Commit | Line | Data |
---|---|---|
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 | 60 | unsigned 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 | */ | |
79 | static 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 |
96 | static 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 |
135 | static 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 | */ | |
205 | static 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 | */ | |
222 | static 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 | */ | |
247 | static 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 | */ | |
269 | static 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 | */ | |
298 | static 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 |
324 | int 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 | ||
8974b15c | 343 | down_read(&ubi->fm_sem); |
801c135c | 344 | vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED; |
8974b15c | 345 | up_read(&ubi->fm_sem); |
d36e59e6 | 346 | err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 0); |
801c135c AB |
347 | |
348 | out_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 |
372 | int 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 | ||
404 | retry: | |
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) { |
049333ce AB |
425 | ubi_warn("corrupted VID header at PEB %d, LEB %d:%d", |
426 | pnum, vol_id, lnum); | |
801c135c AB |
427 | err = -EBADMSG; |
428 | } else | |
429 | ubi_ro_mode(ubi); | |
430 | } | |
431 | goto out_free; | |
432 | } else if (err == UBI_IO_BITFLIPS) | |
433 | scrub = 1; | |
434 | ||
3261ebd7 CH |
435 | ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs)); |
436 | ubi_assert(len == be32_to_cpu(vid_hdr->data_size)); | |
801c135c | 437 | |
3261ebd7 | 438 | crc = be32_to_cpu(vid_hdr->data_crc); |
801c135c AB |
439 | ubi_free_vid_hdr(ubi, vid_hdr); |
440 | } | |
441 | ||
442 | err = ubi_io_read_data(ubi, buf, pnum, offset, len); | |
443 | if (err) { | |
444 | if (err == UBI_IO_BITFLIPS) { | |
445 | scrub = 1; | |
446 | err = 0; | |
d57f4054 | 447 | } else if (mtd_is_eccerr(err)) { |
801c135c AB |
448 | if (vol->vol_type == UBI_DYNAMIC_VOLUME) |
449 | goto out_unlock; | |
450 | scrub = 1; | |
451 | if (!check) { | |
452 | ubi_msg("force data checking"); | |
453 | check = 1; | |
454 | goto retry; | |
455 | } | |
456 | } else | |
457 | goto out_unlock; | |
458 | } | |
459 | ||
460 | if (check) { | |
2ab934b8 | 461 | uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len); |
801c135c AB |
462 | if (crc1 != crc) { |
463 | ubi_warn("CRC error: calculated %#08x, must be %#08x", | |
464 | crc1, crc); | |
465 | err = -EBADMSG; | |
466 | goto out_unlock; | |
467 | } | |
468 | } | |
469 | ||
470 | if (scrub) | |
471 | err = ubi_wl_scrub_peb(ubi, pnum); | |
472 | ||
473 | leb_read_unlock(ubi, vol_id, lnum); | |
474 | return err; | |
475 | ||
476 | out_free: | |
477 | ubi_free_vid_hdr(ubi, vid_hdr); | |
478 | out_unlock: | |
479 | leb_read_unlock(ubi, vol_id, lnum); | |
480 | return err; | |
481 | } | |
482 | ||
483 | /** | |
484 | * recover_peb - recover from write failure. | |
485 | * @ubi: UBI device description object | |
486 | * @pnum: the physical eraseblock to recover | |
487 | * @vol_id: volume ID | |
488 | * @lnum: logical eraseblock number | |
489 | * @buf: data which was not written because of the write failure | |
490 | * @offset: offset of the failed write | |
491 | * @len: how many bytes should have been written | |
492 | * | |
493 | * This function is called in case of a write failure and moves all good data | |
494 | * from the potentially bad physical eraseblock to a good physical eraseblock. | |
495 | * This function also writes the data which was not written due to the failure. | |
496 | * Returns new physical eraseblock number in case of success, and a negative | |
497 | * error code in case of failure. | |
498 | */ | |
499 | static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum, | |
500 | const void *buf, int offset, int len) | |
501 | { | |
502 | int err, idx = vol_id2idx(ubi, vol_id), new_pnum, data_size, tries = 0; | |
503 | struct ubi_volume *vol = ubi->volumes[idx]; | |
504 | struct ubi_vid_hdr *vid_hdr; | |
801c135c | 505 | |
33818bbb | 506 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
9c9ec147 | 507 | if (!vid_hdr) |
801c135c | 508 | return -ENOMEM; |
801c135c AB |
509 | |
510 | retry: | |
b36a261e | 511 | new_pnum = ubi_wl_get_peb(ubi); |
801c135c AB |
512 | if (new_pnum < 0) { |
513 | ubi_free_vid_hdr(ubi, vid_hdr); | |
514 | return new_pnum; | |
515 | } | |
516 | ||
517 | ubi_msg("recover PEB %d, move data to PEB %d", pnum, new_pnum); | |
518 | ||
519 | err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1); | |
520 | if (err && err != UBI_IO_BITFLIPS) { | |
521 | if (err > 0) | |
522 | err = -EIO; | |
523 | goto out_put; | |
524 | } | |
525 | ||
a7306653 | 526 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
801c135c AB |
527 | err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr); |
528 | if (err) | |
529 | goto write_error; | |
530 | ||
531 | data_size = offset + len; | |
4df581f3 | 532 | mutex_lock(&ubi->buf_mutex); |
0ca39d74 | 533 | memset(ubi->peb_buf + offset, 0xFF, len); |
801c135c AB |
534 | |
535 | /* Read everything before the area where the write failure happened */ | |
536 | if (offset > 0) { | |
0ca39d74 | 537 | err = ubi_io_read_data(ubi, ubi->peb_buf, pnum, 0, offset); |
e88d6e10 | 538 | if (err && err != UBI_IO_BITFLIPS) |
4df581f3 | 539 | goto out_unlock; |
801c135c AB |
540 | } |
541 | ||
0ca39d74 | 542 | memcpy(ubi->peb_buf + offset, buf, len); |
801c135c | 543 | |
0ca39d74 | 544 | err = ubi_io_write_data(ubi, ubi->peb_buf, new_pnum, 0, data_size); |
4df581f3 AB |
545 | if (err) { |
546 | mutex_unlock(&ubi->buf_mutex); | |
801c135c | 547 | goto write_error; |
4df581f3 | 548 | } |
801c135c | 549 | |
e88d6e10 | 550 | mutex_unlock(&ubi->buf_mutex); |
801c135c AB |
551 | ubi_free_vid_hdr(ubi, vid_hdr); |
552 | ||
8974b15c | 553 | down_read(&ubi->fm_sem); |
801c135c | 554 | vol->eba_tbl[lnum] = new_pnum; |
8974b15c | 555 | up_read(&ubi->fm_sem); |
d36e59e6 | 556 | ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1); |
801c135c AB |
557 | |
558 | ubi_msg("data was successfully recovered"); | |
559 | return 0; | |
560 | ||
4df581f3 | 561 | out_unlock: |
e88d6e10 | 562 | mutex_unlock(&ubi->buf_mutex); |
4df581f3 | 563 | out_put: |
d36e59e6 | 564 | ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1); |
801c135c AB |
565 | ubi_free_vid_hdr(ubi, vid_hdr); |
566 | return err; | |
567 | ||
568 | write_error: | |
569 | /* | |
570 | * Bad luck? This physical eraseblock is bad too? Crud. Let's try to | |
571 | * get another one. | |
572 | */ | |
573 | ubi_warn("failed to write to PEB %d", new_pnum); | |
d36e59e6 | 574 | ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1); |
801c135c AB |
575 | if (++tries > UBI_IO_RETRIES) { |
576 | ubi_free_vid_hdr(ubi, vid_hdr); | |
577 | return err; | |
578 | } | |
579 | ubi_msg("try again"); | |
580 | goto retry; | |
581 | } | |
582 | ||
583 | /** | |
584 | * ubi_eba_write_leb - write data to dynamic volume. | |
585 | * @ubi: UBI device description object | |
89b96b69 | 586 | * @vol: volume description object |
801c135c AB |
587 | * @lnum: logical eraseblock number |
588 | * @buf: the data to write | |
589 | * @offset: offset within the logical eraseblock where to write | |
590 | * @len: how many bytes to write | |
801c135c AB |
591 | * |
592 | * This function writes data to logical eraseblock @lnum of a dynamic volume | |
89b96b69 | 593 | * @vol. Returns zero in case of success and a negative error code in case |
801c135c AB |
594 | * of failure. In case of error, it is possible that something was still |
595 | * written to the flash media, but may be some garbage. | |
596 | */ | |
89b96b69 | 597 | int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum, |
b36a261e | 598 | const void *buf, int offset, int len) |
801c135c | 599 | { |
89b96b69 | 600 | int err, pnum, tries = 0, vol_id = vol->vol_id; |
801c135c AB |
601 | struct ubi_vid_hdr *vid_hdr; |
602 | ||
603 | if (ubi->ro_mode) | |
604 | return -EROFS; | |
605 | ||
606 | err = leb_write_lock(ubi, vol_id, lnum); | |
607 | if (err) | |
608 | return err; | |
609 | ||
610 | pnum = vol->eba_tbl[lnum]; | |
611 | if (pnum >= 0) { | |
612 | dbg_eba("write %d bytes at offset %d of LEB %d:%d, PEB %d", | |
613 | len, offset, vol_id, lnum, pnum); | |
614 | ||
615 | err = ubi_io_write_data(ubi, buf, pnum, offset, len); | |
616 | if (err) { | |
617 | ubi_warn("failed to write data to PEB %d", pnum); | |
618 | if (err == -EIO && ubi->bad_allowed) | |
89b96b69 AB |
619 | err = recover_peb(ubi, pnum, vol_id, lnum, buf, |
620 | offset, len); | |
801c135c AB |
621 | if (err) |
622 | ubi_ro_mode(ubi); | |
623 | } | |
624 | leb_write_unlock(ubi, vol_id, lnum); | |
625 | return err; | |
626 | } | |
627 | ||
628 | /* | |
629 | * The logical eraseblock is not mapped. We have to get a free physical | |
630 | * eraseblock and write the volume identifier header there first. | |
631 | */ | |
33818bbb | 632 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
633 | if (!vid_hdr) { |
634 | leb_write_unlock(ubi, vol_id, lnum); | |
635 | return -ENOMEM; | |
636 | } | |
637 | ||
638 | vid_hdr->vol_type = UBI_VID_DYNAMIC; | |
a7306653 | 639 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
3261ebd7 CH |
640 | vid_hdr->vol_id = cpu_to_be32(vol_id); |
641 | vid_hdr->lnum = cpu_to_be32(lnum); | |
801c135c | 642 | vid_hdr->compat = ubi_get_compat(ubi, vol_id); |
3261ebd7 | 643 | vid_hdr->data_pad = cpu_to_be32(vol->data_pad); |
801c135c AB |
644 | |
645 | retry: | |
b36a261e | 646 | pnum = ubi_wl_get_peb(ubi); |
801c135c AB |
647 | if (pnum < 0) { |
648 | ubi_free_vid_hdr(ubi, vid_hdr); | |
649 | leb_write_unlock(ubi, vol_id, lnum); | |
650 | return pnum; | |
651 | } | |
652 | ||
653 | dbg_eba("write VID hdr and %d bytes at offset %d of LEB %d:%d, PEB %d", | |
654 | len, offset, vol_id, lnum, pnum); | |
655 | ||
656 | err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr); | |
657 | if (err) { | |
658 | ubi_warn("failed to write VID header to LEB %d:%d, PEB %d", | |
659 | vol_id, lnum, pnum); | |
660 | goto write_error; | |
661 | } | |
662 | ||
393852ec AB |
663 | if (len) { |
664 | err = ubi_io_write_data(ubi, buf, pnum, offset, len); | |
665 | if (err) { | |
049333ce AB |
666 | ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, PEB %d", |
667 | len, offset, vol_id, lnum, pnum); | |
393852ec AB |
668 | goto write_error; |
669 | } | |
801c135c AB |
670 | } |
671 | ||
8974b15c | 672 | down_read(&ubi->fm_sem); |
801c135c | 673 | vol->eba_tbl[lnum] = pnum; |
8974b15c | 674 | up_read(&ubi->fm_sem); |
801c135c AB |
675 | |
676 | leb_write_unlock(ubi, vol_id, lnum); | |
677 | ubi_free_vid_hdr(ubi, vid_hdr); | |
678 | return 0; | |
679 | ||
680 | write_error: | |
681 | if (err != -EIO || !ubi->bad_allowed) { | |
682 | ubi_ro_mode(ubi); | |
683 | leb_write_unlock(ubi, vol_id, lnum); | |
684 | ubi_free_vid_hdr(ubi, vid_hdr); | |
685 | return err; | |
686 | } | |
687 | ||
688 | /* | |
689 | * Fortunately, this is the first write operation to this physical | |
690 | * eraseblock, so just put it and request a new one. We assume that if | |
691 | * this physical eraseblock went bad, the erase code will handle that. | |
692 | */ | |
d36e59e6 | 693 | err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1); |
801c135c AB |
694 | if (err || ++tries > UBI_IO_RETRIES) { |
695 | ubi_ro_mode(ubi); | |
696 | leb_write_unlock(ubi, vol_id, lnum); | |
697 | ubi_free_vid_hdr(ubi, vid_hdr); | |
698 | return err; | |
699 | } | |
700 | ||
a7306653 | 701 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
801c135c AB |
702 | ubi_msg("try another PEB"); |
703 | goto retry; | |
704 | } | |
705 | ||
706 | /** | |
707 | * ubi_eba_write_leb_st - write data to static volume. | |
708 | * @ubi: UBI device description object | |
89b96b69 | 709 | * @vol: volume description object |
801c135c AB |
710 | * @lnum: logical eraseblock number |
711 | * @buf: data to write | |
712 | * @len: how many bytes to write | |
801c135c AB |
713 | * @used_ebs: how many logical eraseblocks will this volume contain |
714 | * | |
715 | * This function writes data to logical eraseblock @lnum of static volume | |
89b96b69 | 716 | * @vol. The @used_ebs argument should contain total number of logical |
801c135c AB |
717 | * eraseblock in this static volume. |
718 | * | |
719 | * When writing to the last logical eraseblock, the @len argument doesn't have | |
720 | * to be aligned to the minimal I/O unit size. Instead, it has to be equivalent | |
721 | * to the real data size, although the @buf buffer has to contain the | |
722 | * alignment. In all other cases, @len has to be aligned. | |
723 | * | |
025dfdaf | 724 | * It is prohibited to write more than once to logical eraseblocks of static |
801c135c AB |
725 | * volumes. This function returns zero in case of success and a negative error |
726 | * code in case of failure. | |
727 | */ | |
89b96b69 | 728 | int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol, |
b36a261e | 729 | int lnum, const void *buf, int len, int used_ebs) |
801c135c | 730 | { |
89b96b69 | 731 | int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id; |
801c135c AB |
732 | struct ubi_vid_hdr *vid_hdr; |
733 | uint32_t crc; | |
734 | ||
735 | if (ubi->ro_mode) | |
736 | return -EROFS; | |
737 | ||
738 | if (lnum == used_ebs - 1) | |
739 | /* If this is the last LEB @len may be unaligned */ | |
740 | len = ALIGN(data_size, ubi->min_io_size); | |
741 | else | |
cadb40cc | 742 | ubi_assert(!(len & (ubi->min_io_size - 1))); |
801c135c | 743 | |
33818bbb | 744 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
745 | if (!vid_hdr) |
746 | return -ENOMEM; | |
747 | ||
748 | err = leb_write_lock(ubi, vol_id, lnum); | |
749 | if (err) { | |
750 | ubi_free_vid_hdr(ubi, vid_hdr); | |
751 | return err; | |
752 | } | |
753 | ||
a7306653 | 754 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
3261ebd7 CH |
755 | vid_hdr->vol_id = cpu_to_be32(vol_id); |
756 | vid_hdr->lnum = cpu_to_be32(lnum); | |
801c135c | 757 | vid_hdr->compat = ubi_get_compat(ubi, vol_id); |
3261ebd7 | 758 | vid_hdr->data_pad = cpu_to_be32(vol->data_pad); |
801c135c AB |
759 | |
760 | crc = crc32(UBI_CRC32_INIT, buf, data_size); | |
761 | vid_hdr->vol_type = UBI_VID_STATIC; | |
3261ebd7 CH |
762 | vid_hdr->data_size = cpu_to_be32(data_size); |
763 | vid_hdr->used_ebs = cpu_to_be32(used_ebs); | |
764 | vid_hdr->data_crc = cpu_to_be32(crc); | |
801c135c AB |
765 | |
766 | retry: | |
b36a261e | 767 | pnum = ubi_wl_get_peb(ubi); |
801c135c AB |
768 | if (pnum < 0) { |
769 | ubi_free_vid_hdr(ubi, vid_hdr); | |
770 | leb_write_unlock(ubi, vol_id, lnum); | |
771 | return pnum; | |
772 | } | |
773 | ||
774 | dbg_eba("write VID hdr and %d bytes at LEB %d:%d, PEB %d, used_ebs %d", | |
775 | len, vol_id, lnum, pnum, used_ebs); | |
776 | ||
777 | err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr); | |
778 | if (err) { | |
779 | ubi_warn("failed to write VID header to LEB %d:%d, PEB %d", | |
780 | vol_id, lnum, pnum); | |
781 | goto write_error; | |
782 | } | |
783 | ||
784 | err = ubi_io_write_data(ubi, buf, pnum, 0, len); | |
785 | if (err) { | |
786 | ubi_warn("failed to write %d bytes of data to PEB %d", | |
787 | len, pnum); | |
788 | goto write_error; | |
789 | } | |
790 | ||
791 | ubi_assert(vol->eba_tbl[lnum] < 0); | |
8974b15c | 792 | down_read(&ubi->fm_sem); |
801c135c | 793 | vol->eba_tbl[lnum] = pnum; |
8974b15c | 794 | up_read(&ubi->fm_sem); |
801c135c AB |
795 | |
796 | leb_write_unlock(ubi, vol_id, lnum); | |
797 | ubi_free_vid_hdr(ubi, vid_hdr); | |
798 | return 0; | |
799 | ||
800 | write_error: | |
801 | if (err != -EIO || !ubi->bad_allowed) { | |
802 | /* | |
803 | * This flash device does not admit of bad eraseblocks or | |
804 | * something nasty and unexpected happened. Switch to read-only | |
805 | * mode just in case. | |
806 | */ | |
807 | ubi_ro_mode(ubi); | |
808 | leb_write_unlock(ubi, vol_id, lnum); | |
809 | ubi_free_vid_hdr(ubi, vid_hdr); | |
810 | return err; | |
811 | } | |
812 | ||
d36e59e6 | 813 | err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1); |
801c135c AB |
814 | if (err || ++tries > UBI_IO_RETRIES) { |
815 | ubi_ro_mode(ubi); | |
816 | leb_write_unlock(ubi, vol_id, lnum); | |
817 | ubi_free_vid_hdr(ubi, vid_hdr); | |
818 | return err; | |
819 | } | |
820 | ||
a7306653 | 821 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
801c135c AB |
822 | ubi_msg("try another PEB"); |
823 | goto retry; | |
824 | } | |
825 | ||
826 | /* | |
827 | * ubi_eba_atomic_leb_change - change logical eraseblock atomically. | |
828 | * @ubi: UBI device description object | |
c63a491d | 829 | * @vol: volume description object |
801c135c AB |
830 | * @lnum: logical eraseblock number |
831 | * @buf: data to write | |
832 | * @len: how many bytes to write | |
801c135c AB |
833 | * |
834 | * This function changes the contents of a logical eraseblock atomically. @buf | |
835 | * has to contain new logical eraseblock data, and @len - the length of the | |
836 | * data, which has to be aligned. This function guarantees that in case of an | |
837 | * unclean reboot the old contents is preserved. Returns zero in case of | |
838 | * success and a negative error code in case of failure. | |
e8823bd6 AB |
839 | * |
840 | * UBI reserves one LEB for the "atomic LEB change" operation, so only one | |
841 | * LEB change may be done at a time. This is ensured by @ubi->alc_mutex. | |
801c135c | 842 | */ |
89b96b69 | 843 | int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol, |
b36a261e | 844 | int lnum, const void *buf, int len) |
801c135c | 845 | { |
89b96b69 | 846 | int err, pnum, tries = 0, vol_id = vol->vol_id; |
801c135c AB |
847 | struct ubi_vid_hdr *vid_hdr; |
848 | uint32_t crc; | |
849 | ||
850 | if (ubi->ro_mode) | |
851 | return -EROFS; | |
852 | ||
60c03153 AB |
853 | if (len == 0) { |
854 | /* | |
855 | * Special case when data length is zero. In this case the LEB | |
856 | * has to be unmapped and mapped somewhere else. | |
857 | */ | |
858 | err = ubi_eba_unmap_leb(ubi, vol, lnum); | |
859 | if (err) | |
860 | return err; | |
b36a261e | 861 | return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0); |
60c03153 AB |
862 | } |
863 | ||
33818bbb | 864 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
865 | if (!vid_hdr) |
866 | return -ENOMEM; | |
867 | ||
e8823bd6 | 868 | mutex_lock(&ubi->alc_mutex); |
801c135c | 869 | err = leb_write_lock(ubi, vol_id, lnum); |
e8823bd6 AB |
870 | if (err) |
871 | goto out_mutex; | |
801c135c | 872 | |
a7306653 | 873 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
3261ebd7 CH |
874 | vid_hdr->vol_id = cpu_to_be32(vol_id); |
875 | vid_hdr->lnum = cpu_to_be32(lnum); | |
801c135c | 876 | vid_hdr->compat = ubi_get_compat(ubi, vol_id); |
3261ebd7 | 877 | vid_hdr->data_pad = cpu_to_be32(vol->data_pad); |
801c135c AB |
878 | |
879 | crc = crc32(UBI_CRC32_INIT, buf, len); | |
84a92580 | 880 | vid_hdr->vol_type = UBI_VID_DYNAMIC; |
3261ebd7 | 881 | vid_hdr->data_size = cpu_to_be32(len); |
801c135c | 882 | vid_hdr->copy_flag = 1; |
3261ebd7 | 883 | vid_hdr->data_crc = cpu_to_be32(crc); |
801c135c AB |
884 | |
885 | retry: | |
b36a261e | 886 | pnum = ubi_wl_get_peb(ubi); |
801c135c | 887 | if (pnum < 0) { |
e8823bd6 AB |
888 | err = pnum; |
889 | goto out_leb_unlock; | |
801c135c AB |
890 | } |
891 | ||
892 | dbg_eba("change LEB %d:%d, PEB %d, write VID hdr to PEB %d", | |
893 | vol_id, lnum, vol->eba_tbl[lnum], pnum); | |
894 | ||
895 | err = ubi_io_write_vid_hdr(ubi, pnum, vid_hdr); | |
896 | if (err) { | |
897 | ubi_warn("failed to write VID header to LEB %d:%d, PEB %d", | |
898 | vol_id, lnum, pnum); | |
899 | goto write_error; | |
900 | } | |
901 | ||
902 | err = ubi_io_write_data(ubi, buf, pnum, 0, len); | |
903 | if (err) { | |
904 | ubi_warn("failed to write %d bytes of data to PEB %d", | |
905 | len, pnum); | |
906 | goto write_error; | |
907 | } | |
908 | ||
a443db48 | 909 | if (vol->eba_tbl[lnum] >= 0) { |
d36e59e6 | 910 | err = ubi_wl_put_peb(ubi, vol_id, lnum, vol->eba_tbl[lnum], 0); |
e8823bd6 AB |
911 | if (err) |
912 | goto out_leb_unlock; | |
801c135c AB |
913 | } |
914 | ||
8974b15c | 915 | down_read(&ubi->fm_sem); |
801c135c | 916 | vol->eba_tbl[lnum] = pnum; |
8974b15c | 917 | up_read(&ubi->fm_sem); |
e8823bd6 AB |
918 | |
919 | out_leb_unlock: | |
801c135c | 920 | leb_write_unlock(ubi, vol_id, lnum); |
e8823bd6 AB |
921 | out_mutex: |
922 | mutex_unlock(&ubi->alc_mutex); | |
801c135c | 923 | ubi_free_vid_hdr(ubi, vid_hdr); |
e8823bd6 | 924 | return err; |
801c135c AB |
925 | |
926 | write_error: | |
927 | if (err != -EIO || !ubi->bad_allowed) { | |
928 | /* | |
929 | * This flash device does not admit of bad eraseblocks or | |
930 | * something nasty and unexpected happened. Switch to read-only | |
931 | * mode just in case. | |
932 | */ | |
933 | ubi_ro_mode(ubi); | |
e8823bd6 | 934 | goto out_leb_unlock; |
801c135c AB |
935 | } |
936 | ||
d36e59e6 | 937 | err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1); |
801c135c AB |
938 | if (err || ++tries > UBI_IO_RETRIES) { |
939 | ubi_ro_mode(ubi); | |
e8823bd6 | 940 | goto out_leb_unlock; |
801c135c AB |
941 | } |
942 | ||
a7306653 | 943 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
801c135c AB |
944 | ubi_msg("try another PEB"); |
945 | goto retry; | |
946 | } | |
947 | ||
6b5c94c6 AB |
948 | /** |
949 | * is_error_sane - check whether a read error is sane. | |
950 | * @err: code of the error happened during reading | |
951 | * | |
952 | * This is a helper function for 'ubi_eba_copy_leb()' which is called when we | |
953 | * cannot read data from the target PEB (an error @err happened). If the error | |
954 | * code is sane, then we treat this error as non-fatal. Otherwise the error is | |
955 | * fatal and UBI will be switched to R/O mode later. | |
956 | * | |
957 | * The idea is that we try not to switch to R/O mode if the read error is | |
958 | * something which suggests there was a real read problem. E.g., %-EIO. Or a | |
959 | * memory allocation failed (-%ENOMEM). Otherwise, it is safer to switch to R/O | |
960 | * mode, simply because we do not know what happened at the MTD level, and we | |
961 | * cannot handle this. E.g., the underlying driver may have become crazy, and | |
962 | * it is safer to switch to R/O mode to preserve the data. | |
963 | * | |
964 | * And bear in mind, this is about reading from the target PEB, i.e. the PEB | |
965 | * which we have just written. | |
966 | */ | |
967 | static int is_error_sane(int err) | |
968 | { | |
786d7831 | 969 | if (err == -EIO || err == -ENOMEM || err == UBI_IO_BAD_HDR || |
756e1df1 | 970 | err == UBI_IO_BAD_HDR_EBADMSG || err == -ETIMEDOUT) |
6b5c94c6 AB |
971 | return 0; |
972 | return 1; | |
973 | } | |
974 | ||
801c135c AB |
975 | /** |
976 | * ubi_eba_copy_leb - copy logical eraseblock. | |
977 | * @ubi: UBI device description object | |
978 | * @from: physical eraseblock number from where to copy | |
979 | * @to: physical eraseblock number where to copy | |
980 | * @vid_hdr: VID header of the @from physical eraseblock | |
981 | * | |
982 | * This function copies logical eraseblock from physical eraseblock @from to | |
983 | * physical eraseblock @to. The @vid_hdr buffer may be changed by this | |
43f9b25a | 984 | * function. Returns: |
6fa6f5bb | 985 | * o %0 in case of success; |
cc831464 | 986 | * o %MOVE_CANCEL_RACE, %MOVE_TARGET_WR_ERR, %MOVE_TARGET_BITFLIPS, etc; |
6fa6f5bb | 987 | * o a negative error code in case of failure. |
801c135c AB |
988 | */ |
989 | int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to, | |
990 | struct ubi_vid_hdr *vid_hdr) | |
991 | { | |
43f9b25a | 992 | int err, vol_id, lnum, data_size, aldata_size, idx; |
801c135c AB |
993 | struct ubi_volume *vol; |
994 | uint32_t crc; | |
801c135c | 995 | |
3261ebd7 CH |
996 | vol_id = be32_to_cpu(vid_hdr->vol_id); |
997 | lnum = be32_to_cpu(vid_hdr->lnum); | |
801c135c | 998 | |
87960c0b | 999 | dbg_wl("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to); |
801c135c AB |
1000 | |
1001 | if (vid_hdr->vol_type == UBI_VID_STATIC) { | |
3261ebd7 | 1002 | data_size = be32_to_cpu(vid_hdr->data_size); |
801c135c AB |
1003 | aldata_size = ALIGN(data_size, ubi->min_io_size); |
1004 | } else | |
1005 | data_size = aldata_size = | |
3261ebd7 | 1006 | ubi->leb_size - be32_to_cpu(vid_hdr->data_pad); |
801c135c | 1007 | |
801c135c | 1008 | idx = vol_id2idx(ubi, vol_id); |
43f9b25a | 1009 | spin_lock(&ubi->volumes_lock); |
801c135c | 1010 | /* |
43f9b25a AB |
1011 | * Note, we may race with volume deletion, which means that the volume |
1012 | * this logical eraseblock belongs to might be being deleted. Since the | |
6fa6f5bb | 1013 | * volume deletion un-maps all the volume's logical eraseblocks, it will |
43f9b25a | 1014 | * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish. |
801c135c | 1015 | */ |
801c135c | 1016 | vol = ubi->volumes[idx]; |
90bf0265 | 1017 | spin_unlock(&ubi->volumes_lock); |
801c135c | 1018 | if (!vol) { |
43f9b25a | 1019 | /* No need to do further work, cancel */ |
87960c0b | 1020 | dbg_wl("volume %d is being removed, cancel", vol_id); |
90bf0265 | 1021 | return MOVE_CANCEL_RACE; |
801c135c AB |
1022 | } |
1023 | ||
43f9b25a AB |
1024 | /* |
1025 | * We do not want anybody to write to this logical eraseblock while we | |
1026 | * are moving it, so lock it. | |
1027 | * | |
1028 | * Note, we are using non-waiting locking here, because we cannot sleep | |
1029 | * on the LEB, since it may cause deadlocks. Indeed, imagine a task is | |
1030 | * unmapping the LEB which is mapped to the PEB we are going to move | |
1031 | * (@from). This task locks the LEB and goes sleep in the | |
1032 | * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are | |
1033 | * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the | |
90bf0265 | 1034 | * LEB is already locked, we just do not move it and return |
e801e128 BP |
1035 | * %MOVE_RETRY. Note, we do not return %MOVE_CANCEL_RACE here because |
1036 | * we do not know the reasons of the contention - it may be just a | |
1037 | * normal I/O on this LEB, so we want to re-try. | |
43f9b25a AB |
1038 | */ |
1039 | err = leb_write_trylock(ubi, vol_id, lnum); | |
1040 | if (err) { | |
87960c0b | 1041 | dbg_wl("contention on LEB %d:%d, cancel", vol_id, lnum); |
e801e128 | 1042 | return MOVE_RETRY; |
801c135c | 1043 | } |
801c135c | 1044 | |
43f9b25a AB |
1045 | /* |
1046 | * The LEB might have been put meanwhile, and the task which put it is | |
1047 | * probably waiting on @ubi->move_mutex. No need to continue the work, | |
1048 | * cancel it. | |
1049 | */ | |
1050 | if (vol->eba_tbl[lnum] != from) { | |
049333ce AB |
1051 | dbg_wl("LEB %d:%d is no longer mapped to PEB %d, mapped to PEB %d, cancel", |
1052 | vol_id, lnum, from, vol->eba_tbl[lnum]); | |
90bf0265 | 1053 | err = MOVE_CANCEL_RACE; |
43f9b25a AB |
1054 | goto out_unlock_leb; |
1055 | } | |
801c135c | 1056 | |
43f9b25a | 1057 | /* |
b77bcb07 | 1058 | * OK, now the LEB is locked and we can safely start moving it. Since |
0ca39d74 | 1059 | * this function utilizes the @ubi->peb_buf buffer which is shared |
90bf0265 | 1060 | * with some other functions - we lock the buffer by taking the |
43f9b25a AB |
1061 | * @ubi->buf_mutex. |
1062 | */ | |
1063 | mutex_lock(&ubi->buf_mutex); | |
87960c0b | 1064 | dbg_wl("read %d bytes of data", aldata_size); |
0ca39d74 | 1065 | err = ubi_io_read_data(ubi, ubi->peb_buf, from, 0, aldata_size); |
801c135c AB |
1066 | if (err && err != UBI_IO_BITFLIPS) { |
1067 | ubi_warn("error %d while reading data from PEB %d", | |
1068 | err, from); | |
6b5c94c6 | 1069 | err = MOVE_SOURCE_RD_ERR; |
43f9b25a | 1070 | goto out_unlock_buf; |
801c135c AB |
1071 | } |
1072 | ||
1073 | /* | |
fd589a8f | 1074 | * Now we have got to calculate how much data we have to copy. In |
801c135c AB |
1075 | * case of a static volume it is fairly easy - the VID header contains |
1076 | * the data size. In case of a dynamic volume it is more difficult - we | |
1077 | * have to read the contents, cut 0xFF bytes from the end and copy only | |
1078 | * the first part. We must do this to avoid writing 0xFF bytes as it | |
1079 | * may have some side-effects. And not only this. It is important not | |
1080 | * to include those 0xFFs to CRC because later the they may be filled | |
1081 | * by data. | |
1082 | */ | |
1083 | if (vid_hdr->vol_type == UBI_VID_DYNAMIC) | |
1084 | aldata_size = data_size = | |
0ca39d74 | 1085 | ubi_calc_data_len(ubi, ubi->peb_buf, data_size); |
801c135c AB |
1086 | |
1087 | cond_resched(); | |
0ca39d74 | 1088 | crc = crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size); |
801c135c AB |
1089 | cond_resched(); |
1090 | ||
1091 | /* | |
90bf0265 | 1092 | * It may turn out to be that the whole @from physical eraseblock |
801c135c AB |
1093 | * contains only 0xFF bytes. Then we have to only write the VID header |
1094 | * and do not write any data. This also means we should not set | |
1095 | * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc. | |
1096 | */ | |
1097 | if (data_size > 0) { | |
1098 | vid_hdr->copy_flag = 1; | |
3261ebd7 CH |
1099 | vid_hdr->data_size = cpu_to_be32(data_size); |
1100 | vid_hdr->data_crc = cpu_to_be32(crc); | |
801c135c | 1101 | } |
a7306653 | 1102 | vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi)); |
801c135c AB |
1103 | |
1104 | err = ubi_io_write_vid_hdr(ubi, to, vid_hdr); | |
6fa6f5bb AB |
1105 | if (err) { |
1106 | if (err == -EIO) | |
90bf0265 | 1107 | err = MOVE_TARGET_WR_ERR; |
43f9b25a | 1108 | goto out_unlock_buf; |
6fa6f5bb | 1109 | } |
801c135c AB |
1110 | |
1111 | cond_resched(); | |
1112 | ||
1113 | /* Read the VID header back and check if it was written correctly */ | |
1114 | err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1); | |
1115 | if (err) { | |
b86a2c56 | 1116 | if (err != UBI_IO_BITFLIPS) { |
049333ce AB |
1117 | ubi_warn("error %d while reading VID header back from PEB %d", |
1118 | err, to); | |
6b5c94c6 | 1119 | if (is_error_sane(err)) |
b86a2c56 AB |
1120 | err = MOVE_TARGET_RD_ERR; |
1121 | } else | |
cc831464 | 1122 | err = MOVE_TARGET_BITFLIPS; |
43f9b25a | 1123 | goto out_unlock_buf; |
801c135c AB |
1124 | } |
1125 | ||
1126 | if (data_size > 0) { | |
0ca39d74 | 1127 | err = ubi_io_write_data(ubi, ubi->peb_buf, to, 0, aldata_size); |
6fa6f5bb AB |
1128 | if (err) { |
1129 | if (err == -EIO) | |
90bf0265 | 1130 | err = MOVE_TARGET_WR_ERR; |
43f9b25a | 1131 | goto out_unlock_buf; |
6fa6f5bb | 1132 | } |
801c135c | 1133 | |
e88d6e10 AB |
1134 | cond_resched(); |
1135 | ||
801c135c AB |
1136 | /* |
1137 | * We've written the data and are going to read it back to make | |
1138 | * sure it was written correctly. | |
1139 | */ | |
0ca39d74 AB |
1140 | memset(ubi->peb_buf, 0xFF, aldata_size); |
1141 | err = ubi_io_read_data(ubi, ubi->peb_buf, to, 0, aldata_size); | |
801c135c | 1142 | if (err) { |
b86a2c56 | 1143 | if (err != UBI_IO_BITFLIPS) { |
049333ce AB |
1144 | ubi_warn("error %d while reading data back from PEB %d", |
1145 | err, to); | |
6b5c94c6 | 1146 | if (is_error_sane(err)) |
b86a2c56 AB |
1147 | err = MOVE_TARGET_RD_ERR; |
1148 | } else | |
cc831464 | 1149 | err = MOVE_TARGET_BITFLIPS; |
43f9b25a | 1150 | goto out_unlock_buf; |
801c135c AB |
1151 | } |
1152 | ||
1153 | cond_resched(); | |
1154 | ||
0ca39d74 | 1155 | if (crc != crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size)) { |
049333ce AB |
1156 | ubi_warn("read data back from PEB %d and it is different", |
1157 | to); | |
6fa6f5bb | 1158 | err = -EINVAL; |
43f9b25a | 1159 | goto out_unlock_buf; |
801c135c AB |
1160 | } |
1161 | } | |
1162 | ||
1163 | ubi_assert(vol->eba_tbl[lnum] == from); | |
8974b15c | 1164 | down_read(&ubi->fm_sem); |
801c135c | 1165 | vol->eba_tbl[lnum] = to; |
8974b15c | 1166 | up_read(&ubi->fm_sem); |
801c135c | 1167 | |
43f9b25a | 1168 | out_unlock_buf: |
e88d6e10 | 1169 | mutex_unlock(&ubi->buf_mutex); |
43f9b25a | 1170 | out_unlock_leb: |
801c135c | 1171 | leb_write_unlock(ubi, vol_id, lnum); |
801c135c AB |
1172 | return err; |
1173 | } | |
1174 | ||
64d4b4c9 AB |
1175 | /** |
1176 | * print_rsvd_warning - warn about not having enough reserved PEBs. | |
1177 | * @ubi: UBI device description object | |
1178 | * | |
41e0cd9d | 1179 | * This is a helper function for 'ubi_eba_init()' which is called when UBI |
64d4b4c9 AB |
1180 | * cannot reserve enough PEBs for bad block handling. This function makes a |
1181 | * decision whether we have to print a warning or not. The algorithm is as | |
1182 | * follows: | |
1183 | * o if this is a new UBI image, then just print the warning | |
1184 | * o if this is an UBI image which has already been used for some time, print | |
1185 | * a warning only if we can reserve less than 10% of the expected amount of | |
1186 | * the reserved PEB. | |
1187 | * | |
1188 | * The idea is that when UBI is used, PEBs become bad, and the reserved pool | |
1189 | * of PEBs becomes smaller, which is normal and we do not want to scare users | |
1190 | * with a warning every time they attach the MTD device. This was an issue | |
1191 | * reported by real users. | |
1192 | */ | |
1193 | static void print_rsvd_warning(struct ubi_device *ubi, | |
a4e6042f | 1194 | struct ubi_attach_info *ai) |
64d4b4c9 AB |
1195 | { |
1196 | /* | |
1197 | * The 1 << 18 (256KiB) number is picked randomly, just a reasonably | |
1198 | * large number to distinguish between newly flashed and used images. | |
1199 | */ | |
a4e6042f | 1200 | if (ai->max_sqnum > (1 << 18)) { |
64d4b4c9 AB |
1201 | int min = ubi->beb_rsvd_level / 10; |
1202 | ||
1203 | if (!min) | |
1204 | min = 1; | |
1205 | if (ubi->beb_rsvd_pebs > min) | |
1206 | return; | |
1207 | } | |
1208 | ||
049333ce AB |
1209 | ubi_warn("cannot reserve enough PEBs for bad PEB handling, reserved %d, need %d", |
1210 | ubi->beb_rsvd_pebs, ubi->beb_rsvd_level); | |
5fc01ab6 AB |
1211 | if (ubi->corr_peb_count) |
1212 | ubi_warn("%d PEBs are corrupted and not used", | |
049333ce | 1213 | ubi->corr_peb_count); |
64d4b4c9 AB |
1214 | } |
1215 | ||
00abf304 RW |
1216 | /** |
1217 | * self_check_eba - run a self check on the EBA table constructed by fastmap. | |
1218 | * @ubi: UBI device description object | |
1219 | * @ai_fastmap: UBI attach info object created by fastmap | |
1220 | * @ai_scan: UBI attach info object created by scanning | |
1221 | * | |
1222 | * Returns < 0 in case of an internal error, 0 otherwise. | |
1223 | * If a bad EBA table entry was found it will be printed out and | |
1224 | * ubi_assert() triggers. | |
1225 | */ | |
1226 | int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap, | |
1227 | struct ubi_attach_info *ai_scan) | |
1228 | { | |
1229 | int i, j, num_volumes, ret = 0; | |
1230 | int **scan_eba, **fm_eba; | |
1231 | struct ubi_ainf_volume *av; | |
1232 | struct ubi_volume *vol; | |
1233 | struct ubi_ainf_peb *aeb; | |
1234 | struct rb_node *rb; | |
1235 | ||
1236 | num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT; | |
1237 | ||
1238 | scan_eba = kmalloc(sizeof(*scan_eba) * num_volumes, GFP_KERNEL); | |
1239 | if (!scan_eba) | |
1240 | return -ENOMEM; | |
1241 | ||
1242 | fm_eba = kmalloc(sizeof(*fm_eba) * num_volumes, GFP_KERNEL); | |
1243 | if (!fm_eba) { | |
1244 | kfree(scan_eba); | |
1245 | return -ENOMEM; | |
1246 | } | |
1247 | ||
1248 | for (i = 0; i < num_volumes; i++) { | |
1249 | vol = ubi->volumes[i]; | |
1250 | if (!vol) | |
1251 | continue; | |
1252 | ||
1253 | scan_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**scan_eba), | |
1254 | GFP_KERNEL); | |
1255 | if (!scan_eba[i]) { | |
1256 | ret = -ENOMEM; | |
1257 | goto out_free; | |
1258 | } | |
1259 | ||
1260 | fm_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**fm_eba), | |
1261 | GFP_KERNEL); | |
1262 | if (!fm_eba[i]) { | |
1263 | ret = -ENOMEM; | |
1264 | goto out_free; | |
1265 | } | |
1266 | ||
1267 | for (j = 0; j < vol->reserved_pebs; j++) | |
1268 | scan_eba[i][j] = fm_eba[i][j] = UBI_LEB_UNMAPPED; | |
1269 | ||
1270 | av = ubi_find_av(ai_scan, idx2vol_id(ubi, i)); | |
1271 | if (!av) | |
1272 | continue; | |
1273 | ||
1274 | ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) | |
1275 | scan_eba[i][aeb->lnum] = aeb->pnum; | |
1276 | ||
1277 | av = ubi_find_av(ai_fastmap, idx2vol_id(ubi, i)); | |
1278 | if (!av) | |
1279 | continue; | |
1280 | ||
1281 | ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) | |
1282 | fm_eba[i][aeb->lnum] = aeb->pnum; | |
1283 | ||
1284 | for (j = 0; j < vol->reserved_pebs; j++) { | |
1285 | if (scan_eba[i][j] != fm_eba[i][j]) { | |
1286 | if (scan_eba[i][j] == UBI_LEB_UNMAPPED || | |
1287 | fm_eba[i][j] == UBI_LEB_UNMAPPED) | |
1288 | continue; | |
1289 | ||
1290 | ubi_err("LEB:%i:%i is PEB:%i instead of %i!", | |
1291 | vol->vol_id, i, fm_eba[i][j], | |
1292 | scan_eba[i][j]); | |
1293 | ubi_assert(0); | |
1294 | } | |
1295 | } | |
1296 | } | |
1297 | ||
1298 | out_free: | |
1299 | for (i = 0; i < num_volumes; i++) { | |
1300 | if (!ubi->volumes[i]) | |
1301 | continue; | |
1302 | ||
1303 | kfree(scan_eba[i]); | |
1304 | kfree(fm_eba[i]); | |
1305 | } | |
1306 | ||
1307 | kfree(scan_eba); | |
1308 | kfree(fm_eba); | |
1309 | return ret; | |
1310 | } | |
1311 | ||
801c135c | 1312 | /** |
41e0cd9d | 1313 | * ubi_eba_init - initialize the EBA sub-system using attaching information. |
801c135c | 1314 | * @ubi: UBI device description object |
a4e6042f | 1315 | * @ai: attaching information |
801c135c AB |
1316 | * |
1317 | * This function returns zero in case of success and a negative error code in | |
1318 | * case of failure. | |
1319 | */ | |
41e0cd9d | 1320 | int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai) |
801c135c AB |
1321 | { |
1322 | int i, j, err, num_volumes; | |
517af48c | 1323 | struct ubi_ainf_volume *av; |
801c135c | 1324 | struct ubi_volume *vol; |
2c5ec5ce | 1325 | struct ubi_ainf_peb *aeb; |
801c135c AB |
1326 | struct rb_node *rb; |
1327 | ||
85c6e6e2 | 1328 | dbg_eba("initialize EBA sub-system"); |
801c135c AB |
1329 | |
1330 | spin_lock_init(&ubi->ltree_lock); | |
e8823bd6 | 1331 | mutex_init(&ubi->alc_mutex); |
801c135c AB |
1332 | ubi->ltree = RB_ROOT; |
1333 | ||
a4e6042f | 1334 | ubi->global_sqnum = ai->max_sqnum + 1; |
801c135c AB |
1335 | num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT; |
1336 | ||
1337 | for (i = 0; i < num_volumes; i++) { | |
1338 | vol = ubi->volumes[i]; | |
1339 | if (!vol) | |
1340 | continue; | |
1341 | ||
1342 | cond_resched(); | |
1343 | ||
1344 | vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int), | |
1345 | GFP_KERNEL); | |
1346 | if (!vol->eba_tbl) { | |
1347 | err = -ENOMEM; | |
1348 | goto out_free; | |
1349 | } | |
1350 | ||
1351 | for (j = 0; j < vol->reserved_pebs; j++) | |
1352 | vol->eba_tbl[j] = UBI_LEB_UNMAPPED; | |
1353 | ||
dcd85fdd | 1354 | av = ubi_find_av(ai, idx2vol_id(ubi, i)); |
517af48c | 1355 | if (!av) |
801c135c AB |
1356 | continue; |
1357 | ||
517af48c | 1358 | ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) { |
2c5ec5ce | 1359 | if (aeb->lnum >= vol->reserved_pebs) |
801c135c AB |
1360 | /* |
1361 | * This may happen in case of an unclean reboot | |
1362 | * during re-size. | |
1363 | */ | |
0bae2887 | 1364 | ubi_move_aeb_to_list(av, aeb, &ai->erase); |
2c5ec5ce | 1365 | vol->eba_tbl[aeb->lnum] = aeb->pnum; |
801c135c AB |
1366 | } |
1367 | } | |
1368 | ||
94780d4d AB |
1369 | if (ubi->avail_pebs < EBA_RESERVED_PEBS) { |
1370 | ubi_err("no enough physical eraseblocks (%d, need %d)", | |
1371 | ubi->avail_pebs, EBA_RESERVED_PEBS); | |
5fc01ab6 AB |
1372 | if (ubi->corr_peb_count) |
1373 | ubi_err("%d PEBs are corrupted and not used", | |
1374 | ubi->corr_peb_count); | |
94780d4d AB |
1375 | err = -ENOSPC; |
1376 | goto out_free; | |
1377 | } | |
1378 | ubi->avail_pebs -= EBA_RESERVED_PEBS; | |
1379 | ubi->rsvd_pebs += EBA_RESERVED_PEBS; | |
1380 | ||
801c135c AB |
1381 | if (ubi->bad_allowed) { |
1382 | ubi_calculate_reserved(ubi); | |
1383 | ||
1384 | if (ubi->avail_pebs < ubi->beb_rsvd_level) { | |
1385 | /* No enough free physical eraseblocks */ | |
1386 | ubi->beb_rsvd_pebs = ubi->avail_pebs; | |
a4e6042f | 1387 | print_rsvd_warning(ubi, ai); |
801c135c AB |
1388 | } else |
1389 | ubi->beb_rsvd_pebs = ubi->beb_rsvd_level; | |
1390 | ||
1391 | ubi->avail_pebs -= ubi->beb_rsvd_pebs; | |
1392 | ubi->rsvd_pebs += ubi->beb_rsvd_pebs; | |
1393 | } | |
1394 | ||
85c6e6e2 | 1395 | dbg_eba("EBA sub-system is initialized"); |
801c135c AB |
1396 | return 0; |
1397 | ||
1398 | out_free: | |
1399 | for (i = 0; i < num_volumes; i++) { | |
1400 | if (!ubi->volumes[i]) | |
1401 | continue; | |
1402 | kfree(ubi->volumes[i]->eba_tbl); | |
7194e6f9 | 1403 | ubi->volumes[i]->eba_tbl = NULL; |
801c135c | 1404 | } |
801c135c AB |
1405 | return err; |
1406 | } |