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
  28 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
  29                                  btrfs_root *extent_root);
  30 static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
  31                        *extent_root);
  32
  33 static int inc_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
  34                          *root, u64 bytenr, u32 blocksize)
  35 {
  36         struct btrfs_path path;
  37         int ret;
  38         struct btrfs_key key;
  39         struct btrfs_leaf *l;
  40         struct btrfs_extent_item *item;
  41         u32 refs;
  42
  43         btrfs_init_path(&path);
  44         key.objectid = bytenr;
  45         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
  46         key.offset = blocksize;
  47         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, &path,
  48                                 0, 1);
  49         if (ret != 0)
  50                 BUG();
  51         BUG_ON(ret != 0);
  52         l = &path.nodes[0]->leaf;
  53         item = btrfs_item_ptr(l, path.slots[0], struct btrfs_extent_item);
  54         refs = btrfs_extent_refs(item);
  55         btrfs_set_extent_refs(item, refs + 1);
  56
  57         BUG_ON(list_empty(&path.nodes[0]->dirty));
  58         btrfs_release_path(root->fs_info->extent_root, &path);
  59         finish_current_insert(trans, root->fs_info->extent_root);
  60         run_pending(trans, root->fs_info->extent_root);
  61         return 0;
  62 }
  63
  64 static int lookup_block_ref(struct btrfs_trans_handle *trans, struct btrfs_root
  65                             *root, u64 bytenr, u32 blocksize, u32 *refs)
  66 {
  67         struct btrfs_path path;
  68         int ret;
  69         struct btrfs_key key;
  70         struct btrfs_leaf *l;
  71         struct btrfs_extent_item *item;
  72         btrfs_init_path(&path);
  73         key.objectid = bytenr;
  74         key.offset = blocksize;
  75         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
  76         ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, &path,
  77                                 0, 0);
  78         if (ret != 0)
  79                 BUG();
  80         l = &path.nodes[0]->leaf;
  81         item = btrfs_item_ptr(l, path.slots[0], struct btrfs_extent_item);
  82         *refs = btrfs_extent_refs(item);
  83         btrfs_release_path(root->fs_info->extent_root, &path);
  84         return 0;
  85 }
  86
  87 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
  88                   struct btrfs_buffer *buf)
  89 {
  90         u64 bytenr;
  91         u32 blocksize;
  92         int i;
  93         int level;
  94
  95         if (!root->ref_cows)
  96                 return 0;
  97
  98         level = btrfs_header_level(&buf->node.header) - 1;
  99         blocksize = btrfs_level_size(root, level);
 100
 101         if (btrfs_is_leaf(&buf->node))
 102                 return 0;
 103
 104         for (i = 0; i < btrfs_header_nritems(&buf->node.header); i++) {
 105                 bytenr = btrfs_node_blockptr(&buf->node, i);
 106                 inc_block_ref(trans, root, bytenr, blocksize);
 107         }
 108
 109         return 0;
 110 }
 111
 112 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
 113                        struct btrfs_root *root)
 114 {
 115         return inc_block_ref(trans, root, root->node->bytenr,
 116                              root->node->size);
 117 }
 118
 119 static int write_one_cache_group(struct btrfs_trans_handle *trans,
 120                                  struct btrfs_root *root,
 121                                  struct btrfs_path *path,
 122                                  struct btrfs_block_group_cache *cache)
 123 {
 124         int ret;
 125         int pending_ret;
 126         struct btrfs_root *extent_root = root->fs_info->extent_root;
 127         struct btrfs_block_group_item *bi;
 128
 129         ret = btrfs_search_slot(trans, root->fs_info->extent_root,
 130                                 &cache->key, path, 0, 1);
 131         BUG_ON(ret);
 132         bi = btrfs_item_ptr(&path->nodes[0]->leaf, path->slots[0],
 133                             struct btrfs_block_group_item);
 134         memcpy(bi, &cache->item, sizeof(*bi));
 135         dirty_tree_block(trans, extent_root, path->nodes[0]);
 136         btrfs_release_path(extent_root, path);
 137         finish_current_insert(trans, root);
 138         pending_ret = run_pending(trans, root);
 139         if (ret)
 140                 return ret;
 141         if (pending_ret)
 142                 return pending_ret;
 143         return 0;
 144
 145 }
 146
 147 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 148                                     struct btrfs_root *root)
 149 {
 150         struct btrfs_block_group_cache *bg;
 151         struct cache_extent *cache;
 152         int err = 0;
 153         int werr = 0;
 154         struct cache_tree *bg_cache = &root->fs_info->block_group_cache;
 155         struct btrfs_path path;
 156         btrfs_init_path(&path);
 157         u64 start = 0;
 158
 159         while(1) {
 160                 cache = find_first_cache_extent(bg_cache, start);
 161                 if (!cache)
 162                         break;
 163                 bg = container_of(cache, struct btrfs_block_group_cache,
 164                                         cache);
 165                 start = cache->start + cache->size;
 166                 if (bg->dirty) {
 167                         err = write_one_cache_group(trans, root,
 168                                                     &path, bg);
 169                         if (err)
 170                                 werr = err;
 171                 }
 172                 bg->dirty = 0;
 173         }
 174         return werr;
 175 }
 176
 177 static int update_block_group(struct btrfs_trans_handle *trans,
 178                               struct btrfs_root *root,
 179                               u64 bytenr, u64 num, int alloc)
 180 {
 181         struct btrfs_block_group_cache *bg;
 182         struct cache_extent *cache;
 183         struct btrfs_fs_info *info = root->fs_info;
 184         u64 total = num;
 185         u64 old_val;
 186         u64 byte_in_group;
 187
 188         while(total) {
 189                 cache = find_first_cache_extent(&info->block_group_cache,
 190                                                 bytenr);
 191                 if (!cache)
 192                         return -1;
 193                 bg = container_of(cache, struct btrfs_block_group_cache,
 194                                         cache);
 195                 bg->dirty = 1;
 196                 byte_in_group = bytenr - bg->key.objectid;
 197                 old_val = btrfs_block_group_used(&bg->item);
 198                 if (total > bg->key.offset - byte_in_group)
 199                         num = bg->key.offset - byte_in_group;
 200                 else
 201                         num = total;
 202                 total -= num;
 203                 bytenr += num;
 204                 if (alloc)
 205                         old_val += num;
 206                 else
 207                         old_val -= num;
 208                 btrfs_set_block_group_used(&bg->item, old_val);
 209         }
 210         return 0;
 211 }
 212
 213 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, struct
 214                                btrfs_root *root)
 215 {
 216         u64 first = 0;
 217         struct cache_extent *pe;
 218         struct cache_extent *next;
 219
 220         pe = find_first_cache_extent(&root->fs_info->pinned_tree, 0);
 221         if (pe)
 222                 first = pe->start;
 223         while(pe) {
 224                 next = next_cache_extent(pe);
 225                 remove_cache_extent(&root->fs_info->pinned_tree, pe);
 226                 free_cache_extent(pe);
 227                 pe = next;
 228         }
 229         root->fs_info->last_insert.objectid = first;
 230         root->fs_info->last_insert.offset = 0;
 231         return 0;
 232 }
 233
 234 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 235                                  btrfs_root *extent_root)
 236 {
 237         struct btrfs_key ins;
 238         struct btrfs_extent_item extent_item;
 239         int ret;
 240         struct btrfs_fs_info *info = extent_root->fs_info;
 241         struct cache_extent *pe;
 242         struct cache_extent *next;
 243         struct cache_tree *pending_tree = &info->pending_tree;
 244
 245         btrfs_set_extent_refs(&extent_item, 1);
 246         btrfs_set_extent_owner(&extent_item, extent_root->root_key.objectid);
 247         ins.offset = 1;
 248         btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
 249         pe = find_first_cache_extent(pending_tree, 0);
 250         while(pe) {
 251                 ins.offset = pe->size;
 252                 ins.objectid = pe->start;
 253
 254                 remove_cache_extent(pending_tree, pe);
 255                 next = next_cache_extent(pe);
 256                 if (!next)
 257                         next = find_first_cache_extent(pending_tree, 0);
 258
 259                 free_cache_extent(pe);
 260                 pe = next;
 261
 262                 ret = btrfs_insert_item(trans, extent_root, &ins, &extent_item,
 263                                         sizeof(extent_item));
 264                 if (ret) {
 265                         btrfs_print_tree(extent_root, extent_root->node);
 266                 }
 267                 BUG_ON(ret);
 268         }
 269         return 0;
 270 }
 271
 272 /*
 273  * remove an extent from the root, returns 0 on success
 274  */
 275 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 276                          *root, u64 bytenr, u64 num_bytes, int pin)
 277 {
 278         struct btrfs_path path;
 279         struct btrfs_key key;
 280         struct btrfs_fs_info *info = root->fs_info;
 281         struct btrfs_root *extent_root = info->extent_root;
 282         int ret;
 283         struct btrfs_extent_item *ei;
 284         u32 refs;
 285
 286         key.objectid = bytenr;
 287         btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
 288         key.offset = num_bytes;
 289
 290         btrfs_init_path(&path);
 291         ret = btrfs_search_slot(trans, extent_root, &key, &path, -1, 1);
 292         if (ret) {
 293                 btrfs_print_tree(extent_root, extent_root->node);
 294                 printf("failed to find %llu\n",
 295                        (unsigned long long)key.objectid);
 296                 BUG();
 297         }
 298         ei = btrfs_item_ptr(&path.nodes[0]->leaf, path.slots[0],
 299                             struct btrfs_extent_item);
 300         BUG_ON(ei->refs == 0);
 301         refs = btrfs_extent_refs(ei) - 1;
 302         btrfs_set_extent_refs(ei, refs);
 303         if (refs == 0) {
 304                 u64 super_bytes_used, root_bytes_used;
 305                 if (pin) {
 306                         int err;
 307                         err = insert_cache_extent(&info->pinned_tree,
 308                                                     bytenr, num_bytes);
 309                         BUG_ON(err);
 310                 }
 311                 super_bytes_used = btrfs_super_bytes_used(info->disk_super);
 312                 btrfs_set_super_bytes_used(info->disk_super,
 313                                             super_bytes_used - num_bytes);
 314                 root_bytes_used = btrfs_root_bytes_used(&root->root_item);
 315                 btrfs_set_root_bytes_used(&root->root_item,
 316                                           root_bytes_used - num_bytes);
 317
 318                 ret = btrfs_del_item(trans, extent_root, &path);
 319                 if (!pin && extent_root->fs_info->last_insert.objectid >
 320                     bytenr)
 321                         extent_root->fs_info->last_insert.objectid = bytenr;
 322                 if (ret)
 323                         BUG();
 324                 ret = update_block_group(trans, root, bytenr, num_bytes, 0);
 325                 BUG_ON(ret);
 326         }
 327         btrfs_release_path(extent_root, &path);
 328         finish_current_insert(trans, extent_root);
 329         return ret;
 330 }
 331
 332 /*
 333  * find all the blocks marked as pending in the radix tree and remove
 334  * them from the extent map
 335  */
 336 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 337                                btrfs_root *extent_root)
 338 {
 339         int ret;
 340         struct cache_extent *pe;
 341         struct cache_extent *next;
 342         struct cache_tree *del_pending = &extent_root->fs_info->del_pending;
 343
 344         pe = find_first_cache_extent(del_pending, 0);
 345         while(pe) {
 346                 remove_cache_extent(del_pending, pe);
 347                 ret = __free_extent(trans, extent_root,
 348                                     pe->start, pe->size, 1);
 349                 BUG_ON(ret);
 350                 next = next_cache_extent(pe);
 351                 if (!next)
 352                         next = find_first_cache_extent(del_pending, 0);
 353                 free_cache_extent(pe);
 354                 pe = next;
 355         }
 356         return 0;
 357 }
 358
 359 static int run_pending(struct btrfs_trans_handle *trans, struct btrfs_root
 360                        *extent_root)
 361 {
 362         del_pending_extents(trans, extent_root);
 363         return 0;
 364 }
 365
 366
 367 /*
 368  * remove an extent from the root, returns 0 on success
 369  */
 370 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 371                       *root, u64 bytenr, u64 num_bytes, int pin)
 372 {
 373         struct btrfs_root *extent_root = root->fs_info->extent_root;
 374         int pending_ret;
 375         int ret;
 376
 377         if (root == extent_root) {
 378                 ret = insert_cache_extent(&root->fs_info->del_pending,
 379                                             bytenr, num_bytes);
 380                 BUG_ON(ret);
 381                 return 0;
 382         }
 383         ret = __free_extent(trans, root, bytenr, num_bytes, pin);
 384         pending_ret = run_pending(trans, root->fs_info->extent_root);
 385         return ret ? ret : pending_ret;
 386 }
 387
 388 static u64 stripe_align(struct btrfs_root *root, u64 val)
 389 {
 390         u64 mask = ((u64)root->stripesize - 1);
 391         u64 ret = (val + mask) & ~mask;
 392         return ret;
 393 }
 394
 395 /*
 396  * walks the btree of allocated extents and find a hole of a given size.
 397  * The key ins is changed to record the hole:
 398  * ins->objectid == block start
 399  * ins->flags = BTRFS_EXTENT_ITEM_KEY
 400  * ins->offset == number of blocks
 401  * Any available blocks before search_start are skipped.
 402  */
 403 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
 404                             *orig_root, u64 total_needed, u64 search_start,
 405                             u64 search_end, struct btrfs_key *ins)
 406 {
 407         struct btrfs_path path;
 408         struct btrfs_key key;
 409         int ret;
 410         u64 hole_size = 0;
 411         int slot = 0;
 412         u64 last_byte = 0;
 413         u64 aligned;
 414         int start_found;
 415         struct btrfs_leaf *l;
 416         struct btrfs_root * root = orig_root->fs_info->extent_root;
 417
 418         if (root->fs_info->last_insert.objectid > search_start)
 419                 search_start = root->fs_info->last_insert.objectid;
 420
 421         search_start = stripe_align(root, search_start);
 422         btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 423
 424 check_failed:
 425         btrfs_init_path(&path);
 426         ins->objectid = search_start;
 427         ins->offset = 0;
 428         start_found = 0;
 429         ret = btrfs_search_slot(trans, root, ins, &path, 0, 0);
 430         if (ret < 0)
 431                 goto error;
 432
 433         if (path.slots[0] > 0)
 434                 path.slots[0]--;
 435
 436         while (1) {
 437                 l = &path.nodes[0]->leaf;
 438                 slot = path.slots[0];
 439                 if (slot >= btrfs_header_nritems(&l->header)) {
 440                         ret = btrfs_next_leaf(root, &path);
 441                         if (ret == 0)
 442                                 continue;
 443                         if (ret < 0)
 444                                 goto error;
 445                         if (!start_found) {
 446                                 aligned = stripe_align(root, search_start);
 447                                 ins->objectid = aligned;
 448                                 ins->offset = (u64)-1 - aligned;
 449                                 start_found = 1;
 450                                 goto check_pending;
 451                         }
 452                         ins->objectid = stripe_align(root,
 453                                                      last_byte > search_start ?
 454                                                      last_byte : search_start);
 455                         ins->offset = (u64)-1 - ins->objectid;
 456                         goto check_pending;
 457                 }
 458                 btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
 459                 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
 460                         goto next;
 461                 if (key.objectid >= search_start) {
 462                         if (start_found) {
 463                                 if (last_byte < search_start)
 464                                         last_byte = search_start;
 465                                 aligned = stripe_align(root, last_byte);
 466                                 hole_size = key.objectid - aligned;
 467                                 if (key.objectid > aligned &&
 468                                     hole_size > total_needed) {
 469                                         ins->objectid = aligned;
 470                                         ins->offset = hole_size;
 471                                         goto check_pending;
 472                                 }
 473                         }
 474                 }
 475                 start_found = 1;
 476                 last_byte = key.objectid + key.offset;
 477 next:
 478                 path.slots[0]++;
 479         }
 480         // FIXME -ENOSPC
 481 check_pending:
 482         /* we have to make sure we didn't find an extent that has already
 483          * been allocated by the map tree or the original allocation
 484          */
 485         btrfs_release_path(root, &path);
 486         BUG_ON(ins->objectid < search_start);
 487         if (find_cache_extent(&root->fs_info->pinned_tree,
 488                                 ins->objectid, total_needed)) {
 489                 search_start = ins->objectid + total_needed;
 490                 goto check_failed;
 491         }
 492         if (find_cache_extent(&root->fs_info->pending_tree,
 493                                 ins->objectid, total_needed)) {
 494                 search_start = ins->objectid + total_needed;
 495                 goto check_failed;
 496         }
 497         root->fs_info->last_insert.objectid = ins->objectid;
 498         ins->offset = total_needed;
 499         return 0;
 500 error:
 501         btrfs_release_path(root, &path);
 502         return ret;
 503 }
 504 /*
 505  * finds a free extent and does all the dirty work required for allocation
 506  * returns the key for the extent through ins, and a tree buffer for
 507  * the first block of the extent through buf.
 508  *
 509  * returns 0 if everything worked, non-zero otherwise.
 510  */
 511 static int alloc_extent(struct btrfs_trans_handle *trans,
 512                         struct btrfs_root *root, u64 owner,
 513                         u64 num_bytes, u64 search_start,
 514                         u64 search_end, struct btrfs_key *ins)
 515 {
 516         int ret;
 517         int pending_ret;
 518         u64 super_bytes_used, root_bytes_used;
 519         struct btrfs_fs_info *info = root->fs_info;
 520         struct btrfs_root *extent_root = info->extent_root;
 521         struct btrfs_extent_item extent_item;
 522
 523         btrfs_set_extent_refs(&extent_item, 1);
 524         btrfs_set_extent_owner(&extent_item, owner);
 525
 526         ret = find_free_extent(trans, root, num_bytes, search_start,
 527                                search_end, ins);
 528         if (ret)
 529                 return ret;
 530
 531         super_bytes_used = btrfs_super_bytes_used(info->disk_super);
 532         btrfs_set_super_bytes_used(info->disk_super, super_bytes_used +
 533                                     num_bytes);
 534         root_bytes_used = btrfs_root_bytes_used(&root->root_item);
 535         btrfs_set_root_bytes_used(&root->root_item, root_bytes_used +
 536                                    num_bytes);
 537         if (root == extent_root) {
 538                 ret = insert_cache_extent(&root->fs_info->pending_tree,
 539                                             ins->objectid, ins->offset);
 540                 BUG_ON(ret);
 541                 goto update_block;
 542         }
 543         ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
 544                                 sizeof(extent_item));
 545
 546         finish_current_insert(trans, extent_root);
 547         pending_ret = run_pending(trans, extent_root);
 548         if (ret)
 549                 return ret;
 550         if (pending_ret)
 551                 return pending_ret;
 552 update_block:
 553         ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
 554         BUG_ON(ret);
 555         return 0;
 556 }
 557 /*
 558  * helper function to allocate a block for a given tree
 559  * returns the tree buffer or NULL.
 560  */
 561 struct btrfs_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 562                                             struct btrfs_root *root,
 563                                             u32 blocksize)
 564 {
 565         struct btrfs_key ins;
 566         int ret;
 567         struct btrfs_buffer *buf;
 568         ret = alloc_extent(trans, root, root->root_key.objectid,
 569                            blocksize, 0, (u64)-1, &ins);
 570         if (ret) {
 571                 BUG();
 572                 return NULL;
 573         }
 574         buf = find_tree_block(root, ins.objectid, blocksize);
 575         btrfs_set_header_generation(&buf->node.header, trans->transid);
 576         btrfs_set_header_bytenr(&buf->node.header, buf->bytenr);
 577         memcpy(buf->node.header.fsid, root->fs_info->disk_super->fsid,
 578                sizeof(buf->node.header.fsid));
 579         dirty_tree_block(trans, root, buf);
 580         return buf;
 581
 582 }
 583
 584 /*
 585  * helper function for drop_snapshot, this walks down the tree dropping ref
 586  * counts as it goes.
 587  */
 588 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 589                           *root, struct btrfs_path *path, int *level)
 590 {
 591         struct btrfs_buffer *next;
 592         struct btrfs_buffer *cur;
 593         u64 bytenr;
 594         int ret;
 595         u32 refs;
 596
 597         ret = lookup_block_ref(trans, root, path->nodes[*level]->bytenr,
 598                                btrfs_level_size(root, *level), &refs);
 599         BUG_ON(ret);
 600         if (refs > 1)
 601                 goto out;
 602         /*
 603          * walk down to the last node level and free all the leaves
 604          */
 605         while(*level > 0) {
 606                 u32 size = btrfs_level_size(root, *level - 1);
 607
 608                 cur = path->nodes[*level];
 609                 if (path->slots[*level] >=
 610                     btrfs_header_nritems(&cur->node.header))
 611                         break;
 612                 bytenr = btrfs_node_blockptr(&cur->node, path->slots[*level]);
 613                 ret = lookup_block_ref(trans, root, bytenr, size, &refs);
 614                 if (refs != 1 || *level == 1) {
 615                         path->slots[*level]++;
 616                         ret = btrfs_free_extent(trans, root, bytenr, size, 1);
 617                         BUG_ON(ret);
 618                         continue;
 619                 }
 620                 BUG_ON(ret);
 621                 next = read_tree_block(root, bytenr, size);
 622                 if (path->nodes[*level-1])
 623                         btrfs_block_release(root, path->nodes[*level-1]);
 624                 path->nodes[*level-1] = next;
 625                 *level = btrfs_header_level(&next->node.header);
 626                 path->slots[*level] = 0;
 627         }
 628 out:
 629         ret = btrfs_free_extent(trans, root, path->nodes[*level]->bytenr,
 630                                 btrfs_level_size(root, *level), 1);
 631         btrfs_block_release(root, path->nodes[*level]);
 632         path->nodes[*level] = NULL;
 633         *level += 1;
 634         BUG_ON(ret);
 635         return 0;
 636 }
 637
 638 /*
 639  * helper for dropping snapshots.  This walks back up the tree in the path
 640  * to find the first node higher up where we haven't yet gone through
 641  * all the slots
 642  */
 643 static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
 644                         *root, struct btrfs_path *path, int *level)
 645 {
 646         int i;
 647         int slot;
 648         int ret;
 649         for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 650                 slot = path->slots[i];
 651                 if (slot <
 652                     btrfs_header_nritems(&path->nodes[i]->node.header)- 1) {
 653                         path->slots[i]++;
 654                         *level = i;
 655                         return 0;
 656                 } else {
 657                         ret = btrfs_free_extent(trans, root,
 658                                         path->nodes[*level]->bytenr,
 659                                         btrfs_level_size(root, *level), 1);
 660                         btrfs_block_release(root, path->nodes[*level]);
 661                         path->nodes[*level] = NULL;
 662                         *level = i + 1;
 663                         BUG_ON(ret);
 664                 }
 665         }
 666         return 1;
 667 }
 668
 669 /*
 670  * drop the reference count on the tree rooted at 'snap'.  This traverses
 671  * the tree freeing any blocks that have a ref count of zero after being
 672  * decremented.
 673  */
 674 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
 675                         *root, struct btrfs_buffer *snap)
 676 {
 677         int ret = 0;
 678         int wret;
 679         int level;
 680         struct btrfs_path path;
 681         int i;
 682         int orig_level;
 683
 684         btrfs_init_path(&path);
 685
 686         level = btrfs_header_level(&snap->node.header);
 687         orig_level = level;
 688         path.nodes[level] = snap;
 689         path.slots[level] = 0;
 690         while(1) {
 691                 wret = walk_down_tree(trans, root, &path, &level);
 692                 if (wret > 0)
 693                         break;
 694                 if (wret < 0)
 695                         ret = wret;
 696
 697                 wret = walk_up_tree(trans, root, &path, &level);
 698                 if (wret > 0)
 699                         break;
 700                 if (wret < 0)
 701                         ret = wret;
 702         }
 703         for (i = 0; i <= orig_level; i++) {
 704                 if (path.nodes[i]) {
 705                         btrfs_block_release(root, path.nodes[i]);
 706                 }
 707         }
 708         return ret;
 709 }
 710
 711 int btrfs_free_block_groups(struct btrfs_fs_info *info)
 712 {
 713         struct btrfs_block_group_cache *bg;
 714         struct cache_extent *cache;
 715
 716         while(1) {
 717                 cache = find_first_cache_extent(&info->block_group_cache, 0);
 718                 if (!cache)
 719                         break;
 720                 bg = container_of(cache, struct btrfs_block_group_cache,
 721                                         cache);
 722                 remove_cache_extent(&info->block_group_cache, cache);
 723                 free(bg);
 724         }
 725         return 0;
 726 }
 727
 728 int btrfs_read_block_groups(struct btrfs_root *root)
 729 {
 730         struct btrfs_path path;
 731         int ret;
 732         int err = 0;
 733         struct btrfs_block_group_item *bi;
 734         struct btrfs_block_group_cache *bg;
 735         struct cache_tree *bg_cache;
 736         struct btrfs_key key;
 737         struct btrfs_key found_key;
 738         struct btrfs_leaf *leaf;
 739         u64 group_size = BTRFS_BLOCK_GROUP_SIZE;
 740
 741         root = root->fs_info->extent_root;
 742         bg_cache = &root->fs_info->block_group_cache;
 743         key.objectid = 0;
 744         key.offset = group_size;
 745         btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
 746         btrfs_init_path(&path);
 747
 748         while(1) {
 749                 ret = btrfs_search_slot(NULL, root->fs_info->extent_root,
 750                                         &key, &path, 0, 0);
 751                 if (ret != 0) {
 752                         err = ret;
 753                         break;
 754                 }
 755                 leaf = &path.nodes[0]->leaf;
 756                 btrfs_disk_key_to_cpu(&found_key,
 757                                       &leaf->items[path.slots[0]].key);
 758                 bg = malloc(sizeof(*bg));
 759                 if (!bg) {
 760                         err = -1;
 761                         break;
 762                 }
 763                 bi = btrfs_item_ptr(leaf, path.slots[0],
 764                                     struct btrfs_block_group_item);
 765                 memcpy(&bg->item, bi, sizeof(*bi));
 766                 memcpy(&bg->key, &found_key, sizeof(found_key));
 767                 key.objectid = found_key.objectid + found_key.offset;
 768                 btrfs_release_path(root, &path);
 769                 bg->cache.start = found_key.objectid;
 770                 bg->cache.size = found_key.offset;
 771                 bg->dirty = 0;
 772                 ret = insert_existing_cache_extent(bg_cache, &bg->cache);
 773                 BUG_ON(ret);
 774                 if (key.objectid >=
 775                     btrfs_super_total_bytes(root->fs_info->disk_super))
 776                         break;
 777         }
 778         btrfs_release_path(root, &path);
 779         return 0;
 780 }
 781
 782 int btrfs_insert_block_group(struct btrfs_trans_handle *trans,
 783                              struct btrfs_root *root,
 784                              struct btrfs_key *key,
 785                              struct btrfs_block_group_item *bi)
 786 {
 787         int ret;
 788         int pending_ret;
 789
 790         root = root->fs_info->extent_root;
 791         ret = btrfs_insert_item(trans, root, key, bi, sizeof(*bi));
 792         finish_current_insert(trans, root);
 793         pending_ret = run_pending(trans, root);
 794         if (ret)
 795                 return ret;
 796         if (pending_ret)
 797                 return pending_ret;
 798         return ret;
 799 }