| 1 | #ifndef _I386_STRING_H_ |
| 2 | #define _I386_STRING_H_ |
| 3 | |
| 4 | #ifdef __KERNEL__ |
| 5 | |
| 6 | /* Let gcc decide whether to inline or use the out of line functions */ |
| 7 | |
| 8 | #define __HAVE_ARCH_STRCPY |
| 9 | extern char *strcpy(char *dest, const char *src); |
| 10 | |
| 11 | #define __HAVE_ARCH_STRNCPY |
| 12 | extern char *strncpy(char *dest, const char *src, size_t count); |
| 13 | |
| 14 | #define __HAVE_ARCH_STRCAT |
| 15 | extern char *strcat(char *dest, const char *src); |
| 16 | |
| 17 | #define __HAVE_ARCH_STRNCAT |
| 18 | extern char *strncat(char *dest, const char *src, size_t count); |
| 19 | |
| 20 | #define __HAVE_ARCH_STRCMP |
| 21 | extern int strcmp(const char *cs, const char *ct); |
| 22 | |
| 23 | #define __HAVE_ARCH_STRNCMP |
| 24 | extern int strncmp(const char *cs, const char *ct, size_t count); |
| 25 | |
| 26 | #define __HAVE_ARCH_STRCHR |
| 27 | extern char *strchr(const char *s, int c); |
| 28 | |
| 29 | #define __HAVE_ARCH_STRLEN |
| 30 | extern size_t strlen(const char *s); |
| 31 | |
| 32 | static __always_inline void *__memcpy(void *to, const void *from, size_t n) |
| 33 | { |
| 34 | int d0, d1, d2; |
| 35 | asm volatile("rep ; movsl\n\t" |
| 36 | "movl %4,%%ecx\n\t" |
| 37 | "andl $3,%%ecx\n\t" |
| 38 | "jz 1f\n\t" |
| 39 | "rep ; movsb\n\t" |
| 40 | "1:" |
| 41 | : "=&c" (d0), "=&D" (d1), "=&S" (d2) |
| 42 | : "0" (n / 4), "g" (n), "1" ((long)to), "2" ((long)from) |
| 43 | : "memory"); |
| 44 | return to; |
| 45 | } |
| 46 | |
| 47 | /* |
| 48 | * This looks ugly, but the compiler can optimize it totally, |
| 49 | * as the count is constant. |
| 50 | */ |
| 51 | static __always_inline void *__constant_memcpy(void *to, const void *from, |
| 52 | size_t n) |
| 53 | { |
| 54 | long esi, edi; |
| 55 | if (!n) |
| 56 | return to; |
| 57 | |
| 58 | switch (n) { |
| 59 | case 1: |
| 60 | *(char *)to = *(char *)from; |
| 61 | return to; |
| 62 | case 2: |
| 63 | *(short *)to = *(short *)from; |
| 64 | return to; |
| 65 | case 4: |
| 66 | *(int *)to = *(int *)from; |
| 67 | return to; |
| 68 | |
| 69 | case 3: |
| 70 | *(short *)to = *(short *)from; |
| 71 | *((char *)to + 2) = *((char *)from + 2); |
| 72 | return to; |
| 73 | case 5: |
| 74 | *(int *)to = *(int *)from; |
| 75 | *((char *)to + 4) = *((char *)from + 4); |
| 76 | return to; |
| 77 | case 6: |
| 78 | *(int *)to = *(int *)from; |
| 79 | *((short *)to + 2) = *((short *)from + 2); |
| 80 | return to; |
| 81 | case 8: |
| 82 | *(int *)to = *(int *)from; |
| 83 | *((int *)to + 1) = *((int *)from + 1); |
| 84 | return to; |
| 85 | } |
| 86 | |
| 87 | esi = (long)from; |
| 88 | edi = (long)to; |
| 89 | if (n >= 5 * 4) { |
| 90 | /* large block: use rep prefix */ |
| 91 | int ecx; |
| 92 | asm volatile("rep ; movsl" |
| 93 | : "=&c" (ecx), "=&D" (edi), "=&S" (esi) |
| 94 | : "0" (n / 4), "1" (edi), "2" (esi) |
| 95 | : "memory" |
| 96 | ); |
| 97 | } else { |
| 98 | /* small block: don't clobber ecx + smaller code */ |
| 99 | if (n >= 4 * 4) |
| 100 | asm volatile("movsl" |
| 101 | : "=&D"(edi), "=&S"(esi) |
| 102 | : "0"(edi), "1"(esi) |
| 103 | : "memory"); |
| 104 | if (n >= 3 * 4) |
| 105 | asm volatile("movsl" |
| 106 | : "=&D"(edi), "=&S"(esi) |
| 107 | : "0"(edi), "1"(esi) |
| 108 | : "memory"); |
| 109 | if (n >= 2 * 4) |
| 110 | asm volatile("movsl" |
| 111 | : "=&D"(edi), "=&S"(esi) |
| 112 | : "0"(edi), "1"(esi) |
| 113 | : "memory"); |
| 114 | if (n >= 1 * 4) |
| 115 | asm volatile("movsl" |
| 116 | : "=&D"(edi), "=&S"(esi) |
| 117 | : "0"(edi), "1"(esi) |
| 118 | : "memory"); |
| 119 | } |
| 120 | switch (n % 4) { |
| 121 | /* tail */ |
| 122 | case 0: |
| 123 | return to; |
| 124 | case 1: |
| 125 | asm volatile("movsb" |
| 126 | : "=&D"(edi), "=&S"(esi) |
| 127 | : "0"(edi), "1"(esi) |
| 128 | : "memory"); |
| 129 | return to; |
| 130 | case 2: |
| 131 | asm volatile("movsw" |
| 132 | : "=&D"(edi), "=&S"(esi) |
| 133 | : "0"(edi), "1"(esi) |
| 134 | : "memory"); |
| 135 | return to; |
| 136 | default: |
| 137 | asm volatile("movsw\n\tmovsb" |
| 138 | : "=&D"(edi), "=&S"(esi) |
| 139 | : "0"(edi), "1"(esi) |
| 140 | : "memory"); |
| 141 | return to; |
| 142 | } |
| 143 | } |
| 144 | |
| 145 | #define __HAVE_ARCH_MEMCPY |
| 146 | |
| 147 | #ifdef CONFIG_X86_USE_3DNOW |
| 148 | |
| 149 | #include <asm/mmx.h> |
| 150 | |
| 151 | /* |
| 152 | * This CPU favours 3DNow strongly (eg AMD Athlon) |
| 153 | */ |
| 154 | |
| 155 | static inline void *__constant_memcpy3d(void *to, const void *from, size_t len) |
| 156 | { |
| 157 | if (len < 512) |
| 158 | return __constant_memcpy(to, from, len); |
| 159 | return _mmx_memcpy(to, from, len); |
| 160 | } |
| 161 | |
| 162 | static inline void *__memcpy3d(void *to, const void *from, size_t len) |
| 163 | { |
| 164 | if (len < 512) |
| 165 | return __memcpy(to, from, len); |
| 166 | return _mmx_memcpy(to, from, len); |
| 167 | } |
| 168 | |
| 169 | #define memcpy(t, f, n) \ |
| 170 | (__builtin_constant_p((n)) \ |
| 171 | ? __constant_memcpy3d((t), (f), (n)) \ |
| 172 | : __memcpy3d((t), (f), (n))) |
| 173 | |
| 174 | #else |
| 175 | |
| 176 | /* |
| 177 | * No 3D Now! |
| 178 | */ |
| 179 | |
| 180 | #define memcpy(t, f, n) \ |
| 181 | (__builtin_constant_p((n)) \ |
| 182 | ? __constant_memcpy((t), (f), (n)) \ |
| 183 | : __memcpy((t), (f), (n))) |
| 184 | |
| 185 | #endif |
| 186 | |
| 187 | #define __HAVE_ARCH_MEMMOVE |
| 188 | void *memmove(void *dest, const void *src, size_t n); |
| 189 | |
| 190 | #define memcmp __builtin_memcmp |
| 191 | |
| 192 | #define __HAVE_ARCH_MEMCHR |
| 193 | extern void *memchr(const void *cs, int c, size_t count); |
| 194 | |
| 195 | static inline void *__memset_generic(void *s, char c, size_t count) |
| 196 | { |
| 197 | int d0, d1; |
| 198 | asm volatile("rep\n\t" |
| 199 | "stosb" |
| 200 | : "=&c" (d0), "=&D" (d1) |
| 201 | : "a" (c), "1" (s), "0" (count) |
| 202 | : "memory"); |
| 203 | return s; |
| 204 | } |
| 205 | |
| 206 | /* we might want to write optimized versions of these later */ |
| 207 | #define __constant_count_memset(s, c, count) __memset_generic((s), (c), (count)) |
| 208 | |
| 209 | /* |
| 210 | * memset(x, 0, y) is a reasonably common thing to do, so we want to fill |
| 211 | * things 32 bits at a time even when we don't know the size of the |
| 212 | * area at compile-time.. |
| 213 | */ |
| 214 | static __always_inline |
| 215 | void *__constant_c_memset(void *s, unsigned long c, size_t count) |
| 216 | { |
| 217 | int d0, d1; |
| 218 | asm volatile("rep ; stosl\n\t" |
| 219 | "testb $2,%b3\n\t" |
| 220 | "je 1f\n\t" |
| 221 | "stosw\n" |
| 222 | "1:\ttestb $1,%b3\n\t" |
| 223 | "je 2f\n\t" |
| 224 | "stosb\n" |
| 225 | "2:" |
| 226 | : "=&c" (d0), "=&D" (d1) |
| 227 | : "a" (c), "q" (count), "0" (count/4), "1" ((long)s) |
| 228 | : "memory"); |
| 229 | return s; |
| 230 | } |
| 231 | |
| 232 | /* Added by Gertjan van Wingerde to make minix and sysv module work */ |
| 233 | #define __HAVE_ARCH_STRNLEN |
| 234 | extern size_t strnlen(const char *s, size_t count); |
| 235 | /* end of additional stuff */ |
| 236 | |
| 237 | #define __HAVE_ARCH_STRSTR |
| 238 | extern char *strstr(const char *cs, const char *ct); |
| 239 | |
| 240 | /* |
| 241 | * This looks horribly ugly, but the compiler can optimize it totally, |
| 242 | * as we by now know that both pattern and count is constant.. |
| 243 | */ |
| 244 | static __always_inline |
| 245 | void *__constant_c_and_count_memset(void *s, unsigned long pattern, |
| 246 | size_t count) |
| 247 | { |
| 248 | switch (count) { |
| 249 | case 0: |
| 250 | return s; |
| 251 | case 1: |
| 252 | *(unsigned char *)s = pattern & 0xff; |
| 253 | return s; |
| 254 | case 2: |
| 255 | *(unsigned short *)s = pattern & 0xffff; |
| 256 | return s; |
| 257 | case 3: |
| 258 | *(unsigned short *)s = pattern & 0xffff; |
| 259 | *((unsigned char *)s + 2) = pattern & 0xff; |
| 260 | return s; |
| 261 | case 4: |
| 262 | *(unsigned long *)s = pattern; |
| 263 | return s; |
| 264 | } |
| 265 | |
| 266 | #define COMMON(x) \ |
| 267 | asm volatile("rep ; stosl" \ |
| 268 | x \ |
| 269 | : "=&c" (d0), "=&D" (d1) \ |
| 270 | : "a" (eax), "0" (count/4), "1" ((long)s) \ |
| 271 | : "memory") |
| 272 | |
| 273 | { |
| 274 | int d0, d1; |
| 275 | #if __GNUC__ == 4 && __GNUC_MINOR__ == 0 |
| 276 | /* Workaround for broken gcc 4.0 */ |
| 277 | register unsigned long eax asm("%eax") = pattern; |
| 278 | #else |
| 279 | unsigned long eax = pattern; |
| 280 | #endif |
| 281 | |
| 282 | switch (count % 4) { |
| 283 | case 0: |
| 284 | COMMON(""); |
| 285 | return s; |
| 286 | case 1: |
| 287 | COMMON("\n\tstosb"); |
| 288 | return s; |
| 289 | case 2: |
| 290 | COMMON("\n\tstosw"); |
| 291 | return s; |
| 292 | default: |
| 293 | COMMON("\n\tstosw\n\tstosb"); |
| 294 | return s; |
| 295 | } |
| 296 | } |
| 297 | |
| 298 | #undef COMMON |
| 299 | } |
| 300 | |
| 301 | #define __constant_c_x_memset(s, c, count) \ |
| 302 | (__builtin_constant_p(count) \ |
| 303 | ? __constant_c_and_count_memset((s), (c), (count)) \ |
| 304 | : __constant_c_memset((s), (c), (count))) |
| 305 | |
| 306 | #define __memset(s, c, count) \ |
| 307 | (__builtin_constant_p(count) \ |
| 308 | ? __constant_count_memset((s), (c), (count)) \ |
| 309 | : __memset_generic((s), (c), (count))) |
| 310 | |
| 311 | #define __HAVE_ARCH_MEMSET |
| 312 | #define memset(s, c, count) \ |
| 313 | (__builtin_constant_p(c) \ |
| 314 | ? __constant_c_x_memset((s), (0x01010101UL * (unsigned char)(c)), \ |
| 315 | (count)) \ |
| 316 | : __memset((s), (c), (count))) |
| 317 | |
| 318 | /* |
| 319 | * find the first occurrence of byte 'c', or 1 past the area if none |
| 320 | */ |
| 321 | #define __HAVE_ARCH_MEMSCAN |
| 322 | extern void *memscan(void *addr, int c, size_t size); |
| 323 | |
| 324 | #endif /* __KERNEL__ */ |
| 325 | |
| 326 | #endif |