1 /******************************************************************************* 2 * Copyright (c) 2013, Intel Corporation 3 * 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions are 8 * met: 9 * 10 * * Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 13 * * Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the 16 * distribution. 17 * 18 * * Neither the name of the Intel Corporation nor the names of its 19 * contributors may be used to endorse or promote products derived from 20 * this software without specific prior written permission. 21 * 22 * 23 * THIS SOFTWARE IS PROVIDED BY INTEL CORPORATION ""AS IS"" AND ANY 24 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL CORPORATION OR 27 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 28 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 29 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 30 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 34 ******************************************************************************** 35 * 36 * Intel SHA Extensions optimized implementation of a SHA-1 update function 37 * 38 * The function takes a pointer to the current hash values, a pointer to the 39 * input data, and a number of 64 byte blocks to process. Once all blocks have 40 * been processed, the digest pointer is updated with the resulting hash value. 41 * The function only processes complete blocks, there is no functionality to 42 * store partial blocks. All message padding and hash value initialization must 43 * be done outside the update function. 44 * 45 * The indented lines in the loop are instructions related to rounds processing. 46 * The non-indented lines are instructions related to the message schedule. 47 * 48 * Author: Sean Gulley <sean.m.gulley@intel.com> 49 * Date: July 2013 50 * 51 ******************************************************************************** 52 * 53 * Example complier command line: 54 * icc intel_sha_extensions_sha1_intrinsic.c 55 * gcc -msha -msse4 intel_sha_extensions_sha1_intrinsic.c 56 * 57 *******************************************************************************/ 58 #include <sys/cdefs.h> 59 __FBSDID("$FreeBSD$"); 60 61 #include <sys/types.h> 62 #include <immintrin.h> 63 64 #include <crypto/aesni/sha_sse.h> 65 66 void intel_sha1_step(uint32_t *digest, const char *data, uint32_t num_blks) { 67 __m128i abcd, e0, e1; 68 __m128i abcd_save, e_save; 69 __m128i msg0, msg1, msg2, msg3; 70 __m128i shuf_mask, e_mask; 71 72 #if 0 73 e_mask = _mm_set_epi64x(0xFFFFFFFF00000000ull, 0x0000000000000000ull); 74 #else 75 (void)e_mask; 76 e0 = _mm_set_epi64x(0, 0); 77 #endif 78 shuf_mask = _mm_set_epi64x(0x0001020304050607ull, 0x08090a0b0c0d0e0full); 79 80 // Load initial hash values 81 abcd = _mm_loadu_si128((__m128i*) digest); 82 e0 = _mm_insert_epi32(e0, *(digest+4), 3); 83 abcd = _mm_shuffle_epi32(abcd, 0x1B); 84 #if 0 85 e0 = _mm_and_si128(e0, e_mask); 86 #endif 87 88 while (num_blks > 0) { 89 // Save hash values for addition after rounds 90 abcd_save = abcd; 91 e_save = e0; 92 93 // Rounds 0-3 94 msg0 = _mm_loadu_si128((const __m128i*) data); 95 msg0 = _mm_shuffle_epi8(msg0, shuf_mask); 96 e0 = _mm_add_epi32(e0, msg0); 97 e1 = abcd; 98 abcd = _mm_sha1rnds4_epu32(abcd, e0, 0); 99 100 // Rounds 4-7 101 msg1 = _mm_loadu_si128((const __m128i*) (data+16)); 102 msg1 = _mm_shuffle_epi8(msg1, shuf_mask); 103 e1 = _mm_sha1nexte_epu32(e1, msg1); 104 e0 = abcd; 105 abcd = _mm_sha1rnds4_epu32(abcd, e1, 0); 106 msg0 = _mm_sha1msg1_epu32(msg0, msg1); 107 108 // Rounds 8-11 109 msg2 = _mm_loadu_si128((const __m128i*) (data+32)); 110 msg2 = _mm_shuffle_epi8(msg2, shuf_mask); 111 e0 = _mm_sha1nexte_epu32(e0, msg2); 112 e1 = abcd; 113 abcd = _mm_sha1rnds4_epu32(abcd, e0, 0); 114 msg1 = _mm_sha1msg1_epu32(msg1, msg2); 115 msg0 = _mm_xor_si128(msg0, msg2); 116 117 // Rounds 12-15 118 msg3 = _mm_loadu_si128((const __m128i*) (data+48)); 119 msg3 = _mm_shuffle_epi8(msg3, shuf_mask); 120 e1 = _mm_sha1nexte_epu32(e1, msg3); 121 e0 = abcd; 122 msg0 = _mm_sha1msg2_epu32(msg0, msg3); 123 abcd = _mm_sha1rnds4_epu32(abcd, e1, 0); 124 msg2 = _mm_sha1msg1_epu32(msg2, msg3); 125 msg1 = _mm_xor_si128(msg1, msg3); 126 127 // Rounds 16-19 128 e0 = _mm_sha1nexte_epu32(e0, msg0); 129 e1 = abcd; 130 msg1 = _mm_sha1msg2_epu32(msg1, msg0); 131 abcd = _mm_sha1rnds4_epu32(abcd, e0, 0); 132 msg3 = _mm_sha1msg1_epu32(msg3, msg0); 133 msg2 = _mm_xor_si128(msg2, msg0); 134 135 // Rounds 20-23 136 e1 = _mm_sha1nexte_epu32(e1, msg1); 137 e0 = abcd; 138 msg2 = _mm_sha1msg2_epu32(msg2, msg1); 139 abcd = _mm_sha1rnds4_epu32(abcd, e1, 1); 140 msg0 = _mm_sha1msg1_epu32(msg0, msg1); 141 msg3 = _mm_xor_si128(msg3, msg1); 142 143 // Rounds 24-27 144 e0 = _mm_sha1nexte_epu32(e0, msg2); 145 e1 = abcd; 146 msg3 = _mm_sha1msg2_epu32(msg3, msg2); 147 abcd = _mm_sha1rnds4_epu32(abcd, e0, 1); 148 msg1 = _mm_sha1msg1_epu32(msg1, msg2); 149 msg0 = _mm_xor_si128(msg0, msg2); 150 151 // Rounds 28-31 152 e1 = _mm_sha1nexte_epu32(e1, msg3); 153 e0 = abcd; 154 msg0 = _mm_sha1msg2_epu32(msg0, msg3); 155 abcd = _mm_sha1rnds4_epu32(abcd, e1, 1); 156 msg2 = _mm_sha1msg1_epu32(msg2, msg3); 157 msg1 = _mm_xor_si128(msg1, msg3); 158 159 // Rounds 32-35 160 e0 = _mm_sha1nexte_epu32(e0, msg0); 161 e1 = abcd; 162 msg1 = _mm_sha1msg2_epu32(msg1, msg0); 163 abcd = _mm_sha1rnds4_epu32(abcd, e0, 1); 164 msg3 = _mm_sha1msg1_epu32(msg3, msg0); 165 msg2 = _mm_xor_si128(msg2, msg0); 166 167 // Rounds 36-39 168 e1 = _mm_sha1nexte_epu32(e1, msg1); 169 e0 = abcd; 170 msg2 = _mm_sha1msg2_epu32(msg2, msg1); 171 abcd = _mm_sha1rnds4_epu32(abcd, e1, 1); 172 msg0 = _mm_sha1msg1_epu32(msg0, msg1); 173 msg3 = _mm_xor_si128(msg3, msg1); 174 175 // Rounds 40-43 176 e0 = _mm_sha1nexte_epu32(e0, msg2); 177 e1 = abcd; 178 msg3 = _mm_sha1msg2_epu32(msg3, msg2); 179 abcd = _mm_sha1rnds4_epu32(abcd, e0, 2); 180 msg1 = _mm_sha1msg1_epu32(msg1, msg2); 181 msg0 = _mm_xor_si128(msg0, msg2); 182 183 // Rounds 44-47 184 e1 = _mm_sha1nexte_epu32(e1, msg3); 185 e0 = abcd; 186 msg0 = _mm_sha1msg2_epu32(msg0, msg3); 187 abcd = _mm_sha1rnds4_epu32(abcd, e1, 2); 188 msg2 = _mm_sha1msg1_epu32(msg2, msg3); 189 msg1 = _mm_xor_si128(msg1, msg3); 190 191 // Rounds 48-51 192 e0 = _mm_sha1nexte_epu32(e0, msg0); 193 e1 = abcd; 194 msg1 = _mm_sha1msg2_epu32(msg1, msg0); 195 abcd = _mm_sha1rnds4_epu32(abcd, e0, 2); 196 msg3 = _mm_sha1msg1_epu32(msg3, msg0); 197 msg2 = _mm_xor_si128(msg2, msg0); 198 199 // Rounds 52-55 200 e1 = _mm_sha1nexte_epu32(e1, msg1); 201 e0 = abcd; 202 msg2 = _mm_sha1msg2_epu32(msg2, msg1); 203 abcd = _mm_sha1rnds4_epu32(abcd, e1, 2); 204 msg0 = _mm_sha1msg1_epu32(msg0, msg1); 205 msg3 = _mm_xor_si128(msg3, msg1); 206 207 // Rounds 56-59 208 e0 = _mm_sha1nexte_epu32(e0, msg2); 209 e1 = abcd; 210 msg3 = _mm_sha1msg2_epu32(msg3, msg2); 211 abcd = _mm_sha1rnds4_epu32(abcd, e0, 2); 212 msg1 = _mm_sha1msg1_epu32(msg1, msg2); 213 msg0 = _mm_xor_si128(msg0, msg2); 214 215 // Rounds 60-63 216 e1 = _mm_sha1nexte_epu32(e1, msg3); 217 e0 = abcd; 218 msg0 = _mm_sha1msg2_epu32(msg0, msg3); 219 abcd = _mm_sha1rnds4_epu32(abcd, e1, 3); 220 msg2 = _mm_sha1msg1_epu32(msg2, msg3); 221 msg1 = _mm_xor_si128(msg1, msg3); 222 223 // Rounds 64-67 224 e0 = _mm_sha1nexte_epu32(e0, msg0); 225 e1 = abcd; 226 msg1 = _mm_sha1msg2_epu32(msg1, msg0); 227 abcd = _mm_sha1rnds4_epu32(abcd, e0, 3); 228 msg3 = _mm_sha1msg1_epu32(msg3, msg0); 229 msg2 = _mm_xor_si128(msg2, msg0); 230 231 // Rounds 68-71 232 e1 = _mm_sha1nexte_epu32(e1, msg1); 233 e0 = abcd; 234 msg2 = _mm_sha1msg2_epu32(msg2, msg1); 235 abcd = _mm_sha1rnds4_epu32(abcd, e1, 3); 236 msg3 = _mm_xor_si128(msg3, msg1); 237 238 // Rounds 72-75 239 e0 = _mm_sha1nexte_epu32(e0, msg2); 240 e1 = abcd; 241 msg3 = _mm_sha1msg2_epu32(msg3, msg2); 242 abcd = _mm_sha1rnds4_epu32(abcd, e0, 3); 243 244 // Rounds 76-79 245 e1 = _mm_sha1nexte_epu32(e1, msg3); 246 e0 = abcd; 247 abcd = _mm_sha1rnds4_epu32(abcd, e1, 3); 248 249 // Add current hash values with previously saved 250 e0 = _mm_sha1nexte_epu32(e0, e_save); 251 abcd = _mm_add_epi32(abcd, abcd_save); 252 253 data += 64; 254 num_blks--; 255 } 256 257 abcd = _mm_shuffle_epi32(abcd, 0x1B); 258 _mm_store_si128((__m128i*) digest, abcd); 259 *(digest+4) = _mm_extract_epi32(e0, 3); 260 } 261 262