| File: | home/bhubbard/working/src/ceph/src/erasure-code/jerasure/jerasure/src/jerasure.c |
| Warning: | line 1326, column 11 Value stored to 'optodo' is never read |
[?] Use j/k keys for keyboard navigation
| 1 | /* * |
| 2 | * Copyright (c) 2014, James S. Plank and Kevin Greenan |
| 3 | * All rights reserved. |
| 4 | * |
| 5 | * Jerasure - A C/C++ Library for a Variety of Reed-Solomon and RAID-6 Erasure |
| 6 | * Coding Techniques |
| 7 | * |
| 8 | * Revision 2.0: Galois Field backend now links to GF-Complete |
| 9 | * |
| 10 | * Redistribution and use in source and binary forms, with or without |
| 11 | * modification, are permitted provided that the following conditions |
| 12 | * are met: |
| 13 | * |
| 14 | * - Redistributions of source code must retain the above copyright |
| 15 | * notice, this list of conditions and the following disclaimer. |
| 16 | * |
| 17 | * - Redistributions in binary form must reproduce the above copyright |
| 18 | * notice, this list of conditions and the following disclaimer in |
| 19 | * the documentation and/or other materials provided with the |
| 20 | * distribution. |
| 21 | * |
| 22 | * - Neither the name of the University of Tennessee nor the names of its |
| 23 | * contributors may be used to endorse or promote products derived |
| 24 | * from this software without specific prior written permission. |
| 25 | * |
| 26 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 27 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 28 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 29 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 30 | * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 31 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, |
| 32 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS |
| 33 | * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED |
| 34 | * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 35 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY |
| 36 | * WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
| 37 | * POSSIBILITY OF SUCH DAMAGE. |
| 38 | */ |
| 39 | |
| 40 | /* Jerasure's authors: |
| 41 | |
| 42 | Revision 2.x - 2014: James S. Plank and Kevin M. Greenan |
| 43 | Revision 1.2 - 2008: James S. Plank, Scott Simmerman and Catherine D. Schuman. |
| 44 | Revision 1.0 - 2007: James S. Plank |
| 45 | */ |
| 46 | |
| 47 | #include <stdio.h> |
| 48 | #include <stdlib.h> |
| 49 | #include <string.h> |
| 50 | #include <assert.h> |
| 51 | |
| 52 | #include "galois.h" |
| 53 | #include "jerasure.h" |
| 54 | |
| 55 | #define talloc(type, num)(type *) malloc(sizeof(type)*(num)) (type *) malloc(sizeof(type)*(num)) |
| 56 | |
| 57 | static double jerasure_total_xor_bytes = 0; |
| 58 | static double jerasure_total_gf_bytes = 0; |
| 59 | static double jerasure_total_memcpy_bytes = 0; |
| 60 | |
| 61 | void jerasure_print_matrix(int *m, int rows, int cols, int w) |
| 62 | { |
| 63 | int i, j; |
| 64 | int fw; |
| 65 | char s[30]; |
| 66 | unsigned int w2; |
| 67 | |
| 68 | if (w == 32) { |
| 69 | fw = 10; |
| 70 | } else { |
| 71 | w2 = (1 << w); |
| 72 | sprintf(s, "%u", w2-1)__builtin___sprintf_chk (s, 2 - 1, __builtin_object_size (s, 2 > 1), "%u", w2-1); |
| 73 | fw = strlen(s); |
| 74 | } |
| 75 | |
| 76 | for (i = 0; i < rows; i++) { |
| 77 | for (j = 0; j < cols; j++) { |
| 78 | if (j != 0) printf(" ")__printf_chk (2 - 1, " "); |
| 79 | printf("%*u", fw, m[i*cols+j])__printf_chk (2 - 1, "%*u", fw, m[i*cols+j]); |
| 80 | } |
| 81 | printf("\n")__printf_chk (2 - 1, "\n"); |
| 82 | } |
| 83 | } |
| 84 | |
| 85 | void jerasure_print_bitmatrix(int *m, int rows, int cols, int w) |
| 86 | { |
| 87 | int i, j; |
| 88 | |
| 89 | for (i = 0; i < rows; i++) { |
| 90 | if (i != 0 && i%w == 0) printf("\n")__printf_chk (2 - 1, "\n"); |
| 91 | for (j = 0; j < cols; j++) { |
| 92 | if (j != 0 && j%w == 0) printf(" ")__printf_chk (2 - 1, " "); |
| 93 | printf("%d", m[i*cols+j])__printf_chk (2 - 1, "%d", m[i*cols+j]); |
| 94 | } |
| 95 | printf("\n")__printf_chk (2 - 1, "\n"); |
| 96 | } |
| 97 | } |
| 98 | |
| 99 | int jerasure_make_decoding_matrix(int k, int m, int w, int *matrix, int *erased, int *decoding_matrix, int *dm_ids) |
| 100 | { |
| 101 | int i, j, *tmpmat; |
| 102 | |
| 103 | j = 0; |
| 104 | for (i = 0; j < k; i++) { |
| 105 | if (erased[i] == 0) { |
| 106 | dm_ids[j] = i; |
| 107 | j++; |
| 108 | } |
| 109 | } |
| 110 | |
| 111 | tmpmat = talloc(int, k*k)(int *) malloc(sizeof(int)*(k*k)); |
| 112 | if (tmpmat == NULL((void*)0)) { return -1; } |
| 113 | for (i = 0; i < k; i++) { |
| 114 | if (dm_ids[i] < k) { |
| 115 | for (j = 0; j < k; j++) tmpmat[i*k+j] = 0; |
| 116 | tmpmat[i*k+dm_ids[i]] = 1; |
| 117 | } else { |
| 118 | for (j = 0; j < k; j++) { |
| 119 | tmpmat[i*k+j] = matrix[(dm_ids[i]-k)*k+j]; |
| 120 | } |
| 121 | } |
| 122 | } |
| 123 | |
| 124 | i = jerasure_invert_matrix(tmpmat, decoding_matrix, k, w); |
| 125 | free(tmpmat); |
| 126 | return i; |
| 127 | } |
| 128 | |
| 129 | /* Internal Routine */ |
| 130 | int jerasure_make_decoding_bitmatrix(int k, int m, int w, int *matrix, int *erased, int *decoding_matrix, int *dm_ids) |
| 131 | { |
| 132 | int i, j, *tmpmat; |
| 133 | int index, mindex; |
| 134 | |
| 135 | j = 0; |
| 136 | for (i = 0; j < k; i++) { |
| 137 | if (erased[i] == 0) { |
| 138 | dm_ids[j] = i; |
| 139 | j++; |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | tmpmat = talloc(int, k*k*w*w)(int *) malloc(sizeof(int)*(k*k*w*w)); |
| 144 | if (tmpmat == NULL((void*)0)) { return -1; } |
| 145 | for (i = 0; i < k; i++) { |
| 146 | if (dm_ids[i] < k) { |
| 147 | index = i*k*w*w; |
| 148 | for (j = 0; j < k*w*w; j++) tmpmat[index+j] = 0; |
| 149 | index = i*k*w*w+dm_ids[i]*w; |
| 150 | for (j = 0; j < w; j++) { |
| 151 | tmpmat[index] = 1; |
| 152 | index += (k*w+1); |
| 153 | } |
| 154 | } else { |
| 155 | index = i*k*w*w; |
| 156 | mindex = (dm_ids[i]-k)*k*w*w; |
| 157 | for (j = 0; j < k*w*w; j++) { |
| 158 | tmpmat[index+j] = matrix[mindex+j]; |
| 159 | } |
| 160 | } |
| 161 | } |
| 162 | |
| 163 | i = jerasure_invert_bitmatrix(tmpmat, decoding_matrix, k*w); |
| 164 | free(tmpmat); |
| 165 | return i; |
| 166 | } |
| 167 | |
| 168 | int jerasure_matrix_decode(int k, int m, int w, int *matrix, int row_k_ones, int *erasures, |
| 169 | char **data_ptrs, char **coding_ptrs, int size) |
| 170 | { |
| 171 | int i, edd, lastdrive; |
| 172 | int *tmpids; |
| 173 | int *erased, *decoding_matrix, *dm_ids; |
| 174 | |
| 175 | if (w != 8 && w != 16 && w != 32) return -1; |
| 176 | |
| 177 | erased = jerasure_erasures_to_erased(k, m, erasures); |
| 178 | if (erased == NULL((void*)0)) return -1; |
| 179 | |
| 180 | /* Find the number of data drives failed */ |
| 181 | |
| 182 | lastdrive = k; |
| 183 | |
| 184 | edd = 0; |
| 185 | for (i = 0; i < k; i++) { |
| 186 | if (erased[i]) { |
| 187 | edd++; |
| 188 | lastdrive = i; |
| 189 | } |
| 190 | } |
| 191 | |
| 192 | /* You only need to create the decoding matrix in the following cases: |
| 193 | |
| 194 | 1. edd > 0 and row_k_ones is false. |
| 195 | 2. edd > 0 and row_k_ones is true and coding device 0 has been erased. |
| 196 | 3. edd > 1 |
| 197 | |
| 198 | We're going to use lastdrive to denote when to stop decoding data. |
| 199 | At this point in the code, it is equal to the last erased data device. |
| 200 | However, if we can't use the parity row to decode it (i.e. row_k_ones=0 |
| 201 | or erased[k] = 1, we're going to set it to k so that the decoding |
| 202 | pass will decode all data. |
| 203 | */ |
| 204 | |
| 205 | if (!row_k_ones || erased[k]) lastdrive = k; |
| 206 | |
| 207 | dm_ids = NULL((void*)0); |
| 208 | decoding_matrix = NULL((void*)0); |
| 209 | |
| 210 | if (edd > 1 || (edd > 0 && (!row_k_ones || erased[k]))) { |
| 211 | dm_ids = talloc(int, k)(int *) malloc(sizeof(int)*(k)); |
| 212 | if (dm_ids == NULL((void*)0)) { |
| 213 | free(erased); |
| 214 | return -1; |
| 215 | } |
| 216 | |
| 217 | decoding_matrix = talloc(int, k*k)(int *) malloc(sizeof(int)*(k*k)); |
| 218 | if (decoding_matrix == NULL((void*)0)) { |
| 219 | free(erased); |
| 220 | free(dm_ids); |
| 221 | return -1; |
| 222 | } |
| 223 | |
| 224 | if (jerasure_make_decoding_matrix(k, m, w, matrix, erased, decoding_matrix, dm_ids) < 0) { |
| 225 | free(erased); |
| 226 | free(dm_ids); |
| 227 | free(decoding_matrix); |
| 228 | return -1; |
| 229 | } |
| 230 | } |
| 231 | |
| 232 | /* Decode the data drives. |
| 233 | If row_k_ones is true and coding device 0 is intact, then only decode edd-1 drives. |
| 234 | This is done by stopping at lastdrive. |
| 235 | We test whether edd > 0 so that we can exit the loop early if we're done. |
| 236 | */ |
| 237 | |
| 238 | for (i = 0; edd > 0 && i < lastdrive; i++) { |
| 239 | if (erased[i]) { |
| 240 | jerasure_matrix_dotprod(k, w, decoding_matrix+(i*k), dm_ids, i, data_ptrs, coding_ptrs, size); |
| 241 | edd--; |
| 242 | } |
| 243 | } |
| 244 | |
| 245 | /* Then if necessary, decode drive lastdrive */ |
| 246 | |
| 247 | if (edd > 0) { |
| 248 | tmpids = talloc(int, k)(int *) malloc(sizeof(int)*(k)); |
| 249 | for (i = 0; i < k; i++) { |
| 250 | tmpids[i] = (i < lastdrive) ? i : i+1; |
| 251 | } |
| 252 | jerasure_matrix_dotprod(k, w, matrix, tmpids, lastdrive, data_ptrs, coding_ptrs, size); |
| 253 | free(tmpids); |
| 254 | } |
| 255 | |
| 256 | /* Finally, re-encode any erased coding devices */ |
| 257 | |
| 258 | for (i = 0; i < m; i++) { |
| 259 | if (erased[k+i]) { |
| 260 | jerasure_matrix_dotprod(k, w, matrix+(i*k), NULL((void*)0), i+k, data_ptrs, coding_ptrs, size); |
| 261 | } |
| 262 | } |
| 263 | |
| 264 | free(erased); |
| 265 | if (dm_ids != NULL((void*)0)) free(dm_ids); |
| 266 | if (decoding_matrix != NULL((void*)0)) free(decoding_matrix); |
| 267 | |
| 268 | return 0; |
| 269 | } |
| 270 | |
| 271 | |
| 272 | int *jerasure_matrix_to_bitmatrix(int k, int m, int w, int *matrix) |
| 273 | { |
| 274 | int *bitmatrix; |
| 275 | int rowelts, rowindex, colindex, elt, i, j, l, x; |
| 276 | |
| 277 | if (matrix == NULL((void*)0)) { return NULL((void*)0); } |
| 278 | |
| 279 | bitmatrix = talloc(int, k*m*w*w)(int *) malloc(sizeof(int)*(k*m*w*w)); |
| 280 | |
| 281 | rowelts = k * w; |
| 282 | rowindex = 0; |
| 283 | |
| 284 | for (i = 0; i < m; i++) { |
| 285 | colindex = rowindex; |
| 286 | for (j = 0; j < k; j++) { |
| 287 | elt = matrix[i*k+j]; |
| 288 | for (x = 0; x < w; x++) { |
| 289 | for (l = 0; l < w; l++) { |
| 290 | bitmatrix[colindex+x+l*rowelts] = ((elt & (1 << l)) ? 1 : 0); |
| 291 | } |
| 292 | elt = galois_single_multiply(elt, 2, w); |
| 293 | } |
| 294 | colindex += w; |
| 295 | } |
| 296 | rowindex += rowelts * w; |
| 297 | } |
| 298 | return bitmatrix; |
| 299 | } |
| 300 | |
| 301 | void jerasure_matrix_encode(int k, int m, int w, int *matrix, |
| 302 | char **data_ptrs, char **coding_ptrs, int size) |
| 303 | { |
| 304 | int i; |
| 305 | |
| 306 | if (w != 8 && w != 16 && w != 32) { |
| 307 | fprintf(stderr, "ERROR: jerasure_matrix_encode() and w is not 8, 16 or 32\n")__fprintf_chk (stderr, 2 - 1, "ERROR: jerasure_matrix_encode() and w is not 8, 16 or 32\n" ); |
| 308 | assert(0)((void) (0)); |
| 309 | } |
| 310 | |
| 311 | for (i = 0; i < m; i++) { |
| 312 | jerasure_matrix_dotprod(k, w, matrix+(i*k), NULL((void*)0), k+i, data_ptrs, coding_ptrs, size); |
| 313 | } |
| 314 | } |
| 315 | |
| 316 | void jerasure_bitmatrix_dotprod(int k, int w, int *bitmatrix_row, |
| 317 | int *src_ids, int dest_id, |
| 318 | char **data_ptrs, char **coding_ptrs, int size, int packetsize) |
| 319 | { |
| 320 | int j, sindex, pstarted, index, x, y; |
| 321 | char *dptr, *pptr, *bdptr, *bpptr; |
| 322 | |
| 323 | if (size%(w*packetsize) != 0) { |
| 324 | fprintf(stderr, "jerasure_bitmatrix_dotprod - size%c(w*packetsize)) must = 0\n", '%')__fprintf_chk (stderr, 2 - 1, "jerasure_bitmatrix_dotprod - size%c(w*packetsize)) must = 0\n" , '%'); |
| 325 | assert(0)((void) (0)); |
| 326 | } |
| 327 | |
| 328 | bpptr = (dest_id < k) ? data_ptrs[dest_id] : coding_ptrs[dest_id-k]; |
| 329 | |
| 330 | for (sindex = 0; sindex < size; sindex += (packetsize*w)) { |
| 331 | index = 0; |
| 332 | for (j = 0; j < w; j++) { |
| 333 | pstarted = 0; |
| 334 | pptr = bpptr + sindex + j*packetsize; |
| 335 | for (x = 0; x < k; x++) { |
| 336 | if (src_ids == NULL((void*)0)) { |
| 337 | bdptr = data_ptrs[x]; |
| 338 | } else if (src_ids[x] < k) { |
| 339 | bdptr = data_ptrs[src_ids[x]]; |
| 340 | } else { |
| 341 | bdptr = coding_ptrs[src_ids[x]-k]; |
| 342 | } |
| 343 | for (y = 0; y < w; y++) { |
| 344 | if (bitmatrix_row[index]) { |
| 345 | dptr = bdptr + sindex + y*packetsize; |
| 346 | if (!pstarted) { |
| 347 | memcpy(pptr, dptr, packetsize); |
| 348 | jerasure_total_memcpy_bytes += packetsize; |
| 349 | pstarted = 1; |
| 350 | } else { |
| 351 | galois_region_xor(dptr, pptr, packetsize); |
| 352 | jerasure_total_xor_bytes += packetsize; |
| 353 | } |
| 354 | } |
| 355 | index++; |
| 356 | } |
| 357 | } |
| 358 | } |
| 359 | } |
| 360 | } |
| 361 | |
| 362 | void jerasure_do_parity(int k, char **data_ptrs, char *parity_ptr, int size) |
| 363 | { |
| 364 | int i; |
| 365 | |
| 366 | memcpy(parity_ptr, data_ptrs[0], size); |
| 367 | jerasure_total_memcpy_bytes += size; |
| 368 | |
| 369 | for (i = 1; i < k; i++) { |
| 370 | galois_region_xor(data_ptrs[i], parity_ptr, size); |
| 371 | jerasure_total_xor_bytes += size; |
| 372 | } |
| 373 | } |
| 374 | |
| 375 | int jerasure_invert_matrix(int *mat, int *inv, int rows, int w) |
| 376 | { |
| 377 | int cols, i, j, k, x, rs2; |
| 378 | int row_start, tmp, inverse; |
| 379 | |
| 380 | cols = rows; |
| 381 | |
| 382 | k = 0; |
| 383 | for (i = 0; i < rows; i++) { |
| 384 | for (j = 0; j < cols; j++) { |
| 385 | inv[k] = (i == j) ? 1 : 0; |
| 386 | k++; |
| 387 | } |
| 388 | } |
| 389 | |
| 390 | /* First -- convert into upper triangular */ |
| 391 | for (i = 0; i < cols; i++) { |
| 392 | row_start = cols*i; |
| 393 | |
| 394 | /* Swap rows if we ave a zero i,i element. If we can't swap, then the |
| 395 | matrix was not invertible */ |
| 396 | |
| 397 | if (mat[row_start+i] == 0) { |
| 398 | for (j = i+1; j < rows && mat[cols*j+i] == 0; j++) ; |
| 399 | if (j == rows) return -1; |
| 400 | rs2 = j*cols; |
| 401 | for (k = 0; k < cols; k++) { |
| 402 | tmp = mat[row_start+k]; |
| 403 | mat[row_start+k] = mat[rs2+k]; |
| 404 | mat[rs2+k] = tmp; |
| 405 | tmp = inv[row_start+k]; |
| 406 | inv[row_start+k] = inv[rs2+k]; |
| 407 | inv[rs2+k] = tmp; |
| 408 | } |
| 409 | } |
| 410 | |
| 411 | /* Multiply the row by 1/element i,i */ |
| 412 | tmp = mat[row_start+i]; |
| 413 | if (tmp != 1) { |
| 414 | inverse = galois_single_divide(1, tmp, w); |
| 415 | for (j = 0; j < cols; j++) { |
| 416 | mat[row_start+j] = galois_single_multiply(mat[row_start+j], inverse, w); |
| 417 | inv[row_start+j] = galois_single_multiply(inv[row_start+j], inverse, w); |
| 418 | } |
| 419 | } |
| 420 | |
| 421 | /* Now for each j>i, add A_ji*Ai to Aj */ |
| 422 | k = row_start+i; |
| 423 | for (j = i+1; j != cols; j++) { |
| 424 | k += cols; |
| 425 | if (mat[k] != 0) { |
| 426 | if (mat[k] == 1) { |
| 427 | rs2 = cols*j; |
| 428 | for (x = 0; x < cols; x++) { |
| 429 | mat[rs2+x] ^= mat[row_start+x]; |
| 430 | inv[rs2+x] ^= inv[row_start+x]; |
| 431 | } |
| 432 | } else { |
| 433 | tmp = mat[k]; |
| 434 | rs2 = cols*j; |
| 435 | for (x = 0; x < cols; x++) { |
| 436 | mat[rs2+x] ^= galois_single_multiply(tmp, mat[row_start+x], w); |
| 437 | inv[rs2+x] ^= galois_single_multiply(tmp, inv[row_start+x], w); |
| 438 | } |
| 439 | } |
| 440 | } |
| 441 | } |
| 442 | } |
| 443 | |
| 444 | /* Now the matrix is upper triangular. Start at the top and multiply down */ |
| 445 | |
| 446 | for (i = rows-1; i >= 0; i--) { |
| 447 | row_start = i*cols; |
| 448 | for (j = 0; j < i; j++) { |
| 449 | rs2 = j*cols; |
| 450 | if (mat[rs2+i] != 0) { |
| 451 | tmp = mat[rs2+i]; |
| 452 | mat[rs2+i] = 0; |
| 453 | for (k = 0; k < cols; k++) { |
| 454 | inv[rs2+k] ^= galois_single_multiply(tmp, inv[row_start+k], w); |
| 455 | } |
| 456 | } |
| 457 | } |
| 458 | } |
| 459 | return 0; |
| 460 | } |
| 461 | |
| 462 | int jerasure_invertible_matrix(int *mat, int rows, int w) |
| 463 | { |
| 464 | int cols, i, j, k, x, rs2; |
| 465 | int row_start, tmp, inverse; |
| 466 | |
| 467 | cols = rows; |
| 468 | |
| 469 | /* First -- convert into upper triangular */ |
| 470 | for (i = 0; i < cols; i++) { |
| 471 | row_start = cols*i; |
| 472 | |
| 473 | /* Swap rows if we ave a zero i,i element. If we can't swap, then the |
| 474 | matrix was not invertible */ |
| 475 | |
| 476 | if (mat[row_start+i] == 0) { |
| 477 | for (j = i+1; j < rows && mat[cols*j+i] == 0; j++) ; |
| 478 | if (j == rows) return 0; |
| 479 | rs2 = j*cols; |
| 480 | for (k = 0; k < cols; k++) { |
| 481 | tmp = mat[row_start+k]; |
| 482 | mat[row_start+k] = mat[rs2+k]; |
| 483 | mat[rs2+k] = tmp; |
| 484 | } |
| 485 | } |
| 486 | |
| 487 | /* Multiply the row by 1/element i,i */ |
| 488 | tmp = mat[row_start+i]; |
| 489 | if (tmp != 1) { |
| 490 | inverse = galois_single_divide(1, tmp, w); |
| 491 | for (j = 0; j < cols; j++) { |
| 492 | mat[row_start+j] = galois_single_multiply(mat[row_start+j], inverse, w); |
| 493 | } |
| 494 | } |
| 495 | |
| 496 | /* Now for each j>i, add A_ji*Ai to Aj */ |
| 497 | k = row_start+i; |
| 498 | for (j = i+1; j != cols; j++) { |
| 499 | k += cols; |
| 500 | if (mat[k] != 0) { |
| 501 | if (mat[k] == 1) { |
| 502 | rs2 = cols*j; |
| 503 | for (x = 0; x < cols; x++) { |
| 504 | mat[rs2+x] ^= mat[row_start+x]; |
| 505 | } |
| 506 | } else { |
| 507 | tmp = mat[k]; |
| 508 | rs2 = cols*j; |
| 509 | for (x = 0; x < cols; x++) { |
| 510 | mat[rs2+x] ^= galois_single_multiply(tmp, mat[row_start+x], w); |
| 511 | } |
| 512 | } |
| 513 | } |
| 514 | } |
| 515 | } |
| 516 | return 1; |
| 517 | } |
| 518 | |
| 519 | /* Converts a list-style version of the erasures into an array of k+m elements |
| 520 | where the element = 1 if the index has been erased, and zero otherwise */ |
| 521 | |
| 522 | int *jerasure_erasures_to_erased(int k, int m, int *erasures) |
| 523 | { |
| 524 | int td; |
| 525 | int t_non_erased; |
| 526 | int *erased; |
| 527 | int i; |
| 528 | |
| 529 | td = k+m; |
| 530 | erased = talloc(int, td)(int *) malloc(sizeof(int)*(td)); |
| 531 | if (erased == NULL((void*)0)) return NULL((void*)0); |
| 532 | t_non_erased = td; |
| 533 | |
| 534 | for (i = 0; i < td; i++) erased[i] = 0; |
| 535 | |
| 536 | for (i = 0; erasures[i] != -1; i++) { |
| 537 | if (erased[erasures[i]] == 0) { |
| 538 | erased[erasures[i]] = 1; |
| 539 | t_non_erased--; |
| 540 | if (t_non_erased < k) { |
| 541 | free(erased); |
| 542 | return NULL((void*)0); |
| 543 | } |
| 544 | } |
| 545 | } |
| 546 | return erased; |
| 547 | } |
| 548 | |
| 549 | void jerasure_free_schedule(int **schedule) |
| 550 | { |
| 551 | int i; |
| 552 | |
| 553 | for (i = 0; schedule[i][0] >= 0; i++) free(schedule[i]); |
| 554 | free(schedule[i]); |
| 555 | free(schedule); |
| 556 | } |
| 557 | |
| 558 | void jerasure_free_schedule_cache(int k, int m, int ***cache) |
| 559 | { |
| 560 | int e1, e2; |
| 561 | |
| 562 | if (m != 2) { |
| 563 | fprintf(stderr, "jerasure_free_schedule_cache(): m must equal 2\n")__fprintf_chk (stderr, 2 - 1, "jerasure_free_schedule_cache(): m must equal 2\n" ); |
| 564 | assert(0)((void) (0)); |
| 565 | } |
| 566 | |
| 567 | for (e1 = 0; e1 < k+m; e1++) { |
| 568 | for (e2 = 0; e2 < e1; e2++) { |
| 569 | jerasure_free_schedule(cache[e1*(k+m)+e2]); |
| 570 | } |
| 571 | jerasure_free_schedule(cache[e1*(k+m)+e1]); |
| 572 | } |
| 573 | free(cache); |
| 574 | } |
| 575 | |
| 576 | void jerasure_matrix_dotprod(int k, int w, int *matrix_row, |
| 577 | int *src_ids, int dest_id, |
| 578 | char **data_ptrs, char **coding_ptrs, int size) |
| 579 | { |
| 580 | int init; |
| 581 | char *dptr, *sptr; |
| 582 | int i; |
| 583 | |
| 584 | if (w != 1 && w != 8 && w != 16 && w != 32) { |
| 585 | fprintf(stderr, "ERROR: jerasure_matrix_dotprod() called and w is not 1, 8, 16 or 32\n")__fprintf_chk (stderr, 2 - 1, "ERROR: jerasure_matrix_dotprod() called and w is not 1, 8, 16 or 32\n" ); |
| 586 | assert(0)((void) (0)); |
| 587 | } |
| 588 | |
| 589 | init = 0; |
| 590 | |
| 591 | dptr = (dest_id < k) ? data_ptrs[dest_id] : coding_ptrs[dest_id-k]; |
| 592 | |
| 593 | /* First copy or xor any data that does not need to be multiplied by a factor */ |
| 594 | |
| 595 | for (i = 0; i < k; i++) { |
| 596 | if (matrix_row[i] == 1) { |
| 597 | if (src_ids == NULL((void*)0)) { |
| 598 | sptr = data_ptrs[i]; |
| 599 | } else if (src_ids[i] < k) { |
| 600 | sptr = data_ptrs[src_ids[i]]; |
| 601 | } else { |
| 602 | sptr = coding_ptrs[src_ids[i]-k]; |
| 603 | } |
| 604 | if (init == 0) { |
| 605 | memcpy(dptr, sptr, size); |
| 606 | jerasure_total_memcpy_bytes += size; |
| 607 | init = 1; |
| 608 | } else { |
| 609 | galois_region_xor(sptr, dptr, size); |
| 610 | jerasure_total_xor_bytes += size; |
| 611 | } |
| 612 | } |
| 613 | } |
| 614 | |
| 615 | /* Now do the data that needs to be multiplied by a factor */ |
| 616 | |
| 617 | for (i = 0; i < k; i++) { |
| 618 | if (matrix_row[i] != 0 && matrix_row[i] != 1) { |
| 619 | if (src_ids == NULL((void*)0)) { |
| 620 | sptr = data_ptrs[i]; |
| 621 | } else if (src_ids[i] < k) { |
| 622 | sptr = data_ptrs[src_ids[i]]; |
| 623 | } else { |
| 624 | sptr = coding_ptrs[src_ids[i]-k]; |
| 625 | } |
| 626 | switch (w) { |
| 627 | case 8: galois_w08_region_multiply(sptr, matrix_row[i], size, dptr, init); break; |
| 628 | case 16: galois_w16_region_multiply(sptr, matrix_row[i], size, dptr, init); break; |
| 629 | case 32: galois_w32_region_multiply(sptr, matrix_row[i], size, dptr, init); break; |
| 630 | } |
| 631 | jerasure_total_gf_bytes += size; |
| 632 | init = 1; |
| 633 | } |
| 634 | } |
| 635 | } |
| 636 | |
| 637 | |
| 638 | int jerasure_bitmatrix_decode(int k, int m, int w, int *bitmatrix, int row_k_ones, int *erasures, |
| 639 | char **data_ptrs, char **coding_ptrs, int size, int packetsize) |
| 640 | { |
| 641 | int i; |
| 642 | int *erased; |
| 643 | int *decoding_matrix; |
| 644 | int *dm_ids; |
| 645 | int edd, *tmpids, lastdrive; |
| 646 | |
| 647 | erased = jerasure_erasures_to_erased(k, m, erasures); |
| 648 | if (erased == NULL((void*)0)) return -1; |
| 649 | |
| 650 | /* See jerasure_matrix_decode for the logic of this routine. This one works just like |
| 651 | it, but calls the bitmatrix ops instead */ |
| 652 | |
| 653 | lastdrive = k; |
| 654 | |
| 655 | edd = 0; |
| 656 | for (i = 0; i < k; i++) { |
| 657 | if (erased[i]) { |
| 658 | edd++; |
| 659 | lastdrive = i; |
| 660 | } |
| 661 | } |
| 662 | |
| 663 | if (row_k_ones != 1 || erased[k]) lastdrive = k; |
| 664 | |
| 665 | dm_ids = NULL((void*)0); |
| 666 | decoding_matrix = NULL((void*)0); |
| 667 | |
| 668 | if (edd > 1 || (edd > 0 && (row_k_ones != 1 || erased[k]))) { |
| 669 | |
| 670 | dm_ids = talloc(int, k)(int *) malloc(sizeof(int)*(k)); |
| 671 | if (dm_ids == NULL((void*)0)) { |
| 672 | free(erased); |
| 673 | return -1; |
| 674 | } |
| 675 | |
| 676 | decoding_matrix = talloc(int, k*k*w*w)(int *) malloc(sizeof(int)*(k*k*w*w)); |
| 677 | if (decoding_matrix == NULL((void*)0)) { |
| 678 | free(erased); |
| 679 | free(dm_ids); |
| 680 | return -1; |
| 681 | } |
| 682 | |
| 683 | if (jerasure_make_decoding_bitmatrix(k, m, w, bitmatrix, erased, decoding_matrix, dm_ids) < 0) { |
| 684 | free(erased); |
| 685 | free(dm_ids); |
| 686 | free(decoding_matrix); |
| 687 | return -1; |
| 688 | } |
| 689 | } |
| 690 | |
| 691 | for (i = 0; edd > 0 && i < lastdrive; i++) { |
| 692 | if (erased[i]) { |
| 693 | jerasure_bitmatrix_dotprod(k, w, decoding_matrix+i*k*w*w, dm_ids, i, data_ptrs, coding_ptrs, size, packetsize); |
| 694 | edd--; |
| 695 | } |
| 696 | } |
| 697 | |
| 698 | if (edd > 0) { |
| 699 | tmpids = talloc(int, k)(int *) malloc(sizeof(int)*(k)); |
| 700 | for (i = 0; i < k; i++) { |
| 701 | tmpids[i] = (i < lastdrive) ? i : i+1; |
| 702 | } |
| 703 | jerasure_bitmatrix_dotprod(k, w, bitmatrix, tmpids, lastdrive, data_ptrs, coding_ptrs, size, packetsize); |
| 704 | free(tmpids); |
| 705 | } |
| 706 | |
| 707 | for (i = 0; i < m; i++) { |
| 708 | if (erased[k+i]) { |
| 709 | jerasure_bitmatrix_dotprod(k, w, bitmatrix+i*k*w*w, NULL((void*)0), k+i, data_ptrs, coding_ptrs, size, packetsize); |
| 710 | } |
| 711 | } |
| 712 | |
| 713 | free(erased); |
| 714 | if (dm_ids != NULL((void*)0)) free(dm_ids); |
| 715 | if (decoding_matrix != NULL((void*)0)) free(decoding_matrix); |
| 716 | |
| 717 | return 0; |
| 718 | } |
| 719 | |
| 720 | static char **set_up_ptrs_for_scheduled_decoding(int k, int m, int *erasures, char **data_ptrs, char **coding_ptrs) |
| 721 | { |
| 722 | int ddf, cdf; |
| 723 | int *erased; |
| 724 | char **ptrs; |
| 725 | int i, j, x; |
| 726 | |
| 727 | ddf = 0; |
| 728 | cdf = 0; |
| 729 | for (i = 0; erasures[i] != -1; i++) { |
| 730 | if (erasures[i] < k) ddf++; else cdf++; |
| 731 | } |
| 732 | |
| 733 | erased = jerasure_erasures_to_erased(k, m, erasures); |
| 734 | if (erased == NULL((void*)0)) return NULL((void*)0); |
| 735 | |
| 736 | /* Set up ptrs. It will be as follows: |
| 737 | |
| 738 | - If data drive i has not failed, then ptrs[i] = data_ptrs[i]. |
| 739 | - If data drive i has failed, then ptrs[i] = coding_ptrs[j], where j is the |
| 740 | lowest unused non-failed coding drive. |
| 741 | - Elements k to k+ddf-1 are data_ptrs[] of the failed data drives. |
| 742 | - Elements k+ddf to k+ddf+cdf-1 are coding_ptrs[] of the failed data drives. |
| 743 | |
| 744 | The array row_ids contains the ids of ptrs. |
| 745 | The array ind_to_row_ids contains the row_id of drive i. |
| 746 | |
| 747 | However, we're going to set row_ids and ind_to_row in a different procedure. |
| 748 | */ |
| 749 | |
| 750 | ptrs = talloc(char *, k+m)(char * *) malloc(sizeof(char *)*(k+m)); |
| 751 | |
| 752 | j = k; |
| 753 | x = k; |
| 754 | for (i = 0; i < k; i++) { |
| 755 | if (erased[i] == 0) { |
| 756 | ptrs[i] = data_ptrs[i]; |
| 757 | } else { |
| 758 | while (erased[j]) j++; |
| 759 | ptrs[i] = coding_ptrs[j-k]; |
| 760 | j++; |
| 761 | ptrs[x] = data_ptrs[i]; |
| 762 | x++; |
| 763 | } |
| 764 | } |
| 765 | for (i = k; i < k+m; i++) { |
| 766 | if (erased[i]) { |
| 767 | ptrs[x] = coding_ptrs[i-k]; |
| 768 | x++; |
| 769 | } |
| 770 | } |
| 771 | free(erased); |
| 772 | return ptrs; |
| 773 | } |
| 774 | |
| 775 | static int set_up_ids_for_scheduled_decoding(int k, int m, int *erasures, int *row_ids, int *ind_to_row) |
| 776 | { |
| 777 | int ddf, cdf; |
| 778 | int *erased; |
| 779 | int i, j, x; |
| 780 | |
| 781 | ddf = 0; |
| 782 | cdf = 0; |
| 783 | for (i = 0; erasures[i] != -1; i++) { |
| 784 | if (erasures[i] < k) ddf++; else cdf++; |
| 785 | } |
| 786 | |
| 787 | erased = jerasure_erasures_to_erased(k, m, erasures); |
| 788 | if (erased == NULL((void*)0)) return -1; |
| 789 | |
| 790 | /* See set_up_ptrs_for_scheduled_decoding for how these are set */ |
| 791 | |
| 792 | j = k; |
| 793 | x = k; |
| 794 | for (i = 0; i < k; i++) { |
| 795 | if (erased[i] == 0) { |
| 796 | row_ids[i] = i; |
| 797 | ind_to_row[i] = i; |
| 798 | } else { |
| 799 | while (erased[j]) j++; |
| 800 | row_ids[i] = j; |
| 801 | ind_to_row[j] = i; |
| 802 | j++; |
| 803 | row_ids[x] = i; |
| 804 | ind_to_row[i] = x; |
| 805 | x++; |
| 806 | } |
| 807 | } |
| 808 | for (i = k; i < k+m; i++) { |
| 809 | if (erased[i]) { |
| 810 | row_ids[x] = i; |
| 811 | ind_to_row[i] = x; |
| 812 | x++; |
| 813 | } |
| 814 | } |
| 815 | free(erased); |
| 816 | return 0; |
| 817 | } |
| 818 | |
| 819 | static int **jerasure_generate_decoding_schedule(int k, int m, int w, int *bitmatrix, int *erasures, int smart) |
| 820 | { |
| 821 | int i, j, x, drive, y, index, z; |
| 822 | int *decoding_matrix, *inverse, *real_decoding_matrix; |
| 823 | int *ptr; |
| 824 | int *row_ids; |
| 825 | int *ind_to_row; |
| 826 | int ddf, cdf; |
| 827 | int **schedule; |
| 828 | int *b1, *b2; |
| 829 | |
| 830 | /* First, figure out the number of data drives that have failed, and the |
| 831 | number of coding drives that have failed: ddf and cdf */ |
| 832 | |
| 833 | ddf = 0; |
| 834 | cdf = 0; |
| 835 | for (i = 0; erasures[i] != -1; i++) { |
| 836 | if (erasures[i] < k) ddf++; else cdf++; |
| 837 | } |
| 838 | |
| 839 | row_ids = talloc(int, k+m)(int *) malloc(sizeof(int)*(k+m)); |
| 840 | ind_to_row = talloc(int, k+m)(int *) malloc(sizeof(int)*(k+m)); |
| 841 | |
| 842 | if (set_up_ids_for_scheduled_decoding(k, m, erasures, row_ids, ind_to_row) < 0) return NULL((void*)0); |
| 843 | |
| 844 | /* Now, we're going to create one decoding matrix which is going to |
| 845 | decode everything with one call. The hope is that the scheduler |
| 846 | will do a good job. This matrix has w*e rows, where e is the |
| 847 | number of erasures (ddf+cdf) */ |
| 848 | |
| 849 | real_decoding_matrix = talloc(int, k*w*(cdf+ddf)*w)(int *) malloc(sizeof(int)*(k*w*(cdf+ddf)*w)); |
| 850 | |
| 851 | /* First, if any data drives have failed, then initialize the first |
| 852 | ddf*w rows of the decoding matrix from the standard decoding |
| 853 | matrix inversion */ |
| 854 | |
| 855 | if (ddf > 0) { |
| 856 | |
| 857 | decoding_matrix = talloc(int, k*k*w*w)(int *) malloc(sizeof(int)*(k*k*w*w)); |
| 858 | ptr = decoding_matrix; |
| 859 | for (i = 0; i < k; i++) { |
| 860 | if (row_ids[i] == i) { |
| 861 | bzero(ptr, k*w*w*sizeof(int)); |
| 862 | for (x = 0; x < w; x++) { |
| 863 | ptr[x+i*w+x*k*w] = 1; |
| 864 | } |
| 865 | } else { |
| 866 | memcpy(ptr, bitmatrix+k*w*w*(row_ids[i]-k), k*w*w*sizeof(int)); |
| 867 | } |
| 868 | ptr += (k*w*w); |
| 869 | } |
| 870 | inverse = talloc(int, k*k*w*w)(int *) malloc(sizeof(int)*(k*k*w*w)); |
| 871 | jerasure_invert_bitmatrix(decoding_matrix, inverse, k*w); |
| 872 | |
| 873 | /* printf("\nMatrix to invert\n"); |
| 874 | jerasure_print_bitmatrix(decoding_matrix, k*w, k*w, w); |
| 875 | printf("\n"); |
| 876 | printf("\nInverse\n"); |
| 877 | jerasure_print_bitmatrix(inverse, k*w, k*w, w); |
| 878 | printf("\n"); */ |
| 879 | |
| 880 | free(decoding_matrix); |
| 881 | ptr = real_decoding_matrix; |
| 882 | for (i = 0; i < ddf; i++) { |
| 883 | memcpy(ptr, inverse+k*w*w*row_ids[k+i], sizeof(int)*k*w*w); |
| 884 | ptr += (k*w*w); |
| 885 | } |
| 886 | free(inverse); |
| 887 | } |
| 888 | |
| 889 | /* Next, here comes the hard part. For each coding node that needs |
| 890 | to be decoded, you start by putting its rows of the distribution |
| 891 | matrix into the decoding matrix. If there were no failed data |
| 892 | nodes, then you're done. However, if there have been failed |
| 893 | data nodes, then you need to modify the columns that correspond |
| 894 | to the data nodes. You do that by first zeroing them. Then |
| 895 | whereever there is a one in the distribution matrix, you XOR |
| 896 | in the corresponding row from the failed data node's entry in |
| 897 | the decoding matrix. The whole process kind of makes my head |
| 898 | spin, but it works. |
| 899 | */ |
| 900 | |
| 901 | for (x = 0; x < cdf; x++) { |
| 902 | drive = row_ids[x+ddf+k]-k; |
| 903 | ptr = real_decoding_matrix + k*w*w*(ddf+x); |
| 904 | memcpy(ptr, bitmatrix+drive*k*w*w, sizeof(int)*k*w*w); |
| 905 | |
| 906 | for (i = 0; i < k; i++) { |
| 907 | if (row_ids[i] != i) { |
| 908 | for (j = 0; j < w; j++) { |
| 909 | bzero(ptr+j*k*w+i*w, sizeof(int)*w); |
| 910 | } |
| 911 | } |
| 912 | } |
| 913 | |
| 914 | /* There's the yucky part */ |
| 915 | |
| 916 | index = drive*k*w*w; |
| 917 | for (i = 0; i < k; i++) { |
| 918 | if (row_ids[i] != i) { |
| 919 | b1 = real_decoding_matrix+(ind_to_row[i]-k)*k*w*w; |
| 920 | for (j = 0; j < w; j++) { |
| 921 | b2 = ptr + j*k*w; |
| 922 | for (y = 0; y < w; y++) { |
| 923 | if (bitmatrix[index+j*k*w+i*w+y]) { |
| 924 | for (z = 0; z < k*w; z++) { |
| 925 | b2[z] = b2[z] ^ b1[z+y*k*w]; |
| 926 | } |
| 927 | } |
| 928 | } |
| 929 | } |
| 930 | } |
| 931 | } |
| 932 | } |
| 933 | |
| 934 | /* |
| 935 | printf("\n\nReal Decoding Matrix\n\n"); |
| 936 | jerasure_print_bitmatrix(real_decoding_matrix, (ddf+cdf)*w, k*w, w); |
| 937 | printf("\n"); */ |
| 938 | if (smart) { |
| 939 | schedule = jerasure_smart_bitmatrix_to_schedule(k, ddf+cdf, w, real_decoding_matrix); |
| 940 | } else { |
| 941 | schedule = jerasure_dumb_bitmatrix_to_schedule(k, ddf+cdf, w, real_decoding_matrix); |
| 942 | } |
| 943 | free(row_ids); |
| 944 | free(ind_to_row); |
| 945 | free(real_decoding_matrix); |
| 946 | return schedule; |
| 947 | } |
| 948 | |
| 949 | int jerasure_schedule_decode_lazy(int k, int m, int w, int *bitmatrix, int *erasures, |
| 950 | char **data_ptrs, char **coding_ptrs, int size, int packetsize, |
| 951 | int smart) |
| 952 | { |
| 953 | int i, tdone; |
| 954 | char **ptrs; |
| 955 | int **schedule; |
| 956 | |
| 957 | ptrs = set_up_ptrs_for_scheduled_decoding(k, m, erasures, data_ptrs, coding_ptrs); |
| 958 | if (ptrs == NULL((void*)0)) return -1; |
| 959 | |
| 960 | schedule = jerasure_generate_decoding_schedule(k, m, w, bitmatrix, erasures, smart); |
| 961 | if (schedule == NULL((void*)0)) { |
| 962 | free(ptrs); |
| 963 | return -1; |
| 964 | } |
| 965 | |
| 966 | for (tdone = 0; tdone < size; tdone += packetsize*w) { |
| 967 | jerasure_do_scheduled_operations(ptrs, schedule, packetsize); |
| 968 | for (i = 0; i < k+m; i++) ptrs[i] += (packetsize*w); |
| 969 | } |
| 970 | |
| 971 | jerasure_free_schedule(schedule); |
| 972 | free(ptrs); |
| 973 | |
| 974 | return 0; |
| 975 | } |
| 976 | |
| 977 | int jerasure_schedule_decode_cache(int k, int m, int w, int ***scache, int *erasures, |
| 978 | char **data_ptrs, char **coding_ptrs, int size, int packetsize) |
| 979 | { |
| 980 | int i, tdone; |
| 981 | char **ptrs; |
| 982 | int **schedule; |
| 983 | int index; |
| 984 | |
| 985 | if (erasures[1] == -1) { |
| 986 | index = erasures[0]*(k+m) + erasures[0]; |
| 987 | } else if (erasures[2] == -1) { |
| 988 | index = erasures[0]*(k+m) + erasures[1]; |
| 989 | } else { |
| 990 | return -1; |
| 991 | } |
| 992 | |
| 993 | schedule = scache[index]; |
| 994 | |
| 995 | ptrs = set_up_ptrs_for_scheduled_decoding(k, m, erasures, data_ptrs, coding_ptrs); |
| 996 | if (ptrs == NULL((void*)0)) return -1; |
| 997 | |
| 998 | |
| 999 | for (tdone = 0; tdone < size; tdone += packetsize*w) { |
| 1000 | jerasure_do_scheduled_operations(ptrs, schedule, packetsize); |
| 1001 | for (i = 0; i < k+m; i++) ptrs[i] += (packetsize*w); |
| 1002 | } |
| 1003 | |
| 1004 | free(ptrs); |
| 1005 | |
| 1006 | return 0; |
| 1007 | } |
| 1008 | |
| 1009 | /* This only works when m = 2 */ |
| 1010 | |
| 1011 | int ***jerasure_generate_schedule_cache(int k, int m, int w, int *bitmatrix, int smart) |
| 1012 | { |
| 1013 | int ***scache; |
| 1014 | int erasures[3]; |
| 1015 | int e1, e2; |
| 1016 | |
| 1017 | /* Ok -- this is yucky, but it's how I'm doing it. You will make an index out |
| 1018 | of erasures, which will be e1*(k+m)+(e2). If there is no e2, then e2 = e1. |
| 1019 | Isn't that clever and confusing. Sorry. |
| 1020 | |
| 1021 | We're not going to worry about ordering -- in other words, the schedule for |
| 1022 | e1,e2 will be the same as e2,e1. They will have the same pointer -- the |
| 1023 | schedule will not be duplicated. */ |
| 1024 | |
| 1025 | if (m != 2) return NULL((void*)0); |
| 1026 | |
| 1027 | scache = talloc(int **, (k+m)*(k+m+1))(int ** *) malloc(sizeof(int **)*((k+m)*(k+m+1))); |
| 1028 | if (scache == NULL((void*)0)) return NULL((void*)0); |
| 1029 | |
| 1030 | for (e1 = 0; e1 < k+m; e1++) { |
| 1031 | erasures[0] = e1; |
| 1032 | for (e2 = 0; e2 < e1; e2++) { |
| 1033 | erasures[1] = e2; |
| 1034 | erasures[2] = -1; |
| 1035 | scache[e1*(k+m)+e2] = jerasure_generate_decoding_schedule(k, m, w, bitmatrix, erasures, smart); |
| 1036 | scache[e2*(k+m)+e1] = scache[e1*(k+m)+e2]; |
| 1037 | } |
| 1038 | erasures[1] = -1; |
| 1039 | scache[e1*(k+m)+e1] = jerasure_generate_decoding_schedule(k, m, w, bitmatrix, erasures, smart); |
| 1040 | } |
| 1041 | return scache; |
| 1042 | |
| 1043 | } |
| 1044 | |
| 1045 | int jerasure_invert_bitmatrix(int *mat, int *inv, int rows) |
| 1046 | { |
| 1047 | int cols, i, j, k; |
| 1048 | int tmp; |
| 1049 | |
| 1050 | cols = rows; |
| 1051 | |
| 1052 | k = 0; |
| 1053 | for (i = 0; i < rows; i++) { |
| 1054 | for (j = 0; j < cols; j++) { |
| 1055 | inv[k] = (i == j) ? 1 : 0; |
| 1056 | k++; |
| 1057 | } |
| 1058 | } |
| 1059 | |
| 1060 | /* First -- convert into upper triangular */ |
| 1061 | |
| 1062 | for (i = 0; i < cols; i++) { |
| 1063 | |
| 1064 | /* Swap rows if we have a zero i,i element. If we can't swap, then the |
| 1065 | matrix was not invertible */ |
| 1066 | |
| 1067 | if ((mat[i*cols+i]) == 0) { |
| 1068 | for (j = i+1; j < rows && (mat[j*cols+i]) == 0; j++) ; |
| 1069 | if (j == rows) return -1; |
| 1070 | for (k = 0; k < cols; k++) { |
| 1071 | tmp = mat[i*cols+k]; mat[i*cols+k] = mat[j*cols+k]; mat[j*cols+k] = tmp; |
| 1072 | tmp = inv[i*cols+k]; inv[i*cols+k] = inv[j*cols+k]; inv[j*cols+k] = tmp; |
| 1073 | } |
| 1074 | } |
| 1075 | |
| 1076 | /* Now for each j>i, add A_ji*Ai to Aj */ |
| 1077 | for (j = i+1; j != rows; j++) { |
| 1078 | if (mat[j*cols+i] != 0) { |
| 1079 | for (k = 0; k < cols; k++) { |
| 1080 | mat[j*cols+k] ^= mat[i*cols+k]; |
| 1081 | inv[j*cols+k] ^= inv[i*cols+k]; |
| 1082 | } |
| 1083 | } |
| 1084 | } |
| 1085 | } |
| 1086 | |
| 1087 | /* Now the matrix is upper triangular. Start at the top and multiply down */ |
| 1088 | |
| 1089 | for (i = rows-1; i >= 0; i--) { |
| 1090 | for (j = 0; j < i; j++) { |
| 1091 | if (mat[j*cols+i]) { |
| 1092 | for (k = 0; k < cols; k++) { |
| 1093 | mat[j*cols+k] ^= mat[i*cols+k]; |
| 1094 | inv[j*cols+k] ^= inv[i*cols+k]; |
| 1095 | } |
| 1096 | } |
| 1097 | } |
| 1098 | } |
| 1099 | return 0; |
| 1100 | } |
| 1101 | |
| 1102 | int jerasure_invertible_bitmatrix(int *mat, int rows) |
| 1103 | { |
| 1104 | int cols, i, j, k; |
| 1105 | int tmp; |
| 1106 | |
| 1107 | cols = rows; |
| 1108 | |
| 1109 | /* First -- convert into upper triangular */ |
| 1110 | |
| 1111 | for (i = 0; i < cols; i++) { |
| 1112 | |
| 1113 | /* Swap rows if we have a zero i,i element. If we can't swap, then the |
| 1114 | matrix was not invertible */ |
| 1115 | |
| 1116 | if ((mat[i*cols+i]) == 0) { |
| 1117 | for (j = i+1; j < rows && (mat[j*cols+i]) == 0; j++) ; |
| 1118 | if (j == rows) return 0; |
| 1119 | for (k = 0; k < cols; k++) { |
| 1120 | tmp = mat[i*cols+k]; mat[i*cols+k] = mat[j*cols+k]; mat[j*cols+k] = tmp; |
| 1121 | } |
| 1122 | } |
| 1123 | |
| 1124 | /* Now for each j>i, add A_ji*Ai to Aj */ |
| 1125 | for (j = i+1; j != rows; j++) { |
| 1126 | if (mat[j*cols+i] != 0) { |
| 1127 | for (k = 0; k < cols; k++) { |
| 1128 | mat[j*cols+k] ^= mat[i*cols+k]; |
| 1129 | } |
| 1130 | } |
| 1131 | } |
| 1132 | } |
| 1133 | return 1; |
| 1134 | } |
| 1135 | |
| 1136 | |
| 1137 | int *jerasure_matrix_multiply(int *m1, int *m2, int r1, int c1, int r2, int c2, int w) |
| 1138 | { |
| 1139 | int *product, i, j, k; |
| 1140 | |
| 1141 | product = (int *) malloc(sizeof(int)*r1*c2); |
| 1142 | for (i = 0; i < r1*c2; i++) product[i] = 0; |
| 1143 | |
| 1144 | for (i = 0; i < r1; i++) { |
| 1145 | for (j = 0; j < c2; j++) { |
| 1146 | for (k = 0; k < r2; k++) { |
| 1147 | product[i*c2+j] ^= galois_single_multiply(m1[i*c1+k], m2[k*c2+j], w); |
| 1148 | } |
| 1149 | } |
| 1150 | } |
| 1151 | return product; |
| 1152 | } |
| 1153 | |
| 1154 | void jerasure_get_stats(double *fill_in) |
| 1155 | { |
| 1156 | fill_in[0] = jerasure_total_xor_bytes; |
| 1157 | fill_in[1] = jerasure_total_gf_bytes; |
| 1158 | fill_in[2] = jerasure_total_memcpy_bytes; |
| 1159 | jerasure_total_xor_bytes = 0; |
| 1160 | jerasure_total_gf_bytes = 0; |
| 1161 | jerasure_total_memcpy_bytes = 0; |
| 1162 | } |
| 1163 | |
| 1164 | void jerasure_do_scheduled_operations(char **ptrs, int **operations, int packetsize) |
| 1165 | { |
| 1166 | char *sptr; |
| 1167 | char *dptr; |
| 1168 | int op; |
| 1169 | |
| 1170 | for (op = 0; operations[op][0] >= 0; op++) { |
| 1171 | sptr = ptrs[operations[op][0]] + operations[op][1]*packetsize; |
| 1172 | dptr = ptrs[operations[op][2]] + operations[op][3]*packetsize; |
| 1173 | if (operations[op][4]) { |
| 1174 | /* printf("%d,%d %d,%d\n", operations[op][0], |
| 1175 | operations[op][1], |
| 1176 | operations[op][2], |
| 1177 | operations[op][3]); |
| 1178 | printf("xor(0x%x, 0x%x -> 0x%x, %d)\n", sptr, dptr, dptr, packetsize); */ |
| 1179 | galois_region_xor(sptr, dptr, packetsize); |
| 1180 | jerasure_total_xor_bytes += packetsize; |
| 1181 | } else { |
| 1182 | /* printf("memcpy(0x%x <- 0x%x)\n", dptr, sptr); */ |
| 1183 | memcpy(dptr, sptr, packetsize); |
| 1184 | jerasure_total_memcpy_bytes += packetsize; |
| 1185 | } |
| 1186 | } |
| 1187 | } |
| 1188 | |
| 1189 | void jerasure_schedule_encode(int k, int m, int w, int **schedule, |
| 1190 | char **data_ptrs, char **coding_ptrs, int size, int packetsize) |
| 1191 | { |
| 1192 | char **ptr_copy; |
| 1193 | int i, tdone; |
| 1194 | |
| 1195 | ptr_copy = talloc(char *, (k+m))(char * *) malloc(sizeof(char *)*((k+m))); |
| 1196 | for (i = 0; i < k; i++) ptr_copy[i] = data_ptrs[i]; |
| 1197 | for (i = 0; i < m; i++) ptr_copy[i+k] = coding_ptrs[i]; |
| 1198 | for (tdone = 0; tdone < size; tdone += packetsize*w) { |
| 1199 | jerasure_do_scheduled_operations(ptr_copy, schedule, packetsize); |
| 1200 | for (i = 0; i < k+m; i++) ptr_copy[i] += (packetsize*w); |
| 1201 | } |
| 1202 | free(ptr_copy); |
| 1203 | } |
| 1204 | |
| 1205 | int **jerasure_dumb_bitmatrix_to_schedule(int k, int m, int w, int *bitmatrix) |
| 1206 | { |
| 1207 | int **operations; |
| 1208 | int op; |
| 1209 | int index, optodo, i, j; |
| 1210 | |
| 1211 | operations = talloc(int *, k*m*w*w+1)(int * *) malloc(sizeof(int *)*(k*m*w*w+1)); |
| 1212 | op = 0; |
| 1213 | |
| 1214 | index = 0; |
| 1215 | for (i = 0; i < m*w; i++) { |
| 1216 | optodo = 0; |
| 1217 | for (j = 0; j < k*w; j++) { |
| 1218 | if (bitmatrix[index]) { |
| 1219 | operations[op] = talloc(int, 5)(int *) malloc(sizeof(int)*(5)); |
| 1220 | operations[op][4] = optodo; |
| 1221 | operations[op][0] = j/w; |
| 1222 | operations[op][1] = j%w; |
| 1223 | operations[op][2] = k+i/w; |
| 1224 | operations[op][3] = i%w; |
| 1225 | optodo = 1; |
| 1226 | op++; |
| 1227 | |
| 1228 | } |
| 1229 | index++; |
| 1230 | } |
| 1231 | } |
| 1232 | operations[op] = talloc(int, 5)(int *) malloc(sizeof(int)*(5)); |
| 1233 | operations[op][0] = -1; |
| 1234 | return operations; |
| 1235 | } |
| 1236 | |
| 1237 | int **jerasure_smart_bitmatrix_to_schedule(int k, int m, int w, int *bitmatrix) |
| 1238 | { |
| 1239 | int **operations; |
| 1240 | int op; |
| 1241 | int i, j; |
| 1242 | int *diff, *from, *b1, *flink, *blink; |
| 1243 | int *ptr, no, row; |
| 1244 | int optodo; |
| 1245 | int bestrow = 0, bestdiff, top; |
| 1246 | |
| 1247 | /* printf("Scheduling:\n\n"); |
| 1248 | jerasure_print_bitmatrix(bitmatrix, m*w, k*w, w); */ |
| 1249 | |
| 1250 | operations = talloc(int *, k*m*w*w+1)(int * *) malloc(sizeof(int *)*(k*m*w*w+1)); |
| 1251 | op = 0; |
| 1252 | |
| 1253 | diff = talloc(int, m*w)(int *) malloc(sizeof(int)*(m*w)); |
| 1254 | from = talloc(int, m*w)(int *) malloc(sizeof(int)*(m*w)); |
| 1255 | flink = talloc(int, m*w)(int *) malloc(sizeof(int)*(m*w)); |
| 1256 | blink = talloc(int, m*w)(int *) malloc(sizeof(int)*(m*w)); |
| 1257 | |
| 1258 | ptr = bitmatrix; |
| 1259 | |
| 1260 | bestdiff = k*w+1; |
| 1261 | top = 0; |
| 1262 | for (i = 0; i < m*w; i++) { |
| 1263 | no = 0; |
| 1264 | for (j = 0; j < k*w; j++) { |
| 1265 | no += *ptr; |
| 1266 | ptr++; |
| 1267 | } |
| 1268 | diff[i] = no; |
| 1269 | from[i] = -1; |
| 1270 | flink[i] = i+1; |
| 1271 | blink[i] = i-1; |
| 1272 | if (no < bestdiff) { |
| 1273 | bestdiff = no; |
| 1274 | bestrow = i; |
| 1275 | } |
| 1276 | } |
| 1277 | |
| 1278 | flink[m*w-1] = -1; |
| 1279 | |
| 1280 | while (top != -1) { |
| 1281 | row = bestrow; |
| 1282 | /* printf("Doing row %d - %d from %d\n", row, diff[row], from[row]); */ |
| 1283 | |
| 1284 | if (blink[row] == -1) { |
| 1285 | top = flink[row]; |
| 1286 | if (top != -1) blink[top] = -1; |
| 1287 | } else { |
| 1288 | flink[blink[row]] = flink[row]; |
| 1289 | if (flink[row] != -1) { |
| 1290 | blink[flink[row]] = blink[row]; |
| 1291 | } |
| 1292 | } |
| 1293 | |
| 1294 | ptr = bitmatrix + row*k*w; |
| 1295 | if (from[row] == -1) { |
| 1296 | optodo = 0; |
| 1297 | for (j = 0; j < k*w; j++) { |
| 1298 | if (ptr[j]) { |
| 1299 | operations[op] = talloc(int, 5)(int *) malloc(sizeof(int)*(5)); |
| 1300 | operations[op][4] = optodo; |
| 1301 | operations[op][0] = j/w; |
| 1302 | operations[op][1] = j%w; |
| 1303 | operations[op][2] = k+row/w; |
| 1304 | operations[op][3] = row%w; |
| 1305 | optodo = 1; |
| 1306 | op++; |
| 1307 | } |
| 1308 | } |
| 1309 | } else { |
| 1310 | operations[op] = talloc(int, 5)(int *) malloc(sizeof(int)*(5)); |
| 1311 | operations[op][4] = 0; |
| 1312 | operations[op][0] = k+from[row]/w; |
| 1313 | operations[op][1] = from[row]%w; |
| 1314 | operations[op][2] = k+row/w; |
| 1315 | operations[op][3] = row%w; |
| 1316 | op++; |
| 1317 | b1 = bitmatrix + from[row]*k*w; |
| 1318 | for (j = 0; j < k*w; j++) { |
| 1319 | if (ptr[j] ^ b1[j]) { |
| 1320 | operations[op] = talloc(int, 5)(int *) malloc(sizeof(int)*(5)); |
| 1321 | operations[op][4] = 1; |
| 1322 | operations[op][0] = j/w; |
| 1323 | operations[op][1] = j%w; |
| 1324 | operations[op][2] = k+row/w; |
| 1325 | operations[op][3] = row%w; |
| 1326 | optodo = 1; |
Value stored to 'optodo' is never read | |
| 1327 | op++; |
| 1328 | } |
| 1329 | } |
| 1330 | } |
| 1331 | bestdiff = k*w+1; |
| 1332 | for (i = top; i != -1; i = flink[i]) { |
| 1333 | no = 1; |
| 1334 | b1 = bitmatrix + i*k*w; |
| 1335 | for (j = 0; j < k*w; j++) no += (ptr[j] ^ b1[j]); |
| 1336 | if (no < diff[i]) { |
| 1337 | from[i] = row; |
| 1338 | diff[i] = no; |
| 1339 | } |
| 1340 | if (diff[i] < bestdiff) { |
| 1341 | bestdiff = diff[i]; |
| 1342 | bestrow = i; |
| 1343 | } |
| 1344 | } |
| 1345 | } |
| 1346 | |
| 1347 | operations[op] = talloc(int, 5)(int *) malloc(sizeof(int)*(5)); |
| 1348 | operations[op][0] = -1; |
| 1349 | free(from); |
| 1350 | free(diff); |
| 1351 | free(blink); |
| 1352 | free(flink); |
| 1353 | |
| 1354 | return operations; |
| 1355 | } |
| 1356 | |
| 1357 | void jerasure_bitmatrix_encode(int k, int m, int w, int *bitmatrix, |
| 1358 | char **data_ptrs, char **coding_ptrs, int size, int packetsize) |
| 1359 | { |
| 1360 | int i; |
| 1361 | |
| 1362 | if (packetsize%sizeof(long) != 0) { |
| 1363 | fprintf(stderr, "jerasure_bitmatrix_encode - packetsize(%d) %c sizeof(long) != 0\n", packetsize, '%')__fprintf_chk (stderr, 2 - 1, "jerasure_bitmatrix_encode - packetsize(%d) %c sizeof(long) != 0\n" , packetsize, '%'); |
| 1364 | assert(0)((void) (0)); |
| 1365 | } |
| 1366 | if (size%(packetsize*w) != 0) { |
| 1367 | fprintf(stderr, "jerasure_bitmatrix_encode - size(%d) %c (packetsize(%d)*w(%d))) != 0\n",__fprintf_chk (stderr, 2 - 1, "jerasure_bitmatrix_encode - size(%d) %c (packetsize(%d)*w(%d))) != 0\n" , size, '%', packetsize, w) |
| 1368 | size, '%', packetsize, w)__fprintf_chk (stderr, 2 - 1, "jerasure_bitmatrix_encode - size(%d) %c (packetsize(%d)*w(%d))) != 0\n" , size, '%', packetsize, w); |
| 1369 | assert(0)((void) (0)); |
| 1370 | } |
| 1371 | |
| 1372 | for (i = 0; i < m; i++) { |
| 1373 | jerasure_bitmatrix_dotprod(k, w, bitmatrix+i*k*w*w, NULL((void*)0), k+i, data_ptrs, coding_ptrs, size, packetsize); |
| 1374 | } |
| 1375 | } |
| 1376 | |
| 1377 | /* |
| 1378 | * Exported function for use by autoconf to perform quick |
| 1379 | * spot-check. |
| 1380 | */ |
| 1381 | int jerasure_autoconf_test() |
| 1382 | { |
| 1383 | int x = galois_single_multiply(1, 2, 8); |
| 1384 | if (x != 2) { |
| 1385 | return -1; |
| 1386 | } |
| 1387 | return 0; |
| 1388 | } |
| 1389 |