swap_32 376 #define atomic_subtract_int atomic_subtract_32 377 378 #define atomic_add_acq_int atomic_add_acq_32 379 #define atomic_fcmpset_acq_int atomic_fcmpset_acq_32 380 #define atomic_clear_acq_int atomic_clear_acq_32 381 #define atomic_cmpset_acq_int atomic_cmpset_acq_32 382 #define atomic_load_acq_int atomic_load_acq_32 383 #define atomic_set_acq_int atomic_set_acq_32 384 #define atomic_subtract_acq_int atomic_subtract_acq_32 385 386 #define atomic_add_rel_int atomic_add_rel_32 387 #define atomic_fcmpset_rel_int atomic_fcmpset_rel_32 388 #define atomic_clear_rel_int atomic_clear_rel_32 389 #define atomic_cmpset_rel_int atomic_cmpset_rel_32 390 #define atomic_set_rel_int atomic_set_rel_32 391 #define atomic_subtract_rel_int atomic_subtract_rel_32 392 #define atomic_store_rel_int atomic_store_rel_32 393 394 #define atomic_add_long atomic_add_64 395 #define atomic_fcmpset_long atomic_fcmpset_64 396 #define atomic_clear_long atomic_clear_64 397 #define atomic_cmpset_long atomic_cmpset_64 398 #define atomic_fetchadd_long atomic_fetchadd_64 399 #define atomic_readandclear_long atomic_readandclear_64 400 #define atomic_set_long atomic_set_64 401 #define atomic_swap_long atomic_swap_64 402 #define atomic_subtract_long atomic_subtract_64 403 404 #define atomic_add_ptr atomic_add_64 405 #define atomic_fcmpset_ptr atomic_fcmpset_64 406 #define atomic_clear_ptr atomic_clear_64 407 #define atomic_cmpset_ptr atomic_cmpset_64 408 #define atomic_fetchadd_ptr atomic_fetchadd_64 409 #define atomic_readandclear_ptr atomic_readandclear_64 410 #define atomic_set_ptr atomic_set_64 411 #define atomic_swap_ptr atomic_swap_64 412 #define atomic_subtract_ptr atomic_subtract_64 413 414 #define atomic_add_acq_long atomic_add_acq_64 415 #define atomic_fcmpset_acq_long atomic_fcmpset_acq_64 416 #define atomic_clear_acq_long atomic_clear_acq_64 417 #define atomic_cmpset_acq_long atomic_cmpset_acq_64 418 #define atomic_load_acq_long atomic_load_acq_64 419 #define atomic_set_acq_long atomic_set_acq_64 420 #define atomic_subtract_acq_long atomic_subtract_acq_64 421 422 #define atomic_add_acq_ptr atomic_add_acq_64 423 #define atomic_fcmpset_acq_ptr atomic_fcmpset_acq_64 424 #define atomic_clear_acq_ptr atomic_clear_acq_64 425 #define atomic_cmpset_acq_ptr atomic_cmpset_acq_64 426 #define atomic_load_acq_ptr atomic_load_acq_64 427 #define atomic_set_acq_ptr atomic_set_acq_64 428 #define atomic_subtract_acq_ptr atomic_subtract_acq_64 429 430 #define atomic_add_rel_long atomic_add_rel_64 431 #define atomic_fcmpset_rel_long atomic_fcmpset_rel_64 432 #define atomic_clear_rel_long atomic_clear_rel_64 433 #define atomic_cmpset_rel_long atomic_cmpset_rel_64 434 #define atomic_set_rel_long atomic_set_rel_64 435 #define atomic_subtract_rel_long atomic_subtract_rel_64 436 #define atomic_store_rel_long atomic_store_rel_64 437 438 #define atomic_add_rel_ptr atomic_add_rel_64 439 #define atomic_fcmpset_rel_ptr atomic_fcmpset_rel_64 440 #define atomic_clear_rel_ptr atomic_clear_rel_64 441 #define atomic_cmpset_rel_ptr atomic_cmpset_rel_64 442 #define atomic_set_rel_ptr atomic_set_rel_64 443 #define atomic_subtract_rel_ptr atomic_subtract_rel_64 444 #define atomic_store_rel_ptr atomic_store_rel_64 445 446 static __inline void 447 atomic_thread_fence_acq(void) 448 { 449 450 dmb(ld); 451 } 452 453 static __inline void 454 atomic_thread_fence_rel(void) 455 { 456 457 dmb(sy); 458 } 459 460 static __inline void 461 atomic_thread_fence_acq_rel(void) 462 { 463 464 dmb(sy); 465 } 466 467 static __inline void 468 atomic_thread_fence_seq_cst(void) 469 { 470 471 dmb(sy); 472 } 473 474 #endif /* _MACHINE_ATOMIC_H_ */ 475 476