1/* 2 * Argon2 source code package 3 * 4 * Written by Daniel Dinu and Dmitry Khovratovich, 2015 5 * 6 * This work is licensed under a Creative Commons CC0 1.0 License/Waiver. 7 * 8 * You should have received a copy of the CC0 Public Domain Dedication along 9 * with 10 * this software. If not, see 11 * <http://creativecommons.org/publicdomain/zero/1.0/>. 12 */ 13 14#include <stdint.h> 15#include <stdlib.h> 16#include <string.h> 17 18#include "argon2-core.h" 19#include "argon2.h" 20#include "blamka-round-ref.h" 21#include "private/common.h" 22 23static void 24fill_block(const block *prev_block, const block *ref_block, block *next_block) 25{ 26 block blockR, block_tmp; 27 unsigned i; 28 29 copy_block(&blockR, ref_block); 30 xor_block(&blockR, prev_block); 31 copy_block(&block_tmp, &blockR); 32 /* Now blockR = ref_block + prev_block and bloc_tmp = ref_block + prev_block 33 Apply Blake2 on columns of 64-bit words: (0,1,...,15), then 34 (16,17,..31)... finally (112,113,...127) */ 35 for (i = 0; i < 8; ++i) { 36 BLAKE2_ROUND_NOMSG( 37 blockR.v[16 * i], blockR.v[16 * i + 1], blockR.v[16 * i + 2], 38 blockR.v[16 * i + 3], blockR.v[16 * i + 4], blockR.v[16 * i + 5], 39 blockR.v[16 * i + 6], blockR.v[16 * i + 7], blockR.v[16 * i + 8], 40 blockR.v[16 * i + 9], blockR.v[16 * i + 10], blockR.v[16 * i + 11], 41 blockR.v[16 * i + 12], blockR.v[16 * i + 13], blockR.v[16 * i + 14], 42 blockR.v[16 * i + 15]); 43 } 44 45 /* Apply Blake2 on rows of 64-bit words: (0,1,16,17,...112,113), then 46 (2,3,18,19,...,114,115).. finally (14,15,30,31,...,126,127) */ 47 for (i = 0; i < 8; i++) { 48 BLAKE2_ROUND_NOMSG( 49 blockR.v[2 * i], blockR.v[2 * i + 1], blockR.v[2 * i + 16], 50 blockR.v[2 * i + 17], blockR.v[2 * i + 32], blockR.v[2 * i + 33], 51 blockR.v[2 * i + 48], blockR.v[2 * i + 49], blockR.v[2 * i + 64], 52 blockR.v[2 * i + 65], blockR.v[2 * i + 80], blockR.v[2 * i + 81], 53 blockR.v[2 * i + 96], blockR.v[2 * i + 97], blockR.v[2 * i + 112], 54 blockR.v[2 * i + 113]); 55 } 56 57 copy_block(next_block, &block_tmp); 58 xor_block(next_block, &blockR); 59} 60 61static void 62fill_block_with_xor(const block *prev_block, const block *ref_block, 63 block *next_block) 64{ 65 block blockR, block_tmp; 66 unsigned i; 67 68 copy_block(&blockR, ref_block); 69 xor_block(&blockR, prev_block); 70 copy_block(&block_tmp, &blockR); 71 xor_block(&block_tmp, 72 next_block); /* Saving the next block contents for XOR over */ 73 /* Now blockR = ref_block + prev_block and bloc_tmp = ref_block + prev_block 74 * + next_block */ 75 /* Apply Blake2 on columns of 64-bit words: (0,1,...,15) , then 76 (16,17,..31)... finally (112,113,...127) */ 77 for (i = 0; i < 8; ++i) { 78 BLAKE2_ROUND_NOMSG( 79 blockR.v[16 * i], blockR.v[16 * i + 1], blockR.v[16 * i + 2], 80 blockR.v[16 * i + 3], blockR.v[16 * i + 4], blockR.v[16 * i + 5], 81 blockR.v[16 * i + 6], blockR.v[16 * i + 7], blockR.v[16 * i + 8], 82 blockR.v[16 * i + 9], blockR.v[16 * i + 10], blockR.v[16 * i + 11], 83 blockR.v[16 * i + 12], blockR.v[16 * i + 13], blockR.v[16 * i + 14], 84 blockR.v[16 * i + 15]); 85 } 86 87 /* Apply Blake2 on rows of 64-bit words: (0,1,16,17,...112,113), then 88 (2,3,18,19,...,114,115).. finally (14,15,30,31,...,126,127) */ 89 for (i = 0; i < 8; i++) { 90 BLAKE2_ROUND_NOMSG( 91 blockR.v[2 * i], blockR.v[2 * i + 1], blockR.v[2 * i + 16], 92 blockR.v[2 * i + 17], blockR.v[2 * i + 32], blockR.v[2 * i + 33], 93 blockR.v[2 * i + 48], blockR.v[2 * i + 49], blockR.v[2 * i + 64], 94 blockR.v[2 * i + 65], blockR.v[2 * i + 80], blockR.v[2 * i + 81], 95 blockR.v[2 * i + 96], blockR.v[2 * i + 97], blockR.v[2 * i + 112], 96 blockR.v[2 * i + 113]); 97 } 98 99 copy_block(next_block, &block_tmp); 100 xor_block(next_block, &blockR); 101} 102 103/* 104 * Generate pseudo-random values to reference blocks in the segment and puts 105 * them into the array 106 * @param instance Pointer to the current instance 107 * @param position Pointer to the current position 108 * @param pseudo_rands Pointer to the array of 64-bit values 109 * @pre pseudo_rands must point to @a instance->segment_length allocated values 110 */ 111static void 112generate_addresses(const argon2_instance_t *instance, 113 const argon2_position_t *position, uint64_t *pseudo_rands) 114{ 115 block zero_block, input_block, address_block, tmp_block; 116 uint32_t i; 117 118 init_block_value(&zero_block, 0); 119 init_block_value(&input_block, 0); 120 121 if (instance != NULL && position != NULL) { 122 input_block.v[0] = position->pass; 123 input_block.v[1] = position->lane; 124 input_block.v[2] = position->slice; 125 input_block.v[3] = instance->memory_blocks; 126 input_block.v[4] = instance->passes; 127 input_block.v[5] = instance->type; 128 129 for (i = 0; i < instance->segment_length; ++i) { 130 if (i % ARGON2_ADDRESSES_IN_BLOCK == 0) { 131 input_block.v[6]++; 132 init_block_value(&tmp_block, 0); 133 init_block_value(&address_block, 0); 134 fill_block_with_xor(&zero_block, &input_block, &tmp_block); 135 fill_block_with_xor(&zero_block, &tmp_block, &address_block); 136 } 137 138 pseudo_rands[i] = address_block.v[i % ARGON2_ADDRESSES_IN_BLOCK]; 139 } 140 } 141} 142 143void 144fill_segment_ref(const argon2_instance_t *instance, argon2_position_t position) 145{ 146 block *ref_block = NULL, *curr_block = NULL; 147 /* Pseudo-random values that determine the reference block position */ 148 uint64_t *pseudo_rands = NULL; 149 uint64_t pseudo_rand, ref_index, ref_lane; 150 uint32_t prev_offset, curr_offset; 151 uint32_t starting_index; 152 uint32_t i; 153 int data_independent_addressing = 1; 154 155 if (instance == NULL) { 156 return; 157 } 158 159 if (instance->type == Argon2_id && 160 (position.pass != 0 || position.slice >= ARGON2_SYNC_POINTS / 2)) { 161 data_independent_addressing = 0; 162 } 163 164 pseudo_rands = instance->pseudo_rands; 165 166 if (data_independent_addressing) { 167 generate_addresses(instance, &position, pseudo_rands); 168 } 169 170 starting_index = 0; 171 172 if ((0 == position.pass) && (0 == position.slice)) { 173 starting_index = 2; /* we have already generated the first two blocks */ 174 } 175 176 /* Offset of the current block */ 177 curr_offset = position.lane * instance->lane_length + 178 position.slice * instance->segment_length + starting_index; 179 180 if (0 == curr_offset % instance->lane_length) { 181 /* Last block in this lane */ 182 prev_offset = curr_offset + instance->lane_length - 1; 183 } else { 184 /* Previous block */ 185 prev_offset = curr_offset - 1; 186 } 187 188 for (i = starting_index; i < instance->segment_length; 189 ++i, ++curr_offset, ++prev_offset) { 190 /*1.1 Rotating prev_offset if needed */ 191 if (curr_offset % instance->lane_length == 1) { 192 prev_offset = curr_offset - 1; 193 } 194 195 /* 1.2 Computing the index of the reference block */ 196 /* 1.2.1 Taking pseudo-random value from the previous block */ 197 if (data_independent_addressing) { 198#pragma warning(push) 199#pragma warning(disable : 6385) 200 pseudo_rand = pseudo_rands[i]; 201#pragma warning(pop) 202 } else { 203 pseudo_rand = instance->region->memory[prev_offset].v[0]; 204 } 205 206 /* 1.2.2 Computing the lane of the reference block */ 207 ref_lane = ((pseudo_rand >> 32)) % instance->lanes; 208 209 if ((position.pass == 0) && (position.slice == 0)) { 210 /* Can not reference other lanes yet */ 211 ref_lane = position.lane; 212 } 213 214 /* 1.2.3 Computing the number of possible reference block within the 215 * lane. 216 */ 217 position.index = i; 218 ref_index = index_alpha(instance, &position, pseudo_rand & 0xFFFFFFFF, 219 ref_lane == position.lane); 220 221 /* 2 Creating a new block */ 222 ref_block = instance->region->memory + 223 instance->lane_length * ref_lane + ref_index; 224 curr_block = instance->region->memory + curr_offset; 225 if (position.pass != 0) { 226 fill_block_with_xor(instance->region->memory + prev_offset, 227 ref_block, curr_block); 228 } else { 229 fill_block(instance->region->memory + prev_offset, ref_block, 230 curr_block); 231 } 232 } 233} 234