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 #ifndef _WIN32 23 #include <arpa/inet.h> 24 #endif 25 26 /* 27 * Macros to extract possibly-unaligned big-endian integral values. 28 */ 29 #ifdef LBL_ALIGN 30 /* 31 * The processor doesn't natively handle unaligned loads. 32 */ 33 #if defined(__GNUC__) && defined(HAVE___ATTRIBUTE__) && \ 34 (defined(__alpha) || defined(__alpha__) || \ 35 defined(__mips) || defined(__mips__)) 36 37 /* 38 * This is a GCC-compatible compiler and we have __attribute__, which 39 * we assume that mean we have __attribute__((packed)), and this is 40 * MIPS or Alpha, which has instructions that can help when doing 41 * unaligned loads. 42 * 43 * Declare packed structures containing a uint16_t and a uint32_t, 44 * cast the pointer to point to one of those, and fetch through it; 45 * the GCC manual doesn't appear to explicitly say that 46 * __attribute__((packed)) causes the compiler to generate unaligned-safe 47 * code, but it apppears to do so. 48 * 49 * We do this in case the compiler can generate code using those 50 * instructions to do an unaligned load and pass stuff to "ntohs()" or 51 * "ntohl()", which might be better than than the code to fetch the 52 * bytes one at a time and assemble them. (That might not be the 53 * case on a little-endian platform, such as DEC's MIPS machines and 54 * Alpha machines, where "ntohs()" and "ntohl()" might not be done 55 * inline.) 56 * 57 * We do this only for specific architectures because, for example, 58 * at least some versions of GCC, when compiling for 64-bit SPARC, 59 * generate code that assumes alignment if we do this. 60 * 61 * XXX - add other architectures and compilers as possible and 62 * appropriate. 63 * 64 * HP's C compiler, indicated by __HP_cc being defined, supports 65 * "#pragma unaligned N" in version A.05.50 and later, where "N" 66 * specifies a number of bytes at which the typedef on the next 67 * line is aligned, e.g. 68 * 69 * #pragma unalign 1 70 * typedef uint16_t unaligned_uint16_t; 71 * 72 * to define unaligned_uint16_t as a 16-bit unaligned data type. 73 * This could be presumably used, in sufficiently recent versions of 74 * the compiler, with macros similar to those below. This would be 75 * useful only if that compiler could generate better code for PA-RISC 76 * or Itanium than would be generated by a bunch of shifts-and-ORs. 77 * 78 * DEC C, indicated by __DECC being defined, has, at least on Alpha, 79 * an __unaligned qualifier that can be applied to pointers to get the 80 * compiler to generate code that does unaligned loads and stores when 81 * dereferencing the pointer in question. 82 * 83 * XXX - what if the native C compiler doesn't support 84 * __attribute__((packed))? How can we get it to generate unaligned 85 * accesses for *specific* items? 86 */ 87 typedef struct { 88 uint16_t val; 89 } __attribute__((packed)) unaligned_uint16_t; 90 91 typedef struct { 92 uint32_t val; 93 } __attribute__((packed)) unaligned_uint32_t; 94 95 static inline uint16_t 96 EXTRACT_16BITS(const void *p) 97 { 98 return ((uint16_t)ntohs(((const unaligned_uint16_t *)(p))->val)); 99 } 100 101 static inline uint32_t 102 EXTRACT_32BITS(const void *p) 103 { 104 return ((uint32_t)ntohl(((const unaligned_uint32_t *)(p))->val)); 105 } 106 107 static inline uint64_t 108 EXTRACT_64BITS(const void *p) 109 { 110 return ((uint64_t)(((uint64_t)ntohl(((const unaligned_uint32_t *)(p) + 0)->val)) << 32 | \ 111 ((uint64_t)ntohl(((const unaligned_uint32_t *)(p) + 1)->val)) << 0)); 112 } 113 114 #else /* have to do it a byte at a time */ 115 /* 116 * This isn't a GCC-compatible compiler, we don't have __attribute__, 117 * or we do but we don't know of any better way with this instruction 118 * set to do unaligned loads, so do unaligned loads of big-endian 119 * quantities the hard way - fetch the bytes one at a time and 120 * assemble them. 121 */ 122 #define EXTRACT_16BITS(p) \ 123 ((uint16_t)(((uint16_t)(*((const uint8_t *)(p) + 0)) << 8) | \ 124 ((uint16_t)(*((const uint8_t *)(p) + 1)) << 0))) 125 #define EXTRACT_32BITS(p) \ 126 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 0)) << 24) | \ 127 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 16) | \ 128 ((uint32_t)(*((const uint8_t *)(p) + 2)) << 8) | \ 129 ((uint32_t)(*((const uint8_t *)(p) + 3)) << 0))) 130 #define EXTRACT_64BITS(p) \ 131 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 56) | \ 132 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 48) | \ 133 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 40) | \ 134 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 32) | \ 135 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 24) | \ 136 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 16) | \ 137 ((uint64_t)(*((const uint8_t *)(p) + 6)) << 8) | \ 138 ((uint64_t)(*((const uint8_t *)(p) + 7)) << 0))) 139 #endif /* must special-case unaligned accesses */ 140 #else /* LBL_ALIGN */ 141 /* 142 * The processor natively handles unaligned loads, so we can just 143 * cast the pointer and fetch through it. 144 */ 145 static inline uint16_t 146 EXTRACT_16BITS(const void *p) 147 { 148 return ((uint16_t)ntohs(*(const uint16_t *)(p))); 149 } 150 151 static inline uint32_t 152 EXTRACT_32BITS(const void *p) 153 { 154 return ((uint32_t)ntohl(*(const uint32_t *)(p))); 155 } 156 157 static inline uint64_t 158 EXTRACT_64BITS(const void *p) 159 { 160 return ((uint64_t)(((uint64_t)ntohl(*((const uint32_t *)(p) + 0))) << 32 | \ 161 ((uint64_t)ntohl(*((const uint32_t *)(p) + 1))) << 0)); 162 163 } 164 165 #endif /* LBL_ALIGN */ 166 167 #define EXTRACT_24BITS(p) \ 168 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 0)) << 16) | \ 169 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 170 ((uint32_t)(*((const uint8_t *)(p) + 2)) << 0))) 171 172 #define EXTRACT_40BITS(p) \ 173 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 32) | \ 174 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 24) | \ 175 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 176 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 8) | \ 177 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 0))) 178 179 #define EXTRACT_48BITS(p) \ 180 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 40) | \ 181 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 32) | \ 182 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 24) | \ 183 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 16) | \ 184 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 8) | \ 185 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 0))) 186 187 #define EXTRACT_56BITS(p) \ 188 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 0)) << 48) | \ 189 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 40) | \ 190 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 32) | \ 191 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 24) | \ 192 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 16) | \ 193 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 8) | \ 194 ((uint64_t)(*((const uint8_t *)(p) + 6)) << 0))) 195 196 /* 197 * Macros to extract possibly-unaligned little-endian integral values. 198 * XXX - do loads on little-endian machines that support unaligned loads? 199 */ 200 #define EXTRACT_LE_8BITS(p) (*(p)) 201 #define EXTRACT_LE_16BITS(p) \ 202 ((uint16_t)(((uint16_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 203 ((uint16_t)(*((const uint8_t *)(p) + 0)) << 0))) 204 #define EXTRACT_LE_32BITS(p) \ 205 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 3)) << 24) | \ 206 ((uint32_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 207 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 208 ((uint32_t)(*((const uint8_t *)(p) + 0)) << 0))) 209 #define EXTRACT_LE_24BITS(p) \ 210 ((uint32_t)(((uint32_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 211 ((uint32_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 212 ((uint32_t)(*((const uint8_t *)(p) + 0)) << 0))) 213 #define EXTRACT_LE_64BITS(p) \ 214 ((uint64_t)(((uint64_t)(*((const uint8_t *)(p) + 7)) << 56) | \ 215 ((uint64_t)(*((const uint8_t *)(p) + 6)) << 48) | \ 216 ((uint64_t)(*((const uint8_t *)(p) + 5)) << 40) | \ 217 ((uint64_t)(*((const uint8_t *)(p) + 4)) << 32) | \ 218 ((uint64_t)(*((const uint8_t *)(p) + 3)) << 24) | \ 219 ((uint64_t)(*((const uint8_t *)(p) + 2)) << 16) | \ 220 ((uint64_t)(*((const uint8_t *)(p) + 1)) << 8) | \ 221 ((uint64_t)(*((const uint8_t *)(p) + 0)) << 0))) 222