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 #if defined(SMP) 78 #if defined(LOCORE) 79 #define MPLOCKED lock ; 80 #else 81 #define MPLOCKED "lock ; " 82 #endif 83 #else 84 #define MPLOCKED 85 #endif 86 87 /* 88 * The assembly is volatilized to demark potential before-and-after side 89 * effects if an interrupt or SMP collision were to occur. 90 */ 91 #define ATOMIC_ASM(NAME, TYPE, OP, V) \ 92 static __inline void \ 93 atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ 94 { \ 95 __asm __volatile(MPLOCKED OP \ 96 : "=m" (*p) \ 97 : "0" (*p), "ir" (V)); \ 98 } 99 100 /* 101 * Atomic compare and set, used by the mutex functions 102 * 103 * if (*dst == exp) *dst = src (all 32 bit words) 104 * 105 * Returns 0 on failure, non-zero on success 106 */ 107 108 #if defined(I386_CPU) 109 static __inline int 110 atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src) 111 { 112 int res = exp; 113 114 __asm __volatile( 115 " pushfl ; " 116 " cli ; " 117 " cmpl %1,%3 ; " 118 " jne 1f ; " 119 " movl %2,%3 ; " 120 "1: " 121 " sete %%al; " 122 " movzbl %%al,%0 ; " 123 " popfl ; " 124 "# atomic_cmpset_int" 125 : "=a" (res) /* 0 (result) */ 126 : "0" (exp), /* 1 */ 127 "r" (src), /* 2 */ 128 "m" (*(dst)) /* 3 */ 129 : "memory"); 130 131 return (res); 132 } 133 #else /* defined(I386_CPU) */ 134 static __inline int 135 atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src) 136 { 137 int res = exp; 138 139 __asm __volatile ( 140 " " MPLOCKED " " 141 " cmpxchgl %2,%3 ; " 142 " setz %%al ; " 143 " movzbl %%al,%0 ; " 144 "1: " 145 "# atomic_cmpset_int" 146 : "=a" (res) /* 0 (result) */ 147 : "0" (exp), /* 1 */ 148 "r" (src), /* 2 */ 149 "m" (*(dst)) /* 3 */ 150 : "memory"); 151 152 return (res); 153 } 154 #endif /* defined(I386_CPU) */ 155 156 #if defined(I386_CPU) 157 /* 158 * We assume that a = b will do atomic loads and stores. 159 * 160 * XXX: This is _NOT_ safe on a P6 or higher because it does not guarantee 161 * memory ordering. These should only be used on a 386. 162 */ 163 #define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \ 164 static __inline u_##TYPE \ 165 atomic_load_acq_##TYPE(volatile u_##TYPE *p) \ 166 { \ 167 return (*p); \ 168 } \ 169 \ 170 static __inline void \ 171 atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ 172 { \ 173 *p = v; \ 174 __asm __volatile("" : : : "memory"); \ 175 } 176 #else 177 178 #define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \ 179 static __inline u_##TYPE \ 180 atomic_load_acq_##TYPE(volatile u_##TYPE *p) \ 181 { \ 182 u_##TYPE res; \ 183 \ 184 __asm __volatile(MPLOCKED LOP \ 185 : "=a" (res), /* 0 (result) */\ 186 "+m" (*p) /* 1 */ \ 187 : : "cc", "memory"); \ 188 \ 189 return (res); \ 190 } \ 191 \ 192 /* \ 193 * The XCHG instruction asserts LOCK automagically. \ 194 */ \ 195 static __inline void \ 196 atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\ 197 { \ 198 __asm __volatile(SOP \ 199 : "+m" (*p), /* 0 */ \ 200 "+r" (v) /* 1 */ \ 201 : : "memory"); \ 202 } 203 #endif /* defined(I386_CPU) */ 204 #endif /* KLD_MODULE */ 205 206 ATOMIC_ASM(set, char, "orb %b2,%0", v) 207 ATOMIC_ASM(clear, char, "andb %b2,%0", ~v) 208 ATOMIC_ASM(add, char, "addb %b2,%0", v) 209 ATOMIC_ASM(subtract, char, "subb %b2,%0", v) 210 211 ATOMIC_ASM(set, short, "orw %w2,%0", v) 212 ATOMIC_ASM(clear, short, "andw %w2,%0", ~v) 213 ATOMIC_ASM(add, short, "addw %w2,%0", v) 214 ATOMIC_ASM(subtract, short, "subw %w2,%0", v) 215 216 ATOMIC_ASM(set, int, "orl %2,%0", v) 217 ATOMIC_ASM(clear, int, "andl %2,%0", ~v) 218 ATOMIC_ASM(add, int, "addl %2,%0", v) 219 ATOMIC_ASM(subtract, int, "subl %2,%0", v) 220 221 ATOMIC_ASM(set, long, "orl %2,%0", v) 222 ATOMIC_ASM(clear, long, "andl %2,%0", ~v) 223 ATOMIC_ASM(add, long, "addl %2,%0", v) 224 ATOMIC_ASM(subtract, long, "subl %2,%0", v) 225 226 ATOMIC_STORE_LOAD(char, "cmpxchgb %b0,%1", "xchgb %b1,%0") 227 ATOMIC_STORE_LOAD(short,"cmpxchgw %w0,%1", "xchgw %w1,%0") 228 ATOMIC_STORE_LOAD(int, "cmpxchgl %0,%1", "xchgl %1,%0") 229 ATOMIC_STORE_LOAD(long, "cmpxchgl %0,%1", "xchgl %1,%0") 230 231 #undef ATOMIC_ASM 232 #undef ATOMIC_STORE_LOAD 233 234 #define atomic_set_acq_char atomic_set_char 235 #define atomic_set_rel_char atomic_set_char 236 #define atomic_clear_acq_char atomic_clear_char 237 #define atomic_clear_rel_char atomic_clear_char 238 #define atomic_add_acq_char atomic_add_char 239 #define atomic_add_rel_char atomic_add_char 240 #define atomic_subtract_acq_char atomic_subtract_char 241 #define atomic_subtract_rel_char atomic_subtract_char 242 243 #define atomic_set_acq_short atomic_set_short 244 #define atomic_set_rel_short atomic_set_short 245 #define atomic_clear_acq_short atomic_clear_short 246 #define atomic_clear_rel_short atomic_clear_short 247 #define atomic_add_acq_short atomic_add_short 248 #define atomic_add_rel_short atomic_add_short 249 #define atomic_subtract_acq_short atomic_subtract_short 250 #define atomic_subtract_rel_short atomic_subtract_short 251 252 #define atomic_set_acq_int atomic_set_int 253 #define atomic_set_rel_int atomic_set_int 254 #define atomic_clear_acq_int atomic_clear_int 255 #define atomic_clear_rel_int atomic_clear_int 256 #define atomic_add_acq_int atomic_add_int 257 #define atomic_add_rel_int atomic_add_int 258 #define atomic_subtract_acq_int atomic_subtract_int 259 #define atomic_subtract_rel_int atomic_subtract_int 260 #define atomic_cmpset_acq_int atomic_cmpset_int 261 #define atomic_cmpset_rel_int atomic_cmpset_int 262 263 #define atomic_set_acq_long atomic_set_long 264 #define atomic_set_rel_long atomic_set_long 265 #define atomic_clear_acq_long atomic_clear_long 266 #define atomic_clear_rel_long atomic_clear_long 267 #define atomic_add_acq_long atomic_add_long 268 #define atomic_add_rel_long atomic_add_long 269 #define atomic_subtract_acq_long atomic_subtract_long 270 #define atomic_subtract_rel_long atomic_subtract_long 271 #define atomic_cmpset_long atomic_cmpset_int 272 #define atomic_cmpset_acq_long atomic_cmpset_acq_int 273 #define atomic_cmpset_rel_long atomic_cmpset_rel_int 274 275 #define atomic_cmpset_acq_ptr atomic_cmpset_ptr 276 #define atomic_cmpset_rel_ptr atomic_cmpset_ptr 277 278 #define atomic_set_8 atomic_set_char 279 #define atomic_set_acq_8 atomic_set_acq_char 280 #define atomic_set_rel_8 atomic_set_rel_char 281 #define atomic_clear_8 atomic_clear_char 282 #define atomic_clear_acq_8 atomic_clear_acq_char 283 #define atomic_clear_rel_8 atomic_clear_rel_char 284 #define atomic_add_8 atomic_add_char 285 #define atomic_add_acq_8 atomic_add_acq_char 286 #define atomic_add_rel_8 atomic_add_rel_char 287 #define atomic_subtract_8 atomic_subtract_char 288 #define atomic_subtract_acq_8 atomic_subtract_acq_char 289 #define atomic_subtract_rel_8 atomic_subtract_rel_char 290 #define atomic_load_acq_8 atomic_load_acq_char 291 #define atomic_store_rel_8 atomic_store_rel_char 292 293 #define atomic_set_16 atomic_set_short 294 #define atomic_set_acq_16 atomic_set_acq_short 295 #define atomic_set_rel_16 atomic_set_rel_short 296 #define atomic_clear_16 atomic_clear_short 297 #define atomic_clear_acq_16 atomic_clear_acq_short 298 #define atomic_clear_rel_16 atomic_clear_rel_short 299 #define atomic_add_16 atomic_add_short 300 #define atomic_add_acq_16 atomic_add_acq_short 301 #define atomic_add_rel_16 atomic_add_rel_short 302 #define atomic_subtract_16 atomic_subtract_short 303 #define atomic_subtract_acq_16 atomic_subtract_acq_short 304 #define atomic_subtract_rel_16 atomic_subtract_rel_short 305 #define atomic_load_acq_16 atomic_load_acq_short 306 #define atomic_store_rel_16 atomic_store_rel_short 307 308 #define atomic_set_32 atomic_set_int 309 #define atomic_set_acq_32 atomic_set_acq_int 310 #define atomic_set_rel_32 atomic_set_rel_int 311 #define atomic_clear_32 atomic_clear_int 312 #define atomic_clear_acq_32 atomic_clear_acq_int 313 #define atomic_clear_rel_32 atomic_clear_rel_int 314 #define atomic_add_32 atomic_add_int 315 #define atomic_add_acq_32 atomic_add_acq_int 316 #define atomic_add_rel_32 atomic_add_rel_int 317 #define atomic_subtract_32 atomic_subtract_int 318 #define atomic_subtract_acq_32 atomic_subtract_acq_int 319 #define atomic_subtract_rel_32 atomic_subtract_rel_int 320 #define atomic_load_acq_32 atomic_load_acq_int 321 #define atomic_store_rel_32 atomic_store_rel_int 322 #define atomic_cmpset_32 atomic_cmpset_int 323 #define atomic_cmpset_acq_32 atomic_cmpset_acq_int 324 #define atomic_cmpset_rel_32 atomic_cmpset_rel_int 325 #define atomic_readandclear_32 atomic_readandclear_int 326 327 #if !defined(WANT_FUNCTIONS) 328 static __inline int 329 atomic_cmpset_ptr(volatile void *dst, void *exp, void *src) 330 { 331 332 return (atomic_cmpset_int((volatile u_int *)dst, (u_int)exp, 333 (u_int)src)); 334 } 335 336 static __inline void * 337 atomic_load_acq_ptr(volatile void *p) 338 { 339 return (void *)atomic_load_acq_int((volatile u_int *)p); 340 } 341 342 static __inline void 343 atomic_store_rel_ptr(volatile void *p, void *v) 344 { 345 atomic_store_rel_int((volatile u_int *)p, (u_int)v); 346 } 347 348 #define ATOMIC_PTR(NAME) \ 349 static __inline void \ 350 atomic_##NAME##_ptr(volatile void *p, uintptr_t v) \ 351 { \ 352 atomic_##NAME##_int((volatile u_int *)p, v); \ 353 } \ 354 \ 355 static __inline void \ 356 atomic_##NAME##_acq_ptr(volatile void *p, uintptr_t v) \ 357 { \ 358 atomic_##NAME##_acq_int((volatile u_int *)p, v);\ 359 } \ 360 \ 361 static __inline void \ 362 atomic_##NAME##_rel_ptr(volatile void *p, uintptr_t v) \ 363 { \ 364 atomic_##NAME##_rel_int((volatile u_int *)p, v);\ 365 } 366 367 ATOMIC_PTR(set) 368 ATOMIC_PTR(clear) 369 ATOMIC_PTR(add) 370 ATOMIC_PTR(subtract) 371 372 #undef ATOMIC_PTR 373 374 static __inline u_int 375 atomic_readandclear_int(volatile u_int *addr) 376 { 377 u_int result; 378 379 __asm __volatile ( 380 " xorl %0,%0 ; " 381 " xchgl %1,%0 ; " 382 "# atomic_readandclear_int" 383 : "=&r" (result) /* 0 (result) */ 384 : "m" (*addr)); /* 1 (addr) */ 385 386 return (result); 387 } 388 389 static __inline u_long 390 atomic_readandclear_long(volatile u_long *addr) 391 { 392 u_long result; 393 394 __asm __volatile ( 395 " xorl %0,%0 ; " 396 " xchgl %1,%0 ; " 397 "# atomic_readandclear_int" 398 : "=&r" (result) /* 0 (result) */ 399 : "m" (*addr)); /* 1 (addr) */ 400 401 return (result); 402 } 403 #endif /* !defined(WANT_FUNCTIONS) */ 404 #endif /* ! _MACHINE_ATOMIC_H_ */ 405