2 * This file contains an ECC algorithm that detects and corrects 1 bit
3 * errors in a 256 byte block of data.
5 * drivers/mtd/nand/nand_ecc.c
7 * Copyright © 2008 Koninklijke Philips Electronics NV.
8 * Author: Frans Meulenbroeks
10 * Completely replaces the previous ECC implementation which was written by:
11 * Steven J. Hill (sjhill@realitydiluted.com)
12 * Thomas Gleixner (tglx@linutronix.de)
14 * Information on how this algorithm works and how it was developed
15 * can be found in Documentation/mtd/nand_ecc.txt
17 * This file is free software; you can redistribute it and/or modify it
18 * under the terms of the GNU General Public License as published by the
19 * Free Software Foundation; either version 2 or (at your option) any
22 * This file is distributed in the hope that it will be useful, but WITHOUT
23 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
24 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
27 * You should have received a copy of the GNU General Public License along
28 * with this file; if not, write to the Free Software Foundation, Inc.,
29 * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
34 * The STANDALONE macro is useful when running the code outside the kernel
35 * e.g. when running the code in a testbed or a benchmark program.
36 * When STANDALONE is used, the module related macros are commented out
37 * as well as the linux include files.
38 * Instead a private definition of mtd_info is given to satisfy the compiler
39 * (the code does not use mtd_info, so the code does not care)
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/module.h>
45 #include <linux/mtd/mtd.h>
46 #include <linux/mtd/nand.h>
47 #include <linux/mtd/nand_ecc.h>
48 #include <asm/byteorder.h>
52 #define EXPORT_SYMBOL(x) /* x */
54 #define MODULE_LICENSE(x) /* x */
55 #define MODULE_AUTHOR(x) /* x */
56 #define MODULE_DESCRIPTION(x) /* x */
62 * invparity is a 256 byte table that contains the odd parity
63 * for each byte. So if the number of bits in a byte is even,
64 * the array element is 1, and when the number of bits is odd
65 * the array eleemnt is 0.
67 static const char invparity[256] = {
68 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
69 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
70 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
71 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
72 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
73 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
74 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
75 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
76 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
77 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
78 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
79 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
80 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1,
81 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
82 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0,
83 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1
87 * bitsperbyte contains the number of bits per byte
88 * this is only used for testing and repairing parity
89 * (a precalculated value slightly improves performance)
91 static const char bitsperbyte[256] = {
92 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4,
93 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
94 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
95 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
96 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
97 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
98 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
99 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
100 1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
101 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
102 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
103 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
104 2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
105 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
106 3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
107 4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8,
111 * addressbits is a lookup table to filter out the bits from the xor-ed
112 * ECC data that identify the faulty location.
113 * this is only used for repairing parity
114 * see the comments in nand_correct_data for more details
116 static const char addressbits[256] = {
117 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
118 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
119 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
120 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
121 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
122 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
123 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
124 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
125 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
126 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
127 0x00, 0x00, 0x01, 0x01, 0x00, 0x00, 0x01, 0x01,
128 0x02, 0x02, 0x03, 0x03, 0x02, 0x02, 0x03, 0x03,
129 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
130 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
131 0x04, 0x04, 0x05, 0x05, 0x04, 0x04, 0x05, 0x05,
132 0x06, 0x06, 0x07, 0x07, 0x06, 0x06, 0x07, 0x07,
133 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
134 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
135 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
136 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
137 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
138 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f,
139 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
140 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f,
141 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
142 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
143 0x08, 0x08, 0x09, 0x09, 0x08, 0x08, 0x09, 0x09,
144 0x0a, 0x0a, 0x0b, 0x0b, 0x0a, 0x0a, 0x0b, 0x0b,
145 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
146 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f,
147 0x0c, 0x0c, 0x0d, 0x0d, 0x0c, 0x0c, 0x0d, 0x0d,
148 0x0e, 0x0e, 0x0f, 0x0f, 0x0e, 0x0e, 0x0f, 0x0f
152 * __nand_calculate_ecc - [NAND Interface] Calculate 3-byte ECC for 256/512-byte
154 * @buf: input buffer with raw data
155 * @eccsize: data bytes per ECC step (256 or 512)
156 * @code: output buffer with ECC
158 void __nand_calculate_ecc(const unsigned char *buf, unsigned int eccsize,
162 const uint32_t *bp = (uint32_t *)buf;
163 /* 256 or 512 bytes/ecc */
164 const uint32_t eccsize_mult = eccsize >> 8;
165 uint32_t cur; /* current value in buffer */
166 /* rp0..rp15..rp17 are the various accumulated parities (per byte) */
167 uint32_t rp0, rp1, rp2, rp3, rp4, rp5, rp6, rp7;
168 uint32_t rp8, rp9, rp10, rp11, rp12, rp13, rp14, rp15, rp16;
169 uint32_t uninitialized_var(rp17); /* to make compiler happy */
170 uint32_t par; /* the cumulative parity for all data */
171 uint32_t tmppar; /* the cumulative parity for this iteration;
172 for rp12, rp14 and rp16 at the end of the
185 * The loop is unrolled a number of times;
186 * This avoids if statements to decide on which rp value to update
187 * Also we process the data by longwords.
188 * Note: passing unaligned data might give a performance penalty.
189 * It is assumed that the buffers are aligned.
190 * tmppar is the cumulative sum of this iteration.
191 * needed for calculating rp12, rp14, rp16 and par
192 * also used as a performance improvement for rp6, rp8 and rp10
194 for (i = 0; i < eccsize_mult << 2; i++) {
257 if (eccsize_mult == 2 && (i & 0x4) == 0)
262 * handle the fact that we use longword operations
263 * we'll bring rp4..rp14..rp16 back to single byte entities by
264 * shifting and xoring first fold the upper and lower 16 bits,
265 * then the upper and lower 8 bits.
276 rp10 ^= (rp10 >> 16);
279 rp12 ^= (rp12 >> 16);
282 rp14 ^= (rp14 >> 16);
285 if (eccsize_mult == 2) {
286 rp16 ^= (rp16 >> 16);
292 * we also need to calculate the row parity for rp0..rp3
293 * This is present in par, because par is now
294 * rp3 rp3 rp2 rp2 in little endian and
295 * rp2 rp2 rp3 rp3 in big endian
297 * rp1 rp0 rp1 rp0 in little endian and
298 * rp0 rp1 rp0 rp1 in big endian
299 * First calculate rp2 and rp3
317 /* reduce par to 16 bits then calculate rp1 and rp0 */
320 rp0 = (par >> 8) & 0xff;
323 rp1 = (par >> 8) & 0xff;
327 /* finally reduce par to 8 bits */
332 * and calculate rp5..rp15..rp17
333 * note that par = rp4 ^ rp5 and due to the commutative property
334 * of the ^ operator we can say:
336 * The & 0xff seems superfluous, but benchmarking learned that
337 * leaving it out gives slightly worse results. No idea why, probably
338 * it has to do with the way the pipeline in pentium is organized.
340 rp5 = (par ^ rp4) & 0xff;
341 rp7 = (par ^ rp6) & 0xff;
342 rp9 = (par ^ rp8) & 0xff;
343 rp11 = (par ^ rp10) & 0xff;
344 rp13 = (par ^ rp12) & 0xff;
345 rp15 = (par ^ rp14) & 0xff;
346 if (eccsize_mult == 2)
347 rp17 = (par ^ rp16) & 0xff;
350 * Finally calculate the ECC bits.
351 * Again here it might seem that there are performance optimisations
352 * possible, but benchmarks showed that on the system this is developed
353 * the code below is the fastest
355 #ifdef CONFIG_MTD_NAND_ECC_SMC
357 (invparity[rp7] << 7) |
358 (invparity[rp6] << 6) |
359 (invparity[rp5] << 5) |
360 (invparity[rp4] << 4) |
361 (invparity[rp3] << 3) |
362 (invparity[rp2] << 2) |
363 (invparity[rp1] << 1) |
366 (invparity[rp15] << 7) |
367 (invparity[rp14] << 6) |
368 (invparity[rp13] << 5) |
369 (invparity[rp12] << 4) |
370 (invparity[rp11] << 3) |
371 (invparity[rp10] << 2) |
372 (invparity[rp9] << 1) |
376 (invparity[rp7] << 7) |
377 (invparity[rp6] << 6) |
378 (invparity[rp5] << 5) |
379 (invparity[rp4] << 4) |
380 (invparity[rp3] << 3) |
381 (invparity[rp2] << 2) |
382 (invparity[rp1] << 1) |
385 (invparity[rp15] << 7) |
386 (invparity[rp14] << 6) |
387 (invparity[rp13] << 5) |
388 (invparity[rp12] << 4) |
389 (invparity[rp11] << 3) |
390 (invparity[rp10] << 2) |
391 (invparity[rp9] << 1) |
394 if (eccsize_mult == 1)
396 (invparity[par & 0xf0] << 7) |
397 (invparity[par & 0x0f] << 6) |
398 (invparity[par & 0xcc] << 5) |
399 (invparity[par & 0x33] << 4) |
400 (invparity[par & 0xaa] << 3) |
401 (invparity[par & 0x55] << 2) |
405 (invparity[par & 0xf0] << 7) |
406 (invparity[par & 0x0f] << 6) |
407 (invparity[par & 0xcc] << 5) |
408 (invparity[par & 0x33] << 4) |
409 (invparity[par & 0xaa] << 3) |
410 (invparity[par & 0x55] << 2) |
411 (invparity[rp17] << 1) |
412 (invparity[rp16] << 0);
414 EXPORT_SYMBOL(__nand_calculate_ecc);
417 * nand_calculate_ecc - [NAND Interface] Calculate 3-byte ECC for 256/512-byte
419 * @mtd: MTD block structure
420 * @buf: input buffer with raw data
421 * @code: output buffer with ECC
423 int nand_calculate_ecc(struct mtd_info *mtd, const unsigned char *buf,
426 __nand_calculate_ecc(buf,
427 mtd_to_nand(mtd)->ecc.size, code);
431 EXPORT_SYMBOL(nand_calculate_ecc);
434 * __nand_correct_data - [NAND Interface] Detect and correct bit error(s)
435 * @buf: raw data read from the chip
436 * @read_ecc: ECC from the chip
437 * @calc_ecc: the ECC calculated from raw data
438 * @eccsize: data bytes per ECC step (256 or 512)
440 * Detect and correct a 1 bit error for eccsize byte block
442 int __nand_correct_data(unsigned char *buf,
443 unsigned char *read_ecc, unsigned char *calc_ecc,
444 unsigned int eccsize)
446 unsigned char b0, b1, b2, bit_addr;
447 unsigned int byte_addr;
448 /* 256 or 512 bytes/ecc */
449 const uint32_t eccsize_mult = eccsize >> 8;
452 * b0 to b2 indicate which bit is faulty (if any)
453 * we might need the xor result more than once,
454 * so keep them in a local var
456 #ifdef CONFIG_MTD_NAND_ECC_SMC
457 b0 = read_ecc[0] ^ calc_ecc[0];
458 b1 = read_ecc[1] ^ calc_ecc[1];
460 b0 = read_ecc[1] ^ calc_ecc[1];
461 b1 = read_ecc[0] ^ calc_ecc[0];
463 b2 = read_ecc[2] ^ calc_ecc[2];
465 /* check if there are any bitfaults */
467 /* repeated if statements are slightly more efficient than switch ... */
468 /* ordered in order of likelihood */
470 if ((b0 | b1 | b2) == 0)
471 return 0; /* no error */
473 if ((((b0 ^ (b0 >> 1)) & 0x55) == 0x55) &&
474 (((b1 ^ (b1 >> 1)) & 0x55) == 0x55) &&
475 ((eccsize_mult == 1 && ((b2 ^ (b2 >> 1)) & 0x54) == 0x54) ||
476 (eccsize_mult == 2 && ((b2 ^ (b2 >> 1)) & 0x55) == 0x55))) {
477 /* single bit error */
479 * rp17/rp15/13/11/9/7/5/3/1 indicate which byte is the faulty
480 * byte, cp 5/3/1 indicate the faulty bit.
481 * A lookup table (called addressbits) is used to filter
482 * the bits from the byte they are in.
483 * A marginal optimisation is possible by having three
484 * different lookup tables.
485 * One as we have now (for b0), one for b2
486 * (that would avoid the >> 1), and one for b1 (with all values
487 * << 4). However it was felt that introducing two more tables
488 * hardly justify the gain.
490 * The b2 shift is there to get rid of the lowest two bits.
491 * We could also do addressbits[b2] >> 1 but for the
492 * performance it does not make any difference
494 if (eccsize_mult == 1)
495 byte_addr = (addressbits[b1] << 4) + addressbits[b0];
497 byte_addr = (addressbits[b2 & 0x3] << 8) +
498 (addressbits[b1] << 4) + addressbits[b0];
499 bit_addr = addressbits[b2 >> 2];
501 buf[byte_addr] ^= (1 << bit_addr);
505 /* count nr of bits; use table lookup, faster than calculating it */
506 if ((bitsperbyte[b0] + bitsperbyte[b1] + bitsperbyte[b2]) == 1)
507 return 1; /* error in ECC data; no action needed */
509 pr_err("%s: uncorrectable ECC error\n", __func__);
512 EXPORT_SYMBOL(__nand_correct_data);
515 * nand_correct_data - [NAND Interface] Detect and correct bit error(s)
516 * @mtd: MTD block structure
517 * @buf: raw data read from the chip
518 * @read_ecc: ECC from the chip
519 * @calc_ecc: the ECC calculated from raw data
521 * Detect and correct a 1 bit error for 256/512 byte block
523 int nand_correct_data(struct mtd_info *mtd, unsigned char *buf,
524 unsigned char *read_ecc, unsigned char *calc_ecc)
526 return __nand_correct_data(buf, read_ecc, calc_ecc,
527 mtd_to_nand(mtd)->ecc.size);
529 EXPORT_SYMBOL(nand_correct_data);
531 MODULE_LICENSE("GPL");
532 MODULE_AUTHOR("Frans Meulenbroeks <fransmeulenbroeks@gmail.com>");
533 MODULE_DESCRIPTION("Generic NAND ECC support");