1 /*- 2 * Copyright (c) 1998 Doug Rabson 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 #ifndef _MACHINE_ATOMIC_H_ 29 #define _MACHINE_ATOMIC_H_ 30 31 /* 32 * Various simple arithmetic on memory which is atomic in the presence 33 * of interrupts and multiple processors. 34 * 35 * atomic_set_char(P, V) (*(u_char*)(P) |= (V)) 36 * atomic_clear_char(P, V) (*(u_char*)(P) &= ~(V)) 37 * atomic_add_char(P, V) (*(u_char*)(P) += (V)) 38 * atomic_subtract_char(P, V) (*(u_char*)(P) -= (V)) 39 * 40 * atomic_set_short(P, V) (*(u_short*)(P) |= (V)) 41 * atomic_clear_short(P, V) (*(u_short*)(P) &= ~(V)) 42 * atomic_add_short(P, V) (*(u_short*)(P) += (V)) 43 * atomic_subtract_short(P, V) (*(u_short*)(P) -= (V)) 44 * 45 * atomic_set_int(P, V) (*(u_int*)(P) |= (V)) 46 * atomic_clear_int(P, V) (*(u_int*)(P) &= ~(V)) 47 * atomic_add_int(P, V) (*(u_int*)(P) += (V)) 48 * atomic_subtract_int(P, V) (*(u_int*)(P) -= (V)) 49 * atomic_readandclear_int(P) (return *(u_int*)P; *(u_int*)P = 0;) 50 * 51 * atomic_set_long(P, V) (*(u_long*)(P) |= (V)) 52 * atomic_clear_long(P, V) (*(u_long*)(P) &= ~(V)) 53 * atomic_add_long(P, V) (*(u_long*)(P) += (V)) 54 * atomic_subtract_long(P, V) (*(u_long*)(P) -= (V)) 55 * atomic_readandclear_long(P) (return *(u_long*)P; *(u_long*)P = 0;) 56 */ 57 58 /* 59 * The above functions are expanded inline in the statically-linked 60 * kernel. Lock prefixes are generated if an SMP kernel is being 61 * built. 62 * 63 * Kernel modules call real functions which are built into the kernel. 64 * This allows kernel modules to be portable between UP and SMP systems. 65 */ 66 #if defined(KLD_MODULE) 67 #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ 68 void atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v) 69 70 int atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src); 71 72 #define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \ 73 u_##TYPE atomic_load_acq_##TYPE(volatile u_##TYPE *p); \ 74 void atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v) 75 76 #else /* !KLD_MODULE */ 77 78 #ifdef __GNUC__ 79 80 /* 81 * For userland, assume the SMP case and use lock prefixes so that 82 * the binaries will run on both types of systems. 83 */ 84 #if defined(SMP) || !defined(_KERNEL) 85 #define MPLOCKED lock ; 86 #else 87 #define MPLOCKED 88 #endif 89 90 /* 91 * The assembly is volatilized to demark potential before-and-after side 92 * effects if an interrupt or SMP collision were to occur. 93 */ 94 #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ 95 static __inline void \ 96 atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ 97 { \ 98 __asm __volatile(__XSTRING(MPLOCKED) OP \ 99 : "+m" (*p) \ 100 : CONS (V)); \ 101 } 102 103 #else /* !__GNUC__ */ 104 105 #define ATOMIC_ASM(NAME, TYPE, OP, CONS, V) \ 106 extern void atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v) 107 108 #endif /* __GNUC__ */ 109 110 /* 111 * Atomic compare and set, used by the mutex functions 112 * 113 * if (*dst == exp) *dst = src (all 32 bit words) 114 * 115 * Returns 0 on failure, non-zero on success 116 */ 117 118 #if defined(__GNUC__) 119 120 #if defined(I386_CPU) || defined(CPU_DISABLE_CMPXCHG) 121 122 static __inline int 123 atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src) 124 { 125 int res = exp; 126 127 __asm __volatile( 128 " pushfl ; " 129 " cli ; " 130 " cmpl %0,%2 ; " 131 " jne 1f ; " 132 " movl %1,%2 ; " 133 "1: " 134 " sete %%al; " 135 " movzbl %%al,%0 ; " 136 " popfl ; " 137 "# atomic_cmpset_int" 138 : "+a" (res) /* 0 (result) */ 139 : "r" (src), /* 1 */ 140 "m" (*(dst)) /* 2 */ 141 : "memory"); 142 143 return (res); 144 } 145 146 #else /* defined(I386_CPU) */ 147 148 static __inline int 149 atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src) 150 { 151 int res = exp; 152 153 __asm __volatile ( 154 " " __XSTRING(MPLOCKED) " " 155 " cmpxchgl %1,%2 ; " 156 " setz %%al ; " 157 " movzbl %%al,%0 ; " 158 "1: " 159 "# atomic_cmpset_int" 160 : "+a" (res) /* 0 (result) */ 161 : "r" (src), /* 1 */ 162 "m" (*(dst)) /* 2 */ 163 : "memory"); 164 165 return (res); 166 } 167 168 #endif /* defined(I386_CPU) */ 169 170 #endif /* defined(__GNUC__) */ 171 172 #if defined(__GNUC__) 173 174 #if defined(I386_CPU) 175 176 /* 177 * We assume that a = b will do atomic loads and stores. 178 * 179 * XXX: This is _NOT_ safe on a P6 or higher because it does not guarantee 180 * memory ordering. These should only be used on a 386. 181 */ 182 #define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \ 183 static __inline u_##TYPE \ 184 atomic_load_acq_##TYPE(volatile u_##TYPE *p) \ 185 { \ 186 return (*p); \ 187 } \ 188 \ 189 static __inline void \ 190 atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ 191 { \ 192 *p = v; \ 193 __asm __volatile("" : : : "memory"); \ 194 } 195 196 #else /* !defined(I386_CPU) */ 197 198 #define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \ 199 static __inline u_##TYPE \ 200 atomic_load_acq_##TYPE(volatile u_##TYPE *p) \ 201 { \ 202 u_##TYPE res; \ 203 \ 204 __asm __volatile(__XSTRING(MPLOCKED) LOP \ 205 : "=a" (res), /* 0 (result) */\ 206 "+m" (*p) /* 1 */ \ 207 : : "memory"); \ 208 \ 209 return (res); \ 210 } \ 211 \ 212 /* \ 213 * The XCHG instruction asserts LOCK automagically. \ 214 */ \ 215 static __inline void \ 216 atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ 217 { \ 218 __asm __volatile(SOP \ 219 : "+m" (*p), /* 0 */ \ 220 "+r" (v) /* 1 */ \ 221 : : "memory"); \ 222 } 223 224 #endif /* defined(I386_CPU) */ 225 226 #else /* !defined(__GNUC__) */ 227 228 extern int atomic_cmpset_int(volatile u_int *, u_int, u_int); 229 230 #define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \ 231 extern u_##TYPE atomic_load_acq_##TYPE(volatile u_##TYPE *p); \ 232 extern void atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v) 233 234 #endif /* defined(__GNUC__) */ 235 236 #endif /* KLD_MODULE */ 237 238 ATOMIC_ASM(set, char, "orb %b1,%0", "iq", v); 239 ATOMIC_ASM(clear, char, "andb %b1,%0", "iq", ~v); 240 ATOMIC_ASM(add, char, "addb %b1,%0", "iq", v); 241 ATOMIC_ASM(subtract, char, "subb %b1,%0", "iq", v); 242 243 ATOMIC_ASM(set, short, "orw %w1,%0", "ir", v); 244 ATOMIC_ASM(clear, short, "andw %w1,%0", "ir", ~v); 245 ATOMIC_ASM(add, short, "addw %w1,%0", "ir", v); 246 ATOMIC_ASM(subtract, short, "subw %w1,%0", "ir", v); 247 248 ATOMIC_ASM(set, int, "orl %1,%0", "ir", v); 249 ATOMIC_ASM(clear, int, "andl %1,%0", "ir", ~v); 250 ATOMIC_ASM(add, int, "addl %1,%0", "ir", v); 251 ATOMIC_ASM(subtract, int, "subl %1,%0", "ir", v); 252 253 ATOMIC_ASM(set, long, "orl %1,%0", "ir", v); 254 ATOMIC_ASM(clear, long, "andl %1,%0", "ir", ~v); 255 ATOMIC_ASM(add, long, "addl %1,%0", "ir", v); 256 ATOMIC_ASM(subtract, long, "subl %1,%0", "ir", v); 257 258 ATOMIC_STORE_LOAD(char, "cmpxchgb %b0,%1", "xchgb %b1,%0"); 259 ATOMIC_STORE_LOAD(short,"cmpxchgw %w0,%1", "xchgw %w1,%0"); 260 ATOMIC_STORE_LOAD(int, "cmpxchgl %0,%1", "xchgl %1,%0"); 261 ATOMIC_STORE_LOAD(long, "cmpxchgl %0,%1", "xchgl %1,%0"); 262 263 #undef ATOMIC_ASM 264 #undef ATOMIC_STORE_LOAD 265 266 #define atomic_set_acq_char atomic_set_char 267 #define atomic_set_rel_char atomic_set_char 268 #define atomic_clear_acq_char atomic_clear_char 269 #define atomic_clear_rel_char atomic_clear_char 270 #define atomic_add_acq_char atomic_add_char 271 #define atomic_add_rel_char atomic_add_char 272 #define atomic_subtract_acq_char atomic_subtract_char 273 #define atomic_subtract_rel_char atomic_subtract_char 274 275 #define atomic_set_acq_short atomic_set_short 276 #define atomic_set_rel_short atomic_set_short 277 #define atomic_clear_acq_short atomic_clear_short 278 #define atomic_clear_rel_short atomic_clear_short 279 #define atomic_add_acq_short atomic_add_short 280 #define atomic_add_rel_short atomic_add_short 281 #define atomic_subtract_acq_short atomic_subtract_short 282 #define atomic_subtract_rel_short atomic_subtract_short 283 284 #define atomic_set_acq_int atomic_set_int 285 #define atomic_set_rel_int atomic_set_int 286 #define atomic_clear_acq_int atomic_clear_int 287 #define atomic_clear_rel_int atomic_clear_int 288 #define atomic_add_acq_int atomic_add_int 289 #define atomic_add_rel_int atomic_add_int 290 #define atomic_subtract_acq_int atomic_subtract_int 291 #define atomic_subtract_rel_int atomic_subtract_int 292 #define atomic_cmpset_acq_int atomic_cmpset_int 293 #define atomic_cmpset_rel_int atomic_cmpset_int 294 295 #define atomic_set_acq_long atomic_set_long 296 #define atomic_set_rel_long atomic_set_long 297 #define atomic_clear_acq_long atomic_clear_long 298 #define atomic_clear_rel_long atomic_clear_long 299 #define atomic_add_acq_long atomic_add_long 300 #define atomic_add_rel_long atomic_add_long 301 #define atomic_subtract_acq_long atomic_subtract_long 302 #define atomic_subtract_rel_long atomic_subtract_long 303 #define atomic_cmpset_long atomic_cmpset_int 304 #define atomic_cmpset_acq_long atomic_cmpset_acq_int 305 #define atomic_cmpset_rel_long atomic_cmpset_rel_int 306 307 #define atomic_cmpset_acq_ptr atomic_cmpset_ptr 308 #define atomic_cmpset_rel_ptr atomic_cmpset_ptr 309 310 #define atomic_set_8 atomic_set_char 311 #define atomic_set_acq_8 atomic_set_acq_char 312 #define atomic_set_rel_8 atomic_set_rel_char 313 #define atomic_clear_8 atomic_clear_char 314 #define atomic_clear_acq_8 atomic_clear_acq_char 315 #define atomic_clear_rel_8 atomic_clear_rel_char 316 #define atomic_add_8 atomic_add_char 317 #define atomic_add_acq_8 atomic_add_acq_char 318 #define atomic_add_rel_8 atomic_add_rel_char 319 #define atomic_subtract_8 atomic_subtract_char 320 #define atomic_subtract_acq_8 atomic_subtract_acq_char 321 #define atomic_subtract_rel_8 atomic_subtract_rel_char 322 #define atomic_load_acq_8 atomic_load_acq_char 323 #define atomic_store_rel_8 atomic_store_rel_char 324 325 #define atomic_set_16 atomic_set_short 326 #define atomic_set_acq_16 atomic_set_acq_short 327 #define atomic_set_rel_16 atomic_set_rel_short 328 #define atomic_clear_16 atomic_clear_short 329 #define atomic_clear_acq_16 atomic_clear_acq_short 330 #define atomic_clear_rel_16 atomic_clear_rel_short 331 #define atomic_add_16 atomic_add_short 332 #define atomic_add_acq_16 atomic_add_acq_short 333 #define atomic_add_rel_16 atomic_add_rel_short 334 #define atomic_subtract_16 atomic_subtract_short 335 #define atomic_subtract_acq_16 atomic_subtract_acq_short 336 #define atomic_subtract_rel_16 atomic_subtract_rel_short 337 #define atomic_load_acq_16 atomic_load_acq_short 338 #define atomic_store_rel_16 atomic_store_rel_short 339 340 #define atomic_set_32 atomic_set_int 341 #define atomic_set_acq_32 atomic_set_acq_int 342 #define atomic_set_rel_32 atomic_set_rel_int 343 #define atomic_clear_32 atomic_clear_int 344 #define atomic_clear_acq_32 atomic_clear_acq_int 345 #define atomic_clear_rel_32 atomic_clear_rel_int 346 #define atomic_add_32 atomic_add_int 347 #define atomic_add_acq_32 atomic_add_acq_int 348 #define atomic_add_rel_32 atomic_add_rel_int 349 #define atomic_subtract_32 atomic_subtract_int 350 #define atomic_subtract_acq_32 atomic_subtract_acq_int 351 #define atomic_subtract_rel_32 atomic_subtract_rel_int 352 #define atomic_load_acq_32 atomic_load_acq_int 353 #define atomic_store_rel_32 atomic_store_rel_int 354 #define atomic_cmpset_32 atomic_cmpset_int 355 #define atomic_cmpset_acq_32 atomic_cmpset_acq_int 356 #define atomic_cmpset_rel_32 atomic_cmpset_rel_int 357 #define atomic_readandclear_32 atomic_readandclear_int 358 359 #if !defined(WANT_FUNCTIONS) 360 static __inline int 361 atomic_cmpset_ptr(volatile void *dst, void *exp, void *src) 362 { 363 364 return (atomic_cmpset_int((volatile u_int *)dst, (u_int)exp, 365 (u_int)src)); 366 } 367 368 static __inline void * 369 atomic_load_acq_ptr(volatile void *p) 370 { 371 /* 372 * The apparently-bogus cast to intptr_t in the following is to 373 * avoid a warning from "gcc -Wbad-function-cast". 374 */ 375 return ((void *)(intptr_t)atomic_load_acq_int((volatile u_int *)p)); 376 } 377 378 static __inline void 379 atomic_store_rel_ptr(volatile void *p, void *v) 380 { 381 atomic_store_rel_int((volatile u_int *)p, (u_int)v); 382 } 383 384 #define ATOMIC_PTR(NAME) \ 385 static __inline void \ 386 atomic_##NAME##_ptr(volatile void *p, uintptr_t v) \ 387 { \ 388 atomic_##NAME##_int((volatile u_int *)p, v); \ 389 } \ 390 \ 391 static __inline void \ 392 atomic_##NAME##_acq_ptr(volatile void *p, uintptr_t v) \ 393 { \ 394 atomic_##NAME##_acq_int((volatile u_int *)p, v);\ 395 } \ 396 \ 397 static __inline void \ 398 atomic_##NAME##_rel_ptr(volatile void *p, uintptr_t v) \ 399 { \ 400 atomic_##NAME##_rel_int((volatile u_int *)p, v);\ 401 } 402 403 ATOMIC_PTR(set) 404 ATOMIC_PTR(clear) 405 ATOMIC_PTR(add) 406 ATOMIC_PTR(subtract) 407 408 #undef ATOMIC_PTR 409 410 #if defined(__GNUC__) 411 412 static __inline u_int 413 atomic_readandclear_int(volatile u_int *addr) 414 { 415 u_int result; 416 417 __asm __volatile ( 418 " xorl %0,%0 ; " 419 " xchgl %1,%0 ; " 420 "# atomic_readandclear_int" 421 : "=&r" (result) /* 0 (result) */ 422 : "m" (*addr)); /* 1 (addr) */ 423 424 return (result); 425 } 426 427 static __inline u_long 428 atomic_readandclear_long(volatile u_long *addr) 429 { 430 u_long result; 431 432 __asm __volatile ( 433 " xorl %0,%0 ; " 434 " xchgl %1,%0 ; " 435 "# atomic_readandclear_int" 436 : "=&r" (result) /* 0 (result) */ 437 : "m" (*addr)); /* 1 (addr) */ 438 439 return (result); 440 } 441 442 #else /* !defined(__GNUC__) */ 443 444 extern u_long atomic_readandclear_long(volatile u_long *); 445 extern u_int atomic_readandclear_int(volatile u_int *); 446 447 #endif /* defined(__GNUC__) */ 448 449 #endif /* !defined(WANT_FUNCTIONS) */ 450 #endif /* ! _MACHINE_ATOMIC_H_ */ 451