fs/reiserfs/stree.c

   1 /*
   2  *  Copyright 2000 by Hans Reiser, licensing governed by reiserfs/README
   3  */
   4
   5 /*
   6  *  Written by Anatoly P. Pinchuk pap@namesys.botik.ru
   7  *  Programm System Institute
   8  *  Pereslavl-Zalessky Russia
   9  */
  10
  11 #include <linux/time.h>
  12 #include <linux/string.h>
  13 #include <linux/pagemap.h>
  14 #include <linux/bio.h>
  15 #include "reiserfs.h"
  16 #include <linux/buffer_head.h>
  17 #include <linux/quotaops.h>
  18
  19 /* Does the buffer contain a disk block which is in the tree. */
  20 inline int B_IS_IN_TREE(const struct buffer_head *bh)
  21 {
  22
  23         RFALSE(B_LEVEL(bh) > MAX_HEIGHT,
  24                "PAP-1010: block (%b) has too big level (%z)", bh, bh);
  25
  26         return (B_LEVEL(bh) != FREE_LEVEL);
  27 }
  28
  29 /* to get item head in le form */
  30 inline void copy_item_head(struct item_head *to,
  31                            const struct item_head *from)
  32 {
  33         memcpy(to, from, IH_SIZE);
  34 }
  35
  36 /*
  37  * k1 is pointer to on-disk structure which is stored in little-endian
  38  * form. k2 is pointer to cpu variable. For key of items of the same
  39  * object this returns 0.
  40  * Returns: -1 if key1 < key2
  41  * 0 if key1 == key2
  42  * 1 if key1 > key2
  43  */
  44 inline int comp_short_keys(const struct reiserfs_key *le_key,
  45                            const struct cpu_key *cpu_key)
  46 {
  47         __u32 n;
  48         n = le32_to_cpu(le_key->k_dir_id);
  49         if (n < cpu_key->on_disk_key.k_dir_id)
  50                 return -1;
  51         if (n > cpu_key->on_disk_key.k_dir_id)
  52                 return 1;
  53         n = le32_to_cpu(le_key->k_objectid);
  54         if (n < cpu_key->on_disk_key.k_objectid)
  55                 return -1;
  56         if (n > cpu_key->on_disk_key.k_objectid)
  57                 return 1;
  58         return 0;
  59 }
  60
  61 /*
  62  * k1 is pointer to on-disk structure which is stored in little-endian
  63  * form. k2 is pointer to cpu variable.
  64  * Compare keys using all 4 key fields.
  65  * Returns: -1 if key1 < key2 0
  66  * if key1 = key2 1 if key1 > key2
  67  */
  68 static inline int comp_keys(const struct reiserfs_key *le_key,
  69                             const struct cpu_key *cpu_key)
  70 {
  71         int retval;
  72
  73         retval = comp_short_keys(le_key, cpu_key);
  74         if (retval)
  75                 return retval;
  76         if (le_key_k_offset(le_key_version(le_key), le_key) <
  77             cpu_key_k_offset(cpu_key))
  78                 return -1;
  79         if (le_key_k_offset(le_key_version(le_key), le_key) >
  80             cpu_key_k_offset(cpu_key))
  81                 return 1;
  82
  83         if (cpu_key->key_length == 3)
  84                 return 0;
  85
  86         /* this part is needed only when tail conversion is in progress */
  87         if (le_key_k_type(le_key_version(le_key), le_key) <
  88             cpu_key_k_type(cpu_key))
  89                 return -1;
  90
  91         if (le_key_k_type(le_key_version(le_key), le_key) >
  92             cpu_key_k_type(cpu_key))
  93                 return 1;
  94
  95         return 0;
  96 }
  97
  98 inline int comp_short_le_keys(const struct reiserfs_key *key1,
  99                               const struct reiserfs_key *key2)
 100 {
 101         __u32 *k1_u32, *k2_u32;
 102         int key_length = REISERFS_SHORT_KEY_LEN;
 103
 104         k1_u32 = (__u32 *) key1;
 105         k2_u32 = (__u32 *) key2;
 106         for (; key_length--; ++k1_u32, ++k2_u32) {
 107                 if (le32_to_cpu(*k1_u32) < le32_to_cpu(*k2_u32))
 108                         return -1;
 109                 if (le32_to_cpu(*k1_u32) > le32_to_cpu(*k2_u32))
 110                         return 1;
 111         }
 112         return 0;
 113 }
 114
 115 inline void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from)
 116 {
 117         int version;
 118         to->on_disk_key.k_dir_id = le32_to_cpu(from->k_dir_id);
 119         to->on_disk_key.k_objectid = le32_to_cpu(from->k_objectid);
 120
 121         /* find out version of the key */
 122         version = le_key_version(from);
 123         to->version = version;
 124         to->on_disk_key.k_offset = le_key_k_offset(version, from);
 125         to->on_disk_key.k_type = le_key_k_type(version, from);
 126 }
 127
 128 /*
 129  * this does not say which one is bigger, it only returns 1 if keys
 130  * are not equal, 0 otherwise
 131  */
 132 inline int comp_le_keys(const struct reiserfs_key *k1,
 133                         const struct reiserfs_key *k2)
 134 {
 135         return memcmp(k1, k2, sizeof(struct reiserfs_key));
 136 }
 137
 138 /**************************************************************************
 139  *  Binary search toolkit function                                        *
 140  *  Search for an item in the array by the item key                       *
 141  *  Returns:    1 if found,  0 if not found;                              *
 142  *        *pos = number of the searched element if found, else the        *
 143  *        number of the first element that is larger than key.            *
 144  **************************************************************************/
 145 /*
 146  * For those not familiar with binary search: lbound is the leftmost item
 147  * that it could be, rbound the rightmost item that it could be.  We examine
 148  * the item halfway between lbound and rbound, and that tells us either
 149  * that we can increase lbound, or decrease rbound, or that we have found it,
 150  * or if lbound <= rbound that there are no possible items, and we have not
 151  * found it. With each examination we cut the number of possible items it
 152  * could be by one more than half rounded down, or we find it.
 153  */
 154 static inline int bin_search(const void *key,   /* Key to search for. */
 155                              const void *base,  /* First item in the array. */
 156                              int num,   /* Number of items in the array. */
 157                              /*
 158                               * Item size in the array.  searched. Lest the
 159                               * reader be confused, note that this is crafted
 160                               * as a general function, and when it is applied
 161                               * specifically to the array of item headers in a
 162                               * node, width is actually the item header size
 163                               * not the item size.
 164                               */
 165                              int width,
 166                              int *pos /* Number of the searched for element. */
 167     )
 168 {
 169         int rbound, lbound, j;
 170
 171         for (j = ((rbound = num - 1) + (lbound = 0)) / 2;
 172              lbound <= rbound; j = (rbound + lbound) / 2)
 173                 switch (comp_keys
 174                         ((struct reiserfs_key *)((char *)base + j * width),
 175                          (struct cpu_key *)key)) {
 176                 case -1:
 177                         lbound = j + 1;
 178                         continue;
 179                 case 1:
 180                         rbound = j - 1;
 181                         continue;
 182                 case 0:
 183                         *pos = j;
 184                         return ITEM_FOUND;      /* Key found in the array.  */
 185                 }
 186
 187         /*
 188          * bin_search did not find given key, it returns position of key,
 189          * that is minimal and greater than the given one.
 190          */
 191         *pos = lbound;
 192         return ITEM_NOT_FOUND;
 193 }
 194
 195
 196 /* Minimal possible key. It is never in the tree. */
 197 const struct reiserfs_key MIN_KEY = { 0, 0, {{0, 0},} };
 198
 199 /* Maximal possible key. It is never in the tree. */
 200 static const struct reiserfs_key MAX_KEY = {
 201         cpu_to_le32(0xffffffff),
 202         cpu_to_le32(0xffffffff),
 203         {{cpu_to_le32(0xffffffff),
 204           cpu_to_le32(0xffffffff)},}
 205 };
 206
 207 /*
 208  * Get delimiting key of the buffer by looking for it in the buffers in the
 209  * path, starting from the bottom of the path, and going upwards.  We must
 210  * check the path's validity at each step.  If the key is not in the path,
 211  * there is no delimiting key in the tree (buffer is first or last buffer
 212  * in tree), and in this case we return a special key, either MIN_KEY or
 213  * MAX_KEY.
 214  */
 215 static inline const struct reiserfs_key *get_lkey(const struct treepath *chk_path,
 216                                                   const struct super_block *sb)
 217 {
 218         int position, path_offset = chk_path->path_length;
 219         struct buffer_head *parent;
 220
 221         RFALSE(path_offset < FIRST_PATH_ELEMENT_OFFSET,
 222                "PAP-5010: invalid offset in the path");
 223
 224         /* While not higher in path than first element. */
 225         while (path_offset-- > FIRST_PATH_ELEMENT_OFFSET) {
 226
 227                 RFALSE(!buffer_uptodate
 228                        (PATH_OFFSET_PBUFFER(chk_path, path_offset)),
 229                        "PAP-5020: parent is not uptodate");
 230
 231                 /* Parent at the path is not in the tree now. */
 232                 if (!B_IS_IN_TREE
 233                     (parent =
 234                      PATH_OFFSET_PBUFFER(chk_path, path_offset)))
 235                         return &MAX_KEY;
 236                 /* Check whether position in the parent is correct. */
 237                 if ((position =
 238                      PATH_OFFSET_POSITION(chk_path,
 239                                           path_offset)) >
 240                     B_NR_ITEMS(parent))
 241                         return &MAX_KEY;
 242                 /* Check whether parent at the path really points to the child. */
 243                 if (B_N_CHILD_NUM(parent, position) !=
 244                     PATH_OFFSET_PBUFFER(chk_path,
 245                                         path_offset + 1)->b_blocknr)
 246                         return &MAX_KEY;
 247                 /*
 248                  * Return delimiting key if position in the parent
 249                  * is not equal to zero.
 250                  */
 251                 if (position)
 252                         return internal_key(parent, position - 1);
 253         }
 254         /* Return MIN_KEY if we are in the root of the buffer tree. */
 255         if (PATH_OFFSET_PBUFFER(chk_path, FIRST_PATH_ELEMENT_OFFSET)->
 256             b_blocknr == SB_ROOT_BLOCK(sb))
 257                 return &MIN_KEY;
 258         return &MAX_KEY;
 259 }
 260
 261 /* Get delimiting key of the buffer at the path and its right neighbor. */
 262 inline const struct reiserfs_key *get_rkey(const struct treepath *chk_path,
 263                                            const struct super_block *sb)
 264 {
 265         int position, path_offset = chk_path->path_length;
 266         struct buffer_head *parent;
 267
 268         RFALSE(path_offset < FIRST_PATH_ELEMENT_OFFSET,
 269                "PAP-5030: invalid offset in the path");
 270
 271         while (path_offset-- > FIRST_PATH_ELEMENT_OFFSET) {
 272
 273                 RFALSE(!buffer_uptodate
 274                        (PATH_OFFSET_PBUFFER(chk_path, path_offset)),
 275                        "PAP-5040: parent is not uptodate");
 276
 277                 /* Parent at the path is not in the tree now. */
 278                 if (!B_IS_IN_TREE
 279                     (parent =
 280                      PATH_OFFSET_PBUFFER(chk_path, path_offset)))
 281                         return &MIN_KEY;
 282                 /* Check whether position in the parent is correct. */
 283                 if ((position =
 284                      PATH_OFFSET_POSITION(chk_path,
 285                                           path_offset)) >
 286                     B_NR_ITEMS(parent))
 287                         return &MIN_KEY;
 288                 /*
 289                  * Check whether parent at the path really points
 290                  * to the child.
 291                  */
 292                 if (B_N_CHILD_NUM(parent, position) !=
 293                     PATH_OFFSET_PBUFFER(chk_path,
 294                                         path_offset + 1)->b_blocknr)
 295                         return &MIN_KEY;
 296
 297                 /*
 298                  * Return delimiting key if position in the parent
 299                  * is not the last one.
 300                  */
 301                 if (position != B_NR_ITEMS(parent))
 302                         return internal_key(parent, position);
 303         }
 304
 305         /* Return MAX_KEY if we are in the root of the buffer tree. */
 306         if (PATH_OFFSET_PBUFFER(chk_path, FIRST_PATH_ELEMENT_OFFSET)->
 307             b_blocknr == SB_ROOT_BLOCK(sb))
 308                 return &MAX_KEY;
 309         return &MIN_KEY;
 310 }
 311
 312 /*
 313  * Check whether a key is contained in the tree rooted from a buffer at a path.
 314  * This works by looking at the left and right delimiting keys for the buffer
 315  * in the last path_element in the path.  These delimiting keys are stored
 316  * at least one level above that buffer in the tree. If the buffer is the
 317  * first or last node in the tree order then one of the delimiting keys may
 318  * be absent, and in this case get_lkey and get_rkey return a special key
 319  * which is MIN_KEY or MAX_KEY.
 320  */
 321 static inline int key_in_buffer(
 322                                 /* Path which should be checked. */
 323                                 struct treepath *chk_path,
 324                                 /* Key which should be checked. */
 325                                 const struct cpu_key *key,
 326                                 struct super_block *sb
 327     )
 328 {
 329
 330         RFALSE(!key || chk_path->path_length < FIRST_PATH_ELEMENT_OFFSET
 331                || chk_path->path_length > MAX_HEIGHT,
 332                "PAP-5050: pointer to the key(%p) is NULL or invalid path length(%d)",
 333                key, chk_path->path_length);
 334         RFALSE(!PATH_PLAST_BUFFER(chk_path)->b_bdev,
 335                "PAP-5060: device must not be NODEV");
 336
 337         if (comp_keys(get_lkey(chk_path, sb), key) == 1)
 338                 /* left delimiting key is bigger, that the key we look for */
 339                 return 0;
 340         /*  if ( comp_keys(key, get_rkey(chk_path, sb)) != -1 ) */
 341         if (comp_keys(get_rkey(chk_path, sb), key) != 1)
 342                 /* key must be less than right delimitiing key */
 343                 return 0;
 344         return 1;
 345 }
 346
 347 int reiserfs_check_path(struct treepath *p)
 348 {
 349         RFALSE(p->path_length != ILLEGAL_PATH_ELEMENT_OFFSET,
 350                "path not properly relsed");
 351         return 0;
 352 }
 353
 354 /*
 355  * Drop the reference to each buffer in a path and restore
 356  * dirty bits clean when preparing the buffer for the log.
 357  * This version should only be called from fix_nodes()
 358  */
 359 void pathrelse_and_restore(struct super_block *sb,
 360                            struct treepath *search_path)
 361 {
 362         int path_offset = search_path->path_length;
 363
 364         RFALSE(path_offset < ILLEGAL_PATH_ELEMENT_OFFSET,
 365                "clm-4000: invalid path offset");
 366
 367         while (path_offset > ILLEGAL_PATH_ELEMENT_OFFSET) {
 368                 struct buffer_head *bh;
 369                 bh = PATH_OFFSET_PBUFFER(search_path, path_offset--);
 370                 reiserfs_restore_prepared_buffer(sb, bh);
 371                 brelse(bh);
 372         }
 373         search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET;
 374 }
 375
 376 /* Drop the reference to each buffer in a path */
 377 void pathrelse(struct treepath *search_path)
 378 {
 379         int path_offset = search_path->path_length;
 380
 381         RFALSE(path_offset < ILLEGAL_PATH_ELEMENT_OFFSET,
 382                "PAP-5090: invalid path offset");
 383
 384         while (path_offset > ILLEGAL_PATH_ELEMENT_OFFSET)
 385                 brelse(PATH_OFFSET_PBUFFER(search_path, path_offset--));
 386
 387         search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET;
 388 }
 389
 390 static int is_leaf(char *buf, int blocksize, struct buffer_head *bh)
 391 {
 392         struct block_head *blkh;
 393         struct item_head *ih;
 394         int used_space;
 395         int prev_location;
 396         int i;
 397         int nr;
 398
 399         blkh = (struct block_head *)buf;
 400         if (blkh_level(blkh) != DISK_LEAF_NODE_LEVEL) {
 401                 reiserfs_warning(NULL, "reiserfs-5080",
 402                                  "this should be caught earlier");
 403                 return 0;
 404         }
 405
 406         nr = blkh_nr_item(blkh);
 407         if (nr < 1 || nr > ((blocksize - BLKH_SIZE) / (IH_SIZE + MIN_ITEM_LEN))) {
 408                 /* item number is too big or too small */
 409                 reiserfs_warning(NULL, "reiserfs-5081",
 410                                  "nr_item seems wrong: %z", bh);
 411                 return 0;
 412         }
 413         ih = (struct item_head *)(buf + BLKH_SIZE) + nr - 1;
 414         used_space = BLKH_SIZE + IH_SIZE * nr + (blocksize - ih_location(ih));
 415
 416         /* free space does not match to calculated amount of use space */
 417         if (used_space != blocksize - blkh_free_space(blkh)) {
 418                 reiserfs_warning(NULL, "reiserfs-5082",
 419                                  "free space seems wrong: %z", bh);
 420                 return 0;
 421         }
 422         /*
 423          * FIXME: it is_leaf will hit performance too much - we may have
 424          * return 1 here
 425          */
 426
 427         /* check tables of item heads */
 428         ih = (struct item_head *)(buf + BLKH_SIZE);
 429         prev_location = blocksize;
 430         for (i = 0; i < nr; i++, ih++) {
 431                 if (le_ih_k_type(ih) == TYPE_ANY) {
 432                         reiserfs_warning(NULL, "reiserfs-5083",
 433                                          "wrong item type for item %h",
 434                                          ih);
 435                         return 0;
 436                 }
 437                 if (ih_location(ih) >= blocksize
 438                     || ih_location(ih) < IH_SIZE * nr) {
 439                         reiserfs_warning(NULL, "reiserfs-5084",
 440                                          "item location seems wrong: %h",
 441                                          ih);
 442                         return 0;
 443                 }
 444                 if (ih_item_len(ih) < 1
 445                     || ih_item_len(ih) > MAX_ITEM_LEN(blocksize)) {
 446                         reiserfs_warning(NULL, "reiserfs-5085",
 447                                          "item length seems wrong: %h",
 448                                          ih);
 449                         return 0;
 450                 }
 451                 if (prev_location - ih_location(ih) != ih_item_len(ih)) {
 452                         reiserfs_warning(NULL, "reiserfs-5086",
 453                                          "item location seems wrong "
 454                                          "(second one): %h", ih);
 455                         return 0;
 456                 }
 457                 prev_location = ih_location(ih);
 458         }
 459
 460         /* one may imagine many more checks */
 461         return 1;
 462 }
 463
 464 /* returns 1 if buf looks like an internal node, 0 otherwise */
 465 static int is_internal(char *buf, int blocksize, struct buffer_head *bh)
 466 {
 467         struct block_head *blkh;
 468         int nr;
 469         int used_space;
 470
 471         blkh = (struct block_head *)buf;
 472         nr = blkh_level(blkh);
 473         if (nr <= DISK_LEAF_NODE_LEVEL || nr > MAX_HEIGHT) {
 474                 /* this level is not possible for internal nodes */
 475                 reiserfs_warning(NULL, "reiserfs-5087",
 476                                  "this should be caught earlier");
 477                 return 0;
 478         }
 479
 480         nr = blkh_nr_item(blkh);
 481         /* for internal which is not root we might check min number of keys */
 482         if (nr > (blocksize - BLKH_SIZE - DC_SIZE) / (KEY_SIZE + DC_SIZE)) {
 483                 reiserfs_warning(NULL, "reiserfs-5088",
 484                                  "number of key seems wrong: %z", bh);
 485                 return 0;
 486         }
 487
 488         used_space = BLKH_SIZE + KEY_SIZE * nr + DC_SIZE * (nr + 1);
 489         if (used_space != blocksize - blkh_free_space(blkh)) {
 490                 reiserfs_warning(NULL, "reiserfs-5089",
 491                                  "free space seems wrong: %z", bh);
 492                 return 0;
 493         }
 494
 495         /* one may imagine many more checks */
 496         return 1;
 497 }
 498
 499 /*
 500  * make sure that bh contains formatted node of reiserfs tree of
 501  * 'level'-th level
 502  */
 503 static int is_tree_node(struct buffer_head *bh, int level)
 504 {
 505         if (B_LEVEL(bh) != level) {
 506                 reiserfs_warning(NULL, "reiserfs-5090", "node level %d does "
 507                                  "not match to the expected one %d",
 508                                  B_LEVEL(bh), level);
 509                 return 0;
 510         }
 511         if (level == DISK_LEAF_NODE_LEVEL)
 512                 return is_leaf(bh->b_data, bh->b_size, bh);
 513
 514         return is_internal(bh->b_data, bh->b_size, bh);
 515 }
 516
 517 #define SEARCH_BY_KEY_READA 16
 518
 519 /*
 520  * The function is NOT SCHEDULE-SAFE!
 521  * It might unlock the write lock if we needed to wait for a block
 522  * to be read. Note that in this case it won't recover the lock to avoid
 523  * high contention resulting from too much lock requests, especially
 524  * the caller (search_by_key) will perform other schedule-unsafe
 525  * operations just after calling this function.
 526  *
 527  * @return depth of lock to be restored after read completes
 528  */
 529 static int search_by_key_reada(struct super_block *s,
 530                                 struct buffer_head **bh,
 531                                 b_blocknr_t *b, int num)
 532 {
 533         int i, j;
 534         int depth = -1;
 535
 536         for (i = 0; i < num; i++) {
 537                 bh[i] = sb_getblk(s, b[i]);
 538         }
 539         /*
 540          * We are going to read some blocks on which we
 541          * have a reference. It's safe, though we might be
 542          * reading blocks concurrently changed if we release
 543          * the lock. But it's still fine because we check later
 544          * if the tree changed
 545          */
 546         for (j = 0; j < i; j++) {
 547                 /*
 548                  * note, this needs attention if we are getting rid of the BKL
 549                  * you have to make sure the prepared bit isn't set on this
 550                  * buffer
 551                  */
 552                 if (!buffer_uptodate(bh[j])) {
 553                         if (depth == -1)
 554                                 depth = reiserfs_write_unlock_nested(s);
 555                         ll_rw_block(REQ_OP_READ, REQ_RAHEAD, 1, bh + j);
 556                 }
 557                 brelse(bh[j]);
 558         }
 559         return depth;
 560 }
 561
 562 /*
 563  * This function fills up the path from the root to the leaf as it
 564  * descends the tree looking for the key.  It uses reiserfs_bread to
 565  * try to find buffers in the cache given their block number.  If it
 566  * does not find them in the cache it reads them from disk.  For each
 567  * node search_by_key finds using reiserfs_bread it then uses
 568  * bin_search to look through that node.  bin_search will find the
 569  * position of the block_number of the next node if it is looking
 570  * through an internal node.  If it is looking through a leaf node
 571  * bin_search will find the position of the item which has key either
 572  * equal to given key, or which is the maximal key less than the given
 573  * key.  search_by_key returns a path that must be checked for the
 574  * correctness of the top of the path but need not be checked for the
 575  * correctness of the bottom of the path
 576  */
 577 /*
 578  * search_by_key - search for key (and item) in stree
 579  * @sb: superblock
 580  * @key: pointer to key to search for
 581  * @search_path: Allocated and initialized struct treepath; Returned filled
 582  *               on success.
 583  * @stop_level: How far down the tree to search, Use DISK_LEAF_NODE_LEVEL to
 584  *              stop at leaf level.
 585  *
 586  * The function is NOT SCHEDULE-SAFE!
 587  */
 588 int search_by_key(struct super_block *sb, const struct cpu_key *key,
 589                   struct treepath *search_path, int stop_level)
 590 {
 591         b_blocknr_t block_number;
 592         int expected_level;
 593         struct buffer_head *bh;
 594         struct path_element *last_element;
 595         int node_level, retval;
 596         int fs_gen;
 597         struct buffer_head *reada_bh[SEARCH_BY_KEY_READA];
 598         b_blocknr_t reada_blocks[SEARCH_BY_KEY_READA];
 599         int reada_count = 0;
 600
 601 #ifdef CONFIG_REISERFS_CHECK
 602         int repeat_counter = 0;
 603 #endif
 604
 605         PROC_INFO_INC(sb, search_by_key);
 606
 607         /*
 608          * As we add each node to a path we increase its count.  This means
 609          * that we must be careful to release all nodes in a path before we
 610          * either discard the path struct or re-use the path struct, as we
 611          * do here.
 612          */
 613
 614         pathrelse(search_path);
 615
 616         /*
 617          * With each iteration of this loop we search through the items in the
 618          * current node, and calculate the next current node(next path element)
 619          * for the next iteration of this loop..
 620          */
 621         block_number = SB_ROOT_BLOCK(sb);
 622         expected_level = -1;
 623         while (1) {
 624
 625 #ifdef CONFIG_REISERFS_CHECK
 626                 if (!(++repeat_counter % 50000))
 627                         reiserfs_warning(sb, "PAP-5100",
 628                                          "%s: there were %d iterations of "
 629                                          "while loop looking for key %K",
 630                                          current->comm, repeat_counter,
 631                                          key);
 632 #endif
 633
 634                 /* prep path to have another element added to it. */
 635                 last_element =
 636                     PATH_OFFSET_PELEMENT(search_path,
 637                                          ++search_path->path_length);
 638                 fs_gen = get_generation(sb);
 639
 640                 /*
 641                  * Read the next tree node, and set the last element
 642                  * in the path to have a pointer to it.
 643                  */
 644                 if ((bh = last_element->pe_buffer =
 645                      sb_getblk(sb, block_number))) {
 646
 647                         /*
 648                          * We'll need to drop the lock if we encounter any
 649                          * buffers that need to be read. If all of them are
 650                          * already up to date, we don't need to drop the lock.
 651                          */
 652                         int depth = -1;
 653
 654                         if (!buffer_uptodate(bh) && reada_count > 1)
 655                                 depth = search_by_key_reada(sb, reada_bh,
 656                                                     reada_blocks, reada_count);
 657
 658                         if (!buffer_uptodate(bh) && depth == -1)
 659                                 depth = reiserfs_write_unlock_nested(sb);
 660
 661                         ll_rw_block(REQ_OP_READ, 0, 1, &bh);
 662                         wait_on_buffer(bh);
 663
 664                         if (depth != -1)
 665                                 reiserfs_write_lock_nested(sb, depth);
 666                         if (!buffer_uptodate(bh))
 667                                 goto io_error;
 668                 } else {
 669 io_error:
 670                         search_path->path_length--;
 671                         pathrelse(search_path);
 672                         return IO_ERROR;
 673                 }
 674                 reada_count = 0;
 675                 if (expected_level == -1)
 676                         expected_level = SB_TREE_HEIGHT(sb);
 677                 expected_level--;
 678
 679                 /*
 680                  * It is possible that schedule occurred. We must check
 681                  * whether the key to search is still in the tree rooted
 682                  * from the current buffer. If not then repeat search
 683                  * from the root.
 684                  */
 685                 if (fs_changed(fs_gen, sb) &&
 686                     (!B_IS_IN_TREE(bh) ||
 687                      B_LEVEL(bh) != expected_level ||
 688                      !key_in_buffer(search_path, key, sb))) {
 689                         PROC_INFO_INC(sb, search_by_key_fs_changed);
 690                         PROC_INFO_INC(sb, search_by_key_restarted);
 691                         PROC_INFO_INC(sb,
 692                                       sbk_restarted[expected_level - 1]);
 693                         pathrelse(search_path);
 694
 695                         /*
 696                          * Get the root block number so that we can
 697                          * repeat the search starting from the root.
 698                          */
 699                         block_number = SB_ROOT_BLOCK(sb);
 700                         expected_level = -1;
 701
 702                         /* repeat search from the root */
 703                         continue;
 704                 }
 705
 706                 /*
 707                  * only check that the key is in the buffer if key is not
 708                  * equal to the MAX_KEY. Latter case is only possible in
 709                  * "finish_unfinished()" processing during mount.
 710                  */
 711                 RFALSE(comp_keys(&MAX_KEY, key) &&
 712                        !key_in_buffer(search_path, key, sb),
 713                        "PAP-5130: key is not in the buffer");
 714 #ifdef CONFIG_REISERFS_CHECK
 715                 if (REISERFS_SB(sb)->cur_tb) {
 716                         print_cur_tb("5140");
 717                         reiserfs_panic(sb, "PAP-5140",
 718                                        "schedule occurred in do_balance!");
 719                 }
 720 #endif
 721
 722                 /*
 723                  * make sure, that the node contents look like a node of
 724                  * certain level
 725                  */
 726                 if (!is_tree_node(bh, expected_level)) {
 727                         reiserfs_error(sb, "vs-5150",
 728                                        "invalid format found in block %ld. "
 729                                        "Fsck?", bh->b_blocknr);
 730                         pathrelse(search_path);
 731                         return IO_ERROR;
 732                 }
 733
 734                 /* ok, we have acquired next formatted node in the tree */
 735                 node_level = B_LEVEL(bh);
 736
 737                 PROC_INFO_BH_STAT(sb, bh, node_level - 1);
 738
 739                 RFALSE(node_level < stop_level,
 740                        "vs-5152: tree level (%d) is less than stop level (%d)",
 741                        node_level, stop_level);
 742
 743                 retval = bin_search(key, item_head(bh, 0),
 744                                       B_NR_ITEMS(bh),
 745                                       (node_level ==
 746                                        DISK_LEAF_NODE_LEVEL) ? IH_SIZE :
 747                                       KEY_SIZE,
 748                                       &last_element->pe_position);
 749                 if (node_level == stop_level) {
 750                         return retval;
 751                 }
 752
 753                 /* we are not in the stop level */
 754                 /*
 755                  * item has been found, so we choose the pointer which
 756                  * is to the right of the found one
 757                  */
 758                 if (retval == ITEM_FOUND)
 759                         last_element->pe_position++;
 760
 761                 /*
 762                  * if item was not found we choose the position which is to
 763                  * the left of the found item. This requires no code,
 764                  * bin_search did it already.
 765                  */
 766
 767                 /*
 768                  * So we have chosen a position in the current node which is
 769                  * an internal node.  Now we calculate child block number by
 770                  * position in the node.
 771                  */
 772                 block_number =
 773                     B_N_CHILD_NUM(bh, last_element->pe_position);
 774
 775                 /*
 776                  * if we are going to read leaf nodes, try for read
 777                  * ahead as well
 778                  */
 779                 if ((search_path->reada & PATH_READA) &&
 780                     node_level == DISK_LEAF_NODE_LEVEL + 1) {
 781                         int pos = last_element->pe_position;
 782                         int limit = B_NR_ITEMS(bh);
 783                         struct reiserfs_key *le_key;
 784
 785                         if (search_path->reada & PATH_READA_BACK)
 786                                 limit = 0;
 787                         while (reada_count < SEARCH_BY_KEY_READA) {
 788                                 if (pos == limit)
 789                                         break;
 790                                 reada_blocks[reada_count++] =
 791                                     B_N_CHILD_NUM(bh, pos);
 792                                 if (search_path->reada & PATH_READA_BACK)
 793                                         pos--;
 794                                 else
 795                                         pos++;
 796
 797                                 /*
 798                                  * check to make sure we're in the same object
 799                                  */
 800                                 le_key = internal_key(bh, pos);
 801                                 if (le32_to_cpu(le_key->k_objectid) !=
 802                                     key->on_disk_key.k_objectid) {
 803                                         break;
 804                                 }
 805                         }
 806                 }
 807         }
 808 }
 809
 810 /*
 811  * Form the path to an item and position in this item which contains
 812  * file byte defined by key. If there is no such item
 813  * corresponding to the key, we point the path to the item with
 814  * maximal key less than key, and *pos_in_item is set to one
 815  * past the last entry/byte in the item.  If searching for entry in a
 816  * directory item, and it is not found, *pos_in_item is set to one
 817  * entry more than the entry with maximal key which is less than the
 818  * sought key.
 819  *
 820  * Note that if there is no entry in this same node which is one more,
 821  * then we point to an imaginary entry.  for direct items, the
 822  * position is in units of bytes, for indirect items the position is
 823  * in units of blocknr entries, for directory items the position is in
 824  * units of directory entries.
 825  */
 826 /* The function is NOT SCHEDULE-SAFE! */
 827 int search_for_position_by_key(struct super_block *sb,
 828                                /* Key to search (cpu variable) */
 829                                const struct cpu_key *p_cpu_key,
 830                                /* Filled up by this function. */
 831                                struct treepath *search_path)
 832 {
 833         struct item_head *p_le_ih;      /* pointer to on-disk structure */
 834         int blk_size;
 835         loff_t item_offset, offset;
 836         struct reiserfs_dir_entry de;
 837         int retval;
 838
 839         /* If searching for directory entry. */
 840         if (is_direntry_cpu_key(p_cpu_key))
 841                 return search_by_entry_key(sb, p_cpu_key, search_path,
 842                                            &de);
 843
 844         /* If not searching for directory entry. */
 845
 846         /* If item is found. */
 847         retval = search_item(sb, p_cpu_key, search_path);
 848         if (retval == IO_ERROR)
 849                 return retval;
 850         if (retval == ITEM_FOUND) {
 851
 852                 RFALSE(!ih_item_len
 853                        (item_head
 854                         (PATH_PLAST_BUFFER(search_path),
 855                          PATH_LAST_POSITION(search_path))),
 856                        "PAP-5165: item length equals zero");
 857
 858                 pos_in_item(search_path) = 0;
 859                 return POSITION_FOUND;
 860         }
 861
 862         RFALSE(!PATH_LAST_POSITION(search_path),
 863                "PAP-5170: position equals zero");
 864
 865         /* Item is not found. Set path to the previous item. */
 866         p_le_ih =
 867             item_head(PATH_PLAST_BUFFER(search_path),
 868                            --PATH_LAST_POSITION(search_path));
 869         blk_size = sb->s_blocksize;
 870
 871         if (comp_short_keys(&p_le_ih->ih_key, p_cpu_key))
 872                 return FILE_NOT_FOUND;
 873
 874         /* FIXME: quite ugly this far */
 875
 876         item_offset = le_ih_k_offset(p_le_ih);
 877         offset = cpu_key_k_offset(p_cpu_key);
 878
 879         /* Needed byte is contained in the item pointed to by the path. */
 880         if (item_offset <= offset &&
 881             item_offset + op_bytes_number(p_le_ih, blk_size) > offset) {
 882                 pos_in_item(search_path) = offset - item_offset;
 883                 if (is_indirect_le_ih(p_le_ih)) {
 884                         pos_in_item(search_path) /= blk_size;
 885                 }
 886                 return POSITION_FOUND;
 887         }
 888
 889         /*
 890          * Needed byte is not contained in the item pointed to by the
 891          * path. Set pos_in_item out of the item.
 892          */
 893         if (is_indirect_le_ih(p_le_ih))
 894                 pos_in_item(search_path) =
 895                     ih_item_len(p_le_ih) / UNFM_P_SIZE;
 896         else
 897                 pos_in_item(search_path) = ih_item_len(p_le_ih);
 898
 899         return POSITION_NOT_FOUND;
 900 }
 901
 902 /* Compare given item and item pointed to by the path. */
 903 int comp_items(const struct item_head *stored_ih, const struct treepath *path)
 904 {
 905         struct buffer_head *bh = PATH_PLAST_BUFFER(path);
 906         struct item_head *ih;
 907
 908         /* Last buffer at the path is not in the tree. */
 909         if (!B_IS_IN_TREE(bh))
 910                 return 1;
 911
 912         /* Last path position is invalid. */
 913         if (PATH_LAST_POSITION(path) >= B_NR_ITEMS(bh))
 914                 return 1;
 915
 916         /* we need only to know, whether it is the same item */
 917         ih = tp_item_head(path);
 918         return memcmp(stored_ih, ih, IH_SIZE);
 919 }
 920
 921 /* prepare for delete or cut of direct item */
 922 static inline int prepare_for_direct_item(struct treepath *path,
 923                                           struct item_head *le_ih,
 924                                           struct inode *inode,
 925                                           loff_t new_file_length, int *cut_size)
 926 {
 927         loff_t round_len;
 928
 929         if (new_file_length == max_reiserfs_offset(inode)) {
 930                 /* item has to be deleted */
 931                 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
 932                 return M_DELETE;
 933         }
 934         /* new file gets truncated */
 935         if (get_inode_item_key_version(inode) == KEY_FORMAT_3_6) {
 936                 round_len = ROUND_UP(new_file_length);
 937                 /* this was new_file_length < le_ih ... */
 938                 if (round_len < le_ih_k_offset(le_ih)) {
 939                         *cut_size = -(IH_SIZE + ih_item_len(le_ih));
 940                         return M_DELETE;        /* Delete this item. */
 941                 }
 942                 /* Calculate first position and size for cutting from item. */
 943                 pos_in_item(path) = round_len - (le_ih_k_offset(le_ih) - 1);
 944                 *cut_size = -(ih_item_len(le_ih) - pos_in_item(path));
 945
 946                 return M_CUT;   /* Cut from this item. */
 947         }
 948
 949         /* old file: items may have any length */
 950
 951         if (new_file_length < le_ih_k_offset(le_ih)) {
 952                 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
 953                 return M_DELETE;        /* Delete this item. */
 954         }
 955
 956         /* Calculate first position and size for cutting from item. */
 957         *cut_size = -(ih_item_len(le_ih) -
 958                       (pos_in_item(path) =
 959                        new_file_length + 1 - le_ih_k_offset(le_ih)));
 960         return M_CUT;           /* Cut from this item. */
 961 }
 962
 963 static inline int prepare_for_direntry_item(struct treepath *path,
 964                                             struct item_head *le_ih,
 965                                             struct inode *inode,
 966                                             loff_t new_file_length,
 967                                             int *cut_size)
 968 {
 969         if (le_ih_k_offset(le_ih) == DOT_OFFSET &&
 970             new_file_length == max_reiserfs_offset(inode)) {
 971                 RFALSE(ih_entry_count(le_ih) != 2,
 972                        "PAP-5220: incorrect empty directory item (%h)", le_ih);
 973                 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
 974                 /* Delete the directory item containing "." and ".." entry. */
 975                 return M_DELETE;
 976         }
 977
 978         if (ih_entry_count(le_ih) == 1) {
 979                 /*
 980                  * Delete the directory item such as there is one record only
 981                  * in this item
 982                  */
 983                 *cut_size = -(IH_SIZE + ih_item_len(le_ih));
 984                 return M_DELETE;
 985         }
 986
 987         /* Cut one record from the directory item. */
 988         *cut_size =
 989             -(DEH_SIZE +
 990               entry_length(get_last_bh(path), le_ih, pos_in_item(path)));
 991         return M_CUT;
 992 }
 993
 994 #define JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD (2 * JOURNAL_PER_BALANCE_CNT + 1)
 995
 996 /*
 997  * If the path points to a directory or direct item, calculate mode
 998  * and the size cut, for balance.
 999  * If the path points to an indirect item, remove some number of its
1000  * unformatted nodes.
1001  * In case of file truncate calculate whether this item must be
1002  * deleted/truncated or last unformatted node of this item will be
1003  * converted to a direct item.
1004  * This function returns a determination of what balance mode the
1005  * calling function should employ.
1006  */
1007 static char prepare_for_delete_or_cut(struct reiserfs_transaction_handle *th,
1008                                       struct inode *inode,
1009                                       struct treepath *path,
1010                                       const struct cpu_key *item_key,
1011                                       /*
1012                                        * Number of unformatted nodes
1013                                        * which were removed from end
1014                                        * of the file.
1015                                        */
1016                                       int *removed,
1017                                       int *cut_size,
1018                                       /* MAX_KEY_OFFSET in case of delete. */
1019                                       unsigned long long new_file_length
1020     )
1021 {
1022         struct super_block *sb = inode->i_sb;
1023         struct item_head *p_le_ih = tp_item_head(path);
1024         struct buffer_head *bh = PATH_PLAST_BUFFER(path);
1025
1026         BUG_ON(!th->t_trans_id);
1027
1028         /* Stat_data item. */
1029         if (is_statdata_le_ih(p_le_ih)) {
1030
1031                 RFALSE(new_file_length != max_reiserfs_offset(inode),
1032                        "PAP-5210: mode must be M_DELETE");
1033
1034                 *cut_size = -(IH_SIZE + ih_item_len(p_le_ih));
1035                 return M_DELETE;
1036         }
1037
1038         /* Directory item. */
1039         if (is_direntry_le_ih(p_le_ih))
1040                 return prepare_for_direntry_item(path, p_le_ih, inode,
1041                                                  new_file_length,
1042                                                  cut_size);
1043
1044         /* Direct item. */
1045         if (is_direct_le_ih(p_le_ih))
1046                 return prepare_for_direct_item(path, p_le_ih, inode,
1047                                                new_file_length, cut_size);
1048
1049         /* Case of an indirect item. */
1050         {
1051             int blk_size = sb->s_blocksize;
1052             struct item_head s_ih;
1053             int need_re_search;
1054             int delete = 0;
1055             int result = M_CUT;
1056             int pos = 0;
1057
1058             if ( new_file_length == max_reiserfs_offset (inode) ) {
1059                 /*
1060                  * prepare_for_delete_or_cut() is called by
1061                  * reiserfs_delete_item()
1062                  */
1063                 new_file_length = 0;
1064                 delete = 1;
1065             }
1066
1067             do {
1068                 need_re_search = 0;
1069                 *cut_size = 0;
1070                 bh = PATH_PLAST_BUFFER(path);
1071                 copy_item_head(&s_ih, tp_item_head(path));
1072                 pos = I_UNFM_NUM(&s_ih);
1073
1074                 while (le_ih_k_offset (&s_ih) + (pos - 1) * blk_size > new_file_length) {
1075                     __le32 *unfm;
1076                     __u32 block;
1077
1078                     /*
1079                      * Each unformatted block deletion may involve
1080                      * one additional bitmap block into the transaction,
1081                      * thereby the initial journal space reservation
1082                      * might not be enough.
1083                      */
1084                     if (!delete && (*cut_size) != 0 &&
1085                         reiserfs_transaction_free_space(th) < JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD)
1086                         break;
1087
1088                     unfm = (__le32 *)ih_item_body(bh, &s_ih) + pos - 1;
1089                     block = get_block_num(unfm, 0);
1090
1091                     if (block != 0) {
1092                         reiserfs_prepare_for_journal(sb, bh, 1);
1093                         put_block_num(unfm, 0, 0);
1094                         journal_mark_dirty(th, bh);
1095                         reiserfs_free_block(th, inode, block, 1);
1096                     }
1097
1098                     reiserfs_cond_resched(sb);
1099
1100                     if (item_moved (&s_ih, path))  {
1101                         need_re_search = 1;
1102                         break;
1103                     }
1104
1105                     pos --;
1106                     (*removed)++;
1107                     (*cut_size) -= UNFM_P_SIZE;
1108
1109                     if (pos == 0) {
1110                         (*cut_size) -= IH_SIZE;
1111                         result = M_DELETE;
1112                         break;
1113                     }
1114                 }
1115                 /*
1116                  * a trick.  If the buffer has been logged, this will
1117                  * do nothing.  If we've broken the loop without logging
1118                  * it, it will restore the buffer
1119                  */
1120                 reiserfs_restore_prepared_buffer(sb, bh);
1121             } while (need_re_search &&
1122                      search_for_position_by_key(sb, item_key, path) == POSITION_FOUND);
1123             pos_in_item(path) = pos * UNFM_P_SIZE;
1124
1125             if (*cut_size == 0) {
1126                 /*
1127                  * Nothing was cut. maybe convert last unformatted node to the
1128                  * direct item?
1129                  */
1130                 result = M_CONVERT;
1131             }
1132             return result;
1133         }
1134 }
1135
1136 /* Calculate number of bytes which will be deleted or cut during balance */
1137 static int calc_deleted_bytes_number(struct tree_balance *tb, char mode)
1138 {
1139         int del_size;
1140         struct item_head *p_le_ih = tp_item_head(tb->tb_path);
1141
1142         if (is_statdata_le_ih(p_le_ih))
1143                 return 0;
1144
1145         del_size =
1146             (mode ==
1147              M_DELETE) ? ih_item_len(p_le_ih) : -tb->insert_size[0];
1148         if (is_direntry_le_ih(p_le_ih)) {
1149                 /*
1150                  * return EMPTY_DIR_SIZE; We delete emty directories only.
1151                  * we can't use EMPTY_DIR_SIZE, as old format dirs have a
1152                  * different empty size.  ick. FIXME, is this right?
1153                  */
1154                 return del_size;
1155         }
1156
1157         if (is_indirect_le_ih(p_le_ih))
1158                 del_size = (del_size / UNFM_P_SIZE) *
1159                                 (PATH_PLAST_BUFFER(tb->tb_path)->b_size);
1160         return del_size;
1161 }
1162
1163 static void init_tb_struct(struct reiserfs_transaction_handle *th,
1164                            struct tree_balance *tb,
1165                            struct super_block *sb,
1166                            struct treepath *path, int size)
1167 {
1168
1169         BUG_ON(!th->t_trans_id);
1170
1171         memset(tb, '\0', sizeof(struct tree_balance));
1172         tb->transaction_handle = th;
1173         tb->tb_sb = sb;
1174         tb->tb_path = path;
1175         PATH_OFFSET_PBUFFER(path, ILLEGAL_PATH_ELEMENT_OFFSET) = NULL;
1176         PATH_OFFSET_POSITION(path, ILLEGAL_PATH_ELEMENT_OFFSET) = 0;
1177         tb->insert_size[0] = size;
1178 }
1179
1180 void padd_item(char *item, int total_length, int length)
1181 {
1182         int i;
1183
1184         for (i = total_length; i > length;)
1185                 item[--i] = 0;
1186 }
1187
1188 #ifdef REISERQUOTA_DEBUG
1189 char key2type(struct reiserfs_key *ih)
1190 {
1191         if (is_direntry_le_key(2, ih))
1192                 return 'd';
1193         if (is_direct_le_key(2, ih))
1194                 return 'D';
1195         if (is_indirect_le_key(2, ih))
1196                 return 'i';
1197         if (is_statdata_le_key(2, ih))
1198                 return 's';
1199         return 'u';
1200 }
1201
1202 char head2type(struct item_head *ih)
1203 {
1204         if (is_direntry_le_ih(ih))
1205                 return 'd';
1206         if (is_direct_le_ih(ih))
1207                 return 'D';
1208         if (is_indirect_le_ih(ih))
1209                 return 'i';
1210         if (is_statdata_le_ih(ih))
1211                 return 's';
1212         return 'u';
1213 }
1214 #endif
1215
1216 /*
1217  * Delete object item.
1218  * th       - active transaction handle
1219  * path     - path to the deleted item
1220  * item_key - key to search for the deleted item
1221  * indode   - used for updating i_blocks and quotas
1222  * un_bh    - NULL or unformatted node pointer
1223  */
1224 int reiserfs_delete_item(struct reiserfs_transaction_handle *th,
1225                          struct treepath *path, const struct cpu_key *item_key,
1226                          struct inode *inode, struct buffer_head *un_bh)
1227 {
1228         struct super_block *sb = inode->i_sb;
1229         struct tree_balance s_del_balance;
1230         struct item_head s_ih;
1231         struct item_head *q_ih;
1232         int quota_cut_bytes;
1233         int ret_value, del_size, removed;
1234         int depth;
1235
1236 #ifdef CONFIG_REISERFS_CHECK
1237         char mode;
1238         int iter = 0;
1239 #endif
1240
1241         BUG_ON(!th->t_trans_id);
1242
1243         init_tb_struct(th, &s_del_balance, sb, path,
1244                        0 /*size is unknown */ );
1245
1246         while (1) {
1247                 removed = 0;
1248
1249 #ifdef CONFIG_REISERFS_CHECK
1250                 iter++;
1251                 mode =
1252 #endif
1253                     prepare_for_delete_or_cut(th, inode, path,
1254                                               item_key, &removed,
1255                                               &del_size,
1256                                               max_reiserfs_offset(inode));
1257
1258                 RFALSE(mode != M_DELETE, "PAP-5320: mode must be M_DELETE");
1259
1260                 copy_item_head(&s_ih, tp_item_head(path));
1261                 s_del_balance.insert_size[0] = del_size;
1262
1263                 ret_value = fix_nodes(M_DELETE, &s_del_balance, NULL, NULL);
1264                 if (ret_value != REPEAT_SEARCH)
1265                         break;
1266
1267                 PROC_INFO_INC(sb, delete_item_restarted);
1268
1269                 /* file system changed, repeat search */
1270                 ret_value =
1271                     search_for_position_by_key(sb, item_key, path);
1272                 if (ret_value == IO_ERROR)
1273                         break;
1274                 if (ret_value == FILE_NOT_FOUND) {
1275                         reiserfs_warning(sb, "vs-5340",
1276                                          "no items of the file %K found",
1277                                          item_key);
1278                         break;
1279                 }
1280         }                       /* while (1) */
1281
1282         if (ret_value != CARRY_ON) {
1283                 unfix_nodes(&s_del_balance);
1284                 return 0;
1285         }
1286
1287         /* reiserfs_delete_item returns item length when success */
1288         ret_value = calc_deleted_bytes_number(&s_del_balance, M_DELETE);
1289         q_ih = tp_item_head(path);
1290         quota_cut_bytes = ih_item_len(q_ih);
1291
1292         /*
1293          * hack so the quota code doesn't have to guess if the file has a
1294          * tail.  On tail insert, we allocate quota for 1 unformatted node.
1295          * We test the offset because the tail might have been
1296          * split into multiple items, and we only want to decrement for
1297          * the unfm node once
1298          */
1299         if (!S_ISLNK(inode->i_mode) && is_direct_le_ih(q_ih)) {
1300                 if ((le_ih_k_offset(q_ih) & (sb->s_blocksize - 1)) == 1) {
1301                         quota_cut_bytes = sb->s_blocksize + UNFM_P_SIZE;
1302                 } else {
1303                         quota_cut_bytes = 0;
1304                 }
1305         }
1306
1307         if (un_bh) {
1308                 int off;
1309                 char *data;
1310
1311                 /*
1312                  * We are in direct2indirect conversion, so move tail contents
1313                  * to the unformatted node
1314                  */
1315                 /*
1316                  * note, we do the copy before preparing the buffer because we
1317                  * don't care about the contents of the unformatted node yet.
1318                  * the only thing we really care about is the direct item's
1319                  * data is in the unformatted node.
1320                  *
1321                  * Otherwise, we would have to call
1322                  * reiserfs_prepare_for_journal on the unformatted node,
1323                  * which might schedule, meaning we'd have to loop all the
1324                  * way back up to the start of the while loop.
1325                  *
1326                  * The unformatted node must be dirtied later on.  We can't be
1327                  * sure here if the entire tail has been deleted yet.
1328                  *
1329                  * un_bh is from the page cache (all unformatted nodes are
1330                  * from the page cache) and might be a highmem page.  So, we
1331                  * can't use un_bh->b_data.
1332                  * -clm
1333                  */
1334
1335                 data = kmap_atomic(un_bh->b_page);
1336                 off = ((le_ih_k_offset(&s_ih) - 1) & (PAGE_SIZE - 1));
1337                 memcpy(data + off,
1338                        ih_item_body(PATH_PLAST_BUFFER(path), &s_ih),
1339                        ret_value);
1340                 kunmap_atomic(data);
1341         }
1342
1343         /* Perform balancing after all resources have been collected at once. */
1344         do_balance(&s_del_balance, NULL, NULL, M_DELETE);
1345
1346 #ifdef REISERQUOTA_DEBUG
1347         reiserfs_debug(sb, REISERFS_DEBUG_CODE,
1348                        "reiserquota delete_item(): freeing %u, id=%u type=%c",
1349                        quota_cut_bytes, inode->i_uid, head2type(&s_ih));
1350 #endif
1351         depth = reiserfs_write_unlock_nested(inode->i_sb);
1352         dquot_free_space_nodirty(inode, quota_cut_bytes);
1353         reiserfs_write_lock_nested(inode->i_sb, depth);
1354
1355         /* Return deleted body length */
1356         return ret_value;
1357 }
1358
1359 /*
1360  * Summary Of Mechanisms For Handling Collisions Between Processes:
1361  *
1362  *  deletion of the body of the object is performed by iput(), with the
1363  *  result that if multiple processes are operating on a file, the
1364  *  deletion of the body of the file is deferred until the last process
1365  *  that has an open inode performs its iput().
1366  *
1367  *  writes and truncates are protected from collisions by use of
1368  *  semaphores.
1369  *
1370  *  creates, linking, and mknod are protected from collisions with other
1371  *  processes by making the reiserfs_add_entry() the last step in the
1372  *  creation, and then rolling back all changes if there was a collision.
1373  *  - Hans
1374 */
1375
1376 /* this deletes item which never gets split */
1377 void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,
1378                                 struct inode *inode, struct reiserfs_key *key)
1379 {
1380         struct super_block *sb = th->t_super;
1381         struct tree_balance tb;
1382         INITIALIZE_PATH(path);
1383         int item_len = 0;
1384         int tb_init = 0;
1385         struct cpu_key cpu_key;
1386         int retval;
1387         int quota_cut_bytes = 0;
1388
1389         BUG_ON(!th->t_trans_id);
1390
1391         le_key2cpu_key(&cpu_key, key);
1392
1393         while (1) {
1394                 retval = search_item(th->t_super, &cpu_key, &path);
1395                 if (retval == IO_ERROR) {
1396                         reiserfs_error(th->t_super, "vs-5350",
1397                                        "i/o failure occurred trying "
1398                                        "to delete %K", &cpu_key);
1399                         break;
1400                 }
1401                 if (retval != ITEM_FOUND) {
1402                         pathrelse(&path);
1403                         /*
1404                          * No need for a warning, if there is just no free
1405                          * space to insert '..' item into the
1406                          * newly-created subdir
1407                          */
1408                         if (!
1409                             ((unsigned long long)
1410                              GET_HASH_VALUE(le_key_k_offset
1411                                             (le_key_version(key), key)) == 0
1412                              && (unsigned long long)
1413                              GET_GENERATION_NUMBER(le_key_k_offset
1414                                                    (le_key_version(key),
1415                                                     key)) == 1))
1416                                 reiserfs_warning(th->t_super, "vs-5355",
1417                                                  "%k not found", key);
1418                         break;
1419                 }
1420                 if (!tb_init) {
1421                         tb_init = 1;
1422                         item_len = ih_item_len(tp_item_head(&path));
1423                         init_tb_struct(th, &tb, th->t_super, &path,
1424                                        -(IH_SIZE + item_len));
1425                 }
1426                 quota_cut_bytes = ih_item_len(tp_item_head(&path));
1427
1428                 retval = fix_nodes(M_DELETE, &tb, NULL, NULL);
1429                 if (retval == REPEAT_SEARCH) {
1430                         PROC_INFO_INC(th->t_super, delete_solid_item_restarted);
1431                         continue;
1432                 }
1433
1434                 if (retval == CARRY_ON) {
1435                         do_balance(&tb, NULL, NULL, M_DELETE);
1436                         /*
1437                          * Should we count quota for item? (we don't
1438                          * count quotas for save-links)
1439                          */
1440                         if (inode) {
1441                                 int depth;
1442 #ifdef REISERQUOTA_DEBUG
1443                                 reiserfs_debug(th->t_super, REISERFS_DEBUG_CODE,
1444                                                "reiserquota delete_solid_item(): freeing %u id=%u type=%c",
1445                                                quota_cut_bytes, inode->i_uid,
1446                                                key2type(key));
1447 #endif
1448                                 depth = reiserfs_write_unlock_nested(sb);
1449                                 dquot_free_space_nodirty(inode,
1450                                                          quota_cut_bytes);
1451                                 reiserfs_write_lock_nested(sb, depth);
1452                         }
1453                         break;
1454                 }
1455
1456                 /* IO_ERROR, NO_DISK_SPACE, etc */
1457                 reiserfs_warning(th->t_super, "vs-5360",
1458                                  "could not delete %K due to fix_nodes failure",
1459                                  &cpu_key);
1460                 unfix_nodes(&tb);
1461                 break;
1462         }
1463
1464         reiserfs_check_path(&path);
1465 }
1466
1467 int reiserfs_delete_object(struct reiserfs_transaction_handle *th,
1468                            struct inode *inode)
1469 {
1470         int err;
1471         inode->i_size = 0;
1472         BUG_ON(!th->t_trans_id);
1473
1474         /* for directory this deletes item containing "." and ".." */
1475         err =
1476             reiserfs_do_truncate(th, inode, NULL, 0 /*no timestamp updates */ );
1477         if (err)
1478                 return err;
1479
1480 #if defined( USE_INODE_GENERATION_COUNTER )
1481         if (!old_format_only(th->t_super)) {
1482                 __le32 *inode_generation;
1483
1484                 inode_generation =
1485                     &REISERFS_SB(th->t_super)->s_rs->s_inode_generation;
1486                 le32_add_cpu(inode_generation, 1);
1487         }
1488 /* USE_INODE_GENERATION_COUNTER */
1489 #endif
1490         reiserfs_delete_solid_item(th, inode, INODE_PKEY(inode));
1491
1492         return err;
1493 }
1494
1495 static void unmap_buffers(struct page *page, loff_t pos)
1496 {
1497         struct buffer_head *bh;
1498         struct buffer_head *head;
1499         struct buffer_head *next;
1500         unsigned long tail_index;
1501         unsigned long cur_index;
1502
1503         if (page) {
1504                 if (page_has_buffers(page)) {
1505                         tail_index = pos & (PAGE_SIZE - 1);
1506                         cur_index = 0;
1507                         head = page_buffers(page);
1508                         bh = head;
1509                         do {
1510                                 next = bh->b_this_page;
1511
1512                                 /*
1513                                  * we want to unmap the buffers that contain
1514                                  * the tail, and all the buffers after it
1515                                  * (since the tail must be at the end of the
1516                                  * file).  We don't want to unmap file data
1517                                  * before the tail, since it might be dirty
1518                                  * and waiting to reach disk
1519                                  */
1520                                 cur_index += bh->b_size;
1521                                 if (cur_index > tail_index) {
1522                                         reiserfs_unmap_buffer(bh);
1523                                 }
1524                                 bh = next;
1525                         } while (bh != head);
1526                 }
1527         }
1528 }
1529
1530 static int maybe_indirect_to_direct(struct reiserfs_transaction_handle *th,
1531                                     struct inode *inode,
1532                                     struct page *page,
1533                                     struct treepath *path,
1534                                     const struct cpu_key *item_key,
1535                                     loff_t new_file_size, char *mode)
1536 {
1537         struct super_block *sb = inode->i_sb;
1538         int block_size = sb->s_blocksize;
1539         int cut_bytes;
1540         BUG_ON(!th->t_trans_id);
1541         BUG_ON(new_file_size != inode->i_size);
1542
1543         /*
1544          * the page being sent in could be NULL if there was an i/o error
1545          * reading in the last block.  The user will hit problems trying to
1546          * read the file, but for now we just skip the indirect2direct
1547          */
1548         if (atomic_read(&inode->i_count) > 1 ||
1549             !tail_has_to_be_packed(inode) ||
1550             !page || (REISERFS_I(inode)->i_flags & i_nopack_mask)) {
1551                 /* leave tail in an unformatted node */
1552                 *mode = M_SKIP_BALANCING;
1553                 cut_bytes =
1554                     block_size - (new_file_size & (block_size - 1));
1555                 pathrelse(path);
1556                 return cut_bytes;
1557         }
1558
1559         /* Perform the conversion to a direct_item. */
1560         return indirect2direct(th, inode, page, path, item_key,
1561                                new_file_size, mode);
1562 }
1563
1564 /*
1565  * we did indirect_to_direct conversion. And we have inserted direct
1566  * item successesfully, but there were no disk space to cut unfm
1567  * pointer being converted. Therefore we have to delete inserted
1568  * direct item(s)
1569  */
1570 static void indirect_to_direct_roll_back(struct reiserfs_transaction_handle *th,
1571                                          struct inode *inode, struct treepath *path)
1572 {
1573         struct cpu_key tail_key;
1574         int tail_len;
1575         int removed;
1576         BUG_ON(!th->t_trans_id);
1577
1578         make_cpu_key(&tail_key, inode, inode->i_size + 1, TYPE_DIRECT, 4);
1579         tail_key.key_length = 4;
1580
1581         tail_len =
1582             (cpu_key_k_offset(&tail_key) & (inode->i_sb->s_blocksize - 1)) - 1;
1583         while (tail_len) {
1584                 /* look for the last byte of the tail */
1585                 if (search_for_position_by_key(inode->i_sb, &tail_key, path) ==
1586                     POSITION_NOT_FOUND)
1587                         reiserfs_panic(inode->i_sb, "vs-5615",
1588                                        "found invalid item");
1589                 RFALSE(path->pos_in_item !=
1590                        ih_item_len(tp_item_head(path)) - 1,
1591                        "vs-5616: appended bytes found");
1592                 PATH_LAST_POSITION(path)--;
1593
1594                 removed =
1595                     reiserfs_delete_item(th, path, &tail_key, inode,
1596                                          NULL /*unbh not needed */ );
1597                 RFALSE(removed <= 0
1598                        || removed > tail_len,
1599                        "vs-5617: there was tail %d bytes, removed item length %d bytes",
1600                        tail_len, removed);
1601                 tail_len -= removed;
1602                 set_cpu_key_k_offset(&tail_key,
1603                                      cpu_key_k_offset(&tail_key) - removed);
1604         }
1605         reiserfs_warning(inode->i_sb, "reiserfs-5091", "indirect_to_direct "
1606                          "conversion has been rolled back due to "
1607                          "lack of disk space");
1608         mark_inode_dirty(inode);
1609 }
1610
1611 /* (Truncate or cut entry) or delete object item. Returns < 0 on failure */
1612 int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,
1613                            struct treepath *path,
1614                            struct cpu_key *item_key,
1615                            struct inode *inode,
1616                            struct page *page, loff_t new_file_size)
1617 {
1618         struct super_block *sb = inode->i_sb;
1619         /*
1620          * Every function which is going to call do_balance must first
1621          * create a tree_balance structure.  Then it must fill up this
1622          * structure by using the init_tb_struct and fix_nodes functions.
1623          * After that we can make tree balancing.
1624          */
1625         struct tree_balance s_cut_balance;
1626         struct item_head *p_le_ih;
1627         int cut_size = 0;       /* Amount to be cut. */
1628         int ret_value = CARRY_ON;
1629         int removed = 0;        /* Number of the removed unformatted nodes. */
1630         int is_inode_locked = 0;
1631         char mode;              /* Mode of the balance. */
1632         int retval2 = -1;
1633         int quota_cut_bytes;
1634         loff_t tail_pos = 0;
1635         int depth;
1636
1637         BUG_ON(!th->t_trans_id);
1638
1639         init_tb_struct(th, &s_cut_balance, inode->i_sb, path,
1640                        cut_size);
1641
1642         /*
1643          * Repeat this loop until we either cut the item without needing
1644          * to balance, or we fix_nodes without schedule occurring
1645          */
1646         while (1) {
1647                 /*
1648                  * Determine the balance mode, position of the first byte to
1649                  * be cut, and size to be cut.  In case of the indirect item
1650                  * free unformatted nodes which are pointed to by the cut
1651                  * pointers.
1652                  */
1653
1654                 mode =
1655                     prepare_for_delete_or_cut(th, inode, path,
1656                                               item_key, &removed,
1657                                               &cut_size, new_file_size);
1658                 if (mode == M_CONVERT) {
1659                         /*
1660                          * convert last unformatted node to direct item or
1661                          * leave tail in the unformatted node
1662                          */
1663                         RFALSE(ret_value != CARRY_ON,
1664                                "PAP-5570: can not convert twice");
1665
1666                         ret_value =
1667                             maybe_indirect_to_direct(th, inode, page,
1668                                                      path, item_key,
1669                                                      new_file_size, &mode);
1670                         if (mode == M_SKIP_BALANCING)
1671                                 /* tail has been left in the unformatted node */
1672                                 return ret_value;
1673
1674                         is_inode_locked = 1;
1675
1676                         /*
1677                          * removing of last unformatted node will
1678                          * change value we have to return to truncate.
1679                          * Save it
1680                          */
1681                         retval2 = ret_value;
1682
1683                         /*
1684                          * So, we have performed the first part of the
1685                          * conversion:
1686                          * inserting the new direct item.  Now we are
1687                          * removing the last unformatted node pointer.
1688                          * Set key to search for it.
1689                          */
1690                         set_cpu_key_k_type(item_key, TYPE_INDIRECT);
1691                         item_key->key_length = 4;
1692                         new_file_size -=
1693                             (new_file_size & (sb->s_blocksize - 1));
1694                         tail_pos = new_file_size;
1695                         set_cpu_key_k_offset(item_key, new_file_size + 1);
1696                         if (search_for_position_by_key
1697                             (sb, item_key,
1698                              path) == POSITION_NOT_FOUND) {
1699                                 print_block(PATH_PLAST_BUFFER(path), 3,
1700                                             PATH_LAST_POSITION(path) - 1,
1701                                             PATH_LAST_POSITION(path) + 1);
1702                                 reiserfs_panic(sb, "PAP-5580", "item to "
1703                                                "convert does not exist (%K)",
1704                                                item_key);
1705                         }
1706                         continue;
1707                 }
1708                 if (cut_size == 0) {
1709                         pathrelse(path);
1710                         return 0;
1711                 }
1712
1713                 s_cut_balance.insert_size[0] = cut_size;
1714
1715                 ret_value = fix_nodes(mode, &s_cut_balance, NULL, NULL);
1716                 if (ret_value != REPEAT_SEARCH)
1717                         break;
1718
1719                 PROC_INFO_INC(sb, cut_from_item_restarted);
1720
1721                 ret_value =
1722                     search_for_position_by_key(sb, item_key, path);
1723                 if (ret_value == POSITION_FOUND)
1724                         continue;
1725
1726                 reiserfs_warning(sb, "PAP-5610", "item %K not found",
1727                                  item_key);
1728                 unfix_nodes(&s_cut_balance);
1729                 return (ret_value == IO_ERROR) ? -EIO : -ENOENT;
1730         }                       /* while */
1731
1732         /* check fix_nodes results (IO_ERROR or NO_DISK_SPACE) */
1733         if (ret_value != CARRY_ON) {
1734                 if (is_inode_locked) {
1735                         /*
1736                          * FIXME: this seems to be not needed: we are always
1737                          * able to cut item
1738                          */
1739                         indirect_to_direct_roll_back(th, inode, path);
1740                 }
1741                 if (ret_value == NO_DISK_SPACE)
1742                         reiserfs_warning(sb, "reiserfs-5092",
1743                                          "NO_DISK_SPACE");
1744                 unfix_nodes(&s_cut_balance);
1745                 return -EIO;
1746         }
1747
1748         /* go ahead and perform balancing */
1749
1750         RFALSE(mode == M_PASTE || mode == M_INSERT, "invalid mode");
1751
1752         /* Calculate number of bytes that need to be cut from the item. */
1753         quota_cut_bytes =
1754             (mode ==
1755              M_DELETE) ? ih_item_len(tp_item_head(path)) : -s_cut_balance.
1756             insert_size[0];
1757         if (retval2 == -1)
1758                 ret_value = calc_deleted_bytes_number(&s_cut_balance, mode);
1759         else
1760                 ret_value = retval2;
1761
1762         /*
1763          * For direct items, we only change the quota when deleting the last
1764          * item.
1765          */
1766         p_le_ih = tp_item_head(s_cut_balance.tb_path);
1767         if (!S_ISLNK(inode->i_mode) && is_direct_le_ih(p_le_ih)) {
1768                 if (mode == M_DELETE &&
1769                     (le_ih_k_offset(p_le_ih) & (sb->s_blocksize - 1)) ==
1770                     1) {
1771                         /* FIXME: this is to keep 3.5 happy */
1772                         REISERFS_I(inode)->i_first_direct_byte = U32_MAX;
1773                         quota_cut_bytes = sb->s_blocksize + UNFM_P_SIZE;
1774                 } else {
1775                         quota_cut_bytes = 0;
1776                 }
1777         }
1778 #ifdef CONFIG_REISERFS_CHECK
1779         if (is_inode_locked) {
1780                 struct item_head *le_ih =
1781                     tp_item_head(s_cut_balance.tb_path);
1782                 /*
1783                  * we are going to complete indirect2direct conversion. Make
1784                  * sure, that we exactly remove last unformatted node pointer
1785                  * of the item
1786                  */
1787                 if (!is_indirect_le_ih(le_ih))
1788                         reiserfs_panic(sb, "vs-5652",
1789                                        "item must be indirect %h", le_ih);
1790
1791                 if (mode == M_DELETE && ih_item_len(le_ih) != UNFM_P_SIZE)
1792                         reiserfs_panic(sb, "vs-5653", "completing "
1793                                        "indirect2direct conversion indirect "
1794                                        "item %h being deleted must be of "
1795                                        "4 byte long", le_ih);
1796
1797                 if (mode == M_CUT
1798                     && s_cut_balance.insert_size[0] != -UNFM_P_SIZE) {
1799                         reiserfs_panic(sb, "vs-5654", "can not complete "
1800                                        "indirect2direct conversion of %h "
1801                                        "(CUT, insert_size==%d)",
1802                                        le_ih, s_cut_balance.insert_size[0]);
1803                 }
1804                 /*
1805                  * it would be useful to make sure, that right neighboring
1806                  * item is direct item of this file
1807                  */
1808         }
1809 #endif
1810
1811         do_balance(&s_cut_balance, NULL, NULL, mode);
1812         if (is_inode_locked) {
1813                 /*
1814                  * we've done an indirect->direct conversion.  when the
1815                  * data block was freed, it was removed from the list of
1816                  * blocks that must be flushed before the transaction
1817                  * commits, make sure to unmap and invalidate it
1818                  */
1819                 unmap_buffers(page, tail_pos);
1820                 REISERFS_I(inode)->i_flags &= ~i_pack_on_close_mask;
1821         }
1822 #ifdef REISERQUOTA_DEBUG
1823         reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
1824                        "reiserquota cut_from_item(): freeing %u id=%u type=%c",
1825                        quota_cut_bytes, inode->i_uid, '?');
1826 #endif
1827         depth = reiserfs_write_unlock_nested(sb);
1828         dquot_free_space_nodirty(inode, quota_cut_bytes);
1829         reiserfs_write_lock_nested(sb, depth);
1830         return ret_value;
1831 }
1832
1833 static void truncate_directory(struct reiserfs_transaction_handle *th,
1834                                struct inode *inode)
1835 {
1836         BUG_ON(!th->t_trans_id);
1837         if (inode->i_nlink)
1838                 reiserfs_error(inode->i_sb, "vs-5655", "link count != 0");
1839
1840         set_le_key_k_offset(KEY_FORMAT_3_5, INODE_PKEY(inode), DOT_OFFSET);
1841         set_le_key_k_type(KEY_FORMAT_3_5, INODE_PKEY(inode), TYPE_DIRENTRY);
1842         reiserfs_delete_solid_item(th, inode, INODE_PKEY(inode));
1843         reiserfs_update_sd(th, inode);
1844         set_le_key_k_offset(KEY_FORMAT_3_5, INODE_PKEY(inode), SD_OFFSET);
1845         set_le_key_k_type(KEY_FORMAT_3_5, INODE_PKEY(inode), TYPE_STAT_DATA);
1846 }
1847
1848 /*
1849  * Truncate file to the new size. Note, this must be called with a
1850  * transaction already started
1851  */
1852 int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,
1853                          struct inode *inode,   /* ->i_size contains new size */
1854                          struct page *page,     /* up to date for last block */
1855                          /*
1856                           * when it is called by file_release to convert
1857                           * the tail - no timestamps should be updated
1858                           */
1859                          int update_timestamps
1860     )
1861 {
1862         INITIALIZE_PATH(s_search_path); /* Path to the current object item. */
1863         struct item_head *p_le_ih;      /* Pointer to an item header. */
1864
1865         /* Key to search for a previous file item. */
1866         struct cpu_key s_item_key;
1867         loff_t file_size,       /* Old file size. */
1868          new_file_size; /* New file size. */
1869         int deleted;            /* Number of deleted or truncated bytes. */
1870         int retval;
1871         int err = 0;
1872
1873         BUG_ON(!th->t_trans_id);
1874         if (!
1875             (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1876              || S_ISLNK(inode->i_mode)))
1877                 return 0;
1878
1879         /* deletion of directory - no need to update timestamps */
1880         if (S_ISDIR(inode->i_mode)) {
1881                 truncate_directory(th, inode);
1882                 return 0;
1883         }
1884
1885         /* Get new file size. */
1886         new_file_size = inode->i_size;
1887
1888         /* FIXME: note, that key type is unimportant here */
1889         make_cpu_key(&s_item_key, inode, max_reiserfs_offset(inode),
1890                      TYPE_DIRECT, 3);
1891
1892         retval =
1893             search_for_position_by_key(inode->i_sb, &s_item_key,
1894                                        &s_search_path);
1895         if (retval == IO_ERROR) {
1896                 reiserfs_error(inode->i_sb, "vs-5657",
1897                                "i/o failure occurred trying to truncate %K",
1898                                &s_item_key);
1899                 err = -EIO;
1900                 goto out;
1901         }
1902         if (retval == POSITION_FOUND || retval == FILE_NOT_FOUND) {
1903                 reiserfs_error(inode->i_sb, "PAP-5660",
1904                                "wrong result %d of search for %K", retval,
1905                                &s_item_key);
1906
1907                 err = -EIO;
1908                 goto out;
1909         }
1910
1911         s_search_path.pos_in_item--;
1912
1913         /* Get real file size (total length of all file items) */
1914         p_le_ih = tp_item_head(&s_search_path);
1915         if (is_statdata_le_ih(p_le_ih))
1916                 file_size = 0;
1917         else {
1918                 loff_t offset = le_ih_k_offset(p_le_ih);
1919                 int bytes =
1920                     op_bytes_number(p_le_ih, inode->i_sb->s_blocksize);
1921
1922                 /*
1923                  * this may mismatch with real file size: if last direct item
1924                  * had no padding zeros and last unformatted node had no free
1925                  * space, this file would have this file size
1926                  */
1927                 file_size = offset + bytes - 1;
1928         }
1929         /*
1930          * are we doing a full truncate or delete, if so
1931          * kick in the reada code
1932          */
1933         if (new_file_size == 0)
1934                 s_search_path.reada = PATH_READA | PATH_READA_BACK;
1935
1936         if (file_size == 0 || file_size < new_file_size) {
1937                 goto update_and_out;
1938         }
1939
1940         /* Update key to search for the last file item. */
1941         set_cpu_key_k_offset(&s_item_key, file_size);
1942
1943         do {
1944                 /* Cut or delete file item. */
1945                 deleted =
1946                     reiserfs_cut_from_item(th, &s_search_path, &s_item_key,
1947                                            inode, page, new_file_size);
1948                 if (deleted < 0) {
1949                         reiserfs_warning(inode->i_sb, "vs-5665",
1950                                          "reiserfs_cut_from_item failed");
1951                         reiserfs_check_path(&s_search_path);
1952                         return 0;
1953                 }
1954
1955                 RFALSE(deleted > file_size,
1956                        "PAP-5670: reiserfs_cut_from_item: too many bytes deleted: deleted %d, file_size %lu, item_key %K",
1957                        deleted, file_size, &s_item_key);
1958
1959                 /* Change key to search the last file item. */
1960                 file_size -= deleted;
1961
1962                 set_cpu_key_k_offset(&s_item_key, file_size);
1963
1964                 /*
1965                  * While there are bytes to truncate and previous
1966                  * file item is presented in the tree.
1967                  */
1968
1969                 /*
1970                  * This loop could take a really long time, and could log
1971                  * many more blocks than a transaction can hold.  So, we do
1972                  * a polite journal end here, and if the transaction needs
1973                  * ending, we make sure the file is consistent before ending
1974                  * the current trans and starting a new one
1975                  */
1976                 if (journal_transaction_should_end(th, 0) ||
1977                     reiserfs_transaction_free_space(th) <= JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD) {
1978                         pathrelse(&s_search_path);
1979
1980                         if (update_timestamps) {
1981                                 inode->i_mtime = current_time(inode);
1982                                 inode->i_ctime = current_time(inode);
1983                         }
1984                         reiserfs_update_sd(th, inode);
1985
1986                         err = journal_end(th);
1987                         if (err)
1988                                 goto out;
1989                         err = journal_begin(th, inode->i_sb,
1990                                             JOURNAL_FOR_FREE_BLOCK_AND_UPDATE_SD + JOURNAL_PER_BALANCE_CNT * 4) ;
1991                         if (err)
1992                                 goto out;
1993                         reiserfs_update_inode_transaction(inode);
1994                 }
1995         } while (file_size > ROUND_UP(new_file_size) &&
1996                  search_for_position_by_key(inode->i_sb, &s_item_key,
1997                                             &s_search_path) == POSITION_FOUND);
1998
1999         RFALSE(file_size > ROUND_UP(new_file_size),
2000                "PAP-5680: truncate did not finish: new_file_size %lld, current %lld, oid %d",
2001                new_file_size, file_size, s_item_key.on_disk_key.k_objectid);
2002
2003 update_and_out:
2004         if (update_timestamps) {
2005                 /* this is truncate, not file closing */
2006                 inode->i_mtime = current_time(inode);
2007                 inode->i_ctime = current_time(inode);
2008         }
2009         reiserfs_update_sd(th, inode);
2010
2011 out:
2012         pathrelse(&s_search_path);
2013         return err;
2014 }
2015
2016 #ifdef CONFIG_REISERFS_CHECK
2017 /* this makes sure, that we __append__, not overwrite or add holes */
2018 static void check_research_for_paste(struct treepath *path,
2019                                      const struct cpu_key *key)
2020 {
2021         struct item_head *found_ih = tp_item_head(path);
2022
2023         if (is_direct_le_ih(found_ih)) {
2024                 if (le_ih_k_offset(found_ih) +
2025                     op_bytes_number(found_ih,
2026                                     get_last_bh(path)->b_size) !=
2027                     cpu_key_k_offset(key)
2028                     || op_bytes_number(found_ih,
2029                                        get_last_bh(path)->b_size) !=
2030                     pos_in_item(path))
2031                         reiserfs_panic(NULL, "PAP-5720", "found direct item "
2032                                        "%h or position (%d) does not match "
2033                                        "to key %K", found_ih,
2034                                        pos_in_item(path), key);
2035         }
2036         if (is_indirect_le_ih(found_ih)) {
2037                 if (le_ih_k_offset(found_ih) +
2038                     op_bytes_number(found_ih,
2039                                     get_last_bh(path)->b_size) !=
2040                     cpu_key_k_offset(key)
2041                     || I_UNFM_NUM(found_ih) != pos_in_item(path)
2042                     || get_ih_free_space(found_ih) != 0)
2043                         reiserfs_panic(NULL, "PAP-5730", "found indirect "
2044                                        "item (%h) or position (%d) does not "
2045                                        "match to key (%K)",
2046                                        found_ih, pos_in_item(path), key);
2047         }
2048 }
2049 #endif                          /* config reiserfs check */
2050
2051 /*
2052  * Paste bytes to the existing item.
2053  * Returns bytes number pasted into the item.
2054  */
2055 int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,
2056                              /* Path to the pasted item. */
2057                              struct treepath *search_path,
2058                              /* Key to search for the needed item. */
2059                              const struct cpu_key *key,
2060                              /* Inode item belongs to */
2061                              struct inode *inode,
2062                              /* Pointer to the bytes to paste. */
2063                              const char *body,
2064                              /* Size of pasted bytes. */
2065                              int pasted_size)
2066 {
2067         struct super_block *sb = inode->i_sb;
2068         struct tree_balance s_paste_balance;
2069         int retval;
2070         int fs_gen;
2071         int depth;
2072
2073         BUG_ON(!th->t_trans_id);
2074
2075         fs_gen = get_generation(inode->i_sb);
2076
2077 #ifdef REISERQUOTA_DEBUG
2078         reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
2079                        "reiserquota paste_into_item(): allocating %u id=%u type=%c",
2080                        pasted_size, inode->i_uid,
2081                        key2type(&key->on_disk_key));
2082 #endif
2083
2084         depth = reiserfs_write_unlock_nested(sb);
2085         retval = dquot_alloc_space_nodirty(inode, pasted_size);
2086         reiserfs_write_lock_nested(sb, depth);
2087         if (retval) {
2088                 pathrelse(search_path);
2089                 return retval;
2090         }
2091         init_tb_struct(th, &s_paste_balance, th->t_super, search_path,
2092                        pasted_size);
2093 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2094         s_paste_balance.key = key->on_disk_key;
2095 #endif
2096
2097         /* DQUOT_* can schedule, must check before the fix_nodes */
2098         if (fs_changed(fs_gen, inode->i_sb)) {
2099                 goto search_again;
2100         }
2101
2102         while ((retval =
2103                 fix_nodes(M_PASTE, &s_paste_balance, NULL,
2104                           body)) == REPEAT_SEARCH) {
2105 search_again:
2106                 /* file system changed while we were in the fix_nodes */
2107                 PROC_INFO_INC(th->t_super, paste_into_item_restarted);
2108                 retval =
2109                     search_for_position_by_key(th->t_super, key,
2110                                                search_path);
2111                 if (retval == IO_ERROR) {
2112                         retval = -EIO;
2113                         goto error_out;
2114                 }
2115                 if (retval == POSITION_FOUND) {
2116                         reiserfs_warning(inode->i_sb, "PAP-5710",
2117                                          "entry or pasted byte (%K) exists",
2118                                          key);
2119                         retval = -EEXIST;
2120                         goto error_out;
2121                 }
2122 #ifdef CONFIG_REISERFS_CHECK
2123                 check_research_for_paste(search_path, key);
2124 #endif
2125         }
2126
2127         /*
2128          * Perform balancing after all resources are collected by fix_nodes,
2129          * and accessing them will not risk triggering schedule.
2130          */
2131         if (retval == CARRY_ON) {
2132                 do_balance(&s_paste_balance, NULL /*ih */ , body, M_PASTE);
2133                 return 0;
2134         }
2135         retval = (retval == NO_DISK_SPACE) ? -ENOSPC : -EIO;
2136 error_out:
2137         /* this also releases the path */
2138         unfix_nodes(&s_paste_balance);
2139 #ifdef REISERQUOTA_DEBUG
2140         reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
2141                        "reiserquota paste_into_item(): freeing %u id=%u type=%c",
2142                        pasted_size, inode->i_uid,
2143                        key2type(&key->on_disk_key));
2144 #endif
2145         depth = reiserfs_write_unlock_nested(sb);
2146         dquot_free_space_nodirty(inode, pasted_size);
2147         reiserfs_write_lock_nested(sb, depth);
2148         return retval;
2149 }
2150
2151 /*
2152  * Insert new item into the buffer at the path.
2153  * th   - active transaction handle
2154  * path - path to the inserted item
2155  * ih   - pointer to the item header to insert
2156  * body - pointer to the bytes to insert
2157  */
2158 int reiserfs_insert_item(struct reiserfs_transaction_handle *th,
2159                          struct treepath *path, const struct cpu_key *key,
2160                          struct item_head *ih, struct inode *inode,
2161                          const char *body)
2162 {
2163         struct tree_balance s_ins_balance;
2164         int retval;
2165         int fs_gen = 0;
2166         int quota_bytes = 0;
2167
2168         BUG_ON(!th->t_trans_id);
2169
2170         if (inode) {            /* Do we count quotas for item? */
2171                 int depth;
2172                 fs_gen = get_generation(inode->i_sb);
2173                 quota_bytes = ih_item_len(ih);
2174
2175                 /*
2176                  * hack so the quota code doesn't have to guess
2177                  * if the file has a tail, links are always tails,
2178                  * so there's no guessing needed
2179                  */
2180                 if (!S_ISLNK(inode->i_mode) && is_direct_le_ih(ih))
2181                         quota_bytes = inode->i_sb->s_blocksize + UNFM_P_SIZE;
2182 #ifdef REISERQUOTA_DEBUG
2183                 reiserfs_debug(inode->i_sb, REISERFS_DEBUG_CODE,
2184                                "reiserquota insert_item(): allocating %u id=%u type=%c",
2185                                quota_bytes, inode->i_uid, head2type(ih));
2186 #endif
2187                 /*
2188                  * We can't dirty inode here. It would be immediately
2189                  * written but appropriate stat item isn't inserted yet...
2190                  */
2191                 depth = reiserfs_write_unlock_nested(inode->i_sb);
2192                 retval = dquot_alloc_space_nodirty(inode, quota_bytes);
2193                 reiserfs_write_lock_nested(inode->i_sb, depth);
2194                 if (retval) {
2195                         pathrelse(path);
2196                         return retval;
2197                 }
2198         }
2199         init_tb_struct(th, &s_ins_balance, th->t_super, path,
2200                        IH_SIZE + ih_item_len(ih));
2201 #ifdef DISPLACE_NEW_PACKING_LOCALITIES
2202         s_ins_balance.key = key->on_disk_key;
2203 #endif
2204         /*
2205          * DQUOT_* can schedule, must check to be sure calling
2206          * fix_nodes is safe
2207          */
2208         if (inode && fs_changed(fs_gen, inode->i_sb)) {
2209                 goto search_again;
2210         }
2211
2212         while ((retval =
2213                 fix_nodes(M_INSERT, &s_ins_balance, ih,
2214                           body)) == REPEAT_SEARCH) {
2215 search_again:
2216                 /* file system changed while we were in the fix_nodes */
2217                 PROC_INFO_INC(th->t_super, insert_item_restarted);
2218                 retval = search_item(th->t_super, key, path);
2219                 if (retval == IO_ERROR) {
2220                         retval = -EIO;
2221                         goto error_out;
2222                 }
2223                 if (retval == ITEM_FOUND) {
2224                         reiserfs_warning(th->t_super, "PAP-5760",
2225                                          "key %K already exists in the tree",
2226                                          key);
2227                         retval = -EEXIST;
2228                         goto error_out;
2229                 }
2230         }
2231
2232         /* make balancing after all resources will be collected at a time */
2233         if (retval == CARRY_ON) {
2234                 do_balance(&s_ins_balance, ih, body, M_INSERT);
2235                 return 0;
2236         }
2237
2238         retval = (retval == NO_DISK_SPACE) ? -ENOSPC : -EIO;
2239 error_out:
2240         /* also releases the path */
2241         unfix_nodes(&s_ins_balance);
2242 #ifdef REISERQUOTA_DEBUG
2243         if (inode)
2244                 reiserfs_debug(th->t_super, REISERFS_DEBUG_CODE,
2245                        "reiserquota insert_item(): freeing %u id=%u type=%c",
2246                        quota_bytes, inode->i_uid, head2type(ih));
2247 #endif
2248         if (inode) {
2249                 int depth = reiserfs_write_unlock_nested(inode->i_sb);
2250                 dquot_free_space_nodirty(inode, quota_bytes);
2251                 reiserfs_write_lock_nested(inode->i_sb, depth);
2252         }
2253         return retval;
2254 }