blob: eed0fcff8d731710dc7723d25b1adb10ecbe6805 [file] [log] [blame]
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001/*
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"
36
37/*
38 * Replay flags.
39 *
40 * REPLAY_DELETION: node was deleted
41 * REPLAY_REF: node is a reference node
42 */
43enum {
44 REPLAY_DELETION = 1,
45 REPLAY_REF = 2,
46};
47
48/**
49 * struct replay_entry - replay tree entry.
50 * @lnum: logical eraseblock number of the node
51 * @offs: node offset
52 * @len: node length
53 * @sqnum: node sequence number
54 * @flags: replay flags
55 * @rb: links the replay tree
56 * @key: node key
57 * @nm: directory entry name
58 * @old_size: truncation old size
59 * @new_size: truncation new size
60 * @free: amount of free space in a bud
61 * @dirty: amount of dirty space in a bud from padding and deletion nodes
62 *
63 * UBIFS journal replay must compare node sequence numbers, which means it must
64 * build a tree of node information to insert into the TNC.
65 */
66struct replay_entry {
67 int lnum;
68 int offs;
69 int len;
70 unsigned long long sqnum;
71 int flags;
72 struct rb_node rb;
73 union ubifs_key key;
74 union {
75 struct qstr nm;
76 struct {
77 loff_t old_size;
78 loff_t new_size;
79 };
80 struct {
81 int free;
82 int dirty;
83 };
84 };
85};
86
87/**
88 * struct bud_entry - entry in the list of buds to replay.
89 * @list: next bud in the list
90 * @bud: bud description object
91 * @free: free bytes in the bud
92 * @sqnum: reference node sequence number
93 */
94struct bud_entry {
95 struct list_head list;
96 struct ubifs_bud *bud;
97 int free;
98 unsigned long long sqnum;
99};
100
101/**
102 * set_bud_lprops - set free and dirty space used by a bud.
103 * @c: UBIFS file-system description object
104 * @r: replay entry of bud
105 */
106static int set_bud_lprops(struct ubifs_info *c, struct replay_entry *r)
107{
108 const struct ubifs_lprops *lp;
109 int err = 0, dirty;
110
111 ubifs_get_lprops(c);
112
113 lp = ubifs_lpt_lookup_dirty(c, r->lnum);
114 if (IS_ERR(lp)) {
115 err = PTR_ERR(lp);
116 goto out;
117 }
118
119 dirty = lp->dirty;
120 if (r->offs == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
121 /*
122 * The LEB was added to the journal with a starting offset of
123 * zero which means the LEB must have been empty. The LEB
124 * property values should be lp->free == c->leb_size and
125 * lp->dirty == 0, but that is not the case. The reason is that
126 * the LEB was garbage collected. The garbage collector resets
127 * the free and dirty space without recording it anywhere except
128 * lprops, so if there is not a commit then lprops does not have
129 * that information next time the file system is mounted.
130 *
131 * We do not need to adjust free space because the scan has told
132 * us the exact value which is recorded in the replay entry as
133 * r->free.
134 *
135 * However we do need to subtract from the dirty space the
136 * amount of space that the garbage collector reclaimed, which
137 * is the whole LEB minus the amount of space that was free.
138 */
139 dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", r->lnum,
140 lp->free, lp->dirty);
141 dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", r->lnum,
142 lp->free, lp->dirty);
143 dirty -= c->leb_size - lp->free;
144 /*
145 * If the replay order was perfect the dirty space would now be
Artem Bityutskiy7d4e9cc2009-03-20 19:11:12 +0200146 * zero. The order is not perfect because the journal heads
Artem Bityutskiy6edbfaf2008-12-30 20:06:49 +0200147 * race with each other. This is not a problem but is does mean
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300148 * that the dirty space may temporarily exceed c->leb_size
149 * during the replay.
150 */
151 if (dirty != 0)
152 dbg_msg("LEB %d lp: %d free %d dirty "
153 "replay: %d free %d dirty", r->lnum, lp->free,
154 lp->dirty, r->free, r->dirty);
155 }
156 lp = ubifs_change_lp(c, lp, r->free, dirty + r->dirty,
157 lp->flags | LPROPS_TAKEN, 0);
158 if (IS_ERR(lp)) {
159 err = PTR_ERR(lp);
160 goto out;
161 }
162out:
163 ubifs_release_lprops(c);
164 return err;
165}
166
167/**
168 * trun_remove_range - apply a replay entry for a truncation to the TNC.
169 * @c: UBIFS file-system description object
170 * @r: replay entry of truncation
171 */
172static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r)
173{
174 unsigned min_blk, max_blk;
175 union ubifs_key min_key, max_key;
176 ino_t ino;
177
178 min_blk = r->new_size / UBIFS_BLOCK_SIZE;
179 if (r->new_size & (UBIFS_BLOCK_SIZE - 1))
180 min_blk += 1;
181
182 max_blk = r->old_size / UBIFS_BLOCK_SIZE;
183 if ((r->old_size & (UBIFS_BLOCK_SIZE - 1)) == 0)
184 max_blk -= 1;
185
186 ino = key_inum(c, &r->key);
187
188 data_key_init(c, &min_key, ino, min_blk);
189 data_key_init(c, &max_key, ino, max_blk);
190
191 return ubifs_tnc_remove_range(c, &min_key, &max_key);
192}
193
194/**
195 * apply_replay_entry - apply a replay entry to the TNC.
196 * @c: UBIFS file-system description object
197 * @r: replay entry to apply
198 *
199 * Apply a replay entry to the TNC.
200 */
201static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
202{
203 int err, deletion = ((r->flags & REPLAY_DELETION) != 0);
204
205 dbg_mnt("LEB %d:%d len %d flgs %d sqnum %llu %s", r->lnum,
206 r->offs, r->len, r->flags, r->sqnum, DBGKEY(&r->key));
207
208 /* Set c->replay_sqnum to help deal with dangling branches. */
209 c->replay_sqnum = r->sqnum;
210
211 if (r->flags & REPLAY_REF)
212 err = set_bud_lprops(c, r);
213 else if (is_hash_key(c, &r->key)) {
214 if (deletion)
215 err = ubifs_tnc_remove_nm(c, &r->key, &r->nm);
216 else
217 err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs,
218 r->len, &r->nm);
219 } else {
220 if (deletion)
221 switch (key_type(c, &r->key)) {
222 case UBIFS_INO_KEY:
223 {
224 ino_t inum = key_inum(c, &r->key);
225
226 err = ubifs_tnc_remove_ino(c, inum);
227 break;
228 }
229 case UBIFS_TRUN_KEY:
230 err = trun_remove_range(c, r);
231 break;
232 default:
233 err = ubifs_tnc_remove(c, &r->key);
234 break;
235 }
236 else
237 err = ubifs_tnc_add(c, &r->key, r->lnum, r->offs,
238 r->len);
239 if (err)
240 return err;
241
242 if (c->need_recovery)
243 err = ubifs_recover_size_accum(c, &r->key, deletion,
244 r->new_size);
245 }
246
247 return err;
248}
249
250/**
251 * destroy_replay_tree - destroy the replay.
252 * @c: UBIFS file-system description object
253 *
254 * Destroy the replay tree.
255 */
256static void destroy_replay_tree(struct ubifs_info *c)
257{
258 struct rb_node *this = c->replay_tree.rb_node;
259 struct replay_entry *r;
260
261 while (this) {
262 if (this->rb_left) {
263 this = this->rb_left;
264 continue;
265 } else if (this->rb_right) {
266 this = this->rb_right;
267 continue;
268 }
269 r = rb_entry(this, struct replay_entry, rb);
270 this = rb_parent(this);
271 if (this) {
272 if (this->rb_left == &r->rb)
273 this->rb_left = NULL;
274 else
275 this->rb_right = NULL;
276 }
277 if (is_hash_key(c, &r->key))
278 kfree(r->nm.name);
279 kfree(r);
280 }
281 c->replay_tree = RB_ROOT;
282}
283
284/**
285 * apply_replay_tree - apply the replay tree to the TNC.
286 * @c: UBIFS file-system description object
287 *
288 * Apply the replay tree.
289 * Returns zero in case of success and a negative error code in case of
290 * failure.
291 */
292static int apply_replay_tree(struct ubifs_info *c)
293{
294 struct rb_node *this = rb_first(&c->replay_tree);
295
296 while (this) {
297 struct replay_entry *r;
298 int err;
299
300 cond_resched();
301
302 r = rb_entry(this, struct replay_entry, rb);
303 err = apply_replay_entry(c, r);
304 if (err)
305 return err;
306 this = rb_next(this);
307 }
308 return 0;
309}
310
311/**
312 * insert_node - insert a node to the replay tree.
313 * @c: UBIFS file-system description object
314 * @lnum: node logical eraseblock number
315 * @offs: node offset
316 * @len: node length
317 * @key: node key
318 * @sqnum: sequence number
319 * @deletion: non-zero if this is a deletion
320 * @used: number of bytes in use in a LEB
321 * @old_size: truncation old size
322 * @new_size: truncation new size
323 *
324 * This function inserts a scanned non-direntry node to the replay tree. The
325 * replay tree is an RB-tree containing @struct replay_entry elements which are
326 * indexed by the sequence number. The replay tree is applied at the very end
327 * of the replay process. Since the tree is sorted in sequence number order,
328 * the older modifications are applied first. This function returns zero in
329 * case of success and a negative error code in case of failure.
330 */
331static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
332 union ubifs_key *key, unsigned long long sqnum,
333 int deletion, int *used, loff_t old_size,
334 loff_t new_size)
335{
336 struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL;
337 struct replay_entry *r;
338
339 if (key_inum(c, key) >= c->highest_inum)
340 c->highest_inum = key_inum(c, key);
341
342 dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key));
343 while (*p) {
344 parent = *p;
345 r = rb_entry(parent, struct replay_entry, rb);
346 if (sqnum < r->sqnum) {
347 p = &(*p)->rb_left;
348 continue;
349 } else if (sqnum > r->sqnum) {
350 p = &(*p)->rb_right;
351 continue;
352 }
353 ubifs_err("duplicate sqnum in replay");
354 return -EINVAL;
355 }
356
357 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
358 if (!r)
359 return -ENOMEM;
360
361 if (!deletion)
362 *used += ALIGN(len, 8);
363 r->lnum = lnum;
364 r->offs = offs;
365 r->len = len;
366 r->sqnum = sqnum;
367 r->flags = (deletion ? REPLAY_DELETION : 0);
368 r->old_size = old_size;
369 r->new_size = new_size;
370 key_copy(c, key, &r->key);
371
372 rb_link_node(&r->rb, parent, p);
373 rb_insert_color(&r->rb, &c->replay_tree);
374 return 0;
375}
376
377/**
378 * insert_dent - insert a directory entry node into the replay tree.
379 * @c: UBIFS file-system description object
380 * @lnum: node logical eraseblock number
381 * @offs: node offset
382 * @len: node length
383 * @key: node key
384 * @name: directory entry name
385 * @nlen: directory entry name length
386 * @sqnum: sequence number
387 * @deletion: non-zero if this is a deletion
388 * @used: number of bytes in use in a LEB
389 *
390 * This function inserts a scanned directory entry node to the replay tree.
391 * Returns zero in case of success and a negative error code in case of
392 * failure.
393 *
394 * This function is also used for extended attribute entries because they are
395 * implemented as directory entry nodes.
396 */
397static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
398 union ubifs_key *key, const char *name, int nlen,
399 unsigned long long sqnum, int deletion, int *used)
400{
401 struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL;
402 struct replay_entry *r;
403 char *nbuf;
404
405 if (key_inum(c, key) >= c->highest_inum)
406 c->highest_inum = key_inum(c, key);
407
408 dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key));
409 while (*p) {
410 parent = *p;
411 r = rb_entry(parent, struct replay_entry, rb);
412 if (sqnum < r->sqnum) {
413 p = &(*p)->rb_left;
414 continue;
415 }
416 if (sqnum > r->sqnum) {
417 p = &(*p)->rb_right;
418 continue;
419 }
420 ubifs_err("duplicate sqnum in replay");
421 return -EINVAL;
422 }
423
424 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
425 if (!r)
426 return -ENOMEM;
427 nbuf = kmalloc(nlen + 1, GFP_KERNEL);
428 if (!nbuf) {
429 kfree(r);
430 return -ENOMEM;
431 }
432
433 if (!deletion)
434 *used += ALIGN(len, 8);
435 r->lnum = lnum;
436 r->offs = offs;
437 r->len = len;
438 r->sqnum = sqnum;
439 r->nm.len = nlen;
440 memcpy(nbuf, name, nlen);
441 nbuf[nlen] = '\0';
442 r->nm.name = nbuf;
443 r->flags = (deletion ? REPLAY_DELETION : 0);
444 key_copy(c, key, &r->key);
445
446 ubifs_assert(!*p);
447 rb_link_node(&r->rb, parent, p);
448 rb_insert_color(&r->rb, &c->replay_tree);
449 return 0;
450}
451
452/**
453 * ubifs_validate_entry - validate directory or extended attribute entry node.
454 * @c: UBIFS file-system description object
455 * @dent: the node to validate
456 *
457 * This function validates directory or extended attribute entry node @dent.
458 * Returns zero if the node is all right and a %-EINVAL if not.
459 */
460int ubifs_validate_entry(struct ubifs_info *c,
461 const struct ubifs_dent_node *dent)
462{
463 int key_type = key_type_flash(c, dent->key);
464 int nlen = le16_to_cpu(dent->nlen);
465
466 if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 ||
467 dent->type >= UBIFS_ITYPES_CNT ||
468 nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 ||
469 strnlen(dent->name, nlen) != nlen ||
470 le64_to_cpu(dent->inum) > MAX_INUM) {
471 ubifs_err("bad %s node", key_type == UBIFS_DENT_KEY ?
472 "directory entry" : "extended attribute entry");
473 return -EINVAL;
474 }
475
476 if (key_type != UBIFS_DENT_KEY && key_type != UBIFS_XENT_KEY) {
477 ubifs_err("bad key type %d", key_type);
478 return -EINVAL;
479 }
480
481 return 0;
482}
483
484/**
485 * replay_bud - replay a bud logical eraseblock.
486 * @c: UBIFS file-system description object
487 * @lnum: bud logical eraseblock number to replay
488 * @offs: bud start offset
489 * @jhead: journal head to which this bud belongs
490 * @free: amount of free space in the bud is returned here
491 * @dirty: amount of dirty space from padding and deletion nodes is returned
492 * here
493 *
494 * This function returns zero in case of success and a negative error code in
495 * case of failure.
496 */
497static int replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
498 int *free, int *dirty)
499{
500 int err = 0, used = 0;
501 struct ubifs_scan_leb *sleb;
502 struct ubifs_scan_node *snod;
503 struct ubifs_bud *bud;
504
505 dbg_mnt("replay bud LEB %d, head %d", lnum, jhead);
506 if (c->need_recovery)
507 sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, jhead != GCHD);
508 else
Artem Bityutskiy348709b2009-08-25 15:00:55 +0300509 sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300510 if (IS_ERR(sleb))
511 return PTR_ERR(sleb);
512
513 /*
514 * The bud does not have to start from offset zero - the beginning of
515 * the 'lnum' LEB may contain previously committed data. One of the
516 * things we have to do in replay is to correctly update lprops with
517 * newer information about this LEB.
518 *
519 * At this point lprops thinks that this LEB has 'c->leb_size - offs'
520 * bytes of free space because it only contain information about
521 * committed data.
522 *
523 * But we know that real amount of free space is 'c->leb_size -
524 * sleb->endpt', and the space in the 'lnum' LEB between 'offs' and
525 * 'sleb->endpt' is used by bud data. We have to correctly calculate
526 * how much of these data are dirty and update lprops with this
527 * information.
528 *
529 * The dirt in that LEB region is comprised of padding nodes, deletion
530 * nodes, truncation nodes and nodes which are obsoleted by subsequent
531 * nodes in this LEB. So instead of calculating clean space, we
532 * calculate used space ('used' variable).
533 */
534
535 list_for_each_entry(snod, &sleb->nodes, list) {
536 int deletion = 0;
537
538 cond_resched();
539
540 if (snod->sqnum >= SQNUM_WATERMARK) {
541 ubifs_err("file system's life ended");
542 goto out_dump;
543 }
544
545 if (snod->sqnum > c->max_sqnum)
546 c->max_sqnum = snod->sqnum;
547
548 switch (snod->type) {
549 case UBIFS_INO_NODE:
550 {
551 struct ubifs_ino_node *ino = snod->node;
552 loff_t new_size = le64_to_cpu(ino->size);
553
554 if (le32_to_cpu(ino->nlink) == 0)
555 deletion = 1;
556 err = insert_node(c, lnum, snod->offs, snod->len,
557 &snod->key, snod->sqnum, deletion,
558 &used, 0, new_size);
559 break;
560 }
561 case UBIFS_DATA_NODE:
562 {
563 struct ubifs_data_node *dn = snod->node;
564 loff_t new_size = le32_to_cpu(dn->size) +
565 key_block(c, &snod->key) *
566 UBIFS_BLOCK_SIZE;
567
568 err = insert_node(c, lnum, snod->offs, snod->len,
569 &snod->key, snod->sqnum, deletion,
570 &used, 0, new_size);
571 break;
572 }
573 case UBIFS_DENT_NODE:
574 case UBIFS_XENT_NODE:
575 {
576 struct ubifs_dent_node *dent = snod->node;
577
578 err = ubifs_validate_entry(c, dent);
579 if (err)
580 goto out_dump;
581
582 err = insert_dent(c, lnum, snod->offs, snod->len,
583 &snod->key, dent->name,
584 le16_to_cpu(dent->nlen), snod->sqnum,
585 !le64_to_cpu(dent->inum), &used);
586 break;
587 }
588 case UBIFS_TRUN_NODE:
589 {
590 struct ubifs_trun_node *trun = snod->node;
591 loff_t old_size = le64_to_cpu(trun->old_size);
592 loff_t new_size = le64_to_cpu(trun->new_size);
593 union ubifs_key key;
594
595 /* Validate truncation node */
596 if (old_size < 0 || old_size > c->max_inode_sz ||
597 new_size < 0 || new_size > c->max_inode_sz ||
598 old_size <= new_size) {
599 ubifs_err("bad truncation node");
600 goto out_dump;
601 }
602
603 /*
604 * Create a fake truncation key just to use the same
605 * functions which expect nodes to have keys.
606 */
607 trun_key_init(c, &key, le32_to_cpu(trun->inum));
608 err = insert_node(c, lnum, snod->offs, snod->len,
609 &key, snod->sqnum, 1, &used,
610 old_size, new_size);
611 break;
612 }
613 default:
614 ubifs_err("unexpected node type %d in bud LEB %d:%d",
615 snod->type, lnum, snod->offs);
616 err = -EINVAL;
617 goto out_dump;
618 }
619 if (err)
620 goto out;
621 }
622
623 bud = ubifs_search_bud(c, lnum);
624 if (!bud)
625 BUG();
626
627 ubifs_assert(sleb->endpt - offs >= used);
628 ubifs_assert(sleb->endpt % c->min_io_size == 0);
629
Artem Bityutskiy2ef13292010-09-19 18:34:26 +0300630 if (sleb->endpt + c->min_io_size <= c->leb_size && !c->ro_mount)
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300631 err = ubifs_wbuf_seek_nolock(&c->jheads[jhead].wbuf, lnum,
632 sleb->endpt, UBI_SHORTTERM);
633
634 *dirty = sleb->endpt - offs - used;
635 *free = c->leb_size - sleb->endpt;
636
637out:
638 ubifs_scan_destroy(sleb);
639 return err;
640
641out_dump:
642 ubifs_err("bad node is at LEB %d:%d", lnum, snod->offs);
643 dbg_dump_node(c, snod->node);
644 ubifs_scan_destroy(sleb);
645 return -EINVAL;
646}
647
648/**
649 * insert_ref_node - insert a reference node to the replay tree.
650 * @c: UBIFS file-system description object
651 * @lnum: node logical eraseblock number
652 * @offs: node offset
653 * @sqnum: sequence number
654 * @free: amount of free space in bud
655 * @dirty: amount of dirty space from padding and deletion nodes
656 *
657 * This function inserts a reference node to the replay tree and returns zero
Artem Bityutskiy6edbfaf2008-12-30 20:06:49 +0200658 * in case of success or a negative error code in case of failure.
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300659 */
660static int insert_ref_node(struct ubifs_info *c, int lnum, int offs,
661 unsigned long long sqnum, int free, int dirty)
662{
663 struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL;
664 struct replay_entry *r;
665
666 dbg_mnt("add ref LEB %d:%d", lnum, offs);
667 while (*p) {
668 parent = *p;
669 r = rb_entry(parent, struct replay_entry, rb);
670 if (sqnum < r->sqnum) {
671 p = &(*p)->rb_left;
672 continue;
673 } else if (sqnum > r->sqnum) {
674 p = &(*p)->rb_right;
675 continue;
676 }
677 ubifs_err("duplicate sqnum in replay tree");
678 return -EINVAL;
679 }
680
681 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
682 if (!r)
683 return -ENOMEM;
684
685 r->lnum = lnum;
686 r->offs = offs;
687 r->sqnum = sqnum;
688 r->flags = REPLAY_REF;
689 r->free = free;
690 r->dirty = dirty;
691
692 rb_link_node(&r->rb, parent, p);
693 rb_insert_color(&r->rb, &c->replay_tree);
694 return 0;
695}
696
697/**
698 * replay_buds - replay all buds.
699 * @c: UBIFS file-system description object
700 *
701 * This function returns zero in case of success and a negative error code in
702 * case of failure.
703 */
704static int replay_buds(struct ubifs_info *c)
705{
706 struct bud_entry *b;
707 int err, uninitialized_var(free), uninitialized_var(dirty);
708
709 list_for_each_entry(b, &c->replay_buds, list) {
710 err = replay_bud(c, b->bud->lnum, b->bud->start, b->bud->jhead,
711 &free, &dirty);
712 if (err)
713 return err;
714 err = insert_ref_node(c, b->bud->lnum, b->bud->start, b->sqnum,
715 free, dirty);
716 if (err)
717 return err;
718 }
719
720 return 0;
721}
722
723/**
724 * destroy_bud_list - destroy the list of buds to replay.
725 * @c: UBIFS file-system description object
726 */
727static void destroy_bud_list(struct ubifs_info *c)
728{
729 struct bud_entry *b;
730
731 while (!list_empty(&c->replay_buds)) {
732 b = list_entry(c->replay_buds.next, struct bud_entry, list);
733 list_del(&b->list);
734 kfree(b);
735 }
736}
737
738/**
739 * add_replay_bud - add a bud to the list of buds to replay.
740 * @c: UBIFS file-system description object
741 * @lnum: bud logical eraseblock number to replay
742 * @offs: bud start offset
743 * @jhead: journal head to which this bud belongs
744 * @sqnum: reference node sequence number
745 *
746 * This function returns zero in case of success and a negative error code in
747 * case of failure.
748 */
749static int add_replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
750 unsigned long long sqnum)
751{
752 struct ubifs_bud *bud;
753 struct bud_entry *b;
754
755 dbg_mnt("add replay bud LEB %d:%d, head %d", lnum, offs, jhead);
756
757 bud = kmalloc(sizeof(struct ubifs_bud), GFP_KERNEL);
758 if (!bud)
759 return -ENOMEM;
760
761 b = kmalloc(sizeof(struct bud_entry), GFP_KERNEL);
762 if (!b) {
763 kfree(bud);
764 return -ENOMEM;
765 }
766
767 bud->lnum = lnum;
768 bud->start = offs;
769 bud->jhead = jhead;
770 ubifs_add_bud(c, bud);
771
772 b->bud = bud;
773 b->sqnum = sqnum;
774 list_add_tail(&b->list, &c->replay_buds);
775
776 return 0;
777}
778
779/**
780 * validate_ref - validate a reference node.
781 * @c: UBIFS file-system description object
782 * @ref: the reference node to validate
783 * @ref_lnum: LEB number of the reference node
784 * @ref_offs: reference node offset
785 *
786 * This function returns %1 if a bud reference already exists for the LEB. %0 is
787 * returned if the reference node is new, otherwise %-EINVAL is returned if
788 * validation failed.
789 */
790static int validate_ref(struct ubifs_info *c, const struct ubifs_ref_node *ref)
791{
792 struct ubifs_bud *bud;
793 int lnum = le32_to_cpu(ref->lnum);
794 unsigned int offs = le32_to_cpu(ref->offs);
795 unsigned int jhead = le32_to_cpu(ref->jhead);
796
797 /*
798 * ref->offs may point to the end of LEB when the journal head points
799 * to the end of LEB and we write reference node for it during commit.
800 * So this is why we require 'offs > c->leb_size'.
801 */
802 if (jhead >= c->jhead_cnt || lnum >= c->leb_cnt ||
803 lnum < c->main_first || offs > c->leb_size ||
804 offs & (c->min_io_size - 1))
805 return -EINVAL;
806
807 /* Make sure we have not already looked at this bud */
808 bud = ubifs_search_bud(c, lnum);
809 if (bud) {
810 if (bud->jhead == jhead && bud->start <= offs)
811 return 1;
812 ubifs_err("bud at LEB %d:%d was already referred", lnum, offs);
813 return -EINVAL;
814 }
815
816 return 0;
817}
818
819/**
820 * replay_log_leb - replay a log logical eraseblock.
821 * @c: UBIFS file-system description object
822 * @lnum: log logical eraseblock to replay
823 * @offs: offset to start replaying from
824 * @sbuf: scan buffer
825 *
826 * This function replays a log LEB and returns zero in case of success, %1 if
827 * this is the last LEB in the log, and a negative error code in case of
828 * failure.
829 */
830static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
831{
832 int err;
833 struct ubifs_scan_leb *sleb;
834 struct ubifs_scan_node *snod;
835 const struct ubifs_cs_node *node;
836
837 dbg_mnt("replay log LEB %d:%d", lnum, offs);
Artem Bityutskiy348709b2009-08-25 15:00:55 +0300838 sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery);
839 if (IS_ERR(sleb)) {
Artem Bityutskiyed43f2f2009-06-29 17:59:23 +0300840 if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery)
841 return PTR_ERR(sleb);
Artem Bityutskiy7d08ae32010-10-17 15:50:19 +0300842 /*
843 * Note, the below function will recover this log LEB only if
844 * it is the last, because unclean reboots can possibly corrupt
845 * only the tail of the log.
846 */
Artem Bityutskiyed43f2f2009-06-29 17:59:23 +0300847 sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300848 if (IS_ERR(sleb))
849 return PTR_ERR(sleb);
850 }
851
852 if (sleb->nodes_cnt == 0) {
853 err = 1;
854 goto out;
855 }
856
857 node = sleb->buf;
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300858 snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
859 if (c->cs_sqnum == 0) {
860 /*
861 * This is the first log LEB we are looking at, make sure that
862 * the first node is a commit start node. Also record its
863 * sequence number so that UBIFS can determine where the log
864 * ends, because all nodes which were have higher sequence
865 * numbers.
866 */
867 if (snod->type != UBIFS_CS_NODE) {
868 dbg_err("first log node at LEB %d:%d is not CS node",
869 lnum, offs);
870 goto out_dump;
871 }
872 if (le64_to_cpu(node->cmt_no) != c->cmt_no) {
873 dbg_err("first CS node at LEB %d:%d has wrong "
874 "commit number %llu expected %llu",
875 lnum, offs,
876 (unsigned long long)le64_to_cpu(node->cmt_no),
877 c->cmt_no);
878 goto out_dump;
879 }
880
881 c->cs_sqnum = le64_to_cpu(node->ch.sqnum);
882 dbg_mnt("commit start sqnum %llu", c->cs_sqnum);
883 }
884
885 if (snod->sqnum < c->cs_sqnum) {
886 /*
887 * This means that we reached end of log and now
888 * look to the older log data, which was already
889 * committed but the eraseblock was not erased (UBIFS
Artem Bityutskiy6edbfaf2008-12-30 20:06:49 +0200890 * only un-maps it). So this basically means we have to
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300891 * exit with "end of log" code.
892 */
893 err = 1;
894 goto out;
895 }
896
897 /* Make sure the first node sits at offset zero of the LEB */
898 if (snod->offs != 0) {
899 dbg_err("first node is not at zero offset");
900 goto out_dump;
901 }
902
903 list_for_each_entry(snod, &sleb->nodes, list) {
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300904 cond_resched();
905
906 if (snod->sqnum >= SQNUM_WATERMARK) {
907 ubifs_err("file system's life ended");
908 goto out_dump;
909 }
910
911 if (snod->sqnum < c->cs_sqnum) {
912 dbg_err("bad sqnum %llu, commit sqnum %llu",
913 snod->sqnum, c->cs_sqnum);
914 goto out_dump;
915 }
916
917 if (snod->sqnum > c->max_sqnum)
918 c->max_sqnum = snod->sqnum;
919
920 switch (snod->type) {
921 case UBIFS_REF_NODE: {
922 const struct ubifs_ref_node *ref = snod->node;
923
924 err = validate_ref(c, ref);
925 if (err == 1)
926 break; /* Already have this bud */
927 if (err)
928 goto out_dump;
929
930 err = add_replay_bud(c, le32_to_cpu(ref->lnum),
931 le32_to_cpu(ref->offs),
932 le32_to_cpu(ref->jhead),
933 snod->sqnum);
934 if (err)
935 goto out;
936
937 break;
938 }
939 case UBIFS_CS_NODE:
940 /* Make sure it sits at the beginning of LEB */
941 if (snod->offs != 0) {
942 ubifs_err("unexpected node in log");
943 goto out_dump;
944 }
945 break;
946 default:
947 ubifs_err("unexpected node in log");
948 goto out_dump;
949 }
950 }
951
952 if (sleb->endpt || c->lhead_offs >= c->leb_size) {
953 c->lhead_lnum = lnum;
954 c->lhead_offs = sleb->endpt;
955 }
956
957 err = !sleb->endpt;
958out:
959 ubifs_scan_destroy(sleb);
960 return err;
961
962out_dump:
Adrian Hunter681947d2009-06-24 09:59:38 +0300963 ubifs_err("log error detected while replaying the log at LEB %d:%d",
Artem Bityutskiy1e517642008-07-14 19:08:37 +0300964 lnum, offs + snod->offs);
965 dbg_dump_node(c, snod->node);
966 ubifs_scan_destroy(sleb);
967 return -EINVAL;
968}
969
970/**
971 * take_ihead - update the status of the index head in lprops to 'taken'.
972 * @c: UBIFS file-system description object
973 *
974 * This function returns the amount of free space in the index head LEB or a
975 * negative error code.
976 */
977static int take_ihead(struct ubifs_info *c)
978{
979 const struct ubifs_lprops *lp;
980 int err, free;
981
982 ubifs_get_lprops(c);
983
984 lp = ubifs_lpt_lookup_dirty(c, c->ihead_lnum);
985 if (IS_ERR(lp)) {
986 err = PTR_ERR(lp);
987 goto out;
988 }
989
990 free = lp->free;
991
992 lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC,
993 lp->flags | LPROPS_TAKEN, 0);
994 if (IS_ERR(lp)) {
995 err = PTR_ERR(lp);
996 goto out;
997 }
998
999 err = free;
1000out:
1001 ubifs_release_lprops(c);
1002 return err;
1003}
1004
1005/**
1006 * ubifs_replay_journal - replay journal.
1007 * @c: UBIFS file-system description object
1008 *
1009 * This function scans the journal, replays and cleans it up. It makes sure all
1010 * memory data structures related to uncommitted journal are built (dirty TNC
1011 * tree, tree of buds, modified lprops, etc).
1012 */
1013int ubifs_replay_journal(struct ubifs_info *c)
1014{
1015 int err, i, lnum, offs, free;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001016
1017 BUILD_BUG_ON(UBIFS_TRUN_KEY > 5);
1018
1019 /* Update the status of the index head in lprops to 'taken' */
1020 free = take_ihead(c);
1021 if (free < 0)
1022 return free; /* Error code */
1023
1024 if (c->ihead_offs != c->leb_size - free) {
1025 ubifs_err("bad index head LEB %d:%d", c->ihead_lnum,
1026 c->ihead_offs);
1027 return -EINVAL;
1028 }
1029
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001030 dbg_mnt("start replaying the journal");
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001031 c->replaying = 1;
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001032 lnum = c->ltail_lnum = c->lhead_lnum;
1033 offs = c->lhead_offs;
1034
1035 for (i = 0; i < c->log_lebs; i++, lnum++) {
1036 if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) {
1037 /*
1038 * The log is logically circular, we reached the last
1039 * LEB, switch to the first one.
1040 */
1041 lnum = UBIFS_LOG_LNUM;
1042 offs = 0;
1043 }
Artem Bityutskiy6599fcb2010-10-18 10:00:40 +03001044 err = replay_log_leb(c, lnum, offs, c->sbuf);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001045 if (err == 1)
1046 /* We hit the end of the log */
1047 break;
1048 if (err)
1049 goto out;
1050 offs = 0;
1051 }
1052
1053 err = replay_buds(c);
1054 if (err)
1055 goto out;
1056
1057 err = apply_replay_tree(c);
1058 if (err)
1059 goto out;
1060
Artem Bityutskiy6edbfaf2008-12-30 20:06:49 +02001061 /*
1062 * UBIFS budgeting calculations use @c->budg_uncommitted_idx variable
1063 * to roughly estimate index growth. Things like @c->min_idx_lebs
1064 * depend on it. This means we have to initialize it to make sure
1065 * budgeting works properly.
1066 */
1067 c->budg_uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
1068 c->budg_uncommitted_idx *= c->max_idx_node_sz;
1069
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001070 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,
Artem Bityutskiye84461a2008-10-29 12:08:43 +02001073 (unsigned long)c->highest_inum);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001074out:
1075 destroy_replay_tree(c);
1076 destroy_bud_list(c);
Artem Bityutskiy1e517642008-07-14 19:08:37 +03001077 c->replaying = 0;
1078 return err;
1079}