1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Bad block management
4 *
5 * - Heavily based on MD badblocks code from Neil Brown
6 *
7 * Copyright (c) 2015, Intel Corporation.
8 */
9
10 #include <linux/badblocks.h>
11 #include <linux/seqlock.h>
12 #include <linux/device.h>
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/stddef.h>
16 #include <linux/types.h>
17 #include <linux/slab.h>
18
19 /**
20 * badblocks_check() - check a given range for bad sectors
21 * @bb: the badblocks structure that holds all badblock information
22 * @s: sector (start) at which to check for badblocks
23 * @sectors: number of sectors to check for badblocks
24 * @first_bad: pointer to store location of the first badblock
25 * @bad_sectors: pointer to store number of badblocks after @first_bad
26 *
27 * We can record which blocks on each device are 'bad' and so just
28 * fail those blocks, or that stripe, rather than the whole device.
29 * Entries in the bad-block table are 64bits wide. This comprises:
30 * Length of bad-range, in sectors: 0-511 for lengths 1-512
31 * Start of bad-range, sector offset, 54 bits (allows 8 exbibytes)
32 * A 'shift' can be set so that larger blocks are tracked and
33 * consequently larger devices can be covered.
34 * 'Acknowledged' flag - 1 bit. - the most significant bit.
35 *
36 * Locking of the bad-block table uses a seqlock so badblocks_check
37 * might need to retry if it is very unlucky.
38 * We will sometimes want to check for bad blocks in a bi_end_io function,
39 * so we use the write_seqlock_irq variant.
40 *
41 * When looking for a bad block we specify a range and want to
42 * know if any block in the range is bad. So we binary-search
43 * to the last range that starts at-or-before the given endpoint,
44 * (or "before the sector after the target range")
45 * then see if it ends after the given start.
46 *
47 * Return:
48 * 0: there are no known bad blocks in the range
49 * 1: there are known bad block which are all acknowledged
50 * -1: there are bad blocks which have not yet been acknowledged in metadata.
51 * plus the start/length of the first bad section we overlap.
52 */
53 int badblocks_check(struct badblocks *bb, sector_t s, int sectors,
54 sector_t *first_bad, int *bad_sectors)
55 {
56 int hi;
57 int lo;
58 u64 *p = bb->page;
59 int rv;
60 sector_t target = s + sectors;
61 unsigned seq;
62
63 if (bb->shift > 0) {
64 /* round the start down, and the end up */
65 s >>= bb->shift;
66 target += (1<<bb->shift) - 1;
67 target >>= bb->shift;
68 sectors = target - s;
69 }
70 /* 'target' is now the first block after the bad range */
71
72 retry:
73 seq = read_seqbegin(&bb->lock);
74 lo = 0;
75 rv = 0;
76 hi = bb->count;
77
78 /* Binary search between lo and hi for 'target'
79 * i.e. for the last range that starts before 'target'
80 */
81 /* INVARIANT: ranges before 'lo' and at-or-after 'hi'
82 * are known not to be the last range before target.
83 * VARIANT: hi-lo is the number of possible
84 * ranges, and decreases until it reaches 1
85 */
86 while (hi - lo > 1) {
87 int mid = (lo + hi) / 2;
88 sector_t a = BB_OFFSET(p[mid]);
89
90 if (a < target)
91 /* This could still be the one, earlier ranges
92 * could not.
93 */
94 lo = mid;
95 else
96 /* This and later ranges are definitely out. */
97 hi = mid;
98 }
99 /* 'lo' might be the last that started before target, but 'hi' isn't */
100 if (hi > lo) {
101 /* need to check all range that end after 's' to see if
102 * any are unacknowledged.
103 */
104 while (lo >= 0 &&
105 BB_OFFSET(p[lo]) + BB_LEN(p[lo]) > s) {
106 if (BB_OFFSET(p[lo]) < target) {
107 /* starts before the end, and finishes after
108 * the start, so they must overlap
109 */
110 if (rv != -1 && BB_ACK(p[lo]))
111 rv = 1;
112 else
113 rv = -1;
114 *first_bad = BB_OFFSET(p[lo]);
115 *bad_sectors = BB_LEN(p[lo]);
116 }
117 lo--;
118 }
119 }
120
121 if (read_seqretry(&bb->lock, seq))
122 goto retry;
123
124 return rv;
125 }
126 EXPORT_SYMBOL_GPL(badblocks_check);
127
128 static void badblocks_update_acked(struct badblocks *bb)
129 {
130 u64 *p = bb->page;
131 int i;
132 bool unacked = false;
133
134 if (!bb->unacked_exist)
135 return;
136
137 for (i = 0; i < bb->count ; i++) {
138 if (!BB_ACK(p[i])) {
139 unacked = true;
140 break;
141 }
142 }
143
144 if (!unacked)
145 bb->unacked_exist = 0;
146 }
147
148 /**
149 * badblocks_set() - Add a range of bad blocks to the table.
150 * @bb: the badblocks structure that holds all badblock information
151 * @s: first sector to mark as bad
152 * @sectors: number of sectors to mark as bad
153 * @acknowledged: weather to mark the bad sectors as acknowledged
154 *
155 * This might extend the table, or might contract it if two adjacent ranges
156 * can be merged. We binary-search to find the 'insertion' point, then
157 * decide how best to handle it.
158 *
159 * Return:
160 * 0: success
161 * 1: failed to set badblocks (out of space)
162 */
163 int badblocks_set(struct badblocks *bb, sector_t s, int sectors,
164 int acknowledged)
165 {
166 u64 *p;
167 int lo, hi;
168 int rv = 0;
169 unsigned long flags;
170
171 if (bb->shift < 0)
172 /* badblocks are disabled */
173 return 1;
174
175 if (bb->shift) {
176 /* round the start down, and the end up */
177 sector_t next = s + sectors;
178
179 s >>= bb->shift;
180 next += (1<<bb->shift) - 1;
181 next >>= bb->shift;
182 sectors = next - s;
183 }
184
185 write_seqlock_irqsave(&bb->lock, flags);
186
187 p = bb->page;
188 lo = 0;
189 hi = bb->count;
190 /* Find the last range that starts at-or-before 's' */
191 while (hi - lo > 1) {
192 int mid = (lo + hi) / 2;
193 sector_t a = BB_OFFSET(p[mid]);
194
195 if (a <= s)
196 lo = mid;
197 else
198 hi = mid;
199 }
200 if (hi > lo && BB_OFFSET(p[lo]) > s)
201 hi = lo;
202
203 if (hi > lo) {
204 /* we found a range that might merge with the start
205 * of our new range
206 */
207 sector_t a = BB_OFFSET(p[lo]);
208 sector_t e = a + BB_LEN(p[lo]);
209 int ack = BB_ACK(p[lo]);
210
211 if (e >= s) {
212 /* Yes, we can merge with a previous range */
213 if (s == a && s + sectors >= e)
214 /* new range covers old */
215 ack = acknowledged;
216 else
217 ack = ack && acknowledged;
218
219 if (e < s + sectors)
220 e = s + sectors;
221 if (e - a <= BB_MAX_LEN) {
222 p[lo] = BB_MAKE(a, e-a, ack);
223 s = e;
224 } else {
225 /* does not all fit in one range,
226 * make p[lo] maximal
227 */
228 if (BB_LEN(p[lo]) != BB_MAX_LEN)
229 p[lo] = BB_MAKE(a, BB_MAX_LEN, ack);
230 s = a + BB_MAX_LEN;
231 }
232 sectors = e - s;
233 }
234 }
235 if (sectors && hi < bb->count) {
236 /* 'hi' points to the first range that starts after 's'.
237 * Maybe we can merge with the start of that range
238 */
239 sector_t a = BB_OFFSET(p[hi]);
240 sector_t e = a + BB_LEN(p[hi]);
241 int ack = BB_ACK(p[hi]);
242
243 if (a <= s + sectors) {
244 /* merging is possible */
245 if (e <= s + sectors) {
246 /* full overlap */
247 e = s + sectors;
248 ack = acknowledged;
249 } else
250 ack = ack && acknowledged;
251
252 a = s;
253 if (e - a <= BB_MAX_LEN) {
254 p[hi] = BB_MAKE(a, e-a, ack);
255 s = e;
256 } else {
257 p[hi] = BB_MAKE(a, BB_MAX_LEN, ack);
258 s = a + BB_MAX_LEN;
259 }
260 sectors = e - s;
261 lo = hi;
262 hi++;
263 }
264 }
265 if (sectors == 0 && hi < bb->count) {
266 /* we might be able to combine lo and hi */
267 /* Note: 's' is at the end of 'lo' */
268 sector_t a = BB_OFFSET(p[hi]);
269 int lolen = BB_LEN(p[lo]);
270 int hilen = BB_LEN(p[hi]);
271 int newlen = lolen + hilen - (s - a);
272
273 if (s >= a && newlen < BB_MAX_LEN) {
274 /* yes, we can combine them */
275 int ack = BB_ACK(p[lo]) && BB_ACK(p[hi]);
276
277 p[lo] = BB_MAKE(BB_OFFSET(p[lo]), newlen, ack);
278 memmove(p + hi, p + hi + 1,
279 (bb->count - hi - 1) * 8);
280 bb->count--;
281 }
282 }
283 while (sectors) {
284 /* didn't merge (it all).
285 * Need to add a range just before 'hi'
286 */
287 if (bb->count >= MAX_BADBLOCKS) {
288 /* No room for more */
289 rv = 1;
290 break;
291 } else {
292 int this_sectors = sectors;
293
294 memmove(p + hi + 1, p + hi,
295 (bb->count - hi) * 8);
296 bb->count++;
297
298 if (this_sectors > BB_MAX_LEN)
299 this_sectors = BB_MAX_LEN;
300 p[hi] = BB_MAKE(s, this_sectors, acknowledged);
301 sectors -= this_sectors;
302 s += this_sectors;
303 }
304 }
305
306 bb->changed = 1;
307 if (!acknowledged)
308 bb->unacked_exist = 1;
309 else
310 badblocks_update_acked(bb);
311 write_sequnlock_irqrestore(&bb->lock, flags);
312
313 return rv;
314 }
315 EXPORT_SYMBOL_GPL(badblocks_set);
316
317 /**
318 * badblocks_clear() - Remove a range of bad blocks to the table.
319 * @bb: the badblocks structure that holds all badblock information
320 * @s: first sector to mark as bad
321 * @sectors: number of sectors to mark as bad
322 *
323 * This may involve extending the table if we spilt a region,
324 * but it must not fail. So if the table becomes full, we just
325 * drop the remove request.
326 *
327 * Return:
328 * 0: success
329 * 1: failed to clear badblocks
330 */
331 int badblocks_clear(struct badblocks *bb, sector_t s, int sectors)
332 {
333 u64 *p;
334 int lo, hi;
335 sector_t target = s + sectors;
336 int rv = 0;
337
338 if (bb->shift > 0) {
339 /* When clearing we round the start up and the end down.
340 * This should not matter as the shift should align with
341 * the block size and no rounding should ever be needed.
342 * However it is better the think a block is bad when it
343 * isn't than to think a block is not bad when it is.
344 */
345 s += (1<<bb->shift) - 1;
346 s >>= bb->shift;
347 target >>= bb->shift;
348 sectors = target - s;
349 }
350
351 write_seqlock_irq(&bb->lock);
352
353 p = bb->page;
354 lo = 0;
355 hi = bb->count;
356 /* Find the last range that starts before 'target' */
357 while (hi - lo > 1) {
358 int mid = (lo + hi) / 2;
359 sector_t a = BB_OFFSET(p[mid]);
360
361 if (a < target)
362 lo = mid;
363 else
364 hi = mid;
365 }
366 if (hi > lo) {
367 /* p[lo] is the last range that could overlap the
368 * current range. Earlier ranges could also overlap,
369 * but only this one can overlap the end of the range.
370 */
371 if ((BB_OFFSET(p[lo]) + BB_LEN(p[lo]) > target) &&
372 (BB_OFFSET(p[lo]) < target)) {
373 /* Partial overlap, leave the tail of this range */
374 int ack = BB_ACK(p[lo]);
375 sector_t a = BB_OFFSET(p[lo]);
376 sector_t end = a + BB_LEN(p[lo]);
377
378 if (a < s) {
379 /* we need to split this range */
380 if (bb->count >= MAX_BADBLOCKS) {
381 rv = -ENOSPC;
382 goto out;
383 }
384 memmove(p+lo+1, p+lo, (bb->count - lo) * 8);
385 bb->count++;
386 p[lo] = BB_MAKE(a, s-a, ack);
387 lo++;
388 }
389 p[lo] = BB_MAKE(target, end - target, ack);
390 /* there is no longer an overlap */
391 hi = lo;
392 lo--;
393 }
394 while (lo >= 0 &&
395 (BB_OFFSET(p[lo]) + BB_LEN(p[lo]) > s) &&
396 (BB_OFFSET(p[lo]) < target)) {
397 /* This range does overlap */
398 if (BB_OFFSET(p[lo]) < s) {
399 /* Keep the early parts of this range. */
400 int ack = BB_ACK(p[lo]);
401 sector_t start = BB_OFFSET(p[lo]);
402
403 p[lo] = BB_MAKE(start, s - start, ack);
404 /* now low doesn't overlap, so.. */
405 break;
406 }
407 lo--;
408 }
409 /* 'lo' is strictly before, 'hi' is strictly after,
410 * anything between needs to be discarded
411 */
412 if (hi - lo > 1) {
413 memmove(p+lo+1, p+hi, (bb->count - hi) * 8);
414 bb->count -= (hi - lo - 1);
415 }
416 }
417
418 badblocks_update_acked(bb);
419 bb->changed = 1;
420 out:
421 write_sequnlock_irq(&bb->lock);
422 return rv;
423 }
424 EXPORT_SYMBOL_GPL(badblocks_clear);
425
426 /**
427 * ack_all_badblocks() - Acknowledge all bad blocks in a list.
428 * @bb: the badblocks structure that holds all badblock information
429 *
430 * This only succeeds if ->changed is clear. It is used by
431 * in-kernel metadata updates
432 */
433 void ack_all_badblocks(struct badblocks *bb)
434 {
435 if (bb->page == NULL || bb->changed)
436 /* no point even trying */
437 return;
438 write_seqlock_irq(&bb->lock);
439
440 if (bb->changed == 0 && bb->unacked_exist) {
441 u64 *p = bb->page;
442 int i;
443
444 for (i = 0; i < bb->count ; i++) {
445 if (!BB_ACK(p[i])) {
446 sector_t start = BB_OFFSET(p[i]);
447 int len = BB_LEN(p[i]);
448
449 p[i] = BB_MAKE(start, len, 1);
450 }
451 }
452 bb->unacked_exist = 0;
453 }
454 write_sequnlock_irq(&bb->lock);
455 }
456 EXPORT_SYMBOL_GPL(ack_all_badblocks);
457
458 /**
459 * badblocks_show() - sysfs access to bad-blocks list
460 * @bb: the badblocks structure that holds all badblock information
461 * @page: buffer received from sysfs
462 * @unack: weather to show unacknowledged badblocks
463 *
464 * Return:
465 * Length of returned data
466 */
467 ssize_t badblocks_show(struct badblocks *bb, char *page, int unack)
468 {
469 size_t len;
470 int i;
471 u64 *p = bb->page;
472 unsigned seq;
473
474 if (bb->shift < 0)
475 return 0;
476
477 retry:
478 seq = read_seqbegin(&bb->lock);
479
480 len = 0;
481 i = 0;
482
483 while (len < PAGE_SIZE && i < bb->count) {
484 sector_t s = BB_OFFSET(p[i]);
485 unsigned int length = BB_LEN(p[i]);
486 int ack = BB_ACK(p[i]);
487
488 i++;
489
490 if (unack && ack)
491 continue;
492
493 len += snprintf(page+len, PAGE_SIZE-len, "%llu %u\n",
494 (unsigned long long)s << bb->shift,
495 length << bb->shift);
496 }
497 if (unack && len == 0)
498 bb->unacked_exist = 0;
499
500 if (read_seqretry(&bb->lock, seq))
501 goto retry;
502
503 return len;
504 }
505 EXPORT_SYMBOL_GPL(badblocks_show);
506
507 /**
508 * badblocks_store() - sysfs access to bad-blocks list
509 * @bb: the badblocks structure that holds all badblock information
510 * @page: buffer received from sysfs
511 * @len: length of data received from sysfs
512 * @unack: weather to show unacknowledged badblocks
513 *
514 * Return:
515 * Length of the buffer processed or -ve error.
516 */
517 ssize_t badblocks_store(struct badblocks *bb, const char *page, size_t len,
518 int unack)
519 {
520 unsigned long long sector;
521 int length;
522 char newline;
523
524 switch (sscanf(page, "%llu %d%c", §or, &length, &newline)) {
525 case 3:
526 if (newline != '\n')
527 return -EINVAL;
528 /* fall through */
529 case 2:
530 if (length <= 0)
531 return -EINVAL;
532 break;
533 default:
534 return -EINVAL;
535 }
536
537 if (badblocks_set(bb, sector, length, !unack))
538 return -ENOSPC;
539 else
540 return len;
541 }
542 EXPORT_SYMBOL_GPL(badblocks_store);
543
544 static int __badblocks_init(struct device *dev, struct badblocks *bb,
545 int enable)
546 {
547 bb->dev = dev;
548 bb->count = 0;
549 if (enable)
550 bb->shift = 0;
551 else
552 bb->shift = -1;
553 if (dev)
554 bb->page = devm_kzalloc(dev, PAGE_SIZE, GFP_KERNEL);
555 else
556 bb->page = kzalloc(PAGE_SIZE, GFP_KERNEL);
557 if (!bb->page) {
558 bb->shift = -1;
559 return -ENOMEM;
560 }
561 seqlock_init(&bb->lock);
562
563 return 0;
564 }
565
566 /**
567 * badblocks_init() - initialize the badblocks structure
568 * @bb: the badblocks structure that holds all badblock information
569 * @enable: weather to enable badblocks accounting
570 *
571 * Return:
572 * 0: success
573 * -ve errno: on error
574 */
575 int badblocks_init(struct badblocks *bb, int enable)
576 {
577 return __badblocks_init(NULL, bb, enable);
578 }
579 EXPORT_SYMBOL_GPL(badblocks_init);
580
581 int devm_init_badblocks(struct device *dev, struct badblocks *bb)
582 {
583 if (!bb)
584 return -EINVAL;
585 return __badblocks_init(dev, bb, 1);
586 }
587 EXPORT_SYMBOL_GPL(devm_init_badblocks);
588
589 /**
590 * badblocks_exit() - free the badblocks structure
591 * @bb: the badblocks structure that holds all badblock information
592 */
593 void badblocks_exit(struct badblocks *bb)
594 {
595 if (!bb)
596 return;
597 if (bb->dev)
598 devm_kfree(bb->dev, bb->page);
599 else
600 kfree(bb->page);
601 bb->page = NULL;
602 }
603 EXPORT_SYMBOL_GPL(badblocks_exit);