extent-tree.c

   1 /*
   2  * Copyright (C) 2007 Oracle.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <stdio.h>
  20 #include <stdlib.h>
  21 #include "kerncompat.h"
  22 #include "radix-tree.h"
  23 #include "ctree.h"
  24 #include "disk-io.h"
  25 #include "print-tree.h"
  26 #include "transaction.h"
  27 #include "crc32c.h"
  28 #include "volumes.h"
  29
  30 #define BLOCK_GROUP_DATA     EXTENT_WRITEBACK
  31 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
  32 #define BLOCK_GROUP_SYSTEM   EXTENT_NEW
  33
  34 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
  35
  36 #define PENDING_EXTENT_INSERT 0
  37 #define PENDING_EXTENT_DELETE 1
  38 #define PENDING_BACKREF_UPDATE 2
  39
  40 struct pending_extent_op {
  41         int type;
  42         u64 bytenr;
  43         u64 num_bytes;
  44         u64 parent;
  45         u64 orig_parent;
  46         u64 generation;
  47         u64 orig_generation;
  48         int level;
  49 };
  50
  51 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
  52                                  btrfs_root *extent_root);
  53 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
  54                                btrfs_root *extent_root);
  55
  56 void maybe_lock_mutex(struct btrfs_root *root)
  57 {
  58 }
  59
  60 void maybe_unlock_mutex(struct btrfs_root *root)
  61 {
  62 }
  63
  64 static int cache_block_group(struct btrfs_root *root,
  65                              struct btrfs_block_group_cache *block_group)
  66 {
  67         struct btrfs_path *path;
  68         int ret;
  69         struct btrfs_key key;
  70         struct extent_buffer *leaf;
  71         struct extent_io_tree *free_space_cache;
  72         int slot;
  73         u64 last = 0;
  74         u64 hole_size;
  75         u64 first_free;
  76         int found = 0;
  77
  78         if (!block_group)
  79                 return 0;
  80
  81         root = root->fs_info->extent_root;
  82         free_space_cache = &root->fs_info->free_space_cache;
  83
  84         if (block_group->cached)
  85                 return 0;
  86
  87         path = btrfs_alloc_path();
  88         if (!path)
  89                 return -ENOMEM;
  90
  91         path->reada = 2;
  92         first_free = block_group->key.objectid;
  93         key.objectid = block_group->key.objectid;
  94         key.offset = 0;
  95         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
  96         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
  97         if (ret < 0)
  98                 return ret;
  99         ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
 100         if (ret < 0)
 101                 return ret;
 102         if (ret == 0) {
 103                 leaf = path->nodes[0];
 104                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 105                 if (key.objectid + key.offset > first_free)
 106                         first_free = key.objectid + key.offset;
 107         }
 108         while(1) {
 109                 leaf = path->nodes[0];
 110                 slot = path->slots[0];
 111                 if (slot >= btrfs_header_nritems(leaf)) {
 112                         ret = btrfs_next_leaf(root, path);
 113                         if (ret < 0)
 114                                 goto err;
 115                         if (ret == 0) {
 116                                 continue;
 117                         } else {
 118                                 break;
 119                         }
 120                 }
 121                 btrfs_item_key_to_cpu(leaf, &key, slot);
 122                 if (key.objectid < block_group->key.objectid) {
 123                         goto next;
 124                 }
 125                 if (key.objectid >= block_group->key.objectid +
 126                     block_group->key.offset) {
 127                         break;
 128                 }
 129
 130                 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
 131                         if (!found) {
 132                                 last = first_free;
 133                                 found = 1;
 134                         }
 135                         if (key.objectid > last) {
 136                                 hole_size = key.objectid - last;
 137                                 set_extent_dirty(free_space_cache, last,
 138                                                  last + hole_size - 1,
 139                                                  GFP_NOFS);
 140                         }
 141                         last = key.objectid + key.offset;
 142                 }
 143 next:
 144                 path->slots[0]++;
 145         }
 146
 147         if (!found)
 148                 last = first_free;
 149         if (block_group->key.objectid +
 150             block_group->key.offset > last) {
 151                 hole_size = block_group->key.objectid +
 152                         block_group->key.offset - last;
 153                 set_extent_dirty(free_space_cache, last,
 154                                  last + hole_size - 1, GFP_NOFS);
 155         }
 156         block_group->cached = 1;
 157 err:
 158         btrfs_free_path(path);
 159         return 0;
 160 }
 161
 162 struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct
 163                                                        btrfs_fs_info *info,
 164                                                        u64 bytenr)
 165 {
 166         struct extent_io_tree *block_group_cache;
 167         struct btrfs_block_group_cache *block_group = NULL;
 168         u64 ptr;
 169         u64 start;
 170         u64 end;
 171         int ret;
 172
 173         bytenr = max_t(u64, bytenr,
 174                        BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
 175         block_group_cache = &info->block_group_cache;
 176         ret = find_first_extent_bit(block_group_cache,
 177                                     bytenr, &start, &end,
 178                                     BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
 179                                     BLOCK_GROUP_SYSTEM);
 180         if (ret) {
 181                 return NULL;
 182         }
 183         ret = get_state_private(block_group_cache, start, &ptr);
 184         if (ret)
 185                 return NULL;
 186
 187         block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
 188         return block_group;
 189 }
 190
 191 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 192                                                          btrfs_fs_info *info,
 193                                                          u64 bytenr)
 194 {
 195         struct extent_io_tree *block_group_cache;
 196         struct btrfs_block_group_cache *block_group = NULL;
 197         u64 ptr;
 198         u64 start;
 199         u64 end;
 200         int ret;
 201
 202         block_group_cache = &info->block_group_cache;
 203         ret = find_first_extent_bit(block_group_cache,
 204                                     bytenr, &start, &end,
 205                                     BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
 206                                     BLOCK_GROUP_SYSTEM);
 207         if (ret) {
 208                 return NULL;
 209         }
 210         ret = get_state_private(block_group_cache, start, &ptr);
 211         if (ret)
 212                 return NULL;
 213
 214         block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
 215         if (block_group->key.objectid <= bytenr && bytenr <
 216             block_group->key.objectid + block_group->key.offset)
 217                 return block_group;
 218         return NULL;
 219 }
 220
 221 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
 222 {
 223         return (cache->flags & bits) == bits;
 224 }
 225
 226 static int noinline find_search_start(struct btrfs_root *root,
 227                               struct btrfs_block_group_cache **cache_ret,
 228                               u64 *start_ret, int num, int data)
 229 {
 230         int ret;
 231         struct btrfs_block_group_cache *cache = *cache_ret;
 232         u64 last;
 233         u64 start = 0;
 234         u64 end = 0;
 235         u64 search_start = *start_ret;
 236         int wrapped = 0;
 237
 238         if (!cache) {
 239                 goto out;
 240         }
 241 again:
 242         ret = cache_block_group(root, cache);
 243         if (ret)
 244                 goto out;
 245
 246         last = max(search_start, cache->key.objectid);
 247         if (cache->ro || !block_group_bits(cache, data)) {
 248                 goto new_group;
 249         }
 250
 251         while(1) {
 252                 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
 253                                             last, &start, &end, EXTENT_DIRTY);
 254                 if (ret) {
 255                         goto new_group;
 256                 }
 257
 258                 start = max(last, start);
 259                 last = end + 1;
 260                 if (last - start < num) {
 261                         continue;
 262                 }
 263                 if (start + num > cache->key.objectid + cache->key.offset) {
 264                         goto new_group;
 265                 }
 266                 *start_ret = start;
 267                 return 0;
 268         }
 269 out:
 270         cache = btrfs_lookup_block_group(root->fs_info, search_start);
 271         if (!cache) {
 272                 printk("Unable to find block group for %llu\n",
 273                         (unsigned long long)search_start);
 274                 WARN_ON(1);
 275         }
 276         return -ENOSPC;
 277
 278 new_group:
 279         last = cache->key.objectid + cache->key.offset;
 280 wrapped:
 281         cache = btrfs_lookup_first_block_group(root->fs_info, last);
 282         if (!cache) {
 283 no_cache:
 284                 if (!wrapped) {
 285                         wrapped = 1;
 286                         last = search_start;
 287                         goto wrapped;
 288                 }
 289                 goto out;
 290         }
 291         cache = btrfs_find_block_group(root, cache, last, data, 0);
 292         cache = btrfs_find_block_group(root, cache, last, data, 0);
 293         if (!cache)
 294                 goto no_cache;
 295
 296         *cache_ret = cache;
 297         goto again;
 298 }
 299
 300 static u64 div_factor(u64 num, int factor)
 301 {
 302         if (factor == 10)
 303                 return num;
 304         num *= factor;
 305         num /= 10;
 306         return num;
 307 }
 308
 309 static int block_group_state_bits(u64 flags)
 310 {
 311         int bits = 0;
 312         if (flags & BTRFS_BLOCK_GROUP_DATA)
 313                 bits |= BLOCK_GROUP_DATA;
 314         if (flags & BTRFS_BLOCK_GROUP_METADATA)
 315                 bits |= BLOCK_GROUP_METADATA;
 316         if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
 317                 bits |= BLOCK_GROUP_SYSTEM;
 318         return bits;
 319 }
 320
 321 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
 322                                                  struct btrfs_block_group_cache
 323                                                  *hint, u64 search_start,
 324                                                  int data, int owner)
 325 {
 326         struct btrfs_block_group_cache *cache;
 327         struct extent_io_tree *block_group_cache;
 328         struct btrfs_block_group_cache *found_group = NULL;
 329         struct btrfs_fs_info *info = root->fs_info;
 330         u64 used;
 331         u64 last = 0;
 332         u64 hint_last;
 333         u64 start;
 334         u64 end;
 335         u64 free_check;
 336         u64 ptr;
 337         int bit;
 338         int ret;
 339         int full_search = 0;
 340         int factor = 10;
 341
 342         block_group_cache = &info->block_group_cache;
 343
 344         if (!owner)
 345                 factor = 10;
 346
 347         bit = block_group_state_bits(data);
 348
 349         if (search_start) {
 350                 struct btrfs_block_group_cache *shint;
 351                 shint = btrfs_lookup_block_group(info, search_start);
 352                 if (shint && !shint->ro && block_group_bits(shint, data)) {
 353                         used = btrfs_block_group_used(&shint->item);
 354                         if (used + shint->pinned <
 355                             div_factor(shint->key.offset, factor)) {
 356                                 return shint;
 357                         }
 358                 }
 359         }
 360         if (hint && !hint->ro && block_group_bits(hint, data)) {
 361                 used = btrfs_block_group_used(&hint->item);
 362                 if (used + hint->pinned <
 363                     div_factor(hint->key.offset, factor)) {
 364                         return hint;
 365                 }
 366                 last = hint->key.objectid + hint->key.offset;
 367                 hint_last = last;
 368         } else {
 369                 if (hint)
 370                         hint_last = max(hint->key.objectid, search_start);
 371                 else
 372                         hint_last = search_start;
 373
 374                 last = hint_last;
 375         }
 376 again:
 377         while(1) {
 378                 ret = find_first_extent_bit(block_group_cache, last,
 379                                             &start, &end, bit);
 380                 if (ret)
 381                         break;
 382
 383                 ret = get_state_private(block_group_cache, start, &ptr);
 384                 if (ret)
 385                         break;
 386
 387                 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
 388                 last = cache->key.objectid + cache->key.offset;
 389                 used = btrfs_block_group_used(&cache->item);
 390
 391                 if (!cache->ro && block_group_bits(cache, data)) {
 392                         if (full_search)
 393                                 free_check = cache->key.offset;
 394                         else
 395                                 free_check = div_factor(cache->key.offset,
 396                                                         factor);
 397
 398                         if (used + cache->pinned < free_check) {
 399                                 found_group = cache;
 400                                 goto found;
 401                         }
 402                 }
 403                 cond_resched();
 404         }
 405         if (!full_search) {
 406                 last = search_start;
 407                 full_search = 1;
 408                 goto again;
 409         }
 410 found:
 411         return found_group;
 412 }
 413
 414 /*
 415  * Back reference rules.  Back refs have three main goals:
 416  *
 417  * 1) differentiate between all holders of references to an extent so that
 418  *    when a reference is dropped we can make sure it was a valid reference
 419  *    before freeing the extent.
 420  *
 421  * 2) Provide enough information to quickly find the holders of an extent
 422  *    if we notice a given block is corrupted or bad.
 423  *
 424  * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 425  *    maintenance.  This is actually the same as #2, but with a slightly
 426  *    different use case.
 427  *
 428  * File extents can be referenced by:
 429  *
 430  * - multiple snapshots, subvolumes, or different generations in one subvol
 431  * - different files inside a single subvolume
 432  * - different offsets inside a file (bookend extents in file.c)
 433  *
 434  * The extent ref structure has fields for:
 435  *
 436  * - Objectid of the subvolume root
 437  * - Generation number of the tree holding the reference
 438  * - objectid of the file holding the reference
 439  * - offset in the file corresponding to the key holding the reference
 440  * - number of references holding by parent node (alway 1 for tree blocks)
 441  *
 442  * Btree leaf may hold multiple references to a file extent. In most cases,
 443  * these references are from same file and the corresponding offsets inside
 444  * the file are close together. So inode objectid and offset in file are
 445  * just hints, they provide hints about where in the btree the references
 446  * can be found and when we can stop searching.
 447  *
 448  * When a file extent is allocated the fields are filled in:
 449  *     (root_key.objectid, trans->transid, inode objectid, offset in file, 1)
 450  *
 451  * When a leaf is cow'd new references are added for every file extent found
 452  * in the leaf.  It looks similar to the create case, but trans->transid will
 453  * be different when the block is cow'd.
 454  *
 455  *     (root_key.objectid, trans->transid, inode objectid, offset in file,
 456  *      number of references in the leaf)
 457  *
 458  * Because inode objectid and offset in file are just hints, they are not
 459  * used when backrefs are deleted. When a file extent is removed either
 460  * during snapshot deletion or file truncation, we find the corresponding
 461  * back back reference and check the following fields.
 462  *
 463  *     (btrfs_header_owner(leaf), btrfs_header_generation(leaf))
 464  *
 465  * Btree extents can be referenced by:
 466  *
 467  * - Different subvolumes
 468  * - Different generations of the same subvolume
 469  *
 470  * When a tree block is created, back references are inserted:
 471  *
 472  * (root->root_key.objectid, trans->transid, level, 0, 1)
 473  *
 474  * When a tree block is cow'd, new back references are added for all the
 475  * blocks it points to. If the tree block isn't in reference counted root,
 476  * the old back references are removed. These new back references are of
 477  * the form (trans->transid will have increased since creation):
 478  *
 479  * (root->root_key.objectid, trans->transid, level, 0, 1)
 480  *
 481  * When a backref is in deleting, the following fields are checked:
 482  *
 483  * if backref was for a tree root:
 484  *     (btrfs_header_owner(itself), btrfs_header_generation(itself))
 485  * else
 486  *     (btrfs_header_owner(parent), btrfs_header_generation(parent))
 487  *
 488  * Back Reference Key composing:
 489  *
 490  * The key objectid corresponds to the first byte in the extent, the key
 491  * type is set to BTRFS_EXTENT_REF_KEY, and the key offset is the first
 492  * byte of parent extent. If a extent is tree root, the key offset is set
 493  * to the key objectid.
 494  */
 495
 496 static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
 497                                           struct btrfs_root *root,
 498                                           struct btrfs_path *path,
 499                                           u64 bytenr, u64 parent,
 500                                           u64 ref_root, u64 ref_generation,
 501                                           u64 owner_objectid, int del)
 502 {
 503         struct btrfs_key key;
 504         struct btrfs_extent_ref *ref;
 505         struct extent_buffer *leaf;
 506         u64 ref_objectid;
 507         int ret;
 508
 509         key.objectid = bytenr;
 510         key.type = BTRFS_EXTENT_REF_KEY;
 511         key.offset = parent;
 512
 513         ret = btrfs_search_slot(trans, root, &key, path, del ? -1 : 0, 1);
 514         if (ret < 0)
 515                 goto out;
 516         if (ret > 0) {
 517                 ret = -ENOENT;
 518                 goto out;
 519         }
 520
 521         leaf = path->nodes[0];
 522         ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
 523         ref_objectid = btrfs_ref_objectid(leaf, ref);
 524         if (btrfs_ref_root(leaf, ref) != ref_root ||
 525             btrfs_ref_generation(leaf, ref) != ref_generation ||
 526             (ref_objectid != owner_objectid &&
 527              ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) {
 528                 ret = -EIO;
 529                 WARN_ON(1);
 530                 goto out;
 531         }
 532         ret = 0;
 533 out:
 534         return ret;
 535 }
 536
 537 static int noinline insert_extent_backref(struct btrfs_trans_handle *trans,
 538                                           struct btrfs_root *root,
 539                                           struct btrfs_path *path,
 540                                           u64 bytenr, u64 parent,
 541                                           u64 ref_root, u64 ref_generation,
 542                                           u64 owner_objectid)
 543 {
 544         struct btrfs_key key;
 545         struct extent_buffer *leaf;
 546         struct btrfs_extent_ref *ref;
 547         u32 num_refs;
 548         int ret;
 549
 550         key.objectid = bytenr;
 551         key.type = BTRFS_EXTENT_REF_KEY;
 552         key.offset = parent;
 553
 554         ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*ref));
 555         if (ret == 0) {
 556                 leaf = path->nodes[0];
 557                 ref = btrfs_item_ptr(leaf, path->slots[0],
 558                                      struct btrfs_extent_ref);
 559                 btrfs_set_ref_root(leaf, ref, ref_root);
 560                 btrfs_set_ref_generation(leaf, ref, ref_generation);
 561                 btrfs_set_ref_objectid(leaf, ref, owner_objectid);
 562                 btrfs_set_ref_num_refs(leaf, ref, 1);
 563         } else if (ret == -EEXIST) {
 564                 u64 existing_owner;
 565                 BUG_ON(owner_objectid < BTRFS_FIRST_FREE_OBJECTID);
 566                 leaf = path->nodes[0];
 567                 ref = btrfs_item_ptr(leaf, path->slots[0],
 568                                      struct btrfs_extent_ref);
 569                 if (btrfs_ref_root(leaf, ref) != ref_root ||
 570                     btrfs_ref_generation(leaf, ref) != ref_generation) {
 571                         ret = -EIO;
 572                         WARN_ON(1);
 573                         goto out;
 574                 }
 575
 576                 num_refs = btrfs_ref_num_refs(leaf, ref);
 577                 BUG_ON(num_refs == 0);
 578                 btrfs_set_ref_num_refs(leaf, ref, num_refs + 1);
 579
 580                 existing_owner = btrfs_ref_objectid(leaf, ref);
 581                 if (existing_owner != owner_objectid &&
 582                     existing_owner != BTRFS_MULTIPLE_OBJECTIDS) {
 583                         btrfs_set_ref_objectid(leaf, ref,
 584                                         BTRFS_MULTIPLE_OBJECTIDS);
 585                 }
 586                 ret = 0;
 587         } else {
 588                 goto out;
 589         }
 590         btrfs_mark_buffer_dirty(path->nodes[0]);
 591 out:
 592         btrfs_release_path(root, path);
 593         return ret;
 594 }
 595
 596 static int noinline remove_extent_backref(struct btrfs_trans_handle *trans,
 597                                           struct btrfs_root *root,
 598                                           struct btrfs_path *path)
 599 {
 600         struct extent_buffer *leaf;
 601         struct btrfs_extent_ref *ref;
 602         u32 num_refs;
 603         int ret = 0;
 604
 605         leaf = path->nodes[0];
 606         ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
 607         num_refs = btrfs_ref_num_refs(leaf, ref);
 608         BUG_ON(num_refs == 0);
 609         num_refs -= 1;
 610         if (num_refs == 0) {
 611                 ret = btrfs_del_item(trans, root, path);
 612         } else {
 613                 btrfs_set_ref_num_refs(leaf, ref, num_refs);
 614                 btrfs_mark_buffer_dirty(leaf);
 615         }
 616         btrfs_release_path(root, path);
 617         return ret;
 618 }
 619
 620 static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
 621                                      struct btrfs_root *root, u64 bytenr,
 622                                      u64 orig_parent, u64 parent,
 623                                      u64 orig_root, u64 ref_root,
 624                                      u64 orig_generation, u64 ref_generation,
 625                                      u64 owner_objectid)
 626 {
 627         int ret;
 628         struct btrfs_root *extent_root = root->fs_info->extent_root;
 629         struct btrfs_path *path;
 630
 631         if (root == root->fs_info->extent_root) {
 632                 struct pending_extent_op *extent_op;
 633                 u64 num_bytes;
 634
 635                 BUG_ON(owner_objectid >= BTRFS_MAX_LEVEL);
 636                 num_bytes = btrfs_level_size(root, (int)owner_objectid);
 637                 if (test_range_bit(&root->fs_info->extent_ins, bytenr,
 638                                 bytenr + num_bytes - 1, EXTENT_LOCKED, 0)) {
 639                         u64 priv;
 640                         ret = get_state_private(&root->fs_info->extent_ins,
 641                                                 bytenr, &priv);
 642                         BUG_ON(ret);
 643                         extent_op = (struct pending_extent_op *)
 644                                                         (unsigned long)priv;
 645                         BUG_ON(extent_op->parent != orig_parent);
 646                         BUG_ON(extent_op->generation != orig_generation);
 647                         extent_op->parent = parent;
 648                         extent_op->generation = ref_generation;
 649                 } else {
 650                         extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
 651                         BUG_ON(!extent_op);
 652
 653                         extent_op->type = PENDING_BACKREF_UPDATE;
 654                         extent_op->bytenr = bytenr;
 655                         extent_op->num_bytes = num_bytes;
 656                         extent_op->parent = parent;
 657                         extent_op->orig_parent = orig_parent;
 658                         extent_op->generation = ref_generation;
 659                         extent_op->orig_generation = orig_generation;
 660                         extent_op->level = (int)owner_objectid;
 661
 662                         set_extent_bits(&root->fs_info->extent_ins,
 663                                         bytenr, bytenr + num_bytes - 1,
 664                                         EXTENT_LOCKED, GFP_NOFS);
 665                         set_state_private(&root->fs_info->extent_ins,
 666                                           bytenr, (unsigned long)extent_op);
 667                 }
 668                 return 0;
 669         }
 670
 671         path = btrfs_alloc_path();
 672         if (!path)
 673                 return -ENOMEM;
 674         ret = lookup_extent_backref(trans, extent_root, path,
 675                                     bytenr, orig_parent, orig_root,
 676                                     orig_generation, owner_objectid, 1);
 677         if (ret)
 678                 goto out;
 679         ret = remove_extent_backref(trans, extent_root, path);
 680         if (ret)
 681                 goto out;
 682         ret = insert_extent_backref(trans, extent_root, path, bytenr,
 683                                     parent, ref_root, ref_generation,
 684                                     owner_objectid);
 685         BUG_ON(ret);
 686         finish_current_insert(trans, extent_root);
 687         del_pending_extents(trans, extent_root);
 688 out:
 689         btrfs_free_path(path);
 690         return ret;
 691 }
 692
 693 int btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
 694                             struct btrfs_root *root, u64 bytenr,
 695                             u64 orig_parent, u64 parent,
 696                             u64 ref_root, u64 ref_generation,
 697                             u64 owner_objectid)
 698 {
 699         int ret;
 700         if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
 701             owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
 702                 return 0;
 703         maybe_lock_mutex(root);
 704         ret = __btrfs_update_extent_ref(trans, root, bytenr, orig_parent,
 705                                         parent, ref_root, ref_root,
 706                                         ref_generation, ref_generation,
 707                                         owner_objectid);
 708         maybe_unlock_mutex(root);
 709         return ret;
 710 }
 711
 712 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 713                                   struct btrfs_root *root, u64 bytenr,
 714                                   u64 orig_parent, u64 parent,
 715                                   u64 orig_root, u64 ref_root,
 716                                   u64 orig_generation, u64 ref_generation,
 717                                   u64 owner_objectid)
 718 {
 719         struct btrfs_path *path;
 720         int ret;
 721         struct btrfs_key key;
 722         struct extent_buffer *l;
 723         struct btrfs_extent_item *item;
 724         u32 refs;
 725
 726         path = btrfs_alloc_path();
 727         if (!path)
 728                 return -ENOMEM;
 729
 730         path->reada = 1;
 731         key.objectid = bytenr;
 732         key.type = BTRFS_EXTENT_ITEM_KEY;
 733         key.offset = (u64)-1;
 734
 735         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 736                                 0, 1);
 737         if (ret < 0)
 738                 return ret;
 739         BUG_ON(ret == 0 || path->slots[0] == 0);
 740
 741         path->slots[0]--;
 742         l = path->nodes[0];
 743
 744         btrfs_item_key_to_cpu(l, &key, path->slots[0]);
 745         BUG_ON(key.objectid != bytenr);
 746         BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY);
 747
 748         item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
 749         refs = btrfs_extent_refs(l, item);
 750         btrfs_set_extent_refs(l, item, refs + 1);
 751         btrfs_mark_buffer_dirty(path->nodes[0]);
 752
 753         btrfs_release_path(root->fs_info->extent_root, path);
 754
 755         path->reada = 1;
 756         ret = insert_extent_backref(trans, root->fs_info->extent_root,
 757                                     path, bytenr, parent,
 758                                     ref_root, ref_generation,
 759                                     owner_objectid);
 760         BUG_ON(ret);
 761         finish_current_insert(trans, root->fs_info->extent_root);
 762         del_pending_extents(trans, root->fs_info->extent_root);
 763
 764         btrfs_free_path(path);
 765         return 0;
 766 }
 767
 768 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 769                          struct btrfs_root *root,
 770                          u64 bytenr, u64 num_bytes, u64 parent,
 771                          u64 ref_root, u64 ref_generation,
 772                          u64 owner_objectid)
 773 {
 774         int ret;
 775         if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
 776             owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
 777                 return 0;
 778         maybe_lock_mutex(root);
 779         ret = __btrfs_inc_extent_ref(trans, root, bytenr, 0, parent,
 780                                      0, ref_root, 0, ref_generation,
 781                                      owner_objectid);
 782         maybe_unlock_mutex(root);
 783         return ret;
 784 }
 785
 786 int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
 787                          struct btrfs_root *root)
 788 {
 789         finish_current_insert(trans, root->fs_info->extent_root);
 790         del_pending_extents(trans, root->fs_info->extent_root);
 791         return 0;
 792 }
 793
 794 int lookup_extent_ref(struct btrfs_trans_handle *trans,
 795                       struct btrfs_root *root, u64 bytenr,
 796                       u64 num_bytes, u32 *refs)
 797 {
 798         struct btrfs_path *path;
 799         int ret;
 800         struct btrfs_key key;
 801         struct extent_buffer *l;
 802         struct btrfs_extent_item *item;
 803
 804         WARN_ON(num_bytes < root->sectorsize);
 805         path = btrfs_alloc_path();
 806         path->reada = 1;
 807         key.objectid = bytenr;
 808         key.offset = num_bytes;
 809         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 810         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
 811                                 0, 0);
 812         if (ret < 0)
 813                 goto out;
 814         if (ret != 0) {
 815                 btrfs_print_leaf(root, path->nodes[0]);
 816                 printk("failed to find block number %Lu\n", bytenr);
 817                 BUG();
 818         }
 819         l = path->nodes[0];
 820         item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
 821         *refs = btrfs_extent_refs(l, item);
 822 out:
 823         btrfs_free_path(path);
 824         return 0;
 825 }
 826
 827 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 828                   struct extent_buffer *orig_buf, struct extent_buffer *buf,
 829                   u32 *nr_extents)
 830 {
 831         u64 bytenr;
 832         u64 ref_root;
 833         u64 orig_root;
 834         u64 ref_generation;
 835         u64 orig_generation;
 836         u32 nritems;
 837         u32 nr_file_extents = 0;
 838         struct btrfs_key key;
 839         struct btrfs_file_extent_item *fi;
 840         int i;
 841         int level;
 842         int ret = 0;
 843         int faili = 0;
 844         int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
 845                             u64, u64, u64, u64, u64, u64, u64, u64);
 846
 847         ref_root = btrfs_header_owner(buf);
 848         ref_generation = btrfs_header_generation(buf);
 849         orig_root = btrfs_header_owner(orig_buf);
 850         orig_generation = btrfs_header_generation(orig_buf);
 851
 852         nritems = btrfs_header_nritems(buf);
 853         level = btrfs_header_level(buf);
 854
 855         if (root->ref_cows) {
 856                 process_func = __btrfs_inc_extent_ref;
 857         } else {
 858                 if (level == 0 &&
 859                     root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
 860                         goto out;
 861                 process_func = __btrfs_update_extent_ref;
 862         }
 863
 864         for (i = 0; i < nritems; i++) {
 865                 cond_resched();
 866                 if (level == 0) {
 867                         btrfs_item_key_to_cpu(buf, &key, i);
 868                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 869                                 continue;
 870                         fi = btrfs_item_ptr(buf, i,
 871                                             struct btrfs_file_extent_item);
 872                         if (btrfs_file_extent_type(buf, fi) ==
 873                             BTRFS_FILE_EXTENT_INLINE)
 874                                 continue;
 875                         bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 876                         if (bytenr == 0)
 877                                 continue;
 878
 879                         nr_file_extents++;
 880
 881                         maybe_lock_mutex(root);
 882                         ret = process_func(trans, root, bytenr,
 883                                            orig_buf->start, buf->start,
 884                                            orig_root, ref_root,
 885                                            orig_generation, ref_generation,
 886                                            key.objectid);
 887                         maybe_unlock_mutex(root);
 888
 889                         if (ret) {
 890                                 faili = i;
 891                                 WARN_ON(1);
 892                                 goto fail;
 893                         }
 894                 } else {
 895                         bytenr = btrfs_node_blockptr(buf, i);
 896                         maybe_lock_mutex(root);
 897                         ret = process_func(trans, root, bytenr,
 898                                            orig_buf->start, buf->start,
 899                                            orig_root, ref_root,
 900                                            orig_generation, ref_generation,
 901                                            level - 1);
 902                         maybe_unlock_mutex(root);
 903                         if (ret) {
 904                                 faili = i;
 905                                 WARN_ON(1);
 906                                 goto fail;
 907                         }
 908                 }
 909         }
 910 out:
 911         if (nr_extents) {
 912                 if (level == 0)
 913                         *nr_extents = nr_file_extents;
 914                 else
 915                         *nr_extents = nritems;
 916         }
 917         return 0;
 918 fail:
 919         WARN_ON(1);
 920 #if 0
 921         for (i =0; i < faili; i++) {
 922                 if (level == 0) {
 923                         u64 disk_bytenr;
 924                         btrfs_item_key_to_cpu(buf, &key, i);
 925                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 926                                 continue;
 927                         fi = btrfs_item_ptr(buf, i,
 928                                             struct btrfs_file_extent_item);
 929                         if (btrfs_file_extent_type(buf, fi) ==
 930                             BTRFS_FILE_EXTENT_INLINE)
 931                                 continue;
 932                         disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 933                         if (disk_bytenr == 0)
 934                                 continue;
 935                         err = btrfs_free_extent(trans, root, disk_bytenr,
 936                                     btrfs_file_extent_disk_num_bytes(buf,
 937                                                                       fi), 0);
 938                         BUG_ON(err);
 939                 } else {
 940                         bytenr = btrfs_node_blockptr(buf, i);
 941                         err = btrfs_free_extent(trans, root, bytenr,
 942                                         btrfs_level_size(root, level - 1), 0);
 943                         BUG_ON(err);
 944                 }
 945         }
 946 #endif
 947         return ret;
 948 }
 949
 950 int btrfs_update_ref(struct btrfs_trans_handle *trans,
 951                      struct btrfs_root *root, struct extent_buffer *orig_buf,
 952                      struct extent_buffer *buf, int start_slot, int nr)
 953
 954 {
 955         u64 bytenr;
 956         u64 ref_root;
 957         u64 orig_root;
 958         u64 ref_generation;
 959         u64 orig_generation;
 960         struct btrfs_key key;
 961         struct btrfs_file_extent_item *fi;
 962         int i;
 963         int ret;
 964         int slot;
 965         int level;
 966
 967         BUG_ON(start_slot < 0);
 968         BUG_ON(start_slot + nr > btrfs_header_nritems(buf));
 969
 970         ref_root = btrfs_header_owner(buf);
 971         ref_generation = btrfs_header_generation(buf);
 972         orig_root = btrfs_header_owner(orig_buf);
 973         orig_generation = btrfs_header_generation(orig_buf);
 974         level = btrfs_header_level(buf);
 975
 976         if (!root->ref_cows) {
 977                 if (level == 0 &&
 978                     root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
 979                         return 0;
 980         }
 981
 982         for (i = 0, slot = start_slot; i < nr; i++, slot++) {
 983                 cond_resched();
 984                 if (level == 0) {
 985                         btrfs_item_key_to_cpu(buf, &key, slot);
 986                         if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
 987                                 continue;
 988                         fi = btrfs_item_ptr(buf, slot,
 989                                             struct btrfs_file_extent_item);
 990                         if (btrfs_file_extent_type(buf, fi) ==
 991                             BTRFS_FILE_EXTENT_INLINE)
 992                                 continue;
 993                         bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
 994                         if (bytenr == 0)
 995                                 continue;
 996
 997                         maybe_lock_mutex(root);
 998                         ret = __btrfs_update_extent_ref(trans, root, bytenr,
 999                                             orig_buf->start, buf->start,
1000                                             orig_root, ref_root,
1001                                             orig_generation, ref_generation,
1002                                             key.objectid);
1003                         maybe_unlock_mutex(root);
1004                         if (ret)
1005                                 goto fail;
1006                 } else {
1007                         bytenr = btrfs_node_blockptr(buf, slot);
1008                         maybe_lock_mutex(root);
1009                         ret = __btrfs_update_extent_ref(trans, root, bytenr,
1010                                             orig_buf->start, buf->start,
1011                                             orig_root, ref_root,
1012                                             orig_generation, ref_generation,
1013                                             level - 1);
1014                         maybe_unlock_mutex(root);
1015                         if (ret)
1016                                 goto fail;
1017                 }
1018         }
1019         return 0;
1020 fail:
1021         WARN_ON(1);
1022         return -1;
1023 }
1024
1025 static int write_one_cache_group(struct btrfs_trans_handle *trans,
1026                                  struct btrfs_root *root,
1027                                  struct btrfs_path *path,
1028                                  struct btrfs_block_group_cache *cache)
1029 {
1030         int ret;
1031         int pending_ret;
1032         struct btrfs_root *extent_root = root->fs_info->extent_root;
1033         unsigned long bi;
1034         struct extent_buffer *leaf;
1035
1036         ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
1037         if (ret < 0)
1038                 goto fail;
1039         BUG_ON(ret);
1040
1041         leaf = path->nodes[0];
1042         bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
1043         write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
1044         btrfs_mark_buffer_dirty(leaf);
1045         btrfs_release_path(extent_root, path);
1046 fail:
1047         finish_current_insert(trans, extent_root);
1048         pending_ret = del_pending_extents(trans, extent_root);
1049         if (ret)
1050                 return ret;
1051         if (pending_ret)
1052                 return pending_ret;
1053         return 0;
1054
1055 }
1056
1057 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
1058                                    struct btrfs_root *root)
1059 {
1060         struct extent_io_tree *block_group_cache;
1061         struct btrfs_block_group_cache *cache;
1062         int ret;
1063         int err = 0;
1064         int werr = 0;
1065         struct btrfs_path *path;
1066         u64 last = 0;
1067         u64 start;
1068         u64 end;
1069         u64 ptr;
1070
1071         block_group_cache = &root->fs_info->block_group_cache;
1072         path = btrfs_alloc_path();
1073         if (!path)
1074                 return -ENOMEM;
1075
1076         while(1) {
1077                 ret = find_first_extent_bit(block_group_cache, last,
1078                                             &start, &end, BLOCK_GROUP_DIRTY);
1079                 if (ret)
1080                         break;
1081
1082                 last = end + 1;
1083                 ret = get_state_private(block_group_cache, start, &ptr);
1084                 if (ret)
1085                         break;
1086                 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
1087                 err = write_one_cache_group(trans, root,
1088                                             path, cache);
1089                 /*
1090                  * if we fail to write the cache group, we want
1091                  * to keep it marked dirty in hopes that a later
1092                  * write will work
1093                  */
1094                 if (err) {
1095                         werr = err;
1096                         continue;
1097                 }
1098                 clear_extent_bits(block_group_cache, start, end,
1099                                   BLOCK_GROUP_DIRTY, GFP_NOFS);
1100         }
1101         btrfs_free_path(path);
1102         return werr;
1103 }
1104
1105 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
1106                                                   u64 flags)
1107 {
1108         struct list_head *head = &info->space_info;
1109         struct list_head *cur;
1110         struct btrfs_space_info *found;
1111         list_for_each(cur, head) {
1112                 found = list_entry(cur, struct btrfs_space_info, list);
1113                 if (found->flags == flags)
1114                         return found;
1115         }
1116         return NULL;
1117
1118 }
1119
1120 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
1121                              u64 total_bytes, u64 bytes_used,
1122                              struct btrfs_space_info **space_info)
1123 {
1124         struct btrfs_space_info *found;
1125
1126         found = __find_space_info(info, flags);
1127         if (found) {
1128                 found->total_bytes += total_bytes;
1129                 found->bytes_used += bytes_used;
1130                 WARN_ON(found->total_bytes < found->bytes_used);
1131                 *space_info = found;
1132                 return 0;
1133         }
1134         found = kmalloc(sizeof(*found), GFP_NOFS);
1135         if (!found)
1136                 return -ENOMEM;
1137
1138         list_add(&found->list, &info->space_info);
1139         found->flags = flags;
1140         found->total_bytes = total_bytes;
1141         found->bytes_used = bytes_used;
1142         found->bytes_pinned = 0;
1143         found->full = 0;
1144         *space_info = found;
1145         return 0;
1146 }
1147
1148
1149 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
1150 {
1151         u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
1152                                    BTRFS_BLOCK_GROUP_RAID1 |
1153                                    BTRFS_BLOCK_GROUP_DUP);
1154         if (extra_flags) {
1155                 if (flags & BTRFS_BLOCK_GROUP_DATA)
1156                         fs_info->avail_data_alloc_bits |= extra_flags;
1157                 if (flags & BTRFS_BLOCK_GROUP_METADATA)
1158                         fs_info->avail_metadata_alloc_bits |= extra_flags;
1159                 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
1160                         fs_info->avail_system_alloc_bits |= extra_flags;
1161         }
1162 }
1163
1164 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
1165                           struct btrfs_root *extent_root, u64 alloc_bytes,
1166                           u64 flags)
1167 {
1168         struct btrfs_space_info *space_info;
1169         u64 thresh;
1170         u64 start;
1171         u64 num_bytes;
1172         int ret;
1173
1174         space_info = __find_space_info(extent_root->fs_info, flags);
1175         if (!space_info) {
1176                 ret = update_space_info(extent_root->fs_info, flags,
1177                                         0, 0, &space_info);
1178                 BUG_ON(ret);
1179         }
1180         BUG_ON(!space_info);
1181
1182         if (space_info->full)
1183                 return 0;
1184
1185         thresh = div_factor(space_info->total_bytes, 7);
1186         if ((space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) <
1187             thresh)
1188                 return 0;
1189
1190         ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags);
1191         if (ret == -ENOSPC) {
1192                 space_info->full = 1;
1193                 return 0;
1194         }
1195
1196         BUG_ON(ret);
1197
1198         ret = btrfs_make_block_group(trans, extent_root, 0, flags,
1199                      BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
1200         BUG_ON(ret);
1201         return 0;
1202 }
1203
1204 static int update_block_group(struct btrfs_trans_handle *trans,
1205                               struct btrfs_root *root,
1206                               u64 bytenr, u64 num_bytes, int alloc,
1207                               int mark_free)
1208 {
1209         struct btrfs_block_group_cache *cache;
1210         struct btrfs_fs_info *info = root->fs_info;
1211         u64 total = num_bytes;
1212         u64 old_val;
1213         u64 byte_in_group;
1214         u64 start;
1215         u64 end;
1216
1217         while(total) {
1218                 cache = btrfs_lookup_block_group(info, bytenr);
1219                 if (!cache) {
1220                         return -1;
1221                 }
1222                 byte_in_group = bytenr - cache->key.objectid;
1223                 WARN_ON(byte_in_group > cache->key.offset);
1224                 start = cache->key.objectid;
1225                 end = start + cache->key.offset - 1;
1226                 set_extent_bits(&info->block_group_cache, start, end,
1227                                 BLOCK_GROUP_DIRTY, GFP_NOFS);
1228
1229                 old_val = btrfs_block_group_used(&cache->item);
1230                 num_bytes = min(total, cache->key.offset - byte_in_group);
1231                 if (alloc) {
1232                         old_val += num_bytes;
1233                         cache->space_info->bytes_used += num_bytes;
1234                 } else {
1235                         old_val -= num_bytes;
1236                         cache->space_info->bytes_used -= num_bytes;
1237                         if (mark_free) {
1238                                 set_extent_dirty(&info->free_space_cache,
1239                                                  bytenr, bytenr + num_bytes - 1,
1240                                                  GFP_NOFS);
1241                         }
1242                 }
1243                 btrfs_set_block_group_used(&cache->item, old_val);
1244                 total -= num_bytes;
1245                 bytenr += num_bytes;
1246         }
1247         return 0;
1248 }
1249
1250 static int update_pinned_extents(struct btrfs_root *root,
1251                                 u64 bytenr, u64 num, int pin)
1252 {
1253         u64 len;
1254         struct btrfs_block_group_cache *cache;
1255         struct btrfs_fs_info *fs_info = root->fs_info;
1256
1257         if (pin) {
1258                 set_extent_dirty(&fs_info->pinned_extents,
1259                                 bytenr, bytenr + num - 1, GFP_NOFS);
1260         } else {
1261                 clear_extent_dirty(&fs_info->pinned_extents,
1262                                 bytenr, bytenr + num - 1, GFP_NOFS);
1263         }
1264         while (num > 0) {
1265                 cache = btrfs_lookup_block_group(fs_info, bytenr);
1266                 WARN_ON(!cache);
1267                 len = min(num, cache->key.offset -
1268                           (bytenr - cache->key.objectid));
1269                 if (pin) {
1270                         cache->pinned += len;
1271                         cache->space_info->bytes_pinned += len;
1272                         fs_info->total_pinned += len;
1273                 } else {
1274                         cache->pinned -= len;
1275                         cache->space_info->bytes_pinned -= len;
1276                         fs_info->total_pinned -= len;
1277                 }
1278                 bytenr += len;
1279                 num -= len;
1280         }
1281         return 0;
1282 }
1283
1284 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
1285 {
1286         u64 last = 0;
1287         u64 start;
1288         u64 end;
1289         struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
1290         int ret;
1291
1292         while(1) {
1293                 ret = find_first_extent_bit(pinned_extents, last,
1294                                             &start, &end, EXTENT_DIRTY);
1295                 if (ret)
1296                         break;
1297                 set_extent_dirty(copy, start, end, GFP_NOFS);
1298                 last = end + 1;
1299         }
1300         return 0;
1301 }
1302
1303 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
1304                                struct btrfs_root *root,
1305                                struct extent_io_tree *unpin)
1306 {
1307         u64 start;
1308         u64 end;
1309         int ret;
1310         struct extent_io_tree *free_space_cache;
1311         free_space_cache = &root->fs_info->free_space_cache;
1312
1313         while(1) {
1314                 ret = find_first_extent_bit(unpin, 0, &start, &end,
1315                                             EXTENT_DIRTY);
1316                 if (ret)
1317                         break;
1318                 update_pinned_extents(root, start, end + 1 - start, 0);
1319                 clear_extent_dirty(unpin, start, end, GFP_NOFS);
1320                 set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
1321         }
1322         return 0;
1323 }
1324
1325 static int finish_current_insert(struct btrfs_trans_handle *trans,
1326                                  struct btrfs_root *extent_root)
1327 {
1328         u64 start;
1329         u64 end;
1330         u64 priv;
1331         struct btrfs_fs_info *info = extent_root->fs_info;
1332         struct btrfs_path *path;
1333         struct btrfs_extent_ref *ref;
1334         struct pending_extent_op *extent_op;
1335         struct btrfs_key key;
1336         struct btrfs_extent_item extent_item;
1337         int ret;
1338         int err = 0;
1339
1340         btrfs_set_stack_extent_refs(&extent_item, 1);
1341         path = btrfs_alloc_path();
1342
1343         while(1) {
1344                 ret = find_first_extent_bit(&info->extent_ins, 0, &start,
1345                                             &end, EXTENT_LOCKED);
1346                 if (ret)
1347                         break;
1348
1349                 ret = get_state_private(&info->extent_ins, start, &priv);
1350                 BUG_ON(ret);
1351                 extent_op = (struct pending_extent_op *)(unsigned long)priv;
1352
1353                 if (extent_op->type == PENDING_EXTENT_INSERT) {
1354                         key.objectid = start;
1355                         key.offset = end + 1 - start;
1356                         key.type = BTRFS_EXTENT_ITEM_KEY;
1357                         err = btrfs_insert_item(trans, extent_root, &key,
1358                                         &extent_item, sizeof(extent_item));
1359                         BUG_ON(err);
1360
1361                         clear_extent_bits(&info->extent_ins, start, end,
1362                                           EXTENT_LOCKED, GFP_NOFS);
1363
1364                         err = insert_extent_backref(trans, extent_root, path,
1365                                                 start, extent_op->parent,
1366                                                 extent_root->root_key.objectid,
1367                                                 extent_op->generation,
1368                                                 extent_op->level);
1369                         BUG_ON(err);
1370                 } else if (extent_op->type == PENDING_BACKREF_UPDATE) {
1371                         err = lookup_extent_backref(trans, extent_root, path,
1372                                                 start, extent_op->orig_parent,
1373                                                 extent_root->root_key.objectid,
1374                                                 extent_op->orig_generation,
1375                                                 extent_op->level, 0);
1376                         BUG_ON(err);
1377
1378                         clear_extent_bits(&info->extent_ins, start, end,
1379                                           EXTENT_LOCKED, GFP_NOFS);
1380
1381                         key.objectid = start;
1382                         key.offset = extent_op->parent;
1383                         key.type = BTRFS_EXTENT_REF_KEY;
1384                         err = btrfs_set_item_key_safe(trans, extent_root, path,
1385                                                       &key);
1386                         BUG_ON(err);
1387                         ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1388                                              struct btrfs_extent_ref);
1389                         btrfs_set_ref_generation(path->nodes[0], ref,
1390                                                  extent_op->generation);
1391                         btrfs_mark_buffer_dirty(path->nodes[0]);
1392                         btrfs_release_path(extent_root, path);
1393                 } else {
1394                         BUG_ON(1);
1395                 }
1396                 kfree(extent_op);
1397         }
1398         btrfs_free_path(path);
1399         return 0;
1400 }
1401
1402 static int pin_down_bytes(struct btrfs_trans_handle *trans,
1403                           struct btrfs_root *root,
1404                           u64 bytenr, u64 num_bytes, int is_data)
1405 {
1406         int err = 0;
1407         struct extent_buffer *buf;
1408
1409         if (is_data)
1410                 goto pinit;
1411
1412         buf = btrfs_find_tree_block(root, bytenr, num_bytes);
1413         if (!buf)
1414                 goto pinit;
1415
1416         /* we can reuse a block if it hasn't been written
1417          * and it is from this transaction.  We can't
1418          * reuse anything from the tree log root because
1419          * it has tiny sub-transactions.
1420          */
1421         if (btrfs_buffer_uptodate(buf, 0)) {
1422                 u64 header_owner = btrfs_header_owner(buf);
1423                 u64 header_transid = btrfs_header_generation(buf);
1424                 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
1425                     header_owner != BTRFS_TREE_RELOC_OBJECTID &&
1426                     header_transid == trans->transid &&
1427                     !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
1428                         clean_tree_block(NULL, root, buf);
1429                         free_extent_buffer(buf);
1430                         return 1;
1431                 }
1432         }
1433         free_extent_buffer(buf);
1434 pinit:
1435         update_pinned_extents(root, bytenr, num_bytes, 1);
1436
1437         BUG_ON(err < 0);
1438         return 0;
1439 }
1440
1441 /*
1442  * remove an extent from the root, returns 0 on success
1443  */
1444 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
1445                          *root, u64 bytenr, u64 num_bytes, u64 parent,
1446                          u64 root_objectid, u64 ref_generation,
1447                          u64 owner_objectid, int pin, int mark_free)
1448 {
1449         struct btrfs_path *path;
1450         struct btrfs_key key;
1451         struct btrfs_fs_info *info = root->fs_info;
1452         struct btrfs_extent_ops *ops = info->extent_ops;
1453         struct btrfs_root *extent_root = info->extent_root;
1454         struct extent_buffer *leaf;
1455         int ret;
1456         int extent_slot = 0;
1457         int found_extent = 0;
1458         int num_to_del = 1;
1459         struct btrfs_extent_item *ei;
1460         u32 refs;
1461
1462         key.objectid = bytenr;
1463         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
1464         key.offset = num_bytes;
1465
1466         path = btrfs_alloc_path();
1467         if (!path)
1468                 return -ENOMEM;
1469
1470         ret = lookup_extent_backref(trans, extent_root, path,
1471                                     bytenr, parent, root_objectid,
1472                                     ref_generation, owner_objectid, 1);
1473         if (ret == 0) {
1474                 struct btrfs_key found_key;
1475                 extent_slot = path->slots[0];
1476                 while(extent_slot > 0) {
1477                         extent_slot--;
1478                         btrfs_item_key_to_cpu(path->nodes[0], &found_key,
1479                                               extent_slot);
1480                         if (found_key.objectid != bytenr)
1481                                 break;
1482                         if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
1483                             found_key.offset == num_bytes) {
1484                                 found_extent = 1;
1485                                 break;
1486                         }
1487                         if (path->slots[0] - extent_slot > 5)
1488                                 break;
1489                 }
1490                 if (!found_extent) {
1491                         ret = remove_extent_backref(trans, extent_root, path);
1492                         BUG_ON(ret);
1493                         btrfs_release_path(extent_root, path);
1494                         ret = btrfs_search_slot(trans, extent_root,
1495                                                 &key, path, -1, 1);
1496                         BUG_ON(ret);
1497                         extent_slot = path->slots[0];
1498                 }
1499         } else {
1500                 btrfs_print_leaf(extent_root, path->nodes[0]);
1501                 printk("Unable to find ref byte nr %llu root %llu "
1502                        " gen %llu owner %llu\n",
1503                        (unsigned long long)bytenr,
1504                        (unsigned long long)root_objectid,
1505                        (unsigned long long)ref_generation,
1506                        (unsigned long long)owner_objectid);
1507                 BUG_ON(1);
1508         }
1509
1510         leaf = path->nodes[0];
1511         ei = btrfs_item_ptr(leaf, extent_slot,
1512                             struct btrfs_extent_item);
1513         refs = btrfs_extent_refs(leaf, ei);
1514         BUG_ON(refs == 0);
1515         refs -= 1;
1516         btrfs_set_extent_refs(leaf, ei, refs);
1517
1518         btrfs_mark_buffer_dirty(leaf);
1519
1520         if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) {
1521                 struct btrfs_extent_ref *ref;
1522                 ref = btrfs_item_ptr(leaf, path->slots[0],
1523                                      struct btrfs_extent_ref);
1524                 BUG_ON(btrfs_ref_num_refs(leaf, ref) != 1);
1525                 /* if the back ref and the extent are next to each other
1526                  * they get deleted below in one shot
1527                  */
1528                 path->slots[0] = extent_slot;
1529                 num_to_del = 2;
1530         } else if (found_extent) {
1531                 /* otherwise delete the extent back ref */
1532                 ret = remove_extent_backref(trans, extent_root, path);
1533                 BUG_ON(ret);
1534                 /* if refs are 0, we need to setup the path for deletion */
1535                 if (refs == 0) {
1536                         btrfs_release_path(extent_root, path);
1537                         ret = btrfs_search_slot(trans, extent_root, &key, path,
1538                                                 -1, 1);
1539                         if (ret < 0)
1540                                 return ret;
1541                         BUG_ON(ret);
1542                 }
1543         }
1544
1545         if (refs == 0) {
1546                 u64 super_used;
1547                 u64 root_used;
1548
1549                 if (pin) {
1550                         ret = pin_down_bytes(trans, root, bytenr, num_bytes, 0);
1551                         if (ret > 0)
1552                                 mark_free = 1;
1553                         BUG_ON(ret < 0);
1554                 }
1555
1556                 /* block accounting for super block */
1557                 super_used = btrfs_super_bytes_used(&info->super_copy);
1558                 btrfs_set_super_bytes_used(&info->super_copy,
1559                                            super_used - num_bytes);
1560
1561                 /* block accounting for root item */
1562                 root_used = btrfs_root_used(&root->root_item);
1563                 btrfs_set_root_used(&root->root_item,
1564                                            root_used - num_bytes);
1565                 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
1566                                       num_to_del);
1567                 if (ret)
1568                         return ret;
1569
1570                 if (ops && ops->free_extent)
1571                         ops->free_extent(root, bytenr, num_bytes);
1572
1573                 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
1574                                          mark_free);
1575                 BUG_ON(ret);
1576         }
1577         btrfs_free_path(path);
1578         finish_current_insert(trans, extent_root);
1579         return ret;
1580 }
1581
1582 /*
1583  * find all the blocks marked as pending in the radix tree and remove
1584  * them from the extent map
1585  */
1586 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
1587                                btrfs_root *extent_root)
1588 {
1589         int ret;
1590         int err = 0;
1591         int mark_free = 0;
1592         u64 start;
1593         u64 end;
1594         u64 priv;
1595         struct extent_io_tree *pending_del;
1596         struct extent_io_tree *extent_ins;
1597         struct pending_extent_op *extent_op;
1598
1599         extent_ins = &extent_root->fs_info->extent_ins;
1600         pending_del = &extent_root->fs_info->pending_del;
1601
1602         while(1) {
1603                 ret = find_first_extent_bit(pending_del, 0, &start, &end,
1604                                             EXTENT_LOCKED);
1605                 if (ret)
1606                         break;
1607
1608                 ret = get_state_private(pending_del, start, &priv);
1609                 BUG_ON(ret);
1610                 extent_op = (struct pending_extent_op *)(unsigned long)priv;
1611
1612                 clear_extent_bits(pending_del, start, end, EXTENT_LOCKED,
1613                                   GFP_NOFS);
1614
1615                 ret = pin_down_bytes(trans, extent_root, start,
1616                                      end + 1 - start, 0);
1617                 mark_free = ret > 0;
1618                 if (!test_range_bit(extent_ins, start, end,
1619                                     EXTENT_LOCKED, 0)) {
1620 free_extent:
1621                         ret = __free_extent(trans, extent_root,
1622                                             start, end + 1 - start,
1623                                             extent_op->orig_parent,
1624                                             extent_root->root_key.objectid,
1625                                             extent_op->orig_generation,
1626                                             extent_op->level, 0, mark_free);
1627                         kfree(extent_op);
1628                 } else {
1629                         kfree(extent_op);
1630                         ret = get_state_private(extent_ins, start, &priv);
1631                         BUG_ON(ret);
1632                         extent_op = (struct pending_extent_op *)
1633                                                         (unsigned long)priv;
1634
1635                         clear_extent_bits(extent_ins, start, end,
1636                                           EXTENT_LOCKED, GFP_NOFS);
1637
1638                         if (extent_op->type == PENDING_BACKREF_UPDATE)
1639                                 goto free_extent;
1640
1641                         ret = update_block_group(trans, extent_root, start,
1642                                                 end + 1 - start, 0, mark_free);
1643                         BUG_ON(ret);
1644                         kfree(extent_op);
1645                 }
1646                 if (ret)
1647                         err = ret;
1648         }
1649         return err;
1650 }
1651
1652 /*
1653  * remove an extent from the root, returns 0 on success
1654  */
1655 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
1656                       *root, u64 bytenr, u64 num_bytes, u64 parent,
1657                       u64 root_objectid, u64 ref_generation,
1658                       u64 owner_objectid, int pin)
1659 {
1660         struct btrfs_root *extent_root = root->fs_info->extent_root;
1661         int pending_ret;
1662         int ret;
1663
1664         WARN_ON(num_bytes < root->sectorsize);
1665         if (root == extent_root) {
1666                 struct pending_extent_op *extent_op;
1667
1668                 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
1669                 BUG_ON(!extent_op);
1670
1671                 extent_op->type = PENDING_EXTENT_DELETE;
1672                 extent_op->bytenr = bytenr;
1673                 extent_op->num_bytes = num_bytes;
1674                 extent_op->parent = parent;
1675                 extent_op->orig_parent = parent;
1676                 extent_op->generation = ref_generation;
1677                 extent_op->orig_generation = ref_generation;
1678                 extent_op->level = (int)owner_objectid;
1679
1680                 set_extent_bits(&root->fs_info->pending_del,
1681                                 bytenr, bytenr + num_bytes - 1,
1682                                 EXTENT_LOCKED, GFP_NOFS);
1683                 set_state_private(&root->fs_info->pending_del,
1684                                   bytenr, (unsigned long)extent_op);
1685                 return 0;
1686         }
1687         ret = __free_extent(trans, root, bytenr, num_bytes, parent,
1688                             root_objectid, ref_generation,
1689                             owner_objectid, pin, pin == 0);
1690         pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
1691         return ret ? ret : pending_ret;
1692 }
1693
1694 static u64 stripe_align(struct btrfs_root *root, u64 val)
1695 {
1696         u64 mask = ((u64)root->stripesize - 1);
1697         u64 ret = (val + mask) & ~mask;
1698         return ret;
1699 }
1700
1701 /*
1702  * walks the btree of allocated extents and find a hole of a given size.
1703  * The key ins is changed to record the hole:
1704  * ins->objectid == block start
1705  * ins->flags = BTRFS_EXTENT_ITEM_KEY
1706  * ins->offset == number of blocks
1707  * Any available blocks before search_start are skipped.
1708  */
1709 static int noinline find_free_extent(struct btrfs_trans_handle *trans,
1710                                      struct btrfs_root *orig_root,
1711                                      u64 num_bytes, u64 empty_size,
1712                                      u64 search_start, u64 search_end,
1713                                      u64 hint_byte, struct btrfs_key *ins,
1714                                      u64 exclude_start, u64 exclude_nr,
1715                                      int data)
1716 {
1717         int ret;
1718         u64 orig_search_start = search_start;
1719         struct btrfs_root * root = orig_root->fs_info->extent_root;
1720         struct btrfs_fs_info *info = root->fs_info;
1721         u64 total_needed = num_bytes;
1722         struct btrfs_block_group_cache *block_group;
1723         int full_scan = 0;
1724         int wrapped = 0;
1725
1726         WARN_ON(num_bytes < root->sectorsize);
1727         btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
1728
1729         if (hint_byte) {
1730                 block_group = btrfs_lookup_first_block_group(info, hint_byte);
1731                 if (!block_group)
1732                         hint_byte = search_start;
1733                 block_group = btrfs_find_block_group(root, block_group,
1734                                                      hint_byte, data, 1);
1735         } else {
1736                 block_group = btrfs_find_block_group(root,
1737                                                      trans->block_group,
1738                                                      search_start, data, 1);
1739         }
1740
1741         total_needed += empty_size;
1742
1743 check_failed:
1744         if (!block_group) {
1745                 block_group = btrfs_lookup_first_block_group(info,
1746                                                              search_start);
1747                 if (!block_group)
1748                         block_group = btrfs_lookup_first_block_group(info,
1749                                                        orig_search_start);
1750         }
1751         ret = find_search_start(root, &block_group, &search_start,
1752                                 total_needed, data);
1753         if (ret)
1754                 goto error;
1755
1756         search_start = stripe_align(root, search_start);
1757         ins->objectid = search_start;
1758         ins->offset = num_bytes;
1759
1760         if (ins->objectid + num_bytes >
1761             block_group->key.objectid + block_group->key.offset) {
1762                 search_start = block_group->key.objectid +
1763                         block_group->key.offset;
1764                 goto new_group;
1765         }
1766
1767         if (test_range_bit(&info->extent_ins, ins->objectid,
1768                            ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) {
1769                 search_start = ins->objectid + num_bytes;
1770                 goto new_group;
1771         }
1772
1773         if (test_range_bit(&info->pinned_extents, ins->objectid,
1774                            ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) {
1775                 search_start = ins->objectid + num_bytes;
1776                 goto new_group;
1777         }
1778
1779         if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start &&
1780             ins->objectid < exclude_start + exclude_nr)) {
1781                 search_start = exclude_start + exclude_nr;
1782                 goto new_group;
1783         }
1784
1785         if (!(data & BTRFS_BLOCK_GROUP_DATA)) {
1786                 block_group = btrfs_lookup_block_group(info, ins->objectid);
1787                 if (block_group)
1788                         trans->block_group = block_group;
1789         }
1790         ins->offset = num_bytes;
1791         return 0;
1792
1793 new_group:
1794         block_group = btrfs_lookup_first_block_group(info, search_start);
1795         if (!block_group) {
1796                 search_start = orig_search_start;
1797                 if (full_scan) {
1798                         ret = -ENOSPC;
1799                         goto error;
1800                 }
1801                 if (wrapped) {
1802                         if (!full_scan)
1803                                 total_needed -= empty_size;
1804                         full_scan = 1;
1805                 } else
1806                         wrapped = 1;
1807         }
1808         cond_resched();
1809         block_group = btrfs_find_block_group(root, block_group,
1810                                              search_start, data, 0);
1811         goto check_failed;
1812
1813 error:
1814         return ret;
1815 }
1816 /*
1817  * finds a free extent and does all the dirty work required for allocation
1818  * returns the key for the extent through ins, and a tree buffer for
1819  * the first block of the extent through buf.
1820  *
1821  * returns 0 if everything worked, non-zero otherwise.
1822  */
1823 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
1824                        struct btrfs_root *root,
1825                        u64 num_bytes, u64 parent,
1826                        u64 root_objectid, u64 ref_generation,
1827                        u64 owner, u64 empty_size, u64 hint_byte,
1828                        u64 search_end, struct btrfs_key *ins, int data)
1829 {
1830         int ret;
1831         int pending_ret;
1832         u64 super_used, root_used;
1833         u64 search_start = 0;
1834         u64 alloc_profile;
1835         u32 sizes[2];
1836         struct btrfs_fs_info *info = root->fs_info;
1837         struct btrfs_root *extent_root = info->extent_root;
1838         struct btrfs_path *path;
1839         struct btrfs_extent_item *extent_item;
1840         struct btrfs_extent_ref *ref;
1841         struct btrfs_key keys[2];
1842
1843         if (info->extent_ops) {
1844                 struct btrfs_extent_ops *ops = info->extent_ops;
1845                 ret = ops->alloc_extent(root, num_bytes, hint_byte, ins);
1846                 BUG_ON(ret);
1847                 goto found;
1848         }
1849
1850         if (data) {
1851                 alloc_profile = info->avail_data_alloc_bits &
1852                                 info->data_alloc_profile;
1853                 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
1854         } else if ((info->system_allocs > 0 || root == info->chunk_root) &&
1855                    info->system_allocs >= 0) {
1856                 alloc_profile = info->avail_system_alloc_bits &
1857                                 info->system_alloc_profile;
1858                 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
1859         } else {
1860                 alloc_profile = info->avail_metadata_alloc_bits &
1861                                 info->metadata_alloc_profile;
1862                 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
1863         }
1864
1865         if (root->ref_cows) {
1866                 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
1867                         ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1868                                              num_bytes,
1869                                              BTRFS_BLOCK_GROUP_METADATA);
1870                         BUG_ON(ret);
1871                 }
1872                 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1873                                      num_bytes + 2 * 1024 * 1024, data);
1874                 BUG_ON(ret);
1875         }
1876
1877         WARN_ON(num_bytes < root->sectorsize);
1878         ret = find_free_extent(trans, root, num_bytes, empty_size,
1879                                search_start, search_end, hint_byte, ins,
1880                                trans->alloc_exclude_start,
1881                                trans->alloc_exclude_nr, data);
1882         BUG_ON(ret);
1883 found:
1884         if (ret)
1885                 return ret;
1886
1887         if (parent == 0)
1888                 parent = ins->objectid;
1889
1890         /* block accounting for super block */
1891         super_used = btrfs_super_bytes_used(&info->super_copy);
1892         btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
1893
1894         /* block accounting for root item */
1895         root_used = btrfs_root_used(&root->root_item);
1896         btrfs_set_root_used(&root->root_item, root_used + num_bytes);
1897
1898         clear_extent_dirty(&root->fs_info->free_space_cache,
1899                            ins->objectid, ins->objectid + ins->offset - 1,
1900                            GFP_NOFS);
1901
1902         if (root == extent_root) {
1903                 struct pending_extent_op *extent_op;
1904
1905                 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
1906                 BUG_ON(!extent_op);
1907
1908                 extent_op->type = PENDING_EXTENT_INSERT;
1909                 extent_op->bytenr = ins->objectid;
1910                 extent_op->num_bytes = ins->offset;
1911                 extent_op->parent = parent;
1912                 extent_op->orig_parent = 0;
1913                 extent_op->generation = ref_generation;
1914                 extent_op->orig_generation = 0;
1915                 extent_op->level = (int)owner;
1916
1917                 set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
1918                                 ins->objectid + ins->offset - 1,
1919                                 EXTENT_LOCKED, GFP_NOFS);
1920                 set_state_private(&root->fs_info->extent_ins,
1921                                   ins->objectid, (unsigned long)extent_op);
1922                 goto update_block;
1923         }
1924
1925         WARN_ON(trans->alloc_exclude_nr);
1926         trans->alloc_exclude_start = ins->objectid;
1927         trans->alloc_exclude_nr = ins->offset;
1928
1929         memcpy(&keys[0], ins, sizeof(*ins));
1930         keys[1].objectid = ins->objectid;
1931         keys[1].type = BTRFS_EXTENT_REF_KEY;
1932         keys[1].offset = parent;
1933         sizes[0] = sizeof(*extent_item);
1934         sizes[1] = sizeof(*ref);
1935
1936         path = btrfs_alloc_path();
1937         BUG_ON(!path);
1938
1939         ret = btrfs_insert_empty_items(trans, extent_root, path, keys,
1940                                        sizes, 2);
1941
1942         BUG_ON(ret);
1943         extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1944                                      struct btrfs_extent_item);
1945         btrfs_set_extent_refs(path->nodes[0], extent_item, 1);
1946         ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1947                              struct btrfs_extent_ref);
1948
1949         btrfs_set_ref_root(path->nodes[0], ref, root_objectid);
1950         btrfs_set_ref_generation(path->nodes[0], ref, ref_generation);
1951         btrfs_set_ref_objectid(path->nodes[0], ref, owner);
1952         btrfs_set_ref_num_refs(path->nodes[0], ref, 1);
1953
1954         btrfs_mark_buffer_dirty(path->nodes[0]);
1955
1956         trans->alloc_exclude_start = 0;
1957         trans->alloc_exclude_nr = 0;
1958         btrfs_free_path(path);
1959         finish_current_insert(trans, extent_root);
1960         pending_ret = del_pending_extents(trans, extent_root);
1961
1962         if (ret) {
1963                 return ret;
1964         }
1965         if (pending_ret) {
1966                 return pending_ret;
1967         }
1968
1969 update_block:
1970         ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0);
1971         if (ret) {
1972                 printk("update block group failed for %llu %llu\n",
1973                        (unsigned long long)ins->objectid,
1974                        (unsigned long long)ins->offset);
1975                 BUG();
1976         }
1977         return 0;
1978 }
1979
1980 /*
1981  * helper function to allocate a block for a given tree
1982  * returns the tree buffer or NULL.
1983  */
1984 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
1985                                              struct btrfs_root *root,
1986                                              u32 blocksize, u64 parent,
1987                                              u64 root_objectid,
1988                                              u64 ref_generation,
1989                                              int level,
1990                                              u64 hint,
1991                                              u64 empty_size)
1992 {
1993         struct btrfs_key ins;
1994         int ret;
1995         struct extent_buffer *buf;
1996
1997         ret = btrfs_alloc_extent(trans, root, blocksize, parent,
1998                                  root_objectid, ref_generation,
1999                                  level, empty_size, hint,
2000                                  (u64)-1, &ins, 0);
2001         if (ret) {
2002                 BUG_ON(ret > 0);
2003                 return ERR_PTR(ret);
2004         }
2005         buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize);
2006         if (!buf) {
2007                 if (parent == 0)
2008                         parent = ins.objectid;
2009                 btrfs_free_extent(trans, root, ins.objectid, blocksize,
2010                                   parent, root->root_key.objectid,
2011                                   ref_generation, level, 0);
2012                 BUG_ON(1);
2013                 return ERR_PTR(-ENOMEM);
2014         }
2015         btrfs_set_buffer_uptodate(buf);
2016         trans->blocks_used++;
2017         return buf;
2018 }
2019
2020 static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans,
2021                                   struct btrfs_root *root,
2022                                   struct extent_buffer *leaf)
2023 {
2024         u64 leaf_owner;
2025         u64 leaf_generation;
2026         struct btrfs_key key;
2027         struct btrfs_file_extent_item *fi;
2028         int i;
2029         int nritems;
2030         int ret;
2031
2032         BUG_ON(!btrfs_is_leaf(leaf));
2033         nritems = btrfs_header_nritems(leaf);
2034         leaf_owner = btrfs_header_owner(leaf);
2035         leaf_generation = btrfs_header_generation(leaf);
2036
2037         for (i = 0; i < nritems; i++) {
2038                 u64 disk_bytenr;
2039
2040                 btrfs_item_key_to_cpu(leaf, &key, i);
2041                 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2042                         continue;
2043                 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
2044                 if (btrfs_file_extent_type(leaf, fi) ==
2045                     BTRFS_FILE_EXTENT_INLINE)
2046                         continue;
2047                 /*
2048                  * FIXME make sure to insert a trans record that
2049                  * repeats the snapshot del on crash
2050                  */
2051                 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
2052                 if (disk_bytenr == 0)
2053                         continue;
2054                 ret = btrfs_free_extent(trans, root, disk_bytenr,
2055                                 btrfs_file_extent_disk_num_bytes(leaf, fi),
2056                                 leaf->start, leaf_owner, leaf_generation,
2057                                 key.objectid, 0);
2058                 BUG_ON(ret);
2059         }
2060         return 0;
2061 }
2062
2063 static void noinline reada_walk_down(struct btrfs_root *root,
2064                                      struct extent_buffer *node,
2065                                      int slot)
2066 {
2067         u64 bytenr;
2068         u64 last = 0;
2069         u32 nritems;
2070         u32 refs;
2071         u32 blocksize;
2072         int ret;
2073         int i;
2074         int level;
2075         int skipped = 0;
2076
2077         nritems = btrfs_header_nritems(node);
2078         level = btrfs_header_level(node);
2079         if (level)
2080                 return;
2081
2082         for (i = slot; i < nritems && skipped < 32; i++) {
2083                 bytenr = btrfs_node_blockptr(node, i);
2084                 if (last && ((bytenr > last && bytenr - last > 32 * 1024) ||
2085                              (last > bytenr && last - bytenr > 32 * 1024))) {
2086                         skipped++;
2087                         continue;
2088                 }
2089                 blocksize = btrfs_level_size(root, level - 1);
2090                 if (i != slot) {
2091                         ret = lookup_extent_ref(NULL, root, bytenr,
2092                                                 blocksize, &refs);
2093                         BUG_ON(ret);
2094                         if (refs != 1) {
2095                                 skipped++;
2096                                 continue;
2097                         }
2098                 }
2099                 mutex_unlock(&root->fs_info->fs_mutex);
2100                 ret = readahead_tree_block(root, bytenr, blocksize,
2101                                            btrfs_node_ptr_generation(node, i));
2102                 last = bytenr + blocksize;
2103                 cond_resched();
2104                 mutex_lock(&root->fs_info->fs_mutex);
2105                 if (ret)
2106                         break;
2107         }
2108 }
2109
2110 /*
2111  * helper function for drop_snapshot, this walks down the tree dropping ref
2112  * counts as it goes.
2113  */
2114 static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
2115                                    struct btrfs_root *root,
2116                                    struct btrfs_path *path, int *level)
2117 {
2118         u64 root_owner;
2119         u64 root_gen;
2120         u64 bytenr;
2121         u64 ptr_gen;
2122         struct extent_buffer *next;
2123         struct extent_buffer *cur;
2124         struct extent_buffer *parent;
2125         u32 blocksize;
2126         int ret;
2127         u32 refs;
2128
2129         WARN_ON(*level < 0);
2130         WARN_ON(*level >= BTRFS_MAX_LEVEL);
2131         ret = lookup_extent_ref(trans, root,
2132                                 path->nodes[*level]->start,
2133                                 path->nodes[*level]->len, &refs);
2134         BUG_ON(ret);
2135         if (refs > 1)
2136                 goto out;
2137
2138         /*
2139          * walk down to the last node level and free all the leaves
2140          */
2141         while(*level >= 0) {
2142                 WARN_ON(*level < 0);
2143                 WARN_ON(*level >= BTRFS_MAX_LEVEL);
2144                 cur = path->nodes[*level];
2145
2146                 if (btrfs_header_level(cur) != *level)
2147                         WARN_ON(1);
2148
2149                 if (path->slots[*level] >=
2150                     btrfs_header_nritems(cur))
2151                         break;
2152                 if (*level == 0) {
2153                         ret = drop_leaf_ref(trans, root, cur);
2154                         BUG_ON(ret);
2155                         break;
2156                 }
2157                 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
2158                 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
2159                 blocksize = btrfs_level_size(root, *level - 1);
2160                 ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs);
2161                 BUG_ON(ret);
2162                 if (refs != 1) {
2163                         parent = path->nodes[*level];
2164                         root_owner = btrfs_header_owner(parent);
2165                         root_gen = btrfs_header_generation(parent);
2166                         path->slots[*level]++;
2167                         ret = btrfs_free_extent(trans, root, bytenr, blocksize,
2168                                                 parent->start, root_owner,
2169                                                 root_gen, *level - 1, 1);
2170                         BUG_ON(ret);
2171                         continue;
2172                 }
2173                 next = btrfs_find_tree_block(root, bytenr, blocksize);
2174                 if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
2175                         free_extent_buffer(next);
2176                         reada_walk_down(root, cur, path->slots[*level]);
2177                         mutex_unlock(&root->fs_info->fs_mutex);
2178                         next = read_tree_block(root, bytenr, blocksize,
2179                                                ptr_gen);
2180                         mutex_lock(&root->fs_info->fs_mutex);
2181                 }
2182                 WARN_ON(*level <= 0);
2183                 if (path->nodes[*level-1])
2184                         free_extent_buffer(path->nodes[*level-1]);
2185                 path->nodes[*level-1] = next;
2186                 *level = btrfs_header_level(next);
2187                 path->slots[*level] = 0;
2188         }
2189 out:
2190         WARN_ON(*level < 0);
2191         WARN_ON(*level >= BTRFS_MAX_LEVEL);
2192
2193         if (path->nodes[*level] == root->node) {
2194                 root_owner = root->root_key.objectid;
2195                 parent = path->nodes[*level];
2196         } else {
2197                 parent = path->nodes[*level + 1];
2198                 root_owner = btrfs_header_owner(parent);
2199         }
2200
2201         root_gen = btrfs_header_generation(parent);
2202         ret = btrfs_free_extent(trans, root, path->nodes[*level]->start,
2203                                 path->nodes[*level]->len, parent->start,
2204                                 root_owner, root_gen, *level, 1);
2205         free_extent_buffer(path->nodes[*level]);
2206         path->nodes[*level] = NULL;
2207         *level += 1;
2208         BUG_ON(ret);
2209         return 0;
2210 }
2211
2212 /*
2213  * helper for dropping snapshots.  This walks back up the tree in the path
2214  * to find the first node higher up where we haven't yet gone through
2215  * all the slots
2216  */
2217 static int noinline walk_up_tree(struct btrfs_trans_handle *trans,
2218                                  struct btrfs_root *root,
2219                                  struct btrfs_path *path, int *level)
2220 {
2221         u64 root_owner;
2222         u64 root_gen;
2223         struct btrfs_root_item *root_item = &root->root_item;
2224         int i;
2225         int slot;
2226         int ret;
2227
2228         for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
2229                 slot = path->slots[i];
2230                 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
2231                         struct extent_buffer *node;
2232                         struct btrfs_disk_key disk_key;
2233                         node = path->nodes[i];
2234                         path->slots[i]++;
2235                         *level = i;
2236                         WARN_ON(*level == 0);
2237                         btrfs_node_key(node, &disk_key, path->slots[i]);
2238                         memcpy(&root_item->drop_progress,
2239                                &disk_key, sizeof(disk_key));
2240                         root_item->drop_level = i;
2241                         return 0;
2242                 } else {
2243                         struct extent_buffer *parent;
2244                         if (path->nodes[*level] == root->node)
2245                                 parent = path->nodes[*level];
2246                         else
2247                                 parent = path->nodes[*level + 1];
2248
2249                         root_owner = btrfs_header_owner(parent);
2250                         root_gen = btrfs_header_generation(parent);
2251                         ret = btrfs_free_extent(trans, root,
2252                                                 path->nodes[*level]->start,
2253                                                 path->nodes[*level]->len,
2254                                                 parent->start, root_owner,
2255                                                 root_gen, *level, 1);
2256                         BUG_ON(ret);
2257                         free_extent_buffer(path->nodes[*level]);
2258                         path->nodes[*level] = NULL;
2259                         *level = i + 1;
2260                 }
2261         }
2262         return 1;
2263 }
2264
2265 /*
2266  * drop the reference count on the tree rooted at 'snap'.  This traverses
2267  * the tree freeing any blocks that have a ref count of zero after being
2268  * decremented.
2269  */
2270 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
2271                         *root)
2272 {
2273         int ret = 0;
2274         int wret;
2275         int level;
2276         struct btrfs_path *path;
2277         int i;
2278         int orig_level;
2279         struct btrfs_root_item *root_item = &root->root_item;
2280
2281         path = btrfs_alloc_path();
2282         BUG_ON(!path);
2283
2284         level = btrfs_header_level(root->node);
2285         orig_level = level;
2286         if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
2287                 path->nodes[level] = root->node;
2288                 extent_buffer_get(root->node);
2289                 path->slots[level] = 0;
2290         } else {
2291                 struct btrfs_key key;
2292                 struct btrfs_disk_key found_key;
2293                 struct extent_buffer *node;
2294
2295                 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
2296                 level = root_item->drop_level;
2297                 path->lowest_level = level;
2298                 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2299                 if (wret < 0) {
2300                         ret = wret;
2301                         goto out;
2302                 }
2303                 node = path->nodes[level];
2304                 btrfs_node_key(node, &found_key, path->slots[level]);
2305                 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
2306                                sizeof(found_key)));
2307         }
2308         while(1) {
2309                 wret = walk_down_tree(trans, root, path, &level);
2310                 if (wret < 0)
2311                         ret = wret;
2312                 if (wret != 0)
2313                         break;
2314
2315                 wret = walk_up_tree(trans, root, path, &level);
2316                 if (wret < 0)
2317                         ret = wret;
2318                 if (wret != 0)
2319                         break;
2320                 /*
2321                 ret = -EAGAIN;
2322                 break;
2323                 */
2324         }
2325         for (i = 0; i <= orig_level; i++) {
2326                 if (path->nodes[i]) {
2327                         free_extent_buffer(path->nodes[i]);
2328                         path->nodes[i] = NULL;
2329                 }
2330         }
2331 out:
2332         btrfs_free_path(path);
2333         return ret;
2334 }
2335
2336 int btrfs_free_block_groups(struct btrfs_fs_info *info)
2337 {
2338         u64 start;
2339         u64 end;
2340         u64 ptr;
2341         int ret;
2342         while(1) {
2343                 ret = find_first_extent_bit(&info->block_group_cache, 0,
2344                                             &start, &end, (unsigned int)-1);
2345                 if (ret)
2346                         break;
2347                 ret = get_state_private(&info->block_group_cache, start, &ptr);
2348                 if (!ret)
2349                         kfree((void *)(unsigned long)ptr);
2350                 clear_extent_bits(&info->block_group_cache, start,
2351                                   end, (unsigned int)-1, GFP_NOFS);
2352         }
2353         while(1) {
2354                 ret = find_first_extent_bit(&info->free_space_cache, 0,
2355                                             &start, &end, EXTENT_DIRTY);
2356                 if (ret)
2357                         break;
2358                 clear_extent_dirty(&info->free_space_cache, start,
2359                                    end, GFP_NOFS);
2360         }
2361         return 0;
2362 }
2363
2364 int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
2365                            struct btrfs_key *key)
2366 {
2367         int ret;
2368         struct btrfs_key found_key;
2369         struct extent_buffer *leaf;
2370         int slot;
2371
2372         ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
2373         if (ret < 0)
2374                 return ret;
2375         while(1) {
2376                 slot = path->slots[0];
2377                 leaf = path->nodes[0];
2378                 if (slot >= btrfs_header_nritems(leaf)) {
2379                         ret = btrfs_next_leaf(root, path);
2380                         if (ret == 0)
2381                                 continue;
2382                         if (ret < 0)
2383                                 goto error;
2384                         break;
2385                 }
2386                 btrfs_item_key_to_cpu(leaf, &found_key, slot);
2387
2388                 if (found_key.objectid >= key->objectid &&
2389                     found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY)
2390                         return 0;
2391                 path->slots[0]++;
2392         }
2393         ret = -ENOENT;
2394 error:
2395         return ret;
2396 }
2397
2398 int btrfs_read_block_groups(struct btrfs_root *root)
2399 {
2400         struct btrfs_path *path;
2401         int ret;
2402         int bit;
2403         struct btrfs_block_group_cache *cache;
2404         struct btrfs_fs_info *info = root->fs_info;
2405         struct btrfs_space_info *space_info;
2406         struct extent_io_tree *block_group_cache;
2407         struct btrfs_key key;
2408         struct btrfs_key found_key;
2409         struct extent_buffer *leaf;
2410
2411         block_group_cache = &info->block_group_cache;
2412
2413         root = info->extent_root;
2414         key.objectid = 0;
2415         key.offset = 0;
2416         btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
2417         path = btrfs_alloc_path();
2418         if (!path)
2419                 return -ENOMEM;
2420
2421         while(1) {
2422                 ret = find_first_block_group(root, path, &key);
2423                 if (ret > 0) {
2424                         ret = 0;
2425                         goto error;
2426                 }
2427                 if (ret != 0) {
2428                         goto error;
2429                 }
2430                 leaf = path->nodes[0];
2431                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2432                 cache = kzalloc(sizeof(*cache), GFP_NOFS);
2433                 if (!cache) {
2434                         ret = -ENOMEM;
2435                         break;
2436                 }
2437
2438                 read_extent_buffer(leaf, &cache->item,
2439                                    btrfs_item_ptr_offset(leaf, path->slots[0]),
2440                                    sizeof(cache->item));
2441                 memcpy(&cache->key, &found_key, sizeof(found_key));
2442                 cache->cached = 0;
2443                 cache->pinned = 0;
2444                 key.objectid = found_key.objectid + found_key.offset;
2445                 btrfs_release_path(root, path);
2446                 cache->flags = btrfs_block_group_flags(&cache->item);
2447                 bit = 0;
2448                 if (cache->flags & BTRFS_BLOCK_GROUP_DATA) {
2449                         bit = BLOCK_GROUP_DATA;
2450                 } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {
2451                         bit = BLOCK_GROUP_SYSTEM;
2452                 } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) {
2453                         bit = BLOCK_GROUP_METADATA;
2454                 }
2455                 set_avail_alloc_bits(info, cache->flags);
2456                 if (btrfs_chunk_readonly(root, cache->key.objectid))
2457                         cache->ro = 1;
2458
2459                 ret = update_space_info(info, cache->flags, found_key.offset,
2460                                         btrfs_block_group_used(&cache->item),
2461                                         &space_info);
2462                 BUG_ON(ret);
2463                 cache->space_info = space_info;
2464
2465                 /* use EXTENT_LOCKED to prevent merging */
2466                 set_extent_bits(block_group_cache, found_key.objectid,
2467                                 found_key.objectid + found_key.offset - 1,
2468                                 bit | EXTENT_LOCKED, GFP_NOFS);
2469                 set_state_private(block_group_cache, found_key.objectid,
2470                                   (unsigned long)cache);
2471         }
2472         ret = 0;
2473 error:
2474         btrfs_free_path(path);
2475         return ret;
2476 }
2477
2478 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
2479                            struct btrfs_root *root, u64 bytes_used,
2480                            u64 type, u64 chunk_objectid, u64 chunk_offset,
2481                            u64 size)
2482 {
2483         int ret;
2484         int bit = 0;
2485         struct btrfs_root *extent_root;
2486         struct btrfs_block_group_cache *cache;
2487         struct extent_io_tree *block_group_cache;
2488
2489         extent_root = root->fs_info->extent_root;
2490         block_group_cache = &root->fs_info->block_group_cache;
2491
2492         cache = kzalloc(sizeof(*cache), GFP_NOFS);
2493         BUG_ON(!cache);
2494         cache->key.objectid = chunk_offset;
2495         cache->key.offset = size;
2496
2497         btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY);
2498         btrfs_set_block_group_used(&cache->item, bytes_used);
2499         btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
2500         cache->flags = type;
2501         btrfs_set_block_group_flags(&cache->item, type);
2502
2503         ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
2504                                 &cache->space_info);
2505         BUG_ON(ret);
2506
2507         bit = block_group_state_bits(type);
2508         set_extent_bits(block_group_cache, chunk_offset,
2509                         chunk_offset + size - 1,
2510                         bit | EXTENT_LOCKED, GFP_NOFS);
2511
2512         set_state_private(block_group_cache, chunk_offset,
2513                           (unsigned long)cache);
2514         ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
2515                                 sizeof(cache->item));
2516         BUG_ON(ret);
2517
2518         finish_current_insert(trans, extent_root);
2519         ret = del_pending_extents(trans, extent_root);
2520         BUG_ON(ret);
2521         set_avail_alloc_bits(extent_root->fs_info, type);
2522         return 0;
2523 }
2524
2525 /*
2526  * This is for converter use only.
2527  *
2528  * In that case, we don't know where are free blocks located.
2529  * Therefore all block group cache entries must be setup properly
2530  * before doing any block allocation.
2531  */
2532 int btrfs_make_block_groups(struct btrfs_trans_handle *trans,
2533                             struct btrfs_root *root)
2534 {
2535         u64 total_bytes;
2536         u64 cur_start;
2537         u64 group_type;
2538         u64 group_size;
2539         u64 group_align;
2540         u64 total_data = 0;
2541         u64 total_metadata = 0;
2542         u64 chunk_objectid;
2543         int ret;
2544         int bit;
2545         struct btrfs_root *extent_root;
2546         struct btrfs_block_group_cache *cache;
2547         struct extent_io_tree *block_group_cache;
2548
2549         extent_root = root->fs_info->extent_root;
2550         block_group_cache = &root->fs_info->block_group_cache;
2551         chunk_objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
2552         total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
2553         group_align = 64 * root->sectorsize;
2554
2555         cur_start = 0;
2556         while (cur_start < total_bytes) {
2557                 group_size = total_bytes / 12;
2558                 group_size = min_t(u64, group_size, total_bytes - cur_start);
2559                 if (cur_start == 0) {
2560                         bit = BLOCK_GROUP_SYSTEM;
2561                         group_type = BTRFS_BLOCK_GROUP_SYSTEM;
2562                         group_size /= 4;
2563                         group_size &= ~(group_align - 1);
2564                         group_size = max_t(u64, group_size, 32 * 1024 * 1024);
2565                         group_size = min_t(u64, group_size, 128 * 1024 * 1024);
2566                 } else {
2567                         group_size &= ~(group_align - 1);
2568                         if (total_data >= total_metadata * 2) {
2569                                 group_type = BTRFS_BLOCK_GROUP_METADATA;
2570                                 group_size = min_t(u64, group_size,
2571                                                    1ULL * 1024 * 1024 * 1024);
2572                                 total_metadata += group_size;
2573                         } else {
2574                                 group_type = BTRFS_BLOCK_GROUP_DATA;
2575                                 group_size = min_t(u64, group_size,
2576                                                    5ULL * 1024 * 1024 * 1024);
2577                                 total_data += group_size;
2578                         }
2579                         if ((total_bytes - cur_start) * 4 < group_size * 5)
2580                                 group_size = total_bytes - cur_start;
2581                 }
2582
2583                 cache = kzalloc(sizeof(*cache), GFP_NOFS);
2584                 BUG_ON(!cache);
2585
2586                 cache->key.objectid = cur_start;
2587                 cache->key.offset = group_size;
2588                 btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY);
2589
2590                 btrfs_set_block_group_used(&cache->item, 0);
2591                 btrfs_set_block_group_chunk_objectid(&cache->item,
2592                                                      chunk_objectid);
2593                 btrfs_set_block_group_flags(&cache->item, group_type);
2594
2595                 cache->flags = group_type;
2596
2597                 ret = update_space_info(root->fs_info, group_type, group_size,
2598                                         0, &cache->space_info);
2599                 BUG_ON(ret);
2600                 set_avail_alloc_bits(extent_root->fs_info, group_type);
2601
2602                 set_extent_bits(block_group_cache, cur_start,
2603                                 cur_start + group_size - 1,
2604                                 bit | EXTENT_LOCKED, GFP_NOFS);
2605                 set_state_private(block_group_cache, cur_start,
2606                                   (unsigned long)cache);
2607                 cur_start += group_size;
2608         }
2609         /* then insert all the items */
2610         cur_start = 0;
2611         while(cur_start < total_bytes) {
2612                 cache = btrfs_lookup_block_group(root->fs_info, cur_start);
2613                 BUG_ON(!cache);
2614
2615                 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
2616                                         sizeof(cache->item));
2617                 BUG_ON(ret);
2618
2619                 finish_current_insert(trans, extent_root);
2620                 ret = del_pending_extents(trans, extent_root);
2621                 BUG_ON(ret);
2622
2623                 cur_start = cache->key.objectid + cache->key.offset;
2624         }
2625         return 0;
2626 }
2627
2628 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
2629                              struct btrfs_root *root,
2630                              u64 bytenr, u64 num_bytes, int alloc,
2631                              int mark_free)
2632 {
2633         return update_block_group(trans, root, bytenr, num_bytes,
2634                                   alloc, mark_free);
2635 }