1 // SPDX-License-Identifier: GPL-2.0+
2 /*
3  * This file is part of UBIFS.
4  *
5  * Copyright (C) 2006-2008 Nokia Corporation.
6  *
7  * Authors: Adrian Hunter
8  *          Artem Bityutskiy (Битюцкий Артём)
9  */
10 
11 /*
12  * This file contains journal replay code. It runs when the file-system is being
13  * mounted and requires no locking.
14  *
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.
21  */
22 
23 #ifdef __UBOOT__
24 #include <log.h>
25 #include <dm/devres.h>
26 #include <linux/compat.h>
27 #include <linux/err.h>
28 #endif
29 #include "ubifs.h"
30 #include <linux/bug.h>
31 #include <linux/list_sort.h>
32 
33 /**
34  * struct replay_entry - replay list entry.
35  * @lnum: logical eraseblock number of the node
36  * @offs: node offset
37  * @len: node length
38  * @deletion: non-zero if this entry corresponds to a node deletion
39  * @sqnum: node sequence number
40  * @list: links the replay list
41  * @key: node key
42  * @nm: directory entry name
43  * @old_size: truncation old size
44  * @new_size: truncation new size
45  *
46  * The replay process first scans all buds and builds the replay list, then
47  * sorts the replay list in nodes sequence number order, and then inserts all
48  * the replay entries to the TNC.
49  */
50 struct replay_entry {
51 	int lnum;
52 	int offs;
53 	int len;
54 	unsigned int deletion:1;
55 	unsigned long long sqnum;
56 	struct list_head list;
57 	union ubifs_key key;
58 	union {
59 		struct qstr nm;
60 		struct {
61 			loff_t old_size;
62 			loff_t new_size;
63 		};
64 	};
65 };
66 
67 /**
68  * struct bud_entry - entry in the list of buds to replay.
69  * @list: next bud in the list
70  * @bud: bud description object
71  * @sqnum: reference node sequence number
72  * @free: free bytes in the bud
73  * @dirty: dirty bytes in the bud
74  */
75 struct bud_entry {
76 	struct list_head list;
77 	struct ubifs_bud *bud;
78 	unsigned long long sqnum;
79 	int free;
80 	int dirty;
81 };
82 
83 /**
84  * set_bud_lprops - set free and dirty space used by a bud.
85  * @c: UBIFS file-system description object
86  * @b: bud entry which describes the bud
87  *
88  * This function makes sure the LEB properties of bud @b are set correctly
89  * after the replay. Returns zero in case of success and a negative error code
90  * in case of failure.
91  */
set_bud_lprops(struct ubifs_info * c,struct bud_entry * b)92 static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b)
93 {
94 	const struct ubifs_lprops *lp;
95 	int err = 0, dirty;
96 
97 	ubifs_get_lprops(c);
98 
99 	lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum);
100 	if (IS_ERR(lp)) {
101 		err = PTR_ERR(lp);
102 		goto out;
103 	}
104 
105 	dirty = lp->dirty;
106 	if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
107 		/*
108 		 * The LEB was added to the journal with a starting offset of
109 		 * zero which means the LEB must have been empty. The LEB
110 		 * property values should be @lp->free == @c->leb_size and
111 		 * @lp->dirty == 0, but that is not the case. The reason is that
112 		 * the LEB had been garbage collected before it became the bud,
113 		 * and there was not commit inbetween. The garbage collector
114 		 * resets the free and dirty space without recording it
115 		 * anywhere except lprops, so if there was no commit then
116 		 * lprops does not have that information.
117 		 *
118 		 * We do not need to adjust free space because the scan has told
119 		 * us the exact value which is recorded in the replay entry as
120 		 * @b->free.
121 		 *
122 		 * However we do need to subtract from the dirty space the
123 		 * amount of space that the garbage collector reclaimed, which
124 		 * is the whole LEB minus the amount of space that was free.
125 		 */
126 		dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
127 			lp->free, lp->dirty);
128 		dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
129 			lp->free, lp->dirty);
130 		dirty -= c->leb_size - lp->free;
131 		/*
132 		 * If the replay order was perfect the dirty space would now be
133 		 * zero. The order is not perfect because the journal heads
134 		 * race with each other. This is not a problem but is does mean
135 		 * that the dirty space may temporarily exceed c->leb_size
136 		 * during the replay.
137 		 */
138 		if (dirty != 0)
139 			dbg_mnt("LEB %d lp: %d free %d dirty replay: %d free %d dirty",
140 				b->bud->lnum, lp->free, lp->dirty, b->free,
141 				b->dirty);
142 	}
143 	lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty,
144 			     lp->flags | LPROPS_TAKEN, 0);
145 	if (IS_ERR(lp)) {
146 		err = PTR_ERR(lp);
147 		goto out;
148 	}
149 
150 	/* Make sure the journal head points to the latest bud */
151 	err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf,
152 				     b->bud->lnum, c->leb_size - b->free);
153 
154 out:
155 	ubifs_release_lprops(c);
156 	return err;
157 }
158 
159 /**
160  * set_buds_lprops - set free and dirty space for all replayed buds.
161  * @c: UBIFS file-system description object
162  *
163  * This function sets LEB properties for all replayed buds. Returns zero in
164  * case of success and a negative error code in case of failure.
165  */
set_buds_lprops(struct ubifs_info * c)166 static int set_buds_lprops(struct ubifs_info *c)
167 {
168 	struct bud_entry *b;
169 	int err;
170 
171 	list_for_each_entry(b, &c->replay_buds, list) {
172 		err = set_bud_lprops(c, b);
173 		if (err)
174 			return err;
175 	}
176 
177 	return 0;
178 }
179 
180 /**
181  * trun_remove_range - apply a replay entry for a truncation to the TNC.
182  * @c: UBIFS file-system description object
183  * @r: replay entry of truncation
184  */
trun_remove_range(struct ubifs_info * c,struct replay_entry * r)185 static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r)
186 {
187 	unsigned min_blk, max_blk;
188 	union ubifs_key min_key, max_key;
189 	ino_t ino;
190 
191 	min_blk = r->new_size / UBIFS_BLOCK_SIZE;
192 	if (r->new_size & (UBIFS_BLOCK_SIZE - 1))
193 		min_blk += 1;
194 
195 	max_blk = r->old_size / UBIFS_BLOCK_SIZE;
196 	if ((r->old_size & (UBIFS_BLOCK_SIZE - 1)) == 0)
197 		max_blk -= 1;
198 
199 	ino = key_inum(c, &r->key);
200 
201 	data_key_init(c, &min_key, ino, min_blk);
202 	data_key_init(c, &max_key, ino, max_blk);
203 
204 	return ubifs_tnc_remove_range(c, &min_key, &max_key);
205 }
206 
207 /**
208  * apply_replay_entry - apply a replay entry to the TNC.
209  * @c: UBIFS file-system description object
210  * @r: replay entry to apply
211  *
212  * Apply a replay entry to the TNC.
213  */
apply_replay_entry(struct ubifs_info * c,struct replay_entry * r)214 static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
215 {
216 	int err;
217 
218 	dbg_mntk(&r->key, "LEB %d:%d len %d deletion %d sqnum %llu key ",
219 		 r->lnum, r->offs, r->len, r->deletion, r->sqnum);
220 
221 	/* Set c->replay_sqnum to help deal with dangling branches. */
222 	c->replay_sqnum = r->sqnum;
223 
224 	if (is_hash_key(c, &r->key)) {
225 		if (r->deletion)
226 			err = ubifs_tnc_remove_nm(c, &r->key, &r->nm);
227 		else
228 			err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs,
229 					       r->len, &r->nm);
230 	} else {
231 		if (r->deletion)
232 			switch (key_type(c, &r->key)) {
233 			case UBIFS_INO_KEY:
234 			{
235 				ino_t inum = key_inum(c, &r->key);
236 
237 				err = ubifs_tnc_remove_ino(c, inum);
238 				break;
239 			}
240 			case UBIFS_TRUN_KEY:
241 				err = trun_remove_range(c, r);
242 				break;
243 			default:
244 				err = ubifs_tnc_remove(c, &r->key);
245 				break;
246 			}
247 		else
248 			err = ubifs_tnc_add(c, &r->key, r->lnum, r->offs,
249 					    r->len);
250 		if (err)
251 			return err;
252 
253 		if (c->need_recovery)
254 			err = ubifs_recover_size_accum(c, &r->key, r->deletion,
255 						       r->new_size);
256 	}
257 
258 	return err;
259 }
260 
261 /**
262  * replay_entries_cmp - compare 2 replay entries.
263  * @priv: UBIFS file-system description object
264  * @a: first replay entry
265  * @a: second replay entry
266  *
267  * This is a comparios function for 'list_sort()' which compares 2 replay
268  * entries @a and @b by comparing their sequence numer.  Returns %1 if @a has
269  * greater sequence number and %-1 otherwise.
270  */
replay_entries_cmp(void * priv,struct list_head * a,struct list_head * b)271 static int replay_entries_cmp(void *priv, struct list_head *a,
272 			      struct list_head *b)
273 {
274 	struct replay_entry *ra, *rb;
275 
276 	cond_resched();
277 	if (a == b)
278 		return 0;
279 
280 	ra = list_entry(a, struct replay_entry, list);
281 	rb = list_entry(b, struct replay_entry, list);
282 	ubifs_assert(ra->sqnum != rb->sqnum);
283 	if (ra->sqnum > rb->sqnum)
284 		return 1;
285 	return -1;
286 }
287 
288 /**
289  * apply_replay_list - apply the replay list to the TNC.
290  * @c: UBIFS file-system description object
291  *
292  * Apply all entries in the replay list to the TNC. Returns zero in case of
293  * success and a negative error code in case of failure.
294  */
apply_replay_list(struct ubifs_info * c)295 static int apply_replay_list(struct ubifs_info *c)
296 {
297 	struct replay_entry *r;
298 	int err;
299 
300 	list_sort(c, &c->replay_list, &replay_entries_cmp);
301 
302 	list_for_each_entry(r, &c->replay_list, list) {
303 		cond_resched();
304 
305 		err = apply_replay_entry(c, r);
306 		if (err)
307 			return err;
308 	}
309 
310 	return 0;
311 }
312 
313 /**
314  * destroy_replay_list - destroy the replay.
315  * @c: UBIFS file-system description object
316  *
317  * Destroy the replay list.
318  */
destroy_replay_list(struct ubifs_info * c)319 static void destroy_replay_list(struct ubifs_info *c)
320 {
321 	struct replay_entry *r, *tmp;
322 
323 	list_for_each_entry_safe(r, tmp, &c->replay_list, list) {
324 		if (is_hash_key(c, &r->key))
325 			kfree(r->nm.name);
326 		list_del(&r->list);
327 		kfree(r);
328 	}
329 }
330 
331 /**
332  * insert_node - insert a node to the replay list
333  * @c: UBIFS file-system description object
334  * @lnum: node logical eraseblock number
335  * @offs: node offset
336  * @len: node length
337  * @key: node key
338  * @sqnum: sequence number
339  * @deletion: non-zero if this is a deletion
340  * @used: number of bytes in use in a LEB
341  * @old_size: truncation old size
342  * @new_size: truncation new size
343  *
344  * This function inserts a scanned non-direntry node to the replay list. The
345  * replay list contains @struct replay_entry elements, and we sort this list in
346  * sequence number order before applying it. The replay list is applied at the
347  * very end of the replay process. Since the list is sorted in sequence number
348  * order, the older modifications are applied first. This function returns zero
349  * in case of success and a negative error code in case of failure.
350  */
insert_node(struct ubifs_info * c,int lnum,int offs,int len,union ubifs_key * key,unsigned long long sqnum,int deletion,int * used,loff_t old_size,loff_t new_size)351 static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
352 		       union ubifs_key *key, unsigned long long sqnum,
353 		       int deletion, int *used, loff_t old_size,
354 		       loff_t new_size)
355 {
356 	struct replay_entry *r;
357 
358 	dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
359 
360 	if (key_inum(c, key) >= c->highest_inum)
361 		c->highest_inum = key_inum(c, key);
362 
363 	r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
364 	if (!r)
365 		return -ENOMEM;
366 
367 	if (!deletion)
368 		*used += ALIGN(len, 8);
369 	r->lnum = lnum;
370 	r->offs = offs;
371 	r->len = len;
372 	r->deletion = !!deletion;
373 	r->sqnum = sqnum;
374 	key_copy(c, key, &r->key);
375 	r->old_size = old_size;
376 	r->new_size = new_size;
377 
378 	list_add_tail(&r->list, &c->replay_list);
379 	return 0;
380 }
381 
382 /**
383  * insert_dent - insert a directory entry node into the replay list.
384  * @c: UBIFS file-system description object
385  * @lnum: node logical eraseblock number
386  * @offs: node offset
387  * @len: node length
388  * @key: node key
389  * @name: directory entry name
390  * @nlen: directory entry name length
391  * @sqnum: sequence number
392  * @deletion: non-zero if this is a deletion
393  * @used: number of bytes in use in a LEB
394  *
395  * This function inserts a scanned directory entry node or an extended
396  * attribute entry to the replay list. Returns zero in case of success and a
397  * negative error code in case of failure.
398  */
insert_dent(struct ubifs_info * c,int lnum,int offs,int len,union ubifs_key * key,const char * name,int nlen,unsigned long long sqnum,int deletion,int * used)399 static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
400 		       union ubifs_key *key, const char *name, int nlen,
401 		       unsigned long long sqnum, int deletion, int *used)
402 {
403 	struct replay_entry *r;
404 	char *nbuf;
405 
406 	dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
407 	if (key_inum(c, key) >= c->highest_inum)
408 		c->highest_inum = key_inum(c, key);
409 
410 	r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
411 	if (!r)
412 		return -ENOMEM;
413 
414 	nbuf = kmalloc(nlen + 1, GFP_KERNEL);
415 	if (!nbuf) {
416 		kfree(r);
417 		return -ENOMEM;
418 	}
419 
420 	if (!deletion)
421 		*used += ALIGN(len, 8);
422 	r->lnum = lnum;
423 	r->offs = offs;
424 	r->len = len;
425 	r->deletion = !!deletion;
426 	r->sqnum = sqnum;
427 	key_copy(c, key, &r->key);
428 	r->nm.len = nlen;
429 	memcpy(nbuf, name, nlen);
430 	nbuf[nlen] = '\0';
431 	r->nm.name = nbuf;
432 
433 	list_add_tail(&r->list, &c->replay_list);
434 	return 0;
435 }
436 
437 /**
438  * ubifs_validate_entry - validate directory or extended attribute entry node.
439  * @c: UBIFS file-system description object
440  * @dent: the node to validate
441  *
442  * This function validates directory or extended attribute entry node @dent.
443  * Returns zero if the node is all right and a %-EINVAL if not.
444  */
ubifs_validate_entry(struct ubifs_info * c,const struct ubifs_dent_node * dent)445 int ubifs_validate_entry(struct ubifs_info *c,
446 			 const struct ubifs_dent_node *dent)
447 {
448 	int key_type = key_type_flash(c, dent->key);
449 	int nlen = le16_to_cpu(dent->nlen);
450 
451 	if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 ||
452 	    dent->type >= UBIFS_ITYPES_CNT ||
453 	    nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 ||
454 	    strnlen(dent->name, nlen) != nlen ||
455 	    le64_to_cpu(dent->inum) > MAX_INUM) {
456 		ubifs_err(c, "bad %s node", key_type == UBIFS_DENT_KEY ?
457 			  "directory entry" : "extended attribute entry");
458 		return -EINVAL;
459 	}
460 
461 	if (key_type != UBIFS_DENT_KEY && key_type != UBIFS_XENT_KEY) {
462 		ubifs_err(c, "bad key type %d", key_type);
463 		return -EINVAL;
464 	}
465 
466 	return 0;
467 }
468 
469 /**
470  * is_last_bud - check if the bud is the last in the journal head.
471  * @c: UBIFS file-system description object
472  * @bud: bud description object
473  *
474  * This function checks if bud @bud is the last bud in its journal head. This
475  * information is then used by 'replay_bud()' to decide whether the bud can
476  * have corruptions or not. Indeed, only last buds can be corrupted by power
477  * cuts. Returns %1 if this is the last bud, and %0 if not.
478  */
is_last_bud(struct ubifs_info * c,struct ubifs_bud * bud)479 static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud)
480 {
481 	struct ubifs_jhead *jh = &c->jheads[bud->jhead];
482 	struct ubifs_bud *next;
483 	uint32_t data;
484 	int err;
485 
486 	if (list_is_last(&bud->list, &jh->buds_list))
487 		return 1;
488 
489 	/*
490 	 * The following is a quirk to make sure we work correctly with UBIFS
491 	 * images used with older UBIFS.
492 	 *
493 	 * Normally, the last bud will be the last in the journal head's list
494 	 * of bud. However, there is one exception if the UBIFS image belongs
495 	 * to older UBIFS. This is fairly unlikely: one would need to use old
496 	 * UBIFS, then have a power cut exactly at the right point, and then
497 	 * try to mount this image with new UBIFS.
498 	 *
499 	 * The exception is: it is possible to have 2 buds A and B, A goes
500 	 * before B, and B is the last, bud B is contains no data, and bud A is
501 	 * corrupted at the end. The reason is that in older versions when the
502 	 * journal code switched the next bud (from A to B), it first added a
503 	 * log reference node for the new bud (B), and only after this it
504 	 * synchronized the write-buffer of current bud (A). But later this was
505 	 * changed and UBIFS started to always synchronize the write-buffer of
506 	 * the bud (A) before writing the log reference for the new bud (B).
507 	 *
508 	 * But because older UBIFS always synchronized A's write-buffer before
509 	 * writing to B, we can recognize this exceptional situation but
510 	 * checking the contents of bud B - if it is empty, then A can be
511 	 * treated as the last and we can recover it.
512 	 *
513 	 * TODO: remove this piece of code in a couple of years (today it is
514 	 * 16.05.2011).
515 	 */
516 	next = list_entry(bud->list.next, struct ubifs_bud, list);
517 	if (!list_is_last(&next->list, &jh->buds_list))
518 		return 0;
519 
520 	err = ubifs_leb_read(c, next->lnum, (char *)&data, next->start, 4, 1);
521 	if (err)
522 		return 0;
523 
524 	return data == 0xFFFFFFFF;
525 }
526 
527 /**
528  * replay_bud - replay a bud logical eraseblock.
529  * @c: UBIFS file-system description object
530  * @b: bud entry which describes the bud
531  *
532  * This function replays bud @bud, recovers it if needed, and adds all nodes
533  * from this bud to the replay list. Returns zero in case of success and a
534  * negative error code in case of failure.
535  */
replay_bud(struct ubifs_info * c,struct bud_entry * b)536 static int replay_bud(struct ubifs_info *c, struct bud_entry *b)
537 {
538 	int is_last = is_last_bud(c, b->bud);
539 	int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start;
540 	struct ubifs_scan_leb *sleb;
541 	struct ubifs_scan_node *snod;
542 
543 	dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d",
544 		lnum, b->bud->jhead, offs, is_last);
545 
546 	if (c->need_recovery && is_last)
547 		/*
548 		 * Recover only last LEBs in the journal heads, because power
549 		 * cuts may cause corruptions only in these LEBs, because only
550 		 * these LEBs could possibly be written to at the power cut
551 		 * time.
552 		 */
553 		sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, b->bud->jhead);
554 	else
555 		sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0);
556 	if (IS_ERR(sleb))
557 		return PTR_ERR(sleb);
558 
559 	/*
560 	 * The bud does not have to start from offset zero - the beginning of
561 	 * the 'lnum' LEB may contain previously committed data. One of the
562 	 * things we have to do in replay is to correctly update lprops with
563 	 * newer information about this LEB.
564 	 *
565 	 * At this point lprops thinks that this LEB has 'c->leb_size - offs'
566 	 * bytes of free space because it only contain information about
567 	 * committed data.
568 	 *
569 	 * But we know that real amount of free space is 'c->leb_size -
570 	 * sleb->endpt', and the space in the 'lnum' LEB between 'offs' and
571 	 * 'sleb->endpt' is used by bud data. We have to correctly calculate
572 	 * how much of these data are dirty and update lprops with this
573 	 * information.
574 	 *
575 	 * The dirt in that LEB region is comprised of padding nodes, deletion
576 	 * nodes, truncation nodes and nodes which are obsoleted by subsequent
577 	 * nodes in this LEB. So instead of calculating clean space, we
578 	 * calculate used space ('used' variable).
579 	 */
580 
581 	list_for_each_entry(snod, &sleb->nodes, list) {
582 		int deletion = 0;
583 
584 		cond_resched();
585 
586 		if (snod->sqnum >= SQNUM_WATERMARK) {
587 			ubifs_err(c, "file system's life ended");
588 			goto out_dump;
589 		}
590 
591 		if (snod->sqnum > c->max_sqnum)
592 			c->max_sqnum = snod->sqnum;
593 
594 		switch (snod->type) {
595 		case UBIFS_INO_NODE:
596 		{
597 			struct ubifs_ino_node *ino = snod->node;
598 			loff_t new_size = le64_to_cpu(ino->size);
599 
600 			if (le32_to_cpu(ino->nlink) == 0)
601 				deletion = 1;
602 			err = insert_node(c, lnum, snod->offs, snod->len,
603 					  &snod->key, snod->sqnum, deletion,
604 					  &used, 0, new_size);
605 			break;
606 		}
607 		case UBIFS_DATA_NODE:
608 		{
609 			struct ubifs_data_node *dn = snod->node;
610 			loff_t new_size = le32_to_cpu(dn->size) +
611 					  key_block(c, &snod->key) *
612 					  UBIFS_BLOCK_SIZE;
613 
614 			err = insert_node(c, lnum, snod->offs, snod->len,
615 					  &snod->key, snod->sqnum, deletion,
616 					  &used, 0, new_size);
617 			break;
618 		}
619 		case UBIFS_DENT_NODE:
620 		case UBIFS_XENT_NODE:
621 		{
622 			struct ubifs_dent_node *dent = snod->node;
623 
624 			err = ubifs_validate_entry(c, dent);
625 			if (err)
626 				goto out_dump;
627 
628 			err = insert_dent(c, lnum, snod->offs, snod->len,
629 					  &snod->key, dent->name,
630 					  le16_to_cpu(dent->nlen), snod->sqnum,
631 					  !le64_to_cpu(dent->inum), &used);
632 			break;
633 		}
634 		case UBIFS_TRUN_NODE:
635 		{
636 			struct ubifs_trun_node *trun = snod->node;
637 			loff_t old_size = le64_to_cpu(trun->old_size);
638 			loff_t new_size = le64_to_cpu(trun->new_size);
639 			union ubifs_key key;
640 
641 			/* Validate truncation node */
642 			if (old_size < 0 || old_size > c->max_inode_sz ||
643 			    new_size < 0 || new_size > c->max_inode_sz ||
644 			    old_size <= new_size) {
645 				ubifs_err(c, "bad truncation node");
646 				goto out_dump;
647 			}
648 
649 			/*
650 			 * Create a fake truncation key just to use the same
651 			 * functions which expect nodes to have keys.
652 			 */
653 			trun_key_init(c, &key, le32_to_cpu(trun->inum));
654 			err = insert_node(c, lnum, snod->offs, snod->len,
655 					  &key, snod->sqnum, 1, &used,
656 					  old_size, new_size);
657 			break;
658 		}
659 		default:
660 			ubifs_err(c, "unexpected node type %d in bud LEB %d:%d",
661 				  snod->type, lnum, snod->offs);
662 			err = -EINVAL;
663 			goto out_dump;
664 		}
665 		if (err)
666 			goto out;
667 	}
668 
669 	ubifs_assert(ubifs_search_bud(c, lnum));
670 	ubifs_assert(sleb->endpt - offs >= used);
671 	ubifs_assert(sleb->endpt % c->min_io_size == 0);
672 
673 	b->dirty = sleb->endpt - offs - used;
674 	b->free = c->leb_size - sleb->endpt;
675 	dbg_mnt("bud LEB %d replied: dirty %d, free %d",
676 		lnum, b->dirty, b->free);
677 
678 out:
679 	ubifs_scan_destroy(sleb);
680 	return err;
681 
682 out_dump:
683 	ubifs_err(c, "bad node is at LEB %d:%d", lnum, snod->offs);
684 	ubifs_dump_node(c, snod->node);
685 	ubifs_scan_destroy(sleb);
686 	return -EINVAL;
687 }
688 
689 /**
690  * replay_buds - replay all buds.
691  * @c: UBIFS file-system description object
692  *
693  * This function returns zero in case of success and a negative error code in
694  * case of failure.
695  */
replay_buds(struct ubifs_info * c)696 static int replay_buds(struct ubifs_info *c)
697 {
698 	struct bud_entry *b;
699 	int err;
700 	unsigned long long prev_sqnum = 0;
701 
702 	list_for_each_entry(b, &c->replay_buds, list) {
703 		err = replay_bud(c, b);
704 		if (err)
705 			return err;
706 
707 		ubifs_assert(b->sqnum > prev_sqnum);
708 		prev_sqnum = b->sqnum;
709 	}
710 
711 	return 0;
712 }
713 
714 /**
715  * destroy_bud_list - destroy the list of buds to replay.
716  * @c: UBIFS file-system description object
717  */
destroy_bud_list(struct ubifs_info * c)718 static void destroy_bud_list(struct ubifs_info *c)
719 {
720 	struct bud_entry *b;
721 
722 	while (!list_empty(&c->replay_buds)) {
723 		b = list_entry(c->replay_buds.next, struct bud_entry, list);
724 		list_del(&b->list);
725 		kfree(b);
726 	}
727 }
728 
729 /**
730  * add_replay_bud - add a bud to the list of buds to replay.
731  * @c: UBIFS file-system description object
732  * @lnum: bud logical eraseblock number to replay
733  * @offs: bud start offset
734  * @jhead: journal head to which this bud belongs
735  * @sqnum: reference node sequence number
736  *
737  * This function returns zero in case of success and a negative error code in
738  * case of failure.
739  */
add_replay_bud(struct ubifs_info * c,int lnum,int offs,int jhead,unsigned long long sqnum)740 static int add_replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
741 			  unsigned long long sqnum)
742 {
743 	struct ubifs_bud *bud;
744 	struct bud_entry *b;
745 
746 	dbg_mnt("add replay bud LEB %d:%d, head %d", lnum, offs, jhead);
747 
748 	bud = kmalloc(sizeof(struct ubifs_bud), GFP_KERNEL);
749 	if (!bud)
750 		return -ENOMEM;
751 
752 	b = kmalloc(sizeof(struct bud_entry), GFP_KERNEL);
753 	if (!b) {
754 		kfree(bud);
755 		return -ENOMEM;
756 	}
757 
758 	bud->lnum = lnum;
759 	bud->start = offs;
760 	bud->jhead = jhead;
761 	ubifs_add_bud(c, bud);
762 
763 	b->bud = bud;
764 	b->sqnum = sqnum;
765 	list_add_tail(&b->list, &c->replay_buds);
766 
767 	return 0;
768 }
769 
770 /**
771  * validate_ref - validate a reference node.
772  * @c: UBIFS file-system description object
773  * @ref: the reference node to validate
774  * @ref_lnum: LEB number of the reference node
775  * @ref_offs: reference node offset
776  *
777  * This function returns %1 if a bud reference already exists for the LEB. %0 is
778  * returned if the reference node is new, otherwise %-EINVAL is returned if
779  * validation failed.
780  */
validate_ref(struct ubifs_info * c,const struct ubifs_ref_node * ref)781 static int validate_ref(struct ubifs_info *c, const struct ubifs_ref_node *ref)
782 {
783 	struct ubifs_bud *bud;
784 	int lnum = le32_to_cpu(ref->lnum);
785 	unsigned int offs = le32_to_cpu(ref->offs);
786 	unsigned int jhead = le32_to_cpu(ref->jhead);
787 
788 	/*
789 	 * ref->offs may point to the end of LEB when the journal head points
790 	 * to the end of LEB and we write reference node for it during commit.
791 	 * So this is why we require 'offs > c->leb_size'.
792 	 */
793 	if (jhead >= c->jhead_cnt || lnum >= c->leb_cnt ||
794 	    lnum < c->main_first || offs > c->leb_size ||
795 	    offs & (c->min_io_size - 1))
796 		return -EINVAL;
797 
798 	/* Make sure we have not already looked at this bud */
799 	bud = ubifs_search_bud(c, lnum);
800 	if (bud) {
801 		if (bud->jhead == jhead && bud->start <= offs)
802 			return 1;
803 		ubifs_err(c, "bud at LEB %d:%d was already referred", lnum, offs);
804 		return -EINVAL;
805 	}
806 
807 	return 0;
808 }
809 
810 /**
811  * replay_log_leb - replay a log logical eraseblock.
812  * @c: UBIFS file-system description object
813  * @lnum: log logical eraseblock to replay
814  * @offs: offset to start replaying from
815  * @sbuf: scan buffer
816  *
817  * This function replays a log LEB and returns zero in case of success, %1 if
818  * this is the last LEB in the log, and a negative error code in case of
819  * failure.
820  */
replay_log_leb(struct ubifs_info * c,int lnum,int offs,void * sbuf)821 static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
822 {
823 	int err;
824 	struct ubifs_scan_leb *sleb;
825 	struct ubifs_scan_node *snod;
826 	const struct ubifs_cs_node *node;
827 
828 	dbg_mnt("replay log LEB %d:%d", lnum, offs);
829 	sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery);
830 	if (IS_ERR(sleb)) {
831 		if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery)
832 			return PTR_ERR(sleb);
833 		/*
834 		 * Note, the below function will recover this log LEB only if
835 		 * it is the last, because unclean reboots can possibly corrupt
836 		 * only the tail of the log.
837 		 */
838 		sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
839 		if (IS_ERR(sleb))
840 			return PTR_ERR(sleb);
841 	}
842 
843 	if (sleb->nodes_cnt == 0) {
844 		err = 1;
845 		goto out;
846 	}
847 
848 	node = sleb->buf;
849 	snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
850 	if (c->cs_sqnum == 0) {
851 		/*
852 		 * This is the first log LEB we are looking at, make sure that
853 		 * the first node is a commit start node. Also record its
854 		 * sequence number so that UBIFS can determine where the log
855 		 * ends, because all nodes which were have higher sequence
856 		 * numbers.
857 		 */
858 		if (snod->type != UBIFS_CS_NODE) {
859 			ubifs_err(c, "first log node at LEB %d:%d is not CS node",
860 				  lnum, offs);
861 			goto out_dump;
862 		}
863 		if (le64_to_cpu(node->cmt_no) != c->cmt_no) {
864 			ubifs_err(c, "first CS node at LEB %d:%d has wrong commit number %llu expected %llu",
865 				  lnum, offs,
866 				  (unsigned long long)le64_to_cpu(node->cmt_no),
867 				  c->cmt_no);
868 			goto out_dump;
869 		}
870 
871 		c->cs_sqnum = le64_to_cpu(node->ch.sqnum);
872 		dbg_mnt("commit start sqnum %llu", c->cs_sqnum);
873 	}
874 
875 	if (snod->sqnum < c->cs_sqnum) {
876 		/*
877 		 * This means that we reached end of log and now
878 		 * look to the older log data, which was already
879 		 * committed but the eraseblock was not erased (UBIFS
880 		 * only un-maps it). So this basically means we have to
881 		 * exit with "end of log" code.
882 		 */
883 		err = 1;
884 		goto out;
885 	}
886 
887 	/* Make sure the first node sits at offset zero of the LEB */
888 	if (snod->offs != 0) {
889 		ubifs_err(c, "first node is not at zero offset");
890 		goto out_dump;
891 	}
892 
893 	list_for_each_entry(snod, &sleb->nodes, list) {
894 		cond_resched();
895 
896 		if (snod->sqnum >= SQNUM_WATERMARK) {
897 			ubifs_err(c, "file system's life ended");
898 			goto out_dump;
899 		}
900 
901 		if (snod->sqnum < c->cs_sqnum) {
902 			ubifs_err(c, "bad sqnum %llu, commit sqnum %llu",
903 				  snod->sqnum, c->cs_sqnum);
904 			goto out_dump;
905 		}
906 
907 		if (snod->sqnum > c->max_sqnum)
908 			c->max_sqnum = snod->sqnum;
909 
910 		switch (snod->type) {
911 		case UBIFS_REF_NODE: {
912 			const struct ubifs_ref_node *ref = snod->node;
913 
914 			err = validate_ref(c, ref);
915 			if (err == 1)
916 				break; /* Already have this bud */
917 			if (err)
918 				goto out_dump;
919 
920 			err = add_replay_bud(c, le32_to_cpu(ref->lnum),
921 					     le32_to_cpu(ref->offs),
922 					     le32_to_cpu(ref->jhead),
923 					     snod->sqnum);
924 			if (err)
925 				goto out;
926 
927 			break;
928 		}
929 		case UBIFS_CS_NODE:
930 			/* Make sure it sits at the beginning of LEB */
931 			if (snod->offs != 0) {
932 				ubifs_err(c, "unexpected node in log");
933 				goto out_dump;
934 			}
935 			break;
936 		default:
937 			ubifs_err(c, "unexpected node in log");
938 			goto out_dump;
939 		}
940 	}
941 
942 	if (sleb->endpt || c->lhead_offs >= c->leb_size) {
943 		c->lhead_lnum = lnum;
944 		c->lhead_offs = sleb->endpt;
945 	}
946 
947 	err = !sleb->endpt;
948 out:
949 	ubifs_scan_destroy(sleb);
950 	return err;
951 
952 out_dump:
953 	ubifs_err(c, "log error detected while replaying the log at LEB %d:%d",
954 		  lnum, offs + snod->offs);
955 	ubifs_dump_node(c, snod->node);
956 	ubifs_scan_destroy(sleb);
957 	return -EINVAL;
958 }
959 
960 /**
961  * take_ihead - update the status of the index head in lprops to 'taken'.
962  * @c: UBIFS file-system description object
963  *
964  * This function returns the amount of free space in the index head LEB or a
965  * negative error code.
966  */
take_ihead(struct ubifs_info * c)967 static int take_ihead(struct ubifs_info *c)
968 {
969 	const struct ubifs_lprops *lp;
970 	int err, free;
971 
972 	ubifs_get_lprops(c);
973 
974 	lp = ubifs_lpt_lookup_dirty(c, c->ihead_lnum);
975 	if (IS_ERR(lp)) {
976 		err = PTR_ERR(lp);
977 		goto out;
978 	}
979 
980 	free = lp->free;
981 
982 	lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC,
983 			     lp->flags | LPROPS_TAKEN, 0);
984 	if (IS_ERR(lp)) {
985 		err = PTR_ERR(lp);
986 		goto out;
987 	}
988 
989 	err = free;
990 out:
991 	ubifs_release_lprops(c);
992 	return err;
993 }
994 
995 /**
996  * ubifs_replay_journal - replay journal.
997  * @c: UBIFS file-system description object
998  *
999  * This function scans the journal, replays and cleans it up. It makes sure all
1000  * memory data structures related to uncommitted journal are built (dirty TNC
1001  * tree, tree of buds, modified lprops, etc).
1002  */
ubifs_replay_journal(struct ubifs_info * c)1003 int ubifs_replay_journal(struct ubifs_info *c)
1004 {
1005 	int err, lnum, free;
1006 
1007 	BUILD_BUG_ON(UBIFS_TRUN_KEY > 5);
1008 
1009 	/* Update the status of the index head in lprops to 'taken' */
1010 	free = take_ihead(c);
1011 	if (free < 0)
1012 		return free; /* Error code */
1013 
1014 	if (c->ihead_offs != c->leb_size - free) {
1015 		ubifs_err(c, "bad index head LEB %d:%d", c->ihead_lnum,
1016 			  c->ihead_offs);
1017 		return -EINVAL;
1018 	}
1019 
1020 	dbg_mnt("start replaying the journal");
1021 	c->replaying = 1;
1022 	lnum = c->ltail_lnum = c->lhead_lnum;
1023 
1024 	do {
1025 		err = replay_log_leb(c, lnum, 0, c->sbuf);
1026 		if (err == 1) {
1027 			if (lnum != c->lhead_lnum)
1028 				/* We hit the end of the log */
1029 				break;
1030 
1031 			/*
1032 			 * The head of the log must always start with the
1033 			 * "commit start" node on a properly formatted UBIFS.
1034 			 * But we found no nodes at all, which means that
1035 			 * someting went wrong and we cannot proceed mounting
1036 			 * the file-system.
1037 			 */
1038 			ubifs_err(c, "no UBIFS nodes found at the log head LEB %d:%d, possibly corrupted",
1039 				  lnum, 0);
1040 			err = -EINVAL;
1041 		}
1042 		if (err)
1043 			goto out;
1044 		lnum = ubifs_next_log_lnum(c, lnum);
1045 	} while (lnum != c->ltail_lnum);
1046 
1047 	err = replay_buds(c);
1048 	if (err)
1049 		goto out;
1050 
1051 	err = apply_replay_list(c);
1052 	if (err)
1053 		goto out;
1054 
1055 	err = set_buds_lprops(c);
1056 	if (err)
1057 		goto out;
1058 
1059 	/*
1060 	 * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable
1061 	 * to roughly estimate index growth. Things like @c->bi.min_idx_lebs
1062 	 * depend on it. This means we have to initialize it to make sure
1063 	 * budgeting works properly.
1064 	 */
1065 	c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
1066 	c->bi.uncommitted_idx *= c->max_idx_node_sz;
1067 
1068 	ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery);
1069 	dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, highest_inum %lu",
1070 		c->lhead_lnum, c->lhead_offs, c->max_sqnum,
1071 		(unsigned long)c->highest_inum);
1072 out:
1073 	destroy_replay_list(c);
1074 	destroy_bud_list(c);
1075 	c->replaying = 0;
1076 	return err;
1077 }
1078