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  * Authors: Adrian Hunter
8  *          Artem Bityutskiy (Битюцкий Артём)
9  */
10 
11 /*
12  * This file implements TNC (Tree Node Cache) which caches indexing nodes of
13  * the UBIFS B-tree.
14  *
15  * At the moment the locking rules of the TNC tree are quite simple and
16  * straightforward. We just have a mutex and lock it when we traverse the
17  * tree. If a znode is not in memory, we read it from flash while still having
18  * the mutex locked.
19  */
20 
21 #include <linux/crc32.h>
22 #include <linux/slab.h>
23 #include "ubifs.h"
24 
25 static int try_read_node(const struct ubifs_info *c, void *buf, int type,
26 			 struct ubifs_zbranch *zbr);
27 static int fallible_read_node(struct ubifs_info *c, const union ubifs_key *key,
28 			      struct ubifs_zbranch *zbr, void *node);
29 
30 /*
31  * Returned codes of 'matches_name()' and 'fallible_matches_name()' functions.
32  * @NAME_LESS: name corresponding to the first argument is less than second
33  * @NAME_MATCHES: names match
34  * @NAME_GREATER: name corresponding to the second argument is greater than
35  *                first
36  * @NOT_ON_MEDIA: node referred by zbranch does not exist on the media
37  *
38  * These constants were introduce to improve readability.
39  */
40 enum {
41 	NAME_LESS    = 0,
42 	NAME_MATCHES = 1,
43 	NAME_GREATER = 2,
44 	NOT_ON_MEDIA = 3,
45 };
46 
47 /**
48  * insert_old_idx - record an index node obsoleted since the last commit start.
49  * @c: UBIFS file-system description object
50  * @lnum: LEB number of obsoleted index node
51  * @offs: offset of obsoleted index node
52  *
53  * Returns %0 on success, and a negative error code on failure.
54  *
55  * For recovery, there must always be a complete intact version of the index on
56  * flash at all times. That is called the "old index". It is the index as at the
57  * time of the last successful commit. Many of the index nodes in the old index
58  * may be dirty, but they must not be erased until the next successful commit
59  * (at which point that index becomes the old index).
60  *
61  * That means that the garbage collection and the in-the-gaps method of
62  * committing must be able to determine if an index node is in the old index.
63  * Most of the old index nodes can be found by looking up the TNC using the
64  * 'lookup_znode()' function. However, some of the old index nodes may have
65  * been deleted from the current index or may have been changed so much that
66  * they cannot be easily found. In those cases, an entry is added to an RB-tree.
67  * That is what this function does. The RB-tree is ordered by LEB number and
68  * offset because they uniquely identify the old index node.
69  */
insert_old_idx(struct ubifs_info * c,int lnum,int offs)70 static int insert_old_idx(struct ubifs_info *c, int lnum, int offs)
71 {
72 	struct ubifs_old_idx *old_idx, *o;
73 	struct rb_node **p, *parent = NULL;
74 
75 	old_idx = kmalloc(sizeof(struct ubifs_old_idx), GFP_NOFS);
76 	if (unlikely(!old_idx))
77 		return -ENOMEM;
78 	old_idx->lnum = lnum;
79 	old_idx->offs = offs;
80 
81 	p = &c->old_idx.rb_node;
82 	while (*p) {
83 		parent = *p;
84 		o = rb_entry(parent, struct ubifs_old_idx, rb);
85 		if (lnum < o->lnum)
86 			p = &(*p)->rb_left;
87 		else if (lnum > o->lnum)
88 			p = &(*p)->rb_right;
89 		else if (offs < o->offs)
90 			p = &(*p)->rb_left;
91 		else if (offs > o->offs)
92 			p = &(*p)->rb_right;
93 		else {
94 			ubifs_err(c, "old idx added twice!");
95 			kfree(old_idx);
96 			return 0;
97 		}
98 	}
99 	rb_link_node(&old_idx->rb, parent, p);
100 	rb_insert_color(&old_idx->rb, &c->old_idx);
101 	return 0;
102 }
103 
104 /**
105  * insert_old_idx_znode - record a znode obsoleted since last commit start.
106  * @c: UBIFS file-system description object
107  * @znode: znode of obsoleted index node
108  *
109  * Returns %0 on success, and a negative error code on failure.
110  */
insert_old_idx_znode(struct ubifs_info * c,struct ubifs_znode * znode)111 int insert_old_idx_znode(struct ubifs_info *c, struct ubifs_znode *znode)
112 {
113 	if (znode->parent) {
114 		struct ubifs_zbranch *zbr;
115 
116 		zbr = &znode->parent->zbranch[znode->iip];
117 		if (zbr->len)
118 			return insert_old_idx(c, zbr->lnum, zbr->offs);
119 	} else
120 		if (c->zroot.len)
121 			return insert_old_idx(c, c->zroot.lnum,
122 					      c->zroot.offs);
123 	return 0;
124 }
125 
126 /**
127  * ins_clr_old_idx_znode - record a znode obsoleted since last commit start.
128  * @c: UBIFS file-system description object
129  * @znode: znode of obsoleted index node
130  *
131  * Returns %0 on success, and a negative error code on failure.
132  */
ins_clr_old_idx_znode(struct ubifs_info * c,struct ubifs_znode * znode)133 static int ins_clr_old_idx_znode(struct ubifs_info *c,
134 				 struct ubifs_znode *znode)
135 {
136 	int err;
137 
138 	if (znode->parent) {
139 		struct ubifs_zbranch *zbr;
140 
141 		zbr = &znode->parent->zbranch[znode->iip];
142 		if (zbr->len) {
143 			err = insert_old_idx(c, zbr->lnum, zbr->offs);
144 			if (err)
145 				return err;
146 			zbr->lnum = 0;
147 			zbr->offs = 0;
148 			zbr->len = 0;
149 		}
150 	} else
151 		if (c->zroot.len) {
152 			err = insert_old_idx(c, c->zroot.lnum, c->zroot.offs);
153 			if (err)
154 				return err;
155 			c->zroot.lnum = 0;
156 			c->zroot.offs = 0;
157 			c->zroot.len = 0;
158 		}
159 	return 0;
160 }
161 
162 /**
163  * destroy_old_idx - destroy the old_idx RB-tree.
164  * @c: UBIFS file-system description object
165  *
166  * During start commit, the old_idx RB-tree is used to avoid overwriting index
167  * nodes that were in the index last commit but have since been deleted.  This
168  * is necessary for recovery i.e. the old index must be kept intact until the
169  * new index is successfully written.  The old-idx RB-tree is used for the
170  * in-the-gaps method of writing index nodes and is destroyed every commit.
171  */
destroy_old_idx(struct ubifs_info * c)172 void destroy_old_idx(struct ubifs_info *c)
173 {
174 	struct ubifs_old_idx *old_idx, *n;
175 
176 	rbtree_postorder_for_each_entry_safe(old_idx, n, &c->old_idx, rb)
177 		kfree(old_idx);
178 
179 	c->old_idx = RB_ROOT;
180 }
181 
182 /**
183  * copy_znode - copy a dirty znode.
184  * @c: UBIFS file-system description object
185  * @znode: znode to copy
186  *
187  * A dirty znode being committed may not be changed, so it is copied.
188  */
copy_znode(struct ubifs_info * c,struct ubifs_znode * znode)189 static struct ubifs_znode *copy_znode(struct ubifs_info *c,
190 				      struct ubifs_znode *znode)
191 {
192 	struct ubifs_znode *zn;
193 
194 	zn = kmemdup(znode, c->max_znode_sz, GFP_NOFS);
195 	if (unlikely(!zn))
196 		return ERR_PTR(-ENOMEM);
197 
198 	zn->cnext = NULL;
199 	__set_bit(DIRTY_ZNODE, &zn->flags);
200 	__clear_bit(COW_ZNODE, &zn->flags);
201 
202 	ubifs_assert(c, !ubifs_zn_obsolete(znode));
203 	__set_bit(OBSOLETE_ZNODE, &znode->flags);
204 
205 	if (znode->level != 0) {
206 		int i;
207 		const int n = zn->child_cnt;
208 
209 		/* The children now have new parent */
210 		for (i = 0; i < n; i++) {
211 			struct ubifs_zbranch *zbr = &zn->zbranch[i];
212 
213 			if (zbr->znode)
214 				zbr->znode->parent = zn;
215 		}
216 	}
217 
218 	atomic_long_inc(&c->dirty_zn_cnt);
219 	return zn;
220 }
221 
222 /**
223  * add_idx_dirt - add dirt due to a dirty znode.
224  * @c: UBIFS file-system description object
225  * @lnum: LEB number of index node
226  * @dirt: size of index node
227  *
228  * This function updates lprops dirty space and the new size of the index.
229  */
add_idx_dirt(struct ubifs_info * c,int lnum,int dirt)230 static int add_idx_dirt(struct ubifs_info *c, int lnum, int dirt)
231 {
232 	c->calc_idx_sz -= ALIGN(dirt, 8);
233 	return ubifs_add_dirt(c, lnum, dirt);
234 }
235 
236 /**
237  * dirty_cow_znode - ensure a znode is not being committed.
238  * @c: UBIFS file-system description object
239  * @zbr: branch of znode to check
240  *
241  * Returns dirtied znode on success or negative error code on failure.
242  */
dirty_cow_znode(struct ubifs_info * c,struct ubifs_zbranch * zbr)243 static struct ubifs_znode *dirty_cow_znode(struct ubifs_info *c,
244 					   struct ubifs_zbranch *zbr)
245 {
246 	struct ubifs_znode *znode = zbr->znode;
247 	struct ubifs_znode *zn;
248 	int err;
249 
250 	if (!ubifs_zn_cow(znode)) {
251 		/* znode is not being committed */
252 		if (!test_and_set_bit(DIRTY_ZNODE, &znode->flags)) {
253 			atomic_long_inc(&c->dirty_zn_cnt);
254 			atomic_long_dec(&c->clean_zn_cnt);
255 			atomic_long_dec(&ubifs_clean_zn_cnt);
256 			err = add_idx_dirt(c, zbr->lnum, zbr->len);
257 			if (unlikely(err))
258 				return ERR_PTR(err);
259 		}
260 		return znode;
261 	}
262 
263 	zn = copy_znode(c, znode);
264 	if (IS_ERR(zn))
265 		return zn;
266 
267 	if (zbr->len) {
268 		err = insert_old_idx(c, zbr->lnum, zbr->offs);
269 		if (unlikely(err))
270 			return ERR_PTR(err);
271 		err = add_idx_dirt(c, zbr->lnum, zbr->len);
272 	} else
273 		err = 0;
274 
275 	zbr->znode = zn;
276 	zbr->lnum = 0;
277 	zbr->offs = 0;
278 	zbr->len = 0;
279 
280 	if (unlikely(err))
281 		return ERR_PTR(err);
282 	return zn;
283 }
284 
285 /**
286  * lnc_add - add a leaf node to the leaf node cache.
287  * @c: UBIFS file-system description object
288  * @zbr: zbranch of leaf node
289  * @node: leaf node
290  *
291  * Leaf nodes are non-index nodes directory entry nodes or data nodes. The
292  * purpose of the leaf node cache is to save re-reading the same leaf node over
293  * and over again. Most things are cached by VFS, however the file system must
294  * cache directory entries for readdir and for resolving hash collisions. The
295  * present implementation of the leaf node cache is extremely simple, and
296  * allows for error returns that are not used but that may be needed if a more
297  * complex implementation is created.
298  *
299  * Note, this function does not add the @node object to LNC directly, but
300  * allocates a copy of the object and adds the copy to LNC. The reason for this
301  * is that @node has been allocated outside of the TNC subsystem and will be
302  * used with @c->tnc_mutex unlock upon return from the TNC subsystem. But LNC
303  * may be changed at any time, e.g. freed by the shrinker.
304  */
lnc_add(struct ubifs_info * c,struct ubifs_zbranch * zbr,const void * node)305 static int lnc_add(struct ubifs_info *c, struct ubifs_zbranch *zbr,
306 		   const void *node)
307 {
308 	int err;
309 	void *lnc_node;
310 	const struct ubifs_dent_node *dent = node;
311 
312 	ubifs_assert(c, !zbr->leaf);
313 	ubifs_assert(c, zbr->len != 0);
314 	ubifs_assert(c, is_hash_key(c, &zbr->key));
315 
316 	err = ubifs_validate_entry(c, dent);
317 	if (err) {
318 		dump_stack();
319 		ubifs_dump_node(c, dent, zbr->len);
320 		return err;
321 	}
322 
323 	lnc_node = kmemdup(node, zbr->len, GFP_NOFS);
324 	if (!lnc_node)
325 		/* We don't have to have the cache, so no error */
326 		return 0;
327 
328 	zbr->leaf = lnc_node;
329 	return 0;
330 }
331 
332  /**
333  * lnc_add_directly - add a leaf node to the leaf-node-cache.
334  * @c: UBIFS file-system description object
335  * @zbr: zbranch of leaf node
336  * @node: leaf node
337  *
338  * This function is similar to 'lnc_add()', but it does not create a copy of
339  * @node but inserts @node to TNC directly.
340  */
lnc_add_directly(struct ubifs_info * c,struct ubifs_zbranch * zbr,void * node)341 static int lnc_add_directly(struct ubifs_info *c, struct ubifs_zbranch *zbr,
342 			    void *node)
343 {
344 	int err;
345 
346 	ubifs_assert(c, !zbr->leaf);
347 	ubifs_assert(c, zbr->len != 0);
348 
349 	err = ubifs_validate_entry(c, node);
350 	if (err) {
351 		dump_stack();
352 		ubifs_dump_node(c, node, zbr->len);
353 		return err;
354 	}
355 
356 	zbr->leaf = node;
357 	return 0;
358 }
359 
360 /**
361  * lnc_free - remove a leaf node from the leaf node cache.
362  * @zbr: zbranch of leaf node
363  */
lnc_free(struct ubifs_zbranch * zbr)364 static void lnc_free(struct ubifs_zbranch *zbr)
365 {
366 	if (!zbr->leaf)
367 		return;
368 	kfree(zbr->leaf);
369 	zbr->leaf = NULL;
370 }
371 
372 /**
373  * tnc_read_hashed_node - read a "hashed" leaf node.
374  * @c: UBIFS file-system description object
375  * @zbr: key and position of the node
376  * @node: node is returned here
377  *
378  * This function reads a "hashed" node defined by @zbr from the leaf node cache
379  * (in it is there) or from the hash media, in which case the node is also
380  * added to LNC. Returns zero in case of success or a negative error
381  * code in case of failure.
382  */
tnc_read_hashed_node(struct ubifs_info * c,struct ubifs_zbranch * zbr,void * node)383 static int tnc_read_hashed_node(struct ubifs_info *c, struct ubifs_zbranch *zbr,
384 				void *node)
385 {
386 	int err;
387 
388 	ubifs_assert(c, is_hash_key(c, &zbr->key));
389 
390 	if (zbr->leaf) {
391 		/* Read from the leaf node cache */
392 		ubifs_assert(c, zbr->len != 0);
393 		memcpy(node, zbr->leaf, zbr->len);
394 		return 0;
395 	}
396 
397 	if (c->replaying) {
398 		err = fallible_read_node(c, &zbr->key, zbr, node);
399 		/*
400 		 * When the node was not found, return -ENOENT, 0 otherwise.
401 		 * Negative return codes stay as-is.
402 		 */
403 		if (err == 0)
404 			err = -ENOENT;
405 		else if (err == 1)
406 			err = 0;
407 	} else {
408 		err = ubifs_tnc_read_node(c, zbr, node);
409 	}
410 	if (err)
411 		return err;
412 
413 	/* Add the node to the leaf node cache */
414 	err = lnc_add(c, zbr, node);
415 	return err;
416 }
417 
418 /**
419  * try_read_node - read a node if it is a node.
420  * @c: UBIFS file-system description object
421  * @buf: buffer to read to
422  * @type: node type
423  * @zbr: the zbranch describing the node to read
424  *
425  * This function tries to read a node of known type and length, checks it and
426  * stores it in @buf. This function returns %1 if a node is present and %0 if
427  * a node is not present. A negative error code is returned for I/O errors.
428  * This function performs that same function as ubifs_read_node except that
429  * it does not require that there is actually a node present and instead
430  * the return code indicates if a node was read.
431  *
432  * Note, this function does not check CRC of data nodes if @c->no_chk_data_crc
433  * is true (it is controlled by corresponding mount option). However, if
434  * @c->mounting or @c->remounting_rw is true (we are mounting or re-mounting to
435  * R/W mode), @c->no_chk_data_crc is ignored and CRC is checked. This is
436  * because during mounting or re-mounting from R/O mode to R/W mode we may read
437  * journal nodes (when replying the journal or doing the recovery) and the
438  * journal nodes may potentially be corrupted, so checking is required.
439  */
try_read_node(const struct ubifs_info * c,void * buf,int type,struct ubifs_zbranch * zbr)440 static int try_read_node(const struct ubifs_info *c, void *buf, int type,
441 			 struct ubifs_zbranch *zbr)
442 {
443 	int len = zbr->len;
444 	int lnum = zbr->lnum;
445 	int offs = zbr->offs;
446 	int err, node_len;
447 	struct ubifs_ch *ch = buf;
448 	uint32_t crc, node_crc;
449 
450 	dbg_io("LEB %d:%d, %s, length %d", lnum, offs, dbg_ntype(type), len);
451 
452 	err = ubifs_leb_read(c, lnum, buf, offs, len, 1);
453 	if (err) {
454 		ubifs_err(c, "cannot read node type %d from LEB %d:%d, error %d",
455 			  type, lnum, offs, err);
456 		return err;
457 	}
458 
459 	if (le32_to_cpu(ch->magic) != UBIFS_NODE_MAGIC)
460 		return 0;
461 
462 	if (ch->node_type != type)
463 		return 0;
464 
465 	node_len = le32_to_cpu(ch->len);
466 	if (node_len != len)
467 		return 0;
468 
469 	if (type != UBIFS_DATA_NODE || !c->no_chk_data_crc || c->mounting ||
470 	    c->remounting_rw) {
471 		crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
472 		node_crc = le32_to_cpu(ch->crc);
473 		if (crc != node_crc)
474 			return 0;
475 	}
476 
477 	err = ubifs_node_check_hash(c, buf, zbr->hash);
478 	if (err) {
479 		ubifs_bad_hash(c, buf, zbr->hash, lnum, offs);
480 		return 0;
481 	}
482 
483 	return 1;
484 }
485 
486 /**
487  * fallible_read_node - try to read a leaf node.
488  * @c: UBIFS file-system description object
489  * @key:  key of node to read
490  * @zbr:  position of node
491  * @node: node returned
492  *
493  * This function tries to read a node and returns %1 if the node is read, %0
494  * if the node is not present, and a negative error code in the case of error.
495  */
fallible_read_node(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_zbranch * zbr,void * node)496 static int fallible_read_node(struct ubifs_info *c, const union ubifs_key *key,
497 			      struct ubifs_zbranch *zbr, void *node)
498 {
499 	int ret;
500 
501 	dbg_tnck(key, "LEB %d:%d, key ", zbr->lnum, zbr->offs);
502 
503 	ret = try_read_node(c, node, key_type(c, key), zbr);
504 	if (ret == 1) {
505 		union ubifs_key node_key;
506 		struct ubifs_dent_node *dent = node;
507 
508 		/* All nodes have key in the same place */
509 		key_read(c, &dent->key, &node_key);
510 		if (keys_cmp(c, key, &node_key) != 0)
511 			ret = 0;
512 	}
513 	if (ret == 0 && c->replaying)
514 		dbg_mntk(key, "dangling branch LEB %d:%d len %d, key ",
515 			zbr->lnum, zbr->offs, zbr->len);
516 	return ret;
517 }
518 
519 /**
520  * matches_name - determine if a direntry or xattr entry matches a given name.
521  * @c: UBIFS file-system description object
522  * @zbr: zbranch of dent
523  * @nm: name to match
524  *
525  * This function checks if xentry/direntry referred by zbranch @zbr matches name
526  * @nm. Returns %NAME_MATCHES if it does, %NAME_LESS if the name referred by
527  * @zbr is less than @nm, and %NAME_GREATER if it is greater than @nm. In case
528  * of failure, a negative error code is returned.
529  */
matches_name(struct ubifs_info * c,struct ubifs_zbranch * zbr,const struct fscrypt_name * nm)530 static int matches_name(struct ubifs_info *c, struct ubifs_zbranch *zbr,
531 			const struct fscrypt_name *nm)
532 {
533 	struct ubifs_dent_node *dent;
534 	int nlen, err;
535 
536 	/* If possible, match against the dent in the leaf node cache */
537 	if (!zbr->leaf) {
538 		dent = kmalloc(zbr->len, GFP_NOFS);
539 		if (!dent)
540 			return -ENOMEM;
541 
542 		err = ubifs_tnc_read_node(c, zbr, dent);
543 		if (err)
544 			goto out_free;
545 
546 		/* Add the node to the leaf node cache */
547 		err = lnc_add_directly(c, zbr, dent);
548 		if (err)
549 			goto out_free;
550 	} else
551 		dent = zbr->leaf;
552 
553 	nlen = le16_to_cpu(dent->nlen);
554 	err = memcmp(dent->name, fname_name(nm), min_t(int, nlen, fname_len(nm)));
555 	if (err == 0) {
556 		if (nlen == fname_len(nm))
557 			return NAME_MATCHES;
558 		else if (nlen < fname_len(nm))
559 			return NAME_LESS;
560 		else
561 			return NAME_GREATER;
562 	} else if (err < 0)
563 		return NAME_LESS;
564 	else
565 		return NAME_GREATER;
566 
567 out_free:
568 	kfree(dent);
569 	return err;
570 }
571 
572 /**
573  * get_znode - get a TNC znode that may not be loaded yet.
574  * @c: UBIFS file-system description object
575  * @znode: parent znode
576  * @n: znode branch slot number
577  *
578  * This function returns the znode or a negative error code.
579  */
get_znode(struct ubifs_info * c,struct ubifs_znode * znode,int n)580 static struct ubifs_znode *get_znode(struct ubifs_info *c,
581 				     struct ubifs_znode *znode, int n)
582 {
583 	struct ubifs_zbranch *zbr;
584 
585 	zbr = &znode->zbranch[n];
586 	if (zbr->znode)
587 		znode = zbr->znode;
588 	else
589 		znode = ubifs_load_znode(c, zbr, znode, n);
590 	return znode;
591 }
592 
593 /**
594  * tnc_next - find next TNC entry.
595  * @c: UBIFS file-system description object
596  * @zn: znode is passed and returned here
597  * @n: znode branch slot number is passed and returned here
598  *
599  * This function returns %0 if the next TNC entry is found, %-ENOENT if there is
600  * no next entry, or a negative error code otherwise.
601  */
tnc_next(struct ubifs_info * c,struct ubifs_znode ** zn,int * n)602 static int tnc_next(struct ubifs_info *c, struct ubifs_znode **zn, int *n)
603 {
604 	struct ubifs_znode *znode = *zn;
605 	int nn = *n;
606 
607 	nn += 1;
608 	if (nn < znode->child_cnt) {
609 		*n = nn;
610 		return 0;
611 	}
612 	while (1) {
613 		struct ubifs_znode *zp;
614 
615 		zp = znode->parent;
616 		if (!zp)
617 			return -ENOENT;
618 		nn = znode->iip + 1;
619 		znode = zp;
620 		if (nn < znode->child_cnt) {
621 			znode = get_znode(c, znode, nn);
622 			if (IS_ERR(znode))
623 				return PTR_ERR(znode);
624 			while (znode->level != 0) {
625 				znode = get_znode(c, znode, 0);
626 				if (IS_ERR(znode))
627 					return PTR_ERR(znode);
628 			}
629 			nn = 0;
630 			break;
631 		}
632 	}
633 	*zn = znode;
634 	*n = nn;
635 	return 0;
636 }
637 
638 /**
639  * tnc_prev - find previous TNC entry.
640  * @c: UBIFS file-system description object
641  * @zn: znode is returned here
642  * @n: znode branch slot number is passed and returned here
643  *
644  * This function returns %0 if the previous TNC entry is found, %-ENOENT if
645  * there is no next entry, or a negative error code otherwise.
646  */
tnc_prev(struct ubifs_info * c,struct ubifs_znode ** zn,int * n)647 static int tnc_prev(struct ubifs_info *c, struct ubifs_znode **zn, int *n)
648 {
649 	struct ubifs_znode *znode = *zn;
650 	int nn = *n;
651 
652 	if (nn > 0) {
653 		*n = nn - 1;
654 		return 0;
655 	}
656 	while (1) {
657 		struct ubifs_znode *zp;
658 
659 		zp = znode->parent;
660 		if (!zp)
661 			return -ENOENT;
662 		nn = znode->iip - 1;
663 		znode = zp;
664 		if (nn >= 0) {
665 			znode = get_znode(c, znode, nn);
666 			if (IS_ERR(znode))
667 				return PTR_ERR(znode);
668 			while (znode->level != 0) {
669 				nn = znode->child_cnt - 1;
670 				znode = get_znode(c, znode, nn);
671 				if (IS_ERR(znode))
672 					return PTR_ERR(znode);
673 			}
674 			nn = znode->child_cnt - 1;
675 			break;
676 		}
677 	}
678 	*zn = znode;
679 	*n = nn;
680 	return 0;
681 }
682 
683 /**
684  * resolve_collision - resolve a collision.
685  * @c: UBIFS file-system description object
686  * @key: key of a directory or extended attribute entry
687  * @zn: znode is returned here
688  * @n: zbranch number is passed and returned here
689  * @nm: name of the entry
690  *
691  * This function is called for "hashed" keys to make sure that the found key
692  * really corresponds to the looked up node (directory or extended attribute
693  * entry). It returns %1 and sets @zn and @n if the collision is resolved.
694  * %0 is returned if @nm is not found and @zn and @n are set to the previous
695  * entry, i.e. to the entry after which @nm could follow if it were in TNC.
696  * This means that @n may be set to %-1 if the leftmost key in @zn is the
697  * previous one. A negative error code is returned on failures.
698  */
resolve_collision(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_znode ** zn,int * n,const struct fscrypt_name * nm)699 static int resolve_collision(struct ubifs_info *c, const union ubifs_key *key,
700 			     struct ubifs_znode **zn, int *n,
701 			     const struct fscrypt_name *nm)
702 {
703 	int err;
704 
705 	err = matches_name(c, &(*zn)->zbranch[*n], nm);
706 	if (unlikely(err < 0))
707 		return err;
708 	if (err == NAME_MATCHES)
709 		return 1;
710 
711 	if (err == NAME_GREATER) {
712 		/* Look left */
713 		while (1) {
714 			err = tnc_prev(c, zn, n);
715 			if (err == -ENOENT) {
716 				ubifs_assert(c, *n == 0);
717 				*n = -1;
718 				return 0;
719 			}
720 			if (err < 0)
721 				return err;
722 			if (keys_cmp(c, &(*zn)->zbranch[*n].key, key)) {
723 				/*
724 				 * We have found the branch after which we would
725 				 * like to insert, but inserting in this znode
726 				 * may still be wrong. Consider the following 3
727 				 * znodes, in the case where we are resolving a
728 				 * collision with Key2.
729 				 *
730 				 *                  znode zp
731 				 *            ----------------------
732 				 * level 1     |  Key0  |  Key1  |
733 				 *            -----------------------
734 				 *                 |            |
735 				 *       znode za  |            |  znode zb
736 				 *          ------------      ------------
737 				 * level 0  |  Key0  |        |  Key2  |
738 				 *          ------------      ------------
739 				 *
740 				 * The lookup finds Key2 in znode zb. Lets say
741 				 * there is no match and the name is greater so
742 				 * we look left. When we find Key0, we end up
743 				 * here. If we return now, we will insert into
744 				 * znode za at slot n = 1.  But that is invalid
745 				 * according to the parent's keys.  Key2 must
746 				 * be inserted into znode zb.
747 				 *
748 				 * Note, this problem is not relevant for the
749 				 * case when we go right, because
750 				 * 'tnc_insert()' would correct the parent key.
751 				 */
752 				if (*n == (*zn)->child_cnt - 1) {
753 					err = tnc_next(c, zn, n);
754 					if (err) {
755 						/* Should be impossible */
756 						ubifs_assert(c, 0);
757 						if (err == -ENOENT)
758 							err = -EINVAL;
759 						return err;
760 					}
761 					ubifs_assert(c, *n == 0);
762 					*n = -1;
763 				}
764 				return 0;
765 			}
766 			err = matches_name(c, &(*zn)->zbranch[*n], nm);
767 			if (err < 0)
768 				return err;
769 			if (err == NAME_LESS)
770 				return 0;
771 			if (err == NAME_MATCHES)
772 				return 1;
773 			ubifs_assert(c, err == NAME_GREATER);
774 		}
775 	} else {
776 		int nn = *n;
777 		struct ubifs_znode *znode = *zn;
778 
779 		/* Look right */
780 		while (1) {
781 			err = tnc_next(c, &znode, &nn);
782 			if (err == -ENOENT)
783 				return 0;
784 			if (err < 0)
785 				return err;
786 			if (keys_cmp(c, &znode->zbranch[nn].key, key))
787 				return 0;
788 			err = matches_name(c, &znode->zbranch[nn], nm);
789 			if (err < 0)
790 				return err;
791 			if (err == NAME_GREATER)
792 				return 0;
793 			*zn = znode;
794 			*n = nn;
795 			if (err == NAME_MATCHES)
796 				return 1;
797 			ubifs_assert(c, err == NAME_LESS);
798 		}
799 	}
800 }
801 
802 /**
803  * fallible_matches_name - determine if a dent matches a given name.
804  * @c: UBIFS file-system description object
805  * @zbr: zbranch of dent
806  * @nm: name to match
807  *
808  * This is a "fallible" version of 'matches_name()' function which does not
809  * panic if the direntry/xentry referred by @zbr does not exist on the media.
810  *
811  * This function checks if xentry/direntry referred by zbranch @zbr matches name
812  * @nm. Returns %NAME_MATCHES it does, %NAME_LESS if the name referred by @zbr
813  * is less than @nm, %NAME_GREATER if it is greater than @nm, and @NOT_ON_MEDIA
814  * if xentry/direntry referred by @zbr does not exist on the media. A negative
815  * error code is returned in case of failure.
816  */
fallible_matches_name(struct ubifs_info * c,struct ubifs_zbranch * zbr,const struct fscrypt_name * nm)817 static int fallible_matches_name(struct ubifs_info *c,
818 				 struct ubifs_zbranch *zbr,
819 				 const struct fscrypt_name *nm)
820 {
821 	struct ubifs_dent_node *dent;
822 	int nlen, err;
823 
824 	/* If possible, match against the dent in the leaf node cache */
825 	if (!zbr->leaf) {
826 		dent = kmalloc(zbr->len, GFP_NOFS);
827 		if (!dent)
828 			return -ENOMEM;
829 
830 		err = fallible_read_node(c, &zbr->key, zbr, dent);
831 		if (err < 0)
832 			goto out_free;
833 		if (err == 0) {
834 			/* The node was not present */
835 			err = NOT_ON_MEDIA;
836 			goto out_free;
837 		}
838 		ubifs_assert(c, err == 1);
839 
840 		err = lnc_add_directly(c, zbr, dent);
841 		if (err)
842 			goto out_free;
843 	} else
844 		dent = zbr->leaf;
845 
846 	nlen = le16_to_cpu(dent->nlen);
847 	err = memcmp(dent->name, fname_name(nm), min_t(int, nlen, fname_len(nm)));
848 	if (err == 0) {
849 		if (nlen == fname_len(nm))
850 			return NAME_MATCHES;
851 		else if (nlen < fname_len(nm))
852 			return NAME_LESS;
853 		else
854 			return NAME_GREATER;
855 	} else if (err < 0)
856 		return NAME_LESS;
857 	else
858 		return NAME_GREATER;
859 
860 out_free:
861 	kfree(dent);
862 	return err;
863 }
864 
865 /**
866  * fallible_resolve_collision - resolve a collision even if nodes are missing.
867  * @c: UBIFS file-system description object
868  * @key: key
869  * @zn: znode is returned here
870  * @n: branch number is passed and returned here
871  * @nm: name of directory entry
872  * @adding: indicates caller is adding a key to the TNC
873  *
874  * This is a "fallible" version of the 'resolve_collision()' function which
875  * does not panic if one of the nodes referred to by TNC does not exist on the
876  * media. This may happen when replaying the journal if a deleted node was
877  * Garbage-collected and the commit was not done. A branch that refers to a node
878  * that is not present is called a dangling branch. The following are the return
879  * codes for this function:
880  *  o if @nm was found, %1 is returned and @zn and @n are set to the found
881  *    branch;
882  *  o if we are @adding and @nm was not found, %0 is returned;
883  *  o if we are not @adding and @nm was not found, but a dangling branch was
884  *    found, then %1 is returned and @zn and @n are set to the dangling branch;
885  *  o a negative error code is returned in case of failure.
886  */
fallible_resolve_collision(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_znode ** zn,int * n,const struct fscrypt_name * nm,int adding)887 static int fallible_resolve_collision(struct ubifs_info *c,
888 				      const union ubifs_key *key,
889 				      struct ubifs_znode **zn, int *n,
890 				      const struct fscrypt_name *nm,
891 				      int adding)
892 {
893 	struct ubifs_znode *o_znode = NULL, *znode = *zn;
894 	int o_n, err, cmp, unsure = 0, nn = *n;
895 
896 	cmp = fallible_matches_name(c, &znode->zbranch[nn], nm);
897 	if (unlikely(cmp < 0))
898 		return cmp;
899 	if (cmp == NAME_MATCHES)
900 		return 1;
901 	if (cmp == NOT_ON_MEDIA) {
902 		o_znode = znode;
903 		o_n = nn;
904 		/*
905 		 * We are unlucky and hit a dangling branch straight away.
906 		 * Now we do not really know where to go to find the needed
907 		 * branch - to the left or to the right. Well, let's try left.
908 		 */
909 		unsure = 1;
910 	} else if (!adding)
911 		unsure = 1; /* Remove a dangling branch wherever it is */
912 
913 	if (cmp == NAME_GREATER || unsure) {
914 		/* Look left */
915 		while (1) {
916 			err = tnc_prev(c, zn, n);
917 			if (err == -ENOENT) {
918 				ubifs_assert(c, *n == 0);
919 				*n = -1;
920 				break;
921 			}
922 			if (err < 0)
923 				return err;
924 			if (keys_cmp(c, &(*zn)->zbranch[*n].key, key)) {
925 				/* See comments in 'resolve_collision()' */
926 				if (*n == (*zn)->child_cnt - 1) {
927 					err = tnc_next(c, zn, n);
928 					if (err) {
929 						/* Should be impossible */
930 						ubifs_assert(c, 0);
931 						if (err == -ENOENT)
932 							err = -EINVAL;
933 						return err;
934 					}
935 					ubifs_assert(c, *n == 0);
936 					*n = -1;
937 				}
938 				break;
939 			}
940 			err = fallible_matches_name(c, &(*zn)->zbranch[*n], nm);
941 			if (err < 0)
942 				return err;
943 			if (err == NAME_MATCHES)
944 				return 1;
945 			if (err == NOT_ON_MEDIA) {
946 				o_znode = *zn;
947 				o_n = *n;
948 				continue;
949 			}
950 			if (!adding)
951 				continue;
952 			if (err == NAME_LESS)
953 				break;
954 			else
955 				unsure = 0;
956 		}
957 	}
958 
959 	if (cmp == NAME_LESS || unsure) {
960 		/* Look right */
961 		*zn = znode;
962 		*n = nn;
963 		while (1) {
964 			err = tnc_next(c, &znode, &nn);
965 			if (err == -ENOENT)
966 				break;
967 			if (err < 0)
968 				return err;
969 			if (keys_cmp(c, &znode->zbranch[nn].key, key))
970 				break;
971 			err = fallible_matches_name(c, &znode->zbranch[nn], nm);
972 			if (err < 0)
973 				return err;
974 			if (err == NAME_GREATER)
975 				break;
976 			*zn = znode;
977 			*n = nn;
978 			if (err == NAME_MATCHES)
979 				return 1;
980 			if (err == NOT_ON_MEDIA) {
981 				o_znode = znode;
982 				o_n = nn;
983 			}
984 		}
985 	}
986 
987 	/* Never match a dangling branch when adding */
988 	if (adding || !o_znode)
989 		return 0;
990 
991 	dbg_mntk(key, "dangling match LEB %d:%d len %d key ",
992 		o_znode->zbranch[o_n].lnum, o_znode->zbranch[o_n].offs,
993 		o_znode->zbranch[o_n].len);
994 	*zn = o_znode;
995 	*n = o_n;
996 	return 1;
997 }
998 
999 /**
1000  * matches_position - determine if a zbranch matches a given position.
1001  * @zbr: zbranch of dent
1002  * @lnum: LEB number of dent to match
1003  * @offs: offset of dent to match
1004  *
1005  * This function returns %1 if @lnum:@offs matches, and %0 otherwise.
1006  */
matches_position(struct ubifs_zbranch * zbr,int lnum,int offs)1007 static int matches_position(struct ubifs_zbranch *zbr, int lnum, int offs)
1008 {
1009 	if (zbr->lnum == lnum && zbr->offs == offs)
1010 		return 1;
1011 	else
1012 		return 0;
1013 }
1014 
1015 /**
1016  * resolve_collision_directly - resolve a collision directly.
1017  * @c: UBIFS file-system description object
1018  * @key: key of directory entry
1019  * @zn: znode is passed and returned here
1020  * @n: zbranch number is passed and returned here
1021  * @lnum: LEB number of dent node to match
1022  * @offs: offset of dent node to match
1023  *
1024  * This function is used for "hashed" keys to make sure the found directory or
1025  * extended attribute entry node is what was looked for. It is used when the
1026  * flash address of the right node is known (@lnum:@offs) which makes it much
1027  * easier to resolve collisions (no need to read entries and match full
1028  * names). This function returns %1 and sets @zn and @n if the collision is
1029  * resolved, %0 if @lnum:@offs is not found and @zn and @n are set to the
1030  * previous directory entry. Otherwise a negative error code is returned.
1031  */
resolve_collision_directly(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_znode ** zn,int * n,int lnum,int offs)1032 static int resolve_collision_directly(struct ubifs_info *c,
1033 				      const union ubifs_key *key,
1034 				      struct ubifs_znode **zn, int *n,
1035 				      int lnum, int offs)
1036 {
1037 	struct ubifs_znode *znode;
1038 	int nn, err;
1039 
1040 	znode = *zn;
1041 	nn = *n;
1042 	if (matches_position(&znode->zbranch[nn], lnum, offs))
1043 		return 1;
1044 
1045 	/* Look left */
1046 	while (1) {
1047 		err = tnc_prev(c, &znode, &nn);
1048 		if (err == -ENOENT)
1049 			break;
1050 		if (err < 0)
1051 			return err;
1052 		if (keys_cmp(c, &znode->zbranch[nn].key, key))
1053 			break;
1054 		if (matches_position(&znode->zbranch[nn], lnum, offs)) {
1055 			*zn = znode;
1056 			*n = nn;
1057 			return 1;
1058 		}
1059 	}
1060 
1061 	/* Look right */
1062 	znode = *zn;
1063 	nn = *n;
1064 	while (1) {
1065 		err = tnc_next(c, &znode, &nn);
1066 		if (err == -ENOENT)
1067 			return 0;
1068 		if (err < 0)
1069 			return err;
1070 		if (keys_cmp(c, &znode->zbranch[nn].key, key))
1071 			return 0;
1072 		*zn = znode;
1073 		*n = nn;
1074 		if (matches_position(&znode->zbranch[nn], lnum, offs))
1075 			return 1;
1076 	}
1077 }
1078 
1079 /**
1080  * dirty_cow_bottom_up - dirty a znode and its ancestors.
1081  * @c: UBIFS file-system description object
1082  * @znode: znode to dirty
1083  *
1084  * If we do not have a unique key that resides in a znode, then we cannot
1085  * dirty that znode from the top down (i.e. by using lookup_level0_dirty)
1086  * This function records the path back to the last dirty ancestor, and then
1087  * dirties the znodes on that path.
1088  */
dirty_cow_bottom_up(struct ubifs_info * c,struct ubifs_znode * znode)1089 static struct ubifs_znode *dirty_cow_bottom_up(struct ubifs_info *c,
1090 					       struct ubifs_znode *znode)
1091 {
1092 	struct ubifs_znode *zp;
1093 	int *path = c->bottom_up_buf, p = 0;
1094 
1095 	ubifs_assert(c, c->zroot.znode);
1096 	ubifs_assert(c, znode);
1097 	if (c->zroot.znode->level > BOTTOM_UP_HEIGHT) {
1098 		kfree(c->bottom_up_buf);
1099 		c->bottom_up_buf = kmalloc_array(c->zroot.znode->level,
1100 						 sizeof(int),
1101 						 GFP_NOFS);
1102 		if (!c->bottom_up_buf)
1103 			return ERR_PTR(-ENOMEM);
1104 		path = c->bottom_up_buf;
1105 	}
1106 	if (c->zroot.znode->level) {
1107 		/* Go up until parent is dirty */
1108 		while (1) {
1109 			int n;
1110 
1111 			zp = znode->parent;
1112 			if (!zp)
1113 				break;
1114 			n = znode->iip;
1115 			ubifs_assert(c, p < c->zroot.znode->level);
1116 			path[p++] = n;
1117 			if (!zp->cnext && ubifs_zn_dirty(znode))
1118 				break;
1119 			znode = zp;
1120 		}
1121 	}
1122 
1123 	/* Come back down, dirtying as we go */
1124 	while (1) {
1125 		struct ubifs_zbranch *zbr;
1126 
1127 		zp = znode->parent;
1128 		if (zp) {
1129 			ubifs_assert(c, path[p - 1] >= 0);
1130 			ubifs_assert(c, path[p - 1] < zp->child_cnt);
1131 			zbr = &zp->zbranch[path[--p]];
1132 			znode = dirty_cow_znode(c, zbr);
1133 		} else {
1134 			ubifs_assert(c, znode == c->zroot.znode);
1135 			znode = dirty_cow_znode(c, &c->zroot);
1136 		}
1137 		if (IS_ERR(znode) || !p)
1138 			break;
1139 		ubifs_assert(c, path[p - 1] >= 0);
1140 		ubifs_assert(c, path[p - 1] < znode->child_cnt);
1141 		znode = znode->zbranch[path[p - 1]].znode;
1142 	}
1143 
1144 	return znode;
1145 }
1146 
1147 /**
1148  * ubifs_lookup_level0 - search for zero-level znode.
1149  * @c: UBIFS file-system description object
1150  * @key:  key to lookup
1151  * @zn: znode is returned here
1152  * @n: znode branch slot number is returned here
1153  *
1154  * This function looks up the TNC tree and search for zero-level znode which
1155  * refers key @key. The found zero-level znode is returned in @zn. There are 3
1156  * cases:
1157  *   o exact match, i.e. the found zero-level znode contains key @key, then %1
1158  *     is returned and slot number of the matched branch is stored in @n;
1159  *   o not exact match, which means that zero-level znode does not contain
1160  *     @key, then %0 is returned and slot number of the closest branch or %-1
1161  *     is stored in @n; In this case calling tnc_next() is mandatory.
1162  *   o @key is so small that it is even less than the lowest key of the
1163  *     leftmost zero-level node, then %0 is returned and %0 is stored in @n.
1164  *
1165  * Note, when the TNC tree is traversed, some znodes may be absent, then this
1166  * function reads corresponding indexing nodes and inserts them to TNC. In
1167  * case of failure, a negative error code is returned.
1168  */
ubifs_lookup_level0(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_znode ** zn,int * n)1169 int ubifs_lookup_level0(struct ubifs_info *c, const union ubifs_key *key,
1170 			struct ubifs_znode **zn, int *n)
1171 {
1172 	int err, exact;
1173 	struct ubifs_znode *znode;
1174 	time64_t time = ktime_get_seconds();
1175 
1176 	dbg_tnck(key, "search key ");
1177 	ubifs_assert(c, key_type(c, key) < UBIFS_INVALID_KEY);
1178 
1179 	znode = c->zroot.znode;
1180 	if (unlikely(!znode)) {
1181 		znode = ubifs_load_znode(c, &c->zroot, NULL, 0);
1182 		if (IS_ERR(znode))
1183 			return PTR_ERR(znode);
1184 	}
1185 
1186 	znode->time = time;
1187 
1188 	while (1) {
1189 		struct ubifs_zbranch *zbr;
1190 
1191 		exact = ubifs_search_zbranch(c, znode, key, n);
1192 
1193 		if (znode->level == 0)
1194 			break;
1195 
1196 		if (*n < 0)
1197 			*n = 0;
1198 		zbr = &znode->zbranch[*n];
1199 
1200 		if (zbr->znode) {
1201 			znode->time = time;
1202 			znode = zbr->znode;
1203 			continue;
1204 		}
1205 
1206 		/* znode is not in TNC cache, load it from the media */
1207 		znode = ubifs_load_znode(c, zbr, znode, *n);
1208 		if (IS_ERR(znode))
1209 			return PTR_ERR(znode);
1210 	}
1211 
1212 	*zn = znode;
1213 	if (exact || !is_hash_key(c, key) || *n != -1) {
1214 		dbg_tnc("found %d, lvl %d, n %d", exact, znode->level, *n);
1215 		return exact;
1216 	}
1217 
1218 	/*
1219 	 * Here is a tricky place. We have not found the key and this is a
1220 	 * "hashed" key, which may collide. The rest of the code deals with
1221 	 * situations like this:
1222 	 *
1223 	 *                  | 3 | 5 |
1224 	 *                  /       \
1225 	 *          | 3 | 5 |      | 6 | 7 | (x)
1226 	 *
1227 	 * Or more a complex example:
1228 	 *
1229 	 *                | 1 | 5 |
1230 	 *                /       \
1231 	 *       | 1 | 3 |         | 5 | 8 |
1232 	 *              \           /
1233 	 *          | 5 | 5 |   | 6 | 7 | (x)
1234 	 *
1235 	 * In the examples, if we are looking for key "5", we may reach nodes
1236 	 * marked with "(x)". In this case what we have do is to look at the
1237 	 * left and see if there is "5" key there. If there is, we have to
1238 	 * return it.
1239 	 *
1240 	 * Note, this whole situation is possible because we allow to have
1241 	 * elements which are equivalent to the next key in the parent in the
1242 	 * children of current znode. For example, this happens if we split a
1243 	 * znode like this: | 3 | 5 | 5 | 6 | 7 |, which results in something
1244 	 * like this:
1245 	 *                      | 3 | 5 |
1246 	 *                       /     \
1247 	 *                | 3 | 5 |   | 5 | 6 | 7 |
1248 	 *                              ^
1249 	 * And this becomes what is at the first "picture" after key "5" marked
1250 	 * with "^" is removed. What could be done is we could prohibit
1251 	 * splitting in the middle of the colliding sequence. Also, when
1252 	 * removing the leftmost key, we would have to correct the key of the
1253 	 * parent node, which would introduce additional complications. Namely,
1254 	 * if we changed the leftmost key of the parent znode, the garbage
1255 	 * collector would be unable to find it (GC is doing this when GC'ing
1256 	 * indexing LEBs). Although we already have an additional RB-tree where
1257 	 * we save such changed znodes (see 'ins_clr_old_idx_znode()') until
1258 	 * after the commit. But anyway, this does not look easy to implement
1259 	 * so we did not try this.
1260 	 */
1261 	err = tnc_prev(c, &znode, n);
1262 	if (err == -ENOENT) {
1263 		dbg_tnc("found 0, lvl %d, n -1", znode->level);
1264 		*n = -1;
1265 		return 0;
1266 	}
1267 	if (unlikely(err < 0))
1268 		return err;
1269 	if (keys_cmp(c, key, &znode->zbranch[*n].key)) {
1270 		dbg_tnc("found 0, lvl %d, n -1", znode->level);
1271 		*n = -1;
1272 		return 0;
1273 	}
1274 
1275 	dbg_tnc("found 1, lvl %d, n %d", znode->level, *n);
1276 	*zn = znode;
1277 	return 1;
1278 }
1279 
1280 /**
1281  * lookup_level0_dirty - search for zero-level znode dirtying.
1282  * @c: UBIFS file-system description object
1283  * @key:  key to lookup
1284  * @zn: znode is returned here
1285  * @n: znode branch slot number is returned here
1286  *
1287  * This function looks up the TNC tree and search for zero-level znode which
1288  * refers key @key. The found zero-level znode is returned in @zn. There are 3
1289  * cases:
1290  *   o exact match, i.e. the found zero-level znode contains key @key, then %1
1291  *     is returned and slot number of the matched branch is stored in @n;
1292  *   o not exact match, which means that zero-level znode does not contain @key
1293  *     then %0 is returned and slot number of the closed branch is stored in
1294  *     @n;
1295  *   o @key is so small that it is even less than the lowest key of the
1296  *     leftmost zero-level node, then %0 is returned and %-1 is stored in @n.
1297  *
1298  * Additionally all znodes in the path from the root to the located zero-level
1299  * znode are marked as dirty.
1300  *
1301  * Note, when the TNC tree is traversed, some znodes may be absent, then this
1302  * function reads corresponding indexing nodes and inserts them to TNC. In
1303  * case of failure, a negative error code is returned.
1304  */
lookup_level0_dirty(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_znode ** zn,int * n)1305 static int lookup_level0_dirty(struct ubifs_info *c, const union ubifs_key *key,
1306 			       struct ubifs_znode **zn, int *n)
1307 {
1308 	int err, exact;
1309 	struct ubifs_znode *znode;
1310 	time64_t time = ktime_get_seconds();
1311 
1312 	dbg_tnck(key, "search and dirty key ");
1313 
1314 	znode = c->zroot.znode;
1315 	if (unlikely(!znode)) {
1316 		znode = ubifs_load_znode(c, &c->zroot, NULL, 0);
1317 		if (IS_ERR(znode))
1318 			return PTR_ERR(znode);
1319 	}
1320 
1321 	znode = dirty_cow_znode(c, &c->zroot);
1322 	if (IS_ERR(znode))
1323 		return PTR_ERR(znode);
1324 
1325 	znode->time = time;
1326 
1327 	while (1) {
1328 		struct ubifs_zbranch *zbr;
1329 
1330 		exact = ubifs_search_zbranch(c, znode, key, n);
1331 
1332 		if (znode->level == 0)
1333 			break;
1334 
1335 		if (*n < 0)
1336 			*n = 0;
1337 		zbr = &znode->zbranch[*n];
1338 
1339 		if (zbr->znode) {
1340 			znode->time = time;
1341 			znode = dirty_cow_znode(c, zbr);
1342 			if (IS_ERR(znode))
1343 				return PTR_ERR(znode);
1344 			continue;
1345 		}
1346 
1347 		/* znode is not in TNC cache, load it from the media */
1348 		znode = ubifs_load_znode(c, zbr, znode, *n);
1349 		if (IS_ERR(znode))
1350 			return PTR_ERR(znode);
1351 		znode = dirty_cow_znode(c, zbr);
1352 		if (IS_ERR(znode))
1353 			return PTR_ERR(znode);
1354 	}
1355 
1356 	*zn = znode;
1357 	if (exact || !is_hash_key(c, key) || *n != -1) {
1358 		dbg_tnc("found %d, lvl %d, n %d", exact, znode->level, *n);
1359 		return exact;
1360 	}
1361 
1362 	/*
1363 	 * See huge comment at 'lookup_level0_dirty()' what is the rest of the
1364 	 * code.
1365 	 */
1366 	err = tnc_prev(c, &znode, n);
1367 	if (err == -ENOENT) {
1368 		*n = -1;
1369 		dbg_tnc("found 0, lvl %d, n -1", znode->level);
1370 		return 0;
1371 	}
1372 	if (unlikely(err < 0))
1373 		return err;
1374 	if (keys_cmp(c, key, &znode->zbranch[*n].key)) {
1375 		*n = -1;
1376 		dbg_tnc("found 0, lvl %d, n -1", znode->level);
1377 		return 0;
1378 	}
1379 
1380 	if (znode->cnext || !ubifs_zn_dirty(znode)) {
1381 		znode = dirty_cow_bottom_up(c, znode);
1382 		if (IS_ERR(znode))
1383 			return PTR_ERR(znode);
1384 	}
1385 
1386 	dbg_tnc("found 1, lvl %d, n %d", znode->level, *n);
1387 	*zn = znode;
1388 	return 1;
1389 }
1390 
1391 /**
1392  * maybe_leb_gced - determine if a LEB may have been garbage collected.
1393  * @c: UBIFS file-system description object
1394  * @lnum: LEB number
1395  * @gc_seq1: garbage collection sequence number
1396  *
1397  * This function determines if @lnum may have been garbage collected since
1398  * sequence number @gc_seq1. If it may have been then %1 is returned, otherwise
1399  * %0 is returned.
1400  */
maybe_leb_gced(struct ubifs_info * c,int lnum,int gc_seq1)1401 static int maybe_leb_gced(struct ubifs_info *c, int lnum, int gc_seq1)
1402 {
1403 	int gc_seq2, gced_lnum;
1404 
1405 	gced_lnum = c->gced_lnum;
1406 	smp_rmb();
1407 	gc_seq2 = c->gc_seq;
1408 	/* Same seq means no GC */
1409 	if (gc_seq1 == gc_seq2)
1410 		return 0;
1411 	/* Different by more than 1 means we don't know */
1412 	if (gc_seq1 + 1 != gc_seq2)
1413 		return 1;
1414 	/*
1415 	 * We have seen the sequence number has increased by 1. Now we need to
1416 	 * be sure we read the right LEB number, so read it again.
1417 	 */
1418 	smp_rmb();
1419 	if (gced_lnum != c->gced_lnum)
1420 		return 1;
1421 	/* Finally we can check lnum */
1422 	if (gced_lnum == lnum)
1423 		return 1;
1424 	return 0;
1425 }
1426 
1427 /**
1428  * ubifs_tnc_locate - look up a file-system node and return it and its location.
1429  * @c: UBIFS file-system description object
1430  * @key: node key to lookup
1431  * @node: the node is returned here
1432  * @lnum: LEB number is returned here
1433  * @offs: offset is returned here
1434  *
1435  * This function looks up and reads node with key @key. The caller has to make
1436  * sure the @node buffer is large enough to fit the node. Returns zero in case
1437  * of success, %-ENOENT if the node was not found, and a negative error code in
1438  * case of failure. The node location can be returned in @lnum and @offs.
1439  */
ubifs_tnc_locate(struct ubifs_info * c,const union ubifs_key * key,void * node,int * lnum,int * offs)1440 int ubifs_tnc_locate(struct ubifs_info *c, const union ubifs_key *key,
1441 		     void *node, int *lnum, int *offs)
1442 {
1443 	int found, n, err, safely = 0, gc_seq1;
1444 	struct ubifs_znode *znode;
1445 	struct ubifs_zbranch zbr, *zt;
1446 
1447 again:
1448 	mutex_lock(&c->tnc_mutex);
1449 	found = ubifs_lookup_level0(c, key, &znode, &n);
1450 	if (!found) {
1451 		err = -ENOENT;
1452 		goto out;
1453 	} else if (found < 0) {
1454 		err = found;
1455 		goto out;
1456 	}
1457 	zt = &znode->zbranch[n];
1458 	if (lnum) {
1459 		*lnum = zt->lnum;
1460 		*offs = zt->offs;
1461 	}
1462 	if (is_hash_key(c, key)) {
1463 		/*
1464 		 * In this case the leaf node cache gets used, so we pass the
1465 		 * address of the zbranch and keep the mutex locked
1466 		 */
1467 		err = tnc_read_hashed_node(c, zt, node);
1468 		goto out;
1469 	}
1470 	if (safely) {
1471 		err = ubifs_tnc_read_node(c, zt, node);
1472 		goto out;
1473 	}
1474 	/* Drop the TNC mutex prematurely and race with garbage collection */
1475 	zbr = znode->zbranch[n];
1476 	gc_seq1 = c->gc_seq;
1477 	mutex_unlock(&c->tnc_mutex);
1478 
1479 	if (ubifs_get_wbuf(c, zbr.lnum)) {
1480 		/* We do not GC journal heads */
1481 		err = ubifs_tnc_read_node(c, &zbr, node);
1482 		return err;
1483 	}
1484 
1485 	err = fallible_read_node(c, key, &zbr, node);
1486 	if (err <= 0 || maybe_leb_gced(c, zbr.lnum, gc_seq1)) {
1487 		/*
1488 		 * The node may have been GC'ed out from under us so try again
1489 		 * while keeping the TNC mutex locked.
1490 		 */
1491 		safely = 1;
1492 		goto again;
1493 	}
1494 	return 0;
1495 
1496 out:
1497 	mutex_unlock(&c->tnc_mutex);
1498 	return err;
1499 }
1500 
1501 /**
1502  * ubifs_tnc_get_bu_keys - lookup keys for bulk-read.
1503  * @c: UBIFS file-system description object
1504  * @bu: bulk-read parameters and results
1505  *
1506  * Lookup consecutive data node keys for the same inode that reside
1507  * consecutively in the same LEB. This function returns zero in case of success
1508  * and a negative error code in case of failure.
1509  *
1510  * Note, if the bulk-read buffer length (@bu->buf_len) is known, this function
1511  * makes sure bulk-read nodes fit the buffer. Otherwise, this function prepares
1512  * maximum possible amount of nodes for bulk-read.
1513  */
ubifs_tnc_get_bu_keys(struct ubifs_info * c,struct bu_info * bu)1514 int ubifs_tnc_get_bu_keys(struct ubifs_info *c, struct bu_info *bu)
1515 {
1516 	int n, err = 0, lnum = -1, offs;
1517 	int len;
1518 	unsigned int block = key_block(c, &bu->key);
1519 	struct ubifs_znode *znode;
1520 
1521 	bu->cnt = 0;
1522 	bu->blk_cnt = 0;
1523 	bu->eof = 0;
1524 
1525 	mutex_lock(&c->tnc_mutex);
1526 	/* Find first key */
1527 	err = ubifs_lookup_level0(c, &bu->key, &znode, &n);
1528 	if (err < 0)
1529 		goto out;
1530 	if (err) {
1531 		/* Key found */
1532 		len = znode->zbranch[n].len;
1533 		/* The buffer must be big enough for at least 1 node */
1534 		if (len > bu->buf_len) {
1535 			err = -EINVAL;
1536 			goto out;
1537 		}
1538 		/* Add this key */
1539 		bu->zbranch[bu->cnt++] = znode->zbranch[n];
1540 		bu->blk_cnt += 1;
1541 		lnum = znode->zbranch[n].lnum;
1542 		offs = ALIGN(znode->zbranch[n].offs + len, 8);
1543 	}
1544 	while (1) {
1545 		struct ubifs_zbranch *zbr;
1546 		union ubifs_key *key;
1547 		unsigned int next_block;
1548 
1549 		/* Find next key */
1550 		err = tnc_next(c, &znode, &n);
1551 		if (err)
1552 			goto out;
1553 		zbr = &znode->zbranch[n];
1554 		key = &zbr->key;
1555 		/* See if there is another data key for this file */
1556 		if (key_inum(c, key) != key_inum(c, &bu->key) ||
1557 		    key_type(c, key) != UBIFS_DATA_KEY) {
1558 			err = -ENOENT;
1559 			goto out;
1560 		}
1561 		if (lnum < 0) {
1562 			/* First key found */
1563 			lnum = zbr->lnum;
1564 			offs = ALIGN(zbr->offs + zbr->len, 8);
1565 			len = zbr->len;
1566 			if (len > bu->buf_len) {
1567 				err = -EINVAL;
1568 				goto out;
1569 			}
1570 		} else {
1571 			/*
1572 			 * The data nodes must be in consecutive positions in
1573 			 * the same LEB.
1574 			 */
1575 			if (zbr->lnum != lnum || zbr->offs != offs)
1576 				goto out;
1577 			offs += ALIGN(zbr->len, 8);
1578 			len = ALIGN(len, 8) + zbr->len;
1579 			/* Must not exceed buffer length */
1580 			if (len > bu->buf_len)
1581 				goto out;
1582 		}
1583 		/* Allow for holes */
1584 		next_block = key_block(c, key);
1585 		bu->blk_cnt += (next_block - block - 1);
1586 		if (bu->blk_cnt >= UBIFS_MAX_BULK_READ)
1587 			goto out;
1588 		block = next_block;
1589 		/* Add this key */
1590 		bu->zbranch[bu->cnt++] = *zbr;
1591 		bu->blk_cnt += 1;
1592 		/* See if we have room for more */
1593 		if (bu->cnt >= UBIFS_MAX_BULK_READ)
1594 			goto out;
1595 		if (bu->blk_cnt >= UBIFS_MAX_BULK_READ)
1596 			goto out;
1597 	}
1598 out:
1599 	if (err == -ENOENT) {
1600 		bu->eof = 1;
1601 		err = 0;
1602 	}
1603 	bu->gc_seq = c->gc_seq;
1604 	mutex_unlock(&c->tnc_mutex);
1605 	if (err)
1606 		return err;
1607 	/*
1608 	 * An enormous hole could cause bulk-read to encompass too many
1609 	 * page cache pages, so limit the number here.
1610 	 */
1611 	if (bu->blk_cnt > UBIFS_MAX_BULK_READ)
1612 		bu->blk_cnt = UBIFS_MAX_BULK_READ;
1613 	/*
1614 	 * Ensure that bulk-read covers a whole number of page cache
1615 	 * pages.
1616 	 */
1617 	if (UBIFS_BLOCKS_PER_PAGE == 1 ||
1618 	    !(bu->blk_cnt & (UBIFS_BLOCKS_PER_PAGE - 1)))
1619 		return 0;
1620 	if (bu->eof) {
1621 		/* At the end of file we can round up */
1622 		bu->blk_cnt += UBIFS_BLOCKS_PER_PAGE - 1;
1623 		return 0;
1624 	}
1625 	/* Exclude data nodes that do not make up a whole page cache page */
1626 	block = key_block(c, &bu->key) + bu->blk_cnt;
1627 	block &= ~(UBIFS_BLOCKS_PER_PAGE - 1);
1628 	while (bu->cnt) {
1629 		if (key_block(c, &bu->zbranch[bu->cnt - 1].key) < block)
1630 			break;
1631 		bu->cnt -= 1;
1632 	}
1633 	return 0;
1634 }
1635 
1636 /**
1637  * read_wbuf - bulk-read from a LEB with a wbuf.
1638  * @wbuf: wbuf that may overlap the read
1639  * @buf: buffer into which to read
1640  * @len: read length
1641  * @lnum: LEB number from which to read
1642  * @offs: offset from which to read
1643  *
1644  * This functions returns %0 on success or a negative error code on failure.
1645  */
read_wbuf(struct ubifs_wbuf * wbuf,void * buf,int len,int lnum,int offs)1646 static int read_wbuf(struct ubifs_wbuf *wbuf, void *buf, int len, int lnum,
1647 		     int offs)
1648 {
1649 	const struct ubifs_info *c = wbuf->c;
1650 	int rlen, overlap;
1651 
1652 	dbg_io("LEB %d:%d, length %d", lnum, offs, len);
1653 	ubifs_assert(c, wbuf && lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
1654 	ubifs_assert(c, !(offs & 7) && offs < c->leb_size);
1655 	ubifs_assert(c, offs + len <= c->leb_size);
1656 
1657 	spin_lock(&wbuf->lock);
1658 	overlap = (lnum == wbuf->lnum && offs + len > wbuf->offs);
1659 	if (!overlap) {
1660 		/* We may safely unlock the write-buffer and read the data */
1661 		spin_unlock(&wbuf->lock);
1662 		return ubifs_leb_read(c, lnum, buf, offs, len, 0);
1663 	}
1664 
1665 	/* Don't read under wbuf */
1666 	rlen = wbuf->offs - offs;
1667 	if (rlen < 0)
1668 		rlen = 0;
1669 
1670 	/* Copy the rest from the write-buffer */
1671 	memcpy(buf + rlen, wbuf->buf + offs + rlen - wbuf->offs, len - rlen);
1672 	spin_unlock(&wbuf->lock);
1673 
1674 	if (rlen > 0)
1675 		/* Read everything that goes before write-buffer */
1676 		return ubifs_leb_read(c, lnum, buf, offs, rlen, 0);
1677 
1678 	return 0;
1679 }
1680 
1681 /**
1682  * validate_data_node - validate data nodes for bulk-read.
1683  * @c: UBIFS file-system description object
1684  * @buf: buffer containing data node to validate
1685  * @zbr: zbranch of data node to validate
1686  *
1687  * This functions returns %0 on success or a negative error code on failure.
1688  */
validate_data_node(struct ubifs_info * c,void * buf,struct ubifs_zbranch * zbr)1689 static int validate_data_node(struct ubifs_info *c, void *buf,
1690 			      struct ubifs_zbranch *zbr)
1691 {
1692 	union ubifs_key key1;
1693 	struct ubifs_ch *ch = buf;
1694 	int err, len;
1695 
1696 	if (ch->node_type != UBIFS_DATA_NODE) {
1697 		ubifs_err(c, "bad node type (%d but expected %d)",
1698 			  ch->node_type, UBIFS_DATA_NODE);
1699 		goto out_err;
1700 	}
1701 
1702 	err = ubifs_check_node(c, buf, zbr->len, zbr->lnum, zbr->offs, 0, 0);
1703 	if (err) {
1704 		ubifs_err(c, "expected node type %d", UBIFS_DATA_NODE);
1705 		goto out;
1706 	}
1707 
1708 	err = ubifs_node_check_hash(c, buf, zbr->hash);
1709 	if (err) {
1710 		ubifs_bad_hash(c, buf, zbr->hash, zbr->lnum, zbr->offs);
1711 		return err;
1712 	}
1713 
1714 	len = le32_to_cpu(ch->len);
1715 	if (len != zbr->len) {
1716 		ubifs_err(c, "bad node length %d, expected %d", len, zbr->len);
1717 		goto out_err;
1718 	}
1719 
1720 	/* Make sure the key of the read node is correct */
1721 	key_read(c, buf + UBIFS_KEY_OFFSET, &key1);
1722 	if (!keys_eq(c, &zbr->key, &key1)) {
1723 		ubifs_err(c, "bad key in node at LEB %d:%d",
1724 			  zbr->lnum, zbr->offs);
1725 		dbg_tnck(&zbr->key, "looked for key ");
1726 		dbg_tnck(&key1, "found node's key ");
1727 		goto out_err;
1728 	}
1729 
1730 	return 0;
1731 
1732 out_err:
1733 	err = -EINVAL;
1734 out:
1735 	ubifs_err(c, "bad node at LEB %d:%d", zbr->lnum, zbr->offs);
1736 	ubifs_dump_node(c, buf, zbr->len);
1737 	dump_stack();
1738 	return err;
1739 }
1740 
1741 /**
1742  * ubifs_tnc_bulk_read - read a number of data nodes in one go.
1743  * @c: UBIFS file-system description object
1744  * @bu: bulk-read parameters and results
1745  *
1746  * This functions reads and validates the data nodes that were identified by the
1747  * 'ubifs_tnc_get_bu_keys()' function. This functions returns %0 on success,
1748  * -EAGAIN to indicate a race with GC, or another negative error code on
1749  * failure.
1750  */
ubifs_tnc_bulk_read(struct ubifs_info * c,struct bu_info * bu)1751 int ubifs_tnc_bulk_read(struct ubifs_info *c, struct bu_info *bu)
1752 {
1753 	int lnum = bu->zbranch[0].lnum, offs = bu->zbranch[0].offs, len, err, i;
1754 	struct ubifs_wbuf *wbuf;
1755 	void *buf;
1756 
1757 	len = bu->zbranch[bu->cnt - 1].offs;
1758 	len += bu->zbranch[bu->cnt - 1].len - offs;
1759 	if (len > bu->buf_len) {
1760 		ubifs_err(c, "buffer too small %d vs %d", bu->buf_len, len);
1761 		return -EINVAL;
1762 	}
1763 
1764 	/* Do the read */
1765 	wbuf = ubifs_get_wbuf(c, lnum);
1766 	if (wbuf)
1767 		err = read_wbuf(wbuf, bu->buf, len, lnum, offs);
1768 	else
1769 		err = ubifs_leb_read(c, lnum, bu->buf, offs, len, 0);
1770 
1771 	/* Check for a race with GC */
1772 	if (maybe_leb_gced(c, lnum, bu->gc_seq))
1773 		return -EAGAIN;
1774 
1775 	if (err && err != -EBADMSG) {
1776 		ubifs_err(c, "failed to read from LEB %d:%d, error %d",
1777 			  lnum, offs, err);
1778 		dump_stack();
1779 		dbg_tnck(&bu->key, "key ");
1780 		return err;
1781 	}
1782 
1783 	/* Validate the nodes read */
1784 	buf = bu->buf;
1785 	for (i = 0; i < bu->cnt; i++) {
1786 		err = validate_data_node(c, buf, &bu->zbranch[i]);
1787 		if (err)
1788 			return err;
1789 		buf = buf + ALIGN(bu->zbranch[i].len, 8);
1790 	}
1791 
1792 	return 0;
1793 }
1794 
1795 /**
1796  * do_lookup_nm- look up a "hashed" node.
1797  * @c: UBIFS file-system description object
1798  * @key: node key to lookup
1799  * @node: the node is returned here
1800  * @nm: node name
1801  *
1802  * This function looks up and reads a node which contains name hash in the key.
1803  * Since the hash may have collisions, there may be many nodes with the same
1804  * key, so we have to sequentially look to all of them until the needed one is
1805  * found. This function returns zero in case of success, %-ENOENT if the node
1806  * was not found, and a negative error code in case of failure.
1807  */
do_lookup_nm(struct ubifs_info * c,const union ubifs_key * key,void * node,const struct fscrypt_name * nm)1808 static int do_lookup_nm(struct ubifs_info *c, const union ubifs_key *key,
1809 			void *node, const struct fscrypt_name *nm)
1810 {
1811 	int found, n, err;
1812 	struct ubifs_znode *znode;
1813 
1814 	dbg_tnck(key, "key ");
1815 	mutex_lock(&c->tnc_mutex);
1816 	found = ubifs_lookup_level0(c, key, &znode, &n);
1817 	if (!found) {
1818 		err = -ENOENT;
1819 		goto out_unlock;
1820 	} else if (found < 0) {
1821 		err = found;
1822 		goto out_unlock;
1823 	}
1824 
1825 	ubifs_assert(c, n >= 0);
1826 
1827 	err = resolve_collision(c, key, &znode, &n, nm);
1828 	dbg_tnc("rc returned %d, znode %p, n %d", err, znode, n);
1829 	if (unlikely(err < 0))
1830 		goto out_unlock;
1831 	if (err == 0) {
1832 		err = -ENOENT;
1833 		goto out_unlock;
1834 	}
1835 
1836 	err = tnc_read_hashed_node(c, &znode->zbranch[n], node);
1837 
1838 out_unlock:
1839 	mutex_unlock(&c->tnc_mutex);
1840 	return err;
1841 }
1842 
1843 /**
1844  * ubifs_tnc_lookup_nm - look up a "hashed" node.
1845  * @c: UBIFS file-system description object
1846  * @key: node key to lookup
1847  * @node: the node is returned here
1848  * @nm: node name
1849  *
1850  * This function looks up and reads a node which contains name hash in the key.
1851  * Since the hash may have collisions, there may be many nodes with the same
1852  * key, so we have to sequentially look to all of them until the needed one is
1853  * found. This function returns zero in case of success, %-ENOENT if the node
1854  * was not found, and a negative error code in case of failure.
1855  */
ubifs_tnc_lookup_nm(struct ubifs_info * c,const union ubifs_key * key,void * node,const struct fscrypt_name * nm)1856 int ubifs_tnc_lookup_nm(struct ubifs_info *c, const union ubifs_key *key,
1857 			void *node, const struct fscrypt_name *nm)
1858 {
1859 	int err, len;
1860 	const struct ubifs_dent_node *dent = node;
1861 
1862 	/*
1863 	 * We assume that in most of the cases there are no name collisions and
1864 	 * 'ubifs_tnc_lookup()' returns us the right direntry.
1865 	 */
1866 	err = ubifs_tnc_lookup(c, key, node);
1867 	if (err)
1868 		return err;
1869 
1870 	len = le16_to_cpu(dent->nlen);
1871 	if (fname_len(nm) == len && !memcmp(dent->name, fname_name(nm), len))
1872 		return 0;
1873 
1874 	/*
1875 	 * Unluckily, there are hash collisions and we have to iterate over
1876 	 * them look at each direntry with colliding name hash sequentially.
1877 	 */
1878 
1879 	return do_lookup_nm(c, key, node, nm);
1880 }
1881 
search_dh_cookie(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_dent_node * dent,uint32_t cookie,struct ubifs_znode ** zn,int * n,int exact)1882 static int search_dh_cookie(struct ubifs_info *c, const union ubifs_key *key,
1883 			    struct ubifs_dent_node *dent, uint32_t cookie,
1884 			    struct ubifs_znode **zn, int *n, int exact)
1885 {
1886 	int err;
1887 	struct ubifs_znode *znode = *zn;
1888 	struct ubifs_zbranch *zbr;
1889 	union ubifs_key *dkey;
1890 
1891 	if (!exact) {
1892 		err = tnc_next(c, &znode, n);
1893 		if (err)
1894 			return err;
1895 	}
1896 
1897 	for (;;) {
1898 		zbr = &znode->zbranch[*n];
1899 		dkey = &zbr->key;
1900 
1901 		if (key_inum(c, dkey) != key_inum(c, key) ||
1902 		    key_type(c, dkey) != key_type(c, key)) {
1903 			return -ENOENT;
1904 		}
1905 
1906 		err = tnc_read_hashed_node(c, zbr, dent);
1907 		if (err)
1908 			return err;
1909 
1910 		if (key_hash(c, key) == key_hash(c, dkey) &&
1911 		    le32_to_cpu(dent->cookie) == cookie) {
1912 			*zn = znode;
1913 			return 0;
1914 		}
1915 
1916 		err = tnc_next(c, &znode, n);
1917 		if (err)
1918 			return err;
1919 	}
1920 }
1921 
do_lookup_dh(struct ubifs_info * c,const union ubifs_key * key,struct ubifs_dent_node * dent,uint32_t cookie)1922 static int do_lookup_dh(struct ubifs_info *c, const union ubifs_key *key,
1923 			struct ubifs_dent_node *dent, uint32_t cookie)
1924 {
1925 	int n, err;
1926 	struct ubifs_znode *znode;
1927 	union ubifs_key start_key;
1928 
1929 	ubifs_assert(c, is_hash_key(c, key));
1930 
1931 	lowest_dent_key(c, &start_key, key_inum(c, key));
1932 
1933 	mutex_lock(&c->tnc_mutex);
1934 	err = ubifs_lookup_level0(c, &start_key, &znode, &n);
1935 	if (unlikely(err < 0))
1936 		goto out_unlock;
1937 
1938 	err = search_dh_cookie(c, key, dent, cookie, &znode, &n, err);
1939 
1940 out_unlock:
1941 	mutex_unlock(&c->tnc_mutex);
1942 	return err;
1943 }
1944 
1945 /**
1946  * ubifs_tnc_lookup_dh - look up a "double hashed" node.
1947  * @c: UBIFS file-system description object
1948  * @key: node key to lookup
1949  * @node: the node is returned here
1950  * @cookie: node cookie for collision resolution
1951  *
1952  * This function looks up and reads a node which contains name hash in the key.
1953  * Since the hash may have collisions, there may be many nodes with the same
1954  * key, so we have to sequentially look to all of them until the needed one
1955  * with the same cookie value is found.
1956  * This function returns zero in case of success, %-ENOENT if the node
1957  * was not found, and a negative error code in case of failure.
1958  */
ubifs_tnc_lookup_dh(struct ubifs_info * c,const union ubifs_key * key,void * node,uint32_t cookie)1959 int ubifs_tnc_lookup_dh(struct ubifs_info *c, const union ubifs_key *key,
1960 			void *node, uint32_t cookie)
1961 {
1962 	int err;
1963 	const struct ubifs_dent_node *dent = node;
1964 
1965 	if (!c->double_hash)
1966 		return -EOPNOTSUPP;
1967 
1968 	/*
1969 	 * We assume that in most of the cases there are no name collisions and
1970 	 * 'ubifs_tnc_lookup()' returns us the right direntry.
1971 	 */
1972 	err = ubifs_tnc_lookup(c, key, node);
1973 	if (err)
1974 		return err;
1975 
1976 	if (le32_to_cpu(dent->cookie) == cookie)
1977 		return 0;
1978 
1979 	/*
1980 	 * Unluckily, there are hash collisions and we have to iterate over
1981 	 * them look at each direntry with colliding name hash sequentially.
1982 	 */
1983 	return do_lookup_dh(c, key, node, cookie);
1984 }
1985 
1986 /**
1987  * correct_parent_keys - correct parent znodes' keys.
1988  * @c: UBIFS file-system description object
1989  * @znode: znode to correct parent znodes for
1990  *
1991  * This is a helper function for 'tnc_insert()'. When the key of the leftmost
1992  * zbranch changes, keys of parent znodes have to be corrected. This helper
1993  * function is called in such situations and corrects the keys if needed.
1994  */
correct_parent_keys(const struct ubifs_info * c,struct ubifs_znode * znode)1995 static void correct_parent_keys(const struct ubifs_info *c,
1996 				struct ubifs_znode *znode)
1997 {
1998 	union ubifs_key *key, *key1;
1999 
2000 	ubifs_assert(c, znode->parent);
2001 	ubifs_assert(c, znode->iip == 0);
2002 
2003 	key = &znode->zbranch[0].key;
2004 	key1 = &znode->parent->zbranch[0].key;
2005 
2006 	while (keys_cmp(c, key, key1) < 0) {
2007 		key_copy(c, key, key1);
2008 		znode = znode->parent;
2009 		znode->alt = 1;
2010 		if (!znode->parent || znode->iip)
2011 			break;
2012 		key1 = &znode->parent->zbranch[0].key;
2013 	}
2014 }
2015 
2016 /**
2017  * insert_zbranch - insert a zbranch into a znode.
2018  * @c: UBIFS file-system description object
2019  * @znode: znode into which to insert
2020  * @zbr: zbranch to insert
2021  * @n: slot number to insert to
2022  *
2023  * This is a helper function for 'tnc_insert()'. UBIFS does not allow "gaps" in
2024  * znode's array of zbranches and keeps zbranches consolidated, so when a new
2025  * zbranch has to be inserted to the @znode->zbranches[]' array at the @n-th
2026  * slot, zbranches starting from @n have to be moved right.
2027  */
insert_zbranch(struct ubifs_info * c,struct ubifs_znode * znode,const struct ubifs_zbranch * zbr,int n)2028 static void insert_zbranch(struct ubifs_info *c, struct ubifs_znode *znode,
2029 			   const struct ubifs_zbranch *zbr, int n)
2030 {
2031 	int i;
2032 
2033 	ubifs_assert(c, ubifs_zn_dirty(znode));
2034 
2035 	if (znode->level) {
2036 		for (i = znode->child_cnt; i > n; i--) {
2037 			znode->zbranch[i] = znode->zbranch[i - 1];
2038 			if (znode->zbranch[i].znode)
2039 				znode->zbranch[i].znode->iip = i;
2040 		}
2041 		if (zbr->znode)
2042 			zbr->znode->iip = n;
2043 	} else
2044 		for (i = znode->child_cnt; i > n; i--)
2045 			znode->zbranch[i] = znode->zbranch[i - 1];
2046 
2047 	znode->zbranch[n] = *zbr;
2048 	znode->child_cnt += 1;
2049 
2050 	/*
2051 	 * After inserting at slot zero, the lower bound of the key range of
2052 	 * this znode may have changed. If this znode is subsequently split
2053 	 * then the upper bound of the key range may change, and furthermore
2054 	 * it could change to be lower than the original lower bound. If that
2055 	 * happens, then it will no longer be possible to find this znode in the
2056 	 * TNC using the key from the index node on flash. That is bad because
2057 	 * if it is not found, we will assume it is obsolete and may overwrite
2058 	 * it. Then if there is an unclean unmount, we will start using the
2059 	 * old index which will be broken.
2060 	 *
2061 	 * So we first mark znodes that have insertions at slot zero, and then
2062 	 * if they are split we add their lnum/offs to the old_idx tree.
2063 	 */
2064 	if (n == 0)
2065 		znode->alt = 1;
2066 }
2067 
2068 /**
2069  * tnc_insert - insert a node into TNC.
2070  * @c: UBIFS file-system description object
2071  * @znode: znode to insert into
2072  * @zbr: branch to insert
2073  * @n: slot number to insert new zbranch to
2074  *
2075  * This function inserts a new node described by @zbr into znode @znode. If
2076  * znode does not have a free slot for new zbranch, it is split. Parent znodes
2077  * are splat as well if needed. Returns zero in case of success or a negative
2078  * error code in case of failure.
2079  */
tnc_insert(struct ubifs_info * c,struct ubifs_znode * znode,struct ubifs_zbranch * zbr,int n)2080 static int tnc_insert(struct ubifs_info *c, struct ubifs_znode *znode,
2081 		      struct ubifs_zbranch *zbr, int n)
2082 {
2083 	struct ubifs_znode *zn, *zi, *zp;
2084 	int i, keep, move, appending = 0;
2085 	union ubifs_key *key = &zbr->key, *key1;
2086 
2087 	ubifs_assert(c, n >= 0 && n <= c->fanout);
2088 
2089 	/* Implement naive insert for now */
2090 again:
2091 	zp = znode->parent;
2092 	if (znode->child_cnt < c->fanout) {
2093 		ubifs_assert(c, n != c->fanout);
2094 		dbg_tnck(key, "inserted at %d level %d, key ", n, znode->level);
2095 
2096 		insert_zbranch(c, znode, zbr, n);
2097 
2098 		/* Ensure parent's key is correct */
2099 		if (n == 0 && zp && znode->iip == 0)
2100 			correct_parent_keys(c, znode);
2101 
2102 		return 0;
2103 	}
2104 
2105 	/*
2106 	 * Unfortunately, @znode does not have more empty slots and we have to
2107 	 * split it.
2108 	 */
2109 	dbg_tnck(key, "splitting level %d, key ", znode->level);
2110 
2111 	if (znode->alt)
2112 		/*
2113 		 * We can no longer be sure of finding this znode by key, so we
2114 		 * record it in the old_idx tree.
2115 		 */
2116 		ins_clr_old_idx_znode(c, znode);
2117 
2118 	zn = kzalloc(c->max_znode_sz, GFP_NOFS);
2119 	if (!zn)
2120 		return -ENOMEM;
2121 	zn->parent = zp;
2122 	zn->level = znode->level;
2123 
2124 	/* Decide where to split */
2125 	if (znode->level == 0 && key_type(c, key) == UBIFS_DATA_KEY) {
2126 		/* Try not to split consecutive data keys */
2127 		if (n == c->fanout) {
2128 			key1 = &znode->zbranch[n - 1].key;
2129 			if (key_inum(c, key1) == key_inum(c, key) &&
2130 			    key_type(c, key1) == UBIFS_DATA_KEY)
2131 				appending = 1;
2132 		} else
2133 			goto check_split;
2134 	} else if (appending && n != c->fanout) {
2135 		/* Try not to split consecutive data keys */
2136 		appending = 0;
2137 check_split:
2138 		if (n >= (c->fanout + 1) / 2) {
2139 			key1 = &znode->zbranch[0].key;
2140 			if (key_inum(c, key1) == key_inum(c, key) &&
2141 			    key_type(c, key1) == UBIFS_DATA_KEY) {
2142 				key1 = &znode->zbranch[n].key;
2143 				if (key_inum(c, key1) != key_inum(c, key) ||
2144 				    key_type(c, key1) != UBIFS_DATA_KEY) {
2145 					keep = n;
2146 					move = c->fanout - keep;
2147 					zi = znode;
2148 					goto do_split;
2149 				}
2150 			}
2151 		}
2152 	}
2153 
2154 	if (appending) {
2155 		keep = c->fanout;
2156 		move = 0;
2157 	} else {
2158 		keep = (c->fanout + 1) / 2;
2159 		move = c->fanout - keep;
2160 	}
2161 
2162 	/*
2163 	 * Although we don't at present, we could look at the neighbors and see
2164 	 * if we can move some zbranches there.
2165 	 */
2166 
2167 	if (n < keep) {
2168 		/* Insert into existing znode */
2169 		zi = znode;
2170 		move += 1;
2171 		keep -= 1;
2172 	} else {
2173 		/* Insert into new znode */
2174 		zi = zn;
2175 		n -= keep;
2176 		/* Re-parent */
2177 		if (zn->level != 0)
2178 			zbr->znode->parent = zn;
2179 	}
2180 
2181 do_split:
2182 
2183 	__set_bit(DIRTY_ZNODE, &zn->flags);
2184 	atomic_long_inc(&c->dirty_zn_cnt);
2185 
2186 	zn->child_cnt = move;
2187 	znode->child_cnt = keep;
2188 
2189 	dbg_tnc("moving %d, keeping %d", move, keep);
2190 
2191 	/* Move zbranch */
2192 	for (i = 0; i < move; i++) {
2193 		zn->zbranch[i] = znode->zbranch[keep + i];
2194 		/* Re-parent */
2195 		if (zn->level != 0)
2196 			if (zn->zbranch[i].znode) {
2197 				zn->zbranch[i].znode->parent = zn;
2198 				zn->zbranch[i].znode->iip = i;
2199 			}
2200 	}
2201 
2202 	/* Insert new key and branch */
2203 	dbg_tnck(key, "inserting at %d level %d, key ", n, zn->level);
2204 
2205 	insert_zbranch(c, zi, zbr, n);
2206 
2207 	/* Insert new znode (produced by spitting) into the parent */
2208 	if (zp) {
2209 		if (n == 0 && zi == znode && znode->iip == 0)
2210 			correct_parent_keys(c, znode);
2211 
2212 		/* Locate insertion point */
2213 		n = znode->iip + 1;
2214 
2215 		/* Tail recursion */
2216 		zbr->key = zn->zbranch[0].key;
2217 		zbr->znode = zn;
2218 		zbr->lnum = 0;
2219 		zbr->offs = 0;
2220 		zbr->len = 0;
2221 		znode = zp;
2222 
2223 		goto again;
2224 	}
2225 
2226 	/* We have to split root znode */
2227 	dbg_tnc("creating new zroot at level %d", znode->level + 1);
2228 
2229 	zi = kzalloc(c->max_znode_sz, GFP_NOFS);
2230 	if (!zi)
2231 		return -ENOMEM;
2232 
2233 	zi->child_cnt = 2;
2234 	zi->level = znode->level + 1;
2235 
2236 	__set_bit(DIRTY_ZNODE, &zi->flags);
2237 	atomic_long_inc(&c->dirty_zn_cnt);
2238 
2239 	zi->zbranch[0].key = znode->zbranch[0].key;
2240 	zi->zbranch[0].znode = znode;
2241 	zi->zbranch[0].lnum = c->zroot.lnum;
2242 	zi->zbranch[0].offs = c->zroot.offs;
2243 	zi->zbranch[0].len = c->zroot.len;
2244 	zi->zbranch[1].key = zn->zbranch[0].key;
2245 	zi->zbranch[1].znode = zn;
2246 
2247 	c->zroot.lnum = 0;
2248 	c->zroot.offs = 0;
2249 	c->zroot.len = 0;
2250 	c->zroot.znode = zi;
2251 
2252 	zn->parent = zi;
2253 	zn->iip = 1;
2254 	znode->parent = zi;
2255 	znode->iip = 0;
2256 
2257 	return 0;
2258 }
2259 
2260 /**
2261  * ubifs_tnc_add - add a node to TNC.
2262  * @c: UBIFS file-system description object
2263  * @key: key to add
2264  * @lnum: LEB number of node
2265  * @offs: node offset
2266  * @len: node length
2267  * @hash: The hash over the node
2268  *
2269  * This function adds a node with key @key to TNC. The node may be new or it may
2270  * obsolete some existing one. Returns %0 on success or negative error code on
2271  * failure.
2272  */
ubifs_tnc_add(struct ubifs_info * c,const union ubifs_key * key,int lnum,int offs,int len,const u8 * hash)2273 int ubifs_tnc_add(struct ubifs_info *c, const union ubifs_key *key, int lnum,
2274 		  int offs, int len, const u8 *hash)
2275 {
2276 	int found, n, err = 0;
2277 	struct ubifs_znode *znode;
2278 
2279 	mutex_lock(&c->tnc_mutex);
2280 	dbg_tnck(key, "%d:%d, len %d, key ", lnum, offs, len);
2281 	found = lookup_level0_dirty(c, key, &znode, &n);
2282 	if (!found) {
2283 		struct ubifs_zbranch zbr;
2284 
2285 		zbr.znode = NULL;
2286 		zbr.lnum = lnum;
2287 		zbr.offs = offs;
2288 		zbr.len = len;
2289 		ubifs_copy_hash(c, hash, zbr.hash);
2290 		key_copy(c, key, &zbr.key);
2291 		err = tnc_insert(c, znode, &zbr, n + 1);
2292 	} else if (found == 1) {
2293 		struct ubifs_zbranch *zbr = &znode->zbranch[n];
2294 
2295 		lnc_free(zbr);
2296 		err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2297 		zbr->lnum = lnum;
2298 		zbr->offs = offs;
2299 		zbr->len = len;
2300 		ubifs_copy_hash(c, hash, zbr->hash);
2301 	} else
2302 		err = found;
2303 	if (!err)
2304 		err = dbg_check_tnc(c, 0);
2305 	mutex_unlock(&c->tnc_mutex);
2306 
2307 	return err;
2308 }
2309 
2310 /**
2311  * ubifs_tnc_replace - replace a node in the TNC only if the old node is found.
2312  * @c: UBIFS file-system description object
2313  * @key: key to add
2314  * @old_lnum: LEB number of old node
2315  * @old_offs: old node offset
2316  * @lnum: LEB number of node
2317  * @offs: node offset
2318  * @len: node length
2319  *
2320  * This function replaces a node with key @key in the TNC only if the old node
2321  * is found.  This function is called by garbage collection when node are moved.
2322  * Returns %0 on success or negative error code on failure.
2323  */
ubifs_tnc_replace(struct ubifs_info * c,const union ubifs_key * key,int old_lnum,int old_offs,int lnum,int offs,int len)2324 int ubifs_tnc_replace(struct ubifs_info *c, const union ubifs_key *key,
2325 		      int old_lnum, int old_offs, int lnum, int offs, int len)
2326 {
2327 	int found, n, err = 0;
2328 	struct ubifs_znode *znode;
2329 
2330 	mutex_lock(&c->tnc_mutex);
2331 	dbg_tnck(key, "old LEB %d:%d, new LEB %d:%d, len %d, key ", old_lnum,
2332 		 old_offs, lnum, offs, len);
2333 	found = lookup_level0_dirty(c, key, &znode, &n);
2334 	if (found < 0) {
2335 		err = found;
2336 		goto out_unlock;
2337 	}
2338 
2339 	if (found == 1) {
2340 		struct ubifs_zbranch *zbr = &znode->zbranch[n];
2341 
2342 		found = 0;
2343 		if (zbr->lnum == old_lnum && zbr->offs == old_offs) {
2344 			lnc_free(zbr);
2345 			err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2346 			if (err)
2347 				goto out_unlock;
2348 			zbr->lnum = lnum;
2349 			zbr->offs = offs;
2350 			zbr->len = len;
2351 			found = 1;
2352 		} else if (is_hash_key(c, key)) {
2353 			found = resolve_collision_directly(c, key, &znode, &n,
2354 							   old_lnum, old_offs);
2355 			dbg_tnc("rc returned %d, znode %p, n %d, LEB %d:%d",
2356 				found, znode, n, old_lnum, old_offs);
2357 			if (found < 0) {
2358 				err = found;
2359 				goto out_unlock;
2360 			}
2361 
2362 			if (found) {
2363 				/* Ensure the znode is dirtied */
2364 				if (znode->cnext || !ubifs_zn_dirty(znode)) {
2365 					znode = dirty_cow_bottom_up(c, znode);
2366 					if (IS_ERR(znode)) {
2367 						err = PTR_ERR(znode);
2368 						goto out_unlock;
2369 					}
2370 				}
2371 				zbr = &znode->zbranch[n];
2372 				lnc_free(zbr);
2373 				err = ubifs_add_dirt(c, zbr->lnum,
2374 						     zbr->len);
2375 				if (err)
2376 					goto out_unlock;
2377 				zbr->lnum = lnum;
2378 				zbr->offs = offs;
2379 				zbr->len = len;
2380 			}
2381 		}
2382 	}
2383 
2384 	if (!found)
2385 		err = ubifs_add_dirt(c, lnum, len);
2386 
2387 	if (!err)
2388 		err = dbg_check_tnc(c, 0);
2389 
2390 out_unlock:
2391 	mutex_unlock(&c->tnc_mutex);
2392 	return err;
2393 }
2394 
2395 /**
2396  * ubifs_tnc_add_nm - add a "hashed" node to TNC.
2397  * @c: UBIFS file-system description object
2398  * @key: key to add
2399  * @lnum: LEB number of node
2400  * @offs: node offset
2401  * @len: node length
2402  * @hash: The hash over the node
2403  * @nm: node name
2404  *
2405  * This is the same as 'ubifs_tnc_add()' but it should be used with keys which
2406  * may have collisions, like directory entry keys.
2407  */
ubifs_tnc_add_nm(struct ubifs_info * c,const union ubifs_key * key,int lnum,int offs,int len,const u8 * hash,const struct fscrypt_name * nm)2408 int ubifs_tnc_add_nm(struct ubifs_info *c, const union ubifs_key *key,
2409 		     int lnum, int offs, int len, const u8 *hash,
2410 		     const struct fscrypt_name *nm)
2411 {
2412 	int found, n, err = 0;
2413 	struct ubifs_znode *znode;
2414 
2415 	mutex_lock(&c->tnc_mutex);
2416 	dbg_tnck(key, "LEB %d:%d, key ", lnum, offs);
2417 	found = lookup_level0_dirty(c, key, &znode, &n);
2418 	if (found < 0) {
2419 		err = found;
2420 		goto out_unlock;
2421 	}
2422 
2423 	if (found == 1) {
2424 		if (c->replaying)
2425 			found = fallible_resolve_collision(c, key, &znode, &n,
2426 							   nm, 1);
2427 		else
2428 			found = resolve_collision(c, key, &znode, &n, nm);
2429 		dbg_tnc("rc returned %d, znode %p, n %d", found, znode, n);
2430 		if (found < 0) {
2431 			err = found;
2432 			goto out_unlock;
2433 		}
2434 
2435 		/* Ensure the znode is dirtied */
2436 		if (znode->cnext || !ubifs_zn_dirty(znode)) {
2437 			znode = dirty_cow_bottom_up(c, znode);
2438 			if (IS_ERR(znode)) {
2439 				err = PTR_ERR(znode);
2440 				goto out_unlock;
2441 			}
2442 		}
2443 
2444 		if (found == 1) {
2445 			struct ubifs_zbranch *zbr = &znode->zbranch[n];
2446 
2447 			lnc_free(zbr);
2448 			err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2449 			zbr->lnum = lnum;
2450 			zbr->offs = offs;
2451 			zbr->len = len;
2452 			ubifs_copy_hash(c, hash, zbr->hash);
2453 			goto out_unlock;
2454 		}
2455 	}
2456 
2457 	if (!found) {
2458 		struct ubifs_zbranch zbr;
2459 
2460 		zbr.znode = NULL;
2461 		zbr.lnum = lnum;
2462 		zbr.offs = offs;
2463 		zbr.len = len;
2464 		ubifs_copy_hash(c, hash, zbr.hash);
2465 		key_copy(c, key, &zbr.key);
2466 		err = tnc_insert(c, znode, &zbr, n + 1);
2467 		if (err)
2468 			goto out_unlock;
2469 		if (c->replaying) {
2470 			/*
2471 			 * We did not find it in the index so there may be a
2472 			 * dangling branch still in the index. So we remove it
2473 			 * by passing 'ubifs_tnc_remove_nm()' the same key but
2474 			 * an unmatchable name.
2475 			 */
2476 			struct fscrypt_name noname = { .disk_name = { .name = "", .len = 1 } };
2477 
2478 			err = dbg_check_tnc(c, 0);
2479 			mutex_unlock(&c->tnc_mutex);
2480 			if (err)
2481 				return err;
2482 			return ubifs_tnc_remove_nm(c, key, &noname);
2483 		}
2484 	}
2485 
2486 out_unlock:
2487 	if (!err)
2488 		err = dbg_check_tnc(c, 0);
2489 	mutex_unlock(&c->tnc_mutex);
2490 	return err;
2491 }
2492 
2493 /**
2494  * tnc_delete - delete a znode form TNC.
2495  * @c: UBIFS file-system description object
2496  * @znode: znode to delete from
2497  * @n: zbranch slot number to delete
2498  *
2499  * This function deletes a leaf node from @n-th slot of @znode. Returns zero in
2500  * case of success and a negative error code in case of failure.
2501  */
tnc_delete(struct ubifs_info * c,struct ubifs_znode * znode,int n)2502 static int tnc_delete(struct ubifs_info *c, struct ubifs_znode *znode, int n)
2503 {
2504 	struct ubifs_zbranch *zbr;
2505 	struct ubifs_znode *zp;
2506 	int i, err;
2507 
2508 	/* Delete without merge for now */
2509 	ubifs_assert(c, znode->level == 0);
2510 	ubifs_assert(c, n >= 0 && n < c->fanout);
2511 	dbg_tnck(&znode->zbranch[n].key, "deleting key ");
2512 
2513 	zbr = &znode->zbranch[n];
2514 	lnc_free(zbr);
2515 
2516 	err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2517 	if (err) {
2518 		ubifs_dump_znode(c, znode);
2519 		return err;
2520 	}
2521 
2522 	/* We do not "gap" zbranch slots */
2523 	for (i = n; i < znode->child_cnt - 1; i++)
2524 		znode->zbranch[i] = znode->zbranch[i + 1];
2525 	znode->child_cnt -= 1;
2526 
2527 	if (znode->child_cnt > 0)
2528 		return 0;
2529 
2530 	/*
2531 	 * This was the last zbranch, we have to delete this znode from the
2532 	 * parent.
2533 	 */
2534 
2535 	do {
2536 		ubifs_assert(c, !ubifs_zn_obsolete(znode));
2537 		ubifs_assert(c, ubifs_zn_dirty(znode));
2538 
2539 		zp = znode->parent;
2540 		n = znode->iip;
2541 
2542 		atomic_long_dec(&c->dirty_zn_cnt);
2543 
2544 		err = insert_old_idx_znode(c, znode);
2545 		if (err)
2546 			return err;
2547 
2548 		if (znode->cnext) {
2549 			__set_bit(OBSOLETE_ZNODE, &znode->flags);
2550 			atomic_long_inc(&c->clean_zn_cnt);
2551 			atomic_long_inc(&ubifs_clean_zn_cnt);
2552 		} else
2553 			kfree(znode);
2554 		znode = zp;
2555 	} while (znode->child_cnt == 1); /* while removing last child */
2556 
2557 	/* Remove from znode, entry n - 1 */
2558 	znode->child_cnt -= 1;
2559 	ubifs_assert(c, znode->level != 0);
2560 	for (i = n; i < znode->child_cnt; i++) {
2561 		znode->zbranch[i] = znode->zbranch[i + 1];
2562 		if (znode->zbranch[i].znode)
2563 			znode->zbranch[i].znode->iip = i;
2564 	}
2565 
2566 	/*
2567 	 * If this is the root and it has only 1 child then
2568 	 * collapse the tree.
2569 	 */
2570 	if (!znode->parent) {
2571 		while (znode->child_cnt == 1 && znode->level != 0) {
2572 			zp = znode;
2573 			zbr = &znode->zbranch[0];
2574 			znode = get_znode(c, znode, 0);
2575 			if (IS_ERR(znode))
2576 				return PTR_ERR(znode);
2577 			znode = dirty_cow_znode(c, zbr);
2578 			if (IS_ERR(znode))
2579 				return PTR_ERR(znode);
2580 			znode->parent = NULL;
2581 			znode->iip = 0;
2582 			if (c->zroot.len) {
2583 				err = insert_old_idx(c, c->zroot.lnum,
2584 						     c->zroot.offs);
2585 				if (err)
2586 					return err;
2587 			}
2588 			c->zroot.lnum = zbr->lnum;
2589 			c->zroot.offs = zbr->offs;
2590 			c->zroot.len = zbr->len;
2591 			c->zroot.znode = znode;
2592 			ubifs_assert(c, !ubifs_zn_obsolete(zp));
2593 			ubifs_assert(c, ubifs_zn_dirty(zp));
2594 			atomic_long_dec(&c->dirty_zn_cnt);
2595 
2596 			if (zp->cnext) {
2597 				__set_bit(OBSOLETE_ZNODE, &zp->flags);
2598 				atomic_long_inc(&c->clean_zn_cnt);
2599 				atomic_long_inc(&ubifs_clean_zn_cnt);
2600 			} else
2601 				kfree(zp);
2602 		}
2603 	}
2604 
2605 	return 0;
2606 }
2607 
2608 /**
2609  * ubifs_tnc_remove - remove an index entry of a node.
2610  * @c: UBIFS file-system description object
2611  * @key: key of node
2612  *
2613  * Returns %0 on success or negative error code on failure.
2614  */
ubifs_tnc_remove(struct ubifs_info * c,const union ubifs_key * key)2615 int ubifs_tnc_remove(struct ubifs_info *c, const union ubifs_key *key)
2616 {
2617 	int found, n, err = 0;
2618 	struct ubifs_znode *znode;
2619 
2620 	mutex_lock(&c->tnc_mutex);
2621 	dbg_tnck(key, "key ");
2622 	found = lookup_level0_dirty(c, key, &znode, &n);
2623 	if (found < 0) {
2624 		err = found;
2625 		goto out_unlock;
2626 	}
2627 	if (found == 1)
2628 		err = tnc_delete(c, znode, n);
2629 	if (!err)
2630 		err = dbg_check_tnc(c, 0);
2631 
2632 out_unlock:
2633 	mutex_unlock(&c->tnc_mutex);
2634 	return err;
2635 }
2636 
2637 /**
2638  * ubifs_tnc_remove_nm - remove an index entry for a "hashed" node.
2639  * @c: UBIFS file-system description object
2640  * @key: key of node
2641  * @nm: directory entry name
2642  *
2643  * Returns %0 on success or negative error code on failure.
2644  */
ubifs_tnc_remove_nm(struct ubifs_info * c,const union ubifs_key * key,const struct fscrypt_name * nm)2645 int ubifs_tnc_remove_nm(struct ubifs_info *c, const union ubifs_key *key,
2646 			const struct fscrypt_name *nm)
2647 {
2648 	int n, err;
2649 	struct ubifs_znode *znode;
2650 
2651 	mutex_lock(&c->tnc_mutex);
2652 	dbg_tnck(key, "key ");
2653 	err = lookup_level0_dirty(c, key, &znode, &n);
2654 	if (err < 0)
2655 		goto out_unlock;
2656 
2657 	if (err) {
2658 		if (c->replaying)
2659 			err = fallible_resolve_collision(c, key, &znode, &n,
2660 							 nm, 0);
2661 		else
2662 			err = resolve_collision(c, key, &znode, &n, nm);
2663 		dbg_tnc("rc returned %d, znode %p, n %d", err, znode, n);
2664 		if (err < 0)
2665 			goto out_unlock;
2666 		if (err) {
2667 			/* Ensure the znode is dirtied */
2668 			if (znode->cnext || !ubifs_zn_dirty(znode)) {
2669 				znode = dirty_cow_bottom_up(c, znode);
2670 				if (IS_ERR(znode)) {
2671 					err = PTR_ERR(znode);
2672 					goto out_unlock;
2673 				}
2674 			}
2675 			err = tnc_delete(c, znode, n);
2676 		}
2677 	}
2678 
2679 out_unlock:
2680 	if (!err)
2681 		err = dbg_check_tnc(c, 0);
2682 	mutex_unlock(&c->tnc_mutex);
2683 	return err;
2684 }
2685 
2686 /**
2687  * ubifs_tnc_remove_dh - remove an index entry for a "double hashed" node.
2688  * @c: UBIFS file-system description object
2689  * @key: key of node
2690  * @cookie: node cookie for collision resolution
2691  *
2692  * Returns %0 on success or negative error code on failure.
2693  */
ubifs_tnc_remove_dh(struct ubifs_info * c,const union ubifs_key * key,uint32_t cookie)2694 int ubifs_tnc_remove_dh(struct ubifs_info *c, const union ubifs_key *key,
2695 			uint32_t cookie)
2696 {
2697 	int n, err;
2698 	struct ubifs_znode *znode;
2699 	struct ubifs_dent_node *dent;
2700 	struct ubifs_zbranch *zbr;
2701 
2702 	if (!c->double_hash)
2703 		return -EOPNOTSUPP;
2704 
2705 	mutex_lock(&c->tnc_mutex);
2706 	err = lookup_level0_dirty(c, key, &znode, &n);
2707 	if (err <= 0)
2708 		goto out_unlock;
2709 
2710 	zbr = &znode->zbranch[n];
2711 	dent = kmalloc(UBIFS_MAX_DENT_NODE_SZ, GFP_NOFS);
2712 	if (!dent) {
2713 		err = -ENOMEM;
2714 		goto out_unlock;
2715 	}
2716 
2717 	err = tnc_read_hashed_node(c, zbr, dent);
2718 	if (err)
2719 		goto out_free;
2720 
2721 	/* If the cookie does not match, we're facing a hash collision. */
2722 	if (le32_to_cpu(dent->cookie) != cookie) {
2723 		union ubifs_key start_key;
2724 
2725 		lowest_dent_key(c, &start_key, key_inum(c, key));
2726 
2727 		err = ubifs_lookup_level0(c, &start_key, &znode, &n);
2728 		if (unlikely(err < 0))
2729 			goto out_free;
2730 
2731 		err = search_dh_cookie(c, key, dent, cookie, &znode, &n, err);
2732 		if (err)
2733 			goto out_free;
2734 	}
2735 
2736 	if (znode->cnext || !ubifs_zn_dirty(znode)) {
2737 		znode = dirty_cow_bottom_up(c, znode);
2738 		if (IS_ERR(znode)) {
2739 			err = PTR_ERR(znode);
2740 			goto out_free;
2741 		}
2742 	}
2743 	err = tnc_delete(c, znode, n);
2744 
2745 out_free:
2746 	kfree(dent);
2747 out_unlock:
2748 	if (!err)
2749 		err = dbg_check_tnc(c, 0);
2750 	mutex_unlock(&c->tnc_mutex);
2751 	return err;
2752 }
2753 
2754 /**
2755  * key_in_range - determine if a key falls within a range of keys.
2756  * @c: UBIFS file-system description object
2757  * @key: key to check
2758  * @from_key: lowest key in range
2759  * @to_key: highest key in range
2760  *
2761  * This function returns %1 if the key is in range and %0 otherwise.
2762  */
key_in_range(struct ubifs_info * c,union ubifs_key * key,union ubifs_key * from_key,union ubifs_key * to_key)2763 static int key_in_range(struct ubifs_info *c, union ubifs_key *key,
2764 			union ubifs_key *from_key, union ubifs_key *to_key)
2765 {
2766 	if (keys_cmp(c, key, from_key) < 0)
2767 		return 0;
2768 	if (keys_cmp(c, key, to_key) > 0)
2769 		return 0;
2770 	return 1;
2771 }
2772 
2773 /**
2774  * ubifs_tnc_remove_range - remove index entries in range.
2775  * @c: UBIFS file-system description object
2776  * @from_key: lowest key to remove
2777  * @to_key: highest key to remove
2778  *
2779  * This function removes index entries starting at @from_key and ending at
2780  * @to_key.  This function returns zero in case of success and a negative error
2781  * code in case of failure.
2782  */
ubifs_tnc_remove_range(struct ubifs_info * c,union ubifs_key * from_key,union ubifs_key * to_key)2783 int ubifs_tnc_remove_range(struct ubifs_info *c, union ubifs_key *from_key,
2784 			   union ubifs_key *to_key)
2785 {
2786 	int i, n, k, err = 0;
2787 	struct ubifs_znode *znode;
2788 	union ubifs_key *key;
2789 
2790 	mutex_lock(&c->tnc_mutex);
2791 	while (1) {
2792 		/* Find first level 0 znode that contains keys to remove */
2793 		err = ubifs_lookup_level0(c, from_key, &znode, &n);
2794 		if (err < 0)
2795 			goto out_unlock;
2796 
2797 		if (err)
2798 			key = from_key;
2799 		else {
2800 			err = tnc_next(c, &znode, &n);
2801 			if (err == -ENOENT) {
2802 				err = 0;
2803 				goto out_unlock;
2804 			}
2805 			if (err < 0)
2806 				goto out_unlock;
2807 			key = &znode->zbranch[n].key;
2808 			if (!key_in_range(c, key, from_key, to_key)) {
2809 				err = 0;
2810 				goto out_unlock;
2811 			}
2812 		}
2813 
2814 		/* Ensure the znode is dirtied */
2815 		if (znode->cnext || !ubifs_zn_dirty(znode)) {
2816 			znode = dirty_cow_bottom_up(c, znode);
2817 			if (IS_ERR(znode)) {
2818 				err = PTR_ERR(znode);
2819 				goto out_unlock;
2820 			}
2821 		}
2822 
2823 		/* Remove all keys in range except the first */
2824 		for (i = n + 1, k = 0; i < znode->child_cnt; i++, k++) {
2825 			key = &znode->zbranch[i].key;
2826 			if (!key_in_range(c, key, from_key, to_key))
2827 				break;
2828 			lnc_free(&znode->zbranch[i]);
2829 			err = ubifs_add_dirt(c, znode->zbranch[i].lnum,
2830 					     znode->zbranch[i].len);
2831 			if (err) {
2832 				ubifs_dump_znode(c, znode);
2833 				goto out_unlock;
2834 			}
2835 			dbg_tnck(key, "removing key ");
2836 		}
2837 		if (k) {
2838 			for (i = n + 1 + k; i < znode->child_cnt; i++)
2839 				znode->zbranch[i - k] = znode->zbranch[i];
2840 			znode->child_cnt -= k;
2841 		}
2842 
2843 		/* Now delete the first */
2844 		err = tnc_delete(c, znode, n);
2845 		if (err)
2846 			goto out_unlock;
2847 	}
2848 
2849 out_unlock:
2850 	if (!err)
2851 		err = dbg_check_tnc(c, 0);
2852 	mutex_unlock(&c->tnc_mutex);
2853 	return err;
2854 }
2855 
2856 /**
2857  * ubifs_tnc_remove_ino - remove an inode from TNC.
2858  * @c: UBIFS file-system description object
2859  * @inum: inode number to remove
2860  *
2861  * This function remove inode @inum and all the extended attributes associated
2862  * with the anode from TNC and returns zero in case of success or a negative
2863  * error code in case of failure.
2864  */
ubifs_tnc_remove_ino(struct ubifs_info * c,ino_t inum)2865 int ubifs_tnc_remove_ino(struct ubifs_info *c, ino_t inum)
2866 {
2867 	union ubifs_key key1, key2;
2868 	struct ubifs_dent_node *xent, *pxent = NULL;
2869 	struct fscrypt_name nm = {0};
2870 
2871 	dbg_tnc("ino %lu", (unsigned long)inum);
2872 
2873 	/*
2874 	 * Walk all extended attribute entries and remove them together with
2875 	 * corresponding extended attribute inodes.
2876 	 */
2877 	lowest_xent_key(c, &key1, inum);
2878 	while (1) {
2879 		ino_t xattr_inum;
2880 		int err;
2881 
2882 		xent = ubifs_tnc_next_ent(c, &key1, &nm);
2883 		if (IS_ERR(xent)) {
2884 			err = PTR_ERR(xent);
2885 			if (err == -ENOENT)
2886 				break;
2887 			kfree(pxent);
2888 			return err;
2889 		}
2890 
2891 		xattr_inum = le64_to_cpu(xent->inum);
2892 		dbg_tnc("xent '%s', ino %lu", xent->name,
2893 			(unsigned long)xattr_inum);
2894 
2895 		ubifs_evict_xattr_inode(c, xattr_inum);
2896 
2897 		fname_name(&nm) = xent->name;
2898 		fname_len(&nm) = le16_to_cpu(xent->nlen);
2899 		err = ubifs_tnc_remove_nm(c, &key1, &nm);
2900 		if (err) {
2901 			kfree(pxent);
2902 			kfree(xent);
2903 			return err;
2904 		}
2905 
2906 		lowest_ino_key(c, &key1, xattr_inum);
2907 		highest_ino_key(c, &key2, xattr_inum);
2908 		err = ubifs_tnc_remove_range(c, &key1, &key2);
2909 		if (err) {
2910 			kfree(pxent);
2911 			kfree(xent);
2912 			return err;
2913 		}
2914 
2915 		kfree(pxent);
2916 		pxent = xent;
2917 		key_read(c, &xent->key, &key1);
2918 	}
2919 
2920 	kfree(pxent);
2921 	lowest_ino_key(c, &key1, inum);
2922 	highest_ino_key(c, &key2, inum);
2923 
2924 	return ubifs_tnc_remove_range(c, &key1, &key2);
2925 }
2926 
2927 /**
2928  * ubifs_tnc_next_ent - walk directory or extended attribute entries.
2929  * @c: UBIFS file-system description object
2930  * @key: key of last entry
2931  * @nm: name of last entry found or %NULL
2932  *
2933  * This function finds and reads the next directory or extended attribute entry
2934  * after the given key (@key) if there is one. @nm is used to resolve
2935  * collisions.
2936  *
2937  * If the name of the current entry is not known and only the key is known,
2938  * @nm->name has to be %NULL. In this case the semantics of this function is a
2939  * little bit different and it returns the entry corresponding to this key, not
2940  * the next one. If the key was not found, the closest "right" entry is
2941  * returned.
2942  *
2943  * If the fist entry has to be found, @key has to contain the lowest possible
2944  * key value for this inode and @name has to be %NULL.
2945  *
2946  * This function returns the found directory or extended attribute entry node
2947  * in case of success, %-ENOENT is returned if no entry was found, and a
2948  * negative error code is returned in case of failure.
2949  */
ubifs_tnc_next_ent(struct ubifs_info * c,union ubifs_key * key,const struct fscrypt_name * nm)2950 struct ubifs_dent_node *ubifs_tnc_next_ent(struct ubifs_info *c,
2951 					   union ubifs_key *key,
2952 					   const struct fscrypt_name *nm)
2953 {
2954 	int n, err, type = key_type(c, key);
2955 	struct ubifs_znode *znode;
2956 	struct ubifs_dent_node *dent;
2957 	struct ubifs_zbranch *zbr;
2958 	union ubifs_key *dkey;
2959 
2960 	dbg_tnck(key, "key ");
2961 	ubifs_assert(c, is_hash_key(c, key));
2962 
2963 	mutex_lock(&c->tnc_mutex);
2964 	err = ubifs_lookup_level0(c, key, &znode, &n);
2965 	if (unlikely(err < 0))
2966 		goto out_unlock;
2967 
2968 	if (fname_len(nm) > 0) {
2969 		if (err) {
2970 			/* Handle collisions */
2971 			if (c->replaying)
2972 				err = fallible_resolve_collision(c, key, &znode, &n,
2973 							 nm, 0);
2974 			else
2975 				err = resolve_collision(c, key, &znode, &n, nm);
2976 			dbg_tnc("rc returned %d, znode %p, n %d",
2977 				err, znode, n);
2978 			if (unlikely(err < 0))
2979 				goto out_unlock;
2980 		}
2981 
2982 		/* Now find next entry */
2983 		err = tnc_next(c, &znode, &n);
2984 		if (unlikely(err))
2985 			goto out_unlock;
2986 	} else {
2987 		/*
2988 		 * The full name of the entry was not given, in which case the
2989 		 * behavior of this function is a little different and it
2990 		 * returns current entry, not the next one.
2991 		 */
2992 		if (!err) {
2993 			/*
2994 			 * However, the given key does not exist in the TNC
2995 			 * tree and @znode/@n variables contain the closest
2996 			 * "preceding" element. Switch to the next one.
2997 			 */
2998 			err = tnc_next(c, &znode, &n);
2999 			if (err)
3000 				goto out_unlock;
3001 		}
3002 	}
3003 
3004 	zbr = &znode->zbranch[n];
3005 	dent = kmalloc(zbr->len, GFP_NOFS);
3006 	if (unlikely(!dent)) {
3007 		err = -ENOMEM;
3008 		goto out_unlock;
3009 	}
3010 
3011 	/*
3012 	 * The above 'tnc_next()' call could lead us to the next inode, check
3013 	 * this.
3014 	 */
3015 	dkey = &zbr->key;
3016 	if (key_inum(c, dkey) != key_inum(c, key) ||
3017 	    key_type(c, dkey) != type) {
3018 		err = -ENOENT;
3019 		goto out_free;
3020 	}
3021 
3022 	err = tnc_read_hashed_node(c, zbr, dent);
3023 	if (unlikely(err))
3024 		goto out_free;
3025 
3026 	mutex_unlock(&c->tnc_mutex);
3027 	return dent;
3028 
3029 out_free:
3030 	kfree(dent);
3031 out_unlock:
3032 	mutex_unlock(&c->tnc_mutex);
3033 	return ERR_PTR(err);
3034 }
3035 
3036 /**
3037  * tnc_destroy_cnext - destroy left-over obsolete znodes from a failed commit.
3038  * @c: UBIFS file-system description object
3039  *
3040  * Destroy left-over obsolete znodes from a failed commit.
3041  */
tnc_destroy_cnext(struct ubifs_info * c)3042 static void tnc_destroy_cnext(struct ubifs_info *c)
3043 {
3044 	struct ubifs_znode *cnext;
3045 
3046 	if (!c->cnext)
3047 		return;
3048 	ubifs_assert(c, c->cmt_state == COMMIT_BROKEN);
3049 	cnext = c->cnext;
3050 	do {
3051 		struct ubifs_znode *znode = cnext;
3052 
3053 		cnext = cnext->cnext;
3054 		if (ubifs_zn_obsolete(znode))
3055 			kfree(znode);
3056 	} while (cnext && cnext != c->cnext);
3057 }
3058 
3059 /**
3060  * ubifs_tnc_close - close TNC subsystem and free all related resources.
3061  * @c: UBIFS file-system description object
3062  */
ubifs_tnc_close(struct ubifs_info * c)3063 void ubifs_tnc_close(struct ubifs_info *c)
3064 {
3065 	tnc_destroy_cnext(c);
3066 	if (c->zroot.znode) {
3067 		long n, freed;
3068 
3069 		n = atomic_long_read(&c->clean_zn_cnt);
3070 		freed = ubifs_destroy_tnc_subtree(c, c->zroot.znode);
3071 		ubifs_assert(c, freed == n);
3072 		atomic_long_sub(n, &ubifs_clean_zn_cnt);
3073 	}
3074 	kfree(c->gap_lebs);
3075 	kfree(c->ilebs);
3076 	destroy_old_idx(c);
3077 }
3078 
3079 /**
3080  * left_znode - get the znode to the left.
3081  * @c: UBIFS file-system description object
3082  * @znode: znode
3083  *
3084  * This function returns a pointer to the znode to the left of @znode or NULL if
3085  * there is not one. A negative error code is returned on failure.
3086  */
left_znode(struct ubifs_info * c,struct ubifs_znode * znode)3087 static struct ubifs_znode *left_znode(struct ubifs_info *c,
3088 				      struct ubifs_znode *znode)
3089 {
3090 	int level = znode->level;
3091 
3092 	while (1) {
3093 		int n = znode->iip - 1;
3094 
3095 		/* Go up until we can go left */
3096 		znode = znode->parent;
3097 		if (!znode)
3098 			return NULL;
3099 		if (n >= 0) {
3100 			/* Now go down the rightmost branch to 'level' */
3101 			znode = get_znode(c, znode, n);
3102 			if (IS_ERR(znode))
3103 				return znode;
3104 			while (znode->level != level) {
3105 				n = znode->child_cnt - 1;
3106 				znode = get_znode(c, znode, n);
3107 				if (IS_ERR(znode))
3108 					return znode;
3109 			}
3110 			break;
3111 		}
3112 	}
3113 	return znode;
3114 }
3115 
3116 /**
3117  * right_znode - get the znode to the right.
3118  * @c: UBIFS file-system description object
3119  * @znode: znode
3120  *
3121  * This function returns a pointer to the znode to the right of @znode or NULL
3122  * if there is not one. A negative error code is returned on failure.
3123  */
right_znode(struct ubifs_info * c,struct ubifs_znode * znode)3124 static struct ubifs_znode *right_znode(struct ubifs_info *c,
3125 				       struct ubifs_znode *znode)
3126 {
3127 	int level = znode->level;
3128 
3129 	while (1) {
3130 		int n = znode->iip + 1;
3131 
3132 		/* Go up until we can go right */
3133 		znode = znode->parent;
3134 		if (!znode)
3135 			return NULL;
3136 		if (n < znode->child_cnt) {
3137 			/* Now go down the leftmost branch to 'level' */
3138 			znode = get_znode(c, znode, n);
3139 			if (IS_ERR(znode))
3140 				return znode;
3141 			while (znode->level != level) {
3142 				znode = get_znode(c, znode, 0);
3143 				if (IS_ERR(znode))
3144 					return znode;
3145 			}
3146 			break;
3147 		}
3148 	}
3149 	return znode;
3150 }
3151 
3152 /**
3153  * lookup_znode - find a particular indexing node from TNC.
3154  * @c: UBIFS file-system description object
3155  * @key: index node key to lookup
3156  * @level: index node level
3157  * @lnum: index node LEB number
3158  * @offs: index node offset
3159  *
3160  * This function searches an indexing node by its first key @key and its
3161  * address @lnum:@offs. It looks up the indexing tree by pulling all indexing
3162  * nodes it traverses to TNC. This function is called for indexing nodes which
3163  * were found on the media by scanning, for example when garbage-collecting or
3164  * when doing in-the-gaps commit. This means that the indexing node which is
3165  * looked for does not have to have exactly the same leftmost key @key, because
3166  * the leftmost key may have been changed, in which case TNC will contain a
3167  * dirty znode which still refers the same @lnum:@offs. This function is clever
3168  * enough to recognize such indexing nodes.
3169  *
3170  * Note, if a znode was deleted or changed too much, then this function will
3171  * not find it. For situations like this UBIFS has the old index RB-tree
3172  * (indexed by @lnum:@offs).
3173  *
3174  * This function returns a pointer to the znode found or %NULL if it is not
3175  * found. A negative error code is returned on failure.
3176  */
lookup_znode(struct ubifs_info * c,union ubifs_key * key,int level,int lnum,int offs)3177 static struct ubifs_znode *lookup_znode(struct ubifs_info *c,
3178 					union ubifs_key *key, int level,
3179 					int lnum, int offs)
3180 {
3181 	struct ubifs_znode *znode, *zn;
3182 	int n, nn;
3183 
3184 	ubifs_assert(c, key_type(c, key) < UBIFS_INVALID_KEY);
3185 
3186 	/*
3187 	 * The arguments have probably been read off flash, so don't assume
3188 	 * they are valid.
3189 	 */
3190 	if (level < 0)
3191 		return ERR_PTR(-EINVAL);
3192 
3193 	/* Get the root znode */
3194 	znode = c->zroot.znode;
3195 	if (!znode) {
3196 		znode = ubifs_load_znode(c, &c->zroot, NULL, 0);
3197 		if (IS_ERR(znode))
3198 			return znode;
3199 	}
3200 	/* Check if it is the one we are looking for */
3201 	if (c->zroot.lnum == lnum && c->zroot.offs == offs)
3202 		return znode;
3203 	/* Descend to the parent level i.e. (level + 1) */
3204 	if (level >= znode->level)
3205 		return NULL;
3206 	while (1) {
3207 		ubifs_search_zbranch(c, znode, key, &n);
3208 		if (n < 0) {
3209 			/*
3210 			 * We reached a znode where the leftmost key is greater
3211 			 * than the key we are searching for. This is the same
3212 			 * situation as the one described in a huge comment at
3213 			 * the end of the 'ubifs_lookup_level0()' function. And
3214 			 * for exactly the same reasons we have to try to look
3215 			 * left before giving up.
3216 			 */
3217 			znode = left_znode(c, znode);
3218 			if (!znode)
3219 				return NULL;
3220 			if (IS_ERR(znode))
3221 				return znode;
3222 			ubifs_search_zbranch(c, znode, key, &n);
3223 			ubifs_assert(c, n >= 0);
3224 		}
3225 		if (znode->level == level + 1)
3226 			break;
3227 		znode = get_znode(c, znode, n);
3228 		if (IS_ERR(znode))
3229 			return znode;
3230 	}
3231 	/* Check if the child is the one we are looking for */
3232 	if (znode->zbranch[n].lnum == lnum && znode->zbranch[n].offs == offs)
3233 		return get_znode(c, znode, n);
3234 	/* If the key is unique, there is nowhere else to look */
3235 	if (!is_hash_key(c, key))
3236 		return NULL;
3237 	/*
3238 	 * The key is not unique and so may be also in the znodes to either
3239 	 * side.
3240 	 */
3241 	zn = znode;
3242 	nn = n;
3243 	/* Look left */
3244 	while (1) {
3245 		/* Move one branch to the left */
3246 		if (n)
3247 			n -= 1;
3248 		else {
3249 			znode = left_znode(c, znode);
3250 			if (!znode)
3251 				break;
3252 			if (IS_ERR(znode))
3253 				return znode;
3254 			n = znode->child_cnt - 1;
3255 		}
3256 		/* Check it */
3257 		if (znode->zbranch[n].lnum == lnum &&
3258 		    znode->zbranch[n].offs == offs)
3259 			return get_znode(c, znode, n);
3260 		/* Stop if the key is less than the one we are looking for */
3261 		if (keys_cmp(c, &znode->zbranch[n].key, key) < 0)
3262 			break;
3263 	}
3264 	/* Back to the middle */
3265 	znode = zn;
3266 	n = nn;
3267 	/* Look right */
3268 	while (1) {
3269 		/* Move one branch to the right */
3270 		if (++n >= znode->child_cnt) {
3271 			znode = right_znode(c, znode);
3272 			if (!znode)
3273 				break;
3274 			if (IS_ERR(znode))
3275 				return znode;
3276 			n = 0;
3277 		}
3278 		/* Check it */
3279 		if (znode->zbranch[n].lnum == lnum &&
3280 		    znode->zbranch[n].offs == offs)
3281 			return get_znode(c, znode, n);
3282 		/* Stop if the key is greater than the one we are looking for */
3283 		if (keys_cmp(c, &znode->zbranch[n].key, key) > 0)
3284 			break;
3285 	}
3286 	return NULL;
3287 }
3288 
3289 /**
3290  * is_idx_node_in_tnc - determine if an index node is in the TNC.
3291  * @c: UBIFS file-system description object
3292  * @key: key of index node
3293  * @level: index node level
3294  * @lnum: LEB number of index node
3295  * @offs: offset of index node
3296  *
3297  * This function returns %0 if the index node is not referred to in the TNC, %1
3298  * if the index node is referred to in the TNC and the corresponding znode is
3299  * dirty, %2 if an index node is referred to in the TNC and the corresponding
3300  * znode is clean, and a negative error code in case of failure.
3301  *
3302  * Note, the @key argument has to be the key of the first child. Also note,
3303  * this function relies on the fact that 0:0 is never a valid LEB number and
3304  * offset for a main-area node.
3305  */
is_idx_node_in_tnc(struct ubifs_info * c,union ubifs_key * key,int level,int lnum,int offs)3306 int is_idx_node_in_tnc(struct ubifs_info *c, union ubifs_key *key, int level,
3307 		       int lnum, int offs)
3308 {
3309 	struct ubifs_znode *znode;
3310 
3311 	znode = lookup_znode(c, key, level, lnum, offs);
3312 	if (!znode)
3313 		return 0;
3314 	if (IS_ERR(znode))
3315 		return PTR_ERR(znode);
3316 
3317 	return ubifs_zn_dirty(znode) ? 1 : 2;
3318 }
3319 
3320 /**
3321  * is_leaf_node_in_tnc - determine if a non-indexing not is in the TNC.
3322  * @c: UBIFS file-system description object
3323  * @key: node key
3324  * @lnum: node LEB number
3325  * @offs: node offset
3326  *
3327  * This function returns %1 if the node is referred to in the TNC, %0 if it is
3328  * not, and a negative error code in case of failure.
3329  *
3330  * Note, this function relies on the fact that 0:0 is never a valid LEB number
3331  * and offset for a main-area node.
3332  */
is_leaf_node_in_tnc(struct ubifs_info * c,union ubifs_key * key,int lnum,int offs)3333 static int is_leaf_node_in_tnc(struct ubifs_info *c, union ubifs_key *key,
3334 			       int lnum, int offs)
3335 {
3336 	struct ubifs_zbranch *zbr;
3337 	struct ubifs_znode *znode, *zn;
3338 	int n, found, err, nn;
3339 	const int unique = !is_hash_key(c, key);
3340 
3341 	found = ubifs_lookup_level0(c, key, &znode, &n);
3342 	if (found < 0)
3343 		return found; /* Error code */
3344 	if (!found)
3345 		return 0;
3346 	zbr = &znode->zbranch[n];
3347 	if (lnum == zbr->lnum && offs == zbr->offs)
3348 		return 1; /* Found it */
3349 	if (unique)
3350 		return 0;
3351 	/*
3352 	 * Because the key is not unique, we have to look left
3353 	 * and right as well
3354 	 */
3355 	zn = znode;
3356 	nn = n;
3357 	/* Look left */
3358 	while (1) {
3359 		err = tnc_prev(c, &znode, &n);
3360 		if (err == -ENOENT)
3361 			break;
3362 		if (err)
3363 			return err;
3364 		if (keys_cmp(c, key, &znode->zbranch[n].key))
3365 			break;
3366 		zbr = &znode->zbranch[n];
3367 		if (lnum == zbr->lnum && offs == zbr->offs)
3368 			return 1; /* Found it */
3369 	}
3370 	/* Look right */
3371 	znode = zn;
3372 	n = nn;
3373 	while (1) {
3374 		err = tnc_next(c, &znode, &n);
3375 		if (err) {
3376 			if (err == -ENOENT)
3377 				return 0;
3378 			return err;
3379 		}
3380 		if (keys_cmp(c, key, &znode->zbranch[n].key))
3381 			break;
3382 		zbr = &znode->zbranch[n];
3383 		if (lnum == zbr->lnum && offs == zbr->offs)
3384 			return 1; /* Found it */
3385 	}
3386 	return 0;
3387 }
3388 
3389 /**
3390  * ubifs_tnc_has_node - determine whether a node is in the TNC.
3391  * @c: UBIFS file-system description object
3392  * @key: node key
3393  * @level: index node level (if it is an index node)
3394  * @lnum: node LEB number
3395  * @offs: node offset
3396  * @is_idx: non-zero if the node is an index node
3397  *
3398  * This function returns %1 if the node is in the TNC, %0 if it is not, and a
3399  * negative error code in case of failure. For index nodes, @key has to be the
3400  * key of the first child. An index node is considered to be in the TNC only if
3401  * the corresponding znode is clean or has not been loaded.
3402  */
ubifs_tnc_has_node(struct ubifs_info * c,union ubifs_key * key,int level,int lnum,int offs,int is_idx)3403 int ubifs_tnc_has_node(struct ubifs_info *c, union ubifs_key *key, int level,
3404 		       int lnum, int offs, int is_idx)
3405 {
3406 	int err;
3407 
3408 	mutex_lock(&c->tnc_mutex);
3409 	if (is_idx) {
3410 		err = is_idx_node_in_tnc(c, key, level, lnum, offs);
3411 		if (err < 0)
3412 			goto out_unlock;
3413 		if (err == 1)
3414 			/* The index node was found but it was dirty */
3415 			err = 0;
3416 		else if (err == 2)
3417 			/* The index node was found and it was clean */
3418 			err = 1;
3419 		else
3420 			BUG_ON(err != 0);
3421 	} else
3422 		err = is_leaf_node_in_tnc(c, key, lnum, offs);
3423 
3424 out_unlock:
3425 	mutex_unlock(&c->tnc_mutex);
3426 	return err;
3427 }
3428 
3429 /**
3430  * ubifs_dirty_idx_node - dirty an index node.
3431  * @c: UBIFS file-system description object
3432  * @key: index node key
3433  * @level: index node level
3434  * @lnum: index node LEB number
3435  * @offs: index node offset
3436  *
3437  * This function loads and dirties an index node so that it can be garbage
3438  * collected. The @key argument has to be the key of the first child. This
3439  * function relies on the fact that 0:0 is never a valid LEB number and offset
3440  * for a main-area node. Returns %0 on success and a negative error code on
3441  * failure.
3442  */
ubifs_dirty_idx_node(struct ubifs_info * c,union ubifs_key * key,int level,int lnum,int offs)3443 int ubifs_dirty_idx_node(struct ubifs_info *c, union ubifs_key *key, int level,
3444 			 int lnum, int offs)
3445 {
3446 	struct ubifs_znode *znode;
3447 	int err = 0;
3448 
3449 	mutex_lock(&c->tnc_mutex);
3450 	znode = lookup_znode(c, key, level, lnum, offs);
3451 	if (!znode)
3452 		goto out_unlock;
3453 	if (IS_ERR(znode)) {
3454 		err = PTR_ERR(znode);
3455 		goto out_unlock;
3456 	}
3457 	znode = dirty_cow_bottom_up(c, znode);
3458 	if (IS_ERR(znode)) {
3459 		err = PTR_ERR(znode);
3460 		goto out_unlock;
3461 	}
3462 
3463 out_unlock:
3464 	mutex_unlock(&c->tnc_mutex);
3465 	return err;
3466 }
3467 
3468 /**
3469  * dbg_check_inode_size - check if inode size is correct.
3470  * @c: UBIFS file-system description object
3471  * @inode: inode to check
3472  * @size: inode size
3473  *
3474  * This function makes sure that the inode size (@size) is correct and it does
3475  * not have any pages beyond @size. Returns zero if the inode is OK, %-EINVAL
3476  * if it has a data page beyond @size, and other negative error code in case of
3477  * other errors.
3478  */
dbg_check_inode_size(struct ubifs_info * c,const struct inode * inode,loff_t size)3479 int dbg_check_inode_size(struct ubifs_info *c, const struct inode *inode,
3480 			 loff_t size)
3481 {
3482 	int err, n;
3483 	union ubifs_key from_key, to_key, *key;
3484 	struct ubifs_znode *znode;
3485 	unsigned int block;
3486 
3487 	if (!S_ISREG(inode->i_mode))
3488 		return 0;
3489 	if (!dbg_is_chk_gen(c))
3490 		return 0;
3491 
3492 	block = (size + UBIFS_BLOCK_SIZE - 1) >> UBIFS_BLOCK_SHIFT;
3493 	data_key_init(c, &from_key, inode->i_ino, block);
3494 	highest_data_key(c, &to_key, inode->i_ino);
3495 
3496 	mutex_lock(&c->tnc_mutex);
3497 	err = ubifs_lookup_level0(c, &from_key, &znode, &n);
3498 	if (err < 0)
3499 		goto out_unlock;
3500 
3501 	if (err) {
3502 		key = &from_key;
3503 		goto out_dump;
3504 	}
3505 
3506 	err = tnc_next(c, &znode, &n);
3507 	if (err == -ENOENT) {
3508 		err = 0;
3509 		goto out_unlock;
3510 	}
3511 	if (err < 0)
3512 		goto out_unlock;
3513 
3514 	ubifs_assert(c, err == 0);
3515 	key = &znode->zbranch[n].key;
3516 	if (!key_in_range(c, key, &from_key, &to_key))
3517 		goto out_unlock;
3518 
3519 out_dump:
3520 	block = key_block(c, key);
3521 	ubifs_err(c, "inode %lu has size %lld, but there are data at offset %lld",
3522 		  (unsigned long)inode->i_ino, size,
3523 		  ((loff_t)block) << UBIFS_BLOCK_SHIFT);
3524 	mutex_unlock(&c->tnc_mutex);
3525 	ubifs_dump_inode(c, inode);
3526 	dump_stack();
3527 	return -EINVAL;
3528 
3529 out_unlock:
3530 	mutex_unlock(&c->tnc_mutex);
3531 	return err;
3532 }
3533