1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2007,2008 Oracle. All rights reserved. 4 */ 5 6 #include <linux/sched.h> 7 #include <linux/slab.h> 8 #include <linux/rbtree.h> 9 #include <linux/mm.h> 10 #include "ctree.h" 11 #include "disk-io.h" 12 #include "transaction.h" 13 #include "print-tree.h" 14 #include "locking.h" 15 #include "volumes.h" 16 #include "qgroup.h" 17 18 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root 19 *root, struct btrfs_path *path, int level); 20 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root, 21 const struct btrfs_key *ins_key, struct btrfs_path *path, 22 int data_size, int extend); 23 static int push_node_left(struct btrfs_trans_handle *trans, 24 struct extent_buffer *dst, 25 struct extent_buffer *src, int empty); 26 static int balance_node_right(struct btrfs_trans_handle *trans, 27 struct extent_buffer *dst_buf, 28 struct extent_buffer *src_buf); 29 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path, 30 int level, int slot); 31 32 static const struct btrfs_csums { 33 u16 size; 34 const char name[10]; 35 const char driver[12]; 36 } btrfs_csums[] = { 37 [BTRFS_CSUM_TYPE_CRC32] = { .size = 4, .name = "crc32c" }, 38 [BTRFS_CSUM_TYPE_XXHASH] = { .size = 8, .name = "xxhash64" }, 39 [BTRFS_CSUM_TYPE_SHA256] = { .size = 32, .name = "sha256" }, 40 [BTRFS_CSUM_TYPE_BLAKE2] = { .size = 32, .name = "blake2b", 41 .driver = "blake2b-256" }, 42 }; 43 44 int btrfs_super_csum_size(const struct btrfs_super_block *s) 45 { 46 u16 t = btrfs_super_csum_type(s); 47 /* 48 * csum type is validated at mount time 49 */ 50 return btrfs_csums[t].size; 51 } 52 53 const char *btrfs_super_csum_name(u16 csum_type) 54 { 55 /* csum type is validated at mount time */ 56 return btrfs_csums[csum_type].name; 57 } 58 59 /* 60 * Return driver name if defined, otherwise the name that's also a valid driver 61 * name 62 */ 63 const char *btrfs_super_csum_driver(u16 csum_type) 64 { 65 /* csum type is validated at mount time */ 66 return btrfs_csums[csum_type].driver[0] ? 67 btrfs_csums[csum_type].driver : 68 btrfs_csums[csum_type].name; 69 } 70 71 size_t __const btrfs_get_num_csums(void) 72 { 73 return ARRAY_SIZE(btrfs_csums); 74 } 75 76 struct btrfs_path *btrfs_alloc_path(void) 77 { 78 return kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS); 79 } 80 81 /* this also releases the path */ 82 void btrfs_free_path(struct btrfs_path *p) 83 { 84 if (!p) 85 return; 86 btrfs_release_path(p); 87 kmem_cache_free(btrfs_path_cachep, p); 88 } 89 90 /* 91 * path release drops references on the extent buffers in the path 92 * and it drops any locks held by this path 93 * 94 * It is safe to call this on paths that no locks or extent buffers held. 95 */ 96 noinline void btrfs_release_path(struct btrfs_path *p) 97 { 98 int i; 99 100 for (i = 0; i < BTRFS_MAX_LEVEL; i++) { 101 p->slots[i] = 0; 102 if (!p->nodes[i]) 103 continue; 104 if (p->locks[i]) { 105 btrfs_tree_unlock_rw(p->nodes[i], p->locks[i]); 106 p->locks[i] = 0; 107 } 108 free_extent_buffer(p->nodes[i]); 109 p->nodes[i] = NULL; 110 } 111 } 112 113 /* 114 * safely gets a reference on the root node of a tree. A lock 115 * is not taken, so a concurrent writer may put a different node 116 * at the root of the tree. See btrfs_lock_root_node for the 117 * looping required. 118 * 119 * The extent buffer returned by this has a reference taken, so 120 * it won't disappear. It may stop being the root of the tree 121 * at any time because there are no locks held. 122 */ 123 struct extent_buffer *btrfs_root_node(struct btrfs_root *root) 124 { 125 struct extent_buffer *eb; 126 127 while (1) { 128 rcu_read_lock(); 129 eb = rcu_dereference(root->node); 130 131 /* 132 * RCU really hurts here, we could free up the root node because 133 * it was COWed but we may not get the new root node yet so do 134 * the inc_not_zero dance and if it doesn't work then 135 * synchronize_rcu and try again. 136 */ 137 if (atomic_inc_not_zero(&eb->refs)) { 138 rcu_read_unlock(); 139 break; 140 } 141 rcu_read_unlock(); 142 synchronize_rcu(); 143 } 144 return eb; 145 } 146 147 /* 148 * Cowonly root (not-shareable trees, everything not subvolume or reloc roots), 149 * just get put onto a simple dirty list. Transaction walks this list to make 150 * sure they get properly updated on disk. 151 */ 152 static void add_root_to_dirty_list(struct btrfs_root *root) 153 { 154 struct btrfs_fs_info *fs_info = root->fs_info; 155 156 if (test_bit(BTRFS_ROOT_DIRTY, &root->state) || 157 !test_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state)) 158 return; 159 160 spin_lock(&fs_info->trans_lock); 161 if (!test_and_set_bit(BTRFS_ROOT_DIRTY, &root->state)) { 162 /* Want the extent tree to be the last on the list */ 163 if (root->root_key.objectid == BTRFS_EXTENT_TREE_OBJECTID) 164 list_move_tail(&root->dirty_list, 165 &fs_info->dirty_cowonly_roots); 166 else 167 list_move(&root->dirty_list, 168 &fs_info->dirty_cowonly_roots); 169 } 170 spin_unlock(&fs_info->trans_lock); 171 } 172 173 /* 174 * used by snapshot creation to make a copy of a root for a tree with 175 * a given objectid. The buffer with the new root node is returned in 176 * cow_ret, and this func returns zero on success or a negative error code. 177 */ 178 int btrfs_copy_root(struct btrfs_trans_handle *trans, 179 struct btrfs_root *root, 180 struct extent_buffer *buf, 181 struct extent_buffer **cow_ret, u64 new_root_objectid) 182 { 183 struct btrfs_fs_info *fs_info = root->fs_info; 184 struct extent_buffer *cow; 185 int ret = 0; 186 int level; 187 struct btrfs_disk_key disk_key; 188 189 WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && 190 trans->transid != fs_info->running_transaction->transid); 191 WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && 192 trans->transid != root->last_trans); 193 194 level = btrfs_header_level(buf); 195 if (level == 0) 196 btrfs_item_key(buf, &disk_key, 0); 197 else 198 btrfs_node_key(buf, &disk_key, 0); 199 200 cow = btrfs_alloc_tree_block(trans, root, 0, new_root_objectid, 201 &disk_key, level, buf->start, 0); 202 if (IS_ERR(cow)) 203 return PTR_ERR(cow); 204 205 copy_extent_buffer_full(cow, buf); 206 btrfs_set_header_bytenr(cow, cow->start); 207 btrfs_set_header_generation(cow, trans->transid); 208 btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV); 209 btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN | 210 BTRFS_HEADER_FLAG_RELOC); 211 if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID) 212 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC); 213 else 214 btrfs_set_header_owner(cow, new_root_objectid); 215 216 write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid); 217 218 WARN_ON(btrfs_header_generation(buf) > trans->transid); 219 if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID) 220 ret = btrfs_inc_ref(trans, root, cow, 1); 221 else 222 ret = btrfs_inc_ref(trans, root, cow, 0); 223 224 if (ret) 225 return ret; 226 227 btrfs_mark_buffer_dirty(cow); 228 *cow_ret = cow; 229 return 0; 230 } 231 232 enum mod_log_op { 233 MOD_LOG_KEY_REPLACE, 234 MOD_LOG_KEY_ADD, 235 MOD_LOG_KEY_REMOVE, 236 MOD_LOG_KEY_REMOVE_WHILE_FREEING, 237 MOD_LOG_KEY_REMOVE_WHILE_MOVING, 238 MOD_LOG_MOVE_KEYS, 239 MOD_LOG_ROOT_REPLACE, 240 }; 241 242 struct tree_mod_root { 243 u64 logical; 244 u8 level; 245 }; 246 247 struct tree_mod_elem { 248 struct rb_node node; 249 u64 logical; 250 u64 seq; 251 enum mod_log_op op; 252 253 /* this is used for MOD_LOG_KEY_* and MOD_LOG_MOVE_KEYS operations */ 254 int slot; 255 256 /* this is used for MOD_LOG_KEY* and MOD_LOG_ROOT_REPLACE */ 257 u64 generation; 258 259 /* those are used for op == MOD_LOG_KEY_{REPLACE,REMOVE} */ 260 struct btrfs_disk_key key; 261 u64 blockptr; 262 263 /* this is used for op == MOD_LOG_MOVE_KEYS */ 264 struct { 265 int dst_slot; 266 int nr_items; 267 } move; 268 269 /* this is used for op == MOD_LOG_ROOT_REPLACE */ 270 struct tree_mod_root old_root; 271 }; 272 273 /* 274 * Pull a new tree mod seq number for our operation. 275 */ 276 static inline u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info) 277 { 278 return atomic64_inc_return(&fs_info->tree_mod_seq); 279 } 280 281 /* 282 * This adds a new blocker to the tree mod log's blocker list if the @elem 283 * passed does not already have a sequence number set. So when a caller expects 284 * to record tree modifications, it should ensure to set elem->seq to zero 285 * before calling btrfs_get_tree_mod_seq. 286 * Returns a fresh, unused tree log modification sequence number, even if no new 287 * blocker was added. 288 */ 289 u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info, 290 struct seq_list *elem) 291 { 292 write_lock(&fs_info->tree_mod_log_lock); 293 if (!elem->seq) { 294 elem->seq = btrfs_inc_tree_mod_seq(fs_info); 295 list_add_tail(&elem->list, &fs_info->tree_mod_seq_list); 296 } 297 write_unlock(&fs_info->tree_mod_log_lock); 298 299 return elem->seq; 300 } 301 302 void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info, 303 struct seq_list *elem) 304 { 305 struct rb_root *tm_root; 306 struct rb_node *node; 307 struct rb_node *next; 308 struct tree_mod_elem *tm; 309 u64 min_seq = (u64)-1; 310 u64 seq_putting = elem->seq; 311 312 if (!seq_putting) 313 return; 314 315 write_lock(&fs_info->tree_mod_log_lock); 316 list_del(&elem->list); 317 elem->seq = 0; 318 319 if (!list_empty(&fs_info->tree_mod_seq_list)) { 320 struct seq_list *first; 321 322 first = list_first_entry(&fs_info->tree_mod_seq_list, 323 struct seq_list, list); 324 if (seq_putting > first->seq) { 325 /* 326 * Blocker with lower sequence number exists, we 327 * cannot remove anything from the log. 328 */ 329 write_unlock(&fs_info->tree_mod_log_lock); 330 return; 331 } 332 min_seq = first->seq; 333 } 334 335 /* 336 * anything that's lower than the lowest existing (read: blocked) 337 * sequence number can be removed from the tree. 338 */ 339 tm_root = &fs_info->tree_mod_log; 340 for (node = rb_first(tm_root); node; node = next) { 341 next = rb_next(node); 342 tm = rb_entry(node, struct tree_mod_elem, node); 343 if (tm->seq >= min_seq) 344 continue; 345 rb_erase(node, tm_root); 346 kfree(tm); 347 } 348 write_unlock(&fs_info->tree_mod_log_lock); 349 } 350 351 /* 352 * key order of the log: 353 * node/leaf start address -> sequence 354 * 355 * The 'start address' is the logical address of the *new* root node 356 * for root replace operations, or the logical address of the affected 357 * block for all other operations. 358 */ 359 static noinline int 360 __tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm) 361 { 362 struct rb_root *tm_root; 363 struct rb_node **new; 364 struct rb_node *parent = NULL; 365 struct tree_mod_elem *cur; 366 367 lockdep_assert_held_write(&fs_info->tree_mod_log_lock); 368 369 tm->seq = btrfs_inc_tree_mod_seq(fs_info); 370 371 tm_root = &fs_info->tree_mod_log; 372 new = &tm_root->rb_node; 373 while (*new) { 374 cur = rb_entry(*new, struct tree_mod_elem, node); 375 parent = *new; 376 if (cur->logical < tm->logical) 377 new = &((*new)->rb_left); 378 else if (cur->logical > tm->logical) 379 new = &((*new)->rb_right); 380 else if (cur->seq < tm->seq) 381 new = &((*new)->rb_left); 382 else if (cur->seq > tm->seq) 383 new = &((*new)->rb_right); 384 else 385 return -EEXIST; 386 } 387 388 rb_link_node(&tm->node, parent, new); 389 rb_insert_color(&tm->node, tm_root); 390 return 0; 391 } 392 393 /* 394 * Determines if logging can be omitted. Returns 1 if it can. Otherwise, it 395 * returns zero with the tree_mod_log_lock acquired. The caller must hold 396 * this until all tree mod log insertions are recorded in the rb tree and then 397 * write unlock fs_info::tree_mod_log_lock. 398 */ 399 static inline int tree_mod_dont_log(struct btrfs_fs_info *fs_info, 400 struct extent_buffer *eb) { 401 smp_mb(); 402 if (list_empty(&(fs_info)->tree_mod_seq_list)) 403 return 1; 404 if (eb && btrfs_header_level(eb) == 0) 405 return 1; 406 407 write_lock(&fs_info->tree_mod_log_lock); 408 if (list_empty(&(fs_info)->tree_mod_seq_list)) { 409 write_unlock(&fs_info->tree_mod_log_lock); 410 return 1; 411 } 412 413 return 0; 414 } 415 416 /* Similar to tree_mod_dont_log, but doesn't acquire any locks. */ 417 static inline int tree_mod_need_log(const struct btrfs_fs_info *fs_info, 418 struct extent_buffer *eb) 419 { 420 smp_mb(); 421 if (list_empty(&(fs_info)->tree_mod_seq_list)) 422 return 0; 423 if (eb && btrfs_header_level(eb) == 0) 424 return 0; 425 426 return 1; 427 } 428 429 static struct tree_mod_elem * 430 alloc_tree_mod_elem(struct extent_buffer *eb, int slot, 431 enum mod_log_op op, gfp_t flags) 432 { 433 struct tree_mod_elem *tm; 434 435 tm = kzalloc(sizeof(*tm), flags); 436 if (!tm) 437 return NULL; 438 439 tm->logical = eb->start; 440 if (op != MOD_LOG_KEY_ADD) { 441 btrfs_node_key(eb, &tm->key, slot); 442 tm->blockptr = btrfs_node_blockptr(eb, slot); 443 } 444 tm->op = op; 445 tm->slot = slot; 446 tm->generation = btrfs_node_ptr_generation(eb, slot); 447 RB_CLEAR_NODE(&tm->node); 448 449 return tm; 450 } 451 452 static noinline int tree_mod_log_insert_key(struct extent_buffer *eb, int slot, 453 enum mod_log_op op, gfp_t flags) 454 { 455 struct tree_mod_elem *tm; 456 int ret; 457 458 if (!tree_mod_need_log(eb->fs_info, eb)) 459 return 0; 460 461 tm = alloc_tree_mod_elem(eb, slot, op, flags); 462 if (!tm) 463 return -ENOMEM; 464 465 if (tree_mod_dont_log(eb->fs_info, eb)) { 466 kfree(tm); 467 return 0; 468 } 469 470 ret = __tree_mod_log_insert(eb->fs_info, tm); 471 write_unlock(&eb->fs_info->tree_mod_log_lock); 472 if (ret) 473 kfree(tm); 474 475 return ret; 476 } 477 478 static noinline int tree_mod_log_insert_move(struct extent_buffer *eb, 479 int dst_slot, int src_slot, int nr_items) 480 { 481 struct tree_mod_elem *tm = NULL; 482 struct tree_mod_elem **tm_list = NULL; 483 int ret = 0; 484 int i; 485 int locked = 0; 486 487 if (!tree_mod_need_log(eb->fs_info, eb)) 488 return 0; 489 490 tm_list = kcalloc(nr_items, sizeof(struct tree_mod_elem *), GFP_NOFS); 491 if (!tm_list) 492 return -ENOMEM; 493 494 tm = kzalloc(sizeof(*tm), GFP_NOFS); 495 if (!tm) { 496 ret = -ENOMEM; 497 goto free_tms; 498 } 499 500 tm->logical = eb->start; 501 tm->slot = src_slot; 502 tm->move.dst_slot = dst_slot; 503 tm->move.nr_items = nr_items; 504 tm->op = MOD_LOG_MOVE_KEYS; 505 506 for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) { 507 tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot, 508 MOD_LOG_KEY_REMOVE_WHILE_MOVING, GFP_NOFS); 509 if (!tm_list[i]) { 510 ret = -ENOMEM; 511 goto free_tms; 512 } 513 } 514 515 if (tree_mod_dont_log(eb->fs_info, eb)) 516 goto free_tms; 517 locked = 1; 518 519 /* 520 * When we override something during the move, we log these removals. 521 * This can only happen when we move towards the beginning of the 522 * buffer, i.e. dst_slot < src_slot. 523 */ 524 for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) { 525 ret = __tree_mod_log_insert(eb->fs_info, tm_list[i]); 526 if (ret) 527 goto free_tms; 528 } 529 530 ret = __tree_mod_log_insert(eb->fs_info, tm); 531 if (ret) 532 goto free_tms; 533 write_unlock(&eb->fs_info->tree_mod_log_lock); 534 kfree(tm_list); 535 536 return 0; 537 free_tms: 538 for (i = 0; i < nr_items; i++) { 539 if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node)) 540 rb_erase(&tm_list[i]->node, &eb->fs_info->tree_mod_log); 541 kfree(tm_list[i]); 542 } 543 if (locked) 544 write_unlock(&eb->fs_info->tree_mod_log_lock); 545 kfree(tm_list); 546 kfree(tm); 547 548 return ret; 549 } 550 551 static inline int 552 __tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, 553 struct tree_mod_elem **tm_list, 554 int nritems) 555 { 556 int i, j; 557 int ret; 558 559 for (i = nritems - 1; i >= 0; i--) { 560 ret = __tree_mod_log_insert(fs_info, tm_list[i]); 561 if (ret) { 562 for (j = nritems - 1; j > i; j--) 563 rb_erase(&tm_list[j]->node, 564 &fs_info->tree_mod_log); 565 return ret; 566 } 567 } 568 569 return 0; 570 } 571 572 static noinline int tree_mod_log_insert_root(struct extent_buffer *old_root, 573 struct extent_buffer *new_root, int log_removal) 574 { 575 struct btrfs_fs_info *fs_info = old_root->fs_info; 576 struct tree_mod_elem *tm = NULL; 577 struct tree_mod_elem **tm_list = NULL; 578 int nritems = 0; 579 int ret = 0; 580 int i; 581 582 if (!tree_mod_need_log(fs_info, NULL)) 583 return 0; 584 585 if (log_removal && btrfs_header_level(old_root) > 0) { 586 nritems = btrfs_header_nritems(old_root); 587 tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), 588 GFP_NOFS); 589 if (!tm_list) { 590 ret = -ENOMEM; 591 goto free_tms; 592 } 593 for (i = 0; i < nritems; i++) { 594 tm_list[i] = alloc_tree_mod_elem(old_root, i, 595 MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS); 596 if (!tm_list[i]) { 597 ret = -ENOMEM; 598 goto free_tms; 599 } 600 } 601 } 602 603 tm = kzalloc(sizeof(*tm), GFP_NOFS); 604 if (!tm) { 605 ret = -ENOMEM; 606 goto free_tms; 607 } 608 609 tm->logical = new_root->start; 610 tm->old_root.logical = old_root->start; 611 tm->old_root.level = btrfs_header_level(old_root); 612 tm->generation = btrfs_header_generation(old_root); 613 tm->op = MOD_LOG_ROOT_REPLACE; 614 615 if (tree_mod_dont_log(fs_info, NULL)) 616 goto free_tms; 617 618 if (tm_list) 619 ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems); 620 if (!ret) 621 ret = __tree_mod_log_insert(fs_info, tm); 622 623 write_unlock(&fs_info->tree_mod_log_lock); 624 if (ret) 625 goto free_tms; 626 kfree(tm_list); 627 628 return ret; 629 630 free_tms: 631 if (tm_list) { 632 for (i = 0; i < nritems; i++) 633 kfree(tm_list[i]); 634 kfree(tm_list); 635 } 636 kfree(tm); 637 638 return ret; 639 } 640 641 static struct tree_mod_elem * 642 __tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq, 643 int smallest) 644 { 645 struct rb_root *tm_root; 646 struct rb_node *node; 647 struct tree_mod_elem *cur = NULL; 648 struct tree_mod_elem *found = NULL; 649 650 read_lock(&fs_info->tree_mod_log_lock); 651 tm_root = &fs_info->tree_mod_log; 652 node = tm_root->rb_node; 653 while (node) { 654 cur = rb_entry(node, struct tree_mod_elem, node); 655 if (cur->logical < start) { 656 node = node->rb_left; 657 } else if (cur->logical > start) { 658 node = node->rb_right; 659 } else if (cur->seq < min_seq) { 660 node = node->rb_left; 661 } else if (!smallest) { 662 /* we want the node with the highest seq */ 663 if (found) 664 BUG_ON(found->seq > cur->seq); 665 found = cur; 666 node = node->rb_left; 667 } else if (cur->seq > min_seq) { 668 /* we want the node with the smallest seq */ 669 if (found) 670 BUG_ON(found->seq < cur->seq); 671 found = cur; 672 node = node->rb_right; 673 } else { 674 found = cur; 675 break; 676 } 677 } 678 read_unlock(&fs_info->tree_mod_log_lock); 679 680 return found; 681 } 682 683 /* 684 * this returns the element from the log with the smallest time sequence 685 * value that's in the log (the oldest log item). any element with a time 686 * sequence lower than min_seq will be ignored. 687 */ 688 static struct tree_mod_elem * 689 tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info, u64 start, 690 u64 min_seq) 691 { 692 return __tree_mod_log_search(fs_info, start, min_seq, 1); 693 } 694 695 /* 696 * this returns the element from the log with the largest time sequence 697 * value that's in the log (the most recent log item). any element with 698 * a time sequence lower than min_seq will be ignored. 699 */ 700 static struct tree_mod_elem * 701 tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq) 702 { 703 return __tree_mod_log_search(fs_info, start, min_seq, 0); 704 } 705 706 static noinline int tree_mod_log_eb_copy(struct extent_buffer *dst, 707 struct extent_buffer *src, unsigned long dst_offset, 708 unsigned long src_offset, int nr_items) 709 { 710 struct btrfs_fs_info *fs_info = dst->fs_info; 711 int ret = 0; 712 struct tree_mod_elem **tm_list = NULL; 713 struct tree_mod_elem **tm_list_add, **tm_list_rem; 714 int i; 715 int locked = 0; 716 717 if (!tree_mod_need_log(fs_info, NULL)) 718 return 0; 719 720 if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0) 721 return 0; 722 723 tm_list = kcalloc(nr_items * 2, sizeof(struct tree_mod_elem *), 724 GFP_NOFS); 725 if (!tm_list) 726 return -ENOMEM; 727 728 tm_list_add = tm_list; 729 tm_list_rem = tm_list + nr_items; 730 for (i = 0; i < nr_items; i++) { 731 tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset, 732 MOD_LOG_KEY_REMOVE, GFP_NOFS); 733 if (!tm_list_rem[i]) { 734 ret = -ENOMEM; 735 goto free_tms; 736 } 737 738 tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset, 739 MOD_LOG_KEY_ADD, GFP_NOFS); 740 if (!tm_list_add[i]) { 741 ret = -ENOMEM; 742 goto free_tms; 743 } 744 } 745 746 if (tree_mod_dont_log(fs_info, NULL)) 747 goto free_tms; 748 locked = 1; 749 750 for (i = 0; i < nr_items; i++) { 751 ret = __tree_mod_log_insert(fs_info, tm_list_rem[i]); 752 if (ret) 753 goto free_tms; 754 ret = __tree_mod_log_insert(fs_info, tm_list_add[i]); 755 if (ret) 756 goto free_tms; 757 } 758 759 write_unlock(&fs_info->tree_mod_log_lock); 760 kfree(tm_list); 761 762 return 0; 763 764 free_tms: 765 for (i = 0; i < nr_items * 2; i++) { 766 if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node)) 767 rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log); 768 kfree(tm_list[i]); 769 } 770 if (locked) 771 write_unlock(&fs_info->tree_mod_log_lock); 772 kfree(tm_list); 773 774 return ret; 775 } 776 777 static noinline int tree_mod_log_free_eb(struct extent_buffer *eb) 778 { 779 struct tree_mod_elem **tm_list = NULL; 780 int nritems = 0; 781 int i; 782 int ret = 0; 783 784 if (btrfs_header_level(eb) == 0) 785 return 0; 786 787 if (!tree_mod_need_log(eb->fs_info, NULL)) 788 return 0; 789 790 nritems = btrfs_header_nritems(eb); 791 tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), GFP_NOFS); 792 if (!tm_list) 793 return -ENOMEM; 794 795 for (i = 0; i < nritems; i++) { 796 tm_list[i] = alloc_tree_mod_elem(eb, i, 797 MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS); 798 if (!tm_list[i]) { 799 ret = -ENOMEM; 800 goto free_tms; 801 } 802 } 803 804 if (tree_mod_dont_log(eb->fs_info, eb)) 805 goto free_tms; 806 807 ret = __tree_mod_log_free_eb(eb->fs_info, tm_list, nritems); 808 write_unlock(&eb->fs_info->tree_mod_log_lock); 809 if (ret) 810 goto free_tms; 811 kfree(tm_list); 812 813 return 0; 814 815 free_tms: 816 for (i = 0; i < nritems; i++) 817 kfree(tm_list[i]); 818 kfree(tm_list); 819 820 return ret; 821 } 822 823 /* 824 * check if the tree block can be shared by multiple trees 825 */ 826 int btrfs_block_can_be_shared(struct btrfs_root *root, 827 struct extent_buffer *buf) 828 { 829 /* 830 * Tree blocks not in shareable trees and tree roots are never shared. 831 * If a block was allocated after the last snapshot and the block was 832 * not allocated by tree relocation, we know the block is not shared. 833 */ 834 if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && 835 buf != root->node && buf != root->commit_root && 836 (btrfs_header_generation(buf) <= 837 btrfs_root_last_snapshot(&root->root_item) || 838 btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC))) 839 return 1; 840 841 return 0; 842 } 843 844 static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans, 845 struct btrfs_root *root, 846 struct extent_buffer *buf, 847 struct extent_buffer *cow, 848 int *last_ref) 849 { 850 struct btrfs_fs_info *fs_info = root->fs_info; 851 u64 refs; 852 u64 owner; 853 u64 flags; 854 u64 new_flags = 0; 855 int ret; 856 857 /* 858 * Backrefs update rules: 859 * 860 * Always use full backrefs for extent pointers in tree block 861 * allocated by tree relocation. 862 * 863 * If a shared tree block is no longer referenced by its owner 864 * tree (btrfs_header_owner(buf) == root->root_key.objectid), 865 * use full backrefs for extent pointers in tree block. 866 * 867 * If a tree block is been relocating 868 * (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID), 869 * use full backrefs for extent pointers in tree block. 870 * The reason for this is some operations (such as drop tree) 871 * are only allowed for blocks use full backrefs. 872 */ 873 874 if (btrfs_block_can_be_shared(root, buf)) { 875 ret = btrfs_lookup_extent_info(trans, fs_info, buf->start, 876 btrfs_header_level(buf), 1, 877 &refs, &flags); 878 if (ret) 879 return ret; 880 if (refs == 0) { 881 ret = -EROFS; 882 btrfs_handle_fs_error(fs_info, ret, NULL); 883 return ret; 884 } 885 } else { 886 refs = 1; 887 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID || 888 btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV) 889 flags = BTRFS_BLOCK_FLAG_FULL_BACKREF; 890 else 891 flags = 0; 892 } 893 894 owner = btrfs_header_owner(buf); 895 BUG_ON(owner == BTRFS_TREE_RELOC_OBJECTID && 896 !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)); 897 898 if (refs > 1) { 899 if ((owner == root->root_key.objectid || 900 root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && 901 !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) { 902 ret = btrfs_inc_ref(trans, root, buf, 1); 903 if (ret) 904 return ret; 905 906 if (root->root_key.objectid == 907 BTRFS_TREE_RELOC_OBJECTID) { 908 ret = btrfs_dec_ref(trans, root, buf, 0); 909 if (ret) 910 return ret; 911 ret = btrfs_inc_ref(trans, root, cow, 1); 912 if (ret) 913 return ret; 914 } 915 new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF; 916 } else { 917 918 if (root->root_key.objectid == 919 BTRFS_TREE_RELOC_OBJECTID) 920 ret = btrfs_inc_ref(trans, root, cow, 1); 921 else 922 ret = btrfs_inc_ref(trans, root, cow, 0); 923 if (ret) 924 return ret; 925 } 926 if (new_flags != 0) { 927 int level = btrfs_header_level(buf); 928 929 ret = btrfs_set_disk_extent_flags(trans, buf, 930 new_flags, level, 0); 931 if (ret) 932 return ret; 933 } 934 } else { 935 if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) { 936 if (root->root_key.objectid == 937 BTRFS_TREE_RELOC_OBJECTID) 938 ret = btrfs_inc_ref(trans, root, cow, 1); 939 else 940 ret = btrfs_inc_ref(trans, root, cow, 0); 941 if (ret) 942 return ret; 943 ret = btrfs_dec_ref(trans, root, buf, 1); 944 if (ret) 945 return ret; 946 } 947 btrfs_clean_tree_block(buf); 948 *last_ref = 1; 949 } 950 return 0; 951 } 952 953 static struct extent_buffer *alloc_tree_block_no_bg_flush( 954 struct btrfs_trans_handle *trans, 955 struct btrfs_root *root, 956 u64 parent_start, 957 const struct btrfs_disk_key *disk_key, 958 int level, 959 u64 hint, 960 u64 empty_size) 961 { 962 struct btrfs_fs_info *fs_info = root->fs_info; 963 struct extent_buffer *ret; 964 965 /* 966 * If we are COWing a node/leaf from the extent, chunk, device or free 967 * space trees, make sure that we do not finish block group creation of 968 * pending block groups. We do this to avoid a deadlock. 969 * COWing can result in allocation of a new chunk, and flushing pending 970 * block groups (btrfs_create_pending_block_groups()) can be triggered 971 * when finishing allocation of a new chunk. Creation of a pending block 972 * group modifies the extent, chunk, device and free space trees, 973 * therefore we could deadlock with ourselves since we are holding a 974 * lock on an extent buffer that btrfs_create_pending_block_groups() may 975 * try to COW later. 976 * For similar reasons, we also need to delay flushing pending block 977 * groups when splitting a leaf or node, from one of those trees, since 978 * we are holding a write lock on it and its parent or when inserting a 979 * new root node for one of those trees. 980 */ 981 if (root == fs_info->extent_root || 982 root == fs_info->chunk_root || 983 root == fs_info->dev_root || 984 root == fs_info->free_space_root) 985 trans->can_flush_pending_bgs = false; 986 987 ret = btrfs_alloc_tree_block(trans, root, parent_start, 988 root->root_key.objectid, disk_key, level, 989 hint, empty_size); 990 trans->can_flush_pending_bgs = true; 991 992 return ret; 993 } 994 995 /* 996 * does the dirty work in cow of a single block. The parent block (if 997 * supplied) is updated to point to the new cow copy. The new buffer is marked 998 * dirty and returned locked. If you modify the block it needs to be marked 999 * dirty again. 1000 * 1001 * search_start -- an allocation hint for the new block 1002 * 1003 * empty_size -- a hint that you plan on doing more cow. This is the size in 1004 * bytes the allocator should try to find free next to the block it returns. 1005 * This is just a hint and may be ignored by the allocator. 1006 */ 1007 static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans, 1008 struct btrfs_root *root, 1009 struct extent_buffer *buf, 1010 struct extent_buffer *parent, int parent_slot, 1011 struct extent_buffer **cow_ret, 1012 u64 search_start, u64 empty_size) 1013 { 1014 struct btrfs_fs_info *fs_info = root->fs_info; 1015 struct btrfs_disk_key disk_key; 1016 struct extent_buffer *cow; 1017 int level, ret; 1018 int last_ref = 0; 1019 int unlock_orig = 0; 1020 u64 parent_start = 0; 1021 1022 if (*cow_ret == buf) 1023 unlock_orig = 1; 1024 1025 btrfs_assert_tree_locked(buf); 1026 1027 WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && 1028 trans->transid != fs_info->running_transaction->transid); 1029 WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && 1030 trans->transid != root->last_trans); 1031 1032 level = btrfs_header_level(buf); 1033 1034 if (level == 0) 1035 btrfs_item_key(buf, &disk_key, 0); 1036 else 1037 btrfs_node_key(buf, &disk_key, 0); 1038 1039 if ((root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && parent) 1040 parent_start = parent->start; 1041 1042 cow = alloc_tree_block_no_bg_flush(trans, root, parent_start, &disk_key, 1043 level, search_start, empty_size); 1044 if (IS_ERR(cow)) 1045 return PTR_ERR(cow); 1046 1047 /* cow is set to blocking by btrfs_init_new_buffer */ 1048 1049 copy_extent_buffer_full(cow, buf); 1050 btrfs_set_header_bytenr(cow, cow->start); 1051 btrfs_set_header_generation(cow, trans->transid); 1052 btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV); 1053 btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN | 1054 BTRFS_HEADER_FLAG_RELOC); 1055 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) 1056 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC); 1057 else 1058 btrfs_set_header_owner(cow, root->root_key.objectid); 1059 1060 write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid); 1061 1062 ret = update_ref_for_cow(trans, root, buf, cow, &last_ref); 1063 if (ret) { 1064 btrfs_abort_transaction(trans, ret); 1065 return ret; 1066 } 1067 1068 if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) { 1069 ret = btrfs_reloc_cow_block(trans, root, buf, cow); 1070 if (ret) { 1071 btrfs_abort_transaction(trans, ret); 1072 return ret; 1073 } 1074 } 1075 1076 if (buf == root->node) { 1077 WARN_ON(parent && parent != buf); 1078 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID || 1079 btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV) 1080 parent_start = buf->start; 1081 1082 atomic_inc(&cow->refs); 1083 ret = tree_mod_log_insert_root(root->node, cow, 1); 1084 BUG_ON(ret < 0); 1085 rcu_assign_pointer(root->node, cow); 1086 1087 btrfs_free_tree_block(trans, root, buf, parent_start, 1088 last_ref); 1089 free_extent_buffer(buf); 1090 add_root_to_dirty_list(root); 1091 } else { 1092 WARN_ON(trans->transid != btrfs_header_generation(parent)); 1093 tree_mod_log_insert_key(parent, parent_slot, 1094 MOD_LOG_KEY_REPLACE, GFP_NOFS); 1095 btrfs_set_node_blockptr(parent, parent_slot, 1096 cow->start); 1097 btrfs_set_node_ptr_generation(parent, parent_slot, 1098 trans->transid); 1099 btrfs_mark_buffer_dirty(parent); 1100 if (last_ref) { 1101 ret = tree_mod_log_free_eb(buf); 1102 if (ret) { 1103 btrfs_abort_transaction(trans, ret); 1104 return ret; 1105 } 1106 } 1107 btrfs_free_tree_block(trans, root, buf, parent_start, 1108 last_ref); 1109 } 1110 if (unlock_orig) 1111 btrfs_tree_unlock(buf); 1112 free_extent_buffer_stale(buf); 1113 btrfs_mark_buffer_dirty(cow); 1114 *cow_ret = cow; 1115 return 0; 1116 } 1117 1118 /* 1119 * returns the logical address of the oldest predecessor of the given root. 1120 * entries older than time_seq are ignored. 1121 */ 1122 static struct tree_mod_elem *__tree_mod_log_oldest_root( 1123 struct extent_buffer *eb_root, u64 time_seq) 1124 { 1125 struct tree_mod_elem *tm; 1126 struct tree_mod_elem *found = NULL; 1127 u64 root_logical = eb_root->start; 1128 int looped = 0; 1129 1130 if (!time_seq) 1131 return NULL; 1132 1133 /* 1134 * the very last operation that's logged for a root is the 1135 * replacement operation (if it is replaced at all). this has 1136 * the logical address of the *new* root, making it the very 1137 * first operation that's logged for this root. 1138 */ 1139 while (1) { 1140 tm = tree_mod_log_search_oldest(eb_root->fs_info, root_logical, 1141 time_seq); 1142 if (!looped && !tm) 1143 return NULL; 1144 /* 1145 * if there are no tree operation for the oldest root, we simply 1146 * return it. this should only happen if that (old) root is at 1147 * level 0. 1148 */ 1149 if (!tm) 1150 break; 1151 1152 /* 1153 * if there's an operation that's not a root replacement, we 1154 * found the oldest version of our root. normally, we'll find a 1155 * MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here. 1156 */ 1157 if (tm->op != MOD_LOG_ROOT_REPLACE) 1158 break; 1159 1160 found = tm; 1161 root_logical = tm->old_root.logical; 1162 looped = 1; 1163 } 1164 1165 /* if there's no old root to return, return what we found instead */ 1166 if (!found) 1167 found = tm; 1168 1169 return found; 1170 } 1171 1172 /* 1173 * tm is a pointer to the first operation to rewind within eb. then, all 1174 * previous operations will be rewound (until we reach something older than 1175 * time_seq). 1176 */ 1177 static void 1178 __tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct extent_buffer *eb, 1179 u64 time_seq, struct tree_mod_elem *first_tm) 1180 { 1181 u32 n; 1182 struct rb_node *next; 1183 struct tree_mod_elem *tm = first_tm; 1184 unsigned long o_dst; 1185 unsigned long o_src; 1186 unsigned long p_size = sizeof(struct btrfs_key_ptr); 1187 1188 n = btrfs_header_nritems(eb); 1189 read_lock(&fs_info->tree_mod_log_lock); 1190 while (tm && tm->seq >= time_seq) { 1191 /* 1192 * all the operations are recorded with the operator used for 1193 * the modification. as we're going backwards, we do the 1194 * opposite of each operation here. 1195 */ 1196 switch (tm->op) { 1197 case MOD_LOG_KEY_REMOVE_WHILE_FREEING: 1198 BUG_ON(tm->slot < n); 1199 fallthrough; 1200 case MOD_LOG_KEY_REMOVE_WHILE_MOVING: 1201 case MOD_LOG_KEY_REMOVE: 1202 btrfs_set_node_key(eb, &tm->key, tm->slot); 1203 btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr); 1204 btrfs_set_node_ptr_generation(eb, tm->slot, 1205 tm->generation); 1206 n++; 1207 break; 1208 case MOD_LOG_KEY_REPLACE: 1209 BUG_ON(tm->slot >= n); 1210 btrfs_set_node_key(eb, &tm->key, tm->slot); 1211 btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr); 1212 btrfs_set_node_ptr_generation(eb, tm->slot, 1213 tm->generation); 1214 break; 1215 case MOD_LOG_KEY_ADD: 1216 /* if a move operation is needed it's in the log */ 1217 n--; 1218 break; 1219 case MOD_LOG_MOVE_KEYS: 1220 o_dst = btrfs_node_key_ptr_offset(tm->slot); 1221 o_src = btrfs_node_key_ptr_offset(tm->move.dst_slot); 1222 memmove_extent_buffer(eb, o_dst, o_src, 1223 tm->move.nr_items * p_size); 1224 break; 1225 case MOD_LOG_ROOT_REPLACE: 1226 /* 1227 * this operation is special. for roots, this must be 1228 * handled explicitly before rewinding. 1229 * for non-roots, this operation may exist if the node 1230 * was a root: root A -> child B; then A gets empty and 1231 * B is promoted to the new root. in the mod log, we'll 1232 * have a root-replace operation for B, a tree block 1233 * that is no root. we simply ignore that operation. 1234 */ 1235 break; 1236 } 1237 next = rb_next(&tm->node); 1238 if (!next) 1239 break; 1240 tm = rb_entry(next, struct tree_mod_elem, node); 1241 if (tm->logical != first_tm->logical) 1242 break; 1243 } 1244 read_unlock(&fs_info->tree_mod_log_lock); 1245 btrfs_set_header_nritems(eb, n); 1246 } 1247 1248 /* 1249 * Called with eb read locked. If the buffer cannot be rewound, the same buffer 1250 * is returned. If rewind operations happen, a fresh buffer is returned. The 1251 * returned buffer is always read-locked. If the returned buffer is not the 1252 * input buffer, the lock on the input buffer is released and the input buffer 1253 * is freed (its refcount is decremented). 1254 */ 1255 static struct extent_buffer * 1256 tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct btrfs_path *path, 1257 struct extent_buffer *eb, u64 time_seq) 1258 { 1259 struct extent_buffer *eb_rewin; 1260 struct tree_mod_elem *tm; 1261 1262 if (!time_seq) 1263 return eb; 1264 1265 if (btrfs_header_level(eb) == 0) 1266 return eb; 1267 1268 tm = tree_mod_log_search(fs_info, eb->start, time_seq); 1269 if (!tm) 1270 return eb; 1271 1272 btrfs_set_path_blocking(path); 1273 btrfs_set_lock_blocking_read(eb); 1274 1275 if (tm->op == MOD_LOG_KEY_REMOVE_WHILE_FREEING) { 1276 BUG_ON(tm->slot != 0); 1277 eb_rewin = alloc_dummy_extent_buffer(fs_info, eb->start); 1278 if (!eb_rewin) { 1279 btrfs_tree_read_unlock_blocking(eb); 1280 free_extent_buffer(eb); 1281 return NULL; 1282 } 1283 btrfs_set_header_bytenr(eb_rewin, eb->start); 1284 btrfs_set_header_backref_rev(eb_rewin, 1285 btrfs_header_backref_rev(eb)); 1286 btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb)); 1287 btrfs_set_header_level(eb_rewin, btrfs_header_level(eb)); 1288 } else { 1289 eb_rewin = btrfs_clone_extent_buffer(eb); 1290 if (!eb_rewin) { 1291 btrfs_tree_read_unlock_blocking(eb); 1292 free_extent_buffer(eb); 1293 return NULL; 1294 } 1295 } 1296 1297 btrfs_tree_read_unlock_blocking(eb); 1298 free_extent_buffer(eb); 1299 1300 btrfs_tree_read_lock(eb_rewin); 1301 __tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm); 1302 WARN_ON(btrfs_header_nritems(eb_rewin) > 1303 BTRFS_NODEPTRS_PER_BLOCK(fs_info)); 1304 1305 return eb_rewin; 1306 } 1307 1308 /* 1309 * get_old_root() rewinds the state of @root's root node to the given @time_seq 1310 * value. If there are no changes, the current root->root_node is returned. If 1311 * anything changed in between, there's a fresh buffer allocated on which the 1312 * rewind operations are done. In any case, the returned buffer is read locked. 1313 * Returns NULL on error (with no locks held). 1314 */ 1315 static inline struct extent_buffer * 1316 get_old_root(struct btrfs_root *root, u64 time_seq) 1317 { 1318 struct btrfs_fs_info *fs_info = root->fs_info; 1319 struct tree_mod_elem *tm; 1320 struct extent_buffer *eb = NULL; 1321 struct extent_buffer *eb_root; 1322 u64 eb_root_owner = 0; 1323 struct extent_buffer *old; 1324 struct tree_mod_root *old_root = NULL; 1325 u64 old_generation = 0; 1326 u64 logical; 1327 int level; 1328 1329 eb_root = btrfs_read_lock_root_node(root); 1330 tm = __tree_mod_log_oldest_root(eb_root, time_seq); 1331 if (!tm) 1332 return eb_root; 1333 1334 if (tm->op == MOD_LOG_ROOT_REPLACE) { 1335 old_root = &tm->old_root; 1336 old_generation = tm->generation; 1337 logical = old_root->logical; 1338 level = old_root->level; 1339 } else { 1340 logical = eb_root->start; 1341 level = btrfs_header_level(eb_root); 1342 } 1343 1344 tm = tree_mod_log_search(fs_info, logical, time_seq); 1345 if (old_root && tm && tm->op != MOD_LOG_KEY_REMOVE_WHILE_FREEING) { 1346 btrfs_tree_read_unlock(eb_root); 1347 free_extent_buffer(eb_root); 1348 old = read_tree_block(fs_info, logical, 0, level, NULL); 1349 if (WARN_ON(IS_ERR(old) || !extent_buffer_uptodate(old))) { 1350 if (!IS_ERR(old)) 1351 free_extent_buffer(old); 1352 btrfs_warn(fs_info, 1353 "failed to read tree block %llu from get_old_root", 1354 logical); 1355 } else { 1356 eb = btrfs_clone_extent_buffer(old); 1357 free_extent_buffer(old); 1358 } 1359 } else if (old_root) { 1360 eb_root_owner = btrfs_header_owner(eb_root); 1361 btrfs_tree_read_unlock(eb_root); 1362 free_extent_buffer(eb_root); 1363 eb = alloc_dummy_extent_buffer(fs_info, logical); 1364 } else { 1365 btrfs_set_lock_blocking_read(eb_root); 1366 eb = btrfs_clone_extent_buffer(eb_root); 1367 btrfs_tree_read_unlock_blocking(eb_root); 1368 free_extent_buffer(eb_root); 1369 } 1370 1371 if (!eb) 1372 return NULL; 1373 btrfs_tree_read_lock(eb); 1374 if (old_root) { 1375 btrfs_set_header_bytenr(eb, eb->start); 1376 btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV); 1377 btrfs_set_header_owner(eb, eb_root_owner); 1378 btrfs_set_header_level(eb, old_root->level); 1379 btrfs_set_header_generation(eb, old_generation); 1380 } 1381 if (tm) 1382 __tree_mod_log_rewind(fs_info, eb, time_seq, tm); 1383 else 1384 WARN_ON(btrfs_header_level(eb) != 0); 1385 WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(fs_info)); 1386 1387 return eb; 1388 } 1389 1390 int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq) 1391 { 1392 struct tree_mod_elem *tm; 1393 int level; 1394 struct extent_buffer *eb_root = btrfs_root_node(root); 1395 1396 tm = __tree_mod_log_oldest_root(eb_root, time_seq); 1397 if (tm && tm->op == MOD_LOG_ROOT_REPLACE) { 1398 level = tm->old_root.level; 1399 } else { 1400 level = btrfs_header_level(eb_root); 1401 } 1402 free_extent_buffer(eb_root); 1403 1404 return level; 1405 } 1406 1407 static inline int should_cow_block(struct btrfs_trans_handle *trans, 1408 struct btrfs_root *root, 1409 struct extent_buffer *buf) 1410 { 1411 if (btrfs_is_testing(root->fs_info)) 1412 return 0; 1413 1414 /* Ensure we can see the FORCE_COW bit */ 1415 smp_mb__before_atomic(); 1416 1417 /* 1418 * We do not need to cow a block if 1419 * 1) this block is not created or changed in this transaction; 1420 * 2) this block does not belong to TREE_RELOC tree; 1421 * 3) the root is not forced COW. 1422 * 1423 * What is forced COW: 1424 * when we create snapshot during committing the transaction, 1425 * after we've finished copying src root, we must COW the shared 1426 * block to ensure the metadata consistency. 1427 */ 1428 if (btrfs_header_generation(buf) == trans->transid && 1429 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) && 1430 !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 1431 btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) && 1432 !test_bit(BTRFS_ROOT_FORCE_COW, &root->state)) 1433 return 0; 1434 return 1; 1435 } 1436 1437 /* 1438 * cows a single block, see __btrfs_cow_block for the real work. 1439 * This version of it has extra checks so that a block isn't COWed more than 1440 * once per transaction, as long as it hasn't been written yet 1441 */ 1442 noinline int btrfs_cow_block(struct btrfs_trans_handle *trans, 1443 struct btrfs_root *root, struct extent_buffer *buf, 1444 struct extent_buffer *parent, int parent_slot, 1445 struct extent_buffer **cow_ret) 1446 { 1447 struct btrfs_fs_info *fs_info = root->fs_info; 1448 u64 search_start; 1449 int ret; 1450 1451 if (test_bit(BTRFS_ROOT_DELETING, &root->state)) 1452 btrfs_err(fs_info, 1453 "COW'ing blocks on a fs root that's being dropped"); 1454 1455 if (trans->transaction != fs_info->running_transaction) 1456 WARN(1, KERN_CRIT "trans %llu running %llu\n", 1457 trans->transid, 1458 fs_info->running_transaction->transid); 1459 1460 if (trans->transid != fs_info->generation) 1461 WARN(1, KERN_CRIT "trans %llu running %llu\n", 1462 trans->transid, fs_info->generation); 1463 1464 if (!should_cow_block(trans, root, buf)) { 1465 trans->dirty = true; 1466 *cow_ret = buf; 1467 return 0; 1468 } 1469 1470 search_start = buf->start & ~((u64)SZ_1G - 1); 1471 1472 if (parent) 1473 btrfs_set_lock_blocking_write(parent); 1474 btrfs_set_lock_blocking_write(buf); 1475 1476 /* 1477 * Before CoWing this block for later modification, check if it's 1478 * the subtree root and do the delayed subtree trace if needed. 1479 * 1480 * Also We don't care about the error, as it's handled internally. 1481 */ 1482 btrfs_qgroup_trace_subtree_after_cow(trans, root, buf); 1483 ret = __btrfs_cow_block(trans, root, buf, parent, 1484 parent_slot, cow_ret, search_start, 0); 1485 1486 trace_btrfs_cow_block(root, buf, *cow_ret); 1487 1488 return ret; 1489 } 1490 1491 /* 1492 * helper function for defrag to decide if two blocks pointed to by a 1493 * node are actually close by 1494 */ 1495 static int close_blocks(u64 blocknr, u64 other, u32 blocksize) 1496 { 1497 if (blocknr < other && other - (blocknr + blocksize) < 32768) 1498 return 1; 1499 if (blocknr > other && blocknr - (other + blocksize) < 32768) 1500 return 1; 1501 return 0; 1502 } 1503 1504 /* 1505 * compare two keys in a memcmp fashion 1506 */ 1507 static int comp_keys(const struct btrfs_disk_key *disk, 1508 const struct btrfs_key *k2) 1509 { 1510 struct btrfs_key k1; 1511 1512 btrfs_disk_key_to_cpu(&k1, disk); 1513 1514 return btrfs_comp_cpu_keys(&k1, k2); 1515 } 1516 1517 /* 1518 * same as comp_keys only with two btrfs_key's 1519 */ 1520 int __pure btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2) 1521 { 1522 if (k1->objectid > k2->objectid) 1523 return 1; 1524 if (k1->objectid < k2->objectid) 1525 return -1; 1526 if (k1->type > k2->type) 1527 return 1; 1528 if (k1->type < k2->type) 1529 return -1; 1530 if (k1->offset > k2->offset) 1531 return 1; 1532 if (k1->offset < k2->offset) 1533 return -1; 1534 return 0; 1535 } 1536 1537 /* 1538 * this is used by the defrag code to go through all the 1539 * leaves pointed to by a node and reallocate them so that 1540 * disk order is close to key order 1541 */ 1542 int btrfs_realloc_node(struct btrfs_trans_handle *trans, 1543 struct btrfs_root *root, struct extent_buffer *parent, 1544 int start_slot, u64 *last_ret, 1545 struct btrfs_key *progress) 1546 { 1547 struct btrfs_fs_info *fs_info = root->fs_info; 1548 struct extent_buffer *cur; 1549 u64 blocknr; 1550 u64 gen; 1551 u64 search_start = *last_ret; 1552 u64 last_block = 0; 1553 u64 other; 1554 u32 parent_nritems; 1555 int end_slot; 1556 int i; 1557 int err = 0; 1558 int parent_level; 1559 int uptodate; 1560 u32 blocksize; 1561 int progress_passed = 0; 1562 struct btrfs_disk_key disk_key; 1563 1564 parent_level = btrfs_header_level(parent); 1565 1566 WARN_ON(trans->transaction != fs_info->running_transaction); 1567 WARN_ON(trans->transid != fs_info->generation); 1568 1569 parent_nritems = btrfs_header_nritems(parent); 1570 blocksize = fs_info->nodesize; 1571 end_slot = parent_nritems - 1; 1572 1573 if (parent_nritems <= 1) 1574 return 0; 1575 1576 btrfs_set_lock_blocking_write(parent); 1577 1578 for (i = start_slot; i <= end_slot; i++) { 1579 struct btrfs_key first_key; 1580 int close = 1; 1581 1582 btrfs_node_key(parent, &disk_key, i); 1583 if (!progress_passed && comp_keys(&disk_key, progress) < 0) 1584 continue; 1585 1586 progress_passed = 1; 1587 blocknr = btrfs_node_blockptr(parent, i); 1588 gen = btrfs_node_ptr_generation(parent, i); 1589 btrfs_node_key_to_cpu(parent, &first_key, i); 1590 if (last_block == 0) 1591 last_block = blocknr; 1592 1593 if (i > 0) { 1594 other = btrfs_node_blockptr(parent, i - 1); 1595 close = close_blocks(blocknr, other, blocksize); 1596 } 1597 if (!close && i < end_slot) { 1598 other = btrfs_node_blockptr(parent, i + 1); 1599 close = close_blocks(blocknr, other, blocksize); 1600 } 1601 if (close) { 1602 last_block = blocknr; 1603 continue; 1604 } 1605 1606 cur = find_extent_buffer(fs_info, blocknr); 1607 if (cur) 1608 uptodate = btrfs_buffer_uptodate(cur, gen, 0); 1609 else 1610 uptodate = 0; 1611 if (!cur || !uptodate) { 1612 if (!cur) { 1613 cur = read_tree_block(fs_info, blocknr, gen, 1614 parent_level - 1, 1615 &first_key); 1616 if (IS_ERR(cur)) { 1617 return PTR_ERR(cur); 1618 } else if (!extent_buffer_uptodate(cur)) { 1619 free_extent_buffer(cur); 1620 return -EIO; 1621 } 1622 } else if (!uptodate) { 1623 err = btrfs_read_buffer(cur, gen, 1624 parent_level - 1,&first_key); 1625 if (err) { 1626 free_extent_buffer(cur); 1627 return err; 1628 } 1629 } 1630 } 1631 if (search_start == 0) 1632 search_start = last_block; 1633 1634 btrfs_tree_lock(cur); 1635 btrfs_set_lock_blocking_write(cur); 1636 err = __btrfs_cow_block(trans, root, cur, parent, i, 1637 &cur, search_start, 1638 min(16 * blocksize, 1639 (end_slot - i) * blocksize)); 1640 if (err) { 1641 btrfs_tree_unlock(cur); 1642 free_extent_buffer(cur); 1643 break; 1644 } 1645 search_start = cur->start; 1646 last_block = cur->start; 1647 *last_ret = search_start; 1648 btrfs_tree_unlock(cur); 1649 free_extent_buffer(cur); 1650 } 1651 return err; 1652 } 1653 1654 /* 1655 * search for key in the extent_buffer. The items start at offset p, 1656 * and they are item_size apart. There are 'max' items in p. 1657 * 1658 * the slot in the array is returned via slot, and it points to 1659 * the place where you would insert key if it is not found in 1660 * the array. 1661 * 1662 * slot may point to max if the key is bigger than all of the keys 1663 */ 1664 static noinline int generic_bin_search(struct extent_buffer *eb, 1665 unsigned long p, int item_size, 1666 const struct btrfs_key *key, 1667 int max, int *slot) 1668 { 1669 int low = 0; 1670 int high = max; 1671 int ret; 1672 const int key_size = sizeof(struct btrfs_disk_key); 1673 1674 if (low > high) { 1675 btrfs_err(eb->fs_info, 1676 "%s: low (%d) > high (%d) eb %llu owner %llu level %d", 1677 __func__, low, high, eb->start, 1678 btrfs_header_owner(eb), btrfs_header_level(eb)); 1679 return -EINVAL; 1680 } 1681 1682 while (low < high) { 1683 unsigned long oip; 1684 unsigned long offset; 1685 struct btrfs_disk_key *tmp; 1686 struct btrfs_disk_key unaligned; 1687 int mid; 1688 1689 mid = (low + high) / 2; 1690 offset = p + mid * item_size; 1691 oip = offset_in_page(offset); 1692 1693 if (oip + key_size <= PAGE_SIZE) { 1694 const unsigned long idx = offset >> PAGE_SHIFT; 1695 char *kaddr = page_address(eb->pages[idx]); 1696 1697 tmp = (struct btrfs_disk_key *)(kaddr + oip); 1698 } else { 1699 read_extent_buffer(eb, &unaligned, offset, key_size); 1700 tmp = &unaligned; 1701 } 1702 1703 ret = comp_keys(tmp, key); 1704 1705 if (ret < 0) 1706 low = mid + 1; 1707 else if (ret > 0) 1708 high = mid; 1709 else { 1710 *slot = mid; 1711 return 0; 1712 } 1713 } 1714 *slot = low; 1715 return 1; 1716 } 1717 1718 /* 1719 * simple bin_search frontend that does the right thing for 1720 * leaves vs nodes 1721 */ 1722 int btrfs_bin_search(struct extent_buffer *eb, const struct btrfs_key *key, 1723 int *slot) 1724 { 1725 if (btrfs_header_level(eb) == 0) 1726 return generic_bin_search(eb, 1727 offsetof(struct btrfs_leaf, items), 1728 sizeof(struct btrfs_item), 1729 key, btrfs_header_nritems(eb), 1730 slot); 1731 else 1732 return generic_bin_search(eb, 1733 offsetof(struct btrfs_node, ptrs), 1734 sizeof(struct btrfs_key_ptr), 1735 key, btrfs_header_nritems(eb), 1736 slot); 1737 } 1738 1739 static void root_add_used(struct btrfs_root *root, u32 size) 1740 { 1741 spin_lock(&root->accounting_lock); 1742 btrfs_set_root_used(&root->root_item, 1743 btrfs_root_used(&root->root_item) + size); 1744 spin_unlock(&root->accounting_lock); 1745 } 1746 1747 static void root_sub_used(struct btrfs_root *root, u32 size) 1748 { 1749 spin_lock(&root->accounting_lock); 1750 btrfs_set_root_used(&root->root_item, 1751 btrfs_root_used(&root->root_item) - size); 1752 spin_unlock(&root->accounting_lock); 1753 } 1754 1755 /* given a node and slot number, this reads the blocks it points to. The 1756 * extent buffer is returned with a reference taken (but unlocked). 1757 */ 1758 struct extent_buffer *btrfs_read_node_slot(struct extent_buffer *parent, 1759 int slot) 1760 { 1761 int level = btrfs_header_level(parent); 1762 struct extent_buffer *eb; 1763 struct btrfs_key first_key; 1764 1765 if (slot < 0 || slot >= btrfs_header_nritems(parent)) 1766 return ERR_PTR(-ENOENT); 1767 1768 BUG_ON(level == 0); 1769 1770 btrfs_node_key_to_cpu(parent, &first_key, slot); 1771 eb = read_tree_block(parent->fs_info, btrfs_node_blockptr(parent, slot), 1772 btrfs_node_ptr_generation(parent, slot), 1773 level - 1, &first_key); 1774 if (!IS_ERR(eb) && !extent_buffer_uptodate(eb)) { 1775 free_extent_buffer(eb); 1776 eb = ERR_PTR(-EIO); 1777 } 1778 1779 return eb; 1780 } 1781 1782 /* 1783 * node level balancing, used to make sure nodes are in proper order for 1784 * item deletion. We balance from the top down, so we have to make sure 1785 * that a deletion won't leave an node completely empty later on. 1786 */ 1787 static noinline int balance_level(struct btrfs_trans_handle *trans, 1788 struct btrfs_root *root, 1789 struct btrfs_path *path, int level) 1790 { 1791 struct btrfs_fs_info *fs_info = root->fs_info; 1792 struct extent_buffer *right = NULL; 1793 struct extent_buffer *mid; 1794 struct extent_buffer *left = NULL; 1795 struct extent_buffer *parent = NULL; 1796 int ret = 0; 1797 int wret; 1798 int pslot; 1799 int orig_slot = path->slots[level]; 1800 u64 orig_ptr; 1801 1802 ASSERT(level > 0); 1803 1804 mid = path->nodes[level]; 1805 1806 WARN_ON(path->locks[level] != BTRFS_WRITE_LOCK && 1807 path->locks[level] != BTRFS_WRITE_LOCK_BLOCKING); 1808 WARN_ON(btrfs_header_generation(mid) != trans->transid); 1809 1810 orig_ptr = btrfs_node_blockptr(mid, orig_slot); 1811 1812 if (level < BTRFS_MAX_LEVEL - 1) { 1813 parent = path->nodes[level + 1]; 1814 pslot = path->slots[level + 1]; 1815 } 1816 1817 /* 1818 * deal with the case where there is only one pointer in the root 1819 * by promoting the node below to a root 1820 */ 1821 if (!parent) { 1822 struct extent_buffer *child; 1823 1824 if (btrfs_header_nritems(mid) != 1) 1825 return 0; 1826 1827 /* promote the child to a root */ 1828 child = btrfs_read_node_slot(mid, 0); 1829 if (IS_ERR(child)) { 1830 ret = PTR_ERR(child); 1831 btrfs_handle_fs_error(fs_info, ret, NULL); 1832 goto enospc; 1833 } 1834 1835 btrfs_tree_lock(child); 1836 btrfs_set_lock_blocking_write(child); 1837 ret = btrfs_cow_block(trans, root, child, mid, 0, &child); 1838 if (ret) { 1839 btrfs_tree_unlock(child); 1840 free_extent_buffer(child); 1841 goto enospc; 1842 } 1843 1844 ret = tree_mod_log_insert_root(root->node, child, 1); 1845 BUG_ON(ret < 0); 1846 rcu_assign_pointer(root->node, child); 1847 1848 add_root_to_dirty_list(root); 1849 btrfs_tree_unlock(child); 1850 1851 path->locks[level] = 0; 1852 path->nodes[level] = NULL; 1853 btrfs_clean_tree_block(mid); 1854 btrfs_tree_unlock(mid); 1855 /* once for the path */ 1856 free_extent_buffer(mid); 1857 1858 root_sub_used(root, mid->len); 1859 btrfs_free_tree_block(trans, root, mid, 0, 1); 1860 /* once for the root ptr */ 1861 free_extent_buffer_stale(mid); 1862 return 0; 1863 } 1864 if (btrfs_header_nritems(mid) > 1865 BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 4) 1866 return 0; 1867 1868 left = btrfs_read_node_slot(parent, pslot - 1); 1869 if (IS_ERR(left)) 1870 left = NULL; 1871 1872 if (left) { 1873 btrfs_tree_lock(left); 1874 btrfs_set_lock_blocking_write(left); 1875 wret = btrfs_cow_block(trans, root, left, 1876 parent, pslot - 1, &left); 1877 if (wret) { 1878 ret = wret; 1879 goto enospc; 1880 } 1881 } 1882 1883 right = btrfs_read_node_slot(parent, pslot + 1); 1884 if (IS_ERR(right)) 1885 right = NULL; 1886 1887 if (right) { 1888 btrfs_tree_lock(right); 1889 btrfs_set_lock_blocking_write(right); 1890 wret = btrfs_cow_block(trans, root, right, 1891 parent, pslot + 1, &right); 1892 if (wret) { 1893 ret = wret; 1894 goto enospc; 1895 } 1896 } 1897 1898 /* first, try to make some room in the middle buffer */ 1899 if (left) { 1900 orig_slot += btrfs_header_nritems(left); 1901 wret = push_node_left(trans, left, mid, 1); 1902 if (wret < 0) 1903 ret = wret; 1904 } 1905 1906 /* 1907 * then try to empty the right most buffer into the middle 1908 */ 1909 if (right) { 1910 wret = push_node_left(trans, mid, right, 1); 1911 if (wret < 0 && wret != -ENOSPC) 1912 ret = wret; 1913 if (btrfs_header_nritems(right) == 0) { 1914 btrfs_clean_tree_block(right); 1915 btrfs_tree_unlock(right); 1916 del_ptr(root, path, level + 1, pslot + 1); 1917 root_sub_used(root, right->len); 1918 btrfs_free_tree_block(trans, root, right, 0, 1); 1919 free_extent_buffer_stale(right); 1920 right = NULL; 1921 } else { 1922 struct btrfs_disk_key right_key; 1923 btrfs_node_key(right, &right_key, 0); 1924 ret = tree_mod_log_insert_key(parent, pslot + 1, 1925 MOD_LOG_KEY_REPLACE, GFP_NOFS); 1926 BUG_ON(ret < 0); 1927 btrfs_set_node_key(parent, &right_key, pslot + 1); 1928 btrfs_mark_buffer_dirty(parent); 1929 } 1930 } 1931 if (btrfs_header_nritems(mid) == 1) { 1932 /* 1933 * we're not allowed to leave a node with one item in the 1934 * tree during a delete. A deletion from lower in the tree 1935 * could try to delete the only pointer in this node. 1936 * So, pull some keys from the left. 1937 * There has to be a left pointer at this point because 1938 * otherwise we would have pulled some pointers from the 1939 * right 1940 */ 1941 if (!left) { 1942 ret = -EROFS; 1943 btrfs_handle_fs_error(fs_info, ret, NULL); 1944 goto enospc; 1945 } 1946 wret = balance_node_right(trans, mid, left); 1947 if (wret < 0) { 1948 ret = wret; 1949 goto enospc; 1950 } 1951 if (wret == 1) { 1952 wret = push_node_left(trans, left, mid, 1); 1953 if (wret < 0) 1954 ret = wret; 1955 } 1956 BUG_ON(wret == 1); 1957 } 1958 if (btrfs_header_nritems(mid) == 0) { 1959 btrfs_clean_tree_block(mid); 1960 btrfs_tree_unlock(mid); 1961 del_ptr(root, path, level + 1, pslot); 1962 root_sub_used(root, mid->len); 1963 btrfs_free_tree_block(trans, root, mid, 0, 1); 1964 free_extent_buffer_stale(mid); 1965 mid = NULL; 1966 } else { 1967 /* update the parent key to reflect our changes */ 1968 struct btrfs_disk_key mid_key; 1969 btrfs_node_key(mid, &mid_key, 0); 1970 ret = tree_mod_log_insert_key(parent, pslot, 1971 MOD_LOG_KEY_REPLACE, GFP_NOFS); 1972 BUG_ON(ret < 0); 1973 btrfs_set_node_key(parent, &mid_key, pslot); 1974 btrfs_mark_buffer_dirty(parent); 1975 } 1976 1977 /* update the path */ 1978 if (left) { 1979 if (btrfs_header_nritems(left) > orig_slot) { 1980 atomic_inc(&left->refs); 1981 /* left was locked after cow */ 1982 path->nodes[level] = left; 1983 path->slots[level + 1] -= 1; 1984 path->slots[level] = orig_slot; 1985 if (mid) { 1986 btrfs_tree_unlock(mid); 1987 free_extent_buffer(mid); 1988 } 1989 } else { 1990 orig_slot -= btrfs_header_nritems(left); 1991 path->slots[level] = orig_slot; 1992 } 1993 } 1994 /* double check we haven't messed things up */ 1995 if (orig_ptr != 1996 btrfs_node_blockptr(path->nodes[level], path->slots[level])) 1997 BUG(); 1998 enospc: 1999 if (right) { 2000 btrfs_tree_unlock(right); 2001 free_extent_buffer(right); 2002 } 2003 if (left) { 2004 if (path->nodes[level] != left) 2005 btrfs_tree_unlock(left); 2006 free_extent_buffer(left); 2007 } 2008 return ret; 2009 } 2010 2011 /* Node balancing for insertion. Here we only split or push nodes around 2012 * when they are completely full. This is also done top down, so we 2013 * have to be pessimistic. 2014 */ 2015 static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans, 2016 struct btrfs_root *root, 2017 struct btrfs_path *path, int level) 2018 { 2019 struct btrfs_fs_info *fs_info = root->fs_info; 2020 struct extent_buffer *right = NULL; 2021 struct extent_buffer *mid; 2022 struct extent_buffer *left = NULL; 2023 struct extent_buffer *parent = NULL; 2024 int ret = 0; 2025 int wret; 2026 int pslot; 2027 int orig_slot = path->slots[level]; 2028 2029 if (level == 0) 2030 return 1; 2031 2032 mid = path->nodes[level]; 2033 WARN_ON(btrfs_header_generation(mid) != trans->transid); 2034 2035 if (level < BTRFS_MAX_LEVEL - 1) { 2036 parent = path->nodes[level + 1]; 2037 pslot = path->slots[level + 1]; 2038 } 2039 2040 if (!parent) 2041 return 1; 2042 2043 left = btrfs_read_node_slot(parent, pslot - 1); 2044 if (IS_ERR(left)) 2045 left = NULL; 2046 2047 /* first, try to make some room in the middle buffer */ 2048 if (left) { 2049 u32 left_nr; 2050 2051 btrfs_tree_lock(left); 2052 btrfs_set_lock_blocking_write(left); 2053 2054 left_nr = btrfs_header_nritems(left); 2055 if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) { 2056 wret = 1; 2057 } else { 2058 ret = btrfs_cow_block(trans, root, left, parent, 2059 pslot - 1, &left); 2060 if (ret) 2061 wret = 1; 2062 else { 2063 wret = push_node_left(trans, left, mid, 0); 2064 } 2065 } 2066 if (wret < 0) 2067 ret = wret; 2068 if (wret == 0) { 2069 struct btrfs_disk_key disk_key; 2070 orig_slot += left_nr; 2071 btrfs_node_key(mid, &disk_key, 0); 2072 ret = tree_mod_log_insert_key(parent, pslot, 2073 MOD_LOG_KEY_REPLACE, GFP_NOFS); 2074 BUG_ON(ret < 0); 2075 btrfs_set_node_key(parent, &disk_key, pslot); 2076 btrfs_mark_buffer_dirty(parent); 2077 if (btrfs_header_nritems(left) > orig_slot) { 2078 path->nodes[level] = left; 2079 path->slots[level + 1] -= 1; 2080 path->slots[level] = orig_slot; 2081 btrfs_tree_unlock(mid); 2082 free_extent_buffer(mid); 2083 } else { 2084 orig_slot -= 2085 btrfs_header_nritems(left); 2086 path->slots[level] = orig_slot; 2087 btrfs_tree_unlock(left); 2088 free_extent_buffer(left); 2089 } 2090 return 0; 2091 } 2092 btrfs_tree_unlock(left); 2093 free_extent_buffer(left); 2094 } 2095 right = btrfs_read_node_slot(parent, pslot + 1); 2096 if (IS_ERR(right)) 2097 right = NULL; 2098 2099 /* 2100 * then try to empty the right most buffer into the middle 2101 */ 2102 if (right) { 2103 u32 right_nr; 2104 2105 btrfs_tree_lock(right); 2106 btrfs_set_lock_blocking_write(right); 2107 2108 right_nr = btrfs_header_nritems(right); 2109 if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) { 2110 wret = 1; 2111 } else { 2112 ret = btrfs_cow_block(trans, root, right, 2113 parent, pslot + 1, 2114 &right); 2115 if (ret) 2116 wret = 1; 2117 else { 2118 wret = balance_node_right(trans, right, mid); 2119 } 2120 } 2121 if (wret < 0) 2122 ret = wret; 2123 if (wret == 0) { 2124 struct btrfs_disk_key disk_key; 2125 2126 btrfs_node_key(right, &disk_key, 0); 2127 ret = tree_mod_log_insert_key(parent, pslot + 1, 2128 MOD_LOG_KEY_REPLACE, GFP_NOFS); 2129 BUG_ON(ret < 0); 2130 btrfs_set_node_key(parent, &disk_key, pslot + 1); 2131 btrfs_mark_buffer_dirty(parent); 2132 2133 if (btrfs_header_nritems(mid) <= orig_slot) { 2134 path->nodes[level] = right; 2135 path->slots[level + 1] += 1; 2136 path->slots[level] = orig_slot - 2137 btrfs_header_nritems(mid); 2138 btrfs_tree_unlock(mid); 2139 free_extent_buffer(mid); 2140 } else { 2141 btrfs_tree_unlock(right); 2142 free_extent_buffer(right); 2143 } 2144 return 0; 2145 } 2146 btrfs_tree_unlock(right); 2147 free_extent_buffer(right); 2148 } 2149 return 1; 2150 } 2151 2152 /* 2153 * readahead one full node of leaves, finding things that are close 2154 * to the block in 'slot', and triggering ra on them. 2155 */ 2156 static void reada_for_search(struct btrfs_fs_info *fs_info, 2157 struct btrfs_path *path, 2158 int level, int slot, u64 objectid) 2159 { 2160 struct extent_buffer *node; 2161 struct btrfs_disk_key disk_key; 2162 u32 nritems; 2163 u64 search; 2164 u64 target; 2165 u64 nread = 0; 2166 struct extent_buffer *eb; 2167 u32 nr; 2168 u32 blocksize; 2169 u32 nscan = 0; 2170 2171 if (level != 1) 2172 return; 2173 2174 if (!path->nodes[level]) 2175 return; 2176 2177 node = path->nodes[level]; 2178 2179 search = btrfs_node_blockptr(node, slot); 2180 blocksize = fs_info->nodesize; 2181 eb = find_extent_buffer(fs_info, search); 2182 if (eb) { 2183 free_extent_buffer(eb); 2184 return; 2185 } 2186 2187 target = search; 2188 2189 nritems = btrfs_header_nritems(node); 2190 nr = slot; 2191 2192 while (1) { 2193 if (path->reada == READA_BACK) { 2194 if (nr == 0) 2195 break; 2196 nr--; 2197 } else if (path->reada == READA_FORWARD) { 2198 nr++; 2199 if (nr >= nritems) 2200 break; 2201 } 2202 if (path->reada == READA_BACK && objectid) { 2203 btrfs_node_key(node, &disk_key, nr); 2204 if (btrfs_disk_key_objectid(&disk_key) != objectid) 2205 break; 2206 } 2207 search = btrfs_node_blockptr(node, nr); 2208 if ((search <= target && target - search <= 65536) || 2209 (search > target && search - target <= 65536)) { 2210 readahead_tree_block(fs_info, search); 2211 nread += blocksize; 2212 } 2213 nscan++; 2214 if ((nread > 65536 || nscan > 32)) 2215 break; 2216 } 2217 } 2218 2219 static noinline void reada_for_balance(struct btrfs_fs_info *fs_info, 2220 struct btrfs_path *path, int level) 2221 { 2222 int slot; 2223 int nritems; 2224 struct extent_buffer *parent; 2225 struct extent_buffer *eb; 2226 u64 gen; 2227 u64 block1 = 0; 2228 u64 block2 = 0; 2229 2230 parent = path->nodes[level + 1]; 2231 if (!parent) 2232 return; 2233 2234 nritems = btrfs_header_nritems(parent); 2235 slot = path->slots[level + 1]; 2236 2237 if (slot > 0) { 2238 block1 = btrfs_node_blockptr(parent, slot - 1); 2239 gen = btrfs_node_ptr_generation(parent, slot - 1); 2240 eb = find_extent_buffer(fs_info, block1); 2241 /* 2242 * if we get -eagain from btrfs_buffer_uptodate, we 2243 * don't want to return eagain here. That will loop 2244 * forever 2245 */ 2246 if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0) 2247 block1 = 0; 2248 free_extent_buffer(eb); 2249 } 2250 if (slot + 1 < nritems) { 2251 block2 = btrfs_node_blockptr(parent, slot + 1); 2252 gen = btrfs_node_ptr_generation(parent, slot + 1); 2253 eb = find_extent_buffer(fs_info, block2); 2254 if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0) 2255 block2 = 0; 2256 free_extent_buffer(eb); 2257 } 2258 2259 if (block1) 2260 readahead_tree_block(fs_info, block1); 2261 if (block2) 2262 readahead_tree_block(fs_info, block2); 2263 } 2264 2265 2266 /* 2267 * when we walk down the tree, it is usually safe to unlock the higher layers 2268 * in the tree. The exceptions are when our path goes through slot 0, because 2269 * operations on the tree might require changing key pointers higher up in the 2270 * tree. 2271 * 2272 * callers might also have set path->keep_locks, which tells this code to keep 2273 * the lock if the path points to the last slot in the block. This is part of 2274 * walking through the tree, and selecting the next slot in the higher block. 2275 * 2276 * lowest_unlock sets the lowest level in the tree we're allowed to unlock. so 2277 * if lowest_unlock is 1, level 0 won't be unlocked 2278 */ 2279 static noinline void unlock_up(struct btrfs_path *path, int level, 2280 int lowest_unlock, int min_write_lock_level, 2281 int *write_lock_level) 2282 { 2283 int i; 2284 int skip_level = level; 2285 int no_skips = 0; 2286 struct extent_buffer *t; 2287 2288 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 2289 if (!path->nodes[i]) 2290 break; 2291 if (!path->locks[i]) 2292 break; 2293 if (!no_skips && path->slots[i] == 0) { 2294 skip_level = i + 1; 2295 continue; 2296 } 2297 if (!no_skips && path->keep_locks) { 2298 u32 nritems; 2299 t = path->nodes[i]; 2300 nritems = btrfs_header_nritems(t); 2301 if (nritems < 1 || path->slots[i] >= nritems - 1) { 2302 skip_level = i + 1; 2303 continue; 2304 } 2305 } 2306 if (skip_level < i && i >= lowest_unlock) 2307 no_skips = 1; 2308 2309 t = path->nodes[i]; 2310 if (i >= lowest_unlock && i > skip_level) { 2311 btrfs_tree_unlock_rw(t, path->locks[i]); 2312 path->locks[i] = 0; 2313 if (write_lock_level && 2314 i > min_write_lock_level && 2315 i <= *write_lock_level) { 2316 *write_lock_level = i - 1; 2317 } 2318 } 2319 } 2320 } 2321 2322 /* 2323 * helper function for btrfs_search_slot. The goal is to find a block 2324 * in cache without setting the path to blocking. If we find the block 2325 * we return zero and the path is unchanged. 2326 * 2327 * If we can't find the block, we set the path blocking and do some 2328 * reada. -EAGAIN is returned and the search must be repeated. 2329 */ 2330 static int 2331 read_block_for_search(struct btrfs_root *root, struct btrfs_path *p, 2332 struct extent_buffer **eb_ret, int level, int slot, 2333 const struct btrfs_key *key) 2334 { 2335 struct btrfs_fs_info *fs_info = root->fs_info; 2336 u64 blocknr; 2337 u64 gen; 2338 struct extent_buffer *tmp; 2339 struct btrfs_key first_key; 2340 int ret; 2341 int parent_level; 2342 2343 blocknr = btrfs_node_blockptr(*eb_ret, slot); 2344 gen = btrfs_node_ptr_generation(*eb_ret, slot); 2345 parent_level = btrfs_header_level(*eb_ret); 2346 btrfs_node_key_to_cpu(*eb_ret, &first_key, slot); 2347 2348 tmp = find_extent_buffer(fs_info, blocknr); 2349 if (tmp) { 2350 /* first we do an atomic uptodate check */ 2351 if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) { 2352 /* 2353 * Do extra check for first_key, eb can be stale due to 2354 * being cached, read from scrub, or have multiple 2355 * parents (shared tree blocks). 2356 */ 2357 if (btrfs_verify_level_key(tmp, 2358 parent_level - 1, &first_key, gen)) { 2359 free_extent_buffer(tmp); 2360 return -EUCLEAN; 2361 } 2362 *eb_ret = tmp; 2363 return 0; 2364 } 2365 2366 /* the pages were up to date, but we failed 2367 * the generation number check. Do a full 2368 * read for the generation number that is correct. 2369 * We must do this without dropping locks so 2370 * we can trust our generation number 2371 */ 2372 btrfs_set_path_blocking(p); 2373 2374 /* now we're allowed to do a blocking uptodate check */ 2375 ret = btrfs_read_buffer(tmp, gen, parent_level - 1, &first_key); 2376 if (!ret) { 2377 *eb_ret = tmp; 2378 return 0; 2379 } 2380 free_extent_buffer(tmp); 2381 btrfs_release_path(p); 2382 return -EIO; 2383 } 2384 2385 /* 2386 * reduce lock contention at high levels 2387 * of the btree by dropping locks before 2388 * we read. Don't release the lock on the current 2389 * level because we need to walk this node to figure 2390 * out which blocks to read. 2391 */ 2392 btrfs_unlock_up_safe(p, level + 1); 2393 btrfs_set_path_blocking(p); 2394 2395 if (p->reada != READA_NONE) 2396 reada_for_search(fs_info, p, level, slot, key->objectid); 2397 2398 ret = -EAGAIN; 2399 tmp = read_tree_block(fs_info, blocknr, gen, parent_level - 1, 2400 &first_key); 2401 if (!IS_ERR(tmp)) { 2402 /* 2403 * If the read above didn't mark this buffer up to date, 2404 * it will never end up being up to date. Set ret to EIO now 2405 * and give up so that our caller doesn't loop forever 2406 * on our EAGAINs. 2407 */ 2408 if (!extent_buffer_uptodate(tmp)) 2409 ret = -EIO; 2410 free_extent_buffer(tmp); 2411 } else { 2412 ret = PTR_ERR(tmp); 2413 } 2414 2415 btrfs_release_path(p); 2416 return ret; 2417 } 2418 2419 /* 2420 * helper function for btrfs_search_slot. This does all of the checks 2421 * for node-level blocks and does any balancing required based on 2422 * the ins_len. 2423 * 2424 * If no extra work was required, zero is returned. If we had to 2425 * drop the path, -EAGAIN is returned and btrfs_search_slot must 2426 * start over 2427 */ 2428 static int 2429 setup_nodes_for_search(struct btrfs_trans_handle *trans, 2430 struct btrfs_root *root, struct btrfs_path *p, 2431 struct extent_buffer *b, int level, int ins_len, 2432 int *write_lock_level) 2433 { 2434 struct btrfs_fs_info *fs_info = root->fs_info; 2435 int ret; 2436 2437 if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >= 2438 BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) { 2439 int sret; 2440 2441 if (*write_lock_level < level + 1) { 2442 *write_lock_level = level + 1; 2443 btrfs_release_path(p); 2444 goto again; 2445 } 2446 2447 btrfs_set_path_blocking(p); 2448 reada_for_balance(fs_info, p, level); 2449 sret = split_node(trans, root, p, level); 2450 2451 BUG_ON(sret > 0); 2452 if (sret) { 2453 ret = sret; 2454 goto done; 2455 } 2456 b = p->nodes[level]; 2457 } else if (ins_len < 0 && btrfs_header_nritems(b) < 2458 BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 2) { 2459 int sret; 2460 2461 if (*write_lock_level < level + 1) { 2462 *write_lock_level = level + 1; 2463 btrfs_release_path(p); 2464 goto again; 2465 } 2466 2467 btrfs_set_path_blocking(p); 2468 reada_for_balance(fs_info, p, level); 2469 sret = balance_level(trans, root, p, level); 2470 2471 if (sret) { 2472 ret = sret; 2473 goto done; 2474 } 2475 b = p->nodes[level]; 2476 if (!b) { 2477 btrfs_release_path(p); 2478 goto again; 2479 } 2480 BUG_ON(btrfs_header_nritems(b) == 1); 2481 } 2482 return 0; 2483 2484 again: 2485 ret = -EAGAIN; 2486 done: 2487 return ret; 2488 } 2489 2490 int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path, 2491 u64 iobjectid, u64 ioff, u8 key_type, 2492 struct btrfs_key *found_key) 2493 { 2494 int ret; 2495 struct btrfs_key key; 2496 struct extent_buffer *eb; 2497 2498 ASSERT(path); 2499 ASSERT(found_key); 2500 2501 key.type = key_type; 2502 key.objectid = iobjectid; 2503 key.offset = ioff; 2504 2505 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0); 2506 if (ret < 0) 2507 return ret; 2508 2509 eb = path->nodes[0]; 2510 if (ret && path->slots[0] >= btrfs_header_nritems(eb)) { 2511 ret = btrfs_next_leaf(fs_root, path); 2512 if (ret) 2513 return ret; 2514 eb = path->nodes[0]; 2515 } 2516 2517 btrfs_item_key_to_cpu(eb, found_key, path->slots[0]); 2518 if (found_key->type != key.type || 2519 found_key->objectid != key.objectid) 2520 return 1; 2521 2522 return 0; 2523 } 2524 2525 static struct extent_buffer *btrfs_search_slot_get_root(struct btrfs_root *root, 2526 struct btrfs_path *p, 2527 int write_lock_level) 2528 { 2529 struct btrfs_fs_info *fs_info = root->fs_info; 2530 struct extent_buffer *b; 2531 int root_lock; 2532 int level = 0; 2533 2534 /* We try very hard to do read locks on the root */ 2535 root_lock = BTRFS_READ_LOCK; 2536 2537 if (p->search_commit_root) { 2538 /* 2539 * The commit roots are read only so we always do read locks, 2540 * and we always must hold the commit_root_sem when doing 2541 * searches on them, the only exception is send where we don't 2542 * want to block transaction commits for a long time, so 2543 * we need to clone the commit root in order to avoid races 2544 * with transaction commits that create a snapshot of one of 2545 * the roots used by a send operation. 2546 */ 2547 if (p->need_commit_sem) { 2548 down_read(&fs_info->commit_root_sem); 2549 b = btrfs_clone_extent_buffer(root->commit_root); 2550 up_read(&fs_info->commit_root_sem); 2551 if (!b) 2552 return ERR_PTR(-ENOMEM); 2553 2554 } else { 2555 b = root->commit_root; 2556 atomic_inc(&b->refs); 2557 } 2558 level = btrfs_header_level(b); 2559 /* 2560 * Ensure that all callers have set skip_locking when 2561 * p->search_commit_root = 1. 2562 */ 2563 ASSERT(p->skip_locking == 1); 2564 2565 goto out; 2566 } 2567 2568 if (p->skip_locking) { 2569 b = btrfs_root_node(root); 2570 level = btrfs_header_level(b); 2571 goto out; 2572 } 2573 2574 /* 2575 * If the level is set to maximum, we can skip trying to get the read 2576 * lock. 2577 */ 2578 if (write_lock_level < BTRFS_MAX_LEVEL) { 2579 /* 2580 * We don't know the level of the root node until we actually 2581 * have it read locked 2582 */ 2583 b = btrfs_read_lock_root_node(root); 2584 level = btrfs_header_level(b); 2585 if (level > write_lock_level) 2586 goto out; 2587 2588 /* Whoops, must trade for write lock */ 2589 btrfs_tree_read_unlock(b); 2590 free_extent_buffer(b); 2591 } 2592 2593 b = btrfs_lock_root_node(root); 2594 root_lock = BTRFS_WRITE_LOCK; 2595 2596 /* The level might have changed, check again */ 2597 level = btrfs_header_level(b); 2598 2599 out: 2600 p->nodes[level] = b; 2601 if (!p->skip_locking) 2602 p->locks[level] = root_lock; 2603 /* 2604 * Callers are responsible for dropping b's references. 2605 */ 2606 return b; 2607 } 2608 2609 2610 /* 2611 * btrfs_search_slot - look for a key in a tree and perform necessary 2612 * modifications to preserve tree invariants. 2613 * 2614 * @trans: Handle of transaction, used when modifying the tree 2615 * @p: Holds all btree nodes along the search path 2616 * @root: The root node of the tree 2617 * @key: The key we are looking for 2618 * @ins_len: Indicates purpose of search, for inserts it is 1, for 2619 * deletions it's -1. 0 for plain searches 2620 * @cow: boolean should CoW operations be performed. Must always be 1 2621 * when modifying the tree. 2622 * 2623 * If @ins_len > 0, nodes and leaves will be split as we walk down the tree. 2624 * If @ins_len < 0, nodes will be merged as we walk down the tree (if possible) 2625 * 2626 * If @key is found, 0 is returned and you can find the item in the leaf level 2627 * of the path (level 0) 2628 * 2629 * If @key isn't found, 1 is returned and the leaf level of the path (level 0) 2630 * points to the slot where it should be inserted 2631 * 2632 * If an error is encountered while searching the tree a negative error number 2633 * is returned 2634 */ 2635 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, 2636 const struct btrfs_key *key, struct btrfs_path *p, 2637 int ins_len, int cow) 2638 { 2639 struct extent_buffer *b; 2640 int slot; 2641 int ret; 2642 int err; 2643 int level; 2644 int lowest_unlock = 1; 2645 /* everything at write_lock_level or lower must be write locked */ 2646 int write_lock_level = 0; 2647 u8 lowest_level = 0; 2648 int min_write_lock_level; 2649 int prev_cmp; 2650 2651 lowest_level = p->lowest_level; 2652 WARN_ON(lowest_level && ins_len > 0); 2653 WARN_ON(p->nodes[0] != NULL); 2654 BUG_ON(!cow && ins_len); 2655 2656 if (ins_len < 0) { 2657 lowest_unlock = 2; 2658 2659 /* when we are removing items, we might have to go up to level 2660 * two as we update tree pointers Make sure we keep write 2661 * for those levels as well 2662 */ 2663 write_lock_level = 2; 2664 } else if (ins_len > 0) { 2665 /* 2666 * for inserting items, make sure we have a write lock on 2667 * level 1 so we can update keys 2668 */ 2669 write_lock_level = 1; 2670 } 2671 2672 if (!cow) 2673 write_lock_level = -1; 2674 2675 if (cow && (p->keep_locks || p->lowest_level)) 2676 write_lock_level = BTRFS_MAX_LEVEL; 2677 2678 min_write_lock_level = write_lock_level; 2679 2680 again: 2681 prev_cmp = -1; 2682 b = btrfs_search_slot_get_root(root, p, write_lock_level); 2683 if (IS_ERR(b)) { 2684 ret = PTR_ERR(b); 2685 goto done; 2686 } 2687 2688 while (b) { 2689 int dec = 0; 2690 2691 level = btrfs_header_level(b); 2692 2693 if (cow) { 2694 bool last_level = (level == (BTRFS_MAX_LEVEL - 1)); 2695 2696 /* 2697 * if we don't really need to cow this block 2698 * then we don't want to set the path blocking, 2699 * so we test it here 2700 */ 2701 if (!should_cow_block(trans, root, b)) { 2702 trans->dirty = true; 2703 goto cow_done; 2704 } 2705 2706 /* 2707 * must have write locks on this node and the 2708 * parent 2709 */ 2710 if (level > write_lock_level || 2711 (level + 1 > write_lock_level && 2712 level + 1 < BTRFS_MAX_LEVEL && 2713 p->nodes[level + 1])) { 2714 write_lock_level = level + 1; 2715 btrfs_release_path(p); 2716 goto again; 2717 } 2718 2719 btrfs_set_path_blocking(p); 2720 if (last_level) 2721 err = btrfs_cow_block(trans, root, b, NULL, 0, 2722 &b); 2723 else 2724 err = btrfs_cow_block(trans, root, b, 2725 p->nodes[level + 1], 2726 p->slots[level + 1], &b); 2727 if (err) { 2728 ret = err; 2729 goto done; 2730 } 2731 } 2732 cow_done: 2733 p->nodes[level] = b; 2734 /* 2735 * Leave path with blocking locks to avoid massive 2736 * lock context switch, this is made on purpose. 2737 */ 2738 2739 /* 2740 * we have a lock on b and as long as we aren't changing 2741 * the tree, there is no way to for the items in b to change. 2742 * It is safe to drop the lock on our parent before we 2743 * go through the expensive btree search on b. 2744 * 2745 * If we're inserting or deleting (ins_len != 0), then we might 2746 * be changing slot zero, which may require changing the parent. 2747 * So, we can't drop the lock until after we know which slot 2748 * we're operating on. 2749 */ 2750 if (!ins_len && !p->keep_locks) { 2751 int u = level + 1; 2752 2753 if (u < BTRFS_MAX_LEVEL && p->locks[u]) { 2754 btrfs_tree_unlock_rw(p->nodes[u], p->locks[u]); 2755 p->locks[u] = 0; 2756 } 2757 } 2758 2759 /* 2760 * If btrfs_bin_search returns an exact match (prev_cmp == 0) 2761 * we can safely assume the target key will always be in slot 0 2762 * on lower levels due to the invariants BTRFS' btree provides, 2763 * namely that a btrfs_key_ptr entry always points to the 2764 * lowest key in the child node, thus we can skip searching 2765 * lower levels 2766 */ 2767 if (prev_cmp == 0) { 2768 slot = 0; 2769 ret = 0; 2770 } else { 2771 ret = btrfs_bin_search(b, key, &slot); 2772 prev_cmp = ret; 2773 if (ret < 0) 2774 goto done; 2775 } 2776 2777 if (level == 0) { 2778 p->slots[level] = slot; 2779 if (ins_len > 0 && 2780 btrfs_leaf_free_space(b) < ins_len) { 2781 if (write_lock_level < 1) { 2782 write_lock_level = 1; 2783 btrfs_release_path(p); 2784 goto again; 2785 } 2786 2787 btrfs_set_path_blocking(p); 2788 err = split_leaf(trans, root, key, 2789 p, ins_len, ret == 0); 2790 2791 BUG_ON(err > 0); 2792 if (err) { 2793 ret = err; 2794 goto done; 2795 } 2796 } 2797 if (!p->search_for_split) 2798 unlock_up(p, level, lowest_unlock, 2799 min_write_lock_level, NULL); 2800 goto done; 2801 } 2802 if (ret && slot > 0) { 2803 dec = 1; 2804 slot--; 2805 } 2806 p->slots[level] = slot; 2807 err = setup_nodes_for_search(trans, root, p, b, level, ins_len, 2808 &write_lock_level); 2809 if (err == -EAGAIN) 2810 goto again; 2811 if (err) { 2812 ret = err; 2813 goto done; 2814 } 2815 b = p->nodes[level]; 2816 slot = p->slots[level]; 2817 2818 /* 2819 * Slot 0 is special, if we change the key we have to update 2820 * the parent pointer which means we must have a write lock on 2821 * the parent 2822 */ 2823 if (slot == 0 && ins_len && write_lock_level < level + 1) { 2824 write_lock_level = level + 1; 2825 btrfs_release_path(p); 2826 goto again; 2827 } 2828 2829 unlock_up(p, level, lowest_unlock, min_write_lock_level, 2830 &write_lock_level); 2831 2832 if (level == lowest_level) { 2833 if (dec) 2834 p->slots[level]++; 2835 goto done; 2836 } 2837 2838 err = read_block_for_search(root, p, &b, level, slot, key); 2839 if (err == -EAGAIN) 2840 goto again; 2841 if (err) { 2842 ret = err; 2843 goto done; 2844 } 2845 2846 if (!p->skip_locking) { 2847 level = btrfs_header_level(b); 2848 if (level <= write_lock_level) { 2849 if (!btrfs_try_tree_write_lock(b)) { 2850 btrfs_set_path_blocking(p); 2851 btrfs_tree_lock(b); 2852 } 2853 p->locks[level] = BTRFS_WRITE_LOCK; 2854 } else { 2855 if (!btrfs_tree_read_lock_atomic(b)) { 2856 btrfs_set_path_blocking(p); 2857 btrfs_tree_read_lock(b); 2858 } 2859 p->locks[level] = BTRFS_READ_LOCK; 2860 } 2861 p->nodes[level] = b; 2862 } 2863 } 2864 ret = 1; 2865 done: 2866 /* 2867 * we don't really know what they plan on doing with the path 2868 * from here on, so for now just mark it as blocking 2869 */ 2870 if (!p->leave_spinning) 2871 btrfs_set_path_blocking(p); 2872 if (ret < 0 && !p->skip_release_on_error) 2873 btrfs_release_path(p); 2874 return ret; 2875 } 2876 2877 /* 2878 * Like btrfs_search_slot, this looks for a key in the given tree. It uses the 2879 * current state of the tree together with the operations recorded in the tree 2880 * modification log to search for the key in a previous version of this tree, as 2881 * denoted by the time_seq parameter. 2882 * 2883 * Naturally, there is no support for insert, delete or cow operations. 2884 * 2885 * The resulting path and return value will be set up as if we called 2886 * btrfs_search_slot at that point in time with ins_len and cow both set to 0. 2887 */ 2888 int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key, 2889 struct btrfs_path *p, u64 time_seq) 2890 { 2891 struct btrfs_fs_info *fs_info = root->fs_info; 2892 struct extent_buffer *b; 2893 int slot; 2894 int ret; 2895 int err; 2896 int level; 2897 int lowest_unlock = 1; 2898 u8 lowest_level = 0; 2899 2900 lowest_level = p->lowest_level; 2901 WARN_ON(p->nodes[0] != NULL); 2902 2903 if (p->search_commit_root) { 2904 BUG_ON(time_seq); 2905 return btrfs_search_slot(NULL, root, key, p, 0, 0); 2906 } 2907 2908 again: 2909 b = get_old_root(root, time_seq); 2910 if (!b) { 2911 ret = -EIO; 2912 goto done; 2913 } 2914 level = btrfs_header_level(b); 2915 p->locks[level] = BTRFS_READ_LOCK; 2916 2917 while (b) { 2918 int dec = 0; 2919 2920 level = btrfs_header_level(b); 2921 p->nodes[level] = b; 2922 2923 /* 2924 * we have a lock on b and as long as we aren't changing 2925 * the tree, there is no way to for the items in b to change. 2926 * It is safe to drop the lock on our parent before we 2927 * go through the expensive btree search on b. 2928 */ 2929 btrfs_unlock_up_safe(p, level + 1); 2930 2931 ret = btrfs_bin_search(b, key, &slot); 2932 if (ret < 0) 2933 goto done; 2934 2935 if (level == 0) { 2936 p->slots[level] = slot; 2937 unlock_up(p, level, lowest_unlock, 0, NULL); 2938 goto done; 2939 } 2940 2941 if (ret && slot > 0) { 2942 dec = 1; 2943 slot--; 2944 } 2945 p->slots[level] = slot; 2946 unlock_up(p, level, lowest_unlock, 0, NULL); 2947 2948 if (level == lowest_level) { 2949 if (dec) 2950 p->slots[level]++; 2951 goto done; 2952 } 2953 2954 err = read_block_for_search(root, p, &b, level, slot, key); 2955 if (err == -EAGAIN) 2956 goto again; 2957 if (err) { 2958 ret = err; 2959 goto done; 2960 } 2961 2962 level = btrfs_header_level(b); 2963 if (!btrfs_tree_read_lock_atomic(b)) { 2964 btrfs_set_path_blocking(p); 2965 btrfs_tree_read_lock(b); 2966 } 2967 b = tree_mod_log_rewind(fs_info, p, b, time_seq); 2968 if (!b) { 2969 ret = -ENOMEM; 2970 goto done; 2971 } 2972 p->locks[level] = BTRFS_READ_LOCK; 2973 p->nodes[level] = b; 2974 } 2975 ret = 1; 2976 done: 2977 if (!p->leave_spinning) 2978 btrfs_set_path_blocking(p); 2979 if (ret < 0) 2980 btrfs_release_path(p); 2981 2982 return ret; 2983 } 2984 2985 /* 2986 * helper to use instead of search slot if no exact match is needed but 2987 * instead the next or previous item should be returned. 2988 * When find_higher is true, the next higher item is returned, the next lower 2989 * otherwise. 2990 * When return_any and find_higher are both true, and no higher item is found, 2991 * return the next lower instead. 2992 * When return_any is true and find_higher is false, and no lower item is found, 2993 * return the next higher instead. 2994 * It returns 0 if any item is found, 1 if none is found (tree empty), and 2995 * < 0 on error 2996 */ 2997 int btrfs_search_slot_for_read(struct btrfs_root *root, 2998 const struct btrfs_key *key, 2999 struct btrfs_path *p, int find_higher, 3000 int return_any) 3001 { 3002 int ret; 3003 struct extent_buffer *leaf; 3004 3005 again: 3006 ret = btrfs_search_slot(NULL, root, key, p, 0, 0); 3007 if (ret <= 0) 3008 return ret; 3009 /* 3010 * a return value of 1 means the path is at the position where the 3011 * item should be inserted. Normally this is the next bigger item, 3012 * but in case the previous item is the last in a leaf, path points 3013 * to the first free slot in the previous leaf, i.e. at an invalid 3014 * item. 3015 */ 3016 leaf = p->nodes[0]; 3017 3018 if (find_higher) { 3019 if (p->slots[0] >= btrfs_header_nritems(leaf)) { 3020 ret = btrfs_next_leaf(root, p); 3021 if (ret <= 0) 3022 return ret; 3023 if (!return_any) 3024 return 1; 3025 /* 3026 * no higher item found, return the next 3027 * lower instead 3028 */ 3029 return_any = 0; 3030 find_higher = 0; 3031 btrfs_release_path(p); 3032 goto again; 3033 } 3034 } else { 3035 if (p->slots[0] == 0) { 3036 ret = btrfs_prev_leaf(root, p); 3037 if (ret < 0) 3038 return ret; 3039 if (!ret) { 3040 leaf = p->nodes[0]; 3041 if (p->slots[0] == btrfs_header_nritems(leaf)) 3042 p->slots[0]--; 3043 return 0; 3044 } 3045 if (!return_any) 3046 return 1; 3047 /* 3048 * no lower item found, return the next 3049 * higher instead 3050 */ 3051 return_any = 0; 3052 find_higher = 1; 3053 btrfs_release_path(p); 3054 goto again; 3055 } else { 3056 --p->slots[0]; 3057 } 3058 } 3059 return 0; 3060 } 3061 3062 /* 3063 * adjust the pointers going up the tree, starting at level 3064 * making sure the right key of each node is points to 'key'. 3065 * This is used after shifting pointers to the left, so it stops 3066 * fixing up pointers when a given leaf/node is not in slot 0 of the 3067 * higher levels 3068 * 3069 */ 3070 static void fixup_low_keys(struct btrfs_path *path, 3071 struct btrfs_disk_key *key, int level) 3072 { 3073 int i; 3074 struct extent_buffer *t; 3075 int ret; 3076 3077 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 3078 int tslot = path->slots[i]; 3079 3080 if (!path->nodes[i]) 3081 break; 3082 t = path->nodes[i]; 3083 ret = tree_mod_log_insert_key(t, tslot, MOD_LOG_KEY_REPLACE, 3084 GFP_ATOMIC); 3085 BUG_ON(ret < 0); 3086 btrfs_set_node_key(t, key, tslot); 3087 btrfs_mark_buffer_dirty(path->nodes[i]); 3088 if (tslot != 0) 3089 break; 3090 } 3091 } 3092 3093 /* 3094 * update item key. 3095 * 3096 * This function isn't completely safe. It's the caller's responsibility 3097 * that the new key won't break the order 3098 */ 3099 void btrfs_set_item_key_safe(struct btrfs_fs_info *fs_info, 3100 struct btrfs_path *path, 3101 const struct btrfs_key *new_key) 3102 { 3103 struct btrfs_disk_key disk_key; 3104 struct extent_buffer *eb; 3105 int slot; 3106 3107 eb = path->nodes[0]; 3108 slot = path->slots[0]; 3109 if (slot > 0) { 3110 btrfs_item_key(eb, &disk_key, slot - 1); 3111 if (unlikely(comp_keys(&disk_key, new_key) >= 0)) { 3112 btrfs_crit(fs_info, 3113 "slot %u key (%llu %u %llu) new key (%llu %u %llu)", 3114 slot, btrfs_disk_key_objectid(&disk_key), 3115 btrfs_disk_key_type(&disk_key), 3116 btrfs_disk_key_offset(&disk_key), 3117 new_key->objectid, new_key->type, 3118 new_key->offset); 3119 btrfs_print_leaf(eb); 3120 BUG(); 3121 } 3122 } 3123 if (slot < btrfs_header_nritems(eb) - 1) { 3124 btrfs_item_key(eb, &disk_key, slot + 1); 3125 if (unlikely(comp_keys(&disk_key, new_key) <= 0)) { 3126 btrfs_crit(fs_info, 3127 "slot %u key (%llu %u %llu) new key (%llu %u %llu)", 3128 slot, btrfs_disk_key_objectid(&disk_key), 3129 btrfs_disk_key_type(&disk_key), 3130 btrfs_disk_key_offset(&disk_key), 3131 new_key->objectid, new_key->type, 3132 new_key->offset); 3133 btrfs_print_leaf(eb); 3134 BUG(); 3135 } 3136 } 3137 3138 btrfs_cpu_key_to_disk(&disk_key, new_key); 3139 btrfs_set_item_key(eb, &disk_key, slot); 3140 btrfs_mark_buffer_dirty(eb); 3141 if (slot == 0) 3142 fixup_low_keys(path, &disk_key, 1); 3143 } 3144 3145 /* 3146 * try to push data from one node into the next node left in the 3147 * tree. 3148 * 3149 * returns 0 if some ptrs were pushed left, < 0 if there was some horrible 3150 * error, and > 0 if there was no room in the left hand block. 3151 */ 3152 static int push_node_left(struct btrfs_trans_handle *trans, 3153 struct extent_buffer *dst, 3154 struct extent_buffer *src, int empty) 3155 { 3156 struct btrfs_fs_info *fs_info = trans->fs_info; 3157 int push_items = 0; 3158 int src_nritems; 3159 int dst_nritems; 3160 int ret = 0; 3161 3162 src_nritems = btrfs_header_nritems(src); 3163 dst_nritems = btrfs_header_nritems(dst); 3164 push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems; 3165 WARN_ON(btrfs_header_generation(src) != trans->transid); 3166 WARN_ON(btrfs_header_generation(dst) != trans->transid); 3167 3168 if (!empty && src_nritems <= 8) 3169 return 1; 3170 3171 if (push_items <= 0) 3172 return 1; 3173 3174 if (empty) { 3175 push_items = min(src_nritems, push_items); 3176 if (push_items < src_nritems) { 3177 /* leave at least 8 pointers in the node if 3178 * we aren't going to empty it 3179 */ 3180 if (src_nritems - push_items < 8) { 3181 if (push_items <= 8) 3182 return 1; 3183 push_items -= 8; 3184 } 3185 } 3186 } else 3187 push_items = min(src_nritems - 8, push_items); 3188 3189 ret = tree_mod_log_eb_copy(dst, src, dst_nritems, 0, push_items); 3190 if (ret) { 3191 btrfs_abort_transaction(trans, ret); 3192 return ret; 3193 } 3194 copy_extent_buffer(dst, src, 3195 btrfs_node_key_ptr_offset(dst_nritems), 3196 btrfs_node_key_ptr_offset(0), 3197 push_items * sizeof(struct btrfs_key_ptr)); 3198 3199 if (push_items < src_nritems) { 3200 /* 3201 * Don't call tree_mod_log_insert_move here, key removal was 3202 * already fully logged by tree_mod_log_eb_copy above. 3203 */ 3204 memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0), 3205 btrfs_node_key_ptr_offset(push_items), 3206 (src_nritems - push_items) * 3207 sizeof(struct btrfs_key_ptr)); 3208 } 3209 btrfs_set_header_nritems(src, src_nritems - push_items); 3210 btrfs_set_header_nritems(dst, dst_nritems + push_items); 3211 btrfs_mark_buffer_dirty(src); 3212 btrfs_mark_buffer_dirty(dst); 3213 3214 return ret; 3215 } 3216 3217 /* 3218 * try to push data from one node into the next node right in the 3219 * tree. 3220 * 3221 * returns 0 if some ptrs were pushed, < 0 if there was some horrible 3222 * error, and > 0 if there was no room in the right hand block. 3223 * 3224 * this will only push up to 1/2 the contents of the left node over 3225 */ 3226 static int balance_node_right(struct btrfs_trans_handle *trans, 3227 struct extent_buffer *dst, 3228 struct extent_buffer *src) 3229 { 3230 struct btrfs_fs_info *fs_info = trans->fs_info; 3231 int push_items = 0; 3232 int max_push; 3233 int src_nritems; 3234 int dst_nritems; 3235 int ret = 0; 3236 3237 WARN_ON(btrfs_header_generation(src) != trans->transid); 3238 WARN_ON(btrfs_header_generation(dst) != trans->transid); 3239 3240 src_nritems = btrfs_header_nritems(src); 3241 dst_nritems = btrfs_header_nritems(dst); 3242 push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems; 3243 if (push_items <= 0) 3244 return 1; 3245 3246 if (src_nritems < 4) 3247 return 1; 3248 3249 max_push = src_nritems / 2 + 1; 3250 /* don't try to empty the node */ 3251 if (max_push >= src_nritems) 3252 return 1; 3253 3254 if (max_push < push_items) 3255 push_items = max_push; 3256 3257 ret = tree_mod_log_insert_move(dst, push_items, 0, dst_nritems); 3258 BUG_ON(ret < 0); 3259 memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items), 3260 btrfs_node_key_ptr_offset(0), 3261 (dst_nritems) * 3262 sizeof(struct btrfs_key_ptr)); 3263 3264 ret = tree_mod_log_eb_copy(dst, src, 0, src_nritems - push_items, 3265 push_items); 3266 if (ret) { 3267 btrfs_abort_transaction(trans, ret); 3268 return ret; 3269 } 3270 copy_extent_buffer(dst, src, 3271 btrfs_node_key_ptr_offset(0), 3272 btrfs_node_key_ptr_offset(src_nritems - push_items), 3273 push_items * sizeof(struct btrfs_key_ptr)); 3274 3275 btrfs_set_header_nritems(src, src_nritems - push_items); 3276 btrfs_set_header_nritems(dst, dst_nritems + push_items); 3277 3278 btrfs_mark_buffer_dirty(src); 3279 btrfs_mark_buffer_dirty(dst); 3280 3281 return ret; 3282 } 3283 3284 /* 3285 * helper function to insert a new root level in the tree. 3286 * A new node is allocated, and a single item is inserted to 3287 * point to the existing root 3288 * 3289 * returns zero on success or < 0 on failure. 3290 */ 3291 static noinline int insert_new_root(struct btrfs_trans_handle *trans, 3292 struct btrfs_root *root, 3293 struct btrfs_path *path, int level) 3294 { 3295 struct btrfs_fs_info *fs_info = root->fs_info; 3296 u64 lower_gen; 3297 struct extent_buffer *lower; 3298 struct extent_buffer *c; 3299 struct extent_buffer *old; 3300 struct btrfs_disk_key lower_key; 3301 int ret; 3302 3303 BUG_ON(path->nodes[level]); 3304 BUG_ON(path->nodes[level-1] != root->node); 3305 3306 lower = path->nodes[level-1]; 3307 if (level == 1) 3308 btrfs_item_key(lower, &lower_key, 0); 3309 else 3310 btrfs_node_key(lower, &lower_key, 0); 3311 3312 c = alloc_tree_block_no_bg_flush(trans, root, 0, &lower_key, level, 3313 root->node->start, 0); 3314 if (IS_ERR(c)) 3315 return PTR_ERR(c); 3316 3317 root_add_used(root, fs_info->nodesize); 3318 3319 btrfs_set_header_nritems(c, 1); 3320 btrfs_set_node_key(c, &lower_key, 0); 3321 btrfs_set_node_blockptr(c, 0, lower->start); 3322 lower_gen = btrfs_header_generation(lower); 3323 WARN_ON(lower_gen != trans->transid); 3324 3325 btrfs_set_node_ptr_generation(c, 0, lower_gen); 3326 3327 btrfs_mark_buffer_dirty(c); 3328 3329 old = root->node; 3330 ret = tree_mod_log_insert_root(root->node, c, 0); 3331 BUG_ON(ret < 0); 3332 rcu_assign_pointer(root->node, c); 3333 3334 /* the super has an extra ref to root->node */ 3335 free_extent_buffer(old); 3336 3337 add_root_to_dirty_list(root); 3338 atomic_inc(&c->refs); 3339 path->nodes[level] = c; 3340 path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING; 3341 path->slots[level] = 0; 3342 return 0; 3343 } 3344 3345 /* 3346 * worker function to insert a single pointer in a node. 3347 * the node should have enough room for the pointer already 3348 * 3349 * slot and level indicate where you want the key to go, and 3350 * blocknr is the block the key points to. 3351 */ 3352 static void insert_ptr(struct btrfs_trans_handle *trans, 3353 struct btrfs_path *path, 3354 struct btrfs_disk_key *key, u64 bytenr, 3355 int slot, int level) 3356 { 3357 struct extent_buffer *lower; 3358 int nritems; 3359 int ret; 3360 3361 BUG_ON(!path->nodes[level]); 3362 btrfs_assert_tree_locked(path->nodes[level]); 3363 lower = path->nodes[level]; 3364 nritems = btrfs_header_nritems(lower); 3365 BUG_ON(slot > nritems); 3366 BUG_ON(nritems == BTRFS_NODEPTRS_PER_BLOCK(trans->fs_info)); 3367 if (slot != nritems) { 3368 if (level) { 3369 ret = tree_mod_log_insert_move(lower, slot + 1, slot, 3370 nritems - slot); 3371 BUG_ON(ret < 0); 3372 } 3373 memmove_extent_buffer(lower, 3374 btrfs_node_key_ptr_offset(slot + 1), 3375 btrfs_node_key_ptr_offset(slot), 3376 (nritems - slot) * sizeof(struct btrfs_key_ptr)); 3377 } 3378 if (level) { 3379 ret = tree_mod_log_insert_key(lower, slot, MOD_LOG_KEY_ADD, 3380 GFP_NOFS); 3381 BUG_ON(ret < 0); 3382 } 3383 btrfs_set_node_key(lower, key, slot); 3384 btrfs_set_node_blockptr(lower, slot, bytenr); 3385 WARN_ON(trans->transid == 0); 3386 btrfs_set_node_ptr_generation(lower, slot, trans->transid); 3387 btrfs_set_header_nritems(lower, nritems + 1); 3388 btrfs_mark_buffer_dirty(lower); 3389 } 3390 3391 /* 3392 * split the node at the specified level in path in two. 3393 * The path is corrected to point to the appropriate node after the split 3394 * 3395 * Before splitting this tries to make some room in the node by pushing 3396 * left and right, if either one works, it returns right away. 3397 * 3398 * returns 0 on success and < 0 on failure 3399 */ 3400 static noinline int split_node(struct btrfs_trans_handle *trans, 3401 struct btrfs_root *root, 3402 struct btrfs_path *path, int level) 3403 { 3404 struct btrfs_fs_info *fs_info = root->fs_info; 3405 struct extent_buffer *c; 3406 struct extent_buffer *split; 3407 struct btrfs_disk_key disk_key; 3408 int mid; 3409 int ret; 3410 u32 c_nritems; 3411 3412 c = path->nodes[level]; 3413 WARN_ON(btrfs_header_generation(c) != trans->transid); 3414 if (c == root->node) { 3415 /* 3416 * trying to split the root, lets make a new one 3417 * 3418 * tree mod log: We don't log_removal old root in 3419 * insert_new_root, because that root buffer will be kept as a 3420 * normal node. We are going to log removal of half of the 3421 * elements below with tree_mod_log_eb_copy. We're holding a 3422 * tree lock on the buffer, which is why we cannot race with 3423 * other tree_mod_log users. 3424 */ 3425 ret = insert_new_root(trans, root, path, level + 1); 3426 if (ret) 3427 return ret; 3428 } else { 3429 ret = push_nodes_for_insert(trans, root, path, level); 3430 c = path->nodes[level]; 3431 if (!ret && btrfs_header_nritems(c) < 3432 BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) 3433 return 0; 3434 if (ret < 0) 3435 return ret; 3436 } 3437 3438 c_nritems = btrfs_header_nritems(c); 3439 mid = (c_nritems + 1) / 2; 3440 btrfs_node_key(c, &disk_key, mid); 3441 3442 split = alloc_tree_block_no_bg_flush(trans, root, 0, &disk_key, level, 3443 c->start, 0); 3444 if (IS_ERR(split)) 3445 return PTR_ERR(split); 3446 3447 root_add_used(root, fs_info->nodesize); 3448 ASSERT(btrfs_header_level(c) == level); 3449 3450 ret = tree_mod_log_eb_copy(split, c, 0, mid, c_nritems - mid); 3451 if (ret) { 3452 btrfs_abort_transaction(trans, ret); 3453 return ret; 3454 } 3455 copy_extent_buffer(split, c, 3456 btrfs_node_key_ptr_offset(0), 3457 btrfs_node_key_ptr_offset(mid), 3458 (c_nritems - mid) * sizeof(struct btrfs_key_ptr)); 3459 btrfs_set_header_nritems(split, c_nritems - mid); 3460 btrfs_set_header_nritems(c, mid); 3461 ret = 0; 3462 3463 btrfs_mark_buffer_dirty(c); 3464 btrfs_mark_buffer_dirty(split); 3465 3466 insert_ptr(trans, path, &disk_key, split->start, 3467 path->slots[level + 1] + 1, level + 1); 3468 3469 if (path->slots[level] >= mid) { 3470 path->slots[level] -= mid; 3471 btrfs_tree_unlock(c); 3472 free_extent_buffer(c); 3473 path->nodes[level] = split; 3474 path->slots[level + 1] += 1; 3475 } else { 3476 btrfs_tree_unlock(split); 3477 free_extent_buffer(split); 3478 } 3479 return ret; 3480 } 3481 3482 /* 3483 * how many bytes are required to store the items in a leaf. start 3484 * and nr indicate which items in the leaf to check. This totals up the 3485 * space used both by the item structs and the item data 3486 */ 3487 static int leaf_space_used(struct extent_buffer *l, int start, int nr) 3488 { 3489 struct btrfs_item *start_item; 3490 struct btrfs_item *end_item; 3491 int data_len; 3492 int nritems = btrfs_header_nritems(l); 3493 int end = min(nritems, start + nr) - 1; 3494 3495 if (!nr) 3496 return 0; 3497 start_item = btrfs_item_nr(start); 3498 end_item = btrfs_item_nr(end); 3499 data_len = btrfs_item_offset(l, start_item) + 3500 btrfs_item_size(l, start_item); 3501 data_len = data_len - btrfs_item_offset(l, end_item); 3502 data_len += sizeof(struct btrfs_item) * nr; 3503 WARN_ON(data_len < 0); 3504 return data_len; 3505 } 3506 3507 /* 3508 * The space between the end of the leaf items and 3509 * the start of the leaf data. IOW, how much room 3510 * the leaf has left for both items and data 3511 */ 3512 noinline int btrfs_leaf_free_space(struct extent_buffer *leaf) 3513 { 3514 struct btrfs_fs_info *fs_info = leaf->fs_info; 3515 int nritems = btrfs_header_nritems(leaf); 3516 int ret; 3517 3518 ret = BTRFS_LEAF_DATA_SIZE(fs_info) - leaf_space_used(leaf, 0, nritems); 3519 if (ret < 0) { 3520 btrfs_crit(fs_info, 3521 "leaf free space ret %d, leaf data size %lu, used %d nritems %d", 3522 ret, 3523 (unsigned long) BTRFS_LEAF_DATA_SIZE(fs_info), 3524 leaf_space_used(leaf, 0, nritems), nritems); 3525 } 3526 return ret; 3527 } 3528 3529 /* 3530 * min slot controls the lowest index we're willing to push to the 3531 * right. We'll push up to and including min_slot, but no lower 3532 */ 3533 static noinline int __push_leaf_right(struct btrfs_path *path, 3534 int data_size, int empty, 3535 struct extent_buffer *right, 3536 int free_space, u32 left_nritems, 3537 u32 min_slot) 3538 { 3539 struct btrfs_fs_info *fs_info = right->fs_info; 3540 struct extent_buffer *left = path->nodes[0]; 3541 struct extent_buffer *upper = path->nodes[1]; 3542 struct btrfs_map_token token; 3543 struct btrfs_disk_key disk_key; 3544 int slot; 3545 u32 i; 3546 int push_space = 0; 3547 int push_items = 0; 3548 struct btrfs_item *item; 3549 u32 nr; 3550 u32 right_nritems; 3551 u32 data_end; 3552 u32 this_item_size; 3553 3554 if (empty) 3555 nr = 0; 3556 else 3557 nr = max_t(u32, 1, min_slot); 3558 3559 if (path->slots[0] >= left_nritems) 3560 push_space += data_size; 3561 3562 slot = path->slots[1]; 3563 i = left_nritems - 1; 3564 while (i >= nr) { 3565 item = btrfs_item_nr(i); 3566 3567 if (!empty && push_items > 0) { 3568 if (path->slots[0] > i) 3569 break; 3570 if (path->slots[0] == i) { 3571 int space = btrfs_leaf_free_space(left); 3572 3573 if (space + push_space * 2 > free_space) 3574 break; 3575 } 3576 } 3577 3578 if (path->slots[0] == i) 3579 push_space += data_size; 3580 3581 this_item_size = btrfs_item_size(left, item); 3582 if (this_item_size + sizeof(*item) + push_space > free_space) 3583 break; 3584 3585 push_items++; 3586 push_space += this_item_size + sizeof(*item); 3587 if (i == 0) 3588 break; 3589 i--; 3590 } 3591 3592 if (push_items == 0) 3593 goto out_unlock; 3594 3595 WARN_ON(!empty && push_items == left_nritems); 3596 3597 /* push left to right */ 3598 right_nritems = btrfs_header_nritems(right); 3599 3600 push_space = btrfs_item_end_nr(left, left_nritems - push_items); 3601 push_space -= leaf_data_end(left); 3602 3603 /* make room in the right data area */ 3604 data_end = leaf_data_end(right); 3605 memmove_extent_buffer(right, 3606 BTRFS_LEAF_DATA_OFFSET + data_end - push_space, 3607 BTRFS_LEAF_DATA_OFFSET + data_end, 3608 BTRFS_LEAF_DATA_SIZE(fs_info) - data_end); 3609 3610 /* copy from the left data area */ 3611 copy_extent_buffer(right, left, BTRFS_LEAF_DATA_OFFSET + 3612 BTRFS_LEAF_DATA_SIZE(fs_info) - push_space, 3613 BTRFS_LEAF_DATA_OFFSET + leaf_data_end(left), 3614 push_space); 3615 3616 memmove_extent_buffer(right, btrfs_item_nr_offset(push_items), 3617 btrfs_item_nr_offset(0), 3618 right_nritems * sizeof(struct btrfs_item)); 3619 3620 /* copy the items from left to right */ 3621 copy_extent_buffer(right, left, btrfs_item_nr_offset(0), 3622 btrfs_item_nr_offset(left_nritems - push_items), 3623 push_items * sizeof(struct btrfs_item)); 3624 3625 /* update the item pointers */ 3626 btrfs_init_map_token(&token, right); 3627 right_nritems += push_items; 3628 btrfs_set_header_nritems(right, right_nritems); 3629 push_space = BTRFS_LEAF_DATA_SIZE(fs_info); 3630 for (i = 0; i < right_nritems; i++) { 3631 item = btrfs_item_nr(i); 3632 push_space -= btrfs_token_item_size(&token, item); 3633 btrfs_set_token_item_offset(&token, item, push_space); 3634 } 3635 3636 left_nritems -= push_items; 3637 btrfs_set_header_nritems(left, left_nritems); 3638 3639 if (left_nritems) 3640 btrfs_mark_buffer_dirty(left); 3641 else 3642 btrfs_clean_tree_block(left); 3643 3644 btrfs_mark_buffer_dirty(right); 3645 3646 btrfs_item_key(right, &disk_key, 0); 3647 btrfs_set_node_key(upper, &disk_key, slot + 1); 3648 btrfs_mark_buffer_dirty(upper); 3649 3650 /* then fixup the leaf pointer in the path */ 3651 if (path->slots[0] >= left_nritems) { 3652 path->slots[0] -= left_nritems; 3653 if (btrfs_header_nritems(path->nodes[0]) == 0) 3654 btrfs_clean_tree_block(path->nodes[0]); 3655 btrfs_tree_unlock(path->nodes[0]); 3656 free_extent_buffer(path->nodes[0]); 3657 path->nodes[0] = right; 3658 path->slots[1] += 1; 3659 } else { 3660 btrfs_tree_unlock(right); 3661 free_extent_buffer(right); 3662 } 3663 return 0; 3664 3665 out_unlock: 3666 btrfs_tree_unlock(right); 3667 free_extent_buffer(right); 3668 return 1; 3669 } 3670 3671 /* 3672 * push some data in the path leaf to the right, trying to free up at 3673 * least data_size bytes. returns zero if the push worked, nonzero otherwise 3674 * 3675 * returns 1 if the push failed because the other node didn't have enough 3676 * room, 0 if everything worked out and < 0 if there were major errors. 3677 * 3678 * this will push starting from min_slot to the end of the leaf. It won't 3679 * push any slot lower than min_slot 3680 */ 3681 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root 3682 *root, struct btrfs_path *path, 3683 int min_data_size, int data_size, 3684 int empty, u32 min_slot) 3685 { 3686 struct extent_buffer *left = path->nodes[0]; 3687 struct extent_buffer *right; 3688 struct extent_buffer *upper; 3689 int slot; 3690 int free_space; 3691 u32 left_nritems; 3692 int ret; 3693 3694 if (!path->nodes[1]) 3695 return 1; 3696 3697 slot = path->slots[1]; 3698 upper = path->nodes[1]; 3699 if (slot >= btrfs_header_nritems(upper) - 1) 3700 return 1; 3701 3702 btrfs_assert_tree_locked(path->nodes[1]); 3703 3704 right = btrfs_read_node_slot(upper, slot + 1); 3705 /* 3706 * slot + 1 is not valid or we fail to read the right node, 3707 * no big deal, just return. 3708 */ 3709 if (IS_ERR(right)) 3710 return 1; 3711 3712 btrfs_tree_lock(right); 3713 btrfs_set_lock_blocking_write(right); 3714 3715 free_space = btrfs_leaf_free_space(right); 3716 if (free_space < data_size) 3717 goto out_unlock; 3718 3719 /* cow and double check */ 3720 ret = btrfs_cow_block(trans, root, right, upper, 3721 slot + 1, &right); 3722 if (ret) 3723 goto out_unlock; 3724 3725 free_space = btrfs_leaf_free_space(right); 3726 if (free_space < data_size) 3727 goto out_unlock; 3728 3729 left_nritems = btrfs_header_nritems(left); 3730 if (left_nritems == 0) 3731 goto out_unlock; 3732 3733 if (path->slots[0] == left_nritems && !empty) { 3734 /* Key greater than all keys in the leaf, right neighbor has 3735 * enough room for it and we're not emptying our leaf to delete 3736 * it, therefore use right neighbor to insert the new item and 3737 * no need to touch/dirty our left leaf. */ 3738 btrfs_tree_unlock(left); 3739 free_extent_buffer(left); 3740 path->nodes[0] = right; 3741 path->slots[0] = 0; 3742 path->slots[1]++; 3743 return 0; 3744 } 3745 3746 return __push_leaf_right(path, min_data_size, empty, 3747 right, free_space, left_nritems, min_slot); 3748 out_unlock: 3749 btrfs_tree_unlock(right); 3750 free_extent_buffer(right); 3751 return 1; 3752 } 3753 3754 /* 3755 * push some data in the path leaf to the left, trying to free up at 3756 * least data_size bytes. returns zero if the push worked, nonzero otherwise 3757 * 3758 * max_slot can put a limit on how far into the leaf we'll push items. The 3759 * item at 'max_slot' won't be touched. Use (u32)-1 to make us do all the 3760 * items 3761 */ 3762 static noinline int __push_leaf_left(struct btrfs_path *path, int data_size, 3763 int empty, struct extent_buffer *left, 3764 int free_space, u32 right_nritems, 3765 u32 max_slot) 3766 { 3767 struct btrfs_fs_info *fs_info = left->fs_info; 3768 struct btrfs_disk_key disk_key; 3769 struct extent_buffer *right = path->nodes[0]; 3770 int i; 3771 int push_space = 0; 3772 int push_items = 0; 3773 struct btrfs_item *item; 3774 u32 old_left_nritems; 3775 u32 nr; 3776 int ret = 0; 3777 u32 this_item_size; 3778 u32 old_left_item_size; 3779 struct btrfs_map_token token; 3780 3781 if (empty) 3782 nr = min(right_nritems, max_slot); 3783 else 3784 nr = min(right_nritems - 1, max_slot); 3785 3786 for (i = 0; i < nr; i++) { 3787 item = btrfs_item_nr(i); 3788 3789 if (!empty && push_items > 0) { 3790 if (path->slots[0] < i) 3791 break; 3792 if (path->slots[0] == i) { 3793 int space = btrfs_leaf_free_space(right); 3794 3795 if (space + push_space * 2 > free_space) 3796 break; 3797 } 3798 } 3799 3800 if (path->slots[0] == i) 3801 push_space += data_size; 3802 3803 this_item_size = btrfs_item_size(right, item); 3804 if (this_item_size + sizeof(*item) + push_space > free_space) 3805 break; 3806 3807 push_items++; 3808 push_space += this_item_size + sizeof(*item); 3809 } 3810 3811 if (push_items == 0) { 3812 ret = 1; 3813 goto out; 3814 } 3815 WARN_ON(!empty && push_items == btrfs_header_nritems(right)); 3816 3817 /* push data from right to left */ 3818 copy_extent_buffer(left, right, 3819 btrfs_item_nr_offset(btrfs_header_nritems(left)), 3820 btrfs_item_nr_offset(0), 3821 push_items * sizeof(struct btrfs_item)); 3822 3823 push_space = BTRFS_LEAF_DATA_SIZE(fs_info) - 3824 btrfs_item_offset_nr(right, push_items - 1); 3825 3826 copy_extent_buffer(left, right, BTRFS_LEAF_DATA_OFFSET + 3827 leaf_data_end(left) - push_space, 3828 BTRFS_LEAF_DATA_OFFSET + 3829 btrfs_item_offset_nr(right, push_items - 1), 3830 push_space); 3831 old_left_nritems = btrfs_header_nritems(left); 3832 BUG_ON(old_left_nritems <= 0); 3833 3834 btrfs_init_map_token(&token, left); 3835 old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1); 3836 for (i = old_left_nritems; i < old_left_nritems + push_items; i++) { 3837 u32 ioff; 3838 3839 item = btrfs_item_nr(i); 3840 3841 ioff = btrfs_token_item_offset(&token, item); 3842 btrfs_set_token_item_offset(&token, item, 3843 ioff - (BTRFS_LEAF_DATA_SIZE(fs_info) - old_left_item_size)); 3844 } 3845 btrfs_set_header_nritems(left, old_left_nritems + push_items); 3846 3847 /* fixup right node */ 3848 if (push_items > right_nritems) 3849 WARN(1, KERN_CRIT "push items %d nr %u\n", push_items, 3850 right_nritems); 3851 3852 if (push_items < right_nritems) { 3853 push_space = btrfs_item_offset_nr(right, push_items - 1) - 3854 leaf_data_end(right); 3855 memmove_extent_buffer(right, BTRFS_LEAF_DATA_OFFSET + 3856 BTRFS_LEAF_DATA_SIZE(fs_info) - push_space, 3857 BTRFS_LEAF_DATA_OFFSET + 3858 leaf_data_end(right), push_space); 3859 3860 memmove_extent_buffer(right, btrfs_item_nr_offset(0), 3861 btrfs_item_nr_offset(push_items), 3862 (btrfs_header_nritems(right) - push_items) * 3863 sizeof(struct btrfs_item)); 3864 } 3865 3866 btrfs_init_map_token(&token, right); 3867 right_nritems -= push_items; 3868 btrfs_set_header_nritems(right, right_nritems); 3869 push_space = BTRFS_LEAF_DATA_SIZE(fs_info); 3870 for (i = 0; i < right_nritems; i++) { 3871 item = btrfs_item_nr(i); 3872 3873 push_space = push_space - btrfs_token_item_size(&token, item); 3874 btrfs_set_token_item_offset(&token, item, push_space); 3875 } 3876 3877 btrfs_mark_buffer_dirty(left); 3878 if (right_nritems) 3879 btrfs_mark_buffer_dirty(right); 3880 else 3881 btrfs_clean_tree_block(right); 3882 3883 btrfs_item_key(right, &disk_key, 0); 3884 fixup_low_keys(path, &disk_key, 1); 3885 3886 /* then fixup the leaf pointer in the path */ 3887 if (path->slots[0] < push_items) { 3888 path->slots[0] += old_left_nritems; 3889 btrfs_tree_unlock(path->nodes[0]); 3890 free_extent_buffer(path->nodes[0]); 3891 path->nodes[0] = left; 3892 path->slots[1] -= 1; 3893 } else { 3894 btrfs_tree_unlock(left); 3895 free_extent_buffer(left); 3896 path->slots[0] -= push_items; 3897 } 3898 BUG_ON(path->slots[0] < 0); 3899 return ret; 3900 out: 3901 btrfs_tree_unlock(left); 3902 free_extent_buffer(left); 3903 return ret; 3904 } 3905 3906 /* 3907 * push some data in the path leaf to the left, trying to free up at 3908 * least data_size bytes. returns zero if the push worked, nonzero otherwise 3909 * 3910 * max_slot can put a limit on how far into the leaf we'll push items. The 3911 * item at 'max_slot' won't be touched. Use (u32)-1 to make us push all the 3912 * items 3913 */ 3914 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root 3915 *root, struct btrfs_path *path, int min_data_size, 3916 int data_size, int empty, u32 max_slot) 3917 { 3918 struct extent_buffer *right = path->nodes[0]; 3919 struct extent_buffer *left; 3920 int slot; 3921 int free_space; 3922 u32 right_nritems; 3923 int ret = 0; 3924 3925 slot = path->slots[1]; 3926 if (slot == 0) 3927 return 1; 3928 if (!path->nodes[1]) 3929 return 1; 3930 3931 right_nritems = btrfs_header_nritems(right); 3932 if (right_nritems == 0) 3933 return 1; 3934 3935 btrfs_assert_tree_locked(path->nodes[1]); 3936 3937 left = btrfs_read_node_slot(path->nodes[1], slot - 1); 3938 /* 3939 * slot - 1 is not valid or we fail to read the left node, 3940 * no big deal, just return. 3941 */ 3942 if (IS_ERR(left)) 3943 return 1; 3944 3945 btrfs_tree_lock(left); 3946 btrfs_set_lock_blocking_write(left); 3947 3948 free_space = btrfs_leaf_free_space(left); 3949 if (free_space < data_size) { 3950 ret = 1; 3951 goto out; 3952 } 3953 3954 /* cow and double check */ 3955 ret = btrfs_cow_block(trans, root, left, 3956 path->nodes[1], slot - 1, &left); 3957 if (ret) { 3958 /* we hit -ENOSPC, but it isn't fatal here */ 3959 if (ret == -ENOSPC) 3960 ret = 1; 3961 goto out; 3962 } 3963 3964 free_space = btrfs_leaf_free_space(left); 3965 if (free_space < data_size) { 3966 ret = 1; 3967 goto out; 3968 } 3969 3970 return __push_leaf_left(path, min_data_size, 3971 empty, left, free_space, right_nritems, 3972 max_slot); 3973 out: 3974 btrfs_tree_unlock(left); 3975 free_extent_buffer(left); 3976 return ret; 3977 } 3978 3979 /* 3980 * split the path's leaf in two, making sure there is at least data_size 3981 * available for the resulting leaf level of the path. 3982 */ 3983 static noinline void copy_for_split(struct btrfs_trans_handle *trans, 3984 struct btrfs_path *path, 3985 struct extent_buffer *l, 3986 struct extent_buffer *right, 3987 int slot, int mid, int nritems) 3988 { 3989 struct btrfs_fs_info *fs_info = trans->fs_info; 3990 int data_copy_size; 3991 int rt_data_off; 3992 int i; 3993 struct btrfs_disk_key disk_key; 3994 struct btrfs_map_token token; 3995 3996 nritems = nritems - mid; 3997 btrfs_set_header_nritems(right, nritems); 3998 data_copy_size = btrfs_item_end_nr(l, mid) - leaf_data_end(l); 3999 4000 copy_extent_buffer(right, l, btrfs_item_nr_offset(0), 4001 btrfs_item_nr_offset(mid), 4002 nritems * sizeof(struct btrfs_item)); 4003 4004 copy_extent_buffer(right, l, 4005 BTRFS_LEAF_DATA_OFFSET + BTRFS_LEAF_DATA_SIZE(fs_info) - 4006 data_copy_size, BTRFS_LEAF_DATA_OFFSET + 4007 leaf_data_end(l), data_copy_size); 4008 4009 rt_data_off = BTRFS_LEAF_DATA_SIZE(fs_info) - btrfs_item_end_nr(l, mid); 4010 4011 btrfs_init_map_token(&token, right); 4012 for (i = 0; i < nritems; i++) { 4013 struct btrfs_item *item = btrfs_item_nr(i); 4014 u32 ioff; 4015 4016 ioff = btrfs_token_item_offset(&token, item); 4017 btrfs_set_token_item_offset(&token, item, ioff + rt_data_off); 4018 } 4019 4020 btrfs_set_header_nritems(l, mid); 4021 btrfs_item_key(right, &disk_key, 0); 4022 insert_ptr(trans, path, &disk_key, right->start, path->slots[1] + 1, 1); 4023 4024 btrfs_mark_buffer_dirty(right); 4025 btrfs_mark_buffer_dirty(l); 4026 BUG_ON(path->slots[0] != slot); 4027 4028 if (mid <= slot) { 4029 btrfs_tree_unlock(path->nodes[0]); 4030 free_extent_buffer(path->nodes[0]); 4031 path->nodes[0] = right; 4032 path->slots[0] -= mid; 4033 path->slots[1] += 1; 4034 } else { 4035 btrfs_tree_unlock(right); 4036 free_extent_buffer(right); 4037 } 4038 4039 BUG_ON(path->slots[0] < 0); 4040 } 4041 4042 /* 4043 * double splits happen when we need to insert a big item in the middle 4044 * of a leaf. A double split can leave us with 3 mostly empty leaves: 4045 * leaf: [ slots 0 - N] [ our target ] [ N + 1 - total in leaf ] 4046 * A B C 4047 * 4048 * We avoid this by trying to push the items on either side of our target 4049 * into the adjacent leaves. If all goes well we can avoid the double split 4050 * completely. 4051 */ 4052 static noinline int push_for_double_split(struct btrfs_trans_handle *trans, 4053 struct btrfs_root *root, 4054 struct btrfs_path *path, 4055 int data_size) 4056 { 4057 int ret; 4058 int progress = 0; 4059 int slot; 4060 u32 nritems; 4061 int space_needed = data_size; 4062 4063 slot = path->slots[0]; 4064 if (slot < btrfs_header_nritems(path->nodes[0])) 4065 space_needed -= btrfs_leaf_free_space(path->nodes[0]); 4066 4067 /* 4068 * try to push all the items after our slot into the 4069 * right leaf 4070 */ 4071 ret = push_leaf_right(trans, root, path, 1, space_needed, 0, slot); 4072 if (ret < 0) 4073 return ret; 4074 4075 if (ret == 0) 4076 progress++; 4077 4078 nritems = btrfs_header_nritems(path->nodes[0]); 4079 /* 4080 * our goal is to get our slot at the start or end of a leaf. If 4081 * we've done so we're done 4082 */ 4083 if (path->slots[0] == 0 || path->slots[0] == nritems) 4084 return 0; 4085 4086 if (btrfs_leaf_free_space(path->nodes[0]) >= data_size) 4087 return 0; 4088 4089 /* try to push all the items before our slot into the next leaf */ 4090 slot = path->slots[0]; 4091 space_needed = data_size; 4092 if (slot > 0) 4093 space_needed -= btrfs_leaf_free_space(path->nodes[0]); 4094 ret = push_leaf_left(trans, root, path, 1, space_needed, 0, slot); 4095 if (ret < 0) 4096 return ret; 4097 4098 if (ret == 0) 4099 progress++; 4100 4101 if (progress) 4102 return 0; 4103 return 1; 4104 } 4105 4106 /* 4107 * split the path's leaf in two, making sure there is at least data_size 4108 * available for the resulting leaf level of the path. 4109 * 4110 * returns 0 if all went well and < 0 on failure. 4111 */ 4112 static noinline int split_leaf(struct btrfs_trans_handle *trans, 4113 struct btrfs_root *root, 4114 const struct btrfs_key *ins_key, 4115 struct btrfs_path *path, int data_size, 4116 int extend) 4117 { 4118 struct btrfs_disk_key disk_key; 4119 struct extent_buffer *l; 4120 u32 nritems; 4121 int mid; 4122 int slot; 4123 struct extent_buffer *right; 4124 struct btrfs_fs_info *fs_info = root->fs_info; 4125 int ret = 0; 4126 int wret; 4127 int split; 4128 int num_doubles = 0; 4129 int tried_avoid_double = 0; 4130 4131 l = path->nodes[0]; 4132 slot = path->slots[0]; 4133 if (extend && data_size + btrfs_item_size_nr(l, slot) + 4134 sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(fs_info)) 4135 return -EOVERFLOW; 4136 4137 /* first try to make some room by pushing left and right */ 4138 if (data_size && path->nodes[1]) { 4139 int space_needed = data_size; 4140 4141 if (slot < btrfs_header_nritems(l)) 4142 space_needed -= btrfs_leaf_free_space(l); 4143 4144 wret = push_leaf_right(trans, root, path, space_needed, 4145 space_needed, 0, 0); 4146 if (wret < 0) 4147 return wret; 4148 if (wret) { 4149 space_needed = data_size; 4150 if (slot > 0) 4151 space_needed -= btrfs_leaf_free_space(l); 4152 wret = push_leaf_left(trans, root, path, space_needed, 4153 space_needed, 0, (u32)-1); 4154 if (wret < 0) 4155 return wret; 4156 } 4157 l = path->nodes[0]; 4158 4159 /* did the pushes work? */ 4160 if (btrfs_leaf_free_space(l) >= data_size) 4161 return 0; 4162 } 4163 4164 if (!path->nodes[1]) { 4165 ret = insert_new_root(trans, root, path, 1); 4166 if (ret) 4167 return ret; 4168 } 4169 again: 4170 split = 1; 4171 l = path->nodes[0]; 4172 slot = path->slots[0]; 4173 nritems = btrfs_header_nritems(l); 4174 mid = (nritems + 1) / 2; 4175 4176 if (mid <= slot) { 4177 if (nritems == 1 || 4178 leaf_space_used(l, mid, nritems - mid) + data_size > 4179 BTRFS_LEAF_DATA_SIZE(fs_info)) { 4180 if (slot >= nritems) { 4181 split = 0; 4182 } else { 4183 mid = slot; 4184 if (mid != nritems && 4185 leaf_space_used(l, mid, nritems - mid) + 4186 data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) { 4187 if (data_size && !tried_avoid_double) 4188 goto push_for_double; 4189 split = 2; 4190 } 4191 } 4192 } 4193 } else { 4194 if (leaf_space_used(l, 0, mid) + data_size > 4195 BTRFS_LEAF_DATA_SIZE(fs_info)) { 4196 if (!extend && data_size && slot == 0) { 4197 split = 0; 4198 } else if ((extend || !data_size) && slot == 0) { 4199 mid = 1; 4200 } else { 4201 mid = slot; 4202 if (mid != nritems && 4203 leaf_space_used(l, mid, nritems - mid) + 4204 data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) { 4205 if (data_size && !tried_avoid_double) 4206 goto push_for_double; 4207 split = 2; 4208 } 4209 } 4210 } 4211 } 4212 4213 if (split == 0) 4214 btrfs_cpu_key_to_disk(&disk_key, ins_key); 4215 else 4216 btrfs_item_key(l, &disk_key, mid); 4217 4218 right = alloc_tree_block_no_bg_flush(trans, root, 0, &disk_key, 0, 4219 l->start, 0); 4220 if (IS_ERR(right)) 4221 return PTR_ERR(right); 4222 4223 root_add_used(root, fs_info->nodesize); 4224 4225 if (split == 0) { 4226 if (mid <= slot) { 4227 btrfs_set_header_nritems(right, 0); 4228 insert_ptr(trans, path, &disk_key, 4229 right->start, path->slots[1] + 1, 1); 4230 btrfs_tree_unlock(path->nodes[0]); 4231 free_extent_buffer(path->nodes[0]); 4232 path->nodes[0] = right; 4233 path->slots[0] = 0; 4234 path->slots[1] += 1; 4235 } else { 4236 btrfs_set_header_nritems(right, 0); 4237 insert_ptr(trans, path, &disk_key, 4238 right->start, path->slots[1], 1); 4239 btrfs_tree_unlock(path->nodes[0]); 4240 free_extent_buffer(path->nodes[0]); 4241 path->nodes[0] = right; 4242 path->slots[0] = 0; 4243 if (path->slots[1] == 0) 4244 fixup_low_keys(path, &disk_key, 1); 4245 } 4246 /* 4247 * We create a new leaf 'right' for the required ins_len and 4248 * we'll do btrfs_mark_buffer_dirty() on this leaf after copying 4249 * the content of ins_len to 'right'. 4250 */ 4251 return ret; 4252 } 4253 4254 copy_for_split(trans, path, l, right, slot, mid, nritems); 4255 4256 if (split == 2) { 4257 BUG_ON(num_doubles != 0); 4258 num_doubles++; 4259 goto again; 4260 } 4261 4262 return 0; 4263 4264 push_for_double: 4265 push_for_double_split(trans, root, path, data_size); 4266 tried_avoid_double = 1; 4267 if (btrfs_leaf_free_space(path->nodes[0]) >= data_size) 4268 return 0; 4269 goto again; 4270 } 4271 4272 static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans, 4273 struct btrfs_root *root, 4274 struct btrfs_path *path, int ins_len) 4275 { 4276 struct btrfs_key key; 4277 struct extent_buffer *leaf; 4278 struct btrfs_file_extent_item *fi; 4279 u64 extent_len = 0; 4280 u32 item_size; 4281 int ret; 4282 4283 leaf = path->nodes[0]; 4284 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 4285 4286 BUG_ON(key.type != BTRFS_EXTENT_DATA_KEY && 4287 key.type != BTRFS_EXTENT_CSUM_KEY); 4288 4289 if (btrfs_leaf_free_space(leaf) >= ins_len) 4290 return 0; 4291 4292 item_size = btrfs_item_size_nr(leaf, path->slots[0]); 4293 if (key.type == BTRFS_EXTENT_DATA_KEY) { 4294 fi = btrfs_item_ptr(leaf, path->slots[0], 4295 struct btrfs_file_extent_item); 4296 extent_len = btrfs_file_extent_num_bytes(leaf, fi); 4297 } 4298 btrfs_release_path(path); 4299 4300 path->keep_locks = 1; 4301 path->search_for_split = 1; 4302 ret = btrfs_search_slot(trans, root, &key, path, 0, 1); 4303 path->search_for_split = 0; 4304 if (ret > 0) 4305 ret = -EAGAIN; 4306 if (ret < 0) 4307 goto err; 4308 4309 ret = -EAGAIN; 4310 leaf = path->nodes[0]; 4311 /* if our item isn't there, return now */ 4312 if (item_size != btrfs_item_size_nr(leaf, path->slots[0])) 4313 goto err; 4314 4315 /* the leaf has changed, it now has room. return now */ 4316 if (btrfs_leaf_free_space(path->nodes[0]) >= ins_len) 4317 goto err; 4318 4319 if (key.type == BTRFS_EXTENT_DATA_KEY) { 4320 fi = btrfs_item_ptr(leaf, path->slots[0], 4321 struct btrfs_file_extent_item); 4322 if (extent_len != btrfs_file_extent_num_bytes(leaf, fi)) 4323 goto err; 4324 } 4325 4326 btrfs_set_path_blocking(path); 4327 ret = split_leaf(trans, root, &key, path, ins_len, 1); 4328 if (ret) 4329 goto err; 4330 4331 path->keep_locks = 0; 4332 btrfs_unlock_up_safe(path, 1); 4333 return 0; 4334 err: 4335 path->keep_locks = 0; 4336 return ret; 4337 } 4338 4339 static noinline int split_item(struct btrfs_path *path, 4340 const struct btrfs_key *new_key, 4341 unsigned long split_offset) 4342 { 4343 struct extent_buffer *leaf; 4344 struct btrfs_item *item; 4345 struct btrfs_item *new_item; 4346 int slot; 4347 char *buf; 4348 u32 nritems; 4349 u32 item_size; 4350 u32 orig_offset; 4351 struct btrfs_disk_key disk_key; 4352 4353 leaf = path->nodes[0]; 4354 BUG_ON(btrfs_leaf_free_space(leaf) < sizeof(struct btrfs_item)); 4355 4356 btrfs_set_path_blocking(path); 4357 4358 item = btrfs_item_nr(path->slots[0]); 4359 orig_offset = btrfs_item_offset(leaf, item); 4360 item_size = btrfs_item_size(leaf, item); 4361 4362 buf = kmalloc(item_size, GFP_NOFS); 4363 if (!buf) 4364 return -ENOMEM; 4365 4366 read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf, 4367 path->slots[0]), item_size); 4368 4369 slot = path->slots[0] + 1; 4370 nritems = btrfs_header_nritems(leaf); 4371 if (slot != nritems) { 4372 /* shift the items */ 4373 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1), 4374 btrfs_item_nr_offset(slot), 4375 (nritems - slot) * sizeof(struct btrfs_item)); 4376 } 4377 4378 btrfs_cpu_key_to_disk(&disk_key, new_key); 4379 btrfs_set_item_key(leaf, &disk_key, slot); 4380 4381 new_item = btrfs_item_nr(slot); 4382 4383 btrfs_set_item_offset(leaf, new_item, orig_offset); 4384 btrfs_set_item_size(leaf, new_item, item_size - split_offset); 4385 4386 btrfs_set_item_offset(leaf, item, 4387 orig_offset + item_size - split_offset); 4388 btrfs_set_item_size(leaf, item, split_offset); 4389 4390 btrfs_set_header_nritems(leaf, nritems + 1); 4391 4392 /* write the data for the start of the original item */ 4393 write_extent_buffer(leaf, buf, 4394 btrfs_item_ptr_offset(leaf, path->slots[0]), 4395 split_offset); 4396 4397 /* write the data for the new item */ 4398 write_extent_buffer(leaf, buf + split_offset, 4399 btrfs_item_ptr_offset(leaf, slot), 4400 item_size - split_offset); 4401 btrfs_mark_buffer_dirty(leaf); 4402 4403 BUG_ON(btrfs_leaf_free_space(leaf) < 0); 4404 kfree(buf); 4405 return 0; 4406 } 4407 4408 /* 4409 * This function splits a single item into two items, 4410 * giving 'new_key' to the new item and splitting the 4411 * old one at split_offset (from the start of the item). 4412 * 4413 * The path may be released by this operation. After 4414 * the split, the path is pointing to the old item. The 4415 * new item is going to be in the same node as the old one. 4416 * 4417 * Note, the item being split must be smaller enough to live alone on 4418 * a tree block with room for one extra struct btrfs_item 4419 * 4420 * This allows us to split the item in place, keeping a lock on the 4421 * leaf the entire time. 4422 */ 4423 int btrfs_split_item(struct btrfs_trans_handle *trans, 4424 struct btrfs_root *root, 4425 struct btrfs_path *path, 4426 const struct btrfs_key *new_key, 4427 unsigned long split_offset) 4428 { 4429 int ret; 4430 ret = setup_leaf_for_split(trans, root, path, 4431 sizeof(struct btrfs_item)); 4432 if (ret) 4433 return ret; 4434 4435 ret = split_item(path, new_key, split_offset); 4436 return ret; 4437 } 4438 4439 /* 4440 * This function duplicate a item, giving 'new_key' to the new item. 4441 * It guarantees both items live in the same tree leaf and the new item 4442 * is contiguous with the original item. 4443 * 4444 * This allows us to split file extent in place, keeping a lock on the 4445 * leaf the entire time. 4446 */ 4447 int btrfs_duplicate_item(struct btrfs_trans_handle *trans, 4448 struct btrfs_root *root, 4449 struct btrfs_path *path, 4450 const struct btrfs_key *new_key) 4451 { 4452 struct extent_buffer *leaf; 4453 int ret; 4454 u32 item_size; 4455 4456 leaf = path->nodes[0]; 4457 item_size = btrfs_item_size_nr(leaf, path->slots[0]); 4458 ret = setup_leaf_for_split(trans, root, path, 4459 item_size + sizeof(struct btrfs_item)); 4460 if (ret) 4461 return ret; 4462 4463 path->slots[0]++; 4464 setup_items_for_insert(root, path, new_key, &item_size, 4465 item_size, item_size + 4466 sizeof(struct btrfs_item), 1); 4467 leaf = path->nodes[0]; 4468 memcpy_extent_buffer(leaf, 4469 btrfs_item_ptr_offset(leaf, path->slots[0]), 4470 btrfs_item_ptr_offset(leaf, path->slots[0] - 1), 4471 item_size); 4472 return 0; 4473 } 4474 4475 /* 4476 * make the item pointed to by the path smaller. new_size indicates 4477 * how small to make it, and from_end tells us if we just chop bytes 4478 * off the end of the item or if we shift the item to chop bytes off 4479 * the front. 4480 */ 4481 void btrfs_truncate_item(struct btrfs_path *path, u32 new_size, int from_end) 4482 { 4483 int slot; 4484 struct extent_buffer *leaf; 4485 struct btrfs_item *item; 4486 u32 nritems; 4487 unsigned int data_end; 4488 unsigned int old_data_start; 4489 unsigned int old_size; 4490 unsigned int size_diff; 4491 int i; 4492 struct btrfs_map_token token; 4493 4494 leaf = path->nodes[0]; 4495 slot = path->slots[0]; 4496 4497 old_size = btrfs_item_size_nr(leaf, slot); 4498 if (old_size == new_size) 4499 return; 4500 4501 nritems = btrfs_header_nritems(leaf); 4502 data_end = leaf_data_end(leaf); 4503 4504 old_data_start = btrfs_item_offset_nr(leaf, slot); 4505 4506 size_diff = old_size - new_size; 4507 4508 BUG_ON(slot < 0); 4509 BUG_ON(slot >= nritems); 4510 4511 /* 4512 * item0..itemN ... dataN.offset..dataN.size .. data0.size 4513 */ 4514 /* first correct the data pointers */ 4515 btrfs_init_map_token(&token, leaf); 4516 for (i = slot; i < nritems; i++) { 4517 u32 ioff; 4518 item = btrfs_item_nr(i); 4519 4520 ioff = btrfs_token_item_offset(&token, item); 4521 btrfs_set_token_item_offset(&token, item, ioff + size_diff); 4522 } 4523 4524 /* shift the data */ 4525 if (from_end) { 4526 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4527 data_end + size_diff, BTRFS_LEAF_DATA_OFFSET + 4528 data_end, old_data_start + new_size - data_end); 4529 } else { 4530 struct btrfs_disk_key disk_key; 4531 u64 offset; 4532 4533 btrfs_item_key(leaf, &disk_key, slot); 4534 4535 if (btrfs_disk_key_type(&disk_key) == BTRFS_EXTENT_DATA_KEY) { 4536 unsigned long ptr; 4537 struct btrfs_file_extent_item *fi; 4538 4539 fi = btrfs_item_ptr(leaf, slot, 4540 struct btrfs_file_extent_item); 4541 fi = (struct btrfs_file_extent_item *)( 4542 (unsigned long)fi - size_diff); 4543 4544 if (btrfs_file_extent_type(leaf, fi) == 4545 BTRFS_FILE_EXTENT_INLINE) { 4546 ptr = btrfs_item_ptr_offset(leaf, slot); 4547 memmove_extent_buffer(leaf, ptr, 4548 (unsigned long)fi, 4549 BTRFS_FILE_EXTENT_INLINE_DATA_START); 4550 } 4551 } 4552 4553 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4554 data_end + size_diff, BTRFS_LEAF_DATA_OFFSET + 4555 data_end, old_data_start - data_end); 4556 4557 offset = btrfs_disk_key_offset(&disk_key); 4558 btrfs_set_disk_key_offset(&disk_key, offset + size_diff); 4559 btrfs_set_item_key(leaf, &disk_key, slot); 4560 if (slot == 0) 4561 fixup_low_keys(path, &disk_key, 1); 4562 } 4563 4564 item = btrfs_item_nr(slot); 4565 btrfs_set_item_size(leaf, item, new_size); 4566 btrfs_mark_buffer_dirty(leaf); 4567 4568 if (btrfs_leaf_free_space(leaf) < 0) { 4569 btrfs_print_leaf(leaf); 4570 BUG(); 4571 } 4572 } 4573 4574 /* 4575 * make the item pointed to by the path bigger, data_size is the added size. 4576 */ 4577 void btrfs_extend_item(struct btrfs_path *path, u32 data_size) 4578 { 4579 int slot; 4580 struct extent_buffer *leaf; 4581 struct btrfs_item *item; 4582 u32 nritems; 4583 unsigned int data_end; 4584 unsigned int old_data; 4585 unsigned int old_size; 4586 int i; 4587 struct btrfs_map_token token; 4588 4589 leaf = path->nodes[0]; 4590 4591 nritems = btrfs_header_nritems(leaf); 4592 data_end = leaf_data_end(leaf); 4593 4594 if (btrfs_leaf_free_space(leaf) < data_size) { 4595 btrfs_print_leaf(leaf); 4596 BUG(); 4597 } 4598 slot = path->slots[0]; 4599 old_data = btrfs_item_end_nr(leaf, slot); 4600 4601 BUG_ON(slot < 0); 4602 if (slot >= nritems) { 4603 btrfs_print_leaf(leaf); 4604 btrfs_crit(leaf->fs_info, "slot %d too large, nritems %d", 4605 slot, nritems); 4606 BUG(); 4607 } 4608 4609 /* 4610 * item0..itemN ... dataN.offset..dataN.size .. data0.size 4611 */ 4612 /* first correct the data pointers */ 4613 btrfs_init_map_token(&token, leaf); 4614 for (i = slot; i < nritems; i++) { 4615 u32 ioff; 4616 item = btrfs_item_nr(i); 4617 4618 ioff = btrfs_token_item_offset(&token, item); 4619 btrfs_set_token_item_offset(&token, item, ioff - data_size); 4620 } 4621 4622 /* shift the data */ 4623 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4624 data_end - data_size, BTRFS_LEAF_DATA_OFFSET + 4625 data_end, old_data - data_end); 4626 4627 data_end = old_data; 4628 old_size = btrfs_item_size_nr(leaf, slot); 4629 item = btrfs_item_nr(slot); 4630 btrfs_set_item_size(leaf, item, old_size + data_size); 4631 btrfs_mark_buffer_dirty(leaf); 4632 4633 if (btrfs_leaf_free_space(leaf) < 0) { 4634 btrfs_print_leaf(leaf); 4635 BUG(); 4636 } 4637 } 4638 4639 /* 4640 * this is a helper for btrfs_insert_empty_items, the main goal here is 4641 * to save stack depth by doing the bulk of the work in a function 4642 * that doesn't call btrfs_search_slot 4643 */ 4644 void setup_items_for_insert(struct btrfs_root *root, struct btrfs_path *path, 4645 const struct btrfs_key *cpu_key, u32 *data_size, 4646 u32 total_data, u32 total_size, int nr) 4647 { 4648 struct btrfs_fs_info *fs_info = root->fs_info; 4649 struct btrfs_item *item; 4650 int i; 4651 u32 nritems; 4652 unsigned int data_end; 4653 struct btrfs_disk_key disk_key; 4654 struct extent_buffer *leaf; 4655 int slot; 4656 struct btrfs_map_token token; 4657 4658 if (path->slots[0] == 0) { 4659 btrfs_cpu_key_to_disk(&disk_key, cpu_key); 4660 fixup_low_keys(path, &disk_key, 1); 4661 } 4662 btrfs_unlock_up_safe(path, 1); 4663 4664 leaf = path->nodes[0]; 4665 slot = path->slots[0]; 4666 4667 nritems = btrfs_header_nritems(leaf); 4668 data_end = leaf_data_end(leaf); 4669 4670 if (btrfs_leaf_free_space(leaf) < total_size) { 4671 btrfs_print_leaf(leaf); 4672 btrfs_crit(fs_info, "not enough freespace need %u have %d", 4673 total_size, btrfs_leaf_free_space(leaf)); 4674 BUG(); 4675 } 4676 4677 btrfs_init_map_token(&token, leaf); 4678 if (slot != nritems) { 4679 unsigned int old_data = btrfs_item_end_nr(leaf, slot); 4680 4681 if (old_data < data_end) { 4682 btrfs_print_leaf(leaf); 4683 btrfs_crit(fs_info, "slot %d old_data %d data_end %d", 4684 slot, old_data, data_end); 4685 BUG(); 4686 } 4687 /* 4688 * item0..itemN ... dataN.offset..dataN.size .. data0.size 4689 */ 4690 /* first correct the data pointers */ 4691 for (i = slot; i < nritems; i++) { 4692 u32 ioff; 4693 4694 item = btrfs_item_nr(i); 4695 ioff = btrfs_token_item_offset(&token, item); 4696 btrfs_set_token_item_offset(&token, item, 4697 ioff - total_data); 4698 } 4699 /* shift the items */ 4700 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr), 4701 btrfs_item_nr_offset(slot), 4702 (nritems - slot) * sizeof(struct btrfs_item)); 4703 4704 /* shift the data */ 4705 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4706 data_end - total_data, BTRFS_LEAF_DATA_OFFSET + 4707 data_end, old_data - data_end); 4708 data_end = old_data; 4709 } 4710 4711 /* setup the item for the new data */ 4712 for (i = 0; i < nr; i++) { 4713 btrfs_cpu_key_to_disk(&disk_key, cpu_key + i); 4714 btrfs_set_item_key(leaf, &disk_key, slot + i); 4715 item = btrfs_item_nr(slot + i); 4716 btrfs_set_token_item_offset(&token, item, data_end - data_size[i]); 4717 data_end -= data_size[i]; 4718 btrfs_set_token_item_size(&token, item, data_size[i]); 4719 } 4720 4721 btrfs_set_header_nritems(leaf, nritems + nr); 4722 btrfs_mark_buffer_dirty(leaf); 4723 4724 if (btrfs_leaf_free_space(leaf) < 0) { 4725 btrfs_print_leaf(leaf); 4726 BUG(); 4727 } 4728 } 4729 4730 /* 4731 * Given a key and some data, insert items into the tree. 4732 * This does all the path init required, making room in the tree if needed. 4733 */ 4734 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, 4735 struct btrfs_root *root, 4736 struct btrfs_path *path, 4737 const struct btrfs_key *cpu_key, u32 *data_size, 4738 int nr) 4739 { 4740 int ret = 0; 4741 int slot; 4742 int i; 4743 u32 total_size = 0; 4744 u32 total_data = 0; 4745 4746 for (i = 0; i < nr; i++) 4747 total_data += data_size[i]; 4748 4749 total_size = total_data + (nr * sizeof(struct btrfs_item)); 4750 ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); 4751 if (ret == 0) 4752 return -EEXIST; 4753 if (ret < 0) 4754 return ret; 4755 4756 slot = path->slots[0]; 4757 BUG_ON(slot < 0); 4758 4759 setup_items_for_insert(root, path, cpu_key, data_size, 4760 total_data, total_size, nr); 4761 return 0; 4762 } 4763 4764 /* 4765 * Given a key and some data, insert an item into the tree. 4766 * This does all the path init required, making room in the tree if needed. 4767 */ 4768 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, 4769 const struct btrfs_key *cpu_key, void *data, 4770 u32 data_size) 4771 { 4772 int ret = 0; 4773 struct btrfs_path *path; 4774 struct extent_buffer *leaf; 4775 unsigned long ptr; 4776 4777 path = btrfs_alloc_path(); 4778 if (!path) 4779 return -ENOMEM; 4780 ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size); 4781 if (!ret) { 4782 leaf = path->nodes[0]; 4783 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); 4784 write_extent_buffer(leaf, data, ptr, data_size); 4785 btrfs_mark_buffer_dirty(leaf); 4786 } 4787 btrfs_free_path(path); 4788 return ret; 4789 } 4790 4791 /* 4792 * delete the pointer from a given node. 4793 * 4794 * the tree should have been previously balanced so the deletion does not 4795 * empty a node. 4796 */ 4797 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path, 4798 int level, int slot) 4799 { 4800 struct extent_buffer *parent = path->nodes[level]; 4801 u32 nritems; 4802 int ret; 4803 4804 nritems = btrfs_header_nritems(parent); 4805 if (slot != nritems - 1) { 4806 if (level) { 4807 ret = tree_mod_log_insert_move(parent, slot, slot + 1, 4808 nritems - slot - 1); 4809 BUG_ON(ret < 0); 4810 } 4811 memmove_extent_buffer(parent, 4812 btrfs_node_key_ptr_offset(slot), 4813 btrfs_node_key_ptr_offset(slot + 1), 4814 sizeof(struct btrfs_key_ptr) * 4815 (nritems - slot - 1)); 4816 } else if (level) { 4817 ret = tree_mod_log_insert_key(parent, slot, MOD_LOG_KEY_REMOVE, 4818 GFP_NOFS); 4819 BUG_ON(ret < 0); 4820 } 4821 4822 nritems--; 4823 btrfs_set_header_nritems(parent, nritems); 4824 if (nritems == 0 && parent == root->node) { 4825 BUG_ON(btrfs_header_level(root->node) != 1); 4826 /* just turn the root into a leaf and break */ 4827 btrfs_set_header_level(root->node, 0); 4828 } else if (slot == 0) { 4829 struct btrfs_disk_key disk_key; 4830 4831 btrfs_node_key(parent, &disk_key, 0); 4832 fixup_low_keys(path, &disk_key, level + 1); 4833 } 4834 btrfs_mark_buffer_dirty(parent); 4835 } 4836 4837 /* 4838 * a helper function to delete the leaf pointed to by path->slots[1] and 4839 * path->nodes[1]. 4840 * 4841 * This deletes the pointer in path->nodes[1] and frees the leaf 4842 * block extent. zero is returned if it all worked out, < 0 otherwise. 4843 * 4844 * The path must have already been setup for deleting the leaf, including 4845 * all the proper balancing. path->nodes[1] must be locked. 4846 */ 4847 static noinline void btrfs_del_leaf(struct btrfs_trans_handle *trans, 4848 struct btrfs_root *root, 4849 struct btrfs_path *path, 4850 struct extent_buffer *leaf) 4851 { 4852 WARN_ON(btrfs_header_generation(leaf) != trans->transid); 4853 del_ptr(root, path, 1, path->slots[1]); 4854 4855 /* 4856 * btrfs_free_extent is expensive, we want to make sure we 4857 * aren't holding any locks when we call it 4858 */ 4859 btrfs_unlock_up_safe(path, 0); 4860 4861 root_sub_used(root, leaf->len); 4862 4863 atomic_inc(&leaf->refs); 4864 btrfs_free_tree_block(trans, root, leaf, 0, 1); 4865 free_extent_buffer_stale(leaf); 4866 } 4867 /* 4868 * delete the item at the leaf level in path. If that empties 4869 * the leaf, remove it from the tree 4870 */ 4871 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, 4872 struct btrfs_path *path, int slot, int nr) 4873 { 4874 struct btrfs_fs_info *fs_info = root->fs_info; 4875 struct extent_buffer *leaf; 4876 struct btrfs_item *item; 4877 u32 last_off; 4878 u32 dsize = 0; 4879 int ret = 0; 4880 int wret; 4881 int i; 4882 u32 nritems; 4883 4884 leaf = path->nodes[0]; 4885 last_off = btrfs_item_offset_nr(leaf, slot + nr - 1); 4886 4887 for (i = 0; i < nr; i++) 4888 dsize += btrfs_item_size_nr(leaf, slot + i); 4889 4890 nritems = btrfs_header_nritems(leaf); 4891 4892 if (slot + nr != nritems) { 4893 int data_end = leaf_data_end(leaf); 4894 struct btrfs_map_token token; 4895 4896 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4897 data_end + dsize, 4898 BTRFS_LEAF_DATA_OFFSET + data_end, 4899 last_off - data_end); 4900 4901 btrfs_init_map_token(&token, leaf); 4902 for (i = slot + nr; i < nritems; i++) { 4903 u32 ioff; 4904 4905 item = btrfs_item_nr(i); 4906 ioff = btrfs_token_item_offset(&token, item); 4907 btrfs_set_token_item_offset(&token, item, ioff + dsize); 4908 } 4909 4910 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot), 4911 btrfs_item_nr_offset(slot + nr), 4912 sizeof(struct btrfs_item) * 4913 (nritems - slot - nr)); 4914 } 4915 btrfs_set_header_nritems(leaf, nritems - nr); 4916 nritems -= nr; 4917 4918 /* delete the leaf if we've emptied it */ 4919 if (nritems == 0) { 4920 if (leaf == root->node) { 4921 btrfs_set_header_level(leaf, 0); 4922 } else { 4923 btrfs_set_path_blocking(path); 4924 btrfs_clean_tree_block(leaf); 4925 btrfs_del_leaf(trans, root, path, leaf); 4926 } 4927 } else { 4928 int used = leaf_space_used(leaf, 0, nritems); 4929 if (slot == 0) { 4930 struct btrfs_disk_key disk_key; 4931 4932 btrfs_item_key(leaf, &disk_key, 0); 4933 fixup_low_keys(path, &disk_key, 1); 4934 } 4935 4936 /* delete the leaf if it is mostly empty */ 4937 if (used < BTRFS_LEAF_DATA_SIZE(fs_info) / 3) { 4938 /* push_leaf_left fixes the path. 4939 * make sure the path still points to our leaf 4940 * for possible call to del_ptr below 4941 */ 4942 slot = path->slots[1]; 4943 atomic_inc(&leaf->refs); 4944 4945 btrfs_set_path_blocking(path); 4946 wret = push_leaf_left(trans, root, path, 1, 1, 4947 1, (u32)-1); 4948 if (wret < 0 && wret != -ENOSPC) 4949 ret = wret; 4950 4951 if (path->nodes[0] == leaf && 4952 btrfs_header_nritems(leaf)) { 4953 wret = push_leaf_right(trans, root, path, 1, 4954 1, 1, 0); 4955 if (wret < 0 && wret != -ENOSPC) 4956 ret = wret; 4957 } 4958 4959 if (btrfs_header_nritems(leaf) == 0) { 4960 path->slots[1] = slot; 4961 btrfs_del_leaf(trans, root, path, leaf); 4962 free_extent_buffer(leaf); 4963 ret = 0; 4964 } else { 4965 /* if we're still in the path, make sure 4966 * we're dirty. Otherwise, one of the 4967 * push_leaf functions must have already 4968 * dirtied this buffer 4969 */ 4970 if (path->nodes[0] == leaf) 4971 btrfs_mark_buffer_dirty(leaf); 4972 free_extent_buffer(leaf); 4973 } 4974 } else { 4975 btrfs_mark_buffer_dirty(leaf); 4976 } 4977 } 4978 return ret; 4979 } 4980 4981 /* 4982 * search the tree again to find a leaf with lesser keys 4983 * returns 0 if it found something or 1 if there are no lesser leaves. 4984 * returns < 0 on io errors. 4985 * 4986 * This may release the path, and so you may lose any locks held at the 4987 * time you call it. 4988 */ 4989 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) 4990 { 4991 struct btrfs_key key; 4992 struct btrfs_disk_key found_key; 4993 int ret; 4994 4995 btrfs_item_key_to_cpu(path->nodes[0], &key, 0); 4996 4997 if (key.offset > 0) { 4998 key.offset--; 4999 } else if (key.type > 0) { 5000 key.type--; 5001 key.offset = (u64)-1; 5002 } else if (key.objectid > 0) { 5003 key.objectid--; 5004 key.type = (u8)-1; 5005 key.offset = (u64)-1; 5006 } else { 5007 return 1; 5008 } 5009 5010 btrfs_release_path(path); 5011 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5012 if (ret < 0) 5013 return ret; 5014 btrfs_item_key(path->nodes[0], &found_key, 0); 5015 ret = comp_keys(&found_key, &key); 5016 /* 5017 * We might have had an item with the previous key in the tree right 5018 * before we released our path. And after we released our path, that 5019 * item might have been pushed to the first slot (0) of the leaf we 5020 * were holding due to a tree balance. Alternatively, an item with the 5021 * previous key can exist as the only element of a leaf (big fat item). 5022 * Therefore account for these 2 cases, so that our callers (like 5023 * btrfs_previous_item) don't miss an existing item with a key matching 5024 * the previous key we computed above. 5025 */ 5026 if (ret <= 0) 5027 return 0; 5028 return 1; 5029 } 5030 5031 /* 5032 * A helper function to walk down the tree starting at min_key, and looking 5033 * for nodes or leaves that are have a minimum transaction id. 5034 * This is used by the btree defrag code, and tree logging 5035 * 5036 * This does not cow, but it does stuff the starting key it finds back 5037 * into min_key, so you can call btrfs_search_slot with cow=1 on the 5038 * key and get a writable path. 5039 * 5040 * This honors path->lowest_level to prevent descent past a given level 5041 * of the tree. 5042 * 5043 * min_trans indicates the oldest transaction that you are interested 5044 * in walking through. Any nodes or leaves older than min_trans are 5045 * skipped over (without reading them). 5046 * 5047 * returns zero if something useful was found, < 0 on error and 1 if there 5048 * was nothing in the tree that matched the search criteria. 5049 */ 5050 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, 5051 struct btrfs_path *path, 5052 u64 min_trans) 5053 { 5054 struct extent_buffer *cur; 5055 struct btrfs_key found_key; 5056 int slot; 5057 int sret; 5058 u32 nritems; 5059 int level; 5060 int ret = 1; 5061 int keep_locks = path->keep_locks; 5062 5063 path->keep_locks = 1; 5064 again: 5065 cur = btrfs_read_lock_root_node(root); 5066 level = btrfs_header_level(cur); 5067 WARN_ON(path->nodes[level]); 5068 path->nodes[level] = cur; 5069 path->locks[level] = BTRFS_READ_LOCK; 5070 5071 if (btrfs_header_generation(cur) < min_trans) { 5072 ret = 1; 5073 goto out; 5074 } 5075 while (1) { 5076 nritems = btrfs_header_nritems(cur); 5077 level = btrfs_header_level(cur); 5078 sret = btrfs_bin_search(cur, min_key, &slot); 5079 if (sret < 0) { 5080 ret = sret; 5081 goto out; 5082 } 5083 5084 /* at the lowest level, we're done, setup the path and exit */ 5085 if (level == path->lowest_level) { 5086 if (slot >= nritems) 5087 goto find_next_key; 5088 ret = 0; 5089 path->slots[level] = slot; 5090 btrfs_item_key_to_cpu(cur, &found_key, slot); 5091 goto out; 5092 } 5093 if (sret && slot > 0) 5094 slot--; 5095 /* 5096 * check this node pointer against the min_trans parameters. 5097 * If it is too old, old, skip to the next one. 5098 */ 5099 while (slot < nritems) { 5100 u64 gen; 5101 5102 gen = btrfs_node_ptr_generation(cur, slot); 5103 if (gen < min_trans) { 5104 slot++; 5105 continue; 5106 } 5107 break; 5108 } 5109 find_next_key: 5110 /* 5111 * we didn't find a candidate key in this node, walk forward 5112 * and find another one 5113 */ 5114 if (slot >= nritems) { 5115 path->slots[level] = slot; 5116 btrfs_set_path_blocking(path); 5117 sret = btrfs_find_next_key(root, path, min_key, level, 5118 min_trans); 5119 if (sret == 0) { 5120 btrfs_release_path(path); 5121 goto again; 5122 } else { 5123 goto out; 5124 } 5125 } 5126 /* save our key for returning back */ 5127 btrfs_node_key_to_cpu(cur, &found_key, slot); 5128 path->slots[level] = slot; 5129 if (level == path->lowest_level) { 5130 ret = 0; 5131 goto out; 5132 } 5133 btrfs_set_path_blocking(path); 5134 cur = btrfs_read_node_slot(cur, slot); 5135 if (IS_ERR(cur)) { 5136 ret = PTR_ERR(cur); 5137 goto out; 5138 } 5139 5140 btrfs_tree_read_lock(cur); 5141 5142 path->locks[level - 1] = BTRFS_READ_LOCK; 5143 path->nodes[level - 1] = cur; 5144 unlock_up(path, level, 1, 0, NULL); 5145 } 5146 out: 5147 path->keep_locks = keep_locks; 5148 if (ret == 0) { 5149 btrfs_unlock_up_safe(path, path->lowest_level + 1); 5150 btrfs_set_path_blocking(path); 5151 memcpy(min_key, &found_key, sizeof(found_key)); 5152 } 5153 return ret; 5154 } 5155 5156 /* 5157 * this is similar to btrfs_next_leaf, but does not try to preserve 5158 * and fixup the path. It looks for and returns the next key in the 5159 * tree based on the current path and the min_trans parameters. 5160 * 5161 * 0 is returned if another key is found, < 0 if there are any errors 5162 * and 1 is returned if there are no higher keys in the tree 5163 * 5164 * path->keep_locks should be set to 1 on the search made before 5165 * calling this function. 5166 */ 5167 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, 5168 struct btrfs_key *key, int level, u64 min_trans) 5169 { 5170 int slot; 5171 struct extent_buffer *c; 5172 5173 WARN_ON(!path->keep_locks && !path->skip_locking); 5174 while (level < BTRFS_MAX_LEVEL) { 5175 if (!path->nodes[level]) 5176 return 1; 5177 5178 slot = path->slots[level] + 1; 5179 c = path->nodes[level]; 5180 next: 5181 if (slot >= btrfs_header_nritems(c)) { 5182 int ret; 5183 int orig_lowest; 5184 struct btrfs_key cur_key; 5185 if (level + 1 >= BTRFS_MAX_LEVEL || 5186 !path->nodes[level + 1]) 5187 return 1; 5188 5189 if (path->locks[level + 1] || path->skip_locking) { 5190 level++; 5191 continue; 5192 } 5193 5194 slot = btrfs_header_nritems(c) - 1; 5195 if (level == 0) 5196 btrfs_item_key_to_cpu(c, &cur_key, slot); 5197 else 5198 btrfs_node_key_to_cpu(c, &cur_key, slot); 5199 5200 orig_lowest = path->lowest_level; 5201 btrfs_release_path(path); 5202 path->lowest_level = level; 5203 ret = btrfs_search_slot(NULL, root, &cur_key, path, 5204 0, 0); 5205 path->lowest_level = orig_lowest; 5206 if (ret < 0) 5207 return ret; 5208 5209 c = path->nodes[level]; 5210 slot = path->slots[level]; 5211 if (ret == 0) 5212 slot++; 5213 goto next; 5214 } 5215 5216 if (level == 0) 5217 btrfs_item_key_to_cpu(c, key, slot); 5218 else { 5219 u64 gen = btrfs_node_ptr_generation(c, slot); 5220 5221 if (gen < min_trans) { 5222 slot++; 5223 goto next; 5224 } 5225 btrfs_node_key_to_cpu(c, key, slot); 5226 } 5227 return 0; 5228 } 5229 return 1; 5230 } 5231 5232 /* 5233 * search the tree again to find a leaf with greater keys 5234 * returns 0 if it found something or 1 if there are no greater leaves. 5235 * returns < 0 on io errors. 5236 */ 5237 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) 5238 { 5239 return btrfs_next_old_leaf(root, path, 0); 5240 } 5241 5242 int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path, 5243 u64 time_seq) 5244 { 5245 int slot; 5246 int level; 5247 struct extent_buffer *c; 5248 struct extent_buffer *next; 5249 struct btrfs_key key; 5250 u32 nritems; 5251 int ret; 5252 int old_spinning = path->leave_spinning; 5253 int next_rw_lock = 0; 5254 5255 nritems = btrfs_header_nritems(path->nodes[0]); 5256 if (nritems == 0) 5257 return 1; 5258 5259 btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1); 5260 again: 5261 level = 1; 5262 next = NULL; 5263 next_rw_lock = 0; 5264 btrfs_release_path(path); 5265 5266 path->keep_locks = 1; 5267 path->leave_spinning = 1; 5268 5269 if (time_seq) 5270 ret = btrfs_search_old_slot(root, &key, path, time_seq); 5271 else 5272 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5273 path->keep_locks = 0; 5274 5275 if (ret < 0) 5276 return ret; 5277 5278 nritems = btrfs_header_nritems(path->nodes[0]); 5279 /* 5280 * by releasing the path above we dropped all our locks. A balance 5281 * could have added more items next to the key that used to be 5282 * at the very end of the block. So, check again here and 5283 * advance the path if there are now more items available. 5284 */ 5285 if (nritems > 0 && path->slots[0] < nritems - 1) { 5286 if (ret == 0) 5287 path->slots[0]++; 5288 ret = 0; 5289 goto done; 5290 } 5291 /* 5292 * So the above check misses one case: 5293 * - after releasing the path above, someone has removed the item that 5294 * used to be at the very end of the block, and balance between leafs 5295 * gets another one with bigger key.offset to replace it. 5296 * 5297 * This one should be returned as well, or we can get leaf corruption 5298 * later(esp. in __btrfs_drop_extents()). 5299 * 5300 * And a bit more explanation about this check, 5301 * with ret > 0, the key isn't found, the path points to the slot 5302 * where it should be inserted, so the path->slots[0] item must be the 5303 * bigger one. 5304 */ 5305 if (nritems > 0 && ret > 0 && path->slots[0] == nritems - 1) { 5306 ret = 0; 5307 goto done; 5308 } 5309 5310 while (level < BTRFS_MAX_LEVEL) { 5311 if (!path->nodes[level]) { 5312 ret = 1; 5313 goto done; 5314 } 5315 5316 slot = path->slots[level] + 1; 5317 c = path->nodes[level]; 5318 if (slot >= btrfs_header_nritems(c)) { 5319 level++; 5320 if (level == BTRFS_MAX_LEVEL) { 5321 ret = 1; 5322 goto done; 5323 } 5324 continue; 5325 } 5326 5327 if (next) { 5328 btrfs_tree_unlock_rw(next, next_rw_lock); 5329 free_extent_buffer(next); 5330 } 5331 5332 next = c; 5333 next_rw_lock = path->locks[level]; 5334 ret = read_block_for_search(root, path, &next, level, 5335 slot, &key); 5336 if (ret == -EAGAIN) 5337 goto again; 5338 5339 if (ret < 0) { 5340 btrfs_release_path(path); 5341 goto done; 5342 } 5343 5344 if (!path->skip_locking) { 5345 ret = btrfs_try_tree_read_lock(next); 5346 if (!ret && time_seq) { 5347 /* 5348 * If we don't get the lock, we may be racing 5349 * with push_leaf_left, holding that lock while 5350 * itself waiting for the leaf we've currently 5351 * locked. To solve this situation, we give up 5352 * on our lock and cycle. 5353 */ 5354 free_extent_buffer(next); 5355 btrfs_release_path(path); 5356 cond_resched(); 5357 goto again; 5358 } 5359 if (!ret) { 5360 btrfs_set_path_blocking(path); 5361 btrfs_tree_read_lock(next); 5362 } 5363 next_rw_lock = BTRFS_READ_LOCK; 5364 } 5365 break; 5366 } 5367 path->slots[level] = slot; 5368 while (1) { 5369 level--; 5370 c = path->nodes[level]; 5371 if (path->locks[level]) 5372 btrfs_tree_unlock_rw(c, path->locks[level]); 5373 5374 free_extent_buffer(c); 5375 path->nodes[level] = next; 5376 path->slots[level] = 0; 5377 if (!path->skip_locking) 5378 path->locks[level] = next_rw_lock; 5379 if (!level) 5380 break; 5381 5382 ret = read_block_for_search(root, path, &next, level, 5383 0, &key); 5384 if (ret == -EAGAIN) 5385 goto again; 5386 5387 if (ret < 0) { 5388 btrfs_release_path(path); 5389 goto done; 5390 } 5391 5392 if (!path->skip_locking) { 5393 ret = btrfs_try_tree_read_lock(next); 5394 if (!ret) { 5395 btrfs_set_path_blocking(path); 5396 btrfs_tree_read_lock(next); 5397 } 5398 next_rw_lock = BTRFS_READ_LOCK; 5399 } 5400 } 5401 ret = 0; 5402 done: 5403 unlock_up(path, 0, 1, 0, NULL); 5404 path->leave_spinning = old_spinning; 5405 if (!old_spinning) 5406 btrfs_set_path_blocking(path); 5407 5408 return ret; 5409 } 5410 5411 /* 5412 * this uses btrfs_prev_leaf to walk backwards in the tree, and keeps 5413 * searching until it gets past min_objectid or finds an item of 'type' 5414 * 5415 * returns 0 if something is found, 1 if nothing was found and < 0 on error 5416 */ 5417 int btrfs_previous_item(struct btrfs_root *root, 5418 struct btrfs_path *path, u64 min_objectid, 5419 int type) 5420 { 5421 struct btrfs_key found_key; 5422 struct extent_buffer *leaf; 5423 u32 nritems; 5424 int ret; 5425 5426 while (1) { 5427 if (path->slots[0] == 0) { 5428 btrfs_set_path_blocking(path); 5429 ret = btrfs_prev_leaf(root, path); 5430 if (ret != 0) 5431 return ret; 5432 } else { 5433 path->slots[0]--; 5434 } 5435 leaf = path->nodes[0]; 5436 nritems = btrfs_header_nritems(leaf); 5437 if (nritems == 0) 5438 return 1; 5439 if (path->slots[0] == nritems) 5440 path->slots[0]--; 5441 5442 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 5443 if (found_key.objectid < min_objectid) 5444 break; 5445 if (found_key.type == type) 5446 return 0; 5447 if (found_key.objectid == min_objectid && 5448 found_key.type < type) 5449 break; 5450 } 5451 return 1; 5452 } 5453 5454 /* 5455 * search in extent tree to find a previous Metadata/Data extent item with 5456 * min objecitd. 5457 * 5458 * returns 0 if something is found, 1 if nothing was found and < 0 on error 5459 */ 5460 int btrfs_previous_extent_item(struct btrfs_root *root, 5461 struct btrfs_path *path, u64 min_objectid) 5462 { 5463 struct btrfs_key found_key; 5464 struct extent_buffer *leaf; 5465 u32 nritems; 5466 int ret; 5467 5468 while (1) { 5469 if (path->slots[0] == 0) { 5470 btrfs_set_path_blocking(path); 5471 ret = btrfs_prev_leaf(root, path); 5472 if (ret != 0) 5473 return ret; 5474 } else { 5475 path->slots[0]--; 5476 } 5477 leaf = path->nodes[0]; 5478 nritems = btrfs_header_nritems(leaf); 5479 if (nritems == 0) 5480 return 1; 5481 if (path->slots[0] == nritems) 5482 path->slots[0]--; 5483 5484 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 5485 if (found_key.objectid < min_objectid) 5486 break; 5487 if (found_key.type == BTRFS_EXTENT_ITEM_KEY || 5488 found_key.type == BTRFS_METADATA_ITEM_KEY) 5489 return 0; 5490 if (found_key.objectid == min_objectid && 5491 found_key.type < BTRFS_EXTENT_ITEM_KEY) 5492 break; 5493 } 5494 return 1; 5495 } 5496