2 * Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
6 * Written by Anatoly P. Pinchuk pap@namesys.botik.ru
7 * Programm System Institute
8 * Pereslavl-Zalessky Russia
12 * This file contains functions dealing with S+tree
27 * pathrelse_and_restore
31 * search_for_position_by_key
33 * prepare_for_direct_item
34 * prepare_for_direntry_item
35 * prepare_for_delete_or_cut
36 * calc_deleted_bytes_number
39 * reiserfs_delete_item
40 * reiserfs_delete_solid_item
41 * reiserfs_delete_object
42 * maybe_indirect_to_direct
43 * indirect_to_direct_roll_back
44 * reiserfs_cut_from_item
46 * reiserfs_do_truncate
47 * reiserfs_paste_into_item
48 * reiserfs_insert_item
51 #include <linux/time.h>
52 #include <linux/string.h>
53 #include <linux/pagemap.h>
54 #include <linux/reiserfs_fs.h>
55 #include <linux/buffer_head.h>
56 #include <linux/quotaops.h>
58 /* Does the buffer contain a disk block which is in the tree. */
59 inline int B_IS_IN_TREE(const struct buffer_head *p_s_bh)
62 RFALSE(B_LEVEL(p_s_bh) > MAX_HEIGHT,
63 "PAP-1010: block (%b) has too big level (%z)", p_s_bh, p_s_bh);
65 return (B_LEVEL(p_s_bh) != FREE_LEVEL);
69 // to gets item head in le form
71 inline void copy_item_head(struct item_head *p_v_to,
72 const struct item_head *p_v_from)
74 memcpy(p_v_to, p_v_from, IH_SIZE);
77 /* k1 is pointer to on-disk structure which is stored in little-endian
78 form. k2 is pointer to cpu variable. For key of items of the same
79 object this returns 0.
80 Returns: -1 if key1 < key2
83 inline int comp_short_keys(const struct reiserfs_key *le_key,
84 const struct cpu_key *cpu_key)
87 n = le32_to_cpu(le_key->k_dir_id);
88 if (n < cpu_key->on_disk_key.k_dir_id)
90 if (n > cpu_key->on_disk_key.k_dir_id)
92 n = le32_to_cpu(le_key->k_objectid);
93 if (n < cpu_key->on_disk_key.k_objectid)
95 if (n > cpu_key->on_disk_key.k_objectid)
100 /* k1 is pointer to on-disk structure which is stored in little-endian
101 form. k2 is pointer to cpu variable.
102 Compare keys using all 4 key fields.
103 Returns: -1 if key1 < key2 0
104 if key1 = key2 1 if key1 > key2 */
105 static inline int comp_keys(const struct reiserfs_key *le_key,
106 const struct cpu_key *cpu_key)
110 retval = comp_short_keys(le_key, cpu_key);
113 if (le_key_k_offset(le_key_version(le_key), le_key) <
114 cpu_key_k_offset(cpu_key))
116 if (le_key_k_offset(le_key_version(le_key), le_key) >
117 cpu_key_k_offset(cpu_key))
120 if (cpu_key->key_length == 3)
123 /* this part is needed only when tail conversion is in progress */
124 if (le_key_k_type(le_key_version(le_key), le_key) <
125 cpu_key_k_type(cpu_key))
128 if (le_key_k_type(le_key_version(le_key), le_key) >
129 cpu_key_k_type(cpu_key))
135 inline int comp_short_le_keys(const struct reiserfs_key *key1,
136 const struct reiserfs_key *key2)
138 __u32 *p_s_1_u32, *p_s_2_u32;
139 int n_key_length = REISERFS_SHORT_KEY_LEN;
141 p_s_1_u32 = (__u32 *) key1;
142 p_s_2_u32 = (__u32 *) key2;
143 for (; n_key_length--; ++p_s_1_u32, ++p_s_2_u32) {
144 if (le32_to_cpu(*p_s_1_u32) < le32_to_cpu(*p_s_2_u32))
146 if (le32_to_cpu(*p_s_1_u32) > le32_to_cpu(*p_s_2_u32))
152 inline void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from)
155 to->on_disk_key.k_dir_id = le32_to_cpu(from->k_dir_id);
156 to->on_disk_key.k_objectid = le32_to_cpu(from->k_objectid);
158 // find out version of the key
159 version = le_key_version(from);
160 to->version = version;
161 to->on_disk_key.k_offset = le_key_k_offset(version, from);
162 to->on_disk_key.k_type = le_key_k_type(version, from);
165 // this does not say which one is bigger, it only returns 1 if keys
166 // are not equal, 0 otherwise
167 inline int comp_le_keys(const struct reiserfs_key *k1,
168 const struct reiserfs_key *k2)
170 return memcmp(k1, k2, sizeof(struct reiserfs_key));
173 /**************************************************************************
174 * Binary search toolkit function *
175 * Search for an item in the array by the item key *
176 * Returns: 1 if found, 0 if not found; *
177 * *p_n_pos = number of the searched element if found, else the *
178 * number of the first element that is larger than p_v_key. *
179 **************************************************************************/
180 /* For those not familiar with binary search: n_lbound is the leftmost item that it
181 could be, n_rbound the rightmost item that it could be. We examine the item
182 halfway between n_lbound and n_rbound, and that tells us either that we can increase
183 n_lbound, or decrease n_rbound, or that we have found it, or if n_lbound <= n_rbound that
184 there are no possible items, and we have not found it. With each examination we
185 cut the number of possible items it could be by one more than half rounded down,
187 static inline int bin_search(const void *p_v_key, /* Key to search for. */
188 const void *p_v_base, /* First item in the array. */
189 int p_n_num, /* Number of items in the array. */
190 int p_n_width, /* Item size in the array.
191 searched. Lest the reader be
192 confused, note that this is crafted
193 as a general function, and when it
194 is applied specifically to the array
195 of item headers in a node, p_n_width
196 is actually the item header size not
198 int *p_n_pos /* Number of the searched for element. */
201 int n_rbound, n_lbound, n_j;
203 for (n_j = ((n_rbound = p_n_num - 1) + (n_lbound = 0)) / 2;
204 n_lbound <= n_rbound; n_j = (n_rbound + n_lbound) / 2)
206 ((struct reiserfs_key *)((char *)p_v_base +
208 (struct cpu_key *)p_v_key)) {
217 return ITEM_FOUND; /* Key found in the array. */
220 /* bin_search did not find given key, it returns position of key,
221 that is minimal and greater than the given one. */
223 return ITEM_NOT_FOUND;
226 #ifdef CONFIG_REISERFS_CHECK
227 extern struct tree_balance *cur_tb;
230 /* Minimal possible key. It is never in the tree. */
231 const struct reiserfs_key MIN_KEY = { 0, 0, {{0, 0},} };
233 /* Maximal possible key. It is never in the tree. */
234 static const struct reiserfs_key MAX_KEY = {
235 __constant_cpu_to_le32(0xffffffff),
236 __constant_cpu_to_le32(0xffffffff),
237 {{__constant_cpu_to_le32(0xffffffff),
238 __constant_cpu_to_le32(0xffffffff)},}
241 /* Get delimiting key of the buffer by looking for it in the buffers in the path, starting from the bottom
242 of the path, and going upwards. We must check the path's validity at each step. If the key is not in
243 the path, there is no delimiting key in the tree (buffer is first or last buffer in tree), and in this
244 case we return a special key, either MIN_KEY or MAX_KEY. */
245 static inline const struct reiserfs_key *get_lkey(const struct treepath
247 const struct super_block
250 int n_position, n_path_offset = p_s_chk_path->path_length;
251 struct buffer_head *p_s_parent;
253 RFALSE(n_path_offset < FIRST_PATH_ELEMENT_OFFSET,
254 "PAP-5010: invalid offset in the path");
256 /* While not higher in path than first element. */
257 while (n_path_offset-- > FIRST_PATH_ELEMENT_OFFSET) {
259 RFALSE(!buffer_uptodate
260 (PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)),
261 "PAP-5020: parent is not uptodate");
263 /* Parent at the path is not in the tree now. */
266 PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)))
268 /* Check whether position in the parent is correct. */
270 PATH_OFFSET_POSITION(p_s_chk_path,
272 B_NR_ITEMS(p_s_parent))
274 /* Check whether parent at the path really points to the child. */
275 if (B_N_CHILD_NUM(p_s_parent, n_position) !=
276 PATH_OFFSET_PBUFFER(p_s_chk_path,
277 n_path_offset + 1)->b_blocknr)
279 /* Return delimiting key if position in the parent is not equal to zero. */
281 return B_N_PDELIM_KEY(p_s_parent, n_position - 1);
283 /* Return MIN_KEY if we are in the root of the buffer tree. */
284 if (PATH_OFFSET_PBUFFER(p_s_chk_path, FIRST_PATH_ELEMENT_OFFSET)->
285 b_blocknr == SB_ROOT_BLOCK(p_s_sb))
290 /* Get delimiting key of the buffer at the path and its right neighbor. */
291 inline const struct reiserfs_key *get_rkey(const struct treepath *p_s_chk_path,
292 const struct super_block *p_s_sb)
294 int n_position, n_path_offset = p_s_chk_path->path_length;
295 struct buffer_head *p_s_parent;
297 RFALSE(n_path_offset < FIRST_PATH_ELEMENT_OFFSET,
298 "PAP-5030: invalid offset in the path");
300 while (n_path_offset-- > FIRST_PATH_ELEMENT_OFFSET) {
302 RFALSE(!buffer_uptodate
303 (PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)),
304 "PAP-5040: parent is not uptodate");
306 /* Parent at the path is not in the tree now. */
309 PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)))
311 /* Check whether position in the parent is correct. */
313 PATH_OFFSET_POSITION(p_s_chk_path,
315 B_NR_ITEMS(p_s_parent))
317 /* Check whether parent at the path really points to the child. */
318 if (B_N_CHILD_NUM(p_s_parent, n_position) !=
319 PATH_OFFSET_PBUFFER(p_s_chk_path,
320 n_path_offset + 1)->b_blocknr)
322 /* Return delimiting key if position in the parent is not the last one. */
323 if (n_position != B_NR_ITEMS(p_s_parent))
324 return B_N_PDELIM_KEY(p_s_parent, n_position);
326 /* Return MAX_KEY if we are in the root of the buffer tree. */
327 if (PATH_OFFSET_PBUFFER(p_s_chk_path, FIRST_PATH_ELEMENT_OFFSET)->
328 b_blocknr == SB_ROOT_BLOCK(p_s_sb))
333 /* Check whether a key is contained in the tree rooted from a buffer at a path. */
334 /* This works by looking at the left and right delimiting keys for the buffer in the last path_element in
335 the path. These delimiting keys are stored at least one level above that buffer in the tree. If the
336 buffer is the first or last node in the tree order then one of the delimiting keys may be absent, and in
337 this case get_lkey and get_rkey return a special key which is MIN_KEY or MAX_KEY. */
338 static inline int key_in_buffer(struct treepath *p_s_chk_path, /* Path which should be checked. */
339 const struct cpu_key *p_s_key, /* Key which should be checked. */
340 struct super_block *p_s_sb /* Super block pointer. */
344 RFALSE(!p_s_key || p_s_chk_path->path_length < FIRST_PATH_ELEMENT_OFFSET
345 || p_s_chk_path->path_length > MAX_HEIGHT,
346 "PAP-5050: pointer to the key(%p) is NULL or invalid path length(%d)",
347 p_s_key, p_s_chk_path->path_length);
348 RFALSE(!PATH_PLAST_BUFFER(p_s_chk_path)->b_bdev,
349 "PAP-5060: device must not be NODEV");
351 if (comp_keys(get_lkey(p_s_chk_path, p_s_sb), p_s_key) == 1)
352 /* left delimiting key is bigger, that the key we look for */
354 // if ( comp_keys(p_s_key, get_rkey(p_s_chk_path, p_s_sb)) != -1 )
355 if (comp_keys(get_rkey(p_s_chk_path, p_s_sb), p_s_key) != 1)
356 /* p_s_key must be less than right delimitiing key */
361 int reiserfs_check_path(struct treepath *p)
363 RFALSE(p->path_length != ILLEGAL_PATH_ELEMENT_OFFSET,
364 "path not properly relsed");
368 /* Drop the reference to each buffer in a path and restore
369 * dirty bits clean when preparing the buffer for the log.
370 * This version should only be called from fix_nodes() */
371 void pathrelse_and_restore(struct super_block *sb,
372 struct treepath *p_s_search_path)
374 int n_path_offset = p_s_search_path->path_length;
376 RFALSE(n_path_offset < ILLEGAL_PATH_ELEMENT_OFFSET,
377 "clm-4000: invalid path offset");
379 while (n_path_offset > ILLEGAL_PATH_ELEMENT_OFFSET) {
380 struct buffer_head *bh;
381 bh = PATH_OFFSET_PBUFFER(p_s_search_path, n_path_offset--);
382 reiserfs_restore_prepared_buffer(sb, bh);
385 p_s_search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET;
388 /* Drop the reference to each buffer in a path */
389 void pathrelse(struct treepath *p_s_search_path)
391 int n_path_offset = p_s_search_path->path_length;
393 RFALSE(n_path_offset < ILLEGAL_PATH_ELEMENT_OFFSET,
394 "PAP-5090: invalid path offset");
396 while (n_path_offset > ILLEGAL_PATH_ELEMENT_OFFSET)
397 brelse(PATH_OFFSET_PBUFFER(p_s_search_path, n_path_offset--));
399 p_s_search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET;
402 static int is_leaf(char *buf, int blocksize, struct buffer_head *bh)
404 struct block_head *blkh;
405 struct item_head *ih;
411 blkh = (struct block_head *)buf;
412 if (blkh_level(blkh) != DISK_LEAF_NODE_LEVEL) {
413 reiserfs_warning(NULL, "reiserfs-5080",
414 "this should be caught earlier");
418 nr = blkh_nr_item(blkh);
419 if (nr < 1 || nr > ((blocksize - BLKH_SIZE) / (IH_SIZE + MIN_ITEM_LEN))) {
420 /* item number is too big or too small */
421 reiserfs_warning(NULL, "reiserfs-5081",
422 "nr_item seems wrong: %z", bh);
425 ih = (struct item_head *)(buf + BLKH_SIZE) + nr - 1;
426 used_space = BLKH_SIZE + IH_SIZE * nr + (blocksize - ih_location(ih));
427 if (used_space != blocksize - blkh_free_space(blkh)) {
428 /* free space does not match to calculated amount of use space */
429 reiserfs_warning(NULL, "reiserfs-5082",
430 "free space seems wrong: %z", bh);
433 // FIXME: it is_leaf will hit performance too much - we may have
436 /* check tables of item heads */
437 ih = (struct item_head *)(buf + BLKH_SIZE);
438 prev_location = blocksize;
439 for (i = 0; i < nr; i++, ih++) {
440 if (le_ih_k_type(ih) == TYPE_ANY) {
441 reiserfs_warning(NULL, "reiserfs-5083",
442 "wrong item type for item %h",
446 if (ih_location(ih) >= blocksize
447 || ih_location(ih) < IH_SIZE * nr) {
448 reiserfs_warning(NULL, "reiserfs-5084",
449 "item location seems wrong: %h",
453 if (ih_item_len(ih) < 1
454 || ih_item_len(ih) > MAX_ITEM_LEN(blocksize)) {
455 reiserfs_warning(NULL, "reiserfs-5085",
456 "item length seems wrong: %h",
460 if (prev_location - ih_location(ih) != ih_item_len(ih)) {
461 reiserfs_warning(NULL, "reiserfs-5086",
462 "item location seems wrong "
463 "(second one): %h", ih);
466 prev_location = ih_location(ih);
469 // one may imagine much more checks
473 /* returns 1 if buf looks like an internal node, 0 otherwise */
474 static int is_internal(char *buf, int blocksize, struct buffer_head *bh)
476 struct block_head *blkh;
480 blkh = (struct block_head *)buf;
481 nr = blkh_level(blkh);
482 if (nr <= DISK_LEAF_NODE_LEVEL || nr > MAX_HEIGHT) {
483 /* this level is not possible for internal nodes */
484 reiserfs_warning(NULL, "reiserfs-5087",
485 "this should be caught earlier");
489 nr = blkh_nr_item(blkh);
490 if (nr > (blocksize - BLKH_SIZE - DC_SIZE) / (KEY_SIZE + DC_SIZE)) {
491 /* for internal which is not root we might check min number of keys */
492 reiserfs_warning(NULL, "reiserfs-5088",
493 "number of key seems wrong: %z", bh);
497 used_space = BLKH_SIZE + KEY_SIZE * nr + DC_SIZE * (nr + 1);
498 if (used_space != blocksize - blkh_free_space(blkh)) {
499 reiserfs_warning(NULL, "reiserfs-5089",
500 "free space seems wrong: %z", bh);
503 // one may imagine much more checks
507 // make sure that bh contains formatted node of reiserfs tree of
509 static int is_tree_node(struct buffer_head *bh, int level)
511 if (B_LEVEL(bh) != level) {
512 reiserfs_warning(NULL, "reiserfs-5090", "node level %d does "
513 "not match to the expected one %d",
517 if (level == DISK_LEAF_NODE_LEVEL)
518 return is_leaf(bh->b_data, bh->b_size, bh);
520 return is_internal(bh->b_data, bh->b_size, bh);
523 #define SEARCH_BY_KEY_READA 16
525 /* The function is NOT SCHEDULE-SAFE! */
526 static void search_by_key_reada(struct super_block *s,
527 struct buffer_head **bh,
528 b_blocknr_t *b, int num)
532 for (i = 0; i < num; i++) {
533 bh[i] = sb_getblk(s, b[i]);
535 for (j = 0; j < i; j++) {
537 * note, this needs attention if we are getting rid of the BKL
538 * you have to make sure the prepared bit isn't set on this buffer
540 if (!buffer_uptodate(bh[j]))
541 ll_rw_block(READA, 1, bh + j);
546 /**************************************************************************
547 * Algorithm SearchByKey *
548 * look for item in the Disk S+Tree by its key *
549 * Input: p_s_sb - super block *
550 * p_s_key - pointer to the key to search *
551 * Output: ITEM_FOUND, ITEM_NOT_FOUND or IO_ERROR *
552 * p_s_search_path - path from the root to the needed leaf *
553 **************************************************************************/
555 /* This function fills up the path from the root to the leaf as it
556 descends the tree looking for the key. It uses reiserfs_bread to
557 try to find buffers in the cache given their block number. If it
558 does not find them in the cache it reads them from disk. For each
559 node search_by_key finds using reiserfs_bread it then uses
560 bin_search to look through that node. bin_search will find the
561 position of the block_number of the next node if it is looking
562 through an internal node. If it is looking through a leaf node
563 bin_search will find the position of the item which has key either
564 equal to given key, or which is the maximal key less than the given
565 key. search_by_key returns a path that must be checked for the
566 correctness of the top of the path but need not be checked for the
567 correctness of the bottom of the path */
568 /* The function is NOT SCHEDULE-SAFE! */
569 int search_by_key(struct super_block *p_s_sb, const struct cpu_key *p_s_key, /* Key to search. */
570 struct treepath *p_s_search_path,/* This structure was
571 allocated and initialized
573 function. It is filled up
575 int n_stop_level /* How far down the tree to search. To
576 stop at leaf level - set to
577 DISK_LEAF_NODE_LEVEL */
580 b_blocknr_t n_block_number;
582 struct buffer_head *p_s_bh;
583 struct path_element *p_s_last_element;
584 int n_node_level, n_retval;
585 int right_neighbor_of_leaf_node;
587 struct buffer_head *reada_bh[SEARCH_BY_KEY_READA];
588 b_blocknr_t reada_blocks[SEARCH_BY_KEY_READA];
591 #ifdef CONFIG_REISERFS_CHECK
592 int n_repeat_counter = 0;
595 PROC_INFO_INC(p_s_sb, search_by_key);
597 /* As we add each node to a path we increase its count. This means that
598 we must be careful to release all nodes in a path before we either
599 discard the path struct or re-use the path struct, as we do here. */
601 pathrelse(p_s_search_path);
603 right_neighbor_of_leaf_node = 0;
605 /* With each iteration of this loop we search through the items in the
606 current node, and calculate the next current node(next path element)
607 for the next iteration of this loop.. */
608 n_block_number = SB_ROOT_BLOCK(p_s_sb);
612 #ifdef CONFIG_REISERFS_CHECK
613 if (!(++n_repeat_counter % 50000))
614 reiserfs_warning(p_s_sb, "PAP-5100",
615 "%s: there were %d iterations of "
616 "while loop looking for key %K",
617 current->comm, n_repeat_counter,
621 /* prep path to have another element added to it. */
623 PATH_OFFSET_PELEMENT(p_s_search_path,
624 ++p_s_search_path->path_length);
625 fs_gen = get_generation(p_s_sb);
627 /* Read the next tree node, and set the last element in the path to
628 have a pointer to it. */
629 if ((p_s_bh = p_s_last_element->pe_buffer =
630 sb_getblk(p_s_sb, n_block_number))) {
631 if (!buffer_uptodate(p_s_bh) && reada_count > 1) {
632 search_by_key_reada(p_s_sb, reada_bh,
633 reada_blocks, reada_count);
635 ll_rw_block(READ, 1, &p_s_bh);
636 wait_on_buffer(p_s_bh);
637 if (!buffer_uptodate(p_s_bh))
641 p_s_search_path->path_length--;
642 pathrelse(p_s_search_path);
646 if (expected_level == -1)
647 expected_level = SB_TREE_HEIGHT(p_s_sb);
650 /* It is possible that schedule occurred. We must check whether the key
651 to search is still in the tree rooted from the current buffer. If
652 not then repeat search from the root. */
653 if (fs_changed(fs_gen, p_s_sb) &&
654 (!B_IS_IN_TREE(p_s_bh) ||
655 B_LEVEL(p_s_bh) != expected_level ||
656 !key_in_buffer(p_s_search_path, p_s_key, p_s_sb))) {
657 PROC_INFO_INC(p_s_sb, search_by_key_fs_changed);
658 PROC_INFO_INC(p_s_sb, search_by_key_restarted);
659 PROC_INFO_INC(p_s_sb,
660 sbk_restarted[expected_level - 1]);
661 pathrelse(p_s_search_path);
663 /* Get the root block number so that we can repeat the search
664 starting from the root. */
665 n_block_number = SB_ROOT_BLOCK(p_s_sb);
667 right_neighbor_of_leaf_node = 0;
669 /* repeat search from the root */
673 /* only check that the key is in the buffer if p_s_key is not
674 equal to the MAX_KEY. Latter case is only possible in
675 "finish_unfinished()" processing during mount. */
676 RFALSE(comp_keys(&MAX_KEY, p_s_key) &&
677 !key_in_buffer(p_s_search_path, p_s_key, p_s_sb),
678 "PAP-5130: key is not in the buffer");
679 #ifdef CONFIG_REISERFS_CHECK
681 print_cur_tb("5140");
682 reiserfs_panic(p_s_sb, "PAP-5140",
683 "schedule occurred in do_balance!");
687 // make sure, that the node contents look like a node of
689 if (!is_tree_node(p_s_bh, expected_level)) {
690 reiserfs_error(p_s_sb, "vs-5150",
691 "invalid format found in block %ld. "
692 "Fsck?", p_s_bh->b_blocknr);
693 pathrelse(p_s_search_path);
697 /* ok, we have acquired next formatted node in the tree */
698 n_node_level = B_LEVEL(p_s_bh);
700 PROC_INFO_BH_STAT(p_s_sb, p_s_bh, n_node_level - 1);
702 RFALSE(n_node_level < n_stop_level,
703 "vs-5152: tree level (%d) is less than stop level (%d)",
704 n_node_level, n_stop_level);
706 n_retval = bin_search(p_s_key, B_N_PITEM_HEAD(p_s_bh, 0),
709 DISK_LEAF_NODE_LEVEL) ? IH_SIZE :
711 &(p_s_last_element->pe_position));
712 if (n_node_level == n_stop_level) {
716 /* we are not in the stop level */
717 if (n_retval == ITEM_FOUND)
718 /* item has been found, so we choose the pointer which is to the right of the found one */
719 p_s_last_element->pe_position++;
721 /* if item was not found we choose the position which is to
722 the left of the found item. This requires no code,
723 bin_search did it already. */
725 /* So we have chosen a position in the current node which is
726 an internal node. Now we calculate child block number by
727 position in the node. */
729 B_N_CHILD_NUM(p_s_bh, p_s_last_element->pe_position);
731 /* if we are going to read leaf nodes, try for read ahead as well */
732 if ((p_s_search_path->reada & PATH_READA) &&
733 n_node_level == DISK_LEAF_NODE_LEVEL + 1) {
734 int pos = p_s_last_element->pe_position;
735 int limit = B_NR_ITEMS(p_s_bh);
736 struct reiserfs_key *le_key;
738 if (p_s_search_path->reada & PATH_READA_BACK)
740 while (reada_count < SEARCH_BY_KEY_READA) {
743 reada_blocks[reada_count++] =
744 B_N_CHILD_NUM(p_s_bh, pos);
745 if (p_s_search_path->reada & PATH_READA_BACK)
751 * check to make sure we're in the same object
753 le_key = B_N_PDELIM_KEY(p_s_bh, pos);
754 if (le32_to_cpu(le_key->k_objectid) !=
755 p_s_key->on_disk_key.k_objectid) {
763 /* Form the path to an item and position in this item which contains
764 file byte defined by p_s_key. If there is no such item
765 corresponding to the key, we point the path to the item with
766 maximal key less than p_s_key, and *p_n_pos_in_item is set to one
767 past the last entry/byte in the item. If searching for entry in a
768 directory item, and it is not found, *p_n_pos_in_item is set to one
769 entry more than the entry with maximal key which is less than the
772 Note that if there is no entry in this same node which is one more,
773 then we point to an imaginary entry. for direct items, the
774 position is in units of bytes, for indirect items the position is
775 in units of blocknr entries, for directory items the position is in
776 units of directory entries. */
778 /* The function is NOT SCHEDULE-SAFE! */
779 int search_for_position_by_key(struct super_block *p_s_sb, /* Pointer to the super block. */
780 const struct cpu_key *p_cpu_key, /* Key to search (cpu variable) */
781 struct treepath *p_s_search_path /* Filled up by this function. */
784 struct item_head *p_le_ih; /* pointer to on-disk structure */
786 loff_t item_offset, offset;
787 struct reiserfs_dir_entry de;
790 /* If searching for directory entry. */
791 if (is_direntry_cpu_key(p_cpu_key))
792 return search_by_entry_key(p_s_sb, p_cpu_key, p_s_search_path,
795 /* If not searching for directory entry. */
797 /* If item is found. */
798 retval = search_item(p_s_sb, p_cpu_key, p_s_search_path);
799 if (retval == IO_ERROR)
801 if (retval == ITEM_FOUND) {
805 (PATH_PLAST_BUFFER(p_s_search_path),
806 PATH_LAST_POSITION(p_s_search_path))),
807 "PAP-5165: item length equals zero");
809 pos_in_item(p_s_search_path) = 0;
810 return POSITION_FOUND;
813 RFALSE(!PATH_LAST_POSITION(p_s_search_path),
814 "PAP-5170: position equals zero");
816 /* Item is not found. Set path to the previous item. */
818 B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_search_path),
819 --PATH_LAST_POSITION(p_s_search_path));
820 n_blk_size = p_s_sb->s_blocksize;
822 if (comp_short_keys(&(p_le_ih->ih_key), p_cpu_key)) {
823 return FILE_NOT_FOUND;
825 // FIXME: quite ugly this far
827 item_offset = le_ih_k_offset(p_le_ih);
828 offset = cpu_key_k_offset(p_cpu_key);
830 /* Needed byte is contained in the item pointed to by the path. */
831 if (item_offset <= offset &&
832 item_offset + op_bytes_number(p_le_ih, n_blk_size) > offset) {
833 pos_in_item(p_s_search_path) = offset - item_offset;
834 if (is_indirect_le_ih(p_le_ih)) {
835 pos_in_item(p_s_search_path) /= n_blk_size;
837 return POSITION_FOUND;
840 /* Needed byte is not contained in the item pointed to by the
841 path. Set pos_in_item out of the item. */
842 if (is_indirect_le_ih(p_le_ih))
843 pos_in_item(p_s_search_path) =
844 ih_item_len(p_le_ih) / UNFM_P_SIZE;
846 pos_in_item(p_s_search_path) = ih_item_len(p_le_ih);
848 return POSITION_NOT_FOUND;
851 /* Compare given item and item pointed to by the path. */
852 int comp_items(const struct item_head *stored_ih, const struct treepath *p_s_path)
854 struct buffer_head *p_s_bh;
855 struct item_head *ih;
857 /* Last buffer at the path is not in the tree. */
858 if (!B_IS_IN_TREE(p_s_bh = PATH_PLAST_BUFFER(p_s_path)))
861 /* Last path position is invalid. */
862 if (PATH_LAST_POSITION(p_s_path) >= B_NR_ITEMS(p_s_bh))
865 /* we need only to know, whether it is the same item */
866 ih = get_ih(p_s_path);
867 return memcmp(stored_ih, ih, IH_SIZE);
870 /* unformatted nodes are not logged anymore, ever. This is safe
873 #define held_by_others(bh) (atomic_read(&(bh)->b_count) > 1)
875 // block can not be forgotten as it is in I/O or held by someone
876 #define block_in_use(bh) (buffer_locked(bh) || (held_by_others(bh)))
878 // prepare for delete or cut of direct item
879 static inline int prepare_for_direct_item(struct treepath *path,
880 struct item_head *le_ih,
882 loff_t new_file_length, int *cut_size)
886 if (new_file_length == max_reiserfs_offset(inode)) {
887 /* item has to be deleted */
888 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
891 // new file gets truncated
892 if (get_inode_item_key_version(inode) == KEY_FORMAT_3_6) {
894 round_len = ROUND_UP(new_file_length);
895 /* this was n_new_file_length < le_ih ... */
896 if (round_len < le_ih_k_offset(le_ih)) {
897 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
898 return M_DELETE; /* Delete this item. */
900 /* Calculate first position and size for cutting from item. */
901 pos_in_item(path) = round_len - (le_ih_k_offset(le_ih) - 1);
902 *cut_size = -(ih_item_len(le_ih) - pos_in_item(path));
904 return M_CUT; /* Cut from this item. */
907 // old file: items may have any length
909 if (new_file_length < le_ih_k_offset(le_ih)) {
910 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
911 return M_DELETE; /* Delete this item. */
913 /* Calculate first position and size for cutting from item. */
914 *cut_size = -(ih_item_len(le_ih) -
916 new_file_length + 1 - le_ih_k_offset(le_ih)));
917 return M_CUT; /* Cut from this item. */
920 static inline int prepare_for_direntry_item(struct treepath *path,
921 struct item_head *le_ih,
923 loff_t new_file_length,
926 if (le_ih_k_offset(le_ih) == DOT_OFFSET &&
927 new_file_length == max_reiserfs_offset(inode)) {
928 RFALSE(ih_entry_count(le_ih) != 2,
929 "PAP-5220: incorrect empty directory item (%h)", le_ih);
930 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
931 return M_DELETE; /* Delete the directory item containing "." and ".." entry. */
934 if (ih_entry_count(le_ih) == 1) {
935 /* Delete the directory item such as there is one record only
937 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
941 /* Cut one record from the directory item. */
944 entry_length(get_last_bh(path), le_ih, pos_in_item(path)));
948 #define JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD (2 * JOURNAL_PER_BALANCE_CNT + 1)
950 /* If the path points to a directory or direct item, calculate mode and the size cut, for balance.
951 If the path points to an indirect item, remove some number of its unformatted nodes.
952 In case of file truncate calculate whether this item must be deleted/truncated or last
953 unformatted node of this item will be converted to a direct item.
954 This function returns a determination of what balance mode the calling function should employ. */
955 static char prepare_for_delete_or_cut(struct reiserfs_transaction_handle *th, struct inode *inode, struct treepath *p_s_path, const struct cpu_key *p_s_item_key, int *p_n_removed, /* Number of unformatted nodes which were removed
956 from end of the file. */
957 int *p_n_cut_size, unsigned long long n_new_file_length /* MAX_KEY_OFFSET in case of delete. */
960 struct super_block *p_s_sb = inode->i_sb;
961 struct item_head *p_le_ih = PATH_PITEM_HEAD(p_s_path);
962 struct buffer_head *p_s_bh = PATH_PLAST_BUFFER(p_s_path);
964 BUG_ON(!th->t_trans_id);
966 /* Stat_data item. */
967 if (is_statdata_le_ih(p_le_ih)) {
969 RFALSE(n_new_file_length != max_reiserfs_offset(inode),
970 "PAP-5210: mode must be M_DELETE");
972 *p_n_cut_size = -(IH_SIZE + ih_item_len(p_le_ih));
976 /* Directory item. */
977 if (is_direntry_le_ih(p_le_ih))
978 return prepare_for_direntry_item(p_s_path, p_le_ih, inode,
983 if (is_direct_le_ih(p_le_ih))
984 return prepare_for_direct_item(p_s_path, p_le_ih, inode,
985 n_new_file_length, p_n_cut_size);
987 /* Case of an indirect item. */
989 int blk_size = p_s_sb->s_blocksize;
990 struct item_head s_ih;
996 if ( n_new_file_length == max_reiserfs_offset (inode) ) {
997 /* prepare_for_delete_or_cut() is called by
998 * reiserfs_delete_item() */
999 n_new_file_length = 0;
1006 p_s_bh = PATH_PLAST_BUFFER(p_s_path);
1007 copy_item_head(&s_ih, PATH_PITEM_HEAD(p_s_path));
1008 pos = I_UNFM_NUM(&s_ih);
1010 while (le_ih_k_offset (&s_ih) + (pos - 1) * blk_size > n_new_file_length) {
1014 /* Each unformatted block deletion may involve one additional
1015 * bitmap block into the transaction, thereby the initial
1016 * journal space reservation might not be enough. */
1017 if (!delete && (*p_n_cut_size) != 0 &&
1018 reiserfs_transaction_free_space(th) < JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD) {
1022 unfm = (__le32 *)B_I_PITEM(p_s_bh, &s_ih) + pos - 1;
1023 block = get_block_num(unfm, 0);
1026 reiserfs_prepare_for_journal(p_s_sb, p_s_bh, 1);
1027 put_block_num(unfm, 0, 0);
1028 journal_mark_dirty (th, p_s_sb, p_s_bh);
1029 reiserfs_free_block(th, inode, block, 1);
1034 if (item_moved (&s_ih, p_s_path)) {
1041 (*p_n_cut_size) -= UNFM_P_SIZE;
1044 (*p_n_cut_size) -= IH_SIZE;
1049 /* a trick. If the buffer has been logged, this will do nothing. If
1050 ** we've broken the loop without logging it, it will restore the
1052 reiserfs_restore_prepared_buffer(p_s_sb, p_s_bh);
1053 } while (need_re_search &&
1054 search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path) == POSITION_FOUND);
1055 pos_in_item(p_s_path) = pos * UNFM_P_SIZE;
1057 if (*p_n_cut_size == 0) {
1058 /* Nothing were cut. maybe convert last unformatted node to the
1066 /* Calculate number of bytes which will be deleted or cut during balance */
1067 static int calc_deleted_bytes_number(struct tree_balance *p_s_tb, char c_mode)
1070 struct item_head *p_le_ih = PATH_PITEM_HEAD(p_s_tb->tb_path);
1072 if (is_statdata_le_ih(p_le_ih))
1077 M_DELETE) ? ih_item_len(p_le_ih) : -p_s_tb->insert_size[0];
1078 if (is_direntry_le_ih(p_le_ih)) {
1079 // return EMPTY_DIR_SIZE; /* We delete emty directoris only. */
1080 // we can't use EMPTY_DIR_SIZE, as old format dirs have a different
1081 // empty size. ick. FIXME, is this right?
1086 if (is_indirect_le_ih(p_le_ih))
1087 n_del_size = (n_del_size / UNFM_P_SIZE) * (PATH_PLAST_BUFFER(p_s_tb->tb_path)->b_size); // - get_ih_free_space (p_le_ih);
1091 static void init_tb_struct(struct reiserfs_transaction_handle *th,
1092 struct tree_balance *p_s_tb,
1093 struct super_block *p_s_sb,
1094 struct treepath *p_s_path, int n_size)
1097 BUG_ON(!th->t_trans_id);
1099 memset(p_s_tb, '\0', sizeof(struct tree_balance));
1100 p_s_tb->transaction_handle = th;
1101 p_s_tb->tb_sb = p_s_sb;
1102 p_s_tb->tb_path = p_s_path;
1103 PATH_OFFSET_PBUFFER(p_s_path, ILLEGAL_PATH_ELEMENT_OFFSET) = NULL;
1104 PATH_OFFSET_POSITION(p_s_path, ILLEGAL_PATH_ELEMENT_OFFSET) = 0;
1105 p_s_tb->insert_size[0] = n_size;
1108 void padd_item(char *item, int total_length, int length)
1112 for (i = total_length; i > length;)
1116 #ifdef REISERQUOTA_DEBUG
1117 char key2type(struct reiserfs_key *ih)
1119 if (is_direntry_le_key(2, ih))
1121 if (is_direct_le_key(2, ih))
1123 if (is_indirect_le_key(2, ih))
1125 if (is_statdata_le_key(2, ih))
1130 char head2type(struct item_head *ih)
1132 if (is_direntry_le_ih(ih))
1134 if (is_direct_le_ih(ih))
1136 if (is_indirect_le_ih(ih))
1138 if (is_statdata_le_ih(ih))
1144 /* Delete object item. */
1145 int reiserfs_delete_item(struct reiserfs_transaction_handle *th, struct treepath *p_s_path, /* Path to the deleted item. */
1146 const struct cpu_key *p_s_item_key, /* Key to search for the deleted item. */
1147 struct inode *p_s_inode, /* inode is here just to update i_blocks and quotas */
1148 struct buffer_head *p_s_un_bh)
1149 { /* NULL or unformatted node pointer. */
1150 struct super_block *p_s_sb = p_s_inode->i_sb;
1151 struct tree_balance s_del_balance;
1152 struct item_head s_ih;
1153 struct item_head *q_ih;
1154 int quota_cut_bytes;
1155 int n_ret_value, n_del_size, n_removed;
1157 #ifdef CONFIG_REISERFS_CHECK
1162 BUG_ON(!th->t_trans_id);
1164 init_tb_struct(th, &s_del_balance, p_s_sb, p_s_path,
1165 0 /*size is unknown */ );
1170 #ifdef CONFIG_REISERFS_CHECK
1174 prepare_for_delete_or_cut(th, p_s_inode, p_s_path,
1175 p_s_item_key, &n_removed,
1177 max_reiserfs_offset(p_s_inode));
1179 RFALSE(c_mode != M_DELETE, "PAP-5320: mode must be M_DELETE");
1181 copy_item_head(&s_ih, PATH_PITEM_HEAD(p_s_path));
1182 s_del_balance.insert_size[0] = n_del_size;
1184 n_ret_value = fix_nodes(M_DELETE, &s_del_balance, NULL, NULL);
1185 if (n_ret_value != REPEAT_SEARCH)
1188 PROC_INFO_INC(p_s_sb, delete_item_restarted);
1190 // file system changed, repeat search
1192 search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path);
1193 if (n_ret_value == IO_ERROR)
1195 if (n_ret_value == FILE_NOT_FOUND) {
1196 reiserfs_warning(p_s_sb, "vs-5340",
1197 "no items of the file %K found",
1203 if (n_ret_value != CARRY_ON) {
1204 unfix_nodes(&s_del_balance);
1207 // reiserfs_delete_item returns item length when success
1208 n_ret_value = calc_deleted_bytes_number(&s_del_balance, M_DELETE);
1209 q_ih = get_ih(p_s_path);
1210 quota_cut_bytes = ih_item_len(q_ih);
1212 /* hack so the quota code doesn't have to guess if the file
1213 ** has a tail. On tail insert, we allocate quota for 1 unformatted node.
1214 ** We test the offset because the tail might have been
1215 ** split into multiple items, and we only want to decrement for
1216 ** the unfm node once
1218 if (!S_ISLNK(p_s_inode->i_mode) && is_direct_le_ih(q_ih)) {
1219 if ((le_ih_k_offset(q_ih) & (p_s_sb->s_blocksize - 1)) == 1) {
1220 quota_cut_bytes = p_s_sb->s_blocksize + UNFM_P_SIZE;
1222 quota_cut_bytes = 0;
1230 /* We are in direct2indirect conversion, so move tail contents
1231 to the unformatted node */
1232 /* note, we do the copy before preparing the buffer because we
1233 ** don't care about the contents of the unformatted node yet.
1234 ** the only thing we really care about is the direct item's data
1235 ** is in the unformatted node.
1237 ** Otherwise, we would have to call reiserfs_prepare_for_journal on
1238 ** the unformatted node, which might schedule, meaning we'd have to
1239 ** loop all the way back up to the start of the while loop.
1241 ** The unformatted node must be dirtied later on. We can't be
1242 ** sure here if the entire tail has been deleted yet.
1244 ** p_s_un_bh is from the page cache (all unformatted nodes are
1245 ** from the page cache) and might be a highmem page. So, we
1246 ** can't use p_s_un_bh->b_data.
1250 data = kmap_atomic(p_s_un_bh->b_page, KM_USER0);
1251 off = ((le_ih_k_offset(&s_ih) - 1) & (PAGE_CACHE_SIZE - 1));
1253 B_I_PITEM(PATH_PLAST_BUFFER(p_s_path), &s_ih),
1255 kunmap_atomic(data, KM_USER0);
1257 /* Perform balancing after all resources have been collected at once. */
1258 do_balance(&s_del_balance, NULL, NULL, M_DELETE);
1260 #ifdef REISERQUOTA_DEBUG
1261 reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
1262 "reiserquota delete_item(): freeing %u, id=%u type=%c",
1263 quota_cut_bytes, p_s_inode->i_uid, head2type(&s_ih));
1265 DQUOT_FREE_SPACE_NODIRTY(p_s_inode, quota_cut_bytes);
1267 /* Return deleted body length */
1271 /* Summary Of Mechanisms For Handling Collisions Between Processes:
1273 deletion of the body of the object is performed by iput(), with the
1274 result that if multiple processes are operating on a file, the
1275 deletion of the body of the file is deferred until the last process
1276 that has an open inode performs its iput().
1278 writes and truncates are protected from collisions by use of
1281 creates, linking, and mknod are protected from collisions with other
1282 processes by making the reiserfs_add_entry() the last step in the
1283 creation, and then rolling back all changes if there was a collision.
1287 /* this deletes item which never gets split */
1288 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1289 struct inode *inode, struct reiserfs_key *key)
1291 struct tree_balance tb;
1292 INITIALIZE_PATH(path);
1295 struct cpu_key cpu_key;
1297 int quota_cut_bytes = 0;
1299 BUG_ON(!th->t_trans_id);
1301 le_key2cpu_key(&cpu_key, key);
1304 retval = search_item(th->t_super, &cpu_key, &path);
1305 if (retval == IO_ERROR) {
1306 reiserfs_error(th->t_super, "vs-5350",
1307 "i/o failure occurred trying "
1308 "to delete %K", &cpu_key);
1311 if (retval != ITEM_FOUND) {
1313 // No need for a warning, if there is just no free space to insert '..' item into the newly-created subdir
1315 ((unsigned long long)
1316 GET_HASH_VALUE(le_key_k_offset
1317 (le_key_version(key), key)) == 0
1318 && (unsigned long long)
1319 GET_GENERATION_NUMBER(le_key_k_offset
1320 (le_key_version(key),
1322 reiserfs_warning(th->t_super, "vs-5355",
1323 "%k not found", key);
1328 item_len = ih_item_len(PATH_PITEM_HEAD(&path));
1329 init_tb_struct(th, &tb, th->t_super, &path,
1330 -(IH_SIZE + item_len));
1332 quota_cut_bytes = ih_item_len(PATH_PITEM_HEAD(&path));
1334 retval = fix_nodes(M_DELETE, &tb, NULL, NULL);
1335 if (retval == REPEAT_SEARCH) {
1336 PROC_INFO_INC(th->t_super, delete_solid_item_restarted);
1340 if (retval == CARRY_ON) {
1341 do_balance(&tb, NULL, NULL, M_DELETE);
1342 if (inode) { /* Should we count quota for item? (we don't count quotas for save-links) */
1343 #ifdef REISERQUOTA_DEBUG
1344 reiserfs_debug(th->t_super, REISERFS_DEBUG_CODE,
1345 "reiserquota delete_solid_item(): freeing %u id=%u type=%c",
1346 quota_cut_bytes, inode->i_uid,
1349 DQUOT_FREE_SPACE_NODIRTY(inode,
1354 // IO_ERROR, NO_DISK_SPACE, etc
1355 reiserfs_warning(th->t_super, "vs-5360",
1356 "could not delete %K due to fix_nodes failure",
1362 reiserfs_check_path(&path);
1365 int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
1366 struct inode *inode)
1370 BUG_ON(!th->t_trans_id);
1372 /* for directory this deletes item containing "." and ".." */
1374 reiserfs_do_truncate(th, inode, NULL, 0 /*no timestamp updates */ );
1378 #if defined( USE_INODE_GENERATION_COUNTER )
1379 if (!old_format_only(th->t_super)) {
1380 __le32 *inode_generation;
1383 &REISERFS_SB(th->t_super)->s_rs->s_inode_generation;
1384 le32_add_cpu(inode_generation, 1);
1386 /* USE_INODE_GENERATION_COUNTER */
1388 reiserfs_delete_solid_item(th, inode, INODE_PKEY(inode));
1393 static void unmap_buffers(struct page *page, loff_t pos)
1395 struct buffer_head *bh;
1396 struct buffer_head *head;
1397 struct buffer_head *next;
1398 unsigned long tail_index;
1399 unsigned long cur_index;
1402 if (page_has_buffers(page)) {
1403 tail_index = pos & (PAGE_CACHE_SIZE - 1);
1405 head = page_buffers(page);
1408 next = bh->b_this_page;
1410 /* we want to unmap the buffers that contain the tail, and
1411 ** all the buffers after it (since the tail must be at the
1412 ** end of the file). We don't want to unmap file data
1413 ** before the tail, since it might be dirty and waiting to
1416 cur_index += bh->b_size;
1417 if (cur_index > tail_index) {
1418 reiserfs_unmap_buffer(bh);
1421 } while (bh != head);
1426 static int maybe_indirect_to_direct(struct reiserfs_transaction_handle *th,
1427 struct inode *p_s_inode,
1429 struct treepath *p_s_path,
1430 const struct cpu_key *p_s_item_key,
1431 loff_t n_new_file_size, char *p_c_mode)
1433 struct super_block *p_s_sb = p_s_inode->i_sb;
1434 int n_block_size = p_s_sb->s_blocksize;
1436 BUG_ON(!th->t_trans_id);
1437 BUG_ON(n_new_file_size != p_s_inode->i_size);
1439 /* the page being sent in could be NULL if there was an i/o error
1440 ** reading in the last block. The user will hit problems trying to
1441 ** read the file, but for now we just skip the indirect2direct
1443 if (atomic_read(&p_s_inode->i_count) > 1 ||
1444 !tail_has_to_be_packed(p_s_inode) ||
1445 !page || (REISERFS_I(p_s_inode)->i_flags & i_nopack_mask)) {
1446 // leave tail in an unformatted node
1447 *p_c_mode = M_SKIP_BALANCING;
1449 n_block_size - (n_new_file_size & (n_block_size - 1));
1450 pathrelse(p_s_path);
1453 /* Permorm the conversion to a direct_item. */
1454 /*return indirect_to_direct (p_s_inode, p_s_path, p_s_item_key, n_new_file_size, p_c_mode); */
1455 return indirect2direct(th, p_s_inode, page, p_s_path, p_s_item_key,
1456 n_new_file_size, p_c_mode);
1459 /* we did indirect_to_direct conversion. And we have inserted direct
1460 item successesfully, but there were no disk space to cut unfm
1461 pointer being converted. Therefore we have to delete inserted
1463 static void indirect_to_direct_roll_back(struct reiserfs_transaction_handle *th,
1464 struct inode *inode, struct treepath *path)
1466 struct cpu_key tail_key;
1469 BUG_ON(!th->t_trans_id);
1471 make_cpu_key(&tail_key, inode, inode->i_size + 1, TYPE_DIRECT, 4); // !!!!
1472 tail_key.key_length = 4;
1475 (cpu_key_k_offset(&tail_key) & (inode->i_sb->s_blocksize - 1)) - 1;
1477 /* look for the last byte of the tail */
1478 if (search_for_position_by_key(inode->i_sb, &tail_key, path) ==
1480 reiserfs_panic(inode->i_sb, "vs-5615",
1481 "found invalid item");
1482 RFALSE(path->pos_in_item !=
1483 ih_item_len(PATH_PITEM_HEAD(path)) - 1,
1484 "vs-5616: appended bytes found");
1485 PATH_LAST_POSITION(path)--;
1488 reiserfs_delete_item(th, path, &tail_key, inode,
1489 NULL /*unbh not needed */ );
1491 || removed > tail_len,
1492 "vs-5617: there was tail %d bytes, removed item length %d bytes",
1494 tail_len -= removed;
1495 set_cpu_key_k_offset(&tail_key,
1496 cpu_key_k_offset(&tail_key) - removed);
1498 reiserfs_warning(inode->i_sb, "reiserfs-5091", "indirect_to_direct "
1499 "conversion has been rolled back due to "
1500 "lack of disk space");
1501 //mark_file_without_tail (inode);
1502 mark_inode_dirty(inode);
1505 /* (Truncate or cut entry) or delete object item. Returns < 0 on failure */
1506 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
1507 struct treepath *p_s_path,
1508 struct cpu_key *p_s_item_key,
1509 struct inode *p_s_inode,
1510 struct page *page, loff_t n_new_file_size)
1512 struct super_block *p_s_sb = p_s_inode->i_sb;
1513 /* Every function which is going to call do_balance must first
1514 create a tree_balance structure. Then it must fill up this
1515 structure by using the init_tb_struct and fix_nodes functions.
1516 After that we can make tree balancing. */
1517 struct tree_balance s_cut_balance;
1518 struct item_head *p_le_ih;
1519 int n_cut_size = 0, /* Amount to be cut. */
1520 n_ret_value = CARRY_ON, n_removed = 0, /* Number of the removed unformatted nodes. */
1521 n_is_inode_locked = 0;
1522 char c_mode; /* Mode of the balance. */
1524 int quota_cut_bytes;
1525 loff_t tail_pos = 0;
1527 BUG_ON(!th->t_trans_id);
1529 init_tb_struct(th, &s_cut_balance, p_s_inode->i_sb, p_s_path,
1532 /* Repeat this loop until we either cut the item without needing
1533 to balance, or we fix_nodes without schedule occurring */
1535 /* Determine the balance mode, position of the first byte to
1536 be cut, and size to be cut. In case of the indirect item
1537 free unformatted nodes which are pointed to by the cut
1541 prepare_for_delete_or_cut(th, p_s_inode, p_s_path,
1542 p_s_item_key, &n_removed,
1543 &n_cut_size, n_new_file_size);
1544 if (c_mode == M_CONVERT) {
1545 /* convert last unformatted node to direct item or leave
1546 tail in the unformatted node */
1547 RFALSE(n_ret_value != CARRY_ON,
1548 "PAP-5570: can not convert twice");
1551 maybe_indirect_to_direct(th, p_s_inode, page,
1552 p_s_path, p_s_item_key,
1553 n_new_file_size, &c_mode);
1554 if (c_mode == M_SKIP_BALANCING)
1555 /* tail has been left in the unformatted node */
1558 n_is_inode_locked = 1;
1560 /* removing of last unformatted node will change value we
1561 have to return to truncate. Save it */
1562 retval2 = n_ret_value;
1563 /*retval2 = p_s_sb->s_blocksize - (n_new_file_size & (p_s_sb->s_blocksize - 1)); */
1565 /* So, we have performed the first part of the conversion:
1566 inserting the new direct item. Now we are removing the
1567 last unformatted node pointer. Set key to search for
1569 set_cpu_key_k_type(p_s_item_key, TYPE_INDIRECT);
1570 p_s_item_key->key_length = 4;
1572 (n_new_file_size & (p_s_sb->s_blocksize - 1));
1573 tail_pos = n_new_file_size;
1574 set_cpu_key_k_offset(p_s_item_key, n_new_file_size + 1);
1575 if (search_for_position_by_key
1576 (p_s_sb, p_s_item_key,
1577 p_s_path) == POSITION_NOT_FOUND) {
1578 print_block(PATH_PLAST_BUFFER(p_s_path), 3,
1579 PATH_LAST_POSITION(p_s_path) - 1,
1580 PATH_LAST_POSITION(p_s_path) + 1);
1581 reiserfs_panic(p_s_sb, "PAP-5580", "item to "
1582 "convert does not exist (%K)",
1587 if (n_cut_size == 0) {
1588 pathrelse(p_s_path);
1592 s_cut_balance.insert_size[0] = n_cut_size;
1594 n_ret_value = fix_nodes(c_mode, &s_cut_balance, NULL, NULL);
1595 if (n_ret_value != REPEAT_SEARCH)
1598 PROC_INFO_INC(p_s_sb, cut_from_item_restarted);
1601 search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path);
1602 if (n_ret_value == POSITION_FOUND)
1605 reiserfs_warning(p_s_sb, "PAP-5610", "item %K not found",
1607 unfix_nodes(&s_cut_balance);
1608 return (n_ret_value == IO_ERROR) ? -EIO : -ENOENT;
1611 // check fix_nodes results (IO_ERROR or NO_DISK_SPACE)
1612 if (n_ret_value != CARRY_ON) {
1613 if (n_is_inode_locked) {
1614 // FIXME: this seems to be not needed: we are always able
1616 indirect_to_direct_roll_back(th, p_s_inode, p_s_path);
1618 if (n_ret_value == NO_DISK_SPACE)
1619 reiserfs_warning(p_s_sb, "reiserfs-5092",
1621 unfix_nodes(&s_cut_balance);
1625 /* go ahead and perform balancing */
1627 RFALSE(c_mode == M_PASTE || c_mode == M_INSERT, "invalid mode");
1629 /* Calculate number of bytes that need to be cut from the item. */
1632 M_DELETE) ? ih_item_len(get_ih(p_s_path)) : -s_cut_balance.
1635 n_ret_value = calc_deleted_bytes_number(&s_cut_balance, c_mode);
1637 n_ret_value = retval2;
1639 /* For direct items, we only change the quota when deleting the last
1642 p_le_ih = PATH_PITEM_HEAD(s_cut_balance.tb_path);
1643 if (!S_ISLNK(p_s_inode->i_mode) && is_direct_le_ih(p_le_ih)) {
1644 if (c_mode == M_DELETE &&
1645 (le_ih_k_offset(p_le_ih) & (p_s_sb->s_blocksize - 1)) ==
1647 // FIXME: this is to keep 3.5 happy
1648 REISERFS_I(p_s_inode)->i_first_direct_byte = U32_MAX;
1649 quota_cut_bytes = p_s_sb->s_blocksize + UNFM_P_SIZE;
1651 quota_cut_bytes = 0;
1654 #ifdef CONFIG_REISERFS_CHECK
1655 if (n_is_inode_locked) {
1656 struct item_head *le_ih =
1657 PATH_PITEM_HEAD(s_cut_balance.tb_path);
1658 /* we are going to complete indirect2direct conversion. Make
1659 sure, that we exactly remove last unformatted node pointer
1661 if (!is_indirect_le_ih(le_ih))
1662 reiserfs_panic(p_s_sb, "vs-5652",
1663 "item must be indirect %h", le_ih);
1665 if (c_mode == M_DELETE && ih_item_len(le_ih) != UNFM_P_SIZE)
1666 reiserfs_panic(p_s_sb, "vs-5653", "completing "
1667 "indirect2direct conversion indirect "
1668 "item %h being deleted must be of "
1669 "4 byte long", le_ih);
1672 && s_cut_balance.insert_size[0] != -UNFM_P_SIZE) {
1673 reiserfs_panic(p_s_sb, "vs-5654", "can not complete "
1674 "indirect2direct conversion of %h "
1675 "(CUT, insert_size==%d)",
1676 le_ih, s_cut_balance.insert_size[0]);
1678 /* it would be useful to make sure, that right neighboring
1679 item is direct item of this file */
1683 do_balance(&s_cut_balance, NULL, NULL, c_mode);
1684 if (n_is_inode_locked) {
1685 /* we've done an indirect->direct conversion. when the data block
1686 ** was freed, it was removed from the list of blocks that must
1687 ** be flushed before the transaction commits, make sure to
1688 ** unmap and invalidate it
1690 unmap_buffers(page, tail_pos);
1691 REISERFS_I(p_s_inode)->i_flags &= ~i_pack_on_close_mask;
1693 #ifdef REISERQUOTA_DEBUG
1694 reiserfs_debug(p_s_inode->i_sb, REISERFS_DEBUG_CODE,
1695 "reiserquota cut_from_item(): freeing %u id=%u type=%c",
1696 quota_cut_bytes, p_s_inode->i_uid, '?');
1698 DQUOT_FREE_SPACE_NODIRTY(p_s_inode, quota_cut_bytes);
1702 static void truncate_directory(struct reiserfs_transaction_handle *th,
1703 struct inode *inode)
1705 BUG_ON(!th->t_trans_id);
1707 reiserfs_error(inode->i_sb, "vs-5655", "link count != 0");
1709 set_le_key_k_offset(KEY_FORMAT_3_5, INODE_PKEY(inode), DOT_OFFSET);
1710 set_le_key_k_type(KEY_FORMAT_3_5, INODE_PKEY(inode), TYPE_DIRENTRY);
1711 reiserfs_delete_solid_item(th, inode, INODE_PKEY(inode));
1712 reiserfs_update_sd(th, inode);
1713 set_le_key_k_offset(KEY_FORMAT_3_5, INODE_PKEY(inode), SD_OFFSET);
1714 set_le_key_k_type(KEY_FORMAT_3_5, INODE_PKEY(inode), TYPE_STAT_DATA);
1717 /* Truncate file to the new size. Note, this must be called with a transaction
1719 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th, struct inode *p_s_inode, /* ->i_size contains new
1721 struct page *page, /* up to date for last block */
1722 int update_timestamps /* when it is called by
1723 file_release to convert
1724 the tail - no timestamps
1725 should be updated */
1728 INITIALIZE_PATH(s_search_path); /* Path to the current object item. */
1729 struct item_head *p_le_ih; /* Pointer to an item header. */
1730 struct cpu_key s_item_key; /* Key to search for a previous file item. */
1731 loff_t n_file_size, /* Old file size. */
1732 n_new_file_size; /* New file size. */
1733 int n_deleted; /* Number of deleted or truncated bytes. */
1737 BUG_ON(!th->t_trans_id);
1739 (S_ISREG(p_s_inode->i_mode) || S_ISDIR(p_s_inode->i_mode)
1740 || S_ISLNK(p_s_inode->i_mode)))
1743 if (S_ISDIR(p_s_inode->i_mode)) {
1744 // deletion of directory - no need to update timestamps
1745 truncate_directory(th, p_s_inode);
1749 /* Get new file size. */
1750 n_new_file_size = p_s_inode->i_size;
1752 // FIXME: note, that key type is unimportant here
1753 make_cpu_key(&s_item_key, p_s_inode, max_reiserfs_offset(p_s_inode),
1757 search_for_position_by_key(p_s_inode->i_sb, &s_item_key,
1759 if (retval == IO_ERROR) {
1760 reiserfs_error(p_s_inode->i_sb, "vs-5657",
1761 "i/o failure occurred trying to truncate %K",
1766 if (retval == POSITION_FOUND || retval == FILE_NOT_FOUND) {
1767 reiserfs_error(p_s_inode->i_sb, "PAP-5660",
1768 "wrong result %d of search for %K", retval,
1775 s_search_path.pos_in_item--;
1777 /* Get real file size (total length of all file items) */
1778 p_le_ih = PATH_PITEM_HEAD(&s_search_path);
1779 if (is_statdata_le_ih(p_le_ih))
1782 loff_t offset = le_ih_k_offset(p_le_ih);
1784 op_bytes_number(p_le_ih, p_s_inode->i_sb->s_blocksize);
1786 /* this may mismatch with real file size: if last direct item
1787 had no padding zeros and last unformatted node had no free
1788 space, this file would have this file size */
1789 n_file_size = offset + bytes - 1;
1792 * are we doing a full truncate or delete, if so
1793 * kick in the reada code
1795 if (n_new_file_size == 0)
1796 s_search_path.reada = PATH_READA | PATH_READA_BACK;
1798 if (n_file_size == 0 || n_file_size < n_new_file_size) {
1799 goto update_and_out;
1802 /* Update key to search for the last file item. */
1803 set_cpu_key_k_offset(&s_item_key, n_file_size);
1806 /* Cut or delete file item. */
1808 reiserfs_cut_from_item(th, &s_search_path, &s_item_key,
1809 p_s_inode, page, n_new_file_size);
1810 if (n_deleted < 0) {
1811 reiserfs_warning(p_s_inode->i_sb, "vs-5665",
1812 "reiserfs_cut_from_item failed");
1813 reiserfs_check_path(&s_search_path);
1817 RFALSE(n_deleted > n_file_size,
1818 "PAP-5670: reiserfs_cut_from_item: too many bytes deleted: deleted %d, file_size %lu, item_key %K",
1819 n_deleted, n_file_size, &s_item_key);
1821 /* Change key to search the last file item. */
1822 n_file_size -= n_deleted;
1824 set_cpu_key_k_offset(&s_item_key, n_file_size);
1826 /* While there are bytes to truncate and previous file item is presented in the tree. */
1829 ** This loop could take a really long time, and could log
1830 ** many more blocks than a transaction can hold. So, we do a polite
1831 ** journal end here, and if the transaction needs ending, we make
1832 ** sure the file is consistent before ending the current trans
1833 ** and starting a new one
1835 if (journal_transaction_should_end(th, 0) ||
1836 reiserfs_transaction_free_space(th) <= JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD) {
1837 int orig_len_alloc = th->t_blocks_allocated;
1838 pathrelse(&s_search_path);
1840 if (update_timestamps) {
1841 p_s_inode->i_mtime = p_s_inode->i_ctime =
1844 reiserfs_update_sd(th, p_s_inode);
1846 err = journal_end(th, p_s_inode->i_sb, orig_len_alloc);
1849 err = journal_begin(th, p_s_inode->i_sb,
1850 JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD + JOURNAL_PER_BALANCE_CNT * 4) ;
1853 reiserfs_update_inode_transaction(p_s_inode);
1855 } while (n_file_size > ROUND_UP(n_new_file_size) &&
1856 search_for_position_by_key(p_s_inode->i_sb, &s_item_key,
1857 &s_search_path) == POSITION_FOUND);
1859 RFALSE(n_file_size > ROUND_UP(n_new_file_size),
1860 "PAP-5680: truncate did not finish: new_file_size %Ld, current %Ld, oid %d",
1861 n_new_file_size, n_file_size, s_item_key.on_disk_key.k_objectid);
1864 if (update_timestamps) {
1865 // this is truncate, not file closing
1866 p_s_inode->i_mtime = p_s_inode->i_ctime = CURRENT_TIME_SEC;
1868 reiserfs_update_sd(th, p_s_inode);
1871 pathrelse(&s_search_path);
1875 #ifdef CONFIG_REISERFS_CHECK
1876 // this makes sure, that we __append__, not overwrite or add holes
1877 static void check_research_for_paste(struct treepath *path,
1878 const struct cpu_key *p_s_key)
1880 struct item_head *found_ih = get_ih(path);
1882 if (is_direct_le_ih(found_ih)) {
1883 if (le_ih_k_offset(found_ih) +
1884 op_bytes_number(found_ih,
1885 get_last_bh(path)->b_size) !=
1886 cpu_key_k_offset(p_s_key)
1887 || op_bytes_number(found_ih,
1888 get_last_bh(path)->b_size) !=
1890 reiserfs_panic(NULL, "PAP-5720", "found direct item "
1891 "%h or position (%d) does not match "
1892 "to key %K", found_ih,
1893 pos_in_item(path), p_s_key);
1895 if (is_indirect_le_ih(found_ih)) {
1896 if (le_ih_k_offset(found_ih) +
1897 op_bytes_number(found_ih,
1898 get_last_bh(path)->b_size) !=
1899 cpu_key_k_offset(p_s_key)
1900 || I_UNFM_NUM(found_ih) != pos_in_item(path)
1901 || get_ih_free_space(found_ih) != 0)
1902 reiserfs_panic(NULL, "PAP-5730", "found indirect "
1903 "item (%h) or position (%d) does not "
1904 "match to key (%K)",
1905 found_ih, pos_in_item(path), p_s_key);
1908 #endif /* config reiserfs check */
1910 /* Paste bytes to the existing item. Returns bytes number pasted into the item. */
1911 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th, struct treepath *p_s_search_path, /* Path to the pasted item. */
1912 const struct cpu_key *p_s_key, /* Key to search for the needed item. */
1913 struct inode *inode, /* Inode item belongs to */
1914 const char *p_c_body, /* Pointer to the bytes to paste. */
1916 { /* Size of pasted bytes. */
1917 struct tree_balance s_paste_balance;
1921 BUG_ON(!th->t_trans_id);
1923 fs_gen = get_generation(inode->i_sb);
1925 #ifdef REISERQUOTA_DEBUG
1926 reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
1927 "reiserquota paste_into_item(): allocating %u id=%u type=%c",
1928 n_pasted_size, inode->i_uid,
1929 key2type(&(p_s_key->on_disk_key)));
1932 if (DQUOT_ALLOC_SPACE_NODIRTY(inode, n_pasted_size)) {
1933 pathrelse(p_s_search_path);
1936 init_tb_struct(th, &s_paste_balance, th->t_super, p_s_search_path,
1938 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
1939 s_paste_balance.key = p_s_key->on_disk_key;
1942 /* DQUOT_* can schedule, must check before the fix_nodes */
1943 if (fs_changed(fs_gen, inode->i_sb)) {
1948 fix_nodes(M_PASTE, &s_paste_balance, NULL,
1949 p_c_body)) == REPEAT_SEARCH) {
1951 /* file system changed while we were in the fix_nodes */
1952 PROC_INFO_INC(th->t_super, paste_into_item_restarted);
1954 search_for_position_by_key(th->t_super, p_s_key,
1956 if (retval == IO_ERROR) {
1960 if (retval == POSITION_FOUND) {
1961 reiserfs_warning(inode->i_sb, "PAP-5710",
1962 "entry or pasted byte (%K) exists",
1967 #ifdef CONFIG_REISERFS_CHECK
1968 check_research_for_paste(p_s_search_path, p_s_key);
1972 /* Perform balancing after all resources are collected by fix_nodes, and
1973 accessing them will not risk triggering schedule. */
1974 if (retval == CARRY_ON) {
1975 do_balance(&s_paste_balance, NULL /*ih */ , p_c_body, M_PASTE);
1978 retval = (retval == NO_DISK_SPACE) ? -ENOSPC : -EIO;
1980 /* this also releases the path */
1981 unfix_nodes(&s_paste_balance);
1982 #ifdef REISERQUOTA_DEBUG
1983 reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
1984 "reiserquota paste_into_item(): freeing %u id=%u type=%c",
1985 n_pasted_size, inode->i_uid,
1986 key2type(&(p_s_key->on_disk_key)));
1988 DQUOT_FREE_SPACE_NODIRTY(inode, n_pasted_size);
1992 /* Insert new item into the buffer at the path. */
1993 int reiserfs_insert_item(struct reiserfs_transaction_handle *th, struct treepath *p_s_path, /* Path to the inserteded item. */
1994 const struct cpu_key *key, struct item_head *p_s_ih, /* Pointer to the item header to insert. */
1995 struct inode *inode, const char *p_c_body)
1996 { /* Pointer to the bytes to insert. */
1997 struct tree_balance s_ins_balance;
2000 int quota_bytes = 0;
2002 BUG_ON(!th->t_trans_id);
2004 if (inode) { /* Do we count quotas for item? */
2005 fs_gen = get_generation(inode->i_sb);
2006 quota_bytes = ih_item_len(p_s_ih);
2008 /* hack so the quota code doesn't have to guess if the file has
2009 ** a tail, links are always tails, so there's no guessing needed
2011 if (!S_ISLNK(inode->i_mode) && is_direct_le_ih(p_s_ih)) {
2012 quota_bytes = inode->i_sb->s_blocksize + UNFM_P_SIZE;
2014 #ifdef REISERQUOTA_DEBUG
2015 reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
2016 "reiserquota insert_item(): allocating %u id=%u type=%c",
2017 quota_bytes, inode->i_uid, head2type(p_s_ih));
2019 /* We can't dirty inode here. It would be immediately written but
2020 * appropriate stat item isn't inserted yet... */
2021 if (DQUOT_ALLOC_SPACE_NODIRTY(inode, quota_bytes)) {
2022 pathrelse(p_s_path);
2026 init_tb_struct(th, &s_ins_balance, th->t_super, p_s_path,
2027 IH_SIZE + ih_item_len(p_s_ih));
2028 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2029 s_ins_balance.key = key->on_disk_key;
2031 /* DQUOT_* can schedule, must check to be sure calling fix_nodes is safe */
2032 if (inode && fs_changed(fs_gen, inode->i_sb)) {
2037 fix_nodes(M_INSERT, &s_ins_balance, p_s_ih,
2038 p_c_body)) == REPEAT_SEARCH) {
2040 /* file system changed while we were in the fix_nodes */
2041 PROC_INFO_INC(th->t_super, insert_item_restarted);
2042 retval = search_item(th->t_super, key, p_s_path);
2043 if (retval == IO_ERROR) {
2047 if (retval == ITEM_FOUND) {
2048 reiserfs_warning(th->t_super, "PAP-5760",
2049 "key %K already exists in the tree",
2056 /* make balancing after all resources will be collected at a time */
2057 if (retval == CARRY_ON) {
2058 do_balance(&s_ins_balance, p_s_ih, p_c_body, M_INSERT);
2062 retval = (retval == NO_DISK_SPACE) ? -ENOSPC : -EIO;
2064 /* also releases the path */
2065 unfix_nodes(&s_ins_balance);
2066 #ifdef REISERQUOTA_DEBUG
2067 reiserfs_debug(th->t_super, REISERFS_DEBUG_CODE,
2068 "reiserquota insert_item(): freeing %u id=%u type=%c",
2069 quota_bytes, inode->i_uid, head2type(p_s_ih));
2072 DQUOT_FREE_SPACE_NODIRTY(inode, quota_bytes);