xref: /openbmc/linux/lib/bch.c (revision ccb01374)
1 /*
2  * Generic binary BCH encoding/decoding library
3  *
4  * This program is free software; you can redistribute it and/or modify it
5  * under the terms of the GNU General Public License version 2 as published by
6  * the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful, but WITHOUT
9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
11  * more details.
12  *
13  * You should have received a copy of the GNU General Public License along with
14  * this program; if not, write to the Free Software Foundation, Inc., 51
15  * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
16  *
17  * Copyright © 2011 Parrot S.A.
18  *
19  * Author: Ivan Djelic <ivan.djelic@parrot.com>
20  *
21  * Description:
22  *
23  * This library provides runtime configurable encoding/decoding of binary
24  * Bose-Chaudhuri-Hocquenghem (BCH) codes.
25  *
26  * Call init_bch to get a pointer to a newly allocated bch_control structure for
27  * the given m (Galois field order), t (error correction capability) and
28  * (optional) primitive polynomial parameters.
29  *
30  * Call encode_bch to compute and store ecc parity bytes to a given buffer.
31  * Call decode_bch to detect and locate errors in received data.
32  *
33  * On systems supporting hw BCH features, intermediate results may be provided
34  * to decode_bch in order to skip certain steps. See decode_bch() documentation
35  * for details.
36  *
37  * Option CONFIG_BCH_CONST_PARAMS can be used to force fixed values of
38  * parameters m and t; thus allowing extra compiler optimizations and providing
39  * better (up to 2x) encoding performance. Using this option makes sense when
40  * (m,t) are fixed and known in advance, e.g. when using BCH error correction
41  * on a particular NAND flash device.
42  *
43  * Algorithmic details:
44  *
45  * Encoding is performed by processing 32 input bits in parallel, using 4
46  * remainder lookup tables.
47  *
48  * The final stage of decoding involves the following internal steps:
49  * a. Syndrome computation
50  * b. Error locator polynomial computation using Berlekamp-Massey algorithm
51  * c. Error locator root finding (by far the most expensive step)
52  *
53  * In this implementation, step c is not performed using the usual Chien search.
54  * Instead, an alternative approach described in [1] is used. It consists in
55  * factoring the error locator polynomial using the Berlekamp Trace algorithm
56  * (BTA) down to a certain degree (4), after which ad hoc low-degree polynomial
57  * solving techniques [2] are used. The resulting algorithm, called BTZ, yields
58  * much better performance than Chien search for usual (m,t) values (typically
59  * m >= 13, t < 32, see [1]).
60  *
61  * [1] B. Biswas, V. Herbert. Efficient root finding of polynomials over fields
62  * of characteristic 2, in: Western European Workshop on Research in Cryptology
63  * - WEWoRC 2009, Graz, Austria, LNCS, Springer, July 2009, to appear.
64  * [2] [Zin96] V.A. Zinoviev. On the solution of equations of degree 10 over
65  * finite fields GF(2^q). In Rapport de recherche INRIA no 2829, 1996.
66  */
67 
68 #include <linux/kernel.h>
69 #include <linux/errno.h>
70 #include <linux/init.h>
71 #include <linux/module.h>
72 #include <linux/slab.h>
73 #include <linux/bitops.h>
74 #include <asm/byteorder.h>
75 #include <linux/bch.h>
76 
77 #if defined(CONFIG_BCH_CONST_PARAMS)
78 #define GF_M(_p)               (CONFIG_BCH_CONST_M)
79 #define GF_T(_p)               (CONFIG_BCH_CONST_T)
80 #define GF_N(_p)               ((1 << (CONFIG_BCH_CONST_M))-1)
81 #define BCH_MAX_M              (CONFIG_BCH_CONST_M)
82 #define BCH_MAX_T	       (CONFIG_BCH_CONST_T)
83 #else
84 #define GF_M(_p)               ((_p)->m)
85 #define GF_T(_p)               ((_p)->t)
86 #define GF_N(_p)               ((_p)->n)
87 #define BCH_MAX_M              15 /* 2KB */
88 #define BCH_MAX_T              64 /* 64 bit correction */
89 #endif
90 
91 #define BCH_ECC_WORDS(_p)      DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 32)
92 #define BCH_ECC_BYTES(_p)      DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 8)
93 
94 #define BCH_ECC_MAX_WORDS      DIV_ROUND_UP(BCH_MAX_M * BCH_MAX_T, 32)
95 
96 #ifndef dbg
97 #define dbg(_fmt, args...)     do {} while (0)
98 #endif
99 
100 /*
101  * represent a polynomial over GF(2^m)
102  */
103 struct gf_poly {
104 	unsigned int deg;    /* polynomial degree */
105 	unsigned int c[0];   /* polynomial terms */
106 };
107 
108 /* given its degree, compute a polynomial size in bytes */
109 #define GF_POLY_SZ(_d) (sizeof(struct gf_poly)+((_d)+1)*sizeof(unsigned int))
110 
111 /* polynomial of degree 1 */
112 struct gf_poly_deg1 {
113 	struct gf_poly poly;
114 	unsigned int   c[2];
115 };
116 
117 /*
118  * same as encode_bch(), but process input data one byte at a time
119  */
120 static void encode_bch_unaligned(struct bch_control *bch,
121 				 const unsigned char *data, unsigned int len,
122 				 uint32_t *ecc)
123 {
124 	int i;
125 	const uint32_t *p;
126 	const int l = BCH_ECC_WORDS(bch)-1;
127 
128 	while (len--) {
129 		p = bch->mod8_tab + (l+1)*(((ecc[0] >> 24)^(*data++)) & 0xff);
130 
131 		for (i = 0; i < l; i++)
132 			ecc[i] = ((ecc[i] << 8)|(ecc[i+1] >> 24))^(*p++);
133 
134 		ecc[l] = (ecc[l] << 8)^(*p);
135 	}
136 }
137 
138 /*
139  * convert ecc bytes to aligned, zero-padded 32-bit ecc words
140  */
141 static void load_ecc8(struct bch_control *bch, uint32_t *dst,
142 		      const uint8_t *src)
143 {
144 	uint8_t pad[4] = {0, 0, 0, 0};
145 	unsigned int i, nwords = BCH_ECC_WORDS(bch)-1;
146 
147 	for (i = 0; i < nwords; i++, src += 4)
148 		dst[i] = (src[0] << 24)|(src[1] << 16)|(src[2] << 8)|src[3];
149 
150 	memcpy(pad, src, BCH_ECC_BYTES(bch)-4*nwords);
151 	dst[nwords] = (pad[0] << 24)|(pad[1] << 16)|(pad[2] << 8)|pad[3];
152 }
153 
154 /*
155  * convert 32-bit ecc words to ecc bytes
156  */
157 static void store_ecc8(struct bch_control *bch, uint8_t *dst,
158 		       const uint32_t *src)
159 {
160 	uint8_t pad[4];
161 	unsigned int i, nwords = BCH_ECC_WORDS(bch)-1;
162 
163 	for (i = 0; i < nwords; i++) {
164 		*dst++ = (src[i] >> 24);
165 		*dst++ = (src[i] >> 16) & 0xff;
166 		*dst++ = (src[i] >>  8) & 0xff;
167 		*dst++ = (src[i] >>  0) & 0xff;
168 	}
169 	pad[0] = (src[nwords] >> 24);
170 	pad[1] = (src[nwords] >> 16) & 0xff;
171 	pad[2] = (src[nwords] >>  8) & 0xff;
172 	pad[3] = (src[nwords] >>  0) & 0xff;
173 	memcpy(dst, pad, BCH_ECC_BYTES(bch)-4*nwords);
174 }
175 
176 /**
177  * encode_bch - calculate BCH ecc parity of data
178  * @bch:   BCH control structure
179  * @data:  data to encode
180  * @len:   data length in bytes
181  * @ecc:   ecc parity data, must be initialized by caller
182  *
183  * The @ecc parity array is used both as input and output parameter, in order to
184  * allow incremental computations. It should be of the size indicated by member
185  * @ecc_bytes of @bch, and should be initialized to 0 before the first call.
186  *
187  * The exact number of computed ecc parity bits is given by member @ecc_bits of
188  * @bch; it may be less than m*t for large values of t.
189  */
190 void encode_bch(struct bch_control *bch, const uint8_t *data,
191 		unsigned int len, uint8_t *ecc)
192 {
193 	const unsigned int l = BCH_ECC_WORDS(bch)-1;
194 	unsigned int i, mlen;
195 	unsigned long m;
196 	uint32_t w, r[BCH_ECC_MAX_WORDS];
197 	const size_t r_bytes = BCH_ECC_WORDS(bch) * sizeof(*r);
198 	const uint32_t * const tab0 = bch->mod8_tab;
199 	const uint32_t * const tab1 = tab0 + 256*(l+1);
200 	const uint32_t * const tab2 = tab1 + 256*(l+1);
201 	const uint32_t * const tab3 = tab2 + 256*(l+1);
202 	const uint32_t *pdata, *p0, *p1, *p2, *p3;
203 
204 	if (WARN_ON(r_bytes > sizeof(r)))
205 		return;
206 
207 	if (ecc) {
208 		/* load ecc parity bytes into internal 32-bit buffer */
209 		load_ecc8(bch, bch->ecc_buf, ecc);
210 	} else {
211 		memset(bch->ecc_buf, 0, r_bytes);
212 	}
213 
214 	/* process first unaligned data bytes */
215 	m = ((unsigned long)data) & 3;
216 	if (m) {
217 		mlen = (len < (4-m)) ? len : 4-m;
218 		encode_bch_unaligned(bch, data, mlen, bch->ecc_buf);
219 		data += mlen;
220 		len  -= mlen;
221 	}
222 
223 	/* process 32-bit aligned data words */
224 	pdata = (uint32_t *)data;
225 	mlen  = len/4;
226 	data += 4*mlen;
227 	len  -= 4*mlen;
228 	memcpy(r, bch->ecc_buf, r_bytes);
229 
230 	/*
231 	 * split each 32-bit word into 4 polynomials of weight 8 as follows:
232 	 *
233 	 * 31 ...24  23 ...16  15 ... 8  7 ... 0
234 	 * xxxxxxxx  yyyyyyyy  zzzzzzzz  tttttttt
235 	 *                               tttttttt  mod g = r0 (precomputed)
236 	 *                     zzzzzzzz  00000000  mod g = r1 (precomputed)
237 	 *           yyyyyyyy  00000000  00000000  mod g = r2 (precomputed)
238 	 * xxxxxxxx  00000000  00000000  00000000  mod g = r3 (precomputed)
239 	 * xxxxxxxx  yyyyyyyy  zzzzzzzz  tttttttt  mod g = r0^r1^r2^r3
240 	 */
241 	while (mlen--) {
242 		/* input data is read in big-endian format */
243 		w = r[0]^cpu_to_be32(*pdata++);
244 		p0 = tab0 + (l+1)*((w >>  0) & 0xff);
245 		p1 = tab1 + (l+1)*((w >>  8) & 0xff);
246 		p2 = tab2 + (l+1)*((w >> 16) & 0xff);
247 		p3 = tab3 + (l+1)*((w >> 24) & 0xff);
248 
249 		for (i = 0; i < l; i++)
250 			r[i] = r[i+1]^p0[i]^p1[i]^p2[i]^p3[i];
251 
252 		r[l] = p0[l]^p1[l]^p2[l]^p3[l];
253 	}
254 	memcpy(bch->ecc_buf, r, r_bytes);
255 
256 	/* process last unaligned bytes */
257 	if (len)
258 		encode_bch_unaligned(bch, data, len, bch->ecc_buf);
259 
260 	/* store ecc parity bytes into original parity buffer */
261 	if (ecc)
262 		store_ecc8(bch, ecc, bch->ecc_buf);
263 }
264 EXPORT_SYMBOL_GPL(encode_bch);
265 
266 static inline int modulo(struct bch_control *bch, unsigned int v)
267 {
268 	const unsigned int n = GF_N(bch);
269 	while (v >= n) {
270 		v -= n;
271 		v = (v & n) + (v >> GF_M(bch));
272 	}
273 	return v;
274 }
275 
276 /*
277  * shorter and faster modulo function, only works when v < 2N.
278  */
279 static inline int mod_s(struct bch_control *bch, unsigned int v)
280 {
281 	const unsigned int n = GF_N(bch);
282 	return (v < n) ? v : v-n;
283 }
284 
285 static inline int deg(unsigned int poly)
286 {
287 	/* polynomial degree is the most-significant bit index */
288 	return fls(poly)-1;
289 }
290 
291 static inline int parity(unsigned int x)
292 {
293 	/*
294 	 * public domain code snippet, lifted from
295 	 * http://www-graphics.stanford.edu/~seander/bithacks.html
296 	 */
297 	x ^= x >> 1;
298 	x ^= x >> 2;
299 	x = (x & 0x11111111U) * 0x11111111U;
300 	return (x >> 28) & 1;
301 }
302 
303 /* Galois field basic operations: multiply, divide, inverse, etc. */
304 
305 static inline unsigned int gf_mul(struct bch_control *bch, unsigned int a,
306 				  unsigned int b)
307 {
308 	return (a && b) ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+
309 					       bch->a_log_tab[b])] : 0;
310 }
311 
312 static inline unsigned int gf_sqr(struct bch_control *bch, unsigned int a)
313 {
314 	return a ? bch->a_pow_tab[mod_s(bch, 2*bch->a_log_tab[a])] : 0;
315 }
316 
317 static inline unsigned int gf_div(struct bch_control *bch, unsigned int a,
318 				  unsigned int b)
319 {
320 	return a ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+
321 					GF_N(bch)-bch->a_log_tab[b])] : 0;
322 }
323 
324 static inline unsigned int gf_inv(struct bch_control *bch, unsigned int a)
325 {
326 	return bch->a_pow_tab[GF_N(bch)-bch->a_log_tab[a]];
327 }
328 
329 static inline unsigned int a_pow(struct bch_control *bch, int i)
330 {
331 	return bch->a_pow_tab[modulo(bch, i)];
332 }
333 
334 static inline int a_log(struct bch_control *bch, unsigned int x)
335 {
336 	return bch->a_log_tab[x];
337 }
338 
339 static inline int a_ilog(struct bch_control *bch, unsigned int x)
340 {
341 	return mod_s(bch, GF_N(bch)-bch->a_log_tab[x]);
342 }
343 
344 /*
345  * compute 2t syndromes of ecc polynomial, i.e. ecc(a^j) for j=1..2t
346  */
347 static void compute_syndromes(struct bch_control *bch, uint32_t *ecc,
348 			      unsigned int *syn)
349 {
350 	int i, j, s;
351 	unsigned int m;
352 	uint32_t poly;
353 	const int t = GF_T(bch);
354 
355 	s = bch->ecc_bits;
356 
357 	/* make sure extra bits in last ecc word are cleared */
358 	m = ((unsigned int)s) & 31;
359 	if (m)
360 		ecc[s/32] &= ~((1u << (32-m))-1);
361 	memset(syn, 0, 2*t*sizeof(*syn));
362 
363 	/* compute v(a^j) for j=1 .. 2t-1 */
364 	do {
365 		poly = *ecc++;
366 		s -= 32;
367 		while (poly) {
368 			i = deg(poly);
369 			for (j = 0; j < 2*t; j += 2)
370 				syn[j] ^= a_pow(bch, (j+1)*(i+s));
371 
372 			poly ^= (1 << i);
373 		}
374 	} while (s > 0);
375 
376 	/* v(a^(2j)) = v(a^j)^2 */
377 	for (j = 0; j < t; j++)
378 		syn[2*j+1] = gf_sqr(bch, syn[j]);
379 }
380 
381 static void gf_poly_copy(struct gf_poly *dst, struct gf_poly *src)
382 {
383 	memcpy(dst, src, GF_POLY_SZ(src->deg));
384 }
385 
386 static int compute_error_locator_polynomial(struct bch_control *bch,
387 					    const unsigned int *syn)
388 {
389 	const unsigned int t = GF_T(bch);
390 	const unsigned int n = GF_N(bch);
391 	unsigned int i, j, tmp, l, pd = 1, d = syn[0];
392 	struct gf_poly *elp = bch->elp;
393 	struct gf_poly *pelp = bch->poly_2t[0];
394 	struct gf_poly *elp_copy = bch->poly_2t[1];
395 	int k, pp = -1;
396 
397 	memset(pelp, 0, GF_POLY_SZ(2*t));
398 	memset(elp, 0, GF_POLY_SZ(2*t));
399 
400 	pelp->deg = 0;
401 	pelp->c[0] = 1;
402 	elp->deg = 0;
403 	elp->c[0] = 1;
404 
405 	/* use simplified binary Berlekamp-Massey algorithm */
406 	for (i = 0; (i < t) && (elp->deg <= t); i++) {
407 		if (d) {
408 			k = 2*i-pp;
409 			gf_poly_copy(elp_copy, elp);
410 			/* e[i+1](X) = e[i](X)+di*dp^-1*X^2(i-p)*e[p](X) */
411 			tmp = a_log(bch, d)+n-a_log(bch, pd);
412 			for (j = 0; j <= pelp->deg; j++) {
413 				if (pelp->c[j]) {
414 					l = a_log(bch, pelp->c[j]);
415 					elp->c[j+k] ^= a_pow(bch, tmp+l);
416 				}
417 			}
418 			/* compute l[i+1] = max(l[i]->c[l[p]+2*(i-p]) */
419 			tmp = pelp->deg+k;
420 			if (tmp > elp->deg) {
421 				elp->deg = tmp;
422 				gf_poly_copy(pelp, elp_copy);
423 				pd = d;
424 				pp = 2*i;
425 			}
426 		}
427 		/* di+1 = S(2i+3)+elp[i+1].1*S(2i+2)+...+elp[i+1].lS(2i+3-l) */
428 		if (i < t-1) {
429 			d = syn[2*i+2];
430 			for (j = 1; j <= elp->deg; j++)
431 				d ^= gf_mul(bch, elp->c[j], syn[2*i+2-j]);
432 		}
433 	}
434 	dbg("elp=%s\n", gf_poly_str(elp));
435 	return (elp->deg > t) ? -1 : (int)elp->deg;
436 }
437 
438 /*
439  * solve a m x m linear system in GF(2) with an expected number of solutions,
440  * and return the number of found solutions
441  */
442 static int solve_linear_system(struct bch_control *bch, unsigned int *rows,
443 			       unsigned int *sol, int nsol)
444 {
445 	const int m = GF_M(bch);
446 	unsigned int tmp, mask;
447 	int rem, c, r, p, k, param[BCH_MAX_M];
448 
449 	k = 0;
450 	mask = 1 << m;
451 
452 	/* Gaussian elimination */
453 	for (c = 0; c < m; c++) {
454 		rem = 0;
455 		p = c-k;
456 		/* find suitable row for elimination */
457 		for (r = p; r < m; r++) {
458 			if (rows[r] & mask) {
459 				if (r != p) {
460 					tmp = rows[r];
461 					rows[r] = rows[p];
462 					rows[p] = tmp;
463 				}
464 				rem = r+1;
465 				break;
466 			}
467 		}
468 		if (rem) {
469 			/* perform elimination on remaining rows */
470 			tmp = rows[p];
471 			for (r = rem; r < m; r++) {
472 				if (rows[r] & mask)
473 					rows[r] ^= tmp;
474 			}
475 		} else {
476 			/* elimination not needed, store defective row index */
477 			param[k++] = c;
478 		}
479 		mask >>= 1;
480 	}
481 	/* rewrite system, inserting fake parameter rows */
482 	if (k > 0) {
483 		p = k;
484 		for (r = m-1; r >= 0; r--) {
485 			if ((r > m-1-k) && rows[r])
486 				/* system has no solution */
487 				return 0;
488 
489 			rows[r] = (p && (r == param[p-1])) ?
490 				p--, 1u << (m-r) : rows[r-p];
491 		}
492 	}
493 
494 	if (nsol != (1 << k))
495 		/* unexpected number of solutions */
496 		return 0;
497 
498 	for (p = 0; p < nsol; p++) {
499 		/* set parameters for p-th solution */
500 		for (c = 0; c < k; c++)
501 			rows[param[c]] = (rows[param[c]] & ~1)|((p >> c) & 1);
502 
503 		/* compute unique solution */
504 		tmp = 0;
505 		for (r = m-1; r >= 0; r--) {
506 			mask = rows[r] & (tmp|1);
507 			tmp |= parity(mask) << (m-r);
508 		}
509 		sol[p] = tmp >> 1;
510 	}
511 	return nsol;
512 }
513 
514 /*
515  * this function builds and solves a linear system for finding roots of a degree
516  * 4 affine monic polynomial X^4+aX^2+bX+c over GF(2^m).
517  */
518 static int find_affine4_roots(struct bch_control *bch, unsigned int a,
519 			      unsigned int b, unsigned int c,
520 			      unsigned int *roots)
521 {
522 	int i, j, k;
523 	const int m = GF_M(bch);
524 	unsigned int mask = 0xff, t, rows[16] = {0,};
525 
526 	j = a_log(bch, b);
527 	k = a_log(bch, a);
528 	rows[0] = c;
529 
530 	/* buid linear system to solve X^4+aX^2+bX+c = 0 */
531 	for (i = 0; i < m; i++) {
532 		rows[i+1] = bch->a_pow_tab[4*i]^
533 			(a ? bch->a_pow_tab[mod_s(bch, k)] : 0)^
534 			(b ? bch->a_pow_tab[mod_s(bch, j)] : 0);
535 		j++;
536 		k += 2;
537 	}
538 	/*
539 	 * transpose 16x16 matrix before passing it to linear solver
540 	 * warning: this code assumes m < 16
541 	 */
542 	for (j = 8; j != 0; j >>= 1, mask ^= (mask << j)) {
543 		for (k = 0; k < 16; k = (k+j+1) & ~j) {
544 			t = ((rows[k] >> j)^rows[k+j]) & mask;
545 			rows[k] ^= (t << j);
546 			rows[k+j] ^= t;
547 		}
548 	}
549 	return solve_linear_system(bch, rows, roots, 4);
550 }
551 
552 /*
553  * compute root r of a degree 1 polynomial over GF(2^m) (returned as log(1/r))
554  */
555 static int find_poly_deg1_roots(struct bch_control *bch, struct gf_poly *poly,
556 				unsigned int *roots)
557 {
558 	int n = 0;
559 
560 	if (poly->c[0])
561 		/* poly[X] = bX+c with c!=0, root=c/b */
562 		roots[n++] = mod_s(bch, GF_N(bch)-bch->a_log_tab[poly->c[0]]+
563 				   bch->a_log_tab[poly->c[1]]);
564 	return n;
565 }
566 
567 /*
568  * compute roots of a degree 2 polynomial over GF(2^m)
569  */
570 static int find_poly_deg2_roots(struct bch_control *bch, struct gf_poly *poly,
571 				unsigned int *roots)
572 {
573 	int n = 0, i, l0, l1, l2;
574 	unsigned int u, v, r;
575 
576 	if (poly->c[0] && poly->c[1]) {
577 
578 		l0 = bch->a_log_tab[poly->c[0]];
579 		l1 = bch->a_log_tab[poly->c[1]];
580 		l2 = bch->a_log_tab[poly->c[2]];
581 
582 		/* using z=a/bX, transform aX^2+bX+c into z^2+z+u (u=ac/b^2) */
583 		u = a_pow(bch, l0+l2+2*(GF_N(bch)-l1));
584 		/*
585 		 * let u = sum(li.a^i) i=0..m-1; then compute r = sum(li.xi):
586 		 * r^2+r = sum(li.(xi^2+xi)) = sum(li.(a^i+Tr(a^i).a^k)) =
587 		 * u + sum(li.Tr(a^i).a^k) = u+a^k.Tr(sum(li.a^i)) = u+a^k.Tr(u)
588 		 * i.e. r and r+1 are roots iff Tr(u)=0
589 		 */
590 		r = 0;
591 		v = u;
592 		while (v) {
593 			i = deg(v);
594 			r ^= bch->xi_tab[i];
595 			v ^= (1 << i);
596 		}
597 		/* verify root */
598 		if ((gf_sqr(bch, r)^r) == u) {
599 			/* reverse z=a/bX transformation and compute log(1/r) */
600 			roots[n++] = modulo(bch, 2*GF_N(bch)-l1-
601 					    bch->a_log_tab[r]+l2);
602 			roots[n++] = modulo(bch, 2*GF_N(bch)-l1-
603 					    bch->a_log_tab[r^1]+l2);
604 		}
605 	}
606 	return n;
607 }
608 
609 /*
610  * compute roots of a degree 3 polynomial over GF(2^m)
611  */
612 static int find_poly_deg3_roots(struct bch_control *bch, struct gf_poly *poly,
613 				unsigned int *roots)
614 {
615 	int i, n = 0;
616 	unsigned int a, b, c, a2, b2, c2, e3, tmp[4];
617 
618 	if (poly->c[0]) {
619 		/* transform polynomial into monic X^3 + a2X^2 + b2X + c2 */
620 		e3 = poly->c[3];
621 		c2 = gf_div(bch, poly->c[0], e3);
622 		b2 = gf_div(bch, poly->c[1], e3);
623 		a2 = gf_div(bch, poly->c[2], e3);
624 
625 		/* (X+a2)(X^3+a2X^2+b2X+c2) = X^4+aX^2+bX+c (affine) */
626 		c = gf_mul(bch, a2, c2);           /* c = a2c2      */
627 		b = gf_mul(bch, a2, b2)^c2;        /* b = a2b2 + c2 */
628 		a = gf_sqr(bch, a2)^b2;            /* a = a2^2 + b2 */
629 
630 		/* find the 4 roots of this affine polynomial */
631 		if (find_affine4_roots(bch, a, b, c, tmp) == 4) {
632 			/* remove a2 from final list of roots */
633 			for (i = 0; i < 4; i++) {
634 				if (tmp[i] != a2)
635 					roots[n++] = a_ilog(bch, tmp[i]);
636 			}
637 		}
638 	}
639 	return n;
640 }
641 
642 /*
643  * compute roots of a degree 4 polynomial over GF(2^m)
644  */
645 static int find_poly_deg4_roots(struct bch_control *bch, struct gf_poly *poly,
646 				unsigned int *roots)
647 {
648 	int i, l, n = 0;
649 	unsigned int a, b, c, d, e = 0, f, a2, b2, c2, e4;
650 
651 	if (poly->c[0] == 0)
652 		return 0;
653 
654 	/* transform polynomial into monic X^4 + aX^3 + bX^2 + cX + d */
655 	e4 = poly->c[4];
656 	d = gf_div(bch, poly->c[0], e4);
657 	c = gf_div(bch, poly->c[1], e4);
658 	b = gf_div(bch, poly->c[2], e4);
659 	a = gf_div(bch, poly->c[3], e4);
660 
661 	/* use Y=1/X transformation to get an affine polynomial */
662 	if (a) {
663 		/* first, eliminate cX by using z=X+e with ae^2+c=0 */
664 		if (c) {
665 			/* compute e such that e^2 = c/a */
666 			f = gf_div(bch, c, a);
667 			l = a_log(bch, f);
668 			l += (l & 1) ? GF_N(bch) : 0;
669 			e = a_pow(bch, l/2);
670 			/*
671 			 * use transformation z=X+e:
672 			 * z^4+e^4 + a(z^3+ez^2+e^2z+e^3) + b(z^2+e^2) +cz+ce+d
673 			 * z^4 + az^3 + (ae+b)z^2 + (ae^2+c)z+e^4+be^2+ae^3+ce+d
674 			 * z^4 + az^3 + (ae+b)z^2 + e^4+be^2+d
675 			 * z^4 + az^3 +     b'z^2 + d'
676 			 */
677 			d = a_pow(bch, 2*l)^gf_mul(bch, b, f)^d;
678 			b = gf_mul(bch, a, e)^b;
679 		}
680 		/* now, use Y=1/X to get Y^4 + b/dY^2 + a/dY + 1/d */
681 		if (d == 0)
682 			/* assume all roots have multiplicity 1 */
683 			return 0;
684 
685 		c2 = gf_inv(bch, d);
686 		b2 = gf_div(bch, a, d);
687 		a2 = gf_div(bch, b, d);
688 	} else {
689 		/* polynomial is already affine */
690 		c2 = d;
691 		b2 = c;
692 		a2 = b;
693 	}
694 	/* find the 4 roots of this affine polynomial */
695 	if (find_affine4_roots(bch, a2, b2, c2, roots) == 4) {
696 		for (i = 0; i < 4; i++) {
697 			/* post-process roots (reverse transformations) */
698 			f = a ? gf_inv(bch, roots[i]) : roots[i];
699 			roots[i] = a_ilog(bch, f^e);
700 		}
701 		n = 4;
702 	}
703 	return n;
704 }
705 
706 /*
707  * build monic, log-based representation of a polynomial
708  */
709 static void gf_poly_logrep(struct bch_control *bch,
710 			   const struct gf_poly *a, int *rep)
711 {
712 	int i, d = a->deg, l = GF_N(bch)-a_log(bch, a->c[a->deg]);
713 
714 	/* represent 0 values with -1; warning, rep[d] is not set to 1 */
715 	for (i = 0; i < d; i++)
716 		rep[i] = a->c[i] ? mod_s(bch, a_log(bch, a->c[i])+l) : -1;
717 }
718 
719 /*
720  * compute polynomial Euclidean division remainder in GF(2^m)[X]
721  */
722 static void gf_poly_mod(struct bch_control *bch, struct gf_poly *a,
723 			const struct gf_poly *b, int *rep)
724 {
725 	int la, p, m;
726 	unsigned int i, j, *c = a->c;
727 	const unsigned int d = b->deg;
728 
729 	if (a->deg < d)
730 		return;
731 
732 	/* reuse or compute log representation of denominator */
733 	if (!rep) {
734 		rep = bch->cache;
735 		gf_poly_logrep(bch, b, rep);
736 	}
737 
738 	for (j = a->deg; j >= d; j--) {
739 		if (c[j]) {
740 			la = a_log(bch, c[j]);
741 			p = j-d;
742 			for (i = 0; i < d; i++, p++) {
743 				m = rep[i];
744 				if (m >= 0)
745 					c[p] ^= bch->a_pow_tab[mod_s(bch,
746 								     m+la)];
747 			}
748 		}
749 	}
750 	a->deg = d-1;
751 	while (!c[a->deg] && a->deg)
752 		a->deg--;
753 }
754 
755 /*
756  * compute polynomial Euclidean division quotient in GF(2^m)[X]
757  */
758 static void gf_poly_div(struct bch_control *bch, struct gf_poly *a,
759 			const struct gf_poly *b, struct gf_poly *q)
760 {
761 	if (a->deg >= b->deg) {
762 		q->deg = a->deg-b->deg;
763 		/* compute a mod b (modifies a) */
764 		gf_poly_mod(bch, a, b, NULL);
765 		/* quotient is stored in upper part of polynomial a */
766 		memcpy(q->c, &a->c[b->deg], (1+q->deg)*sizeof(unsigned int));
767 	} else {
768 		q->deg = 0;
769 		q->c[0] = 0;
770 	}
771 }
772 
773 /*
774  * compute polynomial GCD (Greatest Common Divisor) in GF(2^m)[X]
775  */
776 static struct gf_poly *gf_poly_gcd(struct bch_control *bch, struct gf_poly *a,
777 				   struct gf_poly *b)
778 {
779 	struct gf_poly *tmp;
780 
781 	dbg("gcd(%s,%s)=", gf_poly_str(a), gf_poly_str(b));
782 
783 	if (a->deg < b->deg) {
784 		tmp = b;
785 		b = a;
786 		a = tmp;
787 	}
788 
789 	while (b->deg > 0) {
790 		gf_poly_mod(bch, a, b, NULL);
791 		tmp = b;
792 		b = a;
793 		a = tmp;
794 	}
795 
796 	dbg("%s\n", gf_poly_str(a));
797 
798 	return a;
799 }
800 
801 /*
802  * Given a polynomial f and an integer k, compute Tr(a^kX) mod f
803  * This is used in Berlekamp Trace algorithm for splitting polynomials
804  */
805 static void compute_trace_bk_mod(struct bch_control *bch, int k,
806 				 const struct gf_poly *f, struct gf_poly *z,
807 				 struct gf_poly *out)
808 {
809 	const int m = GF_M(bch);
810 	int i, j;
811 
812 	/* z contains z^2j mod f */
813 	z->deg = 1;
814 	z->c[0] = 0;
815 	z->c[1] = bch->a_pow_tab[k];
816 
817 	out->deg = 0;
818 	memset(out, 0, GF_POLY_SZ(f->deg));
819 
820 	/* compute f log representation only once */
821 	gf_poly_logrep(bch, f, bch->cache);
822 
823 	for (i = 0; i < m; i++) {
824 		/* add a^(k*2^i)(z^(2^i) mod f) and compute (z^(2^i) mod f)^2 */
825 		for (j = z->deg; j >= 0; j--) {
826 			out->c[j] ^= z->c[j];
827 			z->c[2*j] = gf_sqr(bch, z->c[j]);
828 			z->c[2*j+1] = 0;
829 		}
830 		if (z->deg > out->deg)
831 			out->deg = z->deg;
832 
833 		if (i < m-1) {
834 			z->deg *= 2;
835 			/* z^(2(i+1)) mod f = (z^(2^i) mod f)^2 mod f */
836 			gf_poly_mod(bch, z, f, bch->cache);
837 		}
838 	}
839 	while (!out->c[out->deg] && out->deg)
840 		out->deg--;
841 
842 	dbg("Tr(a^%d.X) mod f = %s\n", k, gf_poly_str(out));
843 }
844 
845 /*
846  * factor a polynomial using Berlekamp Trace algorithm (BTA)
847  */
848 static void factor_polynomial(struct bch_control *bch, int k, struct gf_poly *f,
849 			      struct gf_poly **g, struct gf_poly **h)
850 {
851 	struct gf_poly *f2 = bch->poly_2t[0];
852 	struct gf_poly *q  = bch->poly_2t[1];
853 	struct gf_poly *tk = bch->poly_2t[2];
854 	struct gf_poly *z  = bch->poly_2t[3];
855 	struct gf_poly *gcd;
856 
857 	dbg("factoring %s...\n", gf_poly_str(f));
858 
859 	*g = f;
860 	*h = NULL;
861 
862 	/* tk = Tr(a^k.X) mod f */
863 	compute_trace_bk_mod(bch, k, f, z, tk);
864 
865 	if (tk->deg > 0) {
866 		/* compute g = gcd(f, tk) (destructive operation) */
867 		gf_poly_copy(f2, f);
868 		gcd = gf_poly_gcd(bch, f2, tk);
869 		if (gcd->deg < f->deg) {
870 			/* compute h=f/gcd(f,tk); this will modify f and q */
871 			gf_poly_div(bch, f, gcd, q);
872 			/* store g and h in-place (clobbering f) */
873 			*h = &((struct gf_poly_deg1 *)f)[gcd->deg].poly;
874 			gf_poly_copy(*g, gcd);
875 			gf_poly_copy(*h, q);
876 		}
877 	}
878 }
879 
880 /*
881  * find roots of a polynomial, using BTZ algorithm; see the beginning of this
882  * file for details
883  */
884 static int find_poly_roots(struct bch_control *bch, unsigned int k,
885 			   struct gf_poly *poly, unsigned int *roots)
886 {
887 	int cnt;
888 	struct gf_poly *f1, *f2;
889 
890 	switch (poly->deg) {
891 		/* handle low degree polynomials with ad hoc techniques */
892 	case 1:
893 		cnt = find_poly_deg1_roots(bch, poly, roots);
894 		break;
895 	case 2:
896 		cnt = find_poly_deg2_roots(bch, poly, roots);
897 		break;
898 	case 3:
899 		cnt = find_poly_deg3_roots(bch, poly, roots);
900 		break;
901 	case 4:
902 		cnt = find_poly_deg4_roots(bch, poly, roots);
903 		break;
904 	default:
905 		/* factor polynomial using Berlekamp Trace Algorithm (BTA) */
906 		cnt = 0;
907 		if (poly->deg && (k <= GF_M(bch))) {
908 			factor_polynomial(bch, k, poly, &f1, &f2);
909 			if (f1)
910 				cnt += find_poly_roots(bch, k+1, f1, roots);
911 			if (f2)
912 				cnt += find_poly_roots(bch, k+1, f2, roots+cnt);
913 		}
914 		break;
915 	}
916 	return cnt;
917 }
918 
919 #if defined(USE_CHIEN_SEARCH)
920 /*
921  * exhaustive root search (Chien) implementation - not used, included only for
922  * reference/comparison tests
923  */
924 static int chien_search(struct bch_control *bch, unsigned int len,
925 			struct gf_poly *p, unsigned int *roots)
926 {
927 	int m;
928 	unsigned int i, j, syn, syn0, count = 0;
929 	const unsigned int k = 8*len+bch->ecc_bits;
930 
931 	/* use a log-based representation of polynomial */
932 	gf_poly_logrep(bch, p, bch->cache);
933 	bch->cache[p->deg] = 0;
934 	syn0 = gf_div(bch, p->c[0], p->c[p->deg]);
935 
936 	for (i = GF_N(bch)-k+1; i <= GF_N(bch); i++) {
937 		/* compute elp(a^i) */
938 		for (j = 1, syn = syn0; j <= p->deg; j++) {
939 			m = bch->cache[j];
940 			if (m >= 0)
941 				syn ^= a_pow(bch, m+j*i);
942 		}
943 		if (syn == 0) {
944 			roots[count++] = GF_N(bch)-i;
945 			if (count == p->deg)
946 				break;
947 		}
948 	}
949 	return (count == p->deg) ? count : 0;
950 }
951 #define find_poly_roots(_p, _k, _elp, _loc) chien_search(_p, len, _elp, _loc)
952 #endif /* USE_CHIEN_SEARCH */
953 
954 /**
955  * decode_bch - decode received codeword and find bit error locations
956  * @bch:      BCH control structure
957  * @data:     received data, ignored if @calc_ecc is provided
958  * @len:      data length in bytes, must always be provided
959  * @recv_ecc: received ecc, if NULL then assume it was XORed in @calc_ecc
960  * @calc_ecc: calculated ecc, if NULL then calc_ecc is computed from @data
961  * @syn:      hw computed syndrome data (if NULL, syndrome is calculated)
962  * @errloc:   output array of error locations
963  *
964  * Returns:
965  *  The number of errors found, or -EBADMSG if decoding failed, or -EINVAL if
966  *  invalid parameters were provided
967  *
968  * Depending on the available hw BCH support and the need to compute @calc_ecc
969  * separately (using encode_bch()), this function should be called with one of
970  * the following parameter configurations -
971  *
972  * by providing @data and @recv_ecc only:
973  *   decode_bch(@bch, @data, @len, @recv_ecc, NULL, NULL, @errloc)
974  *
975  * by providing @recv_ecc and @calc_ecc:
976  *   decode_bch(@bch, NULL, @len, @recv_ecc, @calc_ecc, NULL, @errloc)
977  *
978  * by providing ecc = recv_ecc XOR calc_ecc:
979  *   decode_bch(@bch, NULL, @len, NULL, ecc, NULL, @errloc)
980  *
981  * by providing syndrome results @syn:
982  *   decode_bch(@bch, NULL, @len, NULL, NULL, @syn, @errloc)
983  *
984  * Once decode_bch() has successfully returned with a positive value, error
985  * locations returned in array @errloc should be interpreted as follows -
986  *
987  * if (errloc[n] >= 8*len), then n-th error is located in ecc (no need for
988  * data correction)
989  *
990  * if (errloc[n] < 8*len), then n-th error is located in data and can be
991  * corrected with statement data[errloc[n]/8] ^= 1 << (errloc[n] % 8);
992  *
993  * Note that this function does not perform any data correction by itself, it
994  * merely indicates error locations.
995  */
996 int decode_bch(struct bch_control *bch, const uint8_t *data, unsigned int len,
997 	       const uint8_t *recv_ecc, const uint8_t *calc_ecc,
998 	       const unsigned int *syn, unsigned int *errloc)
999 {
1000 	const unsigned int ecc_words = BCH_ECC_WORDS(bch);
1001 	unsigned int nbits;
1002 	int i, err, nroots;
1003 	uint32_t sum;
1004 
1005 	/* sanity check: make sure data length can be handled */
1006 	if (8*len > (bch->n-bch->ecc_bits))
1007 		return -EINVAL;
1008 
1009 	/* if caller does not provide syndromes, compute them */
1010 	if (!syn) {
1011 		if (!calc_ecc) {
1012 			/* compute received data ecc into an internal buffer */
1013 			if (!data || !recv_ecc)
1014 				return -EINVAL;
1015 			encode_bch(bch, data, len, NULL);
1016 		} else {
1017 			/* load provided calculated ecc */
1018 			load_ecc8(bch, bch->ecc_buf, calc_ecc);
1019 		}
1020 		/* load received ecc or assume it was XORed in calc_ecc */
1021 		if (recv_ecc) {
1022 			load_ecc8(bch, bch->ecc_buf2, recv_ecc);
1023 			/* XOR received and calculated ecc */
1024 			for (i = 0, sum = 0; i < (int)ecc_words; i++) {
1025 				bch->ecc_buf[i] ^= bch->ecc_buf2[i];
1026 				sum |= bch->ecc_buf[i];
1027 			}
1028 			if (!sum)
1029 				/* no error found */
1030 				return 0;
1031 		}
1032 		compute_syndromes(bch, bch->ecc_buf, bch->syn);
1033 		syn = bch->syn;
1034 	}
1035 
1036 	err = compute_error_locator_polynomial(bch, syn);
1037 	if (err > 0) {
1038 		nroots = find_poly_roots(bch, 1, bch->elp, errloc);
1039 		if (err != nroots)
1040 			err = -1;
1041 	}
1042 	if (err > 0) {
1043 		/* post-process raw error locations for easier correction */
1044 		nbits = (len*8)+bch->ecc_bits;
1045 		for (i = 0; i < err; i++) {
1046 			if (errloc[i] >= nbits) {
1047 				err = -1;
1048 				break;
1049 			}
1050 			errloc[i] = nbits-1-errloc[i];
1051 			errloc[i] = (errloc[i] & ~7)|(7-(errloc[i] & 7));
1052 		}
1053 	}
1054 	return (err >= 0) ? err : -EBADMSG;
1055 }
1056 EXPORT_SYMBOL_GPL(decode_bch);
1057 
1058 /*
1059  * generate Galois field lookup tables
1060  */
1061 static int build_gf_tables(struct bch_control *bch, unsigned int poly)
1062 {
1063 	unsigned int i, x = 1;
1064 	const unsigned int k = 1 << deg(poly);
1065 
1066 	/* primitive polynomial must be of degree m */
1067 	if (k != (1u << GF_M(bch)))
1068 		return -1;
1069 
1070 	for (i = 0; i < GF_N(bch); i++) {
1071 		bch->a_pow_tab[i] = x;
1072 		bch->a_log_tab[x] = i;
1073 		if (i && (x == 1))
1074 			/* polynomial is not primitive (a^i=1 with 0<i<2^m-1) */
1075 			return -1;
1076 		x <<= 1;
1077 		if (x & k)
1078 			x ^= poly;
1079 	}
1080 	bch->a_pow_tab[GF_N(bch)] = 1;
1081 	bch->a_log_tab[0] = 0;
1082 
1083 	return 0;
1084 }
1085 
1086 /*
1087  * compute generator polynomial remainder tables for fast encoding
1088  */
1089 static void build_mod8_tables(struct bch_control *bch, const uint32_t *g)
1090 {
1091 	int i, j, b, d;
1092 	uint32_t data, hi, lo, *tab;
1093 	const int l = BCH_ECC_WORDS(bch);
1094 	const int plen = DIV_ROUND_UP(bch->ecc_bits+1, 32);
1095 	const int ecclen = DIV_ROUND_UP(bch->ecc_bits, 32);
1096 
1097 	memset(bch->mod8_tab, 0, 4*256*l*sizeof(*bch->mod8_tab));
1098 
1099 	for (i = 0; i < 256; i++) {
1100 		/* p(X)=i is a small polynomial of weight <= 8 */
1101 		for (b = 0; b < 4; b++) {
1102 			/* we want to compute (p(X).X^(8*b+deg(g))) mod g(X) */
1103 			tab = bch->mod8_tab + (b*256+i)*l;
1104 			data = i << (8*b);
1105 			while (data) {
1106 				d = deg(data);
1107 				/* subtract X^d.g(X) from p(X).X^(8*b+deg(g)) */
1108 				data ^= g[0] >> (31-d);
1109 				for (j = 0; j < ecclen; j++) {
1110 					hi = (d < 31) ? g[j] << (d+1) : 0;
1111 					lo = (j+1 < plen) ?
1112 						g[j+1] >> (31-d) : 0;
1113 					tab[j] ^= hi|lo;
1114 				}
1115 			}
1116 		}
1117 	}
1118 }
1119 
1120 /*
1121  * build a base for factoring degree 2 polynomials
1122  */
1123 static int build_deg2_base(struct bch_control *bch)
1124 {
1125 	const int m = GF_M(bch);
1126 	int i, j, r;
1127 	unsigned int sum, x, y, remaining, ak = 0, xi[BCH_MAX_M];
1128 
1129 	/* find k s.t. Tr(a^k) = 1 and 0 <= k < m */
1130 	for (i = 0; i < m; i++) {
1131 		for (j = 0, sum = 0; j < m; j++)
1132 			sum ^= a_pow(bch, i*(1 << j));
1133 
1134 		if (sum) {
1135 			ak = bch->a_pow_tab[i];
1136 			break;
1137 		}
1138 	}
1139 	/* find xi, i=0..m-1 such that xi^2+xi = a^i+Tr(a^i).a^k */
1140 	remaining = m;
1141 	memset(xi, 0, sizeof(xi));
1142 
1143 	for (x = 0; (x <= GF_N(bch)) && remaining; x++) {
1144 		y = gf_sqr(bch, x)^x;
1145 		for (i = 0; i < 2; i++) {
1146 			r = a_log(bch, y);
1147 			if (y && (r < m) && !xi[r]) {
1148 				bch->xi_tab[r] = x;
1149 				xi[r] = 1;
1150 				remaining--;
1151 				dbg("x%d = %x\n", r, x);
1152 				break;
1153 			}
1154 			y ^= ak;
1155 		}
1156 	}
1157 	/* should not happen but check anyway */
1158 	return remaining ? -1 : 0;
1159 }
1160 
1161 static void *bch_alloc(size_t size, int *err)
1162 {
1163 	void *ptr;
1164 
1165 	ptr = kmalloc(size, GFP_KERNEL);
1166 	if (ptr == NULL)
1167 		*err = 1;
1168 	return ptr;
1169 }
1170 
1171 /*
1172  * compute generator polynomial for given (m,t) parameters.
1173  */
1174 static uint32_t *compute_generator_polynomial(struct bch_control *bch)
1175 {
1176 	const unsigned int m = GF_M(bch);
1177 	const unsigned int t = GF_T(bch);
1178 	int n, err = 0;
1179 	unsigned int i, j, nbits, r, word, *roots;
1180 	struct gf_poly *g;
1181 	uint32_t *genpoly;
1182 
1183 	g = bch_alloc(GF_POLY_SZ(m*t), &err);
1184 	roots = bch_alloc((bch->n+1)*sizeof(*roots), &err);
1185 	genpoly = bch_alloc(DIV_ROUND_UP(m*t+1, 32)*sizeof(*genpoly), &err);
1186 
1187 	if (err) {
1188 		kfree(genpoly);
1189 		genpoly = NULL;
1190 		goto finish;
1191 	}
1192 
1193 	/* enumerate all roots of g(X) */
1194 	memset(roots , 0, (bch->n+1)*sizeof(*roots));
1195 	for (i = 0; i < t; i++) {
1196 		for (j = 0, r = 2*i+1; j < m; j++) {
1197 			roots[r] = 1;
1198 			r = mod_s(bch, 2*r);
1199 		}
1200 	}
1201 	/* build generator polynomial g(X) */
1202 	g->deg = 0;
1203 	g->c[0] = 1;
1204 	for (i = 0; i < GF_N(bch); i++) {
1205 		if (roots[i]) {
1206 			/* multiply g(X) by (X+root) */
1207 			r = bch->a_pow_tab[i];
1208 			g->c[g->deg+1] = 1;
1209 			for (j = g->deg; j > 0; j--)
1210 				g->c[j] = gf_mul(bch, g->c[j], r)^g->c[j-1];
1211 
1212 			g->c[0] = gf_mul(bch, g->c[0], r);
1213 			g->deg++;
1214 		}
1215 	}
1216 	/* store left-justified binary representation of g(X) */
1217 	n = g->deg+1;
1218 	i = 0;
1219 
1220 	while (n > 0) {
1221 		nbits = (n > 32) ? 32 : n;
1222 		for (j = 0, word = 0; j < nbits; j++) {
1223 			if (g->c[n-1-j])
1224 				word |= 1u << (31-j);
1225 		}
1226 		genpoly[i++] = word;
1227 		n -= nbits;
1228 	}
1229 	bch->ecc_bits = g->deg;
1230 
1231 finish:
1232 	kfree(g);
1233 	kfree(roots);
1234 
1235 	return genpoly;
1236 }
1237 
1238 /**
1239  * init_bch - initialize a BCH encoder/decoder
1240  * @m:          Galois field order, should be in the range 5-15
1241  * @t:          maximum error correction capability, in bits
1242  * @prim_poly:  user-provided primitive polynomial (or 0 to use default)
1243  *
1244  * Returns:
1245  *  a newly allocated BCH control structure if successful, NULL otherwise
1246  *
1247  * This initialization can take some time, as lookup tables are built for fast
1248  * encoding/decoding; make sure not to call this function from a time critical
1249  * path. Usually, init_bch() should be called on module/driver init and
1250  * free_bch() should be called to release memory on exit.
1251  *
1252  * You may provide your own primitive polynomial of degree @m in argument
1253  * @prim_poly, or let init_bch() use its default polynomial.
1254  *
1255  * Once init_bch() has successfully returned a pointer to a newly allocated
1256  * BCH control structure, ecc length in bytes is given by member @ecc_bytes of
1257  * the structure.
1258  */
1259 struct bch_control *init_bch(int m, int t, unsigned int prim_poly)
1260 {
1261 	int err = 0;
1262 	unsigned int i, words;
1263 	uint32_t *genpoly;
1264 	struct bch_control *bch = NULL;
1265 
1266 	const int min_m = 5;
1267 
1268 	/* default primitive polynomials */
1269 	static const unsigned int prim_poly_tab[] = {
1270 		0x25, 0x43, 0x83, 0x11d, 0x211, 0x409, 0x805, 0x1053, 0x201b,
1271 		0x402b, 0x8003,
1272 	};
1273 
1274 #if defined(CONFIG_BCH_CONST_PARAMS)
1275 	if ((m != (CONFIG_BCH_CONST_M)) || (t != (CONFIG_BCH_CONST_T))) {
1276 		printk(KERN_ERR "bch encoder/decoder was configured to support "
1277 		       "parameters m=%d, t=%d only!\n",
1278 		       CONFIG_BCH_CONST_M, CONFIG_BCH_CONST_T);
1279 		goto fail;
1280 	}
1281 #endif
1282 	if ((m < min_m) || (m > BCH_MAX_M))
1283 		/*
1284 		 * values of m greater than 15 are not currently supported;
1285 		 * supporting m > 15 would require changing table base type
1286 		 * (uint16_t) and a small patch in matrix transposition
1287 		 */
1288 		goto fail;
1289 
1290 	if (t > BCH_MAX_T)
1291 		/*
1292 		 * we can support larger than 64 bits if necessary, at the
1293 		 * cost of higher stack usage.
1294 		 */
1295 		goto fail;
1296 
1297 	/* sanity checks */
1298 	if ((t < 1) || (m*t >= ((1 << m)-1)))
1299 		/* invalid t value */
1300 		goto fail;
1301 
1302 	/* select a primitive polynomial for generating GF(2^m) */
1303 	if (prim_poly == 0)
1304 		prim_poly = prim_poly_tab[m-min_m];
1305 
1306 	bch = kzalloc(sizeof(*bch), GFP_KERNEL);
1307 	if (bch == NULL)
1308 		goto fail;
1309 
1310 	bch->m = m;
1311 	bch->t = t;
1312 	bch->n = (1 << m)-1;
1313 	words  = DIV_ROUND_UP(m*t, 32);
1314 	bch->ecc_bytes = DIV_ROUND_UP(m*t, 8);
1315 	bch->a_pow_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_pow_tab), &err);
1316 	bch->a_log_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_log_tab), &err);
1317 	bch->mod8_tab  = bch_alloc(words*1024*sizeof(*bch->mod8_tab), &err);
1318 	bch->ecc_buf   = bch_alloc(words*sizeof(*bch->ecc_buf), &err);
1319 	bch->ecc_buf2  = bch_alloc(words*sizeof(*bch->ecc_buf2), &err);
1320 	bch->xi_tab    = bch_alloc(m*sizeof(*bch->xi_tab), &err);
1321 	bch->syn       = bch_alloc(2*t*sizeof(*bch->syn), &err);
1322 	bch->cache     = bch_alloc(2*t*sizeof(*bch->cache), &err);
1323 	bch->elp       = bch_alloc((t+1)*sizeof(struct gf_poly_deg1), &err);
1324 
1325 	for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++)
1326 		bch->poly_2t[i] = bch_alloc(GF_POLY_SZ(2*t), &err);
1327 
1328 	if (err)
1329 		goto fail;
1330 
1331 	err = build_gf_tables(bch, prim_poly);
1332 	if (err)
1333 		goto fail;
1334 
1335 	/* use generator polynomial for computing encoding tables */
1336 	genpoly = compute_generator_polynomial(bch);
1337 	if (genpoly == NULL)
1338 		goto fail;
1339 
1340 	build_mod8_tables(bch, genpoly);
1341 	kfree(genpoly);
1342 
1343 	err = build_deg2_base(bch);
1344 	if (err)
1345 		goto fail;
1346 
1347 	return bch;
1348 
1349 fail:
1350 	free_bch(bch);
1351 	return NULL;
1352 }
1353 EXPORT_SYMBOL_GPL(init_bch);
1354 
1355 /**
1356  *  free_bch - free the BCH control structure
1357  *  @bch:    BCH control structure to release
1358  */
1359 void free_bch(struct bch_control *bch)
1360 {
1361 	unsigned int i;
1362 
1363 	if (bch) {
1364 		kfree(bch->a_pow_tab);
1365 		kfree(bch->a_log_tab);
1366 		kfree(bch->mod8_tab);
1367 		kfree(bch->ecc_buf);
1368 		kfree(bch->ecc_buf2);
1369 		kfree(bch->xi_tab);
1370 		kfree(bch->syn);
1371 		kfree(bch->cache);
1372 		kfree(bch->elp);
1373 
1374 		for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++)
1375 			kfree(bch->poly_2t[i]);
1376 
1377 		kfree(bch);
1378 	}
1379 }
1380 EXPORT_SYMBOL_GPL(free_bch);
1381 
1382 MODULE_LICENSE("GPL");
1383 MODULE_AUTHOR("Ivan Djelic <ivan.djelic@parrot.com>");
1384 MODULE_DESCRIPTION("Binary BCH encoder/decoder");
1385