1 /* 2 * Copyright (c) 1992, 1993, 1994, 1995, 1996 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that: (1) source code distributions 7 * retain the above copyright notice and this paragraph in its entirety, (2) 8 * distributions including binary code include the above copyright notice and 9 * this paragraph in its entirety in the documentation or other materials 10 * provided with the distribution, and (3) all advertising materials mentioning 11 * features or use of this software display the following acknowledgement: 12 * ``This product includes software developed by the University of California, 13 * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of 14 * the University nor the names of its contributors may be used to endorse 15 * or promote products derived from this software without specific prior 16 * written permission. 17 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED 18 * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF 19 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. 20 */ 21 22 #ifdef __NetBSD__ 23 #include <string.h> 24 25 /* 26 * Do it the portable way and let the compiler optimize the code 27 */ 28 static inline uint16_t EXTRACT_16BITS(const void *p) 29 { 30 uint16_t t; 31 memcpy(&t, p, sizeof(t)); 32 return ntohs(t); 33 } 34 35 static inline uint32_t EXTRACT_24BITS(const void *p) 36 { 37 uint8_t t[3]; 38 memcpy(t, p, sizeof(t)); 39 return 40 ((uint32_t)t[0] << 16) | 41 ((uint32_t)t[1] << 8) | 42 t[2]; 43 } 44 45 static inline uint32_t EXTRACT_32BITS(const void *p) 46 { 47 uint32_t t; 48 memcpy(&t, p, sizeof(t)); 49 return ntohl(t); 50 } 51 52 53 static inline uint64_t EXTRACT_48BITS(const void *p) 54 { 55 uint8_t t[6]; 56 memcpy(t, p, sizeof(t)); 57 return 58 ((uint64_t)t[0] << 40) | 59 ((uint64_t)t[1] << 32) | 60 ((uint64_t)t[2] << 24) | 61 ((uint64_t)t[3] << 16) | 62 ((uint64_t)t[4] << 8) | 63 t[5]; 64 } 65 66 static inline uint64_t EXTRACT_64BITS(const void *p) 67 { 68 uint32_t t[2]; 69 memcpy(&t[0], p, sizeof(t[0])); 70 memcpy(&t[1], (const uint8_t *)p + sizeof(t[0]), sizeof(t[1])); 71 return ((uint64_t)ntohl(t[0]) << 32) | ntohl(t[1]); 72 } 73 74 static inline uint8_t EXTRACT_LE_8BITS(const void *p) 75 { 76 uint8_t t[1]; 77 memcpy(t, p, sizeof(t)); 78 return t[0]; 79 } 80 81 static inline uint16_t EXTRACT_LE_16BITS(const void *p) 82 { 83 uint8_t t[2]; 84 memcpy(t, p, sizeof(t)); 85 return 86 ((uint16_t)t[1] << 8) | 87 t[0]; 88 } 89 90 static inline uint32_t EXTRACT_LE_24BITS(const void *p) 91 { 92 uint8_t t[3]; 93 memcpy(t, p, sizeof(t)); 94 return 95 ((uint32_t)t[2] << 16) | 96 ((uint32_t)t[1] << 8) | 97 t[0]; 98 } 99 100 static inline uint32_t EXTRACT_LE_32BITS(const void *p) 101 { 102 uint8_t t[4]; 103 memcpy(t, p, sizeof(t)); 104 return 105 ((uint32_t)t[3] << 24) | 106 ((uint32_t)t[2] << 16) | 107 ((uint32_t)t[1] << 8) | 108 t[0]; 109 } 110 111 static inline uint64_t EXTRACT_LE_64BITS(const void *p) 112 { 113 uint8_t t[8]; 114 memcpy(&t, p, sizeof(t)); 115 return 116 ((uint64_t)t[7] << 56) | 117 ((uint64_t)t[6] << 48) | 118 ((uint64_t)t[5] << 40) | 119 ((uint64_t)t[4] << 32) | 120 ((uint64_t)t[3] << 24) | 121 ((uint64_t)t[2] << 16) | 122 ((uint64_t)t[1] << 8) | 123 t[0]; 124 } 125 126 #else /* Fast & Loose */ 127 /* 128 * Macros to extract possibly-unaligned big-endian integral values. 129 */ 130 #ifdef LBL_ALIGN 131 /* 132 * The processor doesn't natively handle unaligned loads. 133 */ 134 #if defined(__GNUC__) && defined(HAVE___ATTRIBUTE__) && \ 135 (defined(__alpha) || defined(__alpha__) || \ 136 defined(__mips) || defined(__mips__)) 137 138 /* 139 * This is a GCC-compatible compiler and we have __attribute__, which 140 * we assume that mean we have __attribute__((packed)), and this is 141 * MIPS or Alpha, which has instructions that can help when doing 142 * unaligned loads. 143 * 144 * Declare packed structures containing a uint16_t and a uint32_t, 145 * cast the pointer to point to one of those, and fetch through it; 146 * the GCC manual doesn't appear to explicitly say that 147 * __attribute__((packed)) causes the compiler to generate unaligned-safe 148 * code, but it apppears to do so. 149 * 150 * We do this in case the compiler can generate code using those 151 * instructions to do an unaligned load and pass stuff to "ntohs()" or 152 * "ntohl()", which might be better than than the code to fetch the 153 * bytes one at a time and assemble them. (That might not be the 154 * case on a little-endian platform, such as DEC's MIPS machines and 155 * Alpha machines, where "ntohs()" and "ntohl()" might not be done 156 * inline.) 157 * 158 * We do this only for specific architectures because, for example, 159 * at least some versions of GCC, when compiling for 64-bit SPARC, 160 * generate code that assumes alignment if we do this. 161 * 162 * XXX - add other architectures and compilers as possible and 163 * appropriate. 164 * 165 * HP's C compiler, indicated by __HP_cc being defined, supports 166 * "#pragma unaligned N" in version A.05.50 and later, where "N" 167 * specifies a number of bytes at which the typedef on the next 168 * line is aligned, e.g. 169 * 170 * #pragma unalign 1 171 * typedef uint16_t unaligned_uint16_t; 172 * 173 * to define unaligned_uint16_t as a 16-bit unaligned data type. 174 * This could be presumably used, in sufficiently recent versions of 175 * the compiler, with macros similar to those below. This would be 176 * useful only if that compiler could generate better code for PA-RISC 177 * or Itanium than would be generated by a bunch of shifts-and-ORs. 178 * 179 * DEC C, indicated by __DECC being defined, has, at least on Alpha, 180 * an __unaligned qualifier that can be applied to pointers to get the 181 * compiler to generate code that does unaligned loads and stores when 182 * dereferencing the pointer in question. 183 * 184 * XXX - what if the native C compiler doesn't support 185 * __attribute__((packed))? How can we get it to generate unaligned 186 * accesses for *specific* items? 187 */ 188 typedef struct { 189 uint16_t val; 190 } __attribute__((packed)) unaligned_uint16_t; 191 192 typedef struct { 193 uint32_t val; 194 } __attribute__((packed)) unaligned_uint32_t; 195 196 static inline uint16_t 197 EXTRACT_16BITS(const void *p) 198 { 199 return ((uint16_t)ntohs(((const unaligned_uint16_t *)(p))->val)); 200 } 201 202 static inline uint32_t 203 EXTRACT_32BITS(const void *p) 204 { 205 return ((uint32_t)ntohl(((const unaligned_uint32_t *)(p))->val)); 206 } 207 208 static inline uint64_t 209 EXTRACT_64BITS(const void *p) 210 { 211 return ((uint64_t)(((uint64_t)ntohl(((const unaligned_uint32_t *)(p) + 0)->val)) << 32 | \ 212 ((uint64_t)ntohl(((const unaligned_uint32_t *)(p) + 1)->val)) << 0)); 213 } 214 215 #else /* have to do it a byte at a time */ 216 /* 217 * This isn't a GCC-compatible compiler, we don't have __attribute__, 218 * or we do but we don't know of any better way with this instruction 219 * set to do unaligned loads, so do unaligned loads of big-endian 220 * quantities the hard way - fetch the bytes one at a time and 221 * assemble them. 222 */ 223 #define EXTRACT_16BITS(p) \ 224 ((uint16_t)(((uint16_t)(*((const uint8_t *)(p) + 0)) << 8) | \ 225 ((uint16_t)(*((const uint8_t *)(p) + 1)) << 0))) 226 #define EXTRACT_32BITS(p) \ 227 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 0)) << 24) | \ 228 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 16) | \ 229 ((uint32_t)(*((const uint8_t *)(p) + 2)) << 8) | \ 230 ((uint32_t)(*((const uint8_t *)(p) + 3)) << 0))) 231 #define EXTRACT_64BITS(p) \ 232 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 56) | \ 233 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 48) | \ 234 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 40) | \ 235 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 32) | \ 236 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 24) | \ 237 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 16) | \ 238 ((uint64_t)(*((const uint8_t *)(p) + 6)) << 8) | \ 239 ((uint64_t)(*((const uint8_t *)(p) + 7)) << 0))) 240 #endif /* must special-case unaligned accesses */ 241 #else /* LBL_ALIGN */ 242 /* 243 * The processor natively handles unaligned loads, so we can just 244 * cast the pointer and fetch through it. 245 */ 246 static inline uint16_t 247 EXTRACT_16BITS(const void *p) 248 { 249 return ((uint16_t)ntohs(*(const uint16_t *)(p))); 250 } 251 252 static inline uint32_t 253 EXTRACT_32BITS(const void *p) 254 { 255 return ((uint32_t)ntohl(*(const uint32_t *)(p))); 256 } 257 258 static inline uint64_t 259 EXTRACT_64BITS(const void *p) 260 { 261 return ((uint64_t)(((uint64_t)ntohl(*((const uint32_t *)(p) + 0))) << 32 | \ 262 ((uint64_t)ntohl(*((const uint32_t *)(p) + 1))) << 0)); 263 264 } 265 266 #endif /* LBL_ALIGN */ 267 268 #define EXTRACT_24BITS(p) \ 269 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 0)) << 16) | \ 270 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 271 ((uint32_t)(*((const uint8_t *)(p) + 2)) << 0))) 272 273 #define EXTRACT_40BITS(p) \ 274 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 32) | \ 275 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 24) | \ 276 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 277 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 8) | \ 278 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 0))) 279 280 #define EXTRACT_48BITS(p) \ 281 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 40) | \ 282 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 32) | \ 283 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 24) | \ 284 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 16) | \ 285 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 8) | \ 286 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 0))) 287 288 #define EXTRACT_56BITS(p) \ 289 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 48) | \ 290 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 40) | \ 291 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 32) | \ 292 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 24) | \ 293 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 16) | \ 294 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 8) | \ 295 ((uint64_t)(*((const uint8_t *)(p) + 6)) << 0))) 296 297 /* 298 * Macros to extract possibly-unaligned little-endian integral values. 299 * XXX - do loads on little-endian machines that support unaligned loads? 300 */ 301 #define EXTRACT_LE_8BITS(p) (*(p)) 302 #define EXTRACT_LE_16BITS(p) \ 303 ((uint16_t)(((uint16_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 304 ((uint16_t)(*((const uint8_t *)(p) + 0)) << 0))) 305 #define EXTRACT_LE_32BITS(p) \ 306 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 3)) << 24) | \ 307 ((uint32_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 308 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 309 ((uint32_t)(*((const uint8_t *)(p) + 0)) << 0))) 310 #define EXTRACT_LE_24BITS(p) \ 311 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 312 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 313 ((uint32_t)(*((const uint8_t *)(p) + 0)) << 0))) 314 #define EXTRACT_LE_64BITS(p) \ 315 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 7)) << 56) | \ 316 ((uint64_t)(*((const uint8_t *)(p) + 6)) << 48) | \ 317 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 40) | \ 318 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 32) | \ 319 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 24) | \ 320 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 321 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 322 ((uint64_t)(*((const uint8_t *)(p) + 0)) << 0))) 323 #endif /* __NetBSD__ */ 324