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 /* 12 * This file contains functions dealing with S+tree 13 * 14 * B_IS_IN_TREE 15 * copy_item_head 16 * comp_short_keys 17 * comp_keys 18 * comp_short_le_keys 19 * le_key2cpu_key 20 * comp_le_keys 21 * bin_search 22 * get_lkey 23 * get_rkey 24 * key_in_buffer 25 * decrement_bcount 26 * decrement_counters_in_path 27 * reiserfs_check_path 28 * pathrelse_and_restore 29 * pathrelse 30 * search_by_key_reada 31 * search_by_key 32 * search_for_position_by_key 33 * comp_items 34 * prepare_for_direct_item 35 * prepare_for_direntry_item 36 * prepare_for_delete_or_cut 37 * calc_deleted_bytes_number 38 * init_tb_struct 39 * padd_item 40 * reiserfs_delete_item 41 * reiserfs_delete_solid_item 42 * reiserfs_delete_object 43 * maybe_indirect_to_direct 44 * indirect_to_direct_roll_back 45 * reiserfs_cut_from_item 46 * truncate_directory 47 * reiserfs_do_truncate 48 * reiserfs_paste_into_item 49 * reiserfs_insert_item 50 */ 51 52 #include <linux/config.h> 53 #include <linux/time.h> 54 #include <linux/string.h> 55 #include <linux/pagemap.h> 56 #include <linux/reiserfs_fs.h> 57 #include <linux/smp_lock.h> 58 #include <linux/buffer_head.h> 59 #include <linux/quotaops.h> 60 61 /* Does the buffer contain a disk block which is in the tree. */ 62 inline int B_IS_IN_TREE (const struct buffer_head * p_s_bh) 63 { 64 65 RFALSE( B_LEVEL (p_s_bh) > MAX_HEIGHT, 66 "PAP-1010: block (%b) has too big level (%z)", p_s_bh, p_s_bh); 67 68 return ( B_LEVEL (p_s_bh) != FREE_LEVEL ); 69 } 70 71 // 72 // to gets item head in le form 73 // 74 inline void copy_item_head(struct item_head * p_v_to, 75 const struct item_head * p_v_from) 76 { 77 memcpy (p_v_to, p_v_from, IH_SIZE); 78 } 79 80 81 /* k1 is pointer to on-disk structure which is stored in little-endian 82 form. k2 is pointer to cpu variable. For key of items of the same 83 object this returns 0. 84 Returns: -1 if key1 < key2 85 0 if key1 == key2 86 1 if key1 > key2 */ 87 inline int comp_short_keys (const struct reiserfs_key * le_key, 88 const struct cpu_key * cpu_key) 89 { 90 __u32 n; 91 n = le32_to_cpu(le_key->k_dir_id); 92 if (n < cpu_key->on_disk_key.k_dir_id) 93 return -1; 94 if (n > cpu_key->on_disk_key.k_dir_id) 95 return 1; 96 n = le32_to_cpu(le_key->k_objectid); 97 if (n < cpu_key->on_disk_key.k_objectid) 98 return -1; 99 if (n > cpu_key->on_disk_key.k_objectid) 100 return 1; 101 return 0; 102 } 103 104 /* k1 is pointer to on-disk structure which is stored in little-endian 105 form. k2 is pointer to cpu variable. 106 Compare keys using all 4 key fields. 107 Returns: -1 if key1 < key2 0 108 if key1 = key2 1 if key1 > key2 */ 109 static inline int comp_keys (const struct reiserfs_key * le_key, const struct cpu_key * cpu_key) 110 { 111 int retval; 112 113 retval = comp_short_keys (le_key, cpu_key); 114 if (retval) 115 return retval; 116 if (le_key_k_offset (le_key_version(le_key), le_key) < cpu_key_k_offset (cpu_key)) 117 return -1; 118 if (le_key_k_offset (le_key_version(le_key), le_key) > cpu_key_k_offset (cpu_key)) 119 return 1; 120 121 if (cpu_key->key_length == 3) 122 return 0; 123 124 /* this part is needed only when tail conversion is in progress */ 125 if (le_key_k_type (le_key_version(le_key), le_key) < cpu_key_k_type (cpu_key)) 126 return -1; 127 128 if (le_key_k_type (le_key_version(le_key), le_key) > cpu_key_k_type (cpu_key)) 129 return 1; 130 131 return 0; 132 } 133 134 135 inline int comp_short_le_keys (const struct reiserfs_key * key1, const struct reiserfs_key * key2) 136 { 137 __u32 * p_s_1_u32, * p_s_2_u32; 138 int n_key_length = REISERFS_SHORT_KEY_LEN; 139 140 p_s_1_u32 = (__u32 *)key1; 141 p_s_2_u32 = (__u32 *)key2; 142 for( ; n_key_length--; ++p_s_1_u32, ++p_s_2_u32 ) { 143 if ( le32_to_cpu (*p_s_1_u32) < le32_to_cpu (*p_s_2_u32) ) 144 return -1; 145 if ( le32_to_cpu (*p_s_1_u32) > le32_to_cpu (*p_s_2_u32) ) 146 return 1; 147 } 148 return 0; 149 } 150 151 inline void le_key2cpu_key (struct cpu_key * to, const struct reiserfs_key * from) 152 { 153 int version; 154 to->on_disk_key.k_dir_id = le32_to_cpu (from->k_dir_id); 155 to->on_disk_key.k_objectid = le32_to_cpu (from->k_objectid); 156 157 // find out version of the key 158 version = le_key_version (from); 159 to->version = version; 160 to->on_disk_key.k_offset = le_key_k_offset(version, from); 161 to->on_disk_key.k_type = le_key_k_type(version, from); 162 } 163 164 165 166 // this does not say which one is bigger, it only returns 1 if keys 167 // are not equal, 0 otherwise 168 inline int comp_le_keys (const struct reiserfs_key * k1, const struct reiserfs_key * k2) 169 { 170 return memcmp (k1, k2, sizeof (struct reiserfs_key)); 171 } 172 173 /************************************************************************** 174 * Binary search toolkit function * 175 * Search for an item in the array by the item key * 176 * Returns: 1 if found, 0 if not found; * 177 * *p_n_pos = number of the searched element if found, else the * 178 * number of the first element that is larger than p_v_key. * 179 **************************************************************************/ 180 /* For those not familiar with binary search: n_lbound is the leftmost item that it 181 could be, n_rbound the rightmost item that it could be. We examine the item 182 halfway between n_lbound and n_rbound, and that tells us either that we can increase 183 n_lbound, or decrease n_rbound, or that we have found it, or if n_lbound <= n_rbound that 184 there are no possible items, and we have not found it. With each examination we 185 cut the number of possible items it could be by one more than half rounded down, 186 or we find it. */ 187 static inline int bin_search ( 188 const void * p_v_key, /* Key to search for. */ 189 const void * p_v_base,/* First item in the array. */ 190 int p_n_num, /* Number of items in the array. */ 191 int p_n_width, /* Item size in the array. 192 searched. Lest the reader be 193 confused, note that this is crafted 194 as a general function, and when it 195 is applied specifically to the array 196 of item headers in a node, p_n_width 197 is actually the item header size not 198 the item size. */ 199 int * p_n_pos /* Number of the searched for element. */ 200 ) { 201 int n_rbound, n_lbound, n_j; 202 203 for ( n_j = ((n_rbound = p_n_num - 1) + (n_lbound = 0))/2; n_lbound <= n_rbound; n_j = (n_rbound + n_lbound)/2 ) 204 switch( comp_keys((struct reiserfs_key *)((char * )p_v_base + n_j * p_n_width), (struct cpu_key *)p_v_key) ) { 205 case -1: n_lbound = n_j + 1; continue; 206 case 1: n_rbound = n_j - 1; continue; 207 case 0: *p_n_pos = n_j; return ITEM_FOUND; /* Key found in the array. */ 208 } 209 210 /* bin_search did not find given key, it returns position of key, 211 that is minimal and greater than the given one. */ 212 *p_n_pos = n_lbound; 213 return ITEM_NOT_FOUND; 214 } 215 216 #ifdef CONFIG_REISERFS_CHECK 217 extern struct tree_balance * cur_tb; 218 #endif 219 220 221 222 /* Minimal possible key. It is never in the tree. */ 223 const struct reiserfs_key MIN_KEY = {0, 0, {{0, 0},}}; 224 225 /* Maximal possible key. It is never in the tree. */ 226 const struct reiserfs_key MAX_KEY = { 227 __constant_cpu_to_le32(0xffffffff), 228 __constant_cpu_to_le32(0xffffffff), 229 {{__constant_cpu_to_le32(0xffffffff), 230 __constant_cpu_to_le32(0xffffffff)},} 231 }; 232 233 234 /* Get delimiting key of the buffer by looking for it in the buffers in the path, starting from the bottom 235 of the path, and going upwards. We must check the path's validity at each step. If the key is not in 236 the path, there is no delimiting key in the tree (buffer is first or last buffer in tree), and in this 237 case we return a special key, either MIN_KEY or MAX_KEY. */ 238 static inline const struct reiserfs_key * get_lkey ( 239 const struct path * p_s_chk_path, 240 const struct super_block * p_s_sb 241 ) { 242 int n_position, n_path_offset = p_s_chk_path->path_length; 243 struct buffer_head * p_s_parent; 244 245 RFALSE( n_path_offset < FIRST_PATH_ELEMENT_OFFSET, 246 "PAP-5010: invalid offset in the path"); 247 248 /* While not higher in path than first element. */ 249 while ( n_path_offset-- > FIRST_PATH_ELEMENT_OFFSET ) { 250 251 RFALSE( ! buffer_uptodate(PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)), 252 "PAP-5020: parent is not uptodate"); 253 254 /* Parent at the path is not in the tree now. */ 255 if ( ! B_IS_IN_TREE(p_s_parent = PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)) ) 256 return &MAX_KEY; 257 /* Check whether position in the parent is correct. */ 258 if ( (n_position = PATH_OFFSET_POSITION(p_s_chk_path, n_path_offset)) > B_NR_ITEMS(p_s_parent) ) 259 return &MAX_KEY; 260 /* Check whether parent at the path really points to the child. */ 261 if ( B_N_CHILD_NUM(p_s_parent, n_position) != 262 PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset + 1)->b_blocknr ) 263 return &MAX_KEY; 264 /* Return delimiting key if position in the parent is not equal to zero. */ 265 if ( n_position ) 266 return B_N_PDELIM_KEY(p_s_parent, n_position - 1); 267 } 268 /* Return MIN_KEY if we are in the root of the buffer tree. */ 269 if ( PATH_OFFSET_PBUFFER(p_s_chk_path, FIRST_PATH_ELEMENT_OFFSET)->b_blocknr == 270 SB_ROOT_BLOCK (p_s_sb) ) 271 return &MIN_KEY; 272 return &MAX_KEY; 273 } 274 275 276 /* Get delimiting key of the buffer at the path and its right neighbor. */ 277 inline const struct reiserfs_key * get_rkey ( 278 const struct path * p_s_chk_path, 279 const struct super_block * p_s_sb 280 ) { 281 int n_position, 282 n_path_offset = p_s_chk_path->path_length; 283 struct buffer_head * p_s_parent; 284 285 RFALSE( n_path_offset < FIRST_PATH_ELEMENT_OFFSET, 286 "PAP-5030: invalid offset in the path"); 287 288 while ( n_path_offset-- > FIRST_PATH_ELEMENT_OFFSET ) { 289 290 RFALSE( ! buffer_uptodate(PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)), 291 "PAP-5040: parent is not uptodate"); 292 293 /* Parent at the path is not in the tree now. */ 294 if ( ! B_IS_IN_TREE(p_s_parent = PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset)) ) 295 return &MIN_KEY; 296 /* Check whether position in the parent is correct. */ 297 if ( (n_position = PATH_OFFSET_POSITION(p_s_chk_path, n_path_offset)) > B_NR_ITEMS(p_s_parent) ) 298 return &MIN_KEY; 299 /* Check whether parent at the path really points to the child. */ 300 if ( B_N_CHILD_NUM(p_s_parent, n_position) != 301 PATH_OFFSET_PBUFFER(p_s_chk_path, n_path_offset + 1)->b_blocknr ) 302 return &MIN_KEY; 303 /* Return delimiting key if position in the parent is not the last one. */ 304 if ( n_position != B_NR_ITEMS(p_s_parent) ) 305 return B_N_PDELIM_KEY(p_s_parent, n_position); 306 } 307 /* Return MAX_KEY if we are in the root of the buffer tree. */ 308 if ( PATH_OFFSET_PBUFFER(p_s_chk_path, FIRST_PATH_ELEMENT_OFFSET)->b_blocknr == 309 SB_ROOT_BLOCK (p_s_sb) ) 310 return &MAX_KEY; 311 return &MIN_KEY; 312 } 313 314 315 /* Check whether a key is contained in the tree rooted from a buffer at a path. */ 316 /* This works by looking at the left and right delimiting keys for the buffer in the last path_element in 317 the path. These delimiting keys are stored at least one level above that buffer in the tree. If the 318 buffer is the first or last node in the tree order then one of the delimiting keys may be absent, and in 319 this case get_lkey and get_rkey return a special key which is MIN_KEY or MAX_KEY. */ 320 static inline int key_in_buffer ( 321 struct path * p_s_chk_path, /* Path which should be checked. */ 322 const struct cpu_key * p_s_key, /* Key which should be checked. */ 323 struct super_block * p_s_sb /* Super block pointer. */ 324 ) { 325 326 RFALSE( ! p_s_key || p_s_chk_path->path_length < FIRST_PATH_ELEMENT_OFFSET || 327 p_s_chk_path->path_length > MAX_HEIGHT, 328 "PAP-5050: pointer to the key(%p) is NULL or invalid path length(%d)", 329 p_s_key, p_s_chk_path->path_length); 330 RFALSE( !PATH_PLAST_BUFFER(p_s_chk_path)->b_bdev, 331 "PAP-5060: device must not be NODEV"); 332 333 if ( comp_keys(get_lkey(p_s_chk_path, p_s_sb), p_s_key) == 1 ) 334 /* left delimiting key is bigger, that the key we look for */ 335 return 0; 336 // if ( comp_keys(p_s_key, get_rkey(p_s_chk_path, p_s_sb)) != -1 ) 337 if ( comp_keys(get_rkey(p_s_chk_path, p_s_sb), p_s_key) != 1 ) 338 /* p_s_key must be less than right delimitiing key */ 339 return 0; 340 return 1; 341 } 342 343 344 inline void decrement_bcount( 345 struct buffer_head * p_s_bh 346 ) { 347 if ( p_s_bh ) { 348 if ( atomic_read (&(p_s_bh->b_count)) ) { 349 put_bh(p_s_bh) ; 350 return; 351 } 352 reiserfs_panic(NULL, "PAP-5070: decrement_bcount: trying to free free buffer %b", p_s_bh); 353 } 354 } 355 356 357 /* Decrement b_count field of the all buffers in the path. */ 358 void decrement_counters_in_path ( 359 struct path * p_s_search_path 360 ) { 361 int n_path_offset = p_s_search_path->path_length; 362 363 RFALSE( n_path_offset < ILLEGAL_PATH_ELEMENT_OFFSET || 364 n_path_offset > EXTENDED_MAX_HEIGHT - 1, 365 "PAP-5080: invalid path offset of %d", n_path_offset); 366 367 while ( n_path_offset > ILLEGAL_PATH_ELEMENT_OFFSET ) { 368 struct buffer_head * bh; 369 370 bh = PATH_OFFSET_PBUFFER(p_s_search_path, n_path_offset--); 371 decrement_bcount (bh); 372 } 373 p_s_search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET; 374 } 375 376 377 int reiserfs_check_path(struct path *p) { 378 RFALSE( p->path_length != ILLEGAL_PATH_ELEMENT_OFFSET, 379 "path not properly relsed") ; 380 return 0 ; 381 } 382 383 384 /* Release all buffers in the path. Restore dirty bits clean 385 ** when preparing the buffer for the log 386 ** 387 ** only called from fix_nodes() 388 */ 389 void pathrelse_and_restore ( 390 struct super_block *s, 391 struct path * p_s_search_path 392 ) { 393 int n_path_offset = p_s_search_path->path_length; 394 395 RFALSE( n_path_offset < ILLEGAL_PATH_ELEMENT_OFFSET, 396 "clm-4000: invalid path offset"); 397 398 while ( n_path_offset > ILLEGAL_PATH_ELEMENT_OFFSET ) { 399 reiserfs_restore_prepared_buffer(s, PATH_OFFSET_PBUFFER(p_s_search_path, 400 n_path_offset)); 401 brelse(PATH_OFFSET_PBUFFER(p_s_search_path, n_path_offset--)); 402 } 403 p_s_search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET; 404 } 405 406 /* Release all buffers in the path. */ 407 void pathrelse ( 408 struct path * p_s_search_path 409 ) { 410 int n_path_offset = p_s_search_path->path_length; 411 412 RFALSE( n_path_offset < ILLEGAL_PATH_ELEMENT_OFFSET, 413 "PAP-5090: invalid path offset"); 414 415 while ( n_path_offset > ILLEGAL_PATH_ELEMENT_OFFSET ) 416 brelse(PATH_OFFSET_PBUFFER(p_s_search_path, n_path_offset--)); 417 418 p_s_search_path->path_length = ILLEGAL_PATH_ELEMENT_OFFSET; 419 } 420 421 422 423 static int is_leaf (char * buf, int blocksize, struct buffer_head * bh) 424 { 425 struct block_head * blkh; 426 struct item_head * ih; 427 int used_space; 428 int prev_location; 429 int i; 430 int nr; 431 432 blkh = (struct block_head *)buf; 433 if ( blkh_level(blkh) != DISK_LEAF_NODE_LEVEL) { 434 reiserfs_warning (NULL, "is_leaf: this should be caught earlier"); 435 return 0; 436 } 437 438 nr = blkh_nr_item(blkh); 439 if (nr < 1 || nr > ((blocksize - BLKH_SIZE) / (IH_SIZE + MIN_ITEM_LEN))) { 440 /* item number is too big or too small */ 441 reiserfs_warning (NULL, "is_leaf: nr_item seems wrong: %z", bh); 442 return 0; 443 } 444 ih = (struct item_head *)(buf + BLKH_SIZE) + nr - 1; 445 used_space = BLKH_SIZE + IH_SIZE * nr + (blocksize - ih_location (ih)); 446 if (used_space != blocksize - blkh_free_space(blkh)) { 447 /* free space does not match to calculated amount of use space */ 448 reiserfs_warning (NULL, "is_leaf: free space seems wrong: %z", bh); 449 return 0; 450 } 451 452 // FIXME: it is_leaf will hit performance too much - we may have 453 // return 1 here 454 455 /* check tables of item heads */ 456 ih = (struct item_head *)(buf + BLKH_SIZE); 457 prev_location = blocksize; 458 for (i = 0; i < nr; i ++, ih ++) { 459 if ( le_ih_k_type(ih) == TYPE_ANY) { 460 reiserfs_warning (NULL, "is_leaf: wrong item type for item %h",ih); 461 return 0; 462 } 463 if (ih_location (ih) >= blocksize || ih_location (ih) < IH_SIZE * nr) { 464 reiserfs_warning (NULL, "is_leaf: item location seems wrong: %h", ih); 465 return 0; 466 } 467 if (ih_item_len (ih) < 1 || ih_item_len (ih) > MAX_ITEM_LEN (blocksize)) { 468 reiserfs_warning (NULL, "is_leaf: item length seems wrong: %h", ih); 469 return 0; 470 } 471 if (prev_location - ih_location (ih) != ih_item_len (ih)) { 472 reiserfs_warning (NULL, "is_leaf: item location seems wrong (second one): %h", ih); 473 return 0; 474 } 475 prev_location = ih_location (ih); 476 } 477 478 // one may imagine much more checks 479 return 1; 480 } 481 482 483 /* returns 1 if buf looks like an internal node, 0 otherwise */ 484 static int is_internal (char * buf, int blocksize, struct buffer_head * bh) 485 { 486 struct block_head * blkh; 487 int nr; 488 int used_space; 489 490 blkh = (struct block_head *)buf; 491 nr = blkh_level(blkh); 492 if (nr <= DISK_LEAF_NODE_LEVEL || nr > MAX_HEIGHT) { 493 /* this level is not possible for internal nodes */ 494 reiserfs_warning (NULL, "is_internal: this should be caught earlier"); 495 return 0; 496 } 497 498 nr = blkh_nr_item(blkh); 499 if (nr > (blocksize - BLKH_SIZE - DC_SIZE) / (KEY_SIZE + DC_SIZE)) { 500 /* for internal which is not root we might check min number of keys */ 501 reiserfs_warning (NULL, "is_internal: number of key seems wrong: %z", bh); 502 return 0; 503 } 504 505 used_space = BLKH_SIZE + KEY_SIZE * nr + DC_SIZE * (nr + 1); 506 if (used_space != blocksize - blkh_free_space(blkh)) { 507 reiserfs_warning (NULL, "is_internal: free space seems wrong: %z", bh); 508 return 0; 509 } 510 511 // one may imagine much more checks 512 return 1; 513 } 514 515 516 // make sure that bh contains formatted node of reiserfs tree of 517 // 'level'-th level 518 static int is_tree_node (struct buffer_head * bh, int level) 519 { 520 if (B_LEVEL (bh) != level) { 521 reiserfs_warning (NULL, "is_tree_node: node level %d does not match to the expected one %d", 522 B_LEVEL (bh), level); 523 return 0; 524 } 525 if (level == DISK_LEAF_NODE_LEVEL) 526 return is_leaf (bh->b_data, bh->b_size, bh); 527 528 return is_internal (bh->b_data, bh->b_size, bh); 529 } 530 531 532 533 #define SEARCH_BY_KEY_READA 16 534 535 /* The function is NOT SCHEDULE-SAFE! */ 536 static void search_by_key_reada (struct super_block * s, 537 struct buffer_head **bh, 538 unsigned long *b, int num) 539 { 540 int i,j; 541 542 for (i = 0 ; i < num ; i++) { 543 bh[i] = sb_getblk (s, b[i]); 544 } 545 for (j = 0 ; j < i ; j++) { 546 /* 547 * note, this needs attention if we are getting rid of the BKL 548 * you have to make sure the prepared bit isn't set on this buffer 549 */ 550 if (!buffer_uptodate(bh[j])) 551 ll_rw_block(READA, 1, bh + j); 552 brelse(bh[j]); 553 } 554 } 555 556 /************************************************************************** 557 * Algorithm SearchByKey * 558 * look for item in the Disk S+Tree by its key * 559 * Input: p_s_sb - super block * 560 * p_s_key - pointer to the key to search * 561 * Output: ITEM_FOUND, ITEM_NOT_FOUND or IO_ERROR * 562 * p_s_search_path - path from the root to the needed leaf * 563 **************************************************************************/ 564 565 /* This function fills up the path from the root to the leaf as it 566 descends the tree looking for the key. It uses reiserfs_bread to 567 try to find buffers in the cache given their block number. If it 568 does not find them in the cache it reads them from disk. For each 569 node search_by_key finds using reiserfs_bread it then uses 570 bin_search to look through that node. bin_search will find the 571 position of the block_number of the next node if it is looking 572 through an internal node. If it is looking through a leaf node 573 bin_search will find the position of the item which has key either 574 equal to given key, or which is the maximal key less than the given 575 key. search_by_key returns a path that must be checked for the 576 correctness of the top of the path but need not be checked for the 577 correctness of the bottom of the path */ 578 /* The function is NOT SCHEDULE-SAFE! */ 579 int search_by_key (struct super_block * p_s_sb, 580 const struct cpu_key * p_s_key, /* Key to search. */ 581 struct path * p_s_search_path, /* This structure was 582 allocated and initialized 583 by the calling 584 function. It is filled up 585 by this function. */ 586 int n_stop_level /* How far down the tree to search. To 587 stop at leaf level - set to 588 DISK_LEAF_NODE_LEVEL */ 589 ) { 590 int n_block_number; 591 int expected_level; 592 struct buffer_head * p_s_bh; 593 struct path_element * p_s_last_element; 594 int n_node_level, n_retval; 595 int right_neighbor_of_leaf_node; 596 int fs_gen; 597 struct buffer_head *reada_bh[SEARCH_BY_KEY_READA]; 598 unsigned long reada_blocks[SEARCH_BY_KEY_READA]; 599 int reada_count = 0; 600 601 #ifdef CONFIG_REISERFS_CHECK 602 int n_repeat_counter = 0; 603 #endif 604 605 PROC_INFO_INC( p_s_sb, search_by_key ); 606 607 /* As we add each node to a path we increase its count. This means that 608 we must be careful to release all nodes in a path before we either 609 discard the path struct or re-use the path struct, as we do here. */ 610 611 decrement_counters_in_path(p_s_search_path); 612 613 right_neighbor_of_leaf_node = 0; 614 615 /* With each iteration of this loop we search through the items in the 616 current node, and calculate the next current node(next path element) 617 for the next iteration of this loop.. */ 618 n_block_number = SB_ROOT_BLOCK (p_s_sb); 619 expected_level = -1; 620 while ( 1 ) { 621 622 #ifdef CONFIG_REISERFS_CHECK 623 if ( !(++n_repeat_counter % 50000) ) 624 reiserfs_warning (p_s_sb, "PAP-5100: search_by_key: %s:" 625 "there were %d iterations of while loop " 626 "looking for key %K", 627 current->comm, n_repeat_counter, p_s_key); 628 #endif 629 630 /* prep path to have another element added to it. */ 631 p_s_last_element = PATH_OFFSET_PELEMENT(p_s_search_path, ++p_s_search_path->path_length); 632 fs_gen = get_generation (p_s_sb); 633 634 /* Read the next tree node, and set the last element in the path to 635 have a pointer to it. */ 636 if ((p_s_bh = p_s_last_element->pe_buffer = 637 sb_getblk(p_s_sb, n_block_number)) ) { 638 if (!buffer_uptodate(p_s_bh) && reada_count > 1) { 639 search_by_key_reada (p_s_sb, reada_bh, 640 reada_blocks, reada_count); 641 } 642 ll_rw_block(READ, 1, &p_s_bh); 643 wait_on_buffer(p_s_bh); 644 if (!buffer_uptodate(p_s_bh)) 645 goto io_error; 646 } else { 647 io_error: 648 p_s_search_path->path_length --; 649 pathrelse(p_s_search_path); 650 return IO_ERROR; 651 } 652 reada_count = 0; 653 if (expected_level == -1) 654 expected_level = SB_TREE_HEIGHT (p_s_sb); 655 expected_level --; 656 657 /* It is possible that schedule occurred. We must check whether the key 658 to search is still in the tree rooted from the current buffer. If 659 not then repeat search from the root. */ 660 if ( fs_changed (fs_gen, p_s_sb) && 661 (!B_IS_IN_TREE (p_s_bh) || 662 B_LEVEL(p_s_bh) != expected_level || 663 !key_in_buffer(p_s_search_path, p_s_key, p_s_sb))) { 664 PROC_INFO_INC( p_s_sb, search_by_key_fs_changed ); 665 PROC_INFO_INC( p_s_sb, search_by_key_restarted ); 666 PROC_INFO_INC( p_s_sb, sbk_restarted[ expected_level - 1 ] ); 667 decrement_counters_in_path(p_s_search_path); 668 669 /* Get the root block number so that we can repeat the search 670 starting from the root. */ 671 n_block_number = SB_ROOT_BLOCK (p_s_sb); 672 expected_level = -1; 673 right_neighbor_of_leaf_node = 0; 674 675 /* repeat search from the root */ 676 continue; 677 } 678 679 /* only check that the key is in the buffer if p_s_key is not 680 equal to the MAX_KEY. Latter case is only possible in 681 "finish_unfinished()" processing during mount. */ 682 RFALSE( comp_keys( &MAX_KEY, p_s_key ) && 683 ! key_in_buffer(p_s_search_path, p_s_key, p_s_sb), 684 "PAP-5130: key is not in the buffer"); 685 #ifdef CONFIG_REISERFS_CHECK 686 if ( cur_tb ) { 687 print_cur_tb ("5140"); 688 reiserfs_panic(p_s_sb, "PAP-5140: search_by_key: schedule occurred in do_balance!"); 689 } 690 #endif 691 692 // make sure, that the node contents look like a node of 693 // certain level 694 if (!is_tree_node (p_s_bh, expected_level)) { 695 reiserfs_warning (p_s_sb, "vs-5150: search_by_key: " 696 "invalid format found in block %ld. Fsck?", 697 p_s_bh->b_blocknr); 698 pathrelse (p_s_search_path); 699 return IO_ERROR; 700 } 701 702 /* ok, we have acquired next formatted node in the tree */ 703 n_node_level = B_LEVEL (p_s_bh); 704 705 PROC_INFO_BH_STAT( p_s_sb, p_s_bh, n_node_level - 1 ); 706 707 RFALSE( n_node_level < n_stop_level, 708 "vs-5152: tree level (%d) is less than stop level (%d)", 709 n_node_level, n_stop_level); 710 711 n_retval = bin_search( p_s_key, B_N_PITEM_HEAD(p_s_bh, 0), 712 B_NR_ITEMS(p_s_bh), 713 ( n_node_level == DISK_LEAF_NODE_LEVEL ) ? IH_SIZE : KEY_SIZE, 714 &(p_s_last_element->pe_position)); 715 if (n_node_level == n_stop_level) { 716 return n_retval; 717 } 718 719 /* we are not in the stop level */ 720 if (n_retval == ITEM_FOUND) 721 /* item has been found, so we choose the pointer which is to the right of the found one */ 722 p_s_last_element->pe_position++; 723 724 /* if item was not found we choose the position which is to 725 the left of the found item. This requires no code, 726 bin_search did it already.*/ 727 728 /* So we have chosen a position in the current node which is 729 an internal node. Now we calculate child block number by 730 position in the node. */ 731 n_block_number = B_N_CHILD_NUM(p_s_bh, p_s_last_element->pe_position); 732 733 /* if we are going to read leaf nodes, try for read ahead as well */ 734 if ((p_s_search_path->reada & PATH_READA) && 735 n_node_level == DISK_LEAF_NODE_LEVEL + 1) 736 { 737 int pos = p_s_last_element->pe_position; 738 int limit = B_NR_ITEMS(p_s_bh); 739 struct reiserfs_key *le_key; 740 741 if (p_s_search_path->reada & PATH_READA_BACK) 742 limit = 0; 743 while(reada_count < SEARCH_BY_KEY_READA) { 744 if (pos == limit) 745 break; 746 reada_blocks[reada_count++] = B_N_CHILD_NUM(p_s_bh, pos); 747 if (p_s_search_path->reada & PATH_READA_BACK) 748 pos--; 749 else 750 pos++; 751 752 /* 753 * check to make sure we're in the same object 754 */ 755 le_key = B_N_PDELIM_KEY(p_s_bh, pos); 756 if (le32_to_cpu(le_key->k_objectid) != 757 p_s_key->on_disk_key.k_objectid) 758 { 759 break; 760 } 761 } 762 } 763 } 764 } 765 766 767 /* Form the path to an item and position in this item which contains 768 file byte defined by p_s_key. If there is no such item 769 corresponding to the key, we point the path to the item with 770 maximal key less than p_s_key, and *p_n_pos_in_item is set to one 771 past the last entry/byte in the item. If searching for entry in a 772 directory item, and it is not found, *p_n_pos_in_item is set to one 773 entry more than the entry with maximal key which is less than the 774 sought key. 775 776 Note that if there is no entry in this same node which is one more, 777 then we point to an imaginary entry. for direct items, the 778 position is in units of bytes, for indirect items the position is 779 in units of blocknr entries, for directory items the position is in 780 units of directory entries. */ 781 782 /* The function is NOT SCHEDULE-SAFE! */ 783 int search_for_position_by_key (struct super_block * p_s_sb, /* Pointer to the super block. */ 784 const struct cpu_key * p_cpu_key, /* Key to search (cpu variable) */ 785 struct path * p_s_search_path /* Filled up by this function. */ 786 ) { 787 struct item_head * p_le_ih; /* pointer to on-disk structure */ 788 int n_blk_size; 789 loff_t item_offset, offset; 790 struct reiserfs_dir_entry de; 791 int retval; 792 793 /* If searching for directory entry. */ 794 if ( is_direntry_cpu_key (p_cpu_key) ) 795 return search_by_entry_key (p_s_sb, p_cpu_key, p_s_search_path, &de); 796 797 /* If not searching for directory entry. */ 798 799 /* If item is found. */ 800 retval = search_item (p_s_sb, p_cpu_key, p_s_search_path); 801 if (retval == IO_ERROR) 802 return retval; 803 if ( retval == ITEM_FOUND ) { 804 805 RFALSE( ! ih_item_len( 806 B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_search_path), 807 PATH_LAST_POSITION(p_s_search_path))), 808 "PAP-5165: item length equals zero"); 809 810 pos_in_item(p_s_search_path) = 0; 811 return POSITION_FOUND; 812 } 813 814 RFALSE( ! PATH_LAST_POSITION(p_s_search_path), 815 "PAP-5170: position equals zero"); 816 817 /* Item is not found. Set path to the previous item. */ 818 p_le_ih = B_N_PITEM_HEAD(PATH_PLAST_BUFFER(p_s_search_path), --PATH_LAST_POSITION(p_s_search_path)); 819 n_blk_size = p_s_sb->s_blocksize; 820 821 if (comp_short_keys (&(p_le_ih->ih_key), p_cpu_key)) { 822 return FILE_NOT_FOUND; 823 } 824 825 // FIXME: quite ugly this far 826 827 item_offset = le_ih_k_offset (p_le_ih); 828 offset = cpu_key_k_offset (p_cpu_key); 829 830 /* Needed byte is contained in the item pointed to by the path.*/ 831 if (item_offset <= offset && 832 item_offset + op_bytes_number (p_le_ih, n_blk_size) > offset) { 833 pos_in_item (p_s_search_path) = offset - item_offset; 834 if ( is_indirect_le_ih(p_le_ih) ) { 835 pos_in_item (p_s_search_path) /= n_blk_size; 836 } 837 return POSITION_FOUND; 838 } 839 840 /* Needed byte is not contained in the item pointed to by the 841 path. Set pos_in_item out of the item. */ 842 if ( is_indirect_le_ih (p_le_ih) ) 843 pos_in_item (p_s_search_path) = ih_item_len(p_le_ih) / UNFM_P_SIZE; 844 else 845 pos_in_item (p_s_search_path) = ih_item_len( p_le_ih ); 846 847 return POSITION_NOT_FOUND; 848 } 849 850 851 /* Compare given item and item pointed to by the path. */ 852 int comp_items (const struct item_head * stored_ih, const struct path * p_s_path) 853 { 854 struct buffer_head * p_s_bh; 855 struct item_head * ih; 856 857 /* Last buffer at the path is not in the tree. */ 858 if ( ! B_IS_IN_TREE(p_s_bh = PATH_PLAST_BUFFER(p_s_path)) ) 859 return 1; 860 861 /* Last path position is invalid. */ 862 if ( PATH_LAST_POSITION(p_s_path) >= B_NR_ITEMS(p_s_bh) ) 863 return 1; 864 865 /* we need only to know, whether it is the same item */ 866 ih = get_ih (p_s_path); 867 return memcmp (stored_ih, ih, IH_SIZE); 868 } 869 870 871 /* unformatted nodes are not logged anymore, ever. This is safe 872 ** now 873 */ 874 #define held_by_others(bh) (atomic_read(&(bh)->b_count) > 1) 875 876 // block can not be forgotten as it is in I/O or held by someone 877 #define block_in_use(bh) (buffer_locked(bh) || (held_by_others(bh))) 878 879 880 881 // prepare for delete or cut of direct item 882 static inline int prepare_for_direct_item (struct path * path, 883 struct item_head * le_ih, 884 struct inode * inode, 885 loff_t new_file_length, 886 int * cut_size) 887 { 888 loff_t round_len; 889 890 891 if ( new_file_length == max_reiserfs_offset (inode) ) { 892 /* item has to be deleted */ 893 *cut_size = -(IH_SIZE + ih_item_len(le_ih)); 894 return M_DELETE; 895 } 896 897 // new file gets truncated 898 if (get_inode_item_key_version (inode) == KEY_FORMAT_3_6) { 899 // 900 round_len = ROUND_UP (new_file_length); 901 /* this was n_new_file_length < le_ih ... */ 902 if ( round_len < le_ih_k_offset (le_ih) ) { 903 *cut_size = -(IH_SIZE + ih_item_len(le_ih)); 904 return M_DELETE; /* Delete this item. */ 905 } 906 /* Calculate first position and size for cutting from item. */ 907 pos_in_item (path) = round_len - (le_ih_k_offset (le_ih) - 1); 908 *cut_size = -(ih_item_len(le_ih) - pos_in_item(path)); 909 910 return M_CUT; /* Cut from this item. */ 911 } 912 913 914 // old file: items may have any length 915 916 if ( new_file_length < le_ih_k_offset (le_ih) ) { 917 *cut_size = -(IH_SIZE + ih_item_len(le_ih)); 918 return M_DELETE; /* Delete this item. */ 919 } 920 /* Calculate first position and size for cutting from item. */ 921 *cut_size = -(ih_item_len(le_ih) - 922 (pos_in_item (path) = new_file_length + 1 - le_ih_k_offset (le_ih))); 923 return M_CUT; /* Cut from this item. */ 924 } 925 926 927 static inline int prepare_for_direntry_item (struct path * path, 928 struct item_head * le_ih, 929 struct inode * inode, 930 loff_t new_file_length, 931 int * cut_size) 932 { 933 if (le_ih_k_offset (le_ih) == DOT_OFFSET && 934 new_file_length == max_reiserfs_offset (inode)) { 935 RFALSE( ih_entry_count (le_ih) != 2, 936 "PAP-5220: incorrect empty directory item (%h)", le_ih); 937 *cut_size = -(IH_SIZE + ih_item_len(le_ih)); 938 return M_DELETE; /* Delete the directory item containing "." and ".." entry. */ 939 } 940 941 if ( ih_entry_count (le_ih) == 1 ) { 942 /* Delete the directory item such as there is one record only 943 in this item*/ 944 *cut_size = -(IH_SIZE + ih_item_len(le_ih)); 945 return M_DELETE; 946 } 947 948 /* Cut one record from the directory item. */ 949 *cut_size = -(DEH_SIZE + entry_length (get_last_bh (path), le_ih, pos_in_item (path))); 950 return M_CUT; 951 } 952 953 954 /* If the path points to a directory or direct item, calculate mode and the size cut, for balance. 955 If the path points to an indirect item, remove some number of its unformatted nodes. 956 In case of file truncate calculate whether this item must be deleted/truncated or last 957 unformatted node of this item will be converted to a direct item. 958 This function returns a determination of what balance mode the calling function should employ. */ 959 static char prepare_for_delete_or_cut( 960 struct reiserfs_transaction_handle *th, 961 struct inode * inode, 962 struct path * p_s_path, 963 const struct cpu_key * p_s_item_key, 964 int * p_n_removed, /* Number of unformatted nodes which were removed 965 from end of the file. */ 966 int * p_n_cut_size, 967 unsigned long long n_new_file_length /* MAX_KEY_OFFSET in case of delete. */ 968 ) { 969 struct super_block * p_s_sb = inode->i_sb; 970 struct item_head * p_le_ih = PATH_PITEM_HEAD(p_s_path); 971 struct buffer_head * p_s_bh = PATH_PLAST_BUFFER(p_s_path); 972 973 BUG_ON (!th->t_trans_id); 974 975 /* Stat_data item. */ 976 if ( is_statdata_le_ih (p_le_ih) ) { 977 978 RFALSE( n_new_file_length != max_reiserfs_offset (inode), 979 "PAP-5210: mode must be M_DELETE"); 980 981 *p_n_cut_size = -(IH_SIZE + ih_item_len(p_le_ih)); 982 return M_DELETE; 983 } 984 985 986 /* Directory item. */ 987 if ( is_direntry_le_ih (p_le_ih) ) 988 return prepare_for_direntry_item (p_s_path, p_le_ih, inode, n_new_file_length, p_n_cut_size); 989 990 /* Direct item. */ 991 if ( is_direct_le_ih (p_le_ih) ) 992 return prepare_for_direct_item (p_s_path, p_le_ih, inode, n_new_file_length, p_n_cut_size); 993 994 995 /* Case of an indirect item. */ 996 { 997 int n_unfm_number, /* Number of the item unformatted nodes. */ 998 n_counter, 999 n_blk_size; 1000 __le32 * p_n_unfm_pointer; /* Pointer to the unformatted node number. */ 1001 __u32 tmp; 1002 struct item_head s_ih; /* Item header. */ 1003 char c_mode; /* Returned mode of the balance. */ 1004 int need_research; 1005 1006 1007 n_blk_size = p_s_sb->s_blocksize; 1008 1009 /* Search for the needed object indirect item until there are no unformatted nodes to be removed. */ 1010 do { 1011 need_research = 0; 1012 p_s_bh = PATH_PLAST_BUFFER(p_s_path); 1013 /* Copy indirect item header to a temp variable. */ 1014 copy_item_head(&s_ih, PATH_PITEM_HEAD(p_s_path)); 1015 /* Calculate number of unformatted nodes in this item. */ 1016 n_unfm_number = I_UNFM_NUM(&s_ih); 1017 1018 RFALSE( ! is_indirect_le_ih(&s_ih) || ! n_unfm_number || 1019 pos_in_item (p_s_path) + 1 != n_unfm_number, 1020 "PAP-5240: invalid item %h " 1021 "n_unfm_number = %d *p_n_pos_in_item = %d", 1022 &s_ih, n_unfm_number, pos_in_item (p_s_path)); 1023 1024 /* Calculate balance mode and position in the item to remove unformatted nodes. */ 1025 if ( n_new_file_length == max_reiserfs_offset (inode) ) {/* Case of delete. */ 1026 pos_in_item (p_s_path) = 0; 1027 *p_n_cut_size = -(IH_SIZE + ih_item_len(&s_ih)); 1028 c_mode = M_DELETE; 1029 } 1030 else { /* Case of truncate. */ 1031 if ( n_new_file_length < le_ih_k_offset (&s_ih) ) { 1032 pos_in_item (p_s_path) = 0; 1033 *p_n_cut_size = -(IH_SIZE + ih_item_len(&s_ih)); 1034 c_mode = M_DELETE; /* Delete this item. */ 1035 } 1036 else { 1037 /* indirect item must be truncated starting from *p_n_pos_in_item-th position */ 1038 pos_in_item (p_s_path) = (n_new_file_length + n_blk_size - le_ih_k_offset (&s_ih) ) >> p_s_sb->s_blocksize_bits; 1039 1040 RFALSE( pos_in_item (p_s_path) > n_unfm_number, 1041 "PAP-5250: invalid position in the item"); 1042 1043 /* Either convert last unformatted node of indirect item to direct item or increase 1044 its free space. */ 1045 if ( pos_in_item (p_s_path) == n_unfm_number ) { 1046 *p_n_cut_size = 0; /* Nothing to cut. */ 1047 return M_CONVERT; /* Maybe convert last unformatted node to the direct item. */ 1048 } 1049 /* Calculate size to cut. */ 1050 *p_n_cut_size = -(ih_item_len(&s_ih) - pos_in_item(p_s_path) * UNFM_P_SIZE); 1051 1052 c_mode = M_CUT; /* Cut from this indirect item. */ 1053 } 1054 } 1055 1056 RFALSE( n_unfm_number <= pos_in_item (p_s_path), 1057 "PAP-5260: invalid position in the indirect item"); 1058 1059 /* pointers to be cut */ 1060 n_unfm_number -= pos_in_item (p_s_path); 1061 /* Set pointer to the last unformatted node pointer that is to be cut. */ 1062 p_n_unfm_pointer = (__le32 *)B_I_PITEM(p_s_bh, &s_ih) + I_UNFM_NUM(&s_ih) - 1 - *p_n_removed; 1063 1064 1065 /* We go through the unformatted nodes pointers of the indirect 1066 item and look for the unformatted nodes in the cache. If we 1067 found some of them we free it, zero corresponding indirect item 1068 entry and log buffer containing that indirect item. For this we 1069 need to prepare last path element for logging. If some 1070 unformatted node has b_count > 1 we must not free this 1071 unformatted node since it is in use. */ 1072 reiserfs_prepare_for_journal(p_s_sb, p_s_bh, 1); 1073 // note: path could be changed, first line in for loop takes care 1074 // of it 1075 1076 for (n_counter = *p_n_removed; 1077 n_counter < n_unfm_number; n_counter++, p_n_unfm_pointer-- ) { 1078 1079 cond_resched(); 1080 if (item_moved (&s_ih, p_s_path)) { 1081 need_research = 1 ; 1082 break; 1083 } 1084 RFALSE( p_n_unfm_pointer < (__le32 *)B_I_PITEM(p_s_bh, &s_ih) || 1085 p_n_unfm_pointer > (__le32 *)B_I_PITEM(p_s_bh, &s_ih) + I_UNFM_NUM(&s_ih) - 1, 1086 "vs-5265: pointer out of range"); 1087 1088 /* Hole, nothing to remove. */ 1089 if ( ! get_block_num(p_n_unfm_pointer,0) ) { 1090 (*p_n_removed)++; 1091 continue; 1092 } 1093 1094 (*p_n_removed)++; 1095 1096 tmp = get_block_num(p_n_unfm_pointer,0); 1097 put_block_num(p_n_unfm_pointer, 0, 0); 1098 journal_mark_dirty (th, p_s_sb, p_s_bh); 1099 reiserfs_free_block(th, inode, tmp, 1); 1100 if ( item_moved (&s_ih, p_s_path) ) { 1101 need_research = 1; 1102 break ; 1103 } 1104 } 1105 1106 /* a trick. If the buffer has been logged, this 1107 ** will do nothing. If we've broken the loop without 1108 ** logging it, it will restore the buffer 1109 ** 1110 */ 1111 reiserfs_restore_prepared_buffer(p_s_sb, p_s_bh); 1112 1113 /* This loop can be optimized. */ 1114 } while ( (*p_n_removed < n_unfm_number || need_research) && 1115 search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path) == POSITION_FOUND ); 1116 1117 RFALSE( *p_n_removed < n_unfm_number, 1118 "PAP-5310: indirect item is not found"); 1119 RFALSE( item_moved (&s_ih, p_s_path), 1120 "after while, comp failed, retry") ; 1121 1122 if (c_mode == M_CUT) 1123 pos_in_item (p_s_path) *= UNFM_P_SIZE; 1124 return c_mode; 1125 } 1126 } 1127 1128 /* Calculate number of bytes which will be deleted or cut during balance */ 1129 static int calc_deleted_bytes_number( 1130 struct tree_balance * p_s_tb, 1131 char c_mode 1132 ) { 1133 int n_del_size; 1134 struct item_head * p_le_ih = PATH_PITEM_HEAD(p_s_tb->tb_path); 1135 1136 if ( is_statdata_le_ih (p_le_ih) ) 1137 return 0; 1138 1139 n_del_size = ( c_mode == M_DELETE ) ? ih_item_len(p_le_ih) : -p_s_tb->insert_size[0]; 1140 if ( is_direntry_le_ih (p_le_ih) ) { 1141 // return EMPTY_DIR_SIZE; /* We delete emty directoris only. */ 1142 // we can't use EMPTY_DIR_SIZE, as old format dirs have a different 1143 // empty size. ick. FIXME, is this right? 1144 // 1145 return n_del_size ; 1146 } 1147 1148 if ( is_indirect_le_ih (p_le_ih) ) 1149 n_del_size = (n_del_size/UNFM_P_SIZE)* 1150 (PATH_PLAST_BUFFER(p_s_tb->tb_path)->b_size);// - get_ih_free_space (p_le_ih); 1151 return n_del_size; 1152 } 1153 1154 static void init_tb_struct( 1155 struct reiserfs_transaction_handle *th, 1156 struct tree_balance * p_s_tb, 1157 struct super_block * p_s_sb, 1158 struct path * p_s_path, 1159 int n_size 1160 ) { 1161 1162 BUG_ON (!th->t_trans_id); 1163 1164 memset (p_s_tb,'\0',sizeof(struct tree_balance)); 1165 p_s_tb->transaction_handle = th ; 1166 p_s_tb->tb_sb = p_s_sb; 1167 p_s_tb->tb_path = p_s_path; 1168 PATH_OFFSET_PBUFFER(p_s_path, ILLEGAL_PATH_ELEMENT_OFFSET) = NULL; 1169 PATH_OFFSET_POSITION(p_s_path, ILLEGAL_PATH_ELEMENT_OFFSET) = 0; 1170 p_s_tb->insert_size[0] = n_size; 1171 } 1172 1173 1174 1175 void padd_item (char * item, int total_length, int length) 1176 { 1177 int i; 1178 1179 for (i = total_length; i > length; ) 1180 item [--i] = 0; 1181 } 1182 1183 #ifdef REISERQUOTA_DEBUG 1184 char key2type(struct reiserfs_key *ih) 1185 { 1186 if (is_direntry_le_key(2, ih)) 1187 return 'd'; 1188 if (is_direct_le_key(2, ih)) 1189 return 'D'; 1190 if (is_indirect_le_key(2, ih)) 1191 return 'i'; 1192 if (is_statdata_le_key(2, ih)) 1193 return 's'; 1194 return 'u'; 1195 } 1196 1197 char head2type(struct item_head *ih) 1198 { 1199 if (is_direntry_le_ih(ih)) 1200 return 'd'; 1201 if (is_direct_le_ih(ih)) 1202 return 'D'; 1203 if (is_indirect_le_ih(ih)) 1204 return 'i'; 1205 if (is_statdata_le_ih(ih)) 1206 return 's'; 1207 return 'u'; 1208 } 1209 #endif 1210 1211 /* Delete object item. */ 1212 int reiserfs_delete_item (struct reiserfs_transaction_handle *th, 1213 struct path * p_s_path, /* Path to the deleted item. */ 1214 const struct cpu_key * p_s_item_key, /* Key to search for the deleted item. */ 1215 struct inode * p_s_inode,/* inode is here just to update i_blocks and quotas */ 1216 struct buffer_head * p_s_un_bh) /* NULL or unformatted node pointer. */ 1217 { 1218 struct super_block * p_s_sb = p_s_inode->i_sb; 1219 struct tree_balance s_del_balance; 1220 struct item_head s_ih; 1221 struct item_head *q_ih; 1222 int quota_cut_bytes; 1223 int n_ret_value, 1224 n_del_size, 1225 n_removed; 1226 1227 #ifdef CONFIG_REISERFS_CHECK 1228 char c_mode; 1229 int n_iter = 0; 1230 #endif 1231 1232 BUG_ON (!th->t_trans_id); 1233 1234 init_tb_struct(th, &s_del_balance, p_s_sb, p_s_path, 0/*size is unknown*/); 1235 1236 while ( 1 ) { 1237 n_removed = 0; 1238 1239 #ifdef CONFIG_REISERFS_CHECK 1240 n_iter++; 1241 c_mode = 1242 #endif 1243 prepare_for_delete_or_cut(th, p_s_inode, p_s_path, p_s_item_key, &n_removed, &n_del_size, max_reiserfs_offset (p_s_inode)); 1244 1245 RFALSE( c_mode != M_DELETE, "PAP-5320: mode must be M_DELETE"); 1246 1247 copy_item_head(&s_ih, PATH_PITEM_HEAD(p_s_path)); 1248 s_del_balance.insert_size[0] = n_del_size; 1249 1250 n_ret_value = fix_nodes(M_DELETE, &s_del_balance, NULL, NULL); 1251 if ( n_ret_value != REPEAT_SEARCH ) 1252 break; 1253 1254 PROC_INFO_INC( p_s_sb, delete_item_restarted ); 1255 1256 // file system changed, repeat search 1257 n_ret_value = search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path); 1258 if (n_ret_value == IO_ERROR) 1259 break; 1260 if (n_ret_value == FILE_NOT_FOUND) { 1261 reiserfs_warning (p_s_sb, "vs-5340: reiserfs_delete_item: " 1262 "no items of the file %K found", p_s_item_key); 1263 break; 1264 } 1265 } /* while (1) */ 1266 1267 if ( n_ret_value != CARRY_ON ) { 1268 unfix_nodes(&s_del_balance); 1269 return 0; 1270 } 1271 1272 // reiserfs_delete_item returns item length when success 1273 n_ret_value = calc_deleted_bytes_number(&s_del_balance, M_DELETE); 1274 q_ih = get_ih(p_s_path) ; 1275 quota_cut_bytes = ih_item_len(q_ih) ; 1276 1277 /* hack so the quota code doesn't have to guess if the file 1278 ** has a tail. On tail insert, we allocate quota for 1 unformatted node. 1279 ** We test the offset because the tail might have been 1280 ** split into multiple items, and we only want to decrement for 1281 ** the unfm node once 1282 */ 1283 if (!S_ISLNK (p_s_inode->i_mode) && is_direct_le_ih(q_ih)) { 1284 if ((le_ih_k_offset(q_ih) & (p_s_sb->s_blocksize - 1)) == 1) { 1285 quota_cut_bytes = p_s_sb->s_blocksize + UNFM_P_SIZE; 1286 } else { 1287 quota_cut_bytes = 0 ; 1288 } 1289 } 1290 1291 if ( p_s_un_bh ) { 1292 int off; 1293 char *data ; 1294 1295 /* We are in direct2indirect conversion, so move tail contents 1296 to the unformatted node */ 1297 /* note, we do the copy before preparing the buffer because we 1298 ** don't care about the contents of the unformatted node yet. 1299 ** the only thing we really care about is the direct item's data 1300 ** is in the unformatted node. 1301 ** 1302 ** Otherwise, we would have to call reiserfs_prepare_for_journal on 1303 ** the unformatted node, which might schedule, meaning we'd have to 1304 ** loop all the way back up to the start of the while loop. 1305 ** 1306 ** The unformatted node must be dirtied later on. We can't be 1307 ** sure here if the entire tail has been deleted yet. 1308 ** 1309 ** p_s_un_bh is from the page cache (all unformatted nodes are 1310 ** from the page cache) and might be a highmem page. So, we 1311 ** can't use p_s_un_bh->b_data. 1312 ** -clm 1313 */ 1314 1315 data = kmap_atomic(p_s_un_bh->b_page, KM_USER0); 1316 off = ((le_ih_k_offset (&s_ih) - 1) & (PAGE_CACHE_SIZE - 1)); 1317 memcpy(data + off, 1318 B_I_PITEM(PATH_PLAST_BUFFER(p_s_path), &s_ih), n_ret_value); 1319 kunmap_atomic(data, KM_USER0); 1320 } 1321 /* Perform balancing after all resources have been collected at once. */ 1322 do_balance(&s_del_balance, NULL, NULL, M_DELETE); 1323 1324 #ifdef REISERQUOTA_DEBUG 1325 reiserfs_debug (p_s_sb, REISERFS_DEBUG_CODE, "reiserquota delete_item(): freeing %u, id=%u type=%c", quota_cut_bytes, p_s_inode->i_uid, head2type(&s_ih)); 1326 #endif 1327 DQUOT_FREE_SPACE_NODIRTY(p_s_inode, quota_cut_bytes); 1328 1329 /* Return deleted body length */ 1330 return n_ret_value; 1331 } 1332 1333 1334 /* Summary Of Mechanisms For Handling Collisions Between Processes: 1335 1336 deletion of the body of the object is performed by iput(), with the 1337 result that if multiple processes are operating on a file, the 1338 deletion of the body of the file is deferred until the last process 1339 that has an open inode performs its iput(). 1340 1341 writes and truncates are protected from collisions by use of 1342 semaphores. 1343 1344 creates, linking, and mknod are protected from collisions with other 1345 processes by making the reiserfs_add_entry() the last step in the 1346 creation, and then rolling back all changes if there was a collision. 1347 - Hans 1348 */ 1349 1350 1351 /* this deletes item which never gets split */ 1352 void reiserfs_delete_solid_item (struct reiserfs_transaction_handle *th, 1353 struct inode *inode, 1354 struct reiserfs_key * key) 1355 { 1356 struct tree_balance tb; 1357 INITIALIZE_PATH (path); 1358 int item_len = 0; 1359 int tb_init = 0 ; 1360 struct cpu_key cpu_key; 1361 int retval; 1362 int quota_cut_bytes = 0; 1363 1364 BUG_ON (!th->t_trans_id); 1365 1366 le_key2cpu_key (&cpu_key, key); 1367 1368 while (1) { 1369 retval = search_item (th->t_super, &cpu_key, &path); 1370 if (retval == IO_ERROR) { 1371 reiserfs_warning (th->t_super, 1372 "vs-5350: reiserfs_delete_solid_item: " 1373 "i/o failure occurred trying to delete %K", 1374 &cpu_key); 1375 break; 1376 } 1377 if (retval != ITEM_FOUND) { 1378 pathrelse (&path); 1379 // No need for a warning, if there is just no free space to insert '..' item into the newly-created subdir 1380 if ( !( (unsigned long long) GET_HASH_VALUE (le_key_k_offset (le_key_version (key), key)) == 0 && \ 1381 (unsigned long long) GET_GENERATION_NUMBER (le_key_k_offset (le_key_version (key), key)) == 1 ) ) 1382 reiserfs_warning (th->t_super, "vs-5355: reiserfs_delete_solid_item: %k not found", key); 1383 break; 1384 } 1385 if (!tb_init) { 1386 tb_init = 1 ; 1387 item_len = ih_item_len( PATH_PITEM_HEAD(&path) ); 1388 init_tb_struct (th, &tb, th->t_super, &path, - (IH_SIZE + item_len)); 1389 } 1390 quota_cut_bytes = ih_item_len(PATH_PITEM_HEAD(&path)) ; 1391 1392 retval = fix_nodes (M_DELETE, &tb, NULL, NULL); 1393 if (retval == REPEAT_SEARCH) { 1394 PROC_INFO_INC( th -> t_super, delete_solid_item_restarted ); 1395 continue; 1396 } 1397 1398 if (retval == CARRY_ON) { 1399 do_balance (&tb, NULL, NULL, M_DELETE); 1400 if (inode) { /* Should we count quota for item? (we don't count quotas for save-links) */ 1401 #ifdef REISERQUOTA_DEBUG 1402 reiserfs_debug (th->t_super, REISERFS_DEBUG_CODE, "reiserquota delete_solid_item(): freeing %u id=%u type=%c", quota_cut_bytes, inode->i_uid, key2type(key)); 1403 #endif 1404 DQUOT_FREE_SPACE_NODIRTY(inode, quota_cut_bytes); 1405 } 1406 break; 1407 } 1408 1409 // IO_ERROR, NO_DISK_SPACE, etc 1410 reiserfs_warning (th->t_super, "vs-5360: reiserfs_delete_solid_item: " 1411 "could not delete %K due to fix_nodes failure", &cpu_key); 1412 unfix_nodes (&tb); 1413 break; 1414 } 1415 1416 reiserfs_check_path(&path) ; 1417 } 1418 1419 1420 int reiserfs_delete_object (struct reiserfs_transaction_handle *th, struct inode * inode) 1421 { 1422 int err; 1423 inode->i_size = 0; 1424 BUG_ON (!th->t_trans_id); 1425 1426 /* for directory this deletes item containing "." and ".." */ 1427 err = reiserfs_do_truncate (th, inode, NULL, 0/*no timestamp updates*/); 1428 if (err) 1429 return err; 1430 1431 #if defined( USE_INODE_GENERATION_COUNTER ) 1432 if( !old_format_only ( th -> t_super ) ) 1433 { 1434 __le32 *inode_generation; 1435 1436 inode_generation = 1437 &REISERFS_SB(th -> t_super) -> s_rs -> s_inode_generation; 1438 *inode_generation = cpu_to_le32( le32_to_cpu( *inode_generation ) + 1 ); 1439 } 1440 /* USE_INODE_GENERATION_COUNTER */ 1441 #endif 1442 reiserfs_delete_solid_item (th, inode, INODE_PKEY (inode)); 1443 1444 return err; 1445 } 1446 1447 static void 1448 unmap_buffers(struct page *page, loff_t pos) { 1449 struct buffer_head *bh ; 1450 struct buffer_head *head ; 1451 struct buffer_head *next ; 1452 unsigned long tail_index ; 1453 unsigned long cur_index ; 1454 1455 if (page) { 1456 if (page_has_buffers(page)) { 1457 tail_index = pos & (PAGE_CACHE_SIZE - 1) ; 1458 cur_index = 0 ; 1459 head = page_buffers(page) ; 1460 bh = head ; 1461 do { 1462 next = bh->b_this_page ; 1463 1464 /* we want to unmap the buffers that contain the tail, and 1465 ** all the buffers after it (since the tail must be at the 1466 ** end of the file). We don't want to unmap file data 1467 ** before the tail, since it might be dirty and waiting to 1468 ** reach disk 1469 */ 1470 cur_index += bh->b_size ; 1471 if (cur_index > tail_index) { 1472 reiserfs_unmap_buffer(bh) ; 1473 } 1474 bh = next ; 1475 } while (bh != head) ; 1476 if ( PAGE_SIZE == bh->b_size ) { 1477 clear_page_dirty(page); 1478 } 1479 } 1480 } 1481 } 1482 1483 static int maybe_indirect_to_direct (struct reiserfs_transaction_handle *th, 1484 struct inode * p_s_inode, 1485 struct page *page, 1486 struct path * p_s_path, 1487 const struct cpu_key * p_s_item_key, 1488 loff_t n_new_file_size, 1489 char * p_c_mode 1490 ) { 1491 struct super_block * p_s_sb = p_s_inode->i_sb; 1492 int n_block_size = p_s_sb->s_blocksize; 1493 int cut_bytes; 1494 BUG_ON (!th->t_trans_id); 1495 1496 if (n_new_file_size != p_s_inode->i_size) 1497 BUG (); 1498 1499 /* the page being sent in could be NULL if there was an i/o error 1500 ** reading in the last block. The user will hit problems trying to 1501 ** read the file, but for now we just skip the indirect2direct 1502 */ 1503 if (atomic_read(&p_s_inode->i_count) > 1 || 1504 !tail_has_to_be_packed (p_s_inode) || 1505 !page || (REISERFS_I(p_s_inode)->i_flags & i_nopack_mask)) { 1506 // leave tail in an unformatted node 1507 *p_c_mode = M_SKIP_BALANCING; 1508 cut_bytes = n_block_size - (n_new_file_size & (n_block_size - 1)); 1509 pathrelse(p_s_path); 1510 return cut_bytes; 1511 } 1512 /* Permorm the conversion to a direct_item. */ 1513 /*return indirect_to_direct (p_s_inode, p_s_path, p_s_item_key, n_new_file_size, p_c_mode);*/ 1514 return indirect2direct (th, p_s_inode, page, p_s_path, p_s_item_key, n_new_file_size, p_c_mode); 1515 } 1516 1517 1518 /* we did indirect_to_direct conversion. And we have inserted direct 1519 item successesfully, but there were no disk space to cut unfm 1520 pointer being converted. Therefore we have to delete inserted 1521 direct item(s) */ 1522 static void indirect_to_direct_roll_back (struct reiserfs_transaction_handle *th, struct inode * inode, struct path * path) 1523 { 1524 struct cpu_key tail_key; 1525 int tail_len; 1526 int removed; 1527 BUG_ON (!th->t_trans_id); 1528 1529 make_cpu_key (&tail_key, inode, inode->i_size + 1, TYPE_DIRECT, 4);// !!!! 1530 tail_key.key_length = 4; 1531 1532 tail_len = (cpu_key_k_offset (&tail_key) & (inode->i_sb->s_blocksize - 1)) - 1; 1533 while (tail_len) { 1534 /* look for the last byte of the tail */ 1535 if (search_for_position_by_key (inode->i_sb, &tail_key, path) == POSITION_NOT_FOUND) 1536 reiserfs_panic (inode->i_sb, "vs-5615: indirect_to_direct_roll_back: found invalid item"); 1537 RFALSE( path->pos_in_item != ih_item_len(PATH_PITEM_HEAD (path)) - 1, 1538 "vs-5616: appended bytes found"); 1539 PATH_LAST_POSITION (path) --; 1540 1541 removed = reiserfs_delete_item (th, path, &tail_key, inode, NULL/*unbh not needed*/); 1542 RFALSE( removed <= 0 || removed > tail_len, 1543 "vs-5617: there was tail %d bytes, removed item length %d bytes", 1544 tail_len, removed); 1545 tail_len -= removed; 1546 set_cpu_key_k_offset (&tail_key, cpu_key_k_offset (&tail_key) - removed); 1547 } 1548 reiserfs_warning (inode->i_sb, "indirect_to_direct_roll_back: indirect_to_direct conversion has been rolled back due to lack of disk space"); 1549 //mark_file_without_tail (inode); 1550 mark_inode_dirty (inode); 1551 } 1552 1553 1554 /* (Truncate or cut entry) or delete object item. Returns < 0 on failure */ 1555 int reiserfs_cut_from_item (struct reiserfs_transaction_handle *th, 1556 struct path * p_s_path, 1557 struct cpu_key * p_s_item_key, 1558 struct inode * p_s_inode, 1559 struct page *page, 1560 loff_t n_new_file_size) 1561 { 1562 struct super_block * p_s_sb = p_s_inode->i_sb; 1563 /* Every function which is going to call do_balance must first 1564 create a tree_balance structure. Then it must fill up this 1565 structure by using the init_tb_struct and fix_nodes functions. 1566 After that we can make tree balancing. */ 1567 struct tree_balance s_cut_balance; 1568 struct item_head *p_le_ih; 1569 int n_cut_size = 0, /* Amount to be cut. */ 1570 n_ret_value = CARRY_ON, 1571 n_removed = 0, /* Number of the removed unformatted nodes. */ 1572 n_is_inode_locked = 0; 1573 char c_mode; /* Mode of the balance. */ 1574 int retval2 = -1; 1575 int quota_cut_bytes; 1576 loff_t tail_pos = 0; 1577 1578 BUG_ON (!th->t_trans_id); 1579 1580 init_tb_struct(th, &s_cut_balance, p_s_inode->i_sb, p_s_path, n_cut_size); 1581 1582 1583 /* Repeat this loop until we either cut the item without needing 1584 to balance, or we fix_nodes without schedule occurring */ 1585 while ( 1 ) { 1586 /* Determine the balance mode, position of the first byte to 1587 be cut, and size to be cut. In case of the indirect item 1588 free unformatted nodes which are pointed to by the cut 1589 pointers. */ 1590 1591 c_mode = prepare_for_delete_or_cut(th, p_s_inode, p_s_path, p_s_item_key, &n_removed, 1592 &n_cut_size, n_new_file_size); 1593 if ( c_mode == M_CONVERT ) { 1594 /* convert last unformatted node to direct item or leave 1595 tail in the unformatted node */ 1596 RFALSE( n_ret_value != CARRY_ON, "PAP-5570: can not convert twice"); 1597 1598 n_ret_value = maybe_indirect_to_direct (th, p_s_inode, page, p_s_path, p_s_item_key, 1599 n_new_file_size, &c_mode); 1600 if ( c_mode == M_SKIP_BALANCING ) 1601 /* tail has been left in the unformatted node */ 1602 return n_ret_value; 1603 1604 n_is_inode_locked = 1; 1605 1606 /* removing of last unformatted node will change value we 1607 have to return to truncate. Save it */ 1608 retval2 = n_ret_value; 1609 /*retval2 = p_s_sb->s_blocksize - (n_new_file_size & (p_s_sb->s_blocksize - 1));*/ 1610 1611 /* So, we have performed the first part of the conversion: 1612 inserting the new direct item. Now we are removing the 1613 last unformatted node pointer. Set key to search for 1614 it. */ 1615 set_cpu_key_k_type (p_s_item_key, TYPE_INDIRECT); 1616 p_s_item_key->key_length = 4; 1617 n_new_file_size -= (n_new_file_size & (p_s_sb->s_blocksize - 1)); 1618 tail_pos = n_new_file_size; 1619 set_cpu_key_k_offset (p_s_item_key, n_new_file_size + 1); 1620 if ( search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path) == POSITION_NOT_FOUND ){ 1621 print_block (PATH_PLAST_BUFFER (p_s_path), 3, PATH_LAST_POSITION (p_s_path) - 1, PATH_LAST_POSITION (p_s_path) + 1); 1622 reiserfs_panic(p_s_sb, "PAP-5580: reiserfs_cut_from_item: item to convert does not exist (%K)", p_s_item_key); 1623 } 1624 continue; 1625 } 1626 if (n_cut_size == 0) { 1627 pathrelse (p_s_path); 1628 return 0; 1629 } 1630 1631 s_cut_balance.insert_size[0] = n_cut_size; 1632 1633 n_ret_value = fix_nodes(c_mode, &s_cut_balance, NULL, NULL); 1634 if ( n_ret_value != REPEAT_SEARCH ) 1635 break; 1636 1637 PROC_INFO_INC( p_s_sb, cut_from_item_restarted ); 1638 1639 n_ret_value = search_for_position_by_key(p_s_sb, p_s_item_key, p_s_path); 1640 if (n_ret_value == POSITION_FOUND) 1641 continue; 1642 1643 reiserfs_warning (p_s_sb, "PAP-5610: reiserfs_cut_from_item: item %K not found", p_s_item_key); 1644 unfix_nodes (&s_cut_balance); 1645 return (n_ret_value == IO_ERROR) ? -EIO : -ENOENT; 1646 } /* while */ 1647 1648 // check fix_nodes results (IO_ERROR or NO_DISK_SPACE) 1649 if ( n_ret_value != CARRY_ON ) { 1650 if ( n_is_inode_locked ) { 1651 // FIXME: this seems to be not needed: we are always able 1652 // to cut item 1653 indirect_to_direct_roll_back (th, p_s_inode, p_s_path); 1654 } 1655 if (n_ret_value == NO_DISK_SPACE) 1656 reiserfs_warning (p_s_sb, "NO_DISK_SPACE"); 1657 unfix_nodes (&s_cut_balance); 1658 return -EIO; 1659 } 1660 1661 /* go ahead and perform balancing */ 1662 1663 RFALSE( c_mode == M_PASTE || c_mode == M_INSERT, "invalid mode"); 1664 1665 /* Calculate number of bytes that need to be cut from the item. */ 1666 quota_cut_bytes = ( c_mode == M_DELETE ) ? ih_item_len(get_ih(p_s_path)) : -s_cut_balance.insert_size[0]; 1667 if (retval2 == -1) 1668 n_ret_value = calc_deleted_bytes_number(&s_cut_balance, c_mode); 1669 else 1670 n_ret_value = retval2; 1671 1672 1673 /* For direct items, we only change the quota when deleting the last 1674 ** item. 1675 */ 1676 p_le_ih = PATH_PITEM_HEAD (s_cut_balance.tb_path); 1677 if (!S_ISLNK (p_s_inode->i_mode) && is_direct_le_ih(p_le_ih)) { 1678 if (c_mode == M_DELETE && 1679 (le_ih_k_offset (p_le_ih) & (p_s_sb->s_blocksize - 1)) == 1 ) { 1680 // FIXME: this is to keep 3.5 happy 1681 REISERFS_I(p_s_inode)->i_first_direct_byte = U32_MAX; 1682 quota_cut_bytes = p_s_sb->s_blocksize + UNFM_P_SIZE ; 1683 } else { 1684 quota_cut_bytes = 0 ; 1685 } 1686 } 1687 #ifdef CONFIG_REISERFS_CHECK 1688 if (n_is_inode_locked) { 1689 struct item_head * le_ih = PATH_PITEM_HEAD (s_cut_balance.tb_path); 1690 /* we are going to complete indirect2direct conversion. Make 1691 sure, that we exactly remove last unformatted node pointer 1692 of the item */ 1693 if (!is_indirect_le_ih (le_ih)) 1694 reiserfs_panic (p_s_sb, "vs-5652: reiserfs_cut_from_item: " 1695 "item must be indirect %h", le_ih); 1696 1697 if (c_mode == M_DELETE && ih_item_len(le_ih) != UNFM_P_SIZE) 1698 reiserfs_panic (p_s_sb, "vs-5653: reiserfs_cut_from_item: " 1699 "completing indirect2direct conversion indirect item %h " 1700 "being deleted must be of 4 byte long", le_ih); 1701 1702 if (c_mode == M_CUT && s_cut_balance.insert_size[0] != -UNFM_P_SIZE) { 1703 reiserfs_panic (p_s_sb, "vs-5654: reiserfs_cut_from_item: " 1704 "can not complete indirect2direct conversion of %h (CUT, insert_size==%d)", 1705 le_ih, s_cut_balance.insert_size[0]); 1706 } 1707 /* it would be useful to make sure, that right neighboring 1708 item is direct item of this file */ 1709 } 1710 #endif 1711 1712 do_balance(&s_cut_balance, NULL, NULL, c_mode); 1713 if ( n_is_inode_locked ) { 1714 /* we've done an indirect->direct conversion. when the data block 1715 ** was freed, it was removed from the list of blocks that must 1716 ** be flushed before the transaction commits, make sure to 1717 ** unmap and invalidate it 1718 */ 1719 unmap_buffers(page, tail_pos); 1720 REISERFS_I(p_s_inode)->i_flags &= ~i_pack_on_close_mask ; 1721 } 1722 #ifdef REISERQUOTA_DEBUG 1723 reiserfs_debug (p_s_inode->i_sb, REISERFS_DEBUG_CODE, "reiserquota cut_from_item(): freeing %u id=%u type=%c", quota_cut_bytes, p_s_inode->i_uid, '?'); 1724 #endif 1725 DQUOT_FREE_SPACE_NODIRTY(p_s_inode, quota_cut_bytes); 1726 return n_ret_value; 1727 } 1728 1729 static void truncate_directory (struct reiserfs_transaction_handle *th, struct inode * inode) 1730 { 1731 BUG_ON (!th->t_trans_id); 1732 if (inode->i_nlink) 1733 reiserfs_warning (inode->i_sb, 1734 "vs-5655: truncate_directory: link count != 0"); 1735 1736 set_le_key_k_offset (KEY_FORMAT_3_5, INODE_PKEY (inode), DOT_OFFSET); 1737 set_le_key_k_type (KEY_FORMAT_3_5, INODE_PKEY (inode), TYPE_DIRENTRY); 1738 reiserfs_delete_solid_item (th, inode, INODE_PKEY (inode)); 1739 reiserfs_update_sd(th, inode) ; 1740 set_le_key_k_offset (KEY_FORMAT_3_5, INODE_PKEY (inode), SD_OFFSET); 1741 set_le_key_k_type (KEY_FORMAT_3_5, INODE_PKEY (inode), TYPE_STAT_DATA); 1742 } 1743 1744 1745 1746 1747 /* Truncate file to the new size. Note, this must be called with a transaction 1748 already started */ 1749 int reiserfs_do_truncate (struct reiserfs_transaction_handle *th, 1750 struct inode * p_s_inode, /* ->i_size contains new 1751 size */ 1752 struct page *page, /* up to date for last block */ 1753 int update_timestamps /* when it is called by 1754 file_release to convert 1755 the tail - no timestamps 1756 should be updated */ 1757 ) { 1758 INITIALIZE_PATH (s_search_path); /* Path to the current object item. */ 1759 struct item_head * p_le_ih; /* Pointer to an item header. */ 1760 struct cpu_key s_item_key; /* Key to search for a previous file item. */ 1761 loff_t n_file_size, /* Old file size. */ 1762 n_new_file_size;/* New file size. */ 1763 int n_deleted; /* Number of deleted or truncated bytes. */ 1764 int retval; 1765 int err = 0; 1766 1767 BUG_ON (!th->t_trans_id); 1768 if ( ! (S_ISREG(p_s_inode->i_mode) || S_ISDIR(p_s_inode->i_mode) || S_ISLNK(p_s_inode->i_mode)) ) 1769 return 0; 1770 1771 if (S_ISDIR(p_s_inode->i_mode)) { 1772 // deletion of directory - no need to update timestamps 1773 truncate_directory (th, p_s_inode); 1774 return 0; 1775 } 1776 1777 /* Get new file size. */ 1778 n_new_file_size = p_s_inode->i_size; 1779 1780 // FIXME: note, that key type is unimportant here 1781 make_cpu_key (&s_item_key, p_s_inode, max_reiserfs_offset (p_s_inode), TYPE_DIRECT, 3); 1782 1783 retval = search_for_position_by_key(p_s_inode->i_sb, &s_item_key, &s_search_path); 1784 if (retval == IO_ERROR) { 1785 reiserfs_warning (p_s_inode->i_sb, "vs-5657: reiserfs_do_truncate: " 1786 "i/o failure occurred trying to truncate %K", &s_item_key); 1787 err = -EIO; 1788 goto out; 1789 } 1790 if (retval == POSITION_FOUND || retval == FILE_NOT_FOUND) { 1791 reiserfs_warning (p_s_inode->i_sb, "PAP-5660: reiserfs_do_truncate: " 1792 "wrong result %d of search for %K", retval, &s_item_key); 1793 1794 err = -EIO; 1795 goto out; 1796 } 1797 1798 s_search_path.pos_in_item --; 1799 1800 /* Get real file size (total length of all file items) */ 1801 p_le_ih = PATH_PITEM_HEAD(&s_search_path); 1802 if ( is_statdata_le_ih (p_le_ih) ) 1803 n_file_size = 0; 1804 else { 1805 loff_t offset = le_ih_k_offset (p_le_ih); 1806 int bytes = op_bytes_number (p_le_ih,p_s_inode->i_sb->s_blocksize); 1807 1808 /* this may mismatch with real file size: if last direct item 1809 had no padding zeros and last unformatted node had no free 1810 space, this file would have this file size */ 1811 n_file_size = offset + bytes - 1; 1812 } 1813 /* 1814 * are we doing a full truncate or delete, if so 1815 * kick in the reada code 1816 */ 1817 if (n_new_file_size == 0) 1818 s_search_path.reada = PATH_READA | PATH_READA_BACK; 1819 1820 if ( n_file_size == 0 || n_file_size < n_new_file_size ) { 1821 goto update_and_out ; 1822 } 1823 1824 /* Update key to search for the last file item. */ 1825 set_cpu_key_k_offset (&s_item_key, n_file_size); 1826 1827 do { 1828 /* Cut or delete file item. */ 1829 n_deleted = reiserfs_cut_from_item(th, &s_search_path, &s_item_key, p_s_inode, page, n_new_file_size); 1830 if (n_deleted < 0) { 1831 reiserfs_warning (p_s_inode->i_sb, "vs-5665: reiserfs_do_truncate: reiserfs_cut_from_item failed"); 1832 reiserfs_check_path(&s_search_path) ; 1833 return 0; 1834 } 1835 1836 RFALSE( n_deleted > n_file_size, 1837 "PAP-5670: reiserfs_cut_from_item: too many bytes deleted: deleted %d, file_size %lu, item_key %K", 1838 n_deleted, n_file_size, &s_item_key); 1839 1840 /* Change key to search the last file item. */ 1841 n_file_size -= n_deleted; 1842 1843 set_cpu_key_k_offset (&s_item_key, n_file_size); 1844 1845 /* While there are bytes to truncate and previous file item is presented in the tree. */ 1846 1847 /* 1848 ** This loop could take a really long time, and could log 1849 ** many more blocks than a transaction can hold. So, we do a polite 1850 ** journal end here, and if the transaction needs ending, we make 1851 ** sure the file is consistent before ending the current trans 1852 ** and starting a new one 1853 */ 1854 if (journal_transaction_should_end(th, th->t_blocks_allocated)) { 1855 int orig_len_alloc = th->t_blocks_allocated ; 1856 decrement_counters_in_path(&s_search_path) ; 1857 1858 if (update_timestamps) { 1859 p_s_inode->i_mtime = p_s_inode->i_ctime = CURRENT_TIME_SEC; 1860 } 1861 reiserfs_update_sd(th, p_s_inode) ; 1862 1863 err = journal_end(th, p_s_inode->i_sb, orig_len_alloc) ; 1864 if (err) 1865 goto out; 1866 err = journal_begin (th, p_s_inode->i_sb, 1867 JOURNAL_PER_BALANCE_CNT * 6); 1868 if (err) 1869 goto out; 1870 reiserfs_update_inode_transaction(p_s_inode) ; 1871 } 1872 } while ( n_file_size > ROUND_UP (n_new_file_size) && 1873 search_for_position_by_key(p_s_inode->i_sb, &s_item_key, &s_search_path) == POSITION_FOUND ) ; 1874 1875 RFALSE( n_file_size > ROUND_UP (n_new_file_size), 1876 "PAP-5680: truncate did not finish: new_file_size %Ld, current %Ld, oid %d", 1877 n_new_file_size, n_file_size, s_item_key.on_disk_key.k_objectid); 1878 1879 update_and_out: 1880 if (update_timestamps) { 1881 // this is truncate, not file closing 1882 p_s_inode->i_mtime = p_s_inode->i_ctime = CURRENT_TIME_SEC; 1883 } 1884 reiserfs_update_sd (th, p_s_inode); 1885 1886 out: 1887 pathrelse(&s_search_path) ; 1888 return err; 1889 } 1890 1891 1892 #ifdef CONFIG_REISERFS_CHECK 1893 // this makes sure, that we __append__, not overwrite or add holes 1894 static void check_research_for_paste (struct path * path, 1895 const struct cpu_key * p_s_key) 1896 { 1897 struct item_head * found_ih = get_ih (path); 1898 1899 if (is_direct_le_ih (found_ih)) { 1900 if (le_ih_k_offset (found_ih) + op_bytes_number (found_ih, get_last_bh (path)->b_size) != 1901 cpu_key_k_offset (p_s_key) || 1902 op_bytes_number (found_ih, get_last_bh (path)->b_size) != pos_in_item (path)) 1903 reiserfs_panic (NULL, "PAP-5720: check_research_for_paste: " 1904 "found direct item %h or position (%d) does not match to key %K", 1905 found_ih, pos_in_item (path), p_s_key); 1906 } 1907 if (is_indirect_le_ih (found_ih)) { 1908 if (le_ih_k_offset (found_ih) + op_bytes_number (found_ih, get_last_bh (path)->b_size) != cpu_key_k_offset (p_s_key) || 1909 I_UNFM_NUM (found_ih) != pos_in_item (path) || 1910 get_ih_free_space (found_ih) != 0) 1911 reiserfs_panic (NULL, "PAP-5730: check_research_for_paste: " 1912 "found indirect item (%h) or position (%d) does not match to key (%K)", 1913 found_ih, pos_in_item (path), p_s_key); 1914 } 1915 } 1916 #endif /* config reiserfs check */ 1917 1918 1919 /* Paste bytes to the existing item. Returns bytes number pasted into the item. */ 1920 int reiserfs_paste_into_item (struct reiserfs_transaction_handle *th, 1921 struct path * p_s_search_path, /* Path to the pasted item. */ 1922 const struct cpu_key * p_s_key, /* Key to search for the needed item.*/ 1923 struct inode * inode, /* Inode item belongs to */ 1924 const char * p_c_body, /* Pointer to the bytes to paste. */ 1925 int n_pasted_size) /* Size of pasted bytes. */ 1926 { 1927 struct tree_balance s_paste_balance; 1928 int retval; 1929 int fs_gen; 1930 1931 BUG_ON (!th->t_trans_id); 1932 1933 fs_gen = get_generation(inode->i_sb) ; 1934 1935 #ifdef REISERQUOTA_DEBUG 1936 reiserfs_debug (inode->i_sb, REISERFS_DEBUG_CODE, "reiserquota paste_into_item(): allocating %u id=%u type=%c", n_pasted_size, inode->i_uid, key2type(&(p_s_key->on_disk_key))); 1937 #endif 1938 1939 if (DQUOT_ALLOC_SPACE_NODIRTY(inode, n_pasted_size)) { 1940 pathrelse(p_s_search_path); 1941 return -EDQUOT; 1942 } 1943 init_tb_struct(th, &s_paste_balance, th->t_super, p_s_search_path, n_pasted_size); 1944 #ifdef DISPLACE_NEW_PACKING_LOCALITIES 1945 s_paste_balance.key = p_s_key->on_disk_key; 1946 #endif 1947 1948 /* DQUOT_* can schedule, must check before the fix_nodes */ 1949 if (fs_changed(fs_gen, inode->i_sb)) { 1950 goto search_again; 1951 } 1952 1953 while ((retval = fix_nodes(M_PASTE, &s_paste_balance, NULL, p_c_body)) == 1954 REPEAT_SEARCH ) { 1955 search_again: 1956 /* file system changed while we were in the fix_nodes */ 1957 PROC_INFO_INC( th -> t_super, paste_into_item_restarted ); 1958 retval = search_for_position_by_key (th->t_super, p_s_key, p_s_search_path); 1959 if (retval == IO_ERROR) { 1960 retval = -EIO ; 1961 goto error_out ; 1962 } 1963 if (retval == POSITION_FOUND) { 1964 reiserfs_warning (inode->i_sb, "PAP-5710: reiserfs_paste_into_item: entry or pasted byte (%K) exists", p_s_key); 1965 retval = -EEXIST ; 1966 goto error_out ; 1967 } 1968 1969 #ifdef CONFIG_REISERFS_CHECK 1970 check_research_for_paste (p_s_search_path, p_s_key); 1971 #endif 1972 } 1973 1974 /* Perform balancing after all resources are collected by fix_nodes, and 1975 accessing them will not risk triggering schedule. */ 1976 if ( retval == CARRY_ON ) { 1977 do_balance(&s_paste_balance, NULL/*ih*/, p_c_body, M_PASTE); 1978 return 0; 1979 } 1980 retval = (retval == NO_DISK_SPACE) ? -ENOSPC : -EIO; 1981 error_out: 1982 /* this also releases the path */ 1983 unfix_nodes(&s_paste_balance); 1984 #ifdef REISERQUOTA_DEBUG 1985 reiserfs_debug (inode->i_sb, REISERFS_DEBUG_CODE, "reiserquota paste_into_item(): freeing %u id=%u type=%c", n_pasted_size, inode->i_uid, key2type(&(p_s_key->on_disk_key))); 1986 #endif 1987 DQUOT_FREE_SPACE_NODIRTY(inode, n_pasted_size); 1988 return retval ; 1989 } 1990 1991 1992 /* Insert new item into the buffer at the path. */ 1993 int reiserfs_insert_item(struct reiserfs_transaction_handle *th, 1994 struct path * p_s_path, /* Path to the inserteded item. */ 1995 const struct cpu_key * key, 1996 struct item_head * p_s_ih, /* Pointer to the item header to insert.*/ 1997 struct inode * inode, 1998 const char * p_c_body) /* Pointer to the bytes to insert. */ 1999 { 2000 struct tree_balance s_ins_balance; 2001 int retval; 2002 int fs_gen = 0 ; 2003 int quota_bytes = 0 ; 2004 2005 BUG_ON (!th->t_trans_id); 2006 2007 if (inode) { /* Do we count quotas for item? */ 2008 fs_gen = get_generation(inode->i_sb); 2009 quota_bytes = ih_item_len(p_s_ih); 2010 2011 /* hack so the quota code doesn't have to guess if the file has 2012 ** a tail, links are always tails, so there's no guessing needed 2013 */ 2014 if (!S_ISLNK (inode->i_mode) && is_direct_le_ih(p_s_ih)) { 2015 quota_bytes = inode->i_sb->s_blocksize + UNFM_P_SIZE ; 2016 } 2017 #ifdef REISERQUOTA_DEBUG 2018 reiserfs_debug (inode->i_sb, REISERFS_DEBUG_CODE, "reiserquota insert_item(): allocating %u id=%u type=%c", quota_bytes, inode->i_uid, head2type(p_s_ih)); 2019 #endif 2020 /* We can't dirty inode here. It would be immediately written but 2021 * appropriate stat item isn't inserted yet... */ 2022 if (DQUOT_ALLOC_SPACE_NODIRTY(inode, quota_bytes)) { 2023 pathrelse(p_s_path); 2024 return -EDQUOT; 2025 } 2026 } 2027 init_tb_struct(th, &s_ins_balance, th->t_super, p_s_path, IH_SIZE + ih_item_len(p_s_ih)); 2028 #ifdef DISPLACE_NEW_PACKING_LOCALITIES 2029 s_ins_balance.key = key->on_disk_key; 2030 #endif 2031 /* DQUOT_* can schedule, must check to be sure calling fix_nodes is safe */ 2032 if (inode && fs_changed(fs_gen, inode->i_sb)) { 2033 goto search_again; 2034 } 2035 2036 while ( (retval = fix_nodes(M_INSERT, &s_ins_balance, p_s_ih, p_c_body)) == REPEAT_SEARCH) { 2037 search_again: 2038 /* file system changed while we were in the fix_nodes */ 2039 PROC_INFO_INC( th -> t_super, insert_item_restarted ); 2040 retval = search_item (th->t_super, key, p_s_path); 2041 if (retval == IO_ERROR) { 2042 retval = -EIO; 2043 goto error_out ; 2044 } 2045 if (retval == ITEM_FOUND) { 2046 reiserfs_warning (th->t_super, "PAP-5760: reiserfs_insert_item: " 2047 "key %K already exists in the tree", key); 2048 retval = -EEXIST ; 2049 goto error_out; 2050 } 2051 } 2052 2053 /* make balancing after all resources will be collected at a time */ 2054 if ( retval == CARRY_ON ) { 2055 do_balance (&s_ins_balance, p_s_ih, p_c_body, M_INSERT); 2056 return 0; 2057 } 2058 2059 retval = (retval == NO_DISK_SPACE) ? -ENOSPC : -EIO; 2060 error_out: 2061 /* also releases the path */ 2062 unfix_nodes(&s_ins_balance); 2063 #ifdef REISERQUOTA_DEBUG 2064 reiserfs_debug (th->t_super, REISERFS_DEBUG_CODE, "reiserquota insert_item(): freeing %u id=%u type=%c", quota_bytes, inode->i_uid, head2type(p_s_ih)); 2065 #endif 2066 if (inode) 2067 DQUOT_FREE_SPACE_NODIRTY(inode, quota_bytes) ; 2068 return retval; 2069 } 2070 2071 2072 2073 2074