xref: /openbmc/u-boot/fs/ubifs/recovery.c (revision 419adbfb)
1 /*
2  * This file is part of UBIFS.
3  *
4  * Copyright (C) 2006-2008 Nokia Corporation
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License version 2 as published by
8  * the Free Software Foundation.
9  *
10  * This program is distributed in the hope that it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program; if not, write to the Free Software Foundation, Inc., 51
17  * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18  *
19  * Authors: Adrian Hunter
20  *          Artem Bityutskiy (Битюцкий Артём)
21  */
22 
23 /*
24  * This file implements functions needed to recover from unclean un-mounts.
25  * When UBIFS is mounted, it checks a flag on the master node to determine if
26  * an un-mount was completed sucessfully. If not, the process of mounting
27  * incorparates additional checking and fixing of on-flash data structures.
28  * UBIFS always cleans away all remnants of an unclean un-mount, so that
29  * errors do not accumulate. However UBIFS defers recovery if it is mounted
30  * read-only, and the flash is not modified in that case.
31  */
32 
33 #include "ubifs.h"
34 
35 /**
36  * is_empty - determine whether a buffer is empty (contains all 0xff).
37  * @buf: buffer to clean
38  * @len: length of buffer
39  *
40  * This function returns %1 if the buffer is empty (contains all 0xff) otherwise
41  * %0 is returned.
42  */
43 static int is_empty(void *buf, int len)
44 {
45 	uint8_t *p = buf;
46 	int i;
47 
48 	for (i = 0; i < len; i++)
49 		if (*p++ != 0xff)
50 			return 0;
51 	return 1;
52 }
53 
54 /**
55  * get_master_node - get the last valid master node allowing for corruption.
56  * @c: UBIFS file-system description object
57  * @lnum: LEB number
58  * @pbuf: buffer containing the LEB read, is returned here
59  * @mst: master node, if found, is returned here
60  * @cor: corruption, if found, is returned here
61  *
62  * This function allocates a buffer, reads the LEB into it, and finds and
63  * returns the last valid master node allowing for one area of corruption.
64  * The corrupt area, if there is one, must be consistent with the assumption
65  * that it is the result of an unclean unmount while the master node was being
66  * written. Under those circumstances, it is valid to use the previously written
67  * master node.
68  *
69  * This function returns %0 on success and a negative error code on failure.
70  */
71 static int get_master_node(const struct ubifs_info *c, int lnum, void **pbuf,
72 			   struct ubifs_mst_node **mst, void **cor)
73 {
74 	const int sz = c->mst_node_alsz;
75 	int err, offs, len;
76 	void *sbuf, *buf;
77 
78 	sbuf = vmalloc(c->leb_size);
79 	if (!sbuf)
80 		return -ENOMEM;
81 
82 	err = ubi_read(c->ubi, lnum, sbuf, 0, c->leb_size);
83 	if (err && err != -EBADMSG)
84 		goto out_free;
85 
86 	/* Find the first position that is definitely not a node */
87 	offs = 0;
88 	buf = sbuf;
89 	len = c->leb_size;
90 	while (offs + UBIFS_MST_NODE_SZ <= c->leb_size) {
91 		struct ubifs_ch *ch = buf;
92 
93 		if (le32_to_cpu(ch->magic) != UBIFS_NODE_MAGIC)
94 			break;
95 		offs += sz;
96 		buf  += sz;
97 		len  -= sz;
98 	}
99 	/* See if there was a valid master node before that */
100 	if (offs) {
101 		int ret;
102 
103 		offs -= sz;
104 		buf  -= sz;
105 		len  += sz;
106 		ret = ubifs_scan_a_node(c, buf, len, lnum, offs, 1);
107 		if (ret != SCANNED_A_NODE && offs) {
108 			/* Could have been corruption so check one place back */
109 			offs -= sz;
110 			buf  -= sz;
111 			len  += sz;
112 			ret = ubifs_scan_a_node(c, buf, len, lnum, offs, 1);
113 			if (ret != SCANNED_A_NODE)
114 				/*
115 				 * We accept only one area of corruption because
116 				 * we are assuming that it was caused while
117 				 * trying to write a master node.
118 				 */
119 				goto out_err;
120 		}
121 		if (ret == SCANNED_A_NODE) {
122 			struct ubifs_ch *ch = buf;
123 
124 			if (ch->node_type != UBIFS_MST_NODE)
125 				goto out_err;
126 			dbg_rcvry("found a master node at %d:%d", lnum, offs);
127 			*mst = buf;
128 			offs += sz;
129 			buf  += sz;
130 			len  -= sz;
131 		}
132 	}
133 	/* Check for corruption */
134 	if (offs < c->leb_size) {
135 		if (!is_empty(buf, min_t(int, len, sz))) {
136 			*cor = buf;
137 			dbg_rcvry("found corruption at %d:%d", lnum, offs);
138 		}
139 		offs += sz;
140 		buf  += sz;
141 		len  -= sz;
142 	}
143 	/* Check remaining empty space */
144 	if (offs < c->leb_size)
145 		if (!is_empty(buf, len))
146 			goto out_err;
147 	*pbuf = sbuf;
148 	return 0;
149 
150 out_err:
151 	err = -EINVAL;
152 out_free:
153 	vfree(sbuf);
154 	*mst = NULL;
155 	*cor = NULL;
156 	return err;
157 }
158 
159 /**
160  * write_rcvrd_mst_node - write recovered master node.
161  * @c: UBIFS file-system description object
162  * @mst: master node
163  *
164  * This function returns %0 on success and a negative error code on failure.
165  */
166 static int write_rcvrd_mst_node(struct ubifs_info *c,
167 				struct ubifs_mst_node *mst)
168 {
169 	int err = 0, lnum = UBIFS_MST_LNUM, sz = c->mst_node_alsz;
170 	__le32 save_flags;
171 
172 	dbg_rcvry("recovery");
173 
174 	save_flags = mst->flags;
175 	mst->flags |= cpu_to_le32(UBIFS_MST_RCVRY);
176 
177 	ubifs_prepare_node(c, mst, UBIFS_MST_NODE_SZ, 1);
178 	err = ubi_leb_change(c->ubi, lnum, mst, sz, UBI_SHORTTERM);
179 	if (err)
180 		goto out;
181 	err = ubi_leb_change(c->ubi, lnum + 1, mst, sz, UBI_SHORTTERM);
182 	if (err)
183 		goto out;
184 out:
185 	mst->flags = save_flags;
186 	return err;
187 }
188 
189 /**
190  * ubifs_recover_master_node - recover the master node.
191  * @c: UBIFS file-system description object
192  *
193  * This function recovers the master node from corruption that may occur due to
194  * an unclean unmount.
195  *
196  * This function returns %0 on success and a negative error code on failure.
197  */
198 int ubifs_recover_master_node(struct ubifs_info *c)
199 {
200 	void *buf1 = NULL, *buf2 = NULL, *cor1 = NULL, *cor2 = NULL;
201 	struct ubifs_mst_node *mst1 = NULL, *mst2 = NULL, *mst;
202 	const int sz = c->mst_node_alsz;
203 	int err, offs1, offs2;
204 
205 	dbg_rcvry("recovery");
206 
207 	err = get_master_node(c, UBIFS_MST_LNUM, &buf1, &mst1, &cor1);
208 	if (err)
209 		goto out_free;
210 
211 	err = get_master_node(c, UBIFS_MST_LNUM + 1, &buf2, &mst2, &cor2);
212 	if (err)
213 		goto out_free;
214 
215 	if (mst1) {
216 		offs1 = (void *)mst1 - buf1;
217 		if ((le32_to_cpu(mst1->flags) & UBIFS_MST_RCVRY) &&
218 		    (offs1 == 0 && !cor1)) {
219 			/*
220 			 * mst1 was written by recovery at offset 0 with no
221 			 * corruption.
222 			 */
223 			dbg_rcvry("recovery recovery");
224 			mst = mst1;
225 		} else if (mst2) {
226 			offs2 = (void *)mst2 - buf2;
227 			if (offs1 == offs2) {
228 				/* Same offset, so must be the same */
229 				if (memcmp((void *)mst1 + UBIFS_CH_SZ,
230 					   (void *)mst2 + UBIFS_CH_SZ,
231 					   UBIFS_MST_NODE_SZ - UBIFS_CH_SZ))
232 					goto out_err;
233 				mst = mst1;
234 			} else if (offs2 + sz == offs1) {
235 				/* 1st LEB was written, 2nd was not */
236 				if (cor1)
237 					goto out_err;
238 				mst = mst1;
239 			} else if (offs1 == 0 && offs2 + sz >= c->leb_size) {
240 				/* 1st LEB was unmapped and written, 2nd not */
241 				if (cor1)
242 					goto out_err;
243 				mst = mst1;
244 			} else
245 				goto out_err;
246 		} else {
247 			/*
248 			 * 2nd LEB was unmapped and about to be written, so
249 			 * there must be only one master node in the first LEB
250 			 * and no corruption.
251 			 */
252 			if (offs1 != 0 || cor1)
253 				goto out_err;
254 			mst = mst1;
255 		}
256 	} else {
257 		if (!mst2)
258 			goto out_err;
259 		/*
260 		 * 1st LEB was unmapped and about to be written, so there must
261 		 * be no room left in 2nd LEB.
262 		 */
263 		offs2 = (void *)mst2 - buf2;
264 		if (offs2 + sz + sz <= c->leb_size)
265 			goto out_err;
266 		mst = mst2;
267 	}
268 
269 	dbg_rcvry("recovered master node from LEB %d",
270 		  (mst == mst1 ? UBIFS_MST_LNUM : UBIFS_MST_LNUM + 1));
271 
272 	memcpy(c->mst_node, mst, UBIFS_MST_NODE_SZ);
273 
274 	if ((c->vfs_sb->s_flags & MS_RDONLY)) {
275 		/* Read-only mode. Keep a copy for switching to rw mode */
276 		c->rcvrd_mst_node = kmalloc(sz, GFP_KERNEL);
277 		if (!c->rcvrd_mst_node) {
278 			err = -ENOMEM;
279 			goto out_free;
280 		}
281 		memcpy(c->rcvrd_mst_node, c->mst_node, UBIFS_MST_NODE_SZ);
282 	}
283 
284 	vfree(buf2);
285 	vfree(buf1);
286 
287 	return 0;
288 
289 out_err:
290 	err = -EINVAL;
291 out_free:
292 	ubifs_err("failed to recover master node");
293 	if (mst1) {
294 		dbg_err("dumping first master node");
295 		dbg_dump_node(c, mst1);
296 	}
297 	if (mst2) {
298 		dbg_err("dumping second master node");
299 		dbg_dump_node(c, mst2);
300 	}
301 	vfree(buf2);
302 	vfree(buf1);
303 	return err;
304 }
305 
306 /**
307  * ubifs_write_rcvrd_mst_node - write the recovered master node.
308  * @c: UBIFS file-system description object
309  *
310  * This function writes the master node that was recovered during mounting in
311  * read-only mode and must now be written because we are remounting rw.
312  *
313  * This function returns %0 on success and a negative error code on failure.
314  */
315 int ubifs_write_rcvrd_mst_node(struct ubifs_info *c)
316 {
317 	int err;
318 
319 	if (!c->rcvrd_mst_node)
320 		return 0;
321 	c->rcvrd_mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
322 	c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
323 	err = write_rcvrd_mst_node(c, c->rcvrd_mst_node);
324 	if (err)
325 		return err;
326 	kfree(c->rcvrd_mst_node);
327 	c->rcvrd_mst_node = NULL;
328 	return 0;
329 }
330 
331 /**
332  * is_last_write - determine if an offset was in the last write to a LEB.
333  * @c: UBIFS file-system description object
334  * @buf: buffer to check
335  * @offs: offset to check
336  *
337  * This function returns %1 if @offs was in the last write to the LEB whose data
338  * is in @buf, otherwise %0 is returned.  The determination is made by checking
339  * for subsequent empty space starting from the next min_io_size boundary (or a
340  * bit less than the common header size if min_io_size is one).
341  */
342 static int is_last_write(const struct ubifs_info *c, void *buf, int offs)
343 {
344 	int empty_offs;
345 	int check_len;
346 	uint8_t *p;
347 
348 	if (c->min_io_size == 1) {
349 		check_len = c->leb_size - offs;
350 		p = buf + check_len;
351 		for (; check_len > 0; check_len--)
352 			if (*--p != 0xff)
353 				break;
354 		/*
355 		 * 'check_len' is the size of the corruption which cannot be
356 		 * more than the size of 1 node if it was caused by an unclean
357 		 * unmount.
358 		 */
359 		if (check_len > UBIFS_MAX_NODE_SZ)
360 			return 0;
361 		return 1;
362 	}
363 
364 	/*
365 	 * Round up to the next c->min_io_size boundary i.e. 'offs' is in the
366 	 * last wbuf written. After that should be empty space.
367 	 */
368 	empty_offs = ALIGN(offs + 1, c->min_io_size);
369 	check_len = c->leb_size - empty_offs;
370 	p = buf + empty_offs - offs;
371 
372 	for (; check_len > 0; check_len--)
373 		if (*p++ != 0xff)
374 			return 0;
375 	return 1;
376 }
377 
378 /**
379  * clean_buf - clean the data from an LEB sitting in a buffer.
380  * @c: UBIFS file-system description object
381  * @buf: buffer to clean
382  * @lnum: LEB number to clean
383  * @offs: offset from which to clean
384  * @len: length of buffer
385  *
386  * This function pads up to the next min_io_size boundary (if there is one) and
387  * sets empty space to all 0xff. @buf, @offs and @len are updated to the next
388  * min_io_size boundary (if there is one).
389  */
390 static void clean_buf(const struct ubifs_info *c, void **buf, int lnum,
391 		      int *offs, int *len)
392 {
393 	int empty_offs, pad_len;
394 
395 	lnum = lnum;
396 	dbg_rcvry("cleaning corruption at %d:%d", lnum, *offs);
397 
398 	if (c->min_io_size == 1) {
399 		memset(*buf, 0xff, c->leb_size - *offs);
400 		return;
401 	}
402 
403 	ubifs_assert(!(*offs & 7));
404 	empty_offs = ALIGN(*offs, c->min_io_size);
405 	pad_len = empty_offs - *offs;
406 	ubifs_pad(c, *buf, pad_len);
407 	*offs += pad_len;
408 	*buf += pad_len;
409 	*len -= pad_len;
410 	memset(*buf, 0xff, c->leb_size - empty_offs);
411 }
412 
413 /**
414  * no_more_nodes - determine if there are no more nodes in a buffer.
415  * @c: UBIFS file-system description object
416  * @buf: buffer to check
417  * @len: length of buffer
418  * @lnum: LEB number of the LEB from which @buf was read
419  * @offs: offset from which @buf was read
420  *
421  * This function ensures that the corrupted node at @offs is the last thing
422  * written to a LEB. This function returns %1 if more data is not found and
423  * %0 if more data is found.
424  */
425 static int no_more_nodes(const struct ubifs_info *c, void *buf, int len,
426 			int lnum, int offs)
427 {
428 	struct ubifs_ch *ch = buf;
429 	int skip, dlen = le32_to_cpu(ch->len);
430 
431 	/* Check for empty space after the corrupt node's common header */
432 	skip = ALIGN(offs + UBIFS_CH_SZ, c->min_io_size) - offs;
433 	if (is_empty(buf + skip, len - skip))
434 		return 1;
435 	/*
436 	 * The area after the common header size is not empty, so the common
437 	 * header must be intact. Check it.
438 	 */
439 	if (ubifs_check_node(c, buf, lnum, offs, 1, 0) != -EUCLEAN) {
440 		dbg_rcvry("unexpected bad common header at %d:%d", lnum, offs);
441 		return 0;
442 	}
443 	/* Now we know the corrupt node's length we can skip over it */
444 	skip = ALIGN(offs + dlen, c->min_io_size) - offs;
445 	/* After which there should be empty space */
446 	if (is_empty(buf + skip, len - skip))
447 		return 1;
448 	dbg_rcvry("unexpected data at %d:%d", lnum, offs + skip);
449 	return 0;
450 }
451 
452 /**
453  * fix_unclean_leb - fix an unclean LEB.
454  * @c: UBIFS file-system description object
455  * @sleb: scanned LEB information
456  * @start: offset where scan started
457  */
458 static int fix_unclean_leb(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
459 			   int start)
460 {
461 	int lnum = sleb->lnum, endpt = start;
462 
463 	/* Get the end offset of the last node we are keeping */
464 	if (!list_empty(&sleb->nodes)) {
465 		struct ubifs_scan_node *snod;
466 
467 		snod = list_entry(sleb->nodes.prev,
468 				  struct ubifs_scan_node, list);
469 		endpt = snod->offs + snod->len;
470 	}
471 
472 	if ((c->vfs_sb->s_flags & MS_RDONLY) && !c->remounting_rw) {
473 		/* Add to recovery list */
474 		struct ubifs_unclean_leb *ucleb;
475 
476 		dbg_rcvry("need to fix LEB %d start %d endpt %d",
477 			  lnum, start, sleb->endpt);
478 		ucleb = kzalloc(sizeof(struct ubifs_unclean_leb), GFP_NOFS);
479 		if (!ucleb)
480 			return -ENOMEM;
481 		ucleb->lnum = lnum;
482 		ucleb->endpt = endpt;
483 		list_add_tail(&ucleb->list, &c->unclean_leb_list);
484 	}
485 	return 0;
486 }
487 
488 /**
489  * drop_incomplete_group - drop nodes from an incomplete group.
490  * @sleb: scanned LEB information
491  * @offs: offset of dropped nodes is returned here
492  *
493  * This function returns %1 if nodes are dropped and %0 otherwise.
494  */
495 static int drop_incomplete_group(struct ubifs_scan_leb *sleb, int *offs)
496 {
497 	int dropped = 0;
498 
499 	while (!list_empty(&sleb->nodes)) {
500 		struct ubifs_scan_node *snod;
501 		struct ubifs_ch *ch;
502 
503 		snod = list_entry(sleb->nodes.prev, struct ubifs_scan_node,
504 				  list);
505 		ch = snod->node;
506 		if (ch->group_type != UBIFS_IN_NODE_GROUP)
507 			return dropped;
508 		dbg_rcvry("dropping node at %d:%d", sleb->lnum, snod->offs);
509 		*offs = snod->offs;
510 		list_del(&snod->list);
511 		kfree(snod);
512 		sleb->nodes_cnt -= 1;
513 		dropped = 1;
514 	}
515 	return dropped;
516 }
517 
518 /**
519  * ubifs_recover_leb - scan and recover a LEB.
520  * @c: UBIFS file-system description object
521  * @lnum: LEB number
522  * @offs: offset
523  * @sbuf: LEB-sized buffer to use
524  * @grouped: nodes may be grouped for recovery
525  *
526  * This function does a scan of a LEB, but caters for errors that might have
527  * been caused by the unclean unmount from which we are attempting to recover.
528  *
529  * This function returns %0 on success and a negative error code on failure.
530  */
531 struct ubifs_scan_leb *ubifs_recover_leb(struct ubifs_info *c, int lnum,
532 					 int offs, void *sbuf, int grouped)
533 {
534 	int err, len = c->leb_size - offs, need_clean = 0, quiet = 1;
535 	int empty_chkd = 0, start = offs;
536 	struct ubifs_scan_leb *sleb;
537 	void *buf = sbuf + offs;
538 
539 	dbg_rcvry("%d:%d", lnum, offs);
540 
541 	sleb = ubifs_start_scan(c, lnum, offs, sbuf);
542 	if (IS_ERR(sleb))
543 		return sleb;
544 
545 	if (sleb->ecc)
546 		need_clean = 1;
547 
548 	while (len >= 8) {
549 		int ret;
550 
551 		dbg_scan("look at LEB %d:%d (%d bytes left)",
552 			 lnum, offs, len);
553 
554 		cond_resched();
555 
556 		/*
557 		 * Scan quietly until there is an error from which we cannot
558 		 * recover
559 		 */
560 		ret = ubifs_scan_a_node(c, buf, len, lnum, offs, quiet);
561 
562 		if (ret == SCANNED_A_NODE) {
563 			/* A valid node, and not a padding node */
564 			struct ubifs_ch *ch = buf;
565 			int node_len;
566 
567 			err = ubifs_add_snod(c, sleb, buf, offs);
568 			if (err)
569 				goto error;
570 			node_len = ALIGN(le32_to_cpu(ch->len), 8);
571 			offs += node_len;
572 			buf += node_len;
573 			len -= node_len;
574 			continue;
575 		}
576 
577 		if (ret > 0) {
578 			/* Padding bytes or a valid padding node */
579 			offs += ret;
580 			buf += ret;
581 			len -= ret;
582 			continue;
583 		}
584 
585 		if (ret == SCANNED_EMPTY_SPACE) {
586 			if (!is_empty(buf, len)) {
587 				if (!is_last_write(c, buf, offs))
588 					break;
589 				clean_buf(c, &buf, lnum, &offs, &len);
590 				need_clean = 1;
591 			}
592 			empty_chkd = 1;
593 			break;
594 		}
595 
596 		if (ret == SCANNED_GARBAGE || ret == SCANNED_A_BAD_PAD_NODE)
597 			if (is_last_write(c, buf, offs)) {
598 				clean_buf(c, &buf, lnum, &offs, &len);
599 				need_clean = 1;
600 				empty_chkd = 1;
601 				break;
602 			}
603 
604 		if (ret == SCANNED_A_CORRUPT_NODE)
605 			if (no_more_nodes(c, buf, len, lnum, offs)) {
606 				clean_buf(c, &buf, lnum, &offs, &len);
607 				need_clean = 1;
608 				empty_chkd = 1;
609 				break;
610 			}
611 
612 		if (quiet) {
613 			/* Redo the last scan but noisily */
614 			quiet = 0;
615 			continue;
616 		}
617 
618 		switch (ret) {
619 		case SCANNED_GARBAGE:
620 			dbg_err("garbage");
621 			goto corrupted;
622 		case SCANNED_A_CORRUPT_NODE:
623 		case SCANNED_A_BAD_PAD_NODE:
624 			dbg_err("bad node");
625 			goto corrupted;
626 		default:
627 			dbg_err("unknown");
628 			goto corrupted;
629 		}
630 	}
631 
632 	if (!empty_chkd && !is_empty(buf, len)) {
633 		if (is_last_write(c, buf, offs)) {
634 			clean_buf(c, &buf, lnum, &offs, &len);
635 			need_clean = 1;
636 		} else {
637 			ubifs_err("corrupt empty space at LEB %d:%d",
638 				  lnum, offs);
639 			goto corrupted;
640 		}
641 	}
642 
643 	/* Drop nodes from incomplete group */
644 	if (grouped && drop_incomplete_group(sleb, &offs)) {
645 		buf = sbuf + offs;
646 		len = c->leb_size - offs;
647 		clean_buf(c, &buf, lnum, &offs, &len);
648 		need_clean = 1;
649 	}
650 
651 	if (offs % c->min_io_size) {
652 		clean_buf(c, &buf, lnum, &offs, &len);
653 		need_clean = 1;
654 	}
655 
656 	ubifs_end_scan(c, sleb, lnum, offs);
657 
658 	if (need_clean) {
659 		err = fix_unclean_leb(c, sleb, start);
660 		if (err)
661 			goto error;
662 	}
663 
664 	return sleb;
665 
666 corrupted:
667 	ubifs_scanned_corruption(c, lnum, offs, buf);
668 	err = -EUCLEAN;
669 error:
670 	ubifs_err("LEB %d scanning failed", lnum);
671 	ubifs_scan_destroy(sleb);
672 	return ERR_PTR(err);
673 }
674 
675 /**
676  * get_cs_sqnum - get commit start sequence number.
677  * @c: UBIFS file-system description object
678  * @lnum: LEB number of commit start node
679  * @offs: offset of commit start node
680  * @cs_sqnum: commit start sequence number is returned here
681  *
682  * This function returns %0 on success and a negative error code on failure.
683  */
684 static int get_cs_sqnum(struct ubifs_info *c, int lnum, int offs,
685 			unsigned long long *cs_sqnum)
686 {
687 	struct ubifs_cs_node *cs_node = NULL;
688 	int err, ret;
689 
690 	dbg_rcvry("at %d:%d", lnum, offs);
691 	cs_node = kmalloc(UBIFS_CS_NODE_SZ, GFP_KERNEL);
692 	if (!cs_node)
693 		return -ENOMEM;
694 	if (c->leb_size - offs < UBIFS_CS_NODE_SZ)
695 		goto out_err;
696 	err = ubi_read(c->ubi, lnum, (void *)cs_node, offs, UBIFS_CS_NODE_SZ);
697 	if (err && err != -EBADMSG)
698 		goto out_free;
699 	ret = ubifs_scan_a_node(c, cs_node, UBIFS_CS_NODE_SZ, lnum, offs, 0);
700 	if (ret != SCANNED_A_NODE) {
701 		dbg_err("Not a valid node");
702 		goto out_err;
703 	}
704 	if (cs_node->ch.node_type != UBIFS_CS_NODE) {
705 		dbg_err("Node a CS node, type is %d", cs_node->ch.node_type);
706 		goto out_err;
707 	}
708 	if (le64_to_cpu(cs_node->cmt_no) != c->cmt_no) {
709 		dbg_err("CS node cmt_no %llu != current cmt_no %llu",
710 			(unsigned long long)le64_to_cpu(cs_node->cmt_no),
711 			c->cmt_no);
712 		goto out_err;
713 	}
714 	*cs_sqnum = le64_to_cpu(cs_node->ch.sqnum);
715 	dbg_rcvry("commit start sqnum %llu", *cs_sqnum);
716 	kfree(cs_node);
717 	return 0;
718 
719 out_err:
720 	err = -EINVAL;
721 out_free:
722 	ubifs_err("failed to get CS sqnum");
723 	kfree(cs_node);
724 	return err;
725 }
726 
727 /**
728  * ubifs_recover_log_leb - scan and recover a log LEB.
729  * @c: UBIFS file-system description object
730  * @lnum: LEB number
731  * @offs: offset
732  * @sbuf: LEB-sized buffer to use
733  *
734  * This function does a scan of a LEB, but caters for errors that might have
735  * been caused by the unclean unmount from which we are attempting to recover.
736  *
737  * This function returns %0 on success and a negative error code on failure.
738  */
739 struct ubifs_scan_leb *ubifs_recover_log_leb(struct ubifs_info *c, int lnum,
740 					     int offs, void *sbuf)
741 {
742 	struct ubifs_scan_leb *sleb;
743 	int next_lnum;
744 
745 	dbg_rcvry("LEB %d", lnum);
746 	next_lnum = lnum + 1;
747 	if (next_lnum >= UBIFS_LOG_LNUM + c->log_lebs)
748 		next_lnum = UBIFS_LOG_LNUM;
749 	if (next_lnum != c->ltail_lnum) {
750 		/*
751 		 * We can only recover at the end of the log, so check that the
752 		 * next log LEB is empty or out of date.
753 		 */
754 		sleb = ubifs_scan(c, next_lnum, 0, sbuf);
755 		if (IS_ERR(sleb))
756 			return sleb;
757 		if (sleb->nodes_cnt) {
758 			struct ubifs_scan_node *snod;
759 			unsigned long long cs_sqnum = c->cs_sqnum;
760 
761 			snod = list_entry(sleb->nodes.next,
762 					  struct ubifs_scan_node, list);
763 			if (cs_sqnum == 0) {
764 				int err;
765 
766 				err = get_cs_sqnum(c, lnum, offs, &cs_sqnum);
767 				if (err) {
768 					ubifs_scan_destroy(sleb);
769 					return ERR_PTR(err);
770 				}
771 			}
772 			if (snod->sqnum > cs_sqnum) {
773 				ubifs_err("unrecoverable log corruption "
774 					  "in LEB %d", lnum);
775 				ubifs_scan_destroy(sleb);
776 				return ERR_PTR(-EUCLEAN);
777 			}
778 		}
779 		ubifs_scan_destroy(sleb);
780 	}
781 	return ubifs_recover_leb(c, lnum, offs, sbuf, 0);
782 }
783 
784 /**
785  * recover_head - recover a head.
786  * @c: UBIFS file-system description object
787  * @lnum: LEB number of head to recover
788  * @offs: offset of head to recover
789  * @sbuf: LEB-sized buffer to use
790  *
791  * This function ensures that there is no data on the flash at a head location.
792  *
793  * This function returns %0 on success and a negative error code on failure.
794  */
795 static int recover_head(const struct ubifs_info *c, int lnum, int offs,
796 			void *sbuf)
797 {
798 	int len, err, need_clean = 0;
799 
800 	if (c->min_io_size > 1)
801 		len = c->min_io_size;
802 	else
803 		len = 512;
804 	if (offs + len > c->leb_size)
805 		len = c->leb_size - offs;
806 
807 	if (!len)
808 		return 0;
809 
810 	/* Read at the head location and check it is empty flash */
811 	err = ubi_read(c->ubi, lnum, sbuf, offs, len);
812 	if (err)
813 		need_clean = 1;
814 	else {
815 		uint8_t *p = sbuf;
816 
817 		while (len--)
818 			if (*p++ != 0xff) {
819 				need_clean = 1;
820 				break;
821 			}
822 	}
823 
824 	if (need_clean) {
825 		dbg_rcvry("cleaning head at %d:%d", lnum, offs);
826 		if (offs == 0)
827 			return ubifs_leb_unmap(c, lnum);
828 		err = ubi_read(c->ubi, lnum, sbuf, 0, offs);
829 		if (err)
830 			return err;
831 		return ubi_leb_change(c->ubi, lnum, sbuf, offs, UBI_UNKNOWN);
832 	}
833 
834 	return 0;
835 }
836 
837 /**
838  * ubifs_recover_inl_heads - recover index and LPT heads.
839  * @c: UBIFS file-system description object
840  * @sbuf: LEB-sized buffer to use
841  *
842  * This function ensures that there is no data on the flash at the index and
843  * LPT head locations.
844  *
845  * This deals with the recovery of a half-completed journal commit. UBIFS is
846  * careful never to overwrite the last version of the index or the LPT. Because
847  * the index and LPT are wandering trees, data from a half-completed commit will
848  * not be referenced anywhere in UBIFS. The data will be either in LEBs that are
849  * assumed to be empty and will be unmapped anyway before use, or in the index
850  * and LPT heads.
851  *
852  * This function returns %0 on success and a negative error code on failure.
853  */
854 int ubifs_recover_inl_heads(const struct ubifs_info *c, void *sbuf)
855 {
856 	int err;
857 
858 	ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY) || c->remounting_rw);
859 
860 	dbg_rcvry("checking index head at %d:%d", c->ihead_lnum, c->ihead_offs);
861 	err = recover_head(c, c->ihead_lnum, c->ihead_offs, sbuf);
862 	if (err)
863 		return err;
864 
865 	dbg_rcvry("checking LPT head at %d:%d", c->nhead_lnum, c->nhead_offs);
866 	err = recover_head(c, c->nhead_lnum, c->nhead_offs, sbuf);
867 	if (err)
868 		return err;
869 
870 	return 0;
871 }
872 
873 /**
874  *  clean_an_unclean_leb - read and write a LEB to remove corruption.
875  * @c: UBIFS file-system description object
876  * @ucleb: unclean LEB information
877  * @sbuf: LEB-sized buffer to use
878  *
879  * This function reads a LEB up to a point pre-determined by the mount recovery,
880  * checks the nodes, and writes the result back to the flash, thereby cleaning
881  * off any following corruption, or non-fatal ECC errors.
882  *
883  * This function returns %0 on success and a negative error code on failure.
884  */
885 static int clean_an_unclean_leb(const struct ubifs_info *c,
886 				struct ubifs_unclean_leb *ucleb, void *sbuf)
887 {
888 	int err, lnum = ucleb->lnum, offs = 0, len = ucleb->endpt, quiet = 1;
889 	void *buf = sbuf;
890 
891 	dbg_rcvry("LEB %d len %d", lnum, len);
892 
893 	if (len == 0) {
894 		/* Nothing to read, just unmap it */
895 		err = ubifs_leb_unmap(c, lnum);
896 		if (err)
897 			return err;
898 		return 0;
899 	}
900 
901 	err = ubi_read(c->ubi, lnum, buf, offs, len);
902 	if (err && err != -EBADMSG)
903 		return err;
904 
905 	while (len >= 8) {
906 		int ret;
907 
908 		cond_resched();
909 
910 		/* Scan quietly until there is an error */
911 		ret = ubifs_scan_a_node(c, buf, len, lnum, offs, quiet);
912 
913 		if (ret == SCANNED_A_NODE) {
914 			/* A valid node, and not a padding node */
915 			struct ubifs_ch *ch = buf;
916 			int node_len;
917 
918 			node_len = ALIGN(le32_to_cpu(ch->len), 8);
919 			offs += node_len;
920 			buf += node_len;
921 			len -= node_len;
922 			continue;
923 		}
924 
925 		if (ret > 0) {
926 			/* Padding bytes or a valid padding node */
927 			offs += ret;
928 			buf += ret;
929 			len -= ret;
930 			continue;
931 		}
932 
933 		if (ret == SCANNED_EMPTY_SPACE) {
934 			ubifs_err("unexpected empty space at %d:%d",
935 				  lnum, offs);
936 			return -EUCLEAN;
937 		}
938 
939 		if (quiet) {
940 			/* Redo the last scan but noisily */
941 			quiet = 0;
942 			continue;
943 		}
944 
945 		ubifs_scanned_corruption(c, lnum, offs, buf);
946 		return -EUCLEAN;
947 	}
948 
949 	/* Pad to min_io_size */
950 	len = ALIGN(ucleb->endpt, c->min_io_size);
951 	if (len > ucleb->endpt) {
952 		int pad_len = len - ALIGN(ucleb->endpt, 8);
953 
954 		if (pad_len > 0) {
955 			buf = c->sbuf + len - pad_len;
956 			ubifs_pad(c, buf, pad_len);
957 		}
958 	}
959 
960 	/* Write back the LEB atomically */
961 	err = ubi_leb_change(c->ubi, lnum, sbuf, len, UBI_UNKNOWN);
962 	if (err)
963 		return err;
964 
965 	dbg_rcvry("cleaned LEB %d", lnum);
966 
967 	return 0;
968 }
969 
970 /**
971  * ubifs_clean_lebs - clean LEBs recovered during read-only mount.
972  * @c: UBIFS file-system description object
973  * @sbuf: LEB-sized buffer to use
974  *
975  * This function cleans a LEB identified during recovery that needs to be
976  * written but was not because UBIFS was mounted read-only. This happens when
977  * remounting to read-write mode.
978  *
979  * This function returns %0 on success and a negative error code on failure.
980  */
981 int ubifs_clean_lebs(const struct ubifs_info *c, void *sbuf)
982 {
983 	dbg_rcvry("recovery");
984 	while (!list_empty(&c->unclean_leb_list)) {
985 		struct ubifs_unclean_leb *ucleb;
986 		int err;
987 
988 		ucleb = list_entry(c->unclean_leb_list.next,
989 				   struct ubifs_unclean_leb, list);
990 		err = clean_an_unclean_leb(c, ucleb, sbuf);
991 		if (err)
992 			return err;
993 		list_del(&ucleb->list);
994 		kfree(ucleb);
995 	}
996 	return 0;
997 }
998 
999 /**
1000  * struct size_entry - inode size information for recovery.
1001  * @rb: link in the RB-tree of sizes
1002  * @inum: inode number
1003  * @i_size: size on inode
1004  * @d_size: maximum size based on data nodes
1005  * @exists: indicates whether the inode exists
1006  * @inode: inode if pinned in memory awaiting rw mode to fix it
1007  */
1008 struct size_entry {
1009 	struct rb_node rb;
1010 	ino_t inum;
1011 	loff_t i_size;
1012 	loff_t d_size;
1013 	int exists;
1014 	struct inode *inode;
1015 };
1016 
1017 /**
1018  * add_ino - add an entry to the size tree.
1019  * @c: UBIFS file-system description object
1020  * @inum: inode number
1021  * @i_size: size on inode
1022  * @d_size: maximum size based on data nodes
1023  * @exists: indicates whether the inode exists
1024  */
1025 static int add_ino(struct ubifs_info *c, ino_t inum, loff_t i_size,
1026 		   loff_t d_size, int exists)
1027 {
1028 	struct rb_node **p = &c->size_tree.rb_node, *parent = NULL;
1029 	struct size_entry *e;
1030 
1031 	while (*p) {
1032 		parent = *p;
1033 		e = rb_entry(parent, struct size_entry, rb);
1034 		if (inum < e->inum)
1035 			p = &(*p)->rb_left;
1036 		else
1037 			p = &(*p)->rb_right;
1038 	}
1039 
1040 	e = kzalloc(sizeof(struct size_entry), GFP_KERNEL);
1041 	if (!e)
1042 		return -ENOMEM;
1043 
1044 	e->inum = inum;
1045 	e->i_size = i_size;
1046 	e->d_size = d_size;
1047 	e->exists = exists;
1048 
1049 	rb_link_node(&e->rb, parent, p);
1050 	rb_insert_color(&e->rb, &c->size_tree);
1051 
1052 	return 0;
1053 }
1054 
1055 /**
1056  * find_ino - find an entry on the size tree.
1057  * @c: UBIFS file-system description object
1058  * @inum: inode number
1059  */
1060 static struct size_entry *find_ino(struct ubifs_info *c, ino_t inum)
1061 {
1062 	struct rb_node *p = c->size_tree.rb_node;
1063 	struct size_entry *e;
1064 
1065 	while (p) {
1066 		e = rb_entry(p, struct size_entry, rb);
1067 		if (inum < e->inum)
1068 			p = p->rb_left;
1069 		else if (inum > e->inum)
1070 			p = p->rb_right;
1071 		else
1072 			return e;
1073 	}
1074 	return NULL;
1075 }
1076 
1077 /**
1078  * remove_ino - remove an entry from the size tree.
1079  * @c: UBIFS file-system description object
1080  * @inum: inode number
1081  */
1082 static void remove_ino(struct ubifs_info *c, ino_t inum)
1083 {
1084 	struct size_entry *e = find_ino(c, inum);
1085 
1086 	if (!e)
1087 		return;
1088 	rb_erase(&e->rb, &c->size_tree);
1089 	kfree(e);
1090 }
1091 
1092 /**
1093  * ubifs_recover_size_accum - accumulate inode sizes for recovery.
1094  * @c: UBIFS file-system description object
1095  * @key: node key
1096  * @deletion: node is for a deletion
1097  * @new_size: inode size
1098  *
1099  * This function has two purposes:
1100  *     1) to ensure there are no data nodes that fall outside the inode size
1101  *     2) to ensure there are no data nodes for inodes that do not exist
1102  * To accomplish those purposes, a rb-tree is constructed containing an entry
1103  * for each inode number in the journal that has not been deleted, and recording
1104  * the size from the inode node, the maximum size of any data node (also altered
1105  * by truncations) and a flag indicating a inode number for which no inode node
1106  * was present in the journal.
1107  *
1108  * Note that there is still the possibility that there are data nodes that have
1109  * been committed that are beyond the inode size, however the only way to find
1110  * them would be to scan the entire index. Alternatively, some provision could
1111  * be made to record the size of inodes at the start of commit, which would seem
1112  * very cumbersome for a scenario that is quite unlikely and the only negative
1113  * consequence of which is wasted space.
1114  *
1115  * This functions returns %0 on success and a negative error code on failure.
1116  */
1117 int ubifs_recover_size_accum(struct ubifs_info *c, union ubifs_key *key,
1118 			     int deletion, loff_t new_size)
1119 {
1120 	ino_t inum = key_inum(c, key);
1121 	struct size_entry *e;
1122 	int err;
1123 
1124 	switch (key_type(c, key)) {
1125 	case UBIFS_INO_KEY:
1126 		if (deletion)
1127 			remove_ino(c, inum);
1128 		else {
1129 			e = find_ino(c, inum);
1130 			if (e) {
1131 				e->i_size = new_size;
1132 				e->exists = 1;
1133 			} else {
1134 				err = add_ino(c, inum, new_size, 0, 1);
1135 				if (err)
1136 					return err;
1137 			}
1138 		}
1139 		break;
1140 	case UBIFS_DATA_KEY:
1141 		e = find_ino(c, inum);
1142 		if (e) {
1143 			if (new_size > e->d_size)
1144 				e->d_size = new_size;
1145 		} else {
1146 			err = add_ino(c, inum, 0, new_size, 0);
1147 			if (err)
1148 				return err;
1149 		}
1150 		break;
1151 	case UBIFS_TRUN_KEY:
1152 		e = find_ino(c, inum);
1153 		if (e)
1154 			e->d_size = new_size;
1155 		break;
1156 	}
1157 	return 0;
1158 }
1159 
1160 /**
1161  * ubifs_recover_size - recover inode size.
1162  * @c: UBIFS file-system description object
1163  *
1164  * This function attempts to fix inode size discrepancies identified by the
1165  * 'ubifs_recover_size_accum()' function.
1166  *
1167  * This functions returns %0 on success and a negative error code on failure.
1168  */
1169 int ubifs_recover_size(struct ubifs_info *c)
1170 {
1171 	struct rb_node *this = rb_first(&c->size_tree);
1172 
1173 	while (this) {
1174 		struct size_entry *e;
1175 		int err;
1176 
1177 		e = rb_entry(this, struct size_entry, rb);
1178 		if (!e->exists) {
1179 			union ubifs_key key;
1180 
1181 			ino_key_init(c, &key, e->inum);
1182 			err = ubifs_tnc_lookup(c, &key, c->sbuf);
1183 			if (err && err != -ENOENT)
1184 				return err;
1185 			if (err == -ENOENT) {
1186 				/* Remove data nodes that have no inode */
1187 				dbg_rcvry("removing ino %lu",
1188 					  (unsigned long)e->inum);
1189 				err = ubifs_tnc_remove_ino(c, e->inum);
1190 				if (err)
1191 					return err;
1192 			} else {
1193 				struct ubifs_ino_node *ino = c->sbuf;
1194 
1195 				e->exists = 1;
1196 				e->i_size = le64_to_cpu(ino->size);
1197 			}
1198 		}
1199 		if (e->exists && e->i_size < e->d_size) {
1200 			if (!e->inode && (c->vfs_sb->s_flags & MS_RDONLY)) {
1201 				/* Fix the inode size and pin it in memory */
1202 				struct inode *inode;
1203 
1204 				inode = ubifs_iget(c->vfs_sb, e->inum);
1205 				if (IS_ERR(inode))
1206 					return PTR_ERR(inode);
1207 				if (inode->i_size < e->d_size) {
1208 					dbg_rcvry("ino %lu size %lld -> %lld",
1209 						  (unsigned long)e->inum,
1210 						  e->d_size, inode->i_size);
1211 					inode->i_size = e->d_size;
1212 					ubifs_inode(inode)->ui_size = e->d_size;
1213 					e->inode = inode;
1214 					this = rb_next(this);
1215 					continue;
1216 				}
1217 				iput(inode);
1218 			}
1219 		}
1220 		this = rb_next(this);
1221 		rb_erase(&e->rb, &c->size_tree);
1222 		kfree(e);
1223 	}
1224 	return 0;
1225 }
1226