| 1 |
/* |
| 2 |
* This is an OpenSSL-compatible implementation of the RSA Data Security, Inc. |
| 3 |
* MD4 Message-Digest Algorithm (RFC 1320). |
| 4 |
* |
| 5 |
* Homepage: |
| 6 |
* http://openwall.info/wiki/people/solar/software/public-domain-source-code/md4 |
| 7 |
* |
| 8 |
* Author: |
| 9 |
* Alexander Peslyak, better known as Solar Designer <solar at openwall.com> |
| 10 |
* |
| 11 |
* This software was written by Alexander Peslyak in 2001. No copyright is |
| 12 |
* claimed, and the software is hereby placed in the public domain. |
| 13 |
* In case this attempt to disclaim copyright and place the software in the |
| 14 |
* public domain is deemed null and void, then the software is |
| 15 |
* Copyright (c) 2001 Alexander Peslyak and it is hereby released to the |
| 16 |
* general public under the following terms: |
| 17 |
* |
| 18 |
* Redistribution and use in source and binary forms, with or without |
| 19 |
* modification, are permitted. |
| 20 |
* |
| 21 |
* There's ABSOLUTELY NO WARRANTY, express or implied. |
| 22 |
* |
| 23 |
* (This is a heavily cut-down "BSD license".) |
| 24 |
* |
| 25 |
* This differs from Colin Plumb's older public domain implementation in that |
| 26 |
* no exactly 32-bit integer data type is required (any 32-bit or wider |
| 27 |
* unsigned integer data type will do), there's no compile-time endianness |
| 28 |
* configuration, and the function prototypes match OpenSSL's. No code from |
| 29 |
* Colin Plumb's implementation has been reused; this comment merely compares |
| 30 |
* the properties of the two independent implementations. |
| 31 |
* |
| 32 |
* The primary goals of this implementation are portability and ease of use. |
| 33 |
* It is meant to be fast, but not as fast as possible. Some known |
| 34 |
* optimizations are not included to reduce source code size and avoid |
| 35 |
* compile-time configuration. |
| 36 |
*/ |
| 37 |
|
| 38 |
#include "crypt-port.h" |
| 39 |
|
| 40 |
#if INCLUDE_nt |
| 41 |
|
| 42 |
#include "alg-md4.h" |
| 43 |
|
| 44 |
/* |
| 45 |
* The basic MD4 functions. |
| 46 |
* |
| 47 |
* F and G are optimized compared to their RFC 1320 definitions, with the |
| 48 |
* optimization for F borrowed from Colin Plumb's MD5 implementation. |
| 49 |
*/ |
| 50 |
#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z)))) |
| 51 |
#define G(x, y, z) (((x) & ((y) | (z))) | ((y) & (z))) |
| 52 |
#define H(x, y, z) ((x) ^ (y) ^ (z)) |
| 53 |
|
| 54 |
/* |
| 55 |
* The MD4 transformation for all three rounds. |
| 56 |
*/ |
| 57 |
#define STEP(f, a, b, c, d, x, s) \ |
| 58 |
(a) += f((b), (c), (d)) + (x); \ |
| 59 |
(a) = (((a) << (s)) | (((a) & 0xffffffff) >> (32 - (s)))); |
| 60 |
|
| 61 |
/* |
| 62 |
* SET reads 4 input bytes in little-endian byte order and stores them in a |
| 63 |
* properly aligned word in host byte order. |
| 64 |
* |
| 65 |
* The check for little-endian architectures that tolerate unaligned memory |
| 66 |
* accesses is just an optimization. Nothing will break if it fails to detect |
| 67 |
* a suitable architecture. |
| 68 |
* |
| 69 |
* Unfortunately, this optimization may be a C strict aliasing rules violation |
| 70 |
* if the caller's data buffer has effective type that cannot be aliased by |
| 71 |
* MD4_u32plus. In practice, this problem may occur if these MD4 routines are |
| 72 |
* inlined into a calling function, or with future and dangerously advanced |
| 73 |
* link-time optimizations. For the time being, keeping these MD4 routines in |
| 74 |
* their own translation unit avoids the problem. |
| 75 |
*/ |
| 76 |
#if 0 /* defined(__i386__) || defined(__x86_64__) || defined(__vax__) */ |
| 77 |
#define SET(n) \ |
| 78 |
(*(const MD4_u32plus *)&ptr[(n) * 4]) |
| 79 |
#define GET(n) \ |
| 80 |
SET(n) |
| 81 |
#else |
| 82 |
#define SET(n) \ |
| 83 |
(ctx->block[(n)] = \ |
| 84 |
(MD4_u32plus)ptr[(n) * 4] | \ |
| 85 |
((MD4_u32plus)ptr[(n) * 4 + 1] << 8) | \ |
| 86 |
((MD4_u32plus)ptr[(n) * 4 + 2] << 16) | \ |
| 87 |
((MD4_u32plus)ptr[(n) * 4 + 3] << 24)) |
| 88 |
#define GET(n) \ |
| 89 |
(ctx->block[(n)]) |
| 90 |
#endif |
| 91 |
|
| 92 |
/* |
| 93 |
* This processes one or more 64-byte data blocks, but does NOT update the bit |
| 94 |
* counters. There are no alignment requirements. |
| 95 |
*/ |
| 96 |
static const void *body(MD4_CTX *ctx, const void *data, unsigned long size) |
| 97 |
{ |
| 98 |
const unsigned char *ptr; |
| 99 |
MD4_u32plus a, b, c, d; |
| 100 |
MD4_u32plus saved_a, saved_b, saved_c, saved_d; |
| 101 |
const MD4_u32plus ac1 = 0x5a827999, ac2 = 0x6ed9eba1; |
| 102 |
|
| 103 |
ptr = (const unsigned char *)data; |
| 104 |
|
| 105 |
a = ctx->a; |
| 106 |
b = ctx->b; |
| 107 |
c = ctx->c; |
| 108 |
d = ctx->d; |
| 109 |
|
| 110 |
do { |
| 111 |
saved_a = a; |
| 112 |
saved_b = b; |
| 113 |
saved_c = c; |
| 114 |
saved_d = d; |
| 115 |
|
| 116 |
/* Round 1 */ |
| 117 |
STEP(F, a, b, c, d, SET(0), 3) |
| 118 |
STEP(F, d, a, b, c, SET(1), 7) |
| 119 |
STEP(F, c, d, a, b, SET(2), 11) |
| 120 |
STEP(F, b, c, d, a, SET(3), 19) |
| 121 |
STEP(F, a, b, c, d, SET(4), 3) |
| 122 |
STEP(F, d, a, b, c, SET(5), 7) |
| 123 |
STEP(F, c, d, a, b, SET(6), 11) |
| 124 |
STEP(F, b, c, d, a, SET(7), 19) |
| 125 |
STEP(F, a, b, c, d, SET(8), 3) |
| 126 |
STEP(F, d, a, b, c, SET(9), 7) |
| 127 |
STEP(F, c, d, a, b, SET(10), 11) |
| 128 |
STEP(F, b, c, d, a, SET(11), 19) |
| 129 |
STEP(F, a, b, c, d, SET(12), 3) |
| 130 |
STEP(F, d, a, b, c, SET(13), 7) |
| 131 |
STEP(F, c, d, a, b, SET(14), 11) |
| 132 |
STEP(F, b, c, d, a, SET(15), 19) |
| 133 |
|
| 134 |
/* Round 2 */ |
| 135 |
STEP(G, a, b, c, d, GET(0) + ac1, 3) |
| 136 |
STEP(G, d, a, b, c, GET(4) + ac1, 5) |
| 137 |
STEP(G, c, d, a, b, GET(8) + ac1, 9) |
| 138 |
STEP(G, b, c, d, a, GET(12) + ac1, 13) |
| 139 |
STEP(G, a, b, c, d, GET(1) + ac1, 3) |
| 140 |
STEP(G, d, a, b, c, GET(5) + ac1, 5) |
| 141 |
STEP(G, c, d, a, b, GET(9) + ac1, 9) |
| 142 |
STEP(G, b, c, d, a, GET(13) + ac1, 13) |
| 143 |
STEP(G, a, b, c, d, GET(2) + ac1, 3) |
| 144 |
STEP(G, d, a, b, c, GET(6) + ac1, 5) |
| 145 |
STEP(G, c, d, a, b, GET(10) + ac1, 9) |
| 146 |
STEP(G, b, c, d, a, GET(14) + ac1, 13) |
| 147 |
STEP(G, a, b, c, d, GET(3) + ac1, 3) |
| 148 |
STEP(G, d, a, b, c, GET(7) + ac1, 5) |
| 149 |
STEP(G, c, d, a, b, GET(11) + ac1, 9) |
| 150 |
STEP(G, b, c, d, a, GET(15) + ac1, 13) |
| 151 |
|
| 152 |
/* Round 3 */ |
| 153 |
STEP(H, a, b, c, d, GET(0) + ac2, 3) |
| 154 |
STEP(H, d, a, b, c, GET(8) + ac2, 9) |
| 155 |
STEP(H, c, d, a, b, GET(4) + ac2, 11) |
| 156 |
STEP(H, b, c, d, a, GET(12) + ac2, 15) |
| 157 |
STEP(H, a, b, c, d, GET(2) + ac2, 3) |
| 158 |
STEP(H, d, a, b, c, GET(10) + ac2, 9) |
| 159 |
STEP(H, c, d, a, b, GET(6) + ac2, 11) |
| 160 |
STEP(H, b, c, d, a, GET(14) + ac2, 15) |
| 161 |
STEP(H, a, b, c, d, GET(1) + ac2, 3) |
| 162 |
STEP(H, d, a, b, c, GET(9) + ac2, 9) |
| 163 |
STEP(H, c, d, a, b, GET(5) + ac2, 11) |
| 164 |
STEP(H, b, c, d, a, GET(13) + ac2, 15) |
| 165 |
STEP(H, a, b, c, d, GET(3) + ac2, 3) |
| 166 |
STEP(H, d, a, b, c, GET(11) + ac2, 9) |
| 167 |
STEP(H, c, d, a, b, GET(7) + ac2, 11) |
| 168 |
STEP(H, b, c, d, a, GET(15) + ac2, 15) |
| 169 |
|
| 170 |
a += saved_a; |
| 171 |
b += saved_b; |
| 172 |
c += saved_c; |
| 173 |
d += saved_d; |
| 174 |
|
| 175 |
ptr += 64; |
| 176 |
} while (size -= 64); |
| 177 |
|
| 178 |
ctx->a = a; |
| 179 |
ctx->b = b; |
| 180 |
ctx->c = c; |
| 181 |
ctx->d = d; |
| 182 |
|
| 183 |
return ptr; |
| 184 |
} |
| 185 |
|
| 186 |
void MD4_Init(MD4_CTX *ctx) |
| 187 |
{ |
| 188 |
ctx->a = 0x67452301; |
| 189 |
ctx->b = 0xefcdab89; |
| 190 |
ctx->c = 0x98badcfe; |
| 191 |
ctx->d = 0x10325476; |
| 192 |
|
| 193 |
ctx->lo = 0; |
| 194 |
ctx->hi = 0; |
| 195 |
} |
| 196 |
|
| 197 |
void MD4_Update(MD4_CTX *ctx, const void *data, size_t size) |
| 198 |
{ |
| 199 |
MD4_u32plus saved_lo; |
| 200 |
unsigned long used, available; |
| 201 |
|
| 202 |
saved_lo = ctx->lo; |
| 203 |
if ((ctx->lo = (saved_lo + size) & 0x1fffffff) < saved_lo) |
| 204 |
ctx->hi++; |
| 205 |
ctx->hi += (MD4_u32plus) size >> 29; |
| 206 |
|
| 207 |
used = saved_lo & 0x3f; |
| 208 |
|
| 209 |
if (used) { |
| 210 |
available = 64 - used; |
| 211 |
|
| 212 |
if (size < available) { |
| 213 |
memcpy(&ctx->buffer[used], data, size); |
| 214 |
return; |
| 215 |
} |
| 216 |
|
| 217 |
memcpy(&ctx->buffer[used], data, available); |
| 218 |
data = (const unsigned char *)data + available; |
| 219 |
size -= available; |
| 220 |
body(ctx, ctx->buffer, 64); |
| 221 |
} |
| 222 |
|
| 223 |
if (size >= 64) { |
| 224 |
data = body(ctx, data, size & ~(unsigned long)0x3f); |
| 225 |
size &= 0x3f; |
| 226 |
} |
| 227 |
|
| 228 |
memcpy(ctx->buffer, data, size); |
| 229 |
} |
| 230 |
|
| 231 |
#define OUT(dst, src) \ |
| 232 |
(dst)[0] = (unsigned char)(src); \ |
| 233 |
(dst)[1] = (unsigned char)((src) >> 8); \ |
| 234 |
(dst)[2] = (unsigned char)((src) >> 16); \ |
| 235 |
(dst)[3] = (unsigned char)((src) >> 24); |
| 236 |
|
| 237 |
void MD4_Final(uint8_t result[16], MD4_CTX *ctx) |
| 238 |
{ |
| 239 |
unsigned long used, available; |
| 240 |
|
| 241 |
used = ctx->lo & 0x3f; |
| 242 |
|
| 243 |
ctx->buffer[used++] = 0x80; |
| 244 |
|
| 245 |
available = 64 - used; |
| 246 |
|
| 247 |
if (available < 8) { |
| 248 |
memset(&ctx->buffer[used], 0, available); |
| 249 |
body(ctx, ctx->buffer, 64); |
| 250 |
used = 0; |
| 251 |
available = 64; |
| 252 |
} |
| 253 |
|
| 254 |
memset(&ctx->buffer[used], 0, available - 8); |
| 255 |
|
| 256 |
ctx->lo <<= 3; |
| 257 |
OUT(&ctx->buffer[56], ctx->lo) |
| 258 |
OUT(&ctx->buffer[60], ctx->hi) |
| 259 |
|
| 260 |
body(ctx, ctx->buffer, 64); |
| 261 |
|
| 262 |
OUT(&result[0], ctx->a) |
| 263 |
OUT(&result[4], ctx->b) |
| 264 |
OUT(&result[8], ctx->c) |
| 265 |
OUT(&result[12], ctx->d) |
| 266 |
|
| 267 |
explicit_bzero(ctx, sizeof(*ctx)); |
| 268 |
} |
| 269 |
|
| 270 |
#endif |