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