blob: ccda9387bc8f68d428e454042e43f8a682db309a [file] [log] [blame]
Stefan Roese9eefe2a2009-03-19 15:35:05 +01001/*
2 * This file is part of UBIFS.
3 *
4 * Copyright (C) 2006-2008 Nokia Corporation.
5 *
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License version 2 as published by
8 * the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * more details.
14 *
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc., 51
17 * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 *
19 * Authors: Adrian Hunter
20 * Artem Bityutskiy (Битюцкий Артём)
21 */
22
23/*
24 * This file implements TNC (Tree Node Cache) which caches indexing nodes of
25 * the UBIFS B-tree.
26 *
27 * At the moment the locking rules of the TNC tree are quite simple and
28 * straightforward. We just have a mutex and lock it when we traverse the
29 * tree. If a znode is not in memory, we read it from flash while still having
30 * the mutex locked.
31 */
32
33#include "ubifs.h"
34
35/*
36 * Returned codes of 'matches_name()' and 'fallible_matches_name()' functions.
37 * @NAME_LESS: name corresponding to the first argument is less than second
38 * @NAME_MATCHES: names match
39 * @NAME_GREATER: name corresponding to the second argument is greater than
40 * first
41 * @NOT_ON_MEDIA: node referred by zbranch does not exist on the media
42 *
43 * These constants were introduce to improve readability.
44 */
45enum {
46 NAME_LESS = 0,
47 NAME_MATCHES = 1,
48 NAME_GREATER = 2,
49 NOT_ON_MEDIA = 3,
50};
51
52/**
53 * insert_old_idx - record an index node obsoleted since the last commit start.
54 * @c: UBIFS file-system description object
55 * @lnum: LEB number of obsoleted index node
56 * @offs: offset of obsoleted index node
57 *
58 * Returns %0 on success, and a negative error code on failure.
59 *
60 * For recovery, there must always be a complete intact version of the index on
61 * flash at all times. That is called the "old index". It is the index as at the
62 * time of the last successful commit. Many of the index nodes in the old index
63 * may be dirty, but they must not be erased until the next successful commit
64 * (at which point that index becomes the old index).
65 *
66 * That means that the garbage collection and the in-the-gaps method of
67 * committing must be able to determine if an index node is in the old index.
68 * Most of the old index nodes can be found by looking up the TNC using the
69 * 'lookup_znode()' function. However, some of the old index nodes may have
70 * been deleted from the current index or may have been changed so much that
71 * they cannot be easily found. In those cases, an entry is added to an RB-tree.
72 * That is what this function does. The RB-tree is ordered by LEB number and
73 * offset because they uniquely identify the old index node.
74 */
75static int insert_old_idx(struct ubifs_info *c, int lnum, int offs)
76{
77 struct ubifs_old_idx *old_idx, *o;
78 struct rb_node **p, *parent = NULL;
79
80 old_idx = kmalloc(sizeof(struct ubifs_old_idx), GFP_NOFS);
81 if (unlikely(!old_idx))
82 return -ENOMEM;
83 old_idx->lnum = lnum;
84 old_idx->offs = offs;
85
86 p = &c->old_idx.rb_node;
87 while (*p) {
88 parent = *p;
89 o = rb_entry(parent, struct ubifs_old_idx, rb);
90 if (lnum < o->lnum)
91 p = &(*p)->rb_left;
92 else if (lnum > o->lnum)
93 p = &(*p)->rb_right;
94 else if (offs < o->offs)
95 p = &(*p)->rb_left;
96 else if (offs > o->offs)
97 p = &(*p)->rb_right;
98 else {
99 ubifs_err("old idx added twice!");
100 kfree(old_idx);
101 return 0;
102 }
103 }
104 rb_link_node(&old_idx->rb, parent, p);
105 rb_insert_color(&old_idx->rb, &c->old_idx);
106 return 0;
107}
108
109/**
110 * insert_old_idx_znode - record a znode obsoleted since last commit start.
111 * @c: UBIFS file-system description object
112 * @znode: znode of obsoleted index node
113 *
114 * Returns %0 on success, and a negative error code on failure.
115 */
116int insert_old_idx_znode(struct ubifs_info *c, struct ubifs_znode *znode)
117{
118 if (znode->parent) {
119 struct ubifs_zbranch *zbr;
120
121 zbr = &znode->parent->zbranch[znode->iip];
122 if (zbr->len)
123 return insert_old_idx(c, zbr->lnum, zbr->offs);
124 } else
125 if (c->zroot.len)
126 return insert_old_idx(c, c->zroot.lnum,
127 c->zroot.offs);
128 return 0;
129}
130
131/**
132 * ins_clr_old_idx_znode - record a znode obsoleted since last commit start.
133 * @c: UBIFS file-system description object
134 * @znode: znode of obsoleted index node
135 *
136 * Returns %0 on success, and a negative error code on failure.
137 */
138static int ins_clr_old_idx_znode(struct ubifs_info *c,
139 struct ubifs_znode *znode)
140{
141 int err;
142
143 if (znode->parent) {
144 struct ubifs_zbranch *zbr;
145
146 zbr = &znode->parent->zbranch[znode->iip];
147 if (zbr->len) {
148 err = insert_old_idx(c, zbr->lnum, zbr->offs);
149 if (err)
150 return err;
151 zbr->lnum = 0;
152 zbr->offs = 0;
153 zbr->len = 0;
154 }
155 } else
156 if (c->zroot.len) {
157 err = insert_old_idx(c, c->zroot.lnum, c->zroot.offs);
158 if (err)
159 return err;
160 c->zroot.lnum = 0;
161 c->zroot.offs = 0;
162 c->zroot.len = 0;
163 }
164 return 0;
165}
166
167/**
168 * destroy_old_idx - destroy the old_idx RB-tree.
169 * @c: UBIFS file-system description object
170 *
171 * During start commit, the old_idx RB-tree is used to avoid overwriting index
172 * nodes that were in the index last commit but have since been deleted. This
173 * is necessary for recovery i.e. the old index must be kept intact until the
174 * new index is successfully written. The old-idx RB-tree is used for the
175 * in-the-gaps method of writing index nodes and is destroyed every commit.
176 */
177void destroy_old_idx(struct ubifs_info *c)
178{
179 struct rb_node *this = c->old_idx.rb_node;
180 struct ubifs_old_idx *old_idx;
181
182 while (this) {
183 if (this->rb_left) {
184 this = this->rb_left;
185 continue;
186 } else if (this->rb_right) {
187 this = this->rb_right;
188 continue;
189 }
190 old_idx = rb_entry(this, struct ubifs_old_idx, rb);
191 this = rb_parent(this);
192 if (this) {
193 if (this->rb_left == &old_idx->rb)
194 this->rb_left = NULL;
195 else
196 this->rb_right = NULL;
197 }
198 kfree(old_idx);
199 }
200 c->old_idx = RB_ROOT;
201}
202
203/**
204 * copy_znode - copy a dirty znode.
205 * @c: UBIFS file-system description object
206 * @znode: znode to copy
207 *
208 * A dirty znode being committed may not be changed, so it is copied.
209 */
210static struct ubifs_znode *copy_znode(struct ubifs_info *c,
211 struct ubifs_znode *znode)
212{
213 struct ubifs_znode *zn;
214
215 zn = kmalloc(c->max_znode_sz, GFP_NOFS);
216 if (unlikely(!zn))
217 return ERR_PTR(-ENOMEM);
218
219 memcpy(zn, znode, c->max_znode_sz);
220 zn->cnext = NULL;
221 __set_bit(DIRTY_ZNODE, &zn->flags);
222 __clear_bit(COW_ZNODE, &zn->flags);
223
224 ubifs_assert(!test_bit(OBSOLETE_ZNODE, &znode->flags));
225 __set_bit(OBSOLETE_ZNODE, &znode->flags);
226
227 if (znode->level != 0) {
228 int i;
229 const int n = zn->child_cnt;
230
231 /* The children now have new parent */
232 for (i = 0; i < n; i++) {
233 struct ubifs_zbranch *zbr = &zn->zbranch[i];
234
235 if (zbr->znode)
236 zbr->znode->parent = zn;
237 }
238 }
239
240 atomic_long_inc(&c->dirty_zn_cnt);
241 return zn;
242}
243
244/**
245 * add_idx_dirt - add dirt due to a dirty znode.
246 * @c: UBIFS file-system description object
247 * @lnum: LEB number of index node
248 * @dirt: size of index node
249 *
250 * This function updates lprops dirty space and the new size of the index.
251 */
252static int add_idx_dirt(struct ubifs_info *c, int lnum, int dirt)
253{
254 c->calc_idx_sz -= ALIGN(dirt, 8);
255 return ubifs_add_dirt(c, lnum, dirt);
256}
257
258/**
259 * dirty_cow_znode - ensure a znode is not being committed.
260 * @c: UBIFS file-system description object
261 * @zbr: branch of znode to check
262 *
263 * Returns dirtied znode on success or negative error code on failure.
264 */
265static struct ubifs_znode *dirty_cow_znode(struct ubifs_info *c,
266 struct ubifs_zbranch *zbr)
267{
268 struct ubifs_znode *znode = zbr->znode;
269 struct ubifs_znode *zn;
270 int err;
271
272 if (!test_bit(COW_ZNODE, &znode->flags)) {
273 /* znode is not being committed */
274 if (!test_and_set_bit(DIRTY_ZNODE, &znode->flags)) {
275 atomic_long_inc(&c->dirty_zn_cnt);
276 atomic_long_dec(&c->clean_zn_cnt);
277 atomic_long_dec(&ubifs_clean_zn_cnt);
278 err = add_idx_dirt(c, zbr->lnum, zbr->len);
279 if (unlikely(err))
280 return ERR_PTR(err);
281 }
282 return znode;
283 }
284
285 zn = copy_znode(c, znode);
286 if (IS_ERR(zn))
287 return zn;
288
289 if (zbr->len) {
290 err = insert_old_idx(c, zbr->lnum, zbr->offs);
291 if (unlikely(err))
292 return ERR_PTR(err);
293 err = add_idx_dirt(c, zbr->lnum, zbr->len);
294 } else
295 err = 0;
296
297 zbr->znode = zn;
298 zbr->lnum = 0;
299 zbr->offs = 0;
300 zbr->len = 0;
301
302 if (unlikely(err))
303 return ERR_PTR(err);
304 return zn;
305}
306
307/**
308 * lnc_add - add a leaf node to the leaf node cache.
309 * @c: UBIFS file-system description object
310 * @zbr: zbranch of leaf node
311 * @node: leaf node
312 *
313 * Leaf nodes are non-index nodes directory entry nodes or data nodes. The
314 * purpose of the leaf node cache is to save re-reading the same leaf node over
315 * and over again. Most things are cached by VFS, however the file system must
316 * cache directory entries for readdir and for resolving hash collisions. The
317 * present implementation of the leaf node cache is extremely simple, and
318 * allows for error returns that are not used but that may be needed if a more
319 * complex implementation is created.
320 *
321 * Note, this function does not add the @node object to LNC directly, but
322 * allocates a copy of the object and adds the copy to LNC. The reason for this
323 * is that @node has been allocated outside of the TNC subsystem and will be
324 * used with @c->tnc_mutex unlock upon return from the TNC subsystem. But LNC
325 * may be changed at any time, e.g. freed by the shrinker.
326 */
327static int lnc_add(struct ubifs_info *c, struct ubifs_zbranch *zbr,
328 const void *node)
329{
330 int err;
331 void *lnc_node;
332 const struct ubifs_dent_node *dent = node;
333
334 ubifs_assert(!zbr->leaf);
335 ubifs_assert(zbr->len != 0);
336 ubifs_assert(is_hash_key(c, &zbr->key));
337
338 err = ubifs_validate_entry(c, dent);
339 if (err) {
340 dbg_dump_stack();
341 dbg_dump_node(c, dent);
342 return err;
343 }
344
345 lnc_node = kmalloc(zbr->len, GFP_NOFS);
346 if (!lnc_node)
347 /* We don't have to have the cache, so no error */
348 return 0;
349
350 memcpy(lnc_node, node, zbr->len);
351 zbr->leaf = lnc_node;
352 return 0;
353}
354
355 /**
356 * lnc_add_directly - add a leaf node to the leaf-node-cache.
357 * @c: UBIFS file-system description object
358 * @zbr: zbranch of leaf node
359 * @node: leaf node
360 *
361 * This function is similar to 'lnc_add()', but it does not create a copy of
362 * @node but inserts @node to TNC directly.
363 */
364static int lnc_add_directly(struct ubifs_info *c, struct ubifs_zbranch *zbr,
365 void *node)
366{
367 int err;
368
369 ubifs_assert(!zbr->leaf);
370 ubifs_assert(zbr->len != 0);
371
372 err = ubifs_validate_entry(c, node);
373 if (err) {
374 dbg_dump_stack();
375 dbg_dump_node(c, node);
376 return err;
377 }
378
379 zbr->leaf = node;
380 return 0;
381}
382
383/**
384 * lnc_free - remove a leaf node from the leaf node cache.
385 * @zbr: zbranch of leaf node
386 * @node: leaf node
387 */
388static void lnc_free(struct ubifs_zbranch *zbr)
389{
390 if (!zbr->leaf)
391 return;
392 kfree(zbr->leaf);
393 zbr->leaf = NULL;
394}
395
396/**
397 * tnc_read_node_nm - read a "hashed" leaf node.
398 * @c: UBIFS file-system description object
399 * @zbr: key and position of the node
400 * @node: node is returned here
401 *
402 * This function reads a "hashed" node defined by @zbr from the leaf node cache
403 * (in it is there) or from the hash media, in which case the node is also
404 * added to LNC. Returns zero in case of success or a negative negative error
405 * code in case of failure.
406 */
407static int tnc_read_node_nm(struct ubifs_info *c, struct ubifs_zbranch *zbr,
408 void *node)
409{
410 int err;
411
412 ubifs_assert(is_hash_key(c, &zbr->key));
413
414 if (zbr->leaf) {
415 /* Read from the leaf node cache */
416 ubifs_assert(zbr->len != 0);
417 memcpy(node, zbr->leaf, zbr->len);
418 return 0;
419 }
420
421 err = ubifs_tnc_read_node(c, zbr, node);
422 if (err)
423 return err;
424
425 /* Add the node to the leaf node cache */
426 err = lnc_add(c, zbr, node);
427 return err;
428}
429
430/**
431 * try_read_node - read a node if it is a node.
432 * @c: UBIFS file-system description object
433 * @buf: buffer to read to
434 * @type: node type
435 * @len: node length (not aligned)
436 * @lnum: LEB number of node to read
437 * @offs: offset of node to read
438 *
439 * This function tries to read a node of known type and length, checks it and
440 * stores it in @buf. This function returns %1 if a node is present and %0 if
441 * a node is not present. A negative error code is returned for I/O errors.
442 * This function performs that same function as ubifs_read_node except that
443 * it does not require that there is actually a node present and instead
444 * the return code indicates if a node was read.
445 *
446 * Note, this function does not check CRC of data nodes if @c->no_chk_data_crc
447 * is true (it is controlled by corresponding mount option). However, if
448 * @c->always_chk_crc is true, @c->no_chk_data_crc is ignored and CRC is always
449 * checked.
450 */
451static int try_read_node(const struct ubifs_info *c, void *buf, int type,
452 int len, int lnum, int offs)
453{
454 int err, node_len;
455 struct ubifs_ch *ch = buf;
456 uint32_t crc, node_crc;
457
458 dbg_io("LEB %d:%d, %s, length %d", lnum, offs, dbg_ntype(type), len);
459
460 err = ubi_read(c->ubi, lnum, buf, offs, len);
461 if (err) {
462 ubifs_err("cannot read node type %d from LEB %d:%d, error %d",
463 type, lnum, offs, err);
464 return err;
465 }
466
467 if (le32_to_cpu(ch->magic) != UBIFS_NODE_MAGIC)
468 return 0;
469
470 if (ch->node_type != type)
471 return 0;
472
473 node_len = le32_to_cpu(ch->len);
474 if (node_len != len)
475 return 0;
476
477 if (type == UBIFS_DATA_NODE && !c->always_chk_crc && c->no_chk_data_crc)
478 return 1;
479
480 crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
481 node_crc = le32_to_cpu(ch->crc);
482 if (crc != node_crc)
483 return 0;
484
485 return 1;
486}
487
488/**
489 * fallible_read_node - try to read a leaf node.
490 * @c: UBIFS file-system description object
491 * @key: key of node to read
492 * @zbr: position of node
493 * @node: node returned
494 *
495 * This function tries to read a node and returns %1 if the node is read, %0
496 * if the node is not present, and a negative error code in the case of error.
497 */
498static int fallible_read_node(struct ubifs_info *c, const union ubifs_key *key,
499 struct ubifs_zbranch *zbr, void *node)
500{
501 int ret;
502
503 dbg_tnc("LEB %d:%d, key %s", zbr->lnum, zbr->offs, DBGKEY(key));
504
505 ret = try_read_node(c, node, key_type(c, key), zbr->len, zbr->lnum,
506 zbr->offs);
507 if (ret == 1) {
508 union ubifs_key node_key;
509 struct ubifs_dent_node *dent = node;
510
511 /* All nodes have key in the same place */
512 key_read(c, &dent->key, &node_key);
513 if (keys_cmp(c, key, &node_key) != 0)
514 ret = 0;
515 }
516 if (ret == 0 && c->replaying)
517 dbg_mnt("dangling branch LEB %d:%d len %d, key %s",
518 zbr->lnum, zbr->offs, zbr->len, DBGKEY(key));
519 return ret;
520}
521
522/**
523 * matches_name - determine if a direntry or xattr entry matches a given name.
524 * @c: UBIFS file-system description object
525 * @zbr: zbranch of dent
526 * @nm: name to match
527 *
528 * This function checks if xentry/direntry referred by zbranch @zbr matches name
529 * @nm. Returns %NAME_MATCHES if it does, %NAME_LESS if the name referred by
530 * @zbr is less than @nm, and %NAME_GREATER if it is greater than @nm. In case
531 * of failure, a negative error code is returned.
532 */
533static int matches_name(struct ubifs_info *c, struct ubifs_zbranch *zbr,
534 const struct qstr *nm)
535{
536 struct ubifs_dent_node *dent;
537 int nlen, err;
538
539 /* If possible, match against the dent in the leaf node cache */
540 if (!zbr->leaf) {
541 dent = kmalloc(zbr->len, GFP_NOFS);
542 if (!dent)
543 return -ENOMEM;
544
545 err = ubifs_tnc_read_node(c, zbr, dent);
546 if (err)
547 goto out_free;
548
549 /* Add the node to the leaf node cache */
550 err = lnc_add_directly(c, zbr, dent);
551 if (err)
552 goto out_free;
553 } else
554 dent = zbr->leaf;
555
556 nlen = le16_to_cpu(dent->nlen);
557 err = memcmp(dent->name, nm->name, min_t(int, nlen, nm->len));
558 if (err == 0) {
559 if (nlen == nm->len)
560 return NAME_MATCHES;
561 else if (nlen < nm->len)
562 return NAME_LESS;
563 else
564 return NAME_GREATER;
565 } else if (err < 0)
566 return NAME_LESS;
567 else
568 return NAME_GREATER;
569
570out_free:
571 kfree(dent);
572 return err;
573}
574
575/**
576 * get_znode - get a TNC znode that may not be loaded yet.
577 * @c: UBIFS file-system description object
578 * @znode: parent znode
579 * @n: znode branch slot number
580 *
581 * This function returns the znode or a negative error code.
582 */
583static struct ubifs_znode *get_znode(struct ubifs_info *c,
584 struct ubifs_znode *znode, int n)
585{
586 struct ubifs_zbranch *zbr;
587
588 zbr = &znode->zbranch[n];
589 if (zbr->znode)
590 znode = zbr->znode;
591 else
592 znode = ubifs_load_znode(c, zbr, znode, n);
593 return znode;
594}
595
596/**
597 * tnc_next - find next TNC entry.
598 * @c: UBIFS file-system description object
599 * @zn: znode is passed and returned here
600 * @n: znode branch slot number is passed and returned here
601 *
602 * This function returns %0 if the next TNC entry is found, %-ENOENT if there is
603 * no next entry, or a negative error code otherwise.
604 */
605static int tnc_next(struct ubifs_info *c, struct ubifs_znode **zn, int *n)
606{
607 struct ubifs_znode *znode = *zn;
608 int nn = *n;
609
610 nn += 1;
611 if (nn < znode->child_cnt) {
612 *n = nn;
613 return 0;
614 }
615 while (1) {
616 struct ubifs_znode *zp;
617
618 zp = znode->parent;
619 if (!zp)
620 return -ENOENT;
621 nn = znode->iip + 1;
622 znode = zp;
623 if (nn < znode->child_cnt) {
624 znode = get_znode(c, znode, nn);
625 if (IS_ERR(znode))
626 return PTR_ERR(znode);
627 while (znode->level != 0) {
628 znode = get_znode(c, znode, 0);
629 if (IS_ERR(znode))
630 return PTR_ERR(znode);
631 }
632 nn = 0;
633 break;
634 }
635 }
636 *zn = znode;
637 *n = nn;
638 return 0;
639}
640
641/**
642 * tnc_prev - find previous TNC entry.
643 * @c: UBIFS file-system description object
644 * @zn: znode is returned here
645 * @n: znode branch slot number is passed and returned here
646 *
647 * This function returns %0 if the previous TNC entry is found, %-ENOENT if
648 * there is no next entry, or a negative error code otherwise.
649 */
650static int tnc_prev(struct ubifs_info *c, struct ubifs_znode **zn, int *n)
651{
652 struct ubifs_znode *znode = *zn;
653 int nn = *n;
654
655 if (nn > 0) {
656 *n = nn - 1;
657 return 0;
658 }
659 while (1) {
660 struct ubifs_znode *zp;
661
662 zp = znode->parent;
663 if (!zp)
664 return -ENOENT;
665 nn = znode->iip - 1;
666 znode = zp;
667 if (nn >= 0) {
668 znode = get_znode(c, znode, nn);
669 if (IS_ERR(znode))
670 return PTR_ERR(znode);
671 while (znode->level != 0) {
672 nn = znode->child_cnt - 1;
673 znode = get_znode(c, znode, nn);
674 if (IS_ERR(znode))
675 return PTR_ERR(znode);
676 }
677 nn = znode->child_cnt - 1;
678 break;
679 }
680 }
681 *zn = znode;
682 *n = nn;
683 return 0;
684}
685
686/**
687 * resolve_collision - resolve a collision.
688 * @c: UBIFS file-system description object
689 * @key: key of a directory or extended attribute entry
690 * @zn: znode is returned here
691 * @n: zbranch number is passed and returned here
692 * @nm: name of the entry
693 *
694 * This function is called for "hashed" keys to make sure that the found key
695 * really corresponds to the looked up node (directory or extended attribute
696 * entry). It returns %1 and sets @zn and @n if the collision is resolved.
697 * %0 is returned if @nm is not found and @zn and @n are set to the previous
698 * entry, i.e. to the entry after which @nm could follow if it were in TNC.
699 * This means that @n may be set to %-1 if the leftmost key in @zn is the
700 * previous one. A negative error code is returned on failures.
701 */
702static int resolve_collision(struct ubifs_info *c, const union ubifs_key *key,
703 struct ubifs_znode **zn, int *n,
704 const struct qstr *nm)
705{
706 int err;
707
708 err = matches_name(c, &(*zn)->zbranch[*n], nm);
709 if (unlikely(err < 0))
710 return err;
711 if (err == NAME_MATCHES)
712 return 1;
713
714 if (err == NAME_GREATER) {
715 /* Look left */
716 while (1) {
717 err = tnc_prev(c, zn, n);
718 if (err == -ENOENT) {
719 ubifs_assert(*n == 0);
720 *n = -1;
721 return 0;
722 }
723 if (err < 0)
724 return err;
725 if (keys_cmp(c, &(*zn)->zbranch[*n].key, key)) {
726 /*
727 * We have found the branch after which we would
728 * like to insert, but inserting in this znode
729 * may still be wrong. Consider the following 3
730 * znodes, in the case where we are resolving a
731 * collision with Key2.
732 *
733 * znode zp
734 * ----------------------
735 * level 1 | Key0 | Key1 |
736 * -----------------------
737 * | |
738 * znode za | | znode zb
739 * ------------ ------------
740 * level 0 | Key0 | | Key2 |
741 * ------------ ------------
742 *
743 * The lookup finds Key2 in znode zb. Lets say
744 * there is no match and the name is greater so
745 * we look left. When we find Key0, we end up
746 * here. If we return now, we will insert into
747 * znode za at slot n = 1. But that is invalid
748 * according to the parent's keys. Key2 must
749 * be inserted into znode zb.
750 *
751 * Note, this problem is not relevant for the
752 * case when we go right, because
753 * 'tnc_insert()' would correct the parent key.
754 */
755 if (*n == (*zn)->child_cnt - 1) {
756 err = tnc_next(c, zn, n);
757 if (err) {
758 /* Should be impossible */
759 ubifs_assert(0);
760 if (err == -ENOENT)
761 err = -EINVAL;
762 return err;
763 }
764 ubifs_assert(*n == 0);
765 *n = -1;
766 }
767 return 0;
768 }
769 err = matches_name(c, &(*zn)->zbranch[*n], nm);
770 if (err < 0)
771 return err;
772 if (err == NAME_LESS)
773 return 0;
774 if (err == NAME_MATCHES)
775 return 1;
776 ubifs_assert(err == NAME_GREATER);
777 }
778 } else {
779 int nn = *n;
780 struct ubifs_znode *znode = *zn;
781
782 /* Look right */
783 while (1) {
784 err = tnc_next(c, &znode, &nn);
785 if (err == -ENOENT)
786 return 0;
787 if (err < 0)
788 return err;
789 if (keys_cmp(c, &znode->zbranch[nn].key, key))
790 return 0;
791 err = matches_name(c, &znode->zbranch[nn], nm);
792 if (err < 0)
793 return err;
794 if (err == NAME_GREATER)
795 return 0;
796 *zn = znode;
797 *n = nn;
798 if (err == NAME_MATCHES)
799 return 1;
800 ubifs_assert(err == NAME_LESS);
801 }
802 }
803}
804
805/**
806 * fallible_matches_name - determine if a dent matches a given name.
807 * @c: UBIFS file-system description object
808 * @zbr: zbranch of dent
809 * @nm: name to match
810 *
811 * This is a "fallible" version of 'matches_name()' function which does not
812 * panic if the direntry/xentry referred by @zbr does not exist on the media.
813 *
814 * This function checks if xentry/direntry referred by zbranch @zbr matches name
815 * @nm. Returns %NAME_MATCHES it does, %NAME_LESS if the name referred by @zbr
816 * is less than @nm, %NAME_GREATER if it is greater than @nm, and @NOT_ON_MEDIA
817 * if xentry/direntry referred by @zbr does not exist on the media. A negative
818 * error code is returned in case of failure.
819 */
820static int fallible_matches_name(struct ubifs_info *c,
821 struct ubifs_zbranch *zbr,
822 const struct qstr *nm)
823{
824 struct ubifs_dent_node *dent;
825 int nlen, err;
826
827 /* If possible, match against the dent in the leaf node cache */
828 if (!zbr->leaf) {
829 dent = kmalloc(zbr->len, GFP_NOFS);
830 if (!dent)
831 return -ENOMEM;
832
833 err = fallible_read_node(c, &zbr->key, zbr, dent);
834 if (err < 0)
835 goto out_free;
836 if (err == 0) {
837 /* The node was not present */
838 err = NOT_ON_MEDIA;
839 goto out_free;
840 }
841 ubifs_assert(err == 1);
842
843 err = lnc_add_directly(c, zbr, dent);
844 if (err)
845 goto out_free;
846 } else
847 dent = zbr->leaf;
848
849 nlen = le16_to_cpu(dent->nlen);
850 err = memcmp(dent->name, nm->name, min_t(int, nlen, nm->len));
851 if (err == 0) {
852 if (nlen == nm->len)
853 return NAME_MATCHES;
854 else if (nlen < nm->len)
855 return NAME_LESS;
856 else
857 return NAME_GREATER;
858 } else if (err < 0)
859 return NAME_LESS;
860 else
861 return NAME_GREATER;
862
863out_free:
864 kfree(dent);
865 return err;
866}
867
868/**
869 * fallible_resolve_collision - resolve a collision even if nodes are missing.
870 * @c: UBIFS file-system description object
871 * @key: key
872 * @zn: znode is returned here
873 * @n: branch number is passed and returned here
874 * @nm: name of directory entry
875 * @adding: indicates caller is adding a key to the TNC
876 *
877 * This is a "fallible" version of the 'resolve_collision()' function which
878 * does not panic if one of the nodes referred to by TNC does not exist on the
879 * media. This may happen when replaying the journal if a deleted node was
880 * Garbage-collected and the commit was not done. A branch that refers to a node
881 * that is not present is called a dangling branch. The following are the return
882 * codes for this function:
883 * o if @nm was found, %1 is returned and @zn and @n are set to the found
884 * branch;
885 * o if we are @adding and @nm was not found, %0 is returned;
886 * o if we are not @adding and @nm was not found, but a dangling branch was
887 * found, then %1 is returned and @zn and @n are set to the dangling branch;
888 * o a negative error code is returned in case of failure.
889 */
890static int fallible_resolve_collision(struct ubifs_info *c,
891 const union ubifs_key *key,
892 struct ubifs_znode **zn, int *n,
893 const struct qstr *nm, int adding)
894{
895 struct ubifs_znode *o_znode = NULL, *znode = *zn;
896 int uninitialized_var(o_n), err, cmp, unsure = 0, nn = *n;
897
898 cmp = fallible_matches_name(c, &znode->zbranch[nn], nm);
899 if (unlikely(cmp < 0))
900 return cmp;
901 if (cmp == NAME_MATCHES)
902 return 1;
903 if (cmp == NOT_ON_MEDIA) {
904 o_znode = znode;
905 o_n = nn;
906 /*
907 * We are unlucky and hit a dangling branch straight away.
908 * Now we do not really know where to go to find the needed
909 * branch - to the left or to the right. Well, let's try left.
910 */
911 unsure = 1;
912 } else if (!adding)
913 unsure = 1; /* Remove a dangling branch wherever it is */
914
915 if (cmp == NAME_GREATER || unsure) {
916 /* Look left */
917 while (1) {
918 err = tnc_prev(c, zn, n);
919 if (err == -ENOENT) {
920 ubifs_assert(*n == 0);
921 *n = -1;
922 break;
923 }
924 if (err < 0)
925 return err;
926 if (keys_cmp(c, &(*zn)->zbranch[*n].key, key)) {
927 /* See comments in 'resolve_collision()' */
928 if (*n == (*zn)->child_cnt - 1) {
929 err = tnc_next(c, zn, n);
930 if (err) {
931 /* Should be impossible */
932 ubifs_assert(0);
933 if (err == -ENOENT)
934 err = -EINVAL;
935 return err;
936 }
937 ubifs_assert(*n == 0);
938 *n = -1;
939 }
940 break;
941 }
942 err = fallible_matches_name(c, &(*zn)->zbranch[*n], nm);
943 if (err < 0)
944 return err;
945 if (err == NAME_MATCHES)
946 return 1;
947 if (err == NOT_ON_MEDIA) {
948 o_znode = *zn;
949 o_n = *n;
950 continue;
951 }
952 if (!adding)
953 continue;
954 if (err == NAME_LESS)
955 break;
956 else
957 unsure = 0;
958 }
959 }
960
961 if (cmp == NAME_LESS || unsure) {
962 /* Look right */
963 *zn = znode;
964 *n = nn;
965 while (1) {
966 err = tnc_next(c, &znode, &nn);
967 if (err == -ENOENT)
968 break;
969 if (err < 0)
970 return err;
971 if (keys_cmp(c, &znode->zbranch[nn].key, key))
972 break;
973 err = fallible_matches_name(c, &znode->zbranch[nn], nm);
974 if (err < 0)
975 return err;
976 if (err == NAME_GREATER)
977 break;
978 *zn = znode;
979 *n = nn;
980 if (err == NAME_MATCHES)
981 return 1;
982 if (err == NOT_ON_MEDIA) {
983 o_znode = znode;
984 o_n = nn;
985 }
986 }
987 }
988
989 /* Never match a dangling branch when adding */
990 if (adding || !o_znode)
991 return 0;
992
993 dbg_mnt("dangling match LEB %d:%d len %d %s",
994 o_znode->zbranch[o_n].lnum, o_znode->zbranch[o_n].offs,
995 o_znode->zbranch[o_n].len, DBGKEY(key));
996 *zn = o_znode;
997 *n = o_n;
998 return 1;
999}
1000
1001/**
1002 * matches_position - determine if a zbranch matches a given position.
1003 * @zbr: zbranch of dent
1004 * @lnum: LEB number of dent to match
1005 * @offs: offset of dent to match
1006 *
1007 * This function returns %1 if @lnum:@offs matches, and %0 otherwise.
1008 */
1009static int matches_position(struct ubifs_zbranch *zbr, int lnum, int offs)
1010{
1011 if (zbr->lnum == lnum && zbr->offs == offs)
1012 return 1;
1013 else
1014 return 0;
1015}
1016
1017/**
1018 * resolve_collision_directly - resolve a collision directly.
1019 * @c: UBIFS file-system description object
1020 * @key: key of directory entry
1021 * @zn: znode is passed and returned here
1022 * @n: zbranch number is passed and returned here
1023 * @lnum: LEB number of dent node to match
1024 * @offs: offset of dent node to match
1025 *
1026 * This function is used for "hashed" keys to make sure the found directory or
1027 * extended attribute entry node is what was looked for. It is used when the
1028 * flash address of the right node is known (@lnum:@offs) which makes it much
1029 * easier to resolve collisions (no need to read entries and match full
1030 * names). This function returns %1 and sets @zn and @n if the collision is
1031 * resolved, %0 if @lnum:@offs is not found and @zn and @n are set to the
1032 * previous directory entry. Otherwise a negative error code is returned.
1033 */
1034static int resolve_collision_directly(struct ubifs_info *c,
1035 const union ubifs_key *key,
1036 struct ubifs_znode **zn, int *n,
1037 int lnum, int offs)
1038{
1039 struct ubifs_znode *znode;
1040 int nn, err;
1041
1042 znode = *zn;
1043 nn = *n;
1044 if (matches_position(&znode->zbranch[nn], lnum, offs))
1045 return 1;
1046
1047 /* Look left */
1048 while (1) {
1049 err = tnc_prev(c, &znode, &nn);
1050 if (err == -ENOENT)
1051 break;
1052 if (err < 0)
1053 return err;
1054 if (keys_cmp(c, &znode->zbranch[nn].key, key))
1055 break;
1056 if (matches_position(&znode->zbranch[nn], lnum, offs)) {
1057 *zn = znode;
1058 *n = nn;
1059 return 1;
1060 }
1061 }
1062
1063 /* Look right */
1064 znode = *zn;
1065 nn = *n;
1066 while (1) {
1067 err = tnc_next(c, &znode, &nn);
1068 if (err == -ENOENT)
1069 return 0;
1070 if (err < 0)
1071 return err;
1072 if (keys_cmp(c, &znode->zbranch[nn].key, key))
1073 return 0;
1074 *zn = znode;
1075 *n = nn;
1076 if (matches_position(&znode->zbranch[nn], lnum, offs))
1077 return 1;
1078 }
1079}
1080
1081/**
1082 * dirty_cow_bottom_up - dirty a znode and its ancestors.
1083 * @c: UBIFS file-system description object
1084 * @znode: znode to dirty
1085 *
1086 * If we do not have a unique key that resides in a znode, then we cannot
1087 * dirty that znode from the top down (i.e. by using lookup_level0_dirty)
1088 * This function records the path back to the last dirty ancestor, and then
1089 * dirties the znodes on that path.
1090 */
1091static struct ubifs_znode *dirty_cow_bottom_up(struct ubifs_info *c,
1092 struct ubifs_znode *znode)
1093{
1094 struct ubifs_znode *zp;
1095 int *path = c->bottom_up_buf, p = 0;
1096
1097 ubifs_assert(c->zroot.znode);
1098 ubifs_assert(znode);
1099 if (c->zroot.znode->level > BOTTOM_UP_HEIGHT) {
1100 kfree(c->bottom_up_buf);
1101 c->bottom_up_buf = kmalloc(c->zroot.znode->level * sizeof(int),
1102 GFP_NOFS);
1103 if (!c->bottom_up_buf)
1104 return ERR_PTR(-ENOMEM);
1105 path = c->bottom_up_buf;
1106 }
1107 if (c->zroot.znode->level) {
1108 /* Go up until parent is dirty */
1109 while (1) {
1110 int n;
1111
1112 zp = znode->parent;
1113 if (!zp)
1114 break;
1115 n = znode->iip;
1116 ubifs_assert(p < c->zroot.znode->level);
1117 path[p++] = n;
1118 if (!zp->cnext && ubifs_zn_dirty(znode))
1119 break;
1120 znode = zp;
1121 }
1122 }
1123
1124 /* Come back down, dirtying as we go */
1125 while (1) {
1126 struct ubifs_zbranch *zbr;
1127
1128 zp = znode->parent;
1129 if (zp) {
1130 ubifs_assert(path[p - 1] >= 0);
1131 ubifs_assert(path[p - 1] < zp->child_cnt);
1132 zbr = &zp->zbranch[path[--p]];
1133 znode = dirty_cow_znode(c, zbr);
1134 } else {
1135 ubifs_assert(znode == c->zroot.znode);
1136 znode = dirty_cow_znode(c, &c->zroot);
1137 }
1138 if (IS_ERR(znode) || !p)
1139 break;
1140 ubifs_assert(path[p - 1] >= 0);
1141 ubifs_assert(path[p - 1] < znode->child_cnt);
1142 znode = znode->zbranch[path[p - 1]].znode;
1143 }
1144
1145 return znode;
1146}
1147
1148/**
1149 * ubifs_lookup_level0 - search for zero-level znode.
1150 * @c: UBIFS file-system description object
1151 * @key: key to lookup
1152 * @zn: znode is returned here
1153 * @n: znode branch slot number is returned here
1154 *
1155 * This function looks up the TNC tree and search for zero-level znode which
1156 * refers key @key. The found zero-level znode is returned in @zn. There are 3
1157 * cases:
1158 * o exact match, i.e. the found zero-level znode contains key @key, then %1
1159 * is returned and slot number of the matched branch is stored in @n;
1160 * o not exact match, which means that zero-level znode does not contain
1161 * @key, then %0 is returned and slot number of the closed branch is stored
1162 * in @n;
1163 * o @key is so small that it is even less than the lowest key of the
1164 * leftmost zero-level node, then %0 is returned and %0 is stored in @n.
1165 *
1166 * Note, when the TNC tree is traversed, some znodes may be absent, then this
1167 * function reads corresponding indexing nodes and inserts them to TNC. In
1168 * case of failure, a negative error code is returned.
1169 */
1170int ubifs_lookup_level0(struct ubifs_info *c, const union ubifs_key *key,
1171 struct ubifs_znode **zn, int *n)
1172{
1173 int err, exact;
1174 struct ubifs_znode *znode;
1175 unsigned long time = get_seconds();
1176
1177 dbg_tnc("search key %s", DBGKEY(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 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 */
1305static 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 unsigned long time = get_seconds();
1311
1312 dbg_tnc("search and dirty key %s", DBGKEY(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 */
1401static int maybe_leb_gced(struct ubifs_info *c, int lnum, int gc_seq1)
1402{
1403 /*
1404 * No garbage collection in the read-only U-Boot implementation
1405 */
1406 return 0;
1407}
1408
1409/**
1410 * ubifs_tnc_locate - look up a file-system node and return it and its location.
1411 * @c: UBIFS file-system description object
1412 * @key: node key to lookup
1413 * @node: the node is returned here
1414 * @lnum: LEB number is returned here
1415 * @offs: offset is returned here
1416 *
1417 * This function look up and reads node with key @key. The caller has to make
1418 * sure the @node buffer is large enough to fit the node. Returns zero in case
1419 * of success, %-ENOENT if the node was not found, and a negative error code in
1420 * case of failure. The node location can be returned in @lnum and @offs.
1421 */
1422int ubifs_tnc_locate(struct ubifs_info *c, const union ubifs_key *key,
1423 void *node, int *lnum, int *offs)
1424{
1425 int found, n, err, safely = 0, gc_seq1;
1426 struct ubifs_znode *znode;
1427 struct ubifs_zbranch zbr, *zt;
1428
1429again:
1430 mutex_lock(&c->tnc_mutex);
1431 found = ubifs_lookup_level0(c, key, &znode, &n);
1432 if (!found) {
1433 err = -ENOENT;
1434 goto out;
1435 } else if (found < 0) {
1436 err = found;
1437 goto out;
1438 }
1439 zt = &znode->zbranch[n];
1440 if (lnum) {
1441 *lnum = zt->lnum;
1442 *offs = zt->offs;
1443 }
1444 if (is_hash_key(c, key)) {
1445 /*
1446 * In this case the leaf node cache gets used, so we pass the
1447 * address of the zbranch and keep the mutex locked
1448 */
1449 err = tnc_read_node_nm(c, zt, node);
1450 goto out;
1451 }
1452 if (safely) {
1453 err = ubifs_tnc_read_node(c, zt, node);
1454 goto out;
1455 }
1456 /* Drop the TNC mutex prematurely and race with garbage collection */
1457 zbr = znode->zbranch[n];
1458 gc_seq1 = c->gc_seq;
1459 mutex_unlock(&c->tnc_mutex);
1460
1461 err = fallible_read_node(c, key, &zbr, node);
1462 if (err <= 0 || maybe_leb_gced(c, zbr.lnum, gc_seq1)) {
1463 /*
1464 * The node may have been GC'ed out from under us so try again
1465 * while keeping the TNC mutex locked.
1466 */
1467 safely = 1;
1468 goto again;
1469 }
1470 return 0;
1471
1472out:
1473 mutex_unlock(&c->tnc_mutex);
1474 return err;
1475}
1476
1477/**
1478 * ubifs_tnc_get_bu_keys - lookup keys for bulk-read.
1479 * @c: UBIFS file-system description object
1480 * @bu: bulk-read parameters and results
1481 *
1482 * Lookup consecutive data node keys for the same inode that reside
1483 * consecutively in the same LEB. This function returns zero in case of success
1484 * and a negative error code in case of failure.
1485 *
1486 * Note, if the bulk-read buffer length (@bu->buf_len) is known, this function
1487 * makes sure bulk-read nodes fit the buffer. Otherwise, this function prepares
1488 * maximum possible amount of nodes for bulk-read.
1489 */
1490int ubifs_tnc_get_bu_keys(struct ubifs_info *c, struct bu_info *bu)
1491{
1492 int n, err = 0, lnum = -1, uninitialized_var(offs);
1493 int uninitialized_var(len);
1494 unsigned int block = key_block(c, &bu->key);
1495 struct ubifs_znode *znode;
1496
1497 bu->cnt = 0;
1498 bu->blk_cnt = 0;
1499 bu->eof = 0;
1500
1501 mutex_lock(&c->tnc_mutex);
1502 /* Find first key */
1503 err = ubifs_lookup_level0(c, &bu->key, &znode, &n);
1504 if (err < 0)
1505 goto out;
1506 if (err) {
1507 /* Key found */
1508 len = znode->zbranch[n].len;
1509 /* The buffer must be big enough for at least 1 node */
1510 if (len > bu->buf_len) {
1511 err = -EINVAL;
1512 goto out;
1513 }
1514 /* Add this key */
1515 bu->zbranch[bu->cnt++] = znode->zbranch[n];
1516 bu->blk_cnt += 1;
1517 lnum = znode->zbranch[n].lnum;
1518 offs = ALIGN(znode->zbranch[n].offs + len, 8);
1519 }
1520 while (1) {
1521 struct ubifs_zbranch *zbr;
1522 union ubifs_key *key;
1523 unsigned int next_block;
1524
1525 /* Find next key */
1526 err = tnc_next(c, &znode, &n);
1527 if (err)
1528 goto out;
1529 zbr = &znode->zbranch[n];
1530 key = &zbr->key;
1531 /* See if there is another data key for this file */
1532 if (key_inum(c, key) != key_inum(c, &bu->key) ||
1533 key_type(c, key) != UBIFS_DATA_KEY) {
1534 err = -ENOENT;
1535 goto out;
1536 }
1537 if (lnum < 0) {
1538 /* First key found */
1539 lnum = zbr->lnum;
1540 offs = ALIGN(zbr->offs + zbr->len, 8);
1541 len = zbr->len;
1542 if (len > bu->buf_len) {
1543 err = -EINVAL;
1544 goto out;
1545 }
1546 } else {
1547 /*
1548 * The data nodes must be in consecutive positions in
1549 * the same LEB.
1550 */
1551 if (zbr->lnum != lnum || zbr->offs != offs)
1552 goto out;
1553 offs += ALIGN(zbr->len, 8);
1554 len = ALIGN(len, 8) + zbr->len;
1555 /* Must not exceed buffer length */
1556 if (len > bu->buf_len)
1557 goto out;
1558 }
1559 /* Allow for holes */
1560 next_block = key_block(c, key);
1561 bu->blk_cnt += (next_block - block - 1);
1562 if (bu->blk_cnt >= UBIFS_MAX_BULK_READ)
1563 goto out;
1564 block = next_block;
1565 /* Add this key */
1566 bu->zbranch[bu->cnt++] = *zbr;
1567 bu->blk_cnt += 1;
1568 /* See if we have room for more */
1569 if (bu->cnt >= UBIFS_MAX_BULK_READ)
1570 goto out;
1571 if (bu->blk_cnt >= UBIFS_MAX_BULK_READ)
1572 goto out;
1573 }
1574out:
1575 if (err == -ENOENT) {
1576 bu->eof = 1;
1577 err = 0;
1578 }
1579 bu->gc_seq = c->gc_seq;
1580 mutex_unlock(&c->tnc_mutex);
1581 if (err)
1582 return err;
1583 /*
1584 * An enormous hole could cause bulk-read to encompass too many
1585 * page cache pages, so limit the number here.
1586 */
1587 if (bu->blk_cnt > UBIFS_MAX_BULK_READ)
1588 bu->blk_cnt = UBIFS_MAX_BULK_READ;
1589 /*
1590 * Ensure that bulk-read covers a whole number of page cache
1591 * pages.
1592 */
1593 if (UBIFS_BLOCKS_PER_PAGE == 1 ||
1594 !(bu->blk_cnt & (UBIFS_BLOCKS_PER_PAGE - 1)))
1595 return 0;
1596 if (bu->eof) {
1597 /* At the end of file we can round up */
1598 bu->blk_cnt += UBIFS_BLOCKS_PER_PAGE - 1;
1599 return 0;
1600 }
1601 /* Exclude data nodes that do not make up a whole page cache page */
1602 block = key_block(c, &bu->key) + bu->blk_cnt;
1603 block &= ~(UBIFS_BLOCKS_PER_PAGE - 1);
1604 while (bu->cnt) {
1605 if (key_block(c, &bu->zbranch[bu->cnt - 1].key) < block)
1606 break;
1607 bu->cnt -= 1;
1608 }
1609 return 0;
1610}
1611
1612/**
1613 * validate_data_node - validate data nodes for bulk-read.
1614 * @c: UBIFS file-system description object
1615 * @buf: buffer containing data node to validate
1616 * @zbr: zbranch of data node to validate
1617 *
1618 * This functions returns %0 on success or a negative error code on failure.
1619 */
1620static int validate_data_node(struct ubifs_info *c, void *buf,
1621 struct ubifs_zbranch *zbr)
1622{
1623 union ubifs_key key1;
1624 struct ubifs_ch *ch = buf;
1625 int err, len;
1626
1627 if (ch->node_type != UBIFS_DATA_NODE) {
1628 ubifs_err("bad node type (%d but expected %d)",
1629 ch->node_type, UBIFS_DATA_NODE);
1630 goto out_err;
1631 }
1632
1633 err = ubifs_check_node(c, buf, zbr->lnum, zbr->offs, 0, 0);
1634 if (err) {
1635 ubifs_err("expected node type %d", UBIFS_DATA_NODE);
1636 goto out;
1637 }
1638
1639 len = le32_to_cpu(ch->len);
1640 if (len != zbr->len) {
1641 ubifs_err("bad node length %d, expected %d", len, zbr->len);
1642 goto out_err;
1643 }
1644
1645 /* Make sure the key of the read node is correct */
1646 key_read(c, buf + UBIFS_KEY_OFFSET, &key1);
1647 if (!keys_eq(c, &zbr->key, &key1)) {
1648 ubifs_err("bad key in node at LEB %d:%d",
1649 zbr->lnum, zbr->offs);
1650 dbg_tnc("looked for key %s found node's key %s",
1651 DBGKEY(&zbr->key), DBGKEY1(&key1));
1652 goto out_err;
1653 }
1654
1655 return 0;
1656
1657out_err:
1658 err = -EINVAL;
1659out:
1660 ubifs_err("bad node at LEB %d:%d", zbr->lnum, zbr->offs);
1661 dbg_dump_node(c, buf);
1662 dbg_dump_stack();
1663 return err;
1664}
1665
1666/**
1667 * ubifs_tnc_bulk_read - read a number of data nodes in one go.
1668 * @c: UBIFS file-system description object
1669 * @bu: bulk-read parameters and results
1670 *
1671 * This functions reads and validates the data nodes that were identified by the
1672 * 'ubifs_tnc_get_bu_keys()' function. This functions returns %0 on success,
1673 * -EAGAIN to indicate a race with GC, or another negative error code on
1674 * failure.
1675 */
1676int ubifs_tnc_bulk_read(struct ubifs_info *c, struct bu_info *bu)
1677{
1678 int lnum = bu->zbranch[0].lnum, offs = bu->zbranch[0].offs, len, err, i;
1679 void *buf;
1680
1681 len = bu->zbranch[bu->cnt - 1].offs;
1682 len += bu->zbranch[bu->cnt - 1].len - offs;
1683 if (len > bu->buf_len) {
1684 ubifs_err("buffer too small %d vs %d", bu->buf_len, len);
1685 return -EINVAL;
1686 }
1687
1688 /* Do the read */
1689 err = ubi_read(c->ubi, lnum, bu->buf, offs, len);
1690
1691 /* Check for a race with GC */
1692 if (maybe_leb_gced(c, lnum, bu->gc_seq))
1693 return -EAGAIN;
1694
1695 if (err && err != -EBADMSG) {
1696 ubifs_err("failed to read from LEB %d:%d, error %d",
1697 lnum, offs, err);
1698 dbg_dump_stack();
1699 dbg_tnc("key %s", DBGKEY(&bu->key));
1700 return err;
1701 }
1702
1703 /* Validate the nodes read */
1704 buf = bu->buf;
1705 for (i = 0; i < bu->cnt; i++) {
1706 err = validate_data_node(c, buf, &bu->zbranch[i]);
1707 if (err)
1708 return err;
1709 buf = buf + ALIGN(bu->zbranch[i].len, 8);
1710 }
1711
1712 return 0;
1713}
1714
1715/**
1716 * do_lookup_nm- look up a "hashed" node.
1717 * @c: UBIFS file-system description object
1718 * @key: node key to lookup
1719 * @node: the node is returned here
1720 * @nm: node name
1721 *
1722 * This function look up and reads a node which contains name hash in the key.
1723 * Since the hash may have collisions, there may be many nodes with the same
1724 * key, so we have to sequentially look to all of them until the needed one is
1725 * found. This function returns zero in case of success, %-ENOENT if the node
1726 * was not found, and a negative error code in case of failure.
1727 */
1728static int do_lookup_nm(struct ubifs_info *c, const union ubifs_key *key,
1729 void *node, const struct qstr *nm)
1730{
1731 int found, n, err;
1732 struct ubifs_znode *znode;
1733
1734 dbg_tnc("name '%.*s' key %s", nm->len, nm->name, DBGKEY(key));
1735 mutex_lock(&c->tnc_mutex);
1736 found = ubifs_lookup_level0(c, key, &znode, &n);
1737 if (!found) {
1738 err = -ENOENT;
1739 goto out_unlock;
1740 } else if (found < 0) {
1741 err = found;
1742 goto out_unlock;
1743 }
1744
1745 ubifs_assert(n >= 0);
1746
1747 err = resolve_collision(c, key, &znode, &n, nm);
1748 dbg_tnc("rc returned %d, znode %p, n %d", err, znode, n);
1749 if (unlikely(err < 0))
1750 goto out_unlock;
1751 if (err == 0) {
1752 err = -ENOENT;
1753 goto out_unlock;
1754 }
1755
1756 err = tnc_read_node_nm(c, &znode->zbranch[n], node);
1757
1758out_unlock:
1759 mutex_unlock(&c->tnc_mutex);
1760 return err;
1761}
1762
1763/**
1764 * ubifs_tnc_lookup_nm - look up a "hashed" node.
1765 * @c: UBIFS file-system description object
1766 * @key: node key to lookup
1767 * @node: the node is returned here
1768 * @nm: node name
1769 *
1770 * This function look up and reads a node which contains name hash in the key.
1771 * Since the hash may have collisions, there may be many nodes with the same
1772 * key, so we have to sequentially look to all of them until the needed one is
1773 * found. This function returns zero in case of success, %-ENOENT if the node
1774 * was not found, and a negative error code in case of failure.
1775 */
1776int ubifs_tnc_lookup_nm(struct ubifs_info *c, const union ubifs_key *key,
1777 void *node, const struct qstr *nm)
1778{
1779 int err, len;
1780 const struct ubifs_dent_node *dent = node;
1781
1782 /*
1783 * We assume that in most of the cases there are no name collisions and
1784 * 'ubifs_tnc_lookup()' returns us the right direntry.
1785 */
1786 err = ubifs_tnc_lookup(c, key, node);
1787 if (err)
1788 return err;
1789
1790 len = le16_to_cpu(dent->nlen);
1791 if (nm->len == len && !memcmp(dent->name, nm->name, len))
1792 return 0;
1793
1794 /*
1795 * Unluckily, there are hash collisions and we have to iterate over
1796 * them look at each direntry with colliding name hash sequentially.
1797 */
1798 return do_lookup_nm(c, key, node, nm);
1799}
1800
1801/**
1802 * correct_parent_keys - correct parent znodes' keys.
1803 * @c: UBIFS file-system description object
1804 * @znode: znode to correct parent znodes for
1805 *
1806 * This is a helper function for 'tnc_insert()'. When the key of the leftmost
1807 * zbranch changes, keys of parent znodes have to be corrected. This helper
1808 * function is called in such situations and corrects the keys if needed.
1809 */
1810static void correct_parent_keys(const struct ubifs_info *c,
1811 struct ubifs_znode *znode)
1812{
1813 union ubifs_key *key, *key1;
1814
1815 ubifs_assert(znode->parent);
1816 ubifs_assert(znode->iip == 0);
1817
1818 key = &znode->zbranch[0].key;
1819 key1 = &znode->parent->zbranch[0].key;
1820
1821 while (keys_cmp(c, key, key1) < 0) {
1822 key_copy(c, key, key1);
1823 znode = znode->parent;
1824 znode->alt = 1;
1825 if (!znode->parent || znode->iip)
1826 break;
1827 key1 = &znode->parent->zbranch[0].key;
1828 }
1829}
1830
1831/**
1832 * insert_zbranch - insert a zbranch into a znode.
1833 * @znode: znode into which to insert
1834 * @zbr: zbranch to insert
1835 * @n: slot number to insert to
1836 *
1837 * This is a helper function for 'tnc_insert()'. UBIFS does not allow "gaps" in
1838 * znode's array of zbranches and keeps zbranches consolidated, so when a new
1839 * zbranch has to be inserted to the @znode->zbranches[]' array at the @n-th
1840 * slot, zbranches starting from @n have to be moved right.
1841 */
1842static void insert_zbranch(struct ubifs_znode *znode,
1843 const struct ubifs_zbranch *zbr, int n)
1844{
1845 int i;
1846
1847 ubifs_assert(ubifs_zn_dirty(znode));
1848
1849 if (znode->level) {
1850 for (i = znode->child_cnt; i > n; i--) {
1851 znode->zbranch[i] = znode->zbranch[i - 1];
1852 if (znode->zbranch[i].znode)
1853 znode->zbranch[i].znode->iip = i;
1854 }
1855 if (zbr->znode)
1856 zbr->znode->iip = n;
1857 } else
1858 for (i = znode->child_cnt; i > n; i--)
1859 znode->zbranch[i] = znode->zbranch[i - 1];
1860
1861 znode->zbranch[n] = *zbr;
1862 znode->child_cnt += 1;
1863
1864 /*
1865 * After inserting at slot zero, the lower bound of the key range of
1866 * this znode may have changed. If this znode is subsequently split
1867 * then the upper bound of the key range may change, and furthermore
1868 * it could change to be lower than the original lower bound. If that
1869 * happens, then it will no longer be possible to find this znode in the
1870 * TNC using the key from the index node on flash. That is bad because
1871 * if it is not found, we will assume it is obsolete and may overwrite
1872 * it. Then if there is an unclean unmount, we will start using the
1873 * old index which will be broken.
1874 *
1875 * So we first mark znodes that have insertions at slot zero, and then
1876 * if they are split we add their lnum/offs to the old_idx tree.
1877 */
1878 if (n == 0)
1879 znode->alt = 1;
1880}
1881
1882/**
1883 * tnc_insert - insert a node into TNC.
1884 * @c: UBIFS file-system description object
1885 * @znode: znode to insert into
1886 * @zbr: branch to insert
1887 * @n: slot number to insert new zbranch to
1888 *
1889 * This function inserts a new node described by @zbr into znode @znode. If
1890 * znode does not have a free slot for new zbranch, it is split. Parent znodes
1891 * are splat as well if needed. Returns zero in case of success or a negative
1892 * error code in case of failure.
1893 */
1894static int tnc_insert(struct ubifs_info *c, struct ubifs_znode *znode,
1895 struct ubifs_zbranch *zbr, int n)
1896{
1897 struct ubifs_znode *zn, *zi, *zp;
1898 int i, keep, move, appending = 0;
1899 union ubifs_key *key = &zbr->key, *key1;
1900
1901 ubifs_assert(n >= 0 && n <= c->fanout);
1902
1903 /* Implement naive insert for now */
1904again:
1905 zp = znode->parent;
1906 if (znode->child_cnt < c->fanout) {
1907 ubifs_assert(n != c->fanout);
1908 dbg_tnc("inserted at %d level %d, key %s", n, znode->level,
1909 DBGKEY(key));
1910
1911 insert_zbranch(znode, zbr, n);
1912
1913 /* Ensure parent's key is correct */
1914 if (n == 0 && zp && znode->iip == 0)
1915 correct_parent_keys(c, znode);
1916
1917 return 0;
1918 }
1919
1920 /*
1921 * Unfortunately, @znode does not have more empty slots and we have to
1922 * split it.
1923 */
1924 dbg_tnc("splitting level %d, key %s", znode->level, DBGKEY(key));
1925
1926 if (znode->alt)
1927 /*
1928 * We can no longer be sure of finding this znode by key, so we
1929 * record it in the old_idx tree.
1930 */
1931 ins_clr_old_idx_znode(c, znode);
1932
1933 zn = kzalloc(c->max_znode_sz, GFP_NOFS);
1934 if (!zn)
1935 return -ENOMEM;
1936 zn->parent = zp;
1937 zn->level = znode->level;
1938
1939 /* Decide where to split */
1940 if (znode->level == 0 && key_type(c, key) == UBIFS_DATA_KEY) {
1941 /* Try not to split consecutive data keys */
1942 if (n == c->fanout) {
1943 key1 = &znode->zbranch[n - 1].key;
1944 if (key_inum(c, key1) == key_inum(c, key) &&
1945 key_type(c, key1) == UBIFS_DATA_KEY)
1946 appending = 1;
1947 } else
1948 goto check_split;
1949 } else if (appending && n != c->fanout) {
1950 /* Try not to split consecutive data keys */
1951 appending = 0;
1952check_split:
1953 if (n >= (c->fanout + 1) / 2) {
1954 key1 = &znode->zbranch[0].key;
1955 if (key_inum(c, key1) == key_inum(c, key) &&
1956 key_type(c, key1) == UBIFS_DATA_KEY) {
1957 key1 = &znode->zbranch[n].key;
1958 if (key_inum(c, key1) != key_inum(c, key) ||
1959 key_type(c, key1) != UBIFS_DATA_KEY) {
1960 keep = n;
1961 move = c->fanout - keep;
1962 zi = znode;
1963 goto do_split;
1964 }
1965 }
1966 }
1967 }
1968
1969 if (appending) {
1970 keep = c->fanout;
1971 move = 0;
1972 } else {
1973 keep = (c->fanout + 1) / 2;
1974 move = c->fanout - keep;
1975 }
1976
1977 /*
1978 * Although we don't at present, we could look at the neighbors and see
1979 * if we can move some zbranches there.
1980 */
1981
1982 if (n < keep) {
1983 /* Insert into existing znode */
1984 zi = znode;
1985 move += 1;
1986 keep -= 1;
1987 } else {
1988 /* Insert into new znode */
1989 zi = zn;
1990 n -= keep;
1991 /* Re-parent */
1992 if (zn->level != 0)
1993 zbr->znode->parent = zn;
1994 }
1995
1996do_split:
1997
1998 __set_bit(DIRTY_ZNODE, &zn->flags);
1999 atomic_long_inc(&c->dirty_zn_cnt);
2000
2001 zn->child_cnt = move;
2002 znode->child_cnt = keep;
2003
2004 dbg_tnc("moving %d, keeping %d", move, keep);
2005
2006 /* Move zbranch */
2007 for (i = 0; i < move; i++) {
2008 zn->zbranch[i] = znode->zbranch[keep + i];
2009 /* Re-parent */
2010 if (zn->level != 0)
2011 if (zn->zbranch[i].znode) {
2012 zn->zbranch[i].znode->parent = zn;
2013 zn->zbranch[i].znode->iip = i;
2014 }
2015 }
2016
2017 /* Insert new key and branch */
2018 dbg_tnc("inserting at %d level %d, key %s", n, zn->level, DBGKEY(key));
2019
2020 insert_zbranch(zi, zbr, n);
2021
2022 /* Insert new znode (produced by spitting) into the parent */
2023 if (zp) {
2024 if (n == 0 && zi == znode && znode->iip == 0)
2025 correct_parent_keys(c, znode);
2026
2027 /* Locate insertion point */
2028 n = znode->iip + 1;
2029
2030 /* Tail recursion */
2031 zbr->key = zn->zbranch[0].key;
2032 zbr->znode = zn;
2033 zbr->lnum = 0;
2034 zbr->offs = 0;
2035 zbr->len = 0;
2036 znode = zp;
2037
2038 goto again;
2039 }
2040
2041 /* We have to split root znode */
2042 dbg_tnc("creating new zroot at level %d", znode->level + 1);
2043
2044 zi = kzalloc(c->max_znode_sz, GFP_NOFS);
2045 if (!zi)
2046 return -ENOMEM;
2047
2048 zi->child_cnt = 2;
2049 zi->level = znode->level + 1;
2050
2051 __set_bit(DIRTY_ZNODE, &zi->flags);
2052 atomic_long_inc(&c->dirty_zn_cnt);
2053
2054 zi->zbranch[0].key = znode->zbranch[0].key;
2055 zi->zbranch[0].znode = znode;
2056 zi->zbranch[0].lnum = c->zroot.lnum;
2057 zi->zbranch[0].offs = c->zroot.offs;
2058 zi->zbranch[0].len = c->zroot.len;
2059 zi->zbranch[1].key = zn->zbranch[0].key;
2060 zi->zbranch[1].znode = zn;
2061
2062 c->zroot.lnum = 0;
2063 c->zroot.offs = 0;
2064 c->zroot.len = 0;
2065 c->zroot.znode = zi;
2066
2067 zn->parent = zi;
2068 zn->iip = 1;
2069 znode->parent = zi;
2070 znode->iip = 0;
2071
2072 return 0;
2073}
2074
2075/**
2076 * ubifs_tnc_add - add a node to TNC.
2077 * @c: UBIFS file-system description object
2078 * @key: key to add
2079 * @lnum: LEB number of node
2080 * @offs: node offset
2081 * @len: node length
2082 *
2083 * This function adds a node with key @key to TNC. The node may be new or it may
2084 * obsolete some existing one. Returns %0 on success or negative error code on
2085 * failure.
2086 */
2087int ubifs_tnc_add(struct ubifs_info *c, const union ubifs_key *key, int lnum,
2088 int offs, int len)
2089{
2090 int found, n, err = 0;
2091 struct ubifs_znode *znode;
2092
2093 mutex_lock(&c->tnc_mutex);
2094 dbg_tnc("%d:%d, len %d, key %s", lnum, offs, len, DBGKEY(key));
2095 found = lookup_level0_dirty(c, key, &znode, &n);
2096 if (!found) {
2097 struct ubifs_zbranch zbr;
2098
2099 zbr.znode = NULL;
2100 zbr.lnum = lnum;
2101 zbr.offs = offs;
2102 zbr.len = len;
2103 key_copy(c, key, &zbr.key);
2104 err = tnc_insert(c, znode, &zbr, n + 1);
2105 } else if (found == 1) {
2106 struct ubifs_zbranch *zbr = &znode->zbranch[n];
2107
2108 lnc_free(zbr);
2109 err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2110 zbr->lnum = lnum;
2111 zbr->offs = offs;
2112 zbr->len = len;
2113 } else
2114 err = found;
2115 if (!err)
2116 err = dbg_check_tnc(c, 0);
2117 mutex_unlock(&c->tnc_mutex);
2118
2119 return err;
2120}
2121
2122/**
2123 * ubifs_tnc_replace - replace a node in the TNC only if the old node is found.
2124 * @c: UBIFS file-system description object
2125 * @key: key to add
2126 * @old_lnum: LEB number of old node
2127 * @old_offs: old node offset
2128 * @lnum: LEB number of node
2129 * @offs: node offset
2130 * @len: node length
2131 *
2132 * This function replaces a node with key @key in the TNC only if the old node
2133 * is found. This function is called by garbage collection when node are moved.
2134 * Returns %0 on success or negative error code on failure.
2135 */
2136int ubifs_tnc_replace(struct ubifs_info *c, const union ubifs_key *key,
2137 int old_lnum, int old_offs, int lnum, int offs, int len)
2138{
2139 int found, n, err = 0;
2140 struct ubifs_znode *znode;
2141
2142 mutex_lock(&c->tnc_mutex);
2143 dbg_tnc("old LEB %d:%d, new LEB %d:%d, len %d, key %s", old_lnum,
2144 old_offs, lnum, offs, len, DBGKEY(key));
2145 found = lookup_level0_dirty(c, key, &znode, &n);
2146 if (found < 0) {
2147 err = found;
2148 goto out_unlock;
2149 }
2150
2151 if (found == 1) {
2152 struct ubifs_zbranch *zbr = &znode->zbranch[n];
2153
2154 found = 0;
2155 if (zbr->lnum == old_lnum && zbr->offs == old_offs) {
2156 lnc_free(zbr);
2157 err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2158 if (err)
2159 goto out_unlock;
2160 zbr->lnum = lnum;
2161 zbr->offs = offs;
2162 zbr->len = len;
2163 found = 1;
2164 } else if (is_hash_key(c, key)) {
2165 found = resolve_collision_directly(c, key, &znode, &n,
2166 old_lnum, old_offs);
2167 dbg_tnc("rc returned %d, znode %p, n %d, LEB %d:%d",
2168 found, znode, n, old_lnum, old_offs);
2169 if (found < 0) {
2170 err = found;
2171 goto out_unlock;
2172 }
2173
2174 if (found) {
2175 /* Ensure the znode is dirtied */
2176 if (znode->cnext || !ubifs_zn_dirty(znode)) {
2177 znode = dirty_cow_bottom_up(c, znode);
2178 if (IS_ERR(znode)) {
2179 err = PTR_ERR(znode);
2180 goto out_unlock;
2181 }
2182 }
2183 zbr = &znode->zbranch[n];
2184 lnc_free(zbr);
2185 err = ubifs_add_dirt(c, zbr->lnum,
2186 zbr->len);
2187 if (err)
2188 goto out_unlock;
2189 zbr->lnum = lnum;
2190 zbr->offs = offs;
2191 zbr->len = len;
2192 }
2193 }
2194 }
2195
2196 if (!found)
2197 err = ubifs_add_dirt(c, lnum, len);
2198
2199 if (!err)
2200 err = dbg_check_tnc(c, 0);
2201
2202out_unlock:
2203 mutex_unlock(&c->tnc_mutex);
2204 return err;
2205}
2206
2207/**
2208 * ubifs_tnc_add_nm - add a "hashed" node to TNC.
2209 * @c: UBIFS file-system description object
2210 * @key: key to add
2211 * @lnum: LEB number of node
2212 * @offs: node offset
2213 * @len: node length
2214 * @nm: node name
2215 *
2216 * This is the same as 'ubifs_tnc_add()' but it should be used with keys which
2217 * may have collisions, like directory entry keys.
2218 */
2219int ubifs_tnc_add_nm(struct ubifs_info *c, const union ubifs_key *key,
2220 int lnum, int offs, int len, const struct qstr *nm)
2221{
2222 int found, n, err = 0;
2223 struct ubifs_znode *znode;
2224
2225 mutex_lock(&c->tnc_mutex);
2226 dbg_tnc("LEB %d:%d, name '%.*s', key %s", lnum, offs, nm->len, nm->name,
2227 DBGKEY(key));
2228 found = lookup_level0_dirty(c, key, &znode, &n);
2229 if (found < 0) {
2230 err = found;
2231 goto out_unlock;
2232 }
2233
2234 if (found == 1) {
2235 if (c->replaying)
2236 found = fallible_resolve_collision(c, key, &znode, &n,
2237 nm, 1);
2238 else
2239 found = resolve_collision(c, key, &znode, &n, nm);
2240 dbg_tnc("rc returned %d, znode %p, n %d", found, znode, n);
2241 if (found < 0) {
2242 err = found;
2243 goto out_unlock;
2244 }
2245
2246 /* Ensure the znode is dirtied */
2247 if (znode->cnext || !ubifs_zn_dirty(znode)) {
2248 znode = dirty_cow_bottom_up(c, znode);
2249 if (IS_ERR(znode)) {
2250 err = PTR_ERR(znode);
2251 goto out_unlock;
2252 }
2253 }
2254
2255 if (found == 1) {
2256 struct ubifs_zbranch *zbr = &znode->zbranch[n];
2257
2258 lnc_free(zbr);
2259 err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2260 zbr->lnum = lnum;
2261 zbr->offs = offs;
2262 zbr->len = len;
2263 goto out_unlock;
2264 }
2265 }
2266
2267 if (!found) {
2268 struct ubifs_zbranch zbr;
2269
2270 zbr.znode = NULL;
2271 zbr.lnum = lnum;
2272 zbr.offs = offs;
2273 zbr.len = len;
2274 key_copy(c, key, &zbr.key);
2275 err = tnc_insert(c, znode, &zbr, n + 1);
2276 if (err)
2277 goto out_unlock;
2278 if (c->replaying) {
2279 /*
2280 * We did not find it in the index so there may be a
2281 * dangling branch still in the index. So we remove it
2282 * by passing 'ubifs_tnc_remove_nm()' the same key but
2283 * an unmatchable name.
2284 */
2285 struct qstr noname = { .len = 0, .name = "" };
2286
2287 err = dbg_check_tnc(c, 0);
2288 mutex_unlock(&c->tnc_mutex);
2289 if (err)
2290 return err;
2291 return ubifs_tnc_remove_nm(c, key, &noname);
2292 }
2293 }
2294
2295out_unlock:
2296 if (!err)
2297 err = dbg_check_tnc(c, 0);
2298 mutex_unlock(&c->tnc_mutex);
2299 return err;
2300}
2301
2302/**
2303 * tnc_delete - delete a znode form TNC.
2304 * @c: UBIFS file-system description object
2305 * @znode: znode to delete from
2306 * @n: zbranch slot number to delete
2307 *
2308 * This function deletes a leaf node from @n-th slot of @znode. Returns zero in
2309 * case of success and a negative error code in case of failure.
2310 */
2311static int tnc_delete(struct ubifs_info *c, struct ubifs_znode *znode, int n)
2312{
2313 struct ubifs_zbranch *zbr;
2314 struct ubifs_znode *zp;
2315 int i, err;
2316
2317 /* Delete without merge for now */
2318 ubifs_assert(znode->level == 0);
2319 ubifs_assert(n >= 0 && n < c->fanout);
2320 dbg_tnc("deleting %s", DBGKEY(&znode->zbranch[n].key));
2321
2322 zbr = &znode->zbranch[n];
2323 lnc_free(zbr);
2324
2325 err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
2326 if (err) {
2327 dbg_dump_znode(c, znode);
2328 return err;
2329 }
2330
2331 /* We do not "gap" zbranch slots */
2332 for (i = n; i < znode->child_cnt - 1; i++)
2333 znode->zbranch[i] = znode->zbranch[i + 1];
2334 znode->child_cnt -= 1;
2335
2336 if (znode->child_cnt > 0)
2337 return 0;
2338
2339 /*
2340 * This was the last zbranch, we have to delete this znode from the
2341 * parent.
2342 */
2343
2344 do {
2345 ubifs_assert(!test_bit(OBSOLETE_ZNODE, &znode->flags));
2346 ubifs_assert(ubifs_zn_dirty(znode));
2347
2348 zp = znode->parent;
2349 n = znode->iip;
2350
2351 atomic_long_dec(&c->dirty_zn_cnt);
2352
2353 err = insert_old_idx_znode(c, znode);
2354 if (err)
2355 return err;
2356
2357 if (znode->cnext) {
2358 __set_bit(OBSOLETE_ZNODE, &znode->flags);
2359 atomic_long_inc(&c->clean_zn_cnt);
2360 atomic_long_inc(&ubifs_clean_zn_cnt);
2361 } else
2362 kfree(znode);
2363 znode = zp;
2364 } while (znode->child_cnt == 1); /* while removing last child */
2365
2366 /* Remove from znode, entry n - 1 */
2367 znode->child_cnt -= 1;
2368 ubifs_assert(znode->level != 0);
2369 for (i = n; i < znode->child_cnt; i++) {
2370 znode->zbranch[i] = znode->zbranch[i + 1];
2371 if (znode->zbranch[i].znode)
2372 znode->zbranch[i].znode->iip = i;
2373 }
2374
2375 /*
2376 * If this is the root and it has only 1 child then
2377 * collapse the tree.
2378 */
2379 if (!znode->parent) {
2380 while (znode->child_cnt == 1 && znode->level != 0) {
2381 zp = znode;
2382 zbr = &znode->zbranch[0];
2383 znode = get_znode(c, znode, 0);
2384 if (IS_ERR(znode))
2385 return PTR_ERR(znode);
2386 znode = dirty_cow_znode(c, zbr);
2387 if (IS_ERR(znode))
2388 return PTR_ERR(znode);
2389 znode->parent = NULL;
2390 znode->iip = 0;
2391 if (c->zroot.len) {
2392 err = insert_old_idx(c, c->zroot.lnum,
2393 c->zroot.offs);
2394 if (err)
2395 return err;
2396 }
2397 c->zroot.lnum = zbr->lnum;
2398 c->zroot.offs = zbr->offs;
2399 c->zroot.len = zbr->len;
2400 c->zroot.znode = znode;
2401 ubifs_assert(!test_bit(OBSOLETE_ZNODE,
2402 &zp->flags));
2403 ubifs_assert(test_bit(DIRTY_ZNODE, &zp->flags));
2404 atomic_long_dec(&c->dirty_zn_cnt);
2405
2406 if (zp->cnext) {
2407 __set_bit(OBSOLETE_ZNODE, &zp->flags);
2408 atomic_long_inc(&c->clean_zn_cnt);
2409 atomic_long_inc(&ubifs_clean_zn_cnt);
2410 } else
2411 kfree(zp);
2412 }
2413 }
2414
2415 return 0;
2416}
2417
2418/**
2419 * ubifs_tnc_remove - remove an index entry of a node.
2420 * @c: UBIFS file-system description object
2421 * @key: key of node
2422 *
2423 * Returns %0 on success or negative error code on failure.
2424 */
2425int ubifs_tnc_remove(struct ubifs_info *c, const union ubifs_key *key)
2426{
2427 int found, n, err = 0;
2428 struct ubifs_znode *znode;
2429
2430 mutex_lock(&c->tnc_mutex);
2431 dbg_tnc("key %s", DBGKEY(key));
2432 found = lookup_level0_dirty(c, key, &znode, &n);
2433 if (found < 0) {
2434 err = found;
2435 goto out_unlock;
2436 }
2437 if (found == 1)
2438 err = tnc_delete(c, znode, n);
2439 if (!err)
2440 err = dbg_check_tnc(c, 0);
2441
2442out_unlock:
2443 mutex_unlock(&c->tnc_mutex);
2444 return err;
2445}
2446
2447/**
2448 * ubifs_tnc_remove_nm - remove an index entry for a "hashed" node.
2449 * @c: UBIFS file-system description object
2450 * @key: key of node
2451 * @nm: directory entry name
2452 *
2453 * Returns %0 on success or negative error code on failure.
2454 */
2455int ubifs_tnc_remove_nm(struct ubifs_info *c, const union ubifs_key *key,
2456 const struct qstr *nm)
2457{
2458 int n, err;
2459 struct ubifs_znode *znode;
2460
2461 mutex_lock(&c->tnc_mutex);
2462 dbg_tnc("%.*s, key %s", nm->len, nm->name, DBGKEY(key));
2463 err = lookup_level0_dirty(c, key, &znode, &n);
2464 if (err < 0)
2465 goto out_unlock;
2466
2467 if (err) {
2468 if (c->replaying)
2469 err = fallible_resolve_collision(c, key, &znode, &n,
2470 nm, 0);
2471 else
2472 err = resolve_collision(c, key, &znode, &n, nm);
2473 dbg_tnc("rc returned %d, znode %p, n %d", err, znode, n);
2474 if (err < 0)
2475 goto out_unlock;
2476 if (err) {
2477 /* Ensure the znode is dirtied */
2478 if (znode->cnext || !ubifs_zn_dirty(znode)) {
2479 znode = dirty_cow_bottom_up(c, znode);
2480 if (IS_ERR(znode)) {
2481 err = PTR_ERR(znode);
2482 goto out_unlock;
2483 }
2484 }
2485 err = tnc_delete(c, znode, n);
2486 }
2487 }
2488
2489out_unlock:
2490 if (!err)
2491 err = dbg_check_tnc(c, 0);
2492 mutex_unlock(&c->tnc_mutex);
2493 return err;
2494}
2495
2496/**
2497 * key_in_range - determine if a key falls within a range of keys.
2498 * @c: UBIFS file-system description object
2499 * @key: key to check
2500 * @from_key: lowest key in range
2501 * @to_key: highest key in range
2502 *
2503 * This function returns %1 if the key is in range and %0 otherwise.
2504 */
2505static int key_in_range(struct ubifs_info *c, union ubifs_key *key,
2506 union ubifs_key *from_key, union ubifs_key *to_key)
2507{
2508 if (keys_cmp(c, key, from_key) < 0)
2509 return 0;
2510 if (keys_cmp(c, key, to_key) > 0)
2511 return 0;
2512 return 1;
2513}
2514
2515/**
2516 * ubifs_tnc_remove_range - remove index entries in range.
2517 * @c: UBIFS file-system description object
2518 * @from_key: lowest key to remove
2519 * @to_key: highest key to remove
2520 *
2521 * This function removes index entries starting at @from_key and ending at
2522 * @to_key. This function returns zero in case of success and a negative error
2523 * code in case of failure.
2524 */
2525int ubifs_tnc_remove_range(struct ubifs_info *c, union ubifs_key *from_key,
2526 union ubifs_key *to_key)
2527{
2528 int i, n, k, err = 0;
2529 struct ubifs_znode *znode;
2530 union ubifs_key *key;
2531
2532 mutex_lock(&c->tnc_mutex);
2533 while (1) {
2534 /* Find first level 0 znode that contains keys to remove */
2535 err = ubifs_lookup_level0(c, from_key, &znode, &n);
2536 if (err < 0)
2537 goto out_unlock;
2538
2539 if (err)
2540 key = from_key;
2541 else {
2542 err = tnc_next(c, &znode, &n);
2543 if (err == -ENOENT) {
2544 err = 0;
2545 goto out_unlock;
2546 }
2547 if (err < 0)
2548 goto out_unlock;
2549 key = &znode->zbranch[n].key;
2550 if (!key_in_range(c, key, from_key, to_key)) {
2551 err = 0;
2552 goto out_unlock;
2553 }
2554 }
2555
2556 /* Ensure the znode is dirtied */
2557 if (znode->cnext || !ubifs_zn_dirty(znode)) {
2558 znode = dirty_cow_bottom_up(c, znode);
2559 if (IS_ERR(znode)) {
2560 err = PTR_ERR(znode);
2561 goto out_unlock;
2562 }
2563 }
2564
2565 /* Remove all keys in range except the first */
2566 for (i = n + 1, k = 0; i < znode->child_cnt; i++, k++) {
2567 key = &znode->zbranch[i].key;
2568 if (!key_in_range(c, key, from_key, to_key))
2569 break;
2570 lnc_free(&znode->zbranch[i]);
2571 err = ubifs_add_dirt(c, znode->zbranch[i].lnum,
2572 znode->zbranch[i].len);
2573 if (err) {
2574 dbg_dump_znode(c, znode);
2575 goto out_unlock;
2576 }
2577 dbg_tnc("removing %s", DBGKEY(key));
2578 }
2579 if (k) {
2580 for (i = n + 1 + k; i < znode->child_cnt; i++)
2581 znode->zbranch[i - k] = znode->zbranch[i];
2582 znode->child_cnt -= k;
2583 }
2584
2585 /* Now delete the first */
2586 err = tnc_delete(c, znode, n);
2587 if (err)
2588 goto out_unlock;
2589 }
2590
2591out_unlock:
2592 if (!err)
2593 err = dbg_check_tnc(c, 0);
2594 mutex_unlock(&c->tnc_mutex);
2595 return err;
2596}
2597
2598/**
2599 * ubifs_tnc_remove_ino - remove an inode from TNC.
2600 * @c: UBIFS file-system description object
2601 * @inum: inode number to remove
2602 *
2603 * This function remove inode @inum and all the extended attributes associated
2604 * with the anode from TNC and returns zero in case of success or a negative
2605 * error code in case of failure.
2606 */
2607int ubifs_tnc_remove_ino(struct ubifs_info *c, ino_t inum)
2608{
2609 union ubifs_key key1, key2;
2610 struct ubifs_dent_node *xent, *pxent = NULL;
2611 struct qstr nm = { .name = NULL };
2612
2613 dbg_tnc("ino %lu", (unsigned long)inum);
2614
2615 /*
2616 * Walk all extended attribute entries and remove them together with
2617 * corresponding extended attribute inodes.
2618 */
2619 lowest_xent_key(c, &key1, inum);
2620 while (1) {
2621 ino_t xattr_inum;
2622 int err;
2623
2624 xent = ubifs_tnc_next_ent(c, &key1, &nm);
2625 if (IS_ERR(xent)) {
2626 err = PTR_ERR(xent);
2627 if (err == -ENOENT)
2628 break;
2629 return err;
2630 }
2631
2632 xattr_inum = le64_to_cpu(xent->inum);
2633 dbg_tnc("xent '%s', ino %lu", xent->name,
2634 (unsigned long)xattr_inum);
2635
2636 nm.name = (char *)xent->name;
2637 nm.len = le16_to_cpu(xent->nlen);
2638 err = ubifs_tnc_remove_nm(c, &key1, &nm);
2639 if (err) {
2640 kfree(xent);
2641 return err;
2642 }
2643
2644 lowest_ino_key(c, &key1, xattr_inum);
2645 highest_ino_key(c, &key2, xattr_inum);
2646 err = ubifs_tnc_remove_range(c, &key1, &key2);
2647 if (err) {
2648 kfree(xent);
2649 return err;
2650 }
2651
2652 kfree(pxent);
2653 pxent = xent;
2654 key_read(c, &xent->key, &key1);
2655 }
2656
2657 kfree(pxent);
2658 lowest_ino_key(c, &key1, inum);
2659 highest_ino_key(c, &key2, inum);
2660
2661 return ubifs_tnc_remove_range(c, &key1, &key2);
2662}
2663
2664/**
2665 * ubifs_tnc_next_ent - walk directory or extended attribute entries.
2666 * @c: UBIFS file-system description object
2667 * @key: key of last entry
2668 * @nm: name of last entry found or %NULL
2669 *
2670 * This function finds and reads the next directory or extended attribute entry
2671 * after the given key (@key) if there is one. @nm is used to resolve
2672 * collisions.
2673 *
2674 * If the name of the current entry is not known and only the key is known,
2675 * @nm->name has to be %NULL. In this case the semantics of this function is a
2676 * little bit different and it returns the entry corresponding to this key, not
2677 * the next one. If the key was not found, the closest "right" entry is
2678 * returned.
2679 *
2680 * If the fist entry has to be found, @key has to contain the lowest possible
2681 * key value for this inode and @name has to be %NULL.
2682 *
2683 * This function returns the found directory or extended attribute entry node
2684 * in case of success, %-ENOENT is returned if no entry was found, and a
2685 * negative error code is returned in case of failure.
2686 */
2687struct ubifs_dent_node *ubifs_tnc_next_ent(struct ubifs_info *c,
2688 union ubifs_key *key,
2689 const struct qstr *nm)
2690{
2691 int n, err, type = key_type(c, key);
2692 struct ubifs_znode *znode;
2693 struct ubifs_dent_node *dent;
2694 struct ubifs_zbranch *zbr;
2695 union ubifs_key *dkey;
2696
2697 dbg_tnc("%s %s", nm->name ? (char *)nm->name : "(lowest)", DBGKEY(key));
2698 ubifs_assert(is_hash_key(c, key));
2699
2700 mutex_lock(&c->tnc_mutex);
2701 err = ubifs_lookup_level0(c, key, &znode, &n);
2702 if (unlikely(err < 0))
2703 goto out_unlock;
2704
2705 if (nm->name) {
2706 if (err) {
2707 /* Handle collisions */
2708 err = resolve_collision(c, key, &znode, &n, nm);
2709 dbg_tnc("rc returned %d, znode %p, n %d",
2710 err, znode, n);
2711 if (unlikely(err < 0))
2712 goto out_unlock;
2713 }
2714
2715 /* Now find next entry */
2716 err = tnc_next(c, &znode, &n);
2717 if (unlikely(err))
2718 goto out_unlock;
2719 } else {
2720 /*
2721 * The full name of the entry was not given, in which case the
2722 * behavior of this function is a little different and it
2723 * returns current entry, not the next one.
2724 */
2725 if (!err) {
2726 /*
2727 * However, the given key does not exist in the TNC
2728 * tree and @znode/@n variables contain the closest
2729 * "preceding" element. Switch to the next one.
2730 */
2731 err = tnc_next(c, &znode, &n);
2732 if (err)
2733 goto out_unlock;
2734 }
2735 }
2736
2737 zbr = &znode->zbranch[n];
2738 dent = kmalloc(zbr->len, GFP_NOFS);
2739 if (unlikely(!dent)) {
2740 err = -ENOMEM;
2741 goto out_unlock;
2742 }
2743
2744 /*
2745 * The above 'tnc_next()' call could lead us to the next inode, check
2746 * this.
2747 */
2748 dkey = &zbr->key;
2749 if (key_inum(c, dkey) != key_inum(c, key) ||
2750 key_type(c, dkey) != type) {
2751 err = -ENOENT;
2752 goto out_free;
2753 }
2754
2755 err = tnc_read_node_nm(c, zbr, dent);
2756 if (unlikely(err))
2757 goto out_free;
2758
2759 mutex_unlock(&c->tnc_mutex);
2760 return dent;
2761
2762out_free:
2763 kfree(dent);
2764out_unlock:
2765 mutex_unlock(&c->tnc_mutex);
2766 return ERR_PTR(err);
2767}