xref: /openbmc/linux/lib/bch.c (revision 95e9fd10)
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 #else
82 #define GF_M(_p)               ((_p)->m)
83 #define GF_T(_p)               ((_p)->t)
84 #define GF_N(_p)               ((_p)->n)
85 #endif
86 
87 #define BCH_ECC_WORDS(_p)      DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 32)
88 #define BCH_ECC_BYTES(_p)      DIV_ROUND_UP(GF_M(_p)*GF_T(_p), 8)
89 
90 #ifndef dbg
91 #define dbg(_fmt, args...)     do {} while (0)
92 #endif
93 
94 /*
95  * represent a polynomial over GF(2^m)
96  */
97 struct gf_poly {
98 	unsigned int deg;    /* polynomial degree */
99 	unsigned int c[0];   /* polynomial terms */
100 };
101 
102 /* given its degree, compute a polynomial size in bytes */
103 #define GF_POLY_SZ(_d) (sizeof(struct gf_poly)+((_d)+1)*sizeof(unsigned int))
104 
105 /* polynomial of degree 1 */
106 struct gf_poly_deg1 {
107 	struct gf_poly poly;
108 	unsigned int   c[2];
109 };
110 
111 /*
112  * same as encode_bch(), but process input data one byte at a time
113  */
114 static void encode_bch_unaligned(struct bch_control *bch,
115 				 const unsigned char *data, unsigned int len,
116 				 uint32_t *ecc)
117 {
118 	int i;
119 	const uint32_t *p;
120 	const int l = BCH_ECC_WORDS(bch)-1;
121 
122 	while (len--) {
123 		p = bch->mod8_tab + (l+1)*(((ecc[0] >> 24)^(*data++)) & 0xff);
124 
125 		for (i = 0; i < l; i++)
126 			ecc[i] = ((ecc[i] << 8)|(ecc[i+1] >> 24))^(*p++);
127 
128 		ecc[l] = (ecc[l] << 8)^(*p);
129 	}
130 }
131 
132 /*
133  * convert ecc bytes to aligned, zero-padded 32-bit ecc words
134  */
135 static void load_ecc8(struct bch_control *bch, uint32_t *dst,
136 		      const uint8_t *src)
137 {
138 	uint8_t pad[4] = {0, 0, 0, 0};
139 	unsigned int i, nwords = BCH_ECC_WORDS(bch)-1;
140 
141 	for (i = 0; i < nwords; i++, src += 4)
142 		dst[i] = (src[0] << 24)|(src[1] << 16)|(src[2] << 8)|src[3];
143 
144 	memcpy(pad, src, BCH_ECC_BYTES(bch)-4*nwords);
145 	dst[nwords] = (pad[0] << 24)|(pad[1] << 16)|(pad[2] << 8)|pad[3];
146 }
147 
148 /*
149  * convert 32-bit ecc words to ecc bytes
150  */
151 static void store_ecc8(struct bch_control *bch, uint8_t *dst,
152 		       const uint32_t *src)
153 {
154 	uint8_t pad[4];
155 	unsigned int i, nwords = BCH_ECC_WORDS(bch)-1;
156 
157 	for (i = 0; i < nwords; i++) {
158 		*dst++ = (src[i] >> 24);
159 		*dst++ = (src[i] >> 16) & 0xff;
160 		*dst++ = (src[i] >>  8) & 0xff;
161 		*dst++ = (src[i] >>  0) & 0xff;
162 	}
163 	pad[0] = (src[nwords] >> 24);
164 	pad[1] = (src[nwords] >> 16) & 0xff;
165 	pad[2] = (src[nwords] >>  8) & 0xff;
166 	pad[3] = (src[nwords] >>  0) & 0xff;
167 	memcpy(dst, pad, BCH_ECC_BYTES(bch)-4*nwords);
168 }
169 
170 /**
171  * encode_bch - calculate BCH ecc parity of data
172  * @bch:   BCH control structure
173  * @data:  data to encode
174  * @len:   data length in bytes
175  * @ecc:   ecc parity data, must be initialized by caller
176  *
177  * The @ecc parity array is used both as input and output parameter, in order to
178  * allow incremental computations. It should be of the size indicated by member
179  * @ecc_bytes of @bch, and should be initialized to 0 before the first call.
180  *
181  * The exact number of computed ecc parity bits is given by member @ecc_bits of
182  * @bch; it may be less than m*t for large values of t.
183  */
184 void encode_bch(struct bch_control *bch, const uint8_t *data,
185 		unsigned int len, uint8_t *ecc)
186 {
187 	const unsigned int l = BCH_ECC_WORDS(bch)-1;
188 	unsigned int i, mlen;
189 	unsigned long m;
190 	uint32_t w, r[l+1];
191 	const uint32_t * const tab0 = bch->mod8_tab;
192 	const uint32_t * const tab1 = tab0 + 256*(l+1);
193 	const uint32_t * const tab2 = tab1 + 256*(l+1);
194 	const uint32_t * const tab3 = tab2 + 256*(l+1);
195 	const uint32_t *pdata, *p0, *p1, *p2, *p3;
196 
197 	if (ecc) {
198 		/* load ecc parity bytes into internal 32-bit buffer */
199 		load_ecc8(bch, bch->ecc_buf, ecc);
200 	} else {
201 		memset(bch->ecc_buf, 0, sizeof(r));
202 	}
203 
204 	/* process first unaligned data bytes */
205 	m = ((unsigned long)data) & 3;
206 	if (m) {
207 		mlen = (len < (4-m)) ? len : 4-m;
208 		encode_bch_unaligned(bch, data, mlen, bch->ecc_buf);
209 		data += mlen;
210 		len  -= mlen;
211 	}
212 
213 	/* process 32-bit aligned data words */
214 	pdata = (uint32_t *)data;
215 	mlen  = len/4;
216 	data += 4*mlen;
217 	len  -= 4*mlen;
218 	memcpy(r, bch->ecc_buf, sizeof(r));
219 
220 	/*
221 	 * split each 32-bit word into 4 polynomials of weight 8 as follows:
222 	 *
223 	 * 31 ...24  23 ...16  15 ... 8  7 ... 0
224 	 * xxxxxxxx  yyyyyyyy  zzzzzzzz  tttttttt
225 	 *                               tttttttt  mod g = r0 (precomputed)
226 	 *                     zzzzzzzz  00000000  mod g = r1 (precomputed)
227 	 *           yyyyyyyy  00000000  00000000  mod g = r2 (precomputed)
228 	 * xxxxxxxx  00000000  00000000  00000000  mod g = r3 (precomputed)
229 	 * xxxxxxxx  yyyyyyyy  zzzzzzzz  tttttttt  mod g = r0^r1^r2^r3
230 	 */
231 	while (mlen--) {
232 		/* input data is read in big-endian format */
233 		w = r[0]^cpu_to_be32(*pdata++);
234 		p0 = tab0 + (l+1)*((w >>  0) & 0xff);
235 		p1 = tab1 + (l+1)*((w >>  8) & 0xff);
236 		p2 = tab2 + (l+1)*((w >> 16) & 0xff);
237 		p3 = tab3 + (l+1)*((w >> 24) & 0xff);
238 
239 		for (i = 0; i < l; i++)
240 			r[i] = r[i+1]^p0[i]^p1[i]^p2[i]^p3[i];
241 
242 		r[l] = p0[l]^p1[l]^p2[l]^p3[l];
243 	}
244 	memcpy(bch->ecc_buf, r, sizeof(r));
245 
246 	/* process last unaligned bytes */
247 	if (len)
248 		encode_bch_unaligned(bch, data, len, bch->ecc_buf);
249 
250 	/* store ecc parity bytes into original parity buffer */
251 	if (ecc)
252 		store_ecc8(bch, ecc, bch->ecc_buf);
253 }
254 EXPORT_SYMBOL_GPL(encode_bch);
255 
256 static inline int modulo(struct bch_control *bch, unsigned int v)
257 {
258 	const unsigned int n = GF_N(bch);
259 	while (v >= n) {
260 		v -= n;
261 		v = (v & n) + (v >> GF_M(bch));
262 	}
263 	return v;
264 }
265 
266 /*
267  * shorter and faster modulo function, only works when v < 2N.
268  */
269 static inline int mod_s(struct bch_control *bch, unsigned int v)
270 {
271 	const unsigned int n = GF_N(bch);
272 	return (v < n) ? v : v-n;
273 }
274 
275 static inline int deg(unsigned int poly)
276 {
277 	/* polynomial degree is the most-significant bit index */
278 	return fls(poly)-1;
279 }
280 
281 static inline int parity(unsigned int x)
282 {
283 	/*
284 	 * public domain code snippet, lifted from
285 	 * http://www-graphics.stanford.edu/~seander/bithacks.html
286 	 */
287 	x ^= x >> 1;
288 	x ^= x >> 2;
289 	x = (x & 0x11111111U) * 0x11111111U;
290 	return (x >> 28) & 1;
291 }
292 
293 /* Galois field basic operations: multiply, divide, inverse, etc. */
294 
295 static inline unsigned int gf_mul(struct bch_control *bch, unsigned int a,
296 				  unsigned int b)
297 {
298 	return (a && b) ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+
299 					       bch->a_log_tab[b])] : 0;
300 }
301 
302 static inline unsigned int gf_sqr(struct bch_control *bch, unsigned int a)
303 {
304 	return a ? bch->a_pow_tab[mod_s(bch, 2*bch->a_log_tab[a])] : 0;
305 }
306 
307 static inline unsigned int gf_div(struct bch_control *bch, unsigned int a,
308 				  unsigned int b)
309 {
310 	return a ? bch->a_pow_tab[mod_s(bch, bch->a_log_tab[a]+
311 					GF_N(bch)-bch->a_log_tab[b])] : 0;
312 }
313 
314 static inline unsigned int gf_inv(struct bch_control *bch, unsigned int a)
315 {
316 	return bch->a_pow_tab[GF_N(bch)-bch->a_log_tab[a]];
317 }
318 
319 static inline unsigned int a_pow(struct bch_control *bch, int i)
320 {
321 	return bch->a_pow_tab[modulo(bch, i)];
322 }
323 
324 static inline int a_log(struct bch_control *bch, unsigned int x)
325 {
326 	return bch->a_log_tab[x];
327 }
328 
329 static inline int a_ilog(struct bch_control *bch, unsigned int x)
330 {
331 	return mod_s(bch, GF_N(bch)-bch->a_log_tab[x]);
332 }
333 
334 /*
335  * compute 2t syndromes of ecc polynomial, i.e. ecc(a^j) for j=1..2t
336  */
337 static void compute_syndromes(struct bch_control *bch, uint32_t *ecc,
338 			      unsigned int *syn)
339 {
340 	int i, j, s;
341 	unsigned int m;
342 	uint32_t poly;
343 	const int t = GF_T(bch);
344 
345 	s = bch->ecc_bits;
346 
347 	/* make sure extra bits in last ecc word are cleared */
348 	m = ((unsigned int)s) & 31;
349 	if (m)
350 		ecc[s/32] &= ~((1u << (32-m))-1);
351 	memset(syn, 0, 2*t*sizeof(*syn));
352 
353 	/* compute v(a^j) for j=1 .. 2t-1 */
354 	do {
355 		poly = *ecc++;
356 		s -= 32;
357 		while (poly) {
358 			i = deg(poly);
359 			for (j = 0; j < 2*t; j += 2)
360 				syn[j] ^= a_pow(bch, (j+1)*(i+s));
361 
362 			poly ^= (1 << i);
363 		}
364 	} while (s > 0);
365 
366 	/* v(a^(2j)) = v(a^j)^2 */
367 	for (j = 0; j < t; j++)
368 		syn[2*j+1] = gf_sqr(bch, syn[j]);
369 }
370 
371 static void gf_poly_copy(struct gf_poly *dst, struct gf_poly *src)
372 {
373 	memcpy(dst, src, GF_POLY_SZ(src->deg));
374 }
375 
376 static int compute_error_locator_polynomial(struct bch_control *bch,
377 					    const unsigned int *syn)
378 {
379 	const unsigned int t = GF_T(bch);
380 	const unsigned int n = GF_N(bch);
381 	unsigned int i, j, tmp, l, pd = 1, d = syn[0];
382 	struct gf_poly *elp = bch->elp;
383 	struct gf_poly *pelp = bch->poly_2t[0];
384 	struct gf_poly *elp_copy = bch->poly_2t[1];
385 	int k, pp = -1;
386 
387 	memset(pelp, 0, GF_POLY_SZ(2*t));
388 	memset(elp, 0, GF_POLY_SZ(2*t));
389 
390 	pelp->deg = 0;
391 	pelp->c[0] = 1;
392 	elp->deg = 0;
393 	elp->c[0] = 1;
394 
395 	/* use simplified binary Berlekamp-Massey algorithm */
396 	for (i = 0; (i < t) && (elp->deg <= t); i++) {
397 		if (d) {
398 			k = 2*i-pp;
399 			gf_poly_copy(elp_copy, elp);
400 			/* e[i+1](X) = e[i](X)+di*dp^-1*X^2(i-p)*e[p](X) */
401 			tmp = a_log(bch, d)+n-a_log(bch, pd);
402 			for (j = 0; j <= pelp->deg; j++) {
403 				if (pelp->c[j]) {
404 					l = a_log(bch, pelp->c[j]);
405 					elp->c[j+k] ^= a_pow(bch, tmp+l);
406 				}
407 			}
408 			/* compute l[i+1] = max(l[i]->c[l[p]+2*(i-p]) */
409 			tmp = pelp->deg+k;
410 			if (tmp > elp->deg) {
411 				elp->deg = tmp;
412 				gf_poly_copy(pelp, elp_copy);
413 				pd = d;
414 				pp = 2*i;
415 			}
416 		}
417 		/* di+1 = S(2i+3)+elp[i+1].1*S(2i+2)+...+elp[i+1].lS(2i+3-l) */
418 		if (i < t-1) {
419 			d = syn[2*i+2];
420 			for (j = 1; j <= elp->deg; j++)
421 				d ^= gf_mul(bch, elp->c[j], syn[2*i+2-j]);
422 		}
423 	}
424 	dbg("elp=%s\n", gf_poly_str(elp));
425 	return (elp->deg > t) ? -1 : (int)elp->deg;
426 }
427 
428 /*
429  * solve a m x m linear system in GF(2) with an expected number of solutions,
430  * and return the number of found solutions
431  */
432 static int solve_linear_system(struct bch_control *bch, unsigned int *rows,
433 			       unsigned int *sol, int nsol)
434 {
435 	const int m = GF_M(bch);
436 	unsigned int tmp, mask;
437 	int rem, c, r, p, k, param[m];
438 
439 	k = 0;
440 	mask = 1 << m;
441 
442 	/* Gaussian elimination */
443 	for (c = 0; c < m; c++) {
444 		rem = 0;
445 		p = c-k;
446 		/* find suitable row for elimination */
447 		for (r = p; r < m; r++) {
448 			if (rows[r] & mask) {
449 				if (r != p) {
450 					tmp = rows[r];
451 					rows[r] = rows[p];
452 					rows[p] = tmp;
453 				}
454 				rem = r+1;
455 				break;
456 			}
457 		}
458 		if (rem) {
459 			/* perform elimination on remaining rows */
460 			tmp = rows[p];
461 			for (r = rem; r < m; r++) {
462 				if (rows[r] & mask)
463 					rows[r] ^= tmp;
464 			}
465 		} else {
466 			/* elimination not needed, store defective row index */
467 			param[k++] = c;
468 		}
469 		mask >>= 1;
470 	}
471 	/* rewrite system, inserting fake parameter rows */
472 	if (k > 0) {
473 		p = k;
474 		for (r = m-1; r >= 0; r--) {
475 			if ((r > m-1-k) && rows[r])
476 				/* system has no solution */
477 				return 0;
478 
479 			rows[r] = (p && (r == param[p-1])) ?
480 				p--, 1u << (m-r) : rows[r-p];
481 		}
482 	}
483 
484 	if (nsol != (1 << k))
485 		/* unexpected number of solutions */
486 		return 0;
487 
488 	for (p = 0; p < nsol; p++) {
489 		/* set parameters for p-th solution */
490 		for (c = 0; c < k; c++)
491 			rows[param[c]] = (rows[param[c]] & ~1)|((p >> c) & 1);
492 
493 		/* compute unique solution */
494 		tmp = 0;
495 		for (r = m-1; r >= 0; r--) {
496 			mask = rows[r] & (tmp|1);
497 			tmp |= parity(mask) << (m-r);
498 		}
499 		sol[p] = tmp >> 1;
500 	}
501 	return nsol;
502 }
503 
504 /*
505  * this function builds and solves a linear system for finding roots of a degree
506  * 4 affine monic polynomial X^4+aX^2+bX+c over GF(2^m).
507  */
508 static int find_affine4_roots(struct bch_control *bch, unsigned int a,
509 			      unsigned int b, unsigned int c,
510 			      unsigned int *roots)
511 {
512 	int i, j, k;
513 	const int m = GF_M(bch);
514 	unsigned int mask = 0xff, t, rows[16] = {0,};
515 
516 	j = a_log(bch, b);
517 	k = a_log(bch, a);
518 	rows[0] = c;
519 
520 	/* buid linear system to solve X^4+aX^2+bX+c = 0 */
521 	for (i = 0; i < m; i++) {
522 		rows[i+1] = bch->a_pow_tab[4*i]^
523 			(a ? bch->a_pow_tab[mod_s(bch, k)] : 0)^
524 			(b ? bch->a_pow_tab[mod_s(bch, j)] : 0);
525 		j++;
526 		k += 2;
527 	}
528 	/*
529 	 * transpose 16x16 matrix before passing it to linear solver
530 	 * warning: this code assumes m < 16
531 	 */
532 	for (j = 8; j != 0; j >>= 1, mask ^= (mask << j)) {
533 		for (k = 0; k < 16; k = (k+j+1) & ~j) {
534 			t = ((rows[k] >> j)^rows[k+j]) & mask;
535 			rows[k] ^= (t << j);
536 			rows[k+j] ^= t;
537 		}
538 	}
539 	return solve_linear_system(bch, rows, roots, 4);
540 }
541 
542 /*
543  * compute root r of a degree 1 polynomial over GF(2^m) (returned as log(1/r))
544  */
545 static int find_poly_deg1_roots(struct bch_control *bch, struct gf_poly *poly,
546 				unsigned int *roots)
547 {
548 	int n = 0;
549 
550 	if (poly->c[0])
551 		/* poly[X] = bX+c with c!=0, root=c/b */
552 		roots[n++] = mod_s(bch, GF_N(bch)-bch->a_log_tab[poly->c[0]]+
553 				   bch->a_log_tab[poly->c[1]]);
554 	return n;
555 }
556 
557 /*
558  * compute roots of a degree 2 polynomial over GF(2^m)
559  */
560 static int find_poly_deg2_roots(struct bch_control *bch, struct gf_poly *poly,
561 				unsigned int *roots)
562 {
563 	int n = 0, i, l0, l1, l2;
564 	unsigned int u, v, r;
565 
566 	if (poly->c[0] && poly->c[1]) {
567 
568 		l0 = bch->a_log_tab[poly->c[0]];
569 		l1 = bch->a_log_tab[poly->c[1]];
570 		l2 = bch->a_log_tab[poly->c[2]];
571 
572 		/* using z=a/bX, transform aX^2+bX+c into z^2+z+u (u=ac/b^2) */
573 		u = a_pow(bch, l0+l2+2*(GF_N(bch)-l1));
574 		/*
575 		 * let u = sum(li.a^i) i=0..m-1; then compute r = sum(li.xi):
576 		 * r^2+r = sum(li.(xi^2+xi)) = sum(li.(a^i+Tr(a^i).a^k)) =
577 		 * u + sum(li.Tr(a^i).a^k) = u+a^k.Tr(sum(li.a^i)) = u+a^k.Tr(u)
578 		 * i.e. r and r+1 are roots iff Tr(u)=0
579 		 */
580 		r = 0;
581 		v = u;
582 		while (v) {
583 			i = deg(v);
584 			r ^= bch->xi_tab[i];
585 			v ^= (1 << i);
586 		}
587 		/* verify root */
588 		if ((gf_sqr(bch, r)^r) == u) {
589 			/* reverse z=a/bX transformation and compute log(1/r) */
590 			roots[n++] = modulo(bch, 2*GF_N(bch)-l1-
591 					    bch->a_log_tab[r]+l2);
592 			roots[n++] = modulo(bch, 2*GF_N(bch)-l1-
593 					    bch->a_log_tab[r^1]+l2);
594 		}
595 	}
596 	return n;
597 }
598 
599 /*
600  * compute roots of a degree 3 polynomial over GF(2^m)
601  */
602 static int find_poly_deg3_roots(struct bch_control *bch, struct gf_poly *poly,
603 				unsigned int *roots)
604 {
605 	int i, n = 0;
606 	unsigned int a, b, c, a2, b2, c2, e3, tmp[4];
607 
608 	if (poly->c[0]) {
609 		/* transform polynomial into monic X^3 + a2X^2 + b2X + c2 */
610 		e3 = poly->c[3];
611 		c2 = gf_div(bch, poly->c[0], e3);
612 		b2 = gf_div(bch, poly->c[1], e3);
613 		a2 = gf_div(bch, poly->c[2], e3);
614 
615 		/* (X+a2)(X^3+a2X^2+b2X+c2) = X^4+aX^2+bX+c (affine) */
616 		c = gf_mul(bch, a2, c2);           /* c = a2c2      */
617 		b = gf_mul(bch, a2, b2)^c2;        /* b = a2b2 + c2 */
618 		a = gf_sqr(bch, a2)^b2;            /* a = a2^2 + b2 */
619 
620 		/* find the 4 roots of this affine polynomial */
621 		if (find_affine4_roots(bch, a, b, c, tmp) == 4) {
622 			/* remove a2 from final list of roots */
623 			for (i = 0; i < 4; i++) {
624 				if (tmp[i] != a2)
625 					roots[n++] = a_ilog(bch, tmp[i]);
626 			}
627 		}
628 	}
629 	return n;
630 }
631 
632 /*
633  * compute roots of a degree 4 polynomial over GF(2^m)
634  */
635 static int find_poly_deg4_roots(struct bch_control *bch, struct gf_poly *poly,
636 				unsigned int *roots)
637 {
638 	int i, l, n = 0;
639 	unsigned int a, b, c, d, e = 0, f, a2, b2, c2, e4;
640 
641 	if (poly->c[0] == 0)
642 		return 0;
643 
644 	/* transform polynomial into monic X^4 + aX^3 + bX^2 + cX + d */
645 	e4 = poly->c[4];
646 	d = gf_div(bch, poly->c[0], e4);
647 	c = gf_div(bch, poly->c[1], e4);
648 	b = gf_div(bch, poly->c[2], e4);
649 	a = gf_div(bch, poly->c[3], e4);
650 
651 	/* use Y=1/X transformation to get an affine polynomial */
652 	if (a) {
653 		/* first, eliminate cX by using z=X+e with ae^2+c=0 */
654 		if (c) {
655 			/* compute e such that e^2 = c/a */
656 			f = gf_div(bch, c, a);
657 			l = a_log(bch, f);
658 			l += (l & 1) ? GF_N(bch) : 0;
659 			e = a_pow(bch, l/2);
660 			/*
661 			 * use transformation z=X+e:
662 			 * z^4+e^4 + a(z^3+ez^2+e^2z+e^3) + b(z^2+e^2) +cz+ce+d
663 			 * z^4 + az^3 + (ae+b)z^2 + (ae^2+c)z+e^4+be^2+ae^3+ce+d
664 			 * z^4 + az^3 + (ae+b)z^2 + e^4+be^2+d
665 			 * z^4 + az^3 +     b'z^2 + d'
666 			 */
667 			d = a_pow(bch, 2*l)^gf_mul(bch, b, f)^d;
668 			b = gf_mul(bch, a, e)^b;
669 		}
670 		/* now, use Y=1/X to get Y^4 + b/dY^2 + a/dY + 1/d */
671 		if (d == 0)
672 			/* assume all roots have multiplicity 1 */
673 			return 0;
674 
675 		c2 = gf_inv(bch, d);
676 		b2 = gf_div(bch, a, d);
677 		a2 = gf_div(bch, b, d);
678 	} else {
679 		/* polynomial is already affine */
680 		c2 = d;
681 		b2 = c;
682 		a2 = b;
683 	}
684 	/* find the 4 roots of this affine polynomial */
685 	if (find_affine4_roots(bch, a2, b2, c2, roots) == 4) {
686 		for (i = 0; i < 4; i++) {
687 			/* post-process roots (reverse transformations) */
688 			f = a ? gf_inv(bch, roots[i]) : roots[i];
689 			roots[i] = a_ilog(bch, f^e);
690 		}
691 		n = 4;
692 	}
693 	return n;
694 }
695 
696 /*
697  * build monic, log-based representation of a polynomial
698  */
699 static void gf_poly_logrep(struct bch_control *bch,
700 			   const struct gf_poly *a, int *rep)
701 {
702 	int i, d = a->deg, l = GF_N(bch)-a_log(bch, a->c[a->deg]);
703 
704 	/* represent 0 values with -1; warning, rep[d] is not set to 1 */
705 	for (i = 0; i < d; i++)
706 		rep[i] = a->c[i] ? mod_s(bch, a_log(bch, a->c[i])+l) : -1;
707 }
708 
709 /*
710  * compute polynomial Euclidean division remainder in GF(2^m)[X]
711  */
712 static void gf_poly_mod(struct bch_control *bch, struct gf_poly *a,
713 			const struct gf_poly *b, int *rep)
714 {
715 	int la, p, m;
716 	unsigned int i, j, *c = a->c;
717 	const unsigned int d = b->deg;
718 
719 	if (a->deg < d)
720 		return;
721 
722 	/* reuse or compute log representation of denominator */
723 	if (!rep) {
724 		rep = bch->cache;
725 		gf_poly_logrep(bch, b, rep);
726 	}
727 
728 	for (j = a->deg; j >= d; j--) {
729 		if (c[j]) {
730 			la = a_log(bch, c[j]);
731 			p = j-d;
732 			for (i = 0; i < d; i++, p++) {
733 				m = rep[i];
734 				if (m >= 0)
735 					c[p] ^= bch->a_pow_tab[mod_s(bch,
736 								     m+la)];
737 			}
738 		}
739 	}
740 	a->deg = d-1;
741 	while (!c[a->deg] && a->deg)
742 		a->deg--;
743 }
744 
745 /*
746  * compute polynomial Euclidean division quotient in GF(2^m)[X]
747  */
748 static void gf_poly_div(struct bch_control *bch, struct gf_poly *a,
749 			const struct gf_poly *b, struct gf_poly *q)
750 {
751 	if (a->deg >= b->deg) {
752 		q->deg = a->deg-b->deg;
753 		/* compute a mod b (modifies a) */
754 		gf_poly_mod(bch, a, b, NULL);
755 		/* quotient is stored in upper part of polynomial a */
756 		memcpy(q->c, &a->c[b->deg], (1+q->deg)*sizeof(unsigned int));
757 	} else {
758 		q->deg = 0;
759 		q->c[0] = 0;
760 	}
761 }
762 
763 /*
764  * compute polynomial GCD (Greatest Common Divisor) in GF(2^m)[X]
765  */
766 static struct gf_poly *gf_poly_gcd(struct bch_control *bch, struct gf_poly *a,
767 				   struct gf_poly *b)
768 {
769 	struct gf_poly *tmp;
770 
771 	dbg("gcd(%s,%s)=", gf_poly_str(a), gf_poly_str(b));
772 
773 	if (a->deg < b->deg) {
774 		tmp = b;
775 		b = a;
776 		a = tmp;
777 	}
778 
779 	while (b->deg > 0) {
780 		gf_poly_mod(bch, a, b, NULL);
781 		tmp = b;
782 		b = a;
783 		a = tmp;
784 	}
785 
786 	dbg("%s\n", gf_poly_str(a));
787 
788 	return a;
789 }
790 
791 /*
792  * Given a polynomial f and an integer k, compute Tr(a^kX) mod f
793  * This is used in Berlekamp Trace algorithm for splitting polynomials
794  */
795 static void compute_trace_bk_mod(struct bch_control *bch, int k,
796 				 const struct gf_poly *f, struct gf_poly *z,
797 				 struct gf_poly *out)
798 {
799 	const int m = GF_M(bch);
800 	int i, j;
801 
802 	/* z contains z^2j mod f */
803 	z->deg = 1;
804 	z->c[0] = 0;
805 	z->c[1] = bch->a_pow_tab[k];
806 
807 	out->deg = 0;
808 	memset(out, 0, GF_POLY_SZ(f->deg));
809 
810 	/* compute f log representation only once */
811 	gf_poly_logrep(bch, f, bch->cache);
812 
813 	for (i = 0; i < m; i++) {
814 		/* add a^(k*2^i)(z^(2^i) mod f) and compute (z^(2^i) mod f)^2 */
815 		for (j = z->deg; j >= 0; j--) {
816 			out->c[j] ^= z->c[j];
817 			z->c[2*j] = gf_sqr(bch, z->c[j]);
818 			z->c[2*j+1] = 0;
819 		}
820 		if (z->deg > out->deg)
821 			out->deg = z->deg;
822 
823 		if (i < m-1) {
824 			z->deg *= 2;
825 			/* z^(2(i+1)) mod f = (z^(2^i) mod f)^2 mod f */
826 			gf_poly_mod(bch, z, f, bch->cache);
827 		}
828 	}
829 	while (!out->c[out->deg] && out->deg)
830 		out->deg--;
831 
832 	dbg("Tr(a^%d.X) mod f = %s\n", k, gf_poly_str(out));
833 }
834 
835 /*
836  * factor a polynomial using Berlekamp Trace algorithm (BTA)
837  */
838 static void factor_polynomial(struct bch_control *bch, int k, struct gf_poly *f,
839 			      struct gf_poly **g, struct gf_poly **h)
840 {
841 	struct gf_poly *f2 = bch->poly_2t[0];
842 	struct gf_poly *q  = bch->poly_2t[1];
843 	struct gf_poly *tk = bch->poly_2t[2];
844 	struct gf_poly *z  = bch->poly_2t[3];
845 	struct gf_poly *gcd;
846 
847 	dbg("factoring %s...\n", gf_poly_str(f));
848 
849 	*g = f;
850 	*h = NULL;
851 
852 	/* tk = Tr(a^k.X) mod f */
853 	compute_trace_bk_mod(bch, k, f, z, tk);
854 
855 	if (tk->deg > 0) {
856 		/* compute g = gcd(f, tk) (destructive operation) */
857 		gf_poly_copy(f2, f);
858 		gcd = gf_poly_gcd(bch, f2, tk);
859 		if (gcd->deg < f->deg) {
860 			/* compute h=f/gcd(f,tk); this will modify f and q */
861 			gf_poly_div(bch, f, gcd, q);
862 			/* store g and h in-place (clobbering f) */
863 			*h = &((struct gf_poly_deg1 *)f)[gcd->deg].poly;
864 			gf_poly_copy(*g, gcd);
865 			gf_poly_copy(*h, q);
866 		}
867 	}
868 }
869 
870 /*
871  * find roots of a polynomial, using BTZ algorithm; see the beginning of this
872  * file for details
873  */
874 static int find_poly_roots(struct bch_control *bch, unsigned int k,
875 			   struct gf_poly *poly, unsigned int *roots)
876 {
877 	int cnt;
878 	struct gf_poly *f1, *f2;
879 
880 	switch (poly->deg) {
881 		/* handle low degree polynomials with ad hoc techniques */
882 	case 1:
883 		cnt = find_poly_deg1_roots(bch, poly, roots);
884 		break;
885 	case 2:
886 		cnt = find_poly_deg2_roots(bch, poly, roots);
887 		break;
888 	case 3:
889 		cnt = find_poly_deg3_roots(bch, poly, roots);
890 		break;
891 	case 4:
892 		cnt = find_poly_deg4_roots(bch, poly, roots);
893 		break;
894 	default:
895 		/* factor polynomial using Berlekamp Trace Algorithm (BTA) */
896 		cnt = 0;
897 		if (poly->deg && (k <= GF_M(bch))) {
898 			factor_polynomial(bch, k, poly, &f1, &f2);
899 			if (f1)
900 				cnt += find_poly_roots(bch, k+1, f1, roots);
901 			if (f2)
902 				cnt += find_poly_roots(bch, k+1, f2, roots+cnt);
903 		}
904 		break;
905 	}
906 	return cnt;
907 }
908 
909 #if defined(USE_CHIEN_SEARCH)
910 /*
911  * exhaustive root search (Chien) implementation - not used, included only for
912  * reference/comparison tests
913  */
914 static int chien_search(struct bch_control *bch, unsigned int len,
915 			struct gf_poly *p, unsigned int *roots)
916 {
917 	int m;
918 	unsigned int i, j, syn, syn0, count = 0;
919 	const unsigned int k = 8*len+bch->ecc_bits;
920 
921 	/* use a log-based representation of polynomial */
922 	gf_poly_logrep(bch, p, bch->cache);
923 	bch->cache[p->deg] = 0;
924 	syn0 = gf_div(bch, p->c[0], p->c[p->deg]);
925 
926 	for (i = GF_N(bch)-k+1; i <= GF_N(bch); i++) {
927 		/* compute elp(a^i) */
928 		for (j = 1, syn = syn0; j <= p->deg; j++) {
929 			m = bch->cache[j];
930 			if (m >= 0)
931 				syn ^= a_pow(bch, m+j*i);
932 		}
933 		if (syn == 0) {
934 			roots[count++] = GF_N(bch)-i;
935 			if (count == p->deg)
936 				break;
937 		}
938 	}
939 	return (count == p->deg) ? count : 0;
940 }
941 #define find_poly_roots(_p, _k, _elp, _loc) chien_search(_p, len, _elp, _loc)
942 #endif /* USE_CHIEN_SEARCH */
943 
944 /**
945  * decode_bch - decode received codeword and find bit error locations
946  * @bch:      BCH control structure
947  * @data:     received data, ignored if @calc_ecc is provided
948  * @len:      data length in bytes, must always be provided
949  * @recv_ecc: received ecc, if NULL then assume it was XORed in @calc_ecc
950  * @calc_ecc: calculated ecc, if NULL then calc_ecc is computed from @data
951  * @syn:      hw computed syndrome data (if NULL, syndrome is calculated)
952  * @errloc:   output array of error locations
953  *
954  * Returns:
955  *  The number of errors found, or -EBADMSG if decoding failed, or -EINVAL if
956  *  invalid parameters were provided
957  *
958  * Depending on the available hw BCH support and the need to compute @calc_ecc
959  * separately (using encode_bch()), this function should be called with one of
960  * the following parameter configurations -
961  *
962  * by providing @data and @recv_ecc only:
963  *   decode_bch(@bch, @data, @len, @recv_ecc, NULL, NULL, @errloc)
964  *
965  * by providing @recv_ecc and @calc_ecc:
966  *   decode_bch(@bch, NULL, @len, @recv_ecc, @calc_ecc, NULL, @errloc)
967  *
968  * by providing ecc = recv_ecc XOR calc_ecc:
969  *   decode_bch(@bch, NULL, @len, NULL, ecc, NULL, @errloc)
970  *
971  * by providing syndrome results @syn:
972  *   decode_bch(@bch, NULL, @len, NULL, NULL, @syn, @errloc)
973  *
974  * Once decode_bch() has successfully returned with a positive value, error
975  * locations returned in array @errloc should be interpreted as follows -
976  *
977  * if (errloc[n] >= 8*len), then n-th error is located in ecc (no need for
978  * data correction)
979  *
980  * if (errloc[n] < 8*len), then n-th error is located in data and can be
981  * corrected with statement data[errloc[n]/8] ^= 1 << (errloc[n] % 8);
982  *
983  * Note that this function does not perform any data correction by itself, it
984  * merely indicates error locations.
985  */
986 int decode_bch(struct bch_control *bch, const uint8_t *data, unsigned int len,
987 	       const uint8_t *recv_ecc, const uint8_t *calc_ecc,
988 	       const unsigned int *syn, unsigned int *errloc)
989 {
990 	const unsigned int ecc_words = BCH_ECC_WORDS(bch);
991 	unsigned int nbits;
992 	int i, err, nroots;
993 	uint32_t sum;
994 
995 	/* sanity check: make sure data length can be handled */
996 	if (8*len > (bch->n-bch->ecc_bits))
997 		return -EINVAL;
998 
999 	/* if caller does not provide syndromes, compute them */
1000 	if (!syn) {
1001 		if (!calc_ecc) {
1002 			/* compute received data ecc into an internal buffer */
1003 			if (!data || !recv_ecc)
1004 				return -EINVAL;
1005 			encode_bch(bch, data, len, NULL);
1006 		} else {
1007 			/* load provided calculated ecc */
1008 			load_ecc8(bch, bch->ecc_buf, calc_ecc);
1009 		}
1010 		/* load received ecc or assume it was XORed in calc_ecc */
1011 		if (recv_ecc) {
1012 			load_ecc8(bch, bch->ecc_buf2, recv_ecc);
1013 			/* XOR received and calculated ecc */
1014 			for (i = 0, sum = 0; i < (int)ecc_words; i++) {
1015 				bch->ecc_buf[i] ^= bch->ecc_buf2[i];
1016 				sum |= bch->ecc_buf[i];
1017 			}
1018 			if (!sum)
1019 				/* no error found */
1020 				return 0;
1021 		}
1022 		compute_syndromes(bch, bch->ecc_buf, bch->syn);
1023 		syn = bch->syn;
1024 	}
1025 
1026 	err = compute_error_locator_polynomial(bch, syn);
1027 	if (err > 0) {
1028 		nroots = find_poly_roots(bch, 1, bch->elp, errloc);
1029 		if (err != nroots)
1030 			err = -1;
1031 	}
1032 	if (err > 0) {
1033 		/* post-process raw error locations for easier correction */
1034 		nbits = (len*8)+bch->ecc_bits;
1035 		for (i = 0; i < err; i++) {
1036 			if (errloc[i] >= nbits) {
1037 				err = -1;
1038 				break;
1039 			}
1040 			errloc[i] = nbits-1-errloc[i];
1041 			errloc[i] = (errloc[i] & ~7)|(7-(errloc[i] & 7));
1042 		}
1043 	}
1044 	return (err >= 0) ? err : -EBADMSG;
1045 }
1046 EXPORT_SYMBOL_GPL(decode_bch);
1047 
1048 /*
1049  * generate Galois field lookup tables
1050  */
1051 static int build_gf_tables(struct bch_control *bch, unsigned int poly)
1052 {
1053 	unsigned int i, x = 1;
1054 	const unsigned int k = 1 << deg(poly);
1055 
1056 	/* primitive polynomial must be of degree m */
1057 	if (k != (1u << GF_M(bch)))
1058 		return -1;
1059 
1060 	for (i = 0; i < GF_N(bch); i++) {
1061 		bch->a_pow_tab[i] = x;
1062 		bch->a_log_tab[x] = i;
1063 		if (i && (x == 1))
1064 			/* polynomial is not primitive (a^i=1 with 0<i<2^m-1) */
1065 			return -1;
1066 		x <<= 1;
1067 		if (x & k)
1068 			x ^= poly;
1069 	}
1070 	bch->a_pow_tab[GF_N(bch)] = 1;
1071 	bch->a_log_tab[0] = 0;
1072 
1073 	return 0;
1074 }
1075 
1076 /*
1077  * compute generator polynomial remainder tables for fast encoding
1078  */
1079 static void build_mod8_tables(struct bch_control *bch, const uint32_t *g)
1080 {
1081 	int i, j, b, d;
1082 	uint32_t data, hi, lo, *tab;
1083 	const int l = BCH_ECC_WORDS(bch);
1084 	const int plen = DIV_ROUND_UP(bch->ecc_bits+1, 32);
1085 	const int ecclen = DIV_ROUND_UP(bch->ecc_bits, 32);
1086 
1087 	memset(bch->mod8_tab, 0, 4*256*l*sizeof(*bch->mod8_tab));
1088 
1089 	for (i = 0; i < 256; i++) {
1090 		/* p(X)=i is a small polynomial of weight <= 8 */
1091 		for (b = 0; b < 4; b++) {
1092 			/* we want to compute (p(X).X^(8*b+deg(g))) mod g(X) */
1093 			tab = bch->mod8_tab + (b*256+i)*l;
1094 			data = i << (8*b);
1095 			while (data) {
1096 				d = deg(data);
1097 				/* subtract X^d.g(X) from p(X).X^(8*b+deg(g)) */
1098 				data ^= g[0] >> (31-d);
1099 				for (j = 0; j < ecclen; j++) {
1100 					hi = (d < 31) ? g[j] << (d+1) : 0;
1101 					lo = (j+1 < plen) ?
1102 						g[j+1] >> (31-d) : 0;
1103 					tab[j] ^= hi|lo;
1104 				}
1105 			}
1106 		}
1107 	}
1108 }
1109 
1110 /*
1111  * build a base for factoring degree 2 polynomials
1112  */
1113 static int build_deg2_base(struct bch_control *bch)
1114 {
1115 	const int m = GF_M(bch);
1116 	int i, j, r;
1117 	unsigned int sum, x, y, remaining, ak = 0, xi[m];
1118 
1119 	/* find k s.t. Tr(a^k) = 1 and 0 <= k < m */
1120 	for (i = 0; i < m; i++) {
1121 		for (j = 0, sum = 0; j < m; j++)
1122 			sum ^= a_pow(bch, i*(1 << j));
1123 
1124 		if (sum) {
1125 			ak = bch->a_pow_tab[i];
1126 			break;
1127 		}
1128 	}
1129 	/* find xi, i=0..m-1 such that xi^2+xi = a^i+Tr(a^i).a^k */
1130 	remaining = m;
1131 	memset(xi, 0, sizeof(xi));
1132 
1133 	for (x = 0; (x <= GF_N(bch)) && remaining; x++) {
1134 		y = gf_sqr(bch, x)^x;
1135 		for (i = 0; i < 2; i++) {
1136 			r = a_log(bch, y);
1137 			if (y && (r < m) && !xi[r]) {
1138 				bch->xi_tab[r] = x;
1139 				xi[r] = 1;
1140 				remaining--;
1141 				dbg("x%d = %x\n", r, x);
1142 				break;
1143 			}
1144 			y ^= ak;
1145 		}
1146 	}
1147 	/* should not happen but check anyway */
1148 	return remaining ? -1 : 0;
1149 }
1150 
1151 static void *bch_alloc(size_t size, int *err)
1152 {
1153 	void *ptr;
1154 
1155 	ptr = kmalloc(size, GFP_KERNEL);
1156 	if (ptr == NULL)
1157 		*err = 1;
1158 	return ptr;
1159 }
1160 
1161 /*
1162  * compute generator polynomial for given (m,t) parameters.
1163  */
1164 static uint32_t *compute_generator_polynomial(struct bch_control *bch)
1165 {
1166 	const unsigned int m = GF_M(bch);
1167 	const unsigned int t = GF_T(bch);
1168 	int n, err = 0;
1169 	unsigned int i, j, nbits, r, word, *roots;
1170 	struct gf_poly *g;
1171 	uint32_t *genpoly;
1172 
1173 	g = bch_alloc(GF_POLY_SZ(m*t), &err);
1174 	roots = bch_alloc((bch->n+1)*sizeof(*roots), &err);
1175 	genpoly = bch_alloc(DIV_ROUND_UP(m*t+1, 32)*sizeof(*genpoly), &err);
1176 
1177 	if (err) {
1178 		kfree(genpoly);
1179 		genpoly = NULL;
1180 		goto finish;
1181 	}
1182 
1183 	/* enumerate all roots of g(X) */
1184 	memset(roots , 0, (bch->n+1)*sizeof(*roots));
1185 	for (i = 0; i < t; i++) {
1186 		for (j = 0, r = 2*i+1; j < m; j++) {
1187 			roots[r] = 1;
1188 			r = mod_s(bch, 2*r);
1189 		}
1190 	}
1191 	/* build generator polynomial g(X) */
1192 	g->deg = 0;
1193 	g->c[0] = 1;
1194 	for (i = 0; i < GF_N(bch); i++) {
1195 		if (roots[i]) {
1196 			/* multiply g(X) by (X+root) */
1197 			r = bch->a_pow_tab[i];
1198 			g->c[g->deg+1] = 1;
1199 			for (j = g->deg; j > 0; j--)
1200 				g->c[j] = gf_mul(bch, g->c[j], r)^g->c[j-1];
1201 
1202 			g->c[0] = gf_mul(bch, g->c[0], r);
1203 			g->deg++;
1204 		}
1205 	}
1206 	/* store left-justified binary representation of g(X) */
1207 	n = g->deg+1;
1208 	i = 0;
1209 
1210 	while (n > 0) {
1211 		nbits = (n > 32) ? 32 : n;
1212 		for (j = 0, word = 0; j < nbits; j++) {
1213 			if (g->c[n-1-j])
1214 				word |= 1u << (31-j);
1215 		}
1216 		genpoly[i++] = word;
1217 		n -= nbits;
1218 	}
1219 	bch->ecc_bits = g->deg;
1220 
1221 finish:
1222 	kfree(g);
1223 	kfree(roots);
1224 
1225 	return genpoly;
1226 }
1227 
1228 /**
1229  * init_bch - initialize a BCH encoder/decoder
1230  * @m:          Galois field order, should be in the range 5-15
1231  * @t:          maximum error correction capability, in bits
1232  * @prim_poly:  user-provided primitive polynomial (or 0 to use default)
1233  *
1234  * Returns:
1235  *  a newly allocated BCH control structure if successful, NULL otherwise
1236  *
1237  * This initialization can take some time, as lookup tables are built for fast
1238  * encoding/decoding; make sure not to call this function from a time critical
1239  * path. Usually, init_bch() should be called on module/driver init and
1240  * free_bch() should be called to release memory on exit.
1241  *
1242  * You may provide your own primitive polynomial of degree @m in argument
1243  * @prim_poly, or let init_bch() use its default polynomial.
1244  *
1245  * Once init_bch() has successfully returned a pointer to a newly allocated
1246  * BCH control structure, ecc length in bytes is given by member @ecc_bytes of
1247  * the structure.
1248  */
1249 struct bch_control *init_bch(int m, int t, unsigned int prim_poly)
1250 {
1251 	int err = 0;
1252 	unsigned int i, words;
1253 	uint32_t *genpoly;
1254 	struct bch_control *bch = NULL;
1255 
1256 	const int min_m = 5;
1257 	const int max_m = 15;
1258 
1259 	/* default primitive polynomials */
1260 	static const unsigned int prim_poly_tab[] = {
1261 		0x25, 0x43, 0x83, 0x11d, 0x211, 0x409, 0x805, 0x1053, 0x201b,
1262 		0x402b, 0x8003,
1263 	};
1264 
1265 #if defined(CONFIG_BCH_CONST_PARAMS)
1266 	if ((m != (CONFIG_BCH_CONST_M)) || (t != (CONFIG_BCH_CONST_T))) {
1267 		printk(KERN_ERR "bch encoder/decoder was configured to support "
1268 		       "parameters m=%d, t=%d only!\n",
1269 		       CONFIG_BCH_CONST_M, CONFIG_BCH_CONST_T);
1270 		goto fail;
1271 	}
1272 #endif
1273 	if ((m < min_m) || (m > max_m))
1274 		/*
1275 		 * values of m greater than 15 are not currently supported;
1276 		 * supporting m > 15 would require changing table base type
1277 		 * (uint16_t) and a small patch in matrix transposition
1278 		 */
1279 		goto fail;
1280 
1281 	/* sanity checks */
1282 	if ((t < 1) || (m*t >= ((1 << m)-1)))
1283 		/* invalid t value */
1284 		goto fail;
1285 
1286 	/* select a primitive polynomial for generating GF(2^m) */
1287 	if (prim_poly == 0)
1288 		prim_poly = prim_poly_tab[m-min_m];
1289 
1290 	bch = kzalloc(sizeof(*bch), GFP_KERNEL);
1291 	if (bch == NULL)
1292 		goto fail;
1293 
1294 	bch->m = m;
1295 	bch->t = t;
1296 	bch->n = (1 << m)-1;
1297 	words  = DIV_ROUND_UP(m*t, 32);
1298 	bch->ecc_bytes = DIV_ROUND_UP(m*t, 8);
1299 	bch->a_pow_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_pow_tab), &err);
1300 	bch->a_log_tab = bch_alloc((1+bch->n)*sizeof(*bch->a_log_tab), &err);
1301 	bch->mod8_tab  = bch_alloc(words*1024*sizeof(*bch->mod8_tab), &err);
1302 	bch->ecc_buf   = bch_alloc(words*sizeof(*bch->ecc_buf), &err);
1303 	bch->ecc_buf2  = bch_alloc(words*sizeof(*bch->ecc_buf2), &err);
1304 	bch->xi_tab    = bch_alloc(m*sizeof(*bch->xi_tab), &err);
1305 	bch->syn       = bch_alloc(2*t*sizeof(*bch->syn), &err);
1306 	bch->cache     = bch_alloc(2*t*sizeof(*bch->cache), &err);
1307 	bch->elp       = bch_alloc((t+1)*sizeof(struct gf_poly_deg1), &err);
1308 
1309 	for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++)
1310 		bch->poly_2t[i] = bch_alloc(GF_POLY_SZ(2*t), &err);
1311 
1312 	if (err)
1313 		goto fail;
1314 
1315 	err = build_gf_tables(bch, prim_poly);
1316 	if (err)
1317 		goto fail;
1318 
1319 	/* use generator polynomial for computing encoding tables */
1320 	genpoly = compute_generator_polynomial(bch);
1321 	if (genpoly == NULL)
1322 		goto fail;
1323 
1324 	build_mod8_tables(bch, genpoly);
1325 	kfree(genpoly);
1326 
1327 	err = build_deg2_base(bch);
1328 	if (err)
1329 		goto fail;
1330 
1331 	return bch;
1332 
1333 fail:
1334 	free_bch(bch);
1335 	return NULL;
1336 }
1337 EXPORT_SYMBOL_GPL(init_bch);
1338 
1339 /**
1340  *  free_bch - free the BCH control structure
1341  *  @bch:    BCH control structure to release
1342  */
1343 void free_bch(struct bch_control *bch)
1344 {
1345 	unsigned int i;
1346 
1347 	if (bch) {
1348 		kfree(bch->a_pow_tab);
1349 		kfree(bch->a_log_tab);
1350 		kfree(bch->mod8_tab);
1351 		kfree(bch->ecc_buf);
1352 		kfree(bch->ecc_buf2);
1353 		kfree(bch->xi_tab);
1354 		kfree(bch->syn);
1355 		kfree(bch->cache);
1356 		kfree(bch->elp);
1357 
1358 		for (i = 0; i < ARRAY_SIZE(bch->poly_2t); i++)
1359 			kfree(bch->poly_2t[i]);
1360 
1361 		kfree(bch);
1362 	}
1363 }
1364 EXPORT_SYMBOL_GPL(free_bch);
1365 
1366 MODULE_LICENSE("GPL");
1367 MODULE_AUTHOR("Ivan Djelic <ivan.djelic@parrot.com>");
1368 MODULE_DESCRIPTION("Binary BCH encoder/decoder");
1369