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