1 /* 2 * Copyright (C) 2007,2008 Oracle. All rights reserved. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public 6 * License v2 as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public 14 * License along with this program; if not, write to the 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16 * Boston, MA 021110-1307, USA. 17 */ 18 19 #include <linux/sched.h> 20 #include <linux/slab.h> 21 #include <linux/rbtree.h> 22 #include <linux/mm.h> 23 #include "ctree.h" 24 #include "disk-io.h" 25 #include "transaction.h" 26 #include "print-tree.h" 27 #include "locking.h" 28 29 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root 30 *root, struct btrfs_path *path, int level); 31 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root, 32 const struct btrfs_key *ins_key, struct btrfs_path *path, 33 int data_size, int extend); 34 static int push_node_left(struct btrfs_trans_handle *trans, 35 struct btrfs_fs_info *fs_info, 36 struct extent_buffer *dst, 37 struct extent_buffer *src, int empty); 38 static int balance_node_right(struct btrfs_trans_handle *trans, 39 struct btrfs_fs_info *fs_info, 40 struct extent_buffer *dst_buf, 41 struct extent_buffer *src_buf); 42 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path, 43 int level, int slot); 44 static int tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, 45 struct extent_buffer *eb); 46 47 struct btrfs_path *btrfs_alloc_path(void) 48 { 49 return kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS); 50 } 51 52 /* 53 * set all locked nodes in the path to blocking locks. This should 54 * be done before scheduling 55 */ 56 noinline void btrfs_set_path_blocking(struct btrfs_path *p) 57 { 58 int i; 59 for (i = 0; i < BTRFS_MAX_LEVEL; i++) { 60 if (!p->nodes[i] || !p->locks[i]) 61 continue; 62 btrfs_set_lock_blocking_rw(p->nodes[i], p->locks[i]); 63 if (p->locks[i] == BTRFS_READ_LOCK) 64 p->locks[i] = BTRFS_READ_LOCK_BLOCKING; 65 else if (p->locks[i] == BTRFS_WRITE_LOCK) 66 p->locks[i] = BTRFS_WRITE_LOCK_BLOCKING; 67 } 68 } 69 70 /* 71 * reset all the locked nodes in the patch to spinning locks. 72 * 73 * held is used to keep lockdep happy, when lockdep is enabled 74 * we set held to a blocking lock before we go around and 75 * retake all the spinlocks in the path. You can safely use NULL 76 * for held 77 */ 78 noinline void btrfs_clear_path_blocking(struct btrfs_path *p, 79 struct extent_buffer *held, int held_rw) 80 { 81 int i; 82 83 if (held) { 84 btrfs_set_lock_blocking_rw(held, held_rw); 85 if (held_rw == BTRFS_WRITE_LOCK) 86 held_rw = BTRFS_WRITE_LOCK_BLOCKING; 87 else if (held_rw == BTRFS_READ_LOCK) 88 held_rw = BTRFS_READ_LOCK_BLOCKING; 89 } 90 btrfs_set_path_blocking(p); 91 92 for (i = BTRFS_MAX_LEVEL - 1; i >= 0; i--) { 93 if (p->nodes[i] && p->locks[i]) { 94 btrfs_clear_lock_blocking_rw(p->nodes[i], p->locks[i]); 95 if (p->locks[i] == BTRFS_WRITE_LOCK_BLOCKING) 96 p->locks[i] = BTRFS_WRITE_LOCK; 97 else if (p->locks[i] == BTRFS_READ_LOCK_BLOCKING) 98 p->locks[i] = BTRFS_READ_LOCK; 99 } 100 } 101 102 if (held) 103 btrfs_clear_lock_blocking_rw(held, held_rw); 104 } 105 106 /* this also releases the path */ 107 void btrfs_free_path(struct btrfs_path *p) 108 { 109 if (!p) 110 return; 111 btrfs_release_path(p); 112 kmem_cache_free(btrfs_path_cachep, p); 113 } 114 115 /* 116 * path release drops references on the extent buffers in the path 117 * and it drops any locks held by this path 118 * 119 * It is safe to call this on paths that no locks or extent buffers held. 120 */ 121 noinline void btrfs_release_path(struct btrfs_path *p) 122 { 123 int i; 124 125 for (i = 0; i < BTRFS_MAX_LEVEL; i++) { 126 p->slots[i] = 0; 127 if (!p->nodes[i]) 128 continue; 129 if (p->locks[i]) { 130 btrfs_tree_unlock_rw(p->nodes[i], p->locks[i]); 131 p->locks[i] = 0; 132 } 133 free_extent_buffer(p->nodes[i]); 134 p->nodes[i] = NULL; 135 } 136 } 137 138 /* 139 * safely gets a reference on the root node of a tree. A lock 140 * is not taken, so a concurrent writer may put a different node 141 * at the root of the tree. See btrfs_lock_root_node for the 142 * looping required. 143 * 144 * The extent buffer returned by this has a reference taken, so 145 * it won't disappear. It may stop being the root of the tree 146 * at any time because there are no locks held. 147 */ 148 struct extent_buffer *btrfs_root_node(struct btrfs_root *root) 149 { 150 struct extent_buffer *eb; 151 152 while (1) { 153 rcu_read_lock(); 154 eb = rcu_dereference(root->node); 155 156 /* 157 * RCU really hurts here, we could free up the root node because 158 * it was COWed but we may not get the new root node yet so do 159 * the inc_not_zero dance and if it doesn't work then 160 * synchronize_rcu and try again. 161 */ 162 if (atomic_inc_not_zero(&eb->refs)) { 163 rcu_read_unlock(); 164 break; 165 } 166 rcu_read_unlock(); 167 synchronize_rcu(); 168 } 169 return eb; 170 } 171 172 /* loop around taking references on and locking the root node of the 173 * tree until you end up with a lock on the root. A locked buffer 174 * is returned, with a reference held. 175 */ 176 struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root) 177 { 178 struct extent_buffer *eb; 179 180 while (1) { 181 eb = btrfs_root_node(root); 182 btrfs_tree_lock(eb); 183 if (eb == root->node) 184 break; 185 btrfs_tree_unlock(eb); 186 free_extent_buffer(eb); 187 } 188 return eb; 189 } 190 191 /* loop around taking references on and locking the root node of the 192 * tree until you end up with a lock on the root. A locked buffer 193 * is returned, with a reference held. 194 */ 195 static struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root) 196 { 197 struct extent_buffer *eb; 198 199 while (1) { 200 eb = btrfs_root_node(root); 201 btrfs_tree_read_lock(eb); 202 if (eb == root->node) 203 break; 204 btrfs_tree_read_unlock(eb); 205 free_extent_buffer(eb); 206 } 207 return eb; 208 } 209 210 /* cowonly root (everything not a reference counted cow subvolume), just get 211 * put onto a simple dirty list. transaction.c walks this to make sure they 212 * get properly updated on disk. 213 */ 214 static void add_root_to_dirty_list(struct btrfs_root *root) 215 { 216 struct btrfs_fs_info *fs_info = root->fs_info; 217 218 if (test_bit(BTRFS_ROOT_DIRTY, &root->state) || 219 !test_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state)) 220 return; 221 222 spin_lock(&fs_info->trans_lock); 223 if (!test_and_set_bit(BTRFS_ROOT_DIRTY, &root->state)) { 224 /* Want the extent tree to be the last on the list */ 225 if (root->objectid == BTRFS_EXTENT_TREE_OBJECTID) 226 list_move_tail(&root->dirty_list, 227 &fs_info->dirty_cowonly_roots); 228 else 229 list_move(&root->dirty_list, 230 &fs_info->dirty_cowonly_roots); 231 } 232 spin_unlock(&fs_info->trans_lock); 233 } 234 235 /* 236 * used by snapshot creation to make a copy of a root for a tree with 237 * a given objectid. The buffer with the new root node is returned in 238 * cow_ret, and this func returns zero on success or a negative error code. 239 */ 240 int btrfs_copy_root(struct btrfs_trans_handle *trans, 241 struct btrfs_root *root, 242 struct extent_buffer *buf, 243 struct extent_buffer **cow_ret, u64 new_root_objectid) 244 { 245 struct btrfs_fs_info *fs_info = root->fs_info; 246 struct extent_buffer *cow; 247 int ret = 0; 248 int level; 249 struct btrfs_disk_key disk_key; 250 251 WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) && 252 trans->transid != fs_info->running_transaction->transid); 253 WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) && 254 trans->transid != root->last_trans); 255 256 level = btrfs_header_level(buf); 257 if (level == 0) 258 btrfs_item_key(buf, &disk_key, 0); 259 else 260 btrfs_node_key(buf, &disk_key, 0); 261 262 cow = btrfs_alloc_tree_block(trans, root, 0, new_root_objectid, 263 &disk_key, level, buf->start, 0); 264 if (IS_ERR(cow)) 265 return PTR_ERR(cow); 266 267 copy_extent_buffer_full(cow, buf); 268 btrfs_set_header_bytenr(cow, cow->start); 269 btrfs_set_header_generation(cow, trans->transid); 270 btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV); 271 btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN | 272 BTRFS_HEADER_FLAG_RELOC); 273 if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID) 274 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC); 275 else 276 btrfs_set_header_owner(cow, new_root_objectid); 277 278 write_extent_buffer_fsid(cow, fs_info->fsid); 279 280 WARN_ON(btrfs_header_generation(buf) > trans->transid); 281 if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID) 282 ret = btrfs_inc_ref(trans, root, cow, 1); 283 else 284 ret = btrfs_inc_ref(trans, root, cow, 0); 285 286 if (ret) 287 return ret; 288 289 btrfs_mark_buffer_dirty(cow); 290 *cow_ret = cow; 291 return 0; 292 } 293 294 enum mod_log_op { 295 MOD_LOG_KEY_REPLACE, 296 MOD_LOG_KEY_ADD, 297 MOD_LOG_KEY_REMOVE, 298 MOD_LOG_KEY_REMOVE_WHILE_FREEING, 299 MOD_LOG_KEY_REMOVE_WHILE_MOVING, 300 MOD_LOG_MOVE_KEYS, 301 MOD_LOG_ROOT_REPLACE, 302 }; 303 304 struct tree_mod_move { 305 int dst_slot; 306 int nr_items; 307 }; 308 309 struct tree_mod_root { 310 u64 logical; 311 u8 level; 312 }; 313 314 struct tree_mod_elem { 315 struct rb_node node; 316 u64 logical; 317 u64 seq; 318 enum mod_log_op op; 319 320 /* this is used for MOD_LOG_KEY_* and MOD_LOG_MOVE_KEYS operations */ 321 int slot; 322 323 /* this is used for MOD_LOG_KEY* and MOD_LOG_ROOT_REPLACE */ 324 u64 generation; 325 326 /* those are used for op == MOD_LOG_KEY_{REPLACE,REMOVE} */ 327 struct btrfs_disk_key key; 328 u64 blockptr; 329 330 /* this is used for op == MOD_LOG_MOVE_KEYS */ 331 struct tree_mod_move move; 332 333 /* this is used for op == MOD_LOG_ROOT_REPLACE */ 334 struct tree_mod_root old_root; 335 }; 336 337 static inline void tree_mod_log_read_lock(struct btrfs_fs_info *fs_info) 338 { 339 read_lock(&fs_info->tree_mod_log_lock); 340 } 341 342 static inline void tree_mod_log_read_unlock(struct btrfs_fs_info *fs_info) 343 { 344 read_unlock(&fs_info->tree_mod_log_lock); 345 } 346 347 static inline void tree_mod_log_write_lock(struct btrfs_fs_info *fs_info) 348 { 349 write_lock(&fs_info->tree_mod_log_lock); 350 } 351 352 static inline void tree_mod_log_write_unlock(struct btrfs_fs_info *fs_info) 353 { 354 write_unlock(&fs_info->tree_mod_log_lock); 355 } 356 357 /* 358 * Pull a new tree mod seq number for our operation. 359 */ 360 static inline u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info) 361 { 362 return atomic64_inc_return(&fs_info->tree_mod_seq); 363 } 364 365 /* 366 * This adds a new blocker to the tree mod log's blocker list if the @elem 367 * passed does not already have a sequence number set. So when a caller expects 368 * to record tree modifications, it should ensure to set elem->seq to zero 369 * before calling btrfs_get_tree_mod_seq. 370 * Returns a fresh, unused tree log modification sequence number, even if no new 371 * blocker was added. 372 */ 373 u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info, 374 struct seq_list *elem) 375 { 376 tree_mod_log_write_lock(fs_info); 377 spin_lock(&fs_info->tree_mod_seq_lock); 378 if (!elem->seq) { 379 elem->seq = btrfs_inc_tree_mod_seq(fs_info); 380 list_add_tail(&elem->list, &fs_info->tree_mod_seq_list); 381 } 382 spin_unlock(&fs_info->tree_mod_seq_lock); 383 tree_mod_log_write_unlock(fs_info); 384 385 return elem->seq; 386 } 387 388 void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info, 389 struct seq_list *elem) 390 { 391 struct rb_root *tm_root; 392 struct rb_node *node; 393 struct rb_node *next; 394 struct seq_list *cur_elem; 395 struct tree_mod_elem *tm; 396 u64 min_seq = (u64)-1; 397 u64 seq_putting = elem->seq; 398 399 if (!seq_putting) 400 return; 401 402 spin_lock(&fs_info->tree_mod_seq_lock); 403 list_del(&elem->list); 404 elem->seq = 0; 405 406 list_for_each_entry(cur_elem, &fs_info->tree_mod_seq_list, list) { 407 if (cur_elem->seq < min_seq) { 408 if (seq_putting > cur_elem->seq) { 409 /* 410 * blocker with lower sequence number exists, we 411 * cannot remove anything from the log 412 */ 413 spin_unlock(&fs_info->tree_mod_seq_lock); 414 return; 415 } 416 min_seq = cur_elem->seq; 417 } 418 } 419 spin_unlock(&fs_info->tree_mod_seq_lock); 420 421 /* 422 * anything that's lower than the lowest existing (read: blocked) 423 * sequence number can be removed from the tree. 424 */ 425 tree_mod_log_write_lock(fs_info); 426 tm_root = &fs_info->tree_mod_log; 427 for (node = rb_first(tm_root); node; node = next) { 428 next = rb_next(node); 429 tm = rb_entry(node, struct tree_mod_elem, node); 430 if (tm->seq > min_seq) 431 continue; 432 rb_erase(node, tm_root); 433 kfree(tm); 434 } 435 tree_mod_log_write_unlock(fs_info); 436 } 437 438 /* 439 * key order of the log: 440 * node/leaf start address -> sequence 441 * 442 * The 'start address' is the logical address of the *new* root node 443 * for root replace operations, or the logical address of the affected 444 * block for all other operations. 445 * 446 * Note: must be called with write lock (tree_mod_log_write_lock). 447 */ 448 static noinline int 449 __tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm) 450 { 451 struct rb_root *tm_root; 452 struct rb_node **new; 453 struct rb_node *parent = NULL; 454 struct tree_mod_elem *cur; 455 456 tm->seq = btrfs_inc_tree_mod_seq(fs_info); 457 458 tm_root = &fs_info->tree_mod_log; 459 new = &tm_root->rb_node; 460 while (*new) { 461 cur = rb_entry(*new, struct tree_mod_elem, node); 462 parent = *new; 463 if (cur->logical < tm->logical) 464 new = &((*new)->rb_left); 465 else if (cur->logical > tm->logical) 466 new = &((*new)->rb_right); 467 else if (cur->seq < tm->seq) 468 new = &((*new)->rb_left); 469 else if (cur->seq > tm->seq) 470 new = &((*new)->rb_right); 471 else 472 return -EEXIST; 473 } 474 475 rb_link_node(&tm->node, parent, new); 476 rb_insert_color(&tm->node, tm_root); 477 return 0; 478 } 479 480 /* 481 * Determines if logging can be omitted. Returns 1 if it can. Otherwise, it 482 * returns zero with the tree_mod_log_lock acquired. The caller must hold 483 * this until all tree mod log insertions are recorded in the rb tree and then 484 * call tree_mod_log_write_unlock() to release. 485 */ 486 static inline int tree_mod_dont_log(struct btrfs_fs_info *fs_info, 487 struct extent_buffer *eb) { 488 smp_mb(); 489 if (list_empty(&(fs_info)->tree_mod_seq_list)) 490 return 1; 491 if (eb && btrfs_header_level(eb) == 0) 492 return 1; 493 494 tree_mod_log_write_lock(fs_info); 495 if (list_empty(&(fs_info)->tree_mod_seq_list)) { 496 tree_mod_log_write_unlock(fs_info); 497 return 1; 498 } 499 500 return 0; 501 } 502 503 /* Similar to tree_mod_dont_log, but doesn't acquire any locks. */ 504 static inline int tree_mod_need_log(const struct btrfs_fs_info *fs_info, 505 struct extent_buffer *eb) 506 { 507 smp_mb(); 508 if (list_empty(&(fs_info)->tree_mod_seq_list)) 509 return 0; 510 if (eb && btrfs_header_level(eb) == 0) 511 return 0; 512 513 return 1; 514 } 515 516 static struct tree_mod_elem * 517 alloc_tree_mod_elem(struct extent_buffer *eb, int slot, 518 enum mod_log_op op, gfp_t flags) 519 { 520 struct tree_mod_elem *tm; 521 522 tm = kzalloc(sizeof(*tm), flags); 523 if (!tm) 524 return NULL; 525 526 tm->logical = eb->start; 527 if (op != MOD_LOG_KEY_ADD) { 528 btrfs_node_key(eb, &tm->key, slot); 529 tm->blockptr = btrfs_node_blockptr(eb, slot); 530 } 531 tm->op = op; 532 tm->slot = slot; 533 tm->generation = btrfs_node_ptr_generation(eb, slot); 534 RB_CLEAR_NODE(&tm->node); 535 536 return tm; 537 } 538 539 static noinline int 540 tree_mod_log_insert_key(struct btrfs_fs_info *fs_info, 541 struct extent_buffer *eb, int slot, 542 enum mod_log_op op, gfp_t flags) 543 { 544 struct tree_mod_elem *tm; 545 int ret; 546 547 if (!tree_mod_need_log(fs_info, eb)) 548 return 0; 549 550 tm = alloc_tree_mod_elem(eb, slot, op, flags); 551 if (!tm) 552 return -ENOMEM; 553 554 if (tree_mod_dont_log(fs_info, eb)) { 555 kfree(tm); 556 return 0; 557 } 558 559 ret = __tree_mod_log_insert(fs_info, tm); 560 tree_mod_log_write_unlock(fs_info); 561 if (ret) 562 kfree(tm); 563 564 return ret; 565 } 566 567 static noinline int 568 tree_mod_log_insert_move(struct btrfs_fs_info *fs_info, 569 struct extent_buffer *eb, int dst_slot, int src_slot, 570 int nr_items) 571 { 572 struct tree_mod_elem *tm = NULL; 573 struct tree_mod_elem **tm_list = NULL; 574 int ret = 0; 575 int i; 576 int locked = 0; 577 578 if (!tree_mod_need_log(fs_info, eb)) 579 return 0; 580 581 tm_list = kcalloc(nr_items, sizeof(struct tree_mod_elem *), GFP_NOFS); 582 if (!tm_list) 583 return -ENOMEM; 584 585 tm = kzalloc(sizeof(*tm), GFP_NOFS); 586 if (!tm) { 587 ret = -ENOMEM; 588 goto free_tms; 589 } 590 591 tm->logical = eb->start; 592 tm->slot = src_slot; 593 tm->move.dst_slot = dst_slot; 594 tm->move.nr_items = nr_items; 595 tm->op = MOD_LOG_MOVE_KEYS; 596 597 for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) { 598 tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot, 599 MOD_LOG_KEY_REMOVE_WHILE_MOVING, GFP_NOFS); 600 if (!tm_list[i]) { 601 ret = -ENOMEM; 602 goto free_tms; 603 } 604 } 605 606 if (tree_mod_dont_log(fs_info, eb)) 607 goto free_tms; 608 locked = 1; 609 610 /* 611 * When we override something during the move, we log these removals. 612 * This can only happen when we move towards the beginning of the 613 * buffer, i.e. dst_slot < src_slot. 614 */ 615 for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) { 616 ret = __tree_mod_log_insert(fs_info, tm_list[i]); 617 if (ret) 618 goto free_tms; 619 } 620 621 ret = __tree_mod_log_insert(fs_info, tm); 622 if (ret) 623 goto free_tms; 624 tree_mod_log_write_unlock(fs_info); 625 kfree(tm_list); 626 627 return 0; 628 free_tms: 629 for (i = 0; i < nr_items; i++) { 630 if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node)) 631 rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log); 632 kfree(tm_list[i]); 633 } 634 if (locked) 635 tree_mod_log_write_unlock(fs_info); 636 kfree(tm_list); 637 kfree(tm); 638 639 return ret; 640 } 641 642 static inline int 643 __tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, 644 struct tree_mod_elem **tm_list, 645 int nritems) 646 { 647 int i, j; 648 int ret; 649 650 for (i = nritems - 1; i >= 0; i--) { 651 ret = __tree_mod_log_insert(fs_info, tm_list[i]); 652 if (ret) { 653 for (j = nritems - 1; j > i; j--) 654 rb_erase(&tm_list[j]->node, 655 &fs_info->tree_mod_log); 656 return ret; 657 } 658 } 659 660 return 0; 661 } 662 663 static noinline int 664 tree_mod_log_insert_root(struct btrfs_fs_info *fs_info, 665 struct extent_buffer *old_root, 666 struct extent_buffer *new_root, 667 int log_removal) 668 { 669 struct tree_mod_elem *tm = NULL; 670 struct tree_mod_elem **tm_list = NULL; 671 int nritems = 0; 672 int ret = 0; 673 int i; 674 675 if (!tree_mod_need_log(fs_info, NULL)) 676 return 0; 677 678 if (log_removal && btrfs_header_level(old_root) > 0) { 679 nritems = btrfs_header_nritems(old_root); 680 tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), 681 GFP_NOFS); 682 if (!tm_list) { 683 ret = -ENOMEM; 684 goto free_tms; 685 } 686 for (i = 0; i < nritems; i++) { 687 tm_list[i] = alloc_tree_mod_elem(old_root, i, 688 MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS); 689 if (!tm_list[i]) { 690 ret = -ENOMEM; 691 goto free_tms; 692 } 693 } 694 } 695 696 tm = kzalloc(sizeof(*tm), GFP_NOFS); 697 if (!tm) { 698 ret = -ENOMEM; 699 goto free_tms; 700 } 701 702 tm->logical = new_root->start; 703 tm->old_root.logical = old_root->start; 704 tm->old_root.level = btrfs_header_level(old_root); 705 tm->generation = btrfs_header_generation(old_root); 706 tm->op = MOD_LOG_ROOT_REPLACE; 707 708 if (tree_mod_dont_log(fs_info, NULL)) 709 goto free_tms; 710 711 if (tm_list) 712 ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems); 713 if (!ret) 714 ret = __tree_mod_log_insert(fs_info, tm); 715 716 tree_mod_log_write_unlock(fs_info); 717 if (ret) 718 goto free_tms; 719 kfree(tm_list); 720 721 return ret; 722 723 free_tms: 724 if (tm_list) { 725 for (i = 0; i < nritems; i++) 726 kfree(tm_list[i]); 727 kfree(tm_list); 728 } 729 kfree(tm); 730 731 return ret; 732 } 733 734 static struct tree_mod_elem * 735 __tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq, 736 int smallest) 737 { 738 struct rb_root *tm_root; 739 struct rb_node *node; 740 struct tree_mod_elem *cur = NULL; 741 struct tree_mod_elem *found = NULL; 742 743 tree_mod_log_read_lock(fs_info); 744 tm_root = &fs_info->tree_mod_log; 745 node = tm_root->rb_node; 746 while (node) { 747 cur = rb_entry(node, struct tree_mod_elem, node); 748 if (cur->logical < start) { 749 node = node->rb_left; 750 } else if (cur->logical > start) { 751 node = node->rb_right; 752 } else if (cur->seq < min_seq) { 753 node = node->rb_left; 754 } else if (!smallest) { 755 /* we want the node with the highest seq */ 756 if (found) 757 BUG_ON(found->seq > cur->seq); 758 found = cur; 759 node = node->rb_left; 760 } else if (cur->seq > min_seq) { 761 /* we want the node with the smallest seq */ 762 if (found) 763 BUG_ON(found->seq < cur->seq); 764 found = cur; 765 node = node->rb_right; 766 } else { 767 found = cur; 768 break; 769 } 770 } 771 tree_mod_log_read_unlock(fs_info); 772 773 return found; 774 } 775 776 /* 777 * this returns the element from the log with the smallest time sequence 778 * value that's in the log (the oldest log item). any element with a time 779 * sequence lower than min_seq will be ignored. 780 */ 781 static struct tree_mod_elem * 782 tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info, u64 start, 783 u64 min_seq) 784 { 785 return __tree_mod_log_search(fs_info, start, min_seq, 1); 786 } 787 788 /* 789 * this returns the element from the log with the largest time sequence 790 * value that's in the log (the most recent log item). any element with 791 * a time sequence lower than min_seq will be ignored. 792 */ 793 static struct tree_mod_elem * 794 tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq) 795 { 796 return __tree_mod_log_search(fs_info, start, min_seq, 0); 797 } 798 799 static noinline int 800 tree_mod_log_eb_copy(struct btrfs_fs_info *fs_info, struct extent_buffer *dst, 801 struct extent_buffer *src, unsigned long dst_offset, 802 unsigned long src_offset, int nr_items) 803 { 804 int ret = 0; 805 struct tree_mod_elem **tm_list = NULL; 806 struct tree_mod_elem **tm_list_add, **tm_list_rem; 807 int i; 808 int locked = 0; 809 810 if (!tree_mod_need_log(fs_info, NULL)) 811 return 0; 812 813 if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0) 814 return 0; 815 816 tm_list = kcalloc(nr_items * 2, sizeof(struct tree_mod_elem *), 817 GFP_NOFS); 818 if (!tm_list) 819 return -ENOMEM; 820 821 tm_list_add = tm_list; 822 tm_list_rem = tm_list + nr_items; 823 for (i = 0; i < nr_items; i++) { 824 tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset, 825 MOD_LOG_KEY_REMOVE, GFP_NOFS); 826 if (!tm_list_rem[i]) { 827 ret = -ENOMEM; 828 goto free_tms; 829 } 830 831 tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset, 832 MOD_LOG_KEY_ADD, GFP_NOFS); 833 if (!tm_list_add[i]) { 834 ret = -ENOMEM; 835 goto free_tms; 836 } 837 } 838 839 if (tree_mod_dont_log(fs_info, NULL)) 840 goto free_tms; 841 locked = 1; 842 843 for (i = 0; i < nr_items; i++) { 844 ret = __tree_mod_log_insert(fs_info, tm_list_rem[i]); 845 if (ret) 846 goto free_tms; 847 ret = __tree_mod_log_insert(fs_info, tm_list_add[i]); 848 if (ret) 849 goto free_tms; 850 } 851 852 tree_mod_log_write_unlock(fs_info); 853 kfree(tm_list); 854 855 return 0; 856 857 free_tms: 858 for (i = 0; i < nr_items * 2; i++) { 859 if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node)) 860 rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log); 861 kfree(tm_list[i]); 862 } 863 if (locked) 864 tree_mod_log_write_unlock(fs_info); 865 kfree(tm_list); 866 867 return ret; 868 } 869 870 static inline void 871 tree_mod_log_eb_move(struct btrfs_fs_info *fs_info, struct extent_buffer *dst, 872 int dst_offset, int src_offset, int nr_items) 873 { 874 int ret; 875 ret = tree_mod_log_insert_move(fs_info, dst, dst_offset, src_offset, 876 nr_items); 877 BUG_ON(ret < 0); 878 } 879 880 static noinline void 881 tree_mod_log_set_node_key(struct btrfs_fs_info *fs_info, 882 struct extent_buffer *eb, int slot, int atomic) 883 { 884 int ret; 885 886 ret = tree_mod_log_insert_key(fs_info, eb, slot, 887 MOD_LOG_KEY_REPLACE, 888 atomic ? GFP_ATOMIC : GFP_NOFS); 889 BUG_ON(ret < 0); 890 } 891 892 static noinline int 893 tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb) 894 { 895 struct tree_mod_elem **tm_list = NULL; 896 int nritems = 0; 897 int i; 898 int ret = 0; 899 900 if (btrfs_header_level(eb) == 0) 901 return 0; 902 903 if (!tree_mod_need_log(fs_info, NULL)) 904 return 0; 905 906 nritems = btrfs_header_nritems(eb); 907 tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), GFP_NOFS); 908 if (!tm_list) 909 return -ENOMEM; 910 911 for (i = 0; i < nritems; i++) { 912 tm_list[i] = alloc_tree_mod_elem(eb, i, 913 MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS); 914 if (!tm_list[i]) { 915 ret = -ENOMEM; 916 goto free_tms; 917 } 918 } 919 920 if (tree_mod_dont_log(fs_info, eb)) 921 goto free_tms; 922 923 ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems); 924 tree_mod_log_write_unlock(fs_info); 925 if (ret) 926 goto free_tms; 927 kfree(tm_list); 928 929 return 0; 930 931 free_tms: 932 for (i = 0; i < nritems; i++) 933 kfree(tm_list[i]); 934 kfree(tm_list); 935 936 return ret; 937 } 938 939 static noinline void 940 tree_mod_log_set_root_pointer(struct btrfs_root *root, 941 struct extent_buffer *new_root_node, 942 int log_removal) 943 { 944 int ret; 945 ret = tree_mod_log_insert_root(root->fs_info, root->node, 946 new_root_node, log_removal); 947 BUG_ON(ret < 0); 948 } 949 950 /* 951 * check if the tree block can be shared by multiple trees 952 */ 953 int btrfs_block_can_be_shared(struct btrfs_root *root, 954 struct extent_buffer *buf) 955 { 956 /* 957 * Tree blocks not in reference counted trees and tree roots 958 * are never shared. If a block was allocated after the last 959 * snapshot and the block was not allocated by tree relocation, 960 * we know the block is not shared. 961 */ 962 if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) && 963 buf != root->node && buf != root->commit_root && 964 (btrfs_header_generation(buf) <= 965 btrfs_root_last_snapshot(&root->root_item) || 966 btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC))) 967 return 1; 968 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0 969 if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) && 970 btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV) 971 return 1; 972 #endif 973 return 0; 974 } 975 976 static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans, 977 struct btrfs_root *root, 978 struct extent_buffer *buf, 979 struct extent_buffer *cow, 980 int *last_ref) 981 { 982 struct btrfs_fs_info *fs_info = root->fs_info; 983 u64 refs; 984 u64 owner; 985 u64 flags; 986 u64 new_flags = 0; 987 int ret; 988 989 /* 990 * Backrefs update rules: 991 * 992 * Always use full backrefs for extent pointers in tree block 993 * allocated by tree relocation. 994 * 995 * If a shared tree block is no longer referenced by its owner 996 * tree (btrfs_header_owner(buf) == root->root_key.objectid), 997 * use full backrefs for extent pointers in tree block. 998 * 999 * If a tree block is been relocating 1000 * (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID), 1001 * use full backrefs for extent pointers in tree block. 1002 * The reason for this is some operations (such as drop tree) 1003 * are only allowed for blocks use full backrefs. 1004 */ 1005 1006 if (btrfs_block_can_be_shared(root, buf)) { 1007 ret = btrfs_lookup_extent_info(trans, fs_info, buf->start, 1008 btrfs_header_level(buf), 1, 1009 &refs, &flags); 1010 if (ret) 1011 return ret; 1012 if (refs == 0) { 1013 ret = -EROFS; 1014 btrfs_handle_fs_error(fs_info, ret, NULL); 1015 return ret; 1016 } 1017 } else { 1018 refs = 1; 1019 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID || 1020 btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV) 1021 flags = BTRFS_BLOCK_FLAG_FULL_BACKREF; 1022 else 1023 flags = 0; 1024 } 1025 1026 owner = btrfs_header_owner(buf); 1027 BUG_ON(owner == BTRFS_TREE_RELOC_OBJECTID && 1028 !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)); 1029 1030 if (refs > 1) { 1031 if ((owner == root->root_key.objectid || 1032 root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && 1033 !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) { 1034 ret = btrfs_inc_ref(trans, root, buf, 1); 1035 BUG_ON(ret); /* -ENOMEM */ 1036 1037 if (root->root_key.objectid == 1038 BTRFS_TREE_RELOC_OBJECTID) { 1039 ret = btrfs_dec_ref(trans, root, buf, 0); 1040 BUG_ON(ret); /* -ENOMEM */ 1041 ret = btrfs_inc_ref(trans, root, cow, 1); 1042 BUG_ON(ret); /* -ENOMEM */ 1043 } 1044 new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF; 1045 } else { 1046 1047 if (root->root_key.objectid == 1048 BTRFS_TREE_RELOC_OBJECTID) 1049 ret = btrfs_inc_ref(trans, root, cow, 1); 1050 else 1051 ret = btrfs_inc_ref(trans, root, cow, 0); 1052 BUG_ON(ret); /* -ENOMEM */ 1053 } 1054 if (new_flags != 0) { 1055 int level = btrfs_header_level(buf); 1056 1057 ret = btrfs_set_disk_extent_flags(trans, fs_info, 1058 buf->start, 1059 buf->len, 1060 new_flags, level, 0); 1061 if (ret) 1062 return ret; 1063 } 1064 } else { 1065 if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) { 1066 if (root->root_key.objectid == 1067 BTRFS_TREE_RELOC_OBJECTID) 1068 ret = btrfs_inc_ref(trans, root, cow, 1); 1069 else 1070 ret = btrfs_inc_ref(trans, root, cow, 0); 1071 BUG_ON(ret); /* -ENOMEM */ 1072 ret = btrfs_dec_ref(trans, root, buf, 1); 1073 BUG_ON(ret); /* -ENOMEM */ 1074 } 1075 clean_tree_block(fs_info, buf); 1076 *last_ref = 1; 1077 } 1078 return 0; 1079 } 1080 1081 /* 1082 * does the dirty work in cow of a single block. The parent block (if 1083 * supplied) is updated to point to the new cow copy. The new buffer is marked 1084 * dirty and returned locked. If you modify the block it needs to be marked 1085 * dirty again. 1086 * 1087 * search_start -- an allocation hint for the new block 1088 * 1089 * empty_size -- a hint that you plan on doing more cow. This is the size in 1090 * bytes the allocator should try to find free next to the block it returns. 1091 * This is just a hint and may be ignored by the allocator. 1092 */ 1093 static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans, 1094 struct btrfs_root *root, 1095 struct extent_buffer *buf, 1096 struct extent_buffer *parent, int parent_slot, 1097 struct extent_buffer **cow_ret, 1098 u64 search_start, u64 empty_size) 1099 { 1100 struct btrfs_fs_info *fs_info = root->fs_info; 1101 struct btrfs_disk_key disk_key; 1102 struct extent_buffer *cow; 1103 int level, ret; 1104 int last_ref = 0; 1105 int unlock_orig = 0; 1106 u64 parent_start = 0; 1107 1108 if (*cow_ret == buf) 1109 unlock_orig = 1; 1110 1111 btrfs_assert_tree_locked(buf); 1112 1113 WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) && 1114 trans->transid != fs_info->running_transaction->transid); 1115 WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) && 1116 trans->transid != root->last_trans); 1117 1118 level = btrfs_header_level(buf); 1119 1120 if (level == 0) 1121 btrfs_item_key(buf, &disk_key, 0); 1122 else 1123 btrfs_node_key(buf, &disk_key, 0); 1124 1125 if ((root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && parent) 1126 parent_start = parent->start; 1127 1128 cow = btrfs_alloc_tree_block(trans, root, parent_start, 1129 root->root_key.objectid, &disk_key, level, 1130 search_start, empty_size); 1131 if (IS_ERR(cow)) 1132 return PTR_ERR(cow); 1133 1134 /* cow is set to blocking by btrfs_init_new_buffer */ 1135 1136 copy_extent_buffer_full(cow, buf); 1137 btrfs_set_header_bytenr(cow, cow->start); 1138 btrfs_set_header_generation(cow, trans->transid); 1139 btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV); 1140 btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN | 1141 BTRFS_HEADER_FLAG_RELOC); 1142 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) 1143 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC); 1144 else 1145 btrfs_set_header_owner(cow, root->root_key.objectid); 1146 1147 write_extent_buffer_fsid(cow, fs_info->fsid); 1148 1149 ret = update_ref_for_cow(trans, root, buf, cow, &last_ref); 1150 if (ret) { 1151 btrfs_abort_transaction(trans, ret); 1152 return ret; 1153 } 1154 1155 if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) { 1156 ret = btrfs_reloc_cow_block(trans, root, buf, cow); 1157 if (ret) { 1158 btrfs_abort_transaction(trans, ret); 1159 return ret; 1160 } 1161 } 1162 1163 if (buf == root->node) { 1164 WARN_ON(parent && parent != buf); 1165 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID || 1166 btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV) 1167 parent_start = buf->start; 1168 1169 extent_buffer_get(cow); 1170 tree_mod_log_set_root_pointer(root, cow, 1); 1171 rcu_assign_pointer(root->node, cow); 1172 1173 btrfs_free_tree_block(trans, root, buf, parent_start, 1174 last_ref); 1175 free_extent_buffer(buf); 1176 add_root_to_dirty_list(root); 1177 } else { 1178 WARN_ON(trans->transid != btrfs_header_generation(parent)); 1179 tree_mod_log_insert_key(fs_info, parent, parent_slot, 1180 MOD_LOG_KEY_REPLACE, GFP_NOFS); 1181 btrfs_set_node_blockptr(parent, parent_slot, 1182 cow->start); 1183 btrfs_set_node_ptr_generation(parent, parent_slot, 1184 trans->transid); 1185 btrfs_mark_buffer_dirty(parent); 1186 if (last_ref) { 1187 ret = tree_mod_log_free_eb(fs_info, buf); 1188 if (ret) { 1189 btrfs_abort_transaction(trans, ret); 1190 return ret; 1191 } 1192 } 1193 btrfs_free_tree_block(trans, root, buf, parent_start, 1194 last_ref); 1195 } 1196 if (unlock_orig) 1197 btrfs_tree_unlock(buf); 1198 free_extent_buffer_stale(buf); 1199 btrfs_mark_buffer_dirty(cow); 1200 *cow_ret = cow; 1201 return 0; 1202 } 1203 1204 /* 1205 * returns the logical address of the oldest predecessor of the given root. 1206 * entries older than time_seq are ignored. 1207 */ 1208 static struct tree_mod_elem * 1209 __tree_mod_log_oldest_root(struct btrfs_fs_info *fs_info, 1210 struct extent_buffer *eb_root, u64 time_seq) 1211 { 1212 struct tree_mod_elem *tm; 1213 struct tree_mod_elem *found = NULL; 1214 u64 root_logical = eb_root->start; 1215 int looped = 0; 1216 1217 if (!time_seq) 1218 return NULL; 1219 1220 /* 1221 * the very last operation that's logged for a root is the 1222 * replacement operation (if it is replaced at all). this has 1223 * the logical address of the *new* root, making it the very 1224 * first operation that's logged for this root. 1225 */ 1226 while (1) { 1227 tm = tree_mod_log_search_oldest(fs_info, root_logical, 1228 time_seq); 1229 if (!looped && !tm) 1230 return NULL; 1231 /* 1232 * if there are no tree operation for the oldest root, we simply 1233 * return it. this should only happen if that (old) root is at 1234 * level 0. 1235 */ 1236 if (!tm) 1237 break; 1238 1239 /* 1240 * if there's an operation that's not a root replacement, we 1241 * found the oldest version of our root. normally, we'll find a 1242 * MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here. 1243 */ 1244 if (tm->op != MOD_LOG_ROOT_REPLACE) 1245 break; 1246 1247 found = tm; 1248 root_logical = tm->old_root.logical; 1249 looped = 1; 1250 } 1251 1252 /* if there's no old root to return, return what we found instead */ 1253 if (!found) 1254 found = tm; 1255 1256 return found; 1257 } 1258 1259 /* 1260 * tm is a pointer to the first operation to rewind within eb. then, all 1261 * previous operations will be rewound (until we reach something older than 1262 * time_seq). 1263 */ 1264 static void 1265 __tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct extent_buffer *eb, 1266 u64 time_seq, struct tree_mod_elem *first_tm) 1267 { 1268 u32 n; 1269 struct rb_node *next; 1270 struct tree_mod_elem *tm = first_tm; 1271 unsigned long o_dst; 1272 unsigned long o_src; 1273 unsigned long p_size = sizeof(struct btrfs_key_ptr); 1274 1275 n = btrfs_header_nritems(eb); 1276 tree_mod_log_read_lock(fs_info); 1277 while (tm && tm->seq >= time_seq) { 1278 /* 1279 * all the operations are recorded with the operator used for 1280 * the modification. as we're going backwards, we do the 1281 * opposite of each operation here. 1282 */ 1283 switch (tm->op) { 1284 case MOD_LOG_KEY_REMOVE_WHILE_FREEING: 1285 BUG_ON(tm->slot < n); 1286 /* Fallthrough */ 1287 case MOD_LOG_KEY_REMOVE_WHILE_MOVING: 1288 case MOD_LOG_KEY_REMOVE: 1289 btrfs_set_node_key(eb, &tm->key, tm->slot); 1290 btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr); 1291 btrfs_set_node_ptr_generation(eb, tm->slot, 1292 tm->generation); 1293 n++; 1294 break; 1295 case MOD_LOG_KEY_REPLACE: 1296 BUG_ON(tm->slot >= n); 1297 btrfs_set_node_key(eb, &tm->key, tm->slot); 1298 btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr); 1299 btrfs_set_node_ptr_generation(eb, tm->slot, 1300 tm->generation); 1301 break; 1302 case MOD_LOG_KEY_ADD: 1303 /* if a move operation is needed it's in the log */ 1304 n--; 1305 break; 1306 case MOD_LOG_MOVE_KEYS: 1307 o_dst = btrfs_node_key_ptr_offset(tm->slot); 1308 o_src = btrfs_node_key_ptr_offset(tm->move.dst_slot); 1309 memmove_extent_buffer(eb, o_dst, o_src, 1310 tm->move.nr_items * p_size); 1311 break; 1312 case MOD_LOG_ROOT_REPLACE: 1313 /* 1314 * this operation is special. for roots, this must be 1315 * handled explicitly before rewinding. 1316 * for non-roots, this operation may exist if the node 1317 * was a root: root A -> child B; then A gets empty and 1318 * B is promoted to the new root. in the mod log, we'll 1319 * have a root-replace operation for B, a tree block 1320 * that is no root. we simply ignore that operation. 1321 */ 1322 break; 1323 } 1324 next = rb_next(&tm->node); 1325 if (!next) 1326 break; 1327 tm = rb_entry(next, struct tree_mod_elem, node); 1328 if (tm->logical != first_tm->logical) 1329 break; 1330 } 1331 tree_mod_log_read_unlock(fs_info); 1332 btrfs_set_header_nritems(eb, n); 1333 } 1334 1335 /* 1336 * Called with eb read locked. If the buffer cannot be rewound, the same buffer 1337 * is returned. If rewind operations happen, a fresh buffer is returned. The 1338 * returned buffer is always read-locked. If the returned buffer is not the 1339 * input buffer, the lock on the input buffer is released and the input buffer 1340 * is freed (its refcount is decremented). 1341 */ 1342 static struct extent_buffer * 1343 tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct btrfs_path *path, 1344 struct extent_buffer *eb, u64 time_seq) 1345 { 1346 struct extent_buffer *eb_rewin; 1347 struct tree_mod_elem *tm; 1348 1349 if (!time_seq) 1350 return eb; 1351 1352 if (btrfs_header_level(eb) == 0) 1353 return eb; 1354 1355 tm = tree_mod_log_search(fs_info, eb->start, time_seq); 1356 if (!tm) 1357 return eb; 1358 1359 btrfs_set_path_blocking(path); 1360 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 1361 1362 if (tm->op == MOD_LOG_KEY_REMOVE_WHILE_FREEING) { 1363 BUG_ON(tm->slot != 0); 1364 eb_rewin = alloc_dummy_extent_buffer(fs_info, eb->start); 1365 if (!eb_rewin) { 1366 btrfs_tree_read_unlock_blocking(eb); 1367 free_extent_buffer(eb); 1368 return NULL; 1369 } 1370 btrfs_set_header_bytenr(eb_rewin, eb->start); 1371 btrfs_set_header_backref_rev(eb_rewin, 1372 btrfs_header_backref_rev(eb)); 1373 btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb)); 1374 btrfs_set_header_level(eb_rewin, btrfs_header_level(eb)); 1375 } else { 1376 eb_rewin = btrfs_clone_extent_buffer(eb); 1377 if (!eb_rewin) { 1378 btrfs_tree_read_unlock_blocking(eb); 1379 free_extent_buffer(eb); 1380 return NULL; 1381 } 1382 } 1383 1384 btrfs_clear_path_blocking(path, NULL, BTRFS_READ_LOCK); 1385 btrfs_tree_read_unlock_blocking(eb); 1386 free_extent_buffer(eb); 1387 1388 extent_buffer_get(eb_rewin); 1389 btrfs_tree_read_lock(eb_rewin); 1390 __tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm); 1391 WARN_ON(btrfs_header_nritems(eb_rewin) > 1392 BTRFS_NODEPTRS_PER_BLOCK(fs_info)); 1393 1394 return eb_rewin; 1395 } 1396 1397 /* 1398 * get_old_root() rewinds the state of @root's root node to the given @time_seq 1399 * value. If there are no changes, the current root->root_node is returned. If 1400 * anything changed in between, there's a fresh buffer allocated on which the 1401 * rewind operations are done. In any case, the returned buffer is read locked. 1402 * Returns NULL on error (with no locks held). 1403 */ 1404 static inline struct extent_buffer * 1405 get_old_root(struct btrfs_root *root, u64 time_seq) 1406 { 1407 struct btrfs_fs_info *fs_info = root->fs_info; 1408 struct tree_mod_elem *tm; 1409 struct extent_buffer *eb = NULL; 1410 struct extent_buffer *eb_root; 1411 struct extent_buffer *old; 1412 struct tree_mod_root *old_root = NULL; 1413 u64 old_generation = 0; 1414 u64 logical; 1415 1416 eb_root = btrfs_read_lock_root_node(root); 1417 tm = __tree_mod_log_oldest_root(fs_info, eb_root, time_seq); 1418 if (!tm) 1419 return eb_root; 1420 1421 if (tm->op == MOD_LOG_ROOT_REPLACE) { 1422 old_root = &tm->old_root; 1423 old_generation = tm->generation; 1424 logical = old_root->logical; 1425 } else { 1426 logical = eb_root->start; 1427 } 1428 1429 tm = tree_mod_log_search(fs_info, logical, time_seq); 1430 if (old_root && tm && tm->op != MOD_LOG_KEY_REMOVE_WHILE_FREEING) { 1431 btrfs_tree_read_unlock(eb_root); 1432 free_extent_buffer(eb_root); 1433 old = read_tree_block(fs_info, logical, 0); 1434 if (WARN_ON(IS_ERR(old) || !extent_buffer_uptodate(old))) { 1435 if (!IS_ERR(old)) 1436 free_extent_buffer(old); 1437 btrfs_warn(fs_info, 1438 "failed to read tree block %llu from get_old_root", 1439 logical); 1440 } else { 1441 eb = btrfs_clone_extent_buffer(old); 1442 free_extent_buffer(old); 1443 } 1444 } else if (old_root) { 1445 btrfs_tree_read_unlock(eb_root); 1446 free_extent_buffer(eb_root); 1447 eb = alloc_dummy_extent_buffer(fs_info, logical); 1448 } else { 1449 btrfs_set_lock_blocking_rw(eb_root, BTRFS_READ_LOCK); 1450 eb = btrfs_clone_extent_buffer(eb_root); 1451 btrfs_tree_read_unlock_blocking(eb_root); 1452 free_extent_buffer(eb_root); 1453 } 1454 1455 if (!eb) 1456 return NULL; 1457 extent_buffer_get(eb); 1458 btrfs_tree_read_lock(eb); 1459 if (old_root) { 1460 btrfs_set_header_bytenr(eb, eb->start); 1461 btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV); 1462 btrfs_set_header_owner(eb, btrfs_header_owner(eb_root)); 1463 btrfs_set_header_level(eb, old_root->level); 1464 btrfs_set_header_generation(eb, old_generation); 1465 } 1466 if (tm) 1467 __tree_mod_log_rewind(fs_info, eb, time_seq, tm); 1468 else 1469 WARN_ON(btrfs_header_level(eb) != 0); 1470 WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(fs_info)); 1471 1472 return eb; 1473 } 1474 1475 int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq) 1476 { 1477 struct tree_mod_elem *tm; 1478 int level; 1479 struct extent_buffer *eb_root = btrfs_root_node(root); 1480 1481 tm = __tree_mod_log_oldest_root(root->fs_info, eb_root, time_seq); 1482 if (tm && tm->op == MOD_LOG_ROOT_REPLACE) { 1483 level = tm->old_root.level; 1484 } else { 1485 level = btrfs_header_level(eb_root); 1486 } 1487 free_extent_buffer(eb_root); 1488 1489 return level; 1490 } 1491 1492 static inline int should_cow_block(struct btrfs_trans_handle *trans, 1493 struct btrfs_root *root, 1494 struct extent_buffer *buf) 1495 { 1496 if (btrfs_is_testing(root->fs_info)) 1497 return 0; 1498 1499 /* ensure we can see the force_cow */ 1500 smp_rmb(); 1501 1502 /* 1503 * We do not need to cow a block if 1504 * 1) this block is not created or changed in this transaction; 1505 * 2) this block does not belong to TREE_RELOC tree; 1506 * 3) the root is not forced COW. 1507 * 1508 * What is forced COW: 1509 * when we create snapshot during committing the transaction, 1510 * after we've finished coping src root, we must COW the shared 1511 * block to ensure the metadata consistency. 1512 */ 1513 if (btrfs_header_generation(buf) == trans->transid && 1514 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) && 1515 !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 1516 btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) && 1517 !test_bit(BTRFS_ROOT_FORCE_COW, &root->state)) 1518 return 0; 1519 return 1; 1520 } 1521 1522 /* 1523 * cows a single block, see __btrfs_cow_block for the real work. 1524 * This version of it has extra checks so that a block isn't COWed more than 1525 * once per transaction, as long as it hasn't been written yet 1526 */ 1527 noinline int btrfs_cow_block(struct btrfs_trans_handle *trans, 1528 struct btrfs_root *root, struct extent_buffer *buf, 1529 struct extent_buffer *parent, int parent_slot, 1530 struct extent_buffer **cow_ret) 1531 { 1532 struct btrfs_fs_info *fs_info = root->fs_info; 1533 u64 search_start; 1534 int ret; 1535 1536 if (trans->transaction != fs_info->running_transaction) 1537 WARN(1, KERN_CRIT "trans %llu running %llu\n", 1538 trans->transid, 1539 fs_info->running_transaction->transid); 1540 1541 if (trans->transid != fs_info->generation) 1542 WARN(1, KERN_CRIT "trans %llu running %llu\n", 1543 trans->transid, fs_info->generation); 1544 1545 if (!should_cow_block(trans, root, buf)) { 1546 trans->dirty = true; 1547 *cow_ret = buf; 1548 return 0; 1549 } 1550 1551 search_start = buf->start & ~((u64)SZ_1G - 1); 1552 1553 if (parent) 1554 btrfs_set_lock_blocking(parent); 1555 btrfs_set_lock_blocking(buf); 1556 1557 ret = __btrfs_cow_block(trans, root, buf, parent, 1558 parent_slot, cow_ret, search_start, 0); 1559 1560 trace_btrfs_cow_block(root, buf, *cow_ret); 1561 1562 return ret; 1563 } 1564 1565 /* 1566 * helper function for defrag to decide if two blocks pointed to by a 1567 * node are actually close by 1568 */ 1569 static int close_blocks(u64 blocknr, u64 other, u32 blocksize) 1570 { 1571 if (blocknr < other && other - (blocknr + blocksize) < 32768) 1572 return 1; 1573 if (blocknr > other && blocknr - (other + blocksize) < 32768) 1574 return 1; 1575 return 0; 1576 } 1577 1578 /* 1579 * compare two keys in a memcmp fashion 1580 */ 1581 static int comp_keys(const struct btrfs_disk_key *disk, 1582 const struct btrfs_key *k2) 1583 { 1584 struct btrfs_key k1; 1585 1586 btrfs_disk_key_to_cpu(&k1, disk); 1587 1588 return btrfs_comp_cpu_keys(&k1, k2); 1589 } 1590 1591 /* 1592 * same as comp_keys only with two btrfs_key's 1593 */ 1594 int btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2) 1595 { 1596 if (k1->objectid > k2->objectid) 1597 return 1; 1598 if (k1->objectid < k2->objectid) 1599 return -1; 1600 if (k1->type > k2->type) 1601 return 1; 1602 if (k1->type < k2->type) 1603 return -1; 1604 if (k1->offset > k2->offset) 1605 return 1; 1606 if (k1->offset < k2->offset) 1607 return -1; 1608 return 0; 1609 } 1610 1611 /* 1612 * this is used by the defrag code to go through all the 1613 * leaves pointed to by a node and reallocate them so that 1614 * disk order is close to key order 1615 */ 1616 int btrfs_realloc_node(struct btrfs_trans_handle *trans, 1617 struct btrfs_root *root, struct extent_buffer *parent, 1618 int start_slot, u64 *last_ret, 1619 struct btrfs_key *progress) 1620 { 1621 struct btrfs_fs_info *fs_info = root->fs_info; 1622 struct extent_buffer *cur; 1623 u64 blocknr; 1624 u64 gen; 1625 u64 search_start = *last_ret; 1626 u64 last_block = 0; 1627 u64 other; 1628 u32 parent_nritems; 1629 int end_slot; 1630 int i; 1631 int err = 0; 1632 int parent_level; 1633 int uptodate; 1634 u32 blocksize; 1635 int progress_passed = 0; 1636 struct btrfs_disk_key disk_key; 1637 1638 parent_level = btrfs_header_level(parent); 1639 1640 WARN_ON(trans->transaction != fs_info->running_transaction); 1641 WARN_ON(trans->transid != fs_info->generation); 1642 1643 parent_nritems = btrfs_header_nritems(parent); 1644 blocksize = fs_info->nodesize; 1645 end_slot = parent_nritems - 1; 1646 1647 if (parent_nritems <= 1) 1648 return 0; 1649 1650 btrfs_set_lock_blocking(parent); 1651 1652 for (i = start_slot; i <= end_slot; i++) { 1653 int close = 1; 1654 1655 btrfs_node_key(parent, &disk_key, i); 1656 if (!progress_passed && comp_keys(&disk_key, progress) < 0) 1657 continue; 1658 1659 progress_passed = 1; 1660 blocknr = btrfs_node_blockptr(parent, i); 1661 gen = btrfs_node_ptr_generation(parent, i); 1662 if (last_block == 0) 1663 last_block = blocknr; 1664 1665 if (i > 0) { 1666 other = btrfs_node_blockptr(parent, i - 1); 1667 close = close_blocks(blocknr, other, blocksize); 1668 } 1669 if (!close && i < end_slot) { 1670 other = btrfs_node_blockptr(parent, i + 1); 1671 close = close_blocks(blocknr, other, blocksize); 1672 } 1673 if (close) { 1674 last_block = blocknr; 1675 continue; 1676 } 1677 1678 cur = find_extent_buffer(fs_info, blocknr); 1679 if (cur) 1680 uptodate = btrfs_buffer_uptodate(cur, gen, 0); 1681 else 1682 uptodate = 0; 1683 if (!cur || !uptodate) { 1684 if (!cur) { 1685 cur = read_tree_block(fs_info, blocknr, gen); 1686 if (IS_ERR(cur)) { 1687 return PTR_ERR(cur); 1688 } else if (!extent_buffer_uptodate(cur)) { 1689 free_extent_buffer(cur); 1690 return -EIO; 1691 } 1692 } else if (!uptodate) { 1693 err = btrfs_read_buffer(cur, gen); 1694 if (err) { 1695 free_extent_buffer(cur); 1696 return err; 1697 } 1698 } 1699 } 1700 if (search_start == 0) 1701 search_start = last_block; 1702 1703 btrfs_tree_lock(cur); 1704 btrfs_set_lock_blocking(cur); 1705 err = __btrfs_cow_block(trans, root, cur, parent, i, 1706 &cur, search_start, 1707 min(16 * blocksize, 1708 (end_slot - i) * blocksize)); 1709 if (err) { 1710 btrfs_tree_unlock(cur); 1711 free_extent_buffer(cur); 1712 break; 1713 } 1714 search_start = cur->start; 1715 last_block = cur->start; 1716 *last_ret = search_start; 1717 btrfs_tree_unlock(cur); 1718 free_extent_buffer(cur); 1719 } 1720 return err; 1721 } 1722 1723 /* 1724 * search for key in the extent_buffer. The items start at offset p, 1725 * and they are item_size apart. There are 'max' items in p. 1726 * 1727 * the slot in the array is returned via slot, and it points to 1728 * the place where you would insert key if it is not found in 1729 * the array. 1730 * 1731 * slot may point to max if the key is bigger than all of the keys 1732 */ 1733 static noinline int generic_bin_search(struct extent_buffer *eb, 1734 unsigned long p, int item_size, 1735 const struct btrfs_key *key, 1736 int max, int *slot) 1737 { 1738 int low = 0; 1739 int high = max; 1740 int mid; 1741 int ret; 1742 struct btrfs_disk_key *tmp = NULL; 1743 struct btrfs_disk_key unaligned; 1744 unsigned long offset; 1745 char *kaddr = NULL; 1746 unsigned long map_start = 0; 1747 unsigned long map_len = 0; 1748 int err; 1749 1750 if (low > high) { 1751 btrfs_err(eb->fs_info, 1752 "%s: low (%d) > high (%d) eb %llu owner %llu level %d", 1753 __func__, low, high, eb->start, 1754 btrfs_header_owner(eb), btrfs_header_level(eb)); 1755 return -EINVAL; 1756 } 1757 1758 while (low < high) { 1759 mid = (low + high) / 2; 1760 offset = p + mid * item_size; 1761 1762 if (!kaddr || offset < map_start || 1763 (offset + sizeof(struct btrfs_disk_key)) > 1764 map_start + map_len) { 1765 1766 err = map_private_extent_buffer(eb, offset, 1767 sizeof(struct btrfs_disk_key), 1768 &kaddr, &map_start, &map_len); 1769 1770 if (!err) { 1771 tmp = (struct btrfs_disk_key *)(kaddr + offset - 1772 map_start); 1773 } else if (err == 1) { 1774 read_extent_buffer(eb, &unaligned, 1775 offset, sizeof(unaligned)); 1776 tmp = &unaligned; 1777 } else { 1778 return err; 1779 } 1780 1781 } else { 1782 tmp = (struct btrfs_disk_key *)(kaddr + offset - 1783 map_start); 1784 } 1785 ret = comp_keys(tmp, key); 1786 1787 if (ret < 0) 1788 low = mid + 1; 1789 else if (ret > 0) 1790 high = mid; 1791 else { 1792 *slot = mid; 1793 return 0; 1794 } 1795 } 1796 *slot = low; 1797 return 1; 1798 } 1799 1800 /* 1801 * simple bin_search frontend that does the right thing for 1802 * leaves vs nodes 1803 */ 1804 static int bin_search(struct extent_buffer *eb, const struct btrfs_key *key, 1805 int level, int *slot) 1806 { 1807 if (level == 0) 1808 return generic_bin_search(eb, 1809 offsetof(struct btrfs_leaf, items), 1810 sizeof(struct btrfs_item), 1811 key, btrfs_header_nritems(eb), 1812 slot); 1813 else 1814 return generic_bin_search(eb, 1815 offsetof(struct btrfs_node, ptrs), 1816 sizeof(struct btrfs_key_ptr), 1817 key, btrfs_header_nritems(eb), 1818 slot); 1819 } 1820 1821 int btrfs_bin_search(struct extent_buffer *eb, const struct btrfs_key *key, 1822 int level, int *slot) 1823 { 1824 return bin_search(eb, key, level, slot); 1825 } 1826 1827 static void root_add_used(struct btrfs_root *root, u32 size) 1828 { 1829 spin_lock(&root->accounting_lock); 1830 btrfs_set_root_used(&root->root_item, 1831 btrfs_root_used(&root->root_item) + size); 1832 spin_unlock(&root->accounting_lock); 1833 } 1834 1835 static void root_sub_used(struct btrfs_root *root, u32 size) 1836 { 1837 spin_lock(&root->accounting_lock); 1838 btrfs_set_root_used(&root->root_item, 1839 btrfs_root_used(&root->root_item) - size); 1840 spin_unlock(&root->accounting_lock); 1841 } 1842 1843 /* given a node and slot number, this reads the blocks it points to. The 1844 * extent buffer is returned with a reference taken (but unlocked). 1845 */ 1846 static noinline struct extent_buffer * 1847 read_node_slot(struct btrfs_fs_info *fs_info, struct extent_buffer *parent, 1848 int slot) 1849 { 1850 int level = btrfs_header_level(parent); 1851 struct extent_buffer *eb; 1852 1853 if (slot < 0 || slot >= btrfs_header_nritems(parent)) 1854 return ERR_PTR(-ENOENT); 1855 1856 BUG_ON(level == 0); 1857 1858 eb = read_tree_block(fs_info, btrfs_node_blockptr(parent, slot), 1859 btrfs_node_ptr_generation(parent, slot)); 1860 if (!IS_ERR(eb) && !extent_buffer_uptodate(eb)) { 1861 free_extent_buffer(eb); 1862 eb = ERR_PTR(-EIO); 1863 } 1864 1865 return eb; 1866 } 1867 1868 /* 1869 * node level balancing, used to make sure nodes are in proper order for 1870 * item deletion. We balance from the top down, so we have to make sure 1871 * that a deletion won't leave an node completely empty later on. 1872 */ 1873 static noinline int balance_level(struct btrfs_trans_handle *trans, 1874 struct btrfs_root *root, 1875 struct btrfs_path *path, int level) 1876 { 1877 struct btrfs_fs_info *fs_info = root->fs_info; 1878 struct extent_buffer *right = NULL; 1879 struct extent_buffer *mid; 1880 struct extent_buffer *left = NULL; 1881 struct extent_buffer *parent = NULL; 1882 int ret = 0; 1883 int wret; 1884 int pslot; 1885 int orig_slot = path->slots[level]; 1886 u64 orig_ptr; 1887 1888 if (level == 0) 1889 return 0; 1890 1891 mid = path->nodes[level]; 1892 1893 WARN_ON(path->locks[level] != BTRFS_WRITE_LOCK && 1894 path->locks[level] != BTRFS_WRITE_LOCK_BLOCKING); 1895 WARN_ON(btrfs_header_generation(mid) != trans->transid); 1896 1897 orig_ptr = btrfs_node_blockptr(mid, orig_slot); 1898 1899 if (level < BTRFS_MAX_LEVEL - 1) { 1900 parent = path->nodes[level + 1]; 1901 pslot = path->slots[level + 1]; 1902 } 1903 1904 /* 1905 * deal with the case where there is only one pointer in the root 1906 * by promoting the node below to a root 1907 */ 1908 if (!parent) { 1909 struct extent_buffer *child; 1910 1911 if (btrfs_header_nritems(mid) != 1) 1912 return 0; 1913 1914 /* promote the child to a root */ 1915 child = read_node_slot(fs_info, mid, 0); 1916 if (IS_ERR(child)) { 1917 ret = PTR_ERR(child); 1918 btrfs_handle_fs_error(fs_info, ret, NULL); 1919 goto enospc; 1920 } 1921 1922 btrfs_tree_lock(child); 1923 btrfs_set_lock_blocking(child); 1924 ret = btrfs_cow_block(trans, root, child, mid, 0, &child); 1925 if (ret) { 1926 btrfs_tree_unlock(child); 1927 free_extent_buffer(child); 1928 goto enospc; 1929 } 1930 1931 tree_mod_log_set_root_pointer(root, child, 1); 1932 rcu_assign_pointer(root->node, child); 1933 1934 add_root_to_dirty_list(root); 1935 btrfs_tree_unlock(child); 1936 1937 path->locks[level] = 0; 1938 path->nodes[level] = NULL; 1939 clean_tree_block(fs_info, mid); 1940 btrfs_tree_unlock(mid); 1941 /* once for the path */ 1942 free_extent_buffer(mid); 1943 1944 root_sub_used(root, mid->len); 1945 btrfs_free_tree_block(trans, root, mid, 0, 1); 1946 /* once for the root ptr */ 1947 free_extent_buffer_stale(mid); 1948 return 0; 1949 } 1950 if (btrfs_header_nritems(mid) > 1951 BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 4) 1952 return 0; 1953 1954 left = read_node_slot(fs_info, parent, pslot - 1); 1955 if (IS_ERR(left)) 1956 left = NULL; 1957 1958 if (left) { 1959 btrfs_tree_lock(left); 1960 btrfs_set_lock_blocking(left); 1961 wret = btrfs_cow_block(trans, root, left, 1962 parent, pslot - 1, &left); 1963 if (wret) { 1964 ret = wret; 1965 goto enospc; 1966 } 1967 } 1968 1969 right = read_node_slot(fs_info, parent, pslot + 1); 1970 if (IS_ERR(right)) 1971 right = NULL; 1972 1973 if (right) { 1974 btrfs_tree_lock(right); 1975 btrfs_set_lock_blocking(right); 1976 wret = btrfs_cow_block(trans, root, right, 1977 parent, pslot + 1, &right); 1978 if (wret) { 1979 ret = wret; 1980 goto enospc; 1981 } 1982 } 1983 1984 /* first, try to make some room in the middle buffer */ 1985 if (left) { 1986 orig_slot += btrfs_header_nritems(left); 1987 wret = push_node_left(trans, fs_info, left, mid, 1); 1988 if (wret < 0) 1989 ret = wret; 1990 } 1991 1992 /* 1993 * then try to empty the right most buffer into the middle 1994 */ 1995 if (right) { 1996 wret = push_node_left(trans, fs_info, mid, right, 1); 1997 if (wret < 0 && wret != -ENOSPC) 1998 ret = wret; 1999 if (btrfs_header_nritems(right) == 0) { 2000 clean_tree_block(fs_info, right); 2001 btrfs_tree_unlock(right); 2002 del_ptr(root, path, level + 1, pslot + 1); 2003 root_sub_used(root, right->len); 2004 btrfs_free_tree_block(trans, root, right, 0, 1); 2005 free_extent_buffer_stale(right); 2006 right = NULL; 2007 } else { 2008 struct btrfs_disk_key right_key; 2009 btrfs_node_key(right, &right_key, 0); 2010 tree_mod_log_set_node_key(fs_info, parent, 2011 pslot + 1, 0); 2012 btrfs_set_node_key(parent, &right_key, pslot + 1); 2013 btrfs_mark_buffer_dirty(parent); 2014 } 2015 } 2016 if (btrfs_header_nritems(mid) == 1) { 2017 /* 2018 * we're not allowed to leave a node with one item in the 2019 * tree during a delete. A deletion from lower in the tree 2020 * could try to delete the only pointer in this node. 2021 * So, pull some keys from the left. 2022 * There has to be a left pointer at this point because 2023 * otherwise we would have pulled some pointers from the 2024 * right 2025 */ 2026 if (!left) { 2027 ret = -EROFS; 2028 btrfs_handle_fs_error(fs_info, ret, NULL); 2029 goto enospc; 2030 } 2031 wret = balance_node_right(trans, fs_info, mid, left); 2032 if (wret < 0) { 2033 ret = wret; 2034 goto enospc; 2035 } 2036 if (wret == 1) { 2037 wret = push_node_left(trans, fs_info, left, mid, 1); 2038 if (wret < 0) 2039 ret = wret; 2040 } 2041 BUG_ON(wret == 1); 2042 } 2043 if (btrfs_header_nritems(mid) == 0) { 2044 clean_tree_block(fs_info, mid); 2045 btrfs_tree_unlock(mid); 2046 del_ptr(root, path, level + 1, pslot); 2047 root_sub_used(root, mid->len); 2048 btrfs_free_tree_block(trans, root, mid, 0, 1); 2049 free_extent_buffer_stale(mid); 2050 mid = NULL; 2051 } else { 2052 /* update the parent key to reflect our changes */ 2053 struct btrfs_disk_key mid_key; 2054 btrfs_node_key(mid, &mid_key, 0); 2055 tree_mod_log_set_node_key(fs_info, parent, pslot, 0); 2056 btrfs_set_node_key(parent, &mid_key, pslot); 2057 btrfs_mark_buffer_dirty(parent); 2058 } 2059 2060 /* update the path */ 2061 if (left) { 2062 if (btrfs_header_nritems(left) > orig_slot) { 2063 extent_buffer_get(left); 2064 /* left was locked after cow */ 2065 path->nodes[level] = left; 2066 path->slots[level + 1] -= 1; 2067 path->slots[level] = orig_slot; 2068 if (mid) { 2069 btrfs_tree_unlock(mid); 2070 free_extent_buffer(mid); 2071 } 2072 } else { 2073 orig_slot -= btrfs_header_nritems(left); 2074 path->slots[level] = orig_slot; 2075 } 2076 } 2077 /* double check we haven't messed things up */ 2078 if (orig_ptr != 2079 btrfs_node_blockptr(path->nodes[level], path->slots[level])) 2080 BUG(); 2081 enospc: 2082 if (right) { 2083 btrfs_tree_unlock(right); 2084 free_extent_buffer(right); 2085 } 2086 if (left) { 2087 if (path->nodes[level] != left) 2088 btrfs_tree_unlock(left); 2089 free_extent_buffer(left); 2090 } 2091 return ret; 2092 } 2093 2094 /* Node balancing for insertion. Here we only split or push nodes around 2095 * when they are completely full. This is also done top down, so we 2096 * have to be pessimistic. 2097 */ 2098 static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans, 2099 struct btrfs_root *root, 2100 struct btrfs_path *path, int level) 2101 { 2102 struct btrfs_fs_info *fs_info = root->fs_info; 2103 struct extent_buffer *right = NULL; 2104 struct extent_buffer *mid; 2105 struct extent_buffer *left = NULL; 2106 struct extent_buffer *parent = NULL; 2107 int ret = 0; 2108 int wret; 2109 int pslot; 2110 int orig_slot = path->slots[level]; 2111 2112 if (level == 0) 2113 return 1; 2114 2115 mid = path->nodes[level]; 2116 WARN_ON(btrfs_header_generation(mid) != trans->transid); 2117 2118 if (level < BTRFS_MAX_LEVEL - 1) { 2119 parent = path->nodes[level + 1]; 2120 pslot = path->slots[level + 1]; 2121 } 2122 2123 if (!parent) 2124 return 1; 2125 2126 left = read_node_slot(fs_info, parent, pslot - 1); 2127 if (IS_ERR(left)) 2128 left = NULL; 2129 2130 /* first, try to make some room in the middle buffer */ 2131 if (left) { 2132 u32 left_nr; 2133 2134 btrfs_tree_lock(left); 2135 btrfs_set_lock_blocking(left); 2136 2137 left_nr = btrfs_header_nritems(left); 2138 if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) { 2139 wret = 1; 2140 } else { 2141 ret = btrfs_cow_block(trans, root, left, parent, 2142 pslot - 1, &left); 2143 if (ret) 2144 wret = 1; 2145 else { 2146 wret = push_node_left(trans, fs_info, 2147 left, mid, 0); 2148 } 2149 } 2150 if (wret < 0) 2151 ret = wret; 2152 if (wret == 0) { 2153 struct btrfs_disk_key disk_key; 2154 orig_slot += left_nr; 2155 btrfs_node_key(mid, &disk_key, 0); 2156 tree_mod_log_set_node_key(fs_info, parent, pslot, 0); 2157 btrfs_set_node_key(parent, &disk_key, pslot); 2158 btrfs_mark_buffer_dirty(parent); 2159 if (btrfs_header_nritems(left) > orig_slot) { 2160 path->nodes[level] = left; 2161 path->slots[level + 1] -= 1; 2162 path->slots[level] = orig_slot; 2163 btrfs_tree_unlock(mid); 2164 free_extent_buffer(mid); 2165 } else { 2166 orig_slot -= 2167 btrfs_header_nritems(left); 2168 path->slots[level] = orig_slot; 2169 btrfs_tree_unlock(left); 2170 free_extent_buffer(left); 2171 } 2172 return 0; 2173 } 2174 btrfs_tree_unlock(left); 2175 free_extent_buffer(left); 2176 } 2177 right = read_node_slot(fs_info, parent, pslot + 1); 2178 if (IS_ERR(right)) 2179 right = NULL; 2180 2181 /* 2182 * then try to empty the right most buffer into the middle 2183 */ 2184 if (right) { 2185 u32 right_nr; 2186 2187 btrfs_tree_lock(right); 2188 btrfs_set_lock_blocking(right); 2189 2190 right_nr = btrfs_header_nritems(right); 2191 if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) { 2192 wret = 1; 2193 } else { 2194 ret = btrfs_cow_block(trans, root, right, 2195 parent, pslot + 1, 2196 &right); 2197 if (ret) 2198 wret = 1; 2199 else { 2200 wret = balance_node_right(trans, fs_info, 2201 right, mid); 2202 } 2203 } 2204 if (wret < 0) 2205 ret = wret; 2206 if (wret == 0) { 2207 struct btrfs_disk_key disk_key; 2208 2209 btrfs_node_key(right, &disk_key, 0); 2210 tree_mod_log_set_node_key(fs_info, parent, 2211 pslot + 1, 0); 2212 btrfs_set_node_key(parent, &disk_key, pslot + 1); 2213 btrfs_mark_buffer_dirty(parent); 2214 2215 if (btrfs_header_nritems(mid) <= orig_slot) { 2216 path->nodes[level] = right; 2217 path->slots[level + 1] += 1; 2218 path->slots[level] = orig_slot - 2219 btrfs_header_nritems(mid); 2220 btrfs_tree_unlock(mid); 2221 free_extent_buffer(mid); 2222 } else { 2223 btrfs_tree_unlock(right); 2224 free_extent_buffer(right); 2225 } 2226 return 0; 2227 } 2228 btrfs_tree_unlock(right); 2229 free_extent_buffer(right); 2230 } 2231 return 1; 2232 } 2233 2234 /* 2235 * readahead one full node of leaves, finding things that are close 2236 * to the block in 'slot', and triggering ra on them. 2237 */ 2238 static void reada_for_search(struct btrfs_fs_info *fs_info, 2239 struct btrfs_path *path, 2240 int level, int slot, u64 objectid) 2241 { 2242 struct extent_buffer *node; 2243 struct btrfs_disk_key disk_key; 2244 u32 nritems; 2245 u64 search; 2246 u64 target; 2247 u64 nread = 0; 2248 struct extent_buffer *eb; 2249 u32 nr; 2250 u32 blocksize; 2251 u32 nscan = 0; 2252 2253 if (level != 1) 2254 return; 2255 2256 if (!path->nodes[level]) 2257 return; 2258 2259 node = path->nodes[level]; 2260 2261 search = btrfs_node_blockptr(node, slot); 2262 blocksize = fs_info->nodesize; 2263 eb = find_extent_buffer(fs_info, search); 2264 if (eb) { 2265 free_extent_buffer(eb); 2266 return; 2267 } 2268 2269 target = search; 2270 2271 nritems = btrfs_header_nritems(node); 2272 nr = slot; 2273 2274 while (1) { 2275 if (path->reada == READA_BACK) { 2276 if (nr == 0) 2277 break; 2278 nr--; 2279 } else if (path->reada == READA_FORWARD) { 2280 nr++; 2281 if (nr >= nritems) 2282 break; 2283 } 2284 if (path->reada == READA_BACK && objectid) { 2285 btrfs_node_key(node, &disk_key, nr); 2286 if (btrfs_disk_key_objectid(&disk_key) != objectid) 2287 break; 2288 } 2289 search = btrfs_node_blockptr(node, nr); 2290 if ((search <= target && target - search <= 65536) || 2291 (search > target && search - target <= 65536)) { 2292 readahead_tree_block(fs_info, search); 2293 nread += blocksize; 2294 } 2295 nscan++; 2296 if ((nread > 65536 || nscan > 32)) 2297 break; 2298 } 2299 } 2300 2301 static noinline void reada_for_balance(struct btrfs_fs_info *fs_info, 2302 struct btrfs_path *path, int level) 2303 { 2304 int slot; 2305 int nritems; 2306 struct extent_buffer *parent; 2307 struct extent_buffer *eb; 2308 u64 gen; 2309 u64 block1 = 0; 2310 u64 block2 = 0; 2311 2312 parent = path->nodes[level + 1]; 2313 if (!parent) 2314 return; 2315 2316 nritems = btrfs_header_nritems(parent); 2317 slot = path->slots[level + 1]; 2318 2319 if (slot > 0) { 2320 block1 = btrfs_node_blockptr(parent, slot - 1); 2321 gen = btrfs_node_ptr_generation(parent, slot - 1); 2322 eb = find_extent_buffer(fs_info, block1); 2323 /* 2324 * if we get -eagain from btrfs_buffer_uptodate, we 2325 * don't want to return eagain here. That will loop 2326 * forever 2327 */ 2328 if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0) 2329 block1 = 0; 2330 free_extent_buffer(eb); 2331 } 2332 if (slot + 1 < nritems) { 2333 block2 = btrfs_node_blockptr(parent, slot + 1); 2334 gen = btrfs_node_ptr_generation(parent, slot + 1); 2335 eb = find_extent_buffer(fs_info, block2); 2336 if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0) 2337 block2 = 0; 2338 free_extent_buffer(eb); 2339 } 2340 2341 if (block1) 2342 readahead_tree_block(fs_info, block1); 2343 if (block2) 2344 readahead_tree_block(fs_info, block2); 2345 } 2346 2347 2348 /* 2349 * when we walk down the tree, it is usually safe to unlock the higher layers 2350 * in the tree. The exceptions are when our path goes through slot 0, because 2351 * operations on the tree might require changing key pointers higher up in the 2352 * tree. 2353 * 2354 * callers might also have set path->keep_locks, which tells this code to keep 2355 * the lock if the path points to the last slot in the block. This is part of 2356 * walking through the tree, and selecting the next slot in the higher block. 2357 * 2358 * lowest_unlock sets the lowest level in the tree we're allowed to unlock. so 2359 * if lowest_unlock is 1, level 0 won't be unlocked 2360 */ 2361 static noinline void unlock_up(struct btrfs_path *path, int level, 2362 int lowest_unlock, int min_write_lock_level, 2363 int *write_lock_level) 2364 { 2365 int i; 2366 int skip_level = level; 2367 int no_skips = 0; 2368 struct extent_buffer *t; 2369 2370 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 2371 if (!path->nodes[i]) 2372 break; 2373 if (!path->locks[i]) 2374 break; 2375 if (!no_skips && path->slots[i] == 0) { 2376 skip_level = i + 1; 2377 continue; 2378 } 2379 if (!no_skips && path->keep_locks) { 2380 u32 nritems; 2381 t = path->nodes[i]; 2382 nritems = btrfs_header_nritems(t); 2383 if (nritems < 1 || path->slots[i] >= nritems - 1) { 2384 skip_level = i + 1; 2385 continue; 2386 } 2387 } 2388 if (skip_level < i && i >= lowest_unlock) 2389 no_skips = 1; 2390 2391 t = path->nodes[i]; 2392 if (i >= lowest_unlock && i > skip_level && path->locks[i]) { 2393 btrfs_tree_unlock_rw(t, path->locks[i]); 2394 path->locks[i] = 0; 2395 if (write_lock_level && 2396 i > min_write_lock_level && 2397 i <= *write_lock_level) { 2398 *write_lock_level = i - 1; 2399 } 2400 } 2401 } 2402 } 2403 2404 /* 2405 * This releases any locks held in the path starting at level and 2406 * going all the way up to the root. 2407 * 2408 * btrfs_search_slot will keep the lock held on higher nodes in a few 2409 * corner cases, such as COW of the block at slot zero in the node. This 2410 * ignores those rules, and it should only be called when there are no 2411 * more updates to be done higher up in the tree. 2412 */ 2413 noinline void btrfs_unlock_up_safe(struct btrfs_path *path, int level) 2414 { 2415 int i; 2416 2417 if (path->keep_locks) 2418 return; 2419 2420 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 2421 if (!path->nodes[i]) 2422 continue; 2423 if (!path->locks[i]) 2424 continue; 2425 btrfs_tree_unlock_rw(path->nodes[i], path->locks[i]); 2426 path->locks[i] = 0; 2427 } 2428 } 2429 2430 /* 2431 * helper function for btrfs_search_slot. The goal is to find a block 2432 * in cache without setting the path to blocking. If we find the block 2433 * we return zero and the path is unchanged. 2434 * 2435 * If we can't find the block, we set the path blocking and do some 2436 * reada. -EAGAIN is returned and the search must be repeated. 2437 */ 2438 static int 2439 read_block_for_search(struct btrfs_root *root, struct btrfs_path *p, 2440 struct extent_buffer **eb_ret, int level, int slot, 2441 const struct btrfs_key *key) 2442 { 2443 struct btrfs_fs_info *fs_info = root->fs_info; 2444 u64 blocknr; 2445 u64 gen; 2446 struct extent_buffer *b = *eb_ret; 2447 struct extent_buffer *tmp; 2448 int ret; 2449 2450 blocknr = btrfs_node_blockptr(b, slot); 2451 gen = btrfs_node_ptr_generation(b, slot); 2452 2453 tmp = find_extent_buffer(fs_info, blocknr); 2454 if (tmp) { 2455 /* first we do an atomic uptodate check */ 2456 if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) { 2457 *eb_ret = tmp; 2458 return 0; 2459 } 2460 2461 /* the pages were up to date, but we failed 2462 * the generation number check. Do a full 2463 * read for the generation number that is correct. 2464 * We must do this without dropping locks so 2465 * we can trust our generation number 2466 */ 2467 btrfs_set_path_blocking(p); 2468 2469 /* now we're allowed to do a blocking uptodate check */ 2470 ret = btrfs_read_buffer(tmp, gen); 2471 if (!ret) { 2472 *eb_ret = tmp; 2473 return 0; 2474 } 2475 free_extent_buffer(tmp); 2476 btrfs_release_path(p); 2477 return -EIO; 2478 } 2479 2480 /* 2481 * reduce lock contention at high levels 2482 * of the btree by dropping locks before 2483 * we read. Don't release the lock on the current 2484 * level because we need to walk this node to figure 2485 * out which blocks to read. 2486 */ 2487 btrfs_unlock_up_safe(p, level + 1); 2488 btrfs_set_path_blocking(p); 2489 2490 free_extent_buffer(tmp); 2491 if (p->reada != READA_NONE) 2492 reada_for_search(fs_info, p, level, slot, key->objectid); 2493 2494 btrfs_release_path(p); 2495 2496 ret = -EAGAIN; 2497 tmp = read_tree_block(fs_info, blocknr, 0); 2498 if (!IS_ERR(tmp)) { 2499 /* 2500 * If the read above didn't mark this buffer up to date, 2501 * it will never end up being up to date. Set ret to EIO now 2502 * and give up so that our caller doesn't loop forever 2503 * on our EAGAINs. 2504 */ 2505 if (!btrfs_buffer_uptodate(tmp, 0, 0)) 2506 ret = -EIO; 2507 free_extent_buffer(tmp); 2508 } else { 2509 ret = PTR_ERR(tmp); 2510 } 2511 return ret; 2512 } 2513 2514 /* 2515 * helper function for btrfs_search_slot. This does all of the checks 2516 * for node-level blocks and does any balancing required based on 2517 * the ins_len. 2518 * 2519 * If no extra work was required, zero is returned. If we had to 2520 * drop the path, -EAGAIN is returned and btrfs_search_slot must 2521 * start over 2522 */ 2523 static int 2524 setup_nodes_for_search(struct btrfs_trans_handle *trans, 2525 struct btrfs_root *root, struct btrfs_path *p, 2526 struct extent_buffer *b, int level, int ins_len, 2527 int *write_lock_level) 2528 { 2529 struct btrfs_fs_info *fs_info = root->fs_info; 2530 int ret; 2531 2532 if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >= 2533 BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) { 2534 int sret; 2535 2536 if (*write_lock_level < level + 1) { 2537 *write_lock_level = level + 1; 2538 btrfs_release_path(p); 2539 goto again; 2540 } 2541 2542 btrfs_set_path_blocking(p); 2543 reada_for_balance(fs_info, p, level); 2544 sret = split_node(trans, root, p, level); 2545 btrfs_clear_path_blocking(p, NULL, 0); 2546 2547 BUG_ON(sret > 0); 2548 if (sret) { 2549 ret = sret; 2550 goto done; 2551 } 2552 b = p->nodes[level]; 2553 } else if (ins_len < 0 && btrfs_header_nritems(b) < 2554 BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 2) { 2555 int sret; 2556 2557 if (*write_lock_level < level + 1) { 2558 *write_lock_level = level + 1; 2559 btrfs_release_path(p); 2560 goto again; 2561 } 2562 2563 btrfs_set_path_blocking(p); 2564 reada_for_balance(fs_info, p, level); 2565 sret = balance_level(trans, root, p, level); 2566 btrfs_clear_path_blocking(p, NULL, 0); 2567 2568 if (sret) { 2569 ret = sret; 2570 goto done; 2571 } 2572 b = p->nodes[level]; 2573 if (!b) { 2574 btrfs_release_path(p); 2575 goto again; 2576 } 2577 BUG_ON(btrfs_header_nritems(b) == 1); 2578 } 2579 return 0; 2580 2581 again: 2582 ret = -EAGAIN; 2583 done: 2584 return ret; 2585 } 2586 2587 static void key_search_validate(struct extent_buffer *b, 2588 const struct btrfs_key *key, 2589 int level) 2590 { 2591 #ifdef CONFIG_BTRFS_ASSERT 2592 struct btrfs_disk_key disk_key; 2593 2594 btrfs_cpu_key_to_disk(&disk_key, key); 2595 2596 if (level == 0) 2597 ASSERT(!memcmp_extent_buffer(b, &disk_key, 2598 offsetof(struct btrfs_leaf, items[0].key), 2599 sizeof(disk_key))); 2600 else 2601 ASSERT(!memcmp_extent_buffer(b, &disk_key, 2602 offsetof(struct btrfs_node, ptrs[0].key), 2603 sizeof(disk_key))); 2604 #endif 2605 } 2606 2607 static int key_search(struct extent_buffer *b, const struct btrfs_key *key, 2608 int level, int *prev_cmp, int *slot) 2609 { 2610 if (*prev_cmp != 0) { 2611 *prev_cmp = bin_search(b, key, level, slot); 2612 return *prev_cmp; 2613 } 2614 2615 key_search_validate(b, key, level); 2616 *slot = 0; 2617 2618 return 0; 2619 } 2620 2621 int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path, 2622 u64 iobjectid, u64 ioff, u8 key_type, 2623 struct btrfs_key *found_key) 2624 { 2625 int ret; 2626 struct btrfs_key key; 2627 struct extent_buffer *eb; 2628 2629 ASSERT(path); 2630 ASSERT(found_key); 2631 2632 key.type = key_type; 2633 key.objectid = iobjectid; 2634 key.offset = ioff; 2635 2636 ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0); 2637 if (ret < 0) 2638 return ret; 2639 2640 eb = path->nodes[0]; 2641 if (ret && path->slots[0] >= btrfs_header_nritems(eb)) { 2642 ret = btrfs_next_leaf(fs_root, path); 2643 if (ret) 2644 return ret; 2645 eb = path->nodes[0]; 2646 } 2647 2648 btrfs_item_key_to_cpu(eb, found_key, path->slots[0]); 2649 if (found_key->type != key.type || 2650 found_key->objectid != key.objectid) 2651 return 1; 2652 2653 return 0; 2654 } 2655 2656 /* 2657 * look for key in the tree. path is filled in with nodes along the way 2658 * if key is found, we return zero and you can find the item in the leaf 2659 * level of the path (level 0) 2660 * 2661 * If the key isn't found, the path points to the slot where it should 2662 * be inserted, and 1 is returned. If there are other errors during the 2663 * search a negative error number is returned. 2664 * 2665 * if ins_len > 0, nodes and leaves will be split as we walk down the 2666 * tree. if ins_len < 0, nodes will be merged as we walk down the tree (if 2667 * possible) 2668 */ 2669 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root, 2670 const struct btrfs_key *key, struct btrfs_path *p, 2671 int ins_len, int cow) 2672 { 2673 struct btrfs_fs_info *fs_info = root->fs_info; 2674 struct extent_buffer *b; 2675 int slot; 2676 int ret; 2677 int err; 2678 int level; 2679 int lowest_unlock = 1; 2680 int root_lock; 2681 /* everything at write_lock_level or lower must be write locked */ 2682 int write_lock_level = 0; 2683 u8 lowest_level = 0; 2684 int min_write_lock_level; 2685 int prev_cmp; 2686 2687 lowest_level = p->lowest_level; 2688 WARN_ON(lowest_level && ins_len > 0); 2689 WARN_ON(p->nodes[0] != NULL); 2690 BUG_ON(!cow && ins_len); 2691 2692 if (ins_len < 0) { 2693 lowest_unlock = 2; 2694 2695 /* when we are removing items, we might have to go up to level 2696 * two as we update tree pointers Make sure we keep write 2697 * for those levels as well 2698 */ 2699 write_lock_level = 2; 2700 } else if (ins_len > 0) { 2701 /* 2702 * for inserting items, make sure we have a write lock on 2703 * level 1 so we can update keys 2704 */ 2705 write_lock_level = 1; 2706 } 2707 2708 if (!cow) 2709 write_lock_level = -1; 2710 2711 if (cow && (p->keep_locks || p->lowest_level)) 2712 write_lock_level = BTRFS_MAX_LEVEL; 2713 2714 min_write_lock_level = write_lock_level; 2715 2716 again: 2717 prev_cmp = -1; 2718 /* 2719 * we try very hard to do read locks on the root 2720 */ 2721 root_lock = BTRFS_READ_LOCK; 2722 level = 0; 2723 if (p->search_commit_root) { 2724 /* 2725 * the commit roots are read only 2726 * so we always do read locks 2727 */ 2728 if (p->need_commit_sem) 2729 down_read(&fs_info->commit_root_sem); 2730 b = root->commit_root; 2731 extent_buffer_get(b); 2732 level = btrfs_header_level(b); 2733 if (p->need_commit_sem) 2734 up_read(&fs_info->commit_root_sem); 2735 if (!p->skip_locking) 2736 btrfs_tree_read_lock(b); 2737 } else { 2738 if (p->skip_locking) { 2739 b = btrfs_root_node(root); 2740 level = btrfs_header_level(b); 2741 } else { 2742 /* we don't know the level of the root node 2743 * until we actually have it read locked 2744 */ 2745 b = btrfs_read_lock_root_node(root); 2746 level = btrfs_header_level(b); 2747 if (level <= write_lock_level) { 2748 /* whoops, must trade for write lock */ 2749 btrfs_tree_read_unlock(b); 2750 free_extent_buffer(b); 2751 b = btrfs_lock_root_node(root); 2752 root_lock = BTRFS_WRITE_LOCK; 2753 2754 /* the level might have changed, check again */ 2755 level = btrfs_header_level(b); 2756 } 2757 } 2758 } 2759 p->nodes[level] = b; 2760 if (!p->skip_locking) 2761 p->locks[level] = root_lock; 2762 2763 while (b) { 2764 level = btrfs_header_level(b); 2765 2766 /* 2767 * setup the path here so we can release it under lock 2768 * contention with the cow code 2769 */ 2770 if (cow) { 2771 /* 2772 * if we don't really need to cow this block 2773 * then we don't want to set the path blocking, 2774 * so we test it here 2775 */ 2776 if (!should_cow_block(trans, root, b)) { 2777 trans->dirty = true; 2778 goto cow_done; 2779 } 2780 2781 /* 2782 * must have write locks on this node and the 2783 * parent 2784 */ 2785 if (level > write_lock_level || 2786 (level + 1 > write_lock_level && 2787 level + 1 < BTRFS_MAX_LEVEL && 2788 p->nodes[level + 1])) { 2789 write_lock_level = level + 1; 2790 btrfs_release_path(p); 2791 goto again; 2792 } 2793 2794 btrfs_set_path_blocking(p); 2795 err = btrfs_cow_block(trans, root, b, 2796 p->nodes[level + 1], 2797 p->slots[level + 1], &b); 2798 if (err) { 2799 ret = err; 2800 goto done; 2801 } 2802 } 2803 cow_done: 2804 p->nodes[level] = b; 2805 btrfs_clear_path_blocking(p, NULL, 0); 2806 2807 /* 2808 * we have a lock on b and as long as we aren't changing 2809 * the tree, there is no way to for the items in b to change. 2810 * It is safe to drop the lock on our parent before we 2811 * go through the expensive btree search on b. 2812 * 2813 * If we're inserting or deleting (ins_len != 0), then we might 2814 * be changing slot zero, which may require changing the parent. 2815 * So, we can't drop the lock until after we know which slot 2816 * we're operating on. 2817 */ 2818 if (!ins_len && !p->keep_locks) { 2819 int u = level + 1; 2820 2821 if (u < BTRFS_MAX_LEVEL && p->locks[u]) { 2822 btrfs_tree_unlock_rw(p->nodes[u], p->locks[u]); 2823 p->locks[u] = 0; 2824 } 2825 } 2826 2827 ret = key_search(b, key, level, &prev_cmp, &slot); 2828 if (ret < 0) 2829 goto done; 2830 2831 if (level != 0) { 2832 int dec = 0; 2833 if (ret && slot > 0) { 2834 dec = 1; 2835 slot -= 1; 2836 } 2837 p->slots[level] = slot; 2838 err = setup_nodes_for_search(trans, root, p, b, level, 2839 ins_len, &write_lock_level); 2840 if (err == -EAGAIN) 2841 goto again; 2842 if (err) { 2843 ret = err; 2844 goto done; 2845 } 2846 b = p->nodes[level]; 2847 slot = p->slots[level]; 2848 2849 /* 2850 * slot 0 is special, if we change the key 2851 * we have to update the parent pointer 2852 * which means we must have a write lock 2853 * on the parent 2854 */ 2855 if (slot == 0 && ins_len && 2856 write_lock_level < level + 1) { 2857 write_lock_level = level + 1; 2858 btrfs_release_path(p); 2859 goto again; 2860 } 2861 2862 unlock_up(p, level, lowest_unlock, 2863 min_write_lock_level, &write_lock_level); 2864 2865 if (level == lowest_level) { 2866 if (dec) 2867 p->slots[level]++; 2868 goto done; 2869 } 2870 2871 err = read_block_for_search(root, p, &b, level, 2872 slot, key); 2873 if (err == -EAGAIN) 2874 goto again; 2875 if (err) { 2876 ret = err; 2877 goto done; 2878 } 2879 2880 if (!p->skip_locking) { 2881 level = btrfs_header_level(b); 2882 if (level <= write_lock_level) { 2883 err = btrfs_try_tree_write_lock(b); 2884 if (!err) { 2885 btrfs_set_path_blocking(p); 2886 btrfs_tree_lock(b); 2887 btrfs_clear_path_blocking(p, b, 2888 BTRFS_WRITE_LOCK); 2889 } 2890 p->locks[level] = BTRFS_WRITE_LOCK; 2891 } else { 2892 err = btrfs_tree_read_lock_atomic(b); 2893 if (!err) { 2894 btrfs_set_path_blocking(p); 2895 btrfs_tree_read_lock(b); 2896 btrfs_clear_path_blocking(p, b, 2897 BTRFS_READ_LOCK); 2898 } 2899 p->locks[level] = BTRFS_READ_LOCK; 2900 } 2901 p->nodes[level] = b; 2902 } 2903 } else { 2904 p->slots[level] = slot; 2905 if (ins_len > 0 && 2906 btrfs_leaf_free_space(fs_info, b) < ins_len) { 2907 if (write_lock_level < 1) { 2908 write_lock_level = 1; 2909 btrfs_release_path(p); 2910 goto again; 2911 } 2912 2913 btrfs_set_path_blocking(p); 2914 err = split_leaf(trans, root, key, 2915 p, ins_len, ret == 0); 2916 btrfs_clear_path_blocking(p, NULL, 0); 2917 2918 BUG_ON(err > 0); 2919 if (err) { 2920 ret = err; 2921 goto done; 2922 } 2923 } 2924 if (!p->search_for_split) 2925 unlock_up(p, level, lowest_unlock, 2926 min_write_lock_level, &write_lock_level); 2927 goto done; 2928 } 2929 } 2930 ret = 1; 2931 done: 2932 /* 2933 * we don't really know what they plan on doing with the path 2934 * from here on, so for now just mark it as blocking 2935 */ 2936 if (!p->leave_spinning) 2937 btrfs_set_path_blocking(p); 2938 if (ret < 0 && !p->skip_release_on_error) 2939 btrfs_release_path(p); 2940 return ret; 2941 } 2942 2943 /* 2944 * Like btrfs_search_slot, this looks for a key in the given tree. It uses the 2945 * current state of the tree together with the operations recorded in the tree 2946 * modification log to search for the key in a previous version of this tree, as 2947 * denoted by the time_seq parameter. 2948 * 2949 * Naturally, there is no support for insert, delete or cow operations. 2950 * 2951 * The resulting path and return value will be set up as if we called 2952 * btrfs_search_slot at that point in time with ins_len and cow both set to 0. 2953 */ 2954 int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key, 2955 struct btrfs_path *p, u64 time_seq) 2956 { 2957 struct btrfs_fs_info *fs_info = root->fs_info; 2958 struct extent_buffer *b; 2959 int slot; 2960 int ret; 2961 int err; 2962 int level; 2963 int lowest_unlock = 1; 2964 u8 lowest_level = 0; 2965 int prev_cmp = -1; 2966 2967 lowest_level = p->lowest_level; 2968 WARN_ON(p->nodes[0] != NULL); 2969 2970 if (p->search_commit_root) { 2971 BUG_ON(time_seq); 2972 return btrfs_search_slot(NULL, root, key, p, 0, 0); 2973 } 2974 2975 again: 2976 b = get_old_root(root, time_seq); 2977 level = btrfs_header_level(b); 2978 p->locks[level] = BTRFS_READ_LOCK; 2979 2980 while (b) { 2981 level = btrfs_header_level(b); 2982 p->nodes[level] = b; 2983 btrfs_clear_path_blocking(p, NULL, 0); 2984 2985 /* 2986 * we have a lock on b and as long as we aren't changing 2987 * the tree, there is no way to for the items in b to change. 2988 * It is safe to drop the lock on our parent before we 2989 * go through the expensive btree search on b. 2990 */ 2991 btrfs_unlock_up_safe(p, level + 1); 2992 2993 /* 2994 * Since we can unwind ebs we want to do a real search every 2995 * time. 2996 */ 2997 prev_cmp = -1; 2998 ret = key_search(b, key, level, &prev_cmp, &slot); 2999 3000 if (level != 0) { 3001 int dec = 0; 3002 if (ret && slot > 0) { 3003 dec = 1; 3004 slot -= 1; 3005 } 3006 p->slots[level] = slot; 3007 unlock_up(p, level, lowest_unlock, 0, NULL); 3008 3009 if (level == lowest_level) { 3010 if (dec) 3011 p->slots[level]++; 3012 goto done; 3013 } 3014 3015 err = read_block_for_search(root, p, &b, level, 3016 slot, key); 3017 if (err == -EAGAIN) 3018 goto again; 3019 if (err) { 3020 ret = err; 3021 goto done; 3022 } 3023 3024 level = btrfs_header_level(b); 3025 err = btrfs_tree_read_lock_atomic(b); 3026 if (!err) { 3027 btrfs_set_path_blocking(p); 3028 btrfs_tree_read_lock(b); 3029 btrfs_clear_path_blocking(p, b, 3030 BTRFS_READ_LOCK); 3031 } 3032 b = tree_mod_log_rewind(fs_info, p, b, time_seq); 3033 if (!b) { 3034 ret = -ENOMEM; 3035 goto done; 3036 } 3037 p->locks[level] = BTRFS_READ_LOCK; 3038 p->nodes[level] = b; 3039 } else { 3040 p->slots[level] = slot; 3041 unlock_up(p, level, lowest_unlock, 0, NULL); 3042 goto done; 3043 } 3044 } 3045 ret = 1; 3046 done: 3047 if (!p->leave_spinning) 3048 btrfs_set_path_blocking(p); 3049 if (ret < 0) 3050 btrfs_release_path(p); 3051 3052 return ret; 3053 } 3054 3055 /* 3056 * helper to use instead of search slot if no exact match is needed but 3057 * instead the next or previous item should be returned. 3058 * When find_higher is true, the next higher item is returned, the next lower 3059 * otherwise. 3060 * When return_any and find_higher are both true, and no higher item is found, 3061 * return the next lower instead. 3062 * When return_any is true and find_higher is false, and no lower item is found, 3063 * return the next higher instead. 3064 * It returns 0 if any item is found, 1 if none is found (tree empty), and 3065 * < 0 on error 3066 */ 3067 int btrfs_search_slot_for_read(struct btrfs_root *root, 3068 const struct btrfs_key *key, 3069 struct btrfs_path *p, int find_higher, 3070 int return_any) 3071 { 3072 int ret; 3073 struct extent_buffer *leaf; 3074 3075 again: 3076 ret = btrfs_search_slot(NULL, root, key, p, 0, 0); 3077 if (ret <= 0) 3078 return ret; 3079 /* 3080 * a return value of 1 means the path is at the position where the 3081 * item should be inserted. Normally this is the next bigger item, 3082 * but in case the previous item is the last in a leaf, path points 3083 * to the first free slot in the previous leaf, i.e. at an invalid 3084 * item. 3085 */ 3086 leaf = p->nodes[0]; 3087 3088 if (find_higher) { 3089 if (p->slots[0] >= btrfs_header_nritems(leaf)) { 3090 ret = btrfs_next_leaf(root, p); 3091 if (ret <= 0) 3092 return ret; 3093 if (!return_any) 3094 return 1; 3095 /* 3096 * no higher item found, return the next 3097 * lower instead 3098 */ 3099 return_any = 0; 3100 find_higher = 0; 3101 btrfs_release_path(p); 3102 goto again; 3103 } 3104 } else { 3105 if (p->slots[0] == 0) { 3106 ret = btrfs_prev_leaf(root, p); 3107 if (ret < 0) 3108 return ret; 3109 if (!ret) { 3110 leaf = p->nodes[0]; 3111 if (p->slots[0] == btrfs_header_nritems(leaf)) 3112 p->slots[0]--; 3113 return 0; 3114 } 3115 if (!return_any) 3116 return 1; 3117 /* 3118 * no lower item found, return the next 3119 * higher instead 3120 */ 3121 return_any = 0; 3122 find_higher = 1; 3123 btrfs_release_path(p); 3124 goto again; 3125 } else { 3126 --p->slots[0]; 3127 } 3128 } 3129 return 0; 3130 } 3131 3132 /* 3133 * adjust the pointers going up the tree, starting at level 3134 * making sure the right key of each node is points to 'key'. 3135 * This is used after shifting pointers to the left, so it stops 3136 * fixing up pointers when a given leaf/node is not in slot 0 of the 3137 * higher levels 3138 * 3139 */ 3140 static void fixup_low_keys(struct btrfs_fs_info *fs_info, 3141 struct btrfs_path *path, 3142 struct btrfs_disk_key *key, int level) 3143 { 3144 int i; 3145 struct extent_buffer *t; 3146 3147 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 3148 int tslot = path->slots[i]; 3149 if (!path->nodes[i]) 3150 break; 3151 t = path->nodes[i]; 3152 tree_mod_log_set_node_key(fs_info, t, tslot, 1); 3153 btrfs_set_node_key(t, key, tslot); 3154 btrfs_mark_buffer_dirty(path->nodes[i]); 3155 if (tslot != 0) 3156 break; 3157 } 3158 } 3159 3160 /* 3161 * update item key. 3162 * 3163 * This function isn't completely safe. It's the caller's responsibility 3164 * that the new key won't break the order 3165 */ 3166 void btrfs_set_item_key_safe(struct btrfs_fs_info *fs_info, 3167 struct btrfs_path *path, 3168 const struct btrfs_key *new_key) 3169 { 3170 struct btrfs_disk_key disk_key; 3171 struct extent_buffer *eb; 3172 int slot; 3173 3174 eb = path->nodes[0]; 3175 slot = path->slots[0]; 3176 if (slot > 0) { 3177 btrfs_item_key(eb, &disk_key, slot - 1); 3178 BUG_ON(comp_keys(&disk_key, new_key) >= 0); 3179 } 3180 if (slot < btrfs_header_nritems(eb) - 1) { 3181 btrfs_item_key(eb, &disk_key, slot + 1); 3182 BUG_ON(comp_keys(&disk_key, new_key) <= 0); 3183 } 3184 3185 btrfs_cpu_key_to_disk(&disk_key, new_key); 3186 btrfs_set_item_key(eb, &disk_key, slot); 3187 btrfs_mark_buffer_dirty(eb); 3188 if (slot == 0) 3189 fixup_low_keys(fs_info, path, &disk_key, 1); 3190 } 3191 3192 /* 3193 * try to push data from one node into the next node left in the 3194 * tree. 3195 * 3196 * returns 0 if some ptrs were pushed left, < 0 if there was some horrible 3197 * error, and > 0 if there was no room in the left hand block. 3198 */ 3199 static int push_node_left(struct btrfs_trans_handle *trans, 3200 struct btrfs_fs_info *fs_info, 3201 struct extent_buffer *dst, 3202 struct extent_buffer *src, int empty) 3203 { 3204 int push_items = 0; 3205 int src_nritems; 3206 int dst_nritems; 3207 int ret = 0; 3208 3209 src_nritems = btrfs_header_nritems(src); 3210 dst_nritems = btrfs_header_nritems(dst); 3211 push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems; 3212 WARN_ON(btrfs_header_generation(src) != trans->transid); 3213 WARN_ON(btrfs_header_generation(dst) != trans->transid); 3214 3215 if (!empty && src_nritems <= 8) 3216 return 1; 3217 3218 if (push_items <= 0) 3219 return 1; 3220 3221 if (empty) { 3222 push_items = min(src_nritems, push_items); 3223 if (push_items < src_nritems) { 3224 /* leave at least 8 pointers in the node if 3225 * we aren't going to empty it 3226 */ 3227 if (src_nritems - push_items < 8) { 3228 if (push_items <= 8) 3229 return 1; 3230 push_items -= 8; 3231 } 3232 } 3233 } else 3234 push_items = min(src_nritems - 8, push_items); 3235 3236 ret = tree_mod_log_eb_copy(fs_info, dst, src, dst_nritems, 0, 3237 push_items); 3238 if (ret) { 3239 btrfs_abort_transaction(trans, ret); 3240 return ret; 3241 } 3242 copy_extent_buffer(dst, src, 3243 btrfs_node_key_ptr_offset(dst_nritems), 3244 btrfs_node_key_ptr_offset(0), 3245 push_items * sizeof(struct btrfs_key_ptr)); 3246 3247 if (push_items < src_nritems) { 3248 /* 3249 * don't call tree_mod_log_eb_move here, key removal was already 3250 * fully logged by tree_mod_log_eb_copy above. 3251 */ 3252 memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0), 3253 btrfs_node_key_ptr_offset(push_items), 3254 (src_nritems - push_items) * 3255 sizeof(struct btrfs_key_ptr)); 3256 } 3257 btrfs_set_header_nritems(src, src_nritems - push_items); 3258 btrfs_set_header_nritems(dst, dst_nritems + push_items); 3259 btrfs_mark_buffer_dirty(src); 3260 btrfs_mark_buffer_dirty(dst); 3261 3262 return ret; 3263 } 3264 3265 /* 3266 * try to push data from one node into the next node right in the 3267 * tree. 3268 * 3269 * returns 0 if some ptrs were pushed, < 0 if there was some horrible 3270 * error, and > 0 if there was no room in the right hand block. 3271 * 3272 * this will only push up to 1/2 the contents of the left node over 3273 */ 3274 static int balance_node_right(struct btrfs_trans_handle *trans, 3275 struct btrfs_fs_info *fs_info, 3276 struct extent_buffer *dst, 3277 struct extent_buffer *src) 3278 { 3279 int push_items = 0; 3280 int max_push; 3281 int src_nritems; 3282 int dst_nritems; 3283 int ret = 0; 3284 3285 WARN_ON(btrfs_header_generation(src) != trans->transid); 3286 WARN_ON(btrfs_header_generation(dst) != trans->transid); 3287 3288 src_nritems = btrfs_header_nritems(src); 3289 dst_nritems = btrfs_header_nritems(dst); 3290 push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems; 3291 if (push_items <= 0) 3292 return 1; 3293 3294 if (src_nritems < 4) 3295 return 1; 3296 3297 max_push = src_nritems / 2 + 1; 3298 /* don't try to empty the node */ 3299 if (max_push >= src_nritems) 3300 return 1; 3301 3302 if (max_push < push_items) 3303 push_items = max_push; 3304 3305 tree_mod_log_eb_move(fs_info, dst, push_items, 0, dst_nritems); 3306 memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items), 3307 btrfs_node_key_ptr_offset(0), 3308 (dst_nritems) * 3309 sizeof(struct btrfs_key_ptr)); 3310 3311 ret = tree_mod_log_eb_copy(fs_info, dst, src, 0, 3312 src_nritems - push_items, push_items); 3313 if (ret) { 3314 btrfs_abort_transaction(trans, ret); 3315 return ret; 3316 } 3317 copy_extent_buffer(dst, src, 3318 btrfs_node_key_ptr_offset(0), 3319 btrfs_node_key_ptr_offset(src_nritems - push_items), 3320 push_items * sizeof(struct btrfs_key_ptr)); 3321 3322 btrfs_set_header_nritems(src, src_nritems - push_items); 3323 btrfs_set_header_nritems(dst, dst_nritems + push_items); 3324 3325 btrfs_mark_buffer_dirty(src); 3326 btrfs_mark_buffer_dirty(dst); 3327 3328 return ret; 3329 } 3330 3331 /* 3332 * helper function to insert a new root level in the tree. 3333 * A new node is allocated, and a single item is inserted to 3334 * point to the existing root 3335 * 3336 * returns zero on success or < 0 on failure. 3337 */ 3338 static noinline int insert_new_root(struct btrfs_trans_handle *trans, 3339 struct btrfs_root *root, 3340 struct btrfs_path *path, int level) 3341 { 3342 struct btrfs_fs_info *fs_info = root->fs_info; 3343 u64 lower_gen; 3344 struct extent_buffer *lower; 3345 struct extent_buffer *c; 3346 struct extent_buffer *old; 3347 struct btrfs_disk_key lower_key; 3348 3349 BUG_ON(path->nodes[level]); 3350 BUG_ON(path->nodes[level-1] != root->node); 3351 3352 lower = path->nodes[level-1]; 3353 if (level == 1) 3354 btrfs_item_key(lower, &lower_key, 0); 3355 else 3356 btrfs_node_key(lower, &lower_key, 0); 3357 3358 c = btrfs_alloc_tree_block(trans, root, 0, root->root_key.objectid, 3359 &lower_key, level, root->node->start, 0); 3360 if (IS_ERR(c)) 3361 return PTR_ERR(c); 3362 3363 root_add_used(root, fs_info->nodesize); 3364 3365 memzero_extent_buffer(c, 0, sizeof(struct btrfs_header)); 3366 btrfs_set_header_nritems(c, 1); 3367 btrfs_set_header_level(c, level); 3368 btrfs_set_header_bytenr(c, c->start); 3369 btrfs_set_header_generation(c, trans->transid); 3370 btrfs_set_header_backref_rev(c, BTRFS_MIXED_BACKREF_REV); 3371 btrfs_set_header_owner(c, root->root_key.objectid); 3372 3373 write_extent_buffer_fsid(c, fs_info->fsid); 3374 write_extent_buffer_chunk_tree_uuid(c, fs_info->chunk_tree_uuid); 3375 3376 btrfs_set_node_key(c, &lower_key, 0); 3377 btrfs_set_node_blockptr(c, 0, lower->start); 3378 lower_gen = btrfs_header_generation(lower); 3379 WARN_ON(lower_gen != trans->transid); 3380 3381 btrfs_set_node_ptr_generation(c, 0, lower_gen); 3382 3383 btrfs_mark_buffer_dirty(c); 3384 3385 old = root->node; 3386 tree_mod_log_set_root_pointer(root, c, 0); 3387 rcu_assign_pointer(root->node, c); 3388 3389 /* the super has an extra ref to root->node */ 3390 free_extent_buffer(old); 3391 3392 add_root_to_dirty_list(root); 3393 extent_buffer_get(c); 3394 path->nodes[level] = c; 3395 path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING; 3396 path->slots[level] = 0; 3397 return 0; 3398 } 3399 3400 /* 3401 * worker function to insert a single pointer in a node. 3402 * the node should have enough room for the pointer already 3403 * 3404 * slot and level indicate where you want the key to go, and 3405 * blocknr is the block the key points to. 3406 */ 3407 static void insert_ptr(struct btrfs_trans_handle *trans, 3408 struct btrfs_fs_info *fs_info, struct btrfs_path *path, 3409 struct btrfs_disk_key *key, u64 bytenr, 3410 int slot, int level) 3411 { 3412 struct extent_buffer *lower; 3413 int nritems; 3414 int ret; 3415 3416 BUG_ON(!path->nodes[level]); 3417 btrfs_assert_tree_locked(path->nodes[level]); 3418 lower = path->nodes[level]; 3419 nritems = btrfs_header_nritems(lower); 3420 BUG_ON(slot > nritems); 3421 BUG_ON(nritems == BTRFS_NODEPTRS_PER_BLOCK(fs_info)); 3422 if (slot != nritems) { 3423 if (level) 3424 tree_mod_log_eb_move(fs_info, lower, slot + 1, 3425 slot, nritems - slot); 3426 memmove_extent_buffer(lower, 3427 btrfs_node_key_ptr_offset(slot + 1), 3428 btrfs_node_key_ptr_offset(slot), 3429 (nritems - slot) * sizeof(struct btrfs_key_ptr)); 3430 } 3431 if (level) { 3432 ret = tree_mod_log_insert_key(fs_info, lower, slot, 3433 MOD_LOG_KEY_ADD, GFP_NOFS); 3434 BUG_ON(ret < 0); 3435 } 3436 btrfs_set_node_key(lower, key, slot); 3437 btrfs_set_node_blockptr(lower, slot, bytenr); 3438 WARN_ON(trans->transid == 0); 3439 btrfs_set_node_ptr_generation(lower, slot, trans->transid); 3440 btrfs_set_header_nritems(lower, nritems + 1); 3441 btrfs_mark_buffer_dirty(lower); 3442 } 3443 3444 /* 3445 * split the node at the specified level in path in two. 3446 * The path is corrected to point to the appropriate node after the split 3447 * 3448 * Before splitting this tries to make some room in the node by pushing 3449 * left and right, if either one works, it returns right away. 3450 * 3451 * returns 0 on success and < 0 on failure 3452 */ 3453 static noinline int split_node(struct btrfs_trans_handle *trans, 3454 struct btrfs_root *root, 3455 struct btrfs_path *path, int level) 3456 { 3457 struct btrfs_fs_info *fs_info = root->fs_info; 3458 struct extent_buffer *c; 3459 struct extent_buffer *split; 3460 struct btrfs_disk_key disk_key; 3461 int mid; 3462 int ret; 3463 u32 c_nritems; 3464 3465 c = path->nodes[level]; 3466 WARN_ON(btrfs_header_generation(c) != trans->transid); 3467 if (c == root->node) { 3468 /* 3469 * trying to split the root, lets make a new one 3470 * 3471 * tree mod log: We don't log_removal old root in 3472 * insert_new_root, because that root buffer will be kept as a 3473 * normal node. We are going to log removal of half of the 3474 * elements below with tree_mod_log_eb_copy. We're holding a 3475 * tree lock on the buffer, which is why we cannot race with 3476 * other tree_mod_log users. 3477 */ 3478 ret = insert_new_root(trans, root, path, level + 1); 3479 if (ret) 3480 return ret; 3481 } else { 3482 ret = push_nodes_for_insert(trans, root, path, level); 3483 c = path->nodes[level]; 3484 if (!ret && btrfs_header_nritems(c) < 3485 BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) 3486 return 0; 3487 if (ret < 0) 3488 return ret; 3489 } 3490 3491 c_nritems = btrfs_header_nritems(c); 3492 mid = (c_nritems + 1) / 2; 3493 btrfs_node_key(c, &disk_key, mid); 3494 3495 split = btrfs_alloc_tree_block(trans, root, 0, root->root_key.objectid, 3496 &disk_key, level, c->start, 0); 3497 if (IS_ERR(split)) 3498 return PTR_ERR(split); 3499 3500 root_add_used(root, fs_info->nodesize); 3501 3502 memzero_extent_buffer(split, 0, sizeof(struct btrfs_header)); 3503 btrfs_set_header_level(split, btrfs_header_level(c)); 3504 btrfs_set_header_bytenr(split, split->start); 3505 btrfs_set_header_generation(split, trans->transid); 3506 btrfs_set_header_backref_rev(split, BTRFS_MIXED_BACKREF_REV); 3507 btrfs_set_header_owner(split, root->root_key.objectid); 3508 write_extent_buffer_fsid(split, fs_info->fsid); 3509 write_extent_buffer_chunk_tree_uuid(split, fs_info->chunk_tree_uuid); 3510 3511 ret = tree_mod_log_eb_copy(fs_info, split, c, 0, mid, c_nritems - mid); 3512 if (ret) { 3513 btrfs_abort_transaction(trans, ret); 3514 return ret; 3515 } 3516 copy_extent_buffer(split, c, 3517 btrfs_node_key_ptr_offset(0), 3518 btrfs_node_key_ptr_offset(mid), 3519 (c_nritems - mid) * sizeof(struct btrfs_key_ptr)); 3520 btrfs_set_header_nritems(split, c_nritems - mid); 3521 btrfs_set_header_nritems(c, mid); 3522 ret = 0; 3523 3524 btrfs_mark_buffer_dirty(c); 3525 btrfs_mark_buffer_dirty(split); 3526 3527 insert_ptr(trans, fs_info, path, &disk_key, split->start, 3528 path->slots[level + 1] + 1, level + 1); 3529 3530 if (path->slots[level] >= mid) { 3531 path->slots[level] -= mid; 3532 btrfs_tree_unlock(c); 3533 free_extent_buffer(c); 3534 path->nodes[level] = split; 3535 path->slots[level + 1] += 1; 3536 } else { 3537 btrfs_tree_unlock(split); 3538 free_extent_buffer(split); 3539 } 3540 return ret; 3541 } 3542 3543 /* 3544 * how many bytes are required to store the items in a leaf. start 3545 * and nr indicate which items in the leaf to check. This totals up the 3546 * space used both by the item structs and the item data 3547 */ 3548 static int leaf_space_used(struct extent_buffer *l, int start, int nr) 3549 { 3550 struct btrfs_item *start_item; 3551 struct btrfs_item *end_item; 3552 struct btrfs_map_token token; 3553 int data_len; 3554 int nritems = btrfs_header_nritems(l); 3555 int end = min(nritems, start + nr) - 1; 3556 3557 if (!nr) 3558 return 0; 3559 btrfs_init_map_token(&token); 3560 start_item = btrfs_item_nr(start); 3561 end_item = btrfs_item_nr(end); 3562 data_len = btrfs_token_item_offset(l, start_item, &token) + 3563 btrfs_token_item_size(l, start_item, &token); 3564 data_len = data_len - btrfs_token_item_offset(l, end_item, &token); 3565 data_len += sizeof(struct btrfs_item) * nr; 3566 WARN_ON(data_len < 0); 3567 return data_len; 3568 } 3569 3570 /* 3571 * The space between the end of the leaf items and 3572 * the start of the leaf data. IOW, how much room 3573 * the leaf has left for both items and data 3574 */ 3575 noinline int btrfs_leaf_free_space(struct btrfs_fs_info *fs_info, 3576 struct extent_buffer *leaf) 3577 { 3578 int nritems = btrfs_header_nritems(leaf); 3579 int ret; 3580 3581 ret = BTRFS_LEAF_DATA_SIZE(fs_info) - leaf_space_used(leaf, 0, nritems); 3582 if (ret < 0) { 3583 btrfs_crit(fs_info, 3584 "leaf free space ret %d, leaf data size %lu, used %d nritems %d", 3585 ret, 3586 (unsigned long) BTRFS_LEAF_DATA_SIZE(fs_info), 3587 leaf_space_used(leaf, 0, nritems), nritems); 3588 } 3589 return ret; 3590 } 3591 3592 /* 3593 * min slot controls the lowest index we're willing to push to the 3594 * right. We'll push up to and including min_slot, but no lower 3595 */ 3596 static noinline int __push_leaf_right(struct btrfs_fs_info *fs_info, 3597 struct btrfs_path *path, 3598 int data_size, int empty, 3599 struct extent_buffer *right, 3600 int free_space, u32 left_nritems, 3601 u32 min_slot) 3602 { 3603 struct extent_buffer *left = path->nodes[0]; 3604 struct extent_buffer *upper = path->nodes[1]; 3605 struct btrfs_map_token token; 3606 struct btrfs_disk_key disk_key; 3607 int slot; 3608 u32 i; 3609 int push_space = 0; 3610 int push_items = 0; 3611 struct btrfs_item *item; 3612 u32 nr; 3613 u32 right_nritems; 3614 u32 data_end; 3615 u32 this_item_size; 3616 3617 btrfs_init_map_token(&token); 3618 3619 if (empty) 3620 nr = 0; 3621 else 3622 nr = max_t(u32, 1, min_slot); 3623 3624 if (path->slots[0] >= left_nritems) 3625 push_space += data_size; 3626 3627 slot = path->slots[1]; 3628 i = left_nritems - 1; 3629 while (i >= nr) { 3630 item = btrfs_item_nr(i); 3631 3632 if (!empty && push_items > 0) { 3633 if (path->slots[0] > i) 3634 break; 3635 if (path->slots[0] == i) { 3636 int space = btrfs_leaf_free_space(fs_info, left); 3637 if (space + push_space * 2 > free_space) 3638 break; 3639 } 3640 } 3641 3642 if (path->slots[0] == i) 3643 push_space += data_size; 3644 3645 this_item_size = btrfs_item_size(left, item); 3646 if (this_item_size + sizeof(*item) + push_space > free_space) 3647 break; 3648 3649 push_items++; 3650 push_space += this_item_size + sizeof(*item); 3651 if (i == 0) 3652 break; 3653 i--; 3654 } 3655 3656 if (push_items == 0) 3657 goto out_unlock; 3658 3659 WARN_ON(!empty && push_items == left_nritems); 3660 3661 /* push left to right */ 3662 right_nritems = btrfs_header_nritems(right); 3663 3664 push_space = btrfs_item_end_nr(left, left_nritems - push_items); 3665 push_space -= leaf_data_end(fs_info, left); 3666 3667 /* make room in the right data area */ 3668 data_end = leaf_data_end(fs_info, right); 3669 memmove_extent_buffer(right, 3670 BTRFS_LEAF_DATA_OFFSET + data_end - push_space, 3671 BTRFS_LEAF_DATA_OFFSET + data_end, 3672 BTRFS_LEAF_DATA_SIZE(fs_info) - data_end); 3673 3674 /* copy from the left data area */ 3675 copy_extent_buffer(right, left, BTRFS_LEAF_DATA_OFFSET + 3676 BTRFS_LEAF_DATA_SIZE(fs_info) - push_space, 3677 BTRFS_LEAF_DATA_OFFSET + leaf_data_end(fs_info, left), 3678 push_space); 3679 3680 memmove_extent_buffer(right, btrfs_item_nr_offset(push_items), 3681 btrfs_item_nr_offset(0), 3682 right_nritems * sizeof(struct btrfs_item)); 3683 3684 /* copy the items from left to right */ 3685 copy_extent_buffer(right, left, btrfs_item_nr_offset(0), 3686 btrfs_item_nr_offset(left_nritems - push_items), 3687 push_items * sizeof(struct btrfs_item)); 3688 3689 /* update the item pointers */ 3690 right_nritems += push_items; 3691 btrfs_set_header_nritems(right, right_nritems); 3692 push_space = BTRFS_LEAF_DATA_SIZE(fs_info); 3693 for (i = 0; i < right_nritems; i++) { 3694 item = btrfs_item_nr(i); 3695 push_space -= btrfs_token_item_size(right, item, &token); 3696 btrfs_set_token_item_offset(right, item, push_space, &token); 3697 } 3698 3699 left_nritems -= push_items; 3700 btrfs_set_header_nritems(left, left_nritems); 3701 3702 if (left_nritems) 3703 btrfs_mark_buffer_dirty(left); 3704 else 3705 clean_tree_block(fs_info, left); 3706 3707 btrfs_mark_buffer_dirty(right); 3708 3709 btrfs_item_key(right, &disk_key, 0); 3710 btrfs_set_node_key(upper, &disk_key, slot + 1); 3711 btrfs_mark_buffer_dirty(upper); 3712 3713 /* then fixup the leaf pointer in the path */ 3714 if (path->slots[0] >= left_nritems) { 3715 path->slots[0] -= left_nritems; 3716 if (btrfs_header_nritems(path->nodes[0]) == 0) 3717 clean_tree_block(fs_info, path->nodes[0]); 3718 btrfs_tree_unlock(path->nodes[0]); 3719 free_extent_buffer(path->nodes[0]); 3720 path->nodes[0] = right; 3721 path->slots[1] += 1; 3722 } else { 3723 btrfs_tree_unlock(right); 3724 free_extent_buffer(right); 3725 } 3726 return 0; 3727 3728 out_unlock: 3729 btrfs_tree_unlock(right); 3730 free_extent_buffer(right); 3731 return 1; 3732 } 3733 3734 /* 3735 * push some data in the path leaf to the right, trying to free up at 3736 * least data_size bytes. returns zero if the push worked, nonzero otherwise 3737 * 3738 * returns 1 if the push failed because the other node didn't have enough 3739 * room, 0 if everything worked out and < 0 if there were major errors. 3740 * 3741 * this will push starting from min_slot to the end of the leaf. It won't 3742 * push any slot lower than min_slot 3743 */ 3744 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root 3745 *root, struct btrfs_path *path, 3746 int min_data_size, int data_size, 3747 int empty, u32 min_slot) 3748 { 3749 struct btrfs_fs_info *fs_info = root->fs_info; 3750 struct extent_buffer *left = path->nodes[0]; 3751 struct extent_buffer *right; 3752 struct extent_buffer *upper; 3753 int slot; 3754 int free_space; 3755 u32 left_nritems; 3756 int ret; 3757 3758 if (!path->nodes[1]) 3759 return 1; 3760 3761 slot = path->slots[1]; 3762 upper = path->nodes[1]; 3763 if (slot >= btrfs_header_nritems(upper) - 1) 3764 return 1; 3765 3766 btrfs_assert_tree_locked(path->nodes[1]); 3767 3768 right = read_node_slot(fs_info, upper, slot + 1); 3769 /* 3770 * slot + 1 is not valid or we fail to read the right node, 3771 * no big deal, just return. 3772 */ 3773 if (IS_ERR(right)) 3774 return 1; 3775 3776 btrfs_tree_lock(right); 3777 btrfs_set_lock_blocking(right); 3778 3779 free_space = btrfs_leaf_free_space(fs_info, right); 3780 if (free_space < data_size) 3781 goto out_unlock; 3782 3783 /* cow and double check */ 3784 ret = btrfs_cow_block(trans, root, right, upper, 3785 slot + 1, &right); 3786 if (ret) 3787 goto out_unlock; 3788 3789 free_space = btrfs_leaf_free_space(fs_info, right); 3790 if (free_space < data_size) 3791 goto out_unlock; 3792 3793 left_nritems = btrfs_header_nritems(left); 3794 if (left_nritems == 0) 3795 goto out_unlock; 3796 3797 if (path->slots[0] == left_nritems && !empty) { 3798 /* Key greater than all keys in the leaf, right neighbor has 3799 * enough room for it and we're not emptying our leaf to delete 3800 * it, therefore use right neighbor to insert the new item and 3801 * no need to touch/dirty our left leaft. */ 3802 btrfs_tree_unlock(left); 3803 free_extent_buffer(left); 3804 path->nodes[0] = right; 3805 path->slots[0] = 0; 3806 path->slots[1]++; 3807 return 0; 3808 } 3809 3810 return __push_leaf_right(fs_info, path, min_data_size, empty, 3811 right, free_space, left_nritems, min_slot); 3812 out_unlock: 3813 btrfs_tree_unlock(right); 3814 free_extent_buffer(right); 3815 return 1; 3816 } 3817 3818 /* 3819 * push some data in the path leaf to the left, trying to free up at 3820 * least data_size bytes. returns zero if the push worked, nonzero otherwise 3821 * 3822 * max_slot can put a limit on how far into the leaf we'll push items. The 3823 * item at 'max_slot' won't be touched. Use (u32)-1 to make us do all the 3824 * items 3825 */ 3826 static noinline int __push_leaf_left(struct btrfs_fs_info *fs_info, 3827 struct btrfs_path *path, int data_size, 3828 int empty, struct extent_buffer *left, 3829 int free_space, u32 right_nritems, 3830 u32 max_slot) 3831 { 3832 struct btrfs_disk_key disk_key; 3833 struct extent_buffer *right = path->nodes[0]; 3834 int i; 3835 int push_space = 0; 3836 int push_items = 0; 3837 struct btrfs_item *item; 3838 u32 old_left_nritems; 3839 u32 nr; 3840 int ret = 0; 3841 u32 this_item_size; 3842 u32 old_left_item_size; 3843 struct btrfs_map_token token; 3844 3845 btrfs_init_map_token(&token); 3846 3847 if (empty) 3848 nr = min(right_nritems, max_slot); 3849 else 3850 nr = min(right_nritems - 1, max_slot); 3851 3852 for (i = 0; i < nr; i++) { 3853 item = btrfs_item_nr(i); 3854 3855 if (!empty && push_items > 0) { 3856 if (path->slots[0] < i) 3857 break; 3858 if (path->slots[0] == i) { 3859 int space = btrfs_leaf_free_space(fs_info, right); 3860 if (space + push_space * 2 > free_space) 3861 break; 3862 } 3863 } 3864 3865 if (path->slots[0] == i) 3866 push_space += data_size; 3867 3868 this_item_size = btrfs_item_size(right, item); 3869 if (this_item_size + sizeof(*item) + push_space > free_space) 3870 break; 3871 3872 push_items++; 3873 push_space += this_item_size + sizeof(*item); 3874 } 3875 3876 if (push_items == 0) { 3877 ret = 1; 3878 goto out; 3879 } 3880 WARN_ON(!empty && push_items == btrfs_header_nritems(right)); 3881 3882 /* push data from right to left */ 3883 copy_extent_buffer(left, right, 3884 btrfs_item_nr_offset(btrfs_header_nritems(left)), 3885 btrfs_item_nr_offset(0), 3886 push_items * sizeof(struct btrfs_item)); 3887 3888 push_space = BTRFS_LEAF_DATA_SIZE(fs_info) - 3889 btrfs_item_offset_nr(right, push_items - 1); 3890 3891 copy_extent_buffer(left, right, BTRFS_LEAF_DATA_OFFSET + 3892 leaf_data_end(fs_info, left) - push_space, 3893 BTRFS_LEAF_DATA_OFFSET + 3894 btrfs_item_offset_nr(right, push_items - 1), 3895 push_space); 3896 old_left_nritems = btrfs_header_nritems(left); 3897 BUG_ON(old_left_nritems <= 0); 3898 3899 old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1); 3900 for (i = old_left_nritems; i < old_left_nritems + push_items; i++) { 3901 u32 ioff; 3902 3903 item = btrfs_item_nr(i); 3904 3905 ioff = btrfs_token_item_offset(left, item, &token); 3906 btrfs_set_token_item_offset(left, item, 3907 ioff - (BTRFS_LEAF_DATA_SIZE(fs_info) - old_left_item_size), 3908 &token); 3909 } 3910 btrfs_set_header_nritems(left, old_left_nritems + push_items); 3911 3912 /* fixup right node */ 3913 if (push_items > right_nritems) 3914 WARN(1, KERN_CRIT "push items %d nr %u\n", push_items, 3915 right_nritems); 3916 3917 if (push_items < right_nritems) { 3918 push_space = btrfs_item_offset_nr(right, push_items - 1) - 3919 leaf_data_end(fs_info, right); 3920 memmove_extent_buffer(right, BTRFS_LEAF_DATA_OFFSET + 3921 BTRFS_LEAF_DATA_SIZE(fs_info) - push_space, 3922 BTRFS_LEAF_DATA_OFFSET + 3923 leaf_data_end(fs_info, right), push_space); 3924 3925 memmove_extent_buffer(right, btrfs_item_nr_offset(0), 3926 btrfs_item_nr_offset(push_items), 3927 (btrfs_header_nritems(right) - push_items) * 3928 sizeof(struct btrfs_item)); 3929 } 3930 right_nritems -= push_items; 3931 btrfs_set_header_nritems(right, right_nritems); 3932 push_space = BTRFS_LEAF_DATA_SIZE(fs_info); 3933 for (i = 0; i < right_nritems; i++) { 3934 item = btrfs_item_nr(i); 3935 3936 push_space = push_space - btrfs_token_item_size(right, 3937 item, &token); 3938 btrfs_set_token_item_offset(right, item, push_space, &token); 3939 } 3940 3941 btrfs_mark_buffer_dirty(left); 3942 if (right_nritems) 3943 btrfs_mark_buffer_dirty(right); 3944 else 3945 clean_tree_block(fs_info, right); 3946 3947 btrfs_item_key(right, &disk_key, 0); 3948 fixup_low_keys(fs_info, path, &disk_key, 1); 3949 3950 /* then fixup the leaf pointer in the path */ 3951 if (path->slots[0] < push_items) { 3952 path->slots[0] += old_left_nritems; 3953 btrfs_tree_unlock(path->nodes[0]); 3954 free_extent_buffer(path->nodes[0]); 3955 path->nodes[0] = left; 3956 path->slots[1] -= 1; 3957 } else { 3958 btrfs_tree_unlock(left); 3959 free_extent_buffer(left); 3960 path->slots[0] -= push_items; 3961 } 3962 BUG_ON(path->slots[0] < 0); 3963 return ret; 3964 out: 3965 btrfs_tree_unlock(left); 3966 free_extent_buffer(left); 3967 return ret; 3968 } 3969 3970 /* 3971 * push some data in the path leaf to the left, trying to free up at 3972 * least data_size bytes. returns zero if the push worked, nonzero otherwise 3973 * 3974 * max_slot can put a limit on how far into the leaf we'll push items. The 3975 * item at 'max_slot' won't be touched. Use (u32)-1 to make us push all the 3976 * items 3977 */ 3978 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root 3979 *root, struct btrfs_path *path, int min_data_size, 3980 int data_size, int empty, u32 max_slot) 3981 { 3982 struct btrfs_fs_info *fs_info = root->fs_info; 3983 struct extent_buffer *right = path->nodes[0]; 3984 struct extent_buffer *left; 3985 int slot; 3986 int free_space; 3987 u32 right_nritems; 3988 int ret = 0; 3989 3990 slot = path->slots[1]; 3991 if (slot == 0) 3992 return 1; 3993 if (!path->nodes[1]) 3994 return 1; 3995 3996 right_nritems = btrfs_header_nritems(right); 3997 if (right_nritems == 0) 3998 return 1; 3999 4000 btrfs_assert_tree_locked(path->nodes[1]); 4001 4002 left = read_node_slot(fs_info, path->nodes[1], slot - 1); 4003 /* 4004 * slot - 1 is not valid or we fail to read the left node, 4005 * no big deal, just return. 4006 */ 4007 if (IS_ERR(left)) 4008 return 1; 4009 4010 btrfs_tree_lock(left); 4011 btrfs_set_lock_blocking(left); 4012 4013 free_space = btrfs_leaf_free_space(fs_info, left); 4014 if (free_space < data_size) { 4015 ret = 1; 4016 goto out; 4017 } 4018 4019 /* cow and double check */ 4020 ret = btrfs_cow_block(trans, root, left, 4021 path->nodes[1], slot - 1, &left); 4022 if (ret) { 4023 /* we hit -ENOSPC, but it isn't fatal here */ 4024 if (ret == -ENOSPC) 4025 ret = 1; 4026 goto out; 4027 } 4028 4029 free_space = btrfs_leaf_free_space(fs_info, left); 4030 if (free_space < data_size) { 4031 ret = 1; 4032 goto out; 4033 } 4034 4035 return __push_leaf_left(fs_info, path, min_data_size, 4036 empty, left, free_space, right_nritems, 4037 max_slot); 4038 out: 4039 btrfs_tree_unlock(left); 4040 free_extent_buffer(left); 4041 return ret; 4042 } 4043 4044 /* 4045 * split the path's leaf in two, making sure there is at least data_size 4046 * available for the resulting leaf level of the path. 4047 */ 4048 static noinline void copy_for_split(struct btrfs_trans_handle *trans, 4049 struct btrfs_fs_info *fs_info, 4050 struct btrfs_path *path, 4051 struct extent_buffer *l, 4052 struct extent_buffer *right, 4053 int slot, int mid, int nritems) 4054 { 4055 int data_copy_size; 4056 int rt_data_off; 4057 int i; 4058 struct btrfs_disk_key disk_key; 4059 struct btrfs_map_token token; 4060 4061 btrfs_init_map_token(&token); 4062 4063 nritems = nritems - mid; 4064 btrfs_set_header_nritems(right, nritems); 4065 data_copy_size = btrfs_item_end_nr(l, mid) - leaf_data_end(fs_info, l); 4066 4067 copy_extent_buffer(right, l, btrfs_item_nr_offset(0), 4068 btrfs_item_nr_offset(mid), 4069 nritems * sizeof(struct btrfs_item)); 4070 4071 copy_extent_buffer(right, l, 4072 BTRFS_LEAF_DATA_OFFSET + BTRFS_LEAF_DATA_SIZE(fs_info) - 4073 data_copy_size, BTRFS_LEAF_DATA_OFFSET + 4074 leaf_data_end(fs_info, l), data_copy_size); 4075 4076 rt_data_off = BTRFS_LEAF_DATA_SIZE(fs_info) - btrfs_item_end_nr(l, mid); 4077 4078 for (i = 0; i < nritems; i++) { 4079 struct btrfs_item *item = btrfs_item_nr(i); 4080 u32 ioff; 4081 4082 ioff = btrfs_token_item_offset(right, item, &token); 4083 btrfs_set_token_item_offset(right, item, 4084 ioff + rt_data_off, &token); 4085 } 4086 4087 btrfs_set_header_nritems(l, mid); 4088 btrfs_item_key(right, &disk_key, 0); 4089 insert_ptr(trans, fs_info, path, &disk_key, right->start, 4090 path->slots[1] + 1, 1); 4091 4092 btrfs_mark_buffer_dirty(right); 4093 btrfs_mark_buffer_dirty(l); 4094 BUG_ON(path->slots[0] != slot); 4095 4096 if (mid <= slot) { 4097 btrfs_tree_unlock(path->nodes[0]); 4098 free_extent_buffer(path->nodes[0]); 4099 path->nodes[0] = right; 4100 path->slots[0] -= mid; 4101 path->slots[1] += 1; 4102 } else { 4103 btrfs_tree_unlock(right); 4104 free_extent_buffer(right); 4105 } 4106 4107 BUG_ON(path->slots[0] < 0); 4108 } 4109 4110 /* 4111 * double splits happen when we need to insert a big item in the middle 4112 * of a leaf. A double split can leave us with 3 mostly empty leaves: 4113 * leaf: [ slots 0 - N] [ our target ] [ N + 1 - total in leaf ] 4114 * A B C 4115 * 4116 * We avoid this by trying to push the items on either side of our target 4117 * into the adjacent leaves. If all goes well we can avoid the double split 4118 * completely. 4119 */ 4120 static noinline int push_for_double_split(struct btrfs_trans_handle *trans, 4121 struct btrfs_root *root, 4122 struct btrfs_path *path, 4123 int data_size) 4124 { 4125 struct btrfs_fs_info *fs_info = root->fs_info; 4126 int ret; 4127 int progress = 0; 4128 int slot; 4129 u32 nritems; 4130 int space_needed = data_size; 4131 4132 slot = path->slots[0]; 4133 if (slot < btrfs_header_nritems(path->nodes[0])) 4134 space_needed -= btrfs_leaf_free_space(fs_info, path->nodes[0]); 4135 4136 /* 4137 * try to push all the items after our slot into the 4138 * right leaf 4139 */ 4140 ret = push_leaf_right(trans, root, path, 1, space_needed, 0, slot); 4141 if (ret < 0) 4142 return ret; 4143 4144 if (ret == 0) 4145 progress++; 4146 4147 nritems = btrfs_header_nritems(path->nodes[0]); 4148 /* 4149 * our goal is to get our slot at the start or end of a leaf. If 4150 * we've done so we're done 4151 */ 4152 if (path->slots[0] == 0 || path->slots[0] == nritems) 4153 return 0; 4154 4155 if (btrfs_leaf_free_space(fs_info, path->nodes[0]) >= data_size) 4156 return 0; 4157 4158 /* try to push all the items before our slot into the next leaf */ 4159 slot = path->slots[0]; 4160 space_needed = data_size; 4161 if (slot > 0) 4162 space_needed -= btrfs_leaf_free_space(fs_info, path->nodes[0]); 4163 ret = push_leaf_left(trans, root, path, 1, space_needed, 0, slot); 4164 if (ret < 0) 4165 return ret; 4166 4167 if (ret == 0) 4168 progress++; 4169 4170 if (progress) 4171 return 0; 4172 return 1; 4173 } 4174 4175 /* 4176 * split the path's leaf in two, making sure there is at least data_size 4177 * available for the resulting leaf level of the path. 4178 * 4179 * returns 0 if all went well and < 0 on failure. 4180 */ 4181 static noinline int split_leaf(struct btrfs_trans_handle *trans, 4182 struct btrfs_root *root, 4183 const struct btrfs_key *ins_key, 4184 struct btrfs_path *path, int data_size, 4185 int extend) 4186 { 4187 struct btrfs_disk_key disk_key; 4188 struct extent_buffer *l; 4189 u32 nritems; 4190 int mid; 4191 int slot; 4192 struct extent_buffer *right; 4193 struct btrfs_fs_info *fs_info = root->fs_info; 4194 int ret = 0; 4195 int wret; 4196 int split; 4197 int num_doubles = 0; 4198 int tried_avoid_double = 0; 4199 4200 l = path->nodes[0]; 4201 slot = path->slots[0]; 4202 if (extend && data_size + btrfs_item_size_nr(l, slot) + 4203 sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(fs_info)) 4204 return -EOVERFLOW; 4205 4206 /* first try to make some room by pushing left and right */ 4207 if (data_size && path->nodes[1]) { 4208 int space_needed = data_size; 4209 4210 if (slot < btrfs_header_nritems(l)) 4211 space_needed -= btrfs_leaf_free_space(fs_info, l); 4212 4213 wret = push_leaf_right(trans, root, path, space_needed, 4214 space_needed, 0, 0); 4215 if (wret < 0) 4216 return wret; 4217 if (wret) { 4218 space_needed = data_size; 4219 if (slot > 0) 4220 space_needed -= btrfs_leaf_free_space(fs_info, 4221 l); 4222 wret = push_leaf_left(trans, root, path, space_needed, 4223 space_needed, 0, (u32)-1); 4224 if (wret < 0) 4225 return wret; 4226 } 4227 l = path->nodes[0]; 4228 4229 /* did the pushes work? */ 4230 if (btrfs_leaf_free_space(fs_info, l) >= data_size) 4231 return 0; 4232 } 4233 4234 if (!path->nodes[1]) { 4235 ret = insert_new_root(trans, root, path, 1); 4236 if (ret) 4237 return ret; 4238 } 4239 again: 4240 split = 1; 4241 l = path->nodes[0]; 4242 slot = path->slots[0]; 4243 nritems = btrfs_header_nritems(l); 4244 mid = (nritems + 1) / 2; 4245 4246 if (mid <= slot) { 4247 if (nritems == 1 || 4248 leaf_space_used(l, mid, nritems - mid) + data_size > 4249 BTRFS_LEAF_DATA_SIZE(fs_info)) { 4250 if (slot >= nritems) { 4251 split = 0; 4252 } else { 4253 mid = slot; 4254 if (mid != nritems && 4255 leaf_space_used(l, mid, nritems - mid) + 4256 data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) { 4257 if (data_size && !tried_avoid_double) 4258 goto push_for_double; 4259 split = 2; 4260 } 4261 } 4262 } 4263 } else { 4264 if (leaf_space_used(l, 0, mid) + data_size > 4265 BTRFS_LEAF_DATA_SIZE(fs_info)) { 4266 if (!extend && data_size && slot == 0) { 4267 split = 0; 4268 } else if ((extend || !data_size) && slot == 0) { 4269 mid = 1; 4270 } else { 4271 mid = slot; 4272 if (mid != nritems && 4273 leaf_space_used(l, mid, nritems - mid) + 4274 data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) { 4275 if (data_size && !tried_avoid_double) 4276 goto push_for_double; 4277 split = 2; 4278 } 4279 } 4280 } 4281 } 4282 4283 if (split == 0) 4284 btrfs_cpu_key_to_disk(&disk_key, ins_key); 4285 else 4286 btrfs_item_key(l, &disk_key, mid); 4287 4288 right = btrfs_alloc_tree_block(trans, root, 0, root->root_key.objectid, 4289 &disk_key, 0, l->start, 0); 4290 if (IS_ERR(right)) 4291 return PTR_ERR(right); 4292 4293 root_add_used(root, fs_info->nodesize); 4294 4295 memzero_extent_buffer(right, 0, sizeof(struct btrfs_header)); 4296 btrfs_set_header_bytenr(right, right->start); 4297 btrfs_set_header_generation(right, trans->transid); 4298 btrfs_set_header_backref_rev(right, BTRFS_MIXED_BACKREF_REV); 4299 btrfs_set_header_owner(right, root->root_key.objectid); 4300 btrfs_set_header_level(right, 0); 4301 write_extent_buffer_fsid(right, fs_info->fsid); 4302 write_extent_buffer_chunk_tree_uuid(right, fs_info->chunk_tree_uuid); 4303 4304 if (split == 0) { 4305 if (mid <= slot) { 4306 btrfs_set_header_nritems(right, 0); 4307 insert_ptr(trans, fs_info, path, &disk_key, 4308 right->start, path->slots[1] + 1, 1); 4309 btrfs_tree_unlock(path->nodes[0]); 4310 free_extent_buffer(path->nodes[0]); 4311 path->nodes[0] = right; 4312 path->slots[0] = 0; 4313 path->slots[1] += 1; 4314 } else { 4315 btrfs_set_header_nritems(right, 0); 4316 insert_ptr(trans, fs_info, path, &disk_key, 4317 right->start, path->slots[1], 1); 4318 btrfs_tree_unlock(path->nodes[0]); 4319 free_extent_buffer(path->nodes[0]); 4320 path->nodes[0] = right; 4321 path->slots[0] = 0; 4322 if (path->slots[1] == 0) 4323 fixup_low_keys(fs_info, path, &disk_key, 1); 4324 } 4325 /* 4326 * We create a new leaf 'right' for the required ins_len and 4327 * we'll do btrfs_mark_buffer_dirty() on this leaf after copying 4328 * the content of ins_len to 'right'. 4329 */ 4330 return ret; 4331 } 4332 4333 copy_for_split(trans, fs_info, path, l, right, slot, mid, nritems); 4334 4335 if (split == 2) { 4336 BUG_ON(num_doubles != 0); 4337 num_doubles++; 4338 goto again; 4339 } 4340 4341 return 0; 4342 4343 push_for_double: 4344 push_for_double_split(trans, root, path, data_size); 4345 tried_avoid_double = 1; 4346 if (btrfs_leaf_free_space(fs_info, path->nodes[0]) >= data_size) 4347 return 0; 4348 goto again; 4349 } 4350 4351 static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans, 4352 struct btrfs_root *root, 4353 struct btrfs_path *path, int ins_len) 4354 { 4355 struct btrfs_fs_info *fs_info = root->fs_info; 4356 struct btrfs_key key; 4357 struct extent_buffer *leaf; 4358 struct btrfs_file_extent_item *fi; 4359 u64 extent_len = 0; 4360 u32 item_size; 4361 int ret; 4362 4363 leaf = path->nodes[0]; 4364 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 4365 4366 BUG_ON(key.type != BTRFS_EXTENT_DATA_KEY && 4367 key.type != BTRFS_EXTENT_CSUM_KEY); 4368 4369 if (btrfs_leaf_free_space(fs_info, leaf) >= ins_len) 4370 return 0; 4371 4372 item_size = btrfs_item_size_nr(leaf, path->slots[0]); 4373 if (key.type == BTRFS_EXTENT_DATA_KEY) { 4374 fi = btrfs_item_ptr(leaf, path->slots[0], 4375 struct btrfs_file_extent_item); 4376 extent_len = btrfs_file_extent_num_bytes(leaf, fi); 4377 } 4378 btrfs_release_path(path); 4379 4380 path->keep_locks = 1; 4381 path->search_for_split = 1; 4382 ret = btrfs_search_slot(trans, root, &key, path, 0, 1); 4383 path->search_for_split = 0; 4384 if (ret > 0) 4385 ret = -EAGAIN; 4386 if (ret < 0) 4387 goto err; 4388 4389 ret = -EAGAIN; 4390 leaf = path->nodes[0]; 4391 /* if our item isn't there, return now */ 4392 if (item_size != btrfs_item_size_nr(leaf, path->slots[0])) 4393 goto err; 4394 4395 /* the leaf has changed, it now has room. return now */ 4396 if (btrfs_leaf_free_space(fs_info, path->nodes[0]) >= ins_len) 4397 goto err; 4398 4399 if (key.type == BTRFS_EXTENT_DATA_KEY) { 4400 fi = btrfs_item_ptr(leaf, path->slots[0], 4401 struct btrfs_file_extent_item); 4402 if (extent_len != btrfs_file_extent_num_bytes(leaf, fi)) 4403 goto err; 4404 } 4405 4406 btrfs_set_path_blocking(path); 4407 ret = split_leaf(trans, root, &key, path, ins_len, 1); 4408 if (ret) 4409 goto err; 4410 4411 path->keep_locks = 0; 4412 btrfs_unlock_up_safe(path, 1); 4413 return 0; 4414 err: 4415 path->keep_locks = 0; 4416 return ret; 4417 } 4418 4419 static noinline int split_item(struct btrfs_fs_info *fs_info, 4420 struct btrfs_path *path, 4421 const struct btrfs_key *new_key, 4422 unsigned long split_offset) 4423 { 4424 struct extent_buffer *leaf; 4425 struct btrfs_item *item; 4426 struct btrfs_item *new_item; 4427 int slot; 4428 char *buf; 4429 u32 nritems; 4430 u32 item_size; 4431 u32 orig_offset; 4432 struct btrfs_disk_key disk_key; 4433 4434 leaf = path->nodes[0]; 4435 BUG_ON(btrfs_leaf_free_space(fs_info, leaf) < sizeof(struct btrfs_item)); 4436 4437 btrfs_set_path_blocking(path); 4438 4439 item = btrfs_item_nr(path->slots[0]); 4440 orig_offset = btrfs_item_offset(leaf, item); 4441 item_size = btrfs_item_size(leaf, item); 4442 4443 buf = kmalloc(item_size, GFP_NOFS); 4444 if (!buf) 4445 return -ENOMEM; 4446 4447 read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf, 4448 path->slots[0]), item_size); 4449 4450 slot = path->slots[0] + 1; 4451 nritems = btrfs_header_nritems(leaf); 4452 if (slot != nritems) { 4453 /* shift the items */ 4454 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1), 4455 btrfs_item_nr_offset(slot), 4456 (nritems - slot) * sizeof(struct btrfs_item)); 4457 } 4458 4459 btrfs_cpu_key_to_disk(&disk_key, new_key); 4460 btrfs_set_item_key(leaf, &disk_key, slot); 4461 4462 new_item = btrfs_item_nr(slot); 4463 4464 btrfs_set_item_offset(leaf, new_item, orig_offset); 4465 btrfs_set_item_size(leaf, new_item, item_size - split_offset); 4466 4467 btrfs_set_item_offset(leaf, item, 4468 orig_offset + item_size - split_offset); 4469 btrfs_set_item_size(leaf, item, split_offset); 4470 4471 btrfs_set_header_nritems(leaf, nritems + 1); 4472 4473 /* write the data for the start of the original item */ 4474 write_extent_buffer(leaf, buf, 4475 btrfs_item_ptr_offset(leaf, path->slots[0]), 4476 split_offset); 4477 4478 /* write the data for the new item */ 4479 write_extent_buffer(leaf, buf + split_offset, 4480 btrfs_item_ptr_offset(leaf, slot), 4481 item_size - split_offset); 4482 btrfs_mark_buffer_dirty(leaf); 4483 4484 BUG_ON(btrfs_leaf_free_space(fs_info, leaf) < 0); 4485 kfree(buf); 4486 return 0; 4487 } 4488 4489 /* 4490 * This function splits a single item into two items, 4491 * giving 'new_key' to the new item and splitting the 4492 * old one at split_offset (from the start of the item). 4493 * 4494 * The path may be released by this operation. After 4495 * the split, the path is pointing to the old item. The 4496 * new item is going to be in the same node as the old one. 4497 * 4498 * Note, the item being split must be smaller enough to live alone on 4499 * a tree block with room for one extra struct btrfs_item 4500 * 4501 * This allows us to split the item in place, keeping a lock on the 4502 * leaf the entire time. 4503 */ 4504 int btrfs_split_item(struct btrfs_trans_handle *trans, 4505 struct btrfs_root *root, 4506 struct btrfs_path *path, 4507 const struct btrfs_key *new_key, 4508 unsigned long split_offset) 4509 { 4510 int ret; 4511 ret = setup_leaf_for_split(trans, root, path, 4512 sizeof(struct btrfs_item)); 4513 if (ret) 4514 return ret; 4515 4516 ret = split_item(root->fs_info, path, new_key, split_offset); 4517 return ret; 4518 } 4519 4520 /* 4521 * This function duplicate a item, giving 'new_key' to the new item. 4522 * It guarantees both items live in the same tree leaf and the new item 4523 * is contiguous with the original item. 4524 * 4525 * This allows us to split file extent in place, keeping a lock on the 4526 * leaf the entire time. 4527 */ 4528 int btrfs_duplicate_item(struct btrfs_trans_handle *trans, 4529 struct btrfs_root *root, 4530 struct btrfs_path *path, 4531 const struct btrfs_key *new_key) 4532 { 4533 struct extent_buffer *leaf; 4534 int ret; 4535 u32 item_size; 4536 4537 leaf = path->nodes[0]; 4538 item_size = btrfs_item_size_nr(leaf, path->slots[0]); 4539 ret = setup_leaf_for_split(trans, root, path, 4540 item_size + sizeof(struct btrfs_item)); 4541 if (ret) 4542 return ret; 4543 4544 path->slots[0]++; 4545 setup_items_for_insert(root, path, new_key, &item_size, 4546 item_size, item_size + 4547 sizeof(struct btrfs_item), 1); 4548 leaf = path->nodes[0]; 4549 memcpy_extent_buffer(leaf, 4550 btrfs_item_ptr_offset(leaf, path->slots[0]), 4551 btrfs_item_ptr_offset(leaf, path->slots[0] - 1), 4552 item_size); 4553 return 0; 4554 } 4555 4556 /* 4557 * make the item pointed to by the path smaller. new_size indicates 4558 * how small to make it, and from_end tells us if we just chop bytes 4559 * off the end of the item or if we shift the item to chop bytes off 4560 * the front. 4561 */ 4562 void btrfs_truncate_item(struct btrfs_fs_info *fs_info, 4563 struct btrfs_path *path, u32 new_size, int from_end) 4564 { 4565 int slot; 4566 struct extent_buffer *leaf; 4567 struct btrfs_item *item; 4568 u32 nritems; 4569 unsigned int data_end; 4570 unsigned int old_data_start; 4571 unsigned int old_size; 4572 unsigned int size_diff; 4573 int i; 4574 struct btrfs_map_token token; 4575 4576 btrfs_init_map_token(&token); 4577 4578 leaf = path->nodes[0]; 4579 slot = path->slots[0]; 4580 4581 old_size = btrfs_item_size_nr(leaf, slot); 4582 if (old_size == new_size) 4583 return; 4584 4585 nritems = btrfs_header_nritems(leaf); 4586 data_end = leaf_data_end(fs_info, leaf); 4587 4588 old_data_start = btrfs_item_offset_nr(leaf, slot); 4589 4590 size_diff = old_size - new_size; 4591 4592 BUG_ON(slot < 0); 4593 BUG_ON(slot >= nritems); 4594 4595 /* 4596 * item0..itemN ... dataN.offset..dataN.size .. data0.size 4597 */ 4598 /* first correct the data pointers */ 4599 for (i = slot; i < nritems; i++) { 4600 u32 ioff; 4601 item = btrfs_item_nr(i); 4602 4603 ioff = btrfs_token_item_offset(leaf, item, &token); 4604 btrfs_set_token_item_offset(leaf, item, 4605 ioff + size_diff, &token); 4606 } 4607 4608 /* shift the data */ 4609 if (from_end) { 4610 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4611 data_end + size_diff, BTRFS_LEAF_DATA_OFFSET + 4612 data_end, old_data_start + new_size - data_end); 4613 } else { 4614 struct btrfs_disk_key disk_key; 4615 u64 offset; 4616 4617 btrfs_item_key(leaf, &disk_key, slot); 4618 4619 if (btrfs_disk_key_type(&disk_key) == BTRFS_EXTENT_DATA_KEY) { 4620 unsigned long ptr; 4621 struct btrfs_file_extent_item *fi; 4622 4623 fi = btrfs_item_ptr(leaf, slot, 4624 struct btrfs_file_extent_item); 4625 fi = (struct btrfs_file_extent_item *)( 4626 (unsigned long)fi - size_diff); 4627 4628 if (btrfs_file_extent_type(leaf, fi) == 4629 BTRFS_FILE_EXTENT_INLINE) { 4630 ptr = btrfs_item_ptr_offset(leaf, slot); 4631 memmove_extent_buffer(leaf, ptr, 4632 (unsigned long)fi, 4633 BTRFS_FILE_EXTENT_INLINE_DATA_START); 4634 } 4635 } 4636 4637 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4638 data_end + size_diff, BTRFS_LEAF_DATA_OFFSET + 4639 data_end, old_data_start - data_end); 4640 4641 offset = btrfs_disk_key_offset(&disk_key); 4642 btrfs_set_disk_key_offset(&disk_key, offset + size_diff); 4643 btrfs_set_item_key(leaf, &disk_key, slot); 4644 if (slot == 0) 4645 fixup_low_keys(fs_info, path, &disk_key, 1); 4646 } 4647 4648 item = btrfs_item_nr(slot); 4649 btrfs_set_item_size(leaf, item, new_size); 4650 btrfs_mark_buffer_dirty(leaf); 4651 4652 if (btrfs_leaf_free_space(fs_info, leaf) < 0) { 4653 btrfs_print_leaf(leaf); 4654 BUG(); 4655 } 4656 } 4657 4658 /* 4659 * make the item pointed to by the path bigger, data_size is the added size. 4660 */ 4661 void btrfs_extend_item(struct btrfs_fs_info *fs_info, struct btrfs_path *path, 4662 u32 data_size) 4663 { 4664 int slot; 4665 struct extent_buffer *leaf; 4666 struct btrfs_item *item; 4667 u32 nritems; 4668 unsigned int data_end; 4669 unsigned int old_data; 4670 unsigned int old_size; 4671 int i; 4672 struct btrfs_map_token token; 4673 4674 btrfs_init_map_token(&token); 4675 4676 leaf = path->nodes[0]; 4677 4678 nritems = btrfs_header_nritems(leaf); 4679 data_end = leaf_data_end(fs_info, leaf); 4680 4681 if (btrfs_leaf_free_space(fs_info, leaf) < data_size) { 4682 btrfs_print_leaf(leaf); 4683 BUG(); 4684 } 4685 slot = path->slots[0]; 4686 old_data = btrfs_item_end_nr(leaf, slot); 4687 4688 BUG_ON(slot < 0); 4689 if (slot >= nritems) { 4690 btrfs_print_leaf(leaf); 4691 btrfs_crit(fs_info, "slot %d too large, nritems %d", 4692 slot, nritems); 4693 BUG_ON(1); 4694 } 4695 4696 /* 4697 * item0..itemN ... dataN.offset..dataN.size .. data0.size 4698 */ 4699 /* first correct the data pointers */ 4700 for (i = slot; i < nritems; i++) { 4701 u32 ioff; 4702 item = btrfs_item_nr(i); 4703 4704 ioff = btrfs_token_item_offset(leaf, item, &token); 4705 btrfs_set_token_item_offset(leaf, item, 4706 ioff - data_size, &token); 4707 } 4708 4709 /* shift the data */ 4710 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4711 data_end - data_size, BTRFS_LEAF_DATA_OFFSET + 4712 data_end, old_data - data_end); 4713 4714 data_end = old_data; 4715 old_size = btrfs_item_size_nr(leaf, slot); 4716 item = btrfs_item_nr(slot); 4717 btrfs_set_item_size(leaf, item, old_size + data_size); 4718 btrfs_mark_buffer_dirty(leaf); 4719 4720 if (btrfs_leaf_free_space(fs_info, leaf) < 0) { 4721 btrfs_print_leaf(leaf); 4722 BUG(); 4723 } 4724 } 4725 4726 /* 4727 * this is a helper for btrfs_insert_empty_items, the main goal here is 4728 * to save stack depth by doing the bulk of the work in a function 4729 * that doesn't call btrfs_search_slot 4730 */ 4731 void setup_items_for_insert(struct btrfs_root *root, struct btrfs_path *path, 4732 const struct btrfs_key *cpu_key, u32 *data_size, 4733 u32 total_data, u32 total_size, int nr) 4734 { 4735 struct btrfs_fs_info *fs_info = root->fs_info; 4736 struct btrfs_item *item; 4737 int i; 4738 u32 nritems; 4739 unsigned int data_end; 4740 struct btrfs_disk_key disk_key; 4741 struct extent_buffer *leaf; 4742 int slot; 4743 struct btrfs_map_token token; 4744 4745 if (path->slots[0] == 0) { 4746 btrfs_cpu_key_to_disk(&disk_key, cpu_key); 4747 fixup_low_keys(fs_info, path, &disk_key, 1); 4748 } 4749 btrfs_unlock_up_safe(path, 1); 4750 4751 btrfs_init_map_token(&token); 4752 4753 leaf = path->nodes[0]; 4754 slot = path->slots[0]; 4755 4756 nritems = btrfs_header_nritems(leaf); 4757 data_end = leaf_data_end(fs_info, leaf); 4758 4759 if (btrfs_leaf_free_space(fs_info, leaf) < total_size) { 4760 btrfs_print_leaf(leaf); 4761 btrfs_crit(fs_info, "not enough freespace need %u have %d", 4762 total_size, btrfs_leaf_free_space(fs_info, leaf)); 4763 BUG(); 4764 } 4765 4766 if (slot != nritems) { 4767 unsigned int old_data = btrfs_item_end_nr(leaf, slot); 4768 4769 if (old_data < data_end) { 4770 btrfs_print_leaf(leaf); 4771 btrfs_crit(fs_info, "slot %d old_data %d data_end %d", 4772 slot, old_data, data_end); 4773 BUG_ON(1); 4774 } 4775 /* 4776 * item0..itemN ... dataN.offset..dataN.size .. data0.size 4777 */ 4778 /* first correct the data pointers */ 4779 for (i = slot; i < nritems; i++) { 4780 u32 ioff; 4781 4782 item = btrfs_item_nr(i); 4783 ioff = btrfs_token_item_offset(leaf, item, &token); 4784 btrfs_set_token_item_offset(leaf, item, 4785 ioff - total_data, &token); 4786 } 4787 /* shift the items */ 4788 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr), 4789 btrfs_item_nr_offset(slot), 4790 (nritems - slot) * sizeof(struct btrfs_item)); 4791 4792 /* shift the data */ 4793 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4794 data_end - total_data, BTRFS_LEAF_DATA_OFFSET + 4795 data_end, old_data - data_end); 4796 data_end = old_data; 4797 } 4798 4799 /* setup the item for the new data */ 4800 for (i = 0; i < nr; i++) { 4801 btrfs_cpu_key_to_disk(&disk_key, cpu_key + i); 4802 btrfs_set_item_key(leaf, &disk_key, slot + i); 4803 item = btrfs_item_nr(slot + i); 4804 btrfs_set_token_item_offset(leaf, item, 4805 data_end - data_size[i], &token); 4806 data_end -= data_size[i]; 4807 btrfs_set_token_item_size(leaf, item, data_size[i], &token); 4808 } 4809 4810 btrfs_set_header_nritems(leaf, nritems + nr); 4811 btrfs_mark_buffer_dirty(leaf); 4812 4813 if (btrfs_leaf_free_space(fs_info, leaf) < 0) { 4814 btrfs_print_leaf(leaf); 4815 BUG(); 4816 } 4817 } 4818 4819 /* 4820 * Given a key and some data, insert items into the tree. 4821 * This does all the path init required, making room in the tree if needed. 4822 */ 4823 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans, 4824 struct btrfs_root *root, 4825 struct btrfs_path *path, 4826 const struct btrfs_key *cpu_key, u32 *data_size, 4827 int nr) 4828 { 4829 int ret = 0; 4830 int slot; 4831 int i; 4832 u32 total_size = 0; 4833 u32 total_data = 0; 4834 4835 for (i = 0; i < nr; i++) 4836 total_data += data_size[i]; 4837 4838 total_size = total_data + (nr * sizeof(struct btrfs_item)); 4839 ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1); 4840 if (ret == 0) 4841 return -EEXIST; 4842 if (ret < 0) 4843 return ret; 4844 4845 slot = path->slots[0]; 4846 BUG_ON(slot < 0); 4847 4848 setup_items_for_insert(root, path, cpu_key, data_size, 4849 total_data, total_size, nr); 4850 return 0; 4851 } 4852 4853 /* 4854 * Given a key and some data, insert an item into the tree. 4855 * This does all the path init required, making room in the tree if needed. 4856 */ 4857 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root, 4858 const struct btrfs_key *cpu_key, void *data, 4859 u32 data_size) 4860 { 4861 int ret = 0; 4862 struct btrfs_path *path; 4863 struct extent_buffer *leaf; 4864 unsigned long ptr; 4865 4866 path = btrfs_alloc_path(); 4867 if (!path) 4868 return -ENOMEM; 4869 ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size); 4870 if (!ret) { 4871 leaf = path->nodes[0]; 4872 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]); 4873 write_extent_buffer(leaf, data, ptr, data_size); 4874 btrfs_mark_buffer_dirty(leaf); 4875 } 4876 btrfs_free_path(path); 4877 return ret; 4878 } 4879 4880 /* 4881 * delete the pointer from a given node. 4882 * 4883 * the tree should have been previously balanced so the deletion does not 4884 * empty a node. 4885 */ 4886 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path, 4887 int level, int slot) 4888 { 4889 struct btrfs_fs_info *fs_info = root->fs_info; 4890 struct extent_buffer *parent = path->nodes[level]; 4891 u32 nritems; 4892 int ret; 4893 4894 nritems = btrfs_header_nritems(parent); 4895 if (slot != nritems - 1) { 4896 if (level) 4897 tree_mod_log_eb_move(fs_info, parent, slot, 4898 slot + 1, nritems - slot - 1); 4899 memmove_extent_buffer(parent, 4900 btrfs_node_key_ptr_offset(slot), 4901 btrfs_node_key_ptr_offset(slot + 1), 4902 sizeof(struct btrfs_key_ptr) * 4903 (nritems - slot - 1)); 4904 } else if (level) { 4905 ret = tree_mod_log_insert_key(fs_info, parent, slot, 4906 MOD_LOG_KEY_REMOVE, GFP_NOFS); 4907 BUG_ON(ret < 0); 4908 } 4909 4910 nritems--; 4911 btrfs_set_header_nritems(parent, nritems); 4912 if (nritems == 0 && parent == root->node) { 4913 BUG_ON(btrfs_header_level(root->node) != 1); 4914 /* just turn the root into a leaf and break */ 4915 btrfs_set_header_level(root->node, 0); 4916 } else if (slot == 0) { 4917 struct btrfs_disk_key disk_key; 4918 4919 btrfs_node_key(parent, &disk_key, 0); 4920 fixup_low_keys(fs_info, path, &disk_key, level + 1); 4921 } 4922 btrfs_mark_buffer_dirty(parent); 4923 } 4924 4925 /* 4926 * a helper function to delete the leaf pointed to by path->slots[1] and 4927 * path->nodes[1]. 4928 * 4929 * This deletes the pointer in path->nodes[1] and frees the leaf 4930 * block extent. zero is returned if it all worked out, < 0 otherwise. 4931 * 4932 * The path must have already been setup for deleting the leaf, including 4933 * all the proper balancing. path->nodes[1] must be locked. 4934 */ 4935 static noinline void btrfs_del_leaf(struct btrfs_trans_handle *trans, 4936 struct btrfs_root *root, 4937 struct btrfs_path *path, 4938 struct extent_buffer *leaf) 4939 { 4940 WARN_ON(btrfs_header_generation(leaf) != trans->transid); 4941 del_ptr(root, path, 1, path->slots[1]); 4942 4943 /* 4944 * btrfs_free_extent is expensive, we want to make sure we 4945 * aren't holding any locks when we call it 4946 */ 4947 btrfs_unlock_up_safe(path, 0); 4948 4949 root_sub_used(root, leaf->len); 4950 4951 extent_buffer_get(leaf); 4952 btrfs_free_tree_block(trans, root, leaf, 0, 1); 4953 free_extent_buffer_stale(leaf); 4954 } 4955 /* 4956 * delete the item at the leaf level in path. If that empties 4957 * the leaf, remove it from the tree 4958 */ 4959 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, 4960 struct btrfs_path *path, int slot, int nr) 4961 { 4962 struct btrfs_fs_info *fs_info = root->fs_info; 4963 struct extent_buffer *leaf; 4964 struct btrfs_item *item; 4965 u32 last_off; 4966 u32 dsize = 0; 4967 int ret = 0; 4968 int wret; 4969 int i; 4970 u32 nritems; 4971 struct btrfs_map_token token; 4972 4973 btrfs_init_map_token(&token); 4974 4975 leaf = path->nodes[0]; 4976 last_off = btrfs_item_offset_nr(leaf, slot + nr - 1); 4977 4978 for (i = 0; i < nr; i++) 4979 dsize += btrfs_item_size_nr(leaf, slot + i); 4980 4981 nritems = btrfs_header_nritems(leaf); 4982 4983 if (slot + nr != nritems) { 4984 int data_end = leaf_data_end(fs_info, leaf); 4985 4986 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET + 4987 data_end + dsize, 4988 BTRFS_LEAF_DATA_OFFSET + data_end, 4989 last_off - data_end); 4990 4991 for (i = slot + nr; i < nritems; i++) { 4992 u32 ioff; 4993 4994 item = btrfs_item_nr(i); 4995 ioff = btrfs_token_item_offset(leaf, item, &token); 4996 btrfs_set_token_item_offset(leaf, item, 4997 ioff + dsize, &token); 4998 } 4999 5000 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot), 5001 btrfs_item_nr_offset(slot + nr), 5002 sizeof(struct btrfs_item) * 5003 (nritems - slot - nr)); 5004 } 5005 btrfs_set_header_nritems(leaf, nritems - nr); 5006 nritems -= nr; 5007 5008 /* delete the leaf if we've emptied it */ 5009 if (nritems == 0) { 5010 if (leaf == root->node) { 5011 btrfs_set_header_level(leaf, 0); 5012 } else { 5013 btrfs_set_path_blocking(path); 5014 clean_tree_block(fs_info, leaf); 5015 btrfs_del_leaf(trans, root, path, leaf); 5016 } 5017 } else { 5018 int used = leaf_space_used(leaf, 0, nritems); 5019 if (slot == 0) { 5020 struct btrfs_disk_key disk_key; 5021 5022 btrfs_item_key(leaf, &disk_key, 0); 5023 fixup_low_keys(fs_info, path, &disk_key, 1); 5024 } 5025 5026 /* delete the leaf if it is mostly empty */ 5027 if (used < BTRFS_LEAF_DATA_SIZE(fs_info) / 3) { 5028 /* push_leaf_left fixes the path. 5029 * make sure the path still points to our leaf 5030 * for possible call to del_ptr below 5031 */ 5032 slot = path->slots[1]; 5033 extent_buffer_get(leaf); 5034 5035 btrfs_set_path_blocking(path); 5036 wret = push_leaf_left(trans, root, path, 1, 1, 5037 1, (u32)-1); 5038 if (wret < 0 && wret != -ENOSPC) 5039 ret = wret; 5040 5041 if (path->nodes[0] == leaf && 5042 btrfs_header_nritems(leaf)) { 5043 wret = push_leaf_right(trans, root, path, 1, 5044 1, 1, 0); 5045 if (wret < 0 && wret != -ENOSPC) 5046 ret = wret; 5047 } 5048 5049 if (btrfs_header_nritems(leaf) == 0) { 5050 path->slots[1] = slot; 5051 btrfs_del_leaf(trans, root, path, leaf); 5052 free_extent_buffer(leaf); 5053 ret = 0; 5054 } else { 5055 /* if we're still in the path, make sure 5056 * we're dirty. Otherwise, one of the 5057 * push_leaf functions must have already 5058 * dirtied this buffer 5059 */ 5060 if (path->nodes[0] == leaf) 5061 btrfs_mark_buffer_dirty(leaf); 5062 free_extent_buffer(leaf); 5063 } 5064 } else { 5065 btrfs_mark_buffer_dirty(leaf); 5066 } 5067 } 5068 return ret; 5069 } 5070 5071 /* 5072 * search the tree again to find a leaf with lesser keys 5073 * returns 0 if it found something or 1 if there are no lesser leaves. 5074 * returns < 0 on io errors. 5075 * 5076 * This may release the path, and so you may lose any locks held at the 5077 * time you call it. 5078 */ 5079 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path) 5080 { 5081 struct btrfs_key key; 5082 struct btrfs_disk_key found_key; 5083 int ret; 5084 5085 btrfs_item_key_to_cpu(path->nodes[0], &key, 0); 5086 5087 if (key.offset > 0) { 5088 key.offset--; 5089 } else if (key.type > 0) { 5090 key.type--; 5091 key.offset = (u64)-1; 5092 } else if (key.objectid > 0) { 5093 key.objectid--; 5094 key.type = (u8)-1; 5095 key.offset = (u64)-1; 5096 } else { 5097 return 1; 5098 } 5099 5100 btrfs_release_path(path); 5101 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5102 if (ret < 0) 5103 return ret; 5104 btrfs_item_key(path->nodes[0], &found_key, 0); 5105 ret = comp_keys(&found_key, &key); 5106 /* 5107 * We might have had an item with the previous key in the tree right 5108 * before we released our path. And after we released our path, that 5109 * item might have been pushed to the first slot (0) of the leaf we 5110 * were holding due to a tree balance. Alternatively, an item with the 5111 * previous key can exist as the only element of a leaf (big fat item). 5112 * Therefore account for these 2 cases, so that our callers (like 5113 * btrfs_previous_item) don't miss an existing item with a key matching 5114 * the previous key we computed above. 5115 */ 5116 if (ret <= 0) 5117 return 0; 5118 return 1; 5119 } 5120 5121 /* 5122 * A helper function to walk down the tree starting at min_key, and looking 5123 * for nodes or leaves that are have a minimum transaction id. 5124 * This is used by the btree defrag code, and tree logging 5125 * 5126 * This does not cow, but it does stuff the starting key it finds back 5127 * into min_key, so you can call btrfs_search_slot with cow=1 on the 5128 * key and get a writable path. 5129 * 5130 * This does lock as it descends, and path->keep_locks should be set 5131 * to 1 by the caller. 5132 * 5133 * This honors path->lowest_level to prevent descent past a given level 5134 * of the tree. 5135 * 5136 * min_trans indicates the oldest transaction that you are interested 5137 * in walking through. Any nodes or leaves older than min_trans are 5138 * skipped over (without reading them). 5139 * 5140 * returns zero if something useful was found, < 0 on error and 1 if there 5141 * was nothing in the tree that matched the search criteria. 5142 */ 5143 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key, 5144 struct btrfs_path *path, 5145 u64 min_trans) 5146 { 5147 struct btrfs_fs_info *fs_info = root->fs_info; 5148 struct extent_buffer *cur; 5149 struct btrfs_key found_key; 5150 int slot; 5151 int sret; 5152 u32 nritems; 5153 int level; 5154 int ret = 1; 5155 int keep_locks = path->keep_locks; 5156 5157 path->keep_locks = 1; 5158 again: 5159 cur = btrfs_read_lock_root_node(root); 5160 level = btrfs_header_level(cur); 5161 WARN_ON(path->nodes[level]); 5162 path->nodes[level] = cur; 5163 path->locks[level] = BTRFS_READ_LOCK; 5164 5165 if (btrfs_header_generation(cur) < min_trans) { 5166 ret = 1; 5167 goto out; 5168 } 5169 while (1) { 5170 nritems = btrfs_header_nritems(cur); 5171 level = btrfs_header_level(cur); 5172 sret = bin_search(cur, min_key, level, &slot); 5173 5174 /* at the lowest level, we're done, setup the path and exit */ 5175 if (level == path->lowest_level) { 5176 if (slot >= nritems) 5177 goto find_next_key; 5178 ret = 0; 5179 path->slots[level] = slot; 5180 btrfs_item_key_to_cpu(cur, &found_key, slot); 5181 goto out; 5182 } 5183 if (sret && slot > 0) 5184 slot--; 5185 /* 5186 * check this node pointer against the min_trans parameters. 5187 * If it is too old, old, skip to the next one. 5188 */ 5189 while (slot < nritems) { 5190 u64 gen; 5191 5192 gen = btrfs_node_ptr_generation(cur, slot); 5193 if (gen < min_trans) { 5194 slot++; 5195 continue; 5196 } 5197 break; 5198 } 5199 find_next_key: 5200 /* 5201 * we didn't find a candidate key in this node, walk forward 5202 * and find another one 5203 */ 5204 if (slot >= nritems) { 5205 path->slots[level] = slot; 5206 btrfs_set_path_blocking(path); 5207 sret = btrfs_find_next_key(root, path, min_key, level, 5208 min_trans); 5209 if (sret == 0) { 5210 btrfs_release_path(path); 5211 goto again; 5212 } else { 5213 goto out; 5214 } 5215 } 5216 /* save our key for returning back */ 5217 btrfs_node_key_to_cpu(cur, &found_key, slot); 5218 path->slots[level] = slot; 5219 if (level == path->lowest_level) { 5220 ret = 0; 5221 goto out; 5222 } 5223 btrfs_set_path_blocking(path); 5224 cur = read_node_slot(fs_info, cur, slot); 5225 if (IS_ERR(cur)) { 5226 ret = PTR_ERR(cur); 5227 goto out; 5228 } 5229 5230 btrfs_tree_read_lock(cur); 5231 5232 path->locks[level - 1] = BTRFS_READ_LOCK; 5233 path->nodes[level - 1] = cur; 5234 unlock_up(path, level, 1, 0, NULL); 5235 btrfs_clear_path_blocking(path, NULL, 0); 5236 } 5237 out: 5238 path->keep_locks = keep_locks; 5239 if (ret == 0) { 5240 btrfs_unlock_up_safe(path, path->lowest_level + 1); 5241 btrfs_set_path_blocking(path); 5242 memcpy(min_key, &found_key, sizeof(found_key)); 5243 } 5244 return ret; 5245 } 5246 5247 static int tree_move_down(struct btrfs_fs_info *fs_info, 5248 struct btrfs_path *path, 5249 int *level) 5250 { 5251 struct extent_buffer *eb; 5252 5253 BUG_ON(*level == 0); 5254 eb = read_node_slot(fs_info, path->nodes[*level], path->slots[*level]); 5255 if (IS_ERR(eb)) 5256 return PTR_ERR(eb); 5257 5258 path->nodes[*level - 1] = eb; 5259 path->slots[*level - 1] = 0; 5260 (*level)--; 5261 return 0; 5262 } 5263 5264 static int tree_move_next_or_upnext(struct btrfs_path *path, 5265 int *level, int root_level) 5266 { 5267 int ret = 0; 5268 int nritems; 5269 nritems = btrfs_header_nritems(path->nodes[*level]); 5270 5271 path->slots[*level]++; 5272 5273 while (path->slots[*level] >= nritems) { 5274 if (*level == root_level) 5275 return -1; 5276 5277 /* move upnext */ 5278 path->slots[*level] = 0; 5279 free_extent_buffer(path->nodes[*level]); 5280 path->nodes[*level] = NULL; 5281 (*level)++; 5282 path->slots[*level]++; 5283 5284 nritems = btrfs_header_nritems(path->nodes[*level]); 5285 ret = 1; 5286 } 5287 return ret; 5288 } 5289 5290 /* 5291 * Returns 1 if it had to move up and next. 0 is returned if it moved only next 5292 * or down. 5293 */ 5294 static int tree_advance(struct btrfs_fs_info *fs_info, 5295 struct btrfs_path *path, 5296 int *level, int root_level, 5297 int allow_down, 5298 struct btrfs_key *key) 5299 { 5300 int ret; 5301 5302 if (*level == 0 || !allow_down) { 5303 ret = tree_move_next_or_upnext(path, level, root_level); 5304 } else { 5305 ret = tree_move_down(fs_info, path, level); 5306 } 5307 if (ret >= 0) { 5308 if (*level == 0) 5309 btrfs_item_key_to_cpu(path->nodes[*level], key, 5310 path->slots[*level]); 5311 else 5312 btrfs_node_key_to_cpu(path->nodes[*level], key, 5313 path->slots[*level]); 5314 } 5315 return ret; 5316 } 5317 5318 static int tree_compare_item(struct btrfs_path *left_path, 5319 struct btrfs_path *right_path, 5320 char *tmp_buf) 5321 { 5322 int cmp; 5323 int len1, len2; 5324 unsigned long off1, off2; 5325 5326 len1 = btrfs_item_size_nr(left_path->nodes[0], left_path->slots[0]); 5327 len2 = btrfs_item_size_nr(right_path->nodes[0], right_path->slots[0]); 5328 if (len1 != len2) 5329 return 1; 5330 5331 off1 = btrfs_item_ptr_offset(left_path->nodes[0], left_path->slots[0]); 5332 off2 = btrfs_item_ptr_offset(right_path->nodes[0], 5333 right_path->slots[0]); 5334 5335 read_extent_buffer(left_path->nodes[0], tmp_buf, off1, len1); 5336 5337 cmp = memcmp_extent_buffer(right_path->nodes[0], tmp_buf, off2, len1); 5338 if (cmp) 5339 return 1; 5340 return 0; 5341 } 5342 5343 #define ADVANCE 1 5344 #define ADVANCE_ONLY_NEXT -1 5345 5346 /* 5347 * This function compares two trees and calls the provided callback for 5348 * every changed/new/deleted item it finds. 5349 * If shared tree blocks are encountered, whole subtrees are skipped, making 5350 * the compare pretty fast on snapshotted subvolumes. 5351 * 5352 * This currently works on commit roots only. As commit roots are read only, 5353 * we don't do any locking. The commit roots are protected with transactions. 5354 * Transactions are ended and rejoined when a commit is tried in between. 5355 * 5356 * This function checks for modifications done to the trees while comparing. 5357 * If it detects a change, it aborts immediately. 5358 */ 5359 int btrfs_compare_trees(struct btrfs_root *left_root, 5360 struct btrfs_root *right_root, 5361 btrfs_changed_cb_t changed_cb, void *ctx) 5362 { 5363 struct btrfs_fs_info *fs_info = left_root->fs_info; 5364 int ret; 5365 int cmp; 5366 struct btrfs_path *left_path = NULL; 5367 struct btrfs_path *right_path = NULL; 5368 struct btrfs_key left_key; 5369 struct btrfs_key right_key; 5370 char *tmp_buf = NULL; 5371 int left_root_level; 5372 int right_root_level; 5373 int left_level; 5374 int right_level; 5375 int left_end_reached; 5376 int right_end_reached; 5377 int advance_left; 5378 int advance_right; 5379 u64 left_blockptr; 5380 u64 right_blockptr; 5381 u64 left_gen; 5382 u64 right_gen; 5383 5384 left_path = btrfs_alloc_path(); 5385 if (!left_path) { 5386 ret = -ENOMEM; 5387 goto out; 5388 } 5389 right_path = btrfs_alloc_path(); 5390 if (!right_path) { 5391 ret = -ENOMEM; 5392 goto out; 5393 } 5394 5395 tmp_buf = kvmalloc(fs_info->nodesize, GFP_KERNEL); 5396 if (!tmp_buf) { 5397 ret = -ENOMEM; 5398 goto out; 5399 } 5400 5401 left_path->search_commit_root = 1; 5402 left_path->skip_locking = 1; 5403 right_path->search_commit_root = 1; 5404 right_path->skip_locking = 1; 5405 5406 /* 5407 * Strategy: Go to the first items of both trees. Then do 5408 * 5409 * If both trees are at level 0 5410 * Compare keys of current items 5411 * If left < right treat left item as new, advance left tree 5412 * and repeat 5413 * If left > right treat right item as deleted, advance right tree 5414 * and repeat 5415 * If left == right do deep compare of items, treat as changed if 5416 * needed, advance both trees and repeat 5417 * If both trees are at the same level but not at level 0 5418 * Compare keys of current nodes/leafs 5419 * If left < right advance left tree and repeat 5420 * If left > right advance right tree and repeat 5421 * If left == right compare blockptrs of the next nodes/leafs 5422 * If they match advance both trees but stay at the same level 5423 * and repeat 5424 * If they don't match advance both trees while allowing to go 5425 * deeper and repeat 5426 * If tree levels are different 5427 * Advance the tree that needs it and repeat 5428 * 5429 * Advancing a tree means: 5430 * If we are at level 0, try to go to the next slot. If that's not 5431 * possible, go one level up and repeat. Stop when we found a level 5432 * where we could go to the next slot. We may at this point be on a 5433 * node or a leaf. 5434 * 5435 * If we are not at level 0 and not on shared tree blocks, go one 5436 * level deeper. 5437 * 5438 * If we are not at level 0 and on shared tree blocks, go one slot to 5439 * the right if possible or go up and right. 5440 */ 5441 5442 down_read(&fs_info->commit_root_sem); 5443 left_level = btrfs_header_level(left_root->commit_root); 5444 left_root_level = left_level; 5445 left_path->nodes[left_level] = left_root->commit_root; 5446 extent_buffer_get(left_path->nodes[left_level]); 5447 5448 right_level = btrfs_header_level(right_root->commit_root); 5449 right_root_level = right_level; 5450 right_path->nodes[right_level] = right_root->commit_root; 5451 extent_buffer_get(right_path->nodes[right_level]); 5452 up_read(&fs_info->commit_root_sem); 5453 5454 if (left_level == 0) 5455 btrfs_item_key_to_cpu(left_path->nodes[left_level], 5456 &left_key, left_path->slots[left_level]); 5457 else 5458 btrfs_node_key_to_cpu(left_path->nodes[left_level], 5459 &left_key, left_path->slots[left_level]); 5460 if (right_level == 0) 5461 btrfs_item_key_to_cpu(right_path->nodes[right_level], 5462 &right_key, right_path->slots[right_level]); 5463 else 5464 btrfs_node_key_to_cpu(right_path->nodes[right_level], 5465 &right_key, right_path->slots[right_level]); 5466 5467 left_end_reached = right_end_reached = 0; 5468 advance_left = advance_right = 0; 5469 5470 while (1) { 5471 if (advance_left && !left_end_reached) { 5472 ret = tree_advance(fs_info, left_path, &left_level, 5473 left_root_level, 5474 advance_left != ADVANCE_ONLY_NEXT, 5475 &left_key); 5476 if (ret == -1) 5477 left_end_reached = ADVANCE; 5478 else if (ret < 0) 5479 goto out; 5480 advance_left = 0; 5481 } 5482 if (advance_right && !right_end_reached) { 5483 ret = tree_advance(fs_info, right_path, &right_level, 5484 right_root_level, 5485 advance_right != ADVANCE_ONLY_NEXT, 5486 &right_key); 5487 if (ret == -1) 5488 right_end_reached = ADVANCE; 5489 else if (ret < 0) 5490 goto out; 5491 advance_right = 0; 5492 } 5493 5494 if (left_end_reached && right_end_reached) { 5495 ret = 0; 5496 goto out; 5497 } else if (left_end_reached) { 5498 if (right_level == 0) { 5499 ret = changed_cb(left_root, right_root, 5500 left_path, right_path, 5501 &right_key, 5502 BTRFS_COMPARE_TREE_DELETED, 5503 ctx); 5504 if (ret < 0) 5505 goto out; 5506 } 5507 advance_right = ADVANCE; 5508 continue; 5509 } else if (right_end_reached) { 5510 if (left_level == 0) { 5511 ret = changed_cb(left_root, right_root, 5512 left_path, right_path, 5513 &left_key, 5514 BTRFS_COMPARE_TREE_NEW, 5515 ctx); 5516 if (ret < 0) 5517 goto out; 5518 } 5519 advance_left = ADVANCE; 5520 continue; 5521 } 5522 5523 if (left_level == 0 && right_level == 0) { 5524 cmp = btrfs_comp_cpu_keys(&left_key, &right_key); 5525 if (cmp < 0) { 5526 ret = changed_cb(left_root, right_root, 5527 left_path, right_path, 5528 &left_key, 5529 BTRFS_COMPARE_TREE_NEW, 5530 ctx); 5531 if (ret < 0) 5532 goto out; 5533 advance_left = ADVANCE; 5534 } else if (cmp > 0) { 5535 ret = changed_cb(left_root, right_root, 5536 left_path, right_path, 5537 &right_key, 5538 BTRFS_COMPARE_TREE_DELETED, 5539 ctx); 5540 if (ret < 0) 5541 goto out; 5542 advance_right = ADVANCE; 5543 } else { 5544 enum btrfs_compare_tree_result result; 5545 5546 WARN_ON(!extent_buffer_uptodate(left_path->nodes[0])); 5547 ret = tree_compare_item(left_path, right_path, 5548 tmp_buf); 5549 if (ret) 5550 result = BTRFS_COMPARE_TREE_CHANGED; 5551 else 5552 result = BTRFS_COMPARE_TREE_SAME; 5553 ret = changed_cb(left_root, right_root, 5554 left_path, right_path, 5555 &left_key, result, ctx); 5556 if (ret < 0) 5557 goto out; 5558 advance_left = ADVANCE; 5559 advance_right = ADVANCE; 5560 } 5561 } else if (left_level == right_level) { 5562 cmp = btrfs_comp_cpu_keys(&left_key, &right_key); 5563 if (cmp < 0) { 5564 advance_left = ADVANCE; 5565 } else if (cmp > 0) { 5566 advance_right = ADVANCE; 5567 } else { 5568 left_blockptr = btrfs_node_blockptr( 5569 left_path->nodes[left_level], 5570 left_path->slots[left_level]); 5571 right_blockptr = btrfs_node_blockptr( 5572 right_path->nodes[right_level], 5573 right_path->slots[right_level]); 5574 left_gen = btrfs_node_ptr_generation( 5575 left_path->nodes[left_level], 5576 left_path->slots[left_level]); 5577 right_gen = btrfs_node_ptr_generation( 5578 right_path->nodes[right_level], 5579 right_path->slots[right_level]); 5580 if (left_blockptr == right_blockptr && 5581 left_gen == right_gen) { 5582 /* 5583 * As we're on a shared block, don't 5584 * allow to go deeper. 5585 */ 5586 advance_left = ADVANCE_ONLY_NEXT; 5587 advance_right = ADVANCE_ONLY_NEXT; 5588 } else { 5589 advance_left = ADVANCE; 5590 advance_right = ADVANCE; 5591 } 5592 } 5593 } else if (left_level < right_level) { 5594 advance_right = ADVANCE; 5595 } else { 5596 advance_left = ADVANCE; 5597 } 5598 } 5599 5600 out: 5601 btrfs_free_path(left_path); 5602 btrfs_free_path(right_path); 5603 kvfree(tmp_buf); 5604 return ret; 5605 } 5606 5607 /* 5608 * this is similar to btrfs_next_leaf, but does not try to preserve 5609 * and fixup the path. It looks for and returns the next key in the 5610 * tree based on the current path and the min_trans parameters. 5611 * 5612 * 0 is returned if another key is found, < 0 if there are any errors 5613 * and 1 is returned if there are no higher keys in the tree 5614 * 5615 * path->keep_locks should be set to 1 on the search made before 5616 * calling this function. 5617 */ 5618 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path, 5619 struct btrfs_key *key, int level, u64 min_trans) 5620 { 5621 int slot; 5622 struct extent_buffer *c; 5623 5624 WARN_ON(!path->keep_locks); 5625 while (level < BTRFS_MAX_LEVEL) { 5626 if (!path->nodes[level]) 5627 return 1; 5628 5629 slot = path->slots[level] + 1; 5630 c = path->nodes[level]; 5631 next: 5632 if (slot >= btrfs_header_nritems(c)) { 5633 int ret; 5634 int orig_lowest; 5635 struct btrfs_key cur_key; 5636 if (level + 1 >= BTRFS_MAX_LEVEL || 5637 !path->nodes[level + 1]) 5638 return 1; 5639 5640 if (path->locks[level + 1]) { 5641 level++; 5642 continue; 5643 } 5644 5645 slot = btrfs_header_nritems(c) - 1; 5646 if (level == 0) 5647 btrfs_item_key_to_cpu(c, &cur_key, slot); 5648 else 5649 btrfs_node_key_to_cpu(c, &cur_key, slot); 5650 5651 orig_lowest = path->lowest_level; 5652 btrfs_release_path(path); 5653 path->lowest_level = level; 5654 ret = btrfs_search_slot(NULL, root, &cur_key, path, 5655 0, 0); 5656 path->lowest_level = orig_lowest; 5657 if (ret < 0) 5658 return ret; 5659 5660 c = path->nodes[level]; 5661 slot = path->slots[level]; 5662 if (ret == 0) 5663 slot++; 5664 goto next; 5665 } 5666 5667 if (level == 0) 5668 btrfs_item_key_to_cpu(c, key, slot); 5669 else { 5670 u64 gen = btrfs_node_ptr_generation(c, slot); 5671 5672 if (gen < min_trans) { 5673 slot++; 5674 goto next; 5675 } 5676 btrfs_node_key_to_cpu(c, key, slot); 5677 } 5678 return 0; 5679 } 5680 return 1; 5681 } 5682 5683 /* 5684 * search the tree again to find a leaf with greater keys 5685 * returns 0 if it found something or 1 if there are no greater leaves. 5686 * returns < 0 on io errors. 5687 */ 5688 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path) 5689 { 5690 return btrfs_next_old_leaf(root, path, 0); 5691 } 5692 5693 int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path, 5694 u64 time_seq) 5695 { 5696 int slot; 5697 int level; 5698 struct extent_buffer *c; 5699 struct extent_buffer *next; 5700 struct btrfs_key key; 5701 u32 nritems; 5702 int ret; 5703 int old_spinning = path->leave_spinning; 5704 int next_rw_lock = 0; 5705 5706 nritems = btrfs_header_nritems(path->nodes[0]); 5707 if (nritems == 0) 5708 return 1; 5709 5710 btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1); 5711 again: 5712 level = 1; 5713 next = NULL; 5714 next_rw_lock = 0; 5715 btrfs_release_path(path); 5716 5717 path->keep_locks = 1; 5718 path->leave_spinning = 1; 5719 5720 if (time_seq) 5721 ret = btrfs_search_old_slot(root, &key, path, time_seq); 5722 else 5723 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5724 path->keep_locks = 0; 5725 5726 if (ret < 0) 5727 return ret; 5728 5729 nritems = btrfs_header_nritems(path->nodes[0]); 5730 /* 5731 * by releasing the path above we dropped all our locks. A balance 5732 * could have added more items next to the key that used to be 5733 * at the very end of the block. So, check again here and 5734 * advance the path if there are now more items available. 5735 */ 5736 if (nritems > 0 && path->slots[0] < nritems - 1) { 5737 if (ret == 0) 5738 path->slots[0]++; 5739 ret = 0; 5740 goto done; 5741 } 5742 /* 5743 * So the above check misses one case: 5744 * - after releasing the path above, someone has removed the item that 5745 * used to be at the very end of the block, and balance between leafs 5746 * gets another one with bigger key.offset to replace it. 5747 * 5748 * This one should be returned as well, or we can get leaf corruption 5749 * later(esp. in __btrfs_drop_extents()). 5750 * 5751 * And a bit more explanation about this check, 5752 * with ret > 0, the key isn't found, the path points to the slot 5753 * where it should be inserted, so the path->slots[0] item must be the 5754 * bigger one. 5755 */ 5756 if (nritems > 0 && ret > 0 && path->slots[0] == nritems - 1) { 5757 ret = 0; 5758 goto done; 5759 } 5760 5761 while (level < BTRFS_MAX_LEVEL) { 5762 if (!path->nodes[level]) { 5763 ret = 1; 5764 goto done; 5765 } 5766 5767 slot = path->slots[level] + 1; 5768 c = path->nodes[level]; 5769 if (slot >= btrfs_header_nritems(c)) { 5770 level++; 5771 if (level == BTRFS_MAX_LEVEL) { 5772 ret = 1; 5773 goto done; 5774 } 5775 continue; 5776 } 5777 5778 if (next) { 5779 btrfs_tree_unlock_rw(next, next_rw_lock); 5780 free_extent_buffer(next); 5781 } 5782 5783 next = c; 5784 next_rw_lock = path->locks[level]; 5785 ret = read_block_for_search(root, path, &next, level, 5786 slot, &key); 5787 if (ret == -EAGAIN) 5788 goto again; 5789 5790 if (ret < 0) { 5791 btrfs_release_path(path); 5792 goto done; 5793 } 5794 5795 if (!path->skip_locking) { 5796 ret = btrfs_try_tree_read_lock(next); 5797 if (!ret && time_seq) { 5798 /* 5799 * If we don't get the lock, we may be racing 5800 * with push_leaf_left, holding that lock while 5801 * itself waiting for the leaf we've currently 5802 * locked. To solve this situation, we give up 5803 * on our lock and cycle. 5804 */ 5805 free_extent_buffer(next); 5806 btrfs_release_path(path); 5807 cond_resched(); 5808 goto again; 5809 } 5810 if (!ret) { 5811 btrfs_set_path_blocking(path); 5812 btrfs_tree_read_lock(next); 5813 btrfs_clear_path_blocking(path, next, 5814 BTRFS_READ_LOCK); 5815 } 5816 next_rw_lock = BTRFS_READ_LOCK; 5817 } 5818 break; 5819 } 5820 path->slots[level] = slot; 5821 while (1) { 5822 level--; 5823 c = path->nodes[level]; 5824 if (path->locks[level]) 5825 btrfs_tree_unlock_rw(c, path->locks[level]); 5826 5827 free_extent_buffer(c); 5828 path->nodes[level] = next; 5829 path->slots[level] = 0; 5830 if (!path->skip_locking) 5831 path->locks[level] = next_rw_lock; 5832 if (!level) 5833 break; 5834 5835 ret = read_block_for_search(root, path, &next, level, 5836 0, &key); 5837 if (ret == -EAGAIN) 5838 goto again; 5839 5840 if (ret < 0) { 5841 btrfs_release_path(path); 5842 goto done; 5843 } 5844 5845 if (!path->skip_locking) { 5846 ret = btrfs_try_tree_read_lock(next); 5847 if (!ret) { 5848 btrfs_set_path_blocking(path); 5849 btrfs_tree_read_lock(next); 5850 btrfs_clear_path_blocking(path, next, 5851 BTRFS_READ_LOCK); 5852 } 5853 next_rw_lock = BTRFS_READ_LOCK; 5854 } 5855 } 5856 ret = 0; 5857 done: 5858 unlock_up(path, 0, 1, 0, NULL); 5859 path->leave_spinning = old_spinning; 5860 if (!old_spinning) 5861 btrfs_set_path_blocking(path); 5862 5863 return ret; 5864 } 5865 5866 /* 5867 * this uses btrfs_prev_leaf to walk backwards in the tree, and keeps 5868 * searching until it gets past min_objectid or finds an item of 'type' 5869 * 5870 * returns 0 if something is found, 1 if nothing was found and < 0 on error 5871 */ 5872 int btrfs_previous_item(struct btrfs_root *root, 5873 struct btrfs_path *path, u64 min_objectid, 5874 int type) 5875 { 5876 struct btrfs_key found_key; 5877 struct extent_buffer *leaf; 5878 u32 nritems; 5879 int ret; 5880 5881 while (1) { 5882 if (path->slots[0] == 0) { 5883 btrfs_set_path_blocking(path); 5884 ret = btrfs_prev_leaf(root, path); 5885 if (ret != 0) 5886 return ret; 5887 } else { 5888 path->slots[0]--; 5889 } 5890 leaf = path->nodes[0]; 5891 nritems = btrfs_header_nritems(leaf); 5892 if (nritems == 0) 5893 return 1; 5894 if (path->slots[0] == nritems) 5895 path->slots[0]--; 5896 5897 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 5898 if (found_key.objectid < min_objectid) 5899 break; 5900 if (found_key.type == type) 5901 return 0; 5902 if (found_key.objectid == min_objectid && 5903 found_key.type < type) 5904 break; 5905 } 5906 return 1; 5907 } 5908 5909 /* 5910 * search in extent tree to find a previous Metadata/Data extent item with 5911 * min objecitd. 5912 * 5913 * returns 0 if something is found, 1 if nothing was found and < 0 on error 5914 */ 5915 int btrfs_previous_extent_item(struct btrfs_root *root, 5916 struct btrfs_path *path, u64 min_objectid) 5917 { 5918 struct btrfs_key found_key; 5919 struct extent_buffer *leaf; 5920 u32 nritems; 5921 int ret; 5922 5923 while (1) { 5924 if (path->slots[0] == 0) { 5925 btrfs_set_path_blocking(path); 5926 ret = btrfs_prev_leaf(root, path); 5927 if (ret != 0) 5928 return ret; 5929 } else { 5930 path->slots[0]--; 5931 } 5932 leaf = path->nodes[0]; 5933 nritems = btrfs_header_nritems(leaf); 5934 if (nritems == 0) 5935 return 1; 5936 if (path->slots[0] == nritems) 5937 path->slots[0]--; 5938 5939 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 5940 if (found_key.objectid < min_objectid) 5941 break; 5942 if (found_key.type == BTRFS_EXTENT_ITEM_KEY || 5943 found_key.type == BTRFS_METADATA_ITEM_KEY) 5944 return 0; 5945 if (found_key.objectid == min_objectid && 5946 found_key.type < BTRFS_EXTENT_ITEM_KEY) 5947 break; 5948 } 5949 return 1; 5950 } 5951