blob: 4be671815014ee3b054f2ba8f704781dfbc1f032 [file] [log] [blame]
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001/*
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/*
Artem Bityutskiy85c6e6e2008-07-16 10:25:56 +030022 * The UBI Eraseblock Association (EBA) sub-system.
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040023 *
Artem Bityutskiy85c6e6e2008-07-16 10:25:56 +030024 * This sub-system is responsible for I/O to/from logical eraseblock.
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040025 *
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 *
Artem Bityutskiy85c6e6e2008-07-16 10:25:56 +030030 * 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.
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040036 *
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
Artem Bityutskiye8823bd2007-09-13 14:28:14 +030049/* Number of physical eraseblocks reserved for atomic LEB change operation */
50#define EBA_RESERVED_PEBS 1
51
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040052/**
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040053 * 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 */
60static unsigned long long next_sqnum(struct ubi_device *ubi)
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{
Artem Bityutskiy91f2d532008-01-24 11:23:23 +020081 if (vol_id == UBI_LAYOUT_VOLUME_ID)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040082 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 *
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +020092 * This function returns a pointer to the corresponding &struct ubi_ltree_entry
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040093 * object if the logical eraseblock is locked and %NULL if it is not.
94 * @ubi->ltree_lock has to be locked.
95 */
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +020096static struct ubi_ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
97 int lnum)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +040098{
99 struct rb_node *p;
100
101 p = ubi->ltree.rb_node;
102 while (p) {
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200103 struct ubi_ltree_entry *le;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400104
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200105 le = rb_entry(p, struct ubi_ltree_entry, rb);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400106
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 */
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200135static struct ubi_ltree_entry *ltree_add_entry(struct ubi_device *ubi,
136 int vol_id, int lnum)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400137{
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200138 struct ubi_ltree_entry *le, *le1, *le_free;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400139
Artem Bityutskiyb9a06622008-01-16 12:11:54 +0200140 le = kmalloc(sizeof(struct ubi_ltree_entry), GFP_NOFS);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400141 if (!le)
142 return ERR_PTR(-ENOMEM);
143
Artem Bityutskiyb9a06622008-01-16 12:11:54 +0200144 le->users = 0;
145 init_rwsem(&le->mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400146 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;
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200171 le1 = rb_entry(parent, struct ubi_ltree_entry, rb);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400172
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
Artem Bityutskiy9c9ec142008-07-18 13:19:52 +0300192 kfree(le_free);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400193 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{
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200207 struct ubi_ltree_entry *le;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400208
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{
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200224 struct ubi_ltree_entry *le;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400225
226 spin_lock(&ubi->ltree_lock);
227 le = ltree_lookup(ubi, vol_id, lnum);
228 le->users -= 1;
229 ubi_assert(le->users >= 0);
Artem Bityutskiy23add742008-06-16 13:35:23 +0300230 up_read(&le->mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400231 if (le->users == 0) {
232 rb_erase(&le->rb, &ubi->ltree);
Artem Bityutskiy23add742008-06-16 13:35:23 +0300233 kfree(le);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400234 }
235 spin_unlock(&ubi->ltree_lock);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400236}
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{
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200249 struct ubi_ltree_entry *le;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400250
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
258/**
Artem Bityutskiy43f9b252007-12-18 15:06:55 +0200259 * 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{
Artem Bityutskiy43f9b252007-12-18 15:06:55 +0200271 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);
Artem Bityutskiyb9a06622008-01-16 12:11:54 +0200285 kfree(le);
Artem Bityutskiy23add742008-06-16 13:35:23 +0300286 }
287 spin_unlock(&ubi->ltree_lock);
Artem Bityutskiy43f9b252007-12-18 15:06:55 +0200288
289 return 1;
290}
291
292/**
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400293 * 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{
Artem Bityutskiy3a8d4642007-12-16 12:32:51 +0200300 struct ubi_ltree_entry *le;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400301
302 spin_lock(&ubi->ltree_lock);
303 le = ltree_lookup(ubi, vol_id, lnum);
304 le->users -= 1;
305 ubi_assert(le->users >= 0);
Artem Bityutskiy23add742008-06-16 13:35:23 +0300306 up_write(&le->mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400307 if (le->users == 0) {
308 rb_erase(&le->rb, &ubi->ltree);
Artem Bityutskiyb9a06622008-01-16 12:11:54 +0200309 kfree(le);
Artem Bityutskiy23add742008-06-16 13:35:23 +0300310 }
311 spin_unlock(&ubi->ltree_lock);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400312}
313
314/**
315 * ubi_eba_unmap_leb - un-map logical eraseblock.
316 * @ubi: UBI device description object
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200317 * @vol: volume description object
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400318 * @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 */
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200324int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
325 int lnum)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400326{
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200327 int err, pnum, vol_id = vol->vol_id;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400328
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
343 vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
344 err = ubi_wl_put_peb(ubi, pnum, 0);
345
346out_unlock:
347 leb_write_unlock(ubi, vol_id, lnum);
348 return err;
349}
350
351/**
352 * ubi_eba_read_leb - read data.
353 * @ubi: UBI device description object
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200354 * @vol: volume description object
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400355 * @lnum: logical eraseblock number
356 * @buf: buffer to store the read data
357 * @offset: offset from where to read
358 * @len: how many bytes to read
359 * @check: data CRC check flag
360 *
361 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
362 * bytes. The @check flag only makes sense for static volumes and forces
363 * eraseblock data CRC checking.
364 *
365 * In case of success this function returns zero. In case of a static volume,
366 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
367 * returned for any volume type if an ECC error was detected by the MTD device
368 * driver. Other negative error cored may be returned in case of other errors.
369 */
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200370int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
371 void *buf, int offset, int len, int check)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400372{
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200373 int err, pnum, scrub = 0, vol_id = vol->vol_id;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400374 struct ubi_vid_hdr *vid_hdr;
Jeff Garzika6343af2007-07-17 05:39:58 -0400375 uint32_t uninitialized_var(crc);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400376
377 err = leb_read_lock(ubi, vol_id, lnum);
378 if (err)
379 return err;
380
381 pnum = vol->eba_tbl[lnum];
382 if (pnum < 0) {
383 /*
384 * The logical eraseblock is not mapped, fill the whole buffer
385 * with 0xFF bytes. The exception is static volumes for which
386 * it is an error to read unmapped logical eraseblocks.
387 */
388 dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
389 len, offset, vol_id, lnum);
390 leb_read_unlock(ubi, vol_id, lnum);
391 ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
392 memset(buf, 0xFF, len);
393 return 0;
394 }
395
396 dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
397 len, offset, vol_id, lnum, pnum);
398
399 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
400 check = 0;
401
402retry:
403 if (check) {
Artem Bityutskiy33818bb2007-08-28 21:29:32 +0300404 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400405 if (!vid_hdr) {
406 err = -ENOMEM;
407 goto out_unlock;
408 }
409
410 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
411 if (err && err != UBI_IO_BITFLIPS) {
412 if (err > 0) {
413 /*
414 * The header is either absent or corrupted.
415 * The former case means there is a bug -
416 * switch to read-only mode just in case.
417 * The latter case means a real corruption - we
418 * may try to recover data. FIXME: but this is
419 * not implemented.
420 */
Artem Bityutskiy756e1df2010-09-03 01:30:16 +0300421 if (err == UBI_IO_BAD_HDR_EBADMSG ||
Artem Bityutskiyeb895802010-05-03 09:04:39 +0300422 err == UBI_IO_BAD_HDR) {
Artem Bityutskiyb86a2c52009-05-24 14:13:34 +0300423 ubi_warn("corrupted VID header at PEB "
424 "%d, LEB %d:%d", pnum, vol_id,
425 lnum);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400426 err = -EBADMSG;
427 } else
428 ubi_ro_mode(ubi);
429 }
430 goto out_free;
431 } else if (err == UBI_IO_BITFLIPS)
432 scrub = 1;
433
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300434 ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs));
435 ubi_assert(len == be32_to_cpu(vid_hdr->data_size));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400436
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300437 crc = be32_to_cpu(vid_hdr->data_crc);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400438 ubi_free_vid_hdr(ubi, vid_hdr);
439 }
440
441 err = ubi_io_read_data(ubi, buf, pnum, offset, len);
442 if (err) {
443 if (err == UBI_IO_BITFLIPS) {
444 scrub = 1;
445 err = 0;
446 } else if (err == -EBADMSG) {
447 if (vol->vol_type == UBI_DYNAMIC_VOLUME)
448 goto out_unlock;
449 scrub = 1;
450 if (!check) {
451 ubi_msg("force data checking");
452 check = 1;
453 goto retry;
454 }
455 } else
456 goto out_unlock;
457 }
458
459 if (check) {
Jeff Garzik2ab934b2007-07-17 01:49:56 -0400460 uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400461 if (crc1 != crc) {
462 ubi_warn("CRC error: calculated %#08x, must be %#08x",
463 crc1, crc);
464 err = -EBADMSG;
465 goto out_unlock;
466 }
467 }
468
469 if (scrub)
470 err = ubi_wl_scrub_peb(ubi, pnum);
471
472 leb_read_unlock(ubi, vol_id, lnum);
473 return err;
474
475out_free:
476 ubi_free_vid_hdr(ubi, vid_hdr);
477out_unlock:
478 leb_read_unlock(ubi, vol_id, lnum);
479 return err;
480}
481
482/**
483 * recover_peb - recover from write failure.
484 * @ubi: UBI device description object
485 * @pnum: the physical eraseblock to recover
486 * @vol_id: volume ID
487 * @lnum: logical eraseblock number
488 * @buf: data which was not written because of the write failure
489 * @offset: offset of the failed write
490 * @len: how many bytes should have been written
491 *
492 * This function is called in case of a write failure and moves all good data
493 * from the potentially bad physical eraseblock to a good physical eraseblock.
494 * This function also writes the data which was not written due to the failure.
495 * Returns new physical eraseblock number in case of success, and a negative
496 * error code in case of failure.
497 */
498static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
499 const void *buf, int offset, int len)
500{
501 int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0;
502 struct ubi_volume *vol = ubi->volumes[idx];
503 struct ubi_vid_hdr *vid_hdr;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400504
Artem Bityutskiy33818bb2007-08-28 21:29:32 +0300505 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
Artem Bityutskiy9c9ec142008-07-18 13:19:52 +0300506 if (!vid_hdr)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400507 return -ENOMEM;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400508
509retry:
510 new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
511 if (new_pnum < 0) {
512 ubi_free_vid_hdr(ubi, vid_hdr);
513 return new_pnum;
514 }
515
516 ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum);
517
518 err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
519 if (err && err != UBI_IO_BITFLIPS) {
520 if (err > 0)
521 err = -EIO;
522 goto out_put;
523 }
524
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300525 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400526 err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
527 if (err)
528 goto write_error;
529
530 data_size = offset + len;
Artem Bityutskiy4df581f2008-12-04 20:52:44 +0200531 mutex_lock(&ubi->buf_mutex);
Artem Bityutskiye88d6e102007-08-29 14:51:52 +0300532 memset(ubi->peb_buf1 + offset, 0xFF, len);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400533
534 /* Read everything before the area where the write failure happened */
535 if (offset > 0) {
Artem Bityutskiye88d6e102007-08-29 14:51:52 +0300536 err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset);
537 if (err && err != UBI_IO_BITFLIPS)
Artem Bityutskiy4df581f2008-12-04 20:52:44 +0200538 goto out_unlock;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400539 }
540
Artem Bityutskiye88d6e102007-08-29 14:51:52 +0300541 memcpy(ubi->peb_buf1 + offset, buf, len);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400542
Artem Bityutskiye88d6e102007-08-29 14:51:52 +0300543 err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size);
Artem Bityutskiy4df581f2008-12-04 20:52:44 +0200544 if (err) {
545 mutex_unlock(&ubi->buf_mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400546 goto write_error;
Artem Bityutskiy4df581f2008-12-04 20:52:44 +0200547 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400548
Artem Bityutskiye88d6e102007-08-29 14:51:52 +0300549 mutex_unlock(&ubi->buf_mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400550 ubi_free_vid_hdr(ubi, vid_hdr);
551
552 vol->eba_tbl[lnum] = new_pnum;
553 ubi_wl_put_peb(ubi, pnum, 1);
554
555 ubi_msg("data was successfully recovered");
556 return 0;
557
Artem Bityutskiy4df581f2008-12-04 20:52:44 +0200558out_unlock:
Artem Bityutskiye88d6e102007-08-29 14:51:52 +0300559 mutex_unlock(&ubi->buf_mutex);
Artem Bityutskiy4df581f2008-12-04 20:52:44 +0200560out_put:
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400561 ubi_wl_put_peb(ubi, new_pnum, 1);
562 ubi_free_vid_hdr(ubi, vid_hdr);
563 return err;
564
565write_error:
566 /*
567 * Bad luck? This physical eraseblock is bad too? Crud. Let's try to
568 * get another one.
569 */
570 ubi_warn("failed to write to PEB %d", new_pnum);
571 ubi_wl_put_peb(ubi, new_pnum, 1);
572 if (++tries > UBI_IO_RETRIES) {
573 ubi_free_vid_hdr(ubi, vid_hdr);
574 return err;
575 }
576 ubi_msg("try again");
577 goto retry;
578}
579
580/**
581 * ubi_eba_write_leb - write data to dynamic volume.
582 * @ubi: UBI device description object
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200583 * @vol: volume description object
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400584 * @lnum: logical eraseblock number
585 * @buf: the data to write
586 * @offset: offset within the logical eraseblock where to write
587 * @len: how many bytes to write
588 * @dtype: data type
589 *
590 * This function writes data to logical eraseblock @lnum of a dynamic volume
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200591 * @vol. Returns zero in case of success and a negative error code in case
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400592 * of failure. In case of error, it is possible that something was still
593 * written to the flash media, but may be some garbage.
594 */
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200595int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400596 const void *buf, int offset, int len, int dtype)
597{
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200598 int err, pnum, tries = 0, vol_id = vol->vol_id;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400599 struct ubi_vid_hdr *vid_hdr;
600
601 if (ubi->ro_mode)
602 return -EROFS;
603
604 err = leb_write_lock(ubi, vol_id, lnum);
605 if (err)
606 return err;
607
608 pnum = vol->eba_tbl[lnum];
609 if (pnum >= 0) {
610 dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d",
611 len, offset, vol_id, lnum, pnum);
612
613 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
614 if (err) {
615 ubi_warn("failed to write data to PEB %d", pnum);
616 if (err == -EIO && ubi->bad_allowed)
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200617 err = recover_peb(ubi, pnum, vol_id, lnum, buf,
618 offset, len);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400619 if (err)
620 ubi_ro_mode(ubi);
621 }
622 leb_write_unlock(ubi, vol_id, lnum);
623 return err;
624 }
625
626 /*
627 * The logical eraseblock is not mapped. We have to get a free physical
628 * eraseblock and write the volume identifier header there first.
629 */
Artem Bityutskiy33818bb2007-08-28 21:29:32 +0300630 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400631 if (!vid_hdr) {
632 leb_write_unlock(ubi, vol_id, lnum);
633 return -ENOMEM;
634 }
635
636 vid_hdr->vol_type = UBI_VID_DYNAMIC;
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300637 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
638 vid_hdr->vol_id = cpu_to_be32(vol_id);
639 vid_hdr->lnum = cpu_to_be32(lnum);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400640 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300641 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400642
643retry:
644 pnum = ubi_wl_get_peb(ubi, dtype);
645 if (pnum < 0) {
646 ubi_free_vid_hdr(ubi, vid_hdr);
647 leb_write_unlock(ubi, vol_id, lnum);
648 return pnum;
649 }
650
651 dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d",
652 len, offset, vol_id, lnum, pnum);
653
654 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
655 if (err) {
656 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
657 vol_id, lnum, pnum);
658 goto write_error;
659 }
660
Artem Bityutskiy393852e2007-12-06 18:47:30 +0200661 if (len) {
662 err = ubi_io_write_data(ubi, buf, pnum, offset, len);
663 if (err) {
664 ubi_warn("failed to write %d bytes at offset %d of "
665 "LEB %d:%d, PEB %d", len, offset, vol_id,
666 lnum, pnum);
667 goto write_error;
668 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400669 }
670
671 vol->eba_tbl[lnum] = pnum;
672
673 leb_write_unlock(ubi, vol_id, lnum);
674 ubi_free_vid_hdr(ubi, vid_hdr);
675 return 0;
676
677write_error:
678 if (err != -EIO || !ubi->bad_allowed) {
679 ubi_ro_mode(ubi);
680 leb_write_unlock(ubi, vol_id, lnum);
681 ubi_free_vid_hdr(ubi, vid_hdr);
682 return err;
683 }
684
685 /*
686 * Fortunately, this is the first write operation to this physical
687 * eraseblock, so just put it and request a new one. We assume that if
688 * this physical eraseblock went bad, the erase code will handle that.
689 */
690 err = ubi_wl_put_peb(ubi, pnum, 1);
691 if (err || ++tries > UBI_IO_RETRIES) {
692 ubi_ro_mode(ubi);
693 leb_write_unlock(ubi, vol_id, lnum);
694 ubi_free_vid_hdr(ubi, vid_hdr);
695 return err;
696 }
697
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300698 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400699 ubi_msg("try another PEB");
700 goto retry;
701}
702
703/**
704 * ubi_eba_write_leb_st - write data to static volume.
705 * @ubi: UBI device description object
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200706 * @vol: volume description object
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400707 * @lnum: logical eraseblock number
708 * @buf: data to write
709 * @len: how many bytes to write
710 * @dtype: data type
711 * @used_ebs: how many logical eraseblocks will this volume contain
712 *
713 * This function writes data to logical eraseblock @lnum of static volume
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200714 * @vol. The @used_ebs argument should contain total number of logical
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400715 * eraseblock in this static volume.
716 *
717 * When writing to the last logical eraseblock, the @len argument doesn't have
718 * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent
719 * to the real data size, although the @buf buffer has to contain the
720 * alignment. In all other cases, @len has to be aligned.
721 *
Frederik Schwarzer025dfda2008-10-16 19:02:37 +0200722 * It is prohibited to write more than once to logical eraseblocks of static
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400723 * volumes. This function returns zero in case of success and a negative error
724 * code in case of failure.
725 */
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200726int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
727 int lnum, const void *buf, int len, int dtype,
728 int used_ebs)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400729{
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200730 int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400731 struct ubi_vid_hdr *vid_hdr;
732 uint32_t crc;
733
734 if (ubi->ro_mode)
735 return -EROFS;
736
737 if (lnum == used_ebs - 1)
738 /* If this is the last LEB @len may be unaligned */
739 len = ALIGN(data_size, ubi->min_io_size);
740 else
Kyungmin Parkcadb40c2008-05-22 10:32:18 +0900741 ubi_assert(!(len & (ubi->min_io_size - 1)));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400742
Artem Bityutskiy33818bb2007-08-28 21:29:32 +0300743 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400744 if (!vid_hdr)
745 return -ENOMEM;
746
747 err = leb_write_lock(ubi, vol_id, lnum);
748 if (err) {
749 ubi_free_vid_hdr(ubi, vid_hdr);
750 return err;
751 }
752
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300753 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
754 vid_hdr->vol_id = cpu_to_be32(vol_id);
755 vid_hdr->lnum = cpu_to_be32(lnum);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400756 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300757 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400758
759 crc = crc32(UBI_CRC32_INIT, buf, data_size);
760 vid_hdr->vol_type = UBI_VID_STATIC;
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300761 vid_hdr->data_size = cpu_to_be32(data_size);
762 vid_hdr->used_ebs = cpu_to_be32(used_ebs);
763 vid_hdr->data_crc = cpu_to_be32(crc);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400764
765retry:
766 pnum = ubi_wl_get_peb(ubi, dtype);
767 if (pnum < 0) {
768 ubi_free_vid_hdr(ubi, vid_hdr);
769 leb_write_unlock(ubi, vol_id, lnum);
770 return pnum;
771 }
772
773 dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d",
774 len, vol_id, lnum, pnum, used_ebs);
775
776 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
777 if (err) {
778 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
779 vol_id, lnum, pnum);
780 goto write_error;
781 }
782
783 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
784 if (err) {
785 ubi_warn("failed to write %d bytes of data to PEB %d",
786 len, pnum);
787 goto write_error;
788 }
789
790 ubi_assert(vol->eba_tbl[lnum] < 0);
791 vol->eba_tbl[lnum] = pnum;
792
793 leb_write_unlock(ubi, vol_id, lnum);
794 ubi_free_vid_hdr(ubi, vid_hdr);
795 return 0;
796
797write_error:
798 if (err != -EIO || !ubi->bad_allowed) {
799 /*
800 * This flash device does not admit of bad eraseblocks or
801 * something nasty and unexpected happened. Switch to read-only
802 * mode just in case.
803 */
804 ubi_ro_mode(ubi);
805 leb_write_unlock(ubi, vol_id, lnum);
806 ubi_free_vid_hdr(ubi, vid_hdr);
807 return err;
808 }
809
810 err = ubi_wl_put_peb(ubi, pnum, 1);
811 if (err || ++tries > UBI_IO_RETRIES) {
812 ubi_ro_mode(ubi);
813 leb_write_unlock(ubi, vol_id, lnum);
814 ubi_free_vid_hdr(ubi, vid_hdr);
815 return err;
816 }
817
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300818 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400819 ubi_msg("try another PEB");
820 goto retry;
821}
822
823/*
824 * ubi_eba_atomic_leb_change - change logical eraseblock atomically.
825 * @ubi: UBI device description object
Artem Bityutskiyc63a4912007-12-17 13:21:07 +0200826 * @vol: volume description object
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400827 * @lnum: logical eraseblock number
828 * @buf: data to write
829 * @len: how many bytes to write
830 * @dtype: data type
831 *
832 * This function changes the contents of a logical eraseblock atomically. @buf
833 * has to contain new logical eraseblock data, and @len - the length of the
834 * data, which has to be aligned. This function guarantees that in case of an
835 * unclean reboot the old contents is preserved. Returns zero in case of
836 * success and a negative error code in case of failure.
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300837 *
838 * UBI reserves one LEB for the "atomic LEB change" operation, so only one
839 * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400840 */
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200841int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
842 int lnum, const void *buf, int len, int dtype)
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400843{
Artem Bityutskiy89b96b62007-12-16 20:00:38 +0200844 int err, pnum, tries = 0, vol_id = vol->vol_id;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400845 struct ubi_vid_hdr *vid_hdr;
846 uint32_t crc;
847
848 if (ubi->ro_mode)
849 return -EROFS;
850
Artem Bityutskiy60c03152008-01-24 17:56:14 +0200851 if (len == 0) {
852 /*
853 * Special case when data length is zero. In this case the LEB
854 * has to be unmapped and mapped somewhere else.
855 */
856 err = ubi_eba_unmap_leb(ubi, vol, lnum);
857 if (err)
858 return err;
859 return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0, dtype);
860 }
861
Artem Bityutskiy33818bb2007-08-28 21:29:32 +0300862 vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400863 if (!vid_hdr)
864 return -ENOMEM;
865
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300866 mutex_lock(&ubi->alc_mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400867 err = leb_write_lock(ubi, vol_id, lnum);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300868 if (err)
869 goto out_mutex;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400870
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300871 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
872 vid_hdr->vol_id = cpu_to_be32(vol_id);
873 vid_hdr->lnum = cpu_to_be32(lnum);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400874 vid_hdr->compat = ubi_get_compat(ubi, vol_id);
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300875 vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400876
877 crc = crc32(UBI_CRC32_INIT, buf, len);
Artem Bityutskiy84a92582007-07-04 16:16:51 +0300878 vid_hdr->vol_type = UBI_VID_DYNAMIC;
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300879 vid_hdr->data_size = cpu_to_be32(len);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400880 vid_hdr->copy_flag = 1;
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300881 vid_hdr->data_crc = cpu_to_be32(crc);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400882
883retry:
884 pnum = ubi_wl_get_peb(ubi, dtype);
885 if (pnum < 0) {
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300886 err = pnum;
887 goto out_leb_unlock;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400888 }
889
890 dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d",
891 vol_id, lnum, vol->eba_tbl[lnum], pnum);
892
893 err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr);
894 if (err) {
895 ubi_warn("failed to write VID header to LEB %d:%d, PEB %d",
896 vol_id, lnum, pnum);
897 goto write_error;
898 }
899
900 err = ubi_io_write_data(ubi, buf, pnum, 0, len);
901 if (err) {
902 ubi_warn("failed to write %d bytes of data to PEB %d",
903 len, pnum);
904 goto write_error;
905 }
906
Artem Bityutskiya443db42007-05-21 20:26:05 +0300907 if (vol->eba_tbl[lnum] >= 0) {
Artem Bityutskiy4d88de42008-07-18 12:42:14 +0300908 err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 0);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300909 if (err)
910 goto out_leb_unlock;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400911 }
912
913 vol->eba_tbl[lnum] = pnum;
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300914
915out_leb_unlock:
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400916 leb_write_unlock(ubi, vol_id, lnum);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300917out_mutex:
918 mutex_unlock(&ubi->alc_mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400919 ubi_free_vid_hdr(ubi, vid_hdr);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300920 return err;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400921
922write_error:
923 if (err != -EIO || !ubi->bad_allowed) {
924 /*
925 * This flash device does not admit of bad eraseblocks or
926 * something nasty and unexpected happened. Switch to read-only
927 * mode just in case.
928 */
929 ubi_ro_mode(ubi);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300930 goto out_leb_unlock;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400931 }
932
933 err = ubi_wl_put_peb(ubi, pnum, 1);
934 if (err || ++tries > UBI_IO_RETRIES) {
935 ubi_ro_mode(ubi);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +0300936 goto out_leb_unlock;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400937 }
938
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300939 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400940 ubi_msg("try another PEB");
941 goto retry;
942}
943
944/**
Artem Bityutskiy6b5c94c2009-06-10 11:32:05 +0300945 * is_error_sane - check whether a read error is sane.
946 * @err: code of the error happened during reading
947 *
948 * This is a helper function for 'ubi_eba_copy_leb()' which is called when we
949 * cannot read data from the target PEB (an error @err happened). If the error
950 * code is sane, then we treat this error as non-fatal. Otherwise the error is
951 * fatal and UBI will be switched to R/O mode later.
952 *
953 * The idea is that we try not to switch to R/O mode if the read error is
954 * something which suggests there was a real read problem. E.g., %-EIO. Or a
955 * memory allocation failed (-%ENOMEM). Otherwise, it is safer to switch to R/O
956 * mode, simply because we do not know what happened at the MTD level, and we
957 * cannot handle this. E.g., the underlying driver may have become crazy, and
958 * it is safer to switch to R/O mode to preserve the data.
959 *
960 * And bear in mind, this is about reading from the target PEB, i.e. the PEB
961 * which we have just written.
962 */
963static int is_error_sane(int err)
964{
Artem Bityutskiy786d7832010-04-30 16:50:22 +0300965 if (err == -EIO || err == -ENOMEM || err == UBI_IO_BAD_HDR ||
Artem Bityutskiy756e1df2010-09-03 01:30:16 +0300966 err == UBI_IO_BAD_HDR_EBADMSG || err == -ETIMEDOUT)
Artem Bityutskiy6b5c94c2009-06-10 11:32:05 +0300967 return 0;
968 return 1;
969}
970
971/**
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400972 * ubi_eba_copy_leb - copy logical eraseblock.
973 * @ubi: UBI device description object
974 * @from: physical eraseblock number from where to copy
975 * @to: physical eraseblock number where to copy
976 * @vid_hdr: VID header of the @from physical eraseblock
977 *
978 * This function copies logical eraseblock from physical eraseblock @from to
979 * physical eraseblock @to. The @vid_hdr buffer may be changed by this
Artem Bityutskiy43f9b252007-12-18 15:06:55 +0200980 * function. Returns:
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +0200981 * o %0 in case of success;
Artem Bityutskiy815bc5f2009-06-08 19:28:18 +0300982 * o %MOVE_CANCEL_RACE, %MOVE_TARGET_WR_ERR, %MOVE_CANCEL_BITFLIPS, etc;
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +0200983 * o a negative error code in case of failure.
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400984 */
985int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
986 struct ubi_vid_hdr *vid_hdr)
987{
Artem Bityutskiy43f9b252007-12-18 15:06:55 +0200988 int err, vol_id, lnum, data_size, aldata_size, idx;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400989 struct ubi_volume *vol;
990 uint32_t crc;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400991
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300992 vol_id = be32_to_cpu(vid_hdr->vol_id);
993 lnum = be32_to_cpu(vid_hdr->lnum);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400994
Artem Bityutskiy87960c02009-05-24 11:58:58 +0300995 dbg_wl("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400996
997 if (vid_hdr->vol_type == UBI_VID_STATIC) {
Christoph Hellwig3261ebd2007-05-21 17:41:46 +0300998 data_size = be32_to_cpu(vid_hdr->data_size);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +0400999 aldata_size = ALIGN(data_size, ubi->min_io_size);
1000 } else
1001 data_size = aldata_size =
Christoph Hellwig3261ebd2007-05-21 17:41:46 +03001002 ubi->leb_size - be32_to_cpu(vid_hdr->data_pad);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001003
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001004 idx = vol_id2idx(ubi, vol_id);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001005 spin_lock(&ubi->volumes_lock);
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001006 /*
1007 * Note, we may race with volume deletion, which means that the volume
1008 * this logical eraseblock belongs to might be being deleted. Since the
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +02001009 * volume deletion un-maps all the volume's logical eraseblocks, it will
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001010 * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
1011 */
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001012 vol = ubi->volumes[idx];
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001013 spin_unlock(&ubi->volumes_lock);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001014 if (!vol) {
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001015 /* No need to do further work, cancel */
Artem Bityutskiy87960c02009-05-24 11:58:58 +03001016 dbg_wl("volume %d is being removed, cancel", vol_id);
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001017 return MOVE_CANCEL_RACE;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001018 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001019
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001020 /*
1021 * We do not want anybody to write to this logical eraseblock while we
1022 * are moving it, so lock it.
1023 *
1024 * Note, we are using non-waiting locking here, because we cannot sleep
1025 * on the LEB, since it may cause deadlocks. Indeed, imagine a task is
1026 * unmapping the LEB which is mapped to the PEB we are going to move
1027 * (@from). This task locks the LEB and goes sleep in the
1028 * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
1029 * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001030 * LEB is already locked, we just do not move it and return
1031 * %MOVE_CANCEL_RACE, which means that UBI will re-try, but later.
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001032 */
1033 err = leb_write_trylock(ubi, vol_id, lnum);
1034 if (err) {
Artem Bityutskiy87960c02009-05-24 11:58:58 +03001035 dbg_wl("contention on LEB %d:%d, cancel", vol_id, lnum);
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001036 return MOVE_CANCEL_RACE;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001037 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001038
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001039 /*
1040 * The LEB might have been put meanwhile, and the task which put it is
1041 * probably waiting on @ubi->move_mutex. No need to continue the work,
1042 * cancel it.
1043 */
1044 if (vol->eba_tbl[lnum] != from) {
Artem Bityutskiy87960c02009-05-24 11:58:58 +03001045 dbg_wl("LEB %d:%d is no longer mapped to PEB %d, mapped to "
1046 "PEB %d, cancel", vol_id, lnum, from,
1047 vol->eba_tbl[lnum]);
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001048 err = MOVE_CANCEL_RACE;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001049 goto out_unlock_leb;
1050 }
1051
1052 /*
Zoltan Sogorb77bcb02008-10-29 09:50:02 +01001053 * OK, now the LEB is locked and we can safely start moving it. Since
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001054 * this function utilizes the @ubi->peb_buf1 buffer which is shared
1055 * with some other functions - we lock the buffer by taking the
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001056 * @ubi->buf_mutex.
1057 */
1058 mutex_lock(&ubi->buf_mutex);
Artem Bityutskiy87960c02009-05-24 11:58:58 +03001059 dbg_wl("read %d bytes of data", aldata_size);
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001060 err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001061 if (err && err != UBI_IO_BITFLIPS) {
1062 ubi_warn("error %d while reading data from PEB %d",
1063 err, from);
Artem Bityutskiy6b5c94c2009-06-10 11:32:05 +03001064 err = MOVE_SOURCE_RD_ERR;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001065 goto out_unlock_buf;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001066 }
1067
1068 /*
Anand Gadiyarfd589a82009-07-16 17:13:03 +02001069 * Now we have got to calculate how much data we have to copy. In
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001070 * case of a static volume it is fairly easy - the VID header contains
1071 * the data size. In case of a dynamic volume it is more difficult - we
1072 * have to read the contents, cut 0xFF bytes from the end and copy only
1073 * the first part. We must do this to avoid writing 0xFF bytes as it
1074 * may have some side-effects. And not only this. It is important not
1075 * to include those 0xFFs to CRC because later the they may be filled
1076 * by data.
1077 */
1078 if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
1079 aldata_size = data_size =
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001080 ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001081
1082 cond_resched();
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001083 crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001084 cond_resched();
1085
1086 /*
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001087 * It may turn out to be that the whole @from physical eraseblock
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001088 * contains only 0xFF bytes. Then we have to only write the VID header
1089 * and do not write any data. This also means we should not set
1090 * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
1091 */
1092 if (data_size > 0) {
1093 vid_hdr->copy_flag = 1;
Christoph Hellwig3261ebd2007-05-21 17:41:46 +03001094 vid_hdr->data_size = cpu_to_be32(data_size);
1095 vid_hdr->data_crc = cpu_to_be32(crc);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001096 }
Christoph Hellwig3261ebd2007-05-21 17:41:46 +03001097 vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001098
1099 err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +02001100 if (err) {
1101 if (err == -EIO)
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001102 err = MOVE_TARGET_WR_ERR;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001103 goto out_unlock_buf;
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +02001104 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001105
1106 cond_resched();
1107
1108 /* Read the VID header back and check if it was written correctly */
1109 err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
1110 if (err) {
Artem Bityutskiyb86a2c52009-05-24 14:13:34 +03001111 if (err != UBI_IO_BITFLIPS) {
Artem Bityutskiy6b5c94c2009-06-10 11:32:05 +03001112 ubi_warn("error %d while reading VID header back from "
1113 "PEB %d", err, to);
1114 if (is_error_sane(err))
Artem Bityutskiyb86a2c52009-05-24 14:13:34 +03001115 err = MOVE_TARGET_RD_ERR;
1116 } else
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001117 err = MOVE_CANCEL_BITFLIPS;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001118 goto out_unlock_buf;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001119 }
1120
1121 if (data_size > 0) {
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001122 err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +02001123 if (err) {
1124 if (err == -EIO)
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001125 err = MOVE_TARGET_WR_ERR;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001126 goto out_unlock_buf;
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +02001127 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001128
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001129 cond_resched();
1130
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001131 /*
1132 * We've written the data and are going to read it back to make
1133 * sure it was written correctly.
1134 */
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001135
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001136 err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001137 if (err) {
Artem Bityutskiyb86a2c52009-05-24 14:13:34 +03001138 if (err != UBI_IO_BITFLIPS) {
Artem Bityutskiy6b5c94c2009-06-10 11:32:05 +03001139 ubi_warn("error %d while reading data back "
1140 "from PEB %d", err, to);
1141 if (is_error_sane(err))
Artem Bityutskiyb86a2c52009-05-24 14:13:34 +03001142 err = MOVE_TARGET_RD_ERR;
1143 } else
Artem Bityutskiy90bf0262009-05-23 16:04:17 +03001144 err = MOVE_CANCEL_BITFLIPS;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001145 goto out_unlock_buf;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001146 }
1147
1148 cond_resched();
1149
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001150 if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
Artem Bityutskiy6fa6f5b2008-12-05 13:37:02 +02001151 ubi_warn("read data back from PEB %d and it is "
1152 "different", to);
1153 err = -EINVAL;
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001154 goto out_unlock_buf;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001155 }
1156 }
1157
1158 ubi_assert(vol->eba_tbl[lnum] == from);
1159 vol->eba_tbl[lnum] = to;
1160
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001161out_unlock_buf:
Artem Bityutskiye88d6e102007-08-29 14:51:52 +03001162 mutex_unlock(&ubi->buf_mutex);
Artem Bityutskiy43f9b252007-12-18 15:06:55 +02001163out_unlock_leb:
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001164 leb_write_unlock(ubi, vol_id, lnum);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001165 return err;
1166}
1167
1168/**
Artem Bityutskiy64d4b4c2010-07-30 14:59:50 +03001169 * print_rsvd_warning - warn about not having enough reserved PEBs.
1170 * @ubi: UBI device description object
1171 *
1172 * This is a helper function for 'ubi_eba_init_scan()' which is called when UBI
1173 * cannot reserve enough PEBs for bad block handling. This function makes a
1174 * decision whether we have to print a warning or not. The algorithm is as
1175 * follows:
1176 * o if this is a new UBI image, then just print the warning
1177 * o if this is an UBI image which has already been used for some time, print
1178 * a warning only if we can reserve less than 10% of the expected amount of
1179 * the reserved PEB.
1180 *
1181 * The idea is that when UBI is used, PEBs become bad, and the reserved pool
1182 * of PEBs becomes smaller, which is normal and we do not want to scare users
1183 * with a warning every time they attach the MTD device. This was an issue
1184 * reported by real users.
1185 */
1186static void print_rsvd_warning(struct ubi_device *ubi,
1187 struct ubi_scan_info *si)
1188{
1189 /*
1190 * The 1 << 18 (256KiB) number is picked randomly, just a reasonably
1191 * large number to distinguish between newly flashed and used images.
1192 */
1193 if (si->max_sqnum > (1 << 18)) {
1194 int min = ubi->beb_rsvd_level / 10;
1195
1196 if (!min)
1197 min = 1;
1198 if (ubi->beb_rsvd_pebs > min)
1199 return;
1200 }
1201
1202 ubi_warn("cannot reserve enough PEBs for bad PEB handling, reserved %d,"
1203 " need %d", ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
Artem Bityutskiy5fc01ab2010-09-03 23:08:15 +03001204 if (ubi->corr_peb_count)
1205 ubi_warn("%d PEBs are corrupted and not used",
1206 ubi->corr_peb_count);
Artem Bityutskiy64d4b4c2010-07-30 14:59:50 +03001207}
1208
1209/**
Artem Bityutskiy85c6e6e2008-07-16 10:25:56 +03001210 * ubi_eba_init_scan - initialize the EBA sub-system using scanning information.
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001211 * @ubi: UBI device description object
1212 * @si: scanning information
1213 *
1214 * This function returns zero in case of success and a negative error code in
1215 * case of failure.
1216 */
1217int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
1218{
1219 int i, j, err, num_volumes;
1220 struct ubi_scan_volume *sv;
1221 struct ubi_volume *vol;
1222 struct ubi_scan_leb *seb;
1223 struct rb_node *rb;
1224
Artem Bityutskiy85c6e6e2008-07-16 10:25:56 +03001225 dbg_eba("initialize EBA sub-system");
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001226
1227 spin_lock_init(&ubi->ltree_lock);
Artem Bityutskiye8823bd2007-09-13 14:28:14 +03001228 mutex_init(&ubi->alc_mutex);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001229 ubi->ltree = RB_ROOT;
1230
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001231 ubi->global_sqnum = si->max_sqnum + 1;
1232 num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
1233
1234 for (i = 0; i < num_volumes; i++) {
1235 vol = ubi->volumes[i];
1236 if (!vol)
1237 continue;
1238
1239 cond_resched();
1240
1241 vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int),
1242 GFP_KERNEL);
1243 if (!vol->eba_tbl) {
1244 err = -ENOMEM;
1245 goto out_free;
1246 }
1247
1248 for (j = 0; j < vol->reserved_pebs; j++)
1249 vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
1250
1251 sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
1252 if (!sv)
1253 continue;
1254
1255 ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
1256 if (seb->lnum >= vol->reserved_pebs)
1257 /*
1258 * This may happen in case of an unclean reboot
1259 * during re-size.
1260 */
1261 ubi_scan_move_to_list(sv, seb, &si->erase);
1262 vol->eba_tbl[seb->lnum] = seb->pnum;
1263 }
1264 }
1265
Artem Bityutskiy94780d42007-12-04 21:36:12 +02001266 if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
1267 ubi_err("no enough physical eraseblocks (%d, need %d)",
1268 ubi->avail_pebs, EBA_RESERVED_PEBS);
Artem Bityutskiy5fc01ab2010-09-03 23:08:15 +03001269 if (ubi->corr_peb_count)
1270 ubi_err("%d PEBs are corrupted and not used",
1271 ubi->corr_peb_count);
Artem Bityutskiy94780d42007-12-04 21:36:12 +02001272 err = -ENOSPC;
1273 goto out_free;
1274 }
1275 ubi->avail_pebs -= EBA_RESERVED_PEBS;
1276 ubi->rsvd_pebs += EBA_RESERVED_PEBS;
1277
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001278 if (ubi->bad_allowed) {
1279 ubi_calculate_reserved(ubi);
1280
1281 if (ubi->avail_pebs < ubi->beb_rsvd_level) {
1282 /* No enough free physical eraseblocks */
1283 ubi->beb_rsvd_pebs = ubi->avail_pebs;
Artem Bityutskiy64d4b4c2010-07-30 14:59:50 +03001284 print_rsvd_warning(ubi, si);
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001285 } else
1286 ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
1287
1288 ubi->avail_pebs -= ubi->beb_rsvd_pebs;
1289 ubi->rsvd_pebs += ubi->beb_rsvd_pebs;
1290 }
1291
Artem Bityutskiy85c6e6e2008-07-16 10:25:56 +03001292 dbg_eba("EBA sub-system is initialized");
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001293 return 0;
1294
1295out_free:
1296 for (i = 0; i < num_volumes; i++) {
1297 if (!ubi->volumes[i])
1298 continue;
1299 kfree(ubi->volumes[i]->eba_tbl);
Adrian Hunter7194e6f2009-07-24 17:05:00 +03001300 ubi->volumes[i]->eba_tbl = NULL;
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001301 }
Artem B. Bityutskiy801c1352006-06-27 12:22:22 +04001302 return err;
1303}