xref: /openbmc/u-boot/fs/ubifs/orphan.c (revision 2b7ff261)
1 /*
2  * This file is part of UBIFS.
3  *
4  * Copyright (C) 2006-2008 Nokia Corporation.
5  *
6  * SPDX-License-Identifier:	GPL-2.0+
7  *
8  * Author: Adrian Hunter
9  */
10 
11 #include <linux/err.h>
12 #include "ubifs.h"
13 
14 /*
15  * An orphan is an inode number whose inode node has been committed to the index
16  * with a link count of zero. That happens when an open file is deleted
17  * (unlinked) and then a commit is run. In the normal course of events the inode
18  * would be deleted when the file is closed. However in the case of an unclean
19  * unmount, orphans need to be accounted for. After an unclean unmount, the
20  * orphans' inodes must be deleted which means either scanning the entire index
21  * looking for them, or keeping a list on flash somewhere. This unit implements
22  * the latter approach.
23  *
24  * The orphan area is a fixed number of LEBs situated between the LPT area and
25  * the main area. The number of orphan area LEBs is specified when the file
26  * system is created. The minimum number is 1. The size of the orphan area
27  * should be so that it can hold the maximum number of orphans that are expected
28  * to ever exist at one time.
29  *
30  * The number of orphans that can fit in a LEB is:
31  *
32  *         (c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64)
33  *
34  * For example: a 15872 byte LEB can fit 1980 orphans so 1 LEB may be enough.
35  *
36  * Orphans are accumulated in a rb-tree. When an inode's link count drops to
37  * zero, the inode number is added to the rb-tree. It is removed from the tree
38  * when the inode is deleted.  Any new orphans that are in the orphan tree when
39  * the commit is run, are written to the orphan area in 1 or more orphan nodes.
40  * If the orphan area is full, it is consolidated to make space.  There is
41  * always enough space because validation prevents the user from creating more
42  * than the maximum number of orphans allowed.
43  */
44 
45 static int dbg_check_orphans(struct ubifs_info *c);
46 
47 /**
48  * ubifs_add_orphan - add an orphan.
49  * @c: UBIFS file-system description object
50  * @inum: orphan inode number
51  *
52  * Add an orphan. This function is called when an inodes link count drops to
53  * zero.
54  */
55 int ubifs_add_orphan(struct ubifs_info *c, ino_t inum)
56 {
57 	struct ubifs_orphan *orphan, *o;
58 	struct rb_node **p, *parent = NULL;
59 
60 	orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_NOFS);
61 	if (!orphan)
62 		return -ENOMEM;
63 	orphan->inum = inum;
64 	orphan->new = 1;
65 
66 	spin_lock(&c->orphan_lock);
67 	if (c->tot_orphans >= c->max_orphans) {
68 		spin_unlock(&c->orphan_lock);
69 		kfree(orphan);
70 		return -ENFILE;
71 	}
72 	p = &c->orph_tree.rb_node;
73 	while (*p) {
74 		parent = *p;
75 		o = rb_entry(parent, struct ubifs_orphan, rb);
76 		if (inum < o->inum)
77 			p = &(*p)->rb_left;
78 		else if (inum > o->inum)
79 			p = &(*p)->rb_right;
80 		else {
81 			ubifs_err(c, "orphaned twice");
82 			spin_unlock(&c->orphan_lock);
83 			kfree(orphan);
84 			return 0;
85 		}
86 	}
87 	c->tot_orphans += 1;
88 	c->new_orphans += 1;
89 	rb_link_node(&orphan->rb, parent, p);
90 	rb_insert_color(&orphan->rb, &c->orph_tree);
91 	list_add_tail(&orphan->list, &c->orph_list);
92 	list_add_tail(&orphan->new_list, &c->orph_new);
93 	spin_unlock(&c->orphan_lock);
94 	dbg_gen("ino %lu", (unsigned long)inum);
95 	return 0;
96 }
97 
98 /**
99  * ubifs_delete_orphan - delete an orphan.
100  * @c: UBIFS file-system description object
101  * @inum: orphan inode number
102  *
103  * Delete an orphan. This function is called when an inode is deleted.
104  */
105 void ubifs_delete_orphan(struct ubifs_info *c, ino_t inum)
106 {
107 	struct ubifs_orphan *o;
108 	struct rb_node *p;
109 
110 	spin_lock(&c->orphan_lock);
111 	p = c->orph_tree.rb_node;
112 	while (p) {
113 		o = rb_entry(p, struct ubifs_orphan, rb);
114 		if (inum < o->inum)
115 			p = p->rb_left;
116 		else if (inum > o->inum)
117 			p = p->rb_right;
118 		else {
119 			if (o->del) {
120 				spin_unlock(&c->orphan_lock);
121 				dbg_gen("deleted twice ino %lu",
122 					(unsigned long)inum);
123 				return;
124 			}
125 			if (o->cmt) {
126 				o->del = 1;
127 				o->dnext = c->orph_dnext;
128 				c->orph_dnext = o;
129 				spin_unlock(&c->orphan_lock);
130 				dbg_gen("delete later ino %lu",
131 					(unsigned long)inum);
132 				return;
133 			}
134 			rb_erase(p, &c->orph_tree);
135 			list_del(&o->list);
136 			c->tot_orphans -= 1;
137 			if (o->new) {
138 				list_del(&o->new_list);
139 				c->new_orphans -= 1;
140 			}
141 			spin_unlock(&c->orphan_lock);
142 			kfree(o);
143 			dbg_gen("inum %lu", (unsigned long)inum);
144 			return;
145 		}
146 	}
147 	spin_unlock(&c->orphan_lock);
148 	ubifs_err(c, "missing orphan ino %lu", (unsigned long)inum);
149 	dump_stack();
150 }
151 
152 /**
153  * ubifs_orphan_start_commit - start commit of orphans.
154  * @c: UBIFS file-system description object
155  *
156  * Start commit of orphans.
157  */
158 int ubifs_orphan_start_commit(struct ubifs_info *c)
159 {
160 	struct ubifs_orphan *orphan, **last;
161 
162 	spin_lock(&c->orphan_lock);
163 	last = &c->orph_cnext;
164 	list_for_each_entry(orphan, &c->orph_new, new_list) {
165 		ubifs_assert(orphan->new);
166 		ubifs_assert(!orphan->cmt);
167 		orphan->new = 0;
168 		orphan->cmt = 1;
169 		*last = orphan;
170 		last = &orphan->cnext;
171 	}
172 	*last = NULL;
173 	c->cmt_orphans = c->new_orphans;
174 	c->new_orphans = 0;
175 	dbg_cmt("%d orphans to commit", c->cmt_orphans);
176 	INIT_LIST_HEAD(&c->orph_new);
177 	if (c->tot_orphans == 0)
178 		c->no_orphs = 1;
179 	else
180 		c->no_orphs = 0;
181 	spin_unlock(&c->orphan_lock);
182 	return 0;
183 }
184 
185 /**
186  * avail_orphs - calculate available space.
187  * @c: UBIFS file-system description object
188  *
189  * This function returns the number of orphans that can be written in the
190  * available space.
191  */
192 static int avail_orphs(struct ubifs_info *c)
193 {
194 	int avail_lebs, avail, gap;
195 
196 	avail_lebs = c->orph_lebs - (c->ohead_lnum - c->orph_first) - 1;
197 	avail = avail_lebs *
198 	       ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
199 	gap = c->leb_size - c->ohead_offs;
200 	if (gap >= UBIFS_ORPH_NODE_SZ + sizeof(__le64))
201 		avail += (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
202 	return avail;
203 }
204 
205 /**
206  * tot_avail_orphs - calculate total space.
207  * @c: UBIFS file-system description object
208  *
209  * This function returns the number of orphans that can be written in half
210  * the total space. That leaves half the space for adding new orphans.
211  */
212 static int tot_avail_orphs(struct ubifs_info *c)
213 {
214 	int avail_lebs, avail;
215 
216 	avail_lebs = c->orph_lebs;
217 	avail = avail_lebs *
218 	       ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
219 	return avail / 2;
220 }
221 
222 /**
223  * do_write_orph_node - write a node to the orphan head.
224  * @c: UBIFS file-system description object
225  * @len: length of node
226  * @atomic: write atomically
227  *
228  * This function writes a node to the orphan head from the orphan buffer. If
229  * %atomic is not zero, then the write is done atomically. On success, %0 is
230  * returned, otherwise a negative error code is returned.
231  */
232 static int do_write_orph_node(struct ubifs_info *c, int len, int atomic)
233 {
234 	int err = 0;
235 
236 	if (atomic) {
237 		ubifs_assert(c->ohead_offs == 0);
238 		ubifs_prepare_node(c, c->orph_buf, len, 1);
239 		len = ALIGN(len, c->min_io_size);
240 		err = ubifs_leb_change(c, c->ohead_lnum, c->orph_buf, len);
241 	} else {
242 		if (c->ohead_offs == 0) {
243 			/* Ensure LEB has been unmapped */
244 			err = ubifs_leb_unmap(c, c->ohead_lnum);
245 			if (err)
246 				return err;
247 		}
248 		err = ubifs_write_node(c, c->orph_buf, len, c->ohead_lnum,
249 				       c->ohead_offs);
250 	}
251 	return err;
252 }
253 
254 /**
255  * write_orph_node - write an orphan node.
256  * @c: UBIFS file-system description object
257  * @atomic: write atomically
258  *
259  * This function builds an orphan node from the cnext list and writes it to the
260  * orphan head. On success, %0 is returned, otherwise a negative error code
261  * is returned.
262  */
263 static int write_orph_node(struct ubifs_info *c, int atomic)
264 {
265 	struct ubifs_orphan *orphan, *cnext;
266 	struct ubifs_orph_node *orph;
267 	int gap, err, len, cnt, i;
268 
269 	ubifs_assert(c->cmt_orphans > 0);
270 	gap = c->leb_size - c->ohead_offs;
271 	if (gap < UBIFS_ORPH_NODE_SZ + sizeof(__le64)) {
272 		c->ohead_lnum += 1;
273 		c->ohead_offs = 0;
274 		gap = c->leb_size;
275 		if (c->ohead_lnum > c->orph_last) {
276 			/*
277 			 * We limit the number of orphans so that this should
278 			 * never happen.
279 			 */
280 			ubifs_err(c, "out of space in orphan area");
281 			return -EINVAL;
282 		}
283 	}
284 	cnt = (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
285 	if (cnt > c->cmt_orphans)
286 		cnt = c->cmt_orphans;
287 	len = UBIFS_ORPH_NODE_SZ + cnt * sizeof(__le64);
288 	ubifs_assert(c->orph_buf);
289 	orph = c->orph_buf;
290 	orph->ch.node_type = UBIFS_ORPH_NODE;
291 	spin_lock(&c->orphan_lock);
292 	cnext = c->orph_cnext;
293 	for (i = 0; i < cnt; i++) {
294 		orphan = cnext;
295 		ubifs_assert(orphan->cmt);
296 		orph->inos[i] = cpu_to_le64(orphan->inum);
297 		orphan->cmt = 0;
298 		cnext = orphan->cnext;
299 		orphan->cnext = NULL;
300 	}
301 	c->orph_cnext = cnext;
302 	c->cmt_orphans -= cnt;
303 	spin_unlock(&c->orphan_lock);
304 	if (c->cmt_orphans)
305 		orph->cmt_no = cpu_to_le64(c->cmt_no);
306 	else
307 		/* Mark the last node of the commit */
308 		orph->cmt_no = cpu_to_le64((c->cmt_no) | (1ULL << 63));
309 	ubifs_assert(c->ohead_offs + len <= c->leb_size);
310 	ubifs_assert(c->ohead_lnum >= c->orph_first);
311 	ubifs_assert(c->ohead_lnum <= c->orph_last);
312 	err = do_write_orph_node(c, len, atomic);
313 	c->ohead_offs += ALIGN(len, c->min_io_size);
314 	c->ohead_offs = ALIGN(c->ohead_offs, 8);
315 	return err;
316 }
317 
318 /**
319  * write_orph_nodes - write orphan nodes until there are no more to commit.
320  * @c: UBIFS file-system description object
321  * @atomic: write atomically
322  *
323  * This function writes orphan nodes for all the orphans to commit. On success,
324  * %0 is returned, otherwise a negative error code is returned.
325  */
326 static int write_orph_nodes(struct ubifs_info *c, int atomic)
327 {
328 	int err;
329 
330 	while (c->cmt_orphans > 0) {
331 		err = write_orph_node(c, atomic);
332 		if (err)
333 			return err;
334 	}
335 	if (atomic) {
336 		int lnum;
337 
338 		/* Unmap any unused LEBs after consolidation */
339 		for (lnum = c->ohead_lnum + 1; lnum <= c->orph_last; lnum++) {
340 			err = ubifs_leb_unmap(c, lnum);
341 			if (err)
342 				return err;
343 		}
344 	}
345 	return 0;
346 }
347 
348 /**
349  * consolidate - consolidate the orphan area.
350  * @c: UBIFS file-system description object
351  *
352  * This function enables consolidation by putting all the orphans into the list
353  * to commit. The list is in the order that the orphans were added, and the
354  * LEBs are written atomically in order, so at no time can orphans be lost by
355  * an unclean unmount.
356  *
357  * This function returns %0 on success and a negative error code on failure.
358  */
359 static int consolidate(struct ubifs_info *c)
360 {
361 	int tot_avail = tot_avail_orphs(c), err = 0;
362 
363 	spin_lock(&c->orphan_lock);
364 	dbg_cmt("there is space for %d orphans and there are %d",
365 		tot_avail, c->tot_orphans);
366 	if (c->tot_orphans - c->new_orphans <= tot_avail) {
367 		struct ubifs_orphan *orphan, **last;
368 		int cnt = 0;
369 
370 		/* Change the cnext list to include all non-new orphans */
371 		last = &c->orph_cnext;
372 		list_for_each_entry(orphan, &c->orph_list, list) {
373 			if (orphan->new)
374 				continue;
375 			orphan->cmt = 1;
376 			*last = orphan;
377 			last = &orphan->cnext;
378 			cnt += 1;
379 		}
380 		*last = NULL;
381 		ubifs_assert(cnt == c->tot_orphans - c->new_orphans);
382 		c->cmt_orphans = cnt;
383 		c->ohead_lnum = c->orph_first;
384 		c->ohead_offs = 0;
385 	} else {
386 		/*
387 		 * We limit the number of orphans so that this should
388 		 * never happen.
389 		 */
390 		ubifs_err(c, "out of space in orphan area");
391 		err = -EINVAL;
392 	}
393 	spin_unlock(&c->orphan_lock);
394 	return err;
395 }
396 
397 /**
398  * commit_orphans - commit orphans.
399  * @c: UBIFS file-system description object
400  *
401  * This function commits orphans to flash. On success, %0 is returned,
402  * otherwise a negative error code is returned.
403  */
404 static int commit_orphans(struct ubifs_info *c)
405 {
406 	int avail, atomic = 0, err;
407 
408 	ubifs_assert(c->cmt_orphans > 0);
409 	avail = avail_orphs(c);
410 	if (avail < c->cmt_orphans) {
411 		/* Not enough space to write new orphans, so consolidate */
412 		err = consolidate(c);
413 		if (err)
414 			return err;
415 		atomic = 1;
416 	}
417 	err = write_orph_nodes(c, atomic);
418 	return err;
419 }
420 
421 /**
422  * erase_deleted - erase the orphans marked for deletion.
423  * @c: UBIFS file-system description object
424  *
425  * During commit, the orphans being committed cannot be deleted, so they are
426  * marked for deletion and deleted by this function. Also, the recovery
427  * adds killed orphans to the deletion list, and therefore they are deleted
428  * here too.
429  */
430 static void erase_deleted(struct ubifs_info *c)
431 {
432 	struct ubifs_orphan *orphan, *dnext;
433 
434 	spin_lock(&c->orphan_lock);
435 	dnext = c->orph_dnext;
436 	while (dnext) {
437 		orphan = dnext;
438 		dnext = orphan->dnext;
439 		ubifs_assert(!orphan->new);
440 		ubifs_assert(orphan->del);
441 		rb_erase(&orphan->rb, &c->orph_tree);
442 		list_del(&orphan->list);
443 		c->tot_orphans -= 1;
444 		dbg_gen("deleting orphan ino %lu", (unsigned long)orphan->inum);
445 		kfree(orphan);
446 	}
447 	c->orph_dnext = NULL;
448 	spin_unlock(&c->orphan_lock);
449 }
450 
451 /**
452  * ubifs_orphan_end_commit - end commit of orphans.
453  * @c: UBIFS file-system description object
454  *
455  * End commit of orphans.
456  */
457 int ubifs_orphan_end_commit(struct ubifs_info *c)
458 {
459 	int err;
460 
461 	if (c->cmt_orphans != 0) {
462 		err = commit_orphans(c);
463 		if (err)
464 			return err;
465 	}
466 	erase_deleted(c);
467 	err = dbg_check_orphans(c);
468 	return err;
469 }
470 
471 /**
472  * ubifs_clear_orphans - erase all LEBs used for orphans.
473  * @c: UBIFS file-system description object
474  *
475  * If recovery is not required, then the orphans from the previous session
476  * are not needed. This function locates the LEBs used to record
477  * orphans, and un-maps them.
478  */
479 int ubifs_clear_orphans(struct ubifs_info *c)
480 {
481 	int lnum, err;
482 
483 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
484 		err = ubifs_leb_unmap(c, lnum);
485 		if (err)
486 			return err;
487 	}
488 	c->ohead_lnum = c->orph_first;
489 	c->ohead_offs = 0;
490 	return 0;
491 }
492 
493 /**
494  * insert_dead_orphan - insert an orphan.
495  * @c: UBIFS file-system description object
496  * @inum: orphan inode number
497  *
498  * This function is a helper to the 'do_kill_orphans()' function. The orphan
499  * must be kept until the next commit, so it is added to the rb-tree and the
500  * deletion list.
501  */
502 static int insert_dead_orphan(struct ubifs_info *c, ino_t inum)
503 {
504 	struct ubifs_orphan *orphan, *o;
505 	struct rb_node **p, *parent = NULL;
506 
507 	orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_KERNEL);
508 	if (!orphan)
509 		return -ENOMEM;
510 	orphan->inum = inum;
511 
512 	p = &c->orph_tree.rb_node;
513 	while (*p) {
514 		parent = *p;
515 		o = rb_entry(parent, struct ubifs_orphan, rb);
516 		if (inum < o->inum)
517 			p = &(*p)->rb_left;
518 		else if (inum > o->inum)
519 			p = &(*p)->rb_right;
520 		else {
521 			/* Already added - no problem */
522 			kfree(orphan);
523 			return 0;
524 		}
525 	}
526 	c->tot_orphans += 1;
527 	rb_link_node(&orphan->rb, parent, p);
528 	rb_insert_color(&orphan->rb, &c->orph_tree);
529 	list_add_tail(&orphan->list, &c->orph_list);
530 	orphan->del = 1;
531 	orphan->dnext = c->orph_dnext;
532 	c->orph_dnext = orphan;
533 	dbg_mnt("ino %lu, new %d, tot %d", (unsigned long)inum,
534 		c->new_orphans, c->tot_orphans);
535 	return 0;
536 }
537 
538 /**
539  * do_kill_orphans - remove orphan inodes from the index.
540  * @c: UBIFS file-system description object
541  * @sleb: scanned LEB
542  * @last_cmt_no: cmt_no of last orphan node read is passed and returned here
543  * @outofdate: whether the LEB is out of date is returned here
544  * @last_flagged: whether the end orphan node is encountered
545  *
546  * This function is a helper to the 'kill_orphans()' function. It goes through
547  * every orphan node in a LEB and for every inode number recorded, removes
548  * all keys for that inode from the TNC.
549  */
550 static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
551 			   unsigned long long *last_cmt_no, int *outofdate,
552 			   int *last_flagged)
553 {
554 	struct ubifs_scan_node *snod;
555 	struct ubifs_orph_node *orph;
556 	unsigned long long cmt_no;
557 	ino_t inum;
558 	int i, n, err, first = 1;
559 
560 	list_for_each_entry(snod, &sleb->nodes, list) {
561 		if (snod->type != UBIFS_ORPH_NODE) {
562 			ubifs_err(c, "invalid node type %d in orphan area at %d:%d",
563 				  snod->type, sleb->lnum, snod->offs);
564 			ubifs_dump_node(c, snod->node);
565 			return -EINVAL;
566 		}
567 
568 		orph = snod->node;
569 
570 		/* Check commit number */
571 		cmt_no = le64_to_cpu(orph->cmt_no) & LLONG_MAX;
572 		/*
573 		 * The commit number on the master node may be less, because
574 		 * of a failed commit. If there are several failed commits in a
575 		 * row, the commit number written on orphan nodes will continue
576 		 * to increase (because the commit number is adjusted here) even
577 		 * though the commit number on the master node stays the same
578 		 * because the master node has not been re-written.
579 		 */
580 		if (cmt_no > c->cmt_no)
581 			c->cmt_no = cmt_no;
582 		if (cmt_no < *last_cmt_no && *last_flagged) {
583 			/*
584 			 * The last orphan node had a higher commit number and
585 			 * was flagged as the last written for that commit
586 			 * number. That makes this orphan node, out of date.
587 			 */
588 			if (!first) {
589 				ubifs_err(c, "out of order commit number %llu in orphan node at %d:%d",
590 					  cmt_no, sleb->lnum, snod->offs);
591 				ubifs_dump_node(c, snod->node);
592 				return -EINVAL;
593 			}
594 			dbg_rcvry("out of date LEB %d", sleb->lnum);
595 			*outofdate = 1;
596 			return 0;
597 		}
598 
599 		if (first)
600 			first = 0;
601 
602 		n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
603 		for (i = 0; i < n; i++) {
604 			inum = le64_to_cpu(orph->inos[i]);
605 			dbg_rcvry("deleting orphaned inode %lu",
606 				  (unsigned long)inum);
607 			err = ubifs_tnc_remove_ino(c, inum);
608 			if (err)
609 				return err;
610 			err = insert_dead_orphan(c, inum);
611 			if (err)
612 				return err;
613 		}
614 
615 		*last_cmt_no = cmt_no;
616 		if (le64_to_cpu(orph->cmt_no) & (1ULL << 63)) {
617 			dbg_rcvry("last orph node for commit %llu at %d:%d",
618 				  cmt_no, sleb->lnum, snod->offs);
619 			*last_flagged = 1;
620 		} else
621 			*last_flagged = 0;
622 	}
623 
624 	return 0;
625 }
626 
627 /**
628  * kill_orphans - remove all orphan inodes from the index.
629  * @c: UBIFS file-system description object
630  *
631  * If recovery is required, then orphan inodes recorded during the previous
632  * session (which ended with an unclean unmount) must be deleted from the index.
633  * This is done by updating the TNC, but since the index is not updated until
634  * the next commit, the LEBs where the orphan information is recorded are not
635  * erased until the next commit.
636  */
637 static int kill_orphans(struct ubifs_info *c)
638 {
639 	unsigned long long last_cmt_no = 0;
640 	int lnum, err = 0, outofdate = 0, last_flagged = 0;
641 
642 	c->ohead_lnum = c->orph_first;
643 	c->ohead_offs = 0;
644 	/* Check no-orphans flag and skip this if no orphans */
645 	if (c->no_orphs) {
646 		dbg_rcvry("no orphans");
647 		return 0;
648 	}
649 	/*
650 	 * Orph nodes always start at c->orph_first and are written to each
651 	 * successive LEB in turn. Generally unused LEBs will have been unmapped
652 	 * but may contain out of date orphan nodes if the unmap didn't go
653 	 * through. In addition, the last orphan node written for each commit is
654 	 * marked (top bit of orph->cmt_no is set to 1). It is possible that
655 	 * there are orphan nodes from the next commit (i.e. the commit did not
656 	 * complete successfully). In that case, no orphans will have been lost
657 	 * due to the way that orphans are written, and any orphans added will
658 	 * be valid orphans anyway and so can be deleted.
659 	 */
660 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
661 		struct ubifs_scan_leb *sleb;
662 
663 		dbg_rcvry("LEB %d", lnum);
664 		sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1);
665 		if (IS_ERR(sleb)) {
666 			if (PTR_ERR(sleb) == -EUCLEAN)
667 				sleb = ubifs_recover_leb(c, lnum, 0,
668 							 c->sbuf, -1);
669 			if (IS_ERR(sleb)) {
670 				err = PTR_ERR(sleb);
671 				break;
672 			}
673 		}
674 		err = do_kill_orphans(c, sleb, &last_cmt_no, &outofdate,
675 				      &last_flagged);
676 		if (err || outofdate) {
677 			ubifs_scan_destroy(sleb);
678 			break;
679 		}
680 		if (sleb->endpt) {
681 			c->ohead_lnum = lnum;
682 			c->ohead_offs = sleb->endpt;
683 		}
684 		ubifs_scan_destroy(sleb);
685 	}
686 	return err;
687 }
688 
689 /**
690  * ubifs_mount_orphans - delete orphan inodes and erase LEBs that recorded them.
691  * @c: UBIFS file-system description object
692  * @unclean: indicates recovery from unclean unmount
693  * @read_only: indicates read only mount
694  *
695  * This function is called when mounting to erase orphans from the previous
696  * session. If UBIFS was not unmounted cleanly, then the inodes recorded as
697  * orphans are deleted.
698  */
699 int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only)
700 {
701 	int err = 0;
702 
703 	c->max_orphans = tot_avail_orphs(c);
704 
705 	if (!read_only) {
706 		c->orph_buf = vmalloc(c->leb_size);
707 		if (!c->orph_buf)
708 			return -ENOMEM;
709 	}
710 
711 	if (unclean)
712 		err = kill_orphans(c);
713 	else if (!read_only)
714 		err = ubifs_clear_orphans(c);
715 
716 	return err;
717 }
718 
719 /*
720  * Everything below is related to debugging.
721  */
722 
723 struct check_orphan {
724 	struct rb_node rb;
725 	ino_t inum;
726 };
727 
728 struct check_info {
729 	unsigned long last_ino;
730 	unsigned long tot_inos;
731 	unsigned long missing;
732 	unsigned long long leaf_cnt;
733 	struct ubifs_ino_node *node;
734 	struct rb_root root;
735 };
736 
737 static int dbg_find_orphan(struct ubifs_info *c, ino_t inum)
738 {
739 	struct ubifs_orphan *o;
740 	struct rb_node *p;
741 
742 	spin_lock(&c->orphan_lock);
743 	p = c->orph_tree.rb_node;
744 	while (p) {
745 		o = rb_entry(p, struct ubifs_orphan, rb);
746 		if (inum < o->inum)
747 			p = p->rb_left;
748 		else if (inum > o->inum)
749 			p = p->rb_right;
750 		else {
751 			spin_unlock(&c->orphan_lock);
752 			return 1;
753 		}
754 	}
755 	spin_unlock(&c->orphan_lock);
756 	return 0;
757 }
758 
759 static int dbg_ins_check_orphan(struct rb_root *root, ino_t inum)
760 {
761 	struct check_orphan *orphan, *o;
762 	struct rb_node **p, *parent = NULL;
763 
764 	orphan = kzalloc(sizeof(struct check_orphan), GFP_NOFS);
765 	if (!orphan)
766 		return -ENOMEM;
767 	orphan->inum = inum;
768 
769 	p = &root->rb_node;
770 	while (*p) {
771 		parent = *p;
772 		o = rb_entry(parent, struct check_orphan, rb);
773 		if (inum < o->inum)
774 			p = &(*p)->rb_left;
775 		else if (inum > o->inum)
776 			p = &(*p)->rb_right;
777 		else {
778 			kfree(orphan);
779 			return 0;
780 		}
781 	}
782 	rb_link_node(&orphan->rb, parent, p);
783 	rb_insert_color(&orphan->rb, root);
784 	return 0;
785 }
786 
787 static int dbg_find_check_orphan(struct rb_root *root, ino_t inum)
788 {
789 	struct check_orphan *o;
790 	struct rb_node *p;
791 
792 	p = root->rb_node;
793 	while (p) {
794 		o = rb_entry(p, struct check_orphan, rb);
795 		if (inum < o->inum)
796 			p = p->rb_left;
797 		else if (inum > o->inum)
798 			p = p->rb_right;
799 		else
800 			return 1;
801 	}
802 	return 0;
803 }
804 
805 static void dbg_free_check_tree(struct rb_root *root)
806 {
807 	struct check_orphan *o, *n;
808 
809 	rbtree_postorder_for_each_entry_safe(o, n, root, rb)
810 		kfree(o);
811 }
812 
813 static int dbg_orphan_check(struct ubifs_info *c, struct ubifs_zbranch *zbr,
814 			    void *priv)
815 {
816 	struct check_info *ci = priv;
817 	ino_t inum;
818 	int err;
819 
820 	inum = key_inum(c, &zbr->key);
821 	if (inum != ci->last_ino) {
822 		/* Lowest node type is the inode node, so it comes first */
823 		if (key_type(c, &zbr->key) != UBIFS_INO_KEY)
824 			ubifs_err(c, "found orphan node ino %lu, type %d",
825 				  (unsigned long)inum, key_type(c, &zbr->key));
826 		ci->last_ino = inum;
827 		ci->tot_inos += 1;
828 		err = ubifs_tnc_read_node(c, zbr, ci->node);
829 		if (err) {
830 			ubifs_err(c, "node read failed, error %d", err);
831 			return err;
832 		}
833 		if (ci->node->nlink == 0)
834 			/* Must be recorded as an orphan */
835 			if (!dbg_find_check_orphan(&ci->root, inum) &&
836 			    !dbg_find_orphan(c, inum)) {
837 				ubifs_err(c, "missing orphan, ino %lu",
838 					  (unsigned long)inum);
839 				ci->missing += 1;
840 			}
841 	}
842 	ci->leaf_cnt += 1;
843 	return 0;
844 }
845 
846 static int dbg_read_orphans(struct check_info *ci, struct ubifs_scan_leb *sleb)
847 {
848 	struct ubifs_scan_node *snod;
849 	struct ubifs_orph_node *orph;
850 	ino_t inum;
851 	int i, n, err;
852 
853 	list_for_each_entry(snod, &sleb->nodes, list) {
854 		cond_resched();
855 		if (snod->type != UBIFS_ORPH_NODE)
856 			continue;
857 		orph = snod->node;
858 		n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
859 		for (i = 0; i < n; i++) {
860 			inum = le64_to_cpu(orph->inos[i]);
861 			err = dbg_ins_check_orphan(&ci->root, inum);
862 			if (err)
863 				return err;
864 		}
865 	}
866 	return 0;
867 }
868 
869 static int dbg_scan_orphans(struct ubifs_info *c, struct check_info *ci)
870 {
871 	int lnum, err = 0;
872 	void *buf;
873 
874 	/* Check no-orphans flag and skip this if no orphans */
875 	if (c->no_orphs)
876 		return 0;
877 
878 	buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
879 	if (!buf) {
880 		ubifs_err(c, "cannot allocate memory to check orphans");
881 		return 0;
882 	}
883 
884 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
885 		struct ubifs_scan_leb *sleb;
886 
887 		sleb = ubifs_scan(c, lnum, 0, buf, 0);
888 		if (IS_ERR(sleb)) {
889 			err = PTR_ERR(sleb);
890 			break;
891 		}
892 
893 		err = dbg_read_orphans(ci, sleb);
894 		ubifs_scan_destroy(sleb);
895 		if (err)
896 			break;
897 	}
898 
899 	vfree(buf);
900 	return err;
901 }
902 
903 static int dbg_check_orphans(struct ubifs_info *c)
904 {
905 	struct check_info ci;
906 	int err;
907 
908 	if (!dbg_is_chk_orph(c))
909 		return 0;
910 
911 	ci.last_ino = 0;
912 	ci.tot_inos = 0;
913 	ci.missing  = 0;
914 	ci.leaf_cnt = 0;
915 	ci.root = RB_ROOT;
916 	ci.node = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
917 	if (!ci.node) {
918 		ubifs_err(c, "out of memory");
919 		return -ENOMEM;
920 	}
921 
922 	err = dbg_scan_orphans(c, &ci);
923 	if (err)
924 		goto out;
925 
926 	err = dbg_walk_index(c, &dbg_orphan_check, NULL, &ci);
927 	if (err) {
928 		ubifs_err(c, "cannot scan TNC, error %d", err);
929 		goto out;
930 	}
931 
932 	if (ci.missing) {
933 		ubifs_err(c, "%lu missing orphan(s)", ci.missing);
934 		err = -EINVAL;
935 		goto out;
936 	}
937 
938 	dbg_cmt("last inode number is %lu", ci.last_ino);
939 	dbg_cmt("total number of inodes is %lu", ci.tot_inos);
940 	dbg_cmt("total number of leaf nodes is %llu", ci.leaf_cnt);
941 
942 out:
943 	dbg_free_check_tree(&ci.root);
944 	kfree(ci.node);
945 	return err;
946 }
947