1 // SPDX-License-Identifier: GPL-2.0-only
3 * This file is part of UBIFS.
5 * Copyright (C) 2006-2008 Nokia Corporation.
7 * Authors: Adrian Hunter
8 * Artem Bityutskiy (Битюцкий Артём)
12 * This file contains journal replay code. It runs when the file-system is being
13 * mounted and requires no locking.
15 * The larger is the journal, the longer it takes to scan it, so the longer it
16 * takes to mount UBIFS. This is why the journal has limited size which may be
17 * changed depending on the system requirements. But a larger journal gives
18 * faster I/O speed because it writes the index less frequently. So this is a
19 * trade-off. Also, the journal is indexed by the in-memory index (TNC), so the
20 * larger is the journal, the more memory its index may consume.
24 #include <linux/list_sort.h>
25 #include <crypto/hash.h>
26 #include <crypto/algapi.h>
29 * struct replay_entry - replay list entry.
30 * @lnum: logical eraseblock number of the node
33 * @deletion: non-zero if this entry corresponds to a node deletion
34 * @sqnum: node sequence number
35 * @list: links the replay list
37 * @nm: directory entry name
38 * @old_size: truncation old size
39 * @new_size: truncation new size
41 * The replay process first scans all buds and builds the replay list, then
42 * sorts the replay list in nodes sequence number order, and then inserts all
43 * the replay entries to the TNC.
49 u8 hash[UBIFS_HASH_ARR_SZ];
50 unsigned int deletion:1;
51 unsigned long long sqnum;
52 struct list_head list;
55 struct fscrypt_name nm;
64 * struct bud_entry - entry in the list of buds to replay.
65 * @list: next bud in the list
66 * @bud: bud description object
67 * @sqnum: reference node sequence number
68 * @free: free bytes in the bud
69 * @dirty: dirty bytes in the bud
72 struct list_head list;
73 struct ubifs_bud *bud;
74 unsigned long long sqnum;
80 * set_bud_lprops - set free and dirty space used by a bud.
81 * @c: UBIFS file-system description object
82 * @b: bud entry which describes the bud
84 * This function makes sure the LEB properties of bud @b are set correctly
85 * after the replay. Returns zero in case of success and a negative error code
88 static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b)
90 const struct ubifs_lprops *lp;
95 lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum);
102 if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
104 * The LEB was added to the journal with a starting offset of
105 * zero which means the LEB must have been empty. The LEB
106 * property values should be @lp->free == @c->leb_size and
107 * @lp->dirty == 0, but that is not the case. The reason is that
108 * the LEB had been garbage collected before it became the bud,
109 * and there was not commit inbetween. The garbage collector
110 * resets the free and dirty space without recording it
111 * anywhere except lprops, so if there was no commit then
112 * lprops does not have that information.
114 * We do not need to adjust free space because the scan has told
115 * us the exact value which is recorded in the replay entry as
118 * However we do need to subtract from the dirty space the
119 * amount of space that the garbage collector reclaimed, which
120 * is the whole LEB minus the amount of space that was free.
122 dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
123 lp->free, lp->dirty);
124 dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
125 lp->free, lp->dirty);
126 dirty -= c->leb_size - lp->free;
128 * If the replay order was perfect the dirty space would now be
129 * zero. The order is not perfect because the journal heads
130 * race with each other. This is not a problem but is does mean
131 * that the dirty space may temporarily exceed c->leb_size
135 dbg_mnt("LEB %d lp: %d free %d dirty replay: %d free %d dirty",
136 b->bud->lnum, lp->free, lp->dirty, b->free,
139 lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty,
140 lp->flags | LPROPS_TAKEN, 0);
146 /* Make sure the journal head points to the latest bud */
147 err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf,
148 b->bud->lnum, c->leb_size - b->free);
151 ubifs_release_lprops(c);
156 * set_buds_lprops - set free and dirty space for all replayed buds.
157 * @c: UBIFS file-system description object
159 * This function sets LEB properties for all replayed buds. Returns zero in
160 * case of success and a negative error code in case of failure.
162 static int set_buds_lprops(struct ubifs_info *c)
167 list_for_each_entry(b, &c->replay_buds, list) {
168 err = set_bud_lprops(c, b);
177 * trun_remove_range - apply a replay entry for a truncation to the TNC.
178 * @c: UBIFS file-system description object
179 * @r: replay entry of truncation
181 static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r)
183 unsigned min_blk, max_blk;
184 union ubifs_key min_key, max_key;
187 min_blk = r->new_size / UBIFS_BLOCK_SIZE;
188 if (r->new_size & (UBIFS_BLOCK_SIZE - 1))
191 max_blk = r->old_size / UBIFS_BLOCK_SIZE;
192 if ((r->old_size & (UBIFS_BLOCK_SIZE - 1)) == 0)
195 ino = key_inum(c, &r->key);
197 data_key_init(c, &min_key, ino, min_blk);
198 data_key_init(c, &max_key, ino, max_blk);
200 return ubifs_tnc_remove_range(c, &min_key, &max_key);
204 * inode_still_linked - check whether inode in question will be re-linked.
205 * @c: UBIFS file-system description object
206 * @rino: replay entry to test
208 * O_TMPFILE files can be re-linked, this means link count goes from 0 to 1.
209 * This case needs special care, otherwise all references to the inode will
210 * be removed upon the first replay entry of an inode with link count 0
213 static bool inode_still_linked(struct ubifs_info *c, struct replay_entry *rino)
215 struct replay_entry *r;
217 ubifs_assert(c, rino->deletion);
218 ubifs_assert(c, key_type(c, &rino->key) == UBIFS_INO_KEY);
221 * Find the most recent entry for the inode behind @rino and check
222 * whether it is a deletion.
224 list_for_each_entry_reverse(r, &c->replay_list, list) {
225 ubifs_assert(c, r->sqnum >= rino->sqnum);
226 if (key_inum(c, &r->key) == key_inum(c, &rino->key) &&
227 key_type(c, &r->key) == UBIFS_INO_KEY)
228 return r->deletion == 0;
237 * apply_replay_entry - apply a replay entry to the TNC.
238 * @c: UBIFS file-system description object
239 * @r: replay entry to apply
241 * Apply a replay entry to the TNC.
243 static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
247 dbg_mntk(&r->key, "LEB %d:%d len %d deletion %d sqnum %llu key ",
248 r->lnum, r->offs, r->len, r->deletion, r->sqnum);
250 if (is_hash_key(c, &r->key)) {
252 err = ubifs_tnc_remove_nm(c, &r->key, &r->nm);
254 err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs,
255 r->len, r->hash, &r->nm);
258 switch (key_type(c, &r->key)) {
261 ino_t inum = key_inum(c, &r->key);
263 if (inode_still_linked(c, r)) {
268 err = ubifs_tnc_remove_ino(c, inum);
272 err = trun_remove_range(c, r);
275 err = ubifs_tnc_remove(c, &r->key);
279 err = ubifs_tnc_add(c, &r->key, r->lnum, r->offs,
284 if (c->need_recovery)
285 err = ubifs_recover_size_accum(c, &r->key, r->deletion,
293 * replay_entries_cmp - compare 2 replay entries.
294 * @priv: UBIFS file-system description object
295 * @a: first replay entry
296 * @b: second replay entry
298 * This is a comparios function for 'list_sort()' which compares 2 replay
299 * entries @a and @b by comparing their sequence number. Returns %1 if @a has
300 * greater sequence number and %-1 otherwise.
302 static int replay_entries_cmp(void *priv, const struct list_head *a,
303 const struct list_head *b)
305 struct ubifs_info *c = priv;
306 struct replay_entry *ra, *rb;
312 ra = list_entry(a, struct replay_entry, list);
313 rb = list_entry(b, struct replay_entry, list);
314 ubifs_assert(c, ra->sqnum != rb->sqnum);
315 if (ra->sqnum > rb->sqnum)
321 * apply_replay_list - apply the replay list to the TNC.
322 * @c: UBIFS file-system description object
324 * Apply all entries in the replay list to the TNC. Returns zero in case of
325 * success and a negative error code in case of failure.
327 static int apply_replay_list(struct ubifs_info *c)
329 struct replay_entry *r;
332 list_sort(c, &c->replay_list, &replay_entries_cmp);
334 list_for_each_entry(r, &c->replay_list, list) {
337 err = apply_replay_entry(c, r);
346 * destroy_replay_list - destroy the replay.
347 * @c: UBIFS file-system description object
349 * Destroy the replay list.
351 static void destroy_replay_list(struct ubifs_info *c)
353 struct replay_entry *r, *tmp;
355 list_for_each_entry_safe(r, tmp, &c->replay_list, list) {
356 if (is_hash_key(c, &r->key))
357 kfree(fname_name(&r->nm));
364 * insert_node - insert a node to the replay list
365 * @c: UBIFS file-system description object
366 * @lnum: node logical eraseblock number
370 * @sqnum: sequence number
371 * @deletion: non-zero if this is a deletion
372 * @used: number of bytes in use in a LEB
373 * @old_size: truncation old size
374 * @new_size: truncation new size
376 * This function inserts a scanned non-direntry node to the replay list. The
377 * replay list contains @struct replay_entry elements, and we sort this list in
378 * sequence number order before applying it. The replay list is applied at the
379 * very end of the replay process. Since the list is sorted in sequence number
380 * order, the older modifications are applied first. This function returns zero
381 * in case of success and a negative error code in case of failure.
383 static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
384 const u8 *hash, union ubifs_key *key,
385 unsigned long long sqnum, int deletion, int *used,
386 loff_t old_size, loff_t new_size)
388 struct replay_entry *r;
390 dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
392 if (key_inum(c, key) >= c->highest_inum)
393 c->highest_inum = key_inum(c, key);
395 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
400 *used += ALIGN(len, 8);
404 ubifs_copy_hash(c, hash, r->hash);
405 r->deletion = !!deletion;
407 key_copy(c, key, &r->key);
408 r->old_size = old_size;
409 r->new_size = new_size;
411 list_add_tail(&r->list, &c->replay_list);
416 * insert_dent - insert a directory entry node into the replay list.
417 * @c: UBIFS file-system description object
418 * @lnum: node logical eraseblock number
422 * @name: directory entry name
423 * @nlen: directory entry name length
424 * @sqnum: sequence number
425 * @deletion: non-zero if this is a deletion
426 * @used: number of bytes in use in a LEB
428 * This function inserts a scanned directory entry node or an extended
429 * attribute entry to the replay list. Returns zero in case of success and a
430 * negative error code in case of failure.
432 static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
433 const u8 *hash, union ubifs_key *key,
434 const char *name, int nlen, unsigned long long sqnum,
435 int deletion, int *used)
437 struct replay_entry *r;
440 dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
441 if (key_inum(c, key) >= c->highest_inum)
442 c->highest_inum = key_inum(c, key);
444 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
448 nbuf = kmalloc(nlen + 1, GFP_KERNEL);
455 *used += ALIGN(len, 8);
459 ubifs_copy_hash(c, hash, r->hash);
460 r->deletion = !!deletion;
462 key_copy(c, key, &r->key);
463 fname_len(&r->nm) = nlen;
464 memcpy(nbuf, name, nlen);
466 fname_name(&r->nm) = nbuf;
468 list_add_tail(&r->list, &c->replay_list);
473 * ubifs_validate_entry - validate directory or extended attribute entry node.
474 * @c: UBIFS file-system description object
475 * @dent: the node to validate
477 * This function validates directory or extended attribute entry node @dent.
478 * Returns zero if the node is all right and a %-EINVAL if not.
480 int ubifs_validate_entry(struct ubifs_info *c,
481 const struct ubifs_dent_node *dent)
483 int key_type = key_type_flash(c, dent->key);
484 int nlen = le16_to_cpu(dent->nlen);
486 if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 ||
487 dent->type >= UBIFS_ITYPES_CNT ||
488 nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 ||
489 (key_type == UBIFS_XENT_KEY && strnlen(dent->name, nlen) != nlen) ||
490 le64_to_cpu(dent->inum) > MAX_INUM) {
491 ubifs_err(c, "bad %s node", key_type == UBIFS_DENT_KEY ?
492 "directory entry" : "extended attribute entry");
496 if (key_type != UBIFS_DENT_KEY && key_type != UBIFS_XENT_KEY) {
497 ubifs_err(c, "bad key type %d", key_type);
505 * is_last_bud - check if the bud is the last in the journal head.
506 * @c: UBIFS file-system description object
507 * @bud: bud description object
509 * This function checks if bud @bud is the last bud in its journal head. This
510 * information is then used by 'replay_bud()' to decide whether the bud can
511 * have corruptions or not. Indeed, only last buds can be corrupted by power
512 * cuts. Returns %1 if this is the last bud, and %0 if not.
514 static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud)
516 struct ubifs_jhead *jh = &c->jheads[bud->jhead];
517 struct ubifs_bud *next;
521 if (list_is_last(&bud->list, &jh->buds_list))
525 * The following is a quirk to make sure we work correctly with UBIFS
526 * images used with older UBIFS.
528 * Normally, the last bud will be the last in the journal head's list
529 * of bud. However, there is one exception if the UBIFS image belongs
530 * to older UBIFS. This is fairly unlikely: one would need to use old
531 * UBIFS, then have a power cut exactly at the right point, and then
532 * try to mount this image with new UBIFS.
534 * The exception is: it is possible to have 2 buds A and B, A goes
535 * before B, and B is the last, bud B is contains no data, and bud A is
536 * corrupted at the end. The reason is that in older versions when the
537 * journal code switched the next bud (from A to B), it first added a
538 * log reference node for the new bud (B), and only after this it
539 * synchronized the write-buffer of current bud (A). But later this was
540 * changed and UBIFS started to always synchronize the write-buffer of
541 * the bud (A) before writing the log reference for the new bud (B).
543 * But because older UBIFS always synchronized A's write-buffer before
544 * writing to B, we can recognize this exceptional situation but
545 * checking the contents of bud B - if it is empty, then A can be
546 * treated as the last and we can recover it.
548 * TODO: remove this piece of code in a couple of years (today it is
551 next = list_entry(bud->list.next, struct ubifs_bud, list);
552 if (!list_is_last(&next->list, &jh->buds_list))
555 err = ubifs_leb_read(c, next->lnum, (char *)&data, next->start, 4, 1);
559 return data == 0xFFFFFFFF;
562 /* authenticate_sleb_hash is split out for stack usage */
563 static int noinline_for_stack
564 authenticate_sleb_hash(struct ubifs_info *c,
565 struct shash_desc *log_hash, u8 *hash)
567 SHASH_DESC_ON_STACK(hash_desc, c->hash_tfm);
569 hash_desc->tfm = c->hash_tfm;
571 ubifs_shash_copy_state(c, log_hash, hash_desc);
572 return crypto_shash_final(hash_desc, hash);
576 * authenticate_sleb - authenticate one scan LEB
577 * @c: UBIFS file-system description object
578 * @sleb: the scan LEB to authenticate
580 * @is_last: if true, this is the last LEB
582 * This function iterates over the buds of a single LEB authenticating all buds
583 * with the authentication nodes on this LEB. Authentication nodes are written
584 * after some buds and contain a HMAC covering the authentication node itself
585 * and the buds between the last authentication node and the current
586 * authentication node. It can happen that the last buds cannot be authenticated
587 * because a powercut happened when some nodes were written but not the
588 * corresponding authentication node. This function returns the number of nodes
589 * that could be authenticated or a negative error code.
591 static int authenticate_sleb(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
592 struct shash_desc *log_hash, int is_last)
595 struct ubifs_scan_node *snod;
598 u8 hash[UBIFS_HASH_ARR_SZ];
599 u8 hmac[UBIFS_HMAC_ARR_SZ];
601 if (!ubifs_authenticated(c))
602 return sleb->nodes_cnt;
604 list_for_each_entry(snod, &sleb->nodes, list) {
608 if (snod->type == UBIFS_AUTH_NODE) {
609 struct ubifs_auth_node *auth = snod->node;
611 err = authenticate_sleb_hash(c, log_hash, hash);
615 err = crypto_shash_tfm_digest(c->hmac_tfm, hash,
620 err = ubifs_check_hmac(c, auth->hmac, hmac);
627 err = crypto_shash_update(log_hash, snod->node,
636 * A powercut can happen when some nodes were written, but not yet
637 * the corresponding authentication node. This may only happen on
638 * the last bud though.
642 dbg_mnt("%d unauthenticated nodes found on LEB %d, Ignoring them",
643 n_not_auth, sleb->lnum);
646 dbg_mnt("%d unauthenticated nodes found on non-last LEB %d",
647 n_not_auth, sleb->lnum);
654 return err ? err : n_nodes - n_not_auth;
658 * replay_bud - replay a bud logical eraseblock.
659 * @c: UBIFS file-system description object
660 * @b: bud entry which describes the bud
662 * This function replays bud @bud, recovers it if needed, and adds all nodes
663 * from this bud to the replay list. Returns zero in case of success and a
664 * negative error code in case of failure.
666 static int replay_bud(struct ubifs_info *c, struct bud_entry *b)
668 int is_last = is_last_bud(c, b->bud);
669 int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start;
671 struct ubifs_scan_leb *sleb;
672 struct ubifs_scan_node *snod;
674 dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d",
675 lnum, b->bud->jhead, offs, is_last);
677 if (c->need_recovery && is_last)
679 * Recover only last LEBs in the journal heads, because power
680 * cuts may cause corruptions only in these LEBs, because only
681 * these LEBs could possibly be written to at the power cut
684 sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, b->bud->jhead);
686 sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0);
688 return PTR_ERR(sleb);
690 n_nodes = authenticate_sleb(c, sleb, b->bud->log_hash, is_last);
696 ubifs_shash_copy_state(c, b->bud->log_hash,
697 c->jheads[b->bud->jhead].log_hash);
700 * The bud does not have to start from offset zero - the beginning of
701 * the 'lnum' LEB may contain previously committed data. One of the
702 * things we have to do in replay is to correctly update lprops with
703 * newer information about this LEB.
705 * At this point lprops thinks that this LEB has 'c->leb_size - offs'
706 * bytes of free space because it only contain information about
709 * But we know that real amount of free space is 'c->leb_size -
710 * sleb->endpt', and the space in the 'lnum' LEB between 'offs' and
711 * 'sleb->endpt' is used by bud data. We have to correctly calculate
712 * how much of these data are dirty and update lprops with this
715 * The dirt in that LEB region is comprised of padding nodes, deletion
716 * nodes, truncation nodes and nodes which are obsoleted by subsequent
717 * nodes in this LEB. So instead of calculating clean space, we
718 * calculate used space ('used' variable).
721 list_for_each_entry(snod, &sleb->nodes, list) {
722 u8 hash[UBIFS_HASH_ARR_SZ];
727 if (snod->sqnum >= SQNUM_WATERMARK) {
728 ubifs_err(c, "file system's life ended");
732 ubifs_node_calc_hash(c, snod->node, hash);
734 if (snod->sqnum > c->max_sqnum)
735 c->max_sqnum = snod->sqnum;
737 switch (snod->type) {
740 struct ubifs_ino_node *ino = snod->node;
741 loff_t new_size = le64_to_cpu(ino->size);
743 if (le32_to_cpu(ino->nlink) == 0)
745 err = insert_node(c, lnum, snod->offs, snod->len, hash,
746 &snod->key, snod->sqnum, deletion,
750 case UBIFS_DATA_NODE:
752 struct ubifs_data_node *dn = snod->node;
753 loff_t new_size = le32_to_cpu(dn->size) +
754 key_block(c, &snod->key) *
757 err = insert_node(c, lnum, snod->offs, snod->len, hash,
758 &snod->key, snod->sqnum, deletion,
762 case UBIFS_DENT_NODE:
763 case UBIFS_XENT_NODE:
765 struct ubifs_dent_node *dent = snod->node;
767 err = ubifs_validate_entry(c, dent);
771 err = insert_dent(c, lnum, snod->offs, snod->len, hash,
772 &snod->key, dent->name,
773 le16_to_cpu(dent->nlen), snod->sqnum,
774 !le64_to_cpu(dent->inum), &used);
777 case UBIFS_TRUN_NODE:
779 struct ubifs_trun_node *trun = snod->node;
780 loff_t old_size = le64_to_cpu(trun->old_size);
781 loff_t new_size = le64_to_cpu(trun->new_size);
784 /* Validate truncation node */
785 if (old_size < 0 || old_size > c->max_inode_sz ||
786 new_size < 0 || new_size > c->max_inode_sz ||
787 old_size <= new_size) {
788 ubifs_err(c, "bad truncation node");
793 * Create a fake truncation key just to use the same
794 * functions which expect nodes to have keys.
796 trun_key_init(c, &key, le32_to_cpu(trun->inum));
797 err = insert_node(c, lnum, snod->offs, snod->len, hash,
798 &key, snod->sqnum, 1, &used,
802 case UBIFS_AUTH_NODE:
805 ubifs_err(c, "unexpected node type %d in bud LEB %d:%d",
806 snod->type, lnum, snod->offs);
818 ubifs_assert(c, ubifs_search_bud(c, lnum));
819 ubifs_assert(c, sleb->endpt - offs >= used);
820 ubifs_assert(c, sleb->endpt % c->min_io_size == 0);
822 b->dirty = sleb->endpt - offs - used;
823 b->free = c->leb_size - sleb->endpt;
824 dbg_mnt("bud LEB %d replied: dirty %d, free %d",
825 lnum, b->dirty, b->free);
828 ubifs_scan_destroy(sleb);
832 ubifs_err(c, "bad node is at LEB %d:%d", lnum, snod->offs);
833 ubifs_dump_node(c, snod->node, c->leb_size - snod->offs);
834 ubifs_scan_destroy(sleb);
839 * replay_buds - replay all buds.
840 * @c: UBIFS file-system description object
842 * This function returns zero in case of success and a negative error code in
845 static int replay_buds(struct ubifs_info *c)
849 unsigned long long prev_sqnum = 0;
851 list_for_each_entry(b, &c->replay_buds, list) {
852 err = replay_bud(c, b);
856 ubifs_assert(c, b->sqnum > prev_sqnum);
857 prev_sqnum = b->sqnum;
864 * destroy_bud_list - destroy the list of buds to replay.
865 * @c: UBIFS file-system description object
867 static void destroy_bud_list(struct ubifs_info *c)
871 while (!list_empty(&c->replay_buds)) {
872 b = list_entry(c->replay_buds.next, struct bud_entry, list);
879 * add_replay_bud - add a bud to the list of buds to replay.
880 * @c: UBIFS file-system description object
881 * @lnum: bud logical eraseblock number to replay
882 * @offs: bud start offset
883 * @jhead: journal head to which this bud belongs
884 * @sqnum: reference node sequence number
886 * This function returns zero in case of success and a negative error code in
889 static int add_replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
890 unsigned long long sqnum)
892 struct ubifs_bud *bud;
896 dbg_mnt("add replay bud LEB %d:%d, head %d", lnum, offs, jhead);
898 bud = kmalloc(sizeof(struct ubifs_bud), GFP_KERNEL);
902 b = kmalloc(sizeof(struct bud_entry), GFP_KERNEL);
911 bud->log_hash = ubifs_hash_get_desc(c);
912 if (IS_ERR(bud->log_hash)) {
913 err = PTR_ERR(bud->log_hash);
917 ubifs_shash_copy_state(c, c->log_hash, bud->log_hash);
919 ubifs_add_bud(c, bud);
923 list_add_tail(&b->list, &c->replay_buds);
934 * validate_ref - validate a reference node.
935 * @c: UBIFS file-system description object
936 * @ref: the reference node to validate
938 * This function returns %1 if a bud reference already exists for the LEB. %0 is
939 * returned if the reference node is new, otherwise %-EINVAL is returned if
942 static int validate_ref(struct ubifs_info *c, const struct ubifs_ref_node *ref)
944 struct ubifs_bud *bud;
945 int lnum = le32_to_cpu(ref->lnum);
946 unsigned int offs = le32_to_cpu(ref->offs);
947 unsigned int jhead = le32_to_cpu(ref->jhead);
950 * ref->offs may point to the end of LEB when the journal head points
951 * to the end of LEB and we write reference node for it during commit.
952 * So this is why we require 'offs > c->leb_size'.
954 if (jhead >= c->jhead_cnt || lnum >= c->leb_cnt ||
955 lnum < c->main_first || offs > c->leb_size ||
956 offs & (c->min_io_size - 1))
959 /* Make sure we have not already looked at this bud */
960 bud = ubifs_search_bud(c, lnum);
962 if (bud->jhead == jhead && bud->start <= offs)
964 ubifs_err(c, "bud at LEB %d:%d was already referred", lnum, offs);
972 * replay_log_leb - replay a log logical eraseblock.
973 * @c: UBIFS file-system description object
974 * @lnum: log logical eraseblock to replay
975 * @offs: offset to start replaying from
978 * This function replays a log LEB and returns zero in case of success, %1 if
979 * this is the last LEB in the log, and a negative error code in case of
982 static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
985 struct ubifs_scan_leb *sleb;
986 struct ubifs_scan_node *snod;
987 const struct ubifs_cs_node *node;
989 dbg_mnt("replay log LEB %d:%d", lnum, offs);
990 sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery);
992 if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery)
993 return PTR_ERR(sleb);
995 * Note, the below function will recover this log LEB only if
996 * it is the last, because unclean reboots can possibly corrupt
997 * only the tail of the log.
999 sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
1001 return PTR_ERR(sleb);
1004 if (sleb->nodes_cnt == 0) {
1010 snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
1011 if (c->cs_sqnum == 0) {
1013 * This is the first log LEB we are looking at, make sure that
1014 * the first node is a commit start node. Also record its
1015 * sequence number so that UBIFS can determine where the log
1016 * ends, because all nodes which were have higher sequence
1019 if (snod->type != UBIFS_CS_NODE) {
1020 ubifs_err(c, "first log node at LEB %d:%d is not CS node",
1024 if (le64_to_cpu(node->cmt_no) != c->cmt_no) {
1025 ubifs_err(c, "first CS node at LEB %d:%d has wrong commit number %llu expected %llu",
1027 (unsigned long long)le64_to_cpu(node->cmt_no),
1032 c->cs_sqnum = le64_to_cpu(node->ch.sqnum);
1033 dbg_mnt("commit start sqnum %llu", c->cs_sqnum);
1035 err = ubifs_shash_init(c, c->log_hash);
1039 err = ubifs_shash_update(c, c->log_hash, node, UBIFS_CS_NODE_SZ);
1044 if (snod->sqnum < c->cs_sqnum) {
1046 * This means that we reached end of log and now
1047 * look to the older log data, which was already
1048 * committed but the eraseblock was not erased (UBIFS
1049 * only un-maps it). So this basically means we have to
1050 * exit with "end of log" code.
1056 /* Make sure the first node sits at offset zero of the LEB */
1057 if (snod->offs != 0) {
1058 ubifs_err(c, "first node is not at zero offset");
1062 list_for_each_entry(snod, &sleb->nodes, list) {
1065 if (snod->sqnum >= SQNUM_WATERMARK) {
1066 ubifs_err(c, "file system's life ended");
1070 if (snod->sqnum < c->cs_sqnum) {
1071 ubifs_err(c, "bad sqnum %llu, commit sqnum %llu",
1072 snod->sqnum, c->cs_sqnum);
1076 if (snod->sqnum > c->max_sqnum)
1077 c->max_sqnum = snod->sqnum;
1079 switch (snod->type) {
1080 case UBIFS_REF_NODE: {
1081 const struct ubifs_ref_node *ref = snod->node;
1083 err = validate_ref(c, ref);
1085 break; /* Already have this bud */
1089 err = ubifs_shash_update(c, c->log_hash, ref,
1094 err = add_replay_bud(c, le32_to_cpu(ref->lnum),
1095 le32_to_cpu(ref->offs),
1096 le32_to_cpu(ref->jhead),
1104 /* Make sure it sits at the beginning of LEB */
1105 if (snod->offs != 0) {
1106 ubifs_err(c, "unexpected node in log");
1111 ubifs_err(c, "unexpected node in log");
1116 if (sleb->endpt || c->lhead_offs >= c->leb_size) {
1117 c->lhead_lnum = lnum;
1118 c->lhead_offs = sleb->endpt;
1123 ubifs_scan_destroy(sleb);
1127 ubifs_err(c, "log error detected while replaying the log at LEB %d:%d",
1128 lnum, offs + snod->offs);
1129 ubifs_dump_node(c, snod->node, c->leb_size - snod->offs);
1130 ubifs_scan_destroy(sleb);
1135 * take_ihead - update the status of the index head in lprops to 'taken'.
1136 * @c: UBIFS file-system description object
1138 * This function returns the amount of free space in the index head LEB or a
1139 * negative error code.
1141 static int take_ihead(struct ubifs_info *c)
1143 const struct ubifs_lprops *lp;
1146 ubifs_get_lprops(c);
1148 lp = ubifs_lpt_lookup_dirty(c, c->ihead_lnum);
1156 lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC,
1157 lp->flags | LPROPS_TAKEN, 0);
1165 ubifs_release_lprops(c);
1170 * ubifs_replay_journal - replay journal.
1171 * @c: UBIFS file-system description object
1173 * This function scans the journal, replays and cleans it up. It makes sure all
1174 * memory data structures related to uncommitted journal are built (dirty TNC
1175 * tree, tree of buds, modified lprops, etc).
1177 int ubifs_replay_journal(struct ubifs_info *c)
1179 int err, lnum, free;
1181 BUILD_BUG_ON(UBIFS_TRUN_KEY > 5);
1183 /* Update the status of the index head in lprops to 'taken' */
1184 free = take_ihead(c);
1186 return free; /* Error code */
1188 if (c->ihead_offs != c->leb_size - free) {
1189 ubifs_err(c, "bad index head LEB %d:%d", c->ihead_lnum,
1194 dbg_mnt("start replaying the journal");
1196 lnum = c->ltail_lnum = c->lhead_lnum;
1199 err = replay_log_leb(c, lnum, 0, c->sbuf);
1201 if (lnum != c->lhead_lnum)
1202 /* We hit the end of the log */
1206 * The head of the log must always start with the
1207 * "commit start" node on a properly formatted UBIFS.
1208 * But we found no nodes at all, which means that
1209 * something went wrong and we cannot proceed mounting
1212 ubifs_err(c, "no UBIFS nodes found at the log head LEB %d:%d, possibly corrupted",
1218 lnum = ubifs_next_log_lnum(c, lnum);
1219 } while (lnum != c->ltail_lnum);
1221 err = replay_buds(c);
1225 err = apply_replay_list(c);
1229 err = set_buds_lprops(c);
1234 * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable
1235 * to roughly estimate index growth. Things like @c->bi.min_idx_lebs
1236 * depend on it. This means we have to initialize it to make sure
1237 * budgeting works properly.
1239 c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
1240 c->bi.uncommitted_idx *= c->max_idx_node_sz;
1242 ubifs_assert(c, c->bud_bytes <= c->max_bud_bytes || c->need_recovery);
1243 dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, highest_inum %lu",
1244 c->lhead_lnum, c->lhead_offs, c->max_sqnum,
1245 (unsigned long)c->highest_inum);
1247 destroy_replay_list(c);
1248 destroy_bud_list(c);