fs/btrfs/ctree.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2007,2008 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/sched.h>
   7 #include <linux/slab.h>
   8 #include <linux/rbtree.h>
   9 #include <linux/mm.h>
  10 #include "ctree.h"
  11 #include "disk-io.h"
  12 #include "transaction.h"
  13 #include "print-tree.h"
  14 #include "locking.h"
  15 #include "volumes.h"
  16 #include "qgroup.h"
  17 #include "tree-mod-log.h"
  18
  19 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
  20                       *root, struct btrfs_path *path, int level);
  21 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root,
  22                       const struct btrfs_key *ins_key, struct btrfs_path *path,
  23                       int data_size, int extend);
  24 static int push_node_left(struct btrfs_trans_handle *trans,
  25                           struct extent_buffer *dst,
  26                           struct extent_buffer *src, int empty);
  27 static int balance_node_right(struct btrfs_trans_handle *trans,
  28                               struct extent_buffer *dst_buf,
  29                               struct extent_buffer *src_buf);
  30 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
  31                     int level, int slot);
  32
  33 static const struct btrfs_csums {
  34         u16             size;
  35         const char      name[10];
  36         const char      driver[12];
  37 } btrfs_csums[] = {
  38         [BTRFS_CSUM_TYPE_CRC32] = { .size = 4, .name = "crc32c" },
  39         [BTRFS_CSUM_TYPE_XXHASH] = { .size = 8, .name = "xxhash64" },
  40         [BTRFS_CSUM_TYPE_SHA256] = { .size = 32, .name = "sha256" },
  41         [BTRFS_CSUM_TYPE_BLAKE2] = { .size = 32, .name = "blake2b",
  42                                      .driver = "blake2b-256" },
  43 };
  44
  45 int btrfs_super_csum_size(const struct btrfs_super_block *s)
  46 {
  47         u16 t = btrfs_super_csum_type(s);
  48         /*
  49          * csum type is validated at mount time
  50          */
  51         return btrfs_csums[t].size;
  52 }
  53
  54 const char *btrfs_super_csum_name(u16 csum_type)
  55 {
  56         /* csum type is validated at mount time */
  57         return btrfs_csums[csum_type].name;
  58 }
  59
  60 /*
  61  * Return driver name if defined, otherwise the name that's also a valid driver
  62  * name
  63  */
  64 const char *btrfs_super_csum_driver(u16 csum_type)
  65 {
  66         /* csum type is validated at mount time */
  67         return btrfs_csums[csum_type].driver[0] ?
  68                 btrfs_csums[csum_type].driver :
  69                 btrfs_csums[csum_type].name;
  70 }
  71
  72 size_t __attribute_const__ btrfs_get_num_csums(void)
  73 {
  74         return ARRAY_SIZE(btrfs_csums);
  75 }
  76
  77 struct btrfs_path *btrfs_alloc_path(void)
  78 {
  79         return kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS);
  80 }
  81
  82 /* this also releases the path */
  83 void btrfs_free_path(struct btrfs_path *p)
  84 {
  85         if (!p)
  86                 return;
  87         btrfs_release_path(p);
  88         kmem_cache_free(btrfs_path_cachep, p);
  89 }
  90
  91 /*
  92  * path release drops references on the extent buffers in the path
  93  * and it drops any locks held by this path
  94  *
  95  * It is safe to call this on paths that no locks or extent buffers held.
  96  */
  97 noinline void btrfs_release_path(struct btrfs_path *p)
  98 {
  99         int i;
 100
 101         for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
 102                 p->slots[i] = 0;
 103                 if (!p->nodes[i])
 104                         continue;
 105                 if (p->locks[i]) {
 106                         btrfs_tree_unlock_rw(p->nodes[i], p->locks[i]);
 107                         p->locks[i] = 0;
 108                 }
 109                 free_extent_buffer(p->nodes[i]);
 110                 p->nodes[i] = NULL;
 111         }
 112 }
 113
 114 /*
 115  * safely gets a reference on the root node of a tree.  A lock
 116  * is not taken, so a concurrent writer may put a different node
 117  * at the root of the tree.  See btrfs_lock_root_node for the
 118  * looping required.
 119  *
 120  * The extent buffer returned by this has a reference taken, so
 121  * it won't disappear.  It may stop being the root of the tree
 122  * at any time because there are no locks held.
 123  */
 124 struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
 125 {
 126         struct extent_buffer *eb;
 127
 128         while (1) {
 129                 rcu_read_lock();
 130                 eb = rcu_dereference(root->node);
 131
 132                 /*
 133                  * RCU really hurts here, we could free up the root node because
 134                  * it was COWed but we may not get the new root node yet so do
 135                  * the inc_not_zero dance and if it doesn't work then
 136                  * synchronize_rcu and try again.
 137                  */
 138                 if (atomic_inc_not_zero(&eb->refs)) {
 139                         rcu_read_unlock();
 140                         break;
 141                 }
 142                 rcu_read_unlock();
 143                 synchronize_rcu();
 144         }
 145         return eb;
 146 }
 147
 148 /*
 149  * Cowonly root (not-shareable trees, everything not subvolume or reloc roots),
 150  * just get put onto a simple dirty list.  Transaction walks this list to make
 151  * sure they get properly updated on disk.
 152  */
 153 static void add_root_to_dirty_list(struct btrfs_root *root)
 154 {
 155         struct btrfs_fs_info *fs_info = root->fs_info;
 156
 157         if (test_bit(BTRFS_ROOT_DIRTY, &root->state) ||
 158             !test_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state))
 159                 return;
 160
 161         spin_lock(&fs_info->trans_lock);
 162         if (!test_and_set_bit(BTRFS_ROOT_DIRTY, &root->state)) {
 163                 /* Want the extent tree to be the last on the list */
 164                 if (root->root_key.objectid == BTRFS_EXTENT_TREE_OBJECTID)
 165                         list_move_tail(&root->dirty_list,
 166                                        &fs_info->dirty_cowonly_roots);
 167                 else
 168                         list_move(&root->dirty_list,
 169                                   &fs_info->dirty_cowonly_roots);
 170         }
 171         spin_unlock(&fs_info->trans_lock);
 172 }
 173
 174 /*
 175  * used by snapshot creation to make a copy of a root for a tree with
 176  * a given objectid.  The buffer with the new root node is returned in
 177  * cow_ret, and this func returns zero on success or a negative error code.
 178  */
 179 int btrfs_copy_root(struct btrfs_trans_handle *trans,
 180                       struct btrfs_root *root,
 181                       struct extent_buffer *buf,
 182                       struct extent_buffer **cow_ret, u64 new_root_objectid)
 183 {
 184         struct btrfs_fs_info *fs_info = root->fs_info;
 185         struct extent_buffer *cow;
 186         int ret = 0;
 187         int level;
 188         struct btrfs_disk_key disk_key;
 189
 190         WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
 191                 trans->transid != fs_info->running_transaction->transid);
 192         WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
 193                 trans->transid != root->last_trans);
 194
 195         level = btrfs_header_level(buf);
 196         if (level == 0)
 197                 btrfs_item_key(buf, &disk_key, 0);
 198         else
 199                 btrfs_node_key(buf, &disk_key, 0);
 200
 201         cow = btrfs_alloc_tree_block(trans, root, 0, new_root_objectid,
 202                                      &disk_key, level, buf->start, 0,
 203                                      BTRFS_NESTING_NEW_ROOT);
 204         if (IS_ERR(cow))
 205                 return PTR_ERR(cow);
 206
 207         copy_extent_buffer_full(cow, buf);
 208         btrfs_set_header_bytenr(cow, cow->start);
 209         btrfs_set_header_generation(cow, trans->transid);
 210         btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV);
 211         btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN |
 212                                      BTRFS_HEADER_FLAG_RELOC);
 213         if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
 214                 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC);
 215         else
 216                 btrfs_set_header_owner(cow, new_root_objectid);
 217
 218         write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid);
 219
 220         WARN_ON(btrfs_header_generation(buf) > trans->transid);
 221         if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
 222                 ret = btrfs_inc_ref(trans, root, cow, 1);
 223         else
 224                 ret = btrfs_inc_ref(trans, root, cow, 0);
 225         if (ret) {
 226                 btrfs_tree_unlock(cow);
 227                 free_extent_buffer(cow);
 228                 btrfs_abort_transaction(trans, ret);
 229                 return ret;
 230         }
 231
 232         btrfs_mark_buffer_dirty(cow);
 233         *cow_ret = cow;
 234         return 0;
 235 }
 236
 237 /*
 238  * check if the tree block can be shared by multiple trees
 239  */
 240 int btrfs_block_can_be_shared(struct btrfs_root *root,
 241                               struct extent_buffer *buf)
 242 {
 243         /*
 244          * Tree blocks not in shareable trees and tree roots are never shared.
 245          * If a block was allocated after the last snapshot and the block was
 246          * not allocated by tree relocation, we know the block is not shared.
 247          */
 248         if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
 249             buf != root->node && buf != root->commit_root &&
 250             (btrfs_header_generation(buf) <=
 251              btrfs_root_last_snapshot(&root->root_item) ||
 252              btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
 253                 return 1;
 254
 255         return 0;
 256 }
 257
 258 static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
 259                                        struct btrfs_root *root,
 260                                        struct extent_buffer *buf,
 261                                        struct extent_buffer *cow,
 262                                        int *last_ref)
 263 {
 264         struct btrfs_fs_info *fs_info = root->fs_info;
 265         u64 refs;
 266         u64 owner;
 267         u64 flags;
 268         u64 new_flags = 0;
 269         int ret;
 270
 271         /*
 272          * Backrefs update rules:
 273          *
 274          * Always use full backrefs for extent pointers in tree block
 275          * allocated by tree relocation.
 276          *
 277          * If a shared tree block is no longer referenced by its owner
 278          * tree (btrfs_header_owner(buf) == root->root_key.objectid),
 279          * use full backrefs for extent pointers in tree block.
 280          *
 281          * If a tree block is been relocating
 282          * (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID),
 283          * use full backrefs for extent pointers in tree block.
 284          * The reason for this is some operations (such as drop tree)
 285          * are only allowed for blocks use full backrefs.
 286          */
 287
 288         if (btrfs_block_can_be_shared(root, buf)) {
 289                 ret = btrfs_lookup_extent_info(trans, fs_info, buf->start,
 290                                                btrfs_header_level(buf), 1,
 291                                                &refs, &flags);
 292                 if (ret)
 293                         return ret;
 294                 if (refs == 0) {
 295                         ret = -EROFS;
 296                         btrfs_handle_fs_error(fs_info, ret, NULL);
 297                         return ret;
 298                 }
 299         } else {
 300                 refs = 1;
 301                 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
 302                     btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
 303                         flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
 304                 else
 305                         flags = 0;
 306         }
 307
 308         owner = btrfs_header_owner(buf);
 309         BUG_ON(owner == BTRFS_TREE_RELOC_OBJECTID &&
 310                !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
 311
 312         if (refs > 1) {
 313                 if ((owner == root->root_key.objectid ||
 314                      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) &&
 315                     !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
 316                         ret = btrfs_inc_ref(trans, root, buf, 1);
 317                         if (ret)
 318                                 return ret;
 319
 320                         if (root->root_key.objectid ==
 321                             BTRFS_TREE_RELOC_OBJECTID) {
 322                                 ret = btrfs_dec_ref(trans, root, buf, 0);
 323                                 if (ret)
 324                                         return ret;
 325                                 ret = btrfs_inc_ref(trans, root, cow, 1);
 326                                 if (ret)
 327                                         return ret;
 328                         }
 329                         new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
 330                 } else {
 331
 332                         if (root->root_key.objectid ==
 333                             BTRFS_TREE_RELOC_OBJECTID)
 334                                 ret = btrfs_inc_ref(trans, root, cow, 1);
 335                         else
 336                                 ret = btrfs_inc_ref(trans, root, cow, 0);
 337                         if (ret)
 338                                 return ret;
 339                 }
 340                 if (new_flags != 0) {
 341                         int level = btrfs_header_level(buf);
 342
 343                         ret = btrfs_set_disk_extent_flags(trans, buf,
 344                                                           new_flags, level, 0);
 345                         if (ret)
 346                                 return ret;
 347                 }
 348         } else {
 349                 if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
 350                         if (root->root_key.objectid ==
 351                             BTRFS_TREE_RELOC_OBJECTID)
 352                                 ret = btrfs_inc_ref(trans, root, cow, 1);
 353                         else
 354                                 ret = btrfs_inc_ref(trans, root, cow, 0);
 355                         if (ret)
 356                                 return ret;
 357                         ret = btrfs_dec_ref(trans, root, buf, 1);
 358                         if (ret)
 359                                 return ret;
 360                 }
 361                 btrfs_clean_tree_block(buf);
 362                 *last_ref = 1;
 363         }
 364         return 0;
 365 }
 366
 367 static struct extent_buffer *alloc_tree_block_no_bg_flush(
 368                                           struct btrfs_trans_handle *trans,
 369                                           struct btrfs_root *root,
 370                                           u64 parent_start,
 371                                           const struct btrfs_disk_key *disk_key,
 372                                           int level,
 373                                           u64 hint,
 374                                           u64 empty_size,
 375                                           enum btrfs_lock_nesting nest)
 376 {
 377         struct btrfs_fs_info *fs_info = root->fs_info;
 378         struct extent_buffer *ret;
 379
 380         /*
 381          * If we are COWing a node/leaf from the extent, chunk, device or free
 382          * space trees, make sure that we do not finish block group creation of
 383          * pending block groups. We do this to avoid a deadlock.
 384          * COWing can result in allocation of a new chunk, and flushing pending
 385          * block groups (btrfs_create_pending_block_groups()) can be triggered
 386          * when finishing allocation of a new chunk. Creation of a pending block
 387          * group modifies the extent, chunk, device and free space trees,
 388          * therefore we could deadlock with ourselves since we are holding a
 389          * lock on an extent buffer that btrfs_create_pending_block_groups() may
 390          * try to COW later.
 391          * For similar reasons, we also need to delay flushing pending block
 392          * groups when splitting a leaf or node, from one of those trees, since
 393          * we are holding a write lock on it and its parent or when inserting a
 394          * new root node for one of those trees.
 395          */
 396         if (root == fs_info->extent_root ||
 397             root == fs_info->chunk_root ||
 398             root == fs_info->dev_root ||
 399             root == fs_info->free_space_root)
 400                 trans->can_flush_pending_bgs = false;
 401
 402         ret = btrfs_alloc_tree_block(trans, root, parent_start,
 403                                      root->root_key.objectid, disk_key, level,
 404                                      hint, empty_size, nest);
 405         trans->can_flush_pending_bgs = true;
 406
 407         return ret;
 408 }
 409
 410 /*
 411  * does the dirty work in cow of a single block.  The parent block (if
 412  * supplied) is updated to point to the new cow copy.  The new buffer is marked
 413  * dirty and returned locked.  If you modify the block it needs to be marked
 414  * dirty again.
 415  *
 416  * search_start -- an allocation hint for the new block
 417  *
 418  * empty_size -- a hint that you plan on doing more cow.  This is the size in
 419  * bytes the allocator should try to find free next to the block it returns.
 420  * This is just a hint and may be ignored by the allocator.
 421  */
 422 static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 423                              struct btrfs_root *root,
 424                              struct extent_buffer *buf,
 425                              struct extent_buffer *parent, int parent_slot,
 426                              struct extent_buffer **cow_ret,
 427                              u64 search_start, u64 empty_size,
 428                              enum btrfs_lock_nesting nest)
 429 {
 430         struct btrfs_fs_info *fs_info = root->fs_info;
 431         struct btrfs_disk_key disk_key;
 432         struct extent_buffer *cow;
 433         int level, ret;
 434         int last_ref = 0;
 435         int unlock_orig = 0;
 436         u64 parent_start = 0;
 437
 438         if (*cow_ret == buf)
 439                 unlock_orig = 1;
 440
 441         btrfs_assert_tree_locked(buf);
 442
 443         WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
 444                 trans->transid != fs_info->running_transaction->transid);
 445         WARN_ON(test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
 446                 trans->transid != root->last_trans);
 447
 448         level = btrfs_header_level(buf);
 449
 450         if (level == 0)
 451                 btrfs_item_key(buf, &disk_key, 0);
 452         else
 453                 btrfs_node_key(buf, &disk_key, 0);
 454
 455         if ((root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && parent)
 456                 parent_start = parent->start;
 457
 458         cow = alloc_tree_block_no_bg_flush(trans, root, parent_start, &disk_key,
 459                                            level, search_start, empty_size, nest);
 460         if (IS_ERR(cow))
 461                 return PTR_ERR(cow);
 462
 463         /* cow is set to blocking by btrfs_init_new_buffer */
 464
 465         copy_extent_buffer_full(cow, buf);
 466         btrfs_set_header_bytenr(cow, cow->start);
 467         btrfs_set_header_generation(cow, trans->transid);
 468         btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV);
 469         btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN |
 470                                      BTRFS_HEADER_FLAG_RELOC);
 471         if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
 472                 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC);
 473         else
 474                 btrfs_set_header_owner(cow, root->root_key.objectid);
 475
 476         write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid);
 477
 478         ret = update_ref_for_cow(trans, root, buf, cow, &last_ref);
 479         if (ret) {
 480                 btrfs_tree_unlock(cow);
 481                 free_extent_buffer(cow);
 482                 btrfs_abort_transaction(trans, ret);
 483                 return ret;
 484         }
 485
 486         if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) {
 487                 ret = btrfs_reloc_cow_block(trans, root, buf, cow);
 488                 if (ret) {
 489                         btrfs_tree_unlock(cow);
 490                         free_extent_buffer(cow);
 491                         btrfs_abort_transaction(trans, ret);
 492                         return ret;
 493                 }
 494         }
 495
 496         if (buf == root->node) {
 497                 WARN_ON(parent && parent != buf);
 498                 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
 499                     btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
 500                         parent_start = buf->start;
 501
 502                 atomic_inc(&cow->refs);
 503                 ret = btrfs_tree_mod_log_insert_root(root->node, cow, true);
 504                 BUG_ON(ret < 0);
 505                 rcu_assign_pointer(root->node, cow);
 506
 507                 btrfs_free_tree_block(trans, root, buf, parent_start,
 508                                       last_ref);
 509                 free_extent_buffer(buf);
 510                 add_root_to_dirty_list(root);
 511         } else {
 512                 WARN_ON(trans->transid != btrfs_header_generation(parent));
 513                 btrfs_tree_mod_log_insert_key(parent, parent_slot,
 514                                               BTRFS_MOD_LOG_KEY_REPLACE, GFP_NOFS);
 515                 btrfs_set_node_blockptr(parent, parent_slot,
 516                                         cow->start);
 517                 btrfs_set_node_ptr_generation(parent, parent_slot,
 518                                               trans->transid);
 519                 btrfs_mark_buffer_dirty(parent);
 520                 if (last_ref) {
 521                         ret = btrfs_tree_mod_log_free_eb(buf);
 522                         if (ret) {
 523                                 btrfs_tree_unlock(cow);
 524                                 free_extent_buffer(cow);
 525                                 btrfs_abort_transaction(trans, ret);
 526                                 return ret;
 527                         }
 528                 }
 529                 btrfs_free_tree_block(trans, root, buf, parent_start,
 530                                       last_ref);
 531         }
 532         if (unlock_orig)
 533                 btrfs_tree_unlock(buf);
 534         free_extent_buffer_stale(buf);
 535         btrfs_mark_buffer_dirty(cow);
 536         *cow_ret = cow;
 537         return 0;
 538 }
 539
 540 static inline int should_cow_block(struct btrfs_trans_handle *trans,
 541                                    struct btrfs_root *root,
 542                                    struct extent_buffer *buf)
 543 {
 544         if (btrfs_is_testing(root->fs_info))
 545                 return 0;
 546
 547         /* Ensure we can see the FORCE_COW bit */
 548         smp_mb__before_atomic();
 549
 550         /*
 551          * We do not need to cow a block if
 552          * 1) this block is not created or changed in this transaction;
 553          * 2) this block does not belong to TREE_RELOC tree;
 554          * 3) the root is not forced COW.
 555          *
 556          * What is forced COW:
 557          *    when we create snapshot during committing the transaction,
 558          *    after we've finished copying src root, we must COW the shared
 559          *    block to ensure the metadata consistency.
 560          */
 561         if (btrfs_header_generation(buf) == trans->transid &&
 562             !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) &&
 563             !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
 564               btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) &&
 565             !test_bit(BTRFS_ROOT_FORCE_COW, &root->state))
 566                 return 0;
 567         return 1;
 568 }
 569
 570 /*
 571  * cows a single block, see __btrfs_cow_block for the real work.
 572  * This version of it has extra checks so that a block isn't COWed more than
 573  * once per transaction, as long as it hasn't been written yet
 574  */
 575 noinline int btrfs_cow_block(struct btrfs_trans_handle *trans,
 576                     struct btrfs_root *root, struct extent_buffer *buf,
 577                     struct extent_buffer *parent, int parent_slot,
 578                     struct extent_buffer **cow_ret,
 579                     enum btrfs_lock_nesting nest)
 580 {
 581         struct btrfs_fs_info *fs_info = root->fs_info;
 582         u64 search_start;
 583         int ret;
 584
 585         if (test_bit(BTRFS_ROOT_DELETING, &root->state))
 586                 btrfs_err(fs_info,
 587                         "COW'ing blocks on a fs root that's being dropped");
 588
 589         if (trans->transaction != fs_info->running_transaction)
 590                 WARN(1, KERN_CRIT "trans %llu running %llu\n",
 591                        trans->transid,
 592                        fs_info->running_transaction->transid);
 593
 594         if (trans->transid != fs_info->generation)
 595                 WARN(1, KERN_CRIT "trans %llu running %llu\n",
 596                        trans->transid, fs_info->generation);
 597
 598         if (!should_cow_block(trans, root, buf)) {
 599                 trans->dirty = true;
 600                 *cow_ret = buf;
 601                 return 0;
 602         }
 603
 604         search_start = buf->start & ~((u64)SZ_1G - 1);
 605
 606         /*
 607          * Before CoWing this block for later modification, check if it's
 608          * the subtree root and do the delayed subtree trace if needed.
 609          *
 610          * Also We don't care about the error, as it's handled internally.
 611          */
 612         btrfs_qgroup_trace_subtree_after_cow(trans, root, buf);
 613         ret = __btrfs_cow_block(trans, root, buf, parent,
 614                                  parent_slot, cow_ret, search_start, 0, nest);
 615
 616         trace_btrfs_cow_block(root, buf, *cow_ret);
 617
 618         return ret;
 619 }
 620 ALLOW_ERROR_INJECTION(btrfs_cow_block, ERRNO);
 621
 622 /*
 623  * helper function for defrag to decide if two blocks pointed to by a
 624  * node are actually close by
 625  */
 626 static int close_blocks(u64 blocknr, u64 other, u32 blocksize)
 627 {
 628         if (blocknr < other && other - (blocknr + blocksize) < 32768)
 629                 return 1;
 630         if (blocknr > other && blocknr - (other + blocksize) < 32768)
 631                 return 1;
 632         return 0;
 633 }
 634
 635 #ifdef __LITTLE_ENDIAN
 636
 637 /*
 638  * Compare two keys, on little-endian the disk order is same as CPU order and
 639  * we can avoid the conversion.
 640  */
 641 static int comp_keys(const struct btrfs_disk_key *disk_key,
 642                      const struct btrfs_key *k2)
 643 {
 644         const struct btrfs_key *k1 = (const struct btrfs_key *)disk_key;
 645
 646         return btrfs_comp_cpu_keys(k1, k2);
 647 }
 648
 649 #else
 650
 651 /*
 652  * compare two keys in a memcmp fashion
 653  */
 654 static int comp_keys(const struct btrfs_disk_key *disk,
 655                      const struct btrfs_key *k2)
 656 {
 657         struct btrfs_key k1;
 658
 659         btrfs_disk_key_to_cpu(&k1, disk);
 660
 661         return btrfs_comp_cpu_keys(&k1, k2);
 662 }
 663 #endif
 664
 665 /*
 666  * same as comp_keys only with two btrfs_key's
 667  */
 668 int __pure btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2)
 669 {
 670         if (k1->objectid > k2->objectid)
 671                 return 1;
 672         if (k1->objectid < k2->objectid)
 673                 return -1;
 674         if (k1->type > k2->type)
 675                 return 1;
 676         if (k1->type < k2->type)
 677                 return -1;
 678         if (k1->offset > k2->offset)
 679                 return 1;
 680         if (k1->offset < k2->offset)
 681                 return -1;
 682         return 0;
 683 }
 684
 685 /*
 686  * this is used by the defrag code to go through all the
 687  * leaves pointed to by a node and reallocate them so that
 688  * disk order is close to key order
 689  */
 690 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 691                        struct btrfs_root *root, struct extent_buffer *parent,
 692                        int start_slot, u64 *last_ret,
 693                        struct btrfs_key *progress)
 694 {
 695         struct btrfs_fs_info *fs_info = root->fs_info;
 696         struct extent_buffer *cur;
 697         u64 blocknr;
 698         u64 search_start = *last_ret;
 699         u64 last_block = 0;
 700         u64 other;
 701         u32 parent_nritems;
 702         int end_slot;
 703         int i;
 704         int err = 0;
 705         u32 blocksize;
 706         int progress_passed = 0;
 707         struct btrfs_disk_key disk_key;
 708
 709         WARN_ON(trans->transaction != fs_info->running_transaction);
 710         WARN_ON(trans->transid != fs_info->generation);
 711
 712         parent_nritems = btrfs_header_nritems(parent);
 713         blocksize = fs_info->nodesize;
 714         end_slot = parent_nritems - 1;
 715
 716         if (parent_nritems <= 1)
 717                 return 0;
 718
 719         for (i = start_slot; i <= end_slot; i++) {
 720                 int close = 1;
 721
 722                 btrfs_node_key(parent, &disk_key, i);
 723                 if (!progress_passed && comp_keys(&disk_key, progress) < 0)
 724                         continue;
 725
 726                 progress_passed = 1;
 727                 blocknr = btrfs_node_blockptr(parent, i);
 728                 if (last_block == 0)
 729                         last_block = blocknr;
 730
 731                 if (i > 0) {
 732                         other = btrfs_node_blockptr(parent, i - 1);
 733                         close = close_blocks(blocknr, other, blocksize);
 734                 }
 735                 if (!close && i < end_slot) {
 736                         other = btrfs_node_blockptr(parent, i + 1);
 737                         close = close_blocks(blocknr, other, blocksize);
 738                 }
 739                 if (close) {
 740                         last_block = blocknr;
 741                         continue;
 742                 }
 743
 744                 cur = btrfs_read_node_slot(parent, i);
 745                 if (IS_ERR(cur))
 746                         return PTR_ERR(cur);
 747                 if (search_start == 0)
 748                         search_start = last_block;
 749
 750                 btrfs_tree_lock(cur);
 751                 err = __btrfs_cow_block(trans, root, cur, parent, i,
 752                                         &cur, search_start,
 753                                         min(16 * blocksize,
 754                                             (end_slot - i) * blocksize),
 755                                         BTRFS_NESTING_COW);
 756                 if (err) {
 757                         btrfs_tree_unlock(cur);
 758                         free_extent_buffer(cur);
 759                         break;
 760                 }
 761                 search_start = cur->start;
 762                 last_block = cur->start;
 763                 *last_ret = search_start;
 764                 btrfs_tree_unlock(cur);
 765                 free_extent_buffer(cur);
 766         }
 767         return err;
 768 }
 769
 770 /*
 771  * search for key in the extent_buffer.  The items start at offset p,
 772  * and they are item_size apart.  There are 'max' items in p.
 773  *
 774  * the slot in the array is returned via slot, and it points to
 775  * the place where you would insert key if it is not found in
 776  * the array.
 777  *
 778  * slot may point to max if the key is bigger than all of the keys
 779  */
 780 static noinline int generic_bin_search(struct extent_buffer *eb,
 781                                        unsigned long p, int item_size,
 782                                        const struct btrfs_key *key,
 783                                        int max, int *slot)
 784 {
 785         int low = 0;
 786         int high = max;
 787         int ret;
 788         const int key_size = sizeof(struct btrfs_disk_key);
 789
 790         if (low > high) {
 791                 btrfs_err(eb->fs_info,
 792                  "%s: low (%d) > high (%d) eb %llu owner %llu level %d",
 793                           __func__, low, high, eb->start,
 794                           btrfs_header_owner(eb), btrfs_header_level(eb));
 795                 return -EINVAL;
 796         }
 797
 798         while (low < high) {
 799                 unsigned long oip;
 800                 unsigned long offset;
 801                 struct btrfs_disk_key *tmp;
 802                 struct btrfs_disk_key unaligned;
 803                 int mid;
 804
 805                 mid = (low + high) / 2;
 806                 offset = p + mid * item_size;
 807                 oip = offset_in_page(offset);
 808
 809                 if (oip + key_size <= PAGE_SIZE) {
 810                         const unsigned long idx = get_eb_page_index(offset);
 811                         char *kaddr = page_address(eb->pages[idx]);
 812
 813                         oip = get_eb_offset_in_page(eb, offset);
 814                         tmp = (struct btrfs_disk_key *)(kaddr + oip);
 815                 } else {
 816                         read_extent_buffer(eb, &unaligned, offset, key_size);
 817                         tmp = &unaligned;
 818                 }
 819
 820                 ret = comp_keys(tmp, key);
 821
 822                 if (ret < 0)
 823                         low = mid + 1;
 824                 else if (ret > 0)
 825                         high = mid;
 826                 else {
 827                         *slot = mid;
 828                         return 0;
 829                 }
 830         }
 831         *slot = low;
 832         return 1;
 833 }
 834
 835 /*
 836  * simple bin_search frontend that does the right thing for
 837  * leaves vs nodes
 838  */
 839 int btrfs_bin_search(struct extent_buffer *eb, const struct btrfs_key *key,
 840                      int *slot)
 841 {
 842         if (btrfs_header_level(eb) == 0)
 843                 return generic_bin_search(eb,
 844                                           offsetof(struct btrfs_leaf, items),
 845                                           sizeof(struct btrfs_item),
 846                                           key, btrfs_header_nritems(eb),
 847                                           slot);
 848         else
 849                 return generic_bin_search(eb,
 850                                           offsetof(struct btrfs_node, ptrs),
 851                                           sizeof(struct btrfs_key_ptr),
 852                                           key, btrfs_header_nritems(eb),
 853                                           slot);
 854 }
 855
 856 static void root_add_used(struct btrfs_root *root, u32 size)
 857 {
 858         spin_lock(&root->accounting_lock);
 859         btrfs_set_root_used(&root->root_item,
 860                             btrfs_root_used(&root->root_item) + size);
 861         spin_unlock(&root->accounting_lock);
 862 }
 863
 864 static void root_sub_used(struct btrfs_root *root, u32 size)
 865 {
 866         spin_lock(&root->accounting_lock);
 867         btrfs_set_root_used(&root->root_item,
 868                             btrfs_root_used(&root->root_item) - size);
 869         spin_unlock(&root->accounting_lock);
 870 }
 871
 872 /* given a node and slot number, this reads the blocks it points to.  The
 873  * extent buffer is returned with a reference taken (but unlocked).
 874  */
 875 struct extent_buffer *btrfs_read_node_slot(struct extent_buffer *parent,
 876                                            int slot)
 877 {
 878         int level = btrfs_header_level(parent);
 879         struct extent_buffer *eb;
 880         struct btrfs_key first_key;
 881
 882         if (slot < 0 || slot >= btrfs_header_nritems(parent))
 883                 return ERR_PTR(-ENOENT);
 884
 885         BUG_ON(level == 0);
 886
 887         btrfs_node_key_to_cpu(parent, &first_key, slot);
 888         eb = read_tree_block(parent->fs_info, btrfs_node_blockptr(parent, slot),
 889                              btrfs_header_owner(parent),
 890                              btrfs_node_ptr_generation(parent, slot),
 891                              level - 1, &first_key);
 892         if (!IS_ERR(eb) && !extent_buffer_uptodate(eb)) {
 893                 free_extent_buffer(eb);
 894                 eb = ERR_PTR(-EIO);
 895         }
 896
 897         return eb;
 898 }
 899
 900 /*
 901  * node level balancing, used to make sure nodes are in proper order for
 902  * item deletion.  We balance from the top down, so we have to make sure
 903  * that a deletion won't leave an node completely empty later on.
 904  */
 905 static noinline int balance_level(struct btrfs_trans_handle *trans,
 906                          struct btrfs_root *root,
 907                          struct btrfs_path *path, int level)
 908 {
 909         struct btrfs_fs_info *fs_info = root->fs_info;
 910         struct extent_buffer *right = NULL;
 911         struct extent_buffer *mid;
 912         struct extent_buffer *left = NULL;
 913         struct extent_buffer *parent = NULL;
 914         int ret = 0;
 915         int wret;
 916         int pslot;
 917         int orig_slot = path->slots[level];
 918         u64 orig_ptr;
 919
 920         ASSERT(level > 0);
 921
 922         mid = path->nodes[level];
 923
 924         WARN_ON(path->locks[level] != BTRFS_WRITE_LOCK);
 925         WARN_ON(btrfs_header_generation(mid) != trans->transid);
 926
 927         orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 928
 929         if (level < BTRFS_MAX_LEVEL - 1) {
 930                 parent = path->nodes[level + 1];
 931                 pslot = path->slots[level + 1];
 932         }
 933
 934         /*
 935          * deal with the case where there is only one pointer in the root
 936          * by promoting the node below to a root
 937          */
 938         if (!parent) {
 939                 struct extent_buffer *child;
 940
 941                 if (btrfs_header_nritems(mid) != 1)
 942                         return 0;
 943
 944                 /* promote the child to a root */
 945                 child = btrfs_read_node_slot(mid, 0);
 946                 if (IS_ERR(child)) {
 947                         ret = PTR_ERR(child);
 948                         btrfs_handle_fs_error(fs_info, ret, NULL);
 949                         goto enospc;
 950                 }
 951
 952                 btrfs_tree_lock(child);
 953                 ret = btrfs_cow_block(trans, root, child, mid, 0, &child,
 954                                       BTRFS_NESTING_COW);
 955                 if (ret) {
 956                         btrfs_tree_unlock(child);
 957                         free_extent_buffer(child);
 958                         goto enospc;
 959                 }
 960
 961                 ret = btrfs_tree_mod_log_insert_root(root->node, child, true);
 962                 BUG_ON(ret < 0);
 963                 rcu_assign_pointer(root->node, child);
 964
 965                 add_root_to_dirty_list(root);
 966                 btrfs_tree_unlock(child);
 967
 968                 path->locks[level] = 0;
 969                 path->nodes[level] = NULL;
 970                 btrfs_clean_tree_block(mid);
 971                 btrfs_tree_unlock(mid);
 972                 /* once for the path */
 973                 free_extent_buffer(mid);
 974
 975                 root_sub_used(root, mid->len);
 976                 btrfs_free_tree_block(trans, root, mid, 0, 1);
 977                 /* once for the root ptr */
 978                 free_extent_buffer_stale(mid);
 979                 return 0;
 980         }
 981         if (btrfs_header_nritems(mid) >
 982             BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 4)
 983                 return 0;
 984
 985         left = btrfs_read_node_slot(parent, pslot - 1);
 986         if (IS_ERR(left))
 987                 left = NULL;
 988
 989         if (left) {
 990                 __btrfs_tree_lock(left, BTRFS_NESTING_LEFT);
 991                 wret = btrfs_cow_block(trans, root, left,
 992                                        parent, pslot - 1, &left,
 993                                        BTRFS_NESTING_LEFT_COW);
 994                 if (wret) {
 995                         ret = wret;
 996                         goto enospc;
 997                 }
 998         }
 999
1000         right = btrfs_read_node_slot(parent, pslot + 1);
1001         if (IS_ERR(right))
1002                 right = NULL;
1003
1004         if (right) {
1005                 __btrfs_tree_lock(right, BTRFS_NESTING_RIGHT);
1006                 wret = btrfs_cow_block(trans, root, right,
1007                                        parent, pslot + 1, &right,
1008                                        BTRFS_NESTING_RIGHT_COW);
1009                 if (wret) {
1010                         ret = wret;
1011                         goto enospc;
1012                 }
1013         }
1014
1015         /* first, try to make some room in the middle buffer */
1016         if (left) {
1017                 orig_slot += btrfs_header_nritems(left);
1018                 wret = push_node_left(trans, left, mid, 1);
1019                 if (wret < 0)
1020                         ret = wret;
1021         }
1022
1023         /*
1024          * then try to empty the right most buffer into the middle
1025          */
1026         if (right) {
1027                 wret = push_node_left(trans, mid, right, 1);
1028                 if (wret < 0 && wret != -ENOSPC)
1029                         ret = wret;
1030                 if (btrfs_header_nritems(right) == 0) {
1031                         btrfs_clean_tree_block(right);
1032                         btrfs_tree_unlock(right);
1033                         del_ptr(root, path, level + 1, pslot + 1);
1034                         root_sub_used(root, right->len);
1035                         btrfs_free_tree_block(trans, root, right, 0, 1);
1036                         free_extent_buffer_stale(right);
1037                         right = NULL;
1038                 } else {
1039                         struct btrfs_disk_key right_key;
1040                         btrfs_node_key(right, &right_key, 0);
1041                         ret = btrfs_tree_mod_log_insert_key(parent, pslot + 1,
1042                                         BTRFS_MOD_LOG_KEY_REPLACE, GFP_NOFS);
1043                         BUG_ON(ret < 0);
1044                         btrfs_set_node_key(parent, &right_key, pslot + 1);
1045                         btrfs_mark_buffer_dirty(parent);
1046                 }
1047         }
1048         if (btrfs_header_nritems(mid) == 1) {
1049                 /*
1050                  * we're not allowed to leave a node with one item in the
1051                  * tree during a delete.  A deletion from lower in the tree
1052                  * could try to delete the only pointer in this node.
1053                  * So, pull some keys from the left.
1054                  * There has to be a left pointer at this point because
1055                  * otherwise we would have pulled some pointers from the
1056                  * right
1057                  */
1058                 if (!left) {
1059                         ret = -EROFS;
1060                         btrfs_handle_fs_error(fs_info, ret, NULL);
1061                         goto enospc;
1062                 }
1063                 wret = balance_node_right(trans, mid, left);
1064                 if (wret < 0) {
1065                         ret = wret;
1066                         goto enospc;
1067                 }
1068                 if (wret == 1) {
1069                         wret = push_node_left(trans, left, mid, 1);
1070                         if (wret < 0)
1071                                 ret = wret;
1072                 }
1073                 BUG_ON(wret == 1);
1074         }
1075         if (btrfs_header_nritems(mid) == 0) {
1076                 btrfs_clean_tree_block(mid);
1077                 btrfs_tree_unlock(mid);
1078                 del_ptr(root, path, level + 1, pslot);
1079                 root_sub_used(root, mid->len);
1080                 btrfs_free_tree_block(trans, root, mid, 0, 1);
1081                 free_extent_buffer_stale(mid);
1082                 mid = NULL;
1083         } else {
1084                 /* update the parent key to reflect our changes */
1085                 struct btrfs_disk_key mid_key;
1086                 btrfs_node_key(mid, &mid_key, 0);
1087                 ret = btrfs_tree_mod_log_insert_key(parent, pslot,
1088                                 BTRFS_MOD_LOG_KEY_REPLACE, GFP_NOFS);
1089                 BUG_ON(ret < 0);
1090                 btrfs_set_node_key(parent, &mid_key, pslot);
1091                 btrfs_mark_buffer_dirty(parent);
1092         }
1093
1094         /* update the path */
1095         if (left) {
1096                 if (btrfs_header_nritems(left) > orig_slot) {
1097                         atomic_inc(&left->refs);
1098                         /* left was locked after cow */
1099                         path->nodes[level] = left;
1100                         path->slots[level + 1] -= 1;
1101                         path->slots[level] = orig_slot;
1102                         if (mid) {
1103                                 btrfs_tree_unlock(mid);
1104                                 free_extent_buffer(mid);
1105                         }
1106                 } else {
1107                         orig_slot -= btrfs_header_nritems(left);
1108                         path->slots[level] = orig_slot;
1109                 }
1110         }
1111         /* double check we haven't messed things up */
1112         if (orig_ptr !=
1113             btrfs_node_blockptr(path->nodes[level], path->slots[level]))
1114                 BUG();
1115 enospc:
1116         if (right) {
1117                 btrfs_tree_unlock(right);
1118                 free_extent_buffer(right);
1119         }
1120         if (left) {
1121                 if (path->nodes[level] != left)
1122                         btrfs_tree_unlock(left);
1123                 free_extent_buffer(left);
1124         }
1125         return ret;
1126 }
1127
1128 /* Node balancing for insertion.  Here we only split or push nodes around
1129  * when they are completely full.  This is also done top down, so we
1130  * have to be pessimistic.
1131  */
1132 static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
1133                                           struct btrfs_root *root,
1134                                           struct btrfs_path *path, int level)
1135 {
1136         struct btrfs_fs_info *fs_info = root->fs_info;
1137         struct extent_buffer *right = NULL;
1138         struct extent_buffer *mid;
1139         struct extent_buffer *left = NULL;
1140         struct extent_buffer *parent = NULL;
1141         int ret = 0;
1142         int wret;
1143         int pslot;
1144         int orig_slot = path->slots[level];
1145
1146         if (level == 0)
1147                 return 1;
1148
1149         mid = path->nodes[level];
1150         WARN_ON(btrfs_header_generation(mid) != trans->transid);
1151
1152         if (level < BTRFS_MAX_LEVEL - 1) {
1153                 parent = path->nodes[level + 1];
1154                 pslot = path->slots[level + 1];
1155         }
1156
1157         if (!parent)
1158                 return 1;
1159
1160         left = btrfs_read_node_slot(parent, pslot - 1);
1161         if (IS_ERR(left))
1162                 left = NULL;
1163
1164         /* first, try to make some room in the middle buffer */
1165         if (left) {
1166                 u32 left_nr;
1167
1168                 __btrfs_tree_lock(left, BTRFS_NESTING_LEFT);
1169
1170                 left_nr = btrfs_header_nritems(left);
1171                 if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
1172                         wret = 1;
1173                 } else {
1174                         ret = btrfs_cow_block(trans, root, left, parent,
1175                                               pslot - 1, &left,
1176                                               BTRFS_NESTING_LEFT_COW);
1177                         if (ret)
1178                                 wret = 1;
1179                         else {
1180                                 wret = push_node_left(trans, left, mid, 0);
1181                         }
1182                 }
1183                 if (wret < 0)
1184                         ret = wret;
1185                 if (wret == 0) {
1186                         struct btrfs_disk_key disk_key;
1187                         orig_slot += left_nr;
1188                         btrfs_node_key(mid, &disk_key, 0);
1189                         ret = btrfs_tree_mod_log_insert_key(parent, pslot,
1190                                         BTRFS_MOD_LOG_KEY_REPLACE, GFP_NOFS);
1191                         BUG_ON(ret < 0);
1192                         btrfs_set_node_key(parent, &disk_key, pslot);
1193                         btrfs_mark_buffer_dirty(parent);
1194                         if (btrfs_header_nritems(left) > orig_slot) {
1195                                 path->nodes[level] = left;
1196                                 path->slots[level + 1] -= 1;
1197                                 path->slots[level] = orig_slot;
1198                                 btrfs_tree_unlock(mid);
1199                                 free_extent_buffer(mid);
1200                         } else {
1201                                 orig_slot -=
1202                                         btrfs_header_nritems(left);
1203                                 path->slots[level] = orig_slot;
1204                                 btrfs_tree_unlock(left);
1205                                 free_extent_buffer(left);
1206                         }
1207                         return 0;
1208                 }
1209                 btrfs_tree_unlock(left);
1210                 free_extent_buffer(left);
1211         }
1212         right = btrfs_read_node_slot(parent, pslot + 1);
1213         if (IS_ERR(right))
1214                 right = NULL;
1215
1216         /*
1217          * then try to empty the right most buffer into the middle
1218          */
1219         if (right) {
1220                 u32 right_nr;
1221
1222                 __btrfs_tree_lock(right, BTRFS_NESTING_RIGHT);
1223
1224                 right_nr = btrfs_header_nritems(right);
1225                 if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
1226                         wret = 1;
1227                 } else {
1228                         ret = btrfs_cow_block(trans, root, right,
1229                                               parent, pslot + 1,
1230                                               &right, BTRFS_NESTING_RIGHT_COW);
1231                         if (ret)
1232                                 wret = 1;
1233                         else {
1234                                 wret = balance_node_right(trans, right, mid);
1235                         }
1236                 }
1237                 if (wret < 0)
1238                         ret = wret;
1239                 if (wret == 0) {
1240                         struct btrfs_disk_key disk_key;
1241
1242                         btrfs_node_key(right, &disk_key, 0);
1243                         ret = btrfs_tree_mod_log_insert_key(parent, pslot + 1,
1244                                         BTRFS_MOD_LOG_KEY_REPLACE, GFP_NOFS);
1245                         BUG_ON(ret < 0);
1246                         btrfs_set_node_key(parent, &disk_key, pslot + 1);
1247                         btrfs_mark_buffer_dirty(parent);
1248
1249                         if (btrfs_header_nritems(mid) <= orig_slot) {
1250                                 path->nodes[level] = right;
1251                                 path->slots[level + 1] += 1;
1252                                 path->slots[level] = orig_slot -
1253                                         btrfs_header_nritems(mid);
1254                                 btrfs_tree_unlock(mid);
1255                                 free_extent_buffer(mid);
1256                         } else {
1257                                 btrfs_tree_unlock(right);
1258                                 free_extent_buffer(right);
1259                         }
1260                         return 0;
1261                 }
1262                 btrfs_tree_unlock(right);
1263                 free_extent_buffer(right);
1264         }
1265         return 1;
1266 }
1267
1268 /*
1269  * readahead one full node of leaves, finding things that are close
1270  * to the block in 'slot', and triggering ra on them.
1271  */
1272 static void reada_for_search(struct btrfs_fs_info *fs_info,
1273                              struct btrfs_path *path,
1274                              int level, int slot, u64 objectid)
1275 {
1276         struct extent_buffer *node;
1277         struct btrfs_disk_key disk_key;
1278         u32 nritems;
1279         u64 search;
1280         u64 target;
1281         u64 nread = 0;
1282         u64 nread_max;
1283         struct extent_buffer *eb;
1284         u32 nr;
1285         u32 blocksize;
1286         u32 nscan = 0;
1287
1288         if (level != 1 && path->reada != READA_FORWARD_ALWAYS)
1289                 return;
1290
1291         if (!path->nodes[level])
1292                 return;
1293
1294         node = path->nodes[level];
1295
1296         /*
1297          * Since the time between visiting leaves is much shorter than the time
1298          * between visiting nodes, limit read ahead of nodes to 1, to avoid too
1299          * much IO at once (possibly random).
1300          */
1301         if (path->reada == READA_FORWARD_ALWAYS) {
1302                 if (level > 1)
1303                         nread_max = node->fs_info->nodesize;
1304                 else
1305                         nread_max = SZ_128K;
1306         } else {
1307                 nread_max = SZ_64K;
1308         }
1309
1310         search = btrfs_node_blockptr(node, slot);
1311         blocksize = fs_info->nodesize;
1312         eb = find_extent_buffer(fs_info, search);
1313         if (eb) {
1314                 free_extent_buffer(eb);
1315                 return;
1316         }
1317
1318         target = search;
1319
1320         nritems = btrfs_header_nritems(node);
1321         nr = slot;
1322
1323         while (1) {
1324                 if (path->reada == READA_BACK) {
1325                         if (nr == 0)
1326                                 break;
1327                         nr--;
1328                 } else if (path->reada == READA_FORWARD ||
1329                            path->reada == READA_FORWARD_ALWAYS) {
1330                         nr++;
1331                         if (nr >= nritems)
1332                                 break;
1333                 }
1334                 if (path->reada == READA_BACK && objectid) {
1335                         btrfs_node_key(node, &disk_key, nr);
1336                         if (btrfs_disk_key_objectid(&disk_key) != objectid)
1337                                 break;
1338                 }
1339                 search = btrfs_node_blockptr(node, nr);
1340                 if (path->reada == READA_FORWARD_ALWAYS ||
1341                     (search <= target && target - search <= 65536) ||
1342                     (search > target && search - target <= 65536)) {
1343                         btrfs_readahead_node_child(node, nr);
1344                         nread += blocksize;
1345                 }
1346                 nscan++;
1347                 if (nread > nread_max || nscan > 32)
1348                         break;
1349         }
1350 }
1351
1352 static noinline void reada_for_balance(struct btrfs_path *path, int level)
1353 {
1354         struct extent_buffer *parent;
1355         int slot;
1356         int nritems;
1357
1358         parent = path->nodes[level + 1];
1359         if (!parent)
1360                 return;
1361
1362         nritems = btrfs_header_nritems(parent);
1363         slot = path->slots[level + 1];
1364
1365         if (slot > 0)
1366                 btrfs_readahead_node_child(parent, slot - 1);
1367         if (slot + 1 < nritems)
1368                 btrfs_readahead_node_child(parent, slot + 1);
1369 }
1370
1371
1372 /*
1373  * when we walk down the tree, it is usually safe to unlock the higher layers
1374  * in the tree.  The exceptions are when our path goes through slot 0, because
1375  * operations on the tree might require changing key pointers higher up in the
1376  * tree.
1377  *
1378  * callers might also have set path->keep_locks, which tells this code to keep
1379  * the lock if the path points to the last slot in the block.  This is part of
1380  * walking through the tree, and selecting the next slot in the higher block.
1381  *
1382  * lowest_unlock sets the lowest level in the tree we're allowed to unlock.  so
1383  * if lowest_unlock is 1, level 0 won't be unlocked
1384  */
1385 static noinline void unlock_up(struct btrfs_path *path, int level,
1386                                int lowest_unlock, int min_write_lock_level,
1387                                int *write_lock_level)
1388 {
1389         int i;
1390         int skip_level = level;
1391         int no_skips = 0;
1392         struct extent_buffer *t;
1393
1394         for (i = level; i < BTRFS_MAX_LEVEL; i++) {
1395                 if (!path->nodes[i])
1396                         break;
1397                 if (!path->locks[i])
1398                         break;
1399                 if (!no_skips && path->slots[i] == 0) {
1400                         skip_level = i + 1;
1401                         continue;
1402                 }
1403                 if (!no_skips && path->keep_locks) {
1404                         u32 nritems;
1405                         t = path->nodes[i];
1406                         nritems = btrfs_header_nritems(t);
1407                         if (nritems < 1 || path->slots[i] >= nritems - 1) {
1408                                 skip_level = i + 1;
1409                                 continue;
1410                         }
1411                 }
1412                 if (skip_level < i && i >= lowest_unlock)
1413                         no_skips = 1;
1414
1415                 t = path->nodes[i];
1416                 if (i >= lowest_unlock && i > skip_level) {
1417                         btrfs_tree_unlock_rw(t, path->locks[i]);
1418                         path->locks[i] = 0;
1419                         if (write_lock_level &&
1420                             i > min_write_lock_level &&
1421                             i <= *write_lock_level) {
1422                                 *write_lock_level = i - 1;
1423                         }
1424                 }
1425         }
1426 }
1427
1428 /*
1429  * helper function for btrfs_search_slot.  The goal is to find a block
1430  * in cache without setting the path to blocking.  If we find the block
1431  * we return zero and the path is unchanged.
1432  *
1433  * If we can't find the block, we set the path blocking and do some
1434  * reada.  -EAGAIN is returned and the search must be repeated.
1435  */
1436 static int
1437 read_block_for_search(struct btrfs_root *root, struct btrfs_path *p,
1438                       struct extent_buffer **eb_ret, int level, int slot,
1439                       const struct btrfs_key *key)
1440 {
1441         struct btrfs_fs_info *fs_info = root->fs_info;
1442         u64 blocknr;
1443         u64 gen;
1444         struct extent_buffer *tmp;
1445         struct btrfs_key first_key;
1446         int ret;
1447         int parent_level;
1448
1449         blocknr = btrfs_node_blockptr(*eb_ret, slot);
1450         gen = btrfs_node_ptr_generation(*eb_ret, slot);
1451         parent_level = btrfs_header_level(*eb_ret);
1452         btrfs_node_key_to_cpu(*eb_ret, &first_key, slot);
1453
1454         tmp = find_extent_buffer(fs_info, blocknr);
1455         if (tmp) {
1456                 if (p->reada == READA_FORWARD_ALWAYS)
1457                         reada_for_search(fs_info, p, level, slot, key->objectid);
1458
1459                 /* first we do an atomic uptodate check */
1460                 if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) {
1461                         /*
1462                          * Do extra check for first_key, eb can be stale due to
1463                          * being cached, read from scrub, or have multiple
1464                          * parents (shared tree blocks).
1465                          */
1466                         if (btrfs_verify_level_key(tmp,
1467                                         parent_level - 1, &first_key, gen)) {
1468                                 free_extent_buffer(tmp);
1469                                 return -EUCLEAN;
1470                         }
1471                         *eb_ret = tmp;
1472                         return 0;
1473                 }
1474
1475                 /* now we're allowed to do a blocking uptodate check */
1476                 ret = btrfs_read_buffer(tmp, gen, parent_level - 1, &first_key);
1477                 if (!ret) {
1478                         *eb_ret = tmp;
1479                         return 0;
1480                 }
1481                 free_extent_buffer(tmp);
1482                 btrfs_release_path(p);
1483                 return -EIO;
1484         }
1485
1486         /*
1487          * reduce lock contention at high levels
1488          * of the btree by dropping locks before
1489          * we read.  Don't release the lock on the current
1490          * level because we need to walk this node to figure
1491          * out which blocks to read.
1492          */
1493         btrfs_unlock_up_safe(p, level + 1);
1494
1495         if (p->reada != READA_NONE)
1496                 reada_for_search(fs_info, p, level, slot, key->objectid);
1497
1498         ret = -EAGAIN;
1499         tmp = read_tree_block(fs_info, blocknr, root->root_key.objectid,
1500                               gen, parent_level - 1, &first_key);
1501         if (!IS_ERR(tmp)) {
1502                 /*
1503                  * If the read above didn't mark this buffer up to date,
1504                  * it will never end up being up to date.  Set ret to EIO now
1505                  * and give up so that our caller doesn't loop forever
1506                  * on our EAGAINs.
1507                  */
1508                 if (!extent_buffer_uptodate(tmp))
1509                         ret = -EIO;
1510                 free_extent_buffer(tmp);
1511         } else {
1512                 ret = PTR_ERR(tmp);
1513         }
1514
1515         btrfs_release_path(p);
1516         return ret;
1517 }
1518
1519 /*
1520  * helper function for btrfs_search_slot.  This does all of the checks
1521  * for node-level blocks and does any balancing required based on
1522  * the ins_len.
1523  *
1524  * If no extra work was required, zero is returned.  If we had to
1525  * drop the path, -EAGAIN is returned and btrfs_search_slot must
1526  * start over
1527  */
1528 static int
1529 setup_nodes_for_search(struct btrfs_trans_handle *trans,
1530                        struct btrfs_root *root, struct btrfs_path *p,
1531                        struct extent_buffer *b, int level, int ins_len,
1532                        int *write_lock_level)
1533 {
1534         struct btrfs_fs_info *fs_info = root->fs_info;
1535         int ret = 0;
1536
1537         if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >=
1538             BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) {
1539
1540                 if (*write_lock_level < level + 1) {
1541                         *write_lock_level = level + 1;
1542                         btrfs_release_path(p);
1543                         return -EAGAIN;
1544                 }
1545
1546                 reada_for_balance(p, level);
1547                 ret = split_node(trans, root, p, level);
1548
1549                 b = p->nodes[level];
1550         } else if (ins_len < 0 && btrfs_header_nritems(b) <
1551                    BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 2) {
1552
1553                 if (*write_lock_level < level + 1) {
1554                         *write_lock_level = level + 1;
1555                         btrfs_release_path(p);
1556                         return -EAGAIN;
1557                 }
1558
1559                 reada_for_balance(p, level);
1560                 ret = balance_level(trans, root, p, level);
1561                 if (ret)
1562                         return ret;
1563
1564                 b = p->nodes[level];
1565                 if (!b) {
1566                         btrfs_release_path(p);
1567                         return -EAGAIN;
1568                 }
1569                 BUG_ON(btrfs_header_nritems(b) == 1);
1570         }
1571         return ret;
1572 }
1573
1574 int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path,
1575                 u64 iobjectid, u64 ioff, u8 key_type,
1576                 struct btrfs_key *found_key)
1577 {
1578         int ret;
1579         struct btrfs_key key;
1580         struct extent_buffer *eb;
1581
1582         ASSERT(path);
1583         ASSERT(found_key);
1584
1585         key.type = key_type;
1586         key.objectid = iobjectid;
1587         key.offset = ioff;
1588
1589         ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1590         if (ret < 0)
1591                 return ret;
1592
1593         eb = path->nodes[0];
1594         if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
1595                 ret = btrfs_next_leaf(fs_root, path);
1596                 if (ret)
1597                         return ret;
1598                 eb = path->nodes[0];
1599         }
1600
1601         btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
1602         if (found_key->type != key.type ||
1603                         found_key->objectid != key.objectid)
1604                 return 1;
1605
1606         return 0;
1607 }
1608
1609 static struct extent_buffer *btrfs_search_slot_get_root(struct btrfs_root *root,
1610                                                         struct btrfs_path *p,
1611                                                         int write_lock_level)
1612 {
1613         struct btrfs_fs_info *fs_info = root->fs_info;
1614         struct extent_buffer *b;
1615         int root_lock;
1616         int level = 0;
1617
1618         /* We try very hard to do read locks on the root */
1619         root_lock = BTRFS_READ_LOCK;
1620
1621         if (p->search_commit_root) {
1622                 /*
1623                  * The commit roots are read only so we always do read locks,
1624                  * and we always must hold the commit_root_sem when doing
1625                  * searches on them, the only exception is send where we don't
1626                  * want to block transaction commits for a long time, so
1627                  * we need to clone the commit root in order to avoid races
1628                  * with transaction commits that create a snapshot of one of
1629                  * the roots used by a send operation.
1630                  */
1631                 if (p->need_commit_sem) {
1632                         down_read(&fs_info->commit_root_sem);
1633                         b = btrfs_clone_extent_buffer(root->commit_root);
1634                         up_read(&fs_info->commit_root_sem);
1635                         if (!b)
1636                                 return ERR_PTR(-ENOMEM);
1637
1638                 } else {
1639                         b = root->commit_root;
1640                         atomic_inc(&b->refs);
1641                 }
1642                 level = btrfs_header_level(b);
1643                 /*
1644                  * Ensure that all callers have set skip_locking when
1645                  * p->search_commit_root = 1.
1646                  */
1647                 ASSERT(p->skip_locking == 1);
1648
1649                 goto out;
1650         }
1651
1652         if (p->skip_locking) {
1653                 b = btrfs_root_node(root);
1654                 level = btrfs_header_level(b);
1655                 goto out;
1656         }
1657
1658         /*
1659          * If the level is set to maximum, we can skip trying to get the read
1660          * lock.
1661          */
1662         if (write_lock_level < BTRFS_MAX_LEVEL) {
1663                 /*
1664                  * We don't know the level of the root node until we actually
1665                  * have it read locked
1666                  */
1667                 b = btrfs_read_lock_root_node(root);
1668                 level = btrfs_header_level(b);
1669                 if (level > write_lock_level)
1670                         goto out;
1671
1672                 /* Whoops, must trade for write lock */
1673                 btrfs_tree_read_unlock(b);
1674                 free_extent_buffer(b);
1675         }
1676
1677         b = btrfs_lock_root_node(root);
1678         root_lock = BTRFS_WRITE_LOCK;
1679
1680         /* The level might have changed, check again */
1681         level = btrfs_header_level(b);
1682
1683 out:
1684         p->nodes[level] = b;
1685         if (!p->skip_locking)
1686                 p->locks[level] = root_lock;
1687         /*
1688          * Callers are responsible for dropping b's references.
1689          */
1690         return b;
1691 }
1692
1693
1694 /*
1695  * btrfs_search_slot - look for a key in a tree and perform necessary
1696  * modifications to preserve tree invariants.
1697  *
1698  * @trans:      Handle of transaction, used when modifying the tree
1699  * @p:          Holds all btree nodes along the search path
1700  * @root:       The root node of the tree
1701  * @key:        The key we are looking for
1702  * @ins_len:    Indicates purpose of search:
1703  *              >0  for inserts it's size of item inserted (*)
1704  *              <0  for deletions
1705  *               0  for plain searches, not modifying the tree
1706  *
1707  *              (*) If size of item inserted doesn't include
1708  *              sizeof(struct btrfs_item), then p->search_for_extension must
1709  *              be set.
1710  * @cow:        boolean should CoW operations be performed. Must always be 1
1711  *              when modifying the tree.
1712  *
1713  * If @ins_len > 0, nodes and leaves will be split as we walk down the tree.
1714  * If @ins_len < 0, nodes will be merged as we walk down the tree (if possible)
1715  *
1716  * If @key is found, 0 is returned and you can find the item in the leaf level
1717  * of the path (level 0)
1718  *
1719  * If @key isn't found, 1 is returned and the leaf level of the path (level 0)
1720  * points to the slot where it should be inserted
1721  *
1722  * If an error is encountered while searching the tree a negative error number
1723  * is returned
1724  */
1725 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1726                       const struct btrfs_key *key, struct btrfs_path *p,
1727                       int ins_len, int cow)
1728 {
1729         struct extent_buffer *b;
1730         int slot;
1731         int ret;
1732         int err;
1733         int level;
1734         int lowest_unlock = 1;
1735         /* everything at write_lock_level or lower must be write locked */
1736         int write_lock_level = 0;
1737         u8 lowest_level = 0;
1738         int min_write_lock_level;
1739         int prev_cmp;
1740
1741         lowest_level = p->lowest_level;
1742         WARN_ON(lowest_level && ins_len > 0);
1743         WARN_ON(p->nodes[0] != NULL);
1744         BUG_ON(!cow && ins_len);
1745
1746         if (ins_len < 0) {
1747                 lowest_unlock = 2;
1748
1749                 /* when we are removing items, we might have to go up to level
1750                  * two as we update tree pointers  Make sure we keep write
1751                  * for those levels as well
1752                  */
1753                 write_lock_level = 2;
1754         } else if (ins_len > 0) {
1755                 /*
1756                  * for inserting items, make sure we have a write lock on
1757                  * level 1 so we can update keys
1758                  */
1759                 write_lock_level = 1;
1760         }
1761
1762         if (!cow)
1763                 write_lock_level = -1;
1764
1765         if (cow && (p->keep_locks || p->lowest_level))
1766                 write_lock_level = BTRFS_MAX_LEVEL;
1767
1768         min_write_lock_level = write_lock_level;
1769
1770 again:
1771         prev_cmp = -1;
1772         b = btrfs_search_slot_get_root(root, p, write_lock_level);
1773         if (IS_ERR(b)) {
1774                 ret = PTR_ERR(b);
1775                 goto done;
1776         }
1777
1778         while (b) {
1779                 int dec = 0;
1780
1781                 level = btrfs_header_level(b);
1782
1783                 if (cow) {
1784                         bool last_level = (level == (BTRFS_MAX_LEVEL - 1));
1785
1786                         /*
1787                          * if we don't really need to cow this block
1788                          * then we don't want to set the path blocking,
1789                          * so we test it here
1790                          */
1791                         if (!should_cow_block(trans, root, b)) {
1792                                 trans->dirty = true;
1793                                 goto cow_done;
1794                         }
1795
1796                         /*
1797                          * must have write locks on this node and the
1798                          * parent
1799                          */
1800                         if (level > write_lock_level ||
1801                             (level + 1 > write_lock_level &&
1802                             level + 1 < BTRFS_MAX_LEVEL &&
1803                             p->nodes[level + 1])) {
1804                                 write_lock_level = level + 1;
1805                                 btrfs_release_path(p);
1806                                 goto again;
1807                         }
1808
1809                         if (last_level)
1810                                 err = btrfs_cow_block(trans, root, b, NULL, 0,
1811                                                       &b,
1812                                                       BTRFS_NESTING_COW);
1813                         else
1814                                 err = btrfs_cow_block(trans, root, b,
1815                                                       p->nodes[level + 1],
1816                                                       p->slots[level + 1], &b,
1817                                                       BTRFS_NESTING_COW);
1818                         if (err) {
1819                                 ret = err;
1820                                 goto done;
1821                         }
1822                 }
1823 cow_done:
1824                 p->nodes[level] = b;
1825                 /*
1826                  * Leave path with blocking locks to avoid massive
1827                  * lock context switch, this is made on purpose.
1828                  */
1829
1830                 /*
1831                  * we have a lock on b and as long as we aren't changing
1832                  * the tree, there is no way to for the items in b to change.
1833                  * It is safe to drop the lock on our parent before we
1834                  * go through the expensive btree search on b.
1835                  *
1836                  * If we're inserting or deleting (ins_len != 0), then we might
1837                  * be changing slot zero, which may require changing the parent.
1838                  * So, we can't drop the lock until after we know which slot
1839                  * we're operating on.
1840                  */
1841                 if (!ins_len && !p->keep_locks) {
1842                         int u = level + 1;
1843
1844                         if (u < BTRFS_MAX_LEVEL && p->locks[u]) {
1845                                 btrfs_tree_unlock_rw(p->nodes[u], p->locks[u]);
1846                                 p->locks[u] = 0;
1847                         }
1848                 }
1849
1850                 /*
1851                  * If btrfs_bin_search returns an exact match (prev_cmp == 0)
1852                  * we can safely assume the target key will always be in slot 0
1853                  * on lower levels due to the invariants BTRFS' btree provides,
1854                  * namely that a btrfs_key_ptr entry always points to the
1855                  * lowest key in the child node, thus we can skip searching
1856                  * lower levels
1857                  */
1858                 if (prev_cmp == 0) {
1859                         slot = 0;
1860                         ret = 0;
1861                 } else {
1862                         ret = btrfs_bin_search(b, key, &slot);
1863                         prev_cmp = ret;
1864                         if (ret < 0)
1865                                 goto done;
1866                 }
1867
1868                 if (level == 0) {
1869                         p->slots[level] = slot;
1870                         /*
1871                          * Item key already exists. In this case, if we are
1872                          * allowed to insert the item (for example, in dir_item
1873                          * case, item key collision is allowed), it will be
1874                          * merged with the original item. Only the item size
1875                          * grows, no new btrfs item will be added. If
1876                          * search_for_extension is not set, ins_len already
1877                          * accounts the size btrfs_item, deduct it here so leaf
1878                          * space check will be correct.
1879                          */
1880                         if (ret == 0 && ins_len > 0 && !p->search_for_extension) {
1881                                 ASSERT(ins_len >= sizeof(struct btrfs_item));
1882                                 ins_len -= sizeof(struct btrfs_item);
1883                         }
1884                         if (ins_len > 0 &&
1885                             btrfs_leaf_free_space(b) < ins_len) {
1886                                 if (write_lock_level < 1) {
1887                                         write_lock_level = 1;
1888                                         btrfs_release_path(p);
1889                                         goto again;
1890                                 }
1891
1892                                 err = split_leaf(trans, root, key,
1893                                                  p, ins_len, ret == 0);
1894
1895                                 BUG_ON(err > 0);
1896                                 if (err) {
1897                                         ret = err;
1898                                         goto done;
1899                                 }
1900                         }
1901                         if (!p->search_for_split)
1902                                 unlock_up(p, level, lowest_unlock,
1903                                           min_write_lock_level, NULL);
1904                         goto done;
1905                 }
1906                 if (ret && slot > 0) {
1907                         dec = 1;
1908                         slot--;
1909                 }
1910                 p->slots[level] = slot;
1911                 err = setup_nodes_for_search(trans, root, p, b, level, ins_len,
1912                                              &write_lock_level);
1913                 if (err == -EAGAIN)
1914                         goto again;
1915                 if (err) {
1916                         ret = err;
1917                         goto done;
1918                 }
1919                 b = p->nodes[level];
1920                 slot = p->slots[level];
1921
1922                 /*
1923                  * Slot 0 is special, if we change the key we have to update
1924                  * the parent pointer which means we must have a write lock on
1925                  * the parent
1926                  */
1927                 if (slot == 0 && ins_len && write_lock_level < level + 1) {
1928                         write_lock_level = level + 1;
1929                         btrfs_release_path(p);
1930                         goto again;
1931                 }
1932
1933                 unlock_up(p, level, lowest_unlock, min_write_lock_level,
1934                           &write_lock_level);
1935
1936                 if (level == lowest_level) {
1937                         if (dec)
1938                                 p->slots[level]++;
1939                         goto done;
1940                 }
1941
1942                 err = read_block_for_search(root, p, &b, level, slot, key);
1943                 if (err == -EAGAIN)
1944                         goto again;
1945                 if (err) {
1946                         ret = err;
1947                         goto done;
1948                 }
1949
1950                 if (!p->skip_locking) {
1951                         level = btrfs_header_level(b);
1952                         if (level <= write_lock_level) {
1953                                 btrfs_tree_lock(b);
1954                                 p->locks[level] = BTRFS_WRITE_LOCK;
1955                         } else {
1956                                 btrfs_tree_read_lock(b);
1957                                 p->locks[level] = BTRFS_READ_LOCK;
1958                         }
1959                         p->nodes[level] = b;
1960                 }
1961         }
1962         ret = 1;
1963 done:
1964         if (ret < 0 && !p->skip_release_on_error)
1965                 btrfs_release_path(p);
1966         return ret;
1967 }
1968 ALLOW_ERROR_INJECTION(btrfs_search_slot, ERRNO);
1969
1970 /*
1971  * Like btrfs_search_slot, this looks for a key in the given tree. It uses the
1972  * current state of the tree together with the operations recorded in the tree
1973  * modification log to search for the key in a previous version of this tree, as
1974  * denoted by the time_seq parameter.
1975  *
1976  * Naturally, there is no support for insert, delete or cow operations.
1977  *
1978  * The resulting path and return value will be set up as if we called
1979  * btrfs_search_slot at that point in time with ins_len and cow both set to 0.
1980  */
1981 int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key,
1982                           struct btrfs_path *p, u64 time_seq)
1983 {
1984         struct btrfs_fs_info *fs_info = root->fs_info;
1985         struct extent_buffer *b;
1986         int slot;
1987         int ret;
1988         int err;
1989         int level;
1990         int lowest_unlock = 1;
1991         u8 lowest_level = 0;
1992
1993         lowest_level = p->lowest_level;
1994         WARN_ON(p->nodes[0] != NULL);
1995
1996         if (p->search_commit_root) {
1997                 BUG_ON(time_seq);
1998                 return btrfs_search_slot(NULL, root, key, p, 0, 0);
1999         }
2000
2001 again:
2002         b = btrfs_get_old_root(root, time_seq);
2003         if (!b) {
2004                 ret = -EIO;
2005                 goto done;
2006         }
2007         level = btrfs_header_level(b);
2008         p->locks[level] = BTRFS_READ_LOCK;
2009
2010         while (b) {
2011                 int dec = 0;
2012
2013                 level = btrfs_header_level(b);
2014                 p->nodes[level] = b;
2015
2016                 /*
2017                  * we have a lock on b and as long as we aren't changing
2018                  * the tree, there is no way to for the items in b to change.
2019                  * It is safe to drop the lock on our parent before we
2020                  * go through the expensive btree search on b.
2021                  */
2022                 btrfs_unlock_up_safe(p, level + 1);
2023
2024                 ret = btrfs_bin_search(b, key, &slot);
2025                 if (ret < 0)
2026                         goto done;
2027
2028                 if (level == 0) {
2029                         p->slots[level] = slot;
2030                         unlock_up(p, level, lowest_unlock, 0, NULL);
2031                         goto done;
2032                 }
2033
2034                 if (ret && slot > 0) {
2035                         dec = 1;
2036                         slot--;
2037                 }
2038                 p->slots[level] = slot;
2039                 unlock_up(p, level, lowest_unlock, 0, NULL);
2040
2041                 if (level == lowest_level) {
2042                         if (dec)
2043                                 p->slots[level]++;
2044                         goto done;
2045                 }
2046
2047                 err = read_block_for_search(root, p, &b, level, slot, key);
2048                 if (err == -EAGAIN)
2049                         goto again;
2050                 if (err) {
2051                         ret = err;
2052                         goto done;
2053                 }
2054
2055                 level = btrfs_header_level(b);
2056                 btrfs_tree_read_lock(b);
2057                 b = btrfs_tree_mod_log_rewind(fs_info, p, b, time_seq);
2058                 if (!b) {
2059                         ret = -ENOMEM;
2060                         goto done;
2061                 }
2062                 p->locks[level] = BTRFS_READ_LOCK;
2063                 p->nodes[level] = b;
2064         }
2065         ret = 1;
2066 done:
2067         if (ret < 0)
2068                 btrfs_release_path(p);
2069
2070         return ret;
2071 }
2072
2073 /*
2074  * helper to use instead of search slot if no exact match is needed but
2075  * instead the next or previous item should be returned.
2076  * When find_higher is true, the next higher item is returned, the next lower
2077  * otherwise.
2078  * When return_any and find_higher are both true, and no higher item is found,
2079  * return the next lower instead.
2080  * When return_any is true and find_higher is false, and no lower item is found,
2081  * return the next higher instead.
2082  * It returns 0 if any item is found, 1 if none is found (tree empty), and
2083  * < 0 on error
2084  */
2085 int btrfs_search_slot_for_read(struct btrfs_root *root,
2086                                const struct btrfs_key *key,
2087                                struct btrfs_path *p, int find_higher,
2088                                int return_any)
2089 {
2090         int ret;
2091         struct extent_buffer *leaf;
2092
2093 again:
2094         ret = btrfs_search_slot(NULL, root, key, p, 0, 0);
2095         if (ret <= 0)
2096                 return ret;
2097         /*
2098          * a return value of 1 means the path is at the position where the
2099          * item should be inserted. Normally this is the next bigger item,
2100          * but in case the previous item is the last in a leaf, path points
2101          * to the first free slot in the previous leaf, i.e. at an invalid
2102          * item.
2103          */
2104         leaf = p->nodes[0];
2105
2106         if (find_higher) {
2107                 if (p->slots[0] >= btrfs_header_nritems(leaf)) {
2108                         ret = btrfs_next_leaf(root, p);
2109                         if (ret <= 0)
2110                                 return ret;
2111                         if (!return_any)
2112                                 return 1;
2113                         /*
2114                          * no higher item found, return the next
2115                          * lower instead
2116                          */
2117                         return_any = 0;
2118                         find_higher = 0;
2119                         btrfs_release_path(p);
2120                         goto again;
2121                 }
2122         } else {
2123                 if (p->slots[0] == 0) {
2124                         ret = btrfs_prev_leaf(root, p);
2125                         if (ret < 0)
2126                                 return ret;
2127                         if (!ret) {
2128                                 leaf = p->nodes[0];
2129                                 if (p->slots[0] == btrfs_header_nritems(leaf))
2130                                         p->slots[0]--;
2131                                 return 0;
2132                         }
2133                         if (!return_any)
2134                                 return 1;
2135                         /*
2136                          * no lower item found, return the next
2137                          * higher instead
2138                          */
2139                         return_any = 0;
2140                         find_higher = 1;
2141                         btrfs_release_path(p);
2142                         goto again;
2143                 } else {
2144                         --p->slots[0];
2145                 }
2146         }
2147         return 0;
2148 }
2149
2150 /*
2151  * adjust the pointers going up the tree, starting at level
2152  * making sure the right key of each node is points to 'key'.
2153  * This is used after shifting pointers to the left, so it stops
2154  * fixing up pointers when a given leaf/node is not in slot 0 of the
2155  * higher levels
2156  *
2157  */
2158 static void fixup_low_keys(struct btrfs_path *path,
2159                            struct btrfs_disk_key *key, int level)
2160 {
2161         int i;
2162         struct extent_buffer *t;
2163         int ret;
2164
2165         for (i = level; i < BTRFS_MAX_LEVEL; i++) {
2166                 int tslot = path->slots[i];
2167
2168                 if (!path->nodes[i])
2169                         break;
2170                 t = path->nodes[i];
2171                 ret = btrfs_tree_mod_log_insert_key(t, tslot,
2172                                 BTRFS_MOD_LOG_KEY_REPLACE, GFP_ATOMIC);
2173                 BUG_ON(ret < 0);
2174                 btrfs_set_node_key(t, key, tslot);
2175                 btrfs_mark_buffer_dirty(path->nodes[i]);
2176                 if (tslot != 0)
2177                         break;
2178         }
2179 }
2180
2181 /*
2182  * update item key.
2183  *
2184  * This function isn't completely safe. It's the caller's responsibility
2185  * that the new key won't break the order
2186  */
2187 void btrfs_set_item_key_safe(struct btrfs_fs_info *fs_info,
2188                              struct btrfs_path *path,
2189                              const struct btrfs_key *new_key)
2190 {
2191         struct btrfs_disk_key disk_key;
2192         struct extent_buffer *eb;
2193         int slot;
2194
2195         eb = path->nodes[0];
2196         slot = path->slots[0];
2197         if (slot > 0) {
2198                 btrfs_item_key(eb, &disk_key, slot - 1);
2199                 if (unlikely(comp_keys(&disk_key, new_key) >= 0)) {
2200                         btrfs_crit(fs_info,
2201                 "slot %u key (%llu %u %llu) new key (%llu %u %llu)",
2202                                    slot, btrfs_disk_key_objectid(&disk_key),
2203                                    btrfs_disk_key_type(&disk_key),
2204                                    btrfs_disk_key_offset(&disk_key),
2205                                    new_key->objectid, new_key->type,
2206                                    new_key->offset);
2207                         btrfs_print_leaf(eb);
2208                         BUG();
2209                 }
2210         }
2211         if (slot < btrfs_header_nritems(eb) - 1) {
2212                 btrfs_item_key(eb, &disk_key, slot + 1);
2213                 if (unlikely(comp_keys(&disk_key, new_key) <= 0)) {
2214                         btrfs_crit(fs_info,
2215                 "slot %u key (%llu %u %llu) new key (%llu %u %llu)",
2216                                    slot, btrfs_disk_key_objectid(&disk_key),
2217                                    btrfs_disk_key_type(&disk_key),
2218                                    btrfs_disk_key_offset(&disk_key),
2219                                    new_key->objectid, new_key->type,
2220                                    new_key->offset);
2221                         btrfs_print_leaf(eb);
2222                         BUG();
2223                 }
2224         }
2225
2226         btrfs_cpu_key_to_disk(&disk_key, new_key);
2227         btrfs_set_item_key(eb, &disk_key, slot);
2228         btrfs_mark_buffer_dirty(eb);
2229         if (slot == 0)
2230                 fixup_low_keys(path, &disk_key, 1);
2231 }
2232
2233 /*
2234  * Check key order of two sibling extent buffers.
2235  *
2236  * Return true if something is wrong.
2237  * Return false if everything is fine.
2238  *
2239  * Tree-checker only works inside one tree block, thus the following
2240  * corruption can not be detected by tree-checker:
2241  *
2242  * Leaf @left                   | Leaf @right
2243  * --------------------------------------------------------------
2244  * | 1 | 2 | 3 | 4 | 5 | f6 |   | 7 | 8 |
2245  *
2246  * Key f6 in leaf @left itself is valid, but not valid when the next
2247  * key in leaf @right is 7.
2248  * This can only be checked at tree block merge time.
2249  * And since tree checker has ensured all key order in each tree block
2250  * is correct, we only need to bother the last key of @left and the first
2251  * key of @right.
2252  */
2253 static bool check_sibling_keys(struct extent_buffer *left,
2254                                struct extent_buffer *right)
2255 {
2256         struct btrfs_key left_last;
2257         struct btrfs_key right_first;
2258         int level = btrfs_header_level(left);
2259         int nr_left = btrfs_header_nritems(left);
2260         int nr_right = btrfs_header_nritems(right);
2261
2262         /* No key to check in one of the tree blocks */
2263         if (!nr_left || !nr_right)
2264                 return false;
2265
2266         if (level) {
2267                 btrfs_node_key_to_cpu(left, &left_last, nr_left - 1);
2268                 btrfs_node_key_to_cpu(right, &right_first, 0);
2269         } else {
2270                 btrfs_item_key_to_cpu(left, &left_last, nr_left - 1);
2271                 btrfs_item_key_to_cpu(right, &right_first, 0);
2272         }
2273
2274         if (btrfs_comp_cpu_keys(&left_last, &right_first) >= 0) {
2275                 btrfs_crit(left->fs_info,
2276 "bad key order, sibling blocks, left last (%llu %u %llu) right first (%llu %u %llu)",
2277                            left_last.objectid, left_last.type,
2278                            left_last.offset, right_first.objectid,
2279                            right_first.type, right_first.offset);
2280                 return true;
2281         }
2282         return false;
2283 }
2284
2285 /*
2286  * try to push data from one node into the next node left in the
2287  * tree.
2288  *
2289  * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
2290  * error, and > 0 if there was no room in the left hand block.
2291  */
2292 static int push_node_left(struct btrfs_trans_handle *trans,
2293                           struct extent_buffer *dst,
2294                           struct extent_buffer *src, int empty)
2295 {
2296         struct btrfs_fs_info *fs_info = trans->fs_info;
2297         int push_items = 0;
2298         int src_nritems;
2299         int dst_nritems;
2300         int ret = 0;
2301
2302         src_nritems = btrfs_header_nritems(src);
2303         dst_nritems = btrfs_header_nritems(dst);
2304         push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems;
2305         WARN_ON(btrfs_header_generation(src) != trans->transid);
2306         WARN_ON(btrfs_header_generation(dst) != trans->transid);
2307
2308         if (!empty && src_nritems <= 8)
2309                 return 1;
2310
2311         if (push_items <= 0)
2312                 return 1;
2313
2314         if (empty) {
2315                 push_items = min(src_nritems, push_items);
2316                 if (push_items < src_nritems) {
2317                         /* leave at least 8 pointers in the node if
2318                          * we aren't going to empty it
2319                          */
2320                         if (src_nritems - push_items < 8) {
2321                                 if (push_items <= 8)
2322                                         return 1;
2323                                 push_items -= 8;
2324                         }
2325                 }
2326         } else
2327                 push_items = min(src_nritems - 8, push_items);
2328
2329         /* dst is the left eb, src is the middle eb */
2330         if (check_sibling_keys(dst, src)) {
2331                 ret = -EUCLEAN;
2332                 btrfs_abort_transaction(trans, ret);
2333                 return ret;
2334         }
2335         ret = btrfs_tree_mod_log_eb_copy(dst, src, dst_nritems, 0, push_items);
2336         if (ret) {
2337                 btrfs_abort_transaction(trans, ret);
2338                 return ret;
2339         }
2340         copy_extent_buffer(dst, src,
2341                            btrfs_node_key_ptr_offset(dst_nritems),
2342                            btrfs_node_key_ptr_offset(0),
2343                            push_items * sizeof(struct btrfs_key_ptr));
2344
2345         if (push_items < src_nritems) {
2346                 /*
2347                  * Don't call btrfs_tree_mod_log_insert_move() here, key removal
2348                  * was already fully logged by btrfs_tree_mod_log_eb_copy() above.
2349                  */
2350                 memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0),
2351                                       btrfs_node_key_ptr_offset(push_items),
2352                                       (src_nritems - push_items) *
2353                                       sizeof(struct btrfs_key_ptr));
2354         }
2355         btrfs_set_header_nritems(src, src_nritems - push_items);
2356         btrfs_set_header_nritems(dst, dst_nritems + push_items);
2357         btrfs_mark_buffer_dirty(src);
2358         btrfs_mark_buffer_dirty(dst);
2359
2360         return ret;
2361 }
2362
2363 /*
2364  * try to push data from one node into the next node right in the
2365  * tree.
2366  *
2367  * returns 0 if some ptrs were pushed, < 0 if there was some horrible
2368  * error, and > 0 if there was no room in the right hand block.
2369  *
2370  * this will  only push up to 1/2 the contents of the left node over
2371  */
2372 static int balance_node_right(struct btrfs_trans_handle *trans,
2373                               struct extent_buffer *dst,
2374                               struct extent_buffer *src)
2375 {
2376         struct btrfs_fs_info *fs_info = trans->fs_info;
2377         int push_items = 0;
2378         int max_push;
2379         int src_nritems;
2380         int dst_nritems;
2381         int ret = 0;
2382
2383         WARN_ON(btrfs_header_generation(src) != trans->transid);
2384         WARN_ON(btrfs_header_generation(dst) != trans->transid);
2385
2386         src_nritems = btrfs_header_nritems(src);
2387         dst_nritems = btrfs_header_nritems(dst);
2388         push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems;
2389         if (push_items <= 0)
2390                 return 1;
2391
2392         if (src_nritems < 4)
2393                 return 1;
2394
2395         max_push = src_nritems / 2 + 1;
2396         /* don't try to empty the node */
2397         if (max_push >= src_nritems)
2398                 return 1;
2399
2400         if (max_push < push_items)
2401                 push_items = max_push;
2402
2403         /* dst is the right eb, src is the middle eb */
2404         if (check_sibling_keys(src, dst)) {
2405                 ret = -EUCLEAN;
2406                 btrfs_abort_transaction(trans, ret);
2407                 return ret;
2408         }
2409         ret = btrfs_tree_mod_log_insert_move(dst, push_items, 0, dst_nritems);
2410         BUG_ON(ret < 0);
2411         memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items),
2412                                       btrfs_node_key_ptr_offset(0),
2413                                       (dst_nritems) *
2414                                       sizeof(struct btrfs_key_ptr));
2415
2416         ret = btrfs_tree_mod_log_eb_copy(dst, src, 0, src_nritems - push_items,
2417                                          push_items);
2418         if (ret) {
2419                 btrfs_abort_transaction(trans, ret);
2420                 return ret;
2421         }
2422         copy_extent_buffer(dst, src,
2423                            btrfs_node_key_ptr_offset(0),
2424                            btrfs_node_key_ptr_offset(src_nritems - push_items),
2425                            push_items * sizeof(struct btrfs_key_ptr));
2426
2427         btrfs_set_header_nritems(src, src_nritems - push_items);
2428         btrfs_set_header_nritems(dst, dst_nritems + push_items);
2429
2430         btrfs_mark_buffer_dirty(src);
2431         btrfs_mark_buffer_dirty(dst);
2432
2433         return ret;
2434 }
2435
2436 /*
2437  * helper function to insert a new root level in the tree.
2438  * A new node is allocated, and a single item is inserted to
2439  * point to the existing root
2440  *
2441  * returns zero on success or < 0 on failure.
2442  */
2443 static noinline int insert_new_root(struct btrfs_trans_handle *trans,
2444                            struct btrfs_root *root,
2445                            struct btrfs_path *path, int level)
2446 {
2447         struct btrfs_fs_info *fs_info = root->fs_info;
2448         u64 lower_gen;
2449         struct extent_buffer *lower;
2450         struct extent_buffer *c;
2451         struct extent_buffer *old;
2452         struct btrfs_disk_key lower_key;
2453         int ret;
2454
2455         BUG_ON(path->nodes[level]);
2456         BUG_ON(path->nodes[level-1] != root->node);
2457
2458         lower = path->nodes[level-1];
2459         if (level == 1)
2460                 btrfs_item_key(lower, &lower_key, 0);
2461         else
2462                 btrfs_node_key(lower, &lower_key, 0);
2463
2464         c = alloc_tree_block_no_bg_flush(trans, root, 0, &lower_key, level,
2465                                          root->node->start, 0,
2466                                          BTRFS_NESTING_NEW_ROOT);
2467         if (IS_ERR(c))
2468                 return PTR_ERR(c);
2469
2470         root_add_used(root, fs_info->nodesize);
2471
2472         btrfs_set_header_nritems(c, 1);
2473         btrfs_set_node_key(c, &lower_key, 0);
2474         btrfs_set_node_blockptr(c, 0, lower->start);
2475         lower_gen = btrfs_header_generation(lower);
2476         WARN_ON(lower_gen != trans->transid);
2477
2478         btrfs_set_node_ptr_generation(c, 0, lower_gen);
2479
2480         btrfs_mark_buffer_dirty(c);
2481
2482         old = root->node;
2483         ret = btrfs_tree_mod_log_insert_root(root->node, c, false);
2484         BUG_ON(ret < 0);
2485         rcu_assign_pointer(root->node, c);
2486
2487         /* the super has an extra ref to root->node */
2488         free_extent_buffer(old);
2489
2490         add_root_to_dirty_list(root);
2491         atomic_inc(&c->refs);
2492         path->nodes[level] = c;
2493         path->locks[level] = BTRFS_WRITE_LOCK;
2494         path->slots[level] = 0;
2495         return 0;
2496 }
2497
2498 /*
2499  * worker function to insert a single pointer in a node.
2500  * the node should have enough room for the pointer already
2501  *
2502  * slot and level indicate where you want the key to go, and
2503  * blocknr is the block the key points to.
2504  */
2505 static void insert_ptr(struct btrfs_trans_handle *trans,
2506                        struct btrfs_path *path,
2507                        struct btrfs_disk_key *key, u64 bytenr,
2508                        int slot, int level)
2509 {
2510         struct extent_buffer *lower;
2511         int nritems;
2512         int ret;
2513
2514         BUG_ON(!path->nodes[level]);
2515         btrfs_assert_tree_locked(path->nodes[level]);
2516         lower = path->nodes[level];
2517         nritems = btrfs_header_nritems(lower);
2518         BUG_ON(slot > nritems);
2519         BUG_ON(nritems == BTRFS_NODEPTRS_PER_BLOCK(trans->fs_info));
2520         if (slot != nritems) {
2521                 if (level) {
2522                         ret = btrfs_tree_mod_log_insert_move(lower, slot + 1,
2523                                         slot, nritems - slot);
2524                         BUG_ON(ret < 0);
2525                 }
2526                 memmove_extent_buffer(lower,
2527                               btrfs_node_key_ptr_offset(slot + 1),
2528                               btrfs_node_key_ptr_offset(slot),
2529                               (nritems - slot) * sizeof(struct btrfs_key_ptr));
2530         }
2531         if (level) {
2532                 ret = btrfs_tree_mod_log_insert_key(lower, slot,
2533                                             BTRFS_MOD_LOG_KEY_ADD, GFP_NOFS);
2534                 BUG_ON(ret < 0);
2535         }
2536         btrfs_set_node_key(lower, key, slot);
2537         btrfs_set_node_blockptr(lower, slot, bytenr);
2538         WARN_ON(trans->transid == 0);
2539         btrfs_set_node_ptr_generation(lower, slot, trans->transid);
2540         btrfs_set_header_nritems(lower, nritems + 1);
2541         btrfs_mark_buffer_dirty(lower);
2542 }
2543
2544 /*
2545  * split the node at the specified level in path in two.
2546  * The path is corrected to point to the appropriate node after the split
2547  *
2548  * Before splitting this tries to make some room in the node by pushing
2549  * left and right, if either one works, it returns right away.
2550  *
2551  * returns 0 on success and < 0 on failure
2552  */
2553 static noinline int split_node(struct btrfs_trans_handle *trans,
2554                                struct btrfs_root *root,
2555                                struct btrfs_path *path, int level)
2556 {
2557         struct btrfs_fs_info *fs_info = root->fs_info;
2558         struct extent_buffer *c;
2559         struct extent_buffer *split;
2560         struct btrfs_disk_key disk_key;
2561         int mid;
2562         int ret;
2563         u32 c_nritems;
2564
2565         c = path->nodes[level];
2566         WARN_ON(btrfs_header_generation(c) != trans->transid);
2567         if (c == root->node) {
2568                 /*
2569                  * trying to split the root, lets make a new one
2570                  *
2571                  * tree mod log: We don't log_removal old root in
2572                  * insert_new_root, because that root buffer will be kept as a
2573                  * normal node. We are going to log removal of half of the
2574                  * elements below with btrfs_tree_mod_log_eb_copy(). We're
2575                  * holding a tree lock on the buffer, which is why we cannot
2576                  * race with other tree_mod_log users.
2577                  */
2578                 ret = insert_new_root(trans, root, path, level + 1);
2579                 if (ret)
2580                         return ret;
2581         } else {
2582                 ret = push_nodes_for_insert(trans, root, path, level);
2583                 c = path->nodes[level];
2584                 if (!ret && btrfs_header_nritems(c) <
2585                     BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3)
2586                         return 0;
2587                 if (ret < 0)
2588                         return ret;
2589         }
2590
2591         c_nritems = btrfs_header_nritems(c);
2592         mid = (c_nritems + 1) / 2;
2593         btrfs_node_key(c, &disk_key, mid);
2594
2595         split = alloc_tree_block_no_bg_flush(trans, root, 0, &disk_key, level,
2596                                              c->start, 0, BTRFS_NESTING_SPLIT);
2597         if (IS_ERR(split))
2598                 return PTR_ERR(split);
2599
2600         root_add_used(root, fs_info->nodesize);
2601         ASSERT(btrfs_header_level(c) == level);
2602
2603         ret = btrfs_tree_mod_log_eb_copy(split, c, 0, mid, c_nritems - mid);
2604         if (ret) {
2605                 btrfs_abort_transaction(trans, ret);
2606                 return ret;
2607         }
2608         copy_extent_buffer(split, c,
2609                            btrfs_node_key_ptr_offset(0),
2610                            btrfs_node_key_ptr_offset(mid),
2611                            (c_nritems - mid) * sizeof(struct btrfs_key_ptr));
2612         btrfs_set_header_nritems(split, c_nritems - mid);
2613         btrfs_set_header_nritems(c, mid);
2614
2615         btrfs_mark_buffer_dirty(c);
2616         btrfs_mark_buffer_dirty(split);
2617
2618         insert_ptr(trans, path, &disk_key, split->start,
2619                    path->slots[level + 1] + 1, level + 1);
2620
2621         if (path->slots[level] >= mid) {
2622                 path->slots[level] -= mid;
2623                 btrfs_tree_unlock(c);
2624                 free_extent_buffer(c);
2625                 path->nodes[level] = split;
2626                 path->slots[level + 1] += 1;
2627         } else {
2628                 btrfs_tree_unlock(split);
2629                 free_extent_buffer(split);
2630         }
2631         return 0;
2632 }
2633
2634 /*
2635  * how many bytes are required to store the items in a leaf.  start
2636  * and nr indicate which items in the leaf to check.  This totals up the
2637  * space used both by the item structs and the item data
2638  */
2639 static int leaf_space_used(struct extent_buffer *l, int start, int nr)
2640 {
2641         struct btrfs_item *start_item;
2642         struct btrfs_item *end_item;
2643         int data_len;
2644         int nritems = btrfs_header_nritems(l);
2645         int end = min(nritems, start + nr) - 1;
2646
2647         if (!nr)
2648                 return 0;
2649         start_item = btrfs_item_nr(start);
2650         end_item = btrfs_item_nr(end);
2651         data_len = btrfs_item_offset(l, start_item) +
2652                    btrfs_item_size(l, start_item);
2653         data_len = data_len - btrfs_item_offset(l, end_item);
2654         data_len += sizeof(struct btrfs_item) * nr;
2655         WARN_ON(data_len < 0);
2656         return data_len;
2657 }
2658
2659 /*
2660  * The space between the end of the leaf items and
2661  * the start of the leaf data.  IOW, how much room
2662  * the leaf has left for both items and data
2663  */
2664 noinline int btrfs_leaf_free_space(struct extent_buffer *leaf)
2665 {
2666         struct btrfs_fs_info *fs_info = leaf->fs_info;
2667         int nritems = btrfs_header_nritems(leaf);
2668         int ret;
2669
2670         ret = BTRFS_LEAF_DATA_SIZE(fs_info) - leaf_space_used(leaf, 0, nritems);
2671         if (ret < 0) {
2672                 btrfs_crit(fs_info,
2673                            "leaf free space ret %d, leaf data size %lu, used %d nritems %d",
2674                            ret,
2675                            (unsigned long) BTRFS_LEAF_DATA_SIZE(fs_info),
2676                            leaf_space_used(leaf, 0, nritems), nritems);
2677         }
2678         return ret;
2679 }
2680
2681 /*
2682  * min slot controls the lowest index we're willing to push to the
2683  * right.  We'll push up to and including min_slot, but no lower
2684  */
2685 static noinline int __push_leaf_right(struct btrfs_path *path,
2686                                       int data_size, int empty,
2687                                       struct extent_buffer *right,
2688                                       int free_space, u32 left_nritems,
2689                                       u32 min_slot)
2690 {
2691         struct btrfs_fs_info *fs_info = right->fs_info;
2692         struct extent_buffer *left = path->nodes[0];
2693         struct extent_buffer *upper = path->nodes[1];
2694         struct btrfs_map_token token;
2695         struct btrfs_disk_key disk_key;
2696         int slot;
2697         u32 i;
2698         int push_space = 0;
2699         int push_items = 0;
2700         struct btrfs_item *item;
2701         u32 nr;
2702         u32 right_nritems;
2703         u32 data_end;
2704         u32 this_item_size;
2705
2706         if (empty)
2707                 nr = 0;
2708         else
2709                 nr = max_t(u32, 1, min_slot);
2710
2711         if (path->slots[0] >= left_nritems)
2712                 push_space += data_size;
2713
2714         slot = path->slots[1];
2715         i = left_nritems - 1;
2716         while (i >= nr) {
2717                 item = btrfs_item_nr(i);
2718
2719                 if (!empty && push_items > 0) {
2720                         if (path->slots[0] > i)
2721                                 break;
2722                         if (path->slots[0] == i) {
2723                                 int space = btrfs_leaf_free_space(left);
2724
2725                                 if (space + push_space * 2 > free_space)
2726                                         break;
2727                         }
2728                 }
2729
2730                 if (path->slots[0] == i)
2731                         push_space += data_size;
2732
2733                 this_item_size = btrfs_item_size(left, item);
2734                 if (this_item_size + sizeof(*item) + push_space > free_space)
2735                         break;
2736
2737                 push_items++;
2738                 push_space += this_item_size + sizeof(*item);
2739                 if (i == 0)
2740                         break;
2741                 i--;
2742         }
2743
2744         if (push_items == 0)
2745                 goto out_unlock;
2746
2747         WARN_ON(!empty && push_items == left_nritems);
2748
2749         /* push left to right */
2750         right_nritems = btrfs_header_nritems(right);
2751
2752         push_space = btrfs_item_end_nr(left, left_nritems - push_items);
2753         push_space -= leaf_data_end(left);
2754
2755         /* make room in the right data area */
2756         data_end = leaf_data_end(right);
2757         memmove_extent_buffer(right,
2758                               BTRFS_LEAF_DATA_OFFSET + data_end - push_space,
2759                               BTRFS_LEAF_DATA_OFFSET + data_end,
2760                               BTRFS_LEAF_DATA_SIZE(fs_info) - data_end);
2761
2762         /* copy from the left data area */
2763         copy_extent_buffer(right, left, BTRFS_LEAF_DATA_OFFSET +
2764                      BTRFS_LEAF_DATA_SIZE(fs_info) - push_space,
2765                      BTRFS_LEAF_DATA_OFFSET + leaf_data_end(left),
2766                      push_space);
2767
2768         memmove_extent_buffer(right, btrfs_item_nr_offset(push_items),
2769                               btrfs_item_nr_offset(0),
2770                               right_nritems * sizeof(struct btrfs_item));
2771
2772         /* copy the items from left to right */
2773         copy_extent_buffer(right, left, btrfs_item_nr_offset(0),
2774                    btrfs_item_nr_offset(left_nritems - push_items),
2775                    push_items * sizeof(struct btrfs_item));
2776
2777         /* update the item pointers */
2778         btrfs_init_map_token(&token, right);
2779         right_nritems += push_items;
2780         btrfs_set_header_nritems(right, right_nritems);
2781         push_space = BTRFS_LEAF_DATA_SIZE(fs_info);
2782         for (i = 0; i < right_nritems; i++) {
2783                 item = btrfs_item_nr(i);
2784                 push_space -= btrfs_token_item_size(&token, item);
2785                 btrfs_set_token_item_offset(&token, item, push_space);
2786         }
2787
2788         left_nritems -= push_items;
2789         btrfs_set_header_nritems(left, left_nritems);
2790
2791         if (left_nritems)
2792                 btrfs_mark_buffer_dirty(left);
2793         else
2794                 btrfs_clean_tree_block(left);
2795
2796         btrfs_mark_buffer_dirty(right);
2797
2798         btrfs_item_key(right, &disk_key, 0);
2799         btrfs_set_node_key(upper, &disk_key, slot + 1);
2800         btrfs_mark_buffer_dirty(upper);
2801
2802         /* then fixup the leaf pointer in the path */
2803         if (path->slots[0] >= left_nritems) {
2804                 path->slots[0] -= left_nritems;
2805                 if (btrfs_header_nritems(path->nodes[0]) == 0)
2806                         btrfs_clean_tree_block(path->nodes[0]);
2807                 btrfs_tree_unlock(path->nodes[0]);
2808                 free_extent_buffer(path->nodes[0]);
2809                 path->nodes[0] = right;
2810                 path->slots[1] += 1;
2811         } else {
2812                 btrfs_tree_unlock(right);
2813                 free_extent_buffer(right);
2814         }
2815         return 0;
2816
2817 out_unlock:
2818         btrfs_tree_unlock(right);
2819         free_extent_buffer(right);
2820         return 1;
2821 }
2822
2823 /*
2824  * push some data in the path leaf to the right, trying to free up at
2825  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
2826  *
2827  * returns 1 if the push failed because the other node didn't have enough
2828  * room, 0 if everything worked out and < 0 if there were major errors.
2829  *
2830  * this will push starting from min_slot to the end of the leaf.  It won't
2831  * push any slot lower than min_slot
2832  */
2833 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
2834                            *root, struct btrfs_path *path,
2835                            int min_data_size, int data_size,
2836                            int empty, u32 min_slot)
2837 {
2838         struct extent_buffer *left = path->nodes[0];
2839         struct extent_buffer *right;
2840         struct extent_buffer *upper;
2841         int slot;
2842         int free_space;
2843         u32 left_nritems;
2844         int ret;
2845
2846         if (!path->nodes[1])
2847                 return 1;
2848
2849         slot = path->slots[1];
2850         upper = path->nodes[1];
2851         if (slot >= btrfs_header_nritems(upper) - 1)
2852                 return 1;
2853
2854         btrfs_assert_tree_locked(path->nodes[1]);
2855
2856         right = btrfs_read_node_slot(upper, slot + 1);
2857         /*
2858          * slot + 1 is not valid or we fail to read the right node,
2859          * no big deal, just return.
2860          */
2861         if (IS_ERR(right))
2862                 return 1;
2863
2864         __btrfs_tree_lock(right, BTRFS_NESTING_RIGHT);
2865
2866         free_space = btrfs_leaf_free_space(right);
2867         if (free_space < data_size)
2868                 goto out_unlock;
2869
2870         /* cow and double check */
2871         ret = btrfs_cow_block(trans, root, right, upper,
2872                               slot + 1, &right, BTRFS_NESTING_RIGHT_COW);
2873         if (ret)
2874                 goto out_unlock;
2875
2876         free_space = btrfs_leaf_free_space(right);
2877         if (free_space < data_size)
2878                 goto out_unlock;
2879
2880         left_nritems = btrfs_header_nritems(left);
2881         if (left_nritems == 0)
2882                 goto out_unlock;
2883
2884         if (check_sibling_keys(left, right)) {
2885                 ret = -EUCLEAN;
2886                 btrfs_tree_unlock(right);
2887                 free_extent_buffer(right);
2888                 return ret;
2889         }
2890         if (path->slots[0] == left_nritems && !empty) {
2891                 /* Key greater than all keys in the leaf, right neighbor has
2892                  * enough room for it and we're not emptying our leaf to delete
2893                  * it, therefore use right neighbor to insert the new item and
2894                  * no need to touch/dirty our left leaf. */
2895                 btrfs_tree_unlock(left);
2896                 free_extent_buffer(left);
2897                 path->nodes[0] = right;
2898                 path->slots[0] = 0;
2899                 path->slots[1]++;
2900                 return 0;
2901         }
2902
2903         return __push_leaf_right(path, min_data_size, empty,
2904                                 right, free_space, left_nritems, min_slot);
2905 out_unlock:
2906         btrfs_tree_unlock(right);
2907         free_extent_buffer(right);
2908         return 1;
2909 }
2910
2911 /*
2912  * push some data in the path leaf to the left, trying to free up at
2913  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
2914  *
2915  * max_slot can put a limit on how far into the leaf we'll push items.  The
2916  * item at 'max_slot' won't be touched.  Use (u32)-1 to make us do all the
2917  * items
2918  */
2919 static noinline int __push_leaf_left(struct btrfs_path *path, int data_size,
2920                                      int empty, struct extent_buffer *left,
2921                                      int free_space, u32 right_nritems,
2922                                      u32 max_slot)
2923 {
2924         struct btrfs_fs_info *fs_info = left->fs_info;
2925         struct btrfs_disk_key disk_key;
2926         struct extent_buffer *right = path->nodes[0];
2927         int i;
2928         int push_space = 0;
2929         int push_items = 0;
2930         struct btrfs_item *item;
2931         u32 old_left_nritems;
2932         u32 nr;
2933         int ret = 0;
2934         u32 this_item_size;
2935         u32 old_left_item_size;
2936         struct btrfs_map_token token;
2937
2938         if (empty)
2939                 nr = min(right_nritems, max_slot);
2940         else
2941                 nr = min(right_nritems - 1, max_slot);
2942
2943         for (i = 0; i < nr; i++) {
2944                 item = btrfs_item_nr(i);
2945
2946                 if (!empty && push_items > 0) {
2947                         if (path->slots[0] < i)
2948                                 break;
2949                         if (path->slots[0] == i) {
2950                                 int space = btrfs_leaf_free_space(right);
2951
2952                                 if (space + push_space * 2 > free_space)
2953                                         break;
2954                         }
2955                 }
2956
2957                 if (path->slots[0] == i)
2958                         push_space += data_size;
2959
2960                 this_item_size = btrfs_item_size(right, item);
2961                 if (this_item_size + sizeof(*item) + push_space > free_space)
2962                         break;
2963
2964                 push_items++;
2965                 push_space += this_item_size + sizeof(*item);
2966         }
2967
2968         if (push_items == 0) {
2969                 ret = 1;
2970                 goto out;
2971         }
2972         WARN_ON(!empty && push_items == btrfs_header_nritems(right));
2973
2974         /* push data from right to left */
2975         copy_extent_buffer(left, right,
2976                            btrfs_item_nr_offset(btrfs_header_nritems(left)),
2977                            btrfs_item_nr_offset(0),
2978                            push_items * sizeof(struct btrfs_item));
2979
2980         push_space = BTRFS_LEAF_DATA_SIZE(fs_info) -
2981                      btrfs_item_offset_nr(right, push_items - 1);
2982
2983         copy_extent_buffer(left, right, BTRFS_LEAF_DATA_OFFSET +
2984                      leaf_data_end(left) - push_space,
2985                      BTRFS_LEAF_DATA_OFFSET +
2986                      btrfs_item_offset_nr(right, push_items - 1),
2987                      push_space);
2988         old_left_nritems = btrfs_header_nritems(left);
2989         BUG_ON(old_left_nritems <= 0);
2990
2991         btrfs_init_map_token(&token, left);
2992         old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1);
2993         for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
2994                 u32 ioff;
2995
2996                 item = btrfs_item_nr(i);
2997
2998                 ioff = btrfs_token_item_offset(&token, item);
2999                 btrfs_set_token_item_offset(&token, item,
3000                       ioff - (BTRFS_LEAF_DATA_SIZE(fs_info) - old_left_item_size));
3001         }
3002         btrfs_set_header_nritems(left, old_left_nritems + push_items);
3003
3004         /* fixup right node */
3005         if (push_items > right_nritems)
3006                 WARN(1, KERN_CRIT "push items %d nr %u\n", push_items,
3007                        right_nritems);
3008
3009         if (push_items < right_nritems) {
3010                 push_space = btrfs_item_offset_nr(right, push_items - 1) -
3011                                                   leaf_data_end(right);
3012                 memmove_extent_buffer(right, BTRFS_LEAF_DATA_OFFSET +
3013                                       BTRFS_LEAF_DATA_SIZE(fs_info) - push_space,
3014                                       BTRFS_LEAF_DATA_OFFSET +
3015                                       leaf_data_end(right), push_space);
3016
3017                 memmove_extent_buffer(right, btrfs_item_nr_offset(0),
3018                               btrfs_item_nr_offset(push_items),
3019                              (btrfs_header_nritems(right) - push_items) *
3020                              sizeof(struct btrfs_item));
3021         }
3022
3023         btrfs_init_map_token(&token, right);
3024         right_nritems -= push_items;
3025         btrfs_set_header_nritems(right, right_nritems);
3026         push_space = BTRFS_LEAF_DATA_SIZE(fs_info);
3027         for (i = 0; i < right_nritems; i++) {
3028                 item = btrfs_item_nr(i);
3029
3030                 push_space = push_space - btrfs_token_item_size(&token, item);
3031                 btrfs_set_token_item_offset(&token, item, push_space);
3032         }
3033
3034         btrfs_mark_buffer_dirty(left);
3035         if (right_nritems)
3036                 btrfs_mark_buffer_dirty(right);
3037         else
3038                 btrfs_clean_tree_block(right);
3039
3040         btrfs_item_key(right, &disk_key, 0);
3041         fixup_low_keys(path, &disk_key, 1);
3042
3043         /* then fixup the leaf pointer in the path */
3044         if (path->slots[0] < push_items) {
3045                 path->slots[0] += old_left_nritems;
3046                 btrfs_tree_unlock(path->nodes[0]);
3047                 free_extent_buffer(path->nodes[0]);
3048                 path->nodes[0] = left;
3049                 path->slots[1] -= 1;
3050         } else {
3051                 btrfs_tree_unlock(left);
3052                 free_extent_buffer(left);
3053                 path->slots[0] -= push_items;
3054         }
3055         BUG_ON(path->slots[0] < 0);
3056         return ret;
3057 out:
3058         btrfs_tree_unlock(left);
3059         free_extent_buffer(left);
3060         return ret;
3061 }
3062
3063 /*
3064  * push some data in the path leaf to the left, trying to free up at
3065  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
3066  *
3067  * max_slot can put a limit on how far into the leaf we'll push items.  The
3068  * item at 'max_slot' won't be touched.  Use (u32)-1 to make us push all the
3069  * items
3070  */
3071 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
3072                           *root, struct btrfs_path *path, int min_data_size,
3073                           int data_size, int empty, u32 max_slot)
3074 {
3075         struct extent_buffer *right = path->nodes[0];
3076         struct extent_buffer *left;
3077         int slot;
3078         int free_space;
3079         u32 right_nritems;
3080         int ret = 0;
3081
3082         slot = path->slots[1];
3083         if (slot == 0)
3084                 return 1;
3085         if (!path->nodes[1])
3086                 return 1;
3087
3088         right_nritems = btrfs_header_nritems(right);
3089         if (right_nritems == 0)
3090                 return 1;
3091
3092         btrfs_assert_tree_locked(path->nodes[1]);
3093
3094         left = btrfs_read_node_slot(path->nodes[1], slot - 1);
3095         /*
3096          * slot - 1 is not valid or we fail to read the left node,
3097          * no big deal, just return.
3098          */
3099         if (IS_ERR(left))
3100                 return 1;
3101
3102         __btrfs_tree_lock(left, BTRFS_NESTING_LEFT);
3103
3104         free_space = btrfs_leaf_free_space(left);
3105         if (free_space < data_size) {
3106                 ret = 1;
3107                 goto out;
3108         }
3109
3110         /* cow and double check */
3111         ret = btrfs_cow_block(trans, root, left,
3112                               path->nodes[1], slot - 1, &left,
3113                               BTRFS_NESTING_LEFT_COW);
3114         if (ret) {
3115                 /* we hit -ENOSPC, but it isn't fatal here */
3116                 if (ret == -ENOSPC)
3117                         ret = 1;
3118                 goto out;
3119         }
3120
3121         free_space = btrfs_leaf_free_space(left);
3122         if (free_space < data_size) {
3123                 ret = 1;
3124                 goto out;
3125         }
3126
3127         if (check_sibling_keys(left, right)) {
3128                 ret = -EUCLEAN;
3129                 goto out;
3130         }
3131         return __push_leaf_left(path, min_data_size,
3132                                empty, left, free_space, right_nritems,
3133                                max_slot);
3134 out:
3135         btrfs_tree_unlock(left);
3136         free_extent_buffer(left);
3137         return ret;
3138 }
3139
3140 /*
3141  * split the path's leaf in two, making sure there is at least data_size
3142  * available for the resulting leaf level of the path.
3143  */
3144 static noinline void copy_for_split(struct btrfs_trans_handle *trans,
3145                                     struct btrfs_path *path,
3146                                     struct extent_buffer *l,
3147                                     struct extent_buffer *right,
3148                                     int slot, int mid, int nritems)
3149 {
3150         struct btrfs_fs_info *fs_info = trans->fs_info;
3151         int data_copy_size;
3152         int rt_data_off;
3153         int i;
3154         struct btrfs_disk_key disk_key;
3155         struct btrfs_map_token token;
3156
3157         nritems = nritems - mid;
3158         btrfs_set_header_nritems(right, nritems);
3159         data_copy_size = btrfs_item_end_nr(l, mid) - leaf_data_end(l);
3160
3161         copy_extent_buffer(right, l, btrfs_item_nr_offset(0),
3162                            btrfs_item_nr_offset(mid),
3163                            nritems * sizeof(struct btrfs_item));
3164
3165         copy_extent_buffer(right, l,
3166                      BTRFS_LEAF_DATA_OFFSET + BTRFS_LEAF_DATA_SIZE(fs_info) -
3167                      data_copy_size, BTRFS_LEAF_DATA_OFFSET +
3168                      leaf_data_end(l), data_copy_size);
3169
3170         rt_data_off = BTRFS_LEAF_DATA_SIZE(fs_info) - btrfs_item_end_nr(l, mid);
3171
3172         btrfs_init_map_token(&token, right);
3173         for (i = 0; i < nritems; i++) {
3174                 struct btrfs_item *item = btrfs_item_nr(i);
3175                 u32 ioff;
3176
3177                 ioff = btrfs_token_item_offset(&token, item);
3178                 btrfs_set_token_item_offset(&token, item, ioff + rt_data_off);
3179         }
3180
3181         btrfs_set_header_nritems(l, mid);
3182         btrfs_item_key(right, &disk_key, 0);
3183         insert_ptr(trans, path, &disk_key, right->start, path->slots[1] + 1, 1);
3184
3185         btrfs_mark_buffer_dirty(right);
3186         btrfs_mark_buffer_dirty(l);
3187         BUG_ON(path->slots[0] != slot);
3188
3189         if (mid <= slot) {
3190                 btrfs_tree_unlock(path->nodes[0]);
3191                 free_extent_buffer(path->nodes[0]);
3192                 path->nodes[0] = right;
3193                 path->slots[0] -= mid;
3194                 path->slots[1] += 1;
3195         } else {
3196                 btrfs_tree_unlock(right);
3197                 free_extent_buffer(right);
3198         }
3199
3200         BUG_ON(path->slots[0] < 0);
3201 }
3202
3203 /*
3204  * double splits happen when we need to insert a big item in the middle
3205  * of a leaf.  A double split can leave us with 3 mostly empty leaves:
3206  * leaf: [ slots 0 - N] [ our target ] [ N + 1 - total in leaf ]
3207  *          A                 B                 C
3208  *
3209  * We avoid this by trying to push the items on either side of our target
3210  * into the adjacent leaves.  If all goes well we can avoid the double split
3211  * completely.
3212  */
3213 static noinline int push_for_double_split(struct btrfs_trans_handle *trans,
3214                                           struct btrfs_root *root,
3215                                           struct btrfs_path *path,
3216                                           int data_size)
3217 {
3218         int ret;
3219         int progress = 0;
3220         int slot;
3221         u32 nritems;
3222         int space_needed = data_size;
3223
3224         slot = path->slots[0];
3225         if (slot < btrfs_header_nritems(path->nodes[0]))
3226                 space_needed -= btrfs_leaf_free_space(path->nodes[0]);
3227
3228         /*
3229          * try to push all the items after our slot into the
3230          * right leaf
3231          */
3232         ret = push_leaf_right(trans, root, path, 1, space_needed, 0, slot);
3233         if (ret < 0)
3234                 return ret;
3235
3236         if (ret == 0)
3237                 progress++;
3238
3239         nritems = btrfs_header_nritems(path->nodes[0]);
3240         /*
3241          * our goal is to get our slot at the start or end of a leaf.  If
3242          * we've done so we're done
3243          */
3244         if (path->slots[0] == 0 || path->slots[0] == nritems)
3245                 return 0;
3246
3247         if (btrfs_leaf_free_space(path->nodes[0]) >= data_size)
3248                 return 0;
3249
3250         /* try to push all the items before our slot into the next leaf */
3251         slot = path->slots[0];
3252         space_needed = data_size;
3253         if (slot > 0)
3254                 space_needed -= btrfs_leaf_free_space(path->nodes[0]);
3255         ret = push_leaf_left(trans, root, path, 1, space_needed, 0, slot);
3256         if (ret < 0)
3257                 return ret;
3258
3259         if (ret == 0)
3260                 progress++;
3261
3262         if (progress)
3263                 return 0;
3264         return 1;
3265 }
3266
3267 /*
3268  * split the path's leaf in two, making sure there is at least data_size
3269  * available for the resulting leaf level of the path.
3270  *
3271  * returns 0 if all went well and < 0 on failure.
3272  */
3273 static noinline int split_leaf(struct btrfs_trans_handle *trans,
3274                                struct btrfs_root *root,
3275                                const struct btrfs_key *ins_key,
3276                                struct btrfs_path *path, int data_size,
3277                                int extend)
3278 {
3279         struct btrfs_disk_key disk_key;
3280         struct extent_buffer *l;
3281         u32 nritems;
3282         int mid;
3283         int slot;
3284         struct extent_buffer *right;
3285         struct btrfs_fs_info *fs_info = root->fs_info;
3286         int ret = 0;
3287         int wret;
3288         int split;
3289         int num_doubles = 0;
3290         int tried_avoid_double = 0;
3291
3292         l = path->nodes[0];
3293         slot = path->slots[0];
3294         if (extend && data_size + btrfs_item_size_nr(l, slot) +
3295             sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(fs_info))
3296                 return -EOVERFLOW;
3297
3298         /* first try to make some room by pushing left and right */
3299         if (data_size && path->nodes[1]) {
3300                 int space_needed = data_size;
3301
3302                 if (slot < btrfs_header_nritems(l))
3303                         space_needed -= btrfs_leaf_free_space(l);
3304
3305                 wret = push_leaf_right(trans, root, path, space_needed,
3306                                        space_needed, 0, 0);
3307                 if (wret < 0)
3308                         return wret;
3309                 if (wret) {
3310                         space_needed = data_size;
3311                         if (slot > 0)
3312                                 space_needed -= btrfs_leaf_free_space(l);
3313                         wret = push_leaf_left(trans, root, path, space_needed,
3314                                               space_needed, 0, (u32)-1);
3315                         if (wret < 0)
3316                                 return wret;
3317                 }
3318                 l = path->nodes[0];
3319
3320                 /* did the pushes work? */
3321                 if (btrfs_leaf_free_space(l) >= data_size)
3322                         return 0;
3323         }
3324
3325         if (!path->nodes[1]) {
3326                 ret = insert_new_root(trans, root, path, 1);
3327                 if (ret)
3328                         return ret;
3329         }
3330 again:
3331         split = 1;
3332         l = path->nodes[0];
3333         slot = path->slots[0];
3334         nritems = btrfs_header_nritems(l);
3335         mid = (nritems + 1) / 2;
3336
3337         if (mid <= slot) {
3338                 if (nritems == 1 ||
3339                     leaf_space_used(l, mid, nritems - mid) + data_size >
3340                         BTRFS_LEAF_DATA_SIZE(fs_info)) {
3341                         if (slot >= nritems) {
3342                                 split = 0;
3343                         } else {
3344                                 mid = slot;
3345                                 if (mid != nritems &&
3346                                     leaf_space_used(l, mid, nritems - mid) +
3347                                     data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) {
3348                                         if (data_size && !tried_avoid_double)
3349                                                 goto push_for_double;
3350                                         split = 2;
3351                                 }
3352                         }
3353                 }
3354         } else {
3355                 if (leaf_space_used(l, 0, mid) + data_size >
3356                         BTRFS_LEAF_DATA_SIZE(fs_info)) {
3357                         if (!extend && data_size && slot == 0) {
3358                                 split = 0;
3359                         } else if ((extend || !data_size) && slot == 0) {
3360                                 mid = 1;
3361                         } else {
3362                                 mid = slot;
3363                                 if (mid != nritems &&
3364                                     leaf_space_used(l, mid, nritems - mid) +
3365                                     data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) {
3366                                         if (data_size && !tried_avoid_double)
3367                                                 goto push_for_double;
3368                                         split = 2;
3369                                 }
3370                         }
3371                 }
3372         }
3373
3374         if (split == 0)
3375                 btrfs_cpu_key_to_disk(&disk_key, ins_key);
3376         else
3377                 btrfs_item_key(l, &disk_key, mid);
3378
3379         /*
3380          * We have to about BTRFS_NESTING_NEW_ROOT here if we've done a double
3381          * split, because we're only allowed to have MAX_LOCKDEP_SUBCLASSES
3382          * subclasses, which is 8 at the time of this patch, and we've maxed it
3383          * out.  In the future we could add a
3384          * BTRFS_NESTING_SPLIT_THE_SPLITTENING if we need to, but for now just
3385          * use BTRFS_NESTING_NEW_ROOT.
3386          */
3387         right = alloc_tree_block_no_bg_flush(trans, root, 0, &disk_key, 0,
3388                                              l->start, 0, num_doubles ?
3389                                              BTRFS_NESTING_NEW_ROOT :
3390                                              BTRFS_NESTING_SPLIT);
3391         if (IS_ERR(right))
3392                 return PTR_ERR(right);
3393
3394         root_add_used(root, fs_info->nodesize);
3395
3396         if (split == 0) {
3397                 if (mid <= slot) {
3398                         btrfs_set_header_nritems(right, 0);
3399                         insert_ptr(trans, path, &disk_key,
3400                                    right->start, path->slots[1] + 1, 1);
3401                         btrfs_tree_unlock(path->nodes[0]);
3402                         free_extent_buffer(path->nodes[0]);
3403                         path->nodes[0] = right;
3404                         path->slots[0] = 0;
3405                         path->slots[1] += 1;
3406                 } else {
3407                         btrfs_set_header_nritems(right, 0);
3408                         insert_ptr(trans, path, &disk_key,
3409                                    right->start, path->slots[1], 1);
3410                         btrfs_tree_unlock(path->nodes[0]);
3411                         free_extent_buffer(path->nodes[0]);
3412                         path->nodes[0] = right;
3413                         path->slots[0] = 0;
3414                         if (path->slots[1] == 0)
3415                                 fixup_low_keys(path, &disk_key, 1);
3416                 }
3417                 /*
3418                  * We create a new leaf 'right' for the required ins_len and
3419                  * we'll do btrfs_mark_buffer_dirty() on this leaf after copying
3420                  * the content of ins_len to 'right'.
3421                  */
3422                 return ret;
3423         }
3424
3425         copy_for_split(trans, path, l, right, slot, mid, nritems);
3426
3427         if (split == 2) {
3428                 BUG_ON(num_doubles != 0);
3429                 num_doubles++;
3430                 goto again;
3431         }
3432
3433         return 0;
3434
3435 push_for_double:
3436         push_for_double_split(trans, root, path, data_size);
3437         tried_avoid_double = 1;
3438         if (btrfs_leaf_free_space(path->nodes[0]) >= data_size)
3439                 return 0;
3440         goto again;
3441 }
3442
3443 static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans,
3444                                          struct btrfs_root *root,
3445                                          struct btrfs_path *path, int ins_len)
3446 {
3447         struct btrfs_key key;
3448         struct extent_buffer *leaf;
3449         struct btrfs_file_extent_item *fi;
3450         u64 extent_len = 0;
3451         u32 item_size;
3452         int ret;
3453
3454         leaf = path->nodes[0];
3455         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3456
3457         BUG_ON(key.type != BTRFS_EXTENT_DATA_KEY &&
3458                key.type != BTRFS_EXTENT_CSUM_KEY);
3459
3460         if (btrfs_leaf_free_space(leaf) >= ins_len)
3461                 return 0;
3462
3463         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
3464         if (key.type == BTRFS_EXTENT_DATA_KEY) {
3465                 fi = btrfs_item_ptr(leaf, path->slots[0],
3466                                     struct btrfs_file_extent_item);
3467                 extent_len = btrfs_file_extent_num_bytes(leaf, fi);
3468         }
3469         btrfs_release_path(path);
3470
3471         path->keep_locks = 1;
3472         path->search_for_split = 1;
3473         ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
3474         path->search_for_split = 0;
3475         if (ret > 0)
3476                 ret = -EAGAIN;
3477         if (ret < 0)
3478                 goto err;
3479
3480         ret = -EAGAIN;
3481         leaf = path->nodes[0];
3482         /* if our item isn't there, return now */
3483         if (item_size != btrfs_item_size_nr(leaf, path->slots[0]))
3484                 goto err;
3485
3486         /* the leaf has  changed, it now has room.  return now */
3487         if (btrfs_leaf_free_space(path->nodes[0]) >= ins_len)
3488                 goto err;
3489
3490         if (key.type == BTRFS_EXTENT_DATA_KEY) {
3491                 fi = btrfs_item_ptr(leaf, path->slots[0],
3492                                     struct btrfs_file_extent_item);
3493                 if (extent_len != btrfs_file_extent_num_bytes(leaf, fi))
3494                         goto err;
3495         }
3496
3497         ret = split_leaf(trans, root, &key, path, ins_len, 1);
3498         if (ret)
3499                 goto err;
3500
3501         path->keep_locks = 0;
3502         btrfs_unlock_up_safe(path, 1);
3503         return 0;
3504 err:
3505         path->keep_locks = 0;
3506         return ret;
3507 }
3508
3509 static noinline int split_item(struct btrfs_path *path,
3510                                const struct btrfs_key *new_key,
3511                                unsigned long split_offset)
3512 {
3513         struct extent_buffer *leaf;
3514         struct btrfs_item *item;
3515         struct btrfs_item *new_item;
3516         int slot;
3517         char *buf;
3518         u32 nritems;
3519         u32 item_size;
3520         u32 orig_offset;
3521         struct btrfs_disk_key disk_key;
3522
3523         leaf = path->nodes[0];
3524         BUG_ON(btrfs_leaf_free_space(leaf) < sizeof(struct btrfs_item));
3525
3526         item = btrfs_item_nr(path->slots[0]);
3527         orig_offset = btrfs_item_offset(leaf, item);
3528         item_size = btrfs_item_size(leaf, item);
3529
3530         buf = kmalloc(item_size, GFP_NOFS);
3531         if (!buf)
3532                 return -ENOMEM;
3533
3534         read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf,
3535                             path->slots[0]), item_size);
3536
3537         slot = path->slots[0] + 1;
3538         nritems = btrfs_header_nritems(leaf);
3539         if (slot != nritems) {
3540                 /* shift the items */
3541                 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1),
3542                                 btrfs_item_nr_offset(slot),
3543                                 (nritems - slot) * sizeof(struct btrfs_item));
3544         }
3545
3546         btrfs_cpu_key_to_disk(&disk_key, new_key);
3547         btrfs_set_item_key(leaf, &disk_key, slot);
3548
3549         new_item = btrfs_item_nr(slot);
3550
3551         btrfs_set_item_offset(leaf, new_item, orig_offset);
3552         btrfs_set_item_size(leaf, new_item, item_size - split_offset);
3553
3554         btrfs_set_item_offset(leaf, item,
3555                               orig_offset + item_size - split_offset);
3556         btrfs_set_item_size(leaf, item, split_offset);
3557
3558         btrfs_set_header_nritems(leaf, nritems + 1);
3559
3560         /* write the data for the start of the original item */
3561         write_extent_buffer(leaf, buf,
3562                             btrfs_item_ptr_offset(leaf, path->slots[0]),
3563                             split_offset);
3564
3565         /* write the data for the new item */
3566         write_extent_buffer(leaf, buf + split_offset,
3567                             btrfs_item_ptr_offset(leaf, slot),
3568                             item_size - split_offset);
3569         btrfs_mark_buffer_dirty(leaf);
3570
3571         BUG_ON(btrfs_leaf_free_space(leaf) < 0);
3572         kfree(buf);
3573         return 0;
3574 }
3575
3576 /*
3577  * This function splits a single item into two items,
3578  * giving 'new_key' to the new item and splitting the
3579  * old one at split_offset (from the start of the item).
3580  *
3581  * The path may be released by this operation.  After
3582  * the split, the path is pointing to the old item.  The
3583  * new item is going to be in the same node as the old one.
3584  *
3585  * Note, the item being split must be smaller enough to live alone on
3586  * a tree block with room for one extra struct btrfs_item
3587  *
3588  * This allows us to split the item in place, keeping a lock on the
3589  * leaf the entire time.
3590  */
3591 int btrfs_split_item(struct btrfs_trans_handle *trans,
3592                      struct btrfs_root *root,
3593                      struct btrfs_path *path,
3594                      const struct btrfs_key *new_key,
3595                      unsigned long split_offset)
3596 {
3597         int ret;
3598         ret = setup_leaf_for_split(trans, root, path,
3599                                    sizeof(struct btrfs_item));
3600         if (ret)
3601                 return ret;
3602
3603         ret = split_item(path, new_key, split_offset);
3604         return ret;
3605 }
3606
3607 /*
3608  * This function duplicate a item, giving 'new_key' to the new item.
3609  * It guarantees both items live in the same tree leaf and the new item
3610  * is contiguous with the original item.
3611  *
3612  * This allows us to split file extent in place, keeping a lock on the
3613  * leaf the entire time.
3614  */
3615 int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
3616                          struct btrfs_root *root,
3617                          struct btrfs_path *path,
3618                          const struct btrfs_key *new_key)
3619 {
3620         struct extent_buffer *leaf;
3621         int ret;
3622         u32 item_size;
3623
3624         leaf = path->nodes[0];
3625         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
3626         ret = setup_leaf_for_split(trans, root, path,
3627                                    item_size + sizeof(struct btrfs_item));
3628         if (ret)
3629                 return ret;
3630
3631         path->slots[0]++;
3632         setup_items_for_insert(root, path, new_key, &item_size, 1);
3633         leaf = path->nodes[0];
3634         memcpy_extent_buffer(leaf,
3635                              btrfs_item_ptr_offset(leaf, path->slots[0]),
3636                              btrfs_item_ptr_offset(leaf, path->slots[0] - 1),
3637                              item_size);
3638         return 0;
3639 }
3640
3641 /*
3642  * make the item pointed to by the path smaller.  new_size indicates
3643  * how small to make it, and from_end tells us if we just chop bytes
3644  * off the end of the item or if we shift the item to chop bytes off
3645  * the front.
3646  */
3647 void btrfs_truncate_item(struct btrfs_path *path, u32 new_size, int from_end)
3648 {
3649         int slot;
3650         struct extent_buffer *leaf;
3651         struct btrfs_item *item;
3652         u32 nritems;
3653         unsigned int data_end;
3654         unsigned int old_data_start;
3655         unsigned int old_size;
3656         unsigned int size_diff;
3657         int i;
3658         struct btrfs_map_token token;
3659
3660         leaf = path->nodes[0];
3661         slot = path->slots[0];
3662
3663         old_size = btrfs_item_size_nr(leaf, slot);
3664         if (old_size == new_size)
3665                 return;
3666
3667         nritems = btrfs_header_nritems(leaf);
3668         data_end = leaf_data_end(leaf);
3669
3670         old_data_start = btrfs_item_offset_nr(leaf, slot);
3671
3672         size_diff = old_size - new_size;
3673
3674         BUG_ON(slot < 0);
3675         BUG_ON(slot >= nritems);
3676
3677         /*
3678          * item0..itemN ... dataN.offset..dataN.size .. data0.size
3679          */
3680         /* first correct the data pointers */
3681         btrfs_init_map_token(&token, leaf);
3682         for (i = slot; i < nritems; i++) {
3683                 u32 ioff;
3684                 item = btrfs_item_nr(i);
3685
3686                 ioff = btrfs_token_item_offset(&token, item);
3687                 btrfs_set_token_item_offset(&token, item, ioff + size_diff);
3688         }
3689
3690         /* shift the data */
3691         if (from_end) {
3692                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
3693                               data_end + size_diff, BTRFS_LEAF_DATA_OFFSET +
3694                               data_end, old_data_start + new_size - data_end);
3695         } else {
3696                 struct btrfs_disk_key disk_key;
3697                 u64 offset;
3698
3699                 btrfs_item_key(leaf, &disk_key, slot);
3700
3701                 if (btrfs_disk_key_type(&disk_key) == BTRFS_EXTENT_DATA_KEY) {
3702                         unsigned long ptr;
3703                         struct btrfs_file_extent_item *fi;
3704
3705                         fi = btrfs_item_ptr(leaf, slot,
3706                                             struct btrfs_file_extent_item);
3707                         fi = (struct btrfs_file_extent_item *)(
3708                              (unsigned long)fi - size_diff);
3709
3710                         if (btrfs_file_extent_type(leaf, fi) ==
3711                             BTRFS_FILE_EXTENT_INLINE) {
3712                                 ptr = btrfs_item_ptr_offset(leaf, slot);
3713                                 memmove_extent_buffer(leaf, ptr,
3714                                       (unsigned long)fi,
3715                                       BTRFS_FILE_EXTENT_INLINE_DATA_START);
3716                         }
3717                 }
3718
3719                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
3720                               data_end + size_diff, BTRFS_LEAF_DATA_OFFSET +
3721                               data_end, old_data_start - data_end);
3722
3723                 offset = btrfs_disk_key_offset(&disk_key);
3724                 btrfs_set_disk_key_offset(&disk_key, offset + size_diff);
3725                 btrfs_set_item_key(leaf, &disk_key, slot);
3726                 if (slot == 0)
3727                         fixup_low_keys(path, &disk_key, 1);
3728         }
3729
3730         item = btrfs_item_nr(slot);
3731         btrfs_set_item_size(leaf, item, new_size);
3732         btrfs_mark_buffer_dirty(leaf);
3733
3734         if (btrfs_leaf_free_space(leaf) < 0) {
3735                 btrfs_print_leaf(leaf);
3736                 BUG();
3737         }
3738 }
3739
3740 /*
3741  * make the item pointed to by the path bigger, data_size is the added size.
3742  */
3743 void btrfs_extend_item(struct btrfs_path *path, u32 data_size)
3744 {
3745         int slot;
3746         struct extent_buffer *leaf;
3747         struct btrfs_item *item;
3748         u32 nritems;
3749         unsigned int data_end;
3750         unsigned int old_data;
3751         unsigned int old_size;
3752         int i;
3753         struct btrfs_map_token token;
3754
3755         leaf = path->nodes[0];
3756
3757         nritems = btrfs_header_nritems(leaf);
3758         data_end = leaf_data_end(leaf);
3759
3760         if (btrfs_leaf_free_space(leaf) < data_size) {
3761                 btrfs_print_leaf(leaf);
3762                 BUG();
3763         }
3764         slot = path->slots[0];
3765         old_data = btrfs_item_end_nr(leaf, slot);
3766
3767         BUG_ON(slot < 0);
3768         if (slot >= nritems) {
3769                 btrfs_print_leaf(leaf);
3770                 btrfs_crit(leaf->fs_info, "slot %d too large, nritems %d",
3771                            slot, nritems);
3772                 BUG();
3773         }
3774
3775         /*
3776          * item0..itemN ... dataN.offset..dataN.size .. data0.size
3777          */
3778         /* first correct the data pointers */
3779         btrfs_init_map_token(&token, leaf);
3780         for (i = slot; i < nritems; i++) {
3781                 u32 ioff;
3782                 item = btrfs_item_nr(i);
3783
3784                 ioff = btrfs_token_item_offset(&token, item);
3785                 btrfs_set_token_item_offset(&token, item, ioff - data_size);
3786         }
3787
3788         /* shift the data */
3789         memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
3790                       data_end - data_size, BTRFS_LEAF_DATA_OFFSET +
3791                       data_end, old_data - data_end);
3792
3793         data_end = old_data;
3794         old_size = btrfs_item_size_nr(leaf, slot);
3795         item = btrfs_item_nr(slot);
3796         btrfs_set_item_size(leaf, item, old_size + data_size);
3797         btrfs_mark_buffer_dirty(leaf);
3798
3799         if (btrfs_leaf_free_space(leaf) < 0) {
3800                 btrfs_print_leaf(leaf);
3801                 BUG();
3802         }
3803 }
3804
3805 /**
3806  * setup_items_for_insert - Helper called before inserting one or more items
3807  * to a leaf. Main purpose is to save stack depth by doing the bulk of the work
3808  * in a function that doesn't call btrfs_search_slot
3809  *
3810  * @root:       root we are inserting items to
3811  * @path:       points to the leaf/slot where we are going to insert new items
3812  * @cpu_key:    array of keys for items to be inserted
3813  * @data_size:  size of the body of each item we are going to insert
3814  * @nr:         size of @cpu_key/@data_size arrays
3815  */
3816 void setup_items_for_insert(struct btrfs_root *root, struct btrfs_path *path,
3817                             const struct btrfs_key *cpu_key, u32 *data_size,
3818                             int nr)
3819 {
3820         struct btrfs_fs_info *fs_info = root->fs_info;
3821         struct btrfs_item *item;
3822         int i;
3823         u32 nritems;
3824         unsigned int data_end;
3825         struct btrfs_disk_key disk_key;
3826         struct extent_buffer *leaf;
3827         int slot;
3828         struct btrfs_map_token token;
3829         u32 total_size;
3830         u32 total_data = 0;
3831
3832         for (i = 0; i < nr; i++)
3833                 total_data += data_size[i];
3834         total_size = total_data + (nr * sizeof(struct btrfs_item));
3835
3836         if (path->slots[0] == 0) {
3837                 btrfs_cpu_key_to_disk(&disk_key, cpu_key);
3838                 fixup_low_keys(path, &disk_key, 1);
3839         }
3840         btrfs_unlock_up_safe(path, 1);
3841
3842         leaf = path->nodes[0];
3843         slot = path->slots[0];
3844
3845         nritems = btrfs_header_nritems(leaf);
3846         data_end = leaf_data_end(leaf);
3847
3848         if (btrfs_leaf_free_space(leaf) < total_size) {
3849                 btrfs_print_leaf(leaf);
3850                 btrfs_crit(fs_info, "not enough freespace need %u have %d",
3851                            total_size, btrfs_leaf_free_space(leaf));
3852                 BUG();
3853         }
3854
3855         btrfs_init_map_token(&token, leaf);
3856         if (slot != nritems) {
3857                 unsigned int old_data = btrfs_item_end_nr(leaf, slot);
3858
3859                 if (old_data < data_end) {
3860                         btrfs_print_leaf(leaf);
3861                         btrfs_crit(fs_info,
3862                 "item at slot %d with data offset %u beyond data end of leaf %u",
3863                                    slot, old_data, data_end);
3864                         BUG();
3865                 }
3866                 /*
3867                  * item0..itemN ... dataN.offset..dataN.size .. data0.size
3868                  */
3869                 /* first correct the data pointers */
3870                 for (i = slot; i < nritems; i++) {
3871                         u32 ioff;
3872
3873                         item = btrfs_item_nr(i);
3874                         ioff = btrfs_token_item_offset(&token, item);
3875                         btrfs_set_token_item_offset(&token, item,
3876                                                     ioff - total_data);
3877                 }
3878                 /* shift the items */
3879                 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr),
3880                               btrfs_item_nr_offset(slot),
3881                               (nritems - slot) * sizeof(struct btrfs_item));
3882
3883                 /* shift the data */
3884                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
3885                               data_end - total_data, BTRFS_LEAF_DATA_OFFSET +
3886                               data_end, old_data - data_end);
3887                 data_end = old_data;
3888         }
3889
3890         /* setup the item for the new data */
3891         for (i = 0; i < nr; i++) {
3892                 btrfs_cpu_key_to_disk(&disk_key, cpu_key + i);
3893                 btrfs_set_item_key(leaf, &disk_key, slot + i);
3894                 item = btrfs_item_nr(slot + i);
3895                 data_end -= data_size[i];
3896                 btrfs_set_token_item_offset(&token, item, data_end);
3897                 btrfs_set_token_item_size(&token, item, data_size[i]);
3898         }
3899
3900         btrfs_set_header_nritems(leaf, nritems + nr);
3901         btrfs_mark_buffer_dirty(leaf);
3902
3903         if (btrfs_leaf_free_space(leaf) < 0) {
3904                 btrfs_print_leaf(leaf);
3905                 BUG();
3906         }
3907 }
3908
3909 /*
3910  * Given a key and some data, insert items into the tree.
3911  * This does all the path init required, making room in the tree if needed.
3912  */
3913 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
3914                             struct btrfs_root *root,
3915                             struct btrfs_path *path,
3916                             const struct btrfs_key *cpu_key, u32 *data_size,
3917                             int nr)
3918 {
3919         int ret = 0;
3920         int slot;
3921         int i;
3922         u32 total_size = 0;
3923         u32 total_data = 0;
3924
3925         for (i = 0; i < nr; i++)
3926                 total_data += data_size[i];
3927
3928         total_size = total_data + (nr * sizeof(struct btrfs_item));
3929         ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1);
3930         if (ret == 0)
3931                 return -EEXIST;
3932         if (ret < 0)
3933                 return ret;
3934
3935         slot = path->slots[0];
3936         BUG_ON(slot < 0);
3937
3938         setup_items_for_insert(root, path, cpu_key, data_size, nr);
3939         return 0;
3940 }
3941
3942 /*
3943  * Given a key and some data, insert an item into the tree.
3944  * This does all the path init required, making room in the tree if needed.
3945  */
3946 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3947                       const struct btrfs_key *cpu_key, void *data,
3948                       u32 data_size)
3949 {
3950         int ret = 0;
3951         struct btrfs_path *path;
3952         struct extent_buffer *leaf;
3953         unsigned long ptr;
3954
3955         path = btrfs_alloc_path();
3956         if (!path)
3957                 return -ENOMEM;
3958         ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
3959         if (!ret) {
3960                 leaf = path->nodes[0];
3961                 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
3962                 write_extent_buffer(leaf, data, ptr, data_size);
3963                 btrfs_mark_buffer_dirty(leaf);
3964         }
3965         btrfs_free_path(path);
3966         return ret;
3967 }
3968
3969 /*
3970  * delete the pointer from a given node.
3971  *
3972  * the tree should have been previously balanced so the deletion does not
3973  * empty a node.
3974  */
3975 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
3976                     int level, int slot)
3977 {
3978         struct extent_buffer *parent = path->nodes[level];
3979         u32 nritems;
3980         int ret;
3981
3982         nritems = btrfs_header_nritems(parent);
3983         if (slot != nritems - 1) {
3984                 if (level) {
3985                         ret = btrfs_tree_mod_log_insert_move(parent, slot,
3986                                         slot + 1, nritems - slot - 1);
3987                         BUG_ON(ret < 0);
3988                 }
3989                 memmove_extent_buffer(parent,
3990                               btrfs_node_key_ptr_offset(slot),
3991                               btrfs_node_key_ptr_offset(slot + 1),
3992                               sizeof(struct btrfs_key_ptr) *
3993                               (nritems - slot - 1));
3994         } else if (level) {
3995                 ret = btrfs_tree_mod_log_insert_key(parent, slot,
3996                                 BTRFS_MOD_LOG_KEY_REMOVE, GFP_NOFS);
3997                 BUG_ON(ret < 0);
3998         }
3999
4000         nritems--;
4001         btrfs_set_header_nritems(parent, nritems);
4002         if (nritems == 0 && parent == root->node) {
4003                 BUG_ON(btrfs_header_level(root->node) != 1);
4004                 /* just turn the root into a leaf and break */
4005                 btrfs_set_header_level(root->node, 0);
4006         } else if (slot == 0) {
4007                 struct btrfs_disk_key disk_key;
4008
4009                 btrfs_node_key(parent, &disk_key, 0);
4010                 fixup_low_keys(path, &disk_key, level + 1);
4011         }
4012         btrfs_mark_buffer_dirty(parent);
4013 }
4014
4015 /*
4016  * a helper function to delete the leaf pointed to by path->slots[1] and
4017  * path->nodes[1].
4018  *
4019  * This deletes the pointer in path->nodes[1] and frees the leaf
4020  * block extent.  zero is returned if it all worked out, < 0 otherwise.
4021  *
4022  * The path must have already been setup for deleting the leaf, including
4023  * all the proper balancing.  path->nodes[1] must be locked.
4024  */
4025 static noinline void btrfs_del_leaf(struct btrfs_trans_handle *trans,
4026                                     struct btrfs_root *root,
4027                                     struct btrfs_path *path,
4028                                     struct extent_buffer *leaf)
4029 {
4030         WARN_ON(btrfs_header_generation(leaf) != trans->transid);
4031         del_ptr(root, path, 1, path->slots[1]);
4032
4033         /*
4034          * btrfs_free_extent is expensive, we want to make sure we
4035          * aren't holding any locks when we call it
4036          */
4037         btrfs_unlock_up_safe(path, 0);
4038
4039         root_sub_used(root, leaf->len);
4040
4041         atomic_inc(&leaf->refs);
4042         btrfs_free_tree_block(trans, root, leaf, 0, 1);
4043         free_extent_buffer_stale(leaf);
4044 }
4045 /*
4046  * delete the item at the leaf level in path.  If that empties
4047  * the leaf, remove it from the tree
4048  */
4049 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
4050                     struct btrfs_path *path, int slot, int nr)
4051 {
4052         struct btrfs_fs_info *fs_info = root->fs_info;
4053         struct extent_buffer *leaf;
4054         struct btrfs_item *item;
4055         u32 last_off;
4056         u32 dsize = 0;
4057         int ret = 0;
4058         int wret;
4059         int i;
4060         u32 nritems;
4061
4062         leaf = path->nodes[0];
4063         last_off = btrfs_item_offset_nr(leaf, slot + nr - 1);
4064
4065         for (i = 0; i < nr; i++)
4066                 dsize += btrfs_item_size_nr(leaf, slot + i);
4067
4068         nritems = btrfs_header_nritems(leaf);
4069
4070         if (slot + nr != nritems) {
4071                 int data_end = leaf_data_end(leaf);
4072                 struct btrfs_map_token token;
4073
4074                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
4075                               data_end + dsize,
4076                               BTRFS_LEAF_DATA_OFFSET + data_end,
4077                               last_off - data_end);
4078
4079                 btrfs_init_map_token(&token, leaf);
4080                 for (i = slot + nr; i < nritems; i++) {
4081                         u32 ioff;
4082
4083                         item = btrfs_item_nr(i);
4084                         ioff = btrfs_token_item_offset(&token, item);
4085                         btrfs_set_token_item_offset(&token, item, ioff + dsize);
4086                 }
4087
4088                 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
4089                               btrfs_item_nr_offset(slot + nr),
4090                               sizeof(struct btrfs_item) *
4091                               (nritems - slot - nr));
4092         }
4093         btrfs_set_header_nritems(leaf, nritems - nr);
4094         nritems -= nr;
4095
4096         /* delete the leaf if we've emptied it */
4097         if (nritems == 0) {
4098                 if (leaf == root->node) {
4099                         btrfs_set_header_level(leaf, 0);
4100                 } else {
4101                         btrfs_clean_tree_block(leaf);
4102                         btrfs_del_leaf(trans, root, path, leaf);
4103                 }
4104         } else {
4105                 int used = leaf_space_used(leaf, 0, nritems);
4106                 if (slot == 0) {
4107                         struct btrfs_disk_key disk_key;
4108
4109                         btrfs_item_key(leaf, &disk_key, 0);
4110                         fixup_low_keys(path, &disk_key, 1);
4111                 }
4112
4113                 /* delete the leaf if it is mostly empty */
4114                 if (used < BTRFS_LEAF_DATA_SIZE(fs_info) / 3) {
4115                         /* push_leaf_left fixes the path.
4116                          * make sure the path still points to our leaf
4117                          * for possible call to del_ptr below
4118                          */
4119                         slot = path->slots[1];
4120                         atomic_inc(&leaf->refs);
4121
4122                         wret = push_leaf_left(trans, root, path, 1, 1,
4123                                               1, (u32)-1);
4124                         if (wret < 0 && wret != -ENOSPC)
4125                                 ret = wret;
4126
4127                         if (path->nodes[0] == leaf &&
4128                             btrfs_header_nritems(leaf)) {
4129                                 wret = push_leaf_right(trans, root, path, 1,
4130                                                        1, 1, 0);
4131                                 if (wret < 0 && wret != -ENOSPC)
4132                                         ret = wret;
4133                         }
4134
4135                         if (btrfs_header_nritems(leaf) == 0) {
4136                                 path->slots[1] = slot;
4137                                 btrfs_del_leaf(trans, root, path, leaf);
4138                                 free_extent_buffer(leaf);
4139                                 ret = 0;
4140                         } else {
4141                                 /* if we're still in the path, make sure
4142                                  * we're dirty.  Otherwise, one of the
4143                                  * push_leaf functions must have already
4144                                  * dirtied this buffer
4145                                  */
4146                                 if (path->nodes[0] == leaf)
4147                                         btrfs_mark_buffer_dirty(leaf);
4148                                 free_extent_buffer(leaf);
4149                         }
4150                 } else {
4151                         btrfs_mark_buffer_dirty(leaf);
4152                 }
4153         }
4154         return ret;
4155 }
4156
4157 /*
4158  * search the tree again to find a leaf with lesser keys
4159  * returns 0 if it found something or 1 if there are no lesser leaves.
4160  * returns < 0 on io errors.
4161  *
4162  * This may release the path, and so you may lose any locks held at the
4163  * time you call it.
4164  */
4165 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
4166 {
4167         struct btrfs_key key;
4168         struct btrfs_disk_key found_key;
4169         int ret;
4170
4171         btrfs_item_key_to_cpu(path->nodes[0], &key, 0);
4172
4173         if (key.offset > 0) {
4174                 key.offset--;
4175         } else if (key.type > 0) {
4176                 key.type--;
4177                 key.offset = (u64)-1;
4178         } else if (key.objectid > 0) {
4179                 key.objectid--;
4180                 key.type = (u8)-1;
4181                 key.offset = (u64)-1;
4182         } else {
4183                 return 1;
4184         }
4185
4186         btrfs_release_path(path);
4187         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
4188         if (ret < 0)
4189                 return ret;
4190         btrfs_item_key(path->nodes[0], &found_key, 0);
4191         ret = comp_keys(&found_key, &key);
4192         /*
4193          * We might have had an item with the previous key in the tree right
4194          * before we released our path. And after we released our path, that
4195          * item might have been pushed to the first slot (0) of the leaf we
4196          * were holding due to a tree balance. Alternatively, an item with the
4197          * previous key can exist as the only element of a leaf (big fat item).
4198          * Therefore account for these 2 cases, so that our callers (like
4199          * btrfs_previous_item) don't miss an existing item with a key matching
4200          * the previous key we computed above.
4201          */
4202         if (ret <= 0)
4203                 return 0;
4204         return 1;
4205 }
4206
4207 /*
4208  * A helper function to walk down the tree starting at min_key, and looking
4209  * for nodes or leaves that are have a minimum transaction id.
4210  * This is used by the btree defrag code, and tree logging
4211  *
4212  * This does not cow, but it does stuff the starting key it finds back
4213  * into min_key, so you can call btrfs_search_slot with cow=1 on the
4214  * key and get a writable path.
4215  *
4216  * This honors path->lowest_level to prevent descent past a given level
4217  * of the tree.
4218  *
4219  * min_trans indicates the oldest transaction that you are interested
4220  * in walking through.  Any nodes or leaves older than min_trans are
4221  * skipped over (without reading them).
4222  *
4223  * returns zero if something useful was found, < 0 on error and 1 if there
4224  * was nothing in the tree that matched the search criteria.
4225  */
4226 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
4227                          struct btrfs_path *path,
4228                          u64 min_trans)
4229 {
4230         struct extent_buffer *cur;
4231         struct btrfs_key found_key;
4232         int slot;
4233         int sret;
4234         u32 nritems;
4235         int level;
4236         int ret = 1;
4237         int keep_locks = path->keep_locks;
4238
4239         path->keep_locks = 1;
4240 again:
4241         cur = btrfs_read_lock_root_node(root);
4242         level = btrfs_header_level(cur);
4243         WARN_ON(path->nodes[level]);
4244         path->nodes[level] = cur;
4245         path->locks[level] = BTRFS_READ_LOCK;
4246
4247         if (btrfs_header_generation(cur) < min_trans) {
4248                 ret = 1;
4249                 goto out;
4250         }
4251         while (1) {
4252                 nritems = btrfs_header_nritems(cur);
4253                 level = btrfs_header_level(cur);
4254                 sret = btrfs_bin_search(cur, min_key, &slot);
4255                 if (sret < 0) {
4256                         ret = sret;
4257                         goto out;
4258                 }
4259
4260                 /* at the lowest level, we're done, setup the path and exit */
4261                 if (level == path->lowest_level) {
4262                         if (slot >= nritems)
4263                                 goto find_next_key;
4264                         ret = 0;
4265                         path->slots[level] = slot;
4266                         btrfs_item_key_to_cpu(cur, &found_key, slot);
4267                         goto out;
4268                 }
4269                 if (sret && slot > 0)
4270                         slot--;
4271                 /*
4272                  * check this node pointer against the min_trans parameters.
4273                  * If it is too old, skip to the next one.
4274                  */
4275                 while (slot < nritems) {
4276                         u64 gen;
4277
4278                         gen = btrfs_node_ptr_generation(cur, slot);
4279                         if (gen < min_trans) {
4280                                 slot++;
4281                                 continue;
4282                         }
4283                         break;
4284                 }
4285 find_next_key:
4286                 /*
4287                  * we didn't find a candidate key in this node, walk forward
4288                  * and find another one
4289                  */
4290                 if (slot >= nritems) {
4291                         path->slots[level] = slot;
4292                         sret = btrfs_find_next_key(root, path, min_key, level,
4293                                                   min_trans);
4294                         if (sret == 0) {
4295                                 btrfs_release_path(path);
4296                                 goto again;
4297                         } else {
4298                                 goto out;
4299                         }
4300                 }
4301                 /* save our key for returning back */
4302                 btrfs_node_key_to_cpu(cur, &found_key, slot);
4303                 path->slots[level] = slot;
4304                 if (level == path->lowest_level) {
4305                         ret = 0;
4306                         goto out;
4307                 }
4308                 cur = btrfs_read_node_slot(cur, slot);
4309                 if (IS_ERR(cur)) {
4310                         ret = PTR_ERR(cur);
4311                         goto out;
4312                 }
4313
4314                 btrfs_tree_read_lock(cur);
4315
4316                 path->locks[level - 1] = BTRFS_READ_LOCK;
4317                 path->nodes[level - 1] = cur;
4318                 unlock_up(path, level, 1, 0, NULL);
4319         }
4320 out:
4321         path->keep_locks = keep_locks;
4322         if (ret == 0) {
4323                 btrfs_unlock_up_safe(path, path->lowest_level + 1);
4324                 memcpy(min_key, &found_key, sizeof(found_key));
4325         }
4326         return ret;
4327 }
4328
4329 /*
4330  * this is similar to btrfs_next_leaf, but does not try to preserve
4331  * and fixup the path.  It looks for and returns the next key in the
4332  * tree based on the current path and the min_trans parameters.
4333  *
4334  * 0 is returned if another key is found, < 0 if there are any errors
4335  * and 1 is returned if there are no higher keys in the tree
4336  *
4337  * path->keep_locks should be set to 1 on the search made before
4338  * calling this function.
4339  */
4340 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
4341                         struct btrfs_key *key, int level, u64 min_trans)
4342 {
4343         int slot;
4344         struct extent_buffer *c;
4345
4346         WARN_ON(!path->keep_locks && !path->skip_locking);
4347         while (level < BTRFS_MAX_LEVEL) {
4348                 if (!path->nodes[level])
4349                         return 1;
4350
4351                 slot = path->slots[level] + 1;
4352                 c = path->nodes[level];
4353 next:
4354                 if (slot >= btrfs_header_nritems(c)) {
4355                         int ret;
4356                         int orig_lowest;
4357                         struct btrfs_key cur_key;
4358                         if (level + 1 >= BTRFS_MAX_LEVEL ||
4359                             !path->nodes[level + 1])
4360                                 return 1;
4361
4362                         if (path->locks[level + 1] || path->skip_locking) {
4363                                 level++;
4364                                 continue;
4365                         }
4366
4367                         slot = btrfs_header_nritems(c) - 1;
4368                         if (level == 0)
4369                                 btrfs_item_key_to_cpu(c, &cur_key, slot);
4370                         else
4371                                 btrfs_node_key_to_cpu(c, &cur_key, slot);
4372
4373                         orig_lowest = path->lowest_level;
4374                         btrfs_release_path(path);
4375                         path->lowest_level = level;
4376                         ret = btrfs_search_slot(NULL, root, &cur_key, path,
4377                                                 0, 0);
4378                         path->lowest_level = orig_lowest;
4379                         if (ret < 0)
4380                                 return ret;
4381
4382                         c = path->nodes[level];
4383                         slot = path->slots[level];
4384                         if (ret == 0)
4385                                 slot++;
4386                         goto next;
4387                 }
4388
4389                 if (level == 0)
4390                         btrfs_item_key_to_cpu(c, key, slot);
4391                 else {
4392                         u64 gen = btrfs_node_ptr_generation(c, slot);
4393
4394                         if (gen < min_trans) {
4395                                 slot++;
4396                                 goto next;
4397                         }
4398                         btrfs_node_key_to_cpu(c, key, slot);
4399                 }
4400                 return 0;
4401         }
4402         return 1;
4403 }
4404
4405 /*
4406  * search the tree again to find a leaf with greater keys
4407  * returns 0 if it found something or 1 if there are no greater leaves.
4408  * returns < 0 on io errors.
4409  */
4410 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
4411 {
4412         return btrfs_next_old_leaf(root, path, 0);
4413 }
4414
4415 int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
4416                         u64 time_seq)
4417 {
4418         int slot;
4419         int level;
4420         struct extent_buffer *c;
4421         struct extent_buffer *next;
4422         struct btrfs_key key;
4423         u32 nritems;
4424         int ret;
4425         int i;
4426
4427         nritems = btrfs_header_nritems(path->nodes[0]);
4428         if (nritems == 0)
4429                 return 1;
4430
4431         btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
4432 again:
4433         level = 1;
4434         next = NULL;
4435         btrfs_release_path(path);
4436
4437         path->keep_locks = 1;
4438
4439         if (time_seq)
4440                 ret = btrfs_search_old_slot(root, &key, path, time_seq);
4441         else
4442                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
4443         path->keep_locks = 0;
4444
4445         if (ret < 0)
4446                 return ret;
4447
4448         nritems = btrfs_header_nritems(path->nodes[0]);
4449         /*
4450          * by releasing the path above we dropped all our locks.  A balance
4451          * could have added more items next to the key that used to be
4452          * at the very end of the block.  So, check again here and
4453          * advance the path if there are now more items available.
4454          */
4455         if (nritems > 0 && path->slots[0] < nritems - 1) {
4456                 if (ret == 0)
4457                         path->slots[0]++;
4458                 ret = 0;
4459                 goto done;
4460         }
4461         /*
4462          * So the above check misses one case:
4463          * - after releasing the path above, someone has removed the item that
4464          *   used to be at the very end of the block, and balance between leafs
4465          *   gets another one with bigger key.offset to replace it.
4466          *
4467          * This one should be returned as well, or we can get leaf corruption
4468          * later(esp. in __btrfs_drop_extents()).
4469          *
4470          * And a bit more explanation about this check,
4471          * with ret > 0, the key isn't found, the path points to the slot
4472          * where it should be inserted, so the path->slots[0] item must be the
4473          * bigger one.
4474          */
4475         if (nritems > 0 && ret > 0 && path->slots[0] == nritems - 1) {
4476                 ret = 0;
4477                 goto done;
4478         }
4479
4480         while (level < BTRFS_MAX_LEVEL) {
4481                 if (!path->nodes[level]) {
4482                         ret = 1;
4483                         goto done;
4484                 }
4485
4486                 slot = path->slots[level] + 1;
4487                 c = path->nodes[level];
4488                 if (slot >= btrfs_header_nritems(c)) {
4489                         level++;
4490                         if (level == BTRFS_MAX_LEVEL) {
4491                                 ret = 1;
4492                                 goto done;
4493                         }
4494                         continue;
4495                 }
4496
4497
4498                 /*
4499                  * Our current level is where we're going to start from, and to
4500                  * make sure lockdep doesn't complain we need to drop our locks
4501                  * and nodes from 0 to our current level.
4502                  */
4503                 for (i = 0; i < level; i++) {
4504                         if (path->locks[level]) {
4505                                 btrfs_tree_read_unlock(path->nodes[i]);
4506                                 path->locks[i] = 0;
4507                         }
4508                         free_extent_buffer(path->nodes[i]);
4509                         path->nodes[i] = NULL;
4510                 }
4511
4512                 next = c;
4513                 ret = read_block_for_search(root, path, &next, level,
4514                                             slot, &key);
4515                 if (ret == -EAGAIN)
4516                         goto again;
4517
4518                 if (ret < 0) {
4519                         btrfs_release_path(path);
4520                         goto done;
4521                 }
4522
4523                 if (!path->skip_locking) {
4524                         ret = btrfs_try_tree_read_lock(next);
4525                         if (!ret && time_seq) {
4526                                 /*
4527                                  * If we don't get the lock, we may be racing
4528                                  * with push_leaf_left, holding that lock while
4529                                  * itself waiting for the leaf we've currently
4530                                  * locked. To solve this situation, we give up
4531                                  * on our lock and cycle.
4532                                  */
4533                                 free_extent_buffer(next);
4534                                 btrfs_release_path(path);
4535                                 cond_resched();
4536                                 goto again;
4537                         }
4538                         if (!ret)
4539                                 btrfs_tree_read_lock(next);
4540                 }
4541                 break;
4542         }
4543         path->slots[level] = slot;
4544         while (1) {
4545                 level--;
4546                 path->nodes[level] = next;
4547                 path->slots[level] = 0;
4548                 if (!path->skip_locking)
4549                         path->locks[level] = BTRFS_READ_LOCK;
4550                 if (!level)
4551                         break;
4552
4553                 ret = read_block_for_search(root, path, &next, level,
4554                                             0, &key);
4555                 if (ret == -EAGAIN)
4556                         goto again;
4557
4558                 if (ret < 0) {
4559                         btrfs_release_path(path);
4560                         goto done;
4561                 }
4562
4563                 if (!path->skip_locking)
4564                         btrfs_tree_read_lock(next);
4565         }
4566         ret = 0;
4567 done:
4568         unlock_up(path, 0, 1, 0, NULL);
4569
4570         return ret;
4571 }
4572
4573 /*
4574  * this uses btrfs_prev_leaf to walk backwards in the tree, and keeps
4575  * searching until it gets past min_objectid or finds an item of 'type'
4576  *
4577  * returns 0 if something is found, 1 if nothing was found and < 0 on error
4578  */
4579 int btrfs_previous_item(struct btrfs_root *root,
4580                         struct btrfs_path *path, u64 min_objectid,
4581                         int type)
4582 {
4583         struct btrfs_key found_key;
4584         struct extent_buffer *leaf;
4585         u32 nritems;
4586         int ret;
4587
4588         while (1) {
4589                 if (path->slots[0] == 0) {
4590                         ret = btrfs_prev_leaf(root, path);
4591                         if (ret != 0)
4592                                 return ret;
4593                 } else {
4594                         path->slots[0]--;
4595                 }
4596                 leaf = path->nodes[0];
4597                 nritems = btrfs_header_nritems(leaf);
4598                 if (nritems == 0)
4599                         return 1;
4600                 if (path->slots[0] == nritems)
4601                         path->slots[0]--;
4602
4603                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4604                 if (found_key.objectid < min_objectid)
4605                         break;
4606                 if (found_key.type == type)
4607                         return 0;
4608                 if (found_key.objectid == min_objectid &&
4609                     found_key.type < type)
4610                         break;
4611         }
4612         return 1;
4613 }
4614
4615 /*
4616  * search in extent tree to find a previous Metadata/Data extent item with
4617  * min objecitd.
4618  *
4619  * returns 0 if something is found, 1 if nothing was found and < 0 on error
4620  */
4621 int btrfs_previous_extent_item(struct btrfs_root *root,
4622                         struct btrfs_path *path, u64 min_objectid)
4623 {
4624         struct btrfs_key found_key;
4625         struct extent_buffer *leaf;
4626         u32 nritems;
4627         int ret;
4628
4629         while (1) {
4630                 if (path->slots[0] == 0) {
4631                         ret = btrfs_prev_leaf(root, path);
4632                         if (ret != 0)
4633                                 return ret;
4634                 } else {
4635                         path->slots[0]--;
4636                 }
4637                 leaf = path->nodes[0];
4638                 nritems = btrfs_header_nritems(leaf);
4639                 if (nritems == 0)
4640                         return 1;
4641                 if (path->slots[0] == nritems)
4642                         path->slots[0]--;
4643
4644                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4645                 if (found_key.objectid < min_objectid)
4646                         break;
4647                 if (found_key.type == BTRFS_EXTENT_ITEM_KEY ||
4648                     found_key.type == BTRFS_METADATA_ITEM_KEY)
4649                         return 0;
4650                 if (found_key.objectid == min_objectid &&
4651                     found_key.type < BTRFS_EXTENT_ITEM_KEY)
4652                         break;
4653         }
4654         return 1;
4655 }