1// SPDX-License-Identifier: GPL-2.0 2/* 3 * Generic binary BCH encoding/decoding library 4 * 5 * Copyright �� 2011 Parrot S.A. 6 * 7 * Author: Ivan Djelic <ivan.djelic@parrot.com> 8 * 9 * Description: 10 * 11 * This library provides runtime configurable encoding/decoding of binary 12 * Bose-Chaudhuri-Hocquenghem (BCH) codes. 13 * 14 * Call init_bch to get a pointer to a newly allocated bch_control structure for 15 * the given m (Galois field order), t (error correction capability) and 16 * (optional) primitive polynomial parameters. 17 * 18 * Call encode_bch to compute and store ecc parity bytes to a given buffer. 19 * Call decode_bch to detect and locate errors in received data. 20 * 21 * On systems supporting hw BCH features, intermediate results may be provided 22 * to decode_bch in order to skip certain steps. See decode_bch() documentation 23 * for details. 24 * 25 * Option CONFIG_BCH_CONST_PARAMS can be used to force fixed values of 26 * parameters m and t; thus allowing extra compiler optimizations and providing 27 * better (up to 2x) encoding performance. Using this option makes sense when 28 * (m,t) are fixed and known in advance, e.g. when using BCH error correction 29 * on a particular NAND flash device. 30 * 31 * Algorithmic details: 32 * 33 * Encoding is performed by processing 32 input bits in parallel, using 4 34 * remainder lookup tables. 35 * 36 * The final stage of decoding involves the following internal steps: 37 * a. Syndrome computation 38 * b. Error locator polynomial computation using Berlekamp-Massey algorithm 39 * c. Error locator root finding (by far the most expensive step) 40 * 41 * In this implementation, step c is not performed using the usual Chien search. 42 * Instead, an alternative approach described in [1] is used. It consists in 43 * factoring the error locator polynomial using the Berlekamp Trace algorithm 44 * (BTA) down to a certain degree (4), after which ad hoc low-degree polynomial 45 * solving techniques [2] are used. The resulting algorithm, called BTZ, yields 46 * much better performance than Chien search for usual (m,t) values (typically 47 * m >= 13, t < 32, see [1]). 48 * 49 * [1] B. Biswas, V. Herbert. Efficient root finding of polynomials over fields 50 * of characteristic 2, in: Western European Workshop on Research in Cryptology 51 * - WEWoRC 2009, Graz, Austria, LNCS, Springer, July 2009, to appear. 52 * [2] [Zin96] V.A. Zinoviev. On the solution of equations of degree 10 over 53 * finite fields GF(2^q). In Rapport de recherche INRIA no 2829, 1996. 54 */ 55 56#ifndef USE_HOSTCC 57#include <log.h> 58#include <malloc.h> 59#include <ubi_uboot.h> 60#include <dm/devres.h> 61 62#include <linux/bitops.h> 63#include <linux/printk.h> 64#else 65#include <errno.h> 66#if defined(__FreeBSD__) 67#include <sys/endian.h> 68#elif defined(__APPLE__) 69#include <machine/endian.h> 70#include <libkern/OSByteOrder.h> 71#else 72#include <endian.h> 73#endif 74#include <stdint.h> 75#include <stdlib.h> 76#include <string.h> 77 78#undef cpu_to_be32 79#if defined(__APPLE__) 80#define cpu_to_be32 OSSwapHostToBigInt32 81#else 82#define cpu_to_be32 htobe32 83#endif 84#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d)) 85#define kmalloc(size, flags) malloc(size) 86#define kzalloc(size, flags) calloc(1, size) 87#define kfree free 88#define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0])) 89#endif 90 91#include <asm/byteorder.h> 92#include <linux/bch.h> 93 94#if defined(CONFIG_BCH_CONST_PARAMS) 95#define GF_M(_p) (CONFIG_BCH_CONST_M) 96#define GF_T(_p) (CONFIG_BCH_CONST_T) 97#define GF_N(_p) ((1 << (CONFIG_BCH_CONST_M))-1) 98#else 99#define GF_M(_p) ((_p)->m) 100#define GF_T(_p) ((_p)->t) 101#define GF_N(_p) ((_p)->n) 102#endif 103 104#define BCH_ECC_WORDS(_p) DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 32) 105#define BCH_ECC_BYTES(_p) DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 8) 106 107#ifndef dbg 108#define dbg(_fmt, args...) do {} while (0) 109#endif 110 111/* 112 * represent a polynomial over GF(2^m) 113 */ 114struct gf_poly { 115 unsigned int deg; /* polynomial degree */ 116 unsigned int c[0]; /* polynomial terms */ 117}; 118 119/* given its degree, compute a polynomial size in bytes */ 120#define GF_POLY_SZ(_d) (sizeof(struct gf_poly)+((_d)+1)*sizeof(unsigned int)) 121 122/* polynomial of degree 1 */ 123struct gf_poly_deg1 { 124 struct gf_poly poly; 125 unsigned int c[2]; 126}; 127 128#ifdef USE_HOSTCC 129#if !defined(__DragonFly__) && !defined(__FreeBSD__) && !defined(__APPLE__) 130static int fls(int x) 131{ 132 int r = 32; 133 134 if (!x) 135 return 0; 136 if (!(x & 0xffff0000u)) { 137 x <<= 16; 138 r -= 16; 139 } 140 if (!(x & 0xff000000u)) { 141 x <<= 8; 142 r -= 8; 143 } 144 if (!(x & 0xf0000000u)) { 145 x <<= 4; 146 r -= 4; 147 } 148 if (!(x & 0xc0000000u)) { 149 x <<= 2; 150 r -= 2; 151 } 152 if (!(x & 0x80000000u)) { 153 x <<= 1; 154 r -= 1; 155 } 156 return r; 157} 158#endif 159#endif 160 161/* 162 * same as encode_bch(), but process input data one byte at a time 163 */ 164static void encode_bch_unaligned(struct bch_control *bch, 165 const unsigned char *data, unsigned int len, 166 uint32_t *ecc) 167{ 168 int i; 169 const uint32_t *p; 170 const int l = BCH_ECC_WORDS(bch)-1; 171 172 while (len--) { 173 p = bch->mod8_tab + (l+1)*(((ecc[0] >> 24)^(*data++)) & 0xff); 174 175 for (i = 0; i < l; i++) 176 ecc[i] = ((ecc[i] << 8)|(ecc[i+1] >> 24))^(*p++); 177 178 ecc[l] = (ecc[l] << 8)^(*p); 179 } 180} 181 182/* 183 * convert ecc bytes to aligned, zero-padded 32-bit ecc words 184 */ 185static void load_ecc8(struct bch_control *bch, uint32_t *dst, 186 const uint8_t *src) 187{ 188 uint8_t pad[4] = {0, 0, 0, 0}; 189 unsigned int i, nwords = BCH_ECC_WORDS(bch)-1; 190 191 for (i = 0; i < nwords; i++, src += 4) 192 dst[i] = (src[0] << 24)|(src[1] << 16)|(src[2] << 8)|src[3]; 193 194 memcpy(pad, src, BCH_ECC_BYTES(bch)-4*nwords); 195 dst[nwords] = (pad[0] << 24)|(pad[1] << 16)|(pad[2] << 8)|pad[3]; 196} 197 198/* 199 * convert 32-bit ecc words to ecc bytes 200 */ 201static void store_ecc8(struct bch_control *bch, uint8_t *dst, 202 const uint32_t *src) 203{ 204 uint8_t pad[4]; 205 unsigned int i, nwords = BCH_ECC_WORDS(bch)-1; 206 207 for (i = 0; i < nwords; i++) { 208 *dst++ = (src[i] >> 24); 209 *dst++ = (src[i] >> 16) & 0xff; 210 *dst++ = (src[i] >> 8) & 0xff; 211 *dst++ = (src[i] >> 0) & 0xff; 212 } 213 pad[0] = (src[nwords] >> 24); 214 pad[1] = (src[nwords] >> 16) & 0xff; 215 pad[2] = (src[nwords] >> 8) & 0xff; 216 pad[3] = (src[nwords] >> 0) & 0xff; 217 memcpy(dst, pad, BCH_ECC_BYTES(bch)-4*nwords); 218} 219 220/** 221 * encode_bch - calculate BCH ecc parity of data 222 * @bch: BCH control structure 223 * @data: data to encode 224 * @len: data length in bytes 225 * @ecc: ecc parity data, must be initialized by caller 226 * 227 * The @ecc parity array is used both as input and output parameter, in order to 228 * allow incremental computations. It should be of the size indicated by member 229 * @ecc_bytes of @bch, and should be initialized to 0 before the first call. 230 * 231 * The exact number of computed ecc parity bits is given by member @ecc_bits of 232 * @bch; it may be less than m*t for large values of t. 233 */ 234void encode_bch(struct bch_control *bch, const uint8_t *data, 235 unsigned int len, uint8_t *ecc) 236{ 237 const unsigned int l = BCH_ECC_WORDS(bch)-1; 238 unsigned int i, mlen; 239 unsigned long m; 240 uint32_t w, r[l+1]; 241 const uint32_t * const tab0 = bch->mod8_tab; 242 const uint32_t * const tab1 = tab0 + 256*(l+1); 243 const uint32_t * const tab2 = tab1 + 256*(l+1); 244 const uint32_t * const tab3 = tab2 + 256*(l+1); 245 const uint32_t *pdata, *p0, *p1, *p2, *p3; 246 247 if (ecc) { 248 /* load ecc parity bytes into internal 32-bit buffer */ 249 load_ecc8(bch, bch->ecc_buf, ecc); 250 } else { 251 memset(bch->ecc_buf, 0, sizeof(r)); 252 } 253 254 /* process first unaligned data bytes */ 255 m = ((unsigned long)data) & 3; 256 if (m) { 257 mlen = (len < (4-m)) ? len : 4-m; 258 encode_bch_unaligned(bch, data, mlen, bch->ecc_buf); 259 data += mlen; 260 len -= mlen; 261 } 262 263 /* process 32-bit aligned data words */ 264 pdata = (uint32_t *)data; 265 mlen = len/4; 266 data += 4*mlen; 267 len -= 4*mlen; 268 memcpy(r, bch->ecc_buf, sizeof(r)); 269 270 /* 271 * split each 32-bit word into 4 polynomials of weight 8 as follows: 272 * 273 * 31 ...24 23 ...16 15 ... 8 7 ... 0 274 * xxxxxxxx yyyyyyyy zzzzzzzz tttttttt 275 * tttttttt mod g = r0 (precomputed) 276 * zzzzzzzz 00000000 mod g = r1 (precomputed) 277 * yyyyyyyy 00000000 00000000 mod g = r2 (precomputed) 278 * xxxxxxxx 00000000 00000000 00000000 mod g = r3 (precomputed) 279 * xxxxxxxx yyyyyyyy zzzzzzzz tttttttt mod g = r0^r1^r2^r3 280 */ 281 while (mlen--) { 282 /* input data is read in big-endian format */ 283 w = r[0]^cpu_to_be32(*pdata++); 284 p0 = tab0 + (l+1)*((w >> 0) & 0xff); 285 p1 = tab1 + (l+1)*((w >> 8) & 0xff); 286 p2 = tab2 + (l+1)*((w >> 16) & 0xff); 287 p3 = tab3 + (l+1)*((w >> 24) & 0xff); 288 289 for (i = 0; i < l; i++) 290 r[i] = r[i+1]^p0[i]^p1[i]^p2[i]^p3[i]; 291 292 r[l] = p0[l]^p1[l]^p2[l]^p3[l]; 293 } 294 memcpy(bch->ecc_buf, r, sizeof(r)); 295 296 /* process last unaligned bytes */ 297 if (len) 298 encode_bch_unaligned(bch, data, len, bch->ecc_buf); 299 300 /* store ecc parity bytes into original parity buffer */ 301 if (ecc) 302 store_ecc8(bch, ecc, bch->ecc_buf); 303} 304 305static inline int modulo(struct bch_control *bch, unsigned int v) 306{ 307 const unsigned int n = GF_N(bch); 308 while (v >= n) { 309 v -= n; 310 v = (v & n) + (v >> GF_M(bch)); 311 } 312 return v; 313} 314 315/* 316 * shorter and faster modulo function, only works when v < 2N. 317 */ 318static inline int mod_s(struct bch_control *bch, unsigned int v) 319{ 320 const unsigned int n = GF_N(bch); 321 return (v < n) ? v : v-n; 322} 323 324static inline int deg(unsigned int poly) 325{ 326 /* polynomial degree is the most-significant bit index */ 327 return fls(poly)-1; 328} 329 330static inline int parity(unsigned int x) 331{ 332 /* 333 * public domain code snippet, lifted from 334 * http://www-graphics.stanford.edu/~seander/bithacks.html 335 */ 336 x ^= x >> 1; 337 x ^= x >> 2; 338 x = (x & 0x11111111U) * 0x11111111U; 339 return (x >> 28) & 1; 340} 341 342/* Galois field basic operations: multiply, divide, inverse, etc. */ 343 344static inline unsigned int gf_mul(struct bch_control *bch, unsigned int a, 345 unsigned int b) 346{ 347 return (a && b) ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+ 348 bch->a_log_tab[b])] : 0; 349} 350 351static inline unsigned int gf_sqr(struct bch_control *bch, unsigned int a) 352{ 353 return a ? bch->a_pow_tab[mod_s(bch, 2*bch->a_log_tab[a])] : 0; 354} 355 356static inline unsigned int gf_div(struct bch_control *bch, unsigned int a, 357 unsigned int b) 358{ 359 return a ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+ 360 GF_N(bch)-bch->a_log_tab[b])] : 0; 361} 362 363static inline unsigned int gf_inv(struct bch_control *bch, unsigned int a) 364{ 365 return bch->a_pow_tab[GF_N(bch)-bch->a_log_tab[a]]; 366} 367 368static inline unsigned int a_pow(struct bch_control *bch, int i) 369{ 370 return bch->a_pow_tab[modulo(bch, i)]; 371} 372 373static inline int a_log(struct bch_control *bch, unsigned int x) 374{ 375 return bch->a_log_tab[x]; 376} 377 378static inline int a_ilog(struct bch_control *bch, unsigned int x) 379{ 380 return mod_s(bch, GF_N(bch)-bch->a_log_tab[x]); 381} 382 383/* 384 * compute 2t syndromes of ecc polynomial, i.e. ecc(a^j) for j=1..2t 385 */ 386static void compute_syndromes(struct bch_control *bch, uint32_t *ecc, 387 unsigned int *syn) 388{ 389 int i, j, s; 390 unsigned int m; 391 uint32_t poly; 392 const int t = GF_T(bch); 393 394 s = bch->ecc_bits; 395 396 /* make sure extra bits in last ecc word are cleared */ 397 m = ((unsigned int)s) & 31; 398 if (m) 399 ecc[s/32] &= ~((1u << (32-m))-1); 400 memset(syn, 0, 2*t*sizeof(*syn)); 401 402 /* compute v(a^j) for j=1 .. 2t-1 */ 403 do { 404 poly = *ecc++; 405 s -= 32; 406 while (poly) { 407 i = deg(poly); 408 for (j = 0; j < 2*t; j += 2) 409 syn[j] ^= a_pow(bch, (j+1)*(i+s)); 410 411 poly ^= (1 << i); 412 } 413 } while (s > 0); 414 415 /* v(a^(2j)) = v(a^j)^2 */ 416 for (j = 0; j < t; j++) 417 syn[2*j+1] = gf_sqr(bch, syn[j]); 418} 419 420static void gf_poly_copy(struct gf_poly *dst, struct gf_poly *src) 421{ 422 memcpy(dst, src, GF_POLY_SZ(src->deg)); 423} 424 425static int compute_error_locator_polynomial(struct bch_control *bch, 426 const unsigned int *syn) 427{ 428 const unsigned int t = GF_T(bch); 429 const unsigned int n = GF_N(bch); 430 unsigned int i, j, tmp, l, pd = 1, d = syn[0]; 431 struct gf_poly *elp = bch->elp; 432 struct gf_poly *pelp = bch->poly_2t[0]; 433 struct gf_poly *elp_copy = bch->poly_2t[1]; 434 int k, pp = -1; 435 436 memset(pelp, 0, GF_POLY_SZ(2*t)); 437 memset(elp, 0, GF_POLY_SZ(2*t)); 438 439 pelp->deg = 0; 440 pelp->c[0] = 1; 441 elp->deg = 0; 442 elp->c[0] = 1; 443 444 /* use simplified binary Berlekamp-Massey algorithm */ 445 for (i = 0; (i < t) && (elp->deg <= t); i++) { 446 if (d) { 447 k = 2*i-pp; 448 gf_poly_copy(elp_copy, elp); 449 /* e[i+1](X) = e[i](X)+di*dp^-1*X^2(i-p)*e[p](X) */ 450 tmp = a_log(bch, d)+n-a_log(bch, pd); 451 for (j = 0; j <= pelp->deg; j++) { 452 if (pelp->c[j]) { 453 l = a_log(bch, pelp->c[j]); 454 elp->c[j+k] ^= a_pow(bch, tmp+l); 455 } 456 } 457 /* compute l[i+1] = max(l[i]->c[l[p]+2*(i-p]) */ 458 tmp = pelp->deg+k; 459 if (tmp > elp->deg) { 460 elp->deg = tmp; 461 gf_poly_copy(pelp, elp_copy); 462 pd = d; 463 pp = 2*i; 464 } 465 } 466 /* di+1 = S(2i+3)+elp[i+1].1*S(2i+2)+...+elp[i+1].lS(2i+3-l) */ 467 if (i < t-1) { 468 d = syn[2*i+2]; 469 for (j = 1; j <= elp->deg; j++) 470 d ^= gf_mul(bch, elp->c[j], syn[2*i+2-j]); 471 } 472 } 473 dbg("elp=%s\n", gf_poly_str(elp)); 474 return (elp->deg > t) ? -1 : (int)elp->deg; 475} 476 477/* 478 * solve a m x m linear system in GF(2) with an expected number of solutions, 479 * and return the number of found solutions 480 */ 481static int solve_linear_system(struct bch_control *bch, unsigned int *rows, 482 unsigned int *sol, int nsol) 483{ 484 const int m = GF_M(bch); 485 unsigned int tmp, mask; 486 int rem, c, r, p, k, param[m]; 487 488 k = 0; 489 mask = 1 << m; 490 491 /* Gaussian elimination */ 492 for (c = 0; c < m; c++) { 493 rem = 0; 494 p = c-k; 495 /* find suitable row for elimination */ 496 for (r = p; r < m; r++) { 497 if (rows[r] & mask) { 498 if (r != p) { 499 tmp = rows[r]; 500 rows[r] = rows[p]; 501 rows[p] = tmp; 502 } 503 rem = r+1; 504 break; 505 } 506 } 507 if (rem) { 508 /* perform elimination on remaining rows */ 509 tmp = rows[p]; 510 for (r = rem; r < m; r++) { 511 if (rows[r] & mask) 512 rows[r] ^= tmp; 513 } 514 } else { 515 /* elimination not needed, store defective row index */ 516 param[k++] = c; 517 } 518 mask >>= 1; 519 } 520 /* rewrite system, inserting fake parameter rows */ 521 if (k > 0) { 522 p = k; 523 for (r = m-1; r >= 0; r--) { 524 if ((r > m-1-k) && rows[r]) 525 /* system has no solution */ 526 return 0; 527 528 rows[r] = (p && (r == param[p-1])) ? 529 p--, 1u << (m-r) : rows[r-p]; 530 } 531 } 532 533 if (nsol != (1 << k)) 534 /* unexpected number of solutions */ 535 return 0; 536 537 for (p = 0; p < nsol; p++) { 538 /* set parameters for p-th solution */ 539 for (c = 0; c < k; c++) 540 rows[param[c]] = (rows[param[c]] & ~1)|((p >> c) & 1); 541 542 /* compute unique solution */ 543 tmp = 0; 544 for (r = m-1; r >= 0; r--) { 545 mask = rows[r] & (tmp|1); 546 tmp |= parity(mask) << (m-r); 547 } 548 sol[p] = tmp >> 1; 549 } 550 return nsol; 551} 552 553/* 554 * this function builds and solves a linear system for finding roots of a degree 555 * 4 affine monic polynomial X^4+aX^2+bX+c over GF(2^m). 556 */ 557static int find_affine4_roots(struct bch_control *bch, unsigned int a, 558 unsigned int b, unsigned int c, 559 unsigned int *roots) 560{ 561 int i, j, k; 562 const int m = GF_M(bch); 563 unsigned int mask = 0xff, t, rows[16] = {0,}; 564 565 j = a_log(bch, b); 566 k = a_log(bch, a); 567 rows[0] = c; 568 569 /* buid linear system to solve X^4+aX^2+bX+c = 0 */ 570 for (i = 0; i < m; i++) { 571 rows[i+1] = bch->a_pow_tab[4*i]^ 572 (a ? bch->a_pow_tab[mod_s(bch, k)] : 0)^ 573 (b ? bch->a_pow_tab[mod_s(bch, j)] : 0); 574 j++; 575 k += 2; 576 } 577 /* 578 * transpose 16x16 matrix before passing it to linear solver 579 * warning: this code assumes m < 16 580 */ 581 for (j = 8; j != 0; j >>= 1, mask ^= (mask << j)) { 582 for (k = 0; k < 16; k = (k+j+1) & ~j) { 583 t = ((rows[k] >> j)^rows[k+j]) & mask; 584 rows[k] ^= (t << j); 585 rows[k+j] ^= t; 586 } 587 } 588 return solve_linear_system(bch, rows, roots, 4); 589} 590 591/* 592 * compute root r of a degree 1 polynomial over GF(2^m) (returned as log(1/r)) 593 */ 594static int find_poly_deg1_roots(struct bch_control *bch, struct gf_poly *poly, 595 unsigned int *roots) 596{ 597 int n = 0; 598 599 if (poly->c[0]) 600 /* poly[X] = bX+c with c!=0, root=c/b */ 601 roots[n++] = mod_s(bch, GF_N(bch)-bch->a_log_tab[poly->c[0]]+ 602 bch->a_log_tab[poly->c[1]]); 603 return n; 604} 605 606/* 607 * compute roots of a degree 2 polynomial over GF(2^m) 608 */ 609static int find_poly_deg2_roots(struct bch_control *bch, struct gf_poly *poly, 610 unsigned int *roots) 611{ 612 int n = 0, i, l0, l1, l2; 613 unsigned int u, v, r; 614 615 if (poly->c[0] && poly->c[1]) { 616 617 l0 = bch->a_log_tab[poly->c[0]]; 618 l1 = bch->a_log_tab[poly->c[1]]; 619 l2 = bch->a_log_tab[poly->c[2]]; 620 621 /* using z=a/bX, transform aX^2+bX+c into z^2+z+u (u=ac/b^2) */ 622 u = a_pow(bch, l0+l2+2*(GF_N(bch)-l1)); 623 /* 624 * let u = sum(li.a^i) i=0..m-1; then compute r = sum(li.xi): 625 * r^2+r = sum(li.(xi^2+xi)) = sum(li.(a^i+Tr(a^i).a^k)) = 626 * u + sum(li.Tr(a^i).a^k) = u+a^k.Tr(sum(li.a^i)) = u+a^k.Tr(u) 627 * i.e. r and r+1 are roots iff Tr(u)=0 628 */ 629 r = 0; 630 v = u; 631 while (v) { 632 i = deg(v); 633 r ^= bch->xi_tab[i]; 634 v ^= (1 << i); 635 } 636 /* verify root */ 637 if ((gf_sqr(bch, r)^r) == u) { 638 /* reverse z=a/bX transformation and compute log(1/r) */ 639 roots[n++] = modulo(bch, 2*GF_N(bch)-l1- 640 bch->a_log_tab[r]+l2); 641 roots[n++] = modulo(bch, 2*GF_N(bch)-l1- 642 bch->a_log_tab[r^1]+l2); 643 } 644 } 645 return n; 646} 647 648/* 649 * compute roots of a degree 3 polynomial over GF(2^m) 650 */ 651static int find_poly_deg3_roots(struct bch_control *bch, struct gf_poly *poly, 652 unsigned int *roots) 653{ 654 int i, n = 0; 655 unsigned int a, b, c, a2, b2, c2, e3, tmp[4]; 656 657 if (poly->c[0]) { 658 /* transform polynomial into monic X^3 + a2X^2 + b2X + c2 */ 659 e3 = poly->c[3]; 660 c2 = gf_div(bch, poly->c[0], e3); 661 b2 = gf_div(bch, poly->c[1], e3); 662 a2 = gf_div(bch, poly->c[2], e3); 663 664 /* (X+a2)(X^3+a2X^2+b2X+c2) = X^4+aX^2+bX+c (affine) */ 665 c = gf_mul(bch, a2, c2); /* c = a2c2 */ 666 b = gf_mul(bch, a2, b2)^c2; /* b = a2b2 + c2 */ 667 a = gf_sqr(bch, a2)^b2; /* a = a2^2 + b2 */ 668 669 /* find the 4 roots of this affine polynomial */ 670 if (find_affine4_roots(bch, a, b, c, tmp) == 4) { 671 /* remove a2 from final list of roots */ 672 for (i = 0; i < 4; i++) { 673 if (tmp[i] != a2) 674 roots[n++] = a_ilog(bch, tmp[i]); 675 } 676 } 677 } 678 return n; 679} 680 681/* 682 * compute roots of a degree 4 polynomial over GF(2^m) 683 */ 684static int find_poly_deg4_roots(struct bch_control *bch, struct gf_poly *poly, 685 unsigned int *roots) 686{ 687 int i, l, n = 0; 688 unsigned int a, b, c, d, e = 0, f, a2, b2, c2, e4; 689 690 if (poly->c[0] == 0) 691 return 0; 692 693 /* transform polynomial into monic X^4 + aX^3 + bX^2 + cX + d */ 694 e4 = poly->c[4]; 695 d = gf_div(bch, poly->c[0], e4); 696 c = gf_div(bch, poly->c[1], e4); 697 b = gf_div(bch, poly->c[2], e4); 698 a = gf_div(bch, poly->c[3], e4); 699 700 /* use Y=1/X transformation to get an affine polynomial */ 701 if (a) { 702 /* first, eliminate cX by using z=X+e with ae^2+c=0 */ 703 if (c) { 704 /* compute e such that e^2 = c/a */ 705 f = gf_div(bch, c, a); 706 l = a_log(bch, f); 707 l += (l & 1) ? GF_N(bch) : 0; 708 e = a_pow(bch, l/2); 709 /* 710 * use transformation z=X+e: 711 * z^4+e^4 + a(z^3+ez^2+e^2z+e^3) + b(z^2+e^2) +cz+ce+d 712 * z^4 + az^3 + (ae+b)z^2 + (ae^2+c)z+e^4+be^2+ae^3+ce+d 713 * z^4 + az^3 + (ae+b)z^2 + e^4+be^2+d 714 * z^4 + az^3 + b'z^2 + d' 715 */ 716 d = a_pow(bch, 2*l)^gf_mul(bch, b, f)^d; 717 b = gf_mul(bch, a, e)^b; 718 } 719 /* now, use Y=1/X to get Y^4 + b/dY^2 + a/dY + 1/d */ 720 if (d == 0) 721 /* assume all roots have multiplicity 1 */ 722 return 0; 723 724 c2 = gf_inv(bch, d); 725 b2 = gf_div(bch, a, d); 726 a2 = gf_div(bch, b, d); 727 } else { 728 /* polynomial is already affine */ 729 c2 = d; 730 b2 = c; 731 a2 = b; 732 } 733 /* find the 4 roots of this affine polynomial */ 734 if (find_affine4_roots(bch, a2, b2, c2, roots) == 4) { 735 for (i = 0; i < 4; i++) { 736 /* post-process roots (reverse transformations) */ 737 f = a ? gf_inv(bch, roots[i]) : roots[i]; 738 roots[i] = a_ilog(bch, f^e); 739 } 740 n = 4; 741 } 742 return n; 743} 744 745/* 746 * build monic, log-based representation of a polynomial 747 */ 748static void gf_poly_logrep(struct bch_control *bch, 749 const struct gf_poly *a, int *rep) 750{ 751 int i, d = a->deg, l = GF_N(bch)-a_log(bch, a->c[a->deg]); 752 753 /* represent 0 values with -1; warning, rep[d] is not set to 1 */ 754 for (i = 0; i < d; i++) 755 rep[i] = a->c[i] ? mod_s(bch, a_log(bch, a->c[i])+l) : -1; 756} 757 758/* 759 * compute polynomial Euclidean division remainder in GF(2^m)[X] 760 */ 761static void gf_poly_mod(struct bch_control *bch, struct gf_poly *a, 762 const struct gf_poly *b, int *rep) 763{ 764 int la, p, m; 765 unsigned int i, j, *c = a->c; 766 const unsigned int d = b->deg; 767 768 if (a->deg < d) 769 return; 770 771 /* reuse or compute log representation of denominator */ 772 if (!rep) { 773 rep = bch->cache; 774 gf_poly_logrep(bch, b, rep); 775 } 776 777 for (j = a->deg; j >= d; j--) { 778 if (c[j]) { 779 la = a_log(bch, c[j]); 780 p = j-d; 781 for (i = 0; i < d; i++, p++) { 782 m = rep[i]; 783 if (m >= 0) 784 c[p] ^= bch->a_pow_tab[mod_s(bch, 785 m+la)]; 786 } 787 } 788 } 789 a->deg = d-1; 790 while (!c[a->deg] && a->deg) 791 a->deg--; 792} 793 794/* 795 * compute polynomial Euclidean division quotient in GF(2^m)[X] 796 */ 797static void gf_poly_div(struct bch_control *bch, struct gf_poly *a, 798 const struct gf_poly *b, struct gf_poly *q) 799{ 800 if (a->deg >= b->deg) { 801 q->deg = a->deg-b->deg; 802 /* compute a mod b (modifies a) */ 803 gf_poly_mod(bch, a, b, NULL); 804 /* quotient is stored in upper part of polynomial a */ 805 memcpy(q->c, &a->c[b->deg], (1+q->deg)*sizeof(unsigned int)); 806 } else { 807 q->deg = 0; 808 q->c[0] = 0; 809 } 810} 811 812/* 813 * compute polynomial GCD (Greatest Common Divisor) in GF(2^m)[X] 814 */ 815static struct gf_poly *gf_poly_gcd(struct bch_control *bch, struct gf_poly *a, 816 struct gf_poly *b) 817{ 818 struct gf_poly *tmp; 819 820 dbg("gcd(%s,%s)=", gf_poly_str(a), gf_poly_str(b)); 821 822 if (a->deg < b->deg) { 823 tmp = b; 824 b = a; 825 a = tmp; 826 } 827 828 while (b->deg > 0) { 829 gf_poly_mod(bch, a, b, NULL); 830 tmp = b; 831 b = a; 832 a = tmp; 833 } 834 835 dbg("%s\n", gf_poly_str(a)); 836 837 return a; 838} 839 840/* 841 * Given a polynomial f and an integer k, compute Tr(a^kX) mod f 842 * This is used in Berlekamp Trace algorithm for splitting polynomials 843 */ 844static void compute_trace_bk_mod(struct bch_control *bch, int k, 845 const struct gf_poly *f, struct gf_poly *z, 846 struct gf_poly *out) 847{ 848 const int m = GF_M(bch); 849 int i, j; 850 851 /* z contains z^2j mod f */ 852 z->deg = 1; 853 z->c[0] = 0; 854 z->c[1] = bch->a_pow_tab[k]; 855 856 out->deg = 0; 857 memset(out, 0, GF_POLY_SZ(f->deg)); 858 859 /* compute f log representation only once */ 860 gf_poly_logrep(bch, f, bch->cache); 861 862 for (i = 0; i < m; i++) { 863 /* add a^(k*2^i)(z^(2^i) mod f) and compute (z^(2^i) mod f)^2 */ 864 for (j = z->deg; j >= 0; j--) { 865 out->c[j] ^= z->c[j]; 866 z->c[2*j] = gf_sqr(bch, z->c[j]); 867 z->c[2*j+1] = 0; 868 } 869 if (z->deg > out->deg) 870 out->deg = z->deg; 871 872 if (i < m-1) { 873 z->deg *= 2; 874 /* z^(2(i+1)) mod f = (z^(2^i) mod f)^2 mod f */ 875 gf_poly_mod(bch, z, f, bch->cache); 876 } 877 } 878 while (!out->c[out->deg] && out->deg) 879 out->deg--; 880 881 dbg("Tr(a^%d.X) mod f = %s\n", k, gf_poly_str(out)); 882} 883 884/* 885 * factor a polynomial using Berlekamp Trace algorithm (BTA) 886 */ 887static void factor_polynomial(struct bch_control *bch, int k, struct gf_poly *f, 888 struct gf_poly **g, struct gf_poly **h) 889{ 890 struct gf_poly *f2 = bch->poly_2t[0]; 891 struct gf_poly *q = bch->poly_2t[1]; 892 struct gf_poly *tk = bch->poly_2t[2]; 893 struct gf_poly *z = bch->poly_2t[3]; 894 struct gf_poly *gcd; 895 896 dbg("factoring %s...\n", gf_poly_str(f)); 897 898 *g = f; 899 *h = NULL; 900 901 /* tk = Tr(a^k.X) mod f */ 902 compute_trace_bk_mod(bch, k, f, z, tk); 903 904 if (tk->deg > 0) { 905 /* compute g = gcd(f, tk) (destructive operation) */ 906 gf_poly_copy(f2, f); 907 gcd = gf_poly_gcd(bch, f2, tk); 908 if (gcd->deg < f->deg) { 909 /* compute h=f/gcd(f,tk); this will modify f and q */ 910 gf_poly_div(bch, f, gcd, q); 911 /* store g and h in-place (clobbering f) */ 912 *h = &((struct gf_poly_deg1 *)f)[gcd->deg].poly; 913 gf_poly_copy(*g, gcd); 914 gf_poly_copy(*h, q); 915 } 916 } 917} 918 919/* 920 * find roots of a polynomial, using BTZ algorithm; see the beginning of this 921 * file for details 922 */ 923static int find_poly_roots(struct bch_control *bch, unsigned int k, 924 struct gf_poly *poly, unsigned int *roots) 925{ 926 int cnt; 927 struct gf_poly *f1, *f2; 928 929 switch (poly->deg) { 930 /* handle low degree polynomials with ad hoc techniques */ 931 case 1: 932 cnt = find_poly_deg1_roots(bch, poly, roots); 933 break; 934 case 2: 935 cnt = find_poly_deg2_roots(bch, poly, roots); 936 break; 937 case 3: 938 cnt = find_poly_deg3_roots(bch, poly, roots); 939 break; 940 case 4: 941 cnt = find_poly_deg4_roots(bch, poly, roots); 942 break; 943 default: 944 /* factor polynomial using Berlekamp Trace Algorithm (BTA) */ 945 cnt = 0; 946 if (poly->deg && (k <= GF_M(bch))) { 947 factor_polynomial(bch, k, poly, &f1, &f2); 948 if (f1) 949 cnt += find_poly_roots(bch, k+1, f1, roots); 950 if (f2) 951 cnt += find_poly_roots(bch, k+1, f2, roots+cnt); 952 } 953 break; 954 } 955 return cnt; 956} 957 958#if defined(USE_CHIEN_SEARCH) 959/* 960 * exhaustive root search (Chien) implementation - not used, included only for 961 * reference/comparison tests 962 */ 963static int chien_search(struct bch_control *bch, unsigned int len, 964 struct gf_poly *p, unsigned int *roots) 965{ 966 int m; 967 unsigned int i, j, syn, syn0, count = 0; 968 const unsigned int k = 8*len+bch->ecc_bits; 969 970 /* use a log-based representation of polynomial */ 971 gf_poly_logrep(bch, p, bch->cache); 972 bch->cache[p->deg] = 0; 973 syn0 = gf_div(bch, p->c[0], p->c[p->deg]); 974 975 for (i = GF_N(bch)-k+1; i <= GF_N(bch); i++) { 976 /* compute elp(a^i) */ 977 for (j = 1, syn = syn0; j <= p->deg; j++) { 978 m = bch->cache[j]; 979 if (m >= 0) 980 syn ^= a_pow(bch, m+j*i); 981 } 982 if (syn == 0) { 983 roots[count++] = GF_N(bch)-i; 984 if (count == p->deg) 985 break; 986 } 987 } 988 return (count == p->deg) ? count : 0; 989} 990#define find_poly_roots(_p, _k, _elp, _loc) chien_search(_p, len, _elp, _loc) 991#endif /* USE_CHIEN_SEARCH */ 992 993/** 994 * decode_bch - decode received codeword and find bit error locations 995 * @bch: BCH control structure 996 * @data: received data, ignored if @calc_ecc is provided 997 * @len: data length in bytes, must always be provided 998 * @recv_ecc: received ecc, if NULL then assume it was XORed in @calc_ecc 999 * @calc_ecc: calculated ecc, if NULL then calc_ecc is computed from @data 1000 * @syn: hw computed syndrome data (if NULL, syndrome is calculated) 1001 * @errloc: output array of error locations 1002 * 1003 * Returns: 1004 * The number of errors found, or -EBADMSG if decoding failed, or -EINVAL if 1005 * invalid parameters were provided 1006 * 1007 * Depending on the available hw BCH support and the need to compute @calc_ecc 1008 * separately (using encode_bch()), this function should be called with one of 1009 * the following parameter configurations - 1010 * 1011 * by providing @data and @recv_ecc only: 1012 * decode_bch(@bch, @data, @len, @recv_ecc, NULL, NULL, @errloc) 1013 * 1014 * by providing @recv_ecc and @calc_ecc: 1015 * decode_bch(@bch, NULL, @len, @recv_ecc, @calc_ecc, NULL, @errloc) 1016 * 1017 * by providing ecc = recv_ecc XOR calc_ecc: 1018 * decode_bch(@bch, NULL, @len, NULL, ecc, NULL, @errloc) 1019 * 1020 * by providing syndrome results @syn: 1021 * decode_bch(@bch, NULL, @len, NULL, NULL, @syn, @errloc) 1022 * 1023 * Once decode_bch() has successfully returned with a positive value, error 1024 * locations returned in array @errloc should be interpreted as follows - 1025 * 1026 * if (errloc[n] >= 8*len), then n-th error is located in ecc (no need for 1027 * data correction) 1028 * 1029 * if (errloc[n] < 8*len), then n-th error is located in data and can be 1030 * corrected with statement data[errloc[n]/8] ^= 1 << (errloc[n] % 8); 1031 * 1032 * Note that this function does not perform any data correction by itself, it 1033 * merely indicates error locations. 1034 */ 1035int decode_bch(struct bch_control *bch, const uint8_t *data, unsigned int len, 1036 const uint8_t *recv_ecc, const uint8_t *calc_ecc, 1037 const unsigned int *syn, unsigned int *errloc) 1038{ 1039 const unsigned int ecc_words = BCH_ECC_WORDS(bch); 1040 unsigned int nbits; 1041 int i, err, nroots; 1042 uint32_t sum; 1043 1044 /* sanity check: make sure data length can be handled */ 1045 if (8*len > (bch->n-bch->ecc_bits)) 1046 return -EINVAL; 1047 1048 /* if caller does not provide syndromes, compute them */ 1049 if (!syn) { 1050 if (!calc_ecc) { 1051 /* compute received data ecc into an internal buffer */ 1052 if (!data || !recv_ecc) 1053 return -EINVAL; 1054 encode_bch(bch, data, len, NULL); 1055 } else { 1056 /* load provided calculated ecc */ 1057 load_ecc8(bch, bch->ecc_buf, calc_ecc); 1058 } 1059 /* load received ecc or assume it was XORed in calc_ecc */ 1060 if (recv_ecc) { 1061 load_ecc8(bch, bch->ecc_buf2, recv_ecc); 1062 /* XOR received and calculated ecc */ 1063 for (i = 0, sum = 0; i < (int)ecc_words; i++) { 1064 bch->ecc_buf[i] ^= bch->ecc_buf2[i]; 1065 sum |= bch->ecc_buf[i]; 1066 } 1067 if (!sum) 1068 /* no error found */ 1069 return 0; 1070 } 1071 compute_syndromes(bch, bch->ecc_buf, bch->syn); 1072 syn = bch->syn; 1073 } 1074 1075 err = compute_error_locator_polynomial(bch, syn); 1076 if (err > 0) { 1077 nroots = find_poly_roots(bch, 1, bch->elp, errloc); 1078 if (err != nroots) 1079 err = -1; 1080 } 1081 if (err > 0) { 1082 /* post-process raw error locations for easier correction */ 1083 nbits = (len*8)+bch->ecc_bits; 1084 for (i = 0; i < err; i++) { 1085 if (errloc[i] >= nbits) { 1086 err = -1; 1087 break; 1088 } 1089 errloc[i] = nbits-1-errloc[i]; 1090 errloc[i] = (errloc[i] & ~7)|(7-(errloc[i] & 7)); 1091 } 1092 } 1093 return (err >= 0) ? err : -EBADMSG; 1094} 1095 1096/* 1097 * generate Galois field lookup tables 1098 */ 1099static int build_gf_tables(struct bch_control *bch, unsigned int poly) 1100{ 1101 unsigned int i, x = 1; 1102 const unsigned int k = 1 << deg(poly); 1103 1104 /* primitive polynomial must be of degree m */ 1105 if (k != (1u << GF_M(bch))) 1106 return -1; 1107 1108 for (i = 0; i < GF_N(bch); i++) { 1109 bch->a_pow_tab[i] = x; 1110 bch->a_log_tab[x] = i; 1111 if (i && (x == 1)) 1112 /* polynomial is not primitive (a^i=1 with 0<i<2^m-1) */ 1113 return -1; 1114 x <<= 1; 1115 if (x & k) 1116 x ^= poly; 1117 } 1118 bch->a_pow_tab[GF_N(bch)] = 1; 1119 bch->a_log_tab[0] = 0; 1120 1121 return 0; 1122} 1123 1124/* 1125 * compute generator polynomial remainder tables for fast encoding 1126 */ 1127static void build_mod8_tables(struct bch_control *bch, const uint32_t *g) 1128{ 1129 int i, j, b, d; 1130 uint32_t data, hi, lo, *tab; 1131 const int l = BCH_ECC_WORDS(bch); 1132 const int plen = DIV_ROUND_UP(bch->ecc_bits+1, 32); 1133 const int ecclen = DIV_ROUND_UP(bch->ecc_bits, 32); 1134 1135 memset(bch->mod8_tab, 0, 4*256*l*sizeof(*bch->mod8_tab)); 1136 1137 for (i = 0; i < 256; i++) { 1138 /* p(X)=i is a small polynomial of weight <= 8 */ 1139 for (b = 0; b < 4; b++) { 1140 /* we want to compute (p(X).X^(8*b+deg(g))) mod g(X) */ 1141 tab = bch->mod8_tab + (b*256+i)*l; 1142 data = i << (8*b); 1143 while (data) { 1144 d = deg(data); 1145 /* subtract X^d.g(X) from p(X).X^(8*b+deg(g)) */ 1146 data ^= g[0] >> (31-d); 1147 for (j = 0; j < ecclen; j++) { 1148 hi = (d < 31) ? g[j] << (d+1) : 0; 1149 lo = (j+1 < plen) ? 1150 g[j+1] >> (31-d) : 0; 1151 tab[j] ^= hi|lo; 1152 } 1153 } 1154 } 1155 } 1156} 1157 1158/* 1159 * build a base for factoring degree 2 polynomials 1160 */ 1161static int build_deg2_base(struct bch_control *bch) 1162{ 1163 const int m = GF_M(bch); 1164 int i, j, r; 1165 unsigned int sum, x, y, remaining, ak = 0, xi[m]; 1166 1167 /* find k s.t. Tr(a^k) = 1 and 0 <= k < m */ 1168 for (i = 0; i < m; i++) { 1169 for (j = 0, sum = 0; j < m; j++) 1170 sum ^= a_pow(bch, i*(1 << j)); 1171 1172 if (sum) { 1173 ak = bch->a_pow_tab[i]; 1174 break; 1175 } 1176 } 1177 /* find xi, i=0..m-1 such that xi^2+xi = a^i+Tr(a^i).a^k */ 1178 remaining = m; 1179 memset(xi, 0, sizeof(xi)); 1180 1181 for (x = 0; (x <= GF_N(bch)) && remaining; x++) { 1182 y = gf_sqr(bch, x)^x; 1183 for (i = 0; i < 2; i++) { 1184 r = a_log(bch, y); 1185 if (y && (r < m) && !xi[r]) { 1186 bch->xi_tab[r] = x; 1187 xi[r] = 1; 1188 remaining--; 1189 dbg("x%d = %x\n", r, x); 1190 break; 1191 } 1192 y ^= ak; 1193 } 1194 } 1195 /* should not happen but check anyway */ 1196 return remaining ? -1 : 0; 1197} 1198 1199static void *bch_alloc(size_t size, int *err) 1200{ 1201 void *ptr; 1202 1203 ptr = kmalloc(size, GFP_KERNEL); 1204 if (ptr == NULL) 1205 *err = 1; 1206 return ptr; 1207} 1208 1209/* 1210 * compute generator polynomial for given (m,t) parameters. 1211 */ 1212static uint32_t *compute_generator_polynomial(struct bch_control *bch) 1213{ 1214 const unsigned int m = GF_M(bch); 1215 const unsigned int t = GF_T(bch); 1216 int n, err = 0; 1217 unsigned int i, j, nbits, r, word, *roots; 1218 struct gf_poly *g; 1219 uint32_t *genpoly; 1220 1221 g = bch_alloc(GF_POLY_SZ(m*t), &err); 1222 roots = bch_alloc((bch->n+1)*sizeof(*roots), &err); 1223 genpoly = bch_alloc(DIV_ROUND_UP(m*t+1, 32)*sizeof(*genpoly), &err); 1224 1225 if (err) { 1226 kfree(genpoly); 1227 genpoly = NULL; 1228 goto finish; 1229 } 1230 1231 /* enumerate all roots of g(X) */ 1232 memset(roots , 0, (bch->n+1)*sizeof(*roots)); 1233 for (i = 0; i < t; i++) { 1234 for (j = 0, r = 2*i+1; j < m; j++) { 1235 roots[r] = 1; 1236 r = mod_s(bch, 2*r); 1237 } 1238 } 1239 /* build generator polynomial g(X) */ 1240 g->deg = 0; 1241 g->c[0] = 1; 1242 for (i = 0; i < GF_N(bch); i++) { 1243 if (roots[i]) { 1244 /* multiply g(X) by (X+root) */ 1245 r = bch->a_pow_tab[i]; 1246 g->c[g->deg+1] = 1; 1247 for (j = g->deg; j > 0; j--) 1248 g->c[j] = gf_mul(bch, g->c[j], r)^g->c[j-1]; 1249 1250 g->c[0] = gf_mul(bch, g->c[0], r); 1251 g->deg++; 1252 } 1253 } 1254 /* store left-justified binary representation of g(X) */ 1255 n = g->deg+1; 1256 i = 0; 1257 1258 while (n > 0) { 1259 nbits = (n > 32) ? 32 : n; 1260 for (j = 0, word = 0; j < nbits; j++) { 1261 if (g->c[n-1-j]) 1262 word |= 1u << (31-j); 1263 } 1264 genpoly[i++] = word; 1265 n -= nbits; 1266 } 1267 bch->ecc_bits = g->deg; 1268 1269finish: 1270 kfree(g); 1271 kfree(roots); 1272 1273 return genpoly; 1274} 1275 1276/** 1277 * init_bch - initialize a BCH encoder/decoder 1278 * @m: Galois field order, should be in the range 5-15 1279 * @t: maximum error correction capability, in bits 1280 * @prim_poly: user-provided primitive polynomial (or 0 to use default) 1281 * 1282 * Returns: 1283 * a newly allocated BCH control structure if successful, NULL otherwise 1284 * 1285 * This initialization can take some time, as lookup tables are built for fast 1286 * encoding/decoding; make sure not to call this function from a time critical 1287 * path. Usually, init_bch() should be called on module/driver init and 1288 * free_bch() should be called to release memory on exit. 1289 * 1290 * You may provide your own primitive polynomial of degree @m in argument 1291 * @prim_poly, or let init_bch() use its default polynomial. 1292 * 1293 * Once init_bch() has successfully returned a pointer to a newly allocated 1294 * BCH control structure, ecc length in bytes is given by member @ecc_bytes of 1295 * the structure. 1296 */ 1297struct bch_control *init_bch(int m, int t, unsigned int prim_poly) 1298{ 1299 int err = 0; 1300 unsigned int i, words; 1301 uint32_t *genpoly; 1302 struct bch_control *bch = NULL; 1303 1304 const int min_m = 5; 1305 const int max_m = 15; 1306 1307 /* default primitive polynomials */ 1308 static const unsigned int prim_poly_tab[] = { 1309 0x25, 0x43, 0x83, 0x11d, 0x211, 0x409, 0x805, 0x1053, 0x201b, 1310 0x402b, 0x8003, 1311 }; 1312 1313#if defined(CONFIG_BCH_CONST_PARAMS) 1314 if ((m != (CONFIG_BCH_CONST_M)) || (t != (CONFIG_BCH_CONST_T))) { 1315 printk(KERN_ERR "bch encoder/decoder was configured to support " 1316 "parameters m=%d, t=%d only!\n", 1317 CONFIG_BCH_CONST_M, CONFIG_BCH_CONST_T); 1318 goto fail; 1319 } 1320#endif 1321 if ((m < min_m) || (m > max_m)) 1322 /* 1323 * values of m greater than 15 are not currently supported; 1324 * supporting m > 15 would require changing table base type 1325 * (uint16_t) and a small patch in matrix transposition 1326 */ 1327 goto fail; 1328 1329 /* sanity checks */ 1330 if ((t < 1) || (m*t >= ((1 << m)-1))) 1331 /* invalid t value */ 1332 goto fail; 1333 1334 /* select a primitive polynomial for generating GF(2^m) */ 1335 if (prim_poly == 0) 1336 prim_poly = prim_poly_tab[m-min_m]; 1337 1338 bch = kzalloc(sizeof(*bch), GFP_KERNEL); 1339 if (bch == NULL) 1340 goto fail; 1341 1342 bch->m = m; 1343 bch->t = t; 1344 bch->n = (1 << m)-1; 1345 words = DIV_ROUND_UP(m*t, 32); 1346 bch->ecc_bytes = DIV_ROUND_UP(m*t, 8); 1347 bch->a_pow_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_pow_tab), &err); 1348 bch->a_log_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_log_tab), &err); 1349 bch->mod8_tab = bch_alloc(words*1024*sizeof(*bch->mod8_tab), &err); 1350 bch->ecc_buf = bch_alloc(words*sizeof(*bch->ecc_buf), &err); 1351 bch->ecc_buf2 = bch_alloc(words*sizeof(*bch->ecc_buf2), &err); 1352 bch->xi_tab = bch_alloc(m*sizeof(*bch->xi_tab), &err); 1353 bch->syn = bch_alloc(2*t*sizeof(*bch->syn), &err); 1354 bch->cache = bch_alloc(2*t*sizeof(*bch->cache), &err); 1355 bch->elp = bch_alloc((t+1)*sizeof(struct gf_poly_deg1), &err); 1356 1357 for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++) 1358 bch->poly_2t[i] = bch_alloc(GF_POLY_SZ(2*t), &err); 1359 1360 if (err) 1361 goto fail; 1362 1363 err = build_gf_tables(bch, prim_poly); 1364 if (err) 1365 goto fail; 1366 1367 /* use generator polynomial for computing encoding tables */ 1368 genpoly = compute_generator_polynomial(bch); 1369 if (genpoly == NULL) 1370 goto fail; 1371 1372 build_mod8_tables(bch, genpoly); 1373 kfree(genpoly); 1374 1375 err = build_deg2_base(bch); 1376 if (err) 1377 goto fail; 1378 1379 return bch; 1380 1381fail: 1382 free_bch(bch); 1383 return NULL; 1384} 1385 1386/** 1387 * free_bch - free the BCH control structure 1388 * @bch: BCH control structure to release 1389 */ 1390void free_bch(struct bch_control *bch) 1391{ 1392 unsigned int i; 1393 1394 if (bch) { 1395 kfree(bch->a_pow_tab); 1396 kfree(bch->a_log_tab); 1397 kfree(bch->mod8_tab); 1398 kfree(bch->ecc_buf); 1399 kfree(bch->ecc_buf2); 1400 kfree(bch->xi_tab); 1401 kfree(bch->syn); 1402 kfree(bch->cache); 1403 kfree(bch->elp); 1404 1405 for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++) 1406 kfree(bch->poly_2t[i]); 1407 1408 kfree(bch); 1409 } 1410} 1411