1 /* 2 * Copyright 2009-2015 Samy Al Bahra. 3 * Copyright 2012 João Fernandes. 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28 #ifndef CK_PR_PPC_H 29 #define CK_PR_PPC_H 30 31 #ifndef CK_PR_H 32 #error Do not include this file directly, use ck_pr.h 33 #endif 34 35 #include <ck_cc.h> 36 #include <ck_md.h> 37 38 /* 39 * The following represent supported atomic operations. 40 * These operations may be emulated. 41 */ 42 #include "ck_f_pr.h" 43 44 /* 45 * Minimum interface requirement met. 46 */ 47 #define CK_F_PR 48 49 /* 50 * This bounces the hardware thread from low to medium 51 * priority. I am unsure of the benefits of this approach 52 * but it is used by the Linux kernel. 53 */ 54 CK_CC_INLINE static void 55 ck_pr_stall(void) 56 { 57 58 __asm__ __volatile__("or 1, 1, 1;" 59 "or 2, 2, 2;" ::: "memory"); 60 return; 61 } 62 63 #define CK_PR_FENCE(T, I) \ 64 CK_CC_INLINE static void \ 65 ck_pr_fence_strict_##T(void) \ 66 { \ 67 __asm__ __volatile__(I ::: "memory"); \ 68 } 69 70 CK_PR_FENCE(atomic, "lwsync") 71 CK_PR_FENCE(atomic_store, "lwsync") 72 CK_PR_FENCE(atomic_load, "sync") 73 CK_PR_FENCE(store_atomic, "lwsync") 74 CK_PR_FENCE(load_atomic, "lwsync") 75 CK_PR_FENCE(store, "lwsync") 76 CK_PR_FENCE(store_load, "sync") 77 CK_PR_FENCE(load, "lwsync") 78 CK_PR_FENCE(load_store, "lwsync") 79 CK_PR_FENCE(memory, "sync") 80 CK_PR_FENCE(acquire, "lwsync") 81 CK_PR_FENCE(release, "lwsync") 82 CK_PR_FENCE(acqrel, "lwsync") 83 CK_PR_FENCE(lock, "lwsync") 84 CK_PR_FENCE(unlock, "lwsync") 85 86 #undef CK_PR_FENCE 87 88 #define CK_PR_LOAD(S, M, T, C, I) \ 89 CK_CC_INLINE static T \ 90 ck_pr_md_load_##S(const M *target) \ 91 { \ 92 T r; \ 93 __asm__ __volatile__(I "%U1%X1 %0, %1" \ 94 : "=r" (r) \ 95 : "m" (*(const C *)target) \ 96 : "memory"); \ 97 return (r); \ 98 } 99 100 CK_PR_LOAD(ptr, void, void *, uint32_t, "lwz") 101 102 #define CK_PR_LOAD_S(S, T, I) CK_PR_LOAD(S, T, T, T, I) 103 104 CK_PR_LOAD_S(32, uint32_t, "lwz") 105 CK_PR_LOAD_S(16, uint16_t, "lhz") 106 CK_PR_LOAD_S(8, uint8_t, "lbz") 107 CK_PR_LOAD_S(uint, unsigned int, "lwz") 108 CK_PR_LOAD_S(int, int, "lwz") 109 CK_PR_LOAD_S(short, short, "lhz") 110 CK_PR_LOAD_S(char, char, "lbz") 111 112 #undef CK_PR_LOAD_S 113 #undef CK_PR_LOAD 114 115 #define CK_PR_STORE(S, M, T, C, I) \ 116 CK_CC_INLINE static void \ 117 ck_pr_md_store_##S(M *target, T v) \ 118 { \ 119 __asm__ __volatile__(I "%U0%X0 %1, %0" \ 120 : "=m" (*(C *)target) \ 121 : "r" (v) \ 122 : "memory"); \ 123 return; \ 124 } 125 126 CK_PR_STORE(ptr, void, const void *, uint32_t, "stw") 127 128 #define CK_PR_STORE_S(S, T, I) CK_PR_STORE(S, T, T, T, I) 129 130 CK_PR_STORE_S(32, uint32_t, "stw") 131 CK_PR_STORE_S(16, uint16_t, "sth") 132 CK_PR_STORE_S(8, uint8_t, "stb") 133 CK_PR_STORE_S(uint, unsigned int, "stw") 134 CK_PR_STORE_S(int, int, "stw") 135 CK_PR_STORE_S(short, short, "sth") 136 CK_PR_STORE_S(char, char, "stb") 137 138 #undef CK_PR_STORE_S 139 #undef CK_PR_STORE 140 141 #define CK_PR_CAS(N, T, M) \ 142 CK_CC_INLINE static bool \ 143 ck_pr_cas_##N##_value(M *target, T compare, T set, M *value) \ 144 { \ 145 T previous; \ 146 __asm__ __volatile__("1:" \ 147 "lwarx %0, 0, %1;" \ 148 "cmpw 0, %0, %3;" \ 149 "bne- 2f;" \ 150 "stwcx. %2, 0, %1;" \ 151 "bne- 1b;" \ 152 "2:" \ 153 : "=&r" (previous) \ 154 : "r" (target), \ 155 "r" (set), \ 156 "r" (compare) \ 157 : "memory", "cc"); \ 158 *(T *)value = previous; \ 159 return (previous == compare); \ 160 } \ 161 CK_CC_INLINE static bool \ 162 ck_pr_cas_##N(M *target, T compare, T set) \ 163 { \ 164 T previous; \ 165 __asm__ __volatile__("1:" \ 166 "lwarx %0, 0, %1;" \ 167 "cmpw 0, %0, %3;" \ 168 "bne- 2f;" \ 169 "stwcx. %2, 0, %1;" \ 170 "bne- 1b;" \ 171 "2:" \ 172 : "=&r" (previous) \ 173 : "r" (target), \ 174 "r" (set), \ 175 "r" (compare) \ 176 : "memory", "cc"); \ 177 return (previous == compare); \ 178 } 179 180 CK_PR_CAS(ptr, void *, void) 181 #define CK_PR_CAS_S(a, b) CK_PR_CAS(a, b, b) 182 CK_PR_CAS_S(32, uint32_t) 183 CK_PR_CAS_S(uint, unsigned int) 184 CK_PR_CAS_S(int, int) 185 186 #undef CK_PR_CAS_S 187 #undef CK_PR_CAS 188 189 #define CK_PR_FAS(N, M, T, W) \ 190 CK_CC_INLINE static T \ 191 ck_pr_fas_##N(M *target, T v) \ 192 { \ 193 T previous; \ 194 __asm__ __volatile__("1:" \ 195 "l" W "arx %0, 0, %1;" \ 196 "st" W "cx. %2, 0, %1;" \ 197 "bne- 1b;" \ 198 : "=&r" (previous) \ 199 : "r" (target), \ 200 "r" (v) \ 201 : "memory", "cc"); \ 202 return (previous); \ 203 } 204 205 CK_PR_FAS(32, uint32_t, uint32_t, "w") 206 CK_PR_FAS(ptr, void, void *, "w") 207 CK_PR_FAS(int, int, int, "w") 208 CK_PR_FAS(uint, unsigned int, unsigned int, "w") 209 210 #undef CK_PR_FAS 211 212 #define CK_PR_UNARY(O, N, M, T, I, W) \ 213 CK_CC_INLINE static void \ 214 ck_pr_##O##_##N(M *target) \ 215 { \ 216 T previous; \ 217 __asm__ __volatile__("1:" \ 218 "l" W "arx %0, 0, %1;" \ 219 I ";" \ 220 "st" W "cx. %0, 0, %1;" \ 221 "bne- 1b;" \ 222 : "=&r" (previous) \ 223 : "r" (target) \ 224 : "memory", "cc"); \ 225 return; \ 226 } 227 228 CK_PR_UNARY(inc, ptr, void, void *, "addic %0, %0, 1", "w") 229 CK_PR_UNARY(dec, ptr, void, void *, "addic %0, %0, -1", "w") 230 CK_PR_UNARY(not, ptr, void, void *, "not %0, %0", "w") 231 CK_PR_UNARY(neg, ptr, void, void *, "neg %0, %0", "w") 232 233 #define CK_PR_UNARY_S(S, T, W) \ 234 CK_PR_UNARY(inc, S, T, T, "addic %0, %0, 1", W) \ 235 CK_PR_UNARY(dec, S, T, T, "addic %0, %0, -1", W) \ 236 CK_PR_UNARY(not, S, T, T, "not %0, %0", W) \ 237 CK_PR_UNARY(neg, S, T, T, "neg %0, %0", W) 238 239 CK_PR_UNARY_S(32, uint32_t, "w") 240 CK_PR_UNARY_S(uint, unsigned int, "w") 241 CK_PR_UNARY_S(int, int, "w") 242 243 #undef CK_PR_UNARY_S 244 #undef CK_PR_UNARY 245 246 #define CK_PR_BINARY(O, N, M, T, I, W) \ 247 CK_CC_INLINE static void \ 248 ck_pr_##O##_##N(M *target, T delta) \ 249 { \ 250 T previous; \ 251 __asm__ __volatile__("1:" \ 252 "l" W "arx %0, 0, %1;" \ 253 I " %0, %2, %0;" \ 254 "st" W "cx. %0, 0, %1;" \ 255 "bne- 1b;" \ 256 : "=&r" (previous) \ 257 : "r" (target), \ 258 "r" (delta) \ 259 : "memory", "cc"); \ 260 return; \ 261 } 262 263 CK_PR_BINARY(and, ptr, void, uintptr_t, "and", "w") 264 CK_PR_BINARY(add, ptr, void, uintptr_t, "add", "w") 265 CK_PR_BINARY(or, ptr, void, uintptr_t, "or", "w") 266 CK_PR_BINARY(sub, ptr, void, uintptr_t, "sub", "w") 267 CK_PR_BINARY(xor, ptr, void, uintptr_t, "xor", "w") 268 269 #define CK_PR_BINARY_S(S, T, W) \ 270 CK_PR_BINARY(and, S, T, T, "and", W) \ 271 CK_PR_BINARY(add, S, T, T, "add", W) \ 272 CK_PR_BINARY(or, S, T, T, "or", W) \ 273 CK_PR_BINARY(sub, S, T, T, "subf", W) \ 274 CK_PR_BINARY(xor, S, T, T, "xor", W) 275 276 CK_PR_BINARY_S(32, uint32_t, "w") 277 CK_PR_BINARY_S(uint, unsigned int, "w") 278 CK_PR_BINARY_S(int, int, "w") 279 280 #undef CK_PR_BINARY_S 281 #undef CK_PR_BINARY 282 283 CK_CC_INLINE static void * 284 ck_pr_faa_ptr(void *target, uintptr_t delta) 285 { 286 uintptr_t previous, r; 287 288 __asm__ __volatile__("1:" 289 "lwarx %0, 0, %2;" 290 "add %1, %3, %0;" 291 "stwcx. %1, 0, %2;" 292 "bne- 1b;" 293 : "=&r" (previous), 294 "=&r" (r) 295 : "r" (target), 296 "r" (delta) 297 : "memory", "cc"); 298 299 return (void *)(previous); 300 } 301 302 #define CK_PR_FAA(S, T, W) \ 303 CK_CC_INLINE static T \ 304 ck_pr_faa_##S(T *target, T delta) \ 305 { \ 306 T previous, r; \ 307 __asm__ __volatile__("1:" \ 308 "l" W "arx %0, 0, %2;" \ 309 "add %1, %3, %0;" \ 310 "st" W "cx. %1, 0, %2;" \ 311 "bne- 1b;" \ 312 : "=&r" (previous), \ 313 "=&r" (r) \ 314 : "r" (target), \ 315 "r" (delta) \ 316 : "memory", "cc"); \ 317 return (previous); \ 318 } 319 320 CK_PR_FAA(32, uint32_t, "w") 321 CK_PR_FAA(uint, unsigned int, "w") 322 CK_PR_FAA(int, int, "w") 323 324 #undef CK_PR_FAA 325 326 #endif /* CK_PR_PPC_H */ 327 328