xref: /openbmc/u-boot/lib/bch.c (revision 5044c9cc)
1 /*
2  * Generic binary BCH encoding/decoding library
3  *
4  * SPDX-License-Identifier:	GPL-2.0
5  *
6  * Copyright © 2011 Parrot S.A.
7  *
8  * Author: Ivan Djelic <ivan.djelic@parrot.com>
9  *
10  * Description:
11  *
12  * This library provides runtime configurable encoding/decoding of binary
13  * Bose-Chaudhuri-Hocquenghem (BCH) codes.
14  *
15  * Call init_bch to get a pointer to a newly allocated bch_control structure for
16  * the given m (Galois field order), t (error correction capability) and
17  * (optional) primitive polynomial parameters.
18  *
19  * Call encode_bch to compute and store ecc parity bytes to a given buffer.
20  * Call decode_bch to detect and locate errors in received data.
21  *
22  * On systems supporting hw BCH features, intermediate results may be provided
23  * to decode_bch in order to skip certain steps. See decode_bch() documentation
24  * for details.
25  *
26  * Option CONFIG_BCH_CONST_PARAMS can be used to force fixed values of
27  * parameters m and t; thus allowing extra compiler optimizations and providing
28  * better (up to 2x) encoding performance. Using this option makes sense when
29  * (m,t) are fixed and known in advance, e.g. when using BCH error correction
30  * on a particular NAND flash device.
31  *
32  * Algorithmic details:
33  *
34  * Encoding is performed by processing 32 input bits in parallel, using 4
35  * remainder lookup tables.
36  *
37  * The final stage of decoding involves the following internal steps:
38  * a. Syndrome computation
39  * b. Error locator polynomial computation using Berlekamp-Massey algorithm
40  * c. Error locator root finding (by far the most expensive step)
41  *
42  * In this implementation, step c is not performed using the usual Chien search.
43  * Instead, an alternative approach described in [1] is used. It consists in
44  * factoring the error locator polynomial using the Berlekamp Trace algorithm
45  * (BTA) down to a certain degree (4), after which ad hoc low-degree polynomial
46  * solving techniques [2] are used. The resulting algorithm, called BTZ, yields
47  * much better performance than Chien search for usual (m,t) values (typically
48  * m >= 13, t < 32, see [1]).
49  *
50  * [1] B. Biswas, V. Herbert. Efficient root finding of polynomials over fields
51  * of characteristic 2, in: Western European Workshop on Research in Cryptology
52  * - WEWoRC 2009, Graz, Austria, LNCS, Springer, July 2009, to appear.
53  * [2] [Zin96] V.A. Zinoviev. On the solution of equations of degree 10 over
54  * finite fields GF(2^q). In Rapport de recherche INRIA no 2829, 1996.
55  */
56 
57 #include <common.h>
58 #include <ubi_uboot.h>
59 
60 #include <linux/bitops.h>
61 #include <asm/byteorder.h>
62 #include <linux/bch.h>
63 
64 #if defined(CONFIG_BCH_CONST_PARAMS)
65 #define GF_M(_p)               (CONFIG_BCH_CONST_M)
66 #define GF_T(_p)               (CONFIG_BCH_CONST_T)
67 #define GF_N(_p)               ((1 << (CONFIG_BCH_CONST_M))-1)
68 #else
69 #define GF_M(_p)               ((_p)->m)
70 #define GF_T(_p)               ((_p)->t)
71 #define GF_N(_p)               ((_p)->n)
72 #endif
73 
74 #define BCH_ECC_WORDS(_p)      DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 32)
75 #define BCH_ECC_BYTES(_p)      DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 8)
76 
77 #ifndef dbg
78 #define dbg(_fmt, args...)     do {} while (0)
79 #endif
80 
81 /*
82  * represent a polynomial over GF(2^m)
83  */
84 struct gf_poly {
85 	unsigned int deg;    /* polynomial degree */
86 	unsigned int c[0];   /* polynomial terms */
87 };
88 
89 /* given its degree, compute a polynomial size in bytes */
90 #define GF_POLY_SZ(_d) (sizeof(struct gf_poly)+((_d)+1)*sizeof(unsigned int))
91 
92 /* polynomial of degree 1 */
93 struct gf_poly_deg1 {
94 	struct gf_poly poly;
95 	unsigned int   c[2];
96 };
97 
98 /*
99  * same as encode_bch(), but process input data one byte at a time
100  */
101 static void encode_bch_unaligned(struct bch_control *bch,
102 				 const unsigned char *data, unsigned int len,
103 				 uint32_t *ecc)
104 {
105 	int i;
106 	const uint32_t *p;
107 	const int l = BCH_ECC_WORDS(bch)-1;
108 
109 	while (len--) {
110 		p = bch->mod8_tab + (l+1)*(((ecc[0] >> 24)^(*data++)) & 0xff);
111 
112 		for (i = 0; i < l; i++)
113 			ecc[i] = ((ecc[i] << 8)|(ecc[i+1] >> 24))^(*p++);
114 
115 		ecc[l] = (ecc[l] << 8)^(*p);
116 	}
117 }
118 
119 /*
120  * convert ecc bytes to aligned, zero-padded 32-bit ecc words
121  */
122 static void load_ecc8(struct bch_control *bch, uint32_t *dst,
123 		      const uint8_t *src)
124 {
125 	uint8_t pad[4] = {0, 0, 0, 0};
126 	unsigned int i, nwords = BCH_ECC_WORDS(bch)-1;
127 
128 	for (i = 0; i < nwords; i++, src += 4)
129 		dst[i] = (src[0] << 24)|(src[1] << 16)|(src[2] << 8)|src[3];
130 
131 	memcpy(pad, src, BCH_ECC_BYTES(bch)-4*nwords);
132 	dst[nwords] = (pad[0] << 24)|(pad[1] << 16)|(pad[2] << 8)|pad[3];
133 }
134 
135 /*
136  * convert 32-bit ecc words to ecc bytes
137  */
138 static void store_ecc8(struct bch_control *bch, uint8_t *dst,
139 		       const uint32_t *src)
140 {
141 	uint8_t pad[4];
142 	unsigned int i, nwords = BCH_ECC_WORDS(bch)-1;
143 
144 	for (i = 0; i < nwords; i++) {
145 		*dst++ = (src[i] >> 24);
146 		*dst++ = (src[i] >> 16) & 0xff;
147 		*dst++ = (src[i] >>  8) & 0xff;
148 		*dst++ = (src[i] >>  0) & 0xff;
149 	}
150 	pad[0] = (src[nwords] >> 24);
151 	pad[1] = (src[nwords] >> 16) & 0xff;
152 	pad[2] = (src[nwords] >>  8) & 0xff;
153 	pad[3] = (src[nwords] >>  0) & 0xff;
154 	memcpy(dst, pad, BCH_ECC_BYTES(bch)-4*nwords);
155 }
156 
157 /**
158  * encode_bch - calculate BCH ecc parity of data
159  * @bch:   BCH control structure
160  * @data:  data to encode
161  * @len:   data length in bytes
162  * @ecc:   ecc parity data, must be initialized by caller
163  *
164  * The @ecc parity array is used both as input and output parameter, in order to
165  * allow incremental computations. It should be of the size indicated by member
166  * @ecc_bytes of @bch, and should be initialized to 0 before the first call.
167  *
168  * The exact number of computed ecc parity bits is given by member @ecc_bits of
169  * @bch; it may be less than m*t for large values of t.
170  */
171 void encode_bch(struct bch_control *bch, const uint8_t *data,
172 		unsigned int len, uint8_t *ecc)
173 {
174 	const unsigned int l = BCH_ECC_WORDS(bch)-1;
175 	unsigned int i, mlen;
176 	unsigned long m;
177 	uint32_t w, r[l+1];
178 	const uint32_t * const tab0 = bch->mod8_tab;
179 	const uint32_t * const tab1 = tab0 + 256*(l+1);
180 	const uint32_t * const tab2 = tab1 + 256*(l+1);
181 	const uint32_t * const tab3 = tab2 + 256*(l+1);
182 	const uint32_t *pdata, *p0, *p1, *p2, *p3;
183 
184 	if (ecc) {
185 		/* load ecc parity bytes into internal 32-bit buffer */
186 		load_ecc8(bch, bch->ecc_buf, ecc);
187 	} else {
188 		memset(bch->ecc_buf, 0, sizeof(r));
189 	}
190 
191 	/* process first unaligned data bytes */
192 	m = ((unsigned long)data) & 3;
193 	if (m) {
194 		mlen = (len < (4-m)) ? len : 4-m;
195 		encode_bch_unaligned(bch, data, mlen, bch->ecc_buf);
196 		data += mlen;
197 		len  -= mlen;
198 	}
199 
200 	/* process 32-bit aligned data words */
201 	pdata = (uint32_t *)data;
202 	mlen  = len/4;
203 	data += 4*mlen;
204 	len  -= 4*mlen;
205 	memcpy(r, bch->ecc_buf, sizeof(r));
206 
207 	/*
208 	 * split each 32-bit word into 4 polynomials of weight 8 as follows:
209 	 *
210 	 * 31 ...24  23 ...16  15 ... 8  7 ... 0
211 	 * xxxxxxxx  yyyyyyyy  zzzzzzzz  tttttttt
212 	 *                               tttttttt  mod g = r0 (precomputed)
213 	 *                     zzzzzzzz  00000000  mod g = r1 (precomputed)
214 	 *           yyyyyyyy  00000000  00000000  mod g = r2 (precomputed)
215 	 * xxxxxxxx  00000000  00000000  00000000  mod g = r3 (precomputed)
216 	 * xxxxxxxx  yyyyyyyy  zzzzzzzz  tttttttt  mod g = r0^r1^r2^r3
217 	 */
218 	while (mlen--) {
219 		/* input data is read in big-endian format */
220 		w = r[0]^cpu_to_be32(*pdata++);
221 		p0 = tab0 + (l+1)*((w >>  0) & 0xff);
222 		p1 = tab1 + (l+1)*((w >>  8) & 0xff);
223 		p2 = tab2 + (l+1)*((w >> 16) & 0xff);
224 		p3 = tab3 + (l+1)*((w >> 24) & 0xff);
225 
226 		for (i = 0; i < l; i++)
227 			r[i] = r[i+1]^p0[i]^p1[i]^p2[i]^p3[i];
228 
229 		r[l] = p0[l]^p1[l]^p2[l]^p3[l];
230 	}
231 	memcpy(bch->ecc_buf, r, sizeof(r));
232 
233 	/* process last unaligned bytes */
234 	if (len)
235 		encode_bch_unaligned(bch, data, len, bch->ecc_buf);
236 
237 	/* store ecc parity bytes into original parity buffer */
238 	if (ecc)
239 		store_ecc8(bch, ecc, bch->ecc_buf);
240 }
241 
242 static inline int modulo(struct bch_control *bch, unsigned int v)
243 {
244 	const unsigned int n = GF_N(bch);
245 	while (v >= n) {
246 		v -= n;
247 		v = (v & n) + (v >> GF_M(bch));
248 	}
249 	return v;
250 }
251 
252 /*
253  * shorter and faster modulo function, only works when v < 2N.
254  */
255 static inline int mod_s(struct bch_control *bch, unsigned int v)
256 {
257 	const unsigned int n = GF_N(bch);
258 	return (v < n) ? v : v-n;
259 }
260 
261 static inline int deg(unsigned int poly)
262 {
263 	/* polynomial degree is the most-significant bit index */
264 	return fls(poly)-1;
265 }
266 
267 static inline int parity(unsigned int x)
268 {
269 	/*
270 	 * public domain code snippet, lifted from
271 	 * http://www-graphics.stanford.edu/~seander/bithacks.html
272 	 */
273 	x ^= x >> 1;
274 	x ^= x >> 2;
275 	x = (x & 0x11111111U) * 0x11111111U;
276 	return (x >> 28) & 1;
277 }
278 
279 /* Galois field basic operations: multiply, divide, inverse, etc. */
280 
281 static inline unsigned int gf_mul(struct bch_control *bch, unsigned int a,
282 				  unsigned int b)
283 {
284 	return (a && b) ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+
285 					       bch->a_log_tab[b])] : 0;
286 }
287 
288 static inline unsigned int gf_sqr(struct bch_control *bch, unsigned int a)
289 {
290 	return a ? bch->a_pow_tab[mod_s(bch, 2*bch->a_log_tab[a])] : 0;
291 }
292 
293 static inline unsigned int gf_div(struct bch_control *bch, unsigned int a,
294 				  unsigned int b)
295 {
296 	return a ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+
297 					GF_N(bch)-bch->a_log_tab[b])] : 0;
298 }
299 
300 static inline unsigned int gf_inv(struct bch_control *bch, unsigned int a)
301 {
302 	return bch->a_pow_tab[GF_N(bch)-bch->a_log_tab[a]];
303 }
304 
305 static inline unsigned int a_pow(struct bch_control *bch, int i)
306 {
307 	return bch->a_pow_tab[modulo(bch, i)];
308 }
309 
310 static inline int a_log(struct bch_control *bch, unsigned int x)
311 {
312 	return bch->a_log_tab[x];
313 }
314 
315 static inline int a_ilog(struct bch_control *bch, unsigned int x)
316 {
317 	return mod_s(bch, GF_N(bch)-bch->a_log_tab[x]);
318 }
319 
320 /*
321  * compute 2t syndromes of ecc polynomial, i.e. ecc(a^j) for j=1..2t
322  */
323 static void compute_syndromes(struct bch_control *bch, uint32_t *ecc,
324 			      unsigned int *syn)
325 {
326 	int i, j, s;
327 	unsigned int m;
328 	uint32_t poly;
329 	const int t = GF_T(bch);
330 
331 	s = bch->ecc_bits;
332 
333 	/* make sure extra bits in last ecc word are cleared */
334 	m = ((unsigned int)s) & 31;
335 	if (m)
336 		ecc[s/32] &= ~((1u << (32-m))-1);
337 	memset(syn, 0, 2*t*sizeof(*syn));
338 
339 	/* compute v(a^j) for j=1 .. 2t-1 */
340 	do {
341 		poly = *ecc++;
342 		s -= 32;
343 		while (poly) {
344 			i = deg(poly);
345 			for (j = 0; j < 2*t; j += 2)
346 				syn[j] ^= a_pow(bch, (j+1)*(i+s));
347 
348 			poly ^= (1 << i);
349 		}
350 	} while (s > 0);
351 
352 	/* v(a^(2j)) = v(a^j)^2 */
353 	for (j = 0; j < t; j++)
354 		syn[2*j+1] = gf_sqr(bch, syn[j]);
355 }
356 
357 static void gf_poly_copy(struct gf_poly *dst, struct gf_poly *src)
358 {
359 	memcpy(dst, src, GF_POLY_SZ(src->deg));
360 }
361 
362 static int compute_error_locator_polynomial(struct bch_control *bch,
363 					    const unsigned int *syn)
364 {
365 	const unsigned int t = GF_T(bch);
366 	const unsigned int n = GF_N(bch);
367 	unsigned int i, j, tmp, l, pd = 1, d = syn[0];
368 	struct gf_poly *elp = bch->elp;
369 	struct gf_poly *pelp = bch->poly_2t[0];
370 	struct gf_poly *elp_copy = bch->poly_2t[1];
371 	int k, pp = -1;
372 
373 	memset(pelp, 0, GF_POLY_SZ(2*t));
374 	memset(elp, 0, GF_POLY_SZ(2*t));
375 
376 	pelp->deg = 0;
377 	pelp->c[0] = 1;
378 	elp->deg = 0;
379 	elp->c[0] = 1;
380 
381 	/* use simplified binary Berlekamp-Massey algorithm */
382 	for (i = 0; (i < t) && (elp->deg <= t); i++) {
383 		if (d) {
384 			k = 2*i-pp;
385 			gf_poly_copy(elp_copy, elp);
386 			/* e[i+1](X) = e[i](X)+di*dp^-1*X^2(i-p)*e[p](X) */
387 			tmp = a_log(bch, d)+n-a_log(bch, pd);
388 			for (j = 0; j <= pelp->deg; j++) {
389 				if (pelp->c[j]) {
390 					l = a_log(bch, pelp->c[j]);
391 					elp->c[j+k] ^= a_pow(bch, tmp+l);
392 				}
393 			}
394 			/* compute l[i+1] = max(l[i]->c[l[p]+2*(i-p]) */
395 			tmp = pelp->deg+k;
396 			if (tmp > elp->deg) {
397 				elp->deg = tmp;
398 				gf_poly_copy(pelp, elp_copy);
399 				pd = d;
400 				pp = 2*i;
401 			}
402 		}
403 		/* di+1 = S(2i+3)+elp[i+1].1*S(2i+2)+...+elp[i+1].lS(2i+3-l) */
404 		if (i < t-1) {
405 			d = syn[2*i+2];
406 			for (j = 1; j <= elp->deg; j++)
407 				d ^= gf_mul(bch, elp->c[j], syn[2*i+2-j]);
408 		}
409 	}
410 	dbg("elp=%s\n", gf_poly_str(elp));
411 	return (elp->deg > t) ? -1 : (int)elp->deg;
412 }
413 
414 /*
415  * solve a m x m linear system in GF(2) with an expected number of solutions,
416  * and return the number of found solutions
417  */
418 static int solve_linear_system(struct bch_control *bch, unsigned int *rows,
419 			       unsigned int *sol, int nsol)
420 {
421 	const int m = GF_M(bch);
422 	unsigned int tmp, mask;
423 	int rem, c, r, p, k, param[m];
424 
425 	k = 0;
426 	mask = 1 << m;
427 
428 	/* Gaussian elimination */
429 	for (c = 0; c < m; c++) {
430 		rem = 0;
431 		p = c-k;
432 		/* find suitable row for elimination */
433 		for (r = p; r < m; r++) {
434 			if (rows[r] & mask) {
435 				if (r != p) {
436 					tmp = rows[r];
437 					rows[r] = rows[p];
438 					rows[p] = tmp;
439 				}
440 				rem = r+1;
441 				break;
442 			}
443 		}
444 		if (rem) {
445 			/* perform elimination on remaining rows */
446 			tmp = rows[p];
447 			for (r = rem; r < m; r++) {
448 				if (rows[r] & mask)
449 					rows[r] ^= tmp;
450 			}
451 		} else {
452 			/* elimination not needed, store defective row index */
453 			param[k++] = c;
454 		}
455 		mask >>= 1;
456 	}
457 	/* rewrite system, inserting fake parameter rows */
458 	if (k > 0) {
459 		p = k;
460 		for (r = m-1; r >= 0; r--) {
461 			if ((r > m-1-k) && rows[r])
462 				/* system has no solution */
463 				return 0;
464 
465 			rows[r] = (p && (r == param[p-1])) ?
466 				p--, 1u << (m-r) : rows[r-p];
467 		}
468 	}
469 
470 	if (nsol != (1 << k))
471 		/* unexpected number of solutions */
472 		return 0;
473 
474 	for (p = 0; p < nsol; p++) {
475 		/* set parameters for p-th solution */
476 		for (c = 0; c < k; c++)
477 			rows[param[c]] = (rows[param[c]] & ~1)|((p >> c) & 1);
478 
479 		/* compute unique solution */
480 		tmp = 0;
481 		for (r = m-1; r >= 0; r--) {
482 			mask = rows[r] & (tmp|1);
483 			tmp |= parity(mask) << (m-r);
484 		}
485 		sol[p] = tmp >> 1;
486 	}
487 	return nsol;
488 }
489 
490 /*
491  * this function builds and solves a linear system for finding roots of a degree
492  * 4 affine monic polynomial X^4+aX^2+bX+c over GF(2^m).
493  */
494 static int find_affine4_roots(struct bch_control *bch, unsigned int a,
495 			      unsigned int b, unsigned int c,
496 			      unsigned int *roots)
497 {
498 	int i, j, k;
499 	const int m = GF_M(bch);
500 	unsigned int mask = 0xff, t, rows[16] = {0,};
501 
502 	j = a_log(bch, b);
503 	k = a_log(bch, a);
504 	rows[0] = c;
505 
506 	/* buid linear system to solve X^4+aX^2+bX+c = 0 */
507 	for (i = 0; i < m; i++) {
508 		rows[i+1] = bch->a_pow_tab[4*i]^
509 			(a ? bch->a_pow_tab[mod_s(bch, k)] : 0)^
510 			(b ? bch->a_pow_tab[mod_s(bch, j)] : 0);
511 		j++;
512 		k += 2;
513 	}
514 	/*
515 	 * transpose 16x16 matrix before passing it to linear solver
516 	 * warning: this code assumes m < 16
517 	 */
518 	for (j = 8; j != 0; j >>= 1, mask ^= (mask << j)) {
519 		for (k = 0; k < 16; k = (k+j+1) & ~j) {
520 			t = ((rows[k] >> j)^rows[k+j]) & mask;
521 			rows[k] ^= (t << j);
522 			rows[k+j] ^= t;
523 		}
524 	}
525 	return solve_linear_system(bch, rows, roots, 4);
526 }
527 
528 /*
529  * compute root r of a degree 1 polynomial over GF(2^m) (returned as log(1/r))
530  */
531 static int find_poly_deg1_roots(struct bch_control *bch, struct gf_poly *poly,
532 				unsigned int *roots)
533 {
534 	int n = 0;
535 
536 	if (poly->c[0])
537 		/* poly[X] = bX+c with c!=0, root=c/b */
538 		roots[n++] = mod_s(bch, GF_N(bch)-bch->a_log_tab[poly->c[0]]+
539 				   bch->a_log_tab[poly->c[1]]);
540 	return n;
541 }
542 
543 /*
544  * compute roots of a degree 2 polynomial over GF(2^m)
545  */
546 static int find_poly_deg2_roots(struct bch_control *bch, struct gf_poly *poly,
547 				unsigned int *roots)
548 {
549 	int n = 0, i, l0, l1, l2;
550 	unsigned int u, v, r;
551 
552 	if (poly->c[0] && poly->c[1]) {
553 
554 		l0 = bch->a_log_tab[poly->c[0]];
555 		l1 = bch->a_log_tab[poly->c[1]];
556 		l2 = bch->a_log_tab[poly->c[2]];
557 
558 		/* using z=a/bX, transform aX^2+bX+c into z^2+z+u (u=ac/b^2) */
559 		u = a_pow(bch, l0+l2+2*(GF_N(bch)-l1));
560 		/*
561 		 * let u = sum(li.a^i) i=0..m-1; then compute r = sum(li.xi):
562 		 * r^2+r = sum(li.(xi^2+xi)) = sum(li.(a^i+Tr(a^i).a^k)) =
563 		 * u + sum(li.Tr(a^i).a^k) = u+a^k.Tr(sum(li.a^i)) = u+a^k.Tr(u)
564 		 * i.e. r and r+1 are roots iff Tr(u)=0
565 		 */
566 		r = 0;
567 		v = u;
568 		while (v) {
569 			i = deg(v);
570 			r ^= bch->xi_tab[i];
571 			v ^= (1 << i);
572 		}
573 		/* verify root */
574 		if ((gf_sqr(bch, r)^r) == u) {
575 			/* reverse z=a/bX transformation and compute log(1/r) */
576 			roots[n++] = modulo(bch, 2*GF_N(bch)-l1-
577 					    bch->a_log_tab[r]+l2);
578 			roots[n++] = modulo(bch, 2*GF_N(bch)-l1-
579 					    bch->a_log_tab[r^1]+l2);
580 		}
581 	}
582 	return n;
583 }
584 
585 /*
586  * compute roots of a degree 3 polynomial over GF(2^m)
587  */
588 static int find_poly_deg3_roots(struct bch_control *bch, struct gf_poly *poly,
589 				unsigned int *roots)
590 {
591 	int i, n = 0;
592 	unsigned int a, b, c, a2, b2, c2, e3, tmp[4];
593 
594 	if (poly->c[0]) {
595 		/* transform polynomial into monic X^3 + a2X^2 + b2X + c2 */
596 		e3 = poly->c[3];
597 		c2 = gf_div(bch, poly->c[0], e3);
598 		b2 = gf_div(bch, poly->c[1], e3);
599 		a2 = gf_div(bch, poly->c[2], e3);
600 
601 		/* (X+a2)(X^3+a2X^2+b2X+c2) = X^4+aX^2+bX+c (affine) */
602 		c = gf_mul(bch, a2, c2);           /* c = a2c2      */
603 		b = gf_mul(bch, a2, b2)^c2;        /* b = a2b2 + c2 */
604 		a = gf_sqr(bch, a2)^b2;            /* a = a2^2 + b2 */
605 
606 		/* find the 4 roots of this affine polynomial */
607 		if (find_affine4_roots(bch, a, b, c, tmp) == 4) {
608 			/* remove a2 from final list of roots */
609 			for (i = 0; i < 4; i++) {
610 				if (tmp[i] != a2)
611 					roots[n++] = a_ilog(bch, tmp[i]);
612 			}
613 		}
614 	}
615 	return n;
616 }
617 
618 /*
619  * compute roots of a degree 4 polynomial over GF(2^m)
620  */
621 static int find_poly_deg4_roots(struct bch_control *bch, struct gf_poly *poly,
622 				unsigned int *roots)
623 {
624 	int i, l, n = 0;
625 	unsigned int a, b, c, d, e = 0, f, a2, b2, c2, e4;
626 
627 	if (poly->c[0] == 0)
628 		return 0;
629 
630 	/* transform polynomial into monic X^4 + aX^3 + bX^2 + cX + d */
631 	e4 = poly->c[4];
632 	d = gf_div(bch, poly->c[0], e4);
633 	c = gf_div(bch, poly->c[1], e4);
634 	b = gf_div(bch, poly->c[2], e4);
635 	a = gf_div(bch, poly->c[3], e4);
636 
637 	/* use Y=1/X transformation to get an affine polynomial */
638 	if (a) {
639 		/* first, eliminate cX by using z=X+e with ae^2+c=0 */
640 		if (c) {
641 			/* compute e such that e^2 = c/a */
642 			f = gf_div(bch, c, a);
643 			l = a_log(bch, f);
644 			l += (l & 1) ? GF_N(bch) : 0;
645 			e = a_pow(bch, l/2);
646 			/*
647 			 * use transformation z=X+e:
648 			 * z^4+e^4 + a(z^3+ez^2+e^2z+e^3) + b(z^2+e^2) +cz+ce+d
649 			 * z^4 + az^3 + (ae+b)z^2 + (ae^2+c)z+e^4+be^2+ae^3+ce+d
650 			 * z^4 + az^3 + (ae+b)z^2 + e^4+be^2+d
651 			 * z^4 + az^3 +     b'z^2 + d'
652 			 */
653 			d = a_pow(bch, 2*l)^gf_mul(bch, b, f)^d;
654 			b = gf_mul(bch, a, e)^b;
655 		}
656 		/* now, use Y=1/X to get Y^4 + b/dY^2 + a/dY + 1/d */
657 		if (d == 0)
658 			/* assume all roots have multiplicity 1 */
659 			return 0;
660 
661 		c2 = gf_inv(bch, d);
662 		b2 = gf_div(bch, a, d);
663 		a2 = gf_div(bch, b, d);
664 	} else {
665 		/* polynomial is already affine */
666 		c2 = d;
667 		b2 = c;
668 		a2 = b;
669 	}
670 	/* find the 4 roots of this affine polynomial */
671 	if (find_affine4_roots(bch, a2, b2, c2, roots) == 4) {
672 		for (i = 0; i < 4; i++) {
673 			/* post-process roots (reverse transformations) */
674 			f = a ? gf_inv(bch, roots[i]) : roots[i];
675 			roots[i] = a_ilog(bch, f^e);
676 		}
677 		n = 4;
678 	}
679 	return n;
680 }
681 
682 /*
683  * build monic, log-based representation of a polynomial
684  */
685 static void gf_poly_logrep(struct bch_control *bch,
686 			   const struct gf_poly *a, int *rep)
687 {
688 	int i, d = a->deg, l = GF_N(bch)-a_log(bch, a->c[a->deg]);
689 
690 	/* represent 0 values with -1; warning, rep[d] is not set to 1 */
691 	for (i = 0; i < d; i++)
692 		rep[i] = a->c[i] ? mod_s(bch, a_log(bch, a->c[i])+l) : -1;
693 }
694 
695 /*
696  * compute polynomial Euclidean division remainder in GF(2^m)[X]
697  */
698 static void gf_poly_mod(struct bch_control *bch, struct gf_poly *a,
699 			const struct gf_poly *b, int *rep)
700 {
701 	int la, p, m;
702 	unsigned int i, j, *c = a->c;
703 	const unsigned int d = b->deg;
704 
705 	if (a->deg < d)
706 		return;
707 
708 	/* reuse or compute log representation of denominator */
709 	if (!rep) {
710 		rep = bch->cache;
711 		gf_poly_logrep(bch, b, rep);
712 	}
713 
714 	for (j = a->deg; j >= d; j--) {
715 		if (c[j]) {
716 			la = a_log(bch, c[j]);
717 			p = j-d;
718 			for (i = 0; i < d; i++, p++) {
719 				m = rep[i];
720 				if (m >= 0)
721 					c[p] ^= bch->a_pow_tab[mod_s(bch,
722 								     m+la)];
723 			}
724 		}
725 	}
726 	a->deg = d-1;
727 	while (!c[a->deg] && a->deg)
728 		a->deg--;
729 }
730 
731 /*
732  * compute polynomial Euclidean division quotient in GF(2^m)[X]
733  */
734 static void gf_poly_div(struct bch_control *bch, struct gf_poly *a,
735 			const struct gf_poly *b, struct gf_poly *q)
736 {
737 	if (a->deg >= b->deg) {
738 		q->deg = a->deg-b->deg;
739 		/* compute a mod b (modifies a) */
740 		gf_poly_mod(bch, a, b, NULL);
741 		/* quotient is stored in upper part of polynomial a */
742 		memcpy(q->c, &a->c[b->deg], (1+q->deg)*sizeof(unsigned int));
743 	} else {
744 		q->deg = 0;
745 		q->c[0] = 0;
746 	}
747 }
748 
749 /*
750  * compute polynomial GCD (Greatest Common Divisor) in GF(2^m)[X]
751  */
752 static struct gf_poly *gf_poly_gcd(struct bch_control *bch, struct gf_poly *a,
753 				   struct gf_poly *b)
754 {
755 	struct gf_poly *tmp;
756 
757 	dbg("gcd(%s,%s)=", gf_poly_str(a), gf_poly_str(b));
758 
759 	if (a->deg < b->deg) {
760 		tmp = b;
761 		b = a;
762 		a = tmp;
763 	}
764 
765 	while (b->deg > 0) {
766 		gf_poly_mod(bch, a, b, NULL);
767 		tmp = b;
768 		b = a;
769 		a = tmp;
770 	}
771 
772 	dbg("%s\n", gf_poly_str(a));
773 
774 	return a;
775 }
776 
777 /*
778  * Given a polynomial f and an integer k, compute Tr(a^kX) mod f
779  * This is used in Berlekamp Trace algorithm for splitting polynomials
780  */
781 static void compute_trace_bk_mod(struct bch_control *bch, int k,
782 				 const struct gf_poly *f, struct gf_poly *z,
783 				 struct gf_poly *out)
784 {
785 	const int m = GF_M(bch);
786 	int i, j;
787 
788 	/* z contains z^2j mod f */
789 	z->deg = 1;
790 	z->c[0] = 0;
791 	z->c[1] = bch->a_pow_tab[k];
792 
793 	out->deg = 0;
794 	memset(out, 0, GF_POLY_SZ(f->deg));
795 
796 	/* compute f log representation only once */
797 	gf_poly_logrep(bch, f, bch->cache);
798 
799 	for (i = 0; i < m; i++) {
800 		/* add a^(k*2^i)(z^(2^i) mod f) and compute (z^(2^i) mod f)^2 */
801 		for (j = z->deg; j >= 0; j--) {
802 			out->c[j] ^= z->c[j];
803 			z->c[2*j] = gf_sqr(bch, z->c[j]);
804 			z->c[2*j+1] = 0;
805 		}
806 		if (z->deg > out->deg)
807 			out->deg = z->deg;
808 
809 		if (i < m-1) {
810 			z->deg *= 2;
811 			/* z^(2(i+1)) mod f = (z^(2^i) mod f)^2 mod f */
812 			gf_poly_mod(bch, z, f, bch->cache);
813 		}
814 	}
815 	while (!out->c[out->deg] && out->deg)
816 		out->deg--;
817 
818 	dbg("Tr(a^%d.X) mod f = %s\n", k, gf_poly_str(out));
819 }
820 
821 /*
822  * factor a polynomial using Berlekamp Trace algorithm (BTA)
823  */
824 static void factor_polynomial(struct bch_control *bch, int k, struct gf_poly *f,
825 			      struct gf_poly **g, struct gf_poly **h)
826 {
827 	struct gf_poly *f2 = bch->poly_2t[0];
828 	struct gf_poly *q  = bch->poly_2t[1];
829 	struct gf_poly *tk = bch->poly_2t[2];
830 	struct gf_poly *z  = bch->poly_2t[3];
831 	struct gf_poly *gcd;
832 
833 	dbg("factoring %s...\n", gf_poly_str(f));
834 
835 	*g = f;
836 	*h = NULL;
837 
838 	/* tk = Tr(a^k.X) mod f */
839 	compute_trace_bk_mod(bch, k, f, z, tk);
840 
841 	if (tk->deg > 0) {
842 		/* compute g = gcd(f, tk) (destructive operation) */
843 		gf_poly_copy(f2, f);
844 		gcd = gf_poly_gcd(bch, f2, tk);
845 		if (gcd->deg < f->deg) {
846 			/* compute h=f/gcd(f,tk); this will modify f and q */
847 			gf_poly_div(bch, f, gcd, q);
848 			/* store g and h in-place (clobbering f) */
849 			*h = &((struct gf_poly_deg1 *)f)[gcd->deg].poly;
850 			gf_poly_copy(*g, gcd);
851 			gf_poly_copy(*h, q);
852 		}
853 	}
854 }
855 
856 /*
857  * find roots of a polynomial, using BTZ algorithm; see the beginning of this
858  * file for details
859  */
860 static int find_poly_roots(struct bch_control *bch, unsigned int k,
861 			   struct gf_poly *poly, unsigned int *roots)
862 {
863 	int cnt;
864 	struct gf_poly *f1, *f2;
865 
866 	switch (poly->deg) {
867 		/* handle low degree polynomials with ad hoc techniques */
868 	case 1:
869 		cnt = find_poly_deg1_roots(bch, poly, roots);
870 		break;
871 	case 2:
872 		cnt = find_poly_deg2_roots(bch, poly, roots);
873 		break;
874 	case 3:
875 		cnt = find_poly_deg3_roots(bch, poly, roots);
876 		break;
877 	case 4:
878 		cnt = find_poly_deg4_roots(bch, poly, roots);
879 		break;
880 	default:
881 		/* factor polynomial using Berlekamp Trace Algorithm (BTA) */
882 		cnt = 0;
883 		if (poly->deg && (k <= GF_M(bch))) {
884 			factor_polynomial(bch, k, poly, &f1, &f2);
885 			if (f1)
886 				cnt += find_poly_roots(bch, k+1, f1, roots);
887 			if (f2)
888 				cnt += find_poly_roots(bch, k+1, f2, roots+cnt);
889 		}
890 		break;
891 	}
892 	return cnt;
893 }
894 
895 #if defined(USE_CHIEN_SEARCH)
896 /*
897  * exhaustive root search (Chien) implementation - not used, included only for
898  * reference/comparison tests
899  */
900 static int chien_search(struct bch_control *bch, unsigned int len,
901 			struct gf_poly *p, unsigned int *roots)
902 {
903 	int m;
904 	unsigned int i, j, syn, syn0, count = 0;
905 	const unsigned int k = 8*len+bch->ecc_bits;
906 
907 	/* use a log-based representation of polynomial */
908 	gf_poly_logrep(bch, p, bch->cache);
909 	bch->cache[p->deg] = 0;
910 	syn0 = gf_div(bch, p->c[0], p->c[p->deg]);
911 
912 	for (i = GF_N(bch)-k+1; i <= GF_N(bch); i++) {
913 		/* compute elp(a^i) */
914 		for (j = 1, syn = syn0; j <= p->deg; j++) {
915 			m = bch->cache[j];
916 			if (m >= 0)
917 				syn ^= a_pow(bch, m+j*i);
918 		}
919 		if (syn == 0) {
920 			roots[count++] = GF_N(bch)-i;
921 			if (count == p->deg)
922 				break;
923 		}
924 	}
925 	return (count == p->deg) ? count : 0;
926 }
927 #define find_poly_roots(_p, _k, _elp, _loc) chien_search(_p, len, _elp, _loc)
928 #endif /* USE_CHIEN_SEARCH */
929 
930 /**
931  * decode_bch - decode received codeword and find bit error locations
932  * @bch:      BCH control structure
933  * @data:     received data, ignored if @calc_ecc is provided
934  * @len:      data length in bytes, must always be provided
935  * @recv_ecc: received ecc, if NULL then assume it was XORed in @calc_ecc
936  * @calc_ecc: calculated ecc, if NULL then calc_ecc is computed from @data
937  * @syn:      hw computed syndrome data (if NULL, syndrome is calculated)
938  * @errloc:   output array of error locations
939  *
940  * Returns:
941  *  The number of errors found, or -EBADMSG if decoding failed, or -EINVAL if
942  *  invalid parameters were provided
943  *
944  * Depending on the available hw BCH support and the need to compute @calc_ecc
945  * separately (using encode_bch()), this function should be called with one of
946  * the following parameter configurations -
947  *
948  * by providing @data and @recv_ecc only:
949  *   decode_bch(@bch, @data, @len, @recv_ecc, NULL, NULL, @errloc)
950  *
951  * by providing @recv_ecc and @calc_ecc:
952  *   decode_bch(@bch, NULL, @len, @recv_ecc, @calc_ecc, NULL, @errloc)
953  *
954  * by providing ecc = recv_ecc XOR calc_ecc:
955  *   decode_bch(@bch, NULL, @len, NULL, ecc, NULL, @errloc)
956  *
957  * by providing syndrome results @syn:
958  *   decode_bch(@bch, NULL, @len, NULL, NULL, @syn, @errloc)
959  *
960  * Once decode_bch() has successfully returned with a positive value, error
961  * locations returned in array @errloc should be interpreted as follows -
962  *
963  * if (errloc[n] >= 8*len), then n-th error is located in ecc (no need for
964  * data correction)
965  *
966  * if (errloc[n] < 8*len), then n-th error is located in data and can be
967  * corrected with statement data[errloc[n]/8] ^= 1 << (errloc[n] % 8);
968  *
969  * Note that this function does not perform any data correction by itself, it
970  * merely indicates error locations.
971  */
972 int decode_bch(struct bch_control *bch, const uint8_t *data, unsigned int len,
973 	       const uint8_t *recv_ecc, const uint8_t *calc_ecc,
974 	       const unsigned int *syn, unsigned int *errloc)
975 {
976 	const unsigned int ecc_words = BCH_ECC_WORDS(bch);
977 	unsigned int nbits;
978 	int i, err, nroots;
979 	uint32_t sum;
980 
981 	/* sanity check: make sure data length can be handled */
982 	if (8*len > (bch->n-bch->ecc_bits))
983 		return -EINVAL;
984 
985 	/* if caller does not provide syndromes, compute them */
986 	if (!syn) {
987 		if (!calc_ecc) {
988 			/* compute received data ecc into an internal buffer */
989 			if (!data || !recv_ecc)
990 				return -EINVAL;
991 			encode_bch(bch, data, len, NULL);
992 		} else {
993 			/* load provided calculated ecc */
994 			load_ecc8(bch, bch->ecc_buf, calc_ecc);
995 		}
996 		/* load received ecc or assume it was XORed in calc_ecc */
997 		if (recv_ecc) {
998 			load_ecc8(bch, bch->ecc_buf2, recv_ecc);
999 			/* XOR received and calculated ecc */
1000 			for (i = 0, sum = 0; i < (int)ecc_words; i++) {
1001 				bch->ecc_buf[i] ^= bch->ecc_buf2[i];
1002 				sum |= bch->ecc_buf[i];
1003 			}
1004 			if (!sum)
1005 				/* no error found */
1006 				return 0;
1007 		}
1008 		compute_syndromes(bch, bch->ecc_buf, bch->syn);
1009 		syn = bch->syn;
1010 	}
1011 
1012 	err = compute_error_locator_polynomial(bch, syn);
1013 	if (err > 0) {
1014 		nroots = find_poly_roots(bch, 1, bch->elp, errloc);
1015 		if (err != nroots)
1016 			err = -1;
1017 	}
1018 	if (err > 0) {
1019 		/* post-process raw error locations for easier correction */
1020 		nbits = (len*8)+bch->ecc_bits;
1021 		for (i = 0; i < err; i++) {
1022 			if (errloc[i] >= nbits) {
1023 				err = -1;
1024 				break;
1025 			}
1026 			errloc[i] = nbits-1-errloc[i];
1027 			errloc[i] = (errloc[i] & ~7)|(7-(errloc[i] & 7));
1028 		}
1029 	}
1030 	return (err >= 0) ? err : -EBADMSG;
1031 }
1032 
1033 /*
1034  * generate Galois field lookup tables
1035  */
1036 static int build_gf_tables(struct bch_control *bch, unsigned int poly)
1037 {
1038 	unsigned int i, x = 1;
1039 	const unsigned int k = 1 << deg(poly);
1040 
1041 	/* primitive polynomial must be of degree m */
1042 	if (k != (1u << GF_M(bch)))
1043 		return -1;
1044 
1045 	for (i = 0; i < GF_N(bch); i++) {
1046 		bch->a_pow_tab[i] = x;
1047 		bch->a_log_tab[x] = i;
1048 		if (i && (x == 1))
1049 			/* polynomial is not primitive (a^i=1 with 0<i<2^m-1) */
1050 			return -1;
1051 		x <<= 1;
1052 		if (x & k)
1053 			x ^= poly;
1054 	}
1055 	bch->a_pow_tab[GF_N(bch)] = 1;
1056 	bch->a_log_tab[0] = 0;
1057 
1058 	return 0;
1059 }
1060 
1061 /*
1062  * compute generator polynomial remainder tables for fast encoding
1063  */
1064 static void build_mod8_tables(struct bch_control *bch, const uint32_t *g)
1065 {
1066 	int i, j, b, d;
1067 	uint32_t data, hi, lo, *tab;
1068 	const int l = BCH_ECC_WORDS(bch);
1069 	const int plen = DIV_ROUND_UP(bch->ecc_bits+1, 32);
1070 	const int ecclen = DIV_ROUND_UP(bch->ecc_bits, 32);
1071 
1072 	memset(bch->mod8_tab, 0, 4*256*l*sizeof(*bch->mod8_tab));
1073 
1074 	for (i = 0; i < 256; i++) {
1075 		/* p(X)=i is a small polynomial of weight <= 8 */
1076 		for (b = 0; b < 4; b++) {
1077 			/* we want to compute (p(X).X^(8*b+deg(g))) mod g(X) */
1078 			tab = bch->mod8_tab + (b*256+i)*l;
1079 			data = i << (8*b);
1080 			while (data) {
1081 				d = deg(data);
1082 				/* subtract X^d.g(X) from p(X).X^(8*b+deg(g)) */
1083 				data ^= g[0] >> (31-d);
1084 				for (j = 0; j < ecclen; j++) {
1085 					hi = (d < 31) ? g[j] << (d+1) : 0;
1086 					lo = (j+1 < plen) ?
1087 						g[j+1] >> (31-d) : 0;
1088 					tab[j] ^= hi|lo;
1089 				}
1090 			}
1091 		}
1092 	}
1093 }
1094 
1095 /*
1096  * build a base for factoring degree 2 polynomials
1097  */
1098 static int build_deg2_base(struct bch_control *bch)
1099 {
1100 	const int m = GF_M(bch);
1101 	int i, j, r;
1102 	unsigned int sum, x, y, remaining, ak = 0, xi[m];
1103 
1104 	/* find k s.t. Tr(a^k) = 1 and 0 <= k < m */
1105 	for (i = 0; i < m; i++) {
1106 		for (j = 0, sum = 0; j < m; j++)
1107 			sum ^= a_pow(bch, i*(1 << j));
1108 
1109 		if (sum) {
1110 			ak = bch->a_pow_tab[i];
1111 			break;
1112 		}
1113 	}
1114 	/* find xi, i=0..m-1 such that xi^2+xi = a^i+Tr(a^i).a^k */
1115 	remaining = m;
1116 	memset(xi, 0, sizeof(xi));
1117 
1118 	for (x = 0; (x <= GF_N(bch)) && remaining; x++) {
1119 		y = gf_sqr(bch, x)^x;
1120 		for (i = 0; i < 2; i++) {
1121 			r = a_log(bch, y);
1122 			if (y && (r < m) && !xi[r]) {
1123 				bch->xi_tab[r] = x;
1124 				xi[r] = 1;
1125 				remaining--;
1126 				dbg("x%d = %x\n", r, x);
1127 				break;
1128 			}
1129 			y ^= ak;
1130 		}
1131 	}
1132 	/* should not happen but check anyway */
1133 	return remaining ? -1 : 0;
1134 }
1135 
1136 static void *bch_alloc(size_t size, int *err)
1137 {
1138 	void *ptr;
1139 
1140 	ptr = kmalloc(size, GFP_KERNEL);
1141 	if (ptr == NULL)
1142 		*err = 1;
1143 	return ptr;
1144 }
1145 
1146 /*
1147  * compute generator polynomial for given (m,t) parameters.
1148  */
1149 static uint32_t *compute_generator_polynomial(struct bch_control *bch)
1150 {
1151 	const unsigned int m = GF_M(bch);
1152 	const unsigned int t = GF_T(bch);
1153 	int n, err = 0;
1154 	unsigned int i, j, nbits, r, word, *roots;
1155 	struct gf_poly *g;
1156 	uint32_t *genpoly;
1157 
1158 	g = bch_alloc(GF_POLY_SZ(m*t), &err);
1159 	roots = bch_alloc((bch->n+1)*sizeof(*roots), &err);
1160 	genpoly = bch_alloc(DIV_ROUND_UP(m*t+1, 32)*sizeof(*genpoly), &err);
1161 
1162 	if (err) {
1163 		kfree(genpoly);
1164 		genpoly = NULL;
1165 		goto finish;
1166 	}
1167 
1168 	/* enumerate all roots of g(X) */
1169 	memset(roots , 0, (bch->n+1)*sizeof(*roots));
1170 	for (i = 0; i < t; i++) {
1171 		for (j = 0, r = 2*i+1; j < m; j++) {
1172 			roots[r] = 1;
1173 			r = mod_s(bch, 2*r);
1174 		}
1175 	}
1176 	/* build generator polynomial g(X) */
1177 	g->deg = 0;
1178 	g->c[0] = 1;
1179 	for (i = 0; i < GF_N(bch); i++) {
1180 		if (roots[i]) {
1181 			/* multiply g(X) by (X+root) */
1182 			r = bch->a_pow_tab[i];
1183 			g->c[g->deg+1] = 1;
1184 			for (j = g->deg; j > 0; j--)
1185 				g->c[j] = gf_mul(bch, g->c[j], r)^g->c[j-1];
1186 
1187 			g->c[0] = gf_mul(bch, g->c[0], r);
1188 			g->deg++;
1189 		}
1190 	}
1191 	/* store left-justified binary representation of g(X) */
1192 	n = g->deg+1;
1193 	i = 0;
1194 
1195 	while (n > 0) {
1196 		nbits = (n > 32) ? 32 : n;
1197 		for (j = 0, word = 0; j < nbits; j++) {
1198 			if (g->c[n-1-j])
1199 				word |= 1u << (31-j);
1200 		}
1201 		genpoly[i++] = word;
1202 		n -= nbits;
1203 	}
1204 	bch->ecc_bits = g->deg;
1205 
1206 finish:
1207 	kfree(g);
1208 	kfree(roots);
1209 
1210 	return genpoly;
1211 }
1212 
1213 /**
1214  * init_bch - initialize a BCH encoder/decoder
1215  * @m:          Galois field order, should be in the range 5-15
1216  * @t:          maximum error correction capability, in bits
1217  * @prim_poly:  user-provided primitive polynomial (or 0 to use default)
1218  *
1219  * Returns:
1220  *  a newly allocated BCH control structure if successful, NULL otherwise
1221  *
1222  * This initialization can take some time, as lookup tables are built for fast
1223  * encoding/decoding; make sure not to call this function from a time critical
1224  * path. Usually, init_bch() should be called on module/driver init and
1225  * free_bch() should be called to release memory on exit.
1226  *
1227  * You may provide your own primitive polynomial of degree @m in argument
1228  * @prim_poly, or let init_bch() use its default polynomial.
1229  *
1230  * Once init_bch() has successfully returned a pointer to a newly allocated
1231  * BCH control structure, ecc length in bytes is given by member @ecc_bytes of
1232  * the structure.
1233  */
1234 struct bch_control *init_bch(int m, int t, unsigned int prim_poly)
1235 {
1236 	int err = 0;
1237 	unsigned int i, words;
1238 	uint32_t *genpoly;
1239 	struct bch_control *bch = NULL;
1240 
1241 	const int min_m = 5;
1242 	const int max_m = 15;
1243 
1244 	/* default primitive polynomials */
1245 	static const unsigned int prim_poly_tab[] = {
1246 		0x25, 0x43, 0x83, 0x11d, 0x211, 0x409, 0x805, 0x1053, 0x201b,
1247 		0x402b, 0x8003,
1248 	};
1249 
1250 #if defined(CONFIG_BCH_CONST_PARAMS)
1251 	if ((m != (CONFIG_BCH_CONST_M)) || (t != (CONFIG_BCH_CONST_T))) {
1252 		printk(KERN_ERR "bch encoder/decoder was configured to support "
1253 		       "parameters m=%d, t=%d only!\n",
1254 		       CONFIG_BCH_CONST_M, CONFIG_BCH_CONST_T);
1255 		goto fail;
1256 	}
1257 #endif
1258 	if ((m < min_m) || (m > max_m))
1259 		/*
1260 		 * values of m greater than 15 are not currently supported;
1261 		 * supporting m > 15 would require changing table base type
1262 		 * (uint16_t) and a small patch in matrix transposition
1263 		 */
1264 		goto fail;
1265 
1266 	/* sanity checks */
1267 	if ((t < 1) || (m*t >= ((1 << m)-1)))
1268 		/* invalid t value */
1269 		goto fail;
1270 
1271 	/* select a primitive polynomial for generating GF(2^m) */
1272 	if (prim_poly == 0)
1273 		prim_poly = prim_poly_tab[m-min_m];
1274 
1275 	bch = kzalloc(sizeof(*bch), GFP_KERNEL);
1276 	if (bch == NULL)
1277 		goto fail;
1278 
1279 	bch->m = m;
1280 	bch->t = t;
1281 	bch->n = (1 << m)-1;
1282 	words  = DIV_ROUND_UP(m*t, 32);
1283 	bch->ecc_bytes = DIV_ROUND_UP(m*t, 8);
1284 	bch->a_pow_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_pow_tab), &err);
1285 	bch->a_log_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_log_tab), &err);
1286 	bch->mod8_tab  = bch_alloc(words*1024*sizeof(*bch->mod8_tab), &err);
1287 	bch->ecc_buf   = bch_alloc(words*sizeof(*bch->ecc_buf), &err);
1288 	bch->ecc_buf2  = bch_alloc(words*sizeof(*bch->ecc_buf2), &err);
1289 	bch->xi_tab    = bch_alloc(m*sizeof(*bch->xi_tab), &err);
1290 	bch->syn       = bch_alloc(2*t*sizeof(*bch->syn), &err);
1291 	bch->cache     = bch_alloc(2*t*sizeof(*bch->cache), &err);
1292 	bch->elp       = bch_alloc((t+1)*sizeof(struct gf_poly_deg1), &err);
1293 
1294 	for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++)
1295 		bch->poly_2t[i] = bch_alloc(GF_POLY_SZ(2*t), &err);
1296 
1297 	if (err)
1298 		goto fail;
1299 
1300 	err = build_gf_tables(bch, prim_poly);
1301 	if (err)
1302 		goto fail;
1303 
1304 	/* use generator polynomial for computing encoding tables */
1305 	genpoly = compute_generator_polynomial(bch);
1306 	if (genpoly == NULL)
1307 		goto fail;
1308 
1309 	build_mod8_tables(bch, genpoly);
1310 	kfree(genpoly);
1311 
1312 	err = build_deg2_base(bch);
1313 	if (err)
1314 		goto fail;
1315 
1316 	return bch;
1317 
1318 fail:
1319 	free_bch(bch);
1320 	return NULL;
1321 }
1322 
1323 /**
1324  *  free_bch - free the BCH control structure
1325  *  @bch:    BCH control structure to release
1326  */
1327 void free_bch(struct bch_control *bch)
1328 {
1329 	unsigned int i;
1330 
1331 	if (bch) {
1332 		kfree(bch->a_pow_tab);
1333 		kfree(bch->a_log_tab);
1334 		kfree(bch->mod8_tab);
1335 		kfree(bch->ecc_buf);
1336 		kfree(bch->ecc_buf2);
1337 		kfree(bch->xi_tab);
1338 		kfree(bch->syn);
1339 		kfree(bch->cache);
1340 		kfree(bch->elp);
1341 
1342 		for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++)
1343 			kfree(bch->poly_2t[i]);
1344 
1345 		kfree(bch);
1346 	}
1347 }
1348