1/* 2 * Based on the Mozilla SHA1 (see mozilla-sha1/sha1.c), 3 * optimized to do word accesses rather than byte accesses, 4 * and to avoid unnecessary copies into the context array. 5 */ 6 7#include <string.h> 8#include <arpa/inet.h> 9 10#include "sha1.h" 11 12#if defined(__i386__) || defined(__x86_64__) 13 14/* 15 * Force usage of rol or ror by selecting the one with the smaller constant. 16 * It _can_ generate slightly smaller code (a constant of 1 is special), but 17 * perhaps more importantly it's possibly faster on any uarch that does a 18 * rotate with a loop. 19 */ 20 21#define SHA_ASM(op, x, n) ({ unsigned int __res; __asm__(op " %1,%0":"=r" (__res):"i" (n), "0" (x)); __res; }) 22#define SHA_ROL(x,n) SHA_ASM("rol", x, n) 23#define SHA_ROR(x,n) SHA_ASM("ror", x, n) 24 25#else 26 27#define SHA_ROT(X,l,r) (((X) << (l)) | ((X) >> (r))) 28#define SHA_ROL(X,n) SHA_ROT(X,n,32-(n)) 29#define SHA_ROR(X,n) SHA_ROT(X,32-(n),n) 30 31#endif 32 33/* 34 * If you have 32 registers or more, the compiler can (and should) 35 * try to change the array[] accesses into registers. However, on 36 * machines with less than ~25 registers, that won't really work, 37 * and at least gcc will make an unholy mess of it. 38 * 39 * So to avoid that mess which just slows things down, we force 40 * the stores to memory to actually happen (we might be better off 41 * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as 42 * suggested by Artur Skawina - that will also make gcc unable to 43 * try to do the silly "optimize away loads" part because it won't 44 * see what the value will be). 45 * 46 * Ben Herrenschmidt reports that on PPC, the C version comes close 47 * to the optimized asm with this (ie on PPC you don't want that 48 * 'volatile', since there are lots of registers). 49 * 50 * On ARM we get the best code generation by forcing a full memory barrier 51 * between each SHA_ROUND, otherwise gcc happily get wild with spilling and 52 * the stack frame size simply explode and performance goes down the drain. 53 */ 54 55#if defined(__i386__) || defined(__x86_64__) 56 #define setW(x, val) (*(volatile unsigned int *)&W(x) = (val)) 57#elif defined(__arm__) 58 #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0) 59#else 60 #define setW(x, val) (W(x) = (val)) 61#endif 62 63/* 64 * Performance might be improved if the CPU architecture is OK with 65 * unaligned 32-bit loads and a fast ntohl() is available. 66 * Otherwise fall back to byte loads and shifts which is portable, 67 * and is faster on architectures with memory alignment issues. 68 */ 69 70#if defined(__i386__) || defined(__x86_64__) || \ 71 defined(__ppc__) || defined(__ppc64__) || \ 72 defined(__powerpc__) || defined(__powerpc64__) || \ 73 defined(__s390__) || defined(__s390x__) 74 75#define get_be32(p) ntohl(*(unsigned int *)(p)) 76#define put_be32(p, v) do { *(unsigned int *)(p) = htonl(v); } while (0) 77 78#else 79 80#define get_be32(p) ( \ 81 (*((unsigned char *)(p) + 0) << 24) | \ 82 (*((unsigned char *)(p) + 1) << 16) | \ 83 (*((unsigned char *)(p) + 2) << 8) | \ 84 (*((unsigned char *)(p) + 3) << 0) ) 85#define put_be32(p, v) do { \ 86 unsigned int __v = (v); \ 87 *((unsigned char *)(p) + 0) = __v >> 24; \ 88 *((unsigned char *)(p) + 1) = __v >> 16; \ 89 *((unsigned char *)(p) + 2) = __v >> 8; \ 90 *((unsigned char *)(p) + 3) = __v >> 0; } while (0) 91 92#endif 93 94/* This "rolls" over the 512-bit array */ 95#define W(x) (array[(x)&15]) 96 97/* 98 * Where do we get the source from? The first 16 iterations get it from 99 * the input data, the next mix it from the 512-bit array. 100 */ 101#define SHA_SRC(t) get_be32(data + t) 102#define SHA_MIX(t) SHA_ROL(W(t+13) ^ W(t+8) ^ W(t+2) ^ W(t), 1) 103 104#define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \ 105 unsigned int TEMP = input(t); setW(t, TEMP); \ 106 E += TEMP + SHA_ROL(A,5) + (fn) + (constant); \ 107 B = SHA_ROR(B, 2); } while (0) 108 109#define T_0_15(t, A, B, C, D, E) SHA_ROUND(t, SHA_SRC, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E ) 110#define T_16_19(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E ) 111#define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E ) 112#define T_40_59(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, ((B&C)+(D&(B^C))) , 0x8f1bbcdc, A, B, C, D, E ) 113#define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0xca62c1d6, A, B, C, D, E ) 114 115static void blk_SHA1_Block(blk_SHA_CTX *ctx, const unsigned int *data) 116{ 117 unsigned int A,B,C,D,E; 118 unsigned int array[16]; 119 120 A = ctx->H[0]; 121 B = ctx->H[1]; 122 C = ctx->H[2]; 123 D = ctx->H[3]; 124 E = ctx->H[4]; 125 126 /* Round 1 - iterations 0-16 take their input from 'data' */ 127 T_0_15( 0, A, B, C, D, E); 128 T_0_15( 1, E, A, B, C, D); 129 T_0_15( 2, D, E, A, B, C); 130 T_0_15( 3, C, D, E, A, B); 131 T_0_15( 4, B, C, D, E, A); 132 T_0_15( 5, A, B, C, D, E); 133 T_0_15( 6, E, A, B, C, D); 134 T_0_15( 7, D, E, A, B, C); 135 T_0_15( 8, C, D, E, A, B); 136 T_0_15( 9, B, C, D, E, A); 137 T_0_15(10, A, B, C, D, E); 138 T_0_15(11, E, A, B, C, D); 139 T_0_15(12, D, E, A, B, C); 140 T_0_15(13, C, D, E, A, B); 141 T_0_15(14, B, C, D, E, A); 142 T_0_15(15, A, B, C, D, E); 143 144 /* Round 1 - tail. Input from 512-bit mixing array */ 145 T_16_19(16, E, A, B, C, D); 146 T_16_19(17, D, E, A, B, C); 147 T_16_19(18, C, D, E, A, B); 148 T_16_19(19, B, C, D, E, A); 149 150 /* Round 2 */ 151 T_20_39(20, A, B, C, D, E); 152 T_20_39(21, E, A, B, C, D); 153 T_20_39(22, D, E, A, B, C); 154 T_20_39(23, C, D, E, A, B); 155 T_20_39(24, B, C, D, E, A); 156 T_20_39(25, A, B, C, D, E); 157 T_20_39(26, E, A, B, C, D); 158 T_20_39(27, D, E, A, B, C); 159 T_20_39(28, C, D, E, A, B); 160 T_20_39(29, B, C, D, E, A); 161 T_20_39(30, A, B, C, D, E); 162 T_20_39(31, E, A, B, C, D); 163 T_20_39(32, D, E, A, B, C); 164 T_20_39(33, C, D, E, A, B); 165 T_20_39(34, B, C, D, E, A); 166 T_20_39(35, A, B, C, D, E); 167 T_20_39(36, E, A, B, C, D); 168 T_20_39(37, D, E, A, B, C); 169 T_20_39(38, C, D, E, A, B); 170 T_20_39(39, B, C, D, E, A); 171 172 /* Round 3 */ 173 T_40_59(40, A, B, C, D, E); 174 T_40_59(41, E, A, B, C, D); 175 T_40_59(42, D, E, A, B, C); 176 T_40_59(43, C, D, E, A, B); 177 T_40_59(44, B, C, D, E, A); 178 T_40_59(45, A, B, C, D, E); 179 T_40_59(46, E, A, B, C, D); 180 T_40_59(47, D, E, A, B, C); 181 T_40_59(48, C, D, E, A, B); 182 T_40_59(49, B, C, D, E, A); 183 T_40_59(50, A, B, C, D, E); 184 T_40_59(51, E, A, B, C, D); 185 T_40_59(52, D, E, A, B, C); 186 T_40_59(53, C, D, E, A, B); 187 T_40_59(54, B, C, D, E, A); 188 T_40_59(55, A, B, C, D, E); 189 T_40_59(56, E, A, B, C, D); 190 T_40_59(57, D, E, A, B, C); 191 T_40_59(58, C, D, E, A, B); 192 T_40_59(59, B, C, D, E, A); 193 194 /* Round 4 */ 195 T_60_79(60, A, B, C, D, E); 196 T_60_79(61, E, A, B, C, D); 197 T_60_79(62, D, E, A, B, C); 198 T_60_79(63, C, D, E, A, B); 199 T_60_79(64, B, C, D, E, A); 200 T_60_79(65, A, B, C, D, E); 201 T_60_79(66, E, A, B, C, D); 202 T_60_79(67, D, E, A, B, C); 203 T_60_79(68, C, D, E, A, B); 204 T_60_79(69, B, C, D, E, A); 205 T_60_79(70, A, B, C, D, E); 206 T_60_79(71, E, A, B, C, D); 207 T_60_79(72, D, E, A, B, C); 208 T_60_79(73, C, D, E, A, B); 209 T_60_79(74, B, C, D, E, A); 210 T_60_79(75, A, B, C, D, E); 211 T_60_79(76, E, A, B, C, D); 212 T_60_79(77, D, E, A, B, C); 213 T_60_79(78, C, D, E, A, B); 214 T_60_79(79, B, C, D, E, A); 215 216 ctx->H[0] += A; 217 ctx->H[1] += B; 218 ctx->H[2] += C; 219 ctx->H[3] += D; 220 ctx->H[4] += E; 221} 222 223void blk_SHA1_Init(blk_SHA_CTX *ctx) 224{ 225 ctx->size = 0; 226 227 /* Initialize H with the magic constants (see FIPS180 for constants) */ 228 ctx->H[0] = 0x67452301; 229 ctx->H[1] = 0xefcdab89; 230 ctx->H[2] = 0x98badcfe; 231 ctx->H[3] = 0x10325476; 232 ctx->H[4] = 0xc3d2e1f0; 233} 234 235void blk_SHA1_Update(blk_SHA_CTX *ctx, const void *data, unsigned long len) 236{ 237 int lenW = ctx->size & 63; 238 239 ctx->size += len; 240 241 /* Read the data into W and process blocks as they get full */ 242 if (lenW) { 243 int left = 64 - lenW; 244 if (len < left) 245 left = len; 246 memcpy(lenW + (char *)ctx->W, data, left); 247 lenW = (lenW + left) & 63; 248 len -= left; 249 data += left; 250 if (lenW) 251 return; 252 blk_SHA1_Block(ctx, ctx->W); 253 } 254 while (len >= 64) { 255 blk_SHA1_Block(ctx, data); 256 data += 64; 257 len -= 64; 258 } 259 if (len) 260 memcpy(ctx->W, data, len); 261} 262 263void blk_SHA1_Final(unsigned char hashout[20], blk_SHA_CTX *ctx) 264{ 265 static const unsigned char pad[64] = { 0x80 }; 266 unsigned int padlen[2]; 267 int i; 268 269 /* Pad with a binary 1 (ie 0x80), then zeroes, then length */ 270 padlen[0] = htonl(ctx->size >> 29); 271 padlen[1] = htonl(ctx->size << 3); 272 273 i = ctx->size & 63; 274 blk_SHA1_Update(ctx, pad, 1+ (63 & (55 - i))); 275 blk_SHA1_Update(ctx, padlen, 8); 276 277 /* Output hash */ 278 for (i = 0; i < 5; i++) 279 put_be32(hashout + i*4, ctx->H[i]); 280}