]>
Commit | Line | Data |
---|---|---|
1e51764a AB |
1 | /* |
2 | * This file is part of UBIFS. | |
3 | * | |
4 | * Copyright (C) 2006-2008 Nokia Corporation. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify it | |
7 | * under the terms of the GNU General Public License version 2 as published by | |
8 | * the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along with | |
16 | * this program; if not, write to the Free Software Foundation, Inc., 51 | |
17 | * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
18 | * | |
19 | * Authors: Adrian Hunter | |
20 | * Artem Bityutskiy (Битюцкий Артём) | |
21 | */ | |
22 | ||
23 | /* | |
24 | * This file contains journal replay code. It runs when the file-system is being | |
25 | * mounted and requires no locking. | |
26 | * | |
27 | * The larger is the journal, the longer it takes to scan it, so the longer it | |
28 | * takes to mount UBIFS. This is why the journal has limited size which may be | |
29 | * changed depending on the system requirements. But a larger journal gives | |
30 | * faster I/O speed because it writes the index less frequently. So this is a | |
31 | * trade-off. Also, the journal is indexed by the in-memory index (TNC), so the | |
32 | * larger is the journal, the more memory its index may consume. | |
33 | */ | |
34 | ||
35 | #include "ubifs.h" | |
debf12d5 | 36 | #include <linux/list_sort.h> |
1e51764a | 37 | |
1e51764a | 38 | /** |
debf12d5 | 39 | * struct replay_entry - replay list entry. |
1e51764a AB |
40 | * @lnum: logical eraseblock number of the node |
41 | * @offs: node offset | |
42 | * @len: node length | |
074bcb9b | 43 | * @deletion: non-zero if this entry corresponds to a node deletion |
1e51764a | 44 | * @sqnum: node sequence number |
debf12d5 | 45 | * @list: links the replay list |
1e51764a AB |
46 | * @key: node key |
47 | * @nm: directory entry name | |
48 | * @old_size: truncation old size | |
49 | * @new_size: truncation new size | |
1e51764a | 50 | * |
debf12d5 AB |
51 | * The replay process first scans all buds and builds the replay list, then |
52 | * sorts the replay list in nodes sequence number order, and then inserts all | |
53 | * the replay entries to the TNC. | |
1e51764a AB |
54 | */ |
55 | struct replay_entry { | |
56 | int lnum; | |
57 | int offs; | |
58 | int len; | |
074bcb9b | 59 | unsigned int deletion:1; |
1e51764a | 60 | unsigned long long sqnum; |
debf12d5 | 61 | struct list_head list; |
1e51764a AB |
62 | union ubifs_key key; |
63 | union { | |
64 | struct qstr nm; | |
65 | struct { | |
66 | loff_t old_size; | |
67 | loff_t new_size; | |
68 | }; | |
1e51764a AB |
69 | }; |
70 | }; | |
71 | ||
72 | /** | |
73 | * struct bud_entry - entry in the list of buds to replay. | |
74 | * @list: next bud in the list | |
75 | * @bud: bud description object | |
1e51764a | 76 | * @sqnum: reference node sequence number |
af1dd412 AB |
77 | * @free: free bytes in the bud |
78 | * @dirty: dirty bytes in the bud | |
1e51764a AB |
79 | */ |
80 | struct bud_entry { | |
81 | struct list_head list; | |
82 | struct ubifs_bud *bud; | |
1e51764a | 83 | unsigned long long sqnum; |
af1dd412 AB |
84 | int free; |
85 | int dirty; | |
1e51764a AB |
86 | }; |
87 | ||
88 | /** | |
89 | * set_bud_lprops - set free and dirty space used by a bud. | |
90 | * @c: UBIFS file-system description object | |
074bcb9b AB |
91 | * @b: bud entry which describes the bud |
92 | * | |
93 | * This function makes sure the LEB properties of bud @b are set correctly | |
94 | * after the replay. Returns zero in case of success and a negative error code | |
95 | * in case of failure. | |
1e51764a | 96 | */ |
074bcb9b | 97 | static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b) |
1e51764a AB |
98 | { |
99 | const struct ubifs_lprops *lp; | |
100 | int err = 0, dirty; | |
101 | ||
102 | ubifs_get_lprops(c); | |
103 | ||
074bcb9b | 104 | lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum); |
1e51764a AB |
105 | if (IS_ERR(lp)) { |
106 | err = PTR_ERR(lp); | |
107 | goto out; | |
108 | } | |
109 | ||
110 | dirty = lp->dirty; | |
074bcb9b | 111 | if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) { |
1e51764a AB |
112 | /* |
113 | * The LEB was added to the journal with a starting offset of | |
114 | * zero which means the LEB must have been empty. The LEB | |
074bcb9b AB |
115 | * property values should be @lp->free == @c->leb_size and |
116 | * @lp->dirty == 0, but that is not the case. The reason is that | |
7a9c3e39 AB |
117 | * the LEB had been garbage collected before it became the bud, |
118 | * and there was not commit inbetween. The garbage collector | |
119 | * resets the free and dirty space without recording it | |
120 | * anywhere except lprops, so if there was no commit then | |
121 | * lprops does not have that information. | |
1e51764a AB |
122 | * |
123 | * We do not need to adjust free space because the scan has told | |
124 | * us the exact value which is recorded in the replay entry as | |
074bcb9b | 125 | * @b->free. |
1e51764a AB |
126 | * |
127 | * However we do need to subtract from the dirty space the | |
128 | * amount of space that the garbage collector reclaimed, which | |
129 | * is the whole LEB minus the amount of space that was free. | |
130 | */ | |
074bcb9b | 131 | dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum, |
1e51764a | 132 | lp->free, lp->dirty); |
074bcb9b | 133 | dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum, |
1e51764a AB |
134 | lp->free, lp->dirty); |
135 | dirty -= c->leb_size - lp->free; | |
136 | /* | |
137 | * If the replay order was perfect the dirty space would now be | |
7d4e9ccb | 138 | * zero. The order is not perfect because the journal heads |
6edbfafd | 139 | * race with each other. This is not a problem but is does mean |
1e51764a AB |
140 | * that the dirty space may temporarily exceed c->leb_size |
141 | * during the replay. | |
142 | */ | |
143 | if (dirty != 0) | |
144 | dbg_msg("LEB %d lp: %d free %d dirty " | |
074bcb9b AB |
145 | "replay: %d free %d dirty", b->bud->lnum, |
146 | lp->free, lp->dirty, b->free, b->dirty); | |
1e51764a | 147 | } |
074bcb9b | 148 | lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty, |
1e51764a AB |
149 | lp->flags | LPROPS_TAKEN, 0); |
150 | if (IS_ERR(lp)) { | |
151 | err = PTR_ERR(lp); | |
152 | goto out; | |
153 | } | |
52c6e6f9 AB |
154 | |
155 | /* Make sure the journal head points to the latest bud */ | |
074bcb9b AB |
156 | err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf, |
157 | b->bud->lnum, c->leb_size - b->free, | |
158 | UBI_SHORTTERM); | |
52c6e6f9 | 159 | |
1e51764a AB |
160 | out: |
161 | ubifs_release_lprops(c); | |
162 | return err; | |
163 | } | |
164 | ||
074bcb9b AB |
165 | /** |
166 | * set_buds_lprops - set free and dirty space for all replayed buds. | |
167 | * @c: UBIFS file-system description object | |
168 | * | |
169 | * This function sets LEB properties for all replayed buds. Returns zero in | |
170 | * case of success and a negative error code in case of failure. | |
171 | */ | |
172 | static int set_buds_lprops(struct ubifs_info *c) | |
173 | { | |
174 | struct bud_entry *b; | |
175 | int err; | |
176 | ||
177 | list_for_each_entry(b, &c->replay_buds, list) { | |
178 | err = set_bud_lprops(c, b); | |
179 | if (err) | |
180 | return err; | |
181 | } | |
182 | ||
183 | return 0; | |
184 | } | |
185 | ||
1e51764a AB |
186 | /** |
187 | * trun_remove_range - apply a replay entry for a truncation to the TNC. | |
188 | * @c: UBIFS file-system description object | |
189 | * @r: replay entry of truncation | |
190 | */ | |
191 | static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r) | |
192 | { | |
193 | unsigned min_blk, max_blk; | |
194 | union ubifs_key min_key, max_key; | |
195 | ino_t ino; | |
196 | ||
197 | min_blk = r->new_size / UBIFS_BLOCK_SIZE; | |
198 | if (r->new_size & (UBIFS_BLOCK_SIZE - 1)) | |
199 | min_blk += 1; | |
200 | ||
201 | max_blk = r->old_size / UBIFS_BLOCK_SIZE; | |
202 | if ((r->old_size & (UBIFS_BLOCK_SIZE - 1)) == 0) | |
203 | max_blk -= 1; | |
204 | ||
205 | ino = key_inum(c, &r->key); | |
206 | ||
207 | data_key_init(c, &min_key, ino, min_blk); | |
208 | data_key_init(c, &max_key, ino, max_blk); | |
209 | ||
210 | return ubifs_tnc_remove_range(c, &min_key, &max_key); | |
211 | } | |
212 | ||
213 | /** | |
214 | * apply_replay_entry - apply a replay entry to the TNC. | |
215 | * @c: UBIFS file-system description object | |
216 | * @r: replay entry to apply | |
217 | * | |
218 | * Apply a replay entry to the TNC. | |
219 | */ | |
220 | static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r) | |
221 | { | |
074bcb9b | 222 | int err; |
1e51764a | 223 | |
515315a1 AB |
224 | dbg_mntk(&r->key, "LEB %d:%d len %d deletion %d sqnum %llu key ", |
225 | r->lnum, r->offs, r->len, r->deletion, r->sqnum); | |
1e51764a AB |
226 | |
227 | /* Set c->replay_sqnum to help deal with dangling branches. */ | |
228 | c->replay_sqnum = r->sqnum; | |
229 | ||
074bcb9b AB |
230 | if (is_hash_key(c, &r->key)) { |
231 | if (r->deletion) | |
1e51764a AB |
232 | err = ubifs_tnc_remove_nm(c, &r->key, &r->nm); |
233 | else | |
234 | err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs, | |
235 | r->len, &r->nm); | |
236 | } else { | |
074bcb9b | 237 | if (r->deletion) |
1e51764a AB |
238 | switch (key_type(c, &r->key)) { |
239 | case UBIFS_INO_KEY: | |
240 | { | |
241 | ino_t inum = key_inum(c, &r->key); | |
242 | ||
243 | err = ubifs_tnc_remove_ino(c, inum); | |
244 | break; | |
245 | } | |
246 | case UBIFS_TRUN_KEY: | |
247 | err = trun_remove_range(c, r); | |
248 | break; | |
249 | default: | |
250 | err = ubifs_tnc_remove(c, &r->key); | |
251 | break; | |
252 | } | |
253 | else | |
254 | err = ubifs_tnc_add(c, &r->key, r->lnum, r->offs, | |
255 | r->len); | |
256 | if (err) | |
257 | return err; | |
258 | ||
259 | if (c->need_recovery) | |
074bcb9b | 260 | err = ubifs_recover_size_accum(c, &r->key, r->deletion, |
1e51764a AB |
261 | r->new_size); |
262 | } | |
263 | ||
264 | return err; | |
265 | } | |
266 | ||
267 | /** | |
debf12d5 AB |
268 | * replay_entries_cmp - compare 2 replay entries. |
269 | * @priv: UBIFS file-system description object | |
270 | * @a: first replay entry | |
271 | * @a: second replay entry | |
1e51764a | 272 | * |
debf12d5 AB |
273 | * This is a comparios function for 'list_sort()' which compares 2 replay |
274 | * entries @a and @b by comparing their sequence numer. Returns %1 if @a has | |
275 | * greater sequence number and %-1 otherwise. | |
1e51764a | 276 | */ |
debf12d5 AB |
277 | static int replay_entries_cmp(void *priv, struct list_head *a, |
278 | struct list_head *b) | |
1e51764a | 279 | { |
debf12d5 AB |
280 | struct replay_entry *ra, *rb; |
281 | ||
282 | cond_resched(); | |
283 | if (a == b) | |
284 | return 0; | |
285 | ||
286 | ra = list_entry(a, struct replay_entry, list); | |
287 | rb = list_entry(b, struct replay_entry, list); | |
288 | ubifs_assert(ra->sqnum != rb->sqnum); | |
289 | if (ra->sqnum > rb->sqnum) | |
290 | return 1; | |
291 | return -1; | |
1e51764a AB |
292 | } |
293 | ||
294 | /** | |
debf12d5 | 295 | * apply_replay_list - apply the replay list to the TNC. |
1e51764a AB |
296 | * @c: UBIFS file-system description object |
297 | * | |
debf12d5 AB |
298 | * Apply all entries in the replay list to the TNC. Returns zero in case of |
299 | * success and a negative error code in case of failure. | |
1e51764a | 300 | */ |
debf12d5 | 301 | static int apply_replay_list(struct ubifs_info *c) |
1e51764a | 302 | { |
debf12d5 AB |
303 | struct replay_entry *r; |
304 | int err; | |
1e51764a | 305 | |
debf12d5 | 306 | list_sort(c, &c->replay_list, &replay_entries_cmp); |
1e51764a | 307 | |
debf12d5 | 308 | list_for_each_entry(r, &c->replay_list, list) { |
1e51764a AB |
309 | cond_resched(); |
310 | ||
1e51764a AB |
311 | err = apply_replay_entry(c, r); |
312 | if (err) | |
313 | return err; | |
1e51764a | 314 | } |
debf12d5 | 315 | |
1e51764a AB |
316 | return 0; |
317 | } | |
318 | ||
319 | /** | |
debf12d5 AB |
320 | * destroy_replay_list - destroy the replay. |
321 | * @c: UBIFS file-system description object | |
322 | * | |
323 | * Destroy the replay list. | |
324 | */ | |
325 | static void destroy_replay_list(struct ubifs_info *c) | |
326 | { | |
327 | struct replay_entry *r, *tmp; | |
328 | ||
329 | list_for_each_entry_safe(r, tmp, &c->replay_list, list) { | |
330 | if (is_hash_key(c, &r->key)) | |
331 | kfree(r->nm.name); | |
332 | list_del(&r->list); | |
333 | kfree(r); | |
334 | } | |
335 | } | |
336 | ||
337 | /** | |
338 | * insert_node - insert a node to the replay list | |
1e51764a AB |
339 | * @c: UBIFS file-system description object |
340 | * @lnum: node logical eraseblock number | |
341 | * @offs: node offset | |
342 | * @len: node length | |
343 | * @key: node key | |
344 | * @sqnum: sequence number | |
345 | * @deletion: non-zero if this is a deletion | |
346 | * @used: number of bytes in use in a LEB | |
347 | * @old_size: truncation old size | |
348 | * @new_size: truncation new size | |
349 | * | |
debf12d5 AB |
350 | * This function inserts a scanned non-direntry node to the replay list. The |
351 | * replay list contains @struct replay_entry elements, and we sort this list in | |
352 | * sequence number order before applying it. The replay list is applied at the | |
353 | * very end of the replay process. Since the list is sorted in sequence number | |
354 | * order, the older modifications are applied first. This function returns zero | |
355 | * in case of success and a negative error code in case of failure. | |
1e51764a AB |
356 | */ |
357 | static int insert_node(struct ubifs_info *c, int lnum, int offs, int len, | |
358 | union ubifs_key *key, unsigned long long sqnum, | |
359 | int deletion, int *used, loff_t old_size, | |
360 | loff_t new_size) | |
361 | { | |
1e51764a AB |
362 | struct replay_entry *r; |
363 | ||
515315a1 | 364 | dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs); |
debf12d5 | 365 | |
1e51764a AB |
366 | if (key_inum(c, key) >= c->highest_inum) |
367 | c->highest_inum = key_inum(c, key); | |
368 | ||
1e51764a AB |
369 | r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL); |
370 | if (!r) | |
371 | return -ENOMEM; | |
372 | ||
373 | if (!deletion) | |
374 | *used += ALIGN(len, 8); | |
375 | r->lnum = lnum; | |
376 | r->offs = offs; | |
377 | r->len = len; | |
074bcb9b | 378 | r->deletion = !!deletion; |
1e51764a | 379 | r->sqnum = sqnum; |
074bcb9b | 380 | key_copy(c, key, &r->key); |
1e51764a AB |
381 | r->old_size = old_size; |
382 | r->new_size = new_size; | |
1e51764a | 383 | |
debf12d5 | 384 | list_add_tail(&r->list, &c->replay_list); |
1e51764a AB |
385 | return 0; |
386 | } | |
387 | ||
388 | /** | |
debf12d5 | 389 | * insert_dent - insert a directory entry node into the replay list. |
1e51764a AB |
390 | * @c: UBIFS file-system description object |
391 | * @lnum: node logical eraseblock number | |
392 | * @offs: node offset | |
393 | * @len: node length | |
394 | * @key: node key | |
395 | * @name: directory entry name | |
396 | * @nlen: directory entry name length | |
397 | * @sqnum: sequence number | |
398 | * @deletion: non-zero if this is a deletion | |
399 | * @used: number of bytes in use in a LEB | |
400 | * | |
debf12d5 AB |
401 | * This function inserts a scanned directory entry node or an extended |
402 | * attribute entry to the replay list. Returns zero in case of success and a | |
403 | * negative error code in case of failure. | |
1e51764a AB |
404 | */ |
405 | static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len, | |
406 | union ubifs_key *key, const char *name, int nlen, | |
407 | unsigned long long sqnum, int deletion, int *used) | |
408 | { | |
1e51764a AB |
409 | struct replay_entry *r; |
410 | char *nbuf; | |
411 | ||
515315a1 | 412 | dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs); |
1e51764a AB |
413 | if (key_inum(c, key) >= c->highest_inum) |
414 | c->highest_inum = key_inum(c, key); | |
415 | ||
1e51764a AB |
416 | r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL); |
417 | if (!r) | |
418 | return -ENOMEM; | |
debf12d5 | 419 | |
1e51764a AB |
420 | nbuf = kmalloc(nlen + 1, GFP_KERNEL); |
421 | if (!nbuf) { | |
422 | kfree(r); | |
423 | return -ENOMEM; | |
424 | } | |
425 | ||
426 | if (!deletion) | |
427 | *used += ALIGN(len, 8); | |
428 | r->lnum = lnum; | |
429 | r->offs = offs; | |
430 | r->len = len; | |
074bcb9b | 431 | r->deletion = !!deletion; |
1e51764a | 432 | r->sqnum = sqnum; |
074bcb9b | 433 | key_copy(c, key, &r->key); |
1e51764a AB |
434 | r->nm.len = nlen; |
435 | memcpy(nbuf, name, nlen); | |
436 | nbuf[nlen] = '\0'; | |
437 | r->nm.name = nbuf; | |
1e51764a | 438 | |
debf12d5 | 439 | list_add_tail(&r->list, &c->replay_list); |
1e51764a AB |
440 | return 0; |
441 | } | |
442 | ||
443 | /** | |
444 | * ubifs_validate_entry - validate directory or extended attribute entry node. | |
445 | * @c: UBIFS file-system description object | |
446 | * @dent: the node to validate | |
447 | * | |
448 | * This function validates directory or extended attribute entry node @dent. | |
449 | * Returns zero if the node is all right and a %-EINVAL if not. | |
450 | */ | |
451 | int ubifs_validate_entry(struct ubifs_info *c, | |
452 | const struct ubifs_dent_node *dent) | |
453 | { | |
454 | int key_type = key_type_flash(c, dent->key); | |
455 | int nlen = le16_to_cpu(dent->nlen); | |
456 | ||
457 | if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 || | |
458 | dent->type >= UBIFS_ITYPES_CNT || | |
459 | nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 || | |
460 | strnlen(dent->name, nlen) != nlen || | |
461 | le64_to_cpu(dent->inum) > MAX_INUM) { | |
462 | ubifs_err("bad %s node", key_type == UBIFS_DENT_KEY ? | |
463 | "directory entry" : "extended attribute entry"); | |
464 | return -EINVAL; | |
465 | } | |
466 | ||
467 | if (key_type != UBIFS_DENT_KEY && key_type != UBIFS_XENT_KEY) { | |
468 | ubifs_err("bad key type %d", key_type); | |
469 | return -EINVAL; | |
470 | } | |
471 | ||
472 | return 0; | |
473 | } | |
474 | ||
91c66083 AB |
475 | /** |
476 | * is_last_bud - check if the bud is the last in the journal head. | |
477 | * @c: UBIFS file-system description object | |
478 | * @bud: bud description object | |
479 | * | |
480 | * This function checks if bud @bud is the last bud in its journal head. This | |
481 | * information is then used by 'replay_bud()' to decide whether the bud can | |
482 | * have corruptions or not. Indeed, only last buds can be corrupted by power | |
483 | * cuts. Returns %1 if this is the last bud, and %0 if not. | |
484 | */ | |
485 | static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud) | |
486 | { | |
487 | struct ubifs_jhead *jh = &c->jheads[bud->jhead]; | |
488 | struct ubifs_bud *next; | |
489 | uint32_t data; | |
490 | int err; | |
491 | ||
492 | if (list_is_last(&bud->list, &jh->buds_list)) | |
493 | return 1; | |
494 | ||
495 | /* | |
496 | * The following is a quirk to make sure we work correctly with UBIFS | |
497 | * images used with older UBIFS. | |
498 | * | |
499 | * Normally, the last bud will be the last in the journal head's list | |
500 | * of bud. However, there is one exception if the UBIFS image belongs | |
501 | * to older UBIFS. This is fairly unlikely: one would need to use old | |
502 | * UBIFS, then have a power cut exactly at the right point, and then | |
503 | * try to mount this image with new UBIFS. | |
504 | * | |
505 | * The exception is: it is possible to have 2 buds A and B, A goes | |
506 | * before B, and B is the last, bud B is contains no data, and bud A is | |
507 | * corrupted at the end. The reason is that in older versions when the | |
508 | * journal code switched the next bud (from A to B), it first added a | |
509 | * log reference node for the new bud (B), and only after this it | |
510 | * synchronized the write-buffer of current bud (A). But later this was | |
511 | * changed and UBIFS started to always synchronize the write-buffer of | |
512 | * the bud (A) before writing the log reference for the new bud (B). | |
513 | * | |
514 | * But because older UBIFS always synchronized A's write-buffer before | |
515 | * writing to B, we can recognize this exceptional situation but | |
516 | * checking the contents of bud B - if it is empty, then A can be | |
517 | * treated as the last and we can recover it. | |
518 | * | |
519 | * TODO: remove this piece of code in a couple of years (today it is | |
520 | * 16.05.2011). | |
521 | */ | |
522 | next = list_entry(bud->list.next, struct ubifs_bud, list); | |
523 | if (!list_is_last(&next->list, &jh->buds_list)) | |
524 | return 0; | |
525 | ||
d304820a | 526 | err = ubifs_leb_read(c, next->lnum, (char *)&data, next->start, 4, 1); |
91c66083 AB |
527 | if (err) |
528 | return 0; | |
529 | ||
530 | return data == 0xFFFFFFFF; | |
531 | } | |
532 | ||
1e51764a AB |
533 | /** |
534 | * replay_bud - replay a bud logical eraseblock. | |
535 | * @c: UBIFS file-system description object | |
e76a4526 | 536 | * @b: bud entry which describes the bud |
1e51764a | 537 | * |
e76a4526 AB |
538 | * This function replays bud @bud, recovers it if needed, and adds all nodes |
539 | * from this bud to the replay list. Returns zero in case of success and a | |
540 | * negative error code in case of failure. | |
1e51764a | 541 | */ |
e76a4526 | 542 | static int replay_bud(struct ubifs_info *c, struct bud_entry *b) |
1e51764a | 543 | { |
91c66083 | 544 | int is_last = is_last_bud(c, b->bud); |
e76a4526 | 545 | int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start; |
1e51764a AB |
546 | struct ubifs_scan_leb *sleb; |
547 | struct ubifs_scan_node *snod; | |
1e51764a | 548 | |
91c66083 AB |
549 | dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d", |
550 | lnum, b->bud->jhead, offs, is_last); | |
e76a4526 | 551 | |
91c66083 AB |
552 | if (c->need_recovery && is_last) |
553 | /* | |
554 | * Recover only last LEBs in the journal heads, because power | |
555 | * cuts may cause corruptions only in these LEBs, because only | |
556 | * these LEBs could possibly be written to at the power cut | |
557 | * time. | |
558 | */ | |
efcfde54 | 559 | sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, b->bud->jhead); |
1e51764a | 560 | else |
348709ba | 561 | sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0); |
1e51764a AB |
562 | if (IS_ERR(sleb)) |
563 | return PTR_ERR(sleb); | |
564 | ||
565 | /* | |
566 | * The bud does not have to start from offset zero - the beginning of | |
567 | * the 'lnum' LEB may contain previously committed data. One of the | |
568 | * things we have to do in replay is to correctly update lprops with | |
569 | * newer information about this LEB. | |
570 | * | |
571 | * At this point lprops thinks that this LEB has 'c->leb_size - offs' | |
572 | * bytes of free space because it only contain information about | |
573 | * committed data. | |
574 | * | |
575 | * But we know that real amount of free space is 'c->leb_size - | |
576 | * sleb->endpt', and the space in the 'lnum' LEB between 'offs' and | |
577 | * 'sleb->endpt' is used by bud data. We have to correctly calculate | |
578 | * how much of these data are dirty and update lprops with this | |
579 | * information. | |
580 | * | |
581 | * The dirt in that LEB region is comprised of padding nodes, deletion | |
582 | * nodes, truncation nodes and nodes which are obsoleted by subsequent | |
583 | * nodes in this LEB. So instead of calculating clean space, we | |
584 | * calculate used space ('used' variable). | |
585 | */ | |
586 | ||
587 | list_for_each_entry(snod, &sleb->nodes, list) { | |
588 | int deletion = 0; | |
589 | ||
590 | cond_resched(); | |
591 | ||
592 | if (snod->sqnum >= SQNUM_WATERMARK) { | |
593 | ubifs_err("file system's life ended"); | |
594 | goto out_dump; | |
595 | } | |
596 | ||
597 | if (snod->sqnum > c->max_sqnum) | |
598 | c->max_sqnum = snod->sqnum; | |
599 | ||
600 | switch (snod->type) { | |
601 | case UBIFS_INO_NODE: | |
602 | { | |
603 | struct ubifs_ino_node *ino = snod->node; | |
604 | loff_t new_size = le64_to_cpu(ino->size); | |
605 | ||
606 | if (le32_to_cpu(ino->nlink) == 0) | |
607 | deletion = 1; | |
608 | err = insert_node(c, lnum, snod->offs, snod->len, | |
609 | &snod->key, snod->sqnum, deletion, | |
610 | &used, 0, new_size); | |
611 | break; | |
612 | } | |
613 | case UBIFS_DATA_NODE: | |
614 | { | |
615 | struct ubifs_data_node *dn = snod->node; | |
616 | loff_t new_size = le32_to_cpu(dn->size) + | |
617 | key_block(c, &snod->key) * | |
618 | UBIFS_BLOCK_SIZE; | |
619 | ||
620 | err = insert_node(c, lnum, snod->offs, snod->len, | |
621 | &snod->key, snod->sqnum, deletion, | |
622 | &used, 0, new_size); | |
623 | break; | |
624 | } | |
625 | case UBIFS_DENT_NODE: | |
626 | case UBIFS_XENT_NODE: | |
627 | { | |
628 | struct ubifs_dent_node *dent = snod->node; | |
629 | ||
630 | err = ubifs_validate_entry(c, dent); | |
631 | if (err) | |
632 | goto out_dump; | |
633 | ||
634 | err = insert_dent(c, lnum, snod->offs, snod->len, | |
635 | &snod->key, dent->name, | |
636 | le16_to_cpu(dent->nlen), snod->sqnum, | |
637 | !le64_to_cpu(dent->inum), &used); | |
638 | break; | |
639 | } | |
640 | case UBIFS_TRUN_NODE: | |
641 | { | |
642 | struct ubifs_trun_node *trun = snod->node; | |
643 | loff_t old_size = le64_to_cpu(trun->old_size); | |
644 | loff_t new_size = le64_to_cpu(trun->new_size); | |
645 | union ubifs_key key; | |
646 | ||
647 | /* Validate truncation node */ | |
648 | if (old_size < 0 || old_size > c->max_inode_sz || | |
649 | new_size < 0 || new_size > c->max_inode_sz || | |
650 | old_size <= new_size) { | |
651 | ubifs_err("bad truncation node"); | |
652 | goto out_dump; | |
653 | } | |
654 | ||
655 | /* | |
656 | * Create a fake truncation key just to use the same | |
657 | * functions which expect nodes to have keys. | |
658 | */ | |
659 | trun_key_init(c, &key, le32_to_cpu(trun->inum)); | |
660 | err = insert_node(c, lnum, snod->offs, snod->len, | |
661 | &key, snod->sqnum, 1, &used, | |
662 | old_size, new_size); | |
663 | break; | |
664 | } | |
665 | default: | |
666 | ubifs_err("unexpected node type %d in bud LEB %d:%d", | |
667 | snod->type, lnum, snod->offs); | |
668 | err = -EINVAL; | |
669 | goto out_dump; | |
670 | } | |
671 | if (err) | |
672 | goto out; | |
673 | } | |
674 | ||
c49139d8 | 675 | ubifs_assert(ubifs_search_bud(c, lnum)); |
1e51764a AB |
676 | ubifs_assert(sleb->endpt - offs >= used); |
677 | ubifs_assert(sleb->endpt % c->min_io_size == 0); | |
678 | ||
e76a4526 AB |
679 | b->dirty = sleb->endpt - offs - used; |
680 | b->free = c->leb_size - sleb->endpt; | |
681 | dbg_mnt("bud LEB %d replied: dirty %d, free %d", lnum, b->dirty, b->free); | |
1e51764a AB |
682 | |
683 | out: | |
684 | ubifs_scan_destroy(sleb); | |
685 | return err; | |
686 | ||
687 | out_dump: | |
688 | ubifs_err("bad node is at LEB %d:%d", lnum, snod->offs); | |
689 | dbg_dump_node(c, snod->node); | |
690 | ubifs_scan_destroy(sleb); | |
691 | return -EINVAL; | |
692 | } | |
693 | ||
1e51764a AB |
694 | /** |
695 | * replay_buds - replay all buds. | |
696 | * @c: UBIFS file-system description object | |
697 | * | |
698 | * This function returns zero in case of success and a negative error code in | |
699 | * case of failure. | |
700 | */ | |
701 | static int replay_buds(struct ubifs_info *c) | |
702 | { | |
703 | struct bud_entry *b; | |
074bcb9b | 704 | int err; |
7703f09d | 705 | unsigned long long prev_sqnum = 0; |
1e51764a AB |
706 | |
707 | list_for_each_entry(b, &c->replay_buds, list) { | |
e76a4526 | 708 | err = replay_bud(c, b); |
1e51764a AB |
709 | if (err) |
710 | return err; | |
7703f09d AB |
711 | |
712 | ubifs_assert(b->sqnum > prev_sqnum); | |
713 | prev_sqnum = b->sqnum; | |
1e51764a AB |
714 | } |
715 | ||
716 | return 0; | |
717 | } | |
718 | ||
719 | /** | |
720 | * destroy_bud_list - destroy the list of buds to replay. | |
721 | * @c: UBIFS file-system description object | |
722 | */ | |
723 | static void destroy_bud_list(struct ubifs_info *c) | |
724 | { | |
725 | struct bud_entry *b; | |
726 | ||
727 | while (!list_empty(&c->replay_buds)) { | |
728 | b = list_entry(c->replay_buds.next, struct bud_entry, list); | |
729 | list_del(&b->list); | |
730 | kfree(b); | |
731 | } | |
732 | } | |
733 | ||
734 | /** | |
735 | * add_replay_bud - add a bud to the list of buds to replay. | |
736 | * @c: UBIFS file-system description object | |
737 | * @lnum: bud logical eraseblock number to replay | |
738 | * @offs: bud start offset | |
739 | * @jhead: journal head to which this bud belongs | |
740 | * @sqnum: reference node sequence number | |
741 | * | |
742 | * This function returns zero in case of success and a negative error code in | |
743 | * case of failure. | |
744 | */ | |
745 | static int add_replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead, | |
746 | unsigned long long sqnum) | |
747 | { | |
748 | struct ubifs_bud *bud; | |
749 | struct bud_entry *b; | |
750 | ||
751 | dbg_mnt("add replay bud LEB %d:%d, head %d", lnum, offs, jhead); | |
752 | ||
753 | bud = kmalloc(sizeof(struct ubifs_bud), GFP_KERNEL); | |
754 | if (!bud) | |
755 | return -ENOMEM; | |
756 | ||
757 | b = kmalloc(sizeof(struct bud_entry), GFP_KERNEL); | |
758 | if (!b) { | |
759 | kfree(bud); | |
760 | return -ENOMEM; | |
761 | } | |
762 | ||
763 | bud->lnum = lnum; | |
764 | bud->start = offs; | |
765 | bud->jhead = jhead; | |
766 | ubifs_add_bud(c, bud); | |
767 | ||
768 | b->bud = bud; | |
769 | b->sqnum = sqnum; | |
770 | list_add_tail(&b->list, &c->replay_buds); | |
771 | ||
772 | return 0; | |
773 | } | |
774 | ||
775 | /** | |
776 | * validate_ref - validate a reference node. | |
777 | * @c: UBIFS file-system description object | |
778 | * @ref: the reference node to validate | |
779 | * @ref_lnum: LEB number of the reference node | |
780 | * @ref_offs: reference node offset | |
781 | * | |
782 | * This function returns %1 if a bud reference already exists for the LEB. %0 is | |
783 | * returned if the reference node is new, otherwise %-EINVAL is returned if | |
784 | * validation failed. | |
785 | */ | |
786 | static int validate_ref(struct ubifs_info *c, const struct ubifs_ref_node *ref) | |
787 | { | |
788 | struct ubifs_bud *bud; | |
789 | int lnum = le32_to_cpu(ref->lnum); | |
790 | unsigned int offs = le32_to_cpu(ref->offs); | |
791 | unsigned int jhead = le32_to_cpu(ref->jhead); | |
792 | ||
793 | /* | |
794 | * ref->offs may point to the end of LEB when the journal head points | |
795 | * to the end of LEB and we write reference node for it during commit. | |
796 | * So this is why we require 'offs > c->leb_size'. | |
797 | */ | |
798 | if (jhead >= c->jhead_cnt || lnum >= c->leb_cnt || | |
799 | lnum < c->main_first || offs > c->leb_size || | |
800 | offs & (c->min_io_size - 1)) | |
801 | return -EINVAL; | |
802 | ||
803 | /* Make sure we have not already looked at this bud */ | |
804 | bud = ubifs_search_bud(c, lnum); | |
805 | if (bud) { | |
806 | if (bud->jhead == jhead && bud->start <= offs) | |
807 | return 1; | |
808 | ubifs_err("bud at LEB %d:%d was already referred", lnum, offs); | |
809 | return -EINVAL; | |
810 | } | |
811 | ||
812 | return 0; | |
813 | } | |
814 | ||
815 | /** | |
816 | * replay_log_leb - replay a log logical eraseblock. | |
817 | * @c: UBIFS file-system description object | |
818 | * @lnum: log logical eraseblock to replay | |
819 | * @offs: offset to start replaying from | |
820 | * @sbuf: scan buffer | |
821 | * | |
822 | * This function replays a log LEB and returns zero in case of success, %1 if | |
823 | * this is the last LEB in the log, and a negative error code in case of | |
824 | * failure. | |
825 | */ | |
826 | static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf) | |
827 | { | |
828 | int err; | |
829 | struct ubifs_scan_leb *sleb; | |
830 | struct ubifs_scan_node *snod; | |
831 | const struct ubifs_cs_node *node; | |
832 | ||
833 | dbg_mnt("replay log LEB %d:%d", lnum, offs); | |
348709ba AB |
834 | sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery); |
835 | if (IS_ERR(sleb)) { | |
ed43f2f0 AB |
836 | if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery) |
837 | return PTR_ERR(sleb); | |
7d08ae3c AB |
838 | /* |
839 | * Note, the below function will recover this log LEB only if | |
840 | * it is the last, because unclean reboots can possibly corrupt | |
841 | * only the tail of the log. | |
842 | */ | |
ed43f2f0 | 843 | sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf); |
1e51764a AB |
844 | if (IS_ERR(sleb)) |
845 | return PTR_ERR(sleb); | |
846 | } | |
847 | ||
848 | if (sleb->nodes_cnt == 0) { | |
849 | err = 1; | |
850 | goto out; | |
851 | } | |
852 | ||
853 | node = sleb->buf; | |
1e51764a AB |
854 | snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list); |
855 | if (c->cs_sqnum == 0) { | |
856 | /* | |
857 | * This is the first log LEB we are looking at, make sure that | |
858 | * the first node is a commit start node. Also record its | |
859 | * sequence number so that UBIFS can determine where the log | |
860 | * ends, because all nodes which were have higher sequence | |
861 | * numbers. | |
862 | */ | |
863 | if (snod->type != UBIFS_CS_NODE) { | |
864 | dbg_err("first log node at LEB %d:%d is not CS node", | |
865 | lnum, offs); | |
866 | goto out_dump; | |
867 | } | |
868 | if (le64_to_cpu(node->cmt_no) != c->cmt_no) { | |
869 | dbg_err("first CS node at LEB %d:%d has wrong " | |
870 | "commit number %llu expected %llu", | |
871 | lnum, offs, | |
872 | (unsigned long long)le64_to_cpu(node->cmt_no), | |
873 | c->cmt_no); | |
874 | goto out_dump; | |
875 | } | |
876 | ||
877 | c->cs_sqnum = le64_to_cpu(node->ch.sqnum); | |
878 | dbg_mnt("commit start sqnum %llu", c->cs_sqnum); | |
879 | } | |
880 | ||
881 | if (snod->sqnum < c->cs_sqnum) { | |
882 | /* | |
883 | * This means that we reached end of log and now | |
884 | * look to the older log data, which was already | |
885 | * committed but the eraseblock was not erased (UBIFS | |
6edbfafd | 886 | * only un-maps it). So this basically means we have to |
1e51764a AB |
887 | * exit with "end of log" code. |
888 | */ | |
889 | err = 1; | |
890 | goto out; | |
891 | } | |
892 | ||
893 | /* Make sure the first node sits at offset zero of the LEB */ | |
894 | if (snod->offs != 0) { | |
895 | dbg_err("first node is not at zero offset"); | |
896 | goto out_dump; | |
897 | } | |
898 | ||
899 | list_for_each_entry(snod, &sleb->nodes, list) { | |
1e51764a AB |
900 | cond_resched(); |
901 | ||
902 | if (snod->sqnum >= SQNUM_WATERMARK) { | |
903 | ubifs_err("file system's life ended"); | |
904 | goto out_dump; | |
905 | } | |
906 | ||
907 | if (snod->sqnum < c->cs_sqnum) { | |
908 | dbg_err("bad sqnum %llu, commit sqnum %llu", | |
909 | snod->sqnum, c->cs_sqnum); | |
910 | goto out_dump; | |
911 | } | |
912 | ||
913 | if (snod->sqnum > c->max_sqnum) | |
914 | c->max_sqnum = snod->sqnum; | |
915 | ||
916 | switch (snod->type) { | |
917 | case UBIFS_REF_NODE: { | |
918 | const struct ubifs_ref_node *ref = snod->node; | |
919 | ||
920 | err = validate_ref(c, ref); | |
921 | if (err == 1) | |
922 | break; /* Already have this bud */ | |
923 | if (err) | |
924 | goto out_dump; | |
925 | ||
926 | err = add_replay_bud(c, le32_to_cpu(ref->lnum), | |
927 | le32_to_cpu(ref->offs), | |
928 | le32_to_cpu(ref->jhead), | |
929 | snod->sqnum); | |
930 | if (err) | |
931 | goto out; | |
932 | ||
933 | break; | |
934 | } | |
935 | case UBIFS_CS_NODE: | |
936 | /* Make sure it sits at the beginning of LEB */ | |
937 | if (snod->offs != 0) { | |
938 | ubifs_err("unexpected node in log"); | |
939 | goto out_dump; | |
940 | } | |
941 | break; | |
942 | default: | |
943 | ubifs_err("unexpected node in log"); | |
944 | goto out_dump; | |
945 | } | |
946 | } | |
947 | ||
948 | if (sleb->endpt || c->lhead_offs >= c->leb_size) { | |
949 | c->lhead_lnum = lnum; | |
950 | c->lhead_offs = sleb->endpt; | |
951 | } | |
952 | ||
953 | err = !sleb->endpt; | |
954 | out: | |
955 | ubifs_scan_destroy(sleb); | |
956 | return err; | |
957 | ||
958 | out_dump: | |
681947d2 | 959 | ubifs_err("log error detected while replaying the log at LEB %d:%d", |
1e51764a AB |
960 | lnum, offs + snod->offs); |
961 | dbg_dump_node(c, snod->node); | |
962 | ubifs_scan_destroy(sleb); | |
963 | return -EINVAL; | |
964 | } | |
965 | ||
966 | /** | |
967 | * take_ihead - update the status of the index head in lprops to 'taken'. | |
968 | * @c: UBIFS file-system description object | |
969 | * | |
970 | * This function returns the amount of free space in the index head LEB or a | |
971 | * negative error code. | |
972 | */ | |
973 | static int take_ihead(struct ubifs_info *c) | |
974 | { | |
975 | const struct ubifs_lprops *lp; | |
976 | int err, free; | |
977 | ||
978 | ubifs_get_lprops(c); | |
979 | ||
980 | lp = ubifs_lpt_lookup_dirty(c, c->ihead_lnum); | |
981 | if (IS_ERR(lp)) { | |
982 | err = PTR_ERR(lp); | |
983 | goto out; | |
984 | } | |
985 | ||
986 | free = lp->free; | |
987 | ||
988 | lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC, | |
989 | lp->flags | LPROPS_TAKEN, 0); | |
990 | if (IS_ERR(lp)) { | |
991 | err = PTR_ERR(lp); | |
992 | goto out; | |
993 | } | |
994 | ||
995 | err = free; | |
996 | out: | |
997 | ubifs_release_lprops(c); | |
998 | return err; | |
999 | } | |
1000 | ||
1001 | /** | |
1002 | * ubifs_replay_journal - replay journal. | |
1003 | * @c: UBIFS file-system description object | |
1004 | * | |
1005 | * This function scans the journal, replays and cleans it up. It makes sure all | |
1006 | * memory data structures related to uncommitted journal are built (dirty TNC | |
1007 | * tree, tree of buds, modified lprops, etc). | |
1008 | */ | |
1009 | int ubifs_replay_journal(struct ubifs_info *c) | |
1010 | { | |
1011 | int err, i, lnum, offs, free; | |
1e51764a AB |
1012 | |
1013 | BUILD_BUG_ON(UBIFS_TRUN_KEY > 5); | |
1014 | ||
1015 | /* Update the status of the index head in lprops to 'taken' */ | |
1016 | free = take_ihead(c); | |
1017 | if (free < 0) | |
1018 | return free; /* Error code */ | |
1019 | ||
1020 | if (c->ihead_offs != c->leb_size - free) { | |
1021 | ubifs_err("bad index head LEB %d:%d", c->ihead_lnum, | |
1022 | c->ihead_offs); | |
1023 | return -EINVAL; | |
1024 | } | |
1025 | ||
1e51764a | 1026 | dbg_mnt("start replaying the journal"); |
1e51764a | 1027 | c->replaying = 1; |
1e51764a AB |
1028 | lnum = c->ltail_lnum = c->lhead_lnum; |
1029 | offs = c->lhead_offs; | |
1030 | ||
1031 | for (i = 0; i < c->log_lebs; i++, lnum++) { | |
1032 | if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) { | |
1033 | /* | |
1034 | * The log is logically circular, we reached the last | |
1035 | * LEB, switch to the first one. | |
1036 | */ | |
1037 | lnum = UBIFS_LOG_LNUM; | |
1038 | offs = 0; | |
1039 | } | |
6599fcbd | 1040 | err = replay_log_leb(c, lnum, offs, c->sbuf); |
1e51764a AB |
1041 | if (err == 1) |
1042 | /* We hit the end of the log */ | |
1043 | break; | |
1044 | if (err) | |
1045 | goto out; | |
1046 | offs = 0; | |
1047 | } | |
1048 | ||
1049 | err = replay_buds(c); | |
1050 | if (err) | |
1051 | goto out; | |
1052 | ||
debf12d5 | 1053 | err = apply_replay_list(c); |
1e51764a AB |
1054 | if (err) |
1055 | goto out; | |
1056 | ||
074bcb9b AB |
1057 | err = set_buds_lprops(c); |
1058 | if (err) | |
1059 | goto out; | |
1060 | ||
6edbfafd | 1061 | /* |
b137545c AB |
1062 | * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable |
1063 | * to roughly estimate index growth. Things like @c->bi.min_idx_lebs | |
6edbfafd AB |
1064 | * depend on it. This means we have to initialize it to make sure |
1065 | * budgeting works properly. | |
1066 | */ | |
b137545c AB |
1067 | c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt); |
1068 | c->bi.uncommitted_idx *= c->max_idx_node_sz; | |
6edbfafd | 1069 | |
1e51764a AB |
1070 | ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery); |
1071 | dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, " | |
1072 | "highest_inum %lu", c->lhead_lnum, c->lhead_offs, c->max_sqnum, | |
e84461ad | 1073 | (unsigned long)c->highest_inum); |
1e51764a | 1074 | out: |
debf12d5 | 1075 | destroy_replay_list(c); |
1e51764a | 1076 | destroy_bud_list(c); |
1e51764a AB |
1077 | c->replaying = 0; |
1078 | return err; | |
1079 | } |