xref: /openbmc/u-boot/fs/ubifs/orphan.c (revision ae485b54)
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  * Author: Adrian Hunter
8  */
9 
10 #include <linux/err.h>
11 #include "ubifs.h"
12 
13 /*
14  * An orphan is an inode number whose inode node has been committed to the index
15  * with a link count of zero. That happens when an open file is deleted
16  * (unlinked) and then a commit is run. In the normal course of events the inode
17  * would be deleted when the file is closed. However in the case of an unclean
18  * unmount, orphans need to be accounted for. After an unclean unmount, the
19  * orphans' inodes must be deleted which means either scanning the entire index
20  * looking for them, or keeping a list on flash somewhere. This unit implements
21  * the latter approach.
22  *
23  * The orphan area is a fixed number of LEBs situated between the LPT area and
24  * the main area. The number of orphan area LEBs is specified when the file
25  * system is created. The minimum number is 1. The size of the orphan area
26  * should be so that it can hold the maximum number of orphans that are expected
27  * to ever exist at one time.
28  *
29  * The number of orphans that can fit in a LEB is:
30  *
31  *         (c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64)
32  *
33  * For example: a 15872 byte LEB can fit 1980 orphans so 1 LEB may be enough.
34  *
35  * Orphans are accumulated in a rb-tree. When an inode's link count drops to
36  * zero, the inode number is added to the rb-tree. It is removed from the tree
37  * when the inode is deleted.  Any new orphans that are in the orphan tree when
38  * the commit is run, are written to the orphan area in 1 or more orphan nodes.
39  * If the orphan area is full, it is consolidated to make space.  There is
40  * always enough space because validation prevents the user from creating more
41  * than the maximum number of orphans allowed.
42  */
43 
44 static int dbg_check_orphans(struct ubifs_info *c);
45 
46 /**
47  * ubifs_add_orphan - add an orphan.
48  * @c: UBIFS file-system description object
49  * @inum: orphan inode number
50  *
51  * Add an orphan. This function is called when an inodes link count drops to
52  * zero.
53  */
54 int ubifs_add_orphan(struct ubifs_info *c, ino_t inum)
55 {
56 	struct ubifs_orphan *orphan, *o;
57 	struct rb_node **p, *parent = NULL;
58 
59 	orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_NOFS);
60 	if (!orphan)
61 		return -ENOMEM;
62 	orphan->inum = inum;
63 	orphan->new = 1;
64 
65 	spin_lock(&c->orphan_lock);
66 	if (c->tot_orphans >= c->max_orphans) {
67 		spin_unlock(&c->orphan_lock);
68 		kfree(orphan);
69 		return -ENFILE;
70 	}
71 	p = &c->orph_tree.rb_node;
72 	while (*p) {
73 		parent = *p;
74 		o = rb_entry(parent, struct ubifs_orphan, rb);
75 		if (inum < o->inum)
76 			p = &(*p)->rb_left;
77 		else if (inum > o->inum)
78 			p = &(*p)->rb_right;
79 		else {
80 			ubifs_err(c, "orphaned twice");
81 			spin_unlock(&c->orphan_lock);
82 			kfree(orphan);
83 			return 0;
84 		}
85 	}
86 	c->tot_orphans += 1;
87 	c->new_orphans += 1;
88 	rb_link_node(&orphan->rb, parent, p);
89 	rb_insert_color(&orphan->rb, &c->orph_tree);
90 	list_add_tail(&orphan->list, &c->orph_list);
91 	list_add_tail(&orphan->new_list, &c->orph_new);
92 	spin_unlock(&c->orphan_lock);
93 	dbg_gen("ino %lu", (unsigned long)inum);
94 	return 0;
95 }
96 
97 /**
98  * ubifs_delete_orphan - delete an orphan.
99  * @c: UBIFS file-system description object
100  * @inum: orphan inode number
101  *
102  * Delete an orphan. This function is called when an inode is deleted.
103  */
104 void ubifs_delete_orphan(struct ubifs_info *c, ino_t inum)
105 {
106 	struct ubifs_orphan *o;
107 	struct rb_node *p;
108 
109 	spin_lock(&c->orphan_lock);
110 	p = c->orph_tree.rb_node;
111 	while (p) {
112 		o = rb_entry(p, struct ubifs_orphan, rb);
113 		if (inum < o->inum)
114 			p = p->rb_left;
115 		else if (inum > o->inum)
116 			p = p->rb_right;
117 		else {
118 			if (o->del) {
119 				spin_unlock(&c->orphan_lock);
120 				dbg_gen("deleted twice ino %lu",
121 					(unsigned long)inum);
122 				return;
123 			}
124 			if (o->cmt) {
125 				o->del = 1;
126 				o->dnext = c->orph_dnext;
127 				c->orph_dnext = o;
128 				spin_unlock(&c->orphan_lock);
129 				dbg_gen("delete later ino %lu",
130 					(unsigned long)inum);
131 				return;
132 			}
133 			rb_erase(p, &c->orph_tree);
134 			list_del(&o->list);
135 			c->tot_orphans -= 1;
136 			if (o->new) {
137 				list_del(&o->new_list);
138 				c->new_orphans -= 1;
139 			}
140 			spin_unlock(&c->orphan_lock);
141 			kfree(o);
142 			dbg_gen("inum %lu", (unsigned long)inum);
143 			return;
144 		}
145 	}
146 	spin_unlock(&c->orphan_lock);
147 	ubifs_err(c, "missing orphan ino %lu", (unsigned long)inum);
148 	dump_stack();
149 }
150 
151 /**
152  * ubifs_orphan_start_commit - start commit of orphans.
153  * @c: UBIFS file-system description object
154  *
155  * Start commit of orphans.
156  */
157 int ubifs_orphan_start_commit(struct ubifs_info *c)
158 {
159 	struct ubifs_orphan *orphan, **last;
160 
161 	spin_lock(&c->orphan_lock);
162 	last = &c->orph_cnext;
163 	list_for_each_entry(orphan, &c->orph_new, new_list) {
164 		ubifs_assert(orphan->new);
165 		ubifs_assert(!orphan->cmt);
166 		orphan->new = 0;
167 		orphan->cmt = 1;
168 		*last = orphan;
169 		last = &orphan->cnext;
170 	}
171 	*last = NULL;
172 	c->cmt_orphans = c->new_orphans;
173 	c->new_orphans = 0;
174 	dbg_cmt("%d orphans to commit", c->cmt_orphans);
175 	INIT_LIST_HEAD(&c->orph_new);
176 	if (c->tot_orphans == 0)
177 		c->no_orphs = 1;
178 	else
179 		c->no_orphs = 0;
180 	spin_unlock(&c->orphan_lock);
181 	return 0;
182 }
183 
184 /**
185  * avail_orphs - calculate available space.
186  * @c: UBIFS file-system description object
187  *
188  * This function returns the number of orphans that can be written in the
189  * available space.
190  */
191 static int avail_orphs(struct ubifs_info *c)
192 {
193 	int avail_lebs, avail, gap;
194 
195 	avail_lebs = c->orph_lebs - (c->ohead_lnum - c->orph_first) - 1;
196 	avail = avail_lebs *
197 	       ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
198 	gap = c->leb_size - c->ohead_offs;
199 	if (gap >= UBIFS_ORPH_NODE_SZ + sizeof(__le64))
200 		avail += (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
201 	return avail;
202 }
203 
204 /**
205  * tot_avail_orphs - calculate total space.
206  * @c: UBIFS file-system description object
207  *
208  * This function returns the number of orphans that can be written in half
209  * the total space. That leaves half the space for adding new orphans.
210  */
211 static int tot_avail_orphs(struct ubifs_info *c)
212 {
213 	int avail_lebs, avail;
214 
215 	avail_lebs = c->orph_lebs;
216 	avail = avail_lebs *
217 	       ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
218 	return avail / 2;
219 }
220 
221 /**
222  * do_write_orph_node - write a node to the orphan head.
223  * @c: UBIFS file-system description object
224  * @len: length of node
225  * @atomic: write atomically
226  *
227  * This function writes a node to the orphan head from the orphan buffer. If
228  * %atomic is not zero, then the write is done atomically. On success, %0 is
229  * returned, otherwise a negative error code is returned.
230  */
231 static int do_write_orph_node(struct ubifs_info *c, int len, int atomic)
232 {
233 	int err = 0;
234 
235 	if (atomic) {
236 		ubifs_assert(c->ohead_offs == 0);
237 		ubifs_prepare_node(c, c->orph_buf, len, 1);
238 		len = ALIGN(len, c->min_io_size);
239 		err = ubifs_leb_change(c, c->ohead_lnum, c->orph_buf, len);
240 	} else {
241 		if (c->ohead_offs == 0) {
242 			/* Ensure LEB has been unmapped */
243 			err = ubifs_leb_unmap(c, c->ohead_lnum);
244 			if (err)
245 				return err;
246 		}
247 		err = ubifs_write_node(c, c->orph_buf, len, c->ohead_lnum,
248 				       c->ohead_offs);
249 	}
250 	return err;
251 }
252 
253 /**
254  * write_orph_node - write an orphan node.
255  * @c: UBIFS file-system description object
256  * @atomic: write atomically
257  *
258  * This function builds an orphan node from the cnext list and writes it to the
259  * orphan head. On success, %0 is returned, otherwise a negative error code
260  * is returned.
261  */
262 static int write_orph_node(struct ubifs_info *c, int atomic)
263 {
264 	struct ubifs_orphan *orphan, *cnext;
265 	struct ubifs_orph_node *orph;
266 	int gap, err, len, cnt, i;
267 
268 	ubifs_assert(c->cmt_orphans > 0);
269 	gap = c->leb_size - c->ohead_offs;
270 	if (gap < UBIFS_ORPH_NODE_SZ + sizeof(__le64)) {
271 		c->ohead_lnum += 1;
272 		c->ohead_offs = 0;
273 		gap = c->leb_size;
274 		if (c->ohead_lnum > c->orph_last) {
275 			/*
276 			 * We limit the number of orphans so that this should
277 			 * never happen.
278 			 */
279 			ubifs_err(c, "out of space in orphan area");
280 			return -EINVAL;
281 		}
282 	}
283 	cnt = (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
284 	if (cnt > c->cmt_orphans)
285 		cnt = c->cmt_orphans;
286 	len = UBIFS_ORPH_NODE_SZ + cnt * sizeof(__le64);
287 	ubifs_assert(c->orph_buf);
288 	orph = c->orph_buf;
289 	orph->ch.node_type = UBIFS_ORPH_NODE;
290 	spin_lock(&c->orphan_lock);
291 	cnext = c->orph_cnext;
292 	for (i = 0; i < cnt; i++) {
293 		orphan = cnext;
294 		ubifs_assert(orphan->cmt);
295 		orph->inos[i] = cpu_to_le64(orphan->inum);
296 		orphan->cmt = 0;
297 		cnext = orphan->cnext;
298 		orphan->cnext = NULL;
299 	}
300 	c->orph_cnext = cnext;
301 	c->cmt_orphans -= cnt;
302 	spin_unlock(&c->orphan_lock);
303 	if (c->cmt_orphans)
304 		orph->cmt_no = cpu_to_le64(c->cmt_no);
305 	else
306 		/* Mark the last node of the commit */
307 		orph->cmt_no = cpu_to_le64((c->cmt_no) | (1ULL << 63));
308 	ubifs_assert(c->ohead_offs + len <= c->leb_size);
309 	ubifs_assert(c->ohead_lnum >= c->orph_first);
310 	ubifs_assert(c->ohead_lnum <= c->orph_last);
311 	err = do_write_orph_node(c, len, atomic);
312 	c->ohead_offs += ALIGN(len, c->min_io_size);
313 	c->ohead_offs = ALIGN(c->ohead_offs, 8);
314 	return err;
315 }
316 
317 /**
318  * write_orph_nodes - write orphan nodes until there are no more to commit.
319  * @c: UBIFS file-system description object
320  * @atomic: write atomically
321  *
322  * This function writes orphan nodes for all the orphans to commit. On success,
323  * %0 is returned, otherwise a negative error code is returned.
324  */
325 static int write_orph_nodes(struct ubifs_info *c, int atomic)
326 {
327 	int err;
328 
329 	while (c->cmt_orphans > 0) {
330 		err = write_orph_node(c, atomic);
331 		if (err)
332 			return err;
333 	}
334 	if (atomic) {
335 		int lnum;
336 
337 		/* Unmap any unused LEBs after consolidation */
338 		for (lnum = c->ohead_lnum + 1; lnum <= c->orph_last; lnum++) {
339 			err = ubifs_leb_unmap(c, lnum);
340 			if (err)
341 				return err;
342 		}
343 	}
344 	return 0;
345 }
346 
347 /**
348  * consolidate - consolidate the orphan area.
349  * @c: UBIFS file-system description object
350  *
351  * This function enables consolidation by putting all the orphans into the list
352  * to commit. The list is in the order that the orphans were added, and the
353  * LEBs are written atomically in order, so at no time can orphans be lost by
354  * an unclean unmount.
355  *
356  * This function returns %0 on success and a negative error code on failure.
357  */
358 static int consolidate(struct ubifs_info *c)
359 {
360 	int tot_avail = tot_avail_orphs(c), err = 0;
361 
362 	spin_lock(&c->orphan_lock);
363 	dbg_cmt("there is space for %d orphans and there are %d",
364 		tot_avail, c->tot_orphans);
365 	if (c->tot_orphans - c->new_orphans <= tot_avail) {
366 		struct ubifs_orphan *orphan, **last;
367 		int cnt = 0;
368 
369 		/* Change the cnext list to include all non-new orphans */
370 		last = &c->orph_cnext;
371 		list_for_each_entry(orphan, &c->orph_list, list) {
372 			if (orphan->new)
373 				continue;
374 			orphan->cmt = 1;
375 			*last = orphan;
376 			last = &orphan->cnext;
377 			cnt += 1;
378 		}
379 		*last = NULL;
380 		ubifs_assert(cnt == c->tot_orphans - c->new_orphans);
381 		c->cmt_orphans = cnt;
382 		c->ohead_lnum = c->orph_first;
383 		c->ohead_offs = 0;
384 	} else {
385 		/*
386 		 * We limit the number of orphans so that this should
387 		 * never happen.
388 		 */
389 		ubifs_err(c, "out of space in orphan area");
390 		err = -EINVAL;
391 	}
392 	spin_unlock(&c->orphan_lock);
393 	return err;
394 }
395 
396 /**
397  * commit_orphans - commit orphans.
398  * @c: UBIFS file-system description object
399  *
400  * This function commits orphans to flash. On success, %0 is returned,
401  * otherwise a negative error code is returned.
402  */
403 static int commit_orphans(struct ubifs_info *c)
404 {
405 	int avail, atomic = 0, err;
406 
407 	ubifs_assert(c->cmt_orphans > 0);
408 	avail = avail_orphs(c);
409 	if (avail < c->cmt_orphans) {
410 		/* Not enough space to write new orphans, so consolidate */
411 		err = consolidate(c);
412 		if (err)
413 			return err;
414 		atomic = 1;
415 	}
416 	err = write_orph_nodes(c, atomic);
417 	return err;
418 }
419 
420 /**
421  * erase_deleted - erase the orphans marked for deletion.
422  * @c: UBIFS file-system description object
423  *
424  * During commit, the orphans being committed cannot be deleted, so they are
425  * marked for deletion and deleted by this function. Also, the recovery
426  * adds killed orphans to the deletion list, and therefore they are deleted
427  * here too.
428  */
429 static void erase_deleted(struct ubifs_info *c)
430 {
431 	struct ubifs_orphan *orphan, *dnext;
432 
433 	spin_lock(&c->orphan_lock);
434 	dnext = c->orph_dnext;
435 	while (dnext) {
436 		orphan = dnext;
437 		dnext = orphan->dnext;
438 		ubifs_assert(!orphan->new);
439 		ubifs_assert(orphan->del);
440 		rb_erase(&orphan->rb, &c->orph_tree);
441 		list_del(&orphan->list);
442 		c->tot_orphans -= 1;
443 		dbg_gen("deleting orphan ino %lu", (unsigned long)orphan->inum);
444 		kfree(orphan);
445 	}
446 	c->orph_dnext = NULL;
447 	spin_unlock(&c->orphan_lock);
448 }
449 
450 /**
451  * ubifs_orphan_end_commit - end commit of orphans.
452  * @c: UBIFS file-system description object
453  *
454  * End commit of orphans.
455  */
456 int ubifs_orphan_end_commit(struct ubifs_info *c)
457 {
458 	int err;
459 
460 	if (c->cmt_orphans != 0) {
461 		err = commit_orphans(c);
462 		if (err)
463 			return err;
464 	}
465 	erase_deleted(c);
466 	err = dbg_check_orphans(c);
467 	return err;
468 }
469 
470 /**
471  * ubifs_clear_orphans - erase all LEBs used for orphans.
472  * @c: UBIFS file-system description object
473  *
474  * If recovery is not required, then the orphans from the previous session
475  * are not needed. This function locates the LEBs used to record
476  * orphans, and un-maps them.
477  */
478 int ubifs_clear_orphans(struct ubifs_info *c)
479 {
480 	int lnum, err;
481 
482 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
483 		err = ubifs_leb_unmap(c, lnum);
484 		if (err)
485 			return err;
486 	}
487 	c->ohead_lnum = c->orph_first;
488 	c->ohead_offs = 0;
489 	return 0;
490 }
491 
492 /**
493  * insert_dead_orphan - insert an orphan.
494  * @c: UBIFS file-system description object
495  * @inum: orphan inode number
496  *
497  * This function is a helper to the 'do_kill_orphans()' function. The orphan
498  * must be kept until the next commit, so it is added to the rb-tree and the
499  * deletion list.
500  */
501 static int insert_dead_orphan(struct ubifs_info *c, ino_t inum)
502 {
503 	struct ubifs_orphan *orphan, *o;
504 	struct rb_node **p, *parent = NULL;
505 
506 	orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_KERNEL);
507 	if (!orphan)
508 		return -ENOMEM;
509 	orphan->inum = inum;
510 
511 	p = &c->orph_tree.rb_node;
512 	while (*p) {
513 		parent = *p;
514 		o = rb_entry(parent, struct ubifs_orphan, rb);
515 		if (inum < o->inum)
516 			p = &(*p)->rb_left;
517 		else if (inum > o->inum)
518 			p = &(*p)->rb_right;
519 		else {
520 			/* Already added - no problem */
521 			kfree(orphan);
522 			return 0;
523 		}
524 	}
525 	c->tot_orphans += 1;
526 	rb_link_node(&orphan->rb, parent, p);
527 	rb_insert_color(&orphan->rb, &c->orph_tree);
528 	list_add_tail(&orphan->list, &c->orph_list);
529 	orphan->del = 1;
530 	orphan->dnext = c->orph_dnext;
531 	c->orph_dnext = orphan;
532 	dbg_mnt("ino %lu, new %d, tot %d", (unsigned long)inum,
533 		c->new_orphans, c->tot_orphans);
534 	return 0;
535 }
536 
537 /**
538  * do_kill_orphans - remove orphan inodes from the index.
539  * @c: UBIFS file-system description object
540  * @sleb: scanned LEB
541  * @last_cmt_no: cmt_no of last orphan node read is passed and returned here
542  * @outofdate: whether the LEB is out of date is returned here
543  * @last_flagged: whether the end orphan node is encountered
544  *
545  * This function is a helper to the 'kill_orphans()' function. It goes through
546  * every orphan node in a LEB and for every inode number recorded, removes
547  * all keys for that inode from the TNC.
548  */
549 static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
550 			   unsigned long long *last_cmt_no, int *outofdate,
551 			   int *last_flagged)
552 {
553 	struct ubifs_scan_node *snod;
554 	struct ubifs_orph_node *orph;
555 	unsigned long long cmt_no;
556 	ino_t inum;
557 	int i, n, err, first = 1;
558 
559 	list_for_each_entry(snod, &sleb->nodes, list) {
560 		if (snod->type != UBIFS_ORPH_NODE) {
561 			ubifs_err(c, "invalid node type %d in orphan area at %d:%d",
562 				  snod->type, sleb->lnum, snod->offs);
563 			ubifs_dump_node(c, snod->node);
564 			return -EINVAL;
565 		}
566 
567 		orph = snod->node;
568 
569 		/* Check commit number */
570 		cmt_no = le64_to_cpu(orph->cmt_no) & LLONG_MAX;
571 		/*
572 		 * The commit number on the master node may be less, because
573 		 * of a failed commit. If there are several failed commits in a
574 		 * row, the commit number written on orphan nodes will continue
575 		 * to increase (because the commit number is adjusted here) even
576 		 * though the commit number on the master node stays the same
577 		 * because the master node has not been re-written.
578 		 */
579 		if (cmt_no > c->cmt_no)
580 			c->cmt_no = cmt_no;
581 		if (cmt_no < *last_cmt_no && *last_flagged) {
582 			/*
583 			 * The last orphan node had a higher commit number and
584 			 * was flagged as the last written for that commit
585 			 * number. That makes this orphan node, out of date.
586 			 */
587 			if (!first) {
588 				ubifs_err(c, "out of order commit number %llu in orphan node at %d:%d",
589 					  cmt_no, sleb->lnum, snod->offs);
590 				ubifs_dump_node(c, snod->node);
591 				return -EINVAL;
592 			}
593 			dbg_rcvry("out of date LEB %d", sleb->lnum);
594 			*outofdate = 1;
595 			return 0;
596 		}
597 
598 		if (first)
599 			first = 0;
600 
601 		n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
602 		for (i = 0; i < n; i++) {
603 			inum = le64_to_cpu(orph->inos[i]);
604 			dbg_rcvry("deleting orphaned inode %lu",
605 				  (unsigned long)inum);
606 			err = ubifs_tnc_remove_ino(c, inum);
607 			if (err)
608 				return err;
609 			err = insert_dead_orphan(c, inum);
610 			if (err)
611 				return err;
612 		}
613 
614 		*last_cmt_no = cmt_no;
615 		if (le64_to_cpu(orph->cmt_no) & (1ULL << 63)) {
616 			dbg_rcvry("last orph node for commit %llu at %d:%d",
617 				  cmt_no, sleb->lnum, snod->offs);
618 			*last_flagged = 1;
619 		} else
620 			*last_flagged = 0;
621 	}
622 
623 	return 0;
624 }
625 
626 /**
627  * kill_orphans - remove all orphan inodes from the index.
628  * @c: UBIFS file-system description object
629  *
630  * If recovery is required, then orphan inodes recorded during the previous
631  * session (which ended with an unclean unmount) must be deleted from the index.
632  * This is done by updating the TNC, but since the index is not updated until
633  * the next commit, the LEBs where the orphan information is recorded are not
634  * erased until the next commit.
635  */
636 static int kill_orphans(struct ubifs_info *c)
637 {
638 	unsigned long long last_cmt_no = 0;
639 	int lnum, err = 0, outofdate = 0, last_flagged = 0;
640 
641 	c->ohead_lnum = c->orph_first;
642 	c->ohead_offs = 0;
643 	/* Check no-orphans flag and skip this if no orphans */
644 	if (c->no_orphs) {
645 		dbg_rcvry("no orphans");
646 		return 0;
647 	}
648 	/*
649 	 * Orph nodes always start at c->orph_first and are written to each
650 	 * successive LEB in turn. Generally unused LEBs will have been unmapped
651 	 * but may contain out of date orphan nodes if the unmap didn't go
652 	 * through. In addition, the last orphan node written for each commit is
653 	 * marked (top bit of orph->cmt_no is set to 1). It is possible that
654 	 * there are orphan nodes from the next commit (i.e. the commit did not
655 	 * complete successfully). In that case, no orphans will have been lost
656 	 * due to the way that orphans are written, and any orphans added will
657 	 * be valid orphans anyway and so can be deleted.
658 	 */
659 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
660 		struct ubifs_scan_leb *sleb;
661 
662 		dbg_rcvry("LEB %d", lnum);
663 		sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1);
664 		if (IS_ERR(sleb)) {
665 			if (PTR_ERR(sleb) == -EUCLEAN)
666 				sleb = ubifs_recover_leb(c, lnum, 0,
667 							 c->sbuf, -1);
668 			if (IS_ERR(sleb)) {
669 				err = PTR_ERR(sleb);
670 				break;
671 			}
672 		}
673 		err = do_kill_orphans(c, sleb, &last_cmt_no, &outofdate,
674 				      &last_flagged);
675 		if (err || outofdate) {
676 			ubifs_scan_destroy(sleb);
677 			break;
678 		}
679 		if (sleb->endpt) {
680 			c->ohead_lnum = lnum;
681 			c->ohead_offs = sleb->endpt;
682 		}
683 		ubifs_scan_destroy(sleb);
684 	}
685 	return err;
686 }
687 
688 /**
689  * ubifs_mount_orphans - delete orphan inodes and erase LEBs that recorded them.
690  * @c: UBIFS file-system description object
691  * @unclean: indicates recovery from unclean unmount
692  * @read_only: indicates read only mount
693  *
694  * This function is called when mounting to erase orphans from the previous
695  * session. If UBIFS was not unmounted cleanly, then the inodes recorded as
696  * orphans are deleted.
697  */
698 int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only)
699 {
700 	int err = 0;
701 
702 	c->max_orphans = tot_avail_orphs(c);
703 
704 	if (!read_only) {
705 		c->orph_buf = vmalloc(c->leb_size);
706 		if (!c->orph_buf)
707 			return -ENOMEM;
708 	}
709 
710 	if (unclean)
711 		err = kill_orphans(c);
712 	else if (!read_only)
713 		err = ubifs_clear_orphans(c);
714 
715 	return err;
716 }
717 
718 /*
719  * Everything below is related to debugging.
720  */
721 
722 struct check_orphan {
723 	struct rb_node rb;
724 	ino_t inum;
725 };
726 
727 struct check_info {
728 	unsigned long last_ino;
729 	unsigned long tot_inos;
730 	unsigned long missing;
731 	unsigned long long leaf_cnt;
732 	struct ubifs_ino_node *node;
733 	struct rb_root root;
734 };
735 
736 static int dbg_find_orphan(struct ubifs_info *c, ino_t inum)
737 {
738 	struct ubifs_orphan *o;
739 	struct rb_node *p;
740 
741 	spin_lock(&c->orphan_lock);
742 	p = c->orph_tree.rb_node;
743 	while (p) {
744 		o = rb_entry(p, struct ubifs_orphan, rb);
745 		if (inum < o->inum)
746 			p = p->rb_left;
747 		else if (inum > o->inum)
748 			p = p->rb_right;
749 		else {
750 			spin_unlock(&c->orphan_lock);
751 			return 1;
752 		}
753 	}
754 	spin_unlock(&c->orphan_lock);
755 	return 0;
756 }
757 
758 static int dbg_ins_check_orphan(struct rb_root *root, ino_t inum)
759 {
760 	struct check_orphan *orphan, *o;
761 	struct rb_node **p, *parent = NULL;
762 
763 	orphan = kzalloc(sizeof(struct check_orphan), GFP_NOFS);
764 	if (!orphan)
765 		return -ENOMEM;
766 	orphan->inum = inum;
767 
768 	p = &root->rb_node;
769 	while (*p) {
770 		parent = *p;
771 		o = rb_entry(parent, struct check_orphan, rb);
772 		if (inum < o->inum)
773 			p = &(*p)->rb_left;
774 		else if (inum > o->inum)
775 			p = &(*p)->rb_right;
776 		else {
777 			kfree(orphan);
778 			return 0;
779 		}
780 	}
781 	rb_link_node(&orphan->rb, parent, p);
782 	rb_insert_color(&orphan->rb, root);
783 	return 0;
784 }
785 
786 static int dbg_find_check_orphan(struct rb_root *root, ino_t inum)
787 {
788 	struct check_orphan *o;
789 	struct rb_node *p;
790 
791 	p = root->rb_node;
792 	while (p) {
793 		o = rb_entry(p, struct check_orphan, rb);
794 		if (inum < o->inum)
795 			p = p->rb_left;
796 		else if (inum > o->inum)
797 			p = p->rb_right;
798 		else
799 			return 1;
800 	}
801 	return 0;
802 }
803 
804 static void dbg_free_check_tree(struct rb_root *root)
805 {
806 	struct check_orphan *o, *n;
807 
808 	rbtree_postorder_for_each_entry_safe(o, n, root, rb)
809 		kfree(o);
810 }
811 
812 static int dbg_orphan_check(struct ubifs_info *c, struct ubifs_zbranch *zbr,
813 			    void *priv)
814 {
815 	struct check_info *ci = priv;
816 	ino_t inum;
817 	int err;
818 
819 	inum = key_inum(c, &zbr->key);
820 	if (inum != ci->last_ino) {
821 		/* Lowest node type is the inode node, so it comes first */
822 		if (key_type(c, &zbr->key) != UBIFS_INO_KEY)
823 			ubifs_err(c, "found orphan node ino %lu, type %d",
824 				  (unsigned long)inum, key_type(c, &zbr->key));
825 		ci->last_ino = inum;
826 		ci->tot_inos += 1;
827 		err = ubifs_tnc_read_node(c, zbr, ci->node);
828 		if (err) {
829 			ubifs_err(c, "node read failed, error %d", err);
830 			return err;
831 		}
832 		if (ci->node->nlink == 0)
833 			/* Must be recorded as an orphan */
834 			if (!dbg_find_check_orphan(&ci->root, inum) &&
835 			    !dbg_find_orphan(c, inum)) {
836 				ubifs_err(c, "missing orphan, ino %lu",
837 					  (unsigned long)inum);
838 				ci->missing += 1;
839 			}
840 	}
841 	ci->leaf_cnt += 1;
842 	return 0;
843 }
844 
845 static int dbg_read_orphans(struct check_info *ci, struct ubifs_scan_leb *sleb)
846 {
847 	struct ubifs_scan_node *snod;
848 	struct ubifs_orph_node *orph;
849 	ino_t inum;
850 	int i, n, err;
851 
852 	list_for_each_entry(snod, &sleb->nodes, list) {
853 		cond_resched();
854 		if (snod->type != UBIFS_ORPH_NODE)
855 			continue;
856 		orph = snod->node;
857 		n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
858 		for (i = 0; i < n; i++) {
859 			inum = le64_to_cpu(orph->inos[i]);
860 			err = dbg_ins_check_orphan(&ci->root, inum);
861 			if (err)
862 				return err;
863 		}
864 	}
865 	return 0;
866 }
867 
868 static int dbg_scan_orphans(struct ubifs_info *c, struct check_info *ci)
869 {
870 	int lnum, err = 0;
871 	void *buf;
872 
873 	/* Check no-orphans flag and skip this if no orphans */
874 	if (c->no_orphs)
875 		return 0;
876 
877 	buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
878 	if (!buf) {
879 		ubifs_err(c, "cannot allocate memory to check orphans");
880 		return 0;
881 	}
882 
883 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
884 		struct ubifs_scan_leb *sleb;
885 
886 		sleb = ubifs_scan(c, lnum, 0, buf, 0);
887 		if (IS_ERR(sleb)) {
888 			err = PTR_ERR(sleb);
889 			break;
890 		}
891 
892 		err = dbg_read_orphans(ci, sleb);
893 		ubifs_scan_destroy(sleb);
894 		if (err)
895 			break;
896 	}
897 
898 	vfree(buf);
899 	return err;
900 }
901 
902 static int dbg_check_orphans(struct ubifs_info *c)
903 {
904 	struct check_info ci;
905 	int err;
906 
907 	if (!dbg_is_chk_orph(c))
908 		return 0;
909 
910 	ci.last_ino = 0;
911 	ci.tot_inos = 0;
912 	ci.missing  = 0;
913 	ci.leaf_cnt = 0;
914 	ci.root = RB_ROOT;
915 	ci.node = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
916 	if (!ci.node) {
917 		ubifs_err(c, "out of memory");
918 		return -ENOMEM;
919 	}
920 
921 	err = dbg_scan_orphans(c, &ci);
922 	if (err)
923 		goto out;
924 
925 	err = dbg_walk_index(c, &dbg_orphan_check, NULL, &ci);
926 	if (err) {
927 		ubifs_err(c, "cannot scan TNC, error %d", err);
928 		goto out;
929 	}
930 
931 	if (ci.missing) {
932 		ubifs_err(c, "%lu missing orphan(s)", ci.missing);
933 		err = -EINVAL;
934 		goto out;
935 	}
936 
937 	dbg_cmt("last inode number is %lu", ci.last_ino);
938 	dbg_cmt("total number of inodes is %lu", ci.tot_inos);
939 	dbg_cmt("total number of leaf nodes is %llu", ci.leaf_cnt);
940 
941 out:
942 	dbg_free_check_tree(&ci.root);
943 	kfree(ci.node);
944 	return err;
945 }
946