xref: /openbmc/linux/fs/ubifs/orphan.c (revision a50b854e)
1 // SPDX-License-Identifier: GPL-2.0-only
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 "ubifs.h"
11 
12 /*
13  * An orphan is an inode number whose inode node has been committed to the index
14  * with a link count of zero. That happens when an open file is deleted
15  * (unlinked) and then a commit is run. In the normal course of events the inode
16  * would be deleted when the file is closed. However in the case of an unclean
17  * unmount, orphans need to be accounted for. After an unclean unmount, the
18  * orphans' inodes must be deleted which means either scanning the entire index
19  * looking for them, or keeping a list on flash somewhere. This unit implements
20  * the latter approach.
21  *
22  * The orphan area is a fixed number of LEBs situated between the LPT area and
23  * the main area. The number of orphan area LEBs is specified when the file
24  * system is created. The minimum number is 1. The size of the orphan area
25  * should be so that it can hold the maximum number of orphans that are expected
26  * to ever exist at one time.
27  *
28  * The number of orphans that can fit in a LEB is:
29  *
30  *         (c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64)
31  *
32  * For example: a 15872 byte LEB can fit 1980 orphans so 1 LEB may be enough.
33  *
34  * Orphans are accumulated in a rb-tree. When an inode's link count drops to
35  * zero, the inode number is added to the rb-tree. It is removed from the tree
36  * when the inode is deleted.  Any new orphans that are in the orphan tree when
37  * the commit is run, are written to the orphan area in 1 or more orphan nodes.
38  * If the orphan area is full, it is consolidated to make space.  There is
39  * always enough space because validation prevents the user from creating more
40  * than the maximum number of orphans allowed.
41  */
42 
43 static int dbg_check_orphans(struct ubifs_info *c);
44 
45 static struct ubifs_orphan *orphan_add(struct ubifs_info *c, ino_t inum,
46 				       struct ubifs_orphan *parent_orphan)
47 {
48 	struct ubifs_orphan *orphan, *o;
49 	struct rb_node **p, *parent = NULL;
50 
51 	orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_NOFS);
52 	if (!orphan)
53 		return ERR_PTR(-ENOMEM);
54 	orphan->inum = inum;
55 	orphan->new = 1;
56 	INIT_LIST_HEAD(&orphan->child_list);
57 
58 	spin_lock(&c->orphan_lock);
59 	if (c->tot_orphans >= c->max_orphans) {
60 		spin_unlock(&c->orphan_lock);
61 		kfree(orphan);
62 		return ERR_PTR(-ENFILE);
63 	}
64 	p = &c->orph_tree.rb_node;
65 	while (*p) {
66 		parent = *p;
67 		o = rb_entry(parent, struct ubifs_orphan, rb);
68 		if (inum < o->inum)
69 			p = &(*p)->rb_left;
70 		else if (inum > o->inum)
71 			p = &(*p)->rb_right;
72 		else {
73 			ubifs_err(c, "orphaned twice");
74 			spin_unlock(&c->orphan_lock);
75 			kfree(orphan);
76 			return ERR_PTR(-EINVAL);
77 		}
78 	}
79 	c->tot_orphans += 1;
80 	c->new_orphans += 1;
81 	rb_link_node(&orphan->rb, parent, p);
82 	rb_insert_color(&orphan->rb, &c->orph_tree);
83 	list_add_tail(&orphan->list, &c->orph_list);
84 	list_add_tail(&orphan->new_list, &c->orph_new);
85 
86 	if (parent_orphan) {
87 		list_add_tail(&orphan->child_list,
88 			      &parent_orphan->child_list);
89 	}
90 
91 	spin_unlock(&c->orphan_lock);
92 	dbg_gen("ino %lu", (unsigned long)inum);
93 	return orphan;
94 }
95 
96 static struct ubifs_orphan *lookup_orphan(struct ubifs_info *c, ino_t inum)
97 {
98 	struct ubifs_orphan *o;
99 	struct rb_node *p;
100 
101 	p = c->orph_tree.rb_node;
102 	while (p) {
103 		o = rb_entry(p, struct ubifs_orphan, rb);
104 		if (inum < o->inum)
105 			p = p->rb_left;
106 		else if (inum > o->inum)
107 			p = p->rb_right;
108 		else {
109 			return o;
110 		}
111 	}
112 	return NULL;
113 }
114 
115 static void __orphan_drop(struct ubifs_info *c, struct ubifs_orphan *o)
116 {
117 	rb_erase(&o->rb, &c->orph_tree);
118 	list_del(&o->list);
119 	c->tot_orphans -= 1;
120 
121 	if (o->new) {
122 		list_del(&o->new_list);
123 		c->new_orphans -= 1;
124 	}
125 
126 	kfree(o);
127 }
128 
129 static void orphan_delete(struct ubifs_info *c, struct ubifs_orphan *orph)
130 {
131 	if (orph->del) {
132 		dbg_gen("deleted twice ino %lu", orph->inum);
133 		return;
134 	}
135 
136 	if (orph->cmt) {
137 		orph->del = 1;
138 		orph->dnext = c->orph_dnext;
139 		c->orph_dnext = orph;
140 		dbg_gen("delete later ino %lu", orph->inum);
141 		return;
142 	}
143 
144 	__orphan_drop(c, orph);
145 }
146 
147 /**
148  * ubifs_add_orphan - add an orphan.
149  * @c: UBIFS file-system description object
150  * @inum: orphan inode number
151  *
152  * Add an orphan. This function is called when an inodes link count drops to
153  * zero.
154  */
155 int ubifs_add_orphan(struct ubifs_info *c, ino_t inum)
156 {
157 	int err = 0;
158 	ino_t xattr_inum;
159 	union ubifs_key key;
160 	struct ubifs_dent_node *xent;
161 	struct fscrypt_name nm = {0};
162 	struct ubifs_orphan *xattr_orphan;
163 	struct ubifs_orphan *orphan;
164 
165 	orphan = orphan_add(c, inum, NULL);
166 	if (IS_ERR(orphan))
167 		return PTR_ERR(orphan);
168 
169 	lowest_xent_key(c, &key, inum);
170 	while (1) {
171 		xent = ubifs_tnc_next_ent(c, &key, &nm);
172 		if (IS_ERR(xent)) {
173 			err = PTR_ERR(xent);
174 			if (err == -ENOENT)
175 				break;
176 			return err;
177 		}
178 
179 		fname_name(&nm) = xent->name;
180 		fname_len(&nm) = le16_to_cpu(xent->nlen);
181 		xattr_inum = le64_to_cpu(xent->inum);
182 
183 		xattr_orphan = orphan_add(c, xattr_inum, orphan);
184 		if (IS_ERR(xattr_orphan))
185 			return PTR_ERR(xattr_orphan);
186 
187 		key_read(c, &xent->key, &key);
188 	}
189 
190 	return 0;
191 }
192 
193 /**
194  * ubifs_delete_orphan - delete an orphan.
195  * @c: UBIFS file-system description object
196  * @inum: orphan inode number
197  *
198  * Delete an orphan. This function is called when an inode is deleted.
199  */
200 void ubifs_delete_orphan(struct ubifs_info *c, ino_t inum)
201 {
202 	struct ubifs_orphan *orph, *child_orph, *tmp_o;
203 
204 	spin_lock(&c->orphan_lock);
205 
206 	orph = lookup_orphan(c, inum);
207 	if (!orph) {
208 		spin_unlock(&c->orphan_lock);
209 		ubifs_err(c, "missing orphan ino %lu", (unsigned long)inum);
210 		dump_stack();
211 
212 		return;
213 	}
214 
215 	list_for_each_entry_safe(child_orph, tmp_o, &orph->child_list, child_list) {
216 		list_del(&child_orph->child_list);
217 		orphan_delete(c, child_orph);
218 	}
219 
220 	orphan_delete(c, orph);
221 
222 	spin_unlock(&c->orphan_lock);
223 }
224 
225 /**
226  * ubifs_orphan_start_commit - start commit of orphans.
227  * @c: UBIFS file-system description object
228  *
229  * Start commit of orphans.
230  */
231 int ubifs_orphan_start_commit(struct ubifs_info *c)
232 {
233 	struct ubifs_orphan *orphan, **last;
234 
235 	spin_lock(&c->orphan_lock);
236 	last = &c->orph_cnext;
237 	list_for_each_entry(orphan, &c->orph_new, new_list) {
238 		ubifs_assert(c, orphan->new);
239 		ubifs_assert(c, !orphan->cmt);
240 		orphan->new = 0;
241 		orphan->cmt = 1;
242 		*last = orphan;
243 		last = &orphan->cnext;
244 	}
245 	*last = NULL;
246 	c->cmt_orphans = c->new_orphans;
247 	c->new_orphans = 0;
248 	dbg_cmt("%d orphans to commit", c->cmt_orphans);
249 	INIT_LIST_HEAD(&c->orph_new);
250 	if (c->tot_orphans == 0)
251 		c->no_orphs = 1;
252 	else
253 		c->no_orphs = 0;
254 	spin_unlock(&c->orphan_lock);
255 	return 0;
256 }
257 
258 /**
259  * avail_orphs - calculate available space.
260  * @c: UBIFS file-system description object
261  *
262  * This function returns the number of orphans that can be written in the
263  * available space.
264  */
265 static int avail_orphs(struct ubifs_info *c)
266 {
267 	int avail_lebs, avail, gap;
268 
269 	avail_lebs = c->orph_lebs - (c->ohead_lnum - c->orph_first) - 1;
270 	avail = avail_lebs *
271 	       ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
272 	gap = c->leb_size - c->ohead_offs;
273 	if (gap >= UBIFS_ORPH_NODE_SZ + sizeof(__le64))
274 		avail += (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
275 	return avail;
276 }
277 
278 /**
279  * tot_avail_orphs - calculate total space.
280  * @c: UBIFS file-system description object
281  *
282  * This function returns the number of orphans that can be written in half
283  * the total space. That leaves half the space for adding new orphans.
284  */
285 static int tot_avail_orphs(struct ubifs_info *c)
286 {
287 	int avail_lebs, avail;
288 
289 	avail_lebs = c->orph_lebs;
290 	avail = avail_lebs *
291 	       ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
292 	return avail / 2;
293 }
294 
295 /**
296  * do_write_orph_node - write a node to the orphan head.
297  * @c: UBIFS file-system description object
298  * @len: length of node
299  * @atomic: write atomically
300  *
301  * This function writes a node to the orphan head from the orphan buffer. If
302  * %atomic is not zero, then the write is done atomically. On success, %0 is
303  * returned, otherwise a negative error code is returned.
304  */
305 static int do_write_orph_node(struct ubifs_info *c, int len, int atomic)
306 {
307 	int err = 0;
308 
309 	if (atomic) {
310 		ubifs_assert(c, c->ohead_offs == 0);
311 		ubifs_prepare_node(c, c->orph_buf, len, 1);
312 		len = ALIGN(len, c->min_io_size);
313 		err = ubifs_leb_change(c, c->ohead_lnum, c->orph_buf, len);
314 	} else {
315 		if (c->ohead_offs == 0) {
316 			/* Ensure LEB has been unmapped */
317 			err = ubifs_leb_unmap(c, c->ohead_lnum);
318 			if (err)
319 				return err;
320 		}
321 		err = ubifs_write_node(c, c->orph_buf, len, c->ohead_lnum,
322 				       c->ohead_offs);
323 	}
324 	return err;
325 }
326 
327 /**
328  * write_orph_node - write an orphan node.
329  * @c: UBIFS file-system description object
330  * @atomic: write atomically
331  *
332  * This function builds an orphan node from the cnext list and writes it to the
333  * orphan head. On success, %0 is returned, otherwise a negative error code
334  * is returned.
335  */
336 static int write_orph_node(struct ubifs_info *c, int atomic)
337 {
338 	struct ubifs_orphan *orphan, *cnext;
339 	struct ubifs_orph_node *orph;
340 	int gap, err, len, cnt, i;
341 
342 	ubifs_assert(c, c->cmt_orphans > 0);
343 	gap = c->leb_size - c->ohead_offs;
344 	if (gap < UBIFS_ORPH_NODE_SZ + sizeof(__le64)) {
345 		c->ohead_lnum += 1;
346 		c->ohead_offs = 0;
347 		gap = c->leb_size;
348 		if (c->ohead_lnum > c->orph_last) {
349 			/*
350 			 * We limit the number of orphans so that this should
351 			 * never happen.
352 			 */
353 			ubifs_err(c, "out of space in orphan area");
354 			return -EINVAL;
355 		}
356 	}
357 	cnt = (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
358 	if (cnt > c->cmt_orphans)
359 		cnt = c->cmt_orphans;
360 	len = UBIFS_ORPH_NODE_SZ + cnt * sizeof(__le64);
361 	ubifs_assert(c, c->orph_buf);
362 	orph = c->orph_buf;
363 	orph->ch.node_type = UBIFS_ORPH_NODE;
364 	spin_lock(&c->orphan_lock);
365 	cnext = c->orph_cnext;
366 	for (i = 0; i < cnt; i++) {
367 		orphan = cnext;
368 		ubifs_assert(c, orphan->cmt);
369 		orph->inos[i] = cpu_to_le64(orphan->inum);
370 		orphan->cmt = 0;
371 		cnext = orphan->cnext;
372 		orphan->cnext = NULL;
373 	}
374 	c->orph_cnext = cnext;
375 	c->cmt_orphans -= cnt;
376 	spin_unlock(&c->orphan_lock);
377 	if (c->cmt_orphans)
378 		orph->cmt_no = cpu_to_le64(c->cmt_no);
379 	else
380 		/* Mark the last node of the commit */
381 		orph->cmt_no = cpu_to_le64((c->cmt_no) | (1ULL << 63));
382 	ubifs_assert(c, c->ohead_offs + len <= c->leb_size);
383 	ubifs_assert(c, c->ohead_lnum >= c->orph_first);
384 	ubifs_assert(c, c->ohead_lnum <= c->orph_last);
385 	err = do_write_orph_node(c, len, atomic);
386 	c->ohead_offs += ALIGN(len, c->min_io_size);
387 	c->ohead_offs = ALIGN(c->ohead_offs, 8);
388 	return err;
389 }
390 
391 /**
392  * write_orph_nodes - write orphan nodes until there are no more to commit.
393  * @c: UBIFS file-system description object
394  * @atomic: write atomically
395  *
396  * This function writes orphan nodes for all the orphans to commit. On success,
397  * %0 is returned, otherwise a negative error code is returned.
398  */
399 static int write_orph_nodes(struct ubifs_info *c, int atomic)
400 {
401 	int err;
402 
403 	while (c->cmt_orphans > 0) {
404 		err = write_orph_node(c, atomic);
405 		if (err)
406 			return err;
407 	}
408 	if (atomic) {
409 		int lnum;
410 
411 		/* Unmap any unused LEBs after consolidation */
412 		for (lnum = c->ohead_lnum + 1; lnum <= c->orph_last; lnum++) {
413 			err = ubifs_leb_unmap(c, lnum);
414 			if (err)
415 				return err;
416 		}
417 	}
418 	return 0;
419 }
420 
421 /**
422  * consolidate - consolidate the orphan area.
423  * @c: UBIFS file-system description object
424  *
425  * This function enables consolidation by putting all the orphans into the list
426  * to commit. The list is in the order that the orphans were added, and the
427  * LEBs are written atomically in order, so at no time can orphans be lost by
428  * an unclean unmount.
429  *
430  * This function returns %0 on success and a negative error code on failure.
431  */
432 static int consolidate(struct ubifs_info *c)
433 {
434 	int tot_avail = tot_avail_orphs(c), err = 0;
435 
436 	spin_lock(&c->orphan_lock);
437 	dbg_cmt("there is space for %d orphans and there are %d",
438 		tot_avail, c->tot_orphans);
439 	if (c->tot_orphans - c->new_orphans <= tot_avail) {
440 		struct ubifs_orphan *orphan, **last;
441 		int cnt = 0;
442 
443 		/* Change the cnext list to include all non-new orphans */
444 		last = &c->orph_cnext;
445 		list_for_each_entry(orphan, &c->orph_list, list) {
446 			if (orphan->new)
447 				continue;
448 			orphan->cmt = 1;
449 			*last = orphan;
450 			last = &orphan->cnext;
451 			cnt += 1;
452 		}
453 		*last = NULL;
454 		ubifs_assert(c, cnt == c->tot_orphans - c->new_orphans);
455 		c->cmt_orphans = cnt;
456 		c->ohead_lnum = c->orph_first;
457 		c->ohead_offs = 0;
458 	} else {
459 		/*
460 		 * We limit the number of orphans so that this should
461 		 * never happen.
462 		 */
463 		ubifs_err(c, "out of space in orphan area");
464 		err = -EINVAL;
465 	}
466 	spin_unlock(&c->orphan_lock);
467 	return err;
468 }
469 
470 /**
471  * commit_orphans - commit orphans.
472  * @c: UBIFS file-system description object
473  *
474  * This function commits orphans to flash. On success, %0 is returned,
475  * otherwise a negative error code is returned.
476  */
477 static int commit_orphans(struct ubifs_info *c)
478 {
479 	int avail, atomic = 0, err;
480 
481 	ubifs_assert(c, c->cmt_orphans > 0);
482 	avail = avail_orphs(c);
483 	if (avail < c->cmt_orphans) {
484 		/* Not enough space to write new orphans, so consolidate */
485 		err = consolidate(c);
486 		if (err)
487 			return err;
488 		atomic = 1;
489 	}
490 	err = write_orph_nodes(c, atomic);
491 	return err;
492 }
493 
494 /**
495  * erase_deleted - erase the orphans marked for deletion.
496  * @c: UBIFS file-system description object
497  *
498  * During commit, the orphans being committed cannot be deleted, so they are
499  * marked for deletion and deleted by this function. Also, the recovery
500  * adds killed orphans to the deletion list, and therefore they are deleted
501  * here too.
502  */
503 static void erase_deleted(struct ubifs_info *c)
504 {
505 	struct ubifs_orphan *orphan, *dnext;
506 
507 	spin_lock(&c->orphan_lock);
508 	dnext = c->orph_dnext;
509 	while (dnext) {
510 		orphan = dnext;
511 		dnext = orphan->dnext;
512 		ubifs_assert(c, !orphan->new);
513 		ubifs_assert(c, orphan->del);
514 		rb_erase(&orphan->rb, &c->orph_tree);
515 		list_del(&orphan->list);
516 		c->tot_orphans -= 1;
517 		dbg_gen("deleting orphan ino %lu", (unsigned long)orphan->inum);
518 		kfree(orphan);
519 	}
520 	c->orph_dnext = NULL;
521 	spin_unlock(&c->orphan_lock);
522 }
523 
524 /**
525  * ubifs_orphan_end_commit - end commit of orphans.
526  * @c: UBIFS file-system description object
527  *
528  * End commit of orphans.
529  */
530 int ubifs_orphan_end_commit(struct ubifs_info *c)
531 {
532 	int err;
533 
534 	if (c->cmt_orphans != 0) {
535 		err = commit_orphans(c);
536 		if (err)
537 			return err;
538 	}
539 	erase_deleted(c);
540 	err = dbg_check_orphans(c);
541 	return err;
542 }
543 
544 /**
545  * ubifs_clear_orphans - erase all LEBs used for orphans.
546  * @c: UBIFS file-system description object
547  *
548  * If recovery is not required, then the orphans from the previous session
549  * are not needed. This function locates the LEBs used to record
550  * orphans, and un-maps them.
551  */
552 int ubifs_clear_orphans(struct ubifs_info *c)
553 {
554 	int lnum, err;
555 
556 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
557 		err = ubifs_leb_unmap(c, lnum);
558 		if (err)
559 			return err;
560 	}
561 	c->ohead_lnum = c->orph_first;
562 	c->ohead_offs = 0;
563 	return 0;
564 }
565 
566 /**
567  * insert_dead_orphan - insert an orphan.
568  * @c: UBIFS file-system description object
569  * @inum: orphan inode number
570  *
571  * This function is a helper to the 'do_kill_orphans()' function. The orphan
572  * must be kept until the next commit, so it is added to the rb-tree and the
573  * deletion list.
574  */
575 static int insert_dead_orphan(struct ubifs_info *c, ino_t inum)
576 {
577 	struct ubifs_orphan *orphan, *o;
578 	struct rb_node **p, *parent = NULL;
579 
580 	orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_KERNEL);
581 	if (!orphan)
582 		return -ENOMEM;
583 	orphan->inum = inum;
584 
585 	p = &c->orph_tree.rb_node;
586 	while (*p) {
587 		parent = *p;
588 		o = rb_entry(parent, struct ubifs_orphan, rb);
589 		if (inum < o->inum)
590 			p = &(*p)->rb_left;
591 		else if (inum > o->inum)
592 			p = &(*p)->rb_right;
593 		else {
594 			/* Already added - no problem */
595 			kfree(orphan);
596 			return 0;
597 		}
598 	}
599 	c->tot_orphans += 1;
600 	rb_link_node(&orphan->rb, parent, p);
601 	rb_insert_color(&orphan->rb, &c->orph_tree);
602 	list_add_tail(&orphan->list, &c->orph_list);
603 	orphan->del = 1;
604 	orphan->dnext = c->orph_dnext;
605 	c->orph_dnext = orphan;
606 	dbg_mnt("ino %lu, new %d, tot %d", (unsigned long)inum,
607 		c->new_orphans, c->tot_orphans);
608 	return 0;
609 }
610 
611 /**
612  * do_kill_orphans - remove orphan inodes from the index.
613  * @c: UBIFS file-system description object
614  * @sleb: scanned LEB
615  * @last_cmt_no: cmt_no of last orphan node read is passed and returned here
616  * @outofdate: whether the LEB is out of date is returned here
617  * @last_flagged: whether the end orphan node is encountered
618  *
619  * This function is a helper to the 'kill_orphans()' function. It goes through
620  * every orphan node in a LEB and for every inode number recorded, removes
621  * all keys for that inode from the TNC.
622  */
623 static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
624 			   unsigned long long *last_cmt_no, int *outofdate,
625 			   int *last_flagged)
626 {
627 	struct ubifs_scan_node *snod;
628 	struct ubifs_orph_node *orph;
629 	struct ubifs_ino_node *ino = NULL;
630 	unsigned long long cmt_no;
631 	ino_t inum;
632 	int i, n, err, first = 1;
633 
634 	list_for_each_entry(snod, &sleb->nodes, list) {
635 		if (snod->type != UBIFS_ORPH_NODE) {
636 			ubifs_err(c, "invalid node type %d in orphan area at %d:%d",
637 				  snod->type, sleb->lnum, snod->offs);
638 			ubifs_dump_node(c, snod->node);
639 			return -EINVAL;
640 		}
641 
642 		orph = snod->node;
643 
644 		/* Check commit number */
645 		cmt_no = le64_to_cpu(orph->cmt_no) & LLONG_MAX;
646 		/*
647 		 * The commit number on the master node may be less, because
648 		 * of a failed commit. If there are several failed commits in a
649 		 * row, the commit number written on orphan nodes will continue
650 		 * to increase (because the commit number is adjusted here) even
651 		 * though the commit number on the master node stays the same
652 		 * because the master node has not been re-written.
653 		 */
654 		if (cmt_no > c->cmt_no)
655 			c->cmt_no = cmt_no;
656 		if (cmt_no < *last_cmt_no && *last_flagged) {
657 			/*
658 			 * The last orphan node had a higher commit number and
659 			 * was flagged as the last written for that commit
660 			 * number. That makes this orphan node, out of date.
661 			 */
662 			if (!first) {
663 				ubifs_err(c, "out of order commit number %llu in orphan node at %d:%d",
664 					  cmt_no, sleb->lnum, snod->offs);
665 				ubifs_dump_node(c, snod->node);
666 				return -EINVAL;
667 			}
668 			dbg_rcvry("out of date LEB %d", sleb->lnum);
669 			*outofdate = 1;
670 			return 0;
671 		}
672 
673 		if (first)
674 			first = 0;
675 
676 		ino = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
677 		if (!ino)
678 			return -ENOMEM;
679 
680 		n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
681 		for (i = 0; i < n; i++) {
682 			union ubifs_key key1, key2;
683 
684 			inum = le64_to_cpu(orph->inos[i]);
685 
686 			ino_key_init(c, &key1, inum);
687 			err = ubifs_tnc_lookup(c, &key1, ino);
688 			if (err)
689 				goto out_free;
690 
691 			/*
692 			 * Check whether an inode can really get deleted.
693 			 * linkat() with O_TMPFILE allows rebirth of an inode.
694 			 */
695 			if (ino->nlink == 0) {
696 				dbg_rcvry("deleting orphaned inode %lu",
697 					  (unsigned long)inum);
698 
699 				lowest_ino_key(c, &key1, inum);
700 				highest_ino_key(c, &key2, inum);
701 
702 				err = ubifs_tnc_remove_range(c, &key1, &key2);
703 				if (err)
704 					goto out_ro;
705 			}
706 
707 			err = insert_dead_orphan(c, inum);
708 			if (err)
709 				goto out_free;
710 		}
711 
712 		*last_cmt_no = cmt_no;
713 		if (le64_to_cpu(orph->cmt_no) & (1ULL << 63)) {
714 			dbg_rcvry("last orph node for commit %llu at %d:%d",
715 				  cmt_no, sleb->lnum, snod->offs);
716 			*last_flagged = 1;
717 		} else
718 			*last_flagged = 0;
719 	}
720 
721 	err = 0;
722 out_free:
723 	kfree(ino);
724 	return err;
725 
726 out_ro:
727 	ubifs_ro_mode(c, err);
728 	kfree(ino);
729 	return err;
730 }
731 
732 /**
733  * kill_orphans - remove all orphan inodes from the index.
734  * @c: UBIFS file-system description object
735  *
736  * If recovery is required, then orphan inodes recorded during the previous
737  * session (which ended with an unclean unmount) must be deleted from the index.
738  * This is done by updating the TNC, but since the index is not updated until
739  * the next commit, the LEBs where the orphan information is recorded are not
740  * erased until the next commit.
741  */
742 static int kill_orphans(struct ubifs_info *c)
743 {
744 	unsigned long long last_cmt_no = 0;
745 	int lnum, err = 0, outofdate = 0, last_flagged = 0;
746 
747 	c->ohead_lnum = c->orph_first;
748 	c->ohead_offs = 0;
749 	/* Check no-orphans flag and skip this if no orphans */
750 	if (c->no_orphs) {
751 		dbg_rcvry("no orphans");
752 		return 0;
753 	}
754 	/*
755 	 * Orph nodes always start at c->orph_first and are written to each
756 	 * successive LEB in turn. Generally unused LEBs will have been unmapped
757 	 * but may contain out of date orphan nodes if the unmap didn't go
758 	 * through. In addition, the last orphan node written for each commit is
759 	 * marked (top bit of orph->cmt_no is set to 1). It is possible that
760 	 * there are orphan nodes from the next commit (i.e. the commit did not
761 	 * complete successfully). In that case, no orphans will have been lost
762 	 * due to the way that orphans are written, and any orphans added will
763 	 * be valid orphans anyway and so can be deleted.
764 	 */
765 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
766 		struct ubifs_scan_leb *sleb;
767 
768 		dbg_rcvry("LEB %d", lnum);
769 		sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1);
770 		if (IS_ERR(sleb)) {
771 			if (PTR_ERR(sleb) == -EUCLEAN)
772 				sleb = ubifs_recover_leb(c, lnum, 0,
773 							 c->sbuf, -1);
774 			if (IS_ERR(sleb)) {
775 				err = PTR_ERR(sleb);
776 				break;
777 			}
778 		}
779 		err = do_kill_orphans(c, sleb, &last_cmt_no, &outofdate,
780 				      &last_flagged);
781 		if (err || outofdate) {
782 			ubifs_scan_destroy(sleb);
783 			break;
784 		}
785 		if (sleb->endpt) {
786 			c->ohead_lnum = lnum;
787 			c->ohead_offs = sleb->endpt;
788 		}
789 		ubifs_scan_destroy(sleb);
790 	}
791 	return err;
792 }
793 
794 /**
795  * ubifs_mount_orphans - delete orphan inodes and erase LEBs that recorded them.
796  * @c: UBIFS file-system description object
797  * @unclean: indicates recovery from unclean unmount
798  * @read_only: indicates read only mount
799  *
800  * This function is called when mounting to erase orphans from the previous
801  * session. If UBIFS was not unmounted cleanly, then the inodes recorded as
802  * orphans are deleted.
803  */
804 int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only)
805 {
806 	int err = 0;
807 
808 	c->max_orphans = tot_avail_orphs(c);
809 
810 	if (!read_only) {
811 		c->orph_buf = vmalloc(c->leb_size);
812 		if (!c->orph_buf)
813 			return -ENOMEM;
814 	}
815 
816 	if (unclean)
817 		err = kill_orphans(c);
818 	else if (!read_only)
819 		err = ubifs_clear_orphans(c);
820 
821 	return err;
822 }
823 
824 /*
825  * Everything below is related to debugging.
826  */
827 
828 struct check_orphan {
829 	struct rb_node rb;
830 	ino_t inum;
831 };
832 
833 struct check_info {
834 	unsigned long last_ino;
835 	unsigned long tot_inos;
836 	unsigned long missing;
837 	unsigned long long leaf_cnt;
838 	struct ubifs_ino_node *node;
839 	struct rb_root root;
840 };
841 
842 static bool dbg_find_orphan(struct ubifs_info *c, ino_t inum)
843 {
844 	bool found = false;
845 
846 	spin_lock(&c->orphan_lock);
847 	found = !!lookup_orphan(c, inum);
848 	spin_unlock(&c->orphan_lock);
849 
850 	return found;
851 }
852 
853 static int dbg_ins_check_orphan(struct rb_root *root, ino_t inum)
854 {
855 	struct check_orphan *orphan, *o;
856 	struct rb_node **p, *parent = NULL;
857 
858 	orphan = kzalloc(sizeof(struct check_orphan), GFP_NOFS);
859 	if (!orphan)
860 		return -ENOMEM;
861 	orphan->inum = inum;
862 
863 	p = &root->rb_node;
864 	while (*p) {
865 		parent = *p;
866 		o = rb_entry(parent, struct check_orphan, rb);
867 		if (inum < o->inum)
868 			p = &(*p)->rb_left;
869 		else if (inum > o->inum)
870 			p = &(*p)->rb_right;
871 		else {
872 			kfree(orphan);
873 			return 0;
874 		}
875 	}
876 	rb_link_node(&orphan->rb, parent, p);
877 	rb_insert_color(&orphan->rb, root);
878 	return 0;
879 }
880 
881 static int dbg_find_check_orphan(struct rb_root *root, ino_t inum)
882 {
883 	struct check_orphan *o;
884 	struct rb_node *p;
885 
886 	p = root->rb_node;
887 	while (p) {
888 		o = rb_entry(p, struct check_orphan, rb);
889 		if (inum < o->inum)
890 			p = p->rb_left;
891 		else if (inum > o->inum)
892 			p = p->rb_right;
893 		else
894 			return 1;
895 	}
896 	return 0;
897 }
898 
899 static void dbg_free_check_tree(struct rb_root *root)
900 {
901 	struct check_orphan *o, *n;
902 
903 	rbtree_postorder_for_each_entry_safe(o, n, root, rb)
904 		kfree(o);
905 }
906 
907 static int dbg_orphan_check(struct ubifs_info *c, struct ubifs_zbranch *zbr,
908 			    void *priv)
909 {
910 	struct check_info *ci = priv;
911 	ino_t inum;
912 	int err;
913 
914 	inum = key_inum(c, &zbr->key);
915 	if (inum != ci->last_ino) {
916 		/* Lowest node type is the inode node, so it comes first */
917 		if (key_type(c, &zbr->key) != UBIFS_INO_KEY)
918 			ubifs_err(c, "found orphan node ino %lu, type %d",
919 				  (unsigned long)inum, key_type(c, &zbr->key));
920 		ci->last_ino = inum;
921 		ci->tot_inos += 1;
922 		err = ubifs_tnc_read_node(c, zbr, ci->node);
923 		if (err) {
924 			ubifs_err(c, "node read failed, error %d", err);
925 			return err;
926 		}
927 		if (ci->node->nlink == 0)
928 			/* Must be recorded as an orphan */
929 			if (!dbg_find_check_orphan(&ci->root, inum) &&
930 			    !dbg_find_orphan(c, inum)) {
931 				ubifs_err(c, "missing orphan, ino %lu",
932 					  (unsigned long)inum);
933 				ci->missing += 1;
934 			}
935 	}
936 	ci->leaf_cnt += 1;
937 	return 0;
938 }
939 
940 static int dbg_read_orphans(struct check_info *ci, struct ubifs_scan_leb *sleb)
941 {
942 	struct ubifs_scan_node *snod;
943 	struct ubifs_orph_node *orph;
944 	ino_t inum;
945 	int i, n, err;
946 
947 	list_for_each_entry(snod, &sleb->nodes, list) {
948 		cond_resched();
949 		if (snod->type != UBIFS_ORPH_NODE)
950 			continue;
951 		orph = snod->node;
952 		n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
953 		for (i = 0; i < n; i++) {
954 			inum = le64_to_cpu(orph->inos[i]);
955 			err = dbg_ins_check_orphan(&ci->root, inum);
956 			if (err)
957 				return err;
958 		}
959 	}
960 	return 0;
961 }
962 
963 static int dbg_scan_orphans(struct ubifs_info *c, struct check_info *ci)
964 {
965 	int lnum, err = 0;
966 	void *buf;
967 
968 	/* Check no-orphans flag and skip this if no orphans */
969 	if (c->no_orphs)
970 		return 0;
971 
972 	buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
973 	if (!buf) {
974 		ubifs_err(c, "cannot allocate memory to check orphans");
975 		return 0;
976 	}
977 
978 	for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
979 		struct ubifs_scan_leb *sleb;
980 
981 		sleb = ubifs_scan(c, lnum, 0, buf, 0);
982 		if (IS_ERR(sleb)) {
983 			err = PTR_ERR(sleb);
984 			break;
985 		}
986 
987 		err = dbg_read_orphans(ci, sleb);
988 		ubifs_scan_destroy(sleb);
989 		if (err)
990 			break;
991 	}
992 
993 	vfree(buf);
994 	return err;
995 }
996 
997 static int dbg_check_orphans(struct ubifs_info *c)
998 {
999 	struct check_info ci;
1000 	int err;
1001 
1002 	if (!dbg_is_chk_orph(c))
1003 		return 0;
1004 
1005 	ci.last_ino = 0;
1006 	ci.tot_inos = 0;
1007 	ci.missing  = 0;
1008 	ci.leaf_cnt = 0;
1009 	ci.root = RB_ROOT;
1010 	ci.node = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
1011 	if (!ci.node) {
1012 		ubifs_err(c, "out of memory");
1013 		return -ENOMEM;
1014 	}
1015 
1016 	err = dbg_scan_orphans(c, &ci);
1017 	if (err)
1018 		goto out;
1019 
1020 	err = dbg_walk_index(c, &dbg_orphan_check, NULL, &ci);
1021 	if (err) {
1022 		ubifs_err(c, "cannot scan TNC, error %d", err);
1023 		goto out;
1024 	}
1025 
1026 	if (ci.missing) {
1027 		ubifs_err(c, "%lu missing orphan(s)", ci.missing);
1028 		err = -EINVAL;
1029 		goto out;
1030 	}
1031 
1032 	dbg_cmt("last inode number is %lu", ci.last_ino);
1033 	dbg_cmt("total number of inodes is %lu", ci.tot_inos);
1034 	dbg_cmt("total number of leaf nodes is %llu", ci.leaf_cnt);
1035 
1036 out:
1037 	dbg_free_check_tree(&ci.root);
1038 	kfree(ci.node);
1039 	return err;
1040 }
1041