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 | */ | |
60 | static 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 | */ | |
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 | ||
343 | vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED; | |
344 | err = ubi_wl_put_peb(ubi, pnum, 0); | |
345 | ||
346 | out_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 | |
89b96b69 | 354 | * @vol: volume description object |
801c135c AB |
355 | * @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 | */ | |
89b96b69 AB |
370 | int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum, |
371 | void *buf, int offset, int len, int check) | |
801c135c | 372 | { |
89b96b69 | 373 | int err, pnum, scrub = 0, vol_id = vol->vol_id; |
801c135c | 374 | struct ubi_vid_hdr *vid_hdr; |
a6343afb | 375 | uint32_t uninitialized_var(crc); |
801c135c AB |
376 | |
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 | ||
402 | retry: | |
403 | if (check) { | |
33818bbb | 404 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
405 | 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 | */ | |
eb89580e AB |
421 | if (err == UBI_IO_BAD_HDR_READ || |
422 | err == UBI_IO_BAD_HDR) { | |
b86a2c56 AB |
423 | ubi_warn("corrupted VID header at PEB " |
424 | "%d, LEB %d:%d", pnum, vol_id, | |
425 | lnum); | |
801c135c AB |
426 | 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 | ||
3261ebd7 CH |
434 | ubi_assert(lnum < be32_to_cpu(vid_hdr->used_ebs)); |
435 | ubi_assert(len == be32_to_cpu(vid_hdr->data_size)); | |
801c135c | 436 | |
3261ebd7 | 437 | crc = be32_to_cpu(vid_hdr->data_crc); |
801c135c AB |
438 | 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) { | |
2ab934b8 | 460 | uint32_t crc1 = crc32(UBI_CRC32_INIT, buf, len); |
801c135c AB |
461 | 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 | ||
475 | out_free: | |
476 | ubi_free_vid_hdr(ubi, vid_hdr); | |
477 | out_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 | */ | |
498 | static 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; | |
801c135c | 504 | |
33818bbb | 505 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
9c9ec147 | 506 | if (!vid_hdr) |
801c135c | 507 | return -ENOMEM; |
801c135c AB |
508 | |
509 | retry: | |
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 | ||
3261ebd7 | 525 | vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi)); |
801c135c AB |
526 | err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr); |
527 | if (err) | |
528 | goto write_error; | |
529 | ||
530 | data_size = offset + len; | |
4df581f3 | 531 | mutex_lock(&ubi->buf_mutex); |
e88d6e10 | 532 | memset(ubi->peb_buf1 + offset, 0xFF, len); |
801c135c AB |
533 | |
534 | /* Read everything before the area where the write failure happened */ | |
535 | if (offset > 0) { | |
e88d6e10 AB |
536 | err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset); |
537 | if (err && err != UBI_IO_BITFLIPS) | |
4df581f3 | 538 | goto out_unlock; |
801c135c AB |
539 | } |
540 | ||
e88d6e10 | 541 | memcpy(ubi->peb_buf1 + offset, buf, len); |
801c135c | 542 | |
e88d6e10 | 543 | err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size); |
4df581f3 AB |
544 | if (err) { |
545 | mutex_unlock(&ubi->buf_mutex); | |
801c135c | 546 | goto write_error; |
4df581f3 | 547 | } |
801c135c | 548 | |
e88d6e10 | 549 | mutex_unlock(&ubi->buf_mutex); |
801c135c AB |
550 | 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 | ||
4df581f3 | 558 | out_unlock: |
e88d6e10 | 559 | mutex_unlock(&ubi->buf_mutex); |
4df581f3 | 560 | out_put: |
801c135c AB |
561 | ubi_wl_put_peb(ubi, new_pnum, 1); |
562 | ubi_free_vid_hdr(ubi, vid_hdr); | |
563 | return err; | |
564 | ||
565 | write_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 | |
89b96b69 | 583 | * @vol: volume description object |
801c135c AB |
584 | * @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 | |
89b96b69 | 591 | * @vol. Returns zero in case of success and a negative error code in case |
801c135c AB |
592 | * 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 | */ | |
89b96b69 | 595 | int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum, |
801c135c AB |
596 | const void *buf, int offset, int len, int dtype) |
597 | { | |
89b96b69 | 598 | int err, pnum, tries = 0, vol_id = vol->vol_id; |
801c135c AB |
599 | 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) | |
89b96b69 AB |
617 | err = recover_peb(ubi, pnum, vol_id, lnum, buf, |
618 | offset, len); | |
801c135c AB |
619 | 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 | */ | |
33818bbb | 630 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
631 | if (!vid_hdr) { |
632 | leb_write_unlock(ubi, vol_id, lnum); | |
633 | return -ENOMEM; | |
634 | } | |
635 | ||
636 | vid_hdr->vol_type = UBI_VID_DYNAMIC; | |
3261ebd7 CH |
637 | 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); | |
801c135c | 640 | vid_hdr->compat = ubi_get_compat(ubi, vol_id); |
3261ebd7 | 641 | vid_hdr->data_pad = cpu_to_be32(vol->data_pad); |
801c135c AB |
642 | |
643 | retry: | |
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 | ||
393852ec AB |
661 | 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 | } | |
801c135c AB |
669 | } |
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 | ||
677 | write_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 | ||
3261ebd7 | 698 | vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi)); |
801c135c AB |
699 | 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 | |
89b96b69 | 706 | * @vol: volume description object |
801c135c AB |
707 | * @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 | |
89b96b69 | 714 | * @vol. The @used_ebs argument should contain total number of logical |
801c135c AB |
715 | * 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 | * | |
025dfdaf | 722 | * It is prohibited to write more than once to logical eraseblocks of static |
801c135c AB |
723 | * volumes. This function returns zero in case of success and a negative error |
724 | * code in case of failure. | |
725 | */ | |
89b96b69 AB |
726 | int 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) | |
801c135c | 729 | { |
89b96b69 | 730 | int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id; |
801c135c AB |
731 | 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 | |
cadb40cc | 741 | ubi_assert(!(len & (ubi->min_io_size - 1))); |
801c135c | 742 | |
33818bbb | 743 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
744 | 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 | ||
3261ebd7 CH |
753 | 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); | |
801c135c | 756 | vid_hdr->compat = ubi_get_compat(ubi, vol_id); |
3261ebd7 | 757 | vid_hdr->data_pad = cpu_to_be32(vol->data_pad); |
801c135c AB |
758 | |
759 | crc = crc32(UBI_CRC32_INIT, buf, data_size); | |
760 | vid_hdr->vol_type = UBI_VID_STATIC; | |
3261ebd7 CH |
761 | 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); | |
801c135c AB |
764 | |
765 | retry: | |
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 | ||
797 | write_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 | ||
3261ebd7 | 818 | vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi)); |
801c135c AB |
819 | 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 | |
c63a491d | 826 | * @vol: volume description object |
801c135c AB |
827 | * @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. | |
e8823bd6 AB |
837 | * |
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. | |
801c135c | 840 | */ |
89b96b69 AB |
841 | int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol, |
842 | int lnum, const void *buf, int len, int dtype) | |
801c135c | 843 | { |
89b96b69 | 844 | int err, pnum, tries = 0, vol_id = vol->vol_id; |
801c135c AB |
845 | struct ubi_vid_hdr *vid_hdr; |
846 | uint32_t crc; | |
847 | ||
848 | if (ubi->ro_mode) | |
849 | return -EROFS; | |
850 | ||
60c03153 AB |
851 | 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 | ||
33818bbb | 862 | vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS); |
801c135c AB |
863 | if (!vid_hdr) |
864 | return -ENOMEM; | |
865 | ||
e8823bd6 | 866 | mutex_lock(&ubi->alc_mutex); |
801c135c | 867 | err = leb_write_lock(ubi, vol_id, lnum); |
e8823bd6 AB |
868 | if (err) |
869 | goto out_mutex; | |
801c135c | 870 | |
3261ebd7 CH |
871 | 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); | |
801c135c | 874 | vid_hdr->compat = ubi_get_compat(ubi, vol_id); |
3261ebd7 | 875 | vid_hdr->data_pad = cpu_to_be32(vol->data_pad); |
801c135c AB |
876 | |
877 | crc = crc32(UBI_CRC32_INIT, buf, len); | |
84a92580 | 878 | vid_hdr->vol_type = UBI_VID_DYNAMIC; |
3261ebd7 | 879 | vid_hdr->data_size = cpu_to_be32(len); |
801c135c | 880 | vid_hdr->copy_flag = 1; |
3261ebd7 | 881 | vid_hdr->data_crc = cpu_to_be32(crc); |
801c135c AB |
882 | |
883 | retry: | |
884 | pnum = ubi_wl_get_peb(ubi, dtype); | |
885 | if (pnum < 0) { | |
e8823bd6 AB |
886 | err = pnum; |
887 | goto out_leb_unlock; | |
801c135c AB |
888 | } |
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 | ||
a443db48 | 907 | if (vol->eba_tbl[lnum] >= 0) { |
4d88de4b | 908 | err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 0); |
e8823bd6 AB |
909 | if (err) |
910 | goto out_leb_unlock; | |
801c135c AB |
911 | } |
912 | ||
913 | vol->eba_tbl[lnum] = pnum; | |
e8823bd6 AB |
914 | |
915 | out_leb_unlock: | |
801c135c | 916 | leb_write_unlock(ubi, vol_id, lnum); |
e8823bd6 AB |
917 | out_mutex: |
918 | mutex_unlock(&ubi->alc_mutex); | |
801c135c | 919 | ubi_free_vid_hdr(ubi, vid_hdr); |
e8823bd6 | 920 | return err; |
801c135c AB |
921 | |
922 | write_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); | |
e8823bd6 | 930 | goto out_leb_unlock; |
801c135c AB |
931 | } |
932 | ||
933 | err = ubi_wl_put_peb(ubi, pnum, 1); | |
934 | if (err || ++tries > UBI_IO_RETRIES) { | |
935 | ubi_ro_mode(ubi); | |
e8823bd6 | 936 | goto out_leb_unlock; |
801c135c AB |
937 | } |
938 | ||
3261ebd7 | 939 | vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi)); |
801c135c AB |
940 | ubi_msg("try another PEB"); |
941 | goto retry; | |
942 | } | |
943 | ||
6b5c94c6 AB |
944 | /** |
945 | * 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 | */ | |
963 | static int is_error_sane(int err) | |
964 | { | |
786d7831 | 965 | if (err == -EIO || err == -ENOMEM || err == UBI_IO_BAD_HDR || |
eb89580e | 966 | err == UBI_IO_BAD_HDR_READ || err == -ETIMEDOUT) |
6b5c94c6 AB |
967 | return 0; |
968 | return 1; | |
969 | } | |
970 | ||
801c135c AB |
971 | /** |
972 | * 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 | |
43f9b25a | 980 | * function. Returns: |
6fa6f5bb | 981 | * o %0 in case of success; |
815bc5f8 | 982 | * o %MOVE_CANCEL_RACE, %MOVE_TARGET_WR_ERR, %MOVE_CANCEL_BITFLIPS, etc; |
6fa6f5bb | 983 | * o a negative error code in case of failure. |
801c135c AB |
984 | */ |
985 | int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to, | |
986 | struct ubi_vid_hdr *vid_hdr) | |
987 | { | |
43f9b25a | 988 | int err, vol_id, lnum, data_size, aldata_size, idx; |
801c135c AB |
989 | struct ubi_volume *vol; |
990 | uint32_t crc; | |
801c135c | 991 | |
3261ebd7 CH |
992 | vol_id = be32_to_cpu(vid_hdr->vol_id); |
993 | lnum = be32_to_cpu(vid_hdr->lnum); | |
801c135c | 994 | |
87960c0b | 995 | dbg_wl("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to); |
801c135c AB |
996 | |
997 | if (vid_hdr->vol_type == UBI_VID_STATIC) { | |
3261ebd7 | 998 | data_size = be32_to_cpu(vid_hdr->data_size); |
801c135c AB |
999 | aldata_size = ALIGN(data_size, ubi->min_io_size); |
1000 | } else | |
1001 | data_size = aldata_size = | |
3261ebd7 | 1002 | ubi->leb_size - be32_to_cpu(vid_hdr->data_pad); |
801c135c | 1003 | |
801c135c | 1004 | idx = vol_id2idx(ubi, vol_id); |
43f9b25a | 1005 | spin_lock(&ubi->volumes_lock); |
801c135c | 1006 | /* |
43f9b25a AB |
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 | |
6fa6f5bb | 1009 | * volume deletion un-maps all the volume's logical eraseblocks, it will |
43f9b25a | 1010 | * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish. |
801c135c | 1011 | */ |
801c135c | 1012 | vol = ubi->volumes[idx]; |
90bf0265 | 1013 | spin_unlock(&ubi->volumes_lock); |
801c135c | 1014 | if (!vol) { |
43f9b25a | 1015 | /* No need to do further work, cancel */ |
87960c0b | 1016 | dbg_wl("volume %d is being removed, cancel", vol_id); |
90bf0265 | 1017 | return MOVE_CANCEL_RACE; |
801c135c AB |
1018 | } |
1019 | ||
43f9b25a AB |
1020 | /* |
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 | |
90bf0265 AB |
1030 | * 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. | |
43f9b25a AB |
1032 | */ |
1033 | err = leb_write_trylock(ubi, vol_id, lnum); | |
1034 | if (err) { | |
87960c0b | 1035 | dbg_wl("contention on LEB %d:%d, cancel", vol_id, lnum); |
90bf0265 | 1036 | return MOVE_CANCEL_RACE; |
801c135c | 1037 | } |
801c135c | 1038 | |
43f9b25a AB |
1039 | /* |
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) { | |
87960c0b AB |
1045 | 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]); | |
90bf0265 | 1048 | err = MOVE_CANCEL_RACE; |
43f9b25a AB |
1049 | goto out_unlock_leb; |
1050 | } | |
801c135c | 1051 | |
43f9b25a | 1052 | /* |
b77bcb07 | 1053 | * OK, now the LEB is locked and we can safely start moving it. Since |
90bf0265 AB |
1054 | * this function utilizes the @ubi->peb_buf1 buffer which is shared |
1055 | * with some other functions - we lock the buffer by taking the | |
43f9b25a AB |
1056 | * @ubi->buf_mutex. |
1057 | */ | |
1058 | mutex_lock(&ubi->buf_mutex); | |
87960c0b | 1059 | dbg_wl("read %d bytes of data", aldata_size); |
e88d6e10 | 1060 | err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size); |
801c135c AB |
1061 | if (err && err != UBI_IO_BITFLIPS) { |
1062 | ubi_warn("error %d while reading data from PEB %d", | |
1063 | err, from); | |
6b5c94c6 | 1064 | err = MOVE_SOURCE_RD_ERR; |
43f9b25a | 1065 | goto out_unlock_buf; |
801c135c AB |
1066 | } |
1067 | ||
1068 | /* | |
fd589a8f | 1069 | * Now we have got to calculate how much data we have to copy. In |
801c135c AB |
1070 | * 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 = | |
e88d6e10 | 1080 | ubi_calc_data_len(ubi, ubi->peb_buf1, data_size); |
801c135c AB |
1081 | |
1082 | cond_resched(); | |
e88d6e10 | 1083 | crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size); |
801c135c AB |
1084 | cond_resched(); |
1085 | ||
1086 | /* | |
90bf0265 | 1087 | * It may turn out to be that the whole @from physical eraseblock |
801c135c AB |
1088 | * 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; | |
3261ebd7 CH |
1094 | vid_hdr->data_size = cpu_to_be32(data_size); |
1095 | vid_hdr->data_crc = cpu_to_be32(crc); | |
801c135c | 1096 | } |
3261ebd7 | 1097 | vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi)); |
801c135c AB |
1098 | |
1099 | err = ubi_io_write_vid_hdr(ubi, to, vid_hdr); | |
6fa6f5bb AB |
1100 | if (err) { |
1101 | if (err == -EIO) | |
90bf0265 | 1102 | err = MOVE_TARGET_WR_ERR; |
43f9b25a | 1103 | goto out_unlock_buf; |
6fa6f5bb | 1104 | } |
801c135c AB |
1105 | |
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) { | |
b86a2c56 | 1111 | if (err != UBI_IO_BITFLIPS) { |
6b5c94c6 AB |
1112 | ubi_warn("error %d while reading VID header back from " |
1113 | "PEB %d", err, to); | |
1114 | if (is_error_sane(err)) | |
b86a2c56 AB |
1115 | err = MOVE_TARGET_RD_ERR; |
1116 | } else | |
90bf0265 | 1117 | err = MOVE_CANCEL_BITFLIPS; |
43f9b25a | 1118 | goto out_unlock_buf; |
801c135c AB |
1119 | } |
1120 | ||
1121 | if (data_size > 0) { | |
e88d6e10 | 1122 | err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size); |
6fa6f5bb AB |
1123 | if (err) { |
1124 | if (err == -EIO) | |
90bf0265 | 1125 | err = MOVE_TARGET_WR_ERR; |
43f9b25a | 1126 | goto out_unlock_buf; |
6fa6f5bb | 1127 | } |
801c135c | 1128 | |
e88d6e10 AB |
1129 | cond_resched(); |
1130 | ||
801c135c AB |
1131 | /* |
1132 | * We've written the data and are going to read it back to make | |
1133 | * sure it was written correctly. | |
1134 | */ | |
801c135c | 1135 | |
e88d6e10 | 1136 | err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size); |
801c135c | 1137 | if (err) { |
b86a2c56 | 1138 | if (err != UBI_IO_BITFLIPS) { |
6b5c94c6 AB |
1139 | ubi_warn("error %d while reading data back " |
1140 | "from PEB %d", err, to); | |
1141 | if (is_error_sane(err)) | |
b86a2c56 AB |
1142 | err = MOVE_TARGET_RD_ERR; |
1143 | } else | |
90bf0265 | 1144 | err = MOVE_CANCEL_BITFLIPS; |
43f9b25a | 1145 | goto out_unlock_buf; |
801c135c AB |
1146 | } |
1147 | ||
1148 | cond_resched(); | |
1149 | ||
e88d6e10 | 1150 | if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) { |
6fa6f5bb AB |
1151 | ubi_warn("read data back from PEB %d and it is " |
1152 | "different", to); | |
1153 | err = -EINVAL; | |
43f9b25a | 1154 | goto out_unlock_buf; |
801c135c AB |
1155 | } |
1156 | } | |
1157 | ||
1158 | ubi_assert(vol->eba_tbl[lnum] == from); | |
1159 | vol->eba_tbl[lnum] = to; | |
1160 | ||
43f9b25a | 1161 | out_unlock_buf: |
e88d6e10 | 1162 | mutex_unlock(&ubi->buf_mutex); |
43f9b25a | 1163 | out_unlock_leb: |
801c135c | 1164 | leb_write_unlock(ubi, vol_id, lnum); |
801c135c AB |
1165 | return err; |
1166 | } | |
1167 | ||
64d4b4c9 AB |
1168 | /** |
1169 | * 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 | */ | |
1186 | static 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); | |
1204 | } | |
1205 | ||
801c135c | 1206 | /** |
85c6e6e2 | 1207 | * ubi_eba_init_scan - initialize the EBA sub-system using scanning information. |
801c135c AB |
1208 | * @ubi: UBI device description object |
1209 | * @si: scanning information | |
1210 | * | |
1211 | * This function returns zero in case of success and a negative error code in | |
1212 | * case of failure. | |
1213 | */ | |
1214 | int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si) | |
1215 | { | |
1216 | int i, j, err, num_volumes; | |
1217 | struct ubi_scan_volume *sv; | |
1218 | struct ubi_volume *vol; | |
1219 | struct ubi_scan_leb *seb; | |
1220 | struct rb_node *rb; | |
1221 | ||
85c6e6e2 | 1222 | dbg_eba("initialize EBA sub-system"); |
801c135c AB |
1223 | |
1224 | spin_lock_init(&ubi->ltree_lock); | |
e8823bd6 | 1225 | mutex_init(&ubi->alc_mutex); |
801c135c AB |
1226 | ubi->ltree = RB_ROOT; |
1227 | ||
801c135c AB |
1228 | ubi->global_sqnum = si->max_sqnum + 1; |
1229 | num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT; | |
1230 | ||
1231 | for (i = 0; i < num_volumes; i++) { | |
1232 | vol = ubi->volumes[i]; | |
1233 | if (!vol) | |
1234 | continue; | |
1235 | ||
1236 | cond_resched(); | |
1237 | ||
1238 | vol->eba_tbl = kmalloc(vol->reserved_pebs * sizeof(int), | |
1239 | GFP_KERNEL); | |
1240 | if (!vol->eba_tbl) { | |
1241 | err = -ENOMEM; | |
1242 | goto out_free; | |
1243 | } | |
1244 | ||
1245 | for (j = 0; j < vol->reserved_pebs; j++) | |
1246 | vol->eba_tbl[j] = UBI_LEB_UNMAPPED; | |
1247 | ||
1248 | sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i)); | |
1249 | if (!sv) | |
1250 | continue; | |
1251 | ||
1252 | ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) { | |
1253 | if (seb->lnum >= vol->reserved_pebs) | |
1254 | /* | |
1255 | * This may happen in case of an unclean reboot | |
1256 | * during re-size. | |
1257 | */ | |
1258 | ubi_scan_move_to_list(sv, seb, &si->erase); | |
1259 | vol->eba_tbl[seb->lnum] = seb->pnum; | |
1260 | } | |
1261 | } | |
1262 | ||
94780d4d AB |
1263 | if (ubi->avail_pebs < EBA_RESERVED_PEBS) { |
1264 | ubi_err("no enough physical eraseblocks (%d, need %d)", | |
1265 | ubi->avail_pebs, EBA_RESERVED_PEBS); | |
1266 | err = -ENOSPC; | |
1267 | goto out_free; | |
1268 | } | |
1269 | ubi->avail_pebs -= EBA_RESERVED_PEBS; | |
1270 | ubi->rsvd_pebs += EBA_RESERVED_PEBS; | |
1271 | ||
801c135c AB |
1272 | if (ubi->bad_allowed) { |
1273 | ubi_calculate_reserved(ubi); | |
1274 | ||
1275 | if (ubi->avail_pebs < ubi->beb_rsvd_level) { | |
1276 | /* No enough free physical eraseblocks */ | |
1277 | ubi->beb_rsvd_pebs = ubi->avail_pebs; | |
64d4b4c9 | 1278 | print_rsvd_warning(ubi, si); |
801c135c AB |
1279 | } else |
1280 | ubi->beb_rsvd_pebs = ubi->beb_rsvd_level; | |
1281 | ||
1282 | ubi->avail_pebs -= ubi->beb_rsvd_pebs; | |
1283 | ubi->rsvd_pebs += ubi->beb_rsvd_pebs; | |
1284 | } | |
1285 | ||
85c6e6e2 | 1286 | dbg_eba("EBA sub-system is initialized"); |
801c135c AB |
1287 | return 0; |
1288 | ||
1289 | out_free: | |
1290 | for (i = 0; i < num_volumes; i++) { | |
1291 | if (!ubi->volumes[i]) | |
1292 | continue; | |
1293 | kfree(ubi->volumes[i]->eba_tbl); | |
7194e6f9 | 1294 | ubi->volumes[i]->eba_tbl = NULL; |
801c135c | 1295 | } |
801c135c AB |
1296 | return err; |
1297 | } |