Merge tag 'hwmon-for-linus-v4.7-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
[deliverable/linux.git] / drivers / mtd / ubi / eba.c
CommitLineData
801c135c
AB
1/*
2 * Copyright (c) International Business Machines Corp., 2006
3 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
8 *
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
12 * the GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17 *
18 * Author: Artem Bityutskiy (Битюцкий Артём)
19 */
20
21/*
85c6e6e2 22 * The UBI Eraseblock Association (EBA) sub-system.
801c135c 23 *
85c6e6e2 24 * This sub-system is responsible for I/O to/from logical eraseblock.
801c135c
AB
25 *
26 * Although in this implementation the EBA table is fully kept and managed in
27 * RAM, which assumes poor scalability, it might be (partially) maintained on
28 * flash in future implementations.
29 *
85c6e6e2
AB
30 * The EBA sub-system implements per-logical eraseblock locking. Before
31 * accessing a logical eraseblock it is locked for reading or writing. The
32 * per-logical eraseblock locking is implemented by means of the lock tree. The
33 * lock tree is an RB-tree which refers all the currently locked logical
34 * eraseblocks. The lock tree elements are &struct ubi_ltree_entry objects.
35 * They are indexed by (@vol_id, @lnum) pairs.
801c135c
AB
36 *
37 * EBA also maintains the global sequence counter which is incremented each
38 * time a logical eraseblock is mapped to a physical eraseblock and it is
39 * stored in the volume identifier header. This means that each VID header has
40 * a unique sequence number. The sequence number is only increased an we assume
41 * 64 bits is enough to never overflow.
42 */
43
44#include <linux/slab.h>
45#include <linux/crc32.h>
46#include <linux/err.h>
47#include "ubi.h"
48
e8823bd6
AB
49/* Number of physical eraseblocks reserved for atomic LEB change operation */
50#define EBA_RESERVED_PEBS 1
51
801c135c
AB
52/**
53 * next_sqnum - get next sequence number.
54 * @ubi: UBI device description object
55 *
56 * This function returns next sequence number to use, which is just the current
57 * global sequence counter value. It also increases the global sequence
58 * counter.
59 */
a7306653 60unsigned long long ubi_next_sqnum(struct ubi_device *ubi)
801c135c
AB
61{
62 unsigned long long sqnum;
63
64 spin_lock(&ubi->ltree_lock);
65 sqnum = ubi->global_sqnum++;
66 spin_unlock(&ubi->ltree_lock);
67
68 return sqnum;
69}
70
71/**
72 * ubi_get_compat - get compatibility flags of a volume.
73 * @ubi: UBI device description object
74 * @vol_id: volume ID
75 *
76 * This function returns compatibility flags for an internal volume. User
77 * volumes have no compatibility flags, so %0 is returned.
78 */
79static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
80{
91f2d53c 81 if (vol_id == UBI_LAYOUT_VOLUME_ID)
801c135c
AB
82 return UBI_LAYOUT_VOLUME_COMPAT;
83 return 0;
84}
85
86/**
87 * ltree_lookup - look up the lock tree.
88 * @ubi: UBI device description object
89 * @vol_id: volume ID
90 * @lnum: logical eraseblock number
91 *
3a8d4642 92 * This function returns a pointer to the corresponding &struct ubi_ltree_entry
801c135c
AB
93 * object if the logical eraseblock is locked and %NULL if it is not.
94 * @ubi->ltree_lock has to be locked.
95 */
3a8d4642
AB
96static struct ubi_ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
97 int lnum)
801c135c
AB
98{
99 struct rb_node *p;
100
101 p = ubi->ltree.rb_node;
102 while (p) {
3a8d4642 103 struct ubi_ltree_entry *le;
801c135c 104
3a8d4642 105 le = rb_entry(p, struct ubi_ltree_entry, rb);
801c135c
AB
106
107 if (vol_id < le->vol_id)
108 p = p->rb_left;
109 else if (vol_id > le->vol_id)
110 p = p->rb_right;
111 else {
112 if (lnum < le->lnum)
113 p = p->rb_left;
114 else if (lnum > le->lnum)
115 p = p->rb_right;
116 else
117 return le;
118 }
119 }
120
121 return NULL;
122}
123
124/**
125 * ltree_add_entry - add new entry to the lock tree.
126 * @ubi: UBI device description object
127 * @vol_id: volume ID
128 * @lnum: logical eraseblock number
129 *
130 * This function adds new entry for logical eraseblock (@vol_id, @lnum) to the
131 * lock tree. If such entry is already there, its usage counter is increased.
132 * Returns pointer to the lock tree entry or %-ENOMEM if memory allocation
133 * failed.
134 */
3a8d4642
AB
135static struct ubi_ltree_entry *ltree_add_entry(struct ubi_device *ubi,
136 int vol_id, int lnum)
801c135c 137{
3a8d4642 138 struct ubi_ltree_entry *le, *le1, *le_free;
801c135c 139
b9a06623 140 le = kmalloc(sizeof(struct ubi_ltree_entry), GFP_NOFS);
801c135c
AB
141 if (!le)
142 return ERR_PTR(-ENOMEM);
143
b9a06623
AB
144 le->users = 0;
145 init_rwsem(&le->mutex);
801c135c
AB
146 le->vol_id = vol_id;
147 le->lnum = lnum;
148
149 spin_lock(&ubi->ltree_lock);
150 le1 = ltree_lookup(ubi, vol_id, lnum);
151
152 if (le1) {
153 /*
154 * This logical eraseblock is already locked. The newly
155 * allocated lock entry is not needed.
156 */
157 le_free = le;
158 le = le1;
159 } else {
160 struct rb_node **p, *parent = NULL;
161
162 /*
163 * No lock entry, add the newly allocated one to the
164 * @ubi->ltree RB-tree.
165 */
166 le_free = NULL;
167
168 p = &ubi->ltree.rb_node;
169 while (*p) {
170 parent = *p;
3a8d4642 171 le1 = rb_entry(parent, struct ubi_ltree_entry, rb);
801c135c
AB
172
173 if (vol_id < le1->vol_id)
174 p = &(*p)->rb_left;
175 else if (vol_id > le1->vol_id)
176 p = &(*p)->rb_right;
177 else {
178 ubi_assert(lnum != le1->lnum);
179 if (lnum < le1->lnum)
180 p = &(*p)->rb_left;
181 else
182 p = &(*p)->rb_right;
183 }
184 }
185
186 rb_link_node(&le->rb, parent, p);
187 rb_insert_color(&le->rb, &ubi->ltree);
188 }
189 le->users += 1;
190 spin_unlock(&ubi->ltree_lock);
191
9c9ec147 192 kfree(le_free);
801c135c
AB
193 return le;
194}
195
196/**
197 * leb_read_lock - lock logical eraseblock for reading.
198 * @ubi: UBI device description object
199 * @vol_id: volume ID
200 * @lnum: logical eraseblock number
201 *
202 * This function locks a logical eraseblock for reading. Returns zero in case
203 * of success and a negative error code in case of failure.
204 */
205static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
206{
3a8d4642 207 struct ubi_ltree_entry *le;
801c135c
AB
208
209 le = ltree_add_entry(ubi, vol_id, lnum);
210 if (IS_ERR(le))
211 return PTR_ERR(le);
212 down_read(&le->mutex);
213 return 0;
214}
215
216/**
217 * leb_read_unlock - unlock logical eraseblock.
218 * @ubi: UBI device description object
219 * @vol_id: volume ID
220 * @lnum: logical eraseblock number
221 */
222static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
223{
3a8d4642 224 struct ubi_ltree_entry *le;
801c135c
AB
225
226 spin_lock(&ubi->ltree_lock);
227 le = ltree_lookup(ubi, vol_id, lnum);
228 le->users -= 1;
229 ubi_assert(le->users >= 0);
23add745 230 up_read(&le->mutex);
801c135c
AB
231 if (le->users == 0) {
232 rb_erase(&le->rb, &ubi->ltree);
23add745 233 kfree(le);
801c135c
AB
234 }
235 spin_unlock(&ubi->ltree_lock);
801c135c
AB
236}
237
238/**
239 * leb_write_lock - lock logical eraseblock for writing.
240 * @ubi: UBI device description object
241 * @vol_id: volume ID
242 * @lnum: logical eraseblock number
243 *
244 * This function locks a logical eraseblock for writing. Returns zero in case
245 * of success and a negative error code in case of failure.
246 */
247static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
248{
3a8d4642 249 struct ubi_ltree_entry *le;
801c135c
AB
250
251 le = ltree_add_entry(ubi, vol_id, lnum);
252 if (IS_ERR(le))
253 return PTR_ERR(le);
254 down_write(&le->mutex);
255 return 0;
256}
257
43f9b25a
AB
258/**
259 * leb_write_lock - lock logical eraseblock for writing.
260 * @ubi: UBI device description object
261 * @vol_id: volume ID
262 * @lnum: logical eraseblock number
263 *
264 * This function locks a logical eraseblock for writing if there is no
265 * contention and does nothing if there is contention. Returns %0 in case of
266 * success, %1 in case of contention, and and a negative error code in case of
267 * failure.
268 */
269static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
270{
43f9b25a
AB
271 struct ubi_ltree_entry *le;
272
273 le = ltree_add_entry(ubi, vol_id, lnum);
274 if (IS_ERR(le))
275 return PTR_ERR(le);
276 if (down_write_trylock(&le->mutex))
277 return 0;
278
279 /* Contention, cancel */
280 spin_lock(&ubi->ltree_lock);
281 le->users -= 1;
282 ubi_assert(le->users >= 0);
283 if (le->users == 0) {
284 rb_erase(&le->rb, &ubi->ltree);
b9a06623 285 kfree(le);
23add745
AB
286 }
287 spin_unlock(&ubi->ltree_lock);
43f9b25a
AB
288
289 return 1;
290}
291
801c135c
AB
292/**
293 * leb_write_unlock - unlock logical eraseblock.
294 * @ubi: UBI device description object
295 * @vol_id: volume ID
296 * @lnum: logical eraseblock number
297 */
298static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
299{
3a8d4642 300 struct ubi_ltree_entry *le;
801c135c
AB
301
302 spin_lock(&ubi->ltree_lock);
303 le = ltree_lookup(ubi, vol_id, lnum);
304 le->users -= 1;
305 ubi_assert(le->users >= 0);
23add745 306 up_write(&le->mutex);
801c135c
AB
307 if (le->users == 0) {
308 rb_erase(&le->rb, &ubi->ltree);
b9a06623 309 kfree(le);
23add745
AB
310 }
311 spin_unlock(&ubi->ltree_lock);
801c135c
AB
312}
313
314/**
315 * ubi_eba_unmap_leb - un-map logical eraseblock.
316 * @ubi: UBI device description object
89b96b69 317 * @vol: volume description object
801c135c
AB
318 * @lnum: logical eraseblock number
319 *
320 * This function un-maps logical eraseblock @lnum and schedules corresponding
321 * physical eraseblock for erasure. Returns zero in case of success and a
322 * negative error code in case of failure.
323 */
89b96b69
AB
324int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
325 int lnum)
801c135c 326{
89b96b69 327 int err, pnum, vol_id = vol->vol_id;
801c135c
AB
328
329 if (ubi->ro_mode)
330 return -EROFS;
331
332 err = leb_write_lock(ubi, vol_id, lnum);
333 if (err)
334 return err;
335
336 pnum = vol->eba_tbl[lnum];
337 if (pnum < 0)
338 /* This logical eraseblock is already unmapped */
339 goto out_unlock;
340
341 dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
342
111ab0b2 343 down_read(&ubi->fm_eba_sem);
801c135c 344 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
111ab0b2 345 up_read(&ubi->fm_eba_sem);
d36e59e6 346 err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 0);
801c135c
AB
347
348out_unlock:
349 leb_write_unlock(ubi, vol_id, lnum);
350 return err;
351}
352
353/**
354 * ubi_eba_read_leb - read data.
355 * @ubi: UBI device description object
89b96b69 356 * @vol: volume description object
801c135c
AB
357 * @lnum: logical eraseblock number
358 * @buf: buffer to store the read data
359 * @offset: offset from where to read
360 * @len: how many bytes to read
361 * @check: data CRC check flag
362 *
363 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
364 * bytes. The @check flag only makes sense for static volumes and forces
365 * eraseblock data CRC checking.
366 *
367 * In case of success this function returns zero. In case of a static volume,
368 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
369 * returned for any volume type if an ECC error was detected by the MTD device
370 * driver. Other negative error cored may be returned in case of other errors.
371 */
89b96b69
AB
372int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
373 void *buf, int offset, int len, int check)
801c135c 374{
89b96b69 375 int err, pnum, scrub = 0, vol_id = vol->vol_id;
801c135c 376 struct ubi_vid_hdr *vid_hdr;
a6343afb 377 uint32_t uninitialized_var(crc);
801c135c
AB
378
379 err = leb_read_lock(ubi, vol_id, lnum);
380 if (err)
381 return err;
382
383 pnum = vol->eba_tbl[lnum];
384 if (pnum < 0) {
385 /*
386 * The logical eraseblock is not mapped, fill the whole buffer
387 * with 0xFF bytes. The exception is static volumes for which
388 * it is an error to read unmapped logical eraseblocks.
389 */
390 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
391 len, offset, vol_id, lnum);
392 leb_read_unlock(ubi, vol_id, lnum);
393 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
394 memset(buf, 0xFF, len);
395 return 0;
396 }
397
398 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
399 len, offset, vol_id, lnum, pnum);
400
401 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
402 check = 0;
403
404retry:
405 if (check) {
33818bbb 406 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
801c135c
AB
407 if (!vid_hdr) {
408 err = -ENOMEM;
409 goto out_unlock;
410 }
411
412 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
413 if (err && err != UBI_IO_BITFLIPS) {
414 if (err > 0) {
415 /*
416 * The header is either absent or corrupted.
417 * The former case means there is a bug -
418 * switch to read-only mode just in case.
419 * The latter case means a real corruption - we
420 * may try to recover data. FIXME: but this is
421 * not implemented.
422 */
756e1df1 423 if (err == UBI_IO_BAD_HDR_EBADMSG ||
eb89580e 424 err == UBI_IO_BAD_HDR) {
32608703 425 ubi_warn(ubi, "corrupted VID header at PEB %d, LEB %d:%d",
049333ce 426 pnum, vol_id, lnum);
801c135c 427 err = -EBADMSG;
b388e6a7 428 } else {
1900149c
RW
429 /*
430 * Ending up here in the non-Fastmap case
431 * is a clear bug as the VID header had to
432 * be present at scan time to have it referenced.
433 * With fastmap the story is more complicated.
434 * Fastmap has the mapping info without the need
435 * of a full scan. So the LEB could have been
436 * unmapped, Fastmap cannot know this and keeps
437 * the LEB referenced.
438 * This is valid and works as the layer above UBI
439 * has to do bookkeeping about used/referenced
440 * LEBs in any case.
441 */
442 if (ubi->fast_attach) {
443 err = -EBADMSG;
444 } else {
445 err = -EINVAL;
446 ubi_ro_mode(ubi);
447 }
b388e6a7 448 }
801c135c
AB
449 }
450 goto out_free;
451 } else if (err == UBI_IO_BITFLIPS)
452 scrub = 1;
453
3261ebd7
CH
454 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
455 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
801c135c 456
3261ebd7 457 crc = be32_to_cpu(vid_hdr->data_crc);
801c135c
AB
458 ubi_free_vid_hdr(ubi, vid_hdr);
459 }
460
461 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
462 if (err) {
170505f5 463 if (err == UBI_IO_BITFLIPS)
801c135c 464 scrub = 1;
170505f5 465 else if (mtd_is_eccerr(err)) {
801c135c
AB
466 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
467 goto out_unlock;
468 scrub = 1;
469 if (!check) {
32608703 470 ubi_msg(ubi, "force data checking");
801c135c
AB
471 check = 1;
472 goto retry;
473 }
474 } else
475 goto out_unlock;
476 }
477
478 if (check) {
2ab934b8 479 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
801c135c 480 if (crc1 != crc) {
32608703 481 ubi_warn(ubi, "CRC error: calculated %#08x, must be %#08x",
801c135c
AB
482 crc1, crc);
483 err = -EBADMSG;
484 goto out_unlock;
485 }
486 }
487
488 if (scrub)
489 err = ubi_wl_scrub_peb(ubi, pnum);
490
491 leb_read_unlock(ubi, vol_id, lnum);
492 return err;
493
494out_free:
495 ubi_free_vid_hdr(ubi, vid_hdr);
496out_unlock:
497 leb_read_unlock(ubi, vol_id, lnum);
498 return err;
499}
500
9ff08979
RW
501/**
502 * ubi_eba_read_leb_sg - read data into a scatter gather list.
503 * @ubi: UBI device description object
504 * @vol: volume description object
505 * @lnum: logical eraseblock number
506 * @sgl: UBI scatter gather list to store the read data
507 * @offset: offset from where to read
508 * @len: how many bytes to read
509 * @check: data CRC check flag
510 *
511 * This function works exactly like ubi_eba_read_leb(). But instead of
512 * storing the read data into a buffer it writes to an UBI scatter gather
513 * list.
514 */
515int ubi_eba_read_leb_sg(struct ubi_device *ubi, struct ubi_volume *vol,
516 struct ubi_sgl *sgl, int lnum, int offset, int len,
517 int check)
518{
519 int to_read;
520 int ret;
521 struct scatterlist *sg;
522
523 for (;;) {
524 ubi_assert(sgl->list_pos < UBI_MAX_SG_COUNT);
525 sg = &sgl->sg[sgl->list_pos];
526 if (len < sg->length - sgl->page_pos)
527 to_read = len;
528 else
529 to_read = sg->length - sgl->page_pos;
530
531 ret = ubi_eba_read_leb(ubi, vol, lnum,
532 sg_virt(sg) + sgl->page_pos, offset,
533 to_read, check);
534 if (ret < 0)
535 return ret;
536
537 offset += to_read;
538 len -= to_read;
539 if (!len) {
540 sgl->page_pos += to_read;
541 if (sgl->page_pos == sg->length) {
542 sgl->list_pos++;
543 sgl->page_pos = 0;
544 }
545
546 break;
547 }
548
549 sgl->list_pos++;
550 sgl->page_pos = 0;
551 }
552
553 return ret;
554}
555
801c135c
AB
556/**
557 * recover_peb - recover from write failure.
558 * @ubi: UBI device description object
559 * @pnum: the physical eraseblock to recover
560 * @vol_id: volume ID
561 * @lnum: logical eraseblock number
562 * @buf: data which was not written because of the write failure
563 * @offset: offset of the failed write
564 * @len: how many bytes should have been written
565 *
566 * This function is called in case of a write failure and moves all good data
567 * from the potentially bad physical eraseblock to a good physical eraseblock.
568 * This function also writes the data which was not written due to the failure.
569 * Returns new physical eraseblock number in case of success, and a negative
570 * error code in case of failure.
571 */
572static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
573 const void *buf, int offset, int len)
574{
575 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
576 struct ubi_volume *vol = ubi->volumes[idx];
577 struct ubi_vid_hdr *vid_hdr;
801c135c 578
33818bbb 579 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
9c9ec147 580 if (!vid_hdr)
801c135c 581 return -ENOMEM;
801c135c
AB
582
583retry:
b36a261e 584 new_pnum = ubi_wl_get_peb(ubi);
801c135c
AB
585 if (new_pnum < 0) {
586 ubi_free_vid_hdr(ubi, vid_hdr);
111ab0b2 587 up_read(&ubi->fm_eba_sem);
801c135c
AB
588 return new_pnum;
589 }
590
32608703
TB
591 ubi_msg(ubi, "recover PEB %d, move data to PEB %d",
592 pnum, new_pnum);
801c135c
AB
593
594 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
595 if (err && err != UBI_IO_BITFLIPS) {
596 if (err > 0)
597 err = -EIO;
111ab0b2 598 up_read(&ubi->fm_eba_sem);
801c135c
AB
599 goto out_put;
600 }
601
a7306653 602 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
801c135c 603 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
8fb2a514 604 if (err) {
111ab0b2 605 up_read(&ubi->fm_eba_sem);
801c135c 606 goto write_error;
8fb2a514 607 }
801c135c
AB
608
609 data_size = offset + len;
4df581f3 610 mutex_lock(&ubi->buf_mutex);
0ca39d74 611 memset(ubi->peb_buf + offset, 0xFF, len);
801c135c
AB
612
613 /* Read everything before the area where the write failure happened */
614 if (offset > 0) {
0ca39d74 615 err = ubi_io_read_data(ubi, ubi->peb_buf, pnum, 0, offset);
8fb2a514 616 if (err && err != UBI_IO_BITFLIPS) {
111ab0b2 617 up_read(&ubi->fm_eba_sem);
4df581f3 618 goto out_unlock;
8fb2a514 619 }
801c135c
AB
620 }
621
0ca39d74 622 memcpy(ubi->peb_buf + offset, buf, len);
801c135c 623
0ca39d74 624 err = ubi_io_write_data(ubi, ubi->peb_buf, new_pnum, 0, data_size);
4df581f3
AB
625 if (err) {
626 mutex_unlock(&ubi->buf_mutex);
111ab0b2 627 up_read(&ubi->fm_eba_sem);
801c135c 628 goto write_error;
4df581f3 629 }
801c135c 630
e88d6e10 631 mutex_unlock(&ubi->buf_mutex);
801c135c
AB
632 ubi_free_vid_hdr(ubi, vid_hdr);
633
634 vol->eba_tbl[lnum] = new_pnum;
111ab0b2 635 up_read(&ubi->fm_eba_sem);
d36e59e6 636 ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
801c135c 637
32608703 638 ubi_msg(ubi, "data was successfully recovered");
801c135c
AB
639 return 0;
640
4df581f3 641out_unlock:
e88d6e10 642 mutex_unlock(&ubi->buf_mutex);
4df581f3 643out_put:
d36e59e6 644 ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
801c135c
AB
645 ubi_free_vid_hdr(ubi, vid_hdr);
646 return err;
647
648write_error:
649 /*
650 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
651 * get another one.
652 */
32608703 653 ubi_warn(ubi, "failed to write to PEB %d", new_pnum);
d36e59e6 654 ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
801c135c
AB
655 if (++tries > UBI_IO_RETRIES) {
656 ubi_free_vid_hdr(ubi, vid_hdr);
657 return err;
658 }
32608703 659 ubi_msg(ubi, "try again");
801c135c
AB
660 goto retry;
661}
662
663/**
664 * ubi_eba_write_leb - write data to dynamic volume.
665 * @ubi: UBI device description object
89b96b69 666 * @vol: volume description object
801c135c
AB
667 * @lnum: logical eraseblock number
668 * @buf: the data to write
669 * @offset: offset within the logical eraseblock where to write
670 * @len: how many bytes to write
801c135c
AB
671 *
672 * This function writes data to logical eraseblock @lnum of a dynamic volume
89b96b69 673 * @vol. Returns zero in case of success and a negative error code in case
801c135c
AB
674 * of failure. In case of error, it is possible that something was still
675 * written to the flash media, but may be some garbage.
676 */
89b96b69 677int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
b36a261e 678 const void *buf, int offset, int len)
801c135c 679{
89b96b69 680 int err, pnum, tries = 0, vol_id = vol->vol_id;
801c135c
AB
681 struct ubi_vid_hdr *vid_hdr;
682
683 if (ubi->ro_mode)
684 return -EROFS;
685
686 err = leb_write_lock(ubi, vol_id, lnum);
687 if (err)
688 return err;
689
690 pnum = vol->eba_tbl[lnum];
691 if (pnum >= 0) {
692 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
693 len, offset, vol_id, lnum, pnum);
694
695 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
696 if (err) {
32608703 697 ubi_warn(ubi, "failed to write data to PEB %d", pnum);
801c135c 698 if (err == -EIO && ubi->bad_allowed)
89b96b69
AB
699 err = recover_peb(ubi, pnum, vol_id, lnum, buf,
700 offset, len);
801c135c
AB
701 if (err)
702 ubi_ro_mode(ubi);
703 }
704 leb_write_unlock(ubi, vol_id, lnum);
705 return err;
706 }
707
708 /*
709 * The logical eraseblock is not mapped. We have to get a free physical
710 * eraseblock and write the volume identifier header there first.
711 */
33818bbb 712 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
801c135c
AB
713 if (!vid_hdr) {
714 leb_write_unlock(ubi, vol_id, lnum);
715 return -ENOMEM;
716 }
717
718 vid_hdr->vol_type = UBI_VID_DYNAMIC;
a7306653 719 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
3261ebd7
CH
720 vid_hdr->vol_id = cpu_to_be32(vol_id);
721 vid_hdr->lnum = cpu_to_be32(lnum);
801c135c 722 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
3261ebd7 723 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
801c135c
AB
724
725retry:
b36a261e 726 pnum = ubi_wl_get_peb(ubi);
801c135c
AB
727 if (pnum < 0) {
728 ubi_free_vid_hdr(ubi, vid_hdr);
729 leb_write_unlock(ubi, vol_id, lnum);
111ab0b2 730 up_read(&ubi->fm_eba_sem);
801c135c
AB
731 return pnum;
732 }
733
734 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
735 len, offset, vol_id, lnum, pnum);
736
737 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
738 if (err) {
32608703 739 ubi_warn(ubi, "failed to write VID header to LEB %d:%d, PEB %d",
801c135c 740 vol_id, lnum, pnum);
111ab0b2 741 up_read(&ubi->fm_eba_sem);
801c135c
AB
742 goto write_error;
743 }
744
393852ec
AB
745 if (len) {
746 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
747 if (err) {
32608703 748 ubi_warn(ubi, "failed to write %d bytes at offset %d of LEB %d:%d, PEB %d",
049333ce 749 len, offset, vol_id, lnum, pnum);
111ab0b2 750 up_read(&ubi->fm_eba_sem);
393852ec
AB
751 goto write_error;
752 }
801c135c
AB
753 }
754
755 vol->eba_tbl[lnum] = pnum;
111ab0b2 756 up_read(&ubi->fm_eba_sem);
801c135c
AB
757
758 leb_write_unlock(ubi, vol_id, lnum);
759 ubi_free_vid_hdr(ubi, vid_hdr);
760 return 0;
761
762write_error:
763 if (err != -EIO || !ubi->bad_allowed) {
764 ubi_ro_mode(ubi);
765 leb_write_unlock(ubi, vol_id, lnum);
766 ubi_free_vid_hdr(ubi, vid_hdr);
767 return err;
768 }
769
770 /*
771 * Fortunately, this is the first write operation to this physical
772 * eraseblock, so just put it and request a new one. We assume that if
773 * this physical eraseblock went bad, the erase code will handle that.
774 */
d36e59e6 775 err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
801c135c
AB
776 if (err || ++tries > UBI_IO_RETRIES) {
777 ubi_ro_mode(ubi);
778 leb_write_unlock(ubi, vol_id, lnum);
779 ubi_free_vid_hdr(ubi, vid_hdr);
780 return err;
781 }
782
a7306653 783 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
32608703 784 ubi_msg(ubi, "try another PEB");
801c135c
AB
785 goto retry;
786}
787
788/**
789 * ubi_eba_write_leb_st - write data to static volume.
790 * @ubi: UBI device description object
89b96b69 791 * @vol: volume description object
801c135c
AB
792 * @lnum: logical eraseblock number
793 * @buf: data to write
794 * @len: how many bytes to write
801c135c
AB
795 * @used_ebs: how many logical eraseblocks will this volume contain
796 *
797 * This function writes data to logical eraseblock @lnum of static volume
89b96b69 798 * @vol. The @used_ebs argument should contain total number of logical
801c135c
AB
799 * eraseblock in this static volume.
800 *
801 * When writing to the last logical eraseblock, the @len argument doesn't have
802 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
803 * to the real data size, although the @buf buffer has to contain the
804 * alignment. In all other cases, @len has to be aligned.
805 *
025dfdaf 806 * It is prohibited to write more than once to logical eraseblocks of static
801c135c
AB
807 * volumes. This function returns zero in case of success and a negative error
808 * code in case of failure.
809 */
89b96b69 810int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
b36a261e 811 int lnum, const void *buf, int len, int used_ebs)
801c135c 812{
89b96b69 813 int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
801c135c
AB
814 struct ubi_vid_hdr *vid_hdr;
815 uint32_t crc;
816
817 if (ubi->ro_mode)
818 return -EROFS;
819
820 if (lnum == used_ebs - 1)
821 /* If this is the last LEB @len may be unaligned */
822 len = ALIGN(data_size, ubi->min_io_size);
823 else
cadb40cc 824 ubi_assert(!(len & (ubi->min_io_size - 1)));
801c135c 825
33818bbb 826 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
801c135c
AB
827 if (!vid_hdr)
828 return -ENOMEM;
829
830 err = leb_write_lock(ubi, vol_id, lnum);
831 if (err) {
832 ubi_free_vid_hdr(ubi, vid_hdr);
833 return err;
834 }
835
a7306653 836 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
3261ebd7
CH
837 vid_hdr->vol_id = cpu_to_be32(vol_id);
838 vid_hdr->lnum = cpu_to_be32(lnum);
801c135c 839 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
3261ebd7 840 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
801c135c
AB
841
842 crc = crc32(UBI_CRC32_INIT, buf, data_size);
843 vid_hdr->vol_type = UBI_VID_STATIC;
3261ebd7
CH
844 vid_hdr->data_size = cpu_to_be32(data_size);
845 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
846 vid_hdr->data_crc = cpu_to_be32(crc);
801c135c
AB
847
848retry:
b36a261e 849 pnum = ubi_wl_get_peb(ubi);
801c135c
AB
850 if (pnum < 0) {
851 ubi_free_vid_hdr(ubi, vid_hdr);
852 leb_write_unlock(ubi, vol_id, lnum);
111ab0b2 853 up_read(&ubi->fm_eba_sem);
801c135c
AB
854 return pnum;
855 }
856
857 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
858 len, vol_id, lnum, pnum, used_ebs);
859
860 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
861 if (err) {
32608703 862 ubi_warn(ubi, "failed to write VID header to LEB %d:%d, PEB %d",
801c135c 863 vol_id, lnum, pnum);
111ab0b2 864 up_read(&ubi->fm_eba_sem);
801c135c
AB
865 goto write_error;
866 }
867
868 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
869 if (err) {
32608703 870 ubi_warn(ubi, "failed to write %d bytes of data to PEB %d",
801c135c 871 len, pnum);
111ab0b2 872 up_read(&ubi->fm_eba_sem);
801c135c
AB
873 goto write_error;
874 }
875
876 ubi_assert(vol->eba_tbl[lnum] < 0);
877 vol->eba_tbl[lnum] = pnum;
111ab0b2 878 up_read(&ubi->fm_eba_sem);
801c135c
AB
879
880 leb_write_unlock(ubi, vol_id, lnum);
881 ubi_free_vid_hdr(ubi, vid_hdr);
882 return 0;
883
884write_error:
885 if (err != -EIO || !ubi->bad_allowed) {
886 /*
887 * This flash device does not admit of bad eraseblocks or
888 * something nasty and unexpected happened. Switch to read-only
889 * mode just in case.
890 */
891 ubi_ro_mode(ubi);
892 leb_write_unlock(ubi, vol_id, lnum);
893 ubi_free_vid_hdr(ubi, vid_hdr);
894 return err;
895 }
896
d36e59e6 897 err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
801c135c
AB
898 if (err || ++tries > UBI_IO_RETRIES) {
899 ubi_ro_mode(ubi);
900 leb_write_unlock(ubi, vol_id, lnum);
901 ubi_free_vid_hdr(ubi, vid_hdr);
902 return err;
903 }
904
a7306653 905 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
32608703 906 ubi_msg(ubi, "try another PEB");
801c135c
AB
907 goto retry;
908}
909
910/*
911 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
912 * @ubi: UBI device description object
c63a491d 913 * @vol: volume description object
801c135c
AB
914 * @lnum: logical eraseblock number
915 * @buf: data to write
916 * @len: how many bytes to write
801c135c
AB
917 *
918 * This function changes the contents of a logical eraseblock atomically. @buf
919 * has to contain new logical eraseblock data, and @len - the length of the
920 * data, which has to be aligned. This function guarantees that in case of an
921 * unclean reboot the old contents is preserved. Returns zero in case of
922 * success and a negative error code in case of failure.
e8823bd6
AB
923 *
924 * UBI reserves one LEB for the "atomic LEB change" operation, so only one
925 * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
801c135c 926 */
89b96b69 927int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
b36a261e 928 int lnum, const void *buf, int len)
801c135c 929{
36a87e44 930 int err, pnum, old_pnum, tries = 0, vol_id = vol->vol_id;
801c135c
AB
931 struct ubi_vid_hdr *vid_hdr;
932 uint32_t crc;
933
934 if (ubi->ro_mode)
935 return -EROFS;
936
60c03153
AB
937 if (len == 0) {
938 /*
939 * Special case when data length is zero. In this case the LEB
940 * has to be unmapped and mapped somewhere else.
941 */
942 err = ubi_eba_unmap_leb(ubi, vol, lnum);
943 if (err)
944 return err;
b36a261e 945 return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0);
60c03153
AB
946 }
947
33818bbb 948 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
801c135c
AB
949 if (!vid_hdr)
950 return -ENOMEM;
951
e8823bd6 952 mutex_lock(&ubi->alc_mutex);
801c135c 953 err = leb_write_lock(ubi, vol_id, lnum);
e8823bd6
AB
954 if (err)
955 goto out_mutex;
801c135c 956
a7306653 957 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
3261ebd7
CH
958 vid_hdr->vol_id = cpu_to_be32(vol_id);
959 vid_hdr->lnum = cpu_to_be32(lnum);
801c135c 960 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
3261ebd7 961 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
801c135c
AB
962
963 crc = crc32(UBI_CRC32_INIT, buf, len);
84a92580 964 vid_hdr->vol_type = UBI_VID_DYNAMIC;
3261ebd7 965 vid_hdr->data_size = cpu_to_be32(len);
801c135c 966 vid_hdr->copy_flag = 1;
3261ebd7 967 vid_hdr->data_crc = cpu_to_be32(crc);
801c135c
AB
968
969retry:
b36a261e 970 pnum = ubi_wl_get_peb(ubi);
801c135c 971 if (pnum < 0) {
e8823bd6 972 err = pnum;
111ab0b2 973 up_read(&ubi->fm_eba_sem);
e8823bd6 974 goto out_leb_unlock;
801c135c
AB
975 }
976
977 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
978 vol_id, lnum, vol->eba_tbl[lnum], pnum);
979
980 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
981 if (err) {
32608703 982 ubi_warn(ubi, "failed to write VID header to LEB %d:%d, PEB %d",
801c135c 983 vol_id, lnum, pnum);
111ab0b2 984 up_read(&ubi->fm_eba_sem);
801c135c
AB
985 goto write_error;
986 }
987
988 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
989 if (err) {
32608703 990 ubi_warn(ubi, "failed to write %d bytes of data to PEB %d",
801c135c 991 len, pnum);
111ab0b2 992 up_read(&ubi->fm_eba_sem);
801c135c
AB
993 goto write_error;
994 }
995
36a87e44 996 old_pnum = vol->eba_tbl[lnum];
801c135c 997 vol->eba_tbl[lnum] = pnum;
111ab0b2 998 up_read(&ubi->fm_eba_sem);
e8823bd6 999
36a87e44
RW
1000 if (old_pnum >= 0) {
1001 err = ubi_wl_put_peb(ubi, vol_id, lnum, old_pnum, 0);
1002 if (err)
1003 goto out_leb_unlock;
1004 }
1005
e8823bd6 1006out_leb_unlock:
801c135c 1007 leb_write_unlock(ubi, vol_id, lnum);
e8823bd6
AB
1008out_mutex:
1009 mutex_unlock(&ubi->alc_mutex);
801c135c 1010 ubi_free_vid_hdr(ubi, vid_hdr);
e8823bd6 1011 return err;
801c135c
AB
1012
1013write_error:
1014 if (err != -EIO || !ubi->bad_allowed) {
1015 /*
1016 * This flash device does not admit of bad eraseblocks or
1017 * something nasty and unexpected happened. Switch to read-only
1018 * mode just in case.
1019 */
1020 ubi_ro_mode(ubi);
e8823bd6 1021 goto out_leb_unlock;
801c135c
AB
1022 }
1023
d36e59e6 1024 err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
801c135c
AB
1025 if (err || ++tries > UBI_IO_RETRIES) {
1026 ubi_ro_mode(ubi);
e8823bd6 1027 goto out_leb_unlock;
801c135c
AB
1028 }
1029
a7306653 1030 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
32608703 1031 ubi_msg(ubi, "try another PEB");
801c135c
AB
1032 goto retry;
1033}
1034
6b5c94c6
AB
1035/**
1036 * is_error_sane - check whether a read error is sane.
1037 * @err: code of the error happened during reading
1038 *
1039 * This is a helper function for 'ubi_eba_copy_leb()' which is called when we
1040 * cannot read data from the target PEB (an error @err happened). If the error
1041 * code is sane, then we treat this error as non-fatal. Otherwise the error is
1042 * fatal and UBI will be switched to R/O mode later.
1043 *
1044 * The idea is that we try not to switch to R/O mode if the read error is
1045 * something which suggests there was a real read problem. E.g., %-EIO. Or a
1046 * memory allocation failed (-%ENOMEM). Otherwise, it is safer to switch to R/O
1047 * mode, simply because we do not know what happened at the MTD level, and we
1048 * cannot handle this. E.g., the underlying driver may have become crazy, and
1049 * it is safer to switch to R/O mode to preserve the data.
1050 *
1051 * And bear in mind, this is about reading from the target PEB, i.e. the PEB
1052 * which we have just written.
1053 */
1054static int is_error_sane(int err)
1055{
786d7831 1056 if (err == -EIO || err == -ENOMEM || err == UBI_IO_BAD_HDR ||
756e1df1 1057 err == UBI_IO_BAD_HDR_EBADMSG || err == -ETIMEDOUT)
6b5c94c6
AB
1058 return 0;
1059 return 1;
1060}
1061
801c135c
AB
1062/**
1063 * ubi_eba_copy_leb - copy logical eraseblock.
1064 * @ubi: UBI device description object
1065 * @from: physical eraseblock number from where to copy
1066 * @to: physical eraseblock number where to copy
1067 * @vid_hdr: VID header of the @from physical eraseblock
1068 *
1069 * This function copies logical eraseblock from physical eraseblock @from to
1070 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
43f9b25a 1071 * function. Returns:
6fa6f5bb 1072 * o %0 in case of success;
cc831464 1073 * o %MOVE_CANCEL_RACE, %MOVE_TARGET_WR_ERR, %MOVE_TARGET_BITFLIPS, etc;
6fa6f5bb 1074 * o a negative error code in case of failure.
801c135c
AB
1075 */
1076int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
1077 struct ubi_vid_hdr *vid_hdr)
1078{
43f9b25a 1079 int err, vol_id, lnum, data_size, aldata_size, idx;
801c135c
AB
1080 struct ubi_volume *vol;
1081 uint32_t crc;
801c135c 1082
3261ebd7
CH
1083 vol_id = be32_to_cpu(vid_hdr->vol_id);
1084 lnum = be32_to_cpu(vid_hdr->lnum);
801c135c 1085
87960c0b 1086 dbg_wl("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
801c135c
AB
1087
1088 if (vid_hdr->vol_type == UBI_VID_STATIC) {
3261ebd7 1089 data_size = be32_to_cpu(vid_hdr->data_size);
801c135c
AB
1090 aldata_size = ALIGN(data_size, ubi->min_io_size);
1091 } else
1092 data_size = aldata_size =
3261ebd7 1093 ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
801c135c 1094
801c135c 1095 idx = vol_id2idx(ubi, vol_id);
43f9b25a 1096 spin_lock(&ubi->volumes_lock);
801c135c 1097 /*
43f9b25a
AB
1098 * Note, we may race with volume deletion, which means that the volume
1099 * this logical eraseblock belongs to might be being deleted. Since the
6fa6f5bb 1100 * volume deletion un-maps all the volume's logical eraseblocks, it will
43f9b25a 1101 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
801c135c 1102 */
801c135c 1103 vol = ubi->volumes[idx];
90bf0265 1104 spin_unlock(&ubi->volumes_lock);
801c135c 1105 if (!vol) {
43f9b25a 1106 /* No need to do further work, cancel */
87960c0b 1107 dbg_wl("volume %d is being removed, cancel", vol_id);
90bf0265 1108 return MOVE_CANCEL_RACE;
801c135c
AB
1109 }
1110
43f9b25a
AB
1111 /*
1112 * We do not want anybody to write to this logical eraseblock while we
1113 * are moving it, so lock it.
1114 *
1115 * Note, we are using non-waiting locking here, because we cannot sleep
1116 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
1117 * unmapping the LEB which is mapped to the PEB we are going to move
1118 * (@from). This task locks the LEB and goes sleep in the
1119 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
1120 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
90bf0265 1121 * LEB is already locked, we just do not move it and return
e801e128
BP
1122 * %MOVE_RETRY. Note, we do not return %MOVE_CANCEL_RACE here because
1123 * we do not know the reasons of the contention - it may be just a
1124 * normal I/O on this LEB, so we want to re-try.
43f9b25a
AB
1125 */
1126 err = leb_write_trylock(ubi, vol_id, lnum);
1127 if (err) {
87960c0b 1128 dbg_wl("contention on LEB %d:%d, cancel", vol_id, lnum);
e801e128 1129 return MOVE_RETRY;
801c135c 1130 }
801c135c 1131
43f9b25a
AB
1132 /*
1133 * The LEB might have been put meanwhile, and the task which put it is
1134 * probably waiting on @ubi->move_mutex. No need to continue the work,
1135 * cancel it.
1136 */
1137 if (vol->eba_tbl[lnum] != from) {
049333ce
AB
1138 dbg_wl("LEB %d:%d is no longer mapped to PEB %d, mapped to PEB %d, cancel",
1139 vol_id, lnum, from, vol->eba_tbl[lnum]);
90bf0265 1140 err = MOVE_CANCEL_RACE;
43f9b25a
AB
1141 goto out_unlock_leb;
1142 }
801c135c 1143
43f9b25a 1144 /*
b77bcb07 1145 * OK, now the LEB is locked and we can safely start moving it. Since
0ca39d74 1146 * this function utilizes the @ubi->peb_buf buffer which is shared
90bf0265 1147 * with some other functions - we lock the buffer by taking the
43f9b25a
AB
1148 * @ubi->buf_mutex.
1149 */
1150 mutex_lock(&ubi->buf_mutex);
87960c0b 1151 dbg_wl("read %d bytes of data", aldata_size);
0ca39d74 1152 err = ubi_io_read_data(ubi, ubi->peb_buf, from, 0, aldata_size);
801c135c 1153 if (err && err != UBI_IO_BITFLIPS) {
32608703 1154 ubi_warn(ubi, "error %d while reading data from PEB %d",
801c135c 1155 err, from);
6b5c94c6 1156 err = MOVE_SOURCE_RD_ERR;
43f9b25a 1157 goto out_unlock_buf;
801c135c
AB
1158 }
1159
1160 /*
fd589a8f 1161 * Now we have got to calculate how much data we have to copy. In
801c135c
AB
1162 * case of a static volume it is fairly easy - the VID header contains
1163 * the data size. In case of a dynamic volume it is more difficult - we
1164 * have to read the contents, cut 0xFF bytes from the end and copy only
1165 * the first part. We must do this to avoid writing 0xFF bytes as it
1166 * may have some side-effects. And not only this. It is important not
1167 * to include those 0xFFs to CRC because later the they may be filled
1168 * by data.
1169 */
1170 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1171 aldata_size = data_size =
0ca39d74 1172 ubi_calc_data_len(ubi, ubi->peb_buf, data_size);
801c135c
AB
1173
1174 cond_resched();
0ca39d74 1175 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size);
801c135c
AB
1176 cond_resched();
1177
1178 /*
90bf0265 1179 * It may turn out to be that the whole @from physical eraseblock
801c135c
AB
1180 * contains only 0xFF bytes. Then we have to only write the VID header
1181 * and do not write any data. This also means we should not set
1182 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1183 */
1184 if (data_size > 0) {
1185 vid_hdr->copy_flag = 1;
3261ebd7
CH
1186 vid_hdr->data_size = cpu_to_be32(data_size);
1187 vid_hdr->data_crc = cpu_to_be32(crc);
801c135c 1188 }
a7306653 1189 vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
801c135c
AB
1190
1191 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
6fa6f5bb
AB
1192 if (err) {
1193 if (err == -EIO)
90bf0265 1194 err = MOVE_TARGET_WR_ERR;
43f9b25a 1195 goto out_unlock_buf;
6fa6f5bb 1196 }
801c135c
AB
1197
1198 cond_resched();
1199
1200 /* Read the VID header back and check if it was written correctly */
1201 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1202 if (err) {
b86a2c56 1203 if (err != UBI_IO_BITFLIPS) {
32608703 1204 ubi_warn(ubi, "error %d while reading VID header back from PEB %d",
049333ce 1205 err, to);
6b5c94c6 1206 if (is_error_sane(err))
b86a2c56
AB
1207 err = MOVE_TARGET_RD_ERR;
1208 } else
cc831464 1209 err = MOVE_TARGET_BITFLIPS;
43f9b25a 1210 goto out_unlock_buf;
801c135c
AB
1211 }
1212
1213 if (data_size > 0) {
0ca39d74 1214 err = ubi_io_write_data(ubi, ubi->peb_buf, to, 0, aldata_size);
6fa6f5bb
AB
1215 if (err) {
1216 if (err == -EIO)
90bf0265 1217 err = MOVE_TARGET_WR_ERR;
43f9b25a 1218 goto out_unlock_buf;
6fa6f5bb 1219 }
801c135c 1220
e88d6e10 1221 cond_resched();
801c135c
AB
1222 }
1223
1224 ubi_assert(vol->eba_tbl[lnum] == from);
111ab0b2 1225 down_read(&ubi->fm_eba_sem);
801c135c 1226 vol->eba_tbl[lnum] = to;
111ab0b2 1227 up_read(&ubi->fm_eba_sem);
801c135c 1228
43f9b25a 1229out_unlock_buf:
e88d6e10 1230 mutex_unlock(&ubi->buf_mutex);
43f9b25a 1231out_unlock_leb:
801c135c 1232 leb_write_unlock(ubi, vol_id, lnum);
801c135c
AB
1233 return err;
1234}
1235
64d4b4c9
AB
1236/**
1237 * print_rsvd_warning - warn about not having enough reserved PEBs.
1238 * @ubi: UBI device description object
1239 *
41e0cd9d 1240 * This is a helper function for 'ubi_eba_init()' which is called when UBI
64d4b4c9
AB
1241 * cannot reserve enough PEBs for bad block handling. This function makes a
1242 * decision whether we have to print a warning or not. The algorithm is as
1243 * follows:
1244 * o if this is a new UBI image, then just print the warning
1245 * o if this is an UBI image which has already been used for some time, print
1246 * a warning only if we can reserve less than 10% of the expected amount of
1247 * the reserved PEB.
1248 *
1249 * The idea is that when UBI is used, PEBs become bad, and the reserved pool
1250 * of PEBs becomes smaller, which is normal and we do not want to scare users
1251 * with a warning every time they attach the MTD device. This was an issue
1252 * reported by real users.
1253 */
1254static void print_rsvd_warning(struct ubi_device *ubi,
a4e6042f 1255 struct ubi_attach_info *ai)
64d4b4c9
AB
1256{
1257 /*
1258 * The 1 << 18 (256KiB) number is picked randomly, just a reasonably
1259 * large number to distinguish between newly flashed and used images.
1260 */
a4e6042f 1261 if (ai->max_sqnum > (1 << 18)) {
64d4b4c9
AB
1262 int min = ubi->beb_rsvd_level / 10;
1263
1264 if (!min)
1265 min = 1;
1266 if (ubi->beb_rsvd_pebs > min)
1267 return;
1268 }
1269
32608703 1270 ubi_warn(ubi, "cannot reserve enough PEBs for bad PEB handling, reserved %d, need %d",
049333ce 1271 ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
5fc01ab6 1272 if (ubi->corr_peb_count)
32608703 1273 ubi_warn(ubi, "%d PEBs are corrupted and not used",
049333ce 1274 ubi->corr_peb_count);
64d4b4c9
AB
1275}
1276
00abf304
RW
1277/**
1278 * self_check_eba - run a self check on the EBA table constructed by fastmap.
1279 * @ubi: UBI device description object
1280 * @ai_fastmap: UBI attach info object created by fastmap
1281 * @ai_scan: UBI attach info object created by scanning
1282 *
1283 * Returns < 0 in case of an internal error, 0 otherwise.
1284 * If a bad EBA table entry was found it will be printed out and
1285 * ubi_assert() triggers.
1286 */
1287int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
1288 struct ubi_attach_info *ai_scan)
1289{
1290 int i, j, num_volumes, ret = 0;
1291 int **scan_eba, **fm_eba;
1292 struct ubi_ainf_volume *av;
1293 struct ubi_volume *vol;
1294 struct ubi_ainf_peb *aeb;
1295 struct rb_node *rb;
1296
1297 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1298
1299 scan_eba = kmalloc(sizeof(*scan_eba) * num_volumes, GFP_KERNEL);
1300 if (!scan_eba)
1301 return -ENOMEM;
1302
1303 fm_eba = kmalloc(sizeof(*fm_eba) * num_volumes, GFP_KERNEL);
1304 if (!fm_eba) {
1305 kfree(scan_eba);
1306 return -ENOMEM;
1307 }
1308
1309 for (i = 0; i < num_volumes; i++) {
1310 vol = ubi->volumes[i];
1311 if (!vol)
1312 continue;
1313
1314 scan_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**scan_eba),
1315 GFP_KERNEL);
1316 if (!scan_eba[i]) {
1317 ret = -ENOMEM;
1318 goto out_free;
1319 }
1320
1321 fm_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**fm_eba),
1322 GFP_KERNEL);
1323 if (!fm_eba[i]) {
1324 ret = -ENOMEM;
1325 goto out_free;
1326 }
1327
1328 for (j = 0; j < vol->reserved_pebs; j++)
1329 scan_eba[i][j] = fm_eba[i][j] = UBI_LEB_UNMAPPED;
1330
1331 av = ubi_find_av(ai_scan, idx2vol_id(ubi, i));
1332 if (!av)
1333 continue;
1334
1335 ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
1336 scan_eba[i][aeb->lnum] = aeb->pnum;
1337
1338 av = ubi_find_av(ai_fastmap, idx2vol_id(ubi, i));
1339 if (!av)
1340 continue;
1341
1342 ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
1343 fm_eba[i][aeb->lnum] = aeb->pnum;
1344
1345 for (j = 0; j < vol->reserved_pebs; j++) {
1346 if (scan_eba[i][j] != fm_eba[i][j]) {
1347 if (scan_eba[i][j] == UBI_LEB_UNMAPPED ||
1348 fm_eba[i][j] == UBI_LEB_UNMAPPED)
1349 continue;
1350
32608703 1351 ubi_err(ubi, "LEB:%i:%i is PEB:%i instead of %i!",
5347417e 1352 vol->vol_id, j, fm_eba[i][j],
00abf304
RW
1353 scan_eba[i][j]);
1354 ubi_assert(0);
1355 }
1356 }
1357 }
1358
1359out_free:
1360 for (i = 0; i < num_volumes; i++) {
1361 if (!ubi->volumes[i])
1362 continue;
1363
1364 kfree(scan_eba[i]);
1365 kfree(fm_eba[i]);
1366 }
1367
1368 kfree(scan_eba);
1369 kfree(fm_eba);
1370 return ret;
1371}
1372
801c135c 1373/**
41e0cd9d 1374 * ubi_eba_init - initialize the EBA sub-system using attaching information.
801c135c 1375 * @ubi: UBI device description object
a4e6042f 1376 * @ai: attaching information
801c135c
AB
1377 *
1378 * This function returns zero in case of success and a negative error code in
1379 * case of failure.
1380 */
41e0cd9d 1381int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
801c135c
AB
1382{
1383 int i, j, err, num_volumes;
517af48c 1384 struct ubi_ainf_volume *av;
801c135c 1385 struct ubi_volume *vol;
2c5ec5ce 1386 struct ubi_ainf_peb *aeb;
801c135c
AB
1387 struct rb_node *rb;
1388
85c6e6e2 1389 dbg_eba("initialize EBA sub-system");
801c135c
AB
1390
1391 spin_lock_init(&ubi->ltree_lock);
e8823bd6 1392 mutex_init(&ubi->alc_mutex);
801c135c
AB
1393 ubi->ltree = RB_ROOT;
1394
a4e6042f 1395 ubi->global_sqnum = ai->max_sqnum + 1;
801c135c
AB
1396 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1397
1398 for (i = 0; i < num_volumes; i++) {
1399 vol = ubi->volumes[i];
1400 if (!vol)
1401 continue;
1402
1403 cond_resched();
1404
1405 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1406 GFP_KERNEL);
1407 if (!vol->eba_tbl) {
1408 err = -ENOMEM;
1409 goto out_free;
1410 }
1411
1412 for (j = 0; j < vol->reserved_pebs; j++)
1413 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1414
dcd85fdd 1415 av = ubi_find_av(ai, idx2vol_id(ubi, i));
517af48c 1416 if (!av)
801c135c
AB
1417 continue;
1418
517af48c 1419 ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
2c5ec5ce 1420 if (aeb->lnum >= vol->reserved_pebs)
801c135c
AB
1421 /*
1422 * This may happen in case of an unclean reboot
1423 * during re-size.
1424 */
0bae2887 1425 ubi_move_aeb_to_list(av, aeb, &ai->erase);
d74adbdb
BN
1426 else
1427 vol->eba_tbl[aeb->lnum] = aeb->pnum;
801c135c
AB
1428 }
1429 }
1430
94780d4d 1431 if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
32608703 1432 ubi_err(ubi, "no enough physical eraseblocks (%d, need %d)",
94780d4d 1433 ubi->avail_pebs, EBA_RESERVED_PEBS);
5fc01ab6 1434 if (ubi->corr_peb_count)
32608703 1435 ubi_err(ubi, "%d PEBs are corrupted and not used",
5fc01ab6 1436 ubi->corr_peb_count);
94780d4d
AB
1437 err = -ENOSPC;
1438 goto out_free;
1439 }
1440 ubi->avail_pebs -= EBA_RESERVED_PEBS;
1441 ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1442
801c135c
AB
1443 if (ubi->bad_allowed) {
1444 ubi_calculate_reserved(ubi);
1445
1446 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1447 /* No enough free physical eraseblocks */
1448 ubi->beb_rsvd_pebs = ubi->avail_pebs;
a4e6042f 1449 print_rsvd_warning(ubi, ai);
801c135c
AB
1450 } else
1451 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1452
1453 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1454 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1455 }
1456
85c6e6e2 1457 dbg_eba("EBA sub-system is initialized");
801c135c
AB
1458 return 0;
1459
1460out_free:
1461 for (i = 0; i < num_volumes; i++) {
1462 if (!ubi->volumes[i])
1463 continue;
1464 kfree(ubi->volumes[i]->eba_tbl);
7194e6f9 1465 ubi->volumes[i]->eba_tbl = NULL;
801c135c 1466 }
801c135c
AB
1467 return err;
1468}
This page took 0.63449 seconds and 5 git commands to generate.