2bf3d7f61aec68e99c64ac85f03cb95b8bca6d35
[deliverable/linux.git] / fs / ocfs2 / blockcheck.c
1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * blockcheck.c
5 *
6 * Checksum and ECC codes for the OCFS2 userspace library.
7 *
8 * Copyright (C) 2006, 2008 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License, version 2, as published by the Free Software Foundation.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * General Public License for more details.
18 */
19
20 #include <linux/kernel.h>
21 #include <linux/types.h>
22 #include <linux/crc32.h>
23 #include <linux/buffer_head.h>
24 #include <linux/bitops.h>
25 #include <asm/byteorder.h>
26
27 #include "ocfs2.h"
28
29 #include "blockcheck.h"
30
31
32
33 /*
34 * We use the following conventions:
35 *
36 * d = # data bits
37 * p = # parity bits
38 * c = # total code bits (d + p)
39 */
40 static int calc_parity_bits(unsigned int d)
41 {
42 unsigned int p;
43
44 /*
45 * Bits required for Single Error Correction is as follows:
46 *
47 * d + p + 1 <= 2^p
48 *
49 * We're restricting ourselves to 31 bits of parity, that should be
50 * sufficient.
51 */
52 for (p = 1; p < 32; p++)
53 {
54 if ((d + p + 1) <= (1 << p))
55 return p;
56 }
57
58 return 0;
59 }
60
61 /*
62 * Calculate the bit offset in the hamming code buffer based on the bit's
63 * offset in the data buffer. Since the hamming code reserves all
64 * power-of-two bits for parity, the data bit number and the code bit
65 * number are offest by all the parity bits beforehand.
66 *
67 * Recall that bit numbers in hamming code are 1-based. This function
68 * takes the 0-based data bit from the caller.
69 *
70 * An example. Take bit 1 of the data buffer. 1 is a power of two (2^0),
71 * so it's a parity bit. 2 is a power of two (2^1), so it's a parity bit.
72 * 3 is not a power of two. So bit 1 of the data buffer ends up as bit 3
73 * in the code buffer.
74 */
75 static unsigned int calc_code_bit(unsigned int i)
76 {
77 unsigned int b, p;
78
79 /*
80 * Data bits are 0-based, but we're talking code bits, which
81 * are 1-based.
82 */
83 b = i + 1;
84
85 /*
86 * For every power of two below our bit number, bump our bit.
87 *
88 * We compare with (b + 1) becuase we have to compare with what b
89 * would be _if_ it were bumped up by the parity bit. Capice?
90 */
91 for (p = 0; (1 << p) < (b + 1); p++)
92 b++;
93
94 return b;
95 }
96
97 /*
98 * This is the low level encoder function. It can be called across
99 * multiple hunks just like the crc32 code. 'd' is the number of bits
100 * _in_this_hunk_. nr is the bit offset of this hunk. So, if you had
101 * two 512B buffers, you would do it like so:
102 *
103 * parity = ocfs2_hamming_encode(0, buf1, 512 * 8, 0);
104 * parity = ocfs2_hamming_encode(parity, buf2, 512 * 8, 512 * 8);
105 *
106 * If you just have one buffer, use ocfs2_hamming_encode_block().
107 */
108 u32 ocfs2_hamming_encode(u32 parity, void *data, unsigned int d, unsigned int nr)
109 {
110 unsigned int p = calc_parity_bits(nr + d);
111 unsigned int i, j, b;
112
113 BUG_ON(!p);
114
115 /*
116 * b is the hamming code bit number. Hamming code specifies a
117 * 1-based array, but C uses 0-based. So 'i' is for C, and 'b' is
118 * for the algorithm.
119 *
120 * The i++ in the for loop is so that the start offset passed
121 * to ocfs2_find_next_bit_set() is one greater than the previously
122 * found bit.
123 */
124 for (i = 0; (i = ocfs2_find_next_bit(data, d, i)) < d; i++)
125 {
126 /*
127 * i is the offset in this hunk, nr + i is the total bit
128 * offset.
129 */
130 b = calc_code_bit(nr + i);
131
132 for (j = 0; j < p; j++)
133 {
134 /*
135 * Data bits in the resultant code are checked by
136 * parity bits that are part of the bit number
137 * representation. Huh?
138 *
139 * <wikipedia href="http://en.wikipedia.org/wiki/Hamming_code">
140 * In other words, the parity bit at position 2^k
141 * checks bits in positions having bit k set in
142 * their binary representation. Conversely, for
143 * instance, bit 13, i.e. 1101(2), is checked by
144 * bits 1000(2) = 8, 0100(2)=4 and 0001(2) = 1.
145 * </wikipedia>
146 *
147 * Note that 'k' is the _code_ bit number. 'b' in
148 * our loop.
149 */
150 if (b & (1 << j))
151 parity ^= (1 << j);
152 }
153 }
154
155 /* While the data buffer was treated as little endian, the
156 * return value is in host endian. */
157 return parity;
158 }
159
160 u32 ocfs2_hamming_encode_block(void *data, unsigned int blocksize)
161 {
162 return ocfs2_hamming_encode(0, data, blocksize * 8, 0);
163 }
164
165 /*
166 * Like ocfs2_hamming_encode(), this can handle hunks. nr is the bit
167 * offset of the current hunk. If bit to be fixed is not part of the
168 * current hunk, this does nothing.
169 *
170 * If you only have one hunk, use ocfs2_hamming_fix_block().
171 */
172 void ocfs2_hamming_fix(void *data, unsigned int d, unsigned int nr,
173 unsigned int fix)
174 {
175 unsigned int p = calc_parity_bits(nr + d);
176 unsigned int i, b;
177
178 BUG_ON(!p);
179
180 /*
181 * If the bit to fix has an hweight of 1, it's a parity bit. One
182 * busted parity bit is its own error. Nothing to do here.
183 */
184 if (hweight32(fix) == 1)
185 return;
186
187 /*
188 * nr + d is the bit right past the data hunk we're looking at.
189 * If fix after that, nothing to do
190 */
191 if (fix >= calc_code_bit(nr + d))
192 return;
193
194 /*
195 * nr is the offset in the data hunk we're starting at. Let's
196 * start b at the offset in the code buffer. See hamming_encode()
197 * for a more detailed description of 'b'.
198 */
199 b = calc_code_bit(nr);
200 /* If the fix is before this hunk, nothing to do */
201 if (fix < b)
202 return;
203
204 for (i = 0; i < d; i++, b++)
205 {
206 /* Skip past parity bits */
207 while (hweight32(b) == 1)
208 b++;
209
210 /*
211 * i is the offset in this data hunk.
212 * nr + i is the offset in the total data buffer.
213 * b is the offset in the total code buffer.
214 *
215 * Thus, when b == fix, bit i in the current hunk needs
216 * fixing.
217 */
218 if (b == fix)
219 {
220 if (ocfs2_test_bit(i, data))
221 ocfs2_clear_bit(i, data);
222 else
223 ocfs2_set_bit(i, data);
224 break;
225 }
226 }
227 }
228
229 void ocfs2_hamming_fix_block(void *data, unsigned int blocksize,
230 unsigned int fix)
231 {
232 ocfs2_hamming_fix(data, blocksize * 8, 0, fix);
233 }
234
235 /*
236 * This function generates check information for a block.
237 * data is the block to be checked. bc is a pointer to the
238 * ocfs2_block_check structure describing the crc32 and the ecc.
239 *
240 * bc should be a pointer inside data, as the function will
241 * take care of zeroing it before calculating the check information. If
242 * bc does not point inside data, the caller must make sure any inline
243 * ocfs2_block_check structures are zeroed.
244 *
245 * The data buffer must be in on-disk endian (little endian for ocfs2).
246 * bc will be filled with little-endian values and will be ready to go to
247 * disk.
248 */
249 void ocfs2_block_check_compute(void *data, size_t blocksize,
250 struct ocfs2_block_check *bc)
251 {
252 u32 crc;
253 u32 ecc;
254
255 memset(bc, 0, sizeof(struct ocfs2_block_check));
256
257 crc = crc32_le(~0, data, blocksize);
258 ecc = ocfs2_hamming_encode_block(data, blocksize);
259
260 /*
261 * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no
262 * larger than 16 bits.
263 */
264 BUG_ON(ecc > USHORT_MAX);
265
266 bc->bc_crc32e = cpu_to_le32(crc);
267 bc->bc_ecc = cpu_to_le16((u16)ecc);
268 }
269
270 /*
271 * This function validates existing check information. Like _compute,
272 * the function will take care of zeroing bc before calculating check codes.
273 * If bc is not a pointer inside data, the caller must have zeroed any
274 * inline ocfs2_block_check structures.
275 *
276 * Again, the data passed in should be the on-disk endian.
277 */
278 int ocfs2_block_check_validate(void *data, size_t blocksize,
279 struct ocfs2_block_check *bc)
280 {
281 int rc = 0;
282 struct ocfs2_block_check check;
283 u32 crc, ecc;
284
285 check.bc_crc32e = le32_to_cpu(bc->bc_crc32e);
286 check.bc_ecc = le16_to_cpu(bc->bc_ecc);
287
288 memset(bc, 0, sizeof(struct ocfs2_block_check));
289
290 /* Fast path - if the crc32 validates, we're good to go */
291 crc = crc32_le(~0, data, blocksize);
292 if (crc == check.bc_crc32e)
293 goto out;
294
295 /* Ok, try ECC fixups */
296 ecc = ocfs2_hamming_encode_block(data, blocksize);
297 ocfs2_hamming_fix_block(data, blocksize, ecc ^ check.bc_ecc);
298
299 /* And check the crc32 again */
300 crc = crc32_le(~0, data, blocksize);
301 if (crc == check.bc_crc32e)
302 goto out;
303
304 rc = -EIO;
305
306 out:
307 bc->bc_crc32e = cpu_to_le32(check.bc_crc32e);
308 bc->bc_ecc = cpu_to_le16(check.bc_ecc);
309
310 return rc;
311 }
312
313 /*
314 * This function generates check information for a list of buffer_heads.
315 * bhs is the blocks to be checked. bc is a pointer to the
316 * ocfs2_block_check structure describing the crc32 and the ecc.
317 *
318 * bc should be a pointer inside data, as the function will
319 * take care of zeroing it before calculating the check information. If
320 * bc does not point inside data, the caller must make sure any inline
321 * ocfs2_block_check structures are zeroed.
322 *
323 * The data buffer must be in on-disk endian (little endian for ocfs2).
324 * bc will be filled with little-endian values and will be ready to go to
325 * disk.
326 */
327 void ocfs2_block_check_compute_bhs(struct buffer_head **bhs, int nr,
328 struct ocfs2_block_check *bc)
329 {
330 int i;
331 u32 crc, ecc;
332
333 BUG_ON(nr < 0);
334
335 if (!nr)
336 return;
337
338 memset(bc, 0, sizeof(struct ocfs2_block_check));
339
340 for (i = 0, crc = ~0, ecc = 0; i < nr; i++) {
341 crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size);
342 /*
343 * The number of bits in a buffer is obviously b_size*8.
344 * The offset of this buffer is b_size*i, so the bit offset
345 * of this buffer is b_size*8*i.
346 */
347 ecc = (u16)ocfs2_hamming_encode(ecc, bhs[i]->b_data,
348 bhs[i]->b_size * 8,
349 bhs[i]->b_size * 8 * i);
350 }
351
352 /*
353 * No ecc'd ocfs2 structure is larger than 4K, so ecc will be no
354 * larger than 16 bits.
355 */
356 BUG_ON(ecc > USHORT_MAX);
357
358 bc->bc_crc32e = cpu_to_le32(crc);
359 bc->bc_ecc = cpu_to_le16((u16)ecc);
360 }
361
362 /*
363 * This function validates existing check information on a list of
364 * buffer_heads. Like _compute_bhs, the function will take care of
365 * zeroing bc before calculating check codes. If bc is not a pointer
366 * inside data, the caller must have zeroed any inline
367 * ocfs2_block_check structures.
368 *
369 * Again, the data passed in should be the on-disk endian.
370 */
371 int ocfs2_block_check_validate_bhs(struct buffer_head **bhs, int nr,
372 struct ocfs2_block_check *bc)
373 {
374 int i, rc = 0;
375 struct ocfs2_block_check check;
376 u32 crc, ecc, fix;
377
378 BUG_ON(nr < 0);
379
380 if (!nr)
381 return 0;
382
383 check.bc_crc32e = le32_to_cpu(bc->bc_crc32e);
384 check.bc_ecc = le16_to_cpu(bc->bc_ecc);
385
386 memset(bc, 0, sizeof(struct ocfs2_block_check));
387
388 /* Fast path - if the crc32 validates, we're good to go */
389 for (i = 0, crc = ~0; i < nr; i++)
390 crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size);
391 if (crc == check.bc_crc32e)
392 goto out;
393
394 mlog(ML_ERROR,
395 "CRC32 failed: stored: %u, computed %u. Applying ECC.\n",
396 (unsigned int)check.bc_crc32e, (unsigned int)crc);
397
398 /* Ok, try ECC fixups */
399 for (i = 0, ecc = 0; i < nr; i++) {
400 /*
401 * The number of bits in a buffer is obviously b_size*8.
402 * The offset of this buffer is b_size*i, so the bit offset
403 * of this buffer is b_size*8*i.
404 */
405 ecc = (u16)ocfs2_hamming_encode(ecc, bhs[i]->b_data,
406 bhs[i]->b_size * 8,
407 bhs[i]->b_size * 8 * i);
408 }
409 fix = ecc ^ check.bc_ecc;
410 for (i = 0; i < nr; i++) {
411 /*
412 * Try the fix against each buffer. It will only affect
413 * one of them.
414 */
415 ocfs2_hamming_fix(bhs[i]->b_data, bhs[i]->b_size * 8,
416 bhs[i]->b_size * 8 * i, fix);
417 }
418
419 /* And check the crc32 again */
420 for (i = 0, crc = ~0; i < nr; i++)
421 crc = crc32_le(crc, bhs[i]->b_data, bhs[i]->b_size);
422 if (crc == check.bc_crc32e)
423 goto out;
424
425 mlog(ML_ERROR, "Fixed CRC32 failed: stored: %u, computed %u\n",
426 (unsigned int)check.bc_crc32e, (unsigned int)crc);
427
428 rc = -EIO;
429
430 out:
431 bc->bc_crc32e = cpu_to_le32(check.bc_crc32e);
432 bc->bc_ecc = cpu_to_le16(check.bc_ecc);
433
434 return rc;
435 }
436
437 /*
438 * These are the main API. They check the superblock flag before
439 * calling the underlying operations.
440 *
441 * They expect the buffer(s) to be in disk format.
442 */
443 void ocfs2_compute_meta_ecc(struct super_block *sb, void *data,
444 struct ocfs2_block_check *bc)
445 {
446 if (ocfs2_meta_ecc(OCFS2_SB(sb)))
447 ocfs2_block_check_compute(data, sb->s_blocksize, bc);
448 }
449
450 int ocfs2_validate_meta_ecc(struct super_block *sb, void *data,
451 struct ocfs2_block_check *bc)
452 {
453 int rc = 0;
454
455 if (ocfs2_meta_ecc(OCFS2_SB(sb)))
456 rc = ocfs2_block_check_validate(data, sb->s_blocksize, bc);
457
458 return rc;
459 }
460
461 void ocfs2_compute_meta_ecc_bhs(struct super_block *sb,
462 struct buffer_head **bhs, int nr,
463 struct ocfs2_block_check *bc)
464 {
465 if (ocfs2_meta_ecc(OCFS2_SB(sb)))
466 ocfs2_block_check_compute_bhs(bhs, nr, bc);
467 }
468
469 int ocfs2_validate_meta_ecc_bhs(struct super_block *sb,
470 struct buffer_head **bhs, int nr,
471 struct ocfs2_block_check *bc)
472 {
473 int rc = 0;
474
475 if (ocfs2_meta_ecc(OCFS2_SB(sb)))
476 rc = ocfs2_block_check_validate_bhs(bhs, nr, bc);
477
478 return rc;
479 }
480
This page took 0.041023 seconds and 4 git commands to generate.